Split (1)

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

id string	text string	source string	added string	created string	metadata dict
PMC1142337	Background ========== The cost-effectiveness of insecticide-treated nets in reducing morbidity and mortality in malaria endemic countries has been proven time and again \[[@B1]-[@B3]\]. I It is considered to be one of the most cost-effective ways of reducing the burden of malaria, with an estimated cost per Disability Adjusted Life Year (DALY) averted of under \$50 \[[@B4]\]. However most studies have been undertaken along side trials, and as such they equate to little more than measures of cost-efficacy, leaving a need for better estimates of the true cost-effectiveness of such programmes in practice. With one of the main Roll Back Malaria goals being to achieve at least 60% coverage of pregnant women and children under five years of age with ITNs, evaluations of the cost-effectiveness of different methods of financing and delivering nets on a large scale is desperately needed. Differences in measures of cost-effectiveness in trials and in interventions in practice are understandable, due to the inevitable changes in returns to scale associated with the scaling up of interventions \[[@B5],[@B6]\], the operational difficulties associated with programmes of this size \[[@B7],[@B8]\] and the constant returns to scale normally associated with primary health care delivery systems. \[[@B9],[@B10]\] A number of interventions, that have been evaluated alongside trials, have been shown to be far less effective and often more costly, once they have been scaled up and undertaken outside the confines of the trial setting \[[@B11]\]. The debate as to the best way to achieve long-term shifts in levels of ITN utilization in malaria endemic countries has centered around the trade-off between the need for immediate health impact and the need for long-term sustainability of such a change in coverage. Those who advocate the universal distribution of free nets have prioritized the need for immediate results in terms of health gain, whereas those who argue for the development of domestic markets for ITNs wish to ensure the long-term sustainability of utilization of ITNs. The Malawi model is a third way that combines traditional, social marketing with heavily subsidized highly-targeted distribution through the nationwide network of public health facilities. Social marketing has been defined as the application of \'commercial marketing technologies to the analysis, planning, execution, and evaluation of programmes designed to influence the voluntary behaviour of target audiences in order to improve their personal welfare and that of their society\' \[[@B12]\]. In addition to this, the model has presented itself with the challenge of achieving high coverage levels in a relatively short period of time, while developing and sustaining a local market for ITNs which could sustain itself into the future, when donor money reprioritizes or dries up. While there is a small but growing literature looking at the application of social marketing techniques to ITN distribution \[[@B2],[@B13]-[@B17]\] the same cannot be said for economic analyses. It has been said that social marketing has additional costs that free distribution nets do not, such as advertising, branding, promotion and retailers\' margins, yet there has been only one published, cost-effectiveness study of a social marketing project and that was deemed to be cost-effective \[[@B2]\]. Currently the literature on the cost-effectiveness of ITN distribution interventions is measured using only the immediate, directly relevant health outcomes, and ignores any benefits from developing the market for future accessibility. This is understandable, as conventional forms of economic evaluation tend to overlook issues of sustainability \[[@B18]\]. Nevertheless its value comes in practicality; in the ability to make comparisons between different methodologies with broadly similar goals, and most fundamentally, in recognizing that the resources available for health services are insufficient to meet all the potential uses for them. This evaluation looks at the cost-effectiveness of a specific social marketing ITN intervention in terms of cost per ITN distributed and cost per treated-net-year. In 1998, USAID contracted Population Services International/ Malawi to design and implement a social marketing ITN programme in Blantyre district. The programme strategy adopted has been published in detail elsewhere \[[@B19]\]. In brief, the strategy involves segmenting the market such that a more expensive blue conical net (with insecticide treatment kit) is made available to consumers for \$5--6 through private sector outlets, targeting those who can afford a commercially priced net. A subsidized green rectangular net (with a kit) is made available to pregnant women and children under 5 for \$0.6, through public health facilities. The nets were branded and heavily promoted to the public through a range of mass media and interpersonal communications channels. Over the next three years the delivery of both the commercially available net and the health facility model was expanded nationwide, thanks to a collaborative effort involving the Malawi Government, UNICEF, USAID and DFID. By January 2003, ITNs were being delivered through commercial outlets and public health facilities in all 27 districts of the country. To further improve access to ITNs, in 2003, unbranded green rectangular nets (with a kit) were delivered via community-based groups at the subsidised price of \$1.2. During the latter 12 month period reported here (October 2002 -- September 2003), a total of 942,000 nets were sold of which 8% were blue conical nets, 16% were unbranded green rectangular nets delivered via community-based channels and 76% were green rectangular nets delivered through public health facilities. At the time of writing, the programme continues to sell about 100,000 ITNs per month nationwide with roughly the same proportion of each type of net. Methods ======= The costs were taken from financial expenditure data collected over the course of the first five years of the programme. As the study was retrospective, no research costs are included. The costs relate to Blantyre District only in year one, and nationwide from year two. Costs were broken down into capital costs, which were annualized and discounted across their life span, and recurrent or programme costs. The latter were broken down into direct costs associated with the ITN programme, and shared costs which were apportioned using one of three indicators. The first was budget headings, the second was total volume of product and the third was total sales calls by agents. An appropriate apportionment method was chosen for each type of expenditure, for example, in terms of apportioning warehouse rental, product volume was used, as ITNs take up much more space than condoms or sachets of oral rehydration salts, which are also delivered by PSI/Malawi. Whereas, with cost of sales commissions, the number of sales calls was used, which better represented the relative effort of the sales agents. Capital items were annualized over assumed life spans, taken directly from the KINET study ^2^so that comparison with other studies, the main aim of this study, could be more transparent. The brand was estimated at seven years; billboards and vehicles were eight years, and computers, furniture and the bed-nets themselves were five years. The discount rate used was 3%, and costs were measured in a combination of local currency (Malawi Kwacha) and in US dollars, depending on whether the resources were purchased or paid locally or overseas. All costs were then translated into US dollars on the MK-US\$ exchange rate for July 1^st^of that year. Results ======= Costs ----- The financial cost of the programme, shown in Table [1](#T1){ref-type="table"}, over the first five years was just over \$6 million, with an economic cost from 1998--2003 of just over \$3,500,000. In this time, a little under 1.5 million ITNs and 300,000 re-treatment kits were sold. Table [2](#T2){ref-type="table"} shows the breakdown of the costs by line item, with set-up costs making up 3%, capital costs making up 57% and recurrent costs making up 40%. The biggest cost was that of the nets and the insecticide, which made up 60% of total costs, followed by staff, which made up 10%. Overhead costs, including local overheads, and the cost of technical support from the PSI/Washington office was also 10%. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Financial and economic costs of the programme from 1999--2003 (1999 prices). ::: Total financial cost (US \$) Total economic cost (US \$) ----------------------------------- ------------------------------ ----------------------------- Brand creation / market research 146,801 107,085 Capital costs Vehicles 157,552 50,228 Equipment & furniture 15,468 15,468 ITNs 4,478,365 2,147,400 Subtotal 4,651,385 2,213,096 Recurrent costs Insecticide 191,555 191,555 Staff 357,204 357,204 Fuel/maintenance 339,346 339,346 Office /warehouse rental 45,672 45,672 Advertising & Promotion 272,646 272,646 Supplies/overheads 351,682 351,682 Subtotal 1,558,104 1,558,104 Total cost 6,356,290 3, 878,287 ITNS distributed 1,471,941 Retreatments distributed 287,079 *Treated net years* *879,510* Cost per net distributed 2.63 Cost per treated net year 4.41 ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Annual costs and cost-effectiveness ratios of the programme 1999--2003 (1999 prices) ::: 1999 2000 2001 2002 2003 Average (%) ------------------------------- ------------- ------------- ------------- ------------- --------------- ------------- ---------- Brand creation 20,972 21,161 21,279 21,687 21,986 21,417 3% Capital costs Vehicles 4,056 6,144 8,381 14,595 17,053 10,046 1% Equipment & furniture 1,903 3,288 4,208 2,848 3,221 3,094 0% ITNs 77,394 202,893 348,829 586,770 931,515 429,480 55% Subtotal 83,353 212,325 361,418 604,213 951,789 442,619 57% Recurrent costs Insecticide 16,753 34,335 38,808 58,076 43,582 38,311 4% Staff 81,496 90,572 54,058 64,104 66,974 71,441 10% Fuel/maintenance 37,346 67,105 44,115 94,223 96,558 67,869 9% Office /warehouse rental 806 6,380 10,204 14,202 14,080 9,134 1% Advertising & Promotion 83,960 71,846 48,149 27,603 41,087 54,529 7% Supplies/overheads 38,954 42,492 42,905 82,322 145,008 70,336 10% Subtotal 259,315 312,731 238,240 340,531 407,288 311,621 40% ITNS distributed 72,196 131,881 174,376 372,911 720,577 294,388 Retreatments distributed 22,337 46,731 54,103 81,824 82,084 57,416 *Treated-net-years* 47,267 89,306 114,240 227,368 401,331 175,902 Total cost 363,640 546,217 620,937 966,431 1,381,063 775,657 100% Cost per net distributed 5.04 4.14 3.56 2.59 1.92 2.63 Cost per treated-net-year 7.69 6.12 5.44 4.25 3.44 4.41 ::: Consequences ------------ As this is a retrospective evaluation of an ongoing working programme, rather than a trial, no health impact data was collected, and so the focus is on process outcomes, including the number of nets distributed and the number of treated net years. The latter measure comes from previous literature on CEA of ITNs \[[@B2]\] and does not require direct translation into health benefits. This allows the results from this study to be compared with other studies. The choice of cost per treated-net-year (TNY) is conservative, as it assumes zero benefit from an untreated net and insecticide treatment is assumed to last 6 months. The discussion draws from the literature on the efficacy of ITNs to estimate the likely health impact and more recognizable measures of cost-effectiveness. Cost and effectiveness ---------------------- The average economic cost per net delivered and the average cost per treated-net-year, over the five years, are \$2.63 and \$4.41 respectively. This compares favourably to other studies where estimates of \$8 and \$4 \[[@B2],[@B3]\] have been shown. The interesting aspect of this study is the gains from cost savings from producing at higher levels, or what is often termed scale efficiency savings (SES) that are made, as shown in Table [2](#T2){ref-type="table"}, where the cost per treated-net-year drops from \$7.69 to \$3.44 as throughput rises from 72,196 to 720,577 ITNs. Discussion ========== It is vital for policy-makers to have information on the costs and effects of scaling up malaria interventions, including the provision of ITNs. Recent work in this area has undoubtedly helped to begin addressing this issue. For example, the work by the Commission for Macroeconomics and Health has provided preliminary estimates of costs of scaling up 5 malaria-related interventions including diagnosis and treatment for over-fives, chemoprophylaxis or presumptive treatment for pregnant women, provision of insecticide-treated nets and residual household spraying for malaria prevention \[[@B22]\]. A number of countries are also experimenting with scaling up of interventions designed to improve the home management of malaria. Lessons learnt in Ghana, Uganda, Nigeria, Burkina Faso, Zambia and Kenya are now being shared \[23\]. While there is no doubt that researchers are paying more attention to issues surrounding the costs and effects of scaling up malaria interventions, there is still a way to go, especially with regard to the delivery of ITNs. In particular, the costs of scaling up ITNs are currently restricted to a relatively small number of studies based on the evaluation of trials or research studies that are often of limited scale. These studies also fail to account for the inevitable growth of scaling up over time that is present in implementation of many public health interventions. The purpose of this study is to start to address these gaps by incorporating ongoing fieldwork into the cost-effectiveness debate around the delivery of ITNs. Scaling up of ITN delivery -------------------------- One of the key findings of this study is that there are considerable scale-efficiency savings to be made. Table [3](#T3){ref-type="table"} shows us the scale efficiency savings (SES) over the five years and relates them to increases in throughput of ITNs. SES has been separated into two components, the \'procurement SES\' and the \'distribution SES\'. This allows us to see to what extent the SES are a component of this particular method of distribution and what part of the SES is due primarily to the greater bargaining power of scale. As can be seen from this data, approximately half of the relative efficiency savings over time are due to lowering product or procurement costs. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Breakdown of total scale efficiency savings into \'procurement\' and \'distribution\' costs ::: Financial unit costs 1999 2000 2001 2002 2003 Totals/Average ------------------------------ ---------- ---------- ---------- ---------- ---------- -------------------- *ITN unit cost* 5.36 4.67 4.01 3.05 2.27 3.04 *Retreatment pack u/c* 0.75 0.75 0.75 0.75 0.57 0.70 *ITNs* 72,196 131,881 174,376 372,911 720,577 1,471,941 *Retreatment packs* 22,337 46,731 54,103 81,824 82,084 287,079 *Treated net years* 47,267 89,306 114,240 227,368 401,331 879,510 *Output growth (actual)* 42,040 24,934 113,128 173,963 478,180 Output growth (%) 89% 28% 99% 77% 73% *Procurement U/C* 5.36 4.67 4.01 3.05 2.27 *U/C savings* 0.69 0.66 0.96 0.78 *3.09* SES 13% 14% 24% 26% 58% *Distribution U/C* 3.73 2.33 1.32 0.85 0.56 *U/C savings* 1.40 1.01 0.47 0.30 *3.18* SES 37% 43% 36% 35% 85% *Total U/C* 9.09 7.00 5.33 3.90 2.83 *Savings U/C* 2.09 1.67 1.43 1.08 *6.27* SES 23% 24% 27% 28% 69% ::: It could be estimated that these SES could have been enjoyed by any of the alternative methods of distribution, whereas the distribution SES is probably more closely related to the specifics of the distribution method employed, although this is purely speculation, as there is no comparator. What we can say is that health systems, particularly public sector health systems are not renowned for their economies of scale or for falling marginal costs, and that recent studies looking at opening public services up to private competition has tended to increase cost-efficiency and returns to scale \[[@B20]\]. One thing is certain, there is a growing belief that the reliance on an assumption of constant returns to scale is limiting the value and practicality of cost-effectiveness studies to policy makers \[[@B5],[@B21]\]. Currently, despite the evidence of the relative cost-effectiveness of ITNs and the goals and the political commitment of the RBM partners to increase coverage throughout sub-Saharan Africa beyond the 60% mark, there has not been the speed of progress in scaling up of this intervention that has been required. A recent study by the CDC and UNICEF looked at changes in ITN and bed-net usage in Malawi as a whole between 2000 and 2003 \[[@B22]\]. Households of at least one net have risen from 12% to 43%. In target groups, 35% of children under 5 slept under a net the previous night, up from 8% and similarly, 32% of pregnant women slept under a net the previous night, up from just 8% three years previously. These are significant changes in a relatively short period of time. At this time, this is partly due to a lack of financial commitment which is, in turn, due to a lack of a consensus on: a) how best to undertake this scale up; and b) how much it is likely to cost. The first of these two cannot be answered unless the desired output, be it pure short-term health gain, long-term development of a sustainable net culture, or both, is clarified. The second needs to consider the fact that economies-of-scale may exist. If estimates of the cost of achieving certain targets are based on cost-effectiveness data from small scale trials, and modeled at constant returns to scale there is a chance this could drastically overestimate the true cost, if the evidence presented here on returns to scale is not an anomaly. Without more evaluations of programmes at scale to compare, this can only be speculation. Conclusion ========== The efficacy of insecticide-treated nets for reducing the burden of malaria in sub-Saharan Africa has been repeatedly proven and the debate in this field should, and to some extent has, moved beyond marginal value to one of measuring marginal productivity. The goal for economists and policymakers is to determine the method, or combinations of methods, that can ensure the best way of achieving a sustainable high-level of utilization of this product in communities where the benefits are highest. It is more than likely that with the dual goals of health impact through high coverage levels, and long-term sustainability of the supply of ITNs in those same countries, that a strategy which involves a combination of different methods of distribution will be required. This paper suggests that a combination of standard social marketing techniques, combined with targeting vulnerable groups with highly subsidized ITNs through both the commercial and formal health care sectors, could achieve relatively high-levels of coverage in both urban and rural areas and in vulnerable groups over time with proper investment. In addition, contrary to the weight of evidence on scaling up of public health interventions, and of primary health care in general, it may be possible to achieve these high levels of ITN distribution and rapid increases in coverage while keeping unit costs down and achieving increasing returns to scale. To justify such a conclusion, and to compare and contrast with other hybrid methods of ITN delivery, economic evaluations of large-scale programmes need to be carried out. Authors\' contributions ======================= Warren Stevens compiled and analyzed the data and wrote the first draft of the paper. Virginia Wiseman advised on methodology and analysis and was involved in redrafting of the paper. Juan Ortiz and Desmond Chavasse conceived of the study, were involved in data collection and also in the development of the intervention, and redrafting of the final paper. Acknowledgements ================ The authors would like to acknowledgePSI/Malawi, the Malawi Ministry of Health and Population, the National Malaria Control Programme, UNICEF, the District Health Management Teams, the Blantyre Integrated Malaria InitiativeandMalawi\'s nurses for their roles in the execution of the national ITN programme. John Kadzandira of the Centre or Social Researchprovided theITN coverage data. We are grateful tothe United States Agency for International Department, UNICEF and the UK Department for International Development for financial support of the programme.	PubMed Central	2024-06-05T03:55:57.391363	2005-5-10	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1142337/", "journal": "Malar J. 2005 May 10; 4:22", "authors": [ { "first": "Warren", "last": "Stevens" }, { "first": "Virginia", "last": "Wiseman" }, { "first": "Juan", "last": "Ortiz" }, { "first": "Desmond", "last": "Chavasse" } ] }
