Split (1)

train · ~5.33M rows (showing the first 168k)

id string	text string	source string	added string	created string	metadata dict
PMC1079947	Background ========== Chronic obstructive pulmonary disease (COPD), consisting of variable degrees of pulmonary emphysema and chronic obstructive bronchitis, has a male predominance. However, the prevalence of COPD is steadily increasing among women \[[@B1]\] as a consequence of increased rates of cigarette smoking. Women may be more sensitive to the deleterious effects of smoking than men\[[@B2]\]. Mortality due to COPD is increasing among women,\[[@B3]\] which may not merely reflect the increasing overall prevalence, but also the increase of the most severe forms of COPD. Therefore, novel therapeutic options are needed to cope with COPD. Decreased production of progestins after menopause may result in lack of natural respiratory stimulation and predispose women to respiratory insufficiency. Therefore, progestin replacement therapy could effectively improve ventilation and oxygenation during sleep. Hypoxemia and hypercapnia during sleep herald respiratory insufficiency. Many studies have investigated effects of progestins on breathing in awake subjects but few studies have evaluated the effect of progestin treatment on nocturnal breathing in patients with respiratory insufficiency, mostly due to COPD \[[@B4]-[@B9]\]. In light of the previous observations, the effect of progestin therapy on breathing may be weaker during sleep\[[@B5]\]. A key issue is that the aforementioned studies have recruited almost entirely men. However, effects of progestin therapy may be gender-dependent. We have previously reported of sustained progestin-induced improvement in awake blood gases of postmenopausal women with respiratory insufficiency\[[@B10]\]. We hypothesised that the beneficial effects of medroxyprogesterone acetate (MPA) would also be demonstrated as improved SaO~2~and tcCO~2~during sleep. We therefore wanted to evaluate the degree and duration of MPA effect on sleep and nocturnal breathing in postmenopausal women with chronic respiratory disease and show that MPA effectively improves SaO~2~and tcCO~2~during sleep in postmenopausal women with moderate to severe COPD. Methods ======= Patient selection ----------------- 15 consecutive postmenopausal women with COPD fulfilling the inclusion criteria and willing to volunteer were recruited for the trial by using our hospital data base. The inclusion criteria were forced expiratory volume in one second (FEV1) less than 65 % of predicted value\[[@B11]\]. The criteria for postmenopausal status were age over 50 yr, time since last menstruation at least 12 months, and serum concentrations of follicle-stimulating hormone (FSH) \> 30 IU/l. All patients were receiving optimal therapy and had been in a stable clinical condition for at least a month prior the study. The exclusion criteria included malignancies, heart diseases (except cor pulmonale), insulin dependent diabetes mellitus, previous thromboembolic events, obstructive sleep apnea syndrome, untreated hypothyroidism, any other major diseases, current smoking, long-term oxygen therapy and use of systemic postmenopausal hormone therapy. Patients were on their regular medication that was found optimal by their pulmonologist. Medication varied from patient to patient and included both short- and long-acting inhaled beta-agonists, anticholinergics and inhaled corticosteroids. No changes in medication were allowed throughout the study period. Medication likely to influence sleep or vigilance such as benzodiazepines, psychostimulants, melatonin or antidepressants were not allowed. Vaginal estrogen therapy was allowed since vaginal application does not affect serum estradiol levels. All patients gave their written informed consents. The study protocol was approved by the Joint Commission on Ethics of Turku University and Turku University Central Hospital. Study design ------------ The 12-week study followed a placebo-controlled single-blind design (Fig. [1](#F1){ref-type="fig"}) and included five nights at 3-week intervals in sleep laboratory. Seven days after the baseline measurements, all patients started with placebo treatment for 14 days. After a 7-day interval, MPA treatment for 14 days was started. Patients were then followed-up for six weeks after cessation of MPA. We found it important that the placebo night data would be obtained before the patients had received MPA. Based on our earlier observations,\[[@B10],[@B12],[@B13]\] MPA-induced effects are measurable for weeks after cessation of progestin therapy. To exclude the carry-over effect, more than a 6-week washout period would be needed between the treatments. Knowing that acute exacerbations are frequent in patients with moderate to severe COPD, the longer duration of the study would increase the drop-out rate markedly. Therefore, the placebo and MPA studies were not done in a random sequence. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Study design.Both placebo (PL) and medroxyprogesterone acetate (MPA) periods lasted for 14 days. Arrows indicate the timing of sleep studies, dashed line the timing and duration of the placebo period and solid line thetiming and duration of the MPA period. ::: ![](1465-9921-6-28-1) ::: 30 mg of oral MPA (Lutopolar^®^, Orion Pharma, Espoo, Finland) or placebo tablets provided by the same pharmaceutical company were administered twice in the evening, at 9 PM and 11 PM. Peak serum concentrations are reached within one to three hours followed by a sharp decline thereafter\[[@B14]\]. MPA was divided in two doses given two hours apart to minimize the variation in MPA levels during night. Compliance was assessed by tablet counts, patient interviews and measurements of serum MPA concentrations. The overnight polysomnography was registered five times at 3-week intervals: at baseline, after 14 days on placebo, after 14 days on MPA, and after 3- and 6-week washout periods. Sleep studies were performed at a university sleep laboratory, where the technician was present during the recording. In the evening, patients were weighed, and at each visit blood pressure was measured by the same researcher, using the auscultatory method in seated subjects. Every morning after sleep study patients completed a structured questionnaire with 14 separate items concerning their subjective sleep quality during the past night\[[@B15]\]. Questions were related to initiation and maintenance of sleep (for example, perceptions of sleep latency, frequency of awakenings), events during sleep (for example, restlessness), sleep-related daytime symptoms (for example, naps during the preceding day), and issues possibly disturbing sleep (for example, dyspnoea, sweating). Flow-volume spirometry (Vitalograph Compact II, Vitalograph Ltd, Buckingham, England) was assessed in the morning after the sleep study. An overnight polysomnography included electroencephalogram, electro-oculogram, and chin electromyogram. Arterial oxyhemoglobin saturation (SaO~2~) was measured with pulse oximeter using a finger probe (Ohmeda Biox 3700 Pulse Oximeter, BOC Health Care, U.S.A.). The response time setting for the pulse oximeter was 6 seconds. Partial pressure of carbon dioxide (tcCO~2~) was measured transcutaneously (TINA^®^Transcutaneous pO~2~/pCO~2~Monitoring System, Radiometer, Copenhagen, Denmark). After cleansing the skin with alcohol the skin sensor was placed on the upper part of the chest parasternally and heated up to 43.5°C, at which temperature the sensor could remain attached continuously for 8 hours. Body movements were monitored with a static-charge-sensitive bed\[[@B16],[@B17]\]. Quantitative movement analysis and SaO~2~analyses were performed with custom made analysis modules (UniPlot^®^, Unesta, Turku, Finland). Nasal pressure was recorded with nasal prongs connected to a presser transducer. The nasal pressure signal was analysed with Sullivan Autoset device (Sullivan Autoset^®^, ResCare Limited, Sydney, Australia) used in a diagnostic mode and with a separate inspiratory flow shape analysis software described earlier\[[@B18]\]. Biosignals were recorded and stored with sampling frequency of 250 Hz per channel and with y-resolution of 12 bytes (UniPlot^®^, Unesta, Turku, Finland). The episodes of arterial oxyhemoglobin desaturation of 4 %-units or more per hour (ODI~4~) were calculated with Uniplot^®^software (UniPlot^®^, Unesta, Turku, Finland). EtCO~2~and tcCO~2~signals were recorded with a sampling frequency of 100 Hz throughout the night by Embla^®^system (Embla^®^, Flaga, Reykjavik, Iceland). The SaO~2~and tcCO~2~data were analyzed on a breath-by-breath basis. Each breath was assigned with the sleep stage during which it occurred. This was achieved using an MS-Excel macro, which combined the information from the two files according to time tags. The maximum inspiratory slope during the initial part of inspiration was determined on a breath-by-breath basis from the nasal pressure signal and used as an index of respiratory drive as previously described\[[@B19],[@B20]\]. Sleep stages and arousals were scored according to standard criteria\[[@B21],[@B22]\] and expressed as percentage of total sleep time. Sleep efficiency was defined as percentage sleep during the sleep period time. Statistical analyses -------------------- The overall comparisons between repeated measurements were performed with either nonparametric Friedman\'s test or parametric analysis of variance (ANOVA) for repeated measurements. In nonparametric case Wilcoxon signed ranks test and in parametric case F-test were used. In repeated measurements, Bonferroni correction was used for p-values. All comparisons were made to baseline measurements. Comparisons between the first and the second sessions tested the placebo effect, between the first and the third sessions the immediate effect of MPA, between the first and the fourth and the first and the fifth the sustained effect of MPA. Correlation analyses were performed with Pearson correlation method and p-values were corrected with Bonferroni method. In all tests, p \< 0.05 was considered significant. Statistical computing was performed with SAS statistical package (version 8.01, SAS Institute, Cary, NC). Results ======= 13 out of 15 patients completed the trial. Two subjects discontinued due to acute exacerbation of their COPD after the second visit. None of the subjects had clinical symptoms of OSAS. Mean (± SD) ODI~4~was 1.4 ± 1.9/h at baseline and did not change during the trial. Laboratory assessments were within inclusion criteria in all but one patient. She had low serum FSH (S-FSH 25 IU/L). However, she was considered postmenopausal based on her age (71 years) and her serum concentration of estradiol (36 pmol/L, postmenopausal reference range \< 140 pmol/L), and included in the trial. According to tablet count, all patients were compliant and after the two-week treatment MPA was detectable in all patients confirming their compliance. After a three-week washout MPA was under the detection limit in three patients and near the detection limit in the rest, and after a six-week washout undetectable in all patients. No correlations were found between serum MPA levels and SaO~2~or tcCO~2~. We were not able to analyse the EtCO~2~data due to insufficient plateaus in EtCO~2~signals. Baseline characteristics of patients are presented in Tables [1](#T1){ref-type="table"} and [2](#T2){ref-type="table"}. Weight, dyspnea or other symptom scores, blood pressure, or spirometric values did not differ during the trial. None of the patients had a history of snoring, witnessed apnea or excessive sleepiness. Patients reported that they slept as usual and woke up several times during night. They had no specific complaints disturbing their sleep. Subjective sleep quality and total time in bed did not differ throughout the trial. MPA did not induce any alterations in objectively measured polygraphic sleep parameters. Three women had withdrawal bleeding after cessation of MPA therapy, one of them was on vaginal estrogen therapy. The other two women had benign endometrial findings underlying their withdrawal bleeding. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Characteristics of patients at baseline. ::: N = 15 Mean SD Range ------------------------- ---------- -------- ----------------- Age (yr) 67.5 6.0 56 -- 76 BMI (kg/m^2^) 26.9 4.9 15.4 -- 35.2 Smoking (pack-year) 8.8 13.9 0 -- 50 FEV~1~(L) 0.76 0.3 0.44 -- 1.80 FEV~1~(%) 34 12.4 15 -- 63 FVC (L) 1.25 0.5 0.77 -- 1.69 FVC (%) 44 16.2 26 -- 72 FEV% 61 13.7 48 -- 77 Arterial pH 7.38 0.05 7.25 -- 7.45 PaCO~2~(kPa) 6.0 1.1 5.5 -- 9.9 74.2 PaO~2~(kPa) 9.0 1.2 5.6 -- 12.6 Systolic BP (mmHg) 154 21.8 120 -- 190 Diastolic BP (mmHg) 87 14.1 60 -- 118 BMI = body mass index, FEV~1~= forced expiratory volume in one second, FVC = forced vital capacity, FEV% = 100 × (FEV~1~/FVC), PaCO~2~= partial tension of arterial carbon dioxide, PaO~2~= partial tension of arterial oxygen. ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Objective sleep quality at baseline, N = 15. ::: Sleep Recording Data at Baseline -------------------------------------- -------------------------------- TST 4 h 34 min ± 1 h 13 min Sleep efficiency 66 ± 18 % of sleep period time EEG arousals 2.6 ± 3.3 / h of sleep Movement arousals 9.7 ± 7.1 / h of sleep Latency to S2 sleep 41 ± 41 min Stage S1 sleep 87 ± 60 min (34 ± 24 %) Stage S2 sleep 81 ± 48 min (29 ± 15 %) SWS 73 ± 39 min (26 ± 10 %) REM sleep 34 ± 31 min (11 ± 10 %) Percentages indicate proportion of the total sleep time (TST) unless otherwise indicated. Data presented as mean ± SD. REM = rapid eye movements, SWS = slow wave sleep (sleep stages 3 and 4), TST = total sleep time. ::: No changes in respiratory or sleep parameters were found during placebo period compared to baseline. We analyzed the effects of MPA during sleep and separately according to sleep stages. Both SaO~2~mean and SaO~2~nadir improved in 11 out of 13 (85 %) patients during sleep. tcCO~2~median improved with MPA in 12 out of the 13 women (92 %). At baseline, the SaO~2~mean averaged 90.6 ± 3.2 % (95% Bonferroni corrected confidence interval (CI) 88.3, 92.9; Fig. [2](#F2){ref-type="fig"}). The average increase with MPA was 1.7 ± 1.6 %-units (95% CI 0.56, 2.84; p = 0.044). The median of SaO~2~nadir was 84.8 % (interquartile range, IQR 6.1; 95% CI 66.5, 87.3) at baseline and improved by 3.9 %-units (IQR 4.8; 95% CI 0.24, 10.2; p = 0.024; Fig. [2](#F2){ref-type="fig"}.) with MPA. At baseline, the mean of tcCO~2~median was 6.0 ± 0.9 kPa (CI 95% 5.4, 6.6) and decreased by 0.9 ± 0.5 kPa with MPA (CI 95 % -1.3, -0.5; p = 0.004; Fig. [2](#F2){ref-type="fig"}.) There was a borderline increase in the median of maximum inspiratory slope from 10.4 (IQR 4.7; 95% CI 7.0, 23.7) U/50-1 at baseline to 18.9 (IQR 22.4; CI 95% 11.0, 62.9) U/50-1 with MPA (p = 0.068). ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Changes of SaO~2~nadir, SaO~2~mean, and tcCO~2~median during sleep with MPA and after a 3- and a 6-week follow-up.Changes are absolute percentage values. MPA = medroxyprogesterone acetate, SaO~2~= arterial oxygen saturation, tcCO~2~= transcutaneous partial carbon dioxide tension. The error bars of SaO~2~mean and tcCO~2~median represent standard deviation and those of SaO2 nadir represent interquartile ranges. P-values are corrected with Bonferroni method. \\ = p \< 0.01, \* = p \< 0.05. ::: ![](1465-9921-6-28-2) ::: Three weeks after cessation of MPA, SaO~2~mean and SaO~2~nadir sustained to be higher in 11 out of 13 patients during sleep, when compared to baseline. The average of SaO~2~mean was 1.4 ± 1.8 (95% corrected CI -0.03, 2.8) %-units higher, the difference being statistically not significant after Bonferroni corrections (p = 0.028 prior to the correction; p = 0.112 after Bonferroni correction, n = 11, Fig. [2](#F2){ref-type="fig"}). The median SaO~2~nadir was 2.7 %-units (IQR 4.9; 95% CI 0.06, 18.7) higher than at baseline (p = 0.032). Six weeks after MPA administration SaO~2~nadir or SaO~2~mean did not differ from baseline. Within three weeks after cessation of MPA, tcCO~2~median was lower compared to baseline in 6 out of 13 patients but the mean of tcCO~2~median had returned to baseline (Fig. [2](#F2){ref-type="fig"}). The pattern of MPA effect on SaO~2~nadir (Fig. [3](#F3){ref-type="fig"}), SaO~2~mean (Fig. [4](#F4){ref-type="fig"}), and tcCO~2~median (Fig. [5](#F5){ref-type="fig"}) was quite similar during stages S1, S2, slow wave sleep (SWS; stages S3 and S4) and rapid eye movement (REM) sleep. Importantly, MPA had a marked effect also during REM sleep (Fig. [3](#F3){ref-type="fig"}, [4](#F4){ref-type="fig"}, [5](#F5){ref-type="fig"}). The median of maximum inspiratory slope increased from 10.7 (IQR 4.6; 95% CI 6.5, 30.0) U/50-1 at baseline to 16.8 (IQR 30.7) U/50-1 with MPA (95% CI 11.8, 59.0; p = 0.042) during stage 4 sleep. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Changes of SaO~2~nadir during different sleep stages with MPA and after a 3- and a 6-week follow-up.Changes are absolute percentage values. MPA = medroxyprogesterone acetate, SaO~2~= arterial oxygen saturation. The error bars of SaO~2~nadir represent interquartile ranges. P-values are corrected with Bonferroni method. \\ = p \< 0.01. ::: ![](1465-9921-6-28-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Changes of SaO~2~mean during different sleep stages with MPA and after a 3- and a 6-week follow-up.Changes are absolute percentage values. MPA = medroxyprogesterone acetate, SaO~2~= arterial oxygen saturation. The error bars of SaO~2~mean represent standard deviation. P-values are corrected with Bonferroni method. ::: ![](1465-9921-6-28-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Changes of the mean of tcCO~2~median during different sleep stages with MPA and after a 3- and a 6-week follow-up.Changes are absolute percentage values. MPA = medroxyprogesterone acetate, tcCO~2~= transcutaneous partial carbon dioxide tension. The error bars of tcCO~2~represent standard deviation. P-values are corrected with Bonferroni method. \\ = p \< 0.01, \* = p \< 0.05. ::: ![](1465-9921-6-28-5) ::: Increases in SaO~2~mean and SaO~2~nadir during sleep with MPA were inversely associated with baseline SaO~2~mean (r = -0.70, p = 0.032) and baseline SaO~2~nadir (r = -0.77, p = 0.008), respectively. Decrease in tcCO2 with MPA did not correlate neither with baseline SaO~2~nor tcCO~2~levels. Treatment response in SaO~2~mean, SaO~2~nadir and tcCO~2~levels did not associate with pack-years smoked, age, BMI, baseline spirometric results or sleep variables. Discussion ========== The present data show that MPA effectively improves SaO~2~and tcCO~2~during sleep in postmenopausal women with moderate to severe COPD. The improvement seems to be greater in patients with lower baseline SaO~2~. None of the previous studies has systemically used MPA to stimulate nocturnal breathing in female patients with COPD. The effect seems to be more consistent in female than in male patients with COPD\[[@B5]\]. The pattern of MPA effect was quite similar over the array of sleep stages and MPA had beneficial effects also during REM sleep. Even when administered at higher doses than in ordinary gynecologic indications, MPA was also well tolerated. We used the maximum inspiratory slope as an index of respiratory drive. Although we have used the inspiratory slope index previously in patients with partial upper airway obstruction during sleep\[[@B19],[@B20]\], it can also be used in populations without partial upper airway obstruction such as in patients with COPD. It is a non-invasive index affected also by upper airway dynamics. The different contributing factors are not calculated separately but the index reflects a \"total output\" during the initial part of inspiration. Since this index is insufficiently validated, the results should be considered suggestive and interpreted with caution. At baseline, the maximum slope in this group of women with moderate to severe COPD was compatible with that of healthy postmenopausal women\[[@B20]\]. It was also compatible with the maximum inspiratory slope in postmenopausal women with partial upper airway obstruction during sleep using MPA\[[@B20]\]. Although the maximum inspiratory slope was already at baseline within normal postmenopausal range, MPA almost doubled the slope. This finding is in line with the thinking that the baseline respiratory drive in women with COPD is not decreased compared to healthy postmenopausal women and that exogenous respiratory stimulation increases the respiratory drive above the physiological postmenopausal level in these patients. Our results differ to some extent from the previous studies by other groups. First, in our female patients, the respiratory improvement seems to be more consistent than in the previous studies recruiting almost entirely male patients. SaO~2~mean and SaO~2~nadir increased in 11 out of 13 (85 %) and tcCO~2~median decreased with MPA in 12 out of the 13 women (92 %) during sleep. Previously, in 17 men and two women with COPD, MPA (60 mg/d/1 month) improved SaO~2~during sleep only in five out of 19 patients\[[@B5]\]. The authors did not report, whether both of the studied women were among the responders. In six non-obese men with COPD, MPA (100 mg/d/15 days) decreased the number of hypoxemic episodes during sleep\[[@B6]\]. Second, the CO~2~levels were not monitored during sleep in most previous studies. In the present study, lowering of tcCO~2~was short-lived paralleling to the pattern observed in base excess previously in awake patients\[[@B10]\]. Our findings are consistent with those observed in 17 hypercapnic stable patients with COPD, where nocturnal end-tidal CO~2~decreased by 0.9 kPa with MPA (60 mg/d/2 weeks)\[[@B9]\]. Our results are also in line with observations in awake healthy males, where MPA increased alveolar ventilation and slopes of hypercapnic and hypoxic ventilatory responses\[[@B23]\]. tcCO~2~reflects the metabolic tissue performance\[[@B24]\] and the slow changes of the absolute partial arterial carbon dioxide tension (PaCO~2~)\[[@B25]\]. According to Sanders and co-workers,\[[@B26]\] tcCO~2~does not consistently and accurately reflect PaCO~2~and may even overestimate it\[[@B27]\]. However, tcCO~2~measurements show the direction and rough magnitude of the changes in CO~2~levels. There is no indication that PaCO~2~would be superior to tcCO~2~to monitor the metabolic condition at the tissue level. We used tcCO~2~as a surrogate for PaCO~2~in order to avoid invasive measurements. Nothing suggests so far that transcutaneous CO~2~measurements would affect sleep quality. However, we think that neither end tidal CO~2~nor tcCO~2~represents directly PaCO~2~but that all the three measurements reflect different phenomena. Frequent nocturnal sampling of arterial blood gases would increase our understanding about the blood-tissue interactions of acid-base balance and needs to be considered in future studies. Third, MPA effectively improved oxygenation and tcCO~2~also during REM sleep. In a previous study in patients with COPD, MPA (60 mg/d/4 weeks) decreased PaCO~2~and increased PaO~2~only during non-REM sleep and had a similar trend during REM sleep\[[@B28]\]. Fourth, the possible after-effects of MPA on breathing were usually not evaluated in previous studies. The respiratory effects of a two-week treatment with MPA of 60 mg daily subsided within 14 days in healthy male subjects\[[@B29]\]. In patients with COPD, Wagenaar and coworkers did not observe any after-effects on respiration one month after cessation of combined therapy with MPA and acetazolamide (gender distribution in study population not reported)\[[@B9]\]. Prolonged improvement of daytime PaCO~2~in women with chronic respiratory insufficiency\[[@B10]\] and of nocturnal end tidal CO~2~in postmenopausal women with partial upper airway obstruction during sleep\[[@B13]\] has been a consistent finding in our previous studies. The present study showed a sustained improvement in SaO~2~nadir. The obviously more consistent and prolonged effect of MPA on respiration in our female patients compared to that in male patients in previous studies may be explained by differences in study populations or MPA dosage regimen. We recruited only women. Although both estrogen and progesterone concentrations in postmenopausal women are at the same level than in men, the effects of progestin therapy may remain gender-specific. Sex steroids influence neuromodulatory serotonergic neurons,\[[@B30]\] which are critically involved in the neural control of breathing\[[@B31],[@B32]\]. Serotonin has an excitatory effect on upper airway and phrenic motoneurons \[[@B33]-[@B37]\]. Animal studies demonstrate a greater serotonin activity in female than in male brain\[[@B30],[@B38]\]. The number and distribution of progesterone receptors may differ between genders because androgens down-regulate progesterone and estrogen receptors without affecting progesterone or estradiol concentrations\[[@B39]\]. Estradiol increases the number of progesterone receptors\[[@B40]\]. In ovariectomized rats, MPA did not stimulate breathing until progesterone receptors were upregulated with estrogen\[[@B41]\]. MPA-derived metabolites with intrinsic estrogenic activity\[[@B42]\] may upregulate progesterone receptors more effectively in women than in men. The duration of respiratory effects of MPA may differ between healthy individuals and those with respiratory impairment like COPD or sleep-disordered breathing. The persistent effect of MPA may be due to modification of peripheral or central chemoreceptor action or central processing of the carotid body neural output\[[@B43]\]. MPA may reset the respiratory center for a new response level. The homeostatic regulatory mechanisms aim to maintain the normal function in the body. Therefore, the new response threshold of the respiratory center is maintained longer in subjects with respiratory impairment than in healthy subjects. The elimination of MPA might be slower in diseased patients than in healthy individuals. However, this is unlikely to explain the persistent respiratory effect in our patients, since MPA concentrations were below or near the detection limit within three weeks after cessation of MPA. Our relatively high MPA dose was chosen to ensure effective respiratory stimulation during night. Therefore, we also administered the whole dose of MPA in the evening. MPA reaches its peak serum concentration within one to three hours and the concentrations decline quite rapidly thereafter\[[@B14]\]. 60 mg per day is the most commonly used dose in studies where respiration is stimulated by MPA. However, it is usually divided in three doses administered at 8-hour intervals. MPA-induced persistent respiratory effects may be attributed to the alteration in endocrinological environment, or pharmacodynamics of MPA. MPA alters hormone levels, and their recuperation may take for weeks at least in women\[[@B12]\]. In an animal model, the clearance of MPA-related substances is slow in the lung, skeletal muscle and brain tissues\[[@B44]\]. Although MPA has a stronger progestational activity than progesterone, it does not have greater effects on breathing than progesterone\[[@B29]\]. This observation may indicate that the metabolites of MPA cause the respiratory effects rather than MPA itself. MPA did not have any effect on subjective or objective sleep quality in our patients. In postmenopausal women, MPA (5 mg/d) combined with estrogen had no effect on objective sleep parameters but improved subjective sleep quality\[[@B45]\]. According to previous studies, MPA has induced no consistent effects on sleep\[[@B5],[@B6],[@B13],[@B28],[@B46]-[@B50]\]. The arousal indexes were low. Part of this could be due to true suppression of arousals by hypoxemia and sleep deprivation. On the other hand, many of the arousals initiated a whole epoch of wakefulness, resulting in technically low arousal index, compared to healthy elderly people or patients with COPD. Because our patients spent most of the time in light sleep, arousals easily resulted in awakenings and thereby in lower arousal index than previously reported in healthy elderly people\[[@B51]\] or in patients with COPD\[[@B52],[@B53]\]. Our patients were also hypoxemic and sleep deprived with an average total sleep time less than 5 hours. Both hypoxemia\[[@B54],[@B55]\] and sleep deprivation \[[@B56],[@B57]\] increase arousal threshold resulting in lower arousal index. The single blind study design is a limitation of our study. We refrained from using a cross-over design knowing that acute exacerbations frequently occur in patients with moderate to severe COPD, and therefore the longer duration of the study would increase the drop-out rate markedly. Although a cross-over setting was not feasible, a parallel group design would have strengthened the data and needs to be considered for future trials. Another limitation is the small study population. Although the number of patients recruited was based on our previous trials\[[@B10],[@B13]\], the results of the current study indicated that a greater number of subjects would have strengthened especially the tcCO~2~results. However, we are confident that correction of these shortcomings would only have been strengthened, not undone our findings. Conclusion ========== Taken together, MPA 60 mg daily for 14 days improved breathing during sleep in postmenopausal women with COPD. Some of the effects persisted for weeks after cessation of MPA. Our patients were not homogenous in terms of baseline arterial blood gases and nocturnal SaO~2~. Some of them had values likely to threaten their life expectancy, others had minor impairments. Patients with more severe hypoxemia seemed to improve more than those with milder respiratory impairment. Our results warrant further studies into the long-term efficacy and feasibility of MPA administered either on a cyclical or on a continuous basis to support breathing during sleep in postmenopausal women with clinically significant hypoxia or hypercapnia. List of abbreviations ===================== ANOVA = analysis of variance, CI = confidence interval, COPD = chronic obstructive pulmonary disease, IQR = interquartile range, MPA = medroxyprogesterone acetate, FEV~1~= forced expiratory volume in one second, FSH = follicle-stimulating hormone, SaO~2~= arterial oxyhemoglobin saturation, REM sleep = rapid eye movement sleep, SWS = slow wave sleep, tcCO~2~= transcutaneous carbon dioxide tension Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= TS participated in the design of the study, coordinated and carried out the studies, analyzed the sleep recordings, participated in the analyses and interpretation of data, and drafted the manuscript. TA constructed the analysis model for tcCO~2~signal, participated in the analyses and interpretation of tcCO~2~data and in the statistical analyses, and drafted the manuscript. KU participated in the analyses and interpretation of tcCO~2~data, and drafted the manuscript. OP conceived of the study, and participated in its design and in the interpretation of data, and drafted the manuscript. All authors read and approved the final manuscript. Acknowledgements ================ This work was supported by grants from The Finnish Anti-Tuberculosis Association Foundation, Finnish Sleep Research Society, The Research Foundation for Pulmonary Diseases, The Väinö and Laina Kivi Foundation, Paulo Foundation and The Turku University Foundation. Medication was supplied by the drug company Orion Pharma, Espoo, Finland. Esa Wallius, MSc, is acknowledged for statistical assistance.	PubMed Central	2024-06-05T03:55:55.599951	2005-4-4	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1079947/", "journal": "Respir Res. 2005 Apr 4; 6(1):28", "authors": [ { "first": "Tarja", "last": "Saaresranta" }, { "first": "Tero", "last": "Aittokallio" }, { "first": "Karri", "last": "Utriainen" }, { "first": "Olli", "last": "Polo" } ] }
PMC1079948	Background ========== We and others have shown that one of the mechanisms of growth regulation of small cell lung cancer (SCLC) cell lines and cultured pulmonary neuroendocrine cells (PNEC) is by the binding of agonists to a cell surface receptor of the neuronal nicotinic acetylcholine receptor family comprised of homomeric α7 subunits, which functions as an ion channel with high permeability for Ca^2+^\[[@B1]-[@B8]\]. Binding of agonists to this receptor activates the release of the autocrine growth factor serotonin \[[@B2]-[@B6]\]. In addition, we have shown that the nicotine-derived carcinogenic nitrosamine, 4(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK), is a high affinity agonist for the α7 nicotinic acetylcholine receptor (α7 nAChR) \[[@B5]\]. Binding of NNK to this receptor caused an influx of Ca^2+^from the extracellular environment \[[@B7]\], resulting in the activation of a protein kinase C-dependent Raf-1/MAP kinase-mediated mitogenic pathway \[[@B5],[@B9],[@B10]\]. These findings suggest that the chronic stimulation of this pathway may contribute to the selective development of this histologic cancer type in smokers. Accordingly, components of this signal transduction pathway may be promising targets for cancer intervention studies with selectivity for SCLC. Our goal in the present studies was to determine the extent of α7 mRNA expression in the human lung to determine if previous research with SCLC could be extrapolated to other lung malignancies. Previous research in our laboratory indicated expression of mRNA for the α7 receptor in normal fetal hamster PNEC cells \[[@B6]\]. PNEC cells are one of the possible cells of origin for SCLC \[[@B11],[@B12]\]. We screened multiple small cell and non-small cell lung cancer cell lines, normal cells, and fresh surgical tissue samples from cancer patients for mRNA expression of the α7 receptor. With the exception of NCI-H441 cell line, all cell lines and patient samples tested expressed mRNA for the α7 receptor, and the α7 nicotinic receptor protein is expressed in fewer cell lines. Methods ======= Cell culture ------------ The human SCLC cell lines NCI-H69, NCI-H82, NCI-H146, NCI-H187, NCI-H209, and NCI-H526, the human adenocarcinoma cell lines NCI-H322 and NCI-H441, and A549, the carcinoid cell line NCI-H727, and the squamous cell lines NCI-H226, NCI-H2170, and NCI-H520 were purchased from the American Type Culture Collection (Manassas, VA). The human SCLC cell line WBA \[[@B13]\] was a gift of Dr. G. Krystal, Medical College of Virginia. All cancer cell lines except A549 were maintained in RPMI medium supplemented with fetal bovine serum (10% v/v), L-glutamine (2 mM), penicillin (100 U/ml) and streptomycin (100 μg/ml) at 37°C in an atmosphere of 5% CO~2~. A549 cells were grown in Hams F12 media with supplements as above. Human small airway epithelia cells (SAEC) were purchased from Clonetics/BioWhittaker (Walkersville, MD). These cells were maintained in SAEC basal medium with supplements (Clonetics) at 37°C in an atmosphere of 5% CO~2~. Fresh surgical tissue samples were collected from patients at the University of Tennessee Graduate School of Medicine\'s Cancer Center and processed for reverse transcription polymerase chain reaction (RT-PCR). The collection of tissue was approved by the University of Tennessee Institutional Review Board, and the authors have been certified by the NIH Office of Human Subjects Research. RT-PCR ------ RT-PCR assays were conducted with all cells and tissues. Expression of the α7 receptor by RT-PCR in fetal hamster PNECs has been previously published \[[@B6]\]. RT-PCR was done as described before \[[@B6]\] except new human α7 primers (forward 5\'-gccaatgactcgcaaccactc-3\' and reverse 5\'-ccagcgtacatcgatgtagca-3\' bases 236--571, Genbank accession number X70297) were used. These amplified a 335 bp fragment. Oligonucleotide primers were acquired from Life Technologies (Grand Island, NY). These primer pairs are in areas of the sequence that are homologous between humans, rats, and chick brain \[[@B14]\]. Reactions were run on a MJ Research PTC-200 thermal cycler (Watertown, MA) with the following conditions: 1 cycle of 2 min. at 94°C, 40 cycles of 94°C, 30 sec; 55°C, 30 sec; 72°C, 45 sec, with a final extension for 5 min. at 72°C. Sequencing ---------- Several PCR products were sequenced to verify the integrity of the PCR process. The NCI-H69, NCI-H322, and SAE cells were sequenced using the forward PCR primer for human α7. Sequencing was done with the ABI Terminator Cycle Sequencing reaction kit on an ABI 373 DNA sequencer (Perkin-Elmer, Foster City, CA). Nuclease-protection assay ------------------------- The nuclease protection assay was used to determine differences in expression levels in a representative SCLC cell line (NCI-H82), and in SAE cells. The Lig\'nScribe kit (Ambion, Austin, TX) was used to add a T7 RNA polymerase promoter to the PCR fragment amplified by the gene specific α7 primer, and then this T7-α7 fragment was PCR amplified, and the resulting fragment was used directly in a transcription reaction using the MAXIscript in vitro transcription kit (Ambion). This transcription reaction consisted of the PCR fragment, 10X transcription buffer, 10 mM ATP, UTP and GTP, T7 polymerase, \[α-^32^P\] CTP (800 Ci/mmol, Dupont-NEN, Boston, MA), and nuclease free water to a final volume of 20 μl. A probe for 28S ribosomal RNA was also transcribed and used as an internal control. These reactions were incubated for 1 hour at 37°C. After this incubation, 1 μl of DNase I was added, and the mixture was further incubated for 15 min. at 37°C. The probes were gel purified. The RPA III kit (Ambion) was used. A molar excess of labeled probes (28S and gene specific) were added to 20 μg total RNA or Yeast RNA (Ambion), and the RNA samples and probes were co-precipitated and resuspended in hybridization buffer. After incubation at 95°C for 4 min., and incubation overnight at 42°C, RNase digestion buffer containing RNase A/RNase T1 was added. After digestion for 30 min. at 37°C, RNase inactivation/precipitation solution was added. After precipitation the pellets were resuspended in gel loading buffer, heated at 95°C for 4 min. and loaded onto a 5% polyacrylamide, 8 M urea gel (Bio-Rad, Hercules, CA). In addition RNA Century Markers Plus templates (Ambion) were transcribed and used as markers. After electrophoresis, the gel was transferred to blotting paper, dried, and exposed to Kodak XAR film. Western blots ------------- Cell pellets were collected and membrane protein was isolated with the ReadyPrep protein extraction kit (signal) (Biorad). Protein levels were determined using the RCDC kit (Biorad). Aliquots of 20--30 μg protein were boiled in 3x loading buffer (New England Biolabs, Beverly, MA) for 2 minutes, then loaded onto 12% Tris-glycine-polyacrylamide gels (Cambrex, Rockland, ME), and transferred electrophoretically to nictrocellulose membranes. Membranes were incubated with the primary antibody (alpha 7 nicotinic acetylcholine receptor; Abcam, Cambridge, MA). In all western blots, membranes were additionally probed with an antibody for actin (Sigma) to ensure equal loading of protein between samples. The membranes were then incubated with appropriate secondary antibodies (Rockland, Gilbertsville, PA or Molecular Probes, Eugene OR). The antibody-protein complexes were detected by the LiCor Odyssey infrared imaging system (Lincoln, NE). Cells for some blots were incubated with 100 pM NNK (Midwest Research Institute, Kansas City, MO) for various times. Results ======= The RT-PCR assay demonstrated expression of the α7 nAChR in all seven cultured human SCLC cell lines (Figure [1](#F1){ref-type="fig"}) and in SAE cells (Figure [2](#F2){ref-type="fig"}). Among the two adenocarcinoma cell lines, NCI-H322 demonstrated expression of the α7 nAChR, whereas NCI-H441, yielded negative results (Figure [3](#F3){ref-type="fig"}). Both adenocarcinoma cell lines demonstrated expression of the cyclophilin control indicating that the α7 nicotinic acetylcholine receptor was either not expressed in NCI-H441 cells, or expression levels were below the limit of detection of RT-PCR samples run on an agarose gel. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Agarose gel showing α7 nicotinic acetylcholine receptor expression in SCLC cell lines.cDNA was amplified by PCR using the human α7 primers. SCLC cell lines: 1) WBA; 2) NCI-H69; 3) NCI-H82; 4) NCI-H146; 5) NCI-H187; 6) NCI-H209; 7) NCI-H526. For all gene expression experiments, negative control reactions were performed and found to be negative. The bands were consistent with the expected size, 335 bp. M-100 bp DNA ladder. ::: ![](1465-9921-6-29-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Agarose gel showing α7 nicotinic acetylcholine receptor expression in normal small airway epithelial cells.cDNA was amplified by PCR using the human α7 and cyclophilin primers. 1) α7 primers; 2) cyclophilin primers. For all gene expression experiments, negative control reactions were performed and found to be negative. The bands were consistent with the expected sizes, 335 bp for the α7 primers and 216 bp for the cyclophilin primers. M-100 bp DNA ladder. ::: ![](1465-9921-6-29-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Agarose gel showing α7 nicotinic acetylcholine receptor expression in NCI-H322 but not NCI-H441 adenocarcinoma cell lines.cDNA was amplified by PCR using the human α7 and cyclophilin primers. 1)NCI-H322, α7 primers; 2)NCI-H441, α7 primers; 3)NCI-H322, cyclophilin primers; 4)NCI-H441, cyclophilin primers. For all gene expression experiments, negative control reactions were performed and found to be negative. The bands were consistent with the expected sizes, 335 bp for the α7 primers and 216 bp for the cyclophilin primers. M-100 bp DNA ladder. ::: ![](1465-9921-6-29-3) ::: To verify the RT-PCR results, RT-PCR products from one SCLC cell line (NCI-H69) and the adenocarcinoma cell line NCI-H322 were sequenced using the forward primer used for RT-PCR amplification of the samples (data not shown). The sequences from the PCR products of both NCI-H69 and NCI-H322 were compared to the sequence of the α7 nicotinic acetylcholine receptor (bases 257--571, Genbank accession number X70297) and were found to be 100% homologous. Although we have shown mRNA expression for the α7 nicotinic acetylcholine receptor, it was necessary to show if expression of the α7 nicotinic acetylcholine receptor protein in seen in these cell lines. Using a specific antibody for the α7 nicotinic acetylcholine receptor (Abcam), membrane protein from the cell lines were assessed by western blotting. Expression of α7 protein was seen in 6 of the 7 SCLC cell lines (Figure [4](#F4){ref-type="fig"}). Expression of α7 protein was not seen in the NCI-H187 cell line (Figure [4](#F4){ref-type="fig"}). Actin levels were unchanged in all 7 SCLC cell lines (Figure [4](#F4){ref-type="fig"}). ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Expression of α7 nicotinic acetylcholine receptor protein in SCLC cell lines as assessed by western blot analysis.Protein was isolated with the ReadyPrep protein extraction kit (signal) (Biorad). Nitrocellulose membranes were incubated with the rabbit polyclonal antibody to the alpha 7 nicotinic acetylcholine receptor (Abcam). 1) WBA; 2) NCI-H69; 3) NCI-H82; 4) NCI-H146; 5) NCI-H187; 6) NCI-H209; 7) NCI-H526. The arrow indicates the 56 kDa band (expected size). All SCLC cell lines except NCI-H187 express protein for alpha 7 nicotinic acetylcholine receptor. ::: ![](1465-9921-6-29-4) ::: Additional non-small cell lung cancer cell lines were screened for the presence of α7 nAChR mRNA expression. Expression of α7 nAChR was also seen in A549 adenocarcinoma and NCI-H727 carcinoid cell lines (Figure [5](#F5){ref-type="fig"}). The mRNA for α7 nAChR was also expressed in three squamous cell lines, NCI-H2170, NCI-H226, and NCI-H520 (Figure [5](#F5){ref-type="fig"}). Expression of the α7 nAChR was also seen in nine fresh tissue samples from lung cancer patients, all of which were smokers (Figure [6](#F6){ref-type="fig"}). ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Agarose gel showing α7 nicotinic acetylcholine receptor expression in A-549 adenocarcinoma, NCI-H727 carcinoid, and squamous cell lines NCI-H226, NCI-H520, and NCI-H2170.cDNA was amplified by PCR using the human α7 primers. 1)A549; 2)NCI-727; 3)NCI-H226; 4)NCI-H520; 5)NCI-H2170. For all gene expression experiments, negative control reactions were performed and found to be negative. The bands were consistent with the expected size, 335 bp. M-100 bp DNA ladder. ::: ![](1465-9921-6-29-5) ::: ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Agarose gel showing α7 nicotinic acetylcholine receptor expression in tissue samples from lung cancer patients.cDNA was amplified by PCR using the human α7 primers. 1)large cell carcinoma; 2)carcinoid; 3--5)squamous cell carcinomas; 6--8)adenocarcinomas; 9)adenocarcinoma/carcinoid. For all gene expression experiments, negative control reactions were performed and found to be negative. The bands were consistent with the expected size, 335 bp. M-100 bp DNA ladder. ::: ![](1465-9921-6-29-6) ::: The non-SCLC cell lines and normal cells were also screened for expression of α7 nicotinic acetylcholine receptor protein. SAEC, the adenocarcinoma cell line A549 and the squamous cell line NCI-H226 express protein for α7 (Figure [7](#F7){ref-type="fig"}), whereas the adenocarcinoma cell line NCI-H322, the carcinoid cell line NCI-H727, the squamous cell lines NCI-H520 and NCI-H2170 did not express protein for the α7 nicotinic acetylcholine receptor (Figure [7](#F7){ref-type="fig"}). Actin levels were unchanged in all protein samples (Figure [7](#F7){ref-type="fig"}). ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Expression of α7 nicotinic acetylcholine receptor protein in non-SCLC cell lines and normal SAE cells as assessed by western blot analysis.Protein was isolated with the ReadyPrep protein extraction kit (signal) (Biorad). Nitrocellulose membranes were incubated with the rabbit polyclonal antibody to the alpha 7 nicotinic acetylcholine receptor (Abcam). 1) SAEC; 2) A549; 3) NCI-727; 4) NCI-H226; 5) NCI-H520; 6) NCI-H2170; 7) NCI-H322. The arrow indicates the 56 kDa band (expected size). Only SAEC, A549 and NCI-H226 express protein for alpha 7 nicotinic acetylcholine receptor. ::: ![](1465-9921-6-29-7) ::: To allow for a direct comparison of expression levels of α7 mRNA between a representative SCLC cell line (NCI-H82), and human SAE cells from a non-smoker (information provided by Clonetics), a riboprobe was transcribed from the PCR fragment amplified by the gene specific α7 primer used for RT-PCR and was used in a nuclease protection assay. Expression of mRNA for the α7 nicotinic acetylcholine receptor was demonstrated in the SCLC cell line NCI-H82, however no protected band was detected in the normal small airway epithelial cells (Figure [8](#F8){ref-type="fig"}). Expression of the 28S ribosomal RNA used as an internal control was seen in both samples (Figure [8](#F8){ref-type="fig"}). No protected bands were seen with either probe when annealed to yeast RNA (Figure [8](#F8){ref-type="fig"}). The increased levels of α7 nAChR mRNA as compared with the normal SAE cells suggest that some SCLC cells may have higher levels of α7 nAChR than normal lung epithelial cells. ::: {#F8 .fig} Figure 8 ::: {.caption} ###### Comparison of expression levels of α7 nicotinic acetylcholine receptor between a representative SCLC cell line (NCI-H82), and SAE cells by a nuclease protection assay.A riboprobe transcribed from the human α7 PCR primer was used to compare expression levels in the two cell systems. 1)NCI-H82; 2)SAEC; 3)Yeast RNA. The protected fragments were consistent with the expected sizes, 335 bp for the α7 primers and 115 bp for the 28S ribosomal RNA control primers. The molecular weight markers are indicated by the numbers on the left side of the figure. ::: ![](1465-9921-6-29-8) ::: To determine if the α7 nicotinic acetylcholine receptor protein is a functional protein, we stimulated a representative SCLC cell line, NCI-H69 with NNK. This cell line had been used previously in our laboratories to show α7 nicotinic acetylcholine receptor specific binding \[[@B5]\]. The tobacco carcinogen NNK (100 pM) increased expression of α7 protein levels after 60 and 150 minutes of treatment (Figure [9](#F9){ref-type="fig"}). Actin levels were unchanged between the times protein was collected (Figure [9](#F9){ref-type="fig"}). ::: {#F9 .fig} Figure 9 ::: {.caption} ###### Increases in expression of α7 nicotinic acetylcholine receptor protein in NCI-H69 after treatment with 100 pM of the tobacco carcinogen NNK as assessed by western blot analysis.Protein was isolated with the ReadyPrep protein extraction kit (signal) (Biorad). Nitrocellulose membranes were incubated with the rabbit polyclonal antibody to the alpha 7 nicotinic acetylcholine receptor (Abcam). 1) control; 2) 5 min; 3) 30 min; 4) 60 min; 5) 150 min; 6) 24 hour. The arrow indicates the 56 kDa band (expected size). Increases were seen in expression of protein for alpha 7 nicotinic acetylcholine receptor after 60 min and 150 min of NNK treatment. ::: ![](1465-9921-6-29-9) ::: Discussion ========== Our data supports the ubiquitous expression of the α7 nAChR mRNA in both normal and cancerous lung cells. With the exception of the NCI-H441 adenocarcinoma cell line, the α7 nAChR mRNA was expressed in all normal and cancer cells tested. Previous research in our laboratory indicated expression of mRNA for the α7 receptor in normal fetal hamster PNEC cells \[[@B6]\]. PNEC cells are one of the possible cells of origin for SCLC \[[@B11],[@B12]\]. The expression of the α7 nAChR is not an artifact of cell culture since the expression of the α7 nAChR was seen in nine tumor samples from different patients with lung cancer. We also found that the α7 nAChR is expressed in five distinct types of cancer: squamous, carcinoid, adenocarcinoma, large cell carcinoma, and small cell lung cancer. This is the first report of the expression of the α7 nAChR receptor mRNA in pulmonary squamous, carcinoids or large cell carcinomas. A recent report indicated α7 nAChR receptor mRNA in both human bronchial epithelial cells and airway fibroblasts \[[@B15]\], supporting the hypothesis of ubiquitous expression of the α7 nAChR receptor mRNA in human lung cells. In addition, we have demonstrated expression of the α7 nAChR protein of the correct molecular weight in lung cancer cells for the first time. Of the seven SCLC cell lines tested for the α7 nAChR mRNA expression, only the NCI-H187 cell line did not express a band of the correct molecular weight for the α7 nAChR. In the non-SCLC cells and normal cells that express the α7 nAChR mRNA, α7 nAChR protein expression was more limited than in SCLC. SAEC (normal cells), the adenocarcinoma cell line A549 and the squamous cell line NCI-H226 express protein for α7, whereas the adenocarcinoma cell line NCI-H322, the carcinoid cell line NCI-H727, the squamous cell lines NCI-H520 and NCI-H2170 did not express protein for α7. However, our data are in contrast to another recent study. Carlisle et al. \[[@B15]\] found that α7 nAChR transcripts are frequently found in human bronchial epithelial cells although the protein of the correct size is not. They also postulated that muscle-type nicotinic acetylcholine receptors might be involved in responses to nicotine. Further research is needed to determine if both neuronal and muscle type acetylcholine receptors are involved in signaling events in lung cancer cells. Previous research from our laboratories has indicated that NNK binds with high affinity to alpha-bungarotoxin (agonist for α7 and α8 \[[@B16]\]) sensitive nAChRs in NCI-H69 and NCI-H82 cells, and the NNK affinity for these receptors was several times higher than for nicotine \[[@B5]\]. This binding was inhibited by hexamethonium but not decamethonium \[[@B5]\]. Hexamethonium is a selective agonist for neuronal nAChRs, whereas decamethonium is selective for muscle-type nAChRs \[[@B17]\]. In addition, we have found that NNK activates the Raf-1/MAP kinase pathway resulting in phosphorylation of c-myc \[[@B10]\]. This activation was inhibited by alpha-bungarotoxin. Lending support to this data, in the present study NNK stimulated α7 protein expression in NCI-H69 cells. Although muscle type acetylcholine receptors may be involved in signaling responses in SCLC, our data indicates the α7 neuronal nicotinic acetylcholine receptor is a functional receptor in lung cells and stimulates signaling events in these cells. Using a nuclease protection assay that allows for direct comparisons between samples, we found higher levels of α7 nAChR mRNA receptor in NCI-H82 than in SAEC. The RT-PCR assay showing expression of α7 nAChR receptor in SAE cells is a much more sensitive assay, but does not allow for the comparisons seen in the nuclease protection assay. The increase in α7 nAChR receptor in the SCLC cell line NCI-H82 compared to the normal SAE cells from a non-smoker is consistent with other data from our laboratory and others. In addition, it has been shown that the NCI-H82 cell line synthesized the highest levels of acetylcholine, and addition of nicotinic antagonists slowed growth of the NCI-H82 cells \[[@B18]\]. Acetylcholine is known to act as an autocrine growth factor for SCLC \[[@B19]\]. This data is also consistent with protein levels of the α7 nAChR. The NCI-H82 cell line had the highest level of α7 nAChR protein of the SCLC cell lines tested. Treatment of pregnant hamsters with the tobacco carcinogen, NNK, led to an increase in the α7 nAChR in PNEC isolated from fetal hamsters on the 15^th^day of gestation \[[@B20]\]. As indicated above, previous data from our laboratory has also shown that an associated mitogenic signal transduction pathway is upregulated in SCLC. Similar results were seen in the hamster PNEC model, as protein levels of Raf-1 in NNK treated hamster PNEC were greatly increased compared to control PNEC protein levels, and the NNK treated PNEC protein levels were at similar levels to untreated SCLC cell line NCI-H69 cells \[[@B9],[@B10]\]. In addition, ERK1/2 protein levels were also increased in NNK treated PNEC \[[@B9],[@B10]\]. These findings are in accord with publications which have demonstrated that chronic exposure of brain cells to nicotinic agonists results in a paradoxical upregulation of this receptor \[[@B17],[@B21]-[@B23]\]. Our findings are also supported by a study in monkeys, which has shown that chronic treatment of pregnant monkeys with nicotine caused a pronounced upregulation of the α7 nAChR in the lungs of the newborns \[[@B24]\]. Together these data indicate that the α7 nAChR may play an important role in the development of SCLC, and other lung cancers in which smoking is involved. Although there are many growth factors involved in multiple pathways leading to lung cancer, the effects of signaling through nicotinic receptors needs further investigation to determine its role in the pathogenesis of lung cancer. Further studies with lung cancer cell lines that express both the mRNA and protein for the α7 nAChR are needed. Conclusion ========== Expression of mRNA for the neuronal nicotinic acetylcholine receptor α7 seems to be ubiquitously expressed in all human lung cancer cell lines tested (except for NCI-H441) as well as normal lung cells. The α7 nicotinic receptor protein is expressed in fewer cell lines, and the tobacco carcinogen NNK increases α7 nicotinic receptor protein levels. The α7 nAChR may play an important role in the development of SCLC and other lung cancers in which smoking is involved. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= HP carried out all experiments with the exception of the western blots, participated in the design of the study, and helped draft the manuscript. MD performed the western blot experiments. HS conceived of the study and helped draft the manuscript. Acknowledgements ================ We wish to thank Neil Quigley at the University of Tennessee Molecular Biology Resource Facility for performing the sequencing, and thank Michelle Williams for technical assistance and Tommy Jordan for help with the final figures. The WBA cell line was a gift of Dr. G. Krystal, Medical College of Virginia, Richmond, VA. This research was supported by PHS grants R01-CA 51211, T32ESO7285, and the State of Tennessee Center of Excellence Fund.	PubMed Central	2024-06-05T03:55:55.602606	2005-4-4	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1079948/", "journal": "Respir Res. 2005 Apr 4; 6(1):29", "authors": [ { "first": "Howard K", "last": "Plummer" }, { "first": "Madhu", "last": "Dhar" }, { "first": "Hildegard M", "last": "Schuller" } ] }
PMC1079949	Background ========== There are currently two models for prostate cancer(PCa) which are diametrically opposed to each other. One model\[[@B1]\] proposes that high levels of androgens are responsible for PCa. This model is unable to explain the fact that androgen levels drop with age while the incidence of PCa increases. The other model\[[@B2]\] proposes that high levels of androgen should be effective in the prevention and treatment of PCa. While this model is consistent with the relationship of PCa to age, it does not explain experiments in which increased levels of androgen have been shown to increase the incidence of PCa, such as the correlation of increased incidence of PCa with higher levels of free testosterone(T)\[[@B3]\]. Model ===== Model description ----------------- The Estradiol-Dihydrotestosterone(E-D) model of PCa, proposed in this paper, hypothesizes that 17β-estradiol(E2) is essential for turning on telomerase in PCa, resulting in telomere formation. It is hypothesized that it does this by binding to a heterodimer composed of one estrogen receptor-α(ER-α) and one estrogen receptor-β (ER-β). This model proposes that telomere formation is an essential step for initial PCa growth. It also hypothesizes that 5α-dihydrotestosterone (DHT) is essential for initial PCa growth by binding to the intracellular androgen receptor(iAR), which results in the cell being protected from the apoptosis caused by androgens binding to the membrane xandrogen receptor(mAR). Finally, it proposes that T binds to mAR and upregulates proteins that promote apoptosis, such as U19, ALP1, and Fas. All of the above hormones also have other interactions which affect PCa. The E-D model proposes that when T binds to mAR, it upregulates bcl-2, a protein that protects against apoptosis, whereas when DHT binds to iAR, bcl-2 is downregulated. It is hypothesized that the expression of bcl-2 increases when E2 binds to ER-α and decreases when E2 binds to ER-β. It is known that T is converted to E2 by the enzyme aromatase(Aro) and is converted to DHT by the enzyme 5-α reductase type II(5AR2). Supporting evidence ------------------- It has been observed that mice never get PCa if they have a genetic mutation resulting in a lack of ER-α\[[@B4]\] or Aro\[[@B5]\]. These findings are consistent with the E-D model, which would also predict that mice with a genetic mutation resulting in a lack of ER-β would never get PCa. Aro activity was not observed\[[@B5]\] in normal epithelial prostate cells, but was observed in all of the PCa cell lines tested. E2 has been shown\[[@B6]\] to turn on telomerase activity, with both ER-α and ER-β being involved, in human epithelial cells as well as in tissue cultures of PCa. Some researchers believe that telomerase activity is an essential step for all types of cancer\[[@B7]\]. These findings are consistent with the prediction of the E-D model that E2 is essential for the initiation of PCa growth by bringing about telomere formation. Implantation of PCa xenografts into intact and ovariectomized female mice showed that E2 inhibited the growth of 5 out of the 6 xenografts tested\[[@B8]\]. The authors concluded that \"the observed inhibition of PCa growth may be attributable to direct effects of estrogens via ER-β\". Mice with a genetic mutation lacking ER-β\[[@B9]\] have an overexpression of bcl-2 in their ventral prostate. E2 has been shown\[[@B10]\] to increase the production of bcl-2 in MCF-7, an ER-α positive cell line of breast cancer. This increase is negated by the addition of 4-hydroxytamoxifen(OHT). It is known\[[@B11]\] that OHT acts as an antagonist to ER-α. All of this is consistent with the hypotheses of the E-D model that E2 increases the production of bcl-2 when binding to ER-α and decreases it when binding to ER-β. It has been shown\[[@B12]\] that T binds to mAR to produce apoptosis by upregulating the protein Fas. In the PCa cell line DU145 which lacks iAR, T or DHT alone was sufficient to induce apoptosis. However, in LNCaP, a hormone sensitive PCa cell line which has a functional iAR, T promotes cell growth unless it was given in the form of T-BSA, which does not enter the cell nor bind to the iAR, in which case apoptosis was also induced. This shows that the binding of androgens to iAR can counteract the apoptosis otherwise induced by the binding of androgens to mAR, as proposed in the E-D model. It has been shown\[[@B13]\] that T upregulates a protein, U19, which induces apoptosis in PCa. This apoptosis was inhibited by mibolerone, a synthetic androgen, binding to the iAR. Another protein, ALP1, that was upregulated by T, was also found\[[@B14]\] to induce apoptosis. These findings are all consistent with the E-D model. When finasteride(F), an inhibitor of 5AR2, is added to LNCaP, it inhibits growth in a dose-dependent manner\[[@B15]\]. Apoptosis resulted at the highest dose tested. This indicates that T binding to iAR is not able to completely prevent the apoptosis induced by T binding to mAR, whereas DHT is. This can be represented symbolically as T:mAR \>\>\> T:iAR in the presence of F and DHT:iAR \>\>\> T:mAR in the absence of F. This is consistent with the hypothesis of the E-D model that DHT is essential for initial PCa growth because it protects the PCa from mAR induced apoptosis. It is known\[[@B1]\] that men with a genetic mutation that produces non-functional 5AR2 do not get PCa. Since 5AR2 is found within the prostate cells and converts T to DHT, the result is very little DHT in the prostate. This finding is consistent with the hypothesis that in the absence of DHT, T binding to iAR in PCa is not able to prevent the apoptosis caused by T binding to mAR. It has been shown\[[@B16],[@B17]\] that in a small percentage of men with castrate metastatic PCa, enormous improvement in symptoms occurred following the administration of T. One possible explanation for this is that the PCa in those individuals might have lacked a functional iAR, but retained a functional mAR. The fact that T alone is sometimes capable of causing apoptosis in vivois consistent with the E-D model. When pyrrolidinedithiocarbamate(PDTC), a strong anti-oxidant, is added to LNCaP, it induces apoptosis\[[@B18]\]. It was shown that when 10^-12^M T was added to LNCaP, it increased the growth rate, but when it was added in addition to PDTC, the amount of apoptosis was greater than that induced by PDTC alone. This again shows that T is capable of inducing apoptosis under the proper circumstances. T has been shown\[[@B12]\] to increase bcl-2 production when it binds in the form of T-BSA to the mAR of LNCaP. T-BSA was shown not to bind to iAR. DHT has been shown\[[@B19]\] to decrease bcl-2 production when it binds to the iAR of LNCaP-FGC. This decrease was inhibited by the addition of bicalutamide, an antiandrogen which interferes with the binding of DHT to iAR. This is consistent with mAR being involved in the upregulation of bcl-2 and iAR being involved with the down-regulation of bcl-2, as proposed in the E-D model. Discussion ========== The E-D model, presented in this paper, does not explain how genetic mutations occur that cause PCa, but does explain which factors are essential for PCa to grow. The crucial factors for initial PCa growth are telomere formation and apoptosis avoidance. Since E2 produces telomere formation\[[@B6]\], in the absence of high levels of exogenous E2, Aro activity would initially be needed to supply PCa with large amounts of E2. The binding of normal amounts of DHT to iAR is sufficient for PCa to initially avoid apoptosis. As mutations develop that interfere with the process of apoptosis, DHT binding to iAR may no longer be necessary. The evidence is overwhelming that T is capable of inducing apoptosis in PCa. The in vitroand in vivostudies are unambiguous in this regard. The study with PDTC\[[@B18]\] raises interesting questions that can only be answered with further experimentation. There are three possible explanations as to why T increased the apoptosis caused by PDTC. Either PDTC interfered with the binding of androgens to iAR, or PDTC inhibited 5AR2, or PDTC may have enhanced the ability of T to cause apoptosis when binding to mAR. If the latter is true, it would have important implications in preventing and treating PCa, especially if other anti-oxidants should also be found to exhibit this same sort of enhancement. Assuming that the genetic studies with mice are applicable to humans, then it is clear that E2 is essential in order for PCa to occur. The evidence is only circumstantial that E2 is binding to a heterodimer of ER-α and ER-β. If a dimer were not involved, then one would expect that small amounts of E2 would be sufficient to induce telomerase activity. The heterodimer is consistent with the observation that both ER-α and ER-β are involved in telomere formation\[[@B6]\]. The hypothesis that telomere formation is essential for all cancers to occur\[[@B7]\] is not a proven fact, but it gives a plausible explanation as to why E2 is initially essential for PCa. The genetic study in mice\[[@B9]\] makes it clear that ER-β acts to inhibit bcl-2 production. The evidence that ER-α increases bcl-2 production is more circumstantial. It is known\[[@B20]\] that ER-α and ER-β tend to counteract each other. The increase in bcl-2 in the breast cancer line exposed to E2\[[@B10]\] is consistent with ER-α being responsible for increasing bcl-2. However, this assumes that the effect remains the same when the same hormone binds to the same hormone receptors in breast and prostate cancers. This is an intriguing concept, and would imply that the different overall effects that hormones have on these two cancers could be explained by the different amounts of each hormone receptor present in them. More research is needed to learn if breast cancer also possesses mAR and iAR acting in opposition to each other, as they do in PCa. High dosages of E2 have been used in the treatment of PCa. Transdermal patches of E2 have been shown to produce castrate levels of T within 3 weeks\[[@B21]\]. It is known that castrate levels of T result in apoptosis of most PCa cells, with calcium overload being one of the causes\[[@B22]\]. Calreticulin protects prostate cells from apoptosis due to calcium overload by enhancing the calcium buffering capacity, but castrate levels of T dramatically downregulate calreticulin. More research is needed to determine whether calreticulin is upregulated by mAR, iAR, or both. The fact that Aro activity was found in PCa cell lines, but not in normal prostate cell lines\[[@B5]\], does not mean that it is present in all PCa. Even if E2 is necessary for telomere formation, once telomeres of sufficient length are formed, Aro activity may no longer be necessary. It is interesting to note that the level of Aro activity observed in the PCa cell lines falls within the range observed in breast cancer cell lines\[[@B5]\]. The assumption of the E-D model that when DHT binds to iAR it counters the effect of apoptosis induced when T binds to mAR is supported by the study\[[@B15]\] that showed that increases in F decreased the growth of LNCaP in a dose-dependent manner, with the highest dose causing apoptosis. If the behaviour in vivois the same as this in vitroresult, then it explains why men with defective 5AR2 would not get PCa\[[@B1]\]. There are many questions still to be answered about PCa. Do the progesterone(P) receptor isomers, PRA and PRB, affect bcl-2 production? What is the dose effect of T, DHT, E2, and P on each of their corresponding receptors? For the hormone receptor pairs, to what extent does each receptor work against the other? Another important issue is how mAR is involved in the apoptosis of aged normal prostate cells. The fact\[[@B12]\] that T is capable of causing apoptosis in PCa which lacks iAR indicates that the increased amount of bcl-2 produced by mAR is not sufficient to counteract the apoptotic proteins produced by mAR. The makes it likely that the production of bcl-2 by mAR represents a damping factor designed to protect normal prostate cells from inappropriate apoptosis that might otherwise be caused by local fluctuations in serum T levels. If it turns out that calreticulin\[[@B22]\] is upregulated by mAR, then the absence of mAR should lead to apoptosis through calcium overload. As normal prostate cells age, if they start to lose functionality of mAR, iAR, or both, then the probability of apoptosis occurring should increase as the amount of functional AR decreases. It is beyond the scope of this paper to develop detailed descriptions of treatments for preventing or treating PCa. However, it is clear that treatments that maximize the pro-apoptotic properties of mAR, minimize the anti-apoptotic properties of iAR, and minimize overall bcl-2 production should be of therapeutic value. An example of this would be a combination of T, F, and methyl-piperidino-pyrazole(MPP) along with avoiding foods known to have components that selectively bind to ER-β. MPP has been shown\[[@B23]\] to be an antagonist of ER-α but not of ER-β. Conclusion ========== The E-D model presented here is consistent with known experimental data. Further research is needed to more completely verify and expand its hypotheses. It explains how T can promote either PCa growth or apoptosis and why E2 and DHT are essential for initial PCa growth.	PubMed Central	2024-06-05T03:55:55.604772	2005-3-18	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1079949/", "journal": "Theor Biol Med Model. 2005 Mar 18; 2:10", "authors": [ { "first": "A Edward", "last": "Friedman" } ] }
PMC1079950	Background ========== The paper describes a comprehensive model of a biofeedback system; it adopts a new approach to modeling. Using artificial neural networks (ANN) it is not easy to obtain a dynamic response that reflects dependence on hormone production. Therefore, the authors have endeavoured to design an approach that focuses on the internal state of the subject consequent on biofeedback stimulation. A biofeedback system involves a sensory organ and an appropriate stimulus. The stimulus is mediated through organs derived from specific biosensors \[[@B2]-[@B8]\]. If a subject has disorders involving parenchymal lesions, his or her internal state is likely to indicate exhaustion, as evident from output responses in a conservative system (see below). Thus, it is or may be possible to establish the internal state of the subject from the output responses. The model described in this paper has been developed primarily with a focus on the galvanic skin response (GSR) in biofeedback \[[@B9]\]; galvanic skin response training is also known as the electrodermal response (EDR). The device measures electrical conductance in the skin, which is associated with the activity of the sweat glands \[[@B9],[@B10]\]. Sweat gland activity is due to catecholamine secretion resulting from the stimulation of adrenergic receptors (discussed later). The GSR in a biofeedback system is caused by a stimulus that activates the sweat glands. This activation can be indicated by recording bio-potentials by placing the electrodes on the body surface. The instrumentation for recording consists of a set of amplifiers and filters designed for the purpose \[[@B9],[@B10]\] (Fig. [1](#F1){ref-type="fig"}). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Generalized Galvanic Skin Response. ::: ![](1742-4682-2-11-1) ::: If T~1~is the duration of the rising phase, T~2~is the duration of the decaying phase and ΔV is the residual homeostatic output level, the result from Fig. [1](#F1){ref-type="fig"} is tabulated below (Table [1](#T1){ref-type="table"}). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Records of the measurements of the SCR ::: Measure Measured value Measure Measured value Per unit value -------------------- ---------------- --------------- ---------------- ---------------- SCR latency 3s Peak response 84.5 mV 1 p.u. SCR rise time 9.69s Amplitude 25.5mV 0.3 p.u. Half recovery time 8.75s ::: Before we focus on the design of the biofeedback system, some important terminology needs to be discussed: Topological model, Transduction Phase, Unity biofeedback, Homeostasis, Homeostat, Residual Homeostatic output level, Feedback Control systems, Catecholamine Interactions, Conscious and subconscious parts of the brain and Dissipative & Conservative system. A topological modeloriginates from a root and spreads in tree-like branches. It affords a complete description of the interactions among the different parts of the system considered. The transduction phaseof a subject reflects physiological changes caused by hormone release consequent on stimulation. This phase is characteristic of an individual subject \[[@B2]-[@B7]\]. For example, the transduction phase of a psychosomatic patient is sometimes reflected during a journey in a high-speed vehicle, when the physiological outcome can adversely affect his mental condition, associated with headache and vomiting. Unity biofeedbackmeans that the homeostatic output is directly fed to the brain without going through the transduction phase, which incorporates conservative and dissipative systems. Homeostasisis the set of processes by which constant or \'static\' conditions are maintained within the internal environment of a subject \[[@B6],[@B7],[@B11]\]; a homeostatis a controller involved in maintaining homeostasis. In this paper the residual homeostatic output level, ΔV, has a particular value for each successive response. It can be correlated with the GSR \[[@B9]\]. The residual homeostatic output arises as a result of sustained catecholamine action, which often persists for minutes or hours; control is prolonged, not just instantaneous activation or inhibition \[[@B11]\]. The residual homeostatic output indicated by the GSR response signifies that sweating persists even after the withdrawal of the biofeedback stimulus \[[@B9]\]. Mammals are endowed with a vast network of feedback control systemswith controllers (homeostats) without which survival would be difficult \[[@B11]\]. In this control system a particular neuro-hormone exerts a negative feedback effect, preventing over-secretion of other hormones associated with over-activity of the muscles, unless there is specific disorder in the system \[[@B11]\]. Catecholamine interactionsare very important in biofeedback systems. Catecholamines are excitatory or inhibitory neurotransmitters or hormonal agents. The catecholamine neuro-hormones are epinephrine, norepinephrine, dopamine and serotonin. Epinephrine and norepinephrine function as excitatory hormones. Serotonin functions as an inhibitory hormone, and dopamine is excitatory in some areas and inhibitory in others. Stimulation of sympathetic nerves in the adrenal medullae causes large quantities of epinephrine and norepinephrine to be released into the circulating blood, which carries them to all tissues of the body. Norepinephrine increases the total peripheral resistance and thus elevates the arterial pressure; epinephrine raises the arterial pressure to lesser extent but increases the cardiac output more. Epinephrine has a 5 to 10 times greater metabolic effect than norepinephrine \[[@B11]\]. The adrenergic receptors include α and β receptors. The α-receptors control such physiological activities as vasoconstriction, iris dilatation, intestinal relaxation, intestinal sphincter contraction, pilomotor contraction and bladder sphincter contraction; β-receptors control (e.g.) vasodilatation, cardio-acceleration, increased myocardial strength, intestinal relaxation, uterus relaxation, bronchodilatation, calorigenesis, glycogenesis, lipolysis and bladder wall relaxation. It is therefore evident that both α and β receptors have inhibitory and excitatory functions \[[@B11]\]. Blood pressure transduction phases are associated with activation of α and β receptors \[[@B4]-[@B6]\]. The cerebral cortex, which includes the consciouspart of the brain, never functions alone but always in association with lower centres of the nervous system. In fact, the lower brain centres (or subconsciouspart of the brain) initiate wakefulness in the cerebral cortex \[[@B11]\]. The subconscious part of the brain performs vegetative functions; notably, the hypothalamus controls sympathetic and parasympathetic stimulation \[[@B11]\]. The sweat glands secrete large quantities of sweat when the sympathetic nerves are stimulated; they are controlled primarily by centers in hypothalamus that are usually considered to be parasympathetic centers \[[@B11]\]. Therefore, sweating could be called a parasympathetic function, although it is controlled by nerve fibres that are anatomically distributed through the sympathetic nervous system \[[@B11]\]. The sympathetic innervation of sweat glands results from a developmental change in transmitter phenotype (from catecholaminergic to cholinergic), making parasympathetic stimulation also possible \[[@B13]\]. In biofeedback systems, the subject undergoes different transduction phases. Depending on the nature of transduction phase a system can be classified as dissipativeor conservative. A dissipative system diverges from its original state during biofeedback; it may undergo successive stages during which the response decreases exponentially, with the characteristic features of a normal physiological system. A conservative system, in contrast, has an output characterised by exponentially rising phases due to sustained levels of catecholamines. Nowadays, biofeedback has important clinical applications in at least the following areas. Headache is a psychophysiological disorder associated with disturbances in the homeostatic relationship between mind and body. The classical psychosomatic disorders are included in this category, e.g. peptic ulcer, bronchial asthma, migraine and essential hypertension.\[[@B12]\] In classical migraine (in which the sufferer is sensitive to light and sound stimuli) there are neurological symptoms such as homonymous hemianopia, paresthesias, aphasia and hemiparesis, which precede the unilateral headache (tension headache) and are reflected in the subject\'s muscle activity \[[@B12]\]. Biofeedback is useful for migraine treatment. Stimulation or inhibition of specific adrenergic receptors, mediated through catecholamines, often help relieve the pain, inducing a feeling of drowsiness by a process associated with the smelling of ripe mango or fresh lemon \[[@B4]\]. The digestive system as a whole is governed by innumerable control mechanisms at the cell and tissue levels, whereby a pathway can be activated as needed or inhibited as products accumulate \[[@B12]\]. For example, acetylcholine is an excitatory cholinergic transmitter for smooth muscle fibers in some organs, but an inhibitory transmitter for smooth muscle in others. When acetylcholine excites a muscle fiber, norepinephrine ordinarily inhibits it. Conversely, when acetylcholine inhibits a fiber, norepinephrine usually excites it \[[@B11]\]. Cholinergic (muscarinic) receptors are involved in the parasympathetic activity. Muscarinic receptors are age dependent; their frequency decreases with increasing age. Moreover, the fall of blood pressure and pulse rate during parasympathetic stimulation (discussed later) is due to the combined effects of adrenergic and muscarinic receptors \[[@B14]\]. Adrenergic and cholinergic receptors in the autonomic nervous system play opposite roles. De-activation of the sympathetic innervation (which operates via adrenergic receptors) is followed by enhancement of the cholinergic receptors involved in parasympathetic stimulation in smooth muscle. Conversely, noradrenergic enhancement is diminished as cholinergic neurotransmission becomes established \[[@B14]\]. In the model discussed in this paper, the stimulation of adrenergic receptors diminishes concomitantly with blood pressure and pulse-rate (a dissipative system). This diminishing of the adrenergic receptor effect enhances cholinergic receptor activity automatically in the control of smooth muscle function. Similarly, in a conservative system, adrenergic receptor stimulation is enhanced concomitantly with the blood pressure and the pulse rate. This increasing effect of the adrenergic receptors will diminish the effects of cholinergic receptors automatically in the control of smooth muscle activity. Thus, cholinergic receptors automatically operate in conjunction with adrenergic receptors in the autonomic nervous system control of mammalian smooth muscle. The following extended account of the model focuses on the state of the subject (dissipative or conservative). Biofeedback can be fatal due to cardiac failure for subjects in an exhausted state, unless attention is given. In the paper, emphasis is placed on catecholamine stimulation and a temporal pattern of responses is obtained. It has been established that catecholamine secretion is not only of short duration but also persists for long periods (minutes or even hours) \[[@B11]\]. To take account of this, the authors have designed 1^st^order and 2^nd^order systems. In the 1^st^order system the response decays without oscillation during a short catecholamine secretion phase, whereas the 2^nd^order system represents a prolonged period marked with oscillation, concomitant with adrenergic stimulation leading to vasoconstriction and vasodilatation. A comprehensive biofeedback model consists of a brain, homeostat and transduction phase (Fig.2). The sensory organs are responsible for biofeedback stimulation. Biofeedback stimulates the nervous system concomitantly with homeostatic regulation of the body through hormonal activation. The role of the brain is central, adjusting the system in accordance with the biofeedback stimulus received from the sensory organ. Without the brain there would be no output response. Biofeedback stimulates the subconscious part of the brain, and depends upon the nature of stimulus received from the sensory organ in the subject\'s particular current environment. Both the conscious and subconscious parts of the brain are important in biofeedback. Dreams during sleep are sometimes responsible for locomotor action evoked through stimulation of subconscious parts of the brain. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Biofeedback Circuit. ::: ![](1742-4682-2-11-2) ::: Here, input stimulus to the biofeedback system is a step function while the homeostatic output response is exponential. The input stimulus may be optical (e.g. flash of light), auditory (e.g. tone), tactile (e.g. a blow to the Achilles tendon), or a direct electrical stimulation of some part of the nervous system.\[[@B8]\] Any sinusoidal or ramp input can be simplified by expressing it as a function of step inputs. For this reason the input is taken as a step. In this particular model, the output responses are of two types: exponential rise and exponential decay. Exponential rise signifies that the system is unable to withstand the biofeedback stimulus, depending on the responses of homeostat. Exponential decay signifies a normal homeostatic response. The homeostatic responses are regulated mainly by the functioning of the kidney and heart in tandem. A complex biofeedback output with multiple responses is shown in Fig. [3](#F3){ref-type="fig"}. ΔV is the residual homeostatic output level. In practice, subsequent biofeedback output responses occur, as shown. The residual homeostatic output level at each stage can sometimes exceed the corresponding value in the previous stage, depending on homeostatic responses. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### A biofeedback output with multiple responses. ::: ![](1742-4682-2-11-3) ::: A generalised GSR model was chosen.\[[@B9]\] For a step input, the body\'s biofeedback output response is identical to that illustrated in Fig [1](#F1){ref-type="fig"}. The GSR output was simulated using MATLAB 6.0. Different time constants for the rising and decaying phases were considered for simulation within a fixed interval. Simulation in this model was facilitated by the use of SIMULINK. Knowing that the input is a step and the output exponential, the entire transfer function of the system could be represented by the respective blocks (Fig. [4](#F4){ref-type="fig"}). K~1~and K~2~are the inverse time constants for the rising and decaying phases of the biofeedback output respectively; a~1~is the peak value of the of the biofeedback output response. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Block diagram representation of biofeedback output with single response. ::: ![](1742-4682-2-11-4) ::: Methods and Results =================== The p.u. (per unit) scale values signify normalisation of the curve to correlate a particular physiological phenomenon such as GSR. Qualitatively similar physiological responses can be fitted by a single curve, irrespective of amplitude, if per unit values are chosen. From Figs. [5](#F5){ref-type="fig"}, [6](#F6){ref-type="fig"}, [7](#F7){ref-type="fig"} we see that GSRs, qualitatively identical but of different amplitudes, are fitted by the single curve (Fig. [7](#F7){ref-type="fig"}). ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Galvanic skin response of a subject of a particular age. ::: ![](1742-4682-2-11-5) ::: ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Galvanic skin response of another subject of the same age. ::: ![](1742-4682-2-11-6) ::: ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Fitting (per unit values) of data in Fig 5 and Fig 6. ::: ![](1742-4682-2-11-7) ::: In this model (Fig. [8](#F8){ref-type="fig"}) the output is a single response. The values of K~1~and K~2~are taken as 0.2 and 0.3 and the time periods for the rising and decaying phases are taken as 5s, to correlate with the characteristic GSR response in biofeedback \[[@B9]\]. ::: {#F8 .fig} Figure 8 ::: {.caption} ###### Biofeedback output with single response. ::: ![](1742-4682-2-11-8) ::: From Fig. [8](#F8){ref-type="fig"} the residual homeostatic output level, ΔV, is calculated as 0.142 p.u. Now by keeping K~2~fixed we can change the value of K~1~and observe changes in the value of the residual homeostatic output. For i) K~1~= 0.2, ΔV = 0.1418 p.u; ii) K~1~= 0.25, ΔV = 0.142 p.u; and iii) K~1~= 0.15, ΔV = 0.1422 p.u. We can conclude that the residual homeostatic output level does not depend on the time constant of the rising phase of the biofeedback output response. In a real biofeedback system (in this case GSR), there may be more than one response. In that case the entire transfer function can be represented by a block diagram (Fig. [9](#F9){ref-type="fig"}). ::: {#F9 .fig} Figure 9 ::: {.caption} ###### Block diagram representation of biofeedback output with multiple response. ::: ![](1742-4682-2-11-9) ::: In respect of the homeostatic output level in GSR, the constants a~1,~a~2,~a~3~relate to the peak value; a~2~, a~4~represent residual output level. K~1~, K~3~, K~5~respectively indicate the slopes, i.e. the inverses of the respective time constants of the successive rising phases of the GSR; and K~2~, K~4~, K~6~respectively represent the inverses of the e time constants of the successive decaying phases. These constants are selected so as to represent the GSR attributable to activation of sweat glands concomitant with stimulation through catecholamine \[[@B9],[@B13]\]. The hormonal stimulation helps elicit physiological responses that obey an exponential law with rising and decaying phases. Case-1 ------ In the case of biofeedback with multiple responses, the K~1~and K~2~values for successive responses are taken as 0.2 and 0.3 respectively and K~3~, K~5~and K~4~, K~6~have values identical to K~1~and K~2~(Fig. [10](#F10){ref-type="fig"}). The time periods for the rising and decaying phases of successive responses are matched separately with the characteristic curve of the GSR response. From Fig. [10](#F10){ref-type="fig"} we observe that ΔV increases in successive responses. ::: {#F10 .fig} Figure 10 ::: {.caption} ###### Response vs. Time (Case-1). ::: ![](1742-4682-2-11-10) ::: Case-2 ------ Here (Fig. [11](#F11){ref-type="fig"}) K~1~= 0.2 and K~2~= 0.3; K~3~= 0.1, K~4~= 0.09; K~5~= 0.3, K~6~= 0.5; and the time periods of the 2^nd^and 3^rd^responses are taken to be half of the 1^st^response. ::: {#F11 .fig} Figure 11 ::: {.caption} ###### Response vs. Time (Case-2). ::: ![](1742-4682-2-11-11) ::: Case3 ----- Here (Fig. [12](#F12){ref-type="fig"}) K~1~= 0.2, K~2~= 0.3, K~3~= 0.05, K~4~= 0.03, K~5~= 0.02, K~6~= 0.01; again, the time periods of the 2^nd^and 3^rd^responses are taken to be half of the first response. ::: {#F12 .fig} Figure 12 ::: {.caption} ###### Response vs. Time (Case-3). ::: ![](1742-4682-2-11-12) ::: In all these cases we see that the residual homeostatic output level increases for each successive response \[[@B9]\]. With unity biofeedback the closed loop biofeedback transfer function is given by H(S) = G(S)/(1+G(S)), where G(S) is the open loop transfer function and the biofeedback output is given by Fig. [13](#F13){ref-type="fig"}. Now the whole system can be shown by a block diagram representation in Fig. [14](#F14){ref-type="fig"}. ::: {#F13 .fig} Figure 13 ::: {.caption} ###### Biofeedback output. ::: ![](1742-4682-2-11-13) ::: ::: {#F14 .fig} Figure 14 ::: {.caption} ###### Block diagram representation of closed loop transfer function with unit feedback. ::: ![](1742-4682-2-11-14) ::: Here the unit feedback control system is converted into an open loop control system, where the closed loop transfer function becomes an open loop transfer function. We next studied the output response when the transduction phase was incorporated into the feedback loop of the biofeedback system. The result can again be shown by a block diagram (Fig. [15](#F15){ref-type="fig"}). In the first order transduction phase, the constant \'a\' represents exponential rise or decay during the phase of catecholamine activation \[[@B4]-[@B6]\]. ::: {#F15 .fig} Figure 15 ::: {.caption} ###### Block diagram representation of system incorporating 1^st^order transduction phase. ::: ![](1742-4682-2-11-15) ::: The transduction phase can be either conservative or dissipative. Depending on the nature of the transduction phases, the biofeedback output of a closed loop model as shown in Fig. [16](#F16){ref-type="fig"} will typically show the relevant characteristic responses. The expression for dissipative and conservative systems due to incorporation of the transduction phase is: ::: {#F16 .fig} Figure 16 ::: {.caption} ###### The biofeedback output response when the 1st order transduction phase is incorporated in the feedback loop. ::: ![](1742-4682-2-11-16) ::: Tp(Φ~d~) = Φ~d0~± ∂(ψ~d~)/∂t and Tp(Φc) = Φc~0~± ∫(ψ~c~)dt where Φ~d0~and Φc~0~are the initial states of the dissipative and conservative system respectively, ψ~d~is the time dependent 1^st^order dissipative system and ψ~c~is the time dependent 1^st^order conservative system. Here, the transduction phase signifies the state of the internal environment of the subject \[[@B11]\]. It reflects the topological asymmetry of cellular organization, which shows a relaxation jump associated with hydrophobic linkages among polar heads \[[@B1]\]. Depending on the state of the subject, homeostasis is perturbed in a conservative system. This is the first order system transduction phase where the value of a is taken as 2 and the output appears as Case-I ------ Here peak amplitude = 0.101 p.u and settling time = 17 s From Fig.[16](#F16){ref-type="fig"} we see that the exponentially decaying output phase indicates that the subject returns to the original state within a time frame depending on the duration of the catecholamine signal. When the 2nd order transduction phase is incorporated into the biofeedback loop, the block diagram representation of the system is shown below. To represent the 2^nd^order transduction phase, the constants \'a\' and \'b\' are selected so that there will be simultaneous exponential rise and decay (Fig. [17](#F17){ref-type="fig"}). This is shown in Fig. [18](#F18){ref-type="fig"}, which illustrates the catecholamine activation phase for a normal subject (dissipative system) \[[@B4],[@B5],[@B11],[@B13]\]. Fig. [18](#F18){ref-type="fig"} represents the transduction of blood flow mediated by catecholamine. ::: {#F17 .fig} Figure 17 ::: {.caption} ###### The block diagram representation when the 2^nd^order transduction phase is incorporated in the feedback loop. ::: ![](1742-4682-2-11-17) ::: ::: {#F18 .fig} Figure 18 ::: {.caption} ###### Effect of sympathectomy on blood flow in the arm and the effect of a test dose of norepinephrine before and after sympathectomy (lasting only 1 minute or so), showing supersensitizationof the vasculature to norepinephrine. ::: ![](1742-4682-2-11-18) ::: Assuming a = 1, b = 1 we can have the system response in Fig. [19](#F19){ref-type="fig"}. ::: {#F19 .fig} Figure 19 ::: {.caption} ###### Biofeedback output response when 2^nd^order transduction phase is incorporated in the feedback loop. ::: ![](1742-4682-2-11-19) ::: Case-II ------- Here peak amplitude = 0.129 p.u and settling time = 19 s. Fig. [18](#F18){ref-type="fig"} illustrates the fluctuations of parameters such as blood pressure and pulse rate, which persist for a certain period of time concomitant with the sustained catecholamine signal. Keeping the value of b fixed at 1 and by putting a = 0.5 we obtain the output response shown in Fig. [19](#F19){ref-type="fig"}. Case-III -------- Here (Fig. [20](#F20){ref-type="fig"}) peak amplitude = 0.158 p.u and settling time = 18.3 s ::: {#F20 .fig} Figure 20 ::: {.caption} ###### Response amplitude vs Time (a = 0.5). ::: ![](1742-4682-2-11-20) ::: Case-IV ------- Here (Fig. [21](#F21){ref-type="fig"}) peak amplitude = 0.171 p.u and settling time = 30.2 s ::: {#F21 .fig} Figure 21 ::: {.caption} ###### Response amplitude vs Time (a = 0.3). ::: ![](1742-4682-2-11-21) ::: Case-V ------ Here (Fig. [22](#F22){ref-type="fig"}) peak amplitude = 0.181 p.u. and settling time = 99.2 s ::: {#F22 .fig} Figure 22 ::: {.caption} ###### Response amplitude vs Time (a = 0.1). ::: ![](1742-4682-2-11-22) ::: Figs. [19](#F19){ref-type="fig"}, [20](#F20){ref-type="fig"}, [21](#F21){ref-type="fig"}, [22](#F22){ref-type="fig"} model states with different values of \'a\'. With decreasing \'a\' values, the settling time increases with the increase of oscillations. This is true for a subject with sustained biofeedback. Case-VI ------- Peak amplitude = 2.41 p.u and damping freq = 0.002463Hz (Fig. [23](#F23){ref-type="fig"}). ::: {#F23 .fig} Figure 23 ::: {.caption} ###### Response amplitude vs Time (a = 0.015). ::: ![](1742-4682-2-11-23) ::: Case-VII -------- Here peak amplitude = 1.76 p.u and damped frequency = 1/(126-40.7) = 1/85.3 = 0.01172Hz (Fig. [24](#F24){ref-type="fig"}). ::: {#F24 .fig} Figure 24 ::: {.caption} ###### Response vs Time (when damping is absent, i.e. a = 0). ::: ![](1742-4682-2-11-24) ::: Figs. [23](#F23){ref-type="fig"}, [24](#F24){ref-type="fig"} represent a subject with a permanent disorder; the biofeedback stimuli cause the disorder to be manifest. By putting a = 0 we can have the output response. Here we clearly see that sustained oscillations amplify in a conservative transduction phase due to the prolonged period of catecholamine activation. Conclusion ========== The features of both dissipative and conservative systems are represented in this comprehensive model, which is based on catecholamine activation. The transduction phase of the 2^nd^order system in biofeedback can act as either a dissipative or a conservative system depending on the system dissipation factor (which is related to catecholamine production). For a dissipative system the catecholamine signal is of shorter duration, whereas for a conservative system it survives for a longer period. Biofeedback can sometimes produce complex responses in biological systems depending on how sustained the catecholamine signal is; these complexities are represented by the present model. In the context of this paper, the envelopes of the exponentially rising and decaying phases also represent the stimulation of adrenergic receptors in monotonic phase concomitant with the catecholamine production. Adrenergic and cholinergic receptors have opposing roles in the autonomic nervous system. Downregulation of sympathetic innervation via adrenergic receptor is followed by enhancement of the cholinergic receptors involved in parasympathetic stimulation in smooth muscle. Conversely, noradrenergic enhancement is diminished as cholinergic neurotransmission becomes established. Thus it may be concluded that cholinergic receptors automatically participate, along with adrenergic receptors, in the autonomic nervous system control of mammalian smooth muscle function. In this paper a new conceptual approach has been taken to modeling dynamic responses in biofeedback that depend on hormone activity, by introducing homeostats and transduction phases in the feedback path. Competing Interests =================== As head of the Department of Electrical Engineering, Jadavpur University, Professor Basak requested the University authorities to obtain membership of <http://www.biomedical-engineering-online.com> and the university has given due consideration to this request. Authors\' contributions ======================= Professor T. K. Basak received a third world scientist award from ICTP, Trieste, Italy and worked with Professor A. Glilozzi in the Dept of Biophysics, University of Genoa, Italy in 1985. He furnished the innovative idea in the present paper and provided comprehensive guidance to the team from the outset. After completing his Masters degree in electrical engineering under the supervision of Professor Basak, Mr. Suman Halder began Ph.D. work under the same supervisor and was involved with the work until the completion of the paper. Ms. Madona Kumar and Mrs. Renu Sharma were Masters students under Professor Basak\'s supervision and participated in the completion of the work and the preparation of the manuscript. Ms. Bijoylaxmi Midya\' a lecturer in the Department of Applied Electronics & Instrumentation Engineering, Haldia Institute of Technology, Haldia, is doing Ph.D. work under Prof. Basak and contributed to the completion of the paper. Acknowledgements ================ The authors are grateful to the authorities of Jadavpur University and to Prof. T. K. Ghoshal, ex-head of the Electrical Engineering Department. Professor T. K. Basak is particularly indebted for inspiration received from his late wife, Mala Basak who is in the heavenly abode of Shree Shree Ramakrishna Paramhansa.	PubMed Central	2024-06-05T03:55:55.606226	2005-3-21	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1079950/", "journal": "Theor Biol Med Model. 2005 Mar 21; 2:11", "authors": [ { "first": "Tapas K", "last": "Basak" }, { "first": "Suman", "last": "Halder" }, { "first": "Madona", "last": "Kumar" }, { "first": "Renu", "last": "Sharma" }, { "first": "Bijoylaxmi", "last": "Midya" } ] }
PMC1079951	Introduction ============ Myosins, actin-based motor proteins, are expressed as multiple isoforms in all eukaryotic cells. They are oligomers consisting of one or two heavy chains to which one or more light chains are non-covalently attached. Myosins have been classified into 18 families based on the amino acid sequence differences in the N-terminal head domains, which contain highly conserved regions including actin- and nucleotide-binding sites \[[@B1],[@B2]\]. The tail of myosin is the most variable domain and seems to be responsible for the specific role myosin plays in the cell. Functional activities of most myosins such as actin-dependent ATPase activity or ability to move actin filaments in vitroare regulated in several ways, mainly by phosphorylation of the regulatory light chain, Ca^2+^-binding, or phosphorylation of the heavy chain \[[@B1],[@B3]\] It has been previously claimed that the myosin heavy chain (MHC)undergoes tyrosine phosphorylation during insulin-mediated skeletal muscle differentiation, thus linking signal transduction to highly ordered myosin assembly \[[@B4]\]. Insulin modulates an association of myosin with C-terminal SRC kinase (Csk), a tyrosine kinase signalling molecule, and these interactions are fundamental in skeletal muscle differentiation. Although the claims of tyrosine phosphorylation of MHC in vivoremain somewhat controversial, tyrosine phosphorylation of non-muscle MHC IIa has also been implicated as an early event in human platelet activation \[[@B5]\]. To settle this controversy -and establish the role, if any, of MHC tyrosine phosphorylation it is important to identify sites at which such phosphorylation events may occur. We have mapped potential phosphorylation sites on the skeletal muscle myosin heavy chain utilizing an integrated bioinformatics approach, supporting web-based motif predictions with evolutionary and structural data. Of all the sites analyzed in the bioinformatics approach, the data suggest Y163 and Y1856 as the leading candidates for insulin-mediated tyrosine phosphorylation. Methods ======= Tyrosine Phosphorylation Predictions ------------------------------------ Tyrosine phosphorylation site predictions were made with two different online resources using the sequences described below. NetPhos 2.0 produces neural network predictions based on sequence and structure \[[@B6]\]. Scansite predicts target motifs for different kinases using a positional selectivity matrix based on peptide library screening data \[[@B7],[@B8]\] In addition, Scansite predictions were made for known phosphotyrosine recognition motifs for evidence of downstream signalling events. All Scansite predictions were made on the \'Low Stringency\' setting to identify as many putative sites as possible. These sites were then supported or rejected on the basis of further analysis as described below. Evolutionary Analysis --------------------- Protein sequences for adult skeletal muscle myosin heavy chains (MYHSA) 1 and 2 were extracted from the SwissProt database \[[@B9]\] MYHSA1 \[SwissProt : MYH1\_HUMAN, P12882\]; MYHSA2 \[SwissProt ID: MYH2\_HUMAN, Q9UKX2\] and used as query sequences to extract closely related homologous proteins. First, BLAST \[[@B10]\] was used to search SwissProt-TrEMBL \[[@B9]\] and the known, novel and Genscan-predicted peptides of five EnsEMBL genome databases (Human, Mouse, Rat, Fugu, Zebrafish) \[[@B11]\] Redundant sequences were removed and ALIGN \[[@B12],[@B13]\] was used to make pairwise alignments of each homologue with MYH1\_HUMAN and to calculate the percentage identity across the entire length of the protein. Vertebrate homologues with at least 60% global identity were processed using an in-house homologue processing tool, HAQESAC \[[@B14]\]. Homologues were aligned using CLUSTALW \[[@B15]\] and badly-aligned sequences eliminated from the dataset. A neighbour-joining tree with 1000 bootstrap replicates was constructed using CLUSTALW and the sequences were grouped into subfamilies of orthologous proteins. The clade corresponding to skeletal muscle myosin heavy chains in Amniota (mammals, reptiles and birds) were then used as sequences for tyrosine phosphorylation motif prediction as described above. Secondary Structure Prediction ------------------------------ Secondary structure predictions were made for MYH1\_HUMAN using the PSIPRED V2.3 website \[[@B16]\]. Because of the length of the protein, it was submitted in two overlapping chunks: residues 1--814 and 800 +. 3D Structure Analysis --------------------- 3D structures were obtained from the Protein Data Bank (PDB) \[[@B17]\] and viewed with the RasMol viewer \[[@B18]\]. Three myosin heavy chain structures were identified: 2MYS, Chicken adult skeletal muscle myosin heavy chain; 1BR2, chicken gizzard smooth muscle myosin heavy chain; and 1B7T, Aequipecten irradians(Bay scallop) striated muscle myosin heavy chain. The corresponding SwissProt sequences \[Swiss -Prot :2MYS: MYSS\_CHICK, P13538\]; \[Swiss-Prot1BR2: MYHB\_CHICK, P10587\]; \[Swiss -Prot1B7T: MYS\_AEQIR, P24733\] were downloaded and aligned with Human MYH1\_HUMAN and MYH2\_HUMAN using CLUSTALW. This alignment was used with the skeletal muscle myosin heavy chains (above) to assign putative tyrosine phosphorylation sites to their corresponding residues in the homologous 3D structures. Visualisation with RasMol and DSSP solvent accessibility data \[[@B19]\] was then used to infer whether potential sites of tyrosine phosphorylation were surface-exposed or buried. Results ======= In total, twenty-three myosin heavy chain sequences were used for tyrosine phosphorylation motif prediction, which were divided into five groups of orthologous sequences (Figure [1](#F1){ref-type="fig"}). Important motifs are likely to be conserved during evolution and so we considered only those sites that were predicted to be phosphorylation motifs in all the sequences of at least one orthologous group. Because phosphorylation site predictors have a tendency to over-predict, we increased stringency by accepting only those motifs that received a NetPhos score of 0.8 or higher, or were predicted by both NetPhos and Scansite, in at least one human adult skeletal myosin heavy chain. This yielded fourteen putative sites (Table [1](#T1){ref-type="table"}). Of these, six were predicted by both methods, including two motifs that were conserved across all sequences (MYH1\_HUMAN Y163 and Y1856). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Neighbour-joining phylogeny of MHC homologues, with bootstrap support. PDB structure 2MYS is marked with a black diamond. ::: ![](1742-4682-2-12-1) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Summary of predicted tyrosine phosphorylation sites. ::: Site^a^ NetPhos^b^ Scansite^c^ 2D^d^ Surface Accessibility^e^ --------- ------------ ------------- ------- -------------------------- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ------ -------- ------ ------ ------ 47 \- \- \- Y \- \- \- E(4) (14.5) (21) (16) 18.5 54 \- \- \- \- \- \- Y \- \- \- \- \- \- E(7) (52) (60) 56.0 85 Y Y Y Y Y Y Y C(2) (34.7) (28) 16 26.2 163 Y Y Y Y Y Y Y YP YP YP YP YP YP YP H(9) 13.3 47 8 22.8 286 Y Y Y Y Y Y Y Y Y H(7) (0) 0 0 0.0 313 Y \- YP \- Y Y YP Y Y C(4) 7 5 8 6.7 413 Y Y Y Y \- Y Y \- E(7) (68) 68.0 435 \- Y \- H(9) (0.7) 4 3 2.6 504 Y Y Y Y Y Y Y H(9) 0 6 9 5.0 719 Y Y Y Y Y YP YP YP YP YP YP Y H(6) (21.8) 19 9 16.6 1379 Y Y Y Y Y Y Y H(6) 1464 \- Y \- Y Y Y \- \- \- \- H(7) 1492 Y Y Y Y Y Y Y YP P H(6) 1856 Y Y Y Y Y Y Y Y Y Y Y Y Y Y H(4) a\. Sites are numbered relative to the MHC sequence MYH1\_HUMAN/P12882. b\. Y indicates predicted tyrosine phosphorylation site in all the sequences of orthologous group, with a score of ≥ 0.8 in at least one human sequence. Dashes indicate lack of a tyrosine in that position. c\. Y indicates predicted tyrosine phosphorylation site in all the sequences of orthologous group on \'Low Stringency\'. P indicates predicted phosphotyrosine recognition site in all the sequences of orthologous group on \'Low Stringency\'. Dashes indicate lack of a tyrosine in that position. d\. PSIPRED (McGuffin, Bryson and Jones 2000) secondary structure position for MYH1\_HUMAN. Letters indicate predicted secondary structure (H = helix, E = strand, C = coil). Numbers in brackets are confidence measures (0 = low, 9 = high). e\. Surface accessibility figures are \"numbers of water molecules in contact with this residue \10, or residue water exposed surface in Angstrom\\2\" (Kabsch and Sander 1983). Missing values indicate residues missing from the PDB structure. Values in brackets indicate residues that are not tyrosines in the PDB structure. ::: To be phosphorylated, tyrosine residues must be accessible on the surface of the protein. Although the three-dimensional conformations of homologous myosin molecules will not be identical, the high degree of sequence conservation between human adult skeletal muscle myosin heavy chains and the three myosin sequences present in PDB allowed the inference of solvent accessibility. This was confirmed by the generally good agreement in surface accessibility measures both between models and between the different myosin chains of 1BR2 (data not shown). From these data, two sites (Y286 and Y435) were buried while a further two (Y313 and Y504) had very low solvent accessibility (Table [1](#T1){ref-type="table"}). 3D data were not available for the four tyrosines in the C-terminal of the protein. If tyrosine phosphorylation of MMHC-II is part of a signalling cascade, it is likely that some other protein will interact with the phosphotyrosine. We used Scansite to look for phosphotyrosine interaction motifs and found three SH2 domain recognition motifs that matched potentially exposed phosphorylation sites. Because Scansite also identifies the interacting protein, we interrogated the Gene Cards \[[@B20]\] entry for each kinase and SH2 domain for expression patterns. Only four kinases and two SH2 domains had evidence from UniGene \[[@B21]\] or SAGE \[[@B22]\] of expression in skeletal muscle, while only one kinase (INSR) and one SH2 domain (PIK3R1) had evidence from both (Table [2](#T2){ref-type="table"}). Interestingly, both of the latter pair were predicted to interact with the same, totally conserved, motif (MYH1\_HUMAN Y163). Furthermore, both are involved in insulin-mediated pathways (see Discussion). Two kinases expressed in skeletal muscle were predicted to interact with Y1856. These were an SRC kinase and ABL1, which interacts with SORBS1 following insulin stimulation \[[@B23]\]. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Interacting enzymes predicted by Scansite. ::: Site^a^ Enzyme^b^ Gene Card UniGene^c^ SAGE^c^ Full Name --------- ------------------------- ----------- ------------ --------- ----------------------------------------------------------------------------------------------------------- 163 EGFR Kinase EGFR Yes No EGFR (epidermal growth factor receptor (erythroblastic leukemia viral (v-erb-b) oncogene homolog, avian)) Insulin Receptor Kinase INSR Yes Yes INSR (insulin receptor) p85 SH2 PIK3R1 Yes Yes PIK3R1 (phosphoinositide-3-kinase, regulatory subunit, polypeptide 1 (p85 alpha)) Shc SH2 SHC1 No No SHC1 (SHC (Src homology 2 domain containing) transforming protein 1) 286 EGFR Kinase EGFR Yes No EGFR (epidermal growth factor receptor (erythroblastic leukemia viral (v-erb-b) oncogene homolog, avian)) 313 EGFR Kinase EGFR Yes No EGFR (epidermal growth factor receptor (erythroblastic leukemia viral (v-erb-b) oncogene homolog, avian)) Fgr Kinase FGR No No FGR (Gardner-Rasheed feline sarcoma viral (v-fgr) oncogene homolog) PDGFR Kin PDGFRB No Yes PDGFRB (platelet-derived growth factor receptor, beta polypeptide) Itk SH2 ITK No No ITK (IL2-inducible T-cell kinase) Fgr SH2 FGR No No FGR (Gardner-Rasheed feline sarcoma viral (v-fgr) oncogene homolog) 719 Lck Kinase LCK No No LCK (lymphocyte-specific protein tyrosine kinase) Abl Kinase ABL1 No Yes ABL1 (v-abl Abelson murine leukemia viral oncogene homolog 1) Itk SH2 ITK No No ITK (IL2-inducible T-cell kinase) Src Kinase SRC Yes No SRC (v-src sarcoma (Schmidt-Ruppin A-2) viral oncogene homolog (avian)) 1492 Src Kinase SRC Yes No SRC (v-src sarcoma (Schmidt-Ruppin A-2) viral oncogene homolog (avian)) Lck SH2 LCK No No LCK (lymphocyte-specific protein tyrosine kinase) Fgr SH2 FGR No No FGR (Gardner-Rasheed feline sarcoma viral (v-fgr) oncogene homolog) Src SH2 SRC Yes No SRC (v-src sarcoma (Schmidt-Ruppin A-2) viral oncogene homolog (avian)) 1856 Lck Kinase LCK No No LCK (lymphocyte-specific protein tyrosine kinase) Src Kinase SRC Yes No SRC (v-src sarcoma (Schmidt-Ruppin A-2) viral oncogene homolog (avian)) Abl Kinase ABL1 No Yes ABL1 (v-abl Abelson murine leukemia viral oncogene homolog 1) a\. Sites are numbered relative to the MHC sequence MYH1\_HUMAN/P12882. b\. Enzyme identified by Scansite. c Skeletal muscle expression data from GeneCards (Rebhan et al. 1997). ::: Discussion ========== Bioinformatics alone cannot identify a functional motif; supporting experiments will always be needed for conclusive evidence. Nevertheless, while other sites cannot be categorically excluded, the combined data presented here identify Y163 and Y1856 as the most likely sites for tyrosine phosphorylation events in skeletal muscle. Both NetPhos and Scansite predicted these motifs for all mammalian adult skeletal myosin heavy chain sequences analysed, indicating strong evolutionary conservation (Figure [1](#F1){ref-type="fig"}). A kinase predicted to be responsible for phosphorylation of each site is expressed in skeletal muscle, as was an SH2 domain protein that was predicted to interact with a phosphotyrosine at Y163. Analysis of predicted secondary structures and homologous 3D structures indicates that these sites may be accessible on the protein surface. The position of Y163 as part of an alpha helix within the globular myosin head domain does raise concerns that it is potentially difficult to phosphorylate, even though it is on the surface of the domain. Nevertheless, depending on the relative conformations of the solved in vitrochicken myosin structures compared to in vivo*human myosin, Y163 might still be available for phosphorylation. Y1856 is in region of low predicted secondary structure (Table [1](#T1){ref-type="table"}), indicative of a flexible loop region more usually associated with phosphorylation sites. Myosin heavy chain (MHC) undergoes tyrosine phosphorylation during insulin-mediated differentiation in skeletal muscles and the degree of phosphorylation increases in line with differentiation \[[@B4]\]. Interestingly, for the strongest candidate tyrosine phosphorylation site, Y163, both the kinase and interacting SH2 domain predicted by Scansite are involved in insulin-mediated pathways. The kinase INSR is a transmembrane receptor that binds insulin \[[@B24]\] while the SH2 domain protein PIK3R1 is necessary for the insulin-stimulated increase in glucose uptake and glycogen synthesis in insulin-sensitive tissues \[[@B25]\]. We can therefore conclude that Y163 remains a strong candidate site for insulin-mediated tyrosine phosphorylation of myosin heavy chain, despite concerns over accessibility. As phosphorylation sites are often in the tails of proteins, the tyrosines outside the main globular domains, namely Y1379, Y1492 and Y1856, are also potential candidates for phosphorylation. The strongest of these is Y1856, which is both C-terminal and predicted to be phosphorylated by the kinases SRC and ABL1, which are found in skeletal muscle (Table [2](#T2){ref-type="table"}). Perhaps of most interest is ABL1, a protein known to be associated with \"Sorbin and SH3 domain containing 1\" (SORBS1) during insulin signalling in other cell lines \[[@B23]\]. SORBS1 is highly expressed in skeletal muscle (data not shown) and is involved in formation of actin stress fibres and focal adhesions; its orthologue, CAP, has been identified as an important adaptor during insulin signalling in mice \[[@B26]-[@B28]\]. Furthermore, the SORBS1 gene has been implicated in the pathogenesis of human disorders with insulin resistance \[[@B29]\]. Csk has been shown to be associated with the hormone 1, 25(OH)~2~-vitamin D~3~resulting in the stimulation of the growth-related mitogen-activated protein kinase (MAPK). The phosphorylated form of MAPK is then translocated to the nucleus where it induces the expression of c-myc oncoprotein associated with skeletal muscle proliferation \[[@B30]\]. In addition, Csk has also been implicated in the regulation of integrins and the control of cell attachment and shape \[[@B31]\] Goel et al. showed that insulin can phosphorylate myosin, leading to an association with Csk and thus to a decrease in c-Src activity. This has also been shown in fibroblast cell lines following stimulation of the insulin-like growth factor-I receptor \[[@B32]\]. This demonstrates the potential for skeletal muscle differentiation after phosphorylation of Y163 and/or Y1856 of the MHC. Harney et al. have shown that non-muscle myosin heavy chain type IIA in platelets undergoes tyrosine phosphorylation and subsequent dephosphorylation in a time-dependent manner \[[@B5]\]. In common with other cells, the cytoskeleton of platelets comprises actin filaments, microtubules and myosin molecules. Myosins form rings within the platelet that maintain a spherical shape and several lines of evidence suggest that these rings reorient following platelet activation to permit spreading \[[@B33],[@B34]\]. While myosin function therefore appears critical to platelet spreading, studies using cytoskeleton inhibitors have shown that at least the early events of platelet activation are not dependent on the cytoskeletal changes \[[@B35]\]. Our work is suggestive that tyrosine phosphorylation of myosin heavy chain, whether in skeletal muscle or in platelets, is a significant event that may initiate cytoskeletal reorganization of muscle cells and platelets. Our studies provide a good starting point for further functional analysis of MHC phosphor-signalling events within different cells. Supplementary Information ------------------------- Full prediction results, sequence alignments and links to in-house software used can be found at: <http://www.bioinformatics.rcsi.ie/~redwards/phos/> Competing interests =================== The author(s) declare that they have no competing interests. Acknowledgements ================ The authors would like to thank Dr G. Cagney (GC) and Dr Patricia Maguire (PBM) for many helpful comments during the analysis and preparation of the manuscript. In addition this work was supported in part by a fellowship from Enterprise Ireland (PBM), the Health Research Board of Ireland (PBM) and the Higher Education Authority of Ireland (GC) and by a Science Foundation Ireland award (grant no. 02/IN.1/B117).	PubMed Central	2024-06-05T03:55:55.608836	2005-3-25	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1079951/", "journal": "Theor Biol Med Model. 2005 Mar 25; 2:12", "authors": [ { "first": "DF", "last": "Harney" }, { "first": "RK", "last": "Butler" }, { "first": "RJ", "last": "Edwards" } ] }
PMC1079952	Introduction ============ Blood is a liquid-liquid suspension because erythrocytes exhibit fluid-like behavior under certain shear conditions \[[@B1]\]. The dependence of viscosity on shear rate is one of the most widely used rheological measurements \[[@B2]\]. Normal blood also thins when it is sheared, therefore its apparent viscosity is highly sensitive to shear rates below 100 s^-1^\[[@B2],[@B3]\]. The objective of this research was to determine blood viscosity at yield velocity (and therefore shear stress) equal to zero. Our previous studies have shown that conditions such as blunt trauma to large vessels may lead to boundary layer separation where du/dy = 0, i.e. to the appearance of zones where shear stress equals zero \[[@B4]\]. A further aim of this research was to evaluate quantitative aspects of rouleau formation from erythrocytes when the yield velocity is equal to zero. Methods ======= Various calculations have been made for the viscosity of a mixture and the coefficient of constraint \[[@B5]-[@B7]\]. There is considerable variation in such calculations, resulting from different combinations of phases. This variation apparently reflects the non-Newtonian nature of concentrated viscous disperse mixtures and the insufficiency of the variables ρand μalone (where ρis density and μis viscosity) to determine the mechanical properties of such mixtures. In this regard, experiments over the range of operating parameters are needed for any mixture to determine pressure loss using different rheological models; in particular, the model of a viscous fluid with an effective viscosity coefficient. It must be noted that when f~2~\> 0.1 (where f~2~is the volume fraction of the second phase), not only the shape and size of the erythrocytes but also the irregular arrangement of the particles and their collisions with each other and with the solid walls have substantial effects on the effective viscosity and other rheological characteristics of the mixture \[[@B8],[@B9]\]. The problems mentioned above have led to studies of group sedimentation at f~2~\> 0.1 in the interpenetrating model of two- or multi-phase media \[[@B10]\]. These studies usually deal with either high- or low-concentration mixtures. Mechanisms of sedimentation in moderately concentrated mixtures, which are rather common, have not been fully investigated. Mathematical modeling of group sedimentation of particles (in our case, rouleaux) in two-phase interpenetrating media \[[@B11]\] should take into account not only the Stokes force \[[@B12]\] but also other forces that are given in \[[@B13]\]: ![](1742-4682-2-13-i1.gif) where F~12~^(A)^is a buoyancy force, p-pressure difference, χ^(m)^- coefficient of constraint, ρ- density of the first phase, K^(μ)^-- coefficient of phase interaction, μ~1~and μ~2~-- viscosities of the first and second phases, f~2~-- the volume fraction of the second phase. It is also important to calculate μ, the viscosity of the blood mixture, which depends on the volume fraction of particles. In this case it is possible to determine the force F~12~^(μ)^. F~12~^(μ)^is a frictional force or Stokes force that results from viscous forces involved in the interaction between phases. F~12~^(μ)^is calculated using the difference between velocities (slippage) u~1~- u~2~, the particle size a, the quantities and shapes of inclusions, and the physical properties of the phases (see equation 1). (The effects of the shape and multiplicity of particles, and of some other variables included in the expression for F~12~^(μ)^, are accounted for in coefficients K^(μ)^in (1)). Using all of the above, I shall determine blood viscosity as a variable dependent on a volume fraction of particles. This will allow me to determine blood viscosity at a yield velocity of zero, and the number of rouleaux as a variable dependent on erythrocyte concentration, shear stress and yield velocity. Determination of viscosity of a mixture as a variable dependent on volume fraction of particles ----------------------------------------------------------------------------------------------- Sedimentation of a single particle is based on the Stokes law, according to which a frictional force resulting from the motion of spherical particles with diameter dand velocity Vin a medium of viscosity μis expressed by the equation: ![](1742-4682-2-13-i2.gif) where a-- radius of particles (inclusions) and V-- velocity of particle precipitation. In the general case of a multiphase medium, the frictional force or Stokes force F~12~^(μ)^, which results from viscous forces involved in the interactions between phases, is calculated using the difference between velocities (slippage) u~1~- u~2~, the particle size a, the quantity and shape of inclusions, and the physical properties of the phases. Multiphase models are based on the idea of interpenetrating media, where the system of particles is replaced by a mathematical continuum and particle size is considerably less than the distance over which flow conditions may change \[[@B11]\]. The force of gravity acting on a particle is calculated using the specific gravity of the particle; that is: ![](1742-4682-2-13-i3.gif) where ρ~1~;ρ~2~;gare respectively the density of the fluid, the density of the particle, and the acceleration due to gravity. ![](1742-4682-2-13-i4.gif) is a buoyancy force (Archimedes force); ![](1742-4682-2-13-i5.gif) is a frictional force or Stokes force. Force ![](1742-4682-2-13-i4.gif) causes a particle to accelerate. In addition to gravity, the particle is affected by the frictional force, which acts in the opposite direction and has a value directly proportional to the velocity according to the Stokes law. This means that force ![](1742-4682-2-13-i5.gif) and gravity ![](1742-4682-2-13-i4.gif) tend to cancel each other out. Therefore, the motion proceeds with a constant velocity Vthat can be determined from equations (2) and (3): ![](1742-4682-2-13-i6.gif) where Vs-- velocity of precipitation of a single particle. Sometimes investigators have to deal with the sedimentation of multiple particles in concentrated mixtures. Formulae for the velocity of sedimentation of particles, dependent on the concentration and velocity of a single particle in an infinite fluid, can be derived using statements from the interpenetrating model \[[@B13]\] and the Euler equation \[[@B14]\]. Assuming that a specific volume has two phases differing in specific gravity, the particles with the greater specific gravity will start moving down a channel, so that a process of mutual penetration occurs. The flow of the fluid can be expressed by criterion equations: ![](1742-4682-2-13-i7.gif) where E~u~-- Euler number, A-- coefficient of proportionality, R~e~-- Reynolds number; or: ![](1742-4682-2-13-i8.gif) In the process of sedimentation when the concentration of inclusions is rather high and the particle size is small, flow is laminar; m = - 1 and n = 1 (where m and n are criterion coefficients). Taking into account data from \[[@B13]\]: ![](1742-4682-2-13-i9.gif) where S~i~-- particle surface area; f~1~-- volume fraction of the first phase; f~2~-- volume fraction of the second phase Dividing the continuity equation: V~1~S = V~1i~S~1~ by S, I obtain: V~1~= f~1~V~1i~ where S is the area of the canal section. Therefore: ![](1742-4682-2-13-i10.gif) Using equations (5) and (2), I can transform the last equation into the Kozeny-Carman formula for restrained sedimentation in a laminar flow: ![](1742-4682-2-13-i11.gif) where Alies within the range 80--110. Dividing equation (7) by the number of particles per unit of volume allows the resistance force applied by the fluid to a single particle to be derived as: ![](1742-4682-2-13-i12.gif) Where F\* -- resistance force created by the fluid and acting on a single particle, and χ-- coefficient of resistance for precipitation of multiple particles. The resistance force applied to a single particle during precipitation in a fluid is known to be \[[@B12],[@B15]\]: ![](1742-4682-2-13-i13.gif) For particles suspended in a fluid: F\* = F~12~ therefore from (8) and (9) it follows that: ![](1742-4682-2-13-i14.gif) where β-- the ratio of the velocity of sedimentation of the group of particles to the velocity of sedimentation of a single particle, and χ~c~-- the coefficient of resistance when precipitating a single particle in an infinite fluid. From (10), when f~1~→ 1 it follows that: ![](1742-4682-2-13-i15.gif) when the Reynolds numbers are small: ![](1742-4682-2-13-i16.gif) where c -- constant. Therefore, it can be assumed that: ![](1742-4682-2-13-i17.gif) From equations (10) and (11) it follows that: ![](1742-4682-2-13-i18.gif) where: ![](1742-4682-2-13-i19.gif) where ν-- the coefficient of viscosity. When the motion is laminar, according to the Stokes law: ![](1742-4682-2-13-i20.gif) Substituting this expression in equation (12), it follows that: ![](1742-4682-2-13-i21.gif) If one considers the sedimentation of a particle in a suspension with viscosity μ~m~and density ρ~m~, then the equilibrium equation \[[@B13]\] can be expressed as: ![](1742-4682-2-13-i22.gif) ![](1742-4682-2-13-i23.gif) ρ~m~= f~1~ρ~1i~+ f~2~ρ~2i~ Using equations (14), (15) and (3) and the condition V~1~= 0 it follows that: ![](1742-4682-2-13-i24.gif) Substituting the relative velocity equation (13) into equation (17), it follows that: ![](1742-4682-2-13-i25.gif) When f~1~→ 1 and c = 2.5, this reduces to the Einstein formula: ![](1742-4682-2-13-i26.gif) From the calculation given in Figure [1](#F1){ref-type="fig"}, it follows that equation (18) is consistent with the experimental data (up to f~2~= 0.5 when c = 2.5) obtained by other investigators \[[@B6],[@B7]\] regarding the velocity changes in suspensions for a wide range of fluids and particle sizes as well as particle compositions. Figure [2](#F2){ref-type="fig"} shows the relationship between relative sedimentation velocity and particle concentration. The relationship between relative velocity, viscosity and volume fraction is also consistent with experimental data \[[@B6],[@B7]\]. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### The dependence of a change in relative viscosity on the volume fraction of particles. ::: ![](1742-4682-2-13-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Dependence of relative sedimentation velocity on particle concentration (where βis a change in the relative velocity). ::: ![](1742-4682-2-13-2) ::: Determination of viscosity when yield velocity equals zero ---------------------------------------------------------- The value of viscosity derived in equation (18) describes the sedimentation of solid particles, that is particles that do not form rouleaux. I shall now determine the viscosity of blood when the yield velocity is zero. It is known \[[@B16]\] that if whole blood (in which coagulation is prevented) is placed in a vertically-positioned capillary tube, erythrocytes will aggregate into rouleaux and then sediment. Therefore the viscosity μ~1~must be determined in blood that has minimal numbers of rouleaux, and it is necessary to take into account the effect on rouleau sedimentation of erythrocytes that remain suspended. Such a condition occurs when the yield velocity is high (500 -- 1000 s^-1^) and the number of rouleaux is minimal. This condition can be expressed by equations (18) or (19) when f~1~→ 1and c = 2.5; that is rouleaux do not sediment in plasma but rather in a mixture of erythrocytes, plasma and a certain number of rouleaux. Calculations made according to equations (18) or (19) when f~1~→ 1and c = 2.5 yield the following results: μ~1~= 6.8 mNsm^-2^when concentration of erythrocytes is 28.7% μ~1~= 8.8 mNsm^-2^when concentration of erythrocytes is 48% μ~1~= 10 mNsm^-2^when concentration of erythrocytes is 58.9% These data are consistent with experimental data \[[@B16]\] when the yield velocity ranges from 500 to 1000 s^-1^. Thus, using the effect of the viscosity of the mixture from equations (18) and (19), I can calculate the viscosity of the blood at zero velocity by means of the following equation: ![](1742-4682-2-13-i27.gif) In this equation, when coefficient c = 2.5, there is a minimal number of rouleaux at μ~1~= 3 to 4 mNsm^-2^(the value of viscosity when the maximum yield velocity is more than 500 s^-1^). Figure [3](#F3){ref-type="fig"}, where the viscosity at zero yield velocity is plotted on the Y axis, shows that viscosity increases with increasing concentration. Thus an increase in erythrocyte concentration results in an increase of viscosity. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### The dependence of viscosity on yield velocity. ::: ![](1742-4682-2-13-3) ::: I shall now determine the shear stress at various concentrations and yield velocities. Table [1](#T1){ref-type="table"} shows that an increase of shear stress causes a decrease of viscosity. Thus, an increase in the concentration of erythrocytes will result in an increase of viscosity and a decrease in shear stress. It can be assumed that a maximal number of rouleaux is formed when the yield velocity is zero, since there are no forces that disassemble them. Then I can determine the number of rouleaux at different values of viscosity and shear stress. Table [2](#T2){ref-type="table"} shows these data and indicates that the main source of rouleaux is the erythrocytes themselves. The higher the erythrocyte concentration, the more rouleaux remain in the blood despite an increase in the forces that destroy them. It is also clear that an increase in shear stress results in a decrease of the number of rouleaux. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Relationship between shear stress and viscosity ::: Yield velocity (s^-1^) The volume fraction of the second phase Viscosity (mNsm^-2^) Shear stress (N/m^2^) ---------------------------- --------------------------------------------- -------------------------- --------------------------- 0.2 28.7 13 0.0026 35.9 30 0.006 48 63 0.0126 5 28.7 6 0.03 35.9 8 0.04 48 15 0.075 100 28.7 4 0.4 35.9 5 0.5 48 6 0.6 500 28.7 3 1.5 35.9 3 1.5 48 4 2 ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### The relationship between erythrocyte concentration and number of rouleaux ::: Yield velocity (s^-1^) Concentration % Viscosity (mNsm^-2^) Rouleaux concentration % Concentration of destroyed rouleaux % Shear stress (N/m^2^) ---------------------------- --------------------- -------------------------- ------------------------------ ------------------------------------------- --------------------------- 0.2 28.7 15 65.2 34.8 0.0026 35.9 30 81 19 0.006 48 63 83 17 0.0126 5 28.7 6 26 74 0.03 35.9 8 21.3 78.7 0.04 48 15 20 80 0.075 ::: I can now determine the concentration of rouleaux, assuming that viscosity is determined by the numbers of erythrocytes only at a high yield velocity (since high yield velocities destroy rouleaux). Granted this assumption, the viscosity is determined according to the Einstein equation (18) and (19). Viscosity at decreasing yield velocity is determined by both erythrocytes and newly-formed rouleaux. Then, according to equation (20), I obtain the result presented in Figure [4](#F4){ref-type="fig"}: the number of rouleaux decreases sharply with increasing yield velocity. Therefore, the number of rouleaux depends on the concentration of erythrocytes. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### The relationship between the volume fraction of rouleaux and yield velocity. ::: ![](1742-4682-2-13-4) ::: The quantity of rouleaux depends on shear stress (the higher the shear stress, the lower the rouleaux content of the blood) and erythrocyte concentration (the more erythrocytes, the more rouleaux will be formed). I can now determine whether all rouleaux are interconnected and what kind of cohesive forces operate among them. It is known that at low yield velocities, a greater fraction of the erythrocytes form rouleaux \[[@B16]\]. These long columns of erythrocytes have a certain stiffness and might interweave to form a single structure \[[@B16]\]. It is hypothesized that cohesive forces may vary among rouleaux. This phenomenon makes the properties of blood resemble those of a solid body. When the yield velocity increases, the length of the rouleaux gradually decreases and ultimately only stand-alone erythrocytes are left. To test this hypothesis, an experiment was conducted in which the breaking force and shear stress were those that naturally destroy rouleaux, but the cohesive forces were different. In an aerodynamic tube, a laminar boundary layer was created on a flat surface with the required shear stress on the surface of the wall \[[@B4]\]. On this surface, fine particles of equal diameter were placed (the cohesive force ranged from 0.0027 mN to 0.035 mN). From this information I could determine the destruction, i.e. the detachment and separation of particles from the surface. The results of the experiment are given in Table [3](#T3){ref-type="table"}. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### The relationship between shear stress, particle diameter and damage to the wall ::: Shear stress (N/m^2^) Diameter of particles (mm) Damage (g/s) --------------------------- -------------------------------- ------------------ 0.043 0.25--0.63 0.002 0.051 0.25--0.63 0.03 0.092 0.25--0.63 0.07 0.13 0.25--0.63 0.122 0.13 0.5--0.63 0.05 0.158 0.5--0.63 0.1 ::: Table [3](#T3){ref-type="table"} shows that destruction of rouleaux decreases with increasing particle diameter (which means increasing cohesive force). Conversely, the destruction of rouleaux increases with increasing shear stress. It can be supposed that an increase in shear stress destroys rouleaux that have a cohesive force lower than the breaking force. A further increase in shear stress will lead to the destruction of rouleaux with a greater cohesive force. Summary of results ------------------ ### The following have been determined 1\. The dependence of the viscosity of a mixture on volume fraction during sedimentation of a group of particles (forming no aggregates), confirmed by published experimental data \[[@B7]\] for volume fractions of the second phase (f~2~) up to 0.6. 2\. The dependence of viscosity of a mixture on the volume fraction of erythrocytes during sedimentation of rouleaux when the yield velocity is zero. 3\. Increase in the velocity of a mixture with an increasing concentration of erythrocytes when yield velocity is zero. 4\. An increased erythrocyte concentration results in an increase of viscosity of the mixture, and an increase in shear stress results in a decrease of viscosity of the mixture. 5\. The quantity of rouleaux depends on shear stress (the higher the shear stress, the fewer rouleaux in the blood) and erythrocyte concentration (the more erythrocytes, the more rouleaux are formed). 6\. With an increase in shear stress, those rouleaux are destroyed whose cohesive force is weaker than the breaking force. A further increase in shear stress will start to destroy rouleaux that have a greater cohesive force. Discussion ========== The role of the non-Newtonian viscosity of blood has remained a continuing challenge. Currently, the apparent viscosity at very low shear rates is considered as \"effectively infinite immediately before the substance yields and begins to flow\" \[[@B17]\]. Traditionally, Casson or Herschel-Bulkley models are used to measure both the yield stress of blood and shear thinning viscosity \[[@B18]\]. Human blood however does not comply with Casson\'s equation at a very low shear rate \[[@B13]\]. Other attempts to obtain finite viscosity values failed to take into account the hydrodynamic interactions between particles, or the complications related to aggregates \[[@B2]\]. Although an attempt to estimate blood viscosity at a very low shear rate has been made, no study has estimated the viscosity of blood when yield velocity equals zero. The mathematical model created in this study used the most fundamental differential equations that have ever been derived to estimate blood viscosity. Depending on erythrocyte concentration, this model estimates the blood viscosity at zero yield stress. It takes into account the following factors: (1) Erythrocytes sediment as a group and not as single particles; (2) Erythrocytes interact with each other; (3) Erythrocytes sediment as a rouleaux; (4) Such rouleaux sediment within an erythrocyte-containing medium. In general, abnormal values of blood viscosity can be observed in such pathologies as cancer \[[@B19],[@B20]\], peripheral vascular disease \[[@B19],[@B20]\] and acute myocardial infarction \[[@B19],[@B20]\]. Blood hyperviscosity may impair the circulation and cause ischemia and local necrosis through decreased capillary perfusion \[[@B21]\]. Blood hyperviscosity due to abnormal red cell aggregation has been found in patients with diabetes, hyperlipidemia and cancer \[[@B22]\]. Estimation of blood viscosity is, however, particularly important in trauma patients. It is known that blunt trauma to vascular walls may lead to conditions for boundary layer separation \[[@B4]\]. Physically, this can be explained as follows \[[@B12]\]: flow retarded at the surface has low kinetic energy and cannot enter the high pressure zone, therefore it separates from the vessel wall and moves into the inner flow. It should be noted that under normal physiological conditions, the boundary layer does not separate \[[@B16]\]. Shear stress in the zone of boundary layer separation is equal to zero \[[@B4]\]. Therefore, in accordance with the above, trauma may create transient conditions for the formation of rouleaux or for the interlacing of existing rouleaux that have formed in the flowing blood \[[@B16]\], since there is no breaking force at zero shear and yield velocity. A certain number of rouleaux can then enter the arterial branching zone, where the shear velocity and shear stress on the internal wall are low \[[@B16]\], and these rouleaux might attach to the vessel wall, potentially causing atheromatosis. Such arterial branching zones could also be injured by blunt forces, which will also lead to boundary layer separation \[[@B4]\]. Therefore, rouleaux will be formed with low shear velocity and low shear stress on the internal wall \[[@B16]\], also creating conditions for atheromatosis. Therefore, our understanding of the mechanism of blunt trauma to the vascular wall, which takes into account local hemodynamic and rheological factors, can be summarized in the following way. Trauma leads to the appearance of zones with high shear stress (as the result of injury to part of the vessel) and low or zero shear stress (within the zone of boundary layer separation) \[[@B4]\]. We have reported that high shear stress (exceeding the physiological value) may potentially damage the endothelium \[[@B4]\] and increase platelet aggregation \[[@B23],[@B24]\], possibly leading to thrombus formation. On the other hand, trauma may lead to boundary layer separation, resulting in the appearance of a zone with zero shear stress and zero yield velocity \[[@B4]\]. This may result, according to current research, in an increase of blood viscosity through increased erythrocyte aggregation and rouleaux formation. Such hyperviscosity has been reported in patients with traumatic crush syndrome and also has been studied in animals exposed to traumatic crush \[[@B25]\]. As noted above, hyperviscosity may worsen the blood circulation and cause ischemia and local necrosis through deterioration in capillary perfusion \[[@B21]\]. This work also establishes a quantitative relationship between the extent of rouleaux formation and shear stress. According to current results, the number of rouleaux increases with decreasing shear stress, and this trend becomes more pronounced as the shear stress approaches zero. Rouleaux continue to form inside what I call the \"hemodynamic shade\". This \"hemodynamic shade\" creates a stagnant zone that can be characterized by a secondary flow and a boundary. Hemodynamic stress outside this zone, however, is still significant enough to destroy and entrain rouleaux. The \"hemodynamic shade\" zone can also be characterized by a significant deterioration of mass exchange due to the attachment of rouleaux to the vessel wall. This may decrease the permeability of the endothelium \[[@B16]\] and decrease the rate of removal of lipids and lipoproteins, which in turn can lead to the formation of lipid stripes directed along the blood flow and located in the \"hemodynamic shade\" of the original attached rouleaux. The escalating formation of rouleaux continues within the entire \"hemodymanic shade\" zone. The model of traumatic damage to the vessel that takes into account local rheological and hemodynamic factors could be applied to many internal injuries involving an elastic vessel wall and a blunt traumatic mechanism. One example is traumatic myocardial infarction, which can result from blunt trauma to the coronary vessels. It should be noted that patients with blunt trauma may develop acute myocardial infarction; such patients may benefit from screening procedures such as electrocardiography, which might improve their chances of survival \[[@B8],[@B26]-[@B49]\]. In a large cross-sectional observational study, abdominal, pelvic and blunt cardiac injuries were found to be significantly associated with acute myocardial infarction even after controlling for confounders such as mechanism and severity of injury, age, sex, race, source of payment, alcohol and cocaine use \[[@B50]\]. Intracoronary thrombosis has been suggested as one of the mechanisms of acute myocardial infarction in young people due to trauma, since other \"atherosclerotic\" mechanisms do not apply \[[@B38],[@B42]\]. Nonetheless, the exact mechanism of traumatic myocardial infarction remains unclear. Current research suggests that blunt trauma may result in the appearance of a region of very low or zero shear stress, where hyperviscosity and increased rouleaux formation are likely to appear. Large quantities of rouleaux may be transported in the bloodstream toward the more distal parts of the coronary vessels, causing their occlusion. Caimi et al. \[[@B51]\], for instance, observed that blood viscosity at low shear rate is the only hemorheological factor that significantly increases the risk of acute myocardial infarction in young people. On the other hand, blunt trauma may result in traumatic compression of the vessel wall with high shear stress \[[@B4]\]. Increased shear stress itself may cause rupture of a coronary atherosclerotic plaque \[[@B52]\]. In addition, high shear stress may result in increased platelet aggregation \[[@B23],[@B24]\], often leading to thrombus formation. In summary, there is still a gap in our understanding of all quantitative aspects of the extreme values of viscosity at low and zero shear rates \[[@B3]\]. To the best of my knowledge, the work described in this paper represents one of the few attempts to estimate extreme values of viscosity at low shear rate. An understanding of the precise mechanisms that affect blood viscosity would be of clinical significance. Acknowledgements ================ The author gratefully acknowledges the contribution of Prof. Paul Agutter for his valuable comments.	PubMed Central	2024-06-05T03:55:55.610778	2005-3-30	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1079952/", "journal": "Theor Biol Med Model. 2005 Mar 30; 2:13", "authors": [ { "first": "Rovshan M", "last": "Ismailov" } ] }
PMC1079953	Background ========== Buffering: a paradigm with growing pains ---------------------------------------- Buffering is among the most important mechanisms that help to maintain homeostasis of various physiological parameters in living organisms. This article is concerned with the definition of an appropriate scientific unit, or scale, for the quantitation of buffering action -- a quantity that has been termed \"buffering strength\", \"buffering power\", \"buffer value\", or similarly \[[@B1],[@B2]\]. On the one hand, the concept of \"buffering\" is applied in a growing number of scientific and engineering disciplines. On the other hand, the units that are currently used to measure buffering -- often created on an ad hocbasis -- suffer from fundamental inconsistencies and shortcomings. Comparison with \"mature\" and standardized scientific units, e.g. those of the \"Système International des Unités\" (\"SI\"), highlights the extent of these shortcomings (see below). As a consequence, there are multiple \"local\" theoretical buffering concepts with limited power, and the practical treatment of buffering phenomena is complicated unnecessarily. Thus, rethinking the quantitation of buffering action is not an effort to reinvent the wheel; rather it seems that \"the wheel\" has not been invented yet. Our thesis is that buffering action can be quantitated in a better, simpler, and universal way when buffering is conceived as a purely formal, mathematical principle. In this article, we present such a formal concept of buffering. Compared to existing buffering concepts, its major achievements are formal rigor and scientific richness. ### \"Buffering\" -- a paradigm useful in many fields A look at the current usage of the term \"buffer\" suggests that a corresponding fundamental principle is common to a great variety of disciplines. Buffering concept and terminology originated in acid-base physiology at the end of the 19^th^century when it had become clear that several biological fluids \"undergo much less change in their reaction after addition of acid or alkali than would ordinary salt solutions or pure water\" \[[@B2]\]. Hubert and Fernbach had introduced the term \"buffer\"; Koppel and Spiro suggested the terms \"moderation\" and \"moderators\" instead \[[@B2]\]. The concept of buffering was soon adopted in an increasing number of different contexts, including buffering of other electrolytes (e.g. Ca^++^and Mg^++^), of non-electrolytes, of redox potential, and numerous other quantities inside and outside the realm of chemistry. Examples are presented in Additional file [1](#S1){ref-type="supplementary-material"}. ### Expressing the magnitude of buffering action is problematic The common idea behind these diverse phenomena is that \"buffering\" is present when a certain parameter changes less than expected in response to a given disturbance, i.e., the buffer absorbs or diverts a certain fraction of the disturbance. Very soon after the concept of \"buffering\" had emerged it became apparent that buffering is not just absent or present in a binary sense, but instead may be \"strong\" or \"weak\". In fact, this \"buffering strength\" could differ over a wide range. Moreover, chemists, physiologists, and clinicians realized the great practical importance of this quantitative aspect of buffering \[[@B3]\], and struggled to get a numerical grip on it with the aid of various units or scales. Researchers in other areas followed. By now, buffering strength units are available for some, but not all buffering phenomena. In some cases, e.g. the buffering of ions in aqueous solutions, there exist even multiple units that are used in parallel (Additional file [2](#S2){ref-type="supplementary-material"}, Table [1](#T1){ref-type="table"}). One can thus certainly manage to \"put numbers\" on these buffering phenomena. For many other types of buffering, however, units do not exist at all. For instance, no such scales are available for \"blood pressure buffering\" and for \"cognitive buffering\". Without a buffering strength unit, however, it is obviously difficult to formulate and test quantitative hypotheses regarding buffering phenomena. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Interconversions of units for H^+^buffering strength. ::: *Parameter* *Definition* *f(B)* *f(β)* *f(β~c~) ----------------- ------------------ ---------------------- ------------ ---------------------------- B* (B) β~c~- 1 β~H+~ (B+1) × 2.3 × 10^pH^ (β) β~c~× 2.3 × \[H^+^\]~free~ β~c~ B + 1 (β~c~) B: \"buffering odds\" according to this article, : \"buffering value\" according to Van Slyke \[4\]; : \"buffering coefficient\" according to Saleh et al. \[6\]. First row: given source units; first column: desired target units; intersection of particular row and particular column: transformation, i.e., functions of the respective source unit, that yields the target unit. ::: In the past, researchers have exhibited a surprisingly high degree of tolerance towards the shortcomings and ambiguities inherent to the current approaches to the quantitation of buffering action. However, these drawbacks have already caused problems and confusion, both on a theoretical and practical level, and will become even more problematic and disturbing as the buffering paradigm is applied more widely. The systematic analysis of the available concepts and scales of \"buffering\" presented in Additional file [2](#S2){ref-type="supplementary-material"} substantiates this criticism and points to the features that would be required to make an ideal scale of buffering strength. Briefly, this analysis of the available units of \"buffering strength\" reveals three major problems: i) Intrinsic deficiencies: Scales are second-rate inasmuch as only some of the mathematical operations can be applied to the measurements that would be applicable with different types of scales; ii) Limitedness, both conceptual and practical: Individual units can handle only selected special cases of buffering, whereas other types of buffering require different units or cannot be quantitated at all; iii) Confusion & inconsistencies: A motley multiplicity of units and definitions actually houses disparate things, thus obfuscating the simple, common principle behind the various buffering phenomena. Accordingly, a quantitative measure of buffering would ideally provide i) a scale of the highest possible type, namely a \"ratio scale\". Ratio scales are scales with equal intervals and an absolute zero. For instance, when H^+^ion concentration is expressed in terms of moles per liter, this measure increases by the same amount irrespective of the initial concentration (equal intervals). In contrast, when H^+^ion concentration is expressed, for instance, in terms of pH, this measure of concentration will change only a little at low pH, but much at high pH (non-equal intervals). One example for a scale without an absolute zero, on the other hand, is provided by the Celsius and Fahrenheit scales for temperature where the position of 0° is arbitrary, whereas 0° on the Kelvin scale is an \"absolute\" zero (as would be a probability of zero, a capacitance of 0 Farad, a mass of 0 kg etc.); ii) a scale that is universal, allowing for adequate quantitation of buffering behavior in all its manifestations (i.e., irrespective of its particular physical dimension, and including moderation, amplification, and the complete absence of buffering); iii) a scale that could be used as a general standard, within a given discipline and across different disciplines. The first two properties mentioned (ratio scale and universal applicability) would automatically generate a scale that could serve as such an all-purpose yardstick of buffering strength. However, there is clearly no such scale available to date. A formal and general approach to the quantitation of buffering action ===================================================================== An intuitive introduction of the approach ----------------------------------------- ### Buffering processes as partitioning processes Universal measures of buffering action can be developed if one views the underlying process as a \"partitioning\" process. To explain what we mean by this, consider two arbitrarily shaped vessels that are filled with a fluid and connected via a small tube (Figure [1A](#F1){ref-type="fig"}). The fluid in such a system of communicating vessels will distribute in such a way that the two individual fluid levels become equal. By virtue of hydrostatic pressure, any given total fluid volume is thus associated with a unique partial volume in the first vessel, and with another unique partial volume in the second one. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### *[A simple quantitative criterion of buffering action.]{.underline}(See main text for detailed explanation) A, Communicating vessels model of partitioning processes.In a system of two communicating vessels (A and B), total fluid volume is the sum of the two partial volumes in A and B. In an equilibrated system, the partial volumes in the individual vessels can be described as functions of total fluid volume; these functions are termed \"partitioning functions\". The derivatives of the partitioning functions tell what fraction of a total volume change is conveyed to the respective vessel. B, Partitioning of a quantity in a two-compartment system.A given total change of quantity in the system produces two partial changes in compartments A and B. The greater the partial change in B, the smaller the change in A, and the greater the \"buffering\" of the quantity in A. C, Partitioning of H^+^ions between water and buffer.Free H^+^ions are added to an aqueous solution containing a weak acid (e.g. as strong acid). Some of the added H^+^ions remains free, some become bound to buffer molecules. C, General definition of measures of buffering action.The differential dz/dy, paraphrased as d(buffered)/d(total), is termed the buffering coefficient b. The differential, paraphrased as d(buffered)/d(unbuffered), is termed the buffering ratio B. ::: ![](1742-4682-2-8-1) ::: Now, let us add a small volume of extra fluid into the system. When the system has reached the corresponding new equilibrium state, a portion of the extra fluid is found in the vessel A, another portion in vessel B. Clearly, the volume change in vessel A in response to a given volume load is smaller when this vessel is part of this system of vessels, as compared to vessel A standing alone and subjected to the same load. We can say, the system is able to stabilize or \"buffer\" fluid volume in vessel A in the face of increases or decreases of total volume. This example shows that buffering can be viewed in terms of a partioning process in a system of two complementary compartments. \"Fluid volumes\" are readily replaced by other physical, chemical or other quantities. For instance, the classic case of H^+^buffering can be represented in a straightforward way as the partitioning of H^+^ions into the pool of \"free\" H^+^ions (i.e., H^+^ions bound to water, corresponding to vessel A) and the complementary pool of \"bound\" H^+^ions (i.e., H^+^ions bound to buffer molecules, corresponding to vessel B). #### A simple criterion of buffering strength We now formulate a simple quantitative criterion of buffering action, first in terms of fluid volumes: The more of a given fluid volume added to the system of communicating vessels ends up in vessel B, the greater the stabilization or \"buffering\" of the fluid volume in vessel A. Or in acid-base terms: The more of a given amount of H^+^ions (added, for instance, in the form of strong acid) becomes bound by buffer molecules, the more the concentration of \"free\" H^+^ions is stabilized or buffered. We can easily formulate that criterion in a general form, free of reference to any particular quantity: The greater the change of a given quantity in one individual compartment, the greater the buffering of that quantity in the other compartment (Figure [1B](#F1){ref-type="fig"}). Herein, the magnitude of the change in a compartment may be expressed either relative to the total change, or relative to the complementary change in the other compartment. The example of communication vessels also shows that the magnitude of change (when expressed in either of these ways) is not affected by the direction of the change: it remains the same whether the quantity in question is added to the system, or whether it is subtracted. Unspectacular and intuitive as it may appear, this criterion will lead to conclusions that differ considerably from established views. For instance, it is usually held (on the basis of Van Slyke\'s definition of buffering strength \[[@B4]\]) that a weak acid buffers H^+^ions most strongly when H^+^ion concentration is equal to the acid constant K~A~(i.e., when \[H^+^\] = K~A~). However, this is not where the fraction of added H^+^ions binding to buffers is greatest. Rather, this fraction reaches a maximum when \[H^+^\] approaches zero (Figure [1C](#F1){ref-type="fig"}). According to our simple criterion, that is the point of maximum buffering strength (i.e., when \[H^+^\] = 0). Similarly, when H^+^ions are removed from such a solution (e.g. by addition of strong base), the fraction supplied via deprotonation of buffer molecules (as opposed to a decrease of free \[H^+^\]) is greatest at low total \[H^+^\]. This classic case illustrates the impact of the various buffering strength units on our perception of buffering strength, and is analyzed in detail, together with several further examples, in the accompanying paper (Buffering II[@B10]). Our concept of buffering results, ultimately, from the systematic application of this simple criterion. #### Deriving quantitative measures of buffering strength from this criterion With our simple criterion at hand, all that is left to do in order to quantitate buffering action is to put numbers on the magnitude of the change in the compartment that buffers or stabilizes the other compartment (termed \"buffering compartment\", corresponding to vessel B in Figure [1A](#F1){ref-type="fig"}). This can be done in two equally useful ways (Figure [1D](#F1){ref-type="fig"}): Firstly, change in the \"buffering compartment\" can be expressed with respect to the total change in the system. The resulting measure represents a \"fractional change\", here termed \"buffering coefficient b\" ![](1742-4682-2-8-i8.gif) The buffering coefficient b thus indicates the proportion between one particular part and the whole. Secondly, change in the \"buffering compartment\" can be expressed with respect to the complementary change in the other compartment, termed \"target compartment\" or \"transfer compartment\", to indicate that one views this compartment as the one for which the imposed change is \"intended\" (corresponding to vessel A in Figure [1A](#F1){ref-type="fig"}). We thus obtain a second measure, here termed the \"buffering ratio B\": ![](1742-4682-2-8-i9.gif) The buffering ratio B thus indicates the proportion between the two parts of a whole. This measure is completely analogous to the \"odds\" as used for the quantitation of chance (mainly by epidemiologists) and may therefore be termed synonymously \"buffering odds B\". In the following section, we illustrate a few characteristic types of buffering, using again fluid-filled communicating vessels as an example (Figure [2](#F2){ref-type="fig"}). ::: {#F2 .fig} Figure 2 ::: {.caption} ###### [Communicating vessels as a physical model for a buffered system.]{.underline}* Total fluid volume is taken as x, fluid volume inside vessel A (\"transfer vessel\", red) as the value of the transfer function τ(x), and aggregate fluid volumes in the other vessels (\"buffering vessels\", blue) represent the \"buffering function\" β(x). We can describe these systems in terms of our two measures of buffering action, namely the buffering coefficient b(x) = β\'(x)/\[τ\'(x) + β\'(x)\] and the buffering ratio B(x) = β\'(x)/τ\'(x) (see main text for detailed explanation). *A, Linear buffering, one buffering vessel.The volume changes in A are only half as big as the total volume changes in the system; the volume inside A is \"buffered\", or, more specifically, \"moderated\". The degree of moderation is the same at all fluid levels; b(x) = constant = 0.5 and B(x) = constant = 1. B, Zero buffering, or perfect transfer.Changing total volume in the system translates completely into identical volume changes in vessel A, without \"moderation\" or \"amplification\": b(x) = 0 B(x) = 0. C, Linear buffering, several buffering vessels.Increasing the number of buffering vessels increases buffering action. The four partitioning functions are replaced by a single buffering function β. Buffering parameters are b(x) = 0.8 and B(x) = 4. D, Linear buffering, general case.Same buffering behavior as in C, brought about by a single buffering vessel. E, Non-linear buffering, one buffering vessel.In this system, the individual volume changes are not linear functions of total volume. Consequently, the proportion between volume flow into or out of vessels A is not a constant, but a variable function of the system\'s filling state. F, Non-linear buffering, several buffering vessels.In most buffered systems, buffering is brought about by a multiplicity of buffers (as in C) that are non-linear in their individual ways (as in E). Buffering coefficient and buffering odds provide overall measures of buffering action that neither require nor deliver any knowledge about the individual components. ::: ![](1742-4682-2-8-2) ::: ### Use of buffering coefficient and buffering ratio for the quantitation of buffering action -- some typical examples #### A simple buffered system Consider a system of two communicating vessels, both having identical dimensions and constant cross sectional areas (Figure [2A](#F2){ref-type="fig"}, left panel). We consider vessel A our compartment of interest (i.e., the \"target\" or \"transfer compartment\"), and ask how much the fluid volume inside it is stabilized or \"buffered\". To determine the degree of buffering, we titrate the system up and down by adding or removing fluid. We find that the volume changes in A are always only half as big as the changes of total volume in the system; the volume inside A is \"buffered\". The behavior of the system is repesented graphically on the right hand of Figure [2A](#F2){ref-type="fig"}. Total volume is plotted on the abscissa. The individual volumes in vessels A and B at a given total volume are indicated in this \"area plot\" by the respective heights of the two superimposed areas at that point. Volumes inside vessel A and B are thus expressed as functions of the independent variable \"total volume\". We denote that variable by the letter x. Moreover, the volume in the transfer vessel A expressed as a function of total volume is termed the \"target function\" or \"transfer function\", denoted τ(x), and the volume in the buffering vessel B expressed as a function of total volume is termed the \"buffering function\", denoted β(x). \"Change\" in a compartment then can be defined more specifically as the first derivative of the particular function with respect to the independent variable, notated briefly as τ\'(x) or β\'(x). The buffering coefficient b, defined above as the ratio of \"volume change in vessel B\"over \"total volume change in the system\", can then be expressed more simply and generally as b = β\'(x)/\[τ\'(x) + β\'(x)\]. In this system, total change equals the sum of the individual changes (other systems are covered below), and thus τ\'(x) + β\'(x) = 1, and hence β\'(x)/\[τ\'(x) + β\'(x)\] = β\'(x)/1 = β\'(x). Because the buffering function β(x) equals 0.5·x in this system, we obtain a dimensionless buffering coefficient of b = 0.5. In words, a buffering coefficient of 0.5 says that of the total change imparted to the system, a fraction of 0.5 (or 50%) is directed to the \"buffering compartment\". The buffering ratio B, on the other hand, which was defined above as the ratio of \"volume change in vessel B\" over \"volume change in vessel A\", can then be expressed as B = β\'(x)/τ\'(x). With τ(x) = β(x) = 0.5·x in this system, we find a value of B = 1. In words, a buffering ratio of 1 says that when a certain change is imposed to the system, the change in the target compartment is always associated with a similar sized change in the buffering compartment. In terms of fluid volume: for every drop going into or out of vessel A, another drop goes into or out of vessel B. #### An unbuffered system Figure [2B](#F2){ref-type="fig"} shows a system without a \"buffering vessel\". Accordingly, changes in total volume are completely translated into exactly equal changes of volume in vessel A. Again, the point here is how to express this type of buffering behavior numerically. Change in the transfer vessel A is given by a transfer function τ(x) = x, and change in the buffering vessel, given its non-existence or zero volume, by a buffering function that has a constant value of zero: β(x) = 0. We compute the buffering coefficient b again as b = β\'(x) and find that b = 0, and compute the buffering ratio B as B=β\'(x)/τ\'(x) and find that B = 0. We see that both measures yield scales with an \"absolute zero\", i.e., where the position of \"zero\" does not depend on some arbitrary external reference (as would be the case with electrical or thermodynamical potentials, for instance) or on some similarly arbitrary convention (such as for the Celsius scale for temperature), but follows inescapably from the definition of the unit. Again, it may appear trivial to find zero values for buffering strength in the absence of buffering. However, this desirable property of a buffering strength unit is not the rule, including the widely used H^+^buffering strength unit introduced by Van Slyke. This unit, defined as β = d(Strong Base)/dpH, will always be greater than zero even in the complete absence of buffering; even stranger, the particular numerical value representing the absence of buffering will vary with pH (see detailed discussion in Buffering II[@B10]). #### Multiple buffering vessels vs. an equivalent single one Next, as shown in Figure [2C](#F2){ref-type="fig"}, we add several additional copies of similar buffering vessels (vessels B,C,D,E). Compared to a single buffering vessel B, this alteration results, of course, in increased buffering action. When one compares the initial situation with a single buffering vessel to the system comprising four such vessels, it is reasonable to say that buffering action increases four-fold. However, we are not yet in a position to compute the buffering coefficient of buffering ratio. In principle, the volumes in these vessels can be expressed by several individual functions which may be termed \"partitioning functions\". However, what matters with respect to the stabilization or buffering of the volume in vessel A is only their aggregate volume as a function of total volume. This aggregate function, i.e., the four partitioning functions lumped together into a single function, represents our \"buffering function β(x)\". With respect to buffering, the system in Figure [2C](#F2){ref-type="fig"} is thus perfectly equivalent to the system in Figure [2D](#F2){ref-type="fig"}. In both systems, the buffering function has the value of β(x) = 0.8·x, and we thus find a buffering coefficient of b = 0.8, and a buffering ratio of B = 4. Indeed, the buffering ratio increases accordingly from B = 1 to B = 4. This behavior is typical for a \"ratio scale\", and is a desired property. Ratio scales not only represent the phenomena under study in a particularly intuitive way, they are also the highest type of scale inasmuch they allow meaningful application of the widest range of mathematical operations, including averaging, expression as percentage, and comparison in terms of ratios. In contrast, the buffering coefficient changed from 0.5 to 0.8. Evidently, the buffering coefficient does not yield a ratio scale: the four-fold increase in the number of buffering vessels is reflected in an only 1.6-fold increase of the buffering coefficient. Another four-fold increase from 4 to 16 buffering vessels would entail an even smaller increase of the buffering coefficient, from 0.8 to 0.94, an approximately 1.2-fold increase. #### Systems exhibiting non-constant buffering In the system depicted in Figure [2E](#F2){ref-type="fig"}, the cross-sectional area of the buffering vessel is not constant, but varies with fluid level. As a consequence, the individual volumes in vessels A and B changes are not linear functions of total volume of the type y = constant·x, but may be any arbitrary non-linear function. The proportion between the two individual changes in vessels A and B is therefore not constant, but varies depending on the system\'s filling state. The two measures of buffering action can be computed exactly as indicated above as β\'(x) and β\'(x)/τ\'(x), respectively, but the results are valid only for the given value of x. Consequently, buffering coefficient and buffering ratio must be presented as b(x) and B(x), respectively, where x specifies the filling state of the system. Such variable buffering is found in most buffered systems of scientific interest, including buffering of H^+^and Ca^++^ions in plasma and cytosol. #### Non-constant buffering with multiple irregular buffering vessels Figure [2F](#F2){ref-type="fig"} carries this more realistic version one step further, inasmuch as buffering is also often brought about by several different buffers each of which may be non-linear in its own way. This situation is replicated by a combination of several, irregularly shaped buffering vessels. A buffering function β(x) is again obtained by lumping together the individual partitioning functions of the buffering vessels into a single aggregate buffering function. Buffering coefficient and buffering ratio are then computed in the known way for a given value of x. Buffering coefficient and buffering ratio provide overall measures of buffering action that neither require nor deliver any knowledge about the individual components, and many different combinations of buffering vessels can bring about identical buffering behavior. A formal and general definition of the approach ----------------------------------------------- ### Systems of functions as representations of buffering phenomena The above examples of systems of communicating vessels (Figure [2](#F2){ref-type="fig"}) are useful to become familiar with our approach to the quantitation of buffering action. Indeed, this approach is essentially simple, and the principles illustrated by fluid partitioning between two vessels can be applied immediately to other quantities that distribute between two complementary compartments, for instance to the classical case of H^+^or Ca^++^ions in their complementary pools of \"bound\" and \"free\" ions (Buffering II[@B10]). On the other hand, these examples can illustrate only a fraction of the things one can do in principle with this formal approach to the quantitation of buffering action. This approach has the potential to provide a common language for all types of buffering phenomena, not just for the few cases mentioned. The universal nature of these measures of buffering action, and their various uses can be appreciated and exploited best when the concept is presented in a pure mathematical form. Herein, our buffering concept resembles other formal frameworks such as probability theory or control theory which are, at the core, of purely mathematical nature; specific examples (e.g. flipping coins or control circuit diagrams, respectively) may illustrate these concepts, but cannnot capture them comprehensively and systematically. Emphasizing those aspects that help to use this approach as a \"mathematical tool\", the following paragraphs provide such a systematic framework for the quantitation of buffering action. Herein, combinations of communicating vessels (each with its individual fluid volume depending on the common variable \"total fluid volume\") are replaced by combinations of purely mathematical functions of a common variable. We need the concepts of \"partitioned\", \"two-partitioned\" and \"buffered systems\", of the \"sigma function\" and the distinction between \"conservative\" and \"non-conservative\" partitioned systems, between \"moderation\" and \"amplification\", between \"inverting\" and \"non-inverting\" buffering, and between \"buffering power\" and \"buffering capacity\". All the definitions and concepts set up here will be applied to specific buffering phenomena in the accompanying article (Buffering II[@B10]). Some interesting theoretical aspects are presented in the Additional files. They touch on the question \"What is buffering?\" (as opposed to the question \"How can we quantitate buffering?\"). It will be shown that the definition of \"buffering\" can be reduced to a set of axioms in almost exactly the same way as the concept of \"probability\", and therefore an answer to this question is to be sought on the same spot and with the same mathematical and philosophical approaches. #### Two-partitioned systems In a system of two communicating vessels, the individual fluid volume in one vessel could be described as a function of total fluid volume, and the volume in the other vessel by another function of the same total fluid volume. We are thus dealing with two functions of a single common independent variable. More precisely, with an \"unordered pair\" or a \"combination\" of functions, inasmuch as the two functions are not in a particular order. A combination of two functions of a common independent variable is termed a \"two-partitioned system\", or ^2^P in brief. Its two functions are termed \"partitioning functions\" and denoted π~1~and π~2~. A two-partitioned system can thus be written ^2^P = {π~1~(x), π~2~(x)}, if we let x represent the independent variable. In the following, both functions are assumed to be continuous and differentiable, and x, π~1~(x) and π~2~(x) are all real valued. Importantly, in order to use the buffering paradigm in a meaningful and correct way, a two-partitioned system is a necessary and sufficient condition. As a consequence, one can apply the buffering paradigm outside pure mathematics to \"real world\"-phenomena provided these phenomena are represented mathematically by such a combination of functions. #### Conservative partitioned systems, and the \"sigma function\" The examples above obeyed a conservation law, due to physical or chemical constraints: Fluid distributed into various compartments, but its total volume was constant; H^+^ions added into a solution were bound by buffers or by water, but their total number did not change. More generally, if the quantity in question is neither created or destroyed in the process, the total change imposed onto the system equals the sum of the two partial changes. Analogously, in terms of functions, we use the term \"conservative partitioned system\" to designate a system of partitioning functions whose sum equals the value of the independent variable. That condition, termed \"conservation condition\", can be written as: \[π~1~(x) + π~2~(x)\...+ π~n~(x) \] = ![](1742-4682-2-8-i10.gif) = x. The \"sum\" of the individual functions, given by the expression ![](1742-4682-2-8-i10.gif), can be used to define a function σ (termed \"sigma function\") that lumps together all partitioning functions π~i~of a n-partitioned system: σ: x → ![](1742-4682-2-8-i10.gif). Using this sigma function, we can rewrite the \"conservation condition\" briefly as σ(x) = x. Many important phenomena can be represented and analyzed in terms of a conservative partitioned system. Nonetheless, conservation (in this mathematical sense) is an accidental, not a general feature of partitioned systems. #### Non-conservative partitioned systems We thus drop the conservation condition σ(x) = x, and allow σ to be a continuous function of any type. This generalization will turn out to be very useful (Buffering II[@B10]). On the one hand, it allows one to express conservative systems in alternative, \"parametric\" form. As an example, when one describes bound and free H^+^ions (expressed in terms of \"moles\") as a function of total H^+^ions (added for instance as strong acid), one may readily measure strong acid in terms of \"grams\" or \"milliliters\", instead of \"moles\". Then, the aggregate \"output\" does not equal the \"input\", or σ(x) ≠ x; this inequality characterizes the system as \"non-conservative\". More importantly, the concept of non-conservative systems allows us to deal with functional relationships between completely heterogeneous physical quantities, and to apply the buffering concept to this class of phenomena. Examples include the buffering of organ perfusion in the face of variable perfusion pressure, or systems level buffering (Buffering II[@B10]). Partitioning functions and sigma function can be represented graphically in various ways (Figure [3](#F3){ref-type="fig"}), e.g. as a family of curves or by an area plot. Moreover, partitioned systems with two partitions π~1~and π~2~can be represented by a three-dimensional space curve ![](1742-4682-2-8-i11.gif). For instance, the buffering of H^+^ions in pure water or by weak acids is represented as space curve in the accompanying article (Buffering II[@B10]). ::: {#F3 .fig} Figure 3 ::: {.caption} ###### [Graphical representation of two-partitioned systems of functions.]{.underline}* The unordered combination of two functions π~1~(x), π~2~(x) of a single independent variable x is termed a \"two-partitioned system of functions\". The two functions may represent the two complementary parts of a whole, e.g. \"bound H^+^ions\" vs. \"free H^+^ions\" in an aqueous solution. The sum of the two functions is termed \"sigma function\" σ(x) (see main text for detailed explanation).. *A, Family of curves. The individual functions π~1~(x), π~2~(x), and σ(x) may be plotted individually as a family of curves (this is possible for multi-partitioned systems as well). B & C, Area plots.The individual partitioning functions of partitioned systems can be plotted \"on top of each other\" such that the value of each function is represented by the vertical distance between consecutive curves. In a partitioned system, their order is not constrained, and thus two equally valid representations exist for a two-partitioned system (B,C). A limitation of area plots is that they do not allow visualization of negative-valued partitioning functions. D, Three-Dimensional Space Curve.The independent variable x and the values of the partitioning functions π~1~(x), π~2~(x) of a two-partitioned system may be interpreted as x-, y- and z-coordinates, respectively. This results in a three-dimensional space curve. Such a curve can display both positive and negative values. Again, there are two different, equally valid representations. Projections of that curve on the xy-plane (red) and xz-plane (blue) correspond to the individual partitioning functions π~1~(x) and π~2~(x). Projection of the space curve on the yz-plane (gray) corresponds to a plot of the composite relations π~1~(π~2~(x)) or π~2~(π~1~(x)); these projections are not necessarily single-valued functions. The projection on the yz-plane is suited particularly well to assess the proportion between the individual rates of change of the two functions. Importantly, these proportions provide the clue to the quantitation of \"buffering action\". ::: ![](1742-4682-2-8-3) ::: #### Buffered systems In order to talk about buffering with respect to two communicating vessels, it is necessary to decide which vessel would be considered the buffer of the other one. With respect to H^+^ions, this assignment is conventionally made in such a way that \"free H^+^ion concentration\" is said to be buffered, and \"bound H^+^ion concentration\" that which brings about buffering. More generally, the two partitioning functions in a two-partitioned system must be assigned two different, complementary roles. Which is which must be indicated explicitly; here, this shall be done via the particular order: The first partitioning function is taken as description of the quantity that is being buffered, and termed \"target\" or \"transfer function\". For clarity, we denote the transfer function by τ(x). The second function is taken as to describe the quantity that brings about buffering, and is termed the \"buffering function\" β(x). Obviously, two partitioning functions π~1~(x) and π~2~(x) can be arranged in two different ways, with the resulting \"ordered combinations\" (or \"variations\") written here {π~1~(x), π~2~(x)} and {π~2~(x), π~1~(x)}. An ordered pair of functions is called a \"buffered system\". Briefly, a buffered system B can be written B = {τ(x), β(x)}. ### Quantitative parameters to describe the behavior of buffered systems For every x in an ordered combination of two differentiatable functions τ and β, there are two derivatives τ\'(x) and β\'(x). The proportions between the two derivatives (i.e., \"rates of change\") serve to quantitate \"transfer\" (to the \"target compartment\") and its complement, \"buffering\", according to our simple criterion defined above. In general, there are four ways to express the proportions between two parts of a whole (Figure [4](#F4){ref-type="fig"}). Accordingly, there are four quantitative measures of buffering or transfer in a \"buffered system\". Herein, we also employ the equivalences y↔τ(x) and z↔τ(x) to facilitate geometrical interpretation in terms of partial derivatives of a space curve (Figure [3D](#F3){ref-type="fig"}). ::: {#F4 .fig} Figure 4 ::: {.caption} ###### [Proportions in Two-Partitioned Systems.]{.underline}The buffering measures are dimensionless proportions between two parts of a whole, or between one particular part and the whole. A, Bisection of a straight line.An oriented line ![](1742-4682-2-8-i12.gif) of length σ (row 1) can be divided in several ways into two parts of lengths τ (red) and β (blue), respectively (rows 2--5). Dividing the line at a point that is lying on the line itself gives rise to an \"inner division\" (rows 2 and 3), whereas dividing the line outside the interval ![](1742-4682-2-8-i12.gif) yields an \"outer divison\" (rows 4 and 5). Proportions between the two parts can be expressed in various ways (Figure 4D). For inner divisions, proportions are positive-valued. For outer divisions, \"negative proportions\" and fractional lengths greater than 1 or smaller than 0 are obtained. B, Bisection of a functionThe principle of dividing a quantity into two is also applicable to the values of a function of x at a given value of x. Thus, there are multiple ways to split an entire function σ into two functions τ and β such that the sum of their values τ(x) and β(x) equals the value σ(x) for every x. C, Bisection of a slope, or rate of change.The quantity to be bisected may as well be the slope σ\' of a function σ. Again, there are multiple ways to split a function σ into two functions τ and β such that the sum of their first derivatives τ\'(x) and β\'(x) equals σ\'(x) for every x. For functions and derivatives of functions alike, the proportions between the two parts into which they were split can be expressed by the four measures indicated in Figure 4D. \"Buffering\" relates to the proportion between such partial rates of change of two complementary processes. D, Measures of proportionality between the parts of a bisected slope.Proportions among two partial rates of change τ\' and β\' that result from bisection of a whole rate σ\' can be expressed either as fractions of a part with respect to the whole (τ\'/σ\' and β\'/σ\') or as as ratio of one part over the other (τ\'/β\' and β\'/τ\'). The following terminology is suggested: t, \"transfer coefficient\"; b, \"buffering coefficient\"; T, \"transfer ratio\"; B, \"buffering ratio\". These parameters serve to quantitate buffering action. E, Relation between the four measures of proportion.Any single one of the four measures (t,b,T,B) fully determines the other three. The plot shows b, T, and B as functions of t. F, Trigonometric measure of proportionality between parts of a bisected slope.The two partial rates of change τ\'(x) and β\'(x) may be interpreted as two perpendicular vectors. Their resultant τ\'(x) + β\'(x) encloses an \"buffering angle α\" with τ\'(x). The buffering angle and the four buffering parameters (t, b, T, B) are related by four bijections with the buffering angle ![](1742-4682-2-8-i13.gif). A buffering angle α = 0 is equivalent to zero buffering, a buffering angle of 90° to perfect buffering. This representation of buffering behavior does not have discontinuities at \"infinite\" transfer or buffering odds, and is able to reflect the full range of buffering withing half a unit circle (-45° to 135°). (See Supplement 5 for further details) ::: ![](1742-4682-2-8-4) ::: #### Transfer coefficient t ![](1742-4682-2-8-i14.gif) Thus, the transfer coefficient expresses the proportion between slope of the transfer function and slope of the sigma function, equivalent to a \"fractional\" slope or rate of change (taking the slope of the sigma function as \"total\" slope or rate of change). One might paraphrase the transfer coefficient briefly as the differential ![](1742-4682-2-8-i15.gif). Such a quantity is similar to terms used in specific scientific contexts, such as \"yield\", \"gain\", or \"compliance\". #### Buffering coefficient b ![](1742-4682-2-8-i16.gif) The buffering coefficient thus expresses, analogous and complementary to the transfer coefficient, a \"fractional\" rate of change, namely that of the buffering function with respect to the sigma function. Briefly, the buffering coefficient can be paraphrased by the differential ![](1742-4682-2-8-i17.gif). By analogy to the synonyms of the transfer coefficient, one might call the buffering coefficient also \"fractional loss\" or \"uncompliance\". #### Transfer ratio (or transfer odds) T ![](1742-4682-2-8-i18.gif) In words, the transfer odds reflect the proportion between the two complementary slopes or rates of change, or the differential ![](1742-4682-2-8-i19.gif). Possible synonyms roughly matching current buffering terminology are \"transfer power\" and \"transfer strength\". #### Buffering ratio (or buffering odds) B ![](1742-4682-2-8-i20.gif) Thus, the buffering odds similarly reflect the proportion between two slopes or rates of change, but expressed as the inverse of the transfer odds: buffering function slope over transfer function slope, or the differential ![](1742-4682-2-8-i21.gif). Again, possible synonyms corresponding roughly to current buffering terminology are \"buffering power\" and \"buffering strength\". Importantly, the differentials ![](1742-4682-2-8-i17.gif) = b and ![](1742-4682-2-8-i21.gif) = B are useful as measures of buffering action. They are universal and allow one to quantitate both moderation and amplification. Moreover, the buffering odds B yield the desired absolute ratio scale for buffering action. #### Some properties of the parameters t, b, T, and B The four parameters are completely interdependent, and any single one of these four parameters completely determines the other three and may be used to express the other ones (Figure [4E](#F4){ref-type="fig"} and Additional file [3](#S3){ref-type="supplementary-material"}). The parameters t, b, T, and B are defined for positive, negative and zero-values of x, τ(x), β(x), and of the corresponding proportions between their slopes. For space curves (Figure [3D](#F3){ref-type="fig"}), this property appears trivial. When applied in the context of specific physical sciences, however, this property allows one to describe buffering phenomena that involve negative quantities or zero values. Naturally, this is impossible to achieve with buffering strength units that include logarithmic transforms. In principle, the concept of \"buffered systems\" can be adapted easily to situations where the system state depends not on one single parameter, but on several of them (Additional File [3](#S3){ref-type="supplementary-material"}). Turning the basic approach into a systematic framework ------------------------------------------------------ ### Categories of buffered systems #### Classification according to local behavior We propose to use the term \"buffering\" as a general term for all types of behavior of buffered systems. In contrast, the terms \"moderation\" and \"amplification\" shall denote specific types of behavior of such systems; namely, \"moderation\" for buffering with \\|t(x)\\| \< 1, and \"amplification\" for buffering with \\|t(x)\\| \> 1. Furthermore, we can distinguish \"inverting buffering\" for which t\<0, from \"non-inverting buffering\" for which t\>0. Combining these two criteria, one can distinguish \"inverting amplification\", \"inverting moderation\", \"non-inverting moderation\", and \"non-inverting amplification\". All classical chemical buffers (e.g. for H^+^, Ca^++^or other ions) are \"non-inverting moderators\". #### Classification according to global behavior In addition to the classification by local buffering behavior, one may distinguish buffered systems according to their behavior over the entire definition range. Thus, \"linear buffers\" have transfer and buffering functions that are linear functions of the type τ(x) = a~1~·x + b~1~, and β(x) = a~2~·x + b~2~, where a~1~,a~2~,b~1~,b~2~are constants. For a linear buffer, the proportion between the two partitions is fixed, and the buffering properties are therefore the same at all states of the system. Examples are the partitioning of fluid between two cylindrical vessels (Figure [2A--D](#F2){ref-type="fig"}), partitioning of a solute between the two phases in an oil/water emulsion, or taxation according to a fixed tax rate. \"Nonlinear buffers\" have nonlinear buffering functions of any other type, and thus the proportion between the two partitioning functions varies with x. Thus, the buffering properties of nonlinear buffers depend on the system state, or the value of the independent variable. Examples include the partitioning of fluid between two irregularly shaped vessels (Figure [2E--F](#F2){ref-type="fig"}), the buffering of H^+^ions by water or by weak acids or bases (Buffering II[@B10]), and taxation according to a progressive tax rate. #### Buffering capacity vs. buffering power For the sake of clarity, we would further like to maintain the distinction between intensity termsand capacity termsin the quantitative description of buffering. As differentials, the parameters t, b, T, and B are intensity terms which describe \"fractional rates of change\" or \"proportions between rates of change\". In contrast, a genuine capacity term reflecting an absolute change is obtained by defining a \"buffering capacity\" C~B~as the difference between two particular values z~1~and z~2~of the buffering function: \"buffering capacity\" C~B~≡ Δz = z~2~- z~1~= β(x~2~) - β(x~1~). A \"transfer capacity\" C~T~can be defined analogously as the difference between two particular values y~1~and y~2~of the transfer function. These \"capacities\" are either dimensionless numbers, or they are of the same dimension as y and z. For instance, acid-base physiologists and clinicians use the term \"total body bicarbonate deficit\" (TBBD) in this sense to denote the absolute amount of bicarbonate (indicated in moles or grams) that is needed to increase the present low pH of a patient with metabolic acidosis to the normal value of 7.4. ### Visualizations: communicating vessels, space curve, buffering angle Visualizing the elementary partitioning processes that underlie the buffering phenomena not only has great didactic value, but can as well provide a clear and simple representation of buffering phenomena of genuine scientific interest and that are otherwise complicated or abstract; this may help to avoid or to correct misconceptions. Lack of such direct visual equivalents may have contributed to misconceptions and confusion associated with existing buffering strength units in the past \[[@B5],[@B6]\], and to the persisting difficulties of students with that subject. Fluid-filled communicating vessels can replicate exactly the buffering behavior of virtually all \"classical\" buffering phenomena, and thus provide visible and tangible \"classroom models\". In Figure [2](#F2){ref-type="fig"}, we used this model to illustrate the practical use of the buffering parameters t, b, T, and B. A general method to construct communicating vessel-models of arbitrary buffered systems is described in Additional file [4](#S4){ref-type="supplementary-material"}. The models may be designed in a way that they not only replicate the elementary partitioning process, but also directly visualize buffering strength in terms of the buffering ratio B (or any other buffering parameter, if desired). Three-dimensional space curves are a more general way to represent two-partitioned and buffered systems graphically (Figure [3D](#F3){ref-type="fig"}). When such a space curve is projected parallel to the x-axis onto the yz-plane (defined by the two axes that are used to represent τ(x) and β(x), respectively), then a tangent to the curve will enclose a certain angle α with the τ-axis. This \"buffering angle\" provides another visualization of the proportion between τ\'(x) and β\'(x) and has useful mathematical properties, analogous to the use of trigonometric representations used in electrical engineering (see Additional file [5](#S5){ref-type="supplementary-material"} for details, and Figure [4F](#F4){ref-type="fig"}). ### Axiomatic foundation The measures of buffering introduced in this article (t,b,T,B) are essentially proportions between two elements, albeit with some additional specifications: i) The elements form an ordered pair; ii) The elements and their proportions are not fixed, but may vary as a function of some independent variable; and iii) The elements are rates of change (here: derivatives of differentiable functions). Figure [4](#F4){ref-type="fig"} illustrated some basic aspects of proportions, including \"negative proportions\", and various ways to represent such proportions. In a way, our concept is thus a \"play\" on proportions, and like any game, it is played in accordance with certain rules. Here, elements and rules are purely mathematical objects. For theoreticians, this situation is an invitation to build the buffering concept from scratch on a minimal set of postulates or axioms. For scientific practitioners, motivation to found the concept of buffering on axioms may spring from the experience that seemingly diverse phenomena may be governed by identical principles, and that these more abstract principles allow all of them to be handled with a single mathematical tool. In this sense, Additional file [6](#S6){ref-type="supplementary-material"} is an attempt to demonstrate the common principle behind \"buffering\", \"partitioning\", and \"probability\" in an intuitive way, and to point out the desired properties of an axiomatic formulation of these principles. Additional file [7](#S7){ref-type="supplementary-material"} then presents such an axiomatic formulation of \"buffering\" or \"partitioning\" or \"probability\". These axioms represent the most concise, definitive, and versatile version of our buffering concept. For many readers, it will also offer the most direct approach, especially if they are already familiar with Kolmogorov\'s axiomatic foundation of a probability measure. Importantly, the axiomatic foundation makes the theoretical concept more powerful: Firstly, stringent formalization allows one to decide definitively whether the concept is logically consistent and complete. Secondly, axioms represent the concept in its most general form, and this form is most likely to stimulate free (and correct) use in very diverse, sometimes unanticipated contexts. Moreover, systems of continuous functions are adequate only for a macroscopic description of buffering phenomena. On a microscopic scale, the continuum hypothesis ceases to apply, whereas the quantitative principles that govern these phenomena remain the same. The axioms therefore also provide for systems of discrete functions. Finally, the axiomatic form exposes the striking formal similarity between the concepts of \"probability\" and of \"buffering\" (detailed in Tables 3 and 4 of Additional file [7](#S7){ref-type="supplementary-material"}). This similarity raises deep questions -remaining to be explored- regarding both the interpretation and the mathematical foundation of \"probability\". ### Interconversions and practical rules The general form of the axioms allows one to identify several \"technical\" aspects of buffered systems that are practically relevant. Firstly, there exist equivalencies with respect to buffering properties between systems that present initially in very different forms. Understanding these equivalencies and being able to interconvert such systems allows one to reduce complexity (Additional file [8](#S8){ref-type="supplementary-material"}). Secondly, buffering phenomena can be formalized as partitioned and buffered systems in more than one, formally correct way. The various options at this step allow one to choose the most suitable formalization, and point to some efficient experimental approaches (Additional file [9](#S9){ref-type="supplementary-material"}). ### Practical applications Practical relevance and use of the concept are worked out in the accompanying paper (Buffering II[@B10]), where we apply our definitions and units to various buffering phenomena of genuine scientific interest. These analyses offer some fresh though compelling looks on \"classical\" buffering phenomena (H^+^buffering by pure water or by solutions of weak acids/bases), and demonstrate that our concept affords rigorous quantitative treatment of \"non-classical\" buffering phenomena for which useful measures of buffering strength have been unavailable so far (redox buffering and blood pressure buffering). Finally, a generalization opens the concept to non-stationary systems and thus allows one to quantitate time-dependent buffering, or \"muffling\", and \"systems level buffering\" in an equally rigorous manner. Discussion ========== The introduction of suitable abstractions is our only mental aid to organize and master complexity.-- Edsger W. Dijkstra In this article, we introduced quantitative measures of buffering action based on a purely mathematical concept of \"buffering\". The nucleus of the concept was to describe partitioning processes by means of the proportions between partial and total \"changes\" or \"flows\". On this basis, we could define four interrelated, dimensionless measures: transfer coefficient t, buffering coefficient b, transfer ratio T, and buffering ratio B. Together, they allow one to quantitate the behavior of buffered systems in a way that is analogous to the quantitation of chance using \"probabilities\" and \"odds\". The magnitude of buffering action may thus be measured using the \"buffering coefficient\" which provides a relative scale normalized to 1. Alternatively, one may use the \"buffering ratio\" in order to quantitate buffering action by means of an absolute scale with equal intervals and an absolute zero, the highest scale type possible. \"Buffering\" according to this definition turned out to be an entirely mathematical concept. Phenomena encountered in the \"real world\" may or may not be related to this mathematical concept in exactly the same way in which phenomena may or may not be related to mathematical concepts in general. The concept of exponential decay, for instance, can be stated in purely mathematical terms, but is also exhibited in more or less perfect form by several natural phenomena, such as decaying radioisotopes or chemical reactions on their way to equilibrium. Moreover, our mathematical concept of buffering can also describe amplification phenomena, just as the concept of exponential decay can seamlessly turn into a concept of exponential growth simply by allowing for exponents greater than one. This section discusses the intrinsic, formal properties of the concept. Some of its -- sometimes hidden -- connections to previous work, especially to ideas of Henderson, Van Slyke, or Neher & Augustine are exposed and discussed in Additional file [10](#S10){ref-type="supplementary-material"}. Kolmogorov\'s axiomatic system of probability has been covered extensively elsewhere (see literature in \[[@B7],[@B8]\]). The technical and theoretical implications of our \"Non-Kolmogorov probability measure\" require further study, but cannot be worked out here. Detailed treatments of specific buffering phenomena are presented in the accompanying paper (Buffering II[@B10]). Properties and significance of the general, formal approach to the quantitation of buffering action --------------------------------------------------------------------------------------------------- The introduction and Additional file [2](#S2){ref-type="supplementary-material"} listed a number of major problems associated with the present approaches to the quantitation of buffering action. Our formal, general approach and the four buffering parameters t, b, T, and B, provide a theoretically rigorous and practically useful solution to these problems. ### A ratio scale for buffering strength The buffering odds B provide an absolute, dimensionless ratio scale for buffering action. This is the highest possible type of scientific scale. The advantages of ratio scales, i.e., equal interval scales with an absolute zero, have been pointed out in the introduction and in Additional file [2](#S2){ref-type="supplementary-material"}. Our buffering strength scale is free of scale artefacts and can be used accurately, not just approximately, for small and large, positive and negative values alike. A further benefit gained with a ratio scale is the possibility to build simple, intuitive models of buffering, e.g. with communicating vessels. Of the previous units for buffering strength, only Neher & Augustine\'s \"Ca^++^binding ratio κ~s~\" yields a ratio scale. ### A universal scale for buffering strength Our definitions of buffering and of measures to quantitate buffering are purely formal, mathematical ones, and the measures t, b, T, and B are all dimensionless numbers. Therefore, this conceptual framework is generally applicable and not arbitrarily limited to buffering phenomena of a particular chemical or physical nature. The concept allows one to handle all classical examples of buffering, such as Ca^++^or H^+^buffering. In addition, it can be applied to multiple further buffering phenomena encountered in chemistry, biology, physiology, or elsewhere. Application of our concept to familiar examples shows that it is valid, i.e., it reflects faithfully what is qualitatively understood by the word \"buffering\", and that this will hold generally, not only when certain boundary conditions are met. We stress the formal aspect of buffering: \"Buffering\" is a quantitative pattern abstracted from \"real\" things, not a real thing itself. Therefore, this pattern can be employed freely and in various ways as a tool for the quantitative description of certain aspects of reality. Whether something is the buffer or that which is being buffered is a matter of perspective and thus determined by the analyst, not by reality. Therefore, the definition of a buffer is an operational one: Something that is expressed in terms of a buffering function β constitutes, by definition, a buffer. Conversly, something that cannot be measured by these units cannot and should not be described in terms of buffering terminology. Such an operational definition requires a general, formal concept. In contrast, buffering strength units that are specific for H^+^, Ca^++^, or other particular entities can hardly be wrought into a convincing operational definition of buffering that is not overly exclusive and limited. ### A scale for moderation and amplification The mathematical notation of our buffering concept makes it easy to recognize the common pattern behind \"moderation\" and \"amplification\", and to accommodate this union formally. A single unit is sufficient to deal with both, and such a unit is also called for frequently when both types of buffering behavior occur in a single system. The association of \"buffering\" exclusively with \"attenuation\" or moderation, and the resulting mental divide with respect to the treatment of attenuation vs. amplification phenomena may be related to the historical roots of the buffering concept in acid-base chemistry where \"moderation\" is the most prominent finding. Overcoming that divide greatly enhances the usefulness of the buffering concept. For instance, we can connect our buffering concept directly to systems and control theory, as detailed in the accompanying article (Buffering II[@B10]). This link can enrich systems and control theory by providing a currently lacking rigorous definition of \"systems level buffering\" and an accompanying unit to measure this quantity. On the other hand, this link can expand the application range of the buffering concept to all objects and phenomena already studied by systems and control theory. Most importantly, our approach brings together conceptually and technically \"buffering\" as a homeostatic mechanism, and control theory as the dominant formal language for the description of homeostasis in physiological systems. ### A standard scale for buffering strength? The parallel use of multiple, incommensurate scales for buffering strength has engendered ambiguous terminology, misunderstandings and pseudo-problems. Homonymic usage of the term \"buffering\" might be avoided by agreeing on a certain convention. Ideally, such a convention should codify not just any usage, but the one that affords the highest possible type of scale. Furthermore, a standard scale should cover all possible cases of buffering, and should not arbitrarily exclude some of them. Our concept constitutes such an all-purpose yardstick for buffering strength. It is a valid measure of buffering, universally applicable, and yields a scale of the highest possible type. No other buffering strength unit satisfies these requirements. Moreover, it is also hard to imagine a more compelling, less arbitrary standard scale than an \"absolute ratio scale\" as provided by the buffering odds B: The numerical value of the buffering odds B is completely determined by the general definition of B as the ratio of two derivatives and does not depend on any arbitrary scaling factors or units. In contrast, for instance, the definition of the unit β~H+~= dBase/dpH includes a number of additional, unnecessary \"rules\" that need to be observed in order to obtain the correct value: i) Express the concentration of strong base and of free H^+^ions in multiples of Avogadro\'s number per liter (as opposed to absolute numbers, mass, or others); ii) Carry out a numerical transformation of one quantity (\[H^+^\]~free~into pH), but not of the other (\[Strong Base\]); iii) For the transformation, choose a logarithmic one (as opposed to other transformations, e.g. exponential ones); and iv*) Use the number 10 as the base in this transform (as opposed to e or any other). These procedures are all mere conventions, not compelling formal constraints or the results of particular scientific givens. Another example is de Levie\'s redox buffer strength which includes multiplication by a factor 1/ln(10) for sheer convenience \[[@B9]\]. Conclusion ========== The quantitation of chance can be achieved cleanly and comprehensively by using either probabilities (of occurrence and non-occurrence of an event) or odds (for and against an event). The analogous measures proposed in this article, namely the coefficients (transfer coefficient t and buffering coefficient b) and ratios (transfer ratio T and buffering ratio B), are suited to serve as standard units for buffering action. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Current Usage of the \"Buffering\" Paradigm Outside Acid-Base Chemistry ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 2 Problems with the Current Approaches to the Quantitation of Buffering ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 3 Properties of the Parameters t, b, T, and B ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 4 Constructing Communicating Vessel-Models of Partitioned and Buffered Systems ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 5 A Trigonometric Representation of Buffering Behavior: The \"Buffering Angle\" ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 6 From Galton Desk to Communicating Vessels -- \"Partitioning\" as a Common Pattern Behind Probability and Buffering ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 7 Axiomatic Foundation of the Formal & General Approach ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 8 Conservative and Non-Conservative Partitioned Systems -- Equivalences and Interconversions ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 9 Some Useful General Principles Regarding the Practical Application of the Formal & General Buffering Concept ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 10 Historical Note: Origins of the Formal & General Approach ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ This work was inspired by a stay in the lab of Walter F. Boron, Dept. of Cellular Molecular Physiology, Yale University, New Haven, CT, USA	PubMed Central	2024-06-05T03:55:55.613906	2005-3-15	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1079953/", "journal": "Theor Biol Med Model. 2005 Mar 15; 2:8", "authors": [ { "first": "Bernhard M", "last": "Schmitt" } ] }
PMC1079954	Introduction ============ In the preceding article (Buffering I[@B14]), we presented a formal and general framework for the quantitation of buffering action. The purpose of the present article is to apply that mathematical tool to the analysis of some scientifically important buffering phenomena. Recall that we formulated buffering phenomena as the partitioning of a quantity into two complementary compartments, and then used the proportions between the respective flows as a simple quantitative criterion of buffering strength. The two measures of buffering action were i)the buffering coefficient b, defined as the differential d(buffered)/d(total), and ii)the buffering ratio B, defined as the differential d(buffered)/d(unbuffered). Moreover, the following analyses will make use of the distinction between various categories of buffered systems (e.g. conservative vs. non-conservative partitioned systems), and will exploit the equivalencies and interconversions between these categories. To begin, we revisit a classical case of acid-base buffering: H^+^ion buffering in a solution of a weak acid. This process can be described easily in terms of a conserved quantity (total H^+^ions) that partitions into two complementary compartments or states (bound vs. free). Such a system was termed a \"conservative buffered system\". Conservative buffered systems constitute the most simple buffered systems according to our buffering concept, and they provide a suitable framework to describe further classical buffering phenomena. An important one among them, the so-called \"self-buffering\" of H^+^ions in pure water, is analyzed in [Additional file 2](#S2){ref-type="supplementary-material"}. The concept of a \"conservative buffered system\" can be applied readily and fruitfully to numerous buffering phenomena that involve quantities other than H^+^or Ca^++^ions (\"non-classical\" buffering phenomena). Some examples are presented in the [Additional file 3](#S3){ref-type="supplementary-material"}; these include a straightforward approach to the notoriously difficult quantitation of \"redox buffering\", and examples which demonstrate that the concept of \"buffering\" is by no means limited to the natural sciences. In the second section, we analyze the buffering of organ perfusion in the face of variable blood pressure. Here, the independent variable is blood pressure, whereas the dependent variables are volume flows. Systems that involve different physical dimensions, however, cannot be formalized in terms of \"conservative buffered systems\", the basic form of our buffering concept. Here, the general form of our buffering concept (Buffering I[@B14])proves to provide a rigorous and reliable framework for the treatment of such \"non-conservative\" and \"dimensionally heterogeneous\" buffered systems. The third section extends the buffering concept to time-dependent buffering processes. \"Time\" as a potentially important aspect of buffering becomes evident, for instance, in the Ca^++^concentration transients that are elicited by the brief openings of a calcium channel in the surrounding cytoplasm \[[@B1]\]. It was an important achievement to realize that this blunting of concentration swings represents an independent quantity, and to suggest a term as fitting as \"muffling\" for it \[[@B2]\]. However, for reasons detailed in [Additional file 5](#S5){ref-type="supplementary-material"}, the available units of \"muffling strength\" are not satisfying. We introduce an extension of our buffering concept that clearly satisfies all criteria required for a muffling strength unit and provides a dimensionless ratio scale for this quantity. Furthermore, this unit is able to connect \"muffling\" and \"buffering\" both conceptually and numerically: Steady-state buffering is shown to represent the limiting case of time-dependent muffling for infinitely long time intervals and infinitely small perturbations. Finally, [Additional file 6](#S6){ref-type="supplementary-material"} sketches how our concept of buffering can serve to quantitate \"systems level buffering\" in the context of control systems. Buffering is an important aspect of homeostasis in physiological systems, and control theory provides a powerful general language to describe homeostatic processes. So far, however, the concept of buffering could not be accomodated explicitly in this framework. We show that \"buffering\" and \"control theory\" can be connected conceptually and numerically in a straightforward and meaningful way. To quantitate systems level buffering, we need to exploit simultaneously all possibilities and features of our buffering concept, because control systems may be conservative or non-conservative, dimensionally homogeneous or heterogeneous, and time-invariant or time-dependent. The buffering of H^+^ions by weak acids or bases -- Buffering as partitioning of a conserved quantity and the concept of \"Langmuir buffering\" =============================================================================================================================================== Weak acids in conjunction with their conjugate base, and weak bases in conjunction with their conjugate acids, are the prototypical \"buffers\". They were the first buffers put to action by biochemists in order to stabilize the pH of solutions, and they were also the first buffers to receive thorough theoretical analysis. Numerous textbook definitions explicitly equate \"buffers\" with \"mixtures of weak acids plus conjugate base\" (or vice versa), and this notion became so inextricably woven into our thinking about buffering that the distinction between the chemical substrate of this process and the abstract quantitative pattern manifest in it fell into oblivion. In a wider sense, however, the manifold varieties of ligand binding are indeed responsible for a large number of buffering phenomena encountered in biochemistry or physiology. For instance, ions such as Ca^++^are buffered by physico-chemical processes analogous to H^+^buffering, albeit without the involvement of literal weak acids or bases. Compared to buffering involving other mechanisms, such as blood pressure buffering or systems level buffering (see below), buffering via ligand binding exhibits some distinct quantitative patterns. The terminology of the original acid-base concept of buffering, however, is too specific as to serve as a general framework for the treatment of these phenomena. The following section demonstrates how the quantitative patterns of buffering via ligand binding can be caught with the aid of the four parameters t, b, T, and B. The analysis explores the classic case of a weak monoprotic acid dissolved in water. Mathematical model of free and bound H^+^ion concentrations in a solution of a weak acid ---------------------------------------------------------------------------------------- To obtain an explicit quantitative description of buffering by weak acids, we first recapitulate the mathematical model that describes the concentrations of H^+^ions in an aqueous solution of a weak acid as a function of freeH^+^ion concentration. Subsequently, we reformulate the concentrations of free and bound H^+^ions as functions of totalH^+^ion concentration, and combine these functions into a buffered system. Finally, we derive the four parameters t, b, T, and B from this system. A weak monoprotic acid HA can dissociate into a free H^+^ion and a conjugated base A^-^, to an extent that is dictated by K~A~, the acid constant in water, according to: K~A~× \[HA\] = \[H^+^\]~free~× \[A^-^\]~free~. The total amount of weak acid \[A\]~total~equals the sum \[A^-^\]~free~+ \[HA\] of dissociated and undissociated weak acid; the system is \"conservative\". We can therefore substitute \[A^-^\]~free~by \[A\]~total~- \[HA\] and obtain: ![](1742-4682-2-9-i1.gif) and after several intermediate steps: ![](1742-4682-2-9-i2.gif) For the sake of readability, we express the same relationship in a more general notation: ![](1742-4682-2-9-i3.gif) where c and d stand for the constants \[A\]~total~and K~A~, respectively, and the variables y and z correspond to \[H^+^\]~free~and \[HA\], respectively. This latter equation describes a hyperbola that approaches c as y increases to infinity. It was first used empirically by Hill \[[@B3]\], but became a much more meaningful mathematical model after Langmuir had supplied a mechanistical interpretation, namely in terms of non-cooperative binding of a ligand to a finite number of binding sites \[[@B4]\]. That model is widely applicable to numerous phenomena, e.g. receptor-ligand interactions, adsorption processes at surfaces, or enzyme kinetics, to name but a few. The same rules of non-cooperative binding apply to the binding of H^+^ions to the conjugate base of a monoprotic weak acid. In order to move from specific acid-base terminology to a general ligand binding terminology, we re-interpret the symbols in the equation ![](1742-4682-2-9-i4.gif) as follows: Variable z represents the concentration of bound ligand, y the concentration of free ligand, c the total number of binding sites, and d the equilibrium constant K~d~of the complex with respect its dissociation products. \"Langmuir buffers\" -------------------- So far, we have described bound ligand z as a function of free ligand y. Next, we express the concentration of free ligand y as a function of total ligand x = y+z: ![](1742-4682-2-9-i5.gif) and the concentration of bound ligand z as a function of total ligand x: ![](1742-4682-2-9-i6.gif) The relation between total, free, and bound ligand for non-cooperative binding to a fixed number of binding sites with similar affinity is shown in Figure [1A](#F1){ref-type="fig"}. It is easy to double-check that the sum y(x)+z(x) equals x, i.e., the conservation condition is satisfied. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### *[Buffering via non-cooperative ligand binding: \"Langmuir buffering\"]{.underline}. The prototype of Langmuir buffering is the buffering of H^+^ions in a solution of a weak acid. A, Relation between the three variables of a \"Langmuir\"-type buffer. Concentrations of free ligand (red), bound ligand (blue), and total ligand in a solution of a weak acid. The relations between the three variables are computed from the equation ![](1742-4682-2-9-i7.gif), where K~d~stands for the dissociation constant of the buffer-ligand complex, and \[buffer\] for total buffer concentration. \[buffer\] and K~d~are assumed to be constant. Plotted for \[buffer\] = 5 and K~d~= 1. B, Describing \"Langmuir buffering\" using the four buffering measures t, b, T, and B. Titration of a \"Langmuir buffer\" with increasing concentrations of ligand; constant parameters are: \[buffer\] = 100, K~d~= 10, arbitrary concentration units. Characteristic system states shown are the \"half-saturation point\" of buffer (asterisk), the \"equipartitioning point\" where half of the added ligand remains free, and the other half is bound by the buffer (open circle), and the \"break even point\" where the ligand inside the system is half bound, half free (closed circle). Top panel, left:Transfer function τ, i.e., free ligand concentration (ordinate)as a function of total ligand (ordinate). Top panel, right:buffering function β, i.e., bound ligand concentration as a function of total ligand. Middle panel, left:Transfer coefficient t, i.e., the (differential) fraction of added ligand that enters the pool of free ligand. Middle panel, right:Buffering coefficient b, i.e., the (differential) fraction of added ligand that becomes bound to buffer. Bottom panel, left:Transfer ratio T= d(free)/d(bound), i.e., the differential ratio of additional free ligand over additional bound ligand. Bottom panel, right:Buffering ratio B= d(bound)/d(free), i.e., the differential ratio of additional bound ligand over additional free ligand. The parameters b and B provide two complementary measures of buffering strength. C, Buffering strength of a Langmuir buffer as a function of both total ligand concentration and affinity. Wireframe surface:The buffering ratio B is shown on the vertical axis; affinity expressed as 1/K~d~; concentration of ligand, \[ligand\], and K~d~in arbitrary concentration units. Contours on bottom:Lines connect states of identical buffering strength. For a buffer with a given K~d~, buffering strength decreases monotonically with increasing ligand concentration. However, at a fixed ligand concentration, buffering strength as a function of affinity runs through a maximum. D, Visualizing Langmuir buffering by two-dimensional plots (same data as in Figure C). Left hand, linear plot; white lines, states of identical buffering strength; black lines, states of identical fractional buffer saturation. Right, double-logarithmical plot. black lines, states of identical buffering strength; red lines, states of identical fractional buffer saturation. E, Using the \"buffering angle\" to visualize Langmuir buffering: cylinder plot. As shown in Buffering I, the specific angle α for which \[α = arccos(T) and α = arctan(B)\] can unambiguously represent the buffering parameters t(x), b(x), T(x) and B(x) at a given point on the x axis. Consequently, a curve on the surface of a unit cylinder can represent the buffering behavior for an entire range of x values, yielding a \"state portrait\". State portraits of several Langmuir buffers are shown. Curves with Roman numerals (I-IV) of different color: effect of decreasing ligand affinity at fixed total concentration. Curves with Arabic numerals (1--4) of different size: effect of increasing total buffer concentration. Less intuitively, yet more practically, the cylinder surface may be \"flattened\" out and represented in two dimensions (not shown). Blue segment:buffering angle α for curve 4. F, Using the \"buffering angle\" to visualize Langmuir buffering: polar graph. Alternative form of a buffering state portrait: each point on the curve is characterized by a \"buffering angle α \" with the vertical axis (clockwise) and a radius (here plotted logarithmically), which correspond to buffering angle α and total ligand concentration, respectively. Open circles, equipartitioning points, i.e., where t = b and α = 45°. ::: ![](1742-4682-2-9-1) ::: To turn these two functions into a \"buffered system\", we assign the role of \"transfer function\" τ(x) to the free H^+^concentration y(x), and the role of \"buffering function\" β(x) to the bound H^+^concentration z(x). Because many common and important systems follow this quantitative pattern, it might be useful to have a specific term to refer to them. We suggest the term \"Langmuir-type buffers\"or \"Langmuir-type buffered systems\". Briefly, a Langmuir-type system can be defined as an ordered pair of two functions {y = τ(x), z = β(x)} that satisfies the three conditions (x = y+z) and z = c × y/(d+y) and c, d ∈ ![](1742-4682-2-9-i39.gif)^+^. In systems that involve ion concentrations, both τ(x) and β(x) naturally assume a value of zero at x = 0, i.e., they pass through the origin. However, we can relax this fourth constraint by allowing for offsets τ~0~and β~0~, respectively, without altering the buffering properties (Buffering I [@B14]). Thus, we obtain the general form of a Langmuir buffer B~Langmuir~as: ![](1742-4682-2-9-i8.gif) The buffered system constituted by the solution of a weak acid in water is \"dimensionally homogeneous\": the variables x, y, and z are either all dimensionless (e.g. when expressed as multiples of K~d~or K~A~), or they all have the dimension of a concentration (e.g. when expressed in moles/liter). Similarly, H^+^buffering in pure water is represented by a dimensionally homogeneous buffered system ([Additional file 2](#S2){ref-type="supplementary-material"}). Computing the buffering parameters in Langmuir-type systems ----------------------------------------------------------- Because we find in conservative systems that τ\'(x)+β\'(x) = 1, the transfer and buffering coefficients are simply equal to the respective first derivatives: ![](1742-4682-2-9-i9.gif) These equations have unique solutions as long as the constants c and d are positive; for dissociation constants and concentration terms, this is always warranted. It is easy to verify that, consistent with the conservation condition σ\'(x) = 1, the coefficients t and b always add up unity. From t and b, we can then compute the transfer ratio T and the buffering ratio B as functions of x: ![](1742-4682-2-9-i10.gif) Expression of transfer and buffering ratio as functions of y (instead of x) results in equivalent, yet much simpler forms: ![](1742-4682-2-9-i11.gif) The buffering parameters t, b, T, and B provide a complete description of H^+^buffering by weak acids (Figure [1B](#F1){ref-type="fig"}). They allow us to elucidate the common properties of all Langmuir buffers, both in terms of a communicating-vessels model ([Additional file 1](#S1){ref-type="supplementary-material"}) and mathematically (see following paragraph). ### General properties of Langmuir buffer systems #### Langmuir buffers are \"finite capacity buffers\" At x = 0, the buffering function β(x) has a finite value β~0~∈ R. As x increases, β(x) increases monotonically, asymptotically approaching a finite value c+β~0~. When a Langmuir buffer is modelled by communicating vessels, the buffering vessel has a finite volume, in spite of its infinite height. #### Buffering strength is maximal when ligand concentration is zero In absolute values, we find for buffering coeffient b and buffering ratio B at ligand concentration x = 0: ![](1742-4682-2-9-i12.gif) and ![](1742-4682-2-9-i13.gif) The corresponding values of transfer coefficient t and transfer ratio T are: ![](1742-4682-2-9-i14.gif) and ![](1742-4682-2-9-i15.gif) In the model, the cross-sectional area of the buffering vessel is largest at its base. For the special case of H^+^buffering in a solution of a weak acid, this means: The maximum buffering ratio B is obtained simply by divding the concentration of total weak acid by the acid constant K~A~: ![](1742-4682-2-9-i16.gif) This relationship can be exploited to elegantly determine total concentration A~total~of a buffer with known K~A~or K~d~: The buffering ratio B is determined experimentally at ligand concentrations that are much smaller than K~d~(x\<\<K~d~), from which A~Tot~can be approximated as A~total~≈B × K~d~\[[@B5]\]. #### Langmuir buffers are \"non-linear buffers\" Buffering coefficient b and buffering ratio B decrease monotonically with increasing x (or y), asymptotically approaching zero. In the communicating vessels-model, the buffering vessel is not parallel-walled, but tapers off towards the upper end. #### Langmuir buffers are \"non-inverting moderators\" Over the entire domain ![](1742-4682-2-9-i39.gif)^+^, the buffering coefficient b assumes values between 0 and 1 (0≤b\<1), and the buffering ratio B is always nonnegative (B≥0). In the model, this property is apparent inasmuch as the buffering vessel has fixed walls with positive-valued cross-sectional areas (in fact, \"negative-valued cross-sectional areas\" do not exist, and the vessel model can thus not replicate amplification or inversion). #### Langmuir buffers have a \"break even point\" at x = 2c-2d In the vessel model, \"break even points\" are fluid levels at which transfer and buffering vessel each contain identical fluid volumes. Trivially, this is the case when both vessels are empty, or at x = 0. However, there is a second such system state at x = 2(c-d) if c\>d (based on the definition c, d\>0 and assuming that both functions cross the origin). For x\<2(c-d), the greater part of the quantity is found in the \"buffering compartment\"; for x\>2(c-d), the greater part is in the \"transfer compartment\". \"Break even-point\"may be a suitable term to refer to this point. If however c\<d, then no second break even-point exists, and the transfer compartment contains at all values of x more of the quantity than the buffering compartment. #### Langmuir buffers have a half-saturation point at x = c/2+d In the vessel model, half-saturation of buffer means that the buffering vessel is half full. In terms of total volume x, there is a value x~0.5~for which the buffering function β(x) becomes equal to ![](1742-4682-2-9-i17.gif), namely at ![](1742-4682-2-9-i18.gif), assuming that τ~0~,β~0~= 0. Thus, x~0.5~may be called the \"half-saturation point\"of a given Langmuir-buffer. In terms of the the transfer function y = τ(x), half-saturation of a Langmuir buffer is reached at y = d. This result is a well known property of systems conforming to Langmuir\'s equation. Naturally, infinitecapacity-buffers (e.g. pure water which is not a Langmuir buffer) cannot have a half-saturation point. #### At half-saturation, the buffering ratio B of a Langmuir buffer is one fourth of its maximum When the buffering vessel is half full, its cross-sectional area is one fourth of the cross-sectional area of the transfer vessel. At the half-saturation point x~0.5~, buffering strength has the following values: ![](1742-4682-2-9-i19.gif) and ![](1742-4682-2-9-i20.gif) Thus, for H^+^buffering in a solution of a weak acid, the buffering ratio B at half-saturation is one fourth of the concentration of total weak acid divided by the acid constant: \[A\]~total~/(4 × K~A~). #### Langmuir buffers have an \"equipartitioning point\" at x = c-d In the vessel model, equipartitioning means that the partial flows into the two partitions (buffering and transfer compartment) are equal. This is the case when transfer and buffering vessel have equal cross-sectional areas. Here we find that b = t = 0.5 ∧ B = T = 1. Note the difference between \"break even point\" and \"equipartitioning point\". Comparison with other descriptions of H^+^-buffering by weak acids ------------------------------------------------------------------ Analysis of H^+^buffering by weak acids by means of the buffering coefficient b and buffering ratio B has thus led to conclusions that differ considerably from the standard view of buffering which is based on Van Slyke\'s \"buffer value\". Interestingly, our units will, similarly to Van Slyke\'s buffer value, identify as the strongest H^+^ion buffer for a given pH value that weak acid whose pK~A~equals this pH, in spite of the conflicting conclusions as to the point of maximum buffering strength. [Additional file 1](#S1){ref-type="supplementary-material"} discusses in more detail the impact of different units on our perception of H^+^buffering by weak acids or bases. Other conservative buffered systems ----------------------------------- This \"worked example\" of H^+^buffering by weak acids demonstrated how the concept of conservative buffered systems can be applied in practice. There are multiple other buffering phenomena that conform to that concept and which can be analyzed in exactly the same manner. Among them, H^+^buffering by pure water is of particular interest ([Additional file 2](#S2){ref-type="supplementary-material"}). Oxygen buffering by hemoglobin, a mechanism of great physiological importance, would be another example of the same basic type, but not involving H^+^ions, and with yet different quantitative behavior. The concept of conservative buffered systems can also be applied directly to quantities that are governed by mechanisms unrelated to ligand binding, e.g. to heat energy. Moreover, conservative systems need not be restrained to non-inverting moderation, but may exhibit amplification as well. Such non-classical examples of conservative buffered systems are presented in [Additional file 3](#S3){ref-type="supplementary-material"}. Non-conservative systems are treated in the following section. The buffering of organ perfusion in the face of blood pressure fluctuations -- The concept of \"nonconservative buffered systems\" ================================================================================================================================== The term \"blood pressure buffering\" is used in various ways ------------------------------------------------------------- Some use this term as a synonym to \"autoregulation\" of organ perfusion, i.e., the maintenance of a constant blood flow in the face of variable blood pressure or cardiac output. Others apply it to the mechanisms that blunt the pressure-raising or -lowering effects of physiological maneuvers and drugs; according to a major contributor, this phenomenon is also called \"baroreceptor buffering\" \[[@B6],[@B7]\]. Finally, the term \"blood pressure buffering\" sometimes refers to the attenuation of blood pressure variability, i.e., of the oscillations of mean arterial blood pressure (MAP) around its average \[[@B8]-[@B10]\]. Lack of a quantitative measure of \"blood pressure buffering\" -------------------------------------------------------------- Experimental studies on blood pressure buffering usually report the magnitude of all relevant basic quantities in terms of scientific scales. In contrast, the magnitude of \"blood pressure buffering\" itself is specified merely in terms of \"more\" or \"less\", without attempting to extract from the data some specific numerical value that could serve as a measure of this central quantity. In other words, the currently available scales for blood pressure buffering strength are non-metrical, ordinal scales. These scales are rather primitive and do not allow one to carry out a number of desirable and legitimate scientific operations (e.g. mutual comparisons of the buffering strengths of individual mechanisms that jointly contribute to blood pressure buffering, or comparison of \"blood pressure buffering\" to the buffering of other physiologic parameters such as pH, Ca^++^concentration, or body temperature). This section demonstrates that our concept of buffering readily quantitates \"blood pressure buffering\" in most of its meanings. However, many of these phenomena cannot be described any more in terms of simple \"conservative\" partitioned systems. Rather, the systematic treatment of these buffering phenomena makes it necessary to recall and utilize the distinctions between various categories of buffered systems that were outlined in the preceding article (Buffering I[@B14]): conservative vs. non-conservative, and dimensionally homogeneous vs. heterogeneous systems. Moreover, one type of blood pressure buffering, that of blood pressure variabilitybuffering, will turn out to resist formalization as a \"buffered system\" altogether, suggesting that this paradigm actually refers to something that is essentially different from buffering in the common sense. Autoregulation of flow in the face of variable total flow -- dimensionally homogeneous systems (Figure [2A--C](#F2){ref-type="fig"}) ------------------------------------------------------------------------------------------------------------------------------------ ::: {#F2 .fig} Figure 2 ::: {.caption} ###### [Buffering in conservative and non-conservative buffered systems, illustrated by blood pressure buffering]{.underline}. A-C, Conservative buffered system: Buffering individual flow against variations of total flow. Pipework model of circulation, where cardiac output corresponds to total volume flow Φ, and volume flows in individual organs to volume flows φ~i~in individual tubes. A, Zero buffering. In a circulation comprising a single hydraulic conductor, a total volume flow (Φ) established by a pump (⊗) results in a partial flow (φ~1~) of equal magnitude in the conductor (red). Their quantitative relation can be represented in signal transduction formalism as a \"transfer element\" where input x corresponds to Φ, output y to φ~1~, and the transfer properties are characterized by a constant transfer coefficient of 1. B, Linear buffering. Total volume flow partitioning into two parallel hydraulic conductors. Changes of total flow Φ now elicit smaller changes of the partial flow φ~1~(red)-- due to \"buffering\" by the second conductor (blue). Transfer and buffering behavior with respect to the upper vessel can be expressed in terms of fixed, dimensionless fractions t and b. C, Nonlinear buffering. When one or both vessels have elastic walls, hydraulic conductance and thus responsiveness to changes in Φ will vary with the absolute value of Φ. Transfer and buffering coefficients become nonlinear functions of Φ. D, Non-conservative buffered systems: Buffering individual flow against variations of perfusion pressure. Organ volume flows φ~i~are described as functions of perfusion pressure ΔP. With different physical dimensions for input and output (pressure vs. flow), the transfer coefficient for vessel 1 alone has the dimension of a hydraulic conducance L~P1~. With a second vessel added in series, changes of perfusion pressure translate into smaller changes of volume flow. This effect can be interpreted as \"buffering\" and expressed quantitatively using the buffering parameters t, b, T, and B. If one or both vessels are elastic, transfer and buffering functions become nonlinear functions of perfusion pressure ΔP. ::: ![](1742-4682-2-9-2) ::: We can describe the individual volume flows in several parallel tubes as a function of total volume flow through that system. We then find that flow in a given tube is stabilized or \"buffered\" against a given change of total flow by the parallel tubes. This is one way how one can achieve stabilization of organ perfusion in the face of variable blood flow (e.g. at rest vs. exercise), or its adaptive regulation (e.g. in the skin, via opening or closing of shunting vessels). Such systems can be formalized as conservative buffered system and analyzed in the same way as shown for H^+^buffering (see above)and for other phenomena ([Additional file 3](#S3){ref-type="supplementary-material"}; there, this particular case is also worked out explicitly). Autoregulation of flow in the face of variable pressure -- dimensionally heterogeneous buffered systems (Figure [2D](#F2){ref-type="fig"}) ------------------------------------------------------------------------------------------------------------------------------------------ ### Perfusion in the absence of autoregulation or buffering Rather than using total volume flow as independent variable, volume flows in systems of tubes may be expressed as functions of pressure. According to the basic law of convective volume transport, the flow Φ in a single tube A is a linear function Φ = L~A~× ΔP of the pressure difference ΔP between its inlet and outlet, with hydraulic conductance L~A~as the proportionality factor. Assumptions herein are laminar flow and the absence of other relevant forces such as osmotic gradients. In a black box-approach, we may look at the tube as a transfer element, characterized by an input x, an output y, and a transfer function y = τ(x). Because of the tube\'s rigidity and the resulting constancy of hydraulic conductance, the transfer function τ(x) is a linear function of the type y = b × x. Provided the system comprises only the said tube in the system with no further hydraulic conductance in series, then τ\'(x) corresponds to the hydraulic conductivity L~A~of tube A. We can add a second, \"virtual\" second output z = β(x) to the black box that expresses the flow in a putative second hydraulic resistance in series, induced by a corresponding fraction of the total pressure gradient. Here, there is no second resistance, and z trivially assumes a value of 0. We can thus formulate a buffered system as: ![](1742-4682-2-9-i21.gif) When computing the four buffering parameters t, b, T, and B, it becomes apparent that all four are again dimensionless, even though this buffered system is dimensionally heterogeneous. Under the indicated conditions, these parameters are ![](1742-4682-2-9-i22.gif) Buffering coefficient b and buffering ratio B both equal zero, in agreement with the absence of any buffering. ### Linear buffering of flow against pressure changes Next, we add a second piece of tubing. When connected in parallel, this tube B does not affect the pressure-flow relationship, only the relation between total flow and individual flow. In contrast, when the second piece of tubing is connected in series, the pressure-flow relation is altered profoundly. Analogously, autoregulation of blood flow in living organisms may be brought about via modulation of hydraulic conductance (by constriction or relaxation) of blood vessels that are in series with the capillary bed and that belong to the organ\'s proper vascular bed (i.e., they are located between the two points that used to define the relevant pressure gradient). Whether this resistance is located upstream, downstream, or both is not relevant for the pressure gradient, albeit these variations do affect the transmural pressure. A well-studied example is the autoregulation of glomerular blood flow via afferent and efferent arterioles of renal glomeruli. Thus, this alternative definition equates \"autoregulation\" with the deviation of an observed hydraulic conductance from an expected \"normal\" or \"standard\" value (usually the intrinsic conductance of the isolated hydraulic conductor, e.g. the glomerular capillaries). Consequently, a lumped series resistance (e.g. pre- or postglomerular sphincters) can explain and replicate this type of autoregulation. Moreover, autoregulation in this sense originates in the organ itself (e.g. the kidney) and can therefore be studied in an isolated organ. In quantitative terms, the addition of a second tube of identical dimensions, for instance, halves the volume flow at a given overall pressure difference. Put differently, the associated \"apparent hydraulic permeability\" Δφ~i~/ΔP of tube A is reduced to one half of its original value. Inasmuch as a given pressure change ΔP now results in a smaller change of volume flow as compared to the situation without series resistance, we can say that volume flow is now \"buffered\" against pressure changes. In terms of a buffered system, we represent this situation as ![](1742-4682-2-9-i23.gif) from which the buffering parameters follow immediately as ![](1742-4682-2-9-i24.gif) With an independent variable x having the dimension of a pressure and the corresponding two dependent variables y, z having the dimensions of a flow, the system is dimensionally heterogeneous, and this necessarily implies that it is also non-conservative or \"distorted\" (σ\'(x) ≠ 1). The distortion is a linear one because σ\'(x) = L~A~= constant (Buffering I[@B14]). Importantly, the buffering parameters can be computed only if the sigma function σ(x) is defined explicitly or implicitly. This function, the sum of τ(x) and β(x), specifies the response σ\'(x) of the system in the absence of buffering where β\'(x) = 0 and thus τ\'(x) = σ\'(x). In other words, one can talk meaningfully about buffering only if one is able to identify a reference state where buffering equals zero by definition, and to obtain a quantitative description of the system under these conditions. This step is crucial, but not necessarily trivial, particularly in dimensionally heterogeneous systems. The sigma function provides the clue to the quantitation of buffering in more complicated situations where the unbuffered response itself is non-linear (see paragraph on blood pressure variability buffering in[Additional file 4](#S4){ref-type="supplementary-material"}), or where the second output is a completely virtual, abstract quantity, such as in the context of systems and control theory ([Additional file 6](#S6){ref-type="supplementary-material"}). Even when such a reference state exists, it may be inaccessible experimentally. However, identification of a reference state may as well be impossible as a matter of principle, indicating that rigorous quantitation of buffering strength in this case is inherently impossible and the word \"buffering\" could then be used merely in a metaphorical way. One such example is blood pressure buffering in the sense of \"blood pressure variability buffering\"; [Additional file 4](#S4){ref-type="supplementary-material"} contains our criticism of this term. ### Non-linear buffering at linear pressure-flow relationship In a further modification of our model, we replace the second, rigid \"buffering\" tube by an elastic one, while retaining the first, rigid \"transfer\" tube with its constant hydraulic permeability L~A~. The pressure-flow relation becomes non-linear for both outputs y = τ(x) and z = β(x). Similarly, the buffering parameters t, b, T, and B become dependent on x and must be written as t(x), b(x), T(x), and B(x), respectively. Only in the complete absence of buffering, blood flow will be a linear function of perfusion pressure, implying that σ\'(x) = constant. Note that flow in an elastic tube depends not only on the pressure difference, but on the absolute pressure as well; therefore, it does make a difference whether the second tube is placed upstream or downstream to the first one. In principle, the serial arrangement of one rigid and one elastic tube with an appropriate pressure-conductance profile can reproduce all possible pressure-flow relationships. From a mechanistical point of view, however, the assumption that the unbuffered system should exhibit a perfectly linear pressure-flow relationship (modeled by a rigid tube) appears unrealistic. ### Non-linear pressure-flow relationship -- \"distortion\" Unlike the rigid tube in the foregoing example, the blood vessels of most organs exhibit a highly non-linear relation between pressure difference and volume flow, even upon complete inhibition of vasomotion or other active regulation of hydraulic permeability. Here, non-linearity does not mean \"buffering\". Rather, it reflects solely the passive-elastic properties of the vessels as determined by vessel architecture and material; an appropriate single elastic tube may replicate such a non-linear relationship. One may therefore posit the pressure-flow relationship observed under these conditions as the unbuffered system response. Flow is now the product of pressure difference and a hydraulic conductance that varies non-linearly with absolute pressure: φ~1~= ΔP × L(P). In general terms, independent from hydraulic quantities, the sigma function σ(x) is then a non-linear function σ(x). This means that the system responds non-linearly to changes of the independent variable even in the absence of buffering; this behavior was termed above \"non-linear distortion\". If there exists anything like a \"normal\", \"unbuffered\" response with respect to organ perfusion in response to blood pressure changes, then it may be expressed exactly by such a sigma function. Any modification of that normal response would then constitute \"buffering\"; in the present example, buffering can be brought about, for instance, by mechanisms such as contraction of smooth muscles in pre- or postcapillary sphincters. With an explicit specification of the unbuffered system response (in terms of the sigma function σ(x)), it is then straightforward to derive an explicit quantitative expression of the four buffering parameters from the observed buffered system response (given by the transfer function τ(x): ![](1742-4682-2-9-i25.gif) and ![](1742-4682-2-9-i26.gif) As in the preceding example, the coefficients b(x) and t(x) vary non-linearly with x. Taken together, the formal and general concept of buffering not only allows one to quantitate buffering action in conservative systems, but can be applied with similar rigor to dimensionally heterogeneous systems and to systems with a non-linear response in the unbuffered state. When one wishes to compare different systems in terms of these measures of buffering action, it becomes necessary to recall the distinction between \"normalization in x\" vs. \"normalization in y, z\" (Buffering I[@B14]): The extent of autoregulation in various organs can be compared either at similar pressures(either directly or upon \"normalization in x\" of the respective pressure-flow curves), or at similar volume flows(upon normalization in y, z). Both perspectives may make sense, and it is necessary to explicitly specify which one is used. The concept of \"non-conservative buffered systems\" can be applied analogously to other phenomena. For instance, [Additional file 4](#S4){ref-type="supplementary-material"} applies this concept to electric phenomena, leading among others to a quantitative measure of rectification. Time-dependent buffering of cytoplasmic Ca^++^ions -- The concept of \"muffling\" ================================================================================= The basic concept of \"muffling\" --------------------------------- The buffering parameters t, b, T, or B all describe the relation between the derivatives of two functions of a common independent variable. Invoking the paradigm of \"buffering\" therefore requires that the phenomenon in question indeed exhibits a reproducible, well-defined relationship between three variables. Furthermore, one must be able to identify and express that relationship in an explicit mathematical form -- namely, as an ordered combination of two functions or \"buffered system\". In practice, this usually means that the analysis is restricted to time-independent systems, or to the time-independent equilibrium states of a given time-dependent system. For instance, immediately following the addition of H^+^ions to a solution, ion concentrations will transiently change until a stable and characteristic end-point is reached with respect to the partitioning of total H^+^ions between water (\"free\") and other H^+^acceptors (\"bound\"). Acid-base chemistry is largely occupied with these stable end-points. On the other hand, the presence of buffers not only determines the position of the final equilibrium, but also affects the path and the speed with which this equilibrium is attained. Often enough, the details of these pre-steady state events are practically relevant. For instance, the particular shape of the free Ca^++^concentration transients in response to acute Ca^++^loads can modulate cell signalling in neurons. Strong Ca^++^buffering may protect from cell death under certain pathological conditions \[[@B11]\], but may as well cause or aggravate cell damage in other situations \[[@B5]\]. Clearly, having a quantitative measure of buffering action during the pre-steady state is of similar scientific interest as having such a measure for steady-state buffering. Obviously, the Ca^++^transients in live cells differ profoundly from those observed in pure water or saline. However, the Ca^++^transients also vary considerably from one cell type to another \[[@B5]\]. The two major mechanisms that shape such transients are binding (i.e., to ligands within the volume element under study) and transport (i.e., net flux across the boundaries of the volume element). Transport may be brought about by i)diffusion of free ions, ii)diffusion of ions bound to buffer molecules, and iii)channels or carriers that translocate free or bound ions across membranes. The combined effect of binding and transport on the transients elicited by an acute ion load has been termed \"muffling\" \[[@B2]\]. In many experimental systems, muffling appears to be a rather complex process, and the observed time courses may follow complicated, multiphasic patterns. Is it possible under such circumstances to find a general quantitative measure of \"muffling\"? Ideally, such a measure should again provide a ratio scale and be as general and rigorous as the scale used to quantitate buffering strength under equilibrium conditions. A ratio scale for muffling -------------------------- ### The response to a concentration step in a small volume element We consider a volume element Δx × Δy × Δz assumed to be homogenous with respect to the relevant ion concentrations (Figure [3A](#F3){ref-type="fig"}). Let this volume element contain a solvent and a specific solute, with some solute molecules being bound, others \"free\". The initial number of \"free\" molecules is denoted n~0~. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### [Time-dependent buffering: the \"muffling\" of Ca^++^ions]{.underline}. A, Muffling is brought about by binding and flux. The concentration of free Ca^++^ions in a small volume element (dx × dy × dz)is changed instantaneously by addition Δe~0~further free Ca^++^ions (red arrow). The \"error\" imposed by this acute Ca^++^load either persists, or it is reduced over time. Reduction (blue arrows)may occur via Ca^++^flux across the boundaries of the volume element (right arrow), or by binding of Ca^++^to Ca^++^buffers within the volume element (lower arrow);the time-dependent, combined effect of binding and flux is termed \"muffling\". B, Prototypes of muffling. From top to bottom: 1, Zero muffling in the absence of both flux and binding; 2, Muffling via flux across its boundaries (e.g. diffusion), without binding to buffers within the volume element; 3, Muffling via binding to buffers inside the volume element, without flux; 4, Muffling due to both binding and flux. Note that peak size and magnitude of muffling action are not correlated. C, Measures for time-dependent \"error reduction\", or \"muffling\". A measure that reflects both the size and the duration of the error e(t) is the integral ![](1742-4682-2-9-i27.gif)(red area). Similarly, a measure that reflects both the size and the duration of the \"error reduction\" or \"muffling\" m(t) is given by the integral ![](1742-4682-2-9-i28.gif)(blue area). The proportions between ε(t) and μ(t) can be used to define a \"muffling coefficient\" and \"muffling ratio\" (see main text of BufferingII). These measures are the time-dependent analogs of the time-independent \"buffering coefficient\" and \"buffering ratio\" (Buffering I). ::: ![](1742-4682-2-9-3) ::: Now, the system is acutely disturbed by adding (or removing) instantaneously a certain amount e~0~of the free molecules. Then, the number of free molecules will instantaneously jump from its baseline value n~0~to a value n(t) = n~0~+e~0.~If present, \"muffling\" according to the above definition (i.e., binding to buffer molecules within the volume element, or net flux across its boundaries) will tend to return n(t) towards n~0~over time and thus decrease the absolute deviation \\|e(t)\\| (these considerations are valid for addition as well as for removal of solute, and both e~0~and e(t) may thus have a positive or a negative sign). Depending on the relative contributions of binding vs. flux, three limiting cases can be distinguished (Figure [3B](#F3){ref-type="fig"}). #### A. Zero muffling (Figure [3B](#F3){ref-type="fig"}, top) If there is neither binding nor net flux, the initial disturbance e~0~will persist, and e(t) will remain constant at e(t) = e~0~. A measure that reflects both size and duration of the deviation e(t) is the integral of e(t) over time, which we term ε (Figure [3C](#F3){ref-type="fig"}): ![](1742-4682-2-9-i27.gif). In this case, ε(t) will increase linearly over time. The letters e and ε may be associated with \"error\" or \"extra ions\". #### B. Muffling via flux (Figure [3B](#F3){ref-type="fig"}, 2^nd^from top) If the solute exhibits net flux into adjacent volume elements, but no binding within the volume element itself, then some of the added solute molecules e~0~will leave the volume element and thus decrease e(t). Herein, the final equilibrium concentration of the solute depends on size and composition of the adjacent volume elements, and on the available active transport mechanisms. For instance, most cell types will sooner or later extrude a cytoplasmic Ca^++^load completely, with e(t) approaching zero over time. With efflux, the error integral ε(t) increases more slowly over time than without. The error e(t) = n(t)-n~0~is but one way of looking at such a muffling process. Alternatively, one can measure the extent by which the initial error e~0~is reduced over time; this \"error reduction\" is equal to the difference m(t) ≡ e~0~-e(t). Again, size and duration combined of this error reduction m(t) are faithfully reflected by an integral, i.e., the integral of m(t) over time, which we term μ (Figure [3C](#F3){ref-type="fig"}): ![](1742-4682-2-9-i29.gif) The letters m and μ may be associated with \"muffling\". #### C. Muffling via binding (Figure [3B](#F3){ref-type="fig"}, 3^rd^from top) In the absence of net flux (due to the lack of appropriate gradients or permeabilities), but with binding of solute within the volume element itself, the deviation e(t) of solute concentration from its initial value n~0~will be reduced to some equilibrium value between 0 and e~0~. Again, the situation can be described in terms of the error e(t), error reduction m(t), and their respective time-integrals ε(t) and μ(t). ### Definition of the \"muffling ratio\" M(t) These examples show that the magnitude of muffling depends on several aspects, including not only the size of the initial \"peak\" e~0~, but also the speed by which that disturbance is counteracted and the steady-state error e(t) that remains during the final equilibrium. As a quantitative measure M(t) of muffling at a given time, we can use the ratio of the error reduction integral μ(t) over the error integral ε(t) (Figure [3C](#F3){ref-type="fig"}): ![](1742-4682-2-9-i30.gif) This measure is formed analogously to the buffering ratio B. We term it \"muffling ratio\". From a chemically more realistic perspective, the states of the solute are statistical averages. The muffling ratio can thus be interpreted as the ratio of the average number ![](1742-4682-2-9-i31.gif) of added molecules that existed in a muffled state (i.e., bound or outside the volume element under study) within the particular time window, over the average number ![](1742-4682-2-9-i32.gif) of added molecules found free and inside the volume element during that time: ![](1742-4682-2-9-i33.gif) ### Properties of the muffling ratio M(t) M(t) always is a dimensionless number, and yields an absolute ratio scale for muffling strength. If an error e~0~persists fully, the muffling ratio assumes a value of zero. An error e~0~that is counteracted completely and instantaneously will yield infinite muffling ratio. Importantly, the specific value of M(t) in a given system usually depends on multiple conditions and parameters (initial solute concentration, magnitude and direction of the initial disturbance, integration time), with no a prioriconstraints on any of them. In order to be completely unambiguous, the relevant parameters should be indicated along with the muffling ratio in the form M(n~0~, e~0~, t). In other words, it is not possible to characterize muffling by a single characteristic parameter. In most cases, buffering and transport tend to decreasethe absolute error \\|e(t)\\| over time. Then, the muffling ratio M(t) always has a positive sign, even when ε(t) and μ(t) are both negative. On the other hand, meaningful negative values of M(t) can be obtained under certain conditions: For instance, muffling mechanisms may cause the absolute error \\|e(t)\\| to increase further beyond \\|e~0~\\|, or they may produce an \"undershoot\" or \"overshoot\". Finally, the instantaneous disturbance e~0~may elicit oscillations of n(t), either dampened or undampened ones. The value of M(t) as a function of time will then oscillate, too, and may include negative values. ### Buffering is a limiting case of muffling Importantly, our two concepts of buffering and muffling are linked conceptually and numerically. Recall that \"buffering\" refers to the proportion between the changes of two time-independent partitioning functions τ(x) and β(x) in response to an infinitesimal perturbation, and \"muffling\" to the proportion between two time-dependent functions ε(t) and μ(t) in response to a finite perturbation e~0~. Some (not necessarily all) muffled systems will travel over time towards a unique, time-independent equilibrium upon a particular disturbance e~0~. As demonstrated amply above, such time-independent equilibrium states can be described in terms of time-independent transfer and buffering functions, τ(x) and β(x), and therefore also by the four buffering parameters t, b, T, and B. The initial equilibrium state is given by {τ(x~0~), β(x~0~)}, and the equilibrium after imposing the disturbance e~0~by {τ(x~0~+e~0~), β(x~0~+e~0~)}. Comparing the equilibrium states before and after a finite disturbance e~0~, the transfer function τ(x) changes by an amount Δτ = τ(x~0~+e~0~) - τ(x~0~), and the buffering function β(x) by an amount Δβ = β(x~0~+e~0~) - β(x~0~). Herein, the two partial changes equal the initial disturbance, in keeping with the conservation condition: Δτ+Δβ = Δx = e~0~. The limit of the error integral ε(t) for infinitely long integration time is t × Δτ, and the corresponding limit of the error reduction integral μ(t) is equal to t × Δβ. Thus, the muffling ratio at infinite time is ![](1742-4682-2-9-i34.gif) The ratio Δβ/Δτ reflects the average buffering ratio ![](1742-4682-2-9-i35.gif) in the interval between x~0~and x~0~+Δx, but is different from the true buffering ratio B = dβ/dτ at x = x~0~. In a second limit process, we let the perturbation e~0~decrease progressively from a finite value towards zero. The ratio ![](1742-4682-2-9-i36.gif) will then approach the differential ![](1742-4682-2-9-i37.gif), i.e., the buffering ratio B(x): ![](1742-4682-2-9-i38.gif) Thus, for long integration times and small perturbations e~0~, the muffling ratio becomes equal to the buffering ratio. In other words, buffering is a special limiting case of muffling for t→∞ and e~0~→0. Indeed, these limiting conditions are fundamental in any attempt to determine time-independent buffering power rather than time-dependent muffling, whatever particular buffering unit is to be employed. For instance, determination of H^+^buffering power from the slope of the titration curve \[[@B12]\] requires that the solution be stirred well and for long enough to allow for complete equilibration (t→∞), and to use small amounts of titrant (e~0~→0). Note that \"muffling\" according to our definition provides an empirical measure extracted from concentration transients, but it does not imply any mechanistical assumptions, such as the distinction between binding and transport processes. Inasmuch as buffering is a limiting case of muffling, buffering may similarly comprise both binding and transport. Time-dependent and -independent responses to disturbances are shaped by the combined action of these two mechanisms, and it therefore seems appropriate that the units for buffering and muffling reflect these combined effects. Restricting buffering or muffling by definition to its binding component would leave us without a measure for the effects caused by net transport, and complicate experimental approaches without increasing biological validity. -- The properties of our muffling strength unit are discussed further and compared to previous approaches in the [Additional file 5](#S5){ref-type="supplementary-material"}. ### Muffling can be viewed as \"dynamic disturbance rejection\" In principle, muffling of a Ca^++^constitutes a specific form of \"dynamic disturbance rejection\", which is one aspect of \"control\". Analogously, the effect of H^+^buffers on steady-state pH can be viewed as \"static disturbance rejection\". It is straightforward to extend the buffering concept in order to obtain measures of \"systems level buffering strength\". Specifically, these measures allow one to quantitate the efficiency of \"setpoint tracking\" and of \"disturbance rejection\" by control systems, either static or dynamic. These measures have been sketched elsewhere \[[@B13]\]; a more systematic and comprehensive outline is found in [Additional file 6](#S6){ref-type="supplementary-material"}. Formulating the buffering concept in the language of systems and control theory conveys a tangible quantitative meaning to the term \"resistance to change\", the customary paraphrase of \"buffering\". Conclusion ========== When you can measure what you are talking about and express it in numbers, you know something about it. -Lord Kelvin In this article, we applied the formal and general concept of buffering presented in the preceding paper (Buffering I[@B14])to various types of buffering phenomena. The buffering of H^+^ions in solutions of a weak acid and in pure water could be analyzed simply in terms of \"conservative buffered systems\". The buffering of organ perfusion in the face of variable blood pressure made it necessary to invoke the concept of \"non-conservative\", \"dimensionally heterogeneous\" systems. Describing the response of a cell to an acute Ca^++^ion load could be achieved with a novel quantitative measure of the time-dependent aspects of buffering, also termed \"muffling\". Muffling is equivalent to what is called \"dynamic disturbance rejection\" in systems and control theory, and our general concept of buffering yielded further quantitative measures of control performance. These measures allow one to describe all major aspects of control, namely static or dynamic ones, and disturbance rejection as well as setpoint tracking. Most buffering phenomena may be interpreted in terms of the control paradigm ---------------------------------------------------------------------------- Control systems may exhibit complex buffering behavior, and their description then requires all four measures of control efficiency. Conversly, the control paradigm provides, in the form of these measures of control efficiency, a sufficient framework for the quantitation of allbuffering phenomena, including phenomena that are not routinely viewed in terms of control systems. For instance, the addition of H^+^ions to an aqueous solution can be interpreted as \"disturbance\", and the \"buffering\" of H^+^ions as \"static disturbance rejection\". Both \"buffering\" and \"static disturbance rejection\" can be quantitated in terms of the buffering ratio B(x), which implies that both are just different words for the same thing. Similarly, the addition of Ca^++^ions to the cytoplasm may be interpreted as a \"disturbance\", and the extent to which this Ca^++^load is counteracted in a given time window is reflected in the \"dynamic disturbance rejection ratio\" M(t). It is a very common finding that buffering phenomena in a biological context serve apparent homeostatic or control purposes. Then, \"buffering strength\" is directly proportional to \"control efficiency\". In fact, they are the same thing, and our concept of buffering provides the corresponding unifying formal framework. In contrast, it is impossible to establish any systematic connection between Van Slyke\'s buffering strength unit (which is expressed in terms of \"moles per liter\" \[[@B12]\]) and any known measure of control efficiency. The abstract definition of the buffering concept is compatible with multiple, different interpretations ------------------------------------------------------------------------------------------------------- In many cases, the systems and control paradigm thus provides an intuitive and fruitful interpretation of the static and dynamic buffering parameters. For instance, such an interpretation is widely applicable in the areas of physiology and systems biology. Nonetheless, one should maintain the distinction between the original abstract definitions (stated in Buffering1 -- main text[@B14] and in axiomatic form in Buffering1[@B14]-- Additional file 7) and their subsequent interpretations. \"Control\" is but one out of several possible interpretations of these measures. \"Probability\" would be another natural interpretation, given that these measures were based in our axiomatic approach on a \"signed probability measure\" and were formulated as a \"non-Kolmogorov axiomatic systems of probability\" (Buffering I[@B14]-- Additional file 7). The ambiguous relation between axioms and their interpretations is a general finding; completely analogous to the buffering parameters discussed here, Kolmogorov\'s probability axioms have multiple interpretations, e.g. as relative frequencies, geometric probabilities, degree of individual belief, propensities etc The axioms are logically consistent, but their various interpretations may be (and usually are) in mutual logical conflict. Moreover, no single interpretation allows comprehensive treatment of all phenomena encountered by researchers. Maintaining the distinction between axioms and interpretation of axioms prevents spurious conflicts between interpretations, and allows one to use the axioms in a given situation flexibly and in the most appropriate way. In the words of Bertrand Russel: \"It must be understood that there is here no question of truth or falsehood. Any concept which satisfies the axioms may be taken to bemathematical probability. In fact, it might be desirable to adopt one interpretation in one context, and another in another.\" (\"Human Knowledge\", 1948)* Taken together, we have established a formal and general concept for the quantitation of buffering action, demonstrated the practical usefulness of that concept in a variety of contexts, and provided explicit descriptions of several important buffering phenomena. The formal and general concept is of theoretical interest with its underlying \"non-Kolmogorov\" axiomatic system of probability that is based on a non-Boolean bag algebra and accomodates negative as well as variable probabilities. On the practical level, our concept solves the problem of quantitating buffering action, and provides a common scientific language for a common quantitative pattern that is present in very diverse phenomena, and in many disciplines. Supplementary Material ====================== ::: {.caption} ###### Additional File 2 H^+^Buffering in Pure Water ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 3 Other Conservative Buffered Systems ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 5 Notes on Time-Dependent Buffering or \"Muffling\" ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 6 Buffering and Muffling in Systems and Control Theory ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 1 Further Notes on Langmuir Buffering ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 4 Other Non-Conservative Buffered Systems ::: ::: {.caption} ###### Click here for file :::	PubMed Central	2024-06-05T03:55:55.620929	2005-3-16	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1079954/", "journal": "Theor Biol Med Model. 2005 Mar 16; 2:9", "authors": [ { "first": "Bernhard M", "last": "Schmitt" } ] }
PMC1079955	Diabetes mellitus is a hypercoagulable state \[[@B1]\]. It has been demonstrated that patients with diabetes have abnormally elevated markers of coagulation activation \[[@B2]-[@B12]\], as well as increased risk of mortality from thrombosis \[[@B13]\]. Indeed, virtually all Type II diabetics will die a thrombotic death, whether its origin be cardiac, cerebrovascular or peripheral vascular \[[@B14]\]. What is less clear is how the presence of renal dysfunction affects the hypercoagulable state associated with diabetes. This is an important question; given the fact that renal dysfunction is so prevalent in patients with diabetes \[[@B15]\]. There is clinical and laboratory evidence that uremia associated with renal dysfunction leads to platelet dysfunction and altered hemostasis. \[[@B16]-[@B20]\]. However, we are unaware of any reports that suggest thrombotic risk decreases in diabetics who develop renal dysfunction. Our laboratory has developed a novel whole blood monitoring assay that determines thrombin generation, platelet function and clot structure on a single sample of patient blood. This assay provides a global assessment of clotting function by reporting the markers platelet contractile force (PCF), clot elastic modulus (CEM) and the onset of thrombin generation as measured by thrombin generation time (TGT). PCF is the force produced by platelets during clot retraction, and is a novel measure of platelet function during clotting; CEM reflects the structural integrity of the blood clot; and TGT is the speed at which thrombin is produced in whole blood \[[@B21]\]. Changes in these parameters provide a detailed characterization of the hemostatic effect of various disease states \[[@B22],[@B23]\], antithrombotic and anticoagulant drugs \[[@B21],[@B24],[@B25]\], and hemostatic agents \[[@B26]\] on platelet function and thrombin generation. We report here findings of a hypothesis-generating sub-group analysis that illustrate the effects of diabetes and renal dysfunction on platelet function and thrombin generation time. In addition, we examined the potential impacts of these disease states on the response to the commonly used low-molecular weight heparin enoxaparin. Patients and Methods ==================== This was a hypothesis-generating sub-group analysis of a prospective, open-label, ex vivostudy of enoxaparin in patients with and without renal dysfunction \[[@B27]\]. The Virginia Commonwealth University (VCU) Institutional Review Board approved this study prior to subject enrollment, and this study was conducted in compliance with the Declaration of Helsinki. In this study, a total of 30 subjects were evaluated: 10 healthy controls; 13 subjects with renal dysfunction but without diabetes; and seven patients with concomitant diabetes and renal dysfunction (6 CKD and 1 ESRD). All subjects were anticoagulant naïve, non-thrombosed, and otherwise healthy. Subjects were admitted into this study if they were \> 18 years of age and provided written informed consent. In addition, subjects with ESRD must have received maintenance hemodialysis for at least 3 months through an arterio-venous fistula or graft. Subjects were excluded from this investigation if they had any of the following: active bleeding or thrombotic disorder; pregnancy; recent trauma or surgery (\< 7 days); cirrhosis or other liver abnormality; hematocrit \< 30%; active cancer; had received a blood transfusion within 1 week of study enrollment; thrombocytopenia (platelets \< 100,000/mL); documented history of antithrombin III, protein C or protein S deficiency; concurrent use of anticoagulant or antiplatelet drug therapy. All subjects who provided written informed consent and who met the above criteria underwent screening evaluation, which consisted of a routine physical exam and laboratory evaluation (basic metabolic panel, complete blood count (CBC), international normalized ratio (INR), activated partial thromboplastin time (aPTT), prothrombin time (PT)). Study Procedures ---------------- For each subject, blood was collected via aseptic venipuncture into four evacuated tubes containing 3.2% sodium citrate. The blood was pooled and then aliquotted into five separate samples, which were spiked with increasing enoxaparin antifactor Xa activity concentrations. Final enoxaparin antifactor Xa concentrations in the respective aliquots were 0.0, 0.25, 0.50, 1.0, and 3.0 IU/mL. For each aliquotted blood sample, analysis was performed to determine the corresponding TGT, PCF and CEM at each antifactor Xa activity concentration. Specimen Processing and Analysis -------------------------------- Baseline chemistries (basic metabolic profile, CBC, PT, aPTT, INR) and antifactor Xa concentrations were processed and analyzed at the VCU Health System Department of Pathology. Plasma enoxaparin antifactor Xa activity was measured using a validated, commercially available chromogenic method (STA^®^heparin colorimetric analyzer, Diagnostica Stago, Parsippany, NJ, USA), and the results expressed as IU/mL. The lower limit of antifactor Xa detection for this assay was 0.05 IU/mL; the coefficient of variation (CV) was ± 3%. Whole blood samples were analyzed for TGT, PCF and CEM at the VCU Coagulation Special Studies Laboratory. The TGT, PCF and CEM for each aliquotted whole blood sample were simultaneously measured using the Hemodyne Hemostasis Analysis System™ (Hemodyne, Inc., Richmond, VA, USA) using a previously published, validated method \[[@B27]\]. All samples were run in duplicate. The CV for this assay was ± 7%. Statistical Analysis -------------------- All statistical analyses were performed using JMP statistical software version 5.1 (SAS Institute, Cary, NC, USA). The data were presented based on subject group assignment, as Group I: Control; Group II: subjects with renal dysfunction but without diabetes; and Group III: subjects with concomitant diabetes and renal dysfunction. Descriptive statistics characterized the group demographic data One-way analysis of variance (ANOVA) assessed for intergroup differences in demographic data, and mean PCF, CEM and TGT at baseline and at each spiked enoxaparin antifactor Xa activity concentration for each of the three groups. If intergroup differences were found to be statistically significant, a post-hoc Tukey-Kramer test was used to differentiate which groups were statistically different. The level of significance for all statistical tests was p \< 0.05. Results ======= Thirty subjects completed this study as described above. There were no adverse reactions or dropouts from the study. Table [1](#T1){ref-type="table"} details the subject demographics. Other than differences in the racial makeup of the groups, the groups were similar with respect to age, sex, weight and presence of coronary artery disease (CAD). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Baseline Subject Demographics ::: Group I Group II Group III ------------------------ ------------- -------------- --------------- --------- Age\[mean (S.D.)\] 40.5 (9.6) 46.5 (15.0) 45.1 (11.1) NS Race\[number (%)\] Caucasian 5 (50) 1 (8) 0 Black 3 (30) 12 (92) 6 (86) 0.02^§^ Other 2 (20) 0 1 (14) Sex\[number (%)\] Male 5 (50) 7 (54) 3 (43) NS Female 5 (50) 6 (46) 4 (57) Weight \[mean kg (S.D.)\] 75.0 (13.4) 83.9 (23.1) 86.0 (25.0) NS CAD\[number (%)\] Yes 1 (10) 2 (15) 3 (43) NS No 9 (90) 11 (85) 4 (57) NS -- Not significant § Chi-square test for group homogeneity significantly different ::: Baseline Platelet Function, Clot Structure and Thrombin Generation Time ----------------------------------------------------------------------- The top panel of Table [2](#T2){ref-type="table"} details the baseline PCF, CEM and TGT in Groups I, II and III. Group III (concomitant diabetes and renal dysfunction) exhibited significantly greater platelet activity, as measured by PCF (p = 0.003) and CEM (p = 0.03), relative to the non-diabetic groups I and II. The reported TGT was not significantly different between groups at baseline. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Mean (S.D.) Platelet Contractile Force, Clot Elastic Modulus and Thrombin Generation Time at Baseline and Following Increasing Enoxaparin Antifactor Xa Activity ::: Group I Group II Group III ---------------- --------------- ---------------- --------------- ----------- Baseline PCF^†^ 8.5 (1.4) 9.8 (1.5) 12.0 (2.9)\* 0.003 CEM^‡^ 23.2 (5.4) 29.0 (9.5) 37.3 (15.5)\* 0.03 TGT^§^ 222.0 (60.3) 235.4 (53.2) 270.0 (52.0) NS 0.25 IU/mL PCF 7.0 (1.8) 7.5 (1.6) 8.8 (2.9) NS CEM 21.1 (4.1) 22.6 (4.1) 24.6 (7.1) NS TGT 273.0 (68.5) 327.7 (70.8) 325.7 (76.3) NS 0.50 IU/mL PCF 5.5 (1.9) 5.6 (1.8) 6.3 (2.4) NS CEM 17.2 (5.1) 19.1 (4.4) 19.4 (5.0) NS TGT 318.0 (68.1) 440.0 (97.5)\* 420.0 (86.6) 0.007 1.0 IU/mL PCF 3.6 (1.7) 3.0 (1.7) 4.0 (2.1) NS CEM 11.9 (4.5) 10.0 (4.6) 12.9 (6.1) NS TGT 423.0 (128.4) 588.5 (96.0)\* 552.9 (114.7) 0.005 3.0 IU/mL PCF 1.5 (1.3) 0.6 (0.2) 1.0 (0.9) 0.06 CEM 4.8 (4.4)\* 0.8 (0.8) 0.9 (0.8) 0.003 TGT 780.0 (317.5) 1200.0 (0)\* 1200.0 (0)\* \< 0.0001 ^†^PCF -- Kdynes, ^‡^CEM -- Kdynes/cm^2^,^§^TGT -- seconds NS -- not significant \* Statistically significant ::: Comparative Platelet Function, Clot Structure and Thrombin Generation Times Following Increasing Enoxaparin Antifactor Xa Activity Concentrations ------------------------------------------------------------------------------------------------------------------------------------------------- As the spiked ex vivoenoxaparin antifactor Xa activity concentration increased, there were corresponding changes in PCF, CEM and TGT (Table [2](#T2){ref-type="table"}). As expected, there was an inverse relationship between increasing enoxaparin antifactor Xa activity and platelet function (i.e., PCF) and clot structure (i.e., CEM) in all three groups. Conversely, there was a direct relationship between TGT prolongation and increasing enoxaparin antifactor Xa activity in the groups. The PCF and CEM were not statistically different between groups at antifactor Xa activity concentrations of 0.25, 0.5 or 1.0 IU/mL. However, when the antifactor Xa activity concentration reached 3.0 IU/mL, Group I (controls) had significantly higher CEM (p = 0.003) relative to the two groups with renal dysfunction. There was also a trend of greater PCF in Group I relative to the other groups. The TGT was significantly prolonged in Group II (renal dysfunction without diabetes) relative to Group I (controls) at enoxaparin antifactor Xa activity concentrations of 0.5 (p = 0.07), 1.0 (p = 0.005) and 3.0 IU/mL (p \< 0.0001), respectively. There were no statistical differences between Groups II and III (concomitant diabetes and renal dysfunction) at these antifactor Xa activity concentrations. Discussion ========== This was a hypothesis-generating sub-group analysis performed to assess the pro-coagulant effect of diabetes on the platelet function in patients with CKD. To our knowledge, this is the first study to demonstrate that at baseline, diabetic patients with renal dysfunction have elevated platelet activity. Our review of the literature fails to identify prior studies of platelet function specifically performed in diabetics with renal dysfunction. Given the apparently divergent effects of hyperglycemia and uremia on platelet function, one might hypothesize that depressed platelet function due to uremia may be partially or completely reversed by the enhanced effects of diabetes. At baseline, this appears to be the case. The presence of diabetes increases platelet forces, vis-à-vis PCF and results in greater clot strength, vis-à-vis CEM, compared to the other non-diabetic groups I and II. However, despite the increased platelet activity in diabetics with renal dysfunction noted at baseline, there were no differences in TGT between groups. Previous studies of platelet function in diabetes or uremia have primarily involved measurement of platelet aggregation. Enhanced platelet aggregation has been demonstrated with diabetes \[[@B3],[@B29],[@B30]\] and decreased platelet aggregation has been identified in uremia \[[@B31]-[@B35]\]. Previous studies of PCF noted enhanced function in diabetes \[[@B23]\], but normal values in chronic uremia \[[@B36]\]. These studies are consistent with the current findings. The divergence between aggregation results and PCF findings centers around the fact that one assay (platelet aggregation) is performed with minimal platelet activation and the other (PCF) under conditions of maximal platelet activation (i.e., platelets in the presence of thrombin). Thrombin is such a profound agonist of platelet function that subtle influences noted by platelet aggregation studies are simply overwhelmed. An example of this effect can be seen with the effects of IIb/IIIa blockade on platelet aggregation and PCF. The concentrations of IIb/IIIa blocking agents (e.g., abciximab, tirofiban) required to block PCF production are an order of magnitude higher than those required to suppress platelet aggregation \[[@B24]\]. The data from this analysis, although limited by the relatively small sample size, suggests that there is a \"counter-balancing\" effect of renal dysfunction and diabetes with respect to platelet function and thrombin generation time. This is borne out in Table [2](#T2){ref-type="table"}. For example, subjects in Group II (renal dysfunction but without diabetes) have the greatest prolongations in TGT, compared to the other groups, following increasing antifactor Xa activity. However, Group III (renal dysfunction with diabetes) subjects tend to have a blunted TGT prolongation, though not significantly different compared to Group II. One could argue that given a larger sample size, potential differences may be detected between Groups II and III. Our findings should at least caution against assuming the presence of an \"auto\" anticoagulant or \"auto\" antiplatelet effect of uremia in patients with concomitant diabetes. There are important limitations to these findings. First, this study was a sub-group analysis not designed a priorito detect these differences. Indeed, these differences were detected on data analysis and thought to be important hypothesis-generating data. Secondly, the groups presented were quite heterogeneous, both in terms of numbers and underlying disease states. That is, there were control patients with neither diabetes nor renal dysfunction, there were patients with solely renal dysfunction but no diabetes, and there were patients with concomitant diabetes and renal dysfunction. A more appropriate way to study the effects of diabetes and renal dysfunction on platelet activity and thrombin generation time is prospectively using four study groups: 1) non-diabetics without CKD; 2) non-diabetics with CKD; 3) diabetics without CKD; and 4) diabetics with CKD. This study design would more appropriately describe the effect of diabetes on uremic platelets. Moreover, this would allow for greater statistically analysis using a two-way ANOVA test, that could detect an interaction between the concomitant disease states on altered platelet function. Nevertheless, these data are compelling enough to develop a hypothesis for further study. Conclusion ========== In conclusion, we report hypothesis-generating findings that suggest that at baseline subjects with concomitant diabetes and renal dysfunction have increased platelet activity relative to healthy controls and patients with renal dysfunction but no concomitant diabetes. This may suggest that the presence of uremia does not completely ameliorate the hypercoagulable effect of diabetes. Secondly, it appears that subjects with renal dysfunction but without concomitant diabetes have an enhanced response to enoxaparin compared to the controls, as measured by TGT. Further large-scale studies are warranted to further characterize the effects and interactions of diabetes and renal dysfunction on platelet function. Abbreviations ============= PCF, platelet contractile force; CEM, clot elastic modulus, TGT, thrombin generation time, CKD, chronic kidney disease; ESRD, end-stage renal disease; GFR, glomerular filtration rate; VCU, Virginia Commonwealth University; LMWH, low-molecular weight heparin; CBC, complete blood count; INR, international normalized ratio; aPTT, activated partial thromboplastin time; PT, prothrombin time, CAD, coronary artery disease; ANOVA, analysis of variance. Competing Interests =================== There are no financial or non-financial competing interests related to this manuscript. Authors\' Contributions ======================= DFB was the principal investigator for the study and was responsible for all aspects of its conduct. EJM was the laboratory technologist who performed all analytical assays. TWBG was responsible for recruitment of dialysis subjects and study design. MEC was responsible for overseeing all analytical activities, as well as study design and data analysis. Acknowledgements ================ Supported in part by National Kidney Foundation of the Virginias Grant \#NKFVA018. Presented in abstract form at the 2^nd^International Congress on Clinical Pharmacy, Paris, France, April 28, 2004.	PubMed Central	2024-06-05T03:55:55.626902	2005-3-29	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1079955/", "journal": "Thromb J. 2005 Mar 29; 3:3", "authors": [ { "first": "Donald F", "last": "Brophy" }, { "first": "rika J", "last": "Martin" }, { "first": "Todd WB", "last": "Gehr" }, { "first": "Marcus E", "last": "Carr" } ] }
PMC1079956	Introduction ============ In order to improve the unsatisfying results of current standard approaches to breast cancer treatment, therapeutic strategies tailored to individual patients\' tumor phenotype based on better understanding of breast cancer biology should be considered. For instance, the human epidermal growth factor receptor-2 (HER-2/ErbB-2) is over-expressed in 25--30% of breast and ovarian cancers and has been broadly utilized as a target for passive immunotherapeutical interventions \[[@B1]\] in particular with the humanized anti-HER-2 monoclonal antibody trastuzumab (Herceptin) \[[@B2]\]. The mechanisms of anti-tumor activity of Trastuzumab are complex and still not fully understood. Trastuzumab induces rapid disappearence of HER-2 from the cell surface, thereby reducing heterodimer formation critical for the accumulation of the cyclin-dependent kinase inhibitor p27 resulting in cell cycle arrest \[[@B3]\]. It also appears that Trastuzumab suppresses VEGF expression, inhibits heregulin-mediated angiogenesis both in vitroand in vivo, reverses cytokine resistance, and restores E-cadherin expression \[[@B4]\]. Trastuzumab also inhibits constitutive HER-2 cleavage/shedding by metalloproteases and, consequently, the generation of phosphorylated p95 \[[@B3],[@B5],[@B6]\]. Since Trastuzumab contains a human immunoglobulin G1 (IgG1) Fc region, it is possible that the antibody may participate in complement mediated cytotoxicity (CMC) and/or antibody dependent cell-mediated cytotoxicity (ADCC). However, Trastuzumab cytotoxicity by CMC could not be documented experimentally \[[@B7]\], possibly because of the presence of membrane-bound complement regulatory proteins (mCRP) such as decay accelerating factor (DAF; CD55), membrane cofactor protein (MCP; CD46), or protectin (CD59) on the surface of breast carcinoma cells. However, its cytotoxicity through ADCC has been confirmed \[[@B5],[@B6],[@B8]\]. Herceptin-dependent cell-mediated cytotoxicity action could depend upon activation of FcγRI (CD64) in monocytes and dendritic cells and/or activation of FcγRIII, (CD16) in monocytes/macrofages and natural killer cells, which may release cytotoxic granules such as perforin and granzymes to destroy target tumor cells. Suppression of Herceptin cytotoxicity could be mediated through FcγRII (CD32), constitutively expressed on monocytes, B cells, platelets, dendritic cells, eosinofils, basophiles and neutrophils. It is observed that Trastuzumab is particularly effective in patients with strong (immunohistochemistry score, IHC score = +3) overexpression of the HER-2 receptor, or medium over-expression (IHC score +2) \[[@B9]\]. However, passive immunotherapy of the breast cancer with Herceptin appears promising; it remains unclear why some patients with HER-2 positive tumors do not adequately respond to therapy. Therefore, the aim of this study was to identify in vitro, immune parameters, in the patients with breast carcinoma that may be relevant to anti-tumor immune modulation, and to estimate in vitrothe capability of Herceptin to enhance anti-tumor activity in patients with breast cancer and normal controls. Patients, material and methods ============================== Fourteen HER-2 positive advanced breast cancer patients were included in the study. All patients had histologically proven invasive breast cancer, diagnosed in the metastatic stage of the disease immediately before the inclusion (1 patient), or earlier in the operable clinical stages (13 patients). As the consequence of previous cytotoxic and/or endocrine treatment, most of the patients were postmenopausal at study entry. However, patients had been without any systemic endocrine or cytotoxic therapy at least one month before the serum sampling. All patients had visceral involvement, mostly including the liver metastases. All patients had HER-2 (3+) positive primary tumors according to immunohistochemical examination using DAKO HercepTest on paraffin embedded breast tumor specimens. At the time of inclusion, they were screened for randomization into the clinical study of chemotherapy with or without trastuzumab. Their characteristics are presented on Table [1](#T1){ref-type="table"}. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Patients\' characteristics ::: ---------------------------------------------------------------------- ----------------------------------- -------------------------- -------- Age at study entry range 36--69 (yrs) median 52.5 Menstrual status premenopausal 1 (No. of pts) postmenopausal 13 Previous treatment None 1 (No. of pts) Surgery radical (breast) 13 biopsy only 1 liver metastasectomy 2 Irradiation -- postoperative 4 Systemic therapy adjuvant 11 adj./neo-adj. chemoth 9 adj. endocrine therapy 5 for metastatic disease 7 Min. time elapsed from discontinuation of previous treatments (days) 28 DFI (months) range 0--144 median 26 Histology type of primary tumor IDC 8 (No. of pts) ILC 2 IC 4 Histology grade of primary tumor 1 0 (No. of pts) 2 8 3 2 unknown 4 SR status Negative (ER neg. and PR neg.) 4 (No. of pts) Positive (ER pos. and/or PR pos.) 10 HER-2 status Positive (3+, ICH) 14 (No. of pts) Clinical stage of BC at study entry Stage IV 14 1^st^disease relapse 7 2^nd^disease relapse 5 3^rd^disease relapse 2 Metastatic involvement Visceral 14 (No. of pts) liver involvement 11 other organs involvement 8 ---------------------------------------------------------------------- ----------------------------------- -------------------------- -------- BC = Breast cancer; DFI = Disease-free interval; IDC = Invasive ductal carcinoma; ILC = Invasive lobular carcinoma; IC = Invasive carcinoma of unknown histology type; SR = Steroid receptors; ER = Estrogen receptor; PR = Progesterone receptor; HER-2 = Human epidermal growth factor receptor type 2. ::: The study was approved by Institutional Ethics Committee. All patients signed the written Informed Consent. Twenty healthy donors, age range: 20--55 years, served as controls. Peripheral blood mononuclear cells (PBMC) cytotoxicity and PBMC and Herceptin-dependent cytotoxicity were assessed indirectly, through determination of target, HER-2 positive (3+), breast adenocarcinoma MDA-MB-361 cell survival, by 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide) MTT test \[[@B10]\]. Twenty thousand target MDA-MB-361 cells (T) were mixed with effectors PBMC (E), at effectors to target (E:T) ratios of 0:1 and 5:1 for 48 h, in RPMI 1640 medium with 10% of control serum (pool of healthy sera), or these target cells were incubated in RPMI 1640 medium with 10% autologous serum and PBMC at E:T ratios: 0:1, 1.25:1, 2.5:1 and 5:1 with or without 21 μg/ml of Herceptin. After 48 h 20 μl of MTT solution (5 mg / ml phosphate buffered saline) were added and 4 h later 100 μl of 10% sodium dodecyl sulfate were added. The absorbance at 570nm of MTT stained wells were read at 570 nm 24 h latter. To get the cell survival, the absorbance at 570 nm of MTT stained target cells (A) incubated : • in nutrient medium with autologuous serum (Na), • or in nutrient medium with autologuous serum plus efector cells (Na+E), • or in nutrient medium with control serum (Nc), • or in nutrient medium with control serum plus efector cells (Nc+E), was always devided with the absorbance of control cells (Ac) i.e., target cells incubated in nutrient medium with control serum (Nc) S(%) = A/Ac It was implied that absorbance of corresponding blank AB was always substracted from A of samples, or from A of controls. Therefore final equation was: S (%) = (A (~T+Na+E~)-A~(Na+E)~) / (A~(T+Nc)~-A~Nc~) Percent of CD16^+^cells was determined by direct immunofluorescence assay using the anti-CD16 FITC monoclonal antibodies according to manufacturer\'s instructions (Becton Dickinson, San Jose, Ca, USA). Data collection and analysis were done on FacsCalibur™ flow cytometer (BDIS), using the CellQuest software (BDIS). Statistical analysis was done using two-tailed Student\'s ttest with a minima threshold for significance of P\< 0.05). Results ======= Serum isolated from healthy controls did not alter MDA-MB-361 cell growth compared with pooled serum. Sera from patients did not significantly suppressed the growth of target tumor cells (P = 0,05006) in comparison to their growth in pooled serum The survival of target cells grown in nutrient medium with serum of control healthy people or with serum of patients in the presence of 21 μg/ml of Herceptin was also not significantly affected (Figure [1](#F1){ref-type="fig"}). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### The mean values of the survival (S%) of adenocarcinoma MDA-MB-361 cells, grown for 48 h in nutrient medium with 10% serum from control healthy people, or from breast carcinoma patients with, or without Herceptin (21 μg/ml). ::: ![](1479-5876-3-13-1) ::: PBMC from healthy donors manifested a non-significant trend toward higher cytotoxicity compared to PBMC from cancer patients when control pooled serum was used (as seen on Figure [2](#F2){ref-type="fig"}). Upon replacement of control serum with autologous serum, the cytotoxic activity of patients\' PBMC was higher but not significantly (P= 0.09) than that of controls PBMC (Figure [2](#F2){ref-type="fig"}). Similarly, patients\' PBMC activity was statistically enhanced (P\< 0.003) by the presence of autologous serum compared to their action in pooled control serum (Figure [3](#F3){ref-type="fig"}). These data suggest that autologous serum of patients with breast cancer may contain a soluble facilitator of PBMC-mediated cytotoxicity, which is specific for each individual patient. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### The mean values of survival (S%) of MDA-MB-361 cells co-cultured for 48 h, with effectors, PBMC isolated from control healthy people and from breast carcinoma patients, in nutrient medium with 10% of control, or autologous serum. E:T ratio was 5:1. ::: ![](1479-5876-3-13-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### The mean values of the survival (S%) of MDA-MB-361 cells co-cultured for 48 h, with effectors PBMC, isolated from control healthy people and from the breast carcinoma patients, in nutrient medium with 10% of control, or autologous serum. E:T ratio was 5:1 (\*P\< 0.003). ::: ![](1479-5876-3-13-3) ::: This anti-tumor humoral immunity present in autologous serum, which contributed to the enhancement (≥ 30%) of PBMC cytotoxicity, was seen in 5/14 HER-2 positive breast carcinoma patients, and in 4/13 healthy control subjects Therefore it seems that this is not a breast cancer patient specific phenomenon. The addition of Herceptin to PBMC from controls and from patients with breast cancer, significantly enhanced suppression of target cell survival at the E:T ratio 5:1 (P\< 0.02 and P\< 0.004 respectively) (Figure [4](#F4){ref-type="fig"}). This means that patients PBMC from breast cancer patients have not loosed their ability to kill malignant cells through Herceptin. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Herceptin in the concentration of 21 μg/ml in the incubation mixture of PBMC (from control healthy people and from breast carcinoma patients) and target cells decreased MDA-MB-361 cell survival (S%), in medium with 10% of autologous serum. E:T ratio was 5:1 (\*P\< 0.02, \\P\< 0.004). ::: ![](1479-5876-3-13-4) ::: There were no statistically significant differences between the percentage of CD16^+^cells in patients and controls. The clinical significance of the level of Herceptin dependent cytotoxicity in some patients treated with Herceptin is not available at the moment. Discussion ========== Addition of targeted HER-2-specific passive immune therapy for the treatment of patients with breast cancer may complement conventional therapy \[[@B11]\]. However, some patients do not successfully respond to such treatment. As Herceptin is primarily an immunological agent the potential of Herceptin to enhance anti-cancer immune responses in vitromight guide the selection of patients for Herceptin therapy. Our data suggest that MDA-MB-361 cells, over-expressing HER-2 are not significantly susceptible to direct Herceptin anti-proliferative action. This can be due to the loss of tumor suppressor PTEN by MDA-MB-361 cells which plays a central role in the transmission of growth inhibitory signals induced by Herceptin \[[@B12]\]. In addition, it appears according to our observations that PBMC from patients with breast cancer have non-significantly lower cytotoxic potential against MDA-MB-361 cells compared to PBMC isolated from healthy donors. However, patients PBMC in the presence of autologous serum significantly (P\< 0.003) enhanced their immune reactivity possibly because of spontaneous humoral HER-2 anti-tumor immunity. These findings are in accordance with Disis et al. \[[@B13]\] reporting the existence of humoral immunity in some patients with breast cancers. Unexpectedly, the anti-tumor humoral immunity, which leads to decreased malignant cell survival was also found in a minority of healthy people. The significance of the existence of an effective anti-tumor immunity in vitroin some patients with breast carcinoma which is not effective in destroying tumors in vivoremains unclear. Possibly, some proteins may inhibit the lytic action of perforin and granzymes. One of such proteins could be calreticulin that exerts several inhibitory actions including the inhibition of C1q-dependent complement activation \[[@B14]\]. This protein, together with perforin, is present in the cytolytic granule of NK and cytotoxic T cells protecting them from self destruction by perforin \[[@B15]\]. It has been suggested that calreticulin stabilizes membranes and indirectly prevents polyperforin pore formation, therefore, inhibiting target cell susceptibility to effector cell killing \[[@B16],[@B17]\]. The over-expression of serpins by some tumor cells could also inhibit granzim A or granzyme B lytic activity too \[[@B18],[@B19]\]. Results obtained in this work point to the new experiments aiming to determine the utility of in vitroimmunological tests for the prediction of patients response to immunotherapy; i.e. the relevance of different levels of calreticulin and/or serpins in target cells used in vitroand in cells from malignant tumor specimens in immunological tumor escape needs to be determined. Absence of statistically significant difference between percentage of CD16^+^cells in patients and in controls suggest that ADCC mediated by NK cells is likely preserved in patients with breast carcinoma. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Survival of target tumor MDA/MB-361 cells incubated with breast cancer patients PBMC in nutrient medium with 10% pooled serum (Nc) or with 10% of autologous serum (Na) without or with 21 μg/ml Herceptin (Na+H). Effectors (E) to target tumor cells (T) ratio was 5:1. ::: Patient S ~N~(%) S ~Na~(%) S ~Na+H~(%) CD16 (%) --------- ---------- ----------- ------------- ---------- 1 73 30 21 32.02 2 42 39 13 18.98 3 73 69 30 16.68 4 49 27 24 25.76 5 87 23 12 ND 6 69 62 49 ND 7 55 33 5 28.00 8 53 11 8 25.32 9 49 40 11 5.47 10 51 57 33 4.84 11 24 34 13 ND 12 70 32 10 ND 13 40 35 19 11.81 14 74 14 9 25.59 ND -- non-determined ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Survival of target tumor MDA/MB-361 cells incubated with healthy controls PBMC in nutrient medium with 10% pooled serum (Nc) or with 10% of autologous serum (Na) without or with 21 μg/ml Herceptin (Na+H). Effectors (E) to target tumor cells (T) ratio was 5:1. ::: Control S ~Nc~(%) S ~Na~(%) S ~Na+H~(%) CD16 (%) --------- ----------- ----------- ------------- ---------- 1 ND 103 49 15.00 2 ND 52 24 14.20 3 ND 73 55 10.00 4 ND 92 35 35.45 5 ND 66 17 ND 6 ND 22 26 ND 7 ND 89 58 ND 8 105 85 85 ND 9 58 85 49 17.55 10 74 82 34 16.15 11 20 27 0 16.45 12 93 92 24 ND 13 34 0 0 8.98 14 40 26 26 12.48 15 40 30 21 9.49 16 29 25 26 4.62 17 46 25 22 9.49 18 53 8.8 7 32.19 19 74 29 13 13.2 20 47 16 10 34.58 ND -- non-determined ::: Acknowledgements ================ This work was financed by Ministry of Science and Environmental Protection of Serbia, (grant number 1614). The authors thank Mrs. Tatjana Petrovic and Mr Filip Stojanovic for their excellent technical assistance.	PubMed Central	2024-06-05T03:55:55.628627	2005-3-22	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1079956/", "journal": "J Transl Med. 2005 Mar 22; 3:13", "authors": [ { "first": "Zorica D", "last": "Juranic" }, { "first": "Zora", "last": "Neskovic-Konstantinovic" }, { "first": "Tatjana P", "last": "Stanojkovic" }, { "first": "Zeljko", "last": "Zizak" }, { "first": "Tatjana", "last": "Srdic" }, { "first": "Nevenka", "last": "Stanojevic-Bakic" }, { "first": "Dusanka", "last": "Milosevic" }, { "first": "Danica", "last": "Jovanovic" } ] }
PMC1079957	The launching of Virology Journalcomes at a time of resurgence of interest in the basic biology of the bacteriophages and the impact that these viruses have on earth\'s ecology, evolution of microbial diversity and the control of infectious disease. Since playing an important part in the birth of Molecular Biology more than 50 years ago \[[@B1]\], phage research has continually broken new ground in our understanding of the basic molecular mechanisms of gene action and biological structure \[[@B2]\]. This trend shows no signs of waning. In a recent international meeting entitled The New Phage Biology \[[@B3]\], the program was largely devoted to emerging frontiers of research that have been empowered by a rapid accumulation of genome sequence information from a wide variety of bacteriophages. Phage genomics is revealing novel biochemical mechanisms for replication, maintenance and expression of the genetic material and is providing new insights into origins of infectious disease and the potential use of phage gene products and even whole phage as therapeutic agents. Two reasons why the new era of phage research brims with excitement are the abundance of bacteriophages in nature and the diversity of their genomes. Phage is probably the most widely distributed biological entity in the biosphere, with an estimated population of \>10^30^or \~10 million per cubic centimeter of any environmental niche where bacteria or archaea reside \[[@B4]\]. At one level, there is diversity in the types of phages that infect individual or interrelated bacterial species. At another level, there is diversity among genomically related phages that do not share the same bacterial hosts. One example is the lytic Enterobacterial dsDNA phage T4, which has relatives that are specific to Aeromonas, Vibrio, Acinetobacter, marine and other bacterial species. The genomes of a few T4-like phages have been sequenced and found to indeed share homologies with T4, but to also differ from one another in size, organization of the T4-like genes and content of other putative genes and DNA mobile elements (<http://phage.bioc.tulane.edu>). It appears that phage families like the T4-related phages have learned to cross bacterial species barriers and possess plastic genomes that can acquire and lose genetic cassettes through their travels in the microbial world. In essence, genomes of the dsDNA phages may be repositories of the genetic diversity of all microorganisms in nature. In addition to evolving by serving as traffickers of microbial genes, phage genomes evolve through the accumulation of mutations in both acquired and core genes. Sequence divergence among homologues of the essential genes for phage propagation within a phage family can be used as a source of information about the determinants of specificity of the protein-protein and protein-nucleic-acid interactions that underlie biological function. Phages are excellent sources of many enzymes and biochemical transactions that are broadly represented in all divisions of life. The large numbers of phylogenetic variants of biologically interesting proteins and nucleic acids that one can derive from sequenced phage genomes are treasure troves for studies of biological structure in relation to function. Interest in phage and phage gene products as potential therapeutic agents is also increasing rapidly and is likely to have profound impact on the pharmaceutical industry and biotechnology in general over the coming years. There is a general sense that the best is yet to come out of phage research. Conclusion ========== We anticipate that the pages of Virology Journalwill reflect the excitement of the \"New Phage Biology\" by publishing reports in the areas of Ecology and Taxonomy, Genomics and Molecular Evolution, Regulation of Gene Expression, Genome Replication and Maintenance, Protein and Nucleic Acid Structure, Virus assembly, Biotechnology, Pathogenesis, Therapeutics and more. It would be especially interesting to see submissions of phage genome sequence briefs and their biological implications. Competing interests =================== The author(s) declare that they have no competing interests.	PubMed Central	2024-06-05T03:55:55.632067	2005-3-15	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1079957/", "journal": "Virol J. 2005 Mar 15; 2:19", "authors": [ { "first": "Jim D", "last": "Karam" } ] }
PMC1079958	Introduction ============ Only a very small fraction of the vast number of viral species belonging to the different virus families have been identified and characterized to date. The majority of these uncharacterized viral species are found in host organisms which have not been targeted in biomedical, plant or animal research. However, recent reports have noted an increase in the occurrence of viral diseases, not only in humans, but in animals and plants as well. While some of this rise may reflect more effective surveillance techniques, disease outbreaks caused by novel cross-species infections and/or subsequent virus recombination events have occurred \[[@B1]\]. Therefore, the development of tools for the detection of viruses, the characterization of their genomes and the study of their evolution, becomes important, not only for basic scientific study, but also for the protection of public health and the well-being of the plant and animal life that surrounds us. We have developed a novel technology to identify and characterize distantly related gene sequences based on consensus-degenerate hybrid oligonucleotide primers (CODEHOPs)\[[@B2]\]. CODEHOPs are designed from amino acid sequence motifs that are highly conserved within members of a gene family, and are used in PCR amplification to identify unknown related family members. We have developed and implemented a computer program that is accessible over the World Wide Web to facilitate the design of CODEHOPs from a set of related protein sequences \[[@B3]\]. This site is linked to the Block Maker multiple sequence alignment site \[[@B4]\] on the BLOCKS WWW server \[[@B5]\] hosted at the Fred Hutchinson Cancer Research Center, Seattle, WA. We have utilized the CODEHOP technique to develop novel assays to detect previously unknown viral species by targeting sequence motifs within stable housekeeping genes that are evolutionarily conserved between different members of virus families. Using CODEHOPs derived from conserved motifs within retroviral reverse transcriptases, we have previously identifed a diverse family of retroviral elements in the human genome \[[@B2]\], as well as a novel endogenous pig retrovirus \[[@B6]\], and a new retrovirus in Talapoin monkeys \[[@B7]\]. We have also developed assays to detect unknown herpesviruses by targeting conserved motifs within herpesvirus DNA polymerases. Using this approach, we have identified fourteen previously unknown DNA polymerase sequences from members of the alpha, beta and gamma subfamilies of herpesviruses \[[@B8]\], and have discovered three homologs of the Kaposi\'s sarcoma-associated herpesvirus in macaques \[[@B9],[@B10]\]. We have also used the CODEHOP technique to clone and characterize the entire DNA polymerase gene from these new viruses \[[@B10]\] and to obtain sequences for larger regions of viral genomes containing multiple genes, targeting the divergent locus B of macaque rhadinoviruses \[[@B11]\]. The sequence information obtained from the amplified gene and genomic fragments from these studies has allowed informative phylogenetic characterization of the new viral species, and has provided critical information regarding the gene structure and genetic content of these unknown viral genomes. In this review, the CODEHOP methodology and its utilization in the identification and characterization of novel viral genomes using the herpesvirus family as an example is described. Published CODEHOP assays that we have previously used to identify new herpesviruses are discussed and the latest refined assays and their utility are provided. The use of the CODEHOP methodology for the analysis of larger regions of viral genomes is presented along with the general application of this technology for the identification of viral species and their genes in other virus families. Finally, the software and Web site that we have developed to derive CODEHOP PCR primers from blocks of multiply aligned protein sequences are described. CODEHOP Methodology =================== General CODEHOP Design and PCR Strategy --------------------------------------- CODEHOPs are derived from highly conserved amino acid sequence motifs present in multiple alignments of related proteins from a targeted gene family. Each CODEHOP consists of a pool of primers where each primer contains one of the possible coding sequences across a 3--4 amino acid motif at the 3\' end (degenerate core) (Figure [1A](#F1){ref-type="fig"}) \[[@B2]\]. Each primer also contains a longer sequence derived from a consensus of the possible coding sequences 5\' to the core motif (consensus clamp). Thus, each primer has a different 3\' sequence coding for the amino acid motif and the same 5\' consensus sequence. Hybridization of the 3\' degenerate core with the target DNA template is stabilized by the 5\' consensus clamp during the initial PCR amplification reaction (Figure [1B](#F1){ref-type="fig"}). Hybridization of primers to PCR products during subsequent amplification cycles is driven by interactions through the 5\' consensus clamp. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### CODEHOP description and PCR strategy.(A) A conserved DNA polymerase sequence motif in LOGOS representation \[31\] and a sense-strand CODEHOP (HNLCA) derived from that motif is shown. The 3\' degenerate core contains all possible codons encoding four conserved amino acids and has a degeneracy of 32. The 5\' clamp contains a consensus sequence derived from the most frequently used codons for 5 upstream amino acids within the motif. (B) Schematic description of the CODEHOP PCR strategy illustrating regions of mismatch in primer-to-template annealing during the early PCR cycles and primer-to-product annealing during subsequent cycles. Vertical lines indicate matches between primer (arrow) and template or amplified PCR product. The overall degeneracy of the 3\' degenerate core is the product of the degeneracies at each nucleotide position so that the fraction of primers with sequences identical to the targeted template across the degenerate core = 1/degeneracy. ::: ![](1743-422X-2-20-1) ::: Conserved amino acid motifs used for CODEHOP design are identified by alignment of related proteins from a targeted gene family using computer programs such as the Clustal W multiple alignment program \[[@B12]\]. Optimal blocks contain 3--4 highly conserved amino acids with restricted codon multiplicity from which the 3\' degenerate core is derived; the presence of serines, arginines and leucines are not favored due to the presence of six possible codons for each amino acid. In addition, optimal blocks contain 5 or more conserved amino acids from which the 5\' consensus clamp is derived. These blocks of conserved amino acid sequences should be situated in close enough proximity to allow efficient PCR amplification between blocks yet distant enough to flank a region of significant sequence information. We have developed web-based software to predict CODEHOP PCR primers from blocks of conserved amino acid sequences \[[@B2],[@B13]\]. Multiple related protein sequences from the targeted gene family are provided to the Block Maker program \[[@B4]\] at the BLOCKs WWW server \[[@B5]\] which produces a set of conserved sequence blocks obtained from a multiple sequence alignment. The sequence block output is linked directly to the CODEHOP design software \[[@B3]\] which predicts and scores possible CODEHOP PCR primers. The different CODEHOP PCR primers discussed in this review were either designed manually or with the CODEHOP software, and are listed in Table [1](#T1){ref-type="table"}. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### CODEHOPs developed for herpesvirus screens targeting the DNA polymerase ::: CODEHOPS (degeneracy)^1^ Bias^2^ Sense 5\'\>3\' Sequence(degenerate codons are in lower case)^3^ ------------------------------- -------------------------- ------------- ----------------------------------------------------------- ------------------------------------- \"TVG-IYG\" Assay^4^ DFA (512) All HV (IHV, HHV6,7) NA^5^ \+ Gayttygcnagyytntaycc ILK (1024) All HV \+ TCCTGGACAAGCAGcarnysgcnmtnaa TGV (256) All HV (IHV, HHV6,7) -^6^ \+ TGTAACTCGGTGtayggnttyacnggngt IYG (48) All HV (IHV, AlHV1, RRV) \- \- CACAGAGTCCGTrtcnccrtadat KG1 (128) All HV \- \- GTCTTGCTCACCAGntcnacnccytt \"DFASA-GDTD1B\" Assay^7^ DFASA (256) All HV (IHV, HHV6,7) \- \+ GTGTTCGACttygcnagyytntaycc VYGA (256) All HV (IHV) \- \+ ACGTGCAACGCGGTGtayggnktnacngg GDTD1B (64) All HV \- \- CGGCATGCGACAAACACGGAGTCngtrtcnccrta \"QAHNA\" Assay^7^ QAHNA (48) αHV γHV (IHV, βHV) (CMV) \+ CCAAGTATCathcargcncayaa \"SLYP\" Assay^8^ SLYP1A (64) All HV (CMV, EHV2) \- \+ TTTGACTTTGCCAGCCTGtayccnagyatnat SLYP2A (128) CMV (All other HV) \- \+ TTTGACTTTGCCAGCCTGtayccntcnatnat CODEHOP Predicted^9^ HNLCA (32) All HV (IHV) CODEHOP^10^ \+ TCCATCATCCAGGCCcayaayytntg VYG1A (128) All HV (IHV) CODEHOP \+ GCAACGCGGTGTACggnktnacngg YGDTB (16) All HV CODEHOP^11^ \- CGGCATGCCATGAACATGGAGTCCGTrtcnccrta KGVDB (32) All HV CODEHOP \- CTTCCGCACCAGGTCnacnccytt ^1^The degree of degeneracy, ie the number of individual primers in the pool, is given in parentheses. ^2^Bias indicates the reliance on a specified subset of sequences for determination of the 3\' degenerate core or 5\' consensus clamp. Sequences which are biased against by the choice of nucleotide sequences are indicated in parentheses (see the multiple sequence alignments from which the primers were derived in Figures 3-6). ^3^IUB code: Y = T, C; R = A, G; K = G, T; M = A, C; H = A, C, T not G; N = A, C, G, T. ^4^\[8\] ^5^NA, not applicable ^6^(-), no specific design bias ^7^\[9\] ^8^Primers predicted manually. ^9^Primers predicted using the CODEHOP software. ^10^Clamp sequence was predicted by the CODEHOP software using default codon usage table and thus had no inherent bias design ^11^Underlined sequences have been added to the primer predicted by the CODEHOP software (see legend to Figure 4) Abbreviations: HV, herpesvirus; αHV, alphaherpesvirus; βHV, betaherpesvirus; γHV, gammaherpesvirus; AhlHV1, alcelaphine herpesvirus 1; CMV, cytomegalovirus; EHV2, equine herpesvirus-2, HHV6, human herpesvirus 6; HHV7, human herpesvirus 7; IHV, ictalurid herpesvirus (catfish) ::: CODEHOP PCR Amplification, Product Cloning and Sequence Analysis ---------------------------------------------------------------- CODEHOP PCR amplification has been performed using classical and touch-down approaches with a hot-start initiation \[[@B2]\]. More recently, thermal gradient PCR amplification has been used to empirically determine optimal annealing and amplification conditions for the pool of primers \[[@B11]\]. Different buffers, salt concentrations, and enzymes have been employed with varying success due to differences in DNA template preparation and the unknown nature of the targeted sequence. PCR products are either sequenced directly or after TA-cloning. In this review, sequences were compared by BLAST analysis \[[@B14]\] and multiple alignment using Clustal W \[[@B12]\]. Phylogenetic analysis of the multiply aligned sequences was performed using protein distance and neighbor-joining analysis implemented in the Phylip analysis package \[[@B15]\]. Bootstrap analysis was also performed with 100 replicates and a consensus phylogenetic tree was determined. For the phylogenetic analysis, positions in the multiple alignment containing gaps due to insertions or deletions within the sequence blocks were eliminated. The \"TGV-IYG\" CODEHOP assay to detect novel herpesviruses =========================================================== The Herpesviridaewas chosen as a target virus family to develop assays to detect and characterize new viral members. All members of the herpesvirus family contain a DNA polymerase within their genome which is highly conserved across the different family members. Multiple alignment of different herpesvirus polymerase sequences revealed blocks of conserved amino acids corresponding to many of the functionally important motifs \[[@B16]\], see Figure [2A](#F2){ref-type="fig"}. We have developed and refined PCR strategies using CODEHOP PCR primers derived from these conserved sequence blocks to detect novel herpesviruses and characterize their genomes. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### CODEHOP strategies to identify and molecularly characterize new herpesviruses targeting the DNA polymerase gene.(A) Conserved sequence domains within herpesvirus DNA polymerases. Functional properties of these domains and amino acid (one letter code) motifs present in the domains are indicated. Motifs chosen as targets for the CODEHOP strategy are shown as black boxes. (B) Schematic diagram of the CODEHOP primer positions, the amplification products and their sizes. See Table 1 for primer sequences. ::: ![](1743-422X-2-20-2) ::: Initially, we manually designed a set of nested PCR primers from four of the conserved DNA polymerase blocks (indicated as black boxes in Figure [2A](#F2){ref-type="fig"}) which could be used to identify new viral polymerases and detect the existence of previously unknown or uncharacterized herpesviruses \[[@B8]\]. The primers, \"TGV\", \"IYG\", \"DFA\" and \"KG1\" (Table [1](#T1){ref-type="table"}), and the blocks of multiply aligned sequences from which the primers were derived are shown in Figures [3](#F3){ref-type="fig"}, [4](#F4){ref-type="fig"}, [5](#F5){ref-type="fig"}, [6](#F6){ref-type="fig"}, respectively (letters in the primer name refer to conserved amino acids in the sequence motif). Although these primers were alternately referred to as either \"consensus\" primers or \"degenerate\" primers within the original publication, all except DFA were designed using the general CODEHOP strategy \[[@B2]\]. In the \"TGV-IYG\" herpesvirus assay, the \"DFA\" sense primer was used in an initial PCR amplification with the \"KG1\" anti-sense primer (Figure [2B](#F2){ref-type="fig"}). An additional sense primer \"ILK\" located downstream of the \"DFA\" motif was also added to the initial amplification reaction \[[@B8]\]. The product from this amplification was used as template in a nested amplification reaction using the \"TGV\" sense primer and the \"IYG\" anti-sense primer (Figure [2B](#F2){ref-type="fig"}). This final PCR product was sequenced to obtain the \~165--180 bp region of the DNA polymerase gene located between the two motifs \"TGV\" and \"IYG\". The distance between the two motifs was variable between viral species due to small sequence insertions or deletions. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### CODEHOP PCR primers derived from the VYGF/TGV sequence motif.(A) Multiple sequence alignment of 11 herpesvirus DNA polymerase sequences contained within the conserved VYGF/TGV domain as an output of BlockMaker \[32\]. (B) Sequences from 6 additional herpesvirus species aligned with the conserved sequence block. (C) The consensus amino acid sequence from the VYGF/TGV motif as determined by the CODEHOP algorithm is presented (in bold and boxed) and the other amino acids found at each position are aligned vertically above the consensus amino acid. The sense-strand \"VYG1A\" CODEHOP predicted by the CODEHOP software is indicated with the 5\' consensus clamp in uppercase and the 3\' degenerate core region in lowercase. The sequence, relative position and encoded sequences of the manually designed CODEHOPs, \"TGV\" and \"VYGA\" are also shown (see Table 1). Highlighted amino acids are discussed in the text. The degeneracy of the primer pools is indicated in parentheses. DNA polymerase protein sequences were derived from the following herpesvirus species: HSV1, NC\_001806; VZV, NC\_001348; HHV6, NC\_001664; CMV, AF033184; HHV7, NC\_001716; RhCMV, AF033184; hCMV, AF033184;; HSV2, NC\_001798; RFHVMm, AF005479; MHV68, NC\_001826; KSHV, AF005477; HVS, NC\_001350; AtHV3, NC\_001987; AlHV1, NC\_002531; RRV, AF029302; IHV, NC\_001493; EBV, NC\_001345; EHV2, NC\_001650. ::: ![](1743-422X-2-20-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### CODEHOP PCR primers derived from the IYG/GDTD sequence motif(A)(B) Sequence alignments across the IYG/GDTD motif as described in the legend to Figure 3. (C) The consensus amino acid sequence from the IYG/GDTD motif as determined by the CODEHOP software is presented (in bold and boxed) and the other amino acids found at each position are aligned vertically above the consensus amino acid. The coding strand sequence and the complementary strand corresponding to the \"YGDTB\" CODEHOP predicted by the CODEHOP algorithm are indicated with the sequences of the 5\' consensus clamp in uppercase and the 3\' degenerate core region in lowercase. The consensus sequence shows the extent of the sequence block determined by BlockMaker. The CODEHOP algorithm was unable to determine a 5\' consensus clamp giving the required Tm due to the small size of the block. Therefore, three additional amino acid positions (in italics) were added to the C\' terminal side of the block in (A) and (B) to allow visual inspection of the sequences to manually determine an additional 8 bp of the 5\' consensus clamp which are underlined. The nucleotide sequences, relative positions and encoded amino acid sequences for the manually designed CODEHOPs, \"IYG\" and \"GDTD1B\" are also shown (see Table 1 for the exact nucleotide sequences of these anti-sense strand primers). The degeneracy of the primer pools is indicated in parentheses and the highlighted residues are discussed in the text. The CODEHOP primers, YGDTB, IYG and GDTD1B are all derived from the antisense DNA strand and are shown below the codons for the sense strand. ::: ![](1743-422X-2-20-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### CODEHOP PCR primers derived from the \"DFAS/QAHN\" sequence motif(A)(B) Sequence alignments across the \"DFAS\" motif as described in the legend to Figure 3. The non-conserved amino acids in the IHV sequence are highlighted (C) The consensus amino acid sequence from the \"DFAS\" motif as determined by the CODEHOP algorithm is presented (in bold and boxed) and the other amino acids found at each position are aligned vertically above the consensus amino acid. The sense-strand \"HNLCA\" CODEHOP predicted by the CODEHOP software is indicated with the 5\' consensus clamp in uppercase and the 3\' degenerate core region in lowercase. The sequence, relative position and encoded sequences of the manually designed CODEHOPs, \"DFA\", \"DFASA\", \"QAHNA\" and \"SLYP1A\" are also shown (see Table 1). The degeneracy of the primer pools is indicated in parentheses. The codons found in the different herpesvirus sequences encoding the serine (S), block position 6, in the \"DFAS\" motif were all of the \"AGY\" type serine codons, so the manually derived primers utilized those codons exclusively at that position. ::: ![](1743-422X-2-20-5) ::: ::: {#F6 .fig} Figure 6 ::: {.caption} ###### CODEHOP PCR primers derived from the \"KGV\" sequence motif(A)(B) Sequence alignments across the \"KGV\" motif as described in the legend to Figure 3. (C) The consensus amino acid sequence from the \"KGV\" motif as determined by the CODEHOP algorithm is presented (in bold and boxed) and the other amino acids found at each position are aligned vertically above the consensus amino acid. The sequences of the coding strand and complementary strand corresponding to the \"KGVDB\" CODEHOP predicted by the CODEHOP software is indicated. The nucleotide sequences, relative positions and encoded amino acid sequences of the manually designed CODEHOP, \"KG1\", are also shown (see Table 1 for the exact nucleotide sequences of these anti-sense strand primers). The degeneracy of the primer pools is indicated in parentheses. ::: ![](1743-422X-2-20-6) ::: We have shown the utility of this CODEHOP PCR primer strategy by identifying and characterizing14 previously unknown DNA polymerase sequences from members of the alpha, beta and gamma subfamilies of herpesviruses \[[@B8]\]. Since this original publication, more than 21 additional \"TGV-IYG\" DNA polymerase sequences from previously uncharacterized herpesviruses have been obtained by other investigators using this CODEHOP primer strategy (see Additional File 1; \"TGV-IYG\" assay). In some cases, PCR amplification was performed with modified deoxyinosine-substituted primers \[[@B17]\]. Comparison of the amino acid sequences encoded within the \"TGV-IYG\" region has allowed phylogenetic comparison of the different herpesvirus species from which these sequences were obtained. Figure [7](#F7){ref-type="fig"} shows a phylogenetic tree resulting from the analysis of the sequences obtained from 34 different herpesvirus species identified using the \"TGV-IYG\" CODEHOP strategy and the corresponding sequences of six representative human herpesviruses. Although the number of amino acid comparisons within this region is limited, ie. only 53 amino acids, preliminary assignment of many of the herpesvirus species to one of the three herpesvirus subfamilies has been possible (Figure [7](#F7){ref-type="fig"} and Additional File 1). Values from the bootstrap analysis using 100 replicates are indicated for each branch point. While some of the branch points were not well defined due to the limited amount of sequence data, as indicated by boostrap values less than 50, many groupings were well supported. The analysis shows clearly the grouping of different viral species from evolutionarily related hosts. This is consistent with previous studies which have shown extensive cospeciation of viral species and their host lineages \[[@B18]\]. ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Phylogenetic analysis of DNA polymerase sequences from different herpesvirus species identified with the \"TGV-IYG\" CODEHOP assayThe phylogeny of DNA polymerase sequences (\~53 amino acids in length) from thirty-six herpesviruses identified using the \"TGV-IYG\" assay (see Tables 2 and 3) and the corresponding sequences of six representative human herpesviruses (boxed) was determined using the neighbor joining method (Neighbor) applied to pairwise sequence distances (ProtDist) using the Phylip suite of programs \[15\]. Bootstrap scores (Seqboot) from 100 replicates are indicated and the consensus tree (Consense) is shown. The clustering of the alpha, beta and gamma herpesviruses, including the gamma-1 (Lymphocryptovirus) herpesviruses, and the RV1 and RV2 gamma-2 (Rhadinovirus) lineages are indicated. ::: ![](1743-422X-2-20-7) ::: Parameters for refinement of the \"TVG-IYG\" assay ================================================== Limiting degeneracy to increase sensitivity ------------------------------------------- While the \"TVG-IYG\" herpesvirus assay demonstrated the ability to detect disparate herpesvirus species in high titer virus cultures in vitro, the detection of limiting amounts of virus in tissue samples in vivowas problematic. This was especially true in sections obtained from formalin-fixed, paraffin-embedded tissue blocks which contained small amounts of degraded DNA. The degeneracy of the primer pool, ie. the number of different primers necessary to encode all codon possibilities for the specified block of conserved amino acids, plays a direct role in the sensitivity of the PCR amplification. Whereas highly degenerate primers consisting of pools of hundreds or thousands of primers with different DNA sequences may allow amplification of DNA templates present in high copy number, as found in cultured virus stocks, they are less successful in amplifying low copy numbers of DNA templates found in virus infected tissues in vivo, especially in formalin-fixed tissue. As the degeneracy increases, the concentration of the primer or primers that will participate in the desired amplification reaction decreases and can become suboptimal. Conversely, the vast excess of primers not participating in the amplification of the targeted gene can cause non-specific amplification which can, in turn, inhibit or mask the amplification of the desired target. As indicated in Table [1](#T1){ref-type="table"}, the degeneracy of the primers utilized in the \"TVG-IYG\" assay ranged from 48--1024. This level of degeneracy was driven by the number of nucleotide possibilities encoding the targeted amino acids at each position as well as by the number of amino acid positions allowed to be degenerate. Figure [5A](#F5){ref-type="fig"} shows the DFA/DFAS/QAHN sequence block produced by Block Maker from multiple alignments of 11 different herpesvirus polymerase sequences. Figure [5C](#F5){ref-type="fig"} shows the consensus amino acids at each position, as determined by the CODEHOP algorithm, which are boxed and bolded with the alternate amino acids positioned above. The original primer manually derived from this motif, \"DFA\" is, in fact, completely degenerate, with multiple codons provided for each amino acid position, except the ultimate proline (P) residue, yielding a pool of 512 different primers \[[@B8]\]. Because the performance of this primer was consistently suboptimal in samples with limiting template, the overall structure and degeneracy of the primer was altered by designing a PCR primer \"DFASA\" from the same sequence motif using the CODEHOP methodology. This primer had an 11 bp 5\' consensus region and a 3\' degenerate core containing multiple codons at 5 amino acid positions resulting in a pool of 256 different primers (Figure [5C](#F5){ref-type="fig"}). The \"DFASA\" primer was successfully used to amplify extremely low amounts of viral DNA in a background of genomic DNA from paraffin-embedded formalin-fixed tissue in the discovery of the macaque homolog of Kaposi\'s sarcoma-associated herpesvirus, called retroperitoneal fibromatosis herpesvirus (RFHV) \[[@B9]\]. Subsequent estimates of virus copy number using real-time quantitative PCR indicated a level of RFHV DNA in the available samples that was 1/100--1/1000 of a single copy cellular gene (unpublished observations). The \"DFASA\" primer has been successfully used to identify a number of novel alpha-, beta- and gammaherpesviruses in a wide variety of host organisms (see Additional File 1: \"DFASA-GDTD1B assay\"). Due to the presence of a highly conserved leucine (L) at block position 7 within the \"DFAS\" motif (Figure [5](#F5){ref-type="fig"}) which significantly increased the degeneracy of the primer pool with its six possible codons, an additional CODEHOP was designed from the \"QAHN\" motif immediately downstream of \"DFAS\" to further decrease degeneracy. The \"QAHNA\" primer had an 11 bp 5\'consensus region and a 3\' degenerate core containing multiple codons at 4 amino acid positions resulting in a pool of 48 different primers (Figure [5C](#F5){ref-type="fig"}). This CODEHOP has been successfully used to identify several primate rhadinoviruses related to KSHV in tissue samples with limiting amount of viral DNA \[[@B10],[@B19]\], see also Additional File 1. Primer bias and specificity --------------------------- The primers developed for the \"TGV-IYG\" assay were designed to amplify polymerase fragments from herpesviruses of all three subfamilies based on conserved motifs within the known sequences. However, very few sequence motifs were absolutely conserved between the most divergent herpesviruses. For example, the catfish ictalurid herpesvirus (IHV) lacked the \"KGV\" motif from which the initial \"KGV\" primer was derived (Figure [6](#F6){ref-type="fig"}). Furthermore, numerous sequence differences were present in the IHV DNA polymerase within the DFAS/QAHN motif which was otherwise highly conserved in other herpesvirus species (highlighted residues in Fig. [5B](#F5){ref-type="fig"}). Because of these differences, the IHV sequence was excluded from the primer design of the \"DFA\", \"DFASA\" and \"QAHNA\" PCR primers. As shown in Figure [5C](#F5){ref-type="fig"}, the \"DFA\" and \"DFASA\" primers have mismatches with the IHV sequence at the alanine (A) and leucine (L) codons (Block positions 5 and 7, respectively; Figure [5B](#F5){ref-type="fig"}) and the \"QAHNA\" primer mismatches at three codon positions (Block positions 13--15; Figure [5B](#F5){ref-type="fig"}), all within the 3\' degenerate cores. Figure [8](#F8){ref-type="fig"} shows the presence of nucleotide mismatches with the IHV sequence throughout the different primers (black highlighting). Thus, the lack of the \"KGV\" motif and sequence differences in the \"DFA\" primer strongly biased the \"TGV-IYG\" assay against IHV-like herpesvirus sequences. In order to identify IHV-like herpesviruses, new primers would have to incorporate these sequence differences. ::: {#F8 .fig} Figure 8 ::: {.caption} ###### Alignment of CODEHOP PCR primers with the nucleotide sequences encoding the \"DFAS/QAHN\" sequence block(A) Amino acid consensus sequence -- see Figure 5C (B) Nucleotide sequences encoding the amino acids in the \"DFAS/QAHN\" sequence block from the 11 different herpesvirus species that were used to generate the sequence block. (C) Nucleotide sequences from six additional herpesvirus species. (D) Nucleotide sequences of five manually designed primers \"DFA\", \"DFASA\", \"SLYP1A\", \"SLYP2A and \"QAHNA\", and a primer designed using the CODEHOP software (HNLCA). The codons from two conserved serine positions are boxed and nucleotide sequences mismatched with the different 3\' degenerate cores of the primers are highlighted in black. The subfamily associations of the different viral species are indicated. ::: ![](1743-422X-2-20-8) ::: The \"DFA\" and \"DFASA\" primer pools were originally designed using only the alanine (A) codon at block position 5 in the \"DFAS\" motif and did not include the glutamine (Q) codon found in that position of the motif in HHV6 and HHV7, \"DFQS\" (highlighted, Figure [5A, B](#F5){ref-type="fig"}). The nucleotide mismatches in this region are shown in Figure [8](#F8){ref-type="fig"}. While the \"DFA\" and \"DFASA\" primers are biased by design against HHV6 and HHV7, they have been used successfully to detect betaherpesviruses related to HHV6 and HHV7 \[[@B8]\]. This suggests that mismatches 13--14 nucleotides from the 3\' end of the primer, do not have major affects on the utility of the primers, especially when viral template is not limiting. More significant bias against HHV6- and HHV7-like herpesviruses was present in the \"TGV\" primer used in conjunction with the \"IYG\" primer in the secondary nested PCR reaction in the \"TGV-IYG\" assay (see Figure [2B](#F2){ref-type="fig"}). The \"TGV\" primer contains the partial valine (V) codon \"GT\" at its 3\' end (Block position 11; Figure [3C](#F3){ref-type="fig"}). Since both HHV6 and HHV7 contain alanine (A) (codon = GCN) at this position (highlighted in Fig. [3A, B](#F3){ref-type="fig"}), the \"TGV\" primer would mismatch at the 3\' terminal nucleotide with both HHV6- and HHV7-like sequences. This mismatch occurs at the 3\' end of the \"TGV\" primer and is predicted to significantly impair polymerase extension. To remove this bias, the \"TGV\" primer was redesigned as the \"VYGA\" primer removing the 3\' terminal \"GT\" of the valine codon and the terminal degenerate position of the glycine (G) codon. The \"TGV\" primer contained an additional bias against amplification of HHV6-like sequences due to the use of only the phenylalanine (F) codons (TTY) (Block position 8) at a position encoding valine (V) in both HHV6 and HHV7 (highlighted in Figure [3A](#F3){ref-type="fig"} and [3B](#F3){ref-type="fig"}). To remove this bias, \"VYGA\" was designed to include both the valine (V) and (F) codons at this position. The total degeneracy of the \"TGV\" and \"VYGA\" primer pools remained the same, with 256 different primers, due to the loss of the degenerate codon position in the glycine, block position 10 in \"TGV\" and the gain of the degenerate codon positions in the valine, block position 8 in \"VYGA\". The subsequent cloning and sequence analysis of new herpesvirus DNA polymerases from the rhadinoviruses, rhesus rhadinovirus (RRV) and alcelaphine herpesvirus 1 (AlHV1) \[[@B20],[@B21]\], revealed mismatches with the downstream \"IYG\" primer of the \"TVG-IYG\" herpesvirus assay. The \"IYG\" primer (a reverse orientation primer) includes the codons (ATH) for isoleucine (I) at its 3\' end (Block position 1; Figure [4C](#F4){ref-type="fig"}). Both RRV and AH1 contain a valine (V) codon (GTN) at this position (highlighted in Figure [4A](#F4){ref-type="fig"}). Thus, \"IYG\" is biased against RRV-like or AH1-like rhadinoviruses due to a T-C mismatch at the 3\' end of the primer. To eliminate this bias, the \"IYG\" primer was redesigned as \"GDTD1B\" to remove the isoleucine position within the 3\' degenerate core (Figure [4C](#F4){ref-type="fig"}) and, in addition, the length of the 5\' consensus clamp was increased. Decrease in size of the amplification products ---------------------------------------------- Because typical tissue samples especially paraffin-embedded formalin-fixed tissue contain degraded DNA with sizes averaging near 300--500 bp in length, we decided to decrease the maximal amplification product size of the herpesvirus assay. The initial amplification product of the \"TGV-IYG\" assay (DFA-KG1) was \~800 bp (Fig. [2B](#F2){ref-type="fig"}). To reduce the initial amplification product size, a hemi-nested PCR assay was developed in which the newly designed downstream anti-sense primer \"GDTD1B\" targeting the highly conserved \"YGDT\" motif was used in a primary PCR amplification with the new upstream primer \"DFASA\". This amplification yields an approximate 500 bp PCR product (Figure [2B](#F2){ref-type="fig"}). This initial PCR product is then used as template in a secondary PCR amplification using the nested primer \"VYGA\" with the downstream anti-sense primer \"GDTD1B\". This amplification yields a PCR product of approximately 200 bp (see Figure [2B](#F2){ref-type="fig"}). These modifications produce amplification products close to the average size of degraded DNA present in fixed tissue. The \"DFASA/QAHNA-GDTD1B\" herpesvirus assay: a refinement of the \"TGV-IYG\" assay =================================================================================== We have developed a refined herpesvirus assay using the optimized DNA polymerase CODEHOP PCR primers, discussed above. This assay was designed to use only three CODEHOPs in a hemi-nested PCR assay in which \"DFASA\" and \"GDTD1B\" are used in an initial PCR amplification (Figure [2B](#F2){ref-type="fig"}). The product from that amplification is used as template in a secondary amplification with \"VYGA\" and the original anti-sense primer \"GDTD1B\". A variation of this assay uses the \"QAHNA\" to replace \"DFASA\". Thus, the amplification of novel polymerase sequences required the conservation of only three motifs, rather than five in the original \"TGV-IYG\" assay. Using these assays, we have identified three novel homologs of the newly characterized human herpesvirus, KSHV, in two species of macaques \[[@B9]\] (see Table [1](#T1){ref-type="table"}, RFHVMn, RFHVMm and MneRV2). Phylogenetic analysis of the molecular sequences obtained from these studies provided strong evidence for the existence of two distinct lineages of γ2 rhadinoviruses related to KSHV, called rhadinovirus-1 (RV1) and rhadinovirus-2 (RV2) (Figure [9](#F9){ref-type="fig"}) \[[@B10]\]. Subsequent studies by others using this assay, have identified the presence of additional members of these two lineages in other Old World primates, including African green monkeys \[[@B19]\], mandrills \[[@B22]\], chimpanzees \[[@B23],[@B24]\] and gorillas \[[@B24]\] (see Additional File 1). This data predicts the existence of another human herpesvirus closely related to KSHV belonging to the RV-2 lineage of rhadinoviruses \[[@B10]\]. ::: {#F9 .fig} Figure 9 ::: {.caption} ###### Phylogenetic analysis of DNA polymerase sequences from different herpesvirus species identified with CODEHOP assays targeting the DFAS and YGDT motifsThe phylogeny of DNA polymerase sequences (\~142 amino acids in length) from 25 different herpesvirus species identified using either the \"DFA-IYG\", \"DFASA-GDTD1B\", or QAHNA-GDTD1B assays (see Tables 2 and 3), was determined as described in the legend to Figure 7. ::: ![](1743-422X-2-20-9) ::: The utility of the \"DFASA/QAHNA-GDTD1B\" assays has been demonstrated by these and other studies in which more than 19 novel herpesviruses from the alpha, beta and gamma subfamilies of a wide variety of host species have been identified and molecularly characterized using CODEHOPs (Tables [2](#T2){ref-type="table"} and [3](#T3){ref-type="table"}). Comparison of the amino acid sequences encoded between the \"DFAS\" and \"IYG/GDTD\" motifs has allowed the phylogenetic comparison of the different herpesvirus species from which these sequences were obtained. Figure [9](#F9){ref-type="fig"} shows a phylogenetic tree resulting from the analysis of the sequences obtained from the \"DFA-IYG\", and \"DFASA/QAHNA-GDTD1B\" assays and the corresponding sequences of six representative human herpesviruses. Multiple sequence alignments of the viral sequences were performed and the positions containing gaps were eliminated, leaving 142 amino acid positions for comparison. These sequences were analyzed using protein distances and neighbor-joining analysis implemented in the Phylip analysis package \[[@B15]\]. As shown in Figure [9](#F9){ref-type="fig"}, most of the different viral species could be unambiguously included within either of the three herpesvirus subfamilies as indicated by the high bootstrap scores obtained for most of the branch points. However, the positioning of the branch points for certain viral species could not be reliably determined using the available sequence information. Such uncertainty has been seen in similar analysis of specific herpesvirus species using much larger data sets \[[@B18]\]. The results obtained using the 142 amino acid comparisons confirmed and extended the phylogenic relationships predicted from the \"TVG-IYG\" results derived from only 53 amino acid comparisons. Furthermore, the phylogenetic relationships predicted by the different CODEHOP assays have been subsequently confirmed when substantially more sequence information was obtained from the new viral species, see \[[@B10],[@B11]\]. The phylogenetic relationships shown in Figure [9](#F9){ref-type="fig"} are consistent with the findings that extensive cospeciation of viral species and their host lineages has occurred during evolution \[[@B18]\]. The wide variety of different herpesvirus species identified using the CODEHOPs assays targeting the DNA polymerase gene, as shown in Figures [7](#F7){ref-type="fig"} and [9](#F9){ref-type="fig"}, indicate the wide applicability of the CODEHOPs assays to detect herpesviruses from disparate host lineages. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Alpha- and Betaherpesviruses identified and/or characterized using CODEHOP-based PCR assays targeting the herpesvirus DNA polymerase ::: ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Virus species^1^ Abbrev.^1^ Host Strain Assay Accession (\#aa) Reference ------------------------------------------------------------- ---------------- ----------------------------- ------------ ----------------- ---------------------- --------------- [Alphaherpesvirus]{.underline} Bovine HV-2 BHV2 Cow TGV-IYG^2^ AAC59453 (59aa) \[36\] Canid HV-1 CHV1 Dog D004 TGV-IYG AAC55646 (60aa) \[8\] Caretta caretta HV CcaHV Florida loggerhead turtle TGV-IYG AAD24564 (60aa) \[37\] Chelonia mydas HV-Florida CmyHVf Florida green turtle TGV-IYG\ AAD24565 (60aa)\ \[37\]\ DFASA-GDTD1B^3^ AAC26682 (161aa) \[38\] Chelonia mydas HV-Hawaii CmyHVh Hawaiin green turtle DFASA-GDTD1B AAC26681 (161aa) \[38\] Equid HV-3 EHV3 Horse C-175 TGV-IYG AAD30140 (59aa) \[17\] Felid HV-1 FHV1 Cat C-27 TGV-IYG AAC55649 (60aa) \[8\] Infectious laryngotracheitis virus (Gallid HV-1) ILTV Chicken N-71851 TGV-IYG AAC55650 (59aa) \[8\] Marek\'s disease virus (Gallid HV-3) MDV Chicken GA5 TGV-IYG AAC55651 (59aa) \[8\] Lepidochelys olivacea HV LolHV Olive ridley turtle DFASA-GDTD1B AAC26684 (161aa) \[38\] Psittacid HV-1 PsiHV1 Parrot RSL-1 TGV-IYG AAC55656 (59aa) \[8\] Saimiriine HV-1 SaHV1 S. American squirrel monkey MV-5-4 TGV-IYG AAC55657 (60aa) \[8\] Tursiops truncatus HV-1 TtrHV1 Bottlenose dolphin Heart TGV-IYG^2^ AAF62170 (60aa) Unpublished Tursiops truncatus HV-2 TtrHV2 Bottlenose dolphin Lung TGV-IYG AAF07208 (63aa) Unpublished [Betaherpesvirus]{.underline} African elephant endotheliolytic virus Afeev African elephant Case 2 TGV-IYG AAD24549 (60aa) \[39\] Asian elephant endotheliolytic virus Aseev Asian elephant Case 3 TGV-IYG Not Deposited (60aa) \[39\] Aotine HV-1 AoHV1 Owl monkey S43E TGV-IYG AAC55643 (57aa) \[8\] Chlorocebus aethiops cytomegalovirus (Cercopithecine HV-5) CaeCMV African green monkey CSG TGV-IYG AAC55647 (57aa) \[8\] Mandrill cytomegalovirus MndCMV Mandrill leucophaeus Mnd205 DFASA-GDTD1B AAG39064 (157aa) \[22\] Mandrill HV β MndHVβ Mandrill sphinx Mnd301 DFASA-GDTD1B AAG39065 (159aa) \[22\] ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ^1^Names and abbreviations are usually derived from the original publications although some have been modified to conform to a three letter code derived from the first letter of the genus and the first two letters of the species, ie. Macaca mulatta= Mmu. This was necessary due to the number of different viral species and hosts which could not be distinguished with a two letter code. ^2^Reference \[8\] ^3^Reference \[9\] ^4^Reference \[17\] ^5^Reference \[10\] ^6^Primers modified (see reference) ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Gammaherpesviruses identified and/or characterized using CODEHOP-based PCR assays targeting the herpesvirus DNA polymerase (see legend to Table 2) ::: ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Virus species^1^ Abbrev.^1^ Host Strain Assay Accession (\#aa) Reference ------------------------------------------------------------------------- ---------------- ----------------------------- ------------- ---------------------- ---------------------- ------------------------ [Gammaherpesvirus-1]{.underline} Bovine lymphotrophic HV BLHV Cow DFA-IYG^4^ AAC59451 (160aa) \[36\] Callitrichine HV-3 CalHV3 Marmoset TGV-IYG AAF05882 (58aa) Unpublished Leoporid HV-2 LeHV2 Rabbit TGV-IYG AAC55655 (54aa) \[8\] Rhesus lymphocryptovirus-1 (cercopithecine HV-15) MmuLCV1 Macaque mulatta DFASA-GDTD1B TGV-IYG This study AF091053 This study Unpublished Rhesus lymphocryptovirus-2 MmuLCV2 Macaque mulatta DFASA-GDTD1B This study HV papio (cercopithecine HV-12) HVP Baboon TGV-IYG AAF05878 (58aa) Unpublished Ovine HV 2 OHV2 Sheep DFA-IYG AAC59455 (161aa) \[36\] Porcine lymphotrophic virus-1a PLHV1a Pig 56 DFA-IYG AAD26258 (155aa) \[40\] Porcine lymphotrophic virus-1b PLHV1b Pig 68 DFA-IYG AAD26257 (155aa) \[40\] Saimiriine HV-3 SaHV3 S. American squirrel monkey TGV-IYG AAF98285 (57aa) Unpublished Zalophus californianus HV ZcaHV Sea lion TGV-IYG AAF07188 (55aa) Unpublished [Gammaherpesvirus-2]{.underline} Alcelaphine HV-1 AlHV1 Wildebeest TGV-IYG AAC59452 (58aa) \[36\] Alcelaphine HV-2 AlHV2 Hartebeest TGV-IYG AAG21352 (58aa) Unpublished Caprine HV-2 CapHV2 Goat TGV-IYG AAG21351 (59aa) Unpublished Caprine lymphotropic HV CapLHV Goat TGV-IYG AAG10783 (58aa) Unpublished Deer malignant catarrhal fever virus DMCFV Deer TGV-IYG AAD56945 (59aa) \[41\] Ateline HV-2 AtHV2 S. American spider monkey TGV-IYG AAC55644 (55aa) \[8\] Bovine HV-4 BHV4 Cow DFA-IYG AAC59454 (156aa) \[36\] Callitrichine HV-1 CalHV1 Marmoset TGV-IYG AAC55645 (55aa) \[8\] Chlorocebus rhadinovirus-1 ChRV1 African green monkey Z8 QAHNA-GDTD1B^5^ CAB61753 (151aa) \[19\] Chlorocebus rhadinovirus-2 ChRV2 African green monkey L1 QAHNA-GDTD1B CAB61754 (151aa) \[19\] Equine HV-2 EHV2 Horse TGV-IYG AAC55648 (55aa) \[8\] Equine HV-5 EHV5 Horse TGV-IYG^6^ AAD30141 (56aa) \[17\] Gorilla rhadinoherpesvirus 1 gorRHV1 Gorilla GorGabOmo DFASA-GDTD1B AAG23218 (158aa) \[24\] Kaposi\'s sarcoma-associated HV (HHV8) KSHV Human KS187 DFASA-GDTD1B AAC57974 (151aa) \[9\]\ \[10\] Macaque fascicularis rhadinovirus-2 (Macaque fascicularis gamma virus) MfaRV2 Macaque fascicularis DFASA-GDTD1B AAF23082 (158aa) \[42\] Macaque nemestrina rhadinovirus-2 MneRV2 Macaque nemestrina Mne442N DFASA-GDTD1B AAF81664 (158aa) \[10\] Mandrill rhadinoherpesvirus-1 MndRHV1 Mandrill sphinx Mnd15 DFASA-GDTD1B AAG39066 (158aa) \[22\] Mandrill rhadinoherpesvirus-2 MndlRHV2 Mandrill leucophaeus Mnd205 DFASA-GDTD1B AAG39061 (158aa) \[22\] Mandrill rhadinoherpesvirus-2 MndsRHV2 Mandrill sphinx Mnd13 DFASA-GDTD1B AAG39060 (158aa) \[22\] Pan troglodytes rhadinoherpesvirus-1a panRHV1a Chimpanzee PanCamDja DFASA-GDTD1B AAG23140 (158aa) \[24\] Pan troglodytes rhadinoherpesvirus-1b panRHV1b Chimpanzee PanCamEko DFASA-GDTD1B AAG23142 (158aa) \[24\] Retroperitoneal fibromatosis HVMm RFHVMm Macaque mulatta MmuYN91-224 QAHNA-GDTD1B AAC57976 (151aa) \[9\]\ \[10\] Retroperitoneal fibromatosis HVMn RFHVMn Macaque nemestrina Mne442N DFASA-GDTD1B AAF81662 (158aa) \[9\]\ \[10\] Rhesus rhadinovirus (Macaque mulatta gamma virus) RRV Macaque mulatta DFASA-GDTD1B AAF23083 (158aa) \[42\] Tapirus terrestris HV TteHV Tapir TGV-IYG^6^ AAD30142 (55aa) \[17\] Equus somalicus HV EsoHV Wild ass TGV-IYG^6^ AAD30143 (57aa) \[17\] Equus zebra HV EzeHV Zebra TGV-IYG^6^ AAD30144 (55aa) \[17\] ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ::: The \"SLYP1A-GDTD1B\" herpesvirus assay: a general herpesvirus detection assay ============================================================================== We designed additional primers from the DFAS/QAHN sequence motif using the CODEHOP strategy to develop further assays to detect new herpesviruses. The primer \"SLYP1A\" was one such primer designed to eliminate bias in the 3\' degenerate core of \"DFA\" and \"DFASA\" primers against HHV6 and HHV7, described above. The \"SLYP1A\" primer overlaps the \"DFA\" and \"DFASA\" primers and extends further downstream in a region very well conserved across the different herpesvirus species including HHV6 and HHV7 (Block positions 8--12; Figure [5C](#F5){ref-type="fig"}) \[[@B10]\]. Primer design across this region was based on the similarities in the first two positions for the codons for isoleucine (I) -- (ATA, ATC, ATT) and methionine (M) -- (ATG). These two amino acids are conserved in two positions within this sequence block in all herpesvirus species, including IHV (Block positions 11,12; Figure [5](#F5){ref-type="fig"}) and provide the penultimate and ultimate 3\' codons for the primer. Also, the SLYP1A primer was designed with only one of the two codon types utilized for serine (S) -- (AGY) to minimize degeneracy in the 3\' degenerate core (Block position 10; Figure [5C](#F5){ref-type="fig"}). Serine at this position (Block position 10; Figure [8](#F8){ref-type="fig"}) is encoded by AGY-type codons in all herpesvirus species, except for CMV-like herpesviruses which use TCN-type codons and EHV2 which contains a codon for threonine. A second related primer, SLYP2A was also designed from this region with an identical sequence except that the other serine codons (TCN) were used in the third position. Although this primer was biased for CMV-like sequences, we have successfully amplified KSHV which contains an AGT codon (unpublished results). We have previously used \"SLYP1A\" and \"GDTD1B\" to identify a new herpesvirus related to RRV, called Macaca nemestrinarhadinovirus-2 (MneRV2) in spleen tissue \[[@B10]\]. We subsequently used this assay to screen for herpesviruses in lymphomas from two rhesus macaques, L758 and 881, from the Tulane Regional Primate Research Center. DNA was kindly provided by LS Levy. Strong PCR products were obtained in primary amplification reactions and were cloned and sequenced. The lymphoma from rhesus 881 yielded clones containing a single sequence which was highly related to human EBV. From the lymphoma from rhesus L758, we obtained two distinct EBV-like sequences, one which was identical to the first lymphoma sequence and the other one which contained 10 nucleotide differences across the 475 bp fragment (98% identity). Analysis of the encoded amino acids revealed 3 amino acid differences (98% identity) between the two rhesus EBV-like sequences (MmuLCV1 and MmuLCV2) (Figure [10](#F10){ref-type="fig"}). These sequences clustered closely with human EBV in the γ1 branch of the phylogenetic tree shown in Figure [9](#F9){ref-type="fig"}. The identification of DNA polymerases from two types of EBV-like lymphocryptoviruses corroborates previous reports of the existence of two closely related lymphocryptoviruses in rhesus macaques \[[@B25]\] identified by sequence comparision of two distinct EBNA-2 genes. This is similar to the situation in humans where two different EBV species, EBV1 and EBV2 have been identified \[[@B26]\]. ::: {#F10 .fig} Figure 10 ::: {.caption} ###### Amino acid sequence comparision of two rhesus macaque EBV homologs detected using the \"SLYP1A-GDTD1B\" CODEHOP assayPositions with identity to human EBV are shown as a (.), and unidentified flanking regions or inserted gaps are indicated as (-). Numbering is from the human EBV DNA polymerase sequence. M. mulatta-1 and M. mulatta-2 sequences are listed in Table 1 as MmuLCV1 and MmuLCV2. The Macaca fascicularis, African green monkey (Chlorocebus aethiops) and baboon (Papio hamadryas) EBV-like sequences were published in \[33\] but not deposited in Genbank. The marmoset EBV-like sequence was deposited in Genbank as a AF291653 \[34\]. ::: ![](1743-422X-2-20-10) ::: Using the CODEHOP strategy to determine the complete sequence of novel viral genes ================================================================================== The CODEHOP assays described above targeted a restricted region of one gene and only provided limited sequence information. We have also used CODEHOPs to obtain the complete sequence of targeted genes and identify flanking genes within the unknown viral genome. To obtain the complete sequences of the DNA polymerase genes of the newly identified herpesvirus species of macaques, RFHVMn and RFHVMm, we designed CODEHOP PCR primers from additional conserved sequence blocks within the DNA polymerase (Figure [11](#F11){ref-type="fig"} and Table [4](#T4){ref-type="table"}). The new DNA polymerase-derived CODEHOP PCR primers, \"CVNVA\" and \"YFDKB\" were used in conjunction with gene specific primers derived from within the sequence of the original CODEHOP PCR product \"DFASA-GDTD1B to obtain overlapping PCR products across the majority of the DNA polymerase gene \[[@B10]\]. In all gammaherpesviruses, the DNA polymerase gene (ORF 9) is flanked upstream by ORF 8, the glycoprotein B, the most highly conserved glycoprotein in herpesviruses and downstream by ORF 10, a gene conserved within the gammaherpesviruses with unknown function (Figure [11](#F11){ref-type="fig"}). CODEHOPs were designed from conserved sequence blocks present in ORF 8 -- \"FREYA\" and \"GGMA\" and in ORF 10 \"GDWE2B\" (Table [4](#T4){ref-type="table"}). Using a combination of gene-specific primers obtained from the DNA polymerase sequence obtained above and the new CODEHOPs derived from flanking regions, overlapping PCR products spanning 331 bp of the glycoprotein B genes, 3,039 bp of the DNA polymerase genes, and 27 bp of the ORF 10 gene homolog were obtained for RFHVMn and RFHVMm \[[@B10]\]. ::: {#F11 .fig} Figure 11 ::: {.caption} ###### CODEHOP strategy to determine the complete sequence of a gammaherpesvirus DNA polymerase geneThe conserved linear order of the DNA polymerase gene, ie ORF 9, and the ORF 8 and ORF 10 flanking genes, characteristic of gammaherpesviruses, is shown. The position of the CODEHOP PCR primers used to obtain the sequence of the entire DNA polymerase gene of RFHVMn and RFHVMm is shown. The overlapping PCR products obtained using the CODEHOP and gene-specific primers are shown. ::: ![](1743-422X-2-20-11) ::: ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### CODEHOP and gene-specific primers developed for cloning the complete DNA polymerase gene of novel macaque rhadinoviruses. ::: Primer Gene Target Bias Sense 5\'\>3\' Sequence (degenerate codons are in lower case)^1^ ---------------- ------------- -------- --------- ------------------------------------------------------------ -------------------------------------- CODEHOP^2^ FREYA (32) gB^4^ γHV^3^ KSHV^4^ \+ TTTGACCTGGAGACTATGttymgngartayaa GGMA (128) gB γHV KSHV \+ ACCTTCATCAAAAATCCCttnggnggnatgyt CVNVA (64) DNA pol γHV KSHV \+ GACGACCGCAGCGTGTGCGTGaaygtnttyggnca CVNVB (64) DNA pol γHV KSHV \- TAAAAGTACAGCTCCTGCCCGaanacrttnacrca YFDKB (16) DNA pol γHV KSHV \- TTAGCTACTCCGTGGAGCagyttrtcraarta GDWE2B (8) ORF 10 γHV KSHV \- GAAGTGGCAGTTGGAGAGGCTGACCTCCcartcncc ^1^See legend to Table 1 for the I.U.B. code. ^2^See Figure 11 for the relative positions of the conserved sequence blocks from which the CODEHOPs were derived. The degree of degeneracy, ie the number of individual primers in the pool, is given in parentheses. ^3^The CODEHOPs were derived from the alignment of conserved genes within the gammaherpesvirus subfamily. ^4^The 5\' Clamp region was derived from the KSHV sequence flanking the 3\' core in order to target genes from RFHV, the macaque homolog of KSHV. ::: Using the CODEHOP strategy to characterize genomic regions within novel viral genomes ===================================================================================== Often the linear order of genes within the genomes of related viruses is maintained. Thus, the spacing and orientation of specific genes can be predicted in the genomes of related novel viruses. CODEHOP PCR primers can be utilized to obtain sequences within conserved genes which flank a targeted genomic region. Gene-specific PCR primers derived from these sequences can then used in long-range PCR to obtain the sequence of the entire genomic region between the flanking genes. We have utilized this approach to clone and characterize a portion of the divergent locus B of the genome of the macaque rhadinovirus, RFHVMn \[[@B11]\]. Divergent locus B was identified in KSHV and other rhadinoviruses and contains a number of viral homologs of cellular genes that have been captured during virus evolution \[[@B27]\]. Part of the divergent locus B of KSHV extends upstream of the ORF 9 DNA polymerase gene to a viral homolog of the thymidylate synthase (TS) gene situated approximately 4 kb away (Figure [12A](#F12){ref-type="fig"}). TS is a cellular gene and a non-functional pseudogene is present in humans. Viral TS homologs are well conserved and are found in several herpesvirus species, including KSHV, VZV, EHV2, HVS and AtHV3. To characterize the putative divergent locus B between the DNA polymerase and TS genes of RFHVMn, we targeted the TS gene for PCR amplification using the CODEHOP approach. ::: {#F12 .fig} Figure 12 ::: {.caption} ###### CODEHOP strategy to determine the complete sequence of a region of the divergent locus B of a macaque homolog of KSHV. A) the linear order of genes within the divergent locus B of KSHV \[35\]. Gene size in bp is shown in parantheses. B) The positions of the CODEHOP PCR primers used to obtain the DNA polymerase (GGMA/GDWE2B: see Figure 11) and thymidylate synthase (TS) (DMGLB/RHFGA) sequences are shown. The gene specific primers from the DNA polymerase (PolF1LR) and TS (TSR1LR) genes used in long range PCR are indicated. C) the linear order of genes within the divergent locus B of RFHVMn determined by the CODEHOP technique \[11\]. ::: ![](1743-422X-2-20-12) ::: Two conserved blocks of amino acids within the TS gene family containing 10 and 11 identical amino acids were chosen as candidates for CODEHOP design. The 10 amino acid \"RHFG\" upstream motif (Fig. [13](#F13){ref-type="fig"}) is completely conserved between the viral sequences, the human sequence and the human TS pseudogene. The 11 amino acid \"DMGL\" downstream motif (Fig. [13](#F13){ref-type="fig"}) while completely conserved between the viral and human sequences is not present in the cellular TS pseudogene (data not shown). Since the two motifs in the cellular TS gene are separated from each other by a large intron, CODEHOP PCR amplification of DNA containing a mixture of viral and cellular DNA should only produce a virus-specific \~280 bp PCR product (Fig. [12B](#F12){ref-type="fig"}). ::: {#F13 .fig} Figure 13 ::: {.caption} ###### ClustalW alignment of multiple herpesvirus TS sequences. The ClustalW output was obtained from the five TS sequences shown in Figure 15. The conserved \"RHFG\" and \"DMGL\" motifs which were chosen as targets in the design of the RHFGA (sense orientation) and DMGLB, DMGLXB and DMGLX1B (anti-sense orientation) CODEHOP PCR primers are indicated. ::: ![](1743-422X-2-20-13) ::: The design of the \"DMGLB\" CODEHOP from the conserved \"DMGL\" motif is shown in Figure [14](#F14){ref-type="fig"}. This primer was designed before the CODEHOP prediction program was available. Because RFHVMn is closely related to the gammaherpesvirus, KSHV, the \"DMGLB\" CODEHOP was biased towards gammaherpesviruses, in particular KSHV-like herpesviruses, in order to target the RFHV genomes. In Figure [14](#F14){ref-type="fig"}, the nucleotide sequences encoding the \"DMGL\" motif from the TS genes of KSHV, HVS and EHV2 were multiply aligned with the encoded amino acid sequence. Because \"DMGL\" was the downstream motif, the \"DMGLB\" CODEHOP was designed to be antisense, however, the complementary sequence of the primer is shown to identify codons (Figure [14](#F14){ref-type="fig"}). Thus, the degenerate core of the CODEHOP spans the codons for the aspartic acid (D), methionine (M), glycine (G), and leucine (L) of the motif, and is indicated in lower case letters in Figure [14B](#F14){ref-type="fig"}. The degenerate core provides all possibilities of the codons for these four conserved amino acids and thus has no bias. However, the nucleotides within the consensus region, shown in capitol letters, were chosen at each codon position to be similar to the sequence of KSHV (highlighted in Figure [14A](#F14){ref-type="fig"}), thus biasing the primer towards KSHV-like sequences. ::: {#F14 .fig} Figure 14 ::: {.caption} ###### Alignment of CODEHOPs with the nucleotide sequences of the \"DMGL\" motif in several herpesvirus TS genes. A) Nucleotide sequences encoding the \"DMGL\" motif in several rhadinoviruses. B) Complementary sequences of CODEHOP PCR primers derived from the \"DMGL\" motif. The sequence of the complementary strand of the primer is shown to identify the coding sequence. The actual PCR primer is the complement of the sequence. DMGLB was biased towards KSHV-like sequences by using the codons from the KSHV TS gene in the 5\' clamp region of the primer with KSHV-specific nucleotides highlighted (3\' region of the complementary coding strand shown). DMGLXB was predicted from the amino acid sequence block of the conserved \"DMGL\" motif using the CODEHOP software and utilizes the most common human codons for the amino acids in the 5\' clamp region, and is unbiased in design. The underlined sequence in the 5\' clamp region can form a stem-loop structure, shown in C. The CODEHOP PCR primer, DMGLX1B, is a revised version of DMGLXB to eliminate base pairing in the stem-loop structure by changing the highlighted cytosine (C) in Fig. 13-C. to an adenosine (A), boxed in Fig. 13-B. ::: ![](1743-422X-2-20-14) ::: The TS targeted CODEHOPs \"DMGLB\" and \"RHFGA\" (see Table [5](#T5){ref-type="table"}) were used in PCR amplification reactions with DNA isolated from retroperitoneal fibromatosis (RF) tumor tissue of a pig-tailed macaque, Macaca nemestrina, as described previously \[[@B10]\]. A PCR product of the predicted size (280 bp) was obtained and cloned and sequenced, see Fig. [12B](#F12){ref-type="fig"}. The sequence was 68% identical to the KSHV TS sequence and 64% identical to the TS sequence of RRV, a more distantly related gammaherpesvirus. A TS-specific primer, TSR1LR, derived from this sequence and a DNA polymerase-specific primer, PolF1LR, were chosen to amplify the region between the DNA polymerase and TS genes of RFHV (Table [5](#T5){ref-type="table"} and Figure [12B](#F12){ref-type="fig"}). Long range PCR amplification produced a PCR product of \~4.1 kb which was sequenced. The linear order and sequence of 5 novel genes present in the diverse region B of the RFHVMn virus was obtained (Figure [12C](#F12){ref-type="fig"}). Although region B of RFHV lacked a homolog of KSHV ORF 11, homologs of all the other KSHV genes in this region were present and in the same order within the genome \[[@B10]\]. ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### CODEHOP and gene-specific primers developed for cloning the divergent region B within the RFHV genome ::: Primer Gene Target Bias^2^ Sense 5\'\>3\' Sequence (degenerate codons are in lower case)^3^ ------------------------ ---------------- --------------------------- --------- ------------------------------------------------------------ ----------------------------------- CODEHOP^1^ RHFGA (48) TS^4^gene All cellular and viral TS KSHV^5^ \+ CCTGTTTACGGTTTCcartggagrcayttygg DMGLB (32) TS gene All cellular and viral TS KSHV \- GGCAATGTTAAAAGGAACTccnarncccatrtc RFHVMn-specific^6^ PolF1LR DNA polymerase NA^7^ NA \+ CCACCGTCCCAGACCAACGAAAGCGCCAGA TSR1LR TS gene NA NA \+ GTCTGCCTGGAATCCCGTGGATATACCAAA ^1^CODEHOP, consensus-degenerate hybrid oligonucleotide primers. The degree of degeneracy, ie the number of individual primers in the pool, is given in parentheses. ^2^Bias indicates the reliance on a specified subset of sequences for determination of the 3\' degenerate core or 5\' consensus clamp. ^3^See legend to Table 1 for the IUB code. ^4^TS, thymidylate synthase. ^5^Clamp region derived from the KSHV viral TS gene \[11\] ^6^Primer sequence derived from the RFHVMn sequence obtained by the CODEHOP technique ^7^NA, not applicable -- these are gene-specific primer ::: CODEHOP-mediated PCR -- a general approach to identify novel viral genes ======================================================================== In the previous sections of this review the CODEHOP assays and PCR primers that we have used to identify and characterize novel herpesvirus genes and genomes have been discussed in detail. However, CODEHOP-mediated PCR can also be used to target conserved genes from other virus families. A general flowchart detailing the specific steps involved in the CODEHOP procedure to identify novel viral genes is shown in Figure [15](#F15){ref-type="fig"}. This procedure is based on the CODEHOP prediction software that we have previously developed and made accessible over the internet as part of the BLOCKS database \[[@B2]\]. An example of this procedure is provided below where CODEHOP PCR primers targeting the \"DMGL\" motif of herpesvirus TS genes (introduced above) are designed using the web-based software. ::: {#F15 .fig} Figure 15 ::: {.caption} ###### CODEHOP assay flowchart to identify novel viral genes. The general approach to use CODEHOP-mediated PCR to identify novel viral genomes from a target virus family is shown schematically with links to specific software sites. ::: ![](1743-422X-2-20-15) ::: Using the web-based software to design CODEHOP PCR primers to a conserved viral gene ------------------------------------------------------------------------------------ The amino acid sequences of the TS genes from five herpesviruses were obtained using BLAST analysis of the NCBI protein database with the KSHV TS sequence as probe. The TS sequences from KSHV, VZV, EHV2, HVS and AtHV3 (Figure [16](#F16){ref-type="fig"}) were provided as input to ClustalW \[[@B28]\] and a multiple alignment was obtained. As shown in Figure [13](#F13){ref-type="fig"}, several regions of highly conserved sequences were present in the TS sequence alignment, and the positions of the \"RHFG\" and \"DMGL\" motifs targeted above are indicated. In order to predict CODEHOP PCR primers, the sequences of the TS genes were provided as input to the BlockMaker program of the Blocks Database \[[@B4]\] and a series of conserved sequence blocks were identified (ex., Gibbs Blocks, Figure [17](#F17){ref-type="fig"}). Alternatively, the ClustalW alignment, itself, could be provided as input to the \"Multiple alignment processor\" of the Blocks Database \[[@B29]\]. In order to compare a computer-predicted CODEHOP with the manually derived CODEHOP (DMGLB), the TS Block\_E containing the \"DMGL\" motif (Figure [17](#F17){ref-type="fig"}) was directly input to the CODEHOP program \[[@B3]\] using all default values except that the consensus region was elongated by increasing the temperature setting from the default 60°C to 70°C. The primers predicted from the complement of Block\_E were examined in order to obtain a primer from the complementary strand which could be used in conjunction with the upstream TS primer RHFGA, described above. The underlined primer targeting the \"DMGL\" motif was chosen and named DMGLXB (Figure [18](#F18){ref-type="fig"}) and was compared with the manually designed DMGLB primer in Figure [14](#F14){ref-type="fig"}. Whereas \"DMGLB\" was purposefully biased by using the KSHV sequences in the 5\' consensus clamp, the \"DMGLXB\" is \"unbiased\" in design with the 5\' consensus sequence derived from the most frequently used codons in the human genome. The DMGLXB sequence was examined for potential stem loop structures that could compromise the function of the primer. As shown in Figure [14](#F14){ref-type="fig"}, a putative stem-loop structure was identified which is indicated by the underlined nucleotides in Figure [14B](#F14){ref-type="fig"} and [14C](#F14){ref-type="fig"}. To destablize this structure, the proline codon within the \"DMGLGVP\" motif was changed from the computer predicted \"CCC\", the most frequently used codon in humans, to \"CCA\", another common human codon, as shown in Figure [14](#F14){ref-type="fig"}. This yielded a revised CODEHOP, called \"DMGLX1B\" (shown as the complementary sequence in Figures [14B](#F14){ref-type="fig"} and [14C](#F14){ref-type="fig"}), in which the stem-loop structure was destabilized by substituting an A for the highlighted C in Figure [14C](#F14){ref-type="fig"}. The DMGLX1B antisense primer could then be used in combination with the RHFGA sense primer to amplify unknown TS genes. ::: {#F16 .fig} Figure 16 ::: {.caption} ###### Herpesvirus thymidylate synthase protein sequences. The amino acid sequences of five herpesvirus TS genes used in the prediction of the DMGLXB and DMGLX1B CODEHOP PCR primers by the CODEHOP web-based software. The specific database accession numbers are indicated in the sequence title. ::: ![](1743-422X-2-20-16) ::: ::: {#F17 .fig} Figure 17 ::: {.caption} ###### Output of conserved sequence blocks obtained using the Gibbs method as implemented in the Block Maker program at the Blocks WWW server. Six conserved sequence blocks were identified in the five herpesvirus TS genes shown in Figure 15. Block TS\_E contains the DMGL motif (underlined) from which the DMGLXB and DMGLX1B complementary strand primers were derived. ::: ![](1743-422X-2-20-17) ::: ::: {#F18 .fig} Figure 18 ::: {.caption} ###### Output of the web-based CODEHOP software predicting complementary strand CODEHOP PCR primers for the conserved \"DMGL\" motif of herpesvirus TS genes. The TS\_E block from the BlockMaker output in Figure 17 was provided as input to the CODEHOP software \[3\] and the PCR primers derived from the complementary strand are shown. The predicted consensus amino acid sequence is shown and the DMGL motif is underlined in bold. The complementary strand CODEHOP PCR primer selected for use in amplifying unknown TS genes is underlined in bold. The 3\' degenerate core is shown in lowercase letters and the (len)gth and (degen)eracy are indicated. The 5\' consensus clamp is shown in uppercase letters and the score, (len)gth and predicted melting (temp)erature are indicated. ::: ![](1743-422X-2-20-18) ::: Other examples of CODEHOP PCR primers designed from multiple alignments of the herpesvirus DNA polymerase sequences using the Web-based CODEHOP software are shown in Figures [3](#F3){ref-type="fig"}, [4](#F4){ref-type="fig"}, [5](#F5){ref-type="fig"}, [6](#F6){ref-type="fig"}. The VYG1A primer designed from the conserved VYG motif shown in Figure [3](#F3){ref-type="fig"} is aligned with the original manually designed \"TGV\" and \"VYGA\" primers. The computer-predicted \"YGDTB\" primer designed from the conserved GDTD motif is aligned with the original \"IYG\" and \"GDTD1B\" primers (Figure [4](#F4){ref-type="fig"}). In the prediction of this primer, the conserved sequence block identified by BlockMaker from the sequences shown in Figure [4A](#F4){ref-type="fig"}, extended only from amino acid position 1 -- 10, which was the limit of the conserved sequence block determined by BlockMaker. The CODEHOP software indicated the necessity to add additional nucleotides to the 5\' end of the \"YGDTB\" primer to obtain the minimal length for the 5\' consensus region of the primer. As such, the amino acid sequences of block positions 11--13 were obtained manually and compared in order to derive the eight terminal nucleotides for \"YGDTB\" (overlined in Figure [4C](#F4){ref-type="fig"}). Conclusion ========== In this review, the utility of CODEHOP-mediated PCR for the identification of novel viruses and the characterization of new viral genes and genomic regions is presented. While the focus of this study was on the herpesvirus family, other virus families can be easily targeted using analogous approaches. We have previously developed successful CODEHOP assays targeting the reverse transcriptase genes of retroviruses and lentiviruses \[[@B2],[@B6]\]. Recently, the CODEHOP strategy has been used to develop assays to detect novel papillomaviruses targeting the highly conserved L1 protein \[[@B30]\]. With the CODEHOP strategy, molecular sequence data can be readily obtained for comprehensive virus phylogenies and tracing of evolutionary pathways. Furthermore, comparison of multiple representatives of homologous viral proteins can be of importance for understanding the protein structure and function and provided insight into virus-host relationships. List of Abbreviations ===================== CODEHOP, consensus-degenerate hybrid oligonucleotide primer; PCR, polymerase chain reaction; RFHV, retroperitoneal fibromatosis herpesvirus; KSHV, Kaposi\'s sarcoma-associated herpesvirus. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= Design, conception and preparation of the manuscript (TMR). Acknowledgements ================ The author would like to thank Emily Schultz, Greg Bruce, Lin Bennet, Brian Raden, Jon Ryan, and Kurt Strand for their help in developing the CODEHOP PCR strategy, Jorja and Steve Henikoff, of the Fred Hutchinson Cancer Research Center, for the creation and maintenance of the CODEHOP software and website, and Jeannette Stahli for editing advice.	PubMed Central	2024-06-05T03:55:55.633023	2005-3-15	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1079958/", "journal": "Virol J. 2005 Mar 15; 2:20", "authors": [ { "first": "Timothy M", "last": "Rose" } ] }