PMC1142338	Review ====== 1. Introduction --------------- The solubilization of metals due to the action of microbes and the subsequent recovery of the metals from solution has deep historical roots that have been extensively reviewed \[[@B61],[@B70]\]. Similarly, an indication of the number and sizes of the operations that employ microbes for the recovery of mainly copper, gold, cobalt and uranium has also been reviewed \[[@B61],[@B72]\]. These processes use the action of microbes for one of two purposes. Either to convert insoluble metal sulfides (or oxides) to water soluble metal sulfates or as a pretreatment process to open up the structure of the mineral thereby permitting other chemicals to better penetrate the mineral and solubilize the desired metal. An example of the first type of process is the conversion of insoluble copper present in minerals such as covellite (CuS) or chalcocite (Cu~2~S) to soluble copper sulfate. An example of the second, is the removal of iron, arsenic and sulfur from gold-bearing arsenopyrite so that the gold that remains in the mineral is more easily extracted by subsequent treatment with cyanide. Both are oxidation processes, but where the metal to be recovered is extracted into solution the process is known as bioleaching, whereas when the metal remains in the mineral, bioleaching is an inappropriate term and the process should strictly be referred to as biooxidation. Nevertheless, the term bioleaching is frequently used for both. Not all types of mineral are amenable to biologically-assisted leaching. In general, the mineral should contain iron or a reduced form of sulfur. Alternately, a mineral lacking in these compounds may be leached if it occurs together with another mineral that contains iron and reduced sulfur, provided that the mineral is subject to attack by ferric iron and/or sulfuric acid. Metals in certain non-sulfide minerals may be solubilized by a process of complexation with oxalic, citric or other organic acids. These organic acids are typically produced by certain types of fungi and this type of metal solubilization process will not be discussed in this review \[[@B8]\]. This review will focus on properties that the various types of mineral biooxidation organisms have in common. However, before discussing these general characteristics it is necessary to describe briefly the mechanism of leaching and the technology of the metal recovery processes. 2. Mechanisms of bioleaching ---------------------------- Metal leaching is now recognized as being mainly a chemical process in which ferric iron and protons are responsible for carry out the leaching reactions. The role of the microorganisms is to generate the leaching chemicals and to create the space in which the leaching reactions take place. Microorganisms typically form an exopolysaccharide (EPS) layer when they adhere to the surface of a mineral \[[@B78]\] but not when growing as planktonic cells \[[@B22]\]. It is within this EPS layer rather than in the bulk solution that the biooxidation reactions take place most rapidly and efficiently and therefore the EPS serves as the reaction space \[[@B31],[@B75],[@B78],[@B89]\]. The mineral dissolution reaction is not identical for all metal sulfides and the oxidation of different metal sulfides proceeds via different intermediates \[[@B80]\]. This has also been recently reviewed \[[@B75]\]. Briefly, a thiosulfate mechanismhas been proposed for the oxidation of acid insoluble metal sulfides such as pyrite (FeS~2~) and molybdenite (MoS~2~), and a polysulfide mechanismfor acid soluble metal sulfides such as sphalerite (ZnS), chalcopyrite (CuFeS~2~) or galena (PbS). In the thiosulfate mechanism, solubilization is through ferric iron attack on the acid-insoluble metal sulfides with thiosufate being the main intermediate and sulfate the main end-product. Using pyrite as an example of a mineral, the reactions may be represented as: FeS~2~+ 6 Fe^3+^+ 3 H~2~O → S~2~O~3~^2-^+ 7 Fe ^2+^+ 6 H^+^ (1) S~2~O~3~^2-^+ 8 Fe^3+^+ 5 H~2~O → 2 SO~4~^2-^+ 8 Fe^2+^+ 10 H^+^ (2) In the case of the polysulfide mechanism, solubilization of the acid-soluble metal sulfide is through a combined attack by ferric iron and protons, with elemental sulfur as the main intermediate. This elemental sulfur is relatively stable but may be oxidized to sulfate by sulfur-oxidizing microbes such as Acidithiobacillus thiooxidansor Acidithiobacillus caldus(reaction 5 below). MS + Fe^3+^+ H^+^→ M^2+^+ 0.5 H~2~S~n~+ Fe ^2+^(n ≥ 2) (3) 0.5 H~2~S~n~+ Fe^3+^→ 0.125 S~8~+ Fe^2+^+ H^+^ (4) ![](1475-2859-4-13-i1.gif) The ferrous iron produced in reactions (1) to (4) may be reoxidized to ferric iron by iron-oxidizing microorganisms such as Acidithiobacillus ferrooxidansor bacteria of the genera Leptospirillumor Sulfobacillus. ![](1475-2859-4-13-i2.gif) The role of the microorganisms in the solubilization of metal sulfides is, therefore, to provide sulfuric acid (reaction 5) for a proton attack and to keep the iron in the oxidized ferric state (reaction 6) for an oxidative attack on the mineral. 3. Effect of temperature ------------------------ Bioleaching processes are carried out at a range of temperatures from ambient to a demonstration plant that has been operated at 80°C \[[@B72]\]. As would be expected, the types of iron- and sulfur-oxidizing microbes present differ depending on the temperature range. The types of microbes found in processes that operate from ambient to 40°C tend to be similar irrespective of the mineral, as are those within the temperature ranges 45--55°C and 75--80°C. As described below, there are two broad categories of biologically-assisted mineral degrading processes. An ore or concentrate is either placed in a heap or dump where it is irrigated or a finely milled mineral suspension is placed in a stirred tank where it is vigorously aerated. In general, mineral solubilization processes are exothermic and when tanks are used, cooling is required to keep the processes that function at 40°C at their optimum temperature. At higher temperatures the chemistry of mineral solubilization is much faster and in the case of minerals such as chalcopyrite, temperatures of 75--80°C are required for copper extraction to take place at an economically viable rate. 4. Commercial metal extraction operations ----------------------------------------- ### 4.1. Heap leaching processes Commercial bioleaching can take place using what may be considered to be a low technology process, the irrigation of waste ore dumps \[[@B13]\]. The metal recovery process may be made more efficient by the construction and irrigation of especially-designed heaps rather than by the irrigation of an existing dump that has not been designed to optimize the leaching process \[[@B13],[@B72],[@B81]\]. When building a heap, agglomerated ore is piled onto an impermeable base and supplied with an efficient leach liquor distribution and collection system. Acidic leaching solution is percolated through the crushed ore and microbes growing on the surface of the mineral in the heap produce the ferric iron and acid that result in mineral dissolution and metal solubilization. Aeration in such processes can be passive, with air being draw into the reactor as a result of the flow of liquid, or active with air blown into the heap through piping installed near the bottom. Metal-containing leach solutions that drain from the heap are collected and sent for metal recovery \[[@B81]\]. Heap reactors are cheaper to construct and operate and are therefore more suited to the treatment of lower grade ores. However, compared with tank reactors, heap reactors are more difficult to aerate efficiently and the undesirable formation of gradients of pH and nutrient levels as well as liquor channeling are difficult to manage. Furthermore, although one can rely on the natural movement of microbes to eventually inoculate the heap, initial rates of bioleaching can be improved by effective heap inoculation, but this is difficult to achieve. Copper is the metal recovered in the largest quantity by means of heap reactors \[reviewed in \[[@B61],[@B72]\]\]. Although comparisons are difficult as data are presented in different ways, examples of large copper leaching operations are those by Sociedad Contractual Minera El Abra and the Codelco Division Radimiro Tomic both in Chile and producing 225 000 and 180 000 tonnes Cu per annum respectively. Gold ore is also pretreated by bioleaching in heaps by Newmont Mining, in the Carlin Trend region, Nevada, USA. ### 4.2. Tank leaching processes In stirred tank processes highly aerated, continuous-flow reactors placed in series are used to treat the mineral. Finely milled mineral concentrate or ore is added to the first tank together with inorganic nutrients in the form of ammonia- and phosphate-containing fertilizers. The mineral suspension flows through series of highly-aerated tanks that are pH and temperature-controlled \[[@B23],[@B70],[@B93]\]. Mineral solubilization takes place in days in stirred-tank reactors compared with weeks or months in heap reactors. Stirred tank reactors that operate at 40°C and 50°C have proven to be highly robust and very little process adaptation is required for the treatment of different mineral types \[[@B68]\]. A major constraint on the operation of stirred tank reactors is the quantity of solids (pulp density) that can be maintained in suspension. This is limited to about 20% as at pulp densities \>20%, physical mixing and microbial problems occur. The liquid becomes too thick for efficient gas transfer and the shear force induced by the impellers causes physical damage to the microbial cells. This limitation in solids concentration plus considerably higher capital and running costs in tank compared with heap reactors has meant that the use of stirred reactors has been restricted to high value minerals or mineral concentrates \[[@B72]\]. Stirred tanks are used as a pretreatment process for gold-containing arsenopyrite concentrates with the first of these having been built at the Fairview mine, Barberton, South Africa in 1986 \[[@B73],[@B93]\]. The largest is at Sansu in the Ashanti goldfields of Ghana, West Africa. These two operations currently treat 55 and 960 tonnes of gold concentrate per day respectively. Another example is the use of stirred tanks to treat 240 tonnes of cobalt-containing pyrite in 1300 m^3^tanks at Kasese, Uganda \[\[[@B14]\], reviewed in \[[@B72]\]\]. ### Types of Microorganisms In general, the types of microorganisms found in heap-leaching processes are similar to those found in stirred tank processes, however, the proportions of the microbes may vary depending on the mineral and the conditions under which the heaps or tanks are operated. In processes that operate from ambient temperatures to about 40°C, the most important microorganisms are considered to be a consortium of Gram-negative bacteria. These are the iron- and sulfur-oxidizing Acidithiobacillus ferrooxidans(previously Thiobacillus ferrooxidans), the sulfur-oxidizing Acidithiobacillus thiooxidans(previously Thiobacillus thiooxidans) and Acidithiobacillus caldus(previously Thiobacillus caldus), and the iron-oxidizing leptospirilli, Leptospirillum ferrooxidansand Leptospirillum ferriphilum\[[@B18],[@B29],[@B32],[@B34],[@B94]\]. If ferrous iron is added to the leaching solutions (lixiviants) that are circulated through a heap or dump, then At. ferrooxidansmay dominate the iron-oxidizers. In continuous flow, stirred tank processes, the steady state ferric iron concentration is usually high and under such conditions At. ferrooxidansis less important than a combination of Leptospirillumand At. thiooxidansor At. caldus\[[@B71]\]. Gram-positive iron and sulfur-oxidizing bacteria related to Sulfobacillus thermosulfidooxidanshave also been identified \[[@B29]\]. The consortium of bioleaching microbes frequently includes acidophilic heterotrophic organisms such as bacteria belonging to the genus Acidiphilium\[[@B38]\] or Ferroplasma-like archaea \[[@B33],[@B95]\]. A fluidized-bed reactor operating at 37°C and pH 1.4 was dominated by L. ferriphilumwith a small proportion of Ferroplasma-like archaea \[[@B47]\]. \'Heterotrophically inclined\' microbes are believed to assist the growth of iron-oxidizing bacteria like At. ferrooxidansand the leptospirilli \[[@B36],[@B43]\]. This is thought to be due to their ability to provide essential nutrients or to remove toxic organic compounds or other inhibitory substances. How much this ability contributes to the overall mineral biooxidation efficiency of a microbial consortium in practice is still unclear \[[@B45]\]. There are fewer commercial processes that operate in the 45--50°C range and therefore studies on microorganisms that dominate these bioleaching consortia have been less well reported. Rawlings et al., \[[@B71]\] identified At. caldusand a species of Leptospirillumas being the dominant microbes in a continuous-flow biooxidation tanks processing several mineral ores operating in this temperature range. At. caldus, Sulfobacillus thermosulfidooxidansand bacteria of the informally recognized species \'Sulfobacillus montserratensis\' together with an uncultured thermal soil bacterium were found to dominate the consortium of organisms oxidizing chalcopyrite concentrate at 45°C. The same bacteria dominated the culture irrespective of whether chalcopyrite, pyrite or an arsenic pyrite concentrate was being oxidized \[[@B26]\]. In a pilot scale, stirred-tank operation in which three tanks in series were used to treat a polymetallic sulfide ore at 45°C, At. caldus-like, L. ferriphilum-like and Sulfobacillus-like bacteria were found to dominate the first tank \[[@B59]\]. The proportions of these bacteria decreased in the second tank with the numbers of At. caldusand Ferroplasma-like archaea being equally dominant. The Ferroplasma-like archaea completely dominated the third tank with the number of leptospirilli being reduced to undetectable levels. When combinations of pure cultures were tested, a mixed culture containing both autotrophic (LeptospirillumMT6 and At. caldus) and heterotrophic moderate thermophiles (FerroplasmaMT17) was the most efficient \[[@B60]\]. The presence of Ferroplasma-like organisms is being increasing recognized in bioleaching processes that operate at very low pH (1.4 or less). These archaea appear to be able to oxidize minerals like pyrite in pure culture although not without a small quantity of yeast extract. Species of the gram-positive genus, Acidimicrobium\[[@B16]\] may occur together with sulfobacilli in cultures that grow at 45°C. There are even fewer reports on types of microbes that occur in mineral treatment processes that operate at temperatures \>70°C than at lower temperatures. However, it is clear that these biomining consortia are dominated by archaea rather than bacteria, with species of Sulfolobusand Metallosphaerabeing most prominent \[[@B54],[@B57]\]. Sulfolobus metalicushas been found to dominate at 70°C but this archeaon is probably excluded at higher temperatures with other Metalosphaera-like and Sulfolobus-like archaea dominating at 80°C. Archaea belong to the genus Acidianussuch as Ad. ambivalensior Ad. infernusare also capable of growing at high temperature (90°C for Ad. infernus) on reduced sulfur and at low pH. However, the contribution of these organisms to industrial bioleaching is not well-established \[[@B54]\]. 5. General characteristics of mineral degrading bacteria -------------------------------------------------------- As would be gathered from the above, the most important microbes involved in the biooxidation of minerals are those that are responsible for producing the ferric iron and sulfuric acid required for the bioleaching reactions. These are the iron- and sulfur-oxidizing chemolithrophic bacteria and archaea \[[@B70]\]. Irrespective of the type of process or temperature at which they are employed, these microbes have a number of features in common that make them especially suitable for their role in mineral solubilization. Four of the most important characteristics are; a) they grow autotrophically by fixing CO~2~from the atmosphere; b) they obtain their energy by using either ferrous iron or reduced inorganic sulfur compounds (some use both) as an electron donor, and generally use oxygen as the electron acceptor; c) they are acidophiles and grow in low pH environments (pH.1.4 to 1.6 is typical) and d) they are remarkably tolerant to a wide range of metal ions \[[@B25]\], though there is considerable variation within and between species. Each of the these characteristics will be dealt with in the sections that follow. The modest nutritional requirements of these organisms are provided by the aeration of an iron- and/or sulfur-containing mineral suspension in water or the irrigation of a heap. Small quantities of inorganic fertilizer can be added to ensure that nitrogen, phosphate, potassium and trace element limitation does not occur. A further advantageous characteristic of mineral biooxidation operations is that they are usually not subject to contamination by unwanted microorganisms. In the case of continuous-flow tank leaching processes, the continual wash-out of mineral together with their attached microbes as well as the organisms in suspension provides strong selection for improved microorganisms. 6. Nutrition ------------ ### 6.1 Autotrophy Microorganisms that drive the mineral degradation processes are autotrophic and obtain their carbon for cell mass synthesis from the carbon dioxide in the air used to aerate the process. Heterotrophic microorganisms that live off waste products produced by the autotrophs are usually also present and there is some evidence that these heterotrophs might assist the process \[[@B45]\]. Mineral degradation processes differ from the vast majority of other commercial processes that employ microorganisms where an organic substrate is necessary to provide the carbon source and energy required for microbial growth. If it were necessary to feed the microorganisms required for mineral degradation with a carbon source (e.g. molasses), commercial mineral biooxidation processes would be unlikely to be viable. Bacteria such as the acidothiobacilli and leptospirilli, fix CO~2~by the Calvin reductive pentose phosphate cycle, using the enzyme ribulose 1,5-biphosphate carboxylase (RuBPCase or Rubisco) \[[@B92]\]. The CO~2~concentration present in air is generally sufficient to avoid carbon limitation when bacteria such as Acidithiobacillus ferrooxidansare growing on ferrous iron. This bacterium probably responds to CO~2~limitation by increasing the cellular concentration of RuBPCase \[[@B17]\]. At. ferrooxidansstrain Fe1 has been reported to have two identical copies of the structural genes for RuBPCase (although the flanking regions are different, \[[@B49]\]) which are separated by more than 5 kb \[[@B48]\]. The reason for this duplication has not been tested. At. ferrooxidansis considered to be an obligate autotroph but has been shown to use formic acid as a carbon source provided that it was grown in continuous culture and the formic acid was fed in sufficiently slowly for the concentration to remain low \[[@B65]\]. Similarly, genes for a formate hydrogenlyase complex have been located on the genome of Leptospirillumtype II and it is therefore likely to also grow on formate \[[@B92]\]. However, like CO~2~, formic acid has a single carbon atom and when lysed by the cell formate may be assimilated by the Calvin cycle in much the same way as CO~2~. Whether the ability to use formate is of value in commercial processes is not clear. In the case of several of the other bacteria, such as the moderately thermophilic Sulfobacillus thermosulfidooxidans, 1% v/v CO~2~-enriched air is required for rapid autotrophic growth in pure culture. This may be partly because the solubility of CO~2~is reduced at 50°C and partly because these bacteria are known to be inefficient at CO~2~uptake. Sulfobacillus species are nutritionally versatile and also capable of heterotrophic growth \[[@B16],[@B55]\]. Most members of the archaea are heterotrophic, although certain species of the genus Sulfolobushave been reported to grow autotrophically. Details of the CO~2~-fixation pathway are unknown although it has been suggested that acetyl-CoA carboxylation may be a key step and that the synthesis of biotin carboxylase and biotin-carboxyl-carrier protein are increased under conditions of CO~2~limitation \[[@B54]\]. This complex is encoded by genes adjacent to genes encoding a putative propionyl-CoA carboxyl transferase and together these observations are in agreement with the suggestion that Acidianus brierleyihas a modified 3-hydroxypropionate pathway for CO~2~fixation \[[@B41]\]. Other types of archaea such as the Ferroplasmahave the genes necessary to fix carbon dioxide via the reductive acetyl CoA pathway \[[@B92]\]. Like Sulfobacillusspp., autotrophic growth of Sulfolobusspp. is enhanced in 1% CO~2~-enriched air \[[@B54]\]. ### 6.2 Nitrogen, phosphate and trace elements Based on dry weight, nitrogen is the next most important element after carbon for the synthesis of new cell mass. Ammonium levels of 0.2 mM have been reported to be sufficient to satisfy the nitrogen requirement of At. ferrooxidans\[[@B91]\]. High concentrations of inorganic or organic nitrogen are inhibitory to iron oxidation. Exactly how much nitrogen needs to be present in a growth medium will be dependent on the quantity of cell growth to be supported. Ammonia is highly soluble in acid solutions and it has been found that traces of ammonia present in the air can be readily absorbed into growth media. Therefore determination of the exact nitrogen requirements is difficult to estimate. In commercial operations, inexpensive fertilizer grade ammonium sulfate is typically added to biooxidation tanks or bioleaching heaps to ensure that sufficient nitrogen is available \[[@B23]\]. The ability of At. ferrooxidansto reduce atmospheric dinitrogen to ammonia was reported and the genes for the enzyme nitrogenase (nifHDK) were cloned several years ago \[[@B52],[@B64],[@B69]\]. The ability to fix nitrogen is probably a general property of At. ferrooxidansas at least fifteen strains of At. ferrooxidanshave been shown to contain the nitrogenase genes (Rawlings, unpublished). L. ferrooxidanswas also shown to contain nifHDKgenes, to reduce acetylene to ethylene (a common test for nitrogenase activity) and at the same time to oxidize ferrous to ferric iron at low oxygen concentrations \[[@B56]\]. This activity was repressed by ammonia, a strong indication of the nitrogen fixing activity. The nitrogen fixing (nif) operon and many of the nifregulatory elements of a L. ferrooxidansfrom the Tinto river have been isolated and sequenced \[[@B62],[@B63]\]. Interestingly analysis of the genome of Leptospirillumtype II (L. ferriphilum) indicated the absence of genes for nitrogen fixation in this species \[[@B92]\]. Nitrogenase enzyme activity is inhibited by oxygen. It was found that At. ferrooxidansgrowing on iron did not fix nitrogen when aerated, but began to fix nitrogen once the oxygen concentration had fallen \[[@B52]\]. Therefore, how much nitrogen fixation takes place in highly aerated biooxidation tanks or heaps is uncertain. However, the aeration of heaps is not homogenous and nitrogen fixation could take place in parts of a heap where the oxygen is absent or its concentration is sufficiently low. The sensitivity of nitrogenase to oxygen poses a special problem for leptospirilli because, as far as is known, it uses only iron as its electron donor and is probably obligately aerobic. One mechanism by which nitrogenase can be protected against oxygen is respiratory protection, whereby rapid consumption of oxygen by a cytochrome oxidase is maintains a low oxygen concentration compatible with nitrogen fixation. It has been suggested that cytochrome bdis responsible for respiratory protection in At. ferrooxidans\[[@B10]\]. It has been found that Leptospirillumtype II also has genes encoding both ccb3and bdterminal oxidases even though it has no nitrogenase \[[@B92]\]. One can speculate that if cytochrome bdis also present in L. ferrooxdans, this cytochrome could be responsible for respiratory protection of its nitrogenase. 7. Energy sources ----------------- As described in a previous section, the solubilization of minerals is considered to be a chemical process that results from the action of ferric iron and/or acid, typically sulfuric acid. Therefore, irrespective of the temperatures at which they grow at, the microorganisms that play the major role in the leaching of metals from minerals are either iron- or sulfur-oxidizing organisms. The iron and sulfur serve as electron donors during respiration. ### 7.1 Iron oxidation Ferrous iron is readily oxidized to ferric iron and in this way it can serve as an electron donor. The Fe^2+^/Fe^3+^redox couple has a very positive standard electrode potential (+770 mV at pH 2). As a result only oxygen is able to act as a natural electron acceptor and in the presence of protons with the product of the reaction being water (O~2~/H~2~O +820 mV at pH 7). The use of iron as an electron donor will therefore occur only during aerobic respiration. However, under aerobic conditions, ferrous iron spontaneously oxidizes to ferric iron unless the pH is low. Therefore, extremely acidophilic bacteria are able to use ferrous iron as an electron donor in a manner that is not possible for bacteria that grow at neutral pH. Because the difference in redox potential between the Fe^2+^/Fe^3+^and O~2~/H~2~O redox couples is small and because only one mole of electrons is released per mole of iron oxidized, vast amounts of ferrous iron need to be oxidized to produce relatively little cell mass. These large quantities of iron are not transported through cell membrane but remain outside of the cell and each ferrous iron atom simply delivers its electron to a carrier situated in the cell envelope (see below). The mechanism of iron oxidation has been most extensively studied for the bacterium At. ferrooxidans. A model for iron oxidation is shown in Figure [1](#F1){ref-type="fig"}. This bacterium contains a rusoperon that is proposed to encode for the electron transport chain that is used during the oxidation of ferrous iron \[[@B2]\]. This operon consists of genes for an aa~3~-type cytochrome oxidase, a high molecular weight outer membrane located cytochrome-c(Cyc2) \[[@B97]\], a c~4~-type cytochrome, a low molecular weight copper-containing protein rusticyanin (from which the operon derives its name) and an ORF proposed to encode a periplasmic protein of unknown function. The detection of rusticyanin has been linked to the growth of At. ferrooxidanson iron and it has been shown that the expression of the rusoperon was 5- to 25-fold higher during growth on iron compared with sulfur \[[@B99]\]. Indeed, it has been calculated that up to 5% of the total cell protein of At. ferrooxidanswhen grown on iron consists of rusticyanin \[[@B19]\]. It has been suggested that rusticyanin probably functions as an electron reservoir in such a way that it readily takes up electrons available at the outer membrane and channels them down the respiratory pathway \[[@B76]\]. Rusticyanin serves as redox buffering function ensuring that the outer membrane Cyc2 electron acceptor remains in a fully oxidized state, ready to receive electrons from ferrous iron even in the presence of short-term fluctuations of oxygen. Interestingly aporusticyanin has been implicated in the adhesion of At. ferrooxidanscells to pyrite \[[@B4]\]. Although the rusoperon is clearly involved in iron oxidation, it is not yet known whether the components of the operon are sufficient for iron the electron transport system or whether other components such as the irogene for a high redox potential iron oxidase (HiPIP) might also play a role \[[@B50]\]. HiPIPs might not be present in all strains of At. ferrooxidansand might play a bigger role in sulfur oxidation than iron oxidation. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Model of the iron oxidation electron transport pathway of At. ferrooxidansbased partly on references \[10, 75\]. Electrons are transferred from the membrane-located cytochrome c2 \[97\] to rusticyanin and then along one of two paths. The downhill path is via cytochrome c~4~(Cyt1) to cytochrome aa~3~\[2\] or the uphill, reverse electron transport path via cytochrome c~4~(CytA1) to a bc~1~I complex and a NADH-Q oxidoreductase \[28\]. At. ferrooxidanshas up to twelve cytochromes c\[98\] and a variety of cytochrome oxidases some of which appear to play different roles depending on whether iron or sulfur is being oxidized \[10\]. The NADH is responsible for mercury reduction using a MerA mercuric reductase and the cytochrome aa~3~is required to reduce mercury via the unique iron dependent mechanism discovered in At. ferrooxidans\[84\]. ::: ![](1475-2859-4-13-1) ::: A question that has intrigued researchers is whether the iron-oxidation electron transport chains of different organisms are related. Bob Blake (Xavier University) and colleagues have investigated components of iron oxidation in at least five different acidophilic microorganisms, three bacteria (Acidithiobacillus ferrooxidans, unidentified bacterium m1, Leptospirillum ferrooxidans), and two archaea (Sulfobacillus metallicusand Metallosphaera sedula) \[[@B5],[@B6]\]. In all five organisms the components of the electron transport chain were very different and the conclusion was that the ability to use ferrous iron as an electron donor has probably evolved independently at several times. Although iron oxidation is best studied in At. ferrooxidans, enough is known to suggest that the mechanism in L. ferrooxidans(and presumably L. ferriphilum) must be substantially different. Whereas, At. ferrooxidanswas capable of growth on ferrous iron at redox potentials of up to about +800 mV, L. ferrooxidanswas capable of oxidation at redox potentials of closer to +950 mV \[[@B7],[@B37]\]. The effect of this is that although At. ferrooxidanscan outgrow L. ferrooxidansat high ratios of ferrous to ferric iron (as happens during the earlier stages of iron oxidation), L. ferrooxidansoutcompetes At. ferrooxidansonce the ferric iron concentration becomes high \[[@B74]\]. In a microbial community genome sequencing project, Banfield and co workers \[[@B92]\] reported the assembly of an almost complete genome of Leptospirillumgroup II, thought to be the same as L. ferriphilum. This genome contained a red cytochrome, presumably the same as the red cytochrome previously identified in L. ferrooxidans\[[@B5]\]. Other components typical of electron transport chains included putative cytochrome cbb~3~-type haeme-copper terminal oxidases and cytochrome bd-type quinol oxidases. A putative electron transport chain for Leptospirillumgroup II was constructed for both downhill respiration and uphill NADH synthesis electron flows. ### 7.2 Sulfur as an energy source The acid responsible for the very low pH environment in which extreme acidophiles are found is most often sulfuric acid. This sulfuric acid is produced by the oxidation of RISCs (reduced inorganic sulfur compounds). For biological oxidation to occur, the RISCs serve as an electron donor with oxygen serving as the energetically most favourable electron acceptor. The potential amount of energy that can be made available when a sulfur atom from a sulfide ore is oxidised to sulfate is much greater than when iron is oxidized \[[@B66]\]. Naturally occurring RISCs are present wherever sulfide-containing minerals are exposed to the surface. A variety of RISCs are released as a result of the chemical reaction of sulfide minerals with water, ferric iron and oxygen \[[@B79]\]. Attempts to investigate the pathways involved in sulfur oxidation by acidophilic bacteria have proved challenging. The chemical reactivity of many sulfur intermediates has meant that some intermediates may be produced by a combination of spontaneous and enzymatic reactions \[[@B76],[@B79]\]. Nevertheless, progress has being made. Working with At. ferrooxidans, At thiooxidansand the RISC-oxidizing Acidiphilium acidophilium, Rohwerer and Sand \[[@B76]\] proposed a model for the oxidation of elemental and free sulfide sulfur. Extracellular elemental sulfur is mobilized by the thiol groups of specific outer membrane proteins and transported into the cytoplasm as persulfide sulfane sulfur (see Figure [2](#F2){ref-type="fig"}). This persulfide sulfur is oxidized further to sulfate by a sulfite:acceptor oxidoreductase with the electrons most likely being transferred to cytochromes. Glutathione plays a catalytic role in elemental sulfur activation but is not consumed during enzymic sulfane sulfur oxidation. Sulfide oxidation required the disulfide of glutathione which reacted non-enzymatically with sulfide to give glutathione persulfide prior to enzymic oxidation. Free sulfide is oxidized to elemental sulfur in the periplasm by a separate sulfide:quinone oxidoreductase. Reaction with the thiol groups of the outer membrane proteins keeps the zero valence sulfur from precipitating in the periplasm. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### A composite model of sulfur oxidation electron transport pathway of At. ferrooxidansbased on references \[10, 76, 96\]. Thiol groups of outer membrane proteins are believed to transport the sulfur to the periplasm where it is oxidized by a periplasmic sulfur dioxygenase (SDO) to sulfite and a sulfite acceptor oxidoreductase (SOR) to sulfate \[76\]. Although other cytochrome oxidases are present, a ba~3~cytochrome oxidase and a bc~1~II complex together with a bd-type ubiquinol oxidase are believed to play the major roles during sulfur oxidation \[10, 96\]. Rusticyanin and an iron oxidizing protein (not shown) might also be involved during sulfur oxidation but their exact role is still to be determined \[96\]. ::: ![](1475-2859-4-13-2) ::: In a study of the proteins induced when At. ferrooxidanscells were grown on sulfur compared with iron, it was found that an outer membrane protein, a putative thiosulfate sulfur transfer protein, a putative thiosulfate/sulfate binding protein, a putative capsule polysaccharide export protein and several other proteins of unknown function were induced \[[@B67]\]. The thiosulfate sulfur transfer protein and the thiosulfate/sulfate binding proteins appeared to be transcriptionally linked to a gene for a terminal oxidase. Several other proteins involved in sulfur oxidation have also been identified including a sulfur dioxygenase, a rhodanase and a 40 kD outermembrane protein. However, which proteins are required for the oxidation of different RISCs is far from being understood. Furthermore, studies on the biochemistry of sulfur oxidation including evidence for a bc~1~complex and several cytochrome oxidases (bdand ba~3~) that are produced in higher concentrations when grown on sulfur than iron have been reported \[[@B10]\]. A model in which the components of iron and sulfur oxidation both feed electrons into an aa~3~-type cytochrome coxidase has been proposed to account for biochemical and gene expression data \[[@B96]\]. There are indications that there may be more uniformity in the pathways used by at least the Gram-negative sulfur-oxidizing bacteria \[[@B30],[@B76]\] than there is in iron oxidation pathways. This probably does not stretch to the sulfur-oxidizing archaea where thiol independent systems have been isolated. Irrespective of the pathway used, the ultimate oxidation product of RISCs is sulfate and this results in a decrease in pH. ### 7.3 Other sources of energy Soluble metal ions are frequently present fairly high concentrations in highly acidic environments. Metal ions which exist in more than one oxidation state and which have redox potentials that are more negative than the O~2~/H~2~O redox couple, have the potential to serve as electron donors for acidophilic bacteria. An At. ferrooxidans-like bacterium was reported to directly oxidize Cu^+^to Cu^2+^\[[@B51],[@B53]\] and U^4+^to U^6+^under aerobic conditions and that these oxidation reactions were coupled to CO~2~fixation \[[@B24]\]. However, whenever ferric iron is present, it is difficult to unequivocally demonstrate the biological oxidation of the metal as opposed to chemical oxidation of the metal by ferric iron. Similarly it has been reported that Mo^5+^can be oxidized to Mo^6+^and a molybdenum oxidase has been isolated from cell extracts of At. ferrooxidans\[[@B85]\]. The potential also exists that the oxidation of oxyanions such as As^3+^(AsO~2~^-^) to As^5+^(AsO~4~^3-^) can serve as an alternate electron donor for acidophilic organisms \[[@B83]\]. An analysis of the At. ferrooxidansATCC23270 genome revealed that as many as eleven cytochromes cwere present \[[@B98]\]. One cytochrome cwas specific for growth on sulfur, three were specific for growth on iron and several were produced on both substrates. The large number of cytochrome cmolecules might also be a reflection of the versatility of electron donors (and electron acceptors) that the bacterium is capable of using. The type strain of At. ferrooxidansATCC23270 as well as the two other At. ferrroxidansstrains tested were found to grow by hydrogen oxidation but not At. thiooxidansor L. ferrooxidans\[[@B27]\]. When growing on hydrogen they had a broad pH optimum of pH 3.0 to 5.8 with no growth occurring at pH\<2.2 or pH\>6.5. Hydrogen oxidation appeared to be repressed by the presence of S^0^, Fe^2+^and sulfidic ore. In a later study, only one of six At. ferrooxidansstrains tested could use hydrogen as an electron donor to support CO~2~fixation and cell growth with oxygen as electron acceptor \[[@B58]\]. There is a possibility that some isolates of the genes Leptospirillummight be able to use hydrogen as an electron donor although this has not yet been demonstrated. 8. Relationship to oxygen and alternate electron acceptors ---------------------------------------------------------- The chemolithotrophic acidophiles require large quantities of energy to support their autotrophic lifestyle. As may be expected, their most commonly used terminal electron acceptor is oxygen as this is energetically the most favourable option. As described earlier, the redox potential of the Fe^2+^/Fe^3+^couple is almost as positive as that of O~2~/H~2~O and consequently ferric iron is a potentially suitable alternate electron acceptor. For an autotrophic acidophile to be able to use ferric iron as electron acceptor it must be capable of using RISCs or molecules other than ferrous iron as an electron donor. The oxidation of sulfur and tetrathionate coupled to ferric iron reduction under anaerobic conditions has been shown to occur in the case of At. ferrooxidans\[[@B88]\]. It has also been shown that several though not all isolates of this bacterium can grow by using the H~2~- or S^0^-coupled reduction of ferric iron \[[@B58]\]. Other autotrophic sulfur-oxidizers like At. thiooxidansand At. caldusare apparently unable to catalyze the reduction of ferric iron in the absence of air \[[@B35]\]. Besides the ability to use ferric iron, the At. ferrooxidansis also able to reduce Mo^6+^, Cu^2+^and Co^2+^when using elemental sulfur as an electron donor \[[@B86],[@B87]\]. At. ferrooxidansand At. thiooxidanshave been reported to reduce V^5+^to V^4+^, however, whether the oxidized vanadium served as an electron acceptor for respiration was unclear as the shake flasks were aerated \[[@B11]\]. As described earlier, the large variety of cytochrome cmolecules might reflect the versatility of At. ferrooxidansto use a wide variety of electron acceptor. The potential to grow by ferric iron respiration is even greater amongst the extremely acidophilic heterotrophs since ferric iron reduction can be coupled to the oxidation of many organic compounds. Indeed some Acidiphiliumspecies are able to reduce ferric iron even under aerobic conditions such as in shake flasks and on the surface of agar plates, although ferric iron reduction is enhanced when the oxygen concentrations are relatively low \[[@B44]\]. Furthermore, not only soluble but also insoluble amorphous or crystalline minerals such Fe(OH)~3~and jarosite can be reductively solubilized by AcidiphiliumSJH using ferric iron \[[@B12]\]. Ferric iron respiration has the advantage of regenerating additional ferrous iron electron donor for the iron-oxidizing obligate autotrophs should aerobic conditions again prevail. 