PMC1079959	Background ========== Geminiviruses are a large family of plant viruses with circular, single-stranded DNA (ssDNA) genomes packaged within geminate particles. The family Geminiviridaeis divided into four genera (Mastrevirus, Curtovirus, Topocuvirus, and Begomovirus) according to their genome organizations and biological properties \[[@B1],[@B2]\]. Members of the genus Begomovirushave caused significant yield losses in many crops worldwide \[[@B3]\] and are transmitted by whiteflies (Bemisia tabaci) to dicotyledonous plants. The genome of cassava mosaic geminiviruses (CMGs) in the genus Begomovirusconsists of two DNA molecules, DNA-A and DNA-B, each of about 2.8 kbp \[[@B1]\], which are responsible for different functions in the infection process. DNA-A encodes genes responsible for viral replication \[AC1 (Rep), and AC3 (Ren)\], regulation of gene expression \[AC2 (Trap)\] and particle encapsidation \[AV1 (CP)\]. DNA-B encodes for two proteins, BC1 (MP) and BV1 (NSP) involved in cell-to-cell movement within the plant, host range and symptom modulation \[[@B1]\]. CMGs have been reported from many cassava-growing countries in Africa and the cassava mosaic disease (CMD) induced by them constitutes a formidable threat to cassava production \[[@B4]\]. Representatives of six distinct CMG species have been found to infect cassava in Africa: African cassava mosaic virus(ACMV), East African cassava mosaic virus(EACMV), East African cassava mosaic Cameroon virus(EACMCV), East African cassava mosaic Malawi virus(EACMMV), East African cassava mosaic Zanzibar virus(EACMZV) and South African cassava mosaic virus(SACMV) \[[@B5]\]. Recent studies have uncovered much variation in CMGs including evidence that certain CMGs, when present in mixtures, employ pseudo-recombination or reassortment strategies and recombination at certain hot spots such as the origin of replication \[[@B6]-[@B10]\] resulting in the emergence of \'new\' viruses with altered virulence. For instance, an ACMV-EACMV recombinant component A, designated EACMV-UG2, and a pseudo-recombinant component B, designated EACMV-UG3 \[[@B10]\], have been implicated in the pandemic of severe CMD currently devastating cassava in much of east and central Africa \[[@B4]\]. In 1997, only ACMV and EACMV were known to occur in Tanzania with the former occurring only in the western part of the country \[[@B11]\]. The discovery of EACMZV on the island of Zanzibar \[[@B12]\] together with the recent spread into Tanzania of the EACMV-UG2 associated pandemic of severe CMD \[[@B4],[@B13]\] has aggravated the CMD situation. Consequently, there is much to be learned about the identity, distribution, molecular variability, and the threat that these emerging geminiviruses pose to cassava production in Tanzania and more generally in Africa. In 1997, the first recombination between two species of geminiviruses was recorded \[[@B7],[@B8]\]. This mechanism is now known to be widely used by all geminiviruses and is probably the most important molecular mechanism for generating genetic changes that allow novel geminiviruses to exploit new ecological niches \[[@B2],[@B14]\]. This paper describes the results of a molecular study of the sequences of CMGs collected from the major cassava-growing areas of Tanzania in an effort towards identifying, determining molecular variability and mapping the distribution of CMGs. In addition, because East Africa seems to be unusually rich in virus biodiversity and because the most recent cassava pandemic was first reported in East Africa, we investigated the extent of inter-CMG recombinations and examined their role in the evolution of CMGs in Africa. Results ======= Assessment of CMD symptoms -------------------------- Over 80% of the cassava plants in the fields showed severe CMD symptoms with cassava in the Lake Victoria basin expressing the most severe symptoms followed by that from the southern regions. Symptoms of infected cassava samples collected in the field were reproduced in controlled conditions to examine symptom variability. From a total of 35 selected cuttings planted, 25 (71%) were successfully established in the growth chamber. In all cases, regardless of the cultivar, symptoms expressed in the field, whether moderate or severe, were reproduced in the growth chamber and plants did not recover from the disease even 12 months after planting (Fig. [2](#F2){ref-type="fig"}). Likewise, plants that displayed moderate symptoms in the field showed a similar symptom in the growth chamber as was the case for plants singly-infected with ACMV-\[TZ\] (Fig. [2](#F2){ref-type="fig"}). ::: {#F2 .fig} Figure 2 ::: {.caption} ###### CMD symptoms on naturally infected cassava plants (A, C, E and G) in the field with their corresponding plants raised from field-collected cuttings maintained in the growth chamber (B, D, F and H). Only plants containing single virus infection are shown. Plants A and B contained a single infection of EACMV-\[KE/TZM\], C and D contained ACMV-\[TZ\], E and F were infected by EACMCV-\[TZ1\] and G and H by EACMV-UG2 \[TZ10\]. ::: ![](1743-422X-2-21-2) ::: Detection of viral genomic components ------------------------------------- PCR amplification products (2.7--2.8 kbp) were observed for all the CMG isolates tested using primer UNIF/UNIR (Table [1](#T1){ref-type="table"}) designed to amplify near-full-length DNA-A of CMGs. Bands were not observed with the negative control (nucleic acid preparation from healthy cassava plants). Similarly, a specific (2.7 kbp) product was observed when using abutting primers TZ1B-F/R designed from a 560 bp DNA-B fragment initially PCR-amplified using universal primers EAB555/F and EAB555R for general detection of CMGs DNA-B. DNA-B partial fragments (544--560 kbp) were consistently amplified by PCR using primers EAB555-F and EAB555-R (Table [1](#T1){ref-type="table"}) for all the CMD-diseased samples previously shown to contain EACMV isolates collected from major cassava-growing areas in Tanzania \[[@B13]\]. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### List of the oligonucleotide primers used in this study for amplification of cassava mosaic geminiviruses from Tanzania (^a^nfl = near-full length, ps = partial sequence) ::: Primer name Nucleotide sequence (5\'→3\') Begomovirus isolate DNA component ----------------- ----------------------------------- ------------------------- ------------------- UGT-F TCGTCTAGAACAATACTGATCGGTCTCC EACMV-KE-\[TZT\] DNA-A fl^a^ UGT-R CGGTCTAGAAGGTGATAGCCGAACCGGGA EACMV-KE-\[TZT\] DNA-A fl 3T-F ACGTCTAGAACAATACTGATCGGTCTC EACMV-TZ-\[YV\] DNA-A fl 3T-R GTGCTCTAGAAGGTGATAGCCGAACCGGGA EACMV-TZ-\[YV\] DNA-A fl TZ1B-F GCGCGGAATCACTTGTGAAGCAGTCGT EACMCV-\[TZ1\] DNA-B fl TZ1B-R GCCGGGATTCGGTGAGTGGTTTACATCAC EACMCV-\[TZ1\] DNA-B fl EAB555/F TACATCGGCCTTTGAGTCGCATGG CMGs BC1/CR EAB555/R CTTATTAACGCCTATATAAACACC CMGs BC1/CR UNI/F KSGGGTCGACGTCATCAATGACGTTRTAC CMGs DNA-A nfl UNI/R AARGAATTCATKGGGGCCCARARRGACTGGC CMGs DNA-A nfl AT-F GTGACGAAGATTGCATTCT ACMV-\[TZ\] DNA-A ps AT-R AATAGTATTGTCATAGAAG ACMV-\[TZ\] DNA-A ps ATZ1-F TAAGAAGATGGTGGGAATCC EACMCV-\[TZ1\] DNA-A ps ATZ-R CGATCAGTATTGTTCTGGAAC EACMCV-\[TZ1\] DNA-A ps TZ7-F TGGTGGGAATCCCACCTT EACMCV-\[TZ7\] DNA-A ps TZ7-R GTATTGTTATGGAAGGTGATA EACMCV-\[TZ7\] DNA-A ps TZM-F TATATGATGATGTTGGTC EACMV-UG2Svr-\[TZ10\] DNA-A ps TZ10-R TAGAAGGTGATAGCCGTA EACMV-UG2Svr-\[TZ10\] DNA-A ps TZM-F TATATGATGATGTTGGTC EACMV-KE-\[TZM\] DNA-A ps TZM-R TAGAAGGTGATAGCCGAAC EACMV-KE-TZM\] DNA-A ps ::: Complete nucleotide sequence characteristics of CMGs from Tanzania ------------------------------------------------------------------ The complete DNA-A sequences of seven representative CMGs from the major cassava-growing areas were determined from the representative isolates selected and grown in the growth chambers. An ACMV isolate from Tanzania (ACMV-\[TZ\]) was shown to be most closely related to ACMV-UGMld from Uganda with a sequence identity of 97%. Its DNA-A nucleotide (nt) sequence was established to be 2779 nts in length. It has a high overall sequence identity (\> 90%) with all other published sequences of ACMV isolates (Table [2](#T2){ref-type="table"}) with which it clusters in the phylogenetic tree presented in Figure [3](#F3){ref-type="fig"}. The DNA-A sequence organization was typical of a begomovirus, with two open reading frames (ORFs) (AV2 and AV1) in the virion-sense DNA, and four ORFs (AC1 to AC4) in the complementary sense, separated by an intergenic region (IR). Complete nt sequences of the DNA-A genomes of the different Tanzanian EACMV and ACMV isolates were compared with published sequences (Table [2](#T2){ref-type="table"}). ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Nucleotide sequence identities (percentages) of the DNA-A full-length of cassava mosaic geminiviruses from Tanzania and other geminiviruses from Africa and the Indian sub-continent. Values above 89% are in bold and names of isolates from Tanzania are in bold. ::: Virus Isolate ACMV-\[TZ\] EACMCV-\[TZ1\] EACMCV-\[TZ7\] EACMV-\[KE/TZT\] EACMV-\[KE/TZM\] EACMV-\[TZ/YV\] EACMV-UG2 \[TZ10\] ------------------------ ----------------- -------------------- -------------------- ---------------------- ---------------------- --------------------- ------------------------ ACMV-\[CM\] 95 68 68 70 70 69 73 ACMV-\[CM/DO2\] 95 68 68 70 70 69 73 ACMV-\[IC\] 96 68 68 70 71 70 73 ACMV-\[KE\] 96 68 68 70 70 70 73 ACMV-\[NG\] 95 68 68 70 70 70 73 ACMV-\[NG/Ogo\] 96 68 68 70 70 70 73 ACMV-UGMld 97 68 68 70 71 70 73 ACMV-UGSVr 96 68 68 70 71 70 74 ACMV-\[TZ\] \- 68 68 70 70 70 73 EACMCV-\[CM\] 67 90 89 87 87 85 84 EACMCV-\[CI\] 67 90 90 88 87 86 85 EACMCV-\[TZ1\] 68 \- 96 88 88 87 85 EACMCV-\[TZ7\] 68 96 \- 88 88 87 85 EACMMV-\[K\] 71 81 81 87 88 86 87 EACMMV-\[MH\] 71 81 81 87 88 86 88 EACMV-\[KE/K2B\] 70 88 88 97 96 94 92 EACMV-\[TZ\] 69 88 88 94 94 95 91 EACMV-\[KE/TZT\] 70 88 88 - 95 93 92 EACMV-\[KE/TZM\] 70 88 88 96 \- 94 92 EACMV-\[TZ/YV\] 70 87 87 94 93 - 90 EACMV-UG2 73 85 85 92 92 92 98 EACMV-UG2Mld 73 86 86 93 92 92 99 EACMV-UG2Svr 73 86 86 93 92 92 99 EACMV-UG2 \[TZ10\] 73 85 85 92 92 91 - EACMZV-\[ZB\] 72 80 80 86 86 86 83 EACMZV-\[KE/Kil\] 72 79 79 86 86 85 83 SACMV-\[ZA\] 74 73 73 80 80 79 80 SACMV-\[ZW\] 74 73 73 80 80 80 80 SACMV-\[M12\] 74 73 73 80 80 80 80 SLCMV-\[Col\] 73 67 67 67 67 67 67 TGMV-\[Com\] 58 59 59 59 59 59 59 ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Phylogenetic tree (1000 boot strap replications) showing the DNA-A complete nucleotide sequence relationships between the seven Tanzanian cassava mosaic geminivirus isolates (in blue) and other cassava mosaic geminiviruses. Tomato golden mosaic virus(TGMV-YV) (K02029) was used as the out group. Abbreviations and accession numbers are: ACMV-\[CI\], African cassava mosaic virus-\[Côte d\'Ivoire\] (AF259894); ACMV-\[NG/Ogo\], African cassava mosaic virus-\[Nigeria-Ogo\] (AJ427910); ACMV-\[CM/D02\], African cassava mosaic virus-\[Cameroon D02\] (AF366902); ACMV-\[CM/D03\], African cassava mosaic virus-\[Cameroon D03\] (AY211885); ACMV-\[CM/Mg\], African cassava mosaic virus-\[Cameroon Mg\] (AY211884); ACMV-\[CM\], African cassava mosaic virus-\[Cameroon\] (AF112352); ACMV-\[KE\], African cassava mosaic virus-\[Kenya\] (J02057); ACMV-\[NG\], African cassava mosaic virus-\[Nigeria\] (X17095); ACMV-UGMld, African cassava mosaic virus-Uganda mild (AF126800); ACMV-UGSvr, African cassava mosaic virus-Uganda severe (AF126802); EACMCV-\[CM/KO\], East African cassava mosaic Cameroon virus-\[Cameroon KO\] (AY211887); EACMCV-\[CM\], East African cassava mosaic Cameroon virus-\[Cameroon\] (AF112354); EACMCV-\[CI\], East African cassava mosaic Cameroon virus-\[Côte d\'Ivoire\] (AF259896); EACMMV-\[K\], East African cassava mosaic Malawi virus-\[K\] (AJ006460); EACMMV-\[MH\], East African cassava mosaic Malawi virus-\[MH\] (AJ006459); EACMV-\[KE/k2B\], East African cassava mosaic virus\[Kenya-K2B\] (AJ006458); EACMV-\[TZ\], East African cassava mosaic virus-\[Tanzania\] (Z53256); EACMV-UG2\[2\], East African cassava mosaic virus-Uganda2\[2\] (Z83257); EACMV-UG2Mld, East African cassava mosaic virus-Uganda2 mild (AF126804); EACMV-UG2Svr, East African cassava mosaic virus-Uganda2 severe (AF126806); EACMZV-\[KE/Kil\], East African cassava mosaic Zanzibar virus-\[Kenya -Kil\] (AJ516003); EACMZV-\[ZB\], East African cassava mosaic Zanzibar Virus-- \[Zanzibar\] (AF422174); ICMV-\[Adi2\], Indian cassava mosaic virus-- \[Adivaram 2\] (AJ575819); ICMV-\[Mah\], Indian cassava mosaic virus-- \[Maharashstra\] (AJ314739); ICMV-\[Mah2\], Indian cassava mosaic virus-- \[Maharashstra 2\] (AY730035); ICMV-\[Tri\], Indian cassava mosaic virus-- \[Trivandrum\] (Z24758); SACMV-\[M12\], South African cassava mosaic virus-\[Madagascar M12\] (AJ422132); SACMV-\[ZA\], South African cassava mosaic virus-- \[South Africa\] (AF155806); SACMV-\[ZW\], South African cassava mosaic virus-- \[Zimbabwe\] (AJ575560); SLCMV-\[Adi\], Sri-Lankan cassava mosaic virus-\[Adivaram\] (AJ579307); SLCMV-\[Col\], Sri-Lankan cassava mosaic virus-\[Colombo\] (AF314737); SLCMV-\[Sal\], Sri-Lankan cassava mosaic virus-\[Salem\] (AJ607394). ::: ![](1743-422X-2-21-3) ::: Two isolates, TZ1 and TZ7, with 2798 and 2799 nts respectively, collected from Mbinga district in southwestern Tanzania, were most closely related to isolates of the species East African cassava mosaic Cameroon virusfrom Cameroon and Ivory Coast, West Africa, (EACMCV-\[CM\], -\[CI\]), with 89--90% nt sequence identity. They are clearly isolates of EACMCV and we have named them EACMCV-\[TZ1\] and EACMCV-\[TZ7\] to indicate that they were from Tanzania and to distinguish them from the original EACMCV-\[CM\] isolate from Cameroon. The two isolates were also virtually identical to one another having high overall DNA sequence conservation (93% nt sequence identity). Phylogenetic analysis of the DNA-A nt sequences grouped EACMCV-\[TZ1\] and EACMCV-\[TZ7\] in the same cluster with EACMCV-\[CM\] and EACMCV-\[CI\] (Fig. [3](#F3){ref-type="fig"}). The complete nt sequence of the EACMCV-\[TZ1\] DNA-B component was determined to be 2726 nts long and had the highest sequence identity (85%) with EACMCV-\[CM\] DNA-B with which it is grouped in the phylogenetic tree (Fig. [4](#F4){ref-type="fig"}). It had less than 72% homology with DNA-Bs of other EACMV isolates from East Africa. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Phylogenetic tree (1000 bootstrap replications) obtained from comparison of the complete nucleotide sequence of EACMCV-\[TZ1\] DNA-B, partial B component sequences from Tanzania (TZBx) and available cassava mosaic geminivirus DNA-B component sequences. Tomato golden mosaic virus(TGMV-YV) (K02030) was used as the out-group. Abbreviations and accession numbers are: ACMV-\[CI\], African cassava mosaic virus-\[Côte d\'Ivoire\] (AF259895); ACMV-\[NG/Ogo\], African cassava mosaic virus-\[Nigeria-Ogo\] (AJ427911); ACMV-\[CM/KT\], African cassava mosaic virus-\[Cameroon KT\] (AY211886); ACMV-\[CM\], African cassava mosaic virus-\[Cameroon\] (AF112353); ACMV-\[KE\], African cassava mosaic virus-\[Kenya\] (J02058); ACMV-\[NG\], African cassava mosaic virus-\[Nigeria\] (X17096); ACMV-UGMld, African cassava mosaic virus-Uganda mild (AF126801); ACMV-UGSvr, African cassava mosaic virus-Uganda severe (AF126803); EACMCV-\[CM\], East African cassava mosaic Cameroon virus-\[Cameroon\] (AF112355); EACMCV-\[CI\], East African cassava mosaic Cameroon virus-\[Côte d\'Ivoire\] (AF259897); EACMV-UG3Mld, East African cassava mosaic virus-Uganda3 mild (AF126805); EACMV-UG3Svr, East African cassava mosaic virus-Uganda3 severe (AF126807); EACMZV-\[KE/Kil\], East African cassava mosaic Zanzibar virus-\[Kenya -Kil\] (AJ628732); EACMZV-\[ZB\], East African cassava mosaic Zanzibar Virus-- \[Zanzibar\] (AF422175); ICMV-\[Kat\], Indian cassava mosaic virus-- \[Kattukuda\] (AJ575821); ICMV-\[Ker\], Indian cassava mosaic virus-- \[Kerala\] (AJ575823); ICMV-\[Mah\], Indian cassava mosaic virus-- \[Maharashstra\] (AJ314740); ICMV-\[Mah2\], Indian cassava mosaic virus-- \[Maharashstra 2\] (AY730036); ICMV-\[Tri\], Indian cassava mosaic virus-- \[Trivandrum\] (Z24759); SACMV-\[ZA\], South African cassava mosaic virus-- \[South Africa\] (AF155807); SLCMV-\[Adi\], Sri-Lankan cassava mosaic virus-\[Adivaram\] (AJ579308); SLCMV-\[Col\], Sri-Lankan cassava mosaic virus-\[Colombo\] (AF314738). ::: ![](1743-422X-2-21-4) ::: The complete DNA-A genome of CMG isolates from Yombo Vituka (YV) and Tanga (TZT) in the coastal area of Tanzania were determined to be 2800 and 2801 nts long respectively. Isolate YV showed high (95%) overall nt sequence identity with previously characterized EACMV-\[TZ\] and is therefore named EACMV-\[TZ/YV\] in the Dar-es-Salaam region. It also had high overall sequence identity (87--96%) with other Tanzanian EACMV isolates characterized in this study (Table [2](#T2){ref-type="table"}). Phylogenetic analysis of the complete nt sequence of EACMV-\[TZ/YV\] grouped it with its closest relative, EACMV-TZ (Fig. [3](#F3){ref-type="fig"}). CMG isolate TZT had high sequence identity (96.5%) with EACMV-\[KE/K2B\] from Kenya and is named EACMV-\[KE/TZT\]. Similarly, another CMG isolate (TZM) from the Mara region in the Lake Victoria zone was found to have high overall sequence identity (96%) with EACMV-\[KE/K2B\] and we have named it EACMV-\[KE/TZM\]. This isolate, 2805 nts in length, together with EACMV-\[KE/TZT\], clustered with EACMV-\[KE/K2B\] in the phylogenetic tree (Fig. [3](#F3){ref-type="fig"}). Another isolate from Kagera region in northwestern Tanzania (TZ10) showed very high overall DNA-A nt sequence identity (98.8%) with the published sequence of EACMV-UG2Svr. Its complete DNA-A nt sequence was 2804 nts long and it was named EACMV-UG2 \[TZ10\]. Determination of genetic diversity of EACMV DNA-B using partial sequences ------------------------------------------------------------------------- The diversity of different CMG isolates was analyzed using a partial DNA-B genomic region spanning the N-terminal region of BC1 to the intergenic region (IR). Identities of these sequences with those of the corresponding DNA-B genomic regions of other CMGs in GenBank were determined. Generally, the EACMV isolates showed little genetic divergence amongst one another and isolates collected from the same area displayed high nt sequence identity. Isolates TZB1 and TZB7 from the southern part of Tanzania shared the highest (98%) nt sequence identity followed by TZB3 and TZB8 (94%) as well as TZB and TZB10, all from the east coast area. TZB2 was most closely related to and shared 91% sequence identity with TZB4, both collected from the coastal area. None of the isolates from the south or coastal areas shared \>85% nt sequence identity with those from the Lake Victoria basin (TZB9 and TZB12). The phylogenetic tree generated from a multiple alignment of 13 EACMV isolates with selected bipartite begomovirus sequences and EACMCV-\[TZ1\] B component is shown in Figure [4](#F4){ref-type="fig"}. All 13 Tanzanian isolates studied clustered with the reference EACMVs, with TZB6 being most closely related to Ugandan isolates (EACMV-UG3Svr, EACMV-UG3Mld and EACMV-UG1) (Fig. [4](#F4){ref-type="fig"}) sharing 97% nt sequence identity. Four isolates (TZB3, TZB5, TZB8 and TZB9) formed a closely related group, with TZB8 and TZB9 being the most closely related. Isolates TZMB, TZB5 and TZB11 each grouped separately. None of the EACMV isolates grouped with ICMV and SLCMV from the Indian subcontinent (Fig. [4](#F4){ref-type="fig"}). Capsid protein (CP) gene sequence analysis and comparison with selected viruses ------------------------------------------------------------------------------- The CP gene sequences of the seven CMGs identified in our study were compared to published sequences (Table [3](#T3){ref-type="table"}). ACMV-\[TZ\] shared the highest nt sequence identity (97.4%) with ACMV-UGMld from Uganda followed by ACMV-\[CM\], an isolate from Cameroon. The lowest sequence identity (63.2%) was recorded with TGMV-YV (Table [3](#T3){ref-type="table"}), an American begomovirus. Both EACMCV-\[TZ1\] and EACMCV-\[TZ7\] were more than 92% identical to EACMCV-\[CM\], but they also had very high nt sequence identity (95%) with EACMZV from Zanzibar and EACMV-\[KE/K2B\] (Table [3](#T3){ref-type="table"}) and 96% between each other. Interestingly, EACMV-\[KE/TZT\] and EACMV-\[KE/TZM\] collectively shared high (97%) identity with EACMZV followed by EACMV-\[KE/K2B\](96--97%) and up to 96% between each other. Furthermore the EACMV-\[TZ/YV\] CP gene sequence showed very high identity with EACMV-\[TZ\] (96%) and EACMZV (96%) followed by EACMV-\[KE/K2B\](95%) (Table [3](#T3){ref-type="table"}). The EACMV-UG2 \[TZ10\] sequence shared a very high nt sequence identity (99%) with EACMV-UG2Svr from Uganda and high identity (98--99%) with other Ugandan isolates of EACMV. As expected, EACMV-UG2 \[TZ10\] shared 90% sequence homology with ACMV (Table [3](#T3){ref-type="table"}), suggesting it contained the recombination at the CP gene level previously reported \[[@B7],[@B8]\] for EACMV-UG2. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### CP gene nucleotide sequence identity (%) of cassava mosaic geminiviruses from Tanzania and other published CMG CP sequences. Values above 89% are in bold and names of isolates from Tanzania are in blue. ::: Virus Isolate ACMV-\[TZ\] EACMCV-\[TZ1\] EACMCV-\[TZ7\] EACMV-\[KE/TZT\] EACM-\[KE/TZM\] EACMV-\[TZ/YV\] EACMV-UG2 \[TZ10\] ------------------------ ----------------- -------------------- -------------------- ---------------------- --------------------- --------------------- ------------------------ ACMV-\[CM\] 97 77 77 78 79 77 90 ACMV-\[CI\] 96 77 78 78 79 77 90 ACMV-\[KE\] 97 76 76 77 78 76 90 ACMV-\[NG\] 96 77 77 78 78 77 90 ACMV-UGMld 97 76 77 78 78 76 90 ACMV-\[TZ\] \- 77 77 78 78 77 89 EACMCV-\[CM\] 77 94 94 95 96 93 84 EACMCV-\[TZ1\] 77 \- 97 95 96 94 84 EACMCV-\[TZ7\] 77 97 \- 95 97 95 84 EACMMV-\[K\] 77 80 80 80 80 80 79 EACMMV-\[MH\] 77 79 80 80 80 80 79 EACMV-\[KE/K2B\] 77 95 96 96 97 96 84 EACMV-TZ 77 95 95 96 97 96 85 EACMV-\[KE/TZT\] 78 95 95 \- 97 95 85 EACMV-\[KE/TZM\] 78 96 97 97 - 97 84 EACMV-\[TZ/YV\] 77 94 95 95 96 - 84 EACMV-UG2 90 84 84 85 85 84 99 EACMV-UG2Mld 89 84 84 85 85 84 98 EACMV-UG2Svr 90 84 84 85 85 84 99 EACMV-UG2 \[TZ10\] 89 84 84 85 84 84 - EACMZV-\[ZB\] 78 96 96 97 97 96 85 SACMV-\[ZA\] 77 78 79 80 79 79 73 ICMV-\[Tri\] 74 73 73 74 74 73 64 TGMV-\[Com\] 63 65 65 64 64 65 78 ::: A phylogenetic analysis of the CP of Tanzanian CMGs yielded a tree (Fig. [5](#F5){ref-type="fig"}) that was in agreement with the relationship predicted by pairwise sequence comparison (Table [4](#T4){ref-type="table"}). ACMV-\[TZ\] clustered with other ACMV isolates while EACMV-UG2 \[TZ10\] grouped with Ugandan isolates of EACMV. EACMCV-\[TZ1\], EACMCV-\[TZ7\], EACMV-\[TZ/YV\], and the two viruses, EACMV-\[KE/TZT\] and EACMV-\[KE/TZM\] clustered with other EACMV isolates from either Cameroon or Kenya. No CMG isolate identified in this study clustered with EACMMV from Malawi, SACMV from South Africa, ICMV, or SLCMV from the Indian sub-continent when their CP gene nucleotide sequences were compared (Fig. [5](#F5){ref-type="fig"}). ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Phylogenetic tree of the coat protein gene (CP) nucleotide sequences of the cassava mosaic geminivirus isolates from Tanzania and other cassava begomoviruses (1000 bootstrap replications). Sequence of tomato golden mosaic virus (TGMV-YV) was used as the out-group. Abbreviations and accession numbers can be found in Figure 3. ::: ![](1743-422X-2-21-5) ::: ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Percent similarity (in the upper triangle) in the nucleotide sequence of the common region of East and West African isolates of EACMCV. Values of 89% and above are in bold. ::: Virus isolate EACMCV-\[TZ1\] CRA EACMCV-\[TZ7\] CRA EACMCV-\[TZ1\] CRB EACMCV-\[CM\] CRA EACMCV-\[CM\] CRB EACMCV-\[CI\] CRA EACMCV-\[IC\] CRB ------------------------ ------------------------ ------------------------ ------------------------ ----------------------- ----------------------- ----------------------- ----------------------- EACMCV-\[TZ1\] CRA \\\* 80 80 89 76 82 76 EACMCV-\[TZ7\] CRA \\\* 86 88 74 82 73 EACMCV-\[TZ1\] CRB \\\* 91 80 82 78 EACMCV-\[CM\] CRA \\\* 86 91 83 EACMCV-\[CM\] CRB \\\* 78 97 EACMCV-\[CI\] CRA \\\* 77 EACMCV-\[CI\] CRB \\\* ::: The common regions (CRs) of the Tanzanian CMGs ---------------------------------------------- The conserved nonanucleotide in the hairpin-loop, TAATATTAC, that is characteristic of the members of the family Geminiviridaeand the AC1 TATA box, were identified in the CR sequences of all the Tanzanian CMGs (Fig. [6a,6b](#F6){ref-type="fig"}). The CR of ACMV-\[TZ\] was 170 nts long while those for EACMV were between 152 and 157 nts in length. When the CR sequence of ACMV-\[TZ\] was compared and aligned to the published CR sequences of other cassava-infecting ACMV isolates from Africa (Fig. [6a](#F6){ref-type="fig"}), it was apparent that ACMV-\[TZ\] was virtually identical to all ACMV isolates. The repeated motif upstream the TATA box for all the published ACMV isolates was AATT[GGAGA]{.underline} (Fig. [6a](#F6){ref-type="fig"}). The motif for ACMV-\[TZ\], AATT[GGAGA]{.underline}, was identical. Figure [6b](#F6){ref-type="fig"} presents the alignment of the CRs of the Tanzanian EACMVs with sequences of all published EACMVs. It was found that all the isolates contained the various features characteristic of begomoviruses. The putative Rep-binding sequences (iterons) were [GGTGG]{.underline}AAT[GGGGG]{.underline} for all the Tanzanian isolates except EACMV-\[TZ/YV\] that had different iterons ([GGGGG]{.underline}AAC[GGGGG]{.underline}) and a total of 23 mismatches in the entire CR. It is worth noting that although the genomes of the two isolates of EACMZV are EACMV-based, their CRs are more similar to ACMV than to EACMV and the iteron is AATTGGAGA. ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Alignment of common region (CR) nucleotide sequences of the DNA-A (CRA) and DNA-B (CRB) of ACMV (A) and EACMV (B) isolates from Tanzania with the related isolates of ACMV and EACMV from the database sequences. The TATA box for AC1 is boxed in black. The putative Rep binding iterative sequences (iterons) are boxed in green and purple. The conserved nonanucleotide sequences TAATATTAC together with its stem loop are boxed in blue and green respectively. The conserved sequence 3\'-end of the TATA box is boxed in red and the so-called \"variable region\" is boxed in grey. Virus sequences from Tanzania are written in blue. The accession numbers of the sequences from GenBank are indicated on the right of the virus abbreviation names and the significance of these abbreviations can be found in the legend of Figures 3 and 4. ::: ![](1743-422X-2-21-6) ::: The comparisons of the nt sequences of the CRs of Tanzanian CMGs with other CMGs revealed high sequence identity (\> 90%) of ACMV-\[TZ\] to published sequences of other ACMV isolates and low identity (61--62%) to EACMV species. Similarly, all the Tanzanian EACMV isolates were related with sequence identities of 83--97% between CRs of the DNA-A and DNA-B. The CR of EACMV-\[TZ/YV\] showed a relatively low sequence identity to other isolates. EACMCV-\[TZ1\] (DNA-A and -B) and the EACMCV-\[TZ7\] showed high nt sequence identity to EACMCV (Table [4](#T4){ref-type="table"}). Geographical distribution of the CMGs in Tanzania ------------------------------------------------- The representative isolates sequenced here have been chosen because they represent a range of different RFLP patterns found during a large set of 485 samples collected throughout Tanzania \[[@B13]\]. However, the selection of isolates to sequence was based on the differences in RFLP patterns and not on their frequency of appearance in the country. Figure [7](#F7){ref-type="fig"} shows the different locations of these samples represented by the isolates sequenced here. The EACMCV-\[TZ1\] was the most widespread, found in 50 samples located mainly in the southern part of Tanzania in the Mbinga District of Ruvuma Region. EACMCV-\[TZ7\], the close relative of EACMCV-\[TZ1\], was found only in one sample in the same district of Mbinga. EACMV-\[KE/TZT\] was found only in the coastal areas, in ten samples, mainly in Tanga and Pwani regions. EACMV-\[KE/TZM\] was found in ten samples, only in the Mara Region of the Lake Victoria Basin and to a very limited extent on the island of Ukerewe in Lake Victoria. The rest of the CMGs, EACMV-UG2 \[TZ10\], ACMV-\[TZ\] as well as EACMV-\[TZ/YV\], had a limited geographical distribution (Fig. [7](#F7){ref-type="fig"}). ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Map of the location of the different types of viruses present in Africa and inlay map of Tanzania showing the location of the completely sequenced CMG clones in that study as well as the localization of the distribution of viruses similar to these clones by RFLP mapping \[13\]. On the African map the symbols represent an approximate positioning of the viruses for which we have complete sequence information and not those for which we have either partial sequence information or serological data only. The significance of the different stars and shaded areas and arrows is indicated in the legend boxes in the figure. The solid red arrow represents the current direction of spread of the CMD pandemic, while the faded green and blue arrows represent possible \"routes\" of evolution of EACMV-like viruses and EACMCV in the past. ::: ![](1743-422X-2-21-7) ::: Comparisons of the East African and West African isolates of EACMCV ------------------------------------------------------------------- ### i) Comparisons of the A components of EACMCV-\[TZ\] The East African cassava mosaic Cameroon virusisolates from Tanzania (EACMCV-\[TZ1, TZ7\]) are very typical isolates of the species East African cassava mosaic Cameroon virus. The A component was 89 to 90% identical to the isolates from Cameroon and Ivory Coast and the 300 nts that differ are scattered all along the genome. In addition, the A components from East Africa showed the typical recombination already noted in the West African isolates, i.e.a fragment of about 800 nts not of EACMV origin, covering AC2-AC3 and the C-terminus of AC1 (Fig. [8A](#F8){ref-type="fig"}). ::: {#F8 .fig} Figure 8 ::: {.caption} ###### Pairwise sequence comparisons of EACMCV-\[TZ1, TZ7\] DNA-A (A) and DNA-B (B). Each curve represents a sequence comparison along the linearized virus genomes of a chosen pair of viruses. The correspondance for each colored curve is given in the figure. The dissimilarity index (Y axis) is the percentage of dissimilarity over a window of 50 nucleotides. The curves under 10% represent a pair of isolates of the same species and curves above 10% represent a pair of isolates belonging to different species. A switch between the two types of curves represents a putative recombination between the two viruses or their ancestors. The linearized genome organization of each component is depicted at the bottom. ::: ![](1743-422X-2-21-8) ::: ### ii) Comparisons of the B components of EACMCV The EACMCV West African isolates had only a stretch of 800 nts in the BC1 region in common with EACMV isolates from Uganda, the only B component available for EACMV: the rest of the sequence was completely different. The DNA-B of the East African EACMV isolates is ± 85% homologous to the West African isolates. The pairwise profile (Fig. [8B](#F8){ref-type="fig"}) showed the same recombinant fragment of about 800 nts with above 90% identity with West African isolates of EACMCV and other East African isolates such as EACMV-UG3, EACMZV and SACMV. The rest of the genome showed greater relatedness to the West African isolates of EACMCV, above the \"species threshold\" limit. Overall, the EACMCV-\[TZ1\] B component can be considered a non-closely related strain of the B component of EACMCV-\[CM\], but much closer than the B components of other East African cassava viruses. ### iii) Comparisons of the common regions (CRs) of EACMCVs from Cameroon and Tanzania The common region of A components (CRAs) were 82% to 89% identical to those of West African isolates, which is low but not abnormal as the West African isolates were 91% identical to one another (Table [4](#T4){ref-type="table"}). The differences are mostly in the variable region between the TATA box and the TAATATTAC stem-loop, but also in the rest of the sequence. The CR of B components (CRBs) of the EACMCV-\[TZ1\] isolate was more distantly related, at between 78% and 80% homology to the CRBs of the West African isolates, while they were 97% homologous to one another. The differences were mostly in the variable region. When both (CRAs and CRBs) were compared, it was apparent that CRs of the East African isolates were more similar to the CRAs of West Africa than the CRBs of West Africa. This arises mainly from a deletion of GAAAA, and from a more similar sequence in the region between the TATA box and the stem-loop. The putative replication protein binding sequences (iterons) were [GGTGG]{.underline}-AAT-[GGGGG]{.underline} for all the isolates except for the Bs of West Africa where it is [GGTGG]{.underline}-AAC-[GGGGG]{.underline}. There is a repeat of [GGGGG]{.underline} in the 5\' end of the CRs for all the isolates (Fig. [6B](#F6){ref-type="fig"}). Recombination analysis of cassava mosaic geminiviruses ------------------------------------------------------ The pairwise analysis performed on all African cassava viruses sequenced so far, with two Indian cassava viruses as out-groups, and including the viruses isolated in Tanzania (here described), showed a number of putative recombinant fragments for both components. Figure [9](#F9){ref-type="fig"} shows a genomic map for each component and summarizes the results obtained for the A and B components. ::: {#F9 .fig} Figure 9 ::: {.caption} ###### Recombination linearized map of putative recombinant fragments for the A (top) and B (bottom) components of cassava mosaic geminiviruses. Each horizontal line represents the genotype of one virus isolate and the color-coded boxes represent the tentative origins of the putative recombinant fragments. The length of the genomes is indicated on the top of each diagram and the genome organization is depicted at the bottom, while the abbreviated names of the viruses are listed on the left. The color code for the recombinant fragments is indicated in the boxes at the bottom of each diagram. The vertical arrows indicate the position of possible \"hot spots\" for recombination. On the right side are listed the percentages of EACMV-type and SACMV-type sequences for each virus. ::: ![](1743-422X-2-21-9) ::: ### i) Pairwise analysis of the A components #### African cassava mosaic virus None of the ACMV sequences obtained so far exhibited a possible recombinant fragment. An isolate of ACMV was involved in a recombination between EACMV and ACMV to produce the EACMV-UG2 isolate, which was associated with the epidemic in Uganda in the 90s \[[@B7],[@B8]\]. But it is worth noting that ACMV acted as a donor of DNA, not a receiver, in the recombination. The situation for the EACMV-like viruses is very different, as they exhibit multiple putative recombinations between themselves and also unknown viruses. The A components of all the viruses in East Africa share a common backbone from EACMV and have integrated other pieces of DNA that have been said to originate from the other viruses not identified so far. #### East African cassava mosaic Zanzibar virus Two isolates of EACMZV from Zanzibar and Kenya \[[@B12]\] have most of their genomes from EACMV; approximately 200 nts (2050 to 2250 nts) are similar to SACMV and the rest of the genome, covering AC1, AC4 and the CR, is unique and therefore attributed to EACMZV or an ancestor of EACMZV (Fig. [9A](#F9){ref-type="fig"}). #### East African cassava mosaic Cameroon virus Several EACMCV isolates from Cameroon, Ivory Coast and now Tanzania (this report) belong to the species East African cassava mosaic Cameroon virus(see paragraph 3.6; \[[@B9]\]); all share the same putative recombinant fragment, i.e.a fragment of 800 nts (AC3-AC2-CterAC1), that is unique and therefore attributed to EACMCV (Fig. [9A](#F9){ref-type="fig"}) or a common ancestor. However, the three isolates from West Africa do have a small recombinant fragment (100--250 nts) that is also unique to EACMCV, but this fragment is not present in the Tanzanian isolates. #### East African cassava mosaic Malawi virus Two virus isolates from the species East African cassava mosaic Malawi virusfrom Malawi (EACMMV-\[K\], -\[MH\]) \[[@B15]\] show a similar recombination pattern. The first 1000 nts have either a similar pattern as SACMV-\[M12\] and SACMV-\[ZW\] or share two fragments of 100 and 750 nts with the SACMV-\[ZA\] isolate from South Africa (Fig. [9A](#F9){ref-type="fig"}). The fragments 550--800 and 900--1050 nts are therefore attributed to EACMMV or an ancestor. The major difference with the SACMV isolates resides in the fact that the rest of the genome is purely EACMV-like, with the exception of 100 nts in the AC1 gene (1950--2050 nts). #### South African cassava mosaic virus One virus isolate of the species South African cassava mosaic virusfrom South Africa (SACMV-\[ZA\]) \[[@B16]\] exhibited a putative recombination, i.e.most of the first 1000 nts (CR, AV2 and most of AV1) and then the last 800 nts (NterAC1, AC4 and CR) are unique for this virus and consequently attributed to SACMV, or an ancestor of SACMV. The rest of the genome, covering AC3-AC2 and the C-terminus of AC1, is typical of EACMV (Fig. [9A](#F9){ref-type="fig"}). Another two isolates of SACMV, one from Madagascar (SACMV-\[M12\]) and one from Zimbabwe (SACMV-\[ZW\]), although belonging to the same species as the virus from South Africa, have a different recombination pattern, i.e.the first 1050 nts are similar to EACMMV with portions that are SACMV-type and portions that are EACMMV-type (Fig. [9](#F9){ref-type="fig"}). The SLCMV-\[Col\] and ICMV-\[Mah\] isolates, here used as out-groups \[[@B17]\], exhibited a large recombinant fragment of 1200 nts, possibly originating from ICMV \[[@B18]\] and encompassing NterAC1, AC4 and all the CR. Noticeably, several recombination sites are aligned among the different genomes, possibly indicating \"hot spots\" for recombination and possibly also delineating fragments in which variation led to selective evolutionary advantage. ### ii) Pairwise analysis of the B Components The B components of CMGs also showed the presence of putative recombinant fragments as determined by the pairwise analysis. Unfortunately, some B components, such as those of EACMV-\[TZ\], EACMMV-\[K\] and -\[MH\], have not been cloned yet and therefore we have only partial information. The ACMV B sequences available did not show any recombination. The EACMCV isolates from Cameroon, Ivory Coast and Tanzania all showed the same putative recombinant fragment, i.e.between 1700 and 2300 nts, corresponding to part of the BC1 gene. Interestingly, and a contrarioto the EACMCV A component, most of the B genome is unique and only the recombinant fragment originates from EACMV (Fig. [9B](#F9){ref-type="fig"}); the rest of the genome is therefore marked as the EACMCV-type (Fig. [9B](#F9){ref-type="fig"}). Furthermore, a comparison of the B components of the EACMCV isolates from Cameroon or Ivory Coast with the sequence from Tanzania shows between 250 and 1700 nts and between 2350 and 2800 nts, a different sequence, indicating either another two recombinations with another unknown virus or viruses, or, as supported by the number of point mutations, an extremely old sequence compared to the West African isolates of EACMCV (Fig. [9B](#F9){ref-type="fig"}); therefore it is marked EACMCV-\[TZ\]. On the contrary, the partial sequence of the B component of an isolate from Zanzibar (EACMZV-\[ZB\]) showed almost complete identity with a B component from EACMV-UG3, with a very short EACMZV-type fragment of 150 nts at the end of the genome. Similarly, the sole isolate of a B component of SACMV-\[ZA\] was almost entirely identical to EACMV-UG3, with a 500 nts fragment SACMV-type (1700 -- 2300 nts), mostly corresponding to a non-coding fragment of the virus. ICMV and SLCMV B components, here used as out-groups, were essentially identical with the exception of 200 nts covering the CR of SLCMV and justifying the claim that the SLCMV A component captured the B component of ICMV \[[@B17]\]. Quantification of the percentage of EACMV-type and SACMV-type sequences in each virus ------------------------------------------------------------------------------------- From the recombination analysis and phylogenetic results, it is clear that all EACMV-like viruses share a portion of the EACMV backbone sequence. The recombination map was used to calculate these percentages, indicated in Figure [9](#F9){ref-type="fig"} for each component. This percentage varies from 38 to 100% depending on each virus for the A components and from 22 to 100% for the B components. A similar calculation can be made for sequences that are SACMV-type and the results vary between 0 and 60% for the A components and from 0 to 16% for the B components (Fig. [9](#F9){ref-type="fig"}). Figure [11](#F11){ref-type="fig"} shows a repartition of these percentages according to the different viruses cloned and according to a transect between Uganda and South Africa, indicating that the EACMV backbone sequence decreases towards South Africa while the SACMV-type sequence increases. ::: {#F11 .fig} Figure 11 ::: {.caption} ###### Map of Africa depicting the putative inter-species recombinations of components A and B of cassava mosaic geminivirusess identified in different parts of Africa, either from this study or from GenBank accessions. The significance of the color codes is given in the figure. Where the component B of a particular virus has not been cloned, it is indicated in letters for a different species representative or as a faded drawing for a different isolate. For simplification of the drawing, not all the ACMV isolates have been shown as they are very similar. Similarly, the EACMV-UGs associated with the CMD pandemic now present in several central African countries have not been depicted as they are of very recent introduction (less than 10 years). The solid blue arrows represent the possible \"route\" of evolution of the EACMCV viruses, and the green arrows represent the possible \"route\" of evolution of the EACMV viruses. ::: ![](1743-422X-2-21-11) ::: Discussion ========== The present study confirmed the presence of representatives of 3 species of CMGs in Tanzania: one isolate of ACMV, four isolates of EACMV, and two additional isolates of EACMCV. The complete DNA-A nucleotide sequences of these isolates were determined. ACMV ---- It is apparent from the results of this study that several CMGs exist in Tanzania showing a high genetic diversity. The ACMV characterized from Tanzania was found to have very high overall DNA-A nt sequence identity to all the other isolates of ACMV sequenced so far. As there is no relation between the origin of ACMV isolates and their sequence relationship with other isolates, it is impossible to tell if the one found in Tanzania is more related to one ACMV isolate than another. As it is the first isolate to be sequenced from Tanzania, we named it ACMV-\[TZ\]. This virus, like all the other ACMVs, displayed no detectable recombination in its DNA-A genome. EACMCV ------ EACMCV-\[TZ1\] and EACMCV-\[TZ7\] had high overall DNA-A nt sequence identities, as well as high CP and CR sequence identity to members of the species EACMCV from West Africa, confirming their relatedness to that species. The two isolates from Tanzania are about 8% different, while each of them is more than 10% different to any of the West African isolates. The two Cameroonian isolates are very close to one another (\>99%) and about 3--4% different from the Ivorian isolate. In addition, the Tanzanian viruses showed the same recombination, relative to EACMV-type sequences, as the EACMCVs from Cameroon and Ivory Coast, covering the C-TerAC1-AC2-AC3 region. However, we noted that the Tanzanian isolates have lost or never acquired a small recombinant sequence at the beginning of the genome, as present in the West African isolates. The EACMCV-\[TZ1\] B component showed the same recombination as the EACMCV-\[CM\] and EACMCV-\[CI\] B components, covering part of the BC1 region. However, the EACMCV-\[TZ1\] B component had an additional two putative recombinant fragments (250--1700 nts and 2350--2800 nts) not present in the West African isolates. Considering the overall sequence identity of both components, the fact that sequence differences are scattered all along their genomes and the fact that there are differences in patterns of recombination, it is strongly suggested that the two sets of viruses from East and West Africa have been separated for a very long time and are not the result of a recent introduction in either direction. One recombination in DNA-A and one in DNA-B, as they are identical, pre-date their separation, though it is not possible at this stage to date the separation. EACMCV-\[TZ1\] occurred widely in southern Tanzania, being present in over 98% of CMD-diseased samples collected from the southwestern part of Tanzania in the Ruvuma Region close to Lake Malawi in the same area where EACMCV-\[TZ7\] was found. The fact that the two sequences in Tanzania show from two to three times more sequence variability and two extra recombinant fragments, together with the fact that the parent EACMV has not been found so far in West Africa, suggests an East African origin of this virus species, and therefore a possible spread from the East to the West as indicated in Figure [10](#F10){ref-type="fig"}. ::: {#F10 .fig} Figure 10 ::: {.caption} ###### Graph representing the proportion of EACMV-type and SACMV-type sequences in each virus isolated along a transect from Uganda (Left) to South Africa (Right) for their A components (A) and B components (B). The virus name abbreviations are given in the legend of Figures 3 and 4 and throughout the text. ::: ![](1743-422X-2-21-10) ::: EACMV-TZ, -KE, -UG ------------------ The rest of the CMGs cloned in this study were closely related to those reported in the neighboring countries of Uganda, Kenya or the previously characterized Tanzanian isolate of EACMV. These were EACMV-\[TZ/YV\], which resembled the EACMV-\[TZ\] characterized previously \[[@B19]\], and EACMV-\[KE/TZT\] that showed high sequence identity with EACMV-\[KE/K2B\] from Kenya, on the basis of their overall DNA-A nt sequences. While the CP of EACMV-\[TZ/YV\] showed high sequence identity with EACMV-\[TZ\] and EACMZV-\[ZB\] from the island of Zanzibar \[[@B12]\], EACMV-\[KE/TZT\] from Tanga region showed high nt sequence identity with its close relative EACMV-\[KE/K2B\]. Similarly, EACMV-\[KE/TZM\] also shared high CP nt sequence identity with EACMZV-\[ZB\]. It was found in only ten samples and very localized in spread within the region. Plants singly-infected with EACMV-\[KE/TZM\] expressed very severe symptoms both in the field and growth chamber. Whether this phenotype was a result of the nature of the EACMV-\[KE/TZM\] DNA-A genome remains to be established. The EACMV-UG2 \[TZ10\] shared very high DNA-A and CR sequence identity with EACMV-UG2Svr from Uganda. The CR also showed 100% nt sequence identity with EACMV-UG2Svr as well as high CP sequence identity to ACMV isolates because it has the same recombination as its closest relative, EACMV-UG2Svr, that was proven to involve two viruses (ACMV and EACMV) \[[@B7],[@B8]\]. This CMG was localized in the northwestern part of Tanzania in the post-epidemic area. It is noticeable that this virus, which has invaded a large portion of Central Africa in just a few years \[[@B4]\] (Fig. [8](#F8){ref-type="fig"}), has not yet reached the southern and eastern part of Tanzania. Recombination of A and B components ----------------------------------- Using all the CMG sequences available so far, we have shown that both A and B components of most of the CMGs exhibit putative recombinant fragments from various known or unknown origins. Despite the smaller number of sequences of DNA-B components and the smaller number of putative recombinant fragments, it is interesting to note that, as for the A components, it seems that there are \"hot spots\" for recombination. These apparent hotspots for recombination could result from physical constraints in the virus sequences or could simply result from the functional constraints of having recombinant proteins that keep structural and biological functions. These hotspots have already been mentioned in other general studies of geminivirus recombination \[[@B14]\] as well as in specific studies of particular groups of geminiviruses \[[@B20]\]. Two categories of CMGs in Africa -------------------------------- Based on recombination analyses, it is apparent that there are really two different categories of CMGs. The ACMV group does not have fragments of foreign geminivirus DNA in their genomes. By contrast, all other African CMG species groups show evidence of extensive recombination. It is also significant that EACMCV isolates obtained from each side of the African continent appear to share a similar genetic make-up and recombination pattern. This suggests that these viruses had a common origin, probably in East Africa, but diverged a long time ago. Recombination events have been shown to be key factors in the development of CMD epidemics \[[@B7],[@B8],[@B19]\] and it has been suggested that recombination is a significant contributor to geminivirus evolution \[[@B14]\]. Recombination involving the CP sequence has been reported for EACMV-UG2 from Uganda \[[@B7],[@B8],[@B10]\], a virus that has been associated with the current CMD pandemic that has devastated cassava in eastern and central African countries \[[@B4],[@B21]\], although there is currently no proof that this event has been the key factor driving the pandemic\'s spread. B components of CMGs in Tanzania -------------------------------- The diversity of DNA-B components of EACMV from Tanzania was investigated using partial DNA-B nt sequences (BC1-CR) of \~560 bp. Generally, there was little genetic divergence among the compared isolates with the exception of TZB6 that shared 97% sequence identity to EACMV-UG1 (AF230375) and 96% with EACMV-UG3 from Uganda. Isolate TZB1 and TZB7 clustered with EACMCV-TZ1 and are probably Bs of EACMCV A components. However, for the other isolates that grouped or formed their own group in the phylogenetic analysis, it was difficult to speculate as to what they represent partly because the DNA-Bs of EACMV-\[TZ\], EACMV-\[KE\] and EACMMV have yet to be sequenced. However, these short fragments indicated a clustering, apart from EACMV-UG and EACMCV, into 4 additional clusters that could reflect an even greater molecular diversity in the B components of CMGs in East Africa than we currently recognize. EACMV evolution --------------- The clustering of all the EACMV-like viruses into one species has been the topic of much scientific debate in recent years. ICTV (International Committee on Taxonomy of Viruses) finally decided to split them into 5 species (for now), mostly to comply with the ICTV guidelines for species demarcation, but clearly these viruses are closely related and had common ancestors. All EACMV-like viruses with the exception of EACMCV occur in East Africa, and mostly east of the Rift Valley. Evidence presented here and elsewhere now provides a strong case for an East African origin for the EACMVs. EACMCV is widely-distributed across West Africa, albeit at low incidence \[[@B4]\]. Whilst it seems likely that this is the result of an early introduction or introductions from East Africa, it is not currently clear when such an introduction(s) might have taken place. It is even possible that the spread of this virus occurred in another host, long before cassava was introduced into Africa. Finally, the rapidly expanding EACMV-UG2 associated pandemic of severe CMD in East and Central Africa represents a contrasting, and currently probably unique, scenario in which the combination of a virulent recombinant virus, superabundant vector populations and susceptible local cassava germplasm have led to a rapid expansion in the geographic range of EACMV-UG with a concomitant devastating impact on cassava cultivation. Furthermore, it is significant that when considering the proportion of pure EACMV backbone sequences in the A components of all the EACMV-like viruses, there is a clear gradient from East Africa to South Africa, going from 100 to 38%, suggesting firstly that these viruses are highly related and secondly that the origin of the EACMVs might have been East Africa, hence the green arrows in Figures [8](#F8){ref-type="fig"} and [11](#F11){ref-type="fig"}. Similarly, a reverse gradient for the SACMV-like sequence, going from 8 to 60% from Zanzibar to South Africa, suggests that the SACMV ancestor was located in South Africa. Because recombination can only occur when the two parent viruses are in the same plant, it is logical to expect a spatial relation between the different viruses and their genetic make up. It is, however, the first time that such gradients have been demonstrated for geminiviruses. The situation for the B components is completely different. There is no EACMV-like gradient from North to South, as most of the available sequences show a great proportion of EACMV-like sequences. However, it is evident that EACMCV has captured a B component completely different from EACMV, with only a small EACMV-like fragment. This result is concordant with the idea that B components can be recruited independently from the genetic nature of A components as already suggested for SLCMV and ICMV \[[@B17]\]. East Africa ----------- East Africa has been the cradle for many biological organisms beginning with humanity. From this work, it is also apparent that Tanzania may also be a potential source of origin of the family of EACMVs. The revealed strain diversity further exemplifies the wealth of this part of Africa with respect to cassava geminiviruses. Some of these viruses have been introduced very recently, such as EACMV-UG2 \[TZ10\], while others, such as ACMV and the EACMVs, have clearly been present much longer. The East-African Arc Mountain is known to be the main bio-diversity hot-spot in Africa, and an important refuge for plants and animals \[[@B22]\], therefore it is plausible that some of the geminiviruses that were invading local host plants were spread throughout Africa in their local hosts (for many millions of years), as it was suggested for Rice yellow mottle virus\[[@B23]\] colonizing the domesticated host in very recent history (a few hundred years). The same type of geminiviruses would have colonized cassava wherever they might be, beginning with that crop\'s introduction into the African continent in the XVI^th^century, as these viruses would have had the same potential for such colonization. This might have been the case for EACMCV for which our data presented here suggest an old East African origin for the now widely distributed EACMCV in West Africa. In addition to this scenario, it is certain that cassava geminiviruses have been exchanged throughout the movement of virus infected cassava cuttings via human intervention and by the natural vector Bemisia tabaci. The latter may account for the EACMV/SACMV gradient between East Africa and South Africa, favoured by a natural corridor along the eastern Rift Valley and created by the recombination capacity of CMGs present in the same region. However, more sequences are required in order to compare and contrast variability within and between the virus populations and to strengthen the understanding of their evolutionary interrelationships. The rapid spread of the EACMV-UG2 associated pandemic has been driven through superabundant whitefly populations \[[@B24]\], but other important forces in CMG movement and evolution include movement of cassava cuttings and transmission from and into alternative weed hosts. Although cassava was brought to Africa in the XVI^th^century, it attained its current Africa-wide distribution as recently as the XIX^th^century. Most current movement occurs informally as farmers move cuttings locally. Wider distribution is less frequent but may be more significant in enabling major displacements of CMGs, such as that hypothesized for the introduction of EACMCV from East to West Africa. Although rapid spread of up to 100 km per year has been reported for the EACMV-UG associated pandemic \[[@B4]\], elsewhere there appears to be much less local spread of CMGs by whitefly, and physical barriers including lakes, forests and regions where cassava is not grown, appear to be effective in curtailing local spread of CMG. This would seem to account for the apparent \'island\' of EACMCV in southern Tanzania as well as the absence of ACMV from coastal East Africa. Conclusion ========== In conclusion, we have established the existence of different CMG isolates, strains and species in Tanzania with some isolates resembling those reported previously in East African countries and two isolates very similar to the geographically distant EACMCV from West Africa. This study demonstrates that East Africa is rich in CMGs and could be the cradle for CMG diversification in Africa. It also highlights the urgent need for more information. Only through building a thorough understanding of these important plant pathogens and the evolutionary processes underpinning their emergence can we hope to develop effective and sustainable approaches to managing the disease they cause. Methods ======= Collection of plant samples --------------------------- A total of 510 samples were collected during September 2002 from the northeastern coast (60), east coast (74), southeastern coast (68), southern region (70) and the Lake Victoria Basin (238), representing the major cassava-growing areas in Tanzania. Cassava leaf samples and cuttings (25--30 cm in length) were collected from plants expressing CMD symptoms in fields located at a minimum of 5 km intervals. Leaf samples were kept in a cool box for DNA processing. Selected cassava cuttings were transported to the Donald Danforth Plant Science Center, St. Louis, MO for replanting in controlled growth chambers. Symptom reproduction in the growth chamber ------------------------------------------ Selected cassava cuttings collected from the fields were planted in a growth chamber at 25°C with a 16 hours day length and 50% relative humidity and watered twice weekly. CMD symptoms were recorded daily on the newly formed leaves for the first three months and every three days in the subsequent months for an eight month period. Symptom severity on the top five fully-expanded leaves was scored using a scale described by Fauquet et al\[[@B25]\]. DNA extraction -------------- Total DNA was extracted from the symptomatic cassava leaves collected in the field and growth chamber as described by \[[@B26]\]. Polymerase chain reaction, cloning, and sequencing -------------------------------------------------- Full-length copies of DNA-A were amplified from total cassava plant DNA extracts using sets of primers (Table [1](#T1){ref-type="table"}). UNI/F and UNI/R are degenerate primers with annealing positions in the AC1 gene designed to amplify near full-length DNA-A of CMGs (2.7--2.8 kbp) leaving an unamplified portion of \~17 nts. From the near full-length CMG sequences, primers were designed to amplify the remaining partial DNA-A sequences including the missing 17 nts from the original samples. Partial fragments consisting of a region between the BC1 gene and intergenic region (IR) of DNA-B components of EACMV isolates from different cassava-growing areas were amplified by universal primers EAB555-F and EAB555-R (Table [1](#T1){ref-type="table"}) designed to amplify PCR products of about 540--560 kbp depending on the virus isolate. In order to amplify the DNA-A and DNA-B full-length, PCR was performed with 94°C denaturation followed by 35 cycles of 1 min at 94°C, 59°C for 1 min and 2 min at 72°C. For amplification of the partial DNA-B fragment (BC1/IR), PCR conditions were 30 cycles of 94°C for 1 min, 55°C for 1 min, 72°C for 1 min and an extension cycle of 10 min at 72°C. PCR products of the expected sizes were electrophorezed in a 1% agarose gel in TAE buffer (10 mM tris-acetate, 1mM NaEDTA, pH 8.0), purified, and cloned into the pCR 2.1 vector using the TA cloning kit (Invitrogen, San Diego, CA). Clones containing putative viral sequences were identified by miniprep screening and confirmed positive for inserts by PCR amplification using their respective PCR primers, and inserts were subsequently sequenced in both directions. The complete and partial nucleotide sequences of CMGs were determined by the dideoxynucleotide chain termination method using an ABI automatic sequencer on both orientations at the Protein and Nucleic Acid Chemistry Laboratories (PNACL), Washington University School of Medicine, St. Louis, Missouri, USA (ABI377 DNA sequencer, Perkin Elmer, Foster City, CA). Sequence fragments of \< 600 kbp were generated using M13 universal primers. Moreover, to obtain overlapping data from opposite strands of large or full-length fragments, single primers were constructed for genome walking. Sequences were submitted to GenBank and the accession numbers are as follows: Complete nucleotide sequence of DNA-A named EACMCV-\[TZ1\] (AY795983); EACMCV-\[TZ7\], (AY795984); EACMV-UG2 \[TZ10\], (AY795988); EACMV-\[KE/TZM\] (AY795986); EACMV-\[KE/TZT\], (AY795985); EACMV-\[TZ/YV\], (AY795987); ACMV-\[TZ\] (AY795982); and DNA-B for EACMCV-\[TZ1\](AY795989). Partial DNA-B (BC1/ICR) sequences of EACMV isolates from Tanzania named TZB (AY800251), TZB1 (AY800252), TZB2 (AY800253), TZB3 (AY800254), TZB4 (AY800255), TZB5 (AY800256), TZB6 (AY800257), TZB7 (AY800258), TZB8 (AY800259), TZB9 (AY800260), TZB10 (AY800262), TZB11 (AY800261), and TZB12 (AY800263). Computer analysis of CMG sequences ---------------------------------- Virus sequences were edited using BioEdit Sequence Alignment Editor (Hall, 1999) and SeqEdit (DNAStar, Madison, WI) to obtain a consensus sequence for each. Reference geminiviruses for full length CP and CR sequence alignments were compiled by extracting the complete DNA-A and DNA-B sequences, the CP ORF (approximately 765--777 bp) and CR sequences (approximately 150--170 bp) from sequences available in GenBank (accession \# are provided in the figures). Multiple sequence alignments of the full-length DNA-A, DNA-B, capsid protein (CP) gene and common region (CR) were carried out using the Clustal Program (MegAlign, DNAStar). The phylogenetic trees were constructed from the multiple alignments by the neighbor-joining majority rule consensus. Multiple alignments were analyzed by maximum parsimony with full-length DNA-A, DNA-B and CP phylogenetic trees using Phylogenetic Analysis Using Parsimony (PAUP) \[[@B27]\] and a bootstrap analysis with 1000 replicates was performed. Only values above 50% were reported on the trees in the figures. Virus specific iterons in the CR of selected CMGs were identified and compared with the analogous iterons of the Tanzanian isolates of CMGs. Recombination analysis for cassava mosaic geminiviruses ------------------------------------------------------- The pairwise comparison sequence analysis (PCSA) method compares the profile of a pair of sequences to that of an average profile of sequences that are selected a priori, based on knowledge that the selected sequences are related to the species or the isolate levels \[[@B20]\]. According to the guidelines established by the ICTV GeminiviridaeStudy-Group, two geminivirus sequences sharing more than 89% identity of their A component sequences are considered strains or isolates of the same species. Where homology is less than this, they are considered to be members of different species \[[@B5]\]. However, viruses that share between 80 and 90% sequence identity are often found to be recombinants \[[@B2]\], therefore, in the PCSA, we consider viruses sharing less than 80% identity as different species. PCSA profiles were carried out between sequences of different species and of different isolates and an average profile for the considered cluster of viruses was calculated for these two categories with increments of 50 nts along the genome sequence. A standard deviation value for each segment was calculated and minimum and maximum values corresponding to two standard deviation values were also calculated (Fig. [1](#F1){ref-type="fig"}). Each chosen pairwise analysis for putative recombinant sequences was then compared to the species average profile and the pertaining of each 50 nts fragment to this category is examined. Segments different from more than 2 standard deviation values were considered to be putative recombined fragments. For each PCSA, a putative recombination percentage for the genome is calculated and a corresponding map can be drawn. It is verified (a posteriori) that the particular representatives of species and isolates selected for the \'Species and Isolate Average Curves\' are 100% non-recombinant at the time of the analysis \[[@B20]\]. No statistical test is applied to PCSA. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### \(A) Pairwise analysis of begomoviruses in the Old World that do not exhibit putative recombinant fragments at the species level (green curve) and at the strain level (blue curve). (B) Pairwise analysis of EACMCV-\[TZ1\] A component, paired with the sequence of the A component of other cassava mosaic geminiviruses like EACMCV-\[TZ7\] (blue line), ACMV-\[TZ\] (brown line), EACMV-\[KE\] (red line) and EACMZV-\[ZB\] (green line), showing the recombinant fragment of this virus (1200 -- 2000 nts) as well as the one from EACMZV-\[ZB\] (2000 -- 2900 nts). The linearized genome organization of these geminiviruses is depicted at the bottom of the graph. ::: ![](1743-422X-2-21-1) ::: Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= Design and conception of the study (JN, JPL, TASA, GT, CMF); execution of the experiments (JN); manuscript preparation (JN, JPL, CMF); sequence analysis, alignment and phylogeny (JN, CMF). All authors read and approved the final manuscript. Acknowledgements ================ This study was funded by the Crop Protection Programme of the UK\'s Department for International Development (DFID) through the Tropical Whitefly IPM Project of the System-wide Programme for Integrated Pest Management (SP-IPM). SP-IPM is an inter-centre programme established by the Consultative Group for International Agricultural Research (CGIAR). The senior author was also supported through a graduate fellowship from the International Institute of Tropical Agriculture (IITA) and by the Donald Danforth Plant Science Center, which supported some of the costs in St. Louis. The assistance of Mr. Cyprian Alloyce Rajabu of Plant Protection Division in Mwanza during sample collection is highly appreciated. Special thanks are due to other colleagues for helping in various ways, especially Dr. Justin Pita of the Noble Research Foundation, Ardmore, Oklahoma, USA, and Ismaël Ben F. Fofana of the International Laboratory of Tropical Agricultural Biotechnology (ILTAB), Donald Danforth Plant Science Center, St. Louis, MO, USA for his technical assistance. The views expressed do not necessarily represent those of DFID.	PubMed Central	2024-06-05T03:55:55.640422	2005-3-22	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1079959/", "journal": "Virol J. 2005 Mar 22; 2:21", "authors": [ { "first": "J", "last": "Ndunguru" }, { "first": "JP", "last": "Legg" }, { "first": "TAS", "last": "Aveling" }, { "first": "G", "last": "Thompson" }, { "first": "CM", "last": "Fauquet" } ] }
PMC1079960	Background ========== Sheeppox virus, an epitheliotropic DNA virus, is classified as a member of Capripox virus genus that represent one of eight genera within the chordopox virus subfamily of the Poxviridae. Genus Capripoxvirus is comprised of sheeppox virus, goatpox virus, and lumpy skin disease virus that cause disease in sheep, goats, or cattle, respectively. These viruses are responsible for some of the most economically significant diseases of domestic ruminants in Africa and Asia \[[@B9],[@B10]\]. Live attenuated SPPV and subunit formulations have been used experimentally and in enzootic as well as outbreak areas as vaccines against sheeppox, goatpox, and lumpy skin disease \[[@B8],[@B9]\]. The Poxviridae are the largest known viruses \[[@B10]\] that have strong immunogenic properties. Poxviruses modulate the immune response in infected hosts by inhibiting the synthesis and release of IL-1 from infected cells; encoding soluble cytokine receptors for tumor TNF-α, TNF-β, IL-1, and importantly, IFN-γ; synthesizing virus-encoded cytokines like epidermal growth factor and transforming growth factor, which antagonize the effects of host cytokines mediating the antiviral process \[[@B16],[@B26]\]. In addition, inducing apoptosis in a significant number of antigen-presenting cells \[[@B20]\] as well as inducing IL-10 release that has the capacity to impair the initiation of an acquired immune response \[[@B16],[@B21]\]. If the viruses fail to secrete such immunomodulating proteins, as when the respective genes are deleted or the viruses are inactivated, the strong immunogenicity of the viruses may induce host immune reactions which are no longer inhibited \[[@B19]\]. This is supported by earlier studies revealing enhanced phagocytosis, natural killer (NK) cell activity, and release of IFN-α by the use of inactivated poxviruses \[[@B7],[@B24]\]. Moreover, the secretion of TNF-α, IL-2, and granulocyte-macrophage colony-stimulating factor could also be enhanced \[[@B23],[@B30]\]. This assumption leads to the recommendation of use inactivated poxviruses as prophylactic or metaphylactic tool in reducing susceptibility to infectious diseases \[[@B31]\]. However, it has been reported recently that inactivated parapoxvirus ovis, was able to induce apoptosis of antigen-presenting cells (APC) \[[@B20]\]. In this study, sheeppox virus-induced immunomodulating effects were characterized to elucidate the basic mechanisms responsible for understanding the interaction of SPPV with host immune system. As markers for early immunological reactions, peritoneal cells were tested after in vivo treatment with SPPV for IL-10 release and SOD activities. Markers for late reactions were the proliferation response of splenocytes to PHA-P, IL-12 release, and SOD activity, of cultured splenic macrophages from treated mice. The antibody response to CRBC was also assessed in different treated groups. Results ======= Secretion of IL-10 by peritoneal macrophages -------------------------------------------- At 12 h post treatment, both vaccinated groups showed increased IL-10 (P \< 0.05) in comparison to placebo. Attenuated SPPV vaccinated group showed significant (P \< 0.01) increase in comparison to placebo. No significant variation was observed between the SPPV treated groups Fig. [1](#F1){ref-type="fig"}. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### IL-10 release from cultured peritoneal macrophages 12 h post SPPV immunization. Mice were injected intraperitoneally with PBS, inactivated SPPV, or attenuated SPPV. Peritoneal macrophages were harvested 12 h post inoculation (five/group). Macrophages were co-cultured with LPS 1 μg/ml for 48 h, IL-10 was measured in the culture supernatant. Bars represent mean ± S:E:M: of cytokine. SPPV vaccinated mice are significantly different from controls at \P \< 0.05 or \\P \< 0.01. ::: ![](1743-422X-2-22-1) ::: Secretion of SOD by peritoneal macrophages ------------------------------------------ At 12 h post treatment, both SPPV treated groups showed significant decreased SOD activity (P \< 0.05) in comparison to untreated group. No significant variation was observed between the SPPV treated groups Fig. [2](#F2){ref-type="fig"}. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### SOD activity of cultured peritoneal macrophages 12 h post SPPV immunization. Mice were injected intraperitoneally with PBS, inactivated SPPV, or attenuated SPPV. Peritoneal macrophages were harvested 12 h post inoculation (five/group). Macrophages were co-cultured with LPS1 μg/ml for 48 h, SOD was measured in the culture supernatant. Bars represent mean ± S:E:M: of SOD. SPPV vaccinated mice are significantly different from controls at \P \< 0.05. ::: ![](1743-422X-2-22-2) ::: Secretion of IL-12 by splenic macrophages ----------------------------------------- At 6 day post treatment, both SPPV treated groups showed significant increased IL-12 (P \< 0.05) in comparison to untreated group Fig. [3a](#F3){ref-type="fig"}. Attenuated SPPV treated group showed highly significant value (P \<0.01) than that recorded with placebo mice. At 9 day post inoculation attenuated SPPV treated group showed significant increased IL-12 secretion (P \< 0.01) in comparison to both inactivated SPPV treated group and untreated one Fig. [3b](#F3){ref-type="fig"}. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### IL-12 secretion from cultured splenocytes collected at 6 d (A) ; 9 d (B) post SPPV immunization. Splenic macrophages were harvested from mice (five/group), cultured with LPS 1 μg/ml for 48 h. IL-12 was measured in the culture supernatant. Bars represent mean ± S:E:M: of cytokine. SPPV vaccinated mice are significantly different from controls at \P \< 0.05 or \\P \< 0.01. ::: ![](1743-422X-2-22-3) ::: Splenocytes blastogenic response -------------------------------- No significant variations were detected among different groups at 12 h, 3 and 6 days post treatment. Significant increased splenocytes\' proliferation response to PHA-P (P \< 0.01) was observed at 9, and 12 days post inoculation in attenuated SPPV from placebo. Attenuated group showed significant increased from inactivated group at both 9 (P \> 0.05) and 12 days (P \> 0.01). No significant variation was observed between inactivated SPPV and control group at 12 day post treatment Table. [1](#T1){ref-type="table"}. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### T-cell proliferation based on the MTT dye uptake method of cultured splenocytes at different intervals post SPPV immunization ::: Time post treatment Treatment --------------------- ------------- --------------- ----------------- 12 h 1.22 ± 0.08 1.37 ± 0.14 1.53 ± 0.08 3 d 1.24 ± 0.14 1.39 ± 0.19 1.36 ± 0.18 6 d 1.36 ± 0.09 1.51 ± 0.17 1.45 ± 0.2 9 d 1.16 ± 0.05 1.53 ± 0.08\ 2.18 ± 0.12\\ 12 d 1.78 ± 0.18 1.93 ± 0.05 2.58 ± 0.2\\ Splenocytes were harvested from mice (five/group), cultured with PHA-P (10 μg/well) for 48 h, SI of SPPV treated mice are significantly different from controls at \P \< 0.05 or \\P \< 0.01. ::: Secretion of SOD by splenic macrophages --------------------------------------- No significant variations were detected among different groups at 12 h post treatment. At 3 day post inoculation significant increased SOD activity was observed (P \< 0.01) in attenuated SPPV in comparison to the other groups. Significant increased SOD activity (P \< 0.05) was observed at 6 (P \< 0.05), 9 (P \< 0.01), and 12 (P \< 0.01) days post inoculation in both SPPV treated groups in comparison to untreated group Table. [2](#T2){ref-type="table"}. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### SOD activity of cultured splenocytes\' macrophages at different intervals post SPPV immunization ::: Time post treatment Treatment --------------------- ------------- ------------------ ------------------- 12 h 13.3 ± 1.65 13.6 ± 1.22 12.8 ± 1.35 3 d 12.5 ± 2.8 10.82 ± 2.21 133 ± 23.1\\* 6 d 10.2 ± 1.79 107.7 ± 25 \* 143.7 ± 38\* 9 d 52.3 ± 19.1 253.3 ± 33.4\\ 240.7 ± 42.4 \\ 12 d 47.6 ± 8.5 266 ± 26.6\\ 293 ± 37.1\\ Splenic macrophages were harvested from mice (five/group). Macrophages were cultured with LPS (1 μg/ml) for 48 h and SOD was measured in the culture supernatant. SPPV treated mice are significantly different from controls at \P \< 0.05 or \\P \< 0.01. ::: Immune response to CRBC ----------------------- Both SPPV treated groups showed significant increase in haemagglutinating antibody titers to CRBC (P \< 0.05) in comparison to untreated group Fig. [4](#F4){ref-type="fig"}. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Humoral antibody response to CRBC. Mice were treated with live or inactivated SPPV or placebo. At 7 day, P.I., CRBC were administered by i.p route of inoculation. Seven days later, haemagglutinating abs were measured by HA test. SPPV treated mice are significantly different from controls at \P \< 0.05. ::: ![](1743-422X-2-22-4) ::: Discussion ========== Inactivated poxviruses showed immunostimulating capacity. Such capacity is common to poxviruses of different genera \[[@B11]\]. This fact renders poxviruses common vectors in vaccine development. On the other hand, poxviruses express a wide variety of proteins that are nonessential for virus replication in vitro but help the virus to evade the host response to infection that may in turn impair the immunological response against live viruses. Such nonstructural proteins includes soluble receptors for IFNα, β, TNFα, SOD-like protein,..etc. \[[@B16],[@B26],[@B35]\]. These facts together with the recurrence of SPPV infection among vaccinated flocks \[[@B33]\] let us to study first, the possible pivotal role of SPPV in inducing local immunosuppression as a crucial mechanism of immune escape, and second, to evaluate the potential beneficial systemic effect of SPPV on the host immune system. Early immune response to SPPV at the site of inoculation provides an explanation for the ability of SPPV to induce local suppression at site of inoculation as indicated by increased IL-10 secretion and decreased SOD enzyme activity 12 h after injection. IL-10, a prototypic anti-inflammatory cytokine, that inhibits APC function and ultimately the induction of anti-virus immunity \[[@B12]\], prevents the differentiation of DC from monocytes \[[@B5]\], and inhibits the down regulation of receptor-mediated endocytosis and macropinocytosis following exposure to a soluble immunogen \[[@B25]\]. In addition, IL-10 reduces the production of IL-2, IFN and TNF by murine Th1 cells \[[@B14]\] as well as the IL-12 production by APC \[[@B18]\]. IL-10 may also, intervene at the level of antigen processing within the cell so that antigen is not degraded effectively and MHC class II molecules fail to load with peptide \[[@B13]\]. Accordingly, enhanced secretion of IL-10 by SPPV has the capability to inhibit antigen presenting cell (APC) function as well as innate response and hence impairing the initiation of an acquired immune response. Such effect inhibits the generation of immunological memory necessary to immunity in subsequent exposure \[[@B2]\]. Interestingly, both inactivated and attenuated SPPV showed significant increase in the IL-10 production from peritoneal macrophages. On the other hand, decreased in vitro SOD activity of cultured peritoneal macrophages noticed in SPPV treated groups may also enhance in vivo virus survival in, and in the presence of phagocytes. Superoxide is generated deliberately by phagocytes during the respiratory burst to kill microorganisms \[[@B3]\]. The regulation of cellular SOD in poxvirus-infected cells might disrupt the balance of oxidants and antioxidants. Superoxide radicals arise during numerous oxidations in both living and nonliving systems and can act directly as oxidants or generate other reactive products that are toxic to cells, causing damage to lipid membranes, nucleic acid, carbohydrates, and proteins. SOD scavenges active oxygen species generated during aerobic metabolism. Consequently, aerobic existence is accompanied by a persistent state of oxidative siege, and the survival of a given cell is determined by its balance of reactive oxygen intermediates and antioxidants. Disturbance of this balance can lead to disease \[[@B17]\]. Further, since oxidative stress can induce apoptosis \[[@B6]\], this may aid virus dissemination, a fact recorded with other poxviruses \[[@B20]\]. Additionally, an increase in the cellular oxidant status results in activation of transcriptional factors, such as NF-κB \[[@B28]\], that may be necessary for replication of some viruses \[[@B27],[@B37]\]. Virulent viruses of SPPV may advocate similar immune evasion mechanisms that deflect down regulation and abortion of cell-mediated immunity. In order to gain more insight into the processes underlying the possible immune stimulating effect of vaccination with SPPV, the ex-vivo levels of IL-12, SOD in splenic macrophages, and the magnitude of splenocyte proliferative response to PHA-P as well as the in-vivo effect of SPPV on humoral immune response to TD Ag; CRBC were studied. The obtained data showed that mice successfully vaccinated with SPPV displayed significant increases in IL-12 levels at 6 and 9 days post vaccination as compared to unvaccinated mice. IL-12 was examined at that time as it is likely that a minimum of five days is required to create an effective barrier by poxvirus-mediated T-cell activation and cytokine secretion \[[@B11]\]. Interestingly, SOD, and lymphocyte blastogenesis as well as humoral immune response to CRBC were significantly enhanced in SPPV vaccinated mice especially in group treated with attenuated SPPV that may be related to virus replication. Enhanced splenic T-lymphocyte response to PHA-P in vaccinated mice indicates that SPPV may help in maintaining optimum T-cell responsiveness after vaccination. These effects might also rely on increased amounts of SPPV induced enhancement of IL-12 due to ability of IL-12 to induce early phase of NK and T cell activation \[[@B34]\]. Using CRBC as model of TD Ag, it has been shown that SPPV was capable of enhancing TD Ab responses. Interestingly, enhancement was observed in mice vaccinated with either inactivated or attenuated SPPV. Furthermore, enhancement of lymphocyte blastogenesis and SOD were also recorded. These results are the first to demonstrate an immunostimulant effect of SPPV. Enhanced responses to TD Ags in SPPV vaccinated mice may be due to the enhancement of IL-12 production, increased responsiveness of T-lymphocyte and the SOD activity that were recorded in this study. IL-12 enhancement may initiate such cascade as it has been found to be the dominant factor in development of the Th1 phenotype and also directly, or from its associated release of type-1 cytokines, enhances the activation and production of Th1-associated immunoglobulins \[[@B1]\]. In addition, IL-12 is not only a connective element between accessory cells and lymphocytes, but it is also a key molecule for programming the macrophage and dendritic cell functions \[[@B4]\]. One of the major effects of IL-12 on macrophages and dendritic cells is the induction of IFN-γ, resulting in a positive feedback capable of activating them in different situations \[[@B15]\]. Our observations indicate that the SPPV-induced reduced macrophages\' functions in a local event that occurs at the site of application 12 h after administration. In contrast, 3 till 12 days after injection of either inactivated or attenuated SPPV, enhanced functions of splenic macrophages and increased responsiveness of lymphocytes were found to be significantly increased that was more pronounced in attenuated vaccine rather than inactivated one. The combination of suppressive and stimulatory mechanisms is a complicated blend of viral survival strategy. Conclusion ========== Locally, SPPV shows evidence for an immune escape mechanism that alleviates the host\'s immune response to viral proteins and therefore generates the possibility of replicating in the host in spite of vaccination. Such suggested enhanced replication strategy appears to be essential for the continued existence of SPPV. Surprisingly, injection of inactivated SPPV produced similar effect but to a lower extent that denotes: evading host immune response by SPPV is not dependent on virus infectivity. Later and systemically, the virus protects the host from any fatal consequences of the suppression of the immune system by compensatory enhanced activities of splenic macrophages and lymphocytes. This cascade of immunological events would be an excellent strategy for the virus to survive. Methods ======= Animals ------- Eight-week-old Swiss male mice (Biological Supply Center, Theodar Bilharz Research Institute (TBRI), Cairo, Egypt) were used within this study. Mice were bred conventionally, and received standard laboratory diet as well as water ad libitum. Virus ----- SPPV attenuated vaccine (Vaccine and Sera Production and Research Institute, Abbasia, Cairo, Egypt) was used. The Vaccine was reconstituted in 2 ml sterile PBS (pH 7.4), titrated on the chorioallantoic membrane of 10-day-old specific pathogen free embryonated chicken eggs (Nile SPF, Koom Oshiem, Fayoum, Egypt). Half of the stock virus preparation was inactivated using β propiolactone as previously described \[[@B23]\]. Animal inoculation ------------------ Ninety mice were divided into three groups (30 mice per group). Mice in group 1 and 2 were inoculated with 10^7^EID~50~/0.2 ml of inactivated and attenuated SPPV vaccine, respectively by i.p. route of inoculation. Mice in group 3 were kept as a placebo and inoculated with sterile PBS by the same route. Analyses of IL-10 and SOD from cultured peritoneal macrophages -------------------------------------------------------------- The peritoneal macrophages were collected 12 h post i.p. inoculation of SPPV by peritoneal lavage. Cells were washed twice with sterile PBS, incubated in 24-well plate (Costar, Cramlington, U.K.) at a concentration of 1 × 10^6^cells/ml in RPMI 1640 (Gibco Laboratories, Grand Islands, NY) for 4 h at 37°C in a 5% CO~2~tension. Non adherent cells were removed by three repeated washings and peritoneal macrophages incubated with lipopolysaccharide 1 μg/ml (Sigma Chemical Co., USA) for 48 h at 37°C. At the end of incubation time, culture supernatants were collected, clarified by low speed centrifugation at 250 g for 10 min, and kept at -20°C until processing for IL-10, using mouse IL-10 immunoassay kit (BioSource International Inc. USA) according to manufacture instructions. SOD activity was also assessed in peritoneal macrophages\' culture supernatants by means of the inhibition of pyrogallol autoxidation as described \[[@B22]\]. One unit of SOD activity is defined as the amount of enzyme required to inhibit autoxidation by 50% at 25 °C. Splenocytes preparation ----------------------- Proliferation assay was conducted at 12 h, 3, 6, 9, 12 days post SPPV inoculation. Spleens were aseptically removed and placed in ice cold sterile PBS, (pH 7.4). Each spleen was squeezed with a 5 ml syringe plunger to extrude cells. Cell suspensions were centrifuged at 250 g for 10 min. Pelleted cells were resuspended in 5 ml of lysing buffer (Tris 0.17 M and ammonium chloride 0.16 M, pH 7.2) and incubated at room temperature for 5 min to lyse the red blood cells. Cell suspensions were washed twice with PBS and cell viability determined using trypan blue dye exclusion method. Splenocyte Proliferation Assay ------------------------------ Splenocytes were suspended in RPMI-1640 containing 10% FCS. One hundred micro-liters of suspended cells (1 × 10^5^cells per 100 ul) were added to each well of 96-well microtiter plate (Costar, Cramlington, U.K.). Splenocytes were stimulated with PHA-P (Sigma, Chemical Co.USA) at a final concentration of 10 ug/well. Cell cultures were incubated for 48 h at 37°C with 5% CO~2~tension. Splenocyte proliferation was measured using 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) dye uptake method as described \[[@B32]\]. The lymphocyte blastogenesis was expressed as stimulation index (SI): SI (%) = A2-A0/A1-A0. A2 is the absorption of cultures with PHA-P; A1 is the absorption of cultures without mitogen; A0 is absorption of the blank (culture medium only). Analyses of IL-12 and SOD from cultured splenocytes --------------------------------------------------- The splenic macrophages were incubated in 24-well plate at a concentration of 1 × 10^6^cell/ml in RPMI 1640 for 4 h at 37°C in a 5% CO~2~tension. Non adherent cells were removed by three repeated washings and splenic macrophages were incubated with lipopolysaccharide 1 μg/ml for 48 h at 37°C. At the end of incubation time, culture supernatants were collected, and kept at -20°C until processing for IL-12 using mouse IL-12 immunoassay kit (BioSource International Inc. USA) according to manufacture instructions. The assay recognizes both natural and free p40 subunit. SOD was also assessed as previously described. Effect of SPPV on the immune response to CRBC --------------------------------------------- Mice in all groups were inoculated i.p. with 1 × 10^7^CRBC at day 7. Serum was obtained seven days after CRBC immunization, tested for agglutinins against CRBC by the microhaemagglutination test according to \[[@B36]\]. Statistical analysis -------------------- Analysis of variance (ANOVA) test was done for differences between treated groups according to \[[@B29]\]. List of Abbreviations ===================== Ab, antibody; Ag, antigen; Ags, antigens; attenu., attenuated; APC, antigen presenting cell; CRBC, chicken red blood cells; DC, dendritic cell; Th; T-helper; IL, interleukin; inac., inactivated; INF, interferon; NK, natural killer cell; PHA-P, phytohaemagglutinin-P; SOD, superoxide dismutase; SPPV, sheeppox virus;SI, stimulation index; TD, T-dependent; TNF, tumour necrosis factor. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= Abu-El-Saad participated in the design of the study, carried out the work with the mice, assisted in experimental work and drafting of the manuscript. Abdel-Moneim conceived the study, designed and carried out the experimental work and drafted the manuscript. All authors read and approved the final manuscript.	PubMed Central	2024-06-05T03:55:55.646972	2005-3-22	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1079960/", "journal": "Virol J. 2005 Mar 22; 2:22", "authors": [ { "first": "Abdel-Aziz S", "last": "Abu-EL-Saad" }, { "first": "Ahmed S", "last": "Abdel-Moneim" } ] }
PMC1079961	Background ========== Vaccinia virus (VAC), the prototype for the family Poxviridae, is a large double-stranded DNA virus that encodes numerous enzymes and factors needed for RNA and DNA synthesis, enabling it to replicate in the cytoplasm of infected cells \[[@B1]\]. More than 20 viral proteins including a multi-subunit RNA polymerase and stage specific transcription factors are involved in viral RNA synthesis \[[@B2]\]. Genetic and biochemical studies identified five viral proteins essential for viral DNA replication, namely the viral DNA polymerase \[[@B3]-[@B8]\], polymerase processivity factor \[[@B9],[@B10]\], DNA-independent nucleoside triphosphatase \[[@B11]-[@B13]\], serine/threonine protein kinase \[[@B14]-[@B17]\], and uracil DNA glycosylase \[[@B18]-[@B21]\]. In addition, the virus encoded Holliday junction endonuclease is required for the resolution of DNA concatemers into unit-length genomes \[[@B22]\]. Other proteins that may contribute to viral DNA replication, include DNA type I topoisomerase, single stranded DNA binding protein, DNA ligase, thymidine kinase, thymidylate kinase, ribonucleotide reductase and dUTPase (reviewed in reference \[[@B1]\]). The VAC genome consists of a 192 kbp linear duplex DNA with covalently closed hairpin termini \[[@B23],[@B24]\]. A model for poxvirus DNA replication begins with the introduction of a nick near one or both ends of the hairpin termini, followed by polymerization of nucleotides at the free 3\'-OH end, strand displacement and concatemer resolution \[[@B25],[@B26]\]. Nicking is supported by changes in the sedimentation of the parental DNA following infection, and labeling studies suggested that replication begins near the ends of the genome \[[@B27],[@B28]\]. Efforts to locate a specific origin of replication in the VAC genome led to the surprising conclusion that any circular DNA replicated as head-to-tail tandem arrays in cells infected with VAC \[[@B29],[@B30]\]. Origin-independent plasmid replication was also shown to occur in the cytoplasm of cells infected with other poxviruses including Shope fibroma virus and myxoma virus as well as with African swine fever virus \[[@B30],[@B31]\]. In contrast, studies with linear minichromosomes containing hairpin termini provided evidence for cis-acting elements in VAC DNA replication \[[@B32]\]. It was considered that plasmid replication might be initiating non-specifically, perhaps at random nicks in DNA. Although transfected plasmids were used to study the resolution of poxvirus concatemer junctions \[[@B33]-[@B37]\], the system has not been exploited for studies of viral DNA synthesis. The goal of the present study was to determine how closely plasmid replication mimics viral genome replication. For example, if some viral proteins are needed for initiating DNA synthesis at specific origins near the ends of the viral genome, they might not be required for plasmid replication. In addition, we were curious as to whether synthesis of plasmid DNA occurs diffusely in the cytoplasm, since the transfected DNA enters cells independently of virus and contains no viral targeting sequences. Contrary to these speculations, we found that each of the five viral proteins known to be required for viral genome replication was needed for origin-independent replication of plasmids. Moreover, both plasmid and genome replication occurred in discrete viral cytoplasmic factory areas. Thus, small circular plasmids are useful surrogates for the large viral genome in studying the mechanism of poxvirus DNA replication and the trans-acting factors required. Results ======= Determination of plasmid replication by real-time PCR ----------------------------------------------------- The replication of plasmids and linear minichromosomes, which were transfected into cells infected with VAC, was previously demonstrated by autoradiography following hybridization of ^32^P-labeled probes to Southern blots \[[@B29],[@B30],[@B32]\]. Methylated input DNA prepared in E. coliwas distinguished from unmethylated DNA replicated in infected mammalian cells by digestion with DpnI and MboI, which cleave G^m^ATC and GATC sequences, respectively. DeLange and McFadden \[[@B30]\] had reported an 8-fold net increase of a circular plasmid lacking viral sequences in rabbit cells infected with myxoma virus, whereas Du and Traktman \[[@B32]\] had seen a 2-fold net increase of a linear minichromosome containing VAC genome termini in mouse L cells infected with VAC, but a much lower increase of a circular plasmid lacking viral sequences. We compared the replication of three types of DNA (super coiled circular, linear, and linear minichromosome) in African green monkey BS-C-1 cells, which has become a standard cell line for VAC research. Southern blot analysis of the DpnI-digestion products of DNA isolated from cells infected with VAC and transfected with super coiled pUC13 revealed a prominent high molecular weight band migrating above the 23.1-kbp marker, presumably representing head-to-tail concatemers (Fig. [1A](#F1){ref-type="fig"}). A prominent DpnI-resistant band, migrating between the 4.4 and 6.6 kbp markers, was obtained by digestion of DNA from infected BS-C-1 cells transfected with the covalently closed minichromosome. However, only small digestion products were obtained upon DpnI-treatment of DNA from cells transfected with linear pUC13. In addition, DpnI-resistant bands were not detected by digestion of DNA from mock-infected cells transfected with a linear minichromosome or 10 times more super coiled plasmid (Fig. [1A](#F1){ref-type="fig"}). This experiment confirmed the need for VAC infection and either a circular plasmid or a linear minichromosome template for DNA replication. Moreover, we did not see greater replication of the linear minichromosome than the circular plasmid as had been reported (32). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Replication of transfected DNA in VAC-infected cells. (A)Southern blot of replicated circular plasmid and linear minichromosome. B-SC-1 cells were infected with VAC and 1 h later transfected with equal molar amounts (20 fmol) of super coiled pUC13 (sc pUC), pUC13 linearized by digestion with EcoRI (lin pUC), linear minichromosome containing pUC13 and 1.3 kbp viral telomeric sequences (lin mc). As a control, cells were mock infected and transfected with 20 fmol of linear minichromosome or 10 times that amount (200 fmol) of super coiled pUC13. At 24 h after infection or mock infection, cells were collected and total DNA extracted. Total DNA (2 μg) was digested with DpnI subjected to agarose gel electrophoresis and analyzed by Southern blot hybridization using a ^32^P-labeled pUC13 probe. Samples (0.5 fmol of lin pUC, 0.5 fmol of lin mc, 1 fmol sc pUC) of the DNA used for transfections (input DNA) were also analyzed. The positions of marker DNA (kbp) are shown on the left. (B)Real-time PCR of replicated plasmid. BS-C-1 cells were transfected with the plasmid p716 at 24 h prior to infection with VAC. At indicated hours post infection (hpi), cells were harvested and total DNA extracted. DNA was untreated or treated with DpnI or MboI and analyzed by real-time PCR using primers specific to plasmid DNA. (C)Quantification of Southern blot. DNA described in panel (B) was digested with EcoRI prior to MboI or DpnI treatment. The digested DNA samples were subjected to gel electrophoresis, transferred to a Nylon membrane, hybridized to a ^32^P-labeled p716 probe, and the radioactivity quantified with a phosphoImager. ::: ![](1743-422X-2-23-1) ::: To improve quantification of plasmid replication and to establish a non-radioactive method for rapid analysis of multiple samples, we devised a real-time PCR assay using primers 152 bp apart that flanked two DpnI/MboI sites in a circular plasmid lacking VAC DNA sequences. In initial experiments, we followed the protocol of previous studies by transfecting the plasmid after infection \[[@B29],[@B30]\]. However, MboI-resistant input DNA as well as DpnI-resistant replicated DNA increased with time, suggesting that entry of DNA into the cell occurred continuously even though the medium was changed at 4 h (data not shown). To avoid this problem in subsequent experiments, DNA was transfected 24 h prior to infection. Total DNA was isolated at various times, digested with DpnI, MboI, or left uncut and subjected to real-time PCR. Under these conditions, MboI-resistant DNA did not increase, whereas DpnI-resistant DNA increased \~18 fold between 3 and 6 h and \~400 fold by 24 h (Fig. [1B](#F1){ref-type="fig"}). Moreover, total DNA increased \~10 fold. Increased DpnI-resistant DNA was not detected in mock-infected cells (data not shown). Previous Southern blotting studies had indicated that plasmid replication paralleled genome replication \[[@B30]\]. We compared the kinetics of plasmid replication obtained by real-time PCR with Southern blotting. For the latter analysis, total DNA was first digested with EcoRI to resolve head-to-tail concatemers into linear units followed by digestion with MboI or DpnI. After electrophoresis, the DNA was transferred to a nylon membrane, hybridized to a ^32^P-labeled plasmid probe, and the amount of DNA quantified using a PhosphorImager. The DpnI-resistant and total DNA increased with time, whereas the MboI-resistant DNA did not (Fig. [1C](#F1){ref-type="fig"}). The Southern blot analysis suggested that the amount of replicated plasmid plateaued after 12 h, whereas it continued to increase slightly as determined by PCR (Fig. [1B](#F1){ref-type="fig"}), suggesting that the latter method has the greater dynamic range as well as being more convenient. Determination of the trans-acting factors required for plasmid replication -------------------------------------------------------------------------- The dependence of VAC genome replication on expression of five viral early proteins was previously determined by analysis of conditional lethal mutants. Because of the absence of cis-acting VAC DNA sequences, we considered that plasmid replication might only mimic DNA elongation steps and therefore require only a subset of viral proteins. To test this hypothesis, the plasmid was transfected into BS-C-40 cells (a derivative of BS-C-1 cells that have been passaged at 40°C), which were subsequently infected with a VAC tsmutant under permissive and non-permissive conditions. Plasmid replication was quantified by real-time PCR. Wild type VAC strain WR and Cts16, which has a mutation in the I7 gene encoding a protease required for VAC morphogenesis but not DNA synthesis \[[@B38]\], served as positive controls. Plasmid DNA synthesis was higher at 39.5°C than 31°C for both WR and Cts16 (Fig. [2A](#F2){ref-type="fig"}). In contrast, the reverse was true for each mutation known to impair DNA replication at the non-permissive temperature. Indeed, plasmid replication was barely detected at 39.5°C in cells infected with Cts24, Cts42, and ts185, which have defects in the D5 nucleoside triphosphatase, the E9 DNA polymerase, and the A20 processivity factor (Fig. [2A](#F2){ref-type="fig"}). The reduction in plasmid replication was less complete at 39.5°C in cells infected with Cts25, which has a defect in the B1 serine/threonine protein kinase, which is consistent with previous observations that showed viral genome accumulation was only moderately reduced in BS-C-40 cells at non-permissive temperatures \[[@B15]\]. The relatively low replication of plasmid at 31°C in cells infected with Cts42 and ts185 (Fig. [2A](#F2){ref-type="fig"}) suggested that the mutated DNA polymerase and processivity factor were still somewhat defective even under \"permissive\" conditions. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Viral protein requirements for plasmid replication. (A)Conditional lethal tsmutants. BS-C-40 cells were transfected with p716 and 24 h later were mock infected or infected with 3 PFU per cell of wild type VAC (WR) or tsmutant Cts24, Cts25, ts185, or Cts16 at permissive (31°C) and non-permissive (39.5°C) temperatures for 24 h. DNA was then isolated, digested with DpnI and quantified by real-time PCR. (B)D4R deletion mutant. RKD4R and RK-13 cells were transfected with p716 and 24 h later were infected with 3 PFU of vD4-ZG. At 24 after infection, DNA was isolated, digested with DpnI, and the amount of DNA quantified by real-time PCR. ::: ![](1743-422X-2-23-2) ::: Previous studies had shown that expression of VAC uracil DNA glycosylase was required for genome replication \[[@B39]\]. To determine whether this protein is required for plasmid replication, we used mutant virus vD4-ZG, in which the uracil DNA glycosylase gene was deleted \[[@B39]\], and rabbit cell lines lacking (RK-13) or stably expressing (RKD4R) VAC uracil DNA glycosylase \[[@B39]\]. We found that plasmid DNA replication was only detected in the cell line stably expressing the viral uracil DNA glycosylase (Fig. [2B](#F2){ref-type="fig"}), indicating a requirement for this protein as well as each of the other four factors. Transfected plasmid DNA accumulates in viral factories ------------------------------------------------------ VAC genomic DNA accumulates in specialized cytoplasmic factory areas near the nucleus. However, the intra-cytoplasmic location of plasmid replication had not been determined. We needed a specific tag to distinguish viral and plasmid DNA in order to locate the latter in infected cells. Several studies have used multimerized E. coli lacoperator (lacO) binding sites and lacrepressor (lacI) fusion protein interactions to examine chromatin organization and chromosome dynamics in living cells \[[@B40]-[@B43]\]. To apply this strategy, we transfected cells with a 10.5 kbp plasmid pSV2-dhfr-8.32 \[[@B44]\] containing 256 lacO repeats and infected the cells 24 h later. Initial experiments confirmed that plasmid replication occurred following VAC infection as described above for smaller plasmids (data not shown). Next we transfected cells with pSV2-dhfr-8.32 and then infected them with vV5D4, a recombinant VAC that expresses V5 epitope-tagged uracil DNA glycosylase. At 12 h after infection, DNA in the nucleus and cytoplasmic factories was visualized by Hoechst staining (Fig. [3](#F3){ref-type="fig"}). The plasmid appeared to be excluded from the nucleus and present exclusively in cytoplasmic viral factories as determined by staining the cells with a maltose binding protein (MBP)-lacI fusion protein and an antibody to MBP (Fig. [3](#F3){ref-type="fig"}). In addition, the plasmid sites contained the VAC DNA glycosylase, as shown by staining with antibody to the V5 tag of the latter protein (Fig. [3](#F3){ref-type="fig"}). No MBP staining was detected when a control plasmid lacking lacO sequences was transfected (data not shown). ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Intracellular localization of replicated plasmid containing tandem lacO repeats by staining with an MPB-lacI fusion protein. HeLa cells were transfected with pSV2-dhfr-8.32 containing lacO tandem repeats and 24 h later were infected with 3 PFU per cell of vV5D4 expressing V5-tagged uracil DNA glycosylase. At 12 h after infection, cells were fixed, permeabilized, incubated with MBP-lacI and rabbit antibody to MBP (anti-MBP) and mouse monoclonal antibody to the V5 epitope (anti-V5) followed by Cy5-conjugated donkey anti-mouse IgG and Texas red dye conjugated donkey anti-rabbit IgG. Cells were counterstained with Hoechst dye and analyzed by confocal microscopy. Colors: deep blue, Hoechst dye; red, Texas red; white, Cy5; light blue, overlap of Texas red and Cy5; yellow, overlap of Hoechst and Cy5. ::: ![](1743-422X-2-23-3) ::: Visualization of replicating plasmid DNA in live cells ------------------------------------------------------ In the above experiment, the cells were fixed and stained in order to visualize the plasmid DNA. We considered that these steps might be avoided by expressing a GFP-lacI fusion protein with a nuclear localization signal (NLS). The GFP tag allowed visualization of lacI by fluorescence microscopy while the NLS served to translocate GFP-lacI, which was not specifically bound to lacO sequences in DNA, from the cytoplasm to the nucleus. In order to express the fusion protein prior to and during DNA replication, we constructed the recombinant vGFP-lacIwith the open reading frame encoding the GFP-lacI-NLS fusion protein regulated by a viral early/late promoter. HeLa cells were transfected with pSV2-dhfr-8.32 and infected 24 h later with vGFP-lacI. Bright green fluorescence was detected over the viral factory areas and nuclei, which correlated with Hoechst staining (Fig. [4](#F4){ref-type="fig"}). In cells with multiple viral factories, however, not every one exhibited green fluorescence. The viral factory regions were also visualized by staining with an antibody to viral RNA polymerase, which surrounded and included the DNA sites at 12 h after infection (Fig. [4](#F4){ref-type="fig"}). When a control plasmid (p716) lacking lacO sites was transfected, the green fluorescence was strictly localized to the nucleus (Fig. [4](#F4){ref-type="fig"}). ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Intracellular localization of replicated plasmid containing tandem lacO repeats using a lacI-GFP fusion protein. HeLa cells were transfected with pSV2-dhfr-8.32 containing tandem lacO repeats (top 2 panels) or p716 control plasmid (bottom 2 panels) and infected with vGFP-lacI. At 12 h after infection, cells were fixed, permeabilized, and stained with antibody to VAC RNA polymerase (anti-RNAP), followed by Alexa 594-conjugated goat anti-rabbit IgG. Cells were then stained with Hoechst dye and analyzed by confocal microscopy. Blue, Hoechst; red, Alexa 594; and green, GFP fluorescence. ::: ![](1743-422X-2-23-4) ::: Having established the specificity of the GFP-lacI binding by co-localization, we examined fluorescence of live cells by time-lapse microscopy following transfection with pSV2-dhfr-8.32 and infection with vGFP-lacI. Weak GFP fluorescence was detected at about 5.5 h after infection (not shown) and was largely over the nucleus, reflecting the targeting due to the NLS. A region of juxtanuclear fluorescence corresponding to a viral factory was seen clearly at 7.5 h and over the next several hours increased in intensity (Fig. [5](#F5){ref-type="fig"}). The time course suggested that the factory region was the site of replication as well as accumulation of the plasmid DNA. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Visualization of plasmid replication in live HeLa cells. HeLa cells were transfected with pSV2-dhfr-8.32 containing tandem lacO repeats and infected with vGFP-lacI. Images were made at 5 min intervals starting at 5.5 h and ending at 10 h post infection. Selected images at the indicated time points are shown starting at 6 h time point. Arrow at 7.5 h time point indicates cytoplasmic site of replicated plasmid. ::: ![](1743-422X-2-23-5) ::: Discussion ========== The replication of circular DNA lacking viral sequences as head-to-tail concatemers in the cytoplasm of cells infected with a poxvirus was reported nearly 20 years ago \[[@B29],[@B30]\]. Fortuitous poxviral origins were ruled out by the replication of 5 different circular DNAs and no evidence was obtained for integration into the viral genome by non-homologous recombination. These data strongly suggested autonomous plasmid replication by a rolling circle or theta mechanism. The significance of sequence non-specific DNA replication was called into question by Du and Traktman \[[@B32]\], who reported only low-level replication of a super coiled plasmid compared to a linear minichromosome containing specific telomere sequences \[[@B32]\]. However, our determination of a 10-fold increase in net plasmid DNA compares favorably to the 2-fold increase achieved with the most efficient minichromosome construct \[[@B32]\]. Moreover, our finding was similar to the 8-fold increase in net plasmid DNA reported by DeLange and McFadden \[[@B30]\]. There are several procedural differences that might account for the disparate results. One difference was the type of virus and cell used: Du and Traktman used mouse L cells infected with VAC, DeLange and McFadden principally used rabbit cells infected with myxoma virus or Shope fibroma virus and we used monkey or HeLa cells infected with VAC. A second difference was the method of DNA isolation. Whereas we and DeLange and McFadden proteinase digested whole cell lysates, Du and Traktman lysed cells with cold hypotonic buffer containing a non-ionic detergent and removed nuclei by sedimentation prior to DNA extraction. VAC DNA replication occurs in juxtanuclear factories and loss of high molecular weight protein-DNA complexes, especially those containing long head-to-tail plasmid DNA concatemers upon centrifugation is a concern. Indeed, Du and Traktman \[[@B32]\] reported that the presence of the telomere resolution sequence was required for high efficiency replication of linear minichromosomes and that only monomeric products were recovered. Further studies are needed to determine whether the cis-acting sequences in the linear minichromosomes are serving as origins of replication or as concatemer resolution sites or both. The temporal coincidence of plasmid and viral DNA replication suggested that viral proteins were needed for each. Indeed, we found that each of the five trans-acting viral proteins known to be important for viral genome replication was similarly required for plasmid replication. Either none of these proteins have a sequence-specific role or some have dual roles and are also required for origin-independent replication. The proteins also may have structural roles in assembling the replication complex, the existence of which is suggested by the interaction of A20 with the D4 and D5 proteins \[[@B45]\] and the co-purification of the A20, D4 and E9 proteins with a processive form of DNA polymerase \[[@B46],[@B47]\]. VAC cores containing genomic DNA and an early transcription system travel from the cell entry site along microtubules to the juxtanuclear area where synthesis of early viral proteins and DNA replication result in the formation of discrete factories \[[@B48]\]. It is believed that each factory arises from a single infectious particle \[[@B49]\]. It was interesting to determine whether plasmid replication occurred in factories or dispersed throughout the cell. To investigate this, we transfected cells with a plasmid containing multiple repeats of the E. coli lacO, which tightly binds lacI. In one approach, the lacO DNA was located in discrete juxtanuclear regions by staining fixed and permeabilized cells with an MBP-lacI fusion protein followed by an antibody to MBP. The regions were identified as viral factories by Hoechst DNA staining and localization of the viral uracil DNA glycosylase, a protein required for replication of both plasmid and viral DNA. LacO DNA was not detected in the nucleus or in diffuse areas of the cytoplasm. A second approach involved the construction of a recombinant VAC that expresses a GFP-lacI fusion protein with a NLS to remove unbound protein from the cytoplasm. Again, the lacO DNA was found in viral factories identified with Hoechst staining and viral RNA polymerase antibody. The data suggest that for plasmid replication to occur, the DNA must be at the right place i.e. a site containing viral replication proteins. Presumably the plasmid diffuses into the factory region and is captured by DNA binding proteins. By taking time lapse images of live cells, plasmid DNA was detected in juxtanuclear sites at 6 to 7 h after infection and increased in intensity as the factories enlarged over the next several hours. Factory enlargement appeared to occur from within rather than by fusion of multiple small factories. We suspect that the latter might occur if higher multiplicities of virus were used. In contrast to the cytoplasmic replication of genome and plasmid DNA in VAC-infected cells, Sourvinos et al. \[[@B50]\] visualized nuclear replication of herpes simplex virus amplicons containing tetracycline operator sequence and Fraefel et al. \[[@B51]\] incorporated lacO sites into the genome of adenovirus associated virus and visualized discrete replication sites in the nucleus that fused to form larger structures. The latter study encouraged us to try to incorporate long tandem arrays of lacO repeats in the VAC genome, but they were unstable. Conclusion ========== We described a sensitive and quantitative real-time PCR method of measuring plasmid replication in cells infected with VAC and demonstrated that origin-independent replication requires all known viral replication proteins. In addition, we visualized the plasmid in living and fixed cells by incorporating tandem lacO sequences and determined that replication occurred in cytoplasmic viral factories. This system should be useful for studying the mechanism and minimal requirements of poxvirus DNA replication. Methods ======= Cells, plasmids, and viruses ---------------------------- RK-13, BS-C-1, BS-C-40, HuTK^-^143B, and HeLa cells were maintained in Eagle\'s minimal essential medium (EMEM; Quality Biologicals, Inc. Gaithersburg, MD) or Dulbecco\'s modified Eagle\'s medium (DMEM; Quality Biologicals, Inc.) containing 10% fetal bovine serum (FBS). A rabbit kidney cell line (RKD4R) stably expressing the VAC uracil DNA glycosylase and recombinant VAC vD4-ZG lacking a functional D4R gene \[[@B39]\] were gifts of F.G. Falkner. Plasmid pSV9 contains two copies of a 2.6 kbp insert derived from the VAC concatemer junction and two copies of pUC13 DNA \[[@B33]\]. Linear minichromosomes containing 1.3 kbp of VAC telomere sequences were prepared by ligation of snap cooled, EcoRI digested pSV9 essentially as described by Du and Traktman \[[@B32]\]. Ligation resulted in three products of 8 kbp, 2.6 kbp and 5.3 kbp. The 5.3 kbp minichromosome fragment was isolated by gel electrophoresis and the Qiaex II gel extraction kit (Qiagen). Plasmid p716 \[[@B52]\] was kindly provided by A. McBride; plasmids pSV2-dhfr-8.32 and p3\'SS dimer-Cl-EGFP \[[@B44]\] were gifts of A. Belmont. The temperature sensitive (ts) replication mutants Cts16, Cts24, Cts42, Cts25 with mutations in the I7, D5, E9 and B1 open reading frames, respectively were obtained from R. Condit \[[@B53],[@B54]\]; mut185 has a tsmutation in the A20 ORF \[[@B10]\]. Antibodies ---------- Cy5-conjugated affinipure F(ab\')2 fragment of donkey anti-mouse IgG and Texas red dye conjugated affinipure F(ab\')2 of donkey anti-rabbit IgG were obtained from Jackson ImmunoResearch laboratories. Alexa Fluor 594 goat anti-rabbit IgG was from Molecular probes. New England Biolabs and Invitrogen supplied the rabbit antibody to MBP and mouse anti-V5 monoclonal antibody, respectively. Transfection, infection and isolation of DNA -------------------------------------------- For experiments analyzed by real-time PCR, 0.1 μg of p716 DNA and 3.9 μg of salmon sperm carrier DNA were mixed with 10 μg of lipofectamine 2000 (Invitrogen) and uninfected cells were transfected according to the manufacturer\'s instructions. After 24 h, the cells were infected with VAC strain WR, vD4-ZG or a tsmutant at a multiplicity of 3 PFU per cell. Cells were then washed twice with Opti-MEM (Invitrogen) and overlaid with EMEM with 2.5% FBS. At various times, cells were harvested and the DNA isolated using the Qiamp DNA Blood Kit (Qiagen) according to the manufacturer\'s instructions. DNA was digested with restriction enzymes DpnI or MboI (New England Biolabs). Southern blotting ----------------- DNA (2 μg) was digested with EcoRI and DpnI or MboI, resolved on a 0.8% agarose gel, and transferred to Immobilon-Ny+ (Millipore) transfer membrane. Southern blotting was carried out as described by Maniatis \[[@B55]\]. Plasmid DNA was detected with a DNA probe that was ^32^P-labeled using a random-priming kit (Invitrogen). Pre-hybridizations and hybridizations were carried out using Quik-Hyb (Stratagene) according to the manufacturer\'s recommendation. The blot was exposed to a Phosphor screen and data acquired on a Storm 860 PhosphoImager (Molecular Dynamics, Sunnyvale, CA) and quantified with ImageQuant software (Molecular Dynamics). Real-time PCR ------------- Oligonucleotides P1 (5\'CAACTAAATGTGCAAGCAATGTAATTC3\') and P2 (5\'CATCCTGCCCCTTGCTGT3\') were designed with Primer Express software supplied by Applied Biosystems. Reactions were carried out using SYBR Green PCR master mix (Applied Biosystems), 10 μM of each primer, and 1 ng of DNA in a total volume of 50 μl in an Applied Biosystems Prism 7900HT sequence detection system with v2.1.1 software. For amplification 40 cycles at 95°C for 15 s and 60°C for 60 s were used. Construction of recombinant viruses ----------------------------------- vGFP-lacI: the open reading frame that encodes GFP-lacI was cloned by PCR using primers 5\'CAGGCTGCGCAACTGTTGGGAAGGGCGA3\' and 5\'AAAAGTACTAGCCTGGGGTGCCTAATGAGTGAGC3\' with p3\'SS dimer-Cl-EGFP \[[@B44]\] as a template. The PCR product was digested with XhoI and ScaI and then ligated to XhoI and StuI digested pSC59 \[[@B56]\] to form the plasmid pSC59gfplacI. BS-C-1 cells were infected with VAC strain WR at 0.05 PFU per cell for 1 h and then transfected with 2 μg pSC59gfplacI using 10 μg of Lipofectamine 2000. After 5 h, the medium was replaced with EMEM plus 2.5% FBS and the incubation continued for 2 days. Cells were harvested and lysed, and the diluted lysates were used to infect HuTK^-^143B cell monolayers. The cells were overlaid with medium containing low melting point agarose and 25 μg of 5-bromodeoxyuridine per ml. After three rounds of plaque purification, the viral DNA was screened for the presence of the inserted DNA by PCR. The recombinant virus was propagated and titrated as described previously \[[@B57]\]. vV5D4: primers 5\'ACTAGATACGTATAAAAAGGTATCTAATTTGATATAATGGGTAAGCCTATCCCTAACCCTCTCCTCGGTCTCGATTCTACGAATTCAGTGACTGT3\' and 5\'CTCCTGGACGTAGCCTTCGGG3\' and DNA from plasmid pER-GFP \[[@B21]\] were used to add a V5 tag to the VAC D4R gene. After double digestion of the PCR product and plasmid with SnaBI and SmaI, the products were ligated together to form the new plasmid pERV5-GFP. Approximately 10^6^RKD4R cells were infected with vD4-ZG at a multiplicity of 0.05 PFU per cell for 1 h at 37°C. The infected cells were washed twice with Opti-MEM and transfected with 2 μg of pERV5-GFP using 10 μg of Lipofectamine 2000. After 5 h, the transfection mixture was replaced with EMEM containing 2.5% FBS, and the cells were harvested at 48 h in 0.5 ml of EMEM-2.5% FBS. Lysates were prepared by freezing and thawing the cells three times and sonicating them twice for 30 s. Recombinant viruses that expressed GFP were plaque purified five times on RKD4R cells. The genetic purity of recombinant viruses was confirmed by PCR and sequencing. The recombinant virus was propagated and titrated as described previously \[[@B57]\]. Construction and expression of MBP-lacI --------------------------------------- The lacrepressor gene was PCR amplified using the following primers 5\'CGGAATTCTCATCGGGAAACCTGTCGTGCCAGCTGC3\' and 5\'CGCGGATCCTAGTGAAACCAGTAACGTTATACG3\' and template DNA from p3\'SS dimer-Cl-EGFP. The amplified fragment was cloned into the BamHI and EcoRI sites of the expression vector pMal-c2x (New England Biolabs) resulting in the plasmid pMalc2x-lacI. Luria-Bertani medium (500 ml) supplemented with ampicillin (100 μg/ml) and glucose (0.2% w/v) was inoculated with 5 ml of an overnight culture of the E. coliER2507 (New England Biolabs) containing the recombinant pMalc2x-lacIplasmid. The culture was grown at 37°C to a cell density of 0.5 at A600 nm and the expression of protein was induced for 2 h at 37°C by adding isopropyl-β-D-thiogalactopyranoside to a final concentration of 0.3 mM. The culture was then centrifuged at 4000 × gfor 20 min at 4°C. A cell extract was prepared using B-PER reagent (Pierce) according to the manufacturer\'s recommendation and the protein purified using the pMAL protein fusion and purification kit (New England Biolabs). Confocal microscopy and live cell imaging ----------------------------------------- Cells were plated on glass cover slips in 12 well plates and transfected with 1 μg of pSV2-dhfr-8.32 using 5 μg of Lipofectamine 2000. After 24 h, cells were infected with recombinant VAC at 3 PFU per cell. At 12 h after infection, cells were fixed with cold 4% paraformaldehyde in phosphate buffered saline (PBS) at room temperature for 20 min. Fixed cells were permeabilized for 5 min with PBS containing either 0.2% Triton X-100 at room temperature. Permeabilized cells were incubated with primary antibodies at a 1:100 dilution in10% FBS for 30 min, washed with PBS three times, and then incubated with secondary antibody at a 1:100 dilution in 10% FBS for 30 min at room temperature. After washing with PBS three times, cover slips were incubated with Hoechst dye for 10 min at room temperature to visualize DNA staining. Stained cells were washed extensively with PBS and cover slips mounted in 20% glycerol. Cellular fluorescence was examined under a Leica TCS NT inverted confocal microscope and images were overlaid using Adobe Photoshop version 5.0.2. For live cell imaging, HeLa cells were plated at \~80% confluence onto TC3 dishes (Bioptechs, Inc.) and infected with 3 PFU of virus per cell on the next day. Cells were imaged by either confocal or video microscopy. For video microscopy, a Hammumatsu C5985 camera and controller were attached to a Leica DMIRBE inverted fluorescence microscope. Images were digitized using an IC-PCI video capture card (Coreco Imaging, Inc.) controlled by Image Pro Plus software. Cells were maintained on a heated TC3 stage (Bioptechs, Inc.) with the temperature set at 37°C. Competing interests =================== The author(s) declare they have no competing interests. Authors\' contributions ======================= FDS participated in the design and coordination of the study, acquisition and analysis of data, and preparation of the manuscript. BM designed and coordinated the study, assisted in the data analyses and contributed to the preparation of the manuscript. Acknowledgements ================ We thank Norman Cooper for invaluable assistance with cell culture, Owen Schwartz for helping in confocal microscopy and live cell imaging, and Mike Baxter for his assistance in real-time PCR. A. McBride and A. Belmont provided plasmids and R. Condit and F. Falkner donated mutant viruses and a cell line.	PubMed Central	2024-06-05T03:55:55.648853	2005-3-22	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1079961/", "journal": "Virol J. 2005 Mar 22; 2:23", "authors": [ { "first": "Frank S", "last": "De Silva" }, { "first": "Bernard", "last": "Moss" } ] }
PMC1079962	Background ========== Rotaviruses remain the most common cause of acute gastroenteritis worldwide and cause an estimated 600,000 deaths in children less than 5 years of age \[[@B20]\]. The high disease burden motivated major efforts to develop a suitable rotavirus vaccine. However, the vaccine efficacy is being challenged by the extensive strain diversity of the rotaviruses \[[@B3],[@B7]-[@B9],[@B13],[@B14]\]. Rotaviruses belong to the Reoviridae, and their genome consists of 11 segments of double stranded RNA. The gene segment coding for the VP7 glycoprotein is the basis for genotyping group A rotaviruses into at least 15 G-genotypes. Among them, G1, G2, G3, G4 and G9 are the most common G-types in humans \[[@B5],[@B15],[@B16],[@B19],[@B21],[@B23]\]. The importance of type-specific immunological protection against rotavirus disease is still under discussion \[[@B13]\]. G-genotyping is performed using type-specific-primer-based RT-PCR. Two common primer sets introduced by Gouvea et al. \[[@B6]\] and Das et al. \[[@B2]\] are currently used in rotavirus G-typing surveillance programs \[[@B22]\]. A failure to genotype or mistyping has already been reported from different parts of the world. These reports showed that nucleotide sequence differences between the target region of the respective genes and the primer sequences used for typing led to the genotyping failure \[[@B1],[@B10],[@B11],[@B17]\]. The Dhaka hospital of ICDDR,B, situated in the central Bangladesh, and the Matlab hospital, located 45 km south east of Dhaka respectively treat about 100,000 and 15,000 diarrhoeal patients each year. A hospital surveillance system has been established in these hospitals by ICDDR,B to collect information on clinical, epidemiological and demographic characteristics of the patients attending the hospital since 1978. In Bangladesh, rotavirus strains have previously been typed using a variety of techniques. Serotyping was introduced by Ward et al. \[[@B28]\] with specimens collected during 1985--1986 in Dhaka using neutralization with hyperimmune antisera against prototype rotavirus strains G1, G2, G3 and G4. They concluded that epitopic variations between rotavirus strains influenced the sensitivity of serotyping. Fun et al. \[[@B4]\] detected the major rotavirus types (G1 to G4) by RNA hybridization with serotype-specific synthetic oligonucleotide probes, but this method could not type 33.3% of the Bangladeshi rotaviruses. Likewise, RT-PCR depending on type-specific oligonucleotide primers failed to type a significant portion of rotaviruses. Rotavirus surveillance studies in Bangladesh between 1987 and 1997 reported that 1,095 (43.7%) samples out of 2,515 were G-untypeable \[[@B25]-[@B27]\]. In this study, we characterized rotavirus positive stool specimens collected in the Dhaka and Matlab hospitals during 2002 by using RT-PCR based on the primer set described by Das et al. \[[@B2]\]. We found that a major proportion of the specimens were untypeable. Nucleotide sequences of VP7 genes were performed to investigate the reason why they were untypeable with the routine primer set. The untypeable specimens were further characterized by using a different primer set described by Gouvea et al. \[[@B6]\]. Results and Discussion ====================== Detection of rotavirus strains ------------------------------ In 2002, a total of 3,803 patients with history of watery diarrhea were included in the hospital surveillance system. In Dhaka and Matlab, 535 (27.2%) and 358 (19.4%) specimens were positive for group A rotavirus antigens by enzyme immunoassay. G typing using the Das primer set --------------------------------- Rotavirus G-typing was carried out for all rotavirus positive specimens from Matlab and for every fourth of the rotavirus-positive specimens from Dhaka. Some samples were excluded from this study due to unavailability of sufficient amount of stool specimens for testing. G-typing was performed on 433 rotavirus ELISA-positive stool samples by RT-PCR using the primer set described by Das et al. \[[@B4]\], which was routinely used in our laboratory. The most prevalent G types were G9 (20.5%); G2 (14.6%), and G4 (13.8%). G1 comprised only 11.6% of the isolates and 36.9% of the rotavirus-positive samples were untypeable. VP7 gene sequence analysis -------------------------- We amplified the VP7 genes of five randomly selected untypeable strains (Dhaka162-02, Dhaka18-02, Dhaka164-02 Dhaka165-02 and Matlab26-02) using the VP7 consensus primers Beg9-End9 as described by Gouva et al. \[[@B6]\] and sequenced their complete open reading frame \[GenBank:AY631050, GenBank:AY631054\]. They were typed as G1 rotaviruses by using BLAST homology searches (99--100% nucleotide and amino acid identities with the Indian G1 rotavirus strain, ISO-4). To compare them with the typeable G1 sequences, the VP7 genes of two typeable G1 strains, Dhaka8-02 \[GenBank:AY631049\] and Matlab159-02 \[GenBank:AY631055\] were sequenced. We found that the nucleotide sequences of the typeable and untypeable G1 strains were 100% identical at the G1 primer-binding sites. We aligned the target G1 VP7 sequence with the Das G1 primer sequence (reverse primer, 9T1-1; 5\'-TCTTGTCAAAGCAAATAATG-3\'; nt 176--195, prototype strain Wa \[GenBank:M21843\]) to determine if there was any mismatch between them. Four mismatches were found in the Das G1 primer, 9T1-1, at the 5\' end (Fig. [2](#F2){ref-type="fig"}). Due to these mismatches, the Das G1 primer failed to detect most (75%) of the G1 strains. Since, the primer set had perfect matches at the 3\' end, it could detect 25% of the G1 rotaviruses. When we compared the target sequence with the Gouvea G1 primer sequence (forward primer, aBT1; 5\'-CAAGTACTCAAATCAATGATGG-3\'; nt 314--335, prototype strain Wa), we found only one mismatch (Fig. [2](#F2){ref-type="fig"}). Therefore, the Gouvea G1 primer was found to be more suitable for typing our G1 strains. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Nucleotide mismatches in the primers. The target sequence is the VP7 gene of G1 Bangladeshi strains. The G1 rotavirus VP7 gene specific primers were described by Das et al. \[2\] and Gouvea et al. \[6\]. Mismatches are in red. ::: ![](1743-422X-2-24-2) ::: Distribution of G types using Gouvea primer set ----------------------------------------------- The untypeable specimens were typed using the primer set described by Gouvea et al. \[[@B6]\]. After typing with the Gouvea primer set, the distribution of rotavirus G-types changed dramatically (Fig. [1](#F1){ref-type="fig"}). Type G1 now accounted for 44.8% of the isolates and became the most prevalent genotype, and the number of untypeable strains was reduced from 36.9 to 2.1 %. The other common G types were G9 (21.7%), G2 (15.0%), and G4 (13.8%). The previous studies in Bangladesh reported that G4 strains were the most prevalent strains during 1992--1997 and a significant number of rotavirus strains were untypeable using the Das primer set \[[@B26]\]. It is likely that the Das primer set could not detect most of the G1 rotaviruses in the previous years and that a majority of the untypeable rotaviruses were G1 strains. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Effect of untypeable strains on G typing of Bangladeshi rotavirus strains isolated in 2002 (n = 433). ::: ![](1743-422X-2-24-1) ::: Conclusion ========== Because of the natural variation in the rotaviral gene sequences, close monitoring of rotavirus genotyping methods is important. The findings described in this paper will be important for genotyping strategies in the rotavirus surveillance studies. Materials and methods ===================== Sample collection ----------------- Stool specimens were collected from patients who presented with diarrhea to the Dhaka and Matlab hospitals of ICDDR,B in 2002. In the Dhaka hospital, stool specimens are routinely collected from every 50^th^patient and in Matlab hospital, every patient with diarrhea submits a stool specimen for testing. Rotavirus antigen detection --------------------------- Rotavirus antigens (group A-specific VP6 proteins) were detected in the stool specimen using a solid phase sandwich type enzyme immunoassay modelled after Dakopatts commercial kit incorporating rabbit hyperimmune antisera produced at ICDDR,B and an anti-human rotavirus-horseradish peroxidase conjugate (Dakopatts, Copenhagen, Denmark) using the same criteria for determination of positivity as those used by the Dakopatts kit \[[@B26]\]. RNA extraction -------------- The QIAamp Viral RNA mini kit (Qiagen/Westburg, Leusden, The Netherlands) was used according to the manufacturer\'s instructions for the extraction of rotavirus RNA from the stool samples. RT-PCR ------ A reverse transcriptase-polymerase chain reaction (RT-PCR) was carried out using the Qiagen OneStep RT-PCR Kit (Qiagen/Westburg) as previously described by Das et al. \[[@B2]\] and Gouvea et al. \[[@B6]\] for rotavirus G-types (G1, G2, G3, G4 and G9) using type-specific oligonucleotide primers. The reaction was carried out with an initial reverse transcription step at 45°C for 30 min, followed by 35 cycles of amplification (30 sec at 94°C, 30 sec at 50°C, 1 min at 72°C), and a final extension of 7 min at 72°C in a thermal cycler (Eppendorf, Hamburg, AG). PCR products were run on a 2% agarose gel, and stained with ethidium bromide. Specific segment sizes for the different G types were visualized under UV-light. Nucleotide sequencing --------------------- The amplified PCR products were purified with the QIA quick PCR purification kit (Qiagen/Westburg), and sequenced in both directions using the dideoxy-nucleotide chain termination method with the ABI PRISM^®^BigDye Terminator Cycle Sequencing Reaction kit (Perkin-Elmer Applied Biosystems, Foster City, California) on an automated sequencer (ABI PRISM™ 3100). The Beg9 and End9 RT-PCR primers were used as sequencing primers. Sequence analysis ----------------- The chromatogram sequencing files were inspected using Chromas 2.2 (Technelysium, Queensland, Australia), and consensus sequences were prepared using SeqMan II (DNASTAR, Madison, WI). Multiple sequence alignments were performed using CLUSTALX 1.81 \[[@B24]\]. Sequences were manually edited in the GeneDoc version 2.6.002 alignment editor \[[@B18]\]. Sequence submission ------------------- The nucleotide sequence data were deposited in GenBank using the National Center for Biotechnology Information (NCBI, Bethesda, MD) Sequin 5.15 submission tool <http://www3.ncbi.nlm.nih.gov/> under accession numbers AY631049-AY631055. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= MR carried out the laboratory tests and wrote the manuscript; RS and GP carried out RT-PCR tests; JM performed the sequencing experiments; AF, KZ, RB and DS supervised the rotavirus surveillance program and critically revised the manuscript; MVR and TA supervised the study. Acknowledgements ================ This study was funded by the Program for Appropriate Technology in Health (PATH), grant number GAT 770-790-01451-SPS. ICDDR,B acknowledges with gratitude the commitment of PATH to the Centre\'s research efforts.	PubMed Central	2024-06-05T03:55:55.651454	2005-3-24	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1079962/", "journal": "Virol J. 2005 Mar 24; 2:24", "authors": [ { "first": "Mustafizur", "last": "Rahman" }, { "first": "Rasheda", "last": "Sultana" }, { "first": "Goutam", "last": "Podder" }, { "first": "Abu SG", "last": "Faruque" }, { "first": "Jelle", "last": "Matthijnssens" }, { "first": "Khalequz", "last": "Zaman" }, { "first": "Robert F", "last": "Breiman" }, { "first": "David A", "last": "Sack" }, { "first": "Marc", "last": "Van Ranst" }, { "first": "Tasnim", "last": "Azim" } ] }
PMC1079963	Background ========== The dengue viruses have become recognized as important global pathogens causing dengue haemorrhagic fever not only in Southeast Asia but also in South and Central America and in the Caribbean.\[[@B1],[@B2]\]. There are 4 closely related dengue viruses referred to as DENV-1, DENV-2, DENV-3 and DENV-4\[[@B3]\]. They are mosquito borne viruses with a single stranded positive sense RNA genome around 11 kilobases in length, and are able to infect both mosquito and human hosts. A wide range of cell types from multiple species is susceptible to infection with dengue viruses in vitro. Numerous studies have attempted to identify the cell surface receptor or receptors utilized by the dengue viruses to gain entry into susceptible cells, but multiple approaches using different cell lines and different dengue virus strains have generated many candidate DENV interacting proteins identified in some cases only by molecular mass \[[@B4]-[@B11]\]. Heparan sulfates\[[@B12]\] and the C-type lectins DC-SIGN and L-SIGN have been shown to mediate infection by dengue viruses\[[@B13]\] and most recently, studies using a standard virus overlay protein binding assay (VOPBA) have suggested that in the liver cell line HepG2, different DENV serotypes utilize different cell surface molecules\[[@B14]\]. More specifically, mass spectrometric methods have been used to identify reactive bands using VOPBA and it has been suggested that DENV-2 interacts with GRP78\[[@B15]\] while DENV-1 interacts with the 37 kDa/67 kDa high affinity laminin receptor\[[@B16]\]. In a standard VOPBA, complex protein preparations are separated according to molecular mass in a single dimension and transferred to membranes to be probed with virus antigen\[[@B17],[@B18]\]. When complex mixtures are used there are often many co-migrating proteins that cannot be adequately resolved for accurate interpretation of mass spectrometric data when single dimension separations are used. We have used two dimensional (2D) gel electrophoresis\[[@B19]\] for separation of cell membrane preparations. After electrotransfer of 2D gel-separated proteins to nitrocellulose membranes, we probed the membranes with various dengue virus antigen preparations. This 2D VOPBA approach has facilitated the separation of multiple proteins of similar molecular mass along a pH gradient. Reactive spots recovered using this approach were identified on companion 2D gels using matrix assisted laser desorption ionization-time of flight mass spectrometry (MALDI TOF MS)\[[@B20]\]. A schematic diagram of our experimental design is shown in figure [1](#F1){ref-type="fig"}. To the best of our knowledge, this is the first description of the use of 2D VOPBA for identification of proteins interacting with flaviviruses. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Flowchart of experimental design.This describes the steps in the process of identifying dengue virus reactive proteins by 2D VOPBA to facilitate picking the relevant spots from a gel run under identical conditions and at the same time as those which were used for electrotransfer to nictrocellulose membranes. ::: ![](1743-422X-2-25-1) ::: Results ======= Identification of reference protein spots in detergent extracts of PS Clone D cell monolayer -------------------------------------------------------------------------------------------- The proteomic profile of 1% sodium deoxycholate (NaDOC) soluble proteins from PS Clone D cell monolayers is shown in figure [2a](#F2){ref-type="fig"}. Numerous spots were seen and it was necessary to attempt to fractionate the components of the cell monolayer by sequential treatment with 1% β octyl-glucopyranoside (βOG) followed by 1% NaDOC. Beta octyl-glucopyranoside soluble proteins and post-βOG residual NaDOC-extracted proteins were resolved separately on 2D gels. Major protein spots were picked and subjected to in-gel trypsin digestion followed by MALDI-TOF MS. It was found that the βOG extract contained both cytoplasmic as well as some membrane proteins while the post-βOG residual NaDOC extract contained mainly cytoskeletal and nucleus associated proteins. Figures [2b](#F2){ref-type="fig"} and [2c](#F2){ref-type="fig"} show some of the major spots identified and used as identifying features when comparing overlays of Ponceau S stained blots with VOPBA developed blots. In figure [2b](#F2){ref-type="fig"} showing separation of βOG extracts, BiP or GRP78 was prominent as was calreticulin, alpha enolase, HSP70, PDI and actin. Circled spots were reactive in the 2D VOPBAs described later and are shown here to provide the protein landscape in which these reactive proteins exist. In figure [2c](#F2){ref-type="fig"}, showing separation of βOG-insoluble NaDOC-extracted proteins, BiP/GRP78 was less prominent but vimentin and nucleophosmin were highly prominent. Again the circled spots mark the positions of VOPBA-reactive spots, showing the location of these very sparse proteins in the landscape of much more abundant protein spots. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### 2D gel electrophoresis image of detergent extracts from PS Clone D cell monolayer.The first dimension was run on linear 7 cm IPG strip, pH 3-10. The second dimension was 10% SDS PAGE. The gel was stained with coomasie brilliant blue and spots were picked and subjected to in-gel trypsin digestion. The major spots identified by MALDI TOF are labelled. Circled spots were reactive in the 2D VOPBA. (a) NaDOC-extract (b) βOG-extract (c) NaDOC-extract after removal of βOG-soluble proteins. ::: ![](1743-422X-2-25-2) ::: 2D VOPBA of β octyl-glucopyranoside extract ------------------------------------------- 2D gel blots of βOG extracted PS Clone D cells were probed with a cocktail of antigens prepared from 4 different dengue virus serotypes grown in C6/36 mosquito cells. Bound envelope protein (E) was detected by using the monoclonal antibody 4G2, a flavivirus-reactive anti-E antibody\[[@B21]\]. Replicate blots were also probed with individual dengue serotypes separately. Antigens prepared from uninfected mosquito cells were used as negative controls in order to identify non-dengue specific interactions present in all blots. These blots are shown in figure [3](#F3){ref-type="fig"}. In the 2D VOPBA blots probed with a dengue virus antigen or cocktail, 2 reactive spots were seen. The blots probed with dengue cocktail, DENV-1, DENV-2 and DENV-3 had identical reactivities. The major spot that was clearly reactive was around 45 kD in molecular mass with a pI of 4--5. A weaker, less obvious reaction was seen with a spot at 50--60 kD and a slightly higher pI of around 5. MALDI-TOF MS-generated peptide mass fingerprints identified these spots as 37 kDa/67 kDa laminin binding protein or laminin receptor (LAMR1) and an SH3 domain-containing protein, Hip 55, respectively. The blot probed with DENV-4 did not show any reactivity with LAMR1 but there was a weak reactivity with Hip 55. Furthermore, a clearly reactive pair of spots identified as p47 protein (a cofactor of NSFL1/p97) was seen in the DENV-4 VOPBA. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### 2D VOPBA of β octyl-glucopyranoside soluble proteins from PS Clone D cell monolayer.The antigen preparation used in the VOPBA staining of the 2D blots is labelled on the bottom left of each blot. Two images are superimposed in each panel. The Ponceau S scan of each blot is shown in greyscale and shows the universe of spots transferred to the blot. The spots reactive in the VOPBA analysis are shown in blue. Spots marked with an arrow were identified as Hip-55, those marked with a circle were identified as LAMR1 and the pair of spots in the DENV4 blot marked with a rectangle were identified as p47 protein (NSFL1 cofactor). ::: ![](1743-422X-2-25-3) ::: 2D VOPBA of sodium deoxycholate extract of cells previously extracted with β octyl-glucopyranoside -------------------------------------------------------------------------------------------------- Residual protein from PS Clone D cells treated with βOG was further treated with NaDOC and the resulting extract was used to prepare 2D gel blots. VOPBAs were performed as above and figure [4](#F4){ref-type="fig"} shows the individual VOPBA blots. As with the βOG-extract, LAMR1 was found to be reactive in VOPBA blots probed with dengue cocktail, DENV-1, DENV-2 and DENV-3, but not DENV-4. Instead, the DENV-4-probed blot showed a clear reaction with a protein of higher molecular mass and higher pI than LAMR1. This protein was identified to be lamin B1, a nuclear membrane protein. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### 2D VOPBA of sodium deoxycholate soluble proteins after β octyl-glucopyranoside extraction of PS Clone D cell monolayer.The antigen preparation used in the VOPBA staining of the 2D blots is labelled on the bottom left of each blot. Two images are superimposed in each panel. The Ponceau S scan of each blot is shown in greyscale and shows the universe of spots transferred to the blot. The spots reactive in the VOPBA analysis are shown in blue. Spots marked with a circle were identified as LAMR1 and the spot in the DENV4 blot marked with a rectangle was identified as lamin B1. ::: ![](1743-422X-2-25-4) ::: MALDI TOF MS ------------ Peptide mass fingerprints were obtained using Voyager DE STR (Applied Biosystems, Foster City, CA, USA) and mass lists were submitted for search against the NCBI protein database (National Institutes of Health, Bethesda, MD, USA) using the MASCOT search engine\[[@B22]\]. Spectra and their corresponding mass-lists generated from analysis of the tryptic digests are provided as [additional files 1](#S1){ref-type="supplementary-material"},[2](#S2){ref-type="supplementary-material"},[3](#S3){ref-type="supplementary-material"},[4](#S4){ref-type="supplementary-material"},[5](#S5){ref-type="supplementary-material"}. Immunoblot confirmation of LAMR1 and lamin B1 spots on 2D gel blots ------------------------------------------------------------------- The NaDOC soluble proteins from βOG extracted PS Clone D cells were used to prepare 2D gel blots and these were probed with antibodies specific for LAMR1 and lamin B1. Figure [5a](#F5){ref-type="fig"} shows a Ponceau S stained blot and the dengue virus antigen reactive spots are circled. The locations of specific staining by antibodies to LAMR1 and lamin B1 on the 2D blots were confirmed to be in the positions of the spots identified by MALDI TOF MS as LAMR1 and lamin B1 (figures [5b](#F5){ref-type="fig"} and [5c](#F5){ref-type="fig"}). ::: {#F5 .fig} Figure 5 ::: {.caption} ###### 2D gel immunoblots probed with LAMR1 and lamin B1 specific antisera.Two dimensional separation of βOG-insoluble NaDOC extracts transferred to nitrocellulose and probed with specific antisera shows specific staining of the MALDI TOF MS identified spots. Panel (a) shows a Ponceau S stained blot of a 2D gel prior to probing with specific antibody, (b) shows a blot probed with LAMR1 specific antisera and (c) shows a blot probed with lamin B1 specific antisera. In panel (a) the circle marks the position of LAMR1 and the rectangular box marks the position of lamin B1. ::: ![](1743-422X-2-25-5) ::: LAMR1 is expressed on the surface of PS Clone D cells ----------------------------------------------------- When PS Clone D cells were fixed with paraformaldehyde LAMR1 staining was seen on the surface of the cells. In keeping with observations of de Hoog and coworkers\[[@B23]\] we found that single cells at the edges of the monolayer were consistently displaying brightly fluorescent patches on the cell surfaces showing high levels of expression of LAMR1. In the centres of the monolayers where cells were contact inhibited, LAMR1 staining was more muted and more evenly distributed over the cells (see figure [6](#F6){ref-type="fig"}). ::: {#F6 .fig} Figure 6 ::: {.caption} ###### LAMR1 is expressed on the surface of PS Clone D cells.The top 2 panels and the bottom left panel show individual cells with patches of bright staining of LAMR1 on the surface of paraformaldehyde fixed cells. When contact inhibition occurred as shown in the bottom right panel, LAMR1 was more evenly distributed and the dense patches of LAMR1 were no longer present. ::: ![](1743-422X-2-25-6) ::: Discussion ========== The investigation of early events in the infection of susceptible cells by dengue viruses is important in the quest to understand the ability of this group of mosquito borne viruses to infect both insect and mammalian cells, yet appear to have a restricted tissue tropism in the human host. Determination of the nature of the early interactions of the infecting viruses with molecules on the surface of susceptible cells provides for the possibility that this understanding can lead to the development of therapeutic agents that can be used to inhibit virus infection. The laminin receptor has already been described as a receptor for DENV-1 by single dimension VOPBA followed by MS/MS\[[@B16]\]. The results from the MASCOT\[[@B22]\] search showed multiple hits in this particular study, indicating several possibilities, including ATP synthase β chain, β actin and the laminin receptor, but the authors selected the lower scoring laminin receptor for further investigation. This was a reasonable choice since this molecule has previously been identified as a receptor for the alphaviruses Sindbis virus\[[@B24]\] and Venezuelan equine encephalitis virus (VEEV)\[[@B25]\] and the flavivirus tick-borne encephalitis virus (TBEV)\[[@B26]\]. In our study, 2D VOPBA was used to eliminate the problem of co-migrating bands in a single dimension and the mass list generated from trypsin digestion of the most prominent reactive spot turned up numerous hits unambiguously listing the same protein, variously known as protein 40 kD, laminin-binding protein, 34/67 kDa laminin receptor, laminin receptor 1, LAMR1, Lamr1 protein, 67 kDa laminin receptor, 40S ribosomal protein SA, and so on. Contrary to earlier suggestions that LAMR1 is a DENV-1 specific receptor\[[@B14],[@B16]\], in our hands, DENV-1, DENV-2 and DENV-3 were all shown to interact with the same molecule LAMR1 although DENV-4 did not. It is thus likely that in the PS Clone D cells we have studied, at least 3 of the 4 different dengue serotypes utilize the same surface protein to gain entry into the cells. In our study we also did not find any evidence of DENV-2 binding to BiP/GRP78 as has been shown using single dimension VOPBA\[[@B15]\]. LAMR1 is a non-integrin receptor interacting with the extracellular matrix. It is generally accepted that the 37 kDa form is the precursor to the 67 kDa form although it is still not clear how this transition occurs. We have used a commercially available rabbit polyclonal antibody raised against a recombinant protein corresponding to amino acids 110--250 of human LAMR1 to show that this protein can be found in patches on the surface of PS clone D cells which are not contact inhibited in vitro. This distribution of LAMR1 is consistent with the findings of Donaldson et al\[[@B27]\], and it is thus possible that LAMR1 may be utilized as a receptor by the dengue viruses. Other groups have already shown that LAMR1 is a functional receptor for Sindbis virus, VEEV, TBEV and DENV1\[[@B16],[@B24]-[@B26]\]. We have not included functional studies in this present work, as this study is meant to be an exploratory study of dengue virus interacting proteins in PS Clone D cells using 2D VOPBA as an interrogating tool. There is no doubt that the treatment of the cell extracts limits the ability of this method to identify interactions dependent upon conformational structures, nevertheless, we managed to identify LAMR1 confirming previous observations by other investigators\[[@B27]\]. We have further shown that LAMR1 interaction is not limited to DENV-1 alone, but that DENV-2 and DENV-3 also interact with LAMR1 and that this may be a common receptor for dengue virus entry into cells. Many integrins have been shown to function as receptors for different viruses, for example the α6 integrins mediate human papillomavirus entry\[[@B28]\], β3 integrins mediates cell entry by hantaviruses.\[[@B29]\], α2β1 integrin is a receptor for human echovirus 1.\[[@B30],[@B31]\], α5β1 integrin binds human parvovirus B19\[[@B32]\], α2β1 and αxβ2 mediate rotavirus infection\[[@B33]\] and αvβ3 is the receptor for the flavivirus West Nile virus (WNV).\[[@B34]\]. Many viruses are now known to infect cells through a multistep process involving binding to the cell surface followed by internalization, often through interacting with more than one surface molecule. Outside-in binding of integrins leads also to signal transduction, and this functional activation has been shown to be necessary for internalization as in the example of human parvovirus B19. In the case of adenoviruses, integrin clustering due to receptor binding initiates the signalling events required for internalization\[[@B35]\]. The finding that DENV-1, DENV-2 and DENV-3 interact with the laminin receptor is thus consistent with this growing body of work describing the utilization by viruses of extracellular matrix protein receptors for gaining entry into cells. In this study we have also identified dengue virus envelope protein interaction with an actin binding protein Hip55, which has been shown to be involved in endocytosis, vesicular transport and signal transduction\[[@B36]\]. Hip55 has also been shown to interact with CD2v protein of African swine fever virus (ASFV) and colocalizes to areas surrounding perinuclear virus factories in ASFV infected cells\[[@B37]\]. The role of Hip55 in the life cycle of dengue virus should be further investigated. In our preparations of dengue virus antigens, DENV-4 antigen had the weakest reactivity in ELISA (data not shown) suggesting a lower titre of antigen than the other 3 serotypes, but the DENV-4 2D VOPBA did show a different reactivity pattern, picking out lamin B1 instead of LAMR1. Lamin B1 is not a plasma membrane protein but is part of the nuclear membrane\[[@B38],[@B39]\] and is thus unlikely to be an alternative receptor for DENV-4. The significance of the reaction of DENV-4 envelope protein with lamin B1 is unclear and will be the subject of further studies. It is also interesting that DENV-4 envelope also reacted with another protein involved in vesicle transport and target membrane fusion, the p47 protein cofactor of NSFL1/p97\[[@B40]\]. The similarity of function of the p47 protein with that of Hip 55, which is reactive with all 4 dengue serotypes, suggests that DENV-4 may use a different pathway than DENV-1, DENV-2 and DENV-3 in the course of infection of a particular cell. Conclusion ========== Two-dimensional VOPBA was used to identify cell membrane proteins interacting with dengue virus envelope protein. This approach identified several interactors including LAMR1, a non-integrin laminin binding protein, which has previously been suggested as a receptor for DENV-1 but not other dengue virus serotypes. Using more rigorous tools we have shown clearly that LAMR1 interacts not only with DENV-1 but also with DENV-2 and DENV-3. We have further shown that dengue virus envelope protein from all 4 serotypes also interacts with an actin binding protein Hip55 and that DENV-4 differs from the other three dengue virus serotypes in that it\'s envelope protein interacts with lamin B1 and p47 and does not interact with LAMR1. Methods ======= Preparation of virus antigens ----------------------------- The 4 prototype dengue viruses were used in this study. All viruses were propagated in Aedes albopictusC6/36 cells grown in Leibovitz 15 media supplemented with 5% heat inactivated foetal calf serum, antibiotics and 10% tryptose phosphate broth. Antigens were prepared by inoculating C6/36 cell monolayers with the different DENV serotypes as described previously\[[@B41]\] and harvested when syncytium formation was extensive. Cell culture fluids were clarified by centrifugation before use. Fluids similarly prepared from mock infected C6/36 cells were used as negative antigen controls. Preparation of cell and membrane extracts ----------------------------------------- Just confluent flasks of the porcine kidney cell line PS Clone D were used in the preparation of detergent extracts for separation by 2D gel electrophoresis. Monolayers were washed twice with phosphate buffered saline (PBS) before subjecting to treatment with 1% β octyl-glucopyranoside (βOG) in a hypotonic buffer containing 10 mM HEPES, 1.5 mM MgCl~2~, 5 mM KCl and a protease inhibitor cocktail (Boehringer Mannheim GmbH, Germany), pH 7.5 rocking at 4°C for 1 hour. The solution was removed and designated βOG-extract. The remaining membranes, cytoskeleton and nuclei were then washed for 1 hour at 4°C with a solution containing 2% CHAPS in the same buffer as described above. The resulting solution was discarded and the residual material solubilized by rocking at 4°C for 1 hour in the above buffer containing 1% sodium deoxycholate (NaDOC). The resulting solution was removed and designated βOG-insoluble NaDOC-extract. All extracts were spun in a microfuge at 14,000 rpm for 10 minutes and the supernatants stored at -20°C until use. Sample preparation for 2D gel electrophoresis --------------------------------------------- All samples were prepared for 2D gel electrophoresis using the Ready Prep 2-D Cleanup Kit (BioRad Laboratories, Hercules, CA, USA) according to the manufacturer\'s instructions. 2D gel electrophoresis ---------------------- Protein pellets were resolubilized in IPG strip rehydration solution (8 M urea, 2% CHAPS, 40 mM DTT, 0.5% IPG buffer pH3-10, bromophenol blue) at room temperature for 30 minutes, then spun in a microfuge at 14,000 rpm for 10 min. 125 ul of the resulting supernatant was used for each IPG strip (ReadyStrips pH 3-10, 7 cm, BioRad Laboratories, Hercules, CA, USA) and rehydration was achieved at 50 uA for 15 hours at 20°C using the IPGphor IEF system (Amersham Pharmacia Biotech, Uppsala, Sweden). Subsequently IEF was carried out for 30 minutes at 500 V, 30 minutes at 1000 V and 2 to 2.5 hours at 8000 V with a step-and-hold gradient until a total of 8500 volt-hours had been achieved. IPG strips were then washed with distilled water and then equilibrated by rocking for 20 minutes at room temperature in SDS equilibration buffer (50 mM Tris-HCl pH 8.8, 6 M urea, 30% glycerol, 2% SDS) containing 10 mg/ml DTT, allowing for at least 5 ml of buffer per strip. Strips were then washed with distilled water and placed on the top surface of the second dimension gel which was a 10% SDS polyacrylamide gel polymerized overnight. Molecular weight markers were applied onto small pieces of chromatography paper and inserted next to each strip on the top of each gel, after which the strips and markers were sealed with 0.7% agarose in 0.125 M Tris-HCl pH 6.8. The second dimension separation of proteins by molecular mass was achieved at a constant 140 V (Mini Protean 3, BioRad Laboratories, Hercules, CA, USA). Electrotransfer of 2D gels to nitrocellulose -------------------------------------------- The Hoefer TE series Transfor Electrophoresis Unit (Hoefer Scientific Instruments, San Francisco, CA, USA) was used to electrotransfer proteins from 2D gels to nitrocellulose membranes at 200 mA for 1 hour in ice cold Towbin buffer (25 mM Tris, 192 mM glycine, 20% methanol). Nitrocellulose blots were then stained using Ponceau S. A record of the positions of the visible protein spots on each blot was made by scanning the Ponceau S probed blot using ImageScanner (Amersham Pharmacia Biotech, Uppsala, Sweden) and the software ImageMaster Labscan v3.00 (Amersham Biosciences, UK). After scanning, the Ponceau S was stripped by washing in water and the blots were then blocked by rocking for 1 hour in PBS containing 5% skimmed milk. Virus overlay protein binding assay (VOPBA) ------------------------------------------- The 2D blots were incubated overnight with rocking at room temperature with clarified antigen preparations and mock-infected controls. The blots were then washed with PBS and incubated with the anti-flavivirus monoclonal antibody 4G2 in another overnight incubation at room temperature. After washing with PBS, the blots were incubated with rabbit-anti-mouse Ig HRP (DAKO, Glostrup, Denmark) at 1:1000 dilution in 5% skimmed milk in PBS for 2 hrs at room temperature. The blots were then washed with PBS and reactive protein spots were visualized by developing with the chromogenic substrate, 4-chloro-1-naphthol/hydrogen peroxide. Reaction was stopped after 1 hr by washing with water. The membranes were scanned and compared with the Ponceau S images scanned previously using Adobe Photoshop version 5.0 LE (Adobe Systems Inc., San Jose, CA, USA). In-gel trypsin digestion and analysis by MALDI-TOF MS ----------------------------------------------------- Reactive spots seen on the blots were identified in the Ponceau S scans which had been recorded previously and the corresponding spots in the coomassie blue-stained gel were picked and stored in UHQ water in 0.5 ml microfuge tubes at 4°C. During all steps in the digestion process the buffer used was 5 mM NH~4~HCO~3~. Gel spots were first destained with 50% HPLC grade methanol. Destained spots were dehydrated with acetonitrile for 10 minutes before incubation with 10 mM DTT for 50 minutes at 55°C. This was followed by incubation with 55 mM iodoacetamide (IAA) for 30 minutes at room temperature in the dark. The spots were then washed twice with buffer for 20 minutes each time, dehydrated with acetonitrile for 10 minutes and rehydrated with buffer. Finally the gel spots were dehydrated twice with acetonitrile for 10 minutes each time and dried completely by centrifugation under vacuum (DNA Speed-Vac DNA110, Savant Instruments Inc, Farmingdale, NY, USA) for 10 minutes. Each gel spot was then reswelled in 5 ul of 12.5 ng/ul of sequencing grade trypsin (Promega, Madison, WI, USA) in 5 mM NH~4~HCO~3~for 45 minutes on ice. Excess trypsin solution was then removed, the spots were covered in 5 ul buffer and digestion was allowed to proceed at 37°C overnight. Digests were stored at -20°C until analysed. Analysis by MALDI-TOF MS ------------------------ For MALDI analysis, digests were thawed, spun in a microfuge at 14,000 rpm for 10 minutes. One ul of the supernatant was mixed in a 1:1 ratio with a 1:10 dilution of saturated α-cyano-4-hydroxycinnamic acid (ACCA) matrix in 0.25% trifluoroacetic acid, 50% acetonitrile, 50% water. This mixture was spotted onto MALDI target plates and spectra were acquired using Voyager-DE STR Biospectrometry workstation (Applied Biosystems, Foster City, CA, USA). Peptide mass lists were submitted for search against the NCBI database (National Institutes of Health, Bethesda, MD, USA) using the MASCOT search engine (Matrix Science, London, UK). No constraints were set for species but carbamiodomethylation of cysteine residues and possible missed-cleavages were included. Immunostaining of 2D gel blots ------------------------------ The 2D blots were incubated with polyclonal rabbit antisera against LAMR1 and Lamin B1 diluted 1:200 in PBS with 5% skimmed milk at room temperature, overnight with rocking. After extensive washing with PBS, the bound antibodies were detected with anti rabbit Ig conjugated with horseradish peroxidase, and visualized using the chromogenic substrate 4-chloro-1-naphthol/hydrogen peroxide as described above. Antisera were obtained from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Surface staining of cells and photomicrography ---------------------------------------------- Cells were resuspended at 1 × 10^5^cells per ml in Leibovitz 15 media containing 3% heat inactivated foetal calf serum, antibiotics and tryptose phosphate broth. Resuspended cells were delivered in 25 ul volumes to individual wells of multitest slides (Erie Scientific Co., Portsmouth, NH, USA) and allowed to adhere overnight in a moist box at 37°C. Cells were then washed in PBS and fixed with 3.7% paraformaldehye in PBS at pH 7.4 for 15 minutes followed by a shift to 2% paraformaldehyde in PBS at pH 8.5 for a further 15 minutes. After washing in PBS slides were air dried and stored at -20°C until use. Prior to staining, slides were incubated in 50 mM ammonium chloride in PBS for 5 minutes, washed thoroughly and blocked in 1% foetal calf serum in PBS for 30 minutes. Immunofluorescence staining of the surface of cells was achieved by incubation for 1 hour with polyclonal rabbit antisera against LAMR1 at 1:25 dilution in PBS containing 1% foetal calf serum. After washing with PBS the cells were incubated with anti rabbit Ig conjugated with Alexa Fluor 488 (Molecular Probes, Eugene, OR, USA) at 1:1000 dilution for 30 minutes following by washing with PBS. DAPI was used to counterstain nuclei. Slides were viewed using an Axiovert 200 (Zeiss, Germany) with filter sets appropriate for FITC and DAPI. Photomicrography was achieved using a cooled CCD monochrome 12 bit camera Evolution QEi and Image-Pro 5.0 software (Media Cybernetics Inc., Canada) was used for preparing fluorescence composite images with pseudocolour. Adobe Photoshop version 5.0 LE was used to compose and present the figure collage. Competing interests =================== The author(s) declare no competing interests in relation to this work. Authors\' contributions ======================= PHT and WWJ performed the proteomics work, MJC and PHT prepared the virology reagents and the immunofluorescence. All authors contributed to the analysis and writing of the paper. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Spectra acquired with mass list submitted for searchImage of spectra obtained by MALDI-TOF for the spot from the 2D gel of βOG extracts identified as LAMR1. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 2 Spectra acquired with mass list submitted for searchImage of spectra obtained by MALDI-TOF for the spot from the 2D gel of NaDOC extracts identified as LAMR1. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 3 Spectra acquired with mass list submitted for searchImage of spectra obtained by MALDI-TOF for the spot identified as lamin B1. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 4 Spectra acquired with mass list submitted for searchImage of spectra obtained by MALDI-TOF for the spot identified as Hip 55. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 5 Spectra acquired with mass list submitted for searchImage of spectra obtained by MALDI-TOF for the spot identified as p47 protein. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ This work was funded in part by Venture Technologies Sdn Bhd and by Universiti Malaysia Sarawak.	PubMed Central	2024-06-05T03:55:55.652505	2005-3-25	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1079963/", "journal": "Virol J. 2005 Mar 25; 2:25", "authors": [ { "first": "Phaik Hooi", "last": "Tio" }, { "first": "Wan Wui", "last": "Jong" }, { "first": "Mary Jane", "last": "Cardosa" } ] }
PMC1079964	Background ========== The geriatric population is expanding, and hence the clinical decision making are often confused by effects of ageing. Age should not be the only parameter considered when addressing a medical problem \[[@B1]\]. There is much evidence that the booming elderly population with cancer does not receive potentially curative treatment afforded to younger cancer patient \[[@B2]\]. Recently there has been a rise in the number of articles published related to neoplasm, surgery and elderly \[[@B2]\]. Several series have proven more and more surgeons are auditing and publishing their experience in management of onco-geriatrics. This raising interest is summarised in Figure [1](#F1){ref-type="fig"}. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### This shows the number of Pubmed articles using the search terms; neoplasm, surgery, elderly ::: ![](1477-7819-3-17-1) ::: No significant difference in postoperative mortality and long term survival has been demonstrated, although a slightly higher morbidity between younger patients receiving surgical management was detected \[[@B2]\]. However, one might suspect a selection bias in these series. This highlights the need for prospective data collection, including co-morbidities and patient selection. Outcome of elective cancer operations in elderly could be improved by taking utmost care around peri-operative period. In this overview we raise the concern of epidemiological changes and population dynamics, revisit physiology of aged, look into currently available instruments to assess physiology and functional status of elderly oncological patients, trends in specific organ cancers and their current treatment. Projected epidemiological data of population changes in coming years, cancer prevalence in this subset of population, and current literatures regarding management of cancers in elderly, if any, is persuasive enough to demand for a change. The priority of this issue is to increase awareness among the medical community and stimulate a debate about the urge to update the overall management of elderly oncology patient differently. Complete management of cancer in this population, and their eventual outcome, could be improved by specialist onco-geriatric multidisciplinary team. Preparing medical community to deal with this impending epidemiological time bomb is discussed. Epidemiological outlook ----------------------- Demographic studies in industrialised countries have shown a considerable raise in the average life span and a progressive reduction in birth rates; as a result proportion of elderly patients is continuously increasing \[[@B3],[@B4]\]. In developed countries in general, and England and Wales in particular, more than 15% of population is aged 65 or over 65 years \[[@B3]\].The demographic changes in developing countries are leading to rapid increase in the absolute number of elderly population \[[@B5]\].Average life expectancy in developing countries has increased from 45 years in 1950 to 64 years by 1995 and is expected to reach 72 years by 2020 \[[@B5]\]. In most populous developing countries namely, China, India, and Brazil, people aged 60 years and over were between 5.9 to 6.8 % in 1970 and are expected to reach 11 to 15.5% by 2020 \[[@B5]\]. The risk of developing cancer increases with ageing \[[@B6]\]. The average age of survival at birth was 40 years at the end of 19^th^century and has now doubled to 81 years for females and 76 years for males and projected to be 86.4 years for males and 92.3 years for females by 2050 \[[@B7]\]. In census projection, the high life expectancy series project a US population of 416 million by 2050. In this projection 1% would be over 100 years age (4 million), 7.2 % would be over 85 (30 million), and 23.3% would be over 65 (97 million) \[[@B7]\], there by accounting for 31.3% of population over 65 years. Currently, average age of survival of a 70 years old woman is 15 years. A 65 years male is expected to have an active life expectancy of 12 years and 65 years female 14.09 years \[[@B8]\]. This expansion in life expectancy coupled with increased incidence of cancer is having a profound effect on the prevalence of cancer. The lifetime probability of developing an invasive cancer is almost 45% in men and 38% in women \[[@B9]\]. The rate of death from cancer has raised from 17.7% in 1973 to 23% in 1999, while cardiac disease related death has declined from 40 % in 1973 to 30% in 1999 \[[@B9]\]. Very soon cancer will become the leading cause of death and more than half of new solid cancers cases occur in \>70 years \[[@B9]\]. According to SEER data the prevalence of cancer is 207.4 cases/100,000 in \<65 years old subjects and 2163.9/100,000 in \>65 years in US \[[@B10]\]. As the incidence of cancer increases coupled with improved diagnostic certainty and life expectancy more doctors will be faced with caring for elderly patients with cancer. Despite this epidemiological \"time bomb\" \[[@B11]\], there is concern that the scientific community has not been able to develop a significant amount of evidence-based knowledge. Only a small sub-setting of geriatric patients are being entered into clinical trials \[[@B12]-[@B16]\] thus elderly patients are still being managed on the basis of assumptions based on a younger population group. Physiological changes in elderly -------------------------------- The unstated fear of exposing the frail elderly patient to increased toxicities, unacceptable morbidities and high mortality rate can only be minimised and appraised by improving our insight onto the physiological aspects of geriatrics. Ageing is defined as passage of chronological time whereas senescence is defined as the deteriorative changes with the time during post-maturation life, i.e. passage of biological time \[[@B17]\]. Time of onset is affected by multiple factors like diet, race, sex, physical activity, habits, hormonal effect, etc \[[@B17]\]. The hallmark of senescence is decreased functional reserve of individual organs and reduced ability of these organs to cope with the challenge. The progressive functional inadequacy of physiological systems is variable from species to species and individual to individual within the species \[[@B17]\]. It has significant impact in the peri-operative management of cancer patients, as well as the tolerance to oncological treatments, i.e. chemotherapy, radiotherapy, major curative surgery \[[@B18],[@B37]\]. The organ-specific functional deterioration undoubtedly plays a significant role in the peri-operative management. ### Cardiovascular system Cardiac output is a product of heart rate and ejection fraction. Ejection fraction is affected by myocardial contractility and end diastolic filling. Healthy older individuals fail to increase heart rate to the same extent as younger individuals at exercise \[[@B18]\]. Basal cardiac output is unchanged with ageing \[[@B18],[@B19]\]. Ability to increase cardiac output with ageing is more dependent on ventricular dilatation i.e., preload \[[@B19]\]. Aged myocardium has lowered sensitivity to beta adrenergic modulation, physiologically manifesting as lower heart rate, and lower cardiac dilatation at end diastole and end systole \[[@B18],[@B19]\]. Altered pattern of Calcium regulation allows the older heart myocardium to generate force for a longer time following excitation, hence prolongs of systolic phase of cardiac cycle \[[@B20]\]. This in turn reduces the early diastolic filling rate by half \[[@B20]\]. Incomplete emptying of ventricle at end systole, hence reduction in ejection fraction is the prominent characteristics of old heart \[[@B20]\]. Reduced distensability, superimposed upon stressed heart could impair coronary perfusion and hence lead to cardiac ischemia \[[@B20],[@B21]\]. Perioperative fluid depletion in surgical patients in general and elderly in particular is not uncommon \[[@B20]\]. This depletion of intra-vascular volume during peri-operative period, while undergoing physiological stress is not well tolerated by elderly \[[@B22],[@B23]\]. Combined effect of depletion of intra-vascular volume, impaired response to catecholamine and increased myocardium relaxation time adversely affect the functioning of elderly patient under stress. Cardiovascular disease increases with ageing. Cardiac arrhythmia and conduction abnormalities increase with ageing. Over half of all postoperative deaths in elderly and 11% of postoperative complications are as a result of impaired cardiac function under physiological stress \[[@B23],[@B24]\]. Inexpensive objective tool i.e. supine bicycle exercise test has shown to be very useful for risk stratification for both pulmonary and cardiac complications prior to major abdominal and non cardiac surgery in ≥ 65 years \[[@B25]\]. Identification of this high-risk group is an aim of Preoperative Assessment of Cancer in Elderly(PACE) study and their optimisation prior to surgery is the primary aim of preoperative evaluation. ### Respiratory Changes affecting respiratory functions are as a result of anatomical-physiological changes affecting chest wall, respiratory musculature and lung parenchyma and vasculature \[[@B26],[@B27]\]. With ageing maximal voluntary ventilation, Forced Expiratory Volume 1 (FEV 1), Vital Capacity(VC), decreases \[[@B26]\]. Reductions in blood PO~2~Levels, decreased responsiveness to changes in blood gas levels and impaired airway protective reflexes are noted \[[@B26]\]. The large reserves and capacities of respiratory system allow for significant erosion in function with ageing with minimal impact on normal breathing \[[@B27]\]. However during increased demand e.g., exercise that age associated changes have significant impact on \[[@B26]\]. Reduced lung elastic recoil with increasing chest wall stiffness results in decrease in the compliance and increased functional residual volume \[[@B28],[@B29]\]. With ageing impaired ciliary function in air passage and general host defence mechanisms compounded by changes in the mechanics of breathing increases the risk of atelectasis and postoperative pulmonary infections \[[@B30]\]. Literature quotes 9 to 18% of elderly patients suffer from respiratory complications during postoperative period after major non-cardiac and thoracic surgery \[[@B25],[@B31]\]. Aim of preoperative assessment is to identify this high risk and optimise them prior to major surgery. ### Body composition Changes in the body weight total body water, body fat distribution and muscle mass may affect response drug therapy in general and anaesthetic drug in particular. Proportion of body fat increases with age and is accompanied by concomitant decrease in the total body water and skeletal muscle mass \[[@B32]\]. As the total body water is reduced there is a reduction in the volume of distribution of water soluble drugs. \[[@B1],[@B32]\]. This in association with decreased renal clearance account for higher plasma levels of water soluble non-depolarising muscle relaxants in the aged than in younger patients \[[@B33],[@B34]\]. Similarly increase in the total body fat affect the distribution of lipid soluble drugs and could prolong the effect \[[@B35]\]. ### Fluid, electrolyte and renal physiology Disorders of fluid and electrolyte balance are very common in elderly. The homeostatic reserve decreases with ageing. By the age of 70 years half of the original nephron complement may have been lost \[[@B36]\]. Loss of renal cortical mass reflects in decline of renal function, decrease in the glomerular filtration and tubular function and hence inability to concentrate urine \[[@B36]\]. Serum creatinine may remain stable masking underlying progressive loss of renal function \[[@B37]\]. Philip et al, in 1984 demonstrated decreased thirst sensation in elderly \[[@B37]\]. The renal responsiveness to vasopressin is impaired, similarly renal function and renin-aldosterone decreased \[[@B37]\]. Elderly people under metabolic stress become dehydrated and acidotic. This compounded by reduced plasma flow and impaired handling of nephrotoxic drugs renders them susceptible for acute renal failure \[[@B38]\]. This emphasise the need to identify high-risk elderly individuals during preoperative and Peri-operative time to deliver optimal treatment and to assist them in smooth recovery from surgery. ### Liver functions An elderly patient in general starts with a 20% decrease in plasma albumin concentration. Peri-operative nutritional needs of elderly cancer patient deserve special attention for above reasons. The process of detoxification, conjugation, and extraction of different compounds seems to be remarkably preserved under normal conditions despite actual decrease in the size of the organ. However under hypermetabolic states, the liver may fail in increasing its synthetic and metabolic functions \[[@B39],[@B40]\]. ### Skin and wound healing The changes seen in aged skin are a combination of effects from intrinsic and extrinsic factor \[[@B41]\]. Characteristic changes seen with ageing like dermo-epidermal atrophy, dryness, roughness, sagging, and wrinkling, have implications for wound healing \[[@B42]\]. Wound healing in elderly presents a major clinical and economic problem \[[@B42]\]. Evidence of age-related wound healing has been derived from most past empirical observation \[[@B42]\]. The clinical impact of change in the tensile strength of wounds, rate of wound closure, and accumulation of wound healing in acute wound healing appears to be small. The consensus is that effect of ageing on wound repair is primarily a temporal delay and not an actual impairment in the wound healing. Though there is a temporal delay in healing in aged it is not any qualitatively different to that of younger \[[@B42]\]. Poor healing in chronic wounds more often related to comorbid conditions than age alone. Function Assessment ------------------- Ability to withstand the stress of various forms of treatment for cancer in elderly patient is dependent on the functional reserve and ability to build an appropriate response to the stress. A large number of instruments have been developed over the years by geriatricians, although some of them are time consuming and impractical in our busy clinical setting. Amongst the others, we found the following validated instruments particularly useful in predicting outcomes in a prospective oncogeriatric series receiving chemotherapy (table [1](#T1){ref-type="table"}) \[[@B43]\]) ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Validated instruments in elderly chemotherapy patients ::: ---------------------------------------------------- Mini Mental Score (MMS) \[44\] Activities of Daily Living (ADL) \[45\] Instrumental Activities Daily Living (IADL) \[46\] Geriatric Depression Scale (GDS) \[47\] Brief Fatigue Inventory (BFI) \[48\] ---------------------------------------------------- ::: No scoring method relating to candidacy for surgery has ever been attempted specifically on oncogeriatric population. Surgical risk in elderly ------------------------ Surgeon is often called to make a decision to operate or not to operate on a patient and it is one of the most important decisions that he has to make in surgical practice. This assumes greater importance when dealing with a patient who is a poor surgical candidate. In current day practice there is a high demand for individualised risk assessment to be shared between the surgeon and patient. Risk prediction should be appreciated and disclosed to the patient at the time of consenting. Chronological age is an unreliable predictor of performance of health in an individual. The largest numbers of abnormal laboratory tests are not capable of predicting post-treatment and operative adverse outcome \[[@B49]\]. Over half a century ago Welch reported in his large series of abdominal operation in-patients over 70 years of age, a peri-operative mortality of 20.7%. He concluded that surgery itself is safe but that aged required greater attention in peri-operative management \[[@B50]\]. Present evidence suggests that health of extreme elderly is improving and interventions can be successful at late ages. The Postoperative mortality, morbidity and long term survival after cancer surgery for solid tumours are not affected by chronological age on its own \[[@B51]-[@B53]\]. Currently ongoing multinational study Pre-operative Assessmentof Cancer in Elderly (PACE)is trying to provide a scoring method to assess candidacy for surgery of oncogeriatric populations \[[@B54]\] Widely known American Society of anaesthesiologists (ASA) \[[@B55]\] physical status system is not aimed at measuring operative risk, rather it assess globally the degree of sickness or physical state prior to anaesthesia and surgery. ASA is insensitive to differentiate largest proportion of patients in ASA II and III \[[@B56]\]. Surgeons to assess cardiac risk in non- cardiac surgeries rarely ever use Goldman cardiac risk index (CRI)\[[@B57]\]. Acute Physiology And Chronic Health Evaluation (APACHE)is probably the best known of the physiological scoring systems based on 34 physiological variables taken in the first 24 hours of patient\'s admission. APACHE II using 12 physiological variables is well suited for intensive care unit patients needing ventilatory support \[[@B58]\]. In general surgical patients\' not needing respiratory support its use is limited. Portsmouth modification of Physiological and Operative Severity Score for EnUmaration of Mortality and Morbidity (P-POSSUM)\[[@B59]\] a modification of POSSUM and POSSUM consists of physiological score and operative score. Latter component of POSSUM score is highly relevant to the final outcome; unfortunately necessity of per-operative variables compromises its usefulness as preoperative assessment tool. A comprehensive geriatric assessment (CGA) based on the previous parenthesis might improve our understanding of the surgical risk, allowing a more accurate comparison of surgical series, a careful patient selection, and adequate consenting. CGA has proven to be useful in predicting mortality and morbidity in several clinical settings including hospital geriatric evaluation, inpatient geriatric consultation, home assessment service, hospital home assessment service and outpatient assessment service and in a number of chronic diseases \[[@B60]-[@B63]\]. It is not a mere list of associated medical conditions that have impact on prognosis but it is actually complexity of information that can be gathered through CGA. CGA adds substantial information to the functional assessment of elderly cancer patients routinely collected through performance status (PS)index \[[@B60],[@B6]\]. Preoperative assessment of cancer in elderly (PACE) --------------------------------------------------- An international project has been launched aimed at defining the general health condition of oncogeriatric surgical candidates. PACE is a tool designed to assess the functional activities of geriatric patient there by making an attempt to assess the functional life of an oncogeriatric patient and there by predicting the individualised risk of cancer surgery. Patients age 70 years constitute 90% of the study subjects and present signs of ageing. Patients ≥ 70 years undergoing moderate, major and major + elective surgery whose mini mental score is 18 and above and are able to give written informed consent have been included in this trial. It aims at predicting the probable outcome of cancer surgery treatment preoperatively in elderly. The tools incorporated in the PACE are detailed in table [2](#T2){ref-type="table"}. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Validated Instruments Used with PACE ::: --------------------------------------------------------------------------------------------------- Mini Mental State Examination Satariano\'s Modified Index of Comorbidities Activities of Daily Living Instrumental Activities of Daily Living Geriatric Depression Scale Brief Fatigue Inventory Eastern Co-operative Oncology Group Performance Status American Society of Anesthesiologists Physical Status Physiological and Operative Severity Score for EnUmeration of Mortality and Morbidity -- (POSSUM) Portsmouth POSSUM Modification --------------------------------------------------------------------------------------------------- ::: The surgical outcome is defined by 30 day mortality and 30 day morbidity as assessed by care delivery team. The following complications are included in 30 days morbidity checklist (table [3](#T3){ref-type="table"}). Our pilot study has proven PACE is feasible, inexpensive and well accepted by the patient (table [4](#T4){ref-type="table"}) \[[@B64]\]. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### 30 Days Morbidity check ::: Complications Absent Minor Major ------------------------------ ------------ ----------- ----------- Respiratory Cardiac failure Renal failure Generalized sepsis Stroke & Neurologic Haemorrhage & Bleeding Nutritional problems Other organ failure Wound Infection/Dehiscence/Fistula Thrombo-embolic Hepatic failure Urinary retention Anastomotic Failure Peripheral ischemia Endocrine failure Pressure Sores Analgesic problems Others ::: ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Pilot study results: Association of PACE with postoperative morbidity ::: Components Complications 24 patients No complications 48 patients P value ------------------------ ------------------------------- ---------------------------------- ------------- Median (IQR) Co-morbidities 3.2 (2--5) 3 (2--4) 0.506 MMS 29 (26--30) 27 (26--29) 0.070 GDS 4 (1--7) 2 (1--4) 0.056 BFI 3 (1.1--5.8) 9 (1.3--29.5) 0.216 Number of patients (%) PS = 0 11 (45.8) 41 (85.4) 0.002 ADL (independent) 4 (16.7) 32 (66.6) \<0.001 IADL (dependent) 14 (58.3) 38 (79.2) 0.207 ASA = 1 or 2 17 (70.8) 37 (77.1) 0.681 ::: Performance status was found to be significantly lower in patients who developed morbidity and lower activities of daily living was associated with higher postoperative complications. Recruitment to the study is currently ongoing and full results will be revealed on completion of this study \[[@B64]\]. Solid cancers and surgical management ------------------------------------- Surgery remains the treatment of choice for solid cancers. There is significant under representation of elderly in trials of treatment for lung, colorectal, breast and ovarian cancers \[[@B65]\]. In the following paragraphs, we discuss current trends in the management of common solid tumours encountered by general surgeon. Breast cancer ------------- Breast cancer in elderly women is a significant health problem. Elderly women have 6-fold higher incidence and 8-fold higher mortality rate compared with non-elderly women \[[@B66]\]. Current incidence rate of breast cancer remaining constant, it is projected that a 72% increase in the number of elderly women diagnosed with breast cancer in US by 2025 \[[@B66]\]. Fifty percent of breast cancers occur after the age of 65 years and 25% after the age of 75 years \[[@B67]\]. The number of women diagnosed with breast cancer is likely to increase due to demographic changes. Trends in the cancer mortality for all cancers in elderly have long been unfavourable. Since late 1990s total cancer mortality between ages 65 and 84 has been declining in the European Union (EU) \[[@B68]\]. Breast cancer mortality has declined over the last decade by 8% in US and by 3% in the EU \[[@B68]\]. Our understanding about breast cancer treatment in elderly is mainly based on retrospective and observational studies and on very few randomised clinical trials \[[@B69]\]. Surgery is the main stay of treatment for early breast cancer independent of age and was the usual therapy for all ages till 1970s \[[@B70]\]. Primary tamoxifen treatment was adopted enthusiastically in 1980s with publications of promising early results with tamoxifen in patients over 70 years \[[@B71]\], 81% of elderly women treated with primary tamoxifen appear to develop progressive disease after 12 years of follow up as against 38% with mastectomy alone \[[@B71]\]. The drawback of tamoxifen only treatment was the short duration of response \[[@B72]\]. Patients who relapse face the prospect of second line hormonal treatment or surgery or radiotherapy \[[@B72]\]. A change in treatment plan in favour of surgery is welcome in view of expected active life of 14 years for a 65 years old lady. In general, surgery appears to be well tolerated despite patient age \[[@B71]\]. Large proportions of elderly cancer patients are offered less than conventional treatment. Elderly breast cancer women are offered breast conservative surgery but are less likely to have axillary dissection, postoperative radiation and chemotherapy \[[@B73]-[@B75]\]. Local recurrence rates after conservative surgery without radiotherapy are reported between 3 to 47% \[[@B76]-[@B81]\]. Mortality rates after breast cancer surgery in elderly is \<1% \[[@B82]-[@B84]\]. Predominant morbidities are related to wound complications \[[@B79],[@B83]\]. Fentiman et al, in their multicentric randomised trail investigated quality of life (QOL), survival at 12 months and treatment preferences in elderly patients (≥70 years) with early breast cancer undergoing mastectomy or tumour excision plus tamoxifen \[[@B85]\]. Patient undergoing tumour excision and tamoxifen did not differ from those undergoing mastectomy in terms of fatigue, emotional functioning and fear of recurrence. Conservative breast surgery patients reported fewer arm problems and a borderline shift in the direction of benefit of body image (P = 0.06) \[[@B85]\]. QOL was better after conservative breast surgery and such treatment is to be individualised and to be preferred independent of age \[[@B85]\]. Axillary surgery plays a key role in management of breast cancer surgery either to achieve nodal disease control or to stage the tumour accurately therefore to decide on adjuvant therapy. Denying axillary surgery exposes the patient to increased risk of nodal disease recurrence, on the other hand full axillary clearance increases morbidity. Although the risk of increased morbidity secondary to full axillary surgery is not higher in elderly patient \[[@B86],[@B87]\], axillary surgery is less often undertaken in the elderly patient \[[@B88]\]. Confusion still prevails regarding optimum axillary surgery. To bring new light on this controversy Chetty et al., carried out a randomised clinical trial comparing regional control rate for sampling with axillary clearance. No significant difference was found. Unfortunately elderly patients were excluded from this study \[[@B89]\]. Morbidity related to axillary surgery in elderly was shown to be no greater than young. Further research is needed to guide us in choosing the appropriate axillary surgery for breast cancer in this age group. Primary endocrine treatment has been substituted for surgery based on results from 1980\'s experience when ER status was unknown \[[@B77]\], 100% disease progression has been noted in ER negative group \[[@B78],[@B90]-[@B92]\]. Fennessy concluded that tumour excision decreased the mortality rate in an unselected population of elderly women with operable breast cancer who were fit for the procedure \[[@B93]\]. With the current evidence primary endocrine therapy should be reserved for highly selected ER positive elderly breast cancer patient who are unfit to have surgery or respecting patient wishes. Several retrospective studies have reported age alone should not determine the type of breast reconstruction. Autogenous tissue reconstruction can be a safe successful alternative for women aged 65 year and above \[[@B94]\]. Further research is needed to answer outcome of breast reconstruction after mastectomy for cancer in elderly. The definition of upper age limit for breast cancer screening is a very complex issue. Screening benefit depend on life expectancy. People with \< 5--10 years life expectancy are unlikely to benefit from screening, so it is worth while considering the variability of different ages. Medical system has the ethical obligation to properly inform the population invited for screening. Further research should be encouraged to answer the benefit of mass population breast screening in elderly. The ancient assumption that breast cancer was not worth resecting in the aged population is finally rejected in view of a minute morbidity and mortality rate and obviously improved cancer outcomes for patients who received surgery. It is reassuring to know how the performance of the largest part of breast surgeons is not significantly biased by an ageist mentality. There is a need for research targeting specific needs of elderly patients with breast cancer and develop a specific treatment guideline in this group. Recent survey among breast surgeons in the UK has demonstrated the need of a standardised pre-operative assessment capable of characterising the patient\'s \"functional age\", in order to optimise treatment planning and stratify outcomes on the basis of factors other than chronological age \[[@B96]\]. Melanoma -------- Melanoma is a significant public health problem. Incidence and mortality from cutaneous malignant melanoma continue to rise \[[@B96]\]. Melanoma is often diagnosed late in elderly due to multiple reasons. Many retrospective studies have predicted prognosis of melanoma as age independent \[[@B96]\]. Greater numbers of thick lesions are increasingly seen with rising age \[[@B97]\]. On the contrary, Cohen et al, in 1987 predicted age as an independent poor prognostic factor for death due to melanoma \[[@B98]\]. Literature search did not yield any prospective randomised trial to address effect of age on melanoma. With nose and ear lesions increased \[[@B98]\], the percentage of elderly patients with metastatic disease at initial diagnosis did not vary compared to younger individual \[[@B98]\]. Histotype lentigo maligna lesions were seen with increased frequency in elderly \[[@B97]\]. Chang et al., in their retrospective evaluation of intermediate thickness and T4 (≥ 4 mm) melanoma lesions in ≥ 65 years noted lymph node status is the most important prognostic factor influencing overall survival (OS)and disease free survival (DFS) \[[@B96],[@B99]\]. Surgery plays pivotal role in treating melanoma. The size of resectional safe margins is still being debated in young and elderly alike, but there seems to be no data suggesting a differentiated approach. Results of retrospective study elderly patients with positive nodal status, who received no adjuvant treatment, did significantly worse than historical control \[[@B96],[@B99]\] They concluded treatment for melanoma in ≥ 65 years should be aggressive and should not be denied adjuvant therapy based on age alone \[[@B96],[@B99]\]. There is no consensus regarding optimal nodal surgery for melanoma in elderly. Prospective controlled trial is needed to provide answer to optimum nodal surgery for cutaneous melanoma in elderly. Older patients are assumed to have higher risk of complications from isolated limb perfusion (ILP). Nooda et al., did not find significant difference in complete response rate, loco-regional relapse, limb toxicity, systemic toxicity, local complications and long term morbidity of ILP between \< 75 years and ≥ 75 years \[[@B100]\]. They concluded older age is not a contraindication for ILP. It is prudent to treat melanoma in elderly with same radical approach as in younger patient without age bias \[[@B100]\]. The predictive value of screening diagnosis of melanoma was more than twice as high for men ≥ 50 years with either a changing mole or skin types I and II compared with other participants \[[@B100]\]. The yield of mass screening for melanoma could be improved by outreach to middle aged and older men \[[@B101]\]. Impact of formal assessment of targeted screening warrants a further study. Lung cancer ----------- Lung cancer is the leading cause of cancer-related deaths in population aged over 70 years in the western world \[[@B102]\]. Over half of the people diagnosed with lung cancer are over 65 years old. Compiled data from 33 countries in four continents shows increase of 180 % to 580% in the mortality amongst 65 -- 84 years old males and females with lung cancer respectively from 1955 to 1992 \[[@B103]\]. 25 to 40% of all small cell lung cancer \[[@B103]\] and 40% of non-small cell lung cancer (NSCLC) \[[@B105]\] are in ≥ 70 year. Lung cancer mortality rates have declined over last decade by 8.5% in the EU and by 0.9% in USA \[[@B105]\]. Elderly cancer patients are less likely to enrol (1.3%) in co-operative group than younger patients (3.0%) \[[@B106]\]. Surgery offers the best potential for cure in patients with carcinoma of the lung, as is the case with most solid tumours. Patients with untreated or palliated early stage NSCLC have an average life expectancy of only 1.5 years \[[@B107]\], while individuals in the ninth decade of life have a 50% chance of living an additional 5 to 9 more years \[[@B108]\]. In case of resectable primary lung neoplasms, surgery remains the treatment of choice independent of age, as is the case for most solid tumours. No significant difference in survival or cancer related survival after lobotomy Vs limited resection has been noted between elderly and young \[[@B109]\]. Many clinicians avoid surgery or minimise surgical procedure on the basis of age but recent advances in the preoperative risk assessment and surgical and anaesthetic techniques have resulted in a significant decrease in operative mortality and morbidity \[[@B110]\]. Age is a recognised risk factor for death after thoracotomy in elderly patients with lung cancer. Lung sparing procedures such as segmentectomy and wedge resection are being increasingly performed for lung cancer especially in elderly patients \[[@B111]-[@B113]\]. Recent advances in Video Assisted Thoracic Surgery (VATS)techniques \[[@B114]\], voice controlled robotic lung resections \[[@B115]\], provides an alternative approach to standard thoracotomy in elderly lung cancer patients resulting in decreased recovery time and fewer postoperative complications. In a multi institutional trial of patients with stage I NSCLC undergoing lobectomy by muscle sparing thoracotomy or VATS confirmed that the latter approach decreases the incidence of postoperative complications \[[@B116]\]. Long-term survival after a VATS lobectomy for NSCLC has been reported to be comparable to that achieved by open thoracotomy \[[@B117]\]. VATS lobectomy has been proven to be feasible and relatively safe alternative in-patient with poor cardiopulmonary status for early lung cancer \[[@B118]\]. The treatment of stage III NSCLC is still a matter of debate since the efficacy of surgery decreases and operative mortality increases with stage of disease. The reluctance to offering surgery for the elderly is particularly evident in these advanced stages that require more extensive resections. Combined modality treatments offer an improved outcome for patients with stage III lung cancer \[[@B119]\]. Unfortunately, elderly patients have been under represented in these trials. More studies are warranted in order to define if these conclusions can be extended to the elderly population as well. The multi-disciplinary approach to lung cancer acquires importance when treating elderly patients. The close participation between pulmonologists, oncologists, thoracic surgeons, anaesthesiologists, cardiologists, geriatricians, primary care physicians, physical therapists, and nutritionists on the pre- and postoperative course of the elderly with lung cancer can improve measurable outcomes and decrease their frustrations, therefore improving their quality of life \[[@B118]\]. Oesophageal cancer ------------------ Oesophageal cancer is typically disease of aged man. Estimated annual incidence is 7.7/100,000 in European community inhabitants. In recent years, the number of elderly patients with oesophageal cancer has steadily been increasing. An incidence of 14.5% of all oesophageal cancer was seen in years 70 \[[@B120]\]. Our current limited knowledge regarding management of oesophageal cancer in elderly comes from several retrospective series. Peracchia et al., reported a consecutive series of 1338 oesophageal cancer patients of whom 18% were 70 and above, overall hospital mortality was 6% and five years actuarial survival rate was 30%. This is comparable to the survival of younger patients \[[@B121]\]. The diagnosis is often delayed in elderly \[[@B122]\]. Hence, they present with complications like malnutrition and aspiration pneumonia. Some physicians believe that aggressive surgical approach is imprudent because of alleged higher rates of mortality and morbidity and lower rates of survival than younger patient is. Ellis et al., while reviewing their 27 years experience from January 1970 to Jan 1997, noted a total of 505 patients had surgery for cancer of oesophagus, 29% of these were ≥ 70 years. Actuarial 5-year survival rates were 24.1% in ≥ 70 years as against 22.4% in younger patients. The in-hospital mortality though was higher in elderly group but was not statistically significant \[[@B123]\]. They concluded age should not be a limiting factor in using aggressive surgical approach for management of cancer of oesophagus or cardia in patients aged 70 years or older. Similarly Thomas et al., compared between \<70 years and ≥ 70 years, operative mortality (10.7% Vs 11.2%), Major morbidity (10.7% Vs 13.6%), pulmonary complications (17.9% Vs 20.6%), and 5-year survival rate (17% Vs 18.9%) and concluded that oesophagectomy can be safely performed in septuagenarian patients \[[@B124]\]. No differences in surgical resection rates were noted between patients younger or older than 70 years \[[@B125]\]. Elderly patients (\>80) had an overall higher respiratory and cardiovascular complications than younger \< 70 years. The operative mortality was 4.7% for \< 70 years as against 5.6% for over 80 years. 5-year survival inclusive of operative mortality was similar 25.1% in over 80 as against 19.8% in under 70 years \[[@B125]\]. They concluded that survival benefit is similar to that in younger age groups enforcing the view that oesophagectomy can be safely offered in specialist units with acceptable long-term survival in all ages groups. Long-term survival in oesophageal cancer is related to stage of tumour at presentation \[[@B126]\] (stage I = 73% Vs stage III = 6%). There is an urgent need for targeted research and prospective trial to understand management of oesophageal cancer in elderly. This could be achieved by active recruitment of elderly oesophageal cancer patients into clinical trials. Gastric cancer -------------- Cancer of the stomach is still a common cancer in Europe with an annual rate of 35.7/100.000. Its incidence peaks around the age of 70 and is uncommon before the age of 40 years \[[@B127]\]. In elderly, gastric cancer is more likely to affect the distal part of the stomach \[[@B127]\] and hence there is a higher prevalence of stenosis and anaemia. Elderly patients suffer more from a well-differentiated tumour and frequently of intestinal type \[[@B128],[@B129]\]. There is some evidence that young patients with gastric cancer have worse prognosis and have a higher prevalence of diffuse type \[[@B130]\]. Surgery is the treatment of choice for gastric cancer independent of age. The three-year survival rates among octogenarians having gastric cancer surgery for advanced and early gastric group were superior to no surgery group \[[@B131]\]. The early and long-term outcomes in elderly are comparable to that of younger patient. Age alone should not preclude gastric resection in elderly gastric cancer patients \[[@B132]\]. Due to prevalence of respiratory and cardiovascular co-morbidity in elderly they are often placed at higher ASA classes. Maehara claims that the operative mortality has been significantly reduced in recent years. Reporting on 344 patients who had surgery for gastric cancer, he shows a drop to 0% in most recently operated patients \[[@B133]\]. The ratio of partial to total gastrectomy has declined to 60% Vs 40% while the ratio of R0-R1 Vs R2 -- R3 has increased from 45% Vs 65% to 25% Vs 75%. Though the morbidity is slightly higher than younger age it is not statistically significant. The five-year survival does not seem to be affected by patient\'s age \[[@B133]\] Current practice of gastric cancer treatment in elderly is based upon results from that of younger patient and from retrospective series. The long-term cancer related prognosis of elderly gastric cancer patients does not differ significantly from that of younger patients, elective surgery being feasible with standard morbidity. Colorectal cancer ----------------- Colorectal cancer (CRC) is the second most common cancer in US, primarily a disease of elderly (138,000 new cases / year) \[[@B134],[@B135]\] More than half of deaths from colon and rectal cancer occur in 70 years and above. Its incidence increases with the increasing age. Based on the census projection colon cancer related admissions in US will increase from 192,000 to 448,000 by 2050 in people aged ≥ 60 years \[[@B136]\]. Median ages at diagnosis ranges between 70.5 years for colon and 68.4 for rectal cancer \[[@B136]\]. Association of ColoProctology of Great Britain and Ireland (ACPGBI)in its recently published booklets has confirmed high prevalence of cancer in elderly \[[@B137]\]. It has been shown that patients in the age group of 65--74 were 1.8 times more likely to die following surgery compared with 3.5 times for 75 to 84 years and 5 times for over 85 years. These odds ratios were not, however, adjusted for all the other risk factors (ASA, site, and stage) and should be interpreted with some caution. Right-sided carcinoma shows an increased frequency in the older patient \[[@B1]\]. Advanced colorectal tumours are common in elderly explaining the greater proportion of palliative surgery in this group \[[@B138]\]. Obstructive tumours are significantly more common in patients over 70 years of age. Elderly patients with CRC are still presenting as surgical emergencies (obstruction and/or perforation) in up to 40% of cases. Higher reported incidence of palliative surgeries being performed, and a lower overall utilisation of neo-adjuvant preoperative and adjuvant postoperative therapies; influences long-term cancer related outcome \[[@B139],[@B140]\]. Elderly rectal cancer patients continued to be denied surgery i.e., 11% in ≥ 70 years Vs 1% \< 70 years \[[@B141]\]. There is paucity of data regarding adjuvant treatment of older patients with rectal cancer and hence there is an urgent need to enrol more elderly rectal cancer patients in to trial to fully evaluate the outcomes of cancer therapy in this subpopulation \[[@B142]\]. First line of treatment for colorectal cancer is surgery. Total mesorectal excision (TME) has become new standard of operative treatment for rectal cancer replacing conventional receptions \[[@B143]\]. Quality of life and functional results after low anterior resections and TME in the elderly are of no worse than in younger patient \[[@B144]\], 85% of patients over 75 years, who had sphincter saving rectal surgery denied any significant problem with bowel function or continence \[[@B145]\]. Liver resection for colorectal cancer liver metastases in properly selected elderly patients ≥ 70 years is feasible and age alone is not a contraindication \[[@B146]\]. Clinico-economical evaluation of elective colorectal cancer in the aged independently analysed cost of preoperative investigations, operative real cost postoperative real costs and median total charges. The economic burden when delivering radical surgery to the aged colorectal cancer and younger ones were shown to be statistically no significant \[[@B147]\]. Five-year cancer specific survival disease free rates were not different between young and elderly. Postoperative morbidity and mean hospital stay were not different between young and elderly having colorectal surgery \[[@B138]\]. Evidence available today fails to support the practice of denial of curative colorectal cancer surgery in elderly under elective situation. There is very little rationale for substandard treatment delivery in colorectal cancer surgery, as the long-term cancer specific survival rates do not differ according to patients\' age under elective conditions \[[@B1],[@B146]\]. Minimal invasive surgery like laparoscopic assisted colectomy is emerging as alternative to open surgery for colonic cancer. Early reports shows that it is a viable option and some elderly patients with CRC have been treated with as part of these protocols \[[@B147]\]. The safety and efficacy of laparoscopy assisted colectomy remains unclear, awaiting the final results of the Clinical Outcomes of Surgical therapy (COST)trial designed to examine whether it is an effective alternative to open colectomy in the prevention of recurrence and cancer mortality \[[@B148]\]. The emergency surgery is associated with a significantly higher incidence of operative mortality at any age (15% on emergency Vs 5% Elective surgery \[[@B1],[@B137]\]. No significant difference in mortality is recorded between two age groups i.e. \<70 years and \> 70 years when only elective operations are considered i.e. 4% and 7.4 % respectively. Trans-anal excision of low rectal cancer in selected patients is an acceptable alternative to formal resection. The recent development of trans-anal endoscopic microsurgery (TEMS)has permitted removal of tumours from the upper rectum. Important selection criteria include early T stage, good or moderate differentiation, relatively small tumour size and negative microscopic margins \[[@B150]\] are factors important in curative resections, although criteria for deliberately palliative endorectal resections may be relaxed in selected elderly patients. Local recurrence and survival rates seem comparable to TME in early rectal cancer where TEMS is used with curative intent \[[@B151]-[@B153]\]. Conclusion ========== Surgeons will have to deal with increasing cancers in elderly. Current treatment practices in elderly are based on experience gained from retrospective series and reviews. An attempt to transfer the results of younger patients to this heterogeneous population should be discouraged. Better understanding of our knowledge about this unique age group could be achieved by encouraging active participation in clinical trials and education of medical community. The very finely balanced physiological resources in elderly cancer patients, demands extreme care around pre- and peri-operative. The cost of treatment of cancer in elderly is no expensive than those in younger patient. The enthusiasm to offer cancer treatment to selected oncogeriatric patients in par with young, a practice prevalent among minority of medical community should hence forth not only include oncogeriatric community without selection bias but also spread to wider medical community members. A currently ongoing multinational trial (PACE) is aimed at overcoming this selection bias and assist treating physician to make informed decision about optimum treatment in discussion with patient. Oncogeriatric patients would benefit from complete care with active participation of multi-disciplinary team comprising of Surgeons with special interest in geriatric cancer surgery, Geriatricians, Geriatric oncologists, Anaesthetist with interest in geriatric anaesthesiology and social worker. Further research into tailored treatment of elderly cancer patient from the time of preoperative evaluation to and including optimum surgery for individual cancers and adjuvant therapy is needed. A look at the future -------------------- Improvement in care of oncogeriatric subgroup comes with better understanding of elderly cancer patients by recruiting them into specifically designed clinical trials. Changing the attitude of treating surgeons towards the elderly with cancer comes with widening our knowledge. The benefit of technological advancement should percolate to elderly to improve their quality of life aside long-term survival. The benefit of advancements in science like early detection of cancers, improvement in anaesthetic techniques, and surgical techniques, should be extended based on individual merits independent of age. It is the responsibility of today\'s doctors to train future generations to offer treatment without age bias but on individual merits. This could be achieved by bringing gerontology in par with paediatrics to class room teaching. Proactive participation of practising doctors and up coming surgeons in ongoing national and international educational and scientific meetings should be encouraged. Conflict of interest ==================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= • Hodigere Sripathy Jois Ramesh • Study management -- literature review/update & preparation of manuscript. • Daniel Pope • Statistical analysis & study design • Roberto Gennari • Study material contribution & intellectual input • RAA • Study design & management Funding Source ============== • None Acknowledgements ================ • None	PubMed Central	2024-06-05T03:55:55.655534	2005-3-23	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1079964/", "journal": "World J Surg Oncol. 2005 Mar 23; 3:17", "authors": [ { "first": "Hodigere Sripathy Jois", "last": "Ramesh" }, { "first": "Daniel", "last": "Pope" }, { "first": "Roberto", "last": "Gennari" }, { "first": "Riccardo A", "last": "Audisio" } ] }
PMC1079975	Nanotechnology can be harnessed to address some of the world\'s most critical development problems. However, to our knowledge, there has been no systematic prioritization of applications of nanotechnology targeted toward these challenges faced by the 5 billion people living in the developing world. In this article, we aim to convey three key messages. First, we show that developing countries are already harnessing nanotechnology to address some of their most pressing needs. Second, we identify and rank the ten applications of nanotechnology most likely to benefit developing countries, and demonstrate that these applications can contribute to the attainment of the United Nations Millennium Development Goals (MDGs). Third, we propose a way for the international community to accelerate the use of these top nanotechnologies by less industrialized countries to meet critical sustainable development challenges. Developing Countries Innovate in Nanotechnology {#s2} =============================================== Several developing countries have launched nanotechnology initiatives in order to strengthen their capacity and sustain economic growth \[[@pmed-0020097-b1]\]. India\'s Department of Science and Technology will invest \$20 million over the next five years (2004--2009) for their Nanomaterials Science and Technology Initiative \[[@pmed-0020097-b2]\]. Panacea Biotec (<http://www.panacea-biotec.com/products/products.htm>) (New Delhi, India) is conducting novel drug delivery research using mucoadhesive nanoparticles, and Dabur Research Foundation (Ghaziabad, India) is participating in Phase-1 clinical trials of nanoparticle delivery of the anti-cancer drug paclitaxel \[[@pmed-0020097-b3]\]. The number of nanotechnology patent applications from China ranks third in the world behind the United States and Japan \[[@pmed-0020097-b4]\]. In Brazil, the projected budget for nanoscience during the 2004--2007 period is about \$25 million, and three institutes, four networks, and approximately 300 scientists are working in nanotechnology \[[@pmed-0020097-b5]\]. The South African Nanotechnology Initiative (<http://www.sani.org.za>) is a national network of academic researchers involved in areas such as nanophase catalysts, nanofiltration, nanowires, nanotubes, and quantum dots ([Figure 1](#pmed-0020097-g001){ref-type="fig"}). Other developing countries, such as Thailand, the Philippines, Chile, Argentina, and Mexico, are also pursuing nanotechnology \[[@pmed-0020097-b1]\]. ::: {#pmed-0020097-g001 .fig} Figure 1 ::: {.caption} ###### Quantum Dots for Disease Diagnostics Quantum dots may be used for cheap, efficient handheld diagnostic devices available at point-of-care institutions in developing countries. ::: ![](pmed.0020097.g001) ::: Science and technology alone are not the answer to sustainable development challenges. Like any other science and technology waves, nanoscience and nanotechnology are not "silver bullets" that will magically solve all the problems of developing countries; the social context of these countries must always be considered. Nevertheless, science and technology are a critical component of development \[[@pmed-0020097-b6]\]. The 2001 Human Development Report \[[@pmed-0020097-b7]\] of the UN Development Program clearly illustrates the important roles of science and technology in reducing mortality rates and improving life expectancy in the period 1960--1990, but it did not emphasize nanotechnology specifically. In a report released in early 2005 \[[@pmed-0020097-b8]\], the UN Task Force on Science, Technology and Innovation (part of the process designed to assist UN agencies in achieving the UN MDGs) addresses the potential of nanotechnology for sustainable development. Top Ten Nanotechnologies Contributing to the MDGs {#s3} ================================================= In order to provide a systematic approach with which to address sustainable development issues in the developing world, we have identified and ranked the ten applications of nanotechnology most likely to benefit developing countries. We used a modified Delphi Method, as described in our Top Ten Biotechnologies report \[[@pmed-0020097-b9]\] to identify and prioritize the applications and to achieve consensus among the panelists. We recruited an international panel of 85 experts in nanotechnology who could provide the informed judgments that this study required, of which 63 completed the project ([Table S1](#st001){ref-type="supplementary-material"}). We selected the panelists based on contacts identified in our previous study on nanotechnology in developing countries \[[@pmed-0020097-b1]\]. A conscious effort was made to balance the panel with respect to gender, specialty areas within nanotechnology, and geographic distribution. Of the panelists, 38 (60%) were from developing countries and 25 (40%) from developed countries; 51 panelists (81%) were male and 12 (19%) were female. We posed the following open-ended question: "Which do you think are the nanotechnologies most likely to benefit developing countries in the areas of water, agriculture, nutrition, health, energy, and the environment in the next 10 years?" These areas were identified in the 2002 UN Johannesburg Summit on Sustainable Development \[[@pmed-0020097-b10]\]. We asked the panelists to answer this question using the following criteria derived from our previous Top Ten Biotechnologies study. Impact. How much difference will the technology make in improving water, agriculture, nutrition, health, energy, and the environment in developing countries? Burden. Will it address the most pressing needs? Appropriateness. Will it be affordable, robust, and adjustable to settings in developing countries, and will it be socially, culturally, and politically acceptable? Feasibility. Can it realistically be developed and deployed in a time frame of ten years? Knowledge gap. Does the technology advance quality of life by creating new knowledge? Indirect benefits. Does it address issues such as capacity building and income generation that have indirect, positive effects on developing countries? Three Delphi rounds were conducted using e-mail messages, faxes, and phone calls. In the first round, the panelists proposed examples of nanotechnologies in response to our study question. We analyzed and organized their answers according to common themes and generated a list of twenty distinct nanotechnology applications. This list was reviewed for face and content validity by two nanotechnologists external to the panel. In the second Delphi round, the panelists ranked their top ten choices from the 20 applications provided and gave reasons for their choices. To analyze the data, we produced a summative point score for each application, ranked the list, and summarized the panelists\' reasons. Then we redistributed the top 13 applications, instead of the top ten, to generate a greater number of choices for increased accuracy in the last round. Thus, the highest score possible for an application was 819 (63 × 13). The final Delphi round was devoted to consolidating consensus by re-ranking the top ten of the 13 choices obtained in the previous round and to gathering concrete examples of each application from the panelists. Our results, shown in [Table 1](#pmed-0020097-t001){ref-type="table"}, were compiled from January to July 2004. They display a high degree of consensus with regard to the top four applications: all of the panelists cited at least one of the top four applications in their personal top four rankings, with the majority citing at least three. ::: {#pmed-0020097-t001 .table-wrap} Table 1 ::: {.caption} ###### Correlation between the Top Ten Applications of Nanotechnology for Developing Countries and the UN Millennium Development Goals ::: ![](pmed.0020097.t001) ^a^ The maximum total score an application could receive was 819 ::: To further assess the impact of nanotechnology on sustainable development, we have compared the top ten applications with the UN Millennium Development Goals ([Table 1](#pmed-0020097-t001){ref-type="table"} and [Figure 2](#pmed-0020097-g002){ref-type="fig"}). The MDGs are eight goals that aim to promote human development and encourage social and economic sustainability \[[@pmed-0020097-b11]\]. In 2000, all 189 member states of the UN committed to achieve the MDGs by 2015. The MDGs are: (i) Eradicate extreme poverty and hunger; (ii) Achieve universal primary education; (iii) Promote gender equality and empower women; (iv) Reduce child mortality; (v) Improve maternal health; (vi) Combat HIV/AIDS, malaria, and other diseases; (vii) Ensure environmental sustainability; and (viii) Develop a global partnership for development. As shown in [Table 1](#pmed-0020097-t001){ref-type="table"} and [Figure 2](#pmed-0020097-g002){ref-type="fig"}, the top ten nanotechnology applications can contribute to achieving the UN MDGs. ::: {#pmed-0020097-g002 .fig} Figure 2 ::: {.caption} ###### Comparison between the Millennium Development Goals and the Nanotechnologies Most Likely to Benefit Developing Countries in the 2004--2014 Period ::: ![](pmed.0020097.g002) ::: Addressing Global Challenges Using Nanotechnology {#s4} ================================================= What can the international community do to support the application of nanotechnology in developing countries? In 2002, the National Institutes of Health (NIH) conceptualized a roadmap for medical research to identify major opportunities and gaps in biomedical investigations. Nanomedicine is one of the areas of implementation that has been outlined to address this concern. Several of the applications of nanotechnology that we have identified in our study can aid the NIH in this process by targeting the areas of research that need to be addressed in order to combat some of the serious medical issues facing the developing world. To expand on this idea, we propose an initiative, called "Addressing Global Challenges Using Nanotechnology," to accelerate the use of nanotechnology to address critical sustainable development challenges. We model this proposal on the Foundation for the NIH/Bill and Melinda Gates Foundation\'s Grand Challenges in Global Health \[[@pmed-0020097-b12]\], which itself was based on Hilbert\'s Grand Challenges in Mathematics. A grand challenge is meant to direct investigators to seek a specific scientific or technological breakthrough that would overcome one or more bottlenecks in an imagined path to solving a significant development problem (or preferably, several) \[[@pmed-0020097-b12]\]. A scientific board similar to the one created for the Grand Challenges in Global Health, with strong representation of developing countries, will need to be established to provide guidance and oversee the program. The top ten nanotechnology applications identified in [Table 1](#pmed-0020097-t001){ref-type="table"} are a good starting point for defining the grand challenges. The funding to address global challenges using nanotechnology could come from various sources, including national and international foundations, and from collaboration among nanotechnology initiatives in industrialized and developing countries. These funds could be significantly increased if industrialized nations adopted the target set in February 2004 by Paul Martin, Prime Minister of Canada: that 5% of Canada\'s research and development investment be used to address developing world challenges \[[@pmed-0020097-b13]\]. In parallel to the allocation of public funds, policies should provide incentives for the private sector to direct a portion of their research and development toward funding our initiative. The UN Commission on Private Sector and Development report Unleashing Entrepreneurship: Making Business Work for the Poor \[[@pmed-0020097-b14]\] underscores the importance of partnerships with the private sector, especially the domestic private sectors in developing countries, in working to achieve the MDGs. Perhaps most importantly, our results can provide guidance to the developing countries themselves to help target their growing initiatives in nanotechnology \[[@pmed-0020097-b15]\]. The goal is to use nanotechnology responsibly \[[@pmed-0020097-b16]\] to generate real benefits for the 5 billion people in the developing world. Supporting Information {#s5} ====================== Table S1 ::: {.caption} ###### List of Panel Members (43 KB DOC). ::: ::: {.caption} ###### Click here for additional data file. ::: A Definition of Nanotechnology ------------------------------ Nanotechnology is the study, design, creation, synthesis, manipulation, and application of functional materials, devices, and systems through control of matter at the nanometer scale (1--100 nanometers, one nanometer being equal to 1 × 10^−9^ of a meter), that is, at the atomic and molecular levels, and the exploitation of novel phenomena and properties of matter at that scale. We are grateful to our panelists for providing their expertise, and to W.C.W. Chan and A. Shik for help with our analysis of the nanotechnologies. Grant support was provided by the Canadian Program on Genomics and Global Health (supported by the Ontario Research and Development Challenge Fund, and by Genome Canada through the Ontario Genomics Institute (Toronto, Canada); matching partners can be found at [www.geneticsethics.net](www.geneticsethics.net)). EBC is supported by the Ontario Genomics Institute; DKM is supported by a Career Scientist award from the Ontario Ministry of Health and Long-Term Care; ASD is supported by the McLaughlin Centre for Molecular Medicine; PAS is supported by a Distinguished Investigator award from the Canadian Institutes of Health Research. The University of Toronto Joint Centre for Bioethics (Toronto, Canada) is a PAHO/ WHO Collaborating Center for Bioethics. Citation: Salamanca-Buentello F, Persad DL, Court EB, Martin DK, Daar AS, et al. (2005) Nanotechnology and the developing world. PLoS Med 2(4): e97. NIH : National Institutes of Health MDGs : Millennium Development Goals [^1]: All authors are at the University of Toronto Joint Centre for Bioethics (Toronto, Canada) and the Canadian Program on Genomics and Global Health (Toronto, Canada). Douglas K. Martin is also at the Department of Health Policy, Management and Evaluation, University of Toronto. Abdallah S. Daar is also at the Department of Public Health Sciences and Surgery, University of Toronto, and the McLaughlin Centre for Molecular Medicine (Toronto, Canada). Peter A. Singer is also at Department of Medicine, University of Toronto and University Health Network. [^2]: Competing Interests: Peter A. Singer is on the editorial board of PLoS Medicine.	PubMed Central	2024-06-05T03:55:55.659954	2005-5-12	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1079975/", "journal": "PLoS Med. 2005 May 12; 2(5):e97", "authors": [ { "first": "Fabio", "last": "Salamanca-Buentello" }, { "first": "Deepa L", "last": "Persad" }, { "first": "Erin B", "last": "Court" }, { "first": "Douglas K", "last": "Martin" }, { "first": "Abdallah S", "last": "Daar" }, { "first": "Peter A", "last": "Singer" } ] }
PMC1080125	Background ========== Herpes Simplex virus thymidine kinase gene (HSV-tk) suicide gene therapy has lately been used in a variety of cancer models to sensitize replicating cells to the antiviral drug ganciclovir (GCV) \[[@B1]-[@B3]\]. In this paradigm, tumor cells transfected with the HSV-tkgene become able to mono-phosphorylate ganciclovir, a nucleotide analog. Ganciclovir monophosphate is subsequently bis and tri-phosphorylated by cellular kinases and is then incorporated in the DNA of replicating cells, blocking the cell cycle and inducing apoptosis \[[@B1],[@B4],[@B5]\]. Cells that are transfected and exposed to ganciclovir can also kill adjacent, untransfected cells by the so-called bystander effect. In most tumor types, the bystander effect relies on the transfer of phosphorylated ganciclovir molecules between cells via their gap junctions \[[@B6],[@B7]\], although other mechanisms have been described in some models \[[@B8]-[@B11]\]. Given their resistance to conventional treatments and their confinement to the brain, malignant gliomas have undergone a variety of human trials of HSV-tkgene therapy. The initial enthusiasm has however declined as a large, multicenter phase 3 clinical study failed to demonstrate any survival benefit for patients treated with HSV-TK gene therapy \[[@B12]\]. Efforts are now directed at improving the protocols, especially by modifying the vectors of gene distribution in tumor cells \[[@B13]\], the antiviral drugs \[[@B14]\] and the bystander effect \[[@B15]-[@B17]\]. A few papers have also pointed out that some concomitant treatments reduce the benefit of HSV-tk/ ganciclovir gene therapy \[[@B18]\]. In this view, and given the widespread use of corticosteroids in the symptomatic treatment of malignant gliomas patients, we have assessed the effect of dexamethasone on the bystander effect in this type of cancer. Methods ======= Cells and culture conditions ---------------------------- Rat C6 malignant glioma cells (ATCC \# CCL-107) were grown in Dulbecco\'s modification of Eagle\'s medium supplemented with 3% fetal bovine serum and penicillin (DEM-3%FBS). C6-TK cells were obtained as described previously \[[@B16]\] and maintained in DEM-3% FBS supplemented with 500 μM geneticin. Human U87 (ATCC \# HTB-14) and LN18 cells (a gift from Prof. N de Tribolet) were grown in RPMI 1640 medium supplemented with 10% fetal bovine serum and penicillin (RPMI-10%FBS). U87-TK and LN18-TK cells were obtained as for C6-TK cells \[[@B17]\] and kept in RPMI-10% FBS supplemented with 500 μM geneticin. Gap junction intercellular communication ---------------------------------------- Cells grown for 48 hours in the presence or absence of dexamethasone were loaded either with CMTMR, a gap-junction impermeant dye or with calcein, which diffuses freely through gap junctions. Cells were then mixed for 4 hours, and the percentage of CMTMR-tagged cells having incorporated calcein after this period -- a direct correlate of GJIC- was determined by FACS analysis of 10,000 cells \[[@B17]\]. For microinjection experiments, cells were grown on polyornithine-coated glass coverslips to confluence and then treated or not with dexamethasone (10^-6^M) for 24 hours. The coverslips were then placed in the culture chamber of a Zeiss^®^fluorescence microscope perfusion chamber of a dedicated Zeiss fluorescence microscope, bathed with EA01 buffer (NaCl 137 mM, KCl 5.7 mM, CaCl2 1.8 mM, D-Glucose 22.2 mM, Hepes 10 mM), and injected with a solution of Lucifer Yellow for 30 seconds as previously described \[[@B16]\]. The number of dye-colored cells was then counted one minute after the end of the injection. Western blots ------------- Western blots were performed as previously described \[[@B17]\]. Rabbit polyclonal antibodies to connexin 43 (Zymed, San Francisco, CA), CIAP-2, BCL~XL~and FAS/CD95 (Santa Cruz Biotechnologies, Heidelberg, Germany) were employed for antigen detection while monoclonal antibodies to tubulin (Santa Cruz) or actin (Roche, Mannheim, Germany) were used to assess loading homogeneity. Immunodetection was carried with HRP-coupled secondary antibodies to mouse (Sigma-Aldrich, Bornem, Belgium) or rabbit (Amersham, Uppsala, Sweden) antibodies and a chemoluminescent peroxidase substrate (Pierce, Rockford, IL). Assessment of the bystander effect ---------------------------------- Given proportions of TK^-^and TK^+^cells were seeded in 96-wells plates at a final density of 10,000 cells/well and grown for five days in the presence or absence of ganciclovir (2 -- 10 μM) and/or dexamethasone (1 μM). Media were replaced every 48 hours. Cell survival was then assessed with a 3(4, 5-dimethylthiazol-2-yl)-2, 5-diphenyltetrazolium (MTT) test as described previously \[[@B16]\]. With this test, the cellular viability in each well is directly correlated to its optical density measured at a wavelength of 650 nm (OD~650~). Results are expressed as the ratio of the OD~650~of ganciclovir-treated cells to that of untreated control wells (set as 100 %). Experiments were run 3 times in triplicate in the presence or the absence of dexamethasone, and results are given as mean ± SD for each condition. Sensitivity of TK^+^cells to GCV -------------------------------- C6-TK5 cells were seeded in duplicate in 24-wells plates at a density of 100,000/well and grown for 5 days in the presence of increasing concentrations of ganciclovir. They were then harvested and counted on a hemocytometer using the trypan blue exclusion test. Experiments were run three times. Results are shown as the percentage (± SD) of live cells with respect to live cells in control, untreated wells. For U87-TK and LN18-TK cells, the sensitivity of cells to GCV (10 μM) after 5 days in culture was assessed in the presence or absence of dexamethasone (1 μM) with a MTT test (see previous paragraph for details). Cell growth and survival ------------------------ C6, LN18 and U87 cells were grown for 5 days in the presence or absence of dexamethasone and cell viability was assessed with a MTT test. For thymidine incorporation experiments, C6, LN18 and U87 cells were grown in triplicate in 24-wells plates in culture medium supplemented ^3^H-thymidine (4 μCi/ml, Pharmacia-Amersham, Rosendaal, The Nederlands) for 48 hours. The incorporation was stopped after several PBS washes by digestion of the cells in 1 ml of 0.1 N NaOH, and ^3^H-thymidine activity was recorded with a Wallac^®^1400 scintillation counter. Results were normalized with respect to the protein content of each culture well as assessed with the method of Bradford \[[@B19]\]. Results are shown as the mean of 3 independent experiments ± SD. CD95-induced cell death ----------------------- 100,000 cells were seeded in triplicate in 96-wells culture plates and cultured for 24 hours in the presence of soluble FAS-Ligand (sFAS-L, 2 ng/ml, Alexis, Lausen, Switzerland), dexamethasone (1 μM), ganciclovir (10 μM) and of a combination of these drugs. Cell viability was then assessed with an MTT test. Experiments were run three times in triplicate. Results are expressed as the ratio of cell viability in treated wells to that in untreated wells, ± SD. Statistical analysis -------------------- Statistical analyses were performed as stated in the \'Results\' section with the Statview software version 5.0 (SAS Institute, Cary, NC). Results ======= Dexamethasone decreases the bystander effect in C6, LN18 and U87 cells ---------------------------------------------------------------------- C6 cells were mixed with 5 or 10 % of C6-TK cells, and treated with GCV (2 μM). In the absence of dexamethasone (DEX^-^), the overall cell viability after 5 days in culture was respectively 20.7 ± 9.3 % and 31.5 ± 10.4 % of that of controls. It increased to 41.7 ± 12.3 % and 63.8 ± 20.6 % respectively in the presence of 1 μM dexamethasone (DEX^+^, p \< 0.05, Student\'s t-test, Figure [1A](#F1){ref-type="fig"}). For LN18 cells grown in the presence of 20% LN18-TK cells and treated with 10 μM GCV, the overall survival of DEX^-^cells was 37.94 ± 6.37 % of that of controls but increased to 61.93 ± 12.19 % in DEX^+^wells (p \< 0.05, Student\'s t-test, Figure [1B](#F1){ref-type="fig"}). Similarly, the overall viability of a U87 cells mixed with 10% of U87-TK cells and exposed to 10 μM GCV increased from 45.76 ± 11.1% (DEX^-^) to 69.16 ± 5.51 % in DEX^+^conditions (p \< 0.05, Student\'s t-test, Figure [1C](#F1){ref-type="fig"}). Dexamethasone reduces the gap junction intercellular communication (GJIC) of glioma cells ----------------------------------------------------------------------------------------- In iontophoresis experiments, Lucifer Yellow dye diffused to 10.3 ± 10.6 C6 cells (n = 33) per injected cell in control conditions, and to 7.5 ± 6.5 cells (n = 13) in the presence of 10^-6^M of dexamethasone (NS tendency, Student\'s t-test, mean ± SD, Figure [2A](#F2){ref-type="fig"}). In order to assess the GJIC in much larger cell populations, we analyzed the diffusion of calcein dye in C6, LN18 and U87 cells by flow cytometry (FACS). This technique reproducibly demonstrated a reduction of GJIC following treatment with dexamethasone (1 μM). For example, when CMTMR^+^C6 cells were mixed with calcein-loaded, CMTMR^-^C6 cells, the percentage of CMTMR^+^cells that incorporated calcein after 4 hours in culture decreased from 15.25 ± 0.911 % in control conditions (DEX^-^) to 10.7 ± 1.875 % in the presence of 1 μM dexamethasone (n = 3, p = 0.0179, Student\'s t-test). Using a polyclonal antibody against rat connexin 43, we did not observe any modulation of the expression and phosphorylation of Cx43 in C6 cells treated with dexamethasone for 24 hours (Figure [2B](#F2){ref-type="fig"}). A similar inhibition of GJIC occurred in LN18 and U87 cells treated with 1 μM dexamethasone for 24 hours (Figure [2C](#F2){ref-type="fig"}). LN18 and U87 cells respectively exhibited a very low and high level of GJIC in control conditions (Figure [2C](#F2){ref-type="fig"}). Dexamethasone reduces the sensitivity of TK^+^cells to ganciclovir ------------------------------------------------------------------ The curve that represents the survival of C6-TK cells exposed to increasing concentrations of ganciclovir for 5 days was significantly shifted to the right in the presence of 1 μM dexamethasone (p \< 0.05, multiple ANOVA, Figure [3](#F3){ref-type="fig"}). Similarly, the viability of LN-TK and U87-TK cells treated with 10 μM GCV for 5 days increased 2.35-fold and 1.47-fold respectively in DEX^+^conditions as compared with DEX^-^conditions (n = 3, p \< 0.05, Student t-test). The viability of wild-type C6, U87 and LN18 cells remained unaffected by ganciclovir (data not shown). Dexamethasone alters glioma cells proliferation ----------------------------------------------- The viability of wild-type C6, U87 and LN18 cells after 5 days in culture (logarithmic growth phase) was reduced by respectively 46.75 ± 9.3 %, 33.1 ± 13.25 % and 20.6 ± 7.6 % in the presence of DEX (n = 3 for each condition, p \< 0.005 each, One sample t-test, Figure [4A](#F4){ref-type="fig"}). DEX also decreased the incorporation of ^3^H-thymidine in C6, U87 and LN18 cells by 21.9 ± 8.4 %, 30.7 ± 11 % and 28.5 ± 6.3% respectively (n= 3 for each condition, p \< 0.05 each, Student\'s t-test, Figure [4B](#F4){ref-type="fig"}). Dexamethasone alters CD95-triggered apoptosis in glioma cells ------------------------------------------------------------- GCV (10 μM) significantly sensitized LN-TK and C6-TK cells but not U87-TK cells to the toxic action of soluble FAS-L (sFAS-L, 2 ng/ml). This effect was abolished by DEX (Figure [5A](#F5){ref-type="fig"} and [5B](#F5){ref-type="fig"}, and data not shown). Dexamethasone (1 μM × 24 hours) did not alter the expression of FAS/CD95 and of the apoptosis inhibitor CIAP-2 in LN18, C6 and U87 cells with these cells, as evidenced by Western blot. The expression of the anti-apoptotic protein BCL~XL~however slightly but reproducibly increased in C6 and LN18 cells (Figure [5C](#F5){ref-type="fig"}). Discussion ========== In this study, we assessed the influence of dexamethasone on the bystander effect of HSV-TK/ ganciclovir gene therapy in gliomas. Dexamethasone is indeed commonly used for the symptomatic treatment of brain tumor patients, and the bystander effect is considered a major contributor to the efficacy of type of suicide gene therapy \[[@B20],[@B21]\]. We found that a concentration of 1 μM, i.e. within its therapeutic range \[[@B22]\], dexamethasone significantly inhibited the bystander effect on C6, LN18 and U87 glioma cells co-cultured with small proportions (5--20 %) of C6-TK, LN18-TK and U87-TK cells. This finding is important, since 10 % of TK-expressing cells represent the threshold at which C6 tumors can be cured by HSV-TK suicide gene therapy \[[@B2]\] and since usually less than 5 % of cells are transfected in vivowith currently available vectors \[[@B12],[@B13]\]. The in vitrobystander effect in glioma cells depends on the transfer of phosphorylated ganciclovir molecules trough gap junctions \[[@B23],[@B7]\]. Dexamethasone at 10^-6^M moderately but significantly reduced the GJIC of C6, U87 and LN18 cells, which is a know mechanism through which drugs decrease the bystander effect \[[@B16]\]. This moderate reduction of GJIC by DEX in cells that exhibit a low (LN18), medium (C6) and high (U87) level of GJIC \[[@B17]\] and originate from humans and a rodent, is consistent with previously published results \[[@B24]\]. This modulation of GJIC in C6 cells did not appear to result from a reduced expression or a differential phosphorylation of connexin 43, the major connexin protein of malignant astroglial cell \[[@B25],[@B26]\]. We cannot exclude that some Cx43 phosphorylation changes were not evidenced with the antibody we used. It is also possible that the effect of DEX results from a modification of cell adhesion molecules or other connexin-interacting proteins as has been suggested for other effectors \[[@B27],[@B28]\]. It also inhibited by 23--53 % the log-phase growth of all three types of glioma cells after 5 days in culture, as well as their incorporation of ^3^H-thymidine. Of note, dexamethasone has also been shown to decrease glioma size in humans in vivo, an effect that is well-known to clinicians but is usually transient and lasts for a few weeks only \[[@B29]\]. Since the incorporation of the nucleotide analogue triphosphoganciclovir in replicating DNA strands is required for its toxicity both in TK-expressing and bystander cells, this anti-proliferative effect of dexamethasone may contribute to its inhibition of the bystander effect. Such a mechanism has been described for another inhibitor of DNA replication, Ara-C, that reduces the bystander effect without altering the GJIC in C6 cells \[[@B16]\]. It is somewhat surprising that dexamethasone decreased both cell proliferation and GJIC in our experiments. A decrease of connexin expression and GJIC is indeed usually associated with an enhanced proliferation of tumor cells \[[@B30],[@B18]\]. We however did not observe any dexamethasone-induced alteration of Cx43 expression on our Western blots, and the enhancement of GJIC by itself is less likely to control the growth of tumor cells than is the expression of connexin proteins \[[@B31],[@B25]\]. Finally, glucocorticoids are also known to alter cell proliferation via GJIC-independent mechanisms such as cell cycle blockade \[[@B32]\] and alter the PI3K-Akt pathway \[[@B33]\]. GCV treatment has been shown to induce apotosis in TK-expressing and bystander cells, an effect that notably implies both ligand-dependent and independent CD95 receptor activation \[[@B34]-[@B36]\]. We found that ganciclovir indeed sensitized C6-TK and LN18-TK cells to the lethal action of soluble sFAS-L, and this effect was abolished by dexamethasone. This drug did neither reduce the expression of FAS/CD95 protein expression nor induce that of apoptosis inhibitor CIAP-2 in our hands. This latter result contrasts with the findings of Webster et al. who observed an induction of CIAP-2 by dexamethasone \[[@B37]\]. We however observed an enhanced expression of BCL~XL~, a known modulator of FAS-dependent apoptosis \[[@B38]\] by dexamethasone in our three glioma cell lines, as also described by Gorman et al. \[[@B39]\]. The reasons why ganciclovir failed to enhance the toxicity of sFAS-L on U87-TK cells remain unknown and are beyond the scope of the present work, as is the thorough assessment of the effects of dexamethasone on HSV-TK/GCV-induced apoptosis. Our results nevertheless demonstrate that dexamethasone reduces the bystander effect of HSV-TK/ ganciclovir gene therapy via at least three different mechanisms in rodent and human gliomas in vitro, i.e. a modulation of GJIC, thymidine incorporation and apoptotic pathways. The in vivosignificance of our results must also still be validated. Indeed, although the effects of dexamethasone on HSV-tkgene therapy have partially been studied in vivoon the 9L animal model of glioma \[[@B40],[@B41]\], none of these studies specifically addressed the bystander effect. In one of these studies however \[[@B40]\], dexamethasone tended to reduce the survival of animals inoculated with TK-expressing 9L cells and treated with ganciclovir. Despite a small number of animals in the GCV (n = 9) and in the GCV + dexamethasone (n = 10) treatment arms, and although most of the observations in the control group were censored, this difference almost reached significance (cited P= 0.057). In this study, only 10 % of the animals treated with dexamethasone + GCV remained alive 60 days after tumor implantation (termination of the observation period) versus 55 % of the animals treated with GCV alone, a result that is consistent with our in vitrodata. Given the fundamental role of the bystander effect in clinical situations \[[@B12]\], and since the ultimate evaluation of the effects of dexamethasone on the bystander effect will eventually depend on human clinical trials, the design of upcoming protocols should carefully address this potential drug interaction. Conclusion ========== The overall efficacy of HSV-TK gene therapy is adversely affected by dexamethasone co-treatment in vitrothrough a modulation of apoptosis and gap junction intercellular communication. Although additional in vivoresearch is necesseray to confirm this finding, future HSV-tk/ GCV gene therapy clinical protocols for gliomas should address this interference of corticosteroid treatment. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= PR is responsible for the generation of TK-expressing cells, microinjection experiments and apoptosis experiments and contributed to the Western blots, cell proliferation assays and evaluation of the bystander effect. He is also responsible with VB for the initiation of the project. MTN contributed to the Western blots, cell proliferation assays and evaluation of the bystander effect. OJ is responsible for the FACS experiments. BR contributed to the Western blots, cell proliferation assays and evaluation of the bystander effect and the design of the experiments. MPM contributed to the design of the experiments. All authors read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2407/5/32/prepub> Acknowledgements ================ We wish to thank P. Ernst-Gengoux and B. Coumans for their expert technical assistance, and Dr JM Mangin for his help with some iontophoresis experiments. P.A. Robe, M-P Merville and B. Rogister are respectively post-doctoral researcher, research associate and senior research associate at the Fonds National de Recherches Scientifiques (FNRS) of Belgium. This work was supported by grants from the Fondation Léon Frédéric, the Centre Anti-Cancéreux près l\'ULg, the University Hospital of Liege (FIRS grant \# 2722) and the FNRS of Belgium (grant \# R-0553). Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Effect of dexamethasone on the bystander effect. (A)The viability of C6 cells mixed with 10 % (left) or 5 % (right) of C6-TK cells and treated with ganciclovir (2 μM) for 5 days is shown in the presence (+) or absence (-) of dexamethasone (1 μM, DEX). (B)Viability of LN18 cells mixed with 20 % LN18-TK cells and grown for 5 days with ganciclovir (10 μM) in the presence (+) or absence (-) of DEX. (C)Viability of U87 cells mixed with 10 % U87-TK cells and grown for 5 days with ganciclovir (10 μM) in the presence (+) or absence (-) of DEX. For each condition, the viability in the absence of ganciclovir was set as 100 percents. (\: P\< 0.05, Student\'s t-test) ::: ![](1471-2407-5-32-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Effect of dexamethasone on gap junction intercellular communication. (A)Appearance of confluent cultures of C6 cells after iontophoresis of Lucifer Yellow in one single cell (\) in the absence (left) or presence (right) of dexamethasone (DEX). Photomicrographs are representative of the typical appearance of such injected cells. (B)FACS analysis of calcein dye transfer between rat C6 cells in the presence (right) or absence (left) of DEX. The GJIC is directly correlated with the percentage of CMTMR^+^/Calcein^+^cells (upper right quadrants, see text for details). This experiment is representative of 3 independent experiments each performed on 10,000 cells. A Western blot of Cx43 in C6 cells is shown (left: control conditions, right, DEX treatment. See text for details) (C)FACS analysis of calcein dye transfer between human U87 (above) and LN18 (below) cells in the presence (right) or absence (left) of DEX. The rightmost pictures illustrate the basal GJIC of both cell types in confluent cultures after iontophoresis of Lucifer Yellow in a single cell (\, control conditions). ::: ![](1471-2407-5-32-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Sensitivity of C6-TK cells to ganciclovir. The sensitivity of C6-TK cells is shown in control conditions (▲) and in response to treatment with 1 μM dexamethasone (■). Results are expressed as the percentage of living cells in treated conditions at various concentrations of ganciclovir with respect to ganciclovir-free cultures, and are shown as the mean of 3 independent experiments. ::: ![](1471-2407-5-32-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Cell growth and thymidine incorporation. (A)The total viability of LN18, U87 and C6 cells grown for 5 days with 1 μM dexamethasone (DEX) was determined by way of an MTT test. For each condition, the viability in the absence of DEX was set as 100 percents (n = 3, mean ± SD). (B)The incorporation of ^3^H-thymidine in LN18, U87 and C6 cells grown for 48 hours in the presence of DEX is shown (n = 3, mean ± SD). For each condition, the incorporation in the absence of DEX was set as 100 percents. (\: P\< 0.05; \\ P\< 0.005, Student\'s t-test). ::: ![](1471-2407-5-32-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Effect of dexamethasone on FAS/CD95-triggered apoptosis. (A)Viability of C6-TK cells after 24 hours of growth in the presence of GCV, sFAS-L, sFAS-L + GCV and DEX + sFAS-L + GCV, assessed with a MTT test. This viability was significantly altered in cells treated with sFAS-L and GCV and this sensitization was abolished by DEX. (B)Viability of LN18-TK cells after 24 hours of growth in the presence of GCV, sFAS-L, sFAS-L + GCV and DEX + sFAS-L + GCV. This viability was significantly altered in cells treated with sFAS-L and GCV and this effect was abolished by DEX. For both cell types, the viability in the absence of any drug was set as 100 percents. (\: P\< 0.05, ANOVA). (C)*Western blot analysis of the expression of CIAP-2, FAS/CD95 and BCL~XL~in C6, LN18 and U87 cells following treatment with DEX (1 μM). ::: ![](1471-2407-5-32-5) :::	PubMed Central	2024-06-05T03:55:55.661726	2005-4-2	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1080125/", "journal": "BMC Cancer. 2005 Apr 2; 5:32", "authors": [ { "first": "Pierre A", "last": "Robe" }, { "first": "Minh", "last": "Nguyen-Khac" }, { "first": "Olivier", "last": "Jolois" }, { "first": "Bernard", "last": "Rogister" }, { "first": "Marie-Paule", "last": "Merville" }, { "first": "Vincent", "last": "Bours" } ] }
PMC1080126	Background ========== Evolutionary conflicts of interest between the sexes have been convincingly demonstrated in species with separate sexes \[[@B1]\]. These sexual conflicts often give rise to traits that are advantageous for one sex but harmful to the other. If these detrimental effects are counteracted, a co-evolutionary arms race may ensue in which harmful traits and corresponding counter-adaptations arise repeatedly \[[@B2]\]. Such antagonistic interactions can bring about major changes in the mating behaviour, genital morphology, gametes and seminal products, potentially leading to speciation \[[@B2]\]. Similar arms races seem to occur in hermaphrodites, contrary to Darwin\'s conviction \[[@B3]\] that sexual selection cannot act in hermaphroditic organisms. In fact, theoretical modeling indicates that these processes can become even more extreme in hermaphroditic species (N.K. Michiels and J.M. Koene, unpublished data), mainly because within one mating simultaneous hermaphrodites gain paternity (male fitness) which can outweigh the loss in female fitness. Here we investigate the evolution of a most peculiar reproductive behaviour that occurs in simultaneously hermaphroditic land snails (Stylommatophora), the \"shooting\" of a so-called love-dart into the mating partner. Several explanations have been offered for the evolution of the enigmatic dart shooting behaviour. The dart is made of calcium carbonate and has therefore been proposed to serve as a nuptial gift of calcium for the production of eggs \[[@B4],[@B5]\]. However, in Cantareus aspersus(previously Helix aspersa) the dart does not contain enough calcium to significantly contribute to egg production and darts are only rarely incorporated by the recipient \[[@B6]\]. Likewise, in other investigated species darts are even retained by shooters to be reused on the next mate \[\[[@B7],[@B8]\], J.M. Koene and S. Chiba, unpublished data). Therefore two other hypotheses have been put forward. In the female choice hypothesis the dart represents a sexual signal and recipients select on dart shooting effectiveness \[[@B5],[@B9]\]. The important prediction of this hypothesis is that this can only be beneficial for the recipient if the dart is shot consistently by individuals (assuming that shooting ability is heritable). Tests in C. aspersusdo not support this, because dart shooting of individually-identified animals in consecutive matings is unpredictable (G-test: N= 29 snails, df= 1, G= 6.745, P\< 0.01; J.M. Koene, unpublished data). Besides, in Arianta arbustorumdart shooting seems to be an optional component of courtship \[[@B10]\]. In the last hypothesis, the dart is used to manipulate the mating partner and can thus cause a sexual conflict \[[@B11]\]. This latter hypothesis seems most consistent with findings in the common garden snail C. aspersus, the species in which dart shooting and copulatory behaviour has been extensively studied. During courtship, the stylophore (dart sac) is everted and the single calcareous dart is pierced into the partner. Both mating partners normally shoot a dart before their penises are simultaneously intromitted. During intromission, spermatophores are exchanged and transferred into the partner\'s receiving organ, either directly into the bursa copulatrix or in an associated diverticulum (Figure [1](#F1){ref-type="fig"}). To avoid digestion in the spermatophore receiving organ (SRO), sperm have to actively swim out via the spermatophore\'s tail (formed by the flagellum of the penis) into the vaginal duct to reach the sperm storage site (the spermathecae) \[[@B12]\]. Previous work on C. aspersushas demonstrated that the shooting of the love-dart serves to introduce an allohormone \[[@B13]-[@B15]\], produced by associated glands, into the blood of the partner \[[@B16]\]. This allohormone inhibits digestion of sperm \[[@B17]\], which results in more of the donated sperm reaching the spermathecae \[[@B18]\] and fertilizing eggs \[[@B19],[@B20]\]. This manipulation of the sperm storage process caused by darts can have a negative effect on the recipient\'s (reproductive) fitness because of interference with cryptic female choice. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Schematic morphological drawing of the reproductive morphology of a land snail with one dart and a diverticulum.The love-dart (D) is produced and stored in the stylophore (S, often called dart sac) and shot by a forceful eversion of this organ. The mucus glands (MG) produce the mucus that is deposited on the dart before shooting. The penis (P) is intromitted to transfer the spermatophore. The sperm container is formed in the epiphallus (EP), while the spermatophore\'s tail is formed by the flagellum (FL). When a bursa tract diverticulum (BTD) is present, the spermatophore is received in this organ. Together with the bursa tract (BT) and bursa copulatrix (BC) these form the spermatophore-receiving organ (SRO, indicated in grey), which digest sperm and spermatophores. Sperm swim out via the tail of the spermatophore to enter the female tract and reach the sperm storage organ (SP, spermathecae) within the fertilization pouch (FP)-spermathecal complex. Other abbreviations: AG, albumen gland; G, genital pore; HD, hermaphroditic duct; OT, ovotestis; PRM, penis retractor muscle; SO, spermoviduct; V, vaginal duct; VD, vas deferens. ::: ![](1471-2148-5-25-1) ::: Interestingly, love-darts display an astonishing diversity between species, both in number and shape, ranging from several simple cone-shapes to one elaborately bladed structures \[[@B21],[@B22]\]. The most elaborate darts show surface enlargement with blades that is likely to enhance the transfer of gland product. We therefore predict that dart elaboration should covary with allohormone production, which should be mirrored in the gland morphology as surface enlargement. Moreover, the dart specializations that enhance gland product transfer are potentially more successful at manipulating fertilization. If so, adaptations to counteract this effect are expected, which could give rise to a co-evolutionary arms race. These predictions are tested here using a comparative analysis of dart-possessing land snail species. Note that we are only focussing on the Helicoidea superfamily; we do not include species with non-homologous dart-like structures (see also \[[@B22]\]). Because the phylogeny of land snails is heavily based on reproductive morphology \[[@B23],[@B24]\], we first reconstructed an independent phylogeny based on part of the 28S ribosomal RNA (rRNA) gene \[[@B25]\]. Our findings represent the first comprehensive comparative analysis of reproductive organ characteristics in simultaneously hermaphroditic animals and are consistent with co-evolution and counter-adaptation predicted by sexual conflict theory. Results ======= Reproductive organ morphologies ------------------------------- Within the land snails that possess love-darts, there is a large diversity in reproductive structures and the darts themselves provide an impressive range of shapes. There are species with one dart, while others have several or none. Darts of some species have a simple cone-shape, whereas others show surface enlargement with blades. Additionally, darts vary from straight to curved and contorted. This variety in dart shapes is illustrated in the electron microscopic photographs in Figure [2](#F2){ref-type="fig"} and the line drawings in Figure [3](#F3){ref-type="fig"}. We found similar levels of variation between species in the other reproductive organs. These differences include the number, relative size and placement of both functional and vestigial stylophores; the number, relative size, type of branching, and placement of the glands; the presence, relative length and placement of the diverticulum in the SRO; and the presence and relative length of the flagellum. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Diversity of love-darts.The different shapes of love-darts are illustrated with electron microscopic photographs of side views and cross sections of darts from different species. Scale bars indicate 500 μm for side views and 50 μm for cross-sections. ::: ![](1471-2148-5-25-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Phylogeny of land snails and their love-darts.Cross-section and side views of the darts are shown. For comparability, the line drawings are all at the same size. When two cross-sections are shown, that species possesses two functional darts. The shown phylogeny was obtained by Bayesian inference (BI). Branch lengths correspond to the number of substitutions per site (see scale bar). Maximum likelihood (ML) produced an almost identical tree (see Results). Clade support is given next to the nodes such that values before slashes refer to BI posterior probabilities above 0.5 and values behind slashes to ML bootstrapping results above 50. ::: ![](1471-2148-5-25-3) ::: Phylogenetic analysis --------------------- Both ML and BI yielded a single optimal phylogenetic tree (see Figure [3](#F3){ref-type="fig"} for the BI tree). These two trees had almost identical topologies. The only differences consisted of the exact position of Fruticicola fruticumwithin the Bradybaenidae, of Leptaxis erubescenswithin the Hygromiidae, the relationship of the four major lineages of the Helicidae, and the position of Cernuella cisalpina, C. hydruntina, C. virgata, and Xerosecta cespitumin relation to each other. Moreover, although BI produced high support values for a much larger number of clades than ML bootstrapping, the clades with high ML bootstrap scores also always had high BI posterior probabilities (Figure [3](#F3){ref-type="fig"}). Furthermore, the four main families were all correctly identified (Figure [3](#F3){ref-type="fig"}). The only exception being that Polymita pictawas not grouped with the other species of the Helminthoglyptidae. However, this taxon is found at the end of a comparatively long branch. Hence, its position at the base of the superfamily Helicoidea could be due to long branch attraction to the outgroup and may therefore be unreliable. Both the SH test and BI posterior probabilities indicate a multiple origin of most reproductive organ characteristics (Table [1](#T1){ref-type="table"}). Namely, a single origin was always rejected by BI posterior probabilities except for the presence of the diverticulum when monophyly was assumed within each of the four main families (Helicidae, Bradybaenidae, Helminthoglyptidae, Hygromiidae). The SH test did not confirm a single origin of the number of blades on the darts, the number of stylophores and glands, and the shape of glands if monophyly was assumed across all land snails. However, when monophyly was only hypothesized to occur within the four main families, then the SH test only rejected it for the stylophore number and the shape of the glands. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Assessment of the monophyletic origin of reproductive organ characteristics. ::: H~0~ -ln L Δ P~rell~ P~BI~ ----------------------------------------- --------- -------- ----------- ----------- Monophyly assumed across all snails Optimal tree 3240.98 Dart number 3330.76 89.78 0.031 \<0.001\* Blade number 3538.66 297.68 \<0.001\* \<0.001\* Presence of perpendicular blades 3283.10 42.12 0.399 \<0.001\* Stylophore number 3450.29 209.31 \<0.001\* \<0.001\* Gland number 3492.69 251.71 \<0.001\* \<0.001\* Shape of glands 3705.34 464.36 \<0.001\* \<0.001\* Presence of flagellum 3326.69 85.71 0.044 \<0.001\* Presence of diverticulum 3245.56 4.58 0.905 0.002\* Monophyly assumed within families Optimal tree 3240.98 Dart number 3295.14 54.16 0.234 \<0.001\* Blade number 3351.62 110.64 0.012 \<0.001\* Presence of perpendicular blades 3283.11 42.13 0.399 \<0.001\* Stylophore number 3386.55 145.57 \<0.001\* \<0.001\* Gland number 3373.91 132.92 0.005 \<0.001\* Shape of glands 3455.91 214.93 \<0.001\* \<0.001\* Presence of flagellum 3269.58 28.60 0.574 \<0.001\* Presence of diverticulum 3240.98 0 0.962 0.843 The following ancestral states were assumed: Zero darts, no blades on the dart, no perpendicular blades on the blades, two functional and two vestigial stylophores, many tubular glands, a simple gland shape, and absence of the flagellum of the penis and diverticulum of the spermatophore-receiving organs. Abbreviations: -ln L, negative natural logarithmic likelihood value for the hypothesis, as calculated from a constrained tree using ML in PAUP\. Δ, difference from the optimal tree; P~rell~, probability according to the Shimodaira-Hasegawa test; P~BI~, Bayesian posterior probability; \, Significance after Bonferroni correction. ::: Principal component and correlation analysis -------------------------------------------- The overall PCA (Table [2](#T2){ref-type="table"}) that was performed on the BI dataset revealed that most PICs loaded positively and significantly on the first PC axis (at P\< 0.15 or P\< 0.05 (see \[[@B26]\])) and explained 32.5% of the total variance. This result suggests that the traits all covary. To examine this correlated evolutionary pattern in more detail we performed separate PCAs on the data for darts, stylophores, glands, and spermatophore receiving organs ([Additional file 1](#S1){ref-type="supplementary-material"}) followed by a correlation analysis. These PCAs permitted for successful reduction of the variable number per trait to two PCs. For each organ the first two PCs explained over 70% of the variation for each trait, regardless of the tree used for PIC calculation ([Additional file 1](#S1){ref-type="supplementary-material"}). The interpretation of the PCs can be deduced from the inferred eigenvectors ([Additional file 1](#S1){ref-type="supplementary-material"}) and is mentioned between brackets in the following description of the results from the correlation analyses. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Principal component analysis on phylogenetically independent contrasts of love-dart and reproductive morphology data. ::: Component loadings --------------------------------------------------- -------------------- -------- -------- -------- Love-dart Number of darts 0.807\* -0.212 -0.262 0.050 Number of blades 0.492† 0.544 0.303 0.318 Length of blades 0.664\* 0.381 0.517 0.158 Curvature 0.534\* 0.169 0.127 0.668 Functional stylophore(s) Number 0.781\* -0.204 -0.379 -0.219 Relative size 0.411† 0.498† -0.447 0.393 Placement on vaginal duct 0.536\* 0.053 -0.578 -0.286 Vestigial stylophore(s) Number 0.527† -0.775 0.125 0.013 Relative size 0.361† -0.655 0.258 0.180 Placement on vaginal duct 0.527† -0.753 0.165 0.106 Glands Number 0.709\* -0.339 -0.050 -0.225 Relative size 0.472 0.592 -0.275 -0.089 Type of branching 0.275 0.360 -0.277 0.189 Placement 0.871\* -0.133 -0.050 0.258 Diverticulum of spermatophore receiving organ Length relative to bursa tract 0.457 0.359 0.216 -0.470 Placement 0.554\* 0.308 0.431 -0.342 Relative size 0.184\* 0.454 0.303 -0.167 Flagellum Relative length 0.614\* 0.141 0.138 -0.429 The independent contrasts are based on the trees reconstructed with BI without consideration of five cases of phylogenetic uncertainty. The table shows the loadings for each variable in the different principal components (PC1 to PC4). Collectively the principal components explain over 70% of the total variance in the data set (respectively, 32.5%, 19.3%, 9.6%, and 9.0 %). Tests of significance of variable loadings represent frequency of loadings different in sign to the ones observed, among 1000 bootstrap replicate analyses corrected for axis reversals (\, P \< 0.05; †, P \< 0.15). ::: The correlation analyses in Table [3](#T3){ref-type="table"} shows the relationships between the PCs for the different traits based on the BI tree without phylogenetic uncertainty. Significant correlations indicate co-evolution between organs. The correlations in Figure [4](#F4){ref-type="fig"} are also based on the BI tree without phylogenetic uncertainty, but essentially identical results were obtained for the PICs calculated from the other phylogenetic trees as well as the raw data ([Additional file 2](#S2){ref-type="supplementary-material"}). Additionally, we only show results for the PCAs based on all the variables; the results were essentially identical when only the significantly loading variables were included (data not shown). Most importantly, there are significantly positive correlations between Dart PC1 (dart shape) and Gland PC1 (gland complexity) (Figure [4A](#F4){ref-type="fig"}); Dart PC2 (dart number) and Stylo. PC1 (vestigial stylophores); Gland PC1 (gland complexity) and Stylo. PC1 as well as Stylo. PC2 (functional stylophores) (Figure [4B](#F4){ref-type="fig"}). Additionally, there is a significantly negative correlation between Gland PC2 (gland shape) and Stylo PC1 (vestigial stylophores). Furthermore, the SRO PC1 (SRO complexity) shows significant positive correlations with Dart PC1 (dart shape) and Gland PC1 (gland complexity) (Figure [4C](#F4){ref-type="fig"}). Finally, we also found that the length of the flagellum of the penis (which forms the spermatophore\'s tail) positively correlates with SRO PC1 (Figure [4D](#F4){ref-type="fig"}) as well as with Gland PC1 (gland complexity). ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Pearson correlations (below diagonal) and their significance (above diagonal) for the comparison of principal components (PC) and flagellum. ::: BI* ------------------ -------------- -------------- ---------------- ---------------- --------------- --------------- ------------- ------------- ------------------ Dart PC1 Dart PC2 Stylo. PC1 Stylo. PC2 Gland PC1 Gland PC2 SRO PC1 SRO PC2 Flag. length Dart PC1 1.0000 0.0711 0.0120 \<0.0001\* 0.7586 0.0002\* 0.9030 0.0136 Dart PC2 0.000 \<0.0001\* 0.0330 0.0063 0.1482 0.2417 0.3583 0.3122 Stylo. PC1 0.257 0.631 1.0000 \<0.0001\* 0.0006\* 0.6756 0.0942 0.0459 Stylo. PC2 0.353 0.302 0.000 \<0.0001\* 0.0609 0.0663 0.9023 0.0311 Gland PC1 0.675 0.381 0.532 0.592 1.0000 0.0024\* 0.1921 0.0002\* Gland PC2 0.045 -0.208 -0.468 0.267 0.000 0.2786 0.8307 0.8816 SRO PC1 0.498 -0.169 0.061 0.262 0.420 0.156 1.0000 \<0.0001\* SRO PC2 0.018 -0.133 -0.239 0.018 -0.188 -0.031 0.000 0.2290 Flag. length 0.347 0.146 0.284 0.305 0.502 -0.022 0.591 -0.173 Results are shown for the analysis based on PICs calculated from the BI without phylogenetic uncertainty. \, Significance after Bonferroni correction. ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Graphs illustrating co-evolution and counter-adaptation.*A, Correlated evolution of the darts and glands. B, Correlated evolution between stylophores and glands. C, Counter-adaptation of diverticulum in response to changes in the dart. D, Counter-adaptation of the spermatophore\'s tail (formed by the flagellum) in response to changes in SRO complexity. The principal components used to create the graphs are those obtained from the independent contrasts based on the BI tree without phylogenetic uncertainty. The ellipses represent the 95% confidence interval, Pearson correlations (R) and P-values are indicated in each graph. ::: ![](1471-2148-5-25-4) ::: One potential problem of this detailed analysis is that our character coding may produce a bias in the data when organs are absent, because then the organ itself as well as all related traits are scored as zero. However, this does not have a strong effect on the analysis of darts, stylophores, glands and flagellum, where there are only very few taxa in which the respective organ is absent. The diverticulum is not always present in the SRO, but in a correlation analysis of the data only including presence/absence of this organ (thus removing all irrelevant zero\'s) the same combinations were significant. Discussion ========== Sexual conflict can cause counter-adaptive co-evolution between male and female reproductive organs. Although such conflicts may become costly, when looking at only one species these costs usually remain hidden because the mating partners are well adapted to each other \[[@B27],[@B28]\]. Therefore, an inter-species comparison is required to reliably reveal patterns of counter-adaptive co-evolution driven by sexual conflict \[[@B29]\]. The results from our detailed comparative analysis are indeed most consistent with the presence of a co-evolutionary arms race in simultaneous hermaphrodites, which has generated a diversity of dart shapes and reproductive organ morphologies. This finding corroborates the theoretical prediction that sexual conflict and counter-adaptation can play a major role in the evolution of hermaphroditic mating systems and reproductive morphologies \[[@B30],[@B31]\]. In detail, our findings provide evidence for both repeated and correlated evolution between traits associated with dart shooting and spermatophore receipt. The occurrence of both is considered convincing evidence for a co-evolutionary arms race. The reconstructed phylogenies that we used to test this, were estimated from the 28S rRNA gene with different tree reconstruction methods (see Methods). In agreement with previous studies \[[@B32]-[@B34]\], ML bootstrap values for the inferred clades are generally smaller than the BI posterior probabilities. Hence, BI support for the inferred clades may be overestimated. Although BI is considered to provide a highly consistent framework for phylogenetic inference, its reliability under different evolutionary scenarios and the interpretation of clade posterior probabilities is still under debate \[[@B32]-[@B34]\]. Nevertheless, in our study ML and BI produced highly similar trees. In addition, the results are consistent with morphology-based trees, for example all families form monophyletic groups except for the Helminthoglyptidae \[[@B23],[@B24]\]. Hence, the consistency of results obtained with different methods and between molecular and morphology-based phylogenies suggests that the inferred phylogeny is generally robust. Even though small discrepancies existed between the inferred phylogenies, they did not seem to have a large effect on the subsequent analyses (PCA and correlation analysis). These analyses produced consistent results regardless of the underlying topology or the consideration of phylogenetic uncertainty (Table [3](#T3){ref-type="table"} and [Additional file 2](#S2){ref-type="supplementary-material"}). Furthermore, a comprehensive analysis of computer simulated data highlighted that in comparative analyses the consideration of phylogenies with a limited number of uncertainties still yields considerably more reliable results than no consideration of phylogenetic relationships \[[@B35]\]. Both the SH test and BI posterior probabilities reject a single origin of most reproductive structures when monophyly is assumed across all land snails, as was already suggested for the dart by Tompa \[[@B36]\]. When monophyly is assumed within each of the well-supported families (Helicidae, Bradybaenidae, Helminthoglyptidae, Hygromiidae) monophyly is still rejected in some cases but with less significance (Table [1](#T1){ref-type="table"}). Because BI tends to overestimate reliability of inferred clades (i.e. very high posterior probabilities), it automatically underestimates support for clades, which are not represented in the inferred tree. In this case, hypotheses of their monophyly would be often incorrectly rejected. This could be the reason for our finding that BI posterior probabilities for the specific hypotheses of monophyly were almost always highly significant, even when the SH test statistics were clearly insignificant. Nonetheless, taken together the results strongly indicate that we are dealing with repeated evolution of most characteristics of the reproductive system. Using phylogenetically independent contrasts (PICs), to account for phylogenetic affiliation of the taxa \[[@B37]\], we found clear evidence for correlated evolution. Namely, the overall principal component analysis demonstrates that correlated evolution across the PICs occurs on one co-evolutionary axis, meaning that the traits covary \[[@B26]\]. Examination of this pattern in more detail revealed that when darts become more elaborate the number of stylophores (and thus the number of darts, as previously suggested \[[@B21],[@B22]\]) decreases and the complexity of the glands increases (Figure [4A,B](#F4){ref-type="fig"}). These co-evolutionary patterns lead us to conclude that the changes enhance the transfer of gland product and possibly improve sperm storage, provided that the gland products of the investigated species have similar effects as those seen in C. aspersus\[[@B17],[@B19],[@B18],[@B20]\]. If inhibition of sperm digestion occurs, it will ultimately influence sperm storage and thus fertilization success \[[@B17],[@B19],[@B18],[@B20]\], which is beneficial for the shooter and hence favoured by sexual selection. However, the receiver\'s fertilization processes are manipulated by the mucus while the dart itself causes damage and may increase infection risk. Consequently, counter-measures can be expected. Behavioural counter-adaptations to prevent dart receipt seem unlikely because the tactile information from contact with the partner\'s skin, which cannot be avoided during mating, is essential for dart shooting to occur \[[@B38]\]. Conversely, physiological or morphological counter-adaptations are possible, e.g. changes in allohormone receptor sensitivity, skin thickness, or female reproductive morphology. Here, we report morphological counter-adaptations in the spermatophore receiving organs (Figure [4C](#F4){ref-type="fig"}). The adaptations primarily entail the appearance and subsequent lengthening of a diverticulum. These changes increase the distance sperm need to travel to the spermathecae, presumably this hampering of sperm storage occurs to offset the increased sperm survival caused by the more elaborate darts and glands. The latter change poses an additional complication because sperm are most successful at reaching the storage site when the spermatophore\'s tail is protruding into the vaginal duct \[[@B12]\]. The tail of the spermatophore is formed by the flagellum of the penis, which we used as a measure of the tail\'s length. The significant correlations of the flagellum (spermatophore\'s tail) with both the SRO and gland complexity can therefore also be interpreted as an indication of counter-adaptation. The presented study provides a first step into understanding the diversity of love-darts in land snails and clearly raises several questions for future studies. For example, the previously available experimental data strongly suggest that dart shooting serves to manipulate the partner (manipulation- or sexual conflict hypothesis), whereas alternative hypotheses are not supported (nuptial gift- and female choice hypothesis; see background section). Although the manipulation hypothesis is entirely consistent with the results of the current study, it implies that dart shooting benefits the shooter, as previously demonstrated \[[@B13]-[@B20]\], and that it also negatively affects the receiver, for which no direct evidence is available yet. Hence, an important challenge for the future is to evaluate the possible costs of receiving a dart. In this context, our results also allow to choose closely related species with pronounced differences in dart morphology (e.g. C. hortensisand C. nemoralis) for a more detailed experimental study addressing the causes and consequences of reproductive organ diversity. The examination of variation among closely related species pairs and/or within species may additionally provide novel insights into the dynamics of the co-evolutionary adaptations. Finally, evidence is accumulating that not all species use their dart in the same way. Hence, additional behavioural data are pivotal for a full understanding of the evolution of these darts. Conclusion ========== We found support for both repeated and correlated evolution, which we consider compelling evidence for a co-evolutionary arms race. Furthermore, because empirical findings are most consistent with the manipulation hypothesis, we conclude that the observed co-evolutionary patterns result from a sexual conflict. This comparative study is the first of its kind in simultaneous hermaphrodites. The results strongly suggest that sexually antagonistic co-evolutionary interactions can play an equally important role in hermaphrodites as they do in organisms with separate sexes \[[@B29],[@B39]\]. As such, sexually antagonistic co-evolution may provide an important driving force for the evolution of hermaphroditic mating systems and possibly even speciation. Moreover, it may also account for some of the other bizarre reproductive structures and behaviours found in hermaphrodites like gigantic penises \[[@B40]\] and penis biting \[[@B41]\] in land slugs, hypodermic insemination in tropical flatworms \[[@B42]\], and body piercing in earthworms \[[@B43]\]. Methods ======= Snails material --------------- We obtained data from 51 land snail species from the four main dart-possessing families: the Helicidae, the Hygromiidae, the Helminthoglyptidae, and the Bradybaenidae. We also included one member of both the Polygyridae and Camaenidae, plus three outgroup taxa. Mature specimens of the species were collected by JMK or provided by colleagues ([Additional file 3](#S3){ref-type="supplementary-material"}). Prior to fixation in 80% EtOH the animals were relaxed by drowning, which was also the standard protocol for specimens obtained from the malacological collection of the Academy of Natural Sciences of Philadelphia. In three cases, where we could not obtain snail material ourselves, we used previously published information on reproductive organ morphology \[[@B44]\] and DNA sequence data \[[@B25]\] ([Additional file 3](#S3){ref-type="supplementary-material"}). In 21 other cases, because previously published DNA sequence data were already available \[[@B25]\], we only obtained data on reproductive organ characteristics ([Additional file 1](#S1){ref-type="supplementary-material"}). Reproductive organ morphologies ------------------------------- We examined five different reproductive structures: darts, stylophores, glands, spermatophore receiving organs, and the flagellum of the penis (Figure [1](#F1){ref-type="fig"}). For this, adult specimens of each species were dissected to remove the reproductive tract. Subsequently, the reproductive organs were drawn using a camera lucida. These drawings were used to determine relative organ sizes, thus correcting for body size (see [Additional file 4](#S4){ref-type="supplementary-material"}). The albumen gland, which is seasonally variable because it provisions the eggs, was not included when determining relative sizes. Only adult animals from one location with fully formed darts were included. To avoid damage of the darts, the stylophores were carefully cut out of the reproductive tracts and placed overnight in 1N NaOH, which dissolved all the tissue and mucus but left the dart intact. For cross-sections, darts were carefully broken in two. The intact and broken darts were consecutively prepared for electron microscopy by placing them on small aluminium plates with an electrically conducting adhesive (Leit-Tab, Plano). They were then coated with gold using a Metalloplan (Leitz). The darts were placed under a scanning electron microscope (S-530 SEM, Hitachi) and photos were taken. The characteristics of the reproductive structures were scored as ranks in order of complexity, based on the traditional taxonomic literature (e.g. \[[@B24]\]) ([Additional file 4](#S4){ref-type="supplementary-material"}). Molecular data -------------- The phylogeny of the snails was examined using an analysis of the 5\' end of the 28S rRNA gene. DNA was isolated from snails using a CTAB-based protocol \[[@B45]\]. In detail, snail tissue was ground up with a pestle in 400 μl CTAB buffer (2% (w/v) Cetyl-trimethyl-ammonium-bromid, 0.1 M Tris-HCL pH 8.0, 0.02 M EDTA, 1.4 M NaCl, 0.2% (v/v) β-Mercaptoethanol). Tissue was further digested by addition of 4 μl Proteinase K (10 mg/ml) and incubation at 50°C overnight. DNA was extracted by addition of 2 volumes chloroform:isoamylalcohol (24:1) and centrifugation at 13.000 rotations per minute (rpm) for 15 min. The DNA containing supernatant was recovered and DNA was precipitated by addition of 2/3 volumes of 100% isopropanol, incubation of the mixture at -20°C for 1 h, and subsequent centrifugation at 13.000 rpm for 30 min. The DNA pellet was finally washed in 70% ethanol, left to dry and resuspended in 50--100 μl sterile Millipore H~2~O. The 5\' end of the 28S rRNA gene was amplified via PCR, using primers designed in conserved regions of the ribosomal cistron of molluscs ([Additional file 5](#S5){ref-type="supplementary-material"}). Amplification was performed under standard reaction conditions: 1 U Taq Polymerase (Promega Ltd.), 50 mM KCl, 10 mM Tris-HCl pH 9.0, 0.1% Triton X-100, 2.5 mM MgCl~2~, 0.2 mM of each dNTP, 0.5 μM of each Primer. The following cycling profile was used: 5 min at 95°C, followed by 35 cycles of 20 sec at 95°C, 30 sec at 62.5°C and 1 min at 70°C, and a final extension period of 10 min at 70°C. PCR products were purified using Microcon-50 microconcentrators (Millipore Ltd). DNA sequencing was subsequently performed with the reverse PCR primer and additional internal primers ([Additional file 5](#S5){ref-type="supplementary-material"}), using the ABI Prism BigDye Terminator Cycle Sequencing Kit (Applied Biosystems Ltd) and visualization of results on an ABI310 Genetic Analyser (Applied Biosystems Ltd). All sequences are deposited at the EMBL database under accession numbers AJ550953 to AJ550982. Phylogenetic tree reconstruction -------------------------------- The DNA sequence alignment (EMBL: ALIGN\_000524) was produced with CLUSTALW, using default settings \[[@B46]\], and was subsequently adjusted by eye in cases of obvious errors (e.g. large end gaps) using the program BIOEDIT, version 5.0.9 \[[@B47]\]. The alignment contained 51 taxa and 740 positions, of which 169 were variable (22.84%). All phylogenetic analyses were based on either conventional maximum likelihood (ML) as implemented in the program PAUP\, version 4, beta 10 \[[@B48]\] or Bayesian inference (BI) as implemented in the program MRBAYES, version 2.01 \[[@B49]\]. Both methods permit specification of substitution models to correct for multiple hits and provide a consistent statistical framework for hypothesis testing \[[@B50],[@B51]\]. The Tamura-Nei substitution model with gamma-distributed rate heterogeneity across sites and a proportion of invariable sites (TN-Γ-I) was found to be optimal for the data with the help of likelihood ratio tests following the procedure outlined by Huelsenbeck and Crandall \[[@B52]\] and using the program MODELTEST, version 3.06 \[[@B53]\]. The TN-Γ-I model was employed for all subsequent ML analysis. Since the program MRBAYES currently does not support TN-Γ-I, BI was based on the next more complex model available, the general time reversible substitution model with rate heterogeneity across sites (GTR-Γ). For ML tree estimation, parameters of the substitution model were first optimized using a maximum parsimony (MP) tree, inferred with a heuristic search via branch-swapping by tree bisection and reconnection (TBR). These parameter estimates were then employed in the ML tree search, using the heuristic search options and the MP tree as a starting topology for branch-swapping by TBR \[for a similar approach, see \[[@B45]\]\]. Nodal support was inferred via non-parametric bootstrapping \[[@B45]\] using the same ML settings and 200 replicate data sets. Specific hypotheses on the origin of reproductive organ characteristics were assessed with the Shimodaira-Hasegawa (SH) test \[[@B55]\]. For these tests, trees were first calculated with specified topological constraints, which each represented one of the hypotheses of interest. Tree calculation was again based on ML and the above settings. Thereafter, trees with topological constraints were compared with the optimal topology using the SH test statistics, inferred with the RELL bootstrap option, as implemented in PAUP\ (see \[[@B56]\]). BI was based on the Markov Chain Monte Carlo approach, using 500,000 generations, four chains (one cold and three heated), each started with a random tree, and saving of trees every 10 generations. Stationarity was reached after 40,000 generations. Calculation of consensus trees and summary statistics was thus based on trees obtained from only subsequent generations. BI was performed twice, using the same settings, and produced qualitatively identical results (identical tree topology, highly similar branch lengths and nodal support values) (see \[[@B49]\]). The monophyletic origin of reproductive organ characteristics was evaluated with the help of the inferred posterior probability of the respective clades \[[@B57]\]. Independent contrast and principal component analysis ----------------------------------------------------- The taxa included are related to each other at different degrees. Hence, the data obtained for reproductive organ characteristics are not independent, resulting in increased type I errors in comparisons between taxa \[[@B37],[@B58],[@B59]\]. To correct for phylogenetic affinity, we calculated phylogenetically independent contrasts (PICs) using CONTRAST 3.6 \[[@B37]\]. We treated variables as continuous since the scores for each of them were ranked in order of complexity (for similar approaches see \[[@B60],[@B61]\]). PICs were inferred from four tree topologies: the BI and the ML tree, both either with or without consideration of five cases of phylogenetic uncertainty. These cases refer to differences between BI and ML trees, which also lack support from BI posterior probabilities and ML bootstrapping. Because these differences are characterized by short branches, phylogenetic uncertainty was taken into account by setting branches to zero. This essentially produces hard polytomies, which is equivalent to a very rapid radiation rate (see \[[@B59]\] for a similar approach). Adequate standardization of contrasts was confirmed by plotting the absolute values of contrasts against their standard deviations \[[@B62]\]. Note that we decided against the approach of Pagel \[[@B63]\], as implemented in the program CAIC \[[@B64]\], to take account of phylogenetic uncertainty. This method relies on the a-priorispecification of a predictor variable, which is not applicable for our data set. To assess the main pattern of evolutionary covariance of the traits, we first performed a single principle component analysis (PCA) on all the PICs from the BI tree without phylogenetic uncertainty. Significance of variable loadings was tested by determining the frequency of loadings different in sign to the ones observed, among 1000 bootstrap replicate analyses (generated with SYSTAT 10.2) corrected for axis reversals \[[@B26]\]. We then also examined correlated evolution among pairwise compared traits, in order to obtain more fine-scale information about the co-evolutionary pattern. For this, we first used PCA to reduce the number of related PIC variables that jointly characterize a specific trait. Thus, separate principal components (PC) were created for the dart, stylophores, glands, and SRO. For each of these organs the PCs, which explained over 70% of the variance, were compared in a correlation analysis. PCA and correlation analysis was applied to the raw data and to PICs resulting from the four trees (see above), using the program JMP 4.0.2 (SAS Institute Inc.). Abbreviations ============= BI, Bayesian inference; ML, Maximum likelihood; MP, Maximum parsimony; PC, Principal component; PCA, Principal component analysis; PIC, phylogenetically independent contrasts SH, Shimodaira-Hasegawa test; SRO, spermatophore receiving organ; TBR, Tree bisection and reconnection. Authors\' contributions ======================= All the snail material and reproductive organ morphologies were collected and/or obtained by JMK. The molecular data collection was performed by JMK, while HS did the phylogenetic tree reconstruction and monophyly testing. HS also provided the independent contrasts, that were used by JMK in the principal component and correlation analyses. JMK and HS wrote the paper. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Principal component analyses on raw data and phylogenetically independent contrasts of love-dart and reproductive morphology data.The independent contrasts are based on the raw data and the three alternative trees (see Methods), namely BI or ML with and without consideration of five cases of phylogenetic uncertainty. The ML tree topologies with and without phylogenetic uncertainty were identical. The table shows the eigenvectors for each variable in the different principal components (PC). These vectors are a measure for the weight of the variable in the PC. The eigenvalue and % variance are given for each PC. These values express how much of the total variance in the data is explained by that PC. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 2 Matrix of Pearson correlations (below diagonal) and their significance (above diagonal) for the comparison of the different principal components (PC) and the flagellum based on the raw data and the three other phylogenetic trees.Results are shown for the raw data and the analyses based on PICs calculated from the BI with phylogenetic uncertainty and the ML trees either with or without phylogenetic uncertainty. Note that the ML trees with or without uncertainty are identical. \, Significance after Bonferroni correction. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 3 Information about the location and collectors of the species used in the analysis.Numbers behind collector names represent the catalog number in the malacological collection of the Academy of Natural Sciences of Philadelphia. The EMBL accession numbers for the new and published DNA sequences are also included. Published sequences all derive from Wade et al. 2001 \[[@B25]\]. ‡, The specimens from this location were used for sequencing; § for these species the DNA sequence were available from EMBL but no specimens could be obtained for investigation. Therefore, the morphology data were based on Azuma 1995 \[[@B44]\]. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 4 Description of the data scored for the different reproductive organ characteristics.When traits are absent these are scored as zero. All relative sizes were measured on the camera lucida drawings. When the size is indicated relative to the reproductive organs this refers to the posterior reproductive organs (excluding the albumen gland) thus making sure that they are independent of the size of the penis and bursa tract. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 5 The sequences for PCR and sequencing primers.Primers JOR58F, JOR59F, and JOR28R850 were used for PCR, whereby JOR59F was employed as the forward primer for Cernuella cisalpinaand Euhadra quaesita*, and JOR58F for all remaining species. Primers JOR28F50, JOR28F400, JOR28R401, JOR28F600, JOR28R601, and JOR28R850 were employed in sequencing reactions. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ We are grateful to all snail collectors mentioned in [Additional file 3](#S1){ref-type="supplementary-material"}; Barbara Hassert, Joe Lange, and Cathy Levesque for technical assistance; Göran Arnqvist, Tim Birkhead, Manfred Gahr, Arndt von Haeseler, Andries Ter Maat, Nico Michiels, Igor Muratov, Michael Landolfa, David Rogers, and Lukas Schärer, and the anonymous referees for fruitful discussions and valuable comments. This research was supported by a Casimir-Ziegler Fellowship of the Royal Netherlands Academy of Arts and Sciences (KNAW) and a Jessup-McHenry Award from the Academy of Natural Sciences of Philadelphia to JMK.	PubMed Central	2024-06-05T03:55:55.664363	2005-3-30	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1080126/", "journal": "BMC Evol Biol. 2005 Mar 30; 5:25", "authors": [ { "first": "Joris M", "last": "Koene" }, { "first": "Hinrich", "last": "Schulenburg" } ] }
PMC1080127	Background ========== CCR3, the eotaxin receptor, is the major chemokine receptor expressed on eosinophils, basophils and a subpopulation of Th2 lymphocytes \[[@B1]-[@B10]\]. Recently, CCR3 has been shown to be upregulated on neutrophils and monocytoid U937 cells by interferons in vitroand to be expressed by endothelial cells, epithelial cells and mast cells \[[@B11]-[@B16]\]. The relevance of these findings and the function of CCR3 on non-leukocytes remain to be elucidated since the only cells that consistently accumulate following eotaxin administration in vivoare myeloid cells (primarily eosinophils) \[[@B17]-[@B20]\]. To date, the complete mRNA and genomic organization of only a limited number of chemokine receptors has been described \[[@B21]-[@B26]\]. These studies have shown that the 5\' untranslated region (5\'UTR) can be complex and contain up to 11 exons as in the CXCR2 gene. As a result, alternative splicing and transcription directed by multiple promoters can give rise to variable mRNA isoforms. The function of these 5\' untranslated exons has not been examined except for a single study focused on CCR2, demonstrating a transcriptional role for exon 1 \[[@B24]\]. We have previously characterized the genomic structure and promoter function of the human CCR3 gene \[[@B27]\]. The CCR3 gene contains at least 4 exons that give rise to multiple mRNA species by alternative splicing. The first 1.6 kb of the 5\' flanking region of exon 1 had strong promoter activity in eosinophilic, lymphoid and respiratory epithelial cell lines. Deletion analysis revealed differential regulation of the CCR3 promoter in eosinophilic and epithelial cell lines suggesting the presence of lineage-specific elements. Interestingly, exon 1 enhanced the activity of the promoter. Since our initial characterization, two other groups have studied the CCR3 promoter \[[@B28],[@B29]\]; however, their studies focused on lymphocytic and monocytic cell lines, respectively, rather than eosinophils. Scotet et al. \[[@B29]\] demonstrated that the human CCR3 promoter is active in vitroin lymphocytic cell lines. They also demonstrate a role for chromatin remodeling in the regulation of CCR3 expression in Th2 cells. Vijh et al. \[[@B28]\] demonstrated that the human CCR3 promoter is active in monocytic cell lines and defined the minimal promoter that consists of a downstream promoter element (DPE), a common element in Drosophila genes, but rare in human genes. This element is upstream (50 bp) of the exon 1 sequence studied in the current report. It has been reported that 5\' untranslated exons, and sometimes introns, can regulate the expression of genes in two different ways. An untranslated sequence can act as a tissue-specific translational regulator. A striking example is the gonadotropin releasing hormone gene which is transcribed in multiple tissue, but cannot be translated due to the lack of specific intron removal \[[@B30]\]. Alternatively, untranslated regions (UTR) can facilitate transcription of a gene. Examples include a GATA-1 site in the 5\'-UTR of the γ-globin gene, an HNF-1 site in the plasminogen gene and a C/EBP site in the CCR2 gene \[[@B24],[@B31],[@B32]\]. While the mechanism of action is not completely clear, it is thought that transcription factors binding to untranslated regions affect transcription of the gene through interactions with the RNA transcription complex. The role of untranslated exons in the CCR3 gene has not been studied. In this report, DNase I hypersensitivity identified a major hypersensitive site located in the vicinity of untranslated exon 1. Furthermore, the transcription factor GATA-1 is shown to bind to untranslated exon 1, suggesting a potential mechanism for the regulation of CCR3 transcription by this exon. Finally, using a transgenic approach, we demonstrate that the 1.6 kb 5\' flanking region of CCR3 (including exon 1) has promoter activity in vivo. Results ======= DNase I hypersensitivity in the CCR3 locus ------------------------------------------ We aimed to define regulatory regions in the CCR3 gene by a gene-wide search. Thus, we performed DNase I hypersensitivity of CCR3 in primary human eosinophils. The entire 24 kb of the CCR3 gene were screened for DNase I hypersensitivity using probes specific for Eco RI and Hind III fragments (Figure [1A](#F1){ref-type="fig"} and data not shown). Only one hypersensitive site was noted; this was in the area of exon 1 (Figure [1B](#F1){ref-type="fig"}). Similar results were obtained with nuclei from the eosinophilic AML14.3D10 cell line (data not shown). This cell line does not express CCR3 unless induced with butyric acid and IL-5; suggesting that chromatin remodeling in the HS1 site precedes CCR3 transcription. Thus, our analysis revealed areas of active chromatin remodeling in the vicinity of exon 1 suggesting that this area may be important for CCR3 transcription. Binding of nuclear proteins to exon 1 of the CCR3 gene ------------------------------------------------------ DNase I hypersensitivity indicated that a region consistent with exon 1 is active in CCR3 transcription. Together with our previous data showing that untranslated exon 1 has an important role in CCR3 transcription \[[@B27]\], we hypothesized that nuclear proteins bind to exon 1, and in turn regulate the transcription of CCR3. In order to test this hypothesis, a double-stranded oligonucleotide probe that corresponds to bp +10 to +60 of the CCR3 gene was prepared, referred to as E1-FL (exon 1- full length, Figure [2A](#F2){ref-type="fig"}). This is the exact sequence that was deleted in the CCR3(-exon1).pGL3 plasmid that demonstrated decreased activity compared to the full length 1.6 kb construct \[[@B27]\]. Nuclear extracts from AML14.3D10 cells were incubated with the probe and resolved on a polyacrylamide gel. Two bands were visible (Figure [2B](#F2){ref-type="fig"}). The upper band was eliminated when 150x molar excess of the unlabelled probe was used (CC: E1-FL in Figure [2B](#F2){ref-type="fig"}), indicating that this is the specific band. In order to precisely localize the region responsible for factor binding, overlapping cold competitors were used: E1-A spanning from +10 to +31, E1-B spanning from +25 to +46 and E1-C spanning from +40 to +60. The specific band was eliminated with E1-B and E1-C cold competitors indicating that the factor binds in the region between +25 and +60 (Figure [2B](#F2){ref-type="fig"}). In summary, these data indicate the presence of proteins in the nuclei of AML14.3D10 cells that bind to CCR3 exon 1 between bp 25 and 60. Binding of GATA-1 to exon 1 of the CCR3 gene -------------------------------------------- In order to define the proteins capable of binding to CCR3 exon 1, the exon 1 sequence was analyzed using the publicly available TFSEARCH engine. This analysis indicated the presence of consensus DNA-binding sites in the exon 1 region for several transcription factors (i.e. GATA, AML1, SRY, S8 etc.). Proteins of the GATA family have been detected in eosinophils (specifically GATA-1, GATA-2 and low levels of GATA-3) \[[@B33]\]. Additionally, GATA-1 transactivates the EOS47 promoter, an eosinophil-specific promoter, through a site in the 5\'UTR \[[@B34]\]. Therefore, we hypothesized that proteins of the GATA family bind to sites in exon 1. In order to address the binding of GATA proteins to CCR3 exon 1, we prepared oligonucleotides corresponding to probe E1-B and -C with the GATA site mutated. As demonstrated in Figure [3A](#F3){ref-type="fig"}, when used as cold competitor, the GATA mutant oligonucleotide was not capable of interfering with the binding of nuclear factors to either CCR3 exon 1 probe E1-B or probe E1-C. Additionally, we used a GATA consensus oligonucleotide with binding sites for GATA proteins. This oligonucleotide completely abolished binding of nuclear factors to both CCR3 exon 1 probes E1-B and E1-C (Figure [3A](#F3){ref-type="fig"}). Together, these data indicate that GATA proteins, present in nuclear extracts of AML14.3D10 cells, are capable of binding to CCR3 exon 1 sequence. In order to delineate which GATA factors are binding to exon 1, nuclear extracts from AML14.3D10 cells were incubated with the radiolabelled full-length CCR3 exon 1 probe in the presence or absence of anti-GATA-1 antibody. As seen in Figure [3B](#F3){ref-type="fig"}, the GATA-1 antibody was able to interfere with the binding of the protein factor from the nuclear extracts to the radiolabelled probe. In order to more precisely localize to which part of exon 1 GATA-1 is capable of binding, overlapping oligonucleotide probes were radiolabelled and incubated with nuclear extracts from AML14.3D10 cells in the presence and absence of the GATA-1 antibody. This treatment efficiently inhibited the binding of probes E1-B and -C to the nuclear factor (Figure [3C](#F3){ref-type="fig"}). These data indicate that GATA-1, present in nuclear extracts of AML14.3D10 cells, is at least one of the factors capable of binding the CCR3 exon 1 between bp 25 and 60. Characterization of the CCR3 promoter in vivo --------------------------------------------- We have previously characterized the human CCR3 promoter in vitro\[[@B27]\]. In order to determine if this region had promoter activity in vivoand to assess cell-specificity, we generated transgenic mice expressing the reporter gene EGFP under the control of the CCR3 promoter (Figure [4A](#F4){ref-type="fig"}). 1.6 kb of the promoter and 60 base pairs of exon 1 were cloned upstream of the EGFP gene. Seven transgenic founder lines were identified. Two of the founder lines did not transmit the transgene to their offspring and one line did not show transgene expression as determined by Northern blotting and RT-PCR (data not shown). In four of the lines the level and pattern of transgene expression varied suggesting integration site effects. Two of the lines (4.1 and 4.2) displayed minimal mRNA expression and were not further analyzed. In contrast, the other two lines (3.1 and 3.2) showed high expression of the transgene in multiple tissues. Line 3.1 displayed varying levels of mRNA expression in all organs tested: lungs, kidneys, jejunum, thymus, bone marrow, and spleen, with highest expression in the thymus (by Northern blot analysis, Figure [4B](#F4){ref-type="fig"} and RT-PCR, data not shown). While we expected the transcript to be about 1 kb in size (Figure [4A](#F4){ref-type="fig"}), two bands \~4 and 5 kb were apparent. Since this was not the case with other lines, we suspect it is integration-site specific and another gene may have been transcribed together with GFP. However, the size of the RT-PCR amplified fragment was of the expected size and the protein was detectable by immunohistochemistry (see below) thus implying that this did not affect the translation of the protein and integrity of the epitope for the antibody. Expression in the lungs, jejunum, thymus and kidney was confirmed by anti-GFP immunohistochemistry (Figure [4C](#F4){ref-type="fig"} and data not shown). In the jejunum, staining was present predominantly in stromal cells (Figure [4C](#F4){ref-type="fig"}). Of note, eosinophils normally reside in the gastrointestinal tract and are located in the stroma of the jejunum. However, in the lungs, which are normally devoid of eosinophils at baseline, there was GFP expression that was confined to the alveolar lining consistent with predominant expression in type I pneumocytes (Figure [4C](#F4){ref-type="fig"}). In the thymus, compared to staining without primary antibody (data not shown), all thymocytes stained positive for GFP. Additionally, in the kidney all cells were positive, both in the medulla and cortex (data not shown). Line 3.2 displayed a different pattern of transgene expression. Northern blot analysis revealed expression in multiple organs with the lungs and kidneys expressing the highest levels (Figure [4B](#F4){ref-type="fig"}). Consistent with this, protein was detected by immunohistochemistry at high levels in lungs and kidneys and only marginally in the jejunum (Figure [4C](#F4){ref-type="fig"} and data not shown). While staining in the lungs was comparable to line 3.1, only a minority of stromal cells and no epithelial cells were stained in the jejunum. In the kidney, almost all cells expressed the transgene in the medulla, while in the cortex staining was prominent in the glomeruli. There was no anti-GFP staining observed in wild type mice (Figure [4C](#F4){ref-type="fig"}). No changes in expression were found when the transgene was crossed with CD2.IL-5 transgenic mice (data not shown). In summary, the 1.6 kb of the 5\' flanking region of the CCR3 gene has strong promoter activity in vivo. However, both the level and pattern of expression vary between founder lines and lack cell specificity. Thus, the identified region of the CCR3 promoter contains a broadly active promoter with hematopoietic and non-hematopoietic activity. Discussion ========== In this report, DNase I hypersensitivity implicated untranslated exon 1 in regulating CCR3 transcription. Furthermore, nuclear proteins derived from eosinophilic cells were shown to bind CCR3 exon between nucleotides +25 to +60. Using unlabelled competitors and antibodies, proteins of the GATA family, specifically GATA-1, were shown to bind to this region. Taken together, these data suggest that untranslated exon 1, via GATA-1, has a regulatory role in CCR3 transcription. Finally, we demonstrate that the 1.6-kb CCR3 promoter element, that includes exon 1, is broadly active in vivo. The HS1 site was located in the vicinity of exon 1. Combined with our previous results demonstrating diminished promoter activity when exon 1 is deleted from the promoter construct, these data suggested that 5\' untranslated exon 1 may have a regulatory function. It has been reported that the 5\' untranslated exons may contain sequences that facilitate transcription of the gene. Examples include a GATA-1 site in the 5\'-UTR of the γ-globin gene, an HNF-1 site in the plasminogen gene, a PU.1 site in the PU.1 gene and a C/EBP site in the CCR2 gene \[[@B24],[@B31],[@B32],[@B35]\]. Thus, we hypothesized that nuclear proteins bind to exon 1. Our EMSA analysis, coupled with cold competitors and specific antibodies, indicates that proteins of the GATA family, specifically GATA-1 bind to the CCR3 exon 1. Thus, GATA-1 binding to exon 1 may regulate CCR3 transcription. Alternatively, GATA-1 binding to exon 1 may affect transcription start site function, RNA stability or translation. These possibilities will be addressed in future studies. It is important that this data be viewed with what is known about other myeloid-specific promoters, that have often proven to be difficult to function independently in vivo. For example, constructs using the 5\' flanking region of myeloid-specific genes have not been useful for transgenic work (such as the CD14 promoter \[[@B36]\], the c-kit promoter \[[@B37]\], or the 1.7 kb CD11b promoter \[[@B38]\]). Better success was obtained when the entire gene, including the open reading frame, was used (e.g. the human cathepsin G, chicken lysozyme and c-fps/fes transgenic constructs \[[@B39]-[@B41]\]). These constructs were at least 6 kb in size and contained all exons and introns and several kb of 5\' and 3\' flanking sequence. Presumably, these larger constructs contained the locus control region (LCR)- sequences that have the ability to dominantly control gene expression in any chromosomal region. This in turn allows for a high degree of consistency among independent mouse lines with regard to cell specificity, level of expression and proportionality to gene copy number. These regions may be located at several different sites in the gene, including introns and coding exons; thus, screening with DNase I hypersensitivity is usually the first method employed to identify these regions. Transgenic mice expressing the EGFP reporter gene under the control of the CCR3 promoter demonstrate that the 1.6 kb promoter and 60 bp of exon 1 of the CCR3 gene confer strong promoter activity in vivo. However, these sequences do not contain the entire LCR, since expression of the reporter gene was variable among multiple founder lines. DNase I hypersensitivity studies discovered one hypersensitive site in the CCR3 locus. The HS1 site was apparently not sufficient for integration-site independent effects, since it is contained in the promoter construct used. Thus, future studies will need to broaden the search for the CCR3 LCR. It is important to note that the true cellular specificity of CCR3 has not been established. While this gene product is often considered to be specific for inflammatory cells involved in allergic inflammation (eosinophils, mast cells, and possibly Th2 cells), several reports have documented expression by other cell types including additional leukocytes (e.g. dendritic cells), as well as tissue cells (epithelial and endothelial cells) \[[@B11]-[@B16],[@B42]\]. Thus, it remains to be determined if our observation that the CCR3 promoter has broad activity in vivorepresents the true endogenous activity. Since CCR3 is expressed strongly on eosinophils, analysis of the signals that induce its expression may give insight into the molecular mechanisms for the commitment of myeloid progenitors to the eosinophil lineage. It is generally believed that transcription factors are the final common pathway driving differentiation and that hematopoietic commitment to different lineages is driven by alternative expression of specific combinations of transcription factors \[[@B43],[@B44]\]. Although no eosinophil-specific transcription factors have been reported, eosinophil commitment appears to be regulated by GATA-1, PU-1 and C/EBP proteins \[[@B34],[@B45]-[@B48]\]. Consistent with this, DNA binding sites for these transcription factors are found in several eosinophil-selective promoters, such as the promoter for major basic protein (MBP), IL-5 receptor alpha (IL-5Rα) chain and Charcot-Leyden crystal (CLC) protein. Specifically, ectopic overexpression of GATA-1 in chicken myeloblasts leads to transdifferentiation into eosinophils or thromboblasts depending on the dose used \[[@B47]\]. GATA-1 transactivates several eosinophil-selective promoters \[[@B48]\] including the avian EOS47 promoter in which the GATA-1 site is located downstream of the transcription start site \[[@B34]\]. Conclusion ========== In summary, in this report we have demonstrated that: 1) DNase I hypersensitivity studies implicate untranslated exon 1 in CCR3 transcription; 2) proteins of the GATA family, specifically GATA-1, bind to untranslated exon 1 in the CCR3 gene; and 3) the 1.6 kb 5\' flanking region of the CCR3 gene is broadly active as a promoter in vivo. Methods ======= Cell culture ------------ The AML14.3D10 cell line (kindly provided by C.C. Paul, Dayton VA Medical Center, Dayton, OH) \[[@B49],[@B50]\] was grown in RPMI 1640 (Gibco BRL, Gaithersburg, MD) containing 10% fetal calf serum (FCS, Gibco BRL), 50 μM 2-mercaptoethanol (Sigma, St. Louis, MO), 0.1 mM nonessential amino acids (Gibco BRL), 1 mM sodium pyruvate (Sigma), and penicillin-streptomycin (Gibco BRL). Eosinophils were isolated by anti-CD16 negative selection from granulocyte preparations of healthy or atopic volunteers as described previously \[[@B51]\]. DNase I hypersensitivity ------------------------ Nuclei were derived from cell lines and primary cells using a polyamine buffer containing 0.34 M sucrose, 13.3 mM Tris (pH7.5), 53.2 mM KCl, 13.3 mM NaCl, 2 mM EDTA, 0.5 mM EGTA, 0.133 mM spermine, 0.5 mM spermidine, 0.1% TritonX-100, 2 mM MgCl~2~and freshly prepared 2-mercaptoethanol and phenylmethylsulfonyl fluoride (PMSF). Nuclei were then centrifuged at 2300 g for 30 min over a cushion of 1.2 M sucrose and washed twice prior to resuspension in DNase I digestion buffer (60 mM KCl, 5 mM MgCl~2~, 15 mM Tris (pH7.5), 0.1 mM EGTA, 0.5 mM DTT and 5% glycerol). Nuclei were resuspended at a concentration of 2--4 × 10^6^nuclei/ml and gentle DNase I digestion was carried out in a volume of 0.2 ml with 0 to 100 units of DNase I (Roche) for 5 minutes at 30°C. Following DNase I treatment, nuclei were lysed and DNA purified by phenol/chloroform extraction and ethanol precipitation. DNA was digested with appropriate restriction enzymes (Eco RI and Hind III) and electrophoresed on a 0.8% agarose gel. Following transfer to nylon membranes, hybridization was performed using standard procedures. Probe fragments were made by PCR from genomic DNA. Figure [1](#F1){ref-type="fig"} depicts the CCR3 gene structure and restriction fragments and their corresponding probes that were used to span the entire gene. Preparation of nuclear extracts from cultured cells --------------------------------------------------- Cultured cells were washed twice with ice cold phosphate buffered saline (PBS, Gibco BRL). 2.5 × 10^6^cells were lysed in lysis buffer \[100 mM N-2-hydroxyethylpiperazine-N\'-2-ethanesulfonic acid (HEPES), pH 7.9, 10 mM KCl, 0.1 mM EDTA, 1.5 mM MgCl~2~, 0.2% Nonidet P-40, 1 mM dithiothreitol (DTT), and 0.5 mM PMSF\], briefly vortexed at a moderate speed, then incubated on ice for 5 minutes. Sample was centrifuged and the pellet was next resuspended in 20 μl of extraction buffer (20 mM HEPES, pH 7.9, 420 mM NaCl, 0.1 mM EDTA, 1.5 mM MgCl~2~, 25% glycerol, 1 mM DTT, and 0.5 mM PMSF), mildly vortexed, and incubated on ice for 15 minutes. Samples were centrifuged to pellet the nuclear debris. Supernatants were placed in silicon coated microcentrifuge tubes and stored at -80°C until further use. Synthetic oligonucleotides -------------------------- Single-stranded oligonucleotides based on the sequence of untranslated exon 1 of the CCR3 promoter were synthesized by Integrated DNA Technologies, Inc. (Coralville, IA). One full length 51 base pair (+10 to +60) oligonucleotide (GGTACCACTGGTCTTCTTGTGCT[TATC]{.underline}CGGGCAAGAACT[TATC]{.underline}GAAATACA) and three overlapping oligonucleotides (+10 to +31, +25 to +46, and +40 to +60) and their reverse complements were produced (Figure [2](#F2){ref-type="fig"}). Two of the overlapping fragments contained putative GATA binding sites (underlined). Two mutants and their complements were also made which changed the GATA sequence TATC to TTGA. This mutation does not change the GC content of the oligonucleotides, nor does it create a new transcription factor site for any of the transcription factors represented in the publicly available TFSEARCH engine. Each oligonucleotide was resuspended to a concentration of 50 μM in TE buffer (10 mM Tris-Cl, pH 7.4, and 0.1 mM EDTA, Sigma). Complimentary single-stranded oligonucleotides were annealed at a concentration of 10 mM in restriction enzyme buffer M (10 mM Tris-Cl, 10 mM MgCl~2~, 50 mM NaCl, and 1 mM dithioerythritol, Roche Molecular Biochemicals, Indianapolis, IN). Samples were placed in a 95°C dry heat block for 5 minutes and then the block was removed from the unit and allowed to cool slowly to room temperature. The double-stranded oligomers were diluted to 1 μM with TE buffer and 30 ng were end-labelled using \[γ-^32^P\]ATP (NEN Life Science, Boston, MA) and T4 polynucleotide kinase (Gibco BRL). Probe was purified over Quick Spin G-25 Sephadex columns (Roche) and recovered in a volume of 50 μl. Electrophoretic mobility shift assay (EMSA) ------------------------------------------- Protein content of the nuclear extracts were determined by Bradford (Coomassie) assay (Pierce, Rockford, IL). Total protein (5 μg) was incubated on ice for 10 minutes with 2X EMSA buffer (24% glycerol, 0.08 μg/ml poly dI-dC, 24 mM HEPES, pH 7.9, 8 mM Tris-Cl, pH 7.9, 2 mM EDTA, 2 mM DTT, 50 mM KCl, and 10 mM MgCl~2,~), and when indicated, with 150 fold excess of cold competitor oligonucleotide. Radiolabelled oligo probe was added to each sample and incubation continued for an additional 10 minutes on ice. For antibody supershift assays, anti-GATA-1 antibody (clone C20, Santa Cruz Biotechnology, Santa Cruz, CA) was added following the addition of the probe and samples incubated on ice for one hour. In control experiments, isotype-control antibodies did not have a significant effect. The DNA-protein complexes were then resolved on a non-denaturing 5% acrylamide gel \[29:1 acrylamide/bis-acrylamide, 0.5X TBE buffer (44.5 mM Tris, 44.5 mM borate, and 1 mM EDTA), and 25% glycerol\] at constant current of 30 mA for approximately 60 minutes. Gels were dried on blotting paper and exposed to x-ray film. Generation and analysis of transgenic lines ------------------------------------------- 1.6 kb of the CCR3 promoter, including 60 bp of untranslated exon 1, was subcloned into the pEGFP vector (Clontech, Palo Alto, CA). The transgenic construct containing the promoter, reporter gene enhanced green fluorescent probe (EGFP) and the SV40 polyadenylation signal (Figure [4](#F4){ref-type="fig"}) were liberated by Afl II and Hind III digestion and injected into the pronucleus of fertilized eggs from FVB/N mice by the Transgenic Core Facility at Cincinnati Children\'s Hospital Medical Center. Transgenic mice were identified by Southern blot analysis after digestion with BamHI, using the SV40 poly A fragment as a probe. Lines were tested for expression of the transgene in multiple organs by Northern blotting and RT-PCT. The SV40 poly A fragment was used as a probe for Northern blotting. PCR primers for EGFP were as follows: 5\'ATGGTGAGCCAAGGGCGAG3\' and 5\'CTTGTACAGCTCGTCCATG3\'. RNA integrity and RT efficiency was verified by performing PCR for β-actin on the same cDNA samples. Primers used were as follows: 5\'GGAATCCTGTGGCATCCATGAAACT3\' and 5\'TAAAACGCAGCTCAGTAACAGTCCG3\'. Anti-GFP immunohistochemistry ----------------------------- Immunohistochemistry was performed on frozen sections essentially as described \[[@B52]\]. Briefly, following endogenous peroxidase quenching, slides were blocked and stained with a rabbit-anti-GFP antibody (AB3080 at 1:400 dilution, Chemicon International, Temecula, CA). The slides were washed and incubated with biotinylated goat anti-rabbit antibody and avidin-peroxidase complex (Vectastain ABC Peroxidase Elite kit, Vector Laboratories). The slides were then developed by nickel diaminobenzidine, enhanced nickel cobalt chloride to form a black precipitate and counterstained with nuclear fast red. Authors\' contributions ======================= JLC carried out the EMSA analysis. LEK performed the DNAse hypersensitivity studies. NZ carried out the generation and analysis of transgenic mice, participated in the design and coordination of the overall study and drafted the manuscript. MER participated in the design and coordination of the overall study and helped to draft the manuscript. Acknowledgements ================ The authors wish to thank Drs. Bruce Aronow, Cindy Bachurski and Gurjit Hershey for technical expertise, and Dr. Simon Hogan and Patricia Fulkerson for critical review of the manuscript. This work was supported by the American Heart Association Scientist Development Grant (to NZ). Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### DNase I hypersensitivity studies.In (A), a schematic representation of the CCR3 gene is shown. Exons 1 through 4 (E1 -- E4) are depicted as boxes; open box represents the open reading frame, while the closed box represents the untranslated region. The position of Hind III (H) and Eco RI (RI) restriction sites is depicted with arrows and probes used for DNase I hypersensitivity are depicted as lines above the genomic fragment. The position of hypersensitive site is depicted as HS1. In (B), Southern blot analysis for the HS site is shown. Nuclei from the primary human eosinophils were digested with indicated doses of DNase I for 5 minutes at 30°C. Following DNA purification and digestion with restriction enzymes (shown in figure is Eco RI), DNA was electrophoresed on an agarose gel and transferred to nylon membranes. Following hybridization with genomic probes, membranes were exposed to film. Size markers are shown on the right. Ethidium bromide staining of the gel is also shown. ::: ![](1471-2172-6-7-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Binding of transcription factors to CCR3 exon 1.In (A), a schematic representation of the CCR3 exon region used for electrophoretic mobility shift assays is shown. E1-FL is the exon 1 full-length probe. Overlapping short probes are called E1-A, E1-B and E1-C, respectively. The DNA sequence is shown with GATA sites boxed. In (B), EMSA assay with the full-length probe and extracts from eosinophilic AML14.3D10 cells is shown. As cold competitors (CC), the full-length probe and short probes E1-A through C were used. The arrow depicts the specific band. A representative experiment, of three similar experiments, is shown. ::: ![](1471-2172-6-7-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Binding of GATA factors to CCR3 exon 1.In (A), EMSA assay with probes E1-B (left panel) and E1-C (right panel) and nuclear extracts from AML14.3D10 cells is shown. Three cold competitors were used in each case: the probe itself (E1-B-WT and E1-C-WT, respectively), the probe with the GATA site mutated (E1-B-mut and E1-C-mut, respectively) and the GATA consensus oligonucleotide (GATA cons). A representative experiment, of three similar experiments, is shown. In (B), EMSA assay with the full-length probe (E1-FL) and extracts from eosinophilic AML14.3D10 cells in the presence or absence of antibodies against GATA-1 is shown. The arrow depicts the specific band. In (C), EMSA assay with probes E1-B (left panel) and E1-C (right panel) and nuclear extracts from AML14.3D10 cells and cold competitor probe itself (CC: E1-B and CC: E1-C, respectively) or anti-GATA-1 antibody (GATA-1 atb) is shown. ::: ![](1471-2172-6-7-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### The CCR3/EGFP transgenic mice.In (A), a schematic representation of the transgenic construct is shown. The genomic fragment encoding 1.6 kb of the CCR3 promoter including 60 bp of exon 1 (E1), was cloned into the Hind III and Bam HI sites upstream from the EGFP gene in the pEGFP1 vector. The positions of restriction sites are indicated. In (B), Northern blot analysis of multiple tissues from transgenic mice is shown. RNA was isolated from the spleen, lung, thymus, kidney and jejunum of CCR3/EGFP transgenic mice (lines 3.1 and 3.2). 10 μg total RNA was electrophoresed on an agarose gel, transferred to nylon membranes and probed with the SV40 polyA probe in order to detect expression of the transgene. Position of 28S and 18S is depicted with arrows. Ethidium bromide staining of the gel is also shown. In (C), immunohistochemistry using an anti-GFP antibody is shown. Organs were collected, frozen sections obtained and anti-GFP immunohistochemistry performed. Positive staining is represented as black precipitate. ::: ![](1471-2172-6-7-4) :::	PubMed Central	2024-06-05T03:55:55.669707	2005-4-4	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1080127/", "journal": "BMC Immunol. 2005 Apr 4; 6:7", "authors": [ { "first": "Nives", "last": "Zimmermann" }, { "first": "Jessica L", "last": "Colyer" }, { "first": "Laura E", "last": "Koch" }, { "first": "Marc E", "last": "Rothenberg" } ] }
PMC1082716	Background ========== Human immunodeficiency virus (HIV) enters target cells by forming a ternary complex between the viral envelope protein gp120 and two cellular receptor proteins: CD4 and a chemokine receptor \[\[[@B1]-[@B6]\], reviewed in \[[@B7]\]\]. HIV viral strains have been described which use a wide range of different chemokine receptors, although the majority use either CCR5 (R5 strains), CXCR4 (X4 strains) or both of these receptors. Consistent with a requirement for chemokine receptors as cofactors for viral entry, the chemokine ligands have been reported to reduce HIV infectivity in vitro\[[@B8]-[@B10]\]. Furthermore, mutations in the gene encoding CCR5, such as the CCR5-Δ 32 allele, provide some protection against HIV infection in vivo\[[@B11]-[@B13]\]. Consequently, agents which block HIV interaction with chemokine receptors are candidate antiviral therapies which can be used in conjunction with protease inhibitors and reverse transcriptase inhibitors to attenuate a third phase of the virus life-cycle: cell entry \[[@B7],[@B10],[@B14],[@B15]\], in the same way as the novel fusion inhibitor enfuvrtide \[[@B16]\] Interestingly, the HIV gp120 protein which interacts with the chemokine co-receptor primarily through its V3 loop can induce leukocyte chemotaxis, demonstrating that some intracellular signals are generated through the the virus:receptor interaction \[[@B17],[@B18]\]. This signalling occurs even though the site of the gp120 interaction with the chemokine receptors appears to be only partially overlapping with the natural ligand binding site \[[@B14],[@B19]-[@B22]\]. It has been proposed that this chemotactic signalling might play a role during HIV infection in vivo, possibly by recruiting susceptible T-cells to sites of viral replication \[[@B18]\]. In other retroviruses envelope/receptor interactions are known to be mitogenic \[[@B23]\] and this may facilitate nuclear translocation and integration of the provirus. In HIV, however, it is not known whether the ability to productively engage the chemokine receptors in this way plays any direct role in acute viral entry and subsequent productive infection of the target cell. Guntermann and colleagues showed that pertussis toxin (which blocks G~i~-mediated signalling through chemokine receptors) block cellular infection with HIV in vitro\[[@B24]\]. Montes et al. obtained similar results, and also showed that the MEK inhibitor U0126 could block both chemokine-receptor-induced ERK activity and HIV infection in vitro\[[@B25]\]. However, neither pertussis toxin nor MEK inhibition are specific for chemokine signalling pathways: G~i~and ERKs participate in other intracellular signalling pathways, so it is possible that HIV infection was inhibited because of blockade of downstream pathways not initiated through productive occupancy of the chemokine receptors. Recently, we have described a new class of chemokine inhibitors, termed Broad Spectrum Chemokine Inhibitors (BSCIs) which block chemokine-induced chemotaxis in a range of leukocytes, irrespective of the chemokine used \[[@B26],[@B27]\]. These BSCIs are highly selective for chemokines, however, and have no effect on chemotaxis induced by a range of other chemoattractants such as TGF-β, fMLP or C5a. Importantly, the molecular target of the BSCIs is not the chemokine receptors themselves: BSCIs do not bind to chemokine receptors, do not affect chemokine receptor levels on the cell surface, and do not interefere with the binding of chemokine ligands to the receptors \[[@B27]\]. Instead, they are thought to specifically inhibit intracellular signals required for chemokine-induced migration but not for migration induced by non-chemokine pathways \[[@B27]\], although their molecular target has not yet been published. As a result, members of the BSCI family have been shown to be potentially useful new anti-inflammatory agents in a wide range of diseases \[[@B27]\]. BSCIs provide an ideal tool to probe the importance of chemokine-induced intracellular signalling in HIV infection. Since the effects of these compounds are apparently selective for chemokine receptor-induced signals, if BSCIs interefere with cellular infection by HIV in vitrothis will indicate that productive signalling by the chemokine co-receptor is likely to be important for successful infection. In the present study, we have investigated whether the first BSCI to be described, termed Peptide 3 \[[@B26]\], affects gp120 binding to chemokine receptors or cellular infection by HIV in vitro. Results ======= Effect of Peptide 3 on gp120 binding ------------------------------------ The binding of gp120 to chemokine receptors is likely to involve sequences in the V3 loop of gp120 \[[@B28],[@B20]\]. We therefore synthesised peptide sequences from the V3 loop of the M-tropic BaL strain and the T-tropic IIIb strain and analysed the binding of these biotinylated peptides to the THP-1 and Jurkat cells. Specific (competable) binding of gp120:V3(BaL) to THP-1 cells was detected at 100 μM (Fig [1a](#F1){ref-type="fig"}). In contrast, specific binding of gp120:V3(BaL) to Jurkat cells was not detected even at concentrations up to 500 μM (Fig [1a](#F1){ref-type="fig"}). These observations are consistent with the hypothesis that gp120:V3(BaL) binds specifically to CCR5, which is expressed on the surface of THP-1 monocytic cells but not on Jurkat T-cells. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Specific binding of gp120:V3 loop peptides to THP-1 and Jurkat cells. (a) Binding of gp120:V3(BaL)-biotin 10^6^Jurkat cells or THP-1 cells per reaction were incubated with various concentrations of N-terminal biotinylated peptide in the presence and absence of 10 mM unlabelled peptide. Specific binding is expressed as the absorbance in the absence of unlabelled peptide minus the absorbance in the presence of competitor (b) Binding of gp120:V3(IIIb)-biotin under the same conditions as in (a). All reactions were performed in 100 μl of binding medium at 4°C (see Materials and Methods). Values are mean ± SEM from triplicate determinations. \* p \< 0.05 Student\'s t-test for specific binding. ::: ![](1742-4690-2-23-1) ::: Specific binding of gp120:V3(IIIb) at 100 μM to both Jurkat T-cells and THP-1 cells was detected. There was approximately 5-fold greater specific binding to the Jurkat cells than the THP-1 cells (Fig [1b](#F1){ref-type="fig"}). These observations are consistent with the hypothesis that gp120:V3(IIIb) binds specifically to CXCR4, which is expressed on both THP-1 and Jurkat cells, but at higher levels on the T-cell line. We next incubated THP-1 and Jurkat cells with 100 μM of each biotinylated gp120:V3 peptide in the presence of various concentrations of Peptide 3. Peptide 3 had no effect on the binding of gp120:V3(IIIb) to Jurkat cells (Fig [2a](#F2){ref-type="fig"}), even though it powerfully inhibited SDF-1α induced chemotaxis over the same concentration range (Fig [2b](#F2){ref-type="fig"}). Under the same conditions, the CXCR4 receptor antagonist AMD3100 \[[@B31]\] blocked both gp120:V3(IIIb) binding and SDF-1α-induced migration with similar IC50s (Fig [2c, d](#F2){ref-type="fig"}). Similarly, Peptide 3 had no effect on the binding of gp120:V3(BaL) to THP-1 cells (Fig [2e](#F2){ref-type="fig"}).under conditions where RANTES-induced chemotaxis was powerfully inhibited (Fig [2f](#F2){ref-type="fig"}). Taken together, these data confirm that Peptide 3 blocks chemokine signalling without blocking gp120 interaction with the chemokine receptors, consistent with previous observations that BSCIs such as Peptide 3 do not block chemokine ligand interactions with their receptors \[[@B27]\]. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Effect of Peptide 3 on gp120:V3 loop peptide binding to cells. (a) The binding of 100 μM N-terminal biotinylated gp120:V3(IIIb) to THP-1 cells was measured in the presence of various concentrations of peptide 3. In each case, the non-specific binding (in the presence of 10 mM unlabelled gp120:V3 loop peptide) has been subtracted. (b) Chemotaxis in response to 100 ng/ml SDF1α was measured in the presence of various concentrations of peptide 3. (c) and (d) As for (a) and (b) except that the CXCR4 receptor antagonist AMD3100 was used in place of Peptide 3. (e) As for (a) except that the effect of Peptide 3 on the binding of gp120:V3(BaL) to THP-1 cells was determined. (f) As for (b) except that the effect of Peptide 3 on chemotaxis induced by 25 ng/ml MIP1α was determined. All binding reactions were performed with 10^6^cells in 100 μl of binding medium at 4°C. Chemotaxis assays were performed with 5 × 10^4^cells per well. Values are mean ± SEM of triplicate determinations. ::: ![](1742-4690-2-23-2) ::: Effect of peptides on HIV infection in vitro -------------------------------------------- HIV infection of Jurkat T-cells using the laboratory-adapted T-tropic isolate IIIb was monitored using two different assays. Firstly, Jurkat T-cells in 96-well plates were pre-treated with either Peptide 3, vehicle (as a negative control) or SDF1α (as a positive control) for 4 hours, then exposed to HIV virus (10^6^TCID~50~) and pulsed at 2--3 day intervals with Peptide 3, SDF1α or medium alone as appropriate. After two weeks in culture, the extent of viral infection was assayed by measuring the reverse transcriptase activity in the supernatant, as a measure of viral replication in the culture \[[@B32]\] In six separate experiments, Peptide 3 (100 μM) had no effect on virus replication following HIV exposure (Fig [3a](#F3){ref-type="fig"}), while SDF1α inhibited reverse transcriptase activity by an average of 75%. No effect was seen on cell viability under any of the treatment conditions. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### *Peptide 3 inhibition of HIV infectivity in vitro**. (a) HIV (IIIb) replication in cultures of Jurkat T-cells was estimated by measuring the supernatant reverse transcriptase activity two weeks after infection. Peptide 3 was at 100 μM final concentration and SDF-1α was added at 100 ng/ml final concentration 1 hour prior to exposure to virus. Values are mean ± SEM from 12 wells, expressed as the percentage of the reverse transcriptase activity in the supernatant from the control wells. The experiment shown is typical of six separate experiments. (b) HIV (IIIb) infectivity of Jurkat T-cells was estimated by staining cells treated identically to those in (a) for p24^gag^expression. Values are mean ± S.D. percentage of cells stained for p24^gag^averaged from 12 fields of view from each of two separate wells. (c) HIV (MN) infectivity of THP-1 cells measured as in (b). MIP1α and SDF1α were used at 100 ng/ml final concentration. ::: ![](1742-4690-2-23-3) ::: HIV infection of Jurkat T-cells was also monitored by high sensitivity quantitative immunofluoresence detection of viral p24^gag^expression. Jurkat cells were infected with HIV in the presence or absence of Peptide 3 (100 μM) or SDF1α (100 ng/ml) as described above. Approximately 48 h after infection, the cells were attached to glass slides using a cytospin and then fixed by immersion in ice-cold 70% ethanol for 90 seconds. Expression of p24^gag^was determined using quantitative immunofluoresence as previously described \[[@B33]\], except that the primary antibody was post-fixed to the section using paraformaldehyde to increase the sensitivity of the technique (see Methods). Viral infectivity was expressed as the number of cells stained for p24^gag^expressed as a proportion of the total number of cells (detected using Hoechst 33342 nuclear dye). Consistent with the reverse transcriptase assay results, SDF1α inhibited viral infectivity by more than 80% (Fig [3b](#F3){ref-type="fig"}), while peptide 3 had no effect. Infection of THP-1 cells with M-tropic isolates does not generate high levels of virus particles and hence the reverse transcriptase assay is not sufficiently sensitive to monitor the progress of the infection. However, it was possible to assess HIV infectivity of THP-1 cells using high sensitivity immunofluorescent detection of p24^gag^. THP-1 cells were differentiated with hydrocortisone and PMA, then treated with TNFα, resulting in adherent monolayers on glass chamber slides. The THP-1 cells were then treated with either Peptide 3 (100 μM), MIP1α (100 ng/ml) or SDF1α (100 ng/ml) as for the Jurkat cells. THP-1 cells were infected with HIV strain MN at a concentration previously validated to produce easily detectable infection and grown for 72 h prior to fixation and staining for p24^gag^. In contrast to the findings with HIV strain IIIb infection of Jurkat T cells, Peptide 3 inhibited infection of THP-1 cells by more than 80% (Fig [3c](#F3){ref-type="fig"}), very similar to the effect of MIP1α. SDF-1a had no statistically significant effect on HIV strain MN infection of THP-1 cells, confirming that the infection was entirely CCR5-dependent, even though THP-1 cells express CXCR4. Discussion ========== Taken together our results suggest that, at least under some conditions, the generation of intracellular signals by the chemokine co-receptor during HIV infection might be necessary for productive infection. Since Peptide 3 powerfully inhibited CCR5-dependent HIV infection of THP-1 cells under conditions where gp120 binding to CCR5 was unaffected but chemotaixs in response to RANTES was profoundly blocked, it is likely that at least some of the signals elicted by CCR5 occupation that result in chemotaxis are required for successful infection of the cell by HIV. Since BSCIs, such as Peptide 3, do not block chemokine receptor internalisation induced by ligand binding \[[@B27]\], it seems likely that the HIV successfully entered the cell in the presence of Peptide 3, but that some later stage in the viral life cycle was dependent on one or more intracellular signal generated by chemokine receptor occupancy. These results are consistent with, but extend, the findings of Guntermann \[[@B24]\] and Montes \[[@B25]\] who saw similar effects with pertussis toxin and a MEK inhibitor. It is unclear why Peptide 3 blocked CCR5-dependent HIV infection of THP-1 cells but had no effect on CXCR4-dependent infection of Jurkat T-cells under similar condtions (even though Peptide 3 efficiently blocks SDF1α dependent chemotaxis). It is possible that infection of certain cell types (such as monocyte/macrophage cells) is more dependent on a chemokine receptor-induced intracellular signal than infection of other cell types (such as T-lymphocytes). This may reflect the fact the the Jurkat cells were proliferating at the time of infection, whereas the THP-1 derived macrophages were quiescent. However, it is also possible that this difference is due to the particularly high levels of CXCR4 which are expressed on Jurkat T cells. The high levels of receptor on this cell line might render infection relatively insensitive to intracellular signalling requirements compared with native T-cells or other cell types expressing physiological levels of chemokine co-receptors. Irrespective of the reasons for this difference, our preliminary studies illustrate the need to further investigate the role of intracellular signals induced by co-receptor occupancy as participants in the viral life cycle. Furthermore, the recent discovery of much more potent non-peptide BSCIs, such as the acylaminocaprolactams \[[@B34],[@B35]\], opens up the possibility that interfering with chemokine receptor-induced signalling might offer an alternative therapeutic strategy to blocking chemokine receptor binding. The best acylaminocaprolactam BSCIs are cheap, orally bioavailable and apparently free of acute toxicity \[[@B35]\], and, on the basis of studies such as ours, warrant further investigation as part of an antiviral combination therapy for HIV. Conclusion ========== We conclude that selective inhibition of the intracellular signal(s) generated through interaction of the HIV gp120 envelope protein with its chemokine co-receptor can block productive HIV infection in THP-1 derived macrophages in vitro. However, blockade of CXCR4-mediated signalling in Jurkat T-cells has no effect on HIV infection in vitro. Under certain condtions, therefore, HIV infection may require activation of the chemokine co-receptor signalling pathway. Methods ======= Peptides -------- Peptides were prepared by Affiniti (Exeter, U.K.) using standard solid phase chemistry, followed by reverse-phase HPLC purification to greater than 95% purity. Peptide 3 (derived from amino acids 51--62 of mature human MCP-1) had the sequence EICADPKQKWVQ \[[@B26]\]. Labelled peptides were also synthesised corresponding to the sequence of the full length V3 loop (including terminal cysteine residues) of gp120 from HIV-1 IIIb and HIV-1 BaL with an N-terminal co-synthetic biotin label. All peptides were prepared as TFA salts and dissolved in sterile MilliQ and stored at -20°C until used. Chemotaxis experiments ---------------------- Chemotaxis experiments were performed essentially as previously described \[[@B26],[@B36]\], using the ChemoTx disposable 96-well transwell migration plates, with PVP-free membrane (6 μm pore size). Chemoattractant (lower compartment) and cells upper compartment) were suspended in Gey\'s Balanced salt solution + 1 mg/ml BSA. Consistent with our previous recommendations \[[@B36]\], putative inhibitors were added at equal concentration to both the upper and lower compartments. Migration was allowed to proceed for 2 hours at 37°C. The number of cells which had migrated to the lower compartment was determined using the vital dye MTT and interpolation of a standard curve. Each condition was determined in triplicate, and the number of cells migrating in the absence of chemoattractant was subtracted to determine the chemokine-dependent chemotaxis which was reported. Binding assays -------------- Cells (either Jurkat T-cells or THP-1 cells) were grown in RPMI 1640 medium supplemented with 10% fetal calf serum, 2 mM glutamine, 20 μM β-mercaptoethanol, 100 U/ml penicillin and 100 μg/ml streptomycin and maintained between 2 × 10^5^and 1 × 10^6^cells/ml. Prior to performing a binding assay, cells were spun out (100 × g; 4 mins) and washed 3 times in ice-cold PBS. A volume of cell suspension in PBS containing 10^6^cells was pipetted into each well of a V-bottom 96-well plate (Gibco BRL) and spun out (100 × g; 4 mins). Cells from triplicate wells were then resuspended in 100 μl binding medium (PBS pH 7.2 containing 0.1% fatty-acid free bovine serum albumin (BSA)) containing labelled peptide in the presence or absence of 10 mM unlabelled peptide. The plate was then incubated on ice for 90 minutes. Cells were washed 3 times with 380 μl of ice-cold PBS, spinning out the cells each time (100 × g; 4 mins), and resuspended in 100 μl binding medium containing streptavidin-peroxidase (Amersham International) at 1:1000 dilution. Cells were incubated for a further 15 minutes on ice to allow labelling of any bound biotinylated peptide, then washed 4 times as above. Cells were finally incubated with 200 μl TMB substrate (K-Blue, Bionostics) for 20 minutes at room temperature, and the reaction stopped by addition of 50 μl 2 M HCl. The plate was spun (3,000 × g ; 3 mins) and 200 μl of the coloured product was transferred to an empty 96-well ELISA plate and the absorbance at 450 nm determined. HIV-1 infection and reverse transcriptase assays ------------------------------------------------ HIV-1 stocks were prepared in the following manner. HIV-1 IIIb (provided by the MRC AIDS Reagent Programme) was used to infect Jurkat T cells and the progression of infection was monitored using the reverse transcriptase assay for 10--14 days. Virus was then harvested and assayed for infectivity using a TCID50 assay with Jurkat T cells as targets as previously described \[[@B37]\]. Virus containing supernatants were centrifuged to remove cellular debris and then stored in aliqouts in liquid nitrogen until used. Stocks of HIV-1 MN strain were prepared in a similar manner, except that H9 T cells were used and the progression of infection was monitored by immunofluoresence (because of the low RT activity) in this strain. Experimental infection of Jurkat T cells was performed by incubating cells win the presence of test peptide or chemokine with aliquots of stored virus at the titres described in the text. Thereafter, the cells were fed fresh medium containing test peptide or chemokine where appropriate, every 48 h. Viral replication two weeks after infection was estimated by measuring reverse transcriptase activity in the supernatant using the Potts Mini RT assay as previously described \[[@B32]\]. THP-1 cells were differentiated prior to infection with hydrocortisone and PMA in 8-well chamber slides. Sixteen hours prior to infection, TNFα was added (100 ng/ml). Twelve hours later, the medium was aspirated and replaced with fresh medium containing the test peptides or chemokines as appropriate. After a further four hours, virus from the frozen stocks was adeed to the cells, which were the processed for immunofluoresence between 28 h and 72 h after infection. Immunofluorescence detection of p24^gag^ ---------------------------------------- Jurkat cells following HIV infection were attached to 8-well chamber slides (Becton-Dickinson) by spinning the slides using a plate rotor in a Labofuge centrifuge (Heraeus) at 3,000 × g for 5 minutes. Attached Jurkat cells or THP-1 cells were then fixed by dipping the slides into ice-cold 70% ethanol for 90 seconds. Non-specific binding was blocked by incubation with 3% fatty acid-free BSA in TBS for 1 hour at room temprature. Cells were incubated with the mouse monoclonal anti-HIV-1 p24^gag^antibody EH12E1 (ref 38;AIDS Reagent Program, NIBSC) at 10 μg/ml in 3% BSA in TBS at room temperature overnight. Unbound antibody was removed with 3 × 3 min washes in PBS, and bound antibody was then fixed to the slide by incubation with 3.8% phosphate buffered formalin pH7.2 for 10 mins at room temperature, followed by 3 further 3 min washes in PBS. Bound antibody was visualised using donkey anti-mouse IgG FITC conjugate (715-095-150; Jackson Immunoresearch) at 30 μg/ml in 3% BSA/TBS + 1 ng/ml Hoescht 33342 for 6 hours at room temperature. Twelve fields of view (100× magnification) were captured from each well of the chamber slide using an Olympus Provis AX electronic microscope connected to a Power Macintosh 8500, running OpenLab image analysis software (Improvision), under both FITC illumination conditions (NIBA filter block; λex = 470--490 nm, dichroic mirror = 505 nm, λem = 515--550 nm) and UV illumination conditions (Chroma 31000; λex = 340--380 nm, dichroic mirror = 400 nm, λem = 435--485 nm). Images were acquired with a Hamamatsu C4742-05 monochrome digital camera with 10-bit depth in a 1280-1024 pixel field connected to a DIG Snapper frame grabber. The exposure time, amplifier gain and offset values were controlled by the OpenLab software and were held constant throughout the experiment. A background (an image captured without a slide under the objective) was digitally subtracted from every image. A threshold was then applied to each image which was the lowest threhold that detected \<1% of the pixels of an image of uninfected cells stained under identical conditions. The number of objects exceeding this threshold in each field of view were counted. A similar procedure was used to determine the total number of nuclei in the same field of view, using the image captured under UV illumination conditions. The ratio of positively stained objects to nuclei in each field of view was reported as the percentage of cells stained for p24^gag^. Competing interests =================== DJG is an inventor on a range of patents filed by the University of Cambridge containing composition of matter and pharmaceutical use claims for a wide range of BSCIs, including Peptide 3 used in this manuscript. The patent specifically claims the use of BSCIs for the prevention and/or treatment of HIV infection. An exclusive license to these patents have been granted by the University of Cambridge to Ipsen (Paris, France) and DJG may gain financially from the successful exploitation of this intellectual property. Authors\' contributions ======================= DJG and AML jointly conceived of these studies; AML performed the HIV infection experiments and RT assays; DJG performed the immunofluoresence detection analyses and the in vitrobinding assays and functional migration assays. DJG drafted this manuscript, which was critically reviewed by AML and both authors approve the final version for submission and publication. Acknowledgements ================ This work was supported by grants from NeoRx Corporation (Seattle, Wa., USA) and the Wellcome Trust to D.J.G., who was a Royal Society University Research Fellow. We are grateful to Paul Sheppard and his colleagues at Affiniti (now Biomol International) for help and advice on designing peptides, labelled peptides and their derivatives. The monoclonal antibody to p24^gag*^was obtained from the AIDS Reagent Program at NIBSC.	PubMed Central	2024-06-05T03:55:55.672720	2005-4-4	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1082716/", "journal": "Retrovirology. 2005 Apr 4; 2:23", "authors": [ { "first": "David J", "last": "Grainger" }, { "first": "Andrew ML", "last": "Lever" } ] }
PMC1082883	Background ========== The water column in a stratified lake provides vertical gradients of habitat qualities for zooplankton. Surface layers (epilimnion) and deep layers (hypolimnion) separated by a strong temperature gradient (thermocline) differ very much with respect to temperature, light, food availability and predation risk. Although zooplankton are defined as \"floating\" in the water column, their populations show distinct horizontal and vertical distribution patterns \[[@B1]\]. At least the vertical distribution is the result of active habitat choice. Diel vertical migration is a striking example of habitat shift in response to changing suitability. Large zooplankton leave the warm, lighted and often food-rich epilimnion during the day to dwell in the cold, dark hypolimnion where food may be of low quantity and poor quality, in order to avoid predation by visually hunting predators (fish). They return to the surface layers at night when the predation risk is small \[[@B2]\]. Numerous studies have shown that this pattern is influenced by food conditions \[[@B3]-[@B5]\] and temperature gradient \[[@B6],[@B7]\]. However, there is increasing evidence that the vertical distribution of algal food for herbivorous zooplankton is not always \"typical\". In particular, oligotrophic and mesotrophic lakes often exhibit a deep-water algal maximum, i.e. highest algal densities are not found in the epilimnion but in the upper hypolimnion, below the thermocline \[[@B8]-[@B10]\]. Zooplankton then face a trade-off between high temperature (fast development) but poor food (low reproductive potential) in the epilimnion and low temperature but high food availability in the hypolimnion \[[@B11]\]. In the absence of visual predation risk (e. g. in lakes with scarce populations of planktivorous fish or at night), herbivorous zooplankton should, therefore, distribute vertically so as to optimise the fitness gain depending on the temperature and food gradient. In fact, it has been observed that zooplankton forced to deep layers by fish predation during the day return to the warm epilimnion at night even if their food is more abundant in the hypolimnion \[[@B10],[@B12]\]. It has been suggested \[[@B11]\] and evidence has been given that the vertical distribution of herbivorous zooplankton (Daphnia) faced with the trade-off follows an \"Ideal Free Distribution (IFD) with Costs\" \[[@B13]\]. However, \[[@B11]\] placed a question mark after the title of their paper as, although their experimental results resembled an IFD with costs, the mechanism was not yet clear. Since then, several new experiments have shown that some of the assumptions underlying the IFD model \[[@B14]\] are fulfilled in the system. (1) Daphniaare \"ideal\" organisms as they select the habitat best suited to their survival and reproduction. The proportion of individuals dwelling in a particular habitat is correlated to the relative fitness gain in that habitat \[[@B15]\]. (2) Daphniaare also \"free\" to choose the habitat. Every habitat is equally accessible to all members of the population. The observed distribution is dynamic, i. e. it is the equilibrium result of individuals moving randomly up and down and allocating a proportion of their time to a certain habitat \[[@B16]\]. (3) The IFD model assumes that all individuals are equal. Being members of a clone, Daphniaare genetically equal, but differing in size they may not be metabolically equal. Small differences in depth distributions of different size classes have been found \[[@B17]\], but only during the day (i.e. probably related to predation risk). This problem can be overcome by looking at the response of individual size classes separately. Although several assumptions underlying the IFD have now been tested in this system, the one that is probably most important has not. The IFD model strongly requires that habitat suitability is density dependent. Unfortunately, information on density dependence of Daphniadistributions so far is scarce, and only a weak effect has been found \[[@B18]\]. The purpose of this study was to measure the density dependence of the vertical distribution. The study was designed to test the following hypotheses. (1) The vertical distribution becomes broader as Daphniadensity increases. (2) As density effects are related to food availability, but not to temperature, Daphniawill, at higher densities, spread asymmetrically to deeper layers in order to exploit the deep-water algal maximum better. (3) Differences in the depth distributions of differently sized Daphniacan be explained by intraspecific competition. Results ======= Daphniapopulations in both columns started to grow after a short lag-phase (Fig. [1](#F1){ref-type="fig"}). Total biomass increased steadily from the very beginning indicating that individuals accumulated body mass before they started reproducing. Although the absolute numbers of Daphniadiffered between the two towers, possibly due to small differences in the inoculum, the population growth pattern was similar. The differences between the two towers were due to large numbers of very small individuals in tower 1. As the small individuals contribute less to the total biomass, differences were less pronounced for biomasses than for numbers. The thermocline was stable at 2.5 m, and the deep-water algal maximum was maintained throughout the experimental period with little variation (Fig. [2](#F2){ref-type="fig"}). Particle concentration in the epilimnion varied between 0.05 and 0.19 mg C L^-1^with a trend to higher particle concentrations towards the end of the experiment. There was a significant linear relationship (n = 33, r^2^= 0.624) between the particle concentration in the epilimnion (P~epi~) and the log-transformed Daphniabiomass (B~max~) in the algal maximum (P~epi~= 0.059 log(B~max~) + 0.129). Vertical distributions were assigned to five classes (cf. Fig. [1](#F1){ref-type="fig"} and Methods) of increasing total Daphniabiomass (g dry mass per tower), and the mean distributions are shown in Fig. [2](#F2){ref-type="fig"}. The vertical distribution changed with biomass density. At low population densities, biomass peaked sharply at the thermocline, but the distribution spread out when the total biomass increased (Fig. [2](#F2){ref-type="fig"}). The peak moved below the thermocline at the time when total Daphniabiomass was highest, and the distribution broadened further. However, there were differences in the distributions of different size classes of Daphnia(Fig. [3](#F3){ref-type="fig"}). Small size classes tended to stay higher in the water column than large ones, regardless of the total Daphniabiomass, although their distribution spread out, too, even below the algal maximum. The results of visual inspection of the vertical distributions are confirmed by Principal Component Analysis (PCA, Table [1](#T1){ref-type="table"}). The first three principal components (PC1--PC3) explain more than 95 % of the variation. PC1 contrasts the biomass proportions at the thermocline (port 6, 2.5 m) with the proportions at ports 7 and 8, i.e. the algal maximum. Hence it depicts the downward spreading of the distribution. PC2 contrasts the proportions at the thermocline with those in the epilimnion. Finally, PC3 depicts the distributional shifts within the algal maximum, but its contribution is very small so it is not considered. The ANOVA on the factor scores of PC1 shows significant influences of total biomass (Mass) and Daphniasize class (Size) on the vertical distribution, and a significant interaction between these two factors (Table [2](#T2){ref-type="table"}). This confirms the density dependence of the distribution as well as the different responses of size groups to total density. Total density had no significant effect on the factor scores of PC2, but there were significant effects of Size and Tower, a significant Mass × Tower interaction, and a marginally significant Size × Tower interaction. This shows that PC2 (contrast between thermocline and epilimnion) is mainly influenced by Daphniasize (cf. Fig. [3](#F3){ref-type="fig"}). The significant Mass × Tower and Size × Tower interactions suggest that the tower effect on PC2 is a consequence of differences in size composition of the populations. The descriptive analysis of the shape of the vertical distributions was followed by more quantitative approaches with ungrouped samples. Figure [4](#F4){ref-type="fig"} depicts the positive relationship between the median of the Daphniadepth distribution and total density. The higher the Daphniabiomass per tower, the deeper the median. A linear regression between median depth (MD) and the log-transformed biomass (B) is significant (MD = 0.746 log(B) + 2.72, n = 33, r^2^= 0.754). The proportion of the total population (in terms of biomass) dwelling within the algal maximum increased strongly with total density, but levelled off at ca. 80 % above a total Daphniabiomass of 3 g per tower (Fig. [5](#F5){ref-type="fig"}). The relationship between the log-transformed biomass in the total tower and the log-transformed biomass in the algal maximum was linear (Fig. [6](#F6){ref-type="fig"}). In order to improve homoscedasticity and to obtain a conservative estimate of the slope, the lowest biomass value (initial sample of tower 2) was excluded from the analysis. There was no difference between towers so the values were pooled. The slope of the resulting regression (Table [3](#T3){ref-type="table"}) is significantly larger than 1 (d.f. = 32, F = 57.3, p \< 0.001), i.e. with increasing total biomass, disproportionally more Daphniadwelled in the algal maximum. As Daphniasize affected the response to total biomass (Mass × Size interaction for PC1), the overall regression was broken down to size classes (Table [3](#T3){ref-type="table"}). There was no significant difference between the slopes of the individual regressions, but the intercepts were different and showed a clear trend, increasing with Daphniasize class. Consequently, all size classes moved toward the algal maximum with increasing total density, but large Daphniamoved deeper than small ones. Maximum (peak) Daphniadensity in each vertical profile increased with total biomass, but the relationship was not linear (Fig. [7](#F7){ref-type="fig"}). Rather, an upper limit between 4 and 5 mg L^-1^(approximately 80--90 individuals L^-1^) seems to be approached when the biomass per tower reached high values. This is direct evidence for density dependence of the distribution, and it provides an estimate of the maximum density Daphnia pulicariacan tolerate under these conditions. Discussion ========== Vertical distribution and diel vertical migrations of zooplankton have been studied for a long time \[[@B19]\] but are still not fully understood. This study fills an important gap by revealing an additional mechanism (density dependence) controlling the vertical distribution of zooplankton. Density dependence is a basic determinant of IFD theory \[[@B14],[@B20]\], and the results of this study conform to the theory. Considering that the conditions in this system differ strongly from the original IFD model, this shows that the concept is rather robust, even if it only \"mimics\" an IFD \[[@B21]\]. For simplicity, IFD models have been developed for discrete patches of habitat suitability, where the differences in habitat suitability were caused solely by biotic factors (food availability, competition). Such a situation has been tested when \[[@B22]\] found that Daphniadistributions between food patches complied with IFD predictions. The \"IFD with Costs\" model \[[@B13]\] considers in addition the effect of abiotic factors. This model is more appropriate for the present situation as higher food availability is linked to higher costs (low temperature). But in contrast to water currents, as in \[[@B13]\], temperature does not only affect the net energy gain; it also limits the speed of development, which is a particularly important factor in parthenogenetic animals with continuous reproduction such as Daphnia. Consequently, ignoring interference from competitors, the basic habitat suitability \[[@B14]\] is the result of the total fitness gain, not just resource input. Also, vertical habitat structure in a lake does not represent discrete patches but changes over gradients, which may not be a problem as IFD models have been shown to be applicable for environmental gradients \[[@B23]\]. However, the vertical temperature gradient in a lake is not smooth. The thermal structure rather creates two different habitats (epilimnion and hypolimnion) connected by a steep gradient (thermocline). We may consider this situation as two habitat patches with some overlap. Because of the trade-off, it is possible that the optimum habitat suitability is in the overlap zone, i.e. in the thermocline. In fact, earlier experiments with similar conditions (10°C temperature gradient) \[[@B11],[@B17]\] resulted in pronounced distribution peaks at the thermocline, very much like the distributions found in this study at low Daphniadensities. Population densities in the earlier studies were not controlled, but a re-analysis of the raw data of \[[@B11]\] showed that they ranged from 0.3--2.8 g dry mass per tower, which is in the lower range of this study. The sharp aggregation in the thermocline can be profitable as long as density effects are negligible. Daphniaare not perfectly \"ideal\" organisms since they do not have complete knowledge about the suitability of all habitats. Hence they need to move around and sample the habitats. The nature of the trade-off may also require that they spend some time in both habitats. It has been demonstrated \[[@B16]\] that the vertical distribution is dynamic, i.e. it is the equilibrium result of random movements and different time allocations of the individuals. Since food availability and temperature are uniform in the hypolimnion as well as in the epilimnion, swimming longer distances up and down does not pay off. The steep environmental gradient provides the opportunity to access sufficient food and high temperatures within a short distance, as long as interference from competitors is low. This changes at high densities when competitive interactions become important. The negative effect of high competitor densities inevitably leads to a broadening of the distribution. There are two mechanisms of competitive interactions in filter-feeding Daphnia. Interference may cause a reduction of food intake when encounters between individuals become too frequent and the continuous filtering process is disturbed \[[@B24],[@B25]\]. Exploitation reduces the food concentration through joint filtering activity. The latter process can only be locally important in the tower system as algal concentrations in the hypolimnion were maintained at approximately 1.5 mg C L^-1^, which is above the \"incipient limiting concentration\" where the feeding rate of Daphniabecomes independent of the food concentration \[[@B26]\]. Only at the upper edge of the food maximum (at the thermocline) may exploitation competition play a role due to the high grazing pressure on algae entering the thermocline from below due to small-scale turbulence. The sharp upper edge of the algal distribution is probably a result of grazing. Both types of interactions among Daphnialead to reduced food uptake. On the other hand, density has little negative influence on the temperature effect (eventually through enhanced metabolic activity). This must cause an asymmetric response of the Daphnia. Spreading into the thermally homogeneous food maximum will have a positive effect since avoiding competitors means obtaining more food with no additional temperature costs, except the additional costs of swimming, which are small \[[@B27]\]. Spreading towards the surface will have little or no positive effect as energy input is limited by the low algal concentration, not by competitors. The experimental results conform well with this prediction. There is a significant effect on the shape of the distribution at high Daphniadensities. The vertical distribution becomes broader and shifts into the algal maximum. Although food availability and temperature are identical over the whole hypolimnion, Daphniado not distribute homogeneously, but the distribution is skewed towards the upper edge. Again, this points to the fact that the optimum habitat at low densities is near the thermocline where access to warm water and food is easiest. The Daphniadistribution follows a gradient of costs (to reach warm water), not resources. The shift can be quantified by the positive relationship between Daphniadensity and median depth, and the increasing proportion of Daphniadwelling within the algal maximum. The density effect is as strong as the effect of increasing hypolimetic temperature \[[@B11]\]. One disadvantage of the experimental design is the correlation of Daphniadensity with time. It may be argued that the downward spread of the distribution is the result of some unknown factor changing with time. For example, the epilimnetic habitat suitability may have become worse if the epilimnetic food conditions deteriorated with time. Measurements do not support this caveat. Although the mean particle volume was equivalent to about 0.1 mg C L^-1^, there were nearly no intact Scenedesmuscells present. The measured particles were small and probably comprised material (e.g. empty cell walls) that had been recycled by Daphnia; thus the food quality was low. In fact, the epilimnetic particle concentration even tended to increase slightly towards the end of the experiment, which if the food were good should have resulted in an upward shift. The significant relationship between particle volume in the epilimnion and total Daphniabiomass rather suggests that the increase in epilimnetic particles was due to Daphniafeeding in deep layers and defecating in the epilimnion \[[@B16]\]. A final argument supporting the view that Daphniadensity is the driving force for the downward shift comes from the estimate of maximum local densities. The distribution is not shaped directly by the total biomass of Daphniain the tower, but by the local density at each depth. The maximum local density tends to reach a plateau at 4--5 mg L^-1^, which is an estimate of the upper limit of density tolerance under the given food conditions. IFD theory assumes that all individuals are equally good competitors \[[@B14]\], but there is little evidence to support this \[[@B21]\]. In nature, individuals will have different competitive abilities, and this requires modified concepts \[[@B28]\]. The concept of unequal competitors offers a large range of possible distributions \[[@B28]\], and only a special case yields a distribution with equal proportions of different competitors that \"mimics\" an IFD \[[@B29]\]. Belonging to a single clone, the Daphniain this experiment are at least genetically equal, but they differ in size. Under constant conditions, large Daphniaare considered to be competitively superior to small ones. There is mechanistic evidence for this relationship on the level of interspecific comparisons \[[@B30],[@B31]\]. Mechanistic support for a similar intraspecific relationship comes from an unpublished laboratory study by C. Kreutzer and M. Boersma. They found that large individuals of D. pulicariahad lower food thresholds for growth than small ones, i.e. were competitively superior. Evidence for intraspecific competition is often inferred from segregation of different size classes in field studies \[[@B32],[@B33]\], but the interpretation is hampered by the fact that fish predation has a similar size-selective effect, hence the differing depth distributions of small and large individuals may be due to a trade-off between their ability to exploit resources and their susceptibility to predation \[[@B34]\]. However, fish predation and any cue for the presence of fish was excluded in this tower experiment. The experimental results are consistent with the predictions of the model assuming the competitive superiority of large individuals. A greater proportion of small Daphniathan of large ones is present in the epilimnion (Fig. [3](#F3){ref-type="fig"}). The distributions at low Daphniadensities are very similar to those found under the same conditions by \[[@B17]\], but the differences in the shapes of the distributions become more pronounced at higher Daphniadensities (increased competition). Small Daphniaspread out in both directions, even into the region below the algal maximum (where food is still more abundant than in the epilimnion). The relatively larger fraction of small Daphniain the epilimnion suggests that temperature is a more important factor for small than for large individuals. The smallest size class in this study contained varying proportions of neonates that are probably not so dependent on high food concentrations during their first hours of life, as they still carry some yolk left from the embryonic stage. Small-scale experiments in stratified columns designed to study individual swimming behavior showed, in fact, that neonates crossed the thermocline much less frequently than adults \[[@B35]\]. If the relative competitive ability of unequal competitors does not change in different habitats, the distribution of competitors can be replaced by the distribution of the sum of competitive abilities. Consequently, there are numerous possible combinations of phenotypes in the different habitats \[[@B28]\]. Nevertheless, it is highly probable that a solution is met where the proportions of different phenotypes are identical in all habitats (for review see \[[@B29]\]). The results of this experiment conform with predictions for constant relative competitive abilities in the different habitats. The proportion of individuals dwelling in the algal maximum increases with Daphniadensities in all four size classes (Table [3](#T3){ref-type="table"}) since the regression slopes are all above unity. There is a significant trend in the elevations, indicating that the size classes differ in their mean depth. However, the slopes of the regressions do not differ, i.e. the relative proportions of the size classes in the algal maximum remain constant regardless of the total Daphniadensity. Different size (competitor) classes respond to increasing competition in the same way. Density dependence probably explains the discrepancy between model predictions and experimental results in \[[@B11]\]. A purely physiological model predicted considerably lower proportions of the total population in the hypolimnion than were found experimentally. This was explained by possible food quality effects. However, the physiological model ignores competition completely. The difference between model prediction and experimental results (at 10°C temperature gradient) is relatively small (50 % vs. 60 %) as Daphniadensities were in the lower part of the range covered by the present study. The value of 60 % at low densities is consistent with the results of this study (c.f. Fig. [5](#F5){ref-type="fig"}). At high Daphniabiomasses, the experimental results are about twice as large as the model predicts (80 % vs. 40 %), which shows that density dependence is an important factor in determining the vertical distribution of Daphnia. We are presently not able to incorporate density dependence into a model of Daphniadistribution. Interference competition rather than exploitative competition is important in our system with continuous food renewal, but quantitative physiological data on the effect of interference are not available. However, the ultimate goal is to construct a quantitative model of the vertical fitness distribution and compare it to the distribution adopted by the animals. Only then will it be possible to test if the assumptions of an IFD model are fulfilled. Conclusion ========== Although the assumptions of the IFD theory are very often violated, a large number of tests have found distributions that \"mimic\" an IFD \[[@B21]\]. This study has shown that density dependence is important for the distribution of Daphnia, hence the basic assumptions of the IFD theory are fulfilled. However, there is a need for modifications. Habitat suitability for herbivorous zooplankton is determined by food availability as well as by directed gradients of biotic (predation) and abiotic (temperature, oxygen) factors. The Ideal Free Distribution with Costs model \[[@B13]\] is more appropriate in this case. Further complications arise as Daphniapopulations are size structured \[[@B36]\], i.e. models for unequal competitors \[[@B28]\] need to be applied. The experimental results conform well to the theoretical expectations considering the costs (asymmetric distribution) and differently sized Daphniaas unequal competitors (size specific distributions), although Daphniaare not omniscient and the equilibrium distribution is dynamic. This shows that the theory is rather robust. The IFD with Costs concept can be helpful in developing a general model of vertical distribution of zooplankton. This is important as the vertical location of grazers has consequences for phytoplankton production and species composition \[[@B37],[@B38]\] and, therefore, for ecosystem function. So far, the IFD with costs concept has only been applied to lake situations with a deep-water algal maximum. Although deep-water algal maxima are not rare \[[@B9]\], they are special cases and trade-offs will be different when algal densities are high in the epilimnion, and day-time mortality caused by fish predation \[[@B4]\] or UV \[[@B12]\] is also high. Diel vertical migrations are the logical consequence. However, predation can be considered another type of cost, and applying the IFD with Costs model may be useful for analysing vertical distributions and migrations of zooplankton under a range of environmental conditions. Methods ======= Experimental system ------------------- The basic experimental design has been described in \[[@B11]\]. A large indoor mesocosm system, the Plön Plankton Towers \[[@B39]\], was used to test the hypothesis. These are two stainless steel columns of approximately 11.5 m height and 1 m diameter. The system can be manipulated with a vertical resolution of 50 cm and sampled through 23 vertical ports per tower. After filling the columns with filtered (10 μm) water from a nearby mesotrophic lake (Schöhsee), they were thermally stratified. As in a natural temperate lake in summer, a warm epilimnion was separated from the cool hypolimnion by a steep temperature gradient (thermocline). The same temperature gradient with the thermocline at 2.5 m depth was maintained throughout the experiment in both columns: 20°C in the epilimnion and 10°C in the hypolimnion between 2.5 and 5.1 m. The lower part of the hypolimnion (below 5.1 m) was set to 8°C (cf. Fig. [2](#F2){ref-type="fig"}). This created a stable layer between 2.5 and 5.1 m depth and prevented the water from being mixed into the deep layers. The diel light cycle was 14 hours light and 10 hours dark. Oxygen was monitored with an electrode. Although there was some oxygen depletion with time, the concentration never fell below 3.5 mg L^-1^. Preparation and sampling ------------------------ To create a deep-water algal maximum a pre-cooled algal suspension was injected by tubes into the 10°C layer where it mixed within 2 hours and stayed. The green alga Scenedesmus obliquusMeyen, known to be good food for Daphnia, was cultured in 10-L jars in dilute (1:4) Z4 medium \[[@B40]\] under continuous light. Although the tower system cannot be kept sterile, this alga was almost the only food source for Daphnia. After pre-culture in 100-L containers, Daphnia pulicariaForbes was introduced into the two towers in approximately equal amounts. In the absence of predation, they started building up a population. A vertical profile of small water samples was collected every morning through sampling tubes from each tower, and the algal density was estimated using a particle counter (CASY^®^, Schärfe GmbH, Reutlingen, Germany). The estimated particle volume was converted to particulate carbon using a calibration curve. The amount of algae necessary to maintain a concentration of approximately 1.5 mg carbon L^-1^in the algal layer was then calculated and the missing amount replenished. This algal concentration provided saturating food conditions for Daphnia. Sampling of Daphniastarted the day after inoculation of the towers and was repeated approximately every second day (17 sampling dates in 36 days). To avoid direct light effects \[[@B17]\], vertical distributions of Daphniawere assessed at night (two hours after lights off). Daphniawere sampled by simultaneously pumping 48 liters of water each from sampling ports at 13 depths (between 0.1 and 10.5 m) through glass traps \[[@B39]\], but later only 11 ports (down to 7.5 m) were used for analysis as no Daphniawere found below that depth. Daphniasamples were preserved in sucrose formaldehyde \[[@B41]\]. Preserved zooplankton samples were automatically counted and sized with a bench top model Optical Plankton Counter (Focal Technologies, Dartmouth, Nova Scotia, Canada). Raw size measurements were converted to body length \[[@B42]\], and individuals were grouped into 4 body size classes (0.6 -- 1.0 mm, 1.01 -- 1.5 mm, 1.51 -- 2.0 mm, 2.0 -- 3.5 mm). The first two size classes comprise neonates and juveniles while adults are found in the latter two. Calculations and statistics --------------------------- Depth distributions of Daphniaare given in terms of dry biomass. For each sampling port, numbers of Daphniaof each size class were multiplied by the mean individual mass of the size class obtained from a length-mass relationship \[[@B43]\] using 42 % of carbon in dry mass. Biomasses in each size class were summed to obtain the total biomass at each depth. This yielded vertical biomass profiles for each size class as well as for the total population. Vertical profiles were integrated to calculate the total biomass per tower, which was used as the independent variable for testing density dependence. For further analysis, total biomasses (dry mass per tower) of the 34 data sets (17 dates × 2 towers) were divided into five biomass groups (0.15--0.49 g, 0.5--0.79 g, 0.8--1.8 g, 1.81--3.0 g, \> 3.0 g) with approximately equal (6--7) frequencies. As the total biomass varied between dates, vertical distributions had to be analyzed as percentages of total biomass at each depth. A detailed description of the method for characterizing distribution patterns has been given in \[[@B11],[@B17]\]. Biomass percentages at each sampling port were subjected to a principal components analysis (PCA) based on a variance-covariance matrix. The results of the PCA are linear combinations (principal components, PCs) of the original dependent variables (percentages at different ports). PCs represent different aspects of the distribution in terms of contrasts between the ports. Depending on the amount of variation explained by them, the first few PCs can be used to describe the main aspects of the Daphniadistribution, i.e. to analyze the shape of the distribution. After testing for normality and homogeneity of variances, the factor scores of the PCs explaining most of the variation (\> 90 %) were subjected to a GLM ANOVA with the estimation of the main effects (biomass group, size class, tower) and the two-way interactions between them. The ANOVA identified the experimental factors significantly related to different aspects (PCs) of the distribution, i.e. it showed which factors affected the shape of the distribution. Although the populations within a tower were sampled repeatedly, the samples were considered independent for two reasons. First, the dependent variable was the distribution. The distribution is dynamic, i.e. individuals swim up and down in the water column \[[@B16]\]. Hence, the population had two complete light cycles to redistribute between samplings. Second, populations were growing in numbers. New individuals were born into the population continuously while old ones died. Consequently, the population consisted in part of different individuals at each sampling date. PCA and ANOVA were carried out using the NCSS^®^statistical package \[[@B44]\]. The median depth of the Daphniadistribution and the proportion of the total biomass dwelling in the algal maximum (between 2.5 and 5.1 m) were related to total biomass in order to get a more quantitative description of the effect of Daphniadensity. To avoid the use of percentages in the statistics, a linear regression analysis was performed with the ungrouped, log-transformed absolute Daphniabiomass in the algal maximum versus the total biomass per tower. The null hypothesis was that the slope of the regression was one, i.e. there was no density dependence of the proportion in the algal maximum. Linear regressions were calculated for the sum of all, as well as for the individual, Daphniasize classes, and slopes and elevations were compared to test whether different size classes showed different responses to Daphniadensity (independent variable). Regression analyses were perfomed with Statistix^®^7.0, Analytical Software, Tallahassee, U.S.A Authors\' contributions ======================= W. L. designed and carried out the experiment, participated in data collection, performed data analyses, and wrote the manuscript. Acknowledgements ================ I am grateful to Dieter Albrecht, Maren Volquartsen and Heike Wardenga for technical support, and Kirsten Kessler and John Havel for valuable comments on an earlier draft of this paper. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Population size and total dry mass of D. pulicariain the two mesocosms (closed symbols = tower 1; open symbols = tower 2). Dotted lines delineate the biomass groups used for the analysis. ::: ![](1741-7007-3-10-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Upper left: Mean vertical gradients of temperature (thin line) and algal concentration (closed circles) during the experiment. Thermal variation was too small to be depicted in this graph. Other panels: Vertical biomass distributions of the total population of D. pulicariaat increasing densities. Numbers in different panels indicate the mean dry biomass in g per tower in the five total biomass groups (both towers combined). Broken lines delimit the zone with the algal maximum. Error bars: ± 1 s.e. ::: ![](1741-7007-3-10-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Comparison of the mean vertical biomass distributions of the smallest (open circles) and the largest (closed circles) size groups of D. pulicariaat the lowest (left) and the highest (right) densities (cf. Fig. 2). ::: ![](1741-7007-3-10-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Relationship between the total biomass of D. pulicariaper tower and the median depth of the biomass distribution (both towers combined). Each point represents a single vertical profile. ::: ![](1741-7007-3-10-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Proportion of the total Daphniabiomass dwelling in the algal maximum (cf. Fig. 2) in relation to total density (g per tower). Symbols indicate the two towers. ::: ![](1741-7007-3-10-5) ::: ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Relation between the total mass of Daphniaper tower and in the algal maximum. Symbols indicate the two mesocosms. Line of equality indicated. ::: ![](1741-7007-3-10-6) ::: ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Maximum density of Daphniain a vertical profile (peak) in dependence on the total biomass per tower. ::: ![](1741-7007-3-10-7) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Principal Components Analysis of the vertical biomass distribution of Daphnia. The eigenvectors of the first three principal components (PC1--PC3) account for more than 95 % of the total variation. Bold numbers depict major contrasts. The bottom line shows the cumulative percentage of the total variation explained by an individual PC. ::: Sampling Port Depth (m) PC1 PC2 PC3 ------------------------- --------------- ------------ ------------ ------------ P1 0.1 0.140 -0.304 -0.069 P3 1.2 0.111 -0.331 0.132 P4 1.6 0.104 -0.315 0.136 P5 2.1 0.137 -0.270 0.236 P6 2.5 0.649 0.671 -0.019 P7 3.0 -0.535 0.375 0.636 P8 3.5 -0.460 0.187 -0.505 P10 4.6 -0.137 0.005 -0.493 P12 5.5 -0.007 -0.011 -0.038 P14 6.5 -0.002 -0.007 -0.017 Cumulative Variance (%) 51.1 91.8 95.7 ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Results of a GLM ANOVA on factor scores of PC1 and PC2 with estimates of the main effects of Daphniabiomass (Mass), size class (Size) and experimental tower, and the two-way interactions between these. ::: Factor d.f. MS F p -------------- ---------- -------- ------- --------- PC1 Mass 4, 92 14.99 55.59 \<0.001 Size 3, 92 6.66 24.69 \<0.001 Tower 1, 92 0.92 3.42 0.068 Mass × Size 12, 92 0.74 2.74 0.003 Mass × Tower 4, 92 0.51 1.88 0.121 Size × Tower 3, 92 0.16 0.59 0.168 PC2 Mass 4, 92 0.71 1.89 0.119 Size 3, 92 18.66 49.96 \<0.001 Tower 1, 92 9.74 26.07 \<0.001 Mass × Size 12, 92 0.15 0.40 0.960 Mass × Tower 4, 92 2.06 5.51 \<0.001 Size × Tower 3, 92 1.11 2.97 0.035 ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Linear regression analysis of the effect of total Daphniabiomass (B~tot~) on the biomass (B~max~) dwelling in the algal maximum (cf. Fig. 6). Comparison of the different Daphniasize classes. Model: log(B~max~) = b·log(B~tot~) + a. ::: Size class slope intercept R^2^ p ------------------------------- ----------- --------------- ---------- --------- 1 1.180 -0.548 0.916 \<0.001 2 1.252 -0.365 0.955 \<0.001 3 1.230 -0.240 0.979 \<0.001 4 1.209 -0.208 0.984 \<0.001 total 1.214 -0.244 0.985 \<0.001 Comparison of regressions d.f. F p Difference of slopes 3, 124 0.45 0.719 Difference of elevations 3, 127 70.43 \<0.001 :::	PubMed Central	2024-06-05T03:55:55.675517	2005-4-6	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1082883/", "journal": "BMC Biol. 2005 Apr 6; 3:10", "authors": [ { "first": "Winfried", "last": "Lampert" } ] }
PMC1082884	Background ========== Arabidopsis thalianahas long been considered the foremost model organism in plant biology. It is favored for its short generation time, plentiful seeds, conveniently small stature, and ease of genetic transformation using Agrobacterium tumefaciens. Its comparatively small genome size, estimated at 140 million base pairs, and low repetitive sequence content drove the choice of Arabidopsisas a target for complete genome sequencing in the early nineties. Ten years later, the genome sequence was completed \[[@B1]\], providing a valuable resource for furthering the understanding of Arabidopsisbiology and providing a reference sequence from which results in Arabidopsiscould be extended to other plants. Since its publication, the Arabidopsisgenome has been mined for clues to numerous important metabolic pathways and biological processes, many of which are documented in peer-reviewed publications including the ArabidopsisBook \[[@B2]\]. Additionally, the Arabidopsisgenome has been used extensively as a tool for comparative genomics, both for genome-wide comparisons and to study specific processes among a wide range of plant species, including the gametophytic transcriptome of mosses \[[@B3]\], wood and secondary cell wall formation in woody gymnosperms \[[@B4]\], and legume symbiosis \[[@B5]\]. Unlike the genomic sequence, which is mostly unambiguous and unlikely to change significantly over time, the genome annotation is dynamic and expected to improve further as we better understand the molecular biology of Arabidopsisand related plants. The original Arabidopsisgenome annotation that accompanied the completed genome sequence in 2000 \[[@B1]\] represents the earliest comprehensive depiction of gene content and predicted gene functions. This original annotation was accumulated over the course of the sequencing effort in the form of individually annotated BAC sequences submitted to GenBank by each of the sequencing centers. Due to the diversity of annotation tools and protocols employed by participating centers during this process, and continuing improvements in annotation resources over the several years of the sequencing project, preliminary gene annotations varied considerably in accuracy and quality at the level of both gene structure and gene function. This heterogeneity within the annotation was most visible in the context of gene families constructed upon completion of the entire genome sequence. Related genes often had dissimilar names and predicted functions as well as incongruent gene structures. A coordinated effort was needed to provide a more useful resource to the plant scientific community. Immediately after the initial data release, The Institute for Genomic Research (TIGR) began a reannotation effort \[[@B6]\], with the goal of improving the annotation by refining gene structure and gene function assignments, employing the latest annotation tools and resources, and applying uniform annotation protocols across the entire genome. Over the course of this reannotation effort, which lasted three years and ended in January 2004, five milestone annotation releases were generated and provided to the public by TIGR, hosted additionally by the National Center for Biotechnology Information (NCBI) and The ArabidopsisInformation Resource (TAIR). The fifth annotation release (January, 2004) represents our final major contribution to the Arabidopsisgenome reannotation effort and is the main focus of this manuscript. The primary goals of this reannotation are summarized as follows: • refine gene structures, including the annotation of alternative splicing variants and untranslated regions (UTRs); • manually review gene names and assign genes to Gene Ontology \[[@B7]\] controlled vocabularies describing molecular function, biological process and cellular location; • recreate chromosome sequences accurately, depicting the genome based on the most current BAC tiling path. Here we present a summary of our annotation methods, efforts and history leading to the fifth and final TIGR release of the Arabidopsisgenome annotation. Results and discussion ====================== Contents of Arabidopsisgenome annotation release 5 ---------------------------------------------------- The final TIGR genome reannotation release contains annotations for 26,207 protein-coding genes, 631 tRNAs, 2 rDNA cassettes (18S, 5.8S and 25S rDNA units), 57 snoRNAs, and 15 snRNAs (Table [1](#T1){ref-type="table"}). Of the 26,207 protein coding genes, 2,330 are annotated with alternative splicing isoforms and 18,099 are annotated with UTRs. Genomic regions with homology to open reading frames (ORFs) of transposable elements (2,355) and pseudogenes (1,652) account for an additional 3,786 annotations, and (in contrast to earlier releases) are now separated from the total protein coding gene count. Taking into account alternative splicing variants, the 26,207 protein-coding genes yield 27,855 distinct protein sequences. Nearly 85% of these proteins contain a match to an InterPro \[[@B8]\] accession via PROSITE \[[@B9]\], ProDom \[[@B10]\], PRINTS \[[@B11]\], Pfam \[[@B12]\] or TIGRFAM \[[@B13]\], and nearly 30% are predicted by TMHMM \[[@B14]\] to contain at least one transmembrane domain. The Arabidopsisgenome sequence is essentially complete. The representation of the Arabidopsisgenome sequence as provided in release 5 is illustrated in Fig. [1](#F1){ref-type="fig"}. The sequenced portion of the Arabidopsisgenome now stands at approximately 119 Mbp, including sequences from 1,611 tiled BACs, PACs, YACs, cosmids and PCR products. Unsequenced regions of the genome are restricted to the centromeres of each chromosome, 5S rDNA clusters on chromosomes 4 and 5, and the nucleolar organizer regions (NOR) at the northern ends of chromosomes 2 and 4. With the exception of the NORs and the northern tip of chromosome 5, every other chromosome terminates with either perfect copies of the telomeric repeat (AAACCCT), or degenerate copies of this sequence that are characteristic of sub-telomeric regions. These repeats are found inverted at the bottom of chromosome 3. The regions of overlap between adjacent BACs in each chromosome tiling path were reviewed extensively during our reannotation effort, and the chromosome sequences were generated based on the joining of regions of BAC sequences to yield our most accurate depiction of contiguous chromosomes. A series of 1000 \'N\' characters were inserted into the chromosome sequence at positions representing the unsequenced regions described above, only to provide placeholders for the unsequenced components. The centromere of chromosome 3 includes two internal sequenced contigs each flanked by unsequenced regions. In addition, partially sequenced BACs mapped to centromeric locations are included in both the chromosomal tiling paths and in the representation of the chromosome sequence in order to provide the most comprehensive sequence data possible. How complete a representation of the genome is the version 5 tiling path and pseudomolecules? In the sequencing phase of the ArabidopsisGenome project, it was agreed that each group would continue sequencing up to the region containing intractable centromeric repeats. In order to make the public version of the genome as complete as possible, centromeric BACs for which sequencing was still in progress but the position of which in the tiling path was known were included in builds of pseudomolecules. These sequences are not included in the genome annotation and consist mainly of transposon-related and other centromere-associated sequences. A minimal estimate of the extent of the genome within the centromeres is \~1 Mb per centromere \[[@B15]\] although a recent new estimate of genome size \[[@B16]\] could indicate that the amount of unsequenced genome is larger than this. As reported previously \[[@B6]\], survey sequencing of representative centromeric BACs revealed no firm evidence for previously undetected genes in the centromeric regions. A second view of genome completeness comes from an assessment of the representation of ArabidopsisESTs in the genome sequence. After removal of contaminating human and E. colisequences, approximately 2% of all ESTs did had no cognate match in the genome sequence \[[@B6]\]. Investigation of 20 of these \"missing genes\" by PCR on genomic DNA revealed that only 3 could be detected and all were organellar in origin. Improvements in the annotation from release 1 through 5 ======================================================= Each annotation release represents one or more milestones within our reannotation effort, providing key contributions towards annotation improvement. These are summarized below and elaborated upon in subsequent sections: • Release 1 (August 2001). • The incorporation and assimilation of non-TIGR BAC sequences and annotations into the TIGR ATH1 Sybase relational database. • TIGR XML format was developed and applied to represent the structured contents of ATH1 for public use. • Release 2 (January 2002) • Approximately 5,000 full length (FL) cDNAs from Ceres, Inc. were incorporated into gene models \[[@B17]\] • The annotation was used as the basis for ATH1-Affymetrix Arabidopsiswhole genome microarray chip design \[[@B18]\]. • Release 3 (August 2002) • Incorporation of the RIKEN ArabidopsisFL-cDNA sequence collection \[[@B19]\] into gene structure annotations using the same methods as employed in the incorporation of the Ceres FL-cDNAs. • Comprehensive analysis of intergenic regions using the latest gene finders, incorporating previously missed gene annotations and new hypothetical genes. • Release 4 (April 2003) • The development and application of the FL-cDNA and EST alignment assembly pipeline PASA, incorporating ESTs and FL-cDNAs into gene structure annotations, modeling alternative splicing variants, and maximizing UTR annotations \[[@B20]\]. • Release 5 (January, 2004) • Improved annotation of transposon-homologous regions and pseudogenes. • cDNA sequences, provided pre-publication by Genoscope \[[@B21]\], allowed for the annotation of an additional \~1000 alternatively spliced genes, nearly doubling the count from the previous release. • Completion of GO assignments to all annotated genes. The overall gene density and gene structure statistics differ little from the initial genome annotation. The statistics alone, however, fail to emphasize the improvements that have been made to individual gene annotations over the course of our reannotation effort. Direct comparisons of individual genes between each of the annotation releases provide a more accurate measure of the level of change. Updates performed on gene structures between successive releases of the annotation include modifying individual exon boundaries, splitting single gene structures into two or more genes, merging multiple gene annotations into single genes, deleting poorly supported genes, adding UTR annotations to existing gene models, and creating new gene models. In addition to structural changes, gene names were systematically refined and Gene Ontology assignments were applied. A summary of the contents and changes made between releases is provided in Table [2](#T2){ref-type="table"}. By comparing release 5 to release 1, we find that only 17,975 (67%) of the original gene structures (excluding UTR updates) remain exactly the same. There were 4,241 new genes modeled, 1,130 gene models deleted, 329 genes merged, 253 genes split, and 7,094 updates to existing gene structures. Any protein-coding genes that are still not annotated are likely to be short, to lack homology to known genes, and/or to be compositionally atypical of the majority of Arabidopsisprotein-coding genes. The changes in the sequenced genome size between annotation releases from 115.4 M bp to 119.0 M bp can be attributed to our refinement of the specification of BAC overlaps, the addition of newly sequenced BACs previously absent from the tiling path, the inclusion of partially sequenced centromeric BACs within the tiling path, and the replacement of partially sequenced BACs with more fully sequenced/assembled versions in subsequent GenBank releases. Improving gene structures ------------------------- Gene structure reannotation focused on improving the accuracy of the existing gene structure components, including the refinement of exon boundaries, annotation of UTRs, and identification of alternative splicing variations and pseudogenes. This effort relied primarily on sequence homology, exploiting spliced transcript and protein alignments to infer gene structures. Improved de-novo gene predictors also proved useful in the process of reviewing the annotated gene structures, especially in regard to hypothetical genes, which lack protein homology or EST support. ### Incorporation of full-length cDNAs and ESTs into gene structures Our initial effort to automate gene structure improvements employed \~5,000 FL-cDNAs generated by Ceres, Inc \[[@B17]\]. We developed software tools for modeling genes automatically using alignments of FL-cDNAs, and performed updates to existing gene structure annotations or modeled new genes where none previously existed. FL-cDNA alignments supported structural modifications for approximately 30% of the previously annotated genes, as well as providing UTR annotations for many genes. Our most recent effort to automate gene structure annotation improvements utilized both FL-cDNAs and EST sequences. We developed the Program to Assemble Spliced Alignments (PASA) annotation pipeline to maximally assemble alignments of FL-cDNA and EST sequences and to automatically incorporate the alignment assemblies into the existing gene structure annotations. This included updating exon structures, adding UTRs, modeling new genes, and annotating alternative splice variants where supported by the transcript alignment data \[[@B20]\]. Through the use of the PASA pipeline, the majority of EST and FL-cDNA alignments were incorporated into the Arabidopsisgene annotations. As of 10/08/2003, GenBank included 31,654 FL-cDNAs and 192,671 non-FL sequences. This data set, supplemented with a transcript sequence database from Genoscope comprising an additional 21,508 FL-cDNAs and 8,039 non-FL sequences, totaled 53,162 FL-cDNAs and 200,710 non-FL sequences. Of the 16,250 genes matching a FL-cDNA, 14,555 gene models are now consistent with the FL-cDNA alignments, integrating 43,445 of the FL-cDNAs into the gene structure annotations. In addition, 90% of the ESTs that provide high quality alignments to the genome are also incorporated into gene structure annotations. The FL-cDNAs that were not fully integrated into gene structure annotations include aberrantly spliced transcripts, antisense mRNAs, polycistronic mRNAs, mRNAs encoding short, partial or unidentifiable ORFs, mRNAs with non-consensus splice sites, and mRNAs that did not align well to the genome using the spliced alignment utilities employed. Several of these topics are elaborated upon in subsequent sections. The annotated gene structures integrating FL-cDNA sequence alignments are identified by tags (\"\<CDNA\_SUPPORT\>\") in the TIGR-XML distribution of our annotation, available on our ftp site \[[@B22]\]. Of the 19,117 Arabidopsisgenes matching alignment assemblies, only 2,867 (15%) lack a FL-cDNA match. Thus nearly all Arabidopsisgenes with expression detectable using current cDNA cloning methods are currently represented by a FL-cDNA sequence. Additional sequence-based methods for ascertaining gene expression, including massively parallel signature sequencing (MPSS) and serial analysis of gene expression (SAGE), have provided evidence for approximately 450 additional expressed genes that were previously annotated as hypothetical proteins due to lack of sequence evidence of expression \[[@B23],[@B24]\]. ### Alternative splicing Alternative splicing of mRNAs has many roles that impact biological systems. Variations in protein sequence resulting from alternative splicing can result in altered structures, functions, or subcellular localizations of gene products \[[@B25]-[@B29]\]. Alternative splicing has been given a great deal of attention in the study of mammalian genomes and is thought to be a major factor contributing to the diversity of gene products and gene functions \[[@B30],[@B31]\]. Given its potential biological significance \[[@B32]\], accurate annotation of alternative splicing in Arabidopsisis clearly important. Experimental investigation of splicing variations in Arabidopsishas been limited to a small number of genes (examples in \[[@B33]-[@B35]\]). Over the course of our reannotation effort, analyses of ESTs and cDNAs indicated that alternative splicing in plants is more prevalent than previously thought \[[@B17],[@B20],[@B36],[@B37]\]. Through automated and manual methods, we have identified and annotated large numbers of splicing variations in Arabidopsis. Of the 26,207 protein-coding genes, 2,330 were found to have alternatively spliced forms. Comparisons between sibling transcript isoforms indicate that at least 30% of the variations result in an altered ORF yielding a non-identical protein sequence (Table [3](#T3){ref-type="table"}). The remainder appear to lie exclusively within the UTR, not affecting the annotated protein sequence. Most of the alternative splicing variations are categorized as alternative donor/acceptor splice sites or unspliced introns. Relatively few examples of splicing variations involved exon skipping (and example of which is shown in Figure [2](#F2){ref-type="fig"}) or alternate terminal exons. Most variations affecting alternate terminal exons were restricted to the UTR regions, indicative of alternate transcriptional start and/or stop sites and presumed impacts on splicing patterns. Variations involving skipped exons tended to impact translations in a similar manner to unspliced introns and alternate acceptors/donors, although they occur much less frequently, with only 130 examples currently identified. These splicing variations would be excellent targets for further functional analyses. ### Unspliced, antisense and dicistronic transcripts There are numerous transcript sequences in GenBank that, when analyzed manually in the context of the genome annotation, do not appear to encode complete proteins. Many of these transcripts contain unspliced introns or indicate alternate splice sites that strongly and adversely impact the presumed correct ORF. It is not clear whether these perceptibly corrupted versions of the genes represent biologically meaningful isoforms, mistakes by the splicing machinery that are of no consequence, or artifacts of the cloning and sequencing methods employed. cDNAs with unspliced introns are often presumed to have originated from incompletely processed mRNAs. In the context of genome annotation, unspliced introns often yield stop codons and/or change the reading frame, resulting in a truncated ORF. However, many of these could be the result of regulated mRNA splicing. For example, an alternatively spliced transcript of RPS4 lacks splicing of an intron, which results in the loss of a terminal protein domain. It has been shown that this incompletely spliced isoform is biologically significant and is required, in addition to the completely spliced isoforms, for wild-type disease resistance \[[@B38]\]. There are 3,025 FL-cDNAs derived from 1,565 Arabidopsisgenes that appear to result from splicing aberrations, as they are incompatible with and appear to corrupt our current representations of the full-length protein coding genes, which are presumed more accurate \[[@B39]\]. During the course of re-annotation, we identified 221 genes for which there are expressed sequences that align with the opposite strand and the transcribed orientation of which is confirmed by splice sites \[[@B40]\]; approximately half of these antisense transcripts derive from FL-cDNAs. Independent confirmation for the existence of naturally occurring antisense transcripts comes from two sources. Using an Affymetrix whole genome tiling array, Yamada et al. \[[@B41]\] reported the detection of antisense transcripts from \~7,600 genes. Using MPSS, Meyers et al. \[[@B23]\] reported the expression of antisense transcripts from 4,698 genes (4,298 exonic and 400 intronic). Although the significance of this large number of antisense transcripts in Arabidopsisremains to be determined, there is a growing recognition of the existence and functional significance of antisense transcripts in a variety of systems \[[@B42]-[@B44]\]. The order of magnitude difference between antisense transcripts recognized in cDNA/EST libraries and those detected by expression analysis and MPSS suggests that many of these transcripts are expressed at low levels and are not found in cDNA/EST libraries, or they represent unspliced transcripts that were not examined here because of a lack of confidence in the direction of their transcription. There are at least 20 examples of mRNAs that provide transcripts corresponding to two adjacent genes \[[@B45]\]. Stop codons intervene and separate the two open reading frames within the transcript and, upon manual examination, it is clear that two distinct genes are represented by the single polycistronic transcript. In several cases, FL-cDNAs corresponding to the individual genes exist as well as the unexpected transcript encoding both genes. Dicistronic transcripts have previously been reported in a number of other eukaryotes including D. melanogaster\[[@B46],[@B47]\], C. elegans\[[@B48]-[@B50]\] and H. sapiens\[[@B51]\], and in some cases have been shown to have functional significance \[[@B52]\]. The small number of these polycistronic transcripts identified in Arabidopsisis an indicator of their low frequency of occurrence. The finding of FL-cDNAs corresponding to individual genes of the polycistronic transcripts suggests that the latter may be an aberration resulting from improper transcriptional termination and polyadenylation, although a functional role has not been ruled out. In some cases, the two genes could plausibly be part of the same pathway or process for which coordinated regulation might be advantageous. Examples of genes found here as dicistronic transcripts include H+-transporting two-sector ATPase (At2g25610) and protein phosphatase 2C (At2g25620), prenylated rab acceptor (PRA1) family protein (At3g13710) and putative RNA-binding protein (At3g13700), and putative UDP-glucose 4-epimerase (At4g10960) and lipase class 3 family protein (At4g10955). Studies are needed to ascertain their significance. ### Other plant ESTs and homologous protein alignments The high-quality, near-perfect transcript alignments of ArabidopsiscDNAs/ESTs to their cognate genomic sequence proved largely amenable to automated incorporation into the genome annotation. Lower quality alignments of homologous FL-cDNA and EST sequences from other plants as well as spliced alignments of homologous proteins also served as excellent sources of data from which to infer gene structures. However, they were not as easy to incorporate computationally given that the reliability of the alignment data often varies considerably across their extent. Thus, identification of gene structures conflicting with these spliced alignments was performed automatically, but updates to individual genes based on these spliced alignments were carried out manually using Neomorphic\'s Annotation Station (Figure [2](#F2){ref-type="fig"}) (Neomorphic was acquired by Affymetrix on 10/31/2000). Genes supported only by homologous proteins or cDNAs/ESTs derived from other plants can be retrieved at \[[@B53]\]. ### Comprehensive gene discovery employing gene prediction tools Gene prediction programs have been useful in identifying potentially novel genes, as well as missed or incorrect exons. In the original Arabidopsisgenome annotation, several genomic regions lacked comprehensive gene identification possibly due to the shortcomings of the programs employed. The operational criterion for instantiating a gene model in the Arabidopsisgenome is for a gene structure to be predicted similarly by two different gene-prediction programs. With our latest set of gene prediction programs including GENSCAN+ \[[@B54]\], GeneMark.hmm \[[@B55]\], and glimmerA (glimmerM variant trained for Arabidopsis\[[@B56]\]), we applied this criterion to all genomic regions annotated as intergenic, automatically creating new genes within each region as the minimal criterion was satisfied. To avoid the spurious promotion of numerous small gene predictions, many of which are likely to be false positives, a conservative minimum protein length cutoff of 110 residues was applied in this automated process. This was chosen conservatively to reflect the 5th percentile of the protein length distribution derived from the previously existing, manually curated Arabidopsisprotein-coding gene annotations. Since previous releases of the annotation lacked the comprehensive annotation of transposon-homologous regions, many intergenic regions were found to harbor gene predictions that matched transposon ORFs. These gene models were specifically excluded from the final round of automated gene modeling and were addressed separately. Through our analysis of intergenic regions we annotated 785 new genes, of which 665 had homology to other proteins. The remaining 120 genes were annotated as additional hypothetical genes. The newly annotated genes with homology to known sequences indicate the significant number of gene annotations missed in the original genome annotation. Thus, improved gene prediction programs and increased database content provided us with an additional set of genes worthy of incorporation into the genome annotation and further study. ### Manual refinement of gene structures Throughout the reannotation project, significant effort has been focused on manually refining intron and exon boundaries of gene models predicted by the various automated processes. Initially, the team of 4--6 annotators would progress along BAC sequences and correct, add and delete gene models as necessary. Later, the annotators assessed pre-computed gene families for consistent gene structures concurrent with functional annotation (described below). Intron-exon boundary refinements and UTR additions were performed by annotators viewing alignments generated by the Eukaryotic Genome Control (EGC) computational pipeline (see methods) using the Annotation Station graphical user interface (Figure [2](#F2){ref-type="fig"}). Gene function annotation ------------------------ The primary goal of the functional annotation effort was to produce a high quality, consistently named proteome. The results from numerous bioinformatics analyses such as homology matches and domain hits were made navigable via the MANATEE web interface \[[@B57]\], which interacts with the annotation database. Gene products were assigned descriptive names based on database matches to gene products and protein domains that have been functionally characterized to avoid problems commonly associated with circular annotation. Through MANATEE, annotators were easily able to access the computationally derived data in a compact summary page. Many of the sections in the primary MANATEE display page link out to supplementary pages with more detailed information or specific analysis results, such as alignments and external database descriptions. A set of naming guidelines (see methods) was adopted to provide consistency in functional annotation, but the variable nature of gene families and types of evidence available make it difficult to mandate exact nomenclature. A critical component of the effort was regular communication and discussion of specific annotation examples among the annotators. Choices among multiple possible names and occasional exceptions to the guidelines were made based on consensus decisions by the annotation group as a whole. ### Protein families Arabidopsisproteins were classified into protein families to facilitate and enhance their functional annotation. The identification of putative protein families enables visualization and navigation of relationships between proteins and allows annotators to curate related genes consistently and accurately as a group. Once all the gene structures had been examined, annotators reexamined family members to ensure that members were consistently and appropriately named within the family context. Classification of proteins into families should produce clusters of proteins with common evolutionary history and sequence similarity and hence similar biochemical function \[[@B58],[@B59]\]. There is no single standard for the classification of protein families \[[@B60]-[@B62]\]. Our approach is based upon conserved domain composition, taking into account both previously identified domain signatures in Pfam \[[@B12]\] and TIGRFAM \[[@B13]\] and any remaining potential novel domains identified in the Arabidopsisproteome using independent methods (see Methods). Our protocol differs significantly from the homology-based approach used to calculate paralogs for the Arabidopsiscomplete genome publication \[[@B1]\], which relied on BLASTP matches between proteins with an E value \<1e-20 and extending over at least 80% of the protein length. A benefit of our approach is that related families are easily identified by the fact that they share one or more domains. Using our domain-based protein classification and family construction methods, 18,641 (71%) of the Arabidopsisgene products are classified as members of 2,691 protein families \[[@B63]\]. On average, a family contains 7 members, although large families of kinases, transducins, zinc finger proteins, hydroxyproline-rich glycoproteins, myb family transcription factors, and cytochrome P450s are each represented by more than a hundred proteins and altogether comprise approximately 5% of the proteome. By contrast, the BLASTP method with single linkage clustering produces 18,260 proteins built into 3,142 families. A comparison between the results of protein family building using our domain-based classification scheme and our original BLASTP-based clustering approach is shown in Figure [3](#F3){ref-type="fig"}. Figure [3A](#F3){ref-type="fig"} shows that the distribution of protein families according to size produced by the two methods is quite similar overall. Figure [3B](#F3){ref-type="fig"} illustrates the differences in family sizes built by the two methods on a per protein basis. While most of the proteins in domain-based families are clustered into families of about the same size using single linkage clustering (SLC), this latter approach can produce anomalously large families. For example, the largest SLC family contains 2601 proteins. Using domain-based clustering (DBC) this same set of proteins resolves into 216 families ranging in size from 205 to 2 members. While the largest fraction of genes in the SLC family are protein kinases, other families such as cytochrome P450s, PPR-repeat proteins and calmodulins are included with each group, being linked by sequence similarity to only a subset of the other groups of proteins in the family. These families are well-resolved by the DBC method. Conversely, the SLC method can also produce fragmented families and singletons. This occurs where the functional domain covers only a small percentage of the overall protein size, as for example with many DNA binding and protein interaction domains. While the DBC method groups together proteins with these relatively small domains, the criteria of sequence identity and match length required by SLC is only fulfilled for small subsets of proteins within the domain-based families. For example, one DBC family of 151 members, which represents proteins with a single zinc finger (C3HC4-type RING finger) family domain (PF00097), is split by SLC among 32 families ranging in size from 14 to 2 members and 25 singletons. Clearly there is great diversity in this group of proteins that form a DBC family on the basis of a relatively short domain. However, this can be a useful grouping when no other information is available. The DBC method also over-fragments families under different circumstances. A set of paralogous proteins can contain some members that hit PFAM domains above the trusted cutoff, and some that do not because of divergence and/or lack of plant representatives in the PFAM seed. This results in the creation of Arabidopsis-specific domains that are, in effect, redundant with PFAM domains but are considered distinct, causing inappropriate fragmentation of families. For example, there are 17 proteins in a single SLC cluster that contain the \"seven in absentia\" (SINA) domain (PF03145), but two of these score just below the trusted cut-off. This results in the creation of 3 DBC families of 10, 5, and 2 proteins respectively. The Pfam domain profile can be retuned to include the missing Arabidopsisrepresentatives and remedy any over-fragmentation resulting from the insensitivity of the original domain profile (data not shown). Overall, close to 60% of clustered proteins fall into families whose sizes differ by fewer than 10 members between the two methods of family construction. The domain-based approach produces fewer, slightly larger families, and some anomalously large families are eliminated. ### Duplicated genes (segmental and tandem duplications) The large scale duplications of the Arabidopsisgenome have been extensively analyzed and documented (\[[@B64]-[@B66]\] and references therein). In addition to analyzing genes in the context of gene families, a further analysis of gene names was performed in the context of duplicated genes that may share similar or identical functions. Using approaches and criteria similar to those employed by others, we developed tools to facilitate the identification of segmental and tandem duplicated genes in our latest annotation (web resources at \[[@B67],[@B68]\]). We identified 6,582 protein-coding genes within the segmentally duplicated regions of the genome and 3,737 genes within tandem duplications some of which are found to be within the segmentally duplicated regions. In all, there are 9,533 presumed paralogous protein-coding genes, representing 36% of the Arabidopsisproteome. We then examined the functional annotation of these paralogous groups, verified the uniformity of their annotations and manually resolved any inconsistencies. ### Gene ontology In order to maximize the usability of the annotation data set, Arabidopsisprotein-coding genes were further classified using the controlled vocabularies of the Gene Ontology (GO) \[[@B69]\]. TIGR is a member of the GO Consortium \[[@B70]\], a collaborative international effort to organize and define gene products using standard, species-independent terminology. GO is now widely used in plant, animal and microbial genomics and has become one of the principal tools employed in the annotation of genes and their products \[[@B71]-[@B74]\]. GO consists of dynamic, controlled vocabularies describing three areas of biological systems: molecular function, biological process, and cellular component. Each GO annotation is required to contain an evidence code describing the type of evidence that supports it \[[@B75]\]. The evidence types used in manual GO curation range from direct experimental evidence and published inferences based on experimental data, to annotator inferences from examination of sequence and domain similarities. GO terms were assigned to Arabidopsisgene products based on similarity to functionally characterized proteins and/or functional domains. The majority of the ArabidopsisGO associations fall into the ISS category (inferred from sequence or structural similarity) since there was no published experimental evidence available. These inferences were made by assessing all of the similarity evidence available, including BLASTP results, HMM search results, Prosite and Interpro membership, protein family relationships, and similarity to other gene products having GO annotations. Proteins that were examined and had either weak or partial similarity to functionally characterized proteins were deemed to have too little evidence to warrant functional GO assignments and were given the GO term \"unknown\". This term exists so that annotators can capture the fact that they looked at the evidence available for a specific gene product and could make no assertion about the role this gene product might play in the organism. At TIGR, all GO assignments to Arabidopsisgenes were performed manually with emphasis on molecular function terms, but assignments to biological process and cellular component terms were added when they could easily be inferred from the evidence considered. This work was carried out in coordination with scientists at TAIR \[[@B76]\]. We regularly integrated the manual GO curation provided by TAIR into our dataset in order to minimize redundancy of effort between institutes. However, TAIR associations made automatically through purely computational methods were excluded from our dataset. Of the 49,505 distinct curated associations between 26,207 Arabidopsisgenes and GO terms in the final release, 6,424 associations were contributed uniquely by TAIR, 25,131 loci are annotated with at least one TIGR association, and 4,642 loci are annotated with at least one TAIR association, with 3,566 of these annotated by both centers. Leaving aside the specific GO category \"unknown\", 29,773 specific GO terms are assigned to 14,529 genes. Of these, 17,259 terms (assigned to 13,070 genes) are molecular function, 8,864 terms (7,111 gene assignments) are biological process, and 3,650 terms (3,257 gene assignments) describe cellular component. The GO function term \"unknown\" was assigned to all other genes after confirming the lack of other evidence. The decrease in the proportion of genes with a meaningful GO assignment (55%) compared with the number of genes given a functional assignment at the time of genome completion (69%; \[[@B1]\]) is most likely a reflection of the more rigorous and uniform standards applied during our whole genome reannotation effort As a result of the reannotation effort, each protein-coding gene in the genome has been manually assigned to at least one GO term (data available at \[[@B77]\]). Figure [4](#F4){ref-type="fig"} provides a summary of the current state of functional characterization of the Arabidopsisgenome. Among the most abundant functional role categories, 25 % of the genes are assigned catalytic functions including hydrolase, kinase, or transferase activity; 10 % bind nucleic acids, primarily DNA, including the 6.5 % categorized as transcription factors; 4.4 % are categorized as protein binding, many inferred from the presence of domains implicated in protein-protein interactions such as the RING Zn-finger \[[@B78]\] and leucine-rich repeats \[[@B79]\]; and 5 % are classified as transporters. Transposable element and pseudogene annotations ----------------------------------------------- Transposons and pseudogenes were the last categories of gene models to be systematically addressed by the re-annotation process. Many gene models with similarity to transposons or transposon-related proteins were originally annotated as protein-coding genes. However, the majority of these regions are degenerate, making it difficult or impossible to model ORFs across their entire extent, although shorter ORFs with similarity to parts of transposons may be contained within the boundaries. Thus, the legacy annotation for transposon-related sequences consisted of a mixture of genes and pseudogenes. In release 5.0, all transposon-related sequences were uniformly classified by searching the entire genome against a curated database of protein-coding transposon sequences \[[@B80]\] using the dps alignment utility of the AAT package and automatically applying the corresponding transposon family annotation. Each transposon-related region was defined by a single pair of coordinates and classified into one of the major classes of transposable elements as described in \[[@B81]\], shown in Table [4](#T4){ref-type="table"}. Release 5.0 contains 2,355 loci annotated as transposons, 1,652 matching retrotransposons and 703 matching DNA transposases and (in contrast to all previous releases) these are no longer included in the count of \"protein coding genes\" nor are they represented in that dataset. It should be noted that our transposon annotation has been restricted to elements with protein coding potential. Assimilation of the smaller elements and other classes of repeated sequences into the genome annotation remains a task for the future. Like transposons, pseudogenes are difficult to annotate accurately in an automated manner. Different gene prediction programs will often generate predicted gene structures that are dissimilar to each other and inconsistent with the homologous sequence alignments, introducing introns to circumvent frameshifts and premature stop codons. Pseudogenes are often detected during manual curation of these gene predictions, because the gene model cannot be modeled consistently with homologous protein alignments due to sequence degeneracy that results in stop codons that interrupt the open reading frame. Pseudogenes are often found in transposon-rich regions such as those associated with the pericentromeric regions. In our annotation, pseudogenes, like transposons, are described simply as a single pair of coordinates (5\' and 3\' ends) that span the genomic region in which they are found, and are classified on the basis of sequence homology to known proteins. In the current release, 1,431 loci are classified as non-transposon-related pseudogenes, of which approximately one third are similar to genes of known function. These include kinases, disease resistance proteins, ribosomal proteins, and others found in large gene families in Arabidopsis. The remaining pseudogenes are similar to proteins from Arabidopsisor other species that have no known function and likely represent degenerate genes of hypothetical proteins yet to be characterized. Like transposons, the majority of pseudogenes in the current annotation were named by an automated process. Conclusion ========== With respect to the annotation of gene structure and gene function, our reannotation effort has focused mostly on the protein-coding subset of all Arabidopsisgenes. This reflects a combination of community interest (knowing the entire gene repertoire of an organism) together with databases and gene prediction programs that are relatively effective in identifying and delineating such genes. Without a doubt, the largest contribution to improved gene structure annotation over the last three years has been the generation and release of FL-cDNA sequences by Ceres Inc. \[[@B17]\], by the RIKEN-SSP collaboration \[[@B19],[@B41]\] and by the INRA-Genoscope group \[[@B21]\]. However, because of the bias to annotate genes with presumed functional ORFs, there are likely many genes for regulatory and non-coding RNAs in addition to those already described \[[@B82]-[@B84]\] that remain to be discovered and incorporated into the annotation. Although the accurate annotation of transposable elements is important, our approach was simply to comprehensively identify regions of the genome with homology to transposon ORFs and to explicitly differentiate these from the remaining protein-coding plant genes. More work is needed in this area to improve the resolution and depth of annotation for these complex features, including the deconvolution of polyprotein ORFs, classification of complete, fragmented and degenerate elements, and delineation of repeat structures including long terminal repeats, direct repeats and insertion sites. With this final release from TIGR, primary responsibility for maintaining and updating the Arabidopsisannotation in North America has been assumed by TAIR. It can be anticipated that the annotation will continue to be both improved and enriched. One important distinction between the annotation processes at TIGR and at TAIR is that the former has been entirely sequence-based. This is to some extent historical but also reflects our philosophy that DNA sequence is a public, unambiguous and easily exchanged data type that can for the most part be incorporated into annotation using computational tools. Looking ahead, additional sequence information will permit the refinement of gene structures, while the functional annotation will be enriched both by the availability of new experimental data and by TAIR\'s policy of including results from expression and other kinds of analyses to characterize each gene and its function fully. Methods ======= The TIGR genome annotation pipeline, gene modeling and gene processing ---------------------------------------------------------------------- Prior to beginning our reannotation effort, we incorporated the remainder of the Arabidopsisgenome into our relational database (ATH1) as BAC sequences and annotations derived from the sequencing centers, the MIPS database, and GenBank. The annotation associated with these sequences provided the substrate for annotation improvements. Each BAC sequence was run through our eukaryotic annotation pipeline called Eukaryotic Genome Control (EGC). This pipeline consists of a series of steps during which bioinformatics tools are applied to the genomic sequence. The ArabidopsisEGC pipeline consists of a single Makefile run nightly on a Linux server. The Makefile runs a series of Perl scripts, each a wrapper around a bioinformatics tool responsible for launching an analysis (e.g. BLAST search), parsing the results, and loading the results into ATH1. The pipeline manages two primary tasks: processing the bare genome sequence and processing the individual genes and gene products. The genome sequence processing involves several aspects of gene identification and the gathering of evidence for gene structures. Statistical gene finders including GENSCAN+ \[[@B54]\], GeneMark.hmm \[[@B55]\], and GlimmerA \[[@B56]\] are run to gather gene predictions. The GeneSplicer \[[@B85]\] splice site prediction tool is also run to highlight potential splice sites along the genomic sequence. Transcript and protein spliced alignments provide our greatest resource for accurately identifying and modeling genes, often complemented by the gene predictions described above. We rely heavily on the AAT package \[[@B86]\] to identify genes and resolve gene structures using transcript and protein alignments, and this represents a primary component of EGC. While several other tools exist for generating spliced alignments between transcript sequences (ESTs and FL-cDNAs), including sim4 \[[@B87]\] and BLAT \[[@B88]\], they were not designed for aligning spliced transcripts of diverged species, but rather for accurately mapping near-identical transcript sequences. The AAT package (dds/gap2), although significantly slower than sim4 and BLAT, can generate alignments to divergent transcript sequences. The complete repertoire of TIGR Gene Indices, which includes 22 different plant species, were aligned to each of the ArabidopsisBACs at the nucleotide level using the dds-gap module of the AAT package, providing a great wealth of evidence for identifying conserved plant genes and resolving gene structure components (example shown in Figure [2](#F2){ref-type="fig"}). The AAT package also includes tools (dps/nap) for aligning related protein sequences to the genome, taking into account splice sites and resolving intron/exon boundaries via protein spliced alignments. TIGR\'s in-house non-redundant protein database (NRAA) was searched and aligned to the ArabidopsisBACs using this tool. The AAT package is available at \[[@B89]\]. Following genome sequence processing, the second stage of EGC -- individual gene processing -- begins. For the comprehensive reannotation of the Arabidopsisgenome, all the initial gene structure annotations were derived from the first pass annotation of the completed genome \[[@B1]\]. Each gene annotation is subjected to a series of bioinformatic analyses including: • WU BLASTP \[[@B90]\] search of NRAA. • Pfam \[[@B12]\] and TIGRfam \[[@B13]\] search using HMMER2 \[[@B91]\] • Search of PROSITE, PRINTS and ProDom, followed by Interpro classification including the results from the Pfam and TIGRfam searches using InterProScan \[[@B92]\]. • Transmembrane domain identification using TMHMM \[[@B14]\]. • Cellular localization prediction using TargetP \[[@B93]\]. • Signal peptide prediction using SignalP \[[@B94],[@B95]\]. To ensure that the gene-based searches always reflect the most current gene structure, genes that have been structurally altered during our reannotation were targeted each evening by EGC and reprocessed to gather the latest bioinformatics data. Computing protein families -------------------------- To identify domains in Arabidopsispeptides, the proteome was searched against Pfam and TIGRfam HMM profiles using HMMER2. Any sequence region scoring above the trusted cutoff assigned to the domain profile was designated as representing that domain. These domain sequences were then removed from the protein sequences and the remaining peptide sequences were searched against each other using BLASTP for subsequent clustering and alignment in order to identify potential novel domains not represented in the domain databases. Similar peptide sequences were clustered by creating a link between any two peptide sequences having an identity above 30% over an amino acid span of at least 50 aa. and an Expect value \< 0.001. The Jaccard coefficient of community \[[@B96]\] was calculated for each linked pair of peptide sequences aand bas follows: ![](1741-7007-3-7-i1.gif) with the Jaccard coefficient (J~a,b~), which we also refer to as the link score, providing a measure of similarity between the two proteins. The associations between peptides that had an insufficient link score were dissolved, and the remaining links were used to generate single linkage clusters. The clustered peptides were then aligned using ClustalW \[[@B97]\] and used to develop conserved protein domains not present in the Pfam and TIGRfam databases. A. thaliana-specific domain alignments containing five or more members were considered true domains for the purpose of building families. The peptides in alignments were searched back against the Arabidopsisproteome to seek out additional members that may have been excluded during earlier stages due to the parameters employed. Full length protein sequences were then grouped on the basis of the presence of Pfam/TIGRfam domains and potential novel domains. Proteins with exactly the same domain composition were then classified into putative protein families. The protein family classifications resulting from our analysis are available at \[[@B63]\]. Gene name curation protocol --------------------------- The following naming conventions were developed and followed with only rare exceptions: 1\. If a gene product had an identical match to a functionally characterized protein, then the gene product was given the name of the characterized protein. 2\. If the characterized protein had previously been given a symbol, the symbol was incorporated into the name in parentheses at the end of the name (e.g., holocarboxylate synthetase 1 (HCS1)). Note that the prefix \"At,\" for \"A. thaliana,\" present in some gene symbols in the literature, was omitted since it is redundant. When a functionally characterized protein had multiple names, or aliases, these were included, separated by \'/\' (e.g., phytochrome A specific signal transduction component (PAT3) / far-red elongated hypocotyl protein 1 (FHX1)). In most of these cases, the first name is typically the functionally characterized name followed by the original gene name. While there was a concerted effort to associate aliases and gene names to a particular locus, inevitably some names may have been missed. 3\. If a gene product was not functionally characterized but had a significant match to a functionally characterized protein and was thus believed to be functioning as that protein, then the gene product was designated as putative (e.g., arginine-tRNA ligase, putative). In most cases the Swiss-Prot name was used when there were naming inconsistencies. As in the naming for characterized proteins, aliases were included when they existed. 4\. When a gene product had a significant domain hit or partial yet significant characterized protein matches, or belonged to a characterized family but did not have significant homology to family members that had been functionally characterized, the protein was given the domain or family name and designated as a family member (e.g., DNA-binding family protein). In some cases, the significant domain hit did not imply a function for the gene product; these proteins were named for the domain, but designated as domain-containing proteins (i.e.: DC1 domain-containing protein). In cases where there were no significant domain matches and the gene product had either weak similarity or partial similarity to functionally characterized proteins, gene products were named as the protein but given a \"-related\" designation (i.e.: cysteine protease-related). 5\. Many gene products did not have significant matches to characterized proteins or domain hits and functionality could not be deduced. In such cases, a gene product supported by EST and/or cDNA evidence was designated as an expressed protein, while those supported by gene prediction only were designated hypothetical. Identification of duplicated genes within chromosome segmental duplications --------------------------------------------------------------------------- All-vs-all BLASTP searches were performed for the entire set of protein coding genes. These results were analyzed in the context of chromosome positions, applying a Waterman-Eggert-like alignment algorithm \[[@B98]\] to ordered gene lists. A Java based dot-plot viewer was developed to facilitate the identification and analysis of syntenic or duplicated regions inferred from BLAST matches between pairs of genes, providing rapid visualization and navigation of the data. The viewer includes user-specified filters to exclude matches based on the number of matches or E-value desired (software available at: \[[@B99]\]; note that the viewer has been subsumed by the DAGchainer distribution \[[@B100]\]). Using this tool, the list of tentatively duplicated gene pairs was refined and additional regions were identified manually. The curated list of segmental gene duplicates can be found at \[[@B68]\]. The data are mostly consistent with those reported previously \[[@B65]\]. Identification of tandemly duplicated genes ------------------------------------------- Tandemly duplicated genes were identified as described previously \[[@B1]\]. Neighboring genes were analyzed along each chromosome, and gene pairs having an E-value \<= 1e-20 and separated by not more than one unmatched gene were classified as tandem duplicates. An array of tandem duplicates was allowed to have only one unrelated member within the array. The list of tandem gene arrays can be found at \[[@B67]\]. Specification of sequence overlaps between adjacent BACs in the tiling path and chromosome construction ------------------------------------------------------------------------------------------------------- The tiling path for the Arabidopsisgenome describes the order and orientation of the BACs, YACs, cosmids and other pieces of DNA that collectively represent the sequence of the entire genome. To represent the BAC tiling path, we used a well-known data structure called a double ended queue. Each BAC was represented by a single node in the queue with pointers to the preceding and succeeding BAC. Each node contained additional attributes including the orientation of the BAC sequence, an indication of an overlap or gap between each adjacent BAC, the size of the overlap in base pairs, and the size of any terminal non-overlapping sequence from the overlapping regions to the BAC termini. Each node with pointers was described textually by a single row of a table which exists in ATH1, our Arabidopsisannotation database. Chromosome sequences were constructed by joining the regions of BAC sequences according to their orientation and position of overlap, envisioned as single in-silicorecombination events between the overlapping regions of BAC pairs. One of the major problems in building (and re-building) the composite sequence from the constituent BACs and other molecules is inconsistency of sequence between the two elements of the overlap. Part of this may be due simply to mutations in the BACs sequenced or to sequencing errors. These inconsistencies can lead to different models for the same gene on the two BACs and make merging of these inconsistencies into a single whole genome annotation very difficult to automate. To minimize the amount of poor quality sequence in the chromosome representations and to better automate future builds, we developed the concept of \"high quality overlap regions\" (HQORs). We define an HQOR as a genome sequence region found to align perfectly between two adjacent overlapping BACs. Candidate sequences to represent HQORs were identified using MUMMER \[[@B101],[@B102]\], and a provisional HQOR was chosen as the longest aligned region of perfect sequence identity. To verify the quality of the overlapping region flanking the provisional HQOR, the flanking regions were aligned and assessed using GAP \[[@B103]\]. If use of the provisional HQOR in the chromosome build would result in the incorporation of the model-corrupting base(s) into the sequence, the MUMMER alignments were re-examined and a different HQOR was identified, the use of which would circumvent this problem by shifting the point at which the recombination is made between the overlapping BAC pair. If the provisional HQOR resulted in long flanking sequences within the presumed overlap with low levels of identity suggesting an incorrect automated specification of the overlap, the MUMMER output was reexamined to identify other candidate HQORs that more accurately portray the tiling. This final step addresses potential problems caused by the presence of identical repeats near the ends of the BACs. After constructing each chromosome sequence from the BAC tilings, the coordinate positions of the BACs within the chromosome were utilized in order to copy all BAC annotations to the chromosome with the appropriate coordinates. The BAC tiling data as described here are included in our XML-based data release \[[@B22]\], and navigable from \[[@B104]\]. Authors\' contributions ======================= BJH was responsible for database content and many aspects of data analysis and drafted the manuscript. JRW carried out data analysis and data integrity checks, and assisted in user interface design and implementation and in the drafting of the manuscript. CMR, LIH, RKS, RM, MF and APC performed manual curation of gene structures and functions, including GO assignments and in manual evaluation of computational pipeline outputs. CY developed code for identifying HQORs and for building the pseudomolecules from BACs and for manually curating the tiling path for the entire genome. DW developed and implemented the protocols for generating Arabidopsis-specific protein domains and building paralogous families for facilitating annotation. ORW was involved in database design and the design and implementation of the annotation computational pipeline. CDT contributed to data acquisition, interpretation and analysis and to the drafting of the manuscript. All authors read and approved the final manuscript. Acknowledgements ================ We would like to thank all past members of the Arabidopsisannotation group at TIGR, the IT group led by Vadim Sapiro, the database managers Michael Heaney and Susan Lo, and many members of the bioinformatics staff especially Todd Creasy and Sam Angiuoli for their contributions and support. We gratefully acknowledge productive collaborations with our colleagues at MIPS and at TAIR and with the GO consortium. Finally, we are indebted to NSF for their long and continuous support of this project (Cooperative Agreement. DBI 9813586). Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### The Arabidopsisgenome as depicted in release 5 of the Arabidopsisgenome annotation. Each BAC sequence region within each chromosome is shown colored according to the original sequencing group. The unsequenced NOR and 5SrDNA clusters are colored black and centromeric regions are colored red, both with rounded edges and drawn to scale based on their estimated sizes. ::: ![](1741-7007-3-7-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Screenshot of the Annotation Station gene editor. The evidence for gene identification and gene modeling is viewed using proprietary software called Annotation Station, developed by Neomorphic and maintained now by Affymetrix. This tool, similar to Apollo that was developed at Berkley and Sanger \[105\], is used by human annotators as a genome navigation tool and gene structure modeling tool. The gene models, proteins and transcript alignments are shown for an approximately 4.5 kb window along the minus strand of BAC F10O3 in the region encoding the 3-methylcrotonyl-CoA carboxylase 1 (At1g03090). The curated gene structures are shown in dark green on the white background towards the bottom of the view, with exons filled, and introns and UTRs unfilled. Above this curation within the black background, evidence is shown from bottom to top as follows: splice site predictions, computational gene predictions, protein alignments shown in orange, EST alignments from searching the various plant Gene Indices in varied colors, regions of homology to the genome of Brassica oleraceashown in dark blue at the top of the view, and PASA Arabidopsistranscript alignment assemblies at the top shown in bright pink. The vertical marker line indicates the position of a skipped exon (supported by both PASA FL-cDNA and protein alignments) that results in two protein isoforms. ::: ![](1741-7007-3-7-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Distribution of proteins within families constructed using two distinct family building methods: our currently employed domain composition based clustering versus the single-linkage BLASTP-based clustering method originally described.A: Frequency distribution of family sizes created by the two methods. B: Difference between the two methods evaluated at the protein levelon a per protein basis. The difference in family size between domain-based clustering and the single-linkage clustering method (DBC -- SLC) was calculated for each protein that was included in a family using both methods. The histogram shows the total number of proteins found at each size difference displayed on the abscissa, binned at increments of 10. ::: ![](1741-7007-3-7-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### The distribution of genes in major categories of the Gene Ontologies. Each of the 26,207 protein coding genes was assigned to at least one GO term, with our primary focus the assignment of genes to Molecular Function terms. The ontology categories illustrated correspond to those of the plant GO slim obtained from <ftp://ftp.geneontology.org/pub/go/GO_slims/archived_GO_slims/goslim_plant.2003> ::: ![](1741-7007-3-7-4) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Statistics for Arabidopsisreannotation Release 5. ::: Chr. 1 Chr. 2 Chr. 3 Chr.4 Chr. 5 Total ---------------------------------------------------------- -------- -------- -------- -------- -------- --------- DNA molecules Length (Mb) 30.269 19.702 23.465 18.582 26.978 118.998 \%GC overall 35.9 35.9 36.3 36.2 35.9 36.0 coding 44.1 44.2 44.3 44.2 44.1 44.2 intronic 32.4 32.3 32.6 32.4 32.3 32.4 intergenic 30.8 31.4 31.6 31.6 31.1 31.2 Genes \# genes 6,772 4,104 5,233 3,985 6,113 26,207 gene density (kb/gene) 4.47 4.80 4.48 4.66 4.41 4.5 Avg. gene length (bp)^a^ 2,287 2,156 2,197 2,269 2,227 2,232 Avg. protein length 425 398 417 421 419 417 \# genes in protein families 4,834 2,884 3,803 2,839 4,281 18,641 \#genes duplicated via segmental chromosome duplications 1,868 961 1,315 1,147 1,291 6,582 \#genes found tandemly duplicated 993 545 750 636 813 3,737 \#genes with alt splicing isoforms 600 412 444 357 517 2,330 \#genes with annotated UTRs 4,717 2,936 3,575 2,724 4,147 18,099 \#transposons and pseudogenes 748 817 837 652 732 3,786 \# tRNA genes 240 96 93 79 123 631 Exons \# exons 37,710 21,428 27,937 21,800 33,255 142,130 total length (Mb) 10.378 5.919 7.812 6.011 9.170 39.290 avg exons/gene 5.57 5.22 5.34 5.47 5.44 5.42 avg exon size 275 276 280 276 276 276 Introns \# introns 30,938 17,324 22,704 17,814 27,191 115,921 total length (Mb) 5.060 2.903 3.657 3.016 4.416 19.053 avg size 164 168 161 169 163 164 Proteome \# distinct proteins 7,176 4,451 5,540 4,231 6,457 27,855 \# proteins with interpro domains 6,142 3,686 4,676 3,573 5,441 23,518 \# with TM domain 2,047 1,429 1,599 1,316 1,768 8,159 Signal peptides secretory 1,262 797 974 773 1,103 4,909 chloroplast 1,062 681 845 666 1,021 4,275 mitochondria 820 490 612 430 736 3,088 ^a^Length of genomic sequence from annotated transcriptional start to stop. ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Summary statistics for TIGR Arabidopsisannotation releases. ::: -------------------------------------------------------------------------------------------------------------------------------------------------------------------- Nature (12/00) Release 1 (8/01) Release 2 (1/02) Release 3 (8/02) Release 4 (4/03) Release 5 (1/04) ---------------------------------------------------- ---------------- ------------------ ------------------ ------------------ ------------------ ------------------ Genome size (Mb) 115.410 116.238 117.227 117.077 119.055 118.998 protein-coding genes 25,498 25,554 26,156 27,117 27,170 26,207 transposons and pseudogenes NA 1,274 1,305 1,967 2,218 3,786 Genes annotated as alternatively spliced NA 0 28 162 1,267 2,330 genes with UTRs NA 4,140 10,219 11,691 17,060 18,099 Protein-coding genes similar to transposon ORFs^a^ NA 487 485 528 531 6 gene density (kb per gene) 4.5 4.55 4.48 4.32 4.38 4.54 exons / gene 5.2 5.23 5.25 5.24 5.31 5.42 average exon length (bp) 250 256 265 266 279 276 average intron length (bp) 168 168 167 166 166 164 Gene structures altered since previous release.\ NA \- u: 2,853\ u:1,366\ u: 2,347\ u: 2,858\ (u,a,d,m,s) a: 690\ a: 1,906\ a: 527\ a: 1,393\ d: 231\ d: 221\ d: 143\ d: 730\ m: 14\ m: 62\ m: 153\ m: 169\ s: 167 s: 14 s: 24 s: 28 -------------------------------------------------------------------------------------------------------------------------------------------------------------------- Gene structure modifications from each previous release are represented by u: updated, a: added, d: deleted, m: merged, and s: split. ^a^Annotated protein-coding genes with a BLASTP match containing an E-value \<= 1e-20. ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Genes classified by alternative splicing variation. ::: Splice variation classification Genes with isoform type % alter protein sequence ------------------------------------- ------------------------- ------------------------------ Alternative acceptor and/or donor 1,050 70% Unspliced introns 926 67% Alternate terminal exons 99 28% Exon skipping 130 68% Start or end within intron 520 47% ::: ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Transposon classification. ::: Transposable element classification \# Annotated genomic regions --------------------------------------------- ---------------------------------- Class I (Retrotransposons) 1652 gypsy-like retrotransposon family (Athila) 511 gypsy-like retrotransposon family 374 copia-like retrotransposon family 494 non-LTR retrotransposon family (LINE) 264 other 9 Class II (DNA transposons) 703 hAT-like transposase family (hobo/Ac/Tam3) 77 CACTA-like transposase family (En/Spm) 69 CACTA-like transposase family (Ptta/En/Spm) 127 CACTA-like transposase family (Tnp1/En/Spm) 37 CACTA-like transposase family (Tnp2/En/Spm) 102 Mutator-like transposase family 268 Mariner-like transposase family 9 other 14 :::	PubMed Central	2024-06-05T03:55:55.679544	2005-3-22	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1082884/", "journal": "BMC Biol. 2005 Mar 22; 3:7", "authors": [ { "first": "Brian J", "last": "Haas" }, { "first": "Jennifer R", "last": "Wortman" }, { "first": "Catherine M", "last": "Ronning" }, { "first": "Linda I", "last": "Hannick" }, { "first": "Roger K", "last": "Smith" }, { "first": "Rama", "last": "Maiti" }, { "first": "Agnes P", "last": "Chan" }, { "first": "Chunhui", "last": "Yu" }, { "first": "Maryam", "last": "Farzad" }, { "first": "Dongying", "last": "Wu" }, { "first": "Owen", "last": "White" }, { "first": "Christopher D", "last": "Town" } ] }
PMC1082885	Background ========== Multidecadal fluctuations, or biological regime shifts, have basin-wide effects on sea surface temperature (SST) and upwelling that are associated with large-scale changes in biological productivity \[[@B1],[@B2]\]. They are similar to the 3--4 year cycles of the El Niño Southern Oscillation (ENSO) but on longer time scales, and as background states, may amplify ENSO effects \[[@B3]\]. The Pacific Decadal Oscillation (PDO), which addresses the North Pacific/North American sector, comprises events that persist for 20 to 30 years \[[@B4],[@B5]\]. In the 20^th^century, \"cool\" PDOs prevailed from 1947 to 1976 while \"warm\" PDOs were recorded from 1925 to 1946 and from 1977 through at least the mid-1990s. Chavez and colleagues \[[@B6]\], focusing on the last two regimes, report that in the mid-1970s the Pacific changed from a cool \"anchovy regime\" (ocean temperatures colder than normal, shallow thermocline and strong upwelling, increased nutrient supply and zooplankton and overall productivity, high catches of anchovies and salmon, low sardine abundance) to a warm \"sardine regime\" (ocean temperatures warmer than normal, deep thermocline and weak upwelling, decline in zooplankton and salmon and overall low productivity, high sardine abundance, low anchovy catch), which lasted until the middle to late 1990s before shifting back to an anchovy regime. The changing ocean temperatures associated with these regimes exert strong effects on marine food webs \[[@B7]\]. In the northeastern Pacific many marine birds and mammals flourish during anchovy regimes and are nutritionally stressed during sardine regimes. It is well documented that short-cycle ENSOs cause drastic reductions in survival and reproduction of several marine mammal species in the temperate waters of the Pacific \[[@B8]\]. Non-migrating sea lions and fur seals are especially vulnerable to ocean warming because nursing females alternate feeding and nursing, which ties them closely to their coastal rookeries and results in pup starvation when prey move out of the area \[[@B9]\]. It is not known, however, to what degree these processes impact deep-diving marine mammals that obtain their prey far below the surface as they range widely in the open ocean of the northeastern Pacific far from the coast. The condition of newly weaned northern elephant seals, Mirounga angustirostris, is expected to be sensitive to these processes because of the location where adults feed and the close link in energy exchange between mother and pup. In this species weaning weight reflects total energy transfer from the mother because pups feed solely on mother\'s milk prior to weaning. Because the mother gives birth annually to a single pup and fasts during lactation, her pup\'s weight at weaning correlates positively with her resource accruing ability at sea during the eight-month period of pregnancy \[[@B10],[@B11]\]. The mean weaning weight of the colony reflects the general foraging success of parturient females in that year. It follows that mean weaning weight may oscillate annually with ocean processes that exert strong effects on marine food webs and impact prey availability. A similar logic has been used in studies of the southern congener, M. leonina, that breeds and feeds in Antarctic waters \[[@B12]-[@B16]\] Northern elephant seals are among the deepest diving marine mammals. Females disperse widely in the northeastern Pacific twice a year, from 38 to 60°N and from the coast to 188°E (Fig [1](#F1){ref-type="fig"}), foraging at modal depths of 300--600 m and maximum depths of 1500 m \[[@B17]\] on prey such as mesopelagic squid and Pacific hake, Merluccius productus\[[@B18]\]. After continuous foraging over the 8-month gestation period (mid-April to mid-December), females return to rookeries along the coasts of Mexico and California to give birth and nurse their pups. In four weeks a pup triples its birth weight and is weaned when the mother returns to sea in February for a 2 1/2 month post-partum bout of foraging to recover the one third of the body mass lost during nursing and begin the next pregnancy \[[@B10],[@B19]\]. Milk energy transferred to the pup varies with its sex and the age, mass, blubber reserves and overall condition of the mother \[[@B11]\]. By maternal condition, we mean the total body energy available to the female at the time of parturition. On average, 48.0 ± 3.0% (3490 ± 490 MJ) of a female\'s body reserves is depleted during lactation. The size and blubber reserves obtained by the female during her biannual foraging migrations determine the level of reproductive expenditure in the subsequent breeding cycle. Pup weight at birth and weaning is correlated positively with the mother\'s age and mass \[[@B19]\]. Male pups are about 8% heavier than females at birth and at weaning \[[@B20]\], suggesting that males require more maternal energy than female pups. Pregnant females are capital breeders that store energy in blubber laid down over months of continuous foraging at sea and expend much of this energy later in reproduction \[[@B21]\]. Acquisition of insufficient body reserves owing to low foraging success directly impacts reproductive expenditure resulting in low weaning weight of pups \[[@B11]\]. Additional data supporting these points are reported in studies of southern elephant seals \[[@B13],[@B22]\]. Here we show that annual weaning weights of northern elephant seal pups and resource accrual of adult females associated with a rookery in central California fluctuate with ocean temperature cycles. Results ======= Weaning mass ------------ The pattern of fluctuation in weaning weight over the years was similar in both sexes even though males were, on average, 5% heavier than females (overall mean of 131.7 ± 24.8 kg for males and 125.4 ± 21.4 kg for females). The regression of weaning mass on year showed a statistically significant decline in both males (F = 9.59, df = 1, 25, P = 0.0048) and females (F = 7.50, df = 1, 25, P = 0.0112). The regression slopes of males and females were not significantly different from each other (t = 0.11, df = 50, P \> 0.05). The weaning mass data of both sexes, combined, which we address in the rest of this paper, declined significantly from 1975 to 2004 as illustrated by the scatter plot and regression of weaning mass against year in Fig. [2a](#F2){ref-type="fig"} (F = 19.7, df = 1, 25, P = 0.0002). A Loess smoothing plot, however, reveals two distinct slopes (Fig. [2b](#F2){ref-type="fig"}). Mean weaning weight declined significantly from 1975 to the mid to late 1990s, decreasing from highs of 141 kg in 1975 and 146 kg in 1977 and 1980, to lows of 117, 118 and 115 kg in 1993, 1995 and 1999, respectively. A linear regression line through the data over the 24 year period from 1975 to 1999 (Fig. [2c](#F2){ref-type="fig"}) shows a significant mean decline of 0.89 kg/yr (F = 38.5, df = 1, 20, P = 0.0001). A significant decline is also observed if the end data point is anywhere in the interval 1995--1998. Similarly, the decline is statistically significant when years with a sample size of less than 30 are omitted, i.e. 1975, 1976, 1977 and 1982 (F = 25.93, df = 1, 16, P = 0.0001). In the late mid 1990s, the declining trend ends abruptly and the mean weaning weight increases up to 126 to 130 kg in the new century (Fig. [2b](#F2){ref-type="fig"}). Although the slope of weaning weight from 1999 to 2004, shown in Fig. [2c](#F2){ref-type="fig"}, is positive, it is not statistically significant (F = 2.12, df = 1, 4, P \> 0.05) owing in part to the small sample size. The slope of weaning mass versus year, however, varied significantly between the two regimes (ANCOVA, F = 5.21, df = 1, 19, P = 0.03). The period of declining weaning weight was associated with higher than normal ocean temperatures associated with the long-term sardine decadal cycle and the prevalence of El Niños in the short-term ENSO cycles (Fig. [3](#F3){ref-type="fig"}). Weaning weight declined increasingly with the tenure of the sardine regime, a trend that ended abruptly with the shift to the anchovy regime in the late 1990s. With the regime shift, a reverse trend to increasing weaning weights is evident. Mean weaning weight between the two regimes was not significantly different (mean sardine period = 129 ± 8.2 kg vs mean anchovy period = 127.6 ± 2.3 kg) (t = 0.38, df = 23, P \> 0.05) but this analysis is confounded by the declining slope of weaning mass over the course of the sardine regime and the small and incomplete sample size for the anchovy regime. A similar lack of association was found between weaning mass and PDO. Declining weaning weight over the course of the decadal regime from 1975 to the late 1990s is consistent with the overlapping pattern of short-term ENSO cycles during this period, represented in Fig. [3](#F3){ref-type="fig"} by annual mean monthly values of the standardized Southern Oscillation Index (SOI) \[[@B23]\]. This was clearly a period of extended warmth in the Pacific. Warm El Niños were more frequent than cold La Niñas, the El Niños of 1982--83 and 1997--98 were the most intense of the century, and the El Niño in the early 1990s (1991--1995) was unusually long. The global temperatures in 1997 and 1998 were the warmest years on record \[[@B24]\]. The lowest weaning weights were observed from 1993 to 1999, an unusually warm period dominated by two El Niños coming at the end of the sardine regime. The decline in weaning mass was not linearly related to increasing ocean temperatures during the sardine regime. Although the period from 1975 to 1999 was warmer than usual, ocean temperatures did not increase significantly over the course of this period, as revealed by SOI and PDO annual averages regressed on year (F = 0.02, df = 1, 27, P \> 0.05 and F = 0.08, df = 1, 28, P \> 0.05, respectively). Consequently, a linear regression plot reveals that annual mean weaning mass was not closely associated with annual SOI (F = 0.13, df = 1, 24, P \> 0.05) or annual PDO averages (F = 0.45, df = 1, 25, P \> 0.05). It is notable, however, that the mean weaning weight over the period 1990--1999, which included the longest El Niño and the most intense one of the century, was 122.8 ± 4.8 kg, significantly lower than the mean weaning weight of 136.4 ± 7.2 kg recorded in the period 1975--1985, which included only the strong El Niño of 1982--1983 (t = 4.78, df = 16, P = 0.0002). Female foraging success: foraging duration and mass gain -------------------------------------------------------- To examine the relationship between weaning weight, ocean temperature and female foraging success, we used the spring post-breeding foraging trip duration, and mass gain over this period, as the metric of foraging success. During this foraging trip females can recover approximately 85% of the mass lost during parturition and nursing. We assume that foraging success in the spring is associated with and reflects foraging success in the fall. The duration of the fall foraging trip, however, is an unsuitable measure of foraging effort because its length is set by gestation, and is relatively invariant at 225.7 ± 4.3 days (N = 10). The duration of spring foraging trips of females was inversely related to mass gained (Fig. [4a](#F4){ref-type="fig"}). The regression of mass gain on foraging duration was statistically significant (F = 22.3, df = 1, 10, P = 0.0008). Over the course of the sardine regime, post-partum females spent increasingly more time attempting to accrue resources \[[@B25],[@B26]\], increasing the mean foraging trip duration by approximately 36% from the late 1970s to the mid to late 1990s (Figs. [3](#F3){ref-type="fig"}, [4a](#F4){ref-type="fig"}). With the shift to the anchovy regime in the late 1990s, foraging durations decreased from 90 days or more at sea to approximately 70 days at sea from 2001 to 2003, the same level observed at the beginning of the sardine regime. Association of foraging trip duration with shorter term El Niño cycles is also evident, such as the longest foraging trip and lowest mass gain occurring in 1998, an exceptionally strong El Niño year. Mass gain during foraging trips was inversely related to trip duration (Fig. [4a](#F4){ref-type="fig"}), decreasing from approximately 1 kg/day at sea during the late 1980s to approximately 0.7 kg/day in 1993--1996, to a low of 0.29 ± .36 kg/day in 1998, and returning to 1 kg/day in 1999. In spite of small annual sample sizes, variation around the overall mean was small (mean = .90 ± 0.23 kg/day, n = 52). Weaning weight of pups was inversely related to the foraging trip duration of females (Fig. [4b](#F4){ref-type="fig"}). Linear regression of pup weaning mass on female foraging duration shows that this relationship was statistically significant (F = 10.10, df = 1, 22, P = 0.0044). Weaning weight and mass gain of females covaried in time (Fig. [4c](#F4){ref-type="fig"}) but the sample size was small and the association was not statistically significant (F = 2.87, df = 1, 10, P \> 0.05). Colony composition and sex ratio -------------------------------- The Año Nuevo colony increased over the course of the study period from 605 pups produced in 1975 to 2500 pups produced in 2004. The decline in weaning mass from 1975 to 1999 was not associated with changes in the age composition of the colony such that more young females gave birth to smaller pups. Rather, the opposite was observed; mean age at primiparity of females increased over the study period from a mean of 3.7 years in the early 1970s \[[@B27]\] to over 4 years in the 1990s. The decline in weaning mass was not due to a sex ratio increasingly biased to smaller female pups. There were marginally more females in the sample from 1977 to 1987 (51% of 583 pups), when mean weaning weight was highest, and more males in the sample from 1990 to 1999 (51.6% of 1278 pups) when mean weaning weight was lowest. The sex ratio of weaned pups in the entire colony has not deviated significantly from unity during the last four decades \[[@B20]\]. Discussion ========== Our data show that the weaning mass of northern elephant seal pups varies with ocean temperature cycles. Weaning mass declined over the course of an unusually warm 24-year period that coincided with the warm decadal or sardine regime and included several long and strong short-term warm ENSO events. The decline in weaning mass ended abruptly with the regime shift to the cold anchovy period. Weaning mass increased with the return to the cold regime but monitoring throughout an entire anchovy period is required to reveal the entire pattern and determine whether the positive trend of increasing mass continues. The link between ocean temperature and weaning weight is strengthened by the finding that weaning mass was inversely related to the foraging effort of females and positively related to the mass they gained. Over the course of the warm period from 1975 to about 1999, females spent increasingly longer periods searching for food; the more time spent searching, the less mass they gained. Evidently, females attempted to compensate for reduced foraging success by staying at sea longer searching for food. This would buffer the effect of reduced foraging success and would obviously complicate any expectations of a linear relationship between SOI and weaning mass and rates of female mass gain and pup weaning mass. The ability of a given female to extend her pre-implantation foraging trip, however, is limited by her need to implant in time to give birth during the breeding season and is impacted by her weaning date the previous year. Females do not have the ability to compensate in this way during the gestational foraging trip. The relationship between SOI and weaning mass is further confounded by changes in colony age distribution. As the colony has aged, weaning mass should increase based on our understanding of the relationship between maternal mass and investment \[[@B11]\]. SOI helps define the strength and duration of ENSO cycles but it does not predict weaning mass of pups well except during unusually strong or long El Niños. We emphasize that it is important that we detected such a strong trend in pup weaning mass despite the trip duration compensation by females and the changing colony demographics. This implies that reduction in resources were greater than could be compensated for by increased foraging duration in any given year or by increasing maternal body size with age. By this logic, it is reasonable that we did not observe a linear relationship from SOI and PDO to rate of foraging success to pre-implantation trip duration to weaning mass. It is clear that foraging effort was greatest and foraging success was lowest at the end of the sardine regime during the late 1990s. With the return to the cold regime in the late 1990s, female foraging effort returned to the level observed in the mid-1970s, the time of the previous regime change. We reason that warmer than usual ocean temperatures influenced elephant seal prey, either through recruitment, competitive advantages or predation, effecting a redistribution or diminution of populations. As a result, female seals were less successful forging, and consequently, at parturition and during lactation, they had less energy to transfer to their young. This logic is supported by observations during El Niño years. During the strong El Niño of 1997--98, foraging females spent more time travelling from one prey patch to another and they spent less time on each prey patch than in non-El Niño years \[[@B26]\]; mass gain in 1998 was by far the lowest value we have recorded. Moreover, enduring effects of these \"hard times\" are suggested by the observation that female pups born in 1983, a strong El Niño year, postponed breeding for the first time relative to adjacent cohorts in the years 1977 to 1986 \[[@B25]\]. The link between seal condition and physical measurement of ocean temperature is complicated. The progressive decline in weaning mass over the course of the sardine regime is not simply explained by increasing ocean temperatures. The SOI and PDO indices did not increase systematically over the course of this period. Evidently, other factors such as the duration of exposure to warming or warm temperature spikes during ENSO cycles overlaid on a warm decadal cycle may have been operative. Anthropogenic perturbations, or global warming, may amplify the effect of these natural cycles \[[@B28]\]. In any case, the temporal link between temperature changes, effects on prey, and the weaning mass of pups is unclear. Substantial lag effects may operate, as has been observed in the link between the PDO index and salmon catches \[[@B4]\]. Moreover, female seals spend 10 of the 12 months of the year foraging and it is not clear whether foraging is affected during all or part of this time. An additional complexity is the difficulty of separating the effects of short-term ocean temperature cycles from the long-term cycles. For example, in addition to the long period of warm water associated with the sardine regime, an unusually long El Niño occurred during the early 1990s that had a more severe impact on marine mammals and birds than was predicted from its strength. Overall, the data suggest that the duration and persistence of warm water may have a cumulative negative effect on seal prey and indirectly on seal foraging behaviour and, ultimately, on pup weaning mass. It is also likely that the long-term cycle, as background state, may amplify or dampen the effect of short-term cycles, e.g. the impact of an El Niño on seal prey and foraging may be amplified during a sardine regime and dampened during an anchovy regime. Monitoring weaning mass and foraging behaviour over the entire course of the current cold anchovy regime may shed light on these issues. This study shows that understanding the relationship between ocean climate and biological processes depends on having a long and complete time-series. In the present study, single year records yielded estimates of weaning mass ranging from 147 to 115 kg, a 28 % change from lowest to highest estimate. Even samples of five consecutive years would have provided quite disparate estimates, e.g. means of 140, 130 and 121 kg for the years 1975--1980, 1985--1990 and 1995--2000, respectively. This study, ranging over 29 years, not only provides a more valid and reliable estimate of weaning weight of northern elephant seals from central California, it also presents a wide range of observed values and the circumstances associated with this variability. Even so, the present study suggests that representative sampling of pup weaning weight should cover at least two successive multidecadal regimes, about 50 continuous years! Conclusion ========== Elephant seals, one of the deepest divers in the ocean and a migratory mammal that spends 83% of the year feeding far from the coast in pelagic waters beyond the thermocline in the modal temperature range 4.2--5.2°C \[[@B29]\], are impacted by ocean warming. Females spend more time searching for prey and are less successful in acquiring it when ocean temperatures are warmer than usual. Body stores laid down during long foraging periods at sea set limits on energy transfer to pups. Mean weaning mass of pups, consequently, fluctuates from year to year and reflects the impact of cyclic warming on the availability of elephant seal prey in the northeastern Pacific in that year. Weaning mass declined significantly during the unusually warm period from 1975 to 1999. It is well documented that offspring mass at the time of nutritional independence affects the probability of survival in many animals and is a crucial variable in population growth \[[@B22],[@B30]-[@B32]\]. It is notable that the decline in elephant seal weaning mass we report here paralleled in time the 60 to 80% decline of Steller sea lions, Eumetopias jubatus, northern fur seals, Callorhinus ursinus, and non-migratory harbor seals, Phoca vitulina, in the northern Gulf of Alaska that began in the late 1970s and only abated to approximately 5% in the late 1990s \[[@B33]-[@B36]\]. Our results provide additional support for the hypothesis that nutritional stress \[[@B37]\] is a partial explanation for population declines such as these. Lastly, our findings support the assertion that monitoring aspects of ecosystems, in addition to monitoring climate data, facilitates early identification of regime shifts \[[@B38]\]. Methods ======= Weanling measurements --------------------- We weighed 2750 pups within 10 days of weaning. Mean sample size of all years in the interval 1975--2004 was 108.6 ± 68.2 and the range was 5--224. Sample sizes were smallest in the years 1975 (6), 1976 (5), 1977 (14) and 1982 (5). In every other year the sample size was 30 or more. Weanlings were not weighed in 1979, 1981 or 1983. Selection of individuals for weighing was unbiased and based primarily on convenience with one exception. Pups weighing less than 50 kg were excluded from the analysis because they were orphaned before weaning, suckled briefly and were only 10 kg above mean birthweight \[[@B20]\], were in poor physical condition and injured, and were judged unlikely to survive to four months of age \[[@B39]\]. They were not excluded because their mothers had insufficient resources to nurture them. We weighed all pups on the peripheries of female harems to minimize disturbance to the rookery. We attempted to weigh weanlings as soon as they were accessible to the weighing team; pups usually exited harems within a day or two after weaning. Each animal was captured and weighed in a nylon restraint bag lifted by a winch attached to a scale suspended from an aluminium tripod. Scale precision was ± 1 kg. Since weanlings fast and lose mass, weaning mass was estimated based on known rates of mass lost per day. We used the following equation for back-calculating mass at weaning: Mass at weaning = Measured mass·(e^k·d^), where k = .00596 and d = number of days between weaning and weighing \[[@B19]\]. Estimates of foraging success ----------------------------- Foraging trip duration of 410 adult females was measured annually from 1976 to 2003 (except for 2000) during the post-partum spring foraging trip. Mean sample size was 15.2 ± 12.2 per year (range = 3--54). We recorded departure time from the rookery and return to it by identifying individuals from a combination of dye and bleach marks, VHF transmitters and satellite tags. During the period 1985 to 1999, mass gain of 52 females (3 to 6 per year) was determined over the spring period at sea by weighing the females on the rookery before they departed and when they returned \[[@B21]\]. We determined the foraging trip duration of 10 adult females during the fall migration, which encompasses the entire gestation period. Colony demographics ------------------- Each year we determined the sex ratio of pups produced. We monitored changes in colony size and age composition with weekly censuses, identifying animals flipper-tagged at weaning \[[@B40]\]. Ocean temperature ----------------- Thermal dynamics of the ocean were estimated from monthly Southern Oscillation Index (SOI) values \[[@B41]\] and from the Pacific Decadal Oscillation (PDO) Index \[[@B42],[@B43]\]. Statistical analysis -------------------- We used the nonparametric Loess plot in the Axum 7 software package (MathSoft Engineering and Education, Inc., Cambridge, Massachusetts) to identify slope changes and smooth the scatter plots of annual fluctuations in pup weaning mass, female foraging duration and female mass gain \[[@B44]\]. This method uses locally weighted linear regression to smooth data. The span, which determines the bandwidth as a fraction of the range of the x-axis values of the data, was 0.5. Local linear fits were used and the weight function was symmetric. We present the actual P values for tests that were statistically significant. The alpha level (probability of Type I error) was set in advance at 0.05. Authors\' contributions ======================= BL conceived the study, established data collection protocols, analyzed data, drafted the manuscript, and revised the manuscript after peer review. DC participated in data collection, data analysis and the final revision. Both authors read and approved the final manuscript. Acknowledgements ================ We thank colleagues and numerous students for assistance in weighing pups and tracking females, especially J. Reiter, D. Costa, P. Morris, P. Thorson, G. Oliver, R. Condit and D. Noren. We thank M. Sylvan for statistical advice and Francisco Chavez for comments on an early draft. This work was funded by grants from the National Science Foundation and gifts from George A. Malloch. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Foraging location of adult females.Rectangular mercator projection of spring (yellow) and fall (red) migration tracks of 35 satellite-tagged female elephant seals from Año Nuevo, California (black lined square), during the period 1995--1999. Adapted from previous studies \[17,26\]. ::: ![](1741-7007-3-9-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Fluctuations in annual weaning mass of pups.a) Least squares regression shows a significant decline in weaning mass over the entire study period. b) A Loess plot smoothes the data and reveals two functions, a declining slope and an increasing slope with the breakpoint between the two occurring in the late 1990s. c) A Least squares regression line reveals a significant decline in weaning mass from 1975 to 1999 (solid circles); the regression line from 1999 to 2004 (open circles) shows increasing weaning mass values. The dotted lines are the 95% confidence limits in kilograms. ::: ![](1741-7007-3-9-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Weaning mass in relation to ocean temperature cycles.A Loess plot (blue line) of mean cohort weaning weight (closed circles) in relation to the sardine (thick red line) and anchovy (thick blue line) decadal regimes and the Southern Oscillation Index (SOI) showing colder than normal La Niñas (black) and warmer than normal El Niños (red). ::: ![](1741-7007-3-9-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Foraging effort and foraging success of adult females.a) Loess plots of foraging trip duration of females (blue line, closed circles) show an inverse relationship with mass gain at sea (red line, open circles). b) A Loess plot shows that weaning mass (blue line, closed circles) is inversely related to foraging trip duration of females (red line, open circles). c) Weaning mass (blue line, closed circles) covaries with mass gain of adult females (red line, open circles). ::: ![](1741-7007-3-9-4) :::	PubMed Central	2024-06-05T03:55:55.688423	2005-3-28	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1082885/", "journal": "BMC Biol. 2005 Mar 28; 3:9", "authors": [ { "first": "Burney J", "last": "Le Boeuf" }, { "first": "Daniel E", "last": "Crocker" } ] }
PMC1082904	Background ========== Chickens are known to produce several different proteins which bind biotin in a non-covalent fashion. One of them is avidin, which is expressed by oviduct cells upon progesterone induction and is then transferred to the egg-white where it constitutes a minor fraction of the total protein content of the egg-white \[[@B1]\]. Independently of progesterone, avidin expression is also induced by an inflammation response in almost all of the studied chicken tissues \[[@B2]\]. Another biotin-binder, called literally biotin-binding protein (BBP), is presumably induced by estrogen \[[@B3]\] and secreted from the liver into chicken plasma \[[@B4]\]. From plasma, the BBP is thought to be deposited in egg-yolk \[[@B5]\]. In addition, Seshagiri and Adiga found another egg-white BBP, distinct from avidin, the biochemical characteristics of which resemble those reported for yolk BBP \[[@B6]\]. The affinities that avidin and yolk BBP exhibit toward biotin are extremely high, the dissociation constant being femtomolar for avidin \[[@B1]\] and picomolar for BBP \[[@B7]\]. According to the published data, the yolk BBP serves as a biotin reserve for the developing embryo and hence it is saturated with the vitamin \[[@B5]\]. In contrast, avidin in egg-white is mainly found as an apoprotein and it is assumed to function as an antimicrobial agent that harvests free biotin from its environment \[[@B1],[@B2]\]. Because of its high affinity to biotin, avidin has long been used as a separation, labelling and targeting tool in various bioscience fields \[[@B8]\]. The yolk BBP has been further characterised to consist of two different forms, BBP-I and BBP-II \[[@B3],[@B9]\]. It has been proposed that BBP-I is the primary gene product (67 kDa) that is converted by proteolytic cleavage to BBP-II (19 kDa) \[[@B4]\]. The biological function of BBP-I is believed to be a general biotin transporter in plasma, whereas the actual deposition role in egg-yolk is reserved for BBP-II \[[@B3]\]. BBP-II is a tetrameric protein, like avidin, and is composed of subunits homologous to each other. BBP-I has been thought to be a pseudotetramer containing four binding domains in a polypeptide chain and its gene should, therefore, contain four subsequent repeats encoding for similar peptide sequences \[[@B4],[@B9]\]. The egg-white BBP was also reported to exist in a large form similar to yolk BBP-I \[[@B6]\]. Interestingly, some egg-laying species, such as turkeys and alligators, showed only one type of BBP in the yolk of their eggs \[[@B10]\]. Despite their similar function, some biochemical properties of BBPs and avidin are different. The pI of the chicken yolk BBP (not defined which form) was reported to be 4.6 \[[@B7]\] in contrast to avidin which has a highly basic pI (≈ 10.4) \[[@B1]\]. BBP-I exhibits higher thermostability, being active at 60°C, whereas BBP-II is denatured at temperatures above 40°C \[[@B10]\]. Bush and White III have published the N-terminal amino acid sequences for both chicken yolk BBP forms, which are highly similar to each other and also resemble the avidin N-terminal sequence \[[@B4],[@B10]\]. BBP-I has been proposed to be a glycoprotein \[[@B7]\] whereas BBP-II is shown to be nonglycosylated \[[@B10]\]. Avidin is known to contain one N-linked carbohydrate moiety per subunit \[[@B1]\]. Differences in the radiobiotin exchange rates between these two BBP forms have also been observed: BBP-I showed slower exchange than BBP-II \[[@B10]\]. The published N-terminal sequences, the similar overall sizes of the proteins and the tetrameric appearance of BBP-II as well as the reported high biotin-binding affinities suggest that the BBPs could be related to avidin. Structurally and functionally, avidin is considered to be a member of the larger protein superfamily called calycins \[[@B11]\]. These proteins form a large and divergent family of relatively small extracellular proteins which typically bind small hydrophobic ligands. Two particular groups of calycin protein family, the lipocalins and avidins, are β-barrels composed of eight antiparallel β-strands. The ligand is bound inside the protein at one end of the β-barrel \[[@B12]\]. An interesting feature among the lipocalins and avidins is a structural signature wherein a conserved basic amino acid residue, close to the last β-strand, packs over a specific tryptophan residue on the first β-strand and forms hydrogen bonds with the short 3~10~helix prior to the first β-strand \[[@B11]\]. Because the genes, genomic locations, cDNAs or full amino acid sequences of chicken BBPs are not known, it is impossible to evaluate their true relationship to avidin or, in a broader sense, to the calycin protein superfamily. New hope to solve this enigma aroused when the first draft of the chicken genome was published in March 2004 \[[@B13],[@B14]\]. In addition to the genome project, a comprehensive collection of chicken cDNAs is also in progress \[[@B15]\]. In the current study we searched these databases in order to find the cDNAs and genes for BBPs. Indeed, we found two independent cDNAs whose translated amino acid sequences fitted well to the published N-terminal sequences of BBPs. The genomic fragments corresponding to these cDNAs were also identified and analysed. They showed features similar to those of the avidin gene family members. Interestingly, one of these BBP gene candidates is located together with the avidin gene family in the chicken chromosome Z \[[@B16]\]. In addition, more evidence supporting the previous hypothesis of the high recombination frequency in the avidin gene family is gathered. One of the two putative BBPs was found to significantly resemble avidin, showing a theoretical molecular mass and pI close to those of avidin, whereas the other showed theoretical characteristics fitting more closely to those published for BBPs. Almost all amino acids important for biotin binding in avidin \[[@B17]\] are conserved in both of these supposed BBPs. Neither of the found cDNAs/genes, however, encodes a protein composed of four similar domains as expected for the isolated pseudotetrameric BBP-I \[[@B3]\]. Instead, the encoded proteins show calculated molecular masses corresponding to one BBP domain per polypeptide. In silicoanalysis of these genes as well as modelled structures of the putative BBP proteins are presented. Results ======= Database queries and sequence analyses -------------------------------------- The database searches revealed three cDNA hits which showed significant similarity for the used N-terminal query sequences. The public accession codes for these hits are \[GenBank:BX930135\], \[GenBank:BX932076\] and \[GenBank:BX936151\]. However, detailed analyses suggest that one of the hit sequences, \[GenBank:BX932076\], is identical to sequence \[GenBank:BX936151\] but differentially spliced (first intron being present). The reason for that might be a non-mature cDNA clone obtained during the cDNA library preparation \[[@B15]\]. The two other hits, \[GenBank:BX930135\] and \[GenBank:BX936151\] were therefore chosen for further analysis (Table [1](#T1){ref-type="table"}). For clarity these two sequences are called BBP-A\[GenBank:BX930135\] and BBP-B\[GenBank:BX936151\]. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### The BBP cDNA sequences found from the database and theoretical biochemical characters of the putative BBP proteins. Numbering of the sequences according to avidin sequence. ::: BBP-A BBP-B Method ------------------------------- ------------------- ------------------- -------------- N-glyc (rank) ^17^NMTI^20^(9/9) ^74^NATT^77^(5/9) NetNglyc^b^ Molecular mass (Da)^a^ 13845.8 16404.5 ProtParam^c^ pI^a^ 9.75 5.88 ProtParam Number of residues^a^ 124 148 ProtParam Extinction coeff. (280 nm)^a^ 24160 35660 ProtParam ^a^The signal sequence is cut at the most probable cleavage site according to prediction and published N-terminal sequences \[4\] (BBP-A: 1/2 AS-RKCE; BBP-B: -2/-1 TP-VERK). ^b^\[54\] ^c^\[53\] ::: The gene containing the fully identical sequence to BBP-AcDNA with three introns was found in chicken genome database in Contig166.108. Two contig-sequences (Contig55972.2 and Contig26844.1) containing parts of BBP-Bwere found in the database. These were manually joined together (Figure [1](#F1){ref-type="fig"}). The final product contained five changes in the nucleotide sequence when compared to BBP-BcDNA (\[GenBank:BX936151\]), causing differences in three amino acid residues close to the C-terminal part of the protein (N118I, V119L, F120L). The avidin gene and three avidinrelated genes were also found from the chicken genome database. The gene of BBP-Aand a novel allele of one of the previously cloned AVRs (or a novel avidin-related gene), which we named AVR-Awere found in the same Contig166.108. AVR-Awas similar to AVR2 and AVR6 (Figure [4](#F4){ref-type="fig"}) \[[@B16]\]. BBP-A and AVR-Agenes point towards each other (BBP-A→ ←AVR-A) separated by an intergenic distance of 8.1 kB. A chicken repeat 1 (CR1) element \[[@B18]\] was found between these two genes. The distance between AVR-Aand CR1 was 0.6 kB while the distance between the BBP-Agene and CR1 element was 6.1 kB. CR1 pointed towards BBP-Aand it was in parallel orientation with AVR-A. Previously, Wallén et al. have reported CR1 elements located 1.4--2.1 kB upstream from the 5\'-ends of AVR4and AVR5genes and pointing towards the genes \[[@B19]\]. Contig166.109 contains a partial gene (named AVR-B) clearly resembling AVR4. However, it has a mutation that converts Phe-29 in the AVR4 protein to leucine. In addition, Contig166.110 contains a partial gene (named AVR-C) resembling AVR2with the exception that it codes for Ser and Arg in positions 25 and 26 (as in avidin) instead of Asp and Asn found in AVR2 (Figure [4](#F4){ref-type="fig"}). Finally, Contig166.111 contains the avidin gene (Figure [1](#F1){ref-type="fig"}). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Schematic presentation of the genomic locations and orientations of the genes.The genomic location of the BBP-Bgene is unknown, whereas contigs 166.108-111 reside in the Z-chromosome as schematically shown in lower part of the picture. According to previous studies, the most probable location is in the q21 telomeric region of the Z-chromosome (Alhroth et al., 2000). ::: ![](1471-2164-6-41-1) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Sequence alignment.The sequence alignment of avidin, avidin related proteins AVR2, AVR4 and AVR6 (Laitinen et al., 2002), translated AVR genes AVR-A, AVR-Band AVR-Cfrom chicken genome database and putative biotin-binding proteins BBP-A and BBP-B. Numbering and secondary structures are according to avidin sequence. Completely conserved amino acids are shown in white with black background. Residues with direct contact to biotin in avidin (Livnah et al., 1993) are indicated with black triangles. ::: ![](1471-2164-6-41-4) ::: Alignment of BBPcDNAs with their corresponding DNA contig sequences revealed that both of these genes contain four exons and three introns, as shown for avidin and avidin-related genes \[[@B20]\]. The exon and intron lengths of the BBPgenes and their comparison with the avidin gene structure are shown in Figure [2](#F2){ref-type="fig"}. The fourth exons are cut after the stop-codon, and the first exons (N-terminus) are cleaved before the ATG starting open reading frame. The sizes of the exons are relatively similar with the exception of the fourth exon of BBP-Bwhich encodes 96 amino acids residues, compared to the 30--42 residues in BBP-Aand AVD/AVRs, respectively. The number of variable sites among the exons ranged from 25% (fourth exons) to 58% (second exons). The first intron is similar in size in all compared sequences, whereas the second intron is considerably longer in the avidin and AVRgenes (about 425 bp) than in the BBPgenes (175 bp in BBP-Aand 114 bp in BBP-B). On the contrary, the third intron is longer in the BBP-Bgene (252 bp) than in the avidin/AVRgenes (87 bp) (Figure [2](#F2){ref-type="fig"}). The number of variable sites among intron sequences ranged from 24% in the third intron to 59% in the first intron. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Structural organization of the avidin and biotin-binding protein genes.The coding region of each gene is composed of four exons and three introns (upper panel). The respective exon and intron sizes (in nt) are indicated. Sequences at the exon-intron junctions are presented in the lower part of the figure. AVRsequences are consensus sequences of different alleles/genes reported in this study. ^a^/- is a variable site among AVRs meaning either a gap or adenine. ::: ![](1471-2164-6-41-2) ::: A high similarity among the genes was observed at the exon/intron junctions as shown in Figure [2](#F2){ref-type="fig"}. Sequence divergence (p-distance) among avidin and BBPgenes ranged from 1.4% between AVR4and AVR-Band 48.5% between AVR-Band BBP-B(Table [2](#T2){ref-type="table"}). Similar values were obtained when sequence divergence among exons only or introns only (the combined sequence) were analysed (not shown). ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Pairwise p-distance (below diagonal) and S.E. (above diagonal) between avidin, AVRsand BBP genes obtained using MEGA v.3 \[40\]. ::: --------- ------- -------- -------- -------- --------- --------- --------- --------- --------- AVD AVR2 AVR4 AVR6 AVR-A AVR-B AVR-C BBP-A BBP-B AVD 0.009 0.007 0.008 0.009 0.016 0.010 0.018 0.019 AVR2 0.083 0.007 0.005 0.004 0.015 0.004 0.018 0.019 AVR4 0.054 0.050 0.006 0.007 0.007 0.007 0.018 0.019 AVR6 0.085 0.025 0.052 0.005 0.016 0.006 0.018 0.019 AVR-A 0.093 0.016 0.055 0.024 0.016 0.005 0.018 0.019 AVR-B 0.080 0.070 0.014 0.084 0.077 0.026 0.030 0.031 AVR-C 0.095 0.014 0.047 0.032 0.018 0.065 0.021 0.021 BBP-A 0.412 0.423 0.434 0.426 0.427 0.434 0.431 0.019 BBP-B 0.443 0.440 0.453 0.436 0.441 0.485 0.445 0.452 --------- ------- -------- -------- -------- --------- --------- --------- --------- --------- ::: The phylogenetic relationship of the AVD, AVRsand biotin-binding protein genes is shown graphically in Figure [3](#F3){ref-type="fig"} (the same relationships were obtained from the amino acid sequences; not shown). In the unrooted tree, avidin and avidin related genes formed a well supported cluster, which was the sister group of BBP-A. Finally, basal to the tree, was BBP-B. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Evolutionary relationships of genes.Neighbour joining tree obtained from the gene sequences of avidin, avidin related genes 2, 4 and 6 (AVRgenes are from Ahlroth et al., 2000) and BBP-Aand Bgenes. The tree was obtained from a pairwise p-distance matrix between sequences as implemented in MEGA v.3 (Kumar et al., 2004). Numbers indicate node bootstrap supports. ::: ![](1471-2164-6-41-3) ::: All characterised genes contained a potential promoter region upstream of the coding region according to prediction program used. In the case of AVR-C, the upstream region of the gene was not analysed due to a missing sequence. The promoters of BBP-Aand BBP-Bcontained a TATA sequence TATAAA at position (-30)-(-25) nt upstream of the predicted transcription initiation site. In the case of the AVDand AVR-A/Bgenes, the sequence AATAAAA was detected (-31)-(-25) nt upstream of the predicted transcription initiation site. The putative promoter regions contained possible binding sites for several transcription factors (not shown). Amino acid primary sequence characteristics ------------------------------------------- The most obvious difference between the BBP-B, when compared to avidin and BBP-A, is a C-terminal extension which makes it 18 residues longer than avidin and 22 residues longer than BBP-A. The sequence identity between BBP-A and BBP-B is 49%. The identity between the aligned regions of chicken avidin and BBP-A is 59% and between BBP-B and avidin is 47%. The residues involved in biotin binding in avidin \[[@B17]\] are almost perfectly conserved in both the BBP-forms (Figure [4](#F4){ref-type="fig"}). The only substitutions within these residues are Ser-73 which is replaced with alanine in BBP-A and Ser-75 which is replaced with alanine in BBP-B. Moreover, the T-A-T sequence in avidin (residues 38--40), that forms bonds with the carboxylic tail of biotin, is conserved in BBP-A, but replaced with T-L-A in BBP-B. The SignalP signal prediction program suggested the presence of signal peptide in BBP-A at residues (-21)-(-1) (numbering according to sequence alignment, Figure [4](#F4){ref-type="fig"}) and a cleavage site at position 1/2. Similarly, a signal peptide composed of residues (-21)-(-2) and a cleavage site in position 2/3 \[[@B21]\] or -2/-1 (hidden Markov model \[[@B22]\]) was predicted for BBP-B. It seems, therefore, that these putative BBPs are secreted proteins like avidin \[[@B1]\]. Both BBPs have one possible N-glycosylation site being ^17^Asn-Met-Thr-Ile^20^for BBP-A (identical to avidin) and ^74^Asn-Ala-Thr-Thr^77^for BBP-B. The prediction shows, however, a low probability for glycosylation to occur in BBP-B. The cysteine residues (Cys-4 and Cys-83) which form the intrasubunit disulphide bridge in avidin are conserved in both BBPs. In addition, two cysteine residues are found in the putative signal sequence of BBP-A and one in the signal sequence of BBP-B. Furthermore, there are two additional cysteines in BBP-B, one in the position corresponding to Glu-43 in avidin and one located in the middle of its C-terminal extension. The aromatic amino acids are conserved throughout the sequences. The only exceptions are the two tryptophans found only in BBP-B in the region corresponding to β-sheet 5 in avidin. Secondary and tertiary structure characteristics based on the homology modelling -------------------------------------------------------------------------------- The residues in BBP proteins corresponding to the β-sheet secondary structure elements of avidin are significantly more conserved when compared to the loop regions of avidin. Overall, the homology modelling strongly suggests avidin-like secondary (Figure [4](#F4){ref-type="fig"}) and tertiary (Figure [5](#F5){ref-type="fig"}) structures for both BBPs. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Comparative modelling of BBPs.A) Three-dimensional structure of the avidin subunit with bound ligand determined by X-ray crystallography (Livnah et al., 1993). B) The modelled structure of the BBP-A subunit and C) BBP-B subunit. Secondary structures are shown as cartoons: α-helix is red and β-sheets are grey. Cysteine bridges are shown as yellow sticks D) View of helix-β-barrel contact in RBP structure determined by X-ray crystallography (Cowan et al., 1990). The side-chains forming the contacts are shown as sticks. In E) is an enlargement of the modelled BBP-B helix-β-barrel contact. The conservation of the inner part and changes in the outer part of the modelled barrels of BBP-A (F) and BBP-B (G) are shown when compared to avidin. Conserved side-chains are shown in green and non-conserved are shown in red. The loop regions are omitted from the figure as well as the C-terminal predicted α-helix of BBP-B. H) Avidin tetramer with bound biotins, subunits are numbered according to Livnah et al., 1993. I) Tetrameric model of BBP-B. Proposed α-helices are oriented outwards from the possible tetramer. ::: ![](1471-2164-6-41-5) ::: Based on the modelled structures, both BBPs have the common lipocalin-motif: Gly-Xaa-Trp residues close to the N-terminus (residues 8--10 in alignment) and arginine in the last β-strand. This structural signature indicates that BBPs belong to the calycin superfamily together with avidin and streptavidin (which is a bacterial analogue of chicken avidin) \[[@B12]\]. At the tertiary structure level the most striking feature of the BBPs, when compared to avidin, is the conservation of the amino acid residues forming the inner part of the β-barrel. These amino acids also include almost all biotin-binding contact residues (Figures [5F](#F5){ref-type="fig"}, [5G](#F5){ref-type="fig"}). The hydrogen bond between biotin and Asp-128 in streptavidin \[[@B23]\], and biotin and the analogous residue Asn-118 in avidin (Hytönen VP et al.,unpublished results), are known to be important for their ligand binding. The bonding network including this residue comprises bonds between Gln-24 and Asp-128 in streptavidin \[[@B24]\] and Asn-118 and Asp-13 in avidin \[[@B17]\]. The residue corresponding to Asn-118 in avidin is conserved in both BBPs. The role of the C-terminal extension of BBP-B was hypothesised by modelling. Since there is an orphan cysteine residue near the end of β-strand 4 in BBP-B and another cysteine residue close the end of the C-terminal extension, one could assume a disulphide bridge between these cysteines. Several details support this possibility. Firstly, the region close to the cysteine residue in β-strand 4 in BBP-B seems to be rather hydrophobic (K9L, N17L, T34L, E46I), in comparison with the corresponding region in avidin. This might indicate a presence of a shielding structure in this region (i.e. contact to another protein or peptide) (Figure [5C](#F5){ref-type="fig"}). Secondly, the distance between the end of β-strand 8 and the cysteine in β-strand 4 is in good agreement with the length of the polypeptide sequence. Thirdly, similar structures are found in structurally similar lipocalin family proteins. For example, retinol-binding protein (PDB code: 1RBP) has a similar α-helix connected by a disulphide bridge in corresponding region \[[@B25]\]. Quaternary structure: interface-regions --------------------------------------- All residues forming the 1--2 interface (numbering according to Livnah et al.\[[@B17]\]) in avidin are conserved in all of the studied proteins (Table [3](#T3){ref-type="table"}). The 1--4 interface, being the most extensive, shows interesting similarities and differences when compared to that of avidin. Residues Gln-53, Thr-67, Trp-70, Gln-82, Val-103 and Thr-113 in this interface are conserved in all of the studied proteins. Asn-54 in the 1--4 interface has been shown to have a central role in the structurally important hydrogen-bonding network in the avidin structure \[[@B17],[@B26]\]. Interestingly, histidine is found in this position in the AVR-proteins, which are known to be stable tetramers \[[@B27],[@B28]\]. Glutamine in this position in the BBP model-structures seems to be able to form similar contacts between the subunits over the 1--4 interface. Taken together, only 10 out of 21 interacting residues in the 1--4 interface are conserved in BBP-A when compared to avidin, with the value being 8 out of 21 in BBP-B. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Comparison of interface residues of avidin, putative BBPs and AVR-proteins. Residues at subunit interfaces in avidin are determined according to Livnah et al. \[17\]. Equivalent residues in the other proteins are shown based on their alignment. ::: Secondary structure and type of interaction of amino acid residues of avidin in different subunit interfaces Residue in avidin Differences found in other proteins -------------------------------------------------------------------------------------------------------------- ------------------- ------------------------------------- ------ ------- ------- --------- Secondary M-chain S-chain BBP-A BBP-B AVRs^a^ 1--4 interface β4 H-bonds H50 K^b^ R^c^ L β4 H-bond Q53 \- \- \- β4 H-bonds N54 Q^d^ Q^d^ H B4 H-bond H-bond T55 P^b^ Q^b^ gap L4 H-bond N57 T^d^ S^d^ K L4 H-bonds R59 G^b^ V^b^ A β5 H-bond G65 \- A^d^ \- β5 H-bond H-bond T67 \- \- \- β5 H-bonds N69 Q^e^ W^f^ L/H L5 H-bonds W70 \- \- \- L5 H-bond K71 Q^d^ D^b^ N L5 H-bonds S73 A^d^ \- \- β6 H-phobic V78 \- A^d^ \- β6 H-bond T80 V^f^ A^f^ V β6 H-bonds Q82 \- \- \- β7 H-bond M96 A^b^ T^b^ K β7 H-phobic L98 \- M^d^ \- β7 H-bond R100 \- \- \- β7 H-bond S101 E^g^ E^g^ L L7 H-bond V103 \- \- \- β8 H-bonds T113 \- \- \- 1--2 interface L7 H-phobic W110 \- \- \- B8 H-bond T113 \- \- \- B8 H-bond V115 \- \- \- 1--3 interface B7 H-bond H-phobic M96 A T^b^ K B8 H-phobic V115 \- \- \- B8 H-phobic I117 T R^h^ N/Y ^a^Differences found in all AVR-proteins 1--7 in previous studies \[27, 31\] and current study (AVR-A, B, C) ^b^No interaction according to the model ^c^Salt bridge to D26 in subunit 4 according to the model ^d^Interaction similar to avidin according to the model ^e^Interaction similar to avidin without linking water molecule according to the model ^f^Hydrophobic interaction according to the model ^g^Side-chain hydrogen bond according to the model ^h^Salt bridge to E13 in subunit 3 according to the model ::: In the 1--3 interface, Val-115 is conserved, whereas both Met-96 and Ile-117 show variance both in BBPs and AVRs. The position of Met-96 shows interesting substitutions in other proteins since this residue faces the identical residue from the neighbouring subunit in avidin structure. According to previous mutagenesis studies this residue is known to be important for the tetrameric quaternary structure of avidin \[[@B26],[@B29]\]. According to the model structure, Arg-117 in BBP-B might form an interesting intersubunit salt bridge with Glu-13 from subunit 3. Discussion ========== The circumstantial evidence has indicated that chicken yolk BBPs may be structurally related to avidin and other members of the avidin family. Therefore we were eager to scan the chicken genome data to evaluate the correctness of this hypothesis. The database queries revealed two independent hits showing high similarity to the published N-terminal sequences of yolk BBPs I and II \[[@B4]\] and, indeed, a potential kinship between BBPs and avidin gene family members was revealed. The avidin gene belongs to the gene family that has several other members called AVRs(avidin related genes). Previously, seven different AVR-genes have been cloned \[[@B16],[@B20]\] and the chromosomal location of this gene family has been tracked down to a relatively short region in the telomeric region q21 of the chicken sex chromosome Z \[[@B16]\]. It seems that the number of AVRgenes varies between individual chickens and even between cells within the same chicken \[[@B30]\]. The deposited genome data, analysed in the present study, support this observation demonstrating a novel assembly of 3 AVRgenes together in the same cluster with the avidin gene. Interestingly, the two AVRgenes found in the chicken genome database seem to be novel variants of the formerly cloned AVRs, which also support the previous hypothesis of the high recombination frequency within the avidin gene family \[[@B30],[@B31]\]. Our observations link the BBPs to the avidin family for the first time, at the cDNA and gene level. There are many independent features indicating this. Firstly, the found cDNAs encode proteins that are evidently homologous to avidin. Secondly, the genomic location of the BBP-Agene close to the avidin gene family supports their relationship. Finally, the exon/intron structures of the BBPgenes and avidin family genes are similar to each other. According to the phylogenetic relationships and genome locations of the genes, one scenario for the BBP/avidin evolution is as follow: an initial duplication may have occurred leading to the origin of BBP-Band the precursor of the BBP-A/avidin family, followed by a further duplication leading to the origin of BBP-Aand the precursor of the avidin family. This could have been followed by the formation of AVDand an AVRgene and finally the duplication of the latter in several avidin-related genes. According to molecular modelling, BBP-A and B proteins both showed features that make them suitable for biotin binding. The biotin-binding contact amino acids of avidin \[[@B17]\] were almost perfectly conserved in both BBP sequences. In addition, good conservation of the inner part of the β-barrel in both BBPs is also important for the function of the ligand binding cavity. If we assume that one or both of these putative BBPs represent yolk BBPs I and/or II, other sequence differences should explain their observed weaker dissociation constant for biotin \[[@B7]\]. Similar observation has been done for AVRs which have only a few differences in their biotin-binding residues when compared to avidin, but still exhibit remarkable differences in their biotin-binding affinities \[[@B27]\]. It is probable that both BBP-A and BBP-B form similar tetrameric quaternary structures as avidin. The basis for this assumption is the fact that the conservation of the presumed interface residue patterns were highly similar to those of AVRs, which are also known to form extremely stable tetramers \[[@B27],[@B28]\]. The 1--2 interface, in particular, in which mutations have previously been shown to be extremely important to the stability properties of both avidin and streptavidin \[[@B32]-[@B34]\], was perfectly conserved in both BBPs. Hence the existing changes were concentrated on the 1--3 and 1--4 interfaces which are known to tolerate substitutions in avidin and AVRs \[[@B27]-[@B29]\]. The yolk BBPs have, however, been reported to be clearly less heat-stable than avidin \[[@B10]\] and, therefore either the sequence differences in the interfaces may explain this difference or the tertiary structure of the BBP barrel may be weaker than that of avidin. Overall, interfaces in the BBP models suggested tetramer formation, since the putative interface regions were hydrophobic. Furthermore, differences at the subunit interfaces of the BBP models, when compared to those of avidin, were at least partially complementary. For example the Thr-80-Val mutation at the 1--4 interface of BBP-A seemed to be in a highly hydrophobic environment. The most striking feature that distinguished BBP-B from avidin and BBP-A was its extraneous, approximately 20 amino acid residue-long, C-terminal extension. According to modelling, this stretch could form an α-helix and the cysteine residue at the end of this stretch could form a disulphide bridge with another cysteine residue in β-strand 4 (Figure [5C](#F5){ref-type="fig"}). It is, however, hard to interpret the relevance of this predicted α-helix and the possible effect of the helix and the cysteine bridge to the structural and functional properties of BBP-B. One effect could be that it strengthens the structure of the protein. Interestingly, many members of the lipocalin protein family have similar C-terminal α-helical domains \[[@B11],[@B12]\]. Nonetheless, the exon/intron structure of the avidin gene family is different when compared to the lipocalin family \[[@B35],[@B36]\]. This suggests that even if the overall tertiary structures of these proteins are similar, the evolutionary distance between these lineages is overwhelmingly long, or alternatively these protein families have been developed independently. The manner in which BBP-B has acquired its C-terminal extension, found in lipocalins, remains therefore an enigma. Alternative models for this extension can be done as well; the distance between the beginning of C-terminal extension (Lys-123) of the subunit 1 and the free cysteine in loop 3 of the subunits 3 and subunit 4 is around 40 A (not shown). This is also approximately the length of the fully extended C-terminal extension (Lys-123-Cys-138). Based on these distances both 1--3 and 1--4 intersubunit disulphide-bridges are possible. In the light of the current study, the story of the chicken BBP proteins gets blurred. As before, we have two possible candidates. However, these cDNAs/genes are able to encode proteins having molecular mass of 14--16.5 kDa, which correspond to only one BBP subunit instead of the four subsequent repeats hypothesized earlier \[[@B3],[@B9]\]. This means that either the database still lacks the genuine yolk-BBP gene or that some other phenomenon, like the abovementioned hypothetical disulphide bridges, must explain the previously reported molecular weight properties of BBP-I. Nonetheless, the fact that cDNAs for BBP-A and BBP-B were isolated from the chicken liver library \[[@B15]\], and that the putative proteins they encode have signal sequences, suggest that these cDNAs may really be the yolk BBP cDNAs, which have been reported to be secreted from the liver into the egg \[[@B4]\]. If we try to fit the characteristics of the found BBP candidates to those previously associated with BBPs, BBP-B looks more promising. It has similar low pI \[[@B7]\], its C-terminal extension makes its molecular mass closer to that determined for BBP-II \[[@B10]\] and it is most probably non-glycosylated as is BBP-II \[[@B10]\]. What is then the role of the BBP-A gene? Is it the mysterious BBP isolated from egg-white \[[@B6]\] or some unknown chicken BBP? It is evident that we need to continue the database queries and/or start cloning BBP genes to clarify this puzzle. In addition, we need to produce these found putative BBPs as recombinant proteins to investigate whether their properties are in agreement with the previous findings and the models of the present study. Furthermore, these new proteins can serve as a source for development of new tools for life sciences. For example, it could be possible to construct a chimeric avidin-BBP-dimer \[[@B37]\] to adjust the ligand-binding properties of the resultant hybrid protein. Conclusion ========== We have identified two putative genes and cDNAs for chicken egg-yolk biotin-binding proteins from NCBI database and chicken genome database. The genomic location and the structures of the found genes and the proteins they encode link clearly BBPs to the avidin family and, moreover, give an insight to the evolutionary history of this gene family. Our molecular modelling results support many preceding observations concerning the biochemical properties of BBPs but also impugn some of the previous hypothesis. Most importantly, the gene/cDNA structures provided no evidences of proteolytic processing of pseudotetramers to tetramers that has been presented as a possible maturation process for BBPs, i.e. conversion of BBP-I to BBP-II. Methods ======= Database queries and sequence analyses -------------------------------------- The N-terminal sequences VEIKXQLSGLWENEQDSLMEISALADDGG and VERKXQLSGLWENEQDSLMEISALADDLEN \[[@B4]\] were used to search the deposited collection of chicken cDNAs \[[@B15]\] by using TBlastn at the NCBI web site. The obtained cDNA sequences were used to find the corresponding genes and their genomic locations by searching the chicken genome database at the Ensembl web site \[[@B38]\] using blastn. Furthermore, cDNA of avidin \[[@B39]\] and AVRs\[[@B16]\] were used as search strings from this database. The intron/exon structures of these putative genes were analyzed. DNA sequences of AVD, AVR2, AVR4, AVR6, AVR-A, AVR-B, AVR-Cand BBP-Aand BBP-Bwere aligned exon by exon and intron by intron using Clustal X in multiple alignment mode with default values for both pairwise and multiple alignment parameters. Relationships among avidin and other biotin-binding proteins were obtained by the Neighbour Joining method from the p-distance matrix as implemented in Mega software \[[@B40]\]. The Dragon Promoter Finder v. 1.5 \[[@B41]\] was used to predict the promoter regions of the genes. The located promoter regions were further characterised using the transcription factor analysis implemented in the Dragon Promoter Finder program using Match™ \[[@B42]\] with default parameters. Structural modelling and polypeptide analysis --------------------------------------------- The three-dimensional structure of the avidin-biotin complex (PDB code: 2avi \[[@B17]\]) obtained from the Protein Data Bank \[[@B43]\] was used as a template structure in BBP modelling. Sequence alignment of all proteins was made using the MALIGN \[[@B44]\] multiple alignment tool of BODIL \[[@B45],[@B46]\] by using a structure-based sequence comparison matrix \[[@B47]\] with a gap penalty of 40. Comparative models of BBPs were made with Modeller 6v2 \[[@B48]\] according to alignment: disulphide bridges were forced between cysteine residues 3 and 83 in both BBPs and also between cysteine residues 43 and 138 in BBP-B. Furthermore, carboxy-terminal extension of BBP-B (amino acids 129--137) was forced to adopt α-helix conformation. Visual analysis of obtained models was done with the BODIL molecular modelling program. Alignment representation was made using ALSCRIPT \[[@B49]\] and protein representations were made using PyMOL \[[@B50]\]. The putative signal cleavage sites were analyzed by SignalP \[[@B21],[@B51]\]. The theoretical molecular weights, pIs and extinction coefficients were calculated using the program ProtParam \[[@B52],[@B53]\]. The potential N-glycosylation sites and their qualities were studied by NetNglyc \[[@B54]\]. Abbreviations ============= AVD, avidin protein; AVD, avidin gene; AVR; avidin-related protein; AVR, gene coding for avidin like protein; BBP, biotin binding protein; BBP, gene coding for biotin-binding protein; CR1, chicken repeat 1; pI, isoelectric point Authors\' contributions ======================= EAN did molecular modelling analysis. VPH, HRN and OHL did database searches and sequence alignments together with EAN. VPH carried out promoter analysis. AG did genetical analyses of cDNAs and genes. All authors, including MSK, took part in writing of the manuscript. Acknowledgements ================ Authors thank Dr Mervi Ahlroth for encouraging discussions and Ms Nicola Wakelin for proofreading the manuscript. This work was supported by The National Graduate School in Informational and Structural Biology.	PubMed Central	2024-06-05T03:55:55.690739	2005-3-18	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1082904/", "journal": "BMC Genomics. 2005 Mar 18; 6:41", "authors": [ { "first": "Einari A", "last": "Niskanen" }, { "first": "Vesa P", "last": "Hytönen" }, { "first": "Alessandro", "last": "Grapputo" }, { "first": "Henri R", "last": "Nordlund" }, { "first": "Markku S", "last": "Kulomaa" }, { "first": "Olli H", "last": "Laitinen" } ] }
PMC1082905	Background ========== Chemokines are a family of small chemoattrative peptides that were originally recognized to be involved in host defense as regulators of leukocyte trafficking, but more recently have also been shown to have roles in organogenesis, hematopoiesis, and neuronal communication \[[@B1]\]. Their cognate receptors belong to the Class A subfamily of G-protein coupled receptor superfamily \[[@B2]\]. Chemokines are believed to have originated from gene duplications and these genes underwent selection during recent evolutionary time \[[@B3]\]. All chemokines have a characteristic cysteine motif. Similarly, chemokine receptors may also be derived from a common ancestor through gene duplications. All chemokine receptors share high homology with the prototypical family member, rhodopsin \[[@B2]\]. Chemokines are highly basic proteins, 70 to 125 amino acids long. Sequence identity among chemokines is usually low; however, all share a typical overall tertiary structure, which consists of at least four cysteines that form two disulfide bonds. Chemokines are divided into two major (CC and CXC) and two minor (XC and CX~3~C) subfamilies based on their four conserved cysteines. The first two cysteines in the two major subfamilies are either adjacent (CC) or separated by one amino acid (CXC). The first two cysteines in the CX~3~C chemokines are separated by three amino acids, whereas the XC chemokines contain only two of the cysteines \[[@B4]\]. The CC chemokines can be further divided into two subcategories, MCPs (monocyte chemoattractant proteins) and MIPs (macrophage inflammatory proteins) based on their structural similarities \[[@B5]\]. The members of these two CC subcategories specifically attract mononuclear cells but not neutrophils. The CXC chemokines can also be divided into two subfamilies, one with an ELR (a conserved Glu-Leu-Arg preceding the first cysteine) motif, which is angiogenic and attracts neutrophils, and the other without the ELR motif, whose members do not attract neutrophils \[[@B6]\]. Chemokine genes are characterized by their chromosomal locations and similar gene structure. Most human CC and CXC genes are organized in gene clusters in mammalian genomes, such as human Chromosomes 4 and 17, and mouse Chromosomes 5 and 11 \[[@B7]\]. The genes encoding the CC subfamily contain three exons, whereas the CXC chemokine genes contain four exons \[[@B8],[@B9]\]. The XC subfamily of chemokines contains two members in human but only one in mouse. CX~3~CL1 is the only known member of the CX~3~C subfamily in human, mouse, rat, and monkey. There are extensive conserved syntenies in the chromosomal regions containing chemokine genes between human and mouse. Unlike chemokines, chemokine receptors share a higher degree of sequence identity within a species and between species. These receptors have characteristic seven alpha-helix transmembrane domains with a length between 340--370 amino acids and have up to 80% amino acid identity \[[@B1]\]. They also share an acidic amino terminus, a conserved sequence in the second intracellular loop, and one cysteine in each extracellular domain \[[@B10]\]. Most receptors can bind several chemokines of a single class with high affinity \[[@B11]\]. Like chemokines, most chemokine receptors are also clustered in a few chromosomal regions, such as human Chromosomes 2 and 3 \[[@B2]\]. Most amino acid sequences of chemokine receptors are encoded in one exon. At present, 42 chemokine genes have been identified in human (24 CC, 15 CXC, 1 CX~3~C, and 2 XC) and 36 (21 CC, 13 CXC, 1 CX~3~C, and 1 XC) in mouse, whereas there are 11 receptors for CCLs, 6 for CXCLs, 1 for CX~3~CL, and 1 for XCL in human and mouse. Only 11 chicken chemokines including 4 CXC, 6 CC, and 1 XC and seven chicken chemokine receptors including 2 CXCR and 5 CCR have been reported in the literature \[[@B12]-[@B23]\]. Chicken chemokines share low sequence identity with mammals \[[@B24]\]. Therefore, it was very difficult to assign chicken chemokines to a specific mammalian counterpart based on sequence data alone. Because of limited sequence similarity, most of the reported chicken chemokines were not named in accordance with the systematic nomenclature of mammalian chemokines. The newly available chicken draft genome sequence and a large number ESTs allow systematic identification and annotation of chicken chemokine and cognate receptor genes. The objectives of this study were to systematically identify chemokine and chemokine receptor genes in the chicken genome, to name these genes according to existing systematic nomenclature, and to make ligand-receptor binding inferences based on comparative sequence analysis. The systematic nomenclature for these chicken genes was based on the phylogenetic trees and syntenies of chicken, human, and mouse genes, and ligand-receptor binding inferences were according to \[[@B37]\] and \[[@B38]\]. Results ======= Chicken chemokines and chemokine receptors ------------------------------------------ In addition to the 11 previously reported, 12 new chicken chemokine were identified. These include 7 new CC chemokines named CCL1L1 (BX935885), CCL3L1 (CF258095), CCL/MCP-L2 (CK610423), CCL/MCP-L3 (CK610627), CCL17 (BI067703), CCL19 (BX929857), and CCL21 (CR522995), 4 new chicken CXC chemokines named CXCL13a (BX262175), CXCL13b (BX264625), and CXCL13c (CR352598), CXCL15 (BX929947), and 1 CX~3~CL1 chemokine (assembled from CR389767, BI066258, BM426140, and our sequence: AY730688). Eleven reported chicken genes were also named accordingly as CCL1L2 (L34552), CCL5 (ah294, AY037859), CCL4L1 (MIP-1β, AJ243034), CCL/MCP-L1 (ah221, AY037860/ BX933162), CCL16 (k203, Y18692), CCL20 (ah189, AY037861), CXCL8a (cCAF, M16199), CXCL8b (K60, Y14971), CXCL12 (SDF-1, BX936268), CXCL14 (JSC, AF285876), and XCL1 (lymphotactin, AF006742). In summary, there are 13 CCL, 8 CXCL, 1 CX~3~CL, and 1 XCL genes identified in the chicken genome. The information used for the nomenclature is shown in the comparative genomic maps and phylogenetic trees. Chicken chemokine amino acid sequence alignment shows that all chicken CC chemokines have four conserved cysteines with two adjacent cysteines at the N-terminus (Figure [1](#F1){ref-type="fig"}), whereas all chicken CXC chemokines have the conserved four cysteines with the first two cysteines separated by one amino acid (Figure [2](#F2){ref-type="fig"}). Both chicken CCLs and CXCLs show higher degrees of sequence similarity to each other in the signal peptide sequences and sequence regions containing the last two cysteines. Chicken CXCL8a, CXCL8b, and newly identified chemokine CXCL15 contain the ELR (Glu-Leu-Arg) motif. Only one chicken CX~3~C chemokine was found (Figure [3](#F3){ref-type="fig"}). The number of amino acid residues between conserved cysteines in all chemokines is highly conserved between chicken and human (Table [1](#T1){ref-type="table"}). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Alignment of amino acid sequences of chicken chemokine CC subfamily. Alignment gaps are indicated by dashes. Sequences with identical amino acid in at least 50% of chicken chemokines are highlighted in gray and conserved cysteine residues in dark gray. ::: ![](1471-2164-6-45-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Alignment of amino acid sequences of chicken chemokine CXC subfamily. Alignment gaps are indicated by dashes. Sequences with identical amino acid in at least 50% of chicken chemokines are highlighted in gray and conserved cysteine residues in dark gray. The conserved ELR motifs are underlined. ::: ![](1471-2164-6-45-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Alignment of amino acid sequences of chicken, human, mouse, rat and monkey chemokine CX~3~CL1. Alignment gaps are indicated by dashes. Sequences identical in all species are highlighted in gray. The asterisk represents the conserved cysteine residues. ::: ![](1471-2164-6-45-3) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Chicken and corresponding human chemokine cysteine motifs ::: Families Chemokines Chicken motif Human motif -------------- --------------------------- -------------------- -------------------- CC CCL1L1 CCX~25~CX~15~C CCX~22~CX~15~C CCL1L2 CCX~24~CX~15~C CCX~22~CX~15~C CCL3L1, CCL4L1, CCL5 CCX~22~CX~15~C CCX~22~CX~15~C CCL16 CCX~22~CX~15~C CCX~21~CX~15~C CCL17 CCX~22~CX~15~C CCX~22~CX~15~C CCL19 CCX~25~CX~15~C CCX~24~CX~15~C CCL20 CCX~24~CX~15~C CCX~24~CX~15~C CCL21 CCX~25~CX~15~C CCX~24~CX~17~C CCL/MCP-L1, -L2, -L3 CCX~22~CX~15~C CCX~22~CX~15~C CXC CXCL8a, CXCL8b CXCX~24~CX~15~C CXCX~24~CX~15~C CXCL13a, CXCL13b, CXCL13c CXCX~24~CX~15~C CXCX~24~CX~15~C CXCL12 CXCX~22~CX~15~C CXCX~22~CX~15~C CXCL14 CXCX~23~CX~19~C CXCX~23~CX~20~C CXCL15 CXCX~24~CX~15~C N/A^1^ XC XCL1 CX~36~C CX~36~C CX~3~C CX~3~CL1 CX~3~CX~21~CX~15~C CX~3~CX~21~CX~15~C ^1^CXCL15 is not found in humans. ::: Chicken chemokines have limited amino acid sequence similarity compared to their human counterparts. Generally, chicken CXC chemokines share 27 to 60% amino acid identity with their human homologs except for CXCL12, which share 73% identity with human CXCL12. The length of chicken chemokine CXCL polypeptides ranges from 95 to 107 amino acids. Compared to their human homologs, chicken CXCL chemokine amino acid sequences are shorter except for chicken CXCL8a, CXCL8b, and CXCL12, which are 4, 5, 12 amino acids longer than their respective human homologs. In contrast, the sequence identities between human and chicken CCL chemokines are generally lower than those for CXCL chemokines, ranging from 25 to 56%. Chicken chemokine CCL polypeptides have 89 to 108 amino acids. Chicken CCL1L2, CCL5, and CCL17 have the same amino acid length as their human counterparts. Chicken CCL4L1 is shorter than corresponding human CCL4, whereas chicken CCL1L1, CCL3L1, CCL19, and CCL20 are longer than the corresponding human CCLs. These differences in length between human and chicken chemokines are mostly in the N- and C- termini. Chicken CX~3~CL1 encodes 441 amino acids, longer than all mammalian CX~3~CL1 examined. It shares 20, 22, 20, and 22% amino acid identity with human, mouse, rat, and monkey CX~3~CL1, respectively. Sixty-six amino acid residues in chicken CX~3~CL1 are identical to residues in mammals, but the sequence identity between mammals is, as expected, much higher than that between chickens and mammals (Figure [3](#F3){ref-type="fig"}). There are more identical amino acids between chicken and mammals at both ends of the sequences. Unlike the chemokines, all the chicken chemokine receptor genes were aligned with non-chicken chemokine receptor reference genes in the chicken genome browser. There was at least one chicken EST sequence aligned to each receptor gene except for CCR4. In addition to 7 reported receptors, 7 new chicken chemokine receptors were identified and named as CCR4 (predicted sequences: ENSGALT00000019505.1), CCR6 (CV039916, BU451770, and CK987456), CCR7 (predicted sequence: chr27\_random\_59.1), CCR8a (AJ720982), CXCR2 (BX258468), CXCR5 (AJ450829), CX3CR1 (CF252942, BU204148, and AJ443633). In contrast to chicken chemokines, chicken chemokine receptors share significant amino acid identity with their human receptor counterparts. The percents of amino acid identity between chicken and human chemokine receptors range from 48 to 81%. The lengths of these chicken receptors range from 335 to 382 amino acids. The complete sequence of chicken CXCR2 is unknown due to a sequence gap in the chicken genome sequence. The CXCR2 EST and a partial genome sequence contain the last 170 amino acids of the C-terminus. Fourty-four amino acid residues were highly conservated (\>85% homologies) among all chicken chemokine receptors (Figure [4](#F4){ref-type="fig"}). These receptors all have seven transmembrane helices and three extracellular loops. Of the seven transmembrane helices, helix 1 and 7 show higher degrees of sequence similarity than the other helices. The similarity between the extracellular domains of the chicken receptors is lower, but all have a conserved cysteine residue. In contrast, the intracellular domains (except at the C-terminus) generally have higher degrees of sequence similarity than the extracellular domains. The second intracellular domains contain a highly conserved DRYLAIV sequence. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Alignment of amino acid sequences of chicken chemokine receptors with human CCR1. Alignment gaps are indicated by dashes. Sequences with identical amino acid in at least 50% or 85% of the chicken chemokines are highlighted in gray and dark gray, respectively. Asterisks represent the conserved cysteine residues. ED and ID denote extracellular and intracellular domains, respectively. Seven transmembrane spanning domains of chicken chemokine receptors were predicted using the SMART program and these consensus domains are indicated with a box. The N-terminal sequence of chicken CXCR2 is currently unknown. ::: ![](1471-2164-6-45-4) ::: Chromosomal locations and syntenies ----------------------------------- Comparisons of the chromosomal segments containing chemokines in the human, mouse, and chicken indicate that the organization of the chemokine genes was generally conserved between chickens and mammals (Figure [5](#F5){ref-type="fig"} and [6](#F6){ref-type="fig"}). Chicken CC and CXC chemokines are located on Chromosomes 1, 4, 6, 9, 13, 19, and Z. Like human and mouse, there are two large clusters in the chicken genome, located on Chromosome 19 and containing 9 CCL genes. Two CCL1-like (CCL1L1 and 2) and three chicken MCP-like (CCL/MCP-L1, L2, and L3) genes related to human and mouse MCPs, such as CCL2, 7, 8, 11, and 13, are in one cluster (Figure [5A](#F5){ref-type="fig"}), and CCL5, CCL16, CCL3L1, and CCL4L1 genes in another cluster (Figure [5B](#F5){ref-type="fig"}). Another CCL cluster is located on Chromosome Z containing two genes, CCL19 and CCL21 (Figure [5C](#F5){ref-type="fig"}). Two CXCL gene clusters are located on Chromosome 4 and contained 6 genes, two CXCL8 (CXCL8a and b) and one CXCL15 genes in one cluster (Figure [6A](#F6){ref-type="fig"}) and three CXCL13 (CXCL13a, b, and c) genes in another (Figure [6B](#F6){ref-type="fig"}). Chicken shares the syntenies with mouse and human in all these regions. There is one composite cluster containing one CX~3~CL1 and one CCL17 genes (Figure [6E](#F6){ref-type="fig"}). Synteny was conserved in chicken on one side of this cluster. Chicken CCL20, CXCL12, CXCL14, and XCL1 are individually located on Chromosomes 9 (Figure [5D](#F5){ref-type="fig"}), [6](#F6){ref-type="fig"} (Figure [6C](#F6){ref-type="fig"}), 13 (Figure [6D](#F6){ref-type="fig"}), and 1 (Figure [6F](#F6){ref-type="fig"}), respectively, and the syntenies were highly conserved between chicken, mouse, and human in these four locations. Mammalian CCL25, CCL28, and CXCL16 were not found in the chicken genome, although the syntenies associated with CCL25 and CCL28 were also conserved in chickens. A number of human chemokines including CCL2, 7, 8, 11, 15, 18, 23, 24, and 26, CXCL1, 2, 3, 4, 5, 6, 7, 9, 10, and 11 in chemokine clusters that share syntenies with chicken clusters on Chromosome 4, and 19 were not found in the chicken genome, indicating gene duplications in mammals. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Genomic organization (syntenies) of human, mouse, and chicken CCLs, CX3CLs, and XCLs ::: ![](1471-2164-6-45-5) ::: ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Genomic organization (syntenies) of human, mouse, and chicken CXCLs ::: ![](1471-2164-6-45-6) ::: Chemokine receptor genes were also highly conserved between chicken, human, and mouse, and were similarly clustered. The largest cluster of chicken chemokine receptors was found on Chromosome 2, where 5 receptor genes (CCR2, CCR5, CCR8L1, CCR9, and XCR1) were identified. Another cluster on Chromosome 2 contains CCR4, CCR8, and CX~3~CR1 genes. Chicken CXCR1 and CXCR2 are also clustered as in mammals, but the chromosomal segment is unknown. The remaining CCR and CXCR genes are individually located on Chromosomes 3 (CCR6), 27 (CCR7), 7 (CXCR4), and 24 (CXCR5). Several human chemokine receptors, such as CCR1, CCR3, CCR10, CXCR3, and CXCR6 were not found in the chicken genome, though the syntenies associated with these receptors are present in the chicken genome. Gene structure -------------- According to the chicken genome sequence, chicken chemokine genes share typical three-exon CC and four-exon CXC gene structures with mammals except for CXCL13a and CXCL13b, which have only three exons. Chicken chemokine genes are shorter than the corresponding human genes due to shorter introns in chickens. The gene structure of chemokine receptors was also conserved between chicken and mammals. The EST sequences indicate that chicken chemokine receptor genes could have up to 5 exons, though the complete sequences were not available. However, the expressed sequences show that the amino acid sequences of identified chicken receptors are mostly encoded in a single exon as are most of the mammalian chemokine receptors. Chicken ESTs aligned with the chicken genome sequence indicate that these receptor mRNAs have approximately 2 kb of 5\' UTR, as do those found in humans. Phylogenetic analyses and nomenclatures --------------------------------------- The phylogenetic trees (Figure [7A,7B](#F7){ref-type="fig"}, and [8](#F8){ref-type="fig"}) show that chicken CCL5, 16, 17, 19, and 20 and all seven CXCLs are closely related to single specific human and/or mouse chemokines. The phylogenetic trees together with the syntenies associated with these genes (Figure [5](#F5){ref-type="fig"} and [6](#F6){ref-type="fig"}) strongly indicate that these genes are the orthologs of those found in mammals; therefore, they are named accordingly. The phylogenetic results show that chickens have two CXCL8 and three CXCL13 genes (only one copy each in mammals), indicating gene duplications of these genes in aves. One chicken CXCL related to mouse CXCL15 but not to human CCLs is named as cCXCL15, which is also supported by the synteny of the chemokine cluster (Figure [6A](#F6){ref-type="fig"}). Chicken CCL21 is named according to relatedness to the human and mouse and the highly conserved synteny (Figure [5C](#F5){ref-type="fig"}). According to the phylogenetic tree in Figure [7B](#F7){ref-type="fig"}, two directly linked chicken CCLs are remotely related to human and mouse CCL1. The synteny associated with these genes also indicates that they may be CCL1-like genes (Figure [5A](#F5){ref-type="fig"}); therefore, they are named as CCL1L1 and CCL1L2. Three closely related chicken CCLs that are directly linked to the CCL1-like genes are related to a group of clustered human and mouse MCP CCLs (2, 7, 8, 11, and 13) in the phylogenetic tree. The synteny and phylogenetic tree do not provide information to a specific mammalian ortholog, though these three chicken genes are somewhat more similar to human CCL13 and mouse CCL2. The results indicate that these genes are chicken MCP-like (Figure [5A](#F5){ref-type="fig"} and [7B](#F7){ref-type="fig"}); therefore, they are named as CCL/MCP-like (CCL/MCP-L1, -L2, and -L3). A chicken CCL gene that is directly linked to CCL16 and CCL5 (Figure [5B](#F5){ref-type="fig"}) is distantly related to chicken CCL5 in the tree. This gene has been reported as MIP-1β-like chemokine \[[@B13]\], which is CCL4 in humans and mouse. Therefore, it is named as CCL4L1 in order to conform to the report. Another CCL in this cluster that does not display relatedness to other CCLs in the phylogenetic tree (Figure [7B](#F7){ref-type="fig"}) is named as CCL3L1 because this chemokine displays highest sequence similarity to a human CCL3-like chemokine, and it shares synteny with human CCL3 genes (Figure [5B](#F5){ref-type="fig"}). Overall, CXCLs are more conservative among chicken, human, and mouse than CCLs. ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Phylogenetic trees of the chemokine CC subfamily constructed using the amino acid sequences of chicken, human, and mouse chemokines. The numbers on the branches are bootstrap values (percentage that the simulation supports the original interpretation). Human, mouse, and chicken are abbreviated as h, m, and c, respectively. The scale bar reflects the horizontal distance at which amino acid sequences differ by 20% between two sequences. A and B are the phylogenetic trees of CCLs that are not located on Chromosome 19 or located on Chromosome 19, respectively. ::: ![](1471-2164-6-45-7) ::: ::: {#F8 .fig} Figure 8 ::: {.caption} ###### Phylogenetic tree of the chemokine CXC subfamily constructed using the amino acid sequences of chicken, human, and mouse chemokines. The numbers on the branches are bootstrap values (percentage that the simulation supports the original interpretation). Human, mouse, and chicken are abbreviated as h, m, and c, respectively. The scale bar reflects the horizontal distance at which amino acid sequences differ by 20% between two sequences. ::: ![](1471-2164-6-45-8) ::: Chicken chemokine receptors can also be named according to mammalian nomenclature based on phylogenetic analysis (Figure [7](#F7){ref-type="fig"}) and syntenies. The genetic distances appear to be shorter between chicken and mammalian chemokine receptors than those between chicken and mammalian chemokines, which probably due to highly conserved transmembrane domains in these receptors. Chicken CCR4, CCR6, CCR7, CCR9, CXCR2, CXCR4, CXCR5, CX3CR1, and XCR1 are closely related to a mammalian ortholog based on the phylogenetic analysis. There are two distantly related (relative to the distance between mouse and human CCR8) CCR8 genes in chickens. The one closer to human CCR8 is named as CCR8a and the other as CCR8b. There are also two chicken CCRs closely related to human and mouse CCR2 and CCR5, but the phylogenetic analysis could not distinguish them as either CCR2 or CCR5. Because these two receptors are located in a conserved chromosomal region on chicken, human, and mouse chromosomes, these two chicken CCRs were named as CCR2 and CCR5 based on the synteny in which CCR2 is closer to XCR1 than CCR5. In summary, 23 chemokine and 14 chemokine receptor genes were identified from the chicken genome in this study. Many chicken genes display high degrees of similarity with their human and mouse orthologs in terms of gene structure, sequence homology, and synteny. Chicken has significantly fewer CCLs, CXCLs, CCRs, and CXCRs than mammals, but it has the same number of CX~3~C, XC, and cognate receptors as mouse. The results of phylogenetic analyses generally agree with the comparative chromosomal locations and syntenies of the genes. The independent nomenclature of chicken chemokines and chemokine receptors suggests that the chicken may have ligand-receptor pairings similar to mammals. The organization of these genes suggests that there were a substantial number of these genes present before divergence between aves and mammals and more gene duplications of CC, CXC, CCR, and CXCR subfamilies in mammals than in aves after the divergence. Discussion ========== We systematically searched for chicken chemokine and chemokine receptor genes in the recently available draft chicken genome sequence. Without this information, it may have taken years to find chicken chemokines and their receptors. The independent nomenclature of chicken chemokines and chemokine receptors and mammalian chemokine-receptor binding information suggest that most of the genes have been identified. One exception was CCL25, the only known ligand of CCR9 in mammals, which was not found in this study though its receptor was identified. Likewise, CXCL14, and CXCL15 were identified in both chickens and mice, but their receptors are unknown; therefore, it is very likely that there are additional chicken chemokine and chemokine receptor genes in the chicken genome. Although most of the systematic nomenclature of the chicken genes was unambiguous based on both phylogenetic trees and syntenies, the information that was used to name seven chicken CCLs as CCL1L1, CCL1L2, CCL3L1, CCL4L1, CCL/MCP-L1, CCL/MCP-L2, and CCL/MCP-L3 and to distinguish two chicken chemokine receptors into CCR2 and CCR5 is inadequate. CCR2 and CCR5 are closely related and tightly linked in the human, mouse, and chicken genomes. The phylogenetic analysis indicates these genes were duplicated after the divergence between mammals and aves. Chicken CCL/MCP-L1, -L2, and -L3 were related to a group of clustered mouse and human MCP CCLs; therefore, specific cognate receptors must be tested to distinguish them. In humans, the chemokines of this MCP group and MIPs, such as CCL3, CCL4, and CCL5, can bind to more than one receptor, such as CCR1, 2, 3, and/or 5, but not both CCR2 and CCR5. CCR1 and CCR3 were not found in the chicken genome and probably are not present in the species. Therefore, chicken CCR2 and CCR5 may be two receptors that recognize these two groups of CCL chemokines, such as CCR2 for MCPs and CCR5 for MIPs. Interestingly, two CCL1 like (CCL1L1 and CCL1L2) and two CCL1 receptor (CCR8a and CCR8b) genes were found in the chicken genome. The ligand-receptor binding for these four genes can not be determined in this study. Nerveless, the names assigned based on comparative analysis in this study may prove useful in order to apply the functional and physiological knowledge from other species to chickens. Further lab testing must be carried out to confirm the ligand-receptor binding and to understand their biological functions. Chicken chemokine ESTs are highly represented in the EST database. There are several ESTs aligned to each identified chicken chemokine gene in the UCSC Genome browser. The sequences assembled from ESTs probably contained most, if not all, of the full-length chemokine mRNA sequences. Promoter sequences with a typical TATAA were detected with promoter prediction software (data not shown). However, there were only a few ESTs that partially cover chicken chemokine receptor genes. Some of these EST contain translation start sites. These EST sequences and reported complete coding sequences indicate that the amino acid sequences of chicken chemokine receptors are mostly encoded in one exon. The predicted amino acid sequences were of the expected length and aligned very well with the coding sequences of non-chicken reference genes in the UCSC genome browser. The conserved gene structure of this receptor family and high sequence similarity between chicken and mammals suggest that the predicted coding sequences were very accurate, especially for those with ESTs containing translation start sites. CCR4 is the only predicted gene that does not have a matching EST and CXCR2 is the only identified gene with partial sequence. Further study including sequencing expressed sequences is needed to confirm these genes. Conclusion ========== Based on the organization, syntenies, and phylogenetic trees of chicken, mouse, and human chemokine and chemokine receptor genes, we conclude that there may be a substantial number of chemokine and cognate receptor genes before divergence between aves and mammals. The presence of a few chicken chemokine and chemokine receptor paralogs and orthologs of the mammalian genes indicated that most chicken chemokine and the receptor genes shared common ancestors with the human and mouse genes. There were significantly more gene duplications of CC, CXC, CCR, and CXCR subfamilies in mammals than in aves after the divergence of mammals and aves. The mammalian and chicken genome sequences and the genes identified in this study can be used for further investigation of the molecular evolution of these gene families and as a model for the study of the divergence between aves and mammals. Avian and mammalian species may share similar chemokine-receptor binding patterns. The results of this study may be used as functional inferences for these chicken genes before they are experimentally tested. Methods ======= Gene identification ------------------- To identify syntenies, genes closely linked to human and mouse chemokines were identified and localized on the chicken genome using the UCSC genome browser \[[@B25]\]. Expressed Sequence Tags (ESTs) and chicken mRNA sequences in the corresponding chromosomal regions were then identified and, if necessary, assembled with the CAP3 program \[[@B26],[@B27]\]. These sequences were aligned with the corresponding chicken genomic sequence and any deletions or insertions corrected. Sequences were then submitted to ORF Finder (Open Reading Frame Finder) \[[@B28]\] and the open reading frames were used as queries in BLASTP \[[@B29],[@B30]\] searches against the non-redundant protein database in Genbank \[[@B31]\]. Sequences that produced significant alignments with chemokines were identified as putative chicken chemokine sequences. To identify chicken chemokine receptors, all sequences of putative chicken chemokine receptors including ESTs, mRNAs, and predicted sequences were retrieved from the UCSC Genome Browser. The identified ESTs were used to determine the translation start sites for the receptors. If the translation start sites could not be determined from ESTs, translation start sites were based on the most likely predicted sequences from non-chicken reference genes in the UCSC Genome Browser. Sequence analyses ----------------- Complete amino acid sequences of currently known human and mouse chemokines were retrieved from Genbank. The amino acid sequences of all putative chicken chemokines were predicted based on the open reading frames of the expressed nucleotide sequences (ESTs or mRNAs). The amino acid sequences were grouped according to CC, CXC, and CX~3~C motifs and aligned using the ClustalW program \[[@B32],[@B33]\]. The similarity of the amino acid sequences was determined based on alignments with the most likely human or mouse orthologs. Human CCR1 was included in the multiple alignments of chicken chemokine receptors for comparison. The seven transmembrane domains were predicted using the SMART program \[[@B34]\]. For comparison, human chemokines hCCL1 (GenBank accession number: (NM\_002981), hCCL2 (BC009716), hCCL3 (BC071834), hCCL4 (NM\_002984), hCCL5 (BC008600), hCCL7 (NM\_006273), hCCL8 (NM\_005623), hCCL11 (BC017850), hCCL13 (BC008621), hCCL14(BC045165), hCCL15 (NM\_032964), hCCL16 (NM\_004590), hCCL17 (BC069107), hCCL18 (BC069700), hCCL19 (BC027968), hCCL20 (BC020698), hCCL21 (BC027918), hCCL22 (BC027952), hCCL23 (NM\_145898), hCCL24 (BC069072), hCCL25 (NM\_005624), hCCL26 (BC069394), hCCL27 (AJ243542), hCCL28 (AF220210), hCXCL1 (BC011976), hCXCL2 (BC015753), hCXCL3 (BC065743), hCXCL4 (NM\_002619), hCXCL5 (BC008376), hCXCL6 (BC013744), hCXCL7 (BC028217), hCXCL8 (BC013615), hCXCL9 (BC063122), hCXCL10 (BC010954), hCXCL11 (BC012532), hCXCL12 (BC039893), hCXCL13 (BC012589), hCXCL14 (BC003513), and hCXCL16 (BC017588), and hCX~3~CL1(NM\_002996) and mouse chemokines CCL1 (NM\_011329), mCCL2 (NM\_011333), mCCL3 (NM\_011337), mCCL4 (NM\_013652), mCCL5 (BC033508), mCCL6 (BC002073), mCCL7 (BC061126), mCCL8 (NM\_021443), mCCL9 (NM\_011338), mCCL10 (U15209), mCCL11 (NM\_011330), mCCL12 (BC027520), mCCL17 (BC028505), mCCL19 (BC051472), mCCL20 (BC028504 (NM\_009138), mCCL27 (BC028511), mCCL28 (BC055864), and mCX~3~CL1 (BC054838) were retrieved from the GenBank. Reported chicken chemokines K60 (Y14971), cCAF (M16199), MIP-1β (AJ243034), k203 (Y18692), AH294 (AY037859), AH221 (AY037860), AH189 (AY037861), JSC (AF285876), SDF-1(BX936268), Clone 391 (L34552) and lymphotactin (AF006742) are included in this study. Rat (BC070938) and monkey (AF449286) CX~3~CL1 were also retrieved for CX~3~CL sequence analysis. There are several human chemokine-like genes in the human genome, which were not included in this study. Human and mouse chemokine receptors hCCR1 (NM\_001295), hCCR2 (NM\_000647), hCCR3 (NM\_001837), hCCR4 (NM\_005508), hCCR5 (NM\_000579), hCCR6 (NM\_004367), hCCR7 (NM\_001838), hCCR8 (NM\_005201), hCCR9 (NM\_006641), hCCR10 (AY429103), hCXCR1 (NM\_000634), hCXCR2 (BC037961), hCXCR3 (NM\_001504), hCXCR4 (AY728138), hCXCR5 (NM\_032966), hCXCR6 (NM\_006564), hCX3CR1 (NP\_001328), and hXCR1 (NM\_005283), mCXCR1 (AY749637), mCXCR2 (NM\_009909), mCXCR3 (NM\_009910), mCXCR4 (NM\_009911), mCXCR5 (NM\_007551), mCXCR6 (NM\_030712), mCCR1 (NM\_009912), mCCR2 (NM\_009915), mCCR3 (NM\_009914), mCCR4 (NM\_009916), mCCR5 (NM\_009917), mCCR6 (NM\_009835), mCCR7 (NM\_007719), mCCR8 (NM\_007720), mCCR9 (NM\_0099130), mCCR10 (AF215982), mCX3CR1 (NM\_009987), and mXCR1 (NM\_011798), and reported chicken cCCR2 (CAF28776), cCCR5 (BI393893, CAF28777), cCCR8L1(CAF28778), cCCR9 (CAF28781), cCXCR1 (AAG33964), cCXCR4 (NP\_989948), and cXCR1 (CAF28779), were also retrieved from GenBank for comparisons. Phylogenetic analyses of protein sequences of chicken, human, and mouse chemokines and chemokine receptors were based on the amino acid sequences using neighbor-joining with options selected for bootstrap test, pairwise deletion and Poisson correction, using MEGA3 \[[@B35],[@B36]\]. For ligand-receptor inference, the first 20 amino acids (leading peptide) of all chemokines were removed before the phylogenetic analysis and chicken CCLs were divided into two groups, one group located on Chromosomes 4 and 19 and the other from other chromosomes. Syntenies, phylogenetic trees, and sequence homologies were the combined information used for naming chicken chemokines and their cognate receptors according to the recommendations of the IUIS/WHO Subcommittee on Chemokine Nomenclature \[[@B37]\]. These chicken genes were named according to their closest predicted human or mouse orthologs if all information supports the nomenclature. If there was more than one chicken gene similar to a human and/or mouse gene, these gene was named as in the human and/or mouse followed by a letter with alphabet order. If a specific human or mouse ortholog could not be reliablely determined, the chicken genes were named according to a closest human or mouse ortholog followed by an \"L\" and a number based on the information available. This nomenclature also used the existing systematic names reported in the literature to avoid confusion. Polymerase chain reaction (PCR) and DNA sequencing -------------------------------------------------- Chicken EST or mRNA sequences were identified for all chemokine genes. All sequences contained complete putative open reading frames except for CX~3~CL1. However, partial chicken CX~3~CL1 gene sequences (BM426140, BI066258, and CR389767) were identified, with a gap of 123 nucleotides between the ESTs. Forward (TGTGACATCGGGAGTCGCTAC) and reverse (AAAATCCCCAGCGTTTGCTACT) PCR primers were used to amplify across the gap using cDNA prepared from white blood cells. PCR was performed as follows: An initial denaturation step at 94°C for 2 min and 35 cycles of denaturation, annealing, and extension at 94°C for 30 sec, 59°C for 45 sec, and 72°C for 1 min., and a final extension step was carried out at 72°C for 10 min. Unincorporated nucleotides were removed from amplified PCR products using BioMax spin-50 mini-columns (Millipore, Billerica, MA). BigDye terminator cycle sequencing reaction kits and an ABI Prism 377XL DNA Sequencer (Applied Biosystems) were used for DNA sequencing. List of abbreviations ===================== Abbreviations: cCAF, chicken chemotactic and angiogenic factor; JSC, Jun-suppressed chemokine; SDF-1, stromal cell-derived factor-1; MCPs, monocyte chemoattractant proteins; MIPs, macrophage inflammatory proteins. Authors\' contributions ======================= JW collected most of the data and drafted the manuscript. DLA contributed to the interpretation of the data and final approval of the manuscript. AY performed the DNA sequencing and assisted with the preparation of the manuscript. SHS and YJ designed computer programs to search chicken chemokine sequences in chicken EST database. JJZ provided the conception and design of the study, collected some of the data, conducted phylogenetic analysis, and revised the manuscript. ::: {#F9 .fig} Figure 9 ::: {.caption} ###### Phylogenetic tree of chemokine receptors constructed using the amino acid sequences of chicken, human, and mouse chemokine receptors. The numbers on the branches are bootstrap values (percentage that the simulation supports the original interpretation). Human, mouse, and chicken chemokines are abbreviated as h, m, and c, respectively, followed by the receptor named. The scale bar reflects the horizontal distance at which amino acid sequences differ by 10% between two sequences. ::: ![](1471-2164-6-45-9) ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Systematic names, chromosomal locations (kb), and putative identified cognate receptors of chicken chemokines^1^ ::: Nomenclature Chromosomal location Ligand name Putative receptor Chromosomal location -------------- --------------------------- -------------- -------------------- ------------------------ CCL1L1 chr19:4,491--4,492 I-309/TCA CCR8a and/or CCR8b chr2:43,465--43,469 CCL1L2 chr19:4,493--4,495 chr2:41,804--41,811 CCL/MCP-L1 chr19:4,495--4,496 MCP-? CCL/MCP-L2 chr19:4,498--4,499 MCP-? CCR2 chr2:41,768--41,769 CCL/MCP-L3 chr19:4,507--4,508 MCP-? CCL16 chr19:258--261 HCC CCL3L1 chr19:240--242 MIP-1α CCL4L1 chr19:250--253 MIP-1β CCR5 chr2:41,784--41,786 CCL5 chr19:263--266 RANTES CCL17 chr11:768--771 TARC CCR4 chr2:43,501--43,503 CCL19 chrZ\_random:7,804--7,809 MIP-3β CCR7 chr27\_random:661--673 CCL21 chrZ\_random:7,810--7,810 SLC CCL20 chr9:4,119--4,122 MIP-3α CCR6 chr3:38,589--38,596 Not found^2^ CCL25 CCR9 chr2:41,880--41,882 CXCL8b chr4:51,462--51,466 IL-8 CXCR2 chrUn:136,108--136,109 CXCL8a chr4:51,475--51,479 CXCR1^3^ chrUn:25,460--25,462 CXCL12 chr6:18,184--18,195 SDF-1 CXCR4 chr7:31,441--31,443 CXCL13a chr4:35,453--35,455 BCA-? CXCL13b chr4:35,455--35,457 BCA-? CXCR5 chr24:5,242--5,247 CXCL13c chr4:35,457--35,459 BCA-? CXCL14 chr13:14,231--14,239 BRAK Unknown^4^ CXCL15 chr4:51,500--51,501 Lungkine Unknown^4^ CX~3~CL1 chr11:758--764 Fractalkine CX~3~CR1 chr2:43,480--43,490 XCL1 chr1:780,81--78,086 Lymphotactin XCR1 chr2:41,831--41,833 ^1^The systematic naming, ligand naming, and putative receptors are according to \[37\] and \[38\]. ^2^CCL25 was not identified in this study. ^3^The ligand and receptor binding has been experimentally tested \[39\] ^4^The information is currently not available in humans or mice. ::: Acknowledgements ================ The authors sincerely thank Mr. Suresh Xavier for his assistance in setting up the chicken EST database for this study. This research project was financially supported by Life Science Task Force at Texas A&M University. This research work should also be credited to all persons including the funding agencies who contributed to chicken genome and EST sequencing and the genome browser. Without the chicken genetic information, this research would have taken years and substantial funds to accomplish. The authors also greatly appreciate the inputs from the reviewers of this manuscript.	PubMed Central	2024-06-05T03:55:55.694935	2005-3-24	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1082905/", "journal": "BMC Genomics. 2005 Mar 24; 6:45", "authors": [ { "first": "Jixin", "last": "Wang" }, { "first": "David L", "last": "Adelson" }, { "first": "Ahmet", "last": "Yilmaz" }, { "first": "Sing-Hoi", "last": "Sze" }, { "first": "Yuan", "last": "Jin" }, { "first": "James J", "last": "Zhu" } ] }
PMC1082906	Background ========== Atlantic salmon (Salmo salar) is a fish native to the basin of the North Atlantic Ocean. As farming of Atlantic salmon is growing as an agricultural enterprise, identification of genetic regions and genes responsible for economically important traits might be of importance for future agriculture of salmon. Genome projects in farm animals aim at identifying regions of the genome responsible for traits of economic importance in order to implement this into breeding and management programs. This is for instance achieved by performing QTL studies to identify genomic regions that are linked to traits of economic importance. Genetic and physical maps, publicly available databases and cDNA libraries from various tissues are a few of the tools needed to identify genomic regions responsible for traits of interest. Currently there are two published linkage maps for Atlantic salmon \[[@B2],[@B3]\] and several EST databases available to the public \[[@B4]-[@B8]\]. After a genomic region has been shown to be linked to a trait of economic importance, a high quality BAC library resource is crucial in the identification and functional characterization of the genetic variation. Atlantic salmon along with all fish of the family Salmonidaeshows residual tetraploidity after a duplication event that occurred 25--100 Myr ago \[[@B9]\]. However, rapid chromosome divergence has been observed between different salmonid species, involving Robertsonian changes as well as other structural rearrangements \[[@B9]\]. Hence, the pseudo-tetraploid state is challenging for researchers working with the Atlantic salmon genome. Here we describe the construction and characterization of a highly redundant BAC library of Atlantic salmon. To our knowledge, this is the first reported Atlantic salmon BAC library, which will be an important tool for constructing physical maps and in the identification and sequencing of regions of the Salmon genome. The large insert BAC clones will be useful as to provide better understanding of the pseudo-tetraploid state, identify regions of interest to the aquaculture industry as well as to the basic science community. Results and discussion ====================== The current BAC library, CHORI 214, consists of three segments. The average insert size was estimated based upon analysis of NotI digested DNA isolated from 249, 218, and 220 clones from segment 1, 2 and 3 respectively. The distribution of insert sizes is presented in Figure [1](#F1){ref-type="fig"}. Based on these results, the average insert size was estimated to be 188 kbp. With a total of approximately 299 000 clones the BAC library provides 18.8 haploid genome equivalents (see Table [3](#T3){ref-type="table"}) given a genome size of 3.27 pg (6.55 pg/N) \[[@B10]\] or similar 3.10 pg at the Genome Size Database \[[@B11]\]. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Details of the three segments of the BAC library ::: Segment 1 2 3 All segments ------------------------ -------------- -------------- -------------- -------------- Plate numbers 1--288 289--576 577--816 1--816 Plate count 288 288 240 816 Empty wells 1982 (1.79%) 2830 (2.56%) 2975 (3.23%) 7787 (2.49%) Non-recombinant clones 0 5 4 9 Non-insert clones ^1^ 1.2% ^2^ 2.8% ^3^ 2.3% ^4^ 2.0% Recombinant clones ^5^ \~107 000 \~105 000 \~87 000 \~299 000 Average insert size 189 Kbp 190 Kbp 186 Kbp 188 Kbp Genomic coverage 6.8X 6.6X 5.4X 18.8X Genomic DNA obtained from Salmon sperm was partially digested with EcoRI restriction enzyme and cloned in pTARBAC2.1 vector. ^1^Estimated numbers for each segment is based upon NotI digestion. ^2^3 out of 249 clones were identified as non-insert clones ^3^6 out of 218 clones were identified as non-insert clones ^4^5 out of 220 clones were identified as non-insert clones ^5^Recombinant clones have been estimated from the total well number subtracting non-insert clones, non-recombinant clones and empty wells. ::: ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Size distribution of BAC clone insert sizes in the CHORI-214 library.A total of 249, 218, and 220 clones from each of the segment 1--3 was digested with NotI and analyzed by PFGE. The horizontal axis refers to the range of insert sizes, and the vertical axis indicates percentage of clones corresponding to each size range in each of the three segments. The grey tones represent the segments as indicated ::: ![](1471-2164-6-50-1) ::: In the construction of the library, great care was made to estimate the amount of enzymes to be used for the digestion and dephosphorylation of the vector. Excessive amounts of either enzyme showed an increase in ligation products containing non-insert clones with a smaller vector fragment. These products are possibly due to star activity and deletion in the SacB gene or its promoter region, causing the clones to survive on media containing sucrose. The number of non-insert clones were estimated to represent about 2% of the library based upon the results from the NotI digestion analysis (see Table [3](#T3){ref-type="table"}). Nine non-recombinant clones, which contain the complete BAC vector only, were detected in the entire library by hybridization using pUC19 DNA onto the filter sets. Reducing the frequency of non-recombinant and non-insert clones is important as these clones can easily dominate the clone collection of the library due to reduced transformation efficiency of large DNA fragments in E. coli. High frequency of non-insert and non-recombinant clones will increase the cost for storage and utilization of the library. In addition, large scale random fingerprinting and sequencing will have a significant reduced cost using libraries with low frequency of non-insert and non-recombinant clones. To further characterize the library, a set of 34 oligo probes generated from sequence flanking known non-linked microsatellite markers was selected to be used in screening of the 17 BAC library colony filters. These filters contain the complete BAC library gridded in duplicate. In an attempt to reduce the probability of these probes to contain repeat sequences, each probe was hybridized to one BAC filter in order to check the quality of the probe as a unique marker. From these test hybridizations, in combination with Southern blots of genomic DNA digested with TaqI, PstI and MspI, 12 probes were characterized as unique with 0 to 4 signals onto a single BAC filter and single bands onto the genomic filters (results not shown). The three enzymes TaqI, PstI and MspI were chosen because of a high frequency of restriction sites for these enzymes in the salmon genome. As many as 20 of the 32 probes yielded several hundreds to thousands of positive double signals onto one BAC filter and multiple bands or smear onto the genomic southern blots indicating that the probes contained repetitive sequence elements. The 12 probes, which were assumed not to contain repeat elements based on the initial hybridization results, were hybridized as one pool to the filters and resulted in the identification of 396 positive BAC clones. To identify the relationships between the probes and the clones, a second hybridization experiment was performed for each probe individually onto new small nylon filters containing all 396 clones. Out of the 396 BAC clones identified in the primary screening, 233 were verified in these secondary screening experiments (Table [1](#T1){ref-type="table"}). In addition, all the clones that were not assigned to a probe in the second hybridization experiments, possible false positives, were analyzed individually in PCR experiments with all the microsatellite PCR primer pairs, resulting in the assignment of an additional 91 BAC clones to the microsatellite markers (Table [1](#T1){ref-type="table"}). See [additional file 1](#S1){ref-type="supplementary-material"} for a complete data set of BAC clones linked to each microsatellite marker. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Number of BAC clones detected after hybridization and PCR verification of probes flanking microsatellites ::: GenBank accession no. Mic.sat. clone name No. of clones observed in secondary hybridisation Verified by PCR Positive PCR product on unassigned clones Total no. of clones verified by PCR ------------------------- --------------------- --------------------------------------------------- ----------------- ------------------------------------------- ------------------------------------- AF256671 BHMS-175 28 28 7 35 AF256676 BHMS-189 24 21 5 26 AF256678 BHMS-201 9 9 4 13 AF256741 BHMS-289 19 18 5 23 AF256698 BHMS-304 19 18 6 24 AF256748 BHMS-330 12 12 1 13 AF256746 BHMS-349 23 23 4 27 AF256719 BHMS-429 42 42 29 71 AF256693 BHMS-278 8 \* \* \* AF256848 BHMS-255 24 \# \# \# AF256750 BHMS-337 21 5 29 34 AF256714 BHMS-396 4 3 1 4 Total number of clones: 233 179 91 270 Average 19.4 19.9 9.1 27.0 \* PCR products was not observed from BAC template, and a weak smear was observed using genomic DNA \# Weak smear was observed from BAC template and genomic DNA ::: Furthermore, we also screened the CHORI 214 BAC library using probes specific to known expressed sequences. Hence, 15 overgos were constructed from Atlantic salmon EST sequences (Table [2](#T2){ref-type="table"}). These probes were hybridized to the BAC filters as one pool, and 1203 BAC clones were verified as positive for the 15 probes after a second hybridization experiment. All the BAC clones were fingerprinted in order to provide additional information about their organization. The fingerprinting results generated multiple contigs for 14 of the probes while the clones for the last probe resulted in six single clones. Out of the 1203 clones, 160 clones were not mapped into contigs and are here referred to as singletons. The observation of singletons is either caused by clones containing small insert sizes, possible false positives or lower quality DNA loaded onto the agarose gel. The FPC software will not be able to map small insert clones into contigs using the selected cut off value. For a complete data set of BAC clones linked to each gene marker and the respective fingerprinting contigs see [additional file 1](#S1){ref-type="supplementary-material"}. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Contigs assembled from the hybridization of 15 EST derived overgo probes onto the BAC filters ::: No. of BAC clones in each contig ----------- -------------------------------------------- ------ ------ ----- ---------------------------------- ---- ---- ---- --- --- --- BG 934178 Elongation factor 2 79 19 3 28 18 14 BG934439 Eukaryotic translation initiation factor 5 91 3 2 78 10 BG935804 Ubiquitin 69 8 2 54 3 BG934353 Phosphate transfer protein 108 14 4 47 31 8 4 BG935917 cyclin L ania-6a 6 6 0 AF201470 Retinal rod opsin 100 16 3 64 17 3 AJ344158 Myostatin, isoform II 104 10 4 42 22 17 10 BG936489 Actin related protein (P16arc) 122 10 3 51 39 20 BG933799 AMP deaminase 76 12 5 38 7 7 4 3 BG933794 TAR DNA-binding protein 69 11 2 35 22 BG934675 Helicase 142 25 7 35 25 22 16 9 5 4 BE518514 Translation initiation factor 3 17 0 2 13 4 X14305 Growth hormone 74 8 3 42 20 3 BG935839 Collagen type I 97 7 6 37 28 12 4 4 3 BG935084 Transducin alpha subunit 49 11 3 15 12 9 Average: 80.2 10.7 3.3 Contigs with 3 or more clones are shown ::: An average of 80 BAC clones was observed in the screening of the EST derived overgo probes, ranging from 6 to 142 BAC clones and 3 contigs on average for each of these probes. However, the results obtained using the genomic oligos resulted in an average of 19.4 clones after secondary hybridization screening, and an average of 27 clones from PCR verification (Table [1](#T1){ref-type="table"}). The average observed with the genomic probes corresponds to the estimated genome coverage of 18.8 fold. Because of selection pressure one would expect duplicated expressed DNA sequences to be more preserved than non-coding sequences. Consequently, probes containing expressed sequences might more often identify the pseudo-duplicated regions (in addition to possible pseudo genes) than non-coding probes. However, using fingerprinting, these pseudo-duplicated regions containing restriction enzyme differences might still be mapped to different contigs. Observation of multiple fingerprinting contigs has previously been reported in a BAC library for Rainbow trout (Onchorhynchus mykiss), a member of the Salmonidaefamily, using gene and EST probes in screening of the library \[[@B12]\]. The authors present their results as evidence of locus duplication in Rainbow trout. Our results are in agreement with these findings \[[@B12]\]. Our approach, using both expressed sequences as well as non-coding sequences, might be more suited in characterization of BAC libraries of pseudo-ploidity state genomes than expressed sequences exclusively. Conclusion ========== We have constructed a BAC library from Atlantic salmon (Salmo salar). Based on the number of clones and the observed average insert size of 188 kb, we estimate the library to have in excess of 18-fold genome representation. Providing a publicly available highly redundant genomic large insert library for Atlantic salmon is important for several reasons. First, genomic large insert clones can be used to construct contigs for regions of interest, fingerprinted in large scale and thus create large physical maps of the salmon genome, or be used in shotgun sequencing approaches. Secondly, utilizing one library would be important in exchange of data between researchers. The characterization reported here illustrates the usefulness of the library in identifying genomic clones and the possibility of utilizing clones in fingerprinting analysis. The very large average insert size of the clones combined with the high redundancy will provide researchers with a possibility of obtaining complete gene sequences within a single BAC clone. This might be useful for expression studies of genes and their regulatory elements. The BAC library of Atlantic salmon will be an important tool for future salmon projects and to the salmon industry. Methods ======= Library construction -------------------- The BAC library was constructed following the protocols from \[[@B13]-[@B15]\] using the pTARBAC2.1 vector \[[@B16]\]. Transformation of the ligation products was performed using electrocompetent E. coliDH10B T1 phage resistant cells (ElectroMAX DH10B T1 resistant, Invitrogen). High-density replica filters were prepared as previously described \[[@B14]\]. High molecular weight DNA preparation ------------------------------------- Sperm cells were isolated from one single Atlantic salmon which was provided by Aqua Gen AS \[[@B17]\], and embedded in agarose plugs at a final concentration of 5 × 10^7^cells/ml followed by a sequence of treatments \[[@B15]\]. DNA used in the library construction was partially digested with EcoRI in the presence of EcoRI Methylase, and size fractionated by pulsed-field gel electrophoresis (PFGE) using a CHEF apparatus (BioRad). The size fractioned agarose gels were stored in 0.5 M EDTA until use. Electroelution procedures were used in obtaining the partially digested DNA from the gel slices \[[@B15]\]. BAC vector preparation ---------------------- The pTARBAC2.1 plasmid DNA was isolated using cesium chloride gradient purification \[[@B15]\], digested with ApaLI and EcoRI and treated with calf intestine phosphatase (CIP, from New England Biolabs) \[[@B15]\], and separated on 1.0% agarose CHEF gel. The vector fragment was purified from the gel as previously described \[[@B15]\]. Insert size analysis -------------------- One clone from each 384 plate in the library was inoculated in LB medium containing 20 μg/ml chloramphenicol. The clones were grown for 18 hours and the BAC DNA was purified using the automated plasmid isolation machine AutoGen 960 (AutoGen). BAC DNA was digested with NotI and analyzed by PFGE \[[@B13]\]. Low Range PFG Marker (New England Biolabs) was used as DNA size marker. The molecular weight determination was achieved using Alpha Innotech MultiImage digital imager and AlphaEase computer software (Alpha Innotech). Hybridization screening ----------------------- A set of twelve 36-mer oligo probes from unique genomic sequence flanking twelve different microsatellite markers were end labeled using PNK4 and hybridized as one pool onto the filters. In addition, 15 probes from Atlantic salmon EST sequences were used to design overlapping oligonucleotide probes (overgo; \[[@B18]\]). All probes used are presented in Table [4](#T4){ref-type="table"}. The microsatellite oligos and the EST overgos were used in hybridization as two separate pools. Hybridization was carried out in tubes at 65°C overnight in Church buffer without BSA \[[@B19]\]. The filters were washed 4 times at 65°C each for 15 minutes using 1.5 X SSC and 0.1% SDS. Identification of positive signals was achieved by exposing the filters to either X-ray film (Agfa) or to Phosphor Image cassettes (Amersham Biosciences). All the clones identified in the screening were arrayed into new 384 plates, spotted onto nylon filters using a 384 pin tool, grown overnight on LB agar containing 20 μg/ml chloramphenicol, and processed following established procedures \[[@B19]\]. Each filter was hybridized with the individual probes, and the clones identified were re-arrayed in 96 well plates. Genomic DNA was isolated using established phenol/chloroform extraction procedures \[[@B19]\], digested with TaqI, PstI and MspI and separated on a 1% TBE agarose gel. DNA was transferred to nylon filters using established Southern blot techniques \[[@B19]\]. ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Primer sequences for EST and genomic probes ::: EST probe ---------- -------------------------------------------------- ---------------------------------- 15844006 5\'-AAGCCTGTGCTGATGATGAACAAG-3\' 5\'-GCAGGGCACGGTCCATCTTGTTCA-3\' 15844267 5\'-GTCGTTTATGTCATCCCCTCTTCT-3\' 5\'-GGTATCTTCTGTCTGGAGAAGAGG-3\' 15845632 5\'-GAAGGCATCCCTCCTGATCAGCAG-3\' 5\'-CAGCGAAGATCAACCTCTGCTGAT-3\' 15844181 5\'-TACTCCATGCAGGGACTCTGCAAG-3\' 5\'-CCTCATAGAAGCCAAACTTGCAGA-3\' 15845745 5\'-GCGACCAGCTACATTTACCAAAGC-3\' 5\'-ATTCCACATCACCCAGGCTTTGGT-3\' 7271780 5\'-GCTTCCCCATCAACTTCCTCACGC-3\' 5\'-TCGATGGTGACGTAGAGCGTGAGG-3\' 16604728 5\'-ACTGGATTATTGCCCCTAAGCGCT-3\' 5\'-CAGTAGTTGGCCTTGTAGCGCTTA-3\' 15846317 5\'-TGCCACAAGTTCATGCGCTTCATG-3\' 5\'-TCTCGGCTCTCATCATCATGAAGC-3\' 15843627 5\'-ATGTCTCCGCTCAGCAACAACAGC-3\' 5\'-GGTAGCTGAGGAAGAGGCTGTTGT-3\' 15843622 5\'-GAGCCTAAGCACAATAATAGTAGG-3\' 5\'-CACGATCCATCATTTGCCTACTAT-3\' 15844503 5\'-CCATCAAGAAGGACGAGGACGTGC-3\' 5\'-GGGCAGTTCTTCTTCAGCACGTCC-3\' 15967287 5\'-GAGCTTCCAGCTGGTGGACACTGC-3\' 5\'-AGTCTTCTGCGTCTTGGCAGTGTC-3\' 15845667 5\'-CACTTGCTTAAGCTGGGCTCTATC-3\' 5\'-TCCATTGGTCCTCTCCGATAGAGC-3\' 15843919 5\'-AAGATCCCAGGTGGGCGAGGGAAT-3\' 5\'-TGTGATCCCGCTGACCATTCCCTC-3\' 15844912 5\'-AACATCCTGCAGTCTGCTCTGGCC-3\' 5\'-CCATGCCTCTGATGATGGCCAGAG-3\' PCR primers for microsatellites Forward primer Reverse primer AF256671 5\'-TCACATCCCTTAGCTCCC-3\' 5\'-CCTTTTTTGTGTCTTCAGC-3\' AF256676 5\'-AAACACCCTTCCCTTCAC-3\' 5\'-CAATTCAGGTCAAACCAAC-3\' AF256678 5\'-CCCCATGATGTGTTCTTC-3\' 5\'-CACAATGAGGCTTGACAC-3\' AF256741 5\'-TTGAGCCATCCTCACCTC-3\' 5\'-CACTGGTTTGTTGTTGTTG-3\' AF256698 5\'-CAGAACCGTGATCTGAAG-3\' 5\'-TGGACATTCTCTGGCGTC-3\' AF256748 5\'-CTAGATCACTCACCCAGG-3\' 5\'-GTGCTTTTGGCTTATGTTAG-3\' AF256746 5\'-GCTGTGATTTCTCTCTGC-3\' 5\'-AAAGGTGGGTCCAAGGAC-3\' AF256719 5\'-CCCCTGTCAAACGTCTTC-3\' 5\'-AGCACACTGGATTCAAGG-3\' AF256693 5\'-AGGCACAAACATGCAAGC-3\' 5\'-TCACCCCTGTGTCATCAC-3\' AF256848 5\'-TCCAAACCTGAATCCAGG-3\' 5\'-TTGTAGTGAAAGCCGCTG-3\' AF256750 5\'-TCCCACTGCCAACTACAG-3\' 5\'-GTTTAATCAAAGCATTCGCC-3\' AF256714 5\'-CCTGCCATCATCCAACTC-3\' 5\'-TCCACACCCAACATACTC-3\' Genomic probe sequence AF256671 5\'-GCAGCTCAGTGACTATGACTTCTCCGGTTTCCTGTTCTCT-3\' AF256676 5\'-GCCCTAGAGATTGAAATAGCATCCTCTTTCACGCCATGCA-3\' AF256678 5\'-CTGCAAGACAGAGAACACCATGACACACAGACCTCTGGAT-3\' AF256741 5\'-AGTGAAGACCTCAACCCACAAAGGCGCTATAATCGGCAAC-3\' AF256698 5\'-TCCTGTGTATCTGCAGTCAGTTCCAGGAAATGGAGGAGCA-3\' AF256748 5\'-TGAGGGGGCTTACAAGAGGTCTTCGCTTTGCCCCAGAAAA-3\' AF256746 5\'-CACAGTTGCCAGTTGAGAGAAGAGAAAGACGTTAGGGACA-3\' AF256719 5\'-TGGCAAAGCCTAGAGAGGTTTATCTCAGCACCACATTGCA-3\' AF256693 5\'-TCACCCCTCATTCACACAATCTCCAGCTGTCACATCAAGC-3\' AF256848 5\'-TTCTCGGCTAGATCACTTGCTCTGTCTCTCTTCCCCACTC-3\' AF256750 5\'-TCCTGTAGCATGCTGACATTCTGGCAGTCAGACACACAAG-3\' AF256714 5\'-CGCTGACTTGATTTGCCTTAATGCAGTATGTGTCAACCCC-3\' ::: PCR verification of positive clones ----------------------------------- PCR primers were constructed based on the genomic sequence of the microsatellites used in the screening of the library. Cultures containing 20 μg/ml chloramphenicol and 7% glycerol of the re-arrayed clones identified by the microsatellite oligos were transferred to new 96 well plates containing PCR reaction mix. PCR reactions were carried out in 96 well plates in a total volume of 25 μl, containing 0.2 mM dNTPs (Amersham Biosciences), 1X PCR buffer (Qiagen), 7 pmol of each primer and 0.25 units of Taq DNA polymerase (Qiagen). The PCR reactions were incubated at 95°C for 3 minutes, followed by 35 cycles at 95°C for 30 seconds, annealing at 54°C for 30 seconds, and extension at 72°C for 30 seconds with a final incubation at 72°C for 10 minutes. After thermocycling, 2 μl of each PCR reaction was analyzed on a 1.5% agarose gel (ReadyGel, Amersham Biosciences). Fingerprinting and contig assembly ---------------------------------- DNA from all the clones identified in the screening using the EST overgos was isolated using the AutoGen 960 (AutoGen), digested with HindIII and DNA fragments were separated in a 1.2% agarose gel using established protocols \[[@B20]\]. The gels were stained using SYBRGreen and scanned using a FluorImager 595 (Amersham Biosciences). Gel images were analyzed using Image 3.10 \[[@B6]\] and the data obtained were used with FPC v4.7 \[[@B6]\]. A fixed tolerance of seven, cut-off value of e-14 and a bury value of 10% was used to build contigs. To merge contigs, the cut-off value was raised to e-12 to contigs that corresponded to the individual EST sequences Authors\' contributions ======================= JT performed the experiments and drafted the manuscript. BZ, EF and BH provided supervision. KO, PJD, BFK, WSD and BH coordinated the project. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Probes and associated BAC clones. A list of probes and the associated BAC clones ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ We gratefully acknowledge Barbara Swiatkiewicz and Dorota Kostecka for filter processing assistance. We thank Gery Vessere for assistance with informatics. This work was supported by grants from the Research Council of Norway, NFR 139617/140 Salmon Genome Project, NFR 145847/140 The construction of an Atlantic salmon (Salmo salar) BAC library. This work was also supported by grant from Genome Canada (Genome Research on Atlantic Salmon Project). We would like to thank Aqua Gen AS \[[@B17]\] for providing the sperm used in the library construction.	PubMed Central	2024-06-05T03:55:55.698595	2005-4-4	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1082906/", "journal": "BMC Genomics. 2005 Apr 4; 6:50", "authors": [ { "first": "Jim", "last": "Thorsen" }, { "first": "Baoli", "last": "Zhu" }, { "first": "Eirik", "last": "Frengen" }, { "first": "Kazutoyo", "last": "Osoegawa" }, { "first": "Pieter J ", "last": "de Jong" }, { "first": "Ben F", "last": "Koop" }, { "first": "William S", "last": "Davidson" }, { "first": "Bjørn", "last": "Høyheim" } ] }
PMC1082907	Background ========== Studies have shown that young people with first episode psychosis (FEP) experience lengthy delays between the onset of psychotic symptoms and receipt of treatment \[[@B1],[@B2]\]. The average period from first onset of psychosis to initiation of adequate treatment (duration of untreated psychosis, or DUP) is one to two years \[[@B3]\]. During this untreated period, irreversible biological, social and psychological damage may take place \[[@B2]\], and a delay in treatment is associated with poorer short-term outcome and slower recovery \[[@B1]\]. Early intervention in this \'critical period\' \[[@B4]\] is therefore important for both patients and families. Strategies to reduce DUP include providing early access to specialist mental health services, such as early intervention services (EIS), and improving recognition of FEP by educating primary care professionals \[[@B5]\]. The integration of mental health and primary care services around the world, in developed as well as developing countries, has been widely advocated \[[@B6]-[@B10]\]. As part of this process, a range of national and international policy developments have occurred, with the aim of improving the identification and management of mental illness in primary care \[[@B11]-[@B18]\]. In the United Kingdom (UK), for example, standards two and three of the Mental Health National Service Framework \[[@B16]\] require primary care to provide effective identification, assessment, and treatment of people with mental illness, including appropriate early referral to specialist services. National guidance on schizophrenia \[[@B19]\] and the inclusion of mental health indicators for care of people with serious mental illness in the new GP contract \[[@B20]\] further strengthen the role and responsibilities of primary care by encouraging a more systematic approach to care, including the use of protocols and referral guidelines. As part of the UK policy response to improve the early detection and treatment of FEP, in 2000, the NHS Plan prioritised the development of 50 \"early intervention\" teams across England and Wales to provide specialist mental health services for all young people aged 14--35 with a FEP \[[@B21]\]. General practitioners (GPs) are well placed to play a greater role in the identification and management of FEP, as they are usually the first point of patient contact \[[@B22]\], and GP involvement in the management of psychosis is associated with reduced use of the Mental Health Act\[[@B23]\]. However, primary care health professionals cannot refer to EIS without knowledge of FEP and an understanding of the EIS referral system. The limited evidence base in this area has, however, consistently shown a need for improvement in the detection and management of FEP in primary care. Early detection is a challenge for many GPs, since psychosis does not present in \"neat packages\" and can take several months to emerge \[[@B24]\]. The prodrome largely consists of non-specific psychological and social disturbances of varying intensity, which is possibly why GPs experience difficulty in distinguishing FEP from normal adolescent behaviour \[[@B25],[@B26]\]. GPs also experience uncertainty about how to identify FEP, treat FEP appropriately, and access specialist mental health services \[[@B25],[@B26]\]. These data suggest the importance of training GPs to improve early detection and management of FEP to reduce the delay between onset of symptoms and initiation of treatment. Recent reports indicate that promoting early detection in a community based setting can be successful \[[@B27],[@B28]\]. A Scandinavian community education programme in the identification of FEP, for example, led to reduction in DUP from 1.5 years (mean) to 0.5 years \[[@B27],[@B29]\]. However, the Redirect trial is the first randomised controlled trial aimed at educating GPs about FEP. The Redirect study team designed an evidence based \"complex\" educational intervention that addressed the knowledge, skills, and attitudes of GPs about FEP. A refresher educational intervention was planned and implemented to reinforce knowledge and skills acquired in the initial educational intervention and to promote positive attitudes towards young people with FEP, given that most GPs see only one or two new people with FEP each year. The intervention is being used in a stratified-cluster randomised controlled trial to evaluate the effect of the educational intervention on GP referral rates of young people with FEP to EIS. Study aims ---------- The primary aim of the Redirect trial is to estimate whether an educational intervention targeted at GPs increases the GP referral rate of young people with FEP to EIS. Methods/design ============== Setting, eligibility and recruitment of practices ------------------------------------------------- Ethics Committee approval was obtained from Sandwell & West Birmingham, South Birmingham and East Birmingham Research Ethics Committees. The Redirect trial was conducted in three Primary Care Trusts (PCTs) within Birmingham in the UK, which encompass the areas of Aston, Handsworth, Ladywood, Nechells, Sandwell, Small Heath, Sparkbrook, Sparkhill, Soho, and Washwood Heath. Eighty-nine general practices within these three PCTs with the authority to refer to EIS were eligible for inclusion to the trial. These included 74 general practices located within the catchment area of the Heart of Birmingham Teaching PCT (HoB tPCT), 14 practices within the catchment area of Eastern Birmingham PCT, and one practice within the catchment area of South Birmingham PCT. Practice recruitment was facilitated by a letter to each practice from the study team and from the Chief Executive of the HoB tPCT, presentations by the second author at the HoB tPCT protected \"learning time\" sessions (where the PCT enabled GPs to close their practices and attend by paying for locum doctor cover) and an evening meeting attended by the practices within the catchment area of Eastern Birmingham PCT. Practices received a quarterly newsletter post randomisation during the study period. The newsletter was designed to encourage continued interest and participation in the Redirect trial. Consecutive patients with FEP who are referred by the Redirect trial GPs to the two Birmingham EIS over a two-year period from 1 March 2004 to March 2006 are eligible for inclusion in the study. Patients must also be aged between 14 to 30 years, in line with the Mental Health Policy implementation guide for Early Intervention \[[@B30]\], and have an ICD-10 \[[@B31]\] chart diagnosis of schizophrenia or related disorders (F20, F22, F23, F25, F28, F31). The exclusion criteria are patients with a primary diagnosis of substance use disorder, mood disorder, or organic mental disorder, current criminal proceedings, serious concurrent physical illness, institutional residence, learning disability, or inability to provide informed consent (see Table [1](#T1){ref-type="table"}). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Patient inclusion and exclusion criteria ::: Inclusion criteria ----------------------------------------------------------------------------------- Exclusion criteria Patients with a history of prior psychosis in receipt of antipsychotic medication Patients with a primary diagnosis of substance use disorder or organic disorder Institutional residence Patients with learning disability Inability to provide informed consent Patients with current criminal proceedings Patients with serious concurrent physical illness ::: Design and randomisation of general practices --------------------------------------------- The study is a stratified-cluster randomised controlled trial design \[[@B32],[@B33]\]. The cluster design at practice level reduces the risk that the educational intervention delivered to GPs will be contaminated by interaction between GPs within a single practice. Practices were randomly allocated after stratifying by list size (more or less than 3,500 patients) and PCT (HoB or not). These strata were used to ensure balance on the number of practices where no referrals of FEP might occur and to account for the differential prevalence of FEP across the study area. Practices were randomly allocated to either the intervention (39 practices) or control group (39 practices) by NF, who was blind to practice identity, using computer generated random numbers. Figure [1](#F1){ref-type="fig"} shows the trial profile summarising practice recruitment and retention. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### GP enrolment, randomisation and flow of practices through trial ::: ![](1472-6963-5-19-1) ::: The educational intervention ---------------------------- General practices were randomised to receive either the educational intervention on detecting FEP (intervention practices) or to an alternative educational session on cognitive behaviour therapy for depression (control practices) to control for any possible \"attention\" effect. In developing the educational intervention, the study team followed the five phases of the framework for the design and evaluation of complex health interventions proposed by the Medical Research Council and Campbell et al\[[@B34],[@B35]\], and incorporated evidence from systematic reviews and guidelines on changing professional practice \[[@B36],[@B37]\]. A brief 17-minute video depicting GP consultations with young people with FEP was shown to all members of the intervention practices, and the trial educators (first and second authors) then led a 15-minute question and answer session. The video addressed GP attitudes and negative stereotypes of people with FEP and also included elements of didactic factual teaching on, for example, links between drug use and psychosis, key warning symptoms of psychosis, including recognition of negative symptoms, and how to ask questions about sensitive issues such as hearing voices and suicidal ideas. A booklet summarising the content of the video and a two-sided laminated \'tip\' sheet on FEP symptoms and useful questions to ask patients were also given to all intervention group GPs. All intervention GPs viewed the video between November 2003 and January 2004. Twenty-one general practices (54%) viewed the video during \'protected learning\' time and 18 general practices (46%) viewed the video at their premises during their own time. At the end of the education session, GPs were asked to evaluate three specific elements of the video on a five-point rating scale: key information on FEP, useful questions to ask during consultations, and personal confidence in detecting FEP. GPs were also asked what they found most and least useful about the video. Refresher educational sessions for intervention practices were conducted in small groups, with training events spread over three days from 29 June 2004 to 1 July 2004. The refresher training consisted of personal testimonies from two service users, a service user personal testimony presented in a brief 10 minute video, and a presentation from an EIS representative on the EIS acceptance criteria and referral process. GPs were again asked to provide feedback on specific components of the training using a similar five-point rating scale. Outcomes -------- The primary outcome will be measured at practice level (unit of randomisation) by assessing the difference in the number of referrals of young people with FEP to secondary care services between study groups. Secondary outcomes are DUP, time to recovery, use of the Mental Health Act (at patient level), and GP consultation rate (at practice level). Instruments used in the evaluation are summarised in Table [2](#T2){ref-type="table"}. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Summary of research assessments ::: Instrument Baseline Follow-up -------------------------- ---------- ----------- SCAN \[38\] x PANSS \[41\] x x Insight Scale \[39\] x x Early Signs Scale \[40\] x x DUP \[42\] x PAS \[44\] x Encounter Form \[45\] x ::: Baseline assessments and follow-up ---------------------------------- The two field researchers are approaching all new FEP referrals to the two EIS in inner-city Birmingham to seek informed consent and entry to the trial. Patients meeting eligibility criteria and providing written informed consent are interviewed by the research team at baseline and at a four-month follow-up interview. At baseline, data are collected on socio-demographic factors and psychosis symptoms during the past month using the Schedules for Clinical Assessment in Neuropsychiatry (SCAN) version 2.0 \[[@B38]\]. The SCAN is a set of instruments used to assess, measure, and classify psychopathology associated with adult major psychiatric disorders. Part two, which includes psychosis disorders, was felt most applicable for use in this study. Insight ------- Insight is measured at baseline and at four months using the Insight Scale \[[@B39]\], an eight-item self-report scale designed to be sensitive to changes in levels of insight. The scale captures each of three widely accepted dimensions of insight: awareness of illness, perceived need for treatment, and ability to re-label symptoms as pathological. Higher scores indicate greater levels of insight. The psychometric properties of the scale are excellent and it is a widely used scale in psychosis research. Early signs ----------- The Early Signs Scale \[[@B40]\] is used at baseline and at four months to assess at risk mental states and provide a measure of clinical recovery, determined by ratings of \<20 (including \<10 on scales of incipient psychosis and disinhibition combined). Psychotic symptoms ------------------ Psychotic symptoms are measured at baseline and at four months with the Positive and Negative Syndrome Scale (PANSS; \[[@B41]\]). The PANSS is a 30-item semi-structured interview that consists of seven items assessing positive symptoms (e.g., hallucinations, delusions, conceptual disorganisation), seven items assessing negative symptoms (blunted affect, difficulty in abstract thinking) and 16 items assessing global psychopathology (e.g., depression, anxiety, disorientation). Items are scored between one (not present) and seven (severe). The PANSS is a widely used, valid and reliable measure of mental state. Duration of untreated psychosis (DUP) ------------------------------------- At the four-month follow-up assessment, DUP is assessed with a semi-structured interview following the model of Beiser \[[@B42]\]. DUP is defined as the time interval between the onset of psychotic symptoms and the initiation of treatment, and calculated according to a stringent protocol adapted from criteria developed by Larsen \[[@B43]\]. Premorbid functioning --------------------- Premorbid functioning is assessed at the four-month follow-up assessment with the Premorbid Adjustment Scale (PAS) \[[@B44]\]. The \"premorbid\" period is defined as the period that ends six months before there is any evidence of psychotic symptoms or first psychiatric hospital admission. The PAS measures four areas of development: (1) sociability-isolation, (2) peer relationships, (3) ability to function outside of the nuclear family, and (4) capacity to form intimate socio-sexual ties at each of four life stages: childhood (up to age 11), early adolescence (12--15 years), late adolescence (16--18 years), and adulthood (19 years and older). The PAS includes a \'general\' scale that measures the highest level of functioning attained by the individual before becoming ill. For example, if an individual was aged 20 at the time of completion, but experienced psychotic symptoms at the age of 17, the adult scale would not be completed. Items are scored on a Likert-type scale of zero to six, with lower scores denoting healthier functioning and higher scores indicating greater dysfunctional adjustment. Pathways to care ---------------- Pathways to care is measured at four months follow-up with the Encounter form \[[@B45]\]. This instrument is also used to determine retrospective GP consultation and referral rates and use of the Mental Health Act from primary care records. Reliability of diagnoses and PANSS scores ----------------------------------------- The two field researchers attended a five day, WHO-certified training course in using the SCAN and a local two day training course in administering the PANSS to the training standard of interrater concordance between field researchers and trainers. The interrater reliability method used was descriptive, according to agreement within one rating point on the positive and negative subscales and within three rating points on the general psychopathology subscale. To achieve interrater reliability on the total PANSS scores, agreement between the field researcher and trainer had to be within the 80% range. Throughout the study, we will conduct interrater reliability maintenance checks of the SCAN and PANSS, with live interviews, to avoid drift in scoring across time. Sample size ----------- The primary outcome of the study is the difference between the randomised groups on the number of young people with FEP referred to EIS during the study period, analysed on the basis of intention to treat. As the GPs, rather than their patients, are the subjects of the study, the statistical analysis and thus power calculation accounts for this. Assuming the standard critical value for α (2 sided p = 0.05), and further that, on average, two referrals will be made in control practices in the study period, and that the variance for between practice variability is one, the study has 80% power (1-β) to detect a mean difference of 1.2 referrals between intervention and control practices, and 90% power (1-β) to detect a mean difference of 1.4 referrals. As this study is the first randomised trial of an educational intervention in primary care which aims to influence detection and referral practice of GPs, we have no data upon which to base power calculations for the secondary outcome measures which may be considered exploratory (and which will provide relevant data for any future studies). Blinding -------- The participants and field researchers assessing patient outcomes were blinded to the identity of practices that participated in the educational intervention. Statistical analysis follows a pre-specified analysis plan. Statistical analysis plan ------------------------- The effect of the intervention on the primary outcome will be estimated using a non linear mixed model, where the number of referrals per practice will be modelled using Poisson error, and overdispersion (extra Poissonian variability) will be accounted for by defining the practice as a random effect \[[@B46]\]. Analyses will be conducted using Proc Nlmixed in the SAS statistical programme, version 9 \[[@B47]\]. Secondary outcomes will be analysed using a mixed modelling strategy, accounting for between practice variability by defining practices as random effects. As the subject of the experiment is the practice rather than the patient, the denominator degrees of freedom for the analysis will be derived from that stratum. Practices recruited ------------------- A total of 78 practices out of 89 eligible practices were recruited between July 2003 and October 2003 (Figure [1](#F1){ref-type="fig"}). The 11 practices that declined to participate cited staffing problems and/or time pressures as the main reasons for non-participation. Characteristics of participating practices ------------------------------------------ The practice characteristics of participating and non-participating practices are summarised in Table [3](#T3){ref-type="table"} and suggest that the participating practices are representative and that the randomisation, after stratification by practice list size and PCT, has been effective. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Practice characteristics at baseline ::: Intervention Practices (n= 39) Control Practices (n= 39) Practices Declined (n= 11) -------------------------------------------------- ---------------------------------- ----------------------------- ------------------------------ Mean practice list size (range) 4200 (924--10377) 4083 (1465--9927) 4149 (1300--9598) Single-handed practices 17 (44%) 17 (44%) 6 (55%) Mean No. of senior partners per practice (range) 2.35 (1--8) 2.03 (1--7) 1.81 (1--4) Total number of GPs 80 67 20 ::: Attendance at education intervention and GP feedback ---------------------------------------------------- All 39 practices (100%) were represented by one or more GPs at the initial video-based education sessions. Feedback forms on the video were completed by 53 of the 62 GPs (85% response rate). The majority of GPs who participated in the video training agreed or strongly agreed that the video helped them to identify key information to assist their consultations (89%), highlighted useful questions to ask patients (81%), and would improve their confidence in their ability to detect FEP (66%). Attendance at refresher education and GP feedback ------------------------------------------------- Thirty-five GPs, representing 26 practices (66.7%), attended the refresher education session and one practice (eight GPs) was visited by the first author. Feedback was obtained from 31 GPs (response rate 89%). Training was perceived as effective in raising awareness of FEP (87%) and the referral process to early intervention services (84%), and GPs reported that the training was relevant (84%), enjoyable (87%), and informative (84%). Discussion ========== The Redirect trial has been designed to evaluate the effectiveness of an educational intervention targeted at GPs on the detection of young people with FEP on referrals to EIS. The educational intervention was designed to be feasible to implement in a busy primary care setting and acceptable to GPs. Recruitment of practices into the trial itself was helped by ensuring that GP time and involvement in the study was kept to a minimum, in recognition of the workload pressures that many UK GPs currently face, particularly with the implementation of the new GMS GP contract \[[@B20]\]. All patients are therefore being recruited from the EIS. The full support and co-operation of the PCTs, in particular the Heart of Birmingham PCT who enabled the study team to use the bi-monthly PCT protected learning time session to deliver the video based education, was also key to the success of implementing the intervention. Strengths --------- The practice recruitment and retention rates were excellent, ensuring the generalisability of the study findings. Seventy-eight out of 89 eligible practices were recruited, a response rate of 87.6%. To date, there has been a low drop-out rate (one practice). We used a stratified-cluster design \[[@B32]\] to ensure a balanced randomisation for social deprivation, which we believe may affect patient-level outcomes. We are recruiting patients directly from specialist secondary care services, and patient outcomes are based on structured interviews and self-report. The GPs positively evaluated key elements of the initial education video and refresher education, and the good attendance at both these events suggests that this educational intervention is feasible and acceptable in a primary care setting. Limitations ----------- It was not possible to accurately assess whether the GPs actually used the written materials that accompanied the video (booklet and laminated \'tip\' sheet) during consultations. GPs may have disseminated their copy of the written materials to control group GPs in their area and the extent of contamination between intervention and control group practices is therefore unknown. Conclusion ========== FEP is a devastating diagnosis for the individual and their family. GPs appear to be key players in the referral pathway into EIS and therefore play an important role in the early identification of FEP and potentially in reducing DUP. The Redirect trial, which will report in 2006, will provide robust information about the efficacy of an evidence-based complex educational intervention targeted at GPs on referral rates of young people with FEP to EIS. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= HL, SW, and MB participated in the design of the study and wrote the protocol. LT contributed to the further development of the study design and protocol. NF is involved in the analysis and interpretation of the data. LT and HL participated in the drafting of the manuscript. All authors have contributed to and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1472-6963/5/19/prepub> Acknowledgements ================ We would like to thank Dr Amrit Khera for her contribution to organising the refresher education. We would also like to acknowledge the hard work of the two field researchers, Chris John and Kate Harris, the co-operation of the PCT education and learning staff, the case workers and team leads in the two Birmingham EIS and GPs in the 78 participating practices. Funding for this study was provided by Birmingham and Solihull Mental Health NHS Trust.	PubMed Central	2024-06-05T03:55:55.702632	2005-3-8	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1082907/", "journal": "BMC Health Serv Res. 2005 Mar 8; 5:19", "authors": [ { "first": "Lynda", "last": "Tait" }, { "first": "Helen", "last": "Lester" }, { "first": "Max", "last": "Birchwood" }, { "first": "Nick", "last": "Freemantle" }, { "first": "Sue", "last": "Wilson" } ] }
PMC1082908	Background ========== The clinical importance of yeast infections has increased during the last decades, not only because the number of yeast infections has increased, but also because yeast infections have become a frequent cause of morbidity and mortality in immunocompromised patients \[[@B1]\]. Besides the higher frequency there is also an important change in the spectrum of the species causing clinical infections. Whereas Candida albicanshas long been considered as the clinically most important species of the genus, the occurrence and pathogenic importance of the non-albicans species has been increasing steadily \[[@B2]\]. Because antifungal susceptibility is differing between the species, rapid and reliable identification of the clinical isolates can contribute to an efficient therapy of the patient \[[@B3]\]. The conventional identification of yeast isolates depends on biochemical properties such as assimilation and fermentation reactions and morphology \[[@B4]\], features which are not always stable, which are often not easy to interpret, and which may be time-consuming \[[@B5]\]. Biochemical identification has been standardized and (semi-)automated (e.g. Vitek-2) \[[@B6]\], but this approach equally suffers from the limitations of phenotypic identification in general. As a solution for those problems, several PCR-based methods have been described. Some of these approaches rely on species-specific probes \[[@B7],[@B8]\] -- being limited to those species for which probes are available, but most of them are based on amplification using universal fungal primers, followed by post-amplification analysis like probe hybridization \[[@B9]\], sequencing \[[@B10],[@B11]\], restriction analysis \[[@B12],[@B13]\], temperature gradient gel electrophoresis \[[@B14]\] or -- most simply and least laborious -- fragment length determination \[[@B15]-[@B18]\]. In addition, fingerprinting techniques primarily developed for strain typing are also used for identification: RAPD \[[@B19]\] and AFLP \[[@B20],[@B21]\]. Finally, techniques like real-time PCR \[[@B22],[@B23]\], pyrosequencing \[[@B24]\] and Luminex flow cytometry \[[@B25]\] have been used for identification of yeasts. The principle of Internal Transcribed Spacer 2 (ITS2)-PCR was initially described by Turenne et al.\[[@B13]\]. Amplification of the ITS2-PCR is followed by capillary electrophoresis for the precise determination of the fragment length, whereby the length of the fragment is used for identification. Evaluation, adaptation and expansion of the number of species included has been carried out at our laboratory and has resulted in a publicly available database, listing the ITS2-sizes for 39 yeast species \[[@B16]\]<http://allserv.ugent.be/~mvaneech/Yeasts.pdf>. Subsequently, an interlaboratory evaluation indicated that the ITS2-PCR technique and the database could be exchanged between different laboratories, when using the same electrophoresis platform, i.e. ABI310 (Applied Biosystems, Foster City, Ca.) \[[@B17]\]. Our present aim was to evaluate the interlaboratory exchangeability of the technique between laboratories using different capillary electrophoresis systems, i.e. ABI310 versus CEQ8000 (Beckman Coulter, Fullerton, Ca.), for the size determination of the amplified ITS2-fragments. Results ======= Construction of an ITS2-length library using CEQ8000 ---------------------------------------------------- Since ITS2-fragment length determination on CEQ8000 was expected to be different from that obtained on ABI310, we determined the length of a set of 44 reference strains, representing 33 species and subspecies, using the CEQ8000 capillary electrophoresis apparatus. The obtained lengths on CEQ8000 are listed in Table 1 (Additional file [1](#S1){ref-type="supplementary-material"}) and are compared with the lengths obtained on ABI310, as reported earlier \[[@B16]\], and with the theoretically expected lengths, as derived from published sequences. It could be observed that the values obtained on the CEQ8000 were always higher than on ABI310. The size differences were in the range of 0.3 basepairs (bp) for C. neoformanssubsp. gattii(strain IHEM 04170) to 7.2 bp for Malassezia furfur(strain IHEM 03967), with an average of 2.61 bp (SD 1.31 bp). There was no correlation between fragment size and size difference and there was no constant size difference. As illustrated in Table 1 (Aditional file [1](#S1){ref-type="supplementary-material"}), we could neither establish a correlation between %GC-content of the ITS2-region and migration differences as observed on both machines. The accuracy of fragment length calculation by CEQ8000 seemed to be somewhat higher than that achieved by ABI310, since in 33 of 44 cases the calculated length on CEQ8000 was closer to the theoretical length than that calculated on ABI310 (Table 1, Additional file [1](#S1){ref-type="supplementary-material"}). Although the absolute lengths obtained on CEQ8000 differed from those on ABI310, fragment length determination of the amplified ITS2 region on CEQ8000 enabled discrimination between the same species that could be differentiated on the ABI310. All species included could be identified to species level, except for the members of the genus Trichosporon, for which four of the five tested species had the same ITS2-fragment length. Since none of all other yeast species tested had an ITS2 of this length, the Trichosporonspecies could be identified as a group. Applicability ------------- The obtained ITS2-fragment lengths for the different species on CEQ8000 (Table 1, Additional file [1](#S1){ref-type="supplementary-material"}) were used as reference values with which ITS2-lengths of unknowns were compared. ITS2-fragment length determination on the CEQ8000 enabled the identification of the yeasts present in the cultures of 51 clinical samples, of which 5 samples contained different yeast species simultaneously. The following species and mixtures of species were identified: C. albicans(n = 12), C. albicans+ C. glabrata(1), C. albicans+ C. parapsilosis(1), C. glabrata(11), C. guilliermondii(1), C. kefyr(1), C. parapsilosis(12), C. tropicalis(7), C. tropicalis+ C. albicans(1), C. tropicalis+ C. albicans+ C. glabrata(1), C. tropicalis+ C. glabrata(1), C. krusei(1), and Ustilago maydis(1). No misidentifications occurred, compared to phenotypic identification. Exchangeability of data ----------------------- To evaluate the exchangeability of the technique between two laboratories using a different capillary electrophoresis apparatus, clinical strains were exchanged. Laboratory G identified 20 clinical isolates from the Ghent University Hospital on ABI310 and sent these to laboratory A, where identification was carried out on CEQ8000. Laboratory A identified the clinical isolates from 20 cultured samples from the Onze Lieve Vrouw Ziekenhuis Aalst, using the CEQ8000 and sent these to laboratory G, where identification was carried out on ABI310. The results of this comparison are presented in Table 2 (Additional file [2](#S2){ref-type="supplementary-material"}). The identification obtained with ITS2-PCR in both laboratories was in agreement for 39 of the 40 cultures. One culture yielded another identification, respectively as C. albicansusing the CEQ8000 and as C. kruseiby means of the ABI310. Retesting at laboratory A indicated that the sample contained a mixture of both species, while repeated culture and retesting at laboratory G yielded again only C. krusei. Based on culture results, one sample (OLVA 5397) was thought to contain a mixture of C. albicansand a non-albicans isolate. Both ITS2-PCR identification systems revealed that the mixture contained in fact three species, namely C. albicans, C. tropicalisand C. glabrata. During this study, one isolate was misidentified in both laboratories using ITS2-PCR. An isolate identified biochemically as C. dubliniensis, was identified as C. kruseiby means of ITS2-PCR in both laboratories. ITS2-sequencing, whereby the obtained sequence was compared to all known sequences in Genbank using BLAST, confirmed the biochemical identification as C. dubliniensis. The ITS2-sequence showed highest sequence similarity to C. dubliniensis, but contained a 2-bp indel (insertion/deletion) compared to the ITS2 sequence of the C. dubliniensisreference strains that had been used to construct the ITS2-length library. Genbank contains 17 ITS2 sequences of C. dubliniensis, of which 6 have the 2-bp indel. The ITS2 size library lists C. kruseias having a two bp shorter ITS2 size than that of C. dubliniensisand this explains the misidentification. This finding was added to the library, cautioning that in some cases C. dubliniensisstrains with an aberrant ITS2 fragment length may occur and thus may result in a misidentification as C. krusei. Discussion and conclusions ========================== ITS2-PCR followed by fragment length determination by capillary electrophoresis on ABI310 (Applied Biosystems) has been shown to be a fast and efficient identification technique for clinically important yeast species \[[@B16],[@B18]\] and for which data are exchangeable between laboratories using the same capillary electrophoresis system \[[@B17]\]. Here we evaluated whether comparative discriminatory power could be obtained on another capillary electrophoresis apparatus, in casuon CEQ8000, and whether identifications obtained in different laboratories using different capillary electrophoresis systems were identical. Since the fragment length estimation differs between different capillary electrophoresis systems, it was found necessary to recreate an ITS2 length library for the CEQ8000. The different size estimation most probably results from migration differences between different systems, caused by the usage of different fluorescent labels on the primers and of different polymers in the capillaries. Also the use of different marker fragments may result in differences between calculated lengths. It was observed that size estimations on CEQ8000 were always higher than on ABI310, ranging from 0.3 to 7.2 bp differences, without any notable consistency in the observed differences. However, differences were constant for each species, such that a new library for CEQ8000 could be constructed. It was found that the species that could be differentiated from each other by ITS2 length determination on ABI310 could also be differentiated by using the CEQ8000. ITS2-PCR was not very useful for differentiation between species of the genus Trichosporon. Identification of Trichosporonspecies was also impossible by ITS2-PCR on the ABI310 \[[@B16]\], and even sequencing of the ITS2 region is not always sufficient for final discrimination between all the species \[[@B26]\]. A more reliable identification of Trichosporonspp. can be obtained by sequencing a more variable region like IGS1 (intergenetic spacer region 1) \[[@B27]\] or making use of species specific probes and detection of hybridization in the Luminex flow cytometer \[[@B25]\]. During this study, both an additional problem and an additional possiblity of ITS2 fragment length determination for the identification of yeast species was encountered. Apparently, strains with aberrant ITS2 lengths occur. Here, a strain of which the biochemical identification as C. dubliniensiswas confirmed by ITS2-sequence determination was misidentified by both laboratories as C. krusei, when identification was based on ITS-length determination. It could be shown that this was the result of a 2 bp indel of the ITS2-region of this strain, an indel which is present in 6 of the 17 published C. dubliniensissequences. Further research is needed to find out whether the subdivision that can be made on the basis of ITS2-length corresponds to epidemiological or clinical relevance. An additional advantage of ITS2-PCR was observed when a mixed culture was studied. In order to identify the separate species present in mixed cultures, isolation of the cultures is necessary when identification is based on biochemical features or on sequencing of genes. However, when using ITS2-PCR on either CEQ8000 or ABI310, the different species present can be identified immediately, since one will obtain an ITS2-fingerprint composed of several peaks, each with a length corresponding to the respective species present in the mixed sample. Our results obtained on mixed culture samples indicate that ITS2-PCR could be applied for direct, non culture based, simultaneous detection of different species from a clinical sample. Whether this will be possible, will only depend on the strength of the DNA-extraction method used. Methods ======= Yeast strains ------------- A set of 44 reference strains or well-documented strains belonging to 33 species and subspecies, which were reference and type strains from culture collections or for which the correct identification had been obtained by sequencing of the ITS2 region in previous studies \[[@B16],[@B17]\], were used as reference panel (Table 1, Additional file [1](#S1){ref-type="supplementary-material"}). A series of 40 cultures on Sabouraud agar (Becton Dickinson, Erembodegem, Belgium) -- 20 from each hospital -- of clinical samples with unknown yeast isolates was also studied. Routine identification of these isolates was carried out using Auxacolor (Sanofi-Pasteur, Marnes-la-Coquette, France) at the Ghent University Hospital (GUH) laboratory (Laboratory G) and using Chromagar (Novolab, Geraardsbergen, Belgium) and API 32C (bioMérieux, Marcy-l\'Etoile, France) at the Onze Lieve Vrouw Ziekenhuis Aalst (OLVA) laboratory (Laboratory A). The selection of 40 cultures contained 43 clinical isolates belonging to the following species: C. albicans(n = 9), C. dubliniensis(1), C. glabrata(5), C. guilliermondii(3), C. kefyr(2), C. lusitaniae(1), C. parapsilosis(8), C. rugosa(1), C. tropicalis(6), Debaryomyces hansenii(1), C. krusei(5) and Ustilago maydis(1). All 40 culture plates were exchanged between laboratories, to study whether the identifications obtained in one laboratory could be confirmed in the other. DNA-extraction, PCR, capillary electrophoresis and analysis ----------------------------------------------------------- ### a. DNA-extraction #### Laboratory A DNA was extracted by alkaline lysis. A loopful (1μl inoculation needle) of yeast colonies was suspended in 40 μl of 0.05 M NaOH / 0.25 M SDS and incubated for 15 minutes at 95°C, whereafter 500 μl milliQ water was added. #### Laboratory G DNA-extraction was carried out using a boilling-freezing method. A loopful of yeast colonies was suspended in 250 μl TE-buffer, heated for 15 minutes at 95°C and immediately frozen at -70°C for at least 15 minutes. ### b. PCR #### Laboratory A PCR reactions were carried out in 50 μl reactions containing 10 μl yeast DNA extract, 1 x PCR buffer (Invitrogen, Carlsbad, Ca.), 0.2 mM of each dNTP, 1.5 mM MgCl~2~, 5 U of Taqpolymerase (Invitrogen) and 0.5 μM of primers ITS4 TCCTCCGCTTATTGATATGC) and ITS86 (GTGAATCATCGAATCTTTGAAC). Primer ITS86 was labeled with Beckman dye D4 (Proligo, Boulder, Co.). The amplification was carried out with the following protocol: 5 min at 94°C, 30 cycli of 1 min at 94°C, 1 min at 55°C and 1 min at 72°C and a final extension of 7 min at 72°C, followed by cooling at 4°C. #### Laboratory G PCR reactions were carried out in 25 μl reactions containing 2.5 μl yeast DNA extract, PCR Master Mix (Qiagen, Hilden, Germany) and 0.5 μM of primers ITS4 and ITS86 (sequence see above). Primer ITS86 was HEX-labeled (Applied Biosystems). The amplification was carried out with the following protocol: 5 min at 94°C, 30 cycli of 1 min at 94°C, 1 min at 55°C and 1 min at 72°C and a final extension of 7 min at 72°C, followed by cooling at 4°C. ### c. Capillary electrophoresis #### Laboratory A Capillary electrophoresis was performed on the eight capillary system CEQ8000. PCR products were diluted tenfold in nuclease-free water (Sigma, St. Louis, Mo.). Four μl of the PCR dilution was combined with 40 μl of SLS loading mix (deionized formamide, Analis, Namur, Belgium) and 0.5 μl of the CEQ DNA Size standard kit-600 (Analis). Samples were run on the CEQ8000 under the standard method FRAG-4: capillary temperature of 50°C, denaturing temperature of 90°C for 120 seconds, injection voltage of 2.0 kV for 30 seconds and separation voltage of 4.8 kV for 60 minutes. The total running time for one row of eight samples was 85 minutes. #### Laboratory G The capillary electrophoresis apparatus used was the ABI Prism 310 Genetic Analyser (Applied Biosystems, Foster City, Ca.). Electrophoresis was done in a single capillary, filled with liquid polymer (POP-4: Performance Optimized Polymer, Applied Biosystems). One μl of the PCR-product was added to the electrophoresis mixture (0.1 μl HD 400 marker, 0.3 μl of the ROX-500 marker (both from Applied Biosystem) and 12.1 μl of deionized formamide. All samples were first denaturated (2 min 95°C) before placement on the capillary. Injection of the sample was carried out at 5 kV for 10 sec, followed by an electrophoreis at 60°C and 15 kV, during 35 min. ### d. Size determination of the fragment length, followed by identification At laboratory A, the size and identity of the amplified ITS2-fragment was determined by the Fragment Analysis Module of the CEQ8000. The specific parameters concerning the marker standard used and the reference library, containing the ITS2 fragment lengths of the reference strains needed for identification, have to be selected by the user. The program then automatically calculates the size of the amplified ITS2-fragment by means of the known lengths of the fragments of the CEQ DNA Size standard kit-600 (Analis). This calculated fragment length is then compared to the ITS2 fragment lengths present in the selected library and as a result, the Fragment Analysis Module of the CEQ8000 gives an identification of the unknown strain. At laboratory G, the derivation of the fragment length is carried out by the Gene Scan analysis software (Applied Biosystems). The results were presented in a table indicating length and intensity of the observed fragments. The in-house software BaseHopper \[[@B16]\] enables to compare the obtained fragment length to those lengths present in the ITS2-database. Authors\' contributions ======================= GC, GV, AB and HD were responsible for sample collection and initial biochemical identification. TDB, PVH and AV carried out moleculair biological identification, analysis of the data. TDB, AV and MV drafted the manuscript. All authors read and approved the final manuscript. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Table 1. List of reference strains used, with the ITS2 fragment lengths obtained on CEQ8000 and ABI310 capillaries, and theoretical fragment lengths as calculated using Genbank seqeunces. ^a^: ATCC: American Type Culture Collection, Rockville, Md.; AZB: Algemeen Ziekenhuis Jette Brussel, Belgium; DGG: Veterinary Medicine (Diergeneeskunde), University Ghent, Belgium; HHR: Heilig Hartziekenhuis, Roeselare, Belgium; IHEM: Institute Hygiene Epidemiology Mycology; GUH : Ghent University Hospital, Belgium. ^b^: CEQ-TL: Difference between ITS2-length as determined on CEQ8000 with that as calculated from GenBank sequences (TL: theoretical length); ABI-TL: difference between ITS2-length as determined on ABI310 with that as calculated from GenBank sequences. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 2 Table 2. Comparison of the identifications obtained independently on ABI310 and CEQ8000 in the two different laboratories. ^a^: Confirmed: the identification was first obtained in the other laboratory and confirmed by this laboratory. ^b^: Results of two independent tests are separated by slash. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ We are thankful to Nicole Nolard and Danielle Swinne of the Scientific Institute of Public Health, Mycology Section (Brussels, Belgium) for the use of the reference strains of the IHEM collection. We also would like to thank Richard Summerbell of the Centraal Bureau voor Schimmelculturen, Utrecht, The Netherlands for taxonomic advice. Thierry De Baere is indebted to the Fund for Scientific Research -- Flanders (FWO) for a position as postdoctoral fellow.	PubMed Central	2024-06-05T03:55:55.704989	2005-3-18	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1082908/", "journal": "BMC Microbiol. 2005 Mar 18; 5:14", "authors": [ { "first": "Thierry", "last": "De Baere" }, { "first": "Anne", "last": "Van Keerberghen" }, { "first": "Peter", "last": "Van Hauwe" }, { "first": "Hans", "last": "De Beenhouwer" }, { "first": "An", "last": "Boel" }, { "first": "Gerda", "last": "Verschraegen" }, { "first": "Geert", "last": "Claeys" }, { "first": "Mario", "last": "Vaneechoutte" } ] }
PMC1082909	Introduction ============ Young people are at the forefront of the HIV/AIDS epidemic as it continues to spread worldwide. An estimated 12 million people aged between 15 and 24 years are living with HIV or AIDS around the world. Of the five million new HIV infections in 2001, over half were among youth aged 15--24 \[[@B1]\]. Six thousand young people become infected with HIV every day, and over half of all new HIV infections are related to injecting drug use \[[@B1]\]. In some regions, such as Eastern Europe and Central Asia, nearly all reported HIV infections are linked to drug injection, the majority being young injectors. In some developing and transitional countries, injection drug use is spreading rapidly and the age of initiation of drug injecting is decreasing \[[@B2]\]. Adolescence is an age when critical health behaviours are established, including behaviours related to sex and drug use (Ball, 2000). Most of these behaviors can be predicted from the risk environment, with clusters of risk behaviour being common, such as alcohol abuse and unprotected sex, particularly among marginalised and vulnerable youth. There is some evidence that young injectors think and behave differently to older IDUs and are treated differently within their communities. Specifically, young IDUs have less knowledge about HIV/AIDS, have a lower perception of their risk of acquiring HIV through either drug injecting or sex, and are less likely to identify as being an IDU than older IDUs \[[@B3]\]. Moreover youth have a heightened risk of HIV infection as a result of many factors, including risky sexual behaviour, substance abuse (including injecting drug use), and lack of access to HIV information and prevention services. It is crucial that barriers to accessing services that youth face are recognized by youth health services, including programs to prevent HIV infection. Marginalized young people, including homeless youth and ethnic minorities, may be at an heightened risk due to factors such as stigma (which may prevent access to critical HIV/AIDS information and prevention programs), pressure to engage in unprotected sex in exchange for food, shelter or money and the use of illicit drugs. In an attempt to minimize the HIV epidemic, a range of HIV interventions have been developed \[[@B1]\]. These interventions are designed to change behaviours of individuals who are at risk of acquiring or transmitting HIV infection. Programs for young people offer the greatest potential for changing the course of the epidemic \[[@B1]\]. However, research into HIV prevention in youth is an area that has seemingly been neglected, as most studies focus on adult populations. This remains the case despite global findings that injecting commences during adolescence \[[@B4]\]. Researchers need to redress this neglect of youth if they are to produce evidence necessary to allow an effective global health response. Comparison of programs for effectiveness will allow relative judgments to be made regarding effectiveness, and factors such as cost effectiveness can be taken into account to assist in the allocation of resources, particularly in resource-poor settings. Calls for international standardisation have appeared in the literature since at least 1999 \[[@B5]\]. According to Suishansian et al \[[@B6]\] the need for standardisation, collection, interpretation, and integration of program monitoring data with biological and behavioural surveillance data on HIV/AIDS associated with IDU is critical to informing and guiding appropriate prevention responses. The aim of this paper is to review recent literature of HIV prevention programs for young injecting drug users (IDUs) in an attempt to establish whether this call for standardisation has been heeded by researchers and program managers alike. Recommendations for improvement in evaluations to allow comparisons is provided, to assist in informing policy and program managers in the development of evaluation designs. Method ====== This review of the literature involved assessing the effectiveness of HIV prevention programs for young, and new injecting drugs users. The review also included programs undertaken to prevent initiation of drug injecting and transition from non-injecting to injecting drug use. Databases such as Medline, Psychinfo, Web of Science, Sociofiles, ERIC, Psychofiles and Aidsline were searched. International and local websites of drug addiction and prevention services or agencies and AIDS agencies and libraries were examined. A similar search strategy was also used to cover the grey literature. Unpublished literature, such as conference presentations and agency reports, was drawn from a number of different searches conducted on the worldwide web. Nonetheless, in order to evaluate the efficacy of the HIV programs, only those that provided information regarding the effectiveness of the outcome were included. Not surprisingly, this somewhat restricted the number of studies to be included. Results ======= The comprehensive literature search found five HIV prevention programs for young IDUs that met the criteria as outlined above (see [Additional file 1](#S1){ref-type="supplementary-material"}). More than fifty studies were considered and five met the criteria to be included in the study. The programs included were from Australia \[[@B7],[@B8]\] and the United States \[[@B9]-[@B11]\]. All five studies reported favourable HIV-related outcomes, although inspection of Table 1 reveals the following discrepancies between the studies: All five studies aimed to reduce risk behaviours or decrease incidence of HIV and other BBVIs. Biological outcomes such as an objective measure of HIV/HCV sero-status were absent in these studies. The outcome measured of risk behaviour varied. One outcome measure was BBVI knowledge, which was measured by a questionnaire pre and post-intervention \[[@B8]\] as well as by evaluation feedback questionnaires \[[@B7]\]. A second outcome measure was injecting risk practices, which was measure by self-report questionnaires \[[@B8]-[@B10]\]. Each study employed differing measures of injecting risk behaviours, such as using a new needle/syringe at last injection, the number of sharing partners and the frequency of needle/syringe sharing and other injecting equipment sharing. A third outcome measure used in two studies was sexual risk practices, measured by self-report questions. A fourth outcome used was following through with HIV-related health referrals, measured by using self-report questions. Further outcome measures were used as indicators of a successful program rather than outcomes directly related to changes in risk behaviours. Firstly participants\' satisfaction with the program was measured using focus groups \[7, 11 evaluation sheets \[[@B7]\], in an open-ended interview \[[@B8],[@B11]\]\] and a structured interview \[[@B8]\]. Participants\' perception of the program\'s impact was measured by focus groups and evaluation sheets \[[@B7]\] and structured and open-ended interviews \[[@B8]\]. Lastly, service utilisation was measured. Participants\' use of Harm Reduction Central\'s services such as the needle exchange service was used as an outcome measure by \[[@B11]\]. Service utilisation could also be an indirect measure of reduced syringe sharing; if participants were using the needle exchange services to a greater extent then they were likely to be sharing needles to a lesser extent. Gleghorn et al. \[[@B9]\] included a measure of Outreach Worker contact to determine the effect of different levels of contact on injecting and sexual risk behaviours. The target population varied in each of the HIV program studies. Firstly the definition of a young person varied, from under 26 years old to 12--23 years. However the most common definition of youth in the literature is between the ages of 15 and 24 years \[[@B4]\] although this definition was used in only one program study \[[@B7]\]. Secondly the definition of an injecting drug user varies, from those at risk of commencing injecting to injecting more than once a week. Thirdly the typical target population varied, with some programs targeting specific cultural and socio-economic subsets of the young IDU population, resulting in difficulties in comparing the effectiveness between the studies, as certain studies may vary in effectiveness depending on the target population. The sample sizes of the studies varied from 13 to 1,146 young IDUs. Sample size will impact on the statistical power of the study. Small sample sizes are unlikely to allow differences to be detected. Also, the assessment periods of the studies differed. Three studies were cross-sectional \[[@B9]-[@B11]\] whereas others were longitudinal \[[@B7],[@B8]\]. One cross-sectional study \[[@B9]\] conducted measures pre-intervention at the intervention site and at a comparison site. Measures were then conducted during the program (which was a continuous outreach program) at the program site and two other comparison sites. One cross-sectional study conducted measures during the program at the intervention site and a comparison site only, and another cross-sectional study conducted measures during the program but without a comparison group. The study conducted by Sheaves et al. \[[@B8]\] conducted measures pre and post-intervention and at one-month follow-up but was without a comparison group. The study conducted by Maher et al. \[[@B7]\] conducted assessment post-intervention and at 2-week follow-up, but was without baseline data and a comparison group. Conclusion ========== The United Nations aims to reduce HIV prevalence among 15--24 year olds by a quarter in the most affected countries by 2005 and globally by 2010 \[[@B1]\]. For this to be possible, youth require easy access to a wide range of effective HIV prevention programs. They require information and skills to help them adapt and maintain behaviours that are protective against HIV infection. The studies presented in Table 1 employed disparate methodologies, making a comparison of relative effectiveness impossible. According to the authors, all studies had some positive benefit for reducing HIV-related outcomes in the sample, whether it be increasing participant knowledge of HIV or other BBVIs or reducing needle and syringe sharing. However, they all differed in outcomes measured, instruments used, target population and study design. Due to differences in the programs\' nature and length and other practical constraints such as budgetary factors, it is unrealistic to expect a good level of consistency. Nonetheless, it is not unrealistic to expect better consistency than that presented in Table 1 (see additional file 1). As outlined in the Introduction, standardisation is now recognized as crucial and is in urgent need of implementation. The need for standardisation still needs to be emphasised, as standardized program findings are critical to informing and guiding appropriate prevention responses \[[@B6]\]. Moreover, the push for standardized indicators which can be compared across countries is necessary in order to determine which programs are effective in particular settings. Furthermore it should be noted that all the studies in Table 1 have come from research based in the US or Australia. It is unclear whether the results from these countries can be applied to other countries. The studies presented in Table 1 only measured behavioural outcomes and did not contain biological data. Biological data (HIV seroincidence for example, conducted by \[[@B12]\]) are an objective way of determining the effects of a prevention program. However, conducting a large-scale study with HIV serology, large enough to detect a noticeable effect between the intervention and control group would be costly and may not be feasible. Other objective data that could be used as outcome measures are service utilization measures such as those used by Weiker et al. \[[@B11]\]. It could be inferred, for example, that an increase in syringe distribution at a Needle and Syringe Program may be an indicator of a reduction in needle sharing or a reduction in the time syringes remain in circulation. As self-report data have the problems of recall bias and social desirability, combining these data with biological and/or service utilization data would improve the evidence of the effectiveness of a prevention program. The studies presented in Table 1 additionally lack cost-benefit data, crucial information to aid policy makers and program managers in allocating resources, particularly in resource poor settings. Random sampling was not conducted in any of the studies, thus biases in the results may exist. Although random sampling of individuals in one location for example may not be feasible, random allocation of the intervention to certain clinics may be one solution. Two studies did conduct assessments at an intervention site as well as one or more comparison sites, although these studies did not measure the same participants pre and post intervention, and so were unable to determine the effect of the prevention program on HIV-related outcomes. These short comings resulted in less conclusive data. However the desire to obtain accurate information and the costs and practical issues in collecting such information need to strike a balance and although the ideal should be strived for, it is not always possible. Another reason for the lack of research on youth has been the restrictions imposed by many ethics committees on researchers accessing young people. UNAIDS proposes that HIV prevention programs for youth should have the following characteristics for effectiveness: • To succeed they should respect and involve young people, while being sensitive to their cultures. • Young people need a safe and supportive environment, with sensitive attitudes, policies and legislation at family, community and national levels. • The stigma and discrimination associated with HIV/AIDS needs to be diffused. • Strong and effective education systems are important, yet in many countries education systems are clearly in disarray. • Outreach and peer education programs among young drug users should be expanded, and include steps to improve access to information and prevention equipment such as condoms and needles and syringes and HIV/AIDS care services. Upon searching the literature for recommendations of standardized instruments and study designs to integrate in a new evaluation study of a HIV prevention program, no clear guidelines were found. If standardised, valid and reliable instruments were in the literature, then program managers would be more likely to adopt them in practice. An important confounding concern is that although standardized indicators are needed that can be compared across countries and regions of the world, failure to adapt these indicators to the local setting can weaken our ability to obtain valid information. Key experts in the field need to make clear recommendations of evaluation study design and instruments to use, preferably at a global level. The following recommendations are a result of this review, with respect to study design, outcomes and instruments to use: \(1) Randomise (at clinic level or individual level) \(2) Collect baseline, post-intervention and follow-up data on the same clients \(3) Use control or comparison group \(4) Collect behavioural and biological data \(5) Use standardized instruments to measure HIV outcomes. Competing Interests =================== The author(s) declare that they have no competing interests. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Summary of Studies of programs for young and new injecting drug users ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ Collaborators involved in the review were Susan Kippax, Erica Southgate, Lucas Wiessing, Sylvia Inchaurraga, Nancy Haley, Justeen Hyde, Mary-Jane Rotheram-Borus, Moruf Adelakan, and Suresh Kumar. Julia-Lee Lowe and Lisa Bernstein assisted in the editing of this article.	PubMed Central	2024-06-05T03:55:55.706889	2005-3-17	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1082909/", "journal": "Harm Reduct J. 2005 Mar 17; 2:5", "authors": [ { "first": "Kate A", "last": "Dolan" }, { "first": "Heather", "last": "Niven" } ] }
PMC1082910	Introduction ============ It has become clear that the interaction between the oocyte and granulosa cells is required for the growth of the oocyte. Granulosa cells are influenced by gonadotropin in the blood and are engaged in hormone regulation in the ovary. As follicles grow and an antrum is formed, granulosa cells separate into two anatomically and functionally distinct sub-types: the cumulus granulosa cells, those surrounding and in intimate metabolic contact with the oocyte; and the mural granulosa cells (MGC), the cells lining the follicle wall forming a stratified epithelium with the basal lamina. Oocyte-regulated pathway of granulosa cell differentiation, through the secretion of paracrine growth factors, is towards the cumulus cell phenotype. Cumulus cells display functional characteristics that are markedly distinct from MGC; they have a high rate of proliferation, very low LH receptor expression compared to MGC, and posses the capacity to secrete hyaluronic acid and undergo mucification/expansion which MGC do not \[[@B1]-[@B4]\]. The highly specialized cumulus cells have trans-zonal cytoplasmic processes, which penetrate through the zona pellucida and abut the oocyte membrane \[[@B5]\], forming the cumulus-oocyte complex (COC). Gap junctions at the ends of these processes allow the transfer of small molecular weight molecules between oocyte and cumulus cell, and also between cumulus cells, whereas larger molecules are transported by receptor-mediated endocytosis. This mode of communication is essential for development and fertility \[[@B6]-[@B8]\], and is thought to play a key role in disseminating local and endocrine signals to the oocyte via the cumulus cells. Recently, the COC was suggested to be involved in steroidogenesis \[[@B9],[@B10]\]. Follicular maturation and development are complex processes influenced by both intra- and extra-ovarian events that lead to successful ovulation. It is possible that steroidogenesis in the COC contributes to the quality of the embryo. The purpose of this study is to localize the expression of steroid sulfatase (STS) in the COC and to examine the relationships between the STS expression level and the serum levels of follicle-stimulating hormone (FSH), luteinizing hormone (LH), estradiol and progesterone. Materials and methods ===================== Patients -------- The subject group consisted of 49 women (29 to 44 years old) for whom in vitro fertilization (IVF) treatment was indicated. The women were patients at Showa University Hospital. Informed consent was obtained from all patients. 114 samples of cumulus cells were obtained under microscopic observation. 9 samples of cumulus cells from 3 women were used for the immunohistochemistry, while independently, 105 samples from 49 women were used for RT-PCR. Indications for the IVF procedure in these women included: tubal factor infertility (11 cases, 32 samples); male infertility (10 cases, 30 samples); and unexplained infertility (28 cases, 89 samples). Ovarian stimulation ------------------- The patients received ovarian stimulation therapy with clomiphene citrate (CC, Clomid; Merrell, Cincinnati, OH) + follicle-stimulating hormone (FSH; Fertinorm-P; Serono, SA, Madrid, Spain) or human menopausal gonadotropin (HMG; Humegon; Organon, Oss, The Netherlands). The LH surge was induced by 5,000 IU of human chorionic gonadotropin (hCG; Gonatropin; Teikokuzouki KK, Tokyo, Japan) when at least two follicles reached a diameter of 17 mm, as determined by vaginal ultrasound. Oocytes were retrieved under ultrasonographic guidance 34 to 35 h after the administration of hCG. Oocytes were separated from follicular fluid and flush solution. Then, cumulus cells were separated from oocytes, other cellular debris and blood cells under microscopic observation. Cumulus cells were obtained, and immunohistochemical staining was performed using anti-human STS polyclonal antibody. The phenol-chloroform method was used to extract mRNA from cumulus cells. The expression of STS mRNA was examined by quantitative RT-PCR using primers specific for STS (as described below). Immunohistochemical staining for STS in cumulus cells ----------------------------------------------------- Tissue specimens of 9 samples of cumulus cells were fixed in 10% neutral-buffered formalin for 24 to 72 hours at room temperature, dehydrated through a graded alcohol series, cleared in xylene and embedded in paraffin. Serial sections (3-μm thick) were cut and stained with anti-human STS polyclonal antibody using the ABC method \[[@B11]\]. This antibody was confirmed the cross reactivity by western blott method. Negative controls for the immunostaining were carried out with mouse serum instead of primary antibody. RNA extraction, reverse transcription and real-time PCR ------------------------------------------------------- Cumulus cells obtained for real-time RT-PCR were put into each microtubes, and freeze preservation at -80 degrees was carried out before RNA extraction. Total RNA was extracted from the cells using the phenol-chloroform method. Extracted total RNAs were diluted with 20 μl DEPC water in the absence of Rnase. 5 μl of RNA solutions were used for reverse transcription (RT) using oligo (dT) primers and a TaKaRa RNA PCR Kit (AMV) Ver 2.1 (TAKARA BIO INC, Shiga, Japan) according to the manufacturer\'s instructions. TaqMan Universal PCR MasterMix and Assays-on-Demand Gene Expression probes (Applied Biosystems, Foster City, CA, USA) were used for the PCR step. Amplification and detection were performed using the ABI PRISM 7700 Sequence Detection System (Applied Biosystems) with the following profile : 1 cycle at 94 degrees for 10 min, and 40 cycle each at 95 degrees for 15 sec and 60 degrees for 1 min. The placental RNA with which dilution magnifications differ (×1, ×10, ×10^2^, ×10^3^and ×10^4^) were used and each quantity value were set as ×10^4^, ×10^3^, ×10^2^, ×10, ×1, to make a series of standards curve. The threshold cycle (Ct), which was defined as the cycle at a significant value, was given as the mean value. The relative expression of each mRNA was calculated by the comparative Ct method, using ΔCt (the value obtained by subtracting the Ct value of placental STS and GAPDH mRNA from the Ct value of each mRNAof cumulus cells). Specifically, the quantity of target mRNA relative to placental mRNA was expressed as 2^-(ΔCt)^. Then, we calculated the ratio of the quantity of STS mRNA to that of GAPDH mRNA in each samples, and used the ratio for analysis as the expression values of STS mRNA of cumulus cells \[[@B12]\]. Statistical analysis -------------------- Pearson correlation analysis was done. P value less than 0.05 was considered statistically significant. Serum analysis -------------- The levels of gonadotropins (FSH and LH) and steroids (estradiol and progesterone) in the serum from all 49 womens were determined by a commercial enzyme-linked immunosorbent assay kit (AIA-600II, TOSOH, Tokyo, Japan). The sensitivity, intra-assay coefficient of variation (CV) (n = 1 0) and inter-assay CV(n = 20) were as follows: FSH: 0.07 mIU/ml, 1.6--2.3%, 4.0--4.9%; LH: 0.08 mIU/ml, 1.8--2.3%, 1.9--2.3%; estradiol: 13 pg/ml, 2.4--5.1%, 3.3--7.7%; progesterone: 0.02 ng/ml, 4.4--9.0%, 5.8--10.1%. Results ======= STS expression was present in all 9 samples of human cumulus cells that were immunohistochemically stained. The cytoplasm of the cumulus cells was stained by anti-human STS polyclonal antibody, as shown in Fig [1A](#F1){ref-type="fig"} and [1B](#F1){ref-type="fig"}. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Immunohistochemical staining for STS of cumulus cells1A: ×400, 1B: ×1000 STS expression was present in all 9 samples of human cumulus cells that were immunohistochemically stained. The cytoplasm of the cumulus cells was stained by anti-human STS polyclonal antibody. ::: ![](1743-1050-2-6-1) ::: There was a statistically significant negative correlation between the serum level of FSH and STS mRNA expression (n = 112, p = 0.018, r = -0.22) (Fig [2A](#F2){ref-type="fig"}). On the other hand, there were no significant correlations between STS mRNA and the serum levels of estradiol (Fig [2B](#F2){ref-type="fig"}), progesterone and LH. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Relationship between expressions of STS mRNA and FSH concentrations of serum. There was a statistically significant negative correlation between the serum level of FSH and STS mRNA expression(n = 105, p = 0.018, r = -0.22) ::: ![](1743-1050-2-6-2) ::: Discussion ========== Follicular maturation and development are complex processes influenced by both intra- and extra-ovarian events that lead to successful ovulation. The follicular fluid concentration of progesterone in mature COC was significantly higher than that in the immature stage \[[@B13]\]. It has been suggested that steroid concentrations in the follicular fluid are associated with oocyte maturation and embryo quality \[[@B14]-[@B16]\]. Vanderhyden et al. \[[@B10]\] raised the possibility that mouse oocyte secrete a factor(s) that inhibits progesterone and stimulates estradiol production by cumulus cells by culturing mouse COC and the complex which oocyte was removed microsurgically. The involvement of the oocyte in the regulation of progesterone production suggested a potentially important role for the oocyte in the prevention of premature luteinization of the follicle. Although growth differentiation factor-9 (GDF-9), GDF-9B and/or bone morphogenetic protein-6 (BMP-6) are likely candidate molecules, the identities of these oocyte-secreted growth factors regulating key ovarian functions (prevention of premature luteinization, enabling cumulus cell expansion, extracellular matrix stability, thus facilitating ovulation) remain unknown \[[@B4]\]. Direct production of gonadal steroids from sulfated androgens may be an important alternative or complementary pathway for ovarian steroidogenesis. The conversion of sulfated adrenal androgens, present in serum at micromolar concentrations in adult women, into unconjugated androgens or estrogens requires STS activity. Although STS activity has been observed in granulosa cells from rats and humans \[[@B17],[@B18]\], the existence of STS in the COC of any species has not been reported. This is the first study to use immunohistochemistry and quantitative RT-PCR to demonstrate that STS is localized and expressed in human COC. Kosmath et al. reported that the proofs of aryl sulphatase activity gave positive results in all cell types of the atretic follicle. A very strong reaction was observed in the granulosa cells and the migrating cells of the atretic follicle. They raised the possibility that sulfatase might increase during atresia \[[@B19]\]. In this study, we performed the quantitative RT-PCR of STS mRNA in cumulus cells, however, we didn\'t analyze between the quantity of STS mRNA expression and the maturation of oocyte. We consider it as the future subject to examine the relationship between them. Schoenfelder et al. \[[@B9]\] reported that bovine COCs were able to secrete several steroid hormones during in vitro maturation (IVM) without the support of follicular granulosa or theca cells and that the COCs possessed some important steroidogenic enzymes, such as 3β hydroxysteroid dehydrogenase (3βHSD) and aromatase. Their results indicated that COCs were able to modulate their steroidogenic environment in vitro. And it is possible that steroidogenesis may include STS. Although rapid morphological and steroidogenic changes are described for granulosa cells \[[@B20]\], in culture granulosa cells responded well to LH and FSH by increased estradiol production besides almost unchanged testosterone and progesterone secretion. Supporting such unaffected progesterone levels, 3βHSD mRNA expression showed no significant changes during gonadotropin treatment. In contrast, the initial increase in aromatase mRNA that was caused by either FSH or LH led to delayed secretion of estradiol, whereas co-treatment induced just slightly increased estradiol levels \[[@B9]\]. These results may indicate an estrus cycle-dependent reactivity of granulosa cells to different gonadotropin ratios, which exist in distinct but changing ratios in vivo. It is likely that gonadotropins control steroidal enzymes, and FSH may control STS expression in cumulus cells. J Bonser et al. \[[@B21],[@B22]\] reported that there was STS activity and the ability to utilize dehydroepiandrosterone sulfate (DHEAS) as a precursor for estradiol production in human granulosa cells, and that LH and estrone 3-O-sulfamate inhibited this activity. But, they were unable to find any consistent effects of FSH on the conversion of DHEAS to other steroids \[[@B21],[@B22]\]. It was reported that interleukin (IL)-1β in human endometrium suppressed STS mRNA and STS activity in stromal cell culture \[[@B23]\]. Another report indicated that IL-1β in rat granulosa cells mediated the inhibition of gonadotropin-stimulated steroidogenesis by modulating 20α-hydroxysteroid dehydrogenase. The report suggested a role for IL-1β in mediating the observed decline of these bioactive hormones \[[@B24]\]. In humans, FSH stimulated IL-1β secretion by mononuclear cells isolated from the peripheral blood of women in the follicular phase \[[@B25]\]. Previous studies may support the present results; that is, serum FSH levels might control and suppress the STS mRNA expression in cumulus cells through cytokines such as IL-1β. There might be higher relationship to intrafollicular rather than serum factors. Originally, we should analyze with the hormone levels in follicular fluid, however, we did not perform it in this study. Additionally, because the number of follicles grown by COH in each patients were different and multiple, the local hormonal environment in each follicles would not be reflected in the serum hormonal levels of estradiol and progesterone. Take these speculations into consideration, there may be some correlations with the other hormones of which no correlation was seen between expression of STS mRNA in this study. In the conclusion, these results have demonstrated for the first time the expression of STS in cumulus cells by immunohistological stainings and real-time RT-PCR. STS expression in cumulus cells may be related to the control of the local steroidal environment in the oocyte. Although serum FSH may control STS mRNA expression from the results of RT-PCR, we think further study is needed on this topic, because the correlation was low. After that, additional research will be also needed to analyze with follicular fluid, to identify the responsible mechanism and to determine the relationship between STS expression in cumulus cells, the maturation of oocyte and quality of the embryo. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Relationship between expressions of STS mRNA and estradiol concentrations of serum. There were no significant correlations between STS mRNA and the serum levels of estradiol (n = 105, NS), progesterone (n = 105, NS)(data not shown) and LH (n = 105, NS) (data not shown). ::: ![](1743-1050-2-6-3) ::: Acknowledgements ================ We wish to thank Miss. Momoko Negishi for her technical assistance. We are grateful to TOSOH Corporation and EIKEN CHEMICAL CO., LTD. for providing us with AIA-600II and enzyme immunoassay Kit. This study was supported by Health and Labour Sciences Research Grants.	PubMed Central	2024-06-05T03:55:55.708664	2005-4-11	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1082910/", "journal": "J Exp Clin Assist Reprod. 2005 Apr 11; 2:6", "authors": [ { "first": "Yukiko", "last": "Otsuka" }, { "first": "Atsushi", "last": "Yanaihara" }, { "first": "Shinji", "last": "Iwasaki" }, { "first": "Junichi", "last": "Hasegawa" }, { "first": "Takumi", "last": "Yanaihara" }, { "first": "Takashi", "last": "Okai" } ] }
PMC1082911	Editorial ========= The spontaneous diversification of science is notably increasing the number of niches in which scientists specialize. Therefore, despite the multidisciplinary profile of current scientific approaches, the speedy enrichment of defined conceptual areas generates multiple \"hot spots\" with a particular interest. To this end, updating the scientific and technological record through the publication of Reviews, requires a detailed coverage of such fields, that while keeping a broad interest need to be of additional value for specialists. Bioproduction science is not distinct from this universal stream. The growing number of host cell systems being explored as factories, the rising types of bioproducts, the continuous improvement of vectors and the gene expression setting-up, the increasing awareness of the host cell physiology and stress responses occurring under controlled production, and the implementation of new technologies for strain selection-improvement and for process control-monitoring are demanding more refined, topic-oriented attention. Therefore, apart from the publication of broad standard Reviews, Microbial Cell Factories is adjusting its policy to encourage the publication of highly focused Reviews (eventually miniReviews), covering core aspects of the Cell Factory concept with self-contained interest. Authors are invited to submit proposals for such concise Reviews, that are to be prepared under the general journal Review style <http://www.microbialcellfactories.com/info/instructions/default.asp?txt_jou_id=10001&txt_mst_id=10044>. Such contributions, whilst of general interest in the bioproduction landscape, will particularly fulfill the changing needs within highly specialized technical and scientific areas.	PubMed Central	2024-06-05T03:55:55.710258	2005-4-6	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1082911/", "journal": "Microb Cell Fact. 2005 Apr 6; 4:10", "authors": [ { "first": "Antonio", "last": "Villaverde" } ] }
PMC1082912	Background ========== The prevalence of spontaneous triplet pregnancy is about 1 in 7000 deliveries, but with the increasing availability of assisted reproductive technologies, the rate of high-order multiple pregnancies has risen dramatically over the last 20 years \[[@B1],[@B2]\]. Although multiple births have increased and most of the reported monochorionic triplet pregnancies have been conceived by in-vitro-fertilisation, the monochorionic triplet pregnancy is rare, and is estimated to be approximately 1 in 100,000 births \[[@B3],[@B4]\]. Triplet pregnancies are at an increased risk for pregnancy complications and have higher perinatal morbidity and mortality rates, such as vascular anastomoses and developmental anomalies. In this report we present a case of monochorionic triamniotic triplet pregnancy with a co-triplet discordant for multicystic lung lesion, suggestive of congenital cystic adenomatoid malformation (CCAM). Case presentation ================= A 26 year-old woman was referred to our maternal and fetal unit for detailed ultrasonographic examination because of triplet pregnancy with threatened abortion at 15 weeks of gestation. Her obstetric history included two first trimester abortions. The patient had taken no medication or drug for ovulation induction. An inquiry into the family history revealed that her mother had delivered triplet babies all of whom died in the early neonatal period. In the present case, the attending obstetrician had performed an ultrasonography at 6 weeks of gestation demonstrating a single, 17 × 20 mm gestational sac (chorion) (Figure [1](#F1){ref-type="fig"}). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Ultrasonographic image of the case, monochorionic triamniotic triplet pregnancy, is demonstrating single chorionic cavity at 6 weeks of gestation (arrows). ::: ![](1742-4755-2-2-1) ::: Detailed ultrasonography at our maternal and fetal unit revealed a monochorionic triamniotic triplet pregnancy. Three thin amniotic membranes and an ipsilon zone were detected. The biometric measurements of the three distinct fetuses were appropriate for 15 weeks of gestation. The parents were informed about the risks of a multiple pregnancy and monochorionic placentation. After three days\' hospitalization, the vaginal bleeding ceased and the patient was discharged to follow-up. The obstetric course was unremarkable until 24 weeks, when ultrasonographic examination revealed that one of the triplet fetuses had developed a multicystic lung lesion suggestive of CCAM, with the largest cyst measuring 10 mm in diameter (Figure [2](#F2){ref-type="fig"}). Until 30 weeks of gestation, the fetuses had appropriate growth, and follow-up of the pregnancy was uneventful except that the fetus with CCAM developed mild polyhydramnios. At 30 weeks, the patient presented with preterm uterine contractions that ceased after tocolysis with nifedipine 60 mg per day. Betamethasone (12 mg × 2 doses in 24 hours) was administered intramuscularly to the mother to promote fetal lung maturation. The patient was readmitted to our unit at 34 weeks for uterine contractions and impaired fetal growth. The size of the lung lesion remained the same at that time. At 35 weeks of gestation, the patient underwent a low-transverse caesarean section and delivered three live male babies with birth weights 1560 g, 1580 g and 1590 g. Apgar scores were 7/9, 6/8 and 7/9 at 1 and 5 minutes, respectively. A single placenta weighing 1080 g and three distinct membranes were demonstrated (Figure [3](#F3){ref-type="fig"}). Pathological examination confirmed monochorionic triamniotic placentation. The postnatal course was uneventful and the patient was discharged four days postpartum. Two newborns were admitted to the neonatal intensive care unit for respiratory distress, the third one died due to sepsis on day 7 postpartum. One of the triplets was discharged healthy 42 days postpartum. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Transabdominal ultrasonography is presenting the co-triplet fetus with multicystic lung lesions, suggesting congenital cystic adenomatoid malformation (arrow) at 24 weeks. ::: ![](1742-4755-2-2-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### The photograph is demonstrating a single placenta with marginal cord insertion. ::: ![](1742-4755-2-2-3) ::: The newborn with multicystic lung lesions developed unilateral pneumothorax which was treated by inserting an intercostal drain, and was discharged from the hospital after one month. Computerized tomography (CT) of the infant at 3 months demonstrated two cystic lesions in the middle lobe of the right lung (25 mm and 15 mm in diameter) (Figure [4](#F4){ref-type="fig"}). A repeat CT of the infant at 6 months showed a 30 mm solitary cystic mass (Figure [5](#F5){ref-type="fig"}). ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Computerized tomography of the infant at 3 months: two cystic masses in the middle lobe of the right lung, suggesting congenital cystic adenomatoid malformation (arrows). ::: ![](1742-4755-2-2-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### A repeat computerized tomography of the infant at 6 months presents a 30 mm solitary cystic mass (arrow). ::: ![](1742-4755-2-2-5) ::: Conclusion ========== The rate of monozygotic triplet pregnancies in a triplet population is estimated as 4.5%, but the fraction of monochorionic triplets remains unknown \[[@B5]\]. Splitting of the zygote at various stages of development leads to monozygotic multiple pregnancy. The mechanism of monozygotic twinning is not clear. But it is well known that obstetric outcome and clinical management of multiple pregnancies depend on chorionicity. Chorionicity can be established in the first trimester with ultrasound by defining the number of gestational sacs or ipsilon zone \[[@B2]\]. In our case we determined monochorionic placentation by demonstrating a single gestational sac at 6 weeks and ipsilon zone at 15 weeks, and by pathological examination of the placenta after delivery. The reported incidence of CCAM is approximately one in 10,000--25,000 pregnancies \[[@B6],[@B7]\]. This abnormality is believed to be the result of hamartomatous change in the tertiary bronchioles or an arrest in the embryologic development between 7 and 15 weeks of gestation \[[@B8]\]. It is observed as cystic mass occupying part or the entire fetal lung, predominantly located in the right hemithorax, with up to 15% of cases having bilateral involvement. Prenatal prognostic features for CCAM include size, laterality, progression or regression of the mass, cardiac axis deviation, presentation with hydrops or polyhydramnios \[[@B6]-[@B8]\]. Partial or complete regression of the pulmonary lesion is possible. Conservative management is suggested in cases of fetal CCAM without significant mediastinal compression, hydrops fetalis or severe polyhydramnios \[[@B6]\]. Because of the rarity of monochorionic triplet pregnancies, there is no established guideline for management. The presence of an anomalous fetus further complicates the management of pregnancy. Feto-feto-fetal transfusion, acardiac fetus and conjoined twins in triplet gestations have been reported. However, to our knowledge, there are no reports of prenatal diagnosis of a monochorionic triplet pregnancy with a co-triplet fetus discordant for CCAM of the lung. In our case, a monochorionic triamniotic triplet pregnancy with a co-triplet fetus discordant for CCAM was managed conservatively until 35 weeks of gestation and three live fetuses were delivered. These findings may help in decision-making and counselling of parents. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= AG, HA and AC were the consulting perinatologists associated with the case. AG drafted the manuscript, HA participated in the design of the manuscript and AC participated in editing of the manuscript. YC was the director of the Maternal and Fetal Medicine Unit and participated in the design and revision of the manuscript. AIT was the director of the Reproductive Medicine Unit and participated in the design and revision of the manuscript.	PubMed Central	2024-06-05T03:55:55.710744	2005-4-8	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1082912/", "journal": "Reprod Health. 2005 Apr 8; 2:2", "authors": [ { "first": "Ahmet", "last": "Gul" }, { "first": "Halil", "last": "Aslan" }, { "first": "Altan", "last": "Cebeci" }, { "first": "Yavuz", "last": "Ceylan" }, { "first": "Ali Ismet", "last": "Tekirdag" } ] }
PMC1082913	Background ========== Monocyte/macrophages are one of the major target of dengue virus and responsible for virus dissemination after its initial entry via the mosquito vector \[[@B1]-[@B3]\]. A detailed study of this early virus-monocyte interaction by electron microscopy has not been performed. Since ultrastructural study is one of the important analysis in the interaction virus-cell, we performed electron microscopy studies in DEN2 virus- infected human monocytes at 1, 2, 4 and 6 hours of culture, in order to get more information regarding to morphological aspects of virus, virus replication, cellular alterations and apoptosis. Results and discussion ====================== Virus particles --------------- After 1 hour of culture numerous virus particles were observed attached to plasma membrane, free in the extracellular space and in cytoplasmic vacuoles inside monocytes. The predominant viral particles in infected monocyte cultures were typical viral particles of 35 to 42 nm in diameter (Figures [1A, 1B, 1C](#F1){ref-type="fig"}). Small number of fuzzy coated viral particles (74 to 85 nm) showed a core similar to the usual dengue particles, but they had an envelope with projections, looking like a fuzzy coat (Figures [1D, 1E](#F1){ref-type="fig"}). Typical DEN2 virus particles observed in this study were similar to those reported in mosquito cell cultures \[[@B4]\]. Similar fuzzy coated virus particles have been described by Barth et al \[[@B4],[@B5]\] in DEN2 Brazilian virus-infected C6/36 cell cultures. DEN2 virus used to infect monocytes was New Guinea C virus strain and isolated from virus-infected C6/36 cell cultures, suggesting that the fuzzy coated viral particles are a common feature of DEN2 virus. In addition, fuzzy coated virus particles have also been detected in other virus infections, but their significance remains obscure \[[@B6],[@B7]\]. The presence of DEN2 virus antigens in the cytoplasm of infected monocytes was also investigated by direct immunofluorescence. Using a monoclonal antibody against DEN2 virus a diffuse and patchy patterns of fluorescence were observed in the cytoplasm (Figure [1F](#F1){ref-type="fig"}). It was also observed small electron dense structures (75 to 105 nm) that we called in this report \"dense particles\" (Figure [2](#F2){ref-type="fig"}). In some instances, these dense particles showed a center similar to dengue virus nucleocapsid covered by membrane layers and an electron dense envelope (Figures [2B, 2C](#F2){ref-type="fig"}). Dense particles could represent viral particles covered by a homogenous electron dense material. Since, it was not observe viral replication ultrastructural features in infected monocyte cultures, the contribution of monocytes to the formation of this viral envelope is unclear. However, electron dense material observed on the dense particles could represent a protein matrix obtained after virus replication on mosquito cells. In this regard, a range of variation in one virus after experimental isolation has been reported in other virus \[[@B6],[@B7]\]. In general extracellular viral particles were found as single particles and viral particles forming aggregates were uncommon. Viruses attached to the cell surface and free in the extracellular space were engulfed by mechanisms of phagocytosis or macropicnocytosis via typical cytoplasmic processes (Figure [3](#F3){ref-type="fig"}). During phagocytosis or macropicnocytosis virus particles were engulfed alone or together with cellular debris, so that, intracytoplasmic vacuoles and vesicles containing viral particles or large phagosomes full of an electron dense matrix, cellular debris and viral particles may soon be found inside the cells (Figure [4](#F4){ref-type="fig"}). These data suggest a passive phase leading to virus inactivation. In this regard, previous reports have shown that human immunodeficiency virus entering human macrophages by phagocytosis is noninfectious \[[@B8]\]. Infection of Kupffer cells by dengue virus resulted in no viral progeny \[[@B9]\] and only a small proportion of the monocyte population supports replication of DEN2-virus \[[@B10]\]. Smooth membrane coated vacuoles containing viral particles, membrane fragments and moderated electron dense material were also observed (Figure [1B](#F1){ref-type="fig"}). In some instances, cytoplasmic vesicles containing one or more viral particles showed disruption of the membrane leading to direct communication of viral particles with the cytoplasm (Figure [1C](#F1){ref-type="fig"}), however, no morphological virus-related structures could be detected free in the cytoplasm. Features related to viral replication such as virus absorption by penetrating the cell membrane or by endocytosis by clathrin-coat vesicles, virion precursors on rough endoplasmic reticulum or its cisternae, inside Golgi complex, cytoplasm free viral cores or viral budding from cell membrane were not observed in monocytes. Previous report has shown that DEN2 virus can persistently infect transformed lymphoblastoid cells keeping an intact morphology without any indication of active viral replication \[[@B11]\]. Our data show no indications of viral replication and the induction of apoptosis (see below) makes monocytes unlikely source of persistent dengue virus infection. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Electron microscope morphological observations of DEN2 virus particles. A) Typical viral particle in the extracellular environment (arrow; bar: 200 nm). B) Viral particles engulfed in an intracytoplasmic vacuole (arrow; bar: 50 nm). C) Membrane disruption of a vesicle containing a virus (arrow; bar: 100 nm). D) Fuzzy coated viral particles occur in the extracellular space (arrows; bar: 200 nm) E) A fuzzy coated viral particle showing an envelope with projections (arrow; bar: 100 nm). F) Immunofluorescence staining of DEN2 viral antigens at 4 h of culture. A diffuse and patchy pattern of fluorescence was observed in the cytoplasm (arrows). × 1000. ::: ![](1743-422X-2-26-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Electron microscope morphological observations of dense particles. A) Dense particles close to the cell surface (arrow; bar: 200 nm). B) Aggregated dense particles in the extracellular space (arrow). Note the nucleocapsid like center and the electron dense envelopes (bar: 100 nm). C) Dense particles showing a nucleocapsid like center surrounded by membrane layers and an electron dense material (arrow; bar: 100 nm). ::: ![](1743-422X-2-26-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Ultrastructural features of DEN2 virus-infected monocytes. Prominent formation of cellular lamellipods (A) and engulfing of virus by macropicnocytosis (B) and phagocytosis (C) are observed 1 hour after infection. Note the presence of virus (arrows) and cellular debris in the extracellular space. (A and C bars: 1 μm; B bar: 500 nm). ::: ![](1743-422X-2-26-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Ultrastructural features of DEN2 virus-infected monocytes. A) DEN2 virus-infected monocytes after 2 hours of infection. Observe the presence of virus particles in the extracellular space, on cellular plasma membrane and inside cytoplasmic vacuoles (arrows; bar: 1 μm). B) Monocyte showing huge empty vacuoles and vacuoles containing nuclear debris and myelin structures at 4 hours of culture (arrows; bar: 500 nm). C) Monocyte showing cytoplasmic phagosomes containing cellular debris and viral particles (arrow; bar: 200 nm). D) A huge vacuole containing numerous viral particles and cellular debris (arrow; bar: 500 nm). ::: ![](1743-422X-2-26-4) ::: Monocyte cultures ----------------- As assessed by electron microscopy, monocytes showed high degree of activation after 1 hour of infection. One of the most prominent features in DEN2 virus-infected monocytes was the intense expression of short and long plasma membrane processes (lamellipods), in most of the cases engulfing virus particles, cellular debris and apoptotic cells (Figure [3](#F3){ref-type="fig"}). Engulfing of extracellular elements by pseudopods was also observed (Figure [3C](#F3){ref-type="fig"}). As consequence of this activity, small and huge intracytoplasmic vacuoles and phagosomes containing cellular debris, virus particles and myelin like structures in various stages of digestion were observed (Figures [3](#F3){ref-type="fig"} and [4](#F4){ref-type="fig"}). In some instances, phagosomes or vacuoles were surrounded by lysosomes. (Figure [5A](#F5){ref-type="fig"}). Our data show similar ultrastructural findings than those obtained from DEN1 virus-infected Kupffer cells at 1 hour of culture \[[@B9]\], suggesting a similar cellular response against DEN virus for monocytes and macrophages. In DEN2 virus- infected monocytes mitochondria increased in number and size (Figure [5B](#F5){ref-type="fig"}) and cytoplasmic structures resembling diverse degrees of mitochondrial alterations (Figures [5C, 5D](#F5){ref-type="fig"}) were found. Mitochondria were observed in association with lysosomal granules and vacuoles containing membranous debris, consistent with mitochondrial digestion by lysosomes. Infected monocytes showed extensive proliferation of endoplasmic reticulum and lysosomal granules (Figure [5E](#F5){ref-type="fig"}). Cytoplasmic projections associated with cellular movement (uropods) were also observed (Figure [5F](#F5){ref-type="fig"}). It was not observed syncytia, however as shown in figure [6](#F6){ref-type="fig"} a curious distribution of monocytes in DEN2 virus-infected cultures was found. Empty spaces were surrounded by monocytes looking like \"acinar\" structures. In some instances, a linear electron dense material occurred between the empty space and monocytes, suggesting a previous presence of biological material in the lumen. These findings could represent a reactive response of monocytes around virus particles, cellular debris or virus-infected cells. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Ultrastructural features of DEN2 virus-infected monocytes at 4 hours. A) Cytoplasmic vacuole containing cellular debris in close association with lysosomal granules (arrows; bar 200 nm). B) Increased number and size of mitochondria in the cytoplasm of monocyte (bar: 1 μm). C) Mitochondrial degeneration: normal mitochondria (1), early step of degeneration (2) and late step of degeneration (3). Lysosomal granule (large arrow; bar: 500 nm). D) Lysosomes (arrows) in association with mitochondria an autophagosome containing probably mitochondrial debris (bar: 200 nm). Intense lysosomal and vesicular accumulation in the cytoplasm (bar: 200 nm). F) Leukocyte locomotion; note the formation of uropods (arrows; bar: 2 μm). ::: ![](1743-422X-2-26-5) ::: ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Ultrastructural features of DEN2 virus-infected monocytes. \"Acinar\" like structure. A and B show empty spaces surrounded by monocytes. (A bar 1 μm; B bar: 2 μm). C) In some instances, a moderated electron dense material also delimited the empty space (bar: 1 μm) D) Inset from C shows a lineal electron dense material (large arrow) delimiting the empty space, beyond viral particles (small arrow) and a monocyte are observed (bar: 200 nm). ::: ![](1743-422X-2-26-6) ::: Cellular Death -------------- After 1 hour of infection, electron microscopy revealed cells with morphological features of apoptosis, however, previous report has shown apoptosis in Kupffer cells \[[@B9]\] after 24 hours of DEN-1 virus infection, suggesting different susceptibility of monocyte and macrophage to virus-induced apoptosis or different viral apoptotic effect depending of DEN virus strain. In this regard, the susceptibility to DEN virus infection depending of the differentiation state of monocytic cells has been reported \[[@B12]\]. Apoptotic cells showed chromatin margination in nuclei, nuclear fragmentation, condensation and retraction of cytoplasm and blebbing and budding phenomena (Figures [7](#F7){ref-type="fig"} and [8](#F8){ref-type="fig"}). Numerous vesicles, some of which appeared to be releasing to the extracellular space were observed (Figures [7D](#F7){ref-type="fig"} and [8E](#F8){ref-type="fig"}). The budding phenomenon observed on apoptotic cells led to the formation of apoptotic bodies containing several types of organelles, including nuclear fragments and high number of vesicles. This could represent a common aspect in virus-induced apoptosis, since the formation of vesicular apoptotic bodies has also been reported in monocytic/macrophage lineage infected with bovine leukaemia virus \[[@B13]\]. Blebbing of the plasma membrane was also observed in apoptotic cells. The surface blebbing has also been described in other viral infections and related to a role in the direct cell-to-cell spread of the virus \[[@B14]\] or associated with increased cellular permeability \[[@B15]\]. Some apoptotic cells showed long cisternae structures alongside with the plasma membrane suggesting cytoplasmic splitting (Figure [8G](#F8){ref-type="fig"}). We have no explanation for this finding, but it could be due to the fusion of neighboring cytoplasmic vesicles. Apoptotic cells also showed bundles of intracellular microfibrils (Figures [7G](#F7){ref-type="fig"} and [7H](#F7){ref-type="fig"}), which resembled the contractile structures observed in fibroblasts and some glomerular cells \[[@B16]\]. These structures could be related to the apoptotic process, since, filamentous material, clumping of tonofilaments and MyD88 protein association with fibrillar aggregates containing beta-actin have been associated with apoptosis and apoptotic bodies formation \[[@B17]-[@B19]\]. Huge phagosomes were observed in the cytoplasm of apoptotic cells (Figure [7E](#F7){ref-type="fig"}), and in some instances, vacuoles containing few viral particles associated with an electron dense material were observed (Figures [8E](#F8){ref-type="fig"} and [8F](#F8){ref-type="fig"}). The presence of phagosomes in the cytoplasm of apoptotic cells suggests previous active phagocytosis. Contrarily to non apoptotic cell only scarce number of vacuoles containing virus and degraded material was observed in apoptotic cells, suggesting that the absorption of products of viral degradation could trigger cell death. Several apoptotic monocytes and apoptotic bodies were ingested by neighboring healthy monocytes leading to the formation of huge vacuolar compartments containing different grades of cellular digestion (Figures [8D](#F8){ref-type="fig"} and [9](#F9){ref-type="fig"}). Apoptosis could avoid the release of viral particles \[[@B20]\] and together with the phagocytosis and digestion of apoptotic cells represent mechanisms to prevent viral progeny \[[@B9],[@B21],[@B22]\]. The ultrastructural apoptosis finding was confirmed by detecting intrachromosomal DNA strand breaks using the TUNNEL assay. Untreated cultures showed low levels of TUNEL positive cells compared to higher levels observed in infected monocyte cultures (Control: 0.9 ± 0.15. Infected at 1 h: 6.2 ± 1.5; 2 h: 6.4 ± 1.8; 4 h: 7.4 ± 2.3; 6 h: 16.8 ± 3.3; mean ± SE) (Figure [8H](#F8){ref-type="fig"}). In addition to apoptosis, a cellular alteration accompanied by cellular swelling, plasma membrane disruption and karyolysis was observed (Figure [10](#F10){ref-type="fig"}). Plasma membrane disruption led to increased amount of swelled organelles and cellular debris in the extracellular space and the formation of \"ghost cells\" (Figures [10C](#F10){ref-type="fig"} and [10D](#F10){ref-type="fig"}), with further engulfing by monocytes (Figure [10E](#F10){ref-type="fig"}). These lysed cells could represent nonphagocytized apoptotic cells that have lost the membrane integrity \[[@B23]\]. Since, noninfected controls or heat-inactivated DEN2 virus-infected monocytes showed scarce number of apoptotic cells, apoptosis seems to be linked to virus infection. We can not rule out the role of apoptosis inducer proteins in the apoptosis observed in this study. In this regard, increased production of Tumor Necrosis Factor has been reported in DEN2 virus-infected macrophages which could lead to apoptosis. \[[@B24]\]. ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Ultrastructural features of apoptotic cells in DEN2 virus-infected monocytes at 4 hours. A) The typical features of apoptosis are observed in several monocytes (arrows; bar: 2 μm). B) Apoptotic cell showing cellular shrinkage, nuclear condensation and bundles of microfibrils (arrows; bar: 500 nm). C) Monocyte with dense remnant nucleus and surface blebbing (arrow; bar 500 nm). D) Apoptotic cell showing intense cytoplasmic vacuolization (bar: 500 nm). E) Phagosome in the cytoplasm of apoptotic cell (arrow; bar: 200 nm). F) Nuclear fragmentation in apoptotic cell (arrow). Note beside a healthy monocyte (bar: 2 μm). G) Segment of apoptotic cell showing numerous bundles of cytoplasmic fibrils (bar: 200 nm). H) Bundles of microfibrils (arrow) in the cytoplasm of apoptotic cell (bar: 100 nm). ::: ![](1743-422X-2-26-7) ::: ::: {#F8 .fig} Figure 8 ::: {.caption} ###### Ultrastructural features of apoptotic cells in DEN2 virus-infected monocytes. A) Apoptotic bodies containing nuclear fragments and several organelles (bar: 2 μm). B) Vesicular apoptotic body formation (bar: 1 μm). C) Vesicular apoptotic body in the extracellular space (arrow). Note a partial engulfing of the apoptotic body by monocyte processes (bar: 500 nm). D) Monocyte showing an engulfed vesicular apoptotic body (arrow) and intense accumulation of phagosomes containing cellular and viral material in several degrees of digestion (bar: 2 μm). E) Cytoplasm of apoptotic cell showing intense accumulation of vesicles and releasing of vesicular contents to the extracellular space (black arrows). Note the presence of a vacuole containing viral particles and electron dense material (white arrow; bar: 200 nm). F) Vacuole containing partial digested viral particles (arrow) in the cytoplasm of apoptotic cell (bar: 200 nm). G) Cisternae formation alongside the plasma membrane (arrows). Note a vesicle close to these formations (small arrow; bar: 200 nm). H) TUNEL staining for apoptosis in monocyte cultures infected for 4 hours with DEN-2 virus. Intense green fluorescence was observed in apoptotic nuclei (arrow). × 400. ::: ![](1743-422X-2-26-8) ::: ::: {#F9 .fig} Figure 9 ::: {.caption} ###### Ultrastructural features of apoptotic cells in DEN2 virus-infected monocytes. Different phases of phagocytosis and digestion of apoptotic cells. A) Engulfment of apoptotic cell (arrow) by a monocyte (bar: 2 μm). B) A huge phagosome containing a morphological intact apoptotic cell (arrow; bar: 1 μm). C and D show phagosomes (arrows) containing a partial digested apoptotic cells (C bar: 500 nm; D bar: 1 μm). ::: ![](1743-422X-2-26-9) ::: ::: {#F10 .fig} Figure 10 ::: {.caption} ###### Ultrastructural features of apoptotic cells in DEN2 virus-infected monocytes at 6 hours. A) Swelling of organelles and membrane compartments in an apoptotic cell (bar: 1 μm). B) Release of cellular content from a swelling apoptotic cell. Note numerous viral particles (arrow) probably already present in the extracellular space (bar: 1 μm). C) Advance phase of cellular swelling (ghost cell) showing disruption of plasma membrane (arrow; bar: 1 μm). D) Ghost cell surrounding by numerous viral particles (arrow; bar: 500 nm). E) Monocyte engulfing cellular debris (arrow). Note the presence of a phagosome containing partial digested cellular material (bar: 1 μm). ::: ![](1743-422X-2-26-10) ::: Conclusion ========== This in vitrostudy indicates that the interaction of DEN2 virus with monocytes results in virus engulfment and apoptosis, suggesting that monocytes may protect against DEN2 virus infection by eliminating the virus particles and virus-infected apoptotic cells and this could be important in the rapid clearance of the initial virus input. Methods ======= Preparation of virus stock and virus titration ---------------------------------------------- DEN-2 virus strain New Guinea C was propagated in C6/36HT mosquito cells that were cultured in Eagle\'s MEM medium containing 10% FBS prior to viral monocyte infection. The virus culture medium was harvested after 5 days of incubation and after removal of cell debris by centrifugation, the virus supernatant was aliquoted and stored at -70°C until used. Virus was titrated by plaque formation assays on VERO cells. Cells were planted at 1 × 10^6^cells / well in 24-well plates and subsequently, serial dilutions of virus were added and the mixtures were incubated at 37°C for 7 days. Afterwards, the plaques were visualized by staining with a dye solution composed of 1% crystal violet. Virus concentrations are given as plaque-forming units (PFU) / ml. Virus stock was free of endotoxin as determined by limulus amebocyte lysate assay. Monocyte cultures ----------------- Monocytes were isolated from heparinized peripheral blood obtained from human healthy volunteers (N = 5) by density centrifugation over 1.077 Histopaque (Sigma Chemical Co, St. Louis, MO). Healthy individuals were informed about the study procedures and their consents were obtained before enrollment in the investigation following the ethical committee guidelines of the bioethical committee of Medical School (Universidad del Zulia, Maracaibo, Venezuela). Total mononuclear leukocytes recovered from the interface were washed and resuspended in RPMI 1640, 10 % fetal bovine serum and penicillin/streptomycin. Afterwards, 300 μl / well of a cellular suspension (4 × 10^6^cells / ml) were layered on 8 -well plastic chamber slides (Nunc, Roskilde, Denmark) or 10 ml on 75 cm^3^tissue culture flasks and incubated for 3 hours at 37°C and 5% CO~2~. Non adherents cells were washed out with warm medium and adhered cells were used for experiments. Infection of monocyte cultures ------------------------------ Monocytes were infected with a virus concentration of 4 × 10^4^PFU / ml (MOI: 0.08) and incubated for 1, 2, 4 and 6 hours at 37°C and 5% CO~2~. Controls represent monocytes cultured with supplemented medium without virus. In addition, monocyte cultures were incubated with heat inactivated dengue virus (56°C, 30 min.) at 4 × 10^4^PFU / ml for 6 hours. Electron microscopy studies --------------------------- Monocytes planted on 75 cm^3^tissue culture flasks were incubated for 1, 2, 4 and 6 hours with DEN-2 virus (4 × 10^4^PFU/ml). Afterwards, cells were detached by incubation with a solution of 0.01% EDTA and by using a cell scraper. After centrifugation, infected monocytes and controls were fixed with 2% glutaraldehyde in 0.1 M cacodylate buffer, pH 7.3. Cells were postfixed with 1% osmium tetraoxide, dehydrated in a series of ethanol and embedded in Epon 812. Samples were cut into ultrathin sections, stained with uranyl acetate followed by lead citrate and examined in an electron microscopy JEM 1010 (Jeol, Japan). Direct immunofluorescence for DEN-2 antigens -------------------------------------------- Experiments were performed in 8-well plastic chamber slides. Monocytes were infected by incubation with DEN-2 virus as described above. Monocytes were washed in PBS and fixed with cold acetone for 5 minutes. Intracellular viral antigens were detected by a direct immunofluorescence assay using a fluorescein-conjugated DEN-2 virus-specific monoclonal antibody (CDC, Fort Collins, CO. USA). TUNNEL assay ------------ The method for nick end -labeling of apoptotic cells was adapted from that of Gavrieli et al. \[[@B25]\] with a commercial kit (Pharmigen, San Diego, CA). Adhered monocytes were treated according to the protocol provided with this kit. The assay is based on the preferential binding of the FITC-dUTP by terminal deoxynucleotidyl transferase to 3\' OH ends of the DNA. Positive apoptotic nuclei were assessed by fluorescence microscopy (Axioskop, Zeiss, Germany). Competing interests =================== The authors certify that they have not entered into any agreement that could interfere with their access to the data on the research, or upon their ability to analyze the data independently, to prepare manuscripts, and to publish them. Authors have not any conflicts of interest. Authors\' contributions ======================= JM designed, coordinated and draft the manuscript. JPH performed the ultrastructural procedures. NV, LME, GA performed TUNEL assay, virus isolation, monocytes cultures. All authors read and approved the final manuscript. Acknowledgements ================ We thank Dr. Dwane Gubler (Fort Collins, Centre for Disease Control, Colorado) for the monoclonal antibody anti-dengue virus type 2 that made the viral immunofluorescence studies reported here possible.	PubMed Central	2024-06-05T03:55:55.711730	2005-3-31	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1082913/", "journal": "Virol J. 2005 Mar 31; 2:26", "authors": [ { "first": "Jesus A", "last": "Mosquera" }, { "first": "Juan Pablo", "last": "Hernandez" }, { "first": "Nereida", "last": "Valero" }, { "first": "Luz Marina", "last": "Espina" }, { "first": "German J", "last": "Añez" } ] }