9. Acidophilic properties ------------------------- From an industrial perspective it is essential that biomining microorganisms are able to grow at low pH and tolerate high concentrations of acid. Two important reasons for this are to enable iron cycling and to permit reverse electron transport to take place. A low pH is required for the iron cycle whereby ferrous iron serves as an electron donor under aerobic conditions and ferric iron as an energetically favourable alternate electron acceptor if the concentration of oxygen falls. This has been described above. Ferric iron is almost insoluble at a neutral pH, whereas in acid solutions its solubility is increased. The possibility of using ferric iron as an alternate electron acceptor is therefore readily available to acidophiles but less available to aerobic neutrophiles or moderate acidophiles because ferric iron is almost totally insoluble in neutral, aerobic environments. The external pH of the environment in which extreme acidophiles such as biomining microbes grow is low (e.g. pH 1.0--2.0), whereas the internal cellular pH remains close to neutral \[[@B20]\]. This difference results in a steep pH gradient across the cell membrane. This pH gradient is important for nutritional purposes, especially when using a weak reductant such as ferrous iron as an electron donor. Autotrophic organisms have a high requirement for compounds such as NAD(P)H to reduce their carbon source (CO~2~) to produce the sugars, nucleotides, amino acids and other molecules from which new cell mass is synthesized. Heterotrophic bacteria do not have as high a demand for NAD(P)H as their carbon source is more reduced than CO~2~and hydrogen atoms removed from their source of nutrition may be used to satisfy their lower NAD(P)H requirement. Chemolithotrophic autotrophs require a large transmembrane proton gradient to generate the required proton motive force to energise the synthesis of NAD(P)H. This process is known as reverse electron transport or the \'uphill\' electron transfer pathway \[[@B9]\]. Although this phenomenon has not been studied in many iron- or sulfur-oxidizing chemolithotrophs, strong evidence has been presented that when grown on iron, At. ferrooxdianscontains a unique cytochrome bc~1~complex that functions differently from the bc~1~complex used during the oxidation of sulfur and is specifically involved in the \'uphill\' pathway \[[@B28]\]. One way of viewing this is that growth in acid solutions is a nutritional necessity as a large transmembrane pH gradient is required to produce the hydrogen atoms needed to reduce CO~2~to cell mass. 10. Adaptability and ability to compete in a non-sterile environment -------------------------------------------------------------------- In many industrial processes that are dependent on the use of microorganisms it is important that the process is kept largely free from contamination by undesired organisms. From the description of biomining processes given in the introduction it is clear that \'non-sterile\' open stirred tanks or heaps exposed to the environment are used. Such processes are susceptible to \'contamination\' by microorganisms present on the ores, concentrates, inorganic nutrient solutions, water air etc. Given the huge volumes of mineral that have to be processed, the relatively low value of the product and nature of a mining environment the cost-effective prevention of contamination would be impossible to achieve. Fortunately this is not required. The aim of the process is the biodegradation of the mineral or concentrate and one seeks organisms that are able to do this most effectively. Those microorganisms that are able to degrade the mineral most effectively are also those that grow the quickest and therefore have the fastest doubling times. In a continuous-flow process such as provided by a series of completely mixed leaching tanks, microorganisms in the tanks are continually being washed out. There is thus a strong positive selection for microbes that grow most effectively on the mineral as those microbes that grow and divide the fastest are subjected to less wash out and will dominate the microbial population in the biooxidation tanks. There are few biological fermentation processes that share this advantage with another notable example being activated sludge sewage treatment process where organisms with the capacity to grow most effectively on the waste in the water are selected. Previous unreported research experience by the author has found that after a period of operation, the metabolic capabilities of a population of biomining organisms may improve out of all recognition from the culture originally inoculated into the tanks. One would predict that natural populations of microorganisms are adapted for survival under the highly variable feast or famine conditions that are experienced in nature rather than the optimized, controlled conditions of a biooxidation tank. Early experiments on gold-biooxidation were carried out in a series of three or four continuous-flow, aerated, stirred tank reactors. As these reactors are expensive to construct and operate, the rate of concentrate decomposition has an important effect on the economics of the process \[[@B23]\]. The initial process was very slow because unadapted cultures of biooxidation bacteria were probably not tuned to rapid growth and possibly also because they were sensitive to the arsenic released from the arsenopyrite. Initially a retention time of over twelve days was required for sufficient biooxidation to allow more than 95% gold recovery \[[@B73]\]. However, a period of selection of about two years in a laboratory scale and then pilot plant scale continuous flow process resulted in a reduction in the retention time of concentrate in the reactors to seven days. During the first two years of operation in a full-scale continuous-flow biooxidation plant the growth rate of the bacteria had improved still further so that the retention time had been reduced to about 3.5 days. At the same time the solid concentration in the liquor was increased from 10 to 18% so that the same equipment could be used to treat almost four times the amount of concentrate per day as initially. This process was developed by Gencor SA \[[@B23],[@B93]\] and registered as the Biox process. 11. Metal tolerance and resistance ---------------------------------- An important characteristic of the acidophilic chemolithotrophs is their general tolerance of high concentrations of metallic and other ions. The levels of resistance of several acidophilic bacteria and archaea to As^3+^, Cu^2+^, Zn^2+^, Cd^2+^and Ni^+^have recently been reviewed and will not be covered here in detail \[[@B25]\]. As may be predicted, levels of resistance show considerable strain variation. Adaptation to high levels of metal resistance on exposure to a metal is likely to be responsible for much of the variation. At. ferrooxidansappears to be particularly resistant to metals and the bacterium has been reported to grow in medium containing Co^2+^(30 g/l), Cu^2+^(55 g/l), Ni^2+^(72 g/l), Zn^2+^(120 g/l), U~3~O~8~(12 g/l) and Fe^2+^(160 g/l). In a comparative study of two At. ferrooxidans, two L. ferrooxidansand an At. thiooxidansstrain, it was found that At. ferrooxidansand L. ferrooxidanswere approximately equally resistant to Cu^2+^, Zn^2+^, Al^3+^, Ni^2+^and Mn^2+^, but that L. ferrooxidanswas more sensitive (\<2 g/l) than At. ferrooxidansto Co^2+^\[[@B77]\]. At. thiooxidanswas sensitive to less than 5 g/l of all the cations used in the comparative study with the exception of Zn^2+^(10 g/l). No studies have been carried out on the molecular mechanisms of metal resistance in any of these bacteria. Genome sequencing data on At. ferrooxidansand Leptospirillumtype II plus work from many other groups suggest that metal resistance is due to a combination of genes that are probably present on the chromosomes of most isolates of a bacterial species and mobile genes acquired by specific isolates of a species. An example of genes present on the chromosomes of most species of a genes are the efflux genes for arsenic \[[@B15]\], copper, silver cadmium and several metal cations in At. ferrooxidans(genome sequence data, \[[@B3]\]). Another example of a resistance mechanism that might be present in all members of a species because it is associated with general cell physiology is the polyphosphate mechanism for copper resistance of At. ferrooxidans\[[@B1]\]. These workers presented a model whereby the hydrolysis of polyphophates resulted in the formation of metal-phosphate complexes that are transported out of the cell enhancing resistance to the metal. Mobile genes for metal or metalloid resistance that might be present in certain isolates but not others of the same species are genes present on plasmids or transposons. These genes may be recruited from the horizontal gene pool by the acquisition of a plasmid or the insertion of metal resistance containing transposons into either the chromosome or a plasmid. For example, when ten At. ferrooxidansisolates were screened for Hg^+^resistance, three of the strains contained DNA that hybridized to a Tn501 mergene probe \[[@B82]\]. Bacteria carrying the resistance genes were in general 3--5 times more resistant to Hg^2+^than strains that did not have mergenes. The mergenes of the E-15 strain of At. ferrooxidanswere cloned and sequenced and truncated transposon Tn7-like fragments were found in the vicinity \[[@B39],[@B40]\]. Codon usage analysis suggested that the mergenes had originated from an organism different from At. ferrooxidans\[[@B40]\]. A Tn21-like transposon (Tn5037) that contains mercury resistance genes was isolated from another strain of At. ferrooxidansG66 \[[@B46]\]. Some strains of At. ferrooxidansappear to contain a mercury resistance mechanism that is so far unique to the species. Mercury volatilization in these strains was dependent on Fe^2+^as an electron donor but not NADPH as found with other mercury resistance mechanisms \[[@B42]\]. The cytochrome coxidase appeared to deliver electrons directly to mercury (Figure [1](#F1){ref-type="fig"}) \[[@B84]\]. It was possible to take At ferrooxidansSUG2.2 cells already resistant to 6 μM Hg^2+^and adapt them by successive cultivation to produce At. ferrooxidansstrain MON-1 that was resistant to 20 μM Hg^2+^. This property was maintained after several rounds of cultivation on iron in the absence of Hg^2+^. Interestingly, rusticyanin from mercury resistant cells enhanced Fe^2+^-oxidation actitity of plasma membranes and activated Fe^2+^-dependent mercury volatilization activity \[[@B42]\]. This supports the view of Rohwerder et al. \[[@B75]\] that rusticyanin serves as a channel of electrons from iron. Comparison of cytochrome coxidases from At. ferrooidansstrains that are resistant to Hg^2+^, Mo^5+^, sulfite and 2,4 dinitrophenol with sensitive strains led the authors to suggest that different cytochrome coxidases.might be responsible for resistance to different substances by related mechanism \[[@B84]\]. An example of where resistance genes may be acquired from the horizontal gene pool when needed are the arsenic resistance genes recruited by At. caldus\[[@B21],[@B90]\] and L. ferriphilum(unpublished). These two bacteria have been shown to dominate the biooxidation tanks used to treat gold-bearing arsenopyrite concentrate at the Fairview mine \[[@B71]\]. When microorganisms capable of rapidly oxidizing arsenopyrite concentrate in continuous flow aeration tank were being selected, the rates of oxidation were initially slow. One of the reasons for this is that the organisms were sensitive to arsenic. Once arsenic levels had built up in solution above 1 g/l total arsenic, the process slowed and arsenic had to be precipitated and removed from solution by raising the pH. After arsenic removal and subsequent aeration, biooxidation rates increased until the concentration of arsenic in solution again built up and the arsenic was reprecipitated. After almost two years of selection in continuous-flow laboratory and pilot scale tanks, the microorganisms had become sufficiently resistant to the 13 g/l total arsenic in solution for arsenopyrite biooxidation to take place without the need to remove the arsenic. Unfortunately, the original unadapted arsenic sensitive culture was not maintained and therefore was not available to compare with the highly arsenic resistant culture present in the commercial Biox^®^plant at the Fairview mine in 1996 when arsenic resistance mechanisms were investigated (approximately ten years after it had been commissioned). L. ferriphilumand At. caldusstrains were isolated from the Biox^®^tanks and their arsenic resistance mechanisms examined and compared with those of the same species of bacterium that were not known to have been previously exposed to arsenic. Studies on arsenic resistance genes of six strains of isolates of At. calduswere carried out, three with known exposure to arsenic and three without. Of the three strains previously exposed to arsenic, one strain originated from the Biox^®^plant at Fairview, another from a pilot plant oxidixing arsenopyrite at the University of Cape Town and third from a culture used to treat a nickel-containing ore but which was derived from same culture used in the Fairview plant. Of the three At. caldusisolates not known to have been exposed to arsenic, one originated from Australia and two from the United Kingdom. DNA-DNA hybridization experiments indicated that all six strains contained a set of arsenic resistance genes present on their chromosomes. However, the three arsenic resistant strains contained arsenic resistance genes in addition to those present in all strains. The arsenic resistance genes were present on a transposon belonging to the Tn21family that must have been acquired from the horizontal gene pool. All three resistant strains contained a copy of the TnAtcArstransposon (Figure [3](#F3){ref-type="fig"}) and at least one strain had an additional incomplete copy of the transposon \[[@B90]\]. The arsenic resistance genes were arranged in an unusual manner with the arsA(ATPase) and arsD(regulator and provision of arsenite) being duplicated. In the At. caldusstrain isolated from the nickel plant, the arsAand arsDduplication was absent. Efforts are being made to introduce TnAtcArsinto arsenic sensitive strains of At. caldusto determine the contribution of TnAtcArsto arsenic resistance of the host. A question to be addressed, is from where did the TnAtcArsacquired by the arsenic resistant strains originate? DNA sequencing data indicated that the closest relative to the arsgens is on a transposon present in a heterotrophic bacterium Alcaligenes faecalis. The percentage amino acid sequence identity of proteins associated with arsenic resistance on the two transposons was high (70--95%) but not identical. This suggests that the two transposons originated from the same ancestral plasmid. However, the differences are sufficient to suggest that the two transposons have evolved independently for many years (difficult to allocate a time scale) and that At. caldusand A. faecalisdid not originate from the same gene pool at the time that the arsenic resistant At. caldusstrains were exposed to high levels of arsenic in the early 1980\'s. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### The arsenic resistance gene containing transposon, TnAtcArs, present in highly arsenic resistant strains of At. caldus\[90\]. The arsenic resistance genes are located between the inverted repeat sequences (IR), resolvase (tnpR) and transposase (tnpA) genes of the Tn21-like transposon. R, arsenic resistance regulator; C, arsenate reductase; D, upper-limit arsenic regulator; A, arsenite efflux-dependent ATPase; 7, ORF with a NADH oxidoreductase domain; 8, ORF with a CBS-like domain; B, membrane arsenite efflux transporter. ::: ![](1475-2859-4-13-3) ::: The account of arsenic resistance gene acquisition just described is an illustration of an advantage to be gained by the bioleaching and biooxidation processes being non-sterile, open systems. New organisms will continually enter the system and the iron- and sulfur-oxidizing microbes present will have the opportunity of accessing the horizontal gene pool that these organisms contain and that are selected by growth conditions. 12. Conclusion -------------- The solubilization of metals from minerals or their concentrates is believed to be largely a chemical process that is due to the action of ferric iron and protons depending on the mineral being treated. Like all chemical processes, the rate of reaction is affected by temperature. Some difficult-to-degrade minerals need to be leached at higher temperatures than others for the leaching reactions to proceed at an economically viable rate. Since microorganisms are responsible for producing the leaching reagents and because contact between the microbes and the mineral speeds up the process, there is a need for microorganisms to be able to produce the leaching reagents at a variety of temperatures. As would be expected, the types of microorganisms present in processes used for the recovery of metals vary hugely depending on the temperature at which the process is carried out. Commercial processes that operate at temperatures from ambient to 40°C are dominated by Gram-negative bacteria with some Ferroplasma-like organisms being present if the pH drops below about pH 1.3. There is some overlap with bacteria that dominate processes that operate at 40°C with those at 45--55°C (e.g. L. ferriphilumand At. caldus), but there are also some clear differences. In particular Gram-positive bacteria belonging to the genus Sulfobacillusappear to play a significant role at the higher temperatures and archaea of the Ferroplasmatype are more frequently found. In contrast, microorganisms present in processes that operate at 75--80°C are all archaea. Although there are no commercial processes currently operating in the range 60--70°C suitable organisms almost certainly exist and are likely to be present in low pH hot sulfur springs. The variation in microorganism present in a bioleaching process appears to be more dependent on temperature than on the type of iron-and sulfur-containing mineral being oxidized or on whether tank or heap reactors are being used. In spite of the large variety of potential organisms that can be used, the microbes that play the most important roles tend to have certain properties in common. They obtain their energy by the oxidation of either iron or reduced inorganic sulfur compounds. Although some microorganisms are capable of using both energy sources, a combination of iron-oxidizing and sulfur-oxidizing microbes often works best. The production of sulfuric acid and the need to keep the most important mineral-oxidizing agent (ferric iron) in solution means that the organisms are acid tolerant. The iron- and sulfur-oxidizing organisms are, in general, autotrophic and do not require to be provided with an external carbon source. When in pure culture, some grow better with small amounts of yeast extract or if aerated with CO~2~-enriched air. However, when growing in a mixed microbial consortium, cross-feeding appears to take place so that an extra source of carbon is not required. The microorganisms tend to be resistant to high concentrations of metal ions and where this is lacking they have demonstrated a remarkable ability to become metal-resistant. At least some of this metal resistance is due to the acquisition of metal genes from the horizontal gene pool. At. ferrooxidansis the first bacterium that was recognized as being present in bioleaching environments. This bacterium has been more extensively studied than any other biomining organism and was also the first to have its genome sequenced \[[@B3]\]. Although this bacterium is readily isolated from acid mine drainage and heap reactors operating below 40°C, it appears not to be the most important leaching organism in most high-rate commercial processes. In depth studies on several of the other types of biomining organisms is therefore also needed. The recent gapped genome sequences of L. ferriphilumand a strain of Ferroplasmawere assembled during an environmental metagenome project on the organisms present in acid mine drainage \[[@B92]\]. This and other genome sequencing projects being planned should provide assistance in expanding our knowledge on other important biomining microbes. Acknowledgements ================ The author acknowledges funding from the University of Stellenbosch, BHP-Billiton, the National Research Foundation (Pretoria) Gun2053356 and the EU framework 6 BioMinE project 500329.	PubMed Central	2024-06-05T03:55:57.394828	2005-5-6	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1142338/", "journal": "Microb Cell Fact. 2005 May 6; 4:13", "authors": [ { "first": "Douglas E", "last": "Rawlings" } ] }
PMC1142339	Introduction ============ TRPV1 is a Ca^2+^permeable nonspecific cation channel, located on peripheral sensory neurons, that serves as a molecular detector for heat, capsaicin, protons, and endovanilloids \[[@B1]-[@B4]\]. Moreover, its role as a heat sensor (activation threshold of \~43°C), its influence from trophic/inflammatory agents (i.e. nerve growth factor, bradykinin, prostaglandins, etc.) and its vasodilatory effect on small vessels (by releasing CGRP) make TRPV1 an essential component of the pain pathway \[[@B3],[@B5]-[@B11]\]. In peripheral nerves, insulin does not promote its typical metabolic effects of glucose and amino acid uptake \[[@B12]\]. However, physiologic concentrations can act as neurotrophic factors, in combination with nerve growth factor (NGF), to stimulate neurite outgrowth and survival of both sensory and sympathetic neurons \[[@B13]-[@B15]\]. Furthermore, it has been proposed that insulin and IGFs exert trophic influences on the same neurons that are responsive to NGF by sharing common signaling pathways \[[@B13]\]. Though it is widely believed these neural disturbances are secondary to hyperglycemia \[[@B16]\], this remains controversial. Results from the Diabetes Control and Complications Trial show that intensive glycemic control for 5 years reduced the incidence of neuropathy by 60% in Type I patients \[[@B17]\]. However, the fact that strict glucose regulation does not completely prevent diabetic peripheral neuropathy suggests additional mechanisms as a result of insulin deficiency may be involved. Experiments in humans show that non-diabetic, normoglycemic subjects have heat thresholds that correlate positively with insulin sensitivity \[[@B18]\]. Furthermore, circulating levels of IGFs and NGF are reduced in diabetes and to a greater degree in patients with thermal sensory deficits characteristic of neuropathy \[[@B19]-[@B21]\]. Growth factor replacement therapy improves neuronal regeneration \[[@B22],[@B23]\] and insulin, under conditions of rapid infusion, can restore decreased neurogenic vasodilation \[[@B24],[@B25]\]. Taken together, these studies suggest that insulin and trophic factor deficiencies may alter sensory nerve function \[[@B26]\]. In this study we demonstrate that insulin and IGF-I enhance TRPV1-mediated membrane currents by enhancing receptor sensitivity and translocation from cytosol to plasma membrane in heterologous expression systems and cultured dorsal root ganglion (DRG) neurons in a PKC-dependent manner. Preliminary results of this study have been published elsewhere in an abstract form \[[@B27]\]. Materials and Methods ===================== Oocyte Electrophysiology ------------------------ Oocytes were isolated from tricaine-anesthetized Xenopus laevisand separated from the follicular layer after incubation with collagenase (1--2 mg/ml). 3--7 days following cRNA injections (50--70 nl, 1 μg/μl), oocytes were stored at 18°C and used for experiments. Double electrode voltage clamp (-60 mV) was performed using a Warner amplifier (Warner Instruments, OC725C, Hamden, CT, USA) with 100% d.c. gain, digitized and stored on video tape. Experiments were performed at 21--23°C. Oocytes were placed in a Perspex chamber superfused (5--10 ml min^-1^) with Ca^2+^-free Ringer solution containing (in mM): 100 NaCl, 2.5 KCl, 5 HEPES, pH 7.35. Within figures, current traces were shown as initial response to agonist juxtaposed on currents recorded following incubation with insulin, IGF-I and various inhibitors. Time course experiments consisted of current recording during capsaicin application in the presence of either control (first 20 min) or insulin (last 15 min) containing solutions. In all experiments, fold increase was calculated as agonist induced current amplitude in the presence of trophic factors and inhibitors divided by the agonist induced initial current amplitude under control conditions before their bath application. I-V relationships were measured using 500-ms voltage ramps from -80 mV to +80 mV. DRG Culture and Electrophysiology --------------------------------- Dorsal root ganglia were isolated from embryonic day 18 (E18) rats, triturated and cultured for 5--7 d in Neurobasal/B-27 (Life Technologies; Grand Island, NY) + 10% fetal bovine serum (FBS) on poly-D-lysine-coated glass coverslips. For perforated patch recording, the bath solution contained (in mM): 140 NaGlu, 2.5 KCl, 10 HEPES, 2 MgCl~2~, 1 EGTA, pH 7.35 and the pipette solution contained (in mM): 130 NaGlu, 10 NaCl, 2.5 KCl, 10 HEPES, 1 MgCl~2~, 0.2 EGTA, and amphotericin B (240 μg/ml). For cell-attached recording, the bath solution contained (in mM): 140 Kglu, 10 NaCl, 10 HEPES, 1 EGTA, 2 MgCl~2~, pH 7.35 and the pipette contained (in mM): 140 NaGlu, 2.5 KCl, 10 HEPES, 2 MgCl~2~, pH 7.35. Currents were recorded using a WPC 100 patch clamp amplifier (E. S. F. Electronic, Goettingan, Germany) or Axopatch 200B (Axon Instruments, Union City, CA). Data were digitized (VR-10B, Instrutech Corp.; Great Neck, NY) and stored on video tape. For analysis, data were filtered at 2.5 kHz (-3db frequency with an 8-pole low-pass Bessel filter, Warner Instruments, LPF-8) and digitized at 5 kHz. Data analyses were performed on continuous stretches greater than 20 s from patches that contained one or two channels. Single channel current amplitude and P~o~were calculated from all point amplitude histograms fitted with Gaussian functions (Microcal Origin; Northampton, MA). Ca^2+^Imaging ------------- Cells were grown on glass coverslips, then incubated with 5 μM Fluo-4AM (Molecular Probes; Eugene, OR) for 20 min at 37°C, washed with physiological buffer \[(in mM): 140 NaCl, 5 HEPES, 2 CaCl~2~, 1 MgCl~2~, 2.5 KCl, pH 7.35\], and treated with 500 nM capsaicin +/- 1 μM insulin immediately prior to analysis by confocal microscopy (Fluoview, Olympus; Mellville, NY). Data for each cell was quantified as the fluorescence after treatment (F) divided by initial fluorescence (F~o~) at t = 0 (Fluoview software). Transfection of Cells and Cell Culture -------------------------------------- HEK 293 cells, which endogenously express IGF-I receptors \[[@B29]\] were cultured in Dulbecco\'s modified Eagle\'s medium (high glucose) supplemented with 10% fetal bovine serum, 50 units/ml penicillin, and 50 μg/ml streptomycin (Invitrogen; Carlsbad, CA) and maintained under 95% air/ 5% CO~2~at 37°C. Cells were transiently transfected with wild type TRPV1, TRPV1eGFP, TRPV1 S502/S800A mutant (gifted by D. Julius, N. Kadei, M. Tominaga, respectively), or TRPV1-V5-His tagged plasmid using Lipofectamine2000 reagent (Invitrogen) according to manufacturer\'s protocol. TRPV1-V5-His tagged plasmid was constructed in pcDNA3.1 vector using Topo Cloning kit (Invitrogen). To determine the membrane translocation of TRPV1, confocal images (1 μm sections) of GFP-tagged TRPV1 fluorescence was obtained. The intensity of the brightest membrane fluorescence was selected and quantified using MCID imaging software (Imaging Research Inc; St. Catherines, ONT) Determination of Surface TRPV1 ------------------------------ After 36 h of TRPV1 transfection, the cells (grown on 100 mm tissue culture plate) were insulin-treated and biotinylated with membrane impermeable NHS-LC-biotin (1.5 mg/ml in PBS; Pierce, Rockford, IL) as per manufacturer\'s protocol. Labeled cells were lysed in RIPA buffer and immunoprecipitated overnight with 10 μl of goat anti-VR1 polyclonal antibody/ Protein A/G Plus-agarose (Santa Cruz Biotechnology, Santa Cruz, CA) following manufacturer\'s protocol. Immunoprecipitates were eluted with 2X SDS sample buffer, separated on 7.5% SDS-PAGE, and transferred to PVDF membrane. To detect total TRPV1, blots were probed with rabbit anti-VR1 polyclonal antibody (1:1000; Affinity Bioreagents, Golden, CO). Specific antibody binding was detected using HRP-conjugated anti-rabbit secondary antibody (1:20,000; Jackson Immunoresearch, West Grove, PA) and Super Signal reagent (Pierce). To detect surface TRPV1, blots were stripped and probed with neutravidin-HRP (1:20,000; Pierce). Chemiluminescence was captured in Hitachi CCD Bio Genetic Systems after exposing the blot to Super Signal reagent (Pierce). Data were quantitated using LabWorks analysis software (UVP Inc., Upland, CA) and surface TRPV1 was normalized using total TRPV1. TRPV1-V5-His transfected cells were biotinylated as above and purified using Ni-NTA agarose (Qiagen, Valencia, CA) as per manufacturer\'s protocol. Immunohistochemistry -------------------- Cells were plated on glass coverslips coated with 40 μg ml^-1^poly-L-lysine, treated with insulin/IGF/PDBu after 36 h of transfection, and fixed with 4% paraformaldehyde for 1 h. The cells were probed with rabbit anti-VR1 antibody (1:10,000; Affinity Bioreagents) followed by rhodamine red-X conjugated anti-rabbit secondary antibody (1:50; Jackson Immuno Research) following method described by Santa Cruz Biotechnology. Confocal images (1 μm sections) were captured using 570 nm laser. The brightest membrane fluorescence was selected from control and treated cells and quantified using MCID imaging software (Imaging Research Inc.). Unless otherwise stated, all chemicals were obtained by Sigma (St. Louis, MO). Data are given as mean ± s.e.m. and statistical significance (set at P \< .05) was evaluated using Student\'s t-test or one-way ANOVA. Results ======= Insulin and IGF-I potentiate TRPV1 current in oocytes ----------------------------------------------------- First, we tested whether insulin and/or IGFs influence TRPV1-mediated membrane currents. The effect of insulin and IGF-I on cloned TRPV1 was characterized through dual-electrode voltage clamp (V~m~= -60 mV) of Xenopusoocytes injected with TRPV1 cRNA. Insulin (1 μM) significantly potentiated the response to capsaicin (500 nM) greater than 2 fold (Fig. [1a](#F1){ref-type="fig"}). The time course of insulin-induced potentiation of capsaicin response shows a significant increase in current amplitude for 5, 10 and 15 minute incubation periods compared to initial levels (Fig. [1b](#F1){ref-type="fig"}). Dose response curves of capsaicin show that both the sensitivity (EC~50~shifted from 0.9 to 0.6 μM) and the maximal response (normalized to 1 μM capsaicin induced current before insulin application) increased (from 1.8 to 2.4) after incubation of oocytes with insulin (circles) for 5 min (Fig. [1c](#F1){ref-type="fig"}). A representative current-voltage relationship (I-V curve) confirms the outward rectification characteristic of TRPV1 current and shows that, in the presence of insulin, amplitude is increased at both positive and negative potentials (Fig. [1d](#F1){ref-type="fig"}). In addition, current potentiation was also seen with heat evoked responses (Fig. [1e](#F1){ref-type="fig"}) and protons (Fig. [1f](#F1){ref-type="fig"}), an endogenous TRPV1 agonist. These experiments show that insulin is increasing TRPV1-mediated currents and making the receptor more responsive to exogenous and endogenous activators. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Insulin and IGF-I potentiate TRPV1 currents in XenopusOocytes. a, A representative dual-electrode whole-oocyte experiment showing potentiation of capsaicin induced TRPV1 currents following 15 min incubation of insulin (see Methods). b, Time course for capsaicin induced current (■: 500 nM capsaicin, 30 sec application) before and after insulin exposure at 5 (P\< .05), 10 (P\< .01), and 15 min (P\< .01) incubation times (n= 7). c, Dose response curve of capsaicin before (squares) and after 5 μM insulin (circles) application, the currents were normalized to 1 μM capsaicin before insulin application (n = 2 to 4, before and after insulin). Both the sensitivity (EC~50~shifted from 0.9 to 0.6 μM) and maximal response increased (1.8 to 2.4). d, Representative I-V relationships (1 s, 1 mV step ramp from -80 to +80 mV) before and after insulin treatment that demonstrates the outward rectification typical of TRPV1 channels. e, Potentiation of heat-induced currents by insulin. f, Potentiation of pH induced currents by insulin. g, Potentiation of pH induced currents by IGF-I. h, Tyrphostin A47, an IR / IGFR antagonist, blocked insulin potentiation. i, Summary graph showing fold increase in TRPV1 currents following 10 min incubation with control, insulin (cap: n= 5, P\< .01; pH: n= 4, P\< .01), IGF-I (n= 3, P\< .01), and insulin + tyrophostin A47 (n= 3, P\< .01). Results are expressed as increase in current amplitude relative to initial capsaicin or pH response. ::: ![](1744-8069-1-17-1) ::: Oocytes endogenously express insulin-like growth factor receptors (IGFR) rather than insulin receptors (IR) \[[@B30]\]. Insulin binds to IGFR with 100--1000 times lower affinity than IR. On the other hand, IGF-I binds to IGFR with much higher affinity, so we tested whether lower concentrations of IGF-I could potentiate TRPV1. Indeed, 20 nM IGF-I potentiated pH currents to a similar degree as 1 μM insulin (Fig. [1g](#F1){ref-type="fig"}). Potentiation was dramatically reduced by the addition of tyrphostin A47 (100 μM; 10 min), a selective inhibitor for the insulin family of receptor tyrosine kinases (Fig. [1h](#F1){ref-type="fig"}). Together these data suggest that in oocytes, insulin and IGF-I are acting through IGFR to potentiate TRPV1 (Fig. [1i](#F1){ref-type="fig"}). Insulin and IGF-I potentiate native TRPV1 current in DRG neurons ---------------------------------------------------------------- Since insulin and IGF-I functions vary with cell type, it was important to establish that current potentiation also occurred in sensory neurons that express native TRPV1 \[[@B15],[@B21]\]. Therefore, we determined the effect of insulin and IGF-I on capsaicin-induced TRPV1 responses using cultured DRG neurons. For whole cell recordings, the perforated patch technique was used to prevent desensitization and tachyphylaxis, minimize intracellular disruption, and maintain intact signaling cascades. In this setting, insulin (1 μM) induced a potentiation of the capsaicin (100 nM) response (Fig. [2a](#F2){ref-type="fig"}), which returned to control levels 20 minutes after its removal (Fig. [2b](#F2){ref-type="fig"}, inset). Like oocytes, similar results were seen with IGF-I, signifying that insulin was binding IGFR in DRG neurons (Fig. [2c](#F2){ref-type="fig"}). ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Insulin and IGF-I potentiate native TRPV1 currents in DRG neurons. a, Capsaicin (100 nM, applied at 2 min intervals) response was enhanced by insulin (1 μM) under perforated patch conditions. b, Time course of insulin induced potentiation of capsaicin currents (■: 500 nM capsaicin, 20 sec application) (n= 4, P\< .01 at 5, 10, and 15 min incubation times). Inset shows the time course for experiment in Fig. 2a. c, Summary graph showing fold increase in TRPV1 currents following 10 min treatment with control (n= 15), insulin (1 μM; n= 11, P\< .01) and IGF-I (20 nM; n= 4, P\< .01). d, Under confocal microscopy, insulin (1 μM, 2 min) potentiated intracellular Ca^2+^rise in response to 500 nM capsaicin (n= 10, P\< .05). e, Capsaicin induced single channel current activity recorded at +60 mV was increased by exposure of cell to insulin (1 μM). f, Mean capsaicin induced open probability (P~o~) in the absence and presence of insulin (n= 5, P\< .05). g, Mean current amplitude observed from single channel recordings (n= 5). ::: ![](1744-8069-1-17-2) ::: To confirm electrophysiological experiments, measurements of Ca^2+^-uptake were performed using confocal microscopy (Fig. [2d](#F2){ref-type="fig"}). In DRG neurons, capsaicin (500 nM) produced a Ca^2+^influx that was increased in the presence of insulin. This shows that insulin elevated TRPV1-regulated Ca^2+^mobilization. Together this data, confirm that the potentiation seen with cloned TRPV1 in oocytes is present in native TRPV1 expressing peripheral sensory neurons as well. Insulin increases single channel activity ----------------------------------------- To understand the underlying molecular mechanism responsible for whole-cell potentiation, single channel currents were recorded in cell-attached patches from DRG neurons. Under this configuration, extracellular insulin would require receptor mediated signal transduction to facilitate changes in TRPV1 channel function recorded within the patch area (Fig. [2e](#F2){ref-type="fig"}). Channel open probability (P~o~), a measure of the time the channel spends in the open state, induced by capsaicin (10 nM) increased following bath-application of insulin (1 μM), outside the patch area (Fig. [2e](#F2){ref-type="fig"}, lower graphs). P~o~significantly changed from 0.15 ± 0.05 in control conditions to 0.43 ± 0.14 after insulin (Fig. [2f](#F2){ref-type="fig"}), without altering the single channel amplitude (Fig. [2g](#F2){ref-type="fig"}). Since insulin application was outside the patch, these data support intracellular signaling, as opposed to direct binding to TRPV1 as a mechanism for insulin-mediated potentiation. Signaling cascades utilized by Insulin and IGF-I ------------------------------------------------ IR and IGFR produce their effects via an overlapping set of downstream enzymes \[[@B29]\], so we sought to identify signaling pathways involved in TRPV1 modulation. First, receptor tyrosine kinase (RTK) involvement was assessed by pretreating oocytes with membrane permeable inhibitors before and during insulin application (Fig. [3a](#F3){ref-type="fig"}). The nonspecific RTK blockers genistein (50 μM, 60 min) and lavendustin A (100 μM, 60 min) significantly reduced IGF-I (20 nM, 10 min) potentiation. Second, wortmannin (100 nM, 15 min), a specific phosphatidylinositol 3-kinase (PI(3)K) inhibitor, reduced potentiation as well. Thus, we reasoned that insulin and IGF-I utilize receptor tyrosine kinases to activate PI(3)K and prompt a signaling cascade that leads to TRPV1 current potentiation. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Singnaling cascades utilized by Insulin and IGF-I. a, Summary graph showing fold increase of current in the presence of IGF-I and the inhibitors: genistein (50 μM, n= 6, P\< .05), lavendustin A (100 μM, n= 7, P\< .05), wortmannin (100 nM, n= 5, P \< .05) and BIM (200 nM, n= 5, P\< .05). b, A PKC phosphorylation site TRPV1 mutant (S502A/S800A) (n= 6, P\< .01, upper trace) and cytochalasin D (1 μM, n= 4, P\< .01, lower trace) completely blocked insulin potentiation. ::: ![](1744-8069-1-17-3) ::: PI(3)K mediates some of its effects through various isoforms of protein kinase C (PKC) \[[@B29]\]. Therefore, we tested in oocytes expressing TRPV1 whether insulin and IGF-I could activate PKC, an enzyme known to potentiate TRPV1 through channel phosphorylation \[[@B8],[@B31]\]. Bisindoylmaleimide (BIM; 200 nM, 60 min), a nonspecific PKC inhibitor, significantly decreased IGF-I potentiation (Fig. [3a](#F3){ref-type="fig"}). The importance of PKC was further demonstrated using a mutant TRPV1 (S502A/S800A), which lacks residues acting as substrates for PKC phosphorylation (31) (Fig. [3b](#F3){ref-type="fig"}, upper trace). Two consequences of these mutations were apparent. First, insulin/IGF-I potentiation was abolished. Second, current amplitude was smaller in oocytes injected with mutant TRPV1 compared to wild type, suggesting that phosphorylation might have an intrinsic effect on basal channel function. Thus, it appears that RTK, PI(3)K and PKC activation are required for current potentiation by insulin. Insulin and IGF-I translocate TRPV1 to the plasma membrane ---------------------------------------------------------- In a number of systems, insulin and/or IGF-I can increase surface content of effector molecules \[[@B32]-[@B34]\]. We have used five different approaches to examine whether TRPV1 translocation occurs in response to insulin or IGF-I. First, we used cytochalasin D (1 μM, 60 min) to inhibit actin polymerization and decrease vesicular fusion to the plasma membrane in oocytes (Fig. [3b](#F3){ref-type="fig"}, lower traces). Cytochalasin D almost completely blocked TRPV1 current potentiation by insulin, implicating involvement of protein trafficking affecting translocation of both TRPV1 and PKC. Second, insulin and PDBu not only increased the potency but also the efficacy of capsaicin induced currents. In DRG neurons, at saturating concentrations of capsaicin (20 μM) (Fig. [4a](#F4){ref-type="fig"}~i~) the current amplitude increased (\>50%) following exposure (2--5 min) to insulin (1 μM) (Fig. [4a](#F4){ref-type="fig"}~ii~) or PDBu (1 μM) (Fig. [4a](#F4){ref-type="fig"}~iii~) suggesting recruitment of new channels into the plasma membrane or activation of previously silent channels (see also Fig. [1c](#F1){ref-type="fig"}). Third, Western blot analysis of cell-surface biotinylated TRPV1 expressed in HEK cells was carried out to test whether TRPV1 itself was being translocated to the plasma membrane (Fig. [4b](#F4){ref-type="fig"}). Analysis of band densities indicates that insulin (10 μM, 15 min) doubled surface TRPV1 expression levels relative to controls. Fourth, relative surface to cytosol optical intensities were quantified by immunofluorescence microscopy (Fig. [4c](#F4){ref-type="fig"}). These results, obtained with antibodies specific for TRPV1, show that IGF-I (20 nM, 15 min), insulin (10 μM, 15 min) and PDBu (10 μM, 15 min), all significantly increased surface TRPV1 expressed in HEK cells (Fig. [4c](#F4){ref-type="fig"}). Fifth, similar results were seen in HEK cells transiently transfected with green fluorescence protein (GFP)-tagged TRPV1. Exposure of IGF-I (50 nM) significantly increased the fluorescence intensity of the membrane within five minutes, indicating the accumulation of TRPV1 on the membrane (Fig. [4d](#F4){ref-type="fig"}). Together, these data suggest the involvement of receptor translocation in TRPV1 current potentiation and show that, in the presence of insulin, TRPV1 is mobilized to the plasma membrane from the cytosol. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Insulin and IGF-I translocate TRPV1 to the plasma membrane. a, Potentiation of saturating concentrations of capsaicin response by insulin and PDBu. Application of 5, 10 and 20 μM capsaicin shows saturation of current response (a~i~), exposure (2--5 min) of insulin (a~ii~) or PDBu (a~iii~) caused a 50% increase in current amplitude induced by 20 μM capsaicin (a~iv~) (insulin n = 4 P\< .01; PDBu n = 5 P\> .01). b, Representative Western blot of surface protein and total TRPV1 from control and insulin-treated (10 μM, 15 min) HEK293 cells expressing TRPV1 (probed with anti-TRPV1 antibody). Surface represents fraction of biotinylated TRPV1 and total represents total amount of TRPV1 in immunoprecipitate. Quantitative analysis of insulin\'s effect on surface expression, with mean densities of surface bands normalized to control values for samples run on the same gel (n= 4, P\< .01). c, Immunohistochemistry performed TRPV1 transfected HEK cells that were exposed to IGF-I (20 nM, 15 min), insulin (10 μM, 15 min) and PKC agonist, PDBu (10 μM, 15 min). Quantification of relative optical intensities (ROI, normalized as surface/cytosol for each cell): (control: n= 5; IGF-I: n = 7, P\< .05; insulin: n = 3 P\< .01; and PDBu: n= 7, P\< .01). d, Confocal image showing a significant increase (3.17 ± 0.52 fold, n= 9; P\< .01) in fluorescence intensity at the membrane 5 min after exposure to IGF-1 (50 nM) in HEK cells heterologously expressing GFP-TRPV1 fusion protein as compared to before IGF-1 application. ::: ![](1744-8069-1-17-4) ::: Discussion ========== From these results, we propose that insulin and its related growth factors provide both trophic and sensory support to peripheral nerve endings. Insulin and IGF-I can directly influence nociceptive ion channel function through phosphorylation and receptor translocation. In the presence of these modulators, TRPV1 is more responsive to painful stimuli (capsaicin, pH and heat) by increasing sensitivity and lowering thresholds. Based on this evidence, we conclude that insulin/IGF levels maintain TRPV1 function and their deficiency or resistance leads to deficits in inflammatory thermal sensation. The role of TRPV1 in animal models of diabetes has been suggested \[[@B27],[@B35]-[@B37]\]. The intriguing aspect is that insulin deficiency results in thermal hyperalgesia in streptozotocin induced diabetes. Sensitization of TRPV1 by phosphorylation could account for the hyperalgesic phenotype seen in this animal model of diabetes \[[@B35]\]. The exact mechanism of sensitization of TRPV1 is not clear, but could be due to over compensation by other tropic factors, such as NGF in response to insulin deficiency or elevated PKC activity in diabetes \[[@B38]\]. Using multiple methods, we elucidate molecular mechanism(s) by which insulin/IGF-I potentiate TRPV1 current (Fig. [3](#F3){ref-type="fig"} and [4](#F4){ref-type="fig"}). These neurotrophic factors, operating through RTKs, trigger a signaling cascade leading to PI(3)K and PKC activation \[[@B27],[@B28],[@B35]\]. PKC, an enzyme known to sensitize TRPV1 through phosphorylation \[[@B5],[@B8],[@B31]\], increases channel activity and receptor translocation to the cell surface \[[@B39]\]. Activation of a similar pathway to insulin by NGF via PI(3)K has been shown to robustly potentiate TRPV1 current \[[@B40]\]. Since insulin/IGF-I levels and PKC activity are altered in diabetes, we speculate that abnormalities in TRPV1 function may contribute to neuropathy in diabetes. In DRG sensory neurons, diminished growth factor levels are some of the earliest changes noticed with diabetic neuropathy \[[@B11],[@B20],[@B21]\]. One explanation for diabetic neuropathy states that it is, in essence, a microvascular disturbance \[[@B41]\]. Peripheral C fibers (which express TRPV1) release vasoactive substances like CGRP, causing small vessel dilation to increase cutaneous circulation and nerve terminal viability. Activation of TRPV1 has been shown to induce and enhance the release of CGRP (7). This neural component of microcirculatory control is decreased in diabetic neuropathy, and the consequent reduction in local blood flow contributes to peripheral vascular complications \[[@B41],[@B42]\]. With regards to growth factor deficiency, insulin has a vasodilatory effect that is dependent on CGRP release, which is compromised in diabetes \[[@B43],[@B44]\]. Our findings illustrates insulin/IGFs can cause vasodilation via their influence on TRPV1. A concept is emerging where signals emanating from IR and/or IGFR can activate kinases with the potential to control ion channel phosphorylation, subcellular localization and overall expression. Our work elucidates the mechanism insulin/IGF-I use for TRPV1 sensitization (i.e. RTK→PI(3)K→PKC), but the transduction pathway regulating expression has not been identified. Previously, it was suggested that insulin exerts influences on the same neurons that are responsive to NGF by sharing common pathways \[[@B13]\]. Recent reports demonstrate that NGF and glial cell line-derived neurotrophic factor (GDNF) upregulate TRPV1 expression on DRG neurons using transduction mechanisms common to insulin/IGF-I (i.e. MAPKs, PI(3)K, Ras, etc.) \[[@B11],[@B40],[@B45],[@B46]\]. Taken together, these studies set precedence for growth factor influence on nociceptor levels and implicate signaling cascades, which may be compromised by the absence of insulin/IGF. Insulin potentiates both whole cell and single channel currents mediated by NMDA receptors in Xenopusoocytes in a PKC-dependent manner \[[@B47]\]. Moreover, this effect was found to be due to membrane translocation involving both PI(3)K and PKC \[[@B33],[@B34],[@B38]\]. In DRG neurons insulin increased capsaicin induced cobalt uptake \[[@B48]\]. Along these same lines, insulin can potentiate currents in HEK293 cells through recruitment of GABA~A~receptors to postsynaptic domains \[[@B32]\]. In cultured cerebellar granular cells, IGF-I potentiates kinate receptors through a PI(3)K dependent mechanism \[[@B49]\]. Furthermore, vanilloid receptors (VRs) are present throughout the body, widely believed to have functions other than temperature sensation. TRPV1 expressed in the central terminals of the sensory neurons robustly modify synaptic transmission \[[@B50],[@B51]\]. TRPV1 could be detected using RT-PCR technique throughout the neuroaxis \[[@B52]\] and identification of specific ligands such as NADA in certain brain regions confirms a role in the CNS \[[@B53]\]. The nature of the receptors involved in this response and their role in the CNS are not clearly understood, but suggestive of a direct or indirect role in modifying neurotransmitter release \[[@B54]\]. TRPV1 is also located in vasculature, bronchi and urinary bladder \[[@B7],[@B55],[@B56]\]. Modulation of these receptors by lack of insulin and IGF-I may contribute to CNS disturbances, cardiovascular, respiratory and urinary complications resulting from diabetes. Lastly, these findings emphasize the importance of maintaining proper insulin levels and suggest a potential benefit of IGF-I administration in the treatment and prevention of diabetic complications. We propose that insulin and IGF-I therapy, partially working through TRPV1, can improve complications associated with diabetes mellitus. Acknowledgements ================ We thank R. Khardori for critical reading of the manuscript. cDNA clones were gifts from D. Julius, M. Tominaga and N. Kadei. This work was supported with a grant from NIH (DK065742, NSO42296 to L.S.P) and SIU Excellence in Academic Medicine (M.E.P. and L.S.P).	PubMed Central	2024-06-05T03:55:57.400482	2005-4-27	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1142339/", "journal": "Mol Pain. 2005 Apr 27; 1:17", "authors": [ { "first": "Jeremy J", "last": "Van Buren" }, { "first": "Satyanarayan", "last": "Bhat" }, { "first": "Rebecca", "last": "Rotello" }, { "first": "Mary E", "last": "Pauza" }, { "first": "Louis S", "last": "Premkumar" } ] }
PMC1142340	Background ========== Maternal mortality is the vital indicator with the greatest disparity between developed and developing countries \[[@B1],[@B2]\]. Causes of maternal deaths are similar in these countries, however the distribution of causes differ somewhat from region to region \[[@B3]\]. Measuring maternal mortality is notoriously difficult for both conceptual and practical reasons. The currently available approaches are complex, resource intensive and imprecise; and the results they yield are often misleading \[[@B4]\]. The challenging nature of measuring maternal mortality necessitates to perform an action-oriented means of gathering information on where, how and why deaths are occurring; what kinds of action are needed and have been taken \[[@B5]\]. Assessing the impact of preventive measures demands exact knowledge about how many lives were saved. Often the answer to the \"why\" is not a simple one. The death may occur as a result of a series of interconnected events rather than one single factor. Answering the \"why\" question thus requires a systematic review of each maternal death in order to find information on events surrounding the deaths. A maternal death review is an in-depth investigation of the causes and circumstances surrounding maternal deaths such as problems in accessing care, mismanagement and inadequate routines \[[@B6]\]. Levels of maternal mortality in the Gambia are unacceptably high and are ranked among the highest in Africa, estimated at 1,050 per 100,000 live births and are higher in rural than in urban areas \[[@B7]\]. Obstetric causes of maternal deaths in the Gambia are well documented \[[@B7]-[@B11]\], but little attention is paid to the contributing factors. In order to reach the Millennium Development Goal of reducing maternal mortality, women\'s access to good quality health care embedded in a human rights framework is an important factor. Access to emergency obstetric care and better social status of women are two elements that may contribute significantly towards this goal. Site-specific information may be a key to policy change and action, even if the general causes of continued high maternal mortality rates are well known. The influence of a local maternal care access study on improving service delivery and organization has been demonstrated in Mali \[[@B12]\]. The objectives of the present study were to describe the socio-cultural and health service factors associated with maternal deaths in rural Gambia. We aimed at identifying factors that if avoided, may have prevented these deaths. A key focus of the study is to try to create an operative understanding of the concept of access. The study was carried out in Central and Upper River Divisions between January and September 2002. The main qualitative findings from verbal autopsies performed on 32 maternal deaths are presented. The analysis of socio-demographic background, medical causes and management of cases is presented in another paper \[[@B13]\]. Study Area ---------- Central and Upper River Divisions are located at least 300 km away from the capital city, Banjul, and the roads are generally in poor condition. In 2002 the total population of the two divisions was estimated at 400,917 \[[@B14]\]. The inhabitants are mainly subsistence farmers, and they are generally poor. There are seventeen medical facilities in the area including one referral hospital. This hospital is the only facility providing comprehensive emergency obstetric care (EOC) \[[@B15]\]. The other health facilities provide some basic EOC. General access to basic health care facilities in the country is good, with over 85% of the population living within 3 km of a primary health care or outreach post and 97% within 5 km \[[@B16]\]. The distance from other health facilities in the area to the referral hospital ranges from 17 to 115 km. Antenatal care coverage in the country is exceptionally high, as 96% of pregnant women attend antenatal clinic at least once during pregnancy \[[@B17]\]. Levels of maternal mortality in this region are nevertheless among the highest in the country \[[@B7]\]. Model ----- We used the standard WHO definition of maternal death \[[@B18]\]. Like most health problems, causes of maternal death can be viewed either narrowly or broadly. A broad view would take into account individual, community and health service factors that contributed to the deaths, not merely the medical cause. The \"Three phases of Delay Model\" \[[@B19]\] was chosen to classify factors associated with the maternal deaths in the present study. The model is based on the fact that about 75% of maternal deaths are a result of direct causes: hemorrhage, obstructed labor, sepsis, eclampsia and abortion complications \[[@B4]\]. Most of these deaths are preventable with prompt and adequate medical interventions. Delays in reaching adequate care are prominent factors contributing to maternal deaths. Thaddeus and Maine \[[@B19]\] have argued that not getting adequate care in time is the overwhelming reason why women die in developing countries. Lack of care, they argued, can be related to three factors: a delay in making the decision to seek care when complications develop; a delay in reaching obstetric medical facility once the decision to seek care has been made; or delay in receiving adequate and appropriate care once a medical facility has been reached. Delay in the decision to seek medical care may be influenced by various factors such as the actors involved in the decision-making process, illness characteristics, and experience with the health system or distance to the health facility. Delay in reaching an appropriate medical facility is affected by the distribution of health facilities, availability of transportation, road conditions or cost of transportation. Delay in receiving adequate and appropriate care once the facility is reached is mainly due to operational difficulties in the health care delivery system. Such inadequacies may be characterized by shortages in supplies, equipment, lack of trained personnel, incompetence of the available staff, or uncoordinated emergency services. The delay model helps to identify community and health services factors contributing to maternal deaths and as such it is useful in devising interventions and strategies for preventive measures. Methods ======= Data Collection --------------- This study is part of an in depth study on maternal mortality in the above described area. The main bulk of maternal mortality studies in the Gambia cover other areas \[[@B7]-[@B11]\]. We have, in the chosen time period, identified a total of 42 maternal deaths, by using different community and health facility case finding strategies. The cases identified were women who actually tried to reach or have reached health care services. Medical and social details about the fatal cases are submitted for publication elsewhere. A verbal autopsy technique \[[@B20]\] was possible to perform on 32 of the women who died. We aimed at identifying the various circumstances that contributed to the deaths, often called the \"road to death\" \[[@B21],[@B22]\]. This involved a visit to the home and village of the deceased and all health facilities where she sought care. Key people who had witnessed any stage in this process were interviewed. In the community, interviews were performed with family members or other persons, usually those present during the time the deceased developed the illness; or those who accompanied her to the health facilities and were with her at the time of death. These were mainly in-laws, midwives or the deceased\'s husband. A WHO based, verbal autopsy questionnaire \[[@B10]\] was administered. An interview with the deceased\'s relatives was performed in the deceased\'s home at least one week after, but not later than twenty-five weeks after the death. We often used a group interview, as the whole family participated in revealing their versions of the context of the death, and to give a verbatim account on the deceased\'s final illness up to her death. Specific issues such as the time taken to decide to seek care; places where care was sought; financial constraints; cultural factors influencing care seeking process; means of transportation and time to reach a medical facility were explored and recorded. We also interviewed health care staff who participated in the provision of care to the deceased, ambulance drivers, generator operators and laboratory personnel. Other people interviewed in the community were local taxi drivers and ferry captains. Analysis -------- Interviews with staff were carried out independently. All interviews were tape recorded and transcribed in full text. The transcribed material was categorized and analyzed by using a grounded theory. Grounded theory is an interview text analysis method utilized for qualitative studies in the health sector, where the basis for the analysis is not a theoretical model per se, but the focus is rather on identified items of relevance to the studied topic \[[@B23]\]. Interpretation of the findings with a view to provide possible and plausible explanations was then performed. Ethical approval ---------------- The study was approved by a joint ethics committee of The Gambia Government/Medical Research Council Laboratories and the regional Medical Ethics Committee of Norway. Oral consent was taken from each participant. Results ======= Delay in deciding to seek care ------------------------------ The delay was estimated from the moment somebody, either the woman or her family realized that there was a complication until the decision to seeking care was made. In seven of the 32 cases the process of seeking medical attention was delayed after becoming aware of the complication. The delay ranged from two hours to five days. The reasons mentioned for the delay were underestimation of the severity of the complication, cultural belief or previous unfavorable experience with the health system. Underestimation of the severity of the complication --------------------------------------------------- Previous uncomplicated pregnancies may influence actions taken during the current pregnancy that can lead to a delay in the decision-making process. Women or their relatives used previous pregnancies as a risk- predicting tool. A husband of a deceased woman narrated: \"This was her ninth pregnancy. All