PMC1083412	Background ========== Latent tuberculosis infection (LTBI) is defined as a clinical condition without clinical or radiological signs of active disease, and it is manifested only by a positive tuberculin skin test \[[@B1]\]. Approximately two billion people, or one-third of the world\'s population, have LTBI, and approximately 10% of them will develop active TB during their lifetime. It is also estimated that one-third of all new cases of active TB (about 2.5 million cases reported annually in the world), results from activation of LTBI. The role of LTBI is even greater in persons co-infected with HIV. Currently, 11 million people in this category are registered throughout the world, and at least 10% of them develop active TB every year. All these facts have placed the problem of diagnosis and effective treatment of LTBI in the frontier of current TB research agenda. There is plenty of evidence that the basis for LTBI in humans is persistence of tubercle bacilli in vivofor long periods of time. This status is currently defined as dormancy or non-replicating persistence (NRP) emphasizing that each of these terms is only descriptive, regardless of the mechanisms involved \[[@B2]\]. An important element of this definition is that the suggested terms indicate not only the constant number of tubercle bacilli over time (stationary phase of growth in vitro), but actual lack of multiplication. Besides LTBI, a part of bacterial population in patients with active disease also persists in a non-replicating state, and there is an obvious need for a drug affecting this sub-population -- a drug that would be considered for incorporation into treatment regimens. The in vivofactors that may induce NRP include depletion of nutrients, shifts in pH, accumulation of growth inhibiting products, and depletion of oxygen. Discovery of effective drugs against non-replicating tubercle bacilli requires appropriate in vitromodels of dormancy. In the past, the following in vitromodels were suggested: cultivation at 8°C for 31 days \[[@B3]\], nutrient depletion \[[@B4]\], depletion of oxygen in vigorously shaken broth cultures \[[@B5]\], anaerobic model in submerged broth cultures \[[@B6]\], and cultivation at pH 4.8--5.0 in broth for six weeks \[[@B7]\]. Not all of these models fully comply with the above definition of dormancy, and the most popular among them became the anaerobic model by Wayne \[[@B6]\]. The physiological parameters of NRP of M. tuberculosisin this model were studied intensively \[[@B8]-[@B12]\], which makes the anaerobic conditions the most attractive tool for further studies of NRP. The aim of this study was to analyze the activity of all available anti-tuberculosis drugs in broth cultures in in vitroanaerobic conditions. Only capreomycin was found to be bactericidal in these conditions, and the activity of this drug was at the same level as that of metronidazole. Methods ======= Antimicrobial agents -------------------- Metronidazole was obtained from Searle (Skokie, IL). All other drugs and reagents were obtained from Sigma (St. Louis, MO). Stock solutions of rifampin were made in methanol, ethionamide -- in DMSO, and all other drugs -- in distilled water. Working solutions were prepared from stock solution in 7H9 broth. Test-strains ------------ Three drug-susceptible strains of M. tuberculosis, H~37~Rv, Erdmann, and Atencio, were kept in aliquots of 7H9 broth culture frozen at -70°C. The inocula were prepared by 5--7 days of sub-cultivation in fresh 7H9 broth tubes (4.5 ml) placed in a roller drum for better aeration. These cultures were used as inocula when the turbidity became equal to the optical McFarland standard \#1, which corresponded to about 10^8^CFU/ml. Inoculum for hypoxia in vitromodel ------------------------------------ The so-called \"anaerobic inoculum\" of tubercle bacilli was prepared according to the original description \[[@B6],[@B13]\]. According to these reports, with conditions of gradually decreased supply of oxygen, the tubercle bacilli undergo an orderly metabolic down-shift into a state of dormancy when the bacteria accumulate in the bottom of the unshaken tubes (20 × 125 mm) containing 10 ml of broth. In our experiments, a bacterial suspension (optical density at \#1 McFarland standard) was made from 10--14 day-old 7H11 agar culture. After homogenization with glass beads, the suspension was placed into 25 × 125 mm tubes, 4 ml in each, and an additional 6.0 ml of fresh 7H9 broth was added. After cultivation at 37°C for 19 days in an undisturbed upright position, the caps were loosened by 1/2 turn, and the tubes were placed into an anaerobic jar for another nine days of cultivation. The jars contained indicator strips and GasPackPlus envelopes filled with 10 ml of sterile water. Sediments were removed from tubes by inserting pipettes through supernatants, and then combined into one tube, approximately 8 ml from each four tubes set. Medium for anaerobic cultivation -------------------------------- Standard 7H12 broth in 12B Bactec vials (4.0 ml) was supplemented with three aqueous filter-sterilized reagent solutions, each added in a volume of 0.1 ml per vial. Two of them were oxygen-reducing reagents: Sodium thioglycolate and L-cysteine hydrochlorate. The third was methylene blue, a color indicator of oxygen presence. The final concentration was 0.5 mg/ml of each of these reagents in the medium. Hypoxia in vitromodel: experiments with drugs ----------------------------------------------- We used the Bactec-460 system set in an anaerobic mode for testing activity of drugs in anaerobic conditions. After addition of the above-described reagents, the bottom edges of the caps of the12B vials were sealed by coating with rubber cement, the drug solutions were added (0.1 ml per 4.0-ml vial), and the vials were run through the Bactec machine with an anaerobic setting using Nitrogen gas + 5% CO~2~. The vials were inoculated with the described above bacterial suspension (0.1 ml per vial). The vials were subsequently incubated at 37°C, and checked daily for a possible change in color. Those with blue color (disruption of anaerobic conditions) were discarded. The vials were run through the anaerobic Bactec-460 system every four days to record the cumulative Growth Index (GI). To determine the number of viable bacteria in these cultures, we took samples from alternate vials, starting at day zero, and subsequently every week, for plating appropriate dilutions on 7H11 agar plates. The results were expressed in CFU/ml at each time-point. Statistical analyses -------------------- We have determined the regression coefficients for each of the curves of decline in bacterial counts (CFU/ml) in the presence of all drugs, as well as in the drug-free controls. For each coefficient we determined confidence limits for the 0.95 probability, as well as its level of statistical significance (difference from the \"no-decline\") expressed in the pvalue. In addition, we determined the statistical difference between the regression coefficients for each drug, on one hand, and the same values found for drug-free controls, also expressed as a pvalue. Results ======= Model validation ---------------- We evaluated the kinetics of GI and CFU/ml in three preliminary experiments (with three strains) using vials without any drugs added and in the presence of metronidazole. Cultivation in these preliminary experiments continued up to 29 days, and samples for CFU/ml determination were taken from alternate vials on days 0, 4, 8, 11, 15, 18, 22, 25, and 29. During this period of time, production of CO~2~in cultures remained at a steady low level (with daily GIs less than 10), when cumulative four-day readings on the Bactec-460 machine did not exceed GI = 40. The number of CFU/ml in drug-free vials remained relatively steady during this period, with only a small decline within the first 18 days. The kinetics of CFU/ml in vials containing metranidazole was different: a decline could have been detected at day 8, but it became more evident (by one-to-two log~10~) at day 15. This decline in the number of CFU/ml was more significant in the presence of 32 μg/ml of metronidazole than at 8.0 μg/ml. This difference compared to the drug-free medium and between the effects of two drug concentrations became greater during the subsequent period. Nevertheless, we found these differences within the first 15 days of cultivation quite sufficient and less laborious for the subsequent 15 validation experiments (five with each of the test-strains), in which the number of CFU/ml was determined on days 0, 8, and 15. In addition, we incorporated two more agents along with metronidazole (rifampin 0.5 μg/ml and isoniazid 0.5 μg/ml) into these experiments. The summary of these experiments, presented in Table [1](#T1){ref-type="table"}, demonstrated results similar to that in the original report regarding activity of these agents in anaerobic conditions (13): clear dose-dependent bactericidal effect of metronidazole, very slight effect of rifampin, and no effect of isoniazid. The results of one of these experiments is also shown in Fig. [1](#F1){ref-type="fig"}. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Antimicrobial Effect of Metronidazole (Met), Rifampin (RMP) and Isoniazid (INH)Against Three Strains of M. tuberculosis ::: Drug/ Conc. (μg/ml) Log~10~of CFU/ml on days 0, 8 and 15 --------------------- -------------------------------------- ------ ------ ------ ------ ------ ------ ------ ------ Control 6.76 6.30 6.04 7.26 7.49 7.34 7.08 6.63 6.30 Met. 32.0 6.76 5.18 4.20 7.26 5.73 5.76 7.08 4.70 4.18 Met. 8.0 6.76 5.70 5.72 7.26 6.49 6.20 7.08 5.88 5.18 RMP 0.5 6.76 6.11 5.91 7.26 6.50 6.50 7.08 5.76 5.59 INH 0.5 6.76 6.00 5.30 7.26 7.00 7.00 7.08 6.00 5.87 ::: ::: {#F1 .fig} Figure 1 ::: {.caption} ###### One of experiments confirming the validity of the anaerobic model: kinetics of CFU/ml of M. tuberculosison days 0, 8, and 15 in 7H12 broth in anaerobic conditions: in drug-free control, in the presence of metronidazole (Met, 32 and 8 μg/ml), rifampin (RMP, 0.5 μg/ml), and isoniazid (INH, 0.5 μg/ml). ::: ![](1476-0711-4-6-1) ::: Screening of other drugs ------------------------ In subsequent experiments, each drug was tested in three concentrations (μg/ml): amikacin (8, 4, 2), capreomycin (8, 4, 2), ciprofloxacin (4, 2, 1), clarithromycin (32, 8, 2), ethambutol (8, 4, 2), ethionamide (4, 2, 1), gatifloxacin (2, 0.5, 0.12), isoniazid (2.5, 0.5, 0.1), levofloxacin (2, 1, 0.5), PAS (25 and 6.25), pyrazinamide at pH 6.0 (900, 300, 100), morphazinamide (300 and 100), rifabutin (0.5, 0.25, 0.12), rifampin (0.5, 0.25, 0.12), rifapentine (0.12, 0.06, 0.03), streptomycin (8, 4, 2), sparfloxacin (2, 1, 0.5), thiacetazone (1.2, 0.3, 0.075). Other drugs were tested in single concentrations: amoxicillin, augmentin, clindomycin, ceftriaxone, dapsone, doxycycline, erythromycin, cefoxitin, gentamycin, imipenem, minocycline, pristomycin, sulfamethoxazole, tetracycline, trimethoprim. Each experiment for testing two or three of the above listed drugs always included a drug-free control and metronidazole. The analyses included only those experiments in which clear difference in the kinetics of CFU/ml between drug-free controls and metronidazole have been observed. Other criteria for inclusion were: no appearance of blue color in the media, and no greater than GI = 50 cumulative readings at days 7--8, 14--15, and 21. These screening experiments clearly demonstrated that none of the above listed agents, except capreomycin (see below), could produce any antimicrobial activity in the model used, contrasting with clear bactericidal activity of metronidazole. The results of some of these experiments are shown in Table [2](#T2){ref-type="table"} and Fig. [2](#F2){ref-type="fig"}. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Kinetics of the Number of Viable Bacteria (Log~10~CFU/ml). Initial CFU/ml Contents: H37Rv -- 6.32, Atencio -- 6.75, Erdman -- 6.52. ::: Drug/ Conc. (μg/ml) Log~10~of CFU/ml on days 7, 14 and 21 --------------------- --------------------------------------- ------ ------ ------ ------ ------ ------ ------ ------ Control 6.04 5.49 4.91 5.30 5.62 4.86 5.90 5.32 4.91 MET, 16 4.43 4.46 3.12 4.54 3.84 3.15 4.04 3.67 3.28 EMB, 8 5.86 5.39 4.91 6.17 5.30 4.71 5.82 5.28 4.72 SM, 8 6.06 5.12 4.67 6.07 5.20 4.31 5.75 5.16 4.46 PZA, 900\* 5.96 5.23 4.45 5.96 5.08 4.16 5.59 5.17 4.50 \At pH 6.2 ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Kinetics of CFU/ml of M. tuberculosisin anaerobic conditions on days 0, 7, 14, and 21 in one of experiments with conventional anti-TB drugs compared to that of the drug-free control and effect of metronidazole (Met. 16 μg/ml): ethambutol (EMB, 8 μg/ml), streptomycin (SM, 8 μg/ml), and pyrazinamide (PZA, 900 μg/ml, at pH 6.0). ::: ![](1476-0711-4-6-2) ::: Activity of capreomycin in anaerobic cultures --------------------------------------------- A series of 29 experiments (six of which included capreomycin) were conducted with assessment of the number of CFU/ml at four time-points: days 0, 7, 14, and 21. Some decline in the number of viable bacteria, at one log~10~or less, occurred in drug-free controls during this period. Metronidazole used in a concentration of 16 μg/ml in these experiments produced a more significant effect, with a difference of 1--2 log~10~or more compared with the number of CFU/ml in drug-free controls at day 21 of cultivation (Table [3](#T3){ref-type="table"}). The effect of capreomycin at a concentration of 8.0 μg/ml in these experiments was similar to that of metronidazole, and clear dose-response relations was seen for this effect (Table [3](#T3){ref-type="table"} and Fig. [3](#F3){ref-type="fig"}). ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Antimicrobial Effect of Capreomycin (CM) at Concentrations of 8, 4, and 2 μg/ml Initial CFU/ml Content: H37Rv -- 7.28, Atencio -- 6.31, Erdman -- 6.88 ::: Drug. Conc. Log~10~of CFU/ml on days 7, 14 and 21 ------------- --------------------------------------- ------ ------ ------ ------ ------ ------ ------ ------ Control 6.54 5.85 5.80 6.25 6.30 4.92 6.78 5.20 5.31 MET, 16 5.29 4.71 3.28 5.37 6.54 3.82 6.08 5.00 3.88 CM, 8 5.72 4.57 3.13 5.56 5.97 3.24 5.68 4.88 3.18 CM, 4 5.85 4.83 3.79 5.60 6.13 3.36 5.65 5.02 3.93 CM, 2 6.17 4.82 3.88 5.83 6.26 5.62 4.71 ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Effect of capreomycin (CM, at 8, 4, and 2 μg/ml) in one of experiments against M. tuberculosis(H37Rv) in anaerobic conditions compared to that of metronidazole (Met, 16 μg/ml) ::: ![](1476-0711-4-6-3) ::: Statistical significance ------------------------ Decline in the number of viable bacteria in drug-free controls was not significant, with the exception being for the strain Erdman in experiments presented in Table [1](#T1){ref-type="table"} and [2](#T2){ref-type="table"} (p\< 0.05). Decline in the number of viable bacteria in the presence of RMP and INH (Table [1](#T1){ref-type="table"}) and EMB and SM (Table [2](#T2){ref-type="table"}) was not significant in experiments with strains H37Rv and Atencio (p\> 0.05). Decline in CFU/ml in the presence of metranidazole 32.0 and 16.0 μg/ml was statistically significant (p\< 0.05) in all experiments, with best results in experiments with the H37Rv strain, showing most significant difference from decline in controls. Capreomycin (Table [3](#T3){ref-type="table"}) was most active against strain H37Rv at concentrations of 8.0 and 4.0 μg/ml (p= 0.002 and p= 0.004, correspondingly) with definitive difference from the curve of decline in the control (p= 0.003 and p= 0.009, correspondingly). Discussion ========== We examined the antimicrobial activity of all anti-TB drugs and of other antimicrobial agents against M. tuberculosisin anaerobic conditions. For this purpose, we used an anaerobic in vitromodel based on the same principles suggested by Wayne for submerged broth culture cultivation (6,13). Preparation of the \"anaerobic inoculum\" of M. tuberculosiswas identical to the procedure described in these reports. On the other hand, to evaluate the activity of various drugs, we used a different technology (the Bactec-460 system) for subsequent cultivation in anaerobic conditions. One of the advantages of this system is that it is less laborious for testing a large number of drugs than the original technique. Running the culture vials through the Bactec-460 anaerobic system with a nitrogen gas flow effectively provided anaerobic conditions, enforced by oxygen-reducing agents and controlled by the color indicator present in each vial. In the past, it was established that daily release of CO~2~, which was automatically recorded in this system as radiometric Growth Index (daily GI), correlates well with the increase in the number of viable bacteria. Therefore, detection of any significant amount of CO~2~released in cultures in any of the vials provided an instant indication of possible growth, and such vials could have been immediately removed from the experiment. The validity of this system was confirmed in a series of experiments to test the activity of metronidazole, rifampin, and isoniazid. Results obtained in these experiments were identical to those reported by Wayne and Sramek \[[@B13]\]. Among all drugs listed in the Methods section, only capreomycin exhibited a clear bactericidal effect similar to that of metronidazole against non-replicating M. tuberculosisin anaerobic conditions, and this finding was confirmed in six experiments with three M. tuberculosisstrains. We previously reported that the MICs of capreomycin in aerobic conditions against actively multiplying M. tuberculosiswere within the range of 1.25--2.5 μg/ml in either liquid or solid (7H11 agar) medium, and the MBC/MIC ratio was equal to 2, similar to that of streptomycin, amikacin, and kanamycin \[[@B14],[@B15]\]. New findings that capreomycin can be active against non-replicating M. tuberculosisrequire further investigation, particularly in appropriate animal models. The unusual phenomenon of such activity also requires further studies in regard to the mode of action of this drug in anaerobic conditions. It is well known that M. tuberculosisin NRP state represent a problem of LTBI, but also can be a part of the bacterial population in new patients with active tuberculosis \[[@B16]\]. Therefore, it is quite possible that capreomycin may not only be effective in patients with LTBI, but also may play an important role as an element of an initial treatment regimen in new patients with active tuberculosis. Either of these possibilities should be investigated in various animal models and subsequently considered for confirmation in clinical trials. Setting up and conducting a large-scale trial with LTBI patients would require a substantial period of observation and can be quite expensive. Therefore, before addressing the issue of treatment of LTBI patients, another option that can be considered is evaluation of the potential effect of capreomycin (in a combination with other drugs) for newly diagnosed patients. This can be achieved in pilot clinical trials using various surrogate markers to determine bactericidal and sterilizing activity of drugs or drug combination \[[@B17]-[@B19]\]. It is premature to define the exact avenues of clinical studies until the needed information on the effect of capreomycin in appropriate animal models becomes available. Conclusion ========== Among the known anti-TB drugs and other tested agents, only capreomycin exhibited a bactericidal effect against non-replicating M. tuberculosisin anaerobic conditions in vitro. Further studies are needed to investigate the mode of action of this drug in such a setting, as well as the efficacy in appropriate animal models, and, perhaps, in pilot clinical trials. Testing in an anaerobic dormancy model of some other polypeptides may lead to the discovery of other drugs effective against non-replicating M. tuberculosis*. Authors\' contributions ======================= LH directed the study, analyzed and interpreted the results, drafted the manuscript. JC contributed to the design of the model, and performed experiments with anti-TB drugs. VP contributed to the standardization of the procedures, especially for quality controls of anaerobic conditions, standardized CFU/ml quantitation, repeated a series of experiments with anti-TB drugs and conducted experiment with other antimicrobials. Acknowledgements ================ This work was supported by NIH Grant No. U19-AI40917. We thank Andrew J. Kittelson for the statistical analyses of the experimental data.	PubMed Central	2024-06-05T03:55:55.713882	2005-4-1	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1083412/", "journal": "Ann Clin Microbiol Antimicrob. 2005 Apr 1; 4:6", "authors": [ { "first": "Leonid", "last": "Heifets" }, { "first": "Julie", "last": "Simon" }, { "first": "Van", "last": "Pham" } ] }
PMC1083413	Background ========== The use of microarray technologies to monitor gene expression in model organisms, cell lines and tissues has become an important part of biological research over the last several years. Even though a number of papers have been published on the analysis of microarray data, particularly on normalization, classification and clustering of data in the last few years \[[@B1],[@B2]\], there is limited work on relation between sequence and expression of gene. In past attempts have been made to establish relation between expression and nucleotide sequence of genes \[[@B2]-[@B8]\]. There are studies, which showed the relationship between gene expression and synonymous codon bias \[[@B9]\]. In the past, methods have been developed to predict the expression level of genes from their nucleotide sequences that is based on observation that synonymous codon usage shows an overall bias towards a few codons called major codons \[[@B9]-[@B11]\]. Cogan and Wolf 2000 studied the relationship between mRNA concentration and codon bias in detail and found strong correlation (r = 0.62) between codon adaptation index and gene expression \[[@B9]\]. Recently, Jansen et al. 2003 \[[@B11]\] studied the two commonly used numerical indices to measure the expression of genes; i) \'codon adaptation index\' (CAI) and ii) \'codon usage\' (CU). They improve the performance of two indices using genome wide yeast expression data (15) and achieve correlation r = 0.63 to 0.70 and r = 0.63 to 0.71 of CAI and CU with gene expression level respectively. These studies indicate that it is possible to predict the expression of genes with reasonable accuracy from its nucleotide sequence. There are studies, which indicates directly or indirectly the correlation between amino acid composition and gene expression \[[@B6]-[@B9],[@B12]-[@B14]\]. The question arises if there is correlation than can we use this knowledge to predict the expression level of genes from amino acid sequence of their protein like nucleotide sequence. The aim of this study is two fold; to understand the correlation between expression level of genes and primary structure of protein at genome level, and to examine whether the correlation between amino acid composition and gene expression is sufficient enough to derive rules for predicting gene expression from amino acid composition of a protein. A systematic attempt has been made to analyze the gene expression data of Saccharomyces cerevisiae(Holstege et al., 1998) to detect the relationship between composition of protein and expression level of gene \[[@B15]\]. We select this data because it was analyzed/used in a number of studies in the past so validation and comparison is easy \[[@B9],[@B11]-[@B14]\]. We compute correlation between percent composition and gene expression level, for each residues and observed significant correlation between percent composition and expression level. This means that it is possible to derive rules from proteins whose expression level is known and these rules can be used to predict the expression of other remaining protein in the same organism in the same condition. Similar trend was observed on gene expression data obtained from Jelinsky and Samson, 1999 study \[[@B16]\]. In this study we used a Support Vector Machine (SVM) to learn from known expression data and to predict gene expression of remaining proteins of an organism in the same condition using composition of protein \[[@B17]-[@B21]\]. Initially we took amino acid composition as input vector for a protein that has 20 features. Then we tried dipeptide composition as input vector for the SVM where total features are 400. These features provide local order of sequence with composition \[[@B18],[@B21]\]. The method was more accurate when dipetide composition was used as a feature instead of amino acid composition. The performance was nearly same when we tried relative composition and dipeptide composition (in reference to overall composition of genome) instead of absolute composition. One of the major applications of microarray technology is functional classification of genes where gene expression pattern is used to recognize the functional class of gene\[[@B8],[@B10]\]. It is based on the fact that genes of similar function yield similar gene expression pattern. Brown et al., 2000 developed a SVM based method for predicting five functional classes of genes based on their gene expression in 79 different conditions \[[@B19]\]. We also developed a method based on SVM for recognition of genes belonging to cytoplasmic ribosomes (One of the class used by Brown et al., 2000) using i) gene expression data (79 features); ii) amino acid composition of proteins (20 features) and iii) combination of two. The overall performance in terms of total cost saving \[S(M)\] was 226, 199 and 234 for gene expression data, amino acid composition and combination respectively. This demonstrates that additional amino acid composition information can improve the performance of functional classification methods based on gene expression data. We also developed a web server that allows one to analysis gene expression data to deduce the relation between expression and composition of residues in protein. This server allows one to train and test the SVM on his or her own gene expression data. Results ======= Length correlation ------------------ We examined the correlation between the length of gene and its expression level. A significant negative correlation r = -0.18 was found between the expression and the length of gene. This means that short sequences are expressed more in comparison to long sequences. In order to understand the relationship between expression level and length, we computed the average expression of genes for different length of its protein sequence (Table [1](#T1){ref-type="table"}). The average expression is almost inversely proportional to average length of genes. A similar trend was observed on two alternate datasets, where length correlation was r = -0.15 and -0.18 for set1 and set2, respectively. These results agree with previous observations where researchers have shown that metabolic systems prefer to express those genes that are less costly \[[@B14],[@B24]\]. As shown in Table [1](#T1){ref-type="table"}, genes having protein length less than 100 amino acids have average expression \[e = 15.58\]. There was slightly higher expression \[e = 2.13\] in genes of length more than 1200 in comparison to genes with length in the range of 800--1200. \[However, the number of genes was only 168 in this range.\] The average expression of genes having up to 200 residues is too high in comparison to long genes. As shown in scatter plot between gene expression and length of protein (Figure [1](#F1){ref-type="fig"}), most of the genes whose expression is higher than the average are small proteins. Correlation between gene expression and protein composition ----------------------------------------------------------- In the first step, we computed the percent composition of each protein corresponding to genes in our reference dataset (3462 genes) using equation 1. Thus we have 20 values (one for each type of amino acid) for each protein. In the next step, we calculated the correlation between composition of a residue and expression level of gene, for each type of residue. It was observed that some residue types have significant positive correlation, while some others have negative (Table [2](#T2){ref-type="table"}). We also computed the correlation for only those genes whose proteins are more than 100 residues in order to see the length effect on correlation. A similar trend was observed except that correlation further improves for residues that have positive correlation and slightly decrease for residues having negative correlation. It is interesting to note that correlation (negative/positive) does not show any relationship with the overall composition of residues in whole genome (Table [2](#T2){ref-type="table"}). Following is a brief analysis of both types of residues. ### Positive correlation We further analyzed the residues (Ala, Gly, Arg and Val), which showed more than 0.2 positive correlations with gene expression. It is interesting to note that, in general, these residues are less costly for metabolism so they may be preferred for efficient metabolism. We examined whether the average correlation of these positively correlated residues are effective for all the range or it is only in a specified range. For this, we computed the average expression level for genes whose protein has percent composition in different range. As shown in Table [3](#T3){ref-type="table"}, the average expression level (E. level) is in increasing order, proportional to percent composition except the range \'1--3\' where E. level is higher than the next higher range. We examined the proteins, which have percent composition of these residues in range 1--3 and found that most of the corresponding genes are small. As we observed above (See Table [1](#T1){ref-type="table"}) that expression level is inversely proportional to the size of gene. Thus, the genes in range \'1--3\' have unusually high expression for these residues. This is the reason why the correlation between percent composition and gene expression improved further for most of the residues, which have positive correlation when we analyzed only proteins having more than 100 residues (Table [2](#T2){ref-type="table"}). We also computed the average expression level in different range of amino acid composition for those genes whose proteins have more than 100 residues. As shown in Table [4](#T4){ref-type="table"}, the average expression level of these residues decreased significantly in the range of 1--3% composition, whereas it was nearly unaffected in higher range of composition. These results show that expression level is proportional to composition of these residues over a wide range. ### Negative correlation As shown in table [2](#T2){ref-type="table"}, some residues (Asp, Leu, Asn and Ser) have a negative correlation with expression level. The expression data of these residues were further analyzed and the average expression level of genes having different percent composition of these residues is calculated. As shown in table [5](#T5){ref-type="table"}, the correlation between percent composition and average expression level is very strong. This shows that the expression for genes of proteins having these residues is not preferred in the cell. In contrast to positively correlated residues, negatively correlated residues showed average expression level as per trend even in the range of \'1--3\'. It is because lower percent composition is usually found in small genes and both lower percent composition of these residues and short length of proteins are preferred in gene expression. We also computed the average expression level for those genes whose protein have more than 100 residues and found that the average expression level was slightly decreased in lower range (Data not shown). Correlation on alternative dataset ---------------------------------- ### Dataset 1 As we did with the reference dataset, we computed the correlation between expression level and percent composition on 2693 genes in the alternate Dataset 1. As shown in Table [6](#T6){ref-type="table"}, most of the residues, which have positive/negative correlation with gene expression in the reference dataset, also exhibited the same trend in the alternate dataset1. Among positive correlated residues, Arg showed very poor correlation on this dataset, whereas this residue showed high correlation in the reference dataset. We examined this residue and other residues, which have positive a correlation. It is interesting that all these residues including Arg showed increasing average expression level with the range of percent composition. ### Dataset 2 One of the objectives of this study is to understand the correlation when environment is changed. Here, correlation was computed between gene expression level and percent composition on 2693 genes in alternate dataset 2, after exposure to the alkylating agent methyl methanesulfonate \[[@B2]\]. Overall, similar trend was observed for genes in both alternate dataset (dataset1 & dataset 2) and reference dataset (See table [6](#T6){ref-type="table"}). As shown in Table [6](#T6){ref-type="table"}, normally, the positiveness (or negativeness) of correlation of a residue was same; only the degree of correlation was different. We examined residues, which have positive correlation. It was interesting that all these residues showed increasing average correlation with the range of percent composition (Table [7](#T7){ref-type="table"}). ### Genes whose expression level changes four fold or more As shown above, the correlation between gene expression level and percent composition changes slightly in case of untreated and treated genes. The reason is that the expressions of a number of genes are unaffected after treatment. Thus we examined only those 325 genes whose expression level change significantly (\> 4-fold). As shown in Table [8](#T8){ref-type="table"}, most of residues, which showed high positive or negative correlation in untreated or reference dataset, lose their correlation in treated genes. In other sense, the expression of those genes after treatment increases significantly which favorable residues dominated earlier. ### Correlation between expression change (EC) and percent composition One of the major objectives of microarray is to determine the effect on gene expression in different conditions. Thus, we computed the correlation between log(EC) and percent composition, where as EC is (Expression of treated genes)/(Expression of untreated genes). As shown in Table [9](#T9){ref-type="table"}, some residues have positive correlation with EC, which means that they increase the expression level in treated case genes. In contrast, some other residues have negative correlation. This is interesting that residues (Ala, Gly and Val) that have positive correlation with expression level of genes have negative correlation with expression changes. In contrast, residues having negative correlation with expression level have positive correlation with expression change. These observations indicate that composition of protein have direct relationship with expression of gene and with the change of expression in different conditions. Development of prediction method -------------------------------- The results shown in Table [1](#T1){ref-type="table"} to Table [5](#T5){ref-type="table"} show that there is a strong relationship between primary structure of proteins and expression level of their genes. Based on the above observation, we made a systematic attempt to develop a method for predicting expression level of a gene from its protein sequence; from microarray data of the same organism in a given condition. Based on protein features, we developed two types of prediction methods; one from amino acid composition and the other from dipeptide composition. ### Amino acid composition In this case we developed a method using percent composition of proteins as input feature of vector dimension 20 (for 20-residues). A SVM was trained on a training dataset using percent composition as input and gene expression level as output. The SVM was trained using regression mode with linear, polynomial and RBF kernel and achieved correlation coefficient r = 0.46, 0.60 and 0.66 respectively, between predicted and observed values of gene expression, when evaluated using 5-fold cross-validation (Table [10](#T10){ref-type="table"}). It is known that SVMs perform better if their input and output values are normalized. As the variation of output (expression level) was very high, we normalize the output. Here, two functions were used to normalize the output values; i) logarithm and ii) square root. The performance of SVM method is shown in Table [10](#T10){ref-type="table"} with these two functions. As shown in Table [10](#T10){ref-type="table"}, performance improved significantly when normalized values were used instead of direct value of the expression. The correlation achieved using logarithm and square root functions is r = 0.67 and 0.70 respectively with RBF kernel. The performance of SVM based method with RBF kernel was best when square root was used as the normalization function. ### Dipeptide composition We also developed a SVM based method using dipeptide protein feature. The results of this method are shown in table [10](#T10){ref-type="table"}. The correlation coefficient r = 0.51, 0.59 and 0.58 were obtained between predicted and observed values of gene expression, when direct, logarithm and square root of gene expression was used as output vector for SVM with a linear kernel. The performance of method was further improved with RBF kernel where correlation reaches to r = 0.66, 0.68 and 0.72 respectively for direct, logarithm and square root respectively. We obtained the best performance at parameters \"-c 10 -g 0.01\" for RBF kernel in regression mode \[[@B23]\]. ### Membrane and non-membrane proteins In this study we also tried to develop method for predicting expression level of membrane and non-membrane protein. First we predict membrane proteins in our dataset using popular program SOSUI, which predict 739 membranes, and 2723 non-membrane proteins. We develop SVM based prediction method (RBF kernel with -c 10 -g 0.01, using dipeptide composition) for membrane and non-membrane protein and achieve correlation 0.49 and 0.75 respectively between predicted and actual expression. It is interesting that correlation prediction was too poor for membrane protein, there are two possible reasons one dataset was two small second the amino acid composition of proteins do not exhibit good correlation with gene expression. It was interesting that predictive performance improves from 0.72 to 0.75 for non-membrane proteins despite data set was smaller than original. We also examine the correlation between amino acid composition and gene expression and found that most of residue shows low correlation for membrane and high correlation for non-membrane proteins (See Table [2](#T2){ref-type="table"}). ### Relative composition In addition to absolute composition (described above) we also tried relative composition. Here input was either relative amino acid or dipeptide composition instead of absolute composition (see Materials and Method). We obtained similar results with relative composition (data not shown). Functional classification of genes ---------------------------------- First we developed a SVM based method for predicting functional class of genes from their expression data (79 features). We adopted the same strategy as described by Brown et al., 2000 except that we only applied for one class (cytoplasmic ribosomes) instead of five classes. We used the SVM\_light package whereas they use their GIST package. The performance of our method in term of TP (true positives), FP (False positives), TN (True Negatives) and cost of saving S(M) \[[@B8]\] on cytoplasmic ribosomes is shown in Table [11](#T11){ref-type="table"}. The total cost saving value S(M) of our method was 226 whereas Brown et al. 2000 achieved S(M) value in range of 224 to 229 using various models. The S(M) value achieved by our method was slightly lower than their highest model since they used fine-tuning of parameters and modified the SVM whereas we use the standard SVM with standard RBF kernel. We also developed a method to classify the genes based on their relative amino (See Materials and Methods) acid composition (20 features) and achieved total cost saving value S(M) of 190 (See Table Table [11](#T11){ref-type="table"}). It is interesting to note that by using simple amino acid composition one can achieve such a high value which is even better than some models based on gene expression data such as tried by Brown et al., 2002. We achieved total cost saving value S(M) of 234 when we simply combined the output of two SVM methods described above. Here we have not tuned any parameter. We simply add the SVM score of two methods. This clearly indicates that amino acid composition information can play a vital role in improving the performance of classification methods based on gene expression data. Web server LGEpred for prediction of gene expression level ---------------------------------------------------------- Based on the method described in this study, we developed a server that provides various services to the user via Internet. ### Data analysis This server allows one to perform various type of analysis on microarray data. This may help users in understanding the relationship between expression of genes and amino acid composition of their proteins. Following is the brief description of options. • Correlation coefficient:This allows the user to compute the correlation between amino acid composition and gene expression from microarray data. The user can generate correlation tables on their microarray data like Table [2](#T2){ref-type="table"} and Table [6](#T6){ref-type="table"}. • Bin-wise analysis:One can compute the average expression of genes whose proteins have amino acid composition in a specified range. Basically, it allows comprehensive analysis on binned data. One can generate the average expression tables like Tables [1](#T1){ref-type="table"}, [3](#T3){ref-type="table"}, [4](#T4){ref-type="table"} &[5](#T5){ref-type="table"}. • Scatter plots of gene expression:The user can generate scatter plots between gene expression and amino acid composition or length of protein using this option. This allows visualization of relation between gene expression and amino acid sequence on their own expression data. An example figure created using LGEpred server is shown in Figure [1](#F1){ref-type="fig"}. It provides an option to the user to plot graph by taking expression level on horizontal or vertical axis. • Specific plots of gene expression:The specific plot not only allows us to generate a scatter plot between expression level and amino acid composition but also allows drawing the average expression of genes which have amino acid composition in a specified range (See Figure [2](#F2){ref-type="fig"}). Using these graphs one can easily detect the relation between expression level and composition in various ranges on their own data. • SVM based prediction method: One of the major features of LGEpred is to allow the users to develop a SVM based prediction method on their own microarray data. This has three major options for the prediction of gene or ORF expression. • Training and prediction:This routine builds a SVM model from users\' microarray data using expression level of genes and sequence of proteins. Then it predicts the expression of unknown genes of the same organism in the same condition from their protein sequence using this SVM model. • Evaluation and prediction: This allows users to evaluate the SVM method developed on users\' microarray data using LGEpred server. The evaluation is very important in the area of prediction because it provides confidence to the user in using the method of their choice. • Prediction from model:This allows users to predict the expression of genes from their protein sequence using SVM model built using the above options of LGEpred server. The aim of this server is to provide tools to the users to analyze their own data. All the analyses shown in this manuscript can be performed using LGEpred server. This will allow users to understand their microarray data in depth. This may be used for analyzing cDNA microarray data also where user can provide the expression change instead of expression level in case of oligonucleotide array. This server will also be useful for detecting which residues are preferred in which conditions and why expression of particular genes changes drastically with change of conditions. Discussion ========== Oligonucleotide array is a powerful technique that allows one to study the expression of large number of genes simultaneously \[[@B1],[@B2]\]. Though it is a powerful approach and allows one to study the behavior of genes of an organism in different conditions, it has its own limitations it is expensive, time consuming and has problems in managing and analyzing data. Despite all technical advances it is difficult to study all genes simultaneously of an organism that have a large number of genes like Human Genome. It is also difficult to obtain consistent values in replicates, particularly of those genes whose expression are close to the resolution limits \[[@B1],[@B11],[@B17]\]. In contrast, all the protein sequences of a large number of organisms are available today. The questions arises whether it is possible that we study only limited number of genes \[or take those genes from an oligonucleotide array data whose value are consistent in all the samples (duplicate/triplicate)\] and use this data to predict the expression level of remaining proteins of the organism in the same condition. This may save a lot of researchers\' time and effort in studying whole set of genes of an organism, particularly like human genome. Though there were many studies in the past analyzing the array data, there is no method to predict the expression level of genes. Recently, a paper has been appeared where they describe the procedure to predict the expression of genes \[[@B10]\]. In this paper, they predict the class of genes (genes having the same type of expression behaviors are kept in the same class). They divided the genes into 49 classes and predicted class with 73% accuracy using microarray data from sequences in the 800 bp upstream of genes. From the best of author\'s knowledge, there is no study, which describes the prediction of gene expression from amino acid sequence of their protein. This study is the first attempt in this direction to predict the expression level of genes from their protein sequences. In this study we took Holstege et al. 1998 as reference data because it is well studied, and clean \[[@B11]-[@B15],[@B24]-[@B28]\]. Even though, previous studies indirectly indicate the correlation between amino acid composition and expression level, detailed direct correlations were not shown \[[@B3]-[@B9]\]. We studied the correlation between amino acid composition and expression level in detail. Interestingly, some residues showed positive correlation and most of them were small residues. This agrees with the concept of efficient metabolism where researchers demonstrated that proteins having higher composition of less costly amino acids are preferred \[[@B12]-[@B14]\]. As shown in Table [5](#T5){ref-type="table"}, this correlation trend was shown for whole the range of amino acid composition. The genes having higher range of composition of the positive correlated residues also have higher average expression level. We also observed high correlation between percent composition and expression level for residues Ala, Gly and Val, as previous studies where they found high composition of those residues made of codon GNN in highly expressed genes \[[@B14]\]. There were some exceptions in the case of lowest range, where the average expression level was higher than that of the genes in the next higher range. In fact, most of the genes, which fall in the lowest composition range, belong to category of small genes. As shown in Table [1](#T1){ref-type="table"}, the gene expression level has negative correlation with the length of sequence, where smaller genes have higher average expression level (See Table [1](#T1){ref-type="table"} and [2](#T2){ref-type="table"}). This is the reason why genes having lower percent composition of positively correlated residues have unusually high expression level. The correlation was increased when we considered only large sequences (\> 100 residues). In case of negatively correlated residues trend was more uniform including lower range because low composition of these residues and small sequence are both preferred. Similar trends were observed when we performed our correlation analysis on alternative data sets. Although the aim of this study is to understand the relationship between expression level and residue composition in normal conditions, we also studied effect on correlation if its conditions are changed. Here, we computed the correlation between expression level and residue composition on alternate dataset 2, which provides expression level of genes when samples are treated with alkylated. Even though the observed correlation between the expression and the composition of the sequence of a specific gene can not be described as a general rule, interestingly, the correlation trend was same; both treated and untreated sets show the same relationship except change in the magnitude of correlation. That is, residues having high positive/negative correlation in dataset 1 showed the same trend in and dataset 2, were same, only magnitude was different. We also analyzed genes whose expression level changes significantly. It is interesting that some residues showed high correlation with fold change of expression. This indicates that in the future it would be possible to predict the gene expression level of proteins in different conditions if we understand in a given condition which residues are favored. These observations suggest more studies in this direction to understand the relationship between gene expression level and primary structure of proteins. The correlation analysis performed in this study indicates that amino acid composition has correlation with expression. This also indicates that similar sequence will have similar level of gene expression. Now, the question is how to utilize these observations to predict the gene expression of unknown proteins of the same organism. One of the standard practice is to use similarity search tools like BLAST and FASTA for searching similar sequence in dataset of known proteins (whose expression level is known) \[[@B29]\]. The major problems with these tools are that i) they fail in the absence of significant similarity, ii) it is difficult to obtain similarity when the length of query and target sequence are very different, and iii) it is difficult to predict expression level from similarity scores. The machine learning techniques (like ANN and SVM) can be used to learn the relationship between sequence and expression level. The major problem with these techniques is that they cannot be used directly because there are many variations in protein sequence length and these techniques require fixed length patterns. Alternatively, one needs to generate fixed length patterns from these proteins to learn the relationship (or derive the rules from) in known data to predict the gene expression level of other proteins of the same organism. It has been shown in the past that composition, pseudo composition, and dipeptide composition of protein can be used as input pattern of fixed length for classification of proteins using machine-learning techniques \[[@B21],[@B31],[@B32]\]. In this study, first we used the amino acid composition as input and gene expression level as output to develop a SVM based method for predicting gene expression level form amino acid sequence of proteins. As shown in Table [10](#T10){ref-type="table"}, we achieved a significant correlation of 0.66 (SVM with RBF kernel) between predicted and observed values of expression level when evaluated using 5-fold cross-validation. It is well known that SVM performs better when its output values are normalized. Here we used two functions (natural logarithm and square root) to normalize expression level(output). These normalization functions, logarithm and square root, increased the correlation from 0.66 to 0.67 and to 0.70 respectively. It has been observed in past studies of protein classification that accuracy of classification improves significantly when dipeptide composition is used as input instead of single residue (or amino acid) composition \[[@B18],[@B21],[@B23]\]. We also observed similar trends in this study; the correlation between predicted and observed values increased from 0.66, 0.67 and 0.66 to 0.66, 0.68 and 0.72 respectively for without normalization, normalization with logarithm and square root, when dipeptide composition was used as input instead of amino acid composition. This is because dipeptide provides information about sequence order between neighbor residues instead of simple composition. We also tried tripeptide but the results did not improve further because a certain number of patterns were never observed. Our results agree with previous observations where they found that dipeptide is better feature than amino acid composition for SVM based classification \[[@B18],[@B21],[@B23]\]. Conclusion ========== The results indicate that there is correlation between expression level and amino acid composition of proteins, which can be exploited to predict the expression level of genes. The correlation between expression level and composition is conditions dependent, which explain the failure of earlier methods of gene prediction based on codon usage and CAI index \[[@B11]\]. In these methods they calculate parameter from expression in one condition and implement for all condition. In our case we proposed condition specific prediction where training and testing is performed in same condition and organism. The expression data is commonly used to classify the genes \[[@B8],[@B10]\]. As far as authors know there is no study, which uses the gene expression and amino acid composition information for classifying or clustering the gene. We made the first attempt in this study and found that combined method performs better than the individual methods (Table [11](#T11){ref-type="table"}). We feel that this approach will improve the performance of existing methods in classification and clustering of genes. The web server LGEpred developed in this study not only allows us to predict gene expression level of proteins of the same organism in the same conditions from its amino acid sequence, but also allows one to understand the relationship between protein sequence and expression level. The server allows the user to compute the following type of correlations; i) correlation between length and expression level; ii) the average expression of genes which has a number of residues of its protein in specific ranges (like 100 to 200); iii) the list of residues which have positive, negative and neutral correlation with expression level; iv) correlation coefficient between residue composition and expression level; and v) the average expression level of each residue when composition is in a specified range. Although the computation level is not very complex or novel, authors feel that it may be very useful for experimental research working in the area of gene expression, because it allows computing the various relations/correlation between protein sequence and expression level from known expression data. This will help the users to detect, which residues are preferred and which are not preferred in their gene expression data, or condition in which they measure the expression. We understand that our method on Holstege data for prediction of expression of genes from protein sequence will only be valid for genes of the same organism in the same conditions. As expression level depends on condition and organism\' it is not possible to develop a general method for predicting gene expression. Thus our server allows users to develop their own SVM based method from their known expression data that can be used to predict the expression of genes of the same organism in the same condition. This is a primary study on limited data. In order to understand relationship in depth in various organisms in various conditions, it is needed to analyze all possible available microarray data. In order to assist researchers working with related subjects, we designed LGEpred server by which they can perform various studies on their gene expression data. Methods ======= Reference expression dataset ---------------------------- In this study, the expression data of Holstege et al. (1998) is used as reference dataset, because its results are obtained from careful averaging of many experiments \[[@B1],[@B11],[@B14],[@B17]\]. All the genes whose expression level is less than 0.5 copies/cell were excluded, because they are close to resolution limits. The final reference dataset contains 3462 genes, whose protein sequences are available in Saccharomyces Genome Database (SGD). Alternate dataset ----------------- In addition to the reference dataset, we also performed analysis on an alternate dataset obtained from Jelinsky and Samson (1999). In this dataset they examined about 6200 Saccharomyces cerevisiae gene transcript levels with two different environmental conditions. We used 2693 genes whose expression level is more than 0.5 copies/cell in a normal condition and corresponding protein sequence is available in SGD. This dataset consists of two sets; one consists of gene expression data in normal condition and the other after exposure to the alkylating agent methyl methanesulfonate. We also used a subset of 325 genes whose expression level changes significantly when treated with alkylating agent. Functional classification of genes ---------------------------------- For developing a classification method, we obtained data from Brown et al., 2000 \[[@B19]\]. This data consists of a set of 79-element gene expression vector for 2467 yeast genes. In our study we used 2465 genes whose protein sequence were available in Saccharomyces Genome Database (SGD). Here, we work only one-class protein cytoplasmic ribosomes, which have maximum number of genes 121. Thus our final dataset consists of 121 genes belonging to cytoplamic ribosomes as positive examples and rest 2344 genes as negative examples. Five-fold cross validation -------------------------- The performance of a computational algorithm is often tested by the cross-validation or jackknife method \[[@B21],[@B22]\]. Due to time constraint we evaluate the performance of our method through 5-fold cross validation procedure. In this validation procedure, the dataset was partitioned randomly to 5 equally sized sets. The training and testing of each classifier was carried out five times using one distinct set for testing and rest four for training. Amino acid composition ---------------------- The information of a protein can be encapsulated in a vector of 20 dimensions using amino acid composition of the protein. The composition was used as input in this study, which provides the global information of protein features in the form of fixed length vector. The amino acid composition is the fraction of each amino acid type within a protein. The fractions of all 20 natural amino acids were calculated by using the following equation ![](1471-2105-6-59-i1.gif) where *comp(i)is the fraction of residue or composition of residue of type i. Riand Nare the number of residues of type i, and total the number of residue in protein i(length of protein) respectively. We calculate percent composition by multiplying fraction of residue comp(i)by 100. Dipeptide composition --------------------- Dipeptide composition was used to transform the variable length of proteins to fixed length feature vectors. Dipeptide composition has earlier been used by Bhasin and Raghava (2004) for protein classification \[[@B18],[@B21]\]. We adopted the same dipeptide composition based approach in developing SVM method for predicting gene expression level of proteins. The dipeptide composition gave a fixed pattern length of 400 (20×20) possible dipeptides (e.g., Ala-Ala, Aal-Cys, Ala-Asp etc.). The dipeptide composition encapsulates information about the fraction of amino acids as well as their local order. The dipeptide composition was calculated using the following equation. ![](1471-2105-6-59-i2.gif) where dpep(i)is fraction or composition of dipeptide type i. Diand Nare the number of dipeptide of type iand number of residues in protein i, respectively. We calculate percent dipeptide composition by multiplying fraction dpep(i)by 100. Relative composition -------------------- The composition (amino acid and dipeptide) described above is absolute composition of proteins. In addition to absolute composition, we also tried relative composition. In this case first, we compute the overall composition (on all proteins of Saccharomyces) each type of residue. Then we compute the relative composition (in reference to overall composition) of each gene using the following equation, ![](1471-2105-6-59-i3.gif) ![](1471-2105-6-59-i4.gif) where rcomp(i), ocomp(i)and comp(i)are relative, overall and absolute composition of amino acid types irespectively. Similarly rdpep(i), odpep(i)and dpep(i)are relative, overall and absolute dipeptide composition of dipeptide irespectively. Normalization of gene expression level for SVM learning ------------------------------------------------------- The gene expression level was normalized to represent on scale of 0 to 10. We used following two functions to rescale the value i) log function, where natural log was used for each gene expression level and ii) sort function, where square root of each expression level was calculated. This normalization is very important in training and testing of SVM for better accuracy. SVM training and prediction --------------------------- In this study, SVM simulation was achieved by using the SVM\_light package \[[@B20]\]. This package enables the user to define a number of parameters and to select a choice of inbuilt kernel functions, including Polynomial, RBF, Linear, and Sigmoid. In this study the regression mode of SVM was used. Let us assume that we have N genes xi ∈ R(i= 1, 2,\..., N) with corresponding target value yi ∈ {target value}(i= 1,2,\..., N). The xi corresponds to the representation of amino acid sequence of the proteins to the SVM. Here, target value is a real value (gene expression level) corresponding to proteins. The dimension of the input vector is 20 for amino acid composition, and 400 for dipeptide composition. The decision function implemented by the SVM can be written as follows: ![](1471-2105-6-59-i5.gif) The value of the α~i~is given by the task of quadratic programming task, maximize subject to 0 ≤ α~i~≤ C, where C is the regulatory parameter controlling the trade off between the margin and training error. Choosing a kernel K for SVM is analogous to the problem of choosing architecture for neural network. In the present work, SVM parameters were all set to default, except the kernel function. In case of functional classification of genes, input vector consist of 79 features, each feature represents gene expression in one condition. The dataset consists of 121 positive examples (cytoplasmic ribosomes) and 2344 negative examples (non-cytoplasmic ribosomes). The positive examples are very few in comparison to the total data that leads to imbalance in the number of positive and negative training examples. Thus it is difficult for the SVM to correctly classify these genes. In order to handle this problem we replicate the positive examples to match with negative examples during the training of the SVM. Brown et al 2000 modified the SVM to handle this problem. Performance measures -------------------- The performance of the method has been assessed by computing the correlation coefficient between the actual value of gene expression (experimentally determined) and the predicted value of gene expression \[[@B23]\]. We computed Pearson\'s correlation coefficient (r), which is the ratio of the covariance between the predicted and experimental values to the product of the standard deviations in the two. ![](1471-2105-6-59-i6.gif) where, X and Y are experimental and predicted value of gene expression respectively. N is the total number of genes in the data set. Authors\' Contributions ======================= GPSR conceived the project and developed the computer programs for calculating correlation between gene expression and protein sequence composition. GPSR also developed prediction method and wrote the manuscript. JHH coordinated the project, analysis the data and refined the manuscript written by GPSR. Acknowledgements ================ We are grateful to anonymous referees for their suggestions. The research reported here was supported in part by : the Ministry of Information and Communication (MIC) \[Foreign Scholar Invitation Program\], the Ministry of Science and Technology (MOST) \[National R&D Program -- Fusion Strategy of Advanced Technologies\], and Korea Research Foundation \[BK21 Program\], of the Republic of Korea. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### An example plot between expression of genes and length of their protein sequence on one subset of reference data (692 genes). It is generated using LGEpred server option \"Standard plot\" using option \"gene expression on the X-axis\". ::: ![](1471-2105-6-59-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### An example plot between expression of genes and percent composition of ALA in their protein. Boxes along the X-axis show the range of composition of ALA and height of the box show the average expression of genes in that range. ::: ![](1471-2105-6-59-2) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### The average expression level of genes according to the length of protein. ::: Length of Protein* Total No. of Proteins Average Expression Level (E. Level) ----------------------- --------------------------- ----------------------------------------- 25--100 59 15.58 100--200 561 8.39 200--400 1168 3.71 400--800 1179 2.51 800--1200 327 1.85 1200 168 2.13 ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### The correlation between percent composition of residues and gene expression level. The residues which have more than +0.2 correlation are shown in by underline and residue having correlation (negative) -0.15 are shown by bold letter. Second column have percent amino acid compositions in whole yeast genome. ::: Amino Acid Overall composition All Genes (3462) Proteins (\>100 residues) Membrane Proteins Non-Membrane Proteins ----------------- ------------------------- ---------------------- ------------------------------- ----------------------- --------------------------- [A]{.underline} [5.49]{.underline} [0.336]{.underline} [0.361]{.underline} [0.401]{.underline} [0.340]{.underline} C 1.26 -0.003 -0.102 0.030 0.000 D 5.58 -0.168 -0.152 -0.147 -0.207 E 6.54 -0.061 -0.045 -0.105 -0.088 F 4.42 -0.122 -0.109 -0.093 -0.107 [G]{.underline} [4.98]{.underline} [0.215]{.underline} [0.217]{.underline} [0.127]{.underline} [0.231]{.underline} H 2.23 -0.052 -0.056 -0.131 -0.048 I 6.56 -0.136 -0.116 -0.091 -0.128 K 7.35 0.166 0.158 -0.117 0.182 L 9.50 -0.208 -0.191 -0.131 -0.212 M 2.08 -0.087 -0.098 -0.003 -0.094 N 6.17 -0.210 -0.220 -0.209 -0.226 P 4.37 -0.064 -0.057 0.039 -0.086 Q 3.96 -0.052 -0.061 -0.072 -0.065 [R]{.underline} [4.43]{.underline} [0.204]{.underline} [0.183]{.underline} [-0.260]{.underline} [0.240]{.underline} S 8.98 -0.152 -0.159 0.013 -0.180 T 5.92 0.008 0.003 0.279 -0.036 V 5.56 0.269 0.298 0.214 0.294 W 1.04 -0.072 -0.077 -0.043 -0.055 Y 3.38 -0.009 -0.018 -0.030 0.018 ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### The number of genes and the average expression level of genes which have percent composition of positively correlated residues (e.g. Ala, Gly, Arg & Val) in different bin/range. ::: Percent Ala Gly Arg Val ---------- ------ ------- ------ ------- ------ ------- ------ ------- 1 -- 3 163 2.59 339 2.96 558 3.77 106 2.93 3 -- 5 1074 2.11 1180 2.43 1741 2.92 941 2.48 5 -- 7 1212 2.80 1193 3.81 802 2.86 1498 3.00 7 -- 9 626 5.36 523 6.55 201 8.03 705 5.37 9 -- 11 246 9.41 158 7.82 58 20.34 171 13.22 11 -- 13 64 15.25 36 12.79 27 17.56 24 16.45 13 -- 15 31 15.76 15 12.73 9 29.74 5 22.04 \>15 35 15.78 7 13.54 3 23.90 2 19.55 \* Total number of genes in this range \\ Average expression level of genes in this range ::: ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### The number of genes and the average expression level of genes (only genes having more than 100 residues) which have percent composition of positively correlated residues (e.g. Ala, Gly, Arg & Val) in different bin/range. ::: Percent Composition Ala Gly Arg Val ------------------------- --------- --------- --------- --------- ------ ------- ------ ------- 1 -- 3 152 2.15 333 2.82 543 3.70 543 3.70 3 -- 5 1063 2.05 1166 2.39 1728 2.86 1728 2.86 5 -- 7 1204 2.75 1176 3.61 798 2.85 798 2.85 7 -- 9 613 4.86 510 6.14 194 7.52 194 7.52 9 -- 11 242 9.19 155 7.39 55 19.58 55 19.58 11 -- 13 61 15.19 33 12.64 23 17.30 23 17.30 13 -- 15 30 15.81 15 12.73 5 29.86 5 29.86 \> 15 32 16.52 7 13.54 2 12.85 2 12.85 \* Total number of genes in this range \\ Average expression level of genes in this range ::: ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### The number of genes and the average expression level of negatively correlated residues. ::: Percent Composition Asp Leu Asn Ser ------------------------- --------- --------- --------- --------- ------ ------ ------ ------- 1 -- 3 260 9.81 20 14.06 249 8.69 31 12.52 3 -- 5 847 3.97 114 12.19 1195 5.21 264 7.85 5 -- 7 1505 3.35 467 6.20 1323 2.96 907 5.39 7 -- 9 657 3.10 1033 4.05 478 1.72 1202 3.41 9 -- 11 117 3.15 1118 3.09 148 1.83 645 2.06 11 -- 13 32 2.38 526 2.42 34 3.15 221 2.77 13 -- 15 10 2.86 151 1.98 11 0.95 103 1.98 15 5 1.22 32 2.57 9 1.14 87 3.16 \* Total number of genes in this range \\ Average expression level of genes in this range ::: ::: {#T6 .table-wrap} Table 6 ::: {.caption} ###### The correlation between percent composition of residues and gene expression level on alternate dataset 1 (untreated) and 2 (treated). The residues in reference dataset having positive correlation are shown by underline and negative correlation by bold letter. ::: Amino Acid Alternate Dataset 1 (Untreated) Alternate Dataset 2 (Treated) ----------------- ------------------------------------- ----------------------------------- ---------------------- ---------------------- [A]{.underline} [0.341]{.underline} [0.348]{.underline} [0.331]{.underline} [0.338]{.underline} C -0.052 -0.062 -0.045 -0.051 D -0.084 -0.083 -0.055 -0.057 E 0.004 0.016 -0.012 -0.004 F -0.072 -0.069 -0.057 -0.052 [G]{.underline} [0.156]{.underline} [0.165]{.underline} [0.173]{.underline} [0.184]{.underline} H -0.075 -0.059 -0.064 -0.051 I -0.075 -0.085 -0.060 -0.071 K 0.070 0.062 0.017 0.007 L -0.157 -0.155 -0.137 -0.131 M -0.046 -0.057 -0.053 -0.060 N -0.162 -0.156 -0.159 -0.158 P -0.026 -0.035 -0.006 -0.010 Q -0.037 -0.049 -0.036 -0.047 [R]{.underline} [0.018]{.underline} [0.022]{.underline} [-0.046]{.underline} [-0.040]{.underline} S -0.102 -0.099 -0.108 -0.109 T 0.029 0.022 0.041 0.032 [V]{.underline} [0.178]{.underline} [0.188]{.underline} [0.193]{.underline} [0.209]{.underline} W -0.056 -0.064 -0.034 -0.043 Y -0.031 -0.040 -0.002 -0.009 ::: ::: {#T7 .table-wrap} Table 7 ::: {.caption} ###### The number of genes and the average expression level of genes which have percent composition of positively correlated residues (e.g. Ala, Gly, Arg & Val) in different bin/range on alternate dataset 1 and 2. ::: Percent Composition Ala Gly Arg Val ------------------------- --------- --------- --------- --------- ------- ------- ------ ------- ------- ------ ------- ------- 1 -- 3 156 2.51 3.30 248 3.61 4.24 464 4.24 5.44 95 3.50 4.95 3 -- 5 784 1.90 2.72 868 2.35 3.09 1317 3.15 4.26 715 2.90 3.59 5 -- 7 927 2.44 3.43 965 3.44 4.47 650 2.40 3.27 1173 2.91 3.76 7 -- 9 521 3.93 5.46 433 5.81 6.74 157 5.08 4.98 572 4.46 5.39 9 -- 11 211 9.90 10.72 122 5.30 5.90 32 12.41 8.29 113 11.02 12.20 11 -- 13 47 18.77 14.10 27 9.54 9.55 13 12.03 8.18 16 15.08 16.01 13 -- 15 22 16.86 13.19 13 14.68 13.59 3 51.07 32.87 2 22.45 15.25 15 16 26.69 21.81 5 20.46 20.50 2 16.75 15.25 2 36.30 25.65 \* Number of genes in this range \\ Average expression level of genes in this range ::: ::: {#T8 .table-wrap} Table 8 ::: {.caption} ###### The analysis of genes in alternate dataset 1 & 2, whose expression changes 4 folds or more when treated with Alkylating agent. Residues showed in reference dataset positive and negative correlation are shown by undeline and bold font respectively. ::: Amino Acid Untreated Genes Genes treated with Alkylating agent ----------------- --------------------- ----------------------------------------- [A]{.underline} [0.350]{.underline} [0.265]{.underline} C -0.029 0.023 D -0.099 -0.040 E -0.070 -0.118 F -0.087 0.017 [G]{.underline} [0.300]{.underline} [0.108]{.underline} H 0.026 -0.011 I -0.073 0.067 K 0.075 -0.035 L -0.200 -0.071 M -0.041 0.027 N -0.232 -0.165 P -0.000 0.028 Q -0.159 -0.109 [R]{.underline} [0.073]{.underline} [-0.137]{.underline} S -0.076 -0.137 T 0.002 0.108 [V]{.underline} [0.202]{.underline} [0.190]{.underline} W -0.131 0.035 y -0.047 0.043 ::: ::: {#T9 .table-wrap} Table 9 ::: {.caption} ###### The correlation between amino acid composition and log (EC) where EC is (Expression of Treated Genes)/(Expression of Untreated Genes). Residues shows positive and negative correlations are shown by bold font and by underline respectively. ::: Amino Acid Correlation with log(Fold change) ----------------- --------------------------------------- [A]{.underline} [-0.184]{.underline} C 0.055 D 0.029 E -0.036 F 0.136 [G]{.underline} [-0.118]{.underline} H -0.014 I 0.158 [K]{.underline} [-0.111]{.underline} L 0.220 M 0.046 N 0.143 P 0.064 Q 0.068 R -0.036 [S]{.underline} [-0.113]{.underline} [T]{.underline} [-0.110]{.underline} V -0.068 W 0.185 Y 0.076 ::: ::: {#T10 .table-wrap} Table 10 ::: {.caption} ###### The correlation between predicted and experimentally determined gene expression is shown. The value shown by bold font is average correlation on 5 sets of data using in 5-fold cross-validation. ::: Normalization Function Linear Kernel RBF Kernel Polynomial Kernel ---------------------------- ------------------------------------------ ---------------------------------------- ---------------------------------------- ---------------------------------------- ---------------------------------------- ---------------------------------------- No Function 0.46(0.44, 0.44, 0.48, 0.47, 0.46\) 0.51(0.49, 0.52, 0.51, 0.52, 0.52) 0.66(0.63, 0.60, 0.62, 0.67, 0.75) 0.66(0.63, 0.63, 0.62, 0.66, 0.74) 0.60(0.57, 0.58, 0.60, 0.61, 0.65) 0.62(0.64, 0.62, 0.58, 0.59, 0.66) Natural Logarithm* 0.57(0.58, 0.52, 0.59, 0.58, 0.58) 0.59(0.58, 0.55, 0.60, 0.60, 0.60) 0.67(0.65, 0.64, 0.68, 0.68, 0.72) 0.68(0.68, 0.65, 0.65, 0.67, 0.73) 0.64(0.63, 0.61, 0.64, 0.66, 0.67) 0.66(0.67, 0.53, 0.65, 0.68, 0.73) Square root 0.54(0.53, 0.51, 0.56, 0.55, 0.54) 0.58(0.57, 0.56, 0.58, 0.59, 0.60) 0.70(0.69, 0.67, 0.68, 0.71, 0.74) 0.72(0.72, 0.68, 0.67,0.72,0.80) 0.66(0.63, 0.61, 0.64, 0.66, 0.68) 0.69(0.67, 0.53, 0.66, 0.68, 0.73) \* Correlation achieved for each set ::: ::: {#T11 .table-wrap} Table 11 ::: {.caption} ###### The performance of classification methods on a set of 2465 yeast genes which consists of 121 cytoplasmic ribosomes genes (positive examples) and 2344 other genes (negative examples). GEM and AACM are gene expression based and amino acid composition based SVM methods (RBF kernel) respectively. The SVM parameter for GEM and AACM were \"-c 10 -g 0.03\" and \"-c 10 -g 0.55\" respectively ::: Method FP FN TP TN S(M) ------------ -------- -------- -------- -------- ---------- GEM 4 6 115 2340 226 AACM 8 22 99 2336 190 GEM + AACM 4 2 119 2340 234 :::	PubMed Central	2024-06-05T03:55:55.715798	2005-3-17	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1083413/", "journal": "BMC Bioinformatics. 2005 Mar 17; 6:59", "authors": [ { "first": "Gajendra PS", "last": "Raghava" }, { "first": "Joon H", "last": "Han" } ] }
PMC1083414	Background ========== The retina is a well-defined portion of the Central Nervous System (CNS) that has long been used as a model for CNS development and function \[[@B1]-[@B4]\]. It is susceptible to a variety of diseases that can lead to vision loss or complete blindness. Most of the unique functions of the retina depend upon its tissue-specific transcript sets, suggesting that a systematic definition of retinal transcripts would be an invaluable approach to understanding retinal cell identities and functions. The complete genome sequences of human and mouse provide a new starting point for understanding specific expressed transcripts, especially the sequences associated with development and disease. Expressed sequence tag (ESTs) databases, are the most abundant resource of gene expression data. Recently, Okazaki et al. used available data on ESTs to establish a comprehensive full-length transcript data base \[[@B5]\]. There have been a number of studies of retinal transcripts \[[@B6]-[@B10]\] and initial databases listing some of the retinal transcripts (\[[@B11]\]; <http://neibank.nei.nih.gov/index.shtml>; \[[@B12]\]; <http://www.umich.edu/~igene/>), but none of these have provided a global view of retinal transcripts. Recently, Blackshaw et al used serial analysis of gene expression (SAGE) identified 1,051 genes that showed developmentally dynamic expression in mouse retina \[[@B13]\]. Schulz and coworkers analyzed a set of retina transcripts from a mixed population of different datasets and suggested that about 13,000 transcripts might describe 90% of the adult retinome \[[@B14]\]. Although a systemic analysis of mouse retina ESTs has recently been reported \[[@B15],[@B16]\], the functional specification of retina transcripts has not. Thus, this study provides a complementary view of mouse retina transcripts. In the present study, we have generated mouse retina ESTs from embryonic day 13.5 (E13.5), postnatal day 1 (PN1), and adult (8 weeks old) and then analyzed about 81,000 ESTs along with other three major mouse retina libraries through an approach that integrates computational and retina tissue-specific microarray data to identify a set of candidate genes highly related to retina-specific function and retinal diseases. 33,737 non-redundant sequences were identified as retina transcript clusters (RTCs) and step by step classified into neural, retinal, late-generated retinal, and photoreceptor-enriched RTCs. This study also provides a comprehensive table of mouse retina transcript profiles that will now allow a better understanding of retinal development and function. Results ======= Purification and specification of mouse RTCs -------------------------------------------- We used a series of computational steps (Fig. [1](#F1){ref-type="fig"}, [Additional data file 1](#S1){ref-type="supplementary-material"} and [2](#S2){ref-type="supplementary-material"}) to clean up and reorganize a total of 81,253 mouse retina ESTs from the NCBI database (October 2002). The starting pool of retina ESTs was generated from the total retina ESTs by subtraction of a population (4,848 or 5.9% of the total) containing repeat sequences, fusion sequences, low sequence quality, vector sequences, very short sequences, mitochondrial sequences, and sequences with no BLAST-hit in the mouse genome (data not shown). 33,737 ESTs from this starting pool of 76,467 were identified by applying a program that searched for non-redundant sequences (Fig. [2A](#F2){ref-type="fig"}). We have termed these Retina Transcript Clusters (RTCs). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Scheme of computational and biological procedures. For computational analysis, five mouse EST sets were used (retina, cerebellum, lung, CNS, and whole EST). All ESTs were filtered through a non-redundant procedure (details in [additional data file 1](#S1){ref-type="supplementary-material"} and [2](#S2){ref-type="supplementary-material"}). The whole set of RTCs representing all retina transcripts was called RTC-All (RTC-A). Then RTC-A were subtracted with CNS and whole mouse ESTs to generate RTC-N and RCT-R pool. Cleaned cerebellum and lung EST pool were used as internal control. In microarray assays, tissues from PN21 mouse retina, brain, and other body regions were used for comparison of gene expression to verify RTC-R. Gene expression profiles from PN21/PN1 retina comparisons represented a set of genes involved in late-born retina development. By comparison of this set with RTC-R, RTC related to late-born retina cell development were generated (RTC-L). PN35 wild-type retina was used to compare with rd1mutant retina at same age. Results from this subtraction represent a set of genes whose expression is associated with rod photoreceptors (RTC-P). ::: ![](1471-2164-6-40-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Purification and classification of mouse retina ESTs. (a) Distribution of retina ESTs in different categories. Total numbers of ESTs in mouse retina libraries; ESTs clustered by UniGene; Numbers of UniGene clusters; ESTs not clustered by UniGene; filtered ESTs for TC classification; retina transcript clusters (RTCs); Numbers of RTCs only represented in ESTs (non-annotated RTCs); Numbers of RTCs represented in mRNA references; and Numbers of RTCs represented in GenScan. (b) The percentage of overlap UniGene clusters between two different tissues or among three different tissues at distinct subtraction grades. (c and d) The coverage rates of known retina disease genes in retina libraries at different EST copy numbers and distinct subtraction grades. photoreceptor, photoreceptor related genes; Retina diseases, known retina disease genes; Enriched RTCs, retina specific and enriched RTCs. ::: ![](1471-2164-6-40-2) ::: Among the RTCs, 12,939 (37.4%) have a reference RNA sequence in the NCBI database <ftp://ftp.ncbi.nih.gov/refseq/>, 4,825 (13.9%) have no reference RNA sequence but do have GenScan \[[@B17]\] predicted sequence information, and 16,494 (47.7%) are pure EST (Fig. [2A](#F2){ref-type="fig"}). The whole set of RTCs was further analyzed by comparing their sequences (BLAST score \> = 100) with 60,770 RIKEN mouse full-length cDNAs derived from multiple tissues \[[@B5]\]. 39.1% of the RTCs were not present in the RIKEN mouse full-length pool, indicating the existence of a substantial population with either unique full-length or uniquely spliced transcripts in retina. To find criteria by which we could define tissue-enriched or tissue-specific transcript clusters we carried out a comparative analysis starting with approximately 106,000 lung and 80,000 cerebellum ESTs. These sequence pools were filtered using the same procedures as for retina to give transcript clusters (TCs) for each tissue. We then collected all the sets of mouse EST data and removed from them any set that would contain a contribution from retina, lung or cerebellum. This gave a set of data containing over 2 million ESTs. We then used this total EST dataset and the individual transcript clusters (TCs) to derive a ratio for the number of copies in each TC compared with the number in the total. Any EST in a TC that was not found in the total was set to a value of 100. An EST with 5 copies in a TC and 5 copies in the total would have a ratio of 1 and an EST with 10 copies in a TC and 5 in the total would have a ratio of 2. This analysis was carried out for each of the three TCs. We then asked at what ratio were ESTs in the RTC not found in either the LTC or the CTC using the subset of sequences for which fell into UniGene clusters. At a ratio of 0.3 and above only 1.79% of RTC were shared with lung and only 3 % with cerebellum (Fig [2B](#F2){ref-type="fig"}). By increasing the ratio to 2, the overlap with lung fell to 0% and with cerebellum fell to 0.3%. We have also examined abundance in the EST pool as a criterion for helping define tissue specific transcripts. As a reference for the validity of these criteria we tested the inclusion of a known set of 47 known photoreceptor genes. Using an enrichment ratio of 2 from the above analysis, we found that 70--80% of these known photoreceptor specific genes were included in the RTCs. By increasing the copy number to 2 or more or to 3 or more, the total pool of RTCs decreased from 9,179 to 2,629 and 826 respectively without any substantial loss of detection of the known photoreceptor specific genes (Fig [2C](#F2){ref-type="fig"}). The pool of 826 RTCs is listed in [Additional data file 3](#S3){ref-type="supplementary-material"}. We also carried out this analysis by setting the copy number to \> 3 and examining the effect of varying the enrichment ratios (Fig [2D](#F2){ref-type="fig"}). At this copy number, increasing the enrichment ratio above 0.5 only changed the coverage and total number of ESTs by small amounts. Trend tests of the reduction of RTC numbers associated with increased EST copy number were significant using photoreceptor related genes (P= 6.8 × 10^-59^) as criteria. Together our analysis suggests that about 80% of tissue-specific or -enriched transcripts can be identified using the two criteria, a copy ratio of at least two in the specific tissue and at least three EST copies. Biological approach in identification of enriched retina transcripts -------------------------------------------------------------------- We also tested the biological robustness of the computational data using mouse retina tissue-specific microarrays (9,216 spots and 7,612 UniGene clusters) for experimental confirmation. Two groups of experiments were designed to detect genes enriched in retina or photoreceptors by microarray analysis. In the first group, total RNA from postnatal day 21 (PN21) retina was isolated and compared with RNA from whole brain or pooled RNA from other organs, including heart, lung, spleen, liver, and kidney from the same animals, (designated as \"body\" in this study). The data from Retina/Body, Retina/Brain were analyzed individually and log2 ratio results of the two experiments are shown as a spot-plot graph (Fig. [3A](#F3){ref-type="fig"}). The Retina/Body ratio is represented along the Y-axis and the Retina/Brain ratio along the X-axis. When setting log2 \> = 1 as positive ratio, a double positive population lying at the upper-right represents retina-enriched genes. Using the same protocol, for the second group of experiments we compared RNA from PN21 retina with RNA from PN1 retina as set 1 (PN21retina/PN1retina) and PN35 wild-type retina with PN35 rd-1(photoreceptor-deficient) retina as set 2 (WTretina/rd1retina). In set 1 we are comparing retinas before and after the generation of rod photoreceptors and in set 2 we were comparing wild type retinas with retinas in which the rods had degenerated. Thus, when setting log2 \> = 1 as positive ratio, a double positive gene population lay at the upper-right and was considered as rod-enriched genes (Fig. [3B](#F3){ref-type="fig"}). The microarray results were used to generate two gene-expression profiles specific for later born neurons of retina and photoreceptors, respectively ([Additional data file 4](#S4){ref-type="supplementary-material"}). ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Microarray approach and confirmation of retina enriched RTCs. (a) Spot-plot graph of microarray data from Retina/Brain and Retina/Body. Log2 ratio is used for comparison. Overlap populations on upper-right were retina specific/enriched genes. (b) Spot-plot graph of microarray data from PN21/PN1 and WT/rd1. Log2 ratio is used for comparison. Overlap populations in upper-right were photoreceptor specific/enriched genes. (c and d) The percentage (bar-graph) of RTCs overlapping between computational assay (RTC-R) and microarray assay at distinct microarray ratio, log2 = 1 (C) and log2 = 1.5 (D). Line represents the numbers of RTCs. X-axis is copy number in same RTCs. ::: ![](1471-2164-6-40-3) ::: Computationally enriched ESTs from retina (RTC-R), subtracted with whole body and neuron ESTs, were compared with the microarray data from Retina/Body and Retina/Brain. The overlap between the two approaches was 70% under the microarray criterion of log2 ratio \> = 1 (Fig. [3C](#F3){ref-type="fig"}, 316 RTCs, P= 6.1 × 10^-5^) and 80% using the criterion of 1.5 (Fig. [3D](#F3){ref-type="fig"}, 148 RTCs, P= 10^-4^). The percentage of overlap increased by increasing the EST copy number (Fig. [3C](#F3){ref-type="fig"} and [3D](#F3){ref-type="fig"}), indicating that the accuracy of the enriched detection in specific tissues depends on the EST copy numbers, reflective of mRNA abundance. Functional clusters of retina enriched RTCs ------------------------------------------- RTCs were further specified as RTC-A (RTC-all retina transcripts), RTC-N (RTC-neuron enriched transcripts), RTC-R (RTC-retina enriched transcripts), RTC-L (RTC-late-born retina cell enriched transcripts), and RTC-P (RTC-photoreceptor enriched transcripts) using the methods described in Figures [1](#F1){ref-type="fig"} and [additional data file 1](#S1){ref-type="supplementary-material"}. The number of RTCs in individual RTC categories is shown in Figure [4A](#F4){ref-type="fig"}. RTCs that were present in the Gene Ontology (GO) database \[[@B18]\] were used for functional clustering. As shown in Figure [4B](#F4){ref-type="fig"} within the Biological Processcategory, a significant population (10% to 20%) in the subcategory Response to External Stimulusis found in RTC-L and RTC-P. The majority of RTCs in this category are the genes related to light response. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### GO analysis before and after RTC subtraction. (a) Distribution of RTCs in distinct categories (details described in text). (b) Functional annotation in distinct categories. A, perception of external stimulus; B, nucleobase, nucleoside, nucleotide and nucleic acid metabolism; C, protein metabolism and modification; D, carbohydrate metabolism; E, ion transport; F, protein transport; G, organogenesis; H, catabolism; I, biosynthesis; J, electron transport; K, energy pathways; L, amino acid and derivative metabolism; M, cytoplasmic transport; N, phosphate metabolism; O, hydrogen transport. (c and d) Subcategories of Biological Processcategory Functional clusters of TCs in lung (blue line), cerebellum (red line), and retina (yellow line) before (C) and after (D) subtraction. (e and f) Functional clusters of RTCs enriched in adult (blue line) or in embryonic phase (red line). Subcategories of Biological Processcategory (E) and of Cellular Componentcategory (F). ::: ![](1471-2164-6-40-4) ::: To determine whether the enriched TCs of different tissues contained different functional clusters, we analyzed the EST populations of lung, cerebellum, and retina that are in the GO database. About 45 functional clusters from the Biological Processand Molecular Functioncategories of the GO database were analyzed before subtraction and about 30 functional clusters were analyzed after subtraction and the results shown in radar graphs. Interestingly, before subtraction, a similar distribution of functional clusters appeared among the three libraries in either Biological Process(Fig. [4C](#F4){ref-type="fig"}) or Molecular Function(data not shown) categories. After subtraction, using the criteria of \> 3 EST copies and an enrichment ratio of 2, the distribution of functional clusters shifted in specific directions in Biological Process(Fig. [4D](#F4){ref-type="fig"}, Table [1](#T1){ref-type="table"}) categories. This suggests that using subtracted subsets for classification of ESTs in individual tissues can provide accurate information about functional specificities. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Comparison of Functional Clusters in Three Tissue Specific or Enriched EST Pools\* ::: Retina Lung Cerebellum ---------------------------------------------------------------------------------------- ------------ ---------- ---------------- GO:0007154 cell communication(total gene numbers) 36 8 16 [GO:0009605 response to external stimulus]{.underline} 19 1 0 GO:0009581 perception of external stimulus 18 0 0 GO:0009628 response to abiotic stimulus 18 0 0 GO:0009314 radiation response 18 0 0 GO:0009416 light response 18 0 0 GO:0009582 perception of abiotic stimulus 18 0 0 GO:0009583 perception of light 18 0 0 GO:0009591 perception of mechanical stimulus 1 0 0 GO:0009592 perception of sound 1 0 0 GO:0007600 sensory perception 17 0 0 GO:0007601 vision 17 0 0 GO:0007605 hearing 1 0 0 GO:0007609 mechanosensory perception 1 0 0 GO:0009607 response to biotic stimulus 2 0 0 [GO:0007155 cell adhesion]{.underline} 4 2 3 GO:0016337 cell-cell adhesion 1 0 1 [GO:0007165 signal transduction]{.underline} 22 4 9 GO:0007166 cell surface receptor linked signal transduction 10 3 4 GO:0007167 enzyme linked receptor protein signaling pathway 1 0 0 GO:0007186 G-protein coupled receptor protein signaling pathway 8 0 3 GO:0007187 G-protein signaling, coupled to cyclic nucleotide second messenger 1 0 1 GO:0008277 regulation of G-protein coupled receptor protein signaling pathway 1 0 0 GO:0007212 dopamine receptor signaling pathway 0 0 1 GO:0007214 gamma-amino butyric acid signaling pathway 0 0 1 GO:0016055 Wnt receptor signaling pathway 1 3 1 GO:0007223 frizzled-2 receptor signaling pathway 0 3 0 GO:0007242 intracellular signaling cascade 8 0 4 GO:0019932 second-messenger-mediated signaling 1 0 1 GO:0007243 protein kinase cascade 1 0 0 GO:0007264 small GTPase mediated signal transduction 1 0 0 [GO:0007267 cell-cell signaling]{.underline} 0 3 5 GO:0019226 transmission of nerve impulse 0 0 4 GO:0007268 synaptic transmission 0 0 4 GO:0001505 neurotransmitter maintenance 0 0 1 [GO:0008037 cell recognition]{.underline} 1 1 0 \* Cell communication category is used for this table. Tissue specific or enriched EST pools are under condition of 66% concentrated in individual libraries and at least three EST copies. There are no overlap of genes among each libraries. ::: We also looked at distinct functional profiles at an early phase (embryo) including E13.5, E14.5, and PN1 and an adult phase, using the criteria of \> 3 copies RTCs and 83% (ratio \> 5) enrichment in one phase compared to the other (Fig [4E](#F4){ref-type="fig"} and [4F](#F4){ref-type="fig"}). We found, that in retina-enriched RTC-pools (826 RTCs, \> 3 EST copies, 2 fold enrichment) the majority of the RTCs were found in both phases (495 RTCs, 59.9%). 34.4% (284 RTCs) were found only in the adult phase, and 5.7% (47 RTCs) were found only in the embryo phase. To examine the different functional clusters in these two phases, categories of the GO database were analyzed. In the biological process category, 18% of RTCs belonged to Cell Organization and Biogenesisclusters and 12% to Cell Cycleclusters in the embryo phase compared with 6% and 5%, respectively, in the adult phase. Conversely, about 24% of RTCs were in the Transportclusters in the adult phase and only 8.5% in the embryo phase (Fig. [4E](#F4){ref-type="fig"}). Although there was the same percentage in Metabolismclusters (Fig. [4E](#F4){ref-type="fig"}) in both the embryo (55%) and adult (56%) phases, detailed analysis showed that about 17% of RTCs were in Biosynthesisclusters in the embryo phase compared with 7.7% in the adult phase (Fig. [4F](#F4){ref-type="fig"}). Interestingly, Lipid metabolism(5.8%) and Energy pathways(3.9%) were significantly higher in adult phase and in embryo phase, respectively. As expected, overall comparison of these phases shows a change from functions characteristic of a proliferating epithelium (such as cell cycle) to those characteristic of a mature retina (such as transport). A comprehensive transcript profile of mouse retina -------------------------------------------------- The computational processes described above classified a total of 33,737 RTCs for their statistical probability of being retina specific or retina-enriched, as listed in [Additional data file 5](#S5){ref-type="supplementary-material"}. Because of the possibility of alternative splicing, we did not combine all UniGene clusters together and kept the non-redundant sequences as our cluster units. Thus, there may be some gene redundancy in the 33,737 RTCs. If we ignore this potential redundancy among the RTCs, 25,673 RTCs can be clustered into 14,618 UniGene clusters (57%) and 8,064 RTCs have not yet been clustered. If we assume the same ratio of RTCs to clusters for un-clustered RTCs, we estimate that about 19,000 to 20,000 genes might expressed in mouse retina from embryonic to adult stages. Seven sets of information are provided in the [additional data file 5](#S5){ref-type="supplementary-material"}. First, a set of basic information including RTC I.D. numbers, Genbank I.D. for reference sequences, locus link numbers, UniGene numbers and descriptions; second, chromosomal locations in the mouse (UCSC, mm2, Feb 2002) and human (UCSC, hg13, Nov, 2002) genomes including start and end points within the UCSC golden-path database; third, information about TC numbers in whole mouse ESTs, whole mouse retina libraries, adult retina, and embryonic retina libraries; fourth, ratio of RTC enrichment compared with whole mouse ESTs, whole ESTs from neuronal libraries, and also a comparison between RTC from adult and embryonic libraries; fifth, RTC enriched patterns under different enrichment criteria; sixth, published SAGE information from Blackshaw et al \[[@B6]\], and seventh, human retina transcript information from RetBase \[[@B7]\]. Using the RTC database, we can identify new genes specifically expressed or enriched in the retina for which there is not yet any biological evidence. For example, a total of 37 known genes with homeodomains from the RTC pool are listed in Table [2](#T2){ref-type="table"}. Some of the genes have been well studied like Crx \[[@B19],[@B20]\], Rax \[[@B21]\], Otx2 \[[@B22],[@B23]\], and Prox1 \[[@B24]\]. Most of these genes, however, have not previously been described in retina, such as Og9x \[[@B25]\], Lhx \[[@B26],[@B27]\], and Onecut \[[@B28],[@B29]\]. Three of the sine oculis (so)/Six family of genes were present in the RTC pool (Fig. [5A](#F5){ref-type="fig"}). ESTs of Six3 and Six6 were 60% to 80% enriched in retina. The patterns have been confirmed by biological studies of expression level \[[@B30]\] and function analysis \[[@B31]-[@B33]\] of these genes. Ten members of forkhead/winged helix family appeared in our RNA pools and only one, Foxn4 is highly concentrated in the retina (Fig. [5B](#F5){ref-type="fig"}). This observation has also been confirmed by a recent study \[[@B34]\]. The genes from those two gene families are highly enriched in the embryo phase. Lhx3 and Lhx4, from the LIM homeodomain gene family, are highly enriched in the adult retina (Fig. [5C](#F5){ref-type="fig"}). Similarly, genes such as those encoding members of the guanine nucleotide binding protein family, the ATP-binding cassette protein family, the voltage-dependent calcium channel protein family, and the potassium voltage-gated channel protein family ([additional data file 6](#S6){ref-type="supplementary-material"}) are also present in the retina enriched pool. We do not have biological evidence for all the genes that have been listed in this comprehensive profile of RNAs expressed in retina, yet our gene-tables can be useful tools for analyzing mouse retina transcripts and can provide an overview of genes involved in function and development of retina. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Information of Known Homeodomain Contained Genes from RTCs ::: Symbl % D/E (%) E/D (%) A C RTCs UniGene Gene Name MmChr HsChr Ref ----------- ------- ------------- ------------- ------- ------- ---------- ------------- -------------------------------------------------------- ----------- ----------- --------- Crx 98.7 99.9 0.1 74 1 BU503524 Mm.8008 cone-rod homeobox containing gene chr7 chr19 y Dlx1 50 0 100 1 2 BG808909 Mm.4543 distal-less homeobox 1 chr2 chr2 y Dlx2 100 0 100 1 0 BG805973 Mm.3896 distal-less homeobox 2 chr2 chr2 n Hhex 5.6 100 0 1 17 BB709075 Mm.33896 hematopoietically expressed homeobox chr19 chr10 n Hmx1 100 0 100 3 0 BE949806 Mm.10104 H6 homeo box 1 chr5 chr4 y Hoxc4 5.9 100 0 1 16 BB283935 Mm.1351 homeo box C4 chr15 chr12 n Hoxc8 20 100 0 1 4 BB283726 Mm.6167 homeo box C8 chr15 chr12 n Irx2 1.4 0 100 2 139 BG801773 Mm.28888 Iroquois related homeobox 2 (Drosophila) chr13 chr5 n Irx3 5.7 0 100 2 33 BE951617 Mm.39039 Iroquois related homeobox 3 (Drosophila) chr8 chr16 n Irx5 6.8 0 100 5 69 BE949849 Mm.101153 Iroquois related homeobox 5 (Drosophila) chr8 y Irx6 100 100 0 1 0 BG298876 Mm.137247 Iroquois related homeobox 6 (Drosophila) chr8 chr16 y Isl1 20.6 38.5 61.5 14 68 BF467775 Mm.42242 ISL1 transcription factor, LIM/homeodomain, (islet-1) chr13 chr5 n Lhx1 17.1 100 0 3 17 BB283776 Mm.4965 LIM homeobox protein 1 chr11 chr17 n Lhx2 9.7 33.3 67.7 7 65 BF462761 Mm.142856 LIM homeobox protein 2 chr2 chr9 y Lhx3 66.7 100 0 12 6 BE986454 Mm.15655 LIM homeobox protein 3 chr2 chr9 n Lhx4 66.7 100 0 2 1 BG297508 Mm.103624 LIM homeobox protein 4 chr1 chr1 n Lhx9 30 0 0 3 7 BE982177 Mm.79380 LIM homeobox protein 9 chr1 chr1 n Nkx6-2 3.8 67.7 33.3 7 175 BE949669 Mm.28308 NK6 transcription factor related, locus 2 (Drosophila) chr7 chr10 n Og9x 80 100 0 4 1 BI736847 Mm.142724 OG9 homeobox gene chr11 n Onecut1 33.3 0 100 1 2 BG805378 Mm.3512 one cut domain, family member 1 chr9 chr15 n Onecut3 100 0 0 1 0 BE995314 Mm.221027 one cut domain, family member 3 chr10 chr19 n Otx2 42.9 100 0 9 12 BG404413 Mm.134516 orthodenticle homolog 2 (Drosophila) chr14 chr14 y Pax6 44.2 65.4 34.6 46 49 BQ930162 Mm.3608 paired box gene 6 chr2 chr11 y Pknox1 4.5 0 0 1 21 BM941536 Mm.87619 Pbx/knotted 1 homeobox chr17 chr21 n Pknox2 37.5 0 100 3 5 BG803156 Mm.41577 Pbx/knotted 1 homeobox 2 chr9 chr11 n Prox1 60 100 0 6 4 BM940687 Mm.20429 prospero-related homeobox 1 chr1 chr1 y Prrx1 14.3 100 0 2 12 BB283141 Mm.3869 paired related homeobox 1 chr1 chr1 n Rax 100 66.7 33.3 12 0 BB642844 Mm.3499 retina and anterior neural fold homeobox chr18 chr18 y Six1 2.6 100 0 1 37 BB283914 Mm.4645 sine oculis-related homeobox 1 homolog (Drosophila) chr12 chr14 n Six3 76.9 0 100 10 3 BG807874 Mm.15630 sine oculis-related homeobox 3 homolog (Drosophila) chr17 chr2 y Six6 86.7 42.9 57.1 13 2 BI990712 Mm.57138 sine oculis-related homeobox 6 homolog (Drosophila) chr12 chr14 y Vax2 50 0 100 1 1 BI989827 Mm.57253 ventral anterior homeobox containing gene 2 chr6 chr2 y Vsx1 100 100 0 4 0 BB642331 Mm.207061 visual system homeobox 1 homolog (zebrafish) chr2 y Zfh4 100 100 0 3 0 BB642530 Mm.41522 zinc finger homeodomain 4 chr3 chr8 n Zfhx1a 10.1 42.9 57.1 8 71 BE954320 Mm.3929 zinc finger homeobox 1a chr18 chr10 n Zfhx1b 10.7 100 0 3 25 BI730214 Mm.37676 zinc finger homeobox 1b chr2 chr2 n Zhx1 75 100 0 3 1 BG404047 Mm.37216 zinc fingers and homeoboxes protein 1 chr15 chr8 n \%, percentage of RTCs distributed in retina A, RTCs; C, filtered mouse ESTs D/E (%), percetage of RTCs in adult phase; E/D (%), percetage of RTCs in embryonic phase MmChr, mouse chromosome; HsChr, human chromosome Ref, genes have been study in retina (y) or not (n) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Distribution of enriched RTCs in gene families. (a) The percentage of enriched RTCs of sine oculis (so)/Six family in retina. (b) The percentage of enriched RTCs of forkhead/winged helix family in retina. (c) The percentage of enriched RTCs of LIM-homeobox family in retina. (d) The percentage of enriched RTCs of nuclear receptor family in retina. ::: ![](1471-2164-6-40-5) ::: From the RTC database, we can also extract identify candidate genes for human retinal diseases. As shown in Figure [2C](#F2){ref-type="fig"}, using the criteria of enrichment ratio \> 2 and at least 3 EST copies in RTC, 826 RTCs were identified as a retina enriched gene pool. By homology with the human genome, this RTC pool covers about 80% of known-photoreceptor related genes (29 out of 37, Fig. [2C](#F2){ref-type="fig"} and [2D](#F2){ref-type="fig"}). Through comparison with individual interval gene numbers, numbers of candidate RTCs are concentrated to 1.5% to 0.05% ([additional data file 7](#S7){ref-type="supplementary-material"}). A full list of known human retina disease genes and the subtraction information are shown in [additional data file 8](#S8){ref-type="supplementary-material"}. Among retina-disease-related genes, almost all photoreceptor related genes especially the genes involved in phototransduction processes are covered by our most stringent subtraction and selection criteria (\> 3 EST copies and 2 fold enrichment). However, genes related to systemic diseases or syndromes, genes involved in RNA processing and genes with lower copy numbers in RTCs are not present under these criteria, indicating a limitation of this method. Although candidate genes from the 826 RTCs pool can cover over 41 of 48 mapped human retina disease loci, here we provide only the genes for 27 loci that have been more stringently associated with primarily photoreceptor disorders ([additional data file 7](#S7){ref-type="supplementary-material"} and [9](#S9){ref-type="supplementary-material"}). Comparison with other approaches for retina enriched gene subtraction shows obvious gaps among the various experimental approaches (Table [3](#T3){ref-type="table"}). Here we have compared our enriched gene pool with the pools from SAGE \[[@B6]\] and RetBase \[[@B7]\]. Only genes clustered by UniGene are used using two criteria from SAGE pool with two standards \[[@B6]\], 2 out of 4 (71 genes) and SAGE 264 (264 genes) respectively; one criteria from Retbase (373 genes, \[[@B7]\]); and one criteria with at least 3 EST copies and 2 fold enrichment in RTCs (184 genes). The highest 22% overlaps between our and SAGE pools and the lowest 5% overlaps between Retbase and SAGE pools. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Comparison of predicted retina enriched genes by different system ::: Sage71 Sage264 RetBase374 RTC184 ------------ ------------- --------------- -------------- -------------- Sage71 71(100) 29(40.8) 5(7) 22(30.9) Sage264 29(10.9) 264(100) 4 (1.5) 15(5.7) Retbase374 5(1.3) 4(1.1) 373(100) 14(3.7) RTC184 22(11.9) 15(8.2) 14(7.6) 184(100) Sage 71, 2 out of 4 (Blackshow et al, 2001) Sage264, (Blackshow et al, 2001) RetBase374, (Katsanis et al. 2002) ::: Discussion ========== The full specification of transcripts in an individual tissue or cell type can greatly aid the understanding of the control of cell differentiation and cell function. In the present study, we have integrated computational bioinformatics and microarray experimental approaches to classify the tissue specificity and developmental distribution of mouse retina transcripts. We have defined 33,737 retina transcript clusters (RTCs) as single units with non-redundant sequences, although multiple transcript clusters can be in the same UniGene cluster. Such RTCs may represent different parts of a gene, or splice variants that are not considered in the UniGene database. We have calculated that about 19,000 to 20,000 genes may be expressed during mouse retina development from embryonic day 13.5 to adult. This is about 30% more compared with Schulz\'s Retinome (13,037 genes) \[[@B14]\], although this was restricted to transcripts in the adult phase. Since the nervous system, including the retina, has more prolific RNA splicing (2.5 fold higher that other tissues) \[35\], we suggest that the numbers of unique transcripts found in retina may well be greater than 20,000. More interestingly, since 39.1% of the RTCs are not included in the most comprehensive mouse full-length transcript set made by RIKEN \[[@B5]\], tissue-specific transcripts and splicing may constitute a substantial proportion of the whole transcriptome. On the other hand, there are large numbers of ESTs (about 16,000) that do not appear to encode proteins. Many of these may serve as regulatory RNA or have other unknown functions and need further study \[[@B5],[@B13]\]. A major concern for specification of transcripts from ESTs is how to confirm a reliable result. We have used four control steps to verify the results. First, as an external negative control we compared the specificity in our target tissue, retina, and in other tissues (lung and cerebellum). Second, we used an internal positive control. We used known photoreceptor-specific genes to check their coverage rate during different steps of retina specification. The third control was to identify functional clusters through the Gene Ontology (GO) database before and after specification of ESTs in different tissues. Biological confirmation was the final control experiment in this study and used a mouse retina tissue-specific microarray. Results from all the control experiments fully support the conclusions of this study. Identification of retina disease genes is an important and immediate use for genome-wide study tools. Several approaches have been used for genome-wide hunting of such genes. However, as shown in our results, the data sets show obvious gaps among these various experimental approaches (Table [3](#T3){ref-type="table"}). Katsanis et al performed subtraction against 1.4 million human ESTs with 40,000 human retinal ESTs by a series of computational tools. They found a total of 925 ESTs likely to be specifically or preferentially expressed in the retina \[[@B7]\]. We found a low overlap between RetBase and our enriched pools. In part this is because of the smaller starting sample, the majority of human ESTs are generated from tumors and the emphasis on single copy sequences in the RetBase set. A SAGE analysis identified 264 uncharacterized genes that were specific to or highly enriched in rods \[[@B6]-[@B9]\]. This data set showed more overlap with our results but there are, still obvious gaps among these various experimental approaches, so we suggest that integrating different approaches might be much more valuable for tissue specific and enriched gene prediction. Generation and specification of an entire transcription profile for individual tissues or cells with specific functions and morphologic identities represents the next major task in the genome era. In conclusion, this study complements and extends previous studies in a number of ways. First, we have generated a comprehensive data set of retina transcript profiles with functional and developmental explanation, as the examples shown in Figure [4](#F4){ref-type="fig"} and Table [2](#T2){ref-type="table"}, diverse function information can be generated through our RTC profiles that will now allow a better understanding of retinal development and function. It is not only for hunting retina disease genes, but also for understanding of gene developmental distribution. Second, this study has classified retina transcripts into different grades of retinal specificities that will help us define them as common, neural specific or retinal specific genes. Third, the distinct approaches of this study will allow an easy updating of our mouse retina transcripts databases in future. Methods ======= EST resources and manipulation ------------------------------ Mouse retina ESTs were collected from NCBI and were filtered and cleaned up by a series of programs (Figures [1](#F1){ref-type="fig"}, S1 and S2). Two Dell Precision WorkStation 530\'s running RedHat 7.0 Linux were used for computational processes. The database used was MySQL and the languages for programming were Python, GNU C, awk, and bash. Repeat detection used RepeatMasker <http://repeatmasker.genome.washington.edu/cgi-bin/RepeatMasker> (kindly provided by A. Smith and P. Green). Other analyses used custom programs. Mice and retina sample collection --------------------------------- Mice were purchased from the Jackson Laboratory (C57Bl/6j) or were a gift from Dr. C. Zeiss, Yale University (C3H wtand rd1). Retinas were dissected without contamination from lens, iris, cornea, and ciliary body. 10 to 20 retinas or other organs were pooled for RNA isolation. All animal experiments were conducted in accordance with NIH guidelines and were approved by the IACUC of Yale University School of Medicine. RNA preparation --------------- Total RNA was isolated by TRIzol (Invitrogen) and purified by RNeasy mini kit (QIAGEN). 5 μg total RNA with 280/260 ratios greater than 1.9 was used for array hybridization without amplification. Three to four sets of RNA were prepared from each age of retina and processed individually for microarray analysis. Microarray experiments ---------------------- About 12,000 non-redundant mouse retina ESTs were generated from about 28,000 ESTs that generated from E13.5, PN1 and adult (The NIH-University of Iowa Brain Molecular Anatomy Project). 9,216 purified PCR-amplified inserts were printed by the Yale Keck Microarray Core on poly L-Lysine (Sigma) coated glass slides utilizing a GeneMachines Omnigrid robotic arrayer (GeneMachines). 3DNA Submicro EX Expression Array Detection Kits (Genisphere, PA) were used for RNA labeling. Detailed microarray experimental protocols are shown in supplemental methods. Slides were scanned on a GenePix 4000B scanner and the data were manipulated with GenePix software Version 4.0 (Axon Instruments). Three or four sets of microarray data for each experiment were used for Student t test and gene collection. Gene collection methods are described in text and also please see [additional data file 10](#S1){ref-type="supplementary-material"} for methods detail. Statistics ---------- To test whether the features illustrated in Figures [2](#F2){ref-type="fig"} and [3](#F3){ref-type="fig"} show an increasing trend in terms of covering photoreceptor related genes and retina disease genes through more stringent filtering criteria, we conducted trend tests in the following form: ![](1471-2164-6-40-i1.gif), where Cis the number of classes above the baseline total population, w~i~is the weight for the ith class, and y~i~is the observed percentage related to the feature of interest in the ith class. The value of C= 2, 4, 3, and 3, for figures [2C, 2D](#F2){ref-type="fig"}, [3C](#F3){ref-type="fig"}, and [3D](#F3){ref-type="fig"}, respectively. The value of y~i~is the proportion of the ESTs with a given feature in the ith class, with the feature being either a photoreceptor related gene or a retina disease gene. The weights are 1, 2, \..., C for the C classes. To assess the statistical evidence of a trend in the data based on T, we calculated the mean and variance of Tunder the null hypothesis of no trend conditional on the feature distribution in the baseline total population. It can be shown that when the feature of interest is binary, i.e. a given EST either has or does not have this feature, ![](1471-2164-6-40-i2.gif) where ![](1471-2164-6-40-i3.gif) N~i~is the number of ESTs in the ith class, N~0~is the number of ESTs in the total population, and y~0~is the proportion of ESTs having a given feature in the total population. The statistical significance of the observed increasing trend is ![](1471-2164-6-40-i4.gif) and the statistical significance of the observed decreasing trend is ![](1471-2164-6-40-i5.gif) where Φ is the cumulative function of the standard normal distribution. Statistic calculations were done by R 1.8.1 or R 1.7.1. <http://cran.us.r-project.org> List of abbreviations ===================== RTC, Retina Transcript Cluster; CNS, Central Nervous System; EST, Expressed Sequence Tag; GO, Gene Ontology; PN, Postnatal Authors\' contributions ======================= SSMZ was primarily responsible for the design, coordination, conduct, and all experiments of the studies. XYF and CJB were responsible for coordination of the studies. XX and JL were responsible for computational data analyses and software development. SSMZ and MGL were responsible for microarray experiments and analysis. MGL and HZ were responsible for statistical analysis. SSMZ, MBS, and XYF were responsible for RNA collection and original initiation and generation of mouse retina ESTs. SSMZ and CJB drafted the manuscript and figures. All authors read and approved the final manuscript. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Methods for in silico purification of ESTs. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 2 Methods for in silico specification of RTCs. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 3 RTC list under criteria of RTC-A/C20A3. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 4 Gene list of RTC-L and RTC-P. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 5 Whole RTC information and SAGE, RetBase data comparison. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 6 Samples of the percentage of RTCs in different gene families. a, Heterotrimeric guanine nucleotide-biding proteins. b, ATP-binding cassette (ABC) transporter superfamily. c, Voltage-dependent calcium channel proteins. d, Voltage-gated potassium channel proteins. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 7 Candidate genes for human retina disease loci. a, Gene numbers of chromosome interval and concentrated retina enriched gene pool in human known retina disease gene loci. b, Concentrated ratio between retina enriched gene pool and whole interval genes of the loci for human known retina disease gene. c, Gene numbers of chromosome interval and concentrated retina enriched gene pool in some human known retina disease loci. d, Concentrated ratio between retina enriched gene pool and whole interval genes of the loci for human retina disease. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 8 Known human retina disease gene list used in this study. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 9 Recommended gene candidates for human retina disease loci. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 10 Additional methods. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ We thank Caroline Zeiss for rd1mice and the Yale Keck Microarray Core (Janet Hager and Kenneth Williams) for microarray printing support. We thank Paul Lizardi, Michael Zhang, and Thomas Welte for comments, Lan Ji and Adrienne LaRue for technical assistance. This work is supported by grants EY13607 (XYF), EY 00785 (CJB) and EY13865 (CJB) from the NIH, the David Woods Kemper Memorial Foundation (CJB), the Connecticut Lions Eye Research Foundation and Research to Prevent Blindness Inc.	PubMed Central	2024-06-05T03:55:55.722531	2005-3-18	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1083414/", "journal": "BMC Genomics. 2005 Mar 18; 6:40", "authors": [ { "first": "Samuel Shao-Min", "last": "Zhang" }, { "first": "Xuming", "last": "Xu" }, { "first": "Jinming", "last": "Li" }, { "first": "Mu-Gen", "last": "Liu" }, { "first": "Hongyu", "last": "Zhao" }, { "first": "M Bento", "last": "Soares" }, { "first": "Colin J", "last": "Barnstable" }, { "first": "Xin-Yuan", "last": "Fu" } ] }
PMC1083415	Background ========== Bacterial vaginosis (BV) is a poly-microbial syndrome characterised by a shift in vaginal flora from a predominant population of lactobacilli to their gradual or total replacement with anaerobes such as Gardnerella vaginalis, Prevotella, Bacteroidesand Mobiluncusspecies (spp), and with other bacteria including Mycoplasmaand Ureaplasmaspecies \[[@B1]\]. BV is one of the most frequent conditions encountered in sexually transmitted diseases (STD), genitourinary medicine (GUM) or other reproductive health clinics throughout the world. BV has been strongly associated with poor pregnancy outcomes such as preterm delivery of low-birth-weight infants \[[@B2]\] and several studies have now established associations between BV and HIV \[[@B3]-[@B5]\]. BV appears to be particularly common in sub-Saharan Africa where several studies have reported high prevalence rates, ranging from 20--49% among women presenting to STD clinics with vaginal discharge \[[@B6]-[@B8]\], from 21--52% among pregnant women attending antenatal clinics \[[@B9]-[@B11]\], and from 37--51% in community-based studies \[[@B3],[@B12]\]. These are very much higher than the rates reported from industrialised countries, 13% in GUM clinic attenders in the UK \[[@B13]\], 11% in gynaecology clinic attendees in London \[[@B14]\], and 15% to 30% in studies of non-pregnant women in USA \[[@B15]\]. The reasons for these disparities are not entirely clear, but may arise in part through the use of different case definitions for BV, and because the pattern of vaginal micro-flora associated with this condition may differ in different populations. Earlier African studies have relied on the Amsel\'s clinical definition of BV \[[@B16]\], whilst more recent studies have adopted the microbiological Nugent\'s scoring technique \[[@B17]\]. The latter method relies on the identification of categories of vaginal micro-flora based on quantitative assessment of a vaginal Gram-stained smear. The Nugent\'s method has been extensively validated in industrialised countries where numerous vaginal flora studies have been conducted \[[@B1]\], but little is known of the pattern of vaginal micro-flora associated with BV in Africa. The characterisation of vaginal micro-flora is an important step in understanding the pattern of flora associated with BV. This information may help investigate the significance of this condition in clinical pathology and for targeting treatment. In particular, it is important to know whether vaginal flora changes may enhance HIV acquisition as suggested \[[@B5]\], and to unravel some of the factors that influence such changes, as these could be perhaps modified. Behavioural factors such as vaginal douching or menstrual hygiene practices have been suggested as important factors that might influence vaginal flora composition \[[@B18]\], but little data is available from African populations\[[@B19],[@B20]\]. We have conducted a study aiming at determining the prevalence of BV among women self-presenting with vaginal discharge at a GUM clinic in Fajara, The Gambia. We report here on the vaginal micro-flora patterns and vaginal hygiene practices found in these patients and associations with their HIV serostatus. Methods ======= Study population and sample collection -------------------------------------- This study was part of a WHO-sponsored evaluation of STI syndromic management and of a novel rapid diagnostic test for BV conducted at the GUM clinic of the Medical Research Council (MRC) in Fajara, The Gambia. Study details have been reported elsewhere \[[@B21]\]. Briefly, consecutive and consenting women aged 18 and over, attending the MRC clinic with self-reported symptoms of vaginal discharge and/or vaginal itching were included in the study. Pregnant women were excluded because it was anticipated that their vaginal flora might differ substantially from that of other women attending the clinic. A standardised questionnaire elicited socio-demographic characteristics, reproductive and sexual health history including vaginal douching and menstrual hygiene practices, and current STD symptoms. Women underwent genital examination during which vaginal and cervical swabs were collected. The first vaginal swab was used for aerobic and anaerobic cultures; the second swab was used for direct wet mount microscopy, detection of fishy amine odour (\"whiff\" test) when mixed with 10% potassium hydroxide (KOH) preparation, and vaginal pH determination (range 4.0--7.0); the third swab was rolled onto a slide for Gram staining; two cervical swabs were collected for Neisseria gonorrhoeaeculture on modified Thayer-Martin media, and for Chlamydia trachomatispolymerase chain reaction (PCR) testing using an in-house method \[[@B22]\]. A blood sample was collected for HIV testing as routinely offered at the GUM clinic. Serological diagnosis of HIV infection was done according to a strategy described elsewhere \[[@B23]\]. In brief, sera were screened by the ICEHIV-1.O.2 (Murex Diagnostics Ltd, Dartford, UK) and reactive samples were retested by type-specific ELISAs: Wellcozyme HIV recombinant -1 (Murex) for HIV-1, and ICEHIV-2 test (Murex) for HIV-2. Samples clearly positive in one type were assigned the corresponding serological status; samples positive in both ELISAs were further tested by a synthetic peptide-based strip method, Pepti-Lav 1--2 (Sanofi Diagnostics Pasteur, Marne-La-Coquette, France). Treatment was given to all women according to the Gambian government syndromic management protocols covering all likely vaginal and cervical infections. This included a single dose of 2.0 g of metronidazole to cover Trichomonas vaginalis(TV) and BV. HIV-infected patients were referred internally to our specialist clinic and managed according to local standard guidelines. Microbiological methods for vaginal flora assessment ---------------------------------------------------- The wet preparations were examined microscopically for the presence of motile TV, yeast cells, and \'clue cells\'. Vaginal smear slides were heat fixed, Gram-stained and examined for vaginal flora categories using the Nugent\'s method \[[@B17]\]. The method involves assigning a score between 0 and 10 based on quantitative assessment of the Gram-stain for three different bacterial morphotypes: (i) large Gram-positive rods (indicative of Lactobacillusspp), (ii) small Gram-negative or variable rods (indicative of Gardnerella, Bacteroidesand other anaerobic bacteria), and (iii) curved, Gram-variable rods (indicative of Mobiluncusspp). Scores between 0 and 3 represent \'normal vaginal flora\', between 4 and 6 \'intermediate vaginal flora\', and scores between 7 and 10 are considered diagnostic for \'BV\'. Vaginal swabs were directly inoculated at the clinic onto: (i) Columbia blood agar plates, which were incubated aerobically at 37°C for 24 to 48 hours to isolate aerobic bacteria, including lactobacilli; (ii) Columbia human blood bi-layer agar plates, which were incubated micro-aerophilically at 36°C and read after 48 to 72 hours for Gardnerella vaginalisisolation; (iii) Columbia-base lake horse blood kanamycin agar plates, which were incubated anaerobically at 36°C for 48 to 72 hours to isolate anaerobic bacteria; (iv) Mycoplasmabroths incubated for 48 hrs then sub-cultured onto Mycoplasmaagar, and incubated micro-aerophilically for 48 to 72 hours; (v) Mann Rogosa Sharpe (MRS) medium, which was used for the isolation of Lactobacillusspp after incubation in CO~2~at 37°C for 48 hours; and (vi) Sabouraud\'s agar plates, which were incubated micro-aerophilically at 36°C 24 to 48 hour to isolate Candidaspp. Growth of bacterial isolates was graded as confluent (heavy growth), semi-confluent (moderate pure to mixed growth of bacteria with visible separate single colonies); and scanty (occasional single isolated colonies). Presumptive identification procedures ------------------------------------- ### Lactobacilli and other aerobic flora Lactobacilli were presumptively identified by their ability to grow well on MRS, Gram stain microscopy and catalase reaction. Isolates were further tested for their ability to produce hydrogen peroxide (H~2~O~2~) using a 2,3 tetramethyl benzedine method \[[@B24]\], and classified as positive when they produced blue coloration. The level of H~2~O~2~production was determined by visually grading the intensity of the blue colour produced into low, moderate and high categories. Coliformspp were identified as Gram-negative lactose fermenting rods; Staphylococcusspp were identified by their characteristic colony and Gram stain morphology, then tested for coagulase production (slide test); Gram-positive beta-haemolytic Streptococcusspp isolates were further typed using a rapid latex test according to the manufacturer\'s instructions (Streptex, Murex Biotech Ltd, Darford Kent, UK). Candidaspp were identified as colonies with typical yeast-like morphology and by characteristic morphology on a wet preparation examination (presence of budding cells and/or pseudo-hyphae). ### BV-associated bacteria Gardnerella vaginaliswas identified by beta-haemolytic appearance of the colonies on human blood bilayer agar plate but not on sheep blood agars, Gram stain morphology (Gram-variable pleiomorphic coccobacilli mostly forming clumps) and negative catalase and oxidase reactions. Suspected anaerobic isolates were sub-cultured onto Columbia blood agar without antibiotics and incubated in aerobic and anaerobic conditions. Strict anaerobes were further identified by Gram staining and antibiotic susceptibility to erythromycin, rifampicin, colistin, penicillin, kanamycin and vancomycin (Oxoid Discs, Unipath, Basinstoke, Hampshire UK). Gram-negative anaerobic bacilli were tested and the isolates which grew on bile medium and hydrolysed aesculin were identified as Bacteroidesspp. Isolates which failed to grow on bile or hydrolyse aesculin were identified as Prevotellaspp \[[@B25]\]. Mycoplasmaspp were identified as typical \"fried egg\" colonies and stained with Diene\'s stain. They were further presumptively identified as Mycoplasma hominisby colonial appearance and staining characteristics with a permanent diffused light blue periphery and a darkly blue centre \[[@B25]\]. Diagnosis of bacterial vaginosis -------------------------------- The gold standard microbiological definition of BV was a score of 7--10 by the Nugent\'s method described above. Amsel\'s clinical criteria were also used to make a clinical diagnosis of BV, which included the presence of any three of the following; (i) homogeneous grey adherent vaginal discharge; (ii) vaginal fluid pH ≥ 4.6; (iii) release of fishy amine odour when 10% potassium hydroxide (KOH) solution was added to a sample of vaginal fluid (the \"whiff\" test); and (iv) the presence of \'clue cells\' representing over 20% of vaginal epithelial cells on wet-prep microscopic examination \[[@B16]\]. Ten percent random quality control checks were carried out on the microscopist by an experienced BV microscopist, for Nugent\'s score and wet preparation readings for \'clue cells\'. Statistical methods ------------------- Frequency tables were produced to describe the prevalence of genital infections and of micro-flora isolates. Bacteroides, Prevotella, Peptostreptococcusand other anaerobic isolates were grouped as \'anaerobes\'. Chi-square tests were used to examine the association between isolates and BV diagnosis by Nugent\'s score and Amsel\'s clinical criteria in univariate analyses. Logistic regression was used to estimate the association between each isolate and diagnosis of BV by one method adjusted for (i.e. within categories of) BV diagnosis by the other method; and to test whether the observed associations were different between categories of the other method (i.e. testing for interaction). Chi-square and Fisher\'s Exact (for small numbers) tests were used to examine associations between categorical variables such as: isolation of lactobacilli, particularly H~2~O~2~-producing strains and G vaginalis, anaerobic isolates, Mycoplasma hominis, as well as N gonorrhoeae, C trachomatis, Candidaspp or T vaginalis;HIV and each of the vaginal flora micro-organisms mentioned above, and with BV (Nugent\'s score 7--10); HIV and vaginal hygiene variables; and between BV and vaginal hygiene variables. Ethical issues -------------- The study was approved by the Ethical Committees of the MRC Gambia and the London School of Hygiene & Tropical Medicine, and by the Ethics Review Board of the World Health Organisation. Results ======= Two hundred and thirty women were enrolled in the study over a four-month period in 2000, but data of three women were subsequently excluded because the women had not reported any symptoms consistent with our case definition of vaginal discharge syndrome. The denominators used in subsequent analyses vary slightly depending on completeness of microbiological investigations. The median age of the patients was 26 years (range 18--50), 71% of women were married, and a majority of women (84%) reported only one sexual partner in the last three months. Antibiotic use prior to attending the clinic was reported by 18% of patients. Clinical and aetiological findings ---------------------------------- The prevalence of cervical infections was 6.3% (14/222) for N gonorrhoeaeand 15.0% (34/227) for C trachomatis. A serum sample was obtained from 210 (93%) women and 27/210 (12.8%) of them were HIV infected (19 with HIV-1, 7 with HIV-2, and 1 dually with HIV1 and HIV-2). Using wet preparation microscopy, 10.1% (23/227) of the women were positive for T vaginalis and 22.9% (52/227) for yeast cells, with only two patients having a dual infection. \'Clue cells\' were seen in 56.4% of the wet preparations but in only 23.1% of the Gram-stained smears. A large majority of women (91.2%) had an elevated pH (≥ 4.6). Overall, 30.8% (68/221) of women had a diagnosis for BV according to the Amsel\'s criteria. BV prevalence as determined by Nugent\'s score of 7--10 was 47.6% (108/227), 24.7% (56/227) women had \'intermediate flora\' (score 4--6) and 27.7% (63/227) had \'normal flora\' (score 0--3). Vaginal flora cultures ---------------------- Results of vaginal flora cultures for aerobic, micro-aerophilic and anaerobic bacteria as well as cultures for Mycoplasmaare shown in Table [1](#T1){ref-type="table"}. Variations in the denominators for the types of bacteria isolated were due to cases of either contamination or growth failure. The lower number of anaerobic cultures performed (n = 66) was due to a breakdown in the supply of anaerobic Gas-Packs. We defined two unidentified Gram-positive anaerobic isolates as \'Other\' anaerobic isolates. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Vaginal flora isolates by Nugent\'s score among women presenting with vaginal discharge syndrome at the GUM clinic in Fajara, The Gambia ::: Total Samples Scanty Growth Semi-confluent growth Confluent growth *P-value^2^* --------------------------- ------------------- ------------------- --------------------------- ---------------------- ------------------ ----------- ----------- ----------- ------------ Nugent\'s score^1^ 0--6 7--10 0--6 7--10 0--6 7--10 0--6 7--10 Isolate N N % (n) % (n) % (n) % (n) % (n) % (n) Aerobic flora Lactobacillusspp. 119 106 5.0 (6) 8.5 (9) 16.8 (20) 4.7 (5) 33.6 (40) 4.7 (5) \<0.001 Coliformspp. 119 106 12.6 (15) 9.4 (10) 1.7 (2) 6.6 (7) 5.0 (6) 4.7 (5) 0.868 Staphylococcusspp. 119 105 24.4 (29) 37.1 (39) 9.24 (11) 4.8 (5) 0.8 (1) 0.9 (1) 0.217 Streptococcusspp. 119 105 13.4 (16) 15.2 (16) 11.8 (14) 12.4 (13) 13.4 (16) 10.9 (11) 1.000 Candidaspp. 93 81 11.8 (11) 14.8 (12) 16.1 (15) 7.4 (6) 15.1 (14) 8.6 (7) 0.117 Microaerophilic flora Gardnerella vaginalis 117 106 1.7 (2) 8.5 (9) 6.8 (8) 21.7 (23) 13.7 (16) 38.7 (41) \<0.001 Anaerobic flora Any anaerobic isolate^3^ 34 32 1 (3) 2 (6) 2 (6) 9 (28) 3 (9) 6 (19) 0.005^4^ Bacteroidesspp. 34 32 0 6.2 (2) 0 12.5 (4) 2.9 (1) 12.5 (4) 0.002 Prevotellaspp. 34 32 2.9 (1) 0 2.9 (1) 9.4 (3) 2.9 (1) 12.5 (4) 0.180 Peptostreptococcusspp. 34 32 0 0 0 3.1 (1) 0 0 0.485 Other anaerobes^5^ 34 32 0 0 1 (3) 2 (6) 1 (3) 0 0.801 Mobiluncusspp. 35 33 0 0 0 0 0 0 \- Mycoplasma hominis 115 105 0 0 2.6 (3) 0.9 (1) 14.8 (17) 24.8 (26) 0.142 Note: denominators vary according to the number of samples for which each test was conducted (see Results section). ^1^Nugent\'s categories: score 0--6 = normal and intermediate flora; score 7--10 = BV. ^2^From Fisher\'s Exact test. ^3^Includes Bacteroidesspp., Prevotellaspp. and Peptostreptococcusspp. isolates and other anaerobes. ^4^Anaerobic isolates were found in 21 samples out of 66; 5 women had more than 1 anaerobic isolate in which case the highest level of growth was used for the significance test. ^5^Anaerobic Gram-negative rods, anaerobic Gram-positive or pigmented anaerobes. ::: Gardnerella vaginalis, lactobacilli, streptococci and Mycoplasmaspp were present as confluent or semi-confluent growth. Mobiluncusspp were not isolated. Staphylococcusspp, and Coliformsspp were mostly present in scanty numbers. Bacteroidesand Prevotellaspp were present in semi-confluent to confluent numbers. Seventy percent (42/60) of lactobacilli tested were found to be H~2~O~2~-producers, 12 at high level, 12 at intermediate level, and 18 at low levels of H~2~O~2~production. We could not establish the ability of the remaining 25 lactobacilli isolates to produce H~2~O~2~because these isolates could not be recovered from storage. Staphylococcusspp were tested for coagulase production and 20.5% (17/83) were identified as Staphylococcus aureus. Serotype grouping was done for 61/86 Streptococcusspp isolates which were successfully recovered and the majority of these, 68.9% or 42 strains, were classified as Group B, 2 as group C, 6 as group D, and 8 as group F. Associations between bacterial vaginosis and vaginal flora ---------------------------------------------------------- Patterns of micro-flora isolates according to BV diagnosis by Nugent\'s score and Amsel\'s clinical criteria are summarised in Table [2](#T2){ref-type="table"} and Figure [1](#F1){ref-type="fig"}. Isolation of lactobacilli and H~2~O~2~-producing lactobacilli was negatively associated with BV diagnosis by both Nugent\'s score (OR 0.07, 95%CI 0.03 -- 0.15) and Amsel\'s criteria (OR 0.49, 95%CI 0.26 -- 0.92). Within Amsel\'s diagnostic categories, the Nugent\'s diagnosis of BV was still negatively associated with lactobacilli (adjusted OR \[AOR\] 0.06, 95%CI 0.03 -- 0.15). However, within Nugent\'s diagnostic categories, Amsel\'s diagnosis of BV was not associated with lactobacilli (AOR 1.11, 95%CI 0.52 -- 2.38). G vaginalisand the anaerobes were strongly associated with BV diagnosis by both Nugent\'s score and Amsel\'s criteria. G vaginalisremained significantly associated with BV diagnosis by either method, within the diagnosis categories of the other method. It was not possible to compare the prevalence of anaerobes across the three Nugent categories within Amsel\'s diagnosis using a likelihood ratio test due to the lack of anaerobic isolation in normal subjects. M hominiswas more common in women with BV by both Nugent\'s and Amsel\'s criteria but the association was not statistically significant. Coliformspp, staphylococci and streptococci were not significantly associated with BV by either Nugent\'s score or Amsel\'s criteria (data not shown). ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Association between vaginal micro-flora isolates and Nugent\'s vaginal flora categories and Amsel\'s diagnosis of bacterial vaginosis ::: Association with Nugent\'s vaginal flora categories Association with Amsel\'s diagnosis of BV ---------------------------- -------------- --------------------------------------------------------- ----------------------------------------------- ------------------ ------------------ -------- ----- ------ ------------------ ----------------- Lactobacilli Normal flora 63 74.6 1 1 Interm. 54 35.2 0.18 0.08--0.41 0.18 0.08--0.41 Neg. 152 42.8 1 1 BV 102 16.7 0.07 0.03--0.15 0.06 0.03--0.15 Pos. 67 26.9 0.49 0.26--0.92 1.11 0.52--2.38 P^7^ \<0.001 \<0.001 P^7^ 0.025 0.775 H~2~O~2~- Lactobacilli Normal flora 49 49.0 1 1 Interm. 49 14.3 0.17 0.07--0.46 0.17 0.07--0.47 Neg. 134 24.6 1 1 BV 97 10.3 0.12 0.05--0.28 0.12 0.05--0.31 Pos. 61 13.1 0.46 0.20--1.07 0.94 0.36--2.49 P^7^ \<0.001 \<0.001 P^7^ 0.067 0.909 *G vaginalis* Normal flora 61 11.5 1 1 Interm. 54 35.2 4.19 1.59--11.0 3.96 1.50--10.5 Neg. 150 35.3 1 1 BV 102 71.6 19.42 7.91--47.6 15.45 6.18--38.6 Pos. 67 68.7 4.01 2.17--7.42 2.21 1.10--4.44 P^7^ \<0.001 \<0.001 P^7^ \<0.001 0.026 Anaerobic isolates^8^ Normal flora 21 0 1 1 Interm. 13 38.5 -^9^ -^9^ Neg. 43 18.6 1 1 BV 31 48.4 -^9^ -^9^ Pos. 22 54.6 5.25 1.68--16.38 4.5 1.24--16.37 P^7^ 0.001 -^10^ P^7^ 0.003 0.018 *Mycoplasma hominis* Normal flora 60 13.3 1 1 Interm. 53 22.6 1.90 0.71--5.09 1.85 0.69--4.97 Neg. 148 18.9 1 1 BV 102 25.5 2.22 0.93--5.29 2.00 0.80--4.96 Pos. 67 26.9 1.57 0.80--3.10 1.34 0.65--2.76 P^7^ 0.184 0.282 P^7^ 0.188 0.436 ^1^Nugent\'s categories: normal flora (score 0--3), intermediate flora (score 4--6), BV (score 7--10). ^2^Number of specimens. ^3^Percentage positive for the isolate. ^4^Crude Odds Ratio. ^5^Odds Ratio adjusting for the other method of diagnosing BV. ^6^Amsel\'s diagnosis \'positive\' \[pos.\] if at least 3 out of 4 criteria present (homogenous adherent discharge; pH ≥ 4.6; whiff test positive; clue cells on 20% of epithelial cells); diagnosis is \'negative\' \[neg.\] if \<3 criteria present. ^7^From X^2^test for unadjusted analysis. From likelihood ratio test when adjusting for the other diagnostic method. ^8^Anaerobic isolates comprise Prevotellaspp, Bacteroidesspp, Peptostreptococcusspp and other anaerobes. ^9^Not possible to calculate OR relative to normal flora due to lack of anaerobic isolation in normal subjects. ^10^Not possible to fit model due to lack of anaerobic isolation in normal subjects. ::: ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Prevalence of microflora isolates by Amsel\'s clinical criteria diagnosis and Nugent\'s vaginal flora categories ::: ![](1471-2334-5-12-1) ::: The associations between the presence of lactobacilli, including the presence of H~2~O~2~producing lactobacilli strains, and the isolation of other vaginal or cervical organisms are shown in Table [3](#T3){ref-type="table"}. There were significant 2 to 4 fold decreases in colonisation by G vaginalisor anaerobes in the presence of H~2~O~2~-producing lactobacilli (P \< 0.001 and P = 0.016 respectively). M hominisisolation was less prevalent in the presence of H~2~O~2~-producing lactobacilli, but this was not statistically significant. Overall there was no significant difference for Candidaisolates (P = 0.108) but the lowest prevalence was observed when no lactobacilli were isolated. Detection of C trachomatiswas significantly reduced (p = 0.024) in the presence of H~2~O~2~-producing lactobacilli strains, whilst N gonorrhoeaewas isolated more frequently in the absence of lactobacilli, although this was not statistically significant. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Association between lactobacilli and other genital micro-organisms ::: Genital micro-organisms Prevalence of genital micro-organism in those with: ----------------------------- --------------------------------------------------------- --------- --------- --------- --------- --------- ----------- n/N (%) n/N (%) n/N (%) C trachomatis 21/140 (15.0) 4/18 (22.2) 1/42 (2.4) 0.024 N gonorrhoeae 12/138 (8.7) 0/17 (0) 1/42 (2.4) 0.303 Candidaspp 34/110 (30.9) 6/14 (42.9) 16/32 (50.0) 0.108 T vaginalis 19/140 (13.6) 1/18 (5.6) 3/39 (7.1) 0.509 G vaginalis 78/140 (55.7) 5/18 (27.8) 9/40 (22.5) \<0.001 Anaerobes 16/34 (47.1) 0/3 (0) 3/21 (14.3) 0.016 M hominis 33/137 (24.1) 5/17 (29.4) 5/40 (12.5) 0.214 ^1^From Fisher\'s Exact test because of small numbers in some categories. ::: Associations between vaginal flora, vaginal hygiene practices and HIV --------------------------------------------------------------------- We found that only two anaerobic isolates, Bacteroidesspp and Prevotellaspp, were significantly more common among HIV positive women: 7/57 samples (12.3%) for HIV negative women contained Bacteroidesspp, compared with 3/6 (50%) for HIV-positive women (P = 0.0046); and 6/57 samples (10.5%) for HIV negative women contained Prevotellaspp, compared with 3/6 (50%) for HIV positive women (P = 0.033). In overall crude analysis, there was no association between BV (Nugent\'s score 7--10) and HIV (all types): 12/110 (10.9%) women without BV were HIV-infected vs. 15/100 (15.0%) women with BV (table [4](#T4){ref-type="table"}). ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Prevalence of HIV and BV in symptomatic women, by douching, menstrual hygiene protection method and female genital cutting ::: Prevalence of BV (Nugent\'s score 7--10) Prevalence of HIV ----------------------------------------- ---------------------------------------------- ------------------------------------------- --------- -------- ------ --------- Douching^1^ No 12/22 54.5 0.507 3/22 13.6 1.000 Yes 94/202 46.5 24/185 13.0 Menstrual hygiene protection method Traditional^2^ 70/138 50.7 0.323 17/129 13.2 1.000 Sanitary pads 34/79 43.0 9/71 12.7 Female genital cutting^3^ No 38/68 55.9 0.146 12/63 19.1 0.115 Yes 70/157 44.6 15/145 10.3 BV No 12/110 10.9 0.414 Yes 15/100 15.0 ^1^Douching before and/or after sexual intercourse. ^2^Old cloths washed and re-used as necessary. ^3^Also known as \"female circumcision\"; in the Gambia, the type of female cutting belongs to WHO classification Type II, i.e. removal of all/part of the clitoris and labia minora. ^4^P-values from Fisher\'s Exact Test. ::: Details about vaginal hygiene practices were collected by questionnaire: 37.8% (85/225) of women reported practicing vaginal washing before sex and 89.2% (199/223) after sex, for a combined 90.2% of women (202/224) practising some form of \'douching\' before or after sex. Of the women who douched, 57.6% (114/198) used water, 22.2% (44/198) used soap and water, 18.2% (36/198) used a dry towel, and 2% (4/198) used commercial or other cosmetic products. The source of water was 83% tap water, and 17% water from a protected well. In crude univariate analysis, there was no association between \'douching\' and BV (Nugent\'s score 7--10) (Table [4](#T4){ref-type="table"}), nor between any of the individual vaginal hygiene variables and BV (data not shown). The questionnaire also explored forms of menstrual hygiene: 60.8% (118) of women reported using traditional methods for sanitary protection (i.e. reusable cloths), 34.8% (79) using sanitary pads, 4.0% (9) tampons and 0.4% (1) other methods. Again, there were no associations between these methods and presence of BV (table [4](#T4){ref-type="table"}). The majority of women (157/227, 69.2%) were found circumcised on examination. The dominant form of genital cutting in the Gambia belonged to WHO classification Type II (i.e. removal of all/part of the clitoris and labia minora) as reported previously \[[@B26]\]. Again no association with BV was noted (table [4](#T4){ref-type="table"}). In overall crude univariate analysis, we did not find any association between douching, menstrual hygiene, genital cutting and HIV (table [4](#T4){ref-type="table"}). Discussion ========== The main objective of this study was to determine the prevalence of BV and the pattern of vaginal micro-flora among women with vaginal discharge syndrome in an African setting and to relate this to vaginal hygiene practices and HIV serostatus. Using Nugent\'s score as the gold standard, a BV prevalence of 47.6% was found in this population. This compares to the range (20--49%) reported from other African populations attending STI clinics: 20--23% in Burkina Faso \[[@B7]\] and Malawi \[[@B27]\], 37% in Tanzania \[[@B8]\] and 49% in Kenya \[[@B6]\]. High prevalence of BV (21--29%) has also been observed among pregnant women in Kenya and South Africa \[[@B9],[@B11],[@B28]\], The reasons for higher BV rates in African populations are not known. High BV rates have been reported among African-American women \[[@B15]\], although it has been argued that levels of education and other socio-economic factors were confounding these associations \[[@B15],[@B29]\]. Lifestyle practices such as vaginal douching have also been associated with an increased prevalence of BV \[[@B15],[@B18],[@B19]\], although, the direction of causality is again uncertain, since most studies have been of cross-sectional nature and many potentially confounding factors such as educational, socio-economical and behavioural factors have not always been entirely controlled for. We did not find any association between BV or vaginal micro-organisms and vaginal hygiene practices such as douching before or after sex, the nature of douching compounds used, the source of the water, or with menstrual sanitary protection. This finding perhaps owes to the fact that a very large proportion of women reported these practices, thus any relatively small association with BV would be hard to find with our sample size. On the other hand, additional possible explanations for the high prevalence or incidence of BV in African populations have to be sought -- the role of hormonal factors should be explored. An association between BV and HIV has been reported in several studies \[[@B3],[@B4],[@B30]\], possibly influenced by vaginal hygiene practices \[[@B18],[@B20],[@B31]\]. However, as in our study, not all studies reporting on douching, BV and HIV have found associations between these factors \[[@B19]\]. The relationship between HIV, risk for BV or other STIs is complex, and could be contributed to by high risk sexual behaviour. Our study population consisted only of symptomatic women attending a GUM clinic, thus high-risk behaviours may have blurred any possible association. To our knowledge, this study is one of the first to report on female genital cutting in relation to HIV and vaginal flora in Africa. We did not find any significant impact of circumcision on vaginal flora or HIV serostatus. A comparison of vaginal micro-flora isolates with Nugent\'s score showed significant positive associations between a diagnosis of BV and the isolation of G vaginalis, a significant positive association with anaerobes, and a significant negative association with the presence of lactobacilli. These findings are not surprising, since the method employed for the Nugent\'s score is based on the observation of BV-associated bacterial morphotypes. Nonetheless there have been no reports of vaginal micro flora culture studies in African population aimed at exploring the patterns of BV-associated flora. This study has demonstrated similar vaginal bacterial isolates to those found in the United States \[[@B32],[@B33]\]. These studies found a strong association between BV and the isolation of G vaginalis, anaerobic gram-negative rods belonging to the genera Prevotella, Porphyromonasand Bacteroides, Peptostreptococcusspp, M hominis, Ureaplasma urealyticum, and often Mobiluncusspp. A lower concentration of facultative species of Lactobacillusamong women with BV in comparison to women with a normal flora was noted in this study. Lactobacilli are reported to play an important role in the maintenance of normal vaginal flora \[[@B34],[@B35]\] through the provision of defence mechanisms against pathogenic organisms via hydrogen peroxide (H~2~0~2~) production and the maintenance of an acidic microenvironment generated by lactic acid production. A low pH has been shown to have a direct microbicidal and virucidal effect \[[@B36]\]. Hydrogen peroxide, which is produced by some lactobacilli strains, also has a direct antimicrobial effect. H~2~O~2~can inhibit the growth of Bacteroides, Gardnerella, Mobiluncus, Mycoplasmaand other vaginal organisms by acting directly on these organisms using its toxic effect, or by reacting with halide ions in the presence of vaginal peroxidase as part of the H~2~O~2~-halide peroxidase antibacterial system. Lactobacilli can also adhere onto vaginal epithelial cells \[[@B37]\] thus blocking the attachment of any pathogenic BV-associated bacteria onto these cells, Lactobacilli are known to produce biosurfactant, bacteriocins and coaggregation molecules \[[@B38]\], all of which contribute to the maintenance of a healthy vaginal micro-environment. In our study, a large proportion of lactobacilli isolates (70%) were H~2~O~2~producers. These isolates were associated with a significantly lower prevalence of G vaginalis, anaerobes and C trachomatis, with trends of lower prevalence of N gonorrhoeaeand M hominis, suggesting a protective effect of vaginal/cervical colonisation conferred by these lactobacilli strains. The absence of H~2~0~2~-producing lactobacilli was not associated with growth of Candidaspp. This is different from the findings observed in a study conducted by Hillier et al\[[@B35]\], in which significant associations between H~2~O~2~production and protection against BV and other STI as well as protection against symptomatic candidiasis were reported. In previous studies \[[@B39],[@B40]\], M hominisand Ureaplasma urealyticumhave been associated with BV, although not in all cases \[[@B41]\]. In our study, there appears to be some association between M hominisand BV, defined either by Nugent\'s score or Amsel\'s criteria, but these were not statistically significant. This observed difference may be due to the different detection protocols employed by different researchers. Other bacteria that have been isolated in this study included Staphylococcusspp (38.4%), Coliformspp (20.0%)and Streptococcusspp (38.4%) but they were not associated with any particular vaginal microflora. Most of the staphylococci isolates were coagulase-negative, which are perceived to be normal commensal organisms. Most streptococci were found to belong to group B, an organism which can be highly pathogenic, particularly at time of delivery when it has been associated with neonatal sepsis and premature delivery \[[@B42]\]. The high prevalence of Group B streptococci seen in this study was similar to that reported in previous work done in The Gambia \[[@B43]\]. The latter study, however, failed to demonstrate any association between isolation of Group B streptococci and disease manifestation in neonates. Thus, the high prevalence of group B streptococci isolates reported in this study may not pose any significant clinical problem in this setting. Several logistical problems have arisen in the course of this study, which may limit interpretation of the data on vaginal flora patterns. First, Mobiluncusspp, were not isolated despite observing Mobiluncus-like organisms in some of the vaginal Gram-stained smears. This finding could be attributed to a low prevalence of Mobiluncusin our population, or to an inadequate isolation procedure, which will warrant further investigation. Second, our adoption of a 48-hours incubation period for broth- and agar- Mycoplasmaculture procedures did not favour the isolation of Ureaplasma urealyticum, as the latter is sensitive to the metabolic by-products generated during extended incubation periods. This may have been partly responsible for the lack of Ureaplasmaisolation in this study. Finally, there was a low-recovery rate for a significant number of stored isolates. Optimum storage procedures are not easy to maintain in tropical climates. These logistical problems with fastidious organisms may explain why it is difficult to attempt studies of vaginal flora in developing countries. Another possible limitation of our study was the absence of inclusion of women who did not complain of vaginal discharge syndrome, which would have allowed a more comprehensive description of patterns of vaginal flora in symptomatic and symptomless women. Our study shows, however, that the pattern of organisms cultured in BV in this environment is similar to what has been reported in corresponding populations in industrialised countries, and therefore suggest that different vaginal flora patterns are not the major explanation behind the higher prevalence of BV in Africa. In most STI or GUM clinics in industrialised countries, routine BV diagnosis is made using Nugent\'s score or Amsel\'s clinical criteria. In resource-poor settings good microscopy is not often available, and clinicians rely on clinical judgement or may apply all or part of the Amsel\'s clinical criteria. This study found there was an association between Amsel\'s criteria and Nugent\'s score, a strong correlation between Amsel\'s diagnosis and the presence of G vaginalisand anaerobes, and a negative correlation with the presence of lactobacilli. However, as shown in Figure [1](#F1){ref-type="fig"}, a large number of true BV cases (by Nugent\'s score) were missed by the Amsel\'s method, limiting its utility as a BV diagnostic method. The Amsel\'s method can be highly subjective with regards the description of the discharge and the olfactive component (\'whiff\' test); the wet mount microscopy depends on the experience of the microscopist and can represent a further subjective element, particularly when performed under pressing clinical conditions \[[@B1]\]. Despite regular supervision and the use of the same clinician and microscopist throughout the study (which insured internal consistency of our results), the discrepancies we found between the Amsel\'s criteria and the accepted gold standard diagnosis (Nugent\'s method) in this study appear to preclude the use of this method under routine clinical conditions in our setting. Our study found that where Gram-staining led to a classification of \'intermediate flora\' by Nugent\'s score, this was reflected in the microbiological findings, which were \'intermediate\' quantitatively and qualitatively between \'normal\' and \'BV\' categories (table [2](#T2){ref-type="table"} and figure [1](#F1){ref-type="fig"}) and distinct from them. This supports the validity of the classification and could indicate that the \'intermediate\' flora precedes the development or follow the resolution of frank BV. Longitudinal studies are required to elucidate this phenomenon. Conclusion ========== In this population, BV prevalence was higher than in corresponding populations in industrialised countries, but the pattern of vaginal micro-flora associated with BV was similar. BV or vaginal flora patterns were not associated with HIV nor with any of the vaginal hygiene characteristics. Further studies on the public health significance of BV in this kind of setting are needed to determine future strategies for intervention Competing interests =================== We received donation of FemExam^®^tests from Litmus Ltd (CA, USA) for this study. Authors\' contributions ======================= ED contributed to study design, conducted all laboratory investigations and produced the first draft of the manuscript; LM oversaw data management and conducted statistical analyses; MSL oversaw clinical data collection and provided epidemiological support; AAA contributed to clinical data management; EG conducted the clinical part of the study; PM, RB and BW developed the study protocol and supervised the study with BW overseeing the laboratory aspects of the study; all authors contributed revisions to the manuscripts. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2334/5/12/prepub> Acknowledgements ================ The authors wish to thank the MRC GUM clinic staff (in particular Miss Marie Claire Mendy) and staff at the MRC laboratory (Miss Nkaye Kanyi and Mr Keita Badgie). The study was supported by a grant from the World Health Organisation/Sexually Transmitted Infections Division (WHO/STI). We thank Dr Francis Ndowa (WHO/STI) for his assistance.	PubMed Central	2024-06-05T03:55:55.729009	2005-3-9	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1083415/", "journal": "BMC Infect Dis. 2005 Mar 9; 5:12", "authors": [ { "first": "Edward", "last": "Demba" }, { "first": "Linda", "last": "Morison" }, { "first": "Maarten Schim", "last": "van der Loeff" }, { "first": "Akum A", "last": "Awasana" }, { "first": "Euphemia", "last": "Gooding" }, { "first": "Robin", "last": "Bailey" }, { "first": "Philippe", "last": "Mayaud" }, { "first": "Beryl", "last": "West" } ] }
PMC1083416	Background ========== The genome sequences of Arabidopsis\[[@B1]\] and rice \[[@B2],[@B3]\] have stimulated great advances throughout the plant sciences. Comparisons of these eudicot and monocot genomes have provided many insights into the genome characteristics and evolutionary histories of both lineages \[e.g. \[[@B4]-[@B6]\]\], and comparisons involving additional species are generating a more global picture of angiosperm genome evolution \[[@B7]-[@B9]\]. These multispecies comparisons, and comparative plant sciences more generally, have been aided by the well-supported understanding of evolutionary relationships among flowering plants that has emerged over the last decade \[e.g. \[[@B10]-[@B13]\]\]. Among the most noteworthy phylogenetic results is the well-supported inference that whereas monocots form a clade, the dicots as traditionally circumscribed do not. Rather, monocots are derived from within the \"primitive\" dicot grade, now collectively referred to as basal angiosperms (Fig. [1](#F1){ref-type="fig"}). The \"eudicots\" (or \"tricolpates\" \[[@B14]\]; Fig. [1](#F1){ref-type="fig"}) do form a clade that comprises ca. 75% of all angiosperm species \[[@B15]\], and most of this diversity is found among the \"core\" eudicots, which include the rosids, asterids, and Caryophyllales (Fig. [1](#F1){ref-type="fig"}). Model systems such as Arabidopsis thaliana, tomato (Lycopersicon esculentum), cotton (Gossypium), poplar (Populus), barrel medic (Medicago truncatula) and ice-plant (Mesembryanthemum crystallinum), are all representatives of the core eudicot clade (Fig. [1](#F1){ref-type="fig"}). Rice (Oryza savita), maize (Zea mays), wheat (Triticum aestivum), barley (Hordeum vulgare), sorghum (Sorghum bicolor), and sugarcane (Saccharum officinale) are all members of the grass family (Poaceae), a phylogenetically derived lineage within the monocots (Fig. [1](#F1){ref-type="fig"}). Although comparisons of the rice and Arabidopsisgenomes will undoubtedly identify many features of the ancestral angiosperm genome, this pair-wise comparison alone will not be able to distinguish Arabidopsis-specific attributes from those specifically absent in rice or visa versa. The recent posting of high coverage genome sequence for Populus trichocarpa\[[@B16]\] is a major advance for comparative plant genomics, but even Populus-Arabidopsis-rice comparisons cannot distinguish features common to all angiosperms from those that arose in the most recent common ancestor of eudicots and monocots, which existed at least 125 million years ago \[[@B17]\] and perhaps more than 140 million years ago \[e.g. \[[@B18]-[@B20]\]\]. In general, the resolving power of comparative plant genomics will increase with additional taxa representing key lineages in plant phylogeny (Fig. [1](#F1){ref-type="fig"}). Increased genomic resources for phylogenetically diverse plant species will lead to a better understanding of plant genome evolution, the diversification of gene families, and the origins of reproductive characteristics common to all flowering plants. Basal angiosperms and basal eudicots (e.g., Ranunculales), while comprising a small percentage of the total number of extant angiosperm species, nonetheless encompass an astonishing spectrum of developmental patterns and floral forms \[[@B21],[@B22]\]. In turn, this diversity provides a clear opportunity to reconstruct to the basal condition of angiosperms, and thereby bridge the evolutionary gap between model eudicot and monocot genomes. Understanding the evolution of angiosperm genes and genomes, including the floral transcriptome, requires three-way and higher-order comparisons that extend beyond Arabidopsisand rice. This point is widely appreciated, and comparative plant genomics is being fueled by the availability of genomic resources for a growing number of plant species. The addition of species representing basal angiosperms, basal eudicot, and non-grass monocot lineages will be especially valuable, not only for flowering research, but also for more general \"reconstructomic\" studies of housekeeping and transcription factor functions. A primary objective of the Floral Genome Project (FGP; \[[@B23]\]) is to uncover patterns of conservation and divergence of the floral transcriptome among angiosperms, particularly to elucidate the role of gene duplications and shifting expression patterns in the origin and diversification of angiosperms. The FGP has constructed a large collection of non-normalized nor tissue subtracted cDNA libraries and 5\' EST sets from developing reproductive tissues for selected species of basal eudicots, basal angiosperms and gymnosperms (Table [1](#T1){ref-type="table"}). These species represent not only key nodes in the angiosperm phylogenetic tree and its sister group (gymnosperms), but also a diversity of reproductive structures and developmental patterns. While multi-species comparisons of large sequence data sets are already possible for Poaceae \[[@B9],[@B24]\] Solanaceae \[[@B25]\], and Brassicaceae \[[@B7],[@B26],[@B27]\], the addition of large EST sets for basal angiosperms opens the door to fundamental comparative genomics investigations of the origin and diversification of flowering plants. Results ======= Random 5\' sequencing of cDNAs from basal flowering plants has so far (as of 12/01/04) generated 70,514 ESTs assembled into 48,170 unique gene sequences (Table [1](#T1){ref-type="table"}). These materials should provide essential resources for comparative genomic research because they represent previously poorly sampled genomes placed at crucial points in angiosperm phylogeny. Gene sequences from the gymnosperm Welwitschia, the basalmost angiosperms Amborellaand Nuphar, the basal monocot Acorus, the magnoliids Persea, Liriodendronand Saruma, and the basal eudicot Eschscholziawill aid in placing boundary dates on the origins of florally-expressed gene families, help resolve patterns of gene and genome evolution within the flowering plants, and bridge critical gaps in comparative analyses involving monocot and eudicot model systems. Identification of cDNA clones for finished sequencing has been aided by efficient sorting of EST sequences into putative gene families based on whole Arabidopsis/rice proteome comparison. Phylogenetic analyses of ESTs are providing new insights into the process of gene family evolution in relation to the origin and diversification of the angiosperms. Here we introduce our EST database and provide some examples of broad utility of these data in comparative analyses. PGN website ----------- All FGP EST data and unigene builds are available through the Plant Genome Network (PGN) website \[[@B28]\], linked also through the FGP homepage \[[@B29]\]. PGN was designed as a general-purpose EST analysis pipeline and web-based database that can be readily employed as a \"front end\" for other EST sequencing projects. PGN is a trace file database accepting all standard automated sequencer file formats. Quality information in the raw trace files is used for sequence trimming and assembly, and chromatograms can be visualized through the website. The focus on trace file data distinguishes PGN from other EST databases such as PlantGDB \[[@B30]\] and the TIGR Gene Indices \[[@B31]\]. PGN also provides an EST processing and annotation service for smaller EST projects that may not have the informatics resources to generate a public database, and provides a stable web address for these projects. PGN provides public access to EST library statistics, unigene build details, EST chromatograms, and permits FGP taxon-specific BLAST \[[@B32]\] searches. Tribe analysis -------------- Tentative classification of unigenes has allowed us to identify quickly the genes represented in our EST sets. We created an objectively defined scaffold for classification through cluster analysis of the Arabidopsisand rice proteomes. The PlantTribes database \[[@B33]\] can be searched using BLAST, or by query with Arabidopsisor rice sequence IDs \[[@B34]\], sequence annotations, Pfam accession IDs \[[@B35]\] or keywords. To construct PlantTribes, predicted protein sequences from the Arabidopsis thalianavar. Columbia and Oryza sativavar. japonica(rice) genomes were downloaded from TIGR \[[@B34]\]. The BLASTP program \[[@B32]\] was used to compare all sequences to each other, and the similarity-based clustering procedure TribeMCL \[[@B36],[@B37]\] was used to group proteins into putative gene families within our PlantTribes database. Of the 20,992 tribes identified by MCL cluster analysis of the Arabidopsisand rice proteomes, 60 PlantTribes included at least one of 100 known floral development regulators (Tables [2](#T2){ref-type="table"} and [3](#T3){ref-type="table"}). Unigene overlap --------------- To estimate the complexity of the non-normalized nor tissue subtracted FGP cDNA libraries and the underlying floral transcriptomes, we analyzed predicted functions of the FGP unigenes. Overall, functional classification of FGP unigene assemblies shows that EST sequencing has captured a nearly uniform representation of the sampled transcriptomes. On average, 53% of unigenes from each taxon match Arabidopsisgenes with an e-value of 1.0e^-10^or better. An analysis of GO annotations \[[@B38]\] for these genes shows that the FGP unigene sets provide a remarkably consistent sampling of functional classes defined for the Arabidopsisproteome (Fig. [2](#F2){ref-type="fig"}). Moreover, similar GO classification frequencies were observed in a subset of 11,974 genes that were found to be expressed at moderate-to-high levels (\>100 units) in an Affymetrics microarray analysis of young (stage 3) Arabidopsisinflorescences (Zhang et al. unpublished data). Estimation of unigene overlap is inherently error-prone because best BLAST hits are not necessarily orthologs. Moreover, even when orthology is established through formal phylogenetic analysis, similarity in function does not necessarily follow orthology \[e.g. \[[@B39]\]\]. We used the PlantTribes database to estimate the overlap among our unigene sets at the gene family level. Unigenes were sorted into the tribes if they have best BLASTX hits to any member of the tribe. Each taxon has unigenes sorted to 19--51% of the 60 tribes that include floral development regulators (Tables [2](#T2){ref-type="table"} and [3](#T3){ref-type="table"}). On average, 70% of the gene families represented in one EST set are represented in at least one other EST set. As expected, the most overlap occurs in the largest gene families (Fig. [3](#F3){ref-type="fig"}). Among the 100 Arabidopsisfloral regulatory genes identified from the literature (Tables [2](#T2){ref-type="table"} and [3](#T3){ref-type="table"}), 67 have closely related homologs (best BLAST hit) in at least one FGP EST set. On average, ESTs with a best hit among the 100 listed Arabidopsisflower development genes constitute approximately 1% of each FGP EST set. The average overlap of best hits to these floral regulatory genes was 30.8% between pairs of EST sets. The Amborella, Nupharand Eschscholziaunigene sets shared three-way overlap in best BLAST hits to six floral development genes (AGAMOUS, AGL6, APETALA3, SEPALLATA1, AXR6, and YABBY3; Table [2](#T2){ref-type="table"}) and these species plus Perseashared four-way overlap in best BLAST hits two of these genes (APETALA3and AXR6). In addition, the FGP ESTs/unigenes were found to match on average 4.5% of the apparent single-gene/taxon tribes (representing putative single-copy genes) from Arabidopsis. The average overlap of these putative single copy genes between two FGP taxa was 28%, and three-way overlap of such genes among the Amborella, Nupharand EschscholziaEST sets was 10% (representing 26 genes). The high frequency (66%) of crucial Arabidopsisfloral regulators identified in BLAST searches as best hits for sequences in one or more of our EST sets (Tables [2](#T2){ref-type="table"} and [3](#T3){ref-type="table"}) indicates that FGP EST sets are a valuable resource for comparative floral developmental studies. For example, identifying homologs of genes being investigated in model systems opens the door to broad cross-species comparative analyses of gene function. Of the 1453 Arabidopsisgenes that have recently been identified as having organ-specific expression within developing flowers \[[@B40]\], 388 (27%) are best BLAST matches to genes from at least one of our unigene sets (e.g. genes in Table [4](#T4){ref-type="table"}). We consider this to be a high percentage, given that our cDNA libraries were constructed from a subset of the developmental stages analyzed in the Arabidopsisstudy \[[@B40]\]. Given what is already known about gene duplications in angiosperm history \[e.g. \[[@B41]-[@B43]\]\], the BLAST based measures of overlap are almost certainly underestimates of cross-taxon sampling of orthologous gene sets (including co-orthologs \[[@B44]\]) represented in our EST sets. Whereas the measures of among-taxa overlap in gene families as defined in the PlantTribe database provide a possible upper bound on the degree of overlap among orthologous sets, simple comparison of best BLAST hits in the Arabidopsisproteome provides a likely lower bound. Formal phylogenetic analyses provide a more accurate assessment of orthologous gene sets. For example, within the MIKC MADS-box gene family, phylogenies uncover greater levels of overlap among our EST sets than were inferred from simple BLAST-based analyses. Of the 12 taxa listed in Table [1](#T1){ref-type="table"}, representatives of the DEFICIENS, GLOBOSA, AGAMOUS, FRUITFULL/SQUAMOSA, SEPALLATA, AGL6, and TM8clades have been identified in 8, 6, 5, 3, 7, 9 and 2 unigene sets, respectively. In addition to providing more accurate estimation of overlap among the transcriptomes being sampled in EST studies, phylogenetic analyses of gene families provide insights into the evolutionary history of genes characterized in model systems. For example, recent phylogenetic surveys of MADS box genes have identified gene duplication events that appear to be associated with the origin and rise of angiosperms and the radiation of core eudicots \[e.g., \[[@B45]-[@B50]\]\]. Phylogenetic analyses of poorly understood gene families can also provide valuable insights into both function and phylogenetic history. For example, two of the 18 genes identified as differentially expressed in petals by Wellmer et al. \[[@B40]\] belong to a single gene family identified in PlantTribes \[[@B33]\]. This gene family includes 10 Arabidopsisgenes and 16 rice genes, all containing a plant-specific domain, DUF642 \[[@B35]\], the function of which is unknown. DUF642 homologs were identified in ESTs sampled from Amborella, Nuphar, Persea, Liriodendronas well as 16 additional plant species included in the TIGR plant gene indices \[[@B31]\]. A phylogeny of these sequences reveals one weakly supported and two well supported subfamilies (Fig. [4](#F4){ref-type="fig"}). We will refer to these putative subfamilies as clades A, B, and C, respectively (Fig. [4](#F4){ref-type="fig"}). The well supported placement of a gymnosperm gene (from pine) as sister to all angiosperm genes in clade C indicates that the origin of this subfamily predated the common ancestor of angiosperms and gymnosperms. Asterid, rosid and monocot genes can be identified in each of the three subfamilies; magnoliid genes are placed in clades A and B; and the basal-most angiosperms (Amborellaand the Nymphaeales) are represented in both clades B and C. The phylogeny suggests that one of the two genes identified by Wellmer et al. \[[@B40]\] is a recent duplicate, the sister gene of which is not differentially expressed in petals. Determining the expression patterns of other DUF642 genes in Arabidopsis, rice, and other plant species and mapping this information onto the phylogeny would be a first step toward understanding their current function and functional evolution. Discussion ========== Proof-of-concept: EST coverage ------------------------------ Relative to the Gene Ontology (GO) functional classification, the FGP is detecting new, translated sequences with astonishing similarity to frequencies known for the entire Arabidopsistranscriptome. These gene discovery frequencies support the preliminary hypothesis that the functional complexity of the floral transcriptome roughly equals that of the global plant transcriptome. This point is supported by a comparison of the predicted Arabidopsisproteome and the collection of genes identified as moderate-to-highly expressed in young Arabidopsisinflorescences (Fig. [2](#F2){ref-type="fig"}, Zhang et al. unpublished data). Moreover, these detection rates ensure that FGP sequences will be of great interest to the evolutionary biologists for analyses of gene and genome duplications and selection at the molecular level, as well as to the plant molecular biological community in general. For phylogenetics, FGP unigenes assigned to single-gene PlantTribes (such as FRIGIDAand GIGANTEA) could be used to develop nuclear markers spanning all angiosperms. Indeed, further comparative functional analyses of such single-copy genes could be used to test whether natural selection culls duplicates from plant genomes following segmental or genome-wide duplication events. Proof-of-concept: MADS-box genes -------------------------------- The efficacy of a comparative genomics approach to the discovery of genes involved in floral development is illustrated in an analysis of MIKC-type MADS-box genes. To date, 48 MADS unigenes were identified as likely orthologs to major MADS-box groups, including those of the MIKC-type MADS-box genes that encode well-characterized floral regulators, such as DEFICIENS, GLOBOSA, and AGAMOUS. Further phylogenetic and expression analyses of these newly identified genes from the FGP species promise to yield new insights into the evolution of this gene family critical for flower development. In addition, sequences representing the TM8clade have been identified in our Amborellaand PerseaEST sets. The TM8gene, expressed in developing tomato flower buds, was sequenced at an early point in the study of MADS-box gene function \[[@B51]\]. Although a TM8ortholog, ERAF17, has been associated with female flower development in cucumber \[[@B52]\], no ortholog of TM8has been identified within the Arabidopsisor rice genomes, suggesting that it was lost from the genomes of these species. Identification of TM8orthologs in eudicots suggested that the gene duplication establishing the lineage had at least occurred among ancestral core eudicots \[[@B45]\]. However, phylogenetic placement of Amborellaand Perseagenes within this clade pushes its origin back to the oldest node in angiosperm phylogeny at least 130 Mya on the basis of fossil evidence \[[@B53],[@B54]\] and perhaps 145--208 Mya according to molecular estimates \[[@B20],[@B55]\], suggesting that the TM8-like genes were part of the basic floral \"tool-kit\" in the earliest angiosperms. Had these new orthologs not been identified, the limited understanding of the functions of TM8and ERAF17would have rendered them unlikely targets for candidate gene analyses of basal angiosperms. The high capture rate of known floral development regulators in the MADS-box gene family can be considered a proof-of-concept for the FGP. We expect this frequency to increase for MADS-box and other floral gene families as the size of FGP EST libraries and unigene sets expands. Limitations ----------- The FGP EST sets described here include sequences from non-normalized nor tissue-subtracted cDNA libraries. As such, many of the genes captured are expressed across many tissues, which could in large part account for the evenly proportioned GO classifications among FGP taxa and the Arabidopsis proteome (Fig. [2](#F2){ref-type="fig"}). Additionally, our approach to EST collection has limited the discovery of transcripts known to be rare, such as for the SUPERMANgene \[[@B56]\]. Nonetheless, this limitation can be overcome by either targeted screening of our cDNA libraries or rtPCR. Indeed, primer design for PCR amplification has already been aided by use of alignable sequences observed across multiple FGP EST sets. For example, primer pairs designed from alignments of Amborellaand Liriodendronunigenes with Arabidopsis GIGANTEAand rice homologs have been used successfully to amplify unsampled sequences in Nuphar, Acorus, Eschscholziaand Ribes. Although EST sequencing alone will result in incomplete sampling of genes and gene families across taxa, they offer many possibilities for further experimentation and hypothesis testing. For example, EST resources provide the opportunity to derive finished coding sequences that will be more useful for genetic manipulation as well as for comparative bioinformatics. Finished full-length cDNAs can be used as tools for many research endeavors, ranging from promoter isolation to anchoring of shotgun genomic sequences. Whereas incomplete sequences for sampled genes may reduce resolution and accuracy in some phylogenetic analyses, parsimony methods are relatively robust with respect to missing data when taxon sampling is extensive \[[@B57]\]. Conclusion ========== ESTs and assembled unigenes collected from placeholders for several critical lineages of basal angiosperms are helping to bridge the genomic gap between the eudicot and monocot model plant systems. Initial findings suggest that the basic floral transcriptome, as collected from non-normalized, non-subtractive cDNA libraries, are similar to the inferred Arabidopsistranscriptome in the proportions of its GO functional categories. Moreover, the rates of acquisition of known floral gene homologues among the various basal angiosperm EST sets are high. Finally, in one example of floral gene capture, representation among the ESTs of one lineage of the MADS-box gene family has set the origin of that gene group before the divergence of monocots from basal angiosperms. Together, these results provide strong proofs-of-concept for the Floral Genome Project. We anticipate that these initial findings will afford the opportunity to address a number of conspicuous evolutionary genomic questions, including reproductive organ transcriptome overlap between angiosperms and gymnosperms, genome-wide duplication history, identification of lineage-specific gene duplications and functional divergence, and analyses of adaptive molecular evolution. More generally, plant scientists may use the FGP resources and the comparative method to enhance estimates of sequence/domain conservation as well as hypotheses of function among all of the gene families captured. These resources will also be useful for designing both taxon-specific and universal primer sets for amplification and sequencing of specific genes sampled in one or more of our EST sets. Methods ======= Sampling rationale and molecular methods ---------------------------------------- We selected species for analysis (Fig. [1](#F1){ref-type="fig"}, Table [1](#T1){ref-type="table"}) by balancing the following major criteria: (1) phylogenetic position, (2) diversity of floral-organ structure (but absence of highly specialized floral features), (3) direct relevance to crop or economic plants, (4) diploid with a small genome size, (5) availability of inbred lines, when possible, (6) possession of other desirable properties, such as large numbers of flowers per plant, transformability, and having been the focus of prior flower developmental study, (7) non-duplication of ongoing studies of the floral transcriptome -- i.e., non-duplication of effort with ongoing studies of model plants (Arabidopsis, tomato, maize, rice, alfalfa, soybean, cotton, etc.). Welwitschiaand Zamiaare gymnosperms representing the gymnosperm phyla Gnetophyta and Cycadophyta, respectively. Sequence data from these species and a growing number of conifer species are providing insights into the gene family content in the most recent common ancestor of angiosperms and gymnosperms. Amborella, Nuphar(water lily), and Illicium(star anise) represent the most basal clades of extant angiosperms. Saruma, Liriodendron(tulip poplar), and Persea(avocado) represent three different orders of magnoliids. Perseais a valuable fruit crop and Liriodendronis a transformable timber species. Acorus(sweetflag) is sister to all other monocot phylogeny and Eschscholzia(California poppy) is a transformable species representing the basal eudicots. Asparagus, Cucumis(cucumber), Ribes(currant, gooseberry), and Vaccinium(blueberry) are all crop species holding strategic positions in Angiosperm phylogeny. Finally, Mesembryanthemum(ice plant) is a model for the study of Crassulacean Acid Metabolism (CAM) in drought-resistant plants \[[@B58]\] and represents the Caryophyllales clade of core eudicots. Isolating high quality RNA and building cDNA libraries was particularly challenging for many of the FGP taxa, most of which are non-cultivated. The cDNA libraries from these species were constructed using pre-meiotic immature floral tissues (or reproductive structures in the case of gymnosperms), in order to enrich for early floral regulatory genes important for floral patterning and to prevent heavy representation of the many anther-specific genes that are highly expressed in pollen development. Each library, non-subtractive and non-normalized to be most representative of the floral transcriptome, was made with a signature adaptor-linker sequence to eliminate the possibility of misidentification of clones in the future. Overrepresentation of genes among the libraries was further reduced by sampling mRNA from very young floral buds which have a high diversity of cell types and lack the cells, such as tapetum and pollen, that contain a large number of specific transcripts. Detailed methods used for mRNA extraction, cDNA library construction, and information on library attributes (vector, cloning sites, titer, average insert size, etc.) can be found in Table [1](#T1){ref-type="table"} and on the FGP homepage \[[@B29]\]. Sequence processing pipeline ---------------------------- A standard sequence analysis pipeline was developed consisting of base calling using PHRED \[[@B59]\], vector and E. colisequence contamination screening, and unigene assembly. In addition, a database was developed that also serves as the back end for the Plant Genome Network website \[[@B28]\]. The analysis pipeline and the sequence database are tightly integrated. The quality screen consisted of trimming low-quality sequence, based on PHRED scores, using a custom algorithm. To extend the high-quality sequence as far as possible given a particular quality threshold, the sequence was scanned and, concomitantly, the difference between the quality score and the quality threshold (termed the \"adjusted score\") was integrated over the length of the sequence. The high-quality sequence was defined as the region of sequence in which the integration of the adjusted score was maximal; this region can include small regions of lower-scoring nucleotides if they are \"compensated\" by higher-scoring downstream sequence. Next, putative polyA tails were removed if they contained more than 20 consecutive adenine residues. A contamination screen was performed to remove E. colichromosomal sequences from the dataset. In a final quality screening step, sequences with lengths below a certain threshold (150 bp), sequences with more than 4% ambiguous base calls (Ns), and sequences with a complexity below a given threshold (defined as sequence composed of more than 60% of a given nucleotide) were rejected. The rejected sequences were not used in unigene builds, but were retained in the database along with information as to why they were rejected. Unigene building ---------------- Unigene sets were built for each species by combining the sequences from all available libraries for that species. The sequences were first pre-clustered, and these clusters were then assembled using the cap3 \[[@B60]\] program. The cap3 identify parameter was set at 95%. Unigene sequences were also checked for length, complexity and contamination, and chimera detection was performed. The builds were uploaded to the PGN database, where each unigene was assigned a unique unigene ID. Subsequent unigene builds of the same libraries attributed new IDs to all unigenes. Unigenes from a newer build were then tracked to the older builds through the ESTs that they share, and as such, a complete history of unigene IDs is available on the website for tracking unigenes through the different builds. Annotation of sequence data --------------------------- Several strategies were used to get the best possible annotation for the unigene sequences. First, BLAST was used to find the best match for each unigene sequence in the GenBank NR database, and in the complete coding sequences from Arabidopsis\[[@B34]\]. These annotations were stored in the database and serve as the primary source of FGP sequence annotation. To get a better overview for the annotations, we used the Gene Ontology (GO) system \[[@B38],[@B61]\] to compare the sequence annotations from different species. Gene Ontology is a hierarchical system representing biological knowledge. Many model systems, such as Arabidopsis, are being annotated using GO \[[@B38],[@B61]\]. The ArabidopsisGO annotations were transferred to each unigene that had a match with an Arabidopsissequence with an e value less than 10^-20^. For the comparison of annotations among species, we focused on the biological functional GO category. We selected a number of high-level GO function terms as a GO slim vocabulary, and then mapped the GO annotations to the GO slim terms using the map2slim.pl script provided by the Gene Ontology consortium \[[@B38]\]. Phylogenetic analyses of putative gene families ----------------------------------------------- Floral Genome Project unigene assemblies and finished cDNAs were regularly subjected to phylogenetically-based classification. The general procedure was as follows: (1) for each PlantTribe of interest, use all Arabidopsis, rice and FGP sequences to search public databases \[[@B30],[@B31],[@B61]\] for similar protein coding genes (e \< 10^-20^), (2) machine align all sequences using ClustalW \[[@B64]\], T-Coffee \[[@B65]\] or Muscle \[[@B66]\], (3) manually assess alignment, removing all highly divergent sequences, and then repeat step 2, (4) subject alignment to fast parsimony analysis both with and without branch support estimation \[[@B67],[@B66],[@B69]\]. The parsimony phylogeny shown for the DUF642 gene family (Fig. [4](#F4){ref-type="fig"}) is the strict consensus of 67 equally parsimonious trees estimated for an amino-acid alignment. The parsimony analysis was executed in PAUP\* \[[@B66]\] with 100 random addition replicates and tree bisection-reconnection (TBR) branch swapping. Bootstrap analysis was performed with 250 replicates. The alignment is available in NEXUS format through the FGP website \[[@B29]\]. Two regions of the alignment were deemed questionable and removed from the phylogenetic analysis. Authors\' contributions ======================= VAA, DES, LAM, and JHL-M drafted the manuscript. JEC, WGF, LLL, YH, MB, SK, M-JY, BJB, and XZ acquired most of the data. RP-T and SES made essential contributions to Cucumis and Liriodendron EST library construction. PKW, DCI, TMS, LAM, and JHL-M designed and performed bioinformatic analyses. VAA, DES, JEC, MWF, ST, DGO, PSS, HM, CWD, and JHL-M conceived of the study, participated in its design and coordination, and helped complete the manuscript. All authors read and approved the final manuscript text. Acknowledgements ================ This study was funded through the NSF Plant Genome Research Program (project DBI-0115684) and USDA-IFAFS Grant \#80388 to Mike Havey for AsparagusEST sequences. The Arabidopsismicroarray experiments were supported by a grant from the NSF Plant and Microbial Development Program (IBN-0077832) to H.M. We thank our advisory panel, Ginny Walbot, Elliot Meyerowitz, Rebecca Doerge and Peter Endress for valuable suggestions and encouragement. Sheila Plock and Stephanie Choirean provided valuable technical assistance. We thank Laura Zahn and Hongzhi Kong for assistance in tissue collection. We greatly appreciate the insightful suggestions from Günter Theißen, John Bowers and an anonymous reviewer concerning this manuscript. L.J. Rowland supplied us with the blueberry cDNA library for EST sequencing. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Well- supported evolutionary relationships among FGP species and other genomic models are shown in this phylogenic tree of seed plants. Red taxon names indicate those species for which we aim to sequence 10,000 ESTs, green taxon names indicate species for which we are sequencing 2000 ESTs, and blue taxon names indicate species for which large EST sets are already available in public databases \[24, 25, 28, 29, 61\] or will soon become available \[70\]. ::: ![](1471-2229-5-5-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### The relative frequencies of ESTs assigned to GO Biological process classes are quite similar across our study taxa. Class frequencies are shown for ten EST sets, the inferred Arabidopsis proteome, and Arabidopsis genes with moderate-to-high expression in young inflorescences (stage 3). ::: ![](1471-2229-5-5-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### TribeMCL gene clusters (Tribes) with floral development genes vary in size and tend to include similar numbers of rice and Arabidopsisgenes (left). These gene families are well represented in our EST sets (right and Tables 2 and 3). \The unigene counts (right) for the CLAVATAgene family have been halved. ::: ![](1471-2229-5-5-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### A phylogeny for the DUF642-domain gene family indicates that two Arabidopsisgenes with differential expression in petals \[40\] are not the products of a recent duplication event. Genes with the plant-specific but functionally uncharacterized domain family DUF642 form three clades (A, B, and C). In addition to the Arabidopsisgenes with petal-specific expression patterns, Clade A includes asterid, Vitis, monocot and magnoliid genes, indicating that the clade predates the divergence of these lineages (Fig. 1). The genomes of basal-most angiosperm lineages (Amborellaand Nymphaeales) and gymnosperms may also contain unsampled Clade A genes. Clade B genes were sampled from Nuphar(Nymphaeales) and Clade C genes were identified from Amborellaand Pinus. Bootstrap support values (\>50%) are shown above each branch. Abbreviated taxon names associated with some gene sequences: Os, Oryza sativa(rice); At, Arabidopsis thaliana; Gossyp, Gossypiumspp. (cotton); G.max, Glycine max(soybean); L.japo, Lotus japonicus; M.trun, Medicago truncatula; V.vini, Vitis vinifera(grape); L.sati Lactuca sativa(lettuce); S.tube, Solanum tuberosum(potato); S.escu, Solanum esculentum(tomato); M.crys, Mesembryanthemum crystallinum(ice plant); A.cepa, Allium cepa(onion); T.aest, Triticum aestivum(wheat); H.vulg, Hordeum vulgare(barley); S.bico, Sorghum bicolor; S.offi Saccharum officinarum(sugarcane); Z.mays, Zea mays(maize); L.tuli, Liriodendron tulipifera(tulip poplar); P.amer, Persea americana(avocado); N.adve Nuphar advena; A.tric, Amborella trichopoda. ::: ![](1471-2229-5-5-4) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Current (12/01/04) statistics for Floral Genome Project cDNA libraries, EST sequences and unigene builds. We will perform 10,000 EST sequencing reactions for each taxon listed in the top portion of the table and 2000 ESTs for each taxon in the bottom portion of the table. ::: Taxon* Primary Titre (pfu) Amplified Titre (pfu/ml) Avg Insert size (bp) ESTs to date ^a^Unigenes ^b^Observed Redundancy -------------------------------------------------------------------------------------------- ------------------------- ------------------------------ -------------------------- ------------------ ----------------- ---------------------------- Libraries for deep sequencing Welwitschia mirabilis(m) -- Gymnosperm 3.25E+05 6.00E+08 1382 3732 2771 34.7% Amborella trichopoda(m) -- Basal angiosperm 2.24E+06 1.37E+10 1611 4047 6099 41.5% Amborella trichopoda(f) 4.98E+06 1.40E+10 1031 4442 Nuphar advena (Water lily, Spadderdock) -- Basal angiosperm 2.00E+06 3.20E+10 1134 8442 6205 36.1% Acorus americanus(Sweet flag) -- Basal monocot. 2.80E+06 6.00E+09 1083 5883 3067 28.8% ^c^Asparagus officinalis(m) -- Transformable nongrass monocot 1.50E+06 1.20E+10 1468 5188 4560 61.5% Asparagus officinalis(f) 1.30E+06 1.40E+10 1200 2174 Persea americana(Avocado) -- Cultivated magnoliid 2.74E+06 2.57E+10 1349 8735 5314 41.3% Liriodendron tulipifera(Tulip Poplar, Yellow Poplar) -- Transformable lumber species 3.00E+06 2.00E+10 1346 9531 6520 46.2% Saruma henryi(Upright Wild Ginger) -- member of Aristolochiaceae with bipartite perianth 1.97E+06 1.67E+10 1587 3230 2631 21.8% Eschscholzia californica(California poppy) -- Transformable basal eudicot 7.00E+06 1.68E+11 1702 9079 6015 46.18% Libraries for shallow sequencing ^d^Cucumis sativus (m)(Cucumber) -- rosid \-\-\-\-\-- \-\-\-\-\-- \-\-\-\-\-- 1107 1648 23.5% ^d^Cucumis sativus (f) \-\-\-\-\-- \-\-\-\-\-- \-\-\-\-\-- 928 Ribes americanum (Black currant) 2.58E+06 2.25E+10 1200 2238 1791 25.0% ^e^Vaccinium corymbosum (Blueberry)-- basal asterid \-\-\-\-\-- \-\-\-\-\-- \-\-\-\-\-- 1758 1549 13.5% Total: 70,514 48,170 m indicates library constructed from male tissues; f for female tissues. ^a^Unigene numbers are shown in the first line for taxa with multiple libraries. ^b^Observed redundancy was measured for each taxon as (EST\# -- Unigene \#)/Unigene\#. ^c^5188 ESTs from the male Asparaguslibrary were sequenced in collaboration with Mike Havey (University of Wisconsin) and Chris Town (TIGR). ^d^Male and female Cucumisflower bud libraries described by Perl-Treves et al \[71\] ^e^Vacciniumyoung inflorescence library was provided by Jeannine Rowland (USDA) \[72\]. ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Number of FGP unigenes that are best BLAST hits to Arabidopsisfloral developmental regulation genes, with corresponding tribe ID number and number of Arabidopsisand rice genes in these tribes. Species IDs: Aam, Acorus americanus; Aof, Asparagus officinalis; Ath, Arabidopsis thaliana; Atr, Amborella trichopoda; Eca, Eschscholzia californica; Ltu, Liriodendron tulipifera; Nad, Nuphar advena; Osa, Oryza savita; Pam, Persea americanus; She, Saruma henryi; Wmi, Welwitschia mirabilis. ::: Gene ID Annotation Tribe^a^ Ath^b^ Osa^b^ Aam Aof Atr Eca Ltu Nad Pam She Wmi Tot ------------- ------------------------------------------------ -------------- ------------ ------------ --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- At2g45190 AFO, YABBY1 1010 4 7 3 1 1 3 2 12 At4g18960 AG, AGAMOUS 65 46 51 2 1 1 4 At4g09960 AGL11, MADS-box protein 65 46 51 1 1 At2g45660 AGL20, SOC 65 46 51 1 1 2 At4g24540 AGL24, MADS-box protein 65 46 51 At2g03710 AGL3, MADS-box protein 65 46 51 1 1 2 At2g45650 AGL6, MADS-box protein 65 46 51 1 1 1 2 1 2 9 At4g37750 ANT, AINTEGUMENTA 123 18 38 2 1 1 4 At1g69120 AP1, APETALA 1 65 46 51 1 1 3 At4g36920 AP2, APELATA 2 (FL1, FLOWER1) 123 18 38 1 1 1 3 8 At3g54340 AP3, APETALA 3 65 46 51 1 2 1 2 2 1 9 At1g75950 ASK1 122 19 38 1 1 1 3 At5g42190 ASK2 122 19 38 3 2 7 1 16 At4g02570 AXR6, AUXIN RESISTANT 6 324 10 16 2 2 1 7 1 2 1 17 At1g01040 CAF, CARPEL FACTORY (SUS1) 446 8 13 2 1 3 At1g26310 CAL, CAULIFLOWER 65 46 51 1 1 At1g75820 CLV1, CLAVATA 1 (FASCIATA 3) 8 194 478 2 2 At1g65380 CLV2, CLAVATA 2 8 194 478 1 1 At2g27250 CLV3, CLAVATA 3 10933 1 0 At1g69180 CRC, CRABS CLAW 1010 4 7 At3g61850 DAG1, DOF AFFECTING GERMINATION 1 93 36 36 1 1 At2g33860 ETT, ETTIN 117 26 34 1 3 At3g59380 FTA, FARNESYLTRANSFERASE A 2266 1 5 1 1 At3g30260 FUL, FRUITFULL (AGL8) 65 46 51 At4g20910 HEN1, HUA ENHANCER 1 3162 2 2 1 2 3 At2g06990 HEN2, HUA ENHANCER 2 1435 4 4 1 1 1 1 4 At5g64390 HEN4, HUA ENHANCER 4 601 10 7 2 1 1 1 6 At3g12680 HUA1, ENHANCER OF AG-4 1 469 10 9 1 2 1 4 At5g23150 HUA2, ENHANCER OF AG-4 2 1582 4 4 At1g23420 INO, INNER NO OUTER 1010 4 7 At5g16560 KAN, KANADI 100 25 43 1 1 At5g61850 LFY, LEAFY 7107 1 1 At4g32551 LUG, LEUNIG 1572 2 6 1 1 1 2 1 1 7 At4g10350 NAM, NO APICAL MERISTEM 30 82 105 At1g69490 NAP, NAC-LIKE, ACTIVATED BY AP3/PI 30 82 105 1 1 At1g68640 PAN, PERIANTHIA 253 10 21 1 1 At5g20240 PI, PISTILLATA 65 46 51 1 1 2 1 5 At2g34650 PID, PINOID 87 36 42 1 1 3 At2g28610 PRS, PRESSED FLOWER 626 8 8 At5g35770 SAP, STERILE APETALA 9932 1 0 At5g15800 SEP1, SEPALLATA1 (AGL2) 65 46 51 3 4 1 1 9 At3g02310 SEP2, SEPALLATA2 (AGL4) 65 46 51 1 1 1 3 At1g24260 SEP3, SEPALLATA3 (AGL9) 65 46 51 3 3 1 7 At1g43850 SEU, SEUSS 1762 4 3 2 2 At3g58780 SHP1, SHATTERPROOF 1 (AGL1) 65 46 51 1 1 At2g42830 SHP2, SHATTERPROOF 2 (AGL5) 65 46 51 1 2 3 At1g02065 SPL8, SQUAMOSA PROMOTER BINDING PROTEIN-LIKE 8 284 15 13 1 2 1 5 At3g23130 SUP, SUPERMAN 1084 5 5 At5g03840 TFL1 TERMINAL FLOWER 1 397 6 17 At3g22780 TSO1, 1120 4 6 1 1 2 At1g30950 UFO, UNUSUAL FLORAL ORGANS 4059 1 3 At2g17950 WUS, WUSCHEL 626 8 8 At4g00180 YABBY3 1010 4 7 1 1 2 1 1 6 ^a^Tribe ID\'s are reference numbers for the PlantTribes database \[33\]. ^b^Gene family size is represented by the number of rice (osa) and Arabidopsis(ath) genes in each medium stringency tribe. ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Number of FGP unigenes that are best BLAST hits to Arabidopsisflowering time genes, with corresponding tribe ID number and number of Arabidopsisand rice genes in these tribes (species abbreviations as in Table 2) ::: Gene ID Annotation Tribe Ath Osa Aam Aof Atr Eca Ltu Nad Pam She Wmi Tot ------------- ------------------------------------------------- ----------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- At2g45660 AGL20, SOC 65 46 51 1 1 2 At4g24540 AGL24, MADS-box protein 65 46 51 At2g46830 CCA1, CIRCADIAN CLOCK ASSOCIATED 1 3546 2 1 1 2 At2g25920 ELF3 8701 1 1 At5g11530 EMF1, EMBRYONIC FLOWER 1 10070 1 0 At5g51230 EMF2, EMBRYONIC FLOWER 2 1026 5 6 1 3 4 At4g15880 ESD4, EARLY IN SHORT DAYS 4 10325 1 0 At4g16280 FCA, FCA 2386 2 0 At4g35900 FD, FD 6153 2 0 At1g04400 FHA (CYR2, CRYPTOCHROME 2) 2549 2 3 1 1 At1g68050 FKF1, FLAVIN-BINDING KELCH DOMAIN F BOX PROTEIN 1047 6 5 1 1 At5g10140 FLC, FLOWERING LOCUS F 65 46 51 At2g43410 FPA, FPA 2343 3 3 2 2 At5g24860 FPF1, FLOWERING PROMOTING FACTOR 1 311 11 16 1 1 At4g00650 FRI, FRIGIDA 6545 1 1 1 1 At1g65480 FT FLOWERING LOCUS T 397 6 17 At3g59380 FTA, FARNESYLTRANSFERASE A 2266 1 5 1 1 At3g30260 FUL, FRUITFULL, AGL8 65 46 51 At4g25530 FWA, FWA 230 18 15 At5g13480 FY, FY 7822 1 1 2 2 At1g14920 GAI, GA INSENSITIVE 74 27 62 At1g22770 GI, GIGANTEA 8967 1 1 1 3 2 2 1 10 At4g08920 HY4, ELONGATED HYPOCOTYL 4 (CRY1) 2549 2 3 1 3 5 At2g23380 ICU1, INCURVATA 1 2735 3 2 At4g02560 LD, LUMINIDEPENDENS 8840 1 1 2 2 At5g61850 LFY, LEAFY 7107 1 1 At1g01060 LHY, LATE ELONGATED HYPOCOTYL 3546 2 1 1 1 2 At1g77080 MAF1, MADS AFFECTING FLOWERING 1 65 46 51 At5g65050 MAF2.4, MADS AFFECTING FLOWERING2 65 46 51 At5g65060 MAF3, MADS AFFECTING FLOWERING 3 65 46 51 At5g65070 MAF4.5, MADS AFFECTING FLOWERING 4 VARIANT V 65 46 51 At5g65080 MAF5.2, MADS AFFECTING FLOWERING 5 VARIANT II 65 46 51 At2g19520 NFC4, FVE 1299 6 3 2 1 1 1 3 8 At1g09570 PHYA, FAR RED ELONGATED 1 1254 5 4 2 1 3 1 7 At2g18790 PHYB, PHYTOCHROME B 1254 5 4 1 1 2 1 5 At5g35840 PHYC, PHYTOCHROME DEFECTIVE C 1254 5 4 1 1 2 At4g16250 PHYD, PHYTOCHROME DEFECTIVE D 1254 5 4 At4g18130 PHYE, PHYTOCHROME DEFECTIVE E 1254 5 4 At1g73590 PIN1, PIN-FORMED 1 405 8 14 2 2 At2g01570 RGA1, REPRESSOR OF GA1-3 1 74 27 62 1 1 At1g02065 SPL8, SQUAMOSA PROMOTER BINDING PROTEIN-LIKE 8 284 15 13 1 2 1 5 At3g11540 SPY, SPINDLY 6916 1 1 1 1 2 At2g22540 SVP, SHORT VEGETATIVE PHASE 65 46 51 1 1 1 3 At2g28290 SYD, SPLAYED 135 20 30 1 1 At2g26670 TED4, ELONGATED HYPOCOTYL 1 2358 4 2 1 1 At5g03840 TFL1, TERMINAL FLOWER 1 397 6 17 At5g17690 TFL2, 5181 1 2 1 1 1 4 At5g61380 TOC1, PSEUDO-RESPONSE REGULATOR 1 769 6 8 1 2 2 5 At5g61150 VIP4, VERNALIZATION INDEPENDENCE 4 2513 1 4 2 2 At3g18990 VRN1, REDUCED VERNALIZATION RESPONSE 1 4792 2 1 1 1 At4g16845 VRN2, REDUCED VERNALIZATION RESPONSE 2 1026 5 6 At1g80730 ZFP1, ZFP1 511 6 13 At5g57360 ZTL, ZEITLUPE 1047 2 1 3 ::: ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Distribution of best matches to floral organ-specific Arabidopsisgenes \[40\] among seven unigene sets (species abbreviations as in Table 2). Shown in bold are two DUF642-domain genes with differential expression in petals. ::: Organ Gene ID Annotation Tribe Ath Osa Aam Aof Atr Eca Ltu Nad Pam She Wmi Tot ----------- --------------- ------------------------------------------------------ ----------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- carpel At5g44635 similar to putative CDC21 protein 549 8 10 1 1 1 3 carpel At1g71691 hypothetical protein with GDSL-like motif 40 84 79 1 1 1 2 1 6 carpel At5g53120 spermidine synthase 2331 3 3 1 1 3 1 6 carpel At5g07280 receptor-like protein kinase-like protein 8 194 478 2 1 2 1 7 carpel At3g51860 Ca2+/H+-exchanging protein-like 823 6 7 1 1 1 2 2 1 8 carpel At5g06860 polygalacturonase inhibiting protein (PGIP1) 8 194 478 1 1 1 1 1 8 carpel At5g02540 putative protein 348 11 14 1 1 1 1 1 1 2 1 9 carpel At5g59320 nonspecific lipid-transfer protein precursor -- like 277 11 18 1 1 1 2 2 2 1 10 petal At3g62700 glutathione-conjugate transporter, putative 192 16 22 1 1 2 1 5 petal At5g11420 putative protein (DUF642 Domain) 328 10 16 1 1 2 1 1 6 petal At1g80240 putative protein (DUF642 Domain) 328 10 16 1 1 2 1 1 6 sepal At1g69120 floral homeotic gene APETALA1 65 46 51 1 1 3 stamen At5g14780 formate dehydrogenase (FDH) 4097 1 3 1 1 1 1 3 1 8 stamen At1g52570 phospholipase D, putative 300 10 17 4 1 1 2 1 9 stamen At3g09390 metallothionein-like protein 6057 2 0 1 2 1 2 2 10 stamen At3g03080 putative NADP-dependent oxidoreductase 315 14 13 1 2 1 1 1 3 11 stamen At3g62290 ADP-ribosylation factor-like protein 197 19 17 4 3 2 1 1 2 1 14 stamen At5g43330 cytosolic malate dehydrogenase 1443 4 4 3 2 1 1 3 1 3 14 stamen At1g13950 initiation factor 5A-4 1208 3 7 1 3 1 3 2 5 18 stamen At5g45775 Expressed protein 923 4 7 2 1 7 2 1 3 1 1 18 stamen At5g14670 ADP-ribosylation factor -- like protein 197 19 17 3 2 6 3 3 4 2 24 :::	PubMed Central	2024-06-05T03:55:55.733683	2005-3-30	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1083416/", "journal": "BMC Plant Biol. 2005 Mar 30; 5:5", "authors": [ { "first": "Victor A", "last": "Albert" }, { "first": "Douglas E", "last": "Soltis" }, { "first": "John E", "last": "Carlson" }, { "first": "William G", "last": "Farmerie" }, { "first": "P Kerr", "last": "Wall" }, { "first": "Daniel C", "last": "Ilut" }, { "first": "Teri M", "last": "Solow" }, { "first": "Lukas A", "last": "Mueller" }, { "first": "Lena L", "last": "Landherr" }, { "first": "Yi", "last": "Hu" }, { "first": "Matyas", "last": "Buzgo" }, { "first": "Sangtae", "last": "Kim" }, { "first": "Mi-Jeong", "last": "Yoo" }, { "first": "Michael W", "last": "Frohlich" }, { "first": "Rafael", "last": "Perl-Treves" }, { "first": "Scott E", "last": "Schlarbaum" }, { "first": "Barbara J", "last": "Bliss" }, { "first": "Xiaohong", "last": "Zhang" }, { "first": "Steven D", "last": "Tanksley" }, { "first": "David G", "last": "Oppenheimer" }, { "first": "Pamela S", "last": "Soltis" }, { "first": "Hong", "last": "Ma" }, { "first": "Claude W", "last": "dePamphilis" }, { "first": "James H", "last": "Leebens-Mack" } ] }
PMC1083417	Background ========== Ever since the birth of echocardiography, 50 years ago, this non-invasive diagnostic tool became a milestone in the clinical evaluation of cardiovascular patients, due to its diagnostic accuracy. Owing to advances in technology, the ultrasonic industry has recently produced hand-held devices (HHD) that are miniaturized, compact and battery-equipped echocardiographic systems. Basic versions of HHD usually allow only bidimensional imaging and colour flow Doppler analysis, but M-Mode visualization, continuous and pulsed wave Doppler imaging, ECG synchronization and other tools have recently been integrated in better-equipped HHD. These machines could offer some advantages compared with standard echocardiographic devices (SED), due to their simplicity of use, immediate availability at the patient\'s bedside, transportability and relatively low cost. The potential usefulness of HHD has been successfully assessed in a wide range of clinical conditions. HHD has been reported to improve detection of relevant cardiovascular pathologies \[[@B1]\] or unknown cardiac disorders \[[@B2]\], and to allow a good assessment of cardiac anatomy and function \[[@B3]-[@B5]\]. Furthermore they also have been shown to ensure a reliable assessment of left ventricular hypertrophy \[[@B6]\] and abdominal aortic aneurysm \[[@B7],[@B8]\]. It has also been suggested that internists may use HHD without formal training in echocardiography and after a limited echo-training period \[[@B9]-[@B11]\]. In addition, HHD has demonstrated acceptable accuracy during patient transport \[[@B12]\] and in the context of ultrasound-guided pericardiocentesis and thoracentesis \[[@B13]\]. However, HHD seems to have a narrower diagnostic field when compared with SED in the evaluation of critically ill patients \[[@B14],[@B15]\]. Based on this evidence, HHD may be expected to become an important additional diagnostic tool during outpatients cardiologic consulting or in non-cardiologic hospital sections. Nevertheless, the potential role of HHD in these settings has never been investigated. Methods ======= Eighty-seven consecutive patients (47 males and 40 females, mean age 66.1 ± 15.2 years), who visited our Hospital for cardiologic consulting, have been included in this study. Each participant in the study had one or more referral questions among the following: hypertension, dyspnoea, chest pain, palpitations (Table [1](#T1){ref-type="table"}). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Referral questions in the study population ::: Referral question No. of Patients ---------------------------------- --------------------- Hypertension 45 (51.7%) Dyspnoea 34 (39.1%) Chest pain 30 (34.5%) Palpitations 6 (6.9%) Pericardial effusion control 3 (3.4%) ::: All patients underwent routine physical examination, resting ECG and echocardiographic evaluation using a basic model of HHD (Opti-Go, Philips Medical System). The end-diastolic left ventricular (LV) diameter, interventricular septum and posterior wall thicknesses, the size of the aortic root, bulb and ascending segment, and the end-systolic left atrial antero-posterior diameter were measured by B-mode imaging, using the parasternal long-axis view. An estimation of LV ejection fraction and regional wall motion have also been assessed, with computation of the wall motion score index. A gross assessment of aortic, mitral, tricuspidalic and pulmonic valve features (calcifications, abnormal movements) has been performed. The presence and severity of regurgitations has been estimated using colour flow Doppler imaging. The presence of pericardial effusion has been evaluated using multiple views, and subcostal view was used to measure abdominal aortic diameter. The cardiologist, whenever possible, formulated a diagnosis at the end of the exam. The percentage of subjects for whom the diagnosis was considered satisfactory -- i.e., when findings were judged to be reasonably adequate for final diagnostic and therapeutic conclusions and no further diagnostic evaluation was needed -- was used to quantify the \"conclusiveness\" of HHD evaluation. Within 24 hours, all patients underwent a second echocardiographic evaluation, performed by a second cardiologist with similar experience and echocardiographic competence, blinded to the results of the other investigator. The examination has been performed by using a SED (Agilent Technologies, Sonos 5500). The agreement between the first and the second echocardiographic exam has been assessed by controlling the percentage of concordant diagnostic conclusions between the two evaluations. Data are shown as mean ± SD for continuous variables. The comparison of the examination time between HHD and SED has been assessed using one-way ANOVA. The chi-squared test was used to compare the percentages of conclusiveness between the two ultrasound machines. In the case of expected frequencies ≤ 5, the Fisher\'s exact test was performed. A P value \<0.05 was considered for statistical significance. The statistical analysis was performed using the SPSS (Statistical Package for the Social Sciences, Chicago, Illinois) software. Results ======= Conclusiveness of HHD and SED ----------------------------- Mean examination time was 6.7 ± 1.5 minutes using HHD and 13.6 ± 2.4 minutes using SED (p \< 0.05). The echocardiographic evaluation performed using HHD was considered satisfactory in 74/87 patients, corresponding to 85.1% conclusiveness. After examination using SED, the diagnosis was satisfactory in 83/87 patients, which corresponded to a 95.4% conclusiveness (p = 0.02). Among the 13 patients in whom HHD examination was not conclusive, 6 had poor acoustic window, 4 showed a colour Doppler pattern suggestive of aortic stenosis, and 3 presented critical conditions. All 4 subjects for whom the SED examination were considered to be satisfactory had inadequate acoustic window. Agreement between HHD and SED ----------------------------- Among the 74 patients for whom the examination using the HHD was conclusive, the diagnosis was concordant with that obtained following the examination with the SED in 62 cases (83.8%) (images obtained with HHD and SED both in diseased \[Figure [1](#F1){ref-type="fig"}\] and normal \[Figure [2](#F2){ref-type="fig"}\] patients). The causes of the 12 diagnostic bias were the following: errors in wall thickness measurements (n = 5); incorrect assessment of mitral regurgitation (n = 3); inaccuracy in evaluation of wall motion abnormalities (n = 3) and in measurement of the diameter of the ascending aorta (n = 1). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### A case of dilatative cardiomyopathy. On the left (A), HHD image in a parasternal long-axis view obtained with HHD. On the right (B), the same patient evaluated with SED. ::: ![](1476-7120-3-7-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### A normal patient. On the left, it's shown an apical four-chambers view obtained with HHD (A). On the right (B), the same patient evaluated with SED. ::: ![](1476-7120-3-7-2) ::: Accuracy according to referral question \[Figure [3](#F3){ref-type="fig"}\] --------------------------------------------------------------------------- ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Conclusiveness of echocardiographic examination according to referral question: the diagnostic accuracy of the HHD seems to be affected by the type of referral question. HHD showed a good reliability in the examination of patients referred for arterial hypertension or palpitations, but was less accurate for those who presented dyspnoea or chest pain, as a result of biases in the measurements of wall thicknesses, and in the assessment of valvular diseases and regional wall motion. ::: ![](1476-7120-3-7-3) ::: Examinations performed using either the HHD or the SED were considered satisfactory in all 45 patients referred for arterial hypertension, corresponding to 100% conclusiveness. In 3 of them a diagnostic bias occurred using the HHD, because of incorrect wall thickness measurement. This corresponded to a 93.3% agreement between HHD and SED. Among 34 patients evaluated for dyspnoea, the examination has been considered conclusive in 27 subjects (70.6%) using the HHD and in 31 (91.2%, p = 0.11) using the SED. The causes for unsatisfactory examinations using the HHD were: poor acoustic window (n = 3 for both HHD and SED), suspected aortic stenosis (n = 2) and critical conditions (n = 2). Among the 27 patients with conclusive examination by the HHD, a diagnostic bias has been found in 4, because of errors in wall thickness measurements (n = 1), incorrect mitral regurgitation staging (n = 2), and missed ascending aorta enlargement diagnosis (n = 1). This yielded a 58.8% agreement between HHD and SED results. Among the 33 patient with chest pain, a conclusive examination has been obtained in 27 subjects (81.8%) using the HHD and in 32 (96.9%) using the SED (p = 0.30). Three of the six patients with unsatisfactory examination using the HHD had poor acoustic window (the cause of unsatisfactory examination using SED was a poor acoustic window, too), two had aortic stenosis, whereas in one subject, who showed critical conditions, the examination was not conclusive using HHD. Among the 27 patients with conclusive diagnosis following HHD examination, a diagnostic bias was observed in 5 cases: 3 because of inaccuracy in evaluation of wall motion abnormalities, 1 for inaccuracy in mitralic valve assessment and 1 for errors in wall thickness measurements. This leaded to a percentage of agreement between HHD and SED of 72.7%. Lastly, among 6 patients referred for palpitations, conclusiveness and agreement between the two techniques were 100%. Discussion ========== This is the first study, which evaluates the clinical utility of a basic model of HHD in the context of cardiologic consulting for outpatients or in non-cardiologic hospital sections. The clinical usefulness of the HHD has been reported in previous studies \[[@B16]-[@B21]\]. Potential advantages which could result from the use of a basic HHD include brief examination time, simplicity of use, fast availability at the patient\'s bedside, easy transportability and relatively low cost. Despite limitations due to the lack of M-mode imaging and power, and continuous Doppler analysis, our findings suggest that a basic model of HHD may provide a useful and reliable adjunctive diagnostic tool for cardiologic examination of both outpatients and patients admitted in non-cardiologic sections. The ecocardiographic examination, performed using the HHD, was satisfactory and conclusive in about 85% of subjects, and a good agreement between the diagnosis derived using the HHD and that obtained using the SED was obtained in this subset of patients (83.8%). Thus, a correct diagnosis was made in 71.3% of the total study population. It should also be considered that better-equipped HHD show similar advantages and may ensure further higher diagnostic accuracy in comparison with basic models, thanks to several improvements (M-mode imaging, pulsed and continuous wave Doppler facilities, ECG synchronisation, storage memories, multiple peripheral connections) \[[@B22]\], but they are more expensive. It is also to be expected that such hand-carried ultrasound devices will soon become smaller, simpler to be used and cheaper \[[@B23]\], similar to an \"ultrasound stethoscope\" \[[@B22]\]. Some examinations resulted \"unconclusive\" or unsatisfactory due to a poor acoustic window, either for HHD than SED, so this bias represents a limit for both of them. However, basic HHD should not be considered equivalent to SED. The diagnostic accuracy of the HHD seems to be affected by the type of referral question. The HHD showed a good reliability in the examination of patients referred for arterial hypertension or palpitations -also allowing a rapid screening of left ventricular hypertrophy \[[@B6]\] and of abdominal aortic aneurysm \[[@B7]\] -- but was less accurate for those who presented dyspnoea or chest pain. In the subset of subjects with dyspnoea, the percentage of satisfactory examinations tended to be higher -- although not significantly -- using the SED compared with the HHD. Also, in the subgroups of patients with dyspnoea or chest pain the agreement between the results obtained using the two ultrasound machines was suboptimal, mostly as a result of biases in the measurements of wall thicknesses, and in the assessment of valvular diseases and regional wall motion. These results suggest that the diagnostic information obtained using the HHD should be critically considered in patients with chest pain or dyspnoea. In this study we evaluated HHD comparing its clinical utility to SED: so, SED was our Gold Standard; the number of patients was also very limited (87 patients examined). Moreover, we don\'t know if HHD evaluation combined with ECG and physical examination, vsECG and physical examination only, really allow us to obtain more accurate clinical conclusions. Conclusion ========== HHD real utility is allowing evaluating patients with a gain in terms of time, shortening patients waiting lists, and reducing healthy costs. Moreover, HHD evaluation can help the physician in the choose of the therapy and in the follow-up of the patient. On these basis, HHD may generally allow a reliable cardiologic basic evaluation of outpatient or subjects admitted to non-cardiologic sections. However, in patients with chest pain and dyspnoea, the use of such devices should be performed with caution, and the diagnostic results should be considered critically.	PubMed Central	2024-06-05T03:55:55.742723	2005-3-24	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1083417/", "journal": "Cardiovasc Ultrasound. 2005 Mar 24; 3:7", "authors": [ { "first": "Giovanna", "last": "Giannotti" }, { "first": "Sergio", "last": "Mondillo" }, { "first": "Maurizio", "last": "Galderisi" }, { "first": "Riccardo", "last": "Barbati" }, { "first": "Valerio", "last": "Zacà" }, { "first": "Piercarlo", "last": "Ballo" }, { "first": "Eustachio", "last": "Agricola" }, { "first": "Francesco", "last": "Guerrini" } ] }
PMC1083418	Background ========== U.S. population-based surveys estimate that adults spend 87% of their day indoors, 69% in their place of residence, and 6% in a vehicle \[[@B1]-[@B3]\]. To date, most published disease cluster investigations use static geographies in which individuals are assumed to be sessile. Examples include the use of geocoded place of residence at time of diagnosis, death, and at time of birth (e.g. \[[@B4]\]), as well as the address of the admitting hospital (e.g. \[[@B5]\]) to record locations of health events, even though most researchers acknowledge that residential mobility should be accounted for, especially for diseases with long latencies such as cancer. In a recent review of standard methods for evaluating exposure/hazards, disease mapping and clustering techniques, Bayesian approaches, Markov Chain Monte Carlo (MCMC) and geostatistical methods, Mather et al. \[[@B6]\] identified as substantial weaknesses (1) the lack of temporal referencing of geospatial data and (2) the inability of methods to account for residential histories. A recent meeting of this nation\'s experts recognized the need to account for latency and human mobility as especially pressing in studies of cancer \[[@B7]\]. Boscoe et al. \[[@B8]\] identified residential history information as a primary need for the analysis of cancer data. The representation of individuals as sessile (immobile) rather than vagile (mobile) in part is due to the static world view of GIS software, which is not well suited to representing temporal change \[[@B9],[@B10]\]. Recently, technological advances have resulted in Space Time Intelligence Systems (e.g. \[[@B11]-[@B13]\]) that implement several constructs from Geographic Information Science for representing human mobility (see \[[@B14]\] for a review). The methods presented in this paper build on this body of prior work to produce case-control cluster statistics for residential histories. We begin with a brief background on tests for disease clustering, followed by a summary of approaches to modeling human mobility. We then develop a suite of novel tests for evaluating local, global and focused clustering in residential histories using case-control data. Finally, we illustrate several of the new techniques by quantifying local, global and focused clustering of residential histories in a case-control study of bladder cancer in Michigan. Background on cluster tests --------------------------- Cluster tests work within a hypothesis testing framework that proceeds by calculating a statistic (e.g. clustering metric) to quantify a relevant aspect of spatial pattern in a health outcome (e.g. case/control location, disease incidence, or mortality rate). The numerical value of this statistic is then compared to the distribution of that statistic\'s value under a null spatial model, providing a probabilistic assessment of how unlikely an observed cluster statistic is under the null hypothesis \[[@B15]\]. Waller and Jacquez \[[@B16]\] formalized this approach by identifying five components of a spatial cluster test. The test statistic quantifies a relevant aspect of spatial pattern (e.g. Moran\'s I). The alternative hypothesis describes the spatial pattern that the test is designed to detect. This may be a specific alternative, such as a circular cluster for the scan statistic, or it may be the omnibus \"not the null hypothesis\". The null hypothesis describes the spatial pattern expected when the alternative hypothesis is false (e.g. uniform cancer risk). The null spatial model is a mechanism for generating the reference distribution. This may be based on distribution theory, or it may use randomization (e.g. Monte Carlo) techniques. Most disease cluster tests employ heterogeneous Poisson and Bernoulli models for specifying null hypotheses \[[@B17]\]. The reference distribution is the distribution of the test statistic when the null hypothesis is true. Comparison of the test statistic to the reference distribution allows calculation of the probability of observing that value of the test statistic under the null hypothesis of no clustering. This five-component mechanism underpins most commonly used clustering methods. There are dozens of cluster statistics (see \[[@B17]-[@B19]\] for reviews) that may be categorized for convenience as global, local, and focused tests. Global cluster statistics are sensitive to spatial clustering, or departures from the null hypothesis, that occur anywhere in the study area. Many early tests for spatial pattern, such as Moran\'s I\[[@B20]\] are global tests. While global statistics can determine whether spatial structure (e.g. clustering, autocorrelation, uniformity) exists, they do not identify where the clusters are, nor do they quantify how spatial dependency varies from one place to another. Local statistics such as Local Indicators of Spatial Autocorrelation (LISA) \[[@B21]\] quantify spatial autocorrelation and clustering within the small areas that together comprise the study geography. Local statistics quantify spatial dependency (e.g. not significantly different from the null expectation, cluster of high values, cluster of low values, and high or low spatial outlier) in a given locality. Many local statistics have global counterparts that often are calculated as functions of local statistics. For example, Moran\'s Iis the sum of the scaled local Moran statistics. Focused statistics quantify clustering around a specific location or focus. These tests are particularly useful for exploring possible clusters of disease near potential sources of environmental pollutants. For example, Lawson and Waller \[[@B22],[@B23]\] proposed tests that score each area for the difference between observed and expected disease counts, weighted by exposure to the focus (also see \[[@B24]\] for a review of these approaches). A commonly used exposure function is inverse distance to the focus (1/d). The null hypothesis is no clustering relative to the focus, with expected number of cases calculated as the Poisson expectation using the population at risk in each area and the assumption that risk is uniform over the study area. Hundreds of cluster investigations are recorded in the literature, and several of these have resulted in cancer control activities such as epidemiological studies to understand potential causes. Notable examples of cluster studies include brain cancer \[[@B25]\], liver cancer \[[@B26]\], breast cancer \[[@B27],[@B28]\], prostate cancer \[[@B29]\], colorectal cancer \[[@B30]\], and cancer disparities \[[@B31]\], to name only a few. In studies of lung, breast and colorectal cancer on Cape Cod, stronger evidence for spatial clustering was found once latency was taken into account \[[@B32]\]. In a population-based case-control study Vieira et al. \[[@B33]\] incorporated residential location to evaluate lung cancer risks not explained by age and smoking. Han et al. \[[@B34]\] explored geographic clustering of breast cancer based on place of residence early in life and found space-time clustering in case-control data. They also explored clustering of cases using place of residence at critical time points including at the subject\'s birth, menarche, and at the women\'s first birth. Boscoe et al. \[[@B8]\] recognized representation of residential mobility as a primary need for data used in studies of cancer. But to date and to our knowledge, residential mobility has yet to be directly accounted for in cluster studies. How might one account for residential mobility in cluster studies? Hagerstrand \[[@B35]\] conceptualized the space time pathas an individual\'s continuous physical movement through space and time, and visually represented this as a 3-dimensional graph. Hornsby and Egenhofer \[[@B36]\] recognized that space-time paths mediate individual-level exposure to pathogens and environmental toxins, and that practical application would require a mechanism for representing location uncertainty. A space time prismrefers to the possible locations an individual could feasibly pass through in a specific time interval, given knowledge of their actual locations in the times bracketing that interval. The potential path area\[[@B37]\] shows the locations the individual could occupy given these constraints, and represents places where exposure events might occur. These constructs enabled new research approaches in diverse fields such as student life \[[@B38]\], sports analysis \[[@B39]\], social systems \[[@B40]\], transportation \[[@B37]\], and the analysis of disparities in gender accessibility in households \[[@B41]\]. While these approaches provide a proven mechanism for modeling geospatial lifelines and related constructs, to date and to our knowledge there are no methods for the statistical evaluation of clustering among such lifelines other than the paper by Sinha and Mark \[[@B42]\], who use Minkowski-type metrics to calculate a dissimilarity metric for geospatial lifelines, and then cluster this dissimilarity metric. This paper proposes a novel technique for undertaking the statistical evaluation of clustering of residential histories for case-control data. We first develop the method, and then apply it to an ongoing case-control study of bladder cancer in southeastern Michigan. Setting the Problem ------------------- A naïve approach when considering residential histories is to take an existing test for spatial clustering, and to then apply it repeatedly for different time values. For example, when considering the geographic distribution of bladder cancer, one might use place of residence of individuals in a case-control study from Tyears ago, and then allow Tto vary in a range of several decades. Locations of place of residence will change, as may the numbers of cases and controls extant in the study area. How might results vary depending on when one looks at the system (e.g. on selection of T)? To answer this question we analyzed data from a population-based bladder cancer case-control study currently underway in southeastern Michigan. Cases are recruited from the Michigan State Cancer Registry and diagnosed in the years 2000--2004. Controls are frequency matched to cases by age (± 5 years), race, and gender, and recruited using a random digit dialing procedure from an age-weighted list. This data set is described more fully later in this paper, and is comprised of 63 cases and 182 controls. Using Cuzick and Edwards T~k~statistic with k= 5 nearest neighbors we then analyzed these data at every point in time when the topology of place of residence of the cases and controls changed by having a participant move, enter or leave the study area. The graph of T~k~through time (Figure [1](#F1){ref-type="fig"}) is ascending, reflecting the larger number of cases and controls residing in the study area in later time periods. We found five periods when cases were significantly clustered relative to the controls: January 1 1929 through January 1 1935, January 1 1941 through November 26 1942, January 1 1960 through January 1 1961, August 22 1967 through January 1 1975 and January 1 1995 through January 1 1997. Clearly, results of cluster analyses that rely on single locations may be highly sensitive to the choice of the time at which the analysis is conducted. What are needed are new methods that account for the dynamic topology of cases and controls that arise as a consequence of residential mobility, and that are suited to multi-temporal analyses. The development of such techniques is the focus of this paper. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Cuzick and Edward\'s statistic through time. Graph of Cuzick and Edward\'s T~k~statistic (top) and its Probability (bottom) through time for k= 5. Shown in red are those time intervals in which the probability of T~k~was 0.0l or smaller. ::: ![](1476-069X-4-4-1) ::: Methods ======= We begin by defining an algebra for residential histories, and a matrix representation that describes how spatial nearest neighbor relationships change through time. Next we develop a local case control cluster test, and then extend it to create global, local and focused tests at specific time points, and then for entire residential histories. After completing development of the cluster statistics for residential histories, we next describe exposure traces that account for latency periods and exposure windows. We then develop clustering methods for exposure traces. After that we describe the bladder cancer data set that was analyzed with the new methods. In the Results section we describe application of several of the new cluster tests to evaluate possible clustering of residential histories of cases of bladder cancer in Michigan. Notation -------- Define the coordinate u~i,t~= {x~i,t~, y~i,t~} to indicate the geographic location of the place of residence of the i^th^case or control at time t. Residential histories for individual cases and controls can then be represented as the set of space-time locations: L~i~= (u~i0~, u~i1~,\..., u~iT~) (Equation 1) This defines individual iliving at his or her place of residence found at u~i0~at the beginning of the study (time 0), and moving to location u~i1~at time t= 1. At the end of the study individual imay be found at u~iT~. Tis defined to be the number of unique observation times on all individuals in the study. This bears some emphasis as understanding of how Tis recorded is essential in order to understand the cluster tests for residential histories. In other words, Tis the total number of different observation times across all individuals, and so one might expect several geographic locations in an individual residential history to be the same. For example, suppose we have 2 individuals (iand j) and record their residential histories (Figure [2](#F2){ref-type="fig"}). We record their places of residence at t= 0, the beginning of the study. At some time t= 1 \"i\" moves to a different home, and moves again at time t= 2. \"j\" never moves at all and hence has the location of the same initial place of residence recorded at times t= 0, 1, and 2. In this example T= 2. Notice the duration between t= 0 to t= 1 may not equal the duration from t= 1 to t= 2. This will be important later when we develop duration-weighted versions of the statistics. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Schematic of residential histories. Graphical representation of residential histories from Equation 1 using the instantaneous displacement movement model. Location is on the x-axis, time on the y-axis. Individual imoves from location u~i0~to u~i1~at time t= 1, and stays at that place of residence until t= T. Individual jstays at the same place of residence from t= 0 to t= T. ::: ![](1476-069X-4-4-2) ::: While observations on residential histories occur at a finite number of time points or observation times, these observations do not have to happen at the same time for all individuals under scrutiny. When residential histories are self-reported, these observation times are defined by the \"move\" dates reported by the respondent. We modeled this as an instantaneous displacement from the spatial coordinates for entity iat time t(u~it~) to those at time t+1 (u~it+1~). We defined this instantaneous displacement as occurring at time t+1. We viewed this as an observational model in which the entity is assumed to reside at its known location up until that moment when it is observed elsewhere (e.g. Figure [2](#F2){ref-type="fig"}). Individual residential histories can be associated with time-dependent attributes such as weight, height, disease state, smoking status, case control status, and so on. These attributes may be associated with risk and thereby influence calculation of the latency period and exposure windows defined later. Later we also will use time of diagnosis to define exposure windows during which carcinogenesis was thought to have occurred. For now let us define a case-control identifier, c~i~to be ![](1476-069X-4-4-i1.gif) Define n~a~to be the number of cases and n~b~to be the number of controls. The total number of individuals in the study is then N= n~a~+n~b~. Nearest Neighbor Relationships ------------------------------ Let kindicate the number of nearest neighbors to consider when evaluating nearest neighbor relationships (e.g. \[[@B63]\]), and define a nearest neighbor indicator to be: ![](1476-069X-4-4-i2.gif) We then define a binary matrix of k^th^nearest neighbor relationships at a given time tas: ![](1476-069X-4-4-i3.gif) By convention we define η~i,i,k,t~= 0 (the diagonal elements) since we do not wish to count individuals as nearest neighbors of themselves. This matrix enumerates the knearest neighbors (indicated by a 1) for each of the Nindividuals. The entries of this matrix are 1 (indicating that jis a knearest neighbor of iat time t) or 0 (indicating jis not a knearest neighbor of iat time t). It may be asymmetric about the 0 diagonal since nearest neighbor relationships are not necessarily reflexive (e.g. Imagine 3 people, call them A, B and C, standing in a line. B is in the middle but is closer to person A than to person C. The nearest neighbor to C is B, but the nearest neighbor to B is A. The nearest neighbor relationships are not reflexive). Since two individuals cannot occupy the same location, we assume at any time tthat any individual has kunique k-nearest neighbors. (While it is true that two individuals cannot occupy the exact same location, such as the space occupied by one individuals body, residential history information can assign two individuals the same coordinate when they live in the same house. How might tied nearest neighbor relationships arising from this situation be resolved? Two approaches have been proposed. The first creates fractional nearest neighbor weights \[[@B43]\], the second propagates uncertainty in the nearest neighbor relationships by evaluating the permutations of possible nearest neighbors for the tied nearest neighbor relationships \[[@B44]\]). The row sums thus are equal to k(η~i,•,k,t~= k) although the column sums vary depending on the spatial distribution of case control locations at time t. The sum of all the elements in the matrix is Nk. There exists a 1 × T+1 vector of times denoting those instants in time when either (1) the system is observed and the locations of the entities are recorded, or (2) under continuous observation at least one entity changes geographic location. We can then consider the sequence of Tnearest neighbor matrices given by ![](1476-069X-4-4-i4.gif) This defines the sequence of knearest neighbor matrices for each unique temporal observation recorded in the data set, and thus quantifies how nearest neighbor relationships change through time. This demonstrates one way in which spatial weights (here the nearest neighbor relationship) can be specified from residential histories. We will now use these nearest neighbor relationships to construct case control spatial and space-time cluster tests for residential histories. Spatially and Temporally Local Spatial Cluster Statistic -------------------------------------------------------- A spatially and temporally local case-control cluster statistic is: ![](1476-069X-4-4-i5.gif) This is the count, at time t, of the number of knearest neighbors of case ithat are cases, and not controls (assuming iindeed is a case, if it isn\'t Q~i,k,t~= 0). Since a given individual imay have kunique nearest neighbors, this statistic is in the range 0..k. It always is 0 when iis a control. When iis a case, low values indicate cluster avoidance (e.g. a case surrounded by controls), and large values (near k) indicate a cluster of cases. When Q~i,k,t~= k, at time tall of the knearest neighbors of case iare cases. Probabilities, Null Hypotheses and Randomization ------------------------------------------------ The statistical significance of Q~i,k,t~may be evaluated using conditional randomization that holds the case control identifier for individual ifixed and then allocates the vector of remaining N-1 case-control identifiers across the remaining individuals with a given probability function. If we assume equiprobability such that all individuals have equal disease risk we obtain: ![](1476-069X-4-4-i6.gif) Given the case-control identifier for individual i, this is the probability of individual jbeing a case under Goovaert and Jacquez\'s \[[@B45]\] neutral model Type IV (H~IV~) of spatial independence of risk for a spatially heterogeneous population density. As expressed in Equation 7, the exact number of cases (n~a~) and controls (n~b~) might not be reproduced under probabilistic sampling. Their neutral model type V retains a specified level of spatial autocorrelation and may be simulated using rejection sampling, sequential indicator simulation, or conditional case-control index swapping to achieve the observed level of spatial autocorrelation \[[@B46]\]. Probabilities for neutral model type V are difficult to write in a closed form analogous to Equation 7. Probabilities for neutral model type H~VI~describe the situation where not all individuals have the same probability of being labeled a case. This occurs, for example, when we are concerned with detecting clusters that arise from additional risk above and beyondthat of a background risk that is itself spatially heterogeneous. This may be accomplished in a variety of fashions to model known individual and environmental risk factors. Tests of the significance of Q~i,k,t~are then identifying clusters of cases above and beyond that expected under the neutral model. One calculates the value of the test statistic for each realization of the spatial distribution of cases generated under the chosen neutral model. These values under randomization are retained and used to construct the reference distribution of the statistic under the corresponding null hypothesis. The observed value of the test statistic for the not randomized data (denoted ![](1476-069X-4-4-i7.gif)) is then compared to the reference distribution to calculate the p-value: ![](1476-069X-4-4-i8.gif) Here ais the number of conditional randomizations whose cluster statistic was greater than or equal to that observed for the not randomized data, and bis the total number of randomization runs conducted. A convenient algorithm for conditional randomization under neutral model IV is to hold the case-control identifier for the i^th^individual constant, and to then draw from the 1 × N-1 vector of remaining case-control identifiers new case-control identifiers for the knearest neighbors surrounding i. This sampling is accomplished without replacement. Alternatively, one could populate the k-nearest neighbors about iusing the probabilities from Equation 7. This equation is correct for the first identifier so drawn, but needs to be adjusted for the second, third and so on. For the m^th^identifier the correct probability for sampling without replacement is: ![](1476-069X-4-4-i9.gif) If one assumes sampling with replacement, so that the cases and controls are assumed drawn from a larger population, one can use Equation 7 without modification. This approach does not work for neutral models type V and VI, since spatial structure in the background risk is lost. Instead one calculates the value of the test statistic for each of the N locations, for each realization of the spatial neutral model (of type V or VI) that produces a spatial point pattern of cases and controls with the desired level of spatial autocorrelation. The probability assigned to clusters from these tests (as given by Equation 8) then accounts for the specified background variation in disease risk. Note for each of the approaches listed above, that a reference distribution, test statistic, and corresponding p-value, may be calculated for each of the n~a~case locations. Simes Correction for Local Dependency ------------------------------------- The P-values for the kindividuals surrounding the i^th^case are not independent of one another, as they include one another as their own knearest neighbors. We therefore employ a modified Simes correction \[[@B47]\] to account for the lack of spatial independence in the local Q statistics. The Simes adjustment is calculated as p~i~\' = (k+ 1 - a) p~i~. Here kis the number of p-values being considered (the number of neighbors), and ais the index (starting at 1) indicating the rank in the sorted vector of p values for individual iand its neighbors. We employ this correction later when reporting p-values for the local Q-statistics. Global Test for Spatial Clustering at Time t ---------------------------------------------- A global statistic for spatial clustering at time tmay be constructed as: ![](1476-069X-4-4-i10.gif) This is the time-referenced form of Cuzick and Edward\'s \[[@B43]\] global test for case-control clustering used in Figure [1](#F1){ref-type="fig"}. It is the count, over all cases, of the number of cases that are k-nearest neighbors to those cases at time t. One could divide this statistic, and others to follow, by n~a~to facilitate their interpretation. The test statistic would then be an average number of neighbor cases per case instead of the integer total number of cases, and would facilitate comparison across different studies with different numbers of cases. In this paper we will use the case-count version. The probability of Q~k,t~under H~IV~is evaluated by allocating the case-control id\'s with equal probability over the Nlocations at time t. Q~k,t~is then calculated and this process is repeated btimes to construct the reference distribution and probability (Equation 8). Notice that since this is a global test conditional randomization that holds the case-control id for individual iconstant is not needed. Global Test for Spatial Clustering of Residential Histories ----------------------------------------------------------- A global test for spatial clustering among the Nresidential histories as represented in Equation 1 is ![](1476-069X-4-4-i11.gif) This is the sum, over all T+1 time points, of the global statistic Q~k,t~. It is a measure of the persistence of global clustering and is large when case clustering persists through time. Its reference distribution may be constructed under a randomization procedure in which the case-control ids are allocated with equal probability over the residential histories comprising the set {L~i~, i= 1..N} (Equation 12) This randomization procedure is conditioned on the total number of cases and controls in the data set, so that each data set constructed under randomization has the same number of cases and controls as the original data. Local Test for Spatial Clustering of Residential Histories through Time ----------------------------------------------------------------------- To determine whether cases tend to cluster through time around a specific case we may construct a test statistic: ![](1476-069X-4-4-i12.gif) For the i^th^residential history, this is the sum, over all T+1 time points, of the local spatial cluster statistic Q~i,k,t~. It is the number of cases that are k-nearest neighbors of the i^th^residential history (a case), summed over all T+1 time points. It will be large when cases tend to cluster around the i^th^case through time. Under neutral model type IV, the significance of Q~i,k,t~is evaluated under a conditional randomization that holds the case id for iconstant, and then allocates the remaining case-control id\'s at random over the N-1 remaining residential histories. This statistic is useful for determining whether there is local clustering of residential histories about a specific case. The statistic can be calculated for all cases in the data set to identify those cases whose residential histories form local spatial clusters. However, when calculating significance one should correct for the multiple testing inherent when many spatial locations are evaluated. Focused Test for Spatial Clustering at Time t ----------------------------------------------- Suppose that one suspects that the cases may be clustering about a specific focus defined by the lifeline (e.g. record of business addresses): L~F~= {u~F,0~, u~F,1~,.., u~F,T~} (Equation 14) This records the locations of the focus as it moves about through space-time, and includes situations in which the focus doesn\'t move as a degenerate instance. A test for spatial clustering of cases about the focus at a given time tis then: ![](1476-069X-4-4-i13.gif) Here η~F,j,k,t~is the nearest neighbor index indicating at time twhether the j^th^individual is a k^th^nearest neighbor of the geographic location of the focus defined by u~F,t~. The statistic Q~F,k,t~is then the count of the number of k-nearest neighbors about the focus at time tthat are cases. Under null hypothesis type IV randomization at time tmay be accomplished by allocating the case control identifiers with equal probability over the N-individuals. Since only the k-nearest neighbors are considered it is only necessary to allocate their indices. This may be accomplished by sampling without replacement from the 1 × Nvector of the case-control identifiers, or by drawing the krequired case control identifiers with probabilities defined by Equation 9 (for sampling without replacement) or Equation 7 (for sampling with replacement). Focused Test for Spatial Clustering of Residential Histories about a Mobile Focus --------------------------------------------------------------------------------- A test for focused clustering of residential histories through time is: ![](1476-069X-4-4-i14.gif) This is the count, over the Ttimes, of the number of cases that are knearest neighbors of the focus at each time point. This statistic is large when residential histories that are near the focus are cases. Its maximal value is max(Q~F,k~) = kT. (Equation 17) One drawback of using nearest neighbor relationships for focused tests is that the set of nearest neighbors to the focus are given equal weight in Equations 15 and 16, regardless of their actual geographic distance and direction with respect to the focus. But diffusion and active transport mechanisms that might carry emissions from the focus typically result in higher exposures near the focus, and it thus may make sense to use a maximum distance within which k~i~nearest neighbors are found. In these instances the set of nearest neighbors to the focus will vary (hence the isubscript denoting the i^th^focus) depending on the number of cases and controls found within the specified distance of the focus. Power of the Focused Tests and Specification of the Exposure ------------------------------------------------------------ Notice that the power of the tests given by Equations 15 and 16 decreases as kapproaches Nsince Q~F,k,t~= n~a~when k= N, and its probability is then: P(Q~F,k,t~\\| H~0~, k= N) = 1.0. (Equation 18) When one wishes to search for clustering in instances where kapproaches Npower may be retained by constructing a weight function to model the hypothesized exposure. For geographically localized foci this may be based on proximity to the focus. One choice is ![](1476-069X-4-4-i15.gif) Here r~F,j,t~is the rank indicating proximity of the location of the j^th^individual at time t(as given by u~j,t~) to the location of the focus at time t(u~F,t~). For example, the first nearest neighbor to the focus has rank 1, the second rank 2, and so on. In many situations, such as airborne pollution or groundwater contamination, the magnitude of exposure is a function not only of the proximity to the focus but also of its orientation, since most dispersion processes (i.e. winds, infiltration through porous media) are anisotropic or direction-dependent. Depending on the amount of information available, exposure models of increasing complexity can be built. An easy way to account for anisotropy is to replace the rank value r~F,j,t~by a function of the separation vector h~jF,t~= \\|u~j,t~- u~F,t~\\| joining the location of the jth individual at time tto the location of the focus at time t. Covariance functions seem to be natural choices for the weight functions w~F,j,t~since they incorporate the spatial pattern of dependence of exposure data. For example, one could use the Exponential or Gaussian covariance functions defined as: ![](1476-069X-4-4-i16.gif) ![](1476-069X-4-4-i17.gif) where a(θ) is the practical range of autocorrelation of the covariance models; that is the distance hat which the covariance function equals 0.05. This range is a function of the azimuth of the separation vector h~jF,t~. For example, the range of exposure to an airborne contaminant is expected to be larger in the direction of the prevailing winds. More complex weight functions could be created if a process-based model of dispersion is available. For the example of airborne pollution, an atmospheric dispersion and deposition model could be developed to predict the fate of emissions and dust plumes from targeted facilities \[[@B48]\]. However, such models require many more parameters and assumptions concerning, for example, the emission rate, the meteorological conditions, complex terrain effects, the particle size and density for deposition calculation. A limitation of process-based models is that they fail to provide a measure of uncertainty attached to their predictions and field exposure data are not readily incorporated. Geostatistics \[[@B49]\] provide tools for modeling the spatio-temporal distributions of exposure and assessing the attached uncertainty. Various sources of information can be taken into account, such as measurements at a few monitoring stations, coordinates of major sources of exposure (i.e. factories) and transport characteristics (i.e. wind directions) that could be either directly incorporated into the prediction algorithm \[[@B50]\] or fed into physical models to derive spatial trends \[[@B51]\]. In the latter case, geostatistics are used to model the unexplained or residual part of the variability predicted by the process-based models. The weight function, either based on geographic proximity (as in Equation 19) or derived using a process-based model or geostatistics (as in Equations 20 & 21), is then used to construct the weighted focused test at time tas: ![](1476-069X-4-4-i18.gif) The test for spatial clustering of residential histories about the focus through time is then: ![](1476-069X-4-4-i19.gif) Notice these weighted tests are conducted for the k nearest neighbors being considered. When k= Nthe maximum values are: ![](1476-069X-4-4-i20.gif) Duration-Weighted Tests for Clustering of Residential Histories --------------------------------------------------------------- The number of time points defined by the t= 0 ..Tobservation times, and the frequency with which they are taken, can have some influence on the value of the above statistics. For example, many repeated observations when there is a chance of clustering could lead to spurious significance for the local and global tests for clustering of residential histories. We therefore developed duration-weighted versions of the tests, and these are presented in the Appendix \[see [Additional file 1](#S1){ref-type="supplementary-material"}\]. Accounting for Exposure Windows and Latency Periods --------------------------------------------------- When dealing with cancers, causative exposures may occur during an exposure window (Δ~E~), followed by a latency period (Δ~L~) before cancer is manifested and diagnosed. Given the residential history for case i, L~i~, further denote the space-time coordinate representing place of residence at time of diagnosis as ![](1476-069X-4-4-i21.gif), noting that ![](1476-069X-4-4-i21.gif) ∈ L~i~We can then define that subset of the residential history L~i~over which the exposure window occurred as: ![](1476-069X-4-4-i22.gif) Here t~i,D~is the time of diagnosis for individual i. The term (t~i,D~- Δ~L~) indicates the time prior to diagnosis when the latency period began and (t~i,D~- Δ~L~- Δ~E~) is the time when the causative exposure began. Hence equation 25 denotes that portion of individual i\'s residential history where causative exposures could have occurred. Notice that both the exposure window and latency period could be covariate-adjusted to account for risk factors such as smoking and age (see Discussion). In this instance the latency period and exposure window vary from one individual to another and we write: ![](1476-069X-4-4-i23.gif) Here Δ~i,L~and Δ~i,E~are the latency period and exposure windows for the i^th^individual. In either case (Equations 25 or 26) we call ![](1476-069X-4-4-i24.gif) the exposure tracefor the i^th^individual. Randomization Procedures for Exposure Traces -------------------------------------------- In order to evaluate whether exposure traces of the cases cluster we must first construct a randomization procedure for generating representative times of diagnosis, latency periods, and exposure windows. Once this is accomplished we will be able to determine whether the exposure traces for the cases cluster relative to those so constructed for the controls. For a case, the exposure trace is defined by the time of diagnosis and the latency period, with the latency period potentially dependent on age, gender and other covariates. The procedure proceeds as follows: \(1) Since controls are matched to cases, the \"time of diagnosis\" for each control is set equal to the time of diagnosis for the matched case. \(2) The exposure window and latency period for each control is then defined based on the covariates for each control as was accomplished for that controls matched case. \(3) Completion of steps (1) and (2) will result in exposure traces defined for both cases and controls. Now randomly assign case control identifiers across the residential histories with equiprobability conditioned on the total number of cases and the total number of controls. \(4) Calculate the desired test statistic for clustering of exposure traces. \(5) Repeat steps 3 and 4 a desired number of times to construct the reference distribution of the statistic under randomization. Test statistics for assessing clustering of exposure traces are presented below. Local Case-Control Test for the Spatial Clustering of Exposure Traces at Time t --------------------------------------------------------------------------------- When health events such as cancers are caused by exposure to geographically localized factors we might expect the exposure traces for the cases to cluster relative to the exposure traces that are generated for the controls. The durations of the exposure traces may vary, and we therefore will employ duration-weighted statistics. We would like to know whether exposure traces for the cases exhibit spatial clustering relative to the controls both locally (to identify places where causative exposures occurred) and globally (to ascertain whether the exposure traces for the cases cluster when considered as a group). We also might wish to ask whether exposure traces for the cases exhibit focused clustering. The exposure trace for case i(![](1476-069X-4-4-i24.gif)) records those places where that individual lived during that time when exposures occurred that might have caused cancer later in life. Now define an indicator, e~i,t~, as: ![](1476-069X-4-4-i25.gif) When e~i,t~is 1, let us say the exposure trace is \"active\". A local case-control test for spatial clustering of exposure traces at time tis then: ![](1476-069X-4-4-i26.gif) This is the count, at time t, of the number of knearest neighbors of case i\'s active exposure trace that are cases (and not controls) whose exposure traces also are active. Hence the statistic will be large at those times when exposure traces of a group of cases are active and cluster. Its value is 0 when individual iis a control, and also when individual iis a case with an inactive exposure trace. The duration weighted version of this statistic is: ![](1476-069X-4-4-i27.gif) Local Case-Control Test for the Spatial Clustering of Exposure Traces through Time ---------------------------------------------------------------------------------- We can explore whether active exposure traces of cases tend to cluster spatially through time. A statistic sensitive to this pattern is: ![](1476-069X-4-4-i28.gif) ![](1476-069X-4-4-i29.gif) will tend to be large when active exposure traces for cases tend to cluster around the active exposure trace of the i^th^case. It will be 0 when iis a control, and small when a given case ihas the traces of many controls as its neighbors. The duration-based version of this statistic is: ![](1476-069X-4-4-i30.gif) This statistic will be expressed in case-time units, indicating the number (for example) of case-days over the entire study period for which cases with active traces were k-nearest neighbors of the active trace of case i. Global Case-Control Test for the Spatial Clustering of Exposure Traces at Time t ---------------------------------------------------------------------------------- We can ask whether, as a group, active case traces are spatially clustered relative to the active traces of the controls at a given time t. This is accomplished using the statistic: ![](1476-069X-4-4-i31.gif) This is simply the sum, over all cases, of the local statistic for clustering of case exposure traces at time t. This statistic will be large when active traces of cases tend to be near one another and small when the active traces of cases tend to have controls as their knearest neighbors. The duration-based version is: ![](1476-069X-4-4-i32.gif) Global Case-Control Test for the Spatial Clustering of Exposure Traces through Time ----------------------------------------------------------------------------------- A global test for the spatial clustering of the active exposure traces of cases through time is: ![](1476-069X-4-4-i33.gif) This is the sum, over all time periods, of the global cluster test for the clustering of exposure traces. It will be large when global clustering of active exposure traces tends to persist through time. The duration-based version of this statistic is: ![](1476-069X-4-4-i34.gif) Focused Case-Control Test for the Spatial Clustering of Exposure Traces at Time t ----------------------------------------------------------------------------------- We can also ask whether the exposure traces of cases cluster near putative emission sources. Again, these sources may be mobile, and we accomplish this by assigning larger weights for those cases that are near the focus. Recall from Equation 14 that we can represent a mobile source as L~F~= {u~F,0~, u~F,1~,.., u~F,T~}. The test for spatial clustering of cases about a focus at a given time t(Equation 15) may then be extended to be a focused test for clustering of exposure traces as: ![](1476-069X-4-4-i35.gif) This is the count of the number of cases with active exposure traces that are knearest neighbors of the focus at time t. Significance of this statistic may be evaluated by constructing exposure traces for the controls as described earlier, and by then repeatedly allocating case-control identifiers across the Nlifelines that are knearest neighbors of the focus in order to construct the reference distribution for ![](1476-069X-4-4-i36.gif). The duration weighted version of this statistic is ![](1476-069X-4-4-i37.gif) Focused Test for Spatial Clustering of Exposure Traces about a Mobile Focus through Time ---------------------------------------------------------------------------------------- We can evaluate whether there is statistically significant clustering of exposure traces of cases about a mobile focus through time using the statistic: ![](1476-069X-4-4-i38.gif) This is the count, over T+1 times, of the number of cases that have active exposure traces that are knearest neighbors of the focus at each time point. The maximum value of this statistic is kT, and its significance may be evaluated under randomization by reallocating case-control identities over the exposure traces of the cases and controls as described in the previous section. The duration-weighted version of this statistic is: ![](1476-069X-4-4-i39.gif) Weighted Focused tests for Exposure Traces ------------------------------------------ The power of the k-nearest neighbor based focused test for exposure traces decreases as kapproaches N. Weights such as that suggested in Equations 19--21 may be used to construct a weighted focused test for exposure traces at a given time t: ![](1476-069X-4-4-i40.gif) The test for focused clustering of exposure traces through time is then: ![](1476-069X-4-4-i41.gif) The significance of these statistics is evaluated using randomization across the knearest neighbors of the focus as described earlier. The corresponding duration-weighted versions are ![](1476-069X-4-4-i42.gif) This is the weighted focused test over duration ω~t~. The duration-based weighted focused test for exposure traces through time is ![](1476-069X-4-4-i43.gif) Bladder Cancer in southeastern Michigan --------------------------------------- A population-based bladder cancer case-control study is currently underway in southeastern Michigan. Cases are recruited from the Michigan State Cancer Registry and diagnosed in the years 2000--2004. Controls are frequency matched to cases by age (± 5 years), race, and gender, and recruited using a random digit dialing procedure from an age-weighted list. To be eligible for inclusion in the study, participants must have lived in the eleven county study area for at least the past 5 years and had no prior history of cancer (with the exception of non-melanoma skin cancer). Participants are offered a modest financial incentive and research is approved by the University of Michigan IRB-Health Committee. The data presented here are from 63 cases and 182 controls. As part of the study, participants complete a written questionnaire describing their residential mobility history. The duration of residence and exact street address were obtained, otherwise the closest cross streets were provided. Each residence in the study area was geocoded and assigned a geographic coordinate in ArcGIS; residences outside the study area were not geocoded. Participants resided at 1004 homes within the study area, with time spent averaging 64% of their lifetimes. Residences within the study area were successfully geocoded: 76% automatically matched using ArcGIS settings of spelling sensitivity equal to 75, minimum candidate score equal to 10, and a minimum match score equal to 60. The unmatched addresses were manually matched using cross streets with the assistance of internet mapping services (15%). If cross streets were not provided, best informed guess placed the address on the road (5%), and as a last resort, residence was matched to town centroid (4%). Industrial histories have also been collected for the study area, and will be explored to explain local clustering. Industries reported to or believed to emit contaminants that have been associated with bladder cancer were identified using the Toxics Release Inventory \[[@B52]\] and the Directory of Michigan Manufacturers (Manufacturer Publishing Co., 1946, 1953, 1960, 1969, 1977, 1982). Standard Industrial Classification (SIC) codes were adopted, but prior to SIC coding, industrial classification titles were selected. Characteristics of 268 industries, including, but not limited to, fabric finishing, wood preserving, pulp mills, industrial organic chemical manufacturing, and paint, rubber, and leather manufacturing, were compiled into a database. Industries were geocoded following the same matching procedure as described for residences: 89% matched to the address, 5% were placed on the road using best informed guess, and as a last resort, 6% were matched to town centroid. Each industry was assigned a start year and end year, based on best available data. The data on these industries is used to demonstrate the focused versions of the Qstatistics. Results ======= At the time of this writing, geocoding and data collection are ongoing; hence the results reported in this manuscript are entirely preliminary and should not be used to draw any conclusions regarding the spatial patterns of bladder cancer in Michigan. The analysis undertaken in the manuscript is provided only as an example application of the new Qstatistics. To demonstrate the methods we implemented the local and global Qstatistics for clustering of residential histories, specifically the local test at time t, Q~i,k,t~(Equations 6), and its global counterpart Q~k,t~(Equation 10). We also implemented the local test for clustering of residential histories through time Q~i,k~(Equation 13), and the global test for clustering of residential histories Q~k~(Equation 11). We also were concerned with possible clustering of cases near the industrial facilities, and evaluated this using the focused test at time tQ~F,k,t~(Equation 15) as well as the focused test through time Q~F,k~(Equation 16). In addition we programmed the duration-weighted versions of these statistics, and for the focused tests we also employed exposure weights calculated using the inverse rank distance (Equation 19). Results for Q~kt~ ----------------- These techniques were implemented in TerraSeer\'s STIS software using the Application Programmer\'s Interface. This allowed us to create a methods dynamic linked library with our new techniques that we then invoked using an automatically generated dialog. Time animated maps of the places of residence of the cases and controls, and of the changing geography of the municipal water supplies, were constructed using STIS (Figure [3](#F3){ref-type="fig"} \[see [Additional file 2](#S2){ref-type="supplementary-material"}\] \[see [Additional file 3](#S3){ref-type="supplementary-material"}\]). These display the changing geography of the cases and controls as they move from one place to another, alterations in the geography of the municipal water supplies as they are founded, expand and merge, as well as township boundaries. To verify the methods we compared results using the Qstatistics to those obtained using Cuzick and Edward\'s test in the ClusterSeer software. Specifically, we used STIS to calculate the Q~kt~statistics through time and then exported the data for July 1, 1969. We choose this time point, because Q~kt~reached a local peak of Q~kt~= 77 that was statistically significant (see Figure [1](#F1){ref-type="fig"}). The Cuzick and Edward\'s test in ClusterSeer returned T~5~= 77, confirming the results from STIS. As noted earlier, Cuzick and Edward\'s test is a special case of the Q-statistic for the global test at time t, Q~kt~. Note that Q~kt~is calculated as the sum of the local Qstatistics at time t, Q~ikt~, and thereby provides verification that the statistic Q~ikt~, from which the family of Qstatistics is derived, is being calculated correctly. We must remind the reader that these results are highly preliminary and that data collection is incomplete. In fact, and as noted later in the Discussion, it is likely the observed clustering in these data is due to the geographic ordering in which the data are being collected. Nonetheless, this example demonstrates how plots of the Q~kt~statistics may be used to evaluate geographic case clustering of residential histories. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Still from the animation of residential histories of cases, controls and industry in southeastern Michigan ([additional file 2](#S2){ref-type="supplementary-material"}and[additional file 3](#S3){ref-type="supplementary-material"}). ::: ![](1476-069X-4-4-3) ::: Results for ![](1476-069X-4-4-i44.gif) -------------------------------------- The results reported above were not time standardized. We therefore undertook an analysis using the time-standardized version of Q~kt~called ![](1476-069X-4-4-i45.gif) as per Equation A4. This expresses the amount of clustering at a given time interval in cases per unit time period. STIS reports times down to the second, hence results are recorded in person seconds. Figure [4](#F4){ref-type="fig"} shows a similar overall increasing trend but also a greater variability in the value of the Q statistic through time. This is driven both by the increased number of cases through time and also by differences in the durations between movement events. When these sources of variability are accounted for we find episodic case clustering in approximately the same time intervals as found for the not time weighted statistic. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### ![](1476-069X-4-4-i44.gif)through time. Graph of ![](1476-069X-4-4-i44.gif) (top) and its probability (bottom) through time for k= 5. ![](1476-069X-4-4-i44.gif) is the time weighted version of Q~tki~and is expressed in case-seconds. ::: ![](1476-069X-4-4-4) ::: Results for Q~k~ -------------------- Having found some of the Q~kt~and ![](1476-069X-4-4-i45.gif) statistics to be statistically significant, the question then arises as to whether there is overall global clustering given the multiple time points evaluated. To accomplish this we used the global Q~k~test under the randomization procedure that holds the residential histories of the cases and controls as given and then allocates the n~a~case and n~b~control identifiers across the Nresidential histories. We accomplished this randomization 99 times in the STIS with a resulting p-value of 0.01, and concluded there was global clustering in the residential histories. Results for Q~i,k~to evaluate Clustering of Residential Histories ----------------------------------------------------------------------- The statistics Q~kt~and ![](1476-069X-4-4-i45.gif) are sensitive to a clustering of cases relative to the controls, and are evaluated at each of the T+1 time points in the set of residential histories. We also can ask, whether residential histories of the cases cluster near the residential histories of other cases by using the statistics Q~i,k~(Equation 13) and its duration-weighted version ![](1476-069X-4-4-i46.gif) (Equation A6). Since our analysis above demonstrated the results are not overly sensitive to duration weighting, we report results only for the not-weighted tests. This test will associate a statistic and a p-value with each residential history. A map of the residential histories on April 12, 1997 is shown in Figure [5](#F5){ref-type="fig"}. Note the two red dots that denote the place of residence of the two cases with statistically significant clustering of residential histories. Over the entire time span of the study, these two cases tend to be surrounded by residential histories of other cases, rather than the residential histories of controls. Because of residential mobility, the two red dots move about through time. This animation is quite compelling in the STIS and is approximated by the simpler animation in Figure [3](#F3){ref-type="fig"}. Note the animation in Figure [3](#F3){ref-type="fig"} is sampled from the complete animation created when running the STIS software. This is necessary to create .avi files of small enough size for effective posting on the internet. Periods in which a red dot disappears from the animation denote time periods when that individual moved out of the study area. It is important to note that we have not adjusted these local tests for the multiple testing in the many spatial locations that were evaluated. However, the global Q~k~statistic was statistically significant and the large local statistics observed for the two red dots reference the two residential histories that contributed the most to the global Q~k~. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Map of cases and controls on 4/12/1997. Map of cases and controls on 4/12/1997. Cases are shown as dots within a circle, controls are shown as crosses. The two cases whose residential histories tend to be surrounded by the residential histories of other cases are shown in red. ::: ![](1476-069X-4-4-5) ::: Focused clustering ------------------ To demonstrate the use of the focused versions of the Qstatistic we analyzed possible clustering of the residential histories of cases near the 268 industrial facilities that produced compounds thought to be putative carcinogens for bladder cancer. We undertook two sets of analyses using Q~F,k~(Equation 16). The first evaluated focused clustering of residential histories using k= 5 nearest neighbors. The second only considered those nearest neighbors within 1 kilometer of the focus. When considering the 5 nearest neighbors to each industry, 24 of the 268 industrial facilities had p-values less than 0.05. Thus under the null hypothesis that each person in the study had an equal probability of being labeled a case, these 24 candidate foci had a significant excess of cases among each of their five nearest neighbors, at least at the nominal 0.05 level. Notice that at the 0.05 level, we would have expected 13.4 foci to be significant under this null hypothesis. Using an experiment-wise error approach, and a 5% critical value, the adjusted alpha level of the test is 0.000187 using the Bonferonni correction, and is 0.000191 using Sidak\'s multiplicative inequality. Using 49,999 randomizations, we were able to resolve p-values as small as 0.00005. None of these industries proved to be statistically significant foci once multiple testing was accounted for. We also used the distance-based approach considering those neighbors within 4,000 m of each industrial facility. Under this approach, 10 industrial facilities had p-values \< 0.05, but none of these were significant once multiple testing was accounted for. Discussion ========== This paper presented a new approach for evaluating case control clustering of residential histories. To date and to our knowledge, almost all case control cluster tests rely on the static view, analyzing clustering at one point in time or independently at several points in time. By using the mathematical construct of a residential history in Equation 1, and the notion of super sets of proximity matrices (Equation 5) to represent the changing geometry of place of residence, we have derived local, global and focused tests that are realistic in the sense that they quantify human residential mobility. The results of the analysis of the bladder cancer data are entirely preliminary, and should not be interpreted to reach any inferences or conclusions regarding case-control clustering of bladder cancer in Michigan. At the time of this writing we believe statistically significant spatial clustering of cases is the result of a geographic pattern in the temporal order in which cases are reported. Because of recent implementation of the HIPPA (Health Insurance Portability and Accountability Act) legislation, The University of Michigan hospital systems had been unwilling to release case data until its official position on these requirements was completely formulated. As a result, bladder cancer cases that were treated at the University of Michigan hospitals are only now being recruited to the study\'s data set. Because many of these cases are from the surrounding environs of Washtenaw and Livingston counties, the data set analyzed in this paper has a deficit of cases in these areas. Selection of controls employs a population sample using random digit dialing, and appropriately represents the entire study area. As a result, there is a deficit of cases in Washtenaw and Livingston counties, and a concomitant clustering of cases in the balance of the study area. Further investigation of any bladder cancer clusters would also entail including known bladder cancer risk factors, such as cigarette smoking and occupational exposure history, in the analysis. We intend to revisit this analysis once the data set is complete. Selection of controls through random digit dialing can introduce bias, since not everyone is equally likely to be selected due to different numbers of phones and the likelihood of answering the phone. While such bias might be reduced by first selecting a census block group based on census numbers, adjusted for age and gender, and then doing random digit dialing within that block group, such a procedure has the potential of over-matching on exposure \[[@B53]\]. This would make it very difficult to detect any spatial pattern that arises at a spatial scale greater than the block group. In this study we chose not to match on geography because some of the exposures of interest display a geographic pattern and over-matching on exposure was a possibility. These exposures include regional patterns in arsenic concentration in drinking waters associated with surficial geology and regional differences in household water supply sources \[[@B54],[@B55]\]. Southeastern Michigan includes rural farming areas as well as portions of metropolitan Detroit, and differential response rates under random digit dialing are a concern. We attempted to ensure these areas do not have differential response rates by comparing addresses of responders and non-responders in age-weighted lists. Exposures in early life and over an individual\'s life course may be important risk factors for the onset of cancer \[[@B56],[@B57],[@B34]\], thereby impacting both the date of diagnosis and latency period. But how can such risk factors be accounted for in exposure trace analysis? We need to explicitly model the latency period to take into account not only exposures of direct interest (arsenic in our example) but also additional risk factors (such as smoking) that might decrease the latency period and accelerate disease onset. Many common epidemiological risk-disease measures (e.g. odds ratio) are concerned with whetheran exposure occurred, rather than with whenit occurred, and are thus of little use for estimating relationships between timing of exposure and disease onset \[[@B58]\]. For cohort analyses, Robins and Greenland\[[@B59]\] argued that, when conditioned on age, Years of Life Lost (YLL) due to early exposures cannot be estimated without bias in the absence of causal models for how exposure causes death. This result was demonstrated analytically by Morfeld \[[@B60]\], who developed a framework for causal thinking in epidemiology, and applied it to evaluate the estimability of YLL and related measures. Candidate causal modeling approaches cited by Morfeld include Robin\'s G-estimation procedure \[[@B61],[@B62]\] which can be used to estimate the time period between exposures and outcomes such as death, and thus appear promising for incorporating covariates into models of the latency period. Applications of G-estimation and the YLL procedures of Robins \[[@B61]\] in exposure trace modeling is thus an important future research direction. Discussion of the type of spatial metric to use (nearest neighbor, adjacency, or geographic distance-based) as well as the number of knearest neighbors to analyze is warranted. The approaches detailed in this paper are general in the sense that weights such as inverse distance and adjacency could be used in place of k-nearest neighbor relationships in Equation 4. We chose to work with nearest neighbor measures because we\'ve found them to be more powerful than adjacency- and distance-based measures in some situations (e.g. \[[@B63]\]). As noted earlier in this paper, we used k= 5 because we found in the past that spatial clustering under nearest neighbor methods often may be detected at that level of k. Such a justification is sufficient in analyses conducted purely for demonstration purposes but is deficient in applied settings. In practice, two approaches may be used, which we call a prioriand exploratory. When prior information is available on the scale of clustering this can be used to select a specific number of nearest neighbors to explore. Hence if one wishes to detect clusters of five individuals one might set k= 5. When such prior information is lacking an exploratory approach may be used in which several levels of kare analyzed, and probabilities from the analyses must then be adjusted to account for multiple testing \[[@B63]\]. We could not demonstrate each of the statistics developed in this paper, due to both data and space constraints. We note that exposure traces could be implemented to represent cases and controls of similar ages, in addition to those at a point in time. For example, a researcher may wish to determine whether cases cluster together when they were children, irrespective of year, thereby indicating early-lifetime vulnerability to an environmental exposure in the area. These clustering tools thus can be used to display cancer clusters of similarly-aged participants, as well as clusters based on the years a participant lived at a residence. In this manner, clusters of children can be investigated, whether they are born in the same generation or born in different generations. Conclusion ========== In conclusion, the methods presented in this paper account for residential mobility and are thus far more realistic than existing tests that are founded on static geographic representations. They thus are preferred over clustering methods that ignore human mobility. The techniques demonstrated in this paper have been programmed in a dynamic linked library that can be obtained from the first author and used in conjunction with a STIS. List of Abbreviations ===================== GIS: Geographic Information System HIPPA: Health Insurance Portability and Accountability Act H~IV~: Goovaert and Jacquez\'s \[[@B45]\] neutral model Type IV H~VI~: Goovaert and Jacquez\'s \[[@B45]\] neutral model Type IV IRB: Institutional Review Board LISA: Local Indicators of Spatial Autocorrelations MCMC: Markov Chain Monte Carlo STIS: Space-Time Intelligence System SIC: Standard Industrial Classification code YLL: Years of Life Lost Competing interests =================== Geoffrey Jacquez is President of BioMedware, the software company that is developing the STIS software. Authors\' contributions ======================= GJ derived the methods and drafted the majority of this manuscript. He also accomplished the analysis of the bladder cancer data set. AK programmed and tested the statistical methods in the STIS software. GA and JM provided data and wrote the data set description. PG wrote the sections on geostatistical weighting functions for the focused tests. JN is Principal Investigator on the R01 project that is collecting the bladder cancer data set. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 AppendixThis file contains the article\'s Appendix. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 2 Animation for Figure[3](#F3){ref-type="fig"}in QuickTimeThis is the animation for Figure [3](#F3){ref-type="fig"} in QuickTime format. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 3 Animation forFigure [3](#F3){ref-type="fig"}as an animated GIFThis is the animation for Figure [3](#F3){ref-type="fig"} in GIF format. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ This study was supported by grant R01 CA96002--10, Geographic-Based Research in Cancer Control and Epidemiology, from the National Cancer Institute. Development of the STIS software was funded by grants R43 ES10220 from the National Institutes of Environmental Health Sciences and R01 CA92669 from the National Cancer Institute. The critical comments and suggestions of Martin Kulldorff, Thomas Webster, and Al Ozonoff greatly improved the manuscript. Access to cancer case records was provided by Michigan Cancer Surveillance Program within the Division for Vital Records and Health Statistics, Michigan Department of Community Health. The authors thank the Michigan Public Health Institute for conducting the telephone interviews and Stacey Fedewa and Lisa Bailey for entering written surveys into a database. Thanks to Wanda Angelomatis and Fred Wallace who hosted the first author and his daughter for two weeks at Berkenhead Lake in British Columbia where these methods were originally formulated.	PubMed Central	2024-06-05T03:55:55.743645	2005-3-22	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1083418/", "journal": "Environ Health. 2005 Mar 22; 4:4", "authors": [ { "first": "Geoffrey M", "last": "Jacquez" }, { "first": "Andy", "last": "Kaufmann" }, { "first": "Jaymie", "last": "Meliker" }, { "first": "Pierre", "last": "Goovaerts" }, { "first": "Gillian", "last": "AvRuskin" }, { "first": "Jerome", "last": "Nriagu" } ] }
PMC1083419	Background ========== Compared to older children, children under 2 years are at increased risk of influenza-related hospitalization \[[@B1]-[@B4]\]. In the United States, it is now recommended that children 6--23 months and household contacts of all children under 2 be vaccinated annually against influenza \[[@B5]\]. Older children without high risk medical conditions are at lower risk of influenza-related complications that may result in hospitalization and are not currently recommended to receive influenza vaccination \[[@B1]\]. Recommendations regarding annual influenza vaccination of children may be aided by considering preference data on the potential costs and benefits of such vaccination. Although some information is available for otitis media and pneumonia (potential influenza illness complications in children), no published studies have quantified preferences, either in terms of quality-adjusted life-years saved or in willingness-to-pay, for preventing influenza and influenza vaccination-related events in children. We conducted a survey to measure the value placed by community members on preventing uncomplicated influenza in children and vaccination-related adverse events including the amount of time required by parents when caring for children with these conditions. Methods ======= Study sample ------------ Potential respondents were identified via a random sample of all adults enrolled in Harvard Pilgrim Health Care, a large nonprofit health plan in New England, during 2001. Procedures were approved by the Harvard Pilgrim Health Care Human Subjects Committee. Survey protocol --------------- Participation in the study consisted of completion of a 30-minute telephone survey. At least one week prior to the telephone interview, potential respondents were mailed materials to refer to during the interview. These included a booklet with written descriptions of uncomplicated influenza and one of two vaccination-related adverse events, either anaphylaxis or Guillain-Barré syndrome (Appendix 1 \[see the [additional file 1](#S1){ref-type="supplementary-material"}\]). The booklet also included instructions for the time-tradeoff and willingness-to-pay questions for the interviewer to review with the respondent during the interview. Respondents were interviewed during April-September, 2002. The 30-minute closed-ended interview included time-tradeoff questions \[[@B6]\] for uncomplicated influenza and one vaccination-related adverse event. We used the time-tradeoff method to measure preferences because our pilot tests suggested that this method would be more sensitive than other commonly used procedures, such as standard gamble \[[@B6]\]. Time-tradeoff questions asked the respondent to define the amount of time that the respondent would be willing to lose from their own lives so as to prevent uncomplicated influenza in the respondent\'s child or a hypothetical child (Table [1](#T1){ref-type="table"}). The time-tradeoff algorithm used 4 different starting bids followed by three additional bids titrating up. The respondent was then asked the maximum amount of time he or she would be willing to give up within the final interval. Initial bids were 1 week, 2 weeks, 1 month, and 6 months. The time-tradeoff amounts elicited for this study are preference-based measures \[[@B7]\]. Theoretically, the amounts should be lower for less severe health states (respondents willing to give up less time) and higher for more severe health states. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Sample time-tradeoff and willingness to pay questions ::: ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Introduction: When thinking about the following questions: Do not includethe costs of medical services or prescription medications. Assume these would already be covered by full insurance even if you don\'t have this now. Do includethe value of preventing the following for you or your child: pain and suffering, inconvenience, and lost time for productive activities (paid work or work in the home) or leisure time. Time-tradeoff Question:Think about what portion of your remaining life, if any, you would be willing to trade off from the end of your life to prevent the flu in your 1-year-old child. You can choose any amount of time in days, weeks, months or years. Would you be willing to trade off a portion of your remaining life in order to prevent the flu in your 1-year old child? Remember that we are asking you to imagine what portion of YOUR OWN life you would be willing to trade off to prevent influenza in YOUR CHILD. Willingness to Pay Question:Would you be willing to pay some amount of money to prevent an episode of the flu in your one-year-old child? \[If Yes\] Would you be willing to pay \$200? \[If yes\] Would you be willing to pay \$400? \[If no\] Would you be willing to pay \$100? What is the most you would be willing to pay? ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ::: Respondents were also asked two types of willingness-to-pay questions. First, respondents were asked to state the maximum amount of money they would be willing to pay to prevent uncomplicated influenza or one vaccine adverse event (Table [1](#T1){ref-type="table"}). Second, they were asked their willingness-to-pay for a specific risk reduction in an influenza vaccination-related adverse event (Table [2](#T2){ref-type="table"}). Willingness-to-pay was measured using dichotomous-choice double-bounded questions followed by an open-ended question asking for their maximum willingness-to-pay for both willingness-to-pay questions. Respondents were randomized to four different initial bids to minimize any bias due to the size of the initial bid (\"anchoring bias\"). For example, initial bids for avoiding uncomplicated influenza in a 1-year-old child were \$25, \$50, \$100, \$200. Follow-up bids were double the initial bid if the respondent said yes to the initial bid, and half the initial bid if the respondent said no to the initial bid. See Table [1](#T1){ref-type="table"} for additional details. (For further details on dichotomous-choice questions, see Bateman IJ et al. Economic valuation with stated preference techniques: a manual. Northampton: Edward Elgar Publishing. 2002 or Carson RT. Contingent valuation: A user\'s guide. Environ Sci Technol2000; 34(8):1413--1418\].) ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Sample survey question to elicit willingness-to-pay for avoiding a vaccine adverse event ::: Vaccine 1 Vaccine 2 --------------------------------------------------------------------------------------- --------------------------------------------------------------------------------------- • Reduces risk of influenza from 15 in 100 to 3 in 100 • Reduces risk of influenza from 15 in 100 to 3 in 100 • 3 in 10 risk of sore arm after vaccination • 3 in 10 risk of sore arm after vaccination • 1 in 100 risk of mild side effects such as fever and muscle aches after vaccination • 1 in 100 risk of mild side effects such as fever and muscle aches after vaccination • 1 in 10,000 risk of severe allergic reaction after vaccination • 1 in 10,000 risk of severe allergic reaction after vaccination • *1 in 1 million chance of Guillain-Barré Syndrome after vaccination* • *No chance of Guillain-Barré Syndrome after vaccination* • Costs \$10 • Costs \$? ::: Additional questions included how much time had been spent caring for a child with influenza, sociodemographic characteristics, whether the respondent\'s child had experienced any of the described conditions, and attitude toward childhood vaccination (the respondent was asked how well they agreed with the statement: \"the benefits of vaccines are worth the risks\" using a 5-item Likert scale from \"strongly agree\" to \"strongly disagree\"). To simplify survey design, the order of health state descriptions was not randomized. Respondents were not presented with inconsistencies in their scores, but were asked at the end of the survey if they wished to change any of their answers once the full set of health states had been considered. Analysis of survey data ----------------------- We calculated summary statistics for time-tradeoff and willingness-to-pay amounts, including medians, means, 5th and 95th percentiles, and minimums and maximums. Differences between time-tradeoff and willingness-to-pay responses for avoiding influenza in a 1-year-old as compared with a 14-year-old were evaluated using the sign test for paired observations. Medians were reported along with means because the distributions were skewed toward zero, especially for uncomplicated influenza. For the time-tradeoff questions, we discounted the time-tradeoff amounts. Respondents were instructed to assume that time would be traded off from the end of their life, therefore we calculated the present value for time-tradeoff amounts using the difference between the participant\'s age and life expectancy \[[@B8]\] as the timeframe over which to discount, and used a rate of time preference of 3% per year as the discount rate. The effects of respondent socio-demographic characteristics on time-tradeoff and willingness-to-pay amounts were evaluated using estimated random effects from the generalized linear mixed models (GLMM) version of Poisson regression (SAS v. 8.2, SAS Institute, Cary, NC). This is a two-part model for analyzing data with multiple outcomes from the same respondent. The advantage of using this model for analyzing time-tradeoff and willingness-to-pay data is to account for the correlated responses (in this case the time-tradeoff or willingness-to-pay values) for multiple health states. The underlying assumption being that a respondent\'s responses across health states will be correlated, i.e., if a respondent has a higher than average time-tradeoff value for one health state they are more likely to have higher than average time-tradeoff values for the other health states. (For additional details, see Burton et al., Tutorial in biostatistics: extending the simple linear regression model to account for correlated responses: an introduction to generalized estimating equations and multi-level mixed modeling. Statistics in Medicine1998;17:1261--1291). A Poisson model was selected because time-tradeoff and willingness-to-pay amounts are not distributed normally but are skewed toward zero. Dependent variables included in the model were age, sex, education (college or more: yes/no), marital status (married: yes/no), income (less than three times poverty level or not), health status (good or better: yes/no), whether the respondent has children under 18 in the household, and whether a child they knew has experienced any of the outcomes described in the survey. We also evaluated the effect of the initial bid by including a dummy variable for each of the four initial bids. Results ======= Participants ------------ 493 letters were mailed to potential respondents to invite them to participate in the survey. Of those invited, 23% could not be contacted to schedule an interview, and 2% were unable to be interviewed in English. Of the 373 remaining potential participants, 32% agreed to participate and completed an interview, 2% scheduled an interview, but refused to participate at the time of the interview, 4% scheduled an interview, but could not be contacted to complete the interview, and 62% refused to participate by either returning an opt-out card or declining (either actively or passively) to schedule an appointment at the time of the follow-up phone call. The overall response rate was 26%. There were more women than men in the sample (Table [3](#T3){ref-type="table"}), and respondents were more educated and had higher incomes than the general U.S. population \[[@B10],[@B11]\]. Respondents included 46% who reported having a child who had experienced an episode of influenza, 13% who reported having a child who had experienced a severe allergic reaction, and 1% who reported having a child who had experienced Guillain-Barré Syndrome. This study did not have access to medical records to verify reported events. Seven respondents (6%) who were unable to understand the questions according to interviewer assessment were excluded from further analysis. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Respondent characteristics (N = 112) ::: Characteristics Percent ------------------------------------------ ------------- Female 58.9% Mean age, years (SD) 48 (13.8) Married 66.1% Have Children Under 18 Yes 40.2% No 58.9% Declined to answer 0.9% Race White 90.2% Black 6.3% Asian 3.6% Educational Attainment High school graduate 13.4% Some college 17.0% College degree 33.0% Post graduate training 36.6% Household income in 2001 \$25,000 or less 5.4% \$25,001--\$50,000 17.0% \$50,001--\$75,000 25.0% \$75,001--\$100,000 13.4% More than \$100,000 26.8% Declined to answer 11.6% Did not know 0.9% Benefits of vaccines are worth the risks Strongly agree 44.6% Agree 39.3% Neither agree nor disagree 6.3% Disagree 6.3% Strongly disagree 0.9% Did not know 2.7% Current health (Very good or excellent) 66.9% ::: Time-tradeoff and willingness-to-pay ------------------------------------ Time-tradeoff amounts increased with the severity of the health state (Table [4](#T4){ref-type="table"}). Approximately half of the respondents were not willing to trade off any time from their own life to prevent uncomplicated influenza in a hypothetical child (51% for a 1-year-old child and 60% for a 14-year-old child). The median response for vaccine-induced severe allergic reaction was one month and for Guillain-Barré syndrome 3 years. Discounted mean responses were higher for preventing uncomplicated influenza in a 14-year-old child (41 days) as compared with a 1-year-old (29 days) yet this difference was not significant. An analysis of ranks shows that time-tradeoff responses were consistently higher for uncomplicated influenza in a 1-year-old (despite the higher mean TTO result for uncomplicated influenza in a 14-year-old) (p-value = 0.0596). ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Time-tradeoff amounts for uncomplicated influenza and vaccination-related adverse events ::: a. Undiscounted ---------------------------------------------- ---------- ----------- -------------------------- ---------------------------- Influenza in 1-year old child\* 0 68 days 0 -- 1 year 0 -- 6 years Influenza in 14-year old child\* 0 86 days 0 -- 1 year 0 -- 6 years Severe allergic reaction in 1-year old child 30 days 319 days 0 -- 3 years 0 -- 20 years Guillain-Barré Syndrome in 1-year old child 3 years 5 years 0 -- 20 years 0 -- 25 years b. Discounted Median Mean 5^th^-- 95^th^percentile Range (minimum -- maximum) Influenza in 1-year old child\* 0 29 days 0 -- 129 days 0 -- 2.8 years Influenza in 14-year old child\* 0 41 days 0 -- 230 days 0 -- 2.8 years Severe allergic reaction in 1-year old child 11 days 119 days 0 -- 1.1 years 0 -- 8.3 years Guillain-Barré Syndrome in 1-year old child 352 days 2.1 years 0 -- 7.4 years 0 -- 12.1 years \P-value = 0.0596 using the sign test to assess whether WTP for influenza in a 1-year-old child is greater than WTP for influenza in a 14-year-old. ::: Median willingness-to-pay amounts ranged from \$100 to prevent an episode of influenza in a 14-year-old child to \$5,000 to prevent a case of Guillain-Barré syndrome (Table [5](#T5){ref-type="table"}). Respondents were also willing to pay a premium of \$50 or \$100 for hypothetical influenza vaccines that either had no risk of severe allergic reaction or Guillain-Barré syndrome. ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Willingness-to-pay amounts in 2002 dollars for health states ::: a. For each health state avoided* ------------------------------------------------------------------------------ -------- -------- -------------------------- ---------------------------- Influenza in 1-year old child\* 175 469 0 -- 1500 0 -- 10,000 Influenza in 14-year old child\* 100 288 0 -- 1000 0 -- 5,000 Guillain-Barré Syndrome in 1-year old child 5,000 28,579 100 -- 100,000 10 -- 1,000,000 Severe allergic reaction in 1-year old child 400 4,968 0 -- 10,000 0 -- 200,000 b. For risk reductions Median Mean 5^th^-- 95^th^percentile Range (minimum -- maximum) Influenza vaccine with no risk of Guillain-Barré Syndrome after vaccination 100 341 3 -- 2,000 0 -- 5,000 Influenza vaccine with no risk of severe allergic reaction after vaccination 50 223 10 -- 1,000 0 -- 5,000 \*P-value \< .0001 using the sign test to assess whether WTP for influenza in a 1-year-old child is greater than WTP for influenza in a 14-year-old. ::: According to the Poisson-based regression analysis, one initial bid (1 week) was associated with higher time-tradeoff responses. None of the other respondent-specific variables (e.g., education, income, health status, or having young children) affected either TTO or WTP results. (Tables [6](#T6){ref-type="table"} and [7](#T7){ref-type="table"}) ::: {#T6 .table-wrap} Table 6 ::: {.caption} ###### Effect of respondent characteristics and initial bids on time-tradeoff amounts ::: Variable Type of variable Impact on TTO 95 % CI P value ------------------------------------- ---------------------------------------------- --------------- ------------------ ---------- Independent variables: TTO amounts^1^ Dependent variables: Baseline^2^ Constant 0.5545 (0.2402, 1.2798) 0.1617 Age^3^ 18--34 Binary 1.0385 (0.4904, 2.1994) 0.9200 35--49 Binary 1.1698 (0.6299, 2.1723) 0.6137 50--64 Binary 1.5448 (0.7921, 3.0129) 0.1959 Female Binary 0.8629 (0.5926, 1.2565) 0.4342 Education: Some college or less Binary 1.1579 (0.7537, 1.7789) 0.4966 Not married Binary 0.9204 (0.5815, 1.4569) 0.7188 Health (Worse than average) Binary 1.1043 (0.7343, 1.6606) 0.6279 No children under 18 in household Binary 0.7663 (0.4541, 1.2930) 0.3114 Version (initial bid)^4^ 1 (1 week) 4 binary variables, one for each initial bid 2.8454 (1.7138, 4.7242) \<0.0001 2 (1 month) 1.3331 (0.7943, 2.2374) 0.2694 3 (2 weeks) 1.4705 (0.8741, 2.4739) 0.1415 Income (less than 3X poverty level) Binary 1.0092 (0.5398, 1.8870) 0.9766 Number of disease experienced: None Binary (0 v. 1 or more) 1.2406 (0.7396, 1.9394) 0.3370 ^1^Independent variables included for each respondent were uncomplicated influenza in a 1-year-old, uncomplicated influenza in a 14 year-old, and anaphylaxis or Guillain-Barré syndrome. R-squared = 0.2194. ^2^Baseline refers to a person over 65 years of age, male, college graduate, married, good health, with income three times or greater than poverty level, children under 18 living in the household, familiar with at least one of the conditions in the survey, and responding to survey version 4 (initial bid of 6 months). ^3^Compared to 65 and over. ^4^Compared to version 4 (initial bid of 6 months). ::: ::: {#T7 .table-wrap} Table 7 ::: {.caption} ###### Effect of respondent characteristics and initial bids on willingness-to-pay amounts ::: Variable Type of variable Impact on WTP 95 % CI P value ------------------------------------- ---------------------------------------------- --------------- ------------------ --------- Independent variables: WTP amounts^1^ Dependent variables: Baseline^2^ Constant 0.9368 (0.3763, 2.3324) 0.8864 Age^3^ 18--34 Binary 1.1458 (0.5055, 2.5974) 0.7401 35--49 Binary 1.6824 (0.8564, 3.3049) 0.1265 50--64 Binary 1.3450 (0.6492, 2.7865) 0.4177 Female Binary 0.7866 (0.5222, 1.1851) 0.2444 Education: Some college or less Binary 0.9690 (0.6066, 1.5479) 0.8934 Not married Binary 0.7606 (0.4610, 1.2551) 0.2773 Health (Worse than average) Binary 1.1183 (0.7166, 1.7451) 0.6166 No children under 18 in household Binary 1.1254 (0.6360, 1.9911) 0.6797 Version (initial bid)^4^ 1 (\$100) 4 binary variables, one for each initial bid 1.3809 (0.7945, 2.4001) 0.2460 2 (\$200) 1.0257 (0.5832, 1.8040) 0.9286 3 (\$25) 0.7572 (0.4293, 1.3355) 0.3293 Income (less than 3X poverty level) Binary 0.9811 (0.4957, 1.9416) 0.9554 Number of disease experienced: None Binary (0 v. 1 or more) 0.6862 (0.4216, 1.1169) 0.1254 ^1^Independent variables included for each respondent were uncomplicated influenza in a 1-year-old, uncomplicated influenza in a 14 year-old, and anaphylaxis or Guillain-Barré syndrome. R-squared = 0.1725. ^2^Baseline refers to a person over 65 years of age, male, college graduate, married, good health, with income three times or greater than poverty level, children under 18 living in the household, familiar with at least one of the conditions in the survey, and responding to survey version 4 (initial bid of \$50). ^3^Compared to 65 and over. ^4^Compared to version 4 (initial bid of \$50). ::: Discussion ========== Fewer than half the respondents were willing to trade any time to prevent uncomplicated influenza in a hypothetical child, but many (73%) were willing to give up some time to prevent vaccination-related complications. Most participants indicated a willingness-to-pay to avoid uncomplicated influenza as well as a severe allergic reaction or Guillain-Barré syndrome due to vaccination, but there was substantial variation in the amounts they were willing to pay. This study did not evaluate willingness-to-pay and time-tradeoff amounts for complications of influenza such as otitis media and hospitalization that should be included in an economic evaluation of influenza vaccine. They were not included because values for these conditions were collected in a previous study conducted on a random sample of adults in the United States. In this study, we found that the median time-tradeoff amount for acute otitis media was 4 days, non-hospitalized pneumonia was 65 days, and hospitalization due to pneumonia was 214 days \[[@B12]\]. For preventing a case of uncomplicated influenza in a 1-year-old and a 14-year-old, 51% and 60% of respondents were not willing to trade any time. For the same health states, far fewer respondents reported zero as their willingness-to-pay to prevent uncomplicated influenza in a 1-year-old (13%) and a 14-year-old (14%). Since the smallest unit respondents could trade was one day in the time-tradeoff questions, respondents that might have been willing to trade a fraction of a day might have responded with zero when the true tradeoff value could have been between 0 and 1 day. Allowing respondents to trade minutes or hours could have resulted in fewer non-zero responses. In this study, respondents were not asked about time periods of less than one day and willingness-to-pay appears to be a more sensitive metric for valuing temporary health states. The willingness-to-pay results for the safer (hypothetical) vaccines should be interpreted cautiously. There is considerable evidence that people have difficulty valuing small risk reductions and also are willing to pay more in a hypothetical situation \[[@B13],[@B14]\]. In this study, using the responses from the risk reduction questions results in willingness-to-pay estimates orders of magnitude higher than when respondents directly valued the prevention of one case of either event. Differences could be attributable to (1) risk aversion (the second set of values ignores any premium respondents are willing to pay to avoid a risk), (2) overestimation of small probabilities, and/or (3) the voluntary nature of the risk (because parents voluntarily choose to expose their children to vaccines and may feel responsible for bad outcomes associated with them, they may be willing to pay more to reduce that risk than they would in a situation that included a similar risk of experiencing a condition in a way unrelated to any decision or action by the parent). In any case, the values for questions on risk reductions are sufficiently different to cause some concern about the incorporation of probabilities into contingent valuation questions. There are a number of challenges in measuring the value of health for very young children including the use of parents as proxy respondents, the valuation of temporary health states, and whether or not to include family spillover effects \[[@B15]-[@B17]\]. Applying utilities from standardized instruments such as the Health Utilities Index (HUI) or the EQ-5D which were developed to value chronic health states in adults (and children 6 and older in the case of the HUI) are unlikely to be accurate for valuing temporary or transient health states in very young children \[[@B15],[@B18]\]. There is a small but growing body of literature in the area of valuing temporary health states. Alternatives such as the waiting-tradeoff, conjoint analysis, \"chained\" health states, and other modifications of the time-tradeoff method have been proposed without any clear consensus on a preferred method \[[@B19]-[@B22]\]. This study demonstrates the use of a modified time-tradeoff question that differs from that typically used to value chronic health states, in which respondents choose between years of life with and without a stated condition. (For a discussion of the appropriateness of using time-tradeoff questions to elicit utilities for economic evaluations, see Dolan P. Output measures and valuation in health. In: Economic evaluation in health care: Merging theory with practice. Eds: Drummond M, McGuire A. New York: Oxford University Press. 2001.) Approaches similar to the one used in this study have been employed in previous studies \[[@B12],[@B23],[@B24]\], but clearly more research is needed to reach consensus in the field regarding optimal methods for valuing temporary health states in young children. There has been increasing recognition of family spillover effects (i.e., the effect of one family member\'s illness on other family members) on health-related quality-of-life. The potential importance of including these effects in economic analyses can be quite significant for illnesses in the very young and the very old \[[@B16],[@B17]\]. Our approach of valuing changes in health-related quality-of-life for both parent and child is consistent with the inclusion of family spillover effects in the economic evaluation. Our study evaluated the tradeoff between life in a parent and a temporary health state in their child. The inclusion of loss of quality-of-life for both parent and child prevents the time-tradeoff amounts from being directly comparable to utility values from generic utility instruments for measuring reductions in quality-of-life for chronic health states, such as the Health Utilities Index \[[@B25]\] or the EQ-5D \[[@B26]\]. Clearly more research will be needed to establish the optimal method for valuing family spillover effects. The generalizability of these study results are limited by the small sample size and a relatively homogeneous and geographically-limited respondent population whose characteristics differ from those of the general U.S. population. Sampling from the general membership of a large New England HMO resulted in a group of respondents with little variation in income, and that otherwise differed from characteristics of the general U.S. population. For example, more than 75% of respondents had an annual household income greater than \$50,000. The response rate was somewhat low, but not atypical for similar telephone surveys. Given these limitations, the results are sufficiently robust to justify a larger study for validation that also included additional uncommon severe outcomes of influenza in children, including encephalopathy and death \[[@B27]\]. The annual probability of a one-year-old experiencing an influenza-related illness is approximately 16% in non-pandemic years \[[@B28]-[@B37]\] and varies from 0--23% in a non-pandemic year. The risk of an influenza-related hospitalization is about 3 per thousand for a child at low risk for influenza-related complication \[[@B1]-[@B3]\]. In contrast the risk of anaphylaxis from influenza vaccination is estimated at approximately 1 in 4 million \[[@B38]\] based on surveillance from the 1976 swine flu vaccine program. Guillain-Barré syndrome was associated with receipt of swine flu vaccine in 1976 with a risk of 1 per 100,000 persons vaccinated \[[@B38]\], although children were associated with a lower risk of GBS than adults \[[@B39]\], and studies since 1976 have not found a clear association of GBS with influenza vaccination \[[@B40],[@B41]\]. Most studies of an association of Guillain-Barré syndrome with influenza vaccination have been among adults and not children. Economic analyses can provide information useful in comparing the benefits of vaccination with the risks of adverse events. The variability in preferences and willingness-to-pay observed in this study suggests that different community members may appraise the desirability or cost-effectiveness of influenza vaccination quite differently. The relatively low value many respondents attributed to uncomplicated influenza could provide insight into low coverage rates for influenza vaccination among children. Concern about the safety of vaccination is shown by the premium most respondents were willing to pay for a vaccine with lower risks of adverse events. Information on the costs, benefits, risks, and community preferences can aid policy decisions regarding influenza vaccination. Authors\' contributions ======================= LAP designed the study, developed the survey, oversaw data collection efforts, directed the analysis, and drafted the manuscript. CBB and TMU participated and contributed data during survey development and critically reviewed the manuscript. VHR supervised data collection and participated in statistical analyses. GTR conducted statistical analyses. MIM participated in survey development and critically reviewed the manuscript. MIM, BS, WWT, and KJ participated in survey development. TAL participated in design and helped draft the manuscript. All authors read and approved the final manuscript. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Appendix 1. Health State Descriptions for Outcomes Prevented by Influenza Vaccination ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ We would like to thank Ken Kleinman for providing statistical expertise. Financial support for this study was provided by the Vaccine Safety Datalink Project of the National Immunization Program, Centers for Disease Control and Prevention.	PubMed Central	2024-06-05T03:55:55.749875	2005-3-21	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1083419/", "journal": "Health Qual Life Outcomes. 2005 Mar 21; 3:18", "authors": [ { "first": "Lisa A", "last": "Prosser" }, { "first": "Carolyn Buxton", "last": "Bridges" }, { "first": "Timothy M", "last": "Uyeki" }, { "first": "Virginia H", "last": "Rêgo" }, { "first": "G Thomas", "last": "Ray" }, { "first": "Martin I", "last": "Meltzer" }, { "first": "Benjamin", "last": "Schwartz" }, { "first": "William W", "last": "Thompson" }, { "first": "Keiji", "last": "Fukuda" }, { "first": "Tracy A", "last": "Lieu" } ] }
PMC1083420	Background ========== The c-Myc oncogene is one of the most frequently dysregulated genes in human tumours. Myc was originally identified as the cellular homolog of the transforming part of the viral isolate MC29 \[[@B1]\]. The c-Myc oncogene is a member of the basic-helix-loop-helix-leucine-zipper transcription (bHLH-ZIP) factors, which are essential for different cellular processes \[[@B2]\]. Paradoxically, c-Myc promotes both cell cycle progression and apoptosis under low serum condition \[[@B3],[@B4]\]. c-Myc regulates the cellular processes by controlling a large number of target genes \[[@B5],[@B6]\] through heterodimerization with its biological partner Max \[[@B7]-[@B9]\]. The abundance of the Myc-Max heterodimer is effectively controlled by the short lived Myc protein \[[@B10]\]. The Myc protein is under tight and complex control mechanisms \[[@B11]\]. Critical phosphorylation events determining the protein half life occur in Myc homology box I (aa45-aa65) \[[@B10]\]. These detrimental events involve the hierarchical phosphorylation of S62 and T58 by ERK1/2 MAPK and GSK3β, respectively \[[@B12]\]. It is widely accepted that these kinases are involved in the phosphorylation events at these residues although other reports question the role of MAPK \[[@B13]\]. These two kinases are part of two different Ras effector pathways. The presence of different Ras isoforms provides for selective activation of specific Ras effector pathway, although this can only be shown in vivo \[[@B14]\]. It has been reported that PI-3 kinase is most effectively activated by M-Ras and R-Ras and to a less extent by H-Ras \[[@B15],[@B16]\]. On the other hand, Raf-1 is most effectively activated by K-Ras \[[@B17],[@B18]\]. This selective activation of different Ras effector pathways has opposing effects on Myc controlled functions. Whereas the activation of Raf fails to suppress Myc induced apoptosis, the activation of PI-3 Kinase can effectively suppress it \[[@B19]\]. A key component of the PI3-kinase/Akt (PKB) pro-survival pathway is GSK3 \[[@B20]\], whereas the active phosphorylated form of ERK1/2 MAPK is a downstream signal in the signalling cascade Ras/Raf/MEK \[[@B21]\]. The ERK1/2 MAPK is one of three major MAPK signalling pathways, which also includes JNK/SAPK and p38 kinase. Constitutive activation of MEK/ERK has been reported in cancer cells \[[@B22],[@B23]\], with a possible role in cell transformation and oncogenesis \[[@B24]\]. The constitutive activation of MAPK ERK1/2 could be linked to the mitogen independence reported for oncogenes like Ras \[[@B25]\], Raf \[[@B26]\], Jun \[[@B27]\] and Myc \[[@B4]\]. Therefore, one of the aims of this study was to examine the status of active ERK2 in Myc transformed chick embryo fibroblasts (CEF), the ideal model for Myc induced transformation. Our second aim was to examine the possibility of a cross talk between ERK2 and GSK3 in Myc transformed fibroblasts using LiCl to inhibit GSK3. Reports on signalling between GSK3 and ERK1/2 are very scarce. Nonetheless, a recent report has demonstrated that GSK3β was a natural activator of the JNK/SAPK pathway \[[@B28]\]. Furthermore, it has been demonstrated that GSK3β could be phosphorylated on Ser9 and therefore inactivated by ERK1/2 mediated pathways, mainly through p90^rsk^but also through a novel mechanism downstream of ERK1/2 \[[@B29]\]. These findings need to be verified in transformed phenotype. Results and Discussion ====================== We have found that v-Myc (MC29) transformed fibroblasts have almost non-detectable active ERK2 (Figure [1A](#F1){ref-type="fig"}). A control experiment using the SFCV vector without an insert was performed in parallel with every experiment to exclude any effect for the transfection procedure. Cells transfected with the control vector gave identical results to the non-transfected control CEF cells. The addition of 100 mM LiCl was very successful in restoring (not fully) the levels of active ERK2 in v-Myc transformed fibroblasts to those found in non-transformed fibroblasts within the time scale of the experiment. The barely detectable basal levels of phosphorylated ERK2 in v-Myc transformed fibroblasts showed an increase after the addition of LiCl at the earliest time point of 20 minutes (31% of basal levels in non-transformed control CEF). These levels were almost completely restored to the levels found in non-transformed CEF after 80 minutes (83% of basal levels in non-transformed control CEF). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### The effect of LiCl on the levels of active ERK in Myc transformed and non transformed fibroblasts.(A)A time course for the effect of LiCl on the levels of active ERK2 in v-Myc transformed fibroblasts (panel 1). Total ERK levels were not affected and were used as a loading control (panel 2). Panel 3 is a graphical representation of the ERK2 levels in Li^+^and K^+^treated v-Myc cells after normalisation to ERK2 levels in non-transformed control CEF (identical to CEF transfected with empty vector) growing under normal conditions. (B)A time course for the effect of LiCl on the levels of active ERK2 in c-Myc transformed fibroblasts (panel 1). Total ERK levels were not affected and were used as a loading control (panel 2). Panel 3 is a graphical representation of the ERK2 levels in Li^+^and K^+^treated c-Myc cells after normalisation to ERK2 levels in non-transformed control CEF growing under normal conditions. (C)A time course for the effect of LiCl on the levels of active ERK2 in non-transformed control CEF (panel 1). Total ERK levels were not affected and were used as a loading control (panel 2). Panel 3 is a graphical representation of the ERK2 levels in Li^+^and K^+^treated cells after normalisation to ERK2 levels in non-transformed control CEF growing under normal conditions. The same pattern of expression was seen in several independent experiments for all the panels. ::: ![](1476-4598-4-13-1) ::: On the other hand, c-Myc transformed fibroblasts have shown attenuated but detectable active ERK2 levels compared to the non-transformed CEF. The addition of 100 mM LiCl fully restored the levels of active ERK2 to those found in non-transformed CEF (Figure [1B](#F1){ref-type="fig"}). The reduced basal levels of ERK2 in c-Myc transformed fibroblasts showed an increase at the earliest time point of 20 minutes (88% of basal levels in non-transformed control CEF) after the addition of LiCl and were comparable to the levels seen in the non-transformed CEF after 60 minutes. In the non-transformed CEF, the addition of LiCl enhanced the levels of active ERK2 considerably (Figure [1C](#F1){ref-type="fig"}). The increase in the levels of active ERK2 in CEF after the addition of LiCl was detectable after 20 minutes (153% of basal levels in non-transformed control CEF) and peaked after 40 minutes (350% of basal levels in non-transformed control CEF). In addition, adding 100 mM of LiCl increased the levels of inactive phosphorylated GSK3 α/β in a time dependent manner (Figure [2](#F2){ref-type="fig"}) in agreement with the pattern seen for the restored levels of active ERK2. We confirmed the activation of the pro-survival pathway PI3K signalling pathway after the addition of LiCl by the inhibition of apoptosis in Myc transformed fibroblasts (Figure [3](#F3){ref-type="fig"}). Compared to the KCl control, LiCl treatment resulted in 2.1, 3.6 and 2.4 fold reduced apoptosis in v-Myc, c-Myc and non-transformed control fibroblast cells, respectively. The different cell populations showed variable number of apoptotic cells after serum starvation. This is expected since v-Myc is a stronger inducer of proliferation and apoptosis than c-Myc \[[@B30],[@B31]\]. Other researchers have also demonstrated that inhibition of GSK3 using LiCl was contributory to apoptosis inhibition\[[@B32]\]. Other GSK3 inhibitors can be used to further support these findings. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### The effect of LiCl on the levels of GSK3 in non-transformed fibroblasts.(A)A western blot showing the levels of inactive or phosphorylated GSK3 α/β in non-transformed fibroblasts after the addition of Li^+^or K^+^salt control for the time points indicated. (B)A western blot showing the levels of phosphorylated and non-phosphorylated GSK3 α/β of the same samples in (A) above after stripping and re-probing of the blot with the appropriate antibody. ::: ![](1476-4598-4-13-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### The effect of LiCl on the apoptosis levels of serum starved Myc transformed and non-transformed fibroblasts.(A)A graphical representation of the percentage of apoptotic v/c-Myc transformed and non transformed fibroblasts 17 hours after serum deprivation and addition of Li^+^or K^+^. The error bars are the standard error of three independent experiments. The total number of cells counted for each experiment was 50 cells. (B)&(C)Representative sections of Hoeschst 33258 stained c-Myc cells after K^+^and Li^+^treatment, respectively. ::: ![](1476-4598-4-13-3) ::: Similar to what we have observed in our Myc transformed cells, previous researchers have demonstrated that ERK1/2 activity was repressed in c-Raf-1 (Raf22W), v-Ha-Ras, and v-Src transformed cells by a single-specificity tyrosine phosphatase \[[@B33]\]. A more recent report has also demonstrated attenuated levels of ERK2 in v-Jun transformed CEF cells, which was attributed to inefficient signalling between Ras and Raf, and increased levels of MAPK phosphatase \[[@B34]\]. In light of the roles of ERK and GSK3 in Myc protein phosphorylation and stability, we investigated the effect of LiCl addition on Myc protein half life in c-/v-Myc transformed fibroblasts. Figures [4A](#F4){ref-type="fig"} and [4B](#F4){ref-type="fig"} show that the addition of LiCl results in dramatic stabilisation of Myc in the non-transformed and c-Myc transformed cells, respectively. Surprisingly, although endogenous Myc is hardly detectable in non-transformed CEF (t~1/2~\< 1 minute), it was 30 fold more stable after the addition of LiCl (Figure [4A](#F4){ref-type="fig"}). We verified this using immunoprecipitation (data not shown). In comparison, c-Myc protein was 4 fold more stable after LiCl treatment. Not surprisingly, both endogenous Myc in non-transformed control fibroblasts and exogenously expressed Myc in c-Myc transformed fibroblasts had similar half life values after the addition of LiCl (30 and 32 minutes, respectively). However, LiCl failed to further stabilise Myc in v-Myc transformed fibroblasts (Figure [4C](#F4){ref-type="fig"}). Myc protein half life was 40 minutes compared to 38 minutes in the control cells. Since MC29 v-Myc has T58\>M (T61\>M in chicken), we did not expect any effect for LiCl on Myc protein half life in these cells, although it was a necessary control. Other researchers documented a small effect for LiCl on Myc stability (2 fold) in immortalised cell lines \[[@B35]\].. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Myc protein turnover after the addition of LiCl in Myc transformed and non transformed fibroblasts.(A)Endogenous Myc protein in Li+ and K^+^treated non-transformed fibroblasts. This was identical to fibroblasts transfected with empty vector. Panel 1 is the scanned image of a western blot autoradiograph. Panel 2 shows the half life values of the Myc protein under the different conditions. CEF represents the non-transformed control fibroblasts (B)Turnover of the Myc protein in Li+ and K^+^treated c-Myc transformed fibroblasts. Panel 1 is the scanned image of a western blot autoradiograph. Panel 2 shows the half life values of the Myc protein under the different conditions. CEF represents the non-transformed control fibroblasts (C)Turnover of the Myc protein in Li^+^and K^+^treated v-Myc transformed fibroblasts. Panel 1 is the scanned image of a western blot autoradiograph. Panel 2 shows the half life values of the Myc protein under the different conditions. CEF represents the non-transformed control fibroblasts. The experiments were independently repeated three times and one representative experiment is shown. ::: ![](1476-4598-4-13-4) ::: We can conclude that LiCl has direct effects on the hierarchical phosphorylation of S62 and T58 (S65 and T61 in chicken) by controlling the levels of active ERK2 and GSK3, respectively. The results in this study show that this is important for the Myc half life in the non-transformed and c-Myc transformed fibroblasts but not in the v-Myc transformed cells. In this context, other researchers have found that S62 phosphorylation was necessary for Myc stabilization following Ras activation or serum stimulation\[[@B36]\]. Conclusion ========== In this short communication we provided important findings about Myc induced transformation. The abrogation of active MAPK in Myc transformed cells can potentially provide an insight into the mechanism of Myc induced transformation. Clarification of the mechanism of ERK2 inactivation in Myc transformed CEF is needed. Furthermore, it is critical to examine the implications of the differences in active ERK2 levels between v-Myc and c-Myc transformed cells and the possible role this has in Myc induced transformation and protein stability. Last, we need to elucidate on the possibility of a cross-talk between GSK3 and ERK, as this could be a very important mechanism for controlling the Myc protein. Methods ======= Cell Culture, Transfection and Inhibition Studies ------------------------------------------------- Cell culture and transfection of the appropriate SFCV-Myc construct (10 μg) together with RCAN(A) helper (4 μg) into secondary CEF were performed as described previously \[[@B37]\]. A control experiment using the SFCV vector without an insert was used with every experiment as a transfection control. After G418 Neomycin selection (BDH, UK), cultures were expanded and used for the subsequent studies. At this stage, cells transfected with vectors containing either c-Myc or v-Myc were fully transformed as determined by anchorage independent growth and visible transformation characteristics, such as metamorphosis (data not shown). LiCl was added for 30 minutes at a final concentration of 100 mM to exponentially growing CEF, c-Myc or v-Myc cells before harvesting for western blotting. KCl was used in all the experiments as a salt control. For the Myc protein turnover studies, a protein synthesis inhibitor (emetine from Sigma, UK) was added 30 minute after the addition of either LiCl or KCL at a final concentration of 0.1 mM \[[@B38]\] for the indicated times shown in the figure. Apoptosis Induction and Measurement ----------------------------------- Apoptosis was induced by incubating the cells in a medium containing 0.2% serum for 17 hours. Serum starvation for longer periods of time resulted in apoptosis in almost all of the v-Myc cell population. To measure the percentage of the apoptotic cells, the cell population was divided into adherent and suspended cells. The adherent cells were trypsinised, washed in 1× phosphate buffered saline (PBS) and fixated in ice cold 3:1 glacial acetic acid/methanol solution. Then, the cells were permealised at room temperature using a solution of 1× PBS/0.1% triton X-100 for 5 minutes. The cells were then stained in a solution of 2.5 μg/μl Hoeschst 33258 (Sigma, UK) in 1× PBS/0.1 % triton X-100 on ice and protected from light. After that, the cells were washed twice in 1× PBS/0.1% triton X-100, made adherent onto a slide using a cytospin, and viewed and counted under an epi-fluorescent microscope using a DAPI filter. We treated the suspended cells in exactly the same way with the exception that they did not need trypsinisation. To calculate the total number of apoptotic cells in both adherent and suspended cells, we used the following formula: ![](1476-4598-4-13-i1.gif) SDS PAGE Western Blotting and Protein Half Life Measurement ----------------------------------------------------------- Cell lysates were prepared by lysing cultures in SDS-sample buffer containing 1% SDS without bromophenol blue or mercaptoethanol. Protein concentration was measured using Micro BCA reagent (Pierce, UK) before loading onto 7.5% SDS-PAGE gels. Transfer to nitrocellulose and western blotting was performed essentially as described previously \[[@B37]\], except that incubation with the primary antibody was performed in 2.5% BSA in TBS-Tween20 for ERK western blots. Active phosphorylated ERK was detected using rabbit polyclonal antibodies (catalogue number 9101, New England Biolabs, UK) and total phosphorylated and non-phosphorylated ERK levels were detected using rabbit polyclonal antibody (catalogue number 71--1800, Zymed, UK). Inactive phosphorylated GSK3 α/β (Ser21/9) protein was detected using a rabbit polyclonal antibody (catalogue number 9331S, Cell Signaling Technology, UK) and phosphorylated and non-phosphorylated GSK3 α/β levels were detected using a rabbit polyclonal antibody raised against GSK3 β but detected both GSK3 α and β (catalogue number 9332, Cell Signaling Technology, UK). Full length Myc protein was expressed in our laboratory and was used to raise rabbit polyclonal antibodies against. To re-probe a blot, it was first submerged in a solution containing 100 mM 2-mercaptoethanol, 2% SDS and 62.5 mM Tris HCl pH 6.7 at 50°C for 1 hour, with agitation. Then, the blot was washed in a solution containing TBS/0.1% Tween20 for 10 minutes three times at room temperature. Last, the blot was blocked and probed as above with the appropriate antibody. Equal loading of lanes was determined by staining the polyacrylamide gel after transfer onto a nitrocellulose membrane in coomassie blue solution for 2 hours (50% methanol, 10% acetic acid, 0.25% coomassie blue R-250) and de-staining in a solution containing 10% methanol and 5% acetic acid for 4 hours. The different band intensities were analysed using the Kodak 1D image analysis software and then they were plotted on a linear graph for the ERK levels. For the calculation of the Myc protein half life, the densitometric values were plotted on a semi-logarithmic graph against time and then fitted to an exponential line. The half life of the protein was calculated from the equation: ![](1476-4598-4-13-i2.gif)	PubMed Central	2024-06-05T03:55:55.753117	2005-4-5	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1083420/", "journal": "Mol Cancer. 2005 Apr 5; 4:13", "authors": [ { "first": "Osama", "last": "Al-Assar" }, { "first": "Dorothy H", "last": "Crouch" } ] }
PMC1083421	Background ========== Psychophysical procedures such as the visual analogue scale are commonly used to assess human pain magnitude. However, any of those assessments seem not to be suitable for evaluation of a change in thresholds of each class of afferent fibers, despite of the fact that quantitative analysis of each afferent class is essential for understanding underlying mechanisms of pathological states and evaluation of analgesic chemicals. It now becomes evident that hyperalgesia or allodynia produced by various chronic pain conditions is due to not only a change in threshold of fine afferent fibers but also participation of large myelinated afferents in producing allodynia in inflammation and sciatic nerve transection models \[[@B1]-[@B3]\]. Recently, one method that allows us to clarify a change in threshold of each afferent fiber quantitatively has been developed (Neurometer) \[[@B4]\]. That is a method to measure selectively the thresholds of three classes of afferent fibers by applying transcutanous sine-wave stimulation at three frequencies of 2000, 250 and 5 Hz via surface electrodes at a current intensity in the range of 0.01--9.9 mA to the skin. Neurometer is now widely used clinically to evaluate peripheral nerve sensitization and dysfunction in various painful states, including neuropathic pain or an effect of treatment of analgesic drugs \[[@B5]-[@B9]\]. Although several lines of evidence suggest the selective stimulation of three distinct afferent fibers by Neurometer, the validity of the method at single cellular level has not been made. There are three types of sensory afferent fibers that send sensory information to the CNS; unmyelinated C fibers send a long lasting delayed painful sensation, thinly myelinated Aδ fibers send a short and fast painful sensation and the thickly myelinated Aβ fibers send tactile information. Several attempts have been made to identify the types of peripheral sensory fibers activated by the sine-wave stimuli using behavioral pharmacological approaches such as blockade of nerve conduction by local anesthetics or use of topical application of capsaicin, etc. \[[@B10]-[@B12]\]. However, no studies have been made to clarify the single neuronal properties such as action potential generation or its frequency of peripheral fibers. Moreover, responses of spinal dorsal horn neurons have not been examined by peripheral stimulation with Neurometer. The purpose of this study was to clarify the selectivity of the sine-wave stimuli on the activation of sensory afferent fibers in detail. First, we performed behavioral tests with Neurometer to measure the withdrawal thresholds by transcutaneous application of three frequencies of stimuli. Next, we performed intracellular recordings from acutely isolated DRG neurons with attached dorsal roots to examine firing properties of C, Aδ and Aβ fibers. Moreover, using in vivopatch-clamp recordings, we analyzed excitatory synaptic responses evoked by the transcutaneous stimuli in SG neurons. Results ======= Behavioral response ------------------- We first examined the escape and/or vocalization behavior of rats to transcutaneous stimuli using Neurometer. When 250 and 5 Hz stimuli were applied to the left hind limb, rats escaped from the stimuli or vocalized. The thresholds for 250 and 5 Hz stimuli were quite similar, and were averaged to be 0.74 ± 0.06 mA (T250, n = 12) and 0.84 ± 0.12 mA (T5, n = 12), respectively. Next, 2000 Hz stimulation which is used to evaluate Aβ fibers was applied. Although Aβ fibers were thought to convey innocuous information, escape and/or vocalization behavior was also observed at higher intensity. The threshold for 2000 Hz stimulation (2.14 ± 0.28 mA, T2000, n = 12) was more than two times higher than T250 and T5. Classification of DRG neurons ----------------------------- Intracellular recordings were made from 53 DRG neurons. Dorsal root stimulation with a suction electrode elicited antidromic APs with latencies in the range of 0.2 -- 15.0 ms. Fast-conducting neurons had a brief action potential duration (APD), whereas slow-conducting ones had a relatively broad APD. Based on the threshold stimulus intensity (TSI), conduction velocity (CV) and APD, DRG neurons were divided into three subgroups, C, Aδ and Aβ fibers (Table [1](#T1){ref-type="table"}). These results were consistent with those reported previously \[[@B1],[@B13]-[@B17]\]. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Comparison of electrophysiological properties of DRG neurons ::: RMP (mV) CV (m/s) TSI (mA) APD (ms) ----------- -------------- -------------- -------------- -------------- C fibers -63.0 ± 1.5 0.6 ± 0.1 3.1 ± 0.5 1.38 ± 0.01 (n = 15) (51.0--74.0) (0.4--0.7) (2.5--4.6) (0.81--2.85) Aδ fibers -67.7 ± 2.4 5.9 ± 0.8 1.8 ± 0.6 0.63 ± 0.05 (n = 18) (55.0--71.0) (1.4--12.5) (1.6--2.1) (0.30--0.99) Aβ fibers -65.1 ± 1.8 19.6 ± 1.3 0.8 ± 0.2 0.29 ± 0.01 (n = 20) (52.0--76.0) (14.1--23.4) (0.04--1.34) (0.19--0.43) RMP, resting membrane potential; CV, conduction velocity; TSI, threshold of stimulus intensity. APD, duration of action potential measured at its half-maximal amplitude; Values are shown as means ± SEM. ::: Action potential generated in DRG neurons by sine-wave stimuli -------------------------------------------------------------- After classification of DRG neurons, C, Aδ and Aβ neurons were stimulated by sine-wave stimuli with gradual increase in intensity using Neurometer. ### 2000 Hz Stimulation As shown in figure [1A](#F1){ref-type="fig"}, no AP was generated even when the intensity was increased to more than 2 mA in 10 out of 15 C neurons. In the remaining 5 C neurons, one or two APs were observed at the point of intensity change at T5/T250. However, further increase in stimulus intensity did not produce any firing in these C neurons (data not shown). In all Aδ neurons tested, phasic firing with several APs was elicited at high intensities of \~2 mA (Figure [1B](#F1){ref-type="fig"}). On the other hand, tonic firing which lasted throughout the stimulation was produced in all Aβ neurons examined at near T5/T250 (Figure [1C](#F1){ref-type="fig"}). Figure [1D](#F1){ref-type="fig"} demonstrates the AP frequency plotted against stimulus intensity. The AP frequency of C neurons was near zero at intensities less than 3 mA. In Aδ neurons, AP was elicited at 1.3 mA and the frequency was increased to \~17 Hz at 2.8 mA. In Aβ neurons, AP was generated at 0.8 mA and the frequency increased steeply and reached \~140 Hz at the intensity of 2.7 mA. These data indicate that 2000 Hz stimulation at low intensity (near T5/250) selectively activates Aβ neurons. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Representative firing patterns of C (A), Aδ (B) and Aβ (C) neurons in response to 2000 Hz sine-wave stimulation.In this and subsequent figures 2 and 3, the stimulus intensities were shown above the traces. Action potentials recorded from three types of afferent fibers were shown on the right of the traces. The frequencies of APs were plotted against the stimulus intensity (D). In this and subsequent figures of 2 and 3, ○: C fiber (n = 15), △: Aδ fiber (n = 18), □: Aβ fiber (n = 20). ::: ![](1744-8069-1-13-1) ::: ### 250 Hz Stimulation 250 Hz stimulation at intensity below T5/T250 initiated APs in Aδ and Aβ neurons but not in C neurons examined (Figures [2A--C](#F2){ref-type="fig"}). As shown in figures [2D](#F2){ref-type="fig"}, the AP frequency of Aβ and Aδ neurons started to increase from 0.3 mA and reached plateau at 2 mA. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Representative firing patterns of C (A), Aδ (B) and Aβ (C) neurons in response to 250 Hz sine-wave stimulation.The frequencies of APs were plotted against the stimulus intensity (D). ::: ![](1744-8069-1-13-2) ::: ### 5 Hz Stimulation 5 Hz stimulation produced tonic firing in all three classes of DRG neurons examined (Figures [3A--C](#F3){ref-type="fig"}). As shown figure [3D](#F3){ref-type="fig"}, in C neurons, APs were elicited at 0.4 mA and the frequency reached plateau at \~1 mA. In Aδ neurons, APs were generated at 0.3 mA and the frequency reached the maximum at 0.8 mA. In Aβ neurons, APs were produced at 0.1 mA and the frequency reached plateau at 0.2 mA. Taken together these three studies shown in figures [1](#F1){ref-type="fig"},[2](#F2){ref-type="fig"},[3](#F3){ref-type="fig"}, the data indicate that C fibers are activated only by 5 Hz stimulation among the three specific frequencies. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Representative firing patterns of C (A), Aδ (B) and Aβ (C) neurons in response to 5 Hz sine-wave stimulation.The frequencies of APs were plotted against the stimulus intensity (D). ::: ![](1744-8069-1-13-3) ::: Excitatory synaptic responses evoked in SG neurons in vivoby cutaneous noxious stimuli ---------------------------------------------------------------------------------------- It is well known that SG neurons in the spinal dorsal horn receive synaptic inputs predominantly from C and Aδ fibers \[[@B18]-[@B20]\]. Therefore, SG is thought to be specifically implicated in nociceptive processing. As shown previously \[[@B21]-[@B24]\], SG neurons exhibited miniature and spontaneous excitatory postsynaptic currents (mEPSCs and sEPSC, respectively) under voltage-clamp condition at a holding potential of -70 mV. The large amplitude of EPSCs were reversibly blocked by TTX (1 μM) applied to the surface of the spinal cord leaving mEPSCs intact (data not shown). Therefore, the large amplitude of EPSCs (sEPSCs) were mediated by spontaneous firings of primary afferents or interneurons. Pinch stimulation applied to the skin of the hind limb produced a barrage of EPSCs (defined as evoked EPSCs; eEPSCs) in all 24 SG neurons examined (Figure [4](#F4){ref-type="fig"}). ::: {#F4 .fig} Figure 4 ::: {.caption} ###### *EPSCs in SG neurons occurring spontaneously and in response to pinch stimulation in vivo(A).Pinch stimulation was applied to the skin of the hind limb. EPSCs indicated by arrowheads and pinch-evoked EPSCs were shown an expanded time scale in B and C, respectively. ::: ![](1744-8069-1-13-4) ::: Excitatory synaptic responses evoked in SG neurons in vivoby transcutaneous sine-wave stimuli ----------------------------------------------------------------------------------------------- When the sine-wave stimuli was applied to the skin of the receptive field, 250 and 5 Hz stimuli evoked a barrage of EPSCs at a stimulus intensity of \~0.8 mA (Figures [5A](#F5){ref-type="fig"} and [5B](#F5){ref-type="fig"}). On the other hand, SG neurons showed an increase in frequency of large amplitude of EPSCs with 2000 Hz stimulation at higher intensities (Figure [5C](#F5){ref-type="fig"}) in spite of few inputs from Aβ fibers to SG, suggesting that the EPSCs elicited by 2000 Hz stimulation were mediated by the activation of Aδ afferents. Figure [5D](#F5){ref-type="fig"} shows the frequency of eEPSCs (\> 50 pA) evoked at the behavioral thresholds of the three stimuli. The average frequency of 5 Hz-evoked EPSCs at the behavioral threshold was increased to 378 ± 30 % of control (1.8 ± 0.3 Hz, n = 7) and that of 250 Hz-evoked EPSCs to 473 ± 69 % (2.9 ± 0.7 Hz, n = 8), and that of 2000 Hz-evoked EPSCs to 602 ± 198 % (2.9 ± 0.8 Hz at the behavioral threshold T2000, n = 6). The large amplitude EPSCs evoked by sine-wave stimuli were blocked by TTX (1 μM) and all EPSCs were abolished by a glutamatergic AMPA receptor antagonist CNQX (20 μM; Figures [6A](#F6){ref-type="fig"} and [6B](#F6){ref-type="fig"}). ::: {#F5 .fig} Figure 5 ::: {.caption} ###### EPSCs in SG neurons evoked by 5 (A), 250 (B) and 2000 (C) Hz transcutaneous sine-wave stimuli.In A-C, the stimulus intensities were shown above the traces. Lower two records in Awere shown in an expanded time scale. Relative frequencies of large amplitude of EPSCs evoked by the stimuli at intensities of T5, T250 or T 2000 (D). \P \< 0.05, \\P \< 0.01. ::: ![](1744-8069-1-13-5) ::: ::: {#F6 .fig} Figure 6 ::: {.caption} ###### EPSCs in SG neurons recorded in the presence of TTX (1 μM, A) or CNQX (20 μM, B). ::: ![](1744-8069-1-13-6) ::: Discussion ========== The present study demonstrates that 2000 Hz sine-wave stimulation activates selectively Aβ fibers, and 250 and 5 Hz stimuli excite noxious transmission to SG neurons mediated mainly through Aδ and C fibers, respectively, though 250 Hz stimulation also activates Aβ fibers. Our study is the first to examine the selective activation of sensory transmission by the sine-wave stimuli at the cellular level, and the results support that transcutaneous sine-wave stimuli can be applied to examine clinically the changes of threshold in each subtype of afferent fibers following pathological conditions or after treatment with analgesics based on the combination of three sine-wave stimuli. Activation of nociceptive transmission by sine-wave stimulation --------------------------------------------------------------- We observed that the stimulus thresholds producing the escape and/or vocalization behavior in rats with 250 and 5 Hz stimuli were not significantly different, whereas the threshold for 2000 Hz stimulation was two times higher than those of 250 and 5 Hz stimuli. The values of these thresholds are consistent with those of previous study \[[@B10]\]. These observations suggest that transcutaneous stimuli at frequency of 250 and 5 Hz induce pain sensation, and the 2000 Hz stimulation may also induce pain at higher intensity. From the analysis in C and Aδ fibers, which predominantly convey noxious information, we found that 5 Hz stimuli activated both C and Aδ fibers, 250 Hz stimulation activated only Aδ at low intensity (\~0.8 mA). Interestingly, Aδ fibers were also activated by 2000 Hz stimulation, although 2000 Hz stimulation is thought to activate Aβ fibers. However, the activation of Aδ fibers by 2000 Hz was observed only at high stimulus intensity (\> 2 mA). In in vivoanalysis of EPSCs evoked in SG neurons, which receive monosynaptic inputs predominantly from C and Aδ fibers \[[@B18]-[@B20]\], we found that 2000, 250 and 5 Hz stimuli significantly increased the frequency of evoked EPSCs at their behavioral thresholds. Therefore, these results indicate that 250 and 5 Hz stimuli, produce nociceptive responses by activation of Aδ fibers and by activation of C and Aδ fibers, respectively. On the other hand, 2000 Hz stimulation at high intensity could activate Aδ fibers and thus initiate noxious sensation. There was no significant difference in the frequencies of EPSCs evoked by 250, 5 and 2000 Hz stimuli at the behavioral threshold (Figure [5D](#F5){ref-type="fig"}). In the previous behavioral pharmacological study, topical application of capsaicin to the skin, a specific desensitizer of small fibers \[[@B25]\], increased behavioral thresholds for 250 and 5 Hz but not 2000 Hz stimuli, indicating that 250 and 5 Hz stimuli activate small fibers \[[@B10]\]. This observation is consistent with our results. Discharge ability of sensory fibers ----------------------------------- We observed that C fibers were activated only by 5 Hz stimulation with a maximum frequency of \~3 Hz, Aδ fibers were activated by 250 and 5 Hz (maximum frequency of \~22 Hz with 250 Hz stimulation) and Aβ fibers were activated by all three stimuli (maximum frequency of \~140 Hz with 2000 Hz stimulation, Figures [1](#F1){ref-type="fig"},[2](#F2){ref-type="fig"},[3](#F3){ref-type="fig"}). One of the most important factors for affecting the maximum frequency of sensory fibers is the inactivation of voltage-dependent sodium channels. In other words, discharge frequency of fibers is controlled by the time taken to recover the sodium channels from their inactivation state. Recent studies have shown that TTX-sensitive sodium channels expressed in large diameter DRG neurons exhibit faster recovery from the inactivation than the TTX-resistant ones expressed in small DRG neurons \[[@B26]-[@B29]\]. Therefore, it can be speculated that high frequency sine-wave stimuli are unable to initiate APs in small C fibers due to slow repriming of TTX-resistant sodium channels. Passive membrane property is also an important factor for the ability of neurons to discharge. Previous report has shown that the membrane properties of the three subgroups of rat DRG neurons were varied \[[@B17]\]. In particular, the membrane time constant of C neurons (6.5 ms) is much longer than those of Aδ (2.6 ms) and Aβ (1.8 ms) neurons. On the other hand, the durations of single sine-wave stimuli at frequencies of 2000 Hz and 250 Hz are 0.5 ms and 4 ms, respectively. The membrane time constant is the time required for the voltage change across the membrane to reach \~63% of its final value \[[@B30]\]. Therefore, the longer membrane time constant might contribute to the failure of AP in C neurons with 2000 and 250 Hz stimuli, since the membrane potential is slowly depolarized, resulting that the membrane potential can not reach AP threshold within the short depolarizing cycles with the high frequency stimuli as compared to 5 Hz one. In C fiber of rat DRG neurons, AP failure was increased when repetitive stimulation was applied at high frequencies \[[@B1]\]. Consistent with the lines of evidence, in the present study, there was no firing in most C fibers with 250 Hz and 2000 Hz stimuli even when the high intensity stimuli were applied as described in the results. Selectivity of specific frequencies of sine-wave stimuli to activate primary afferents -------------------------------------------------------------------------------------- ### 2000 Hz Stimulation We found that 2000 Hz stimulation selectively activated Aβ fibers at low intensity (\< 2 mA). Only at higher intensities (\> 2 mA), 2000 Hz stimulation produced APs in Aδ fibers. On the other hand, 2000 Hz stimulation increased EPSC frequency in SG neurons, which are known to receive a few Aβ inputs, at higher stimulus intensity. Therefore, it is conceivable that EPSCs in SG neurons by 2000 Hz stimulation at higher intensities are produced predominantly by activation of Aδ fibers. ### 250 Hz Stimulation 250 Hz stimulation produced APs in Aδ fibers (with a max discharge frequency of 22.3 ± 5.9 Hz) and Aβ fibers (with a max discharge frequency of 133.4 ± 13.3 Hz). These values of AP frequencies were consistent with those evoked by physiological stimuli \[[@B31]\]. These results suggest that 250 Hz stimulation initiates both nociceptive sensation via Aδ fibers and tactile sensation via Aβ fibers. However, in general when both painful and tactile sensations are produced simultaneously, the painful sensation would only be perceived. Therefore, 250 Hz is thought to be effective stimulus frequency for activation of Aδ fibers initiating noxious sensation. ### 5 Hz Stimulation The present study showed that C fibers were activated only by 5 Hz stimulation, but not by other stimulus frequencies. This indicates that 5 Hz stimulation is effective to activate C fibers among the three specific frequencies. Although both Aδ and Aβ fibers were also activated by 5 Hz stimulation, it is conceivable that the activation of Aδ or Aβ fibers by 5 Hz stimulation does not reach sufficient frequency to induce functional sensation. In the present study with 5 Hz stimulation, the maximum AP frequencies were \~3 Hz in C fibers, \~4 Hz in Aδ fibers and \~20 Hz in Aβ fibers (Figures [1](#F1){ref-type="fig"},[2](#F2){ref-type="fig"},[3](#F3){ref-type="fig"}). On the other hand, it has been reported that AP frequencies produced by natural physiological stimuli in rat DRG neurons are: \~5 Hz in C fibers by mechanical nociceptive stimuli, \~15 Hz in Aδ fibers by high threshold mechanical stimuli and \~100 Hz in Aβ fibers by slowly adapting tactile stimuli \[[@B31]\]. Although the AP frequency by 5 Hz stimulation in C fibers in our results was almost similar to the value of AP frequency in C fibers by natural mechanical noxious stimulation, the AP frequencies in Aδ and Aβ fibers by the 5 Hz stimulation were much less than those produced by natural stimuli. Furthermore, the number of C fibers in rat DRG neurons is quite larger than those of Aδ and Aβ fibers (C fiber, \~63%; Aδ fiber; \~25%; Aβ fiber, \~12%) \[[@B32]\]. Based on the present observations, the following analysis could be made (Figure [7](#F7){ref-type="fig"}). 2000 Hz sine-wave stimulation selectively activates Aβ afferents. This result could make it possible to analyze a change in threshold of Aβ afferent. If this stimulation protocol could produce pain sensation, it would be suggested that tactile stimuli produce pain, i.e. allodynia. 250 Hz stimulation activates both Aδ and Aβ fibers. In combination with 2000 Hz stimulation, a change of threshold in Aδ afferent could be examined. However, a substantial number of Aδ afferents convey non-noxious sensation, and therefore, it will be difficult to differentiate convincingly whether a change in sensitivity of Aδ afferent causes allodynia or hyperalgesia. Although, 5 Hz stimulation activates not only C but also Aδ and Aβ afferents, a change in C afferent threshold could be analyzed by excluding Aβ and Aδ afferent changes which can be performed by both 2000 and 250 Hz stimuli. ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Schematic diagram of activation of afferent fibers with sine-wave stimuli.The pie showed the number of three types of rat DRG neurons as reported by Djouhri et al \[32\]. The height represented normalized AP frequencies produced by sine-wave stimuli in the three types of fibers in our study relative to those by natural stimuli reported previously \[31\]. This figure shows the selectivity of the sine-wave stimuli to activate afferent fibers as described in \'Discussion\'. ::: ![](1744-8069-1-13-7) ::: In conclusion, we provided electrophysiological evidence that the sine-wave stimulation could differentiate changes in threshold of Aβ, Aδ and C afferent fibers by combining data obtained from 2000, 250 and 5 Hz stimulus protocol. This analysis using Neurometer would be useful for understanding not only changes in primary afferent fibers due to pathological conditions but also the evaluation of clinical treatment of analgesics. Methods ======= All experiments involving the in vivostudy were approved by the Committee of the Ethics on Animal Experiments of Kyushu University and conducted in accordance with the Guiding Principles of the Care and Use of Animals in the Field of Physiological Science of the Physiological Society of Japan. Animals and Behavioral test --------------------------- Male Sprague-Dawley rats (6--9-week old), weighing 250--300 g were used for behavioral tests, intracellular recordings and in vivopatch-clamp recordings. A skin patch dispersion electrode was fixed to attach the skin of left hind limb of rats. Transcutaneous nerve stimuli were applied through this electrode using the nociceptive mode of the Neurometer^®^(NUROTRON INCORPORATED, Baltimore, USA). In this mode, the intensity was automatically increased in step from 0 to 9.99 mA (29 steps for 5 Hz stimulation, 20 steps for 250 Hz and 2000 Hz stimuli). The durations were 2.5 sec for 5 Hz and 2.1 sec for 250 Hz and 0.72 sec for 2000 Hz stimuli. When rats escaped and/or vocalized, the stimulation was immediately stopped and the stimulus intensity was defined as behavioral threshold. Intracellular recording from DRG neurons ---------------------------------------- The methods used for the current experiment were similar to those described previously \[[@B1]\]. Under anesthesia with diethyl ether, a rat was decapitated and the lumbar laminectomy was performed. L4-6 DRGs with an attached proximal dorsal root were isolated from the animal. The isolated DRG was submerged in Krebs solution (in mM: NaCl 117, KCl 3.6, CaCl~2~2.5, MgCl~2~1.2, NaH~2~PO~4~1.2, NaHCO~3~2.5 and glucose 11) equilibrated with 95% O~2~-5% CO~2~and maintained at 36 ± 1°C and set in a recording chamber. Intracellular recordings of APs were made from DRG neurons with glass-microelectrodes having a DC tip resistance of 50--100 MΩ, filled with 4 M potassium-acetate. Signals were amplified with a high input-impedance bridge amplifier (Axoclamp 2B; Axon Instruments, Foster City CA, USA). Artifacts were minimized with low-pass (1000 Hz) or notch (250 Hz) filters using pCLAMP program (Axon Instruments). Data from neurons with resting membrane potentials less than -- 50 mV and AP amplitudes smaller than 60 mV were excluded in the present study. AP duration was determined at half of the peak amplitude of the AP. Antidromic stimulation (duration 100 μsec) was given to the central end of the dorsal root with a suction electrode. The stimulus intensity was monitored with a digitized output isolator (ss-202J; Nihon Kohden, Tokyo, Japan). The value of conduction velocity was calculated from the latency of AP and the length of the dorsal root retained. In vivopatch-clamp recording from SG in dorsal horn ----------------------------------------------------- The methods for in vivopatch-clamp recording from SG neurons have been described in detailed elsewhere \[[@B22],[@B24]\]. Briefly, under artificial ventilation, a lumbar laminectomy was performed at the level of L4--L5 and the animal was then placed in a stereotaxic apparatus (Model ST-7, Narishige, Japan). Under a binocular microscope with ×8 \~ × 40 magnification, the dura mater was cut and reflected, and then either L4 or L5 dorsal root was shifted laterally using a glass hook. The pia-arachnoid membrane was cut to make a window to allow the patch electrode into the spinal cord. The patch pipette had a tip resistance of 8--10 MΩ when filled with a solution of the following composition (in mM): Cs~2~SO~4~110, CaCl~2~0.5, MgCl~2~2, EGTA 5, HEPES 5, Mg-ATP 5, tetraethylammonium 5. Signals were acquired with an Axopatch 200B amplifier (Axon Instruments), low-pass-filtered at 5 kHz. Data were analyzed using Axograph program (Axon Instruments). The surface of the exposed spinal cord was perfused with Krebs solution equilibrated with 95% O~2~-5% CO~2~at 38 ± 0.5°C. Drugs were dissolved in Krebs solution and applied to the perfusing line. Whole-cell patch-clamp recordings could be obtained from in vivopreparations for more than 8 hours and stable recordings were made from single SG neurons for up to 2 hours. After the end of the experiments the animals were killed by exanguination. Statistical analysis -------------------- Data were presented as mean ± SEM. Statistical significance was determined as P\< 0.05 using Dunnett\'s multiple comparison. Acknowledgements ================ We thank Dr. K. Honda for technical advice with behavioral test, and Dr. J. Yagi and Dr. H. Inokuchi for useful advice on intracellular recording analysis. We also thank Asahikasei Pharma Corporation to provide a Neurometer. This study was supported by the Grants-in Aid for Scientific Research from the Ministry of Education, Science, Sports and Culture to M.Y. and H.F. and in part by a Grant-in-aid for the 21^st^Century COE Program of Japan.	PubMed Central	2024-06-05T03:55:55.754645	2005-3-25	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1083421/", "journal": "Mol Pain. 2005 Mar 25; 1:13", "authors": [ { "first": "Kohei", "last": "Koga" }, { "first": "Hidemasa", "last": "Furue" }, { "first": "Md Harunor", "last": "Rashid" }, { "first": "Atsushi", "last": "Takaki" }, { "first": "Toshihiko", "last": "Katafuchi" }, { "first": "Megumu", "last": "Yoshimura" } ] }
PMC1083422	Background ========== Melanomas are a heterogeneous group of tumors that arise in the skin, eye, meninges and other parts of the body. While early stages of cutaneous melanomas are recognized by changes in the size, shape or color of black nevi, most cancer cells grow downward from the skin and invade neighboring tissues before they are detected as highly metastatic tumors in lymph nodes or other organs \[[@B1]\]. Patients with malignant melanomas do not respond well to the currently available therapies and most treatments remain ineffective. To gain a better understanding of mechanisms involved in the growth, differentiation and transformation of cells, we have studied a fully differentiated (pigmented) highly malignant cat melanoma cell line CT1413, that trans-differentiates into neuronal cells, 48--72 hours after infection with the endogenous cat retrovirus RD114 in vitro\[[@B2],[@B3]\]. The trans-differentiated cells exhibit long multipolar cytoplasmic extensions that form a network of connections with giant multinucleated neuronal cells and smaller glial-like cells (Fig. [1A, B](#F1){ref-type="fig"} &[1C](#F1){ref-type="fig"}). This transformation is a highly specific event for CT1413 cells and its interaction with RD114 virus since no morphological change occurs when these cells are treated with various chemicals known to induce cell differentiation (dexamethasone, insulin, isobutyl-methyl-xathine) or they are infected separately with primate retroviruses that use the same neutral amino acid (NAA) transporter proteins as receptors as does RD114 virus \[[@B4],[@B3]\]. In addition, human, mouse and hamster melanoma cell lines infected with RD114 virus, or CT1413 cells infected with the feline or murine leukemia viruses also fail to exhibit any change in cellular morphology. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Photographs of the cat melanoma cell cultures before and after trans-differentiation.A:Uninfected control melanoma cells 48 hours after culture. B & C: Two different cultures of trans-differentiated neuronal cells, 48--72 hours after infection with the endogenous feline retrovirus RD114. ::: ![](1479-5876-3-14-1) ::: Since RD114 virus does not replicate in the terminally trans-differentiated neuronal cells, we hypothesized that binding of virus to its receptor on the melanoma cell membrane could induce conformational changes that generate a cascade of molecular interactions through distinct signals which arrest the tumor cell growth and drive them toward neurogenetic pathways. To test this hypothesis we have analyzed comprehensive protein profiles of both the malignant melanoma and its counterpart trans-differentiated neuronal cells by 2-dimensional gel electrophoresis (2DGE) and identified differentially expressed proteins in the two cell types by Matrix-Assisted-Laser-Desorption-Ionization-Time-Of-Flight mass spectrometry (MALDI-TOF-MS). Herein we show that the trans-differentiation of melanoma into neuronal cells is directly associated with de novoexpression of pro-inflammatory cytokines, neuro-regulatory enzymes/kinases, neurotrophic factors and concomitant suppression of growth-promoting proteins. This repertoire of proteins is not only responsible for the generation of neuronal cells but it is also involved in the reversion of a highly malignant tumor into non-cancerous neuronal cells. Results and Discussion ====================== We have compared 3129 protein spots in 15 gels derived from RD114-infected and uninfected melanoma cells from two independent experiments and analyzed peptide fingerprints of 467 differentially expressed (up-regulated and downregulated) protein spots by MALDI-TOF-MS. A total of 46 proteins were confirmed unambiguously from 302 spots excised from multiple gels of both experiments (Figure [2](#F2){ref-type="fig"}). The remaining 165 spots did not identify any protein from the SWISS-PROT database or proteins were not identified reproducibly from corresponding spots in different gels. Since species-specific isomers and protein-protein interactions in cat cells may be functionally different from those of human or other species, all 46 proteins reported here were confirmed from the feline protein database. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### 46 proteins identified by MALDI-TOF-MS from multiple gels derived both from melanoma and neuronal cells. X-axis= Protein names (abbreviations): Y-axis= frequency or \# of times a protein is detected and confirmed by mass spectrometry from multiple gels. Pinkbars represent proteins in neuronal cells and Bluebars indicate proteins present in melanoma cells. Protein names, abbreviations and accession numbers are from SWISS-PROT database used for protein identification (See Table 1-3 for details). One asterisk (\) indicates that the protein is detected in single, non-complexed form only in neuronal cells and two asterisks (\\) indicate that this protein is detected as single protein only in melanoma cells. A: From left to right, expression of 12 proteins detected exclusively in trans-differentiated neuronal cells; 3 proteins were detected in single (non-complexed) form only and are marked by (\) and 9 proteins were expressed as complexes with other proteins. B: Two proteins were detected in melanoma cells only (i.e. NOT detected in neuronal cells); CDNA was expressed as a single protein (\\) and MMO7 was present in a complex form. C:Expression frequencies of 32 proteins expressed both in melanoma and neuronal cells. Left side of the graph displays proteins that are detected at higher frequency in melanoma and the right side shows proteins that are detected at a higher frequency in trans-differentiated neuronal cells. ::: ![](1479-5876-3-14-2) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Proteins Expressed Exclusively in Neuronal Cells ::: Protein Name Abbreviation Accession \# Functional Category ----------------------------------------------------------- ------------------ ------------------ ----------------------------------- Acetylcholinesterase precursor ACES O62763 Membrane Enzyme Beta-2 adrenergic receptor B2AR Q9TST5 Membrane Receptor Beta-hexosaminidase beta chain precursor HEXB P49614 Lysosomal Enzyme Interleukin-1 receptor beta precursor IL1B P41687 Cytokine Receptor Interleukin-2 receptor alpha chain precursor IL2A\* P41690 Cytokine Receptor Interleukin-6 precursor IL6\* P41683 Cytokine Interferon gamma precursor ING P46402 Cytokine Macrophage colony stimulating factor I receptor precursor KFMS P13369 Membrane tyrosine kinase Receptor Major allergen I polypeptide chain 2 precursor FEL2 P30440 Cytoplasmic Cat Allergen Lipoprotein lipase precursor LIPL P55031 Membrane Enzyme Red-sensitve opsin OPSR O18913 Sensory Retinal protein Proto-oncogene serine/threonine-protein kinase pim-1 PIM1\* Q95LJ0 Serine Threonine Kinase B: Proteins Expressed Exclusively in Melanoma cells Protein Name Abbreviation Accession \# Functional Category Cyclin-dependent kinase inhibitor 1 CDNA\\ O19002 Nuclear Cell Cycle Protein Matrilysin precursor MMO7 P55032 Membrane Enzyme A:shows 12 proteins that were expressed exclusively in neuronal cells (ACES, B2AB, FEL-2, HEXB, IL1B, IL2A\, IL6\, ING, KFMS, LIPL, CPSR and PIM1). One asterisk (\) indicates that the protein is detected in single, non-complexed form in neuronal cells only. B:lists two proteins (CDNA\\* and MMO7) that were expressed exclusively in melanoma (i.e. not detected in trans-differentiated neuronal cells. Two asterisks (\\) indicate that the protein is detected in single, non-complexed form in melanoma cells only. ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Higher Detection Frequency in Neuronal Cells ::: Protein Name Abbreviation Accession \# Functional Category ------------------------------------------------------ ------------------ ------------------ ------------------------------------------------ Alkaline Phosphatase PPBT Q29486 Membrane Phosphatase Aminopeptidase N AMPN P79171 Membrane Enzyme Arylsulfatase B precursor ARSB P33727 Lysosomal Enzyme Brain-derived neurotrophic factor precursor BDNF\* Q9TST3 Neurotrophic Factor Cellular tumor suppressor p53 P53\* P41685 Nuclear Dipeptidyl peptidase IV DPP4 Q9N2I7 Membrane Enzyme Glutamate decarboxylase, 67 kDa isoform DCEI P14748 Cytoplasmic Enzyme Integrin beta-1 precursor ITB1 P53713 Membrane Receptor Interleukin-1 beta convertase precursor ICE1 Q9MZV6 Cytoplasmic Enzyme Lysosomal alpha-mannosidase precursor M2B1 O46432 Lysosomal Enzyme Major prion protein precursor PRIO O18754 Membrane Prion Protein Mast/stem cell growth factor receptor precursor KIT Q28889 Membrane Tyrosine Kinase Receptor NADH-Ubiquinone oxidoreductase chain 5 NU5M P48921 Mitochondrial Enzyme Polycomb complex protein BMI-1 BMI1 Q9TST0 Nuclear Protein Pyruvate kinase, M1 isozyme KPYM P11979 Pyruvate Kinase Serum albumin precursor\\\* ALBU\* P49064 Cytoplasmic Albumin Protein Sodium/calcium exchanger 1 precursor NAC1 P48767 Membrane Calcium Ion Channel (Plasma Membrane) T-cell surface glycoprotein CD8 beta chain precursor CD8B P79336 Membrane Receptor Toll-like receptor 4 precursor TLR4 P58727 Membrane Receptor Transferrin receptor protein 1 TFR1 Q9MYZ3 Membrane Receptor Zona pellucida sperm-binding protein 2 precursor ZP2\* P47984 Membrane Receptor Zona pellucida sperm-binding protein B precursor ZP4 P48834 Membrane Receptor Proteins with higher frequency of detection in complex forms in neuronal cells compared to melanoma. Ten proteins were detected in single as well as in complex forms in both melanoma and neuronal cells. These proteins displayed a higher frequency of detection in neuronal cells than in melanoma. Four proteins Fel d 2 /ALBU\, BDNF\, tumor suppressor protein (P53\) and zona pellucida-2 (ZP2\) were detected both in single (\) and complex forms only in neuronal cells. These proteins were present only in complexes in melanoma cells. The ALBU and P53\ proteins also showed increased frequency of detection in the neuronal cells than in melanoma (also see Fig 2C\) and BDNF and ZP2 showed similar detection frequencies in the two cell types. ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Lower Detetion Frequency in Neuronal Cells ::: Protein Name* Abbreviation Accession \# Functional Category -------------------------------------------------------------- ------------------ ------------------ ------------------------------------------------------ Beta-glucuronidase precursor BGLR O97524 Lysosomal Enzyme Cathepsin W precursor CATW Q9TST1 Lysosomal Enzyme Neurotrophin-3 precursor NT3 Q9TST2 Neurotrophic Favtor Phosducin PHOS P41686 Cytoplasmic Phosducin Platelet-derived growth factor, B chain precursor PDGB P12919 Growth Factor Proto-oncogene tyrosine-protein kinase FES P14238 Tyrosine Kinase (Cytoplasmic/Transmembrane) Sarcoplasmic/endoplasmic reticulum calcium ATPase 2 ATA2 Q00779 Membrane Calcium Ion Channel (Endoplasmic Reticulum) Serine -- pyruvate aminotransferase, mitochondrial precursor SPYA P41689 Mitochondrial Enzyme Tumor necrosis factor ligand superfamily member 5 TNF5 O97605 Cytokine Zona pellucida sperm-binding protein 3 precursor ZP3 P48832 Membrane Receptor Proteins that were detected at lower frequencies in neuronal cells (also see Fig 2 C) ::: By mass spectrometry each gel spot contained either a single protein per spot in a non-complexed form or as a complex of 2--8 proteins in one spot. The frequency of distribution for the 46 proteins among the 302 spots included 82 single protein spots (32 in melanoma and 50 in neuronal cells) and 220 spots containing protein complexes (103 in melanoma and 117 in neuronal cells. Among the differentially identified proteins, 12 were expressed de novoexclusively in neuronal cells, two were detected only in melanoma cells and 32 of 46 (\>69%) proteins were shared between the two cell types (Table [1](#T1){ref-type="table"}, [2](#T2){ref-type="table"}, [3](#T3){ref-type="table"} and Fig [2](#F2){ref-type="fig"}). Three of the 12 de novoexpressed proteins in neuronal cells were detected in single, non-complexed form (designated by \* in Table [1](#T1){ref-type="table"} and Fig. [2A](#F2){ref-type="fig"}) and 9 proteins were detected in complexes with other proteins (Fig. [2A](#F2){ref-type="fig"}). All newly synthesized or upregulated proteins belonged to families of cytokines, neuro-regulatory enzymes/ kinases, sensory, and other signaling proteins (Tables [1](#T1){ref-type="table"}, [2](#T2){ref-type="table"}, [3](#T3){ref-type="table"}). The two proteins that were expressed only in melanoma but not in any of the gels derived from neuronal cells were cyclin dependent kinase inhibitor 1 (P21-CDNA\\) expressed in a non complexed /single form and matrilysin or matrix metalloproteinase-7 (MMO7) (Fig. [2B](#F2){ref-type="fig"}; Table [1B](#T1){ref-type="table"}). These proteins have been shown to be critical for maintaining cell growth and metastasis respectively \[[@B5],[@B6]\]. The binomial probability for the distribution of the 12 newly expressed proteins in the trans-differentiated cells compared to melanoma was statistically significant (Fisher\'s exact test p = 0.007). Tables [2](#T2){ref-type="table"} and [3](#T3){ref-type="table"} show proteins that are expressed both in melanoma and neuronal cells. Of the 32 shared proteins in melanoma and trans-differentiated neuronal cells, 10 were detected in single as well as in complex forms, 21 proteins were present in complex forms and one protein, the T-cell surface glycoprotein CD8 beta chain (CD8B) was detected in a single form in both cell types but the complex form of this protein was present only in the neuronal cells (Tables [2](#T2){ref-type="table"} &[3](#T3){ref-type="table"} ; Fig. [2C](#F2){ref-type="fig"}). Each of the 32 proteins derived from both melanoma and neuronal cells have been compared and grouped according to their mean/-normalized quantities (i.e. upregulated or down regulated) or according to their higher or lower frequency of detection (i.e. \# of times each protein is detected in multiple gel spots). However, no functional significance has been implied in grouping these proteins (Tables [1](#T1){ref-type="table"}, [2](#T2){ref-type="table"},[3](#T3){ref-type="table"}). Analyses of known physiological functions for each of the 46 proteins indicated that 44 of 46 proteins (\>95%) expressed in neuronal cells have been shown previously to be associated with the early stages of brain development, differentiation and regulation of nervous system. Based on these biological functions, we have proposed possible roles of these 46 proteins in signaling antiviral responses, arrest in tumor cell growth, modulation of cell membranes and in altering the signal transduction pathways from malignancy to early neurogenesis and cell differentiation (Figure [3](#F3){ref-type="fig"}). ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Tentative Signal Transduction Pathways involved in suppression of tumorigenesis and induction of neurogenetic pathways. Proposed protein functions are based on previously published reports. ::: ![](1479-5876-3-14-3) ::: Antiviral Responses and Arrest in Tumor Growth ---------------------------------------------- The first line of defense against the RD114 virus infection begins with conformational changes in the melanoma cell membrane due to binding of the viral envelope glycoproteins to its receptor, the NAA transporter protein \[[@B4]\]. The presence of metalloprotease /aminopeptidase (AMPN) on the melanoma cell surface can increase heterodimer formation as it cleaves NAAs from the N-terminus of proteins \[[@B7]\]. The toll-like receptor-4 (Tl-4) recognizes pattern changes in membrane proteins and triggers the innate immunity while the antiviral responses are activated by de novoexpression of interferon gamma (ING) in the presence of tumor necrosis factor (TNF5) and interleukin-1 beta converting enzyme (ICE1) that were already present in melanoma cells (Fig. [2A](#F2){ref-type="fig"}&[2C](#F2){ref-type="fig"}). This activates de novosynthesis of IL-1 receptor beta (IL-1B), IL-2 receptor alpha chain (IL2A) and IL-6, all of which produce anti-tumorigenic and inflammatory responses. Co-expression of ING, IL-6 and other newly synthesized cytokines prompts proteases (AMPN, ICE1 and dipeptidyl peptidase VI (DPP IV/CD26) to functionally activate these proteins by cleavage \[[@B7]\]. A series of interactions are then initiated by the activated IL-1B and IL-6 toward the suppression of metastatic activity and arrest of melanoma cell growth \[[@B8],[@B9]\]. At this time, two other proteins CDNA/ p21 and MM07 that are essential for tumor growth and metastasis respectively are downregulated and are not detected in trans-differentiated neuronal cells i.e. after the exposure of cat melanoma cells to RD114 virus (Fig. [2B](#F2){ref-type="fig"}). Suppression of these proteins was associated with concurrent expression of P53 tumor suppressor protein as a distinct single protein in neuronal cells indicating its inhibitory effect on the tumor growth (Fig [2C](#F2){ref-type="fig"} \). Thus, the antiviral responses of the newly induced cytokines and growth regulatory proteins affect the arrest of tumor cell growth, which then trigger production of neurotrophic factors that alter the pathways responsible for modulation of neurogenesis. Activation of kinases, neuro-regulatory enzymes and neurotrophic factors ------------------------------------------------------------------------ Both IL-1B and IL-6 are multipotential cytokines capable of activating kinases necessary for phosphorylation of factors involved in the transcription of new genes, synthesis of novel proteins, up-regulation or down-regulation of many existing proteins that enhance signals for tumor suppression and neurogenesis. The cytoplasmic kinase of IL-6 acts synergistically in stimulating production of other kinases and receptors in a ligand-independent manner \[[@B10]\]. In addition, the Pim1 serine/ threonine kinase is expressed as a single non-complexed protein exclusively in neuronal cells (Fig. [2A](#F2){ref-type="fig"} \; Table [1](#T1){ref-type="table"} group A). This kinase phosphorylates proteins in dendritic and nuclear compartments of stimulated neurons and it is required for long-term potentiation of these cells \[[@B11]\]. The red-sensitive opsin receptor (OPSR) that is normally found in retinal rods is expressed de novoin transdifferentiated neuronal cells after exposure of melanoma cells to the RD114 virus (Fig. [2A](#F2){ref-type="fig"}). This integral membrane protein is phosphorylated on most of its serine and threonine residues present at its C-terminus \[[@B12]\]. It is possible that proteins that are phosphorylated at the serine or threonine residues may also be phosphorylated at tyrosine residues at acidic pH and these may generate distinct neurogenerative pathways \[[@B8]\]. While both melanoma and neuronal cells expressed the mast/stem cell growth factor receptor tyrosine kinase Kit (C-Kit), the frequency of detection for this kinase was significantly higher in neuronal cells compared to melanoma (Fig. [2C](#F2){ref-type="fig"}). Co-expression of c-Kit with IL-6, IL-2A and IL-B in trans-differentiated cells can further retard tumor growth while promoting cell signaling, differentiation and potentiation of synapses in conjunction with other proteins . The membrane tyrosine kinase receptor (KFMS) for the macrophage colony stimulating factor-1 (CSF-1) was expressed exclusively in neuronal cells and this protein has been shown to protect neural cells from degeneration \[[@B13],[@B14]\]. The FES tyrosine protein kinase was already present in melanoma cells at the time of trans-differentiation and this proto-oncogene has been implicated in morphological differentiation of neuronal cells\[[@B15]\]. The signals produced by IL6 and other cytokines activate brain-derived neurotrophic factor (BDNF\), a protein critical for intracellular protein-protein interactions that result in the development, differentiation, plasticity, and regeneration of neuronal network in the brain \[[@B15]\]. BDNF has recently been shown to be critical for neuronal self-repair following ischemia due to stroke and it contributes to the functional recovery of neurons \[[@B16]-[@B18]\]. BDNF binds to tyrosine kinase receptor with a high affinity \[[@B19]\]. It can also auto-phosphorylate and activate other signaling molecules. In our model system BDNF was over-expressed in trans-differentiated cells and it was detected in gel spots both as a single protein and in the form of complexes (Table [2](#T2){ref-type="table"}: Figure [2C](#F2){ref-type="fig"}). Concurrent expression of several kinases, cytokines and growth factors in neuronal cells appears to have enhanced a network of protein-protein interactions between different proteins present in the cell. This is evident by the detection of a large number of detergent-soluble complexes of proteins in these cells (Fig. [2A,B](#F2){ref-type="fig"}&[2C](#F2){ref-type="fig"}). Analyses of all complexed proteins containing BDNF (Fig. [4A](#F4){ref-type="fig"}&[4B](#F4){ref-type="fig"}) indicated that 4 of 5 complexes (80%) in neuronal cells and only 1 of 7 complexes (14%) in melanoma interacted with tyrosine kinase Kit (p= 0.072 based on 2-sided Fisher\'s exact test). It is interesting to note that 1 of the 5 BDNF complexes in neuronal cells that did not contain Kit (not shown in Fig. [4](#F4){ref-type="fig"}), was found to be associated with ZP3, PDGB and TL-4 which acts as an adapter-like protein for BDNF indicating that it was autophosphorylated in the complex without the Kit. This interaction not only stimulates BDNF activity but together these proteins increase length of neurites \[[@B20],[@B21]\]. Although the biological significance of all the recovered complexes from melanoma and neuronal cells is not clear yet, our data suggest that most of the detergent soluble complexes that we have isolated represent functionally active rather than randomly aggregated proteins. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Protein complexes as identified by mass spectrometry in single spots. All proteins identified in one complex are drawn by a single color. Arepresents 4 of the 5 complexes found in trans-differentiated neuronal cells. Each complex contains brain-derived neurotrophic factor (BDNF\) and its interacting partner proteins. Each color in the diagram represents one complex of BDNF with other interacting proteins as identified by mass spectrometry. NOTEthat all BDNF-interactive proteins in the 4 complexes in neuronal cells contain Kit tyrosine kinase. Bdisplays proteins in 5 complexes derived from melanoma cells. NOTEthat only one complex in melanoma cells contains Kit with no other protein in this complex. ::: ![](1479-5876-3-14-4) ::: Production of BDNF stimulates expression of acetylcholinesterase (ACES), an enzyme vital for neurite formation and transmission of signals in both the sympathetic and parasympathetic arms of the nervous system \[[@B22]\]. Expression of ACES in glial cells has been associated with the induction of prion (PRIO) protein present in most neural cells \[[@B23]\]. However, since PRIO was already present in melanoma cells and ACES was expressed only after RD114 virus infection, it indicates that ACES does not induce PRIO in these neuronal cells (Fig. [2A](#F2){ref-type="fig"}&[2C](#F2){ref-type="fig"}). BDNF also activates beta-2 adrenergic receptor (B2AR) in neuronal cells and promotes protein-protein interactions by enhanced binding of different kinases, which trigger astrogliosis and neuronal cell survival \[[@B24]\]. Both beta-hexosaminidase (HEXB) and lipoprotein lipase (LIPL) were expressed de novoin trans-differentiated cells (Fig. [2A](#F2){ref-type="fig"}). These enzymes not only retard cancer cell growth but are also critical for synaptic remodeling in the brain and affect differentiation of neuronal cells respectively \[[@B25]-[@B27]\]. The tissue nonspecific isozyme of alkaline phosphatase (PPBT) was expressed with high frequency. This neural cell marker binds to prion protein and it enhances neurotransmission and developmental plasticity \[[@B28]\]. Although PPBT was expressed in both melanoma and neuronal cells its frequency of detection was significantly higher in neuronal cells compared to the tumor cells (Fig. [2 C](#F2){ref-type="fig"}). Another important enzyme, arylsulfatase (ARSB) that desulfates proteins post-translationally and renders them functional was more frequently expressed in neuronal cells compared to melanoma. This enzyme is important for new glial cells and developing neurons in the brain rather than in mature cells \[[@B29]\]. The pyruvate kinase isoenzyme M1 (KPYM) was downregulated in trans-differentiated cells but its presence is considered important for nerve endings \[[@B30]\]. Likewise the down regulation of neurotrophic factor-3 (NT-3) in neuronal cells is noteworthy since this factor is necessary for the development of mature neurons and enteric nervous system but it is not essential for the generation of newly differentiated neural cells \[[@B31]\]. However, presence of NT-3 has been reported to be essential for enhancement of invasive potential of these tumor cells \[[@B17]\]. Glutamate decarboxylase 67 (DCE1) is a key regulatory enzyme for the neurogenic activity induced by neurotrophic factors \[[@B32]\] and the polycomb complex protein BMI1 is required for stem cell renewal \[[@B33]\] together with Kit, PDGB, NT-3, BDNF and others. One of the most frequently detected cell membrane proteins, in both melanoma and neuronal cells was integrin VLA-4 beta I subunit (ITB/ CD29). This protein was present in both single and complex forms in both cell types (Fig. [2C](#F2){ref-type="fig"}). Integrins are cell-surface receptors with multifunctional domains that bind to numerous ligands and transduce signals through a cascade of intracellular events that help in the development of laminae in the central nervous system \[[@B34]\]. In particular, Kit and beta-1 integrins collaborate in modulating cellular functions \[[@B35]\]. During neurogenesis, the synthesis and mobility of integrin is increased by the presence of both BDNF and PDGB \[[@B36],[@B37]\]. These proteins also redistribute transferrin (TFR1) to the membranes of immature neurons, a process essential for neurite formation, axonal cell growth and for regulation of IL-1B production \[[@B38],[@B39]\]. The sarcoplasmic/endoplasmic reticulum calcium ATPase 2 (ATA2) enzyme was detected with a high frequency in both melanoma and neuronal cells and sodium/calcium exchanger 1 (NAC1) a plasma membrane pump was expressed at a slightly higher frequency in neuronal cells than in melanoma (Fig. [2C](#F2){ref-type="fig"}). Both proteins are essential for neural development, regulation of intracellular concentration of calcium and maintenance of plasticity and synapses, \[[@B40]\]. Expression of different isomers of zona pellucida (ZP2, ZP3 and ZP4) in both melanoma and neuronal cells was totally unanticipated. Although exact function of these sperm-binding receptor proteins in melanoma or neuronal cells is not known, recent reports suggest that ZP domains particularly those of ZP2 that is upregulated in neuronal cells may transduce intracellular signals necessary for polymerization of proteins into neural filaments required for mechano-sensory dendrites \[[@B41],[@B42]\]. It is also possible that ZP proteins may participate in the formation of multipolar extensions in trans-differentiated cells during early neurogenesis. Phosducin (PHOS) is an abundant sensory protein and opsin, is a red-sensitive protein expressed in retinal rods, synapses and photoreceptor cells \[[@B43]\]. Expression of both of these neuro-sensory proteins may be involved in transduction of neurogenic signal in trans-differentiated cells. Conclusion ========== We have shown that RD114 virus infection of cat melanoma cells induces a repertoire of novel proteins that suppress its tumorigenic potential and promote neuronal cell differentiation in vitro. To our knowledge, this is the first demonstration that epigenetic stimuli from the cell surface of a naturally occurring melanoma can generate neurogenetic signals that can alter the genome-wide transcriptional and translational programs such that its oncogenic pathways are halted and pro-neuronal auto-regulatory mechanisms are prompted toward neural cell differentiation. By immunohistochemistry numerous markers of neural cell neoplasms (S-100 protein, neurofilament, epithelial membrane antigen, Leu-7 (CD57), neural specific enolase, neuro-peptide substance P, the low-affinity nerve growth factor receptor (p75NGFR), neural cell adhesion molecule (CD56/N-CAM) and growth-associated phosphoprotein-43) have been shown to enhance malignant potential of melanomas and neuronal cell-derived tumors \[[@B44]-[@B46]\]. In addition, peripherin, an intermediate filament protein is expressed during the transition of a neuronal cell to a Schwann cell phenotype and during the normal maturation of epithelioid melanocytes or melanocytic nevi \[[@B47]\]. Based on the immunohistochemical staining, many different tumor types such as neurotropic or desmoplastic melanomas, neuroepithelial and malignant granular cell tumors have also been grouped together as malignant peripheral nerve sheath tumors \[[@B48]\]. None of these proteins was detected in the melanoma or trans-differentiated neuronal cells that we have studied. This could be because we have primarily focused on proteins and profiles (upregulated, down regulated or de novoexpressed) that distinguish highly malignant melanoma cells from their counterpart trans-differentiated neuronal cells. Our proteomics studies also emphasize that although a significant number of the expressed proteins are shared between the two cell types, distinct protein-protein interactions are operative in altering signals that lead cells toward tumorigenic or neurogenic pathways. Expression of \>69% of germ-line associated sensory or neurogenic signaling proteins, neural-specific enzymes, cytokines, neurotrophic/ growth factors, serine, threonine and tyrosine kinases in cat melanoma cells as well as in trans-differentiated counterpart cells, supports the stem cell origin of this tumor and suggests that the precursor of this melanoma is a stem-like cell rather than the primitive melanoblast committed to melanocytic differentiation only. These cells are unique in that they are fully differentiated (pigmented) and they have the ability to give rise to new cell types that express differentiation-specific proteins. In addition, these cells express many proteins involved in self-renewal of cells. These include the stem cell growth factor receptor protein (C-Kit), PDGB, BDNF, NT-3, polycomb protein 1 and others. Recently, BMI1 has been shown to be essential for efficient self-renewal of neural stem cells \[[@B34],[@B49]\]. As can be seen in Fig [2C](#F2){ref-type="fig"}, BMI 1 and other protein involved in self-renewal are expressed both in melanoma and trans-differentiated neuronal cells but these are upregulated in neuronal cells. Since both melanocytes and neuronal cells arise from the neural crest of the embryo our studies indicate that CT1413 melanoma is a stem cell tumor. Based on our findings it can also be speculated that a high incidence of cutaneous melanomas and atypical melanocytic nevi in patients with hereditary neurofibroma, astrocytoma, glioma, meningioma, and other neuronal tumors \[[@B50]\], may also arise due to the expression or suppression of specific embryonic stem cell proteins in these cells. Correlation of cellular phenotypes with signature patterns of proteins in response to epigenetic or environmental changes would now be critical in establishing molecular definitions for classifying clinically related tumors. To gain a better insight into the complex processes of oncogenesis or neurogenesis, it would also be important to identify proteins in various natural biological settings since most substrate proteins are cleaved by proteases prior to enzyme interactions. Defining the specificity of natural targets may throw light on how biological specificities are achieved in nature. This model system offers a unique opportunity to study protein profiles in relation to specific ligand-binding interactions that are capable of inducing genetic or phenotypic changes in these cells. Characterization of protein-protein interaction domains by the use of constitutively active and dominantly negative constructs and screening of small peptides or interfering RNAs would provide new knowledge into functional significance of each of the proteins that direct cells toward distinct pathways (i.e. oncogenic versus neurogenic). These studies would also have important clinical implications for understanding neurogenerative and neuroprotective mechanisms for neurologic disorders since activated proteins may induce changes in protein subunits in a cell-type specific manner. This information would help in identifying selective inhibitors for tumors and activators/co-activators of neurogenic proteins involved in neural cell development. Materials and Methods ===================== Cell Cultures ------------- A cat melanoma cell line (CT1413) was established in vitrofrom a highly metastatic and melanotic tumor of the lung \[[@B2]\]. Cells were grown in minimal essential medium (MEM) with 10% fetal bovine serum and 2 mM glutamine (growth medium) and were plated in 36 flasks (75 cm^2^), at a density of 2 × 10^6^cells/ flask in medium containing 2-ug/ml polybrene \[[@B2]\]. After 24 hrs, culture medium was removed and fresh growth medium containing RD114 virus was added. RD114 is an endogenous retrovirus originally derived from the brain of a young cat and grown in human rhabdomyosarcoma cells \[[@B51]\]. This virus replicates in a wide range of different mammalian cells and does not induce any pathology in healthy cats or in cultures derived from cat embryo cells in vitro\[[@B52]\]. The RD114 virus was added to 12 flasks with a multiplicity of infection (MOI) of 1 and another set of 12 cultures was exposed to MOI of 0.25. An additional set of 12 flasks of melanoma cells was cultured without the virus as uninfected controls. After 48 hours post-virus infections, cultures exposed to MOI of 1 showed neuronal cell morphology and this experiment was highly reproducible \[[@B3]\]. To validate protein profiles two independent cell culture experiments were conducted almost 12 months apart and proteomes of both cell types were analyzed separately for differential expression profiles. Proteomics Studies ------------------ Proteins were extracted from 12 flasks each of the trans-differentiated and counterpart melanoma cells from each of the two experiments. Cells were washed 3× with phosphate buffered saline (PBS) to remove serum and other proteins, and 2 × 10^7^cells from each set of experimental and control melanoma cells were removed by the use of a cell-scraper or a \"rubber policeman\" (i.e. not treated with trypsin to detach cells). Each cell pellet was washed again 2× with PBS and proteins were extracted sequentially by a modified two-step solubilization procedure using two reagents containing different concentrations of 5--8 M Urea, 2%--4% CHAPS, 2 M thiourea and other non-ionic and/zwitterionic detergents (BioRad Sequential extraction kit CA, USA). The most soluble membrane proteins were removed in the first extraction and the less soluble proteins were separated in the second fraction. In order to isolate proteins as close to their natural states in cells as possible, each solubilization step was modified by rapid lysis technique (only 10--15 seconds\' cell lysis with 2-seconds\' sonication) followed by centrifugation for 90 minutes at 100,000 × g prior to analyzing soluble proteins present in the supernatant by isoelectric focusing (IEF). All protein fractions from melanoma controls and trans-differentiated neuronal cells were analyzed separately by 2-dimensional gel electrophoresis (2DGE). In the first dimension proteins were separated by IEF on a pH gradient 3--10 and then size fractionated in the second dimension on 6--18% gradient of SDS-polyacrylamide gels. Proteins were stained with Coomassie blue and by the use of a CCD camera and an image-processing analytical program (PDQuest from BioRad) all protein spots were evaluated from each of the 15 gels (9 gels from the first experiment and 6 gels from the second experiments). This program compares both the quality and normalized quantity of each spot across 15 gels and creates a master gel-image of well-calibrated and quantifiable spots using internal reference proteins. A master gel containing 3,129 spots was used to compare each of the corresponding spots in all gels derived from melanoma controls and trans-differentiated neuronal cell. All differentially expressed (i.e. upregulated and down-regulated) proteins were identified in each gel and 467 spots including some spots common to both cell types were excised from multiple gels. Proteins were digested using ultra-pure trypsin and peptide fingerprints of each in-gel digest were analyzed by MALDI-TOF-MS. Full scan mass spectra were acquired and high mass accuracy criteria were used throughout the study with high resolution and strict trypsin specificity. Spectra were submitted to the Protein database (SWISS-PROT) for searching protein identification. In addition to using these stringent criteria, we used manual acquisitions of spectra, which yielded more reliable and reproducible results compared to automated acquisitions using PS1 software. The confidence level in our protein identification was high because almost all proteins were confirmed in corresponding spots in multiple gels and by duplicating the entire experiment for validation. In this study we included only those proteins that were most reproducible in the high stringency Feline Protein database and results were confirmed in other mammalian species (SWISS-PROT). Thus far we have identified 46 proteins among 302 spots from multiple gels tested. All spots in which proteins were not identified reproducibly by mass spectrometry from the same or different gels were NOT included in any analysis. Statistical Analyses -------------------- The frequency distribution between melanoma and neuronal cells was compared using Chi-square test or 2-sided Fisher\'s exact test. The mean quantity for each protein was compared between melanoma and neuronal cells using the Mann-Whitney rank-sum test and the probability of a protein occurrence in the trans-differentiated neuronal versus melanoma cells was examined based on the binomial probability distribution with an expected proportion less than 0.05. Competing Interests =================== The author(s) declare that they have no competing interests. Acknowledgements ================ We thank Drs. S. Martin and J-P. Faulon of Sandia National Laboratories, Albuquerque, NM, for reading the manuscript and their enthusiasm about this research; summer students of 2003--2204 (Steve Doo, Chrissy Hsiao, Arti Ayer, Arvan Chan and Alex Lau, for their technical assistance and excitement while conducting experiments in vitro; David Wei and Arpan Patel for literature searches for protein functions and analyses of protein complexes; Vicky Moreno for assistance in preparing Figure [1A](#F1){ref-type="fig"}&[1B](#F1){ref-type="fig"} and Elizabeth Parra for preparing tables. The Rasheed Research and Endowment Funds at USC supported this research.	PubMed Central	2024-06-05T03:55:55.757458	2005-3-22	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1083422/", "journal": "J Transl Med. 2005 Mar 22; 3:14", "authors": [ { "first": "Suraiya", "last": "Rasheed" }, { "first": "Zisu", "last": "Mao" }, { "first": "Jane MC", "last": "Chan" }, { "first": "Linda S", "last": "Chan" } ] }
PMC1084244	Background ========== It has been well documented that immune functions declines with aging in both humans and experimental rodents. Especially the T cell dependent functions are compromised \[[@B1]-[@B3]\]. The immune functions are known to play an important role in host defense mechanisms \[[@B4]\]. Hence, it can be considered that enhancement of the immune functions contributes to the primary prevention of infectious illness, incidence of malignancy and autoimmune diseases in the elderly stage. It is well known that dietary factors play an important role in enhancement of health status and physically strength in human. Epidemiological data suggest that ingestion of some constituents from vegetables and fruits may contribute to a reduction in cancer incidence in humans \[[@B5],[@B6]\]. In experimental studies, several investigators reported that ingestion of extracts and several components including Ayurvedic food supplements from fruits and vegetables suppress carcinogenesis \[[@B7]-[@B9]\]. In the past, foods had been evaluated by both nutritional function (primary function) and gustatory function (secondary function). Recently, it has been shown that certain foods have a host defense function related to the immune system \[[@B10]-[@B12]\] and anti-oxidation \[[@B13],[@B14]\] and anti-tumor \[[@B15],[@B16]\] activity. The immune system plays an important role in physical and chemical carcinogenesis \[[@B17],[@B18]\] and in tumor-bearing hosts \[[@B19]\]. The role of the host immune function has become increasingly important in our understanding of the mechanisms that are involved in the body\'s ability to prevent cancer. Although the inter-relationship between diet, immune function and carcinogenesis is not clear, there is increasing evidence that dietary alteration of the host\'s immune functions is a key component of chemoprevention \[[@B20],[@B21]\]. Maharishi Amrit Kalash (MAK) 5, one of the Ayurvedic food supplements, belongs to a group of substances known as Rasayana \[[@B22]\]. MAK5 is a commercially available Rasayana that is composed of a variety of herbs, minerals and daily products \[[@B23]\]. MAK5 and other Rasayanas are believed to enhance the body\'s resistance to infections and disease, and enhance longevity \[[@B22]\]. Recently, few investigators have examined the effects of several Indian Ayurvedic products on chemically induced mammary tumors in rats \[[@B7],[@B23]\]. Vimal and his co-workers \[[@B24]\] reported that ingestion of MAK4 reduces Lewis Lung Carcinoma (LLC) metastases in mice. Such Indian Ayurvedic agents (MAK4 and MAK5) are also able to induce differentiation of several tumor cell lines \[[@B25],[@B26]\]. Although the physiological significance of the above findings is unknown, it might be speculated that Indian Ayurvedic products reduces certain cancer. One of the mechanisms by which these agents inhibit tumor metastasis and growth could be by enhancing macrophage and lymphocyte functions \[[@B23],[@B27],[@B28]\]. Since Ayurvedic drugs and food supplements contain a variety of herbs, minerals and dairy products, the presence of such compounds is certainly possible. Hence, to our knowledge, it is very difficult to elucidate which component has the prior effect compared to the other ingredients. In order to elucidate the mechanism(s) of anti-cancer effects of MAK5, the effects of MAK5 on macrophage and lymphocyte functions in mice were reported in our previous studies \[[@B11],[@B27]-[@B29]\]. We found that oral administration of MAK5 enhances phagocytic activity in the reticuloendothelial system, digestive and elimination activities of macrophage as primary stage of the host defense system, and also augments proliferative responses to Con A in young mice. However, less attention has been paid to the effects of MAK5 on immune function, such as macrophage and lymphocyte functions in old populations. Therefore, the purpose of this study was to determine if MAK5 can modulate macrophage and lymphocyte functions in aged mice. We tested this hypothesis by MAK5 treated mice for 2-month period. In the present study, we have investigated whether gastric intubation of MAK5 modifies (enhances) immune function, especially macrophage functions, splenic lymphocyte proliferation and cytokine production, in aged mice. Macrophage function was evaluated by measuring glucose consumption and production of NO by peritoneal macrophages as an indicator. Mitogenic response and production of IL-2, IFN-γ and IL-4 of spleen cells were examined for assessment of lymphocyte function. Materials ========= Chemicals --------- RPMI 1640 medium containing 10% heat-inactivated fetal bovine serum (FBS), 25 mM HEPES buffer, 100 units/ml penicillin, 0.1 mg/ml streptomycin, 1 mM L-glutamine and 0.1 M 2-mercaptoethanol was prepared by Gibco BRL (Gibco Laboratories Life Technologies, Inc., New York, USA); the pH was adjusted to 7.4 with NaHCO~3~. This medium was filtered sterilely with a 0.22-μm filter (Millex-GV; Millipore, Bedford, MA). Lipopolysaccharide (LPS, E. coli050: B5), concanavalin A (Con A) and other common chemicals for the NO~2~^-^and mitogenicity assay were purchased from Sigma (Sigma Chemical Co., St. Louis, USA). Animals ------- Specific pathogen-free (SPF) inbred male C3H/He N mice were used in the present study. Young mice (2 months old, weighing 24 to 26 g, n= 10) and old mice (22 months old, weighing 30 to 34 g, n = 40) were obtained from Japan SLC Inc. (Hamamatsu City, Japan). They were housed, five per cage, with pelleted basal diet, CE-2 (CLEA Japan, Inc., Tokyo, Japan) and water ad libitum, in an animal room under a 12 hours light-dark cycle at a temperature of 22 ± 1°C and a humidity of 60 ± 5%. After a week acclimation, they were used for the experiment. The old mice were divided into 4 groups. Each group consisted of 10 old mice. Ingredients of MAK 5, and treatment ----------------------------------- MAK5 were obtained from Maharishi Ayurveda Products International (Lancaster, USA). The ingredients of MAK5 have been described by Sharma et al.\[[@B7]\]. The ingredients in MAK5 are: gymnema aurentiacum (meda milkweed), black musale, heart-leaved moonseed, sphaeranthus indicus, butterfly pea, licorice, vanda spatulatum, elephant creeper and indian wild pepper. The exact composition of various ingredients in MAK 5 is not disclosed by the supplier, but the quality control (e. g., minimal variation from batch to batch) was assured. MAK5 suspended in distilled water was given to old mice p.o.at 50, 100 or 200 mg/kg per day (3 days/week) for 2-month. Due to the limited number of published researches on MAK, a standardized experimental protocol has not been established. We assumed that every day\'s MAK application would possibly cause an acute effect on the physiological functions of the animals, leading to death in a large number of aged mice. In order to prevent this possibility, the animals were given MAK5 food supplement every other day over the course of two months treatment. Old control mice and young control mice were given water as the vehicle (0.1 ml/10 g of body weight). In order to remove the acute effects of the treatment of MAK5, the animals were sacrificed by bleeding 24 hours after the last administration under ether anesthesia for the following experiments. The experimental protocol, animal care and treatment were approved by the Committee for Animal Studies at Gifu University School of Medicine. Isolation of macrophages ------------------------ All procedures were conducted under aseptic conditions. From each group 6 mice were selected for this part of experiment. Peritoneal exudate cells (PECs) were obtained from each mouse. Mice were sacrificed by bleeding under ether anesthesia, following a method previously described \[[@B30]\]. The abdomen was cleaned with 70% ethanol, the abdominal skin was carefully dissected without opening the peritoneum, and 5 ml of Hanks\' medium adjusted to pH 7.4 was injected intraperitoneally. The abdomen was massaged and 90 -- 95% of the injected volume was recovered. The peritoneal resting macrophages in the PECs suspension were isolated by the cell adhesion method. The PECs were suspended in RPMI 1640 medium containing 10% heat-inactivated FBS and incubated in a culture plate (Corning Laboratory Sciences Co., New York, USA) for 2 h at 37°C in a 5% CO~2~incubator. After removing non-adherent cells by washing the plate with Hanks\' medium, the adherent cells were harvested from the bottom using a cell scraper and resuspended in 10% FBS-RPMI 1640 medium. These cells were counted, checked for viability in trypan blue, and used in in vitroassays for glucose consumption and production of nitric oxide (NO~2~^-^). Preparation of splenic lymphocytes ---------------------------------- All procedures were conducted under aseptic conditions. For this part of the experiment, 4 mice were used from each group. Mice were sacrificed by bleeding under ether anesthesia and the single cell suspension was prepared by pressing the spleen between two slide glasses. The cell suspensions were passed through a 200-gauge stainless steel sieve and then let to stand to remove tissue fragments. Contaminating red blood cells were lysed by suspending cells in 0.85% NH~4~in Tris-HCl buffer. The cell suspensions were centrifuged (600 × g for 10 min) and then resuspended gently in FBS-RPMI 1640. These cells were counted, checked for viability in trypan blue, and used in in vitroassays for proliferative responses, production of IL-2, IFN-γ and IL-4. Assay of glucose consumption in peritoneal macrophages ------------------------------------------------------ Glucose consumption in peritoneal macrophages was determined by the method reported previously \[[@B29]\]. Briefly, the supernatants (20 μl) obtained from the macrophage culture sampled (1 × 10^5^cells/well) for 48 and 72 hours at 37°C were incubated with 3.0 ml of color reagent for 20 min at 37°C. The optical density at 505 nm of the solution was measured and the remaining glucose was determined from a calibration curve with standard glucose solution. The results were expressed as percent glucose consumption, calculated from the following formula: \[1 - (glucose content in culture medium cultured with macrophages / glucose content in culture medium without macrophages)\] × 100. Production of nitric oxide (NO~2~^-^) ------------------------------------- Macrophages (1 × 10^5^cells/well) were cultured at 37°C with 5% CO~2~in humidified air for 24 h with LPS (10 μg/ml). The accumulation of NO (as measured by the metabolite NO~2~^-^) in the culture supernatants was measured using the assay system described by Ding et al. \[[@B31]\]. Briefly, at the end of cell culture, 100 μl of supernatant were removed from each well onto empty the 96-well plate. After the addition of 100 μl Griess reagent (1:1, v/v, N-1-naphthylethylene diamine dihydrochloride 0.1% in H~2~O, sulfanilamide 1% in 5% H~3~PO~4~) to each well at room temperature for 10 min, the absorbance at 550 nm was measured using the microplate reader (Nalge Nunc International Co., Ltd., Immuno mino NJ-2300, Osaka, Japan). NO~2~^-^was determined by using sodium nitrite as a standard (range of 0 -- 100 μM). The samples were frozen and stored at -80°C until use. Mitogenicity assay ------------------ Assays were done in microplates (Corning Laboratory Sciences Co., New York, USA) as previously described \[[@B28]\]. Briefly, 2 × 10^5^ cells in 50 μl RPMI 1640 supplemented with 10% FBS were stimulated with optimal concentration of Con A (10 μg/ml). The plates were then placed in a 5% CO~2~incubator (37°C) for 72 hours. The proliferation of spleen cells was assayed using 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT). Six hours before the end of the incubation, 10 μl of 0.5% MTT dissolved in Ca^2+^and Mg^2+^-free phosphate buffered saline (PBS; pH 7.4) was added to the well. After the incubation, 150 μl of 0.04N HCl-isopropanol was added to each well. Then, the optical density at 570 nm was measured by a microtiter plate reader. The experiments were done in triplicate sets. The stimulation index (S. I.) was calculated using the following equation. S. I. = mean optical density of the cells stimulated with Con A / mean optical density of the cells not stimulated with Con A. This assay method has been reported to yield results similar to those obtained from the traditional ^3^H-thymidine incorporation method \[[@B32]\]. Production of Interleukin-2 (IL-2), interferon-gamma (IFN-γ) and interleukin-4 (IL-4) ------------------------------------------------------------------------------------- The 100 μl of spleen cell suspension at a concentration of 4 × 10^6^cells/ml were incubated at 37°C with 5% CO~2~in humidified air for 24 hours with concanavalin A (Con A) 5 μg/ml. IL-2, IFN-γ and IL-4 activities in culture supernatants were determined by using the ELISA kit (Endogen, Inc., Woburn, USA), respectively. The samples were frozen and stored at -80°C until use. The sensitivity of IL-2, IFN-γ and IL-4 assay were \<3 pg/ml, \<10 pg/ml and \<5 pg/ml, respectively. Statistical analysis -------------------- Results were presented as the means ± SE. Data were analyzed statistically using one-way analysis of variance (ANOVA) and post-hocScheffé\'s test for multiple comparisons. Significance levels were set at P \<0.05. Results ======= Body weight and food intake --------------------------- The body weights of the old control mice were 33.6 ± 1.0 g at 22 months of age and 32.3 ± 0.9 g at 24 months of age. The body weights of the young mice were 27.5 ± 0.2 g at 2 months of age and 29.9 ± 0.3 g at 4 months of age. The food intake of the old control mice was 4.1 ± 0.1 g/day/mouse. MAK5 administration scarcely affected the body weight gain and the food intake (data not shown). Glucose consumption capacity ---------------------------- The results of glucose consumption capacity in peritoneal macrophages are shown in Figure [1](#F1){ref-type="fig"}. Glucose consumption of peritoneal macrophages from old mice treated with MAK5 at all doses and incubated for 48 and 72 h were significantly greater than that in the control group (P\< 0.01 -- 0.001). Glucose consumption of peritoneal macrophages from young mice without treatment incubated for 48 and 72 h were significantly greater than those in the old mice treated with and without MAK5 (P\< 0.001). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Effects of Maharishi Amrit Kalash 5 (MAK5) on glucose consumption capacity of peritoneal macrophages from mice cultured for 48--72 h. Values are means ± SE. a\\ P\< 0.01, a\\\* P\< 0.001, with respect to old control mice; b\\\* P\< 0.001, with respect to young control mice. ::: ![](1472-6882-5-8-1) ::: Production of NO~2~^-^ ---------------------- The results of NO production (as measured by the metabolic NO~2~^-^) by peritoneal macrophages cultured for 24 h are illustrated in Figure [2](#F2){ref-type="fig"}. NO~2~^-^production of peritoneal macrophages stimulated by LPS in old mice treated with MAK5 at all doses was significantly greater (about two fold) than that in the old control group (P\< 0.001). NO~2~^-^production of peritoneal macrophages stimulated by LPS in the young control mice without treatment was significantly lower than that of the old mice treated with MAK5 (P\< 0.001). NO~2~^-^production of peritoneal macrophages stimulated by LPS in the old control mice was slightly higher than that in the young control mice; however, the difference was not statistically significant. MAK5 at any dose did not enhance spontaneous NO~2~^-^production by unstimulated peritoneal macrophages (data not shown). ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Effects of Maharishi Amrit Kalash 5 (MAK5) on nitric oxide (NO~2~^-^)production of peritoneal macrophages stimulated by LPS in mice. Macrophages from control (old and young) and MAK5 treated mice were incubated with LPS (10 μg/ml) for 24 h. The accumulation of NO~2~^-^in culture supernatants was measured by Griess reagent as described in METHOD. Values are means ± SE. a\\\* P\< 0.001, with respect to old control mice; b\\\* P\< 0.001, with respect to young control mice. ::: ![](1472-6882-5-8-2) ::: Splenocytes proliferative responses ----------------------------------- The results of Con A-stimulated splenocytes proliferative responses are shown in Figure [3](#F3){ref-type="fig"}. S.I. in old mice gavaged with MAK5 at all doses was significantly higher than that in the control group (P\< 0.01 -- 0.001). S.I. in young mice was significantly higher than those in the old mice treated with and without MAK5 (P\< 0.001). MAK5 at any dose did not enhance spontaneous proliferation of splenic lymphocyte without Con A stimulation in old mice (data not shown). ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Effects of Maharishi Amrit Kalash 5 (MAK5) on the Con A stimulated splenocyte proliferative response in mice. Splenic lymphocytes from control (old and young) and MAK5 treated mice were incubated with Con A (10 μg/ml) for 72 h. The proliferation of splenic lymphocytes was assayed using 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT). Values are means ± SE. a\\ P\< 0.01, a\\\* P\< 0.001, with respect to old control mice; b\\\* P\< 0.001, with respect to young control mice. ::: ![](1472-6882-5-8-3) ::: Production of IL-2 and IFN-γ ---------------------------- The results of assays of IL-2 and IFN-γ assays are indicated in Figures [4](#F4){ref-type="fig"} and [5](#F5){ref-type="fig"}. When spleen cells cultured without Con A, the production of IL-2 and IFN-γ was not detectable in the supernatant. However, cells cultured with Con A produced significant amounts of IL-2 and IFN-γ. The amounts of IL-2 stimulated by Con A in old mice given MAK5 at all doses were significantly high when compared with that in the control group (P\< 0.05 -- 0.001). The amounts of IFN-γ stimulated by Con A in old mice given MAK5 at doses of 100 mg/kg and 200 mg/kg were significantly high when compared with that in the control group (P\< 0.05). The amounts of IL-2 and IFN-γ stimulated by Con A in young mice without treatment were significantly high when compared with those in the old mice treated with and without MAK5 (P\< 0.001). ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Effects of Maharishi Amrit Kalash 5 (MAK5) on IL-2 production of splenic lymphocytes stimulated by Con A in mice. Splenic lymphocytes from control (old and young) and MAK5 treated mice were incubated with Con A (5 μg/ml) for 24 h. Production of IL-2 in culture supernatants was measured by ELISA system. Values are means ± SE. a\* P\< 0.05, a\\\* P\< 0.001, with respect to old control mice; b\\\* P\< 0.001, with respect to young control mice. ::: ![](1472-6882-5-8-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Effects of Maharishi Amrit Kalash 5 (MAK5) on IFN-γ production of splenic lymphocytes stimulated by Con A in mice. Splenic lymphocytes from control (old and young) and MAK5 treated mice were incubated with Con A (5 μg/ml) for 24 h. Production of IFN-γ in culture supernatants was measured by ELISA system. Values are means ± SE. a\* P\< 0.05, a\\\* P\< 0.001, with respect to old control mice; b\\\* P\< 0.001, with respect to young control mice. ::: ![](1472-6882-5-8-5) ::: IL-4 production --------------- The data on IL-4 assay are shown in Figure [6](#F6){ref-type="fig"}. As in the case of IL-2 and IFN-γ production in the spleen cells, MAK5 treatment at any dose did not enhance spontaneous IL-4 production by splenocytes in mice. The IL-4 production of splenic lymphocyte stimulated by Con A in old mice given MAK5 at dose levels of 50 and 200 mg/kg were significantly high when compared with that in the control group (P\< 0.05 -- 0.001). There were no significant differences in the IL-4 production of splenic lymphocyte stimulated by Con A between the old control mice and the young mice without treatment. ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Effects of Maharishi Amrit Kalash 5 (MAK5) on IL-4 production of splenic lymphocytes stimulated by Con A in mice. Splenic lymphocytes from control (old and young) and MAK5 treated mice were incubated with Con A (5 μg/ml) for 24 h. Production of IL-4 in culture supernatants was measured by ELISA system. Values are means ± SE. a\* P\< 0.05, a\\\* P\< 0.001, with respect to old control mice; b\* P\< 0.05, b\\\* P\< 0.001, with respect to young control mice. ::: ![](1472-6882-5-8-6) ::: Discussion ========== The results in the current study clearly indicate that oral administration of MAK5 effectively enhances both macrophage and lymphocyte functions in aged mice. A recent study suggested involvement of the immune response in chemically induced carcinogenesis \[[@B33]\]. In that study, the number of immune cells positive for dendritic cell and macrophage common markers was significantly greater in a N-methyl-N\'-nitro-N-nitrosoguanidine-resistant rat strain (Buffalo) than that in a carcinogen-sensitive rat strain (ACI). Thus, the findings in this study suggest that immunomodulation by MAK5 may be partially responsible for inhibiting chemically induced carcinogenesis. It is now known that the immunosuppression linked to aging is due to a decline in the response of lymphocytes, but not to a decline in the non-specific function of macrophages. Indeed, phagocyte function of macrophage could somewhat counteract the decreased specific immune response in old age \[[@B34]\]. Macrophages are known to play an important role in host defense mechanisms for protection from microbial invaders \[[@B4]\]. Macrophage glucose consumption increases as a result of the activation of macrophages \[[@B35]\]. In the present study, glucose consumption capacity of peritoneal macrophages from old mice treated with MAK5 at all doses and incubated for up to 72 h was significantly higher compared with the control group. However, glucose consumption of peritoneal macrophages from young mice without treatment incubated for 48 and 72 h were significantly greater than those in the old mice treated with MAK5. These results suggest that MAK5 activates peritoneal macrophages from old mice, though it was not as young mice. Since glucose consumption by macrophages is related to the pentose phosphate pathway of glycolysis \[[@B35]\], it is likely that MAK5 activates the pentose phosphate pathway in the peritoneal macrophages. Some researchers have suggested that NO is a cytotoxic effector molecule of macrophage-mediated tumorcidal actions \[[@B36],[@B37]\]. Murine peritoneal macrophages in culture synthesize significant quantities of NO in response to LPS \[[@B31],[@B38]\]. Dileepan et al.\[[@B23]\] reported that production of NO by LPS-activated macrophages from MAK5 treated mice (3 months old) was significantly higher than those from control. In the present study, we found that NO~2~^-^production of peritoneal macrophages stimulated by LPS in old mice treated with MAK5 at the doses of 50, 100 and 200 mg/kg were significantly increased (about two fold) compared with that in the old control group. Therefore, it is speculated that the immunosuppression linked to ageing is not due to a decline in the non-specific function of macrophages \[[@B34]\] and MAK5 ingestion enhances the stored capacity of macrophage function even in old mice. In this study, MAK5 treatment alone did not cause spontaneous elevation in NO~2~^-^production of peritoneal macrophages. These findings suggest that MAK5 treated causes priming of macrophages for enhanced sensitivity to other activating triggers, such as LPS and IFN-γ. Various cytokines such as IFN-γ, interleukin-2 (IL-2), and tumor necrosis factor-α either independently or in combination can prime macrophages for enhanced cytolysis \[[@B39]\]. It has generally been found that production of NO by macrophages is an important mediator of tumoricidal and microbicidal activity \[[@B40]\]. Other researchers reported that supplementation of a similar Ayurvedic herbal food supplement (Maharishi Amrit Kalash 4) in experimental animals rodents resulted in decreased incidence of tumor growth and proliferation and decrease metastases \[[@B7],[@B24]\]. The mechanism(s) in which antitumor activity in made to increase by MAK is not well documented yet. Although previous studies \[[@B7],[@B24]\] did not examine any immune functions, the ability of MAK 5 to induce NO~2~^-^production from macrophages may be one of the mechanisms of cancer chemoprevention. Lectins are known to possess mitogenic activity after binding to lectin receptors \[[@B41]\]. T lymphocyte mitogens such as Con A are thought to act through several subsequent steps, initially inducing IL-1 secretion in macrophages. Also, it is known that mitogenic activity, which reflects an early stage in the immune response, has been measured as a first screening of immunomodulatory activity \[[@B42]\]. We \[[@B28]\] reported that the stimulation index of spleen cells by Con A was increased significantly by the treatment with MAK5 at the doses of 50, 100 and 200 mg/kg in young mice. As shown in the present experiment, MAK5 at the doses of 50, 100 and 200 mg/kg exerted an augmentative effect on spleen cell proliferative responses to Con A in old mice, though it was not as young mice. It is accepted that cytokines are major factors involved in the regulation of the immune response to antigens and infectious agents. Recently, T helper cells are divided into Th1 and Th2 cells from the profiles of cytokine secretion \[[@B43]\]. It is known that Th1 cells are able to produce IL-2 and IFN-γ, whereas Th2 cells can produce IL-4 and IL-10. Th1 cells upregulates mainly cell-mediated immunity and downregulate humoral immunity, whereas Th2 cells act oppositely \[[@B44]\]. A Th1/Th2 imbalance is found in cancer patients \[[@B45],[@B46]\]. In this study, we observed that the amounts of IL-2 and IFN-γ, but not the amount of IL-4, stimulated by Con A in young mice without treatment were significantly high when compared with those in the old mice. These results support the hypothesis that the function of Th-1 cells declines in aged mice \[[@B47]\]. In the current study, MAK5 enhanced the production of IL-2 at doses of 50, 100 and 200 mg/kg, IFN-γ at doses of 100 and 200 mg/kg and IL-4 at doses of 50 and 200 mg/kg. We previously reported that the macrophage functions (glucose consumption, enzyme activity) and lymphocyte proliferation in young mice (10 weeks old) is significantly higher in groups treated with MAK at the doses of over 50 mg/kg than in controls \[[@B28],[@B29]\]. The result of the aged mice in the present study is in good agreement with the findings of the previous studies \[[@B28],[@B29]\]. In the present study, we could not demonstrate dose-response relationship by the applied MAK5 dosages (50, 100 and 200 mg/kg). Further research with different MAK5 dosages should be undertaken to possibly overcome this failure. Our results suggest that oral administration of MAK5 may affect the production of cytokines not only from Th1 cells, such as IL-2 and IFN-γ but also from Th2 cells, such as IL-4 in aged mice. Conclusion ========== The results described here may support the hypothesis that MAK5 directly activates macrophage activities in aged mice, whereas it only primes lymphocytes to display a greater immune response following interaction of splenic lymphocytes with another stimulus, such as Con A. Our findings suggest that MAK5 may contribute to the prevention of the immunosenescence. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= RI and HS carried out the study, wrote the paper, and provided overall coordination of the project. HS and SMM participated in data analysis. HS and SMM participated in study design. All authors read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1472-6882/5/8/prepub>	PubMed Central	2024-06-05T03:55:55.760962	2005-3-25	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084244/", "journal": "BMC Complement Altern Med. 2005 Mar 25; 5:8", "authors": [ { "first": "Ryoichi", "last": "Inaba" }, { "first": "Seyed Mohammad", "last": "Mirbod" }, { "first": "Haruo", "last": "Sugiura" } ] }
PMC1084245	Background ========== Peutz-Jeghers syndrome (PJS; MIM 175200) is a rare autosomal dominant disorder typified by hamartomatous polyposis of the gastrointestinal tract and melanin pigmentation of the oro-facial region \[[@B1]\]. Although germline mutations in the coding sequence of the serine-threonine kinase gene STK11/LKB1have been found to cause PJS \[[@B2],[@B3]\], such mutations only account for up to 80% of cases \[[@B4]-[@B12]\]. In addition to locus heterogeneity \[[@B13]\] mutations in regulatory sequences of STK11/LKB1may cause PJS. Identifying regulatory genomic sequences through functional assays is time consuming and problematic. As natural selection is more likely to tolerate sequence changes in redundant, non-functional sequence than in functionally important sequences, regulatory elements in non-coding sequence will be highly conserved through evolution. Comparison of sequence between both closely related and highly divergent species therefore allows for the identification of non-coding sequence that has a high probability of being important to the regulation of gene transcription. This alternative approach to the identification of promoter elements is termed \"phylogenetic foot printing\" \[[@B14]\]. Here we describe the phylogenetic foot printing of the 5\' upstream region of STK11/LKB1and the sequence analysis of this region in a series of PJS cases in whom exonic and splice site mutations in the gene had been excluded. Results and discussion ====================== Figure [1](#F1){ref-type="fig"} shows the evolutionary conserved region 5\' of the coding sequence of STK11as identified by the phylogenetic foot-printing programs ECR Browser \[[@B15]\] and CONSITE \[[@B16]\]. There was a high degree of concordance between the predictions from the programs. ECR Browser predicted three regions of high conservation upstream of STK11, encompassing nucleotides -981 to -1668 (nucleotide 0 representing the STK11translation initiation signal). Sixteen transcription factor binding sites (TFBS) predicted by rVista \[[@B17]\] resided between nucleotides -1053 and -1472. ConSite predicted 27 TFBSs conserved between human and mouse between nucleotides -1090 and -1605. Through the integration of phylogenetic foot printing and TFBS prediction data from ECR browser/ rVista and ConSite, a consensus region containing predicted TFBS positions was identified between nucleotides -1090 to -1472. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### in-silicoidentification of the putative STK11/LKB1promoter. (a) ECR browser output. Regions of high sequence conservation 5\' of STK11 exon 1 are annotated with transcription factor binding sites as predicted by rVista v2.0 between nucleotides -981 to -1668. (c) ConSite output. Sequence of high conservation is annotated with predicted transcription factor binding sites (TFBS) as predicted by the JASPER TFBS database between nucleotides -1104 to -1613. (b) Ideogram showing consensus conserved region, defined at the 3\' end by the E74A TFBS predicted ConSite, and at the 5\' end by the CMYB TFBS predicted by rVista. Arrows indicate the genomic sequence analysed. ::: ![](1471-2164-6-38-1) ::: Additional analysis using the TFBS prediction programs Transplorer (Biobase Biological Databases, Wolfenbüttel, Germany) and Proscan v1.7 \[[@B18]\] identified 9 binding sites between nucleotides -1142 and -1724, and 76 binding sites between nucleotides -881 and -1572 respectively, confirming the presence of regulatory elements within the consensus region. ECR Browser/ rVista and ConSite have previously been shown to correctly identify 88% and 66% of TFBS in functionally verified promoter elements respectively\[[@B15],[@B16]\] and it is highly likely, therefore, that the consensus region encompassing nucleotides -1090 to -1472, contains STK11/LKB1promoter elements. DNA from a series of 33 PJS patients that did not harbour germline STK11/LKB1mutations (mean age at diagnosis 14 years, range 1--38) was studied for mutations in the region between nucleotides -1001 to -1815, encompassing all TFBSs predicted by rVista and ConSite. Seven of the cases had a documented family history of PJS. The diagnosis of PJS in all cases was based on established criteria \[[@B1]\]. Three sequence changes were identified. The change G → C at position -1566 was found in four cases (three heterozygotes and one homozygote) and represented a previously documented single nucleotide polymorphism (rs3795061). An additional single nucleotide change at position -1268 (G → T) was identified in eight cases (seven heterozygotes and four homozygotes). A single sample of control DNA used as a sequence reference was also found to be homozygous for the change. Finally, a two base pair deletion coupled with a single base pair insertion at position -1709 (n-1709insTdelCC) was identified in one case. A series of healthy population controls (n = 92) was screened by High Performance Liquid Chromatography (HPLC) and one individual (1/92) was found to harbour the sequence change. None of these three sequence changes identified were therefore deemed to be potentially pathogenic. There is a high degree of redundancy in promoter elements of genes, however point mutations in promoter regions of PTEN and MLH1 have been reported to be disease causing \[[@B19],[@B20]\]. To investigate the possibility of large-scale deletions or insertions undetectable by straightforward PCR primers P1Fwd and P3Rev were used to amplify an 814 bp fragment with products visualised on a 2% agarose gel. No large-scale deletions or insertions were detected in any of the patients. Conclusion ========== As understanding of the contribution of coding sequence changes to disease becomes clearer, attention will focus on regulatory elements of genes. Phylogenetic foot printing using programs such as ECR browser and ConSite present potentially powerful tools in identifying regulatory elements, enabling the analysis of these sequences without time consuming functional studies. Although the efficiency of in-silico delineation of promoter elements has not been rigorously evaluated, ECR Browser/ rVista and ConSite have been shown to correctly identify 88% and 66% of TFBS respectively\[[@B15],[@B16]\]. On the basis of our findings, however, it appears unlikely (upper 95% confidence interval, prevalence; 9%) that mutations in the promoter region of STK11/LKB1are responsible for PJS cases not attributable to exonic sequence changes. Methods ======= Bioinformatics -------------- Phylogenetic foot printing of 3 Kb of sequence upstream of human and mouse STK11/LKB1(NT\_011255) was performed using the promoter predication programs ECR Browser \[[@B15]\] and CONSITE \[[@B16]\]. For both programs a conservation cut-off of 70% identity over 100 base pairs was adopted, in accordance with published criteria \[[@B21],[@B22]\]. Regions identified by ECR Browser with greater than 70% conservation were analysed using the TFBS prediction program rVista v2.0 \[[@B17]\]. Conserved regions identified by ConSite were annotated with TFBSs based on the JASPAR database \[[@B23]\] and links provide sequence information. Further analysis was carried out using the TFBS prediction programs Proscan version1.7 \[[@B18]\] and Transplorer (Biobase Biological Databases, Wolfenbüttel, Germany). The promoters and TFBS predicted by each program were aligned to delineate a consensus region indicative of a putative promoter. Mutational analysis ------------------- The possession of germline mutations in STK11/LKB1, including a large-scale deletion of the gene, was excluded using methods previously described \[[@B24]\]. Mutational analysis of the minimal consensus conserved region was carried out by direct DNA sequencing in both directions using the BigDye v3.1 Terminator Cycle Sequencing Ready Reaction Kit in conjunction with an ABI3100 semi-automated genetic analyser (Applied Biosystems, Foster City, USA). Overlapping oligonucleotides amplifying three fragments spanning the region were designed using the Primer3 program \[[@B25]\]: P1Fwd -- GCACAGGAGGGTTCAATATTTTC, P1Rev -- TTGCGGACCTGGAAGGAG, P2Fwd -- ACTGGAATTGGCCACTTTGT, P2Rev -- GATACAGCGCGCTCATTG, P3Fwd -- GTCTCCCCATGCCTGCTTC, P3Rev -- GGCCCAGCCCATCCAAGG. Predicted PCR products were subjected to searches of the genome using the BLAST program \[[@B26]\] to confirm specificity. Chromatograms were analysed by two independent researchers using the programs Chromas \[[@B27]\] and MutationSurveyor (SoftGenetics, State College, USA). High performance liquid chromatography was performed using the WAVE system (Transgenomics, Omaha, USA). Authors\' contributions ======================= NCMH performed all bioinformatical and molecular studies and drafted the manuscript. IT, WL, VM, ES, GL, RP, PL, JO, RT, PJM, AN, RCT, SH and AL contributed cases of PJS to the study. RSH conceived the study and drafted the manuscript, and participated in the study\'s design and coordination. All authors read and approved the final manuscript.	PubMed Central	2024-06-05T03:55:55.763322	2005-3-17	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084245/", "journal": "BMC Genomics. 2005 Mar 17; 6:38", "authors": [ { "first": "Nicholas CM", "last": "Hearle" }, { "first": "Ian", "last": "Tomlinson" }, { "first": "Wendy", "last": "Lim" }, { "first": "Victoria", "last": "Murday" }, { "first": "Edwin", "last": "Swarbrick" }, { "first": "Guan", "last": "Lim" }, { "first": "Robin", "last": "Phillips" }, { "first": "Peter", "last": "Lee" }, { "first": "John", "last": "O'Donohue" }, { "first": "Richard C", "last": "Trembath" }, { "first": "Patrick J", "last": "Morrison" }, { "first": "Andrew", "last": "Norman" }, { "first": "Rohan", "last": "Taylor" }, { "first": "Shirley", "last": "Hodgson" }, { "first": "Anneke", "last": "Lucassen" }, { "first": "Richard S", "last": "Houlston" } ] }
PMC1084246	Background ========== Reviewers and users of reviews draw conclusions about the overall quality of the evidence that is reviewed. Similarly, people making recommendations and users of those recommendations draw conclusions about the strength of the recommendations that are made. Systematic approaches to doing this can help protect against errors by both doers and users, and can facilitate critical appraisal and communication of the conclusions that are made. The GRADE Working Group began as an informal collaboration of people with an interest in addressing shortcomings in systems for grading evidence and recommendations. We report elsewhere a critical appraisal of six prominent systems for grading evidence and recommendations \[[@B1]\]. Based on this critical appraisal and a series of discussions, we reached agreement on the key attributes of a system that would address the major shortcomings that we identified. Based on the critical assessment of existing approaches, the agreement we had reached about the key elements that should be included in an approach for grading the level of evidence and strength of recommendations and our previous experiences we put together a suggestion for a grading system. We then applied the suggested system to a series of examples and discussed and revised the system based on this experience and the consideration of other examples. Examples were selected to challenge our thinking. All of the examples used in this pilot study were questions about interventions. We describe here the pilot study of this system. The aims of the pilot study were to test whether the approach is sensible relative to diverse examples of evidence and recommendations, and to agree on necessary changes to the approach, decision rules, and changes in how the evidence profiles used in the pilot study were constructed. The revised approach is described elsewhere \[[@B16]\]. Methods ======= Seventeen people independently judged the quality of evidence, the balance between benefits and harms, and the formulation of a recommendation for 12 examples. The 17 judges all had experience using other approaches to grade evidence and recommendations. Evidence profiles ----------------- For each example we prepared an evidence profile. Each evidence profile was made based on information available in a systematic review and consists of two tables, one for quality assessment of the available information and one table that presents a summary of the findings (Table [1](#T1){ref-type="table"} and Table [2](#T2){ref-type="table"}). For the purpose of testing our grading approach in this pilot study we made the assumption that the systematic reviews that we used were all well conducted. The examples we used and presented here were selected to test our new approach, not with an intention of making actual recommendations for a specific setting based on up-to-date systematic reviews. The quality assessment table was designed such that the quality of each outcome was evaluated separately. For each outcome, the table contained information regarding the number of studies that had reported the outcome, information about the study design (RCTs or observational studies) and the quality of the studies that reported on that outcome (was there any limitations in the design or conduct of these studies). Also included in the quality assessment table was information about the consistency of the results across studies for each outcome and information regarding directness of the study population, outcome measure, intervention and comparison. The summary of findings table was also designed such that each outcome was presented separately. For each outcome information are presented about both the experimental and the control group patients, for dichotomous outcomes the number of events and the total number of participants, and for continuous outcomes means (standard deviation) and the number of patients were presented. Also included in the summary of findings table is information about the effect, relative effect (95% confidence interval) and absolute effect for each outcome. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Example of an evidence profile quality assessment given to the evaluators for them to grade in the pilot study. Example question: Should depressed patients in primary care be treated with SSRIs rather than tricyclics? ::: ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Outcome: Depression severity(measured with Hamilton Depression Rating Scale) ---------------------------------------------------------------------------------- ------------------------------------------ --------------------------------- ---------------------------- ---------------------------------------------------- Studies Design Quality Consistency Directness 8 trials Citalopram\ RCTs No serious flaws No important inconsistency Some uncertainty about relevance (outcome measure) 38 trials Fluoxetine\ 25 trials Fluvoxamine\ 2 trials Nefazodone\ 18 trials Paroxetine\ 4 trials Sertaline\ 4 trials Velafaxine Outcome: Transient side effects(drop-out from 6 week treatment) 8 trials Citalopram\ RCTs No serious flaws No important inconsistency Some uncertainty about relevance (outcome measure) 50 trials Fluoxetine\ 27 trials Fluvoxamine\ 4 trials Nefazodone\ 23 trials Paroxetine\ 6 trials Sertaline\ 5 trials Velafaxine Outcome: Poisoning fatalities Office for National Statistics (British) Observational data (national statistics) Serious flaw, population based\ Only one study Direct Reporting bias ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Example of an evidence profile summary of findings given to the evaluators for them to grade in the pilot study. Example question: Should depressed patients in primary care be treated with SSRIs rather than tricyclics? ::: Outcome SSRI tricyclics Effect Quality Relative importance ------------------------ --------------------------------- ---------------------------------- ----------------------------- --------------- --------------------- -- Depression Severity 5044 patients 4510 patients WMD 0.034 (-0.007 to 0.075) No difference Transient side effects 1948/7032 (28%) 2072/6334 (33%) RRR 13% (5% to 20%) 20 Poisoning fatalities\* 1/100,000 per year of treatment 58/100,000 per year of treatment RRR 98% 1754 \* Uncertainty about baseline risk: Fatality data may be influenced by which pills are given to whom, and it is uncertain if changing antidepressant would deter suicide attempts ::: Instructions and a form for recording each judgement were included with each example \[see [Additional file 1](#S1){ref-type="supplementary-material"}\]. The judges were instructed to apply the approach without second guessing the information presented in the evidence profile or the approach. They were asked to note problems that they encountered and judgements that did not make sense to them when they adhered to the approach as instructed. Questions and judgements ------------------------ The 12 examples were selected to include a variety of health care interventions, types of evidence and types of recommendations. The questions that were posed in the 12 examples were: • Should depressed patients in primary care be treated with SSRIs or tricyclics? \[[@B2]\] • Should patients with atrial fibrillation be treated with warfarin or aspirin for prevention of stroke? \[[@B3]\] • Should patients with pain believed to be due to degenerative arthritis be treated with non-steroidal anti-inflammatory drugs (NSAIDs) or paracetamol? \[[@B4]\] • Should patients who have had a myocardial infarction be given antiplatelet therapy to reduce all cause mortality? \[[@B5]\] • Should patients who have had a myocardial infarction be offered exercise rehabilitation? \[[@B5]\] • Should patients with deep venous thrombosis be treated with Low Molecular Weight Heparin (LMWH) or IV unfractionated heparin for prevention of pulmonary embolism? \[[@B6]\] • Should antibiotics be used to treat acute maxillary sinusitis? \[[@B7]\] • Should BCG vaccine be used to prevent tuberculosis? \[[@B8]\] • Should surgical discectomy be recommended for patients with sciatica due to lumbar disc prolapse? \[[@B9]\] • Should community water fluoridation be used to reduce dental caries? \[[@B10],[@B11]\] • Should distribution of child safety seats and education programs be used to increase correct use of child safety seats? \[[@B12]\] • Should hormone replacement therapy be given to prevent cardiovascular heart disease in healthy post menopausal women? \[[@B13]\] For each example each person made judgements about; • the quality of evidence for each outcome, scored as high, intermediate, low, or very low; • the relative importance of each outcome, scored as critical to the decision (7--9), important but not critical to the decision (4--6), or not important to the decision (1--3); • the overall quality of all the critical outcomes, scored as high, intermediate, low, or very low; • the balance between benefits and harms, scored as net benefit, trade offs, uncertain net benefit, or not net benefit; and • the recommendation, scored as do it, probably do it, toss up, probably don\'t do it, or don\'t do it. For each example the judgements made by all 17 people were collected and summarised as illustrated in Table [3](#T3){ref-type="table"}. Disagreements were discussed at a meeting attended by 15 of the 17 judges. Because of a lack of time, the last two examples were discussed at another meeting attended by six of the 17 judges, but all 17 raters provided judgements for all of the 12 examples. For each example the kappa agreement was calculated \[[@B14]\] for the 17 graders across the four levels for the quality of evidence across outcomes for each example (number of outcomes per example range from two to seven), across all outcomes (46) and for the judgements about overall quality of the evidence (12). ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Summary of the judgements made by the 17 evaluators for Example 1 of the pilot study. Should depressed patients in primary care be treated with SSRIs rather than tricyclics? ::: Rater Quality of outcome Relative importance of outcome Overall quality Balance benefits vs harm Recommendation ------- --------------------- ------------------------ ---------------------- -------------------------------- ------------------------ ---------------------- ----------------- -------------------------- ---------------- Depression severity Transient side effects Poisining fatalities Depression severity Transient side effects Poisining fatalities 1 H H Vl 7 9 9 H or Vl Uncertain net benefit Don\'t do it 2 M M Vl 9 6 7 Vl Net benefit Probably do it 3 H H M 8 7 9 H Uncertain net benefit Toss up 4 H H L 6 5 6 5 M M M 9 6 8 M Net benefit Do it 6 M M Vl 9 6 9 Vl Net benefit Do it 7 M M L 8 7 8 L Net benefit Do it 8 H H Vl 9 5 3 H Net benefit Probably do it 9 M M L 9 6 8 L Net benefit Probably do it 10 M M L 9 7 8 L Net benefit Probably do it 11 H H L 8 5 7 L Trade offs Probably do it 12 H H L 8 5 7 L Trade offs Probably do it 13 M H M 9 7 9 M Net benefit Do it 14 M M L 9 9 5 OR 9 M Net benefit Probably do it 15 M M Vl 9 6 8 Vl Uncertain net benefit Toss up 16 M M Vl 9 5 9 Vl Not net benefit Don\'t do it 17 M M M 9 9 9 M Net benefit Toss up ::: Sensibility and understandability --------------------------------- After grading all 12 examples, the judges were asked 16 questions regarding the sensibility and understandability of the approach. Each question consisted of a statement and five response options: strongly disagree, disagree, not sure, agree, and strongly agree. Eleven people completed this questionnaire. The questionnaire was adapted from Feinstein \[[@B15]\] and the 16 statements were: 1\. The approach is applicable to different types of interventions, including drugs, surgery, counselling, and community-based interventions. 2\. The approach is clear and simple to apply 3\. The information that is needed is generally available. 4\. Subjective decisions are generally not needed. 5\. All of the components included in each of the five types of judgements should be included 6\. There are not important components that are missing for any of the five types of judgements. 7\. The ways in which the components are aggregated for each of the five types of judgements are clear and simple. 8\. The ways in which the included components are aggregated are appropriate for each of the five types of judgements. 9\. The categories are sufficient to discriminate between different grades for each of the five types of judgements. 10\. The approach successfully discriminates between different grades of evidence. 11\. The approach successfully discriminates between different grades of recommendations. 12\. The overall quality of evidence is clear and understandable. 13\. The balance between the benefits and harms is clear and understandable. 14\. The recommendation is clear and understandable. 15\. The way in which the overall quality of evidence was graded is better than other ways of doing this with which I am familiar. 16\. The way in which the recommendation was graded is better than other ways of doing this with which I am familiar. Results ======= Quality of evidence for each outcome ------------------------------------ The quality of evidence for each outcome as assessed by the 17 graders are shown in Table [4](#T4){ref-type="table"}. Much of the disagreement was due to lacking information in the evidence summaries that we prepared based on the information available in the chosen examples. We agreed that the evidence summaries should include footnotes explaining the basis for judgements about study quality, consistency and directness. We also agreed that it was necessary to include information about baseline risk and the setting as part of the background information since different assumptions about these factors also explained some of the disagreement. It was possible to reach a consensus about the quality of evidence for most outcomes when we discussed our judgements. Of the 48 outcomes that were included across the 12 examples, we were not able to reach a consensus regarding five. The lack of consensus resulted from disagreement about whether there was sparse data for three outcomes and because of insufficient information for two outcomes. ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Results, summary of the judgements made by the 17 evaluators of the quality for each of the outcomes presented in the 12 examples in the pilot study. ::: Outcome High Moderate Low Very low Consensus Comments ------------------------------- ------- ---------- ------- ---------- ----------- ------------------------------------------------- Depression severity 6/17 10/17 \- \- Moderate Transient side effects 7/17 10/17 \- \- High Changed to little uncertainty Poisoning fatalities \- 4/17 7/17 6/17 Moderate Upgraded for very strong association Stroke 15/17 2/17 \- \- High Extracranial hemorrage 16/16 \- \- \- High All cause mortality 12/17 5/17 \- \- \- Agreed to remove this outcome Pain at rest 16/17 1/17 \- \- High Pain ay motion 15/17 2/17 \- \- Moderate Uncertainty about directness of outcome measure Mobility 3/17 14/17 \- \- Moderate Quality of life 1/17 11/17 5/17 \- Moderate Dropout due to side effects 14/17 3/17 \- \- High Serious gi complications \- 3/17 8/17 6/17 \- Need more information before consensus All cause mortality 2/17 15/17 \- \- Moderate Non-fatal stroke 17/17 \- \- \- High Non-fatal MI 17/17 \- \- \- High Death 13/17 4/17 \- \- High Non-fatal MI 11/16 5/16 \- \- High All cause death 12/17 5/17 \- \- Moderate If reporting bias, otherwise high Major bleeding 15/17 2/17 \- \- High Recurrent thromboembolism 6/17 11/17 \- \- High Clinical cure \- 13/17 4/17 \- Moderate Dropout due to side effects \- 10/17 7/17 \- Moderate Relapse 2/17 11/17 4/17 \- Moderate Tuberculosis 2/17 10/17 5/17 \- Moderate TB death 8/17 8/17 1/17 \- High TB meningitis 1/17 4/17 12/17 \- Moderate Strong association Serious adverse events 1/12 \- \- 11/12 \- No data, outcome removed Condition unchanged 5/17 19/17 2/17 \- \- No consensus regarding sparse data Poor outcome- surgeon rated 9/17 8/17 \- \- \- Need bias information before consensus 2^nd^procedure needed 9/17 8/17 \- \- Moderate No success -- objective rater 5/17 10/17 2/17 \- \- No consensus regarding sparse data Risks & side effects 1/15 2/15 2/15 10/15 \- No data, outcome removed Dental caries -- start \- \- 8/17 9/17 Very low Dental caries -- stop \- \- 5/17 12/17 Very low Dental florosis \- 1/17 3/17 13/17 Very low Bone fracture \- 1/17 3/17 13/17 Very low Cancer mortality \- 1/17 2/17 14/17 Very low All injuries \- 1/17 12/17 4/17 Very low Question changed Correct use early 8/17 5/17 4/17 \- High Question changed Correct use follow up 2/17 8/17 6/17 1/17 High Question changed Possession of seat 7/17 6/17 3/17 \- High Question changed CHD 11/13 2/13 \- \- High Breast cancer 11/13 2/13 \- \- High Stroke 11/13 2/13 \- \- High Colorectal cancer 11/13 2/13 \- \- High Endometrial cancer 11/13 2/13 \- \- High Hip fracture 11/13 2/13 \- \- High ::: We found that in addition to study design, quality, consistency and directness, other quality criteria also influenced judgements about evidence. These additional criteria were sparse data, strong associations, publication bias, dose response, and situations where all plausible confounders strengthened rather than weakened our confidence in the direction of the effect. Concequently, the consistency with which we considered these additional issues were affected and disagreements regarding the quality of evidence for each outcome were reduced. Relative importance of each outcome ----------------------------------- Specification of outcomes in the question that each example addressed resulted in some confusion regarding the relative importance of each outcome and the overall quality of evidence across outcomes. We therefore agreed that outcomes should not be included in the questions and that all important outcomes should be considered. There was good agreement about the relative importance of the 48 outcomes that were considered. We reached a consensus about the relative importance of all but two of the outcomes. This was due to uncertainty and true disagreement about the importance of these two outcomes, dental fluorosis and bone fractures, in relation to the question about water fluoridation. Overall quality of important outcomes ------------------------------------- There was a lack of agreement about the overall quality of evidence across the critical outcomes for each question (Table [5](#T5){ref-type="table"}). This poor agreement reflected an accumulation of disagreements about the quality of evidence and importance of the individual outcomes that were considered for each question. In addition, we found that it did not make sense to downgrade the overall quality of evidence because of lower quality evidence for one of several critical outcomes when all of the outcomes showed effect in the same direction. We therefore agreed that the overall quality of evidence should be based on the higher quality evidence, rather than the lowest quality of evidence, when all of the results are in favour of the same option. ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Results, summary of the judgements made by the 17 evaluators of the overall quality in the 12 examples in the pilot study ::: Example High Moderate Low Very low Consensus Comments ------------- ---------- -------------- --------- -------------- -------------------------------------------------- ----------------------- 1 2/15 4/15 5/15 4/15 Moderate 2 12/17 5/17 \- \- High 3 1/17 6/17 5/17 5/17 Need more information before consensus 4 4/17 13/17 \- \- High Based on the new rule 5 12/16 4/16 \- \- High 6 7/17 10/17 \- \- High Based on new rule 7 \- 11/17 6/17 \- Moderate 8 \- 6/17 3/17 8/17 High Based on new rule 9 2/16 3/16 5/16 6/16 High/Moderate depending if there are fatal flaws 10 \- 1/17 4/17 12/17 Very low 11 1/17 3/17 8/17 5/17 High Changed question 12 11/13 2/13 \- \- High ::: The kappa statistics for each question are shown in Table [6](#T6){ref-type="table"}. The number of outcomes per example range from two to seven and the kappa ranged from 0 to 0.82. In some instances, the agreement among the graders was slightly worse than by chance as indicated by the negative kappa values seen in Table [6](#T6){ref-type="table"}. The kappa across the 46 outcomes included in the calculation was 0.395 (SE 0.008). Kappa for agreement beyond chance for the 12 final judgements about the quality of evidence was 0.270 (SE 0.015). ::: {#T6 .table-wrap} Table 6 ::: {.caption} ###### Results, kappa agreement among the evaluators for each of the 12 examples in the pilot study ::: Example No of outcomes P Kappa (SE) ------------- -------------------- ------- ----------- ---------- 1 3 0.436 0.149 0.031 2 3 0.769 0.075 0.053 3 6 0.643 0.441 0.024 4 3 0.926 0.823 0.050 5 2 0.608 -0.044 0.065 6 3 0.618 0.163 0.050 7 3 0.520 -0.028 0.044 8 3 0.451 0.146 0.036 9 4 0.441 -0.022 0.037 10 5 0.579 0.005 0.034 11 4 0.377 0.112 0.027 12 7 0.718 -0.083 0.043 ::: Balance between benefits and harms ---------------------------------- The graders assessments about the balance between benefits and harms are shown in Table [7](#T7){ref-type="table"}. There is visibly a poor agreement, this can, in part, be explained by the accumulation of all the previous differences in grading of the quality and importance of the evidence. Some of the judges made assumptions or considered information that was not included in the evidence profiles. When we discussed these judgements, we reached a consensus about the balance between benefits and harms for all but three questions. For one question we found we needed more information. For the second judgement we disagreed about the importance of two of the outcomes. For the third judgement we disagreed about the relative values we attached to the benefits and the harms. ::: {#T7 .table-wrap} Table 7 ::: {.caption} ###### Results, summary of the judgements made by the 17 evaluators about the balance between benefits and harms for each of the 12 examples in the pilot study ::: Example Net benefit Trade offs Uncertain net benefits Not net benefits Consensus --------- ------------- ------------ ------------------------ ------------------ ------------------------------ 1 10/16 2/16 3/16 1/16 Net benefit 2 11/16 4/16 1/16 \- Net benefit 3 2/17 8/17 7/17 \- Need more information 4 15/16 \- 1/16 \- Net benefit 5 13/17 \- 4/17 \- Net benefit 6 13/17 2/17 2/17 \- Net benefit 7 4/17 3/17 9/17 1/17 Uncertain net benefits 8 7/16 \- 9/16 \- Net benefit 9 2/16 8/16 6/16 \- Uncertain benefit/trade offs 10 2/17 4/17 10/17 1/17 No consensus 11 12/17 \- 5/17 \- Net benefit 12 \- 2/13 1/13 10/17 No consensus ::: Recommendation -------------- The graders individual considerations about the recommendations are shown in Table [8](#T8){ref-type="table"}. During the discussion, we reached a consensus on a recommendation for the nine examples where we agreed on the balance between benefits and harms. We found that first agreeing on the balance between the benefits and harms clarified our judgements about recommendations and facilitated a consensus. There was not a one-to-one correspondence between our judgements about trade-offs and our judgements about recommendations, because the latter took into account additional considerations. ::: {#T8 .table-wrap} Table 8 ::: {.caption} ###### Results, summary of the recommendations made the 17 evaluators for each of the 12 examples in the pilot study ::: Example Do it Probably do it Toss up Probably don\'t do it Don\'t do it Consensus --------- ------- ---------------- --------- ----------------------- -------------- ------------------------------ 1 4/16 7/16 3/16 \- 2/16 Probably do it 2 6/16 8/16 2/16 \- \- Do it 3 \- 6/15 7/15 2/15 \- Need more information 4 13/15 2/15 \- \- \- Do it 5 11/16 5/16 \- \- \- Do it 6 11/17 5/17 1/17 \- \- Do it 7 1/17 7/17 2/17 6/17 1/17 Probably do it 8 2/15 7/15 4/15 2/15 \- Do it 9 1/17 4/17 8/17 4/17 \- Probably don\'t do it/Tossup 10 \- 2/17 6/17 7/17 2/17 No consensus 11 7/17 8/17 2/17 \- \- Do it 12 \- \- \- 4/13 9/13 No consensus ::: Sensibility and understandability --------------------------------- Eleven raters provided feedback on the sensibility and understandability of the GRADE system for grading evidence and formulating recommendations. Nine of the 11 respondents agreed or strongly agreed that the judgements about the overall quality of evidence were clear and understandable, and that the judgements about the balance between benefits and harms were clear and understandable using the GRADE approach. Everyone agreed or strongly agreed that the judgements about recommendations were clear and understandable. Eight of the judges agreed or strongly agreed that the GRADE approach to judging the overall quality of evidence was better than other grading systems with which they were familiar. Two disagreed and one was not sure. Eight also agreed that the GRADE approach to formulating recommendations was better than approaches with which the raters were familiar. Three raters were not sure about whether the GRADE approach was superior to other approaches of formulating recomendations. Nine of the 11 respondents agreed or strongly agreed that the GRADE approach was applicable to different types of interventions, and that the approach was clear and simple to apply. Five judges disagreed that the information that is needed is generally available, two were not sure and four agreed. Six of the eleven judges disagreed or strongly disagreed that subjective decisions were generally not needed, four were not sure and one agreed. Ten of the eleven judges agreed or strongly agreed that all the components included in each of the four types of judgements should be included; one judge was not sure. Five of the judges were unsure if there were not important components that were missing from any of the four types of judgements, one disagreed and three agreed or strongly agreed. Eight judges agreed or strongly agreed that the ways in which the components were aggregated for each of the four types of judgements were clear and simple; three were unsure. Seven judges agreed or strongly agreed that the ways in which the included components were aggregated were appropriate for each of the four types of judgements, two were unsure and two disagree. Ten of the eleven judges agreed or strongly agreed that the categories were sufficient to discriminate between different grades for each of the four types of judgements; one disagreed. All the eleven judges agreed or strongly agreed that the GRADE approach successfully discriminated between different quality of evidence, and between different grades of recommendations. Discussion ========== This pilot study of the GRADE approach to grading the quality of evidence and strength of recommendations helped to identify problems with the approach and enabled us to address these. We found that it was possible to resolve most of the disagreements we had when making judgements independently and there was agreement that this approach warrants further development and evaluation. Many of the disagreements were a direct result of a lack of information. We concluded that there is a need for detailed additional information in evidence profiles, and have modified the evidence profiles accordingly. When we have found an empirical basis or compelling arguments, we have also provided precise definitions. For example, we have agreed on a basis for defining strong and very strong associations. However, in many cases we continue to rely on judgement. We have addressed this by always including the rationale for such judgements in footnotes attached to the evidence profile. The evidence profiles used in the pilot study were based on systematic reviews. \[[@B2]-[@B13]\] Much of the information we found lacking was missing in these original systematic reviews, particularly information about harms and side effects. It was outside of the scope of this study to systematically collect this information. However, systematic reviews of evidence of harms, as well as benefits, are essential for guidelines development panels. If reviews, such as Cochrane reviews, are going to meet the needs of guideline development panels, and others making decisions about health care, it is essential that evidence of adverse effects is systematically included in these. An important benefit of the approach to grading evidence and recommendations that we used in this study is that it clarifies the source of true disagreements, as well as helping to resolve disagreements through discussing each type of judgement sequentially. Judgements about the relative importance of different outcomes and about trade-offs, as well as about the quality of evidence, are made explicitly, rather than implicitly. This facilitates discussion and clarification of these judgements. It may be helpful to guideline panels and others to use this approach before making decisions and recommendations. The most common source of disagreement that we encountered was differences in what we consider to be sparse data. We have not reached a consensus on a definition of sparse data, but have acknowledged that we have different thresholds and now recognize this when we make judgements about the quality of evidence \[[@B16]\]. We have as a result of this pilot study been able to make considerable improvements to our system for grading the quality of evidence and strength of recommendations. The evidence profiles used in the pilot study have been modified and now include information that was missing and was found to be an important source of disagreement, as illustrated in Table [9](#T9){ref-type="table"} and Table [10](#T10){ref-type="table"} and the criteria used for grading the quality of evidence for each important outcome have been modified as summarised in Table [11](#T11){ref-type="table"}. Guideline generation includes judgement. Individual, residual judgements will impact on the agreement we measured in this study. Thus, lower kappa values are expected. Further refinement of the GRADE system and additional instructions will improve agreement. ::: {#T9 .table-wrap} Table 9 ::: {.caption} ###### Example of a modified GRADE evidence profile quality assessment. Table 9 and 10 is what Table 1 and 2 became when including the improvements made based on the pilot study experience. Question: Should depressed patients be treated with SSRIs rather than tricyclics? Setting: Primary care Patients: Moderately depressed adult patients Reference: North of England Evidence Based Guideline Development Project. Evidence based clinical practice guideline: the choice of antidepressants for depression in primary care. Newcastle upon Tyne: Centre for Health Services Research, 1997. ::: ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ Outcome: Depression severity(measured with Hamilton Depression Rating Scale after 4 to 12 weeks) ------------------------------------------------------------------------------------------------------ -------------------- ------------------------ ---------------------------- ------------------------------------------------------- ---- ---- ---- ---- ---- Studies Design Quality Consistency Directness SD SA RB DR PC 8 trials Citalopram\ RCTs No serious limitations No important inconsistency Some uncertainty about directness (outcome measure)\* No No No No No 38 trials Fluoxetine\ 25 trials Fluvoxamine\ 2 trials Nefazodone\ 18 trials Paroxetine\ 4 trials Sertaline\ 4 trials Velafaxine Outcome: Transient side effects resulting in discontinuation of treatment 8 trials Citalopram\ RCTs No serious limitations No important inconsistency Direct No No No No No 50 trials Fluoxetine\ 27 trials Fluvoxamine\ 4 trials Nefazodone\ 23 trials Paroxetine\ 6 trials Sertaline\ 5 trials Velafaxine Outcome: Poisoning fatalities Office for National Statistics (British) Observational data Serious limitation\\ Only one study Direct No ++ No No No ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ \There was uncertainty about the directness of the outcome measure because of the short duration of the trials. \\It is possible that people at lower risk were more likely to have been given SSRI\'s and it is uncertain if changing antidepressant would have deterred suicide attempts. SD = Sparse data (Yes or No) SA = Strong association (No, + = strong, ++ = very strong) RB = Reporting bias (Yes or No) DR = Dose response (Yes or No) PC = All plausible confounders would have reduced the effect (Yes or No) CI = confidence interval WMD = weighted mean difference RRR = relative risk reduction ::: ::: {#T10 .table-wrap} Table 10 ::: {.caption} ###### Example of a modified GRADE evidence profile summary of findings. Table 9 and 10 is what Table 1 and 2 became when including the improvements made based on the pilot study experience ::: Outcome* SSRI Tricyclics Effect Quality Importance ---------------------------- --------------------------------- ---------------------------------- ----------------------------- --------------- ---------------- ---------- Depression severity 5044 patients 4510 patients WMD 0.034 (-0.007 to 0.075) No difference Moderate Critical Transient side effects 1948/7032 (28%) 2072/6334 (33%) RRR 13% (5% to 20%) 5 per 100 High Critical Poisoning fatalities\\\* 1/100,000 per year of treatment 58/100,000 per year of treatment RRR 98% (97% to 99%) 6 per 10,000 Moderate Critical \\\There is uncertainty about the baseline risk for poisoning fatalities. ::: ::: {#T11 .table-wrap} Table 11 ::: {.caption} ###### Modified GRADE quality assessment criteria ::: ------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Quality of evidence* Study design Lower if \* Higher if \* ------------------------- ------------------------ ------------------------------------------- -------------------------------------------------------------------------- High Randomised trial Study quality:\ Strong association:\ -1-Serious limitations\ +1-Strong, no plausible confounders, consistent and direct evidence\\\ -2-Very serious limitations\ +2-Very strong, no major threats to validity and direct evidence\\\\ -1-Important inconsistency\ +1-Evidence of a Dose responsegradient\ Directness:\ +1-All plausible confounderswould have reduced the effect -1-Some uncertainty\ -2-Major uncertainty\ -1-Sparse data\ -1-High probability of Reporting bias* Moderate Quasi-randomised trial Low Observational study Very low Any other evidence ------------------------------------------------------------------------------------------------------------------------------------------------------------------------- \* 1 = move up or down one grade (for example from high to moderate) 2 = move up or down two grades (for example from high to low) The highest possible score is High (4) and the lowest possible score is Very low (1). Thus, for example, randomised trials with a strong association would not move up a grade. \\ A relative risk of \>2 (\< 0.5), based on consistent evidence from two or more observational studies, with no plausible confounders \\\* A relative risk of \> 5 (\< 0.2) based on direct evidence with no major threats to validity ::: Judgements about confidence in evidence and recommendations are complex. The GRADE system represents our current thinking about how to reduce errors and improve communication of these complex judgements. Ongoing developments include: • Exploring the extent to which the same system should be applied to public health and health policy decisions as well as clinical decisions • Developing guidance for when and how costs (resource utilisation) should be considered • Developing guidance for judgements regarding sparse data • Adapting the approach to accommodate recommendations about diagnostic tests when these are based on evidence of test accuracy • Incorporating considerations about equity • Preparing tools to support the application of the GRADE system Plans for further development include studies of the reliability and sensibility of this approach and a study comparing alternative ways of presenting these judgements \[[@B17]\]. We invite other organisations responsible for systematic reviews of the effects of healthcare or practice guidelines to work with us to further develop and evaluate the system described here. Conclusion ========== Based on the results of this pilot study we have been able to considerably improve our system for grading the quality of evidence and strength of recommendation \[[@B16]\]. Competing interests =================== DA has competing interests with the US Preventive Services Task Force (USPSTF), PAB has a competing interest with the US Task Force on Community Preventive Services (USTFCPS), ME has and JM had competing interests with the National Institute for Clinical Excellence (NICE), GHG, RJ and HS have competing interests with the American College of Chest Physicians (ACCP), RTH has competing interests with the Scottish Intercollegiate Guidelines Network (SIGN), SH and DO\'C have competing interests with the Australian National Health and Medical Research Council (ANHMRC), BP has competing interests with the Oxford Centre for Evidence-Based Medicine (OCEBM). Most of the other members of the GRADE Working Group have experience with the use of one or more systems for grading evidence and recommendations. Authors\' contributions ======================= DA, PAB, ME, SF, GHG, RTH, SH, RJ, AL, NM, JM, DO\'C, ADO, BP, HS, TTTE, GEV & JWJ Jr as members of the GRADE Working Group have contributed to the preparation of this manuscript and the development of the ideas contained herein, participated in the pilot study, and read and commented on drafts of this article. All authors except GEV judged the quality of the evidence and strength of recommendation based on information presented in the evidence profiles. GEV prepared the first draft of this article, had primary responsibility for preparing the evidence profiles used in the study, and coordinated the study. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1472-6963/5/25/prepub> Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Instructions and form for judgements used in the pilot study ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ We wish to thank Stephen Walter for his help with the kappa statistics. The National Institute for Clinical Excellence (NICE) for England and Wales and the Polish Institute for Evidence-Based Medicine (PIEBM) have provided support for meetings of the GRADE working group. The institutions with which members of the Working Group are affiliated have provided intramural support. Opinions expressed in this paper do not necessarily represent those of the institutions with which the authors are affiliated.	PubMed Central	2024-06-05T03:55:55.764353	2005-3-23	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084246/", "journal": "BMC Health Serv Res. 2005 Mar 23; 5:25", "authors": [ { "first": "David", "last": "Atkins" }, { "first": "Peter A", "last": "Briss" }, { "first": "Martin", "last": "Eccles" }, { "first": "Signe", "last": "Flottorp" }, { "first": "Gordon H", "last": "Guyatt" }, { "first": "Robin T", "last": "Harbour" }, { "first": "Suzanne", "last": "Hill" }, { "first": "Roman", "last": "Jaeschke" }, { "first": "Alessandro", "last": "Liberati" }, { "first": "Nicola", "last": "Magrini" }, { "first": "James", "last": "Mason" }, { "first": "Dianne", "last": "O'Connell" }, { "first": "Andrew D", "last": "Oxman" }, { "first": "Bob", "last": "Phillips" }, { "first": "Holger", "last": "Schünemann" }, { "first": "Tessa Tan-Torres", "last": "Edejer" }, { "first": "Gunn E", "last": "Vist" }, { "first": "John W", "last": "Williams" } ] }
PMC1084247	Background ========== Respiratory syncytial virus (RSV) is the most common cause of viral bronchiolitis and pneumonia in infants and children under two years of age. Premature infants, children with chronic lung diseases (e.g. bronchopulmonary dysplasia), children with congenital heart disease and immunodeficiency patients have high risk of a severe disease. There are no efficient active vaccines and causal therapy, therefore prophylactic measures are especially important. Prospective surveillance of RSV infection and consequent implication of hygienic measures for preventing transmission and spread of RSV contribute effectively to reduction of hospital-acquired infections. Moreover, a humanized monoclonal antibody palivizumab (Synagis^®^, Abbott) was licensed in September 1999 in the European Union for passive immunoprophylaxis of RSV. In view of the relatively low risk of RSV rehospitalisation among premature infants and high costs of immunoprophylaxis, additional individual risk factors and especially the local epidemiologic situation have to be considered in the case of Synagis^®^-administration \[[@B1]-[@B3]\]. RSV epidemics occur seasonally and last from four to six months. North of the equator the peak incidence of disease is observed from December to April \[[@B4]-[@B6]\]. In Europe, regular biennial patterns of RSV epidemics were described in which a weaker late season was followed by a more severe early season \[[@B7]-[@B11]\]. The discovery of a regular pattern of RSV epidemics seemed to allow the prediction of outbreaks. We characterized RSV epidemics in southern Germany with a focus on the Stuttgart area -- onset, offset, peak, intensity, duration of outbreaks and seasonal rhythm -- and compared them with already published data. The aim of this work was to use obtained data to fix a time limit for Synagis^®^prophylaxis by precisely determining the period with high risk of RSV. Methods ======= Samples ------- Nasopharyngeal samples were sent from paediatric hospitals and practitioners from children with suspected RSV infection. Samples were sent cooled in transport medium Veal Infusion Broth (BD Difco™, USA). Detection of RSV ---------------- From January 1996 to April 2001 the detection of RSV was carried out by RT-PCR and ELISA Pathfinder (Bio-Rad, USA). 70% of samples were tested by both methods, 20% of samples were tested only by ELISA and 10% only by RT-PCR. Since May 2001 only RT-PCR has been used for RSV testing because of its higher sensitivity and specificity. Extraction of Nucleic Acid (NA) ------------------------------- From January 1996 up to September 2001 NA extraction was performed manually using the High Pure Viral Nucleic Acid Kit (Roche Diagnostics GmbH, Mannheim, Germany). Briefly: Virus lysis is performed by incubation of the samples in a lysis/binding buffer in the presence of proteinase K and poly (A) carrier RNA. NA are bound to glass fibers and were subsequently recovered in 50 μl elution buffer by centrifugation through a glass fleece. From September 2001 on NA extraction was performed on the MagNA Pure LC automated instrument (Roche Diagnostics GmbH, Mannheim, Germany) using the MagNA Pure LC Total Nucleic Acid Isolation Kit (Roche Diagnostics GmbH, Mannheim, Germany) for all preparations. The setting and preparation of the instrument were performed according to the manufacturer\'s instructions. Briefly, 200 μl of each sample was transferred to the sample cartridge and loaded onto the workstation together with the kit reagents. The instrument automatically performs all isolation and purification steps such as addition of lysis/binding buffer and magnetic glass particles (MGPs), binding of DNA to the MGPs, washing steps, elution of the pure NA (60 μl), and transfer of the NA to a cooled storage cartridge. NA extracted in this way is ready for use in RT-PCR applications. RT-PCR ------ A Dig-ELISA RT-PCR protocol was used from September 1996 up to October 1999 together with a commercial PCR-ELISA kit (Roche Diagnostics GmbH, Mannheim, Germany) for amplicon detection. A real-time RT-PCR protocol replaced the Dig-ELISA RT-PCR from October 1999 on. For real-time RT-PCR, the LightCycler (LC) instrument (Roche Diagnostics GmbH, Mannheim, Germany) and primers and probes from the viral N gene, previously described by \[[@B12]\] were used. To enable rapid non-type specific simultaneous detection of both RSV subtypes, primer pairs A21/A102 for subtype A and B17/B120 for subtype B were used in combination with the probes APB48 for subtype A and BPB45 for subtype B in one reaction mixture. Probes contained the 5\' reporter dye 6FAM and the 3\' quencher dye TAMRA (TIB MOLBIOL, Berlin, Germany). LC RT-PCR was performed in LC capillaries in a final reaction volume of 20 μl. Master mixes were based on the commercial QuantiTect™ Probe RT-PCR kit (QIAGEN, Hilden, Germany) supplemented with primers A21/A102/B17/B120 (f.c. 1,0 μM each), the probes APB48 and BPB45 (f.c. 0,25 μM each) and heat labile UNG (0,2 U) (Roche Diagnostics GmbH, Mannheim, Germany). Use of this kit allows both reverse transcription and PCR to take place in a single capillary, so there is no need to open the capillary once reverse transcription is accomplished. After pipetting of 13,2 μl of this reaction mixture into LC capillaries, 6,8 μl of the crude NA preparation was added. A 5 min. step at room temperature to enable UNG activity is followed by 20 min. of reverse transcription at 50°C. PCR is started with a 15 min. step at 95°C followed by 10 cycles of denaturation (95°C for 5 sec), annealing (with a 1°C decrease per cycle from 65°C to 55°C for 20 sec) and extension (72°C for 15 sec), followed by 40 cycles of denaturation (95°C for 0 sec), annealing (55°C for 20 sec) and extension (72°C for 30 sec). Epidemiological analysis ------------------------ Epidemiological analysis included nine RSV seasons from January 1996 to June 2004. Only samples sent by hospitals and practitioners in southern Germany (Baden-Württemberg and Bavaria) were analyzed in this study. 80% of samples were sent from hospitals in the Stuttgart area. The following data were extracted from the laboratory database: laboratory identification number, age, sample withdrawal date, sender address and RSV antigen detection test results. Onset, offset, peak and duration of outbreaks were determined for each year. In order to distinguish between the beginning/ending of an outbreak and background activity, the number of sporadic RSV cases in the summer months from June to October were calculated. A maximum of 5 cases in a month and 2 cases in a week were registered in this time. Therefore, onset month (week) of outbreak was defined as the first of two consecutive months (weeks) with at least 5(2) positive RSV findings. Offset month was defined as the last of the final two consecutive months (weeks) with at least 5(2) positive RSV findings. Duration was defined as the number of months (weeks) between the calculated onset and offset months (weeks) inclusive of those months (weeks). Outbreak peak was defined as the month (week) after onset with the highest number of positive test results. Outbreaks were graphically depicted in two different forms: an epidemic curve by plotting the number of cases on the vertical axis and time on the horizontal axis and a graphic showing percentage of positive test results in a month. The total number of positive RSV findings was considered for estimation of outbreak intensity. Results ======= Nasopharyngeal samples from a total of 3577 patients were tested in the period from January 1996 to June 2004. 32% of samples (1154) were found to be positive. Our customers, paediatric hospitals and practitioners, who sent the samples, remained the same over the time period. 80% of samples were sent in by children\'s hospitals, 10% by practitioners and 10% by other private medical-diagnostic laboratories. The age distribution of patients tested positive reflected the distribution for the whole material sent in: approximately 40% of RSV were infants younger than 3 months, 30% of those aged 3 to 6 months, 20% 7--12 months and 10 % 1 to 3 years. Patients older then 3 years were rarely tested (less then 1%). From Januar 1996 to April 2001 a total of 1653 samples were tested both by EIA and RT-PCR. The sensitivity of EIA versus RT-PCR was 41% (112/273), specificity 98% (1358/1380). The epidemic curve (Fig. [1](#F1){ref-type="fig"}) shows the seasonal distribution of acquired positive RSV results by month. The highest RSV activity was registered in the period from late autumn to spring. A clear biennial cycle of successive RSV outbreaks can be seen: Late weak seasons 1995/96, 1997/98, 1999/00, and 2001/02 (late-low phase, minor peaks) followed by more severe early seasons 1996/97, 1998/99, 2000/01 and 2002/03 (early-high phase, major peaks). Table [1](#T1){ref-type="table"} shows that the onsets, offsets and durations of outbreaks varied significantly from season to season. The late season can start relatively early in week 50 or relative late in week 6. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Absolute number of total RSV investigations and positive results by month, from January 1996 to May 2004. ::: ![](1471-2334-5-20-1) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Onset, peak, offset and duration (weeks) of RSV Outbreaks in Stuttgart, from 1996 to 2004. ::: Season Onset Peak Offset Duration ------------------------ ------- ------ -------- ---------- 1995/1996 late season 1 12 17 17 1996/1997 early season 49 8 13 16 1997/1998 late season 2 13 20 19 1998/1999 early season 48 5 18 24 1999/2000 late season 50 14 19 23 2000/2001 early season 48 6 15 21 2001/2002 late season 6 20 23 18 2002/2003 early season 45 51 6 14 2003/2004 late season 1 14 19 19 ::: Fig. [2](#F2){ref-type="fig"} presents the seasonality of RSV detection as a percentage of samples tested positive per month. Year-to-year alternation between late and early seasons can also be seen in Fig. [2](#F2){ref-type="fig"} as in a Fig. [1](#F1){ref-type="fig"}. The monthly percentage of positive results in early-high seasons was the same as in late-low seasons or actually lower. Therefore any difference of the epidemics intensity cannot be established. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Percentage of RSV positive cases by month, from January 1996 to May 2004. ::: ![](1471-2334-5-20-2) ::: The comparison of seasonal rhythms with literature and surveillance data showed that RSV outbreaks in Stuttgart followed the same pattern as in Freiburg and in Switzerland. In Kiel the seasons from 1994/95 to 1996/97 started late. Since 1997/1998 the same 2-year cycle could be detected in Kiel as in Stuttgart and Freiburg. RSV epidemics in Finland and Sweden occurred synchronously with regular rhythm (high-early versus low-late phase), but were in antiphase with Germany RSV outbreaks (Table [2](#T2){ref-type="table"}). ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Comparison of seasonal rhythm in European areas. ::: RSV Seasons Finland \[7\] Sweden \[8\] Switzerland \[11\] Germany, Freiburg \[9\] Germany, Kiel \[10,18\] Germany, Stuttgart ------------- --------------- -------------- -------------------- ------------------------- ------------------------- -------------------- 1982/1983 late 1983/1984 early 1984/1985 late late 1985/1986 early early 1986/1987 late late 1987/1988 early early 1988/1989 late late early 1989/1990 early early late 1990/1991 late early early 1991/1992 early late late 1992/1993 late early early 1993/1994 early late late 1994/1995 early early late 1995/1996 late late late 1996/1997 early early late early 1997/1998 late late late 1998/1999 early early early 1999/2000 late late late 2000/2001 early early 2001/2002 late late 2002/2003 early early 2003/2004 late late ::: Discussion ========== Age distribution analysis of our patients with diagnosed RSV infection confirmed once more that very small infants younger than three months have the highest risk of hospitalisation and therefore of being tested for RSV \[[@B9]\]. Sensitivity of EIA calculated versus RT-PCR was found to be very low, only 41 %. In the few studies in which Pathfinder RSV EIA was evaluated the sensitivity (up to100%) of this assay was calculated only in comparison with other less sensitive methods than PCR, such as cell culture, shell vial assay or direct immunofluorescence \[[@B13]\]. The low sensitivity of EIA in our study can also be explained with the instability of RSV, even though appropriate collection and transport conditions were assured. In addition, it must be considered that the nasal quantity of RSV correlates with disease severity in hospitalised infants \[[@B14]\]. In our study, the use of EIA did not affect the proportion of positives, since only a small number of samples were tested by this method only. The discovered in Stuttgart long-term regular biennial rhythm allows prediction of whether the next outbreak will be late or early and whether RSV activity will be strong or weak. However, the value of this prediction is limited. The precise time of increase and decrease of RSV activity is not foreseeable (Table [1](#T1){ref-type="table"}). Only local monitoring of RSV activity allows definition of the period with high risk of infection. Therefore, since November 2003 Laboratory Enders & Partner has published weekly from late autumn to spring the actual RSV epidemiological situation in the Stuttgart area on its website \[[@B15]\]. It is noted that our monitoring focused mainly on hospitalised children. With regard to RSV, this approach is acceptable while the proportion of RSV infections among inpatients is higher than in outpatients. However, a population-based approach used in the PRI.DE study may be more sensitive in the detection of offsets because it includes outpatients \[[@B16]\]. There is no generally accepted method for surveillance of RSV outbreaks. In order to compare the RSV epidemiological situation in Stuttgart with that in other areas, we investigated primarily which methods of RSV outbreak analysis are usually used. In Europe, Waris \[[@B7]\] described nine RSV seasons. He ascertained the number of RSV laboratory diagnosis made at the Department of Virology, University of Turku, Finland per month in the time period from January 1981 to March 1990, and discovered a regular periodic biennial pattern of occurrence of RSV outbreaks. A weaker late season was followed by a more severe early season. Other similar epidemic curves were depicted by Reyes (seasons 1984--1993, Karolinska Institute, Sweden) \[[@B8]\], Berner (seasons 1988--1999, University of Freiburg, south Germany) \[[@B9]\], Weigl (seasons 1994--2001 University of Kiel, north Germany 2002) \[[@B10]\] and Duppenthaler (seasons 1988--2001, Switzerland) \[[@B11]\]. In the USA, the national Respiratory and Enteric Virus Surveillance System (NREVSS) monitors and summarizes the RSV surveillance results starting in July 1985 and publishes them on the NREVSS website \[[@B17]\]. The RSV activity in NREVSS reports is shown as median percent of positive for RSV results per month (week) and region. Onsets/offsets are defined as the first/last of two consecutive weeks (months) of at least 10% positive test results with at least two positive samples in the numerator of the reports of both weeks (months) \[[@B4],[@B5]\]. In our study both methods of outbreak analysis were used. We acquired the number and percentage of positive cases per month. Monitoring of RSV activity by calculation of the percentage of positive RSV detection in comparison to the number of positive cases has the advantage that the results show the real RSV situation independently of changes in population. Also, in our study, in outbreaks months at least 10% of samples were positive. However, in summer weeks, in which only few samples were sent in, the positive rate could be over 10% by only one positive sample. Therefore, it is also important to consider the number of sporadic cases in order to distinguish between the beginning/ending of an outbreak and background activity. Onset, offset and peak months on the plot with the percentage curve were nearly the same as on the plot presenting the absolute number of positive cases. Early and late seasons were recognisable on our plot with the percentage curve. These fluctuations of RSV season onset weeks can also be observed in two Omaha Nebraska NREVSS surveillance laboratories \[[@B5]\]. It is noted that on the plot with percentage curve the intensity of epidemics cannot be seen. In seasons with low RSV intensity (small number of samples) the positive rate was as high (50--60%) as in seasons with high RSV activity. Therefore, this information is missing in NRVSS reports. Only the combination of both approaches to the monitoring of RSV outbreaks supplies complete information for the estimation of the epidemiological situation. European surveillance centres present RSV data in different ways: The Pediatric Infectious Diseases Network on Acute Respiratory Tract Infections (PID-ARI.Net) Web&Warnsystem \[[@B18]\] (Kiel, Freiburg and Mainz, Germany) shows tables and incidence plot, the North Saxon State Health Office (NLGA) (Hanover, Germany) \[[@B19]\] uses percentage plot, the Swedish Institute for Infectious Disease Control (SMI) \[[@B20]\] presents tables, the Infectious Diseases Information System (ISIS) (the Netherlands) \[[@B21]\] shows different plots and tables. The use of different methods complicates the interpretation of data. As far as possible we compared seasonal RSV rhythm in Stuttgart with other European reports (Table [2](#T2){ref-type="table"}). This comparison showed that RSV seasonality in Stuttgart was in accordance with other areas of Germany and Switzerland \[[@B9]-[@B11],[@B16],[@B18]\]. The explanation for the absence of a regular pattern in Kiel before 1996 may be the low total number of samples tested and, therefore, the possibility of missed cases. The other possible explanation may be found by analyzing RSV epidemics in northern Europe. RSV epidemics in Finland and Sweden were also synchronized in time as outbreaks in Germany, but the early-high phase in northern Europe corresponded with late-low phase in southern Germany. Kiel is possibly located at the borderline between two areas with different rhythms \[[@B22]\]. Even in areas with a regular rhythm it may shift. Such a switchover can be observed in Chile. The figure depicted by Avendano, which presents RSV detection from 1989 to 2000, shows that after two cycles of late/early seasons the following season 1994 began repeatedly early, then the rhythm was resumed \[[@B23]\]. Also, in Gambia a regular pattern of outbreaks during 6 consecutive annual seasons was disrupted by 2 years of irregular outbreaks, followed by another 2 years of regular seasonal outbreaks \[[@B24]\]. Conclusion ========== The long-term regular biennial rhythm allows predicting whether the next outbreak will be late or early and whether RSV activity will be strong or weak. Not foreseeable, however, is the precise time of increase and decrease of RSV activity. Thus, activity of RSV has to be monitored every year to define the period with high risk of infection. Moreover, different rhythms of epidemics in Europe and the theoretical possibility of their shift demonstrate that the present regionally limited monitoring is not sufficient. An RSV surveillance system in Europe similar to the existing influenza surveillance system is required. Analysis of spatio-temporal European data will contribute to the understanding of RSV epidemiology and predicting outbreaks. Correct identification of outbreaks will permit the limiting of the time period for Synagis^®^administration and thus save costs. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= ET conceived, carried out the study and wrote the manuscript. GE supervised the study. GS carried out the assay development. ME conceived the study and critically reviewed the manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2334/5/20/prepub> Acknowledgements ================ We thank Herbert Ladwig und Robert Hugman for critical reading of the manuscript.	PubMed Central	2024-06-05T03:55:55.773079	2005-3-31	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084247/", "journal": "BMC Infect Dis. 2005 Mar 31; 5:20", "authors": [ { "first": "Elena", "last": "Terletskaia-Ladwig" }, { "first": "Gisela", "last": "Enders" }, { "first": "Gunnar", "last": "Schalasta" }, { "first": "Martin", "last": "Enders" } ] }
PMC1084248	Background ========== A growing number of health care report cards are now available, and they are increasingly being used to increase the profiles of high performing hospitals amongst consumers \[[@B1]-[@B3]\]. However, the recent proliferation of health care report cards has occurred in the absence of an ethical framework to guide their development and implementation. An ethical framework is a consistent and comprehensive theoretical foundation in ethics, and is formed by integrating ethical theories, relevant literature, and other critical information, such as the views of stakeholders. It provides guidance for developing new practices and for challenging and evaluating existing ones. Gormley and Weimer\[[@B4]\] developed a normative framework for organizational report cards through their own expertise as policy analysts, and input from scholars and individuals involved in the design and implementation of organizational report cards. However it was not derived from ethical theory or grounded in the systematically described views of stakeholders. Health care report cards are most well developed in cardiac care. An ethical framework would provide necessary guidance for those generating cardiac report cards (CRCs) and may help them avoid a number of difficult issues associated with existing report cards. \'Gaming\' and uncertainty about the quality of report card data have been cited as impediments to reliable outcomes measures and a reason to limit the public release of report card data \[[@B5],[@B6]\]. Uncertainty also exists as to whether report cards have empowered patients and/or improved health care quality\[[@B7]-[@B10]\]. Other ethical and practical issues, such as balancing the public\'s desire for provider-specific outcomes measures with cardiac care providers\' desires to limit the amount and type of information released to the public, affect the content and legitimacy of CRCs. An ethical framework can identify points of ethical concern for practitioners, patients, policy makers and researchers. And it can aid in the development, implementation and improvement of future generations of CRCs. The purpose of this study is to develop an ethical framework for CRCs. Methods ======= Design ------ Forming this ethical framework has been a three step process. First, we analyzed the relevant ethical issues in an earlier article \[[@B11]\]. A summary of these issues is presented in table [1](#T1){ref-type="table"}. Next we described stakeholders\' views in two previous papers \[[@B12]\] (a paper on patients\' views has been submitted for publication). Finally, this study used a Delphi method with a panel of stakeholders to synthesize these insights into an ethical framework. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Summary of the ethical issues concerning cardiac report cards (adapted from Nast S, Richard SA, Martin DK. Ethical issues related to cardiac report cards. Can J Cardiol. Mar 1 2004;20(3):325--328.) ::: Ethical issue Description ------------------- --------------------------------------------------------------------------------------------------------------------- Quality •Quality operationalizes the ethical principles of beneficence and non-maleficence •Report cards may improve quality of care through external pressure from an informed public •Report cards may impede improvements to quality by generating anger and defensiveness Informed Consent •Informed consent operationalizes the ethical principle of autonomy •To make health care decisions, patients need and want information about their medical options •Report cards have the potential to provide this information and thus facilitate informed consent Equity •Equity operationalizes the ethical principle of justice •Health equity between regions is an important consideration in publicly funded health systems •Report cards must address policy makers to affect regional inequities in health care Legitimacy •Legitimacy operationalizes the ethical principle of justice, in this case deliberative forms of democratic justice •Report card authors must ensure that report cards are and perceived to be legitimate •The legitimacy of report cards will depend on their ability to meet stakeholders\' reasonable expectations ::: The Delphi method allows a panel of stakeholders to generate ideas on a given topic and to reach a consensus on the relative importance of those ideas \[[@B13],[@B14]\]. This study used a three-round modified Delphi method to identify and reach agreement on the elements of an ethical framework for CRCs. Participants and sampling ------------------------- Participants were selected from two previous studies conducted by this research team. The first study described the views of cardiac care administrators, cardiac surgeons, cardiac nurses, cardiac patients, cardiologists, members of the media, and outcomes researchers about CRCs \[[@B12]\]. The second study described the views of cardiac patients (a paper on this study has been submitted for publication). We selected participants for our Delphi panel so that the views and interests of each stakeholder group were represented. Our final panel consisted of 13 panelists: 5 cardiac patients, 2 administrators, 2 cardiac nurses, 2 cardiologists, 1 member of the media, and 1 outcomes researcher. We included a critical mass of patients in order to balance potential or perceived power differentials and enhance the comfort level of participating patients. Data collection and analysis ---------------------------- The Delphi process consisted of three rounds. Round 1 was conducted using electronic communication. We provided three papers based on previous research to each panelist in order to provide background information on the topic. The first paper, Ethical issues related to cardiac report cards, identifies ethical issues related to CRCs, and provided panelists with an ethical analysis \[[@B11]\]. The other two papers, Stakeholders\' views about cardiac report cards\[[@B12]\] and Patients\' views about cardiac report cards, describe stakeholders\' views relating to CRCs (a paper on patients\' views has been submitted for publication). We asked the panelists to identify key issues that emerged from the papers, and to suggest other issues that may have been missing but were relevant to the ethics of CRCs. We synthesized the feedback from the panelists and organized the items into categories that originated from the panelists\' feedback. We combined the items into a draft ethical framework, using the ethical language of the panelists. Round 2 was conducted as a face-to-face round-table discussion. First, we disseminated the draft ethical framework developed in Round 1 to the panelists by mail and email. Then the panelists came together for a discussion of key issues from the draft ethical framework. The discussion was facilitated by a member of the research team who encouraged panelists to debate the issues amongst themselves and develop a consensus on the key items. Two other members of the research team independently recorded the round table discussion. We used the data gathered at the meeting to refine the framework\'s key items. After we organized the feedback into their corresponding categories, we integrated new comments into the existing draft framework and examined it to identify any inconsistencies. At the end of Round 2, we had a refined draft ethical framework. Round 3 was conducted electronically, we disseminated the new draft ethical framework by mail and email to the panelists for final refinements. We organized the panelists\' feedback according to the refined list of categories from Round 2. We integrated new comments into a refined, ethical framework and re-examined it for inconsistencies. The result of Round 3 was a final ethical framework for CRCs. Research ethics --------------- This study was approved by the Committee on Research with Human Subjects at the University of Toronto. We obtained informed consent from each panelist and kept all data confidential and anonymous to those who were not directly involved in the project. Results ======= Figure [1](#F1){ref-type="fig"} provides an overview of the ethical framework. In this section we will describe the items that form the framework and discuss the interrelationships between items. Although the panelists decided that improving the quality of cardiac care was the overriding principle in this ethical framework, they also decided that the principles should not be ranked because they are interrelated, not independent. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Ten principles for cardiac report cards. ::: ![](1472-6939-6-3-1) ::: Improving quality of care ------------------------- Improving the quality of cardiac care should be the fundamental objective driving the development and implementation of CRCs. In meeting this objective, report cards should also maximize the public good by ensuring equitable quality of care within and between regions. Providing stakeholders with information on the quality of cardiac care may be a necessary and effective impetus for improving the quality of cardiac care. All of the other principles described below should contribute to the overall goal of improving the quality of cardiac care. Informed understanding ---------------------- CRCs should help inform patients about the quality of cardiac care provided within regions, by healthcare organizations, teams and individual health providers. They should provide information that users identify as relevant to their needs and wants. In particular, the information contained within CRCs must be available and comprehensible to the public -- i.e. outcome measures should be provided with an adequate amount of relevant context to interpret data. Providing patients with information is a useful end in itself, even if that information is not utilized by patients in making informed choices. For example, a patient can derive comfort from knowing that his/her health care provider meets an acceptable standard of cardiac care. Report cards should contain physician specific qualitative information, such as a description of patient experiences. Valid and reliable qualitative measures as to the quality of cardiac care will have to be developed. However, physician-specific quantitative measures should not be reported within report cards unless data meet reliable criteria, such as originating from a sufficiently large sample size over a sufficient period of time, utilizing commonly accepted risk-adjustment methods, and including a validation process. Further individual physicians should not be ranked based on the outcomes of these measures. Rather physicians should be identified as meeting or not meeting an acceptable standard of care -- which will have to be defined. Access to information --------------------- Information that is collected for and presented in CRCs must remain public property, and access to this information must be free. Public ownership of report cards may prevent them from perpetuating disparities in access to information. Panelists saw public ownership of report cards as a way of engaging the public and other relevant stakeholders in improving the quality of cardiac care. There should be two types of report cards: one for health care professionals and policy makers, and one for patients and the public. There are a number of reasons why this is necessary. The two audiences have different interests, which necessitates different content, and different levels of understanding, which necessitates different formats. However both types of report cards should remain public property. Those responsible for CRCs should use multiple dissemination methods that will reach all stakeholders. These methods should include physician-patient discussions, open forums and discussion panels. Equity ------ CRCs should address issues of equity. They should include information on the quality of cardiac care and allocation of resources in different health care institutions and geographic regions. Equity interplays with the principle of access to information. Ineffective dissemination methods may exacerbate existing disparities amongst cardiac patients. Living in rural areas or having limited access to computers and the internet should not prevent patients from having access to CRCs. The principle of equity insists that all cardiac patients have access to the information contained within CRCs. Transparency ------------ CRCs should be available to the public because they enhance transparency regarding how the cardiac care system functions. By noting strengths and deficiencies within the cardiac care system, CRCs will enable health care professionals and health policy makers to make necessary changes to enhance the quality of cardiac care, and will enhance public accountability for quality cardiac care. Transparency is also necessary for patients to achieve an informed understanding of the quality of cardiac care, by ensuring consistent access to relevant and accurate information. In this way the principles of transparency, informed understanding, access to information, and quality of information are interrelated. Thus transparency is instrumental because it facilitates other desired outcomes. Public accountability --------------------- Public accountability refers to the obligation on the part of health care professionals to accept responsibility for the quality of care they provide. Public accountability can be operationalized in two ways: (1) accountability to the public independent of public participation; and (2) accountability to the public enforced through a partnership with patients. CRCs can help signify health care providers\' willingness to be accountable for the quality of care they provide and satisfy the public\'s need for accountability. Transparency and access to information are key aspects of public accountability. Multi-stakeholder collaboration ------------------------------- Report cards should be developed in collaboration with all stakeholders, including patients. Those who develop and implement CRCs must involve stakeholders early and throughout the process of report card development, and balance the needs of different stakeholders. In their development and implementation, CRCs can foster collaboration and reduce anger and defensiveness. Multi-stakeholder collaboration and transparency are a means of ensuring that CRCs are viewed as legitimate to all relevant stakeholders. Legitimacy ---------- Legitimacy, the moral authority of CRCs, must be a key aspect of the development and implementation of CRCs. Patients\' views on what is needed in CRCs are necessary to support the principle of legitimacy because patients are at the heart of cardiac care. Patients\' views are inherently important, and outcome measures that address patients\' wants, such as patient experience measures, should be included in report cards. Traditional quality indicators of morbidity and mortality will be insufficient to meet these needs. Like morbidity and mortality, patient experience measures must also be of high quality, realizing the principle of quality of information. The legitimacy of report cards is also derived from transparency and public ownership of the information contained within them. Quality of information ---------------------- Report card data must be of high quality. This means that data must be without bias and risk-adjusted and should also be produced and disseminated by an independent and objective third party. Report cards should provide stakeholders with information on matters that can be addressed by the individuals, groups, or organizations which are being evaluated. Similar institutions should be compared to each other -- for example, an urban teaching hospital should be compared to a similar urban teaching hospital. This sensitivity to the similarity of institutions should also encompass differences in patients\' geographical proximity to institutions. In addition, CRCs must protect patient privacy. The quality of report card data and access to that data are important if report cards are to help initiate changes for improvement in practice sites, or be tied to government funding. Evaluation and continuous quality improvement (of reporting) ------------------------------------------------------------ CRCs should be useful to stakeholders. Continuous quality improvement and effectiveness monitoring initiatives should be in place to ensure report cards meet their intended goals. The measures contained within CRCs should be constantly reviewed, through multi-stakeholder collaborations, to ensure that they provide a fair assessment of the quality of care being evaluated, which requires transparency of the information gathered. Access to report card data by individuals from each stakeholder group is necessary for continuous quality improvement to be achieved. Discussion ========== This study developed an ethical framework to guide the development and implementation of CRCs (see Figure [1](#F1){ref-type="fig"}). To our knowledge, this is the first ethical framework developed for CRCs, or for health care report cards in general. Gormley and Weimer developed a normative framework for report cards that helped identify some key issues \[[@B4]\]. However, its impact was limited because it was not explicitly grounded in moral theory or the systematically described views of stakeholders. Our framework is an advance because it is grounded in ethical theory and in the systematically described views of stakeholders, and thus can provide guidance and is applicable in real life policy and practice. Improving the quality of cardiac care is the primary ethical objective of CRCs. There was consensus on this across all stakeholder groups. Thus, efforts to develop and disseminate a CRC ought to be congruent with this fundamental objective. The panelists identified nine other elements: access to information, informed understanding, equity, transparency, public accountability, multi-stakeholder collaboration, legitimacy, quality of information, and the evaluation and continuous quality improvement of reporting. This ethical framework for CRCs can be used to facilitate the development of report cards. It can provide guidance for addressing difficult issues associated with existing report cards. For example, report cards have been met with anger and defensiveness from clinicians. In response, this framework suggests that clinicians and other stakeholders be involved throughout the process of report card development, giving them opportunities to identify their concerns. The public release of quality of care data has been met with considerable controversy, particularly in the United States \[[@B15]\]. Studies suggest that clinicians are skeptical of data contained in report cards and that such data has little impact on referral decisions \[[@B8]\]. Further, cardiac patients have made limited use of previous CRCs to inform decision-making on their care \[[@B16],[@B17]\]. Questions about the legitimacy of earlier CRCs may be one source of these issues. This framework provides guidance for enhancing the scientific, political and moral legitimacy of report card data. For example, for report cards to be more scientifically and morally legitimate to clinicians, physician-specific quantitative measures should not be reported within report cards unless data meet reliable criteria, such as originating from a sufficiently large sample size over a sufficient period of time, utilizing commonly accepted risk-adjustment methods, and including a validation process. These methods will help ensure the subjects of CRCs are evaluated fairly. And for report cards to be legitimate to patients, outcome measures addressing their information needs must be included, and patients must be involved in the development and implementation of CRCs. The success of qualitative measures that are currently being employed in other areas of health care suggests such measures are possible \[[@B18]\]. The framework suggests that individual providers should be identified as meeting or not meeting an acceptable standard of care, and that physician-specific qualitative information be provided to assist cardiac patients in developing an informed understanding of their care. This view is congruent with findings that patients value information on the interpersonal aspects of care, such as communication and timeliness \[[@B19]\]. Since the framework suggests that clinicians be reported as meeting an acceptable or unacceptable standard of care, and not ranked individually, future report cards should significantly curb some of the incentives for \'gaming\' \[[@B5]\]. In addition, this ethical framework recommends that report card authors involve health care providers in the development of outcomes measures and implement continuous quality reviews of outcomes measures to ensure that they provide a fair assessment of the quality of care being evaluated. The framework suggests that information contained in report cards must remain publicly owned, that the public ought to have access to this information without charge, and that such information remain in the public domain. Public ownership of reports cards was seen as a method of ensuring that report cards do not perpetuate disparities in access to information. It was further viewed as a means of engaging the public and other relevant stakeholders in the process of improving quality of care. In that sense, public ownership can be seen as one means of facilitating public accountability. The items identified in this ethical framework are interrelated. Transparency enhances public accountability by enabling health care professionals and policy makers to identify and correct deficiencies in the quality of cardiac care. Moreover, it is required if patients are to achieve an informed understanding of the quality of cardiac care. Transparency also plays an important role in the legitimacy of CRCs. Together multi-stakeholder collaboration and transparency are a means of securing the legitimacy of CRCs amongst stakeholders. The legitimacy of report cards is also derived from public ownership of the information contained within them. In this ethical framework, the quality of and access to report card data are important if report cards are to help initiate changes for improvement in practice sites, or be tied to government funding. A CRC both describing and comparing the availability of health care to citizens by region can serve, in part, as an indicator of how fairly health care resources are being allocated. Thus CRCs should include information on the quality of cardiac care and allocation of resources in different health care institutions and geographic regions. They are also necessary for continuous quality improvement of report card data. The primary strengths of this study are that it is grounded in ethical theory and in the experiences and views of relevant stakeholders in cardiac care. Our Delphi panel consisted of a diverse range of participants from cardiologists, to cardiac patients, to members of the media. The diversity of our panel lends confidence to the validity of our findings \[[@B13]\]. Although a substantial body of literature on health care report cards exists, our framework is the first to provide ethical guidance on the development and dissemination of such reports. Further, it is the first to be constructed through multi-stakeholder collaboration. Limitations ----------- The primary limitation of this research is its generalizability. Our framework reflects the views of the thirteen panelists who participated in our Delphi rounds, and may not be generalizable to other contexts. However, many of the items in our framework are congruent with the work of Marshall, Romano, and Davies who described strategies to maximize the impact of health care report cards \[[@B20]\], and with Gormley and Weimer\'s normative framework for organizational report cards \[[@B4]\]. This would suggest that our framework might be applicable in other contexts and to other types of health care report cards. Conclusion ========== We have developed an ethical framework to guide the development, implementation and improvement of CRCs. CRCs ought to improve the quality of cardiac care and should be informed by the following items: access to information, informed understanding, equity, transparency, public accountability, multi-stakeholder collaboration, legitimacy, quality of information, and the evaluation and continuous quality improvement of reporting. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= DKM conceived the study and participated in the data analysis and writing the paper. SAR had primary responsibility for data collection and analysis, and writing the paper. SR participated in data collection and analysis, and writing. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1472-6939/6/3/prepub> Acknowledgements ================ This study was supported by grants from The Heart and Stroke Foundation of Canada (HSF) and The Canadian Institutes for Health Research\'s (CIHR) Interdisciplinary Health Research Team Program (IHRT) to the Canadian Cardiovascular Outcomes Research Team (CCORT). DKM is supported by an Ontario Ministry of Health And Long-Term Care Career Scientist award.	PubMed Central	2024-06-05T03:55:55.775967	2005-3-28	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084248/", "journal": "BMC Med Ethics. 2005 Mar 28; 6:3", "authors": [ { "first": "Shawn A", "last": "Richard" }, { "first": "Shail", "last": "Rawal" }, { "first": "Douglas K", "last": "Martin" } ] }
PMC1084249	Background ========== Fear of falling and associated activity restriction are common in elderly persons, both in those older persons who have experienced a fall and those who have not \[[@B1],[@B2]\]. Studies in older people living in the community showed that about 20 to 60 percent of these persons experience at least some fear of falling \[[@B2]-[@B6]\] and about 20 to 55 percent report activity restriction due to fear of falling \[[@B1],[@B5],[@B7]-[@B9]\]. In this paper, we focus on the design of a randomised controlled trial evaluating a cognitive behavioural group intervention, which aims to reduce fear of falling and associated activity restriction in elderly persons. In one of the first reported studies on fear of falling, Vellas and colleagues indicated that fear of falling may lead to a debilitating spiral marked by loss of confidence and reduced activity, resulting ultimately in a loss of independence \[[@B10]\]. In later studies this observation was strengthened. Fear of falling was found to be associated with several adverse factors, including decreased quality of life \[[@B3],[@B11]\], decreased mobility \[[@B3],[@B11]\], functional decline \[[@B3],[@B11],[@B12]\], falls \[[@B13]\], and institutionalisation \[[@B11]\]. These factors may not only have an adverse influence on the physical health status of elderly persons, but on the social and mental health status as well. Therefore, reducing fear of falling and associated activity restriction in older persons may improve their health status. However, until now only a few interventions have been developed and evaluated specifically to reduce fear of falling in elderly people living in the community \[[@B14]-[@B16]\]. In our trial we will evaluate one of these interventions in the Netherlands, a cognitive behavioural group intervention called A Matter of Balance (AMB). This intervention has originally been developed and evaluated in the US and aims to reduce fear of falling and associated activity restriction in elderly persons \[[@B15]\]. The intention-to-treat analysis showed statistically significant effects for increased activity (mobility and intended activity) directly after the intervention. The on-treatment analysis (including participants who attended five or more out of eight sessions) showed statistically significant improvement in falls efficacy, perceived ability to manage falls, and activity (mobility) directly after the intervention. At 12 months, this latter group showed statistically significant improvement in falls efficacy and activity (mobility range and social function). Based on the reported effectiveness of AMB in the US and the aspiration to implement AMB in the Dutch health care system, we decided to perform a trial to assess its effectiveness in the Netherlands. For this purpose, the intervention protocol of AMB was translated and adapted for the Dutch setting (AMB-NL) (Zijlstra et al., development of intervention protocol, submitted). The current paper presents the design of a randomised controlled trial evaluating AMB-NL in Dutch older persons living in the community. Aims ---- The primary aim of the effect evaluation was to study the effects of AMB-NL on fear of falling, avoidance of activity due to fear of falling, and daily activity in older persons living in the community in the Netherlands. The secondary aim was to study the effects of this intervention on perceived general health, self-rated life satisfaction, activities of daily life, feelings of anxiety, symptoms of depression, social support interactions, feelings of loneliness, falls, perceived consequences of falling, and perceived risk of falling. The aim of the process evaluation was to gain insight into factors potentially influencing the effectiveness of the intervention and factors facilitating future implementation of AMB-NL in the Dutch health care setting, if the intervention proves to be effective. Methods/design ============== Study design ------------ A two-group randomised controlled trial with participants being randomly allocated to either an intervention or a control group has been developed to evaluate AMB-NL. Selecting potential participants, conducting the intervention, and collecting data were performed in five consecutive cycles. The first cycle started in November 2002 and the last cycle in July 2003. Each cycle lasted about 18 months and included respectively: screening for eligible participants, baseline measurement, randomisation (allocation to the intervention or control group), the intervention period, a follow-up measurement directly after the intervention period, a booster session at six months after the intervention, a follow-up at six months after the intervention (directly after the booster session), and a final follow-up at 12 months after the intervention. The Medical Ethics Committee of the Maastricht University/Academic Hospital Maastricht granted approval for conducting this trial. The study design is presented in figure [1](#F1){ref-type="fig"}. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Study design ::: ![](1471-2458-5-26-1) ::: Settings and locations ---------------------- Two communities, Heerlen and Maastricht, situated in the southeastern part of the Netherlands were selected for participation in the trial. As screening for potential participants and conducting the intervention was performed in five cycles, both communities were divided into five sections proportional to the number of potential participants in contiguous neighbourhoods. Intervention sessions were conducted in local community centres or homes for the elderly preferably located in the centre of those neighbourhoods. Transportation to the intervention location was offered to participants who expected difficulties to reach the location, for example due to health problems. Participants ------------ Elderly persons were eligible for study if they met all of the following criteria: 1) reporting at least some fear of falling; 2) reporting at least some associated avoidance of activity; 3) living in the community; and 4) being 70 years of age or over. Elderly persons confined to bed, restricted by permanent use of wheelchair, or waiting for nursing home admission were excluded for study. Immediately before screening for eligible participants in each of the five cycles, addresses of older people living in the community who were 70 years of age or over at 1 January 2003 were randomly selected by municipal registry offices. Selected persons were sent information about the trial and a brief self-administered screening questionnaire. This questionnaire assessed socio-demographic and fall-related variables, all inclusion and exclusion criteria, and willingness to participate. Persons interested in participating in the trial were asked to sign an informed consent form enclosed in the questionnaire. All selected persons were requested to complete and return the questionnaire, even if they lacked interest in participating in the trial. A postage free envelope was enclosed for returning the questionnaire. If the questionnaire was not returned in a fortnight, a reminder letter to return the questionnaire was sent. Those persons who signed the informed consent form, were willing to participate, and met all other eligibility criteria were invited to participate in the study. Randomisation ------------- Randomisation was carried out directly after baseline measurement and was performed per community to ensure having both an intervention group and a control group in each of the two communities. During each cycle two intervention groups were composed per community. Per cycle, approximately half of the participants were allocated to the intervention group (with a maximum of 15 participants per intervention group). Participants allocated to the intervention group were then randomly allocated to one of the two groups in their own community. Participants allocated to the control group received no intervention as a result of this trial. An independent researcher conducted randomisation by selecting random samples using SPSS12.0 for Windows. Intervention ------------ The intervention AMB-NL is a translated and adapted version (Zijlstra, et al., development of intervention protocol, submitted) of a cognitive behavioural group intervention for older persons living in the community developed by Tennstedt and colleagues \[[@B15]\]. This intervention has been developed to reduce fear of falling and promote physical, social, and functional activity in elderly persons living in the community. Principles of cognitive restructuring \[[@B17]\] are applied by focusing on changing attitudes and self-efficacy beliefs with respect to falling before attempting to change actual behaviour. To attain a reduction in fear of falling, the intervention aims to increase self-efficacy beliefs with regard to falling as well as the sense of control over falling. Four strategies are used to accomplish these aims: (1) restructuring misconceptions to promote a view of fall risk and fear of falling as controllable; (2) setting realistic goals for increasing activity; (3) changing the environment to reduce fall risk; and (4) promoting physical exercise to increase strength and balance \[[@B15]\]. The intervention consists of eight weekly group sessions lasting two hours. Six months after the eighth session a booster session is scheduled. The main topics in each of the sessions of AMB-NL, presented in table [1](#T1){ref-type="table"}, were discussed similarly: (1) introduction; (2) participant\'s point of view; (3) positive and negative aspects concerning the topic; (4) association with falls or fear of falling; and (5) implementation in the participant\'s daily life. A more extensive description of the intervention will be published elsewhere (Zijlstra et al., development of intervention protocol, submitted). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Main contents of the Dutch version of AMB (AMB-NL) ::: ----------------------------------------------------------------------------------------------------- Session contents of AMB-NL Session 1: Introduction to the Program Starting a group intervention (e.g. getting acquainted) Background information on fear of falling (e.g. incidence, impact) Beliefs and disbeliefs about fear of falling Shifting from negative to positive thinking patterns Finding personal solutions to deal with fear of falling Session 2: Exploring Thoughts and Concerns about Falling Attitudes related to fear of falling and challenging them Adaptive responses to counter misconceptions about falls Unhelpful thoughts and their effects regarding to feelings and behaviour Shifting from self-defeating to self-motivating thoughts Session 3: Exercise and Fall Prevention Misconceptions regarding physical exercise for elderly people Potential consequences of inactivity and benefits of physical activity Staying or becoming physically active to prevent falls Recognising and overcoming barriers to stay or become physically active Appropriate physical exercises for elderly people and fall prevention Practicing simple physical exercises Session 4: Assertiveness and Fall Prevention Association between assertiveness and fall prevention Potential benefits of being assertive Reducing fall risk by being assertive in a proper fashion Addressing physical risk factors for falls The influence of physical exercise on physical characteristics (e.g. blood pressure) Practicing physical exercises Session 5: Managing Concerns about Falling Developing and implementing a personal physical exercise program Shifting from self-defeating to self-motivating thoughts regarding physical activity and fall risk Practicing physical exercises Midcourse evaluation to review all main topics Session 6: Recognising Fall-ty Habits Identifying and managing risk-taking behaviour in daily life Prioritising fall risk behaviours Searching for suitable, personal solutions to change risk-taking behaviour into safe actions Planning behaviour change strategies Setting goals for activities one would like to carry out Shifting from negative thoughts associated with planned activities to positive responses Practicing physical exercises Discussing falls and seeking help after a fall Session 7: Recognising Fall Hazards in the Home and Community Potential fall hazards in homes and community Recognising and eliminating environmental hazards by finding simple solutions Discussing displayed assistive devices which improve safety Practicing physical exercises Session 8: Practicing No Fall-ty Habits Practicing assertiveness skills for locating and utilising resources to prevent falls Understanding that risk-taking behaviour can be eliminated Practicing physical exercises Booster session Discussing personal experiences with falls and fear of falling Shifting from self-defeating to self-motivating thoughts Exercise and fall prevention Potential fall hazards in homes and community Change risk-taking behaviour into safe actions Practicing physical exercises ----------------------------------------------------------------------------------------------------- ::: Nurses qualified in the field of geriatrics and working for home care agencies were trained as facilitators of the intervention. Except for the first session when two facilitators were present, each intervention session was conducted by one facilitator. Monthly meetings with facilitators and researchers were scheduled to evaluate and discuss the progress of the trial, the intervention and associated matters. Participants were informed to notify the facilitators or researchers if they were unable to attend a session. After the session, facilitators contacted those participants who were absent and briefly discussed the topics of that session. However, those participants who were absent in all of the first three sessions were excluded from further participation in the intervention due to an unbridgeable deficiency in knowledge. Measures -------- ### Effect evaluation #### Primary outcome variables The primary outcomes of the effect evaluation are fear of falling, avoidance of activity due to fear of falling, and daily activity. Fear of fallingwas assessed by three different measures. First, respondents indicated the frequency of fear of falling when asked \"Are you afraid of falling?\" (1 = never to 5 = very often). Second, on the 10-item Falls Efficacy Scale (FES) respondents indicated how worried they are about falling while carrying out several indoor activities of daily living (1 = not at all worried to 4 = very worried) \[[@B7],[@B18]\]. Four items about outdoor activities \[[@B19]\] were added. Finally, a four-item scale was used to assess the respondent\'s perceived control over falling (PCOF) \[[@B20]\]. This scale focuses on the perceived control over the environment and one\'s own mobility and ability to do things to prevent falls and reduce fear of falling. The frequency of avoidance of activitydue to fear of falling was assessed by the question \"Do you avoid certain activities due to fear of falling?\" (1 = never to 5 = always). Daily activitywas assessed by the Frenchay Activities Index (FAI) \[[@B21],[@B22]\]. The FAI measures the frequency in which daily activities that reflect the broader everyday activities of normal living are performed \[[@B21]\]. An overview of the primary and secondary outcomes measured during the course of the study is presented in table [2](#T2){ref-type="table"}. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Primary and secondary outcome measures of the effect evaluation ::: Variables No. of items Range\* S B FU1 FU2 FU3 ------------------------------------------------------------- -------------- ------------------------ -- ---- -- ----- -- ----- -- ----- -- ----- Primary outcome measures fear of falling 1 [1]{.underline} to 5 SQ Q Q Q Q fall-related self-efficacy (FES) \[7\] 10 [10]{.underline} to 40 \- TI TI TI TI outdoor items fall-related self-efficacy \[19\] 4 [4]{.underline} to 16 \- TI TI TI TI perceived control over falling (PCOF) \[20\] 4 [4]{.underline} to 20 \- Q Q Q Q avoidance of activity due to fear of falling 1 [1]{.underline} to 5 SQ Q Q Q Q daily activity (FAI) \[21, 22\] 15 15 to [60]{.underline} \- Q Q Q Q Secondary outcome measures perceived general health (MOS SF-20 item one) \[23, 24\] 1 [1]{.underline} to 5 SQ Q Q Q Q self-rated life satisfaction \[25\] 1 1 to [7]{.underline} \- Q Q Q Q activities of daily life (ADL subscale of the GARS) \[26\] 11 [11]{.underline} to 44 \- TI TI TI TI feelings of anxiety (HADS) \[27, 28\] 7 [0]{.underline} to 21 \- Q Q Q Q symptoms of depression (HADS) \[27, 28\] 7 [0]{.underline} to 21 \- Q Q Q Q social support interactions (SLL12-I) \[29\] 12 12 to [48]{.underline} \- Q Q Q Q feelings of loneliness 1 1 to [6]{.underline} \- Q Q Q Q number of falls in the previous 6 months 1 [1]{.underline} to 6 SQ \- \- Q Q number of falls in the previous 2 months 1 [1]{.underline} to 6 \- Q Q \- \- number of indoor falls 1 N/A \- C\> C\> C\> C\> number of outdoor falls 1 N/A \- C\> C\> C\> C\> number of times medical attention required due to falls 1 N/A \- C\> C\> C\> C\> perceived consequences of falling (CoF) \[30\] 12 [12]{.underline} to 48 \- TI TI TI TI perceived risk of falling (RoF) \[30\] 3 [3]{.underline} to 12 \- TI TI TI TI \* The underlined scores indicate the most favourable scores. S = screening; B = baseline; FU1 = direct follow-up; FU2 = 6-month follow-up; FU3 = 12-month follow-up SQ = screening questionnaire; Q = questionnaire; TI = telephone interview; C\> = calendar (continuous registration) N/A = not applicable ::: #### Secondary outcome variables The secondary outcomes that were assessed are: perceived general health(item one of the MOS SF-20) \[[@B23],[@B24]\], self-rated life satisfaction(seven point satisfaction rating) \[[@B25]\], activities of daily life(ADL subscale of the Groningen Activity Restriction Scale (GARS)) \[[@B26]\], feelings of anxiety(Hospital Anxiety and Depression Scale (HADS)) \[[@B27],[@B28]\], symptoms of depression(HADS) \[[@B27],[@B28]\], social support interactions(SSL12-I) \[[@B29]\], feelings of loneliness(6 point Likert scale), falls, perceived consequences of falling(CoF) \[[@B30]\], and perceived risk of falling(RoF) \[[@B30]\]. Feelings of lonelinesswere assessed by the question \'During the past four weeks, how often did you feel lonely?\' (1 = all the time to 6 = never). Fallswere registered by both a one-item question and a fall calendar. The one-item question assessed how frequently the participant had fallen during the past few months (1 = never to 6 = 5 or more falls). The fall calendar was used for continuous registration of falls during the course of the trial. If a fall occurred, participants indicated on the calendar: (a) the location of the fall (indoor or outdoor); and (b) the number of times medical care was received due to the fall. #### Additional variables Several additional variables were assessed to provide insight into the population, to interpret the outcomes of the trial, and to study the underlying mechanisms of the intervention. Socio-demographic variablesassessed during the process of screening were: age, gender, marital status, living condition, living alone or not, and educational level. Health-related variablesassessed during the telephone interview of the baseline measurement were: chronic medical conditions (a 19-item checklist) \[[@B31]\], cognitive status (modified version of the Telephone Interview for Cognitive Status (TICS)) \[[@B32]\], and impaired vision and hearing (a four-item questionnaire) \[[@B33]\]. Other health-related variables assessed at baseline and follow-up telephone interviews were: use of healthcare (for example, number of visits to the general practitioner) and use of assistive devices. Additional dataassessed with baseline and follow-up questionnaires were: general self-efficacy (GSE) \[[@B34],[@B35]\], physical self-efficacy (PSE) \[[@B36]\], social self-efficacy (SSE) \[[@B34]\], and mastery \[[@B37]\]. An overview of the additional health-related scales and self-efficacy and mastery scales is presented in table [3](#T3){ref-type="table"}. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Additional measures during the trial ::: Description of the variables No. of items Range\* B FU1 FU2 FU3 ---------------------------------------- -------------- ------------------------ -- ---- -- ----- -- ----- -- ----- chronic medical conditions \[31\] 19 [0]{.underline} to 19 TI \- \- \- cognitive status (TICS) \[32\] 25 0 to [41]{.underline} TI \- \- \- impaired vision and hearing \[33\] 4 [4]{.underline} to 16 TI \- \- \- use of health care 6 N/A TI TI TI TI use of assistive devices 14 N/A TI TI TI TI general self-efficacy (GSE) \[34, 35\] 16 16 to [80]{.underline} Q Q Q Q physical self-efficacy (PSE) \[36\] 10 10 to [50]{.underline} Q Q Q Q social self-efficacy (SSE) \[34\] 6 6 to [30]{.underline} Q Q Q Q mastery \[37\] 7 7 to [35]{.underline} Q Q Q Q \* The underlined scores indicate the most favourable scores. B = baseline; FU1 = direct follow-up; FU2 = 6-month follow-up; FU3 = 12-month follow-up Q = questionnaire; TI = telephone interview N/A = not applicable ::: ### Process evaluation The process evaluation was aimed at gaining insight into factors potentially influencing the effectiveness of the intervention and factors facilitating future implementation of the intervention. Four main outcome measures were identified: (1) performance of the intervention according to protocol; (2) attendance of participants; (3) adherence of participants; and (4) opinion of participants and facilitators about the intervention. Table [4](#T4){ref-type="table"} provides an overview of the outcomes of the process evaluation during the course of the trial. ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Outcome measures of the process evaluation ::: Variables BDI FU1 FU2 FU3 ----------------------------------------------------------- ------- ----------- ------- ------ Performance intervention according to protocol duration of the sessions R^f^ \- \- \- deviations from protocol R^f^ \- \- \- Attendance of participants reasons for refusal before the start of the intervention TI^p^ \- \- \- number of sessions visited by each subject R^f^ \- R^f^ \- reasons for stopping during the intervention period TI^p^ \- TI^p^ \- Adherence of participants adherence to homework assignments \- Q^p^/Q^f^ \- \- adherence to physical exercise \- Q^p^/Q^f^ Q^p^ Q^p^ Opinion about intervention overall judgement about the intervention \- Q^p^/Q^f^ \- \- judgement about the facilitators \- Q^p^/Q^f^ \- \- benefit experienced by participants \- Q^p^/Q^f^ Q^p^ Q^p^ strong and weak aspects of the intervention \- Q^p^/Q^f^ \- \- recommendations for improvement \- Q^p^/Q^f^ Q^f^ \- BDI = before or during intervention; FU1 = direct follow-up; FU2 = 6-month follow-up; FU3 = 12-month follow-up R = registration form filled in after each session; Q = questionnaire; TI = telephone interview Data collected from: ^f^= facilitator; ^p^= participant ::: Data collection --------------- Data for the effect evaluation were gathered by means of self-administered questionnaires, fall calendars, and telephone interviews. Trained interviewers, who were blinded for group allocation, conducted the interviews. After baseline measurement participants received a fall calendar; every three months a sheet of the calendar was to be returned to the research team. Data for the process evaluation collected from the participants were completed by means of self-administered questionnaires and short telephone interviews. Registration forms and self-administered questionnaires were used to gather data from the facilitators. These data were discussed and illustrated by the facilitators in two evaluative meetings. As recommended by Hollis and Campbell \[[@B38]\], non-compliant participants of the intervention group were approached for all follow-up measurements and participants with missing data were contacted to ensure completion of data. At five and 11 months after the intervention, newsletters were sent to keep the participants interested in participating in the trial and to notify them about its progress. Sample size and power --------------------- Sample size calculations were based on experiences in evaluative studies among older persons in the Netherlands \[[@B39]\] and the US \[[@B15]\] where fall-related self-efficacy \[[@B7]\] was assessed. To detect a mean difference of at least 2.5 points on fall-related self-efficacy between the intervention and control group (SD is 6.0; equivalent to an effect size of .42), at least 117 participants in each group were necessary to meet a power of 80 percent at alpha .01 (one-sided). Taking a dropout rate of 30 percent during the trial into account, a total number of 180 participants per group (n = 360) were needed to enrol in the trial. Based on unpublished data of Van Haastregt and colleagues (study described in \[[@B39]\]), we estimated that approximately 14.5 percent of the older persons who returned the screening questionnaire would meet all inclusion criteria and exclusion criteria and would be interested in participating in the trial. With an estimated response rate of 55 percent, a minimum sample of 4500 older persons aged 70 or over needed to be approached with a screening questionnaire. Analysis -------- Descriptive techniques will be used to describe participants of the trial. To detect differences between intervention and control group, baseline variables will be compared using the appropriate tests. Baseline variables of compliant and non-compliant participants of the intervention group will be compared as well. Compliant participants are participants who attended five or more sessions \[[@B15]\], not including the booster session. Data of the effect evaluation will be analysed according the intention-to-treat and on-treatment principle. Univariate and multivariate techniques will be applied to examine the differences in the intervention and control group with regard to the primary and secondary outcome measures at the follow-up measurements. Effect sizes \[[@B40]\] will be calculated to quantify the size of the difference between both groups. Subgroup analysis will be performed with several potential effect modifiers, like cognitive status or educational level. Data of the process evaluation will be analysed using descriptive techniques. Discussion ========== Progress of the study --------------------- A random selection of older persons aged 70 or over living in the community was screened for eligibility in five cycles between November 2002 and July 2003. A total of 7431 older persons rightfully received the screening questionnaire. The response rate was 58.9 percent. During the course of the study the number of people who were sent a screening questionnaire was increased based on two grounds. First, in contrast to our estimate of 14.5 percent, only about 10 percent of the participants in the first cycle met our eligibility criteria and were willing to participate. Second, about 25 percent of the persons participating in the baseline measurement dropped out before randomisation. In total, 540 participants were included in the trial; 260 participants were allocated to the intervention group and 280 to the control group. The data collection was completed in February 2005. Currently, preparations for the analyses of the data of the screening and process evaluation are being made. Data of the effect evaluation will be available in 2005. Future implementation --------------------- Implementation in the Dutch setting has been taken into account throughout the development of AMB-NL (Zijlstra et al., development of intervention protocol, submitted). If the results of the trial show the effectiveness of AMB-NL, recommendations will be developed to implement the intervention in the Dutch health care setting and a manual of the intervention, updated with the experiences of the trial, will be made available. Abbreviations ============= AMB -- A Matter of Balance AMB-NL -- the Dutch version of A Matter of Balance Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= All authors contributed to the development of the design of this study. G.A.R. Zijlstra drafted the manuscript with input from the other authors. All authors read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2458/5/26/prepub> Acknowledgements ================ We would like to thank the Faculty of Health Sciences of the Universiteit Maastricht, the Care and Public Health Research Institute, and ZonMw -- The Netherlands Organisation for Health Research and Development (grant 014-91-052) for funding this study.	PubMed Central	2024-06-05T03:55:55.778217	2005-3-21	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084249/", "journal": "BMC Public Health. 2005 Mar 21; 5:26", "authors": [ { "first": "GAR", "last": "Zijlstra" }, { "first": "JCM", "last": "van Haastregt" }, { "first": "JThM", "last": "van Eijk" }, { "first": "GIJM", "last": "Kempen" } ] }
PMC1084250	Background and clinical perspectives ==================================== Normally coronary blood flow can increase approximately four-to-six fold to meet increasing myocardial oxygen demands. This effect is mediated by vasodilation at the arteriolar bed, which reduces vascular resistance, thereby augmenting flow. The coronary reserve (CFR) represents the capacity of the coronary circulation to dilate following an increase in myocardial metabolic demands and can be expressed by the difference between the hyperemic flow and the resting flow curve (figure [1](#F1){ref-type="fig"}). In 1974, Lance K Gould \[[@B1]\] proposed the relationship between the anatomic condition and behaviour of coronary hyperaemic flow (figure[2](#F2){ref-type="fig"}), whereby an inverse curvilinear relationship exists between the narrowing of lumen of coronary artery and hyperaemic capability, up to a completely abolished coronary reserve for stenosis \>90%. This experimental paradigm can be accurately reproduced clinically in highly selected series of patients with single vessel disease, no myocardial infarction, no coronary collateral circulation, normal baseline function, no left ventricular hypertrophy, without evidence of coronary vasospasm, and off therapy at the time of testing. The perfect, predictable relationship found in the experimental animal or in a very selected patient population \[[@B2]\] is not so perfect in clinical practice \[[@B3],[@B4]\], where many variables can modulate the imperfect match between epicardial coronary artery stenosis and coronary flow reserve (figure [2](#F2){ref-type="fig"}), such as the geometric characteristics of the stenosis, the presence of coronary collateral circulation, the microvascular component of coronary resistance, the presence of left ventricular hypertrophy modulating the myocardial extravascular component of coronary resistance, the viable or necrotic state of the myocardium distal to the stenosis, the presence of coronary macrovascular or microvascular spasm, and, last but not least, the presence of concomitant anti-ischemic therapy. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Schematic representation of coronary flow velocity profile obtained with tranthoracic Doppler of distal left anterior descending coronary artery : in diastole the flow velocity is higher than in systole. ::: ![](1476-7120-3-8-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Relationship between the true increments of the flow signal obtained with the currently available imaging techniques. Modified from Gould KL, ref.1. On the abscissa are represented different narrowing of the coronary vessel. ::: ![](1476-7120-3-8-2) ::: Coronary flow reserve in the stress- echo lab: a new approach to CAD ==================================================================== Up to now, coronary flow reserve has been evaluated invasively in the cath. lab and in nuclear medicine through perfusion imaging. Only recently has coronary flow reserve entered the echo lab, with the combination of coronary flow assessment by Doppler and vasodilator stress. With either TEE (sampling proximal tract) (figure [3](#F3){ref-type="fig"}) or TTE (exploring mid-distal tract) (figure [4](#F4){ref-type="fig"}), the coronary blood flow velocity profile recoded with pulsed wave Doppler is consistent with the pathophysiological premises. Accordingly, coronary flow velocity by Doppler assessment appears to be biphasic, with a lower peak during systole and a higher peak during diastole (figure [1](#F1){ref-type="fig"}). Myocardial extravascular resistance is higher in systole and lower in diastole due to the effect of myocardial contraction. The flow velocity variations are proportional to the total blood flow if the vessel lumen is kept constant, a reasonable assumption with the administration of drugs such as dipyridamole or adenosine. The coronary flow velocity variation between the baseline and peak effect of a coronary vasodilator allows a coronary flow reserve index in the left anterior descending artery territory to be derived. Peak diastolic flow is the simplest parameter to be measured and the most easily obtained, in addition to being the most reproducible and the one with the closest correlation with coronary perfusion reserve measured by positron emission tomography. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Visualization of left main and bifurcation of left anterior descending coronary artery and circumflex assessed by transesophageal approach. The color-Doppler trace the flow inside the proximal tract of left coronary artery Left: the Pulsed wave Doppler highlights the typical biphasic flow velocity coronary pattern ::: ![](1476-7120-3-8-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Artist\'s drawing illustrating transducer beam orientations to the left anterior descending coronary artery (LAD) with the corresponding echocardiographic images of the mid-distal tract of LAD color flow. ::: ![](1476-7120-3-8-4) ::: The coronary flow signal on LAD was first made possible by TEE (figure [3](#F3){ref-type="fig"}) -- with excellent diagnostic results \[[@B5],[@B6]\] -- but only more recently has there been an increase in clinical interest due to the development of the transthoracic method \[[@B7]-[@B12]\]. There were technological factors which allowed the totally noninvasive transthoracic imaging of mid-distal LAD (figure [4](#F4){ref-type="fig"}): second harmonic imaging, allowing better definition of smaller structures, such as LAD; high frequency transducers (up to 8 MHz in second harmonic), leading to improved resolution imaging of near-field structures. The availability of contrast agents also improved the signal-to-noise ratio, thereby increasing the feasibility of transthoracic imaging of LAD above the threshold of potential clinical impact, although it is true that after a training period its use may not be necessary. The Doppler assessment of coronary flow reserve has some limitations. The assessment of absolute blood velocity can be limited in some patients by the large incident angle between the Doppler beam and blood flow. However, calculation of the flow reserve allows assessment of flow patterns without the need for absolute values. More importantly, the velocity ratio is used as a surrogate of flow reserve: flow within the coronary artery is not calculated because cross-sectional visualization of the vessel does not allow an accurate measurement of the diameter of the vessel. The estimated flow reserve can be accurate if the coronary functions only as a conduit, without changing in diameter during drug infusion. This assumption is reasonable with dipyridamole \[[@B6]\] and less valid with dobutamine: this is an additional reason to stress coronary flow reserve with vasodilators. Coronary flow reserve: a new diagnostic power ============================================= The use of CFR as a \"stand-alone\" diagnostic criterion suffers from so many structural limitations as to make it little more than an academic somersault: firstly, only LAD is sampled; secondly, the coronary flow reserve cannot distinguish between microvascular and macrovascular coronary disease \[[@B13]\]. Therefore, it is much more interesting (and clinically realistic) to evaluate the additive value over conventional wall motion for LAD detection. The assessment of CFR adds sensitivity for LAD disease -- with a modest loss in specificity. In reality, the inherently quantitative information of LAD flow reserve allows a stratification of the response, integrating many different tests into one: greatly reduced CFR (\<1.5) yields extraordinary specificity whilst mildly reduced CFR (\<2.0) offers outstanding sensitivity (Table [1](#T1){ref-type="table"}). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Diagnostic value of 2D Echo and coronary flow reserve ::: Sensitivity CI 95% Specificity CI 95% Accuracy CI 95% ------------------------------------------- ----------------- ------------ ----------------- ------------ -------------- ------------ 2D Echo 74% 64--84% 91% 87--96% 86% 82--91% Coronary Flow Reserve (cut-off = 2) 89% 81--96% 77% 71--84% 81% 76--86% Coronary Flow Reserve (cut-off = 1,9) 81% 72--90% 84% 79--90% 83% 79--88% Coronary Flow Reserve (cut-off = 1,8) 69% 58--79% 90% 85--95% 83% 79--88% Coronary Flow Reserve (cut-off = 1,7) 63% 52--74% 97% 94--99% 86% 82--91% Coronary Flow Reserve (cut-off = 1,6) 50% 38--61% 100% \- 85% 80--89% Coronary Flow Reserve (cut-off = 1,5) 30% 19--41% 100% \- 79% 73--84% 2D Echo / CFR cut-off = 1,9 90% 81--98% 94% 91--98% 93% 89--97% ::: In addition, the flow information is relatively unaffected by concomitant antianginal therapy, which markedly reduces the sensitivity of ischemia-dependent regional wall motion abnormality \[[@B14]\] and does not influence coronary flow reserve, or does so only to a limited extent \[[@B15],[@B16]\]. As a result, CFR can already help in the difficult task of identifying patients with coronary artery disease in accordance with the classic ischemic cascade (figure [5](#F5){ref-type="fig"}). ::: {#F5 .fig} Figure 5 ::: {.caption} ###### The classical ischemic cascade, triggered by coronary vasospasm and/or epicardial stenosis. The various markers are usually ranked according to a well-defined time sequence. ::: ![](1476-7120-3-8-5) ::: Clinical application of coronary flow reserve ============================================= In our experience of a single center since 2000, the combination of the CFR study on LAD and the analysis of left ventricular wall motion abnormalities represents the best choice in echo-lab evaluation of the flow-function relationship \[[@B17]\]. This can be done with the employment of a high frequency (5--7 MHz) probe aimed at investigating the coronary flow on LAD, and a low frequency (1,8--2,5 MHz) probe aimed at investigating the deeper wall motion behaviour. The presence of microvasculature dysfunction can create a bias between these two parameters: as a result, the WMSI guarantees good specificity while the CFR guarantees an improvement in sensitivity (Table [1](#T1){ref-type="table"}). When the CFR cut-off value was decreased to below 2, the close relationship became even more evident (Table [1](#T1){ref-type="table"}). Coronary flow reserve integrates and complements, but cannot be considered alternative to, classical stress echo based on regional wall motion analysis. Coronary flow reserve in the echo lab is not a \"stand-alone\" variable. In the echo lab, there is already stress echo information on wall motion: every additional piece of information, which adds to the overall complexity, should be evaluated in terms of its ability to provide additional information on wall motion analysis, which has high diagnostic accuracy and strong prognostic stratification power. Theoretically, and on the basis of the classic and alternative cascade (figure [6](#F6){ref-type="fig"}), coronary flow reserve information can be especially helpful for mild to moderate stenosis (capable of reducing flow reserve, but to subischemic levels) and in identifying patients with microvascular disease (reduced flow reserve and normal coronary arteries). ::: {#F6 .fig} Figure 6 ::: {.caption} ###### The alternative ischemic cascade, triggered by microvasculature dysfunction. The various markers are a different time sequence in comparison with the classical ischemic cascade. ::: ![](1476-7120-3-8-6) ::: The combination of regional wall motion and coronary flow reserve identifies distinct patterns with 2 different parameters, coronary flow reserve and regional wall motion analysis. At one end of the spectrum, there is the totally normal pattern, with normal left ventricular function and normal coronary flow reserve, which is highly predictive of normal coronary anatomy and normal physiological response of coronary micro and macrocirculation. At the opposite end of the spectrum, there is abnormal left ventricular functional and abnormal coronary flow response, which is highly predictive of diseased epicardial coronary anatomy and impaired flow reserve. In between these extreme black-and-white responses, \"gray zone\" responses can be found, with mild-to-moderate abnormal coronary flow reserve and normal function with normal coronary flow reserve. From PubMed search, we identified all papers with head-to-head comparison of dipyridamole stress echo (0.84 mg/kg ± atropine, with standard wall motion analysis) versus dipyridamole stress echo (0.84 mg/kg in 10\' plus atropine or 0.84 mg/kg in 6\' without atropine) with wall motion analysis and CFR evaluation. A total of 5 papers (from Italy, Argentina, and Japan) were found, involving 725 patients (355 with LAD stenosis and 370 without CAD \[[@B17]-[@B21]\](table [2](#T2){ref-type="table"}). ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Meta-analysis on CFR and WMSI diagnostic value ::: SENSITIVITY (%) SPECIFICITY (%) ACCURACY (%) ----------------------------------- --------------------- --------------------- ------------------ ------------------- ----------- -------------------- *DIP-2D* *Dip-2D + CFR* *DIP-2D* *Dip-2D +CFR* *DIP* *Dip-2D + CFR* Rigo et al, Am J Cardiol 2003 74 90 91 94 82 93 Lowenstein et al, JASE 2003 69 87 91 73 81 80 Nohtomi et al, JASE 2003 72 93 95 70 82 83 Chirillo et al, AJC 2004 67 93 91 93 71 93 Ascione et al, Int J cardiol 2004 51 83 96 98 78 94 67 ± 9 90 ± 3 93 ± 2 86 ± 12 79 ± 5 89 ± 7 ::: If we consider the latest papers on the diagnostic role of the Dipirydamole stress-test, it becomes clear that by adding the CFR-LAD evaluation to wall motion analysis we significantly increase the sensitivity of the test, whilst maintaining an excellent value in terms of specificity. As a result, diagnostic accuracy value is also extremely good. Up to now, this kind of relationship has only been valid for the LAD coronary artery, although this of course represents the most important coronary vessel in terms of clinical and prognostic impact. Coronary flow reserve: the importance of being quantitative =========================================================== Coronary circulation has in itself the possibility of modulating blood flow variation in response to metabolic requests. The magnitude of this adaptive variation may be objectively expressed as a number such as coronary flow velocity reserve. The potential impact of this number still has to be clarified in the clinical arena. Since 2000 we have evaluated the CFR-LAD behaviour of 1235 patients consecutively in our stress echo lab. We obtained the values summarized in Table [3](#T3){ref-type="table"}, ranging from athletes to normal subjects and in the different patient subsets. It is important to underline the fact that different pathologies can give the same values in terms of CFR. Consequently, this functional parameter also has its own gray zone where it is often not possible to discriminate between dysfunction due to microvasculature or epicardial coronary disease. In our experience we found a significant impairment of CFR (\< 2), in the three major cardiac pathologies with microvasculature dysfunction; in percentage terms, in 26% for Syndrome X, in 21% for Hypertrophic Cardiomyopathy and in 41% for Dilated Cardiomyopathy. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Results on CFR in different pathologies ::: Number Male/Female Mean age(years) CFR ---------------------- -------- ------------- ----------------- ------------- Normal patients 76 47/29 39 ± 12 3,32 ± 0,3 Syndrome X 97 24/73 57 ± 17 2,27 ± 0,3 LAD (≥ 70%) 223 171/152 63 ± 16 1,38 ± 0,2 LAD (\<70%) 128 84/44 62 ± 16 2,2 ± 0,24 Hypertensive pts 323 72/251 56 ± 17 2,46 ± 0,44 DC 48 29/19 64 ± 112 1,94 ± 0,24 HCM 44 35/9 53 ± 11 2,21 ± 0,23 Aortic stenosis 22 6/14 74 ± 13 2,18 ± 0,34 Aortic insufficiency 12 5/7 68 ± 12 2,57 ± 0,40 PCI- LAD (\>3 mo.) 72 51/21 61 ± 16 2,52 ± 0,45 Graft-IMA (\>3 mo.) 56 41/15 64 ± 14 2,60 ± 0,38 Post-AMI (\>3 mo.) 93 69/24 68 ± 17 1,98 ± 0,41 Athletes 41 41 34 ± 12 4,5 ± 0,45 LAD = Left anterior descending coronary artery; Syndrome X = microvasculature dysfunction; DC = Dilated cardiomyopathy; HCM = Hypertrophic cardiomyopathy; PCI-LAD = after LAD angioplasty (\>3 months); Graft-IMA = By-pass with internal mammary artery on LAD; Post-Ami = after anterior myocardial infarction not revascularized. ::: Prognostic role of coronary flow reserve ======================================== The additional diagnostic value of coronary flow reserve is represented by the possibility of monitoring different heart pathologies objectively and thereby obtaining important functional information over time in patient follow-up. Although the importance of the prognostic role played by stress-echo in patients with coronary artery disease has been demonstrated, there has also been emphasis on how the presence of antiischemic therapy at the time of testing can heavily modulate the predictive value of pharmacological stress-echo. In fact, a positive test in therapy is more prognostically malignant, and a negative test less prognostically benign \[[@B22]\]. The potential prognostic role of coronary flow reserve has recently been tested in predicting different clinical situations through invasive and nonivasive approaches. In particular, it has been demonstrated that the patency over time of coronary vessel disease after coronary angioplasty (PCI) can be accurately predicted by evaluating the functional status of the revascularized coronary artery just after the procedure \[[@B23]-[@B26]\]. Recently, we have emphasized the added prognostic role of an impairment of CFR despite normal wall motion contractility during combined dipyidamole stress-echo: these patients showed a worsening outcome during a mean follow-up of 24 months \[[@B27]\]. A further important application of CFR is as a good predictor of adverse events regarding the relationship with left ventricular remodelling after anterior myocardial infarction treated with coronary angioplasty \[[@B26]-[@B29]\]. In patients with coronary microvasculature dysfunction such as Dilated Cardiomyopathy (DC) \[[@B30]-[@B32]\] and Hypertrophic cardiomyopathy (HCM) \[[@B32],[@B33]\], an impairment of CFR allows us to identify those patients with a worsening outcome and therefore represents an important guide to the efficient management of these patients. Even if we need confirmation of the prognostic role of CFR through further and larger study, this parameter could have a very useful role to play in daily clinical practice. Multi coronary flow reserve: the next ultrasound challenge ========================================================== At present, the impossibility of discriminating pathological behaviour of CFR due to microvascular or epicardial stenosis has represented a major limitation in the final diagnosis, as has the impossibility of investigating all the three major coronary arteries simultaneously \[[@B34]\]. We are trying to overcome this bias by introducing the contemporary analysis of at least two coronary arteries \[[@B35],[@B36]\]. The most promising application regards right coronary evaluation, consisting of an apical off-axis 2 chamber approach: a high frequency probe (2\^ Harmonic ;7 MHz receiving) \[[@B37]\] is rotated in a counter-clockwise position compared to the classical 3--4 chamber approach (figure [7](#F7){ref-type="fig"}). By applying this method in 658 consecutive patients (389 males; age 64,3 ± 13 years) referred for stress echocardiography, we were able to recognize a good right coronary flow (Additional file [1](#S1){ref-type="supplementary-material"}) and reserve in 429 pts (66%). We recognized the LAD-CFR (Additional file [2](#S2){ref-type="supplementary-material"}) in 637 pts (98%) from the same group: for Circumflex-CFR (Additional file [3](#S3){ref-type="supplementary-material"}) we were able to recognize a color flow signal in 344 pts (53%) but a good Pulse -Wave Doppler spectral in only 43% of patients. This last coronary artery creates the most difficulty in detecting a good pulse-Doppler signal during the peak test. This is due to the need to employ a low frequency probe which guarantees a better color Doppler signal but not such a good pulse Doppler signal due to wall noise interference. Coronary flow reserve evaluation in clinical practice ===================================================== My 5-year personal experience of evaluating CFR added to wall motion abnormalities during a vasodilator stress-echo has led me to reach some final considerations. Such an integrated analysis is feasible and non time-consuming in any stress-echo Lab., after a short training period. By integrating these two pieces of information, we can improve our diagnostic and prognostic accuracy as together they offer a better and more complete pathophysiologic finding. This allows us to choose the best treatment for patients with coronary artery disease and to follow them up over time. I believe that soon we will be in a position to make an accurate and exhaustive study of the posterior coronary artery, too and therefore to obtain a more complete functional cardiac evaluation. It follows that the integrated study of coronary flow and reserve and contractility will, in the near future, become the test of choice in guiding decision-making in the treatment of coronary stenosis before undergoing interventional procedures. List of Abbreviations ===================== CFR-coronary flow reserve LAD-left anterior descending artery RCA-right coronary artery WMSI-wall motion score index ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Imaging and projections for transthoracic imaging of right coronary and posterior descending coronary arteries. ::: ![](1476-7120-3-8-7) ::: Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Visualization by color-Doppler of the mid-distal tract of the right coronary artery and posterior descending with perforans branches. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 2 Imaging of the blood flow inside the left anterior descending coronary artery depicted by color-Doppler in real anatomical direction. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 3 Visualization by color-Doppler of the proximal-mid tract blood flow of Circumflex coronary artery in real anatomical direction. ::: ::: {.caption} ###### Click here for file :::	PubMed Central	2024-06-05T03:55:55.781449	2005-3-25	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084250/", "journal": "Cardiovasc Ultrasound. 2005 Mar 25; 3:8", "authors": [ { "first": "Fausto", "last": "Rigo" } ] }
PMC1084330	Introduction {#s1} ============ Upon infection by poliovirus and many other positive-strand RNA viruses, dramatic changes are rapidly induced in the cellular environment. Poliovirus infection causes the massive rearrangement of intracellular membranes, with double-membraned vesicles 200--400 nm in diameter accumulating in the cytoplasm \[[@pbio-0030156-b01],[@pbio-0030156-b02]\]. Immunoelectron microscopy has revealed that the cytoplasmic surfaces of these membranous vesicles are the sites of viral RNA replication \[[@pbio-0030156-b03],[@pbio-0030156-b04],[@pbio-0030156-b05]\] and, indeed, all known positive-strand RNA viruses of eukaryotic cells replicate their RNA on cytoplasmic membranes. It is thought that one function for membrane localization of viral RNA replication proteins is to promote their oligomerization \[[@pbio-0030156-b06],[@pbio-0030156-b07]\]. It has been noted previously that several of the features displayed by the vesicles induced during poliovirus infection \[[@pbio-0030156-b01],[@pbio-0030156-b02]\] are known to be shared with cellular membranous structures termed autophagosomes. During cellular autophagy, cells break down cytoplasmic proteins and organelles within autophagosomes, double-membraned structures that become degradative upon maturation. Originally identified as a process induced by cellular starvation, autophagy is now appreciated as a cellular response to a variety of stimuli, including hormone treatment, and is a feature of normal development in several organisms (reviewed in \[[@pbio-0030156-b08]\]). In mammalian cells, two chemical inducers of autophagy are tamoxifen \[[@pbio-0030156-b09],[@pbio-0030156-b10]\] and rapamycin \[[@pbio-0030156-b11],[@pbio-0030156-b12],[@pbio-0030156-b13]\]. Cells that express the estrogen receptor respond to tamoxifen treatment by accumulating large amounts of autophagosomes \[[@pbio-0030156-b09],[@pbio-0030156-b10]\]. Rapamycin inhibits the function of the mammalian target of rapamycin, a known repressor of the autophagic pathway (reviewed in \[[@pbio-0030156-b14]\]). In yeast, formation and maturation of autophagosomes requires the functions of many genes (reviewed in \[[@pbio-0030156-b15]\]). Human homologs of several of the yeast autophagy genes have been recently identified, including LC3, the human homolog of Atg8p, and it is likely that much of the autophagic pathway is conserved \[[@pbio-0030156-b16],[@pbio-0030156-b17],[@pbio-0030156-b18]\]. In mammalian cells, the formation and maturation of autophagosomes involves the stepwise acquisition of proteins from disparate cellular compartments (reviewed in \[[@pbio-0030156-b19],[@pbio-0030156-b20],[@pbio-0030156-b21]\]). Nascent autophagosomes form either de novo \[[@pbio-0030156-b22],[@pbio-0030156-b23]\] or from the endoplasmic reticulum (ER) \[[@pbio-0030156-b24],[@pbio-0030156-b25]\] and comprise cellular cytoplasm surrounded by two lipid bilayers that fuse from a C-shaped intermediate. Although relatively protein-poor \[[@pbio-0030156-b22],[@pbio-0030156-b26]\], they have been shown to contain a modified, lipidated form of LC3 \[[@pbio-0030156-b16]\]. As maturation proceeds, the late autophagosome acquires LAMP1 before lysosomal fusion \[[@pbio-0030156-b16],[@pbio-0030156-b25],[@pbio-0030156-b27],[@pbio-0030156-b28]\], which defines the "mature autolysosome." Mature autolysosomes are no longer surrounded by double membranes and, having degraded the inner membrane and much of the cytosolic contents, become electron-dense \[[@pbio-0030156-b27],[@pbio-0030156-b29]\]. Various intermediates among these stages have also been visualized, including incompletely fused double membranes and incompletely degraded inner membranes \[[@pbio-0030156-b23],[@pbio-0030156-b24],[@pbio-0030156-b25]\]. Although poliovirus-induced vesicles display several hallmarks of autophagosomes, other origins for the poliovirus-induced vesicles have been suggested \[[@pbio-0030156-b30],[@pbio-0030156-b31],[@pbio-0030156-b32]\]. Poliovirus RNA replication is known to be inhibited by brefeldin A, which has led to the suggestion that the virally induced vesicles might derive from the COPI pathway, a known target of brefeldin A \[[@pbio-0030156-b32],[@pbio-0030156-b33],[@pbio-0030156-b34]\]. However, this would be inconsistent with the apparent origin of the poliovirus-induced vesicles from the ER: many images consistent with the budding of the poliovirus-induced vesicles directly from the ER have been reported \[[@pbio-0030156-b03],[@pbio-0030156-b31]\]. Furthermore, poliovirus proteins 2BC and 3A induce the formation of vesicles that are biochemically and ultrastructurally similar to those formed in poliovirus-infected cells \[[@pbio-0030156-b35]\]. The 3A protein and 2C, a proteolytic product of 2BC, localize to the ER when expressed in isolation \[[@pbio-0030156-b35],[@pbio-0030156-b36]\]. Therefore, it is likely that poliovirus proteins 3A and 2BC, or larger precursors, localize to the ER early in infection and subsequently promote vesiculation from ER membranes. Poliovirus-induced vesicles have been shown to contain human COPII proteins Sec13p and Sec31p early in their formation, leading to the hypothesis that they are modified COPII vesicles, subverted from the anterograde transport pathway \[[@pbio-0030156-b31]\]. Although this is an attractive suggestion, it is incongruent with other findings concerning the poliovirus-induced vesicles: their frequently observed double-membraned morphology, their cytosolic contents, and the abundance of components from throughout the secretory pathway, including LAMP1, a marker of late endosomes and lysosomes \[[@pbio-0030156-b01],[@pbio-0030156-b02],[@pbio-0030156-b35]\]. Furthermore, in Saccharomyces cerevisiae, components of the COPII pathway, encoded by the SEC12, 16, 23, and 24 genes, are known to be required for autophagy \[[@pbio-0030156-b37],[@pbio-0030156-b38],[@pbio-0030156-b39]\], making it possible that the Sec13p and Sec31p proteins could have other functions in addition to their known roles in anterograde traffic. The size of the poliovirus-induced vesicles, at 200--400 nm in diameter, does not immediately suggest whether they are related to the COPII or autophagosomal pathways: typical diameters of COPII vesicles and autophagosomes are 50--100 nm \[[@pbio-0030156-b40],[@pbio-0030156-b41]\] and 500--1,000 nm, respectively \[[@pbio-0030156-b24],[@pbio-0030156-b25]\]. Finally, the idea that poliovirus infection induces autophagosome formation to facilitate viral growth is surprising in view of existing data that autophagy is an effective antimicrobial host response in many cases (reviewed in \[[@pbio-0030156-b42],[@pbio-0030156-b43]\]). During herpes virus infection, there is good correlation between the induction of autophagy and the presence of effective antiviral responses. The process of autophagy is thought to be destructive to herpes virus because mutant viruses that fail to inhibit autophagy and other antiviral responses showed large decreases in yield \[[@pbio-0030156-b44]\]. During Streptococcus infection, genetic disruption of the autophagy pathway in the infected host cells resulted in increased bacterial yield, consistent with a role for autophagy in bacterial clearance \[[@pbio-0030156-b45]\]. Specific markers for autophagosome formation have become available, facilitating the identification of membranes derived from the autophagic pathway (reviewed in \[[@pbio-0030156-b43],[@pbio-0030156-b46]\]). Here, we explore the role of several constituents of autophagosome machinery in poliovirus- and rhinovirus-infected cells by monitoring: the presence of autophagosomal protein LC3 in virally induced vesicles, the acquisition of colocalization of LC3 and LAMP1 in virally infected cells, the viral induction of punctate structures that stain with monodansylcadaverine (MDC), and the effects of perturbing the autophagosomal pathway pharmacologically and via RNA interference on intracellular and extracellular virus yield. Our data support the hypothesis that poliovirus, and likely the closely related rhinovirus, induce the formation of autophagosome-like structures to serve as the membrane scaffolds for RNA replication. We further suggest that double-membraned vesicles, by forming a luminal cytoplasmic compartment, may facilitate prelytic viral exit from infected cells. Results {#s2} ======= LC3, a Marker of Cellular Autophagosomes, Colocalizes with Proteins of the Poliovirus RNA Replication Complex {#s2a} ------------------------------------------------------------------------------------------------------------- To ask whether the membranes on which poliovirus RNA replication complexes assemble contain constituents of autophagosomes, we monitored the localization of both LC3, a specific marker of autophagosomes, and 3A, a critical component of the poliovirus RNA replication complex, in infected cells. LC3 was expressed via DNA transfection as an amino-terminal fusion with green fluorescent protein (GFP). As can be seen in [Figure 1](#pbio-0030156-g001){ref-type="fig"}, the punctate GFP--LC3 signal colocalized with that of poliovirus RNA replication protein 3A, as visualized by immunofluorescence. No colocalization with poliovirus 3A protein was seen when GFP was not fused to LC3. Therefore, the membranous structures to which 3A and other components of the poliovirus RNA replication complex localize \[[@pbio-0030156-b03]\] can recruit LC3. ::: {#pbio-0030156-g001 .fig} Figure 1 ::: {.caption} ###### Simultaneous Visualization of GFP-LC3 and Poliovirus Protein 3A (Red) MCF-7 cells were transfected either with a plasmid that expresses a GFP-LC3 fusion protein or GFP alone as indicated. Forty-eight hours posttransfection, cells were infected with poliovirus, then fixed and stained using a monoclonal antibody to poliovirus 3A protein and a rhodamine-conjugated secondary antibody. ::: ![](pbio.0030156.g001) ::: LAMP1 and LC3 Colocalize in Both Tamoxifen-Treated, Rapamycin-Treated, and Poliovirus-Infected Cells {#s2b} ---------------------------------------------------------------------------------------------------- In nonautophagic cells, LC3, originally identified as a microtubule-associated protein \[[@pbio-0030156-b47]\], and LAMP1, a marker of late endosomes and lysosomes, do not colocalize. However, during autophagy, LC3 has been shown to colocalize with lysosomal protein LAMP1, defining the maturation of the nascent autophagosome to the autophagosome \[[@pbio-0030156-b28],[@pbio-0030156-b48],[@pbio-0030156-b49]\]. This process can be seen in [Figure 2](#pbio-0030156-g002){ref-type="fig"}A, in which LC3 and LAMP1 were shown to colocalize almost completely in MCF-7 cells upon treatment with tamoxifen, a known inducer of autophagy. Similarly, rapamycin treatment of HeLa cells led to the colocalization of GFP--LC3 and LAMP-1, while they remained distinct in untreated cells ([Figure 2](#pbio-0030156-g002){ref-type="fig"}A). By 3 h postinfection with poliovirus, the LAMP1 and GFP--LC3 signals in HeLa cells also began to colocalize ([Figure 2](#pbio-0030156-g002){ref-type="fig"}B); this time corresponds to the beginning of viral RNA synthesis in infected HeLa cells \[[@pbio-0030156-b50]\]. By 4.5 h postinfection, the LAMP1 and GFP--LC3 signals had merged almost completely ([Figure 2](#pbio-0030156-g002){ref-type="fig"}B). The colocalization of GFP--LC3 and LAMP1 characteristic of autophagosome formation was also seen in cells infected with either of two different serotypes of human rhinovirus ([Figure 2](#pbio-0030156-g002){ref-type="fig"}C). ::: {#pbio-0030156-g002 .fig} Figure 2 ::: {.caption} ###### Simultaneous Visualization of GFP-LC3 and Resident Lysosomal Protein LAMP1 (Red) during Autophagic Induction, Poliovirus Infection, and Rhinovirus 14 Infection \(A) MCF-7 cells were transfected with a plasmid that expresses GFP-LC3 fusion protein and treated with 10 μM tamoxifen in DMSO/EtOH or with DMSO/EtOH alone as indicated. H1-HeLa cells were transfected with the GFP-LC3-expressing plasmid and treated with rapamycin for 5 h or left untreated as indicated. \(B) A time course of poliovirus infection was performed in GFP-LC3-transfected MCF-7 cells, followed by visualization of LAMP1 and GFP-LC3 localization. Infections were with Mahoney type 1 poliovirus at a multiplicity of infection (MOI) of 50 PFU/cell for 5 h at 37 °C. \(C) H1 HeLa cells were transfected with an LC3-GFP-expressing plasmid and either mock-infected or infected with human rhinoviruses as indicated at 50 infectious units/cell for 6 h at 33.5 °C. ::: ![](pbio.0030156.g002) ::: Simultaneous Expression of Poliovirus Proteins 2BC and 3A Induces the Colocalization of GFP--LC3 and LAMP1 {#s2c} ---------------------------------------------------------------------------------------------------------- Expression of poliovirus proteins 2BC and 3A in isolation can induce the formation of double-membraned vesicles that display biochemical and ultrastructural similarity to those formed in poliovirus-infected cells \[[@pbio-0030156-b35]\]. However, expression of poliovirus protein 2BC alone has been shown to induce the formation of single-membraned vesicles in both mammalian cells and S. cerevisiae \[[@pbio-0030156-b04],[@pbio-0030156-b35],[@pbio-0030156-b51],[@pbio-0030156-b52]\]. Expression of poliovirus 3A protein in isolation, on the other hand, reduces the rate of ER-to-Golgi traffic and distends ER membranes \[[@pbio-0030156-b36],[@pbio-0030156-b53]\]. To test whether 2BC, 3A, or both were sufficient to induce colocalization of GFP--LC3 and LAMP1 as observed in poliovirus-infected cells, proteins 2BC and 3A were expressed singly or in combination in 293T cells in the presence of GFP--LC3. As shown in [Figure 3](#pbio-0030156-g003){ref-type="fig"}, colocalization of GFP--LC3 and LAMP1 was not observed in control cells or in cells that expressed either 2BC or 3A in isolation. However, coexpression of 2BC and 3A caused GFP--LC3 and LAMP1 to colocalize ([Figure 3](#pbio-0030156-g003){ref-type="fig"}A), as was seen in cells infected with poliovirus or rhinoviruses treated with tamoxifen or treated with rapamycin (see [Figure 2](#pbio-0030156-g002){ref-type="fig"}). ::: {#pbio-0030156-g003 .fig} Figure 3 ::: {.caption} ###### Simultaneous Visualization of GFP-LC3 and Resident Lysosomal Protein LAMP1 (Red) in Cells that Express Poliovirus 2BC and 3A Proteins 293T cells were transfected with vectors expressing 2BC, 3A, or both and a GFP-LC3 expressing vector for 48 h at 37 °C. Nonexpressing vector DNA was used to ensure that all transfections contained the same amount of DNA. ::: ![](pbio.0030156.g003) ::: Poliovirus Infection and Autophagic Induction Induce Punctate MDC Staining {#s2d} -------------------------------------------------------------------------- To examine further whether the membranes induced during poliovirus infection display additional characteristics of autophagosomes, we employed fluorescent staining with MDC. Under specific fixation conditions, MDC is retained in autophagosomal membranes \[[@pbio-0030156-b54],[@pbio-0030156-b55]\]. As shown in [Figure 4](#pbio-0030156-g004){ref-type="fig"}, treatment of MCF7 human breast tumor cells with tamoxifen caused the accumulation of punctate structures that retained staining with MDC after fixation. Similar punctate staining was observed in cells infected for 5 h with rhinovirus 14 ([Figure 4](#pbio-0030156-g004){ref-type="fig"}A) or for 3 h or more with poliovirus (see [Figure 4](#pbio-0030156-g004){ref-type="fig"}B). ::: {#pbio-0030156-g004 .fig} Figure 4 ::: {.caption} ###### MDC Staining of MCF7 Cells upon Tamoxifen Treatment, Rhinovirus 14 Infection, or Poliovirus Infection \(A) Cells that were treated with 10 μM tamoxifen in DMSO/EtOH or with DMSO/EtOH alone for 48 h at 37 °C, or that were mock-infected or infected with human rhinovirus 14 as in [Figure 3](#pbio-0030156-g003){ref-type="fig"}, were incubated with 100 μM MDC, fixed, and visualized by deconvolution microscopy in the UV channel as described in [Materials and Methods](#s4){ref-type="sec"}. \(B) Cells were infected with poliovirus at an MOI of 50 PFU/cell for 0, 1.5, 3, or 4.5 h at 37 °C; for each time point, MDC incubation was begun 1 h prior to fixation and visualization. ::: ![](pbio.0030156.g004) ::: To test the hypothesis that the MDC-stained structures observed in [Figure 4](#pbio-0030156-g004){ref-type="fig"} were the same as the GFP--LC3-containing structures observed in [Figure 2](#pbio-0030156-g002){ref-type="fig"}, we monitored the localization of GFP--LC3 with MDC. As a live stain, MDC is a poor marker for autophagosomes, because both autophagic membranes and lysosomes are visualized \[[@pbio-0030156-b46]\]. However, if cells are subjected to a defined fixation protocol after staining with MDC, the dye can be removed from single-membraned, but not double-membraned, vesicles \[[@pbio-0030156-b54],[@pbio-0030156-b55]\]. As shown in [Figure 5](#pbio-0030156-g005){ref-type="fig"}, good colocalization of MDC and LC3 was observed in poliovirus-infected cells, and no evidence of the punctate MDC staining expected of lysosomal membranes could be seen in uninfected cells. ::: {#pbio-0030156-g005 .fig} Figure 5 ::: {.caption} ###### Simultaneous Visualization of GFP-LC3 Localization and MDC Staining (Red) in Uninfected and Poliovirus-Infected Cells Forty-eight hours posttransfection of MCF-7 cells with a plasmid that expresses a GFP-LC3 fusion protein, cells were mock-infected or infected with poliovirus at an MOI of 50 PFU/cell for 5 h at 37 °C. After fixation, deconvolution microscopy was used to visualize fluorescence from both the MDC and GFP fluorescent molecules. MDC incubation was begun 1 h prior to fixation and visualization. ::: ![](pbio.0030156.g005) ::: Compounds That Stimulate and Inhibit Cellular Autophagy Affect Yield of Intracellular Poliovirus {#s2e} ------------------------------------------------------------------------------------------------ To determine whether the autophagosome-like membranes induced during poliovirus infection perform an antiviral function or facilitate viral replication, we tested the effect on poliovirus yield of pretreating H1--HeLa cells with either tamoxifen or rapamycin, known inducers of autophagy. Yield of intracellular virus increased approximately 4-fold when cells were pretreated with tamoxifen, and 3-fold upon pretreatment with rapamycin ([Figure 6](#pbio-0030156-g006){ref-type="fig"}A and [6](#pbio-0030156-g006){ref-type="fig"}B). Conversely, when cells were treated with 3-methyladenine, a pharmacological inhibitor of autophagy \[[@pbio-0030156-b56]\], yield of intracellular virus was decreased ([Figure 6](#pbio-0030156-g006){ref-type="fig"}C). Therefore, during poliovirus infection, the activity of the autophagy pathway correlates with productive viral replication, not with viral destruction. ::: {#pbio-0030156-g006 .fig} Figure 6 ::: {.caption} ###### Intracellular Viral Yields from Poliovirus Infections Performed in the Presence of Pharmacological Inducers and Inhibitors of Autophagy \(A) H1 HeLa cells were treated with 10 μM tamoxifen in DMSO/EtOH or in DMSO/EtOH alone for 48 h at 37 °C. Cell numbers were determined, and triplicate plates were infected with poliovirus at an MOI of 0.1 PFU/cell for the times indicated. (B) H1 HeLa cells were treated with 50 nM rapamycin in DMSO/EtOH or DMSO/EtOH alone for 3 h before infection with poliovirus as in (A). \(C) H1 HeLa cells were treated with 10 mM 3-methyladenine in DMSO/EtOH or DMSO/EtOH alone for 3 h before infection with poliovirus as in (A). Viral yields were determined by plaque assay in H1-HeLa cells and expressed as PFU/cell. ::: ![](pbio.0030156.g006) ::: A Potential Role for Autophagosome-Like Membranes in Extracellular Delivery of Viruses {#s2f} -------------------------------------------------------------------------------------- To extend the studies with pharmacological agents, we tested the effect of reducing the amount of intracellular autophagy proteins LC3 and Atg12p on poliovirus yield, using double-stranded RNA oligonucleotides designed to target their mRNAs for destruction by RNA interference. Pools of double-stranded oligonucleotides were synthesized to target ATG12 mRNA, and to target both LC3A and LC3B mRNAs (see [Materials and Methods](#s4){ref-type="sec"}). As shown in [Figure 7](#pbio-0030156-g007){ref-type="fig"}, the intracellular abundance of Atg12p and LC3 proteins could be reduced to 70% and 90%, respectively, of their abundance in cells treated with control double-stranded small interfering RNAs (siRNAs) known to target firefly luciferase mRNA. These reductions in Atg12p and LC3 protein concentration resulted in 3-fold and 4-fold reductions, respectively, in yields of intracellular poliovirus. Even stronger effects of the RNA interference (RNAi)-mediated reduction in autophagy proteins on viral yield were observed, however, when the yields of extracellular virus were examined. The effects of RNAi against ATG12 and LC3 on extracellular virus were 9-fold and 20-fold, respectively ([Figure 7](#pbio-0030156-g007){ref-type="fig"}). Two potential mechanisms for this preferential reduction in extracellular, as opposed to intracellular virus, seemed possible. The first possibility was that the RNAi-mediated reductions in the intracellular concentrations of Atg12p and LC3 protein reduced cell lysis and therefore lytic release of virus; links between autophagy and apoptosis have been reported (reviewed in \[[@pbio-0030156-b57],[@pbio-0030156-b58]\]). The possibility of decreased cellular susceptibility to lysis upon RNAi treatment is difficult to exclude. However, early cell lysis in the presence of reduced Atg12p or LC3 concentrations seemed unlikely because the amounts of virus released were very small, and the observed reductions in extracellular virus release were seen at early time points, before lysis was expected. ::: {#pbio-0030156-g007 .fig} Figure 7 ::: {.caption} ###### Intracellular and Extracellular Viral Yields from Poliovirus Infections of Cells Treated with Small RNA Duplexes to Reduce the Intracellular Concentrations of LC3 and Atg12p Proteins \(A) H1 HeLa cells (1 × 10^6^) were transfected with 12.5 pm each of eight different RNA duplexes targeted to the LC3A and LC3B mRNAs, or with 100 pm of an RNA duplex targeted to firefly luciferase for 24 h at 37 °C (see [Materials and Methods](#s4){ref-type="sec"}). Cell numbers were determined and triplicate plates were infected with poliovirus at an MOI of 0.1 PFU/cell for the times indicated. Viral yields were determined by plaque assay in H1-HeLa cells and expressed as PFU/cell for the intracellular virus, and as total PFU/plate for the extracellular virus. The relative abundance of LC3 protein in the cells incubated with the control and LC3-targeted RNAi molecules was determined by immunoblot using polyclonal antibodies directed against LC3 and GAPDH. \(B) H1 HeLa cells were transfected with 25 pm each of four RNA duplexes targeted to ATG12 mRNA, or with 100 pm of an RNA duplex targeted to firefly luciferase, for 24 h at 37 °C (see [Materials and Methods](#s4){ref-type="sec"}). Poliovirus infections were performed as in (A). The relative abundance of Atg12p protein was determined by immunoblot using antibodies against human Atg12p and GAPDH. ::: ![](pbio.0030156.g007) ::: A second possibility for the greater reduction in extracellular than intracellular virus is that double-membraned structures with the cytoplasm provide a topologically reasonable mechanism for nonlytic release of cytoplasmic contents. Ultrastructural analysis studies of cells infected with poliovirus \[[@pbio-0030156-b01],[@pbio-0030156-b02]\] and with rhinovirus 14 ([Figure 8](#pbio-0030156-g008){ref-type="fig"}A) have revealed the presence of virions and other cytoplasmic material within the lumen of double-membraned vesicles. Presumably, these are double-membraned vesicles that formed relatively late in infection and were therefore able to trap viruses already present in the nearby cytoplasm. The known proximity of RNA replication, protein expression, and virion packaging to the membrane-associated replication complexes \[[@pbio-0030156-b59],[@pbio-0030156-b60]\] should facilitate such events. [Figure 8](#pbio-0030156-g008){ref-type="fig"}B displays an image consistent with the release of the packets of cytosol expected if the outer bilayer of a multilamellar membrane structure were to fuse with the plasma membrane. [Figure 8](#pbio-0030156-g008){ref-type="fig"}C shows that, in poliovirus-infected cells, such blebs can contain LC3, and [Figure 8](#pbio-0030156-g008){ref-type="fig"}D shows that they can also contain VP1, a viral capsid protein. ::: {#pbio-0030156-g008 .fig} Figure 8 ::: {.caption} ###### Ultrastructure of H1-HeLa Cells Infected with Human Rhinovirus 14 (HRV14) and Immunoelectron Microscopy of Cells Infected with Poliovirus \(A) Cells were infected with rhinovirus 14 as in [Figure 3](#pbio-0030156-g003){ref-type="fig"} and prepared for electron microscopy by high-pressure freezing. Examples of readily discernable double lipid bilayers are designated with large arrowheads; vesicles that contain intralumenal viral particles are denoted with small arrows. (B and C) Cells were transfected with a plasmid that expresses GFP-LC3 and subsequently infected with poliovirus for 5 h as in [Figure 2](#pbio-0030156-g002){ref-type="fig"}. GFP-LC3 was visualized using a secondary antibody coupled to 10-nm gold particles; examples of such particles are denoted with arrows. An arrowhead identifies apparently extracellular packets of cytosol. \(D) Cells transfected with an GFP-LC3-expressing plasmid and infected with poliovirus as in (C) were immunostained using an antibody directed against VP1, a viral capsid protein and visualized using a secondary antibody conjugated to 10-nm gold particles; examples of such particles are identified with arrows. ::: ![](pbio.0030156.g008) ::: Discussion {#s3} ========== In this work, we have shown that several hallmarks of cellular autophagosomes, including the localization of GFP--LC3 into discrete punctate structures and staining with MDC followed by fixation, can be observed in human cells whether they are treated with tamoxifen or infected with poliovirus or rhinovirus. Furthermore, we have employed a new criterion for autophagy, the colocalization of LC3 and LAMP1, to demonstrate that this intracellular rearrangement was observed in cells treated with tamoxifen, infected with picornaviruses (poliovirus, rhinovirus 2, or rhinovirus 14) or transfected with plasmids that express poliovirus proteins 2BC and 3A (see [Figures 2](#pbio-0030156-g002){ref-type="fig"} and [8](#pbio-0030156-g008){ref-type="fig"}). Taken together with the previously observed double-membraned morphology, cytoplasmic contents, and complex origin of the membranous structures induced during poliovirus infection \[[@pbio-0030156-b01],[@pbio-0030156-b02],[@pbio-0030156-b35]\], we argue that the viruses utilize components of the autophagosome formation pathway to form the characteristic double-membraned vesicles seen during infection. Do the autophagosome-like membranous structures induced by poliovirus act as scaffolds for RNA replication, or are they part of the host antiviral response? The numerous positive correlations between the functional presence of autophagosomal pathways and increased viral yield lead us to conclude that the autophagosome-like structures observed during poliovirus infection are not antiviral. Instead, we argue that the double-membraned vesicles induced during poliovirus infection facilitate poliovirus replication, and we hypothesize that poliovirus, rhinovirus 2, and rhinovirus 14 subvert the constituents of the cellular autophagy pathway to form membranous scaffolds on which RNA replication complexes can assemble. Other positive-strand RNA viruses that have been shown to localize their RNA replication complexes to double-membraned vesicles in the cytoplasm of infected cells are equine artirivirus \[[@pbio-0030156-b61]\], murine hepatitis virus \[[@pbio-0030156-b62]\], and SARS virus \[[@pbio-0030156-b63]\]. The membranes associated with murine hepatitis virus infection have been shown recently to contain LC3 protein \[[@pbio-0030156-b64]\]. Strikingly, infection of murine ES cell lines deficient for APG5 with murine hepatitis virus was shown to result in a large decrease in the extracellular yield of this enveloped virus; the effect on intracellular viral particles or RNA was not reported. Nevertheless, the large reduction in yield of extracellular virus seen in the absence of Atg5p protein argues that this component of the cellular autophagy pathway is crucial for some step in the formation or egress of infectious virions of this murine coronavirus \[[@pbio-0030156-b64]\]. For both poliovirus and equine artirivirus, molecular inducers of double-membraned vesicle formation have been identified. Specifically, coexpression of poliovirus proteins 2BC and 3A is required to accumulate double-membraned vesicles \[[@pbio-0030156-b35]\] and to elicit the colocalization of GFP--LC3 and LAMP1 that correlates with the formation of autophagosomes (see [Figure 3](#pbio-0030156-g003){ref-type="fig"}A). For equine artirivirus, the coexpression of viral proteins nsp2 and nsp3 is sufficient to induce the formation of double-membraned vesicles \[[@pbio-0030156-b65]\]. We anticipate that these viral proteins, likely to be capable of mimicking, intercepting, or corrupting the pathway of cellular autophagy, will prove to be useful tools to decipher its mechanism. Recent work has highlighted an important role for autophagy in the innate immune response of vertebrates to intracellular pathogens. For example, induction of autophagy has been shown to promote clearance of Mycobacterium tuberculosis from infected macrophage \[[@pbio-0030156-b66]\]. Furthermore, during infection with Shigella flexneri, the wild-type function of the bacterial icsB gene was shown to be required to prevent autophagic degradation, a process that the authors argue is specifically induced by a bacterial protein, the product of the virG gene \[[@pbio-0030156-b67]\]. Like S. flexneri, successful microorganisms often display strategies to evade potent host defenses. Furthermore, some microorganisms actively subvert otherwise effective host defense responses for their own benefit: for example, the growth of mink focus-forming virus requires apoptotic caspase activity for the maturation of a nonstructural protein \[[@pbio-0030156-b68]\], and murine cytomegalovirus encodes a chemokine homolog to attract cells of the immune system to the site of infection, which then promote viral dissemination (reviewed in \[[@pbio-0030156-b69]\]). Similarly, precedents are beginning to be established in which the autophagic pathway or its constituents may be subverted by intracellular pathogens to benefit their own replication (reviewed in \[[@pbio-0030156-b42],[@pbio-0030156-b43]\]). As shown in [Figure 9](#pbio-0030156-g009){ref-type="fig"}, the pathway of autophagosome formation ends with the acquisition of the lysosomal proteases and lipases that render the autophagosome a degradative organelle \[[@pbio-0030156-b27],[@pbio-0030156-b28]\]. Mature autolysosomes are no longer bounded by double membranes because the inner membrane and luminal contents are degraded, becoming electron-dense and compact \[[@pbio-0030156-b24]\]. In Legionella infection, several genes, termed Dot or Icm genes, are required to retard the progression of autophagosome maturation, presumably to benefit bacterial growth within organelles that subvert components of the autophagosome \[[@pbio-0030156-b70],[@pbio-0030156-b71],[@pbio-0030156-b72]\]. We argue that poliovirus, rhinovirus 2, rhinovirus 14, equine encephalitis virus \[[@pbio-0030156-b65]\], and murine hepatitis virus \[[@pbio-0030156-b64]\] have similarly evolved a mechanism to accumulate autophagosome-like membranes in the cytoplasm for the duration of the infection. To maintain most of these structures in their double-membraned form, viral infection may both induce their formation and prevent their maturation into degradative organelles. ::: {#pbio-0030156-g009 .fig} Figure 9 ::: {.caption} ###### Pathway of Autophagosome Formation, Autophagic Degradation, and Proposed Steps of Pathway Subversion by Poliovirus and Related Viruses Double-membraned autophagosomes form either from ER membrane or de novo, encapsulating cytosol; the action of many gene products, including Atg5p and Atg12p, are required. LC3 protein (the Atg8p homolog) is associated with "sequestration crescents" as well as fully formed double-membraned autophagosomes. LAMP1 acquisition is a hallmark of the maturation of these structures, which eventually fuse with lysosomes to produce mature autophagosomes with single membranes and electron-dense contents. We hypothesize that infection by poliovirus or rhinovirus induces accumulation of autophagosomes to promote viral RNA replication by accelerating the formation of autophagosome-like structures from ER membranes, blocking the maturation of these structures into degradative organelles, or both (upper dotted line). The double-membraned topology makes the extracellular release of virions trapped in the cytosolic lumen topologically plausible, providing a mechanism for viral release in the absence of cell lysis. This could occur either from a double-membraned structure or from one in which only one of the membranes remained (dotted arrows). ::: ![](pbio.0030156.g009) ::: Why would a virus choose a double-membraned, autophagosome-like vesicle on which to replicate its RNA? Not all positive-strand RNA viruses utilize such structures. For example, Flock House virus RNA replication complexes assemble on outer mitochondrial membranes \[[@pbio-0030156-b73]\]. When the RNA replication complexes of a subgenomic Flock House virus RNA were experimentally redirected to the cytoplasmic surface of the ER of S. cerevisiae, the yield of replicated RNA actually increased \[[@pbio-0030156-b73]\]. Therefore, viral RNA replication displayed no mitochondrion-specific requirement for specific lipids, proteins, or processes. Presumably, however, there are reasons why particular positive-strand RNA viruses target particular intracellular membranes on which to assemble their replication complexes, some of which might not be assayed under single-cycle growth conditions in tissue culture. A larger effect on extracellular than intracellular virus yield was observed when the abundance of autophagy proteins Atg12p and LC3 was reduced by RNAi (see [Figure 7](#pbio-0030156-g007){ref-type="fig"}). Possible explanations for the observed preferential decrease in extracellular virions are that the reduction in autophagosome machinery decreased cell lysis early in infection, or that reduced abundance of autophagosomal machinery decreased nonlytic viral escape. Although our data do not yet distinguish between these hypotheses, we will discuss the latter possibility because nonlytic delivery of cytosol to the extracellular milieu could be a unique characteristic of multilamellar vesicles, and we were able to obtain ultrastructural images consistent with this interpretation (see [Figure 8](#pbio-0030156-g008){ref-type="fig"}). Nominally lytic viruses are often assumed to spread exclusively via cell lysis. However, the possibility of nonlytic viral release, even for nonenveloped viruses such as poliovirus, has been suggested from numerous reports of persistently infected cell lines that continuously secreted infectious particles \[[@pbio-0030156-b74],[@pbio-0030156-b75],[@pbio-0030156-b76],[@pbio-0030156-b77]\]. Even more convincingly, when polarized Caco-2 cultures, growing as intact monolayers, as shown by resistance to the passage of dyes and electric current between the apical and basolateral surfaces, were infected with poliovirus, newly synthesized virus was shown to emerge from only the apical surface \[[@pbio-0030156-b78]\]. That the egress of this virus did not correspond to a detectable breach in the monolayer argued that a nonlytic, and polarized exit route of unknown origin had been utilized. We suggest a potential mechanism for this nonlytic release of cytosolic viral particles via the formation of double-membraned vesicles throughout the course of infection. Early in infection, the double-membraned structures would entrap cytosol, but this cytosol would be free of virions. However, at later stages of infection, the cytosol trapped by newly generated double-membraned structures would often contain viral particles. Poliovirions and related enteroviruses are relatively resistant to the low pH and active proteolysis that would prevail within the lumen of these vesicles should they mature \[[@pbio-0030156-b79],[@pbio-0030156-b80]\]. As depicted in [Figure 9](#pbio-0030156-g009){ref-type="fig"}, the fusion of the outer membrane of an intact double-membraned structure would result in the release of membrane-bound packets of cytoplasm into the extracellular milieu, whereas fusion of the membrane of a mature autophagosome would result in the direct release of cytosolic contents, and incompletely resolved double-membrane structures would result in the formation of more complex topologies. The presence of both LC3 and poliovirus capsid protein VP1 in extracellular structures adjacent to poliovirus-infected cells (see [Figure 8](#pbio-0030156-g008){ref-type="fig"}C and [8](#pbio-0030156-g008){ref-type="fig"}D) is consistent with the release of at least partially intact packets of cytoplasm during poliovirus infection. We speculate that the formerly intracellular membranes surrounding these packets of cytoplasm would be short-lived outside the cell, with free virus being the eventual result. Recently, HIV has been shown to exit human macrophages via the fusion of multivesicular bodies with the plasma membrane, rather than by directly budding from the cell surface as in HIV-infected T cells \[[@pbio-0030156-b81],[@pbio-0030156-b82],[@pbio-0030156-b83]\]. HIV gag protein directs the targeting of HIV particles to multivesicular bodies via direct binding to cellular protein Tsg101 \[[@pbio-0030156-b84],[@pbio-0030156-b85]\], a component of the ESCRT complex required for the sorting of proteins into endosomes \[[@pbio-0030156-b86],[@pbio-0030156-b87],[@pbio-0030156-b88]\]. Similarly, the intracellular formation of double-membraned vesicles in poliovirus-infected cells provides a topologically reasonable mechanism for the extracellular delivery of cytosolic contents in the absence of cell lysis. We speculate that a mechanism to spread virus within tissues of infected hosts without cell lysis could provide an advantage to those positive-strand viruses that subvert constituents of the cellular pathway of autophagosome formation. Materials and Methods {#s4} ===================== {#s4a} ### Viruses, plasmids, and cells. {#s4a1} Poliovirus Mahoney type 1 was isolated following transfection with an infectious cDNA \[[@pbio-0030156-b89]\] and propagated as previously described \[[@pbio-0030156-b34]\]. "Tet-off" MCF7 cells were obtained from Clontech (Palo Alto, California, United States) and propagated in DMEM + 10% FBS; MCF7 cells are highly inducible for autophagosome formation in response to tamoxifen \[[@pbio-0030156-b09]\]. Poliovirus stocks were titered on both HeLa H1 and MCF7 cells. The multiplicities of infections indicated refer to titers on the appropriate cell lines. H1--HeLa cells were used due to their permissiveness for rhinovirus infection. Rhinovirus stocks were obtained from the American Type Culture Collection ( ATCC, Manassas, Virginia, United States). Virus stocks were prepared in H1--HeLa cells ( ATCC) grown in EMEM supplemented with 0.2 M HEPES and 0.1 M MgCl~2~, and viral titers were measured by TCID~50~ as described previously \[[@pbio-0030156-b90]\]. To construct the GFP--LC3 fusion protein-expressing plasmid, LC3B sequences were amplified by PCR from a Human Lung Library (ResGen, Carlsbad, California, United States) using targeted primers containing EcoRI sites ( ACTGAATTCCCATGCCGTCGGAGAAG and TTTGAATTCTTACACTGACAATTTCA). The LC3A coding region was then inserted into the EcoRI site of pEGFP-C3 (Clontech, Palo Alto, California, United States) to create an EGFP--LC3 fusion protein under the control of the CMV immediate-early promoter. Expression of poliovirus proteins 2BC and 3A was performed as described previously \[[@pbio-0030156-b35]\]. The 293T cells, which express SV40 T antigen, were used to facilitate expression from the SV40 promoter in these plasmids. ### Immunofluorescence {#s4a2} Cells were fixed using freshly made 4% formaldehyde in PBS for 10 min at room temperature. Poliovirus protein 3A and cellular protein LAMP1 were visualized by indirect immunofluorescence. For visualization of poliovirus 3A protein in the presence of GFP--LC3, cells were washed twice in PBS and incubated in a PBS solution that also contained 0.5% saponin, 10 mM sodium azide, 0.125% BSA, 3A monoclonal tissue culture supernatant at a dilution of 1:30, and rhodamine-linked antimouse secondary antibody (SC-2084, Santa Cruz Biotechnologies, Santa Cruz, California, United States) at a dilution of 1:200. Cells were incubated at 4 °C for 45 min, washed twice with PBS, and placed under Vectashield mounting medium (Vector Laboratories, Burlingame, California, United States). Visualization of LAMP1 was performed using a monoclonal LAMP1 antibody (Transduction Laboratories, Lexington, Kentucky, United States) at a dilution of 1:200. ### MDC staining and costaining. {#s4a3} MDC (Sigma, St. Louis, United States) was stored at −20 °C under desiccant. A fresh stock solution of 5 mM MDC was made in 1:1 DMSO/EtOH immediately prior to adding to cultures. At 1 h before fixation, fresh medium that contained either 10 μM MDC in DMSO/EtOH or an equivalent volume of DMSO/EtOH was added to the cells. The cells were then fixed using a freshly made 4% formaldehyde solution in PBS for 10 min at room temperature and imaging was performed immediately. To ensure that membrane vesiculation induced by methods other than the induction of autophagy did not show similar MDC staining patterns, Golgi vesiculation was induced with 5 μM ilimaquinone \[[@pbio-0030156-b91]\] or 5 μM nocodazole \[[@pbio-0030156-b92]\]; the disappearance of intact Golgi was confirmed using BODIPY-C~5~ ceramide (Molecular Probes, Eugene, Oregon, United States). Under these conditions, punctate MDC staining was not observed (data not shown). ### Microscopy and deconvolution analysis. {#s4a4} Microscopic analysis was carried out on an Olympus IX70 at 100X magnification. Images were captured and deconvolved using SoftWorx 2.50 on an SGI Octane workstation. MDC staining was detected at 360 nm excitation/457 nm emission. GFP--LC3 expression was detected at 490 nm excitation/528nm emission. Rhodamine was detected at 555 nm excitation/617 nm emission. Individual images from each stack were saved as TIFF files and processed in Adobe Photoshop 7.0. ### RNA interference to reduce intracellular concentrations of LC3 and ATG12 {#s4a5} siRNA SMARTpools, consisting of four RNA duplexes targeting the gene of interest, and a control siRNA targeting firefly luciferase, were purchased from Dharmacon (Lafayette, Colorado, United States). For LC3, both LC3A and LC3B RNAs \[[@pbio-0030156-b93]\] were targeted, bringing the total number of transfected duplexes to eight. Pools consisted of an equal amount of each duplex. The siRNA sequences are given as sense/antisense pairs. The siRNA sequences for ATG12 were: GGGAAGGACUU ACGGAUGUUU/5′P-ACAUCCGUAAGUCCUUCCCUU; GAACACCAAGUUUCACUGUUU/5′P- ACAGU GAAACUUGGUGUUCUU; GCAGU AGAGCGAACACGAAUU/5′P-UUCGUGUUCGCUCUACUGCUU; and UGUU GCAGCUUCCUACUUCUU/5′P- GAAGU AGGAAGCU GCAACAUU. The siRNA sequences for LC3A were: GGACGGCUUCCUCUAUAUGUU/5′P-CAUAU AGAGGAAGCCGUCCUU; CGGUGAUCAU CGAGCGCUAUU/5′P-U AGCGCUCGAUGAU CACCGUU; ACAU GAGCGAGUUGGUCAAUU/5′P-UU GACCAACUCGCUCAUGUUU; and CGCCCAU CGCGGACAUCUAUU/5′P-UAGAUGU CCGCGAU GGGCGUU. The siRNA sequences for LC3B were: CAAAGUUCCUUGUACCUGAUU/5′P-U CAGGU ACAAGGAACUUUGUU; GAUAAU AGAACGAU ACAAGUU/5′P-CUUGUAUCGUUCUAUUAUCUU; GU AGAAGAUGU CCGACUUAUU/5′P-UAAGU CGGACAUCUUCUACUU; and AGGAGACGUU CGGGAUGAAUU/ 5′P-UUCAU CCCGAACGUCUCCUUU. Cells were grown to densities of 1---5 × 10^5^ per 6-cm dish in 2.5 ml EMEM without antibiotics, and transfected using Lipofectamine 2000 (Invitrogen, Carlsbad, California, United States) according to manufacturer\'s instructions. For each 6-cm dish, 100 total pmol of pooled siRNA was diluted in 250 μl of serum-free OptiMEM medium (Invitrogen) and, separately, 5 μl of Lipofectamine 2000 was diluted in 250 μl of OptiMEM. After an incubation of 5 min at room temperature, the diluted RNA and Lipofectamine 2000 were combined and incubated for 20 min at room temperature. The 500-μl mixture was then added to each dish and gently rocked to spread the lipid--RNA complexes. Growth curves and immunoblots were performed 48 h after transfection. Total cell extracts were made using RSB-NP40 extraction buffer (10 mM Tris \[pH 7.5\], 10 mM NaCl, 1.5 mM MgCl~2~, 1% NP-40) supplemented with protease inhibitors (Roche, Mannheim, Germany). Extract was separated on a 15% Laemmli gel and transferred to a PVDF membrane for Western blotting. Anti-LC3 immunoblotting was performed using rabbit antibody raised commercially (AnaSpec, San Jose, California, United States) against a peptide comprising the first 16 amino acids of murine LC3 (MPSEKTFKQRRSFEQR). Anti-Atg12p and anti-GAPDH immunoblotting was performed using antibodies from Zymed (South San Francisco, California, United States) and Research Diagnostics (Flanders, New Jersey, United States), respectively. Antibodies were diluted 1:3,000 in a PBS solution that also contained 0.1% Tween-20 and 2% BSA, and detected with alkaline-phosphatase conjugated goat antirabbit antibody, at a dilution of 1:10,000 using the ECF reagent from Amersham Biosciences (Piscataway, New Jersey, United States). ### High-pressure freezing, freeze-substitution, and electron microscopy (EM). {#s4a6} For cryofixation and EM analysis, H1--HeLa cells were grown in EMEM supplemented with 0.2 M HEPES and 0.1 M MgCl~2~ in flasks. Cells were infected with virus at an MOI of 50 PFU/cell for poliovirus or 50 TCID~50~/cell for rhinovirus, then washed three times with PBS, trypsinized, and collected by centrifugation. The cell pellet was resuspended in 0.15 M mannitol in PBS and collected by centrifugation. Aliquots of the resulting pellet were frozen in a Balzers HPM 10 high-pressure freezing apparatus as described previously \[[@pbio-0030156-b02]\] and stored in liquid nitrogen. To observe cellular ultrastructure, samples were freeze-substituted in 0.1% tannic acid in acetone at −80 °C, rinsed in acetone, then warmed at −20 °C in the presence of 2% osmium tetroxide in acetone for 16 h, followed by incubation at 4 °C for 4 h. After rinsing in acetone at 4 °C, samples were embedded in Epon-Araldite resin. Thin sections were stained with 2% uranyl acetate and lead citrate and imaged at 80 kV in a JEOL 100C or Philips CM10 EM. For immunostaining, high-pressure frozen samples were freeze-substituted in 0.1% glutaraldehyde-0.05% uranyl acetate in acetone, embedded, stained, and visualized as described previously \[[@pbio-0030156-b35]\]. ### Accession Numbers {#s4a7} The SwissProt (<http://us.expasy.org/sprot/>) accession numbers for the gene products discussed in this article are 2BC and 3A (P03299), Atg12p (O94817), Atg8p (P38182), LAMP1 (P11279), LC3 (Q9GZQ8), mTor (P42345), Sec12p (P11615), Sec13p (P5573), Sec16p (P48415), Sec23p (P15303), Sec24p (P40482), and Sec31p (O94979). We are indebted to Patrice Codogno and Skip Virgin for experimental suggestions and advice and to Peter Sarnow and Katherine Shim for helpful comments on the manuscript. We are grateful to Susan Palmieri and Jon Mulholland with the Stanford Cell Science Imaging Facility for assistance with deconvolution microscopy. This work was supported by the Hutchison Consortium for Translation Research and the National Institutes of Health. WTJ was supported by a National Research Service Award from the National Heart, Lung, and Blood Institute and a Stanford Dean\'s Fellowship. Competing interests. The authors have declared that no competing interests exist. Author contributions. WTJ, THG, RRK, and KK conceived and designed the experiments. WTJ, THG, and MPT performed the experiments. WTJ and KK analyzed the data. WTJ, MPT, SM, MR, and RRK contributed reagents/materials/analysis tools. KK and WTJ wrote the paper. Citation: Jackson WT, Giddings TH Jr, Taylor MP, Mulinyawe S, Rabinovitch M, et al. (2005) Subversion of cellular autophagosomal machinery by RNA viruses. PLoS Biol 3(5): e156. ER : endoplasmic reticulum GFP : green fluorescent protein MDC : monodansylcadaverine siRNA : small interfering RNA	PubMed Central	2024-06-05T03:55:55.784110	2005-4-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084330/", "journal": "PLoS Biol. 2005 May 26; 3(5):e156", "authors": [ { "first": "William T", "last": "Jackson" }, { "first": "Thomas H", "last": "Giddings" }, { "first": "Matthew P", "last": "Taylor" }, { "first": "Sara", "last": "Mulinyawe" }, { "first": "Marlene", "last": "Rabinovitch" }, { "first": "Ron R", "last": "Kopito" }, { "first": "Karla", "last": "Kirkegaard" } ] }
PMC1084331	Introduction {#s1} ============ Neuronal differentiation is a protracted process during which newly generated neurons express distinct cellular and molecular programs at precise times during their maturation: long-distance axon outgrowth, subsequent terminal branching, and finally synaptogenesis. Many important aspects of neuronal character appear to be acquired through the expression of transcription factors at progenitor cell stages, whereas others depend on expression immediately upon cell cycle exit \[[@pbio-0030159-b01]\]. But whether the orderly expression and activity of transcriptional programs at much later developmental stages, well after cell cycle exit, is an essential step in the progression of neuronal differentiation and circuit assembly has yet to be resolved. The differentiation of sensory neurons of dorsal root ganglia (DRG) has been studied extensively with respect to inductive events that specify neuronal fate \[[@pbio-0030159-b02],[@pbio-0030159-b03]\], as well as the involvement of late target-derived neurotrophic factors in the control of neuronal survival \[[@pbio-0030159-b04]\]. Recent evidence has begun to emerge that target-derived factors are also involved in regulating later aspects of neuronal differentiation \[[@pbio-0030159-b05],[@pbio-0030159-b06],[@pbio-0030159-b07]\]. In particular, genetic experiments have addressed the survival-independent role of neurotrophic factors during development by exploiting strains of mice defective both in neurotrophin signaling and in the function of the proapoptotic gene Bax \[[@pbio-0030159-b08],[@pbio-0030159-b09]\]. These studies, for example, have revealed that neurotrophin signaling controls the acquisition of peptidergic traits in nociceptive DRG neurons and the control of target innervation \[[@pbio-0030159-b08],[@pbio-0030159-b09]\]. The onset of some transcriptional programs in neurons, however, has also been shown to occur long after neurons exit the cell cycle. An emerging principle from work in Drosophila and vertebrates is that target-derived factors play a crucial role in the induction of these transcriptional programs \[[@pbio-0030159-b10]\]. In Drosophila, retrograde BMP signals from the target region control the terminal differentiation of a subpopulation of peptidergic neurons expressing Apterous and Squeeze \[[@pbio-0030159-b11],[@pbio-0030159-b12]\]. In vertebrates, peripheral neurotrophic signals have been shown to direct the onset of expression of the ETS transcription factors Er81 and Pea3 in DRG sensory neurons and motor neuron pools several days after these neurons have become post-mitotic \[[@pbio-0030159-b09],[@pbio-0030159-b13],[@pbio-0030159-b14],[@pbio-0030159-b15],[@pbio-0030159-b16]\]. Moreover, the induction of Er81 expression in proprioceptive afferents is known to be mediated by peripheral neurotrophin 3 (NT-3) \[[@pbio-0030159-b09]\]. These two ETS proteins control late aspects of spinal monosynaptic circuit assembly, with Er81 directing proprioceptive sensory neuron differentiation and Pea3 directing motor neuron pool differentiation, respectively \[[@pbio-0030159-b14],[@pbio-0030159-b15]\]. In particular, in the absence of Er81, achieved by mutation in the gene or by deprivation of peripheral neurotrophin signaling, group Ia proprioceptive afferents fail to invade the ventral spinal cord and to make effective synaptic connections with motor neurons \[[@pbio-0030159-b09],[@pbio-0030159-b14]\]. The involvement of target-derived signals in induction of ETS transcription factor expression raises the question of the necessity for the observed delay in the onset of ETS signaling for neuronal maturation. Would precocious expression of ETS proteins in post-mitotic neurons also direct the appropriate sensory neuron developmental programs? In this study, we have used mouse genetics to test this general idea, by investigating whether the precise timing of onset of ETS transcription factor signaling is essential for normal sensory neuron development. We have assessed the biological effects of inducing ETS signaling either at the correct developmental time, or precociously. We find that within proprioceptive sensory neurons, the late onset of ETS signaling is essential for the establishment of normal sensory afferent projections in the spinal cord. Precocious initiation of ETS signaling in post-mitotic DRG neurons leads to abnormal DRG neuron differentiation characterized by neurotrophin-independent neurite outgrowth and inappropriate profiles of gene expression. Our findings reveal that target-triggered inductive signals provide an effective means of ensuring the late onset of expression of transcription factors, and thus an orderly temporal transcriptional sequence that is crucial for neuronal maturation and circuit assembly. Results {#s2} ======= To test the hypothesis that a temporal delay in the onset of transcriptional programs is crucial for the control of appropriate neuronal maturation, we studied the development of proprioceptive DRG neurons, since transcriptional effectors regulated by target-derived signals, as well as some of their downstream biological actions, have been identified for these neurons. Er81 controls proprioceptive afferent connectivity \[[@pbio-0030159-b14]\], and we therefore sought to identify an ETS transcriptional regulator that, when expressed over the normal time course of Er81 expression, is able to substitute for Er81 function within group Ia afferent sensory neurons. With this reference point, we then designed experiments to examine the effects of precocious post-mitotic expression of the same ETS transcription factor on sensory neuron differentiation. EWS-Pea3 Can Replace Er81 Function in Controlling Ia Afferent Projections {#s2a} ----------------------------------------------------------------------------- We first defined an ETS transcription regulator that is able to replace the function of Er81 within proprioceptive afferents to direct projections into the ventral spinal cord. Er81, Pea3, and Erm constitute the Pea3 subfamily of ETS transcription factors, show a high degree of amino acid identity, and bind to very similar DNA target sequences \[[@pbio-0030159-b17],[@pbio-0030159-b18],[@pbio-0030159-b19]\]. Nevertheless, when introduced into the Er81 locus in analogy to a previously used targeting strategy (data not shown; \[[@pbio-0030159-b14]\]), neither Pea3 nor Erm could rescue Ia proprioceptive afferent projections to extensively invade the ventral horn of the spinal cord (data not shown). These findings prompted us to analyze mice in which we integrated EWS-Pea3, a break-point fusion product between the amino-terminal domain of the Ewing sarcoma (EWS) gene and the Pea3 DNA binding domain \[[@pbio-0030159-b20],[@pbio-0030159-b21]\], into the Er81 locus ([Figure 1](#pbio-0030159-g001){ref-type="fig"}). We found that in a luciferase-enzyme-based cell culture transfection assay, EWS-Pea3 showed stronger transactivation activity than Er81 or Pea3 ([Figure 1](#pbio-0030159-g001){ref-type="fig"}J; data not shown), in agreement with previous studies \[[@pbio-0030159-b22],[@pbio-0030159-b23],[@pbio-0030159-b24]\]. Moreover, transactivation by EWS-Pea3 was abolished by mutation of ETS-binding sites in the reporter plasmid, demonstrating ETS-binding-site dependence (data not shown). ::: {#pbio-0030159-g001 .fig} Figure 1 ::: {.caption} ###### Replacement of Er81 by EWS-Pea3 \(A) Generation of Er81^EWS-Pea3^ mutant mice. Above is the organization of the Er81 genomic locus in the region targeted by homologous recombination in analogy to \[[@pbio-0030159-b14]\]. Exons 1--4 are shown as light blue boxes, and the Er81 start codon in exon 2 is indicated as ATG. The probe used to detect homologous recombination is shown as a grey box. Below is replacement of Er81 by EWS-Pea3 through the integration of EWS-Pea3 in frame with the endogenous start codon of the Er81 locus in exon 2 (in analogy to \[[@pbio-0030159-b14]\]). \(B) PCR and Southern blot analysis of Er81^EWS-Pea3^ wild-type (+/+), heterozygous (+/−), and homozygous (−/−) genomic DNA to detect the mutant allele. PCR primer pairs (EWS-Pea3ki) were used to detect specifically the recombined allele, and a primer pair in exon2 was used to detect the presence of the wild-type allele \[[@pbio-0030159-b14]\]. (C--E) Analysis of Er81 expression in lumbar DRG neurons of E16.5 wild-type (C), Er81^−/−^ (D), and Er81^EWS-Pea3/−^ (E) embryos. Inset in lower right corner of each panel shows Isl1 expression in the respective DRG. (F--H) PV expression in lumbar DRG of E16.5 wild-type (F), Er81^−/−^ (G), and Er81^EWS-Pea3/−^ (H) embryos. Confocal scans were performed with equal gain intensity. \(J) Transcriptional transactivation of luciferase expression from a minimal reporter construct containing five consensus ETS DNA-binding sites ( GCCGGAAGC; \[[@pbio-0030159-b18],[@pbio-0030159-b19]\]) and a minimal TK promoter upon transient transfection of Er81 (n ≥ 7; 3.03 ± 0.66) or EWS-Pea3 (n ≥ 7; 20.3 ± 2.7). Relative luciferase activity normalized to control (Con). Scale bar: 80 μm. ::: ![](pbio.0030159.g001) ::: Expression of Er81 in DRG neurons of embryos containing integration of EWS-Pea3 in the Er81 locus (Er81^EWS-Pea3/−^) was abolished ([Figure 1](#pbio-0030159-g001){ref-type="fig"}E), and the expression level of the calcium-binding protein Parvalbumin (PV) in proprioceptive afferents, which is decreased approximately 5- to 10-fold in Er81 mutants \[[@pbio-0030159-b14]\], was comparable to wild-type levels in Er81^EWS-Pea3/−^ embryos ([Figure 1](#pbio-0030159-g001){ref-type="fig"}F--[1](#pbio-0030159-g001){ref-type="fig"}H). To further define DRG neuron differentiation in the presence of EWS-Pea3 in proprioceptive afferents in vivo, we assessed whether replacement of Er81 by EWS-Pea3 had an influence on neuronal survival or on the expression of proprioceptive-afferent-specific genes. Er81^EWS-Pea3/−^ mice did not differ from wild-type in the number of proprioceptive afferent cell bodies within the DRG of L1 to L5, the expression of several genes normally expressed by proprioceptive afferents, and the lack of expression of genes not normally expressed in proprioceptive afferents ([Figure S1](#sg001){ref-type="supplementary-material"}; data not shown). Together, these findings suggest that the expression of EWS-Pea3 from the normal time of onset mimics the function of Er81 as assessed by induction and maintenance of gene expression within proprioceptive afferents. To determine the extent of rescue of Ia proprioceptive afferent projections into the ventral spinal cord of Er81 mutant mice achieved by expression of EWS-Pea3, we traced intraspinal afferent projections by axonal labeling of PV ([Figure 2](#pbio-0030159-g002){ref-type="fig"}A--[2](#pbio-0030159-g002){ref-type="fig"}C). In addition, to analyze axon ingrowth independent of the level of PV expression in DRG neurons, we used anterograde labeling of afferent fibers by applying fluorescently labeled dextran to cut dorsal roots ([Figure 2](#pbio-0030159-g002){ref-type="fig"}D--[2](#pbio-0030159-g002){ref-type="fig"}F). Using both assays, we found extensive rescue of the projections into the ventral horn of the spinal cord in Er81^EWS-Pea3/−^ mice ([Figure 2](#pbio-0030159-g002){ref-type="fig"}C and [2](#pbio-0030159-g002){ref-type="fig"}F). Within the ventral horn, Ia afferents in both wild-type and Er81^EWS-Pea3/−^ mice formed vGlut1^+^ terminals that were absent in Er81 mutant mice ([Figure 2](#pbio-0030159-g002){ref-type="fig"}G--[2](#pbio-0030159-g002){ref-type="fig"}I). To assess whether synapses between Ia afferents and motor neurons are functional in Er81^EWS-Pea3/−^ mice, we performed intracellular recordings from identified quadriceps motor neurons after stimulation of nerves innervating the quadriceps muscle group. We found no significant difference in the input amplitude to quadriceps motor neurons when comparing wild-type to Er81^EWS-Pea3/−^ mice ([Figure S2](#sg002){ref-type="supplementary-material"}; wild-type, 10.6 ± 0.9 mV, n = 11; Er81^EWS-Pea3/−^, 10.9 ± 1 mV, n = 8). Together, these findings suggest that in the absence of Er81, EWS-Pea3 can direct the complex biological process of correct laminar termination within the ventral spinal cord and the formation of synapses with motor neurons ([Figure 2](#pbio-0030159-g002){ref-type="fig"}J--[2](#pbio-0030159-g002){ref-type="fig"}L), thus identifying an ETS transcription factor suitable for heterochronic expression experiments in DRG neurons. ::: {#pbio-0030159-g002 .fig} Figure 2 ::: {.caption} ###### Rescue of Ia Proprioceptive Afferent Projections into the Ventral Spinal Cord in Er81^EWS-Pea3^ Mutants (A--F) Morphological analysis of central projections at lumbar level L3 of PV^+^ DRG neurons (A--C) or all DRG sensory afferents after application of fluorescently labeled dextran to individual dorsal roots (D--F) in P0.5 (A--C) or P5 (D--F) wild-type (A and D), Er81^−/−^ (B and E), and Er81^EWS-Pea3/−^ (C and F) mice. Red dotted line indicates intermediate level of spinal cord. (G--I) Analysis of vGlut1 immunocytochemistry in the ventral horn of P0.5 wild-type (G), Er81^−/−^ (H), and Er81^EWS-Pea3/−^ (I) mice. Yellow dotted box in (A) indicates size of images shown in (G--I). (J--L) Schematic summary diagrams of the morphological rescue of Ia proprioceptive afferent projections (blue) into the ventral spinal cord observed in wild-type (J), Er81^−/−^ (K), and Er81^EWS-Pea3/−^ (L) mice. DRG indicated by dotted grey line; motor neurons are shown in black. Scale bar: (A--C), 150 μm; (D--F), 160 μm; (G--I), 70 μm. ::: ![](pbio.0030159.g002) ::: Precocious Expression of EWS-Pea3 in DRG Neurons Leads to Axonal Projection Defects {#s2b} ------------------------------------------------------------------------------------- To address the consequences of precocious ETS signaling for proprioceptive afferent differentiation, we next expressed EWS-Pea3 in DRG neurons as soon as they became post-mitotic. We used a binary mouse genetic system based on Cre-recombinase-mediated excision of a transcriptional stop cassette flanked by loxP sites. Targeting cassettes were integrated into the Tau locus to generate two strains of mice conditionally expressing either EWS-Pea3 or a membrane-targeted green fluorescent protein (mGFP) to trace axonal projections of DRG neurons ([Figure S3](#sg003){ref-type="supplementary-material"}; \[[@pbio-0030159-b25]\]). Embryos positive for either Isl1^Cre/+^ and Tau^EWS-Pea3/+^ or Isl1^Cre/+^ and Tau^mGFP/+^ alleles showed efficient activation of the silent Tau allele in 95% or more of all DRG neurons, including proprioceptive afferents, at all segmental levels ([Figure S3](#sg003){ref-type="supplementary-material"}). We first assessed the influence of EWS-Pea3 expression in early post-mitotic DRG neurons on the establishment of afferent projections into the spinal cord using the Tau^mGFP/+^ allele or a Thy1-promoter-driven synaptophysin green fluorescent protein (spGFP) with an expression profile restricted to DRG sensory neurons at embryonic day (E) 13.5 (Thy1^spGFP^; \[[@pbio-0030159-b25]\]) ([Figure 3](#pbio-0030159-g003){ref-type="fig"}). In contrast to wild-type proprioceptive afferent projections ([Figure 3](#pbio-0030159-g003){ref-type="fig"}A--[3](#pbio-0030159-g003){ref-type="fig"}C), GFP^+^ sensory afferents in Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos failed to invade the spinal cord and instead were found in an extreme lateral position at the dorsal root entry zone, a phenotype observed at least up to E18.5 ([Figure 3](#pbio-0030159-g003){ref-type="fig"}A--[3](#pbio-0030159-g003){ref-type="fig"}C and 3G--3I; data not shown). We next visualized the path of sensory afferent projections towards the dorsal root entry zone in Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos by injecting fluorescently labeled dextran into an individual DRG (L3; n = 3; [Figure 3](#pbio-0030159-g003){ref-type="fig"}M--[3](#pbio-0030159-g003){ref-type="fig"}Q). Sensory afferents in E13.5 wild-type embryos bifurcated at their lateral spinal entry point, and projected rostrally and caudally over six or more segmental levels while gradually approaching the midline ([Figure 3](#pbio-0030159-g003){ref-type="fig"}M). Sensory afferents in Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos also bifurcated at the entry point, although approximately 5% of afferent fibers continued to grow towards the midline ([Figure 3](#pbio-0030159-g003){ref-type="fig"}O and [3](#pbio-0030159-g003){ref-type="fig"}Q). While rostro-caudal projections were present in Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos, afferent fibers failed to approach the midline at distal segments and continued to occupy an extreme lateral position ([Figure 3](#pbio-0030159-g003){ref-type="fig"}O), consistent with the analysis of transverse sections. ::: {#pbio-0030159-g003 .fig} Figure 3 ::: {.caption} ###### Defects in the Establishment of Sensory Afferent Projections upon Precocious Expression of EWS-Pea3 in DRG Neurons (A--C and G--I) Visualization of sensory afferent projections (green) into the spinal cord of wild-type (A--C) and Tau^EWS-Pea3/+^ Isl1^Cre/+^ (G--I) embryos at E13.5 (A, C, G, and I) and E16.5 (B and H) by Cre-recombinase-mediated activation of mGFP expression from the Tau locus (A, B, G, and H) or by a Thy1^spGFP^ transgene (C and I; \[[@pbio-0030159-b25]\]). Grey arrows indicate normal pattern of afferent projections into the spinal cord, whereas red arrows show aberrant accumulation of sensory afferents at the lateral edge of the spinal cord in Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos. (D--F and J--L) Analysis of sensory afferent projections (green) into the skin (D and J) or muscle (E and K; red, α-Bungarotoxin, BTX) of wild-type (D--F) and Tau^EWS-Pea3/+^ Isl1^Cre/+^ (J--L) embryos at E16.5 by Cre-recombinase-mediated activation of mGFP (D, E, J, and K) expression from the Tau locus. (F and L) show Egr3 expression in intrafusal muscle fibers using in situ hybridization (consecutive sections to \[E and K\] are shown). (M--Q) Analysis of bifurcation of sensory afferent projections towards the spinal cord in E13.5 wild-type (M) and Tau^EWS-Pea3/+^ Isl1^Cre/+^ (O and Q) embryos after injection of fluorescently labeled dextran (green) into one DRG (lumbar level L3). Confocal scanning plane for (M and O) is schematically illustrated in (N). Inset in (O) is also shown at a deeper confocal scanning plane (P and Q) to visualize aberrant axonal projections. Scale bar: (A and G), 60 μm; (B and H), 80 μm; (C and I), 100 μm; (D and J), 160 μm; (E, F, K, and L), 70 μm; (M, O, and Q), 240 μm. ::: ![](pbio.0030159.g003) ::: We next examined the establishment of peripheral projections upon precocious EWS-Pea3 expression in DRG neurons. While sensory axons in Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos reached the skin and established major nerve trunks by E16.5, only rudimentary sensory axon branching was established within the skin ([Figure 3](#pbio-0030159-g003){ref-type="fig"}D and [3](#pbio-0030159-g003){ref-type="fig"}J). In addition, there was a significant reduction in the number of muscle spindles in Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos (approximately 25% of wild-type complement; n = 3) as assessed by innervation and expression of genes specific for intrafusal muscle fibers such as Egr3 ([Figure 3](#pbio-0030159-g003){ref-type="fig"}E, [3](#pbio-0030159-g003){ref-type="fig"}F, [3](#pbio-0030159-g003){ref-type="fig"}K, and 3L; \[[@pbio-0030159-b26]\]). In summary, whereas isochronic expression of EWS-Pea3 promoted the establishment of proprioceptive afferent projections into the ventral spinal cord, precocious expression of the same ETS signaling factor in DRG neurons interfered with establishment of projections into the spinal cord as well as to peripheral targets. Precocious EWS-Pea3 Expression Promotes Neurotrophin-Independent Survival and Neurite Outgrowth {#s2c} ------------------------------------------------------------------------------------------------- To begin to address the cellular and molecular mechanisms involved in the distinct biological actions of EWS-Pea3 at different developmental stages, we first turned to in vitro culture experiments. These experiments permit assessment of whether precocious ETS transcription factor signaling influences neuronal survival and in vitro neurite outgrowth of DRG neurons, two parameters prominently influenced by target-derived neurotrophic factors and their receptors. We cultured E13.5 whole DRG explants from wild-type and Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos in the presence of NGF or NT-3 or in the absence of neurotrophins and analyzed neuronal survival and neurite outgrowth on matrigel substrate after 48 h in vitro. Without neurotrophic support, very few wild-type DRG neurons survived ([Figure 4](#pbio-0030159-g004){ref-type="fig"}A). In contrast, culturing wild-type DRG with neurotrophic factors led to neuronal survival and neurite outgrowth. Addition of NGF, which supports survival of cutaneous afferents, resulted in straight and unbranched neurite outgrowth ([Figure 4](#pbio-0030159-g004){ref-type="fig"}B), while cultures grown in the presence of NT-3, which supports survival of proprioceptive afferents, resulted in a highly branched neurite outgrowth pattern after 48 h in vitro ([Figure 4](#pbio-0030159-g004){ref-type="fig"}C). Surprisingly, DRG neurons isolated from Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos and cultured without neurotrophic support survived after 48 h in vitro and had established long and highly branched neurites ([Figure 4](#pbio-0030159-g004){ref-type="fig"}D). Neither the pattern of neurite outgrowth nor neuronal survival changed significantly after application of either NGF or NT-3 ([Figure 4](#pbio-0030159-g004){ref-type="fig"}E and [4](#pbio-0030159-g004){ref-type="fig"}F). ::: {#pbio-0030159-g004 .fig} Figure 4 ::: {.caption} ###### Neurotrophin-Independent Neurite Outgrowth In Vitro of DRG Neurons Expressing EWS-Pea3 Precociously E13.5 lumbar DRG from wild-type (A, B, and C), Tau^EWS-Pea3/+^ Isl1^Cre/+^ (D, E, and F), or Bax^−/−^ (G, H, and I) embryos cultured for 48 h without neurotrophic support (A, D, and G) or in the presence of NGF (B, E, and H) or NT-3 (C, F, and I) were stained for expression of neurofilament to visualize axonal extensions. Scale bar: 130 μm. ::: ![](pbio.0030159.g004) ::: To directly compare neurotrophin dependence of DRG neurons expressing EWS-Pea3 from the Tau locus at a precocious versus isochronic time of onset, we generated a strain of mice in which Cre recombinase is expressed from the PV locus ([Figure S4](#sg004){ref-type="supplementary-material"}). The expression of GFP in Tau^mGFP/+^ PV^Cre/+^ was restricted to PV^+^ proprioceptive DRG neurons and mirrored the onset of expression of PV at approximately E14 ([Figure S4](#sg004){ref-type="supplementary-material"}; data not shown). We next cultured E14.5 whole DRG explants from Tau^EWS-Pea3/+^ PV^Cre/+^ and Tau^mGFP/+^ PV^Cre/+^ mice for 48 h in vitro in the presence or absence of NT-3 ([Figure 5](#pbio-0030159-g005){ref-type="fig"}). We found that DRG neurons from both genotypes survived and extended neurites only in the presence of NT-3, whereas they died in the absence of NT-3 ([Figure 5](#pbio-0030159-g005){ref-type="fig"}). Together, these findings suggest that only precocious but not isochronic ETS signaling in DRG neurons is capable of uncoupling survival and neurite outgrowth from a requirement for neurotrophin signaling normally observed in wild-type DRG. ::: {#pbio-0030159-g005 .fig} Figure 5 ::: {.caption} ###### DRG Neurons Expressing EWS-Pea3 Isochronically Depend on Neurotrophins for Survival E14.5 lumbar DRG from Tau^mGFP/+^ PV^Cre/+^ (A and B) and Tau^EWS-Pea3/+^ PV^Cre/+^ (C and D) embryos cultured for 48 h without neurotrophic support (A and C) or in the presence of NT-3 (B and D) were stained for expression of neurofilament (red) and LacZ (green) to visualize axonal extensions and survival of PV-expressing proprioceptive afferents. Scale bar: 150 μm. ::: ![](pbio.0030159.g005) ::: To determine whether neuronal survival of DRG neurons from Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos in the absence of neurotrophic support is sufficient to explain the observed neuronal outgrowth, we analyzed DRG isolated from mice mutant in the proapoptotic gene Bax \[[@pbio-0030159-b27]\]. Consistent with previous results, Bax^−/−^ DRG neurons survived without neurotrophic support \[[@pbio-0030159-b28]\]. In contrast, neurite outgrowth of Bax^−/−^ DRG neurons was significantly less (see [Figure 4](#pbio-0030159-g004){ref-type="fig"}G) than that of either DRG from Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos cultured in the absence of neurotrophic support (see [Figure 4](#pbio-0030159-g004){ref-type="fig"}D) or Bax^−/−^ DRG neurons cultured in the presence of neurotrophic support (see [Figure 4](#pbio-0030159-g004){ref-type="fig"}H and [4](#pbio-0030159-g004){ref-type="fig"}I). These findings suggest that in addition to mediating neurotrophin-independent neuronal survival, expression of EWS-Pea3 in early post-mitotic neurons also promotes neurite outgrowth in a neurotrophin-independent manner. To begin to assess at which step of the neurotrophin signaling cascade DRG neurons from Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos become unresponsive to the addition of neurotrophins, we assayed the expression of neurotrophin receptors in Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos ([Figure 6](#pbio-0030159-g006){ref-type="fig"}). Whereas expression of the neurotrophin receptors TrkA, TrkB, and TrkC marks afferents of distinct sensory modalities in DRG of wild-type embryos ([Figure 6](#pbio-0030159-g006){ref-type="fig"}A--[6](#pbio-0030159-g006){ref-type="fig"}C) \[[@pbio-0030159-b04],[@pbio-0030159-b29]\], Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos showed complete absence of expression of TrkA, TrkB, and TrkC in DRG neurons at E16.5 ([Figure 6](#pbio-0030159-g006){ref-type="fig"}G--[6](#pbio-0030159-g006){ref-type="fig"}I). This absence of Trk receptor expression in DRG of Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos provides a likely explanation for the lack of responsiveness of these neurons to the addition of neurotrophic factors. ::: {#pbio-0030159-g006 .fig} Figure 6 ::: {.caption} ###### Loss of Trk Receptor Expression and Increased Survival in DRG Neurons upon Precocious ETS Signaling (A--C and G--I) In situ hybridization analysis of TrkA (A and G), TrkB (B and H), and TrkC (C and I) expression in E16.5 lumbar DRG of wild-type (A--C) and Tau^EWS-Pea3/+^ Isl1^Cre/+^ (G--I) embryos. (D--F and J--L) Analysis of lumbar DRG of wild-type (D), Tau^mGFP/+^ Isl^Cre/+^ (E and F), and Tau^EWS-Pea3/+^ Isl1^Cre/+^ (J, K, and L) embryos for (1) neuronal cell death at E13.5 by TUNEL (green; D and J), (2) cell survival and proliferation at E16.5 by LacZ (blue) wholemount staining (E and K; lumbar levels L1 and L2 are shown), and (3) BrdU (green)/LacZ (red) double labeling (F and L). (M and N) Quantitative analysis (n ≥ 3 independent experiments) of the mean number of apoptotic events relative to wild-type levels is shown in (M) and neuronal survival in (N) as percent of wild-type of DRG at lumbar levels L1 to L5 as quantified on serial sections. \(O) Western blot analysis of protein extracts isolated from lumbar DRG of E16.5 wild-type (wt) and Tau^EWS-Pea3/+^ Isl1^Cre/+^ (mut) embryos using the following antibodies: Akt, p-Akt (Ser473), CREB, p-CREB (Ser133), Bax, Bcl2, and Bcl-xl. \(P) Quantitative analysis of protein levels relative to wild-type in percent is shown on the right (n = 3 independent experiments). Scale bar: (A--C and G--I), 35 μm; (D and J), 40 μm; (E and K), 200 μm; (F and L), 50 μm. ::: ![](pbio.0030159.g006) ::: We next assayed whether the complete absence of Trk receptor expression in Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos had an influence on naturally occurring cell death in vivo using TUNEL on DRG sections. Surprisingly, we found that apoptosis was decreased by approximately 50% (n = 3 embryos, average of \>50 sections) in DRG of Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos in comparison to wild-type ([Figure 6](#pbio-0030159-g006){ref-type="fig"}D, [6](#pbio-0030159-g006){ref-type="fig"}J, and 6M). Quantitative analysis of the number of neurons in lumbar DRG of Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos revealed a significant increase to approximately 170% of wild-type levels ([Figure 6](#pbio-0030159-g006){ref-type="fig"}E, [6](#pbio-0030159-g006){ref-type="fig"}K, and 6N). Moreover, BrdU pulse-chase experiments ruled out the possibility that DRG neurons in Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos reenter the cell cycle (no BrdU^+^/LacZ^+^ cells, n = 3 embryos, analysis of \>50 sections each; [Figure 6](#pbio-0030159-g006){ref-type="fig"}F and [6](#pbio-0030159-g006){ref-type="fig"}L). Together with the in vitro culture experiments, these findings suggest that DRG neurons from Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos remain post-mitotic but fail to become sensitive to naturally occurring cell death, and survive in the absence of Trk receptors and neurotrophic support. We next analyzed whether changes in the expression of proteins known to be involved in the regulation of neuronal survival or cell death could be detected in DRG of Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos. We found no significant quantitative changes in the level of Akt/p-Akt or CREB/p-CREB in DRG ([Figure 6](#pbio-0030159-g006){ref-type="fig"}O and [6](#pbio-0030159-g006){ref-type="fig"}P) both of which have been shown to be key regulators of neuronal survival \[[@pbio-0030159-b29]\]. Moreover, the level of the proapoptotic Bcl2 family member Bax was not significantly reduced ([Figure 6](#pbio-0030159-g006){ref-type="fig"}O and [6](#pbio-0030159-g006){ref-type="fig"}P). In contrast, the expression level of the anti-apoptotic Bcl2 family members Bcl-xl and Bcl2 was significantly increased when compared to wild-type levels (Bcl2, 157%; Bcl-xl, 259%; average of n = 3 independent experiments; [Figure 6](#pbio-0030159-g006){ref-type="fig"}O and [6](#pbio-0030159-g006){ref-type="fig"}P), providing a potential molecular explanation for the enhanced neuronal survival of DRG neurons of Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos in the absence of Trk receptor expression \[[@pbio-0030159-b30]\]. Only Precocious but Not Isochronic ETS Signaling in DRG Neurons Interferes with Neuronal Fate Acquisition {#s2d} --------------------------------------------------------------------------------------------------------- The observed differences in neuronal survival and neurite outgrowth between precocious and isochronic expression of EWS-Pea3 prompted us to perform a direct comparative analysis of gene expression between mice with precocious EWS-Pea3 expression and mice in which the expression of EWS-Pea3 is initiated in DRG sensory neurons from the time of normal onset of Er81 expression. Moreover, to rule out the possibility that a differential effect may be due to the different genetic strategies by which expression of EWS-Pea3 in proprioceptive afferents is achieved, we performed this analysis both in Er81^EWS-Pea3/−^ and Tau^EWS-Pea3/+^ PV^Cre/+^ embryos. We first analyzed expression of TrkC, a gene downregulated in DRG neurons of Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos ([Figure 7](#pbio-0030159-g007){ref-type="fig"}A and [7](#pbio-0030159-g007){ref-type="fig"}B). The level of expression of TrkC was indistinguishable from wild-type in DRG neurons of Er81^EWS-Pea3/−^ and Tau^EWS-Pea3/+^ PV^Cre/+^ embryos ([Figure 7](#pbio-0030159-g007){ref-type="fig"}A, [7](#pbio-0030159-g007){ref-type="fig"}C, and 7D). Moreover, PV was not expressed in DRG neurons of Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos ([Figure S5](#sg005){ref-type="supplementary-material"}) but was expressed by proprioceptive afferents in both wild-type and Er81^EWS-Pea3/−^ embryos (see [Figures 1](#pbio-0030159-g001){ref-type="fig"} and S5) \[[@pbio-0030159-b14]\]. We also found several genes that were ectopically upregulated in DRG neurons of Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos ([Figure 7](#pbio-0030159-g007){ref-type="fig"}). Calretinin and Calbindin, two different calcium-binding proteins expressed by subpopulations of DRG neurons in wild-type, Er81^EWS-Pea3/−^, and Tau^EWS-Pea3/+^ PV^Cre/+^ embryos ([Figure 7](#pbio-0030159-g007){ref-type="fig"}E, [7](#pbio-0030159-g007){ref-type="fig"}G, and 7H; data not shown) \[[@pbio-0030159-b31],[@pbio-0030159-b32]\], were induced in more than 95% of all DRG neurons of Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos ([Figures 7](#pbio-0030159-g007){ref-type="fig"}F and S5). These findings suggest that DRG neurons in Tau^EWS-Pea3/+^ Isl1^Cre/+^ embryos fail to differentiate to a normal fate and instead acquire an aberrant identity distinct from any subpopulation of wild-type DRG neurons. Finally, to assess whether EWS-Pea3 expressed precociously acts exclusively cell-autonomously or whether it may also influence neighboring DRG neurons, we activated expression of EWS-Pea3 using Hb9^Cre^ mice \[[@pbio-0030159-b33]\]. Due to a transient and rostro-caudally graded expression of Hb9 at neural plate stages, very few DRG neurons at brachial levels and increasingly more neurons progressing caudally undergo recombination in Tau^EWS-Pea3/+^ Hb9^Cre/+^ and Tau^mGFP/+^ Hb9^Cre/+^ embryos ([Figure 8](#pbio-0030159-g008){ref-type="fig"}). Nevertheless, downregulation of Trk receptor expression or upregulation of Calretinin is restricted exclusively to neurons that have undergone recombination and cannot be observed in Tau^mGFP/+^ Hb9^Cre/+^ embryos ([Figure 8](#pbio-0030159-g008){ref-type="fig"}). Together, these results and the findings obtained from in vitro culture experiments (see [Figures 4](#pbio-0030159-g004){ref-type="fig"} and [5](#pbio-0030159-g005){ref-type="fig"}) demonstrate that precocious or isochronic expression of EWS-Pea3 in the same neurons leads to significantly different cell-autonomous cellular responses with respect to gene expression, neuronal survival, and neurite outgrowth ([Figure 9](#pbio-0030159-g009){ref-type="fig"}). ::: {#pbio-0030159-g007 .fig} Figure 7 ::: {.caption} ###### Gene Expression Analysis upon Induction of Precocious or Isochronic ETS Signaling (A--H) Analysis of TrkC expression by in situ hybridization (A--D), or Calretinin (red) and LacZ (green) expression by immunohistochemistry (E--H), on E16.5 lumbar DRG of wild-type (A and E), Tau^EWS-Pea3/+^ Isl1^Cre/+^ (B and F), Er81^EWS-Pea3/−^ (C and G), and Tau^EWS-Pea3/+^ PV^Cre/+^ (D and H) embryos. \(I) Summary diagram illustrating deregulation of TrkC (red arrows, downregulation) and Calretinin (green arrows, upregulation) expression upon precocious (B and F) induction of EWS-Pea3 expression in DRG neurons (B and F; E10--E11, i.e., shortly after cell cycle exit, E9.5--E10). In contrast, activation of EWS-Pea3 from the endogenous Er81 locus (C and G; E12.5--E13) or via Cre recombinase expression from the PV locus activating late expression from the Tau locus (D and H; E14.5) does not interfere with the normal expression of TrkC and Calretinin (shown in grey). Scale bar: (A--D), 65 μm; (E--H), 80 μm. ::: ![](pbio.0030159.g007) ::: ::: {#pbio-0030159-g008 .fig} Figure 8 ::: {.caption} ###### Precocious ETS Signaling Induces Gene Expression Changes Cell-Autonomously (A--D) Expression of TrkA (A and C; green) or TrkC (B and D; green), and LacZ (red), in E16.5 lumbar DRG of Tau^mGFP/+^ Hb9^Cre/+^ (A and B) and Tau^EWS-Pea3/+^ Hb9^Cre/+^ (C and D) embryos. (E--L) Expression of Calretinin (green), LacZ (red), and Isl1 (F, J, H, and L; blue) in E16.5 brachial (E--H) and lumbar (I--L) DRG of Tau^mGFP/+^ Hb9^Cre/+^ (E, F, I, and J) and Tau^EWS-Pea3/+^ Hb9^Cre/+^ (G, H, K, and L) embryos. Scale bar: (A--D), 80 μm; (E--L), 70 μm. ::: ![](pbio.0030159.g008) ::: ::: {#pbio-0030159-g009 .fig} Figure 9 ::: {.caption} ###### Progressive Neuronal Specification Is Paralleled by a Developmental Shift in Response to ETS Transcription Factor Signaling Schematic summary diagram illustrating the importance of temporally appropriate upregulation of transcription factor expression during specification of DRG neurons for late aspects of neuronal differentiation and circuit assembly. (A--D) Expression of EWS-Pea3 from the endogenous Er81 locus can rescue anatomical defects observed in Er81^−/−^ mice, and no change in expression of TrkC (green) or Calretinin (CR; grey) is observed in proprioceptive afferents (A, B, and D). In contrast, precocious ETS signaling leads to severe defects in the establishment of DRG neuronal projections accompanied by inappropriate gene expression changes (C; upregulation of CR (red) and downregulation of TrkC \[grey\]). \(E) Precocious ETS signaling (red) during progressive specification of proprioceptive sensory neurons leads to aberrant neuronal differentiation (red dashed line). In contrast, the isochronic, target-induced (green; peripheral signal) onset of ETS transcription factor signaling (black) induces appropriate terminal neuronal differentiation (blue). ::: ![](pbio.0030159.g009) ::: Discussion {#s3} ========== Target-derived signals exhibit a conserved role in the induction of defined programs of transcription factor expression late in post-mitotic neuronal differentiation \[[@pbio-0030159-b10]\]. This study provides evidence that the late onset of transcription factor expression is essential for many later aspects of neuronal differentiation and circuit formation. Our data indicate that DRG neurons undergo a temporal change in their competence to respond to ETS transcription factor signaling, as assessed by changes in gene expression and axonal target invasion ([Figure 9](#pbio-0030159-g009){ref-type="fig"}). Our findings argue for the necessity of target-induced, and therefore temporally controlled, upregulation of ETS transcription factor signaling. More generally, they suggest that temporally regulated activation of transcriptional programs coupled to a particular fate induced in neurons at early developmental stages represents an important mechanism of neuronal maturation. One striking observation of this study is that precocious induction of ETS signaling promotes neuronal survival without a requirement for neurotrophic support and in complete absence of Trk receptor expression. In contrast, ETS signaling at the normal time of onset of Er81 expression does not result in enhanced neuronal survival in the absence of neurotrophins and also does not lead to downregulation of TrkC expression in proprioceptive afferents. These findings demonstrate very distinct activities of one transcriptional regulator at different developmental steps within a committed post-mitotic neuronal lineage. The absence of Trk receptor expression upon precocious induction of ETS signaling can only partially explain the observed phenotype in axonal projections. Elimination of TrkA receptor signaling in Bax mutant mice perturbs establishment of peripheral projections of cutaneous afferents, whereas establishment of central projections does not appear to be affected \[[@pbio-0030159-b08]\]. In the absence of NT-3 signaling, development of central as well as peripheral proprioceptive afferent projections is perturbed \[[@pbio-0030159-b09]\]. In contrast, upon precocious induction of ETS signaling, we found more pronounced defects in the establishment of central rather than peripheral projections for all DRG neurons. Induction of Er81 expression in proprioceptive afferents is controlled by peripheral NT-3 as axons reach the vicinity of target muscles, and thus occurs only approximately 3 d after proprioceptive neurons become post-mitotic \[[@pbio-0030159-b09],[@pbio-0030159-b14]\]. This temporally delayed and target-induced upregulation of ETS transcription factor expression several days after a neuronal lineage of a specific identity first emerges is not restricted to DRG sensory neurons, but is also found in motor neuron pools \[[@pbio-0030159-b13]\]. Target-derived factors have also been implicated in controlling neuronal maturation of predetermined neurons in Drosophila, in which expression of members of the BMP family in the target region is essential for the induction of mature peptidergic properties in a subpopulation of neurons marked by the coordinate expression of the two transcription factors Apterous and Squeeze \[[@pbio-0030159-b11],[@pbio-0030159-b12]\]. Thus, both in Drosophila and vertebrates, target-derived factors appear to act permissively to induce the expression of transcriptional programs involved in terminal neuronal maturation. Our findings are compatible with a model in which DRG neurons acquire their mature fate by sequential and temporally controlled addition of lineage-specific features ([Figure 9](#pbio-0030159-g009){ref-type="fig"}). Target-derived factors act on predetermined neuronal lineages to switch their developmental programs to become compatible with processes such as target invasion and branching. Such a transition state in the acquisition of a defined neuronal fate would be accompanied by the induction of appropriate transcriptional programs through the expression of specific transcription factors. Mechanisms such as chromosomal remodeling that restrict or expand access to certain target genes \[[@pbio-0030159-b34]\] or activation by cofactors responsible for changing the action of particular transcription factors \[[@pbio-0030159-b35]\] could represent possible mechanisms by which the downstream transcriptional profile of a transcription factor could be temporally shifted towards the selection and control of distinct target genes. Interestingly, several ETS transcription factors are activated through release of autoinhibition via interaction with cofactors and/or via post-translational modifications \[[@pbio-0030159-b35],[@pbio-0030159-b36],[@pbio-0030159-b37]\]. The fusion of EWS with Pea3 could circumvent a need for activation through specific cofactors while still maintaining ETS site dependence, thus rendering EWS-Pea3 less sensitive to the cellular context than endogenous ETS transcription factors. Using this fusion protein, our experiments demonstrate a profound change in the action of ETS signaling at the level of transcriptional regulation within post-mitotic DRG neurons over time. Moreover, the observed transcriptional shift in ETS signaling is paired with the onset of appropriate regulation of neuronal subtype specification and establishment of axonal projections into the target area. Recent experiments addressing the temporal constraints of transcription factor action in proliferating neural progenitor cells adds to the idea that defined temporal windows, during which transcription factors act to control distinct downstream target genes, are of key importance to neuronal fate acquisition. During Drosophila neuroblast generation, the transcription factor Hunchback controls specification and differentiation of early-born neuroblasts \[[@pbio-0030159-b38]\]. Over time, however, neuroblasts progressively lose their competence to generate cells of an early fate in response to Hunchback expression \[[@pbio-0030159-b39]\]. These findings thus also argue for a change in cellular competence to respond to a specific transcription factor over time albeit in an early precursor context. More generally, during the differentiation of hematopoietic lineages, several transcription factors have also been shown to exhibit distinct functions at progressive steps of lineage specification \[[@pbio-0030159-b40]\]. Analysis of the mechanisms by which transcription factor programs can be shifted over time to control different complements of downstream genes and thus different aspects of neuronal and cellular fates in progenitor cells or post-mitotic neurons may provide further insight into the way in which transcription factors act to control the assembly of neuronal circuits. Materials and Methods {#s4} ===================== {#s4a} ### Generation of transgenic mice and mouse genetics {#s4a1} Er81^EWS-Pea3^ mice were generated following a strategy similar to that described for the generation of Er81^NLZ^ mice \[[@pbio-0030159-b14]\]. In brief, a targeting vector with a cDNA coding for EWS-Pea3 was inserted in frame with the endogenous start ATG into exon 2 of the Er81 genomic locus and used for homologous recombination in W95 ES cells. EWS-Pea3 represents a fusion gene between the amino terminal of EWS and the ETS domain of Pea3 \[[@pbio-0030159-b20]\]. The primer pair used to specifically detect the Er81^EWS-Pea3^ allele was 5′- CAGCCACTGCACCTACAAGAC-3′ and 5′- CTTCCTGCTTGATGTCTCCTTC-3′. For the generation of Tau^mGFP^ and Tau^EWS-Pea3^ mice, lox-STOP-lox-mGFP-IRES-NLS-LacZ-pA and lox-STOP-lox-EWS-Pea3-IRES-NLS-LacZ-pA targeting cassettes were integrated into exon 2 of the Tau genomic locus (the endogenous start ATG was removed in the targeting vectors; details available upon request). mGFP was provided by P. Caroni \[[@pbio-0030159-b25]\]. ES cell recombinants were screened by Southern blot analysis using the probe in the 5′ region as described previously \[[@pbio-0030159-b41]\]. Frequency of recombination in 129/Ola ES cells was approximately 1/3 for both Tau constructs. For the generation of PV^Cre^ mice, mouse genomic clones were obtained by screening a 129SV/J genomic library (Incyte, Wilmington, Delaware, United States). For details on the genomic structure of the mouse PV locus see \[[@pbio-0030159-b42]\]. An IRES-Cre-pA targeting cassette \[[@pbio-0030159-b33]\] was integrated into the 3′ UTR of exon 5, and ES cell recombinants were screened with a 5′ probe (oligos, 5′- GAGATGACCCAGCCAGGATGCCTC-3′ and 5′- CTGACCACTCTCGCTCCGGTGTCC-3′; genomic DNA, HindIII digest). The frequency of recombination in 129/Ola ES cells was approximately 1/20. Recombinant clones were aggregated with morula stage embryos to generate chimeric founder mice that transmitted the mutant alleles. In all experiments performed in this study, animals were of mixed genetic background (129/Ola and C57Bl6). Thy1^spGFP^ transgenic mice were generated in analogy to De Paola et al. \[[@pbio-0030159-b25]\], and for these experiments a strain of mice with early embryonic expression was selected. Isl1^Cre^ and Hb9^Cre^ mouse strains have been described \[[@pbio-0030159-b33],[@pbio-0030159-b43]\] and Bax^+/−^ animals were from Jackson Laboratory (Bar Harbor, Maine, United States) \[[@pbio-0030159-b27]\]. Timed pregnancies were set up to generate embryos of different developmental stages with all genotypes described throughout the study. ### Transcriptional transactivation assays {#s4a2} The following plasmids were used for transcriptional transactivation assays: pRc/RSV (Invitrogen, Carlsbad, California, United States), pRc/RSV-Er81, pRc/RSV-EWS-Pea3, pTP-5xETS, and pTP-5xETSmut. pRc/RSV-Er81 and pRc/RSV-EWS-Pea3 were obtained by insertion of the cDNAs for Er81 or EWS-Pea3 (gift from J. A. Hassell) into pRc/RSV. pTP-5xETS was constructed by inserting a cassette of five repetitive copies of high-affinity Pea3 binding sites (5′- GCCGGAAGC-3′) \[[@pbio-0030159-b18],[@pbio-0030159-b19]\] into a modified version of pTK-Luc. pTP-5xETSmut was generated as pTP-5xETS but using a mutated complement of the Pea3 binding sites (5′- GCCTATGGC-3′). A control plasmid to normalize for transfection efficiency (placZ) and pTK-Luc were a gift from D. Kressler. COS-7 cells were co-transfected with 1--1.2 μg of total DNA including one of the effector plasmids pRc/RSV-empty, pRc/RSV-Er81, or pRc/RSV-EWS-Pea3; one of the reporter plasmids pTP-5xETS or pTP-5xETSmut; and placZ. Cells were harvested after 25 h and processed for assays to determine luciferase and LacZ activity as described previously \[[@pbio-0030159-b44]\]. Luciferase values normalized to LacZ activity are referred to as luciferase units. ### In situ hybridization and immunohistochemistry {#s4a3} For in situ hybridization analysis, cryostat sections were hybridized using digoxigenin-labeled probes \[[@pbio-0030159-b45]\] directed against mouse TrkA or TrkB, or rat TrkC (gift from L. F. Parada). Antibodies used in this study were as follows: rabbit anti-Er81 \[[@pbio-0030159-b14]\], rabbit anti-Pea3 \[[@pbio-0030159-b14]\], rabbit anti-PV \[[@pbio-0030159-b14]\], rabbit anti-eGFP (Molecular Probes, Eugene, Oregon, United States), rabbit anti-Calbindin, rabbit anti-Calretinin (Swant, Bellinzona, Switzerland), rabbit anti-CGRP (Chemicon, Temecula, California, United States), rabbit anti-vGlut1 (Synaptic Systems, Goettingen, Germany), rabbit anti-Brn3a (gift from E. Turner), rabbit anti-TrkA and -p75 (gift from L. F. Reichardt), rabbit anti-Runx3 (Kramer and Arber, unpublished reagent), rabbit anti-Rhodamine (Molecular Probes), mouse anti-neurofilament (American Type Culture Collection, Manassas, Virginia, United States), sheep anti-eGFP (Biogenesis, Poole, United Kingdom), goat anti-LacZ \[[@pbio-0030159-b14]\], goat anti-TrkC (gift from L. F. Reichardt), and guinea pig anti-Isl1 \[[@pbio-0030159-b14]\]. Terminal deoxynucleotidyl transferase-mediated biotinylated UTP nick end labeling (TUNEL) to detect apoptotic cells in E13.5 DRG on cryostat sections was performed as described by the manufacturer (Roche, Basel, Switzerland). Quantitative analysis of TUNEL^+^ DRG cells was performed essentially as described \[[@pbio-0030159-b27]\]. BrdU pulse-chase experiments and LacZ wholemount stainings were performed as previously described \[[@pbio-0030159-b46]\]. For anterograde tracing experiments to visualize projections of sensory neurons, rhodamine-conjugated dextran (Molecular Probes) was injected into single lumbar (L3) DRG at E13.5 or applied to whole lumbar dorsal roots (L3) at postnatal day (P) 5 using glass capillaries. After injection, animals were incubated for 2--3 h (E13.5) or overnight (P5). Cryostat sections were processed for immunohistochemistry as described \[[@pbio-0030159-b14]\] using fluorophore-conjugated secondary antibodies (1:1,000, Molecular Probes). Images were collected on an Olympus (Tokyo, Japan) confocal microscope. Images from in situ hybridization experiments were collected with an RT-SPOT camera (Diagnostic Instruments, Sterling Heights, Michigan, United States), and Corel (Eden Prairie, Minnesota, United States) Photo Paint 10.0 was used for digital processing of images. ### In vitro cultures of DRG {#s4a4} Individual lumbar DRG were dissected from E13.5 or E14.5 embryos and placed on Matrigel (BD Biosciences, San Jose, California, United States) coated coverslips in DMEM/F12 (Gibco, San Diego, California, United States), 2 mM L-Gln (Gibco), N2 (Gibco), and 1 mg/ml BSA (Sigma, St. Louis, Missouri, United States) without neurotrophins, or supplemented with either NGF (100 ng/ml, Gibco) or NT-3 (20 ng/ml, Sigma). DRG explants (n ≥ 20 for each condition) were cultured for 48 h, processed for immunocytochemistry, and analyzed using confocal microscopy. ### Western blot analysis {#s4a5} Lumbar DRG from E16.5 embryos were isolated, mechanically disrupted, homogenized using glass beads (Sigma), and lysed in standard lysis buffer supplemented with protease and phosphatase inhibitors as described \[[@pbio-0030159-b47]\]. Protein extracts were resolved by SDS-PAGE, and immunoblotting was performed using antibodies against Akt, p-Akt (Ser473), CREB, p-CREB (Ser133), Bax, Bcl-xl (Cell Signaling Technology, Beverly, Massachusetts, United States), and Bcl2 (BD Pharmingen, San Diego, California, United States). For quantification, films (X-OMAT AR, Eastman Kodak, Rochester, New York, United States) were scanned and densitometry was performed using IMAGEQUANT 5.2 (Molecular Dynamics, Amersham, Uppsala, Sweden). ### Electrophysiology {#s4a6} Electrophysiological analysis was performed as previously described \[[@pbio-0030159-b48]\]. Briefly, intracellular recordings from identified quadriceps motor neurons were made using sharp electrodes (75--120 MΩ, 3M KCl). Average responses (10--20 trials) from suprathreshold nerve stimulation (1.5 times the strength that evokes maximal monosynaptic response) of the quadriceps nerve were acquired with LTP software \[[@pbio-0030159-b49]\]. Only cells with stable resting potentials more negative than −50 mV were considered for analysis. Monosynaptic amplitudes were determined offline using custom routines in the Matlab environment (The Mathworks, Natick, Massachusetts, United States) as previously described \[[@pbio-0030159-b48]\]. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### Gene Expression Analysis in DRG Neurons of Er81^EWS-Pea3^ Mice Analysis of TrkC expression by in situ hybridization (A and E), and Runx3 (red; B and F), Brn3A (red; C and G), Isl1 (red; D and H), p75 (red; I and M), TrkA (red; J and N), CGRP (red; K and O), Calretinin (CR; red; L and P), and LacZ (green; B--D and F--P) by immunohistochemistry on E16.5 lumbar DRG of Er81^NLZ/+^ (A--D and I--L) and Er81^EWS-Pea3/−^ (E--H and M--P) embryos. Scale bar: (A, B, E, F, L, and P), 70 μm; (C, D, G, H, I--K, and M--O), 75 μm. (4.9 MB CDR). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### Ia Proprioceptive Afferents Make Functional Connections with Motor Neurons in Er81^EWS-Pea3^ Mutants (A and B) Representative traces from intracellular recordings measuring Ia afferent monosynaptic input to quadriceps motor neurons evoked by suprathreshold stimulation of the quadriceps nerve in wild-type (A) and Er81^EWS-Pea3/−^ mutant (B) animals. \(C) Average monosynaptic amplitudes (± standard error of the mean) from all recorded cells (wild-type, n = 11; mutant, n = 8). (20 KB CDR). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### Generation of Mice Expressing EWS-Pea3 or mGFP in Early Post-Mitotic DRG Neurons \(A) Top panel shows organization of the Tau genomic locus in the region targeted by homologous recombination in analogy to Tucker et al. \[[@pbio-0030159-b41]\]. Exons 1--3 are shown as light blue boxes, and the Tau start codon in exon 2 is indicated as ATG. The probe used to detect homologous recombination is shown as a grey box. Middle and bottom panels show Tau locus after homologous recombination to integrate targeting cassettes (green) into exon 2 with coincident elimination the endogenous Tau start codon. The integrated targeting cassettes allow for conditional expression of EWS-Pea3 and NLS-LacZ (NLZ) (middle) or mGFP and NLS-LacZ (bottom) upon Cre-recombinase-mediated activation. In the absence of Cre recombinase, a transcriptional stop sequence flanked by loxP sites inhibits expression of the respective transgenes from their start codons (ATG in grey). \(B) Southern blot analysis of Tau^EWS-Pea3/+^ and Tau^mGFP/+^ genomic DNA to detect the mutant allele. \(C) In the presence of Cre recombinase, the transcriptional stop sequence in the cassettes integrated into the Tau locus is removed. Expression of EWS-Pea3 and NLS-LacZ (top) or mGFP and NLS-LacZ (bottom) can now occur in neurons coincidently expressing Cre recombinase and Tau (indicated as ATG in green). (D--L) Expression of Isl1 (D, G, and J), EWS-Pea3 (E and H), GFP (K), or LacZ (F, I, and L), in E12 (D--I) or E13.5 (J--L) DRG neurons of wild-type (D--F), Tau^EWS-Pea3/+^ Isl1^Cre/+^ (G--I), and Tau^mGFP/+^ Isl1^Cre/+^ (J--L) embryos. Scale bar: (D--F), 40 μm; (G--I), 35 μm; (J--K), 50 μm. (1.5 MB CDR). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S4 ::: {.caption} ###### Generation of PV^Cre^ Mice \(A) Above is the organization of the PV genomic locus. Exons are schematically illustrated as light blue boxes, where exon 2 contains the start codon (ATG) and exon 5 contains the stop codon (STOP). Probe to screen for homologous recombination is shown as grey box. Below is a schematic diagram to show the PV locus after the integration of an IRES-Cre cassette (green) 3′ to the translational stop codon of PV using homologous recombination in ES cells. \(B) Southern blot analysis of PV^Cre^ wild-type (+/+) and heterozygous (+/−) genomic DNA using the probe indicated in (A). (C and D) Expression of GFP (green) and LacZ (red; C) or PV (red; D) in P0 Tau^mGFP/+^ PV^Cre/+^ mice. Note that more than 90% of PV^+^ neurons coexpress GFP (D; data not shown). Scale bar: (C and D), 50 μm. (2 MB CDR). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S5 ::: {.caption} ###### Gene Expression Analysis upon Precocious Induction of EWS-Pea3 in DRG Neurons Immunohistochemical analysis of PV (A and D), Calretinin (B and E), and Calbindin (C and F) expression on E16.5 lumbar DRG of wild-type (A--C) and Tau^EWS-Pea3/+^ Isl1^Cre/+^ (D--F) embryos. Scale bar: 80 μm. (950 KB CDR). ::: ::: {.caption} ###### Click here for additional data file. ::: The generation and initial characterization of Er81^EWS-Pea3/−^ mice was performed in the laboratory of Thomas Jessell at Columbia University, with expert assistance from Barbara Hahn and Monica Mendelsohn. We thank Thomas Jessell for encouragement and many stimulating discussions and thank him and Pico Caroni for helpful comments on the manuscript. We thank Jean-Francois Spetz, Patrick Kopp, Bernard Kuchemann, and Monika Mielich for excellent technical assistance, Ina Kramer for providing Runx3 antibodies, Pico Caroni for providing mGFP cDNA, John A. Hassell for providing EWS-Pea3 cDNA and fruitful discussions on ETS gene function, and Y. A. Barde and K. Tucker for providing the Tau targeting construct. SH, EV, MS, TP, CL, DRL, and SA were supported by a grant from the Swiss National Science Foundation, by the Kanton of Basel-Stadt, and by the Novartis Research Foundation. In addition, DRL was supported by a research fellowship from the Roche Research Foundation. Conflict of interest. The authors have declared that no conflicts of interests exist. Author contributions. SH and SA conceived and designed the experiments. SH, EV, MS, TP, CL, DRL, and SA performed the experiments. SA wrote the paper. Citation: Hippenmeyer S, Vrieseling E, Sigrist M, Portmann T, Laengle C, et al. (2005) A developmental switch in the response of DRG neurons to ETS transcription factor signaling. PLoS Biol 3(5): e159. DRG : dorsal root ganglion/ganglia E : embryonic day EWS : Ewing sarcoma mGFP : membrane-targeted green fluorescent protein NT-3 : neurotrophin 3 P : postnatal day PV : Parvalbumin spGFP : synaptophysin green fluorescent protein	PubMed Central	2024-06-05T03:55:55.787835	2005-4-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084331/", "journal": "PLoS Biol. 2005 May 26; 3(5):e159", "authors": [ { "first": "Simon", "last": "Hippenmeyer" }, { "first": "Eline", "last": "Vrieseling" }, { "first": "Markus", "last": "Sigrist" }, { "first": "Thomas", "last": "Portmann" }, { "first": "Celia", "last": "Laengle" }, { "first": "David R", "last": "Ladle" }, { "first": "Silvia", "last": "Arber" } ] }
PMC1084332	Building an embryo is like building a house: everything has to be done at the right time and the right place if the plans are to be translated faithfully. On the building site, if the roofer comes along before the bricklayer has finished, the result may be a bungalow instead of a two-story residence. In the embryo, if the neurons, for example, start to make connections prematurely, the resultant animal may lack feeling in its skin. On the building site, the project manager passes messages to the subcontractors, and they tell the laborers what to do and where. In the embryo, the expression of specific transcription factors (molecules that tell the cell which DNA sequences to convert into proteins) at different stages of development and in different places controls the orderly construction of the body. Silvia Arber and her colleagues are studying the protracted process of neuronal differentiation in mice. Early in development, neurons are generated from dividing progenitor cells. Cell division stops soon after, and long extensions called axons grow out of the neurons in specific directions. When these axons reach their targets---peripheral tissues like the skin at one end, in the case of sensory neurons, and the spinal cord at the other---they form characteristic terminal branches. Finally, the nerves form contacts with other neurons so that they can pass messages on to the brain. Many aspects of neuronal character are acquired through the expression of transcription factors in the progenitor cells or immediately after cell division stops. But Arber and her colleagues have been investigating whether an even later wave of transcription programs is needed for neuronal differentiation and circuit assembly in the sensory neurons of the dorsal root ganglia (DRG), structures containing the cell bodies of the sensory neurons. Previous work indicates that the release of molecules called neurotrophic factors by the neuron\'s target tissues directs the late expression of Er81 and Pea3. These ETS transcription factors (so called because they contain a region known as the erythroblast-transformation-specific domain) control late aspects of the differentiation of DRG neurons. What would happen, the researchers asked, if ETS proteins were expressed earlier? Would precocious ETS expression in DRG neurons also direct the appropriate neuronal developmental programs? Arber\'s team made genetically engineered mice in which ETS signaling occurred either at the correct time or earlier, and examined the development of the proprioceptive sensory neurons, which are involved in the coordination of body balance. In vivo, they found that early initiation of ETS signaling disrupts the axonal growth of the DRG neurons, both to their peripheral targets and into the spinal cord, and perturbs the acquisition of terminal differentiation markers. In vitro, premature ETS signaling allows the DRG neurons to survive and grow in the absence of the neurotrophins normally required for these processes. Arber and her coworkers conclude that the late onset of expression of ETS transcription factors induced by target-derived signals is essential for many of the later aspects of neuronal differentiation and circuit formation. During their differentiation, the researchers suggest, DRG neurons undergo a temporal switch in their ability to respond to ETS signaling. Further analysis of the mechanisms by which responses to transcription factor programs are altered over time during development will advance our understanding not only of neuronal differentiation but of other aspects of embryogenesis.	PubMed Central	2024-06-05T03:55:55.791970	2005-4-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084332/", "journal": "PLoS Biol. 2005 May 26; 3(5):e180", "authors": [] }
PMC1084333	Against the constant threat of infection by bacteria or viruses, one line of defense for the eukaryotic cell is the autophagosome. This double-membrane structure, which buds off from the endoplasmic reticulum, traps cytoplasmic intruders and, upon maturation, merges with a lysosome to destroy them. In this issue, however, Karla Kirkegaard and colleagues show that for one class of viruses, the autophagosome is not a holding cell but a breeding ground, and may even provide a novel escape route. The viruses in question are picornaviruses, which include polioviruses and rhinoviruses. Infection of human cells with poliovirus is known to induce proliferation of double-membrane cytoplasmic vesicles that are morphologically similar to autophagosomes, but the origin and ultimate identity of these vesicles has not been resolved. To test whether these viral-laden vesicles are truly autophagosomes, the authors visualized two proteins: LC3, a specific marker for autophagosomes, and 3A, a part of the poliovirus RNA replication complex. After infection, these proteins colocalized, indicating the poliovirus was indeed within the autophagosome-like vesicles. LC3 also colocalized with LAMP1, a marker for lysosomes, indicating these vesicles mature in a manner similar to that of autophagosomes. This same effect could be induced simply by expressing two viral proteins. All these results indicate that the virus stimulates production of vesicles that bear the traits of autophagosomes and contain the virus, but they don\'t indicate what the consequence is for viral replication. To determine that, the authors increased autophagosome production with two known stimulators of autophagy, tamoxifen and rapamycin. But rather than protecting the cell, this stimulation increased viral yield either 4-fold, in the case of tamoxifen, or 3-fold, for rapamycin. Conversely, inhibiting autophagosome production reduced viral yield. From these results, it seems the virus has subverted the components of the autophagy pathway for its own uses. Inhibiting autophagosome production reduced viral yield inside the cell, but even more so outside. While they were not able to exclude other mechanisms, the authors argue that one possible explanation is that these vesicles are used by the virus to exit from the cell. Supporting this view, they produced electron micrographic images consistent with the fusion of the autophagosome with the plasma membrane and the extracellular release of its contents. This suggests that the virus, which is known to lyse cells to release new viral particles, has another, less lethal means of escape. This may increase the virus\'s chance of avoiding immune system detection as it infects new cells.	PubMed Central	2024-06-05T03:55:55.792432	2005-4-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084333/", "journal": "PLoS Biol. 2005 May 26; 3(5):e195", "authors": [] }
PMC1084334	Introduction {#s1} ============ Organelle function is controlled primarily by the regulation of nuclear gene expression in response to developmental and environmental cues. In turn, organelles signal to the nucleus, in a process termed retrograde signaling, to coordinate the biological activities of the two subcellular compartments. For example, in animals and yeast, mitochondria-to-nucleus and ER-to-nucleus signaling have a dramatic impact on cellular activities under a variety of conditions \[[@pbio-0030151-b01],[@pbio-0030151-b02]\]. In plants, plastid-to-nucleus signaling significantly alters the expression of nuclear genes that encode chloroplast-localized proteins involved in photosynthesis and leaf morphogenesis \[[@pbio-0030151-b03],[@pbio-0030151-b04],[@pbio-0030151-b05],[@pbio-0030151-b06]\]. Therefore, signals originating from plastids play major roles in photoautotrophic growth. Genetic and physiological studies indicate that the accumulation of the chlorophyll precursors Mg-protoporphyrin IX (Mg-Proto) and Mg-protoporphyrin IX monomethyl ester (Mg-ProtoMe) act as a plastid signal that regulates nuclear gene expression in plants and algae \[[@pbio-0030151-b06],[@pbio-0030151-b07],[@pbio-0030151-b08],[@pbio-0030151-b09],[@pbio-0030151-b10],[@pbio-0030151-b11]\]. The current hypothesis proposes that the plastid exports Mg-Proto and/or Mg-ProtoMe, which then interact with a cytoplasmic signaling pathway that ultimately regulates nuclear gene expression \[[@pbio-0030151-b06],[@pbio-0030151-b09],[@pbio-0030151-b11]\]. This proposed model is not without precedent; heme, a tetrapyrrole that bears a striking resemblance to Mg-Proto, regulates gene expression in animal, yeast, and bacterial cells by binding to transcription factors or to kinases that regulate translation \[[@pbio-0030151-b12],[@pbio-0030151-b13],[@pbio-0030151-b14],[@pbio-0030151-b15],[@pbio-0030151-b16]\]. Moreover, the bulk of cellular heme is produced in chloroplasts, which is then transported to other cellular compartments \[[@pbio-0030151-b16],[@pbio-0030151-b17]\]. Because of the molecular similarity between heme and Mg-Proto, it is reasonable to assume that the cellular machinery used to export heme from the chloroplast may be similar to the machinery used for Mg-Proto export. In a search for mutants that affect communication between chloroplasts and the nucleus, a number of mutants, called gun mutants, were identified that have defects in plastid-to-nucleus signaling pathways. These plastid-to-nucleus signaling pathways repress the transcription of nuclear genes that encode proteins active in photosynthesis when chloroplast development is blocked \[[@pbio-0030151-b18],[@pbio-0030151-b19]\]. A number of GUN genes were found to encode factors that participate in Mg-Proto metabolism. Among these were subunits of Mg-chelatase, the enzyme that synthesizes Mg-Proto from protoporphyrin IX (Proto), and indeed, we have shown that buildup of Mg-Proto is a signal that regulates nuclear gene expression \[[@pbio-0030151-b06],[@pbio-0030151-b18]\]. GUN4 participates in the same Mg-Proto signaling pathway that Mg-chelatase does, but GUN4 is not related to any previously described Mg-chelatase subunit or any gene with a known function \[[@pbio-0030151-b20]\]. Purification of a GUN4 complex from Arabidopsis thylakoids revealed that a fraction of GUN4 is tightly associated with GUN5 \[[@pbio-0030151-b20]\], also called ChlH, which is the 140-kDa subunit of Mg-chelatase \[[@pbio-0030151-b18],[@pbio-0030151-b21]\]. Although GUN4 is not essential for Mg-Proto synthesis in vitro or in Arabidopsis, GUN4 is required for chlorophyll accumulation in Arabidopsis under normal growth conditions, GUN4 binds porphyrins in vitro, and GUN4 stimulates Mg-chelatase in vitro \[[@pbio-0030151-b20]\]. Thus, GUN4, like GUN5, is a key participant in the generation of a plastid signal. Also, because GUN4 appears to be monomeric or associated with heterogeneous complexes in fractionated chloroplasts, and because GUN4 binds to Mg-Proto more tightly than GUN5 in Synechocystis, it is reasonable to expect that GUN4 might perform additional functions in porphyrin metabolism \[[@pbio-0030151-b20]\]. For instance, GUN4 might participate in Mg-Proto trafficking or shield Mg-Proto from collisions with molecular oxygen and light, collisions that could result in the production of reactive oxygen species. Alternatively, GUN4 might protect Mg-Proto from catabolic enzymes found in the plastid \[[@pbio-0030151-b22]\], or it might participate in other tetrapyrrole biosynthetic reactions localized to plastids \[[@pbio-0030151-b23]\]. To better understand the porphyrin-binding mechanism and Mg-chelatase stimulatory activity of GUN4, we determined the crystal structure of a GUN4 homolog from the model cyanobacterium Synechocystis sp. PCC 6803 (hereafter referred to as Synechocystis 6803) and refined the three-dimensional model to 1.78 Å resolution. The crystal structure reveals a novel fold that bears no resemblance to known porphyrin-binding proteins. We subsequently used nuclear magnetic resonance (NMR) to map the porphyrin-binding site through analysis of chemical shift data. Quantitative analysis of the putative porphyrin binding site on GUN4 using fluorescence quenching and enzymatic assays has allowed us to determine the energetic contribution of key residues for porphyrin binding, as well as for the enhancement of Mg^2+^ incorporation into metal-free porphyrins. Taken together, these data paint a picture of a novel enzymatic cofactor that enhances Mg-Proto biosynthesis and additionally, may play a role in Mg-Proto shuttling and chemical protection within the cell. Results {#s2} ======= Crystallization, Structure Determination, and Refinement {#s2a} -------------------------------------------------------- The crystallized protein (residues 1--233) comprises the entire Synechocystis GUN4 (SynGUN4) protein. Full-length protein was shown to be competent for binding to both deuteroporphyrin IX (Deutero) and Mg-deuteroporphyrin IX (Mg-Deutero). It was also shown to significantly enhance Mg^2+^ incorporation into Deutero in the presence of the Synechocystis Mg-chelatase, which contains the ChlD, ChlH, and ChlI subunits \[[@pbio-0030151-b21]\]. The SynGUN4 structure was solved by multiple isomorphous replacement using crystals soaked with methyl mercury (II) chloride (Hg) and potassium tetrachloroplatinate II-containing compounds. The resultant model was built and refined to a crystallographic R ~cryst~ and R ~free~ of 22.1% and 25.9%, respectively, using all Hg data extending to 1.78 Å resolution ([Table 1](#pbio-0030151-t001){ref-type="table"}). ::: {#pbio-0030151-t001 .table-wrap} Table 1 ::: {.caption} ###### Statistics from the Crystallographic Analysis ::: ![](pbio.0030151.t001) ^a^ Number in parenthesis is for highest resolution shell ^b^ R ~sym~ = Σ\\|I~h~ − \<I~h~\>\\|/ ΣI~h~, where \<I~h~\> is the average intensity over symmetry equivalent reflections ^c^ R-factor = Σ\\|F~obs~ -- F~calc~\\| / Σ~Fobs~, where summation is over the data used for refinement ^d^ R ~free~ was calculated using 5% of the reflection data chosen and excluded from refinement ::: Structure of GUN4 {#s2b} ----------------- The crystal structure of SynGUN4 reveals a two-domain protein linked by a 12-residue loop ([Figure 1](#pbio-0030151-g001){ref-type="fig"}A). The C-terminal domain, which we refer to herein as the GUN4 core domain, is the conserved domain among all GUN4 family members. The N-terminal domain of SynGUN4 is composed of the first five helices of the full-length protein. The α1′ to α4′ helices fold into a right-handed, up-and-down helix bundle, and the addition of the α5′ helix gives the domain an elongated cross-section. Overall, the N-terminal helical bundle is reminiscent in appearance to other entirely helical domains such as the TPR domain \[[@pbio-0030151-b24]\] or the 14--3-3 domain \[[@pbio-0030151-b25]\]. The entire helical bundle is held together through a hydrophobic core consisting entirely of centrally located and interdigitated leucines, isoleucines, and valines provided by each helix. In contrast to the N-terminal domain\'s core, the surface is highly charged ([Figure 1](#pbio-0030151-g001){ref-type="fig"}B). Structurally, the α2′ and α3′ helices are the linchpins of this tertiary architecture as they bridge one end of the bundle to the other. ::: {#pbio-0030151-g001 .fig} Figure 1 ::: {.caption} ###### Overall Structure of Synechocystis GUN4 \(A) Orthogonal views of the crystal structure of the full length (residues 1--233) Synechocystis GUN4 protein (SynGUN4). Helices are shown as red cylinders and loop regions are displayed as gray loops. SynGUN4 contains two distinct domains linked by a flexible loop. The helices of the N-terminal domain are labeled with apostrophes to distinguish them from the helices making up the C-terminal domain. All structure figures were made with MOLSCRIPT \[[@pbio-0030151-b57]\] and POV-Ray (<http://www.povray.org>). \(B) Orthogonal views of the GRASP \[[@pbio-0030151-b58]\] representation of the SynGUN4 solvent-accessible surface colored to approximately reflect the underlying electrostatic potential, where blue is positive, red is negative, and white is neutral. ::: ![](pbio.0030151.g001) ::: Interestingly, analysis of the known GUN4 sequences from other organisms shows that the N-terminal region is the most highly variable sequence and is unique to only some of the prokaryotic family members. SynGUN4\'s N-terminal region shares homology with GUN4 family members from Gracilaria tenuistipitata var. liui, Porphyra purpurea, Nostoc sp. 7120, Trichodesmium erythraeum, Anabaena variabilis, Thermosynechococcus elongatus, and Cyanidium caldarium ([Figure 2](#pbio-0030151-g002){ref-type="fig"}A). The Cyanidioschyzon merolae sequence lacks the N-terminal extension altogether. Additionally, within the Arabidopsis and rice GUN4 sequences, the N-terminal domain is replaced with a chloroplast transit peptide, which is removed upon import into the chloroplast \[[@pbio-0030151-b26]\]. What role this N-terminal helical bundle plays within prokaryotes is unknown, but most likely it does not participate in GUN4 functions conserved between prokaryotic and eukaryotic organisms. ::: {#pbio-0030151-g002 .fig} Figure 2 ::: {.caption} ###### Sequence Alignment of GUN4 and GUN4-like Proteins \(A) Alignment of the N-terminal portions of GUN4 family members whose N-terminal domains show sequence homology to SynGUN4. Residues contributing to the hydrophobic core of the five-helix bundle are highlighted (pink). GUN4 sequences isolated from plants thus far all have a plastid transit peptide in place of the N-terminal domain found in SynGUN4. The Chlamydomonas reinhardtii sequence was derived from sequence data produced by the United States Department of Energy Joint Genome Institute (<http://www.jgi.doe.gov/>). The N-terminal sequence of C. reinhardtii is not yet known but it most likely contains a chloroplast transit peptide. \(B) Sequence alignment of possible GUN4 core domains. Residues that form the "palm" of the "cupped hand" are highlighted in pink. Residues from the α6/α7 loop that structure the loop and protrude into the core are highlighted in yellow. Arg214 and Arg217, predicted to be important for binding to porphyrins, are highlighted in blue. Residues that disrupt proper folding when mutated and expressed in E. coli are denoted by an asterisk (\). ::: ![](pbio.0030151.g002) ::: The N-terminal domain links to the C-terminal domain through a long loop (residues 81--93) that connects the α5′ helix of the N-terminal helical bundle to the α1 helix within the GUN4 core domain ([Figure 2](#pbio-0030151-g002){ref-type="fig"}B). The interaction between the two domains is not extensive and is mediated by van der Waal\'s interactions between Pro51 and Leu53 from the α4′ helix and Trp151 and Leu152 from the α4 helix within the C-terminal domain. Additionally, the carbonyl oxygen of Gly101 on the α1 helix is within hydrogen bonding distance of Asn60 and Arg63 located on the α4′ helix. In all, a total of 1,183 Å^2^ of surface area is buried between the two domains, which suggests that the particular arrangement displayed in the crystal packing may be one of several possible orientations juxtaposing the two domains in solution. The C-terminal domain of SynGUN4, the GUN4 core domain, appears to have no currently identified structural homologs, as indicated by a lack of any structural matches from a search of the DALI server \[[@pbio-0030151-b27]\]. The GUN4 core domain maintains two distinct architectural regions: a highly structured helical section that forms the majority of the domain, and two loops (α2/α3 loop and α6/α7 loop), which constitute one side of the domain ([Figure 3](#pbio-0030151-g003){ref-type="fig"}A). The helices adopt a concave shell shape resembling a "cupped hand" on one side of the domain, with the back of the "hand" facing toward the N-terminal helical bundle. Helices α1 and α2 extend like a thumb and index finger to form one side of the "cupped hand." The α3 and α4 helices compose the middle finger of the hand, with the α3/α4 loop forming a knuckle. The α5 and α6 and the α7 and α8 helices form the remaining fingers, respectively, on the opposite side of the "cupped hand" arrangement. ::: {#pbio-0030151-g003 .fig} Figure 3 ::: {.caption} ###### Close-Up View of the GUN4 Core Domain\'s "Cupped Hand" Architecture \(A) Rendered skeletal view of the GUN4 core domain. Helices are shown as red cylinders, and coiled regions are depicted as gray loops. The overall shape resembles that of a "cupped hand." \(B) Rendered view of the solvent-accessible surface of the GUN4 core domain, colored gold. The α6/α7 loop is colored gray and is bound by the remainder of the domain. The "cupped hand" grips this loop. ::: ![](pbio.0030151.g003) ::: In total, the helical section constitutes 70% of the GUN4 core domain. The buried surface within the concave section of the "cupped hand" is highly hydrophobic. Phe105 and Leu116 on the α2 helix; Val135 and Phe138 from the α3 helix; Leu143, Ile146, and Trp150 from the α4 helix; Phe160, Val162, Val166, and Trp167 from the α5 helix; Phe174, Leu177, and Trp178 from the α6 helix; Val218, Ala219, and Tyr223 from the α7 helix; and Trp228 from the α8 helix form an extensive hydrophobic surface or "greasy palm" of the "cupped hand." Lying loosely across this palm is the α6/α7 loop, which is itself very hydrophobic ([Figure 3](#pbio-0030151-g003){ref-type="fig"}B). The loop is striking in that it lacks any clear secondary structure, yet it very neatly folds in on itself via van der Waal\'s interactions between hydrophobic residues. Residues Ile181, Trp183, Trp189, Pro193, Phe196, and Trp198 form a hydrophobic groove within which sits Leu207 and Leu209. The extensive hydrophobicity and intricate arrangement of residues suggest that this loop forms a stable structure that would resist unraveling. In contrast to the intraloop packing, only two residues (Pro208 and Leu210) protrude from the α6/α7 loop into the "greasy palm" of the GUN4 core domain\'s "cupped hand." Mutation of either position to alanine leads to the expression of misfolded protein as determined by inclusion body production in E. coli during attempted purification of recombinant samples (unpublished data). The necessity of both Pro208 and Leu210 in maintaining protein stability is not surprising, given the lack of interaction between the α6/α7 loop and the "cupped hand." Analysis of this region of the structure reveals that the juncture of the α6/α7 loop with the "greasy palm" forms an extended hydrophobic surface that is shielded by the "thumb" (α2 helix) and the "middle finger" (α3 and α4 helices). In total, this structural design forms a cavity that is hydrophobic in nature with a volume of about 5,000 Å^3^. Additionally, analysis of 2Fo-Fc electron density maps contoured at 1σ reveals several well-ordered water molecules within the confines of the greasy palm, which, given the high degree of hydrophobicity of this space, are unusual. Mapping the Porphyrin Binding Site {#s2c} ---------------------------------- In an effort to determine the binding site for porphyrin within SynGUN4, we used NMR to analyze the full length protein in the absence and presence of Deutero. Comparison of spectra obtained from ^1^H-^15^N transverse relaxation-optimized spectroscopy (TROSY) experiments of SynGUN4 in the absence and presence of 1--2 mM Deutero reveals several shifting peaks that were picked with the program CARA ([Figure 4](#pbio-0030151-g004){ref-type="fig"}A) \[[@pbio-0030151-b28]\]. Chemical shifts were calculated for those peaks whose position changes in the presence of Deutero. Partial sequence-specific assignment for the backbone atoms of shifting peaks was obtained using TROSY-type triple resonance, and ^15^N-resolved \[^1^H,^1^H\] nuclear overhauser effect spectroscopy (NOESY) experiments with a ^2^H, ^13^C, ^15^N-labeled sample of SynGUN4 (1--233) ([Figure 4](#pbio-0030151-g004){ref-type="fig"}B). ::: {#pbio-0030151-g004 .fig} Figure 4 ::: {.caption} ###### Analysis of SynGUN4 by NMR \(A) Comparison of spectra obtained from ^1^H-^15^N TROSY experiments of SynGUN4 in the absence (black) and presence (red) of 2 mM deuteroporphyrin. \(B) Normalized chemical shifts for those ^1^H-^15^N cross peaks whose positions change in the presence of 2 mM deuteroporphyrin. In general, the largest shifts cluster for residues on the α6/α7 loop. The remaining positions with significant chemical shifts reside on the "greasy palm" region of SynGUN4. \(C) Rendered ribbon diagram of the Gun4 core domain with the position of the shifting ^1^H-^15^N cross peaks mapped onto the backbone structure of SynGUN4. The magnitude of the chemical shift changes shown corresponds to the color bar at the bottom. Briefly, shifts larger than 2.5 parts per million (ppm) are shown in red, shifts between 2 and 2.5 ppm are shown in orange, shifts between 1.5 and 2 ppm are shown in yellow, and shifts of 1.5 ppm and less are shown in green. ::: ![](pbio.0030151.g004) ::: The vast majority of the residues whose environments drastically change are within the GUN4 core domain ([Figure 4](#pbio-0030151-g004){ref-type="fig"}C). Of the 13 residues that were found to exhibit significant chemical shift perturbations in the presence of Deutero, only two (Tyr59 and Thr72) are located within the nonconserved N-terminal helical bundle domain. Of the remaining residues, three residues (Val135, Val218, and Trp228) form part of the "greasy palm," six residues (Asn187, Arg191, Trp198, Asp199, Ser201, and Pro208) are localized within the α6/α7 loop, and the remaining two residues (Lys180 and Arg217) form the hinge regions at the ends of the α6/α7 loop. The chemical shifts offer a very precise measure of the chemical environment around the backbone nitrogen atoms at these positions. The observed chemical shift changes upon addition of Deutero suggest that the α6/α7 loop is undergoing a significant conformational change and/or that an electron-rich molecule, such as Deutero, lies within close proximity. As mentioned earlier, the interaction between this α6/α7 loop and the hydrophobic "greasy palm" of the GUN4 core domain is not extensive, and rearrangements of either or both necessary to accommodate the porphyrin molecule are feasible in both a dynamic and architectural sense. Taken together, the NMR data and the crystal structure strongly suggest that the porphyrin binding region lies within the pocket formed at the intersection of the α6/α7 loop and the "greasy palm" of the GUN4 core domain\'s "cupped hand" region. Deuteroporphyrin IX and Mg-Deuteroporphyrin IX Binding {#s2d} ------------------------------------------------------ First, the affinity of wild-type SynGUN4 for various porphyrins was investigated ([Figure 5](#pbio-0030151-g005){ref-type="fig"}A). The dissociation constants (K* ~d~) for protoporphyrin analogs, Mg-Deutero (0.449 ± 0.045 μM) and Deutero (0.865 ± 0.146 μM) were measured. Additionally, SynGUN4\'s dissociation constants for deuteroporphyrin IX 2,4-(4,2) hydroxyethyl-vinyl-(deutero-divinyl) (3.94 ± 0.739 μM), hemin (4.73 ± 1.16 μM), N-methyl mesoporphyrin IX (NMMP) (11.0 ± 0.673 μM) and cobalt (III) protoporphyrin IX (Co-Proto) (2.67 ± 0.856 μM) were determined. SynGUN4 displayed the highest affinity for Mg-Deutero and Deutero with a preference for the metal-bound porphyrin. The reported dissociation constants for the Synechocystis Mg-chelatase enzyme, ChlH, are 1.22 ± 0.420 μM for Deutero and 2.43 ± 0.460 μM for Mg-Deutero \[[@pbio-0030151-b29]\]. ::: {#pbio-0030151-g005 .fig} Figure 5 ::: {.caption} ###### Quantitative Analysis of Porphyrin Binding by SynGUN4 \(A) Comparison of the binding of SynGUN4 to analogs of both Proto and Mg-Proto. Both Mg-Deutero and Deutero quench endogenous tryptophan fluorescence upon binding (inset). A single binding site was assumed for the fitted line. \(B) Relative dissociation constants were determined for each mutant and compared to the wild-type dissociation constant for both Deutero (red bars) and Mg-Deutero (green bars). The difference between these two sets of constants was calculated (blue bars). \(C) Rendered ribbon diagram of the GUN4 core domain with the relative dissociation constants of each mutant for Deutero mapped onto the structure. While in some cases several different amino acid replacements were tested at particular positions, only the results obtained for the alanine mutations are mapped on the backbone structure shown. The x-fold change in the magnitude of the affinity of Deutero for each mutant is color-coded, as depicted by the scale shown at the bottom. Most amino acid changes did not alter binding affinity, as shown by the preponderance of light blue. Of the mutations that measurably alter binding affinity, the majority reside on the "greasy palm" of the GUN4 core domain. Several other energetic hotspots reside on the α2/α3 and α6/α7 loops. Positions of mutations that exhibit a greater than 10-fold decrease in affinity are labeled. Positions colored black failed to produce properly folded protein when mutated to alanine and expressed in E. coli. \(D) Rendered ribbon diagram of the GUN4 core domain with the relative dissociation constants of each mutant for Mg-Deutero mapped onto the structure. Color coding is the same as for (C). In contrast to Deutero binding, many more mutants alter in vitro binding as shown by the lesser amount of light blue and the prominence of green and yellow color coding. ::: ![](pbio.0030151.g005) ::: Analysis of SynGUN4\'s affinity for other porphyrins provides a larger context from which to deduce the determinants of binding specificity. Both deuteroporphyrins are smaller than the other porphyrins examined in that they lack two ethylene groups. Hemin and Co-Proto, which closely mimic the size of the naturally occurring substrate Mg-Proto, bind with slightly weaker affinities. This suggests that Mg-Proto and Proto may also bind with slightly weaker affinity then the their Deutero versions. Nonetheless, both hemin and Co-Proto bind as well as deutero-divinyl, which accurately mimics the size of the naturally occurring metal free Proto. This further supports that the less flexible metal bound porphyrins are indeed high affinity ligands for SynGUN4. Interestingly, NMMP, which mimics distorted porphyrins binds with very weak affinity. NMMP is a potent inhibitor of enzymes known to catalyze insertion of metals into porphyrins, including ferrochelatase \[[@pbio-0030151-b21],[@pbio-0030151-b30]\]. SynGUN4, however, appears to favor the more planar metal-bound porphyrins. To quantify the energetic contributions of specific residues in SynGUN4 involved in porphyrin binding, a series of putative porphyrin-binding site mutants were made and dissociation constants were measured. Most of the side chains were mutated to alanine; however, at some positions, other amino acids were investigated to explore the importance of electrostatics, side chain volume, or hydrophobicity in greater detail. Endogenous tryptophan fluorescence quenching of SynGUN4 by Deutero and Mg-Deutero was used to determine the K ~d~ for all point mutants for which protein could be expressed and purified ([Table 2](#pbio-0030151-t002){ref-type="table"}). Relative values of all the dissociation constants for each mutant were determined by comparing the K ~d~ of each mutant to that measured for wild-type SynGUN4 ([Figure 5](#pbio-0030151-g005){ref-type="fig"}B). ::: {#pbio-0030151-t002 .table-wrap} Table 2 ::: {.caption} ###### Dissociation Constants of SynGUN4 Proteins for Deutero and Mg-Deutero ::: ![](pbio.0030151.t002) Endogenous tryptophan fluorescence quenching (330 nm) was used to analyze each SynGUN4 mutant\'s dissociation constant for both Deutero and Mg-Deutero. More than 60 mutants were made, of which those listed above produces stable protein for binding analysis. Quenching data were fit to a single site model using GraphPad Prism version 4.0a for Macintosh (GraphPad Software, <http://www.graphpad.com>) ::: Comparison of the relative values for binding to both Deutero and Mg-Deutero shows that the majority of positions affecting binding cluster in neighboring regions of the GUN4 core domain. The positions that appear to be most critical for binding to both porphyrins reside in the α2/α3 and α6/α7 loops, as well as on the faces of the α2, α3, and α7 helices that form significant parts of the "greasy palm" within the "cupped hand" of SynGUN4. Calculation of the difference between the two sets of the relative K ~d~ values shows that certain residues appear to be more important for binding Deutero instead of Mg-Deutero, and vice versa. For example, Val135 and Val218 appear to be essential for Deutero binding but not for Mg-Deutero binding. Alternatively, Ser125, Gln126, Ile146, Phe160, and A219 appear to be essential for specifically involved in binding Mg-Deutero. Comparison of Mg-Deutero binding to Deutero binding shows that binding of the metal-bound porphyrin is much more sensitive to changes within the GUN4 core domain ([Figure 5](#pbio-0030151-g005){ref-type="fig"}C and [5](#pbio-0030151-g005){ref-type="fig"}D). This difference most likely reflects the highly specialized architecture involved in binding the more rigid Mg-Deutero. Modeling of Porphyrin Binding {#s2e} ----------------------------- Armed with an energy map of relevant positions for porphyrin binding, we generated a model of SynGUN4 bound to Mg-Proto ([Figure 6](#pbio-0030151-g006){ref-type="fig"}). In this model, the porphyrin molecule sits over Leu210 within the α6/α7 loop, deep in the "greasy palm" of SynGUN4. The solvent-exposed section of the α6/α7 and α2/α3 loops bracket the porphyrin, burying it deep within SynGUN4 core domain. Significantly, the carboxyl moieties of the porphyrin insert between Arg214 and Arg217, which would be complementary to the charge of the carboxylic acid groups extending from the porphyrin scaffold. Analysis of the Bacillus subtilis ferrochelatase structure bound to NMMP shows that this porphyrin-binding chelatase uses a pair of conserved arginines to bind the carboxyl moieties extending off the porphyrin scaffold ([Figure 6](#pbio-0030151-g006){ref-type="fig"}A). The Arg214 and Arg217 positions within the α6/α7 loop of SynGUN4 closely resemble the arginine motif found on ferrochelatase, suggesting that this motif in SynGUN4 may function in an analogous fashion upon porphyrin binding ([Figure 6](#pbio-0030151-g006){ref-type="fig"}A). ::: {#pbio-0030151-g006 .fig} Figure 6 ::: {.caption} ###### Model of a Putative SynGUN4 Porphyrin Complex Compared to an Experimentally Determined Structure for Ferrochelatase Bound to NMMP \(A) Comparison of the crystal structure of the B. subtilis ferrochelatase bound to NMMP to the model of the SynGUN4 core domain bound to Mg-Proto. The SynGUN4 core domain • Mg-Proto model was generated by GOLD \[[@pbio-0030151-b54]\]. The carboxylic acid moieties of the porphyrin were staggered between the δ-guanido side chains of Arg214 and Arg217. The position of the arginine loop used to tether the carboxyl moieties of the porphyrin bound to ferrochelatase served as the fixed point for the structural alignment of SynGUN4 and ferrochelatase. \(B) Close-up view of the structural alignment between Mg-Proto (gold) and NMMP (lavender). Attempts to strictly superimpose all of the atoms of the two porphyrins resulted in at least one corner of the porphyrin scaffold residing out of the plane defined by the flat Mg-Proto complex, because of the pucker of NMMP. ::: ![](pbio.0030151.g006) ::: Arg214 is absolutely essential for binding to both porphyrins, with both alanine and glutamate substitutions completely abolishing binding ([Table 2](#pbio-0030151-t002){ref-type="table"}). Sequence comparisons of plant and bacterial GUN4s show that Arg214 is highly conserved. Synechocystis 6803 contains two genes that encode proteins related to Arabidopsis GUN4, previously named SynGUN4b and SynGUN4c, that are less similar to Arabidopsis GUN4 than is SynGUN4 \[[@pbio-0030151-b20]\]. The SynGUN4b sequence has an asparagine at the Arg214 position, which could functionally substitute for the arginine. The SynGUN4c sequence has a tyrosine at this position, however. This leads us to propose that SynGUN4c does not efficiently bind to Proto or Mg-Proto, but rather may interact with other cellular metabolites. An interesting difference between Deutero and Mg-Deutero binding is seen at the Arg217 position. Alanine substitution at this position has no effect on either Deutero or Mg-Deutero binding. However, glutamate substitution at this position drastically reduces affinity for only Mg-Deutero by as much as 15-fold. The Arg217 is not as highly conserved, frequently being replaced by glutamine. The glutamine substitution, however, retains the overall aliphatic nature of the arginine side chain, as well as an amide available for interactions with the carboxyl moiety of Proto or Mg-Proto. In an effort to address this difference, NMMP and Mg-Proto were structurally aligned in the putative SynGUN4 porphyrin binding pocket. Analysis of the NMMP molecule shows that one of its corners lies below the plane of the other three. An effort to model NMMP onto Mg-Proto always results in at least one of the four corners of the NMMP molecule lying outside the flattened plane of the porphyrin structural alignment. If the corners that are most buried within the protein are aligned and held fixed, this superpositioning produces a noticeable difference in the position of the carboxyl moieties between the metal-bound and metal-free porphyrin ([Figure 6](#pbio-0030151-g006){ref-type="fig"}B). Specifically, the position of the carboxyl moiety closest to Arg217 is altered. We propose that the metal-free porphyrins most likely can avoid the electrostatic conflict produced by substituting Arg217 with glutamate through its ability to ruffle or pucker. The metal-bound porphyrin\'s enforced rigidity does not allow it to avoid electrostatic conflict with the glutamate, which results in a loss of binding affinity. These observations collectively suggest that, although Arg217 is not essential for binding to Mg-Proto, it nonetheless plays some role in selectivity of planar porphyrins. Enhancing Mg-Chelatase Activity {#s2f} ------------------------------- The effect of SynGUN4 on the activity of the Mg-chelatase was next investigated. Stoichiometric amounts of SynGUN4 dramatically increased the rate of Mg^2+^ incorporation into Deutero, as determined by an increase in fluorescence emission of Mg-Deutero ([Figure 7](#pbio-0030151-g007){ref-type="fig"}A). The apparent Michaelis constant (K ~m~) for Deutero was 7.46 ± 3.24 μM in a Mg-chelatase reaction containing 200 nM ChlH. The measured K ~m~ for Deutero corresponded well with that previously reported for Synechocystis 6803 Mg-chelatase \[[@pbio-0030151-b29]\]. In Mg-chelatase reactions containing SynGUN4, the apparent K ~m~ for Deutero was 1.48 ± 0.371 μM at 200 nM SynGUN4, and 1.65 ± 0.357 μM at 400 nM SynGUN4. ::: {#pbio-0030151-g007 .fig} Figure 7 ::: {.caption} ###### Quantitative Analysis of the GUN4 Stimulation of Mg^2+^ Incorporation into Deutero \(A) The K ~m~ values for GUN4 assisted Mg-Deutero biosynthesis were determined using both 0.2 μm (red) and 0.4 μM SynGUN4 (green). The resultant values contrast with those obtained for Mg^2+^ incorporation by the Mg-chelatase complex in the absence of SynGUN4 (blue). \(B) Relative K ~m~ values (compared to wild-type SynGUN4) were determined for each SynGun4 mutant previously examined for Deutero binding. \(C) Rendered ribbon diagrams of orthogonal views of the SynGUN4 core domain with the relative K ~m~ values of each SynGUN4 mutant mapped onto the structure. The color-coded scale for each mutation\'s effect on Mg^2+^ incorporation is shown at the bottom. Several mutants that altered Mg^2+^ incorporation activity but previously did not affect binding to deuteroporphyrin IX were uncovered including a R217A mutation and the Trp129 and Tyr131 residing on the α2/α3 loop. Mutants of Val218, the later of which is critical for binding Deutero but not for binding Mg-Deutero showed no effect on chelatase activity while mutants of Ala219, the later of which is essential for binding to Mg-Deutero, completely failed to stimulate Mg-chelatase activity. Only those mutant SynGUN4s exhibiting a greater than 10-fold change in K ~m~ are labeled. Shown in black are residues that, when mutated to alanine, failed to produce properly folded protein upon expression in E. coli. ::: ![](pbio.0030151.g007) ::: Interestingly, the Mg-chelatase complex from Rhodobacter sphaeroides, which carries out anoxygenic photosynthesis and lacks a GUN4 homolog, has a much lower K ~m~ than that measured for Synechocystis Mg-chelatase. This seems to suggest that the development of GUN4 activity coincided with a decrease in the basal catalytic efficiency of the Mg-chelatase complex in the absence of GUN4. Oxygenic phototrophs may require a porphyrin-binding protein such as GUN4 to minimize oxidative stress induced by free Proto and Mg-Proto in the presence of O~2~ and bright light. Mg-Proto shielding may be a critical event in the evolution of photosynthesis in oxygen-rich environments. Next, each point mutant previously tested for Deutero and Mg-Deutero binding was assayed for its ability to enhance Mg^2+^ incorporation carried out by the Mg-chelatase complex ([Figure 7](#pbio-0030151-g007){ref-type="fig"}B). Measurements at several time points were analyzed, and the resulting values were compared to the K ~m~ determined for SynGUN4 enhancement of Mg^2+^ incorporation. Many of the mutants that had no energetic effect on porphyrin binding also did not dramatically affect SynGUN4\'s stimulatory activity on the holo Mg-chelatase. Several mutants did demonstrate significant effects on enzymatic activity, however. Many of the mutants, not surprisingly, are clustered in the α6/α7 loop, including residues that are essential for binding porphyrin ([Figure 7](#pbio-0030151-g007){ref-type="fig"}C). Specifically, a region of the α2/α3 loop (Trp129, Leu130, and Tyr131) appears to be essential for Mg^2+^ incorporation. This loop region is moderately important for porphyrin binding as well. No mutants deficient in Mg-Proto binding enhanced Mg-chelatase complex activity. These data support a mechanism in which porphyrin binding by SynGUN4 underlies its stimulatory effect on the Mg-chelatase complex. A mechanism in which SynGUN4 stimulates the Mg-chelatase complex by assisting with product release, as opposed to localizing substrates, is further supported by the fact that the Val218 position, which is essential for binding to Deutero but not for Mg-Deutero, had no effect on Mg^2+^ incorporation. The Ala219 position, which is essential for Mg-Deutero binding, did have a significant effect on stimulation of Mg-chelatase activity. Conversely, the Ala219 position is not essential for binding to the metal-free porphyrin. Nonetheless, some mutants that had only moderate to no effect on binding appear to dramatically alter enzymatic activity. Among these are two residues (Cys117 and Ile146) within the putative porphyrin binding pocket; one residue (Phe132) on the α3 helix; and Arg217. Serine substitution at the Cys117 position only moderately affects porphyrin binding while completely abolishing Mg-chelatase activity. This cysteine lies directly behind the α2/α3 loop, which has been shown here to be critical for GUN4-assisted Mg^2+^ incorporation activity. The role of the sulfhydryl group in stabilizing loop rearrangements is not fully understood, but clearly, substitution of a hydroxyl group at this position does not suffice to replace wild-type activity. Interestingly, mutation of Phe132 to alanine moderately affects binding to both porphyrins. Mutation to tyrosine and tryptophan restores normal binding activity. Nonetheless, the tryptophan mutation at this position has severely compromised enzymatic activity. This suggests that this section of SynGUN4 may modulate interactions with the Mg-chelatase complex. The more bulky hydrophobic groups may help to shield the Arg217-porphyrin interaction from the solvent. However, a too-large side chain at this position, such as a tryptophan, may prevent the GUN4 • Mg-chelatase interactions necessary to assist in product release from the chelatase enzyme. Finally, the Arg217 substitution to alanine abolishes GUN4\'s stimulatory effect on the Mg-chelatase complex while having no apparent effect on binding of either Deutero or Mg-Deutero to GUN4. This suggests that Arg217 plays a role in the selection of chelatase products that is dependent on the presence of the chelatase enzymes. Discussion {#s3} ========== Mg-Proto accumulation is a plastid signal that regulates the expression of a large number of nuclear genes encoding proteins localized to the chloroplast \[[@pbio-0030151-b06]\]. GUN4 is a recently discovered Mg-Proto and Mg-chelatase subunit binding protein that appears to participate in plastid-to-nucleus signaling and chlorophyll synthesis in Arabidopsis \[[@pbio-0030151-b20]\]. Here, we describe the novel three-dimensional structure of SynGUN4, an Arabidopsis GUN4 homolog, as well as the porphyrin binding properties of GUN4. Three GUN4 relatives in Synechocystis 6803, previously named SynGUN4a (referred to as SynGUN4 in this paper), SynGUN4b, and SynGUN4c \[[@pbio-0030151-b20]\], correspond to loci sll0558, sll1380, and slr1958, respectively \[[@pbio-0030151-b31]\]. In contrast, GUN4 is a single-copy gene in plant genomes \[[@pbio-0030151-b20]\]. Because the amino acid sequence derived from SynGUN4/sll0558 is more similar to Arabidopsis GUN4 than are those derived from the other two Synechocystis 6803 GUN4 relatives, it was concluded that SynGUN4/sll0558 was most likely the ortholog of Arabidopsis GUN4 \[[@pbio-0030151-b20]\]. Recently, a null allele of SynGUN4/sll0558 was reported to cause a severe chlorophyll-deficient phenotype, while a null allele of SynGUN4b/sll1380 was reported to have no effect on chlorophyll levels \[[@pbio-0030151-b23]\]. An Arabidopsis null mutant, gun4--2, does not accumulate chlorophyll under normal growth conditions \[[@pbio-0030151-b20]\], which suggests that SynGUN4/sll0558 is the true ortholog of Arabidopsis GUN4. Therefore, SynGUN4 residues that are important for porphyrin binding and Mg-chelatase stimulation are likely to be important in plant GUN4 proteins as well. GUN4 appears to play a role in photoprotection, since Arabidopsis plants containing the null allele, gun4--2, accumulate chlorophyll only in very dim light conditions \[[@pbio-0030151-b20]\]. Our results are consistent with GUN4 having a photoprotective role within the chloroplast. Biophysical and biochemical data support a model wherein the hydrophobic "greasy palm" of the GUN4 core domain, formed by helices that comprise the fingers of the "cupped hand," wrap around the porphyrin molecule, potentially shielding it from the bulk solvent of the plant or bacterial cell. Our model suggests that porphyrins are buried deep within the GUN4 core domain, where they would be protected from collisions with molecular oxygen. The unique "cupped hand" fold of the GUN4 core domain could serve as a novel vehicle for shuttling porphyrin within the chloroplast while inhibiting cellular damage caused by porphyrin-mediated generation of reactive oxygen species formed in the presence of bright light and oxygen. In addition to shielding porphyrin from the oxygen-rich environment of the cell, GUN4 also appears to be a necessary component of the Mg-chelatase in oxygenic photosynthetic organisms. For example, in Rhodobacter capsulatus, which carries out anoxygenic photosynthesis, bacteriochlorophyll synthesis is rapidly shut down in the presence of molecular oxygen and bright light by a mechanism that involves both gene repression and Mg-chelatase inhibition \[[@pbio-0030151-b21]\]. Under these conditions, the largest subunit of R. capsulatus Mg-chelatase, BchH, forms covalent adducts with bound Proto, which inactivates the enzyme \[[@pbio-0030151-b32]\]. Species that perform anoxygenic photosynthesis do not appear to contain GUN4-like genes \[[@pbio-0030151-b20]\]. The largest subunit of the Mg-chelatase complex from both Rhodobacter (BchH) and Synechocystis (ChlH) bind to both Mg-Proto and Proto, as well as to their analogs, Mg-Deutero and Deutero. However, in contrast to ClhH, BchH binds to the metal-free porphyrin (Deutero) with higher affinity than to Mg-Deutero. It is possible that GUN4 enables ChlH to bind Mg-Proto tightly, like BchH, while simultaneously providing photoprotective functions required for species that carry out oxygenic photosynthesis. Distortion of metal-bound porphyrin molecules requires more energy (65--130 kJ/mole) than is needed to distort metal-free porphyrins (48 kJ/mole) \[[@pbio-0030151-b33]\]. In ferrochelatase, once the metal ion is inserted, the porphyrin becomes much more difficult to bend, leading to a decrease in ferrochelatase affinity and ultimately leading to metallized porphyrin release. Conversely, GUN5 maintains some affinity for its metal-containing porphyrin products, creating a potentially deleterious situation in which inhibitory adducts form under conditions of oxygenic photosynthesis. This energetic observation suggests a mechanism in which GUN4 assists in product release from the Mg-chelatase enzyme through its preference for the more planar and rigid Mg-Proto species. In turn, GUN4 may prevent the formation of covalent adducts between the largest subunit of the Mg-chelatase and bound porphyrins in oxygenic photosynthetic organisms by stimulating product release while additionally protecting the Mg-chelatase enzyme active site from molecular oxygen and/or bright light. The mechanism in which SynGUN4 stimulates the Mg-chelatase complex by assisting with product release is supported by the fact that the Val218 position, which is essential for binding to Deutero but not Mg-Deutero had no effect on Mg^2+^ incorporation, whereas the Ala219 position, which is essential for Mg-Deutero binding, did have a significant effect on stimulation of Mg-chelatase activity. The question arises as to how GUN4 discriminates between different metal-bound porphyrins. Quantum mechanical calculations predict that the energy required to distort porphyrin bound to Fe^2+^ and Mg^2+^ are nearly equivalent \[[@pbio-0030151-b33]\]. The implication of this calculation is that both metals distort the porphyrin scaffold equally and confer comparable rigidity upon the tetrapyrrole once either metal cation is inserted. The biophysical models, and the observation that GUN4 binds porphyrins containing different metal ions with different affinities, intimate the existence of a mechanism mediating preferential selection of specific metal-bound porphyrins based on the bound cation rather than on any structural differences exerted by the metal onto the porphyrin scaffold. Although we were unable to test Mg-Proto directly because of its decreased solubility (unpublished data), we were able to determine that Co-Proto binds with higher affinity than hemin. This in vitro experiment further suggests that, indeed, there may be some intrinsic difference between metal-bound porphyrins that SynGUN4 is able to distinguish. Copurification with GUN5 and the lower K ~d~ for Mg-Deutero suggest that GUN4 functions specifically in Mg-porphyrin biosynthesis and trafficking, but currently we cannot rule out a role for GUN4 in the biosynthesis and trafficking of porphyrins that contain other metals (e.g., heme). Other porphyrin-binding proteins have been described, among them the mammalian mitochondrial (peripheral-type) benzodiazepine receptor (PBR) and the tryptophan-rich sensory protein (TspO) from R. sphaeroides \[[@pbio-0030151-b34],[@pbio-0030151-b35],[@pbio-0030151-b36],[@pbio-0030151-b37]\]. PBR is found in most tissues but is expressed at high levels in steroid-synthesizing tissues such as the liver, where it is thought to act as a cholesterol transporter in steroidogenesis \[[@pbio-0030151-b38],[@pbio-0030151-b39],[@pbio-0030151-b40]\]. Interestingly, Proto has been shown to inhibit binding of the high affinity PBR agonist 1-(2-chlorophenyl)-N-methyl-(1-methylpropyl)-3-isoquinoline carboxamide (PK 11195) \[[@pbio-0030151-b36]\]. Additionally, E. coli cells expressing PBR uptake Proto, and this activity can be inhibited by cholesterol \[[@pbio-0030151-b41]\]. Another Proto-binding protein, TspO, has been demonstrated to negatively regulate the transcription of specific photosynthetic genes in R. sphaeroides 2.4.1 \[[@pbio-0030151-b37],[@pbio-0030151-b42]\]. The physiological significance of Proto binding to these proteins is not fully understood, although both proteins are believed to be involved in tetrapyrrole homeostasis. TspO and PBR are highly related to one another \[[@pbio-0030151-b43]\]. Interestingly, sequence alignment of these proteins with GUN4 reveals 40% similarity to TspO and 44% similarity to PBR (see [Figure 2](#pbio-0030151-g002){ref-type="fig"}B). What is notable from the sequence alignment is that the homology is clustered in the helical regions. More specifically, many of the residues that constitute the "greasy palm" of the cupped hand of the GUN4 core domains are conserved. Additionally, the Arg214 position is conserved, which is consistent with the purported ability of PBR and TspO to bind to Proto. Both of these proteins are thought to exist as integral membrane transport proteins consisting of five helices. Indeed, only four of the five predicted helical regions show homology to those of GUN4. Notably, the homologous regions are critically involved in porphyrin binding in GUN4. Additionally, PBR and TspO lack the α7 and α8 helices seen in SynGUN4, which, if absent, would create a significant pore through the GUN4 architecture. This pore might represent an essential feature of the transport functions of PBR and/or TspO. The GUN4 core domain fold presented here describes the structure of a dedicated noncatalytic porphyrin-binding protein. Moreover, the combined use of NMR and x-ray crystallography viewed in the context of subsequent quantitative analysis of porphyrin binding allowed us to propose a mechanism for porphyrin sequestration by GUN4 and GUN4-like proteins. Together, these structural and energetic analyses form a solid foundation for understanding GUN4 activity, and they add to our understanding of the role of GUN4 in the stimulation of Mg-chelatase activity, as well as in Mg-Proto retrograde signaling. Materials and Methods {#s4} ===================== {#s4a} ### Cloning, mutagenesis, and purification {#s4a1} The SynGUN4 used in these studies was isolated from Synechocystis genomic DNA \[[@pbio-0030151-b20]\] and was subcloned into the NcoI and BamHI restriction sites in the pHIS8--3 vector \[[@pbio-0030151-b44]\]. E. coli strain BL21 (DE3) was transformed (Stratagene, La Jolla, California, United States) and cells were grown at 37 °C to an OD~600~ of 0.8. Expression was induced by adding IPTG (Boehringer Mannheim, Amsterdam, Netherlands) to 0.1 mM, and cells were allowed to grow for an additional 6 h at 20 °C. Bacteria were harvested by centrifugation (8,000 × g) at 5 °C; cell pellets were stored at −70 °C. Cell pellets were thawed and resuspended in 50 mM Tris-Cl (pH 8.0), 500 mM NaCl, 10 mM imidazole (pH 8.0), 10% (v/v) glycerol, 1% (v/v) Tween-20, and 10 mM β-mercaptoethanol (β-ME) (Sigma, St. Louis, Missouri, United States) at 4 °C. Resuspended cells were sonicated, and lysates were cleared by centrifugation (100,000 × g) at 4 °C. His-tagged SynGUN4 was purified from clarified supernatants by Ni^2+^-chelation chromatography (Qiagen, Valencia, California, United States). Bound protein was eluted in 50 mM Tris-Cl (pH 8.0), 500 mM NaCl, 250 mM imidazole (pH 8.0), 10% (v/v) glycerol, and 10 mM β-ME. The N-terminal octa-histidine tag was removed by thrombin (Sigma) digestion during dialysis against 50 mM Tris-Cl (pH 8.0), 500 mM NaCl, and 20 mM β-ME at 4 °C for 24 h. The sample was further purified over a Superdex 200 26/60 gel filtration column (Amersham Biosciences, Little Chalfont, United Kingdom) equilibrated in dialysis/thrombin cleavage buffer. Peak fractions were collected and dialyzed against 5 mM Tris-Cl (pH 8.0) containing 10 mM β-ME, concentrated to 30 mg/ml using an Amicon Centricon 10 column (Millipore, Billerica, Massachusetts, United States), and stored at −70 °C. All site-directed mutants of SynGUN4 used for activity assays were constructed using the QuikChange (Stratagene) protocol. Wild type and mutants for binding and activity assays were expressed using pHIS8--3 and purified as described above, except the thrombin cleavage step was eliminated. NMR samples were prepared with aforementioned expression procedures in minimal media containing 4 g of L^−1^ ^13^C-D-glucose and 1 g of L^−1^ ^15^N-ammonium sulfate (Cambridge Isotopes Laboratories, Andover, Massachusetts, United States). Protein purification was carried out as described \[[@pbio-0030151-b45]\]. A final yield of ^13^C, ^15^N-labeled Synechocystis GUN4 at a final concentration of 0.6 mM was obtained from 2 l of expression media. ### Crystallization and structure determination {#s4a2} Crystals of SynGUN4 were grown by the hanging drop vapor diffusion method at 4 °C by mixing 1.0 μl of SynGUN4 (30 mg/ml) with 1.0 μl of a reservoir solution containing 7% PEG 8,000 (Sigma), 200 mM NaBr, 50 mM HEPES-Na^+^ (pH 7.5), and 2 mM DTT. Crystals were stabilized in 15% (v/v) ethylene glycol, 5% (v/v) PEG 8,000, 200 mM NaBr, 50 mM HEPES-Na^+^ (pH 7.5), and 2 mM DTT, and rapidly frozen in a 100-K stream of nitrogen gas. Native (2.2 Å) and derivative data were collected on a MacScience Imaging plate detector, DIP 2020k (MacScience Corporation, Yokohama, Japan), using double focusing Pt/Ni coated mirrors and CuKα X-rays from a MacScience M18XHF generator operating at 4.5 kW (50 kV and 90 mA). Data were processed with DENZO and SCALEPACK \[[@pbio-0030151-b46]\]. The crystals contain one molecule per asymmetric unit (52.9% solvent) and belong to the space group P2~1~2~1~2~1~ (a = 64.21 Å, b = 71.19 Å, c = 72.89 Å; α = β = γ = 90.0°). Hg and Pt sites from crystals soaked in 1 mM methyl mercury (II) chloride (Hg) and 1 mM potassium tetrachloroplatinate (II) (Pt), respectively, were located and used for phasing with SOLVE/RESOLVE \[[@pbio-0030151-b47]\]. The figure of merit (FOM) was 0.3 after MIRAS phasing and 0.60--0.67 after density modification \[[@pbio-0030151-b48]\]. Automated model building was carried out in SOLVE/RESOLVE \[[@pbio-0030151-b49]\]. The SOLVE/RESOLVE-generated model was manually improved using the experimental electron density maps displayed in O \[[@pbio-0030151-b50]\]. The resulting model was positionally refined against all of the Hg data, using the default bulk solvent model in CNS with maximum likelihood targets \[[@pbio-0030151-b51]\]. The model consists of residues 1--233 of SynGUN4 and 249 water molecules. PROCHECK \[[@pbio-0030151-b52]\] revealed a total of 91% of the residues in the most favored region of the Ramachandran plot and 9% in the additionally allowed region. Main chain and side chain structural parameters were consistently better than average (overall G value of 0.46). ### NMR experiments {#s4a3} All of the NMR spectra were recorded at 22 °C on a Bruker 700-MHz spectrometer equipped with four radiofrequency channels and a triple-resonance cryoprobe with shielded z-gradient coil. The NMR samples contained 0.6 mM ^13^C, ^15^N-labeled SynGUN4 in 5 mM Tris-Cl (pH 8.0), 300 mM NaCl, 1 mM DTT/95% H~2~O/5% D~2~O. Partial sequence assignment and structure determination were performed with the standard protocol for ^13^C, ^15^N-labeled samples \[[@pbio-0030151-b53]\]. ^1^H, ^13^C, and ^15^N backbone resonances were assigned using the triple-resonance experiments HNCA and CBCA(CO)NH and three-dimensional ^15^N-resolved \[^1^H, ^1^H\]-NOESY experiments. Chemical-shift perturbation experiments were carried out with NMR samples in the absence and presence of 2 mM deuteroporphyrin dissolved in 5 mM Tris-Cl (pH 8.0). ### Binding analysis {#s4a4} Endogenous tryptophan quenching of SynGUN4 was measured using fluorescence spectroscopy as described previously \[[@pbio-0030151-b29]\]. Briefly, serial dilutions of both Deutero and Mg-Deutero were made in DMSO. SynGUN4 was diluted into 50 mM MOPS-KOH (pH 7.7), 150 mM NaCl, and 1 mM DTT. Fluorescence quenching data were collected on a PTI Alphascan spectrofluorimeter (Photon Technology Instruments, Santa Clara, California) by monitoring 200 nM GUN4 tryptophan fluorescence at 332 nm. Dissociation equilibrium constants for GUN4-porphyrin interactions were determined by fitting titration data to a one-site binding model. ### Modeling {#s4a5} The program GOLD was used to dock the Mg-Proto in the GUN4 core domain of SynGUN4 \[[@pbio-0030151-b54]\]. A 16-Å cavity was defined around the carbon atom of the terminal methyl group of residue Leu210 in the crystal structure of SynGUN4. The majority of outputs from GOLD were similar to each other in the position of features such as the porphyrin carboxyl moieties. Minor manual adjustments of the GOLD solution were made using the programs Chimera \[[@pbio-0030151-b55]\] and O \[[@pbio-0030151-b51]\]. ### Mg-chelatase activity {#s4a6} Mg-chelatase subunits were expressed and purified as previously described \[[@pbio-0030151-b20],[@pbio-0030151-b56]\]. Reactions contained indicated concentrations of SynGUN4. Concentrations of the Mg-chelatase components used were 0.1 μM ChlH, 0.1 μM ChlD and 0.2 μM ChlI, which were similar to those used previously \[[@pbio-0030151-b56]\]. Reactions were carried out using Deutero, as previously described, in 50 mM MOPS-KOH (pH 7.7), 20 mM MgCl~2~, 1 mM DTT, and 5 mM ATP \[[@pbio-0030151-b20]\]. Each 100-μl reaction mixture was incubated at 34 °C and stopped by addition of 900 μl of acetone/water/32% (v/v) ammonia (80:20:1, by volume) \[[@pbio-0030151-b56]\]. Reactions were then centrifuged at 15,000 × g for 5 min at room temperature. The Mg-Deutero concentration was determined by fluorescence spectroscopy. The K ~m~ was determined by analysis of the rise in the slope of Mg-Deutero synthesis at several different starting substrate concentrations using the program PRISM 4.0 (GraphPad Software, San Diego, California, United States). Supporting Information {#s5} ====================== Accession Numbers {#s5a1} ----------------- Coordinates for Synechocystis GUN4 (accession number 1Y6I) have been deposited in the Protein Data Bank (<http://www.rcsb.org>). This work was supported by National Institutes of Health/National Cancer Institute grant CA54418 to JPN, by the National Science Foundation grant 0236027 to JPN, and by a grant from the United States Department of Energy to JC. JC is an investigator of the Howard Hughes Medical Institute. Competing interests. The authors have declared that no competing interests exist. Author contributions. MAV, RML, JLF, RR, JC, and JPN conceived and designed the experiments. MAV, RL, JLF, RR, and JC performed the experiments. MAV, RML, JLF, RR, JC, and JPN analyzed the data, contributed reagents/materials/analysis tools, and wrote the paper. ¤a Current address: Department of Neurobiology and Behavior, State University of New York, Stony Brook, New York, United States of America ¤b Current address: MSU-DOE Plant Research Laboratory, Michigan State University, East Lansing, Michigan, United States of America Citation: Verdecia MA, Larkin R, Ferrer JL, Riek R, Chory J, et al. (2005) Structure of the Mg-chelatase cofactor GUN4 reveals a novel hand-shaped fold for porphyrin binding. PLoS Biol 3(5): e151. NMMP : N-methyl mesoporphyrin NMR : nuclear magnetic resonance PBR : peripheral-type benzodiazepine receptor TROSY : transverse relaxation-optimized spectroscopy TspO : tryptophan-rich sensory protein ppm : parts per million	PubMed Central	2024-06-05T03:55:55.793170	2005-4-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084334/", "journal": "PLoS Biol. 2005 May 26; 3(5):e151", "authors": [ { "first": "Mark A", "last": "Verdecia" }, { "first": "Robert M", "last": "Larkin" }, { "first": "Jean-Luc", "last": "Ferrer" }, { "first": "Roland", "last": "Riek" }, { "first": "Joanne", "last": "Chory" }, { "first": "Joseph P", "last": "Noel" } ] }