previous pregnancies were delivered at home. She always gave birth without even calling for help from the traditional birth attendant. We thought she will deliver this time without a problem\". In another case a mother in-law narrated: \"She swept the house and prepared breakfast. At midday she was lying in the room complaining of labor pains. We thought she will deliver without a problem as in her last pregnancy. It was after 2:00 pm she did not deliver then we decided to look for transport to take her to the health centre\". Cultural belief --------------- In Gambian society pregnancy and childbirth are generally regarded entirely as women\'s entity. Older women in their menopause are seen as experts on pregnancy and childbirth, particularly in rural areas of the country. These women are consulted if a complication is noticed during pregnancy, labor or during the puerperium. When consulted, they usually decide what should be done and their advice is taken. Words of elders are hardly challenged in Gambian societies. In the case below an older woman advised a woman in labor to wait until after the next Muslim praying time (after three hours) before seeking medical attention. The rationale for the delay was that it is believed to evaluate progress in labor at specific times corresponding to the Muslim praying times: \"Labor and child birth takes place at certain times\... \... \....these times correspond with the Muslim praying times. It was around midday and the next praying time was 2:00 pm so we thought she will deliver by then. It was after then when she did not deliver we decided to take her to the health centre\". Experience with the health care system -------------------------------------- Other testimonies indicated that structural factors in maternal health care provision discouraged women from seeking care. Prenatal care is provided on specific days in each community during week days only. This gives the impression that maternal health services are only available on days when clinics are held. A mother in-law narrated: \"On Thursday evening she complained of abdominal pains\... \... \...throughout the weekend she was with severe abdominal pain but we had to wait until the following Monday as it is the day on which pregnant women are attended to. The clinic is closed on Saturdays and Sundays\". Poor provider attitude, fear of punishment by health care providers based on previous experiences or just gossip can lead to delays in the decision making process. A midwife narrated \"She was vomiting throughout the night, the following morning the husband decided to take her to the health centre but she refused\... \... \...she has not yet get an antenatal care card. She feared the nurses because if she goes to complain about the vomiting she will be asked the card and without it they \[nurses\] will tell her all salty words. She may be insulted or may even not be given medicine\". Information from care providers that is not clearly understood, can lead to delay in seeking medical care. A woman with twin pregnancy was advised to deliver at the hospital. However, the information provided lead to the following situation: \"She was told she had twin pregnancies by the nurse. She was told by the nurse to report to the hospital when in labor. When labor began we decided to go to the hospital. She was not told to report to the hospital before labor began\". Barriers to seeking care may not appear as such to care providers when they make recommendations. The woman may not see them as an issue, as she is not given room to discuss these with the nurse. Lack of money and refusal to receive medical attention were not identified as factors affecting health care seeking process. In twenty-two out of the thirty-two women, no funds were available when the complication developed. In all these cases the woman was taken to a medical facility without money and a relative was left behind to raise money in the community, to be able to pay at a later stage. Delay in reaching an appropriate medical facility ------------------------------------------------- Once a decision to seek medical care has been made, other obstacles had to be overcome in reaching a medical facility. Twenty-seven of the 32 women were delayed in reaching an appropriate medical facility. The reasons for this delay can be grouped into three subcategories: lack of transport, prolonged transportation and seeking care at more than one facility. Lack of transportation ---------------------- Transportation difficulties, such as poor road conditions, lack of readily available transport and/or inadequate means of transportation were mentioned. The relatives often expressed shortage of transport as serious obstacles. Lack of motorized transport forced some families to opt for alternative means of transport such as using a cart (donkey, ox or horse) or in extreme instances, they walked. A husband explained: \"She started pouring blood late in the evening just after evening prayers \[5:00 pm\]\....we took her to the main road \[tarred road\] to look for transport. We were there \[main road\] up to twelve midnight but couldn\'t get transport. All the vehicles that came were full. We went back home and woke up early morning the following day to catch the first transports\". Transportation difficulties were experienced even after reaching the first medical facility, as some of the health facilities were without an ambulance. If a facility has an ambulance it usually serves multiple purposes and may not be available at certain times. A midwife narrated: \"The patient came to the health centre at around 4:00 pm\... \....she cannot be managed here because she may need an operation \[caesarean section\]. We planned to evacuate her to the hospital but our ambulance had a breakdown a week ago. We looked for transport in the village throughout the night but could not get one. The following morning we went to the agricultural department to look for transport but their vehicle had already left for trek. It returned around 11:00 am and thereafter it came to transport the patient to the hospital\". Lack of fuel for the ambulance was also mentioned. In such occasions relatives or escorts met the fuel cost. A husband narrated: \"I took my wife to the health centre\... \...two hours later the nurse told me that she \[my wife\] will be transferred to the hospital but that the ambulance had no fuel. I was asked to buy fuel for the ambulance to take my wife. I bought twenty liters of diesel\". Some communities in the Gambia have -- with the assistance of the health authorities -- tried to set up community based emergency transport systems, such as horse carts or bicycle ambulances, but it is difficult to make them sustainable. Prolonged transportation ------------------------ Long distance, visiting different health facilities, poor road and vehicle conditions contributed to prolonged traveling time. Several testimonies highlighted this. A husband explained: \"She was admitted in the hospital for two weeks and discharged on a Monday. On her return to our village \[85 km away from the hospital\] she fell down unconscious. We took her to the health centre in our village where she was transferred to another health centre \[20 km\]. She was again transferred to the hospital \[60 km away\]. She spent few hours at the hospital and died\". Seeking care at more than one medical facility ---------------------------------------------- Seeking care at an inappropriate level of facility actually delays access to appropriate treatment. The inability to provide comprehensive obstetrical services forces peripheral health facilities to refer all women needing such services to the nearest hospital: 26 of the 32 women visited more than one medical facility during the care seeking process, 18 of the 26 women visited two health facilities while the other 8 women visited three different facilities. Thus, the women accessed a health care facility, but not appropriate health care. The husband of a deceased narrated: \"We took her to the health centre in the village\... \.....she was examined by the nurse who later transferred her to another health centre \[44 km away\]. There she spent the night and the following morning she was again transferred to the hospital \[36 km away\]. On our way to the hospital we had to cross the river at two different crossing points. Immediately after we reached the hospital she died\". Delay in receiving prompt and appropriate care after reaching the hospital -------------------------------------------------------------------------- Thirty-one women experienced delay in receiving prompt and adequate obstetric care at the hospital level. Lack of blood transfusions and basic medical supplies were mentioned in the testimonies. A mother in-law explained: \"When we reached the hospital, they \[the doctor and the nurses\] told us to find two bottles of blood for her \[our patient\]. We went to the laboratory but the man at the lab said there was no blood. I donated one bottle and bought another in the lab. After giving her \[patient\] the blood we were asked to get another two bottles. We went back to the lab but the man at the lab insisted there was no blood. I paid him D300.00 \[equivalent to US\$12.00\] before getting the two bottles of blood\". A husband of a deceased woman narrated: \"She was pouring blood at home so we took her to the health centre. There we were told she urgently needed blood but blood bags were not available. She was then transferred to the hospital \[60 km away\]. At the hospital blood bags were finished. She was in the hospital from mid-day up to the following day in the evening but had not received blood. Late at night she died\". A laboratory officer narrated: \"Here patients are escorted to the hospital by old women who are not fit to donate blood. In addition most men in this area are reluctant to donate blood and prefer to buy blood\". Delay in providing prompt and adequate care by the medical team was also highlighted in the testimonies. A midwife narrated: \"She was brought to the hospital on the 13^th^at around 9:00 am from another health centre. The doctor saw her and diagnosed hand-presentation. He \[doctor\] asked us \[midwives\] to observe her. No action was taken by the doctors up to the 15^th^late in the evening \[48 hours later\] when they took her to the theatre. He \[doctor\] first tried external cephalic version which failed before a caesarean section was performed. The patient was wheeled dead from the theatre\". Poor management of staff availability, particularly doctors, has been mentioned as a factor contributing to poor care. A midwife narrated: \"There used to be four doctors in the maternity unit but in July all three went on leave together. Now only one doctor is available for the unit. He does ward rounds, performs operations and runs the out-patients clinic. Even when there were four doctors we usually have problems with them \[doctors\] because there is no duty rooster for doctors in place. After normal working hours when there is an emergency it is always difficult to see them\". Discussion ========== The Gambia was the first country to implement a sisterhood approach to measure maternal mortality rates \[[@B9]\]. Former studies of maternal deaths in the Gambia indicated a decrease in numbers \[[@B10]\]. However, factors related to health care delivery could contribute to further improvement \[[@B16]\], as substandard care has been demonstrated as a contributing factor to poor survival \[[@B8]\]. We used multiple sources of information, such as health workers\' identification, community leaders\' knowledge, hospital files and post partum follow up visits to identify maternal deaths that took place in the health care facilities. Maternal death is often a consequence of a long and complex chain of delays, and only in few cases death can be attributed to a specific event \[[@B24]-[@B26]\]. Any one delay could be fatal to a woman with obstetrical complications. Contrary to the common belief, that women do not seek care and die in the community, we identified a number of women who initially intended to deliver at home, but tried to get assistance once a complication occurred. The problems encountered in trying to do so, reveal major obstacles in access to appropriate care within an acceptable time. Delay in deciding to seek medical care -------------------------------------- Delay in deciding to seek medical care on the part of the woman or her relative is usually regarded entirely as patient factor. First, the illness or complication must be recognized and classified as abnormal. Recognition of an illness may be influenced by factors such as the prevalence of the condition \[[@B27]\]. In a study among pregnant women in Senegal, 13% regarded fever, pallor and dizziness as normal signs of pregnancy because these conditions were common among pregnant women in that area \[[@B28]\]. In Tanzania, rural women seem to avoid going to the hospital because of fear of discrimination, geographical and financial barriers and different interpretation of danger signs \[[@B29]\]. Raising awareness is a health education issue for health care workers and the community. One role of appropriate antenatal care is to address these issues and to offer care seeking solutions in advance. Access to skilled attendance at childbirth includes improved technical skills as well as skills in attitude, communication, information and early advice on referral \[[@B30]\]. Brown \[[@B31]\] defined culture as a \'complex whole\' that refers to the learnt pattern of thoughts and behavior characteristics of a social group. It involves religion, kinship, knowledge, belief, art, morals and child bearing practices. The tendency to act or not in the presence of a complication is also influenced by the interpretation supported by cultural beliefs. Several studies carried out in Africa and elsewhere \[[@B24],[@B32],[@B33]\] have highlighted how culture influenced health care seeking process. Religious belief was mentioned to have influenced the care seeking process in our study. Jansen \[[@B34]\] asserted that religion, medicine and magic are closely interwoven. If the barriers to care are too overwhelming, a culturally based reassurance that \"things most likely will go well\" may cause a hesitation in recognizing early signs of complications. Health service related factors were mentioned to have constrained the decision-making process in this study. Bad experience with the health system will mostly lead to reluctance or non-utilization of health care services. Poor provider attitude towards patients has been identified as a major factor to low utilization of services in Kigoma \[[@B21]\] and to low compliance to a referral hospital by high-risk pregnant women \[[@B28]\]. The communication barriers between lay people\'s concepts and those of professional care providers may lead to serious misinterpretations. Women in the Gambia often resort to home delivery assisted by a traditional birth attendant or a relative as their first option. Sundari \[[@B25]\] identified unfamiliar setting at the health facility, being attended to by strangers, lack of family support, attendant being a male care provider, reduced autonomy, lack of sympathy and understanding on the part of the health care personnel and not seeing the need for care as some of the factors contributing to non-utilization of health services during labor and childbirth. Delay in reaching an appropriate medical facility ------------------------------------------------- Lack of public transportation systems in rural areas requires that communities need to form partnership with the local commercial transport owners in addressing the transport problem. This strategy was adopted in North-western Nigeria \[[@B35]\] and had contributed to the reduction of maternal deaths and cost of transportation. Major health centers are strategically located in the Gambia, but accessing them does not necessarily mean to receive appropriate care. Sometimes using these sites as the entry point to health services can delay further attempts of accessing adequate care. Efforts to transfer health centers into fully functional basic obstetric emergency units could reduce the delay caused by long transportation time. Being unable to meet the costs for immediate health care was not seen as a main obstacle. Some health facilities supported by non-governmental organizations (NGOs) and local associations have implemented a system of cost sharing in order to provide quality health care. Delay in receiving prompt and appropriate care after reaching the hospital -------------------------------------------------------------------------- A multi-centre study from three West-African countries, reported that most of the women classified as \"near misses\" were referred from another facility \[[@B36]\], highlighting the need to differentiate between those who arrive in a critical condition and those who develop one. Inadequacy in health care may be due to one or a chain of the following events: shortage of medical supplies, lack of equipment, lack of trained personnel, and incompetence of the available staff. Health system failures have been identified as a major contributing factor to maternal deaths \[[@B8],[@B10],[@B21],[@B37],[@B38]\]. Conclusion ========== The failure to get adequate treatment in time may be seen in a \"right to access health care \"context. Women\'s access to appropriate services is a concern in the Gambia. This study reveals that women do try to reach adequate health services when an emergency occurs, but that there are many obstacles that delay this process. Improving accessibility and quality of EOC services in the area is necessary if maternal deaths are to be prevented Competing interests =================== There are no competing interests for this study. The study was partly financed by a Norwegian government quota grant for students from developing countries for higher studies in Norway. This project was also funded by the Participatory Health Population and Nutrition Project (PHPNP) of the Department of State for Health of The Gambia. Authors\' contributions ======================= Mamady Cham did the data collection in the field, the main analysis, wrote the first draft of the paper and reviewed the final document. Johanne Sundby planned the current study and introduced the methodology, and participated in the writing of the first draft, and wrote the final version of the document. Siri Vangen assisted in the analysis of data as the second reviewer of the text, contributed to the first writing of the paper and reviewed the final document. Acknowledgements ================ We would like to express our profound gratitude to the relatives and family members of the deceased for their willingness and courage in narrating what had happened. We also express our thanks and appreciation to all the health staffs in Central and Upper River Divisions. Our special thanks to the research assistants and driver.	PubMed Central	2024-06-05T03:55:57.402789	2005-5-4	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1142340/", "journal": "Reprod Health. 2005 May 4; 2:3", "authors": [ { "first": "Mamady", "last": "Cham" }, { "first": "Johanne", "last": "Sundby" }, { "first": "Siri", "last": "Vangen" } ] }