PMC1084335	Introduction {#s1} ============ Somatic cells within an organism possess genomes that are, with only a few minor exceptions, identical. However, various cell types may possess different epigenomes including the variation of DNA methylation and histone modification patterns. Epigenome variability accounts for cell-type-specific gene expression and silencing patterns in multicellular organisms. The impact of higher-order nuclear architecture on these patterns is not yet known \[[@pbio-0030157-b01]\]. Studies of higher-order chromatin arrangements in numerous cell types from different species form an indispensable part of a comprehensive approach to understanding epigenome evolution and cell-type-specific variability. Numerous research groups have attempted to map the large-scale organization and distribution of chromatin in cycling and postmitotic cell types (for reviews see \[[@pbio-0030157-b02],[@pbio-0030157-b03],[@pbio-0030157-b04],[@pbio-0030157-b05],[@pbio-0030157-b06],[@pbio-0030157-b07],[@pbio-0030157-b08]\]). Reliable topological maps, however, for the three-dimensional (3D) and 4D (3D plus spatiotemporal) arrangements of the two haploid chromosome complements in a diploid somatic cell nucleus have been lacking so far. Such 3D and 4D maps would provide the necessary foundation for studying the effect of higher-order chromatin distribution on nuclear functions, and are needed for different cell types at various stages of the cell cycle and at various stages of terminal differentiation. In addition to their importance for epigenome research, these maps should also help to understand karyotype evolution \[[@pbio-0030157-b09],[@pbio-0030157-b10],[@pbio-0030157-b11],[@pbio-0030157-b12]\] and the formation of chromosomal rearrangements in irradiated or cancer cells \[[@pbio-0030157-b13],[@pbio-0030157-b14],[@pbio-0030157-b15],[@pbio-0030157-b16],[@pbio-0030157-b17]\]. In a 2D analysis of human fibroblast prometaphase rosettes, Nagele et al. \[[@pbio-0030157-b18],[@pbio-0030157-b19]\] measured distances and angular separations for a number of chromosomes. These authors concluded that the maternal and paternal chromosome sets were separate, and that the heterologous chromosomes in each set showed highly nonrandom distributions. Subsequent studies further emphasized a highly ordered chromosome territory (CT) pattern for the nuclei of polarized human bronchial epithelial cells \[[@pbio-0030157-b20]\] and for nuclei of quiescent (G0) diploid (46, XY) human fibroblasts in culture \[[@pbio-0030157-b21]\]. Koss \[[@pbio-0030157-b20]\] reported that angles between the center of the nucleus and homologous pairs of Chromosome 1, 7, and X CTs were nearly identical in about two-thirds of bronchial epithelial cell nuclei to the angles reported by Nagele et al. for the same chromosome pairs in fibroblast prometaphase rosettes \[[@pbio-0030157-b18]\]. In contrast, Allison and Nestor \[[@pbio-0030157-b22]\] found a relatively random array of chromosomes on the mitotic ring of prometaphase and anaphase cells in cultured human diploid fibroblasts, diploid cells from human lung tissue, and human lymphocytes. The causes of these discrepancies have so far remained elusive. For nuclei of human lymphocytes, phytohemagglutinin-stimulated lymphoblasts, and lymphoblastoid cell lines, several groups have consistently reported a preferential positioning of gene-rich CTs (e.g., Homo sapiens chromosome \[HSA\] 19) towards the center of the nucleus, and of gene-poor CTs (e.g., HSAs 18 and Y) towards the nuclear periphery \[[@pbio-0030157-b23],[@pbio-0030157-b24],[@pbio-0030157-b25],[@pbio-0030157-b26]\]. We recently confirmed this gene-density-correlated radial CT positioning for several other normal and malignant human cell types \[[@pbio-0030157-b26]\]. Bickmore and colleagues \[[@pbio-0030157-b23],[@pbio-0030157-b27]\] also reported gene-density-correlated CT arrangements for cycling human fibroblasts. In contrast, Sun et al. \[[@pbio-0030157-b28]\] and our group \[[@pbio-0030157-b23],[@pbio-0030157-b24],[@pbio-0030157-b25],[@pbio-0030157-b26]\] provided support for chromosome-size-correlated radial arrangements in quiescent fibroblasts. Although Sun et al. refer to nuclei studied in the G1-phase of the cell cycle, we believe that most of the cells included in their analysis were in a quiescent state (G0), since fibroblasts were grown on coverslips to 90%--95% confluence. Bridger et al. \[[@pbio-0030157-b27]\] reported that Chromosome 18 CTs were significantly closer to the nuclear periphery in S-phase fibroblasts than in quiescent fibroblasts. These findings suggest that cycling and noncycling fibroblasts differ in higher-order chromatin organization. We tested this hypothesis further in the present study. To overcome some of the technical limitations of previous studies, and to explore some of their inconsistencies, we employed 3D fluorescence in situ hybridization (FISH) protocols that allowed the differential coloring of all 24 chromosome types (22 autosomes plus X and Y) simultaneously within a population of human male fibroblasts (46, XY) under conditions preserving the 3D nuclear shape and structure to the highest possible degree \[[@pbio-0030157-b29],[@pbio-0030157-b30]\]. In addition, we performed a series of two-color 3D FISH experiments in semi-confluent cultures, and determined the radial 3D positions of a subset of CTs (HSAs 1, 17--20, and Y) in quiescent (G0) and cycling (early S-phase) fibroblasts. Our data demonstrate unequivocally that the 3D arrangements of chromosomes in quiescent and cycling human fibroblasts follow probabilistic rules, and suggest that nuclear functions in human fibroblasts do not require a deterministic neighborhood pattern of homologous and heterologous chromosomes. Throughout, when we use the term "probabilistic chromosome order," we mean an order that cannot be explained simply as a consequence of geometrical constraints that affect the distribution of chromosomes in mitotic rosettes or of CTs in cell nuclei. Constraints may enforce an arrangement of large and small chromosomes or CTs that deviates significantly from the prediction of a random order of points without any functional implications. Our long-term goal is to contribute to the elucidation of the set of rules (most likely a combination of probabilistic and deterministic) that generate cell-type-specific, functionally relevant higher-order chromatin arrangements. Results {#s2} ======= Differential Coloring of All 24 Chromosome Types in Nuclei of Human Male Diploid Fibroblasts {#s2a} -------------------------------------------------------------------------------------------- Early-passage human fibroblast cultures (46, XY) were grown to confluence and maintained at this stage for several days before being fixed with buffered 4% paraformaldehyde. Under these conditions, the overwhelming majority (\>99%) of cells were postmitotic (G0), as demonstrated by a lack of both pKi67 staining and incorporation of thymidine analogs (data not shown). Two 3D multiplex FISH (M-FISH) protocols were used for the differential coloring of all 24 human chromosome types (22 autosomes plus X and Y). The first approach was based on 3D M-FISH with 24 chromosome paint probes. Probes were differentially labeled using a combinatorial labeling scheme with seven different haptens/fluorochromes \[[@pbio-0030157-b31]\]. DAPI was used to stain nuclear DNA. Light-optical serial sections were separately recorded for each fluorochrome using digital wide-field epifluorescence microscopy ([Figure 1](#pbio-0030157-g001){ref-type="fig"}). A second approach, called ReFISH \[[@pbio-0030157-b32]\], achieved differential staining of all 24 human chromosome types in two sequential FISH experiments with triple-labeled probe subsets. Light-optical serial sectioning of the same nuclei with laser confocal microscopy was performed after both the first and the second hybridization. Both approaches provided stringent accuracy for color classification of all CTs, and yielded the same results. Therefore, we combined data from 31 nuclei studied with the first approach and from 23 nuclei studied with the second approach (54 nuclei in total). ::: {#pbio-0030157-g001 .fig} Figure 1 ::: {.caption} ###### 24-Color 3D FISH Representation and Classification of Chromosomes in a Human G0 Fibroblast Nucleus \(A) A deconvoluted mid-plane nuclear section recorded by wide-field microscopy in eight channels: one channel for DAPI (DNA counterstain) and seven channels for the following fluorochromes: diethylaminocoumarin (Deac), Spectrum Green (SG), and the cyanine dyes Cy3, Cy3.5, Cy5, Cy5.5, and Cy7. Each channel represents the painting of a CT subset with the respective fluorochrome. The combinatorial labeling scheme is described in [Materials and Methods](#s4){ref-type="sec"}. RGB images of the 24 differently labeled chromosome types (1--22, X, and Y) were produced by superposition of the seven channels (bottom right). \(B) False color representation of all CTs visible in this mid-section after classification with the program goldFISH. \(C) 3D reconstruction of the complete CT arrangement in the nucleus viewed from different angles. \(D) Simulation of a human fibroblast model nucleus according to the SCD model (see [Materials and Methods](#s4){ref-type="sec"}). The first image shows 46 statistically placed rods representing the 46 human chromatids. The next three images simulate the decondensation process and show the resulting CT arrangement obtained after different numbers of Monte Carlo relaxation steps (200, 1,000, and 400,000). This set of figures is taken from [Video S1](#sv001){ref-type="supplementary-material"}. ::: ![](pbio.0030157.g001) ::: Following careful correction for chromatic shifts, and image deconvolution in the case of wide-field microscopy ([Figure S1](#sg001){ref-type="supplementary-material"}), we performed overlays of the corresponding light-optical sections from all channels with voxel accuracy. CT classification was carried out on these overlays by the computer program goldFISH \[[@pbio-0030157-b33]\] ([Figures 1](#pbio-0030157-g001){ref-type="fig"}B and S1C). This program classifies chromosomes by virtue of differences in the combinatorial fluorescent labeling schemes. [Figure 1](#pbio-0030157-g001){ref-type="fig"}C shows the 3D reconstruction of a nucleus with all CTs viewed from different angles. Although the present experiments were not designed to address the issue of chromatin intermingling from neighboring CTs, it is obvious that goldFISH should have led to numerous misclassifications if there were excessive, widespread intermingling (for further discussion of CT boundaries, see \[[@pbio-0030157-b34]\]). For each individual CT the classification achieved by goldFISH was confirmed or rejected by careful visual inspection of light-optical sections. Any CT that could not be classified with certainty was omitted from further consideration. We were thus able to identify 2,030 CTs (82%) from a total of 2,484 CTs present in the 54 diploid fibroblast nuclei. As reference points for all distance and angle measurements reported below, we determined the 3D location of the fluorescence intensity gravity centers (IGCs) of individual painted CTs and the IGC of the nucleus (CN). Unless stated otherwise, when we describe below the position of a CT or prometaphase chromosome (PC) and report distance and angle measurements, we are referring to the 3D position of the CT\'s or PC\'s IGC. As a control for the reliability of the CT localizations, we subjected nuclei first studied by 24-color 3D FISH to a sequential five-color FISH experiment with individually labeled paint probes for Chromosomes 1 (Cy5), 3 (Cy3), 10 (FITC), 12 (Cy3.5), and 20 (Cy5.5). We were able to retrieve 11 of the 31 originally studied nuclei and to determine whether 3D positions of CTs first classified in the 24-color 3D FISH experiment could be confirmed after the second hybridization. In 96% of the re-hybridized CTs, the 3D position of the IGC differed by less than 1 μm, the range being between 0.01 and 1.3 μm. Size-Correlated Radial CT Positions in Nuclei of Quiescent (G0) Fibroblasts {#s2b} --------------------------------------------------------------------------- For every identified CT we measured the 3D radial CN--CT distance (from the CN to the CT\'s IGC). For a graphic overview of the location of each CT in 2D nuclear projections, the 3D positions of all IGCs obtained for a given CT were normalized and drawn into an ellipse representing the nuclear rim ([Figure 2](#pbio-0030157-g002){ref-type="fig"}). As representative examples, [Figure 2](#pbio-0030157-g002){ref-type="fig"}A shows nuclear projections of the normalized 3D IGC locations of CTs of HSAs 1, 7, 11, 18, 19, and Y, while [Figure 2](#pbio-0030157-g002){ref-type="fig"}B shows cumulative 3D CN--CT graphs for the same CTs. Figures [S2](#sg002){ref-type="supplementary-material"} and [S4](#sg004){ref-type="supplementary-material"} provide the respective data for the entire chromosome complement. Notably, 3D radial CN--CT distance measurements did not reveal a significant difference between the positions of the gene-poor HSA 18 and the gene-rich HSA 19, although distinctly peripheral and interior locations, respectively, have been found for these two chromosomes in the spherical nuclei of lymphocytes and several other cell types (see Introduction). In summary, our data ([Figures 2](#pbio-0030157-g002){ref-type="fig"}B, [S2](#sg002){ref-type="supplementary-material"}, [S4](#sg004){ref-type="supplementary-material"}, and [S7](#sg007){ref-type="supplementary-material"} \[left panel\]) demonstrate that the territories of all small chromosomes---independent of their gene density---were preferentially found close to the center of the nucleus, while the territories of large chromosomes were preferentially located towards the nuclear rim. [Figure 3](#pbio-0030157-g003){ref-type="fig"} displays the positive correlation obtained in quiescent human fibroblasts for the mean normalized radial CN--CT distances and the DNA content of the chromosomes. The broad variability of radial CT positions seen in the set of 54 G0 nuclei indicates that radial CT arrangements in quiescent fibroblasts follow probabilistic, not deterministic, rules. ::: {#pbio-0030157-g002 .fig} Figure 2 ::: {.caption} ###### Radial Chromosome Positions Correlate with Chromosome Size in Quiescent Human Fibroblasts (G0) \(A) Two-dimensional projections of the IGCs of CTs 1, 7, 11, 18, 19, and Y studied in 54 nuclei are represented by dots. Ellipses represent the 2D shape of a fibroblast nucleus normalized for shape and size and rotated so that the long axis of each nucleus evaluated lies on the abscissa. Projections of IGCs for all CTs are shown in [Figure S2](#sg002){ref-type="supplementary-material"}. Note that we were not able to distinguish in nuclei between a "north" and "south" pole of the short axis or a "west" and "east" pole of the long axis. Either fibroblast nuclei do not possess such compass polarizations or we lack markers to recognize them. Accordingly, distance comparisons between IGCs located in different quadrants of the ellipse are not meaningful. \(B) Cumulative 3D distance graphs of the CT distribution within a normalized nucleus taken from the data (A). The abscissa represents the normalized radial 3D distances of CTs 1, 7, 11, 18, 19, and Y from the center of the nucleus (CN; IGC of the DAPI-stained nucleus) to the IGC of a specific CT. The origin represents the CN, and "1" represents the nuclear periphery. The Ordinate represents the cumulative percentage of normalized 3D CN--CT distances. Cumulative graphs for the entire set of chromosomes are shown in [Figure S4](#sg004){ref-type="supplementary-material"}. \(C) Cumulative 3D distance graphs of PC distribution within a normalized prometaphase rosette. Abscissa and ordinate are as in (B), with PC being the IGC of a prometaphase chromosome and CR the center of the prometaphase rosette. IGC projections and cumulative graphs for all PCs are shown in Figures [S3](#sg003){ref-type="supplementary-material"} and [S6](#sg006){ref-type="supplementary-material"}, respectively. ::: ![](pbio.0030157.g002) ::: ::: {#pbio-0030157-g003 .fig} Figure 3 ::: {.caption} ###### Normalized Radial Chromosome Distances in G0 Fibroblast Nuclei, SCD Model Nuclei, and Prometaphase Rosettes \(A) Normalized 3D radial CN--CT distances (filled squares) show a positive correlation with chromosome size (indicated by DNA content): CTs of small chromosomes were preferentially located in the center of the nucleus, whereas CTs of large chromosomes were found more often at the nuclear periphery. Open circles 1 and 21 indicate the endpoints of SCD model data simulating a statistical placement of CTs. \(B) SCD model data indicate that geometrical constraints result in a reverse pattern of CT distributions, i.e., modeled small CTs show a significantly higher probability of being localized at the nuclear periphery, while modeled large CTs adopt a more internal localization. \(C) In agreement with normalized 3D CN--CT distances, normalized 3D CR--PC distances also show a positive correlation with the DNA content or size of chromosomes: small PCs were preferentially located near the CR, large chromosomes at the rosette periphery. ::: ![](pbio.0030157.g003) ::: To visualize the relative average positions of the IGCs of all heterologous CTs, we generated multidimensional scaling (MDS) plots \[[@pbio-0030157-b35],[@pbio-0030157-b36]\] based on the mean of all normalized 3D CT--CT distances ([Figure 4](#pbio-0030157-g004){ref-type="fig"}). Consistent with the data shown in [Figure 3](#pbio-0030157-g003){ref-type="fig"}A, we found CTs from small chromosomes preferentially clustering towards the center of the nucleus, while CTs from large chromosomes were preferentially located towards the periphery. ::: {#pbio-0030157-g004 .fig} Figure 4 ::: {.caption} ###### MDS Plots of Experimental and Simulated Heterologous Distances \(A) The MDS plot provides a 2D distance map of the mean locations of the IGCs of all heterologous CTs established from 54 G0 nuclei (for further explanations see [Materials and Methods](#s4){ref-type="sec"}). The units (dimension 1 and dimension 2) are arbitrary. \(B) MDS plot for all heterologous PCs in 28 prometaphase rosettes. \(C) MDS plot for statistically placed CTs in 50 SCD model nuclei. \(D) MDS plot for 50 model nuclei with points randomly placed with a random number generator. ::: ![](pbio.0030157.g004) ::: The acrocentric chromosomes (13--15, 21, and 22) carry nucleolar organizer regions (NORs) on their short arms, and active NORs are associated with the nucleoli. Since nucleoli are generally located away from the nuclear envelope in the inner nuclear space, we expected that normalized 3D CN--CT distances for all acrocentric chromosomes should be significantly shorter on average than 3D CN--CT distances for the largest chromosomes. [Figure 5](#pbio-0030157-g005){ref-type="fig"} confirms this expectation in the sample of 54 3D evaluated nuclei, emphasizing the sensitivity of the IGC approach. We also found a highly significant difference (p \< 0.001 to p \< 0.0001; two-tailed Kolmogorov-Smirnov \[K-S\] test) between the mean radial CT positions of the five largest chromosomes (HSAs 1--5) and those of smaller chromosomes (HSAs 11--22 and Y). The radial positioning of the X territory did not differ significantly from the distribution of the five largest CTs. ::: {#pbio-0030157-g005 .fig} Figure 5 ::: {.caption} ###### Significance Levels for Pairwise Comparisons of 3D Distance Measurements Performed in 54 G0 Fibroblast Nuclei and in 50 SCD Model Nuclei Significance levels determined by the two-tailed K-S test are indicated by colors. Green, not significant, p \> 0.01; yellow, p \< 0.01; orange, p \< 0.001; red, p \< 0.0001. A minus (or plus) sign in a colored field indicates that the 3D CN--CT (A and B) or 3D CT--CT distance (C and D) indicated at the left (row) revealed a significantly shorter (or greater) mean radial distance than the CT indicated at the top (column). \(A) Comparison of 3D CN--CT distances in G0 fibroblast nuclei. \(B) Comparison of 3D CT--CN distances in fibroblast nuclei (vertical row) with 3D CN--CT distances in SCD model nuclei (horizontal row). \(C) Comparison of 3D CT--CT distances between homologous chromosomes in G0 fibroblast nuclei. \(D) Comparison of 3D CT--CT distances between homologous chromosomes in SCD model nuclei. ::: ![](pbio.0030157.g005) ::: Size-Correlated Radial Chromosome Positions in Human Fibroblast Prometaphase Rosettes {#s2c} ------------------------------------------------------------------------------------- Growing fibroblast cultures (46, XY) were subjected to 24-color 3D FISH. The shape of prometaphase rosettes identified in these cultures resembled an intermediate between a flat cylinder and a hemisphere. [Figure S5](#sg005){ref-type="supplementary-material"} shows a typical 3D-reconstructed rosette, as well as light-optical sections before and after deconvolution and after classification of PCs. In a sample of 28 rosettes with a total of 1,288 PCs, we determined the IGC of the prometaphase rosette (CR) and were able to classify 990 PCs (77%) by goldFISH and visual inspection. Consistent with the findings described above for CT arrangements in quiescent fibroblast nuclei, we observed a size-correlated radial PC distribution, i.e., IGCs of small PCs were typically clustered closer to the center of the rosette, while IGCs of large PCs were preferentially located towards the rim (see [Figures 2](#pbio-0030157-g002){ref-type="fig"}C, [3](#pbio-0030157-g003){ref-type="fig"}C, [4](#pbio-0030157-g004){ref-type="fig"}B, [S3](#sg003){ref-type="supplementary-material"}, [S5](#sg005){ref-type="supplementary-material"}, [S6](#sg006){ref-type="supplementary-material"}, and [S7](#sg007){ref-type="supplementary-material"} \[right panel\]). Voxel-Based Comparison of Radial CT Positions in Nuclei from Quiescent (G0) and Cycling Human Fibroblasts {#s2d} --------------------------------------------------------------------------------------------------------- For a direct comparison of the radial CT positioning of gene-poor and gene-rich chromosomes, we performed two-color 3D FISH experiments and painted CTs of HSAs 18/19 and HSA Y/17 in nuclei of quiescent and cycling fibroblasts ([Figures 6](#pbio-0030157-g006){ref-type="fig"} and S8). Semi-confluent cell cultures were fixed for this purpose at a density that yielded large fractions of both quiescent and cycling cells. Cultures were pulse-labeled with BrdU immediately before 3D fixation. Cells showing a BrdU labeling pattern typical of early S-phase \[[@pbio-0030157-b37]\] were chosen for 3D evaluation of cycling cells. Quiescent cells (G0) identified in the same culture lacked pKi67 and BrdU staining (data not shown). We recorded light-optical serial sections from 20 G0 nuclei and 20 S-phase nuclei. Since the use of IGCs for the location of individual CTs has obvious limitations, we used a voxel-based evaluation. After thresholding, each voxel attributed to a given CT was located in one of 25 concentric 3D nuclear shells. To define these shells, the radii in the x, y, and z directions were subdivided into 25 nuclear segments of equal thickness. The innermost shell comprised the smallest volume, and the outermost shell the largest volume. [Figure 6](#pbio-0030157-g006){ref-type="fig"}A and [6](#pbio-0030157-g006){ref-type="fig"}D show the results for HSA 18 CTs (red) and HSA 19 CTs (green). [Figure 6](#pbio-0030157-g006){ref-type="fig"}B and [6](#pbio-0030157-g006){ref-type="fig"}E show the respective results for the Y territory (red) and HSA 17 CTs (green). In each experiment the normalized CT-specific radial distribution curves were compared with the curve obtained for the TOPRO-3-stained total nuclear DNA (blue). The latter curve was taken as a rough approximation for a uniform radial DNA distribution (neglecting differences in the radial distribution of heterochromatin and euchromatin). This comparison revealed for both quiescent and cycling cells that Chromosome 17, 18, 19, and Y CTs, in spite of the pronounced differences in gene density, were all located significantly closer to the nuclear center than expected in the case of a uniform radial distribution (p \< 0.05; two-tailed K-S test). In both quiescent and cycling fibroblast nuclei the voxel distribution of the gene-poor HSA 18 CTs was slightly shifted towards the 3D nuclear border compared to the gene-rich HSA 19 CTs ([Figure 6](#pbio-0030157-g006){ref-type="fig"}A and [6](#pbio-0030157-g006){ref-type="fig"}D). This shift, however, was nonsignificant for both quiescent and cycling fibroblasts (p \> 0.05; Mann-Whitney U-test \[U-test\]). In contrast, the gene-poor Y territory was slightly more shifted towards the nuclear interior than the gene-rich HSA 17 CTs ([Figure 6](#pbio-0030157-g006){ref-type="fig"}B and [6](#pbio-0030157-g006){ref-type="fig"}E). This shift was significant for cycling fibroblasts (p \< 0.05; U-test), but not for quiescent fibroblasts. We also compared the radial voxel distributions of the CTs of HSAs 1 and 20 as an example of a large and a small chromosome having similar average gene densities ([Figure 6](#pbio-0030157-g006){ref-type="fig"}C and [6](#pbio-0030157-g006){ref-type="fig"}F). In agreement with the hypothesis of a size-correlated radial arrangement, CTs of HSA 20 were significantly shifted towards the nuclear interior compared to the DNA counterstain and to the CTs of HSA 1 (p \< 0.001; U-test). ::: {#pbio-0030157-g006 .fig} Figure 6 ::: {.caption} ###### Comparison of Radial CT Positions in Quiescent and Proliferating Cell Nuclei The abscissa shows the radial 3D distribution of CT-specific and whole-DNA-specific voxels in 25 concentric nuclear shells. The origin represents the center of the nucleus, and "100" represents the nuclear border. The ordinate indicates the mean frequency of the intensity-weighted voxels above threshold in each shell (in percent). Example images are shown in [Figure S8](#sg008){ref-type="supplementary-material"}. Bars correspond to standard errors of the mean. (A--F) A semi-confluent human diploid fibroblast (HDF) culture containing both quiescent cells (G0) and proliferating cells was used for two-color 3D FISH with different pairs of chromosome paint probes. In each experiment light-optical serial sections were recorded from 20 G0 nuclei (A--C) and 20 nuclei at early S-phase (D--F). All voxels attributed to a segmented CT were used to describe its radial position (red and green curves). Blue curves show the voxel-based radial distribution of whole DNA stained with TOPRO-3. Comparison of (A and D) gene-poor HSA 18 (red curve) and gene-rich HSA 19 (green curve), (B and E) gene-poor HSA Y (red curve) and gene-rich HSA 17 (green curve), and (C and F) large HSA 1 (red curve) and small HSA 20 (green curve). Irrespective of their gene content the small HSAs 17, 18, 19, 20, and Y were all located significantly closer towards the nuclear interior compared to the distribution of the whole nuclear DNA (p \< 0.05, two-tailed K-S test). The difference between the radial location of pair HSA 18/19 CTs both in quiescent and cycling cell nuclei, however, was not significant. For the pair HSA 17/Y, no significant difference was obtained between G0 and S-phase nuclei (p \> 0.05, Students t-test \[t-test\]). In contrast, HSA 1 was significantly shifted towards the nuclear border compared to HSA 20 (p \< 0.001, t-test), although HSAs 1 and 20 have a similar overall gene density. \(G) Comparison of HSA 18 (red curve) and HSA 19 (green curve) in nuclei of proliferating amniotic fluid cells during S-phase (n = 18). Although there is a slight excess of CT 19 voxels towards the nuclear interior compared to CT 18 voxels, the two curves are not significantly different (p \> 0.05; t-test). ::: ![](pbio.0030157.g006) ::: A novel approach (J. von Hase and C. Cremer, unpublished data) allowed us to measure the nearest 3D distance of each voxel attributed to HSA 18 CTs and HSA 19 CTs, respectively, from the segmented 3D nuclear periphery. We found that a fraction of the voxels attributed to HSA 18 CTs was located closer to the top and bottom part of the nuclear envelope than voxels attributed to HSA 19 CTs. The difference between the two voxel-based distance distributions, however, was only significant in cycling fibroblasts (p \< 0.05; K-S test). Comparison of Higher-Order Chromatin Arrangements in Nuclei of Human Fibroblasts, Amniotic Fluid Cells, and Lymphocytes {#s2e} ----------------------------------------------------------------------------------------------------------------------- All previously studied cell types that showed gene-density-correlated radial CT arrangements had a more spherical nuclear shape \[[@pbio-0030157-b26]\]. This prompted the question of whether other cell types, in addition to fibroblasts, with flat-ellipsoidal nuclei would also reveal a chromosome-size-related radial CT arrangement. To answer this question, we performed two-color 3D FISH experiments with HSA 18 and HSA 19 paint probes in cultured human amniotic fluid cells in the same way as described above ([Figure 6](#pbio-0030157-g006){ref-type="fig"}G). Such cells have a fibroblast-like phenotype with a flat-ellipsoidal nucleus, but they are typically derived from epithelial cells of the urinary tract \[[@pbio-0030157-b38]\]. Consistent with the fibroblast data described above, we observed that HSA 18 and 19 CTs were similarly distributed (p \> 0.05; U-test), but located significantly closer to the nuclear center than expected in the case of a uniform radial distribution (p \< 0.01; two-tailed K-S test). For a direct comparison of the 3D distribution of gene-rich and gene-poor chromatin in nuclei of human fibroblasts and lymphocytes from peripheral blood, we studied the 3D nuclear distribution of the repetitive DNA family of interspersed Alu sequences. The density of Alu sequences corresponds to GC-richness and density of housekeeping genes along human chromosomes \[[@pbio-0030157-b39],[@pbio-0030157-b40]\]. Accordingly, HSA 18 has a low and HSA 19 a high content of these sequences ([Figure 7](#pbio-0030157-g007){ref-type="fig"}A). Three-dimensional FISH experiments with a consensus Alu probe demonstrated for both cell types a concentration of Alu-rich chromatin in the nuclear interior, while Alu-poor chromatin formed a shell attached to the nuclear envelope ([Figure 7](#pbio-0030157-g007){ref-type="fig"}B--[7](#pbio-0030157-g007){ref-type="fig"}D). We conclude from this experiment that fibroblast and lymphocyte nuclei---in spite of profound differences in their shapes and radial CT arrangements---share this aspect of a gene-density-correlated, nonrandom, higher-order chromatin arrangement. Notably, expansions from the more interior Alu-rich chromatin into the Alu-poor peripheral shell were seen on closer inspection ([Figure 7](#pbio-0030157-g007){ref-type="fig"}E), indicating complex spatial interactions of chromatin regions with high and low gene density. Our findings support the hypothesis that HSA 18 CTs in fibroblast nuclei are more intensely connected to the top or bottom part of the nuclear envelope than HSA 19 CTs are. ::: {#pbio-0030157-g007 .fig} Figure 7 ::: {.caption} ###### Localization of Alu Sequences in Nuclei of Fibroblasts and Lymphocytes \(A) Karyotype from a female human lymphocyte (46, XX). Chromosomes were hybridized with a probe for Alu sequences (green) and counterstained with TOPRO-3 (red). Alu sequences were used as a marker for chromosomes and chromosome bands rich in genes. (B and C) Confocal serial sections were obtained from a human G0 fibroblast nucleus (B) and a G0 lymphocyte nucleus from peripheral blood (C) after 3D FISH with the Alu probe (green) and TOPRO-3 counterstaining (red). As examples, sections made at the top, middle, and bottom of the nuclei (separated by about 1 μm) are shown from left to right. Scale bars, 5 μm. \(D) Enlarged confocal mid-section through the human G0 fibroblast nucleus. Scale bar, 5 μm. \(E) Enlargement of the boxed sector in (D). The color image in the middle reflects the merged images left (TOPRO-3 counterstaining, red) and right (Alu staining, green). Arrows indicate chromatin rich in Alu sequences expanding into the TOPRO-3-stained, Alu-poor nuclear rim. Scale bar, 2 μm. ::: ![](pbio.0030157.g007) ::: Chromosome Proximity Patterns in Human Fibroblast Nuclei {#s2f} -------------------------------------------------------- Notwithstanding the gene-density-correlated features of higher-order chromatin arrangements described above, both IGC- and voxel-based evaluation procedures consistently demonstrated that radial CT arrangements in nuclei of quiescent and cycling fibroblasts correlate with chromosome size. In an attempt to quantify patterns of proximity of homologous and heterologous chromosomes, we measured 3D CT--CT and 3D PC--PC distances, as well as 3D CT--CN--CT and 3D PC--CR--PC angles within the nuclear space. To assess the possibilities and limitations of such measurements, let us imagine that we could perform "orbital" walks around the nucleus in a series of concentric 3D shells (like planets orbiting around the sun). A priori the evidence for a nonrandom radial CT order is compatible with both random and nonrandom "orbital" arrangements of CTs. Combined 3D CT--CT and 3D CT--CN--CT (or 3D PC--PC and 3D PC--CR--PC) measurements provide an opportunity to discriminate between the two possibilities. If size-correlated radial arrangements were the only principle of an ordered CT or PC positioning in human fibroblasts, we would expect a broad range of angles from very small ones to 180°, independent of the preferred radial position for any pair of CTs/PCs. In contrast, consistently small or large angles for a given chromosome pair would strongly support the notion of nonrandom associations or separations of the two chromosomes. ### Proximity patterns of homologous chromosomes {#s2f1} A broad range of normalized 3D CT--CT and 3D PC--PC distances was observed between all homologous chromosome pairs in the 54 G0 fibroblast nuclei and 28 prometaphase rosettes (data not shown). Consistent with this variability 3D CT--CN--CT angles ([Figure S9](#sg009){ref-type="supplementary-material"}) and 3D PC--CR--PC angles ([Figure S10](#sg010){ref-type="supplementary-material"}) ranged from very small to large angles near 180°. [Figure 8](#pbio-0030157-g008){ref-type="fig"} illustrates the broad angle distributions obtained for homologous chromosome pairs 7, 15, and 22 in both G0 nuclei and prometaphase rosettes. [Figure 5](#pbio-0030157-g005){ref-type="fig"}C shows the significance levels obtained in G0 fibroblast nuclei for all possible comparisons of normalized distances between two different pairs of homologous CTs. From a total of 276 possible comparisons, 19 yielded a significant difference; 12 of these suggested that the two CTs of HSA 5 were located significantly more distant from each other than the homologous CTs of HSAs 10, 11, 13, and 15--22, as well as the CTs of the two gonosomes, X and Y. Further studies will tell us whether the preferential involvement of HSA 5 is a true finding or a statistical artifact occasionally expected when large numbers of comparisons are statistically tested. In conclusion, these data indicate highly variable proximity patterns for homologous chromosomes in human fibroblast nuclei and prometaphase rosettes. However, we cannot firmly exclude at present the possibility that some homologous chromosome pairs occupy nonrandom "orbital" positions with regard to each other. ::: {#pbio-0030157-g008 .fig} Figure 8 ::: {.caption} ###### Relative Spatial Distributions of Homologous Chromosomes of HSAs 7, 15, and 22 Schematic outlines of fibroblast nuclei (as ellipses, A) and prometaphase rosettes (as circles, B) with normalized size and shape. The IGC of one randomly selected homolog was placed along the negative long axis. The IGC of the other homolog was marked at the corresponding nuclear position. Gray dots represent data obtained by confocal microscopy, open circles by wide-field microscopy. Angle measurements for all pairs of homologous chromosomes are presented in [Figure S9](#sg009){ref-type="supplementary-material"} for G0 nuclei and in [Figure S10](#sg010){ref-type="supplementary-material"} for rosettes. ::: ![](pbio.0030157.g008) ::: ### Proximity patterns of heterologous chromosomes {#s2f2} We performed 3D distance measurements between all possible combinations (n = 42,988) of heterologous CTs/PCs. These measurements suggested a certain degree of nonrandom proximity: 7,651 (18%) pairwise 3D CT--CT distance comparisons in G0 nuclei and 3,657 pairwise 3D PC--PC distance comparisons (8.5%) in prometaphase rosettes were statistically significant (p \< 0.01; two-tailed K-S test). The observed percentages of nonrandom proximity may be explained mostly as a consequence of the nonrandom radial CT/PC arrangements described above, but may also to some degree reflect nonrandom orbital arrangements of heterologous CTs/PCs. Consistent with the finding that 3D CN--CT distances for acrocentric chromosomes were significantly shorter than 3D CN--CT distances for the largest chromosomes (HSA 1--5; see above), we also found a significantly smaller mean 3D CT--CT distance for acrocentric chromosomes than for the five largest chromosomes ([Table 1](#pbio-0030157-t001){ref-type="table"}). The mean 3D CT--CT distance for the five largest chromosomes was also significantly larger than the mean distance for the other small metacentric/submetacentric autosomes (HSAs 16--20) that do not carry NORs. In contrast, the mean 3D CT--CT distances for acrocentric chromosomes were not significantly different from the mean distances for these other small chromosomes. ::: {#pbio-0030157-t001 .table-wrap} Table 1 ::: {.caption} ###### Comparison of Mean 3D CT--CT Distances and Mean 3D CT--CN--CT Angles between Groups of Large (HSAs 1--5) and Small (HSAs 13--22) Chromosomes ::: ![](pbio.0030157.t001) Chromosome group nomenclature A--F corresponds to ISCN \[[@pbio-0030157-b63]\]. Group A, HSAs 1--3; group B, HSAs 4 and 5; group D, HSAs 13--15; group E, HSAs 16--18; group F, HSAs 19 and 20; group G, HSAs 21, 22, and Y. Note that the Y chromosome was excluded from this comparison. Significance levels were tested with a t-test using the mean CT--CT and angle values, respectively, for the combined groups. For a comparison of 3D CT--CN--CT angles between individual pairs of large and small heterologous chromosomes, see Figure S11 ::: The mean 3D CT--CN--CT angle obtained for the five largest chromosomes was not significantly different from the mean angle for the small metacentric and submetacentric chromosomes, while a modestly significant difference (p \< 0.05) was noted in mean angle between the latter and the acrocentric chromosomes ([Table 1](#pbio-0030157-t001){ref-type="table"}). [Table 2](#pbio-0030157-t002){ref-type="table"} presents the mean angles with standard deviations and the angle range measured for all possible pairwise combinations of the largest chromosomes (HSAs 1--5), the acrocentric chromosomes (HSAs 13--15, 21 and 22), and the other small chromosomes (HSAs 16--20). For all comparisons we noted a large angle range from a few degrees to more than 170 degrees. [Figure S11](#sg011){ref-type="supplementary-material"} presents a comparison of CT--CN--CT angle distributions between different pairs of heterologous CTs. With few exceptions these comparisons did not reveal significant differences. ::: {#pbio-0030157-t002 .table-wrap} Table 2 ::: {.caption} ###### Angular Separation of the Largest and Smallest Heterologous CTs in G0 Nuclei ::: ![](pbio.0030157.t002) ns, not significant ::: While the findings described above support the possibility of nonrandom proximity between chromosomes in fibroblast nuclei, the observed cell-to-cell variability emphasizes the probabilistic nature of these patterns. This large variability of proximity is further emphasized in [Figure S12](#sg012){ref-type="supplementary-material"} for CTs of HSAs 7 and 8. It shows projections of confocal image stacks from a series of 50 randomly chosen G0 fibroblast nuclei after two-color 3D FISH with the respective paint probes. Quantitative evaluation of this experiment ([Figure S13](#sg013){ref-type="supplementary-material"}) confirmed that the angular distribution for these CTs did not significantly differ from that calculated for points randomly placed in an ellipsoid. In conclusion, our results do not support the existence of a highly ordered orbital arrangement of the paternal and maternal chromosome sets. Comparison of CT Positions in Fibroblast Nuclei (G0) with Modeled CT Positions {#s2g} ------------------------------------------------------------------------------ As a first and very simple attempt to create a model for the random distribution of the IGCs from all CTs present in a diploid human fibroblast nucleus, we placed 46 points randomly in ellipsoids with similar shapes, and performed distance and angle measurements between the points and the centers of the ellipsoids (random point distribution model). Twenty-seven out of 30 comparisons of heterologous CT--CN--CT angle distributions measured in fibroblast nuclei did not reveal a significant difference from the random point distribution model ([Table 2](#pbio-0030157-t002){ref-type="table"}). The three exceptions were the 3D CT--CN--CT angles between CTs of HSAs 1 and 2 (p \< 0.05), CTs of HSAs 2 and 4 (p \< 0.01), and CTs of HSAs 13 and 15 (p \< 0.01). In prometaphase rosettes (see [Figure S10](#sg010){ref-type="supplementary-material"}), none of the angle distributions for homologous chromosome pairs differed significantly from the distributions obtained for randomly distributed points. The random point distribution model does not take into account geometrical constraints, which likely affect the distribution of small and large CTs within the limited nuclear space. Thus, a significant difference between model and experimental results could simply result from geometrical constraints imposed on the distribution of CTs with various sizes. To determine the possible influence of geometrical constraints, we used the spherical 1-Mbp chromatin domain (SCD) model \[[@pbio-0030157-b41]\], which assumes that CTs are built up from 1-Mbp domains, and that the relative fraction of the nuclear volume occupied by each CT is directly proportional to the number of these domains that constitute a chromosome. Chromosomes of male diploid fibroblasts (46, XY) in telophase/early G1 were modeled as rods with lengths reflecting their relative DNA content, since estimates of CT volumes supported a rough positive correlation with DNA content (data not shown). These rods were statistically placed with random orientations into an ellipsoid with the x, y, and z diameters corresponding to the average main diameters of the fibroblast nuclei. Using Monte Carlo calculations, we allowed the model chromosomes in these starting configurations to "decondense" to CTs in early G1 nuclei (see [Figure 1](#pbio-0030157-g001){ref-type="fig"}D; [Video S1](#sv001){ref-type="supplementary-material"}). Since any modifications of the starting arrangements of rods in a model nucleus may influence the final CT distribution profoundly, this approach should be considered only as an attempt to model the arrangements of all CTs in fibroblast model nuclei with ellipsoidal shape. Note that it was not our goal here to model the real anaphase/telophase chromosome distribution or the chromatin movements involved in the formation of G1 nuclei. Instead, we wished to study whether geometrical constraints acting during the process of decondensation strongly influence the final positions of decondensed CTs compared to the positions of statistically placed condensed chromosomes. In a set of 50 SCD model nuclei with fully decondensed model CTs, we determined the 3D centers of gravity for the 46 simulated CTs, and performed distance and angle measurements. In contrast to fibroblast nuclei and prometaphase rosettes (see [Figure 3](#pbio-0030157-g003){ref-type="fig"}A and [3](#pbio-0030157-g003){ref-type="fig"}C), 3D radial distances measured in these 50 SCD model nuclei revealed an inverse correlation of radial positioning and chromosome size (see [Figure 3](#pbio-0030157-g003){ref-type="fig"}B). Large modeled CTs were preferentially located in the interior and small CTs towards the rim of SCD model nuclei. [Figure 5](#pbio-0030157-g005){ref-type="fig"}B shows the significance levels obtained for the pairwise comparison of 3D CN--CT distances in SCD model nuclei with the corresponding 3D CT--CN distances in fibroblast nuclei. Consistent with this inverse correlation the mean 3D CN--CT (radial) distances measured for the largest chromosomes in real fibroblast nuclei were significantly greater than the mean radial distances for their modeled counterparts in SCD model nuclei. For the smallest chromosomes, we found numerous instances where the mean radial distances measured in fibroblast nuclei were significantly smaller than the SCD model distances. These results suggest that the size-correlated radial CT arrangements observed in fibroblast nuclei cannot be explained simply as a consequence of geometrical constraints affecting the radial positions of small and large CTs. From a total of 253 possible pairwise comparisons of 3D CT--CT distances between homologous CT pairs in SCD model nuclei, only three (1.2%) were significant (p \< 0.01; see [Figure 5](#pbio-0030157-g005){ref-type="fig"}D). The significantly greater 3D distance between modeled CT 21 pairs compared to the 3D distances between modeled CT 1 and CT 2 homologs is in line with the more peripheral location of small versus large modeled CTs. Pairwise comparisons between heterologous 3D CT--CT distances in SCD model nuclei yielded 837 (2%) significant (p \< 0.01) differences out of 42,988 pairwise distance comparisons. These small percentages emphasize our effort to model a random neighborhood of CTs except for the influence of different geometrical constraints affecting the positions of large versus small CTs. Pairwise comparisons of 3D distances measured in 50 random point distribution model nuclei yielded only 45 out of 42,988 significant differences at a significance level p \< 0.01, a fraction expected by chance. This comparison emphasizes that geometrical constraints affected the distribution of CTs in SCD model nuclei. The comparison of angle distributions obtained for the 22 homologous autosome pairs and the XY pair in the 50 SCD model nuclei with the respective angle distributions described above for G0 fibroblast nuclei showed no significant differences except for Chromosomes 4 (p \< 0.01), 7 (p \< 0.05), 10 (p \< 0.05), 20 (p \< 0.01), and 22 (p \< 0.05) (see [Figure S9](#sg009){ref-type="supplementary-material"}). In G0 fibroblast nuclei the number of significant differences (p \< 0.01) was considerably larger for heterologous distance pairs (18%; 7,651/42,988) than for homologous distance pairs (7.5%; 19/253). In conclusion, the considerably higher percentage of nonrandom distance comparisons in G0 fibroblast nuclei (see above) than in SCD model nuclei supports the hypothesis that a certain degree of nonrandom CT proximity exists that is not expected if proximity were simply the result of geometrical constraints imposed on crowded small and large objects (CTs) within a limited volume with special shape (the nucleus). Other mechanisms must be invoked to explain the observed chromosome distributions. Discussion {#s3} ========== Probabilistic Chromosome Order in Human Fibroblasts {#s3a} --------------------------------------------------- Mapping the positions of all chromosomes simultaneously in a single nucleus provides a major step forward towards the goal of elucidating general and cell-type-specific rules that govern chromosome positioning in cycling and postmitotic cells. Reviewing the evidence for an "inherently imperfect and probabilistic nature" of CT positioning, Parada et al. \[[@pbio-0030157-b42]\] came to the following conclusion: "Chromosome positioning patterns are statistical representations of chromosome positions but do not provide information about the precise coordinates of a given chromosome in a given nucleus. It is important to realize that, although significant non-random chromosome positions can be described, they contain a significant degree of uncertainty and merely indicate a preferred, probabilistic position of a given chromosome in the cell nucleus." This view is corroborated by the present study and cannot be reconciled with previous reports suggesting a precise pattern of chromosomal neighborhood in human fibroblasts \[[@pbio-0030157-b18],[@pbio-0030157-b19]\]. Conclusive evidence for preferred, probabilistic rather than absolute CT positioning patterns does not exclude the possibility that higher-order chromatin architecture at large follows determininistic rules to some extent. The observation that a shell of gene-poor chromatin domains is located beneath the nuclear envelope of all cell types studied so far indicates such a deterministic feature. NOR-bearing CTs placed around nucleoli provide a telling example of how probabilistic and deterministic rules may act together to bring about this well known nuclear topography. The probabilistic combination of NORs from several acrocentric chromosomes gives rise to the formation of the several nucleoli typically found in human fibroblast nuclei. This combination is subject to changes from cell cycle to cell cycle. Consistent with this conclusion we found a large variability of 3D CT--CT distances and CT--CN--CT angles for pairs of homologous and heterologous acrocentric chromosomes. Notwithstanding this variability, distances between CTs of acrocentric chromosomes were significantly smaller than between CTs of the five largest chromosomes, reflecting frequent spatial association of acrocentric chromosomes with the more internally located nucleoli as compared to the high probability of large chromosomes to be located close to the nuclear rim. A probabilistic order of acrocentric CTs as a general feature of human cells is consistent with a quantitative analysis of interchanges between all possible heterologous pairs of chromosomes in human lymphoblast nuclei after damage by sparsely ionizing radiation in vitro \[[@pbio-0030157-b13]\]. A deterministic feature of this order, however, is introduced by the fact that the formation of a nucleolus at a given nuclear site requires the presence of at least one NOR-bearing chromosome. Recently, Misteli and coworkers \[[@pbio-0030157-b17],[@pbio-0030157-b43]\] determined the nearest neighbors of CTs in cell types of different tissues and found statistically significant cell-type-specific proximity patterns for CT clusters, as well as a correlation between tissue-specific spatial proximity patterns and tissue-specific translocations during tumor development \[[@pbio-0030157-b06]\]. Considering significant differences between the probabilistic arrangements of CTs/PCs observed in fibroblast nuclei and prometaphase rosettes and the arrangements predicted by random point and SCD model distributions, we must take into account the limited statistical power of the sample sizes evaluated in the present experiments. Repeated experiments with larger sample sizes are necessary to distinguish biologically significant probabilistic proximity patterns unequivocally from potential statistical artifacts. Since we were not able to distinguish paternal from maternal homologs, the question of a separation of the parental genomes could not be settled by the present experiments. Although the wide range of angles which we measured for each homologous or heterologous chromosome pair does not necessarily exclude the possibility of a spatial separation of the parental genomes, this range clearly argues against a consistent separation of homologous PCs at opposite sites of the rosette \[[@pbio-0030157-b18],[@pbio-0030157-b19]\]. Features of CT Arrangements Correlated with Chromosome Size and Gene Density {#s3b} ---------------------------------------------------------------------------- In agreement with the study of Sun et al. \[[@pbio-0030157-b28]\], we found that small chromosomes (independent of gene density) were preferentially located towards the center of fibroblast nuclei, while large chromosomes were preferentially positioned towards the nuclear rim. Evidence for a chromosome-size-correlated radial CT arrangement was also found in nuclei of quiescent and cycling human amniotic fluid cell nuclei (\[[@pbio-0030157-b25],[@pbio-0030157-b44],[@pbio-0030157-b45]\]; present study), as well as in nuclei of cultured orangutan fibroblasts (M. Neusser, T. Cremer, and S. Müller, unpublished data). The finding of chromosome-size-correlated radial CT arrangements in two cell types with flat-ellipsoidal nuclei contrasts with the gene-density-correlated radial CT arrangements consistently reported for human and other primate cell nuclei with approximately spherical nuclear shapes. There is an apparent discrepancy of our data with reports of the Bickmore group, who described a gene-density-correlated CT arrangement in nuclei of cycling human fibroblasts \[[@pbio-0030157-b24]\]. This group also reported a central shift of HSA 18 CT positions when fibroblasts became quiescent \[[@pbio-0030157-b27]\]. In agreement with this finding we also observed HSA 18 CTs somewhat closer to the nuclear periphery in cycling than in quiescent fibroblasts (see [Figure 6](#pbio-0030157-g006){ref-type="fig"}C). It is not clear at present why the observed differences were much more pronounced in the analysis performed by Bridger et al. \[[@pbio-0030157-b27]\]. We noted one obvious difference between Bridger et al.\'s study and our study: while Bridger et al. studied quiescent fibroblasts in serum-starved cultures, we compared quiescent and cycling cells that coexisted in the same semi-confluent culture supplemented with 10% fetal calf serum (FCS). Serum-starved fibroblasts exhibit extremely flat nuclei compared to cells grown in the presence of FCS (unpublished data). It seems therefore possible that differences in nuclear shape and/or other unknown factors affected the distribution of CTs detected in G0 nuclei from serum-starved fibroblasts compared to G0 nuclei from cells grown in FCS-supplemented medium. A more complex picture has emerged from our study, namely that both size- and gene-density-correlated features of higher-order chromatin arrangements coexist in human fibroblast nuclei. Chromosome-size-correlated radial CT arrangements are most obvious in 2D nuclear projections (see [Figures 2](#pbio-0030157-g002){ref-type="fig"} and S2). In contrast, measurements of HSA 18 and HSA 19 CT positions along the optical axis suggest a location of the gene-poor HSA 18 closer to the top or bottom part of the nuclear envelope than for HSA 19. This hypothesis is supported by our observation that a layer of Alu- and gene-poor chromatin coats the interior side of the nuclear envelope, whereas chromatin rich in Alu sequences and genes was preferentially found in the interior of the nucleus. Considering the fact that the resolution along the optical axis is lower than the lateral resolution, these findings do, of course, not exclude direct contacts of HSA 19 chromatin with the top or bottom part of the nuclear envelope. A detailed comparison of similarities and differences of higher-order chromatin arrangements in nuclei from different cell types is made difficult by the possibility that additional levels of complexity may exist, which were not considered so far. Note, for example, that invaginations of the nuclear membrane may yield a direct contact of chromatin domains or nucleoli apparently detected in the nuclear interior with the nuclear envelope. Interdependence between Nuclear Shape and CT Arrangements {#s3c} --------------------------------------------------------- The presently available data indicate that CT arrangements differ between cell types with flat-ellipsoidal nuclei, such as human fibroblasts and amniotic fluid cells, and cell types containing nuclei with a more roundish shape \[[@pbio-0030157-b46]\]. Whether these cell-type-specific differences are generally correlated with or even caused by differences in nuclear shape remains an open question. Does a nuclear shape transformation suffice to change size-correlated CT arrangements observed in flat-ellipsoidal nuclei into gene-density-correlated CT arrangements observed in roundish nuclei, and vice versa? For example, let us consider a population of flat-ellipsoidal nuclei, where the CTs of both HSA 18 and HSA 19 stay close to the nuclear center. Let us further assume that HSA 18 CTs are strongly attached to the lamina, while HSA 19 CTs are not. Accordingly, both of the HSA 18 CTs but not the HSA 19 CTs should follow the nuclear envelope during shape transformation. At the end of the transformation from a flat-ellipsoidal to a spherical nucleus, we would expect the two HSA 18 CTs to now be located in the nuclear periphery far away from the 3D nuclear center. Both HSA 18 CTs may be still close together or located at opposite sites of the now spherical nucleus. Accordingly, we would expect strongly bimodal angle distributions for the HSA 18 CTs, with one peak at small angles and a second peak at large angles. When we recorded the angular separation of the two peripherally located HSA 18 CTs in a set of 82 lymphocyte nuclei from peripheral human blood, we found indeed two maxima, one around 10° and a second at 140° (J. Hase and M. Cremer, unpublished data). Let us now consider the opposite transformation of a spherical precursor nucleus, with both HSA 18 CTs located peripherally either close to each other or on opposite sides, to a flat-ellipsoidal nucleus. There are an infinite number of possible mid-planes through the center of a spherical nucleus. In order to achieve the desired location of both CTs close to the 3D center, it is essential to choose the correct mid-plane for this transformation. Consider a population of spherical nuclei with two HSA 18 CTs located at opposite sides. We need to choose mid-planes roughly perpendicular to the connection between the IGCs of the two CTs. Selection of a mid-plane along this connection would yield a transformation with both CTs located at the nuclear rim. The considerations above emphasize that nuclear shape transformations may affect CT arrangements \[[@pbio-0030157-b47]\], and inspire a number of questions for future experiments. Do shape changes enforce changes of CT arrangements or vice versa? And are changes of nuclear shape causally connected with changes of gene expression patterns? Mechanisms Responsible for the Establishment of Cell-Type-Specific CT Arrangements {#s3d} ---------------------------------------------------------------------------------- We expect that complex genetic and epigenetic mechanisms at various hierarchical levels act in concert in order to establish, maintain, or alter higher-order chromatin arrangements as required for proper nuclear functions. A search for such mechanisms must take into account the fact that chromosome arrangements---at least in the cell types studied so far---follow probabilistic and not deterministic rules \[[@pbio-0030157-b42]\]. Candidate mechanisms include the genetically regulated timing of centromere separation and anaphase movements of individual chromosomes \[[@pbio-0030157-b48]\], interactions of constitutive and facultative heterochromatic domains with each other and the nuclear lamina, and interactions of chromatin domains comprising active genes with nuclear bodies located in the interchromatin compartment \[[@pbio-0030157-b04]\]. The mechanisms producing a size-correlated radial positioning in fibroblast nuclei during telophase and early G1 may differ from the mechanisms acting to establish this arrangement in mitotic rosettes. The central part of a mitotic spindle in vertebrates consists of a tight bundle of microtubules stretched between the two centrioles. The exclusion of chromosomes from this tight bundle leads to the formation of a centromere ring around this bundle. Accordingly, the localization of the IGCs of smaller chromosomes towards the center of the rosette may simply reflect their small size. Geometrical constraints may force large chromosomes to expand their arms to the periphery of the rosette, and accordingly, their IGCs are shifted towards the periphery. To some extent the spindle may reinforce a size-correlated radial arrangement of chromosomes (see \[[@pbio-0030157-b12]\] and references therein). Persistence of this order from the prometaphase rosette to the anaphase rosette could yield a starting configuration for the decondensation of chromosomes in telophase and early G1 consistent with size-correlated radial CT arrangements in the flat-ellipsoidal fibroblast nuclei. If cycling lymphoblasts were to show the same size-correlated radial chromosome arrangement in mitotic rosettes---a possibility not yet experimentally tested---the gene-density-correlated radial arrangements during interphase would require major chromosome movements during telophase/early G1. Evolutionary Aspects {#s3e} -------------------- Systematic comparisons of nuclear architecture in cell types from evolutionarily distant species must be performed to distinguish evolutionarily conserved features from species-specific ones. One feature of nuclear architecture has already been detected that has been strongly conserved during evolution. In animal species, ranging from Hydra to mammals \[[@pbio-0030157-b37],[@pbio-0030157-b49],[@pbio-0030157-b50]\], and plant species \[[@pbio-0030157-b51]\], DNA replication is linked to a typical nuclear topology of early-, mid-, and late-replicating chromatin. The evolutionary conservation of a layer of constitutive and facultative heterochromatin at the nuclear periphery argues for a selective advantage. In 1975, T. C. Hsu proposed the "bodyguard" hypothesis for the function of constitutive heterochromatin \[[@pbio-0030157-b52]\]. Hsu argued that heterochromatin localized at the nuclear periphery might protect the centrally localized euchromatin against mutagens, clastogens, and viruses entering the nucleus. If this hypothesis is correct, we should observe DNA damage preferentially in the peripheral chromatin shield. More recently, a protective, "buffering" role for redundant DNA was suggested \[[@pbio-0030157-b53]\]. However, experimental evidence to support these hypotheses has not been provided so far. Finally, we may consider the possibility that a layer of dense facultative and constitutive heterochromatin intimately connected to the nuclear lamina may provide more stability to the nucleus, just as certain proteins beneath the cell membrane provide stability against mechanical stress. This protection may be particularly important in cell types, such as muscle cells, that are the frequent subject of mechanical stress, affecting the shape of both the cell and its nucleus \[[@pbio-0030157-b54],[@pbio-0030157-b55]\]. Materials and Methods {#s4} ===================== {#s4a} ### Cells, fixation procedure, and 3D FISH pretreatment {#s4a1} A vigorously growing primary human fibroblast culture was established from a skin biopsy of a 2-y-old boy. Chromosome banding and M-FISH analyses performed after the second passage (1:2 split) showed a normal male karyotype (46, XY). Surplus cultures were kindly provided by the Abteilung Medizinische Genetik, Munich University, Germany; cells were further grown in our laboratory in DMEM medium supplemented with 10% FCS, and aliquots were frozen at about 5--7 passages in liquid nitrogen. Cells from these aliquots were further propagated for 2--4 wk, subcultivated (1:2) every 5 d, and routinely checked for absence of mycoplasm contaminations \[[@pbio-0030157-b56]\]. For 3D FISH and ReFISH experiments, cells were seeded on coverslips (26 × 76 mm, thickness 0.17 ± 0.01 mm). For studies of quiescent cell populations, cultures were grown to confluence and maintained for 1 wk before 3D fixation was performed in 4% paraformaldehyde/1× PBS for 10 min \[[@pbio-0030157-b30]\]. Control experiments with BrdU pulse labeling (1 h) and immunostaining of the cell-cycle-specific nuclear protein Ki67 indicated that more than 99.5% of the cells were in a quiescent state (G0) under these conditions. To investigate G0 and S-phase cells simultaneously on the same coverslip, cells were fixed at approximately 40%--50% confluence. Nuclei from cells in and out the cell cycle were discriminated by BrdU pulse labeling (45 min) and pKi67 staining. Growing fibroblast cultures were also used to obtain prometaphase rosettes. Nuclei in S-phase showed pKi67 staining and incorporation of BrdU, while G0 nuclei lacked both signals (data not shown). An amniotic fluid cell culture from a female fetus (46, XX) was established following diagnostic amniocentesis, and a growing early-passage culture was pulse-labeled with BrdU (45 min) prior to 3D fixation. Permeabilization steps performed prior to 3D FISH included treatments with 0.5% Triton-X100 (20 min), 20% glycerol in PBS (30 min), repeated freeze/thawing in liquid nitrogen, and incubation in 0.1 M HCl (5 min) or pepsin (0.002% in 0.01 M HCl). Slides were stored at 4 °C in 50% formamide/2× SSC until 3D FISH was performed. ### DNA probes, labeling protocols, 3D FISH, and probe detection {#s4a2} Whole chromosome painting probes were kindly provided by Malcolm Ferguson-Smith (Cambridge University, United Kingdom). Probes were established from flow-sorted human chromosomes and amplified by DOP-PCR. For 24-color 3D FISH experiments in a single assay, chromosome paint probes for all 24 chromosome types (HSAs 1--22, X, and Y) were labeled using a combinatorial labeling scheme with seven differentially labeled nucleotides including diethylaminocoumarine (DEAC; NEN Life Science Products, Zaventem, Belgium), Fluorogreen (Amersham Pharmacia Biotech, Piscataway, New Jersey, United States), TexasRed (Molecular Probes, Eugene, Oregon, United States), Cy3, Cy5 (Amersham Pharmacia Biotech), biotin-dUTP (Rockland Immunochemicals, Gilbertsville, Pennsylvania, United States) and digoxigenin-dUTP (Roche, Basel, Switzerland) \[[@pbio-0030157-b31]\]. A mix containing the 24 labeled probes in 50% formamide/10% dextran sulfate/1× SSC was hybridized for 3 d at 37 °C as previously described \[[@pbio-0030157-b30]\]. Post-hybridization washes were performed three times with 0.5× SSC at 60 °C. Avidin-Cy5.5 was used for detection of biotin and anti-dig-Cy7 (custom-made) for the immunodetection of digoxigenin. For the differential coloring of all 24 chromosome types in ReFISH experiments \[[@pbio-0030157-b30],[@pbio-0030157-b32]\], chromosome-specific paint probes from all 24 human chromosome types were labeled with either biotin-dUTP, TAMRA-dUTP, or Spectrum Green--dUTP. Two different hybridization mixtures were prepared in a way that allowed the unequivocal discrimination of each chromosome type in two subsequent hybridization experiments. After hybridization of the first probe subset, biotinylated probes were detected by Avidin-Cy5.5. Confocal image stacks were acquired as described below, and the cell coordinates were recorded. Thereafter, the same cells were re-hybridized with the second probe subset, followed again by detection with Avidin-Cy5.5 and confocal microscopy. In two-color 3D FISH experiments aimed at the simultaneous visualization of two pairs of homologous CTs, we used biotin- and digoxigenin-labeled chromosome paint probes for HSAs 18/19, HSAs 17/Y, and HSAs 1/20. ### Microscopy {#s4a3} After 3D FISH of all 24 chromosome types in a single assay, nuclei were imaged with an epifluorescence wide-field microscope (DMRXA, Leica, Wetzlar, Germany), equipped with a Plan Apo 63×/1.4 oil immersion objective, an 8-filter wheel with narrow-band filters (Chroma Group, San Bruno, California, United States), and an automated z-step motor \[[@pbio-0030157-b57]\]. For image capturing, a Sensys CCD camera (PhotoMetrics, Huntington Beach, California, United States) was used. Both camera and microscope were controlled by Leica QFluoro software. Stacks of optical sections with an axial distance of 250 nm were collected from nuclei with a regular shape showing apparently complete and specific hybridization signals in all channels. For each optical section images were collected sequentially for all fluorochromes. Stacks of 8-bit gray-scale 2D images were obtained with a pixel size of 110 nm in the x and y directions, and an image size of 256 × 256 pixels. In ReFISH experiments, images for the three fluorochromes were obtained from the same nuclei after the first and second hybridization with a Zeiss (Oberkochen, Germany) LSM 410 confocal microscope equipped with a 63× Plan Apo objective and filters for FITC, Cy3, and Cy5. Scans were sequentially performed for the three fluorochromes on each light-optical section. An alignment of the two image stacks obtained for each nucleus after the first and second hybridization was performed on a Silicon Graphics (Mountain View, California, United States) workstation (OS Irix 6.2) using the program Correlator \[[@pbio-0030157-b58]\] in the integrated development environment Khoros (Khoral, Albuquerque, New Mexico, United States). This procedure allowed the fitting of the two image stacks with subvoxel accuracy. A comparison of the nuclear shape after the first and second hybridization did not reveal a notable difference, but we observed a slight increase in volume, which was corrected by the computer algorithm. ### Deconvolution and image processing {#s4a4} Each fluorochrome channel was normalized to a maximum intensity value of 255, and subjected to deconvolution by the software Huygens (Scientific Volume Imaging, Hilversum, The Netherlands). Thereafter, chromosomes in the image stacks were classified according to their labeling scheme using the software goldFISH \[[@pbio-0030157-b33]\] running on a Silicon Graphics workstation. This software carried out an automated classification of fluorescently stained areas in each light-optical nuclear section on the basis of the combinatorial labeling scheme. A false color representing the classified chromosome type was allocated to each classified territory. The software calculated the 3D IGCs of each classified territory as well as the CN or the CR by means of the DAPI image stack. Maximum intensity projections of image stacks were made with ImageJ software (National Institutes of Health, Bethesda, Maryland, United States). Displayed overlays were processed with Adobe Photoshop (Adobe Systems, San Jose, California, United States). Three-dimensional reconstructions of image stacks were performed using Amira 2.3 (Mercury Computer Systems, Chelmsford, Massachusetts, United States). Overlap of CTs in nuclei or of chromosomes in 3D fixed mitotic rosettes can be a source of misclassification. Because of the remarkable flatness of fibroblast nuclei in G0 (maximum height approximately 6 μm, with CTs often expanded from the bottom of the nucleus to the top), errors due to CT overlaps are less likely than in spherical nuclei. ### Data evaluation {#s4a5} For each classified PC and CT we determined its 3D IGC, together with the CR and the CN, respectively. IGC coordinates were imported into Excel (Microsoft, Redmond, Washington, United States), and the following distances and angles were measured: (1) 3D CR--PC and 3D CN--CT distances (3D radial distances), (2) 3D PC--PC and 3D CT--CT distances between all possible pairs of homologous and heterologous PCs or CTs, and (3) 3D PC--CR--PC and 3D CT--CN--CT angles. For comparison of different nuclei, 3D distances were normalized using the following procedure. A coordinate system was applied to each individual nucleus with the nuclear center as the origin (polar coordinates). The angle α between the longer cell axis and the x-axis of the coordinate system was determined, and the coordinates were recalculated following a rotation: x′ = cosα x − sinα y and y′ = sinα x − cosα y. For size measurement of nuclei, the DNA was stained with DAPI or TOPRO-3. The diameters in X and Y were then measured using the light-optical section with maximum lateral nuclear expansion. The height was measured between the upper and lower plane showing the most peripheral DAPI staining along the z-axis. The relative radial distance r of a PC or CT was calculated as r = (r ~1~/r ~0~)·100, where r ~1~ represents the distance CN--CT or CR--PC, respectively, and r ~0~ denotes either the distance between the CR and the prometaphase edge or between the CN and the nuclear edge, drawing a line through the IGC of the analyzed chromosome. Angles were calculated between the IGCs of homologous CTs or PCs using the CN or CR as the midpoint. The relative radial distance between heterologous PCs or CTs was calculated as a fraction of the nuclear diameter. Distance and angle measurements obtained for IGCs of CTs and PCs were compared with distances and angles calculated between points statistically placed by a random number generator ("random point distribution model"). MDS plots were generated by SPSS 11 (SPSS, Chicago, Illinois, United States). This program provided a 2D distance map taking into account the normalized mean heterologous 3D CT--CT distances calculated for all possible combinations of heterologous CTs. All distances were normalized to the diameter of the nucleus before generating the plots. The units after the transformation to a MDS map are arbitrary since the distances are merely relative. For a quantitative 3D evaluation of CT distributions in two-color 3D FISH experiments, the 3D-RRD computer program was used (see \[[@pbio-0030157-b25]\] and \[[@pbio-0030157-b59]\] for detailed description). Briefly, the program determines (1) the center of gravity of a given nucleus and its borders, on the basis of the DNA counterstain, and (2) all voxels of painted chromosome territories. The nuclear radius in any direction from the nuclear center of gravity to the segmented nuclear edge was normalized to 100%, and the nuclear space was divided into 25 concentric shells. All shells possessed the same thickness along each possible radial vector from the center of the nucleus to the periphery. Accordingly, the thickness of these shells in flat ellipsoidal nuclei was much larger along the x- and y-axis than along the z-axis. Thus, the distribution of the DNA of painted CTs was measured and expressed as a function of the relative distances of each shell from the center of the nucleus. Significance tests were carried out with either SPSS 11 or Sigma Stat (SPSS). If not otherwise stated a K-S significance test was applied. ### Modeling of human fibroblast cell nuclei with statistical CT distributions {#s4a6} To simulate the statistical distribution of CTs in human fibroblast nuclei, the SCD model was applied \[[@pbio-0030157-b25],[@pbio-0030157-b41],[@pbio-0030157-b60]\]. The DNA content of individual human chromosomes \[[@pbio-0030157-b61]\] was used to estimate the number of 1-Mbp chromatin domains constituting a given model chromosome. These domains were represented by 500-nm diameter spheres. Model nuclei were generated with an ellipsoidal shape with half-axes representing the average half-axes of human G0 fibroblast nuclei determined from light-optical stacks (x = 10 μm, y = 5 μm, and z = 2.5 μm). Briefly, starting configurations representing a statistical chromatid distribution in male diploid human fibroblasts (46, XY) at late anaphase/telophase were established as follows. The location of the center of gravity of each chromatid was initially represented by the mass center of a small inelastic sphere (iS). The volume of a given iS was proportional to the DNA content of its natural chromosome counterpart, while its radius was very small compared with the half-axes of the ellipsoid. The total volume of the 46 iSs (representing the 46 chromosomes of the diploid human complement) comprised 22% of the total volume of the ellipsoidal model nucleus. The mass centers of the 46 iSs were statistically placed into the ellipsoid as follows. Using a random-number generator, the 3D coordinates for the mass centers of all iSs were generated iteratively in a nonoverlapping fashion. If the addition of a new iS yielded any overlap with the position of already existing iS, the 3D coordinates were discarded and new randomly generated 3D coordinates were tested. This process was repeated until all 46 iSs were located in the model nucleus. In a second step, chromatids were modeled as small rods with a spherical cross section of 500 nm (see [Figure 1](#pbio-0030157-g001){ref-type="fig"}D). Each rod represents a linear chain of spherical 1-Mbp chromatin domains. To model the dense packaging of 1-Mbp domains in chromatids, a distance of 13 nm was simulated between the centers of gravity of any two adjacent 1-Mbp domains along the chromatid axis. For example, a rod representing a chromatid with a DNA content of 100 Mbp had a length of 1,300 nm. In this way we modeled highly compacted and rigid chromatids. Rods were placed with a random orientation inside the 46 iSs in such a way that their centers of gravity coincided with the centers of the iSs from step one. As a third step, the relaxation of the statistically placed chromatids into decondensed CTs was simulated. For this purpose Monte Carlo relaxation loops were carried out with about 400,000 steps to obtain thermodynamic equilibrium configurations \[[@pbio-0030157-b14]\]. For decondensed CTs we assumed 120-kb linker connections between neighboring 1-Mbp chromatin domains, representing higher-order clusters of 100-kb loop domains \[[@pbio-0030157-b62]\]. These connections were modeled by entropic spring potentials enforcing a mean distance of 600 nm between the centers of gravity of adjacent domains. For distances between the centers of two adjacent 1-Mbp domains of 500 nm or more, we assumed that their repulsive potential was zero. When distances became smaller than 500 nm, the repulsive potential became increasingly positive, resulting in an increasing mutual repulsion between the two domains. Long-term Monte Carlo relaxation loops showed that the two assumptions of a spring potential and a repulsive potential are not sufficient to maintain the experimentally observed compactness of CTs. To achieve model CTs with diameters comparable to CTs in fibroblast nuclei, we introduced a weak potential barrier around each simulated chromatin domain chain representing a given chromosome. In each Monte Carlo step the 3D coordinates of a randomly chosen 1-Mbp domain were changed slightly for each CT. The new coordinates were accepted if the resulting chromatin domain configuration came closer to the thermodynamic equilibrium. Otherwise they were rejected and the process repeated. The 3D positions of the CT centers of gravity after 400,000 Monte Carlo steps represented the statistical arrangement of CTs in model nuclei. The further evaluation of distances and angles between model CTs was performed as described for experimental fibroblast nuclei. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### Male Human Fibroblast Nucleus (G0): Deconvolution 3D Microscopy, CT Classification, and 3D Reconstruction Images selected from a stack of 20 light-optical serial sections. Sections from a fibroblast nucleus following chromosome painting with 24-color M-FISH were recorded by wide-field epifluorescence microscopy. Images from left to right show optical sections obtained from the bottom to the top of the nucleus. \(A) RGB images without deconvolution. \(B) RGB images after deconvolution. \(C) False color images after classification. \(D) Still shot from [Video S1](#sv001){ref-type="supplementary-material"}, which shows the simulation of CT expansion in a fibroblast model nucleus according to the SCD model (compare with [Figure 1](#pbio-0030157-g001){ref-type="fig"}). (766 KB JPG). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### Positions of IGCs of Each CT in G0 Human Fibroblast Nuclei Each point represents the 2D projection of the ICG of an individual CT. The x- and y-half-axes of the ellipses shown represent the mean x- and y-half-axes of all the 54 nuclei evaluated. Each nucleus was rotated until its long axis fitted the abscissa of the coordinate system. The inner ellipsoid lines were drawn to facilitate the comparison of the ICG locations from different CTs. (1.5 MB JPG). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### Prometaphase Rosettes: Positions of IGCs of Each Chromosome 2D projections of the IGCs determined for all chromosome types (1--22, X, and Y) present in 28 prometaphase rosettes of male diploid human fibroblasts (compare to [Figure S2](#sg002){ref-type="supplementary-material"}). (1.2 MB JPG). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S4 ::: {.caption} ###### Cumulative Normalized 3D Radial Distances in Human Fibroblast Nuclei (G0) Cumulative frequency of normalized 3D CN--CT distances for all chromosome types (1--22, X, and Y) measured in 54 G0 human fibroblast nuclei of the 2D projections shown in [Figure S2](#sg002){ref-type="supplementary-material"}. The origin on the abscissa corresponds to the CN, and "1" corresponds to the maximum 3D distance from CN to any site of the nuclear periphery. Distances were sorted into ten classes representing ten concentric 3D nuclear shells. To define these shells, normalized 3D radii were subdivided into ten equal segments. (919 KB JPG). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S5 ::: {.caption} ###### Prometaphase Rosette: Deconvolution 3D Microscopy, PC Classification, and 3D Reconstruction Prometaphase rosette from a male fibroblast following chromosome painting with the 24-color M-FISH protocol. (A--C) Typical sections selected from a stack of 20 light-optical serial sections obtained by wide-field epifluorescence microscopy (A) before deconvolution, (B) after deconvolution, and (C) after classification. \(D) 3D reconstruction from the entire stack of deconvoluted light-optical images. \(E) Mid-plane section recorded in eight channels. Note that the Cy7 channel was not analyzable in this experiment. As a consequence, Chromosomes 5 and 19 were both labeled solely by Spectrum Green, while Chromosomes 8 and 20 were labeled solely by Cy3. Despite these limitations, differences in size and labeling intensity allowed the classification of these PCs. (876 KB JPG). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S6 ::: {.caption} ###### Cumulative Normalized 3D Radial Distances in Human Fibroblast Prometaphase Rosettes Cumulative frequency plot of normalized 3D CR--PC distances for all chromosome types (1--22, X, and Y) measured in 28 rosettes of the 2D projections shown in [Figure S3](#sg003){ref-type="supplementary-material"}. The origin on the abscissa corresponds to the CR, and "1" corresponds to the maximum 3D distance from CR to any site of the rosette periphery. Distances were sorted into ten classes as in [Figure S4](#sg004){ref-type="supplementary-material"}. (922 KB JPG). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S7 ::: {.caption} ###### Mean Normalized 3D CN--CT/CR--PC Distances and Standard Deviation of All Chromosome Types in Human Male Fibroblasts Left: normalized mean radial 3D CN--CT distances from the 54 evaluated 3D fibroblast nuclei. Right: normalized mean radial 3D CR--PC distances of the 28 evaluated 3D prometaphase rosettes. SD, standard deviation. (424 KB JPG). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S8 ::: {.caption} ###### Comparison of Radial CT Positions in Quiescent and Proliferating Fibroblasts (A--C) Left: for quality control chromosome paint probes were hybridized to metaphase spreads from phytohemagglutinin-stimulated human lymphocytes. For example, shown are metaphase spreads after two-color 3D FISH with the following pairs of chromosome paint probes: (A) HSA 18 (red) and HSA 19 (green); (B) HSA 17 (green) and HSA Y (red); and (C) HSA 1 (red) and HSA 20 (green). Painted metaphase chromosomes show a rather homogeneous coverage, except for the centromere region, which remained unstained because of the signal suppression with C~o~t-1 DNA. Right: maximum intensity projections of confocal image stacks from selected nuclei in quiescent (G0) and proliferating (early S-phase) human male fibroblasts demonstrate the variability of CT positioning. CT colors are the same as described for metaphase chromosomes. \(D) To compare the voxel-based analysis of 3D radial CT 18 and CT 19 distributions in G0 and S-phase fibroblasts (see [Figure 6](#pbio-0030157-g006){ref-type="fig"}A and [6](#pbio-0030157-g006){ref-type="fig"}D) with an IGC-based 3D analysis, we allocated CT 18 IGCs and CT 19 IGCs to one of five concentric nuclear shells with equal volume. Shell 1 corresponds to the most peripheral shell. The data was analyzed in this way to compare with [Figure 1](#pbio-0030157-g001){ref-type="fig"} in Bridger et al \[[@pbio-0030157-b27]\], except that 3D distances were used here. In both quiescent and S-phase nuclei a higher fraction of CT 19 IGCs was found in the two central shells 4 and 5 than for CT 18 IGCs. In contrast, in S-phase nuclei the fraction of CT 18 IGCs was larger in the more peripheral shells 1--3. In quiescent nuclei only shells 2 and 3 revealed a higher fraction of IGCs from HSA 18 CTs, while the fraction of IGCs from HSA 19 CTs was slightly higher in the most peripheral shell, shell 1. The results are consistent with a voxel-based approach (see [Figure 6](#pbio-0030157-g006){ref-type="fig"}). \(E) Comparison of the median and the mean 3D CN--CT distances obtained for HSA 18 CTs and HSA 19 CTs in G0 and S-phase nuclei. A plus sign indicates a slight shift of CT 18 IGCs towards the nuclear periphery in early S-phase fibroblast nuclei compared to G0 nuclei, whereas a minus sign indicates a shift to the interior of the CT 19 IGCs. However the measured shifts were statistically not significant (ns). \(F) Median and mean normalized 3D CN--CT 18 distances were not significantly different from CN--CT 19 distances in G0 fibroblast nuclei, but showed a marginal difference (\* p \< 0.05) in early S-phase nuclei. (1.3 MB JPG). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S9 ::: {.caption} ###### G0 Nuclei: Mean Angular Separation between 3D IGCs of Homologous CTs and the CN CT--CN--CT angles were measured between the IGCs of homologous CTs and the center of the nucleus (CN) in 54 G0 fibroblast nuclei. Sample sizes (n) indicate the number of nuclei in which CT--CN--CT angles could be measured for a given pair of homologous CTs. The experimental distribution did not deviate from a normal distribution (p \> 0.05; one-tailed K-S test of goodness of fit). With few exceptions pairwise comparisons of the mean angular separation between a pair of homologous CTs with the respective mean angle distribution in 60 random point distribution model nuclei did not show a significant difference (p \> 0.05; two-tailed K-S test). Significant differences (p \< 0.05) are indicated by an asterisk. The comparison of mean angular separation of homologous CTs with statistically placed homologous CTs in 50 SCD model nuclei also did not reveal significant differences except for 4-CN--4 and 20-CN--20 angles, which were smaller in the experiment than in the model (\\p \< 0.01). (503 KB JPG). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S10 ::: {.caption} ###### Prometaphase Rosettes: Mean Angular Separation between 3D IGCs of Homologs and the CR PC--CR--PC angles between the IGCs of homologous PCs and the rosette center (CR) measured in 28 prometaphase rosettes. No significant difference was detected in comparison with angular separations found in the random point distribution model (p \> 0.05; two-tailed K-S test). (426 KB JPG). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S11 ::: {.caption} ###### Significance Levels for Pairwise Comparisons between Heterologous 3D CT--CN--CT Angles in 54 G0 Fibroblast Nuclei Significance levels were determined by the two-tailed K-S test. Green, not significant, p \> 0.05; yellow, p \< 0.05; red, p \< 0.01. Minus/plus signs in a colored field indicate that the chromosome pair given on the left shows a significantly shorter/greater mean radial distance than the chromosome pair presented at the top. \(A) Large chromosomes: HSAs 1--5. \(B) Small, acrocentric chromosomes: HSAs 13, 14, 15, 21, and 22. \(C) Other small chromosomes: HSAs 16--20. (292 KB JPG). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S12 ::: {.caption} ###### Arrangements of HSA 7 and HSA 8 CTs in 50 Fibroblast Nuclei Nuclei of G0 fibroblasts were subjected to 3D FISH with painting probes for HSA 7 and HSA 8, labeled with dUTP-Cy3 (green) and dUTP-FITC (red), respectively. Maximum intensity projections of confocal image stacks from 50 scanned nuclei are shown to demonstrate the variability of proximity patterns (for quantitative measurements see [Figure S13](#sg013){ref-type="supplementary-material"}). (870 KB JPG). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S13 ::: {.caption} ###### Quantitative Evaluation of the Angular Separation of Homologous and Heterologous Pairs of HSA 7 and HSA 8 CTs in 50 Fibroblast Nuclei Upper left: Ellipses represent the normalized 2D shape of the nuclei and show 2D projections of the radial IGC locations of HSA 7 and HSA 8 CTs. Upper right: The IGCs of CT 7 and CT 8 pairs, respectively, were rotated around the nuclear center until one IGC lay on the positive abscissa (closed circles). Open circles show the IGC position of the corresponding homolog. Below are the mean 3D CT--CN--CT angles between homologous and heterologous HSA 7 and HSA 8 CTs, their ranges, and their standard deviations. Angles between homologous and heterologous pairs showed a normal distribution. Comparisons of experimental data with mean 3D angle distributions in the random point distribution model or SCD model did not show a significant difference (p \> 0.05; two-tailed K-S test). (328 KB JPG). ::: ::: {.caption} ###### Click here for additional data file. ::: Video S1 ::: {.caption} ###### Model Nucleus: CT Simulation The video shows the simulation of CT expansion in a fibroblast model nucleus according to the SCD model (compare with [Figure 1](#pbio-0030157-g001){ref-type="fig"}). (567 KB MPG). ::: ::: {.caption} ###### Click here for additional data file. ::: The authors dedicate this paper to the 80th anniversary of Professor F. Vogel. We thank Marion Cremer for kindly contributing her 3D analysis of radial HSA 18 and 19 CT arrangements in nuclei of cycling amniotic fluid cells ([Figure S8](#sg008){ref-type="supplementary-material"}D). We appreciate the technical support of Leica (Cambridge, United Kingdom), SVI (Hilversum, Netherlands), ZeissVision (Hallbergmoos, Germany), and T.I.L.L.-Photonics (Munich, Germany), and we thank Adrian Sumner (North Berwick, United Kingdom) for editorial services. We acknowledge very helpful discussions with Joachim Walter (T.I.L.L.-Photonics), Rainer Heintzmann (MPI for Biophysical Chemistry, Göttingen, Germany), and Jörg Langowski (DKFZ, Heidelberg, Germany). We are indebted to Joanna Bridger and Wallace Marshall, as well as two anonymous reviewers for their helpful comments. This work was supported by the Deutsche Forschungsgemeinschaft (Sp460/2--1, Cr59/20, and Cr60/19) and the German--Israeli Foundation for Research and Technology (G-112--207.04/97). Competing interests. The authors have declared that no competing interests exist. Author contributions. AB, GK, IS, CF, SM, RE, CC, MRS, and TC conceived and designed the experiments. AB, GK, DK, and SM performed the experiments. AB, DK, KS, and MRS analyzed the data. AB and SM contributed reagents/materials/analysis tools. AB, GK, IS, MRS, and TC wrote the paper. ¤ Current address: Cancer Research UK, London, United Kingdom Citation: Bolzer A, Kreth G, Solovei I, Koehler D, Saracoglu K, et al. (2005) Three-dimensional maps of all chromosomes in human male fibroblast nuclei and prometaphase rosettes. PLoS Biol 3(5): e157. \[number\]D : \[number\]-dimensional CN : intensity gravity center of the nucleus CR : intensity gravity center of the prometaphase rosette CT : chromosome territory FCS : fetal calf serum FISH : fluorescence in situ hybridization HSA : Homo sapiens chromosome IGC : intensity gravity center iS : inelastic sphere K-S : Kolmogorov-Smirnov MDS : multidimensional scaling M-FISH : multiplex fluorescence in situ hybridization NOR : nucleolar organizer region PC : prometaphase chromosome SCD : spherical 1-Mbp chromatin domain t-test : Students t-test U-test : Mann-Whitney U-test	PubMed Central	2024-06-05T03:55:55.798122	2005-4-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084335/", "journal": "PLoS Biol. 2005 May 26; 3(5):e157", "authors": [ { "first": "Andreas", "last": "Bolzer" }, { "first": "Gregor", "last": "Kreth" }, { "first": "Irina", "last": "Solovei" }, { "first": "Daniela", "last": "Koehler" }, { "first": "Kaan", "last": "Saracoglu" }, { "first": "Christine", "last": "Fauth" }, { "first": "Stefan", "last": "Müller" }, { "first": "Roland", "last": "Eils" }, { "first": "Christoph", "last": "Cremer" }, { "first": "Michael R", "last": "Speicher" }, { "first": "Thomas", "last": "Cremer" } ] }
PMC1084336	Eukaryotic cells have an organizational problem. The specialized proteins found in cellular organelles (structures with specific functions such as energy production) are mostly encoded within the nucleus. To build and maintain a cell that works efficiently under all conditions, each type of organelle needs to be able to send signals to the nucleus to say "Send more protein X" or "hold back on enzyme Y." Think of it as the cellular version of grocery store clerks\' restocking orders to the warehouse. In plant cells, the chloroplasts (the photosynthetic organelles that convert light excitation energy into chemical energy) send signals to the nucleus to control the expression of the genes that encode chloroplast-localized proteins such as the enzymes that fix carbon dioxide, make chlorophyll, or perform photosynthesis. The accumulation of the chlorophyll precursor Mg-protoporphyrin IX provides one of these signals. A protein called GUN4 both enhances the activity of Mg-chelatase, the enzyme that makes Mg-protoporphyrin IX, and plays a role in the chloroplast-to-nucleus signaling activity of Mg-protoporphyrin IX in Arabidopsis, a well-studied plant. To discover how GUN4 has these effects, Mark Verdecia in Joseph Noel\'s laboratory and Rob Larkin, formally in Joanne Chory\'s laboratory, determined the crystal structure of the GUN4 equivalent in the cyanobacteria Synechocystis. Cyanobacteria are the evolutionary ancestors of chloroplasts, so whatever GUN4 does in these cells is likely to be important in plant cells. The researchers\' crystallographic studies, together with nuclear magnetic resonance and other studies, indicate that the porphyrin-binding region of Synechocystis GUN4 has a unique three-dimensional shape that resembles a cupped hand, the inner concave surface of which is highly hydrophobic. Because of this tendency to repel water, the researchers call this region the "greasy palm" of the cupped hand. This structure suggests how GUN4 is involved in the chloroplast-to-nucleus signaling activity of Mg-protoporphyrin. By wrapping Mg-protoporphyrin IX in its cupped, greasy palm, the GUN4 structure provides a novel vehicle for binding Mg-protoporphyrin IX and may be involved in transporting signals from the chloroplast to the nucleus. In addition, the structure also suggests that GUN4 may be involved in photoprotection. Although light drives photosynthesis, which is essential to green plants, light has a downside---porphyrins combine with the oxygen released during photosynthesis to generate reactive oxygen species, which are generally damaging to the cell. GUN4, by cocooning Mg-protoporphyrin IX in its protective hand, may provide a way to safely move porphyrin around the chloroplast without exposing it to oxygen. Finally, the detailed structural and functional studies described by Noel and colleagues explain how GUN4 enhances the activity of Mg-chelatase. GUN4 binds Mg-protoporphyrin IX, the product of the chelatase, much better than protoporphyrin IX and so will tend to enhance the enzymatic reaction by removing its product.[](#pbio-0030167-g001){ref-type="fig"} ::: {#pbio-0030167-g001 .fig} ::: {.caption} ###### A "greasy palm" in the cupped hand of a porphyrin-binding protein reveals how this hydrophobic protein repels water. (Illustration: Mark Verdecia) ::: ![](pbio.0030167.g001) :::	PubMed Central	2024-06-05T03:55:55.804952	2005-4-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084336/", "journal": "PLoS Biol. 2005 May 26; 3(5):e167", "authors": [] }
PMC1084337	On this, theologians, philosophers, and biologists can agree: we are more than the sum of our genes. Biological complexity arises not from gene number but from patterns of gene expression, which change under the direction of both genetic and so-called epigenetic mechanisms. Epigenetics, broadly defined, concerns heritable changes in gene function that don\'t involve changes in DNA sequence. Until recently, studies of heritable traits have focused largely on mutations in DNA. But it\'s become increasingly clear that how DNA is packaged in the nucleus also impacts heritability. Epigenetic changes are mediated largely by proteins that shape and remodel chromatin---the association of DNA and histone proteins that condenses the genome into compact bundles inside the nucleus. Different cell types have different chromatin arrangements during development and cell differentiation that appear to regulate gene expression, which possibly accounts for the unique gene expression patterns associated with specific cell types. Such phenomena have been well-studied for specific genes or chromosomal regions, but to understand the full impact of epigenetic mechanisms on gene regulation, we need a more panoramic view of gene organization within the nucleus. In a new study, Thomas Cremer together with Andreas Bolzer and an interdisciplinary team of German physicists, bioinformaticians, and geneticists created 3D positional maps of each human chromosome simultaneously in a single nucleus to investigate the link between chromatin structure and cell-specific gene expression. Working with human fibroblasts, cultured from a skin biopsy from a two-year-old boy, the authors were able to visualize and study the order of the full genetic complement within a human nucleus.[](#pbio-0030188-g001){ref-type="fig"} ::: {#pbio-0030188-g001 .fig} ::: {.caption} ###### In this karotype from a female human lymphocyte, the gene-rich areas are stained green and the gene-poor areas are red (Photo: Irina Solovei) ::: ![](pbio.0030188.g001) ::: Cremer and colleagues first produced a 3D topological map of all 46 chromosomes in different cell types at key points in the cell cycle---a landmark achievement---using a fluorescent staining technique that preserves chromosome shape during visual inspection under the microscope. Next, they established that small chromosomes in quiescent (nondividing) fibroblasts hewed close to the center of the nucleus while the large chromosomes were preferentially found at the nuclear rim, regardless of their gene density. Nuclei from cells entering the prometaphase stage of the cell cycle---just before chromosomes are aligned along the center of the nucleus prior to segregation---revealed a size-correlated chromosomal distribution akin to that seen in the quiescent nuclei. Statistical modeling analyses indicated that these size correlations do not simply reflect the geometric constraints of fitting into the nucleus, but likely hint at some degree of functional order within the nucleus. Because previous studies of cells with sphere-like nuclei correlated chromosomal arrangements with gene density, the authors investigated how shape affects chromosome position along the nuclear radius. Fibroblast nuclei are somewhat flat and ellipsoidal. Chromosomes in similarly shaped amniotic fluid cells assumed the same size-related positions taken by chromosomes in fibroblast nuclei. But when the authors examined the higher-order chromatin arrangements in fibroblasts and lymphocytes, they found that, even though the cell types differ in nuclear shape and radial chromosomal arrangements, they both show a nonrandom higher-order chromatin architecture correlated with gene density. Many questions remain concerning the functional and physiological significance of these observations: Do shape changes produce changes in chromosomal arrangements and vice versa? Do shape changes produce changes in gene expression patterns? Cremer and colleagues conclude that, although nonrandom chromosome positions occur, these appear to be governed by a degree of uncertainty and more likely reflect probabilistic preferences inside the nucleus. Still, deterministic mechanisms in higher-order chromatin structure may exist---sequestering gene-rich chromatin areas in the nuclear interior, for example, protected from malevolent agents entering the nucleus. And given the coexistence of size-correlated features with gene-density-correlated features seen in this study, it may well be that both random and deterministic factors combine to create the nuclear landscape.	PubMed Central	2024-06-05T03:55:55.805594	2005-4-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084337/", "journal": "PLoS Biol. 2005 May 26; 3(5):e188", "authors": [] }
PMC1084338	Introduction {#s1} ============ Apoptosis is considered the primary means by which physiologic cell death occurs \[[@pbio-0030185-b01]\]. The fate of apoptotic material is rapid clearance and degradation by phagocytes. There is, however, growing evidence that apoptotic death need not be an endpoint, and that dying cells are capable of transferring antigen to the immune system for the induction of T cell immunity \[[@pbio-0030185-b02],[@pbio-0030185-b03]\]. We have previously demonstrated that human dendritic cells (DCs) phagocytose apoptotic cells, and rather than degrading the internalized material, the DCs are capable of generating peptide epitopes for major histocompatibility (MHC) I molecules and activating viral- and tumor-antigen-specific CD8^+^ T cells \[[@pbio-0030185-b04]--[@pbio-0030185-b06]\]. This pathway has been referred to as cross-presentation for its ability to "cross" classically defined restrictions for MHC I antigen presentation \[[@pbio-0030185-b07]\]. Our work has offered a physiologically relevant mechanism for the in vivo phenomenon of cross-presentation, which accounts for both the cross-priming and cross-tolerization of tissue-restricted antigen-specific CD8^+^ T cells \[[@pbio-0030185-b08]--[@pbio-0030185-b10]\]. We have demonstrated that antigen capture occurs via receptor-mediated phagocytosis \[[@pbio-0030185-b05],[@pbio-0030185-b11]\], and that internalized apoptotic material can be located within the MHC II−containing compartment \[[@pbio-0030185-b12]\]; however, the trafficking of antigen from the apoptotic cell to the DC for generation of MHC I/peptide (MHC I/pep) complexes has not been fully characterized. To define the cellular machinery required for cross-presentation, several studies have focused on the use of cells expressing vector-encoded gene products to test whether proteasomal substrates (e.g., intact proteins) or chaperoned peptides serve as the source of antigen \[[@pbio-0030185-b13]--[@pbio-0030185-b15]\]. While these studies conclude that cellular proteins are the major source of antigen transferred to the immune system, the inability to demonstrate transfer of cell-associated antigen, or the lack of processed antigen within the dying cell, may have skewed the observed results. In other studies, the use of exogenous antigen bound to latex beads \[[@pbio-0030185-b16],[@pbio-0030185-b17]\], derived from internalized immune complexes \[[@pbio-0030185-b18]\] or whole protein \[[@pbio-0030185-b19]\], does not permit antigen processing to occur prior to capture by an antigen-presenting cell (APC); therefore, it is not surprising that these models demonstrate that the phagosome-to-cytosol pathway is the dominant means by which antigen is trafficked for processing and presentation onto MHC I. To examine whether antigen processed within the dying cell can be transferred to the DC, we designed in vitro and in vivo experiments to track the activation of polyclonal influenza-reactive T cell responses stimulated by DCs cross-presenting antigen from haplotype-mismatched apoptotic cells. We report the utilization of two independent pathways by which internalized antigen may access MHC I within the DC: in one, the substrate for cross-presentation is whole or partially degraded protein, which must be further processed by the DC; and in the other, we find evidence for processed antigen accessing the MHC I pathway of the DC. Concerning the in vivo presentation of viral antigen, this latter pathway seems dominant, thus permitting efficient loading of MHC I/pep complexes by the DC. Results {#s2} ======= In order to dissect the pathway or pathways by which antigens derived from apoptotic cells are processed and presented by DCs, we established in vivo and in vitro systems that permit monitoring of direct presentation and cross-presentation of antigen ([Protocol S1](#sd001){ref-type="supplementary-material"}; [Figure S1](#sg001){ref-type="supplementary-material"}). Using these model systems, we tested the hypothesis that dying cells participate in the processing of antigen for cross-presentation. To restrict the DCs\' capacity to process antigen, we employed bone-marrow-derived DCs prepared from mice deficient in TAP-1. The generation of antigen-specific MHC I/pep complexes was assayed based on the ability to stimulate influenza-reactive T cells. In all experiments, interleukin-12 (IL-12) was added to the DC/CD8^+^ T cell cultures to bypass the requirement for CD4^+^ T cell help \[[@pbio-0030185-b08]\]. To ensure that the influenza antigens being monitored required transporter activity for the generation of MHC I/pep complexes, we directly infected DCs prepared from TAP^−/−^ mice ([Figure 1](#pbio-0030185-g001){ref-type="fig"}). As has previously been reported, no T cell activation was evident when infected TAP^−/−^ DCs were employed ([Figure 1](#pbio-0030185-g001){ref-type="fig"}A and [1](#pbio-0030185-g001){ref-type="fig"}B). Furthermore, we established that the TAP^−/−^ DCs efficiently engulfed apoptotic bodies and that the kinetics of uptake were similar to those evident in wild-type (WT) DCs ([Figure S2](#sg002){ref-type="supplementary-material"}). DCs that had internalized dying cells were tested for their ability to cross-present antigen; in contrast to their ability to present antigen via the "classical" MHC I pathway, the TAP^−/−^ DCs were able to cross-present influenza antigen derived from MHC-mismatched apoptotic cells as efficiently as WT DCs ([Figure 1](#pbio-0030185-g001){ref-type="fig"}D and [1](#pbio-0030185-g001){ref-type="fig"}E). To rule out the possibility that transporter activity was simply being transferred via fusion of membranes between the apoptotic cells and the DC during phagocytosis, we tested whether TAP^−/−^ DCs that had previously internalized uninfected TAP-expressing apoptotic cells could now present antigen after direct infection with influenza (flu-\[TAP^−/−^ DC x/p 3T3\]; [Figure 1](#pbio-0030185-g001){ref-type="fig"}C). We detected background levels of T cell activation in these assays as compared to the robust stimulation observed using influenza-infected WT DCs ([Figure 1](#pbio-0030185-g001){ref-type="fig"}C), arguing against such a mechanism and supporting the possibility that TAP^−/−^ DCs are capturing processed antigen from apoptotic cells. ::: {#pbio-0030185-g001 .fig} Figure 1 ::: {.caption} ###### TAP^−/−^ DCs Cross-Present Antigen Derived from Apoptotic Cells (A and B) Antigen presentation via the endogenous pathway was evaluated in WT DCs and TAP^−/−^ DCs by directly infecting cells with influenza virus and assaying for T cell activation. IFN-γ production and T cell precursor frequency were determined using an ELISPOT assay. \(C) To evaluate transfer of TAP activity from the dying cells to DCs, WT or TAP^−/−^ DCs after capture of apoptotic cells were directly infected and tested for their respective ability to activate CD8^+^ T cells via the endogenous pathway. (D and E) WT DCs (D) or TAP^−/−^ DCs (E) were co-cultured for 36--48 h with influenza-infected or uninfected allogeneic cells in the presence of TNF-α. As above, mature DCs were harvested and assayed for their ability to stimulate influenza-reactive CD8^+^ T cells. To bypass the requirement for CD4^+^ T cell help in the activation of CD8^+^ T cells via the exogenous pathway, IL-12 was added to the cultures. Spot-forming cells (SFCs) per 10^6^ T cells are reported. Data are representative of three experiments. Values are averages of triplicate wells with error bars indicating standard deviation. ::: ![](pbio.0030185.g001) ::: To further establish the role for antigen processing in the dying cell, we generated apoptotic cells that expressed influenza proteins but were unable to process the influenza viral antigens. Prior to infection with influenza and the induction of apoptosis, 3T3 cells were treated with the proteasome inhibitor lactacystin \[[@pbio-0030185-b20]\]. We found that the lactacystin-treated cells expressed levels of influenza antigen similar to untreated cells ([Figure 2](#pbio-0030185-g002){ref-type="fig"}A). The inhibition of proteasome activity was confirmed functionally using treated cells as stimulators for HA-reactive T cells restricted to H-2^d^, the MHC haplotype of the dying 3T3 cells ([Figure 2](#pbio-0030185-g002){ref-type="fig"}B). Parallel cultures were triggered to undergo apoptosis and were co-cultured with WT or TAP^−/−^ DCs as described in the [Materials and Methods](#s4){ref-type="sec"}. Cross-presentation of antigen by the DCs was evaluated based on the activation of influenza-reactive CD8^+^ T cells. We found that the lactacystin-treated apoptotic cells were competent to serve as a source of antigen for WT DCs ([Figure 2](#pbio-0030185-g002){ref-type="fig"}C); however, the lack of processed antigen in lactacystin-treated 3T3 cells prevented cross-presentation by TAP^−/−^ DCs ([Figure 2](#pbio-0030185-g002){ref-type="fig"}D). This result further establishes that transporter activity is not passed from the dying cell to the TAP^−/−^ DC as a result of an ill-defined fusion event. Instead, it is an active process whereby processed antigen within the dying cell is being utilized by the DC for the generation of MHC I/pep complexes. ::: {#pbio-0030185-g002 .fig} Figure 2 ::: {.caption} ###### Processed Antigen from the Dying Cell Is Required for MHC I Presentation in TAP^−/−^ DCs To generate apoptotic cells lacking processed antigen, lactacystin pretreatment of influenza-infected H-2^d^ 3T3 cells was performed. Expression of influenza antigen was evaluated by intracellular FACS analysis using influenza NP mAbs followed by PE-conjugated goat anti-mouse mAb (A). Expression of MHC I/pep complexes in the lactacystin-treated 3T3 cells was evaluated by monitoring the activation of H-2^d^-restricted influenza hemagglutanin-reactive T cells. The K^d^-restricted immunodominant peptide (HA~210--219~) derived from hemagglutanin was pulsed onto 3T3 cells and served as a positive control (B). The influenza-infected H-2^d^ 3T3 cells were then induced to undergo apoptosis, and co-cultures were generated using C57BL/6 WT DCs (C) or TAP^−/−^ DCs (D). To evaluate T cell activation and expansion, DCs cross-presenting antigen were cultured with CD8^+^ T cells in the presence of IL-12 for 7--8 d. T cells were then harvested and tested for influenza reactivity in a 20-h IFN-γ ELISPOT. H-2^b^ EL4 cells with or without influenza infection served as the stimulators in the ELISPOT assay as above. Data are representative of two experiments. Values are averages of triplicate wells with error bars indicating standard deviation. ::: ![](pbio.0030185.g002) ::: To determine the importance of antigen access to the dying cell\'s endoplasmic reticulum (ER), we used RMA/s cells, which are deficient in TAP-2, as a source of influenza antigen. These cells express influenza proteins upon infection ([Figure S3](#sg003){ref-type="supplementary-material"}A), but do not facilitate peptide transport into the ER, as established by their inability to re-stimulate an influenza-reactive CD8^+^ T cell line ([Figure S3](#sg003){ref-type="supplementary-material"}B). When WT DCs were employed as the APC, RMA/s served as a source of antigen for the generation of MHC I/pep complexes ([Figure 3](#pbio-0030185-g003){ref-type="fig"}A); in contrast, when TAP^−/−^ DCs were used, no activation of CD8^+^ T cells was observed ([Figure 3](#pbio-0030185-g003){ref-type="fig"}B). While the absence of TAP in both the dying cell and the DC prevented loading of MHC I, antigen presentation on MHC II was unaffected, as equivalent stimulation of influenza-reactive CD4^+^ T cells was observed when using WT or TAP^−/−^ DCs ([Figure 3](#pbio-0030185-g003){ref-type="fig"}A and [3](#pbio-0030185-g003){ref-type="fig"}B). ::: {#pbio-0030185-g003 .fig} Figure 3 ::: {.caption} ###### Transporter Activity Is Required in Either the Apoptotic Cell or the DC for Efficient Cross-Presentation of Antigen RMA/S cells were infected with influenza and irradiated with UVB to allow for antigen loading and the induction of apoptotic death. Co-cultures were generated as described in the [Materials and Methods](#s4){ref-type="sec"} using WT DCs (A) or TAP^−/−^ DCs (B). DCs were harvested and tested for their ability to cross-present antigen and activate influenza-reactive CD8^+^ T cells as measured in a 40-h ELISPOT assay. To evaluate loading of MHC II, WT DCs and TAP^−/−^DCs that had captured apoptotic antigen were tested for their ability to activate influenza-reactive CD4+ T cells. Data are representative of four experiments. Values are averages of triplicate wells with error bars indicating standard deviation. ::: ![](pbio.0030185.g003) ::: Together, these results suggest the presence of two independent pathways by which antigen may be cross-presented. The first is a pathway that requires transporter activity in the DC---presumably relying on the transport of exogenous antigen from the phagosome to the cytosol---and accounts for the requirement for TAP-sufficient DCs to generate MHC I/pep complexes in conditions where the peptide is derived from whole or partially degraded protein that is present within internalized apoptotic lactacystin-treated 3T3 cells (see [Figure 2](#pbio-0030185-g002){ref-type="fig"}C) or TAP^−/−^ RMA/S cells (see [Figure 3](#pbio-0030185-g003){ref-type="fig"}B). In the second, the DCs are able to capture processed antigen present within proteasome- and TAP-competent dying cells; notably, this latter antigen source may be cross-presented without a need for further transporter activity within the DC (see [Figure 1](#pbio-0030185-g001){ref-type="fig"}E). To evaluate the in vivo relevance of these findings, we took advantage of a recent observation made by Woodland and colleagues regarding the unique ability of DCs to generate MHC I peptides derived from the influenza A/PR/8 acid polymerase (PA) protein \[[@pbio-0030185-b21]\]. They demonstrated that while most cells are capable of processing influenza A/PR/8 nucleoprotein (NP), only DCs process and present the epitope PA~224--233~. We confirmed this result and demonstrated that both epitopes required transporter activity in the infected DC---in other words, influenza-infected TAP^−/−^ DCs activated neither NP~366--374~- nor PA~224--233~-specific T cells (data not shown). We next analyzed the T cell repertoire generated after priming C57BL/6 (H-2^b^) mice with live influenza versus influenza-infected apoptotic 3T3 (H-2^d^) cells. Given that the apoptotic cells express both NP and PA protein, but process only NP~366--374~, the in vivo activation of NP~366--374~- and PA~224--233~-specific T cells would suggest that the DC was responsible for processing the antigen, whereas a response to NP~366--374~ in the absence of a response to PA~224--233~ would imply that processed antigen was the preferred source of antigen for cross-presentation. As reported by Crowe et al. \[[@pbio-0030185-b21]\], priming the mice with 300 hemaglutanin units (HAU) of live influenza resulted in the activation of both NP~366--374~- and PA~224--233~-reactive T cells ([Figure 4](#pbio-0030185-g004){ref-type="fig"}A). However, when apoptotic influenza-infected 3T3 cells were injected into naïve mice, we observed a robust NP~366--374~-specific response and only weak reactivity to PA~224--233~ ([Figure 4](#pbio-0030185-g004){ref-type="fig"}B). To examine whether the response was indeed due to the processing of antigen within the dying cell, we primed naïve mice using lactacystin-treated influenza-infected 3T3 cells (prepared as above), and observed a marked reduction in the efficiency of NP~366--374~-reactive T cell priming ([Figure 4](#pbio-0030185-g004){ref-type="fig"}C). Similarly, we demonstrated that infected kidney epithelial cells derived from β~2~m-deficient but not TAP^−/−^ mice were capable of priming NP~366--374~-specific T cells ([Figure 5](#pbio-0030185-g005){ref-type="fig"}). These data again highlight the importance of processed antigen within the ER of the dying cell. In sum, we demonstrated the in vivo relevance of DCs capturing processed antigen derived from dying cells for the cross-priming of CD8^+^ T cells, and suggest that in some experimental conditions, this pathway may be more efficient than the cross-priming of whole protein. ::: {#pbio-0030185-g004 .fig} Figure 4 ::: {.caption} ###### Immunization with Apoptotic Cells Results in the Selective Priming of T Cells Reactive to Processed Antigen (A and B) C57BL/6 mice were immunized intraperitoneally with 300 HAU of influenza (A), or 5 × 10^6^ infected apoptotic 3T3 cells (B). After 14 d, splenocytes were harvested, and CD8^+^ T cells were purified. To assay for the specificity of these cells, an IFN-γ ELISPOT was performed using the following stimulators: DCs alone or DCs pulsed with either 1 μM NP~366--374~ or 1 μM PA~224--233~ peptide. \(C) C57BL/6 mice were immunized intraperitoneally with 5 × 10^6^ untreated versus lactacystin-treated influenza-infected apoptotic 3T3 cells. As above, 14 d after priming, splenocytes were harvested, and CD8^+^ T cells were purified and assayed for their reactivity to NP~366--374~ versus PA~224--233~. In this experiment, peptide-pulsed EL4 cells were employed as the stimulators. Data are representative of two experiments. Values are averages of triplicate wells with error bars indicating standard deviation. ::: ![](pbio.0030185.g004) ::: ::: {#pbio-0030185-g005 .fig} Figure 5 ::: {.caption} ###### Processed Antigen within the Dying Cell Is Required for Efficient In Vivo Priming C57BL/6 mice were immunized intraperitoneally with 300 HAU of influenza (A), or 2 × 10^6^ infected apoptotic kidney epithelial cells derived from β~2~m-deficient (B) or TAP-deficient mice (C). After 14 d, splenocytes were harvested, and CD8^+^ T cells were purified. To assay for the specificity of these cells, an IFN-γ ELISPOT was performed using the following stimulators: EL4 cells alone or EL4 cells pulsed with either 1 μM NP~366--374~ or 1 μM PA~224--233~ peptide. Values are averages of triplicate wells with error bars indicating standard deviation. ::: ![](pbio.0030185.g005) ::: Discussion {#s3} ========== Several models for antigen cross-presentation have evaluated a role for transporter activity, and most report that COOH terminal processing by the proteasome and utilization of TAP by the APC is essential \[[@pbio-0030185-b22],[@pbio-0030185-b23]\]. This is understandable when exogenous antigen is derived from internalized immune complexes \[[@pbio-0030185-b18]\], antigen-coated latex beads \[[@pbio-0030185-b16],[@pbio-0030185-b17]\], or whole protein \[[@pbio-0030185-b19]\]. In these instances, there is no ability for antigen processing to occur in a manner that would permit loading of DC MHC I in the absence of transporter activity. With respect to in vitro systems that have reported a TAP-independent pathway, the antigens successfully presented seem limited to peptides immediately COOH-terminal to an ER targeting sequence, or those within secreted or transmembrane proteins that are processed by still undefined ER proteases \[[@pbio-0030185-b24],[@pbio-0030185-b25]\]. It has also been demonstrated that at high levels of antigen challenge, it is possible for peptide epitopes to be generated by Cathepsin S within the phagolysosome \[[@pbio-0030185-b26]\]. In this study, we restricted our analysis to physiologically relevant levels of antigen (all of which require proteasome processing and TAP activity), and asked whether dying cells serve as a source of whole protein or whether they may also participate in antigen presentation by generating processed antigen that may be transferred to DCs. As has been previously shown, our data demonstrated that WT DCs can process antigen from cells that contain whole or partially processed protein (see [Figures 2](#pbio-0030185-g002){ref-type="fig"}C and [3](#pbio-0030185-g003){ref-type="fig"}A). To assess the ability of the dying cell to process the antigen, TAP^−/−^ DCs were used. Applying this strategy, we identified the existence of an antigen cross-presentation pathway that utilizes proteasome and transporter activity present in the dying cell (see [Figures 1](#pbio-0030185-g001){ref-type="fig"}, [2](#pbio-0030185-g002){ref-type="fig"}, and 3). Importantly, when lactacystin-treated or TAP^−/−^ apoptotic cells were the source of antigen, the TAP^−/−^ DCs were no longer capable of cross-presenting antigen (see [Figures 2](#pbio-0030185-g002){ref-type="fig"}D and [3](#pbio-0030185-g003){ref-type="fig"}B). Furthermore, we demonstrated activation of CD4^+^ T cells in all experimental conditions, illustrating that DCs indeed captured dying cells expressing influenza antigen, even in situations where MHC I presentation was inhibited. While this in vitro system allowed us to carefully control the nature of the antigen present in the dying cells (whole protein or processed antigen) and permitted us to ensure the transfer of cell-associated protein to the DCs, we also tested the ability to prime T cells in vivo. Taking advantage of the differential processing of influenza antigen by DCs versus other cell types, we demonstrated that in situations of direct infection, DCs processed the antigen. When dying cells were the source of antigen, we observed a preferential skewing of T cell cross-priming toward the protein that could be processed by the apoptotic cell. Lactacystin treatment of the influenza-infected cell or the use of TAP^−/−^ cells confirmed the requirement for proteasome processing and transporter activity within the dying cell (see [Figures 4](#pbio-0030185-g004){ref-type="fig"} and [5](#pbio-0030185-g005){ref-type="fig"}). Furthermore, the ability to cross-prime influenza-specific T cells with β~2~m-deficient but not TAP-deficient cells indicated that the antigen was originating from the ER of the dying cell (see [Figure 5](#pbio-0030185-g005){ref-type="fig"}). When greater numbers of apoptotic cells were used (10--50×), it was possible to observe cross-priming of whole or partially degraded protein (data not shown). Our findings support the in vitro work shown here and that of Serna et al. \[[@pbio-0030185-b27]\]---cross-priming of influenza antigen favors the processed antigen within the dying cell ([Figure 6](#pbio-0030185-g006){ref-type="fig"}). Indeed, apoptotic cells may play an active role in antigen presentation through the delivery of processed antigen, in turn allowing for efficient generation of MHC I/pep complexes by the DC. ::: {#pbio-0030185-g006 .fig} Figure 6 ::: {.caption} ###### An Active Role for Apoptotic Cells in the Transfer of Antigen to DCs We propose that apoptotic cells play an active role through the transfer of processed antigen to DCs for the generation of MHC I/pep complexes. This pathway may be dominant in the presentation of infectious antigen as the virus may co-opt cellular translational machinery, resulting in high levels of viral protein, and the upregulation of stress proteins, as well as inducing apoptotic cell death. Defective ribosomal initiation products chaperoned by HSPs offer a potential source of antigen. Within the DC, HSPs derived from the internalized apoptotic cell may traffic via a retrograde transport pathway, shuttled to the trans-Golgi and then the ER via binding to KDEL receptors (A). Alternatively, the evidence for phagosome--ER (PHAGO-ER) fusion and/or the recycling of MHC I from the plasma membrane offers the possibility that processed antigen may interact directly with the DC\'s MHC I (B). As ER chaperones within the phagocytosed cell would be bound to the pool of peptides derived from newly synthesized proteins, these pathways offer the DC an accurate representation of what occurred immediately prior to death (A and B). At high concentrations of protein, we also find evidence for the DC processing the cross-presented antigen. This likely occurs via a phago--ER-to-cytosol pathway as has been previously described (C). ::: ![](pbio.0030185.g006) ::: The identification of this pathway raises the intriguing possibility that ER chaperones within the apoptotic cell are facilitating delivery of peptide epitopes to the DC (see [Figure 6](#pbio-0030185-g006){ref-type="fig"}). Notably, the heat shock proteins (HSPs) GP96 and calreticulin have been shown to associate with newly processed cytosolic-derived epitopes, and when injected in vivo, they cross-prime cytotoxic T lymphocytes \[[@pbio-0030185-b28]--[@pbio-0030185-b30]\]. The recent reports of ER--phagosome fusion \[[@pbio-0030185-b31]\] suggest that HSP/peptide complexes may be capable of direct interaction with the DC\'s MHC I; alternatively, HSPs containing a KDEL motif (Lys-Asp-Glu-Leu) may employ a retrograde transport pathway to directly access the ER \[[@pbio-0030185-b24]\]. As ER chaperones within the phagocytosed cell would be bound to the pool of peptides derived from newly synthesized proteins \[[@pbio-0030185-b32]\], they offer the DC an accurate representation of what was being translated immediately prior to death. As an interesting alternative to HSPs, the processed antigen transferred may be the pool of peptides bound to chromatin \[[@pbio-0030185-b33]\]. As the nucleus lacks efficient peptidase activity, antigen may be protected within the nuclear remnants of an apoptotic body. If this were occurring, we predict that loading of the MHC I in the DC would occur in the phagolysosome. Our findings are of particular interest when placed in the context of three recent studies that report that cell-associated whole protein is the primary source of antigen for cross-priming \[[@pbio-0030185-b13]--[@pbio-0030185-b15]\]. We fully appreciate that our study may reflect the choice of a viral model for cross-priming and acknowledge that the experimental details will influence the conclusions. In this light, it is important to consider the differences between the chosen model systems. In the work of Shen and Rock \[[@pbio-0030185-b13]\], lysates prepared from ovalbumin-transfected cell lines were used as a source of antigen, testing different subcellular fractions for their ability to prime animals. This study argues that intact cellular protein, rather than peptides or HSP/peptide complexes, is the main source of antigen for cross-presentation. Considering their use of nitrogen cavitation as the method for disrupting cells, which has been reported to dissociate antigenic peptides from HSP70 and GP96 \[[@pbio-0030185-b34],[@pbio-0030185-b35]\], it would be expected that the HSPs within their lysates would indeed be inert. As a result, this model may have been biased toward the cross-presentation of whole antigen. Wolkers et al. \[[@pbio-0030185-b15]\] demonstrated that peptides present in the secretory domain of nascent proteins are not efficiently cross-presented, while the stable epitope within the mature protein is indeed transferred to APCs. The in vivo studies presented seem to be tracking the cross-presentation of secreted protein, not cell-associated protein. Use of H-2^d^ × H-2^b^ F~1~ mice to demonstrate cross-presentation from P815 cells supports the requirement for the P815 (H-2^d^) to remain alive. An alternate interpretation of their experiments is that soluble proteins produced (in large quantities) by growing tumors resulted in the observed in vivo T cell activation. As a result, there may have been little opportunity for processed cell-associated protein to gain access to a DC. Finally, the study from Norbury et al. \[[@pbio-0030185-b14]\] reported that proteasome substrates (rather than peptides) are critical for achieving antigen transfer for cross-presentation. These studies rely heavily on the use of lactacystin to inhibit proteasome activity. However, while they show the persistence of whole protein, they do not demonstrate that lactacystin prevents the generation of processed peptides in the experimental models used. In the studies described herein, 100 μM lactacystin with a maintenance dose of 1 μM, to inhibit newly synthesized proteasomes (during infection and antigen expression), is required to block the generation of peptides secondary to influenza infection. Given the possibility of newly synthesized proteasomes and/or proteasome-independent processing acting on their artificial constructs for the generation of peptide epitopes, it is critical that functional studies be used to exclude the production of small amounts of processed antigen. In sum, while cell-associated whole protein is important in cross-priming, previous studies have not excluded proteasome products or HSP/peptide complexes as substrates for in vivo cross-priming. As shown here, the apoptotic cell may in fact play a critical role in processing antigen for cross-presentation, in essence preselecting immunologically important antigenic determinants. A comprehensive model accounting for antigen derived from whole protein as well as processed antigens from apoptotic cells is needed to more clearly define the pathways of antigen cross-priming in physiologic (resting) as well as pathologic (stress) situations. Materials and Methods {#s4} ===================== {#s4a} ### Mice. {#s4a1} WT and TAP-1-deficient C57BL/6 mice were purchased from Jackson Laboratory (Bar Harbor, Maine, United States). In all experiments, 4- to 6-wk-old female mice were employed. ### Antibodies, cell lines, and reagents. {#s4a2} All FACS antibodies used in this study were obtained from BD Biosciences Pharmingen (San Diego, California, United States); reagents for the ELISPOT assays were obtained from Mabtech (Stockholm, Sweden). PC3 cells, a human prostate cancer cell line, were obtained from American Type Culture Collection ( ATCC) (Manassas, Virginia, United States). BALB/3T3 cells clone A31 (3T3) were obtained from ATCC. RMA/S, a TAP-deficient T cell lymphoma cell line derived from the Rauscher murine leukemia virus--induced RBL-5 cell line, was employed \[[@pbio-0030185-b36]\]. β~2~m- and TAP-deficient kidney epithelial cells were derived from organ culture followed by a 2-wk in vitro expansion. All cell lines were grown in DMEM containing 10% fetal bovine serum, supplemented with nonessential amino acids, sodium pyruvate, glutamine, 2β-mercaptoethanol, and gentamicin (D-10). Human influenza A/PR/8 was provided as allantoic fluid from Charles River Laboratories (Wilmington, Massachusetts, United States) and used at a 1:3 dilution to infect PC3, 3T3, or RMA/S cells (1,000 HAU/10^6^ cells) or 1:10 dilution to infect DCs (300 HAU/10^6^ cells). Recombinant mouse TNF-α and IL-12 were obtained from R&D Systems (Minneapolis, Minnesota, United States). ### Preparation of antigen-loaded DCs. {#s4a3} Bone-marrow-derived DCs were prepared as previously described \[[@pbio-0030185-b37]\]. In brief, bone marrow obtained from tibia and femurs was lysed of red blood cells and cultured at a density of 3 × 10^6^ cells/well in six-well plates with RPMI containing 10% fetal bovine serum, nonessential amino acids, sodium pyruvate, glutamine, 2β-mercaptoethanol, gentamicin (R-10), and in the presence of GM-CSF (provided by J558L cells transduced with GM-CSF, used 3% vol/vol). Fresh GM-CSF-supplemented medium was added to the wells on days 2, 4, and 6. On day 7, DCs were harvested and plated in fresh wells with or without apoptotic cells. In addition, a maturation stimulus, 80 ng/ml rmTNF-α, was added. To generate influenza-infected apoptotic cells, living cells were first infected with influenza for 1 h at 37 °C in serum-free medium. To allow for expression of viral proteins, 3--5 × 10^6^ infected cells per well of a six-well plate were cultured for 5 h at 37 °C. Cells were washed three times with 3 ml of PBS and were UVB irradiated (120 mJ/cm^2^) in 0.5 ml of PBS, and 0.5 ml of R-10 was added. Cells were allowed to undergo apoptosis for 8--10 h prior to adding 10^6^ immature DCs. Non-adherent cells were harvested 36 h later, and mature DCs were purified to greater than 95% purity using anti-CD11c microbeads and LS^+^ columns (Miltenyi Biotec, Bergisch Gladbach, Germany). DCs were monitored by FACS and found to express high levels of I-A^b^ and CD40. To generate influenza-infected DCs, day 9 mature DCs were infected with influenza for 1 h at 37 °C in serum-free medium. These cells were washed three times in serum containing medium, counted, and used directly. ### In vitro cross-presentation studies. {#s4a4} Four- to six-week-old mice were infected intraperitoneally with 200--300 HAU of influenza A/PR/8--1976 (Charles River, North Franklin, Connecticut, United States). After 2--4 wk, CD4^+^ and CD8^+^ T cells were isolated using MACS purification (Miltenyi Biotec). These cells served as responders in antigen cross-presentation ELISPOT assays. Stimulators included WT or TAP^−/−^ DCs presenting antigen via the endogenous or exogenous pathway. To achieve a ratio of 30 T cells to one DC, 2 × 10^5^ T cells were added to 6.6 × 10^3^ DCs. Cultures were incubated in the plates for 20--36 h at 37 °C, after which cells were washed out of the ELISPOT plates using a mild detergent followed by incubation with 1 μg/ml biotin-conjugated α-interferon-γ (α-IFN-γ) monoclonal antibody (mAb) (BD Biosciences Pharmingen, clone XMG1.2). Wells were then developed using the Vectastain Elite Kit as per manufacturer\'s instructions (Vector Laboratories, Burlingame, California, United States). Colored spots represent IFN-γ-releasing cells and are reported as spot-forming cells per 10^6^ cells. The ELISPOT plate evaluation was performed in a blinded fashion by an independent evaluation service (ZellNet Consulting, Fort Lee, New Jersey, United States) using an automated ELISPOT reader (Carl Zeiss, Thornwood, New York, United States) with KS Elispot 4.3 software. ### In vivo cross-priming studies. {#s4a5} Mice were primed intraperitoneally using influenza A/PR/8-infected apoptotic cells. Ten to 14 d post-immunization, spleens were harvested and CD8^+^ T cells were isolated using MACS purification (Miltenyi Biotec). A total of 2 × 10^5^ T cells were added to 2 × 10^4^ peptide-pulsed DCs or EL4 cells (haplotype-matched cell line) in the ELISPOT plates, pre-coated with 5 μg/ml of a primary α-IFN-γ mAb (Mabtech, clone AN18). Cultures were incubated for 20--36 h at 37 °C, and developed as above. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### DC Cross-Presenting Apoptotic Cells Prime CD8^+^ T Cells In Vivo and Serve As Targets for Effector CD8^+^ T Cells In Vitro \(A) C57BL/6 mice were immunized intra-footpad with 10^5^ DCs cross-presenting apoptotic, influenza-infected PC3 cells, or apoptotic influenza-infected PC3 cells alone. Six days following this single immunization, draining lymph nodes were harvested, CD8^+^ T cells were immediately purified and tested for their ability to respond to syngeneic stimulator cells with or without antigen in a 20-hr IFN-γ ELISPOT assay. SFCs per 10^6^ CD8+ T cells are reported. \(B) As shown in the schematic representation, surface expression of influenza-peptide-loaded MHC I by DCs was monitored using a modified cytotoxicity assay. After charging the DCs with antigen via the endogenous or exogenous pathways, they were loaded with ^51^Cr and used as targets for previously activated influenza-specific cytotoxic T lymphocytes. This assay is designed to evaluate the surface expression of MHC I/pep complexes. DCs were loaded via the exogenous pathway with influenza-infected allogeneic 3T3 cells (filled black squares) or uninfected 3T3 cells (open black squares). Alternatively, DCs were directly infected, thus presenting antigen via the endogenous pathway (filled red squares) or left uninfected (open red squares). After antigen expression or capture of the apoptotic material, respectively, DCs were loaded with ^51^Cr and tested as targets. After 5 h, supernatants were collected and percent cytotoxicity was calculated. Percent cytotoxicity = (experimental-well ^51^Cr release − spontaneous release)/(total release − spontaneous release) × 100. (95 KB EPS). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### TAP^−/−^ DCs Efficiently Phagocytose Apoptotic Cells Immature DCs from C57BL/6 WT or TAP-1-deficient (TAP^−/−^) mice were prepared as above and labeled with the PKH-67-GL (green) fluorescent cell linker (A and B). These cells were next added to PKH-26-GL-labeled (red) apoptotic cells (ACs) for 7 h at a ratio of one DC to five apoptotic cells in the presence or absence of EDTA. FACS Calibur analysis allowed for the detection of double-positive cells (A), indicating that the green DCs captured red apoptotic material. Phagocytosis was calculated as the percent double-positive cells per total population of DCs (A). Samples of DCs alone and apoptotic cells alone were used for setting the parameters of the flow cytometer. The kinetics of phagocytosis were monitored throughout the experiment, and the percent double-positive cells is reported (B). (54 KB PPT). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### RMA/s Cells Express but Do Not Present Influenza Antigen on MHC I \(A) RMA or RMA/S cells were infected with influenza and incubated at 37 °C for 5 h to allow for viral antigen expression. Cells were fixed, permeabilized, and stained using NP mAbs followed by PE-conjugated goat anti-mouse mAb. Analysis was performed on a FACS Calibur, and histograms are shown. \(B) Infected RMA or RMA/s cells were used as stimulators in an ELISPOT assay, testing for their ability to stimulate influenza-reactive CD8^+^ T cells. The influenza A/PR/8 NP~366--374~ peptide restricted for D^b^ was pulsed onto RMA or RMA/s cells and served as a positive control. Values are averages of triplicate wells with error bars indicating standard deviation. (205 KB EPS). ::: ::: {.caption} ###### Click here for additional data file. ::: Protocol S1 ::: {.caption} ###### DCs Capture Apoptotic Cells and Cross-Present Antigen to CD8^+^ T Cells (31 KB DOC). ::: ::: {.caption} ###### Click here for additional data file. ::: Accession Numbers {#s5a1} ----------------- The NCBI Entrez Protein (<http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=Protein>) accession numbers for the gene products discussed in this paper are influenza PA (AAA43619), influenza NP (B36754), and K^d^-restricted immunodominant peptide (HA~210--219~) derived from hemagglutanin (NP040980). The authors would like to thank D. Mithal, H. Morris, and H. Saklani for their technical help. This work was supported by The Pasteur Foundation (NEB), the National Institutes of Health (grant R01 CA85784 to RBD), the Howard Hughes Medical Institute (RBD), the Burroughs Wellcome Fund (MLA and RBD), and INSERM Avenir-AV0201 (MLA). We would also like to thank the reviewer who offered the insight of Heisenberg and the wisdom of Solomon. Competing interests. The authors have declared that no competing interests exist. Author contributions. NEB, RBD, and MLA conceived and designed the experiments. NEB performed the experiments. NEB, RBD, and MLA analyzed the data. RBD and MLA contributed reagents/materials/analysis tools. MLA wrote the paper. Citation: Blachère NE, Darnell RB, Albert ML (2005) Apoptotic cells deliver processed antigen to dendritic cells for cross-presentation. PLoS Biol 3(6): e185. APC : antigen-presenting cell DC : dendritic cell ER : endoplasmic reticulum HAU : hemaglutanin units HSP : heat shock protein IFN-γ : interferon-γ IL-12 : interleukin-12 MHC : major histocompatibility mAb : monoclonal antibody MHC I/pep : major histocompatibility class I/peptide NP : A/PR/8 nucleoprotein PA : A/PR/8 acid polymerase WT : wild-type	PubMed Central	2024-06-05T03:55:55.806198	2005-4-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084338/", "journal": "PLoS Biol. 2005 Jun 26; 3(6):e185", "authors": [ { "first": "Nathalie E", "last": "Blachère" }, { "first": "Robert B", "last": "Darnell" }, { "first": "Matthew L", "last": "Albert" } ] }
PMC1084339	For a cell, suicide is a perfectly acceptable response to stress. Programmed cell death, or apoptosis, is a normal part of development: it occurs 131 times over the course of worm development, for example, and throughout the life of an organism to replenish worn out cells in a variety of tissues. Disease and infection also trigger apoptosis. The dying cell shrinks into a membrane-bound package that attracts the attention of phagocytes, which quickly ingest and degrade the cell\'s contents. Dendritic cells, specialized phagocytes, patrol the body, capturing molecules (or antigens) expressed by a host of pathogens, parasites, environmental substances, and apoptotic cells, and present their findings to the helper and killer T cells that launch an immune response. Before dendritic cells can sound the immune alarm, they must capture antigens, chop them into peptide fragments, and bind the antigenic peptides to molecules of the major histocompatibility complex (MHC). Though dendritic cells can present to helper or killer T cells, it had been thought that the MHC I--bound antigens (recognized by killer T cells) must first enter the dendritic cell\'s cytosol and undergo proteasome degradation. It was also thought that antigenic peptides needed the transporter activity of TAP (transporter associated with antigen processing) proteins to escort them into the cellular compartment where the MHC I molecules reside. A new study by Nathalie Blachère, Robert Darnell, and Matthew Albert lends support to an emerging view that antigen processing may be more flexible and suggests that dendritic cells can generate MHC I peptides that activate killer T cells through cross-presentation. (The phenomenon is called cross-presentation because the pathway crosses established rules of MHC I antigen presentation.) Working with a mouse model to study killer T cell activation in response to influenza antigens derived from apoptotic cells, the authors show that two distinct pathways mediate antigen processing. In the first pathway, antigen processing in dendritic cells depends on TAP, following the established model. But in the second pathway, antigens processed by a dying, infected cell can bypass the dendritic cells\' degradation machinery yet still permit generation of MHC I/peptide complexes that can engage killer T cells. T cells can recognize millions of different antigens because each T cell is outfitted with a unique T cell antigen receptor that recognizes a unique MHC/peptide complex. In this study, the authors used the activation of influenza-reactive T cells as evidence of antigen presentation. To test the hypothesis that dying cells can process antigen for cross-presentation, Blachère et al. limited the dendritic cells\' ability to process antigen by using cells extracted from the bone marrow of mice lacking functional TAP proteins. If the cells could provoke a T cell response, then antigen-specific MHC I/peptide complexes would have to arise through an alternate pathway. When TAP-deficient dendritic cells were directly infected with influenza, they could not elicit a T cell response. But when infected dying cells were internalized by the TAP-deficient dendritic cells, the dendritic cells were able to cross-present influenza antigen and activate the T cells. To investigate the mechanism of cross-presentation, the authors generated cells with deficient proteasomes that could not degrade viral proteins, and also used another cell line inhibited in TAP activity. When either of these cell types were infected with influenza, caused to undergo apoptosis, and engulfed by dendritic cells, it became clear that the dendritic cells needed functional TAP to successfully present the influenza antigen to T cells. The authors then went on to validate the operation of these pathways in vivo.[](#pbio-0030197-g001){ref-type="fig"} ::: {#pbio-0030197-g001 .fig} ::: {.caption} ###### Antigenic proteins in the cytosol are degraded by the proteasome and transported by TAP into the MHC1 compartment for MHC1 loading (Image: Giovanni Maki) ::: ![](pbio.0030197.g001) ::: Altogether, these results point to two MHC I pathways: one relies on transporter activity in the dendritic cell; the other allows dendritic cells to capture antigens already processed in the dying cell. Though the details of the cross-presentation pathway remain unclear, the conditions that give rise to it are not: when dendritic cells are directly infected, they mediate antigen processing. When dendritic cells ingest already infected, apoptotic cells, the dying cells deliver processed antigen and set the stage for cross-presentation. It\'s possible that stressful conditions, like an influenza infection, trigger this cross-presentation pathway, allowing the dying cell to alert the immune system to a pathogenic agent that requires immediate action.	PubMed Central	2024-06-05T03:55:55.809150	2005-4-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084339/", "journal": "PLoS Biol. 2005 Jun 26; 3(6):e197", "authors": [] }
PMC1084341	Background ========== Ketamine was patented in 1966 \[[@B1]\], and has long been known to be associated with short-term analgesia \[[@B2]\]. Considerable interest was renewed in ketamine with the discovery that it could block the NMDA receptor and therefore it has a potential role in the management of windupand prevention of subsequent spinal cord sensitisation. To date, clinical trials that have investigated its use as an analgesic drug have often described its adverse effects. This has led some authors to question its use in the management of postoperative pain \[[@B3]\]. Several animal studies have suggested that the mechanisms for a synergistic interaction between ketamine and opioids might exist \[[@B4]\] and \[[@B5],[@B6]\]. that combinations of opioids and NMDA receptor antagonists might result in an enhanced effect \[[@B7]\] -- as might be predicted by the different mechanisms of action of these classes of drugs \[[@B8],[@B9]\]. The current investigation explored the interaction between ketamine and the opioid fentanyl in the anticipation that a low dose of ketamine might potentiate the analgesic effect of fentanyl. Furthermore, it was hypothesised that the interaction of these drugs might be associated with selective potentiation of analgesia without associated increased sedation; that is that potentiation might occur in the context of a very low dose of ketamine that was not otherwise associated with brain effects such as sedation. It was hoped that the identification of such doses of ketamine may enable better future management of both opioid sensitive physiological pain and NMDA receptor mediated sensitisation without the disadvantage of increased sedation. Methods ======= This study was conducted using a double blinded, randomised, placebo controlled, crossover methodology to determine whether a low dose of ketamine potentiated the antinociceptive effect of fentanyl without potentiating the sedative effect of fentanyl. A battery of tests was assembled to assess both nociception and sedation. Tests of nociception used electrical current, pressure, and thermal stimuli. Sedation was assessed by a subjective and objective score in addition to psychometric tests. Saline was used as the control and propofol was used to validate the tests of sedation. Cardiovascular and respiratory parameters were also monitored in order to detect the occurrence of adverse events. This investigation was approved by the Southern Health Human Research and Ethics Committee (Project number 96022A and 97074A) in accordance with the guidelines of the National Health and Medical Research Council, Australia (NHMRC). Ten healthy male volunteers were recruited via bulletin board advertisements. The volunteers were trained in the test procedures employed and medically screened. Volunteers were excluded if they had a history of cardiac, neurological, or musculoskeletal disease. Other exclusion criteria included a history of drug abuse, pain syndromes, myasthenia gravis, acute narrow angle glaucoma, asthma, or heart failure, concurrent use of any analgesics, sedatives, erythromycin, MAO inhibitors, or allergy to propofol, fentanyl, or ketamine. The ten volunteers each attended five three-hour laboratory sessions on separate occasions. In each session, the volunteer received either one of the drug treatments or saline (Table [1](#T1){ref-type="table"}). Therefore, each volunteer was exposed to each of the five treatments, over five sessions, with the order of treatment randomised for each volunteer. During each session, the test battery was performed prior to drug administration as a measure of \'baseline\' and then repeated when each of the four targeted concentrations were reached. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Drug Concentrations Targeted ::: Baseline Concentration 1 Concentration 2 Concentration 3 Concentration 4 -------------------- ---------- ----------------- ----------------- ----------------- ----------------- Placebo (saline) \- \- \- \- \- Propofol (μg/ml) 0.00 0.15 0.30 0.60 0.90 Ketamine (ng/ml) 0.00 15.00 30.00 60.00 120.00 Fentanyl (ng/ml) 0.00 0.20 0.40 0.80 1.20 Ketamine (ng/ml) & 0.00 30.00 30.00 30.00 30.00 Fentanyl (ng/ml) 0.00 0.20 0.40 0.80 1.20 ::: The orders of tests within the test battery were not varied. The physiological measures were conducted first, followed by the sedation tests, and lastly the nociception tests. Before and after each test battery was performed, a blood sample was taken to establish the drug serum concentration. The duration of each test battery was approximately 20--30 minutes, and between each battery the volunteer was instructed to rest for approximately 20 minutes while the drug serum concentration was increased according to the administration protocol. The drugs were administered by serum target controlled intravenous infusions (Stanpump; Shafer, CA. 94304, USA, revision November 5, 1996; and a Harvard 22 syringe pump). This method was utilised to maintain a stable serum concentration for the duration of the test battery. Two identical computer and syringe pump systems using opaque intravenous tubing were employed and operated in parallel at all times. The syringe pumps were hidden from both the investigator and the volunteer within an opaque sound-proofed box. Two intravenous catheters (20--22 gauge) were inserted -- one for the infusion of the study drug and the other for the withdrawal of blood for serum concentration assay. Blood samples were transferred to a SST gel and clot activator vacutainer (Becton Dickinson, Franklin Lakes, NJ, USA) and allowed to clot. Immediately after clotting, the samples were centrifuged at 3500 rpm for 10 minutes. The serum was then frozen in liquid nitrogen before being stored in a refrigerator -- fentanyl and ketamine samples were stored at -4°C whereas propofol samples were refrigerated at -20°C. A scientist who was blinded to the serum concentrations targeted conducted the analysis of the serum samples. Propofol was analysed using a method similar to Plummer \[[@B10]\]. Fentanyl and ketamine concentrations were analysed using a method based on that of Bjorkman and Stanski. \[[@B11]\]. The pain threshold to electrical current was determined using a computer controlled constant current stimulator (Amlab International Pty Ltd, NSW 2113, Australia). A train of five 1 ms unipolar rectangular pulses, at a frequency of 200 Hz, lasting 25 ms was delivered using single-use disposable Silver-silver chloride electrodes (9013S0241, Medtronic Dantec, NSW, Australia) with a contact area of 0.54 cm^2^applied 2 cm apart to the medial non-dominant wrist. The increasing and decreasing staircase method was used to determine the electrical pain threshold, which was defined as the minimum amount of current resulting in a stimulation that was graded as \"painful\". The mean of three consecutive measurements was used in subsequent analysis. The pain threshold to contact heat was determined using the ascending ramp method (Somedic Thermotest, Somedic AB, Sweden; applied to the volunteer\'s non-dominant wrist). A contact thermistor (Hewlett Packard patient monitor M1165A, Model 54S Mass. 02254, USA) was used to monitor the temperature of the volunteer\'s skin (adjacent to the wrist on the non-dominant forearm) during the session. The pain threshold to pressure was measured using the ascending ramp method (Somedic Algometer, Somedic, Sweden; applied to the non-dominant middle-finger nail bed). Pressure was applied at the increasing rate of 40 kPa/s over a contact surface area of 1 cm^2^. For both tests, the mean of five consecutive measurements was used in subsequent analysis. The volunteers were asked to rate the symptom of sedation with a visual analogue score by placing a vertical mark through a horizontal 100 mm line with a pencil. The caption \"I feel drowsy\" was printed above the horizontal line and the left and right ends of the line were labelled with the statements, \"not at all\" and \"extremely\" respectively. The investigator -- who was unaware of the nature of the drug treatment -- assessed the volunteers level of sedation using the Observer Assessment of Alertness/Sedation Scale (OASS) \[[@B12]\]. These scores were measured once during each test battery. The Symbol Digit Modalities Test (SDMT) is a pen and paper test in which nine symbols are paired with digits and 110 blank squares associated with digits are required to be filled in within 90 seconds \[[@B13]\]. Parallel forms were used in random order. The number of errors that were made during the SDMT test was recorded and an incidental recall task immediately following the primary test was performed. For this, the volunteers were given a new sheet composed of a line of 15 symbols in which all nine symbols were included at least once. The volunteer was then asked to fill in the number associated with the symbol. Where a symbol appeared more than once, and the volunteer correctly identified the number on one occasion and incorrectly on another occasion, credit for the correct identification was given. A simple auditory reaction time was measured by computer (Amlab International Pty Ltd, NSW 2113, Australia). In this test, the volunteer pushed a micro-switch button in response to a tone using a handset in the dominant hand. The Finger Tapping test was included to assess the effect of the study drugs on the motor nervous system. The volunteers were asked to tap a micro-switch contained within a modified computer serial mouse (Microsoft Corporation, USA) as rapidly as possible over ten seconds and the number of taps were counted. For both of these tests the mean of five consecutive measurements was used in the subsequent analysis. During each test battery a Hewlett Packard patient monitor (M1165A, Model 54S, Mass. 02254, USA) was used to measure the volunteers\' blood pressure, pulse rate, and pulse oximetry. In addition, a small-volume circuit consisting of a mouthpiece, Wrights Respirometer, and an in-line Capnometer sensor (Hewlett Packard, Model 14360A, Mass. 02254, USA) was used to determine the respiratory rate, tidal volume (averaged from ten consecutive breaths), and the end-tidal carbon dioxide. The minute volume was calculated from the tidal volume and the respiratory rate. The results were analysed by repeated-measures analysis of variance (ANOVA), incorporating the Greenhouse-Geisser adjustment for multisample asphericity (compound asymmetry). For each outcome variable the analysis was conducted across the five treatments (placebo, propofol, ketamine, fentanyl, and fentanyl and ketamine combined), and over the five ascending doses. Within the ANOVA, the hypothesis tested was that provided by the within-subject interaction term Drug by Dose. The result of testing this factor was that this method tested for parallelism of the response curves across ascending dosesaccording to drug treatment. Thus, it identified differences between the drugs with respect to the slope or profile of each dose response curve. Because the above hypothesis was tested for each of the 18 outcome measures subjected to ANOVA, the familywise Type I error-rate was controlled by the Ryan-Holm step-down Bonferroni procedure. A value of P≤ 0.05 was regarded as statistically significant. Further analysis and illustration was confined to those dose response curves that were significantly different between the study drugs. When a significant difference between the dose response curves was determined, the dose response curves were ranked according to similarity in order to establish the origin of the significant difference. This was achieved by a stepwise addition of each dose response curve until a significant difference was detected by ANOVA testing. The OAAS scores were analysed by Friedman\'s test because of the lack of variance of the OAAS scores at baseline. The profile of the response curve of each volunteer to the increasing drug concentrations was summarised by the sum of the OAAS scores across the five concentrations of each drug. The summary of each drug profile was then blocked by the term volunteer. Where a significant difference between the drug profiles was detected, the profiles were classified according to similarity in order to establish the origin of the significant difference. This was achieved by a stepwise addition of profile data with similar appearances until a significant difference was detected by the Friedman test in a similar manner as previously described with ANOVA. Statistical calculations were conducted using Minitab (Release 13.31) for the Friedman Test and SPSS for Windows (Release 10.0.7) for all other procedures. Results ======= Ten volunteers completed all of the experiments and none were excluded or withdrawn. The mean age, weight, and height (standard deviation) of the group were 23 (4.2) years, 66.3 (4.6) kg, and 173.5 (33.5) cm. The study drugs had a dose-dependent effect on the electrical current and pressure pain, SDMT, subjective sedation VAS and OASS, reaction time, and finger-tapping tests (Table [2](#T2){ref-type="table"} and Table [3](#T3){ref-type="table"}). The remaining tests did not demonstrate a difference between the study drugs and placebo. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Differences between the Profiles of the Study Drug Dose Response Curves ::: Outcome Variable Pvalue ------------------------------------------- ---------- ---- Nociception Tests Electrical pain threshold \<0.01 \* Pressure pain threshold \<0.01 \* Heat pain threshold 0.99 Skin temperature 0.59 Sedation Score Subjective sedation visual analogue score 0.01 \* OASS \<0.01 \* Psychometric Tests SDMT 0.03 \* SDMT errors 0.17 SDMT recall 0.99 Reaction time \<0.01 \* Finger tapping test \<0.01 \* Physiological Tests Systolic blood pressure 0.34 Diastolic blood pressure 0.99 Heart rate 0.99 Respiratory rate 0.99 Tidal volume 0.73 Pulse oximetry 0.99 End tidal carbon dioxide 0.99 Minute volume 0.99 SDMT denotes symbol digit modalities test; OASS, observer assessment of alertness/sedation scale; \, P≤ 0.05 ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Origins of the Significant Differences between the Study Drugs with Respect to the Profile of the Dose Response Curves (P ≤ 0.05) ::: Outcome Variable Origins of Difference in Profile by Drug --------------------------- ------------------------------------------------------------------------ Nociception Tests* Electrical pain threshold fentanyl & ketamine versus ketamine, fentanyl versus saline Pressure pain threshold fentanyl & ketamine, fentanyl, versus ketamine, saline Sedation Score Sedation visual analogue ketamine, fentanyl, propofol, fentanyl & ketamine versus saline OASS propofol versus ketamine, fentanyl, fentanyl & ketamine, versus saline Psychometric Tests SDMT propofol versus ketamine, fentanyl, fentanyl & ketamine versus saline Reaction time propofol versus ketamine, fentanyl, fentanyl & ketamine, saline Finger tapping test propofol versus ketamine, fentanyl, fentanyl & ketamine, saline SDMT denotes symbol digit modalities test; OASS, observer assessment of alertness/sedation scale. ::: All drugs resulted in a dose dependent increase in the electrical pain threshold compared with placebo (Figure [1](#F1){ref-type="fig"}). The dose response curve of fentanyl combined with ketamine was markedly steeper than the dose response curves of fentanyl alone (P\< 0.05; ANOVA). While a ketamine serum concentration of 30 ng/ml did not result in a change in electrical pain threshold when administered alone (P= 0.32; two-way ANOVA), when it was added to a fentanyl serum concentration of 0.4 ng/ml the combination resulted in greater increase in pain threshold than that of fentanyl administered alone (P\< 0.01; two-way ANOVA). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### The Antinociceptive Effect Measured using ECT. The electrical pain threshold is expressed as the mean electrical pain threshold of all volunteers (n= 10) when a steady serum concentration of the treatment drug had been achieved. These values are standardised by the threshold value at baseline in order to illustrate the difference in profiles between each treatment arm. The serum concentrations targeted are listed in Table 1 and enumerated on the abscissa using a logarithmic scale. ::: ![](1471-2253-5-2-1) ::: Fentanyl, both alone and in combination with ketamine, produced a dose dependent increase in pressure pain threshold compared with saline, whereas ketamine alone was ineffective at all doses (Figure [2](#F2){ref-type="fig"}). The dose response profiles of fentanyl alone and in combination with ketamine were not different (P= 0.35; ANOVA). No difference was seen between the study drugs (including placebo) when assessed with heat (Figure [3](#F3){ref-type="fig"}). ::: {#F2 .fig} Figure 2 ::: {.caption} ###### The Antinociceptive Effect Measured using Pressure Algometry. The pressure pain threshold was expressed as the mean pressure pain threshold of all volunteers (n= 10) when a steady plasma level of the treatment drug had been achieved. These values were then standardised by the threshold value at baseline in order to illustrate the difference in profiles between each treatment arm. The serum concentrations targeted are listed in Table 1 and enumerated on the abscissa using a logarithmic scale. ::: ![](1471-2253-5-2-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### The Antinociceptive Effect Measured using Heat. The heat pain threshold was expressed as the mean heat pain threshold of all volunteers (n= 10) when a steady plasma level of the treatment drug had been achieved. These values were then standardised by the threshold value at baseline in order to illustrate the difference in profiles between each treatment arm. The serum concentrations targeted are listed in Table 1 and enumerated on the abscissa using a logarithmic scale. ::: ![](1471-2253-5-2-3) ::: All of the study drugs were associated with increased sedation compared with placebo (Table [1](#T1){ref-type="table"}) when assessed by the objective psychometric tests (Figure [4](#F4){ref-type="fig"}, [5](#F5){ref-type="fig"}, and [6](#F6){ref-type="fig"}) and the scores of sedation (Figure [7](#F7){ref-type="fig"} and [8](#F8){ref-type="fig"}). Propofol was associated with a more marked subjective sedative effect than either fentanyl or ketamine -- alone or in combination (Figure [7](#F7){ref-type="fig"}Error! Reference source not found.). There was no difference between the sedative effect of fentanyl and fentanyl in combination with ketamine as assessed by the subjective sedation VAS, OASS, and SDMT (Figure [4](#F4){ref-type="fig"}Error! Reference source not found.). Specifically, fentanyl 0.4 ng/ml combined with ketamine 30 ng/ml was not associated with increased sedation in comparison with fentanyl 0.4 ng/ml alone. While Propofol was associated with a reduced number of finger taps and an increase in reaction time compared with placebo, none of the other study drugs had any significant effect on these measures of sedation. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### The Psychometric Effect Measured by SDMT The SDMT. score was expressed as the mean score of all volunteers (n= 10) when a steady serum concentration of the treatment drug had been achieved. These values are standardised by the threshold value at baseline in order to illustrate the difference in profiles between each treatment arm. The serum concentrations targeted are listed in Table 1 and enumerated on the abscissa using a logarithmic scale. ::: ![](1471-2253-5-2-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Psychometric Effect Measured by Reaction Time. The reaction time was expressed as the mean of all volunteers (n= 10) when a steady serum concentration of the treatment drug had been achieved. These values are standardised by the threshold value at baseline in order to illustrate the difference in profiles between each treatment arm. The serum concentrations targeted are listed in Table 1 and enumerated on the abscissa using a logarithmic scale. ::: ![](1471-2253-5-2-5) ::: ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Psychometric Effect Measured by the Finger Tapping Test. The number of finger taps was expressed as the mean of all volunteers (n= 10) when a steady serum concentration of the treatment drug had been achieved. These values are standardised by the threshold value at baseline in order to illustrate the difference in profiles between each treatment arm. The serum concentrations targeted are listed in Table 1 and enumerated on the abscissa using a logarithmic scale. ::: ![](1471-2253-5-2-6) ::: ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Subjective Sedation Measured by Visual Analogue Score. This figure shows the effect of the five treatments on the visual analogue score (VAS) given to the statement, \"I feel drowsy\". The VAS was expressed as the mean score of all volunteers (n= 10) when a steady serum concentration of the treatment drug had been achieved. All drug treatments were associated an increase in VAS compared to saline. The serum concentrations targeted are listed in Table 1 and enumerated on the abscissa using a logarithmic scale. ::: ![](1471-2253-5-2-7) ::: ::: {#F8 .fig} Figure 8 ::: {.caption} ###### Sedation Measured by OASS. The OASS was expressed as the median score of all volunteers (n= 10) when a steady serum concentration of the treatment drug had been achieved. The serum concentrations targeted are listed in Table 1 and enumerated on the abscissa using a logarithmic scale. ::: ![](1471-2253-5-2-8) ::: The serum concentrations remained steady at each concentration targeted and the ketamine serum concentration remained steady throughout the duration of the experiment when it was combined with fentanyl in a fixed dose (Table [4](#T4){ref-type="table"}). ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Measured Serum Drug Concentrations ::: Baseline Concentration 1 Concentration 2 Concentration 3 Concentration 4 ------------------ ---------- ----------------- ----------------- ----------------- ----------------- --------- ------- --------- ------- --------- Propofol (μg/ml) 0.00 (0.00) 0.10 (0.04) 0.26 (0.07) 0.58 (0.18) 0.92 (0.24) 0.11 (0.03) 0.31 (0.11) 0.62 (0.16) 0.91 (0.25) Ketamine (ng/ml) 0.00 (0.00) 11.49 (4.31) 24.64 (6.68) 47.87 (18.03) 85.75 (31.87) 12.61 (4.85) 27.48 (11.62) 51.86 (21.54) 93.88 (38.26) Fentanyl (ng/ml) 0.00 (0.00) 0.28 (0.03) 0.53 (0.09) 1.01 (0.15) 1.48 (0.26) 0.31 (0.07) 0.55 (0.13) 1.02 (0.21) 1.53 (0.25) Ketamine (ng/ml) 0.00 (0.00) 25.59 (13.26) 30.29 (13.94) 30.73 (16.98) 29.88 (12.84) 25.41 (5.76) 35.41 (15.90) 29.82 (12.26) 31.58 (14.23) & Fentanyl (ng/ml) 0.00 (0.00) 0.27 (0.11) 0.49 (0.13) 1.05 (0.17) 1.57 (0.26) 0.29 (0.06) 0.57 (0.15) 1.09 (0.24) 1.57 (0.29) SD denotes standard deviation. The upper and lower data in each cell represents the serum concentration measured at the beginning and at the end of each test battery respectively. ::: Discussion ========== Pre-clinical investigation of the antinociceptive effect of ketamine and morphine by Chapman and Dickenson \[[@B8]\] led Schmid, Sandler and Katz \[[@B14]\] to hypothesize that there may be a dose of ketamine that has no analgesic potency on its own, but when used in combination with an opioid might produce superior pain relief than either drug alone. To test this hypothesis, the present study constructed a dose response curve for ketamine, from which a dose of ketamine that had no antinociceptive effect was identified (30 ng/ml). When this dose of ketamine was combined with fentanyl the resulting antinociception, as assessed by electrical current pain threshold, was greater than that of either drug alone and clear potentiation was demonstrated. In addition, while ketamine potentiated the antinociceptive effect of fentanyl, the sedative effect was not increased. Therefore, the data presented here supports the hypothesis put forward by Schmid et al\[[@B14]\]. A review of the recent preclinical work investigating the interaction between opioid agonists and NMDA antagonists shows mixed results. Some authors have demonstrated that NMDA antagonists enhance the antinociceptive effect of opioids \[[@B15],[@B16]\]. whereas others have been unable to confirm these findings \[[@B17],[@B18]\]. Some authors such as Redwine et al\[[@B19]\] have gone so far as to suggest that NMDA receptors are not involved in acute opiate mediated analgesia. In order to resolve these difficulties, a number of investigators have suggested that factors other than receptor interactions may have resulted in the heterogeneous results seen. These factors may include; the opioid agonist and the NMDA antagonist chosen for investigation \[[@B19]-[@B23]\], the nociceptive test chosen \[[@B16]\], and the species used for investigation \[[@B24]\]. The findings of this current investigation support the hypothesis that the nociceptive test used is an important determinant of the outcome of investigations assessing the interaction between ketamine and fentanyl, that is, the interaction between ketamine and fentanyl is stimulus dependent. Despite fentanyl-mediated analgesia being clearly potentiated by ketamine when assessed by ECT, this was not seen when it was assessed by pressure algometry. The inability of pressure to demonstrate potentiation was consistent with other investigators who have found that the antinociceptive activity of ketamine on pressure nociception is dependent on preconditioning, such as preconditioning which produces neuropathic pain \[[@B25]\]. Clearly such preconditioning was absent in the current study. This study encountered similar difficulties to that of Sethna et alwho found that a large variation in the assessment of C-fibre mediated pain made capsaicin -- and in our case heat -- a difficult modality to use in the study of nociception \[[@B26]\]. The global failure of all of the drug infusions to increase the heat pain thresholds suggests that either the assessment of heat pain threshold was inadequate or the drug doses used were too low. The heat pain thresholds were assessed by the ascending ramp method with a continuous stimulus, which increased intensity at a constant rate. This method was more susceptible to provoking conditioned timed responses in the volunteers compared with the discrete stimuli used for the assessment of electrical current pain thresholds. Other factors that weakened the assessment of heat pain thresholds included the mechanical stimulation caused by contact of the thermode, which may produce Aβ-fibre mediated interference of the C-fibre sensory input. The authors hypothesize that the use of discrete heat stimuli -- such as that produced by laser \[[@B27]\] -- may have improved heat stimulation\'s ability to demonstrate the nature of the interaction between fentanyl and ketamine using this paradigm. Clinical studies assessing the interaction of ketamine and opioids have also produced mixed results however, consistent with the findings of the current study, a review of the interaction between ketamine and opioids in alleviating clinical pain has demonstrated that ketamine enhances opioid-mediated analgesia \[[@B28]\]. Schmid et alin their review of the clinical use of ketamine for the management of postoperative pain differentiated between a high and a low dose range of ketamine \[[@B14]\]. Both the sedative and the psychotomimetic effects of ketamine are known to be dose dependent \[[@B29]\]. As opposed to idiosyncratic drug effects, the dose responsive nature of the unwanted sedative and psychotomimetic effects of ketamine provided the opportunity that a low dose of ketamine could be used to provide analgesia without their occurrence. However, due to the large overlap in the respective ketamine dose response curves for analgesia and sedation, separation of these effects was difficult when ketamine was used alone. Clinical studies using ketamine alone for analgesia have often reported adequate analgesia with associated psychotomimetic effects, or have reported an acceptable level of adverse effects but poor analgesia \[[@B30]-[@B33]\]. The primary advantage of combinational therapy relies on the improvement of the desired effect without a concomitant increase in its adverse effects, which some authors have seen as a requirement for the demonstration of a synergistic interaction \[[@B26]\]. That is, it is considered that a synergistic interaction exists between two drugs when the effect of a combination is greater than the sum of the effects produced by each drug given alone. However, it is clear that whilst an increase of the desired effects is desirable, what is critical for clinical advantage is that the occurrence of adverse effects is not increased in a similar manner. Some authors consider that an increase in the beneficial effects of a combination without an increase in the adverse effects is more important than the existence of synergy per se\[[@B34]\]. Human dose ranging studies of ketamine are scarce and consequently the doses and methods of administration of ketamine in clinical studies are numerous. Schmid et alproposed that dose ranging studies were required as part of a research program to explore the clinical usefulness of ketamine as an adjunct to opioid analgesia in 1999 \[[@B14]\]. In this study, the highest serum concentrations of ketamine targeted were related to regimens that have been used for analgesia clinically \[[@B35]-[@B37]\]. While the human laboratory paradigm is one step removed from the clinical arena, it is well suited to dose ranging studies that are vital for the rational planning of clinical investigations. Clinical and human laboratory studies investigating ketamine for the management of postoperative pain are heterogenous in design and varied in their appraisal of the effect of ketamine in conjunction with opioids for the management of pain \[[@B38]-[@B52]\]. In these trials ketamine has been given by the epidural, intramuscular, and intravenous routes, in addition to the description of intrathecal use \[[@B53]-[@B59]\]. Javery et aldescribed reduced pain ratings and morphine consumption following microdiscectomy when patients received a combination of ketamine and morphine rather than morphine alone \[[@B44]\]. By contrast most studies have investigated pain following abdominal surgery and the results have been less encouraging \[[@B45],[@B46]\]. While these studies may yet lead to the conclusion that the surgical procedure conducted is important in determining whether ketamine augments postoperative opioid analgesia or not; other factors such as the dosing regimen also varies across these studies and appears to be a dominant reason for such variable results. Studies that have combined ketamine with morphine in an intravenous Patient Controlled Analgesia (PCA) pump have usually failed to demonstrate a reduction in morphine consumption \[[@B45]-[@B48]\]. This may reflect the mismatch of ketamine\'s and morphine\'s respective time response curves for analgesia. Compared with morphine, ketamine is a short acting drug. Ketamine\'s short duration of action was demonstrated when it reduced ischaemic arm pain for less than 5 and 10 minutes respectively when used intravenously in doses of 125 μg/kg and 250 μg/kg \[[@B2]\]. Therefore, if the use of the PCA pump is predominantly determined by the longer acting morphine, the combination of ketamine and morphine may result in troughs of ketamine\'s serum concentration, and occasional peaks when the PCA pump is triggered. This is consistent with the finding that ketamine co-administered with morphine in PCA pumps did not improve patients\' VAS in six studies overall when assessed by weighted mean difference \[[@B28]\]. Therefore, because ketamine has a small therapeutic window within which analgesia is not accompanied by adverse effects, great importance should be placed on the dosing regimen chosen. Some studies have used steady intravenous infusions of ketamine in addition to PCA pump morphine to relieve abdominal postoperative-pain. Both manual infusions and computer controlled serum targeted infusions of ketamine have been studied \[[@B38]-[@B41],[@B49]-[@B52]\]. However, while it may be expected that a steady manual infusion should result in a stable serum concentration of ketamine, this has not always been shown to be the case. Owen et alstudied a regimen of ketamine consisting of a bolus and an infusion which lasted 24 hours \[[@B39]\]. The subsequent analysis of the serum concentrations showed that a constant serum concentration was achieved in fewer than half of the thirty patients studied and that in the majority of patients the serum concentrations rose continuously. Therefore, while the published literature describing the use of ketamine suggests that the dose of ketamine needs to be carefully controlled, a manual infusion may not always achieve the level of stability required. As dosage is important in avoiding ketamine\'s side effects and ensuring adequate NMDA receptor antagonism, the potentiation of analgesia resulting from the combination of ketamine with fentanyl that was found in this study is likely to require a similar attention to dosing. In keeping with this suggestion, Adriaenssens et alhave shown improved pain ratings and decreased morphine consumption following laparotomy by targeting a ketamine serum concentration of 100 ng/ml using the same pharmacokinetic software used in this study \[[@B38]\]. This present study suggests that ketamine may provide enhanced opioid analgesia at a third of this dose without adverse effects. Notwithstanding this, in contrast to ketamine administered in PCA pumps, ketamine co-administered with morphine infusions have been shown to improve patients\' VAS in seven studies overall when assessed by weighted mean difference \[[@B28]\]. An advantage of human laboratory investigation is the ability to use greater number of drug doses and combinations than is usually practical in clinical investigations. There are few barriers to applying the knowledge gained from this study to present day clinical medicine. Doses and serum concentrations of ketamine have been identified using a freely available pharmacokinetic program and using the racemic form of ketamine, which is licensed for clinical use internationally. The findings of this study and those from published clinical studies suggest that accurate and consistent administration of low doses of ketamine are required in order to enhance the antinociceptive effect of opioids and avoid adverse effects. Conclusion ========== The effect of intravenous ketamine and fentanyl was studied in human volunteers using nociceptive and sedative tests. A serum concentration of ketamine that did not alter indices of sedation potentiated the antinociceptive effect of fentanyl. This potentiation of antinociception occurred without an increase in sedation suggesting that low doses of ketamine might be combined with μ opioid agonists to improve their analgesic effect in a clinical setting. Abbreviations ============= ANOVA analysis of variance MAO Monoamine oxidase NHMRC National Health and Medical Research Council, Australia NMDA N-methyl-D-aspartate OASS Observers Assessment of Alertness and Sedation score PCA patient controlled analgesia SD standard deviation SDMT Symbol Digit Modalities test SPSS statistical package for the social sciences VAS visual analogue score Competing interests =================== The authors have not received reimbursements, fees, funding, or salary from an organisation that may in any way gain or lose financially from this publication, nor is such an organisation financing the article-processing charge for this article. The authors do not hold stocks or shares in an organisation that may gain or lose financially from the publication of this paper. Authors\' contributions ======================= AT, RN, and CG conceived of the study. AT designed and coordinated the study, custom-built equipment and wrote the software, conducted the experimental sessions, performed the data analysis, and wrote the manuscript. YIK assisted with the conduct of the study. All authors read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2253/5/2/prepub> Acknowledgements ================ The authors gratefully acknowledge Tattersalls for their financial support of the Monash University Department of Anaesthesia.	PubMed Central	2024-06-05T03:55:55.809803	2005-4-2	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084341/", "journal": "BMC Anesthesiol. 2005 Apr 2; 5:2", "authors": [ { "first": "Adam P", "last": "Tucker" }, { "first": "Yong Ik", "last": "Kim" }, { "first": "Raymond", "last": "Nadeson" }, { "first": "Colin S", "last": "Goodchild" } ] }
PMC1084342	Background ========== DNA cytosine-5-methyltransferases (Dnmts) catalyze the methyl transfer from S-adenosyl-L-methionine to the cytosine in CpG dinucleotides. Four active DNA methyltransferases, Dnmt1, Dnmt2, Dnmt3a, and Dnmt3b, have been reported in mammals. These four enzymes contain highly conserved DNA methyltransferase motifs; however, genetic and biochemical studies have suggested functional differences between them. Dnmt1 is considered a maintenance DNA methyltransferase, though some de novo methylation activity has been described in vitro \[[@B1]\]. Dnmt3a and Dnmt3b are generally regarded as de novo DNA methyltransferases, even though it has been proposed that these enzymes may play a role in the maintenance of methylation in ES cells \[[@B2]\]. While the biological role of Dnmt2 remains unknown, it has been reported to methylate DNA at a very low level, and it may be involved in non-CpG methylation \[[@B3]-[@B6]\]. De novo methylation mostly occurs during embryonic development and plays an important role in genomic imprinting, X-inactivation, and transposon suppression in mammals. De novo methylation has been described to occur frequently in cancer \[for a review, see \[[@B7]\]\] and at sites of integrated exogenous DNA \[[@B8]\]. It has been proposed that de novo methyltransferases may participate in the maintenance of methylation based on the findings that Dnmt3a and/or Dnmt3b may play roles in restoring methylation at sites missed by Dnmt1 during replication in ES cells \[[@B2]\]. A Dnmt1-deficient cancer cell line showed little loss of DNA methylation, suggesting that the de novo methyltransferases can maintain a high level of DNA methylation in the absence of Dnmt1 in neoplastic cells \[[@B9]\]. A more dramatic loss of DNA methylation was observed when Dnmt1 and Dnmt3b were both absent from that cell line, suggesting the possible cooperation between these two enzymes \[[@B10]\]. Another study using antisense and RNAi strategies suggested that Dnmt1 is responsible for global and CpG island methylation in cancer cells \[[@B11]\]. These studies raise the possibility that Dnmt1 and the two de novo methyltransferases may work together in maintaining, as well as establishing, DNA methylation patterns. Dnmt3a has been shown to have biochemical characteristics different than Dnmt1. Dnmt3a shows no preference for hemimethylated over fully unmethylated DNA substrates, whereas Dnmt1 has a strong preference for hemimethylated DNA \[[@B12],[@B13]\]. Dnmt1 has no known preference other than CpG sites in vitro \[[@B14]\], whereas Dnmt3a appears to prefer CpG sites flanked by pyrimidines in vitro \[[@B15]\] and shows less specificity for CpG sites than Dnmt1 both in vitro and in vivo \[[@B13],[@B16]\]. Dnmt3a generates asymmetrical methylation pattern on two DNA strands in vitro \[[@B15]\]. These differences reflect the methylation maintenance function of Dnmt1 and de novo methylation function of Dnmt3a. It has been reported that pre-existing cytosine methylation at CpG as well as at non-CpG sites can stimulate de novo methylation activity of native and recombinant Dnmt1 on single-stranded oligonucleotides in vitro \[[@B1],[@B17]\]. It has been hypothesized that single-stranded DNA formed during replication or repair might serve as a nucleation site for de novo methylation by Dnmt1 \[[@B17]\]. Purified murine Dnmt1 overexpressed in E. colialso showed increased de novo methylation activity when pre-existing cytosine methylation was present on double-stranded oligonucleotide substrates \[[@B18]\]. In another study, introduction of random CpG methylation into plasmid DNA stimulated methylation activity of DNA methyltransferases partially-purified from human placenta and murine liver \[[@B19]\]. A more recent study showed that fully methylated oligonucleotides can also stimulate Dnmt1 de novo methylation activity in trans \[[@B20]\]. These studies demonstrate that pre-existing cytosine methylation on oligonucleotides as well as on plasmids can stimulate the de novo methylation activity of Dnmt1 in vitro, even though this activity has not been confirmed in vivo. This stimulation may be due to the allosteric activation of the catalytic domain by the binding of Dnmt1 to methylated DNA \[[@B20]\]. In our previous studies \[[@B21]-[@B23]\], de novo methylation of the episome by endogenous Dnmt1 has not been detected in human cells regardless of the methylation status of the episome (CpG methylation at HhaI or HpaII sites, or non-CpG methylation at Dcm sites of the plasmid). It is clear that Dnmt3a is not stimulated by pre-existing cytosine methylation at CpG sites on only one of the DNA strands because it does not prefer hemimethylated DNA over fully unmethylated DNA. While Dnmt3a does not appear to be stimulated by methylation at HpaII sites on a 310 bp DNA fragment \[[@B24]\], extensive testing has not been carried out. One study showed stimulation of Dnmt3a by hemimethylated CCG and fully methylated CWG; no stimulation by fully methylated CCG was detected \[[@B25]\]. It is curious that Dnmt3a is not stimulated by hemimethylated CpG, fully methylated CpG, and fully methylated CCG, but it is stimulated by hemimethylated CCG and fully methylated CWG in vitro \[[@B25]\]. In this study, we carried out experiments to determine whether Dnmt3a has similar de novo methylation activity on single-stranded substrates and whether it can be stimulated by pre-existing cytosine methylation, as is the case for Dnmt1. Results ======= Dnmt3a has substrate preferences and does not act on single-stranded DNA substrates ----------------------------------------------------------------------------------- The activity of murine Dnmt3a purified from human cells overexpressing this enzyme was tested in an in vitro methylation assay with in vitro ^3^H-incorporation using fully unmethylated, hemimethylated, and fully methylated 30 bp or 122 bp oligonucleotide substrates. All reactions were carried out in duplicate with no protein and no DNA controls, and all experiments were done multiple times. The no protein controls routinely showed a very low level of ^3^H counts; therefore, the level of radioactivity detected in the no DNA control was treated as background and subtracted from reactions with DNA substrates. Dnmt1 was used as a control in similar reactions. Consistent with previous reports \[[@B12],[@B24]\], Dnmt1 is strongly stimulated by hemimethylated substrates showing approximately a 13-fold higher activity on these substrates than on fully unmethylated DNA and showed nearly no activity on fully methylated DNA (Fig. [1A](#F1){ref-type="fig"}). Also consistent with the lack of in vitro site preference reported previously \[[@B14]\], virtually no difference in Dnmt1 activity between the 30 bp and the 122 bp substrates was detected (Fig. [1A](#F1){ref-type="fig"}). This validates the assay and the substrates used. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### **In vitromethylation of oligonucleotides by DNA methyltransferases.*A) Duplex 30-mer and 122-mer DNA substrates with hemimethylated top strand (HMTop), hemimethylated bottom strand (HMBot), both strands methylated (DM), and fully unmethylated (UM) were used in the in vitro ^3^H-incorporation assay (counts per minute, CPM) to test the activity of purified Dnmt1. B) Duplex 30-mer and 122-mer DNA substrates with hemimethylated top strand (HMTop), hemimethylated bottom strand (HMBot), both strands methylated (DM), and fully unmethylated (UM) were used in the in vitro ^3^H-incorporation assay (counts per minute, CPM) to test the activity of purified Dnmt3a. C) The single-stranded 122-mer DNA substrates, unmethylated top strand (SSTop), unmethylated bottom strand (SSBot), methylated top strand (SMTop), and methylated bottom strand (SMBot), were used to assay Dnmt1 activity in vitro. D) Single-stranded 122-mer DNA substrates, unmethylated top strand (SSTop), unmethylated bottom strand (SSBot), methylated top strand (SMTop), and methylated bottom strand (SMBot), were used to assay Dnmt3a activity in vitro. ::: ![](1471-2091-6-6-1) ::: Consistent with previous reports \[[@B12],[@B13]\], Dnmt3a showed a very low activity on all of the 30 bp substrates, though the highest activity was detected with fully unmethylated DNA (UM); and nearly no activity was detected with fully methylated substrate (DM) (Fig. [1B](#F1){ref-type="fig"}). A Dnmt3a catalytic mutant \[[@B15]\] showed background levels of radioactivity, confirming the activity of Dnmt3a (data not shown). In contrast, Dnmt3a showed a 20-fold increase in activity on unmethylated 122 bp substrates over the unmethylated 30 bp substrate (UM, Fig. [1B](#F1){ref-type="fig"}). While near background activity was detected on fully methylated substrates (DM), Dnmt3a displayed the highest activity on fully unmethylated duplex DNA (UM), and its activity on hemimethylated substrates (HMTop and HMBot) was two- to three-fold lower (Fig. [1B](#F1){ref-type="fig"}). The activity of Dnmt3a on hemimethylated and unmethylated 122 bp substrates is comparable when the number of available unmethylated CpG sites is taken into account. The 30 bp and the 122 bp oligonucleotide substrates have 3 and 4 CpG sites on each strand, respectively. Therefore, the difference in Dnmt3a activity is not proportional to the number of available CpG sites on the 30 bp and the 122 bp substrates. Consistent with our previous findings \[[@B15]\] and results using several other substrates (data not shown), these findings suggest that Dnmt3a has substrate preferences while Dnmt1 does not. A de novo methylation activity of Dnmt1 has been described as acting on single-stranded substrates at low efficiency in vitro \[[@B1],[@B17]\]. We were interested in determining whether Dnmt3a, with its primary function as a de novo methyltransferase, can also use single-stranded DNA as a substrate. In vitro methylation assays were carried out using unmethylated and methylated single-stranded oligonucleotides as substrates for the analysis of Dnmt1 and Dnmt3a. Dnmt1 showed moderate activity on unmethylated single-stranded DNA (Fig. [1C](#F1){ref-type="fig"}; SSTop and SSBot). This activity is five-fold lower than its activity on hemimethylated double-stranded DNA substrates (compare with Fig. [1A](#F1){ref-type="fig"}; HMTop and HMBot) and is two-fold higher than its activity on fully unmethylated duplex DNA (compare with Fig. [1A](#F1){ref-type="fig"}; UM). These findings are consistent with previous reports \[[@B1],[@B18]\]. Unlike Dnmt1, Dnmt3a showed very little activity on unmethylated single-stranded DNA substrates (Fig. [1D](#F1){ref-type="fig"}; SSTop and SSBot) and this activity is ten-fold lower than its activity on fully unmethylated double-stranded DNA (compare with Fig. [1B](#F1){ref-type="fig"}; UM). Both enzymes showed near background activity on methylated single-stranded DNA substrates (SMTop and SMBot in Fig. [1C](#F1){ref-type="fig"} and Fig. [1D](#F1){ref-type="fig"}). These findings clearly indicate that Dnmt1 has a stronger de novo methylation activity on single-stranded DNA substrates than on unmethylated duplex DNA, and the de novo methylation activity of Dnmt3a primarily targets duplex DNA substrates with unmethylated CpG sites and not single-stranded DNA. It is also clear that the de novo methylation activity of these two enzymes is quite different in vitro. Pre-existing cytosine methylation on the plasmid does not stimulate methylation activity of Dnmt3a in vitro ----------------------------------------------------------------------------------------------------------- Previous work indicated that partially-purified human and mouse Dnmt1 has a higher level of methylation activity on partially-methylated plasmids than fully unmethylated plasmids in vitro \[[@B19]\]. Dnmt3a overproduced in E. colishowed no increased activity in vitro when three HpaII sites were methylated on a 310 bp substrate, while Dnmt1 showed a 2.9-fold increase in activity, even though the activity reported was low for Dnmt3a \[[@B24]\]. Another report indicated that hemimethylated CCG can stimulate Dnmt3a but not Dnmt1; fully methylated CCG fails to stimulate either enzyme; and fully methylated CWG can stimulate both enzymes \[[@B25]\]. We were interested in further testing whether murine Dnmt3a is stimulated by pre-existing cytosine methylation on the plasmid in vitro. Plasmid DNA is often prepared using Dcm-positive strains of E. coli; therefore, it carries methylated cytosines at the internal C of CC(A/T)GG site. For this experiment, plasmid p220.2 DNA was prepared from a Dcm-negative strain (the plasmid DNA thus derived is designated p220.2-Dcm) and from a Dcm-positive strain (plasmid designated p220.2+Dcm) for comparison. The p220.0-Dcm DNA was further methylated in vitro using HhaI (GCGC, bold indicates methylation site) methylase, HpaII (CCGG) methylase, HaeIII (GGCC) methylase, and MspI (CCGG) methylase individually. The p220.2+Dcm DNA was also further methylated in vitro using HhaI methylase, HpaII methylase, BamHI methylase (GGATCC), and HaeIII methylase individually. BamHI, HaeIII, and MspI methylases methylate cytosines at non-CpG sites, while HhaI and HpaII methylases methylate cytosines at CpG sites. These in vitro methylated substrates were used in the in vitro methylation assay of Dnmt3a with p220.2+Dcm and p220.0-Dcm DNA as controls. There is very little difference in the Dnmt3a methylation activity between plasmid DNA with or without Dcm methylation, indicating the lack of stimulation by fully methylated CWG sites. Although there are minor variations detected for substrates methylated with different methylases, no stimulation by pre-existing cytosine methylation was observed (Fig. [2](#F2){ref-type="fig"}). These findings indicate that Dnmt3a is not stimulated by pre-existing symmetrical cytosine methylation at CpG sites or at non-CpG sites on the plasmid substrates in vitro. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Pre-existing cytosine methylation does not stimulate Dnmt3a in vitro.Plasmid DNA extracted from a Dcm-negative E. colistrain and a Dcm-positive E. colistrain was used in the in vitro ^3^H-incorporation methylation assay (CPM). The average Dnmt3a activities on each plasmid are summarized in the table above the histogram. Both Dcm+ and Dcm-DNA with no CpG methylation (Un-Me), CpG methylation at 32 HhaI sites (meHhaI) or 40 HpaII sites (meHpaII), and non-CpG methylation at 51 HaeIII sites (meHaeIII) were tested. Dcm+ DNA with non-CpG methylation at a single BamHI site (meBamHI) and Dcm- DNA with non-CpG methylation at 40 MspI sites (meMspI) were also tested. ND represents data point not determined. ::: ![](1471-2091-6-6-2) ::: Pre-existing CpG methylation on a minichromosome does not lead to increased Dnmt3a methylation at HhaI or HpaII sites in human cells ------------------------------------------------------------------------------------------------------------------------------------ To test whether pre-existing CpG methylation can stimulate Dnmt3a activity in vivo, a previously established assay system was used \[[@B22]\]. It has been described previously \[[@B22]\], cotransfection of Dnmt3a expression vector with a target plasmid, such as p220.2, into 293/EBNA1 cells leads to methylation of the target plasmid while no methylation on the target plasmid was detected when the target plasmid was transfected alone or with a Dnmt3a catalytic mutant expression vector into 293/EBNA1 cells. When the transfected plasmid DNA is digested with a methylation-sensitive restriction enzyme, such as HhaI or HpaII, the fraction of the increased size fragments can be compared to the fully digestable fragments in the same lane to assess the activity of Dnmt3a on the plasmid. Any loading variation between lanes does not interfere with this assessment because no comparisons of fragments across lanes are made. p220.2-Dcm plasmid DNA was methylated with HhaI methylase, HpaII methylase, HaeIII methylase, and MspI methylase in vitro before being co-transfected with the Dnmt3a expression vector into 293/EBNA1 cells. The unmethylated p220.2-Dcm and p220.2+Dcm were used as controls in this experiment. Seven days after transfection, the plasmid DNA was harvested using the Hirt method and analyzed by Southern blotting after restriction digestion. There is no appreciable difference in Dnmt3a methylation activity on plasmids with and without Dcm methylation when the methylation sensitive enzyme HhaI is used to assess DNA methylation on p220.2 (Fig. [3A](#F3){ref-type="fig"}, lanes 1 and 2). There is also little difference in methylation at HhaI sites between HpaII-methylated, HaeIII-methylated, and MspI-methylated p220.2-Dcm plasmid DNA versus plasmid DNA without in vitro methylation (Fig. [3A](#F3){ref-type="fig"}, lanes 2, 3, 4, and 5). No apparent differences were detected in Dnmt3a methylation activity at HpaII sites between unmethylated plasmid and HhaI-methylated, HaeIII-methylated, and MspI-methylated plasmid DNA (Fig. [3B](#F3){ref-type="fig"}). When HhaI, HpaII, BamHI, and HaeIII methylases were used to methylate p220.2+Dcm DNA, similar results were observed in the transfection experiments (data not shown). Similar levels of Dnmt3a protein was detected at two days after transfection by Western blot using the Myc antibody in the experiments where Dnmt3a expression was monitored (data not shown). When the same experiments were carried out using 3a-5 cells that overexpress murine Dnmt3a, similar results were seen (data not shown). These results indicate that pre-existing symmetrical methylation at CpG sites or non-CpG sites does not stimulate Dnmt3a methylation at HhaI or HpaII sites in human cells. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Southern blot analysis of de novo methylation on plasmid DNA with various types of pre-existing cytosine methylation.A) Southern hybridization of HhaI-digested plasmid DNA harvested eight days after transfection into human cells. Plasmid DNA with no Dcm methylation (p220.2-Dcm) and with Dcm methylation (p220.2+Dcm) showed no detectable difference in the amount of methylation acquired at HhaI sites. When p220.2-Dcm DNA is methylated in vitro with HpaII methylase (me-HpaII p220.2), HaeIII methylase (me-HaeIII p220.2), or MspI methylase (me-MspI p220.2), and then used in the transfection experiment, no detectable difference due to either p220.2+Dcm or p220.2-Dcm was observed. B) Southern hybridization of HpaII-digested plasmid DNA harvested eight days after transfection into human cells. No detectable difference was observed between p220.2-Dcm, p220.2+Dcm, me-HaeIII p220.2, me-MspI p220.2 and plasmid in vitro methylated with HhaI methylase (me-HhaI p220.2). ::: ![](1471-2091-6-6-3) ::: Pre-existing CpG methylation at HpaII or HhaI sites on the minichromosome does not stimulate Dnmt3a methylation at other CpG sites in human cells ------------------------------------------------------------------------------------------------------------------------------------------------- To examine whether methylation at CpG sites other than HhaI or HpaII sites is stimulated by pre-existing cytosine methylation, sodium bisulfite sequencing of three different regions in the EBNA1 gene of the plasmid harvested from transfected human cells was carried out. The unmethylated p220.2, HhaI-methylated p220.2, and HpaII-methylated p220.2 plasmids harvested from transfected cells were regionally sequenced for comparison. Methylation at the HhaI and the HpaII sites was excluded from the calculation of methylation because of the pre-existing methylation at these sites on the HhaI-methylated and HpaII-methylated plasmids. Unmethylated p220.2 showed 50.8% methylation in the EBNA1 region 2, 36.3% methylation in the EBNA1 region 4, and 6.3% methylation in the EBNA1 region 6 (Fig. [4A](#F4){ref-type="fig"}). HhaI-methylated plasmid p220.2 had 69.3%, 52.8%, and 5.6% of methylation in EBNA1 regions 2, 4, and 6, respectively (Fig. [4B](#F4){ref-type="fig"}). Similar levels of methylation (49.4%, 53.8%, and 10.4% in EBNA1 regions 2, 4, and 6, respectively) were found on HpaII-methylated p220.2 (Fig. [4C](#F4){ref-type="fig"}). The patterns of methylation in all three regions of the EBNA1 gene were also similar for these three plasmids (Fig. [4](#F4){ref-type="fig"}). Previously, 55.3 % and 10.6% of methylation were detected in the EBNA1 region 2 and region 6, respectively, in similar experiments \[[@B15]\]. The percentage of overall methylation is higher in region 2 of HhaI-methylated p220.2 than unmethylated and HpaII-methylated p220.2. The percentage of overall methylation in region 4 is also higher in both HhaI-methylated and HpaII-methylated p220.2 than unmethylated p220.2. However, the patterns of methylation are similar for all plasmids in all three regions. We reasoned that if preexisting methylation HhaI or HpaII sites can stimulate Dnmt3a in vivo, a larger fraction of highly methylated molecules should be observed in the transfected HhaI-methylated and HpaII-methylated p220.2 plasmid when compared with unmethylated p220.2. Therefore, the sequenced molecules were sorted into two groups, a group with more than 50% of the CpG sites methylated and one with less than 50% of the CpG sites methylated. A chi-square test was then carried out to test the hypothesis that there is no significant difference between the number of molecules in these two groups from p220.2, HhaI-methylateed-p220.2, and HpaII-methylated p220.2 plasmids. Fisher\'s exact test was used to test the same hypothesis for results from region 6 because no molecules from unmethylated and HhaI-methylated p220.2 had more than 50% of the CpG sites methylated. The Pvalues of the statistical tests were 0.17, 0.13, and 0.32 for regions 2, 4, and, 6, respectively (Fig. [4D](#F4){ref-type="fig"}), indicating that the difference is not significant. The same tests were carried out for regions 2 and 4 with a different cut-off for the sorting to ensure that this does not change the outcome. When the sequenced molecules from region 2 were sorted as one group with 5 to 10 methylated CpG sites and another group with 0 to 4 methylated CpG sites, the Pvalue of the chi-square test is 0.37. When the sequenced molecules from region 4 were sorted into a group with 4 to 8 methylated CpG sites and a group with 0 to 3 methylated sites, the Pvalue of the chi-square test is 0.22. Therefore, the higher level of methylation observed in the EBNA1 regions 2 and 4 on HhaI-methylated plasmid and in the EBNA1 regions 2 and 6 on HpaII-methylated plasmid is considered within normal experimental variation and not due to stimulation of Dnmt3a by the pre-existing methylation at HhaI of HpaII sites. These findings indicate that there is no stimulation of methylation at other CpG sites in human cells when HhaI or HpaII sites carry pre-existing cytosine methylation. It is noteworthy that no methylation at non-CpG sites was detected in all three regions sequenced. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Sodium bisulfite sequencing of three different regions of the EBNA1 gene from transfected plasmids.A) Fully unmethylated p220.2, B) HhaI methylated p220.2, and C) HpaII methylated p220.2 were harvested eight days after co-transfection into human cells and sodium bisufite sequenced. Regions 2, 4, and 6 of the EBNA1 gene as designated previously \[[@B14]\] were sequenced. There is a single HhaI site in region 2 and region 6, and there are three HpaII sites in region 4 as indicated on the top of the CpG sites and designated with a box. The percentage of methylation was calculated by dividing the total number of methylation sites detected (excluding methylation found in the HhaI and HpaII sites) by the total number of CpG sites examined (also excluding the HhaI and the HpaII sites). The white rectangular box indicates the premethylated HhaI and HpaII sites on the plasmids. D) Sorting and statistical analysis of the two groups of sequenced molecules from each region. The degrees of freedom (df) are equal to 2. Column A represents molecules from unmethylated p220.2, column B represents molecules from HhaI-methylated p220.2, and column C represents molecules from HpaII-methylated p220.2. ::: ![](1471-2091-6-6-4) ::: Discussion ========== It has been described that the de novo methylation activity of Dnmt1 can act on single-stranded DNA \[[@B1],[@B17]\]. This leads to the hypothesis that single-stranded DNA formed during replication or repair might serve as a nucleation site for de novo methylation \[[@B17]\]. Several studies suggest that CpG islands may serve as DNA replication origins \[for a review see \[[@B26]\]\]. As first described by Baylin in 1986 \[[@B27]\], hypermethylation of CpG islands has been well recognized to occur in cancer. It is possible that the single-stranded DNA methylation activity of Dnmt1 targets it to CpG islands that are DNA replication origins and initiates de novo methylation in the genome. Since the primary function of Dnmt3a is de novo methylation, it is important to know whether Dnmt3a has such an activity. Unlike Dnmt1, we find here that Dnmt3a does not act on single-stranded oligonucleotides in vitro. Although one previous study \[[@B28]\] reported Dnmt3a methylation of single-stranded oligonucleotide substrates, the conclusion remains unclear based on the substrate used in that study. The substrate used in Yokochi and Robertson \[[@B28]\] has 10 CpG sites on an oligomer which is 34 nucleotides in length and which has GATC on both ends; therefore, it can form double-stranded regions over substantial lengths both intra- and inter-molecularly under low salt and moderately high temperature conditions. In contrast, the 122 bases oligonucleotide substrate used in this study does not form CpG pairing either intra- or inter-molecularly based on several oligonucleotide analysis software packages. Furthermore, no other conformation was observed when the radioactive-labeled oligonucleotide was resolved on a native gel either as single-stranded or double-stranded configuration (data not shown). In our study, Dnmt3a showed a much higher activity on the 122 bp double-stranded DNA substrate than on the 30 bp substrate, indicating some form of substrate preference. We have tested other substrates ranging from 83 bp to 130 bp and the substrate length or the potential processivity of the enzyme is not the basis for these differences (data not shown), indicating that local DNA sequence is the basis. These findings suggest functional differences for the de novo methylation activity of these two enzymes. Based on our study here, if de novo methylation occurs at CpG sites within DNA with single-stranded features in vivo, then Dnmt3a is not likely to be the methyltransferase responsible for this process. Also, Dnmt3a may prefer certain sequences while Dnmt1 does not discriminate among its targets. It has been shown that pre-existing cytosine methylation can stimulate de novo methylation activity of Dnmt1 in vitro both in cis and in trans \[[@B17]-[@B19]\]. This suggests that Dnmt1 can be stimulated by hemimethylated DNA as well as symmetrically methylated DNA. In a previous study, Dnmt3a overproduced in E. colishowed no increased activity when three HpaII sites were methylated on a 310 bp substrate, even though the activity of Dnmt3a was low \[[@B24]\]. However, another study showed stimulation of Dnmt3a by hemimethylated CCG and fully methylated CWG \[[@B25]\]. Here, we performed both in vitro and in vivo testing of whether pre-existing cytosine methylation can stimulate Dnmt3a, including fully methylated CCG and CWG sites. We find that Dcm+ plasmid DNA does not stimulate Dnmt3a activity in vitro when compared with plasmid DNA without methylation at these Dcm sites, suggesting the lack of stimulation by methylation at CWG sites. This finding is at odds with the finding of Kim et al. \[[@B25]\]. There are several non-CpG sites on the methylated CWG and CCG substrates used in Kim et al. \[[@B25]\]. It is possible that CWG methylation does not stimulate CpG methylation but stimulates a non-CpG methylation activity in those reactions, since the Dnmt3a used in Kim et al. was demonstrated to have non-CpG methylation activity \[[@B25]\]. The Dnmt3a used in our study has no detectable non-CpG methylation activity; therefore, no increased activity was observed with pre-existing cytosine methylation. However, it is intriguing that Dnmt3a is not stimulated by hemimethylated CpG, fully methylated CpG, or fully methylated CCG, and yet it is stimulated by hemimethylated CCG and fully methylated CWG in vitro \[[@B25]\]. Whether this difference is due to the expression of the enzyme in insect cells versus human cells would need further investigation. There is also no stimulation of the Dnmt3a activity in vitro by cytosine methylation at other CpG or non-CpG sites on the plasmid with or without Dcm methylation. These findings indicate a clear difference between the de novo methylation activity of Dnmt1 and Dnmt3a in vitro. Transfection experiments also showed that pre-existing cytosine methylation at CpG sites or non-CpG sites does not stimulate Dnmt3a to methylate CpGs at HhaI and HpaII sites on the minichromosome. Sodium bisulfite sequencing of the transfected minichromosome further showed that methylation at HhaI or HpaII sites does not lead to increased methylation at other CpG sites. Methylation at non-CpG sites by Dnmt3a was not detected in any of the molecules sequenced, indicating the lack of non-CpG methylation by Dnmt3a in this particular human cell line. These findings demonstrate that pre-existing cytosine methylation does not stimulate Dnmt3a activity for either CpG or non-CpG methylation in human cells. It has been proposed that Dnmt1 plays a role in DNA methylation spreading in mammalian cells \[[@B24]\]. In our previous transfection experiments using patch-methylated minichromosomes, no spreading of methylation from the methylated patch was observed \[[@B23]\]. We also have not observed de novo methylation on the transfected minichromosomes with methylation at HhaI or HpaII sites by bisulfite sequencing (data not shown). These findings indicate the lack of de novo methylation activity on these minichromosomes without overexpression of Dnmt3a in human cells. Although it remains unclear which methyltransferases are involved in methylation spreading, the fact that Dnmt3a is not stimulated by hemimethylated DNA or by symmetrically methylated DNA strongly suggests that Dnmt3a does not participate in this process. Consistent with the findings in Fatemi et al \[[@B24]\], we have found that plasmid DNA premethylated with Dnmt3a stimulates Dnmt1 activity more than two-fold in vitro (data not shown). In our previous study, Dnmt3a generates an asymmetrical methylation pattern on the two DNA strands in vitro \[[@B15]\]. Therefore, it is very likely that one of the processes for de novo methylation is for Dnmt3a to initiate methylation on one DNA strand; then this hemimethylated DNA would lead to the stimulation of Dnmt1 for further methylation (Fig. [5](#F5){ref-type="fig"}). This activity of Dnmt1 would be in addition to its known methylation of hemimethylated sites after DNA replication (Fig. [5](#F5){ref-type="fig"}). ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Model for one of the de novo methylation processes carried out by Dnmt3a and Dnmt1.*Dnmt3a initiates methylation on one DNA strand at its preferred CpG sites (filled circles). After DNA replication, Dnmt1 methylates the other strand (open circles) at these hemimethylated sites. The hemimethylated DNA may also stimulate Dnmt1 for further methylation at previous unmethylated CpG sites (open circles indicated with ?). ::: ![](1471-2091-6-6-5) ::: Conclusion ========== In this study, we demonstrated several novel differences between Dnmt1 and Dnmt3a de novo methylation activity. Unlike Dnmt1, Dnmt3a produced in human cells has very little activity on single-stranded DNA in vitro. While Dnmt1 shows higher de novo methylation activity on single-stranded substrates than on unmethylated duplex DNA, Dnmt3a primarily acts on unmethylated sites on duplex DNA. Dnmt1 shows no preference for substrates with different sequence or length, whereas Dnmt3a shows preference for different substrates. In contrast to Dnmt1, Dnmt3a is not stimulated by pre-existing cytosine methylation on plasmids either at CpG sites or non-CpG sites in vitro. This is the first study demonstrating that pre-existing cytosine methylation does not stimulate the activity of Dnmt3a at CpG or non-CpG sites in human cells. The de novo methylation activity of these two enzymes is clearly different. Methods ======= Oligonucleotide substrates -------------------------- The 30-mer double-stranded oligonucleotide substrates used in this study are identical to the ones used in Pardhan et al \[[@B1]\] and Okano et al \[[@B12]\] with the addition of one hemimethylated substrate with the CpG methylation on the anti-parallel DNA strand. The 30-mer oligonucleotides were synthesized by Operon. The top strand of the 122-mer substrate, 5\'-GATACATATTTGAATGTATTTAGAAAAATAAACAAATAGGGGTTC CGCGCACATTTCCCCGAAAAGTGCCACCTGACGTCTAAGAAACCATTATT ATCATGACATTAACCTATAAAAATAGG-3\', was derived from the EBNA1 sequence that contains 4 CpG sites. The bottom strand of the substrate is the complementary sequence of the top strand. The 122-mer oligonucleotides were synthesized and purified by the Microchemical Core facility of the Norris Cancer Center. The double-stranded substrates were generated by adding equimolar amounts of the complementary strands in 10 mM TrisHCl (pH 8.0) and 100 mM NaCl, boiling for 10 min, and slowly cooling in a bath of 300 ml water at room temperature. The annealing of the oligonucleotides was checked by PAGE. The hemimethylated HMTop substrates were generated using a top strand oligonucleotide with methyl cytosines at all the CpG sites and a fully unmethylated bottom strand oligonucleotide. The hemimethylated HMBot substrates were made with unmethylated top strand and fully methylated bottom strand oligonucleotides. The fully methylated substrates (DM) contain methylated top and methylated bottom strand oligonucleotides. The fully unmethylated substrates (UM) were generated with unmethylated top and bottom strand oligonucleotides. Enzymes and in vitro ^3^H-incorporation assay --------------------------------------------- Human Dnmt1 was purchased from New England Biolabs. The GST-Dnmt3a (GST-3a) and GST-Dnmt3a mutant (GST-3aMut) fusion proteins were expressed in 293T cells and purified using glutathione-agarose beads as described previously \[[@B15]\]. In vitro methylation activity of Dnmt1, Dnmt3a, and Dnmt3a mutant was measured by ^3^H-incorporation assay. In vitro methylation was carried out for Dnmt3a and the Dnmt3a mutant in a 20 ul reaction with 314 fmol of enzyme in 10 mM TrisHCl (pH 8.0), 1 mM EDTA, 1 mM DTT, and 1.1 uM ^3^H-AdoMet (New England Nuclear, 14.7 Ci/mmol) at 37°C overnight (16 hours). The same assay was carried out for Dnmt1 (New England Biolabs) in a 25 ul reaction with 1.1 uM ^3^H-AdoMet in the manufacturer\'s recommended buffer using 2.5 units of the enzyme. In each reaction, 0.25 pmol of the duplex substrate or 0.5 pmol of the single-stranded substrate was used. The in vitro methylation reaction was treated with Proteinase K at 55°C for 1 h before being spotted onto the DE-81 filters and washed as described previously \[[@B29]\]. The radioactivity retained on the air-dried DE-81 filters was measured by scintillation counting (Packard Tri Carb 2100TR) with 2 ml of scintillation fluid. Plasmid and in vitro methylation of plasmid ------------------------------------------- The plasmid p220.2 \[[@B30]\] was used throughout the study. Plasmid p220.2 DNA without Dcm methylation, designated p220.2-Dcm, was prepared from a Dcm-negative strain, GM127. Plasmid DNA with Dcm methylation, designated p220.2+Dcm, was prepared from a Dcm-positive strain, DH10B. The p220.0-Dcm DNA was further methylated in vitro using HhaI methylase, HpaII methylase, HaeIII methylase, and MspI methylase individually. The p220.2+Dcm DNA was further methylated in vitro using prokaryotic methylase, HhaI methylase, HpaII methylase, BamHI methylase, or HaeIII methylase individually. MeHhaI-p220.2 was in vitro methylated at all HhaI sites using the HhaI-methylase (New England Biolabs) under the conditions recommended by the manufacturer. MeHpaII-p220.2 was methylated at HpaII sites according to the conditions suggested by the manufacturer using the HpaII methylase (New England Biolabs). Two cytosine methylases that methylate non-CpG sites, BamHI methylase and HaeIII methylase, were used to generate meBamHI-p220.2 and meHaeIII-p220.2, respectively. There are 32 HhaI sites, 40 HpaII sites, a single BamHI site, 51 HaeIII sites, and 40 MspI sites on p220.2. After methylation, DNA was extracted with phenol-chloroform followed by ethanol precipitation. The methylation status was verified by restriction enzyme digestion using the appropriate enzyme for the specific methylase used. Cell lines and transfection --------------------------- The 293/EBNA1 and 3a-5 cells were used in this study. The 3a-5 cells were derived by integrating a Dnmt3a expression construct into the 293/EBNA1 cells \[[@B22]\]. Throughout the study, the calcium phosphate transfection method was used to introduce DNA into the human cell line. In some experiments, wildtype or mutant Dnmt3a expression vector \[[@B22]\], pMT3aMyc and pMT3aMut, was cotransfected with the assay plasmid into 293/EBNA1 cells. DNA recovery and analysis ------------------------- Plasmid DNA was harvested from the cells by the Hirt method \[[@B31]\] when the cells reached confluence after transfection. A small fraction of the cells was re-seeded for later harvests. The plasmid DNA harvested from the transfected cells was digested with HhaI or HpaII enzyme, fractionated on 1% agarose gel, and analyzed by Southern blot analysis using the entire plasmid as a probe. Bisulfite genomic sequencing and statistical analysis ----------------------------------------------------- Bisulfite genomic sequencing was carried out as described previously \[[@B32]\] with minor modifications. Three different regions, regions 2, 4, and 6, of the EBNA1 gene, on the plasmid p220.2 were amplified and ligated into the TOPO TA cloning vector (Invitrogen). The locations of these three regions as well as primer sequences and PCR conditions have been described previously \[[@B14]\]. Both strands of each clone were sequenced using the EXCEL II sequencing kit (Epicentre) and analyzed on the Li-Cor 4200 sequencer. Bisulfite sequencing results for each of the three regions from p220.2, HhaI-methylated p220.2, and HpaII-methylated p220.2 were compared. Molecules were sorted into two groups: one with more than 50% of the CpG sites methylated and one with less than 50% of the CpG sites methylated. A chi-square test was carried out to test the hypothesis that there is no significant difference between the number of molecules in each of the two groups observed from p220.2, HhaI-methylateed-p220.2, and HpaII-methylated p220.2 plasmids. A Fisher\'s exact test was used to test the same hypothesis for results from region 6 because nearly all of the molecules had less than 50% of the CpG sites methylated. Acknowledgements ================ I thank S. Groshen for statistical analysis consultation, which is supported by the Norris Cancer Center Core grant. I also thank J.-Y. Wang, K. Yu, and M. R. Lieber, for helpful discussion and critical reading of the manuscript. This work was supported by NIH grant GM60237.	PubMed Central	2024-06-05T03:55:55.813284	2005-3-30	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084342/", "journal": "BMC Biochem. 2005 Mar 30; 6:6", "authors": [ { "first": "Chih-Lin", "last": "Hsieh" } ] }
PMC1084343	Background ========== The spread of microarray technology has made possible the routine and simultaneous measurement of expression profiles for tens of thousands of genes. In the case of photolithographically synthesized high-density oligonucleotide arrays as described in \[[@B1]\], the technology for hybridizing RNA on chips and quantitating fluoresence-intensity data has been highly standardized and automated. The results are then related to the biology of interest, both through exploratory methods (e.g. \[[@B2]\]) and a large and growing number of sophisticated prediction and classification algorithms (e.g. \[[@B3]\]). Yet the very first step on which these procedures rely is still open to discussion: the derivation of a numerical summary value that is both representative of a gene\'s relative expression level and reasonably free of technical variation, summarily referred to as low-level analysis. The need for a summary function is due to the setup of high-density oligonucleotide arrays, where each gene is probed by a set of paired oligonucleotides: one of each pair matches the target sequence on the probed gene perfectly (perfect match or PM oligo), the other has one altered central base-pair (mismatch or MM oligo), where the MMs serve to establish a reference for non-specific hybridisation. While the full set of PMs has been used successfully for detecting differential expression \[[@B4]\], there is usually a strong interest in having one number that represents the relative abundance of a gene on a chip. The most common summary measures use a non-model-based robust averaging of measurements in a probe set, such as Affymetrix\'s MAS5 expression value \[[@B5]\], or a model-based expression index (MBEI \[[@B6]\]) or a log-additive robust-multichip-average (RMA \[[@B7]\]) across chips. The second crucial aspect of low-level analysis is the control of technical variation between chips, which is introduced by the measurement process during sample preparation, labelling, hybridization, and scanning. Technical variation of this kind and the need for a corrective normalization procedure are not specific to high-density oligonucleotide arrays, but are a general feature of mRNA measurement, e.g. for cDNA microarrays \[[@B8]\], northern-blot analysis or RT-PCR \[[@B9]\]. Numerous procedures have been suggested, differing in their assumptions on what feature of the data remains constant across chips and can therefore be used for normalization \[[@B10]\]. Comparative evaluation of different approaches to low-level analysis has so far been limited to artificial data sets, where differential expression is due to spiked-in RNA or mixtures and dilutions of RNA from different sources \[[@B4],[@B10],[@B11]\]. This has the obvious advantage that the true expression ratios are known (up to experimental error). Consequently, different approaches can be compared in regard to bias (when estimating fold change) and variance (when testing for differential expression). Results so far indicate that there is generally a trade-off between the two, and it seems fair to say that no current method is optimal under all circumstances. The choice of low-level analysis and especially the choice of normalization have severe impact on the subsequent analysis of the expression data \[[@B12]\]. Given the wide range of methods available, it would be useful to have a method for assessing their relative merits for a concrete data set, without reference to an external spike-in or dilution data set. This is especially true if we have to assume that our data set is not as well behaved as artificial data, either in terms of the percentage of differentially expressed genes or in terms of RNA quality, or both, as for the clinical data set on breast cancer described in the Methods section. In this paper, we propose that by studying coregulation or correlations between random pairs of genes, we can compare different summary measures and assess the effect of different normalization procedures. Our underlying hypothesis is that given a modern large-scale chip covering a large percentage of a species\' genome, randomly selected pairs of genes will be on averageuncorrelated. Note that we do not claim the absence of all biological correlation between genes, but rather that the number of connections between genes in regulatory pathways is small compared to the number of all possible combinations of genes; this argument is given more detail in the Discussion. Consequently, a low-level analysis strategy will be deemed suitable for a given data set, if the resulting normalized expression values are on average uncorrelated for randomly chosen pairs of genes. Lack of correlation is not assessed via formal tests, but by easily adaptable graphical tools that do not rely on stringent conditions for validity. We proceed as follows: first, we establish relationships between lack of normalization and correlations between randomly selected genes for three important summary measures; then we show that the default normalization schemes associated with these summary measures do remove the correlations to a large degree, but not completely, with varying amounts of residual correlation. We also show that where available, housekeeping gene normalization is inferior to default normalization in removing random correlation, and we relate random correlation to the number of unexpressed genes in the data. We conclude by discussing the results and the underlying assumption of our approach as well as considerations for its practical implementation, and point out both limitations and possible extensions. Results ======= Lack of normalization is associated with random correlation ----------------------------------------------------------- We first calculated raw unnormalized MAS5, RMA, and MBEI expression values for the breast cancer, dilution, and spike-in data sets as described in the Methods section. The breast cancer data set is an example of a clinical data set from a real patient population, which is expected to have greater biological variation than the dilution and spike-in data sets. We then computed the Pearson correlation coefficients for 5000 random pairs of probes for each data set. As shown in the upper part of Figure [1](#F1){ref-type="fig"}, the distributions of the correlation coefficients are centered far away from zero for each data set and expression measure. There is clearly a large amount of excess correlation that is unrelated to biological relationships between genes. The similarity of expression between random pairs of genes across chips is due to technical differences between chips which have not been normalized out. This is a striking example of statistical confounding, where genes are apparently correlated for some underlying non-biological reason. We have also found that the technical correlation between genes is inversely related to the variability of the genes involved. This can be seen in the lower part of Figure [1](#F1){ref-type="fig"}, where the correlations between the random pairs are plotted against the product of their standard deviations: the average correlation (shown in blue) is highest for genes with small standard deviations and decreases with increasing variability. This fits well with what we would expect from assuming a simple additive chip effect as the source of chip-to-chip variation; even though this is certainly an oversimplification, the corresponding model fits the general shape of the data well enough (shown as red line in Figure [1](#F1){ref-type="fig"}; see Methods). Default normalization removes excess correlation ------------------------------------------------ We calculated the same expression measures for the same data sets as above, but applied the default normalization procedure suggested for each expression measure: for MAS5 expression values, we normalized to the global mean within each array, for RMA values, we applied the quantile normalization, for MBEI we applied the invariant set normalization, see Methods. The upper part of Figure [2](#F2){ref-type="fig"} shows that in all cases, the default normalization step was sufficient to remove excess correlation and center the distribution of the correlation coefficients at zero. In the following, we will refer to unwanted correlation artifacts after normalization as residual correlation. Although we observed no residual correlation for the whole set of genes, there was no guarantee that this would hold for certain subsets of genes: an ideal normalization should remove the residual correlation for any sufficiently large subset of genes. Therefore, we investigated the pattern of correlations for pairs of genes with different intensity and variability across chips. Genes with low variability are poorly normalized by RMA and MBEI ---------------------------------------------------------------- We previously described the systematic inverse relationship between correlation and variability. Although the default normalizations strongly reduced the scale of this correlation for all three expression measures, we still found a significant relationship between correlations and variability for RMA and MBEI, especially for the breast cancer data. The lower part of Figure [2](#F2){ref-type="fig"} shows the average correlations between genes grouped by the product of their standard deviations; this is the same summary line as in Figure [1](#F1){ref-type="fig"}, but without plotting the individual points contributing to it. The residual correlations were smaller than before normalization, but the approximate confidence intervals show them to be highly significant. The shape of the relationship also changed and did no longer follow any simple model. We found that the residual correlations were both absolutely larger and more significant for RMA than for MAS5. For MAS5, only the subset of genes with the lowest variability showed significant correlation, all of it positive and less than 0.05. In contrast, for RMA and MBEI, several of the low-variability classes showed significant positive correlation, up to 0.2 for the breast cancer data set. In addition, we observed small but significant negative correlations for genes in the middle range of variability for the breast cancer and dilution data. Thus the analysis shows that RMA and MBEI do not provide properly normalized expression values for genes with low variability, particularly for the clinical data. We will explain this pattern later in terms of absence and intensity of genes. Normalization on housekeeping genes fails to remove excess correlation ---------------------------------------------------------------------- The HGU133A chips that were used for the breast cancer study contain 100 probes for generic housekeeping genes, whose expression is assumed to be constant on average for most or all experimental conditions. Consequently, it has been suggested to use these housekeeping genes for normalization, by adjusting the expression level on each chip so that the average expression of the housekeeping genes is constant across chips (see Methods). To date, there is no convincing evidence whether this method actually works or not, and it seems that some research groups are using it. The correlation test given in Figure [3](#F3){ref-type="fig"} shows that for the MAS5, RMA and MBEI methods of computing expression values, the housekeeping gene normalization failed to remove the excess correlation. There was nonzero average correlation over all genes, indicating a general failure of normalization. The systematic inverse relationship between correlation and variability were at higher levels throughout the range of variability compared to the default normalizations. The failure of housekeeping gene normalization was particularly severe for RMA. Note that even if the amount of residual correlation shown in Figure [3](#F3){ref-type="fig"} for MAS5 housekeeping-genenormalized values looks small, the impact on the subsequent high-level analysis can be serious. Figure [4](#F4){ref-type="fig"} shows the distribution of 22283 gene-wise t-statistics for the housekeeping-normalized and the global-mean-normalized breast cancer data. Each t-statistic compares the mean expression level between (a) postmenopausal women who are users of hormone replacement therapy (HRT) versus (b) those who are not; see (for personal communications see Hall P, Ploner A, Bjöhle J et al.). The t-statistics for the housekeeping-normalized values are globally shifted below zero, indicating a genome-wide down-regulation of thousands of genes. In contrast, the t-statistics based on global-mean-normalized values are centered around zero, suggesting a much less pronounced difference between HRT users and non-users. In this example, the global-mean-normalized results are biologically much more plausible. Absent genes are poorly normalized by RMA and MBEI -------------------------------------------------- In each tissue, only a limited number of genes will be expressed in quantities above the detection limit, usually much fewer than the number of genes available on modern large-scale chips. The purpose of pairing PM and MM probes is to detect which genes are reliably expressed (present genes), and for which genes the observed intensities are dominated by technical and biological noise (absent genes). The most common method of classifying genes as either present or absent is based a non-parametric test for the PM/MM pairs (Affymetrix\'s detection calls \[[@B13]\]). There is currently no consensus on how to use these detection calls. All methods report expression values of all genes including the absent genes, so in principle the analyst might ignore the issue of absent genes and treat all genes as present. Intuitively the absent genes will be measured with a lot of noise, but will they be properly normalized, i.e., will the measurements be unbiased? In order to study the success of normalization of measured expression of absent genes, we classified all genes as either present or absent based on Affymetrix\'s present calls (see Methods). For all data sets, genes were most frequently either completely absent or completely present across all chips (Figure [5](#F5){ref-type="fig"}). Consequently, the pairs of genes in our random samples could naturally be divided into three classes: those averaging few or no present calls between them, those averaging almost a 100% present calls, and those averaging around 50% present calls (upper part of Figure [6](#F6){ref-type="fig"}). These classes correspond naturally to pairs where both genes were mostly absent, or both mostly present, or where one was mostly absent and the other mostly present; by cutting at 33% and 67% average present calls as indicated in the histograms in the upper part of Figure [6](#F6){ref-type="fig"}, we managed to separate these groups evenly. To provide more information, the average correlation for each subset was again plotted against variability; see lower half of Figure [6](#F6){ref-type="fig"}. Generally, the average correlation was highest for pairs of absent genes, indicating failure of normalization of measured expression of absent genes. This was most serious for RMA: excess correlations were consistently and strongly positive for absent pairs and negative for absent/present pairs for all data sets. Only for present pairs, correlations were mostly non-significant and small in absolute value. Correlations for MAS5 were throughout smaller and less significant, with no clear pattern between the three groups of pairs. MBEI showed the same pattern as RMA, though somewhat weaker. This result implies that, at least in case of RMA and MBEI, measured expressions of absent genes were poorly normalized, so analyses of absent genes should be avoided or at least viewed with caution. This interpretation is supported by Figure [7](#F7){ref-type="fig"}, which shows the distribution of t-statistics comparing HRT-users and non-users as above, but only for genes that were not detected (absent) on all 159 chips (n= 4371); the distributions for MBEI and especially RMA indicate strong and wide-spread regulatory effects of HRT, which seems biologically implausible, especially for genes measured at the detection limit throughout the data set. While the absence or presence of a gene could be assessed via other potential quality control measures, the Affymetrix detection call seems to provide useful information for gene filtering. Note that the summary curves of mean correlation shown in Figure [2](#F2){ref-type="fig"} are the weighted means of the curves by presence status shown in Figure [6](#F6){ref-type="fig"}. We can, for example, explain that the high correlations at low variability for RMA in Figure [2](#F2){ref-type="fig"} are mainly due to absent/absent pairs in the expression data. The slight negative dip for genes at the middle range of variability in Figure [2](#F2){ref-type="fig"} is the effect of an incomplete cancelation between the positive correlations for absent/absent pairs and the negative correlations for absent/present pairs in this range. Residual correlation is only weakly related to the expression level of genes ---------------------------------------------------------------------------- Detection of a gene is trivially related to the relative abundance of its mRNA in the sample. Thus, genes that are expressed at the lower end of the detection range are much more likely to be absent. This might indicate that the relationship between the absence/presence of genes and their residual correlation is in fact due to their difference in abundance, and that by focusing on genes with a minimum expression level, we could avoid residual correlation altogether. Figure [8](#F8){ref-type="fig"} shows that this is not the case: when plotting correlations against standard deviations grouped by intensity in the breast cancer data, we found that the pattern of correlation depends more on the percentage of present calls than on the intensity level. The pattern we saw previously in Figure [6](#F6){ref-type="fig"} was observed at different levels of intensity: (i) pairs where both genes are mostly absent tend to be positively correlated, (ii) pairs with one gene mostly absent and one gene mostly present tend to be negatively correlated, and (iii) genes where both partners are mostly present tend to be almost uncorrelated. This pattern is most pronounced at low and medium intensities, and it is stronger for RMA and MBEI, but it is consistently seen, also at high intensities and for MAS5 values. In summary it seems strongly preferable to define a gene filter according to absent/present calls than according to the gene intensity levels. Note that correlation between the intensity and presence of genes is reflected by the number of pairs that contribute to each curve in Figure [8](#F8){ref-type="fig"}: there were relatively more present/present gene pairs and less absent/absent pairs at high intensities, and vice versa for low intensities; curves with lower pair counts have correspondingly wider confidence intervals. Filtering out absent genes reduces residual correlation ------------------------------------------------------- Figure [9](#F9){ref-type="fig"} demonstrates for the breast cancer data how the filtering of genes with a large number of absent calls can reduce residual correlation for normalized expression values. In this case, the 5000 pairs of genes were randomly sampled from subsets of genes with an increasing percentage of present calls. Already by excluding genes that are always absent, the level of systematic correlation was reduced below 0.04 for all expression measures, though the pattern of positive correlations for genes with low variability was still present; by considering only genes with at least 20% present calls, we found that this pattern is reversed for RMA and MBEI, but not for MAS5. Further restrictions did not change this pattern, but increased the absolute level of residual correlation. Discussion ========== The assumption of zero correlation ---------------------------------- As some genes are connected in biochemical pathways, the hypothesis that random pairs of genes will be on average uncoregulated or uncorrelated seems counterintuitive, but it is really a question of scale. For a moderately large chip of 10000 probe sets, there are about 50 million possible pairwise correlations, the huge majority of which will be extremely unlikely to be biological. Any random sample of probe set pairs will contain only a small percentage of pairs representing an unequivocal biological relationship, and additionally, negative and positive correlations will tend to cancel each other out during averaging. We can demonstrate this for the breast cancer data set. On the Affymetrix HGU133A chip, we find represented 124 KEGG pathways, organising 3137 probe sets or 14% of all probe sets on the chip (\[[@B14]\], build 2004/03). This constitutes an as highly-organised subset of the genome as we can currently hope to select, with numerous probe sets appearing in multiple pathways, thereby establishing numerous cross-correlations between pathways. Figure [10](#F10){ref-type="fig"} shows the boxplots of correlations for 5000 randomly selected pairs of genes from this subset, firmly centered at zero for all three expression measures. So even for this special subset of many coregulated genes, the average correlation of a random pair of genes is zero. The simple model of lack of normalization ----------------------------------------- The model described in the Methods section only assumes differences in mean intensity between chips. This corresponds to the simple global mean normalization commonly used for the MAS5 expression values. Figure [1](#F1){ref-type="fig"} confirms that this model (shown in red) describes the average behavior of the correlations (shown in blue) adequately for all data sets, suggesting that global mean normalization is indeed suitable for MAS5 data. Apart from MAS5, the model fits adequately only for the RMA-based correlations in the breast cancer data, suggesting that global mean normalization on the probe-set level may be attempted in this case, but that it is not generally suitable for RMA and MBEI data. Still, Figure [1](#F1){ref-type="fig"} shows that correlations decrease systematically with the variability of the gene pairs for all expression measures, and it may be possible to describe this relationship by extending the simple model, e.g. by allowing the array effect θin Equation 1 to be correlated with the gene effects ψ~i~. The bad performance of housekeeping genes ----------------------------------------- The use of housekeeping genes seems reasonable when studying a small number of genes under controlled experimental settings, or where the choice of one or several housekeeping genes can be motivated biologically. For the breast cancer data, which was collected in a real clinical setting, where samples are both genetically heterogenous and potentially genomically unstable, it is much harder to believe in the common expression of housekeeping genes. Therefore we argue that the failure of housekeeping normalization in this example is not due to the procedure per se, but to our inability to identify a suitable set of housekeeping genes, and the use of the generic set of genes suggested by the chip manufacturer. Even for northern-blot analysis and RT-PCR, where housekeeping normalization is the default, an uncritical use of housekeeping genes has been shown to lead to unacceptable results \[[@B9]\]. Comparison of MAS5, RMA and MBEI -------------------------------- It has been suggested that the generally much lower variability of RMA and MBEI for low-intensity probe sets is a clear advantage of these model-based expression measures over the simpler MAS5 \[[@B7],[@B10]\]. Our results however indicate that this low variability may well be misleading: RMA and MBEI values for absent probe sets, which constitute the vast majority of low-intensity probe sets, show the strongest residual correlation. This indicates that RMA and MBEI values for low-intensity probe sets that are reported without regard for their absence/presence status will be compromised by lack of normalization (Figures [6](#F6){ref-type="fig"} and [8](#F8){ref-type="fig"}). It seems therefore that RMA and MBEI estimate expression of low-abundance genes in a biased, but very precise manner. Minimizing variability as much as possible only makes sense for unbiased estimato rs: if the variability of the estimate becomes small relative to the bias, we get a dangerous sense of confidence in an estimate that is not quite what we think it is. In the same way, the large variability of the MAS5 values at low intensities may well hide an amount of bias comparable to that of RMA and MBEI: as long as the variability of MAS5 is large compared to the bias, we will not be lead to make inappropriate conclusions based on possibly biased estimates; in that sense the MAS5 estimates for low-intensity genes are more honest and better normalized than the corresponding RMA and MBEI values. It is interesting to note that Bolstad et al. have already described the choice between different low-level approaches in terms of bias (when estimating fold change between conditions) and variance (when testing for differential expression between conditions) \[[@B10]\]. Our results suggest that a) the same trade-off applies when looking directly at the expression values, instead of comparing aggregated fold changes and test statistics between different biological conditions, and b) the trade-off is more disadvantageous for the model-based expression measures than generally thought. The underlying lack of normalization of RMA and MBEI for absent genes could be due to the computation of the expression values, or the normalization step, or a combination thereof. Preliminary results (not shown) indicate that the first step, the summarization of the individual probe intensities through the expression measure, seems to be responsible in both cases. If this can be confirmed, a possible explanation would be that the models used (log-linear for RMA and and multiplicative for MBEI) may not be appropriate for absent genes (but see also below). Improving low-level analysis ---------------------------- In a recent paper, Choe et al. have evaluated the performance of a wide range of low-level analysis methods and test procedures in detecting differential expression in a carefully constructed spike-in data set \[[@B15]\]. They report 70% sensitivity at 10% false discovery rate for their top-ranking combinations clearly there is still ample room for improvement in current low-level methodology. We want to outline here shortly how our approach could be used to guide this effort. The authors of \[[@B15]\] found that an additional second step of normalization on the probe-set level improved the performance of MAS5, RMA, and MBEI in detecting differential expression (indeed, MAS5 with the second round of normalization was one of the top-ranking combinations). We have applied the same renormalization to our data sets (see Methods for details), the results are shown in Figure [11](#F11){ref-type="fig"}. We found that renormalization reduced residual correlation for all data sets and all expression measures. Indeed, for MAS5 the correlations are not significantly different from zero at any lag, indicating perfect normalization as measured by our criterion. RMA and MBEI show strongly reduced levels of residual correlation, but are still well above the levels of the original MAS5 as seen in Figure [2](#F2){ref-type="fig"}. It is interesting that the ranking of the original and renormalized expression measures in terms of normalization quality (i.e. renormalized MAS5 is best, followed by the original MAS5, followed by renormalized RMA and MBEI, followed by the original RMA and MBEI) corresponds closely to the ranking by performance in detecting differential expression found by Choe et al. (\[[@B15]\], Figure [7f](#F7){ref-type="fig"}). This suggests that the lack of normalization that our method is able to measure is indeed relevant for the ability to detect regulated genes. Additionally, Figure [11](#F11){ref-type="fig"} gives an indication of how the newly renormalized expression measures may be further improved. E.g. for the renormalized MAS5, there is clearly little need to work on the normalization aspect; modifications of the expression measure could instead aim at reducing the variability of MAS5 values, possibly by using the information in the MM probes as weights in the summary measure. Renormalized RMA and MBEI on the other hand still suffer from insufficient normalization; as we perform already normalization steps on both the probe and the probe set level, it seems promising to focus on the intermediate steps like the fitting of the multi-chip model and to study whether these steps are prone to systematic biases. Limitations ----------- The only condition for using the correlation test is a fairly large chip, with probes covering a wide range of the genome under study. For chips that are designed to study only a few related pathways or highly specialized tissues with only a couple of hundred probe sets, the zero correlation assumption may not hold, because the genes from which we want to sample randomly have already been pre-selected by the chip design. The example of the KEGG probe sets on the U133A chip suggests though that several thousand probe sets organized in a hundred and some pathways is a safe size. It should be pointed out that this approach is not limited to high-density oligonucleotide chips. The same argument for between-chip normalization holds in principle for cDNA or any other two-color microarray system, although the usual intensity-based normalization between dye channels on the same chip simplifies the situation somewhat \[[@B16]\]. Conclusion ========== We have presented a simple graphical method for assessing the quality of low-level analysis of oligonucleotide array expression data. The main advantage of our approach lies in the fact that we do not make use of external reference data, but instead exploit the internal correlation structure of large expression data sets. This allows us to select, evaluate, and modify low-level procedures for specific data sets. In order to demonstrate the use of and usefulness of our approach, we have applied it to three large data sets and three widely used low-level methods (MAS5, RMA, MBEI). We found a number of interesting results: a) For a large breast cancer data set, normalizing to housekeeping genes does not work at all, regardless of expression measure; b) normalization quality for all three data sets and all three expression measures is closer related to the absence/presence status of a probe set than to its intensity level; c) RMA and MBEI normalize absent probe sets poorly for all three data sets; d) removing pre-dominantly absent probe sets improves normalization for all data sets and all expression measures. The cutoff percentage of absent calls for a probe set to be included in the analysis can be chosen based on our graphical criterion. We have also evaluated the effect of a second round of normalization on the probe set level data. We found that this improved normalization significantly for all three data sets, in a manner consistent with the observed improvements in the detection of gene regulation \[[@B15]\]. Methods ======= Data ---- We used three data sets, two of which are publicly available from GeneLogic \[[@B17]\]. (1) The dilution data set is a collection of 75 HGU95Av2 chips, on which RNA from two different sources (liver and nervous system) was hybridized in different concentrations and mixture ratios. (2) The spike-in data set consists of 94 HGU95Avl chips, for which eleven bacterial cRNA fragments were added in different concentrations and combinations to a base sample from an AML cell line. Both of these data sets have been widely used for assessing normalization methods and expression measures \[[@B4],[@B7],[@B10]\]. In contrast the RNA for the third data set was extracted from tumor tissue collected from a population-based breast cancer cohort at Karolinska Hospital, Stockholm. After processing the RNA, several quality control steps, and screening the patients on medical criteria resulted in data from 159 HGU133A chips. Details on data preparation, patient selection, and the definition of clinical parameters like hormone replacement therapy are given in \[for personal communications see Hall P, Ploner A, Bjöhle J et al.\]. Expression measures and normalization methods --------------------------------------------- MAS5 expression values were computed as described in \[[@B5]\]. We used global mean normalization for the logged expression values as default, assuming that the mean across the logged expression values of all probes should be constant across all chips, and adjusting the level of each chip by adding a corrective constant to all probes. This is roughly equivalent to using the standard Affymetrix scaling factors on the raw data, but estimation of the corrective term on the log-scale has been found to be less variable \[[@B18]\]. The corrected log values were used for the analysis. For RMA, the individual PM probe values were background-corrected and quantilenormalized before computing the expression values, as described in \[[@B7]\]. The computation of MBEI expression values followed \[[@B6]\]: PM and MM values were normalized separately to a baseline array of average PM and MM intensities. The baseline array was obtained via smoothing an empirically identified set of rank-invariant probes. A multiplicative model was fitted to the difference between normalized PM and MM values. Expression values were logged, with non-positive values set to missing. Housekeeping gene normalization was based on the probes with suffix 2000\_ on the HGU133A chip. The same principle as with the global mean normalization was employed, except that the correction constant was based on the average of the housekeeping genes. For the MAS5 values, two variants were considered: a) multiplicative correction (scaling factor) of the un-logarithmized expression values, referred to as \'raw housekeeping\' in the legends for Figures [3](#F3){ref-type="fig"} and [4b](#F4){ref-type="fig"}) additive correction of the logarithmized expression values, as for the global mean normalization, referred to as \' log housekeeping\' in the figure legends. Housekeeping normalization for RMA and MBEI was done additively for the logarithmized expression values calculated from the unnormalized probe data; therefore, these are also addressed as \'log housekeeping\' values in Figure [3](#F3){ref-type="fig"}. The renormalization of the expression measures described in the Discussion was performed as in \[[@B15]\], using the iterated pairwise intensity-based normalization via smoothing loess curves described previously in \[[@B10]\]. The renormalization was performed at the probe set level on the expression measures computed and normalized as described above; correspondingly, renormalized RMA and MBE values have been normalized on both the probe level (originally) and the probe set level (second round), whereas MAS5 has been normalized twice on the probe set level. All computations were done using the open source statistical software package R \[[@B19]\] and the package affy of the Bioconductor project \[[@B20]\]. Calculation of correlations and summary curves ---------------------------------------------- For each data set, we randomly selected 5000 pairs of probe sets from the collection of probe sets available on the different chip types. For each pair, we computed the Pearson correlation coefficient between the two probe sets across all chips in the underlying data set, resulting in a random sample of 5000 correlation coefficients. At the same time, we calculated for each pair of probe sets the product of the two standard deviations across all chips in the data set; the scatter plots in the lower part of Figure [1](#F1){ref-type="fig"} show the resulting 5000 pairs of (product of standard deviation, correlation). The summary curves shown on top (like in Figure [1](#F1){ref-type="fig"}) or instead (all other Figures) of the point scatter describe the average behaviour of the scattered cloud. They were produced by taking the range of values for the product of the standard deviations in the sample and splitting it into intervals containing an equal number of observations, typically around 500. For each interval, the mean of the correlations was plotted against the median of the product of standard deviations. The 95% confidence intervals of the means shown from Figure [2](#F2){ref-type="fig"} onwards were computed based on normal approximation. A simple model for lack of normalization ---------------------------------------- We assume as experimental unit one microarray chip with the associated samples from the biological population under study. Each chip yields observations y~i~for i= 1 \... ngenes specified by the array design. We can write this as a random variable Y~i~= θ+ ψ~i~+ ε~i~ (1) where θis a random array effect, ψ~i~is a random gene effect, and ε~i~is the gene-specific measurement error. Note that this assumes random gene effects only in so far as we sample from the population, and we do not specify any treatment or experimental structure. We assume that the random components are independent, and that the errors have expectation zero. The covariance between the observable expression values for two genes then simplifies to ![](1471-2105-6-80-i1.gif) i.e. the covariance between the unobservable \'real\' gene expressions plus the variance of the array effect. Let\'s designate the variance of any Y~i~as ![](1471-2105-6-80-i2.gif). Now the correlations between the observable expression values for two genes can be written as ![](1471-2105-6-80-i3.gif) The first term is the contribution of the array effect to the correlation, which is the source of the correlation artifact. The second term varies across all possible pairs of genes, and we expect it to have zero average. Thus we get the inverse relationship ![](1471-2105-6-80-i4.gif) We can investigate this empirically. Given a set of microarrays, we can take a sample of random pairs of genes (i, j), then calculate their correlations r~ij~and standard deviations s~i~and s~j~. Under our hypothesis of zero average correlation and assuming that our simple model holds, the underlying pattern in the plot of (s~i~s~j~, r~ij~) should follow this inverse relationship. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Correlations between the unnormalized expression values of 5000 randomly selected pairs of genes. Top: The distribution of the correlation coefficients is centered far away from zero for all expression measures and data sets. Bottom: Scatterplots of the correlations versus the product of the standard deviations of the genes for the randomly selected pairs. Local mean correlations are shown in blue and indicate that correlations decrease with variability. Shown in red is a simple model for lack of normalization that was fitted to the data. ::: ![](1471-2105-6-80-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Correlations between the normalized expression values of the 5000 randomly selected pairs of genes. The most commonly used normalization procedure was chosen for each expression measure: global mean normalization for MAS5, quantile normalization for RMA, and invariant set normlization for MBEI, see Methods. Top: The distributions of the correlation coefficients are now centered at zero for all expression measures and data sets. Bottom: Summary curves plotting mean correlations versus median product of standard deviations for pairs of genes in non-overlapping intervals along the horizontal axis; 95% confidence intervals are shown as vertical bars. Normalization has removed most of the technical correlation (as seen in the reduced vertical scale compared to Figure 1), but there are still systematic relationships between correlations and standard deviations for RMA and MBEI, which are most pronounced for the Breast Cancer data set. ::: ![](1471-2105-6-80-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Correlations for the Breast Cancer data set normalized using a predefined set of 100 houskeeping genes. The MAS5 expression values were normalized both before and after taking the logarithm, corresponding to MAS5 (raw) and MAS5 (log). The RMA and MBEI expression values were normalized after caculating the logarithmized expression values, see Methods. Top left: The distributions of the correlation coeffcients are not centered at zero. Top right and bottom: Mean correlations versus median product of standard deviations are shown as in Figure 2, comparing the residual correlation for the housekeeping normalization and the default procedure for each expression measure (as indicated in the legends). Housekeeping normalization does much worse than the default in removing the systematic relationships. In case of the MAS5 values, both variants do worse than the default, and the normalization of the raw (un-logged) values doing worse than that of the logarithmized values. ::: ![](1471-2105-6-80-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Distributions of t-statistics calculated from differently normalized MAS5 expression values in the Breast Cancer data. The normalization procedures applied were the same as in Figure 3: global mean as default and housekeeping normalization before and after taking logarithms (raw and log respectively in the legend.) The t-statistics compare expression values between users and non-users of hormone replacement therapy (HRT) in the sample. The density curves show the smoothed histograms of the test statistics for all 22283 probe sets on the chip. The default curve is centered at zero, indicating balanced up- and down-regulation in reaction to HRT, whereas the curves for the housekeeping-normalized data are shifted to the left, indicating a massive down-regulation of thousands of genes in HRT users, which is biologically much less plausible. This effect is more pronounced for the raw than for the logged normalization, same as with the residual correlations shown in Figure 3. ::: ![](1471-2105-6-80-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Percentage of Affymetrix present calls across all chips for all three data sets. For each probe set, the percentage of present calls across all chips in its data set was calculated. The histograms show the frequencies of these percentages. For each data set, about 30% of all probes are absent on all chips, and between 20 and 30% are present on all chips, with the balance equally distributed between the extremes. ::: ![](1471-2105-6-80-5) ::: ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Correlations of randomly sampled pairs of genes by percentage of Affymetrix present calls. Top: Histograms of the average number of present calls for the 5000 pairs of genes, showing three peaks at the ends and in the middle of the range. The red lines seperate the peaks and indicate the grouping of the percentages used in the bottom part of the figure. Looking at Figure 5, we identify the groups with pairs of genes where both genes are mostly absent (left peak), where both genes are mostly present (right peak), and where one gene is mostly absent and the other mostly present (central peak). Bottom: Summary curves for the correlations as a function of the product of standard deviations. The different curves correspond to the three groups of pairs of genes as identified in the histograms above. Note how correlations are consistently positive for absent/absent pairs and consistently negative for absent/present pairs in case of RMA and MBEI. Only the present/present pairs appear to be reasonably uncorrelated throughout. ::: ![](1471-2105-6-80-6) ::: ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Distributions of t-statistics calculated from different expression measures for genes that are completely absent in the Breast Cancer data. The same default normalization procedures as in Figure 2 were applied. The t-statistics compare expression values between users and non-users of hormone replacement therapy (HRT) in the sample as in Figure 4. The density curves show the smoothed histograms of the test statistics for 4371 probe sets that had absent calls on all 159 chips. The MAS5 curve is centered at zero and close to a standard normal distribution expected approximately if no gene is differentially expressed between HRT users and non-users. The RMA curve is strongly shifted to the left, indicating wide-spread down-regulation of numerous absent genes in non-users; the MBEI curve shows moderate asymmetry and heavy tails at both sides, indicating more moderate, but still common gene regulation. Given the low signal quality of these genes that were classified as absent throughout the data set, small or no detectable regulation effects seem biologically most plausible. ::: ![](1471-2105-6-80-7) ::: ::: {#F8 .fig} Figure 8 ::: {.caption} ###### Correlations of the randomly sampled pairs of genes by percentage of present calls and mean intensity. The summary curves show the average correlation of the pairs of genes sampled from the breast cancer data. The differently colored curves correspond to the groups of absent/absent, absent/present and present/present pairs that have been defined in the text and are shown in Figure 6. Additionally, the pairs of genes have been arranged by their average mean intensity across chips: the range of average intensities was split into three intervals with an equal number of pairs, from the lowest third on the left to the highest third on the right. Grouping the pairs by their absence/presence status yields distinctly different patterns of correlations for RMA and MBEI, these patterns are however highly consistent between the intensities. ::: ![](1471-2105-6-80-8) ::: ::: {#F9 .fig} Figure 9 ::: {.caption} ###### Correlations of pairs of genes sampled from subsets with an increasing minimum percentage of present calls for probe sets. From the breast cancer data, 5000 pairs of genes were randomly sampled from the subset of probe sets that were present on at least one of the chips, the subset of probe sets that were present on at least 20% of the chips etc., finally from the subset of probe sets present on all chips. The three curves represent the three expression measures. Excluding the probe sets that are always absent already reduces the residual correlation considerably, excluding the probe sets that are more than 80% absent breaks the pattern of highest correlation at smallest variability for RMA and MBEI seen in Figures 2, 3, 6, and 7. Using only probe sets that are always present however leads to a notable increase in residual correlation, especially for MAS5. ::: ![](1471-2105-6-80-9) ::: ::: {#F10 .fig} Figure 10 ::: {.caption} ###### Correlations for 5000 random pairs of probe sets selected from KEGG pathways. Pairs were sampled from the subset of probe sets in the breast cancer data that appear in at least one KEGG pathway. Even for this smaller (n= 3137) and well-connected subset, the correlations are centered around zero. ::: ![](1471-2105-6-80-10) ::: ::: {#F11 .fig} Figure 11 ::: {.caption} ###### Correlations after renormalizing the expression measures using the pairwise loess normalization on the probe set level. Except for the second normalization step, this is identical to the lower half of Figure 2, including the scaling of the axis to facilitate the comparison. The MAS5 values appear now well normalized throughout, whereas RMA and MBEI still show a similar pattern of residual correlation, although on a clearly reduced level. ::: ![](1471-2105-6-80-11) :::	PubMed Central	2024-06-05T03:55:55.816620	2005-3-31	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084343/", "journal": "BMC Bioinformatics. 2005 Mar 31; 6:80", "authors": [ { "first": "Alexander", "last": "Ploner" }, { "first": "Lance D", "last": "Miller" }, { "first": "Per", "last": "Hall" }, { "first": "Jonas", "last": "Bergh" }, { "first": "Yudi", "last": "Pawitan" } ] }
PMC1084344	Background ========== Numerous genome-sequencing projects have caused a rapid growth of the protein databases. In contrast to the pre-genomic era, when the selection of sequences was highly biased towards known and characterized genes, the systematic exploration of genomes now allows to assign more and precise functional properties in the majority of cases. However, manual annotation of sequences is laborious and expensive. Thus, there is a strong interest in developing reliable methods for the automatic functional classification of genome sequences employing evolutionary sequences as reflected in using sequence homology to predict functional properties. The identification of protein families, defined as set of proteins with significant sequence similarity encoding for at least related but often identical function between members, is a very important subtask to achieve this fundamental goal. Indeed, the fact that proteins with high sequence similarity share a common evolutionary history is accepted as the basis for functional assignment \[[@B1]\]. Among the different methods proposed to organize the sequence space into protein families, several approaches based on clustering using sequence similarity scores were successfully established (see e.g., \[[@B2]-[@B4]\]). However, the multi-domain composition of proteins, as well as the presence of promiscuous domains can influence the accuracy of such methods. Recently, an efficient algorithm for large-scale detection of protein families based on the Markov cluster algorithm, TRIBE-MCL, was proposed \[[@B5]\]. This algorithm simulates random walks within a graph by iterative alternation of two operators called expansion (explores intra-cluster structure) and inflation (eliminates flow between different clusters). In comparison to other clustering algorithms, the TRIBE-MCL produces clusters that resist contamination by promiscuous domains that could provide significant problems for other clustering algorithms as is discussed elsewhere \[[@B6]\]. TRIBE-MCL was tested using large databases of manually annotated protein sequences such as SwissProt \[[@B7]\] and SCOP \[[@B8]\] and has already been widely used in bioinformatics (about 50 publications referred to this algorithm since its publication in 2002\[[@B5]\] according to \[[@B9]\]). Thus this method is one of the recognized bioinformatic tools and its results can be used as an established standard for comparison of new algorithms. Moreover, we have already used the TRIBE-MCL algorithm for the analysis of the SIMAP database \[[@B10]\]. Several clustering methods have appeared in recent years. One of these, the Super Paramagnetic Clustering (SPC) has received considerable attention in microarray data analysis \[[@B11],[@B12]\]. This algorithm provides clustering of input data \[[@B13]\] based on analogy to the physics of an inhomogeneous ferromagnet. The method detects natural (physical) clusters present in the data and is able to efficiently cluster difficult test examples, such as concentric circles. The SPC algorithm was also successfully used in a supervised setting to analyze protein sequences and classify SCOP and CATH proteins according to their FSSP scores \[[@B14]\]. Following our first successful application of SPC to a database of RING-finger domains \[[@B15]\] and our approach to project expression data to known functional modules \[[@B16]\], the present study further investigates the power of SPC to cluster protein sequences of two large databases, SwissProt and SCOP. We compare its performance with the TRIBE-MCL algorithm. Since both these databases do not contain complete genome sequences required for an unbiased comparison of the methods, we additionally analyzed protein sequences from four bacterial genomes, namely Bacillus subtilis, Helicobacter pylori, Listeria innocuaand Listeria monocytogenesmanually annotated at MIPS according to FunCat \[[@B17]-[@B19]\]. We also introduce an extension of this algorithm, global SPC or gSPC, which performs step-wise clustering on different levels of connectivity between points and provides significantly improved performance to the annotation of whole genomes compared to both the original SPC algorithm and TRIBE-MCL. Results ======= Clustering of a simulated data ------------------------------ We tested the ability of the SPC method to determine the physical number of clusters in the data using synthetic data. The model problem of Figure [1](#F1){ref-type="fig"} consisted of n= 60 points in D= 2 dimensions. The data points were generated using three normal distributions N(x~i~, σ = 1.5), with centers x~0~(1; 2) (n= 60 samples), x~1~(10; 2) (n= 30) and x~2~(14; 2) (n= 30). The data points generated by the second and third normal distributions are overlapping. In addition three points, x= (4.5 + 1.5\j;2), j = {0,1,2}were added to simulate an artificial link between the data points from the 1^st^and 2^nd^distributions. Figure [1](#F1){ref-type="fig"} demonstrates that SPC (K= 10) was able to correctly determine the presence of three clusters in the data. Two splits of clusters at temperature T= 0.054 and T= 0.084 are observed. The first split corresponds to a separation of clusters formed by the 1^st^and two other distributions. The second split corresponds to a separation of clusters formed by 2^nd^and 3^rd^distributions. Following these two breaks, the cluster melts on singletons. Thus, the hierarchical structure of data was uncovered and physical clusters present in these data were found. The noise between the data points from distributions 1 and 2 (green circles) did not affect the clustering results. In contrast to SPC, the TRIBE-MCL algorithm has some difficulties in correctly determine the structure of the data. For example, for inflation parameter 2.1, the algorithm subclusters points generated by normal distributions 1 into 2 different subclusters. For the inflation value of 5, one can already observe 5 different clusters. One of the largest clusters contains 21 points, including 6 and 15 points generated by distribution 2 and 3, respectively. This cluster remains stable even for inflation parameter 20. Thus TRIBE-MCL could not detect the physical structure of this data set. Of course, one should not draw a general conclusion about the relative performance of both algorithms following only a single simulated example. Comparison of algorithms using SwissProt and SCOP databases ----------------------------------------------------------- The performance of the algorithms was investigated using SwissProt \[[@B7]\] and SCOP \[[@B3],[@B8]\] databases. The accuracy of SPC clustering for the SwissProt database was assessed by analysis of InterPro domains \[[@B20]\] and Swiss keywords of members in calculated clusters. Sequences without any annotation were used for data clustering but were not considered to estimate the performance of the methods. Ideally, all members of each detected cluster should have exactly the same annotation in terms of InterPro domains and Swiss keywords. Analogous to the previous analysis \[[@B5]\] only clusters that contained at least 4 or more annotated sequences were considered. The domain (or keyword) combination detected for ≥50% annotated sequences in the cluster was used as the (consensus) annotation of the cluster. Since some proteins had more than one domain (or keyword), we measured the performance of the method by counting the number of correctly assigned domains rather than the number of correctly classified proteins. This procedure avoids ambiguities in cases where, for instance, the annotation of three out of five domains was predicted correctly. The number of true positive (TP) domains/keywords was determined as the count of domains/keywords that coincided with the cluster annotation. The number of false negatives (FN), i.e. domains/keywords observed for a particular protein but absent in the cluster annotation, and false positive (FP), i.e. domains/keywords presented in the cluster annotation but missed for some particular proteins, were calculated. These numbers were used to compute the sensitivity = TP/(TP+ FN) and specificity = TP/(TP+ FP) of the clustering algorithms analyzed. The sensitivity is equivalent to the probability of correctly predicting some classifier while specificity is defined as the probability that the provided prediction is correct \[[@B21]\]. Not all proteins initially used in the evaluation will get a chance to be annotated by clustering. Some proteins will not be clustered at all, because they do not have significant hits to other proteins. These proteins can be either treated as false negatives (indeed, their categories were not predicted) or simply excluded from the analysis (since they cannot be clustered and the clustering algorithm explicitly \"refused\" to annotate them). The sensitivity that is calculated taking clustered proteins appears to be more relevant. Indeed, if a protein was clustered, the sensitivity determined in our study will indicate how many of the existing categories of the protein analyzed are expected to be correctly predicted. This definition deals with a posteriorisensitivity, i.e. it should be used only after clustering of the protein families. The sensitivity determined by considering all non-clustered proteins as false positive, corresponds to a priorisensitivity. Indeed, this number indicates how many categories of the given protein will be correctly predicted when there is no knowledge if a protein will be clustered or not. Since each of these two definitions of sensitivity has its own advantage (e.g. the later allows for a more straightforward comparison of methods) we calculated them both. Notice, that this definition of specificity does not include FNand thus it is not affected by which definition is used. SCOP database analysis ---------------------- Sequences from the PDB database \[[@B22]\] (Release from 01/07/2003) were clustered after removing redundant entries. These sequences were annotated using the SCOP database v. 1.63 \[[@B3],[@B8]\]. The TRIBE-MCL results were calculated using the inflation value of 5 \[[@B5]\]. The total number of proteins used for analysis was 15,605 and 12,961 of the sequences had assigned SCOP domains. The total number of manually curated domains was 13,070 domains. Both annotated and non-annotated proteins were used for clustering. Obviously the method performance was calculated only for the annotated cases. The SPC covered 6% fewer sequences for K= 20 but resulted in higher a posteriorispecificity and sensitivities (Table [1](#T1){ref-type="table"}). The TRIBE-MCL, however, resulted in higher a priorisensitivity. For this value of K, both the SPC and TRIBE-MCL clustered data into approximately the same number of clusters. Larger values of the parameter Kfurther increased the number of covered sequences but decreased the performance of SPC. For example, an increase of Kfrom 20 to the use of all connections (\"all NN\") covered an additional 330 sequences. The number of true positive predictions increased by 183. However, this increase was accompanied by an additional 173 false negative and 147 false positive predictions, thus decreasing the overall performance. Not all sequences were identical for both cases. The \"20\" clusters contained 300 sequences that were absent in \"all NN\" clusters. Correspondingly, there were 630 sequences that were present in \"all NN\" clusters but were absent in \"20\" clusters. The performance of the algorithm for these 630 new sequences was 569 true positives, 63 false negatives and 61 false positives. This corresponds to 90% in sensitivity and specificity. Thus, the performance of the clustering method using all connections did not dramatically decrease due to the addition of the new sequences, but rather due to worse prediction of some sequences that were already clustered using K= 20. Therefore by joining results calculated using variable Kvalues and preserving results calculated for the sequences that were clustered in each preceding step, one can expect to increase both sensitivity and specificity of the method. Indeed, the use of the gSPC method provided a considerable increase in clustering performance. The number of false positive and false negative for the \"all NN\" clusters was lower than the numbers calculated using fixed value of K= 20, but as many as 610 new sequences were covered. The gSPC method outperformed the TRIBE-MCL and SPC in terms of both a prioriand a posteriorisensitivities. Thus the performance of the gSPC was considerably better than the other methods both in terms of covered sequences and quality of annotation. This improvement is of great importance for an automatic annotation of protein sequences. SwissProt analysis ------------------ The InterPro domains (v. 6.1) covered 112,935 (89%) sequences from the SwissProt database Release 41.9 (126,798 sequences). The total number of 235,672 domains was calculated for this set. The TRIBE-MCL clustered 97,792 sequences into 6,200 clusters that contained at least 4 annotated proteins (Table [2](#T2){ref-type="table"}). The consensus annotation provided 93.8% of specificity for these data. A similar specificity of SPC, 94.1%, was calculated for K= 20 nearest neighbors. The number of covered sequences by the SPC algorithm was 96,716. The use of the gSPC made it possible to cover about 7,000 additional sequences with overall specificity and \" a posteriori\" sensitivity of 95.2% and 94%, respectively. The SwissProt database provides a controlled vocabulary of 878 keywords that has been used by many researches to test different classification algorithms. The total number of 490,065 keyword instances was assigned to 125,248 proteins. As for the analysis of the InterPro domains, the gSPC analysis provided higher performance and covered a larger number of protein sequences compared to the use of any single K-value or TRIBE-MCL algorithm (Table [3](#T3){ref-type="table"}). Comparison of algorithms for annotation of bacterial genomes ------------------------------------------------------------ For this analysis 11,502 protein sequences from four completely sequenced genomes, Bacillus subtilis, Helicobacter pylori, Listeria innocuaand Listeria monocytogenes, were used. The annotation of the genomes was done at MIPS using FunCat 1.3 \[[@B17]\]. The FunCat is an annotation scheme for the functional description of proteins from prokaryotes, unicellular eukaryotes, plants and animals \[[@B17],[@B18]\]. Taking into account the broad and highly diverse spectrum of known protein functions, FunCat consists of 28 main functional categories (or branches) that cover general fields like cellular transport, metabolism and signal transduction. The main branches exhibit a hierarchical, tree like structure with up to six levels of increasing specificity. In total, the FunCat 1.3 included 1,445 functional categories and a total number of 403 distinct categories were available for analyzed bacterial genomes. The manual functional classifications were presented for 6,354 proteins. An estimation of performance for proteins that have similar but not exact annotation represent some challenge. Let us consider an example of a cluster containing three proteins. The first protein that has annotation 01.01.01.01.02, biosynthesis of the glutamate group 01.05.01, C-compound and carbohydrate utilization 40.03, cytoplasm the second protein has annotation 01.01.01.01.02.01, biosynthesis of proline 40.03, cytoplasm and the third protein is annotated with one category only: 01.01.01.01.02, biosynthesis of the glutamate group The annotation of all three proteins is similar but the annotation of the second protein is more detailed for the metabolism (01) category while the first protein contains additional category 01.05.01. The third protein does not have an annotation category 40, subcellular localization. We measured the performance of annotation by counting the number of all non-redundant subcategories, i.e. 01, 01.01, 01.01.01, etc. In the above example the consensus annotation of this cluster is 01.01.01.01.02 and 40.03 categories (including all their sub-categories). The number of TPannotations is 19 = 7 + 7 + 5. The number of FNis 3. These are sub-categories 01.05, 01.05.01, and 01.01.01.01.02.01, for the first and second proteins, respectively. There are also two FPannotations, 40 and 40.03 observed for the third protein. The performance of the three algorithms is shown in Table [4](#T4){ref-type="table"}. The total number of non-redundant subcategories for this analysis was 44,531. The methods calculated similar performance, but the gSPC algorithm covered a larger number of sequences. Therefore the performance of gSPC was remarkably higher in terms of a priorisensitivity. Table [5](#T5){ref-type="table"} summarizes the comparison of the three methods in terms of covered sequences and total error. Overall, the use of the gSPC algorithm resulted in higher performance for all examples and covered a larger number of sequences. Discussion ========== We have described and demonstrated the use and performance of SPC and of its extension, gSPC, for the clustering of protein sequences using sequence similarity only. For the first time, the SPC algorithms for clustering of protein sequences was employed in a large-scale study. Our results confirm that this method is a valuable, reliable tool for the automatic functional classification of protein sequences. The use of the step-wise clustering approach, gSPC, improved sensitivity and specificity of the original method and allowed us to get a higher accuracy compared to the TRIBE-MCL and original SPC algorithms using data sets from PDB and SwissProt databases. The performance of the gSPC for annotation of Swiss Keywords favorably compares with supervised approaches. For example, the percentage of sequences covered by gSPC for 93.7% (K= 6) and 93% (K= 20) specificity were 62% and 79% respectively. The first number is similar to that calculated by the supervised classification approach using the C4.5 algorithm, where 58% of sequences were covered with 94% of specificity \[[@B23]\]. Thus the SPC algorithm classification performance is comparable to the state-of-art supervised machine learning classification approach used by \[[@B23]\]. Notice, that the latter method used different sequence attributes, such as PFAM domain and PROSITE composition, and thus explicitly took into consideration the multi-domain organization of proteins. The SPC algorithm, contrary to that, used the sequence similarity only. This result indicates high prediction ability of the annotation using gSPC clustering. The specificity of the gSPC algorithm using all connections (\"all-NN\"), 92.8%, is also slightly superior to the results of another supervised approach, the adaptive algorithm for automated annotation \[[@B24]\] that calculated 90.4%. These results, however, are indicative only since our and previous approaches are different with respect to their types (i.e. supervised and unsupervised ones). In addition similar but not identical datasets were used. For example, the performance of the adaptive algorithm \[[@B24]\] was tested using 500 probe sequences, randomly chosen from SwissProt while the performance of the C4.5 algorithm was tested using 10-fold cross-validation. A significant advantage of unsupervised clustering approaches over the supervised ones is the ability of the former methods to detect as yet unobserved relations between proteins. Such results could be used to find new protein families that currently do not have functional annotations or have incomplete or inconsistent ones. The unsupervised methods are also not subject to the risk of overfitting problems. Overfitting impairs the predictive power of supervised approaches for samples that were not represented in the training set \[[@B25]-[@B27]\]. The gSPC resulted in higher sensitivity and specificity for all analyzed datasets. The use of a priorisensitivity made it possible to compare results of all methods using a fixed number of samples and made the comparison more straightforward. However, from a practical point of view, particularly for the implementation of annotation pipe lines of complete genomes the specificity and a posteriorisensitivity rather than a priorisensitivity are the most relevant factors of the automatic annotation. Indeed, it is preferable to annotate a smaller number of sequences in automatic mode with a higher accuracy than to annotate more of them with a lower accuracy. The sequences not automatically annotated by the algorithm can be annotated subsequently by complementary approaches or by careful manual analysis of domain structure of the sequences. Any attempt to increase the a priorisensitivity and thus cover a larger number of sequences by lowering specificity may result in an unacceptable performance for annotation purposes. The SPC algorithm calculates hierarchical tree-structures of clusters for each Kvalue. In our analysis we identified and considered only the leaf clusters. The upper part of the tree was ignored. Such analysis was possible since for the data analyzed in this study the largest number of protein sequences were clustered in leafs and only a few additional sequences could be still clustered considering the whole tree structure. For example, using K= 6 (Table [1](#T1){ref-type="table"} &[2](#T2){ref-type="table"}, SPC results) only 56 and 627 additional sequences could be clustered for the PDB and SwissProt data sets, respectively. These numbers corresponded to about 1% sequences in each database and only marginally influenced the method performance. The gSPC algorithm, as it was already mentioned in the Method section, clustered such sequences at higher Kvalues. This improved its prediction performance compared to the original SPC method. The performance of any algorithm analyzing sequence relations depends on the selection of the reference data sets. The composition of such references data sets could bias the performance, since in reality the selection of sequences from known genomes or databases is not a representative random sample from the sequence space. For example, SwissProt and SCOP databases are very often used as \"a gold standard\" for annotation and classification methods. However, these databases represent a biased selection and do not cover entire genomes. Therefore, analysis of the performance of clustering methods is biased by the composition of the reference data sets. For example, the gSPC decreased missassignments by 2--5% for these two sets. Since the error rate for clustered sequences from these data sets was about 10--15% (Table [5](#T5){ref-type="table"}), the relative gain in the performance was 10--40%. In other words, the automatic annotation of sequences clustered with gSPC will make up to 40% fewer erroneous annotations (false positive or false negative annotations) compared to other methods. On the other hand, all three methods returned similar results for the bacterial genomes. For this set gSPC covered about 12% of additional sequences which is very important for genome annotation. The gSPC method developed is fast. A complete analysis of PDB and SwissProt datasets using all Klevels took on moderate PC computer (AMD 1.3 GHz) less than 14 and 120 minutes, respectively. The speed of the original SPC algorithm scales approximately linearly with the number of connections. This number increases approximately as N^2^with the number of samples. However, since gSPC uses step-wise clustering, the actual number of samples remaining for clustering using large Kvalues is small. Therefore, the gSPC speed is mainly determined by clustering using small, K= 2--6, values and is in practice approximately linear with the number of samples. In fact, the method will be fast enough to efficiently cluster datasets with millions of sequences; clustering computational requirements is therefore small compared to the computation of the sequence similarity scores. The computational efficiency is clearly an advantage of the gSPC method compared to the TRIBE-MCL. The complexity of the later scales as O(N^2^) were Nis number of non-zero elements in the matrix. Therefore, this method performs slower if an increasing volume of genomic data needs to be processed. Conclusion ========== The gSPC calculated with higher accuracy or covered a larger number of sequences than the TRIBE-MCL algorithm for the analyzed datasets. The accuracy of annotation of gSPC for the SwissProt database was comparable to that of supervised methods. Thus it is a useful approach for automatic detection of protein families and annotation of full genomes. Methods ======= Clustering of sequences using the properties of super paramagnetic systems -------------------------------------------------------------------------- SPC performs a hierarchical clustering strategy \[[@B13]\]. The input data for SPC are represented as a distance matrix d~ij~(specified by the user) between data points i = 1,\...,N. This matrix is used to construct a graph, whose vertices are the data points and edges corresponding to the connections between neighboring points. Each two points are considered to be neighbors (and thus have an edge), if they are within Knearest neighbors of each other (K-mutual-neighbor criterion). In the ferromagnetic model each point is considered to have a Potts spin, equivalent to one of qinteger values, s= 1,2,\...q. Pairs of neighboring points iand jthat have the same spin s~i~= s~j~are interacting with strength J~ij~, which is a function of an initial distance matrix d~ij~\[[@B13]\]. A Monte Carlo simulation using the Swendsen-Wang method (SW) \[[@B28]\] is used to determine clusters of points. The simulation starts with assigning a random Potts spin to each data point in the dataset. The neighboring points with identical spins interact with each other. The probability that these points belong to the same SW cluster (i.e., instant cluster resulting from an iteration) is calculated as P~ij~= (1 - exp(-J~ij~/T)), where Tis some fixed temperature. The points from the same SW cluster receive identical spin (selected by chance) for the next iteration. The iterations are performed until convergence observed using, e.g. autocorrelation time \[[@B29]\]. For clustering, the strengths of the edges of the network are calculated using the spin-spin correlation function G~ij~. This function is estimated as the ratio of iterations when points i,jbelong to the same SW cluster versus the total number of iterations during the simulation. Notice, that if P~ij~values determine an instant probability of two points with the same spin to belong to the same instant cluster (i.e., local connectivity), the spin-spin correlation function takes into account the multiple interactions amid points, i.e. global connectivity of the network. Actually the global connectivity determines the probability two points will have the same spin. At very low temperatures even small J~ij~cause neighboring points to have the same spin, G~ij~≅ 1, i.e. the system is in the ferromagnetic state. If the temperature is very high then all J~ij~\<\<Tand the probability of finding each pair of neighboring points in the same state decreases to the value expected by chance G~ij~≅ 1/q. Thus the system is unordered, i.e. it is in the paramagnetic state. If coupling values are equal J~ij~= J, only two these states are stable. The system has a sharp transition from the ferromagnetic to the paramagnetic state at a certain temperature. In case of an inhomogeneous system, some magnetic \"grains\" could be observed. Such clusters of points have strong connections between the members and only weak connections to all points outside the cluster. Each grain has its own transition temperature from the ferromagnetic to the paramagnetic state. At some temperature range, the so-called superparamagnetic state, the system can contain some points in a locally ordered ferromagnetic state and as well as others in a locally unordered paramagnetic state. At the end of simulation, all points that had spin-spin correlation G~ij~\> 0.5 form one cluster. The points that had G~ij~\< 0.5 can be connected to their neighbors with maximal G~ij~. The results of the analysis at different temperatures are combined and provide a hierarchical clustering of data. A more detailed description of the SPC can be found elsewhere \[[@B13]\]. Selection of free parameters of SPC algorithm --------------------------------------------- Free parameters of the SPC algorithm include the number of spins qin the Potts model, the number of nearest neighbors Kfor the K-mutual-neighbor criterion, and the number of iterations Ifor the SW algorithm. An increase of qrequired larger numbers of iterations for convergence but otherwise did not affect the performance of the algorithm \[[@B30]\]. For q= 20 the convergence of SW algorithm was usually observed using I= 1000 iterations for the vast majority of the cases investigated. The same values also resulted in convergence of the algorithm for the datasets used in our study. As described above, SW simulations start with random spin configuration. After subsequent iterations, the algorithm converges into a stable state that does not depend on the initial configuration. Given convergence at a certain temperature, one can expect that the configuration should be similar to the converged state of the next step in respect to the increased temperature. Thus if SW simulations are started from a converged state calculated for a similar temperature, the convergence to the new state should be faster. Without loss of precision, we used this fact to speed up SW simulations. The configuration after completion of each simulation cycle was saved and used it as the initial one for the next temperature. It was possible to decrease the number of iterations from I= 1000 to I= 200 without affecting the quality of the results. Global clustering and the gSPC algorithm ---------------------------------------- The K-mutual-neighbor criterion controls the resolution of the clustering. A low Kvalue takes into consideration only the closest pairs of mutually connected points (\"i.e. short connections\"). As its values increases, long-range connections are also taken into consideration and they participate in the clustering process. In previous studies this parameter was selected in the range of 5--20 \[[@B13],[@B30]\]. In a recent study \[[@B31]\] an algorithm to determine optimal Kvalues was also proposed. The optimal value of Kcorresponds to the maximum stable partitions characterized by maximum widths of the superparamagnetic domains. The optimization of Kis critical in absence of reliable classification information to allow for supervised learning. In this study, reliable protein family assignment for test and training data was available. Thus, it was possible to directly analyze the performance of the SPC algorithm as a function of the parameter K. Using a constant value for Kis reasonable if the data set is rather homogeneous. In this case, the data points have the same connectivity on average. In reality, the problem of sequence based family assignment is complex. Some protein families apparently underlay constraints concerning the variation of their sequence and contain a large number of highly similar sequences while other families contain only few members displaying low sequence similarity. Using Kvalues optimal for conserved families may bias the clustering process and lower the sensitivity to discover the more diverging ones. Thus, it is inappropriate to use a single K-value for clustering of all data. In this article we introduce a step-wise analysis using different K-values. The data were consequently clustered using K= 2, 6, 20, 64 and K= \"all\" connections. This choice of Kapproximately corresponded to a 10-fold increase in the data connectivity that is proportional to K^2^. For each Kwe varied temperature from a minimum, T= 10^-4^, to a maximum, T= 2, values. At the maximum value of Tno cluster was observed. As soon as a subset of sequences was clustered using some K, they were removed from any further analysis except one representative sequence. The representative sequence was the one that had the minimal sum of distances to the other sequences in the cluster. These representative sequences as well as all other remaining non-clustered sequences were used for the next round, i.e. larger K, of analysis. Such bottom-up clustering allowed us to cover different types of clusters covering a wide range of homogeneity and to improve the performance of the method as shown in the Result section. In order to distinguish the initial SPC algorithm from the step-wise based approach proposed in this study, we will refer to the latter as the global SPC (gSPC). Processing of hierarchic data ----------------------------- The SPC algorithm provided a hierarchical clustering of data on each step of the gSPC analysis. Let us consider the clustering procedure of increasing the temperature. For T= 0 all sequences form one cluster that starts to melt and break into subclusters as the temperature increases. Clusters of different sizes can appear and then melt at higher temperatures. Some clusters may contain only 2--3 sequences while other clusters consist of more than a thousand sequences. A number of clusters will lose one or several members for some particular temperatures while others will break into two or more sub-clusters. In case the data set contains only few clusters, one can inspect the cluster size or susceptibility, that is the variance of magnetization of the system as a function of temperature \[[@B13],[@B30]\]. If different stable sub-phases are determined, e.g. by changes in the cluster sizes, it is possible to select a particular temperature in each region of interest for further analysis. Such a method, however, is unfeasible for the purpose of our study when as many as thousands clusters can be simultaneously observed for each particular temperature. Thus, the cluster composition for each temperature step should be considered for the analysis and some automated approach for the analysis of such results should be performed. In the following paragraphs we describe a method that performs such an analysis in a completely automatic way. Let us limit a minimal size of any cluster to some parameter C. All points in clusters that have size less than Cmembers were considered as unclustered points or singletons. The singletons did not form separate clusters but belonged to the cluster from which they were generated. At some particular temperature a cluster can break (melt) into a number of subclusters and singletons. However, if and only if these subclusters contained at least two subclusters each having at least Csequences, the parent cluster was considered to be subdivided into subclusters. Otherwise the cluster was considered to be unchanged. A leaf cluster was a cluster that did not form subclusters (i.e., it broke into singletons). The singletons were clustered again by the same procedure for larger K. In other words, the result of the SPC clustering using a single Kwas a hierarchical tree of clusters (or several disjoined trees) generated by analyzing the data points over the whole temperature range (from minimal to maximal temperature). The tree was processed to detect the leaf clusters, which were identified at different temperatures. Figure [2](#F2){ref-type="fig"} illustrates an example of the detection of leaf clusters for a single Kand Figure [3](#F3){ref-type="fig"} demonstrates processing of data for different K. The data flow of the algorithm (see Figure [4](#F4){ref-type="fig"}) is summarized as follows: In the very beginning the sequence similarities are downloaded from the SIMAP database as FASTA scores. Then, the algorithm clusters sequences using K= 2. During the analysis the temperature is scanned from a minimum T= 10^-4^(when practically all points form one large cluster) to a maximum T= 2 value (when almost all points are singletons). The leaf clusters are detected and used to calculate the performance of the algorithm for this K(see Tables [1](#T1){ref-type="table"},[2](#T2){ref-type="table"},[3](#T3){ref-type="table"},[4](#T4){ref-type="table"}). The sequences from the leaf clusters are then eliminated from further analysis except the representative ones (see above). Subsequently, we increase Kand perform the next round of analysis. The analysis is terminated when the maximal specified Kvalue is reached. The representative sequences may be assigned to some clusters with different Kin later rounds of iteration. The annotation of these sequences in the cluster with smallest Kwas used to count the method performance. The gSPC procedure, as it was correctly noticed by the anonymous reviewer, also depends on the temperature step used to analyze clusters for each single K. For example, using large steps, ΔT = 0.3, one would detect different leaf clusters in Figure [2](#F2){ref-type="fig"}. As it is clear from the same figure, the sensitivity to this parameter is not a particular feature of gSPC but of the original SPC method that is used to determine the leaf clusters. Thus, the identification of leaf clusters can be ambiguous in some cases. In our article we used ΔT = 0.01 and did not observe significant changes in the performance of the algorithm for smaller values of this parameter. TRIBE-MCL algorithm ------------------- The TRIBE-MCL algorithm was downloaded from \[[@B32]\]. The values of the inflation parameter used for analysis of data were selected as indicated in the original study \[[@B5]\]. Data representation ------------------- A FASTA file containing sequences that were used for clustering was compared to itself using BLAST \[[@B33]\] for the SwissProt database implemented in PEDANT \[[@B34]\]. For the bacterial genomes the FASTA pair-wise scores for were retrieved from the SIMAP database \[[@B10]\]. The all-against-all sequence similarities generated were parsed and used as input for both algorithms. The input data for TRIBE-MCL were represented as -log~10~(E-value). The input values for the SPC were distance values calculated as -1./log~10~(E-value). Pairwise scores with E-value\> 0.1 were excluded from the analysis. List of abbreviations ===================== PEDANT -- Protein Extraction, Description and ANalysis Tool FunCat -- MIPS Functional Catalog TRIBE-MCL -- Markov Cluster Algorithm SPC -- Super Paramagnetic Clustering gSPC -- global SPC SIMAP -- Similarity Matrix of Proteins Authors\' contributions ======================= IVT programmed the algorithm and performed data analysis. AF developed graphical interface of the algorithm. AR annotated and provided the data for bacterial genomes and participated in their analysis. HWM conceived and supervised the project. All authors read and approved the final manuscript. Acknowledgements ================ We thank the MIPS annotation group, Goar Frishman, Barbara Brauner, Gisela Fobo, Irmtraud Dunge and Corinna Montrone, for annotation of bacterial genomes. We also thank Sabina Tornow who introduced us the SPC algorithm, Louise Riley and anonymous reviewers for valuable remarks. This work was supported in part by grant 031U212C BFAM (BMBF) to HWM and grant TE 380/1-1 (DFG) to IVT/HWM. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Clustering of Artificial Data Set using SPC and TRIBE-MCL algorithm. The red, blue and yellow colors correspond to three distributions used to generate the data. The noise between the data points from distributions 1 and 2 is indicated as green circles. A) Cluster size plot of SPC algorithm. The vertical line indicates temperature when calculation of the distance matrix B) was performed. B) The distance matrix calculated for the SPC clusters at temperature T= 0.087. More intense colors correspond to smaller distances between points. The diagonal and off-diagonal elements correspond to inter- and intra- cluster distances, respectively. Each horizontal block on the left side of Figure corresponds to one cluster and the colors are used to indicate composition of samples from different distributions. C) and D) are the distance matrices calculated for the TRIBE-MCL clusters obtained using inflation parameter 2.1 and 5.1, respectively. ::: ![](1471-2105-6-82-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Detection of leaf clusters shown as ellipsoids. The minimal number of points in the leaf cluster is C= 3. The initial clusters contain 10 points each at T= 0.01 for both panel A) and B). On panel A) the cluster melts on two leaf clusters shown by ellipses. On panel B) the leaf cluster is the initial cluster. The leaf clusters are identified at different temperatures. ::: ![](1471-2105-6-82-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### gSPC processing of simulated data using different K. The clusters detected for different Kare shown as ellipsoids. On the first level K= 2, three most compact clusters are detected by the algorithm. The representative cases from these clusters will stay for further step of analysis, K= 6. For K= 6 two clusters are detected and, eventually for K= 20 one cluster will be detected. Thus the gSPC algorithm detects embedded clusters. The representative cases are used to restore hierarchical organization of clusters shown at the bottom of the figure. Notice, that for the upper clusters, the same sequence is the representative one for K= 2 and K= 6. ::: ![](1471-2105-6-82-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Data-flow of the gSPC algorithm. ::: ![](1471-2105-6-82-4) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Clustering of PDB sequences using SPC, gSPC and TRIBE-MCL algorithms ::: K* Cases clusters true positive false positive false negative specificity a posteriorisensitivity a priorisensitivity ----------- ------- ---------- --------------- ---------------- ---------------- ------------- --------------------------- ----------------------- SPC 2 2472 479 2466 46 18 98.2 99.3 18.9 6 7332 1079 7107 276 274 96.3 96.3 54.4 20 8666 875 8324 413 401 95.3 95.4 63.7 all NN^1^ 8996 740 8507 586 548 93.6 93.9 65.1 TRIBE-MCL 9208 964 8654 510 614 94.4 93.4 66.2 gSPC 6 7432 880 7252 277 239 96.3 96.8 55.5 20 8961 233 8709 377 314 95.9 96.5 66.6 all NN^1^ 9276 28 9009 392 329 95.8 96.5 68.9 1- the SPC analysis was performed using the complete similarity matrix and thus all Nearest Neighbors (NN) participated to the algorithm training. ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Analysis of InterPro domain composition in clusters calculated for SwissProt database ::: K cases clusters true positive false positive false negative specificity a posteriorisensitivity a priorisensitivity ----------- -------- ---------- --------------- ---------------- ---------------- ------------- --------------------------- ----------------------- SPC 2 18960 4423 39414 1239 988 96.7 97.6 16.7 6 78441 11418 159201 6045 8474 96.3 94.9 67.6 20 96716 6635 185012 11514 20464 94.1 90.0 78.5 64 98568 3739 170364 18864 36143 90.0 82.5 72.3 all NN 91452 3420 155239 16206 32472 90.6 82.7 65.9 TRIBE-MCL 97792 6755 191500 12764 16939 93.8 91.9 81.3 gSPC 6 79406 9543 162210 6208 7709 96.3 95.5 68.8 20 100458 2803 201766 9634 12402 95.4 94.2 86.6 64 103585 427 207131 10312 13264 95.3 94.0 87.9 All NN 103729 30 207339 10362 13304 95.2 94.0 88 ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Analysis of SwissProt keywords composition in clusters calculated for the SwissProt database ::: K cases clusters true positive false positive false negative specificity a posteriorisensitivity a priorisensitivity ----------- -------- ---------- --------------- ---------------- ---------------- ------------- --------------------------- ----------------------- SPC 2 19161 4473 78782 4421 3373 94.7 95.9 16.1 6 79642 11643 308068 20598 261446 93.7 92.2 62.9 20 98276 6875 354980 35282 53628 91.0 86.9 72.4 64 100177 3953 331422 47215 80772 87.5 80.4 67.6 all NN 93601 3601 308433 41834 74472 88.1 80.6 62.9 TRIBE-MCL 99636 7015 364554 36783 49333 90.8 88.1 74.4 gSPC 6 80617 9755 314537 21080 23838 93.7 93.0 64.1 20 101805 2857 388434 29379 35042 93.0 91.7 79.3 64 105248 445 400100 31061 37325 92.8 91.5 81.6 all NN 105339 19 400363 31142 37391 92.8 91.5 81.7 ::: ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Clustering of sequences of bacterial genomes using SPC, gSPC and TRIBE-MCL algorithms ::: K cases clusters true positive false positive false negative specificity a posteriorisensitivity a priorisensitivity ----------- ------- ---------- --------------- ---------------- ---------------- ------------- --------------------------- ----------------------- SPC 2 646 157 5365 357 99 93.8 98.2 12.1 6 4652 794 34699 2635 2624 92.9 93.0 78.2 20 5072 637 35862 3840 5396 86.9 90.3 80.1 all NN 4993 592 34375 4160 6241 89.2 84.6 77.5 TRIBE-MCL 4517 704 34563 2475 3042 93.3 91.9 77.9 gSPC 6 4612 710 34631 2472 2365 93.3 93.6 78.2 20 4988 115 37574 2836 2948 93.0 92.7 84.7 All NN 5043 18 37862 2968 3105 92.7 92.4 85.4 ::: ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Comparison of different clustering algorithms ::: Analyzed data set SPC MCL GSPC ------------------------------- --------------- ------ ------- ------ --------------- ------ SCOP domains 8666 (94%)^2^ 9.3 9208 12.2 9276 (101%) 7.7 SwissProt InterPro domains 96716 (99%) 15.6 97792 14.3 103729 (106%) 10.7 SwissProt keywords 98276 (99%) 21.8 99636 20.8 105339 (106%) 15.7 Bacterial genomes, FunCat 1.3 4652 (103%) 14.1 4517 14.7 5043 (112%) 14.8 1-The error is defined as error=100%\(FP+FN)/(TP+FN*). 2-In parentheses the percentage of clustered sequences relative to the corresponding numbers of the MCL algorithm (100%) are indicated. :::	PubMed Central	2024-06-05T03:55:55.820701	2005-4-1	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084344/", "journal": "BMC Bioinformatics. 2005 Apr 1; 6:82", "authors": [ { "first": "Igor V", "last": "Tetko" }, { "first": "Axel", "last": "Facius" }, { "first": "Andreas", "last": "Ruepp" }, { "first": "Hans-Werner", "last": "Mewes" } ] }
PMC1084345	Background ========== The growth and culturability of the actinobacteria is controlled by a family of secreted or membrane-associated proteins \[[@B1]\]. The Rpf protein of Micrococcus luteuswas the founder member of this family, which now comprises more than forty representatives \[[@B2]-[@B4]\]. Rpf is required for the resuscitation of dormant cells of M. luteusand for the growth of sparsely inoculated cultures of this organism in nutrient-poor media. M. luteusseems to contain only one rpfgene, whose product appears to be essential for bacterial growth \[[@B5]\]. In contrast, most organisms contain several rpf-like genes, whose products are functionally redundant \[[@B3],[@B6]-[@B8]\]. All the proteins so far tested show cross-species activity in bioassays using laboratory cultures of several different organisms, including M. luteus, Rhodococcus rhodochrous, Mycobacterium tuberculosis, Mycobacterium bovis(BCG) and Mycobacterium smegmatis\[[@B4],[@B7],[@B9],[@B10]\]. Since they are active at minute concentrations, it was suggested that they might be involved in inter-cellular signalling \[[@B1],[@B3],[@B4]\]. Rpf-like proteins are not found in firmicutes (low G+C Gram-positive bacteria), although some distantly related proteins are found in Staphylococcusand Oenococcus(see below). In this article we report the results of comparative genomic and domain analyses indicating that the firmicutes contain a cognate protein family related to the actinobacterial Rpf proteins by a process of \"non-orthologous domain displacement\". The available evidence strongly suggests that both the firmicute and actinobacterial proteins have a catalytic function, which may be responsible for their observed activity in improving the culturability of the organisms that produce them. Results ======= The Rpf domain -------------- Bacterial genome sequencing projects have uncovered many genes whose products share with M. luteusRpf a ca. 70-residue segment that we have called the Rpf domain. This segment of M. luteusRpf is both necessary and sufficient for biological activity, indicating that it corresponds to a functional protein domain \[[@B5]\]. The Rpf-like proteins appear to be restricted to several genera within the actinobacteria, including Corynebacterium, Micrococcus, Mycobacterium, Saccharopolysporaand Streptomyces, but they appear to be absent from some others, such as Bifidobacterium, Thermobifidaand Tropheryma(Table [1](#T1){ref-type="table"}). An alignment of 44 Rpf-like domains revealed that a central region of between 6 and 9 residues accounts for almost all of the observed variation in length of this domain (see [additional data file 1](#S1){ref-type="supplementary-material"}). SignalP \[[@B11]\] and TMHMM \[[@B12]\] predictions suggest that all of the Rpf-like gene products so far uncovered are either secreted, or membrane-associated, with the exception of one instance of an Rpf-like domain within a mycobacteriophage tape measure protein \[[@B13]\]. The Rpf domain also contains two highly conserved cysteine residues. Modelling has suggested that they lie in close proximity and may form a disulphide bridge (A. Murzin, personal communication) \[[@B14]\]. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Organisms containing rpf-like genes ::: Part A: genes encoding proteins containing a Rpf domain --------------------------------------------------------------------------------------------- ------------------ -------------- -------------------------- Organism Genome size (Mb) No. of genes Genome Accession Number Corynebacterium diphtheriae 2.5 3 NC\_002935 Corynebavterium glutamicum 3.3 2 NC\_003450 Corynebacterium efficiens 3.1 2 NC\_004369 Micrococcus luteus 2.3 1 Mukamolova et al, 1998 Mycobacterium avium 4.7 4 NC\_002944 Mycobacterium bovis 4.3 5 NC\_002945 Mycobacterium leprae 3.3 3 NC\_002677 Mycobacterium marinum 6.5 4 NC\_004506 (unfinished) Mycobacterium smegmatis 7.0 4 NC\_002974 (unfinished) Mycobacterium tuberculosisH37Rv 4.4 5 NC\_000962 Streptomyces coelicolor 8\. 7 5 NC\_003888 Streptomyces avermitilis 9.0 6 NC\_003155 Part B: genes encoding proteins containing a domain distantly related to the Rpf domain Bifidobacterium longumNCC2705 2.3 3 NC\_004307 Tropheryma whippleistrain Twist 0.9 2 NC\_004572 Streptomyces coelicolor 8\. 7 2 NC\_003888 Streptomyces avermitilis 9.0 3 NC\_003155 Staphylococcus carnosus \- 2 \- Staphylococcus aureusN315 2.8 1 NC\_002745 Staphylococcus epidermidis 2.6 1 NC\_004461 Oenococcus oeni 0.3 1 NZ\_AABJ02000001 M. luteusand S. carnosusgenomes are not yet sequenced M. luteusgenome size taken from Murayama et al. \[78\] ::: HMMs were used to create profiles of the Rpf domain alignment and these were employed to perform local and global searches of the SWISS-PROT and TrEMBL databases (downloaded from the European Bioinformatics Institute website \[[@B15]\]). In addition to the previously known Rpf domains in the various actinobacterial Rpf-like proteins, which were detected with highly significant E-values (5.7·10^-56^-- 4.8·10^-39^), these searches also identified two Staphylococcus carnosusprotein precursors, SceD and SceA (054493 and 054494), with much higher, but nevertheless statistically significant E-values (7.1·10^-4^and 3.9·10^-2^). These proteins contain a domain more distantly related to the Rpf domain. Additional hits above the level of statistical significance (E-values more than 0.1) included many c-type lysozyme precursors, which shared similarity with a 24-residue segment towards the C-terminus of the Rpf domain, as has been reported previously \[[@B2],[@B14],[@B16]\]. A PSI-BLAST search was also performed (Blosum62 matrix and a 0.005 E-value threshold) using the Rpf domain of M. luteusRpf for the first iteration <http://www.ncbi.nlm.nih.gov/BLAST/>. No new hits were found after 3 iterations. In addition to the known Rpf-like gene products and the more distantly related SceA & SceD proteins of S. carnosus, this search revealed SceD orthologues in two strains of Staphylococcus aureus(NP\_646837.1 & NP\_372619.1; E-values 2·10^-3^& 3·10^-3^) and Staphylococcus epidermidis(NP\_765249.1; E-value 9·10^-4^) in addition to a previously undetected gene product from Oenococcus oeni(ZP\_00069230.1; E-value 3·10^-13^). These proteins containing a domain distantly related to the Rpf domain are found in the firmicutes, whereas proteins containing the Rpf domain appear to be restricted to the actinobacteria. Rpf protein subfamilies ----------------------- Analysis of the various Rpf-like proteins for low complexity regions using SEG, which can separate discrete protein domains \[[@B17]\], and for common motifs using MEME, which can indicate orthologous domains \[[@B18],[@B19]\], indicated that they form ten discrete subfamilies, reflecting their multi-domain architecture. M. tuberculosiscontains representatives of five of these families, denoted RpfA-E in Fig. [1](#F1){ref-type="fig"}\[[@B7]\]. A sixth family, containing proteins with the peptidoglycan-binding motif, LysM \[[@B20]\], is restricted to the non-mycolate actinomycetes. A seventh family contains only corynebacterial proteins, while an eighth family contains two short proteins from Corynebacterium glutamicumand Streptomyces coelicolor, comprising only an Rpf domain. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Domain structure of the Rpf proteins grouped into their subfamilies. Proteins are from the genomes listed in Table 1. Proteins from organisms whose genome is not yet annotated (M. marinumand M. smegmatis) have been given the name of the subfamily to which they belong. ::: ![](1471-2164-6-39-1) ::: Proteins more distantly related to Rpf have been grouped together in two additional families. One of these includes the O. oeniprotein mentioned above; it has an inverse domain organisation compared with that of M. luteusRpf and Rpf-like proteins from Streptomyces. The other family of proteins distantly related to Rpf contains two proteins identified following a PSI\_BLAST search (3 iterations), using the large N-terminal region of M. tuberculosisRpfB (Rv1009) for the first iteration. This protein segment contains three repeats of PFAM-B DUF348 ([d]{.underline}omain of [u]{.underline}nknown [f]{.underline}unction) and a G5 domain (also of unknown function, which is found in various proteins involved in cell wall metabolism). The search detected all the previously known RpfB homologues, as well as the two additional gene products from Bifidobacterium longum(BL0658 and BL1227; E-values 2·10^-59^and 9·10^-32^). Several firmicute proteins were also detected (see below). The C-terminal region of the two previously undetected B. longumproteins was similar to part (the N-terminal portion) of the Rpf domain (Fig. [1](#F1){ref-type="fig"}). It was used to search the genpept database downloaded from the National Centre for Biotechnology Information website \[[@B21]\] and this revealed multiple hits in B. longum, Streptomyces avermitilis, S. coelicolorand Tropheryma whipplei. The search also detected the S. carnosusSceA protein, although this hit was not statistically significant. The actinobacterial gene products detected in these searches are grouped together as a subfamily of proteins distantly related to Rpf in Fig. [1](#F1){ref-type="fig"}. They were not detected in the original searches using HMMs of the profile of the Rpf domain alignment because similarity with the Rpf domain is restricted to its N-terminal portion (see [additional data file 1](#S1){ref-type="supplementary-material"}). Proteins similar to RpfB are found in firmicutes ------------------------------------------------ The link between actinobacterial RpfB and a family of firmicute proteins was noted several years ago, when FASTA was used to search the then available database with Rv1009 (M. tuberculosisRpfB) as a query sequence (R. McAdam, personal communication). This detected a B. subtilisprotein (YabE) of unknown function (23% identity and 38% similarity over 283 residues encompassing the DUF348 repeats and the G5 domain). A HMM model of this protein segment was used to search the TrEMBL and SWISS-PROT databases. In addition to the actinobacterial RpfB proteins, significant hits (E-value range 10^-5^-- 10^-28^) were found to a range of DUF348-containing proteins from various bacilli and clostridia (YabE-like proteins). In these firmicute proteins, the C-terminal Rpf domain is replaced by region of similar size (ca. 60 aa) but totally unrelated sequence. Significantly, rpfBand yabE(and the gene encoding the distantly related B. longumprotein) are found in a similar genomic context in the actinobacteria and the firmicutes (Fig. [2](#F2){ref-type="fig"}). ::: {#F2 .fig} Figure 2 ::: {.caption} ###### *Genomic context of some rpfBand yabEgenes. The sco3152, cgl0883, cpe2521and lin0223genes represented by an empty arrow are hypothetical proteins unrelated to each other. See the text for the designations of the remaining genes. ::: ![](1471-2164-6-39-2) ::: YabE is a member of an extended firmicute protein family -------------------------------------------------------- A tBLASTN search against the translated GenBank database using the C-terminal segment of YabE as query, revealed similar sequences in more than 40 proteins, suggesting that this is a distinct domain, which we have denoted Sps (Stationary phase survival -- see below). This region is also recognized as an uncharacterised conserved domain in the cluster of orthologous groups of proteins COG3584 and has recently been annotated in Pfam (see below). As for the Rpf domain, an HMM profile was created using the newly identified Sps domains and employed to search the TrEMBL database. In addition to the previously identified proteins in bacilli and clostridia, which were detected with highly significant E-values (4.4·10^-65^-- 4.3·10^-35^), these searches also identified some more distantly related proteins with higher, although still significant E-values (4.6·10^-7^-- 2.1·10^-2^). These hits include additional proteins OB0947 from Oceanobacillus ieheyensis, CAC2045 from Clostridium acetobutylicum, DR0488 from Deinococcus radioduransand TM0568 from Thermotoga maritima. The last two hits are the only examples of Sps-like proteins outside the firmicute phylum. Significantly (see below), the CAC2045 gene of C. acetobutylicumis annotated as an MltA (membrane-bound lytic transglycosylase A) homologue. Indeed, additional hits above the level of statistical significance in both standard similarity (BLAST) and HMM searches included several lytic transglycosylases from various proteobacteria (see below). Sps proteins are not found in organisms that contain Rpf proteins (Table [2](#T2){ref-type="table"}). ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Organisms containing sps-like genes ::: Part A: genes encoding proteins containing a Sps domain --------------------------------------------------------------------------------------------- ------------------ -------------- ------------------------- Organism Genome size (Mb) No. of genes Genome Accession Number Bacillus anthracisstrain A2012 5.1 5 NC\_003995 Bacillus anthracisstrain Ames 5.2 6 NC\_003997 Bacillus cereusATCC 10987 5.2 6 NC\_003939 Bacillus cereusATCC 14579 5.4 5 NC\_004722 Bacillus halodurans* 4.2 3 NC-002570 Bacillus subtilis 4.2 4 NC\_000964 Oceanobacillus iheyensis 3.6 4 NC\_004193 Listeria innocua 3.0 2 NC\_003212 Listeria monocytogenesEGD-e 2.9 2 NC\_003210 Enterococcus faecalisV583 3.2 1 NC\_004668 Lactococcus lactissubsp lactis 2.4 1 NC\_002662 Clostridium acetobutylicum 3.9 2 NC\_003030 Clostridium botulinumA 3.9 2 NC\_003223 (unfinished) Clostridium perfringensstr 13 3.0 3 NC\_003366 Clostridium tetaniE88 2.8 2 NC\_004557 Clostridium thermocellum 3.7 4 AABG03000000 Desulfitobacterium hafniense 4.9 1 AAAW00000000 Thermoanaerobacter tengcongensis 2.7 1 NC\_003869 Phage SPβc2 0.1 1 NC\_001884 Part B: genes encoding proteins containing a domain distantly related to the Sps domain Oceanobacillus iheyensis 3.6 1 NC\_004193 Deinococcus radiodurans 3.1 1 NC\_001263, NC\_001264 Thermotoga maritima 1.9 1 NC\_000853 ::: Sps protein subfamilies ----------------------- SignalP \[[@B11]\] and TMHMM \[[@B12]\] predictions suggest that all of the Sps-like gene products so far uncovered are likely to be either secreted, or membrane-associated, with the exception of Clostridium thermocellumCHTE712 (Fig. [3](#F3){ref-type="fig"}). The Sps proteins were also analysed using PFAM \[[@B22]\] and SMART \[[@B23],[@B24]\] for the presence of additional domains. Based on their domain architecture, and the chromosomal context of the encoding genes, they fell into eight subfamilies (Fig. [3](#F3){ref-type="fig"}). B. subtiliscontains four genes encoding representatives of four distinct subfamilies. The SpsB subfamily is characterised by the presence of two or three DUF348 domains and a G5 domain, both of which are common to the RpfB subfamily (cf. Fig. [1](#F1){ref-type="fig"}). The only exceptions are DESU7026 from Desulfitobacterium hafniense, which does not have DUF348 domains (but contains a G5 domain and shares the same genomic context as the other members of the SpsB subfamily), together with CPE1504 and CTC01185, from Clostridium perfringensand Clostridium tetani, respectively. These last two organisms appear to contain two yabE-like genes, one in the usual chromosomal context, and another elsewhere (in different positions in the two organisms). The SpsA subfamily is notable as a null mutant of its founder member from B. subtilisshows a substantial reduction in post-exponential phase survival (Ravagnani et al, ms. in preparation). These proteins are characterised by the presence of two copies of the peptidoglycan-binding motif LysM \[[@B20]\] (PG1 in the case of Bacillus haloduransBH3322), suggesting an association with the cell envelope. Members of the SpsA subfamily do not have a conserved chromosomal context. The other two subfamilies found in B. subtilisare the SpsC subfamily, whose members cluster on the basis of their sequence similarity outside the Sps domain and their identical genomic context, and YorM, which is located within the SPβ prophage and is therefore absent from strains that lack this genetic element. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Domain structure of the Sps proteins grouped into their subfamilies. Proteins are from the organisms listed in Table 2 with the exception of B. anthracisstrain A2012 and B. cereusATCC 14579, which contain the same proteins as B. anthracisstrain Ames and B. cereusATCC 10987 (apart from BA0685 and BCE3743, respectively). Proteins from C. botulinum, whose genome is not yet annotated, have been given the name of the subfamily to which they belong. YabE-like clostridial proteins that do not occupy the conserved chromosomal context represented in Figure 2 are indicated with an asterisk. ::: ![](1471-2164-6-39-3) ::: Two more subfamilies not represented in B. subtilisare of particular interest as they provided evidence for a link between the Sps proteins and muralytic enzymes. Bacillus anthracisand Bacillus cereusare the only organisms containing multiple spsgenes that do not contain members of the spsBsubfamily. Instead, they have gene products containing two copies of the SH3b domain (SpsE). In bacteria this domain is found in a number of muralytic enzymes, including endopeptidases and amidases. Several Sps proteins from a variety of firmicutes were clustered in another subfamily (SpsD) because they all contain a copy of the putative COG3883 domain. This uncharacterised conserved domain is also shared by a number of muralytic enzymes. O. ieheyensisOB0947, D. radioduransDR0488 and T. maritimaTM0568 are grouped together because they contain a domain that is only distantly related to the Sps domain (see above). DR0488 is the only known example of an Sps-like protein in an organism with high mole % GC DNA -- note however, that D. radioduransis not closely related to the Rpf-containing actinobacteria. The domain structure of TM0568, which has LysM and M23 peptidase domains, in addition to the Sps module, is reminiscent of the Rpf5 proteins from S. coelicolorand S. avermitilisthat contain LysM and M23 peptidase domains in addition to the Rpf module (Fig. [1](#F1){ref-type="fig"}), and provides another link between these proteins and cell-wall metabolism. The MltA-like proteins ---------------------- Three proteins from Clostridium thermocellumand one from Clostridium acetobutylicumrepresent the eighth subfamily of Sps proteins (Fig. [3](#F3){ref-type="fig"}). In these proteins, the Sps domain overlaps with a region of strong similarity to the Gram-negative membrane-bound lytic transglycosylase, MltA (Pfam E-value = 10^-6^-- 10^-7^). Indeed, Pfam predicted potential matches with MltA for all the Sps proteins, although with lower E-values (10^-2^-- 10^-3^). HMM profiles were built from the known lytic transglycosylases using the classification proposed by Blackburn and Clarke \[[@B25]\]. Local and global searches of the B. subtilisgenome using these profiles detected two known and six new putative lytic transglycosylases. Five of these (YjbJ, YomI, YqbO, YddH and YkdO) were similar to the family 1 of goose-type lysozymes. The remaining three, which are similar to the MltA-type family 2, are the Sps proteins, YocH, YuiC and YabE (E-values in local searches 4.1·10^-5^, 5.6·10^-6^& 2.3·10^-2^, respectively). The fourth B. subtilisSps protein, YorM, which lies within the SPβ prophage, was not detected. Blackburn and Clarke \[[@B25]\] distinguish six motifs within the MltA-type family 2 consensus sequence. The Sps domain encompasses motif 6 and part of motif 5. This region contains three conserved aspartate residues that may be involved in catalysis \[[@B25]\]. Significantly, these residues are absolutely conserved amongst all the 46 known Sps domains (Fig. [4A](#F4){ref-type="fig"}) as recently recognised in Pfam (3D domain). ::: {#F4 .fig} Figure 4 ::: {.caption} ###### T-Coffee alignment of MltA & Sps (A) and EmtA & Rpf (B) proteins. Residues shaded in black are present in 100% of the sequences, dark grey in 80% and light grey in 60%. Bars above the sequences indicate conserved motifs V (partial) & VI from MltA proteins (in Part A) and I, II and III from EmtA proteins (in Part B), as described by Blackburn and Clarke \[25\]. Putative catalytic residues are marked with an asterisk. Abbreviations are as follows: B. halodurans(BH), B. subtilis(Bsu), Brucella suis(BS), Candidatus Blochmannia floridanus(CBF), C. acetobutylicum(CA), C. perfringens(CP), C. tetani(CT), C. diphtheriae(CD), C. efficiens(CE), C. glutamicum(CG), D. hafniense(DH), E. faecalis(EF), E. coli(EC), E. coliO157:H7 (ECH7), E. coliO6 (ECO6), Haemophilus ducreyi(HD), L. innocua(LI), L. monocytogenes(LM), M. luteus(MLu), M. avium(MA), M. leprae(ML), M. tuberculosis(MT), O. iheyensis(OI), Photorhabdus luminescens(PL), Pseudomonas aeruginosa(PA), Rhizobium loti(RL), Salmonella typhi(Sti), Salmonella typhimurium(ST), Shigella flexneri(SF), S. avermitilis(SA), S. coelicolor(SC) and Yersinia pestis(YP). ::: ![](1471-2164-6-39-4) ::: These observations acquire even greater significance in the light of the weak similarity that has been noted between the Rpf domain and the goose-type lysozymes \[[@B2],[@B14],[@B16]\]. Blackburn and Clarke \[[@B25]\] identified four motifs in the consensus sequence of this type of lytic transglycosylase, and divided the family into five subclasses according to two more variable motifs 3 and 4. The C-terminus of the Rpf domain encompasses motifs 1 and 2 of the EmtA-type family 1e, which includes the absolutely conserved catalytic glutamyl residue (Fig. [4B](#F4){ref-type="fig"}). Discussion ========== We have presented evidence indicating that the firmicutes contain a family of proteins functionally equivalent to the actinobacterial Rpf family. The original link between the two protein families was provided by M. tuberculosisRpfB and B. subtilisYabE, which share a large N-terminal region containing DUF348 and G5 domains. In spite of this striking similarity, YabE lacks a C-terminal Rpf domain and contains instead a domain of similar size that we have called Sps (see above). Although the Rpf and Sps domains are totally unrelated in both sequence and secondary structure (see [additional data files 1](#S1){ref-type="supplementary-material"} and [2](#S2){ref-type="supplementary-material"}), we have presented evidence that they have a similar biological function. According to the definition proposed by Koonin et al. \[[@B26]\], an event of non-orthologous gene displacement can be suspected when the same function is fulfilled by unrelated or distantly related proteins. The RpfB and YabE proteins provide an example of a related phenomenon applicable to protein domains that we have called \"non-orthologous domain displacement\". Phylogenetic trees constructed using only the shared N-terminal region of RpfB-like and YabE-like (SpsB) proteins (Fig. [5](#F5){ref-type="fig"}) resemble trees generated with 16S rRNA, suggesting that these proteins have undergone vertical transmission from a common ancestor and that the Rpf domain displaced the Sps domain (or vice versa) sometime after the actinobacterial and firmicute lineages diverged. Most probably, this event has been followed by duplication and diversification within each lineage to create paralogues of the Rpf proteins in the actinobacteria and the Sps proteins in the firmicutes. Other instances of what could be referred to as non-orthologous domain displacement have been documented previously, e.g. aminoacyl tRNA synthetases. Bacterial and eukaryotic glutamyl-tRNA synthetases have generally similar domain architectures but they contain unrelated anticodon-binding domains \[[@B27],[@B28]\]. Similarly, eukaryotic tyrosyl tRNA synthetases contain two domains that are unrelated to those of their bacterial counterparts \[[@B28],[@B29]\]. The DnaG-like primases of bacteria and their phages differ from their archaeal orthologues in that the former contain a Zn-finger DNA-binding domain, whereas the latter contain a helicase-derived domain probably involved in the same function \[[@B30],[@B31]\]. Protein domains are considered as the basic units of folding, function and evolution \[[@B32]-[@B35]\] and we suspect that the phenomenon of non-orthologous domain displacement could be quite widespread. Moreover, it might have predictive value in cases where the function of only one of a pair of non-orthologous domains is already known. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Phylogenetic analysis of the RpfB and SpsB proteins. Phylogenetic trees based on the N-terminal moieties (DUF348 & G5 domains) from RpfB and SpsB proteins (right) and 16S rRNA sequences of organisms that contain RpfB and SpsB proteins (left). Trees were constructed by neighbour joining methods using MEGA v2.1 \[77\]. Bootstrap values are shown at the branch points. ::: ![](1471-2164-6-39-5) ::: Most rpfBand spsBgenes lie within a very similar genomic context flanked by tatDand ksgA(with rnmVinserted between spsBand ksgAin firmicutes). The only exceptions are the duplicate spsBgenes found in C. perfringensand C. tetani, one of which is located elsewhere in both organisms. Statistical analysis of the enormous amount of genome sequence information that has become available in recent years has shown that conservation of genome context may often be employed to infer functional relationships between neighbouring genes \[[@B36]\]. In our case, a functional association is indeed predicted by the SNAP algorithm (Similarity Neighbourhood APproach \[[@B37],[@B38]\]), though it is not obvious what the relationship might be. TatD is a Mg^2+^-dependent deoxyribonuclease of unknown function \[[@B39]\], RnmV is a ribonuclease M5/primase-related protein involved in maturation of the 5S rRNA \[[@B40],[@B41]\] and KsgA is a 16S rRNA methyltransferase that may play a role in translation initiation \[[@B42]\]. In B. subtilisthe tatD(yabD) gene does not appear to be expressed during either vegetative growth or sporulation, whereas the rnmV(yabF) and ksgAgenes appear to be co-transcribed during vegetative growth. They are highly expressed at the beginning of exponential phase and their expression declines sharply shortly afterwards, an almost identical pattern to that of yabE(data from the B. subtilisGenome Database \[[@B43]\]. These observations may reflect a connection between protein synthesis (RnmV, KsgA) and cell wall expansion (RpfB or SpsB -- see below) as would be required when a cell restarts growth after dormancy (in the case of Rpf) or prolonged stationary phase (in the case of Sps). The SNAP algorithm also predicts a functional association between RpfB/SpsB and the 4-diphosphocytidyl-2C-methyl-D-erythritol kinase. The gene encoding this protein (ispE) is located immediately downstream of ksgAin actinobacteria and two to four genes downstream of ksgAin Listeriaand Bacillusspp., respectively (however, it appears to have a scattered distribution in clostridia). The 4-diphosphocytidyl-2C-methyl-D-erythritol kinase participates in the non-mevalonate pathway for isoprenoid synthesis, which is involved in cell wall biosynthesis in E. coliand B. subtilis\[[@B44]\]. A functional relationship between neighbouring genes is normally inferred when they also show the same phylogenetic profile. This is not universally true in the present case, since some firmicutes, e.g. S. aureus, Streptococcus agalactiae, Streptococcus pyogenes, B. anthracisand B. cereus, contain neither rpfBnor spsBalthough the other genes normally associated with them, tatD, ksgAand rnmV(in firmicutes) are present in the same relative order. Presumably, rpfBor yabEhave been lost from these organisms (the alternative, necessitating several independent gene acquisition events, seems less likely). This is particularly evident in the mollicutes, where the occurrence of the genes in question is patchy. None of the strains sequenced contain rpfB/spsB(these organisms lack a cell wall), but some contain rnmV-ksgA(Mycoplasma capricolumand Mycoplasma mycoydes-- D14983 and NC\_005364, respectively), some contain tatD-ksgA(Mycoplasma pulmonis, NC\_002771) and some contain only ksgA(Mycoplasma genitalium, Mycoplasma gallisepticum, Mycoplasma penetransand Mycoplasma pneumoniae-- NC\_000908, NC\_004829, NC\_004432 and NC\_000912, respectively). As mollicutes are believed to derive from bacilli by reductive evolution \[[@B45]\], it seems that this group has lost rpfB/spsBand is in the process of loosing the remaining genes in the string. Note that rpfB, yabEand ksgAare non-essential genes \[[@B6],[@B8],[@B46]\] (Ravagnani et al., in preparation), as are tatDand rnmVin B. subtilis\[[@B41],[@B43]\]). Information from gene fusions may also be used to predict gene function. The \"Rosetta stone\" \[[@B47]\] and \"guilt by association\" \[[@B48]\] approaches propose that if a combination of domains A and B is detected in one protein and a combination of domains B and C in another, then it may be predicted that domains A, B and C are functionally related. The \"Rosetta stone\" hypothesis suggests that the function of one protein domain may be predicted on the basis of its fusion to another domain of known function. Since we do not know the function of the domains connecting RpfB and SpsB (DUF348 & G5), it might be more correct to invoke \"guilt by association\" in the present case. More recently, a new method based on consideration of genomic context has been employed to predict orthologous relationships between genes on the basis of anti-correlating occurrences of genes across species \[[@B49]\]. Given three genes A, B and C, if A is always present in a particular group of organisms in association with either B or C, but B and C are never found in the same organism, it can be predicted that B and C fulfil the same function. Extending this approach to protein domains, we may predict that the Rpf domain of RpfB and the Sps domain of SpsB have the same function, as they are both fused to the same DUF348- and G5-containing region, but never occur in the same organism (or, at least, in those so far sequenced). In bacteria, the DUF348 domain appears to be restricted to proteins containing either Rpf or Sps domains (but it is also found in the yeast Myb-like protein Snt1). B. anthracisand B. cereusare the only organisms containing multiple spsgenes that do not have an spsBgene, despite conservation of the genes with which it is normally associated (tatD, rnmVand ksgA). These bacteria have instead four and three copies, respectively, of spsEgenes encoding proteins containing two SH3b domains. SH3b is the equivalent of the eukaryotic SH3 (Src homology 3) domain, which is found in a variety of membrane-associated and cytoskeletal proteins and mediates protein-protein interactions by typically binding proline-rich polypeptides \[[@B50]\]. In bacteria, SH3b domains are found in various cell wall amidases and peptidases. Although their function is unknown, the SH3b-containing region of Staphylococcus simulanslysostaphin, which cleaves peptidoglycan, mediates binding to the S. aureuscell wall \[[@B51]\]. Such a function would be consistent with the occurrence of this domain in muralytic enzymes. It is tempting to suggest that the DUF348 domain has a role similar to that of the SH3b domain. Whatever their functions might be, invoking again the principle of \"guilt by association\" \[[@B48]\], the association of the Sps domains with other domains present in muralytic enzymes (SH3b, COG3883, LysM) points very strongly to a role for the Sps proteins in cell wall metabolism. This hypothesis is also supported by the occurrence of an M23 peptidase domain in S. coelicolorand S. avermitilisRpf5, Thermotoga maritimaTM0568 and some lytic transglycosylases, such as B. subtilisYomI. The sequence similarity between the C-terminal region of the Sps domain and that of the Gram-negative membrane-bound lytic transglycosylase, MltA, serves to reinforce this connection. Figure [4A](#F4){ref-type="fig"} shows that the similarity between Sps and MltA encompasses all three aspartate residues that have been highlighted as potential catalytic residues for the lytic transglycosylase family 2 -- classification according to Blackburn and Clarke \[[@B25]\]. In parallel with this, there is also sequence similarity between the Rpf domain and the N-terminal region of the Gram-negative endo membrane-bound lytic transglycosylase, EmtA \[[@B2],[@B14]\]. Although quite limited, the similarity in this case encompasses the absolutely conserved catalytic glutamate residue of the lytic transglycosylase family 1 (Fig. [4B](#F4){ref-type="fig"}). Lytic transglycosylases are enzymes that catalyse cleavage of the β-1,4-glycosidic bond between N-acetylmuramic acid and N-acetylglucosamine in the peptidoglycan backbone. Unlike lysozyme, they also catalyse an intramolecular glycosyltransferase reaction to form terminal 1,6-anhydromuramic acid-containing products. The exact function of these enzymes is unknown, but they are thought to be involved in cleavage of the peptidoglycan to permit the insertion of newly synthesised material during cell elongation and division. Remodelling of the cell envelope requires the concerted action of both hydrolases and synthetases, which may form large multienzyme complexes \[[@B52],[@B53]\]. Consistent with this, physical interactions between some E. colilytic transglycosylases and penicillin-binding proteins (enzymes involved in the synthesis of peptidoglycan) have been demonstrated experimentally \[[@B54],[@B55]\]. In E. colithere are at least six lytic transglycosylases, one soluble and five membrane-bound \[[@B56]-[@B60]\], with different substrate specificities. Due to the high degree of redundancy, no obvious effect on growth is observed after deletion of their genes \[[@B60]\]. This is in agreement with the results obtained after disruption of three of the five rpf-like genes in S. coelicolor\[[@B2]\] and the five rpf-like genes of M. tuberculosis\[[@B6],[@B8]\]. In contrast, there is evidence for essentiality of the apparently unique rpfgene of M. luteus, whose chromosomal copy could be disrupted only in the presence of an extra plasmid-encoded copy of the gene \[[@B5]\]. However, definitive proof of essentiality would require the construction of a conditional mutant and this technology is not currently available for M. luteus. In B. subtilisthe spsgenes are not essential, but a clear phenotype is associated with disruption of yocHand this is much accentuated by the disruption of all four spsgenes: these mutants show reduced survival after prolonged stationary phase (Ravagnani et al., ms. in preparation). This phenotype has been observed previously, associated with disruption of genes involved in cell wall metabolism, such as the E. coli nlpD, encoding an M23 endopeptidase \[[@B61]\], and surA, encoding a peptidyl-prolyl isomerase \[[@B62]\]. The latter is required for the correct folding of extracytoplasmic proteins and it has been proposed to be necessary for the assembly of the murein-synthesizing complex, of which lytic transglycosylases are a component \[[@B62]\]. In the Gram-positive bacteria, rpfBor spsBoccupy a highly conserved genomic context, within a group of genes including ksgA(see above). Interestingly, in E. coliand related enteric bacteria, ksgAlies within the same transcription unit as surA(surA-pdxA-ksgA-apaG-apaH), suggesting again a possible association between protein synthesis and cell wall expansion. The assignment of a muralytic function to the Sps and Rpf domains is entirely consistent with the presence of an Sps protein, YorM, in the B. subtilisprophage SPβ, and the recent discovery of the Rpf domain in a large mycobacteriophage \"tape measure protein\" \[[@B13]\]. Muralytic transglycosylase activity is often associated with bacteriophage virions and confers upon them the highly localised muralytic activity that is required for the process of phage infection, without provoking premature lysis of the host \[[@B63]\]. The bioinformatic evidence in favour a role for the Rpf and Sps proteins in peptidoglycan metabolism is now compelling. This prediction has recently been confirmed; both M. luteusRpf and B. subtilisYocH have murein hydrolase activity in zymograms (Mukamolova et al., ms. in preparation; Ravagnani et al., ms. in preparation). Conclusions =========== As a result of the observed catalytic activity of the Sps and Rpf proteins, our views on the nature of bacterial non-culturability are changing. The various models of non-culturability we have developed over the years \[[@B1],[@B64],[@B65]\] might be explained by the disappearance of nascent peptidoglycan and its gradual replacement by inert peptidoglycan in the bacterial cell wall. This has recently been proposed as a key feature of the mechanism that determines the position of growth zones in the bacterial cell wall \[[@B66]-[@B68]\]. We suggest that the walls of non-culturable organisms may contain such a preponderance of inert peptidoglycan that their envelope has effectively become a \"cocoon\", requiring the action of specialised muralytic enzymes to make a restricted number of scissions, before growth and wall expansion can resume. The Sps and Rpf proteins may have been recruited to serve this function. Resumption of cell wall synthesis might therefore be regarded as one of the \"core processes\" (see above), along with re-initiation of protein synthesis, that would need to be activated by cells emerging from dormancy (in the case of Rpf) or prolonged stationary phase (in the case of Sps). Signalling could be part of such a resuscitation mechanism, mediated perhaps by a small molecule released from murein as a result of the action of Rpf / Sps proteins. This hypothesis is currently being tested. Methods ======= Database searching was carried out using either the position-specific iterative BLAST (PSI-BLAST) method \[[@B69]\] or the Hidden Markov model (HMM) database searching algorithm of HMMER 2.2 g <http://hmmer.wustl.edu/>. Both local and global profiles of aligned sequences were generated, and searches were carried out using the default parameters. For one application, FASTA \[[@B70]\] was employed. Domain analysis was undertaken using COG \[[@B71]-[@B73]\], MEME \[[@B18],[@B19]\], PFAM \[[@B22]\], SEG \[[@B17]\], SignalP \[[@B11]\], TMHMM \[[@B12]\] and SMART \[[@B23],[@B24]\]. Sequence alignments were generated using ClustalX version 1.81 \[[@B74]\] and T-coffee \[[@B75],[@B76]\]. Phylogenetic trees were generated using MEGA v2.1 \[[@B77]\]. T-coffee-aligned sequences were analysed using the neighbour-joining method (options: p-distance model, compete removal of gaps, 10,000 bootstrap replications). Authors\' contributions ======================= AR carried out the bioinformatic analysis of the Sps proteins and drafted the manuscript. CLF carried out the bioinformatic analysis of the Rpf proteins. MY supervised the project and contributed to drafting of the manuscript. All authors read and approved the final manuscript. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Sequence alignment of the Rpf domainsClustal X alignment of Rpf domains (A) and of domains distantly related to the Rpf domain (B). ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 2 Sequence alignment of the Sps domainsClustal X alignment of Sps domains (A) and of domains distantly related to the Sps domain (B). ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ This work was funded by the UK BBSRC. C.L.F. was the grateful recipient of a BBSRC studentship. We are grateful to Tim Langdon and Joe Ironside for many helpful discussions and to Eugene Koonin for drawing other examples of non-orthologous domain displacement to our attention.	PubMed Central	2024-06-05T03:55:55.826855	2005-3-17	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084345/", "journal": "BMC Genomics. 2005 Mar 17; 6:39", "authors": [ { "first": "Adriana", "last": "Ravagnani" }, { "first": "Christopher L", "last": "Finan" }, { "first": "Michael", "last": "Young" } ] }
PMC1084346	Background ========== Insulin resistance is associated with obesity especially when centrally distributed \[[@B1],[@B2]\]. The connection between increased adiposity and insulin resistance is still poorly understood, although recent evidence has suggested that adipose tissue-released cytokines like adiponectin, resistin, leptin and Tumour Necrosis Factor-Alpha (TNF alpha) may be contributory factors \[[@B3]-[@B6]\]. In this context, the activation of the adipocyte expressed nuclear receptor PPAR-γ2, by the new insulin-sensitising drugs thiazolidinediones \[[@B7],[@B8]\], was suggested to increase plasma adiponectin levels, which contributed to the promotion of fatty acid oxidation and insulin sensitivity in muscle and liver \[[@B9],[@B10]\] and also to the inhibition of the expression of TNF-α and resistin \[[@B4],[@B11]-[@B13]\]. There is a frequent non synonymous (cytosine to guanine) single nucleotide polymorphism (SNP) in PPAR-γ2exon 2 \[[@B14]\]. This variation results in a Proline to Alanine substitution at the codon 12, which has been found to modulate the transcriptional activity of the gene \[[@B15],[@B16]\]. Several studies have reported an association between Ala12, increased insulin sensitivity and a reduced risk of type 2 diabetes \[[@B17]-[@B20]\] which has been supported by meta-analysis \[[@B17]\]. However, results assessing the risk of the Pro12Ala in obesity were controversial \[[@B21],[@B22]\]. We previously studied the effect of this variant in a rather limited sample of French Caucasians \[[@B23]\] and found no association between the Pro12Ala and T2D or obesity in this population. Given the modest impact of the Pro12Ala in the T2D risk \[[@B17]\], the previous study may have probably lacked the statistical power for replication. In order to achieve a definite conclusion about this gene variant and \"diabesity\" in French Caucasians, this present study has analysed large sample sets representative of the condition and as well as controls (altogether 2126 cases and 1124 controls). We designed three independent case/control studies to assess the presence of any association between the Pro12Ala SNP and T2D, childhood obesity or adulthood obesity. In addition the putative effect of this SNP on insulin sensitivity and insulin secretion indexes in normal glucose tolerant (NGT) lean subjects and in NGT obese subjects was evaluated, and as well any possible interaction between the SNP and adiposity on insulin sensitivity, as recently reported in the US population, was also assessed \[[@B24]\]. Methods ======= Subjects -------- The DNA samples were extracted from EDTA whole-blood samples using Puregene Kit (Gentra, Minneapolis, M N). Six samples sets were used for association studies and for analysis of variance of phenotypic traits: Association studies with childhood obesity were performed using a set of 195 unrelated lean children from the \"Fleurbaix-Laventie Ville Santé\" and a set of 396 unrelated obese children chosen from the cohort of 554 obese children available. The pool of obese children used in association studies was constituted by a first set of 278 unrelated obese children collected from 278 pedigrees with at least one obese child at the CNRS-Institut Pasteur Unit and at the Jeanne de Flandres Hospital in Lille, a second set of 90 unrelated obese children recruited at the Children\'s Hospital, Toulouse \[[@B25]\], and a third set of 28 unrelated obese children recruited through the \"Fleurbaix-Laventie Ville Santé\" study. Children with a BMI greater than the 97^th^percentile of BMI for age and sex reported on the tables of Rolland-Cachera et al. \[[@B26]\] (French general population) were defined as obese as recommended by the European Childhood Obesity Group (ECOG) \[[@B27]\]. From the 376 lean children available in the Fleurbaix-Laventie study, only 195 unrelated children were included as control, since all children with at least one obese sibling were excluded. Adult obesity association studies were performed using two sets of adult obese patients: the first constituted by 450 moderately obese adults and the second constituted by 652 morbidly adult obese patients. These individuals were collected at the Department of Nutrition of the Hôtel Dieu Hospital in Paris or at the CNRS-Institut Pasteur Unit in Lille. Obesity status was defined as BMI ≥ 30 in adults. We used as control group a set of 611 unrelated non obese and non diabetic subjects recruited at the CNRS-Institut Pasteur Unit in Lille (N = 345) and through the \"Fleurbaix-Laventie Ville Santé\" study (N = 266) \[[@B28]\]. Association studies with T2D were performed using a set of 628 type 2 diabetic subjects. Diabetic state was informed by a fasting and/or glucose-tolerance test according to the WHO 1999 criteria. Patients are part of a publicity advertised campaign for \"200 families to overcome diabetes\", of CNRS-Institut Pasteur de Lille recruitment (N = 365) or recruited at the Sud Francilien Hospital in Corbeil-Essonnes (N = 263). We used a group of 318 non diabetic unrelated spouses from T2D and obesity families, aged more than 50 years old with a fasting glycaemia less than 5.6 mmol/l and recruited by a multimedia campaign at the Institut Pasteur of Lille. Quantitative traits were calculated using normal glucose tolerant subjects selected from each of the initial sample groups previously described (Table [1](#T1){ref-type="table"}). Normal Glucose Tolerance was defined by fasting glycaemia lower than 6.1 mmol/l and by a 2-hour post OGTT glycaemia lower than 7.8 mmol/l, according to the WHO 1999 criteria when available. For quantitative trait analysis, a set of 362 NGT lean children, 507 NGT obese adults and 525 NGT obese children were analysed. The 865 non obese NGT adults cohort was constituted by pooling 547 non obese NGT subjects and 318 non diabetic subjects. Genotype-BMI interaction effect on insulin resistance was calculated separately in NGT children and adults. In each interaction test, obese and lean subjects were pooled together. The phenotypic characteristics of all the studied populations are displayed in Table [1](#T1){ref-type="table"}. Clinical parameters ------------------- The Body Mass Index (BMI) was calculated as weight (Kg) divided by height (m) squared. The Z score of BMI was calculated using Cole\'s last mean square method \[[@B29]\]. Quantitative measurements of plasma insulin were carried out using double-antibody radio immunoassays. Serum glucose concentrations were measured using a glucose oxidase procedure. HOMA-IR and HOMA-B were calculated according to Matthews et al \[[@B30]\]. Genotyping ---------- Genotyping of the Pro12Ala of the PPAR gamma 2 gene was performed using the Taqman Allelic discrimination (AD) Assay (Applied Biosystem). The Taqman genotyping reaction was amplified on a GeneAmp PCR system 9600 (95°C for 10 minutes, followed by 40 cycles of 92°C for 15 seconds, and 60°C for 1 minute), and fluorescence was detected on an ABI Prism 7900 sequence detector (Applied Biosystem). The mix used in the Taqman experiment contained 2.5 μl of the master mix, 0.25 μl of Primers, 0.25 μl of water and 2 μl of DNA with a concentration of 10 ng/μl. Statistical analysis -------------------- Statistical analysis were carried out using SPSS 10.0 program (SPSS, Chicago, IU, USA). The Pro12Ala variant was complied with Hardy-Weinberg proportions. Fisher\'s exact test was applied to compare allelic frequencies between case and controls \[[@B31]\]. Quantitative traits were studied using the general linear model (GLM), taking into account gender, age, BMI and genotype effect (Pro12Pro versusPro12Ala/Ala12Ala-dominant model). The BMI quantitative trait was adjusted only by gender and age. The genotype-BMI and genotype-obesity status interaction effects on insulin resistance were also tested using a GLM procedure. HOMA-IR was the dependant variable. Age, gender, BMI (or obesity status), and Pro12Ala polymorphism were the explicative factors. A term of interaction BMI\Pro12Ala, or obesity status\Pro12Ala was introduced. The Z score of BMI was used in analyses including obese children. Results ======= Phenotypic characteristics of the studied populations ----------------------------------------------------- Phenotypic characteristics of the six populations are shown in Table [1](#T1){ref-type="table"}. Compared to lean children, obese children have higher fasting glycaemia and fasting insulinemia (11.88 UI/l versus5.77 UI/l, p \< 0.001). Obese adult subjects have higher glycaemia and fasting insulinemia (16.11 UI/l versus5.77 UI/l, p \< 0.001) than non obese subjects. T2D subjects have higher BMI and fasting insulinemia than non diabetic subjects (11.18 UI/l versus7.40 UI/l, p \< 0.001). Effect of the Pro12Ala SNP on obesity ------------------------------------- The genotypic distributions of the Pro12Ala did not significantly deviate from the Hardy Weinberg equilibrium in any of the six groups of subjects. There was no significant difference in the allelic frequencies between the four analysed sample sets as shown in Table [2](#T2){ref-type="table"}, ruling out any association between this SNP and severe childhood or adulthood obesity in the French Caucasian population. Effect of the Pro12Ala SNP on T2D --------------------------------- As previously shown in other ethnic groups, the \"at risk\" Pro allele was modestly but significantly more frequent in T2D subjects than in non diabetic controls (0.87 in control versus0.90 in T2D subjects, respectively, p = 0.039, OR = 1.37, see Table [3](#T3){ref-type="table"} and Figure [1](#F1){ref-type="fig"}). When T2D subjects were stratified according to their obesity status (defined by a BMI = 30), the association with T2D was only significant in the obese diabetic subgroup (p = 0.030) where it was found to display a rather strong genetic risk (OR = 1.81) than the non obese diabetic subgroup (OR = 1.28, p = 0.12, Table [3](#T3){ref-type="table"}, Figure [1](#F1){ref-type="fig"}). Effect of the Pro12Ala SNP on insulin sensitivity and secretion indexes in lean normal glucose tolerant (NGT) French Caucasians ------------------------------------------------------------------------------------------------------------------------------- The effect of the Pro12Ala variant on diabesity associated quantitative phenotypes was then assessed under a dominant model in the lean NGT controls (865 adults, and 362 children). The Ala allele was not significantly associated with fasting insulinemia, glycaemia, BMI or Z score of BMI, HOMA-IR (insulin resistance index) or with HOMA-B (insulin secretion index) in any of these two sample sets (Table [4](#T4){ref-type="table"}). Effect of the Pro12Ala SNP on insulin sensitivity and secretion in obese NGT French Caucasians ---------------------------------------------------------------------------------------------- The impact of the Pro12Ala SNP in the cohort of 1032 normal glucose tolerant obese subjects was then evaluated. In the adult subgroup (n = 507), subjects carrying the \"protective\" 12Ala allele showed a significant increase in their insulin sensitivity (p = 0.003), and consistently, a decrease in fasting insulinemia (p = 0.01) and of glycaemia (p = 0.04, see Table [4](#T4){ref-type="table"}). In obese children (n = 525), the same trend was observed although it did not reach statistical significance. No effect on insulin secretion index (HOMA-B), Z score of BMI or BMI was found (Table [4](#T4){ref-type="table"}). Interaction between the Pro12Ala SNP and adiposity on insulin resistance ------------------------------------------------------------------------ The relationship between the Pro12Ala SNP, HOMA-IR and the Z score of BMI or BMI in NGT children and adults respectively was evaluated. Thus, we analysed separately 887 lean or obese NGT children and 1372 NGT adults (obese and lean together). The general linear model (GLM) analysis taking into account the continuous trait Z score of BMI or BMI supported the null hypothesis of no interaction between the Pro12Ala and the corpulence on HOMA-IR values in both ages (p = 0.59 and p = 0.32, see figures [2A](#F2){ref-type="fig"} and [2B](#F2){ref-type="fig"}). A similar GLM analysis using the obesity status binary trait (BMI ≥ 97^th^percentile in children and BMI ≥ 30 kg/m^2^in adults) revealed no evidence for interaction (p = 0.21) in NGT children, but a borderline interaction effect (p = 0.06) between the Pro12Ala SNP and the obesity status with respect to the HOMA-IR index in normal glucose tolerant adults. Discussion ========== In the present study, and in contrast to what we have previously reported \[[@B23]\], T2D subjects were found to possess a significantly higher frequency of the Pro12 allele risk than non diabetic controls, thus supporting a role for PPAR-γ2in the genetic risk for type 2 diabetes in French Caucasians. In contrast, no association with severe forms of obesity was reported in both adults and children confirming our previous finding. Interestingly, in the obese subgroup of diabetics, the Pro12 allele was found to nearly double the risk for T2D, although it has a weaker effect in lean individuals. This discrepancy might be due to an obesity-dependent effect of the \"at risk\" allele on insulin sensitivity. Indeed, obese adults carrying the Pro12Pro genotype were more insulin resistant than subjects carrying at least one Ala12 allele, with a similar trend seen in obese children, although no such effect was found in lean controls. Li et al \[[@B24]\] reported that the detrimental effect of the Pro12 allele on insulin sensitivity was stronger in adult subjects with the highest BMI in the white Caucasian general population, with similar trends observed in children. However, in contrast with these data, we did not find in the entire cohorts of French adults or children formal evidence for an interaction between adiposity (evaluated by the BMI or the Z score of BMI) and the PPAR-γ2Pro12Ala SNP on the variation of the insulin sensitivity index. Consistent with our results, Buzzetti et al. also failed to confirm this interaction in an Italian general population \[[@B32]\]. We then used the obesity status rather than the continuous trait BMI or the Z score of BMI in the analysis of interaction effect. Indeed, the BMI range observed in our samples was bimodal (due to the pooling effect of non obese and obese individuals) and highly skewed regarding a population-based cohort, suggesting that the study of a binary trait could be more relevant in our experimental design. We didn\'t detect any interaction in NGT children group (p = 0.21), but a borderline interaction effect (p = 0.06) between the Pro12Ala SNP and the obese status with respect to the HOMA-IR index in normal glucose tolerant adults. Our results, according to the hypothesis advanced by Li et al \[[@B24]\], confirmed that an obesity background could worsen the detrimental effect of the PPAR-γ2Pro12 allele on insulin sensitivity in adult subjects, but not in children. Clément\'s negative data with T2D in the same ethnic group \[[@B23]\] may highlight the fact that studies with modest sample size can fail to detect true associations. Risch et al. \[[@B33]\] and Lohmueller et al\[[@B34]\] both state that inadequate power in replication studies may contribute to the large number of non replications, hence underlining the importance of working with large enough sized populations in order to show a modest variant\'s effect in a multifactorial disease context. Indeed, meta-analysis are in line with our present data \[[@B17]\]. In our T2D population, the frequency of the protective Ala12 variant was under-represented compared with adult population controls (0.10 versus0.13, respectively) but is very similar with lean children\'s cohort (0.10 versus0.10, respectively). Thus, in French children, the prevalence of the \"at risk\" allele is surprisingly higher than in middle-aged non diabetic individuals. However, our results confirm those recently obtained by Doney et al in a Scottish cohort \[[@B35]\]. Indeed, they suggested that the protective effect of Ala12 against T2D and other metabolic diseases which increase the risk for premature coronary heart disease, could result in the selective recruitment of elderly Ala carriers in control populations. The Ala12 allele resulted in a reduction in the transcriptional activity of PPAR-γ2 \[[@B15]\] and was associated with an increase of insulin sensitivity. Interestingly, data from murine models support these findings; Moderate reduction of PPAR-γ activity in heterozygous PPARγ-deficient mice prevented adipocyte hypertrophy, increased fatty acid combustion and decreased lipogenesis, which resulted in increased insulin sensitivity with regards to both glucose disposal and suppression of glucose production \[[@B36]-[@B38]\]. Moreover, these mice showed lower fasting plasma insulin, higher leptin and adiponectin serum levels \[[@B39]\]. Stumvoll et al. showed that alterations in the transcriptional activity of the Ala variant in adipocytes could primarily enhance insulin action on suppression of lipolysis, resulting in a decreased release of free fatty acids (FFAs) \[[@B40]\]. Boden et al. demonstrated that reduced availability of FFAs would permit muscle to utilize more glucose and the liver to suppress glucose production more efficiently upon insulin stimulation \[[@B41]\]. Taken together, these two results can explain the improvement of the insulin sensitivity as a consequence of the Ala allele. Conclusion ========== In summary, we confirm that, in the French Caucasian population, the PPAR-γ212Ala SNP allele confers a reduced risk for T2D, and decreased obesity-associated insulin resistance although it was not associated with childhood or adulthood obesity. In this regard PPAR-γ2can be considered as a \"diabesity\" gene. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= Maya Ghoussaini and Stéphane Lobbens genotyped the Pro12Ala polymorphism of the PPAR-gamma2 gene in the studied populations. David Meyre and Maya Ghoussaini performed the statistical analyses to evaluate the Pro12Ala effect. Philippe Froguel and David Meyre have directed the study and the redaction of the article that was written by Maya Ghoussaini. DNA was provided by Guillaume Charpentier, Karine Clément, Marie-Aline Charles, Maïté Tauber and Jacques Weill. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2350/6/11/prepub> Acknowledgements ================ We are indebted to all families who participated to this study. We also thank \"le Conseil National de la Recherche Scientifique Libanais (CNRS-L)\", \"200 Familles pour Vaincre le Diabète et l\'Obésite \" and \"Association Française des Diabétiques\" for their financial support and Christopher G. Bell for the helpful comments and the corrections in the manuscript. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Allelic frequency of the Pro12Ala polymorphism in non diabetic and type 2 diabetic subjects (T2D) stratified according to obesity status. ::: ![](1471-2350-6-11-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Interaction between the Pro12Ala SNP and adiposity on HOMA-IR. The interaction was tested using a GLM procedure. HOMA-IR was the dependant variable. Age, gender, BMI or obesity status, and Pro12Ala polymorphism were the explicative factors. A term of interaction BMI\Pro12Ala or obesity status\Pro12Ala was introduced. The Z score of BMI was used rather than BMI in analyses including obese children. (A) No evidence for interaction was found neither for Z score of BMI\Pro12Ala (p = 0.59) nor obesity status\Pro12Ala (p = 0.21) in 887 normal glucose tolerant lean and obese children (p = 0.58). (B) No evidence for interaction was found for BMI\Pro12Ala (p = 0.32) in 1372 normal glucose tolerant obese and lean adults subjects, but a borderline significant interaction was found for obesity status\Pro12Ala (p = 0.06). ::: ![](1471-2350-6-11-2) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Phenotypic characteristics of the studied populations. ::: --------------------------------------------------------------------------------------------------------------------------------------------- Phenotypic traits Obese Children Lean Children Obese Adults Non Obese Adults Type II Diabetics Non Diabetic Adults ---------------------------- ---------------- ----------------- ---------------- ------------------ ------------------- --------------------- Sex (M/F) 264/290\ 198/178\ 357/745\ 234/377\ 353/276\ 123/195\ N = 554 N = 376 N = 1102 N = 611 N = 628 N = 318 Age (years) 10.44 ± 0.16\ 13.61 ± 0.13\ 47.15 ± 0.41\ 51.04 ± 0.52\ 59.35 ± 0.45\ 62.03 ± 0.66\ N = 554 N = 376 N = 1102 N = 611 N = 625 N = 318 BMI (kg/m^2^) 28.16 ± 0.27\ 18.14 ± 0.13\ 42.32 ± 0.265\ 22.95 ± 0.09\ 26.84 ± 0.14\ 25.69 ± 0.30\ N = 554 N = 365 N = 1102 N = 604 N = 616 N = 318 Z score of BMI 4.10 ± 0.06\ \- 0.24 ± 0.05\ \- \- \- \- N = 554 N = 365 Fasting glycaemia (mmol/l) 4.93 ± 0.02\ 4.71 ± 0.02\ 6.73 ± 0.08\ 5.03 ± 0.24\ 9.45 ± 0.13\ 4.95 ± 0.02\ N = 546 N = 364 N = 1056 N = 571 N = 611 N = 318 Fasting insulinemia (UI/l) 11.88 ± 0.48\ 5.77 ± 0.21\ 16.11 ± 0.40\ S.77±0.22\ 11.18 ± 0.57\ 7.40 ± 0.36\ N = 491 N = 362 N = 1003 N = 570 N = 304 N = 318 --------------------------------------------------------------------------------------------------------------------------------------------- Values are expressed as mean ± mean standard error. ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Comparison of Genotypic and Allelic Distribution of Pro12Ala polymorphism in obese versusnon obese subjects. Case/Control studies show that the Pro12Ala polymorphism is not associated with childhood or adult obesity in the French Caucasian population (p = 0.19, p = 0.92 respectively). ::: Populations PPAR genotypes Allelic frequencies P-value of allelic frequencies ---------------------------- ---------------- --------------------- -------------------------------- ------ ------ ---------------------------------------- Pro/Pro Pro/Ala Ala/Ala Pro Ala Obese children (n = 396) 304 (76.8%) 84 (21.2%) 8 (2.0%) 0.87 0.13 P = 0.19 OR = 0.77 CI = \[0.52--1.14\] Lean children (n = 195) 156 (77.9%) 39 (21.0%) 0 (1.1%) 0.90 0.10 Obese adults (n = 1102) 857 (77.9%) 231 (21.0%) 12 (1.1%) 0.88 0.12 P = 0.92 OR = 1.01 CI = \[0.81--1.26\] Non obese adults (n = 611) 478 (78.2%) 123 (20.2%) 10 (1.6%) 0.88 0.12 OR: Odd Ratio and CI: Confidence Interval. ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Comparison of Genotypic and Allelic Distribution of Pro12Ala polymorphism in diabetic versusnon diabetic subjects. Case/Control studies show the association of the Pro12Ala polymorphism with T2D in the French Caucasian population (p = 0.04). The stratification of diabetic subjects on the obesity status shows that the association previously found is rather due to obese T2D subgroup (p = 0.03) than non obese T2D subgroup (p = 0.12). ::: Populations PPAR genotypes Allelic frequencies P-value of allelic frequencies ---------------------------------- ---------------- --------------------- -------------------------------- ------ ------ -------------------------------------------- Pro/Pro Pro/Ala Ala/Ala Pro Ala T2D patients (n = 628) 511 (81.4%) 113 (18.0%) 4 (0.6%) 0.90 0.10 P = 0.04OR = 1.37 C.I = \[1.02--1.85\] T2D obese patients (n = 114) 98 (86.0%) 15 (13.1%) 1 (0.9%) 0.93 0.07 P = 0.03OR = 1.81 C.I = \[1.05--3.13\] T2D non obese patients (n = 503) 403 (80.1%) 97 (19.3%) 3 (0.6%) 0.90 0.10 P = 0.12 OR = 1.28 C.I = \[0.94--1.74\] Non diabetic subjects (n = 318) 246 (77.4%) 63 (19.8%) 9 (2.8%) 0.87 0.13 OR: Odd Ratio and CI: Confidence Interval. ::: ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Biochemical parameters of subjects according to Pro12Ala genotypes. ::: Populations Genotypes Fasting insulin (UI/l) Fasting glucose (mmol/l) Body Mass Index (Kg/m^2^) Z score of BMI HOMA-IR HOMA-B ---------------------------------- -------------- ------------------------ -------------------------- --------------------------- ---------------- -------------- ---------------- NGT non obese adults (n = 865) P/P = 673 6.28 ± 0.24 4.92 ± 0.02 23.61 ± 0.14 \- 1.39 ± 0.05 92.99 ± 3.44 P/A+AA = 192 6.23 ± 0.31 4.94 ± 0.03 24.02 ± 0.26 \- 1.38 ± 0.07 95.11 ± 7.19 P-value 0.92 0.65 0.16 \- 0.90 0.78 NGT non obese children (n = 362) P/P = 291 5.39 ± 0.17 4.70 ± 0.21 18.12 ± 0.15 \- 0.24 ± 0.06 1.13 ± 0.04 95.84 ± 3.47 P/A+AA = 71 6.21 ± 0.42 4.71 ± 0.04 18.19 ± 0.28 \- 0.27 ± 0.12 1.315 ± 0.09 111.58 ± 9.62 P-value 0.06 0.67 0.84 0.84 0.06 0.62 NGT obese adults (n = 507) P/P = 397 14.35 ± 0.60 5.30 ± 0.02 40.82 ± 0.42 \- 3.41 ± 0.15 171.07 ± 7.60 P/A+AA = 110 11.93 ± 0.72 5.19 ± 0.05 40.08 ± 0.70 \- 2.75 ± 0.16 165.89 ± 17.31 P-value 0.01 0.04 0.40 - 0.003 0.76 NGT obese children (n = 525) P/P = 402 12.00 ± 0.59 4.90 ± 0.22 27.92 ± 0.30 4.06 ± 0.06 2.659 ± 0.14 178.71 ± 8.66 P/A+AA = 123 10.77 ± 0.74 4.89 ± 0.42 28.55 ± 0.62 4.20 ± 0.14 2.368 ± 0.16 169.31 ± 16.48 P-value 0.18 0.61 0.33 0.37 0.18 0.37 Data are given as mean ± mean standard error. Quantitative traits are shown according to the Pro12Ala genotypes in healthy and NGT obese French populations under a dominant model. Z score of BMI was used rather than BMI in children. Quantitative traits were studied using a general linear model procedure, taking into account gender, age, BMI and genotype effect. BMI was adjusted only by gender and age. Z score of BMI was compared according to the Pro12Ala genotypes using the T-test. Pro allele is indicated as P and Ala allele as A. :::	PubMed Central	2024-06-05T03:55:55.830817	2005-3-22	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084346/", "journal": "BMC Med Genet. 2005 Mar 22; 6:11", "authors": [ { "first": "Maya", "last": "Ghoussaini" }, { "first": "David", "last": "Meyre" }, { "first": "Stéphane", "last": "Lobbens" }, { "first": "Guillaume", "last": "Charpentier" }, { "first": "Karine", "last": "Clément" }, { "first": "Marie-Aline", "last": "Charles" }, { "first": "Maïté", "last": "Tauber" }, { "first": "Jacques", "last": "Weill" }, { "first": "Philippe", "last": "Froguel" } ] }
PMC1084347	Background ========== Type-III secretion is one of five different types of protein secretion employed by Gram-negative bacteria to move proteins from the cytoplasm across two membranes to the external milieu \[[@B1]-[@B5]\]. Any given type-III secretion system (T3SS) consists of a multi-protein complex that spans both the inner and outer membranes and the periplasm so that proteins are delivered to the exterior in an ATPase-dependent fashion without a periplasmic intermediate. Type-III secretion systems can be classified into two major groups: those associated with flagellar biosynthesis and those associated with interactions between bacteria and eukaryotic cells \[[@B5]\]. Type-III secretion is thus central to our understanding of bacterial motility, virulence, symbiosis, and ecology. Type-III secretion also provides an attractive drug and vaccine target \[[@B6]\] and has been exploited in the biotechnology arena as a antigen delivery system \[[@B7],[@B8]\] The important human pathogens, enteropathogenic and enterohaemorrhagic Escherichia coli(EPEC and EHEC respectively) utilise type-III secretion to subvert eukaryotic signalling pathways by injecting bacterial effector proteins into the host cell cytoplasm \[[@B1],[@B9]-[@B12]\]. Within these pathovars, a well-characterised T3SS is responsible for the development of a characteristic attaching-effacing (AE) lesion and for other effects on enterocyte function \[[@B9],[@B11]-[@B13]\] (Figure [1](#F1){ref-type="fig"}). In common with most other T3SSs, this system is encoded by a \"pathogenicity island\" (in this case termed the \"locus of enterocyte effacement\" or LEE), which contains virulence genes clustered on the chromosome and acquired en bloc by horizontal gene transfer \[[@B13]-[@B16]\]. Some strains of the rarely isolated putative pathogen of humans, Escherichia albertii(formerly misidentified as Hafnia alvei), and of the mouse pathogen Citrobacter rodentium, which causes transmissible murine colonic hyperplasia, have been shown to induce AE lesion formation and to possess the LEE \[[@B17]-[@B20]\]. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### *1a: A comparison between the E. coliO157:H7 LEE and the Yersinia enterocoliticaYop-Ysc clusters. 1b: Graphical representation of the LEE-encoded type-III secretion system.The LEE region between EspF and Ler is shown (4589000--4620000 of the E. coliO157:H7 RIMD 0509952 genome. Accession number: dbj\\|BA000007.2). The Yop-Ysc cluster between YopM and YscL is shown (15000--40000 of the Yersinia enterocoliticaplasmid pYVe227 Accession number: dbj\\|BA000007.2gb\\|AF102990). Genes shaded in the same solid colour exhibit previously known homologies between the LEE and Ysc/Yop cluster. Genes shaded in the same colour but with horizontal hatching are homologues where we have added information in this paper. Genes which show no apparent homology to any the Ysc/Yop system, but which encode proteins that appear in our graphical representation are shaded with grey/white diagonal hatching (espB, espA, sepZ). Uncharacterized genes where there is no homology, and where their gene products do not appear in our cartoon are shaded solid grey. In the cartoon of the LEE-encoded type-III secretion system proteins are shaded in the same colour as their corresponding genes in the map of the gene cluster. Model based on the type III secretion model from KEGG <http://www.genome.jp/kegg/pathway/map/map03070.html>, and additional published data and images \[70, 73, 145, 156, 157\]. IM, inner membrane; PG, peptidoglycan layer; OM, outer membrane; EM, eukaryotic membrane. ::: ![](1471-2180-5-9-1) ::: When applied to biological sequences, homology is defined as similarity that arises because of descent from a common ancestor \[[@B21]\]. Assignments of sequence homology have a predictive value, in that sequence homology implies structural homology, and, less certainly, functional similarity. Although homologous sequences can diverge in function, or be recruited to a different physiological role or cellular compartment, the discovery of homology allows functional assignments to be transferred from a protein or gene that has undergone experimental investigation to a protein or gene that has not been studied in the laboratory (albeit tentatively and allowing for various common pitfalls in this process \[[@B22]\]). Any such functional assignment should always be treated as a working hypothesis rather than as an established fact. However, this approach can lead to new predictions about biological function that can then be tested in the laboratory, informing an existing programme of experimental work, or even opening up an entire line of enquiry where none existed before \[[@B22],[@B23]\]. At its most basic, an assertion of homology calls to our attention the \"null hypothesis\" -- that proteins with homologous sequences should behave in similar ways -- and, whenever this does not* appear to be true, provokes the question, why are these proteins behaving differently? Searches for homology between sequences are reliable only when the search program provides statistics that allow one to estimate how likely it is that randomly composed sequences could yield alignment scores at least as high as that obtained between the two sequences in question (an example of such a statistical evaluation is the e value reported by BLAST \[[@B24]\]). When the complete sequence of the LEE was first reported, an attempt was made to create a standard rational nomenclature based on experimental findings and functional predictions. E. coli-secreted proteins were given the generic name Esp (EspA, EspB, EspD) \[[@B11]\]. Components of the secretion apparatus that had obvious homologues in the better-characterized plasmid-encoded Ysc-Yop T3SS of Yersiniasp. were given the prefix Esc and the same suffix as their Yersinia homologues (EscC, EscD, EscJ, EscN, EscR, EscS, EscT, EscU and EscV). Proteins that apparently lacked Yersinia homologues, but appeared to be involved in type III secretion, retained the designation Sep (for \"secretion of E. coliproteins\"): originally SepZ for rOrf9 and SepQ for Orf17, but with the designations SepD for rorf6 and SepL for Orf23 subsequently seeing widespread use. The chaperone for the secretion of EspD was named cesD(\"chaperone for E. colisecreted protein D\"). The intimin gene was named eae(for \"E. coliattaching and effacing\") and the intimin receptor named Tir (for \"translocated intimin receptor\"). Since these original designations were made, several other LEE-encoded proteins have been characterized and re-named according to these conventions ((EspF, EspG, EspH; CesF, CesAB/CesA, CesD2), while some others have acquired names outside the original nomenclature, based on functional properties (Ler, Map, GrlR, GrlA) \[[@B25]-[@B33]\] The original assertions of homology for the LEE-encoded proteins -- and the functional assignments that flowed from them -- were based on the results of unsophisticated searches using the gapped BLAST program \[[@B11]\]. However, in the search for distant homologues, a considerable body of evidence now confirms that simple BLAST searches are far outperformed by more advanced, iterative methods such as PSI-BLAST \[[@B24],[@B34]-[@B39]\]. In addition, the growth of domain databases such as PFAM \[[@B40]\] and the steady accumulation of new sequence data on type-III secretion systems, especially from genome sequencing \[[@B2],[@B23],[@B41]\], provide a new backdrop against which original claims of homology (or lack of homology) can now be judged. Bearing these facts in mind -- and prompted by a recent assertion that \"nearly half of the LEE genes have no homologues \[[@B33]\]\" -- we have therefore undertaken a fresh bioinformatics analysis of the proteins encoded by the LEE, using highly sensitive methods for the detection of homology. Given the recent discoveries of numerous T3SS effectors secreted through the LEE-encoded T3SS \[[@B33],[@B42]-[@B46]\], for reasons of space we have opted to restrict our analysis to LEE-encoded components of the secretion and translocation apparatus and soluble cytoplasmic proteins associated therewith (i.e. chaperones and regulators). We will leave homology analyses of effectors (Tir, EspF, EspG, EspH, Map, SepZ \[[@B46]\] encoded within the LEE; potentially many dozens encoded outside the LEE) and functionally related proteins (intimin) for later publications. We conclude that the vast majority of LEE-encoded proteins do indeed possess homologues and that homology data can be used in combination with experimental data to make fresh functional predictions (Figure [1](#F1){ref-type="fig"}, Table [1](#T1){ref-type="table"}). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Summary of known characteristics of LEE-encoded genes and their associated proteins, with novel homology-based functional predictions. New names as suggested in body text of this article. Summary of effects on gene expression, type-III secretion and virulence in mice drawn from Deng et al \[33\]. Yeast two-hybrid results drawn from Creasey et al \[145\]. Predictions drawn from combination of homology data and cited works. T3S: involved in type-III secretion. ::: Gene (new name) Effect of deletion mutation on secretion Effect on pedestal formation / virulence Known interactions Homologues Domains Known functions New functional prediction --------------------- ---------------------------------------------- ---------------------------------------------- --------------------------------- ----------------------------------- ------------------------------------- ------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ *rorf1* \+ +/+ EspD YjiK SdiA-regulated domain (NHL repeats) SdiA-regulated? *espG* \+ +/+ VirA Translocated effector *ler* No expression -/- H-NS, StpA Positive regulator Might form heterodimers with H-NS, StpA *orf2 (escE)* \- Orf29 YscE, PscE, SsaE, CV2595, YPO0259 T3S Interacts with C-terminus of SepL; Functionally important conserved residues can be identified by scrutiny of homologues *cesAB* ± EspA, EspB -/- EspA, EspB none Chaperone for EspA and EspB *orf4* \- -/- YPO0264, SsaK, CV2589 T3S Functionally important conserved residues can be identified by scrutiny of homologues *orf5 (escL)* \- -/- YscL, FliH T3S Binds to and regulates activity of ATPase EscN *escR* \- -/- EscR, EscS, EscU, SepZ EspD YscR, FliP PF00813 T3S *escS* \- -/- EscR, EspD YscS, FliQ PF01313 T3S *escT* \- -/- YscT, FliR PF01311 T3S Physically associated in the basal body with EscU in a 1:1 ratio; C-terminus is cytoplasmic \[158\] *escU* \- -/- EscrR, EscI, EspD Yscu, FlhB PF00771 T3S Physically associated in the basal body with EscT in a 1:1 ratio \[158\]; cleaved at NPTH motif \[159\]; regulates substrate specificity of secretion system \[160\]. *rorf3 (etgA)* ± ±/± EscI lytic transglycosylase IagB, IpgF Lytic transglycosylase needed to open gap in peptidoglycan for assembly of secretion system \[58\]; potential drug target *grIR* \+ +/++ GrlR/GrlA Bongori regulator Negative regulator Will work in concert with GrlA *grIA* No expression -/- GrlR Bongori regulator, CaiF Positive regulator Will work in concert with GrlR *cesD* -/EspD ±/- Other TPR chaperones Chaperone for EspD TPR chaperone \[115\] *escC* \- -/- EscD YscC, MxiD, InvG T3S Interacts with peptidoglycan \[161\], Dsb-dependent domain missing *sepD* -/Translocators -/- SepL None Switches translocator/effector secretion Works in concert with SepL *escJ* \- -/- YscJ, PrgK, FliF PF01514 T3S EscD and EscJ form ring-like structure in inner membrane \[66, 162, 163\] *rorf8 (escI)* \- -/- EscU, rORF3 PrgJ, MxiI, YscI T3S Forms inner rod within base of needle complex \[65, 68-70\]. *sepZ* \+ +/± EscR None Translocated effector *orf12* \- -/- SsaM T3S Might play a role in switching translocator/effector secretion *escN* \- -/- CesT, Tir \[50\] YscN T3S ATPAse Forms hexameric complex *escV* \- -/- LcrD, FlhA PF00771 T3S C-terminal cytoplasmic domain might play a role in gene regulation \[164, 165\]. *orf15* \- -/- None T3S *orf16* ±/Translocators ±/± None Secretion of translocators FliK/YscP analogue? *sepQ (escQ)* \- -/- YscQ, FliN COG1887 T3S Forms ring within basal body \[56\]; interacts with EscL \[54\] *espH* \+ +/++ Translocated effector *cesF* ±/EspF? +/+ Chaperone for EspF *map* \+ +/++ CesT Translocated effector *tir* \+ -/- CesT Tir Translocated effector *cesT* ±/Tir -/- Map, EspF, Tir, CesT Many Chaperone for Tir Possible regulatory role, by analogy with ExsC? *eae* \+ -/- Adhesin (intimin) *escD* \- -/- YscD, PrgH YscD: FHA and BON domains T3S EscD and EscJ form ring-like structure in inner membrane \[66, 162, 163\]; BON domains mediate binding to phospholipids; FHA domain mediates protein-protein interactions and signalling *sepL* -/Translocators -/- SepD YopN, TyeA Switches translocator/effector secretion single protein mediating effects of YopN and TyeA; internal interaction between YopN- and TyeA-like moieties; also a translocated effector? *espA* \+ -/- CesAB flagellin Translocator Half-flagellin model (see text); dimerizes before polymerizes; lacks D0 and D3 domains *espD* \+ -/- EspD, EscRSU, CesD, rORF1, SepZ YopB etc Translocator *espB* \+ -/- CesAB YopD etc Translocator *cesD2* ±/EspD? +/± LcrR Chaperone for EspD LcrR will bind YopB *escF* \- -/- YscF Major needle component *orf29* \- -/- Orf2 YPO0261, SsaI, H, CV2586 T3S Functionally important conserved residues can be identified by scrutiny of homologues *espF* \+ +/+ Translocated effector ::: Results and discussion ====================== The conserved type-III secretion apparatus ------------------------------------------ Several LEE-encoded proteins have already been assigned Esc designations to reflect their similarities to the Ysc proteins in the Yersinia Ysc-Yop system \[[@B47]\]. BLASTP-based comparisons of the predicted components of basal complex confirms significant near-full-length homology between EscC, EscD, EscJ, EscN, EscR, EscS, EscT, EscU and EscV and the respective Ysc proteins and homologues (Figure [1](#F1){ref-type="fig"}, Table [1](#T1){ref-type="table"}), although, confusingly, YscV is more commonly known as LcrD, and EscD is sometimes termed Pas \[[@B1],[@B47],[@B48]\]. In the cases of EcsN, EscR, EscS, EscT, EscU and EscV, homology clearly extends to components of the flagellar apparatus \[[@B1]\]. These unequivocal homologues suggest that observations on these proteins in other systems can safely be generalised to the LEE-encoded system -- as has been confirmed in several recent papers \[[@B49],[@B50]\] -- and can be used to frame novel hypotheses about their function (Table [1](#T1){ref-type="table"}, Figure [1](#F1){ref-type="fig"}). However, it is important to remain on guard for minor but, perhaps, important differences between systems: for example, the C-terminal \~100-residue domain from YscC, which is thought to house two DsbA-dependent disulfide bridges is missing from EscC and many other homologues \[[@B51]\]. As noted previously, YscD contains a cytoplasmic FHA domain \[[@B52]\]. PSI-BLAST searches confirm that EscD also contains a cytoplasmic FHA domain (Figure [2](#F2){ref-type="fig"}). In addition, these searches revealed the presence of at least one putative phospholipid-binding domain (also called a BON domain \[[@B53]\]) in the periplasmic portion of the protein (Figure [3](#F3){ref-type="fig"}). Two compelling hypotheses arise from these observations. The first of these is that, given the established role of FHA domains in signalling \[[@B52]\], it is tempting to speculate that the cytoplasmic FHA domain of EscD might be involved in signal transduction (alternatively, it might represent a molecular fossil, now fully adapted to a purely structural role). Secondly, as BON domains are thought to mediate binding to phospholipids in a variety of other proteins \[[@B53]\], it is likely that they play a similar role in EscD and its homologues. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Multiple alignment of the FHA domain from EscD with other FHA domains.Alignment is presented using default CHROMA \[155\] settings: Consensus abbreviations (amino acids): a, aromatic (FHWY, blue lettering on a dark yellow background); b, big (EFHIKLMQRWY, blue on light yellow); h, hydrophobic (ACFGHILMTVWY, black on dark yellow); l, aliphatic (ILV, grey on dark yellow); p, polar (CDEHKNQRST, blue on white); s, small (ACDGNPSTV, dark green on white); t, tiny (AGS, light green on white); -, negatively-charged (DE, red on white); and, +, positively-charged (KR, blue on white), c, charged (DEKRH, pink on white. Organism and gene name abbreviations used: EHEC (Escherichia coliO157:H7) EscD (ECs4558/dbj\\|BAB37981), Vp (Vibrio parahaemolyticus) PscD(ref\\|NP\_798074), Pa(Pseudomonas aeruginosa) PscD (ref\\|NP\_250408), Ah (Aeromonas hydrophila) AscD(gb\\|AAS91829), Pl (Photorhabdus luminescens) LscD(gb\\|AAO18032), Yp (Yersinia enterocolitica) YscD(gb\\|AAC37021), Dv (Desulfovibrio vulgaris) YscD(ref\\|YP\_009153), Ct (Chlamydia trachomatis) CT664 (NP\_220183), Cp (Chlamydophila pneumoniae) CpB0739(gb\\|AAP98668), Xa (Xanthomonas axonopodis) HrpD5 (gb\\|BAD29996), Ea (Erwinia amylovora) HrpQ(gb\\|AAB06000). The remaining 12 sequences are representative members of the SMART FHA domain (SM00240). Gene names, numbering and alignment are as presented in the SMART FHA family alignment <http://www.smart.ox.ac.uk/>. ::: ![](1471-2180-5-9-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### A multiple alignment of the phospholipid-binding (BON) domain from EscD with other BON domains.The most obvious BON domain from EscD is shown. However, from patterns of residue conservation centred on conserved glycines in the EscD family of proteins (data not shown), we suspect that there may be one or two more BON-like domains within the cytoplasmic portion of EscD. For sequences 1--11 organism and gene name abbreviations are as for Figure 2 legend. The remaining 16 sequences are representative members of the Pfam BON domain (PF04972). Gene names, numbering and alignment are as presented in the Pfam:BON family alignment <http://pfam.wustl.edu/>. Alignment is presented using default CHROMA settings (see Figure 2 legend). ::: ![](1471-2180-5-9-3) ::: The missing EscL and EscQ proteins ---------------------------------- Although, as noted above, the majority of the \"Ysc proteins\" that form conserved components of the secretion apparatus in Yersinia have easily recognisable \"Esc protein\" counterparts in the LEE-encoded system, there are two salient exceptions -- in the current annotation of the LEE, there are no equivalents of YscQ and YscL. YscQ is a member of the FliN family of proteins and is thought to be a component of the basal secretion complex \[[@B54]\]. Although PSI-BLAST searches fail to find any YscQ homologue among the LEE-encoded proteins, a search of the NCBI\'s CDD database \[[@B55]\] identifies a FliN domain (COG1886) in the C-terminal half of SepQ (Figure [4](#F4){ref-type="fig"}). This suggests that SepQ plays a similar role to YscQ and FliN -- allowing new hypotheses to be framed (Table [1](#T1){ref-type="table"}) -- and should be re-named EscQ. Although, the high degree of divergence of the EscQ C-terminal domain from other YscQ/FliN family members is puzzling, it is clear from scrutiny of the recently solved structure of the homologous domain from HrcQB \[[@B56]\] that SepQ possesses most of the conserved motifs common to this domain family, and must adopt a similar fold (Figure [4](#F4){ref-type="fig"}). ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Multiple alignment of SepQ/EscQ with related type-III secretion proteins.The sequence and secondary structure of the crystal structure of the C-terminal domain of the HrcQb protein from Pseudomonas Syringae Pv. Phaseolicola(pdb\\|1OY9) are shown 56. Sequences 2--21 (including SepQ) are from COG1886. 1OY9 and the remaining 10 sequences are representative members of Pfam:SpoA. Organism and gene name abbreviations are abbreviated as follows: EHEC (Escherichia coliO157:H7) SepQ (ECs4565/gb\\|AAG58829), Yp (Yersinia pestisCO92) YscQ (ref\\|NP\_403921), Ct YscQ (ref\\|NP\_220191), At (Agrobacterium tumefaciensstr. C58) AGRc988p(ref\\|NP\_353589), Ba (Buchnera aphidicola) FliN (sp\\|P57183), Ci (Caulobacter vibrioides) FliN(sp\\|Q03593), Cc (Caulobacter crescentus) FliN(ref\\|NP\_420978), Cj (Campylobacter jejuni) FliY (ref\\|NP\_281274) and FliN (ref\\|NP\_281542), EHEC EpaO (ECs3726/sp\\|Q8X6F0), Pa (Pseudomonas aeruginosa) PscQ (ref\\|NP\_250385), Rs (Ralstonia solanacearum) HrcQ (ref\\|NP\_522422), St (Salmonella typhimurium) SsaQ(sp\\|P74860) and SpaO (sp\\|P40699), Yp FliN(ref\\|NP\_404342.1), Aa (Aquifex aeolicus) FliN (sp\\|O67495), Li (Listeria innocua) Lin0701 (ref\\|NP\_470044), Lin0706 (ref\\|NP\_470049), and Lin0708 (ref\\|NP\_470051), Ml (Mesorhizobium loti) hrcQ (ref\\|NP\_106868), At FliM (sp\\|Q44457), Hp (Helicobacter pylori) FliM (sp\\|O25675), Bs (Bacillus subtilis) FliM (sp\\|P23453), Bb (Borrelia burgdorferi) FliM(sp\\|57511), Tp (Treponema pallidum) FliM (sp\\|P74927), Vp FliM (sp\\|Q9Z6GI), Rp (Rhodobacter sphaeroides) FliM (O85118), Sf (Shigella flexneri) SpaO (sp\\|P35534), Bp (Burkholderia pseudomallei) SctQ (Q9ZGR1), Ea (Erwinia amylovora) HrcQ (Q46645). Alignment is presented using default CHROMA settings (see Figure 2 legend). ::: ![](1471-2180-5-9-4) ::: YscL is a member of the FliH family of proteins. In flagellar systems, FliH binds to and regulates the activity of the ATPase FliI \[[@B57]\], and YscL is also known to bind to YscN \[[@B54]\]. PSI-BLAST searches across the NCBI\'s non-redundant database with YscL fail to identify any homologue in the LEE system, because of contamination with low-complexity eukaryotic proteins. However, if a PSI-BLAST search with Orf5 is restricted to bacterial proteins, after one iteration, YscL appears in the results list (15% identity 29/184 residues; e value 0.002). After two iterations, several more FliH homologues are found. Furthermore, consistent with the recent suggestion that the YscL-YscN interaction mirrors similar interactions in other ATPases \[[@B54]\], weak similarity is also reported between Orf5 and several F^0^ATPase beta subunits (data not shown). A multiple alignment confirms the presence of conserved residues within FliH, YscL and Orf5 (Figure [5](#F5){ref-type="fig"}). It thus seems likely that Orf5 is a homologue of YscL and FliH, plays a similar role (Table [1](#T1){ref-type="table"}, Figure [1](#F1){ref-type="fig"}) and should be re-named EscL. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Multiple alignment of Orf5/EscL with related type-III secretion proteins.Organism and gene name abbreviations: EHEC (Escherichia coliO157:H7) EscL (ECs4584/sp\\|O85643), Yp(Yersinia pestisCO92) YscL(ref\\|NP\_395195), Bb (Borrelia burgdorferi) FliH(ref\\|NP\_212423), Tm(Thermotoga maritim) FliH(ref\\|NP\_228034), Ph(Pyrococcus horikoshiiOT3) Ph1978(ref\\|NP\_143803), E. coliK-12 FliH (ref\\|NP\_416450), Tm TM1614(ref\\|NP\_229414). Note that Ph1978 and TM1614 are annotated as a hypothetical protein and an ATPase synthase F0, subunit b, respectively. Alignment is presented using default CHROMA settings (see Figure 2 legend). ::: ![](1471-2180-5-9-5) ::: A lytic trans-glycosylase: a novel drug target? ----------------------------------------------- Both domain searches and PSI-BLAST searches show that rOrf3 encodes a lytic transglycosylase. Similar enzymes, which occur in type II, type III and type IV secretion systems, are responsible for enlarging gaps in the peptidoglycan meshwork to allow efficient assembly and anchoring of supramolecular transport complexes in the cell envelope \[[@B58]\]. The presence of a dedicated transglycosylase within this type-III secretion system provides an obvious target for the development of anti-microbial agents that specifically inhibit type-III secretion, particularly as structural data is available for homologous enzymes \[[@B59]-[@B61]\]. We propose that rOrf3 be re-named EtgA for E. colitransglycosylase. Curiously, we have been unable to find a homologue of EtgA in the Ysc-Yop system. The needle ---------- Despite its name, EscF, a major component of the needle, does not show significant similarity to YscF on a simple BLASTP search (19% identity; e value of only 0.88), but the unimpressive significance score can be explained by the short length of the sequence (only 73 residues); the e value rises to a more respectable 0.003 within two iterations of a PSI-BLAST search. Furthermore, it is clear from published experimental work that YscF and EscF play equivalent roles as needle components \[[@B62]-[@B64]\]. In the Spi-1 and Mxi-Spa systems, two small proteins are associated with the needle -- PrgI/PrgJ and MxiH/MxiI respectively \[[@B65]-[@B68]\]. There is experimental evidence from both systems to suggest that only one of the proteins (PrgI/MxiH) is the major subunit component of the needle, while the function of the second protein (PrgJ/MxiI) was, until recently, unclear -- it was initially thought to form a cap for the needle \[[@B65],[@B68],[@B69]\]. However, the function of PrgJ (and by extrapolation, of its homologues) has now been elucidated -- it forms the inner rod within the base of the needle \[[@B70]\]. A PrgJ-like component has yet to be described in the LEE-encoded system. However, PSI-BLAST searches indicate that rORF8 is homologous to EprJ, PrgJ, MxiI and to an uncharacterised protein from the Ysc-Yop system, YscI (Figure [6](#F6){ref-type="fig"}). This suggests that rORF8 and YscI play similar roles to PrgJ/MxiI as components of the inner rod and that rORF8 should be re-named EscI. Interestingly, PSI-BLAST searches also suggest that the PrgI-like and PrgJ-like proteins are paralogous. ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Multiple alignment of rOrf8/EscI with related type-III secretion proteins.Organism and gene name abbreviations as follows: EHEC (Escherichia coliO157:H7) EscI (ECs4572/sp\\|O85634) and EprJ (ECs3717/sp\\|Q8X6G5), Sf (Shigella flexineri) MxiI (sp\\|Q06080) and MxiH (sp\\|Q06079), St (Salmonella typhimurium) PrgJ (sp\\|P41785) and PrgI (sp\\|P41784), and Yp (Yersinia pestisCO92) YscI (Q00933). Alignment is presented using default CHROMA settings (see Figure 2 legend). ::: ![](1471-2180-5-9-6) ::: Homology between EspA and Flagellin ----------------------------------- The LEE-encoded T3SS is so far unique in possessing a filamentous organelle, the EspA pilus, that resembles the flagellum, forms an extension to the needle complex and is thought to act as a molecular syringe providing a continuous protein-translocation channel between the bacterial and host cells \[[@B71]-[@B73]\]. The pilus primarily consists of EspA, but also requires EspD for its biosynthesis \[[@B72]\]. EspB and EspD form a pore in the host cell plasma membrane \[[@B74]\]. PSI-BLAST searches on the ViruloGenome site with EspA reveal homology within two iterations to numerous flagellins, including the Salmonella flagellin for which a structure is available \[[@B75],[@B76]\] (Figure [7](#F7){ref-type="fig"}). Although the precise alignment of the N-terminal region of EspA with flagellin appears critically dependent on the PSI-BLAST settings used, the alignment of the C-terminal region of EspA with flagellin appears robust. This allows certain hypotheses to be framed about the structure of EspA and the EspA filament. ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Multiple alignment of EspA with other EspA-like proteins and flagellins.The sequence and secondary structure of S. typhimuriumSjw1665 R-type straight flagellar filament is shown (pdb\\|1UCU) \[76\]. Organism and gene names are abbreviated as follows: EHEC (Escherichia coliO157:H7) EspA (dbj\\|BAB37979), EPEC (Escherichia coliE2348/69) EspA (gb\\|AAC38394), Et (Edwardsiella tarda) EspA (gb\\|AAN52733), Cr (Citrobacter rodentium) EspA(gb\\|AAL06381), Cv (Chromobacterium violaceum) CseB1 (gb\\|AAQ60253), Cv CseB2 (gb\\|AAQ60252), Cv CseB3 (gb\\|AAQ60250), Cv CseB4 (gb\\|AAQ60249), Se (Salmonella enterica Typhi) SseB (ref\\|NP\_456130), Bb (Bartonella bacilliformis) Fla1 (gb\\|AAO33576), Rm (Rhizobium meliloti) Fla2(gb\\|AAA26277), Ap (Aquifex pyrophilus) FlaA(gb\\|AAA88923), Bo (Bordetella bronchiseptica) FlaA (gb\\|AAA22977), Lm, (Listeria moncytogenes) FlaA (gb\\|EAL07249), Hm (Helicobacter mustelae) FlaB(gb\\|AAA25017), Vp (Vibrio parahaemolyticus) LafA (dbj\\|BAC62891), Bh (Bacillus halodurans) Hag (dbj\\|BAB07335), Bs (Bacillus subtilis) Hag (gb\\|AAA22437), Le (Legionella micdadei) Fla (emb\\|CAA59172), Th (Treponema hydodysenteriae) FlaB2 (emb\\|CAA45081), Pm (Proteus mirabilis) FliC1(gb\\|AAA62396), E. coliK-12 FliC(gb\\|AAC74990). Alignment is presented using default CHROMA settings (see Figure 2 legend). ::: ![](1471-2180-5-9-7) ::: Firstly, surprisingly, the coiled-coil domain at the C-terminus of EspA appears to correspond notto the coiled-coil at the extreme C-terminus of flagellin, which forms one half of the flagellin D0 domain, but insteadto the C-terminal portion of the D1 domain (CD1) \[[@B75],[@B76]\]. In other words, EspA appears to lack a D0 domain. In the flagellar filament, D0 forms the inner tube, while D1 forms the outer tube \[[@B75]\]. Although D0 interactions are important in filament stability, mutant flagellins that lack the D0 domain can still assemble into straight filaments, albeit with a structure, termed Lt, distinct from the native flagellar filaments \[[@B77]\]. It is thus entirely plausible that the EspA pilus is assembled without the need for a D0 domain and resembles the Lt filament structure. One problem with this is that, unlike the native flagellar filament and the EspA filament, the Lt filament appears to lack a central pore. However, this is probably an artefact (it is hard to see how flagellin monomers could be exported without a pore) and by analogy with flagellin \[[@B76]\], one could predict that the central pore in the EspA filament is lined by conserved polar residues close to extreme C-terminus of EspA (e.g. Lys-192). Secondly, EspA appears to lack a D3 domain (and probably also anything homologous to the D2 domain, although it must contain an analogous surface-exposed domain). This fits in with the observation that D3 is highly variable in size and in sequence among flagellins and is not essential for flagellar filament formation in Salmonella\[[@B78]\]. However, it is interesting to note that two EspA homologues from Chromobacterium violaceumpossess a central insertion, which is likely to fold into an extensive surface-exposed domain, perhaps analogous to the flagellar D3 domain \[[@B41]\]. A third, crucial point concerns the role of the coiled-coil domains in EspA. The coiled-coil domains at the N- and C-termini of flagellin were initially thought to facilitate filament assembly by mediating inter-molecular interactions between neighbouring flagellin subunits \[[@B79],[@B80]\]. Based on the hypothesis that coiled coils might play an analogous role in the assembly of the EspA pilus \[[@B72],[@B81]\], mutagenesis experiments were performed and showed that the C-terminal coiled coil domain of EspA is required for filament assembly \[[@B82]\]. However, the interpretation of these experiments must now be revised in the light of the flagellar structure and homology data. When the flagellar filament structure was solved, it became clear that, contrary to expectation, the coiled-coil domains mediated interactions withinsubunits, rather than betweenthem \[[@B76]\]. Thus, by extension, it is likely that, for both EspA and flagellin, the coiled-coil interaction is a necessary but not sufficient requirement for filament assembly. Coiled-coil prediction programs and PSI-BLAST searches suggest that, in addition to lacking the D0 and D3 domains, EspA lacks any homologue of the N-terminal helix (ND1) of the coiled-coil that forms the D1 domain. However, recent structural information on EspA complexed with its chaperone CesA indicates that EspA contains a short N-terminal coiled-coil domain in addition to the already recognized much longer C-terminal coiled-coil domain \[[@B83]\]. It is thus possible that EspA folds up as a \"mini-flagellin\", so that together these two coiled-coil domains mediate an intra-subunit interaction similar to that seen in the flagellar monomer. However, this does not fit well with the apparent discrepancy in length between the two domains. An alternative possibility is that EspA acts as \"half-a-flagellin\", so that it dimerizes through an inter-subunit coiled-coil interaction, and then the dimer polymerizes, like flagellin, through hydrophobic interactions but between dimers rather than monomers. The \"mini-flagellin\" and \"half-flagellin\" models could be distinguished in the laboratory. For example if the half-flagellin model holds true, it should be possible through mutagenesis to identify mutations that disrupt polymerisation but leave the coiled-coil-mediated dimerization potential intact (an approach that might make it possible to grow crystals for structure determination). Furthermore, it should be possible to distinguish dimerization from subsequent polymerisation by physicochemical methods. Finally, under this model, there is a potential for the formation of heterodimers -- an idea supported by the existence of four C. violaceumEspA homologues, encoded by two pairs of adjacent genes. In each pair, the first gene encodes a short conventional EspA homologue, while the second gene encodes a long EspA homologue containing a central D3-like insertion \[[@B41]\]. According to the half-flagellin model, the short and the long EspA-like proteins would have the potential to form heterodimers, which would then polymerise into a filament. This prediction could be tested experimentally, e.g. by showing that these EspA homologues form long/short dimers but do not engage in self-self interactions. The Toll-like receptor 5 (TLR5) recognizes the D1 domain of flagellin \[[@B84]\]. Although EspA lacks several of the residues identified as critical in binding (data not shown), the overall conservation of the CD1 domain raises the question of whether TLR5 might recognize EspA. However, recent experimental investigations suggest it does not (K. D. Smith, personal communication). Yet, this in turn raises the question of whether this lack of interaction is the result of genetic drift or of positive selective pressure on EspA to avoid recognition by TLR5. PSI-BLAST searches with EspB prove unhelpful, as the compositional bias of the protein, particularly in its coiled-coil domains, attracts numerous simple-sequence proteins, even when a filter and composition-based statistics are used \[[@B81],[@B85]\]. However, a comparison of domain architecture and genetic location with translocation-pore proteins/genes from other systems (e.g. SipC, IpaB, YopD) suggests functional and structural similarities. PSI-BLAST searches with EspD prove more fruitful, especially when composition-based statistics are employed -- they reveal significant similarities to several other T3SS translocon-pore proteins. For example, when EspD is used as a query sequence for a PSI-BLAST search on the NCBI NR database under default settings, YopB is found in the first round with an e value (0.001) within the PSI-BLAST threshold, while SseC achieves significance (4e-07) on the second PSI-BLAST round. Homologies and Gene fusions in the SepL/YopN-TyeA family -------------------------------------------------------- SepL is a LEE-encoded protein that has been said to have no homologues and is required for AE lesion formation and secretion of translocation proteins and for the translocation of -- but not the secretion of -- effectors \[[@B86],[@B87]\]. PSI-BLAST searches reveal homology to two proteins from the Ysc-Yop system: most of the SepL sequence (up to residue 267) is homologous to YopN (sometimes called LcrE), while the C-terminal 83 amino acids are homologous to TyeA (Figure [8](#F8){ref-type="fig"}). This assertion of homology between SepL and YopN/TyeA is supported by several other lines of evidence (i) the yopNand tyeAgenes are adjacent to one another; (ii) the YopN and TyeA proteins interact with one another; (iii) it is well established that genes encoding interacting domains often occur adjacent to one another and can undergo gene fusion in some genomes, (iv) experimental evidence has recently emerged showing that a chimaeric YopN/TyeA protein can be produced by programmed frame shifting in some Yersinia strains \[[@B88]-[@B91]\]. ::: {#F8 .fig} Figure 8 ::: {.caption} ###### Multiple alignment of SepL with related type-III secretion proteins, including YopN-like and TyeA-like proteins.The N-terminal (YopN-like) and C-terminal (TyeA-like) domains are shown separately for clarity. Organism and gene name abbreviations as follows: EHEC(Escherichia coliO157:H7) SepL (ECs4557/dbj\\|BAB37980), Se (Salmonella enterica Typhi) SsaL (gb\\|AAL20336), Cv (Chromobacterium violaceum) InvE (gb\\|AAQ60301), St (Salmonella typhimurium) InvE (pir\\|\\|A46138), Sf (Shigella flexineri) MxiC (gb\\|AAL72332), Er (Erwinia chrysanthemi) HrpJ (gb\\|AAO34609), Pt (Pectobacterium atrosepticum) HrpJ (emb\\|CAD43175), Bp (Bordetella pertussis) BopN (emb\\|CAE42532), Cp (Chlamydophila pneumoniae) CopN (gb\\|AAD18473), Cc (Chlamydophila caviae) CopN (gb\\|AAP05204), Ah (Aeromonas hydrophila) AopN(gb\\|AAR26331) and Acr1(gb\\|AAR26332), Pl (Photorhabdus luminescens) LopN(gb\\|AAO18045) and LssA(gb\\|AAO18046), Pa (Pseudomonas aeruginosa) PopN(gb\\|AAC45939) and Pcr1(gb\\|AAC45940), Ye (Yersinia enterocolitica) YopN(gb\\|AAD16823) and TyeA(gb\\|AAN37519), Vp (Vibrio parahaemolyticus) PopN(dbj\\|BAC59930) and VP1666(dbj\\|BAC59929). Alignment is presented using default CHROMA settings (see Figure 2 legend). ::: ![](1471-2180-5-9-8) ::: This new-found homology also fuels two predictions: (i) the C-terminal residues 268--351 of SepL, corresponding to TyeA, interact with the residues immediately upstream (229--267) in the region of SepL corresponding to YopN residues 242--280; (ii) the interaction between YopN and TyeA must occur in a 1:1 ratio -- this could be confirmed by creating a yopN-tyeAfusion that lacks a start codon for tyeA, and then showing that this construct could complement a yopN/tyeAmutant to give normal type-III secretion. Further insights into type-III secretion arise when additional PSI-BLAST searches are applied to the homologues retrieved when SepL is used as the query sequence. Firstly, it is clear that the Ysc-Yop system is unusual in separating out the two SepL-related sequences into two proteins. This arrangement occurs only in systems closely related to the Ysc-Yop system (i.e. T3SSs from Pseudomonas aeruginosa, Photorhabdus luminescens, Aeromonas salmonicida, Vibrio parahaemolyticus); in other systems, there is a single protein corresponding to a YopN-TyeA chimaera (e.g. SsaL in Spi-2, HrpJ in Erwiniaand related species, BopN in Bordetella, CopN in Chlamydia\[[@B92]\]). Secondly, through careful use of PSI-BLAST, it is possible, through repeated searching, to establish unequivocal homology between SepL-like proteins from each of the well-characterised animal-pathogen-related T3SSs -- not just between SepL and YopN-TyeA, but also encompassing InvE from the Spi-1 system and MxiC from the Shigella system (Figure [8](#F8){ref-type="fig"}). This implies that functional assignments for any one of these proteins should be generalisable to all the others -- a prediction borne out by the similar phenotypes ascribed to these proteins, which all appear to play a role in regulating the translocation of effectors \[[@B86],[@B87],[@B93]-[@B98]\]. Indeed, it is even possible that some reported differences in phenotype (e.g. the surface location of YopN versus the apparent cytoplasmic location of InvE; the fact that YopN appears to control secretion of all Yops, whereas InvE is more selective in controlling secretion of translocators rather than effectors \[[@B98]\]) might represent idiosyncrasies of experimental approach than genuine biological differences. However, even if these do represent true differences, this begs the question: how and why have homologous proteins come to acquire such divergent functions? SepL interacts with another protein of unknown function SepD \[[@B87]\]. Unfortunately, homology searches afforded no insights into the function of SepD. However, given that sepDis located next to escCin the LEE, while yscBoccupies an analogous position (next to yscC) in the ysccluster, it is tempting to speculate that they might play similar roles. Also, no homologues could be found in the LEE-encoded system of LcrV and LcrG, which are thought to regulate entry into the Ysc secretion apparatus from the cytoplasm \[[@B99],[@B100]\]. Similarly, we could not detect any LEE-encoded homologues of the needle-length molecular ruler YscP \[[@B101]\], even though, as we have reported elsewhere \[[@B23]\], PSI-BLAST searches do reveal sequence similarity at the C-terminus between YscP and the flagellar hook-length-determining protein FliK, implying that they employ similar mechanisms \[[@B102],[@B103]\]. Nor could we find any homologue of InvJ/Spa32, which, although showing no detectable sequence similarity to YscP or FliK, is thought to play a similar role to these proteins in needle-length determination in the Spi-1 and Mxi-Spa systems \[[@B67],[@B104]\]. Despite the frustrating lack of sequence similarities, two pieces of tantalising circumstantial evidence link FliK/YscP with InvJ/Spa32, and hint at the identity of an equivalent protein in the LEE-encoded system. Firstly, all four proteins are characterised by the presence of large amounts of random coil on secondary structure predictions (S. Lloyd, personal communication). Secondly, the genes for these proteins always occupy an equivalent position within the type-III secretion gene cluster: always next door but one to and downstream of the gene encoding the ATPase. When these criteria are applied to the LEE, an obvious candidate for needle-length determination emerges: Orf16 (28% random coil according to SOPMA \[[@B105]\]). Almost nothing is known about this gene/protein, which has no known homologues outside of the LEE. However, a mutant in this gene in Citrobacteris deficient in translocator secretion, but is still able to secrete effectors normally, and is severely attenuated in the mouse \[[@B33]\]. These results are consistent with the hypothesis that Orf16 controls the switch between export of needle components and components of translocation apparatus. However, in the absence of structural or experimental data, it is premature to make an assertion of homology between Orf16 and YscP/FliK, particularly as tentative assignments tend to harden into dogma and then preclude consideration of other possibilities i.e. that different systems employ different mechanisms for the same purpose. Clearly, additional experimental data are needed to resolve this issue. Chaperones ---------- Five LEE-encoded proteins have been designated chaperones on the basis of experimental work: CesT, CesF, CesD, CesD2 and CesAB \[[@B29]-[@B31]\], \[[@B106]-[@B108]\]. For one of these, CesT, which chaperones the effectors Tir and Map, the structure is available \[[@B109]\]. PSI-BLAST searches with the CesT sequence link it to CesF and to two other virulence-related T3SS chaperones for which structural information is available (SycE, SicP, SigE \[[@B109]-[@B112]\]) and also identify numerous other uncharacterised T3SS-associated proteins as chaperones (Figure [9](#F9){ref-type="fig"}). Interestingly, the results show homology between CesT/CesF and ExsC from P. aeruginosa. As this protein has recently been reported to act as an anti-anti-activator, controlling transcription of T3SS genes in P. aeruginosa\[[@B113],[@B114]\], it is plausible that CesT and/or CesF might play similar roles in gene regulation. ::: {#F9 .fig} Figure 9 ::: {.caption} ###### Multiple alignment of CesT with related type-III secretion chaperones.Organism and gene name abbreviations as follows: As(Aeromonas salmonicida) orfX(gb\\|AAK83051), Pa(Pseudomonas aeruginosa) PA3842(ref\\|NP\_252531), Cv(Chromobacterium violaceum) CV0975(ref\\|NP\_900645), Pl(Photorhabdus luminescens) unkown(gb\\|AAO18077), Ye(Yersinia enterocolitica) SycT(ref\\|NP\_783658), Bt(Bordetella pertussis) BP0499(ref\\|NP\_879351), Pl plu4749(ref\\|NP\_931911), Ps(Pseudomonas syringae) SchF(ref\\|NP\_790352), Sf(Shigella flexineri) IpgE(gb\\|AAP78997), Ye YsaK(gb\\|AAB69191), Ea(Erwinia amylovora) DspF(gb\\|AAC04851), Ps AvrF(ref\\|NP\_791203), EPEC(Escherichia coliE2348/69) CesF(gb\\|AAC38388), Ps Pspto1369(ref\\|NP\_791196) and Pspto4589(ref\\|NP\_794340), Pl plu0829 (ref\\|NP\_928172), Ps Pspto4721(ref\\|NP\_794464), Bt BP2236(ref\\|NP\_880880), Pa ExsC(gb\\|AAC46214), Pl plu3753(ref\\|NP\_930959), Ye Orf155(ref\\|NP\_783720), Vp VP1687(ref\\|NP\_798066), Ps SchA(ref\\|NP\_795083), Dv Dvu2312(ref\\|YP\_011525), Lm(Leishmania major) L4830.13(emb\\|CAC22615), Lm P883.10(emb\\|CAC37204), Ct CT043(ref\\|NP\_219546), Xa(Xanthomonas axonopodis) HpaB(ref\\|NP\_640751), Ps Pspto1374(ref\\|NP\_791201), Vp(Vibrio parahaemolyticus) VP1684(ref\\|NP\_798063), Bt BP2265(ref\\|NP\_880908), Ye YscB(ref\\|NP\_052412), Pa PscB(gb\\|AAC44773), Pl SctB(ref\\|NP\_930983), Vp NosA(ref\\|NP\_798076), Dv(Desulfovibrio vulgaris) CesT(ref\\|YP\_011605), Ea ORFA(gb\\|AAF63396), Ct(Chlamydia trachomatis) SycE(ref\\|NP\_219591), Ct CT663(ref\\|NP\_220182), Ye SycH(ref\\|NP\_863547), Vp VP1682(ref\\|NP\_798060), Ye SycN(ref\\|NP\_052398), As(Aeromonas salmonicida) Acr2(emb\\|CAD30215), Pl plu3764(ref\\|NP\_930970), Pa Pcr2(gb\\|AAC45941), Vp VP1665(ref\\|NP\_798044), and Bt BP2258(ref\\|NP\_880901). Alignment is presented using default CHROMA settings (see Figure 2 legend). ::: ![](1471-2180-5-9-9) ::: CesD is a member of the \"chaperones of the translocators\" family common to many T3SSs. We have already shown that this family is characterised by the presence of three tandem tetratricopeptide repeats \[[@B115]\]. CesAB (previously Orf3) is a newly described chaperone that binds to EspA and EspB \[[@B31]\]. Homology searches fail to reveal any homologues of this protein, even in the two other systems that possess obvious EspA homologues. CesD2 (previously Orf27), is a recently described chaperone that, like CesD, binds to EspD \[[@B29]\]. PSI-BLAST searches show that it is homologous to SseE from the Spi-2 system, CseE from the Cpi-2 system of Chromobacterium violaceum, LcrR from the Ysc-Yop system, AcrR from Aeromonas salmonicidaand LssR from Photorhabdus luminescens. None of these homologues has been characterised aside from LcrR. An initial report suggested that a mutation in lcrRhad a regulatory effect \[[@B116]\], although this was subsequently shown to be the result of a polar effect on the downstream gene lcrG\[[@B117]\]. However, an lcrRmutant was detected in a signature-tagged mutagenesis screen of Y. pseudotuberculosisand shown to be attenuated both after oral and intra-peritoneal infection in Balb-C mice, even though the mutation did not have a measurable effect on Yop secretion or function in cell culture assays \[[@B118]\]. These findings stand in contrast to a report that a Y. pseudotuberculosis lcrRmutant had no virulence phenotype after oral infection of Swiss albino mice \[[@B119]\]. The homology between LcrR and CesD2 and between EspD and YopB suggests that LcrR is likely to bind to YopB (Table [1](#T1){ref-type="table"}). Regulators ---------- Ler is a regulatory protein encoded with the LEE1 operon, which has been extensively studied for its effects on gene expression within the LEE-encoded system \[[@B120]-[@B123]\]. It is known to be homologous to the global regulator H-NS, and is likely to adopt a similar fold, including a dimeric structure, particularly as key residues in the N-terminal dimerization domain are conserved in both proteins \[[@B124]-[@B126]\]. H-NS is known to form heterodimers with other proteins from the H-NS family \[[@B124],[@B127],[@B128]\] and to form complexes with members of the Hha family of regulators \[[@B129]\]. As the attaching and effacing pathogens possess two other H-NS-like proteins besides Ler (H-NS and StpA), and possess Hha-like proteins (Hha, YdgT), it is tempting to ask whether Ler could, under any circumstances, form complexes with H-NS or StpA or with the Hha-like proteins (and if not, why not?). Quorum sensing through the AI-2 pathway is known to influence type-III secretion in the LEE-encoded system \[[@B130]-[@B134]\]. In addition, there is some limited evidence to support an impact on type-III secretion of quorum-sensing through the AI-1/acyl-homoserine lactone route \[[@B135]\]. However, this latter route is less plausible, as E. colicannot produce these signalling molecules, although it can probably sense signals produced by other bacteria through the LuxR-like protein SdiA \[[@B136]\]. It is thus surprising that domain searches with the protein sequence from rOrf1 of the LEE identify a PFAM domain (PF06977) entitled \"SdiA-regulated\", which represents a conserved region approximately 100 residues long common to YjiK from E. coli K-12 and several other bacterial outer-membrane proteins, including some Salmonellaproteins that are regulated by SdiA \[[@B137]\] (data not shown). This hints at a link between SdiA-mediated sensing of other bacteria and LEE-mediated type-III secretion, which, twinned with growing evidence from Salmonellaand Vibriofor an influence of SdiA/LuxR on virulence and type-III secretion \[[@B136],[@B138]-[@B141]\], suggests that it might be worth taking a fresh look at the influence of SdiA on LEE-mediated type-III secretion. Exhaustive PSI-BLAST searches suggest that the \"SdiA-regulated domain\" in fact consists of several NHL-like repeats (PF01436), which are likely to fold up into a six-bladed beta-propeller \[[@B142],[@B143]\] (data not shown). GrlR and GlrA (previously known as Orf10/L0044 and Orf11 respectively) are a pair of recently characterised regulatory proteins encoded by adjacent genes within the LEE \[[@B33],[@B144]\]. Similarity has already been reported between GlrA and CaiF, a transcriptional regulator from E. coli and to a Salmonellaprotein \[[@B33]\]. PSI-BLAST searches confirm this and report similarity to several other putative transcriptional regulators (e.g. the R721-plasmid-encoded regulator YheC from E. coli K-12). PSI-BLAST searches with GlrR identify only one homologue outside the LEE -- a protein encoded within the unfinished Salmonella bongorigenome (data not shown). Interestingly, the S. bongori glrRhomologue is located next to a grlAhomologue, suggesting a functional link between the two proteins and adding weight to the finding from yeast two-hybrid studies that GlrA and GlrR interact with each other \[[@B145]\]. The remaining coding sequences ------------------------------ Six coding sequences, which initially appeared to be orphans, have recently been shown to essential for LEE-encoded type-III secretion: Orf2, Orf4, Orf12, Orf15, Orf16 and Orf29 \[[@B33]\]. Thus, something can now be said about all of them based on homology and/or experimental data. PSI-BLAST searches show that Orf2 has homologues in several other systems: ssaEin Spi-2, Cv2595 in Cpi-2 and yscEin the Ysc-Yop system (data not shown). Orf2 should thus be renamed EscE. The ssaEgene was found to be induced after invasion of a murine cultured cell line \[[@B146]\]. Like YscE in Yersinia \[[@B147]\], Orf2/EscE is necessary for type-III secretion through the LEE-encoded T3SS \[[@B33]\]. YscE interacts with TyeA \[[@B148]\] -- thus it would seem likely that Orf2/EscE interacts with the C-terminal portion of SepL (Table [1](#T1){ref-type="table"}). YscE also interacts with YscG, a protein with no detectable homologue in the LEE-encoded system \[[@B149]\]. Orf4, Orf12 and Orf29 possess homologues in T3SSs closely related to the LEE-encoded system, although not in the Ysc-Yop system (Table [1](#T1){ref-type="table"}). None of the Orf4 or Orf29 homologues has been investigated. However, the homologue of Orf12 in the Spi-2 system, SsaM, has recently been characterised in S. enterica\[[@B150]\]. An ssaMdeletion mutant was attenuated in vivo-- in vitroit over-secreted, but failed to translocate, the effector SseJ. Furthermore, it also failed to secrete translocators. In addition, SsaM was shown to interact with another effector SpiC within the bacterial cell. The story told by this complex set of phenotypes, reminiscent of the multi-faceted YopN-like proteins or of the FliK/YscP-like proteins, stands at odds with the simpler global defect in type-III secretion reported in an orf12deletion mutant \[[@B33]\]. As with the YopN-like and FliK/YscP-like proteins, it remains to be seen whether these apparently dissimilar functional properties of homologous proteins from different T3SSs will converge into a unified mechanistic picture or represent genuine evolutionary divergence of function. Given that Orf29 is known to interact with Orf2/EscE \[[@B145]\], it is tempting to speculate that, even in absence of homology, Orf29 and the chief binding partner of YscE, YscG might play similar roles. As noted above, Orf16 is a candidate for needle-length control, while the function and evolutionary origin of Orf15 remain mysterious. Conclusion ========== In this survey, we have discovered several new homologous relationships and woven together threads of evidence from our own in silicosurveys and from published experimental studies to craft numerous novel functional predictions about the LEE-encoded and other T3SSs that can now be tested in the laboratory. We invite the scientific community to test our predictions. Furthermore, in this sequence-rich, post-genomic era \[[@B2],[@B23]\], we believe the time is right to roll out similar bioinformatics analyses over all known type-III secretion systems and invite others to join us in this enterprise on our new web site, 3Base, dedicated to type-III secretion <http://3base.bham.ac.uk>. Methods ======= Sequences --------- Peptide sequences of the 44 LEE-encoded proteins from the genome-sequenced enterohaemorrhagic E. coliO157:H7 Sakai strain (encoded by genes ECs4548-ECs4591 in the Sakai nomenclature) were extracted from the relevant GenBank file (accession number dbj\\|BA000007). Analyses were performed on a Macintosh Dual 2 GHz G5 running OSX10.3. Homology searches ----------------- PSI-BLAST searches \[[@B24]\] were performed on LEE-encoded proteins against a composite database that included the NCBI\'s non-redundant (NR) protein database and a database of predicted protein sequences from unfinished genomes (VGE-PEPT: searchable on our web site <http://www.vge.ac.uk>). To explore all possibilities, multiple PSI-BLAST searches were run for the LEE-encoded proteins using different combinations of substitution matrices (BLOSUM45, BLOSUM62, BLOSUM80), low-complexity filtering (ON, OFF), compositional-based statistics (ON, OFF), E-value inclusion threshold (E = 0.002, E = 0.02) and database (VGE-PEPT + NR, VGE-PEPT alone). PSI-BLAST results were parsed using home-produced Perl scripts and any searches that uncovered significant or near-significant matches to other type III secretion associated proteins were flagged for closer examination and potential homologues retrieved and analysed further. In some cases, PSI-BLAST searches were supplemented with iterative Hmmer searches \[[@B151]\] and searches of the PFAM, CDD and SMART databases \[[@B40],[@B55],[@B152]\]. Most LEE proteins produced statistically significant matches to other Type III associated proteins of known and unknown function under PSI-BLAST default conditions (except for the absence of filtering and compositional based statistics) within 1 to 5 iterations. In some cases (e.g. ECs4584/ORF5), a significant PSI-BLAST hit was dependent on an absence of low-complexity eukaryote sequences in the target database, whereas in others (e.g. ECs4572/rORF8) significant hits were only identified using the BLOSUM80 matrix. Alignments ---------- In most cases alignments were generated from homologues identified in PSI-BLAST searches with T-coffee under default settings \[[@B153]\]. In the case of the EscD alignment, the FHA-like and BON-like regions were aligned separately using T-coffee. The former was then aligned to the FHA domain profile (SM00240) from the SMART database and the latter to the BON domain profile from PFAM (PF04972), respectively, using hmmalign from the Hmmer package. Similarly, the SepQ alignment was generated by aligning the COG1886 and Pfam:SpoA domain profiles with hmmalign. The EscI alignment was generated using T-coffee to align homologues parsed from the PSI-BLAST searches before the FliH sequence from E. coliK-12 was aligned to it using hmmalign. All multiple alignments were manually curated to ensure that secondary structure elements were not broken and to minimise the number of misaligned regions (as assessed using ClustalX and T-coffee \[[@B153],[@B154]\]). Alignments were coloured using default CHROMA settings \[[@B155]\]. Authors\' contributions ======================= MJP performed the initial PSI-BLAST and CDD searches, managed the project and wrote the paper. SAB performed exhaustive PSI-BLAST searches and other bioinformatics analyses, prepared multiple alignments and compiled these into figures. CMB prepared Figure [1](#F1){ref-type="fig"}. Acknowledgements ================ We thank the MRC for funding SB\'s Bioinformatics Fellowship; we thank the Division of Infection and Immunity and the Medical School, University of Birmingham for funding CMB and the 3Base project. We thank the British Biotechnology and Biologivcal Sciences Research Council for funding other experimental and bioinformatics work within the Pallen group on E. coli and type-III secretion (grant references EGA16107, D13414, 02B1D08033). We thank an anonymous reviewer for pointing out the similar genetic locations of sepDand yscB.	PubMed Central	2024-06-05T03:55:55.834190	2005-3-9	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084347/", "journal": "BMC Microbiol. 2005 Mar 9; 5:9", "authors": [ { "first": "Mark J", "last": "Pallen" }, { "first": "Scott A", "last": "Beatson" }, { "first": "Christopher M", "last": "Bailey" } ] }
PMC1084348	Background ========== In diseased skeletal muscle, damaged myofibers can sometimes be replaced or repaired by mononucleate muscle precursor cells that can commit to myogenesis and fuse to form multinucleate myofibers. In some muscle diseases, progressive loss of muscle function may be due at least in part to a decreasing rate of repair that becomes insufficient to replace lost myofibers. The mononucleate cells in skeletal muscle include the relatively abundant muscle satellite cells and their myoblast progeny, as well as infiltrating inflammatory cells and the much less numerous muscle stem cells \[[@B1]-[@B4]\]. In this study, we examine how different subsets of muscle-derived mononucleate cells are affected in the diseased skeletal muscles of laminin-α2-deficient (Lama2^-/-^) and dystrophin-deficient (mdx) mice. We examined these two diseases because both the extracellular laminin-α2 and the intracellular dystrophin proteins interact with the dystroglycan complex of membrane proteins, yet the pathologies of the two diseases are significantly different. Mice lacking laminin-α2 due to mutations in the Lama2gene have greatly shortened lifespan, poor muscle growth, and poor muscle regeneration compared to normal littermates \[[@B5],[@B6]\]. Laminin-α2 is a subunit of the basement membrane component known as merosin (or laminin 2/4) and is highly expressed in skeletal muscle \[[@B7],[@B8]\]. Lama2^-/-^mice are a model for the severely disabling human disease, primary laminin-α2-deficient Congenital Muscular Dystrophy Type 1A \[[@B9],[@B10]\], which is also known as merosin-deficient congenital muscular dystrophy \[[@B11],[@B12]\]. The mdxmice carry a point mutation in the dystrophin gene that leads to loss of functional dystrophin protein. Muscle pathology in 2--4 month old mdxmice is characterized by widespread muscle fiber degeneration and regeneration \[[@B13]\]. Unlike Lama2-null mice, mdxmice have almost normal life spans and their muscles show evidence of continuing, successful regeneration \[[@B14],[@B15]\]. The human disease, Duchenne muscular dystrophy, is also caused by mutations in the dystrophin gene. For this study, we classified mononucleate cells derived from skeletal muscles based on expression of two cell surface proteins: Sca-1 and CD45. Sca-1 ([S]{.underline}tem [c]{.underline}ell [a]{.underline}ntigen 1) was first described as a hematopoietic stem cell marker \[[@B16]\] and subsequently found to be expressed by a rare subset of the mononucleate cells in skeletal muscles \[[@B17]-[@B19]\]. CD45 is a cell surface tyrosine phosphatase found on all nucleated cells of hematopoietic origin and is also found on rare muscle-derived cells \[[@B1]\]. The expression patterns of these two proteins define four subsets of cells. The double-negative subset includes muscle satellite cells and their myoblast progeny \[[@B20]\]. The double-positive subset includes cells that appear to be multi-potential stem cells \[[@B1]\]. The Sca-1^-^/CD45^+^subset includes infiltrating inflammatory cells; and different subsets of this heterogeneous group of cells are found in dystrophin-deficient and laminin-α2-deficient muscles \[[@B21]-[@B27]\]. The Sca-1^+^/CD45^-^subset includes cells with myogenic and endothelial potential \[[@B28]\]. Previous studies suggested that satellite cells and their myoblast progeny have higher than normal proliferation in mdxmuscles \[[@B15],[@B31]\], but proliferation rates in Lama2^-/-^muscles had not been well studied. In addition, it was known that injury induced by cardiotoxin leads to an increase in the CD45^+^cells in skeletal muscles \[[@B3],[@B4]\]. No previous studies, however, had examined how dystrophin-deficiency and laminin-α2-deficiency might alter the proliferation and composition of different subsets of muscle-derived mononucleate cells. In this study, therefore, we examined proliferation rates and compositions of the Sca-1 and CD45 defined subsets of skeletal muscle cells. Our results show that dystrophic muscles (both mdxand Lama2^-/-^) have a similarly increased percentage of CD45^+^cells compared to muscles of wild-type mice. Furthermore, we found that most cell subsets in mdxmuscles in vivohad increased proliferation relative to normal muscles, whereas most cell subsets in Lama2^-/-^muscles had lower than normal proliferation. The contrasting proliferation profiles of the different subsets of mononucleate cells in dystrophin-deficient and laminin-α2-deficient muscles may explain aspects of the distinct pathologies of the two diseases. Methods ======= Mice, breeding, and genotyping ------------------------------ For dystrophin-deficient muscles, we used mice of the C57BL/10ScSn-Dmd^mdx^/J genotype (mdx); and as a normal muscle control for the dystrophin-deficient mice, we used C57BL/10ScSn mice (Jackson Laboratory, Bar Harbor ME). For laminin-α2-deficient muscles, we used mice that carry a targeted LacZ insertion that inactivates the Lama2gene which encodes laminin-α2 \[[@B5],[@B32]\]. This targeted allele is termed dy-W\[[@B5]\]. Heterozygous Lama2^dy-W/+^mice were a gift of Dr. Eva Engvall (Burnham Institute, La Jolla CA). Breeding of these heterozygotes in our laboratory was used to obtain homozygous, laminin-α2-null (Lama2^-/-^) progeny, as well as heterozygous (Lama2^+/-^) and normal (Lama2^+/+^) littermates. Littermates were genotyped by analysis of the Lama2^dy-W^targeted mutation and the wild-type allele by PCR of DNA obtained from tail biopsies as described \[[@B5]\]. Muscles from normal littermates served as controls for Lama2^-/-^muscles. The Lama^2dy-W/dy-W^homozygotes express only very small amounts of a truncated laminin-α2 that lacks domain VI \[[@B6]\], and they have a severe neuromuscular disease in which about most mice die by 6 weeks of age \[[@B32]\]. All animal experiments were reviewed and approved by the Institutional Animal Care and Use Committee at the Boston Biomedical Research Institute. Muscle cell isolation and fluorescence-activated cell sorting ------------------------------------------------------------- To prepare cells for FACS, Lama2^-/-^(n = 3 at 3 weeks old), mdx(n = 2 at 4 weeks old or n = 3 at 7.5 week old), and the appropriate normal control mice (n = 2 for 4 week old controls and n = 3 for 3 week and 7.5 week old controls) were injected intra-peritoneally with 100 μl of 5-bromo-2\'-deoxyuridine (BrdU) (BrdU Flow Kit, BD Biosciences) solution (1 mg/ml in PBS) at 24 hours and again at 16 hours before sacrifice. Limb muscles were dissected, cleaned, and thoroughly minced. The tissue was digested in 0.2% Pronase (Calbiochem) in HBSS at 37°C for one hour and successively passed through nylon mesh filters with cut-offs of 100 μm, 40 μm, and 10 μm. After the final filtering, cells were re-suspended in 3% heat-inactivated fetal bovine serum in PBS. Biotinylated Sca-1 antibody (E13-161.7) and phycoerythrin (PE)-conjugated CD45 antibody (clone 30-F11) (both from Pharmingen, San Diego CA) were added to the cells at a concentration of 2 μg/ml, and the cells were incubated on ice for one hour. Cells were washed in PBS and incubated with the secondary detecting reagent, allophycocyanin (APC)-streptavidin, at 1 μg/ml for 20 minutes on ice. After three additional washes in PBS, the cells were fixed for 30 minutes with a DPBS buffer containing 4% paraformaldehyde and the detergent saponin, then washed and stored overnight in staining buffer (from the BrdU flow kit) supplemented with 3% fetal bovine serum. The following day, cells were refixed as above, treated with 300 μg /ml DNase for one hour at 37°C, incubated for 20 minutes with fluorescein isothiocyanate (FITC) conjugated anti-BrdU antibody at a 1:50 dilution of the stock solution provided with the kit, and stained with 20 μl of a DNA binding dye, 7-amino-actinomycinD (7-AAD) (BrdU flow kit) for 1 hour. Separate aliquots of cells were stained with appropriate isotype control antibodies and used as negative controls. Flow cytometric analysis was performed using a dual-laser cell sorter (FACSCalibur, BD Biosciences) to determine the distribution and proliferation dynamics of the muscle-derived cells. Cell Quest Pro software (BD Biosciences) was used for data acquisition and analysis. Debris was first gated out by analyzing particle morphology based on side and forward scatter. In vivoproliferation was determined by analyzing the percent of all the cells that incorporated BrdU. Also, we determined the percent of cells within each sub-population that were BrdU positive. Statistical analyses -------------------- All data are expressed as mean ± standard deviation. Comparison among groups was done by analysis of variance (ANOVA) and individual differences between pairs was determined by unpaired two-tailed t-test performed with the Instat program (v2.03, GraphPad, San Diego CA). Statistical analyses were done either using raw data (prior to conversion to percentages) or using arcsin-transformed percentage data \[[@B33]\]. Differences among the disease groups or between diseased and normal controls was considered significant when p ≤ 0.05. Results ======= We used FACS analysis to identify subsets of the mononucleate cells that reside in skeletal muscle and then determined how the distribution of cells across the subsets was affected by disease. We identified the cell subsets based on expression or lack of expression of two cell surface markers: CD45 and Sca-1. Thus, we partitioned the muscle-derived cells into four subsets: Sca-1^+^/CD45^+^, Sca-1^+^/CD45^-^, Sca-1^-^/Cd45^+^and Sca-1^-^/CD45^-^. Cells for sorting were obtained from muscles of mdxmice and their 10ScSn normal controls at 4 weeks and 7.5 weeks after birth and from muscles of Lama2^-/-^and normal littermates at 3 weeks after birth. The respective myopathies are well in progress at these times \[[@B31],[@B32]\]. It was not possible to analyze older Lama2^-/-^mice for Sca-1 and CD45 expression, as most died by 5--6 weeks of age and the small muscles did not consistently provide sufficient cells for the necessary FACS analyses. We found that cells from muscles of the two types of normal control mice that we used (C57Bl/10ScSn at 4 and 7.5 weeks old for mdxand C57BL/6 at 3 weeks old for Lama 2^-/-^) had similar distributions of cells across the four cell subsets (Figs. [1](#F1){ref-type="fig"} and [2](#F2){ref-type="fig"}; Table [1](#T1){ref-type="table"}). In all healthy muscles, the majority (\>75% of the total) of mononucleate cells were negative for both Sca-1 and CD45. As noted \[[@B20]\], satellite cells and their myoblast progeny form multinucleate myotubes in culture, are Sca-1^-^/CD45^-^, and thus are included in this double-negative subset. We confirmed this result, because we found that (i) the double-negative cells at the time of isolation also expressed CD34, as expected for quiescent satellite cells \[[@B2]\]; (ii) \>98% of the double-negative cells expressed desmin, an intermediate filament protein found in proliferating satellite cells, after 3 days in culture; and (iii) the double-negative cells were able to form myotubes in culture when switched to low serum medium (not shown). The rarest mononucleate cells (\<1.5% of the total) in healthy muscles were those that expressed both Sca-1 and CD45. The remaining two cell subsets were more abundant than the double-positive cells, but still much less abundant than the double-negative cells. Cells that expressed Sca-1 but not CD45 amounted to 2.5 -- 12% of the total in healthy muscles; and cells that expressed CD45 but not Sca-1 amounted to 4.7 -- 8.3% of the total. The two Sca-1^+^subsets were more abundant in the 3 and 4 week old than in the 7.5 week old healthy muscles. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### The composition of mononucleate cells was altered in Lama2^-/-^and mdxdiseased muscles compared to healthy muscles. Mononucleate cells were prepared from muscles of Lama2^-/-^(3 week old), mdx(7.5 weeks old), and age- and strain-matched control mice and analyzed by FACS for expression of Sca-1 and CD45. A. Scatter-plots of a representative FACS study of the different cell subsets defined by Sca-1 and CD45 expression patterns in normal C57BL/10ScSn (left) and mdx(right) muscles. B. Scatter-plots of a representative FACS study of the different cell subsets in normal C57BL/6J (right) and Lama2^-/-^(left) muscles. Significant differences were observed in the distribution of cells between the wild type and disease conditions (see Fig. 2 and Table 1 for quantitative and statistical analyses). As also noted in the text, mdxmuscles contained several-fold more mononucleate cells per gram of tissue than Lama2^-/-^muscles. ::: ![](1471-2377-5-7-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### The composition of mononucleate cells was altered in diseased mdxand Lama2^-/-^compared to control muscles. As in Fig. 1, mononucleate cells were prepared from muscles of Lama2^-/-^(3 week old), mdx(4 weeks or 7.5 weeks old), and appropriate age- and strain-matched control mice and analyzed by FACS for expression of Sca-1 and CD45. A. Quantitative analysis of the cell subsets defined by Sca-1 and CD45 expression in mdxand normal control muscles showed that the percentage of cells in the two CD45^+^subsets was signficantly higher in mdxmuscles at both 4 weeks and 7.5 weeks of age. B. Similarly, in Lama2^-/-^muscles, the percentage of cells in the two CD45^+^subsets was also significantly higher than in normal healthy muscle. See text for statistical analyses. As also noted in the text, mdxmuscles contained several-fold more mononucleate cells per gram of tissue than Lama2^-/-^muscles. Error bars represent standard deviation with n = 3. \* indicates that diseased and healthy control samples are significantly different at p \< 0.05 (see Table 1 for quantitative and statistical analyses). ::: ![](1471-2377-5-7-2) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Patterns of expression of Sca-1 and CD45 by mononucleate cells derived from healthy and diseased skeletal muscles. ::: Percentage of cells with indicated expression pattern^1^ ----------- -------------- ---------------------------------------------------------- ------------------ ------------------ ------------------ Age Genotype Sca-1^+^/CD45^-^ Sca-1^+^/CD45^+^ Sca-1^-^/CD45^+^ Sca-1^-^/CD45^-^ 3 weeks C57BL/6 5.9 ± 3.5 1.0 ± 0.4 4.7 ± 1.7 88.0 ± 5.3 Lama2^-/-^ 5.6 ± 1.8 3.3 ± 0.4 38.9 ± 15.9 52.2 ± 15.6 P \> 0.1 P \< 0.03\* P \< 0.01\* P \< 0.01\* 4 weeks C57BL/10 12.0 ± 6.7 1.2 ± 0.2 8.3 ± 2.2 78.6 ± 4.3 mdx 21.0 ± 7.4 6.4 ± 2.4 55.8 ± 11.3 16.7 ± 6.2 P \> 0.1 P = 0.09 P \< 0.03\* P \< 0.01\* 7.5 weeks C57BL/10 2.5 ± 0.6 0.2 ± 0.1 5.1 ± 0.2 92.0 ± 0.6 mdx 1.8 ± 0.5 0.8 ± 0.3 28.9 ± 3.6 68.5 ± 4.1 P \> 0.1 P = 0.03\* P \< 0.01\* P \< 0.01\* ^1^Data are presented as ave. ± SD with n = 3 for all except 4 week old values where n = 2. Unpaired, two-tailed t-tests (see Methods) were used to compare values from each pair of healthy and diseased muscles that were of the same age and expression pattern. P values are shown below the pair of tested values. P ≤ 0.05 was considered significant and indicated by an asterisk. ::: We next found that the distribution profiles of cells from dystrophin-deficient and laminin-α2-deficient muscles differed from those of cells from normal muscles, with the diseased muscles showing a marked increase in the CD45^+^cell subsets (Figs. [1](#F1){ref-type="fig"} and [2](#F2){ref-type="fig"}, Table [1](#T1){ref-type="table"}). The changes in the percentage of cells in each subset were similar for cells derived from Lama2^-/-^and mdxmuscles. In particular, the rarest subset of double-positive CD45^+^/Sca-1^+^cells composed a significantly higher percentage of the mononucleate cells in diseased muscles (Figs. [1](#F1){ref-type="fig"} and [2](#F2){ref-type="fig"}; Table [1](#T1){ref-type="table"}). The percentage of double-positive cells was increased 3--5X in diseased compared to healthy muscles. As one example, the double-positive cells amounted to 3.3 ± 0.4% of the total cells from diseased Lama2^-/-^muscles, but only 1 ± 0.4% of the cells from healthy C57BL/6 muscles (n = 3, P \< 0.03). The subset of CD45^+^cells that did not express Sca-1, which was moderately abundant in healthy muscle and includes inflammatory cells, also accounted for a higher percentage of the total mononucleate cells in diseased muscles than in normal muscles (Figs. [1](#F1){ref-type="fig"} and [2](#F2){ref-type="fig"}; Table [1](#T1){ref-type="table"}). In particular, the Sca-1^-^/CD45^+^cells accounted for a 5 -- 8X higher percentage of the mononucleate cells in diseased than in healthy muscles. As one example, the 7.5 week old mdxmuscles had a 5.7-fold increase in Sca-1^-^/CD45^+^cells compared to healthy control muscle (28.9 ± 3.6% of the cells in mdxsamples vs. 5.1 ± 0.2% in control muscles; n = 3, P \< 0.01). These increased percentages reflect the increased numbers of inflammatory cells that are known to occur in the skeletal muscles of these disease models \[[@B21]-[@B27]\]. These Sca-1^-^/CD45^+^inflammatory cells are known to be heterogeneous (e. g., macrophages, eosinophils, neutrophils, B-cells, T-cells; all of which are Sca-1^-^/CD45^+^); and the subsets of inflammatory cells that are present differ between dystrophin-deficient and laminin-α2-deficient muscles \[[@B21]-[@B27],[@B29],[@B30]\]. Because these different subsets of inflammatory cells have been well-characterized in dystrophin-deficient and laminin-α2-deficient muscles \[[@B21]-[@B27]\], we did not re-examine this heterogeneity for this study. The increased percentage of CD45^+^cells in diseased muscles was accompanied by a significantly decreased percentage of cells in the most abundant double-negative subset, whereas the percentage of cells in the Sca-1^+^/CD45^-^subset was unchanged by disease (Figs. [1](#F1){ref-type="fig"} and [2](#F2){ref-type="fig"}, Table [1](#T1){ref-type="table"}). Thus, in both dystrophin-deficient and laminin-α2-deficient diseased muscles, the double negative cells, which included satellite cells and their myoblast progeny, became relatively less abundant, whereas cells expressing the hematopoietic marker CD45 became relatively more abundant. As in previous studies, however, the absolute number of Sca-1^-^/CD45^-^cells (muscle precursor cells) obtained from a given amount of mdxmuscle was considerably larger than the number obtained from the same amount of control muscle, whereas fewer muscle precursor cells were obtained from Lama2^-/-^than control muscles. For example, in one experiment using muscles from 5--6 week old mdxand control mice, healthy control muscles yielded 1.3 ± 0.4 × 10^5^double-negative cells per gram (the number of cells that can be obtained decreases with increasing age), whereas mdxmuscles yielded many more at 4.8 ± 3.0 × 10^5^double-negative cells per gram (ave. ± SD, n = 4). In contrast, in an experiment using muscles from 3 week old Lama2^-/-^and control mice, we obtained 2.6 ± 0.6 × 10^5^double-negative cells per gram of control muscle, but only 1.6 ± 0.3 × 10^5^double-negative cells per gram of Lama2^-/-^muscle (ave. ± SD, n = 3). Thus, the similar increases in the percentage of CD45^+^cells in the two diseases occurred despite the different directions of change in total cell numbers. Though both mdxand Lama2^-/-^muscles showed similar increases in CD45^+^cells, the two diseases showed opposite changes in mononucleate cell proliferation. In muscles of the two control strains (C57Bl/6J, the control for Lama2^-/-^muscles; and C57BL/10ScSn, the control for mdxmuscles), there was a similar several-fold decrease in BrdU incorporation from 3--4 weeks to 7.5 weeks after birth, which is consistent with the slowing of muscle growth found during this postweaning period (Fig. [3A](#F3){ref-type="fig"}). At both 4 weeks and 7.5 weeks after birth, BrdU incorporation was significantly higher in mdxthan in control muscles of the same age (Fig. [3B](#F3){ref-type="fig"}). In mdxmuscles at 7.5 weeks after birth, for example, the percentage of all mononucleate cells that incorporated BrdU during 24 h in vivowas 6.9 ± 2.6%, which was significantly higher than 1.9 ± 1.1% that incorporated BrdU in the 10ScSn control muscles of the same age (n = 3, p \< 0.05) (Fig. [3B](#F3){ref-type="fig"}). Consistent with this general increase of proliferation in mdxmuscles, three of the four Sca-1/CD45 subsets of muscle-derived mdxcells showed increased BrdU incorporation, indicating a greater number of proliferating cells compared to normal controls. The exception was the Sca-1^-^/CD45^+^cell subset which had decreased BrdU incorporation in mdxcompared to normal muscles (Fig. [4A](#F4){ref-type="fig"}). ::: {#F3 .fig} Figure 3 ::: {.caption} ###### The percentage of cells that incorporated BrdU decreased with increased age and was lower in Lama2^-/-^muscles, but higher in mdxmuscles, than in normal control muscles. A. The percentage of the total population of mononucleate muscle cells that incorporated BrdU was decreased as postnatal development proceeded. In muscles of two control strains (C57Bl/6J, the control for Lama2^-/-^muscles; and C57BL/10ScSn, the control for mdxmuscles), there was a similar several-fold decrease in BrdU incorporation from 3--4 weeks to 7.5 weeks after birth, which is consistent with the slowing of muscle growth found during this period. B. The percentage of all cells that incorporated BrdU was significantly lower in 3 week old Lama2^-/-^muscles than in normal control muscles of the same age. In contrast, the percentage of all that incorporated BrdU was signifcantly higher in 4 week old and 7.5 week old mdxmuscles than in normal control muscles of the same age. Error bars represent standard deviation with n = 3. \* indicates diseased and healthy control samples differ significantly at P \< 0.05. ::: ![](1471-2377-5-7-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### In most cell subsets, the percentage of cells that incorporated BrdU was lower in Lama2^-/-^muscles, but higher in mdxmuscles, than in healthy control muscles. A. In all cell subsets except Sca-1^-^/CD45^+^, the percentage of cells that incorporated BrdU during 24 h in vivowas significantly increased in mdxcells compared to normal cells (see text for statistics). B.The percentage of cells that incorporated BrdU in Lama2^-/-^muscles was significantly decreased compared to normal muscles in all subsets of Lama2^-/-^cells with the exception of the rare double-positive Sca-1^+^/CD45^+^subset in which incoporation was similar in diseased and healthy muscles (see text for statistics). Error bars represent standard deviation with n = 3. \* indicates that the value for the mdxor Lama2^-/-^muscles differed significantly from the corresponding control at P \< 0.05. ::: ![](1471-2377-5-7-4) ::: In contrast, at 3 weeks after birth, a much lower percentage of cells incorporated BrdU in Lama2^-/-^muscle than in normal muscle (Fig. [3B](#F3){ref-type="fig"}) (6.8 ± 1.7% for Lama2^-/-^vs. 23.5 ± 10.0% for normal. n = 3, p \< 0.05). Consistent with this general decrease in proliferation, three of four Sca-1/CD45 subsets of muscle-derived Lama2^-/-^cells showed decreased BrdU incorporation, indicating fewer proliferating cells compared to normal controls (Fig. [4B](#F4){ref-type="fig"}). The exception in this case was the double-positive Sca-1^+^/CD45^+^cell subset which had approximately the same BrdU incorporation in Lama2^-/-^muscles as in normal muscles (Fig. [4B](#F4){ref-type="fig"}). Discussion ========== We found that the composition and proliferation dynamics of mononucleate cells were significantly different in the diseased muscles of mdxand Lama2^-/-^mice than in normal healthy muscles. Dystrophin-deficient mdxmuscles and laminin-α2-deficient Lama2^-/-^muscles showed similar changes in cell compositions, with CD45^+^cells accounting for a significantly higher percentage of the total in diseased than in healthy muscles. In contrast, cells in mdxand Lama2^-/-^muscles showed opposite changes in proliferation kinetics, with most cell subsets showing increased BrdU incorporation in mdxmuscles, but decreased incorporation in Lama2^-/-^muscles. With the recognition that adult skeletal muscles contain rare cells with stem cell properties, a number of markers, including Sca-1, CD34, CD45, Bcl-2, and c-kit, have been used alone or in combination to identify and purify the cells \[[@B1],[@B18],[@B19],[@B28],[@B34]-[@B37]\]. In our study, we subdivided muscle-derived stem cells on the basis of expression or lack of expression of Sca-1, which is commonly used to purify hematopoietic stem cells, and CD45, which is a pan-hematopoietic cell marker. Neither Sca-1 nor CD45 is expressed by muscle satellite cells and their myoblast progeny (muscle precursor cells) \[[@B20]\], whereas muscle-derived cells with hematopoietic potential are almost all CD45^+^\[[@B1],[@B20]\]. Though the distribution of cells among the four cell subsets did not significantly change during early postnatal (3 to 7.5 weeks) development in normal muscles (compare controls in Figs. [2A](#F2){ref-type="fig"} and [2B](#F2){ref-type="fig"}), there was a significant shift in the distribution profile in response to muscle disease. Specifically, diseased muscles had a large increase in the percentage of cells in both CD45^+^subsets, i. e., Sca-1^+^/CD45^+^and Sca-1-/ CD45^+^(Fig. [2](#F2){ref-type="fig"}, Table [1](#T1){ref-type="table"}). The double-positive Sca-1^+^/CD45^+^cells were likely putative stem cells with hematopoietic potential \[[@B1]\], whereas the Sca-1^-^/CD45^+^cells included the heterogeneous group of inflammatory cells. The increase in CD45^+^cells was accompanied by a decrease in the double-negative Sca-1^-^/CD45^-^cells, which included satellite cells and their myoblast progeny. Though the double-negative cells were usually the most abundant subset, we found that the Sca-1^-^/CD45^+^inflammatory cells became the most abundant subset in younger four week old mdxmuscles, which is consistent with the rapid myofiber degeneration at this age. A previous study showed that double-positive Sca-1^+^/CD45^+^cells also increase during regeneration of injured muscle and that a small proportion of the double-positive cells from injured muscle, unlike those from uninjured muscles, give rise to myogenic cells in culture \[[@B3],[@B4]\]. Diseased muscles tend to undergo extensive degeneration followed by attempted regeneration. At least initially, regeneration is largely successful in mdxmuscles but abortive in Lama2^-/-^muscles \[[@B5],[@B15],[@B31]\]. Because the double-positive Sca-1^+^/CD45^+^cells increased in both types of diseased muscle (this study), it is likely that, as in injured muscles \[[@B3],[@B4]\], these cells play a role in attempted muscle regeneration. In contrast to the similar increases in CD45^+^cells in both types of diseased muscles, the mononucleate cells in mdxand Lama-2^-/-^muscles showed opposite changes in proliferation in vivo, with proliferation increased in mdxmuscles but decreased in Lama-2^-/-^muscles compared to healthy control muscles. Total cell proliferation in mdxmuscles was known to be elevated relative to healthy controls \[[@B31]\]. Here we examined proliferation in subsets of the muscle mononucleate cells and found that, consistent with the increase in overall proliferation, a higher than normal percentage of cells in three of the four mdxsubsets incorporated BrdU. Both the double-negative cells, which included satellite cells, and the double-positive cells with hematopoietic potential, showed increased proliferation. Only cells of the Sca-1^-^/CD45^+^subset failed to show increased BrdU incorporation in mdxmuscles; and these were likely to be mostly inflammatory cells (e. g., macrophages, eosinophils, neutrophils, T-cells; all of which are Sca-1^-^/CD45^+^) which do not have hematopoietic stem cell potential but are known to migrate into and thus accumulate in diseased muscles \[[@B1],[@B21]-[@B27]\]. The increased percentage of proliferating mdxcells in vivois likely due to the altered environment in mdxmuscles which includes increased levels of growth factors; increased extracellular matrix proteins such as collagen, fibronectin, and laminin which can promote myoblast proliferation; and perhaps decreased levels of growth inhibitors \[[@B38]-[@B42]\]. Whether cells in different subsets respond to the same or different growth regulators remains to be determined. In Lama2^-/-^muscles, overall BrdU incorporation was significantly decreased compared to that in healthy control muscles. BrdU incorporation was decreased in every Lama2^-/-^cell subset except the rare double-positive Sca-1^+^/CD45^+^subset that includes cells with hematopoietic potential \[[@B1]\]. The decrease could have arisen if, relative to normal cells, the Lama2^-/-^cells had longer cell cycles or lower probability of initiating replication from G~0~, if there was more cell death among dividing than non-dividing Lama2^-/-^cells \[[@B43]\], or if some combination of these occurred. The relative lack of Lama2^-/-^muscle cell proliferation could in part explain why regeneration is relatively unsuccessful in laminin-α2-deficient muscles \[[@B5]\], though other mechanisms such as increased apoptosis of both myoblasts and newly formed myofibers are also likely to be important. Laminin-α2 is expressed in skeletal muscle myoblasts and myotubes \[[@B44]\] where it promotes the survival of satellite cells in vivoand in vitroas well as myoblast fusion and myotube formation \[[@B5],[@B45],[@B46]\]. We found that the most abundant, double-negative subset of muscle cells, which includes muscle precursor cells, showed significantly less BrdU incorporation in Lama2^-/-^muscles. It is likely, therefore, that laminin-α2 not only promotes survival of myotubes as shown previously \[[@B45],[@B46]\] but also influences division of muscle precursor cells as early as three weeks after birth, which is a time when extensive myofiber growth occurs in normal muscles. A recent study has demonstrated that laminin-2 (α2, β1, γ1) improves proliferation of an epithelial cell line through an integrin/ERK pathway \[[@B47]\]. A similar mechanism could occur in skeletal muscle cells, because merosin, of which laminin-α2 is a component, influences the level of expression and localization of α7β 1D integrin at the sarcolemma \[[@B46]\] and a lack of laminin-α2 could, therefore, disrupt merosin/integrin-mediated signals that regulate proliferation. Pathogenesis follows distinct courses in Lama2^-/-^and mdxmice, as well as in humans with corresponding diseases \[[@B48]\]. Nonetheless, CD45^+^cells are similarly increased in both mouse diseases. The role of the double-positive Sca-1^+^/CD45^+^cells in muscles needs to be defined, perhaps by tracing fates of their progeny; and the likely heterogeneity within the cell subsets we studied should be explored. For example, one study with a specific mAb suggested that satellite cells are \<15% of the mononucleate cells in healthy, post-weaning skeletal muscles \[[@B49]\], suggesting that there might be heterogeneity within the double-negative Sca-1^-^/CD45^-^cell subset which we and others \[[@B3]\] found to contain more than half of the mononucleate cells and to contain satellite cells. Some differences between studies might be expected because assessment of cell type percentages likely depends on the particular enzymes used for tissue dissociation, the antibodies and staining protocols used, and the selection of fluorescent cut-off limits for cells sorting. Despite these caveats, the general agreement between our study of healthy vs. diseased muscles and an earlier study \[[@B3]\] of healthy vs. cardiotoxin-injured muscles provides confidence in the conclusion that the CD45^+^cell subsets become relatively more abundant in both injured muscles and in the two types of diseased muscles analyzed here. Further experiments in which particular cell subsets have their fates determined by lineage tracking or are experimentally increased or decreased, perhaps by specific growth factors or conditional ablation, could determine how the changes in cell subsets that we observed here may affect pathogenesis. The Sca-1^+^/CD45^-^cell subset, for example, has been reported to contain cells with myogenic and endothelial lineage potentials \[[@B28]\]. Inhibition of apoptosis ameliorates myopathology and produces a several-fold increase in lifespan of Lama2-nullmice \[[@B50]\], and could affect mononuclear cell subsets. That not only Sca-1^-^/CD45^-^cells (which included satellite cells), but also other cell subsets, showed opposite changes of proliferation in Lama2^-/-^and mdxmuscle, suggests that growth requirements may be shared among the different subsets of cells. In addition, the relatively poor proliferation of Lama2^-/-^cells may be one of the mechanisms underlying the lack of successful regeneration in Lama2^-/-^muscles. The comparative significance to Lama2^-/-^pathogenesis of poor proliferation vs. other mechanisms such as increased apoptosis remains to be defined. Conclusion ========== The similar increases in CD45^+^cells, but opposite changes in proliferation of muscle precursor cells, may underlie aspects of the distinct pathologies in dystrophin-deficiency and laminin-α2-deficiency. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= MG and CAK carried out all of the experimental studies. MG and JBM conceived the study, carried out statistical analyses and drafted the manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2377/5/7/prepub> Acknowledgements ================ We thank Dr. Eva Engvall for supplying Lama2^+/dy-W^mice. This work was supported by grants to J.B.M. from the Muscular Dystrophy Association, the National Institutes of Health (AR49496, ES11384, HL6464), and the United States Department of Agriculture.	PubMed Central	2024-06-05T03:55:55.838873	2005-4-7	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084348/", "journal": "BMC Neurol. 2005 Apr 7; 5:7", "authors": [ { "first": "Mahasweta", "last": "Girgenrath" }, { "first": "Christine A", "last": "Kostek" }, { "first": "Jeffrey Boone", "last": "Miller" } ] }
PMC1084349	Background ========== Sweet and umami taste involve TAS1R receptors, which belong to the subclass 3 of the GPCR super family and are distantly related to the V2R pheromone-, metabotropic glutamate- and calcium sensing receptors \[[@B1]-[@B3]\]. They have a large N-terminal extracellular domain that is linked by a cysteine-rich domain to the seven transmembrane domains \[[@B2]\]. Heterologous expression and functional analysis showed that cells co-transfected with Tas1R2-TAS1R3responded to a variety of natural and artificial sweeteners, while the combination TAS1R1-TAS1R3responded to L-amino acids \[[@B3]-[@B5]\]. In line with observed differences in the perception of these compounds between humans and rodents \[[@B6]-[@B8]\], the artificial sweeteners aspartame and cyclamate, and the sweet proteins brazzein, monellin, and thaumatin only activate the human but not the rodent TAS1R2-TAS1R3 heteromer \[[@B2],[@B3]\]. Further studies of mouse-human interspecies chimeras and point-mutated receptors revealed that cyclamate interacts with the part of hTAS1R3 encompassing the seven transmembrane regions, whereas aspartame interacts with the N-terminal domain of hTAS1R2 \[[@B9]\]. Moreover, the cysteine-rich motif in the N-terminal domain of hTAS1R3 is required for the activation of the sweet receptor by brazzein \[[@B2]\]. These results and modeling studies on the basis of the solved X-ray structure of the mGLUR1 ligand binding domain strongly suggest that multiple sites in both receptor subunits are involved in the activation of the sweet receptor by its ligands \[[@B9]-[@B12]\]. Interestingly, differences in taste perception are not limited to sweet compounds but also have been observed for 2-(4-methoxphenoxy) propanoic acid sodium salt (lactisole), which suppresses the sweet taste of various compounds in humans but not in rats \[[@B3],[@B13]-[@B17]\]. Functional studies of the human and rat sweet taste receptor recently showed that lactisole inhibits the human TAS1R2-TAS1R3 heteromer but not that of rats \[[@B3]\]. Moreover it could be shown that that the lactisole sensitivity of the human sweet taste receptor involves the TAS1R3 subunit \[[@B9]\]. Thus, the inhibition of the sweet response by lactisole in humans is due to structural differences between the rat and human sweet taste receptor. Using mutational analysis we investigated the molecular basis for these differences. Results and discussion ====================== Application of the artificial sweetener saccharin to HEK293T-Gα~16gust44~cells transfected with the cDNAs for the rat sweet receptor rTas1r3-rTas1r2 or its human counterpart hTAS1R3-hTAS1R2 elevated intracellular calcium levels (Fig. [1a](#F1){ref-type="fig"}). The responses were dose-dependent with half maximal effective concentrations (EC~50~) of 0.17 ± 0.04 mM for the human heteromer and of 1.28 ± 0.16 mM for the rat receptor heteromer. Thus, the rodent receptor is \~7 fold less sensitive to saccharin than the human receptor. This shows subtle functional differences between the human and the rat sweet receptor. In addition to saccharin also other sweet tasting compounds of various chemical classes such as the disacchride sucrose, the dipeptide derivate aspartame, the sulfamate cyclamate, the dihydrochalcone neohesperidine dihydrochalcone, and the terpenoid glycoside stevioside the activated the human receptor (Fig. [1b](#F1){ref-type="fig"}). Like its human counterpart, the rat sweet taste receptor could be robustly activated by saccharin, sucrose, and stevioside (Fig. [1c](#F1){ref-type="fig"}). But in contrast to the human receptor, aspartame, cyclamate, and neohesperidine dihydrochalcone did not activate the rat receptor (Fig. [1c](#F1){ref-type="fig"}) which reflects species specific variations in sweet taste perception \[[@B7],[@B8]\]. Lactisole inhibited the responses of cells transfected with the cDNAs for the human receptor to all tested sweeteners (Fig. [1a](#F1){ref-type="fig"}, [1b](#F1){ref-type="fig"}) while it did not inhibit the responses of cells transfected with the rat sweet receptor (Fig. [1a](#F1){ref-type="fig"}, [1c](#F1){ref-type="fig"}). This is well in line with previous studies of the receptor function \[[@B3],[@B9]\]. Based on the inhibitory effect of lactisole on the human but not on the rat receptor heteromer, we next asked which amino acids cause the functional difference of the receptors. We therefore analyzed cells co-transfected with cDNAs of the interspecies sweet taste receptor combinations, i.e. hTAS1R3-rTas1r2or rTas1r3-hTAS1R2. The combination hTAS1R2-rTas1r3 responded to aspartame, saccharin, sucrose, and stevioside but not to cyclamate and neohesperidine dihydrochalcone (Fig. [1d](#F1){ref-type="fig"}). The responses induced by all four agonists were lactisole insensitive (Fig. [1d](#F1){ref-type="fig"}). The combination rTas1r2-hTAS1R3 clearly responded to the sweeteners cyclamate and neohesperidine dihydrochalcone, although with markedly reduced amplitudes. Aspartame, saccharin, sucrose, and stevioside failed to activate this receptor combination (Fig. [1e](#F1){ref-type="fig"}). This indicates that the hTAS1R3 subunit is involved in the response of the human sweet taste receptor to cyclamate and neohesperidine dihydrochalcone. Notably, the responses were lactisole sensitive (Fig. [1e](#F1){ref-type="fig"}). In line with previous observations \[[@B9]\] we therefore conclude that the lactisole sensitivity is solely mediated by the hTAS1R3 subunit. Next we wanted to know which part of hTAS1R3 mediates the sensitivity to lactisole. To address this point we constructed two chimeric receptors. In the receptor chimera tas1r3-rN-hTM we fused the N-terminal domain of rTas1r3 to the C-terminal part containing the transmembrane regions of hTAS1R3. In the complementary chimera tas1r3-hN-rTM the human N-terminal domain was fused to the segment of the rat subunit encompassing the transmembrane regions. Cells co-transfected with plasmids encoding hTAS1R2 and tas1r3-hN-rTM responded to aspartame, saccharin, sucrose, and stevioside but not to cyclamate and neohesperidine dihydrochalcone (Fig. [2a](#F2){ref-type="fig"}). The responses to all four agonists were lactisole insensitive (Fig. [2a](#F2){ref-type="fig"}). Cells co-transfected with plasmids encoding hTAS1R2 and tas1r3-rN-hTM responded to all sweeteners tested (Fig. [2b](#F2){ref-type="fig"}). The responses to all six agonists were lactisole sensitive (Fig. [2b](#F2){ref-type="fig"}). Based on these findings we conclude that the activation the human sweet receptor by cyclamate and neohesperidine dihydrochalcone involves the hTAS1R3 transmembrane regions. Moreover, the sensitivity towards lactisole is determined by the hTAS1R3 transmembrane regions. Sequence comparisons (Fig. [2c](#F2){ref-type="fig"}) between hTAS1R3 and its rat and mouse orthologs revealed 75 amino acid differences in the transmembrane regions. As mice and rats are lactisole insensitive while humans are lactisole sensitive, we reasoned that the amino acids responsible for this species difference are conserved in the rodent receptors but variable between rodents and humans. We therefore excluded 27 of the 75 amino acids because they vary between rat and mouse. Due to the polar nature of lactisole we assumed that it is membrane impermeable and should therefore interact with amino acids that are accessible from the extracellular site. Based on this assumption we identified 17 amino acids (Fig. [2c](#F2){ref-type="fig"}, asterisks) as the most promising candidates. We assumed that the conversion of one or several of these amino acids from the rodent to the human form should confer the lactisole sensitivity to the rat receptor. To test this hypothesis we created seven rTas1r3 variants (m1 -- m7) that contained 1--4 amino acids of hTAS1R3 (Fig. [2c](#F2){ref-type="fig"}, boxes). The functional analysis of these receptor variants revealed that receptor mutant m6 confered lactisole sensitivity to the heteromer, while all other rTas1r3mutants did not (Fig. [3a](#F3){ref-type="fig"}). Thus, the 3 amino acids exchanges L735F, V738A, and I740A of mutant m6 are sufficient to confer lactisole sensitivity to rTas1r3. To test whether all exchanges in m6 are equally important, we created three variants, m6/1, m6/2 and m6/3 (Fig. [2c](#F2){ref-type="fig"}, boxes). In each of these, we reconverted one of the three human amino acids present in the mutant m6 back into that of the rat receptor. In case the amino acid is important for the inhibition of the sweet receptor, the variant should loose its lactisole sensitivity. Only mutant m6/2, L735F and I740A, was lactisole insensitive, while the two other variants m6/1, V738A and I740A, and m6/3, L735F and V738A remained lactisole sensitive (Fig. [3b](#F3){ref-type="fig"}). Thus, we concluded that the alanine to valine exchange in position 738 in transmembrane segment 5 is crucial for the lactisole sensitivity. To confirm this, we created a rTas1r3 variant where we only substituted the valine at amino acid position 738 by an alanine. Functional analysis of this receptor variant, rTas1r3-V738A, co-transfected with rTas1r2 revealed that it was activated by saccharin, sucrose, and stevioside, but not by aspartame, cyclamate and neohesperidine dihydrochalcone (Fig. [3c](#F3){ref-type="fig"}). Thus, rTas1r3-V738A shows the same agonist profile as the wild type rat receptor (Fig. [3c](#F3){ref-type="fig"}, [1b](#F1){ref-type="fig"}). However, in marked contrast to the rat wild type receptor, the response of rTas1r3-V738A was always lactisole sensitive (Fig. [3c](#F3){ref-type="fig"}). To test if rTas1r3-V738A displays the same degree of lactisole sensitivity as the human receptor, we determined the half maximal inhibitory concentration (IC~50~) of lactisole on rTas1r3-V738A and compared it with the IC~50~values of the human receptor and the mutants m6, m6/1, m6/2, and m6/3. Although all variants except the lactisole insensitive mutant m6/2 were inhibited by lactisole, we observed subtle differences (Fig. [3d](#F3){ref-type="fig"}). The IC~50~values were 0.05 ± 0.01 for the human receptor, 0.06 ± 0.01 for m6, 0.11 ± 0.01 m6/1, 0.05 ± 0.02 for m6/3 and 0.16 ± 0.03 for rTas1r3-V738A (Fig. [3d](#F3){ref-type="fig"}). Thus, the human receptor and the mutants m6 and m6/3 behave identical while the mutants m6/1 and rTas1r3-V738A are \~2 fold less lactisole sensitive. This result shows that the replacement of lysine 735 by phenylalanine is required for full lactisole sensitivity of the mutant rat receptor. Notably, the gain of lactisole sensitivity in the rodentTas1r3 variants m6, m6/3, m6/1, is accompanied by a \~50% reduction of the signal amplitude compared to the wild type rTas1r3 (Fig. [3a](#F3){ref-type="fig"},[3b](#F3){ref-type="fig"}). In marked contrast, the loss of lactisole sensitivity by the rTas1r3 mutant m6/2 led to an increase in the signal amplitude (Fig. [3b](#F3){ref-type="fig"}). The diminished signal amplitudes are not caused by individual differences in the expression rates of the receptor variants (Fig. [4](#F4){ref-type="fig"}). Moreover the V738A receptor variant, that contains just a single amino acid exchange also showed a \~50 % reduced amplitude compared to the native rat sweet taste receptor (Fig. [3b](#F3){ref-type="fig"}, [3c](#F3){ref-type="fig"}). This reduction is independent of the sweet taste agonist (Fig. [3c](#F3){ref-type="fig"}). The activation of the TAS1R2-TAS1R3 receptor by different sweeteners involves several receptor sites \[[@B9]-[@B12]\]. As the V738A amino acid exchange leads to an agonist independent amplitude reduction, this observation in combination with its localization in the fifth transmembrane region argues for a global or regional alteration of the receptor structure instead of a specific change at the agonist binding sites. Conclusion ========== The lactisole insensitivity of rat sweet taste receptor is caused by valine 738 in transmembrane region five. The additional replacement of a lysine residue by a phenylalanine at amino acid position 735 again in transmembrane region five is sufficient to increases the lactisole sensitivity of the mutant rat receptors to the level of the human counterpart. Thus, transmembrane region five is critical for the interaction of lactisole with the sweet receptor. The observation that the mutant receptors displays a generally reduced sensitivity towards all agonists suggests that the lactisole insensitivity of the rodent receptor might be more likely caused by an inaccessibility of the lactisole binding site rather then by its direct disruption. Instead the lactisole insensitivity of the rodent receptor more is likely a side effect that is caused by a steric alteration in the fifth transmembrane region of the rat TAS1R3 subunit induced by the valine at amino acid 738. Recently functional studies of the mouse and human sweet taste receptor in combination with mutational analysis led to a model of the lactisole binding site in human TAS1R3 \[[@B18]\]. This model predicts that the alanine 733 in human TAS1R3 which corresponds to valine 738 of rat TAS1R3 is not directly involved in lactisole binding although it lies in close proximity to the lactisole binding pocket. The proximity of this amino acid to the predicted lactisole binding pocket further supports our conclusion that valine 738 in rat sweet taste receptor prevents the access of lactisole to its binding site. Methods ======= Cloning of the TAS1R cDNAs -------------------------- hTAS1R2was constructed via amplification of all exons from genomic DNA by recombinant polymerase chain reactions (PCR) \[[@B19]\]. hTAS1R3was cloned from testis cDNA. rTas1r2was cloned from rat vallate papillae cDNA. PCR products were cloned into a pcDNA5/FRT expression vector (Invitrogen), fused with a C-terminal FLAG-epitope for immuncytochemical detection, sequenced, and checked for comparable protein expression by immuncytochemistry of transfected HEK293T cells. Construction of chimeric receptor cDNAs --------------------------------------- The receptor chimeras of Tas1r3 were constructed using recombinant PCR. In the first reactions the N-terminal and the C-terminal part containing the seven transmembrane domains of hTAS1R3and rTas1r3were amplified from plasmid DNA using Ultra Pfu Polymerase (Stratagene). For the hTAS1R3and rTas1r3N-termini the forward vector-anchored primer (CMV: CGCAAATGGGCGGTAGGCGTG) and a reverse gene specific primer (CACAGCCGGCTCGCCC) were used. The hTAS1R3and rTas1r3C-terminal part were amplified with forward gene specific primer (GGGCGAGCCGGCTGTG) and a reverse vector-anchored primer (TAGAAGGCACAGTCGAGG). PCR conditions were: 5 min, 94°C, 24 cycles 1 min 94°C, 0.5 min 64°C, 2 min 72°C; 5 min 72°C. The amplicons were purified using the Highpure PCR Product Purification Kit (Roche) and used as template for the second PCR. To create the chimeric receptor tas1r3-hN-rTM, equimolar amounts (\~50 ng) of the hTAS1R3N-terminus and rTas1r3transmembrane domains were combined without additional oligonucleotides. PCR conditions were: 5 min 94°C; 15 cycles of 2 min 64°C, 2 min 72°C and 0.5 min 94°C; followed by 5 min 64°C and 10 min 72°C. The identical protocol and combining the rTas1r3N-terminus with the human transmembrane regions resulted in receptor chimera tas1r3-rN-hTM. These chimeras were cloned into pcDNA5/FRT. Construction of Tas1r3 mutants ------------------------------ Receptor mutants were generated by site directed mutagenesis according to the QuikChange protocol (Stratagene). The forward and reverse complement primers contained the desired mutations and annealed to the same sequence on opposite strands of the plasmids. The following rTas1r3receptor variants were generated: m1: S576L; m2: R632Q, S635P, S637Q; m3: M644L A645S; m4: A702T; m5: Q717H, V718M, V723A, E725V; m6: L735F, V738A, I740A. m6/1: V738A, I740A, m6/2: L735F, I740A, m6/3: L735F, V738A. m7: A793V, Y794L, Q795R. Subsequently a rTas1r3-V738A variant was generated. All constructs were checked by sequencing. Functional expression --------------------- The cDNAs were transiently transfected into HEK293T cells stably expressing the chimeric G-protein subunit Gα~16gust44~\[[@B20]\] using Lipofectamine 2000 (Invitrogen) according to the manufacturer\'s protocol. 3--4 hours after transfection, DMEM was replaced by low-glucose DMEM supplemented with GlutaMAX and 10% dialyzed FBS (Invitrogen). 22--44 hours later, cells were loaded for 1 hour with the calcium sensitive dye Fluo4-AM (2 μg/ml in DMEM, Molecular Probes). Cells were washed 3x in solution C1 (130 mM NaCl, 5 mM KCl, 10 mM Hepes, 2 mM CaCl~2,~and 5 mM Glucose, pH 7,4). Calcium mobilization was monitored by an automated fluorometric imaging plate reader (FLIPR, Molecular Devices). Ligands (Sigma-Aldrich, Merck) were dissolved in C1 solution. All data were collected from at least two independent experiments carried out in triplicate. The obtained calcium signals were corrected for the response of mock transfected cells and normalized to the fluorescence of cells prior to the stimulus using ΔF/F=(F-F0)/F0. Concentration-response curves and EC~50~and IC~50~values were calculated in SigmaPlot by nonlinear regression using the function f = ((a-d)/(1+(x/EC~50~)^nH^)+d) and f = (a-b)/ \[1+(x/IC50)^nH^\]+b respectively. Immuncytochemistry ------------------ HEK-293T/G~α16gust44~cells were seeded on coverslips coated with 10 μg/ml poly-D-lysine and transfected with the respective cDNAs. 48 h after transfection cells were washed with PBS and fixed and permeabilized for 5 min in acetone:methanol (1:1). Non-specific binding was reduced by incubating the cells in 5% goat serum for 1 h. To detect the receptors, antiserum against the FLAG-epitope (anti-FLAG M2 (Sigma), 1:2000 in 3% goat serum) was added to the cells for 1 h at room temperature (RT). After washing the cells three times with PBS we added Alexa488-conjugated goat antiserum against mouse IgG ((Molecular Probes), 1:1000 in 3% goat serum) for 1 h at RT. The cells were embedded in Fluorescent Mounting Medium (Dako) and analyzed using a fluorescence microscope (Zeiss Axioplan, Jena) and a camera (RT Slide, Visitron Systems, Munich). List of abbreviations ===================== GPCR G-protein coupled receptor HEK293T/ Gα~16gust44~Human embryonic kidney cells stably expressing the large T antigen and the G protein chimera Gα~16gust44~ EC~50~half maximal effective concentration IC~50~half maximal inhibitory concentration FLIPR fluorometric imaging plate reader Authors\' contributions ======================= MW cloned most of the TAS1R subunits, constructed the chimeric TAS1R3 receptors and carried out all functional studies. BB created the receptor variants m1--m7, participated in the study design and drafted the manuscript. W.M conceived the study, participated in its design and coordination and helped to draft the manuscript. All authors read and approved the final manuscript. Acknowledgements ================ We thank Jay Slack (Cincinnati) for the HEK293T Gα~16gust44~cell line, Hartwig Schmale and Nicole Burhenne (Hamburg) for the rTas1r3 cDNA, and Konrad Hoppe for providing stevioside, cyclamate, and aspartame. This project was supported through DFG-grants ME 1024 / 1--2 and ME 1024 / 2-1. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Calcium responses in cells co-transfected with different rat and human TAS1Rsubunits to stimulation with various sweeteners in the presence and absence of lactisole. (a) Intracellular calcium traces recorded in the FLIPR. Upper panel: Responses of cells cotransfected with cDNAs hTAS1R2 and hTAS1R3. Lower panel: Response of cells co-transfected with cDNAs rTas1r2 and rTas1r3. The cells were challenged with 3 mM saccharin containing 0.0, 0.01, 0.1, or 1.0 mM lactisole. Arrows denote the time of application. Signals were averaged over three independent wells of a FLIPR plate. Vertical scale, 1000 counts; horizontal scale, 1 min. (b) Cells co-transfected with hTAS1R2-hTAS1R3, (c) rTas1r2-rTas1r3, (d) hTAS1R2-rTas1r3, or (e) rTas1r2-hTAS1R3. c, response of mock-transfected control cells; lac, 1.0 mM lactisole; asp, 15 mM aspartame; cyc, 30 mM cyclamate; neo, 0.1 mM neohesperidine dihydrochalcone; sac, 10 mM saccharin; suc, 100 mM sucrose; stev, 1 mM stevioside; -, response without lactisole; +, response in the presence of 1.0 mM lactisole. All sweeteners were used at concentrations which were close to the maximal obtainable signal. ::: ![](1471-2202-6-22-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Effect of lactisole on Tas1r3 variants. (a) Calcium responses of cells co-transfected with hTAS1R2and a Tas1r3chimera comprising the human N-terminus and the rat transmembrane region or hTAS1R2and a Tas1r3chimera comprising the rat N-terminus and the human transmembrane region (b) to stimulation with various sweeteners in the presence and absence of lactisole. c, response of mock-transfected control cells; lac, 1.0 mM lactisole; asp, 15 mM aspartame; cyc, 30 mM cyclamate; neo, 0.1 mM neohesperidine dihydrochalcone; sac, 10 mM saccharin; suc, 100 mM sucrose; stev, 1 mM stevioside; -, response without lactisole; +, response in the presence of 1.0 mM lactisole. All sweeteners were used at concentrations which were close to the maximal obtainable signal. (c) Alignment of amino acid sequences of the human, rat, and mouse Tas1r3 transmembrane regions. Color code: white, identical; grey, conservative; dark-grey, similar; black, non-similar. TM, predicted transmembrane domain, indicated by a black bar. Dots indicate the location of the primer used to generate the chimeric rat-human Tas1r3 receptors. Asterisks denote amino acid positions that are identical in rat and mouse but variable between humans and rodents and are located in the extracellular loops or the upper part of the transmembran regions. Boxes indicate the several rTas1r3 mutants m1 -- m7, including the amino acid substitutions. m6/1 -- m6/3 signified the reconverted amino acid of mutant m6. ::: ![](1471-2202-6-22-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Effect of lactisole on Tas1r3 variants. (a, b) Responses of cells co-transfected with rTas1r2and rTas1r3mutants to stimulation with 1 mM stevioside. (c) Calcium responses of cells co-transfected with rTas1r2and the rTas1r3-V738A receptor variant to stimulation with various sweeteners in the presence and absence of lactisole. c, response of mock-transfected control cells; lac, 1.0 mM lactisole; asp, 15 mM aspartame; cyc, 30 mM cyclamate; neo, 0.1 mM neohesperidine dihydrochalcone; sac, 10 mM saccharin; suc, 100 mM sucrose; stev, 1 mM stevioside; -, response without lactisole; +, response in the presence of 1.0 mM lactisole. All sweeteners were used at concentrations which were close to the maximal obtainable signal. (d) Concentration-dependent inhibition of calcium responses by lactisole in cells transfected with hTAS1R2-hTAS1R3(open circles, solid line), rTas1r2-rTas1r3 V738A(squares, dash-dot-dotted line), rTas1r2-rTas1r3 m6(open triangles up, long dashed line), rTas1r2-rTas1r3 m6/1(filled circles, short dashed line), rTas1r2-rTas1r3 m6/2(triangles down, medium dashed line) or rTas1r2-rTas1r3 m6/3(open diamond, dotted line) that were challenged with 3 mM stevioside. ::: ![](1471-2202-6-22-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Expression of receptor variants. The table shows the overall expression in % of all receptor constructs detected by immuncytochemistry (a), transmission and fluorescence pictures of HEK293T/ G~α16gust44~cells transfected with rTas1r3(b), rTas1r3-V738A(c) or rTas1r3 m6/3(d) DNA. Scale, 50 μm. ::: ![](1471-2202-6-22-4) :::	PubMed Central	2024-06-05T03:55:55.841805	2005-4-7	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084349/", "journal": "BMC Neurosci. 2005 Apr 7; 6:22", "authors": [ { "first": "Marcel", "last": "Winnig" }, { "first": "Bernd", "last": "Bufe" }, { "first": "Wolfgang", "last": "Meyerhof" } ] }
PMC1084350	Background ========== Women and infants of a multiple pregnancy are recognised to be at increased risk of adverse outcome when compared with singletons. The greatest risk to infants of a multiple pregnancy is being born preterm, with preterm birth, defined as birth less than 37 weeks gestation, and very preterm birth less than 32 weeks gestation. The preterm birth rate less than 37 weeks for women with a singleton pregnancy is 6.3% versus 97% for women with a triplet pregnancy \[[@B1]\] the mean gestational age of birth for infants of a triplet pregnancy being 31.9 weeks, with 39.3% if infants born before 32 weeks gestation, and a further 57.7% between 32 and 36 weeks gestation \[[@B1]\]. Infants of a triplet pregnancy are at increased risk of poor intrauterine growth, with the mean birth weight of a triplet infant being 1668 grams, compared 3398 with grams in singleton infants \[[@B1]\]. At birth, 15.9% of triplet infants weigh less than 1000 grams, 35.9% less than 1500 grams, and 92.9% less than 2500 grams \[[@B1]\]. Infants of a higher order multiple pregnancy are at increased risk of perinatal death, with a rate of 53.0/1000, almost 7 times greater than that observed in singletons \[[@B1]\]. Several studies have demonstrated a favourable effect of bed rest for women with a triplet pregnancy on fetal growth \[[@B2],[@B3]\]. However, advice regarding the timing and duration of bed rest has varied, including hospitalisation from 28 to 30 weeks gestation until birth \[[@B4]\], from 24 weeks until the beginning of the third trimester \[[@B5]\], or only at the onset of complications \[[@B6]\]. Hospital admission has been advocated in the past for women with a twin pregnancy, as a means of reducing the risk of preterm birth and improving fetal growth \[[@B7]\]. However, the Cochrane Systematic Review assessing the role of hospitalisation and bed rest for women with an uncomplicated twin pregnancy has found the practice to be associated with an increase in the risk of preterm birth, and should not be offered as part of routine care \[[@B8]\]. The value of admission to hospital for rest in triplet or higher order multiple pregnancy is uncertain, with little consistent information available. Several retrospective studies assessing bed rest for women with a triplet pregnancy suggest a reduction in the risk of preterm birth \[[@B9],[@B10]\], while others have not demonstrated a prolongation in gestation \[[@B4],[@B11]\]. The effect of bed rest on perinatal mortality is similarly associated with inconsistent findings, some authors reporting a reduction in mortality \[[@B4],[@B9],[@B10]\], others not \[[@B11]\]. In the only small, randomised study to date in 19 triplet pregnancies, hospitalisation for rest suggests a beneficial trend in reducing the incidence of preterm birth and of increased birthweight in the hospitalised group \[[@B12]\]. All of these beneficial results are compatible with chance variation. There is a need for further evaluation of the effects of admission to hospital for rest in women with a triplet pregnancy. Any potentially beneficial effects of hospitalisation and bed rest for infant health outcomes must be considered in light of the physical and psychosocial effects on the pregnant woman \[[@B13],[@B14]\]. The separation from family members and the practical issues related to this separation have been identified as a considerable stressor associated with hospitalisation \[[@B13]\]. This randomised controlled trial was designed to assess the effects of hospitalisation from 24 to 30 weeks gestation for women with a triplet pregnancy on the risk of preterm birth. Our primary hypotheses were that routine hospitalisation of women with a triplet pregnancy from 24 to 30 weeks gestation would be associated with a reduction in the incidence of preterm birth (defined as birth less than 37 weeks), and very preterm birth (defined as birth less than 34 weeks). Methods ======= Women with a triplet pregnancy, with ultrasound confirmed gestational age of less than 24 weeks were approached from the antenatal clinic of the Women\'s and Children\'s Hospital for participation in the study. Women with a triplet pregnancy and any other condition requiring hospitalisation (for example, placenta praevia) were excluded from participation in the trial. Approval was obtained from the research and ethics committee of the Women\'s and Children\'s Hospital. Recruitment commenced in 1996 and was terminated 2003 due to a lack of success in securing research funding to support multicentred collaboration, and difficulties experienced in recruiting sufficient women. Women with a triplet pregnancy were identified early in their antenatal care, provided with the trial information, and asked to discuss their participation with a family member. Women who provided informed written consent were then randomised to either antenatal hospitalisation (hospitalised group), or to routine antenatal care (control group). The randomisation schedule used variable blocks with stratification by parity, and was prepared by an investigator not involved in clinical care. After completion of the trial entry details, an independent researcher responsible for treatment allocation was contacted by telephone, and the next in a series of consecutively numbered, opaque, sealed envelopes opened. The treatment allocation was stated over the telephone for either hospitalisation for bed rest (hospitalised group) or not (control group) according to the instructions enclosed in the envelope. Women allocated to the hospitalised group were admitted to hospital from 24 weeks gestation until 30 weeks gestation, after which time, women were discharged home and encouraged to obtain as much rest as possible. All women were able to ambulate within the hospital, received a normal hospital diet, and fortnightly routine antenatal assessments. Women were allowed leave from the ward over weekend periods to assist with compliance with continued hospitalisation. Women allocated to the control group were encouraged to continue with their normal activities at home, and were reviewed fortnightly in the antenatal clinic. They were admitted to hospital if any complications developed, such as preterm labour, preterm prelabour ruptured membranes, and pregnancy induced hypertension. Baseline characteristic were obtained to describe the two groups at randomisation, and included maternal age, booking weight and height, smoking and alcohol use, mode of conception (spontaneous conception versus conception via assisted reproductive techniques), previous pregnancy outcomes, and use of antenatal corticosteroids for fetal lung maturation. The primary study outcomes were the incidence of preterm birth (defined as birth less than 37 weeks gestation) and very preterm birth (defined as birth less than 34 weeks gestation), and the development of maternal pregnancy induced hypertension (defined as blood pressure greater than 140/90 mmHg or an increase in the diastolic blood pressure of more than 15 mmHg from booking. Secondary outcome measures included tocolytic use, mode of birth, infant Apgar score of less than seven at five minutes, infant birth weight less than 2500 grams, infant birth weight less than 1500 grams, admission to the neonatal unit, length of stay in the neonatal unit more than seven days, perinatal death (stillbirth and neonatal death), and the occurrence of neonatal morbidity (including respiratory distress syndrome, intraventricular haemorrhage, and necrotising enterocolitis). The previous randomised controlled trial assessing hospitalisation and bed rest for women with a triplet pregnancy suggests a reduction in the occurrence of preterm birth less than 34 weeks gestation from 44% to 30% \[[@B12]\]. A sample of 400 women would be able to detect this difference, at a level of statistical significance of 5%, and power of 80%. Using preterm birth less than 37 weeks gestation, a sample of 52 women would be able to detect a reduction from 100% to 80% (p = 0.05; power = 80%). Baseline characteristics were compared, to assess comparability of the treatment groups at trial entry. All randomised women were included in the analysis on an intention to treat basis. Relative risks and 95% confidence intervals were calculated for primary and secondary outcomes. These results were then incorporated into a systematic review and meta-analysis with the previous trial of hospitalisation and bed rest for women with a triplet pregnancy \[[@B12]\]. Results ======= Seven women with a triplet pregnancy at less that 24 weeks gestation were recruited to the trial, with three women randomised to the hospitalisation group, and four women to the control group. Baseline characteristics between the two groups were comparable (Table [1](#T1){ref-type="table"}). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Baseline antenatal characteristics ::: Characteristic Hospitalisation Group N = 3 Control Group N = 4 -------------------------------------------- --------------------------------- ------------------------- Gestational age at randomisation (weeks)\* 23.4 ± 1.7 22.0 ± 1.8 Maternal Age (years)\* 33.2 ± 0.8 36.2 ± 13.2 Booking Weight (kg)\* 62.3 ± 11.0 61.3 ± 3.2 Height (cm)\* 170.0 ± 1.5 164.5 ± 3.5 Caucasian 3 4 Married / Defacto 3 4 Primigravid 2 2 Smoker 1 0 Alcohol use in pregnancy 1 0 Conception spontaneous 1 2 \mean and standard deviation ::: Of the women in the hospitalisation group, two developed antenatal complications (one pregnancy induced hypertension, and one recurrent antepartum haemorrhage), two gave birth by caesarean section (both prelabour procedures), and two developed postnatal complications (one primary postpartum haemorrhage, and one raised blood pressure). All three women were administered antenatal corticosteroids, but none were administered antenatal tocolytic agents. All three women gave birth at less than 37 weeks gestation, and two at less than 34 weeks gestation (Table [2](#T2){ref-type="table"}). ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Primary outcome measures ::: Outcome* Hospitalisation Group N = 3 Control Group N = 4 Relative Risk 95% CI -------------------------------- --------------------------------- ------------------------- ------------------- --------------- Birth \<37 weeks gestation 3 4 Not estimable Birth \<34 weeks gestation 3 2 2.00 0.75 to 5.33 Pregnancy induced hypertension 1 1 1.33 0.13 to 13.74 Figures are numbers ::: Of the women in the control group, three were admitted during their antenatal course (one for threatened preterm labour, one for pregnancy induced hypertension, and one for recurrent antepartum haemorrhage and one for intrauterine growth restriction), three of the four women were administered antenatal corticosteroids and one woman received antenatal tocolytic agents. All four women gave birth by caesarean section (three elective prelabour procedures), and three developed postnatal complications (one postpartum haemorrhage and two raised blood pressure). All four women gave birth at less than 37 weeks gestation, and two women gave birth at less than 34 weeks gestation. There were no statistically significant differences between the two groups for any of the infant outcomes (Table [3](#T3){ref-type="table"}). There were no stillbirths and two neonatal deaths in one triplet grouping in the control group (one due to tracheal stenosis and the other to necrotising enterocolitis and bowel infarction). ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Secondary outcome measures ::: Outcome Hospitalisation Group N = 9 Control Group N = 12 Relative Risk 95% CI ------------------------------------ --------------------------------- -------------------------- ------------------- -------------------- Gestational age at birth (weeks)\* 33.5 ± 2.7 33.5 ± 3.5 -0.01 -4.60 to 4.53 Birth weight (grams)\* 1892 ± 251.8 1810 ± 551.8 82.00 -314.28 to 478.28 Birth weight \<2500 grams 9 8 1.12 0.89 to 1.42 Birth weight \<1500 grams 1 3 0.33 0.04 to 2.63 Placental weight (grams)\* 906.7 ± 136.5 1210.5 ± 557.9 -303.83 -1082.18 to 474.52 Apgar score \<7 at 5 mins 0 1 0.33 0.04 to 2.63 Neonatal Death 0 2 0.20 0.01 to 3.66 Neonatal Morbidity 0 1 0.33 0.02 to 7.24 Figures are numbers and relative risk or \mean and standard deviation, and weighted mean difference ::: When the results of this trial are incorporated into a meta-analysis with the previous randomised controlled trial assessing hospitalisation and bed rest for women with a triplet pregnancy \[[@B12]\], a total sample size of 26 women and 78 infants is obtained. There were no statistically significant differences identified between the two groups for the outcomes reported (Table [4](#T4){ref-type="table"}). ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Meta-analysis of two randomised trials assessing bed rest for triplets ::: Outcome* Number Trials Number Participants Relative Risk 95% Confidence Intervals ------------------------------------ ------------------- ------------------------- ------------------- ------------------------------ Gestational age at birth (weeks)\* 2 26 0.57 -1.36 to 2.51 Birth \<37 weeks gestation 2 26 0.80 0.59 to 1.09 Birth \<34 weeks gestation 2 26 0.87 0.36 to 2.08 Caesarean section 2 26 1.18 0.47 to 2.96 Maternal hypertension 2 26 0.43 0.11 to 1.72 Birth weight \<2500 grams 2 75 0.87 0.36 to 2.08 Birth weight \<1500 grams 2 75 1.18 0.47 to 2.96 Apgar score \<7 at 5 mins 2 75 0.43 0.11 to 1.72 Neonatal unit admission 2 71 0.90 0.74 to 1.09 Neonatal stay \>7 days 2 71 1.39 0.80 to 2.42 Perinatal Death 2 78 2.71 0.12 to 63.84 Neonatal Death 2 75 0.19 0.02 to 1.53 \*weighted mean difference ::: Discussion ========== The results of this small randomised trial and meta-analysis suggest no benefit of routine hospitalisation and bed rest for women with a triplet pregnancy in reducing the risk of preterm birth and improving fetal growth. The effect of the additional seven women recruited from this trial has been to reduce the magnitude of potential beneficial trends identified previously \[[@B12]\], with the point estimates for outcomes coming closer to unity. While the combined sample size in the meta-analysis of 26 women remains underpowered to be able to reliably detect differences in rates of preterm birth and in particular, perinatal mortality, it is unlikely that there will be further attempts to answer the question of the value of hospitalisation and bed rest in women with triplet pregnancies in the form of randomised controlled trials. In any case, the reduction in magnitude of potential benefits somewhat reduces the degree of uncertainty that remains in clinical practice. The difficulties encountered in recruitment to this trial highlight a number of issues. Randomised trials where the target population (in this case, women with a triplet pregnancy) comprise a small proportion of the obstetric population will, by necessity, require multicentred collaboration. While this trial received ethics approval from a number of collaborating centres, all women recruited were from the coordinating centre. This may reflect an inability of busy clinicians to actively recruit women due to time constraints within their practice, or lack of specific funding, thereby relying on collaborators to act altruistically, with little financial or other rewards for multicentre collaboration \[[@B15]\]. While the current health care system may not be conducive to active participation in clinical research, it is unlikely that the structure will change sufficiently in the short term. Another equally difficult approach may be to change the attitude of obstetricians, with greater emphasis, particularly during training, on the need for research. If clinical research is then viewed as a \"normal\" component of clinical practice, participation and recruitment to clinical trials may be facilitated \[[@B15]\]. After all, in the face of uncertainty about a clinical intervention or treatment, the most ethical response on the part of the clinician is to offer participation in a clinical trial \[[@B16]\], which essentially involves \"choice under uncertainty, plus data collection\" \[[@B17]\]. Conclusion ========== The results of this trial and meta-analysis suggest no benefit of routine hospitalisation and bed rest for women with a triplet pregnancy in terms of reducing the risk of preterm birth and improving fetal growth. The adoption or continuation of a policy of routine hospitalisation and bed rest for women with an uncomplicated triplet pregnancy cannot be recommended. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= CAC conceived the study and was involved in the study coordination, data analysis and revisions of the manuscript. JMD was involved in verification of the data, data analysis, drafted the original manuscript and contributed to subsequent revision. Both authors read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2393/5/8/prepub>	PubMed Central	2024-06-05T03:55:55.843581	2005-4-4	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084350/", "journal": "BMC Pregnancy Childbirth. 2005 Apr 4; 5:8", "authors": [ { "first": "Jodie M", "last": "Dodd" }, { "first": "Caroline A", "last": "Crowther" } ] }
PMC1084351	Background ========== Urinary tract infection (UTI) is one of the most common sources of infection in children under 5. In a small proportion of children UTI may lead to renal scarring \[[@B1],[@B2]\]. This outcome of infection is of concern as it is associated with significant future complications and ultimately with end stage renal disease\[[@B3]\]. Prompt diagnosis and treatment is therefore important to reduce the risk of future scarring. Clinical history and examination is the first step in any diagnosis and is the means of identifying children with suspected UTI. Elements of the clinical examination have also been evaluated as diagnostic tests for UTI but there is little data available on these. Urine tests are commonly used for the diagnosis of UTI. The reference standard for the diagnosis of UTI in children is considered to be any bacterial growth on a culture of urine obtained by suprapubic aspiration\[[@B4]\]. Culture has the disadvantage of taking at least 48 hours to give a result. More rapid methods of UTI diagnosis are therefore desirable. The most widely used rapid tests are dipsticks. Analytes commonly tested by dipsticks include leukocyte esterase, nitrite, blood and protein\[[@B4]\]. Dipstick tests have the advantage of being quick and easy to perform and can be carried out in primary care giving an immediate result. Microscopic examination of urine samples for leukocytes or bacteria \[[@B4]\] is considerably more time consuming and labour intensive than the dipstick method\[[@B5]\]. However, unlike culture, it can be used to give results within the primary care setting. An uncontaminated sample is necessary to reach an accurate diagnosis. Obtaining this is a particular issue when investigating young children. Table [1](#T1){ref-type="table"} presents a summary of the advantages and disadvantages of these tests. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Details of tests evaluated in the review ::: ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Test Details Advantages Disadvantages ----------------------------------- -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- -------------------------------------------------- ---------------------------------------------------------------------------------------- Urine sampling Suprapubic aspiration (SPA) Needle attached to syringe inserted through lower abdomen into bladder. Least risk of contamination Invasive Transurethral catheterisation Catheter inserted through the urethra into the bladder. Less invasive than SPA Invasive, causes pain and distress to child Clean voided urine (CVU) Midstream sample collected in sterile container. Non-invasive, easy to obtain Difficult in younger children Urine bags Bag applied to perineum. Suitable for babies and infants Risk of contamination Urine pads Absorbent pad placed in nappy. Dipstick Nitrite Gram-negative bacteria reduce dietary nitrate to nitrites. Very easy and quick to perform, relatively cheap Less accurate than culture Leukocyte esterase (LE) Glucose Leukocyte esterase is an enzyme that suggests the presence of leukocytes. Normal urine contains small amount of glucose. Bacteria metabolise glucose and so this test tests for the absence of glucose. Requires morning fasting urine specimen. Not commercially available, not suitable for non-potty trained children Microscopy Pyuria Urine examined through microscope for presence of white blood cells. Samples may be centrifuged before examination Quicker than culture More time consuming than dipstick, more expensive than dipstick and culture Bacteriuria Urine examined for presence of bacteria.\ Urine may be Gram-stained. Culture Standard Culture Reference standard test for UTI. Involves streaking urine on enrichment and selective media. Very accurate Time consuming: takes 48 hours to give a result, has to be performed in the laboratory ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ::: A wide range of other tests have been evaluated for the diagnosis of UTI. These include dipslide and rapid culture methods, colorimetric tests, headspace gas analysis, impedance, bio- and chemical luminescence, immunologic tests (e.g. ELISA), enzyme tests, bacterial oxygen consumption, and turbidimetry. However, these are not in widespread use and will not be discussed in this paper. This review aims to determine the diagnostic accuracy of dipstick and microscopy, and different methods of urine sampling, for detecting UTI in children under five years of age. Two previous reviews have addressed a similar objective\[[@B6],[@B7]\]. These were published over 2 years ago and did not assess urine sampling. They also included fewer studies (48 and 26 compared to 70), possibly as a result of less extensive literature searches and tighter inclusion criteria, than this review. This review therefore presents the most up to date and extensive systematic review of the topic area. Methods ======= We searched 16 electronic databases from inception to between October 2002 and February 2003. Update searches were conducted in May 2004. To identify additional published and unpublished studies we searched the internet, hand searched 12 key journals, screened reference lists of included papers and contacted experts in the field. We did not apply any language restrictions. Full details of the search strategy will be reported elsewhere\[[@B8]\]. Studies had to meet the following criteria to be included in the review: Study design: diagnostic cohort (single sample) studies Population: at least some children aged \<5 years with suspected UTI Index tests:microscopy or dipstick tests used to diagnose UTI or an evaluation of urine sampling methods. Reference standard:culture or culture combined with other tests Outcome measures:sufficient information to construct a 2 × 2 table Two reviewers independently screened titles and abstracts for relevance, we resolved disagreements by consensus. One reviewer performed inclusion assessment; data extraction and quality assessment and a second reviewer checked this. We extracted 2 × 2 data and used this to calculate measures of diagnostic performance. We used QUADAS to assess study quality\[[@B9]\]. Individual QUADAS items were used to investigate heterogeneity and to present a detailed assessment of quality to the reader. For each test, or test combination, we calculated the range in sensitivity, specificity, positive (LR+) and negative (LR-) likelihood ratios, and diagnostic odds ratios (DOR). We selected likelihood ratios as the measure of test performance for further analysis as these measures are easier to interpret than sensitivity and specificity \[[@B10]\]. For tests investigated in more than two studies, we used random effects models to pool positive and negative likelihood ratios \[[@B11]\]. Where studies presented more than one estimate of test performance for the same test, for example at different cut-off points or for different patient subgroups, we only included one estimate in the pooled analysis. We aimed to select the data set most similar to the estimates provided by the other studies in terms of population, test manufacturer or population. Heterogeneity of likelihood ratios was investigated using the Q statistic \[[@B12]\] and through visual examination of forest plots of study results \[[@B13]\]. We presented individual studies results graphically by plotting estimates of sensitivity and specificity in receiver operating characteristic (ROC) space. Where sufficient data were available, we used regression analysis to investigate heterogeneity. We extended the summary ROC (sROC) model \[[@B14]\], estimated by regressing D (log DOR) against S (logit true positive rate -- logit false positive rate), weighted according to sample size, to include covariates relating to patient age (\<2 years, \<5 years, \<12 years and \<18 years), geographic region and each of the 14 QUADAS items. In addition, for microscopy for pyuria and bacteriuria a variable on whether the sample was centrifuged was included, and for microscopy for bacteriuria a variable for Gram stain was included. Results ======= The literature searches identified over 10 000 references of which 70 studies were included. Figure [1](#F1){ref-type="fig"} shows the flow of studies through the review process. A summary of the results of all 70 studies included in the review is provided \[see [Additional file 1](#S1){ref-type="supplementary-material"}\]. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Flow chart of studies through review process. ::: ![](1471-2431-5-4-1) ::: Quality ------- The median number of the 14 QUADAS items fulfilled was 8 (range 5--13). The main limitation with the studies was the failure to include an appropriate patient spectrum (\<40%) or to report inclusion criteria. Studies also failed to report sufficient details to judge whether clinical review bias (the availability of clinical information to the person interpreting the test results), diagnostic review bias (the availability of the results of the index test to the person interpreting the reference standard) and test review bias (the availability of the results of the reference standard to the person interpreting the index test) were avoided. Withdrawals and handling of uninterpretable results were also poorly reported. Figure [2](#F2){ref-type="fig"} illustrates the number of studies that answered \"yes\", \"no\" and \"not stated\" to each of the 14 QUADAS items. A summary of the results of the quality assessment for each study is provided \[see [Additional file 2](#S1){ref-type="supplementary-material"}\]. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Results of the quality assessment. ::: ![](1471-2431-5-4-2) ::: Urine sampling -------------- Thirteen studies, with a total of 17 different test evaluations, compared the diagnostic accuracy of different methods obtaining urine for testing \[[@B15]-[@B27]\]. These studies compared the results of culture from urine obtained by different sampling methods. Five studies reporting seven data sets assessed the diagnostic accuracy of a clean voided urine (CVU) sample, using a supra-pubic aspiration (SPA) urine sample as the reference standard \[[@B15]-[@B19]\]. When both samples were cultured the agreement between the two sampling methods was good. There was considerable heterogeneity in positive likelihood ratios (p \< 0.0001). However, the negative likelihood ratios were statistically homogeneous (p = 0.531). The pooled positive likelihood ratio for a CVU sample was 8.8 (95% CI: 2.6, 29.6) and the pooled negative likelihood ratio was 0.23 (95% CI: 0.18, 0.30). Overall, there were insufficient data to draw any conclusions regarding the appropriateness of using urine samples obtained from bags (4 studies)\[[@B16],[@B20],[@B21],[@B27]\] or pads/nappies (4 studies) \[[@B22]-[@B25]\]. Dipstick tests -------------- A total of 39 studies reporting 107 data sets evaluated dipstick tests for the diagnosis of UTI \[[@B28]-[@B66]\]. These studies assessed the utility of dipstick tests for nitrite, leukocyte esterase (LE), protein, glucose and blood, alone and in combination. Table [2](#T2){ref-type="table"} summarises the results of these studies. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Summary of results for studies of dipstick tests ::: Dipstick positive for: Number of studies Range in LR+ Pooled LR+ (95 % CI)\* Range in LR- Pooled LR- (95 % CI)\* -------------------------------- ----------------------- ------------------ ---------------------------- ------------------ ---------------------------- Nitrite 23 2.5 -- 439.6 15.9 (10.7, 23.7) 0.12 -- 0.86 0.51 (0.43, 0.60) LE 14 2.6 -- 32.2 5.5 (4.1, 7.3) 0.02 -- 0.66 0.26 (0.18, 0.36) Nitrite or LE 15 3.0 -- 32.2 6.1 (4.3, 8.6) 0.03 -- 0.39 0.20 (0.16, 0.26) Nitrite and LE 9 6.3 -- 197.1 28.2 (17.3--46.0) 0.07 -- 0.86 0.37 (0.26, 0.52) Glucose 4 25.2 -- 156.1 66.3 (20.0, 219.6) 0.02 -- 0.38 0.07 (0.01, 0.83) Protein 2 1.7 & 1.8 na 0.78 & 0.96 na Blood 1 2.3 na 0.84 na LE and protein 1 17.4 na 0.12 na Nitrite, blood, or protein 1 2.7 na 0.28 na Nitrite, blood, or LE 1 1.3 na 0.50 na Nitite, blood and LE 1 3.5 na 0.19 na Nitrite, LE and protein 2 3.1 & 69.2 na 0.05 & 0.17 na Nitrite, LE, or protein 1 1.9 na 0.05 na Nitrite, LE, protein, or blood 1 8.0 na 0.19 na \* There was significant heterogeneity in all pooled estimates therefore these should be interpreted with caution ::: Figure [3](#F3){ref-type="fig"} shows the estimates of sensitivity and 1-specificity plotted in ROC space for glucose, and dipstick tests for nitrite and LE, alone and in combination. This graph suggests that glucose is considerably better than the other tests, both for ruling in and ruling out disease, this is supported by the pooled likelihood ratios. However, the confidence intervals around the pooled likelihood ratios are very large, especially for the negative likelihood ratios (ruling out disease), suggesting considerable uncertainty in these estimates. It should also be noted that very few studies of glucose tests were available and that they were all conducted over 30 years ago and the test used (\"Uriglox\") \[[@B58]\] is no longer commercially available. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Sensitivity and specificity plotted in ROC space for different dipstick tests. ::: ![](1471-2431-5-4-3) ::: Nitrite alone has a relatively high pooled positive likelihood ratio (15.9, 95% CI: 10.7, 23.7) and so may be useful for ruling in disease. However, it has a relatively poor negative likelihood ratio (0.51, 95% CI: 0.43, 0.60) suggesting that it may not be a useful test for ruling out disease. LE alone appears to be a relatively poor test both for ruling in (pooled LR+ = 5.5, 95% CI: 4.1, 7.3) and ruling out disease (pooled LR- = 0.26, 95% CI: 0.18, 0.36). A strategy which combines the results of LE and nitrite testing appears to offer the best performance both for ruling in and ruling out disease. A dipstick test positive for both nitrite and LE has the highest positive likelihood ratio (28.2, 95% CI: 17.3, 46.0) suggesting that this test combination may be used to rule in disease. A dipstick test negative for both LE and nitrite has the best negative likelihood ratio (0.20, 95% CI: 0.16, 0.26) suggesting that this test combination may be used to rule out disease. A dipstick test positive for either LE or nitrite and negative for the other is less informative for the diagnosis of UTI. Such a test result could be seen as an \"indeterminate\" test result requiring further investigation. It is difficult to draw conclusions about the overall accuracy of dipstick tests given the heterogeneity between studies in some areas, and the lack of data in others. There was insufficient information to make any judgement regarding the overall diagnostic accuracy of dipstick tests for protein, blood, or for combinations of three different dipstick tests (e.g. combination of LE, nitrite and blood). A regression analysis found that only clinical review bias showed an association with the diagnostic accuracy of nitrite dipstick (the DOR was 3.1 (95% CI: 0.97, 9.95) times higher in studies that avoided clinical review bias, i.e. in those studies that reported that the same clinical information was available to those interpreting the test results as would be available in practice). A higher DOR indicates higher overall accuracy. None of the items investigated, including age, showed a significant association with the DOR in the regression analysis for dipstick for LE, or for dipstick for LE or nitrite positive. Regression analysis was not carried out to investigate heterogeneity for other tests, as insufficient data were available. Microscopy ---------- A total of 39 studies reporting 101 data sets evaluated microscopy for diagnosing UTI \[[@B15]-[@B18],[@B26],[@B28]-[@B33],[@B35],[@B36],[@B38],[@B42],[@B52],[@B59]-[@B61],[@B64],[@B67]-[@B84]\]. Microscopy was used to determine the presence of pyuria or bacteriuria, or combinations of the two. Table [3](#T3){ref-type="table"} summarises the results of these studies. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Summary of results for studies of microscopy ::: Microscopy positive for: Number of studies Range in LR+ Pooled LR+ (95 % CI)\* Range in LR- Pooled LR- (95 % CI)\* ------------------------------ ----------------------- ------------------ ---------------------------- ------------------ ---------------------------- Pyuria 28 1.3 -- 27.7 5.9 (4.1, 8.5) 0.04 -- 0.68 0.27 (0.20, 0.37) Bacteriuria 22 1.6 -- 304.8 14.7 (8.6, 24.9) 0.01 -- 0.48 0.19 (0.14, 0.24) Pyuria or bacteriuria 8 1.5 -- 5.9 4.2 (2.3, 7.6) 0.02 -- 0.27 0.11 (0.05, 0.23) Pyuria and bacteriuria 8 2.7 -- 281.0 37.0 (11.0, 125.9) 0.07 -- 0.56 0.21 (0.13, 0.36) \* There was significant heterogeneity in all pooled estimates therefore these should be interpreted with caution ::: Figure [4](#F4){ref-type="fig"} shows the estimates of sensitivity and 1-specificity plotted in ROC space for all studies. This graph suggests that bacteriuria is considerably better than pyuria both for ruling out and ruling in disease. The diagnostic performance of bacteriuria may be improved when combined with pyuria. The pooled positive likelihood ratios are highest for pyuria and bacteriuria combined (37.0, 95% CI: 11.0, 125.9), where a positive result was defined as both tests positive, supporting the suggestion that the combination of a positive result for both of these tests may be useful for ruling in disease. Conversely, the lowest negative likelihood ratio resulted from the combination of pyuria and bacteriuria (0.11, 95% CI: 0.14, 0.24), where a negative result was defined as both tests negative, and this may be useful for ruling out disease. However, the confidence intervals around the pooled estimates are large, and in combination with the observed heterogeneity suggest considerable uncertainty in these estimates. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Sensitivity and specificity plotted in ROC space for different microscopy evaluations. ::: ![](1471-2431-5-4-4) ::: Regression analysis showed that centrifugation of the sample, reporting of selection criteria, reporting of details of reference standard execution, and reporting of uninterpretable results showed a significant association with the DOR in the studies of microscopy for pyuria. All of these items, with the exception of centrifugation, relate to the quality of reporting. The DOR was 6.25 (95% CI: 3.44, 11.11) times greater in samples that were not centrifuged; 3.19 (95% CI: 1.76, 5.79) times higher in studies that adequately reported selection criteria; 6.6 (95% CI: 2.43, 17.94) times higher in studies that reported sufficient details of reference standard execution; and 2.99 (95% CI: 1.50, 5.94) times higher in studies that reported on uninterpretable results. The association for centrifugation is not what we anticipated, as we would expect centrifugation of the sample to lead to improved test accuracy. In the analysis of microscopy for bacteriuria, Gram stain, incorporation bias and reporting of selection criteria showed a significant association with the DOR. The DOR was 5.96 (95% CI: 2.99, 11.89) times greater in samples that were Gram stained; 50.0 (95% CI: 6.67, 1000) times greater in studies in which incorporation bias was not present (i.e. studies in which the index test did not form part of the reference standard); and 2.46 (95% CI: 1.26, 8.41) times greater in studies that reported selection criteria. We would expect Gram staining to increase test performance as found in the analysis. However, we would expect the absence of incorporation bias to decrease test performance. The observed association may be explained by the fact that, for the purposes of the regression analysis, studies scoring \"unclear\" for a quality item were grouped with those scoring \"no\". For this analysis only one study scored \"no\" (i.e. incorporation bias was present) and the other studies grouped with this scored as \"unclear\". The association may therefore reflect quality of reporting, as does the association with reporting of selection criteria. Combinations of tests from different categories ----------------------------------------------- Nine studies including a total of 14 data sets examined the accuracy of different combinations of microscopy and dipstick tests for the diagnosis of UTI \[[@B30],[@B32],[@B35],[@B36],[@B42],[@B52],[@B67],[@B79],[@B85]\]. Given the results of individual tests, the test combination that appears to be potentially the most interesting is dipstick for LE and nitrite, and microscopy for pyuria and bacteriuria. Five studies investigated different permutations of these tests. \[[@B32],[@B35],[@B42],[@B67],[@B79]\] Three studies evaluated the accuracy of a positive result in one of these four tests (i.e. dipstick positive for LE or nitrite or microscopy positive for pyuria or bacteriuria) \[[@B32],[@B35],[@B67]\]. The results varied considerably between studies with positive likelihood ratios ranging from 0.8 to 35.9, and negative likelihood ratios ranging from 0.01 to 5.38. It is therefore not possible to draw overall conclusions from these studies. One study examined the combination of a positive result for all four tests \[[@B35]\]. This study reported a very high positive likelihood ratio (35.9) i.e. the combination was found to be very good for ruling in disease, but the negative likelihood ratio was less good at 0.28. These results might be expected given the results from the studies that examined combinations of dipstick tests, or combinations of microscopy tests. The other test combinations evaluated by these studies differed widely, and none were repeated between studies. Test combinations investigated included LE and nitrite dipstick test combined with microscopy for bacteriruria\[[@B42]\] or pyuria \[[@B42],[@B52],[@B85]\], dipstick for LE, nitrite and blood combined with microscopy for pyuria,\[[@B36]\] and dipstick for nitrite combined with microscopy for pyuria\[[@B30]\]. As most test combinations were only evaluated by one study and the definition of a positive test varied for the tests investigated by more than one study, it was not possible to draw conclusions regarding the diagnostic accuracy of these test combinations. Comparison of different tests ----------------------------- Comparison of the pooled likelihood ratios suggests that the microscopy combinations may be more accurate than the dipstick combinations. Only one study evaluated both dipstick positive for nitrite and LE and microscopy positive for bacteriuria and pyuria \[[@B59]\]. This study found that the dipstick combination was best for ruling in disease (LR+ was 18.9 for the dipstick combination compared to 11.6 for the microscopy combination). Five studies examined dipstick negative for nitrite and LE, and microscopy negative for pyuria and bacteriuria \[[@B32],[@B35],[@B40],[@B42],[@B59]\]. All but one found that microscopy was better for ruling out disease than dipstick. What do these results mean? --------------------------- If we take an estimate for the prevalence of UTI in children presenting to their GP with symptoms of possible UTI (the pre-test probability of disease), i.e. children in whom tests to diagnose UTI are likely to be used, likelihood ratios can be used to calculate the post-test probability of UTI. We were unable to find reliable estimates of the pre-test probability of UTI in the literature, and therefore used the results from the included studies to provide an estimate. Only studies that included an appropriate patient spectrum were included in this analysis. UTI prevalence varied greatly between studies (3--73%). As the distribution was highly skewed we used the median prevalence, which was 20%. Figure [5](#F5){ref-type="fig"} shows how the probability of UTI changes after testing. In a typical primary care setting in which the pre-test probability of disease is estimated to be around 20%, a negative likelihood ratio of 0.20 translates to a post-test probability of UTI of about 4%. In other words, children who receive a dipstick test negative for both nitrite and LE have a 4% probability of having a UTI. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Likelihood ratio nomogram for dipstick tests. ::: ![](1471-2431-5-4-5) ::: Discussion ========== An accurate and prompt diagnosis is important to inform patient management decisions in young children with suspected UTI. The first step in the diagnostic process is to identify children presenting to the GPs surgery who may have a UTI. This will inevitably involve a clinical assessment. It is very difficult, if not impossible, to capture all the signs and symptoms that a GP might use to develop a clinical suspicion of UTI and decide to test a child for UTI. Further research to accurately define from which children urine samples should be taken to test for UTI may be useful. Following clinical examination, the next step is to collect a suitable urine sample to test for the presence of infection. Different methods of urine sampling may be differently susceptible to contamination and hence to false positive results. The issue of appropriate urine sampling techniques is of particular concern in young children, where the collection of a sterile, mid-stream sample can be problematic. Suprapubic aspiration has been regarded as the reference standard collection method. This procedure is invasive and may require the use of ultrasound guidance to ensure that the needle is inserted into the bladder. The identification of an alternative sampling method with acceptable diagnostic performance, which can readily be applied in the GP\'s surgery, and which is more acceptable to children and parents, is therefore desirable. The studies on urine sampling showed reasonably good agreement between clean voided urine (CVU) and suprapubic aspiration (SPA) samples, suggesting that this is an appropriate routine method of urine collection. CVU samples are difficult to collect in young children who are not potty trained. A number of alternative collection methods have been developed, including bag, pad and nappy specimens. There is currently insufficient data available to determine whether bag or nappy/pad specimens may be used as substitutes for SPA Further work is needed in this area. The main types of urine testing evaluated for the diagnosis of UTI were dipstick and microscopy. Culture is generally considered to be the reference standard for UTI diagnosis. The logistics of urine culture represent a significant drawback; culture takes approximately 48 hours to give a result, is generally performed in the laboratory and is more expensive than other methods. For this reason alternative, more rapid tests are needed to guide the prompt initiation of treatment. Dipsticks have the advantage of providing an immediate result, and of being both cheap and easy to perform and interpret. The studies of dipstick tests showed considerable heterogeneity and so the results should be interpreted with caution. The results suggest that a dipstick test that is positive for both LE and nitrite is good for ruling in disease whilst one that is negative for both LE and nitrite is good for ruling out disease. An additional dipstick test that provided interesting results was the estimation of urinary glucose, where a negative urinary glucose is regarded as a positive test for UTI. Only four studies of this test were identified, and all were conducted more than 30 years ago. All studies reported excellent specificity for this test. Sensitivity was also very high in three of the studies but was lower, at 64% in the fourth. This last study was conducted in children aged less than one year, suggesting that the test maybe less useful in very young children. This difference in performance of the test with patient age may be explained by its apparent dependence on an overnight, fasting sample; such a sample would be impossible to obtain in children who are not toilet trained. However, given the limited results reported, this test appears to be potentially useful for the diagnosis of UTI in toilet trained children. Further studies are needed. Although, in practice, microscopy and culture are generally requested in combination, microscopy has the advantage of being quicker to provide a result. It may be that microscopy has some potential as a test that could be performed in the GP surgery. However, it remains more expensive than a dipstick test and requires some degree of expertise to perform. The studies of microscopy showed considerable heterogeneity, in terms of results, cut-off points, types of urine samples and population. A urine sample that was positive for both pyuria and bacteriuria on microscopy was found to be very good for ruling in disease. Similarly, a urine sample that was negative for both pyuria and bacteriuria on microscopy was found to be very good for ruling out disease. The possibility of publication bias remains a potential problem in this review. It is possible, and indeed likely, that studies reporting higher estimates of test performance are more often published, but the extent to which this occurs is unclear. There is evidence that publication bias is a particular problem for studies of small sample size, although these data are general and does not come from the diagnostic literature\[[@B86],[@B87]\]. We restricted this review to studies that included at least 20 children, meaning that this type of publication bias is less likely to be a problem. We are unaware of any articles on publication bias in diagnostic tests or on methods to formally assess publication bias in a diagnostic systematic review. We chose likelihood ratios as the primary effect measure as these are the measure that physicians find easiest to interpret \[[@B88]\]. We used pooled likelihood ratios and estimates of the pre-test probability of disease to calculate estimates of the post-test probability of disease. These measures provide a simple illustration of how the results of a test change the probability of disease and help the reader to determine how useful a test is likely to be in practice. The main limitation of this approach was the considerable heterogeneity in pooled likelihood ratios; it is debatable whether it is appropriate to pool these estimates. It is important that pooled estimates are interpreted with caution and that the heterogeneity between studies is considered when interpreting these results. A further problem with this analysis is that positive and negative likelihood ratios were pooled individually. These measures are likely to be correlated within an individual study and ignoring this correlation may be problematic\[[@B89]\]. We conducted a regression analysis to investigate possible explanations for the observed heterogeneity. This analysis was carried out according to standard methods for pooling studies of diagnostic accuracy using the summary ROC approach\[[@B14]\]. Using the DOR for further investigation of heterogeneity means that we can only assess whether the factors investigated are associated with the DOR and not with sensitivity and specificity, or with positive and negative likelihood ratios. Often factors that lead to an increase in sensitivity will lead to a decrease in specificity and vice versa, possibly with no effect on the DOR. A further limitation of this analysis was that we could only investigate the effect of variables at the study level. One factor that may impact on the accuracy of the diagnostic tests investigated is patient age. However, as the majority of studies investigated included children aged 0--16 or 18 years and did not report results separately for younger age groups it was not possible to carry out appropriate sub-group analyses to investigate the effects of age on estimates of test accuracy. This is an area where further investigation is required. Conclusion ========== Based on the results of this review dipstick negative for LE and nitrite, or microscopic analysis negative for pyuria and bacteriuria of a CVU, bag, or nappy/pad specimen may reasonably be used to rule out UTI. These patients can then be excluded from further investigation, without the need for confirmatory culture. Similarly, combinations of positive tests could be used to rule in UTI, and trigger further investigation. In the latter case, however, confirmation by culture may be preferred prior to the initiation of further, possibly invasive, investigations. Additional information on antibiotic sensitivities, which can be provided by culture, may also be a significant consideration. If combinations of rapid tests were routinely used to rule in and/or rule out disease, as described, then a cost saving in the number of cultures ordered would be expected. In addition it is likely that the number of children without disease exposed to inappropriate antibiotic therapy, whilst awaiting culture results, would be reduced. This may have implications for antibiotic resistance at a population level. The quality assessment highlighted several areas that could be improved upon in future diagnostic accuracy studies, in particular in relation to reporting. Future studies should follow the STARD guidelines for reporting of diagnostic accuracy studies \[[@B90]\]. The review also highlighted the following specific areas requiring further research for the diagnosis of UTI: • urine sampling methods in younger children • accuracy of the glucose test, and its practical applicability • handling of indeterminate nitrite and LE dipstick test results • accuracy of microscopy in combination with a dipstick test Abbreviations ============= CI Confidence interval CVU Clean voided urine DOR Diagnostic odds ratio ELISA Enzyme linked immunosorbant assay LE Leukocyte esterase LR Likelihood ratio QUADAS Quality assessment of diagnostic accuracy studies ROC Receiver operating characteristic SPA Suprapubic aspiration STARD Standards of reporting of diagnostic accuracy studies UTI Urinary tract infection Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= All authors contributed towards the conception and design of the study and the interpretation of the data. They also read and approved the final manuscript. PW and MW participated in data extraction, the analysis of data, and drafted the article. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2431/5/4/prepub> Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Microsoft Word file.doc containing a table of the results of individual studies included in the review. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 2 Microsoft Word file.doc containing a table of the results of the quality assessment of included studies. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ We would like to thank Professor Martin Bland, (Department of Health Sciences, University of York) for statistical advice, Julie Glanville (Centre for Reviews and Dissemination, University of York) for the conduct of electronic searches and management of the reference database, and Alison Booth (Centre for Reviews and Dissemination, University of York) for advice on the dissemination of results. This project was funded by the Health Technology Assessment Programme (project number 01/66/01). The views and opinions expressed in this paper are those of the authors and do not necessarily reflect those of the Department of Health.	PubMed Central	2024-06-05T03:55:55.845613	2005-4-5	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084351/", "journal": "BMC Pediatr. 2005 Apr 5; 5:4", "authors": [ { "first": "Penny", "last": "Whiting" }, { "first": "Marie", "last": "Westwood" }, { "first": "Ian", "last": "Watt" }, { "first": "Julie", "last": "Cooper" }, { "first": "Jos", "last": "Kleijnen" } ] }
PMC1084352	Background ========== Tuberculosis is still a great challenge for public health in Brazil and worldwide. In Recife, a large city in the Northeast Region of the country, the annual incidence of tuberculosis is high, about 100 per 100 000 inhabitants in 2003. Early detection of infectious tuberculosis cases followed by effective treatment is extremely important in controlling the disease, mainly because it can be assumed that the time of infection is likely to pre-date the positive sputum test by some time \[[@B1]\]. Because of this, recent studies have investigated reasons for delays in treatment \[[@B2]-[@B8]\]. The starting point from which the delays are measured is uncertain and there is no agreed definition as to what constitutes an acceptable delay. The cut off point in studies of risk factors for an acceptable delay has been defined in two ways: either a panel of experts agrees on a reasonable period of time or, alternatively, the median delay in observed data is used. Panels of expert have agreed on an acceptable total delay of 30 days \[[@B9],[@B10]\] or 60 days \[[@B11],[@B12]\]. Ward et al. \[[@B13]\] have criticized this approach, suggesting that what is an acceptable delay should depend on the local health services and the local epidemiological situation, with a shorter delay to be expected when incidence is high. Studies using the median observed total delay as the cut off point found medians ranging from 2 to 4 months \[[@B2],[@B3],[@B8],[@B14]-[@B16]\]. Studies using the median patient delay found medians ranging from 1 to 4 months \[[@B2],[@B3],[@B7],[@B13],[@B9],[@B17],[@B10]\]. Studies restricted to smear-positive cases have tended to find lower median patient delays \[[@B8],[@B16]\]. Risk factors identified in the literature as associated with long total delays before treatment include: those relating to perception and knowledge (inadequate knowledge about TB) \[[@B5],[@B7],[@B9]\], lack of perception of the need to seek health care \[[@B18]\]; those relating to the difficulty of access to health care (living in rural areas)\[[@B14]\], and distance from health care units \[[@B7],[@B8]\]; those relating to the severity or specificity of symptoms (vagueness of symptoms, absence of haemoptysis \[[@B5],[@B14],[@B9],[@B16]\], negative smears \[[@B2]\]; those relating to old age \[[@B2],[@B13],[@B14],[@B16]\], female sex \[[@B11],[@B6],[@B19],[@B20]\], ethnic group (white patients) \[[@B6]\] and alcoholism \[[@B8]\]. The only study of delay between the onset of symptoms and treatment in patients with pulmonary tuberculosis carried out in Brazil did not look at risk factors \[[@B21]\]. The aim of this study was to investigate biological, clinical, social, life-style and health-care factors associated with total treatment delay among cases of tuberculosis in residents of Recife, to identify the main constraints on early detection of patients and thus help with deciding which measures to use to decrease tuberculosis transmission in Recife. Methods ======= This was an observational study consisting of a cohort of cases of pulmonary TB, aged 18 years or more, diagnosed in Recife between May 2001 and May 2003. Case finding, mostly passive, is one of the main activities of the national tuberculosis control program (TCP). Decentralization has been underway since 2000, with the progressive transfer of activities from tuberculosis (TB) reference units to primary care units, as part of the Family Health Program (FHP) \[[@B22]\]. One of the activities of FHP staff is to detect people with respiratory symptoms (a cough persisting for three or more weeks) and to have their sputum examined. Patients were invited to participate in the study when the decision to start treatment was taken, i.e. when the diagnosis of tuberculosis was given according to the criteria established by the Ministry of Health. After signing the informed consent form they were interviewed by trained assistant nurses using a standard questionnaire, and had sputum and blood samples collected for examination. Patients were asked how long ago they first showed symptoms of pulmonary tuberculosis. In this way, the total delay (from onset of symptoms to start of tuberculosis treatment) was ascertained directly from the patients. Patients were not aware if they had acceptable or unacceptable delay; thus, recall bias, if it occurred, was likely to be non-differential. To define acceptable total delay we worked with two cut off points: sixty and ninety days. An analysis of risk factors for delays of ninety days or more (the median of our data) found essentially the same factors in the univariable analysis as with sixty days, but only smoking and loss of weight remained in the final multivariable model. We concluded that a cut off point of sixty days enables more potential public health interventions to be identified. Exposures studied included biological factors (age and sex), alcohol consumption, social factors (employment status, marital status, literacy; numbers of dwellers in residence, and income of the head of the household), history of contact with other cases of tuberculosis (knowing somebody with TB, early contact with TB patient, and living in the same house as a TB patient), clinical factors (previous tuberculosis treatment, cough, weight loss, acid-fast bacilli smear result, HIV co-infection, and haemoptysis), the Health District in which unit of treatment is located, and access to health services (number of health units attended with this complaint before start of treatment; whether unit of treatment is the local health center and whether it is in the district of residence; whether the Family Health Program had been implemented in the area of residence). The univariable OR of unacceptable treatment delay was estimated for biological factors, alcohol consumption, social factors, history of contact with other cases of tuberculosis, clinical factors, the health district in which treatment was carried out, and access to health services. The significance of these associations was tested using the chi-squared test and p-values were calculated. All variables with a univariable association with a p-value ≤ 0.20 were included in a multiple logistic regression analysis. Crude and adjusted ORs and 95% CIs were calculated. The attributable-risk percentages in the final model were calculated as ![](1471-2458-5-25-i1.gif), where RR was the adjusted OR for each variable. Results ======= Of 1126 cases of pulmonary TB in the study population, 21 (1.9 %) were excluded because of lack of information about the time of onset of symptoms, leaving 1105 patients for the study. The patients excluded were similar to those included in terms of characteristics identified as predictors of delay in the univariable analysis, with the exception of weight loss. The mean total delay was 142 days, the median 90 days (120 in men and 90 in women). Sixty-two percent of patients had a total delay of longer than 60 days (which we consider an unacceptable delay in this analysis). In 18%, the delay was longer than 181 days. Tables [1](#T1){ref-type="table"} to [5](#T5){ref-type="table"} show the results of univariable analysis of risk factors for unacceptable total delay. No association with delay was found in the case of: sex, age, neither heavy nor light alcohol consumption (table [1](#T1){ref-type="table"}), literacy, income of head of household, marital status of case, size of household (table [2](#T2){ref-type="table"}); history of contact with a case of tuberculosis (table [3](#T3){ref-type="table"}); any clinical factor other than weight loss (including smear-positive test result) (table [4](#T4){ref-type="table"}); and any factor concerning access to health services, except for district of treatment (table [5](#T5){ref-type="table"}). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Biological factors, alcohol and tobacco consumption and total delay in pulmonary tuberculosis treatment. Recife, 2001--2003. ::: Total delay in treatment --------------------- -------------------------- ------ ----- ------ ------ -------------- -------- Sex Male 470 63.4 271 36.6 1.00 \- \- Female 216 59.3 148 40.7 1.18 0.91 -- 1.53 0.1884 Total 686 \- 419 \- \- \- \- Age (years) 0.1810 18--39 344 60.2 227 39.8 1.00 \- \- 40--59 280 65.4 148 34.6 1.24 0.96 -- 1.61 0.0949 ≥ 60 62 58.5 44 41.5 0.92 0.61 -- 1.41 0.7349 Total 686 \- 419 \- \- \- \- Alcohol consumption 0.5575 None 338 60.7 219 39.3 1.00 \- \- Light drinker 214 61.5 134 38.5 1.03 0.78 -- 1.36 0.8075 Heavy drinker 83 65.9 43 34.1 1.25 0.83 -- 1.87 0.2800 Total 635 \- 396 \- \- \- \- Cigarette smoking 0.0010 Never smoked 213 68.9 96 31.1 1.00 \- \- Given up smoking 297 56.7 227 43.3 0.58 0.43 -- 0.79 0.0005 Ever smoked 175 65.3 93 34.7 0.84 0.59 -- 1.20 0.3545 Total 685 \- 416 \- \- \- \- ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Socio-economic factors and total delay in pulmonary tuberculosis treatment. Recife, 2001--2003. ::: Total delay in treatment ------------------------------------- -------------------------- ------ ----- ------ ------ -------------- -------- Employment Employed 282 57.3 210 42.7 1.00 \- \- Unemployed 404 65.9 209 34.1 1.43 1.12 -- 1.83 0.0035 Total 686 \- 419 \- \- \- \- Literacy Yes 540 61.6 337 38.4 1.00 \- \- No 146 64.0 82 36.0 1.11 0.82 -- 1.50 0.4951 Total 686 \- 419 \- \- \- \- Income of Head of household (MW) 0.7356 ≥ 5 22 59.5 15 40.5 1.00 \- \- 2 to \< 5 75 66.4 38 33.6 1.34 0.62 -- 1.88 0.4461 1 to \< 2 184 63.7 105 36.3 1.19 0.59 -- 2.40 0.6176 \< 1 327 61.5 205 38.5 1.08 0.55 -- 2.14 0.8086 Total 608 \- 363 \- \- \- \- Civil status 0.9600 Married 251 62.4 151 37.6 1.00 \- \- Divorced, widowed or separated 64 62.7 38 37.3 1.01 0.64 -- 1.58 0.9544 Single 341 61.7 212 38.3 0.96 0.74 -- 1.26 0.8078 Total 656 \- 401 \- \- \- \- Number of individuals per Household 0.3897 1 -- 2 160 65.3 85 34.7 1.00 \- \- 3 -- 5 345 60.2 228 39.8 0.80 0.59 -- 1.09 0.1700 ≥ 6 165 61.6 103 38.4 0.85 0.59 -- 1.22 0.3803 Total 670 \- 416 \- \- \- \- ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Contact with tuberculosis and total delay in pulmonary tuberculosis treatment. Recife, 2001--2003. ::: Total delay in treatment ----------------------------------------------- -------------------------- ------ ----- ------ ------ -------------- -------- Knows someone with tuberculosis Yes 305 63.0 179 37.0 1.00 \- \- No 375 61.4 236 38.6 0.93 0.72 -- 1.19 0.5782 Total 680 \- 415 \- \- \- \- Recent contact with someone with tuberculosis 0.8554 \< 1 year 151 63.4 87 36.6 1.00 \- \- ≥ 1 year 128 61.8 79 38.2 0.93 0.63 -- 1.37 0.7263 Not known 375 61.4 236 38.6 0.91 0.67 -- 1.24 0.5769 Total 654 \- 402 \- \- \- \- Household contact with a case of tuberculosis Yes 104 62.7 62 37.3 1.00 \- \- No 198 63.1 116 36.9 1.01 0.68 -- 1.50 0.9301 Total 302 \- 178 \- \- \- \- ::: ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Clinical features and symptoms and total delay in pulmonary tuberculosis treatment. Recife, 2001--2003. ::: Total delay in treatment ----------------------------- -------------------------- ------ ----- ------ ------ -------------- -------- Previous TB treatment Yes 147 60.5 96 39.5 1.00 \- \- No 537 62.5 322 37.5 1.08 0.81 -- 1.45 0.5666 Total 684 \- 418 \- \- \- \- Productive cough Yes 418 62.4 252 37.6 1.00 \- \- No 259 61.1 165 38.9 0.94 0.73 -- 1.21 0.6655 Total 686 \- 419 \- \- \- \- Haemoptysis Yes 94 56.6 72 43.4 1.00 \- \- No 585 62.9 345 37.1 1.29 0.92 -- 1.81 0.1256 Total 679 \- 417 \- \- \- \- Loss of weight Yes 451 59.6 306 40.4 1.00 \- \- No 225 68.8 102 31.2 1.49 1.13 -- 1.97 0.0041 Total 676 \- 408 \- \- \- \- HIV co-infection 0.6964 Yes 219 63.8 124 36.2 1.00 \- \- No 18 64.3 10 35.7 0.98 0.43 -- 2.19 0.9631 Not known 449 61.3 284 38.7 0.87 0.39 -- 1.92 0.7467 Total 686 \- 418 \- \- \- \- Acid fast bacilli in sputum Yes 477 64.1 267 35.9 1.00 \- \- No 102 63.4 59 36.6 0.96 0.67 -- 1.37 0.8557 Total 579 \- 326 \- \- \- \- ::: ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Access to health care and characteristics of health service (HS) and total delay in pulmonary tuberculosis treatment. Recife, 2001--2003. ::: Total delay in treatment ---------------------------------- -------------------------- ------ ----- ------ ------ -------------- -------- Numbers of health units only 1 176 63.1 103 36.9 1.00 \- \- ≥ 2 493 61.2 312 38.8 0.92 0.69 -- 1.22 0.5859 Total 669 \- 415 \- \- \- \- HS in same district of residence Yes 431 61.3 272 38.7 1.00 \- \- No 255 63.6 146 36.4 1.10 0.85 -- 1.42 0.4523 Total 686 \- 418 \- \- \- \- HS in neighborhood of residence Yes 111 63.4 64 36.6 1.00 \- \- No 575 61.9 354 38.1 0.93 0.67 -- 1.30 0.7012 Total 686 \- 418 \- \- \- \- Residence in areas of FHP visits Yes 258 60.4 169 39.6 1.00 \- \- No 406 63.4 234 36.6 1.13 0.88 -- 1.46 0.3196 Total 664 \- 403 \- \- \- \- Health District of treatment 0.0001 HD II 170 52.8 152 47.2 1.00 \- \- HD I 57 62.6 34 37.4 1.49 0.92 -- 2.41 0.0960 HD III 139 72.8 52 27.2 2.39 1.62 -- 3.51 0.0000 HD IV 111 61.7 69 38.3 1.43 0.99 -- 2.08 0.0553 HD V 172 67.7 82 32.3 1.87 1.33 -- 2.64 0.0003 HD VI 37 55.2 30 44.8 1.10 0.64 -- 1.87 0.7170 Total 686 \- 419 \- \- \- \- ::: Table [6](#T6){ref-type="table"} presents the results of the multivariable analysis. All variables that were significant in the univariable analysis remained so in the multivariable analysis. The factors significantly associated with delay were: being unemployed (OR = 1.4; 95% CI: 1.09 -- 1.81); not having weight loss as one of the symptoms (OR = 1.54; 95% CI: 1.16 -- 2.04); and treatment in two of the six districts (OR = 2.34; 95% CI: 1.57 to 3.47 and OR = 1.92 95% CI: 1.35 to 2.73). The unacceptable delay attributable to each of these factors was: unemployment, 28.6%, not having weight loss, 35.1%, and being treated at health districts III and V, 57.3% and 47.9% respectively. ::: {#T6 .table-wrap} Table 6 ::: {.caption} ###### Final model of characteristics of patients and health services and total delay in pulmonary tuberculosis treatment. Recife, 2001--2003. ::: Adjusted OR 95% CI P value ---------------------- ------------- -------------- --------- Employment Yes 1.00 \- \- No 1.40 1.09 -- 1.81 0.0079 Loss of weight Yes 1.00 \- \- No 1.54 1.16 -- 2.04 0.0028 Health District (HD) 0.0001 HD II 1.00 \- \- HD I 1.35 0.83 -- 2.21 0.2189 HD III 2.34 1.57 -- 3.47 0.0000 HD IV 1.39 0.95 -- 2.04 0.0863 HD V 1.92 1.35 -- 2.73 0.0002 HD VI 0.97 0.56 -- 1.68 0.9395 \* variables excluded from model: sex, age and haemoptysis because they were not statistically significant. ::: Discussion ========== In our setting there was no statistically significant association between unacceptably long total delays and living in a different district from the unit of treatment, negative smears, older age, female sex and alcoholism. We found independent, statistically significant association with long delays and being unemployed, having loss of weight as one of the symptoms, and living in two of the six health districts. One of the limitations of the study is that we were not able to separate patient delay from health services delay. It is, therefore, possible that some associations are specific to one of these kinds of delay. Some of the associations absent (for example with being a heavy drinker) \[[@B8]\] may be due to methodological limitations. We did not use the CAGE standard questionnaire to ascertain alcoholism. Another limitation of the study of the impact of the Family Health Program is that we did not ask whether there was a visit from the FHP after the onset of the symptoms or even whether symptoms were addressed during a visit. In contrast to results reported in the literature, it was found that a history of contact with a patient with tuberculosis did not decrease delays. We expected this to act as a proxy variable for knowledge about the disease \[[@B5],[@B7],[@B9]\]. However, none of the variables relating to previous contact with a patient with TB was associated with treatment delay. Again in contrast to other studies \[[@B2],[@B7],[@B16]\], being smear-positive was not associated with a shorter delay and most of those patients (64.1%) did not start early treatment and continued to contribute to transmission. The only symptom that was associated with shorter delays was loss of weight. Productive cough, even when associated with haemoptysis, did not lead to shorter delays in the population studied, in contrast to the findings of other authors \[[@B8],[@B14],[@B16]\]. Our findings, taken as a whole, may suggest that it is not specific knowledge of tuberculosis per se, but rather a general awareness of health and attitude towards prevention and early care that leads to shorter delays. If one assumes that the risk of delay associated with literacy, age and sex is mediated by difficulty of access, one could say that, in this case, access to care does not appear to be a problem. In contrast to results reported in the literature \[[@B2],[@B4],[@B8],[@B13]-[@B15]\], we found no association with age, sex or literacy. Associations have been reported between travelling time, distance to health service etc. \[[@B7],[@B8],[@B17]\] in the literature. In our study, factors directly relating to health care, such as whether the treatment unit was in the same neighborhood of residence, showed no association. This lack of association, however, could also be interpreted to mean that access is not difficult in this urban setting, but that ease of access did not decrease delays, as those attending health units outside their neighborhood or district of residence show the same length of delay. Decentralization has been introduced to improve access and we expected those patients being treated in their own neighborhood of residence to exhibit shorter delays. Another of the health service variables, however, did prove to be significant: patients attending health units located in districts III and V showed longer delays. This did not disappear in the multivariable analysis, suggesting that the delays were not a result of characteristics of patients but of the way the health service is organized. Further analysis (not presented here) showed that, in district V, this was mainly caused by delays in diagnosing smear-negative patients: only 10% had a delay of less than 60 days, compared with roughly 30% of the smear-positive cases. This is, therefore, more likely to reflect the internal organization of care (in this specific case, of diagnosis of smear-negative cases) rather than a patient\'s ability to appear at a health care unit. In summary, 62% of patients started treatment after 60 days and half after 90 days. In our setting, delays do not appear to be caused by difficulty of access, age, sex or alcoholism. In this urban setting, delays seem to be linked to unemployment and the general attitude towards health, and reflect the way the health services are organized. Conclusion ========== There is a need to monitor the impact of the FHP and modify staff training to include active identification of symptoms that may indicate tuberculosis, and to increase population awareness of tuberculosis symptoms (building on the fact that the population already recognizes weight loss), with emphasis on developing general health awareness. A review of practices in the districts with short and long delays, to improve health services organization, should also be carried out. List of abbreviations ===================== TCP Tuberculosis control Programme FHP Family Health Programme TB Tuberculosis HIV Human Immunodeficiency virus Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= MDPS, OMSD, ARLC, NLCLS, AMK and URM trained the field workers and participated in data entry and analysis. MDPS, MFPMA, CB, RAAX, URM, WVS and LCR interpreted the results and drafted the manuscript. MFPMA, RAAX, MDPS, LCR conceived of the study, and participated in its design and coordination. All authors read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2458/5/25/prepub>	PubMed Central	2024-06-05T03:55:55.849505	2005-3-18	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084352/", "journal": "BMC Public Health. 2005 Mar 18; 5:25", "authors": [ { "first": "Martinho APS", "last": "dos Santos" }, { "first": "Maria FPM", "last": "Albuquerque" }, { "first": "Ricardo AA", "last": "Ximenes" }, { "first": "Norma LCL", "last": "Lucena-Silva" }, { "first": "Cynthia", "last": "Braga" }, { "first": "Antônio RL", "last": "Campelo" }, { "first": "Odimariles MS", "last": "Dantas" }, { "first": "Ulisses R", "last": "Montarroyos" }, { "first": "Wayner V", "last": "Souza" }, { "first": "Alexandre M", "last": "Kawasaki" }, { "first": "Laura C", "last": "Rodrigues" } ] }
PMC1084353	Background ========== The first probable case of severe acute respiratory syndrome (SARS) in Taiwan was identified on March 14, 2003 \[[@B1]\]. The epidemic became elevated at the end of April and reached its peak in May and June. Finally, the epidemic ended when Taiwan was officially removed from the World Health Organization (WHO) list of SARS affected countries on July 5, 2003 \[[@B1]-[@B3]\]. The fear of SARS spread over all of Taiwan and was mainly directed at those regional hospitals and medical centers where outbreaks had occurred and where SARS patients were being treated. People\'s fears originated with the novelty of the disease, its rapid nosocomial transmission, and the apparent vulnerability of hospitals and health care workers. Many people started to wonder whether the overwhelming fears of SARS directed at the regional hospitals and medical centers changed a patient\'s care seeking behavior and even physicians\' treatment patterns, and thereby had possible health consequences, especially for those whose survival or state of health depended on routine and/or continuous medical care. Delivery of a child is a classic example. Under normal circumstances, distance and quality are believed to be two major determinants in a patient\'s choice for obstetrics care \[[@B4]-[@B7]\]. In this study, we sought to estimate the influence of people\'s fears of SARS on people\'s choice of provider, mode of child delivery, and length of hospital stay before, during and after the delivery. The people\'s fears include fears of patients and fears of doctors as choice of mode of delivery and length of stay tend to be a joint decision. The fears of doctors to SARS may also reduce cesarean section rate and shorten the length of stay for lowering the possibility of acquiring SARS. Considering the fact that hospital levels \[[@B5],[@B8]-[@B15]\], (mode of delivery \[[@B16],[@B17]\], and length of maternity stay \[[@B18]-[@B25]\] are significantly associated with postpartum maternal and neo-natal health status, the question whether the fears of SARS led patients to transfer from a more advanced hospital to a less advanced hospital becomes an important quality of care concern. Health consequences resulting from these behavioral changes due to people\'s fears of SARS should not be overlooked \[[@B26]-[@B29]\]. The results could provide public health agencies an important reference when assessing the consequences of the SARS epidemic on quality of care. So that when SARS re-emerges or other similar new infectious disease emerges, it can guide obstetricians and public health professionals to prevent avoidable health consequences because of people\'s fears concerning these new and strongly infectious diseases. Methods ======= Obstetrics services under the NHI program ----------------------------------------- Since its implementation in March 1995, the National Health Insurance (NHI) program provides a mandatory and comprehensive universal health care coverage for all Taiwanese residents. For obstetrics care, a wide range of services including pre-natal, delivery, and neo-natal care are extensively covered under the NHI program. Both vaginal and cesarean deliveries are covered and no co-insurance is required for child delivery. The program offers patients complete freedom of choice among providers and methods of delivery. In order to contain the in-patient costs for Western medical services, the NHI program has instituted the case-payment method for 50 clearly defined medical conditions, and they include both vaginal delivery and cesarean delivery. Case payment is similar to the DRG (diagnosis-related groups) payment system in the U.S. It bundles itemized medical services essential for each condition and provides financial incentives for a more efficient delivery process. The case payment system reimburses the provider on a per-case basis, and the payment varies by accreditation level of the provider. It also assumes that the complexity or quality of care also varies by provider level. The current accreditation system categorizes medical institutions into 4 levels, from most advanced to most basic: medical center, regional hospital, district hospital, and clinic. The accreditation criteria include infrastructure, capacity, manpower, volume, management and administrative processes. For vaginal delivery, the case payment ranges from NT\$17,420 per case in medical centers to NT\$15,100 per case in clinics. For cesarean delivery, the range is larger (from NT\$32,330 per case in medical centers to NT\$27,170 per case in clinics). Furthermore, the National Health Insurance Law prohibits its contracted providers from any direct/extra billing for any service covered by the NHI program. Study design and statistical analyses ------------------------------------- This study conducted a population-based descriptive analysis of changes in market share, cesarean rate, and length of maternity stay, for each of the four levels of provider, before, during and after the SARS epidemic. This study retrieved all 448,365 NHI in-patient claims for childbirth from January 01, 2002 to December 31, 2003. Vaginal deliveries were those coded with the NHI Case Payment-DRG 0373A and 0373C. Cesarean deliveries were those coded with the NHI Case Payment-DRG 0371A and 0373B. Each claim cites the patient, provider, diagnoses, treatment procedures, mode of delivery, and admission and discharge dates. A unique provider identification number was used to link the NHI provider file, which identifies provider accreditation level. As the SARS epidemic started in March, elevated in May, and finally ended in July, we divided the study period into 5 sub-periods: pre-SARS period from January 2002 to February 2003 (t~0~), initial SARS period from March 2003 to April 2003 (t~1~), peak SARS period from May 2003 to June 2003 (t~2~), final SARS period from July 2003 to August 2003 (t~3~) and post-SARS period from September 2003 to December 2003 (t~4~). Monthly average estimates for market share, cesarean rate, length of maternity stay, all by accreditation level were calculated and expressed in numbers and percentages. Mean differences in average cesarean rate and average length of maternity stay between the pre-SARS and the peak periods (t~2~- t~0~), and between the pre-SARS and post-SARS periods (t~4~- t~0~) were determined by 2-tailed t tests. The data were managed and analyzed by using SAS, Version 8.2. All analyses were tested for significance by using an alpha of .05. Since no human participants were involved, no IRB approval is necessary. The confidentiality assurances were addressed by abiding the data regulations of the Bureau of National Health Insurance, which stipulated the data only for use of this research. Results ======= Changes in market share ----------------------- Figure [1](#F1){ref-type="fig"} shows the changes in the market share of delivery services by accreditation level of provider in the five sub-periods. In terms of market share, before the SARS epidemic, although clinics still had the largest market share, they were slowly losing their share to their powerful competitors, the medical centers and the regional hospitals. However, people\'s fears of SARS reversed this situation during the SARS epidemic, as patients switched to clinics during the SARS epidemic to deliver their child, rather than in the larger academic medical centers and regional hospitals. A dramatic drop in market share was observed for both the medical centers and the regional hospitals during the SARS epidemic. While at the same time both district hospitals and clinics gained market share. In addition, Table [1](#T1){ref-type="table"} shows the changes in monthly average number of deliveries, and the market share of delivery services by accreditation level of provider in the 3 key sub-periods (pre-SARS, peak-SARS, and post-SARS periods). Generally speaking, over the past several years Taiwan has experienced a constant decreasing fertility rate. On average, there were 2,189 fewer deliveries each month during the peak SARS period than during the pre-SARS period. Compared to the pre-SARS period (t~0~), medical centers and regional hospitals dropped 5.2% and 4.1% in market shared during the peak SARS period (t~2~), but district hospitals and clinics increased 2.1% and 7.1%, respectively. Although medical centers and regional hospitals regained some of their lost market share during the post-SARS period (t~4~), they still did not fully regain their market shares at the pre-SARS level (medical centers: 2.0%; regional hospitals: 0.8%). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Trends in market shares of childbirth service in Taiwan by provider\'s accreditation level, January 2002--December 2003. ::: ![](1471-2458-5-30-1) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Average Volumes of Childbirth Service in Taiwan by Provider\'s Accreditation Level and Time Period, January 2002--December 2003. ::: ------------------------- ------------------------------------------ -------------------------------------- -------------------------------------- ---------------- ---------------- ------- --------- ------- --------- ------- t~0~ t~2~ t~4~ t2-t0 t4-t0 Pre-SARS Period (Jan 2002\~Feb 2003) Peak SARS Period (May\~Jun 2003) Post-SARS Period (Sep\~Dec 2003) Difference Difference Accreditation Level No. % No. % No. % No. % No. % Medical Center 3,044 15.9 1,820 10.7 2,574 13.9 -1,225 -5.2 -471 -2.0 Regional Hospital 4,731 24.7 3,501 20.6 4,415 23.8 -1,230 -4.1 -316 -0.8 District Hospital 5,067 26.4 4,853 28.5 5,156 27.8 -214 2.1 90 1.4 Clinic 6,346 33.1 6,826 40.2 6,382 34.4 480 7.1 36 1.4 Total 19,188 100.0 16,999 100.0 18,527 100.0 -2,189 -661 ------------------------- ------------------------------------------ -------------------------------------- -------------------------------------- ---------------- ---------------- ------- --------- ------- --------- ------- ::: Changes in cesarean rate and length of maternity stay ----------------------------------------------------- Table [2](#T2){ref-type="table"} compares the changes in monthly average cesarean section rates and length of stay (LOS) in the pre-, peak- and post-SARS periods (t~0~, t~2~, and t~4~). Compared to the pre-SARS period, we observed only a marginal decrease in overall cesarean rate (1.0%), but a significantly larger increase in cesarean rate in medical centers (2.2%) and significantly decrease in clinics (1.9%) during the peak SARS period (t~2~). One plausible explanation is that as normal or less complicated deliveries shifted to lower level hospitals or clinics, those cases that remained in the medical centers tended to be complicated ones which required cesarean sections. Hence, although women shifted their place of delivery from higher level hospitals to lower level hospitals/clinics due to a greater risk of exposure to SARS at these higher level hospitals, their choice of delivery method did not seem to change. In addition, the average cesarean section rate in medical centers returned to the pre-SARS level during the post-SARS period while the cesarean section rates in lower level hospitals and clinics showed the opposite. The cesarean section rates in regional, district hospitals and clinics dropped significantly during the post-SARS period from their levels before the SARS epidemic. Furthermore, Table [2](#T2){ref-type="table"} and Figure [2](#F2){ref-type="fig"} present changes in average LOS among different provider levels. As expected, in order to reduce their potential risk of exposure to SARS, people tried to minimize their maternal stay in a hospital as much as possible. The overall average length of stay decreased from 3.40 days to 3.25 days from the pre-SARS period (t~0~) to the peak SARS period (t~2~), and then returned to 3.39 days after the SARS epidemic (t~4~). More specifically, significant reductions in average LOS were observed in all hospital levels except for clinics at t~2~. Average length of stay was shortened by 0.21 days in medical centers (5.6%), 0.21 days in regional hospitals (5.8%), and 0.13 days in district hospitals (3.8%). Average LOS in clinics remained basically unchanged. However, as soon as the SARS epidemic ended, average LOS in most hospitals and clinics not only returned to the pre-SARS level, but became slightly longer than it was prior to SARS, which is not significant statistically speaking. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Average Cesarean Section Rates and Length of Maternity Stay in Taiwan by Provider\'s Accreditation Level and Time Period, January 2002--December 2003. ::: t~0~ t~2~ t~4~ t2-t0 t4-t0 ------------------------- ---------- ---------- ---------- ------------------- ----------------- C-Section Rate % % % % (S.E) % (S.E) Medical Center 36.1 38.3 36.1 2.2 (0.7) \* 0.0 (0.6) Regional Hospital 32.8 32.3 31.7 -0.5 (0.6) -1.1 (0.5) \* District Hospital 34.3 33.2 30.5 -1.1 (0.5) \* -3.7 (0.4) \\ Clinic 34.5 32.6 31.3 -1.9 (0.5) \\ -3.2 (0.4) \\ Total 34.3 33.3 31.9 -1.0 (0.4) \* -2.4 (0.3) \\ Average LOS Day Day Day Day (S.E) Day (S.E) Medical Center 3.73 3.52 3.74 -0.21 (0.05) \\ 0.01 (0.02) Regional Hospital 3.65 3.44 3.67 -0.21 (0.05) \\ 0.02 (0.03) District Hospital 3.39 3.26 3.36 -0.13 (0.04) \\ -0.03 (0.03) Clinic 3.07 3.07 3.08 0.00 (0.04) 0.01 (0.03) Total 3.40 3.25 3.39 -0.15 (0.04) \\ -0.01 (0.02) \P \< 0.05, two-tailed test; \\*P \< 0.01, two-tailed test ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Trends in length of maternity stay in Taiwan by provider\'s accreditation level, January 2002--December 2003. ::: ![](1471-2458-5-30-2) ::: Discussion ========== Our results, based on a population-based study, demonstrate that if the use of medical care required was essential and could not be deferred, such as childbirth, the fears of SARS, including both fears of patients and fears of doctors, had a significant influence on patients\' preference for provider and on their length of in-patient stay. However, it did not necessarily influence their choice of therapies during the SARS epidemic. Even though the SARS influence upon a patient\'s decision regarding the length of stay ended with the end of the epidemic, the fear of SARS remained influential on how people chose their place of delivery. During the SARS epidemic, large amounts of patients shifting from more advanced hospitals to less advanced hospitals and substantially, reducing their length of maternity stay simply out of fear could result in a serious concern for quality of care, especially for a patient\'s accessibility to quality care and the continuity of that care. For example, patients transferring to a less advanced hospital could severely compromise their access to more sophisticated technologies and interrupt their continuity of care. This in turn could endanger the lives of the mothers and their babies \[[@B5],[@B8]-[@B14]\]. (Significantly shorter length of stay increases the risk of a premature discharge and compromise the quality of care, which again could lead to adverse health outcomes for both mother and baby \[[@B23]-[@B25],[@B30]\]. Furthermore, not only the physicians who may directly encounter with SARS patients need to be alert and better trained for this type of outbreak, obstetricians should also be aware and extremely cautious about suddenly substantial patient shift and strong demand of shorter length of maternity stay posed by the public\'s fear of such an outbreak. As these sudden changes in people\'s behaviors are likely to hinder patient\'s continuity of care and cause adverse maternal and neonatal health consequences, it is essential for obstetricians to be well prepared to deal with these consequences during an outbreak. A few study limitations should be noted. First, due to data and time constraints, this study only shows the immediate impacts of the fears of SARS on the shifting of patients from one facility to another, as well as their length of stay. This study cannot confirm whether these changes led to any long term adverse maternity or perinatal outcomes. If in the post-SARS period, mother\'s and child\'s morbidity and mortality remained the same, the residual increase of child births in clinics and district hospitals would be a positive impact since it reduces health care costs without compromising health outcomes. Future research with a longer post-SARS observational period and more detailed maternal and perinatal outcome information could help to advance our knowledge in this regard. Second, the significant amount of patient shifting observed among provider levels suggests that the fear of SARS changed people\'s preference as to their choice of provider during the SARS epidemic. Whether this influence will persist remains to be seen. Third, since we only focused on childbirth in this study, the results may not be generalizable to other medical conditions. Patient behavior may very well differ for medical conditions/diseases with different levels of severity and/or medical urgency. Furthermore, the influence of the fear of SARS on other important patient behaviors, such as their decision to seek care or not during the SARS epidemic if their medical needs could be deferred or suspended to a later time, remains uncertain. There are many questions that remain to be answered concerning the possible impacts of SARS on various aspects of health care. Conclusion ========== In terms of policy implications, the medical centers and regional hospitals in Taiwan took on the majority of the responsibility in caring for the more severe SARS cases, and by doing so lost their market share to district hospitals and clinics because of people\'s fear of SARS. The BNHI negotiated with these hospitals to compensate them for their loss of revenue during the SARS epidemic, and bring it up to the previous year\'s level. It is expected that by minimizing a hospital\'s financial loss it will increase that hospitals\' willingness to admit SARS patients and secure people\'s access to proper medical care in the likelihood of a re-emergence of a SARS epidemic or any other pandemic. Strategies which help to restore people\'s confidence in those hospitals that have admitted SARS patients should be part of the long term solution. Finally, while internationally all of the global attention is focused on the direct causalities of SARS, serious quality of care concerns resulting from people\'s behavioral changes due to their fears of SARS should not be overlooked. Taiwan\'s experience could provide valuable lessons to other countries in assessing full impacts of the SARS epidemic and help to minimize adverse health consequences when SARS or other similar pandemic emerge. List of abbreviations ===================== SARS: severe acute respiratory syndrome NHI: the National Health Insurance LOS: length of stay C-section: cesarean section No: number Jan: January Feb: February Mar: March Apr: April Jun: June Jul: July Aug: August Sep: September Oct: October Nov: November Dec: December Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= CHL planned the study and supervised all aspects of its implementation. NH assisted with the study and led the writing. HJC synthesized analyses and contributed to the writing of the article. YJH and YJC contributed to the design, analyzed the data, commented on the interpretation of the results. MCW assisted with the study and completed the statistical analyses. All authors helped to conceptualize ideas, interpret findings, and review drafts of the manuscript. All authors read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2458/5/30/prepub> Acknowledgements ================ The project is supported by the National Scientific Council grant NSC-93-2320-B-010-037.	PubMed Central	2024-06-05T03:55:55.853754	2005-4-4	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084353/", "journal": "BMC Public Health. 2005 Apr 4; 5:30", "authors": [ { "first": "Cheng-Hua", "last": "Lee" }, { "first": "Nicole", "last": "Huang" }, { "first": "Hong-Jen", "last": "Chang" }, { "first": "Yea-Jen", "last": "Hsu" }, { "first": "Mei-Chu", "last": "Wang" }, { "first": "Yiing-Jenq", "last": "Chou" } ] }
PMC1084354	Background ========== Personal Digital Assistants (PDAs) or handheld computers are now commonly used devices in clinical medicine and other professions where immediate access to information is required away from the comforts of a traditional office and desk top computer \[[@B1]\]. Several specialities of clinical medicine are using PDAs for the delivery of information at the point of care including anaesthesia \[[@B2]\], surgery \[[@B3]\], paediatrics \[[@B4]\], general practice\[[@B5]\] and obstetrics \[[@B6]\]. PDAs have also been used to collect patient information and improve clinical records, for administrative functions such as electronic prescribing \[[@B7]\], coding and tracking \[[@B8]\], in research projects. \[[@B9]-[@B11]\] and in medical education including the monitoring of training \[[@B12]\]. This paper describes the piloting of a PDA in the on call health protection service. Health protection is a part of public health and is defined as those public health activities intended to protect individuals, groups, and populations from infectious diseases, environmental hazards such as chemical contamination, and from radiation \[[@B13]\]. The on call service for health protection varies in different parts of England. The UK Faculty of Public Health in collaboration with other UK professional and governmental bodies including the Health Protection Agency (HPA), Department of Health, Association of Directors of Public Health and Public Health Medicine Environmental Group has recently published recommendations for the organisation and delivery of out of hours service to protect the public\'s health \[[@B14]\]. In Hertfordshire, as in many parts of the UK, the on call service is provided by public health registrars and communicable disease control nurses as the first tier of response backed by public health/health protection consultants as the second tier. The first on call was previously required to carry a large bag of papers containing both local and national guidance on the management of common cases/incidents. Managing incidents may require the responder to visit incident sites located away from a normal office setting where access to information is easy. The reasons for developing an electronic on call pack in Hertfordshire were to improve portability by reducing the size and weight of the on call pack and to enable easier updating. The training group of the Local and Regional Services Division of the HPA has developed a number of initiatives to facilitate and improve the quality of training provided for staff of the Agency. This on call device may contribute to the effort of the Agency in delivering this function by providing an additional option for those on call to be more effective. The aim of this study was to develop and pilot the use of a portable electronic on call pack for use by the public health on-call team in Hertfordshire to deliver the health protection function. The advantages, disadvantages and feasibility of using this system were assessed. Methods ======= Developing the on-call pack --------------------------- The East and North Hertfordshire paper based on call pack was converted into electronic text files using Microsoft Frontpage^©^in January 2003. Software and hardware --------------------- The on call pack was loaded from the desktop computer where it was initially developed unto a pocket PC based IPAQ^©^PDA by synchronising. All guidelines were in portable document format (.pdf) and could be view ed with an Adobe Reader^©^for pocket PC. Local information files were all Microsoft Word^©^documents and could therefore be easily updated. The choice of Pocket PC operating system was based on ease of adaptation of documents already prepared for use with Microsoft software. The process of adaptation entails automatically converting files from a PC based version, for instance, Microsoft Word documents were converted to a pocket PC equivalent and where appropriate the formatting and content were changed. A Palm based on-call pack is currently under development. Structure and content --------------------- The on-call pack consists of an index page, which can either be on a laptop or a PDA. The main index is divided into a disease information column and an administrative information column (figure [1](#F1){ref-type="fig"}). Each disease page was divided in to three sections, 1. on-call action summarising key steps to be taken when responding to a call (figure [2](#F2){ref-type="fig"}), 2. additional information for managing cases/incidents and 3. links to further national or local guidelines for the management of the specific condition that can be accessed without connecting to the internet (figure [3](#F3){ref-type="fig"}). The administrative information section consists of contact details, on call rotas for different organisations and how to deal with the media. A user guide was included linked to the administrative information column. The main differences between the PDA and laptop versions of the on call pack include the smaller screen on the PDA and therefore bigger font required and different versions of the software used. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Structure of the on call pack as viewed on a PDA ::: ![](1471-2458-5-35-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Immediate on call action for meningococcal disease ::: ![](1471-2458-5-35-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### UK National guidelines for meningococcal disease ::: ![](1471-2458-5-35-3) ::: Sources of information ---------------------- The information in the on-call pack is based primarily on data already in the paper version of the pack in three health protection units in England (Hertfordshire, Cambridgeshire and Bedfordshire). We used national guidelines for the management of communicable disease cases and outbreaks, chemical and radiological incidents and emergency planning guidance as published on the HPA \[[@B15]\], National Electronic Library for Infections \[[@B16]\] and UK Department of Health\'s \[[@B17]\] websites. We obtained further information from textbooks commonly used for the public health management of communicable diseases including Communicable Disease Control Handbook \[[@B18]\] and Chin\'s Communicable Disease Control Manual \[[@B19]\]. Although the authors assessed each source prior to inclusion no formal evaluation of the sources was conducted. Updating information in the on-call pack ---------------------------------------- The administrative information documents such as the on-call rota can be updated by either replacing the word document or cutting and pasting content into the document. The content of disease-based pages can be updated with any text editor such as Microsoft Notepad^©^or with web editing software such as Frontpage^©^or Dreamweaver^©^. Arrangements for updating the content of the on-call pack have not been agreed and are currently undertaken on an ad-hoc basis. Ideally a formal process for regular updating should be implemented because the current ad-hoc system is not satisfactory. Pilot ----- The use of the on-call pack was evaluated in a pilot study over the first year of its development initially as a tool on a laptop computer and subsequently on a PDA. The components of the evaluation includes a qualitative and a quantitative aspect. Qualitative ----------- ### District The on-call pack was presented at a training meeting for trainees and consultants on the district on-call rota held in February 2003. The presentation was used to introduce the on-call pack on a laptop computer and to obtain feedback regarding its potential usefulness and whether individuals on the rota were willing to use the pack as an alternative to the current system. Attendees asked questions and gave comments at the end of the presentation which were used to improve the next version of the pack. ### Regional The on call pack was presented at a regional training event held in April 2004 for general public health consultants and directors of public health. The content of the pack was explained and attendees were allowed to review the PDA. A questionnaire was given asking for their views on how useful they thought it would be. The evaluation questionnaire was based on a Likert scale of good, fair, neutral, weak and poor and free text comments. Quantitative ------------ This aspect of the evaluation was carried out over a period of five weeks, each person on duty was given the on-call pack at the beginning of the week. Three PDAs were used so individuals had to return the device at the end of the on-call period. Each person on-call was given an evaluation questionnaire to be completed at the end of the week. Questions were asked to comment on the frequency of use, type of incidents, ease of access, quality of information and portability. Responses were collated and analysed in a spreadsheet. Quantitative measures were summarised using an average score and range, while free text comments were summarised in a table. Results ======= Qualitative ----------- ### District Several comments were received from the district training day. The main suggestions were; firstly, there is a need for a back up facility as with any electronic device as failures can happen. Secondly, a search function to identify key documents without the need to scroll through several pages would be helpful. Thirdly, there were concerns about confidentiality issues if a PDA or laptop was to be used for recording patient identifiable information and fourthly, participants were concerned about how the on-call pack would be updated. ### Regional Twenty individuals attended the regional training meeting and 12 completed evaluation forms. Eleven of the 12 respondents scored the on-call pack \"good\", while one gave it a \"fair\". Comments included a suggestion that training on using the on call pack should be provided for anyone going on the rota and that contents should be available on a web page to synchronise with individual PDA\'s. Potential difficulties highlighted include problems with batteries running down, the need for a paper backup and difficulty in reading some documents with very small font. Quantitative ------------ Five individuals trialled the PDA based on-call pack over a five week period, four of whom completed and returned evaluation forms (participants A, B, C, D). The mean number of calls answered was 1.5 (range 0--3 calls). Incidents tackled during the week varied and included both infections and chemical incidents. Table one summarises the response of the participants. Assessment was based on a Likert scale of one to five where one was poor, three was same as paper version and five was very good. All three respondents gave a score of good for ease of use and very good for portability. The quality of content was rated four out of five by all three respondents. The coordinator of the evaluation had difficulties with handover of the limited number of PDA\'s and with synchronising PDAs to the desktop computer. Discussion ========== We have developed and piloted the use of a handheld computer based on-call pack for health protection out of hours duty. As far as we are aware, this is the first use and evaluation of a hand held device for public health on call duties. The idea of using an electronic portable pack was considered good and acceptable by most individuals on the rota and the information was found to be easy to access and use. PDAs have been found to be useful in delivering information at the point of use in several specialities of clinical medicine\[[@B1]\]. The findings of this study indicate that many of the benefits of using PDAs in clinical medicine could be applicable to public health practice. Availability of information is especially important because of the increasing volume of knowledge and the expectations that guidelines will be followed. Other aspects of the basic functions of a PDA can provide added benefit to the reference function intended in our application. These functions include the address book, scheduler, to do list and memo \[[@B1]\]. There is also a voice recorder that can be used to record notes for subsequent typing. There are several advantages of using a PDA identified from this and other studies. Portability was found to be one of the main positive attributes of the on-call pack as users found the device more convenient to carry compared to the paper based version. This was the biggest strength of the device given the amount of information required for dealing with incidents. An advantage cited by users is the ease of access to information. This has been reported in other studies\[[@B20],[@B21]\] It would be useful to compare the outcome of incidents and assess whether using a handheld device improves outcome. We have not conducted an outcome evaluation due to the long-term nature of results of public health measures and resource limitation. The results of the evaluation should be interpreted cautiously due to the small sample size. Ideally each person on call should be provided with an individual PDA but cost considerations are important. Each PDA used cost one hundred and ninety nine pounds, but simpler versions are available for less. There was no additional cost for software. The time of a specialist registrar was required to update and maintain the information. The costs of the paper based on-call pack included the cost of a bag, papers, printing and time of the person updating and printing the information. Disadvantages of using a PDA based on call pack include occasional problems with reading information due to small size of the device. The quality of the information in the pack was rated good rather than very good which means it can benefit from further improvement. The content of the pack would benefit from an editorial group. The on-call action for each disease can be replaced with agreed standard operating procedures. Other areas that need to be addressed before such a scheme is used more widely include the provision of a back up facility and the availability of an efficient system to update the pack. The previous paper based on call pack contained a large volume of guidelines which made it difficult to comprehensively catalogue and update. The initial update during conversion of the pack to an electronic format made explicit the need for continual updating. The most feasible way to update a regional or national on-call pack would be through a web based password protected server. The administrative content can be updated locally. The provision of training and good administrative and information technology support will also be essential as reported by other authors \[[@B20],[@B21]\]. Any support system developed should be linked to local services. Comparing the use of PDAs and laptop computers in our study suggests that PDAs were more likely to be used because of portability. Although smaller than the paper based on call pack, most laptop computers are too large to carry around. PDAs are significantly cheaper than laptops, while laptop computers do not require synchronisation because data can be updated directly. The use of PDAs is widespread among junior doctors especially in North America. Evidence suggests there is greater use among younger physicians \[[@B1],[@B22]\]. This has implications for use in health protection. A significant number of future public health doctors will start public health training with previous experience of using PDAs as these devices are increasing used at all levels \[[@B23]\] and therefore, facilitate access. Conversely, older doctors may find it difficult to adapt to the change. Reasons given by older physicians for not using PDAs include the small size of the touch screen typing device which is difficult to use \[[@B24]\] and lack of familiarity with computer systems. In a recent qualitative study, McAlearney et al found the main barriers to the use of handheld computers to fall in two broad categories, physical (older age, poor sight and finger size) and perceptual (low comfort with technology and devices, preference for paper or desktop computers) constraints \[[@B20]\]. The difficulty with visualising text on PDAs can be aided by using a larger font while that with data entry can be tackled by either using a portable keyboard or with the aid of a voice recording device. Further functions can be added to the device to enhance it use including a search function, internet access and data entry facilitating surveillance and potentially contributing to electronic medical records in the future. The most significant constraint to data entry will be security. There are software available to encrypt data on a PDA \[[@B23]\]. The pack was specifically developed for public health on call staff and general public heath consultants. Due to the potential for memory expansion a significantly larger volume of reference texts and books could be added making the device useful for all cadres of health protection staff. Increasing number of reference texts are becoming available via PDAs. For example, textbooks such as the Oxford Handbooks and the British National Formulary are available on a hand held device. Several journals and Pubmed have also developed PDA compatible access. We conclude that a PDA could be a useful addition to tools available for efficiently delivering information at the point of use for out of hours health protection on call duty. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= IA conceived the idea and developed the on call pack and drafted the manuscript, CJW carried out the evaluation. All the authors participated in writing the paper. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Summary of responses from participants ::: ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ Participant A B C D -------------------------------------- --------------------------------------- ----------------------------------------------- ------------------------------------------------------------------------------------------------------------------------------------- --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Number of calls/week 0 1 2 3 Frequency of use 0 2\* 2 2 Ability to find specific information N/A Yes Yes Yes Information used N/A Measles advice Immunoglobulin & vaccination advice, E. coli, Hep B, measles, chicken pox, useful numbers chemical incidents and meningitis Comments \"Contacts was missing on this PDA\"\ \"Needs regular updating as no paper backup\" \"IPAQ is the way forward rather than carrying a laptop I have no computer at home PDA makes it easier to have a life and go out\"\ \"Extremely good idea. Information easy to find. Some files not accessible, contact info not available Some tables in guidelines (measles) too small to read and couldn\'t magnify\"\ \"PDF files difficult to read\"\ \"Would like some training on all functions\" \"Using PDA far preferable to laptop\" \"Travel immunisation advice useful\" ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ \* including one use for familiarisation ::: Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2458/5/35/prepub> Acknowledgements ================ We are grateful to Di Timoney for her contribution to the development of the on call pack and to Laura Hay for comments on the manuscript.	PubMed Central	2024-06-05T03:55:55.856285	2005-4-11	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084354/", "journal": "BMC Public Health. 2005 Apr 11; 5:35", "authors": [ { "first": "Ibrahim", "last": "Abubakar" }, { "first": "Christopher J", "last": "Williams" }, { "first": "Marian", "last": "McEvoy" } ] }
PMC1084355	Background ========== Autism (MIM 209850) is a heritable neurodevelopmental disorder characterized by delayed language acquisition, impaired socialization, and repetitive or stereotyped behaviors. Frequent associated features include sensory abnormalities and lack of motor coordination. The recurrence risk for siblings is approximately 30 times that of the general population, and twin studies have documented a higher concordance rate in monozygotic (60--91%) than in dizygotic twins (0--6%) \[[@B1]\]. According to a summary of 32 epidemiological studies, the male:female ratio is \~4:1, and up to 75% of autistic subjects are mentally retarded to some degree \[[@B2]\]. Associated medical disorders were estimated to be present in \~6.4 % of autistic individuals in that review. Four studies also indicate that siblings of autistic probands are at increased risk for Pervasive Developmental Disorder Not Otherwise Specified (PDDNOS) and Asperger\'s Disorder, suggesting that these disorders exist on a continuum \[[@B3]-[@B6]\]. In fact, a mutation in the Neuroligin 3gene on Xq13 has been identified in a sibship where one brother has autism and the other has Asperger\'s \[[@B7]\]. This mutation was inherited from the mother and changes a highly conserved arginine residue to cysteine (R451C), resulting in retention of the protein in the endoplasmic reticulum and reduced binding with its presynaptic partner beta neurexin-1 \[[@B8]\]. Although mutations in this gene may not explain a substantial portion of the population risk for autism, the finding provides definitive proof for the genetic basis of autism spectrum disorders (ASD) and encouragement for ongoing genetic studies such as ours. Multiple genes and environmental factors likely contribute to the etiology of autism, making it difficult to isolate disease genes. There is a substantial fall-off in risk of autism for second- and third-degree relatives of autistic probands \[[@B9]\], and some predict that alleles of 10 to 100 or more genes, each of small effect, may underlie the autistic phenotype \[[@B10],[@B11]\]. Hirschprung\'s disease, or congenital megacolon, is the prototype of a complex genetic trait and displays an inheritance pattern that could well be representative of autism. To date, there are more than nine susceptibility genes for this disorder, and different combinations of alleles interact to produce the phenotype. The penetrance of Hirschprung\'s disease is low, and the phenotype varies widely in severity within and among families. This pattern is also seen in the family mentioned above, where a sibling with autism and a sibling with Asperger\'s share at least one identical genetic determinant. Two principle strategies have been used to try to limit the genetic heterogeneity underlying such complex phenotypes. One proven method is to identify more genetically homogeneous subphenotypes. With respect to autism, language delay and obsessive compulsive behaviors are traits that segregate within both affected and unaffected members of a family \[[@B12],[@B13]\]. In practice, subsetting samples according to the presence of language delay has strengthened evidence for linkage to the same regions on chromosomes 2q and 7q in several different studies \[[@B14]-[@B17]\]. Another, albeit less successful, means of reducing genetic heterogeneity is to study these disorders in isolated founder populations. These populations are derived from a small number of original founders and then grow exponentially, in isolation, such that the small number of disease alleles introduced by the founders will be widely distributed in the present day. For instance, genes for Hirschprung\'s disease have been identified in the genetically isolated population of the old order Mennonites, although some degree of heterogeneity was present even in this population \[[@B18]\]. The CVCR is another well known founder population widely studied by geneticists interested in neuropsychiatric disorders and other complex traits like asthma \[[@B19]\]. The CVCR was founded \~15 generations ago by a small number of families from southern Spain, and the local, predominantly female, Amerindians. These assertions, based largely on civil and church registries, have been confirmed at the genetic level using haplotype data from mitochondrial DNA (mtDNA), Y-chromosome, and autosomal markers \[[@B20]\]. Mitochondrial data indicate that the four major Native American mtDNA lineages are enriched in females from the CVCR. Data from the Y chromosome shows that the male ancestry in CVCR is about 0.86 European, with a small contribution from West African and Amerind populations. Importantly, an isolated founder population in Antioquia, Colombia shares the same demographic history as the CVCR, and autism case collection is currently underway in that region using the same study protocol as ours. The only published studies of ASD in an isolated population are ongoing in Finland \[[@B21]-[@B23]\]. A comparison of gene diversity between Finland and the CVCR shows that they are similar at the autosomal level \[[@B20]\]. Previously, genetic studies of autism were not undertaken in the CVCR because the Autism Diagnostic Interview-Revised (ADI-R) \[[@B24]\] and the Autism Diagnostic Observation Schedule (ADOS) \[[@B25]\], the standard research instruments used to diagnose autism, were available only in English. Furthermore, application of the ADI-R and ADOS for research studies requires formal certification training in English. The ADI-R is a structured diagnostic interview administered to parents that evaluates social abilities, verbal and non-verbal communication, and repetitive and stereotyped behaviors in possibly autistic children. In our population, the test takes anywhere from 2 to 3.5 hours, depending upon the parent\'s level of awareness. The ADOS is a semi-structured standardized assessment of social interaction, communication, and play that can distinguish autism and PDDNOS from non-spectrum disorders. The test is 30 minutes long, and one of four modules is administered to the case, according to their level of expressive language. Fortunately, a Spanish-language version of the ADI-R is now available, and our group participated in the recent translation of the ADOS. According to the National Census Register of Costa Rica, the CVCR harbors approximately 250,000 children between 0--4 years, another 250,000 between 5--9 years, and a total of one million individuals under the age of 20, with a slight preponderance of males in all age groups. (These numbers represent just under two-thirds of the totals for the entire country, as that is the estimated percentage of people living in the Central Valley). Assuming that published prevalence estimates in other countries apply to Costa Rica, as has been documented for Tourette\'s Syndrome and Bipolar Disorder \[[@B19]\], and taking a conservative value of 1 in 1,000 \[[@B2]\] as the prevalence estimate, there should be \~1,000 cases of autism between ages 0--20. Given that few cases will be seen before two or three years of age, and that we require Central Valley ancestry (not just birth) for entry into the study, the number of eligible cases may be anywhere from 300--500. With the appropriate instruments and research-trained bilingual clinicians, we have begun recruiting autistic subjects and their parents from the CVCR for a genetic study. The first stage of the genetic study will be to perform highthroughput cytogenetic screening using a microarray-based Comparative Genomic Hybridization assay specially designed to detect microdeletions and duplications (microdels/dups), including sub-telomeric abnormalities, too small to be identified by standard karyotyping. Should we have a suitable sample size of cytogenetically normal cases after this evaluation, possibly augmented by samples from the genetically related population in Antioquia, we intend to perform a genome-wide single nucleotide polymorphism (SNP) screen to identify autism susceptibility genes. We also intend to use our samples to perform fine-mapping of autism loci identified by other investigators. Here we report complete clinical data gathered on a representative 35 cases recently assessed in the CVCR. Although this is a preliminary assessment, our findings related to the problems of IQ testing and recruitment in this population may be helpful to other Latin American investigators planning similar studies. Methods ======= This project was approved under the guidelines of the Ministry of Health of Costa Rica, the Ethical committee of the National Children\'s Hospital in San Jose, and the Institutional Review Board at Mount Sinai School of Medicine. These approvals remain active. All adult participants provided written informed consent. Parents of participating minors provided written informed consent; minors who were capable also provided signed assent. Translation of the ADOS ----------------------- Dr. Valeria Nanclares completed a preliminary translation of the ADOS. She is the Autism Program Coordinator at the Pediatric Developmental Center, Advocate Illinois Masonic Medical Center, and also the translator of the ADI-R. Next, two professional translators read the document separately in English and Spanish and documented instances where the translation was weak or ambiguous. This approach avoids the introduction of synonyms and colloquialisms and is thus preferable to simple back-translation. Subsequently, native Spanish-speaking clinical experts from Latin America and Spain reviewed the document for content and accuracy of clinical terminology and achieved a final consensus with the original translator. The Latin American expert is Dr. Elina R. Manghi (ERM), who is also the lead best-estimator for this study. She is a clinical psychologist and the Clinical Service Director for the Family Clinic, Department of Disability and Human Development at the University of Illinois at Chicago. She has over 20 years of experience in the field of psychological testing and for the last 10 years has worked with developmental disorders, including autism. She is a certified off-site trainer for both the English and Spanish versions of the ADI-R and is certified for the research use of the ADOS in the United States. The expert from Spain was Dr. Amaia Hervas, who has worked extensively with Dr. Michael Rutter\'s group on various autism research projects. Ascertainment of cases ---------------------- The Hospital Nacional de Ninos (HNN) is located in San Jose, the capital city of Costa Rica, and is the only major hospital for children in the CVCR. The HNN has an inpatient Neurodevelopmental Unit and recently added a small special education school providing services for children with mental retardation (MR) and developmental delay but not specifically autism. All four child psychiatrists serving the CVCR are also at the HNN. The leading clinician on our study, Dr. Patricia Jimenez Gonzalez (PJG), is Chief of the inpatient Neurodevelopmental Unit at the HNN and has an extensive private practice. She also has an active relationship with the Autism Parents\' Association of Costa Rica, a group that is highly motivated to participate in research and promote community education. PJG has passed Dr. Catherine Lord\'s certification course for research use of the ADI-R and ADOS. A recent study showed that the number of autism and autism spectrum disorder (ASD) cases being diagnosed in the United States increased precipitously following the introduction of Federal special education laws that included autism spectrum disorders as a disability category in 1991 \[[@B26]\]. The diagnostic criteria for autism were also broadened in DSM III-R in 1987 and DSM IV in 1994, and one study shows that both the changes in the formal diagnosis of autism, as well as the laws providing Federal support for children with ASD, can account for the increased incidence of this disorder in Olmsted County, Minnesota \[[@B27]\]. Therefore, it is not surprising that awareness of autism is only a recent phenomenon in Costa Rica, as this country lacks sufficient financial support for treatment of children with ASD. As a result, we began our recruitment with a list of 245 potential cases known to Dr. Jimenez from the HNN and her outpatient practice. These cases were selected because of documented developmental delay (cognitive/language and motor) and/or extremely disruptive behaviors characteristic of autism that lead parents to seek consultation in the first place. However, we have formulated a strategy to find the cases of high functioning autism, with normal or near normal IQ, that are not being seen by clinicians or treatment providers, or are being seen but not properly diagnosed. For instance, one study recently reported that an increase in the diagnosis of ASD between 1983 and 1996 appeared to be due to greater recognition of ASD in higher functioning children initially presenting with attention deficit hyperactivity disorder (ADHD) \[[@B28]\]. Another study surveyed the medical and psychiatric records of 101 Swedish children of normal intelligence (mean age 9 years, 8 months) diagnosed with PDD, in order to identify cases that might actually represent high functioning autism, and found that three-quarters had symptoms compatible with mild to severe ADHD or problems with attention and motor control \[[@B29]\]. These children also tended to be doing poorly in school or to be in special education programs despite their normal intelligence. Therefore, we will offer training to child psychiatrists, pediatricians, and special education teachers, under the auspices of the Costa Rican Education Ministry, to help them recognize potential cases of high functioning autism. ERM is actively involved in providing this training in Illinois. She will demonstrate this training model, which can be repeated several times a year. ERM has generated a large number of referrals for evaluation of autism at her bilingual Spanish Autism Clinic, which is part of the Family Clinic, Department of Disability and Human Development, University of Illinois at Chicago. We will also invite members of the Autism Parent Association of Costa Rica to participate in this training. We anticipate this training strategy will increase the awareness of high functioning autism, resulting in referrals for the project. Diagnosis of autism and autism spectrum disorders ------------------------------------------------- Families of individuals with possible autism contact us, or are contacted by us after expressing interest in the study, and are formally asked to participate using established informed consent criteria. They are first screened by PJG using an abbreviated version of the ADI-R (Autism Screening Questionnaire) for the presence of symptoms consistent with a developmental disorder, as well as for pertinent medical information. All interviews and exams take place in the Neurodevelopmental Unit of the HNN. Possibly affected individuals and their parents are interviewed by PJG or Dr. Marietha Fallas Delgado (a pediatrician under her supervision) using the ADI-R. Either PJG or Sylvia Monge Monge, M.A., who has a master\'s degree in psychoeducation, administers the ADOS. Both of these assessments are videotaped for independent scoring by the best estimators. An IQ score and an adaptive behavior score are required for the diagnosis of MR. Intelligence is assessed using age and developmentally appropriate instruments available in Spanish and currently in use in Costa Rica. However, a major problem with these IQ tests is that there are often no norms specific for individual Latin American countries. A few IQ tests have norms derived in Spain and Mexico, yet, according to Domingo Campos, Director, Instituto de Investigaciones Psicologicas de la Universidad de Costa Rica (Psychological Research Institute, University of Costa Rica), Costa Rica has no internal data of its own to assess potential scoring biases that might be introduced using foreign norms. The tests we use include the Bayley Scales of Infant Development II (1993) \[[@B30]\] (0--42 months developmental age), the McCarthy Scales of Children\'s Abilities (1972) \[[@B31]\] (2-1/2 to 8-1/2 years), the Leiter-R (1997) \[[@B32]\] (2 years to 20 years), the Wechsler Preschool and Primary Scale of Intelligence-Third Edition (2002) (WPPSI-III) \[[@B33]\], which is divided into two subtests for ages 2:6--3:11 years and 4:0--7:3 years, and the Wechsler Intelligence Scale for Children, Third Edition (1991) (WISC III) \[[@B34]\]. The IQ reflects how children answer verbal reasoning questions and respond to visual-spatial reasoning tasks. Because language and attention are impaired in children with autism, assessments of their IQs are notoriously unreliable. Therefore, we also administer the Vineland Adaptive Behavior Scales/VABS \[[@B35]\] to each proband. The VABS is a structured interview of the parent concerning observed communication and social and daily living skills at home and in the community. Both the IQ and VABS are administered by our psychologist, Licenciada Marcela Esquivel Pla., who was trained in these instruments by ERM. ### Physical examination and laboratory testing Each autistic proband undergoes a complete physical examination, including a neurological examination, an assessment for dysmorphic features, and a dermatological evaluation. Skin is carefully examined to look for signs of neurocutaneous disorders, including tuberous sclerosis (using a Wood\'s lamp) and hypomelanosis of Ito, as these disorders are strongly associated with autistic symptomatology. (Of note, one of the world\'s experts on hypomelanosis of Ito is doing a study at the HNN). Probands with gastrointestinal problems are sent to Internal Medicine for evaluation as at least one study has shown that an unusual complement-mediated lymphocytic colitis may be unique to severely affected autistic subjects \[[@B36]\]. All probands are karyotyped. Testing of auditory brainstem-evoked responses is requested in some cases, and computerized tomography scans are also available on some probands. Fragile X testing is done on all cases. Phenylketonuria (PKU) testing is also performed when required, as it was not instituted as a routine procedure in the CVCR until \~10 years ago. ### Medical and social history Please see the Clinical History form \[see [Additional file 1](#S1){ref-type="supplementary-material"}\] for complete details of this evaluation. Sources of information for medical and social history include parents and medical records. In brief, we record parental occupational and educational history, followed by family neuropsychiatric history, obstetric data, including any adverse events during pregnancy and APGAR scores, the developmental and medical history of the proband, any medication and/or education the proband has received, and results of the physical exam and laboratory testing. As the family history of psychiatric disorders relies on parental report, it cannot be regarded as fully reliable. Therefore, we try to validate the diagnoses by asking about hospitalizations and medication treatments when possible. In both the immediate and extended family we also specifically ask about the presence of language delay and learning deficits, which could be manifestations of the broader autism phenotype. We also ask about autoimmune disorders, as data suggests that at least some of the autistic phenotype may be induced by autoimmune mechanisms operating during pregnancy \[[@B37]\]. ### Verification of genealogy Don Eduardo Fournier is a Costa Rican historian with extensive experience in tracing genealogies who collaborates on a majority of the genetic studies in Costa Rica. He screens the Civil Registry and, if necessary, birth, marriage, and death records in churches and civic centers to determine the birthplace of all grandparents and great-grandparents. Previous genetic studies in the CVCR have shown that probands whose great-grandparents were born in the CVCR will most probably be descendants of the original Spanish founder population. We require that cases have six of eight great grandparents born in the CVCR for inclusion in the study. ### Best estimate diagnostic process The ADI-R and ADOS are independently scored by our bilingual best-estimators, ERM or Pamela Bondy, M.S./CCC-L (PB). PB is a bilingual speech and language therapist with over 10 years of experience in the field of developmental disabilities. She was trained in the use of the ADI-R and ADOS by ERM. Discrepancies, if any, were resolved and final diagnoses reached after review of collateral historical information, clinical exam, and laboratory testing, according to standard best-estimate diagnostic procedure \[[@B38]\]. Results ======= To date we have ascertained 76 subjects and enrolled 70 families in our study. One family later declined to participate because of the mother\'s health, two cases have been excluded by genealogy, one case was too retarded to determine a diagnosis, one case has RETT syndrome (AU043), and the other excluded case is an autistic child with Moebius syndrome (AU044). We are screening the RETT syndrome case for mutations in the MECP2gene. The Moebius syndrome case offers fascinating insights into the etiological basis of autism and will be reported in detail in a subsequent publication. Extrapolating from the entire data set of 70 cases, the male to female ratio appears to be \~6:1. Our goal is to collect over 200 trios, with no major or minor cytogenetic abnormalities, suitable for genetic mapping. As the diagnostic evaluation takes two to three appointments, we have only complete best-estimated data on 35 cases to date. Karyotypes ordered for cases AU025, AU031, and AU040, with family histories of MR, and for case AU001, who has a cleft palate, were all negative at ≥ 550 bands. Interestingly, case AU025 and case AU031 are related to each other by a distance of seven generations, and a connection with case AU040 has not been ruled out. Case AU008, suspected of Angelman\'s syndrome, had one normal karyotype, and testing for a chromosomal 15q11-q13 abnormality will be performed. The rest of the karyotypes are pending. Table [1](#T1){ref-type="table"} in the text contains abbreviated descriptive statistics on the first complete 35 cases \[see [Additional file 2](#S2){ref-type="supplementary-material"}\]. Ages range from 3 years to 13 years of age, and 31 of the 35 cases are male. Mothers and fathers of the autistic children have at least some high school on average. The average mother is a homemaker and the fathers are skilled laborers; overall, the socioeconomic status of the parents is low. The mean total score for the social domain of the ADI-R is 25.3 ± 5.4 in our sample; the nonverbal and verbal communication domain mean scores are 12.6 ± 1.8 and 17.8 ± 3.4, respectively; and the mean score for the repetitive behaviors and stereotyped patterns domain is 6.6 ± 2.3. The ADOS module scores reflect the severity of the ADI-R scores in all cases. Vineland adaptive behavioral profile composite scores range from 20--80, excluding the value of 127 for case AU029 with Asperger\'s, and the mean score is 48.5 ± 19.4. The majority of full-scale IQ scores are less than 50. Delayed onset of phrase speech is present in 24 of the 35 cases. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### is an abbreviated version of descriptive data obtained for each proband, including their sex and age at interview. Scores for each domain of the ADI-R are provided, as well as the full-scale IQ score and the Vineland Adaptive Behavior composite score. The superscript numerals next to the IQ score indicate the specific IQ test that was used. ::: Case Social total Communi-cation nonverbal total Communi-cation verbal total Behavior total Age at first single words Age at first phrases Gender Age at interview (ADI-R) Full-scale IQ Vineland Adaptive Behavior composite score -------- -------------- -------------------------------- ----------------------------- ---------------- --------------------------- ---------------------- -------- -------------------------- --------------- -------------------------------------------- 001 29 18 10 1 1 M 7.83 49^6^ 37 002 29 20 5 0 0 M 5.67 \<50^8^ 53 003 28 25 6 0 1 M 6.75 87^6^ 60 004 27 14 8 1 1 M 7.50 \<50^5^ 30 005 26 22 10 1 1 F 5.42 57^6^ 51 006 9 20 5 1 1 M 6.25 67^7^ 54 007 21 20 10 0 1 M 10.00 37^7^ 20 008^1^ 29 12 8 0 0 M 5.08 NA 45 009 27 19 9 0 1 M 8.08 34^7^ 46 010 27 21 5 0 1 M 9.25 43^6^ 35 011 27 16 6 0 1 M 10.50 NA 50 012 20 8 14 8 1 1 M 5.25 72^7^ 70 013 29 14 16 6 0 0 M 4.33 \<50^5^ 44 014 29 13 19 6 1 1 M 6.50 \<50^5^ 51 015 30 14 9 0 1 F 5.03 \<50^5^ 46 016 28 12 20 12 1 1 M 13.00 60^6^ 39 018 27 11 7 1 0 M 3.67 \<50^5^ 67 019 29 17 7 0 0 M 7.25 66^6^ 63 025 28 12 8 0 0 M 7.83 \<50^5^ 24 027 24 18 6 1 1 M 4.25 \<50^5^ 55 028 29 14 3 0 1 M 5.42 \<50^5^ 47 029^2^ 16 9 2 1 1 M 6.83 113^6^ 127 030 27 13 5 1 0 M 5.66 \<50^5^ 43 031 29 15 4 1 1 M 7.08 70^6^ 52 032 16 11 4 1 0 M 5.42 \<50^5^ 42 033 18 8 6 0 0 F 4.83 \<50^5^ 46 035 18 15 5 1 1 M 7.08 \<50^5^ 45 036 29 14 3 0 0 M 6.33 \<50^5^ 32 039 30 14 5 0 0 M 9.82 \<50^5^ 21 040 15 15 5 1 1 M 3.08 75^5^ 80 049 29 14 9 1 1 M 13.25 \<50^5^ 20 050 21 13 6 1 1 M 2.92 pending 60 051 30 14 8 0 1 F 5.50 \<50^5^ 39 053 29 18 8 0 0 M 5.83 105^6^ 55 056 28 21 7 ? 1 M 6.00 56^6^ 49 043^3^ 22 12 5 0 0 F 4.83 NA 41 044^4^ 27 11 2 1 1 F 5.75 NA 50 ^1^Angelman\'s syndrome ^2^Asperger\'s syndrome ^3^Rett syndrome ^4^Moebius syndrome ^5^Bayley ^6^WPPSI ^7^Leiter ^8^McCarthy NA indicates that the IQ test could not be completed because the child is too mentally retarded or was too distracted. ::: The Family History of Medical Disorders chart is available in the supplementary data \[see [Additional file 3](#S3){ref-type="supplementary-material"}\]. AU033 developed epilepsy at the age of 2 years and 9 months, but is the only proband with this condition. Six of the cases have minor congenital defects, including a cleft palate, hip rotation, visual problems, megacisterna magna, and esophageal reflux/cardiomyopathy. Twelve of the cases have first- or second-degree relatives with language delay or learning deficits. Developmental delay or autism is present in the first-degree or second-degree relatives of four cases. Four of the cases have a second-degree relative with non-specific MR, one case has a second-degree relative with Down\'s syndrome, five cases have a second-degree relative with epilepsy, and one case has a sister with a minor congenital defect (cleft palate). Six cases have first-degree relatives with a major psychiatric disorder, including schizophrenia and major depression. No cases have tested positively for Fragile X or phenylketonuria, and no history of autoimmune disorder was documented for any case. We have obtained extensive obstetric data on our cases and will be reporting it in full in a subsequent publication. In brief, although only 4 of the mothers were older than 35, 21 of them had obstetric complications and/or the infant had an APGAR score less than 6 at one minute. Fully twenty of the cases were the product of difficult labor, resulting in fetal distress and a caesarean section in eight instances. Discussion ========== On the whole, our first cases have mean ADI-R scores comparable in severity to those of other autism genetic studies, including a sample from the Autism Genetic Resource Exchange (AGRE) (N = 364 cases from 212 sibships) \[[@B12]\] and a sample of 43 sibships from the International Molecular Genetic Study of Autism Consortium \[[@B16]\]. IQ and adaptive behavioral scores were not reported for the AGRE cases, but our mean VABS composite score is 48.5 ± 19.4 versus a mean of 47.44 ± 16.15 for the IMGSAC cases. However, it is clear that the cases of autism known to the HNN are more severe than those seen in other non-Hispanic populations studied recently, although perhaps not before 1991. Eighty-three percent of our cases are mentally retarded, while 60% have IQs \<50. In contrast, epidemiological data from a range of countries, although none of them Latin American, estimate that 75% of autistic individuals have IQs \<70 and 40% have IQs \<50 \[[@B2]\]. However, a closer look at all the major epidemiological studies to date, reviewed by Fombonne in 2003, shows that the median prevalence of autism was only 4.4/10,000, and the reported percentages of normal IQs were, in general, lower in studies done before 1991 when there was also less awareness of autism. In contrast, the median prevalence of autism is 12.7/10,000, and the percentages of normal IQs is increasingly higher, in studies done more recently. This is in part due to a broader definition and recognition of autism, but also, at least in the United States, it is possible that early intervention services, instituted over 15 years ago, have improved the cognitive performance of autistic children. For instance, some studies report that cases of autism with low IQs can achieve marked improvements in intelligence, language, learning, or visuo-spatial skills \[[@B39],[@B40]\] with a minimum of two years of intervention \[[@B41]\], although others feel that outcome measures have not been properly evaluated \[[@B42]\]. Costa Rica now provides some special education services; however, these are not specific to children with autism and are not offered to children more than once weekly. Moreover, teachers are not trained in autism interventions, and many children complete only primary education. We must also consider the possibility that the IQs of our cases are underestimated due to the use of psychometric tests without Costa Rican norms. A recent study done in Argentina illustrates the effects of low socioeconomic status and lack of internal norms on IQ testing results. This group set out to perform a longitudinal study of the effects of poverty and malnutrition on the psychological development of children in a province of Buenos Aires \[[@B43]\]. The study took place from 1984 to 1995 and followed 100 children of low socioeconomic status with or without malnutrition. Cases between ages 0--2 with malnutrition but without other neurological or medical deficits were ascertained from an ambulatory Nutritional Rehabilitation Unit of a local hospital and matched with well-nourished cases of the same age and sex from the same socioeconomic background. A pediatrician, a pediatric neurologist, a psychologist, and an anthropologist performed the evaluations. Ninety-two children were reassessed at 2 years, and 44 were reassessed at the 10-year point. Malnutrition was rectified in that group during the two-year interval between assessments. For the initial evaluation, the group used the Escala de O. Brunet-I. Lezine, which is used to assess global psychomotor development of 0--2 year olds. At the next two time points, they used the Terman-Merril (Forma L-M) intelligence test to obtain IQs. This test was normed on children in the United States, and there were no norms available for Argentina. At the first evaluation, the malnourished children did have significantly lower developmental quotients than the well nourished group, the majority of which had normal ratings. However, at time point 2, after which the malnourished children had been treated, there was no difference between groups, but both were lower than the standard. At time point 3, when the children were \~10 years of age, there was again no difference in IQ between groups, but \~50% of both groups had IQs = 70. The authors of the study drew two major conclusions from these results. The first was that poverty and lack of early intellectual stimulation had profound effects on IQ, even without malnutrition, and that these effects became more prominent with time. The second conclusion concerned the use of IQ tests without norms reflecting the general population of Argentina. The raters felt that the IQ scores of both groups were probably somewhat underestimated because the norms they were comparing them to were derived from children with much higher socioeconomic status. The raters were very experienced in administration of the IQ tests, which were commonly used at universities and elsewhere in Argentina, so misuse of the tests was not an issue. Unfortunately, Latin American governments do not invest in psychological research, and the corporations that sell tests in the United States don\'t see Latin America as a strong market and thus put no money into norming translated tests. (We wish to note that Argentina now has its own norms for some IQ tests but not all.) As further evidence of problems with IQ testing, and perhaps ascertainment bias, in the CVCR, we note that in genetic studies of Tourette\'s Syndrome and ADHD ongoing in the CVCR, 20--25% of cases evaluated for either phenotype have been excluded because of MR (PJG, unpublished data). In contrast, 3.9 % of children with Tourette\'s Syndrome (TS) had MR (range 1--14%) in a large-scale analysis of this disorder in 22 countries (no specific data for Spanish-speaking countries) \[[@B44]\]. Additionally, although children with ADHD have, on average, IQs 7--12 points lower than comparison children, the rate of MR observed in the CVCR is certainly not characteristic of ADHD in North America \[[@B45],[@B46]\]. Of course, an ascertainment bias towards severe cases could also explain the CVCR IQ data for ADHD and TS. However, data concerning the prevalence of epilepsy in autism and ASD also suggests that the IQs of our subjects may be underestimated. For example, severe MR is often seen in cases of autism with a comorbid medical disorder. Early seizures, before two years of age, were present in 60% of this population in one study, and 20% of autistic subjects with severe MR but no comorbid medical disorders experienced seizures before the age of five \[[@B47]\]. As the mean age of our probands is 6.8 years and only one of them has experienced seizures, it seems less likely that autism is secondary to other medical factors in the majority of our cases. Finally, the median male to female sex ratio of 11 previous epidemiological studies, wherein the majority of cases were moderately to severely retarded, was 1.9:1, versus 5.75:1 for studies of predominantly normal IQ probands \[[@B2]\]. We also observe a family history of MR in 4 of our cases, but our male to female sex ratio is \~6:1 (also true for the 70 cases collected to date). Accurate assessment of IQ is important, as data suggests that that there may be different genetic and environmental determinants for autism based on level of cognition, although cases of autism with low IQ may still have a strong genetic basis. For instance, Baird & August \[[@B48]\] observed a familial loading in cognitive impairment confined to autistic probands with severe or profound MR. To determine if autism with severe MR was a distinctly different autistic endophenotype associated with increased familial cognitive impairment, another group \[[@B49]\] studied 47 autistic probands with an IQ \<50 and contrasted their findings with a sample of 99 autistic subjects, the majority of whom had IQs \>50 \[[@B6]\]. Importantly, they found that increased loading for ASD was comparable in both samples, supporting prior evidence that the same genetic liability applies to autism and ASD within families, regardless of IQ level. Interestingly, a recent study also shows that co-occurrence of lower IQ and ADHD reflect common genetic determinants and not necessarily the presence of a primary medical or neurological disorder \[[@B46]\]. We do note frequent obstetric complications in our sample and are researching the rates of these complications in the CVCR hospitals for comparison. We are also looking at factors such as geographic distance from a hospital, which might contribute to suboptimal labor and delivery. However, recent data suggests that these complications are commonly associated with autism and ASD. A large-scale epidemiological study of autism in Western Australia found that cases diagnosed with autism, as well as their unaffected siblings, had more obstetric complications than a control group \[[@B50]\]. The authors suggest that the increased prevalence of obstetric complications among autism cases is most likely due to underlying genetic factors, possibly exacerbated by environmental stressors. There are many possible underlying cytogenetic or other medical disorders that may be accompanied by autistic behavior, and the presence of severe MR can make those cases suspect even if karyotypes are normal. Disturbingly, a group of investigators recently reported that they found microdups/dels, as well as other genetic abnormalities, in 272 cases from the Paris Autism Research International Sibpair (PARIS) sample that had already been included in genetic linkage and association studies \[[@B51]\]. These include a 22q11 deletion without the associated features of DiGeorge/Velocardiofacial syndrome in two autistic brothers with severe MR, and their mother, as well as duplications of 15q11-13 (in two different singletons) and of 17p11.2 and 7q11.2. Therefore, we have developed a collaboration to screen our cases using microarray-based Comparative Genomic Hybridization technology. We will hybridize our samples to a whole genome array containing 2632 BAC (bacterial artificial chromosome) clones at a resolution of \~1 Mb spanning the 22 autosomes and the X and Y chromosomes. In this way we will find any known and novel microdups/dels, as well as subtelomeric abnormalities, in our cases before we proceed with any genetic association analyses. Thus, we feel confident that our cases, regardless of IQ, will be as free of genetic abnormalities as can be determined given current technology. Conclusion ========== We have successfully used the Spanish versions of the ADI-R and ADOS to diagnose autism for genetic study in the CVCR. The majority of the cases reported here have IQ \<50. This reflects an ascertainment bias, yet it is also possible that these scores are artificially low. Nevertheless, we are most likely missing cases of autism with normal or near normal IQ due to a lack of government-sponsored services. Therefore, we are implementing a strategy, used successfully by one of our clinical experts to find both English-speaking and Spanish-speaking higher functioning autistic subjects in the United States, to ascertain these cases. We believe that similar problems with IQ testing and the need for specific strategies to find high functioning autism cases will be important issues to address for studies of autism planned or ongoing in other Latin American countries. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= LAM secured the funding for the study, played a major role in designing the study, contributed to the data analysis, and wrote the manuscript. PJG is the leading clinician in Costa Rica and performed the majority of the diagnostic interviews using the ADI-R and ADOS, as well as the neurological and medical evaluations. She contributed to the study design and to the data analysis and ensures compliance with the Bioethics Committee at the HNN. ME coordinated the study in Costa Rica and contributed to the data analysis. She administered informed consents and IQ tests, arranged appointments with subjects. ERM is the lead best estimator for our study and performed the majority of data evaluations. She trained the Costa Rican team in the use of current psychological instruments. She also contributed to the manuscript. MFD performed diagnostic interviews using the ADI-R and contributed to the data analysis. SSM assisted PJG in performing the ADOS and contributed to the data analysis. EF traced the genealogies of the probands and is responsible for establishing familial relationships between the cases as well as providing population data for the CVCR essential to the manuscript. PB performed best-estimate data evaluations and contributed to the data analysis. KC coordinated the study in the USA, entered diagnostic information into a database for future statistical analysis, contributed to the manuscript and ensured compliance with the Institutional Review Board at Mount Sinai School of Medicine. All authors read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-244X/5/15/prepub> Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Formulario de historia clinica This is the clinical history form that we designed in Spanish. The table begins with parental occupational and educational history followed by family neuropsychiatric history, obstetric data, including APGAR scores, the developmental and medical history of the proband, education the proband has received and results of the physical exam and laboratory testing. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 2 Results of Diagnostic Evaluations for Autism This table contains scores for the individual scores subdomains of the ADI-R, the ADOS scores, the IQ scores and information regarding the parents\' occupations and education. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 3 Family History of Medical Disorders This table contains details of the family and personal medical history for each proband. We specifically asked about the presence of autism, language delay, learning deficits, MR, epilepsy or other neurological abnormality, cerebral palsy, major psychiatric disorders and deafness. For instance, if a first cousin on the maternal side had epilepsy, we typed \'M first cousin\' in the appropriate box. The letters \'M\' or \'P\' indicate whether the relative is on the maternal or paternal side, respectively. If the mother had a complication during pregnancy such as bleeding or a urinary tract infection, this is noted with an x in the appropriate box. APGAR scores are recorded, as well as the presence of any congenital defects in the proband. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ LAM is supported by NINDS grant number R01 043540. We would like to thank Andrea Vargas and Haya Wasserstein for help with recording subject interviews, the physicians at the HNN who served as consultants for this project, including the dermatologist Dr. Mario Sancho, and the audiologist Dr. Juan José Madriz, Dr. Olga Arguedas from the Bioethical Board, Drs. Isabel Castro and Patricia Cuenca for performing chromosomal testing, and Dr. Abdon Castro, Presidente de la Fundación Pro Hospital Nacional de Ninos \"Dr Carlos Sáenz Herrera\". We also thank Safiana Katz, Eric Mesh, and Zhan Liu for their careful processing of DNA samples and immortalized cell lines. Above all, we are grateful to the families who have participated in the study and to the Autism Parents\' Association of San José, Costa Rica.	PubMed Central	2024-06-05T03:55:55.858304	2005-3-21	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084355/", "journal": "BMC Psychiatry. 2005 Mar 21; 5:15", "authors": [ { "first": "L Alison", "last": "McInnes" }, { "first": "Patricia Jiménez", "last": "González" }, { "first": "Elina R", "last": "Manghi" }, { "first": "Marcela", "last": "Esquivel" }, { "first": "Silvia", "last": "Monge" }, { "first": "Marietha Fallas", "last": "Delgado" }, { "first": "Eduardo", "last": "Fournier" }, { "first": "Pamela", "last": "Bondy" }, { "first": "Kathryn", "last": "Castelle" } ] }
PMC1084356	Introduction ============ Tissue Doppler echocardiography (DTI) has been introduced as a new method to quantify regional myocardial function. It provides an analysis of myocardial velocities and regional myocardial displacement, deformation and rate of deformation \[[@B1]-[@B5]\]. In routine practice the diagnosis of an acute coronary syndrome is based on guidelines using clinical, electrocardiographic and biological patterns \[[@B6]\]. Nevertheless to improve speed, pertinence and efficiency of the diagnosis in many patients, a subjective interpretation of myocardial regional motion and thickness is frequently performed in the intensive care unit (ICU), therefore not necessarily in the best conditions. The, wall motion score assessing regional contractility has been validated. Reproducibility of this method is high for expert observer but it requires dedicated training and remains subjective. Thus, DTI may offer an alternative quantitative technique that can be used in routine practice. Several parameters have been described in experimental settings \[[@B7]-[@B11]\]. We therefore sought, to look at most of these previously proposed quantitative parameters in a clinical setting to assess their relevance compared to coronary angiographic findings. Methods ======= This study was a single-center observational study. Informed consent was obtained prior to each coronary angiography. The study was designed in compliance with the ethic principles of our institution. A series of consecutive acute myocardial infarction (AMI) patients treated by primary percutaneous coronary angioplasty (PTCA) of the left anterior descending coronary artery (15 patients, 66.5 ± 17 years of age) or the right coronary artery (13 patients, 54 ± 13 years of age) were prospectively analyzed. The TIMI 3 flow was obtained in all patients. The exclusion criteria were: AMI related to circonflex artery (this coronary artery might be responsible of too variable size and localization to constitute an homogeneous group), any other coronary artery stenosis than the culprit lesion, left bundle branch block, pace maker, significant valve disease, atrial fibrillation, dilated or hypertrophied cardiomyopathy and cases in which more than twelve hours had elapsed between the beginnings of the chest pain to the coronary angiography. We did not included patient with a circonflex artery lesion in this study because the myocardial territory vascularized by this artery might be too much different from one patient to another one. The control group consisted of 17 patients without any known myocardial disease who had been referred to the catheterism laboratory for chest pain. Each of these patients had a normal coronary angiography and normal conventional trans-thoracic echocardiography. For the 2 AMI groups, the echocardiographic exam was performed in the 24 hours following the admission and the PTCA. Each of these patients had no other known pathology likely to impact on myocardial function and the treated artery (right coronary and left anterior descending coronary artery) was the only abnormal artery on the coronary angiography. The angiographic result of each PTCA performed was satisfactory and a TIMI 3 flow was recorded in all patients at the end of the procedure. The same investigator performed all the echocardiographic recordings and analysis. A Vivid Five ultrasound scanner (GE VingMed, Milwaukee, Wisconsin) with a 2.5 Mhz phased array transducer was used. The patients were imaged in the left lateral decubitus position in the ICU. Standard 2-dimensional, M-mode, pulsed and color Doppler were saved on digital files (Echopac -- GEVingMed, Milwaukee, Wisconsin). In apical four and two-chamber, color DTI Cineloops of 3 cardiac cycles (from a gating EKG) were recorded with a frame rate between 80 and 115 frames/s depending on the sector width. The image was adjusted to ensure a parallel alignment of the sampling window with the myocardial wall studied. The pulse repetition frequency was between 500 Hz and 1 KHz, resulting in an aliasing velocity about 16 to 32 cm/s. Three cardiac beats were digitized and stored on a magneto-optical disk for further analysis on Echopac (GE VingMed, Milwaukee, Wisconsin). Data processing and analysis ---------------------------- Myocardial Doppler velocity profile signals were reconstructed offline from the DTI color images. Curve analysis was performed on an averaged analysis of the three recorded cardiac beats. A regional analysis was then performed, positioning an index at the mitral annulus, basal, mid and apical segments of the septal, lateral, anterior and inferior left ventricle myocardial wall (figure [1](#F1){ref-type="fig"}). The systolic (S); early diastolic (E), late diastolic (A) and also the isovolumic contraction and relaxation peaks were recorded. Tissue tracking systolic peak displacement was recorded for each of the 4 indices placed on each wall (figure [2](#F2){ref-type="fig"}). The strain rate analysis was considered, but we chose to not report any results. The signal to noise ratio was not acceptable. Frequently, it was challenging to distinguish peaks in systole and diastole. Nevertheless, we considered the systolic peak of strain and the delay of apparition of that peak of contraction for the analysis (figure [3](#F3){ref-type="fig"}). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Example of a DTI curve analysis. It was manually positioned a region of interest (ROI) at the level of the mitral annulus and three others in the 3 segments of each wall (following the ASE left ventricular segmentation) IVC: peak of velocity recorded at the isivolumic contraction time S: peak of velocity recorded in systole IVR: peak velocity recorded in isvolumic relaxation time E: peak velocity in early diastole A: peak velocity in end-diastole ::: ![](1476-7120-3-10-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Example of Tissue tracking curves displayed on Echopac. It was there easy to record the peak (S) of systolic displacement of the same 4 ROI described in figure 1. ::: ![](1476-7120-3-10-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Example of Strain analysis; It was there possible to record the degree of systolic shortening or lengthening of the 4 studied ROI. ::: ![](1476-7120-3-10-3) ::: To evaluate left ventricular filling pressures, the ratio mitral inflow E-wave peak velocity / early diastolic velocity Ea of the mitral annulus (septal part) was calculated. Ea being the early diastolic pulsed DTI peak velocity recorded in pulsed DTI \[[@B12]\]. Statistics ---------- Before comparison, the 3 grouped datasets were tested for normal distribution and equality of standard deviation (SD). Normally distributed data are given as mean and SD, and were compared using a parametric analysis of variance (ANOVA, completed by a Scheffe post-hoc analysis). Tests resulting in p values below 0.05 are considered statistically significant. Receiver operating characteristics and logistic regression analysis were used to find and optimize cut-off values of the most relevant parameters for distinguishing normally vascularized myocardial segments and myocardial segments just after revascularization by emergency coronary angioplasty for signs of AMI. To assess the intra and inter-observer variability, each echographic parameter was determined again by the original observer and an independent observer in 10 patients from the 3 populations. Results ======= Clinical and echocardiographic characteristics ---------------------------------------------- Table 1 ([additional file 1](#S1){ref-type="supplementary-material"}) summarizes the clinical characteristics of the three populations. Pharmacologic therapy is identical in the two AMI populations. Each patient was receiving Metoprolol Succinate. The mean dosage was 100 mg per day or Atenolol 50 to 100 mg per day. In control subjects: The isovolumic contraction (IVC) was a positive wave. The systolic (S) wave was positive too. The isovolumic relaxation time motion (IVR) was negative as were E (early relaxation motion) and A (late diastolic motion). Figure [4](#F4){ref-type="fig"} shows the mean values of the most pertinent DTI-derived parameters in each wall for the 4 points that were studied (annulus, base, mid-segment and apex). This figure [4](#F4){ref-type="fig"} illustrates that even in non-ischemic segments from the 2 AMI populations (the inferior and the anterior one), velocities and displacements were significantly altered in comparison to the control population. Potentially linked to these findings, estimated left ventricular filling pressures were significantly higher in the two AMI groups: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Representation of IVR (isovolumic relaxation), IVC (isovolumic contraction) peak velocities, early diastolic tissue Doppler velocity (E), systolic tissue Doppler velocity (S), systolic displacement (tissue tracking) and strain in the three segments of each myocardial walls analyzed in the anterior (ANT) myocardial infarction population, in the inferior myocardial infarction one (POST) as in the control group (normal). ::: ![](1476-7120-3-10-4) ::: E/Ea = 18 ± 12 (cm/s) in the Anterior AMI group; E/Ea = 16 ± 10 (cm/s) in the Posterior AMI group; E/Ea = 8 ± 2 (cm/s) in the control group. Analysis of the differences between each studied parameters using variance analysis ----------------------------------------------------------------------------------- The E and the S waves of the DTI appeared relevant to discriminate the myocardial properties of the three studied populations, especially in the basal and mid segments of each wall (p-value always ≤ 0.03) (additional file - table 2). The IVR peak velocities were highly discriminative when the control subjects were compared to each of the two AMI populations, considering the values observed in the basal and mid segments. P-value was \<0.005 for each wall except the inferior wall for which IVR did not discriminate either of the two AMI populations in the mid segment and p-values were only 0.02 in the basal segment. The peak of systolic displacement analyzed in tissue tracking mode was extremely relevant, always discriminating the three myocardial status with p-values \< 0.003. There was no substantial change in the peak IVC, the delay of S-wave and the A-wave between the control population and each of the two AMI populations. The systolic peak of strain was not discriminative enough to distinguish normal from abnormal myocardium (p = ns). To define the best parameter or the best combination of parameters against the coronary angiography results, we performed a logistic regression analysis -------------------------------------------------------------------------------------------------------------------------------------------------------- To discriminate the control subject from the pathologic subjects, the peak of systolic displacement measured by tissue tracking appeared to be the best parameter at the annulus and the basal segment level (β = -14.27, p = 0.0018 ; β = -10.36, p = 0.005; β = -17.75, p = 0.01; β = -12.09, p = 0.01 in the septal, lateral, inferior and anterior wall respectively). Results were less clinically relevant in the mid and apical segments. The multivariate analysis did not provide one parameter, but three different ones form one segment to another. To discriminate anterior from inferior AMI, peak E was the best parameter at the mitral annulus and peak IVR was the best in the basal and mid-segments of each left ventricular walls (β = -1.14, p = 0.011; β = 0.94, p = 0.014; β = -1.74, p = 0.025; β = 1.97, p = 0.059 in the septal, lateral, inferior and anterior walls respectively). To complete the determination of the most clinically relevant parameters to distinguish normal from recently revascularized segments, it was performed a ROC curve analysis. At this point, it was chosen to implement a qualitative approach based on our experience of DTI signal analysis. Thus, it was added in the analysis the parameter compiling IVC peak \> 0 and IVR \< 1. The qualitative parameter was added because of the observation that we did during the analysis of curves and quantitative parameters. This \"IVC peak \> 0 and a IVR peak \< 1\" enabled to exclude coronary artery disease for the studied segment with 82% sensitivity and 85% specificity. The predictive positive value was 68% and negative predictive value 92%. For IVR peak velocity, the area under the curve was 0.73; sensitivity was 72% and specificity 69% for a cut-off value of less than 0.12 cm/s to predict the absence of coronary artery lesion. For the peak VTI (tissue tracking), a cut-off value of less than 0.78 cm offered 76% sensitivity for 51% specificity in predicting the presence of a coronary lesion treated by PTCA in the last 24 hours. The area under the curve was 0.68. Intra and inter observer variabilitiesare shown on table 3 (additional file 3). Average variability over each wall and each segment are shown. No data are reported for the strain rate in the table. Relative strain rate variability was around 10% in the basal segments and reached 50% of variability in some of the apical ones. Discussion ========== In the current study, the main findings were: (1) in an AMI population, myocardial velocities and displacement were found to be deteriorated in the AMI territory but were also deteriorated in the non-ischemic left ventricular myocardial walls. (2) isovolumic relaxation peak velocity seemed to be relevant taken alone and even more in conjunction with isovolumic contraction peak velocity. (3) Peak systolic displacement provided by tissue tracking was easy to obtain and relevant to detect the AMI population and/or AMI territories 12 hours after acute revascularization of an anterior or inferior AMI. (4) Strain and strain rate analysis appeared to be less relevant than expected from a physio-pathologic state points and taking into account studies done under stringent experimental conditions. A qualitative approach based on visual assessment of myocardial wall motion has well-documented limitations and even with eyes of an expert may fail to identify subtle ischemia induced changes in regional myocardial mechanics \[[@B13]\]. In our practice a hypokinesia was observed in one or two segments of the territory of the treated coronary artery. Nevertheless, the hyperkinesia frequently observed in adjacent segments often altered our capacity to precisely define what could be called the \"region at risk\". This limit and the observer dependency were our initial motivation in studying DTI parameters. The curve analysis and especially the qualitative analysis of isovolumic relaxation and contraction peak velocity answer to our initial goal. It becomes possible to record images and to discuss between physicians with the help of this curves analysis. Furthermore we showed that no AMI segments were having also abnormal values of DTI parameters compared to the control group myocardial segment. Age is probably not the explanation to this difference, as evoked by the E/Ea ratio; left ventricular end-diastolic pressure may have an impact on regional myocardial function even in non AMI segments \[[@B14]\]. Most studies dealing until now, with DTI abilities have been conducted in stringent experimental settings; also most of them postulated that one specific parameter or one specific DTI -- derived post-treatment was sufficient to distinguish stunned or ischemic myocardial segments from others, normally vascularized or scarred \[[@B15]-[@B19]\]. The most recent studies have emphasized the great capability of deformation indices derived from strain and strain rate analysis. The accuracy of such ultrasonic strain measurements has been validated in a comparison with sonomicrometry in experimental ischemic models, and ranges of normal values have been recently established in healthy volunteers \[[@B20],[@B21]\]. The main advantages of these myocardial regional deformation parameters comes from the fact that they are less affected by global cardiac motion and segmental tethering than velocity measurements \[[@B22]-[@B25]\]. Nevertheless, they remain sensitive to noise and reproducibility does not reach sufficient robustness. An expertise might be request to perform the interpretation of strain and even more, strain rate curve analysis and one has to keep in mind that our study was performed in the intensive care unit characterized by its monitoring of EKG, pressure, SaO~2~. New softwares, dealing with higher frame-rate images might offer new opportunities to use strain rate, but it remains in our mind a research tool difficult to use in the clinical routine \[[@B25],[@B26]\]. Tissue tracking and, thus the measurement of peak systolic regional displacement is a more recently available post-treatment \[[@B27],[@B28]\]. The obtained curves are much easier to interpret. This might encourage the performance of this measurement in routine practice to discriminate AMI myocardial walls even if the tethering effect, due to the fact that one analyzes velocities (integrated over time), might decrease its theoretical sensitivity. This signal processing method has several advantages: it can be used for off-line measurements and it could also be used on-line to display a color map (parametric imaging) presentation of the degree of systolic longitudinal displacement of each studied myocardial segment. Tissue tracking peak systolic regional displacement has been validated in ischemic heart disease and in stress echocardiography interpretation \[[@B29],[@B30]\]. In our opinion it needs no specific-training, and it has been demonstrated to be relevant especially at the basal part of the left ventricle \[[@B29]\]. Isovolumic peaks, especially isovolumic relaxation peak (IVR) velocity, have been reported in the past and appeared to be relevant \[[@B11],[@B31]\]. In this present study, we underscored the advantage of looking at the positivity or the negativity of the relaxation and contraction isovolumic peaks, without any precise measurement. Pislaru et alreported, in an experimental model, that the presence of a positive IVC velocity wave in the ischemic wall was a marker of a less severe ischemic insult, while its absence identified a more severe ischemia \[[@B31]\]. IVC tissue velocities were more sensitive in terms of the trans-mural extension of necrosis than other systolic and diastolic regional velocity parameters. They also reported that the systolic DTI wave was of little value to discriminate ischemic from non ischemic myocardial segment. Edvarsen et alreported previously IVC and IVR velocity waves relevance in differentiating ischemic from non-ischemic myocardium \[[@B11]\]. These isovolumic indices take advantage of being fewer loads dependent and perhaps also easier to assess than strain and strain rate signals. Limitations ----------- Our control population is not matched in age. Ethic consideration prevent to include patient without any indication for a coronary angiography. We were nevertheless able to compare ischemic segments from non-ischemic ones in the two AMI groups. We do not reported the wall motion score analysis because it was not our purpose. Our daily practice provided us doubts in the ability of wall motion score performed within the first 24-hour in the intensive care unit to really discriminate the \"region at risk\". Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Epidemiological and general echocardiography characteristics of the three studies group and comparison of them. (Ant MI: anterior myocardial infarction; Inf MI: inferior myocardial infarction; LV: left ventricle; IVRT: isovolumic relaxation time; PVF: pulmonary vein flow; TEI: Tei index or Myocardial performance index; E: peak of the velocity recorded in early diastole at the mitral inflow; Ea: peak of velocity recorded in early diastole at the level of the septal part of the mitral annulus in pulsed DTI; E-wave DT: deceleration time of the E-wave). The differencies between the three group by the one-way ANOVA was highly significant: \\ = p \< 0.001 or significant:\* = p \< 0.05. It is also display the differences group per group when the ANOVA showed a significant difference (Scheffe post-hoc analysis). ::: ::: {.caption} ###### Click here for file :::	PubMed Central	2024-06-05T03:55:55.863352	2005-4-8	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084356/", "journal": "Cardiovasc Ultrasound. 2005 Apr 8; 3:10", "authors": [ { "first": "Erwan", "last": "Donal" }, { "first": "Pascale", "last": "Raud-Raynier" }, { "first": "Damien", "last": "Coisne" }, { "first": "Joseph", "last": "Allal" }, { "first": "Daniel", "last": "Herpin" } ] }
PMC1084357	Review ====== The past two decades have seen considerable interest in the development and evaluation of health status outcome measures \[[@B1]-[@B14]\]. Although the assessment of reliability and cross-sectional validity is straightforward, the same cannot be said about the evaluation of a measure\'s ability to detect change. Investigators have often expressed uncertainty in the choice of study design and analysis, and statements such as the following are common: \"Because there is not yet agreement on the optimal design and analysis strategies for a responsiveness study, the authors evaluated the responsiveness of the FRI and RM-18 using two methods\" \[[@B10]\]; \"A variety of statistics have been used to assess responsiveness and no single one is superior\" \[[@B2]\]; and \"The purpose of this study was to determine if different indices of responsiveness provided similar rank orderings of scales in terms of responsiveness\" \[[@B5]\]. It is likely that the absence of a gold standard for change in health status plays a prominent role in stimulating uncertainty in the choice of analysis. The solution to the expressed conundrum has often been the application of a \"shotgun analysis\" where multiple change coefficients are applied to a common dataset \[[@B2],[@B4],[@B5],[@B9]-[@B14]\]. In this paper we provide a brief review of prominent study designs and change coefficients, and illustrate the conflict in applying change coefficients from different \"families of analytic methods\" to the same data. Methodological shortcomings --------------------------- The methodological sophistication and standards for reporting clinical intervention trials stand in sharp contrast to those evident for longitudinal validity studies of sensitivity to change. Agencies funding clinical trials demand a clearly stated research question, evidence -- often in the form of a pilot study -- supporting the sample size, and a statement justifying the analysis. Journal editors require equal clarity and rigor when manuscripts pertaining to clinical trials are considered for publication. All too frequently reports of sensitivity to change of various health status measures appear to be \"studies of opportunity,\" rather than carefully planned investigations. Notably absent from many studies are a clear statement of purpose, elaboration of design details including the expected extent to which the sample\'s true change is likely to be homogeneous or heterogeneous (we will subsequent refer to this as the sample\'s change characteristic), justification of sample size, and a commitment to the most appropriate analysis \[[@B2],[@B5],[@B6],[@B9],[@B11],[@B14],[@B15]\]. The importance of specifying the change characteristic of the sample is that it dictates the choice of change coefficient, or at least the family from which the change coefficient will be selected. Study designs and sample change characteristics ----------------------------------------------- Previous monographs have provided comprehensive reviews of popular designs for sensitivity to change studies \[[@B16]-[@B18]\] and it is not our intent to repeat these discussions. However, to set the stage we identify three popular designs and their corresponding samples\' change characteristics: (1) patients who are expected to truly change by approximately the same amount are assessed at two points in time \[[@B1]\]; (2) two or more identifiable subgroups of patients who are expected to change by different amounts are assessed at two points in time \[[@B19]\]; and (3) patients, many of whom, are expected to truly change by different amounts are assessed at two points in time \[[@B20]\]. To distinguish between Designs 1 and 3 we will refer to Design 1 as being homogeneous and Design 3 as being heterogeneous with respect to change. Design 2 shares the characteristics of Designs 1 and 3. Consistent with Design 1 is the assumption that patients within a subgroup truly change by approximately the same amount and the extent to which differences occur is attributed to measurement error. Like Design 3, the ability of a measure to detect true change is reflected by the extent to which the measure is capable of differentiating the amount of change between units that truly change by different amounts. The units are groups of patients for Design 2 and individual patients for Design 3. Although the three study designs are conceptually simple, ascertaining a sample\'s change characteristic is more demanding. Perhaps the most popular method, particularly for Designs 2 and 3, has been the retrospective global rating of change \[[@B4],[@B5],[@B13],[@B14],[@B21]\]. Here, at the follow-up assessment patients provide their impression of global change in addition to completing the measure of interest. This single item global rating of change is then used as the standard for assessing the measure\'s ability to detect change. Norman and colleagues \[[@B22]\] have challenged this approach on three counts: (1) the notion that the measurement properties of the single item global rating are superior to the multi-item measure under investigation; (2) judgments of change are psychologically difficult and therefore suspect; and (3) correlated measurement error between the global rating and the measure under investigation inflates the true association between the two ratings. With respect to the last point, Norman et al \[[@B22]\], showed that the retrospective global rating of change can result in declaring a measure responsive in a sample of stable patients. An alternative to the retrospective rating is the prognostic rating of change \[[@B19],[@B23]-[@B25]\]. This approach is not subject to errors of recall or correlated error; however, it is dependent on the ability of the rater to accurately estimate the extent of change that might occur. As the name suggests, the essential feature of the prognostic rating method is an a priori declaration of the sample\'s change characteristic. Sensitivity to change studies have applied three designs using prognostic ratings of change: (1) randomized trials where interventions of known effectiveness are compared to placebo or weaker interventions \[[@B19]\]; (2) cohort studies where a known prognostic variable is used to classify patients into groups that are expected to change by different amounts \[[@B25]\]; and (3) clinicians assign expected change scores to patients at their initial visit \[[@B23],[@B24]\]. Meenan et al \[[@B19]\], in a three group (placebo, oral gold, injectable gold) randomized controlled trial, investigated the sensitivity to change of the Arthritis Impact Measurement Scales and several other clinical measures. Consistent with a priori hypotheses, the measures demonstrated a gradient in treatment effects with the injectable gold group demonstrating the greatest change and the placebo group showing least change. Stratford and Binkley \[[@B25]\] applied a cohort design where the natural history of patients with low-back pain was used to established two groups of patients with different change characteristics. Specifically, these investigators theorized that patients with low-back pain of less than 2-weeks duration would change more over the subsequent 2-week interval than would patients who presented with low-back pain of 2 or more weeks duration. Westaway et al \[[@B23]\] investigated the sensitivity to change of the Neck Disability Index (NDI) \[[@B26]\] and Patient Specific Functional Scale (PSFS) \[[@B27]\]. These investigators theorized that seasoned clinicians\' would be able to distinguish among patients who would change by different amounts over an interval of several weeks. At the initial assessment clinicians rated patients\' prognoses on a 5-point scale. Prognostic ratings were based on clinical judgment alone. The results demonstrated significant correlations between the prognostic rating of change and the measures\' change scores. Study designs and their respective families of analytic methods --------------------------------------------------------------- Sensitivity to change studies are rich with descriptions of change coefficients \[[@B2],[@B5],[@B6],[@B9]-[@B12],[@B15]\] which we place in the following three groups or families according to study design: Design 1, coefficients based on homogeneity of patients change characteristics; (2) Design 2, between group contrast coefficients; (3) Design 3, correlation coefficients. Homogeneous patient change -------------------------- This design and analysis is based on the premise that the sample consists of patients who are expected to change by approximately the same amount over the study period. Of interest is not what accounts for the change -- it could be the natural history or the application of an effective intervention -- but rather that the amount of change is expected to be reasonably homogeneous among patients. The ability of a measure to assess change is quantified by dividing the mean change (signal) by the variation in change or sample characteristics at baseline (noise). The standardized response mean (SRM = mean change/standard deviation of change) \[[@B1]\] is a frequently reported change coefficient associated with this design. Statistical tests include the paired t-test and repeated measures ANOVA with one within patient factor (occasion at 2 levels: baseline and follow-up) and no between patient factor. Of the three designs, this one is considered to be the weakest because it does not challenge a measure\'s ability to discriminate among different amounts of change \[[@B16],[@B17]\]. Heterogeneous patient composition: between group contrast --------------------------------------------------------- This design is based on the premise that identifiable subgroups of patients who change by different amounts exist. Change coefficients include area under receiver operating characteristic (ROC) curves \[[@B18]\] and Norman\'s S~repeat~\[[@B28]\]. Statistical analyses for this design include the z-statistic for the area under a ROC curve \[[@B16]\], t-test for independent sample means of change scores, and repeated measures analysis of variance (ANOVA) with one within patient factor (occasion at 2 levels) and one grouping factor (amount of change at 2 or more levels: small change, large change) \[[@B28]\]. Heterogeneous patient composition: among patient contrast --------------------------------------------------------- Like the first design, this one investigates a single group of patients. However, rather than the patients being reasonably homogeneous with respect to change, the patients are expected to truly change by different amounts. Moreover, an essential aspect of this design is that an external standard is applied, the change scores of which are compared to the change scores of the measure of interest. A measure\'s ability to detect change is based on a correlation analysis \[[@B5],[@B16],[@B17]\]. Problem clarification --------------------- Investigators have often applied analyses and change coefficients from the three families of tests to the same patient sample \[[@B4],[@B6],[@B9],[@B11]\], apparently without realizing that the coefficients are based on different, and at times conflicting assumptions concerning the sample\'s change characteristic. For example, Kopec et al \[[@B4]\] reported a study that was conceived to \"determine whether the Quebec scale (a functional status measure for patients with low-back pain) is a reliable, valid, and responsive measure of disability, in back pain, and to compare it with other disability scales.\" The sample was diverse in that it included patients from physical therapy clinics, physiatry centers, rheumatology clinics, family practice groups, and pain clinics. Statistical tests included the paired t-test, repeated measures ANOVA with one grouping factor (amount of change), and a correlation of the Quebec\'s change scores with those of a retrospective global rating of change. Change coefficients included the SRM \[[@B1]\], Norman\'s S~repeat~\[[@B28]\], and an unnamed correlation coefficient. The three analyses were applied to the same group of patients. To underscore the theoretical conflict in applying these coefficients to the same patient sample we will link the coefficients reported by Kopec et al \[[@B4]\] through repeated measures and regression ANOVA tables. Illustrative comparison of change coefficients ---------------------------------------------- To facilitate discussion, we will make reference to the dataset displayed in Table [1](#T1){ref-type="table"}. These data represent the results from a hypothetical study where a health status measure was administered to 20 patients at their baseline assessment and at follow-up 2-months later. The investigator believed that patients would improve over this interval. Also, at the follow-up visit patients provided a global rating of change on a 15-point scale (-7 to 7) \[[@B21]\]. Furthermore, the investigator dichotomized the patients\' global ratings using a cut-point of 5 on the global rating. The investigator did not declare detailed a priori assumptions concerning the extent to which patients were expected to change by different amounts. Three analyses are presented: (1) a repeated measures ANOVA with no grouping factor and 1-within patient factor; (2) a repeated measures ANOVA with 1-grouping factor and 1-within patient factor; and (3) a correlation of the measure\'s change scores with those of the retrospective global rating of change. Although our illustration represents a hypothetical study, the design and analyses are consistent with the approach of Kopec et al \[[@B4]\] and many other studies reported in the literature \[[@B6],[@B9],[@B12]\]. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Summary of synthetic data ::: Patient ID Initial Assessment Follow-up Assessment Change Initial-follow-up Global Rating Global Rating Dichotomized ---------------- ------------------------ -------------------------- ------------------------------ ------------------- -------------------------------- 1 16 14 2 3 0 2 14 14 0 1 0 3 13 5 8 6 1 4 19 10 9 5 1 5 14 8 6 6 1 6 3 4 -1 2 0 7 18 12 6 3 0 8 8 6 2 5 1 9 10 5 5 7 1 10 12 9 3 6 1 11 14 10 4 5 1 12 9 9 0 3 0 13 10 8 2 6 1 14 14 10 4 4 0 15 17 16 1 1 0 16 9 6 3 2 0 17 9 8 1 1 0 18 13 7 6 6 1 19 8 4 4 4 0 20 13 8 5 7 1 Mean 12.15 8.65 3.50 4.15 St. dev. 3.92 3.38 2.71 2.03 ::: Homogeneous patient change analysis ----------------------------------- The first analysis presented is a repeated measures ANOVA with no grouping factor and 1-within patient factor, occasion, at 2-levels (baseline and follow-up) \[[@B29]\]. The results from this analysis are shown in Table [2](#T2){ref-type="table"}. The statistical analysis is equivalent to a paired t-test and the F-value of 33.49 is equal to the square of the paired t-value. The SRM \[[@B1]\] is typically defined as: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Repeated measures ANOVA with one within patient factor and no grouping factor ::: Source DF SS MS F, (p) --------------------- -------- ------------ -------- ------------------ Between Patients 19 439.60 23.14 Within Patients 20 192.00 Occasion 1 122.50 122.50 33.49, (\<0.001) Error 19 69.50 3.66 ::: ![](1477-7525-3-23-i1.gif) However, it can also be calculated from the repeated measures ANOVA shown in Table [2](#T2){ref-type="table"}: ![](1477-7525-3-23-i2.gif) where MSO is the mean square occasions, MSE is the mean square error, and n is the number of patients. Heterogeneous patient composition: between group analysis --------------------------------------------------------- This analysis is based on a repeated measures ANOVA with 1-between patient grouping factor at 2-levels (amount of change: a small amount or a large amount according to the dichotomized retrospective global rating of change) and the same within patient grouping factor as in the previous analysis \[[@B29]\]. The results are reported in Table [3](#T3){ref-type="table"}. The group-by-occasion interaction term represents the extent to which the two groups changed by different amounts. The F-value for this term, F~1,18~= 8.62, is the square of the t-value that would have been obtained had a t-test for independent sample means based on change scores been applied. Norman\'s S~repeat~\[[@B28]\] is calculated from the following information provided in Table [3](#T3){ref-type="table"}: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Repeated measures ANOVA with one within patient factor and one grouping factor ::: Source DF SS MS F, (p) Variance ---------------------- -------- ------------ -------- ------------------ -------------- Between Patients 19 439.60 Group 1 3.60 3.60 Error 18 436.00 24.22 Within Patients 20 192.00 Occasion 1 122.50 122.50 46.92, (\<0.001) Group by occasion 1 22.50 22.50 8.62, (0.009) 1.99 Error 18 47.00 2.61 2.61 ::: ![](1477-7525-3-23-i3.gif) Heterogeneous patient composition: among patient analysis --------------------------------------------------------- This analysis represents a correlation of change scores with patients\' retrospective global ratings of change. To show the location of the sources of variation, we generated the correlation coefficient from a regression analysis \[[@B30]\]. Also, we provide an intermediate analysis which replicates the previous identifiable subgroup analysis. Here, \"group\" was coded as a dummy variable (0 or 1): it is the dichotomized rating of change shown in Table [1](#T1){ref-type="table"}. Notice that the F-value in Table [4](#T4){ref-type="table"} is identical to that for the group-by-occasion interaction term reported in Table [3](#T3){ref-type="table"}. Table [5](#T5){ref-type="table"} presents the results from the correlation of change scores with the raw retrospective global ratings of change. ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Regression analysis with group as a dummy variable ::: Source DF SS MS F, (p) Correlation ------------ -------- -------- -------- --------------- ----------------- Regression 1 45.00 45.00 8.62, (0.009) r = 0.57 Residual 18 122.80 6.82 ::: ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Regression analysis with raw global rating change scores ::: Source DF SS MS F, (p) Correlation ------------ -------- -------- -------- ---------------- ----------------- Regression 1 56.30 56.30 12.25, (0.003) r = 0.64 Residual 18 139.00 4.60 ::: Source of conflict among analyses --------------------------------- An examination of the sum of squares terms (SS) in the ANOVA tables exposes the deficiency in applying these tests to the same dataset. Notice that when a repeated measures ANOVA with no grouping factor is applied, its SS error term contains both the group-by-occasion interaction term and the residual error from the repeated measures ANOVA with a grouping factor. Thus, to the extent that identifiable subgroups of patients exist, their presence drives down the magnitude of the SRM: the signal has become noise. The regression analyses reveal that this phenomenon extends to situations where patients truly differ in their change scores. Moreover, a comparison of the two regression analyses and correlation coefficients demonstrates that to the extent individual differences in change scores truly exist among patients, a between group analysis will under-estimate the ability of a measure to detect change. Reasons for \"Agreement\" among coefficients -------------------------------------------- A natural question is if the signal for the between group and among patient change scores is contained in the noise portion of the SRM, how is it possible to obtain a change coefficient that differs from zero for this analysis? There are at least three answers. First, the reported coefficients may not truly differ from zero. This statement is based on the observation that change coefficients are often presented as point estimates \[[@B2],[@B4],[@B5]\]. Without knowledge of a confidence interval or hypothesis test, one cannot ascertain the chance that a reported point estimate truly differs from zero. As a matter of interest, the 95% confidence interval for the reported SRM of 1.29 in our example is 0.91 to 1.92, confirming that it is highly likely that it differs from zero. The second explanation considers a situation similar to that of our data where the SRM is greater than zero. Although self-evident, it is important to acknowledge that investigators interested in evaluating a measure\'s ability to detect change select patients who, in most instances, are expected to truly improve. Accordingly, the mean change for the group will be greater than zero even when some patients remain stable or get worse. When the mean change is greater than zero, the SRM will be greater than zero, even when subgroups or individual patients truly change by different amounts. The third explanation addresses the situation where apparent patient differences in change scores, as represented by a correlation with another measure, are observed in a sample that is truly homogeneous with respect to change. In this case the design premise applied most frequently by investigators is that change scores on the measure under investigation will correlate with patients\' retrospective global ratings of change. To the extent that in clinical practice clinicians ask patients about their perceptions of change, this methodology seems reasonable. However, the major limitation associated with this approach is that it spuriously inflates the observed correlation coefficient. To understand the mechanism of this apparent association, a brief review of the relationship and assumptions of observed, true, and error scores is necessary \[[@B31]\]. In this example, the observed scores are those reported by patients on the measure under investigation and the retrospective global rating of change. True scores are unknown values that represent the scores that would be obtained in the absence of measurement error. Error scores are the differences between observed scores and true scores. The framework for comparing the change scores of a measure to the global rating of change is that of parallel assessments of the same attribute. A fundamental assumption is that the measure\'s error scores and the global rating\'s error scores are uncorrelated \[[@B31]\]. However, it is extremely unlikely that the error scores are independent when a patient provides both the measure\'s change score and that of the global rating \[[@B22]\]. The consequence is that the observed correlation will be greater than zero even when the correlation between the true scores is zero. Conclusion ========== The absence of a gold standard combined with multiple change coefficients has created uncertainty for those who investigate the sensitivity to change of health status measures. In an attempt to increase confidence in a measure\'s ability to detect change investigators have often reported multiple change coefficients derived from the same patient sample, the belief being that uniform findings among coefficients adds to the confidence in the results. We contend that this approach is inconsistent with theory: the signal for some coefficients is included in the noise of others. We suggest that rather than calculating multiple change coefficients, a more theoretically sound approach is to devote more preparatory work to determine the likely change characteristics of the patients of interest. Once the sample\'s change characteristic is established, the choice of change coefficient should be clear. Moreover, when the opportunity presents, investigators are encouraged to select the more rigorous designs which not only allow the assessment of change, but also challenge a measure\'s ability to differentiate among patients or groups of patients who change by different amounts. Conflict of interest ==================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= Both authors contributed to the conceptualization and writing of this manuscript.	PubMed Central	2024-06-05T03:55:55.865386	2005-4-5	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084357/", "journal": "Health Qual Life Outcomes. 2005 Apr 5; 3:23", "authors": [ { "first": "Paul W", "last": "Stratford" }, { "first": "Daniel L", "last": "Riddle" } ] }
PMC1084358	Introduction ============ Breast cancer is one of the most common cancers, it is estimated that one in eight women in the USA will develop breast cancer during their lifetime \[[@B1]-[@B4]\]. Furthermore, 25--30% of breast cancers are found in pre-menopausal women \[[@B1]\]. Currently mammography is the best available approach for the early detection of breast cancer in the general population with a sensitivity of 75--90% \[[@B2]\]. However, the positive predictive value is only 25% \[[@B3],[@B4]\]. In addition to mammography, non invasive new modalities have been developed to allow the early detection of breast cancer in all age groups and more importantly in young women with dense breast tissue and women who have high risk of developing breast cancer such as, women with strong family history and carriers of BRCA1 and/or BRCA2 genes. Currently, magnetic resonance imaging (MRI) is being studied for the early detection of breast cancer. Its sensitivity in high risk women has been found to be much higher than mammography but with a lower specificity \[[@B5],[@B6]\]. Kriege et al observed a higher sensitivity for MRI in detection of breast cancer in women with a genetic predisposition or at high risk compared to (71% vs. 41 %) but with lower specificity (90% vs. 95%) \[[@B6]\]. Electrical impedance scanning (EIS) is another modality under development for breast cancer detection especially in young women with dense breasts \[[@B7]\]. The basic science behind its use is the fact that malignant tumours have lower electrical impedance than the surrounding normal tissue. However, separation between malignant and benign lesions needs further investigations \[[@B8]\]. Furthermore, mammary ductoscopy (MD) and visualization of mammary ducts and proteomics of nipple aspirate fluid (NAF) and serum are promising screening modalities that require further evaluation \[[@B9],[@B10]\]. The limitations of mammography as a screening modality especially in young women with dense breasts necessitated the development of novel and more effective screening strategies with a high sensitivity and specificity. This article focuses on the dynamic thermal analysis as an evolving non invasive and a safe method in breast cancer detection in pre-menopausal women with dense breast tissue and women at high risk due to family history or genetic predisposition. Breast and circadian rhythm \[[@B1]\] ===================================== It is now recognised that the establishment and growth of a tumour depend on neovascularization. This successful recruitment of new blood vessels into a tumour; also known as angiogenesis is dependent on angiogenic growth factors produced by the tumour cells \[[@B11]\]. Such new vessels grow adjacent to the tumour presumably to increase its nutrient supply \[[@B12]\]. These new vessels lack smooth muscles rendering them unreceptive to control by epinephrine \[[@B13],[@B14]\]. The lack of receptivity produce a more constant blood flow, thus increasing the local temperature. Earlier technology for assessing thermal abnormalities in the breast focussed on the presence of the abnormal temperature as a crucial marker \[[@B15]-[@B17]\]. In a study conducted by Gantherine et al, 21.3% of patients who had abnormal thermograms but no abnormality on physical examination and mammography developed breast cancer within the next 3 years \[[@B17]\]. In another study of women who had thermal abnormalities on initial examination using infrared technology, long term follow up (2--10 years) revealed that 33% of these women developed breast cancer, a rate six times higher than that expected in the normal population \[[@B18]\]. This relationship between breast skin temperature and breast cancer was thoroughly examined by Gros et al \[[@B15],[@B16]\]. They found that the differences between the characteristics of rhythmic changes in skin temperature of clinically healthy and cancerous breasts were real and measurable. Despite these interesting observations thermography as a general screening tool for the detection of women at risk of breast cancer did not find a wide spread acceptance due to low sensitivity of the test and the subjective nature of the test interpretations. The superficial thermal patterns measured on the surface of the breast seem to be related to tissue metabolism and vascularization within the underlying tissue. Such thermal patterns change significantly as a result of normal phenomena including menstrual cycle, pregnancy and more importantly the pathologic process itself. Additionally, cancer development represents the summation of a large number of mutations that occur over years, each with its own particular histologic phenotype \[[@B19]-[@B23]\]. Such changes appear to generate their own thermal signature and the complexity of these signatures may be a reflection of their degree of development \[[@B24]-[@B28]\]. Temperature in a normal breast increases from the skin into the deep tissue and heat conductivity in the healthy breasts is constant in most cases and generally can be characterized in terms of circadian rhythm periodicity \[[@B29]\]. In contrast, the rhythms associated with malignant cells proliferation are largely non circadian and suggest that a circadian to ultradian shift may be a general correlation to neoplasia. Heat production by the tumour under the influence of angiogenesis should be therefore re-examined in terms of absence of normal circadian fluctuations. Due to the increased blood flow and the lack of receptivity in the newly formed vessels in malignancy, temperature production exhibits circadian rhythmic variations to a far lesser degree than is evident in the healthy breasts \[[@B13]\]. It has been found that independent of a tumour\'s size, relatively small tumours (\>/= 0.5 cm in diameter), poorly vascularized rapidly growing tumors can produce increases in regional heat. The explanation for this effect is unclear but it may be due to the chronic inflammatory response around developing breast tumours. With increasing evidence that inflammation can enhance tumor growth and is associated with a poor prognosis, this suggestion implies that thermal analysis may have considerable value \[[@B30]\]. Furthermore, the unique relationship between the thermal circadian rhythm and mitotic activity could be considered as a first warning of tumour development, which can be detected using a safe and non-invasive technology. The genes that drive the circadian rhythm are emerging as central players in gene regulation throughout the organism, particularly for cell-cycle regulatory genes and the genes of apoptosis \[[@B31]\]. Dynamic thermal analysis ======================== Recent technological advances have facilitated the recording of circadian rhythm variations of the breast and analysing the recorded data using highly complicated computer statistical software. A miniaturized microprocessor has been developed to record and store thermal information collected from eight separate sites of each breast. Sensors are placed in anatomically critical positions elicited by data obtained from tumour registries as to where cancers are most likely to develop. In the First Warning System (FWS, Lifeline Biotechnologies, Florida, USA), thermal data are collected every five minutes for a period of 48 hours during which time women are encouraged to maintain their daily activities. 9000 pieces of data are recorded by microprocessors during the test period and analysed using specially developed statistical software. Temperature points from each contralateral sensor are plotted against each other to form a thermal motion picture of a lesion\'s physiological activity. Such a technology was first used by Farrar et al who examined a cohort of 138 women who had been scheduled for open breast biopsies based on the finding of physical examination and mammography \[[@B14]\]. A total of 23 women (17%) were found to have breast cancer, of these, 20 (87%) were characterized by the monitor as being high risk. The other 3 patients (13%) who were missed by the monitor had ductal carcinoma. Mammography was positive or suspicious in only 19 patients (83%). Of the 4 cancers missed by mammography (3 of them were pre menopausal), the monitor correctly characterized 3 women as being high risk. Figures [1](#F1){ref-type="fig"} and [2](#F2){ref-type="fig"} demonstrate the thermal graphs in two patients with a fibroadenoma (Fig. [1](#F1){ref-type="fig"}) and a T1 breast cancer (Fig. [2](#F2){ref-type="fig"}). A neural net algorithm was subsequently developed and evaluated by the authors because of its value in analysing the non-linear data such as these recorded by the breast\'s monitors. Using this neural net algorithm reduced the number of false positives (18% vs. 30%)) and improved sensitivity (91% vs. 87%). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Dynamic thermal analysis in a patient with fibroadenoma. ::: ![](1477-7800-2-8-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Dynamic thermal analysis in a patient with T1 breast cancer. ::: ![](1477-7800-2-8-2) ::: One of the main challenges to this technology is the false positive cases; confusion could be created in these women who are characterized as being positive or high risk by dynamic thermal analysis in the absence of physical and mammographical signs. This group of women may or may not have cancer in its earliest stages. Further retrospective analysis of the thermal data using a refined neural net algorithm may increase the sensitivity and reduce the number of false positives. Also this group of patients may well benefit from the new advances in the nipple aspirate fluid analysis and proteomic profiling technologies. Research is currently ongoing on this subject and the initial results are promising \[[@B9]\]. The Future ========== Dynamic thermal analysis of the breast is a safe, non invasive approach that seems to be sensitive for the early detection of breast cancer especially in young women where the conventional mammography is of limited value. Such a technology could become the initial breast screening test in pre-menopausal women and those who are classified as positive can then be selected for anatomical imaging with mammography, MRI and/or ultrasonography. Further refinement of the neural net algorithm is required in order to shorten the period of data recording and improve specificity. Prospective multi-centre trials are then required to validate these promising observations. The issue of false positives require further investigation using molecular genetic markers of malignancy and novel techniques such as mammary ductoscopy \[[@B10]\]. Finally, a better understanding of the circadian rhythm biology \[[@B1],[@B30]\] and clearer definition of the thermal activity boundaries for various pathological conditions of the breast will open the door to a new and more precise screening method for breast cancer.	PubMed Central	2024-06-05T03:55:55.868035	2005-4-8	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084358/", "journal": "Int Semin Surg Oncol. 2005 Apr 8; 2:8", "authors": [ { "first": "M", "last": "Salhab" }, { "first": "W", "last": "Al Sarakbi" }, { "first": "K", "last": "Mokbel" } ] }
PMC1084359	1. Introduction =============== Autoimmune Thyroid Diseases (AITD) are one of the most frequent organ-specific autoimmune diseases affecting more than 3% of the total population worldwide. The clinical spectrum can be divided into two major subtypes, (i) glandular hyperfunction characterized by Graves\' disease, and (ii) glandular hypofunction as Hashimoto\'s thyroiditis. In both diseases, there is a breakdown in tolerance and the generation of an immunoglobulin G response directed against thyrotropin receptor (predominantly in Graves\' disease \[[@B1],[@B2]\]), thyroglobulin, and thyroid peroxidase (TPO). Autoantibodies (aAbs) against TPO, a common denominator of AITD, are present in 90% of Hashimoto\'s thyroiditis and 75% of Graves\' disease patients\' sera \[[@B3]\]. Since anti-TPO aAbs are present at high concentration in patients\' sera \[[@B4],[@B5]\], they are invaluable makers of the thyroid autoimmune response and thus extensively used to diagnose such pathologies. In vitroand in vivocytotoxic effector functions have also been described such as C3 complement activation \[[@B6]-[@B8]\] and antibody-dependent cell cytotoxicity \[[@B9]-[@B12]\] leading to the maintenance and amplification of thyroid cell destruction in Hashimoto\'s disease. Moreover, anti-TPO aAbs are likely to play a more important role in presenting TPO to T cells \[[@B5],[@B13],[@B14]\]. To understand the role of anti-TPO aAbs in the pathogenesis of AITD and to shed new light on how the TPO molecule is seen by the immune system, the delineation of anti-TPO B-cell epitopes has been the goal of several studies during the last decades. These findings enabled a better localization of the discontinuous immunodominant region (IDR) with the description of several amino acid residues taking part in this highly complex structure. All together, these data could be a great interest to rationally design competitors (such as peptides) which could be used in combination with other immunotherapies such as systemic antibody treatment, antigen-specific immunization or others generating antigen-specific regulatory T cells capable to block at least for a period of time, an ongoing autoimmune process and may synergize to delay hypothyroiditis. 2. TPO : structure and function =============================== TPO is a membrane-bound, glycosylated hemoprotein that plays a key role in thyroid hormone synthesis by catalyzing both the iodination of thyroglobulin and the coupling of some of the iodotyrosyl residues to generate the thyroid hormones T3 and T4. Human TPO (hTPO) is a 933 amino acid type I integral apical membrane protein that contains a large extra-cytoplasmic domain orientated toward the follicular lumen, a short membrane-spanning region, and a 61 amino acids cytoplasmic tail. The extra-cellular region consists of 848 amino acids and five potential glycosylation sites. Alignment studies and structural homologies have shown that hTPO is formed by three distinct domains: a myeloperoxidase (MPO)-like, a complement control protein (CCP)-like, and an epidermal growth factor (EGF)-like domains, from the N- to the C-terminal extremities \[[@B15]\] (Figure [1](#F1){ref-type="fig"}). The high homology between the MPO and TPO molecules has allowed the prediction of the secondary structure and domain organization of TPO \[[@B16]\] which is composed mainly of alpha-helical structure with relatively little beta-sheet structures. More recently, the structure of each domain has been partially elucidated by three-dimensional modeling \[[@B17],[@B18]\]. The three-dimensional structure of TPO, however, remains unknown, even though low resolution crystals have been obtained \[[@B19],[@B20]\]. TPO is known to exist as multiple species and migrates on SDS PAGE analysis as a closely spaced doublet, which derives from endogenous proteolysis \[[@B21],[@B22]\]. The multiplicity of N-terminal truncated forms of TPO contributing to the diversity of the polypeptide chain, and the flexibility observed for the hinge regions leading to an instable form of TPO, have made difficult to obtain high resolution diffracting crystals for three dimensional structural analysis. Probably, the only way to solve the three-dimensional structure of TPO will be to \"freeze\" the molecule in a stable conformation by crystallizing a complex TPO/anti-TPO Fab. This strategy would be also greatly informative to define all residues involved in the interaction. Nevertheless, because of the difficulty in obtaining diffracting crystals, other strategies, such as competition between anti-TPO aAbs, site-directed mutagenesis or construction of chimera, have been used to delineate B-cell autoepitopes on the TPO molecule \[[@B17],[@B23]\]. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Ribbon diagram of the structure of human TPO.(Reproduced with permission 14). The MPO- (residues 142--733), CCP- (residues 739--795) and EGF-like (residues 796--841) domains are indicated with a black arrow. This representation corresponds to the juxtaposition of the three-dimensional model of each domain. ::: ![](1740-2557-2-2-1) ::: 3. Anti-TPO autoantibodies and immunodominant region ==================================================== Anti-TPO aAbs are specific serological markers for diagnosing AITD. Anti-TPO aAbs are produced by B lymphocytes via a T cell-dependent mechanism. Since several years, it is well known that anti-TPO aAbs recognize conformational epitopes that are highly dependent on the three-dimensional structure of the TPO molecule \[[@B2]\]. These TPO aAbs are known to be restricted to two immunodominant regions (IDR) containing different but adjacent surface epitope. In 1989, Ruf and Carayon\'s group \[[@B24]\] pioneered in the localization on antigenic domains on hTPO. Using a panel of monoclonal antibodies (mAbs) with different affinities and epitopic specificities for hTPO, they studied the distribution of epitopes on the surface of the hTPO molecule. These mAbs were found to interact with four domains (termed A to D) on native TPO. But, in competitive experiments with sera from patients suffering from AITD, only the binding of mAbs interacting with the two major antigenic domains (named A and B) was highly inhibited by anti-TPO aAbs from patients\' sera, demonstrating that these two regions form the IDR on the hTPO. During the last two decades, an explosion in the development of combinatorial libraries technologies enable the production of anti-TPO human mAb fragment from several phage display antibody libraries constructed by using thyroid infiltrating B-cells from Graves\' disease or Hashimoto\'s thyroiditis patients. Using a panel of four anti-TPO antibody fragments (Fabs), named SP1-5, WR1-7, TR1-8 and TR1-9, Chazenbalk and co-workers \[[@B25]\] showed that these four recombinant human Fabs inhibit more than 90% of anti-TPO aAbs present in patients\' sera affected by AITD. Thus, they have been concluded that these four Fabs recognize epitopes that delineate the IDR of hTPO. Interestingly, the IDR was found to be composed by two overlapping regions (named A and B domains) as previously described with the mouse mAbs (but the domains were described with an inverted nomenclature) \[[@B25]-[@B27]\]. Our group also contributed to increase the number and diversity of human monoclonal anti-TPO aAbs using the phage display technology. Forty-four human anti-TPO aAbs were selected from one \"in cell\" library and three random combinatorial libraries constructed from different B-cell subsets extracted from thyroid biopsies from Graves\' disease patients \[[@B28],[@B29]\]. These recombinant human anti-TPO aAbs, mimicking the human repertoire, strongly displace the binding of anti-TPO aAbs present in patients\' sera (80% inhibition) to hTPO, and predominantly recognize the IDR. Competitive experiments between human anti-TPO aAbs from patients\' sera and anti-TPO aAbs (human Fabs or mouse mAbs), were used to draw contours of the IDR/A and IDR/B on the TPO molecule. However competition studies do not allow the precise localization of the IDR as well as the identification of the amino acid residues involved in the interaction. Consequently, other approaches were investigated with the aim of (i) defining the contribution of each domains (A and B) in the IDR, (ii) identifying key residues involved in the IDR/A and IDR/B-specific human anti-TPO aAbs epitopes, and (iii) understanding the relationship that could exist between both domains. 4. Contribution of the MPO-like and CCP-like domains in the IDR =============================================================== The MPO-like domain has been clearly found to contribute in the folding of the IDR (IDR/A and IDR/B). It is important to note that the A domain defined by the murine mAbs corresponds to the B domain defined by the human Fabs and vice versa. Since we have used human anti-TPO aAb in our experiments as did McLachlan and Rapoport\'s group \[[@B30]\], we use here the nomenclature of the IDR/A and B defined by this group. Banga and Gardas\' groups chose to produce a panel of polyclonal anti-sera by immunizing rabbits with a series of overlapping or adjacent peptide sequences exposed at the surface of TPO (based upon the TPO model). By using a rabbit antiserum against the peptide P14 (aa 599--617), they successfully located the IDR/A epitope on hTPO in the MPO-like domain. Importantly, this region is included in the recombinant fragment of hTPO (aa 589--633) which was previously identified as an autoantigenic determinant \[[@B31]\]. Amongst the panel of mAbs produced by Ruf and Carayon\'s group \[[@B24]\], only mAb 47 binds to both, denatured and native TPO. This mAb recognizes a linear epitope first located between amino acid residues 713--721 \[[@B32]\] and later restricted to residues 713--717 \[[@B33]\]. Interestingly, mAb 47 was shown to strongly competes with IDR/B-specific human anti-TPO aAbs thus defining for the first time a short immunoreactive IDR/B-specific region \[[@B30],[@B37]\]. Furthermore, mutations between the positions 713--717 specifically affected the binding of patients\' sera to hTPO, definitively demonstrating that this region in MPO-like domain is involved in the IDR/B epitope. In the past five years, three other regions (210--225, 353--363, 549--563) \[[@B15],[@B33],[@B34]\], structuring the IDR/B have been identified in the MPO-like region (Figure [2A](#F2){ref-type="fig"}). Finally using a chimeric molecule between TPO and MPO molecules, the 121 amino-terminal residues of TPO were excluded from the IDR \[[@B35]\]. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Location of the regions participating in the IDR.(A) Regions involved in the binding with aAbs are indicated on a three-dimensional diagram of the human TPO : positions 353--363 in yellow, 377--386 in red, 506--514 in green, 713--720 in blue, 599--617 in orange in the MPO-like domain, 737--740 in pink, and 766--775 in orange in the CCP-like domain. The flexibility of the hinge regions is represented by a white arrow. (B) Ribbon diagram representing the possible folding of the CCP-like domain onto the MPO-like domain. An immunodominant binding surface is virtually represented by three dotted lines and the distance (in Å) between the regions 377--387 and 713--720 is shown. The model was adjusted by using Swiss-PDB viewer 3.7b2 freeware available at <http://www.expasy.ch/spdbv>. ::: ![](1740-2557-2-2-2) ::: On the other hand, the participation of the CCP-like domain in the IDR was more difficult to demonstrate. In 1998, Estienne et al, \[[@B36]\] located a conformational B-cell epitope at the C-terminal end of TPO near the membrane anchorage domain of the molecule (amino acids 742--848). Since the majority of anti-TPO aAbs from Hashimoto patients\' sera recognized a recombinant polypeptide encompassing this region, the authors thought that this region was part of the IDR. One year later, using reticulocyte lysate cell-free translation approach of TPO, Grennan et al\[[@B37]\] observed that truncation of TPO downstream of amino acid residue 771 had little effect on aAb recognition in all patients\' sera tested, suggesting that the IDR recognized by human aAbs should lie between amino acid residues 742 and 771. This observation was in agreement with the early study from Estienne et al, but later, Xiong et al\[[@B38]\], using the cell-free translation approach, concluded that the region 742--771 does not contain the IDR. These contradictory data underline the difficulty of determining the contribution of the CCP-like domain in the IDR, specially when the CCP-like domain is expressed and studied outside the three-dimensional context of the hTPO (as exemplified with the in vitrotranslation experiments). To overcome such problems, and with the aim to localize the discontinuous IDR of the TPO, we have combined two technological advances (i) the selection of mimotopes by screening phage display peptides libraries on an IDR/B-specific human recombinant anti-TPO aAb (T13) mimicking aAbs from patients\' sera, and (ii) the sequence alignment of the selected mimotopes on the primary sequence of hTPO \[[@B33]\]. We identified four distinct regions distributed between the MPO- and CCP-like domains. The sequences 353--363, 377--386, and 713--720, are located in the MPO-like domain while the last one, in position 766--775, is located in the CCP-like domain. More interestingly, mutation of the region 766--775 abrogated the binding of human anti-TPO aAbs from Hashimoto\'s and Graves\' disease patients. In the contrary, the EGF-like domain of the TPO molecule was clearly excluded from the IDR \[[@B39]\], as well as the contact region between the two TPO monomers during the dimerisation of the molecule \[[@B35]\]. All together, these data pointed out (i) the participation of several regions in the MPO-like domain and at least one region (766--775) in the CCP-like domain for the folding of the IDR and (ii) that one major region (599--617) belongs to the IDR/A and five regions (210--225, 353--363, 549--563, 713--720, and 766--775) structure the IDR/B. This emphasizes the discontinuous nature as well as the complexity of the IDR and provides new insights into the MPO-like and CCP-like domains positioning. It is now obvious that the MPO-like and CCP-like domains have to be closed in the three-dimensional structure of hTPO to form an immunodominant binding surface recognized by the majority of anti-TPO aAbs, thus one could hypothesize that the CCP-like domain may be able to fold onto the MPO-like domain to enable the binding of a single aAb simultaneously to the MPO- and CCP-like domains (Figure [2B](#F2){ref-type="fig"}) \[[@B33]\]. 5. Amino acid residues structuring the IDR ========================================== To improve our knowledge of IDR and to understand how the TPO is seen by the immune system during the pathogenesis of AITD, it was also important to determine the amino acid residues involved in the epitopes recognized by human anti-TPO aAbs. Among the major regions identified over time, contributing amino acid residues have been identified. By expressing recombinant hTPO mutants with single amino acid replacements, several groups participated to the characterization of critical amino acid residues. \[[@B17],[@B34],[@B40],[@B41]\]. Although, seven key residues (R225, Y772, and K713 to D717) involved in the IDR/B-specific anti-TPO aAbs epitope were identified, only one amino acid residue (K627) has been clearly assigned to the IDR/A-specific human anti-TPO aAb epitope. 6. Respective positioning of IDR/A and IDR/B ============================================ The two main discontinuous IDR (known as IDR/A and IDR/B) have been studied for several years to identify the TPO amino acid sequences involved in the binding of human anti-TPO aAbs. Over time, a body of evidence demonstrated that IDR/A and IDR/B, although different, overlap in part \[[@B25],[@B27]\]. Thus, in spite of the growing number of data available for delimiting each domain (A and B) forming the entire IDR, it was unclear whether these domains (i) form two separate clusters recognized by either IDR/A- or IDR/B-specific human anti-TPO aAbs or (ii) are structurally similar (involving the same surface area on the TPO molecule). Using all the reference anti-TPO Abs (recombinant human Fabs, mouse mAbs and rabbit polyclonal anti-peptides), with the same approach, the relationship between IDR/A and IDR/B domains has been investigated \[[@B42]\]. We have demonstrated that human IDR/A and IDR/B-specific anti-TPO aAbs recognize essentially the same region on the surface of hTPO. However, IDR/A-specific aAbs use the region 599--617 as anchor point for their binding to hTPO much more forcefully (Figure [3](#F3){ref-type="fig"}). ::: {#F3 .fig} Figure 3 ::: {.caption} ###### IDR/A and IDR/B-specific anti TPO aAbs mainly recognize the same regions on hTPO: the binding strength makes the difference.IDR/A- and IDR/B- specific epitopes (left and right panel respectively) are globally composed by the same peptidic regions on hTPO. Region 599--617 (in orange) represents the amino acid sequence predominantly recognized by IDR/A-specific anti-TPO aAbs (as illustrated by three yellow lines on the left panel). On the hand, IDR/B-specific anti-TPO aAbs bind several regions widely spread on the IDR. The immunodominant binding surface (in blue) and region 599--617, composed the IDR/B-specific epitopes as shown by the yellow lines which illustrate the contact points on hTPO. As not shown in the figure, the CCP-like domain may contribute to the immunodominant epitopes. ::: ![](1740-2557-2-2-3) ::: Such observations emphasize, once again, the high complexity of the IDR on the TPO molecule. It has been now well demonstrated that the IDR is stretched along the MPO- and CCP-like domains of hTPO. Consequently, some regions such as the region 599--617, 713--720, and 772--775 seem to be too far, on the three dimensional model of hTPO, to compose a single epitope, albeit discontinuous. Furthermore, a rapid inspection of the three dimensional TPO model shows that region 599--617 (the main anchor point for the IDR/A-specific aAbs) is located on the opposite side of the MPO-like domain with regard to the immunodominant binding surface composed of three auto-reactive TPO peptide sequences (353--363, 713--720 or 766--775) (Figure [2B](#F2){ref-type="fig"}). Until now, the only three-dimensional structure of hTPO known is a computer model and the whole structure of hTPO could be more compact than the model suggested. Thus, a flexibility in the hinge region between the MPO-, CCP- and EGF-like domains would lead to the formation of a more closely folded molecule \[[@B17],[@B33]\]. Furthermore, the membrane-bound TPO, at the surface of thyrocytes as well as CHO cells, exists as a disulfide-linked dimer \[[@B43]\]. Thus the presence of TPO dimer could favor a spatial gathering of the IDRs recognized by IDR/A and IDR/B-specific anti-TPO aAbs. Since, anti-TPO aAbs epitopes are largely restricted to one facet of the native molecule \[[@B24],[@B25],[@B33]\] this may enable the formation of a dimer by the other side, as previously modeled by homology with the crystal structure of the MPO molecule \[[@B15]\]. 7. Concluding remarks ===================== Numerous efforts have been made to precisely characterize the regions, as well as the amino acid residues forming the IDR of hTPO. This could lead to the rational design of therapeutic peptides able to modulate or block immune responses such as antigen presentation. Thus, it would be of great interest to determine in vivo, by using animals models for AITD, whether it will be possible to influence the diseases\' course (rapidity, severity, etc\...) by using these peptidic immunomodulators interacting with antigen presentation of B-cells to T-cells. However, a major question remains: What is the clinical significance of the anti-TPO aAbs which are present in the two well characterized AITD (Graves\' disease and Hashimoto\'s thyroiditis)? Although, there is a general consensus that one of the most important markers of thyroid autoimmunity is the production of such thyroid-specific aAbs. Their role during the pathogenesis of AITD is still controversial. Several lines of evidence show that there is no difference in the epitope recognized by anti-TPO aAbs produced during Graves\' disease, Hashimoto\'s thyroiditis and Euthyroiditis \[[@B44]-[@B46]\], and it is well-known that epitope spreading phenomenon does not occur in patients suffering from these diseases, at least for the B-cell epitopes \[[@B5]\]. Since, the components of the IDR/A and IDR/B of hTPO are now well characterized, it would be interesting to evaluate by longitudinal studies, whether the epitopic pattern of individual patients is changing (from an IDR/A to IDR/B pattern for example or vice versa) during the course of the diseases or after treatment. List of Abbreviations ===================== Thyroid Peroxidase; TPO, Autoantibody; aAb, Autoantigen; aAg, Immunodominant Region ; IDR, Autoimmune Thyroid Disease; AITD, Complement Control Protein; CCP, Myeloperoxidase; MPO, Epidermal Growth Factor; EGF. Competing interests =================== The author(s) declare that they have no competing interests. Acknowledgements ================ The authors thank Dr. I. Robbins for carefully reading the manuscript. D. Bresson is a recipient of a postdoctoral fellowship from the Juvenile Diabetes Research Foundation. S.A. Rebuffat is financed by the Ligue Nationale Contre le Cancer.	PubMed Central	2024-06-05T03:55:55.869050	2005-3-15	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084359/", "journal": "J Autoimmune Dis. 2005 Mar 15; 2:2", "authors": [ { "first": "Damien", "last": "Bresson" }, { "first": "Sandra A", "last": "Rebuffat" }, { "first": "Sylvie", "last": "Péraldi-Roux" } ] }
PMC1084360	Introduction ============ Infertility patients are a vulnerable group that often seek a non-medical solution for their failure to conceive. The Cochrane Collaboration\'s definition of Complementary and Alternative Medicine (CAM) is \"a broad domain of healing resources that encompasses all health systems, modalities, and practices and their accompanying theories and beliefs other than those intrinsic to the politically dominant health systems of a particular society or culture in a given historical period\". This survey sought to compare use of CAM by infertile couples in NHS and private settings. Methods ======= The University of Bristol operates clinics in two settings: one is an NHS reproductive medicine clinic based at St. Michael\'s Hospital which treats \~500 new couples each year. Patients are seen for routine investigation and treatment of infertility, but no assisted conception is provided at this facility. The other clinic is a non-NHS office at the Centre for Reproductive Medicine, where patients are seen for private investigation and treatment of their infertility including in vitro fertilisation and donor insemination. This private service provides between 500--600 cycles of assisted conception and \~220 cycles of donor insemination each year. Between February and April 2001, all patients (men and women) attending a routine visit at either center were invited to complete an anonymous questionnaire regarding the patient\'s use of CAM (see additional file [1](#S1){ref-type="supplementary-material"}). The invitation was offered until 200 patients in each location had completed the survey. Although couples often attend together, men and women were approached separately. Each study participant completed only one questionnaire. Fishers exact test was used for comparison of proportions between groups. Results ======= The survey was simple to complete and the receptionist asked the patients for their replies at the end of the appointment, hence response rates in both services were high: 181 respondents from 200 NHS patients (124 women and 57 men) and 157 respondents of 200 private patients (120 women and 37 men), or 83% overall. While these clinics only treat couples, for many of the routine visits it is the female partner who comes unaccompanied for procedures such as ultrasound scanning, post coital testing and donor insemination, thus explaining the higher frequency of women responders. Mean age of respondents was 36 and 34 for men and women in the private setting, respectively, and 39.8 and 35.5 in the NHS setting. Overall, in the private sector 13% of men and 40% of women had used CAM for their infertility compared with 12% of men and 23% of women in the NHS clinic (see table). It was also noted that overall 19% of patients had used CAM for other health problems in the past Of patients who had used CAM, 10% thought it had been helpful for infertility, 13% felt it had helped them psychologically and that they had done everything possible, and 22% felt it had helped them to relax. In the space offered for additional comments (free text), several respondents indicated that they would like advice from their doctor specifically about CAM and infertility. Discussion and Conclusion ========================= Infertility patients in our survey accessed CAM for their infertility more frequently than the overall use of CAM in the general population estimated by telephone and postal surveys. In 1999, a BBC telephone poll of 1204 randomly selected British adults were asked about the use of CAM in the preceding year \[[@B1]\]. Although the poll did not ask whether CAM had been accessed via a practitioner or by over the counter sales, it did reveal that 20% responders had accessed CAM over the preceding year. The most commonly used CAM was herbalism (34%), followed by aromatherapy (21%), homeopathy (17%), acupuncture (14%), reflexology (6%) and massage (6%). A postal survey of 5,010 randomly selected adults (53% response rate) showed that 13.6% had visited a CAM therapist in the preceding 12 months and 28.3% had bought an over the counter remedy and / or seen a CAM therapist \[[@B2]\]. As the use of CAM in the past by the infertile patients in our survey was consistent with the above reports, it seems reasonable to suggest that infertile patients in general make a greater use of CAM for their infertility than the general population. We observed the highest use of CAM among women from the private clinic (40%). This may reflect a greater ability of these couples to afford the cost of CAM, even though they have the additional cost of their fertility treatment. There are various forms of CAM that may offer different things to different people. Although the CAM therapies used by the infertility patients are presented in table [1](#T1){ref-type="table"}, there are many others including: acupressure, chiropractor, naturopathy, cranial osteopathy, osteopathy, Alexander technique, environmental medicine, kinesiology, Reiki, anthroposophic medicine, aromatherapy, autogenic training, visualisation, shiatsu, ayurveda massage, therapeutic touch, mediation and yoga. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Use of complementary medicine for infertility as measured by questionnaire among NHS and private clinic patients (n= 400). ::: NHS Male n= 57 NHS female n= 124 Private male n= 37 Private female n= 120 ------------------------------ ------------------ --------------------- ---------------------- ------------------------- nutritional advice 2 (0.3) 9 (7) 2 (5) 21 (18) Reflexology 0 11 (9) 1 (3) 19 (16) Acupuncture 1 (0.2) 10 (8) 1 (3) 15 (13) traditional Chinese medicine 0 (0.3) 5 (4) 1 (3) 6 (5) herbalism 2 (0.3) 6 (5) 0 6 (5) hypnosis 0 1 (1) 0 5 (4) spiritual healing 2 (0.3) 5 (4) 0 2 (2) CAM (all types) 7 (12) 29 (23) 5 (13)^a^ 48 (40)^b^ Note: data presented as patient number (%) ^a^CAM use by females at private clinic vs. males at private clinic, p\< 0.01 by Fishers exact test. ^b^CAM use by females at private clinic vs. females at NHS, p\< 0.01 by Fishers exact test ::: A search of the CISCOM Complementary Medicine database (see <http://www.rccm.org.uk/ciscom/>) revealed 105 reporting on different types of CAM for infertility and miscarriage. However, the studies were generally of a poor quality with no study providing prospective randomised controlled evidence of clinical efficacy for any form of CAM to improve the prognosis for infertile couples. Given this paucity of data, there is great need for properly conducted and appropriately controlled research in this area. Objective evaluation of the effectiveness of CAM in infertility would help to answer the questions that prompted this survey, i.e., is CAM effective, and how many people are using it? It is important that physicians be familiar with and consider CAM; patients should also understand the importance of evaluating efficacy of any intervention, including CAM. It is interesting to speculate why infertility patients would access CAM despite the lack of evidence of efficacy. Indeed, the patients in our study population registered scepticism regarding CAM themselves. Apart from pregnancy, there are other valuable aspects to any infertility treatment. Quality of life measures and sense of well-being are valid, especially in a disability which may become chronic, such as infertility. This condition has a deeply distressing impact on how a woman or man feels about her/himself at the level of core identity. Patients often describe an encounter with a CAM provider in terms of a someone who was \"really interested\", a person \"who listened really carefully to what I was saying\" and \"who seemed to understand how I feel\". From this, it may be offered that traditional doctors would benefit by refining listening and counselling skills -- an integral part of training, especially in general practice and now recognised as a core skill in RCOG and other specialist training. These skills can be enhanced by Balint groups and seminars offered by the Institute of Psychosexual Medicine, as well as formal training in psychotherapy. While counselling can address some of these issues, doctors should strive to listen more carefully to patients so the consultation experience is satisfying and possibly, in itself, therapeutic. The effect of laying on of hands should not be underestimated; as physicians, we have the privilege of examining the bodies as well as the minds of our patients. Although it is important to ensure the treatment we provide to patients is safe and effective, it remains vital to consider the patients foremost when managing infertility. As CAM is used so frequently (unsubstantiated claims of efficacy notwithstanding) there is a clear need for further research on this topic \[[@B3]\]. However, treatment offered for infertility should be patient, not doctor, centred. Patients should be treated holistically, respecting their own views and moral/ ethical framework. Our study suggests that CAM may be addressing a need that is not fully met by traditional medical practices. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= CC and JJJ contributed equally to this manuscript. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Questionnaire on Alternative or Complementary Medicine ::: ::: {.caption} ###### Click here for file :::	PubMed Central	2024-06-05T03:55:55.870999	2005-4-4	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084360/", "journal": "J Exp Clin Assist Reprod. 2005 Apr 4; 2:5", "authors": [ { "first": "Catherine", "last": "Coulson" }, { "first": "Julian", "last": "Jenkins" } ] }
PMC1084361	Introduction ============ The analysis of the transcriptome of Plasmodium falciparumhas revealed that during the intraerythrocytic development of the parasite, genes coding for enzymes and proteins that are involved in complex cellular functions such as transcription, replication or energy metabolism, each involving many gene products, are transcribed in a coordinated fashion, supporting the notion that all components must be present at the right time to allow for optimal function \[[@B1]-[@B3]\]. While this may be true in general, it has already been found that scrutinizing the details of specific metabolic functions reveal some significant departures from this paradigm \[[@B4]\]. Such scrutiny has also provided some intriguing peculiarities that through detailed biochemical studies may reveal some parasite-specific functions that may enlighten our understanding of parasitism or even provide for novel targets for chemotherapeutic intervention. The present analysis explores of the transcriptome to fathom additional metabolic pathways. During the erythrocytic stage the malaria parasite is engaged in intensive synthesis of nucleotides and is subjected to endogenously produced oxidative radicals that must be detoxified. Like other cells, in order to perform its anabolism, the parasite needs not only energy (ATP): it also needs reducing power, under the form of NADPH. Enzymes that function primarily in the reductive direction utilize NADP^+^/NADPH pair as co-factors as opposed to oxidative enzymes that utilize the NAD^+^/NADH cofactor pair. The conversion of ribonucleotides to deoxyribonucleotides (through the action of ribonucleotide reductase) requires NADPH as the electron source. Thus, any cell that proliferates rapidly requires large quantities of NADPH. NADPH can be produced during glucose-6-phosphate oxidation through the pentose-phosphate pathway (PPP; Figure [1](#F1){ref-type="fig"}). This pathway also produces ribose-5-phosphate (R5P), the sugar component of nucleic acids. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Pentose phosphate pathway. The no-entry symbol indicates that the gene coding for this enzyme could not be found in the genome of P. falciparum, but biochemical evidence suggests that the enzyme activity must be present. P represents phosphate. ::: ![](1475-2875-4-17-1) ::: The reactions of the PPP operate exclusively in the cytoplasm. PPP has both an oxidative and a non-oxidative arm. The oxidation steps, utilizing glucose-6-phosphate (G6P) as the substrate, occur at the beginning of the pathway and are the reactions that generate NADPH. Thus, the first carbon of glucose-6-phosphate is first oxidized to a lactone (catalyzed by glucose-6-phosphate dehydrogenase) concomitantly releasing two electrons that reduce one molecule of NADP^+^to NADPH. The ensuing decarboxylation of 6-phospho-D-gluconate (catalyzed by 6-phosphogluconate dehydrogenase) releases two additional electrons, which reduce a second molecule of NADP^+^. A five-carbon sugar, D-ribulose-5-phosphate, is produced in the reaction. By isomerization, D-ribulose-5-phosphate is transformed into D-ribose-5-phosphate (R5P). To be used in nucleic acid synthesis, R5P is transformed into 5-Phospho-?-D-ribose 1-pyrophosphoric acid (PRPP) by ribose-phosphate diphosphokinase (EC: 2.7.6.1). The non-oxidative reactions of the PPP are primarily designed to generate R5P. Equally important reactions of the PPP are to convert dietary 5 carbon sugars or D-ribose-1-phosphate generated in the salvage of purines (that can be slowly converted to R5P by phosphoglucomutase; EC: 5.4.2.2) into both 6 (fructose-6-phosphate) and 3 (glyceraldehyde-3-phosphate) carbon sugars which can then be utilized by the pathways of glycolysis. In the first reaction, R5P will accept two carbon atoms from xylulose-5-phosphate (obtained by epimerization of ribulose-5-P), yielding sedoheptulose-7-phosphate and glyceraldehyde-3-phosphate (catalyzed by transketolase). Sedoheptulose-7-phosphate transfers three carbons to glyceraldehyde-3-phosphate (catalyzed by transaldolase), yielding fructose-6-phosphate (F6P) and erythrose-4-phosphate. Erythrose-4-phosphate then accepts two carbon atoms from a second molecule of xylulose-5-phosphate (catalyzed again by transketolase), yielding a second molecule of F6P and a glyceraldehyde-3-P (GAP) molecule. F6P and a GAP can then enter glycolysis and eventually produce ATP. The intermediate erythrose-4-phosphate is a substrate for the shikimate pathway <http://sites.huji.ac.il/malaria/maps/shikimatebiopath.html>. The non-oxidative part of PPP can also work in the reverse direction utilizing fructose-6-phosphate glyceraldehyde-3-phosphate generated by glycolysis to produce ribose-5-phosphate. In essence, the PPP can operate in different modes (Figure [2](#F2){ref-type="fig"}). When both R5P and NADPH are needed (Figure [2](#F2){ref-type="fig"} mode 1) only the oxidative path will operate: all the ribulose 5-phosphate is isomerized to R5P and the pathway is completed. When more ribose-5-phosphate than NADPH is needed (Figure [2](#F2){ref-type="fig"} mode 2), R5P is synthesized by a reversion of the non-oxidative path from F6P and GAP generated by glycolysis. When the cell needs both NADPH and ATP but not R5P (Figure [2](#F2){ref-type="fig"} mode 3): ribulose-5-phosphate is converted to F6P and GAP for glycolysis to make ATP and NADH. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Different modes of PPP action. The PPP can function in different modes depending on the needs of the cell. Mode 1: Both ribose-5-phosphate and NADPH needed -- predominating reactions are shown. -- All the ribulose 5-phosphate is isomerized to ribose 5-phosphate, which is used for the synthesis of PRPP. Mode 2: More ribose-5-phosphate is needed than NADPH Ribose 5-phosphate is synthesized by the non-oxidative arm using fructose-6-phosphate and glyceraldehydes-3-phosphate supplied by glycolysis. Mode 3: The cell needs NADPH and ATP but notribose-5-phosphate. Ribulose-5-phosphate is converted to fructose-6-phosphate and glyceraldehydes-3-phosphate which are channeled into glycolysis ::: ![](1475-2875-4-17-2) ::: The genes that code for enzymes participating in ancillary processes that produce NADP^+^and thiamine pyrophosphate, that serves as a co-factor for transketolase activity, should be expressed in coordination with the enzymes of the PPP. The details of the mentioned pathways can be also grasped at (<http://sites.huji.ac.il/malaria/maps/nicotinatemetpath.html>; <http://sites.huji.ac.il/malaria/maps/thiaminemetpath.html>, respectively) and the time-dependent transcription of the genes coding for the different enzymes will be discussed below. PPP activity in P. falciparum-infected erythrocytes has been measured \[[@B5]-[@B7]\]. The reverse activity of the non-oxidative arm of PPP has also been demonstrated by measuring the incorporation of radiolabel from \[1-^14^C\]glucose into nucleotides \[[@B6],[@B8]\]. The former investigators have shown that 4/5 of the glucose incorporated into parasite nucleic acids comes from the condensation of F6P and GAP in the reverse action of this arm. Atamna et al, reported that infected cells have large increase of PPP activity where 82 % is contributed by the parasite while the host cell\'s PPP activity is activated some 24-fold as a result of the oxidative stress that the parasite generates and impinges on the host cell \[[@B9]\]. The gene coding for G6PD has been cloned \[[@B10]\] and the biochemical properties of the isolated enzyme have been characterized \[[@B11],[@B12]\]. Molecular investigations have revealed that G6PD is coded by a hybrid gene that contains also the sequence of 6-phosphogluconolactonase (the second enzyme of PPP) \[[@B13],[@B14]\]. Activity of 6-phosphogluconate dehydrogenase in Plasmodium-infected erythrocytes has been detected indirectly and alluded to a parasite enzyme, but not characterized \[[@B15]-[@B17]\]. Ribose-phosphate diphosphokinase of P. falciparumactivity has been characterized and levels of its product 5\'-phosphoribosyl-pyrophospate (PRPP) were measured in P. falciparum-infectederythrocytes \[[@B8],[@B18]\]. The levels of PRPP were found to be increased 56-fold in infected cells at the trohozoite stage compared to uninfected erythrocytes. The parasite contains substantial levels of the pyridine nucleotides NAD^+^and NADP^+^and their reduced forms \[[@B19]\] and jointly with the presence of genes coding for the enzymes necessary for their synthesis in the genome of P. falciparum(but see below the discussion concerning the absence of NAD^+^kinase-coding gene), indicate the ability of the parasite to produce these co-factors. NADP^+^/NADPH are used by several biochemical reactions in the parasite (Table [1](#T1){ref-type="table"}), but the major role of NADPH is probably in the antioxidant defense of the parasite \[[@B20],[@B21]\] (see <http://sites.huji.ac.il/malaria/maps/redoxmetpath.html> for details). Interestingly enough, the biochemical (such as the intermediate products, enzyme activities for each step, timing of production?) details of NAD(P)^+^synthesis in the parasite such as the intermediate products, enzyme activities for each step, timing of production, were not investigated. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### NADP utilizing enzymes Enzymes are arranged by their sequential functional order. They are given by their name, their EC numbers, the gene identification (PfID) in the Plasmodiumgenome database (PlasmoDB), the time (in hours post invasion (HPI)) in the parasite\'s developmental cycle when they are maximally transcribed obtained from the IDC database and the metabolic function of the enzyme. ::: Enzyme EC number PfID Peak (HPI) Metabolic function ------------------------------------------------- ----------- --------------------- ------------ ------------------------ 3-oxoacy1-\[acy1-carrier-protein\] reductase 1.1.1.100 PFI1125c 28 Fatty acid synthesis 1-deoxy-D-xylulose-5-phosphate reductoisomerase 1.1.1.267 PF14\_0641 30 Isoprenoid metabolism Isocitrate dehydrogenase (NADP+); 1.1.1.42 PF13\_0242 28 TCA cycle Phosphogluconate dehydrogenase 1.1.1.44 PF14\_0520 26 PPP Glucose-6-phosphate 1-dehydrogenase; 1.1.1.49 PF14\_0511 48 PPP Pyruvate dehydrogenase (acetyltransferring). 1.2.4.1 PF11 0256 PF14 0441 28 28 Pyruvate metabolism Glutamate dehydrogenase (NADP) 1.4.1.4 PF14 0164 PF14 0286 1 30 Glutamate metabolism Pyrroline-5-carboxylate reductase 1.5.1.2 MAL13P1.284 18 Proline metabolism Dihydrofolate reductase 1.5.1.3 PFD0830w 26 Folate synthesis NAD(P)+ transhydrogenase (B-specific); 1.6.1.1 PF14\_0508 32 ? NADPH hemoprotein reductase 1.6.2.4 PFI1140w 13 Flavoprotein reduction Dihydrolipoamide dehydrogenase 1.8.1.4 PF08 0066 PFL1550w 28 11 Pyruvate metabolism Dihydrolipoamide S-acetyltransferase; 2.3.1.12 PF10 0407 28 Pyruvate metabolism ::: In this in silicoanalysis, the stage-dependent transcription of genes that code for enzymes that are involved in the PPP activity of the parasite will be analyzed in a functional context. Materials and methods ===================== The expression data used in this study was obtained from the transcriptome database <http://malaria.ucsf.edu/> of the P. falciparumintraerythrocytic developmental cycle as described \[[@B2]\]. This database contains the relative mRNA abundance for every hour of the intraerythrocytic cycle of parasite development based on the 70-mer oligonucleotide microarray \[[@B22]\]. The expression profile of each transcript is represented by an array of ratios between the mRNA level in the time point sample versus a fixed mRNA level in the control RNA pool. Loess smoothed expression profiles used in this study were calculated as described (see \[[@B2]\]). The peak hour induction was calculated from these profiles. This represents more accurate estimate of the timing of maximal expression since the loess smoothing provides and average mRNA abundance values over a time interval, correcting possible minor fluctuations in the raw value profiles. Each expression profile was subsequently normalized by its peak value. It is assumed that for each metabolic pathway there is a stoichiometric relationship between its individual enzymes and therefore the normalized values are more meaningful for functional evaluation. All enzymes discussed in this essay with their EC numbers and gene ID\'s are shown in Table [2](#T2){ref-type="table"}. Amplification of transcription was calculated by dividing the maximum value by the minimum value of the profile of relative mRNA abundance. A fair agreement between the time-dependent transcription data of the DeRisi transcriptome and Winzeler\'s data has been observed (data not shown). ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Enzymes of the pentose phosphate pathway and pyridine nucleotide metabolism. Enzymes are grouped into pathways. They are given by their name, their EC numbers, the gene identification (PfID) in the Plasmodiumgenome database (PlasmoDB)), the time (in hours post invasion (HPI)) in the parasite\'s developmental cycle when they are maximally transcribed obtained from the IDC database and the metabolic function of the enzyme. ::: Pathway Enzyme EC number Pf ID Peak (HPI) -------------------------------- ---------------------------------------------------------------- ----------------- --------------------- ------------ Pentose phosphate pathway Glucose-6-phosphate dehydrogenase 1.1.1.49 PF14\_0511 48 6-phosphogluconolactonase 3.1.1.31 PF14 0511 48 6-phosphogluconate dehydrogenase 1.1.1.44 PF14\_0520 26 Ribulose 5-phosphate 3-epimerase 5.1.3.1 PFL0960w 29 Ribose 5-phosphate isomerase 5.3.1.6 PFE0730c 21 Transketolase 2.2.1.1 MAL6P1.110 17 Ribose phosphate diphosphokinase 2.7.6.1 PF13 0143 PF13 0157 21 21 Deoxyribose phosphate aldolase 4.1.2.4 PF10 0210 21 Phosphoglucomutase 5.4.2.2 PF10 0122 21 Pyridine nucleotide metabolism Nicotinamidase 3.5.1.19 chr3.glm 243 18, 38 Nicotinatephosphoribosyl transferase 2.4.2.11 chr6.glm\_337 25 Nicotinate-nucleotide adenylyl transferase 2.7.7.18 chr13.glm\_329 20, 40 NAD^+^synthetase 6.3.5.1 PFI1310w 28 Pyridine nucleotide transhydrogenase 1.6.1.1. PF14\_0508 32 Thiamine metabolism Hydroxyethylthiazole kinase 2.7.1.50 PFL1920c 22 Hydroxymethylpyrimidine kinase/ Phosphomethylpyrimidine kinase 2.7.149 2.7.4.7 PFE1030c 33 Thiamine-phosphate diphosphorylase 2.5.1.3 MAL6P1.285 1 Thiamine diphosphokinase 2.7.6.2 PFI1195c 28 ::: Results and Discussion ====================== At the onset of the present analysis, it should be underscored that time-dependent transcription does not always overlaps translation. Hence, transcript levels cannot be directly extrapolated to levels of their translated product. Moreover, transient transcription does not divulge on the stability of the translated products. Indeed, a recent analysis has shown for several genes that while the transcript peaks at the trophozoite stage and declines thereafter, the translated protein continues to accumulate \[[@B23]\]. Significant discrepancies between mRNA and protein abundance in P. falciparumwere also reported by LeRoche et al. \[[@B24]\]. This was shown to be due to a delay between the maximum detection of an mRNA transcript and that of its cognate protein. Surely, a protein cannot be produced if its cognate gene is not transcribed. Therefore, the present analysis can at best set a time for possible translation but not for its actual occurrence. Thus, while transcript levels are informative for the concerted action of metabolically related enzymes, post-transcriptional mechanisms for controlling protein levels and protein stability must also be considered. In this analysis, the transcription of genes coding for enzymes that constitute the PPP will be discussed first followed by those that are involved in the synthesis of NAD(P)^+^and of thiamine pyrophosphate. The pentose phosphate pathway ----------------------------- The pathway is shown in Figure [1](#F1){ref-type="fig"} and the time-dependent transcription of the different genes is depicted in Figure [3](#F3){ref-type="fig"}. Whereas G6PD (EC: 1.1.1.49) and hence its conjoint 6-phosphogluconolactonase (EC: 3.1.1.31) are transcribed at high levels during the early stages of parasite development, the transcription of 6-phosphogluconate dehydrogenase (EC: 1.1.1.44) seems to be biphasic (first peak at 9 hours post invasion (HPI) while the major peak is observed at 26 HPI). The two other enzymes that need to be present for full synthesis of PRPP are ribose phosphate isomerase (EC: 5.3.1.6) and ribose-phosphate diphosphokinase (EC: 2.7.6.1). The first has a small peak at 12 HPI and a major one at 21 HPI. The parasite contains 2 genes that code for ribose-phosphate diphosphokinase (PF13\_0143 and PF13\_0157). Both peak at 21 HPI. Thus, the oxidative arm seems to be fully activated to meet the requirement of ribonucleotide synthesis according to mode 1 of Figure [2](#F2){ref-type="fig"}. The genes that code for this process are coordinately transcribed starting immediately after invasion and peak at 12 HPI \[[@B2]\]. Biochemical evidence indicates that ribonucleotide synthesis starts at the mature ring stage \[[@B25]\]. Functioning according to mode 1 (Figure [2](#F2){ref-type="fig"}) would also supply reducing power to neutralize the toxic reactive oxidative species that are maximally produced during the hemoglobin-rich cytosol of the host cell \[[@B9]\]. It seems that the synthesis of PRPP and its utilization in the synthesis purine and pyrimidine nucleotides serves as a metabolic sink that could accelerate the pace of the oxidative arm. Contrary to this judicious coordination, the transcription of genes coding for enzymes needed for the synthesis of NADP^+^that is essential for the operation of the oxidative arm, lags considerably behind (see below). This is a crucial paradox that waits to be resolved. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Time dependent transcription of genes coding for enzymes involved in the pentose phosphate pathway. Glucose-6-phosphate dehydrogenase/6-phosphogluconolactoase -- black filled circles; 6-phosphogluconate dehydrogenase -- red filled circles; Ribose phosphate isomerase -- blue filled circles; Ribulose 5-phosphate epimerase -- green filled circles. Transketolase -- purple filled circles; Ribose phosphate diphosphokinase (PF13\_0157) -- gold filled circles and (PF13\_0143) -- dark green filled circles. Glucose-6-phosphate dehydrogenase and 6-phosphogluconolactoase were grouped together because they constitute a chimerical protein. ::: ![](1475-2875-4-17-3) ::: The situation of the non-oxidative arm genes is different. The gene that codes for ribose phosphate isomerase (EC: 5.1.3.1) peaks at 21 HPI and this seems to restrict the expression of the non-oxidative arm, since transketolase (EC: 2.2.1.1) is transcribed maximally immediately following invasion. Most importantly, the gene that codes for transaldolase (EC: 2.2.1.2) could not be found in the genome of P. falciparum(or in any other Plasmodiumspecies sequenced so far, or in any other Apicomplexans for that matter) thus precluding the completion of the analysis of the non-oxidative arm transcription. As mentioned above, biochemical evidence indicates that this arm is active in the parasite. If it is indeed fully activated when ribose phosphate isomerase is transcribed (and supported by the transcription of the genes that code for ribose-phosphate diphosphokinase that peak at 21 HPI), it would match the transcription of the deoxyribonucleotide synthesis-related genes that starts at 18 HPI and peaks at 30 HPI \[[@B2]\] as well as those related to NADP^+^synthesis. When the synthesis of ribonucleotides and deoxyribonucleotides ebbs off towards the end of the intraerythrocytic life cycle of the parasite, the PPP probably functions according to mode 3 (Figure [2](#F2){ref-type="fig"}) providing both NADPH and ATP. A feature that emerges from this analysis is that the transcription of the genes that code for enzymes acting in the non-oxidative arm are coordinated as shown for other clusters of genes whose products are functionally related \[[@B1]-[@B3]\]. However, it remains to be seen if the cognate enzymes of these transcripts are similarly coordinated since significant discrepancies between mRNA and protein abundance has been recently reported \[[@B24]\]. NAD(P) biosynthesis ------------------- The pathway is shown in Figure [4](#F4){ref-type="fig"} and the time-dependent transcription of the different genes is depicted in Figure [5](#F5){ref-type="fig"}. Nicotine adenine dinucleotide (NAD^+^) and NAD phosphate (NADP^+^) play key roles in glycolysis and PPP as well as many other enzymatic reactions (see Tables [3](#T3){ref-type="table"} and [1](#T1){ref-type="table"}, respectively). NAD synthesis starts by converting nicotinic acid to nicotinate D-ribonucleotide by nicotinate phosphoribosyltransferase (NAPRT; EC: 2.4.2.11). Nicotinate D-ribonucleotide is then adenylated to deamino-NAD+ in a reaction catalyzed by nicotinate-nucleotide adenylyltransferase (EC: 2.7.7.18) then this intermediate is amidated to NAD^+^by NAD synthetase (EC: 6.3.5.1). Nicotinamide can be used for NAD synthesis either by being deamidated to nicotininc acid by nicotinamidase (EC: 3.5.1.19) or by being converted to nicotinamide mononucleotide by nicotinamide phosphoribosyltransferase (NPRT; EC: 2.4.2.12) and then to NAD by NAD pyrophosphorylase (EC: 2.7.7.1). Genes coding for the two latter enzymes could not be found in the genome of P. falciparum, although a measurable activity of NPRT has been detected \[[@B26]\]. These enzymes are present in normal erythrocytes and at physiological conditions the production of NAD^+^from nicotinamide seems to be more important than that of NADP^+^\[[@B27]\]. Zerez et al. \[[@B26]\] demonstrated 15-fold increase in the levels of NAD^+^(NAD^+^+ NADH) in P. falciparum-infected erythrocytes, as well as 3-fold increase in NAPRT activity. These results indicate that the parasite is capable of NAD synthesis although neither the activity of nicotinate-nucleotide adenylyltransferase or of NAD synthetase were increased in infected cells compared to uninfected erythrocytes. This observation suggests that these enzymes must be active in the parasite since the digestion of host cell cytosol during parasite development should have reduced their activities. Abundant nicotinamidase activity was also detected in infected cells implying that the parasite can synthesize NAD^+^from both nicotinic acid and nicotinamide (both present in RPMI-1640 medium used for parasite cultivation). The concentrations of NADP^+^in infected cells are 10-fold lower that those of NAD^+^. A gene coding for NAD^+^kinase could not be found in the genome of P. falciparumor in any other Plasmodiumspecies or in other Apicomplexans, but could be detected in Giardia lamblia. No information about the action of NAD kinase in the parasite could be found. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Nicotine nucleotide metabolism. The no-entry symbol indicates that the gene coding for this enzyme could not be found in the genome of P. falciparum, but biochemical evidence suggests that the enzyme activity must be present. ::: ![](1475-2875-4-17-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Time dependent transcription of genes coding for enzymes involved in the biosynthesis of NAD(P)^+^. Nicotinamidase -- purple filled circles; Nicotinate phosphoribosyl transferase -- green filled circles; Nicotinate-nucleotide adenylyltransferase -- red filled circles; NAD^+^synthetase -- black filled circles; NAD(P) transhydrogenase -- blue filled circles. ::: ![](1475-2875-4-17-5) ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### NAD^+^-utilizing enzymes. Enzymes are arranged by their sequential functional order. They are given by their name, their EC numbers, the gene identification (PfID) in the Plasmodiumgenome database (PlasmoDB), the time (in hours post invasion (HPI)) in the parasite\'s developmental cycle when they are maximally transcribed obtained from the IDC database, the metabolic function of the enzyme. ::: Enzyme EC number PfID Peak (HPI) Metabolic function --------------------------------------------------- ----------- -------------------- ------------ -------------------------------------------- Pyrroline carboxylate reductase 1.5.1.2 MAL13P1.284 18 Methionine polyamine metabolism Ferrodoxin reductase-like protein 1.7.1.4 PF07\_0085 21 Nitrogen metabolism Glutamate dehydrogenase 1.4.1.2 PF08\_0132 20 Glutamine metabolism Aminomethyltransferase 2.1.2.10 PF13 0345 22 Folate biosynthesis L-lactate dehydrogenase 1.1.1.27 PF13 0141 22 Glycolysis Enoyl-acyl carrier reductase 1.3.1.9 MAL6P 1.275 30 Fatty acid synthesis 2-oxoglutarate dehydrogenase el component 1.2.4.2 PF08\_0045 30 TCA cycle Lipoamide dehydrogenase, putative 1.8.1.4 PF08\_0066 28 TCA cycle Pyruvate dehydrogenase El component, ?-subunit 1.2.4.1 PF11\_0256 28 Fatty acid synthesis NADH-cytochrome b5 reductase 1.6.2.2 PF13\_0353 32 Electron transport Malate dehydrogenase, putative 1.1.1.37 MAL6P 1.242 34 Pyruvate metabolism Glycerol-3-phosphate dehydrogenase 1.1.1.8 PF11 0157 PFL0780w 37 19 Glycolysis; Glycerol metabolism GDP-mannose 4,6-dehydratase 1.1.1.187 PF08\_0077 39 Mannose and fructose metabolism Inosine-5 \'-monophosphate dehydrogenase 1.1.1.205 PFI1020c 12 Purine metabolism 3-methyl-2-oxobutanoate dehydrogenase (lipoamide) 1.2.4.4 PFE0225w 32 Leucine, isoleucine and valine degradation Nitrate reductase 1.7.1.1 chr13.glm\_739 32 Nitrogen metabolism ::: Inspection of the time-dependent transcription of genes coding enzymes that are involved in NAD(P)^+^biosynthesis reveal an unusual pattern. Transcriptional profiles of at least three transcripts show two peaks, namely those of nicotinamidase, of nicotinate-nucleotide adenylyltransferase and of NAD(P)^+^transhydrogenase, as compared to most other transcripts in the parasite transcriptome that are monophasic. The reason for this pattern is unclear as biochemical data are uanavilable for any physiological intepretation, but it may suggest that the the enzymes are not stable. Assuming that the first smaller peak of nicotinate-nucleotide adenylyltransferase (20 HPI) is sufficient for adequate expression of enzymatic activity, the biosynthesis of NAD(P)^+^should peak when NAD^+^synthetase is at its peak, i.e., at 28 HPI. Thus, by 28 HPI the transcripts of all enzymes necessary for NAD^+^synthesis are fully deployed and somewhat later, synthesis itself is probably fully deployed. This time pattern means that important activities, such as glycolysis (the transcription of genes coding for glycolytic enzymes starts to peak at 9 HPI and starts to decrease at 24 HPI \[[@B2]\] At earlier stages glycolysis and PPP activity probably depend on the pool of NAD(P)^+^that were present in the invading merozoite. For the full activation of glycolysis at the trophozoite stage \[[@B28]\], the pool of NAD(P)^+^has to be significantly amplified. As can be seen in Table [3](#T3){ref-type="table"}, several genes coding for enzymes requiring NAD(H) are transcribed earlier than the complete transcription of genes related to NAD^+^synthesis. Such an outstanding exception is that of inosine-5\'-monophosphate dehydrogenase, an essential enzyme in the purine metabolic pathway: it is transcribed too early to allow its translated product to be functionally useful. The functional meaning of the second peaks of nicotinamidase and of nicotinate-nucleotide adenylyltransferase is enigmatic since the transcription of all other enzyme-coding genes declines to a minimum when they peak. For being meaningful physiologically, the stability of all other enzymes must be sustained with time. Indeed, it has been recently shown that the levels of some proteins such as methionine adenosyltransferase, ornithine aminotransferase, lactate dehydrogenase, glyceraldehyde 3-phosphate dehydrogenase and enolase, are not only increased following transcription, but further increase after their transcript levels decline \[[@B23]\]. Finally, NAD^+^kinase is essential for the production of NADP^+^. Its absence from the genome is indeed very perplexing because biochemical evidence indicate that the parasite is able to synthesize NADP^+^\[[@B19]\] and the presence of many enzymes that need it as a cofactor, implicate that its synthesis is essential for parasite growth. As long as the gene (or some surrogate mecahnism) is not identified, nothing can be said about the coordination of NADP^+^biosynthesis and the expression of enzymes that utilize it (Table [1](#T1){ref-type="table"}). Transhydrogenase operates at an important interface between NAD(H) and NADP(H) and between the mitochondrial proton electrochemical gradient ??; \[[@B29]\]. Under regular physiological conditions, the enzyme is a consumer of ??: NADH + NADP^+^+ H^+^~out\_~NAD^+^+ NADPH + H^+^~in~. The energy of the gradient can drive the \[NADPH\]\[NAD^+^\]/\[NADP^+^\]\[NADH\] ratio to values \>400. Transhydrogenation can also function in the reverse direction from NADPH to NAD^+^. This is accompanied by outward proton translocation and formation of ??. In this mode, the enzyme utilizes substrate binding energy for proton pumping. Therefore, in terms of energy transduction, the transhydrogenase works in principle like the ATP synthase complex of mitochondria, the proton ATPase. Given the fact that the parasite genome does not have the full complement of genes coding for the mitochondrial proton ATPase \[[@B30]\], it is tempting to suggest that the transhydrogenase could fulfill such role. The transcription of the gene coding for transhydrogenase shows two very distinct peaks at 20 and 30 HPI, with no detectable transcription between them. It may well be that the second peak is adjusted to the time of elongation and division of the single mitochondrion \[[@B31]\]. Thiamine biosynthesis --------------------- The essential product of this pathway, thiamine diphosphate, is a cofactor of many enzymes (Table [4](#T4){ref-type="table"}). The pathway is shown in Figure [6](#F6){ref-type="fig"} and the time-dependent transcription of the different genes is depicted in Figure [7](#F7){ref-type="fig"}. Thiamine (vitamin B~1~) is the precursor of the coenzyme thiamine pyrophosphate that is involved in the action of many decarboxylating enzymes and relevant to the present analysis, in the function of transketolase. P. berghei-infected erythrocytes contain higher levels of thiamine than their uninfected counterparts \[[@B32]\] and thiamine deficiency retards the propagation of this parasite in vivo\[[@B33]\]. The incidence of thiamine deficiency in adults admitted to hospital with malaria in Thailand has been examined \[[@B34]\] and it was observed that in hospitalized thiamine deficiency commonly complicates acute falciparum malaria, particularly in severe infections, and could contribute to dysfunction of the central nervous system. Thiamine could be obtained by the parasite from the extracellular space (like in animal cells) and pyrophosphorylated by thiamine diphosphokinase (EC: 2.7.6.2). However the presence of genes coding for enzymes that participate in the eukaryote biosynthetic pathway of thiamine phosphate, suggest that the parasite is able to synthesize its own precursor. Nevertheless, the absence of a gene coding for thiamine phosphate kinase (EC: 2.7.4.16) must be resolved and direct biochemical demonstration of activity must be performed before a firm statement can be made on this metabolic pathway. It is not unlikely that thiamine phosphate kinase is not needed, as is the situation in yeast \[[@B35]\]. There it is suggested that thiamine phosphate is hydrolyzed to thiamine by a non-specific phosphatase and thiamine is then converted to thiamine pyrophosphate by thiamine diphosphokinase. ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Thiamine diphosphate utilizing enzymes. Enzymes are arranged by their sequential functional order. They are given by their name, their EC numbers, the gene identification (PfID) in the Plasmodiumgenome database (PlasmoDB), the time (in hours post invasion (HPI)) in the parasite\'s developmental cycle when they are maximally transcribed obtained from the IDC database, the metabolic function of the enzyme. ::: -------------------------------------------------------------------------------------------------------------------------- Enzyme EC number PfID HPI Metabolic function ----------------------------------------- ----------- ----------------- ----- -------------------------------------------- Pyruvate dehydrogenase E1 1.2.4.1 PF11\_0256 (a)\ 28\ Apicoplast fatty acid synthesis PF14\_0441 (b) 28 Transketolase 2.2.1.1 MAL6P1.110 17 Pentose phosphate pathway Oxalyl-CoA decarboxylase 4.1.18 MAL6P1.231\ 26\ Glyoxylate and dicarboxylate metabolism PFF0945c NA 1-deoxy-D-xylulose-5-phosphate synthase 2.2.1.7 PF13\_0207 21 Isoprenoid biosynthesis 3-methyl-2-oxobutanoate dehydrogenase 1.2.4.4 PFE0225w 32 Leucine, isoleucine and valine degradation Oxoglutarate dehydrogenase 1.2.4.2 PF08\_0045 32 Mitochondrial TCA cycle -------------------------------------------------------------------------------------------------------------------------- ::: ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Biosynthesis of thiamine diphosphate. The no entry symbol indicates that the gene coding for this enzyme could not be found in the genome of P. falciparum. No biochemical evidence exists to confirm or refute enzyme activity. It has been added to the scheme because genes coding for enzymes preceding it in the pathway are present in the genome. ::: ![](1475-2875-4-17-6) ::: ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Time dependent transcription of genes coding for enzymes involved in the biosynthesis of thiamine diphosphate. Hydroxyethylthiazole kinase -- red filled circles; Hydroxymethylpyrimidine kinase/phosphomethylpyrimidine kinase -- black filled circles; Thiamine-phosphate diphosphorylase -- blue filled circles; Thiamine diphosphokinase -- green filled circles. Hydroxymethylpyrimidine kinase and phosphomethylpyrimidine kinase were grouped together because they probably constitute a chimerical protein. ::: ![](1475-2875-4-17-7) ::: The transcription of all genes coding for enzymes involved in thiamine pyrophosphate seems to be coordinated, single phased and peaking between 20 and 30 HPI. If thiamine is obtained from the host and thiamine pyrophosphate is synthesized by the single step mediated by thiamine diphosphokinase, the peak transcription of this gene at 28 HPI lags by several hours after that of transketolase (17 HPI), but since transketoalse is not the expression time-setter of PPP, this lag does not seem to limit the full activity of PPP. However, if most of the thiamine is generated endogenously by the parasite, the supply of this precursor will peak only at 33 HPI, thus limiting full PPP activity. It is not unlikely that both processes occur in tandem or that at advanced stage of parasite development, the need for thiamine cannot be met anymore by exogenous supply and the endogenous synthesis joins the game. Conclusion ========== The analysis of time-dependence transcription of parasite genes concluded that the parasite has evolved a highly specialized mode of transcriptional regulation that produces a continuous cascade of gene expression, beginning with genes corresponding to general cellular processes, such as protein synthesis, and ending with Plasmodium-specific functionalities, such as genes involved in erythrocyte invasion \[[@B2]\]. However, a meticulous analysis shows marked and important deviations from this prototype that reveal a lack of coordinated transcription of genes coding for enzymes of the same metabolic pathway and between pathways. There are three most straightforward explanations for these apparent discrepancies. First, there are additional enzymes facilitating the \"missing\" activities and their identity was not revealed by the present annotations due to their diverse amino acid sequence. Second, the misaligned transcriptional regulation reflect an intricate interplay of the biosynthetic pathways where delayed production of metabolites in one pathway functions as a rate limiting factor for other pathway, which is otherwise fully deployed. Last but not least, post-transcriptional regulation may also play a role. All theories create an intriguing possibility for further studies. Clearly this effort will be enhanced by substantial progress in proteomics and most importantly, direct biochemical demonstrations of activities of individual enzymes and entire pathways. Authors\' contributions ======================= Each author contributed equally to this investigation. Acknowledgements ================ We thank Professor I. Ohad for critical reading of the manuscript.	PubMed Central	2024-06-05T03:55:55.872434	2005-3-18	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084361/", "journal": "Malar J. 2005 Mar 18; 4:17", "authors": [ { "first": "Zbynek", "last": "Bozdech" }, { "first": "Hagai", "last": "Ginsburg" } ] }
PMC1084362	Background ========== Although the major steps in reverse transcription have been known for some time, the process by which the tRNA primer is specifically selected from the intracellular milieu by the virus is less well understood. Even though different retroviruses select different tRNA primers for reverse transcription, within a group of retroviruses, the tRNA primer selected is conserved \[[@B1],[@B2]\]. For example, murine leukemia virus (MuLV) selects tRNA^Pro^, while avian leukosis virus (ALV) selects tRNA^Trp^\[[@B3],[@B4]\]. Human immunodeficiency virus type 1 (HIV-1), as do all lentiviruses, selects tRNA^Lys,3^for use as the primer for reverse transcription \[[@B5],[@B6]\]. How and why HIV-1 exclusively selects tRNA^Lys,3^as the primer for reverse transcription is not known. Pseudovirions composed of Gagand Gag-polare enriched for tRNA^Lys^, including tRNA^Lys,3^, that is used for initiation of reverse transcription \[[@B2],[@B7],[@B8]\]. Additional studies have shown that the specific incorporation of lysyl tRNA synthetase and its interaction with Gagcould also be important for the specific capture of tRNA^Lys,3^used for initiation of reverse transcription \[[@B9]-[@B11]\]. Substitution of the primer-binding site (PBS) to be complementary to alternative tRNAs results in the capacity of HIV-1 to transiently use these tRNAs for initiation of reverse transcription \[[@B12]-[@B14]\], even though upon extended culture, these viruses all reverted back to specifically utilize tRNA^Lys,3^for initiation of reverse transcription. In some instances, mutation of a region 5\' of the PBS so as to be complementary to the anticodon of certain tRNAs, in conjunction with mutations of the PBS, results in the virus stably utilizing these alternative tRNAs for replication \[[@B15]-[@B19]\]. Interestingly, analysis of the virion tRNAs of a virus that stably utilized tRNA^His^for replication did not show a difference in composition compared to that of the wild type virus that utilized tRNA^Lys,3^, indicating that tRNAs in the HIV-1 virion did not necessarily reflect the tRNA selected for initiation of reverse transcription \[[@B20]\]. The fact that HIV-1 can select different tRNAs for replication suggests the virus has access to a variety of tRNAs. Recent studies from this laboratory demonstrated that the tRNA selected by HIV-1 for replication have undergone nuclear-to-cytoplasmic transport. Furthermore, the tRNAs that are aminoacylated, indicating inclusion in translation, are most efficiently selected as primers \[[@B21]\]. The realization that tRNA biogenesis and translation might intersect with primer selection has prompted us to re-examine the stability and replication of HIV-1 with a PBS complementary to alternative tRNAs in a relevant cell type peripheral blood mononuclear cells (PBMC). In a previous study, we found that HIV-1 in which the PBS was altered to be complementary to tRNA^Lys1,2^or tRNA^His^reverted to utilize tRNA^Lys,3^upon extended culture in PBMC \[[@B22]\]. Viruses could be generated which stably utilized these tRNAs for replication if additional mutations within the U5, consisting of nucleotides complementary to the anticodon regions, were also included in the viral genomes. Interestingly, viruses which utilize tRNA^Lys1,2^had further adapted to utilize this tRNA, exhibiting replication characteristics similar to the wild type virus following extended in vitroreplication in human PBMC. Similar results have been recently reported for HIV-1 in which the PBS and a second region upstream, the primer activation site (PAS), has been altered to be complementary to tRNA^Lys1,2^\[[@B23]\]. In this case, the virus stably utilized tRNA^Lys1,2^for an extended culture period. A mutation in the RNase H domain of the reverse transcriptase was also found, although the major determinant of the stability of the PBS was correlated with the mutations in the PAS and PBS. In the current study, we have further examined the preference of HIV-1 for certain tRNAs. A previous study from this laboratory has shown that viruses with a PBS complementary to tRNA^Pro^or tRNA^Ile^were unstable following replication in SupT1 cells, an immortalized, continuous, human T cell line \[[@B24]\]. However, during the process of reversion to a PBS complementary to tRNA^Lys,3^, we noted several different anomalies with the PBS, including the isolation of viruses with multiple PBS complementary to other tRNAs and a virus in which the PBS was complementary to tRNA^Met^. Further characterization of this virus revealed that it stably utilized tRNA^Met^as the initiation primer following additional mutations in which the U5 was made complementary to the anticodon region of tRNA^Met^\[[@B15],[@B19]\]. We have now analyzed the replication and stability of the PBS of viruses in which the PBS alone was altered to be complementary to tRNA^Met^, tRNA^Pro^or tRNA^Ile^, as well as viruses with both the PBS and U5 region altered to be complementary to the 3\' 18-nucleotides and anticodon of these tRNAs. Clear differences were identified with respect to primer preference that correlated with virus replication. The results of these studies therefore establish that preferences for selection of certain tRNAs to be used in reverse transcription by HIV-1 do exist and are more evident following replication in PBMC than in continuous T cell lines. Results ======= Construction and characterization of HIV-1 proviral genomes complementary to tRNA^Met^, tRNA^Pro^and tRNA^Ile^ -------------------------------------------------------------------------------------------------------------- In previous studies, we have described the construction of HIV-1 proviral genomes in which the PBS was made complementary to alternative tRNAs \[[@B15]-[@B19]\]. For these studies, the proviral genomes were based on HXB2, which allows high-level replication in continuous T cell lines (eg. SupT1s). For the current studies, we have transferred the 5\' LTR up to the BssHII site (nucleotide 233) from these clones into the NL4-3 proviral clone of HIV. The NL4-3 proviral clone of HIV-1, in contrast to the HXB clone, contains open reading frames for all the accessory proteins and replicates to high levels in PBMC. The U5-PBS regions of the subsequent proviral constructs, named pNL4-3-Met, pNL4-3-Pro and pNL4-3-Ile were sequenced prior to analysis to confirm that the constructs were isogenic with the wild type with the exception of the 3\' 18-nucleotide PBS region (Figure [1A](#F1){ref-type="fig"}). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### U5-PBS sequence and infectivity levels of HIV-1 NL4-3 viral mutants at start of PBMC infection. Panel A. HIV-1 U5 and PBS sequence shown (from 5\' to 3\'). Viral primer binding site (PBS) sequence was altered to be complementary to the 3\' terminal 18 nucleotides of tRNA^Ile^, tRNA^Met^and tRNA^Pro^. The PBS sequence is shadowed. Panel B. Comparison of infectivity of NL4-3 PBS mutants. HIV-1 NL4-3 proviral clones were transfected into 293T cells, incubated for 48 hours, and supernatants were measured for infectious units. For a given sample, the number of infectious units per microliter is equal to the number of blue cells in a well divided by the dilution factor for that well and represents the average of at least two wells. Wild type infectivity levels were set at 100% and mutant virus infectivity was reported as a percentage of wild type. All viruses with altered PBS sequences showed reduced levels of infectivity as compared to wild type. Results presented are representative of three experiments. ::: ![](1742-4690-2-21-1) ::: To characterize these viruses, we first measured the production of infectious virus and p24 antigen following transfection into 293T cells. Since 293T cells do not support HIV-1 replication, this analysis would provide us with the inherent infectivities of viruses prior to undergoing reverse transcription/replication in PBMC. In our previous studies, we noted that there was no substantial difference in the production of virus (as measured by p24 antigen) as a result of altering the PBS in the HXB2 proviral constructs \[[@B15]-[@B19]\]. For the current studies, we transfected the proviral clones into 293T cells and determined the amount of infectious units using the JC53BL assay; virus production was then measured using a p24 antigen capture ELISA. Infectivity was determined as the ratio of infectious units to p24 antigen. The values are presented relative to the infectivity of the wild type virus, with a PBS complementary to tRNA^Lys,3^(Figure [1B](#F1){ref-type="fig"}). All of the viruses with altered PBS had infectivities lower than the wild type. The virus NL4-3-Ile, with a PBS complementary to tRNA^Ile^, was consistently the most infectious of the mutants with a level approximately 40% that of wild type, while the other viruses were 10--20% as infectious as the wild type virus. Stability of PBS following replication in PBMC ---------------------------------------------- We next wanted to determine the effects of alteration of the PBS on the replication of these viruses in PBMC. Infections were initiated with 200 pg of p24 and were allowed to proceed with re-feeding of PBMC every 14 days for periods of time exceeding 50 days of in vitroculture. The cultures were sampled periodically, supernatants were assayed for p24 antigen and cells were processed to extract high molecular weight DNA to determine the stability of the PBS. All of the viruses with an altered PBS showed an initial delay in production of p24 antigen compared to the wild type virus, consistent with the initial reduced infectivity compared to wild type (Figure [2](#F2){ref-type="fig"}). The NL4-3-Met virus had replication kinetics most similar to wild type virus in that during the first 10 days of culture we observed a rapid rise in p24 antigen, followed by a plateau at a level similar to that for wild type. The NL4-3-Ile virus replicated more slowly, with a gradual rise in p24 antigen before finally reaching a level similar to wild type. Finally, the NL4-3-Pro virus showed minimal replication in the first 21 days of culture, followed by a rapid increase in the next 14 days to reach levels similar to that of the wild type virus (Figure [2](#F2){ref-type="fig"}). Although all of the viruses replicated in PBMC, the kinetics did not correlate with the initial infectivities from the 293T transfection supernatants. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Replication of HIV-1 with PBS sequence altered to be complementary to the 3\' 18 nucleotides of tRNA^Ile^, tRNA^Met^and tRNA^Pro^in human peripheral blood mononuclear cells (PBMC). Infections were initiated with transfection supernatant containing approximately 200 pg of p24 antigen in a volume of 10 mLs of media, giving a final p24 level of 20 pg/mL on day zero. At 14 day intervals, 5 × 10^6^fresh PHA stimulated PBMC were added to each culture. Supernatants were assayed for p24 viral antigen using an ELISA. Two additional separate infections produced similar replication patterns for each virus. Squares are wild type NL4-3; diamonds are NL4-3-Met; open circles are NL4-3-Ile and closed circles are NL4-3-Pro. ::: ![](1742-4690-2-21-2) ::: During the culture period, we collected the DNA from the cells to determine the status of the integrated virus PBS. The wild type virus, as expected, maintained a PBS complementary to tRNA^Lys,3^throughout the entire culture period (data not shown). In contrast, viruses with a PBS complementary to tRNA^Ile^initially used tRNA^Ile^for reverse transcription, but by day 21 from the initiation of culture had reverted to be complementary to tRNA^Lys,3^(Table [1](#T1){ref-type="table"}). Viruses in which the PBS was altered to be complementary to tRNA^Pro^(NL4-3-Pro) appeared to be slightly more stable and maintained the PBS complementary to this tRNA through day 21 of culture, before reverting to wild type by day 35. However, the subsequent rapid replication of the virus corresponded with the presence of a PBS complementary to tRNA^Lys,3^(Table [1](#T1){ref-type="table"}). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Stability of PBS following extended culture in PBMC ::: Virus PBS Sequence Time to Reversion^1^(days) ------------------------------- ------------------ -------------------------------- [NL4-3-lle]{.underline} Lys,3^2^ 21 [NL4-3-Met]{.underline} Met^3^ \-\--^4^ [\>NL4-3-Pro]{.underline} Lys,3 35 1\. PBS analyzed at the time of in vitroculture in PBMC and found to be wild type, complementary to tRNA^Lys,3^. 2\. PBS complementary to tRNA^Lys,3^. 3\. PBS complementary to tRNA^Met^. 4\. Analysis of 34 TA clones of the PBS following 63 days in culture revealed all maintained a PBS complementary to tRNA^Met^. ::: Surprisingly, the viruses in which the PBS was complementary to tRNA^Met^remained stable for use of tRNA^Met^throughout the complete culture period (in this case, up to 63 days post initiation of culture). Analysis of 34 individual TA clones of the PBS from these viruses revealed that all were complementary to tRNA^Met^(Table [1](#T1){ref-type="table"}). This is the first instance in which we have found a virus that stably maintains a PBS complementary to an alternative tRNA (not tRNA^Lys,3^) following extensive in vitroreplication that did not have additional mutations in U5. Previously, analysis of this virus in the HXB2 proviral clone revealed that the PBS was unstable following replication of the virus in vitroin SupT1 cells and reverted back to use tRNA^Lys,3^\[[@B15],[@B18]\]. We further characterized the replication of this virus compared to wild type virus at different times during the culture period. Analysis of p24 antigen production from this virus at day 56 post initiation of culture revealed that it replicated similar to the wild type virus, albeit with slightly lower levels of p24 antigen (data not shown). The infectivity of the virus obtained after long-term culture, which utilized tRNA^Met^, was approximately 50--80% of the wild type virus (data not shown). Collectively, the results of these studies establish that HIV-1 has a preference for certain tRNAs, such as tRNA^Met^, for the selection as primer for reverse transcription. Effect of mutations in U5 on replication of viruses that use alternative tRNAs ------------------------------------------------------------------------------ In previous studies, we have found that with a PBS complementary to tRNA^Lys1,2^, tRNA^Met^, tRNA^Glu^and tRNA^His^, the additional mutation in which the U5 region was made complementary to the anticodon stabilized the HXB2 proviral clones to allow continuous use of the alternative tRNA during replication \[[@B16],[@B19],[@B24],[@B25]\]. In contrast, in the HXB2 provirus, modification of the U5 region for viruses in which the PBS was made complementary to tRNA^Pro^or tRNA^Ile^did not result in virus that could stably utilize these tRNAs following replication \[[@B24]\]. To determine if this would be case for viruses that were grown in PBMC, we constructed HIV-1 in which both the U5 and PBS were made complementary to tRNA^Met^, tRNA^Pro^or tRNA^Ile^(Figure [3A](#F3){ref-type="fig"}). The initial infectivities of the viruses were analyzed following transfection of the proviral clones into 293T cells. Similar to what we observed for viruses with just the PBS altered to be complementary to these tRNAs, the viruses with both the U5 and PBS altered demonstrated infectivities lower than wild type virus. In this case, the levels ranged from a low of 5% (NL4-3-Pro-AC) to a high of 30% (NL4-3-Met-AC) (Figure [3B](#F3){ref-type="fig"}). We initiated infections in PBMC with the same amount of p24 antigen. We noted a delay in the production of p24 antigen in the cultures of viruses in which both the PBS and A loop were mutated to be complementary to the alternative tRNA^Met^or tRNA^Ile^, relative to the wild type virus (Figure [4](#F4){ref-type="fig"}). By day 21, the viruses derived from pNL4-3-Met-AC had p24 antigen levels in the culture supernatants similar to that for the wild type virus. Viruses derived from pNL4-3-Ile-AC replicated at levels approximately 1/10 that of the wild type virus, while viruses derived from pNL4-3-Pro-AC did not replicate well (or at all), as evidenced by p24 levels that did not increase substantially over the culture period (Figure [4](#F4){ref-type="fig"}). ::: {#F3 .fig} Figure 3 ::: {.caption} ###### U5-PBS sequence and Infectivity levels of HIV-1 NL4-3 viral mutants with altered U5 and PBS sequences at start of PBMC infection. Panel A. HIV-1 U5 and PBS sequence (shown 5\' to 3\'). Viral primer binding site (PBS) and U5 A-loop sequences were altered to be complementary to the 3\' terminal 18 nucleotides and anticodon loop of tRNA^Ile^, tRNA^Met^, tRNA^Pro^, and tRNA^Trp^, respectively. U5 A-loop sequence and PBS are shadowed. Panel B. Comparison of the infectivity of U5-PBS mutant NL4-3 viruses. HIV-1 NL4-3 proviral clones were transfected into 293T cells, incubated for 48 hours, and supernatants were measured for infectious units. For a given sample, the number of infectious units per microliter is equal to the number of blue cells in a well divided by the dilution factor for that well and represents the average of at least two wells. Wild type infectivity levels were set at 100%, and mutant virus infectivity was reported as a percentage of wild type. All viruses with altered U5 and PBS sequences had reduced levels of infectivity as compared to wild type virus. The data presented are representative for three independent experiments. ::: ![](1742-4690-2-21-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Replication of HIV-1 with U5 and PBS sequence altered to be complementary to the anticodon loop and 3\' 18 nucleotides of tRNA^Ile^, tRNA^Met^, tRNA^Pro^and tRNA^Trp^in PBMC. Infections were initiated with transfection supernatant containing approximately 200 pg of p24 antigen in a volume of 10 mLs of media, giving a final p24 level of 20 pg/mL on day zero. Supernatants were assayed for p24 viral antigen every 7 days for a period of 42 days. At 14-day intervals, 5 × 10^6^fresh PHA stimulated PBMC were added to each culture. Two additional separate infections produced very similar replication patterns for each virus (data not shown). Squares are NL4-3 wild type; diamonds are NL4-3-Met-AC; open circles are NL4-3-Ile-AC; closed circles are NL4-3-Pro-AC. ::: ![](1742-4690-2-21-4) ::: We next analyzed the PBS of the viruses. Consistent with our previous studies, we found that the viruses in which both the U5 and PBS were complementary to tRNA^Met^remained stable during the culture period (Table [2](#T2){ref-type="table"}). Sequence analysis of the virus that stably utilized tRNA^Met^(NL4-3-Met-AC) revealed a few nucleotide changes outside of the PBS. Previous studies from our laboratory have reported single nucleotide changes, noting that these changes might be important in stabilizing RNA structures to facilitate more effective primer selection \[[@B15],[@B17]-[@B20],[@B24]\]. Further experiments will be needed to address this issue. Characterization of NL4-3-Met-AC after extended culture revealed that it had infectivities that were still lower than that of the wild type virus (data not shown). In fact, the infectivities of the virus derived from pNL4-3-Met-AC were generally lower than those from the virus derived from pNL4-3-Met (data not shown). In contrast, viruses in which the PBS and U5 region were made complementary to tRNA^Ile^were not stable and reverted to utilize tRNA^Lys,3^following in vitroreplication (Table [2](#T2){ref-type="table"}). Thus, the A loop modification did not stabilize the virus to continuously use tRNA^Ile^for replication in PBMC. Virus with both the U5 and PBS altered to be complementary to tRNA^Pro^replicated poorly in the in vitroculture. Amplification of the region containing the PBS required use of a double PCR method in which the initial PCR product was used as the template of the second reaction (double PCR). Sequence analysis revealed that NL4-3-Pro-AC had maintained a PBS complementary to tRNA^Pro^(data not shown). Viruses in which the U5 and PBS were made complementary to tRNA^Pro^also reverted to use wild type following replication in SupT1 cells; in this case, we found viruses which contained multiple PBS, some of which were complementary to tRNA^Lys,3^. Following replication in PBMC, though, we did not isolate viruses with multiple PBS and all the viruses isolated contained a PBS complementary to tRNA^Pro^. Collectively, the results of these studies establish that HIV-1 does have a preference for tRNA^Met^over tRNA^Pro^with respect to the selection of the tRNA primer for replication. Furthermore, tRNA^Ile^is not favored for selection by HIV-1 even if compensatory mutations are provided in which the U5 region has been made complementary to the anticodon region. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Analysis of U5-PBS from viruses following extended in vitroculture in PBMC ::: Virus U5-PBS Days Post-Culture ---------------------------------- ----------------------------------------------------------------------------------------------------------------- ------------------ ----------------------- Ile^3^ [NL4-3-Ile-AC]{.underline} ^1^5\' AGTCAGTGTTTATCAGCTCTAGCAG ^2^TGGTGGCCCGTACGGGGATTGAAA 3\' Input^5^ 0 Ile ^4^5\' \\\\\\\\\\\\\\\\\\\\\\\\\* \\\\\\\\\\\\\\\\\\ \\\\\\ 3\' PCR Product^6^ 21 Lys, 3 5\' \\\\\\\\\\\\\\\\\\\\\\\\\* TGGCGCCCGAACAGGGAC-\\\\\* 3\' 6/7 TA Clones^7^ 35 Lys,1,2 5\' \\\\\\\\\\\\\\\\\\\\\\\\\* TGGCGCCCAACGTGGGGC-\\\\\* 3\' 1/7 TA Clones 35 Lys, 3 5\' \\\\\\\\\\\\\\\\\\\\\\\\\* TGGCGCCCGAACAGGGAC-\\\\\* 3\' PCR Product 73 Met [NL4-3-Met-AC 5]{.underline} 5\' AGTCAGTGTTGTGAGACTGTAGCAG TGGTGCCCCGTGTGAGGCGAAAGC 3\' Input 0 Met 5\' \\\\\\\\\\\\\\\\\\\\\\\\\* \\\\\\\\\\\\\\\\\\ \\\\\\ 3\' 5/10 TA Clones 35 5\' \\\\\\\\\\\\\\\\\\\\\\\\\* \\\\\\\\\\\\\\\\\*A\\\\\\ 3\' 3/10 TA Clones 35 5\' \\\\\\\\\\\\\\\\\\\\\* \A\ \* \\\\\\\\\\\\\\\\\\ \\\\\\ 3\' 1/10 TA Clones 35 5\' \\\\\\\\\\\\\\\\\\\\\\\\\* \\\\\\\*T\\\\\\\\\\ \\\\\\ 3\' 1/10 TA Clones 35 Met 5\' \\\\\\\\\\\\\\\\\\\\\\\\\* \\\\\\\\\\\\\\\\\\ A\\\\\* 3\' 6/9 TA Clones 63 5\' \\\\\\\\\\\\\\\\\\\\\\\\\* \\\\\\\\\\\\\\\\\*AA\\\\\* 3\' 2/9 TA CIones 63 5\' \\\\\\\\\C\\\\\G\\\\\\\\\* \\\\\\\\\\\\\\\\\\ A\\\\\* 3\' 1/9 TA CIones 63 1\. The U5-loop is in bold type. 2\. Spaces separate the PBS (indicated in bold type) from flanking sequence. 3\. PBS complementary to the 3\' terminal 18-nucleotide sequence of the indicated host tRNA. 4\. Asterisks represent conserved nucleotides. 5\. \"Input\" refers to the clone that was used to initiate viral infection in PBMCs. 6\. PCR product that was sequenced directly. 7\. Refers to TA clones of PCR product that is cloned into the Promega, P-Gem T-Easy Vector System I, to isolate individual colonies for sequencing. ::: Discussion ========== In previous studies, we have described HIV-1 in which the PBS and U5 have been altered to be complementary to tRNA^Met^, tRNA^Pro^and tRNA^Ile^\[[@B18],[@B24]\]. All of the viruses with only a PBS complementary to these tRNAs were replication competent but reverted to the wild type following infection in SupT1. To extend these studies to a more relevant cell type, we cloned the mutant PBS into the NL4-3 background, which replicates well in PBMC, reaching high levels of p24 antigen in the culture supernatant. Analysis of the effect of altering the PBS on infectivity of proviral clones revealed that these viruses were 10--40% as infectious as the wild type virus, with the virus containing a PBS complementary to tRNA^Ile^being the most infectious. However, analysis of the growth of these viruses revealed a clear preference for the viruses with a PBS complementary to tRNA^Met^compared to the virus with a PBS complementary to tRNA^Ile^. Virus with a PBS complementary to tRNA^Pro^had a rapid increase in p24 antigen after 21 days in culture and subsequently replicated similar to wild type and viruses with a PBS complementary to tRNA^Met^. As we had found in our previous studies, the sequence analysis of the PBS from both of these viruses at different times of culture revealed the reversion of the PBS to wild type \[[@B24]\]. The unexpected result from our studies was the distinct preference for HIV-1 to utilize tRNA^Met^as evidenced by the stability of the PBS following long-term culture. The preference of HIV-1 for the selection of tRNA^Met^was noted in a previous study in which we found a PBS complementary to this tRNA following analysis of the reversion of viruses that initially had a PBS complementary to tRNA^Trp^\[[@B24]\]. A subsequent study found that HXB2 derived viruses in which only the PBS was mutated to be complementary to tRNA^Met^reverted back to the wild type PBS; a virus that could stably use tRNA^Met^was obtained by additional mutations in the U5 \[[@B15]\]. Thus, the results of our current study are unique in that the NL4-3-Met, without mutations in the U5, was stable and replicated well in PBMC, at a level comparable to the wild type virus. Further characterization of the viruses obtained from these two cell types will be needed to resolve the reason for differences in stability of the PBS. It is possible that differences in nucleotide concentrations or tRNA availability between the SupT1 or PBMC could influence the stability of the PBS. Further experiments using an endogenous RT reaction \[[@B18]\] and analysis of virus tRNA^Met^content could be informative. With respect to the latter point though, our previous studies have not shown differences in tRNA content of virions that use alternative primers for reverse transcription \[[@B20]\]. How does this relate to the process of primer selection? In recent studies, we have found that HIV-1 most effectively selects tRNAs that have undergone the steps in tRNA biogenesis that result in transport from the nucleus to the cytoplasm \[[@B21]\]. Once in the cytoplasm, the tRNAs interact with a myriad of proteins involved in translation \[[@B26]\]. At any one time, the tRNA selected by HIV-1 as a primer for reverse transcription has been channeled into the translational process, supporting the idea of coupling of translation and primer selection. One possibility could be the coupling of primer selection with the synthesis of the Gag-polpolyprotein. Previous studies have shown that pseudovirions composed of Gagand Gag-polcontain the appropriate ratios of tRNA^Lys^found in intact wild type virions \[[@B7],[@B8]\]. That is, during the translation of Gag-pol, the tRNAs available for selection might be enriched for tRNA^Lys,3^and tRNA^Met^; conversely, tRNA^Ile^may not be favored because of the absence of isoleucine during translation of Gag-pol. This is not because isoleucine is excluded from the Gag-polprotein. Rather, it is possible that a translational event in the production of Gag-pol, possibly at or during the frame shifting, could influence the local amounts of tRNA so as to favor some (e.g., tRNA^Lys,3^, tRNA^Met^) while not others (e.g., tRNA^Ile^). Without tRNA^Ile^to occupy the PBS, there would be greater access by tRNA^Lys,3^to facilitate reversion back to the wild type PBS, complementary to tRNA^Lys,3^. Viruses with a PBS complementary to tRNA^Pro^, and from previous studies those with PBS complementary to tRNA^His^, tRNA^Lys1,2^or tRNA^Glu^, initially replicated slowly but reverted to use tRNA^Lys,3^, whereupon they exhibited rapid replication. We would predict that the local availability of these tRNAs would be sufficient to allow the limited replication. However, given the selective pressure for the use of tRNA^Lys,3^, the virus would have a propensity to revert to wild type if the tRNAs were not present at levels similar to tRNA^Lys,3^or tRNA^Met^. Coupling of the synthesis of Gag-polwith primer tRNA selection and encapsidation might provide all of the necessary components for the generation of infectious virus within the same intracellular locale. Further studies will be needed to explore the relationship between the synthesis of Gag-poland primer selection using the unique viruses described in this study. The results of our studies in which we included additional regions of complementarity between the tRNA and U5 further substantiates a role for this interaction in the selection of the tRNA primer \[[@B24]\]. In a recent study, we found that viruses with PBS and U5 complementary to tRNA^Lys1,2^or tRNA^His^were stable after extended replication in PBMC, similar to what we found for NL4-3-Met-AC \[[@B22]\]. In this study, the virus derived from NL4-3-Pro-AC replicated poorly and in contrast to NL4-3-Pro, did not revert to wild type following extensive in vitroculture in PBMC. This finding supports the idea that the complementarity between the U5 and tRNA can impact the selection process. Most probably, the NL4-3-Pro-AC remains stable because it can more effectively select tRNA^Pro^, or exclude tRNA^Lys,3^, from binding to the PBS complementary to tRNA^Pro^. If tRNA^Lys,3^is used, the PBS generated during plus strand synthesis would be complementary to tRNA^Lys,3^, which could facilitate reversion upon subsequent replication. The results of the current study and others are consistent with the concept that multiple interactions between the viral RNA genome and tRNA occur during the selection process \[[@B16],[@B19],[@B23],[@B24]\]. A recent study found that a virus that stably used tRNA^Lys1,2^could be generated by changing the PBS and a region upstream, different from the A loop, designated as the primer activation site (PAS) \[[@B23]\]. Interestingly, a virus with similar mutations to facilitate the use of tRNA^Pro^was not stable, consistent with the results presented in our study. We suspect that the U5-PBS interactions are more important for tRNA selection in primary cells (e.g., PBMC) where the availability of the tRNAs in the intracellular environment might be different. Further experiments will be needed to address this issue. In summary, the results of our studies analyzing the replication in PBMC of HIV-1 with PBS complementary to alternative tRNAs has revealed a clear preference for certain tRNAs to be selected for replication. The tRNA^Met^is highly favored for selection, slightly less than the wild type tRNA^Lys,3^, while tRNA^Ile^is not favored for selection as evidenced by the fact that viruses with this U5-PBS revert to use tRNA^Lys,3^after short term culture. Viruses that use tRNAs such as tRNA^Pro^, tRNA^Lys1,2^or tRNA^His^replicate poorly in PBMC compared to the wild type virus and the virus that uses tRNA^Met^\[[@B22]\]. These results suggest that HIV-1 can select the tRNA primer from a pool of tRNAs, with certain tRNAs favored over others, further substantiating a link between viral protein translation and primer selection. Conclusion ========== The results of our study provide new insights into the tRNA selection process by HIV-1. For the first time, we have described a unique HIV-1 that utilizes a tRNA primer (tRNA^Met^) that does not require additional mutations within the U5. This virus replicates well in human PBMC, similar to the wild type virus. In contrast, the virus did not prefer to select tRNA^Ile^as evidenced by the fact that this virus was unstable with or without additional mutations within U5. This result highlights that different tRNAs are available in PBMC for capture by HIV-1 for use as the primer for reverse transcription. The importance of additional mutations within U5 that are complementary to the anti-codon region of tRNAs for selection was highlighted by the studies with viruses in which the PBS was made complementary to tRNA^Pro^. In this case, the virus was unstable with only the PBS complementary to tRNA^Pro^while the additional U5 mutation did not allow reversion but severely impacted on the subsequent replication capacity of the virus, demonstrating that complex RNA-RNA interactions occur within the U5-PBS during primer selection. Collectively, the results of our studies demonstrate, for the first time, that distinct preferences exist for the selection of tRNAs to be used as the primer for HIV-1 reverse transcription. Coupled with our previous studies, we conclude there is most probably a link between viral translation and primer selection. The exclusive use of tRNA^Lys,3^by HIV-1 is most probably due to inherent features of this tRNA as well as the intracellular availability during viral translation. Methods ======= Construction of NL4-3 proviruses containing modified PBS regions ---------------------------------------------------------------- We previously reported the construction of pHXB2 (Met and Met-AC) and pHXB2 (Ile and Ile-AC) and PHXB2 (Pro and Pro-AC) with PBS and PBS-U5 changes complimentary to the respective tRNA 3\' and anticodon sequences \[[@B15],[@B18],[@B19],[@B24]\]. These proviral mutants were constructed in the pHXB2 molecular clone of HIV-1. In this study, the NL4-3 molecular clone of HIV-1 was used as the proviral backbone DNA for the U5-PBS mutants \[[@B27]\]. Proviral clones pHXB2 (Met, Met-AC, Pro, Pro-AC, Ile and Ile-AC) from these previous studies were digested with HpaI and BssHII restriction enzymes (New England Biolabs, Beverly, MA) to release an 868-bp fragment that contained the 5\' LTR, PBS, and leader region from the gaggene of HXB2. The HpaI site is located upstream of the 5\' LTR within the flanking sequence, and the BssHII site is located downstream of the PBS within the viral genome, in the proximity of nucleotide 255 (5\'GCGCGC-3\'). Digests were run on a 1% Agarose gel (Amresco, Solon, OH) to separate the 868 bp U5-PBS fragment from the pHXB2 proviral DNA fragment. U5-PBS fragments were isolated using the Qiagen Gel Extraction kit (Qiagen, Valencia, CA) and cloned into the NL4-3 HIV-1 proviral plasmid using the same BssHII and HpaI restriction sites. All resulting NL4-3 constructs were verified by DNA sequencing to ensure the identity of the mutated sequence and the successful ligation of the U5-PBS fragment into the pNL4-3. Tissue Culture and DNA transfections ------------------------------------ Transfections were performed according to the protocol for the Fugene 6 Transfection Reagent (Roche Molecular Biochemicals, Indianapolis, IN). Briefly, 2 μg of proviral plasmid DNA and 3 μL Fugene reagent were added to 100 μL of Dulbecco\'s modified Eagle\'s Medium (no Fetal Bovine Serum) (Cellgro by Mediatech, Herndon, VA). This mixture was incubated at room temperature for approximately 45 minutes then added drop-wise to one well of a 6-well plate containing 60% confluent 293T cells in DMEM with 10% Fetal Bovine Serum (FBS). The transfections were incubated overnight at 37°C and the medium was replaced with fresh DMEM containing 10% FBS (Hyclone, Logan, UT). After 48-hours, all supernatants were collected and stored at -80°C. Supernatants from transfected cells were assayed for HIV-1 p24 antigen (Beckman Coulter, Miami, FL) and infectivity \[[@B28]\]. PBMC Infections --------------- Human peripheral blood mononuclear cells (PBMC) were collected, stimulated using rIL-2 phytohemagglutinin (PHA) (Sigma, St. Louis, MO) and maintained as described previously \[[@B22]\]. Infections were performed by innoculating 20 × 10^6^PHA-stimulated PBMC with a volume of transfection supernatant containing 200 pg of p24 antigen and incubating for 2 hours at 37°C and 5% CO~2~. Virus/PBMC mixtures were transferred to 25 cm^2^tissue culture flasks, and the final volumes were adjusted to 10 mL with RPMI 1640, 1× (Cellgro by Mediatech, Herndon, VA) containing 15% FBS (Hyclone, Logan, UT) and 30 U/mL rIL-2 (Sigma, St. Louis, MO). Infected PBMC cultures were maintained for 10 weeks by replacing half the volume of medium every 7 days, without removing PBMC. Every 7 days, 1 mL of cell suspension was removed and centrifuged in an Eppendorf microcentrifuge at 24,000 × gfor 2 minutes. Supernatant was separated from the cell pellet and stored at -80°C for further analysis by p24 ELISA and JC53BL infectivity assays. Cell pellets were also stored at -80°C for isolation of high molecular weight DNA. Every 14 days an additional 5 × 10^6^PHA-stimulated PBMC were added to each culture. Infectivity Assay ----------------- Levels of infectious virus (IU/μL) in both 293T and PBMC culture supernatants were determined using the JC53BL assay as previously described \[[@B22],[@B28]\]. For a given test sample, the number of infectious units per microliter is equal to the number of blue cells in a well divided by the dilution factor for that well and represents the average of at least two wells. PCR analysis of integrated PBS-containing proviral DNA ------------------------------------------------------ Cell pellets from virus cultures were stored at -80°C, and isolation of high molecular weight genomic DNA was performed as described previously \[[@B22]\]. Approximately 2 μg of each genomic DNA sample was PCR amplified using 5 pmole/μL EcoRI(5\'-CGGAATTCTCTCCTTCTAGCCTCCGCTAGTC-3\') and 5 pmole/μL SphI(5\'-CCTTGAGCATGCGATCTACCACACACAAGGC-3\') primers (Gibco BRL, Rockville, MD) with 2.5 mM dNTP, 50 mM MgC1~2~(Invitrogen, Grand Island, NY), and 5 U/μL Recombinant TAQ DNA polymerase (Invitrogen, Grand Island, NY). The PCR program used to amplify genomic DNA had a denaturation temperature of 94°C and an annealing temperature of 56°C. The resulting PCR product was isolated and purified as described previously \[[@B22]\]. In cases of low virus replication (eg. pNL4-3-Pro-AC), the PCR product was used as a template for an additional PCR reaction (referred to as double PCR). Subcloning of PCR products and DNA Sequencing --------------------------------------------- Purified PCR product was sequenced for the U5-PBS region of the viral genome using the EcoRIprimer (Invitrogen, Grand Island, NY). DNA sequencing was performed on an automated DNA sequencer. PCR products that resulted in accordant sequence throughout the U5-PBS region were considered to be a homogenous infection of virus, using the same tRNA primer. PCR products that resulted in discordant sequence in the PBS region were considered to be a heterogeneous population of virus, using more than one tRNA to prime reverse transcription, and was therefore, subjected to further TA cloning in order to isolate sequences of individual viruses. This PCR product was subcloned according to the Promega pGEM-T Easy Vector System I (Promega, Madison, WI) to prepare the DNA for efficient and accurate sequencing. PCR product was ligated into the pGEM-T Easy plasmid vector at 4°C overnight. Ligations were then transformed into DH5α Escherichia colicells (Invitrogen, Grand Island, NY) and grown overnight on LB with 100 μg/mL ampicillin and 20 mg/mL Xgal. White colonies (indicating successful ligation) were picked and grown in LB-Amp100 μg/mL broth overnight at 37°C. DNA was harvested using the Qiagen QIAprep Spin miniprep kit, according to protocol (Qiagen, Valencia, CA). To assure that the TA clone DNA contained the PCR product insert, samples were digested by EcoRIto release the ligated fragment. Digests were run on a 1% agarose gel to verify the presence of a band of approximately 750 bp size. The TA clone DNA was then sequenced for the U5-PBS region using the EcoRIprimer. Acknowledgements ================ We would like to thank members of the Morrow laboratory for helpful discussion and Adrienne Ellis for preparation of the manuscript. KLM-R was supported by training grant (AI 07493). The UAB Center for AIDS Research Molecular Biology Core is acknowledged for help with the construction of the proviral clones (AI 27767). DNA Sequencing was carried out by Maria Salazar in the UAB CFAR DNA Sequencing Core (AI 27767). CDM acknowledges the helpful discussions from MAR. This research was supported by a grant from the NIH to CDM (AI 34749).	PubMed Central	2024-06-05T03:55:55.876295	2005-3-24	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084362/", "journal": "Retrovirology. 2005 Mar 24; 2:21", "authors": [ { "first": "Kenda L", "last": "Moore-Rigdon" }, { "first": "Barry R", "last": "Kosloff" }, { "first": "Richard L", "last": "Kirkman" }, { "first": "Casey D", "last": "Morrow" } ] }
PMC1084363	Background ========== Critical-care patients, such as premature newborns or adults suffering from severe respiratory illness, often need high concentrations of oxygen to maintain adequate blood oxygenation. Yet, high oxygen concentrations for extended periods of time are toxic and can lead to impaired lung function, respiratory failure and even death \[[@B1]-[@B3]\]. The sequence of events resulting in acute lung injury has been described in different models and involves alveolar edema, surfactant dysfunction, interstitial fibrosis \[[@B4]-[@B6]\], and subsequent deterioration of lung function \[[@B7]\]. At the cellular level, hyperoxia generates oxygen-derived free radicals that lead to damaged lung parenchymal, epithelial and endothelial cells \[[@B3]\]. Insulin-like growth factor-I (IGF-I) and its cognate receptor, the type 1 insulin-like growth factor receptor (IGF-1R), are known to be key modulators of growth and injury repair processes in several organs, including the lung. IGF-1R is essential for normal growth, development and differentiation, and mediates signals for the suppression of apoptosis and promotion of mitogenesis \[[@B8]\]. Interestingly, we \[[@B9]\] and others \[[@B10]\] have observed that IGF-1R expression was more intense in areas of increased proliferation in the developing lung of rodents. Furthermore, exposure of adult rats to prolonged hyperoxia was associated with an increase in IGF-1R immunostaining in the lung, in the vicinity of cells expressing IGF-I \[[@B11]\]. Likewise, the expression pattern of IGF-I and IGF-1R in the lung is altered in various diseases that require oxygen therapy, including cystic fibrosis \[[@B12],[@B13]\], lung fibrosis \[[@B14]-[@B16]\], acute respiratory distress syndrome or bronchopulmonary dysplasia \[[@B14]\]. Together, these observations have led to the suggestion that IGF signaling may be important in the pathogenesis of oxygen-induced lung injury. Interestingly, a link between IGF signaling and resistance to oxidative stress was found in studies on the biological mechanisms of aging. Holzenberger et al.recently showed that mutant mice with only one functional copy of the Igf-1rgene had an extended lifespan \[[@B17]\]. Remarkably, increased longevity in these mutants was associated with enhanced resistance to paraquat-induced oxidative stress, while there was a slight reduction in body growth and no alteration in metabolism, food intake, body temperature or onset of sexual maturation and fertility. These findings suggest that the final common pathway of IGF-1R action in the aging process may be modulation of the cellular response to oxidative damage. In this study, we sought to determine whether decreased functional levels of IGF-1R could prevent deleterious effects of oxygen exposure in the lung. Since IGF-1R null mutant mice die at birth from respiratory failure, we generated compound heterozygous mice harboring a hypomorphic (Igf-1r^neo^) allele and a knockout (Igf-1r^-^) allele \[[@B18]\]. These IGF-1R^neo/-^mice, strongly deficient in IGF-1R, were subjected to hyperoxia and analyzed for survival time, ventilatory function, pulmonary histopathology, morphometry, lung edema and vascular permeability. Methods ======= Animals ------- The mice were studied according to European recommendations on animal ethics. They were kept under standard conditions (average ambient temperature 23°C, 12/12-h light/dark cycle) with water and food provided ad libitumthroughout the study. Experiments were performed on male and female mice with a 129/Sv genetic background. They were either deficient in IGF-1R (IGF-1R^neo/-^) or had normal levels of IGF-1R (IGF-1R^+/+^). The hypomorphic IGF-1R^neo^allele was created by introducing a neomycin resistance cassette (neo) in intron 2 of the receptor gene using homologous recombination, and the receptor null allele (IGF-1R^-^) resulted from the complete elimination of exon 3, which encodes the type 1 IGF receptor ligand-binding domain \[[@B18]\]. We compared heterozygous IGF-1R^neo/-^mice (n = 78) with their wild-type IGF-1R^+/+^littermates (n = 84), which served as controls. Body weight and body temperature were systematically measured prior to each experiment. Seven-week-old mice were analyzed at three different time periods: under basal conditions, after hyperoxic exposure and/or during post-hyperoxia recovery. Genotyping of the mice was performed by PCR. Amplification conditions were as follows: 95°C for 3 min, and 35 cycles at 95°C for 35 s, 55°C for 45 s and 72°C for 1 min, with a final elongation step at 72°C for 7 min. To detect IGF-1R^neo^(hypomorphic), IGF-1R^-^(knockout) and IGF-1R^+^(wild-type) alleles simultaneously, three oligonucleotides were used. The primer sequences were 5\'-CATGGGTGTTAAATGTTAATGGC-3\' (Nex, sense-oriented), 5-ATGAATGCTGGTGAGGGTTGTCTT-3\' (Nvl, sense-oriented) and 5\'-ATCTTGGAGTGGTGGGTCTGTTTC-3\' (Nmt2, antisense-oriented). Hyperoxic exposure and post-hyperoxia recovery ---------------------------------------------- To induce hyperoxic lung injury, mice were exposed to 90% O~2~for 72 h in a sealed Plexiglas chamber of 9.7 liters, at 22--24°C. The O~2~and CO~2~levels in the chamber were monitored by O~2~and CO~2~analyzers (OM-11 and LB-2, respectively; Beckman Instruments, Schiller Park, IL, USA). Oxygen flow rate was adjusted to 1.5 L/min. The CO~2~level was always kept below 0.5%. Following hyperoxia, the mice were then placed in normoxic basal conditions. The survival study was performed in 31 IGF-1R^neo/-^and 30 IGF-1R^+/+^mice. They were checked every 2 h. Ventilatory physiology ---------------------- Analyses were performed sequentially in the same 5 IGF-1R^neo/-^and 6 IGF-1R^+/+^mice. There were three experimental conditions: before hyperoxia (in the basal, normoxic conditions), after hyperoxia, and during the post-hyperoxia recovery. Ventilatory parameters were recorded in awake and unrestrained mice placed in a whole-body plethysmograph, by using the barometric method \[[@B19]\]. The pressure signal resulting from breathing was detected using a differential pressure transducer (Validine DP103/12; Validine, Northridge, CA, USA) connected to the animal chamber (400 mL) and to a reference chamber of the same volume. The spirogram was recorded and stored on a computer using respiratory acquisition software (CIO-DAS 1602/16 interface and ELPHY software) for analysis off-line. Calibration was performed at the beginning of experiments by several injections of 50 μL air into the chamber. Each animal was weighed and placed in the chamber. A thermistal probe (BIO-BIT14) was inserted rectally and secured in place at the base of the tail and on the wall of the chamber. A protective muff was placed around the mouse to prevent stress. The protocol consisted of recording ventilatory parameters and rectal temperature before and immediately after the 72 h of hyperoxia. Measurements were made every 5 min while the mouse was breathing the following gas mixtures: 20 min in hyperoxia (100% O~2~) to obtain baseline values, 20 min in normoxia (21% O~2~), 10 min in mild hypoxia (12% O~2~) and 10 min in severe hypoxia (10% O~2~). The mouse was then allowed to recover in normoxia for 24 h prior to changing the experimental conditions, and the same protocol of recording was again applied. The CO~2~concentration in the chamber was always \< 1% at the end of each session. The following variables were measured and calculated by a computer-assisted method: minute ventilation (V~E~) and its two components: tidal volume (V~T~) and respiratory frequency (f~R~). For each 5 min recording, values were averaged on 50--100 contiguous breaths. Measurement of relative lung weight and lung histology analysis --------------------------------------------------------------- Mice were weighed and anesthetized with sodium pentobarbital (60 mg/kg). After bleeding, the lung was either immediately removed to measure the relative lung weight, or perfused in situwith neutral-buffered 10% formalin (15 cm H~2~O pressure) for histological analysis. The ratio of lung weight to body weight was determined in 11 IGF-1R^neo/-^and 16 IGF-1R^+/+^mice under basal conditions. For histological analysis, the trachea was ligated, the lungs were removed and fixed overnight in 10% formalin, and embedded in paraffin. Lung histology was documented both under basal conditions and after hyperoxic exposure, using 4 IGF-1R^neo/-^and 4 IGF-1R^+/+^mice for each condition. Three random 4-μm paraffin-embedded tissue sections from four different lungs from each group were stained with hematoxylin-eosin. The histopathology was reviewed in a blinded manner with respect to which group or mouse was being reviewed, using characteristic pathological traits including: alveolar destruction, edema, hemorrhage and hyaline membrane deposition. Morphometric analysis was performed under basal conditions and after hyperoxic exposure, using 4 IGF-1R^neo/-^and 4 IGF-1R^+/+^mice for each condition. Images were captured using the software TRIBVN ICS (release 2.04; IMAGIC, Bildverarbeitung AG, Glattbrug, Zürich, Switzerland). Six randomly selected fields (magnification 20×) were assessed in each mouse for (i) intra-alveolar hemorrhage and (ii) fibrin deposition which are easily identifiable pathological processes. Each process was scored on a scale of 0 to 4. Scores were assessed twice in a blinded manner. Zero with no apparent injury; 1, mild injury with \<25% lung involvement; 2, moderate injury with 25--50% lung involvement; 3, severe injury with 50--75% lung involvement; 4, very severe injury with \>75% lung involvement. An overall score of hyperoxia-induced lung injury was obtained by summation of scores of hemorrhage and of fibrin deposition. A mean ± SEM was generated from each group. In addition, five fields from each mouse were selected for the presence of perivascular edema, and used to quantify the edema by the ratio between the surface of edema and the surface of the vessel. Surface was estimated using a one-dimensional grid of points spaced 50 μm apart which was overlaid on each field. Lung edema analysis and vascular permeability by Evans Blue assay ----------------------------------------------------------------- Pulmonary edema was determined from the ratio between fresh lung weight at autopsy and dry weight. Lungs were dried at 80°C until their weight remained constant over 24 h. The wet-to-dry lung weight was assessed under basal conditions using 4 IGF-1R^neo/-^and 4 IGF-1R^+/+^mice, and after hyperoxic exposure using 8 IGF-1R^neo/-^and 7 IGF-1R^+/+^mice for each experiment. Vascular permeability related to lung injury was measured using the Evans Blue dye extravasation assay, as previously described \[[@B20]\]. The animals were anesthetized with a combination of ketamine (24 mg/kg) and xylazine (36 mg/kg), and received 30 mg/kg Evans Blue dye dissolved in 0.9% saline at a final concentration of 20 mg/ml by jugular vein injection, 10 min before sacrifice. The lungs were then perfused with 5 ml PBS containing 5 mM EDTA until all the blood has been removed. They were then dissected out, weighed and incubated at room temperature for 24 h in 4 ml formamide per g tissue (Sigma, St. Louis, MO, USA). The extravasation of Evans Blue-labeled albumin from the tracheobronchial microcirculation was quantified by measuring the optical density (OD) of the formamide extracts at wavelength of 620 nm. The quantity of Evans Blue dye extravasated in the airway tissues, expressed in ng/mg of dry tissue weight, was interpolated from a standard curve of Evans Blue concentrations (0.5--10 μg/ml). The Evans Blue dye assay was assessed both under basal conditions using 5 IGF-1R^neo/-^and 6 IGF-1R^+/+^mice, and after hyperoxic exposure using 6 IGF-1R^neo/-^and 7 IGF-1R^+/+^mice for each experiment. Statistical analysis -------------------- The values for all animals within each experimental group were averaged and standard deviations (SD) calculated. Statistical comparison of the results was performed between IGF-1R^+/+^and IGF-1R^neo/-^mice, and between normoxia and hyperoxia-treated groups, using the unpaired ANOVA t-test and the multiple comparison test (Bonferroni). The significance between the survival rate of two groups was determined by Kaplan-Meier analysis using the Cox statistical test. Differences were considered significant if P\< 0.05. Results ======= Survival of compound IGF-1R^neo/-^mice -------------------------------------- To assess the importance of IGF-1R for postnatal lung function, compound mice that harbored a hypomorphic (Igf-1r^neo^) allele and a knockout (Igf-1r^-^) allele were generated. As previously described \[[@B21]\], IGF-1R^neo/-^mice constitutively express only 25% of the IGF-1R levels typically found in wild-type (IGF-1R^+/+^) mice. In the IGF-1R^neo/-^mice, survival up to 6 months and fertility in young adults were unaltered (data not shown). Body weight of IGF-1R^neo/-^and IGF-1R^+/+^female mice did not differ significantly (-3.8%, NS), while the body weight of IGF-1R^neo/-^males was slightly reduced compared to IGF-1R^+/+^littermates (-16.6%, P\< 0.05). In both female and male mice, the lung to body weight ratio was similar in both genotypes (Fig. [1A](#F1){ref-type="fig"}). Under basal normoxic conditions, histological examination of the lungs of IGF-1R^neo/-^and IGF-1R^+/+^mice did not reveal fibrosis, nor inflammation, nor remodeling of intrapulmonary airways and lung parenchyma (Fig. [1B](#F1){ref-type="fig"}). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Lung weight and histology under basal conditions.(A) Mean values (×100) of lung-to-body weight ratio (± SD) in IGF-1R^+/+^mice (male: n = 11; female: n = 5) and IGF-1R^neo/-^mice (male: n = 7; female: n = 4). (B) Representative histology of hematoxylin-eosin stained lung sections from IGF-1R^+/+^and IGF-1R^neo/-^mice. Magnification: 10× objective. ::: ![](1465-9921-6-31-1) ::: To investigate whether IGF-1R contributed to hyperoxia-induced lung injury, IGF-1R^neo/-^mice and their wild-type littermate controls were exposed to 90% O~2~for 72 h and their survival was assessed during a recovery period of 48 h under normoxic conditions (Fig. [2](#F2){ref-type="fig"}). Forty-eight hours actually corresponded to the critical period during which mice either developed lethal respiratory symptoms or survived. The percentage of survivors following the initial period of hyperoxia was significantly greater in the IGF-1R^neo/-^mice (77%) than in IGF-1R^+/+^mice (53%) (P\< 0.05). Moreover, mortality predominantly affected mice within the initial 24-h period after hyperoxia (Fig. [2](#F2){ref-type="fig"}). ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Survival rate following exposure to 90% oxygen.The survival times of IGF-1R^+/+^mice (male: n = 15; female: n = 15) and IGF-1R^neo/-^mice (male: n = 16; female: n = 15) were measured following exposure to 90% oxygen. Data are expressed as percentage of mice alive at each time point. The asterisk indicates significant differences between IGF-1R^+/+^and IGF-1R^neo/-^mice, as determined by Kaplan-Meier analysis; P\< 0.05. ::: ![](1465-9921-6-31-2) ::: Ventilatory parameters in IGF-1R^neo/-^mice ------------------------------------------- We recorded ventilation in conscious IGF-1R^neo/-^and IGF-1R^+/+^mice placed in a whole-body plethysmograph while breathing 90% O~2~, 21% O~2~or under hypoxia (at 12% and 10% O~2~). Baseline ventilation and control of breathing during hypoxia were assessed by measuring the respiratory parameters such as minute ventilation (V~E~) and its two components: respiratory frequency (f~R~) and tidal volume (V~T~) before hyperoxia, after hyperoxia and during recovery in room air. Before hyperoxia (control conditions), baseline minute ventilation (V~E~) was slightly greater in IGF-1R^neo/-^mice than in IGF-1R^+/+^mice because of a greater tidal volume (V~T~), while respiratory frequency (f~R~) was similar in the two groups (Table [1](#T1){ref-type="table"}). During hypoxia, V~E~significantly increased under 12% and 10% O~2~in both groups, this effect being essentially due to a large increase in f~R~(Fig. [4](#F4){ref-type="fig"}). During this protocol, mean values of V~E~were constantly greater in IGF-1R^neo/-^mice than in IGF-1R^+/+^mice. However, the magnitude of the response to hypoxia -- which was obtained by calculating the per cent change in V~E~during severe hypoxia (10% O~2~) relative to hyperoxia -- was similar in the two groups. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Ventilatory parameters at baseline (90% O~2~) before hyperoxia, after hyperoxia and during recovery. ::: IGF-1R^+/+^ IGF-1R^neo/-^ ------ ------------- ------------------ ----------------- ---------------- -------------------- ----------------- V~E~ 2.03 ± 0.17 2.35 ± 0.32 ^\#^ 2.84 ± 0.69^\#^ 2.27 ± 0.22 \* 2.93 ± 0.37 \^\#^ 3.30 ± 0.62^\#^ V~T~ 10.0 ± 0.5 15.8 ± 2.0^\#^ 14.4 ± 2.2^\#^ 11.2 ± 0.8 \ 15.2 ± 1.1^\#^ 14.8 ± 1.1^\#^ f~R~ 203 ± 9 142 ± 30^\#^ 196 ± 32 203 ± 21 192 ± 18 \^\#^ 222 ± 35 Values are mean parameters ± SD in IGF-1R^+/+^mice (n = 6) and IGF-1R^neo/-^mice (n = 5) mice; V~E~, minute ventilation expressed in ml/min/g, V~T~, tidal volume in μl/g, f~R~, respiratory frequency in c/min; \# significant difference relative to control before hyperoxia; \ significant difference between the two genotypes (P\< 0.05). ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Ventilatory responses before and after hyperoxia, and during recovery.Minute ventilation (V~E~) in response to decreasing levels of inspired O~2~. In each one of the three conditions, mean values (± SD) of VE in IGF-1R^+/+^and IGF-1R^neo/-^mice were recorded under 90% O~2~, 21% O~2~and at two levels of hypoxia (12% and 10% O~2~). Values with asterisks are significantly different from control values at 90% O~2~. Crosses indicate significant differences from normoxic control values, and asterisks indicate significant differences between IGF-1R^+/+^and IGF-1R^neo/-^mice; P\< 0.05. ::: ![](1465-9921-6-31-4) ::: Following hyperoxia, the mice of both genotypes manifested an abnormal pattern of breathing. This pattern was characterized by a deep and slow breathing with expiratory pauses, particularly in the IGF-1R^+/+^mice (Fig. [3](#F3){ref-type="fig"}). Mean values of baseline V~E~were substantially increased compared with control conditions in both groups, this effect being due to a large increase in V~T~, whereas f~R~was moderately decreased (Table [1](#T1){ref-type="table"}). However, the hyperoxia-induced increase in V~E~was greater in IGF-1R^neo/-^mice because of a less pronounced diminution of f~R~. During hypoxia, V~E~did not increase under 12% or 10% O~2~in either of the genotypes, which indicates that prolonged hyperoxic exposure abolished the ventilatory response to hypoxia (Fig. [4](#F4){ref-type="fig"}). ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Respiratory pattern before and after hyperoxia.Typical respiratory pattern in a control IGF-1R^+/+^mouse (upper traces) and in an IGF-1R^neo/-^mouse (lower traces) under basal conditions and following hyperoxia. Expiratory pauses were observed in most control mice but not in IGF-1R^neo/-^mice. ::: ![](1465-9921-6-31-3) ::: During recovery in room air, V~E~and V~T~remained elevated while f~R~recovered its control values in both IGF-1R^neo/-^and IGF-1R^+/+^mice (Table [1](#T1){ref-type="table"}), and the stimulation of ventilation by hypoxia was restored (Fig. [4](#F4){ref-type="fig"}). Again, mean values of V~E~at 12% and 10% O~2~were significantly greater in IGF-1R^neo/-^mice than in IGF-1R^+/+^mice, but the magnitude of the response was similar in both groups. IGF-1R deficiency confers protection in 90% O~2~ ------------------------------------------------ As shown in Fig. [5](#F5){ref-type="fig"}, exposure to 90% O~2~for 72 h caused histological changes in the lungs of both IGF-1R^+/+^and IGF-1R^neo/-^mice, but injury was more severe in the lungs taken from IGF-1R^+/+^mice. The histological lesions included alveolar destruction, hyaline membrane formation, hemorrhage, and perivascular and peribronchiolar edema (Fig. [5](#F5){ref-type="fig"}). In contrast, hyperoxia-induced lung lesions were remarkably milder in IGF-1R^neo/-^mice, with less edema and hemorrhage. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Histology following hyperoxic injury.Hematoxylin-eosin stained lung sections of (a) IGF-1R^+/+^, and (b) IGF-1R^neo/-^mice following 72 h of hyperoxic exposure, illustrating perivascular and peribronchiolar edema as indicated by asterisks. Magnification: 10× objective. Representative histology of lung sections demonstrates (c) focal alveolar hemorrhages (black arrow) and hyaline membrane formation (gray arrow) in IGF-1R^+/+^lungs, and (d) minimal lesions in IGF-1R^neo/-^mice. Magnification: 40× objective. ::: ![](1465-9921-6-31-5) ::: Lung morphometry was used to measure the degree of hyperoxic injury. Analysis of a cumulative score which reflects both intra-alveolar hemorrhage and fibrin deposition revealed a significantly reduced score in hyperoxic-exposed IGF-1R^neo/-^mice as compared to IGF-1R^+/+^(Fig. [6A](#F6){ref-type="fig"}). Likewise, quantitative assessment of perivascular edema, as shown in Fig. [6B](#F6){ref-type="fig"}, confirmed that IGF-1R^neo/-^were protected from hyperoxia induced lung injury, and indistinguishable from IGF-1R^+/+^mice under basal conditions. ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Hyperoxia-induced lung injury, lung edema and vascular permeability.Parameters of lung damage were considered under basal conditions and following hyperoxia. (A) Lung injury scores from IGF-1R^+/+^(n = 4) and IGF-1R^neo/-^(n = 4) mice under basal conditions (normoxia) and under hyperoxia. A cumulative score was based on evaluation of hemorrhage and intraalveolar deposition (6 random sections per lung). (B) Quantitated perivascular edema from IGF-1R^+/+^(n = 4) and IGF-1R^neo/-^(n = 4) mice under basal conditions (normoxia) and under hyperoxia. Surface of perivascular edema was reported to the surface of the vessel, and noted S~E~/S~V~(5 selected sections per lung). (C) Lung edema was determined as mean values (± SD) of wet-to-dry lung weight ratio under basal conditions (normoxia) in IGF-1R^+/+^(n = 4) and IGF-1R^neo/-^(n = 4) mice, and under hyperoxia in IGF-1R^+/+^(n = 7) and IGF-1R^neo/-^(n = 8) mice. (D) Evans Blue dye extravasation was assessed as mean values (± SD) in IGF-1R^+/+^(n = 13) and IGF-1R^neo/-^(n = 11) mice. Crosses indicate significant differences between hyperoxic and normoxic values in IGF-1R^+/+^mice, and asterisks indicate significant differences between IGF-1R^+/+^and IGF-1R^neo/-^mice under hyperoxic conditions; P\< 0.05. ::: ![](1465-9921-6-31-6) ::: We quantified lung damage in IGF-1R^neo/-^and IGF-1R^+/+^mice using two parameters which are either relevant to edema or to vascular permeability (Fig. [6](#F6){ref-type="fig"}). Wet-to-dry lung weight ratio, which was measured as a supplementary mean to quantify edema, was noticeably reduced in IGF-1R^neo/-^mice (-26%, P\< 0.05) (Fig. [6C](#F6){ref-type="fig"}). This data was consistent with histomorphometric results (Fig. [6B](#F6){ref-type="fig"}). Moreover, lung vascular permeability, which was quantified by measuring the extravasation of Evans Blue dye, was reduced in IGF-1R^neo/-^mice (-49%, P\< 0.05) (Fig. [6D](#F6){ref-type="fig"}). Discussion ========== The main aim of the present study was to evaluate the importance of IGF-1R in acute hyperoxic lung injury by studying survival time, respiratory physiology and pulmonary histology. Since complete inactivation of Igf-1ris lethal at birth due to respiratory failure \[[@B18],[@B22]\], we combined a knockout allele (Igf-1r^-^) and a hypomorphic allele (Igf-1r^neo^) in order to generate viable mice with markedly reduced levels of IGF-1R \[[@B21]\]. Although haploinsufficiency of the Igf-1Rgene in humans has been suspected to be associated with lung hypoplasia \[[@B23]\], the low levels of IGF-1R expressed in IGF-1R^neo/-^mice were sufficient to ensure normal lung growth (Fig. [1](#F1){ref-type="fig"}). Thus, these mice provided a new way of exploring the in vivoeffects of a major deficiency in IGF-1R in the adult lung. The most striking result of this study was the improved survival of the IGF-1R^neo/-^mutant mice following hyperoxic injury, associated with increased protection against pulmonary damage. These results support the idea that IGF-1R plays a critical role in the regulation of pulmonary resistance to oxidative stress. Recently, using paraquat-induced stress which is known to cause multi-organ failure, Holzenberger et al.have shown that deficiency in IGF-1R is associated with improved oxidative stress resistance, as assessed by prolonged survival \[[@B17]\]. In this study, we further evaluated IGF-1R as one component of the mechanistic determinants of the lung injury process, and focused our attention on survival after hyperoxic-lung injury. IGF-1R^neo/-^mutant mice and their wild-type littermate controls were exposed to hyperoxia-induced oxidative stress, which has been used extensively in rodents as a valuable model of acute respiratory distress. Because sensitivity to hyperoxia has been described to be strain-dependent \[[@B24]\], these experiments were performed using IGF-1R^-^and IGF-1R^neo^alleles in a pure 129/Sv genetic background. Consistent with previous findings \[[@B25]\], acute respiratory distress was observed in our mice when they were exposed to \> 90% oxygen for 72 h. Overall, IGF-1R^neo/-^mice were more resistant to hyperoxic stress than littermate controls (Fig. [2](#F2){ref-type="fig"},[3](#F3){ref-type="fig"},[4](#F4){ref-type="fig"},[5](#F5){ref-type="fig"},[6](#F6){ref-type="fig"}). These results, together with the induction of IGF-1R in response to high oxygen concentration \[[@B11],[@B14]\], directly implicate IGF-1R in the development of oxidant-mediated lung injury, and indicate that deficiency in IGF-1R is associated with a higher capacity to endure and to recover from oxidant-induced injury. There are numerous mechanisms that can potentially interfere with resistance to hyperoxia, involving antioxidant enzymes, inflammatory mediators, cell cycle progression regulators, apoptosis and anti-apoptosis factors, endothelial cell-specific factors and the extracellular matrix repair system. Although assessment of the importance of IGF-1R in each of these mechanisms warrants specific attention, we made it our priority to determine which of the physiological traits in these mice is most likely to account for the phenotypic differences in response to acute lung injury. Hyperoxic exposure is known to induce severe lung dysfunction, which can dramatically affect gas exchange. Indeed, several studies have shown that prolonged hyperoxia causes a reduction in lung compliance \[[@B26],[@B27]\], and markedly alters the control of breathing through attenuating both carotid chemosensory responses and chemoreflex responses to hypoxia \[[@B28]\]. We hypothesized that such alterations may influence the resistance to oxidative stress, and subsequent survival of the mice. We found that hyperoxic exposure of the mice of either genotype, IGF-1R^neo/-^and IGF-1R^+/+^, abolished the ventilatory response to hypoxia, mainly caused by the absence of responsiveness of respiratory frequency (Fig. [4](#F4){ref-type="fig"}). These results clearly suggest that prolonged hyperoxia markedly alter carotid chemosensory activity and are consistent with previous studies in unanesthetized cats \[[@B1]\] and rats \[[@B29]\]. Interestingly, we observed also that the deleterious effects of hyperoxia on the hypoxic ventilatory drive were fully reversible under recovery conditions in room air. Thus, the blunted ventilatory response to hypoxia immediately after hyperoxic injury probably plays a critical role in the development and progression of respiratory distress exhibited by the mice. It is noteworthy that striking differences in pattern of breathing and ventilation were observed between IGF-1R^neo/-^and IGF-1R^+/+^mice, which may well contribute to the improved survival of the former group after hyperoxia. In particular, the abnormal pattern of breathing induced by hyperoxia, which limited the ventilatory response to hypoxia and precipitated respiratory failure, was more often elicited in the control mice than in the IGF-1R^neo/-^mice (Fig. [3](#F3){ref-type="fig"}). Also, minute ventilation at baseline, as well as in response to hypoxia, was greater in the IGF-1R^neo/-^mice. This latter finding suggests a pulmonary role for IGF-1R in response to hypoxia. This is consistent with a previous observation in which IGF-1R mRNA levels were specifically increased by 200% in the lung in response to hypoxia \[[@B30]\]. Pathogenesis of lung injury after hyperoxia has been studied extensively in mice. Prominent features include extensive alveolar and endothelial cell death, leading to disruption of the alveolo-capillary barrier, high pulmonary microvascular permeability and pulmonary edema. The hyperoxic morphological phenotype that we quantified by morphometry at 72 h of hyperoxia was consistently characterized by pulmonary edema, intra-alveolar hemorrhage and hyaline membrane formation in IGF-1R^+/+^, but was remarkably milder in the IGF-1R^neo/-^mice (Figs. [5](#F5){ref-type="fig"} and [6](#F6){ref-type="fig"}). These results are directly supportive of decreased sensitivity to hyperoxia being associated with IGF-1R deficiency. Protective mechanisms in which IGF-1R is suspected to interfere may include changes in the defense against accumulation of oxygen free radicals, although a recent study in patients with respiratory distress syndrome or bronchopulmonary dysplasia indicated only moderate induction of the antioxidant defense system \[[@B31]\]. Likewise, the response of mice to prolonged hyperoxia was not associated with altered expression of classical antioxidant enzymes such as catalase, MnSOD and Cu-Zn SOD \[[@B32]\]. Other differences may reside in the inflammatory response that leads to an influx of protein-rich fluid to the airspaces. By measuring changes in the ratio of wet-to-dry lung weight and in the Evans Blue extravasation, we found decreased permeability of the alveolo-capillary membrane in the lungs of IGF-1R^neo/-^mice compared with IGF-1R^+/+^mice (Fig. [6](#F6){ref-type="fig"}). These results further led to consider that IGF-1R may exert its influence on the integrity of alveolo-capillary membrane through endothelial cells. Increased integrity of the alveolo-capillary membrane in the mutant mice may result from enhanced survival and/or proliferation of the alveolar and endothelial cells in response to oxidative stress \[[@B10]\]. Studies have indicated that IGF-1R is present in vascular endothelium and vascular smooth muscle of oxygen-exposed lungs, both in patients and in animal models \[[@B11],[@B14]\]. To date, most studies on the regulation of vascular permeability have focused on vascular endothelial growth factor (VEGF) and related molecules. In the lung, however, their role in the control of permeability remains controversial \[[@B33]-[@B35]\]. Moreover, in vivostudies on the retinal neovascularization process have shown that VEGF is regulated through the activation of IGF-1R. This was demonstrated using either IGF-1R neutralizing antibody \[[@B36]\] or an endothelial cell-specific knockout of IGF-1 receptor \[[@B37]\]. Thus, it can be speculated that the deficiency of IGF-1R in IGF-1R^neo/-^mice contributes to the protection of lung endothelium from oxidant injury -- either directly or through intermediary modulation of VEGF expression. Taken together, these findings highlight the need for additional experiments to elucidate the true role of IGF-1R in the regulation of pulmonary permeability. Conclusion ========== Understanding of the molecular mechanisms allowing protection against pulmonary oxygen toxicity is crucial in order to enable the development of targeted and effective interventions aimed at preventing or alleviating oxidant-induced acute lung injury \[[@B38]\]. Our results support the novel notion that IGF signaling, for which numerous interfering drugs are under clinical development \[[@B39]\], plays an important role in the sensitivity to and recovery from oxygen lung injury. We consider that the findings of our study will precipitate additional research efforts to assess whether interference with IGF-1R signaling may be a way of reducing or preventing the development of acute respiratory distress syndrome. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= KA conceived of the experiment and carried out all animal studies. RE conceived of the design of the study, assisted in the acquisition of data and helped to draft the manuscript. MH provided the mouse model of IGF-1R deficiency, provided expert advice, and critically revised the manuscript. MB performed all ventilatory measurements and helped to draft the manuscript. JFF performed the histological analyses and provided expert advice and interpretation of the results. JP carried out the design of hyperoxic exposure. AC participated in the direction of the study and critically revised the manuscript. AHC conceived of the study and participated in its design and coordination, drafted and edited the manuscript. All authors read and approved the final manuscript. Acknowledgements ================ We specifically thank Pr. J.F. Bernaudin for expert help in the morphometric study. This work was supported by grants from Inserm, Assistance Publique Hôpitaux de Paris, Université Pierre et Marie Curie and Chancellerie des Universités (Legs Poix).	PubMed Central	2024-06-05T03:55:55.879528	2005-4-8	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084363/", "journal": "Respir Res. 2005 Apr 8; 6(1):31", "authors": [ { "first": "Karmene", "last": "Ahamed" }, { "first": "Ralph", "last": "Epaud" }, { "first": "Martin", "last": "Holzenberger" }, { "first": "Monique", "last": "Bonora" }, { "first": "Jean-François", "last": "Flejou" }, { "first": "Julien", "last": "Puard" }, { "first": "Annick", "last": "Clement" }, { "first": "Alexandra", "last": "Henrion-Caude" } ] }
PMC1084364	1. Introduction =============== The analysis of growth is an important component of many clinical and biological studies. The evolution of such mathematical functions as Gompertz, logistic, Richards, Weibull and Von Bertalanffy to describe population growth clearly indicates how this field has developed over the years. These models have proved useful for a wide range of growth curves \[[@B1]\]. In the logistic model, the growth curve is symmetric around the point of maximum growth rate and has equal periods of slow and fast growth. In contrast, the Gompertz model does not incorporate the symmetry restriction and has a shorter period of fast growth. Both the logistic and Gompertz have points of inflection that are always at a fixed proportion of their asymptotic population values. A number of recent publications have utilized some of these models. Kansal \[[@B2]\] developed a cellular automation model of proliferative brain tumor growth. This model is able to simulate Gompertzian tumor growth over nearly three orders of magnitude in radius using only four microscopic parameters. Brisbin \[[@B3]\] observed that the description of alligator growth by fixed-shape sigmoid models such as logistic, Gompertz or Von Bertalanffy curves may not be adequate because of the failure of the assumption that a constant curve shape holds across treatment groups. There are many applications of Gompertz, logistic and Von Bertalanffy models to multicellular tumor spheroid (MTS) growth curves \[[@B4]-[@B9]\]. Yin \[[@B10]\] introduced the beta growth function for determinate growth and compared it to the logistic, Gompertz, Weibull and Richards models. He showed that the beta function shares several characteristics with the four classic models, but was more suitable for accurate estimation of final biomass and duration of growth. Ricklef \[[@B18]\] investigated the biological implications of the Weibull and Gompertz models of aging. Castro \[[@B19]\] studied a Gompertzian model for cell growth as a function of phenotype using six human tumor cell lines. They concluded that cell growth kinetics can be a phenotypic organization of attached cells. West \[[@B20],[@B21]\] introduced an ontogenetic theory of growth, which is based on first principles of energy conservation and allocation. A review of these studies reveals that the sigmoid character of the classical three or more parameter growth functions, such as the logistic or Von Bertalanffy, may not adequately fit three-dimensional tumor cell cultures, which often show complex growth patterns. Models that have been found to provide the best fit were modified or generalized versions of the Gompertz or logistic functions. The 1949 data on the polio epidemic \[[@B11]\] provide another classic example of a situation in which none of the above models fit the data very well. Our purpose is to introduce three new growth models that have flexible inflection points and can fit data with different shapes. We apply our proposed models to the 1949 polio epidemic data \[[@B11]\] and Deisboeck\'s MTS volume data \[[@B9]\] and compare their fit with four classical models: logistic \[[@B12]\], Richards \[[@B13]\], Gompertz \[[@B14]\] and Weibull \[[@B15]\]. 2. The Hyperbolastic Model H1 ============================= First, we start by considering the following growth curve, which produces flexible asymmetric curves through nonlinear ordinary differential equations of the form ![](1742-4682-2-14-i1.gif) or ![](1742-4682-2-14-i2.gif) with initial condition P(t~0~) = P~0~ where P(t) represents the population size at time t, βis the parameter representing the intrinsic growth rate, θis a parameter, and Mrepresents the maximum sustainable population (carrying capacity), which is assumed to be constant, though in general the carrying capacity may change over time. For growth curves, βhas to be positive, leading to an eventually increasing curve with an asymptote at M; βcan be negative only for eventual inhibition curves or decay profiles. We refer to growth rate model (1) as the hyperbolastic differential equation of type I. If θ= 0, then the model (1) reduces to a logistic differential equation and equation (2) reduces to a general logistic model \[[@B12]\]. Solving the equation (1) for the population Pgives ![](1742-4682-2-14-i3.gif) where ![](1742-4682-2-14-i4.gif) and arcsinh(t) is the inverse hyperbolic sine function of t. We call the function P(t) in equation (2) the hyperbolastic growth model of type I or simply H1. To reduce the number of parameters, observed values of P~0~and t~0~are used to obtain an approximate value of α. Notice that the asymptotic value of P(t) is ![](1742-4682-2-14-i5.gif) From equation (1) we calculate the second derivative ![](1742-4682-2-14-i6.gif) If we set θ= 0, then the second derivative ![](1742-4682-2-14-i7.gif) when ![](1742-4682-2-14-i8.gif) In other words, when the population Preaches half the carrying capacity M, the growth ![](1742-4682-2-14-i9.gif) is most rapid and then starts to diminish toward zero. If we assume θ≠ 0, then the growth ![](1742-4682-2-14-i9.gif) is most rapid at the time t\, such that t\ satisfies the following equation ![](1742-4682-2-14-i10.gif) If the carrying capacity changes at discrete phases of a hyperbolastic growth, then a bi-hyperbolastic or multi-hyperbolastic model may be appropriate. 3. The Hyperbolastic Model H2 ============================= Now we consider an alternative growth curve through a nonlinear hyperbolastic differential equation of the form ![](1742-4682-2-14-i11.gif) with initial condition P(t~0~) = P~0~and γ\> 0, where tanh stands for hyperbolic tangent function, Mis the carrying capacity, and βand γare parameters. As in the H1 model, parameter βhas to be positive for increasing growth curves with an asymptote at Mand is negative only for decay profiles. We refer to the growth rate model (3) as the hyperbolastic differential equation of type II. Solving equation (3) for population size Pgives the three parameter model ![](1742-4682-2-14-i12.gif) where ![](1742-4682-2-14-i13.gif) We call the function P(t) in equation (4) the hyperbolastic growth model of type II or simply H2. As in the H1 model, observed values of P~0~and t~0~are used to obtain an approximate value of αand to reduce the number of parameters. Notice from equation (4) that for positive values of β, P(t) approaches Mas ttends to infinity and for negative values of β, P(t) approaches zero as ttends to infinity. Moreover, from equation (3), we calculate the second derivative ![](1742-4682-2-14-i14.gif) where csch and coth represent hyperbolic cosecant and hyperbolic cotangent, respectively. The growth rate ![](1742-4682-2-14-i9.gif) is most rapid at time t\* provided that t= t\* satisfies the following equation ![](1742-4682-2-14-i15.gif) If γ= 1, then the growth rate ![](1742-4682-2-14-i9.gif) is most rapid at time t= t\* if the following equality is true ![](1742-4682-2-14-i16.gif) 4. The Hyperbolastic Model H3 ============================= Finally, we consider a third growth curve through the following nonlinear hyperbolastic differential equation of the form ![](1742-4682-2-14-i17.gif) with initial condition P(t~0~) = P~0~where M is the carrying capacity and β, γand θare parameters. We refer to model (5) as the hyperbolastic ordinary differential equation of type III. The solution to equation (5) is a four parameter model P(t) = M- αEXP\[-βt^γ^- arcsinh(θt)\] (6) where α= (M- P~0~) EXP\[βt~0~^γ^+ arcsinh(θt~0~)\]. We call the function P(t) in equation (6) the hyperbolastic growth model of type III or simply H3. If θ= 0, then this model reduces to the Weibull function \[[@B15]\]. The growth rate ![](1742-4682-2-14-i9.gif) is most rapid at time t\* such that ![](1742-4682-2-14-i18.gif) If we define the a(t) as the rate of generation of new tumor cells and b(t) as the rate of loss of tumor cells, for instance, then ![](1742-4682-2-14-i19.gif) and ![](1742-4682-2-14-i20.gif). The growth rate can then be written as ![](1742-4682-2-14-i21.gif) If no tumor cells are lost (b(t) = 0), the tumor size P(t) follows the equation P(t) = M\[βt^γ^- arcsinh(θt)\]. 5. Application of Hyperbolastic Models ====================================== Statistical Analysis -------------------- We analyze two data sets by fitting the general logistic model of the form ![](1742-4682-2-14-i22.gif)\[[@B12]\], where ![](1742-4682-2-14-i23.gif) the Richards model of the form ![](1742-4682-2-14-i24.gif)\[[@B13]\], where ![](1742-4682-2-14-i25.gif) the Gompertz model of the form P(t) = MEXP\[-αEXP(-Mβt)\] \[[@B14]\], where ![](1742-4682-2-14-i26.gif) the Weibull model of the form P(t) = M- αEXP(-βt^γ^) \[[@B15]\], where α= (M- P~0~) EXP(βt~0~^γ^) and the hyperbolastic models H1, H2, and H3 described above. Obviously some of these models are closely related. Nonetheless, the parameter values may be quite different when these models are fitted to a single set of data. The logistic model used here is a two parameter symmetric model, while the Richards model generalizes the logistic model by introducing an additional parameter (γ) to the equation to deal with asymmetrical growth. The Richards function reduces to the logistic equation if γ= 1. The Gompertz equation, which is a two parameter asymmetric equation, attains its maximum growth rate at an earlier time than the logistic. In the Weibull equation, βand γare constants defining the shape of the response. In all seven models Mis a constant, the maximum value or the upper asymptote, which is estimated by non-linear regression. In each instance we express one model parameter (α) as the function of the other parameters and initial observed value P~0~at time t~0~, which allows us to reduce the number of parameters to be estimated and also anchors the first predicted value to the original value observed at the initial time point. The mean squared error (MSE) and the R^2^value from the nonlinear regression, as well as the absolute value of the relative error (RE), which was defined as ![](1742-4682-2-14-i27.gif) were used to indicate the prediction accuracy or goodness of fit for all seven fitted models. All models were fitted using SAS v.9.1 PROC NLIN (SAS Institute Inc., Cary, NC) and SPSS v.12.0.1 (SPSS Inc., Chicago, IL). The best fitting functions and their derivatives were plotted using Mathematica v.4.2 (Wolfram Research Inc., Champaign, IL) to find the growth rates and accelerations. Analysis of the Polio Epidemic Data ----------------------------------- In 1949, the United States experienced the second worst polio epidemic in its history. Table [1](#T1){ref-type="table"} gives the cumulative number or incidence of polio cases diagnosed on a monthly basis \[[@B11]\] and the number of cases predicted by each of the seven models. The data originally appeared in the 1949 Twelfth Annual Report of the National Foundation for Infantile Paralysis. Absolute values of RE, MSE and R^2^for the seven tested models are given in Table [2](#T2){ref-type="table"}. Average RE and MSE plots for the seven polio epidemic models are graphically presented in Figure [1](#F1){ref-type="fig"}. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Number of observed and predicted polio cases using seven models. ::: Month Polio Cases H1 H2 Weibull H3 Richards Logistic Gompertz --------- --------------- ---------- ---------- ----------- ---------- ------------ ------------ ------------ 0 494 494 494 494 494 494 494 494 1 759 242.47 544.99 494.15 467.57 838.76 901.13 1112.71 2 1016 278.44 720.94 505.28 452.96 1424.12 1632.93 2224.48 3 1215 526.24 1153.82 635.10 563.29 2418.00 2924.19 4016.77 4 1619 1279.87 2218.78 1334.67 1262.75 4105.48 5130.07 6649.96 5 2964 3506.99 4917.77 3769.25 3730.08 6970.11 8697.85 10223.41 6 8489 9510.12 11444.30 9859.95 9872.33 11824.88 13987.47 14754.76 7 22377 21110.00 23267.60 20665.19 20687.08 19913.74 20898.37 20177.15 8 32618 33011.70 34655.40 32802.34 32777.73 31513.43 28575.53 26352.24 9 38153 39160.25 39855.70 39775.43 39753.72 39660.86 35708.48 33093.26 10 41462 41203.34 41247.00 41264.73 41277.59 41382.42 41316.48 40190.97 11 42375 41763.32 41522.90 41339.95 41358.99 41573.16 45174.27 47437.25 ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Absolute value of the relative error(s), MSE and R^2^for seven tested models with polio data. ::: Month RE(H1) RE(H2) RE(W) RE(H3) RE(R) RE(L) RE(G) --------- -------------- -------------- --------------- --------------- --------------- ---------------- ---------------- 0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1 0.68 0.28 0.35 0.38 0.11 0.19 0.47 2 0.73 0.29 0.50 0.55 0.40 0.61 1.19 3 0.57 0.05 0.48 0.54 0.99 1.41 2.31 4 0.21 0.37 0.18 0.22 1.54 2.17 3.11 5 0.18 0.66 0.27 0.26 1.35 1.93 2.45 6 0.12 0.35 0.16 0.16 0.39 0.65 0.74 7 0.06 0.04 0.08 0.08 0.11 0.07 0.10 8 0.01 0.06 0.01 0.01 0.03 0.12 0.19 9 0.03 0.05 0.04 0.04 0.04 0.06 0.13 10 0.01 0.01 0.01 0.00 0.00 0.00 0.03 11 0.01 0.02 0.02 0.02 0.02 0.07 0.12 MSE 6.61 × 10^5^ 8.72 × 10^5^ 11.09 × 10^5^ 12.45 × 10^5^ 50.21 × 10^5^ 111.12 × 10^5^ 223.65 × 10^5^ R^2^ 0.9983 0.9978 0.9969 0.9969 0.9864 0.9667 0.9336 ::: ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Bar graphs represent mean(s) of the relative error(s) and mean squared error for the polio models. ::: ![](1742-4682-2-14-1) ::: The results show that the H1 ![](1742-4682-2-14-i28.gif) and H2 ![](1742-4682-2-14-i29.gif) models provide the best fit to the polio incidence data, followed by Weibull ![](1742-4682-2-14-i30.gif) and H3 ![](1742-4682-2-14-i31.gif) models. The Richards ![](1742-4682-2-14-i32.gif), logistic ![](1742-4682-2-14-i33.gif) and Gompertz ![](1742-4682-2-14-i34.gif) models are clearly inadequate to describe the polio incidence growth pattern (Figures [1](#F1){ref-type="fig"} and [2](#F2){ref-type="fig"}). The second derivative of the fitted H1 function suggests that the highest incidence of the polio epidemic cases occurred between July and August of 1949 (Figure [3](#F3){ref-type="fig"}). ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Area represents the error between the observed and predicted polio cases for the seven tested models in the following order starting from top left: H1, H2, Weibull, H3, Richards, logistic and Gompertz. ::: ![](1742-4682-2-14-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Curves represent a) predicted number of polio cases using best fitting H2 model b) first derivative of the previous function or the growth rate of the polio outbreak and c) second derivative or acceleration of the polio outbreak. ::: ![](1742-4682-2-14-3) ::: Analysis of the MTS Growth Data ------------------------------- In 2001, Deisboeck et al. \[[@B9]\] studied the development of multicellular tumor spheroids (MTS) by creating a microtumor model. They claimed that a highly malignant brain tumor is an opportunistic, self-organizing and adaptive complex dynamic bio-system rather than an unorganized cell mass. Mature MTS possess a well-defined structure, comprising a central core of dead cells surrounded by a layer of non-proliferating, quiescent cells, with proliferating cells restricted to the outer, nutrient-rich layer of the tumor. Angiogenesis is a process by which new blood vessels are created from existing ones. A cell, which would be malignant, detaches from the tumor and uses the new blood supply to travel throughout the body. These authors suggested that such growth can be described by both the Gompertz and logistic functions. Using Deisboeck\'s MTS with \"heterotype attractor\" data, the four classical models were compared with the hyperbolastic ones to identify which model predicted the MTS volume most accurately. The observed and predicted MTS volume values are presented in Table [3](#T3){ref-type="table"}. The absolute values of RE, MSE and R^2^for each model are given in Table [4](#T4){ref-type="table"}. The average RE and MSE plots for the seven cancer volume models are graphically presented in Figure [4](#F4){ref-type="fig"}. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Observed and predicted MTS volume using seven models. ::: Time(hr) Volume H3 Weibull H1 H2 Gompertz Logistic Richards ------------ ---------- ------- ----------- ------- ------- ------------ ------------ ------------ 0 0.087 0.087 0.087 0.087 0.087 0.087 0.087 0.087 24 0.080 0.080 0.088 0.067 0.089 0.099 0.107 0.108 48 0.082 0.083 0.096 0.093 0.099 0.116 0.132 0.134 72 0.129 0.127 0.125 0.133 0.125 0.140 0.162 0.165 96 0.188 0.189 0.184 0.186 0.182 0.177 0.200 0.202 120 0.255 0.256 0.259 0.251 0.261 0.234 0.245 0.245 144 0.318 0.317 0.317 0.320 0.316 0.327 0.302 0.297 ::: ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Absolute value of the relative error(s), MSE and R^2^for seven tested models with MTS volume data. ::: Time(hr) REH3 REW REH1 REH2 REG REL RER ------------ --------------- ---------------- ---------------- ---------------- ----------------- ----------------- ------------------ 0 0.000 0.000 0.000 0.000 0.000 0.000 0.000 24 0.002 0.096 0.166 0.109 0.239 0.339 0.353 48 0.016 0.168 0.132 0.201 0.416 0.607 0.634 72 0.019 0.030 0.029 0.028 0.089 0.257 0.277 96 0.007 0.020 0.009 0.034 0.059 0.061 0.072 120 0.002 0.016 0.016 0.022 0.083 0.038 0.038 144 0.002 0.004 0.007 0.006 0.030 0.052 0.066 MSE 3.33 × 10^-6^ 74.21 × 10^-6^ 82.83 × 10^-6^ 109.2 × 10^-6^ 464.35 × 10^-6^ 802.06 × 10^-6^ 1374.42 × 10^-6^ R^2^ 0.9998 0.9974 0.9974 0.9957 0.9808 0.9100 0.8972 ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Bar graphs represent mean(s) of the relative error(s) and mean squared error for the MTS volume growth models. ::: ![](1742-4682-2-14-4) ::: The results indicate that the H3 model ![](1742-4682-2-14-i35.gif) has superior prediction accuracy for this particular data set. It is followed by the Weibull ![](1742-4682-2-14-i36.gif), H1 ![](1742-4682-2-14-i37.gif) and H2 ![](1742-4682-2-14-i38.gif) models, which predict with similar accuracy. Finally, the Gompertz ![](1742-4682-2-14-i39.gif), logistic ![](1742-4682-2-14-i40.gif) and Richards ![](1742-4682-2-14-i41.gif) models resulted in the least precise fit of the seven (Figures [4](#F4){ref-type="fig"} and [5](#F5){ref-type="fig"}). Even though the Weibull model was the second best, the mean relative error associated with it was almost seven times the mean relative error for the best-fitting H3 model. Over 144 hours, MTS growth follows decelerating growth dynamics with some shrinking during early stages (Figure [6](#F6){ref-type="fig"}). The first derivative of P(t) (growth rate) indicates that the MTS volume growth rate is zero at t = 4.90 hours and t = 34.27 hours (Figure [6](#F6){ref-type="fig"}). The second derivative of the fitted H3 function shows that the acceleration is slowest at t = 15.27 hours and fastest when t = 103.03 hours (Figure [6](#F6){ref-type="fig"}). ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Area represents the error between the observed and predicted MTS volume for the seven tested models in the following order starting from top left: H3, Weibull, H1, H2, Gompertz, logistic and Richards. ::: ![](1742-4682-2-14-5) ::: Figure [7](#F7){ref-type="fig"} compares the MTS rate of generation of new tumor cells (a(t)) to the rate of loss of tumor cells (b(t)). One can clearly see that the gap between the two rates becomes smaller and gradually approaches zero. Notice that as the gap approaches zero, the growth rate ![](1742-4682-2-14-i9.gif) also approaches zero. ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Functions represent the rate of generation of new tumor cells a(t) and rate of loss of tumor cells b(t). ::: ![](1742-4682-2-14-7) ::: 6. Discussion ============= Obviously no model can accurately describe every biological phenomenon that researchers encounter in their practice and the same is true for our models. Many models have been developed to deal with sigmoid growth \[[@B16]\] and new ones are continuously being proposed. The logistic function is symmetric around the point of inflection. The Richards function is more flexible and can fit asymmetric growth patterns \[[@B10],[@B17]\]; however, it has more parameters than the logistic function. The Gompertz function has the same number of parameters as the logistic function and the Weibull function has the same number of parameters as the Richards function and both can fit asymmetric growth, but they are not very flexible \[[@B10]\]. The H1 function has one more parameter than the logistic and Gompertz functions, but it is more flexible and can fit asymmetric growth patterns as well as increasing and decreasing growth, as shown in the MTS volume example. The H2 function has the same number of parameters as H1 and can fit asymmetric curves, but it cannot fit decreasing growth patterns, so it is less flexible. The H3 function has the same flexibility as the H1 function at the expense of one more parameter, similar to the Weibull and Richards equations. Some of the flexibility of the H1, H2 and H3 functions is illustrated in Figure [8](#F8){ref-type="fig"}. The logistic and Gompertz functions have two parameters that are easily interpretable. Like Yin \[[@B10]\], we encountered problems in trying to provide initial parameter values in the Weibull function. One can arrive at a satisfactory solution by trial and error, or using a grid search in SAS PROC NLIN by providing a range of starting values. These functions can be easily implemented in SPSS or SAS PROC NLIN (see [Additional file 1](#S1){ref-type="supplementary-material"}) or other readily available software packages. Non-linear function parameters that have biological meaning are more advantageous for statistical parameterization of such equations. The same can be said for some of the parameters in the three proposed models, which can be determined by summarizing the data or using the above suggestions. Table [5](#T5){ref-type="table"} provides estimates for the parameters of the H1, H2 and H3 models. If necessary, an additional parameter called the shift parameter may be added to a model to improve the fit of the data to a model. ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Parameter estimates (with standard errors in parentheses) for H1, H2 and H3 models applied in two examples. ::: Model Parameter Polio Estimate MTS Estimate --------- ------------- ------------------------------ ---------------------------- H1 M 41951.4 (603.8) 0.5633 (0.16) β 3.9 × 10^-5^(3.06 × 10^-6^ 0.0395 (0.02) θ -2.6851 (0.29) -0.2171 (0.05) H2 M 41574.9 (670.6) 0.3360 (0.03) β 2.9 × 10^-6^(7.02 × 10^-7^) 1.9 × 10^-5^(4.7 × 10^-5^) γ 1.8865 (0.13) 2.6784 (0.55) H3 M 41359.6 (817.5) 0.5871 (0.02) β 4.11 × 10^-6^(5.62 × 10^-6^) 0.0371 (0.01) γ 6.18 (0.67) 0.8575 (0.05) θ -0.00065 (0.01) -0.0256 (0.003) ::: While the results presented are valid only for the data sets used in this study, these models can have much wider application than shown here. We successfully applied them to several other data sets including craniofacial and stem cell growth data and the results indicate supreme prediction accuracy for the hyperbolastic models. Based on the results presented in this paper and others not shown here, we can say that the H3 model performs the best with cancer cell, craniofacial and stem cell growth data. However, it is reasonable to compare models for fit before deciding on the selection of the \"best\" one. With appropriate parameter adjustments in H1 or H2, one can derive regression type models for dichotomous or polytomous response variables, and use these models in survival data problems, reliability studies, business applications and many other situations. Finally, our hyperbolastic models show very promising results. In both the above discussed data sets, they fitted the data with smaller MSE, smaller mean RE and higher prediction accuracy than the logistic, Richards and Gompertz, which were the worst fit models in both cases. Our models are accurate and simple and two of them generalize the logistic and Weibull models. They can be easily implemented and tested in readily available software packages or routines. We strongly believe that choosing a flexible and highly accurate predictive model such as hyperbolastic can significantly improve the outcome of a study and it is the accuracy of a model that determines its utility. We strongly recommend usage of such models to the scientific community and practitioners and urge comparison of them with classical models before decisions on model selection are made. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= MT carried out the mathematical derivations, programming and testing of the models and the drafting and reviewing of the manuscript. DKW was involved in verifying the mathematical derivations, programming the models and reviewing the manuscript. ZB participated in the derivations, verification and formatting of the functions, programming and testing of the models and writing the manuscript. ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Curves represent a) predicted MTS volume using best fitting H3 model b) first derivative of the previous function or the growth rate of the MTS volume and c) second derivative or the acceleration of the MTS volume growth. ::: ![](1742-4682-2-14-6) ::: ::: {#F8 .fig} Figure 8 ::: {.caption} ###### Functions illustrate the flexibility of the H1, H2 and H3 models. One parameter is varied while the others are held constant to demonstrate the capability of the models to fit different growth or decay patterns. In all examples parameter M is held constant at 100. ::: ![](1742-4682-2-14-8) ::: Supplementary Material ====================== ::: {.caption} ###### Additional File 1 SAS code used to fit H1, H2 and H3 models to MTS volume data. ::: ::: {.caption} ###### Click here for file :::	PubMed Central	2024-06-05T03:55:55.882715	2005-3-30	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1084364/", "journal": "Theor Biol Med Model. 2005 Mar 30; 2:14", "authors": [ { "first": "Mohammad", "last": "Tabatabai" }, { "first": "David Keith", "last": "Williams" }, { "first": "Zoran", "last": "Bursac" } ] }
PMC1087199	Mental health is perhaps the most neglected area of health policy and programming. According to the 2001 World Health Report, "some 450 million people suffer from a mental or behavioral disorder, yet only a small minority of them receive even the most basic treatment" \[[@pmed-0020071-b1]\]. More than 40% of countries have no mental health policy and over 30% have no mental health program. Over 90% of countries have no mental health policy that includes children and adolescents \[[@pmed-0020071-b1]\]. According to the World Health Organization (WHO), mental and behavioral disorders are estimated to account for 12% of the global burden of disease, yet the mental health budgets of the majority of countries constitute less than 1% of their total health expenditures \[[@pmed-0020071-b1]\]. The relationship between disease burden and disease spending is clearly disproportionate. Those few who do receive services often fare just as badly. Mental Disability Rights International (MDRI; Washington, D.C., United States), a human rights group dedicated to the promotion of rights of the mentally disabled, has documented how, in many countries, severely mentally disabled individuals become targets of stigma, discrimination, and other human rights abuses. Routinely, children and adults with mental disabilities are arbitrarily detained in psychiatric facilities, social care homes, orphanages and other closed institutions. Out of public view, they are subject to the most extreme forms of inhuman and degrading treatment experienced by any population ([Figure 1](#pmed-0020071-g001){ref-type="fig"}). In Kosovo, MDRI learned that women were raped in psychiatric facilities in plain view of local staff and international humanitarian relief workers \[[@pmed-0020071-b2]\]. In Hungary and Paraguay, MDRI found people locked in cages \[[@pmed-0020071-b3],[@pmed-0020071-b4]\]. In Turkey, Peru, and Bulgaria, MDRI investigators learned of a practice called "unmodified ECT"---the use of without any form of anaesthesia or muscle relaxants---a practice that is both painful and dangerous \[[@pmed-0020071-b5],[@pmed-0020071-b6]\]. ::: {#pmed-0020071-g001 .fig} Figure 1 ::: {.caption} ###### Bathroom in the Men\'s Chronic Ward of Larco Herrera Hospital, a Government-Supported Psychiatric Hospital in Lima, Peru An investigation of the men\'s chronic ward, by MDRI and Asociación Pro Derechos Humanos (a Peruvian nonprofit organization that works toward human rights), found that "conditions were stark, the bathrooms filthy, and severe regimentation denied patients\' basic autonomy" \[[@pmed-0020071-b5]\]. (Photo: Mental Disability Rights International) ::: ![](pmed.0020071.g001) ::: Defining a Human Rights Approach to Mental Health Policy {#s2} ======================================================== The starting point for the development of a human-rights based policy on mental health is that mentally ill individuals are full human beings who are entitled to rights. Although seemingly obvious, in practice MDRI has found that the implications of these premises challenge predominant biomedical approaches to mental illness, as well as health services paradigms \[[@pmed-0020071-b7]\]. In a rights framework, "mental health needs" are not analyzed (as they are in many studies) in terms of the application of given diagnostic criteria in isolation from the social context that leads to use of the mental health sector, and mentally disabled persons are treated as more than patients who need services \[[@pmed-0020071-b7],[@pmed-0020071-b8]\]. They have rights to exercise agency in their own lives and to participate as members of their communities and societies, and these rights trump other concerns such as general attitudes toward risk containment in society \[[@pmed-0020071-b9]\]. Thus, a human rights framework calls for changes that go beyond quality of care to include both legal and services reforms. Further, a human rights approach demands that we develop policies and take actions to end discrimination in the overall society that has a direct effect on the health and well-being of the mentally disabled. Legal Reform and Accountability {#s3} =============================== Suspicion of mental illness cannot mean untrammeled discretion to disregard due process concerns in detention. Whether or not ideological factors are at play, civil commitment laws must provide for minimum substantive and procedural protections that protect mentally ill individuals\' fundamental agency \[[@pmed-0020071-b10]\]. This is often not the case. For example, MDRI found that civil commitment laws in Uruguay and Mexico allow commitment upon medical certification of "mental illness," which MDRI found, in many cases, to be questionable. There are no requirements that a patient be dangerous or in need of psychiatric treatment. These laws do not require a right to counsel or a periodic review of commitment, as international law requires \[[@pmed-0020071-b11],[@pmed-0020071-b12]\]. Thus, people who are found "mentally ill" can be deprived of their liberty indefinitely in these---and many other---countries. Once mental health is construed in terms of human rights, all states are required, at a very minimum, to establish a normative framework consistent with international law \[[@pmed-0020071-b13]\]. Such a normative framework provides for procedural protections; it also provides for human rights oversight and remedies in the event of abuses. Mentally disabled persons have the same right to redress for violations of their fundamental rights as other people do \[[@pmed-0020071-b13],[@pmed-0020071-b14],[@pmed-0020071-b15]\]. Recourse may imply judicial remedies but in some cases a human rights ombudsman can be equally effective \[[@pmed-0020071-b16]\]. As mentally disabled individuals are often not in a position to avail themselves of remedies, proactive monitoring and enforcement is also necessary. In short, instituting legal reform, accountability procedures, and effective mechanisms to provide human rights oversight becomes a cornerstone of a human rights-based approach to mental health policy. Services Reform: Community Integration and Participation {#s4} ======================================================== Treating mentally ill people as full human beings implies that they have rights to participate in their communities and societies. This, in turn, calls for community integration and service system reform instead of programs that merely rebuild segregated institutions. The WHO also recognizes that it is important to provide treatment in the community, but its reasoning is largely utilitarian. The WHO argues that "community care has a better effect than institutional treatment on the outcome and quality of life of individuals with chronic mental disorders. Shifting patients from mental hospitals to care in the community is also cost-effective" \[[@pmed-0020071-b1]\]. From a human rights perspective, people are entitled to live in and receive care in the community not because it is more efficient, but because all human beings develop their identities within social contexts, and have rights to work and study, as well as be with family and friends. > People with mental disabilities are often denied the right to work outside the home, to marry or have children A rights-based approach calls not only for the location of care in the community, but also for the transfer of planning and decision-making power to the individuals and communities that the health system is supposed to serve. In this case, consumers and family members must be integrally involved in the policy-making and programming decisions \[[@pmed-0020071-b13],[@pmed-0020071-b17],[@pmed-0020071-b18],[@pmed-0020071-b19]\]. MDRI has repeatedly found that funds are misdirected toward rebuilding psychiatric institutions and orphanages. Further, international institutions often undermine rights-based approaches to policy. MDRI has documented how European governments, development banks and international humanitarian relief organizations fund projects to build new psychiatric institutions and orphanages throughout the Americas and Eastern Europe, rather than focusing on community care \[[@pmed-0020071-b9]\]. Non-Discrimination: Within and beyond the Health Sector {#s5} ======================================================= Nondiscrimination is the most fundamental tenet in human rights. Under international law, discrimination need not be intentional nor de jure (in law) to constitute a violation of various relevant treaties, but merely needs to have the "effect of nullifying or impairing the equal enjoyment or exercise" of rights (paragraph 11 of \[[@pmed-0020071-b20]\]). MDRI has found that discrimination against people with mental disabilities is pervasive, and takes many forms. In Peru, for example, the public health insurance scheme does not cover mental disorders \[[@pmed-0020071-b5]\]. However, it is critical to recognize that discrimination outside the health sector also affects well-being. MDRI has seen that children with intellectual disabilities are denied equal access to education in Peru. They are placed in programs that in effect warehouse them and assume that they are unable to learn \[[@pmed-0020071-b5]\]. As adults, people with mental disabilities are often denied the right to work outside the home, to marry or have children, or to take part in the religious and social activities that define people as adult members of society \[[@pmed-0020071-b21]\]. A human rights approach to mental health policy demands that special attention be placed on remedying such inequities---both within and beyond the health sector---which affect the physical, mental, and social well-being of persons with mental disabilities \[[@pmed-0020071-b22]\]. Such an approach depends upon multisectoral strategies including education, housing, and work, and establish that people with mental disabilities are full citizens. Conclusions and Reflections on First Steps {#s6} ========================================== MDRI works through a human rights framework that links the improvement of mental health services with broader questions of social justice and nondiscrimination relating to the full spectrum of rights set out in international instruments \[[@pmed-0020071-b13],[@pmed-0020071-b14]\]. As an immediate first step, all states---regardless of resources---can develop national mental health policies and plans of action with measurable targets, which provide for open public discussion \[[@pmed-0020071-b13],[@pmed-0020071-b23]\]. Devising a national policy is a precondition to creating rights-based programs that address the multivalent problems faced by persons with mental disability \[[@pmed-0020071-b1]\]. Stakeholder participation in the process affirms that mentally disabled people are rights-worthy \[[@pmed-0020071-b13],[@pmed-0020071-b17]\]. Both international agencies and professional associations can play critical roles in providing technical assistance to countries to develop rights-based national mental health policies \[[@pmed-0020071-b1],[@pmed-0020071-b4]\]. Bilateral and multilateral donors should encourage such rights-based policies through their funding prerogatives. Citation: Yamin AE, Rosenthal E (2005) Out of the shadows: Using human rights approaches to secure dignity and well-being for people with mental disabilities. PLoS Med 2(4): e71. This is the first in a series of two articles commissioned to coincide with the April 2005 conference at Emory University, "Lessons Learned from Rights Based Approaches to Health" (<http://humanrights.emory.edu>). MDRI : Mental Disability Rights International WHO : World Health Organization [^1]: Alicia Ely Yamin is a human rights attorney, an instructor at the Harvard School of Public Health, Boston, Massachusetts, United States of America, and a board member of Mental Disability Rights International, Washington, D.C., United States of America. Eric Rosenthal is the founder and executive director of Mental Disability Rights International. [^2]: Competing Interests: The authors declare that they have no competing interests.	PubMed Central	2024-06-05T03:55:55.885888	2005-4-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1087199/", "journal": "PLoS Med. 2005 Apr 26; 2(4):e71", "authors": [ { "first": "Alicia Ely", "last": "Yamin" }, { "first": "Eric", "last": "Rosenthal" } ] }
PMC1087200	Introduction {#s1} ============ An increasing prevalence of drug-resistant parasites now threatens the efficacy of the antimalarial drugs chloroquine (CQ), sulphadoxine-pyrimethamine (SP), and amodiaquine (AQ) in sub-Saharan Africa \[[@pmed-0020092-b01],[@pmed-0020092-b02],[@pmed-0020092-b03]\]. In response, there is currently a concerted effort to evaluate new combinations of antimalarials \[[@pmed-0020092-b04]\]. In addition to direct therapeutic benefits, there is the important opportunity to select combination therapy that reduces post-treatment parasite transmission so as to counteract the selection and transmission advantage of drug-resistant parasites \[[@pmed-0020092-b04],[@pmed-0020092-b05],[@pmed-0020092-b06]\]. Recently published data show that the appearance in many geographical locations of parasites resistant to sulphadoxine and to pyrimethamine is not necessarily due to multiple de novo events, but rather that new resistant haplotypes have arisen extremely rarely and subsequently spread by selection pressure \[[@pmed-0020092-b07]\]. The goal of reducing this spread therefore requires that combination therapy minimises the number of treated individuals harbouring gametocytes, the transmissible sexual stages of the Plasmodium life cycle, and thus does not favour the survival and transmission of drug-resistant genotypes \[[@pmed-0020092-b08],[@pmed-0020092-b09]\]. In a series of clinical trials in Farafenni, The Gambia, we have evaluated the effects of combination treatment on the subsequent infectiousness to mosquitoes of children treated for uncomplicated falciparum malaria. Children who received three daily doses of artesunate (AS) in addition to SP \[[@pmed-0020092-b06]\] or CQ \[[@pmed-0020092-b09],[@pmed-0020092-b10],[@pmed-0020092-b11]\] had significantly reduced gametocyte carriage and reduced infectiousness to mosquitoes compared to those who received either SP or CQ alone. However, the addition of three doses of artesunate did not wholly prevent gametocyte emergence in either case, and thus onward transmission occurred \[[@pmed-0020092-b06],[@pmed-0020092-b09],[@pmed-0020092-b10]\]. As SP-resistant parasites occur in The Gambia at low frequency \[[@pmed-0020092-b12]\], and CQ-resistant parasites are common \[[@pmed-0020092-b08]\], both SP/AS and CQ/AS will permit the transmission of resistant parasites from treated patients \[[@pmed-0020092-b09],[@pmed-0020092-b10]\]. The co-formulation of artemether and lumefantrine, known as co-artemether (Coartem, Riamet), is an alternative artemisinin-based combination currently being evaluated in African clinical trials \[[@pmed-0020092-b13],[@pmed-0020092-b14]\]. In areas of drug-resistant malaria, the World Health Organization (WHO) has recommended that co-artemether be used as a six-dose regimen, replacing the commonly used four-dose regimen \[[@pmed-0020092-b15],[@pmed-0020092-b16]\]. Although no trials of the six-dose regimen have been reported in Africa, low gametocyte carriage rates have been observed in efficacy studies using the short treatment course \[[@pmed-0020092-b13],[@pmed-0020092-b14]\]. There have been no studies of the effect of co-artemether treatment on transmission of parasites to the mosquito host. A non-artemisinin combination under consideration in a number of African countries is CQ/SP, largely due to its low cost \[[@pmed-0020092-b17]\]. In The Gambia, where CQ has been front-line treatment policy up until 2004, clinical treatment failures with CQ monotherapy are approaching 20% in in vivo tests of drug efficacy \[[@pmed-0020092-b11]\]. It is hoped to extend the useful life of CQ by combining this drug with SP, which retains clinical efficacy of over 95% in The Gambia \[[@pmed-0020092-b12]\]. However, both of these drugs, when used alone, lead to a high prevalence of gametocyte carriage among treated patients \[[@pmed-0020092-b06],[@pmed-0020092-b10]\]. Therefore, in contrast to artemisinin-based combinations, we are concerned that CQ/SP will not prevent transmission of CQ- and SP-resistant parasites from treated cases, resulting in a rapid deterioration in efficacy. In this paper we present a study of Plasmodium falciparum transmission in Gambian children with uncomplicated malaria who were randomised to receive either six doses of co-artemether or CQ/SP. The major endpoints presented are gametocyte carriage rates, peripheral blood gametocyte density, and infectiousness to Anopheles gambiae mosquitoes of children carrying gametocytes 7 d after treatment. Methods {#s2} ======= The study design comprised a single-blind, open-label, randomised, controlled trial in children with acute uncomplicated falciparum malaria, who were treated with either a combination of CQ and SP or six doses of co-artemether. Major endpoints were post-treatment gametocyte carriage and infectiousness to mosquitoes of children carrying gametocytes 7 d after treatment. The study protocol was approved by both the Joint Gambia Government/Medical Research Council Ethics Committee and the London School of Hygiene and Tropical Medicine Ethics Committee ([Protocol S1](#sd001){ref-type="supplementary-material"}) and is wholly compatible with the revised Helsinki Declaration of 1983 ([Protocol S2](#sd002){ref-type="supplementary-material"}; [Table S1](#st001){ref-type="supplementary-material"}). Study Children and Treatment {#s2a} ---------------------------- Children 1--10 years of age attending Farafenni Health Centre, The Gambia, were screened at our recruiting clinic over a 12-wk period from September to December, 2002. Children with a temperature greater than 37.5 °C, or a history of fever and who were blood-film positive for P. falciparum at a density greater than 500 parasites per μl, and for whom signed informed consent was obtained, were eligible for recruitment into the study. Exclusion criteria were: an inability to take drugs orally, treatment with antimalarial chemotherapy within the past 2 wk, carriage of circulating gametocytes at presentation, any evidence of chronic disease or acute infection other than malaria, domicile outside the study area (approximately 10-km radius), or any signs or symptoms of severe malaria. These latter included severe anaemia (peripheral blood packed cell volume \[PCV\] less than 20%), hyper-parasitaemia (more than 250,000 per μl of peripheral blood), respiratory distress (respiratory rate over 40 with two of the following: nasal flaring, intercostal indrawing, subcostal recession, or grunting), repeated generalized convulsions (three or more per 24 h or two witnessed seizures in 24 h), haemoglobinuria (dark red/black urine), jaundice, prostration, or circulatory collapse. Children were randomized to receive co-artemether or CQ/SP in the ratio 4.5:1 (see below). Children in the CQ/SP group were given three daily oral doses of 10 mg of chloroquine base per kilogram of body weight, plus one half-tablet of SP (12.5 mg of pyrimethamine with 250 mg of sulphadoxine per half-tablet), and an additional one-quarter tablet for each 5 kg of body weight over 10 kg. Children in the co-artemether group received at each dose one half-tablet of co-artemether (20 mg of artemether with 120 mg of lumefantrine) per 5 kg body weight up to 24 kg (two tablets). Children 25 kg or heavier received the adult dose of three tablets. This treatment was given six times, at 0 h (i.e., in the clinic at approximately midday), 8 h (i.e., in the evening of the recruitment day), 20 h, 32 h, 44 h, and 56 h (± 90 min) after the initiation of treatment. This regimen is consistent with the manufacturer\'s recommendations. Absorption of co-artemether is facilitated by fatty food, so the first dose was given to all children in both treatment groups with food (bread with a local oily sauce of beans) under the supervision of staff at the recruitment clinic. Each child was observed for 20 min after the first dose, and any child who vomited was administered a full replacement dose. All subsequent doses of both treatments were administered by a visiting field assistant at the child\'s home, accompanied by food provided by the child\'s family. Each child also received 10 mg/kg paracetamol under supervision at recruitment, and parents or guardians were instructed to administer further doses every 6 h until the child\'s symptoms had subsided. Open-label chloroquine (Alkaloida), pyrimethamine-sulphadoxine (Pharmamed), and paracetamol (Echo International Services) were used. co-artemether was purchased from Novartis (Switzerland). The treatment order was randomised by a Medical Research Council (MRC) statistician not engaged on the trial, such that two CQ/SP treatments occurred in each block of 11 assignments (see below). The treatment order, on slips of paper, was placed in identical envelopes numbered 1 to 550, each corresponding to a study subject number that was assigned to a single child by a study clinician prior to opening of the envelope. All staff engaged in the trial were blinded as to the treatment group of each child, apart from those in the recruiting clinic and the field assistants responsible for supervising administration of medication. Trial clinicians could not influence the choice of drug given. A physician at MRC Farafenni, who was not engaged in the study, fulfilled the role of trial monitor and held an additional copy of the randomised treatment code. A major endpoint of the trial was infectiousness of patients to Anopheles mosquitoes, measured 7 d after treatment, so the study was designed so that approximately equal numbers of gametocyte carriers would be identified in both treatment groups at day 7 of follow-up. The treatment allocation ratio of 4.5:1 was based on an expected post-treatment day 7 gametocyte prevalence of 40%--50% among children receiving CQ/SP, as was observed in the previous year (Sutherland et al., unpublished data), and an estimate of day 7 gametocyte prevalence among co-artemether-treated children of between 5% and 10% \[[@pmed-0020092-b13],[@pmed-0020092-b14]\]. The recruitment of 533 individuals was expected to provide at least 60 children divided equally between the two groups who were gametocytaemic on day 7 and prepared to donate venous blood for membrane-feeding, after gametocyte carriers were excluded at recruitment. These exclusions were necessary because children who present with peripheral gametocytaemia are very likely to carry gametocytes during the post-treatment period and to be infectious to mosquitoes, independent of the treatment regimen used \[[@pmed-0020092-b06],[@pmed-0020092-b10],[@pmed-0020092-b18]\]. This sample size would require dissection of approximately 1,200 mosquitoes. We have previously found that 5.3% of mosquitoes fed on blood donated by children treated with CQ/SP and carrying gametocytes at day 7 became infected, and that 20% of gametocyte carriers were infectious (Sutherland et al., unpublished data). This study was designed to provide a power of 90% at the 95% confidence level to detect a 3.5-fold reduction in the proportion of mosquitoes infected by gametocyte carriers on day 7. Finger-prick peripheral blood from all children attending the recruitment clinic was obtained for thick blood-film examination, PCV determination, and, in the case of recruits, collection of a filter paper sample for subsequent nucleic acid extraction. Parasitological admission criteria were checked at the recruitment clinic using thick blood films stained with rapid Field\'s stain; 25 high-power fields were examined. Duplicate thick blood films from all recruits were returned to the laboratory, allowed to dry for 24 h, and stained with Giemsa. Two reads of 100 high-power fields were performed for asexual and sexual stage parasites. Parasitaemia was calculated as described \[[@pmed-0020092-b19]\]. Transport was provided to bring children to the MRC Field Station at Farafenni 7 d after treatment to be screened for circulating gametocytes as previously described \[[@pmed-0020092-b10]\]. Finger-prick samples for thick-blood film and haematocrit were obtained, and a field assistant fluent in four local languages posed a fixed list of questions to each guardian and/or child to determine if clinical signs or symptoms had resolved and if there had been any of a range of adverse events. Answers were recorded on a specific report form designed for this purpose. Thick-blood films were stained with Field\'s stain, and 100 fields were examined immediately for the presence of gametocytes. Guardians of any gametocyte carriers were asked if approximately 3 ml of venous blood could be taken from the child\'s arm for membrane-feeding of mosquitoes. Day 7 follow-up and gametocyte screening were performed on day 6 in the case of eight individuals for whom day 7 fell on a religious festival. Children who were not available to be brought to the MRC laboratory on the morning of day 7 were followed up at home by a field assistant whenever possible, and the same fixed list of questions posed. However, such children were not able to be venous blood donors for mosquito feeding. Field assistants made further visits to each child on post-treatment days 14 and 28 to obtain finger-prick peripheral blood for thick films and dried filter-paper blood spots for parasite DNA extraction. At each contact with study participants, field staff asked the patient and/or guardians whether any symptoms indicative of clinical malaria had persisted or returned since treatment, and whether the child had experienced any other adverse events. This information was recorded on a follow-up form. Participants were encouraged to attend the recruiting clinic at any time should the child become sick. Estimating Duration of Gametocyte Carriage {#s2b} ------------------------------------------ Gametocyte carriage was estimated by scoring each individual as gametocyte positive or not on each of the follow-up days (0, 7, 14, or 28) \[[@pmed-0020092-b20]\]. Given the reported half-life of gametocytes of 2.5 d \[[@pmed-0020092-b21]\], a single gametocyte-positive slide was assumed to represent gametocytes circulating in the peripheral blood for 2.5 d. When patients had blood films that were gametocyte positive in successive weeks, the estimated time of gametocyte carriage was the interval between observations plus 2.5 d. It was assumed that a single negative or missing slide between two positive slides had a greater than 50% chance of being positive \[[@pmed-0020092-b22]\]. Between-group comparison of mean duration of gametocyte carriage was by the rank-sum test of Wilcoxon. Membrane Feeding and Mosquito Dissection {#s2c} ---------------------------------------- Venous blood samples were obtained from children who were gametocytaemic (limit of detection, 5 gametocytes/μl) on day 7, who had a PCV greater than 20%, and whose parent or guardian gave specific consent for the procedure. If the number of gametocyte-positive patients was greater than the number of cages of mosquitoes available for feeding on a particular day, consenting gametocyte carriers were selected in the order in which they were finger-pricked, which was not dependent on study identification numbers or treatment group. The mosquito feeds were performed as described elsewhere \[[@pmed-0020092-b06],[@pmed-0020092-b10]\]. In brief, venous blood in citrate-phosphate dextrose (anticoagulant) was centrifuged and the plasma removed. After being washed, the red blood cell pellet was resuspended to a PCV of 33% in pooled type AB serum from European donors with no history of malaria exposure. This allowed us to reduce the possible effects of residual drug or of antibody in the blood meal \[[@pmed-0020092-b10]\], although some CQ may be carried over within erythrocytes. Each suspension then was fed to approximately 50 5-d-old female A. gambiae s.s. mosquitoes via an artificial membrane attached to a water-jacketed glass feeder maintained at 37 °C. Mosquitoes were from an established colony maintained by feeding on rabbits. Mosquito mid-guts were dissected 7--8 d later, and the number of malaria oocysts on the mid-gut recorded. Molecular Genotyping {#s2d} -------------------- P. falciparum genotypes circulating in children with detectable asexual parasitaemia during post-treatment follow-up were compared with those present in the same child prior to treatment. DNA was extracted from dried blood spots using a Chelex-based method \[[@pmed-0020092-b08]\]. Alleles of the polymorphic locus pfmsp2 were compared between pre- and post-treatment parasite isolates by PCR \[[@pmed-0020092-b08],[@pmed-0020092-b23]\]. The procedure of Cattamanchi et al. \[[@pmed-0020092-b24]\] was followed, in that indeterminate samples in which a majority of novel bands appeared in the post-treatment infection were scored as new infections. There were only seven such indeterminate infections among the 44 successfully typed. Data Analysis {#s2e} ------------- Data were analysed using Stata, version 8 (StataCorp, Texas). Proportions were compared using the chi-square statistic or Fisher\'s exact test. In the case of mosquito dissection data, the confidence intervals (CIs) of proportions were estimated by bootstrapping to take account of clustering within feed-cages. Normally distributed continuous variables were compared using the t-test of Student. Gametocyte density was compared between groups using the ratio of arithmetic means, as in a previous study \[[@pmed-0020092-b10]\]. This was done by fitting generalized linear models with a logarithmic link function (the "nbreg" command in Stata), i.e., modelling the logarithm of the (arithmetic) mean gametocyte density \[[@pmed-0020092-b25]\]. Ratios of mean densities were obtained by exponentiating the relevant regression coefficients. We favour this approach because the test statistic (mean ratio) is intuitively simple, the negative binomial function gives a good fit to the typically skewed distributions observed, and because all data points (including zeroes) are included. Thus differences in gametocyte prevalence, as well as density, contribute to the mean ratio. Results {#s3} ======= Of 3,167 febrile children, 1,328 were infected with microscopically-confirmed P. falciparum and did not have circulating gametocytes. Eighty-three children (6.3%) harboured both gametocytes and trophozoites, and were not recruited. A total of 497 eligible children were enrolled in the trial and randomised between the treatment groups. This was a smaller number than predicted, because of lower rainfall than in previous seasons, and thus reduced malaria transmission. The reason for ineligibility was recorded for 82% of non-recruits. These were: residence outside the study area (26.1%), parasitaemia below the threshold (16.5%), consent withheld (15.9%), anaemia (PCV below 20%) (7.6%), participation in another research study (4.4%), recent antimalarial use (3.8%), hyperparasitaemia (2.9%), other signs of severe malaria (1.6%), vomiting (0.7%), other disease present (0.6%) and other reasons not included in the above (19.9%). Ninety-one children received CQ/SP and 406 received co-artemether. Baseline characteristics of these 497 enrolments are presented in [Table 1](#pmed-0020092-t001){ref-type="table"}. Two children enrolled and treated in the CQ/SP group, and six in the co-artemether group, were found to have zero parasitaemia upon subsequent examination of Giemsa-stained duplicate slides. Similarly, four (4.4%) of the CQ/SP group and 26 (6.5%) of the co-artemether group were found to have been carrying peripheral gametocytes at presentation that had not been identified in the recruitment clinic. These and all other subjects randomised between treatments were retained in the primary analysis under our intention-to-treat protocol. Eighty-eight (96.7%) of the CQ/SP group, and 379 (93.4%) of the co-artemether group, were visited five times by a field assistant in the 56 h following recruitment, and received the full regimen allotted under supervision. Seventy-two (79.1%) and 347 (85.5%) of the CQ/SP and co-artemether groups, respectively, completed follow-up to day 28, although 12 and 46 of these, respectively, skipped day 7, day 14, or both. A profile of the study is given in [Figure 1](#pmed-0020092-g001){ref-type="fig"}. ::: {#pmed-0020092-g001 .fig} Figure 1 ::: {.caption} ###### Profile of the Study Ten and 38 children in the CQ/SP and co-artemether groups, respectively, who were not seen on day 7 were seen on day 14. Two and four children, respectively, who were not seen on day 7 or day 14 were seen on day 28. Ten children in the co-artemether group not seen on day 14 were seen on day 28. ::: ![](pmed.0020092.g001) ::: ::: {#pmed-0020092-t001 .table-wrap} Table 1 ::: {.caption} ###### Baseline Characteristics of Enrolled Patients ::: ![](pmed.0020092.t001) ^a^ Samples containing at least five per μl of peripheral blood ::: Adverse Advents {#s3a} --------------- No serious adverse events occurred in either treatment group, and there were no deaths among children recruited into the study. Minor complaints noted at the day 7 laboratory visit included headache (12% and 11% in the CQ/SP and co-artemether groups, respectively), anorexia (12% and 16%), diarrhoea (7% and 4%), abdominal pain (5% and 5%), and pruritis (1% and 1%). Gametocyte Prevalence and Density {#s3b} --------------------------------- Children treated with co-artemether were significantly less likely to carry gametocytes during follow-up than children treated with CQ/SP ([Figure 2](#pmed-0020092-g002){ref-type="fig"}; p \< 0.001 at each of days 7, 14, and 28). In the CQ/SP group, 42 of 86 (48.8%) evaluable children had detectable peripheral gametocytaemia at some point over 28 d of post-treatment follow-up, compared to 30 of 378 (7.9%) evaluable children in the co-artemether group (relative risk 6.15; 95% CI, 4.10--9.23; p \< 0.001). The mean duration of gametocyte carriage was also significantly lower in the co-artemether group (0.3 d compared to 4.2 d in the CQ/SP group; range in both groups 2.5--23.5 d; p \< 0.0001). ::: {#pmed-0020092-g002 .fig} Figure 2 ::: {.caption} ###### Gametocyte Carriage after Treatment with CQ/SP or Co-Artemether Point-prevalence of gametocyte-positive thick films in both treatment groups at 7, 14, and 28 d post-treatment (limit of detection: five gametocytes per μl of peripheral blood). Error bars represent 95% CI calculated from the binomial distribution. Denominators for CQ/SP and co-artemether (CoArt) groups are, respectively, 74 and 336 (day 7), 79 and 356 (day 14), and 72 and 355 (day 28). ::: ![](pmed.0020092.g002) ::: The ratio of mean peripheral gametocyte density in all children at each follow-up visit is presented in [Table 2](#pmed-0020092-t002){ref-type="table"}. Six of the eight children in the co-artemether group who were gametocyte-positive at day 28 also harboured trophozoites, suggesting a new infection emerging from the liver, whereas only one of 15 gametocyte carriers in the CQ/SP group also harboured trophozoites at day 28 (odds ratio = 42; 95% CI, 2.4--2,061; Fisher\'s exact p = 0.002). ::: {#pmed-0020092-t002 .table-wrap} Table 2 ::: {.caption} ###### Ratio of Means of Gametocyte Density in Peripheral Blood after Treatment with CQ/SP or Co-Artemether ::: ![](pmed.0020092.t002) The ratio of arithmetic mean of gametocyte numbers per μl of peripheral blood from both treatment groups at 7, 14, and 28 d post-treatment. Denominators are as in Figure [2](#pmed-0020092-g002){ref-type="fig"} ::: Infectiousness of Treated Children {#s3c} ---------------------------------- 399 children were seen at the MRC laboratories on day 7, 73 in the CQ/SP group and 326 in the co-artemether group. There were 27 children (37.0%) in the CQ/SP group and 16 children (4.9%) in the co-artemether group identified as gametocyte carriers by our rapid screening procedure (p \< 0.001). Subsequent definitive microscopy with the Giemsa-stained duplicate thick films gave good concordance with the screening result (positive predictive value 90.2%; specificity 98.8%). Explicit consent for venous blood donation for membrane-feeding of mosquitoes was obtained from 29 (67.4%) of these 43 children: 19 (67.9%) of the gametocyte positives in the CQ/SP group and 10 (62.5%) of those in the co-artemether group. Mean ages among these 29 children were 3.3 years (95% CI, 2.4--4.3) and 4.2 years (2.7--5.7) in the two treatment groups, respectively. Results of the transmission experiments are given in [Table 3](#pmed-0020092-t003){ref-type="table"}. Whereas seven children (36.9%) among the CQ/SP gametocyte donors were infectious to mosquitoes, none of the gametocyte donors in the co-artemether group was infectious. The infectious proportion observed in the CQ/SP group was higher than expected from an earlier study in which 20% of CQ/SP treated children were infectious, perhaps due to the use here of an established blood-fed mosquito colony, which leads to better insect survival. In the current study, we dissected an average of 18.6 mosquitoes (95% CI, 16.8--20.4) per infection experiment, compared to 14.8 mosquitoes (95% CI, 12.5--17.1; two-tailed p = 0.021) in a similar study performed in 2001 that used F1 progeny of wild-caught female Anopheles. The data in [Table 3](#pmed-0020092-t003){ref-type="table"} show that gametocyte donors in the co-artemether group carried fewer gametocytes, and were as a result significantly less infectious than those in the CQ/SP group, consistent with previous studies \[[@pmed-0020092-b06],[@pmed-0020092-b10]\] ::: {#pmed-0020092-t003 .table-wrap} Table 3 ::: {.caption} ###### Profile of 29 Children Who Donated Gametocyte-Positive Blood at Day 7 and Results of Transmission Experiments ::: ![](pmed.0020092.t003) ^a^ Student\'s t-test, two-sided ^b^ Ratio of arithmetic means (95% CI); significance tested by negative binomial regression ^c^ Fisher\'s exact test of equal proportions, one-sided ^d^ Bootstrap (50,000 replicates) ^e^ Negative binomial regression is not informative with one mean at zero ::: Clinical and Parasitological Efficacy of CQ/SP and Co-Artemether {#s3d} ---------------------------------------------------------------- Among eligible children both regimens employed were highly efficacious against acute uncomplicated falciparum malaria. Clinical treatment failure, defined as the persistence or recurrence of malaria symptoms with detectable peripheral trophozoites, occurred in 8 cases (1.6%), all of which were in the co-artemether treatment group. One of these patients received only a single observed dose of co-artemether at the recruiting clinic, the other seven cases received the full six-dose regimen under supervision. All eight clinical failures occurred 2 wk or more after treatment. The cumulative 28-d clinical failure rate among children under 5 years of age was 2.55%, and among those 5 years and over 0.45%; (relative risk 5.65; 95% CI, 0.70--45.6; one-sided Fisher\'s exact p = 0.064). Parasitological treatment failure, defined as carriage of asexual parasites (trophozoites) during post-treatment follow-up, occurred in fewer than 5% of patients in both treatment groups up to day 14, but rose to approximately 12% in both groups by day 28 ([Figure 3](#pmed-0020092-g003){ref-type="fig"}). ::: {#pmed-0020092-g003 .fig} Figure 3 ::: {.caption} ###### Crude Parasitological Failure after Treatment with CQ/SP or Co-Artemether Point-prevalence of trophozoite-positive thick films in both treatment groups at 7, 14, and 28 d post-treatment (limit of detection: five trophozoites per μl of peripheral blood). Error bars represent 95% CI calculated from the binomial distribution. Denominators for CQ/SP and co-artemether (CoArt) groups are, respectively, 74 and 336 (day 7), 79 and 356 (day 14), and 72 and 355 (day 28). ::: ![](pmed.0020092.g003) ::: Interpretable PCR data were obtained for 44 of the 60 post-treatment asexual parasite infections identified by microscopy. Parasites harboured at the time of treatment failure were indistinguishable from pre-treatment parasites by msp2 PCR genotyping in 15 of the 44 cases for which the data were interpretable. Post-treatment parasites in five of the nine children in the CQ/SP group were caused by recrudescent infections compared to ten of 35 in the co-artemether group. After correction of the 28-d cumulative parasitological failure rate using these ratios, it was estimated that co-artemether was 96.1% efficacious over 28 d, resulting in a total of 14 recrudescent infections, compared to 91.1% for CQ/SP, resulting in seven recrudescent infections (odds ratio = 0.413; 95% CI, 0.150--1.26; p = 0.058). Discussion {#s4} ========== Co-artemether, administered to children with uncomplicated falciparum malaria as a six-dose regimen, is highly effective at reducing the prevalence and duration of gametocyte carriage, the density of peripheral gametocytaemia among carriers, and the infectiousness to mosquitoes of carriers at day 7 compared to a combination of CQ and SP. In the co-artemether group, none of the ten children who carried gametocytes at day 7 and donated blood for membrane-feeding of mosquitoes was found to be infectious. In contrast, 7 (36.9%) of the 19 gametocyte donors in the CQ/SP group were infectious, largely due to a significantly higher gametocyte density than in the co-artemether group. This suggests that co-artemether has a specific action against developing gametocytes, as any gametocytes that appeared on or before day 7 must have been committed to sexual differentiation at the time of treatment, because of the period of 8--10 d of sequestered sexual development that occurs before gametocytes are detectable in the periphery \[[@pmed-0020092-b26]\]. This anti-gametocyte activity is additional to the anti-trophozoite and anti-schizont effects of co-artemether, which contribute to the very low gametocyte carriage rates observed later in follow-up. Importantly, we also found evidence that post-treatment carriage of both trophozoites and gametocytes in the co-artemether group, but not the CQ/SP group, was due mainly to new emergent infections, rather than recrudescent infections, which suggests that artemether-lumefantrine reduces transmission of recrudescent, resistant parasites compared to CQ/SP. Genotyping of these infections at resistance-associated loci is underway to test this further. The potential of artemisinin-containing combination therapies to reduce gametocyte carriage rates, and therefore transmission of falciparum malaria, was first shown by Price et al. \[[@pmed-0020092-b05]\]. This study, among refugee camps on the Thailand-Burma border, showed that deploying the combination of mefloquine with artesunate significantly reduced the incidence of malaria cases over a few years in this low-transmission area where asymptomatic infections are rare and thus drug pressure is very high. We have extended this work to a setting in sub-Saharan Africa where transmission is high and seasonal, asymptomatic infections are common, and as few as 20% of slide-positive infections are treated \[[@pmed-0020092-b27]\]. Transmission reduction at the population level through the treatment of clinical cases with artemisinin-containing combination therapy is unlikely in this setting, and so our strategy has been to use membrane-feeding of Anopheles gambiae mosquitoes to measure the infectiousness of individual treated children who become gametocyte carriers after treatment \[[@pmed-0020092-b06],[@pmed-0020092-b10]\]. Those children who are carrying gametocytes prior to treatment are likely to become infectious regardless of the treatment received. We can estimate the proportion of treated children estimated to be infectious by multiplying the proportion of gametocyte donors that were infectious by the prevalence of gametocyte carriage at day 7. Using the data in [Table 3](#pmed-0020092-t003){ref-type="table"}, % (95% CI, 0--1.9% ) of children treated with six doses of co-artemether will be infectious to mosquitoes at day 7. For CQ/SP, the corresponding estimate is 15.8% (5.6%--26.0%). We have previously estimated that the mean infectious proportion in control serum membrane-feeding experiments for SP plus artesunate (three doses) in 1999 was 4%, and for CQ plus artesunate (three doses) was also 4% \[[@pmed-0020092-b06],[@pmed-0020092-b10]\]. Six doses of co-artemether produced the lowest gametocyte carriage rates we have seen in 5 years of consecutive transmission studies in Farafenni \[[@pmed-0020092-b06],[@pmed-0020092-b10]\], excluding patients harbouring gametocytes at day 0. Although 2002 had lower rainfall than any of the previous 10 years, there were nevertheless significant numbers of gametocyte carriers, many of them infectious, in the CQ/SP group in this study. From a public health point of view, the effect of each regimen on all children presenting for treatment, including gametocyte carriers, is of interest. In co-artemether-treated children who had unidentified circulating gametocytes at recruitment, seven of 20 (29.2%) carried gametocytes at 7 d, compared to 13 of 312 (4.2%) of those without circulating gametocytes at the time of treatment (Fisher\'s exact test, p \< 0.01). This demonstrates that children presenting with gametocytaemia at day 0 make a disproportionate contribution to day 7 gametocytaemia. Inclusion, hypothetically, of the 83 non-recruited gametocyte carriers in addition to the 30 that were, predicts that only 8% of children in the co-artemether group would have harboured day 7 gametocytes, compared to 45% in the CQ/SP group. Co-artemether is thus likely to be highly effective at reducing post-treatment gametocyte carriage among children irrespective of gametocytaemia at presentation. At day 28, gametocyte carriage in the co-artemether group was associated with evidence of a new infection recently emerged from the liver. This was not the case with CQ/SP. Therefore, at day 28 a significant proportion of children treated with CQ/SP will harbour gametocytes that originated from the treated clinical infection. These gametocytes and their asexual progenitors have been under sustained drug selection, given the elimination half-lives of CQ, des-ethylchloroquine (an active metabolite of CQ), sulphadoxine and pyrimethamine: 5.9 d, 8.5 d, 10.9 d, and 2.9 d, respectively when given simultaneously as CQ/SP \[[@pmed-0020092-b28]\]. CQ continues to be gradually released back into the circulation from host tissues for at least 30 d \[[@pmed-0020092-b29]\]. In contrast, artemether and lumefantrine are cleared from the body with terminal elimination half-lives of 0.8 h and 4.5 d, respectively, after administration of co-artemether \[[@pmed-0020092-b30]\]. Thus, parasites detected in co-artemether-treated children during follow-up have, on average, experienced a shorter period of selective drug pressure than those in CQ/SP-treated children. Nevertheless, lumefantrine is "unprotected" for several days following the elimination of artemether and its metabolites, and so we cannot rule out the emergence of resistance to lumefantrine under widespread use of co-artemether. We found that co-artemether is better than CQ/SP at preventing recrudescence of treated parasites. However, as seen with artemisinins in other areas of moderate to high malaria transmission, co-artemether has the disadvantage of permitting rapid re-infection, which can lead to new clinical attacks. This is due to absence of the prophylactic effect afforded by a pharmacological "tail," such as that seen with both SP and CQ. From our co-artemether data, the 28 d new infection rate in our study population is approximately 10% with fewer than a fifth of these experiencing a new clinical attack. Although this reinfection/clinical infection rate may be even higher in other settings, we believe it would be offset by the ability of co-artemether to reduce selection and transmission of drug resistant parasites, and thus make possible sustainable improvements in both therapeutic and public health outcomes. In contrast, as our data show, the pharmacological persistence of CQ and SP eliminates sensitive parasites emerging from the liver, and so permits a much smaller proportion of new infections (3.5%). We predict that CQ/SP thus selects for any resistant mutants present in the starting population, leading to increased onward transmission of resistant parasites, and reduced onward transmission of sensitive parasites. There are difficulties with the introduction of co-artemether as a first-line treatment for malaria in Africa. Firstly, as the six-dose regimen is the most likely to provide the treatment and transmission-reducing benefits needed, there may be problems with treatment compliance. Secondly, the requirement of oily food for adequate absorption may lead to inadequate drug levels in the blood of many treated individuals. Effectiveness studies that examine these indicators under conditions as close as possible to the normal clinical setting, and with both therapeutic and transmission endpoints, are thus of great importance. Nevertheless, the therapeutic success of co-artemether as a front-line treatment policy has been demonstrated in KwaZulu--Natal, South Africa, where it replaced SP \[[@pmed-0020092-b31]\]. The impact of this policy on post-treatment transmission in this region is yet to be determined. We have shown that, compared to CQ/SP, co-artemether significantly reduces post-treatment gametocyte carriage, peripheral gametocyte density and infectiousness of gametocyte carriers. There is little recrudescence of parasites under co-artemether, and the majority of gametocyte carriers are associated with new infections that will have been subject to little or no drug pressure. We are not suggesting that introduction of co-artemether as first-line treatment policy in The Gambia or elsewhere is likely to reduce overall transmission of P. falciparum within the community, but rather that selective transmission of resistant parasites in treated patients would be greatly reduced. The benefit is because drug pressure is exerted only on these treated infections. Therefore, the use of co-artemether is likely to have measurable public health benefits in reducing the impact of drug resistance. We predict that these benefits would prove far more robust than those transiently enjoyed after the initial deployment of SP in Africa. Supporting Information {#s5} ====================== Trial Registration {#s5a} ------------------ The Wellcome Trust Project number is 061910. An International Standard Randomised Controlled Trial Number (ISRCTN) is being applied for. Protocol S1 ::: {.caption} ###### Ethics Permission (283 KB DOC). ::: ::: {.caption} ###### Click here for additional data file. ::: Protocol S2 ::: {.caption} ###### Study Protocol (99 KB DOC). ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### Consort Table (55 KB DOC). ::: ::: {.caption} ###### Click here for additional data file. ::: ### Patient Summary {#sb1} #### {#sb1a} ##### Background {#sb1a1} Treatment studies of malaria often focus on the whether parasites have become resistant to anti-malarial treatment. However, an ideal treatment would be one that not only cured the patient but also prevented further transmission of the parasite, especially those that carry resistance, to mosquitoes and hence to other individuals. ##### What Did the Researchers Do? {#sb1a2} They randomised 497 children with uncomplicated falciparum malaria (the most serious kind) in Gambia to either a combination of two drugs (chloroquine and sulphadoxine-pyrimethamine \[CQ/SP\]) to which many malarial parasites have become resistant, or a new drug, artemether, in fixed combination with lumefantrine (co-artemether, brand names Coartem and Riamet). Within the 4 weeks following treatment they found that patients treated with co-artemether were much less likely to contain the stage of the malarial parasite that is infectious, that they carried the parasites for a shorter time, and that blood from the children treated with co-artemether was less able to infect mosquitoes. ##### What Do These Findings Mean? {#sb1a3} The new combination substantially reduces the chances of malaria being transmitted from infected patients, and this finding should be taken into account when deciding which drugs are the preferred choice for treatment. ##### Where Can I Get More Information? {#sb1a4} The Wellcome Trust has information on malaria: <http://www.wellcome.ac.uk/en/malaria/> The World Health Organisation has a programme called Roll Back Malaria, which has various initiatives: <http://www.who.int/topics/malaria/en/> The Gates foundation funds research on malaria, one arm of which is the Gates Malaria Partnership: <http://www.lshtm.ac.uk/gmp/> This work was supported by Wellcome Trust Project \#061910 awarded to CJS, the Gates Malaria Partnership, and the Medical Research Council, United Kingdom. CJD is supported by a Wellcome Trust Research Fellowship (\#063516). None of these funding sources played any role in study design, or in collection, analysis, and interpretation of data. We thank Angela Hunt-Cooke for over-seeing the microscopy work of Kebba Konteh, Simon Corea, Momodou Jallow, and others. We are grateful to Vasee Moorthy for acting as trial monitor and to Paul Milligan for providing the randomisation. Yorro Bah and the field-work team provided excellent follow-up of patients. We are grateful to staff of Farafenni Maternal and Child Health clinic, and to staff and management of Farafenni hospital for clinical support. We also acknowledge Folosade Olodude and Kalifa Bojang for cooperation in the recruitment clinic and for helpful discussions. Brian Greenwood kindly provided detailed comments on the manuscript. This study would have been impossible without the continuing good will and participation of the people of Farafenni and surrounding villages. Citation: Sutherland CJ, Ord R, Dunyo S, Jawara M, Drakeley CJ, et al. (2005) Reduction of malaria transmission to Anopheles mosquitoes with a six-dose regimen of co-artemether. PLoS Med 2(4): e92. AQ : amodiaquine AS : artesunate CQ : chloroquine MRC : Medical Research Council PCV : packed cell volume SP : sulphadoxine-pyrimethamine [^1]: Competing Interests: The authors have declared that no competing interests exist. [^2]: Author Contributions: CJS, SD, CJD, NA, RC, MP, GW, and GATT designed the study. CJS, RO, SD, and MJ performed the experiments. CJS, MJ, CJD, NA, MP, and GATT analyzed the data. CJS, RO, MJ, RC, MP, and GW enrolled patients. CJS, CJD, NA, RC, MP, GW, and GATT contributed to writing the paper.	PubMed Central	2024-06-05T03:55:55.887157	2005-4-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1087200/", "journal": "PLoS Med. 2005 Apr 26; 2(4):e92", "authors": [ { "first": "Colin J", "last": "Sutherland" }, { "first": "Rosalynn", "last": "Ord" }, { "first": "Sam", "last": "Dunyo" }, { "first": "Musa", "last": "Jawara" }, { "first": "Christopher J", "last": "Drakeley" }, { "first": "Neal", "last": "Alexander" }, { "first": "Rosalind", "last": "Coleman" }, { "first": "Margaret", "last": "Pinder" }, { "first": "Gijs", "last": "Walraven" }, { "first": "Geoffrey A. T", "last": "Targett" } ] }
PMC1087201	One of the most fundamental human rights is the assumption that each person matters, and everyone deserves to be treated with dignity---this is the tenet from which all other human rights flow. Another is that those who are most vulnerable deserve special protection. However, in many developing countries, vast numbers of children are born but never counted, and their health and welfare throughout their lives remains unknown. And because single-mean measures of population health mask inequalities among the best-off and worst-off, the health of vulnerable populations is not effectively documented and acknowledged. Health information systems can play an important role in supporting these rights by documenting and tracking health and health inequities, and by creating a platform for action and accountability. A human rights approach to health information systems also supports effective health development. To effectively improve population health, governments and communities need access to socioeconomically disaggregated population health data. Because the relationship between such information and human rights has received little attention, the two areas of health information systems and human rights have done little to support each other. The Health Metrics Network {#s2} ========================== The Health Metrics Network (HMN) is a global collaboration focused on strengthening country health information systems to generate sound data for decision-making at country and global levels (see <http://www.who.int/healthmetrics/en/>). Its interim secretariat is based at the World Health Organisation. The development group of HMN recently considered strategies for strengthening health information systems within countries. The Equity Working Group of HMN made recommendations that outlined the content of equity-sensitive information systems, identified opportunities for minimizing collection burdens, and suggested strategies to foster an equity-oriented decision-making culture. Although the recommendations were implicitly focused on human rights and on improving opportunity for the worst-off populations, making that framework explicit helps us to acknowledge and clarify the values on which decisions for developing information systems are made. The framework is based on several principles rooted in human rights and their implied actions (see Sidebar). Principles to Guide a Human Rights-Oriented Framework for Integrating Equity into Health Information Systems ------------------------------------------------------------------------------------------------------------ Each person has dignity and each matters---Count everyone in society from birth to death*Everyone should have opportunities for health and the means to improve health; vulnerable populations deserve special attention---Collect and analyze information related to inequalities in health status and determinants of health among various better-off and worse-off subpopulationsGovernments are accountable to the public, communities have a right to the information they need to make healthy decisions, and individual autonomy should be supported---Release information to the public in a meaningful formGovernments, communities, and individuals are all responsible for promoting health and health opportunities---Support capacity for and cultures of human rights-oriented decision-making based on health information Every Individual Matters {#s3} ======================== Despite the acknowledged importance of counting everyone, only 57 of the 192 WHO member states, almost all of which are developed countries, have vital registration systems that report on births and deaths for at least 90% of the population. Consequently, a primary recommendation from the Equity Working Group is that health information systems should support the most basic acknowledgment of human rights---one\'s existence---by counting births and deaths in every country through a vital registration system ([Figure 1](#pmed-0020102-g001){ref-type="fig"}). ::: {#pmed-0020102-g001 .fig} Figure 1 ::: {.caption} ###### Every Birth Must Be Registered This photo comes from the development organization Plan ([www.plan-international.org](www.plan-international.org)), which has launched an international campaign for universal birth registration ([www.writemedown.com](www.writemedown.com)). The campaign\'s Web site states that a third of babies born each year are not registered (this statistic comes from Unicef), and that many of these unregistered children have no legal right to health care, education, or the state\'s protection. (Photo: Adam Hinton/Plan) ::: ![](pmed.0020102.g001) ::: Opportunities and the Means to Health {#s4} ===================================== A person\'s human rights are recognized to include "a standard of living adequate for the health and well-being of himself and of his family, including food, clothing, housing, and medical care and necessary social services, and the right to security in the event of...lack of livelihood in circumstances beyond his control" \[[@pmed-0020102-b1]\]. In other words, the means necessary to achieve health are part and parcel of a right to health. However, many country studies have repeatedly shown that there are large gaps between the opportunities of advantaged and disadvantaged sub-populations. For example, in South Africa, average household expenditure for whites was five times the rate for blacks in 1995, and female children in Bangladesh were less likely to be brought to clinic than their male counterparts \[[@pmed-0020102-b2]\]. Examples demonstrating inequalities between various economic, ethnic, gender-based, and other social groups are replicated time and time again in all countries \[[@pmed-0020102-b2],[@pmed-0020102-b3]\]. Of 192 countries, 39 have a health information system sufficient to support basic analysis of socioeconomic inequalities and health. Such a health information system would include the vital registration system coupled with a major household survey. Ninety countries have only a census, an old household survey, or no data at all, whereas the remaining countries fall into a "middling" category (a census and a recent household survey, or two household surveys) (L. Bambas, P. Braveman, N. Dachs, I. Delgado, E. Gakidou, et al., unpublished data). The two issues of health inequalities and human rights overlap in a number of ways that enrich a perspective on health as a human right. A human rights framework can inform a minimally acceptable level of data collection in health information systems---that is, a core set of equity indicators---as well as the conditions surrounding the release and use of that data. In contrast to individual measures of health inequalities, populations as units of analysis are especially useful for examining achievement of human rights (as well as identifying health inequities) since they indicate patterns of differential opportunity in society among various social strata. A human rights-oriented indicator of health inequalities, then, would be composed of two measures: (1) A health measure, including health status, health care, and other determinants of health or the social/ economic consequences of ill health; and (2) A measure of social position/advantage (also called an equity stratifier or social stratifier) that defines strata in a hierarchy, e.g., income or economic assets, education, sex, or ethnic group. Health differences across social strata in comparison with the most advantaged group or an absolute standard suggest inequities in health \[[@pmed-0020102-b4],[@pmed-0020102-b5]\]. This interpretation of health inequities directly coincides with a human rights framework since it focuses on a broad concept of health as well as on means to and opportunities for health. Critical issues of methodology for using these indicators also need attention, including identification of appropriate measures of social position and how to meaningfully compare indicators across countries. Consequently, the Equity Working Group recommended that health information systems should include the information necessary to create equity indicators, and that research into methodologies should be given due attention. On a practical level, equity strata and health indicators can be integrated into a number of existing data sources, including censuses, vital registries, household surveys, small areas data, and administrative data sources. Such empirical information not only clarifies distributions of health and achievement of rights but can also identify barriers to health and provide insights on multi-sectoral approaches in planning and interventions to support the most vulnerable populations. Accountability and Autonomy {#s5} =========================== When equity-sensitive information is collected, access to information is often restricted to the government and is rarely disaggregated to show differences between socioeconomic groups. But a human rights approach to health information implies not only particular content but also mechanisms to promote the effective use of information, including the public release of data in a useful form. Confidentiality and privacy issues arise in relation to information disaggregated by equity stratifiers, especially in the context of the public release of such data. There are strong arguments for the public release of health equity information, including the fact that such information is a determinant of health, and that civil society can play a vital role in improving health opportunities, both directly and by influencing governmental priorities. The public\'s general lack of knowledge regarding patterns of health inequalities and their causes within societies further supports the need for the public release of such information. If information is disaggregated for very small populations, such as within a village, particular individuals or households may be identifiable and feel their privacy rights were impinged upon. Therefore, the public release of equity-oriented information on health should be explicitly planned for in the development of health information systems. Principles to guide release of disaggregated data should be followed, and communities should have a voice in the decision to release highly disaggregated information when privacy rights might be compromised. The HMN Equity Working Group suggests development of international standards for collection and sharing of disaggregated data and its use. Given the potential conflict between the two interests, we should continue to investigate possible technologic and/or policy solutions. Mutual Responsibility {#s6} ===================== Another mechanism for promoting the effective use of information is to support cultures of equity-oriented decision-making. In addition to the public release of information, strategies should include supporting research on pathways of health inequities and interventions; building capacity for analyzing information and developing interventions; encouraging demand for equity-sensitive data in government and the public; and supporting broad participation in the promotion of health equity. Standards for improvements and target dates for achieving a minimally acceptable information system should take into account differences in the resources and needs of those implementing changes. The Equity Working Group recommended that target dates be developed with countries to define and integrate core equity indicators into routine information sources. All countries should be able to achieve, within the next 5--15 years, at least the middle-level information system, which could be considered a minimum standard. However, the rationale for this standard is predicated on the assumption of significant financial and technical support being given to the effort, and on the least well-off countries receiving the most support (L. Bambas, P. Braveman, N. Dachs, I. Delgado, E. Gakidou, et al., unpublished data). Conclusion {#s7} ========== These recommendations provide a strategy for strengthening decision- and policy-making by providing a stronger empirical base for human rights considerations. This equity-oriented empirical base could strengthen health rights, not only through health sector decision-making, but also through decision-making in sectors related to determinants. The initial development stage of the HMN has now ended, and hopefully the refinement and implementation stages of the effort will begin within the next year. Regardless of the existence of a centralized effort, countries would greatly promote health rights by integrating equity issues into health information systems, releasing that information publicly, and supporting participation and decision-making attentive to the concerns of health equity and human rights. Citation:* Bambas L (2005) Integrating equity into health information systems: A human rights approach to health and information. PLoS Med 2(4): e102. This is the second of two articles commissioned to coincide with the April 2005 conference at Emory University "Lessons Learned from Rights Based Approaches to Health" (<http://humanrights.emory.edu>). HMN : Health Metrics Network [^1]: Lexi Bambas is the former Coordinator of the Global Equity Gauge Alliance, and currently works as a consultant from San Antonio, Texas, United States. E-mail: <lexibambas@hotmail.com> [^2]: Competing Interests: The author declares that she has no competing interests.	PubMed Central	2024-06-05T03:55:55.890984	2005-4-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1087201/", "journal": "PLoS Med. 2005 Apr 26; 2(4):e102", "authors": [ { "first": "Lexi", "last": "Bambas" } ] }
PMC1087202	Buruli ulcer is a disease of skin and soft tissue with the potential to leave sufferers scarred and disabled. It is caused by an environmental pathogen, Mycobacterium ulcerans, that produces a destructive toxin. The exact mode of transmission is unclear. The main burden of disease falls on children living in sub-Saharan Africa, but healthy people of all ages, races, and socioeconomic classes are susceptible. History and Epidemiology {#s2} ======================== M. ulcerans is the third most common mycobacterial pathogen of humans, after M. tuberculosis and M. leprae (which cause tuberculosis and leprosy, respectively). The definitive description of M. ulcerans was published in 1948, when MacCallum and others in Australia reported six cases of an unusual skin infection caused by a mycobacterium that could only be cultured when the incubation temperature was set lower than for M. tuberculosis \[[@pmed-0020108-b1]\]. In Africa, large ulcers almost certainly caused by M. ulcerans had been described by Sir Albert Cook in 1897 and by Kleinschmidt in northeast Congo during the 1920s \[[@pmed-0020108-b2]\]. > The main burden of disease falls on children living in sub-Saharan Africa. Prior to the 1980s, foci of M. ulcerans infection were reported in several countries in sub-Saharan Africa including Congo \[[@pmed-0020108-b3]\], Uganda \[[@pmed-0020108-b4]\], Gabon, Nigeria \[[@pmed-0020108-b5]\], Cameroon, and Ghana \[[@pmed-0020108-b6]\]. The Uganda Buruli Group coined the name "Buruli ulcer" because the cases they described were first detected in Buruli county, near lake Kyoga \[[@pmed-0020108-b7]\]. Since 1980, dramatic increases in the incidence of Buruli ulcer have been reported from the West African countries of Benin \[[@pmed-0020108-b8]\], Côte d\'Ivoire \[[@pmed-0020108-b9]\], and Ghana \[[@pmed-0020108-b10]\]. New foci were also discovered recently in Togo \[[@pmed-0020108-b11]\] and Angola \[[@pmed-0020108-b12]\]. A characteristic of Buruli ulcer is its focal distribution even within endemic regions, and obtaining accurate disease burden estimates is difficult. However, in some highly endemic districts in Ghana, point prevalence has been estimated to be as high as 150.8/100,000 individuals \[[@pmed-0020108-b10]\], and in southern Benin, a recent study has reported detection rates of 21.5/100,000 per year, higher than for either tuberculosis or leprosy \[[@pmed-0020108-b8]\]. In West Africa, about 25% of people affected by the disease, mostly children, are left with permanent disabilities. The disease is also endemic in several other countries outside Africa, including rural areas of Papua New Guinea, Malaysia, French Guiana, and Mexico ([Figure 1](#pmed-0020108-g001){ref-type="fig"}). In Australia, the disease remains uncommon, but there have been increases in both incidence and the number of endemic areas in the last 15 years \[[@pmed-0020108-b13],[@pmed-0020108-b14]\]. ::: {#pmed-0020108-g001 .fig} Figure 1 ::: {.caption} ###### Countries Reporting Buruli Ulcer The boundaries and names shown and the designations used on this map do not imply the expression of any opinion whatsoever on the part of the World Health Organization concerning the legal status of any country, territory, city, or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. Dotted lines on maps represent approximate border lines for which there may not yet be full agreement. Data Source: WHO/Global Buruli Ulcer Initiative Map Production: Public Health Mapping & GIS Communicable Diseases (CDS) World Health Organization. (Photo: World Health Organization) ::: ![](pmed.0020108.g001) ::: Causative Organism and Pathology {#s3} ================================ Mycobacterium ulcerans is a slow-growing environmental mycobacterium that can be cultured from human lesions on mycobacterial medium at 30--32 °C \[[@pmed-0020108-b15]\]. Histological specimens typically show large clumps of extracellular acid-fast organisms surrounded by areas of necrosis and a poor or absent inflammatory response \[[@pmed-0020108-b16]\]. Subcutaneous fat is particularly affected, but underlying bone may also become involved in advanced cases \[[@pmed-0020108-b15]\]. The pathogenesis and histological appearance is explained by a recently identified diffusible lipid toxin, mycolactone \[[@pmed-0020108-b17]\]. Later in the natural history of the disease, the immunosuppressive effect of the toxin is somehow overcome by the host, immunity develops, and healing commences. Clinical Features {#s4} ================= The classic lesion is a necrotic skin ulcer with deeply undermined edges ([Figure 2](#pmed-0020108-g002){ref-type="fig"}). Any part of the body can be affected, but most lesions occur on limbs. The ulcers are slowly progressive and usually painless, and the patient is usually systemically well, which may explain why sufferers often delay seeking medical assistance. Early Buruli lesions may initially appear as a mobile subcutaneous nodule, a papule, or a raised plaque. ::: {#pmed-0020108-g002 .fig} Figure 2 ::: {.caption} ###### M. ulcerans Infection of the Shin of an 11-Year-Old Boy, Coastal Victoria, Australia A closer view, revealing deep undermining, is shown in the second panel. (Photo: Paul D. R. Johnson) ::: ![](pmed.0020108.g002) ::: A subgroup of patients present with rapidly progressive oedema of a whole limb, abdominal wall, or side of the face without an obvious focal lesion. Part or all of the affected area will subsequently ulcerate, although anecdotal reports suggest that timely antibiotic therapy may greatly reduce the resulting necrosis \[[@pmed-0020108-b18]\]. Treatment {#s5} ========= The aim of treatment is to halt the infection and repair existing damage. Large ulcers are usually treated surgically to remove necrotic tissue and to graft the resulting defect. Relapse after surgery may occur in 18%--47% of cases \[[@pmed-0020108-b19]\], so surgeons commonly ensure wide excision margins in the hope of curing the infection. Traditionally, drug therapy has been considered ineffective, but recent data suggest that combinations of anti-mycobacterial antibiotics that include rifampicin and either streptomycin or amikacin are able to kill M. ulcerans in human lesions \[[@pmed-0020108-b14],[@pmed-0020108-b20]\]. Provisional guidelines now recommend the use of selected anti-mycobacterial drugs, usually combined with surgery, for the treatment of Buruli ulcer \[[@pmed-0020108-b21]\]. Prevention of Disabilities {#s6} ========================== Untreated Buruli ulcer will eventually subside with the gradual development of host immunity in most cases. However, by this time, tissue damage may be very extensive and healing by scar can lead to permanent functional and cosmetic deformity ([Figure 3](#pmed-0020108-g003){ref-type="fig"}). Successful treatment will shorten the course of the disease and minimise deformity. Skilled surgery, expert post-operative nursing care, and restorative physiotherapy are often required to achieve good outcomes. The cost of this may be beyond the means of local rural health services. Even in Australia, where there is universal access to health care, the cost and complexity of treating M. ulcerans infections can be considerable \[[@pmed-0020108-b22]\]. ::: {#pmed-0020108-g003 .fig} Figure 3 ::: {.caption} ###### Long-Term Sequelae of M. ulcerans Infection, Benin, West Africa (Photo: Kingsley Asiedu) ::: ![](pmed.0020108.g003) ::: Public Health Efforts {#s7} ===================== Case control studies have suggested that farming activities close to rivers in endemic areas are a risk factor for Buruli ulcer \[[@pmed-0020108-b9]\], but for farmers involved in subsistence agriculture, avoidance of riverine areas is difficult. A recent study from Ghana has suggested that swimming in rivers may also be an independent risk factor \[[@pmed-0020108-b23]\]. To date, the main focus of public health efforts has been on early detection and treatment, which greatly reduce morbidity and cost \[[@pmed-0020108-b8],[@pmed-0020108-b24]\]. Why Has Buruli Ulcer Been Neglected until Now? {#s8} ============================================== Despite its long history, Buruli ulcer has gone largely unnoticed until recently. Buruli ulcer typically occurs in poor rural communities with little economic or political influence. Rural isolation may mean that national surveillance systems do not immediately detect the appearance of new outbreaks. Affected populations may believe that there is no effective medical treatment for the disease, which discourages them from seeking assistance \[[@pmed-0020108-b25],[@pmed-0020108-b26]\]. In the developed world, Buruli ulcer is frequently omitted from standard medical texts and undergraduate medical courses. The absence of a potentially profitable market has meant that there has been little private investment to date in drug and vaccine development or in research to improve prospects for better control. Future Directions and the End of Obscurity {#s9} ========================================== Global Buruli Ulcer Initiative. {#s9a} ------------------------------- In December 1997, Hiroshi Nakajima, then Director-General of the World Health Organization (WHO), announced that WHO would take the lead to mobilise the world\'s expertise and resources to fight the emergence of Buruli ulcer as a serious public health problem. In 1998, WHO launched the Global Buruli Ulcer Initiative to coordinate control and research efforts, and organised the first International Conference on Buruli ulcer control and research in Yamoussoukro, Côte d\'Ivoire. The resulting "Yamoussoukro Declaration on Buruli Ulcer" drew attention to the severity of the disease as an emerging public health problem and expressed concern about its many poorly understood features. In May 2004, the World Health Assembly adopted a resolution on Buruli ulcer that called for increasing surveillance and control, and for intensified research to develop tools to diagnose, treat, and prevent the disease \[[@pmed-0020108-b27]\]. The attention of the affected countries, donor agencies, nongovernmental organisations, and the research community has been captured by these and other initiatives, and we are now entering an exciting period of rapidly expanding knowledge and interest in the disease. These developments will ensure that Buruli ulcer is not neglected again. Recent research on transmission. {#s9b} -------------------------------- For over 50 years we have known that proximity to marshes and wetlands, often created as a result of some human environmental disturbance, is a risk factor for infection \[[@pmed-0020108-b28]\], but the exact mode of transmission remains an enigma. M. ulcerans was first detected in the environment in the 1990s by Australian researchers using polymerase chain reaction (PCR) \[[@pmed-0020108-b29],[@pmed-0020108-b30]\]. Subsequently, PCR was used by others to identify M. ulcerans in aquatic insects obtained from endemic areas in Africa \[[@pmed-0020108-b31]\], leading to the hypothesis that M. ulcerans may be transmitted by biting water bugs of the insect order Hemiptera (Naucoridae and Belostomatidae; [Figure 4](#pmed-0020108-g004){ref-type="fig"}). ::: {#pmed-0020108-g004 .fig} Figure 4 ::: {.caption} ###### Semi-Aquatic Hemiptera That Have Tested Positive for M. ulcerans The top row is Macrocoris sp., 1.0 centimeter in body length (Family Naucoridae), and the bottom row is Appasus sp., about 2.5 centimeters in body length (Family Belastomatidae). The ventral and dorsal views are in the left and right panels, respectively. ::: ![](pmed.0020108.g004) ::: In support of this hypothesis, M. ulcerans has been detected in the salivary glands of Naucoris sp., and has been transmitted to laboratory mice via this aquatic insect \[[@pmed-0020108-b32],[@pmed-0020108-b33]\]. There is additional evidence that M. ulcerans DNA can be detected by PCR in other aquatic insect predators (e.g., Odonata and Coleoptera), as well as in aquatic snails, small fish, and the biofilm of aquatic plants \[[@pmed-0020108-b34]\]. Despite this, only two pure cultures of M. ulcerans have been obtained from environmental sources. In Australia, it has been postulated that aerosols arising from contaminated water may disseminate M. ulcerans and infect humans via the respiratory tract, or through contamination of skin lesions and minor abrasions \[[@pmed-0020108-b35],[@pmed-0020108-b36]\], but this has yet to be proven. Recent progress has been rapid, but the exact mode of transmission, the key important reservoir species, and transmission of M. ulcerans through the aquatic food chain remain to be elucidated. Immune response. {#s9c} ---------------- The immune mechanisms involved in protection against Buruli ulcer are also largely unknown at present. Interestingly, peripheral blood mononuclear cells obtained from people with a past or current M. ulcerans infection typically show a strong T helper (Th)--2 cytokine response when exposed in vitro to M. ulcerans. In contrast, samples obtained from their household contacts (exposed healthy controls) exhibit a Th-1 immune response, suggesting that natural resistance may be determined by cell-mediated immune mechanisms directed against intracellular organisms \[[@pmed-0020108-b37]\]. In one fascinating case study, it has been shown that the development of ulcerative M. ulcerans disease is associated with a shift from the Th-1 to the Th-2 phenotype \[[@pmed-0020108-b38]\]. Interleukin-10 may be a key cytokine that mediates local Th phenotype switching within nodules and ulcers \[[@pmed-0020108-b39]\]. Antibodies may also have a protective role against M. ulcerans, as the pathogen is extracellular during active disease. Experimental infection of mice genetically inactivated in various compartments of the immune response (B lymphocytes, Th cells, and cytolytic T lymphocytes, cytokines, and monokines) will help us to understand how host immunity is acquired. Developing new drugs. {#s9d} --------------------- M. ulcerans is susceptible to several anti-mycobacterial drugs in vitro, but the most promising results in the mouse footpad model were obtained with a combination of rifampicin and amikacin \[[@pmed-0020108-b40]\]. A human trial has recently shown that early nodular lesions may be rendered culture-negative after a minimum of four weeks therapy with rifampicin plus streptomycin \[[@pmed-0020108-b14],[@pmed-0020108-b20]\]. Further research to identify cheap, safe, and effective oral combinations that can be used as an adjuvant to surgery or that could even replace surgery for early lesions is urgently required. At least one new compound, which appears safe for humans in early phase I trials, has remarkable activity in vitro against many mycobacterial species including M. tuberculosis and M. ulcerans \[[@pmed-0020108-b41]\]. M. ulcerans toxin. {#s9e} ------------------ M. ulcerans makes a family of toxic macrolides, the mycolactones, that are required for virulence \[[@pmed-0020108-b17]\]. Mycolactone causes cells in cell culture assays to undergo apoptosis and necrosis and produces a lesion that closely resembles Buruli ulcer when injected directly into guinea pig skin \[[@pmed-0020108-b42]\]. Although toxic lipid molecules are relatively commonly produced by mycobacteria, the synthesis of mycolactone itself appears to be restricted to M. ulcerans \[[@pmed-0020108-b43]\]. Prospects for developing a vaccine. {#s9f} ----------------------------------- There is no specific vaccine against M. ulcerans, but the M. bovis BCG vaccine offers some protection, albeit short lived \[[@pmed-0020108-b44],[@pmed-0020108-b45]\]. BCG may possibly provide more enduring protection against the most severe forms of Buruli ulcer \[[@pmed-0020108-b46]\]. Current prospects for better vaccines include improved or repeated BCG vaccination, rational attenuation of a live M. ulcerans isolate, or subunit vaccines aimed at protein antigens or the toxin mycolactone itself \[[@pmed-0020108-b47]\]. Genome, bacterial population structure, and serodiagnosis. {#s9g} ---------------------------------------------------------- The expected publication of the whole M. ulcerans genome sequence in 2005 will mark a major milestone for Buruli ulcer researchers. Already the project has uncovered the presence of a large virulence plasmid that encodes mycolactone production proteins \[[@pmed-0020108-b48]\]. Biosynthesis of mycolactone requires three polyketide synthase enzymes and at least two accessory enzymes, all of which are located within a 110-kb cluster on this plasmid. The genome project has also revealed a remarkably high copy number of two insertion sequences, accounting for more than 5% of the total genome. There is evidence of considerable genome decay, with many potential pseudogenes and DNA deletions. These data coincide with accumulating evidence that suggests a reservoir in insects or other aquatic species and indicate that M. ulcerans may be passing through an evolutionary bottleneck as it adapts to life in a specialised niche environment. Molecular typing of M. ulcerans isolates has revealed a clonal population structure within specific geographical regions. Innovative genetic fingerprinting methods will be required to reveal local transmission pathways and environmental reservoirs. In endemic areas, clinical diagnosis of advanced Buruli ulcer lesions by experienced clinicians is quite reliable. Diagnostic confidence can be increased by detecting acid-fast bacilli in smears or biopsies, and the diagnosis confirmed by culture or PCR \[[@pmed-0020108-b49]\]. A recent report describes a new dry-reagent PCR for Buruli ulcer that could be used in small regional centres, and reports from Ghana suggest a sensitivity of 95% \[[@pmed-0020108-b50]\]. Diagnosis of Buruli ulcer outside endemic areas or of pre-ulcerative lesions can be challenging. The genome sequence will greatly assist the development of a noninvasive serodiagnostic assay based on M. ulcerans--specific antigens. Conclusion {#s10} ========== Buruli ulcer is now emerging from long years of neglect: interest and momentum are growing. However, there is much to do if we are to understand why the disease is becoming more common and how this relates to human activity. The current control strategy of early detection and treatment should be scaled up in the affected countries. Our ultimate goal is the development of an effective and safe vaccine able to provide long-lasting protection for those who live in endemic areas. A video about Buruli ulcer is freely available at <http://www.who.int/gtb-buruli/publications/video.htm>. Citation: Johnson PDR, Stinear T, Small PLC, Pluschke G, Merritt RW, et al. (2005) Buruli ulcer (M. ulcerans infection): New insights, new hope for disease control. PLoS Med 2(4): e108. PCR : polymerase chain reaction Th : T helper WHO : World Health Organization [^1]: Paul D. R. Johnson is at the Department of Infectious Diseases, Austin Health, Heidelberg, Australia, at the University of Melbourne, Australia, and at the Department of Microbiology, Monash University, Melbourne, Australia. Timothy Stinear is at the Department of Microbiology, Monash University, Melbourne, Australia. Pamela L. C. Small is at the Department of Microbiology, University of Tennessee, Knoxville, Tennessee, United States of America. Gerd Pluschke is at the Department of Molecular Immunology, Swiss Tropical Institute, Basel, Switzerland. Richard W. Merritt is at the Department of Entomology, Michigan State University, East Lansing, Michigan, United States of America. Francoise Portaels is at the Department of Microbiology, Institute of Tropical Medicine, Antwerp, Belgium. Kris Huygen is at the Department of Mycobacterial Immunology, WIV-Pasteur Institute Brussels, Belgium. John A. Hayman is at the Department of Anatomy and Cell Biology, Monash University, Melbourne, Australia. Kingsley Asiedu is at the Department of Communicable Diseases, World Health Organization, Geneva, Switzerland. [^2]: Competing Interests: PLCS is listed as an inventor on a patent for mycolactone; PDRJ, TS, and PLCS are listed as inventors on a provisional patent for the PKS locus that makes mycolactone. However, any financial benefit that may arise from these patents would be in the field of either cancer medicine (mycolactone may inhibit some cell functions) or combinatorial chemistry, rather than Buruli ulcer.	PubMed Central	2024-06-05T03:55:55.892384	2005-4-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1087202/", "journal": "PLoS Med. 2005 Apr 26; 2(4):e108", "authors": [ { "first": "Paul D. R", "last": "Johnson" }, { "first": "Timothy", "last": "Stinear" }, { "first": "Pamela L. C", "last": "Small" }, { "first": "Gerd", "last": "Pluschke" }, { "first": "Richard W", "last": "Merritt" }, { "first": "Francoise", "last": "Portaels" }, { "first": "Kris", "last": "Huygen" }, { "first": "John A", "last": "Hayman" }, { "first": "Kingsley", "last": "Asiedu" } ] }
PMC1087203	Hypertension is common in affluent societies and a major risk factor for heart disease. In Canada, hypertension is the leading primary diagnosis for patient visits to physicians\' offices. Beyond recommending lifestyle changes such as losing weight, quitting smoking, and lowering salt and alcohol intake, prescription drugs are indicated in many patients. As a consequence, antihypertensive drugs are the leading category of prescription drugs in Canada, accounting for 20% of prescription drug sales. Several classes of drugs are available for treatment, including diuretics, ACE inhibitors, and calcium channel blockers. First-line treatment with thiazide diuretics---the oldest and by far the cheapest drug class---has been shown in randomized trials to reduce serious cardiovascular morbidity and mortality with benefits at least as great as first-line treatment with other drug classes.[](#pmed-0020113-g001){ref-type="fig"} ::: {#pmed-0020113-g001 .fig} ::: {.caption} ###### Prescription trends for hypertension ::: ![](pmed.0020113.g001) ::: Steve Morgan and colleagues set out to examine whether prescribing practices were in accordance with this evidence. They analyzed administrative claims data from a public drug plan for seniors (residents of age 65 and older) to determine trends in first-line hypertension drug use. During the period from 1993 to 2000, over 82,000 seniors were identified as new users of hypertension drugs. Less than a third of these patients received thiazides (alone or as part of a combination regimen) as a first-line treatment. The share of new patients receiving a thiazide increased over the study period, but did not exceed 45% at any point. Women were more likely than men, and older patients were more likely than younger ones, to receive thiazides. Comorbidities also influenced prescribing practices: patients without concurrent diagnoses were more likely to receive thiazides. While for some comorbidities (such as previous acute myocardial infarction) evidence suggested that there were good reasons to prescribe drugs other than thiazides, no such evidence existed for many of the other conditions that nevertheless were associated with lower prescription of thiazides. Changes in drug availability and existing evidence during the period studied make it difficult to calculate the exact extent to which thiazides were under-prescribed. However, the study shows that many patients received drugs that had previously been found to be no better at treating hypertension than much cheaper alternatives. Drug prices changed over the study period as well, but even comparing the lowest price for any of the alternatives to thiazides, \$0.34 per day, with the constant cost of less than \$ 0.01 per day for a thiazide makes it clear that a lot of money was wasted. On a more positive note, prescription of thiazides as a first-line therapy rose over the study period---from 25% to 42% in patients without comorbidities. Most of the increase occurred shortly after a specific local education campaign. This suggests that repeated targeting of prescribing physicians---many of whom receive regular marketing material from pharmaceutical companies and subscribe to the general view that newer drugs are better---with current evidence-based information should be considered. One of the most influential studies comparing antihypertensive drugs, the ALLHAT study, also supported the use of thiazides as first-line drugs. ALLHAT was published in 2003, and its results widely publicized. According to Steve Morgan, "Anecdotal evidence suggests that ALLHAT has had an influence on prescription practices, but I am not aware of a large-scale analysis yet."	PubMed Central	2024-06-05T03:55:55.894162	2005-4-26	{ "license": "Creative Commons Zero - Public Domain - https://creativecommons.org/publicdomain/zero/1.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1087203/", "journal": "PLoS Med. 2005 Apr 26; 2(4):e113", "authors": [] }
PMC1087204	As the HIV-1 epidemic continues to grow, mutations in the virus that confer drug resistance are becoming increasingly important in the clinical management of patients worldwide. Of all the different virus subtypes (A, B, C, D, F, G, H, J, and K) and a rapidly increasing number of established and emerging recombinant viruses, it is subtype B that predominates in Western Europe, the United States, and the rest of the industrialized world. Antiretroviral drugs were developed by studying subtype B, and most data on the genetic mechanisms of HIV drug resistance are also from subtype B. However, worldwide, subtype B is in the minority (\~10% of the infected population). In Africa, for example, where there is broad viral diversity, there is a greater spread of subtypes, with subtype C being the most common, representing over half of all infections. Although it seems that current drugs---developed against subtype B virus---are active against non-subtype-B virus, one critical issue is whether viruses from some subtypes or particular regions are more likely than others to develop resistance against certain drugs. Another crucial issue is to identify the mutations that confer drug resistance in non-B subtypes. Answering these questions might determine whether initial treatment strategies should be different for people with non-subtype-B viruses, and also could help decide how patients with non-subtype-B virus who fail antiretroviral therapy should be managed.[](#pmed-0020117-g001){ref-type="fig"} ::: {#pmed-0020117-g001 .fig} ::: {.caption} ###### There are many subtypes of HIV-1 worldwide ::: ![](pmed.0020117.g001) ::: In a paper in this month\'s PLoS Medicine, Rami Kantor and colleagues from a worldwide collaboration have looked at the mutations found in 3,686 people with non-subtype-B HIV-1 infections compared with those in 4,769 people with subtype B infections. They wanted to answer two questions: first, whether the mutations that cause drug resistance in subtype B viruses also develop in non-subtype-B viruses exposed to antiretroviral drugs, and second, whether novel mutations (i.e., not previously seen in subtype B virus) develop in non-subtype-B viruses when they fail to respond to antiretroviral drugs. What they found was that all of the 55 drug-resistance mutations that have been known to occur in subtype B also occurred in at least one non-subtype-B isolate, and most of these mutations were also statistically associated with antiretroviral treatment in at least one non-B subtype. Conversely, of the 67 mutations associated with antiretroviral therapy in at least one non-B subtype, 61 were also associated with antiretroviral therapy in subtype B isolates. So it appears that few novel mutations are arising in non-subtype-B viruses exposed to the current antiretroviral drugs and that the present focus on subtype B mutations for global surveillance and genotypic assessments of drug resistance is a reasonable approach. However, the authors emphasize that differences in the types and patterns of drug-resistance mutations are likely to differ between the subtypes, and that larger numbers of samples and further analyses are needed to exclude the possibility of new and/or rare subtype-specific mutations.	PubMed Central	2024-06-05T03:55:55.894767	2005-4-26	{ "license": "Creative Commons Zero - Public Domain - https://creativecommons.org/publicdomain/zero/1.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1087204/", "journal": "PLoS Med. 2005 Apr 26; 2(4):e117", "authors": [] }
PMC1087205	The debate between a libertarian bioethicist and a communitarian bioethicist \[[@pmed-0020121-b1]\] illustrates why American bioethics is becoming increasingly marginalised and irrelevant to the democratic society that it intends to serve. Both participants in the debate, Arthur Caplan and Carl Elliott, explicitly reject the notion of "human nature" as a foundation for bioethics. But without human nature, on what grounds can advances in biomedical knowledge be called good or bad, right or wrong, or even harmful or beneficial? Clearly Caplan and Elliott have to accept something as a touchstone of their bioethical discourse, or it will lapse into windy incoherence. Although they approach it from different angles, this benchmark is informed consent, with Elliott placing the stress on "informed" and Caplan on "consent". As a result, their lively disagreement over enhancement technology is just verbal sparring and not a battle of ideas. Caplan believes that the consumer-patient is sufficiently mature to weigh up the dangers; Elliott is more sceptical. Neither appears to think that it makes any sense to argue that technology should be suited to human nature. This belief seems to be widespread in the bioethics community. Ruth Macklin, a bioethicist at Albert Einstein College of Medicine, argued recently, for instance, that "human dignity" is an empty and meaningless concept \[[@pmed-0020121-b2]\]. However, academic discourse has failed to dislodge from the heads of the hoi polloi the conviction that the starting point of ethics is not consent but happiness. The man in the street, the ultimate consumer of bioethics, still believes in human nature. The notion that human dignity is meaningless would be regarded by nearly all Americans as not merely absurd but reprehensible. What I find odd in the writings of many bioethicists is that they skirt around the question that the average person wants to ask: will this enhancement make me happy in a deeply satisfying and fulfilling way? He or she is much less interested in whether all the boxes on the informed consent form have been ticked properly. Consequently, as the Caplan--Elliott bunfight demonstrates, bioethicists are now reduced to arguing that human enhancement is good if people want it---even if they want it mainly because powerful commercial interests have persuaded them to, even if it is weird and kinky, even if it won\'t make them happy. Elliott\'s fascinating book Better than Well \[[@pmed-0020121-b3]\] is evidence that exercising a right to enhancement still leaves many lives hollow and unhappy. Sooner or later people will ask why they hadn\'t been warned, and a lot of bioethicists will be looking for jobs. Citation: Cook M (2005) Is American bioethics lost in the woods? PLoS Med 2(4): e121. [^1]: Competing Interests: The author has declared that no competing interests exist.	PubMed Central	2024-06-05T03:55:55.895255	2005-4-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1087205/", "journal": "PLoS Med. 2005 Apr 26; 2(4):e121", "authors": [ { "first": "Michael", "last": "Cook" } ] }
PMC1087206	DESCRIPTION of CASE {#s1} =================== A 75-year-old right-handed woman presented with a two-day history of symptoms suggestive of a right hemispheric stroke (slurred speech and left facial and left arm weakness). She had no previous cerebrovascular symptoms, such as symptoms of a previous transient ischaemic attack or amaurosis fugax (loss of vision in one eye due to a temporary lack of blood flow to the retina). Past medical history included long-standing hypertension and chronic obstructive pulmonary disease. She was on amlodipine, 10 mg once daily, and salbutamol and fluticasone inhalers. On examination, the patient had a Glasgow Coma Score of 15, she was apyrexial, her pulse rate was regular, at 80 per min, and her blood pressure was 176/99 mm Hg. There was no cardiac murmur or carotid bruits. She had left-sided weakness. What Investigation Is Indicated at This Stage? {#s1a} ---------------------------------------------- Brain imaging is necessary for two main reasons. The first is to exclude a brain haemorrhage (responsible for 25% of all strokes \[[@pmed-0020079-b1]\]); against this diagnosis was the absence of headache and a normal Glasgow Coma Score. The second is to rule out a brain tumour. Computed tomography (CT) brain scanning on admission showed two areas of low density within the right cerebral hemisphere, one in the right parietal lobe and one in the posterior right frontal lobe ([Figure 1](#pmed-0020079-g001){ref-type="fig"}), most likely ischaemic in nature. Small low-density lesions consistent with lacunar infarcts were also seen in both basal ganglia, the most prominent ones seen within the left basal ganglia. There was also marked frontal atrophy, and atrophy of the brain stem structures. The CT scan showed no evidence of haemorrhagic transformation of the infarct, a condition that is a contraindication for anticoagulation. ::: {#pmed-0020079-g001 .fig} Figure 1 ::: {.caption} ###### CT Brain Scan on Admission Showing an Infarct in the Posterior Right Frontal Lobe ::: ![](pmed.0020079.g001) ::: Routine blood tests (full blood count, urea and electrolytes, and clotting), an electrocardiogram, and a chest X ray were performed before the CT brain scan. The electrocardiogram showed no arrhythmia or changes suggestive of an old or new myocardial infarction. The normal electrocardiogram raised the possibility of embolisation from a large artery (such as the right carotid artery or the aortic arch), rather than from the heart. How Did We Identify the Source of Embolisation? {#s1b} ----------------------------------------------- A carotid ultrasound showed a calcified, haemodynamically significant plaque at the right carotid bifurcation. A similar lesion seen at the left carotid bifurcation was not haemodynamically significant. Carotid angiogram showed a tight stenosis of the distal right common carotid artery ([Figure 2](#pmed-0020079-g002){ref-type="fig"}), and occlusion of the left internal carotid artery. ::: {#pmed-0020079-g002 .fig} Figure 2 ::: {.caption} ###### Digital Subtraction Angiography Showing a Tight Stenosis of the Distal Right Common Carotid Artery ::: ![](pmed.0020079.g002) ::: What Was the Management Plan at This Stage? {#s1c} ------------------------------------------- Bendrofluazide, 2.5 mg once daily, and perindropril, 2 mg once daily, were prescribed to treat the patient\'s hypertension. Antiplatelets (aspirin, 75 mg once daily, and clopidogrel, 75 mg once daily) and a statin (simvastatin, 40 mg at night) were also prescribed. Carotid endarterectomy was scheduled. Bisoprolol, 2.5 mg once daily, was prescribed to reduce the risk of periprocedural myocardial ischaemia \[[@pmed-0020079-b2]\]. The patient\'s symptoms gradually improved and she was discharged two days later. A contrast-enhanced CT brain scan performed nine days after the onset of symptoms confirmed the previous findings. The dose of perindopril was gradually increased from 2 mg to 6 mg daily to achieve satisfactory control of the patient\'s hypertension. After a Hemispheric Stroke, How Soon Should Carotid Endarterectomy Be Performed? {#s1d} -------------------------------------------------------------------------------- Six weeks after the stroke, the patient underwent elective right carotid endarterectomy under general anaesthesia with the use of a shunt. Although some surgeons favour early carotid endarterectomy, most agree that this should not be performed earlier than six weeks, to allow the autoregulative mechanism of the brain to recover \[[@pmed-0020079-b3],[@pmed-0020079-b4]\]. Arteriotomy was closed with a Dacron patch. No neurological deficits or cranial nerve palsy were noted postoperatively. The patient\'s discharge was postponed until the seventh postoperative day because of a mild urinary tract infection and an episode of syncope. Histology revealed an atheroma producing near total lumen occlusion. Eleven months after the operation no new neurological events have occurred. DISCUSSION {#s2} ========== Causes of Hemispheric Ischaemic Stroke {#s2a} -------------------------------------- Mohr et al. classified the causes of ischaemic stroke into three broad categories: embolism to the brain of cardiac or aortic origin, cerebral ischaemia due to perfusion failure and artery-to-artery embolism, and cerebral artery thrombosis \[[@pmed-0020079-b5]\]. Embolism to the brain of cardiac or aortic origin includes myocardial infarction, atrial fibrillation, valvular disease in native, prosthetic, or repaired cardiac valves (including mitral valve prolapse), embolism of aortic arch origin, and myxoma of the heart and from venous thrombi via a patent foramen ovale. Cerebral ischaemia due to perfusion failure and artery-to-artery embolism includes large artery atherosclerotic plaque, vasculitis, and other arterial disease and small artery occlusion. Thrombosis is caused by prothrombotic states. Carotid stenosis accounts for about 20% of all cases of ischaemic stroke \[[@pmed-0020079-b6]\], and is considered as the single most preventable cause of stroke. Like all atherosclerotic diseases, the most common risk factors for carotid stenosis are smoking, hypertension, hyperlipidaemia, and diabetes mellitus. Investigation of Carotid Artery Bifurcation Stenosis {#s2b} ---------------------------------------------------- Carotid ultrasound duplex is the imaging method of choice for the initial investigation of a patient with suspected carotid artery stenosis \[[@pmed-0020079-b7]\]. It is non-invasive and low-cost, and can be easily repeated if necessary. Anatomical criteria at the point of stenosis (cross-sectional area reduction or diameter reduction) should always be applied to ensure that stenosis is present and that flow velocity is not increased secondary to a vessel curve. A limitation of using anatomical criteria to estimate the degree of carotid stenosis is that in the presence of heavy acoustic shadowing (due to calcification) interrupting flow visualisation, no intraluminal diameter reduction can be calculated ([Figure 3](#pmed-0020079-g003){ref-type="fig"}). ::: {#pmed-0020079-g003 .fig} Figure 3 ::: {.caption} ###### Carotid Artery Ultrasound Showing a Completely Calcified Atherosclerotic Plaque (Geroulakos Type 5) Although post-stenotic velocities are increased, accurate grading with ultrasound is not possible. ::: ![](pmed.0020079.g003) ::: Because of these problems, intrastenotic velocity measurements are widely used; in cases of acoustic shadowing due to calcification, these measurements should be performed just distally to the acoustic shadowing, at the point of the flow jet. Long acoustic shadowing, known also as the Gibraltar sign \[[@pmed-0020079-b8]\], can result in falsely reduced velocity measurements and downgrade the stenosis. It has been found that velocity measurements are affected by different ultrasound scanners, physiological changes, and the presence of contralateral occlusion or "tandem" lesions \[[@pmed-0020079-b9]\]. To overcome this problem, intrastenotic flow velocities are "normalised" using the common carotid or distal internal carotid flow velocities as a reference point, and the resulting ratio is used. There are potential problems related to the sampling point and to patients with heart failure with falsely low end-diastolic velocity of the common carotid artery \[[@pmed-0020079-b10]\]. In order to overcome potential limitations of individual criteria, the use of combined criteria has been suggested ([Figure 4](#pmed-0020079-g004){ref-type="fig"}) \[[@pmed-0020079-b11],[@pmed-0020079-b12]\]. ::: {#pmed-0020079-g004 .fig} Figure 4 ::: {.caption} ###### Carotid Artery Ultrasound Showing an Echolucent Atherosclerotic Plaque (Geroulakos Type 2) This plaque is causing approximately 80% stenosis, as determined using diameter reduction and velocity ratio methods. ::: ![](pmed.0020079.g004) ::: Selective arteriogram is nowadays rarely indicated, because the procedure itself can cause stroke, and non-invasive alternative methods (magnetic resonance angiography) are available. There is no consensus on the optimum method to grade carotid stenosis ([Figure 5](#pmed-0020079-g005){ref-type="fig"}) \[[@pmed-0020079-b13],[@pmed-0020079-b14]\]. ::: {#pmed-0020079-g005 .fig} Figure 5 ::: {.caption} ###### Angiographic Methodology of Grading Carotid Stenosis Grading can be done in relation to the carotid bulb (ECST method) or the distal internal carotid stenosis (NASCET method). CCA, common carotid artery; ECA, external carotid artery; ICA, internal carotid artery. ::: ![](pmed.0020079.g005) ::: The Role of Carotid Endarterectomy in Preventing a Recurrent Event {#s3} ================================================================== Carotid reconstruction was first performed by Eastcott et al. at St. Mary\'s Hospital, London, in 1954 \[[@pmed-0020079-b15]\]. However, it took nearly four decades until trial evidence became available to show that carotid endarterectomy was better than best medical treatment in patients with amaurosis fugax or hemispheric symptoms, transient ischaemic attacks, or stroke who had made a good recovery and whose symptoms were caused by severe carotid bifurcation stenosis (\>70% with the North American Symptomatic Carotid Endarterectomy Trial \[NASCET\] method or \>80% with the European Carotid Surgery Trial \[ECST\] method) \[[@pmed-0020079-b13],[@pmed-0020079-b14]\]. The two-year risk of stroke in the medical arm of NASCET was 26% compared with 9% in those who underwent endarterectomy \[[@pmed-0020079-b13]\]. Subsequently, the NASCET trialists reported that endarterectomy reduces the five-year risk of stroke in moderate stenosis (50%--69%) from 22.2% to 15.7% \[[@pmed-0020079-b16]\]. Longer follow-up also showed that the long-term risk of stroke after carotid endarterectomy is about 1% per year. A recent meta-analysis of the NASCET and ECST trials showed that benefit from surgery was greatest in men, patients aged 75 years or older, and those randomised within two weeks after their last ischaemic event, and fell rapidly with increasing delay \[[@pmed-0020079-b17]\]. Surgery is usually performed at six weeks if there is good recovery, but there is a tendency to perform it earlier in patients with transient ischaemic attacks or strokes with good recovery when CT brain scan shows no infarct. Surgery reduces the risk of stroke by 50% even if the event occurred more than six months previously, as shown by the Medical Research Council Asymptomatic Carotid Surgery Trial \[[@pmed-0020079-b18]\], but because of the low incidence of stroke at five years and the relatively small number of patients in the trial, benefit was only marginally significant \[[@pmed-0020079-b18]\]. Best Medical Treatment {#s3a} ---------------------- While recovering from stroke and awaiting carotid endarterectomy, aspirin even at a low dose of 75 mg daily reduces the risk of recurrence. This is improved with dipyradamole \[[@pmed-0020079-b19]\], but not clopidogrel \[[@pmed-0020079-b20]\]. The Heart Protection Study has recently shown that regardless of pre-treatment lipid levels, lipid-lowering agents are beneficial in secondary prevention of stroke \[[@pmed-0020079-b21]\]. In our patient\'s case, total cholesterol levels after treatment were 3.6 mmol/l. Carotid Endarterectomy under Local Anaesthesia, or Carotid Angioplasty and Stenting? {#s3b} ------------------------------------------------------------------------------------ Minimally invasive treatment is considered nowadays a preferable mode of delivering health care, and carotid disease is no exception. Meta-analysis of previous studies has shown that carotid endarterectomy under local anaesthesia has less surgical hazards than under general anaesthesia. This was the basis for the GALA trial---a multicentre randomised trial assessing the relative risks of stroke, cardiac events, and death with these two different treatments (<http://www.galatrial.com/>). The results are not yet available. ### Key Learning Points Although carotid stenosis accounts for about 20% of all cases of ischaemic stroke, it has been considered as the single most preventable cause of stroke.Carotid ultrasound is the method of choice for the initial investigation of a patient with suspected carotid artery stenosis.Carotid endarterectomy has been proven to reduce the incidence of recurrent stroke, mainly in severe stenosis.Carotid stenting has recently emerged as an alternative procedure.Early intervention as soon as possible is desirable to reduce the risk of recurrent stroke. Carotid angioplasty and stenting with the aid of distal protection devices has recently emerged as a good alternative to endarterectomy, being equivalent or slightly better, when periprocedural complications are considered \[[@pmed-0020079-b22],[@pmed-0020079-b23]\]. In the SAPPHIRE trial, performed in selected high-risk patients, those allocated to the stenting group had fewer cardiovascular events than those undergoing surgery \[[@pmed-0020079-b23]\]. Citation: Kakkos SK, Geroulakos G (2005) A 75-year-old woman with a hemispheric stroke. PLoS Med 2(4): e79. CT : computed tomography ECST : European Carotid Surgery Trial NASCET : North American Symptomatic Carotid Endarterectomy Trial [^1]: Stavros K. Kakkos is a Clinical Vascular Fellow with a particular interest in carotid artery disease and George Geroulakos is a Consultant Vascular Surgeon and Senior Lecturer, Vascular Unit, Ealing Hospital, London, United Kingdom. [^2]: Competing Interests: The authors declare that they have no competing interests.	PubMed Central	2024-06-05T03:55:55.895776	2005-4-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1087206/", "journal": "PLoS Med. 2005 Apr 26; 2(4):e79", "authors": [ { "first": "Stavros K", "last": "Kakkos" }, { "first": "George", "last": "Geroulakos" } ] }
PMC1087207	Michael Makover\'s Viewpoint: We Should Use High-Sensitivity Carotid Ultrasound to Detect Very Early Atherosclerosis and Treat Aggressively {#s1} =========================================================================================================================================== Atherosclerotic vascular disease is the greatest cause of death and disability in developed countries. The best way to prevent these outcomes is to detect disease at the earliest possible stage and to attack it with pinpoint-targeted aggressive treatment. B-mode ultrasonography of the carotid artery to measure the intima-media thickness (IMT) ([Figure 1](#pmed-0020098-g001){ref-type="fig"}) is the most effective way to do that. ::: {#pmed-0020098-g001 .fig} Figure 1 ::: {.caption} ###### Ultrasonography of the Carotid Artery to Measure the IMT This figure provides a framework for understanding the multiple, quantitative calculations undertaken for each patient\'s carotid ultrasound scan. Two images of carotid artery without plaques (left) and two with plaques (right) are shown at the top. ROI indicates the region of interest in which thickness measurements are made. These were obtained from a single patient. Schematics for each image are provided below the video images. The grey lines indicate the measured IMT within the defined region of interest. Regarding the plaques (PL~1~ and PL~2~), M indicates maximum plaque thickness, and L indicates minimum lumen diameter at the site of maximum plaque thickness. Percent diameter stenosis is calculated as \[M/(M + L)\] × 100%. The graph (bottom) indicates length in millimeters on the x-axis and thickness in millimeters on the y-axis. The areas from the far wall of the left (L) and right (R) common carotid arteries are measured in all patients. AIMT designates the average IMT derived from the left and right measurements. PL~1~ and PL~2~ are the lesion lengths of the respective plaques, and M~1~ and M~2~ are the respective maximal thicknesses. APT~1~ and APT~2~ designate the respective average lesion thicknesses for each plaque. The total area is given by the sum of all the areas shown on the graph in units of square millimeters. The average total thickness (ATT) is given by the total area divided by the total length shown on the x-axis. PL~n~ on the x-axis is intended to indicate that the calculations are undertaken using these concepts, irrespective of the number of plaques identified in a given study. (Figure from \[[@pmed-0020098-b28]\] with permission of all four authors.) ::: ![](pmed.0020098.g001) ::: It is time to end the mindset that thinks of atherosclerosis as an inevitable function of aging. It is not: it is a disease, whether it causes an acute event or gradual decline. As with any disease, the earlier and more intensively we attack it---with medications and lifestyle changes---the more successful we will be in containing and reversing it. Ultrasonography of the carotid to detect plaque and increased intimal wall thickness is the best, safest, and easiest early detector we have. Patients at risk for atherosclerotic cardiovascular disease are identifiable only up to a point using traditional methods, including the Framingham risk score \[[@pmed-0020098-b1]\], the European SCORE (Systemic Coronary Risk Evaluation) \[[@pmed-0020098-b2]\], and the C-reactive protein level \[[@pmed-0020098-b3]\]. These are particularly effective for people in the highest risk groups, but they have serious flaws. They do not take family history of premature coronary artery disease into account; they do not benefit from new assays that can differentiate lipoproteins by particle size and number, both of which are important factors in atherogenicity; and they do not include the quality of the patient\'s diet, abdominal fat content, racial and ethnic genetic differences, confounding medical conditions, certain gender differences, and other factors. Thus, many patients deemed "low risk" for acute events within a period of time might still develop disease. All of the current risk models were developed within the obsolete gestalt of life spans of perhaps 60 to 70 years, whereas people are now increasingly living productively into their 80s and 90s. Risk tables are based only on cardiac events over a relatively short time period of ten years. Progress in medical science is now on the steepest slope of exponential growth. It is reasonable to expect that those who maintain healthy circulation will experience good quality life beyond the 100-year mark in the surprisingly near future, and perhaps to 120 years not long after. Furthermore, risk escalates with age. Thus, an initial low risk profile becomes far more significant when considered in the context of long lives. Equally important, gradual impairment of blood flow that nourishes tissues can lead to peripheral vascular disease, increased susceptibility to infection, cognitive decline, frailty, and other changes that impair life quality and longevity. While risk factors help individualize treatment, they do not address the most critical issue: identifying atherosclerosis at the very earliest stage to try to stop and reverse the process before any damage has occurred. Risk factors identify those who might develop disease. The "smoking gun" is to identify disease when it actually first develops. Atherosclerosis begins as a lipid and inflammatory cell deposition in the intimal subendothelial space, which expands outward, with only a slight inward intrusion on the lumen. This does not affect blood flow unless the lumen-side plaque cap ruptures, and a clot forms and expands into the blood stream. Most clots are limited by thrombolytic mechanisms, but significant blockage usually results as the clot is scarred over. Some clots expand all the way across to the other side of the artery, completely occluding it, causing a heart attack or stroke \[[@pmed-0020098-b4]\]. Thus, if we could discover plaque at the very earliest stage, we could aggressively attack it with our full armamentarium of lifestyle modification, addressing all the factors noted above, and medications---statins, ACE inhibitors, aspirin, niacin, ezetimibe, and others---to stabilize and then reverse the process before a rupture occurs. The REVERSAL trial, among many others, has shown that atherosclerosis is reversible with aggressive treatment \[[@pmed-0020098-b5]\]. Carotid ultrasound is an easy, safe, noninvasive method for detecting early, focal plaques and early thickening of the inner lining of the artery \[[@pmed-0020098-b6]\]. There is ample evidence that this is an effective and reproducible detector, and a predictor of progression and symptomatic disease \[[@pmed-0020098-b7]\]. In contrast, angiography is invasive and is insensitive to early changes in the artery wall thickness. Magnetic resonance imaging is expensive, cumbersome, and still experimental. Calcium scoring by computed tomography is less sensitive and less reproducible and suggests plaque presence only indirectly \[[@pmed-0020098-b8]\]. Angiography and calcium scoring entail considerable X ray exposure. Using high-sensitivity carotid ultrasound as a primary screening tool would add modest cost, but that would be more than offset by sharply reducing the enormous worldwide toll of atherosclerosis. Screening tests should be easy, affordable, widely available, and predictive. Carotid ultrasound meets all these criteria and should be used to screen everyone early on who is at any increased risk by measure of the expanded risk factors noted above. Shah Ebrahim\'s Viewpoint: Reducing Risk Factors across the Population Is Better Than Identifying Only Those at Highest Risk {#s2} ============================================================================================================================ The "high risk" approach to reducing heart disease involves identifying those at highest risk by means of risk factor scoring, dominated by the Framingham equations \[[@pmed-0020098-b9]\], followed by aggressive risk factor control. This approach must be contrasted with the "population" approach to prevention ([Figure 2](#pmed-0020098-g002){ref-type="fig"}), which aims to shift the whole distribution of risk factors in a beneficial direction and thereby reduce the incidence of cardiovascular disease in the whole population. In the population approach, all have to take part in making the population healthy but only relatively few will benefit---the "prevention paradox" \[[@pmed-0020098-b10]\]. ::: {#pmed-0020098-g002 .fig} Figure 2 ::: {.caption} ###### The High Risk Versus the Population Approach to Prevention (Figure: John Emberson) ::: ![](pmed.0020098.g002) ::: Population change does require political commitment---as Virchow commented, "Mass diseases require mass solutions" \[[@pmed-0020098-b11]\]. There are several politically difficult population actions that many public health agencies would wish to implement to reduce heart disease. These include extension of restrictions on the use and sale of tobacco products, reduction of "hidden" saturated fats in processed foods, and an increase in the opportunities for regular physical exercise. Some of these policies cut across the vested interests of international corporations that have powerful means of controlling elected governments, leading to a laissez-faire approach of emphasising individual "choice" as the means of improving public health \[[@pmed-0020098-b12]\]. Although stopping smoking and lowering blood pressure and blood cholesterol are beneficial, the crucial question is whether using risk factor screening tools to identify people at high risk for disease and then treating them is more effective than focusing on individual risk factors. In common with many screening enterprises, there is only limited evidence to help determine whether screening for those at highest risk of heart disease is worthwhile. This reflects both the difficulty of mounting trials of screening (which are costly, large scale, and of long duration) and the "culture" of evaluation of screening, which has been dominated by Wilson and Jungner\'s out-dated criteria that do not emphasise the fundamental importance of randomised trials in evaluation \[[@pmed-0020098-b13]\]. Screening, like any other medical technology, requires robust evaluation \[[@pmed-0020098-b14]\]. The evidence available from randomised controlled trials of the high risk approach is not encouraging. The impact on clinical events in community populations is very limited---the best estimate is of no real effect at all, although an effect as small as a 10% risk reduction may have been missed \[[@pmed-0020098-b15]\]. Since medications for reducing coronary disease morbidity and mortality are clearly effective, it is tempting to look for better screening tests for early disease, such as carotid ultrasound with plaque characterisation, fast computed tomography scanning of the coronary arteries, and genetic "SNPing" (looking for single nucleotide polymorphisms associated with coronary disease) \[[@pmed-0020098-b16]\]. But none of the new screening technologies currently recommended \[[@pmed-0020098-b17]\] have been adequately assessed in randomised controlled trials. Ultrasound IMT measures are capable of identifying people at increased risk. However, in evaluations of the receiver--operator curves comparing the accuracy of ultrasound measures with conventional risk factor scoring, no extra predictive power is obtained from ultrasound \[[@pmed-0020098-b18],[@pmed-0020098-b19]\]. This is not surprising as increased IMT and plaque are simply the downstream consequences of conventional risk factors operating over many years. Even if screening tools were highly accurate, high risk approaches can never match the health gains achieved by small downward movements in the distribution of risk factors. This is because the high risk group is (by definition) small relative to the rest of the population and consequently even though these people are at higher relative risk, the absolute numbers of coronary events will always be greater in the much larger "low risk" population. Of course, one way around this problem is to define the whole population as "high risk" and give everyone several medications (a "polypill") to lower their risk \[[@pmed-0020098-b20]\]. Estimates of how much can be achieved by treating all those at a 15% or greater risk of suffering a cardiovascular event over 10 years (about half the older adult population) with a combination of effective drugs, show that, at best, only half of the events would be prevented. By contrast, small downward movements in blood pressure and blood cholesterol of 10%--15% of the mean will reduce coronary heart disease events by the same amount \[[@pmed-0020098-b21]\]. Improving the accuracy of risk prediction with new screening technologies will not alter this relationship between high risk and population approaches. If cost is factored into any policy-making process, it is obvious that investing in widespread use of new screening technologies will, inevitably, be less cost-effective than implementing population policies. In privatised, individualised, health-care systems of the United States it is much more likely that new screening developments will arise without evaluation and will prove popular with naïve, but worried, well people. In these circumstances, the onus is on the scientific community to demand better standards of evidence, as many governments will not. Makover\'s Response to Ebrahim\'s Viewpoint {#s3} =========================================== While the "population" approach is obviously very important, there is no reason not to use the "high risk" approach as well. "High risk" is the wrong term---all arteries matter over a 120-year life span. We must attack atherosclerosis at the very onset, not wait until it is "high risk." All previous approaches were designed to detect those at allegedly higher risk by various criteria. This misses the point. Once plaque exists, risk exists, especially when viewed over a full lifetime, not the current ten-year standard. And although the aggressive approach I outlined in my viewpoint would obviously have a cost, the savings (a reduction in the financial toll of cardiovascular disease) would be greater than the cost. Ultrasonography of the carotid artery intimal wall is effective and highly predictive according to the great preponderance of studies \[[@pmed-0020098-b8]\]. Ebrahim cites two studies to support his argument that measuring carotid IMT offers no additional predictive power over conventional screening tools \[[@pmed-0020098-b18],[@pmed-0020098-b19]\], but the first of these has serious flaws (such as a significant dropout rate that the authors assumed was random, a short follow-up period, and the inclusion only of people of an advanced age) \[[@pmed-0020098-b18]\], and the other concluded that carotid IMT measurement "substantially improved prediction of future coronary heart disease" \[[@pmed-0020098-b19]\]. Millions of people are at risk now and cannot wait for endless studies to prove over and over what we already know. It is not "naïve" for well people to want to remain well and to maintain their arteries to enable healthy older lives. Nor should we disparage as "worried" the very human desire to remain that way. Population-wide planning is important, but doctors take care of each patient as an individual. Carotid ultrasonography is by far the best way to design treatment for every patient exactly as needed. I believe that there is already sufficient evidence and sensible rationale to meld already well-proven technologies and approaches into a comprehensive, aggressive attack on atherosclerosis. Ebrahim\'s Response to Makover\'s Viewpoint {#s4} =========================================== Declines in the risk of coronary heart disease have been reported in many countries that had experienced cardiovascular disease "epidemics" during the middle of the 20th century \[[@pmed-0020098-b22]\]. The onset of these beneficial trends pre-dated the introduction of both risk factor screening and widespread use of effective treatment of hypertension and hypercholesterolaemia. These downward trends in cardiovascular mortality have contributed to increases in life expectancy. In autopsies done during the Korean and Vietnam wars, atherosclerosis was found in 77% and 45% of young men, respectively \[[@pmed-0020098-b23],[@pmed-0020098-b24]\]. More recently, the origins of cardiovascular disease have been pushed back to fetal development and early childhood \[[@pmed-0020098-b25],[@pmed-0020098-b26]\]. Raised levels of risk factors are set in early life, and if earlier intervention means better outcomes, it is at this stage (i.e., in early life) that treatment and prevention must be targeted. Plaques are remarkably common in middle-aged people screened using ultrasound \[[@pmed-0020098-b27]\], but most are stable and unlikely to rupture. Presence of plaque in otherwise healthy people does not identify those who are destined to suffer an event or benefit from treatment. In the same way as risk factors provide a probabilistic evaluation of the chances of suffering a clinical event, so too does presence of plaque, with no greater accuracy than risk factor scoring methods \[[@pmed-0020098-b19]\]. Using carotid ultrasound screening to define a subgroup of those with plaque but normal or low levels of cardiovascular risk factors may be useful (see [Table 1](#pmed-0020098-t001){ref-type="table"}). However, studies have yet to be done to confirm whether plaque, in the absence of raised risk factors, is harmful. Importantly, trials are needed to determine whether treating such people aggressively---lowering their already low risk factors---results in better outcomes. It is possible that imaging of arteries may provide extra psychological incentives for patient adherence to lifestyle and drug regimens, but this too remains to be tested in well-designed randomised trials. ::: {#pmed-0020098-t001 .table-wrap} Table 1 ::: {.caption} ###### Treatment Decisions in Relationship to Carotid Ultrasound and Cardiovascular Risk Screening: Will Ultrasound Be of Value in Deciding Whether Some Low Risk Patients Should Be Treated? ::: ![](pmed.0020098.t001) ::: Citation: Makover ME, Ebrahim S (2005) What is the best strategy for reducing deaths from heart disease? PLoS Med 2(4): e98. IMT : intima-media thickness [^1]: Michael E. Makover is on the faculty of the New York University School of Medicine, New York, New York, United States of America, and is in private practice in internal medicine with a special interest in preventive medicine. E-mail: <mmakover@optonline.net> [^2]: Shah Ebrahim is a professor of epidemiology and ageing in the Department of Social Medicine, University of Bristol, United Kingdom. E-mail: <shah.ebrahim@bristol.ac.uk> [^3]: Competing Interests: The authors declare that no competing interests exist.	PubMed Central	2024-06-05T03:55:55.897235	2005-4-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1087207/", "journal": "PLoS Med. 2005 Apr 26; 2(4):e98", "authors": [ { "first": "Michael E", "last": "Makover" }, { "first": "Shah", "last": "Ebrahim" } ] }
PMC1087208	Introduction {#s1} ============ Advances in transplant-based therapy for type 1 diabetes mellitus \[[@pmed-0020103-b01]\] and a dearth of cadaveric pancreatic islets of Langerhans have focused interest on developing renewable sources of transplant-ready islet-replacement tissues. Pancreatic islets derive from embryonic endoderm, but display features of neurons, including a shared set of cell-autonomous developmental regulators \[[@pmed-0020103-b02]\]. Islets are the principal source of insulin in humans, but in some invertebrate species, such as Drosophila, brain neurons are the main source of circulating insulin \[[@pmed-0020103-b03],[@pmed-0020103-b04]\]. The similarities between islet cells and neurons are further underscored by the demonstration of insulin gene transcription in the vertebrate brain \[[@pmed-0020103-b05]\], although it remains unclear whether these vertebrate neurons produce or secrete insulin protein \[[@pmed-0020103-b06]\]. Moreover, recent studies also suggest that in some settings, mouse pancreatic epithelial cells can engender neuron-like cells \[[@pmed-0020103-b07],[@pmed-0020103-b08]\]. These and other findings \[[@pmed-0020103-b02]\] suggest that common, ancient developmental programs may govern the differentiation of islet cells and neurons. Methods promoting neural differentiation by embryonic stem (ES) cells have been adapted \[[@pmed-0020103-b09],[@pmed-0020103-b10],[@pmed-0020103-b11],[@pmed-0020103-b12],[@pmed-0020103-b13]\] to derive insulin-producing cells (IPCs), but it remains unclear whether such IPCs were derived from neural progenitors, or whether a neural-based strategy can generate transplantable human IPCs. Neural cells derive from multipotent progenitor cells, and recently, clonal human neural progenitor cells that remained multipotent and karyotypically stable through at least 15 passages were purified by flow-cytometry-based methods \[[@pmed-0020103-b14]\]. The availability of human neural progenitor cells allowed us to investigate whether inductive signals involved in normal pathways of islet development could direct these neural progenitors to develop into glucose-responsive IPCs. Here we describe experimental strategies for developing IPCs directly from human neural stem cells. Methods {#s2} ======= All studies were performed in accordance with appropriate human participants\' approval according to Stanford University Institutional Review Board guidelines. Cell Lines and Culture Conditions {#s2a} --------------------------------- The human neurosphere (NS) cell lines (line numbers 1651, 1664, and 1673) were established following enrichment for AC133^+^ cells, as previously described \[[@pmed-0020103-b14]\], although none of the cell lines was expanded from a single cell, and they are therefore not clonal. The different cell lines were from distinct donors and karyotypically normal. They gave comparable results, but line 1664 produced less insulin by stage 4 in our protocol than the other cell lines. Following cell thaw, NSs were expanded in suspension culture, and in each case were used between passages 5 and 7. Undifferentiated human NS cells (stage 1) were cultured and expanded in medium containing X-VIVO15 (Cambrex, Walkersville, Maryland, United States), N2 supplement (Invitrogen, Carlsbad, California, United States), 2 μg/ml heparin, 10 ng/ml leukemia inhibitory factor (Chemicon, Temecula, California, United States), 20 ng/ml epidermal growth factor (R&D Systems, Minneapolis, Minnesota, United States), and 20 ng/ml fibroblast growth factor-2 (R&D Systems), and they were frozen after enzymatic dissociation. At stage 2, we thawed and cultured 10^7^ cells in medium made from a 1:1 mixture of glucose-free DMEM (Invitrogen) and F-12 medium (Invitrogen) containing 0.2% bovine serum albumin (Sigma, St. Louis, Missouri, United States), N2 supplement, 2 μg/ml heparin, 10 ng/ml leukemia inhibitory factor, 20 ng/ml epidermal growth factor, and 20 ng/ml fibroblast growth factor-2. The final concentration of glucose in stage 2 was 5 mM. Cultures were fed weekly for up to 2 wk. Cells formed NSs during stages 1 and 2. After 2 wk, we transferred 40--80 NSs containing a total of approximately 2--4 × 10^5^ cells to single wells in plates coated with 0.005% poly-L-ornithine (Sigma) and 3 μg/ml fibronectin (Invitrogen). These were cultured for 2 wk (stage 3) in a 1:1 mixture of low-glucose DMEM and F-12 media containing 100 mg/l apo-transferrin (Sigma), 288 mg/l glucose, 73 mg/l [L]{.smallcaps}-glutamine, 1.69 g/l sodium bicarbonate, 25 μg/ml bovine insulin (Sigma), 20 nM progesterone, 100 μM putrescine, 30 nM sodium selenite, and penicillin/streptomycin. Stage 3 medium was supplemented with 2 μM all-trans retinoic acid (RA) (Sigma) and had a final glucose concentration of either 5 or 17 mM. At stage 4, cells were cultured in N2 medium \[[@pmed-0020103-b10]\] supplemented with 10 mM nicotinamide (Sigma) and 10 nM insulin-like growth factor-1 (R&D Systems) for up to 6 d. Stage 4 medium included 17.3 mM glucose. During stage 3 and 4, medium was changed every other day. To examine the effect of Sonic hedgehog (Shh) (R&D Systems), we added 300 nM mouse Shh in PBS every other day during stage 4. To quantify neurite outgrowths per cluster, 20 Shh-treated or control stage 4 IPC clusters were photographed under light microscopy, and digitized images were used to score neurites. Data are presented as the average ± the standard error of the mean. Two-tailed t tests were conducted to determine statistical significance. RT-PCR {#s2b} ------ Total RNA was prepared by using TRIZOL (Invitrogen) and RQ1 RNase-free DNase (Promega, Madison, Wisconsin, United States). For cDNA synthesis, random sequence primers were used to prime reverse transcription reactions and synthesis was carried out by Thermoscript RT (Invitrogen). A total of 35 cycles of PCR were performed using Platinum Taq High Fidelity DNA Polymerase (Invitrogen). GAPDH expression was used to normalize input template cDNA to analyze relative gene expression. Primer sequences and gene accession numbers are listed in [Table 1](#pmed-0020103-t001){ref-type="table"}. To confirm their identity, DNA products from PCR reactions were excised, cloned, and sequenced. ::: {#pmed-0020103-t001 .table-wrap} Table 1 ::: {.caption} ###### Specific PCR Primers and Gene Accession Numbers ::: ![](pmed.0020103.t001) ^a^ National Center for Biotechnology Information ([www.ncbi.nlm.nih.gov/](www.ncbi.nlm.nih.gov/)) accession numbers ::: Immunohistochemistry and Morphometry {#s2c} ------------------------------------ Cell clusters were fixed in 4% paraformaldehyde, embedded in HistoGel (Richard-Allan Scientific, Kalamazoo, Michigan, United States), and then embedded in paraffin. We performed immunohistochemistry on 6-μm tissue sections. We used primary antibodies at the following dilutions: guinea pig anti-insulin, 1:200 (Linco Research, St. Charles, Missouri, United States), mouse anti-β-tubulin III, 1:500 (Sigma); rabbit anti-C-peptide, 1:100 (Linco Research); mouse anti-Nestin, 1:200 (Chemicon); mouse anti-Ki67, 1:100 (Novocastra, Newcastle, United Kingdom); rabbit anti--cleaved caspase-3, 1:200 (Cell Signaling, Beverly, Massachusetts, United States); rabbit anti-glucagon, 1:200 (Dako, Carpinteria, California, United States); rabbit anti-Glut-2, 1:200 (ADI, San Antonio, Texas, United States); mouse anti-proinsulin, 1:500 (O. Madsen, Hagedorn, Gentofte, Denmark); mouse anti-glucokinase, 1:200 (C. Newgard, Duke University, Durham, North Carolina, United States); mouse anti-GFAP, 1:1,000 (Dako); mouse anti-MAP2, 1:500 (Sigma); and rabbit anti-Olig2, 1:1,000 (H. Takebayashi, National Institute for Physiological Sciences, Okazaki, Japan). Confocal immunofluorescence microscopy with an optical slice thickness of 0.6 μm was performed on a Bio-Rad (Hercules, California, United States) MRC1000. Cell counting and point-counting morphometry were performed using standard morphometric techniques \[[@pmed-0020103-b15]\]. To obtain representative results, all quantification of immunostaining was performed by counting numbers of positive-stained cells and dividing by the area of total tissues using a standard 10 × 10 microscope grid. For quantification of cells expressing Nestin, Ki67, β-tubulin III, activated caspase-3, insulin, or C-peptide, NS-derived cell clusters were fixed, and sectioned to generate 7-μm-thick tissue sections. Appropriately stained cells were counted in a minimum of ten random microscopic fields obtained from at least ten cell clusters per condition. In Situ Hybridization {#s2d} --------------------- We used human insulin cDNA (320 bp) cloned in plasmid pCR4-TOPO (Invitrogen) as a template for in vitro transcription to produce riboprobes with a digoxigenin-RNA labeling kit (Ambion, Austin, Texas, United States), performed hybridization with IPC cluster sections as described \[[@pmed-0020103-b16]\], and used DIG Nucleic Acid Detection Kit (Roche, Indianapolis, Indiana, United States) for development according to the manufacturer\'s instructions. Insulin C-Peptide Quantification and In Vitro Insulin Secretion Assay {#s2e} --------------------------------------------------------------------- At each stage, 300 NS or IPC clusters were handpicked, washed with PBS, and homogenized; then intracellular C-peptide content was measured with a C-peptide ELISA kit (American Laboratory Products Company, Windham, New Hampshire, United States). Stage 4 IPC clusters were cultured in glucose-free RPMI (Invitrogen) supplemented with 2.8 mM glucose, 20 mM HEPES, and 10% newborn calf serum (Invitrogen) for at least 2--3 h. At the end of these washes, no insulin was detectable in the wash supernatant using an ELISA kit (American Laboratory Products Company). Fifty IPC clusters were handpicked, transferred to a 24-well plate, and incubated for 2 h at 37 °C in the presence of 2.8 mM glucose, 25 mM glucose, or 25 mM sucrose. Supernatants were then harvested for enzyme-linked immunosorption assay (ELISA)--based quantification of released insulin. IPC Transplantation and Physiologic Tests {#s2f} ----------------------------------------- All animal studies were performed in accordance with Stanford University Animal Care and Use Guidelines. Under general anesthesia, 9- to 10-wk-old male NOD scid mice (purchased from Jackson Laboratories, Bar Harbor, Maine, United States) were engrafted with 1,000 handpicked IPC clusters in the right and left subcapsular renal space (500 IPC clusters each) or received a sham transplant of saline solution. Transplantation of more IPC clusters in a renal graft site was not feasible because of the size of the clusters. Two weeks after IPC cluster transplantation, we subjected mice to an overnight fast. Then we measured serum human C-peptide before and 30 min after intraperitoneal injection with 3 g of glucose per kilogram of body weight using a human C-peptide ELISA kit. When stage 4 IPC clusters were transplanted, tumors were not observed 4 wk after engraftment, the maximum period of observation (n = 5). All data represent the average (from the indicated number of samples) ± standard error of the mean. Two-tailed t tests were conducted to determine statistical significance. Results {#s3} ======= Glucose Restriction Initiates IPC Development {#s3a} --------------------------------------------- Culture media with high glucose concentrations have been useful for isolating and maintaining progenitor cell populations from the central and peripheral nervous systems \[[@pmed-0020103-b14],[@pmed-0020103-b17]\] and for permitting differentiation by neurogenic cells \[[@pmed-0020103-b18]\]. We routinely expanded NS cultures in medium containing 37 mM glucose, a level 5- to 6-fold higher than physiologic glucose concentration. Isolated human neural progenitors spontaneously aggregate in these conditions to produce a cell cluster called a NS, and 85%--90% of NS cells produce the intermediary filament Nestin ([Figures 1](#pmed-0020103-g001){ref-type="fig"} and [2](#pmed-0020103-g002){ref-type="fig"}), a putative marker of neural stem and progenitor cells \[[@pmed-0020103-b19]\]. Prolonged exposure to high glucose levels can also severely reduce insulin expression in pancreatic β-cells \[[@pmed-0020103-b20]\]. Thus, we postulated that glucose reduction might blunt neurogenic programs by NSs and promote alternate fates, including development of IPCs. Therefore, we switched NS culture conditions from high glucose (37 mM, "stage 1") to low glucose (5 mM, "stage 2"; see [Figure 1](#pmed-0020103-g001){ref-type="fig"}A) to initiate development. ::: {#pmed-0020103-g001 .fig} Figure 1 ::: {.caption} ###### Development of Cells Expressing Islet Markers from Human Neural Progenitor Cells \(A) Outline of four-stage differentiation protocol, essential factor manipulations at each stage, and stage-specific cell cluster morphology. Original magnification, 100×. \(B) RT-PCR analysis of gene expression by stages 1--4 cell clusters, in stage 4 samples with RT omitted (--RT) and in human fetal spinal cord as a control ("C"). \(C) RT-PCR analysis of expression during stages 1--4 of pancreatic and endodermal markers. Fetal pancreas RNA served as a control ("C"). ::: ![](pmed.0020103.g001) ::: ::: {#pmed-0020103-g002 .fig} Figure 2 ::: {.caption} ###### Immunohistochemical Detection of Neural Markers and Insulin during Stages 1--4 of Human Neural Progenitor Cell Differentiation Immunofluorescent images were obtained by confocal microscopy and are representative of at least ten samples for each antibody. We detected insulin in only stage 4 IPCs, consistent with RT-PCR results. Distribution of insulin staining is localized in the cytoplasm. Immunofluorescent detection of Ki67, a nuclear marker of proliferating cells, showed stage 4 IPCs are predominantly non-proliferating, similar to mature pancreatic islets. Original magnification, 630×. ::: ![](pmed.0020103.g002) ::: To assess developmental changes resulting from glucose reduction and later culture modifications, we examined NS cell composition, and expression of gene products known to regulate or define in vivo fates of embryonic neural or islet cells. A β-tubulin III isoform produced in differentiating neurons was not detected in stage 1 cells, whereas expression of Ki67, an S-phase-associated nuclear antigen, was expressed in the majority of stage 1 cells (see [Figure 2](#pmed-0020103-g002){ref-type="fig"}). Genes specifying transcription factors essential for in vivo differentiation of lineage-restricted neural progenitors and their differentiated progeny \[[@pmed-0020103-b21],[@pmed-0020103-b22],[@pmed-0020103-b23]\], including En1, Hb9, Isl1, Hoxc6, NRSF/REST, and Nkx6.1, were expressed at low or undetectable levels in stage 1 (see [Figure 1](#pmed-0020103-g001){ref-type="fig"}B--[1](#pmed-0020103-g001){ref-type="fig"}C), consistent with the reported multipotency of proliferating NSs grown in high glucose \[[@pmed-0020103-b14]\]. After 2 wk at stage 2, NS-derived cells remained in clusters (see [Figure 1](#pmed-0020103-g001){ref-type="fig"}A), but Nestin and Ki67 were expressed in fewer than 35% of cells (see [Figure 2](#pmed-0020103-g002){ref-type="fig"}), indicating possible differentiation of cells toward lineage-restricted fates. Consistent with this view, we detected increased expression of Nkx6.1, NRSF/REST, and β-tubulin III (see [Figures 1](#pmed-0020103-g001){ref-type="fig"}B and [2](#pmed-0020103-g002){ref-type="fig"}). We also detected increased expression of Neurogenin3 (ngn3) (see [Figure 1](#pmed-0020103-g001){ref-type="fig"}C), a transcription factor expressed in differentiating neurons, and required for development of pancreatic islet progenitor cells \[[@pmed-0020103-b24],[@pmed-0020103-b25],[@pmed-0020103-b26]\]. Nkx6.1 and ngn3 expression were never observed in NS cultures maintained in 37 mM glucose (see [Figure 1](#pmed-0020103-g001){ref-type="fig"}B; data not shown). These data suggest that glucose restriction initiates developmental programs known to promote differentiation of mammalian neural and neuroendocrine cells. As shown below, insulin expression in NS-derived cells required culture in low-glucose medium at stage 2, but we did not consistently detect insulin^+^ cells by immunohistochemistry until stage 4. RA Promotes IPC Development {#s3b} --------------------------- RA induces development of primitive endodermal cells from a subset of embryonal carcinoma cell lines, and is an endogenous signal that directs development of posterior organs like the pancreas from embryonic endoderm \[[@pmed-0020103-b27],[@pmed-0020103-b28]\]. In vivo exposure to RA is sufficient to induce ectopic development of insulin-expressing tissue in the anterior foregut \[[@pmed-0020103-b27]\]. Moreover, stage 2 NS-derived cells expressed RA receptors including RAR-α and RAR-γ (see [Figure 1](#pmed-0020103-g001){ref-type="fig"}C), indicating competence for RA signals. To investigate whether RA could stimulate development of NS-derived cells toward an IPC fate, we measured expression of Pdx1, Cdx1, and FoxA3, transcription factors that regulate gastrointestinal organ development \[[@pmed-0020103-b29],[@pmed-0020103-b30]\]. Pdx1, Cdx1, and FoxA3 expression was induced in "stage 3," when cell clusters were allowed to adhere to pre-coated wells at a density of 40--80 NSs/cm^2^ and cultured in low-glucose medium with 2 μM RA for 2 wk (see [Figure 1](#pmed-0020103-g001){ref-type="fig"}A and [1](#pmed-0020103-g001){ref-type="fig"}C). We then tested other concentrations of RA and other plating densities. NS cells exposed to 100 nM RA at 20, 40, or 80 NSs/cm^2^ and NS cells exposed to 2 μM RA at a plating density of 20 NSs/cm^2^ did not express these markers or insulin at a later stage ("stage 4"; [Figure 3](#pmed-0020103-g003){ref-type="fig"}A; data not shown). Thus, higher doses of RA induced markers of posterior fate in NS-derived cells, a dosage response like that seen during RA patterning of posterior gut development in vivo. Moreover, this response was sensitive to plating density, revealing that additional cell-non-autonomous factors may regulate this response. Mesoderm develops in close association with gastrointestinal endoderm both in vivo and in vitro \[[@pmed-0020103-b31],[@pmed-0020103-b32]\]. Thus, detection of gut markers like Pdx1 and FoxA3 in NS-derived cell cultures raised the possibility that mesodermal cells might co-develop. However, we did not detect expression of known markers of mesoderm formation, including brachyury, flk-1, myosin light chain-2, and β-globin, at any stage in our cultures (see [Figure 1](#pmed-0020103-g001){ref-type="fig"}B). By contrast, mesoderm development invariably accompanies endoderm development in embryoid bodies derived from human ES cells or embryonic germ cells \[[@pmed-0020103-b31],[@pmed-0020103-b32],[@pmed-0020103-b33]\]. In stage 3 cultures, Nestin and Ki67 expression was nearly extinguished (see [Figure 2](#pmed-0020103-g002){ref-type="fig"}), while expression of En1, ngn3, and Hb9 increased (see [Figure 1](#pmed-0020103-g001){ref-type="fig"}B and [1](#pmed-0020103-g001){ref-type="fig"}C). While their numbers increased slightly during stage 3, β-tubulin III^+^ cells composed fewer than 10% of cells in NS-derived clusters (see [Figure 2](#pmed-0020103-g002){ref-type="fig"}), suggesting that the majority of differentiated cells had acquired non-neural fates. Consistent with this interpretation, we also did not detect expression of homeodomain proteins Hoxc6 or Hoxb9 (see [Figure 1](#pmed-0020103-g001){ref-type="fig"}B; data not shown), which are induced by RA in spinal cord neuronal precursor cells \[[@pmed-0020103-b21]\]. ::: {#pmed-0020103-g003 .fig} Figure 3 ::: {.caption} ###### The Sequence of Glucose Concentration Changes, Cell Density, and Absence of Hh Signals Are Essential for Development of IPCs \(A) RT-PCR analysis of insulin gene expression for different glucose concentrations and cell densities during stages 2--4. High glucose is designated "H"; low glucose designated "L." Thus, maintenance of glucose at high concentration in stages 1--4 is abbreviated as "HHHH"; exposure to high glucose in stage 1, followed by reduction of glucose in stages 2--4 is abbreviated "HLLL"; and so forth. Cell clusters analyzed here were cultured at 40--80 clusters/cm^2^ except where indicated ("low density"). Fetal pancreas RNA served as a control ("C"). \(B) RT-PCR analysis of Hh signaling factors during stages 1--4 and in fetal spinal cord as a control ("C"). \(C) RT-PCR analysis of stage 4 IPC expression of Shh, Desert hedgehog (Dhh), and Indian hedgehog (Ihh). Control ("C") is a spinal cord sample. \(D) Neurite outgrowth in control stage 4 IPC clusters, and in stage 4 IPC clusters exposed to 300 nM Shh. Original magnification, 100×. \(E) Quantification of neurite outgrowth per control or Shh-treated stage 4 IPC cluster. \, P* \< 0.001. \(F) RT-PCR analysis of IPC cluster expression of Ptc, FoxA3, Pdx1, and insulin following treatment with Shh at stage 4. Control ("C") is fetal pancreas sample. ::: ![](pmed.0020103.g003) ::: RA can activate signaling through the Hedgehog (Hh) pathway, and recent studies show that Hh signals control development of embryonic pancreatic islets in vivo \[[@pmed-0020103-b16],[@pmed-0020103-b34]\]. To elucidate the mechanisms underlying RA-induced expression of markers like FoxA3 and Pdx1, we examined Hh signaling during IPC development. We detected expression of the Hedgehog receptors Patched (Ptc) and Smoothened (Smo) at all stages ([Figure 3](#pmed-0020103-g003){ref-type="fig"}B), consistent with the possibility that NS-derived cells are competent for Hh signaling. However, we did not detect NS expression of genes encoding Hh ligands like Shh ([Figure 3](#pmed-0020103-g003){ref-type="fig"}B), similar to previous studies \[[@pmed-0020103-b19]\]. Increased Ptc transcription is a known consequence of Shh signaling, but we did not detect changes of Ptc expression during stages 1--4, consistent with the absence of detectable expression of Hh ligands. To test directly whether the absence of Hh signaling was required for expression of endodermal or islet markers, including FoxA3, Pdx1, and insulin, we added Shh protein to stage 4 cultures, which do not express Shh or genes encoding other Hh ligands like Desert hedgehog or Indian hedgehog ([Figure 3](#pmed-0020103-g003){ref-type="fig"}C). At 300 nM, a dose used to induce neuronal development in ES cell cultures \[[@pmed-0020103-b21]\], Shh produced a significant increase in neurite outgrowth in NS-derived cells ([Figure 3](#pmed-0020103-g003){ref-type="fig"}D and [3](#pmed-0020103-g003){ref-type="fig"}E; see Methods), and resulted in increased Ptc expression ([Figure 3](#pmed-0020103-g003){ref-type="fig"}F). Moreover, Shh treatment during stage 4 eliminated expression of Pdx1, FoxA3, and insulin ([Figure 3](#pmed-0020103-g003){ref-type="fig"}F). These data are reminiscent of prior studies demonstrating that excess endodermal Shh signaling disrupts pancreas development in vivo \[[@pmed-0020103-b16],[@pmed-0020103-b34]\]. Thus, in contrast to the response of ES cells \[[@pmed-0020103-b21]\], RA treatment of NS-derived cells does not stimulate endogenous Hh signaling. The lack of Hh signals in NS-derived cells blunts neural differentiation and permits differentiation toward endocrine-like cell fates. Insulin Expression by IPCs {#s3c} -------------------------- ngn3, Pdx1, and Hb9 are essential factors for mammalian β-cell development \[[@pmed-0020103-b22]\], and their expression in stage 3 suggested the potential for deriving IPCs. Insulin expression has not, to our knowledge, been reported in NS-derived cells \[[@pmed-0020103-b19],[@pmed-0020103-b35]\]. Moderate hyperglycemia is a potent stimulus for β-cell differentiation and expansion in vivo and in utero \[[@pmed-0020103-b20]\], so we tested whether exposure of stage 3 cells to elevated glucose levels could induce insulin expression. Semi-quantitative RT-PCR measures of insulin mRNA expression showed that exposure of cultures grown in 5 mM glucose at stage 3 to 17 mM glucose at stage 4 induced insulin expression ([Figure 3](#pmed-0020103-g003){ref-type="fig"}A). Growth of cells at stage 2 in 5 mM glucose followed by exposure in stages 3 and 4 to 17 mM glucose also resulted in insulin expression at stage 4, but at lower levels. By contrast, no insulin was expressed by cultures maintained in 5 mM glucose or 17 mM glucose throughout stages 2--4 ([Figure 3](#pmed-0020103-g003){ref-type="fig"}A). Thus, reduction of glucose in stage 2 growth medium followed by a later increase in glucose at stage 3 or 4 was essential for development of IPCs. Insulin-like growth factor-1 may promote insulin secretion and prevent apoptosis of β-cells in vivo \[[@pmed-0020103-b36]\], and addition of this factor at stage 4 reduced apoptosis approximately 10-fold (see [Figure 1](#pmed-0020103-g001){ref-type="fig"}A; data not shown). Nicotinamide, which can stimulate in vivo pancreatic endocrine cell differentiation \[[@pmed-0020103-b37]\], did not affect cell survival, but did increase insulin C-peptide expression approximately 1.5- to 2-fold (data not shown). Thus, simultaneous exposure to high glucose levels, insulin-like growth factor-1, and nicotinamide optimized insulin expression at stage 4. Comparable results were obtained with three independently derived NS lines. Coinciding with the onset of high insulin expression levels in mouse islets, ngn3 expression is virtually extinguished \[[@pmed-0020103-b25]\], while expression of Pdx1, FoxA3, Isl1, and Nkx6.1 in islet β-cells is maintained or increased \[[@pmed-0020103-b22]\]. Similarly, we noted that IPC maturation from stage 3 to 4 was accompanied by increased insulin expression, reduced ngn3 expression, and increased or maintained levels of Isl1, FoxA3, and Pdx1 expression (see [Figure 1](#pmed-0020103-g001){ref-type="fig"}B and [1](#pmed-0020103-g001){ref-type="fig"}C). Thus, tissues derived from neural progenitor cells express insulin, Pdx1, and FoxA3, markers typically co-expressed in foregut-derived tissues. Expression of these markers suggests differentiation of some neural progenitor cells toward an endoderm-like fate. We were unable to detect expression of Pdx1 and FoxA3 using immunohistochemical methods in stage 4 IPCs, (data not shown). We also did not detect expression of Nkx6.1 or islet amyloid polypeptide, markers of mature pancreatic β-cells (see [Figure 1](#pmed-0020103-g001){ref-type="fig"}C; data not shown). Thus, the sequence of gene expression accompanying formation of IPCs from NSs was similar but not identical to that described for differentiating pancreatic β-cells, and further molecular studies are required to determine the degree of endoderm-like differentiation of cells in IPCs. We detected expression of insulin protein by immunohistochemistry in 90% of stage 4 cell clusters ([Figure 4](#pmed-0020103-g004){ref-type="fig"}). An average of 26% of cells in stage 4 clusters were insulin^+^, and appeared healthy, with abundant cytoplasm and a well-defined nucleus delineated by the nuclear stain 7AAD ([Figure 4](#pmed-0020103-g004){ref-type="fig"}A). Of the cells composing stage 4 clusters, 45% were β-tubulin III^+^, none of which contained insulin ([Figure 5](#pmed-0020103-g005){ref-type="fig"}A--[5](#pmed-0020103-g005){ref-type="fig"}C). Insulin^+^ cells showed little or no co-expression of neural stem cell markers like Nestin (see [Figure 1](#pmed-0020103-g001){ref-type="fig"}), or other markers of neural lineages known to develop from human neural progenitors, like MAP2 (differentiated neurons; [Figure 5](#pmed-0020103-g005){ref-type="fig"}D--[5](#pmed-0020103-g005){ref-type="fig"}F), Olig2 (multipotent precursors, bipotent glial precursors, and oligodendrocytes; [Figure 5](#pmed-0020103-g005){ref-type="fig"}J--[5](#pmed-0020103-g005){ref-type="fig"}L; see \[[@pmed-0020103-b38]\]), or myelin basic protein (data not shown). Occasional rare insulin^+^ cells expressed detectable GFAP (a marker of astrocytes; [Figure 5](#pmed-0020103-g005){ref-type="fig"}G--[5](#pmed-0020103-g005){ref-type="fig"}I). Thus, while we cannot completely rule out the possibility that some neural cell types expressed insulin, these data suggest that nearly all insulin^+^ cells produced in stage 4 cultures were non-neuronal. ::: {#pmed-0020103-g004 .fig} Figure 4 ::: {.caption} ###### Stage 4 IPCs Express Characteristic Pancreatic β-Cell Markers and Are Glucose-Responsive (A--C) Immunofluorescent images of stage 4 IPCs were obtained by confocal microscopy and are representative of at least ten samples for each probe. Shown is simultaneous immunofluorescent staining of insulin and either the nuclear marker 7AAD (A) or Ki67 (B). (C) Lack of immunostaining upon omission of the anti-insulin primary antibody. (D--G) Detection of insulin C-peptide in cells stained by anti-insulin antibodies. There is complete overlap between immunostained C-peptide and insulin in the cytoplasm. (H and I) Representative in situ hybridization with antisense riboprobes specific for human insulin on sectioned human fetal pancreatic islet cells (H) or a stage 4 IPC cluster (I). \(J) Sense control probe in situ hybridization on a stage 4 IPC cluster. (K--O) Simultaneous immunofluorescent detection of insulin and cleaved caspase-3 (K), glucagon (L), Glut-2 (M), glucokinase (N), or proinsulin (O). \(P) Intracellular C-peptide content of IPCs during stages 1--4 by human C-peptide-specific ELISA. \(Q) In vitro secretion assay of stage 4 IPCs. Insulin release followed a step increase from 2.8 mM to 25 mM glucose, but not following exposure to 25 mM sucrose, which served as an osmotic control. \, p* \< 0.001. \(R) Serum human C-peptide detection 2-wk after sham or IPC cluster transplantation in NOD scid mice. Samples were obtained before or 30 min after intraperitoneal glucose challenge (IPGTT). \(S) Lack of tumor growth after renal transplantation of stage 4 IPCs (graft indicated by white circle). \(T) Immunohistochemical detection of human C-peptide expression in stage 4 IPCs (cells stained brown) 2-wk after engraftment; bar is 20 μm. Original magnification of (H--J) and (T) was 250×. All other photomicrographs\' original magnification was 630×. ::: ![](pmed.0020103.g004) ::: ::: {#pmed-0020103-g005 .fig} Figure 5 ::: {.caption} ###### Stage 4 IPCs Do Not Express Markers of Differentiated Neural Cells (A--C) Immunohistochemical detection of β-tubulin III (A) and insulin (B) in stage 4 cells, and a merge of both images (C). (D--F) Immunohistochemical detection of MAP2 (D) and insulin (E) in stage 4 cells, and a merge of both images (F). (G--I) Immunohistochemical detection of GFAP (G) and insulin (H) in stage 4 cells, and a merge of both images (I). (J--L) Immunohistochemical detction of Olig2 at stage 1 (J) and stage 4 (K) and merged view of Olig2^+^ and insulin^+^ cells at stage 4 (L). Original magnification was 630×. ::: ![](pmed.0020103.g005) ::: Nuclei in insulin^+^ cells were not stained by an antibody to Ki67 (see [Figures 2](#pmed-0020103-g002){ref-type="fig"} and [4](#pmed-0020103-g004){ref-type="fig"}B), suggesting that insulin^+^ cells at stage 4 are post-mitotic, like mature pancreatic β-cells. Immunostaining with specific antibodies revealed proinsulin and C-peptide, an internal portion of the proinsulin translation product, in the cytoplasm of all insulin^+^ cells (see [Figure 4](#pmed-0020103-g004){ref-type="fig"}D--[4](#pmed-0020103-g004){ref-type="fig"}G and 4O), supporting our conclusion that insulin protein was produced in stage 4 IPCs. In situ hybridization with antisense riboprobes specific for human insulin mRNA labeled 25%--30% of stage 4 cells (see [Figure 4](#pmed-0020103-g004){ref-type="fig"}H--[4](#pmed-0020103-g004){ref-type="fig"}J), providing further evidence of insulin production. We observed only rare insulin^+^ cells stained by antibodies specific for activated caspase-3 (\<0.3%; see [Figure 4](#pmed-0020103-g004){ref-type="fig"}K) or TUNEL assay (data not shown). In stage 4 cell clusters we did not consistently detect expression of glucagon (see [Figure 4](#pmed-0020103-g004){ref-type="fig"}L), pancreatic polypeptide, or somatostatin by immunohistochemistry. The absence of glucagon gene expression in our IPC cultures was further confirmed by RT-PCR (data not shown). Thus, like in pancreatic islets, insulin was the principal hormone produced by IPCs. However, the composition of IPCs was distinct from that of pancreatic islets, and further studies are required to determine the basis for this difference. To quantify insulin expression in IPCs, we used a human insulin C-peptide-specific ELISA (see [Figure 4](#pmed-0020103-g004){ref-type="fig"}P). By ELISA, we did not detect C-peptide in stage 3 or stage 4 culture media, which contained supplementary bovine insulin. Thus, use of a C-peptide assay did not detect medium-derived bovine insulin in measures of IPC insulin content \[[@pmed-0020103-b39]\]. C-peptide levels at stages 1--3 were low or undetectable (see [Figure 4](#pmed-0020103-g004){ref-type="fig"}P), consistent with the lack of immunostainable insulin (see [Figure 2](#pmed-0020103-g002){ref-type="fig"}) in cells at these earlier stages. Similarly, in cultures maintained in 17 mM glucose during stages 1--4 ("HHHH"), C-peptide levels were not detectable (see [Figure 4](#pmed-0020103-g004){ref-type="fig"}P). In cell clusters exposed to 5 mM glucose in stages 2--3 then switched to 17 mM glucose at stage 4 ("HLLH"), insulin C-peptide levels were 0.36 fmol per cluster, and the average number of C-peptide^+^ cells per cluster was 200. Thus, we estimate 0.0018 fmol C-peptide per NS-derived IPC. A single β-cell contains approximately 0.6 fmol C-peptide and insulin \[[@pmed-0020103-b40]\], so we calculate that insulin C-peptide content in one stage 4 IPC is approximately 0.3% of the level in isolated human β-cells. Collectively, these experiments confirm that IPCs transcribe and translate insulin, and rule out that insulin measured in IPCs is derived from medium, as shown in another system \[[@pmed-0020103-b39]\]. IPCs Are Glucose Responsive {#s3d} --------------------------- Glucokinase and glucose transporters like Glut-2 are essential regulators of glucose responses in pancreatic β-cells, and expression of these regulators in stage 4 insulin^+^ cells (see [Figures 1](#pmed-0020103-g001){ref-type="fig"}C, [4](#pmed-0020103-g004){ref-type="fig"}M, and 4N) suggested that IPCs could sense and respond to glucose. To test whether IPCs respond appropriately to glucose stimulation, we measured insulin release in static batch in vitro assays. We found that IPC insulin release increased markedly following a step increase of glucose from 2.8 to 25 mM (see [Figure 4](#pmed-0020103-g004){ref-type="fig"}Q). To examine whether IPCs are similarly responsive to glucose stimulation in vivo, we performed a series of IPC grafting experiments in immunocompromised recipient mice. Two weeks following engraftment (1,000 IPC-containing clusters/mouse), recipient mice and sham-transplanted controls were fasted overnight, and human C-peptide levels measured. Human C-peptide was undetectable in mouse sera prior to glucose challenge, or following glucose challenge of sham-transplanted controls (see [Figure 4](#pmed-0020103-g004){ref-type="fig"}R). In contrast, 30 min after intraperitoneal glucose challenge of IPC recipients we detected 6.1 ± 1.2 pmol/l human C-peptide (p \< 0.001), approximately 0.5%--1% of serum C-peptide levels observed following engraftment of 2,000 human islet equivalents in a prior study \[[@pmed-0020103-b41]\]. Thus, IPCs respond to glucose challenge in vivo by releasing insulin C-peptide. Analysis of IPC graft sites 4 wk after transplantation revealed nests of transplanted cells without obvious tumor formation. Histologic analysis and immunohistochemical detection revealed that C-peptide-expressing cells persisted in the graft site (see [Figure 4](#pmed-0020103-g004){ref-type="fig"}S and [4](#pmed-0020103-g004){ref-type="fig"}T). Thus, engrafted IPCs remained differentiated and survived up to a month following transplantation. Discussion {#s4} ========== There is widespread interest in developing tissue replacement strategies for treatment of human disorders like diabetes mellitus and Parkinson disease. While much attention has been focused on the promise of ES cells for tissue replacement, recent work suggests that neural stem cells, like ES cells, may have an unusually broad differentiation potential. For example, Gage and colleagues demonstrated the conversion of mouse neural stem cells to the endothelial lineage, indicating that plasticity is a bona fide property of cultured neural stem cells \[[@pmed-0020103-b42]\]. These results were also unexpected because endothelial cells and neuronal cells normally derive, respectively, from mesoderm and ectoderm, distinct embryonic germ layers. Here we show that human neural stem cells have a similarly broad differentiation potential, and that specific in vitro culture conditions can divert neural stem-derived cells from neural lineages toward a fate with endocrine and endodermal characteristics. Our study shows that application of endogenous signals governing pancreas development to human neural progenitor cells can generate glucose-responsive IPCs. Systematic variation of the identity, concentration, and sequence of these signals led to discovery of methods culminating in IPC formation. However, there are limitations to our findings that should be noted. First, our molecular analysis of differentiating IPCs showed that they remain distinct from mature pancreatic islet β-cells; in other words, the insulin^+^ cells derived by our methods are not mature β-cells. For example, reverse transcriptase polymerase chain reaction confirmed that IPCs were enriched for a combination of gene products that approximate those expressed in developing pancreatic islet cells. However, the temporal sequence of expression of some of these products, like glucokinase, Glut-2, and Pdx1 does not precisely recapitulate that observed in the embryonic pancreas (reviewed in \[[@pmed-0020103-b22]\]), and transcription of other genes typically expressed in β-cells, like Nkx6.1, was not detected in later stage IPCs (see [Figure 1](#pmed-0020103-g001){ref-type="fig"}). Thus, the insulin^+^ cells derived in this study are not bona fide β-cells. Further experimentation may elucidate the basis for these differences, revealing the extent of neural progenitor cell development toward an endocrine cell fate. Recent studies suggest that neural cells can be derived from adult pancreatic epithelium \[[@pmed-0020103-b07],[@pmed-0020103-b08]\], adding to a growing body of data demonstrating numerous similarities in neural and pancreatic endocrine development (reviewed in \[[@pmed-0020103-b43]\]). In light of these similarities, we speculate that methods leading to the production of β-cell factors like Nkx6.1 in stage 4 clusters may enhance the β-cell-like qualities of neural-progenitor-derived insulin^+^ cells, and thereby permit differentiation of tissues that more closely resemble endoderm-derived islet cells. Second, our immunohistochemical analysis shows that stage 4 insulin-expressing cells do not express markers like Nestin, β-tubulin III, MAP2, or Olig2, suggesting that these insulin^+^ cells are non-neuronal, but we cannot formally rule out the alternate possibility that our methods produced neural cells capable of secreting insulin. Third, we have not demonstrated fully that the stimulus--secretion coupling apparatus in IPCs is similar to that in pancreatic islets. While insulin release by IPCs produced here appears to be glucose-sensitive, both in vitro and in vivo, future work should elaborate whether IPC insulin is stored in dense-core secretory vesicles, and whether secretogogues or secretion potentiators other than glucose (like amino acids or sulfonylureas) stimulate insulin release by IPCs, like in pancreatic islets. Lastly, we have not ameliorated glucose regulation in diabetic animal models with the human IPCs described here. Serum levels of insulin C-peptide expression achieved following IPC transplantation in glucose-challenged mice were less than 1% of normal (see [Figure 4](#pmed-0020103-g004){ref-type="fig"}). Thus, we did not attempt IPC transplantations in overtly diabetic animals, since prior studies suggest that insulin production at 10% of normal (or greater) may be required to improve glucose regulation in diabetic patients and animal models \[[@pmed-0020103-b10],[@pmed-0020103-b41]\]. Additional studies to test the impact of transplanted IPCs in animal models of diabetes therefore await production of IPCs capable of insulin secretion at levels higher than achieved here. Nevertheless, compared to other methods for IPC development from human stem cells \[[@pmed-0020103-b13],[@pmed-0020103-b44]\], our methods produced insulin at the highest levels yet achieved from an expandable, human stem-cell-derived tissue. Multiple experimental approaches were taken to demonstrate that these IPCs transcribe, translate, and secrete insulin, and to rule out the possibility that insulin measured in IPCs derived from the culture media. In vitro studies demonstrate that IPCs release insulin in a glucose-responsive manner, like islets. Moreover, IPCs transplanted into mice remained differentiated and released circulating human insulin in a glucose-dependent manner. Thus, for the first time, we demonstrate moderately efficient production of glucose-responsive IPCs from an expandable population of human stem cells. Current islet transplantation methods require an estimated 5 × 10^8^ to 10^9^ β-cells per recipient \[[@pmed-0020103-b45]\]. From 10^6^ cells at stage 2, we produced an average of 200--400 clusters with approximately 2,000 cells per cluster; approximately 25% of these cells are C-peptide^+^ and release 0.5%--1% of C-peptide secreted by β-cells. Based strictly on these yields, we would need to expand IPC production approximately 10^5^-fold to meet the need for one transplantation. Undifferentiated NSs can be readily expanded through at least a dozen passages \[[@pmed-0020103-b14]\], suggesting sufficient cell numbers could be generated to "scale up" this protocol for transplant-based therapies. We did not exhaust all possible factor combinations in our study: we speculate that further method refinements may improve the efficiency of NS conversion into IPCs, as well as IPC insulin synthesis and stimulus--secretion coupling, the hallmark functions of mature β-cells. If so, then human neural stem cells may serve as a valuable model for elucidating the mechanisms and factors that regulate neuroendocrine cell differentiation. For instance, addition of glucagon-like peptide-1, TGF-β ligands, or other factors that potentiate β-cell maturation, growth, and insulin secretion \[[@pmed-0020103-b46],[@pmed-0020103-b47],[@pmed-0020103-b48]\] may improve the methods described here. Because our method is based solely on extracellular factor modulation, in the absence of genetic manipulations, it could serve as the basis for developing replacement islets from a wide range of human stem cells, including neural stem cells and ES cells. Supporting Information {#s5} ====================== Accession Numbers {#s5a1} ----------------- The National Center for Biotechnology Information ([www.ncbi.nlm.nih.gov/](www.ncbi.nlm.nih.gov/)) accession numbers for the genes and gene products discussed in this paper are brachyury (AF012130), Cdx1 (NM\_000209), Desert hedgehog (NM\_021044), En1 (NM\_001426), flk-1 (AF035121), FoxA3 (NM\_004497), GAPDH (NM\_002046), GFAP (NM\_002055), glucokinase (M90299), Hb9 (NM\_005515), Hoxb9 (NM\_024017), Hoxc6 (NM\_004503), human insulin cDNA (J00265), Indian hedgehog (NM\_002181), Isl1 (NM\_002202), MAP2 (U01828), myelin basic protein (M13577), myosin light chain-2 (X57542), Nestin (NM\_006617), ngn3 (AF234829), Nkx6.1 (NM\_006168), NRSF/REST (U13879), Olig2 (NM\_005806), Pdx1 (NM\_000209), Ptc (U59464), RAR-α (NM\_000964), RAR-β (BC060794), RAR-γ (NM\_000966), Shh (L38518), Smo (AH007453), β-globin (NM\_000518), and β-tubulin III (BC000748). ### Patient Summary {#sb1} #### {#sb1a} ##### Background {#sb1a1} The transplantation of insulin-secreting cells has proved effective in treating severe cases of type 1 diabetes, but currently such cells have to be collected from donors after death. The process of isolation is very laborious and does not provide sufficient numbers of cells to permit wider use of this transplant-based treatment. A renewable source of these cells would be very valuable. Previous work has shown that some nerve cells can also produce insulin. ##### What Did the Researchers Do? {#sb1a2} They took stem cells that had come from human brains and treated them with a variety of different substances, including glucose, to see if they could alter the way the cells normally develop. They managed to manipulate the stem cells so that they developed into more mature cells that secreted insulin, and could respond to glucose. They put the cells into mice, and found that there, too, they could respond to glucose by producing insulin, albeit at very low levels. ##### What Do the Results Mean for Patients? {#sb1a3} These findings are a first step towards producing a renewable source of insulin-producing cells; however, the amount of insulin produced was quite low. In addition much more work will need to be done to ensure the safety of the procedure over the long term. ##### Where Can I Get More Information? {#sb1a4} The National Diabetes Information Clearinghouse, part of the National Institute of Diabetes and Digestive and Kidney Diseases has an information page on islet cell transplantation: <http://diabetes.niddk.nih.gov/dm/pubs/pancreaticislet/> Diabetes UK, a charity, also has information: <http://www.diabetes.org.uk/islets/> We thank Graeme McLean, Ursula Ehmann, and Dongping He for expert technical assistance, and StemCells for providing human neural progenitor cell lines. Drs. Irving L. Weissman, Nobuko Uchida, Ann Tsukamoto, Matthias Hebrok, James Wells, and Julie Baker, and Mr. Jeremy Heit provided valuable advice and comments on the manuscript. YH is a Juvenile Diabetes Research Foundation (JDRF) physician postdoctoral fellow. This work was supported by grants from the JDRF, Riva Foundation, Hillblom Foundation, and Programs in Molecular and Genetic Medicine and Interdisciplinary Translational Research at Stanford University. The funders of this work had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Citation: Hori Y, Gu X, Xie X, Kim SK (2005) Differentiation of insulin-producing cells from human neural progenitor cells. PLoS Med 2(4): e103. ELISA : enzyme-linked immunosorption assay ES : embryonic stem Hh : Hedgehog IPC : insulin-producing cell NS : neurosphere RA : retinoic acid Shh : Sonic hedgehog [^1]: Competing Interests: SK has a consulting relationship with and owns shares of the company StemCells. [^2]: Author Contributions: YH, XG, XX, and SKK designed the study. YH, XG, and XX performed the experiments. YH and SKK contributed to writing the paper. [^3]: ¤ Current address: Department of Gastroenterological Surgery, 21st Century COE Program, Kobe University Graduate School of Medicine, Kobe, Japan	PubMed Central	2024-06-05T03:55:55.898925	2005-4-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1087208/", "journal": "PLoS Med. 2005 Apr 26; 2(4):e103", "authors": [ { "first": "Yuichi", "last": "Hori" }, { "first": "Xueying", "last": "Gu" }, { "first": "Xiaodong", "last": "Xie" }, { "first": "Seung K", "last": "Kim" } ] }
PMC1087209	Filariasis {#s1} ========== The filarial nematodes Wuchereria bancrofti, Brugia spp., and Onchocerca volvulus are insect-borne parasites that cause lymphatic or cutaneous filariasis. Lymphangitis, hydrocele, and elephantiasis are pathologies that result from W. bancrofti and Brugia spp. infections. O. volvulus infections can present with severe skin pathologies (acute and chronic dermatitis, atrophy) and blindness (onchocerciasis, or river blindness). The nematodes infect more than 140 million people in 90 mostly tropical countries. An additional one billion people are at risk of contracting the diseases caused by these nematodes \[[@pmed-0020110-b1],[@pmed-0020110-b2]\]. Current control efforts, both vector control and mass antifilarial chemotherapies, have shown initial success, but sustainability is uncertain. For example, vector control, used for onchocerciasis in west African savannah areas, has become too expensive and therefore been stopped. Current and planned elimination programs will rely on mass administration of antifilarial drugs that require annual administration for 5--10 years for lymphatic filariasis and more than 20 years for onchocerciasis. Control programs based on drug administration require long treatment durations because the adult female worms, which produce thousands of larvae daily, survive many years (over 14 years for onchocerciasis) and are not killed by current drugs \[[@pmed-0020110-b1]\]. The success of such control programs, as shown by mathematical modeling, depends on a minimum of 60% of the people in an endemic area participating every year \[[@pmed-0020110-b3]\]. This is probably too optimistic an estimate, as a recent review of onchocerciasis therapy in regions that have had 10--12 years of ivermectin treatment still show infection levels of 2%--3%. These levels are enough to establish the infection within a few years after the end of annual drug administration \[[@pmed-0020110-b4]\]. Additionally, there is evidence that some geographic areas have worms with apparent resistance to ivermectin \[[@pmed-0020110-b5]\]. Therefore, it is essential that we find new drugs that kill or sterilize adult worms. Enter Wolbachia---Endosymbionts of Filarial Nematodes {#s2} ===================================================== For almost 30 years, it has been known that filarial nematodes contain endosymbiotic bacteria. These endobacteria are found in the hypodermis of male and female worms, and in the oocytes, embryos, and larval stages ([Figure 1](#pmed-0020110-g001){ref-type="fig"}). As in many animal filarial species, endobacteria are present in the human filariae W. bancrofti, Brugia spp., and O. volvulus \[[@pmed-0020110-b1],[@pmed-0020110-b6],[@pmed-0020110-b7]\], but not in Loa loa \[[@pmed-0020110-b8],[@pmed-0020110-b9]\]. Recently, these endosymbionts were classified at the molecular level to be of the genus Wolbachia, a genus of bacteria that are common endosymbionts of arthropods. The next nearest relatives of Wolbachia are Rickettsia, Ehrlichia, Cowdria, and Anaplasma \[[@pmed-0020110-b10]\]. ::: {#pmed-0020110-g001 .fig} Figure 1 ::: {.caption} ###### Cross-Section of a Female Worm from an Extirpated Nodule from a Patient with Onchocerciasis Wolbachia, endosymbiotic bacteria of filarial nematodes important for embryo development, are targets for antifilarial treatment. The endobacteria cause inflammation and contribute to the pathology of lymphatic filariasis and onchocerciasis. The bacteria here were stained with antibodies against Yersinia Hsp60. The bacteria are located in the hypodermis and oocytes of the worm (arrows). h, hypodermis; i, intestine; m, muscle; u, uterine tube. (Photo: D.W. Büttner, Bernhard Nocht Institute for Tropical Medicine, Hamburg, Germany) ::: ![](pmed.0020110.g001) ::: Studies of the effect of antirickettsials such as tetracycline and rifampicin in animals infected with filarial nematodes have shown, using immunohistochemistry, that these drugs deplete the Wolbachia from the worms. After the Wolbachia are depleted, the worms develop a distinct phenotype. Monitoring the microfilaria (Mf) levels in the blood showed that the number of Mf in the treated animals was lower than the number in the control groups, and that the number of Mf in the treated group neared zero with time. Examination of the adult worms showed that embryogenesis was blocked and the uteri contained degenerated embryos \[[@pmed-0020110-b1]\]. A study done in cattle infected with O. ochengi even showed killing of adult worms \[[@pmed-0020110-b11]\]. The success of antibiotic treatment against animal filariae has been extended to human filarial infections. Trials of doxycycline have been completed for populations infected with O. volvulus and W. bancrofti. For both species, larval levels were zero or near zero after treatment \[[@pmed-0020110-b6],[@pmed-0020110-b12]\]. Examination of worms from extirpated nodules of patients with onchocerciasis showed that, as in the animal studies, depletion of Wolbachia led to a block in embryogenesis that appears to be permanent \[[@pmed-0020110-b2]\]. Most recently, there has been evidence for the killing of adult W. bancrofti by doxycycline therapy \[[@pmed-0020110-b13]\]. Wolbachia stimulate inflammatory responses via Toll-like receptor (TLR) 2 and TLR4 \[[@pmed-0020110-b14]\], pattern recognition receptors that recognize a variety of bacterial molecules \[[@pmed-0020110-b15]\]. Working with a mouse model of O. volvulus--induced corneal inflammation (keratitis), it has been shown that Wolbachia antigens presented via TLR4 are required for the development of pathology \[[@pmed-0020110-b16]\]. Wolbachia have also been associated with adverse reactions seen in infected patients after antifilarial therapy. Recent studies supporting a role for Wolbachia in adverse reactions after antifilarial treatment have shown that doxycycline given before ivermectin reduced Wolbachia loads as well as the number and severity of adverse reactions in patients with lymphatic filariasis (\[[@pmed-0020110-b17]\]; J. Turner, S. Mand, A. Y. Debrah, J. Muehlfeld J, K. M. Pfarr, et al., unpublished data). Thus, doxycycline fulfills the role for a new antifilarial therapy in that it produces sterility in lymphatic filariasis and onchocerciasis, kills adult worms in lymphatic filariasis, and prevents or lessens adverse reactions due to the rapid killing of Mf by microfilaricidal drugs. However, the treatment time of four weeks is still longer than that desired for new antifilarial therapies. Wolbachia are ideal targets for antifilarial drugs that have the same effect as doxycycline, but that work in a shorter interval. Potential drug targets may be found by analyzing the genome of Wolbachia. Sequencing the Wolbachia Genome {#s3} =============================== As part of the effort to find antiwolbachial drugs that act in less time than the current four-week regime for doxycycline, a consortium was established to sequence the genomes of the Wolbachia species that inhabit human filarial nematodes. In the April 2005 issue of PLoS Biology, Foster and colleagues report on the completion of the sequencing and annotation of the genome from the Wolbachia of Brugia malayi (wBm) ([Figure 2](#pmed-0020110-g002){ref-type="fig"}) \[[@pmed-0020110-b18]\]. This is the first complete Wolbachia genome from a filarial nematode. The authors compare the wBm genome to the first sequenced genome of the Wolbachia of Drosophila melanogaster (wMel), which is strictly parasitic, and the genomes of other endosymbiotic bacteria, pointing out potential metabolites that wBm may supply to the nematode. ::: {#pmed-0020110-g002 .fig} Figure 2 ::: {.caption} ###### Annotation of the Complete Wolbachia Genome (Figure from \[[@pmed-0020110-b18]\]) ::: ![](pmed.0020110.g002) ::: Features of the Genome and Metabolites that wBm May Provide to Its Host {#s4} ========================================================================= The genome of wBm is 1.1 million nucleotides, smaller than wMel and Rickettsia prowazekii, but larger than that of Mycobacterium. The reduced genome size is indicative of long-term symbiosis, and reflects the loss of all genes necessary to make all but one amino acid, genes needed to infect new hosts, and genes needed to evade the host immune system. Also lacking from the wBm genome are several genes needed for DNA repair and genes required for RNA modification. Many of these genes have also been lost from other endosymbiotic bacteria. wBm cannot synthesize lipopolysaccharide, a component of the cell membrane in most bacteria. This is astounding because, as noted above, Wolbachia stimulate an inflammatory reaction via TLR4. Endobacteria antigens presented via TLR4 are also responsible for adverse reactions after antifilarial treatment \[[@pmed-0020110-b17]\], and for pathology in onchocerciasis \[[@pmed-0020110-b16]\]. wBm lack the genes necessary to cross-link and degrade the carbohydrate backbone of peptidoglycan. The Wolbachia peptidoglycan structure resembles the peptidoglycan-derived cytotoxins produced by Neisseria gonorrhoeae and Bordetella pertussis, which are made up of muramyl peptides \[[@pmed-0020110-b19]\]. These muramyl peptides are known to stimulate an inflammatory response and pathology via TLR2. > The completion of the wBm genome offers a wealth of information for both basic and applied science. However, to understand the Wolbachia--filaria endosymbiosis, it is not only important what wBm has lost from the genome during its long symbiosis with B. malayi, but also what has been kept. The endobacteria have retained all of the genes necessary for the synthesis of the co-factors riboflavin, flavin adenine dinucleotide, and heme. The genome also contains the genes necessary to make glutathione, although it lacks the genes needed for glutathione-mediated methylglyoxal detoxification \[[@pmed-0020110-b20]\], which is the most common reason intracellular bacteria need glutathione. Finally, in contrast to most endosymbiotic bacteria, wBm have retained the genes necessary to make all nucleotides. As the wBm genome has the genes for a type IV secretion system, used by intracellular bacteria for exporting molecules to nonbacterial (e.g., host) cells, the above described metabolites could be supplied to the nematode host by Wolbachia. To date, there is no evidence of genes for riboflavin and heme synthesis in the B. malayi genome \[[@pmed-0020110-b21]\]. Heme from Wolbachia could be vital to worm embryogenesis, as there is evidence that molting and reproduction are controlled by ecdysteroid-like homones \[[@pmed-0020110-b22]\], whose synthesis requires heme. Depletion of Wolbachia would therefore halt production of these hormones and block embryogenesis. Wolbachia could be a source of glutathione which, aside from its role in the detoxification of methylglyoxal, is required for protection against oxidative stress from oxygen-reactive species secreted by mammalian immune cells \[[@pmed-0020110-b23]\]. Finally, Wolbachia may be an essential source of nucleotides during embryogenesis. Wolbachia as a source of the above metabolites would explain the block in embryogenesis and the sterility seen in worms after depletion of the endobacteria. Conclusion {#s5} ========== The completion of the wBm genome offers a wealth of information for both basic and applied science. With the completion of this genome, one can now compare close relatives that infect different hosts and have different symbiotic lifestyles, i.e., parasitic versus mutualistic. Such a comparison of the differences could help to define genes necessary for invading host cells. Examining the genome of wBm will help us understand the molecular basis for the endosymbiosis between filarial nematodes and Wolbachia. Researchers now know which metabolites the endobacteria require from the nematode host and which might be provided by Wolbachia to the nematode. This is exciting because it opens up the possibility to find and test drugs already registered for use in humans that might inhibit key biochemical pathways in the Wolbachia---and lead to sterility or killing of the adult worms in shorter treatment times---and that could be given to all infected persons. Given the huge challenges that still lie ahead for the programs to eliminate filariasis, such a need has been and will be repeatedly stated. The sequencing and annotation of the wBm genome will aid researchers in fulfilling this goal by focusing research on those biochemical events that are essential to the mutualistic symbiosis between filarial nematodes and their Wolbachia. Citation: Pfarr K, Hoerauf A (2005) The annotated genome of Wolbachia from the filarial nematode Brugia malayi: What it means for progress in antifilarial medicine. PLoS Med 2(4): e110. Mf : microfilaria or microfilariae TLR : Toll-like receptor wBm : Wolbachia of Brugia malayi wMel : Wolbachia of Drosophila melanogaster [^1]: Kenneth Pfarr and Achim Hoerauf are at the Institute for Medical Parasitology, University of Bonn, Germany. [^2]: Competing Interests: The authors declare that no competing interests exist.	PubMed Central	2024-06-05T03:55:55.902515	2005-4-26	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1087209/", "journal": "PLoS Med. 2005 Apr 26; 2(4):e110", "authors": [ { "first": "Kenneth", "last": "Pfarr" }, { "first": "Achim", "last": "Hoerauf" } ] }
PMC1087210	There are 300 million cases of malaria each year worldwide, causing one million deaths. Around 90% of these deaths occur in Africa, mostly in young children. One of the greatest challenges facing Africa in the fight against malaria is drug resistance; resistance to chloroquine (CQ), the cheapest and most widely used antimalarial, is common throughout Africa, and resistance to sulfadoxine-pyrimethamine (SP), the first-developed and least expensive alternative to CQ, is also increasing in eastern and southern Africa. These trends have forced many countries to change their treatment policies and use more expensive drugs, including drug combinations that will hopefully slow the development of resistance. One avenue of research is to identify combinations that minimize gametocyte emergence in treated cases and prevent selective transmission of parasites resistant to any of the partner drugs. In this month\'s PLoS Medicine Colin Sutherland and colleagues tested two leading combination therapies in children with uncomplicated malaria. One regimen was an artemisinin-based combination consisting of artemether and lumefantrine (co-artemether, trade names CoArtem and Riamet). The other was a combination of CQ and SP---currently under consideration in several African countries, largely due to its low cost. In this randomized, controlled trial, 497 children with acute uncomplicated falciparum malaria were given either a combination of CQ and SP or six doses of co-artemether (91 received CQ/SP and 406 received co-artemether), and their blood was tested for infectivity to mosquitoes seven days after treatment. During follow up at seven, 14, and 28 days the team found that children treated with co-artemether were significantly less likely to carry gametocytes in their blood than children treated with CQ and SP---7.9% compared with 48.8%.[](#pmed-0020114-g001){ref-type="fig"} ::: {#pmed-0020114-g001 .fig} ::: {.caption} ###### Thick blood films from patients in Gambia with malaria (Photo: Elisa Meier) ::: ![](pmed.0020114.g001) ::: Altogether, the six-dose regimen of co-artemether was highly effective at reducing the prevalence and duration of gametocyte carriage. The numbers of gametocytes and the infectiousness to mosquitoes at day 7 were also reduced compared to a combination of CQ and SP, said the authors. Other studies have already shown the potential of co-artemether combination therapy to both cure malaria and reduce gametocyte carriage, acknowledged the authors. However, this study is the first to demonstrate the treatment\'s potential to markedly reduce the infectiousness of patients to mosquitoes, and has done so in a sub-Saharan African setting with highly seasonal transmission and where asymptomatic infections are common. Do the results mean co-artemether should be introduced as a first-line treatment for malaria in Africa? The authors are hesitant and suggest there might be compliance issues with the six-dose regimen. The requirement of oily food for adequate absorption might also lead to inadequate drug levels in the blood of many treated individuals. The authors suggest that co-artemether as a first-line treatment is not likely to reduce overall transmission of Plasmodium falciparum within the community but rather would reduce selective transmission of resistant parasites in treated patients. Hence, co-artemether could have a public health benefit by reducing the impact of drug resistance.	PubMed Central	2024-06-05T03:55:55.903733	2005-4-26	{ "license": "Creative Commons Zero - Public Domain - https://creativecommons.org/publicdomain/zero/1.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1087210/", "journal": "PLoS Med. 2005 Apr 26; 2(4):e114", "authors": [] }

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.