PMC1142341	Findings ======== Three types of Primate T-cell lymphotropic viruses (PTLVs) have been discovered so far in primates \[[@B1]\]. While two of them i.e. PTLV-1 and PTLV-2 include human (HTLV-1, HTLV-2) and simian (STLV-1, STLV-2) viruses, the third type (STLV-3) consists only, so far, of simian strains. Sequence comparisons of STLV-3 proviruses indicated that these strains are highly divergent from HTLV-1 (60% nucleotide similarity), HTLV-2 (62%), or STLV-2 (62%) prototype sequences. In all phylogenetic analyses, STLV-3 viruses cluster in a highly supported group, indicating an evolutionary lineage independent from PTLV-1 and PTLV-2. Nevertheless, STLV-3 lineage is composed of at least three subtypes that are corresponding more or less to the geographical origin of the virus (East, West or Central Africa) \[[@B2]-[@B9]\]. Most of the viruses belonging to the PTLV-1 type cannot be separated into distinct phylogenetic lineages according to their species of origin. Their intermixing has therefore been inferred as an evidence for past or recent interspecies transmission episodes. The hypothesis of viral transmission from monkeys to humans is supported by an increasing number of observations \[[@B1]\]. Thus, it has been proposed that HTLV strains related to STLV-3 might infect human populations living in areas where STLV-3 is present. Cameroon has a remarkable diversity of retroviruses. All the subtypes of HIV-1 group M (A to H) are present, subtype-recombinant strains co-circulate, and HIV-1 groups O and N have been reported. Besides, HTLV-1 subtypes B and D as well as HTLV-2 type A and B are also present in Cameroonian individuals, while STLV-1 and STLV-3 strains have been isolated from several non-human primates (NHPs) species living in this region \[[@B3],[@B4],[@B8]\]. We therefore conducted a study to search for HTLV variants in Cameroonian individuals with HTLV-1/2 indeterminate serology. This survey was approved by both the national (Cameroon Ministry of Health and their National Ethics committee) and local authorities (village chief) with information to each participant. An oral informed consent was obtained from each participant (adults or parents for minors). A series of 240 blood samples was obtained from Bakola (n = 64) and Baka (n = 65) Pygmies, while others (n = 111) were obtained from Bantous (mainly from the Fang, Mvae and Ngumba tribes). All these individuals (117 women and 123 men, mean age 44, range 10--75 years) live in remote villages in the rain forest area of the Southern part of Cameroon. The 240 plasma were tested at a 1/40 dilution for the presence of HTLV-1/2 antibodies with a highly sensitive immunofluorescence test (IF), that uses MT2 and C19 as HTLV-1 and HTLV-2 viral antigen producing cells respectively. This test also allows the detection of STLV-3 positive samples \[[@B4],[@B5]\]. The 48 plasma that were IF reactive on MT2, C19 or both, were further tested by western blot (WB HTLV BLOT 2.4; Genelabs Diagnostics, Singapore). Among the 48 samples tested, 4 and 11 WB patterns were very evocative of HTLV-1 and HTLV-2 infection respectively, while 27 exhibited diverse HTLV incomplete patterns, including some HTLV-1 indeterminate gag profile (HGIP). Six samples were WB negative. High-molecular weight DNA was extracted from the 48 blood samples and was first subjected to PCR using human β-globin specific primers, to ensure that DNA was amplifiable. They were then subjected to two series of PCR using degenerated taxand polprimers designed on highly conserved regions that are common to all PTLVs. The taxprimers are the following: SCTaxoutse: 5\'-CTHTAYGGRTACCCHGTCTACGT-3\' and SCTaxoutas: 5\'-AGGGGAGBCGAGGGATAAGG-3\' corresponding to nucleotides 7279 to 7301 and 7455 to 7474 respectively of the prototype STLV-3~PHA969~sequence (GenBank accession number Y07616). The polprimers are SCPOL1outse: 5\'-TTAAACCDGARCGCCTCCAGGC-3\' (nt 2485 to 2506) SCPOL1outas: 5\'-GGDGTDCCYTTRGAGACCCA-3\' (nt 3201 to 3220) and SCPOL1inse: 5\'-TAYHHAGGRCCAGGMAATAACCC-3\' (nt 2556 to 2578). HTLV-1 and HTLV-2 taxsequences were obtained for 4 and 11 samples which exhibited complete HTLV-1 and HTLV-2 WB profiles respectively, but none of the WB indeterminate sample gave a PCR signal. Consistent results were obtained for these HTLV-1 and HTLV-2 strains with the polsemi-nested PCR. However, a faint band (665 bp) was also obtained for one sample (Pyl43), which was previously found to be taxPCR negative. Sequencing of this fragment indicated the presence of an HTLV polsequence that is highly related to STLV-3 strains (86.6% to 99.2% nucleotide identity). Based on an alignment of different STLV-3 sequences, a taxsemi-nested PCR was then designed using SCTaxoutse (see above) and Mac4 followed by Mac2 and Mac4 as inner primers \[[@B10]\]. This allowed the amplification of a 279 bp fragment which was also found to be highly homologous to STLV-3 strains (92.4% to 99.6% nucleotide identity). We did two phylogenetic analyses (taxand pol) with the neighbor joining method. Assessment of a 180-bp taxsequence (Figure [1](#F1){ref-type="fig"}) or of a 665-bp polregion (data not shown) demonstrated a strong relationship between Pyl43 and STLV-3 strains from Cameroon. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### HTLV-3 is closely related to STLV-3.Unrooted phylogenetic tree generated with the Neighbor-joining method, performed in the PAUP program (v4.0b10), on a 180 bp fragment of the taxgene using all full length PTLV-1/2 available sequences and all published STLV-3 taxsequences. The PTLV-1/2/3 strains, including (in bold), the novel sequence generated in this work (Pyl43), were aligned with the DAMBE program (version 4.2.13). The final alignment was submitted to the Modeltestprogram (version 3.6) to select, according to the Akaike Information Criterion (AIC), the best model to apply to phylogenetic analyses. The selected model was the TrN+G. Bootstrap support (1,000 replicates) is noted on the branches of the tree. The branch lengths are drawn to scale, with the bar indicating 0.1 nucleotide replacement per site. ::: ![](1742-4690-2-30-1) ::: The HTLV-3 sample originated from a 62 years old Bakola Pygmy living in a remote settlement in the ocean department of Southern Cameroon. His plasma was reactive on MT2 cells (titer: 1/320) but was negative on C19 cells. The HTLV BLOT 2.4 WB \[[@B11]\] exhibited a strong reactivity to p19 and a faint one to MTA-1 (Figure [2A](#F2){ref-type="fig"}). On the INNO-LIA strip (Innogenetics, Ghent, Belgium) \[[@B12]\], it reacted faintly (+/-) with the generic p19 (I/II), but strongly to the generic envgp46 (I/II) and to the specific HTLV-2 gp46 (Figure [2B](#F2){ref-type="fig"}). Surprisingly, the close molecular relationship between Pyl43 and STLV-3 is thus not paralleled by the serological results, as most of the STLV-3 infected monkeys have an \"HTLV-2 like\" WB pattern (p24 \> to p19 with or without K55) (Figure [2A](#F2){ref-type="fig"}, lanes 3--4) \[[@B2]-[@B9]\]. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Serological pattern of the person infected by the HTLV-3 Pyl43 strain. (A) Western Blot from Genelabs Diagnostics (HTLV BLOT 2.4 version) and (B) a line immunoassay (INNO-LIA HTLV confirmation Immunogenetics). The HTLV 2.4 western blot kit is based on strips incorporating HTLV-1/2 native viral antigens (originating from HTLV-1 infected cells) to which HTLV-1 (MTA-1) or HTLV-2 (K55) gp46s or HTLV-1 and HTLV-2 (GD21) gp21 recombinant proteins have been added \[11\]. The INNO LIA kit uses only recombinant antigens and synthetic peptides derived from both HTLV-1 and HTLV-2 proteins sequences. Whereas gag p19 I/II corresponds both to a recombinant protein and synthetic peptides being recognized by anti HTLV-1 and HTLV-2 immune sera, env gp46 I/II corresponds only to synthetic peptides recognized by anti HTLV-1 and HTLV-2 immune sera. env gp46 II corresponds to synthetic peptides specific of HTLV-2 \[12\]. (A, B) Lane 1: HTLV-1 positive control; lane 2: HTLV-2 positive control; lane 3: STLV-3 positive control (STLV-3~604~strain); lane 4: STLV-3 positive control (STLV-3~F3~); lane 5: HTLV-1/2 negative control; lane 6: plasma from the person infected by HTLV-3 (Pyl43 strain). ::: ![](1742-4690-2-30-2) ::: In conclusion, we have demonstrated in this report the presence of a new human retrovirus in the peripheral blood cells of a Central African native. This virus is closely related to STLV-3. In the context of multiple interspecies transmissions that have occurred in the past and led to the present-day distribution of the PTLV-1 \[[@B1]\], we suggest that HTLV-3 might be widespread, throughout the African continent. HTLV-3 infection seems to be reflected by an HTLV indeterminate serological WB pattern. This raises an important public health question regarding the effectiveness of the current commercially available screening and confirmation tests for detecting this new HTLV type. Key research priorities are now to investigate the transmission modes of this virus as well as possible pathogenic associations. List of abbreviations used ========================== PTLV: Primate T Lymphotropic Viruses HTLV: Human T Cell Lymphotropic Virus PCR: Polymerase Chain Reaction WB: western-blot NHPs: Non Human Primates HGIP: HTLV Gag Indeterminate Profile Nucleotide accession number =========================== The taxand polaccession number for the sequences determined in this study are: \[GenBank:[DQ020492](DQ020492), GenBank:[DQ020493](DQ020493)\] respectively. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= SC and SAC performed the laboratory work. SB did the serological assay and RD the phylogenetic analyses. AF and AG organized and performed the field studies, AG and RM designed, implemented and coordinated the study, wrote the manuscript. All authors have read and approved the manuscript. Note ==== Wolfe et al. recently reported in an abstract the presence of two novel HTLV viruses \[[@B13]\]. Whether or not these viruses are related to the new strain described here (HTLV-3 Pyl43) remains to be determined by further comparative studies. Acknowledgements ================ This work was supported by a grant from l\'Association de Recherche sur le Cancer (ARC \# 4781) to RM and fellowships from le Ministère de la Recherche to SAC and Virus Cancer Prévention association to SC. RM is supported by INSERM. SC and SAC contributed equally to the laboratory work. RM and AG share senior authorship on this work. We also thank Dr Timothy Stinear for his critical comments.	PubMed Central	2024-06-05T03:55:57.405997	2005-5-9	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1142341/", "journal": "Retrovirology. 2005 May 9; 2:30", "authors": [ { "first": "Sara", "last": "Calattini" }, { "first": "Sébastien Alain", "last": "Chevalier" }, { "first": "Renan", "last": "Duprez" }, { "first": "Sylviane", "last": "Bassot" }, { "first": "Alain", "last": "Froment" }, { "first": "Renaud", "last": "Mahieux" }, { "first": "Antoine", "last": "Gessain" } ] }
PMC1142342	Background ========== Systems or computational biology -- the use of mathematical analysis to examine complex biological systems -- is becoming increasingly important \[[@B1],[@B2]\]. Biological signals are complex, with fractal or even multi-fractal characteristics, and health is associated with fractal timing sequences \[[@B3]\]. For example, normal sinus rhythm is multi-fractal and the onset of congestive failure significantly attenuates this complex signal\[[@B4]\]. Respiratory sinus arrhythmia (RSA) -- the increase in heart rate with inspiration and decrease with expiration -- is one component of this complexity. It represents a dynamic interaction between respiratory, brain stem and cardiac oscillators that is physiologically advantageous. Hayano et al.\[[@B5]\]showed in dogs that positive RSA is associated with lower shunt fraction and lower dead space ventilation. Negative RSA -- a decrease in heart rate with inspiration and increase with expiration -- increases shunt fraction and dead space ventilation. Thus positive RSA improves ventilation/perfusion matching. Godin and Buchman\[[@B6]\] have proposed that the loss of RSA that occurs in illness is a consequence of uncoupling of important biological oscillators. Interventions that restore or enhance coupling would therefore be advantageous. We have developed a so-called biologically variable mechanical ventilator, which uses a normal fractal-breathing pattern (a form of 1/f noise) to generate physiological rates and volumes. Long recordings of normal respiration are processed to drive the ventilator using computer hardware and software. In numerous models, fractal ventilation (FV) has been shown to improve gas exchange (increases arterial oxygen tension) and respiratory mechanics (more compliant lung and lower peak inspiratory pressure) with decreases in shunt fraction and dead space ventilation compared to conventional ventilation (CV) which delivers monotonously regular respiratory rates and tidal volumes \[[@B7]-[@B10]\]. Mechanisms for such improvement include increased surfactant phospholipid\[[@B11]\] and stochastic resonance -- the addition of noise to an input signal to enhance an output in a nonlinear system\[[@B12]\]. We hypothesized that the imposition of a fractal respiratory signal with its added physiological noise could influence the cardiorespiratory oscillators and manifest as enhanced RSA. We measured RSA during FV and CV in pigs under control conditions and after oleic acid induced acute lung injury. Methods ======= The Committee for Animal Experimentation at the University of Manitoba approved the study. The study was done in ten, 30--40 kg pigs. Pigs were sedated with midazolam (0.5 mg/kg)/ketamine (12 mg/kg)/atropine (0.6 mg) IM. When sedated, anaesthesia was induced with isoflurane 5% in oxygen by a tight fitting nose cone and the pig was intubated endotracheally. Anaesthesia was switched to a total intravenous technique using propofol (8 mg/kg/hr)/ketamine (2 mg/kg/hr) with doxacurium (10 mg/hr) for muscle relaxation. Arterial blood gases, airway pressures and respiratory system compliance (Crs) were determined at baseline. A Respironics Esprit^®^ventilator was used for the duration of the experiment. The fractal-breathing file used to program the ventilator is shown in Figure [1](#F1){ref-type="fig"}. Data were obtained from an awake spontaneously breathing teenager sitting at rest. The mono-fractal analysis of the breathing file is shown in Figure [2](#F2){ref-type="fig"}. With FV the ventilator functioned as a volume divider -- meaning the instantaneous minute ventilation product was constant. When rate was high, tidal volume was low and vice versa. Thus, minute ventilation was unchanged on converting from CV to FV or vice versa. The healthy animals were randomly allocated at recording period 1 to either FV or CV and then switched to the other mode for recording period 2. Simultaneous measurements of delivered tidal volume (ml), inspiratory and expiratory flow and electrocardiogram (ECG) were recorded at 400 samples/sec. RSA was determined as follows: the longest R-R interval in expiration was subtracted from the shortest R-R interval in inspiration (msec) -- from data recorded to a digital acquisition system. If the difference in R-R interval was a positive value this was considered \'positive RSA\'; if the difference determined had a negative value, then this was considered \'negative RSA\'. If the preceding beat (R wave) was initiated in inspiration then this beat interval was considered to have occurred during inspiration and vice versa for expiration (Figure [3](#F3){ref-type="fig"}). Eighty to 100 breathing cycles were analysed in each ventilation mode. Mean respiratory rate was set to 20 breaths/min so data collection was over 4 -- 5 min in each mode. The lungs were then injured by infusion of oleic acid intravenously to simulated acute respiratory distress syndrome. Blood gas, airway pressures and Crs were repeated. After injury, animals were randomly allocated to either of the two ventilation modes for recording period 3 and switched to the other mode for period 4. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Spontaneous human breathing pattern. Breathing pattern used to program the fractal ventilator -- data from awake human. Red line -- mean rate. ::: ![](1465-9921-6-41-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Monofractal Analysis of the breathing pattern. Monofractal or residual dispersion (RD) analysis -- one of a number of ways that fractal behaviour can be analysed \[25\]. RD is defined as standard deviation/mean × 100. This analysis uses data from Figure 1 with pooled data (or window sizes) based on powers of 2^n^following a log-log transform. A linear relationship indicates fractal behaviour, if the slope of the line lies between -- 0.5 and 0. If these conditions are met the data fits an equation of the form y = 1/x^α^defining a power law; with slope = α. The fractal dimension (D) is defined a 1 - α. D was 1.27 over 2.71 decades (log 512) with R^2^= 0.97. D of 1.0 defines a completely homogeneous data set, and 1.5 defines a completely random set (white noise). In between these two boundary conditions the data has fractal characteristics. ::: ![](1465-9921-6-41-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Calculation of Respiratory Sinus Arrhythmia (RSA). Computer generated strip chart recording from data acquisition system showing how respiratory sinus arrhythmia (RSA) was determined during mechanical ventilation. Top panel -- delivered tidal volume (millilitres), middle panel -- inspiratory and expiratory flow, bottom panel -- electrocardiogram (ECG). Below each heartbeat is the time in milliseconds (msec) of the beat interval. If the preceding beat was initiated in inspiration then this beat interval was considered to have occurred during inspiration and vice versa for expiration. In this respiratory cycle the RSA was positive at 120 msec (817 -- 697 msec). ::: ![](1465-9921-6-41-3) ::: Results ======= No cross-over effects were seen for measurements of arterial blood gases, airway pressures or Crs between groups at either baseline or following infusion of oleic acid. One animal died following administration of oleic acid. Data were pooled for baseline and oleic acid time periods to demonstrate the nature of the lung injury. At baseline, over the measurement period, tidal volume was modestly higher with FV than CV; 225 mL versus 204 mL respectively (p = 0.039). Following lung injury, tidal volume did not differ; CV 188 mL versus 202 mL with FV (p = 0.26). When mean tidal volume was determined over a longer time interval (500 -- 800 breath cycles), there were no differences between tidal volumes at baseline; CV, 203 ± 13 mL and FV, 208 ± 17 mL, or following oleic acid lung injury; CV, 188 ± 16 mL and FV, 185 ± 12 mL. At baseline PaO~2~was 248 ± 22 mm Hg, and decreased to 121 ± 19 mm Hg following oleic acid infusion; PaCO~2~was 43 ± 4 mm Hg and increased to 54 ± 6 mm Hg with oleic acid; Crs decreased from 1.00 ± 0.09 mL/cm H~2~O/kg to 0.55 ± 0.08 mL/cm H~2~O/kg after lung injury. As calculated, RSA was usually positive by our definition. Mean RSA was the global average of both positive and negative determinations with each breathing cycle in a measurement period of 80 -- 100 breaths. For further analysis, the incidence of mean positive RSA (RSA determined when only positive; presumed beneficial) and mean negative RSA (RSA determined when only negative; presumed detrimental) were examined. Mean RSA and mean positive RSA were nearly double with FV, both at baseline and following lung injury. At baseline, mean RSA = 18.6 msec with CV and 36.8 msec with FV (n = 10; p = 0.043); post oleic acid, mean RSA = 11.1 msec with CV and 21.8 msec with FV (n = 9; p = 0.028); at baseline, mean positive RSA = 20.8 msec with CV and 38.1 msec with FV (p = 0.047); post oleic acid, mean positive RSA = 13.2 msec with CV and 24.4 msec with FV (p = 0.026). The percent positive RSA at baseline was 88 ± 10% with CV and 95 ± 6% with FV (p = 0.051 between groups); following oleic acid injury the percent positive RSA moderately decreased in both groups to 79 ± 26% with CV and 84 ± 18% with FV (p = 0.079 between groups). Percent negative RSA was (100% -- calculated positive RSA%) for each measurement period. At baseline the percent negative RSA was 12 ± 10% with CV and 5 ± 6% with FV (p = 0.051); following oleic acid injury the percent negative RSA was 21 ± 26% with CV and 16 ± 18% with FV (P = 0.079). The differences in mean RSA between groups is examined in more detail in Figure [4](#F4){ref-type="fig"} and [5](#F5){ref-type="fig"}. In Figure [4](#F4){ref-type="fig"}, a regression through the origin (blue line) was fit to baseline data, using weighted least squares with weights proportional to 1/(RSA with CV). Estimated slope was 1.95 ± 0.34. FV points are above the red line of identity in 8 of 10 cases. In Figure [5](#F5){ref-type="fig"}, post oleic acid injury, estimated slope = 1.90 ± 0.46. We also analysed heart rate variability over more than 500 breaths. At baseline the coefficient of variation for heart rate was 2.2% during CV and 4.0% during FV. Following oleic acid the variation was 2.1 vs. 5.6% respectively. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### RSA with Fractal Ventilation (FV) vs. RSA with Conventional Ventilation (CV) at baseline. Regressions through the origin (blue line) were fit at baseline, using weighted least squares with weights proportional to 1/(RSA with CV). The estimated slope = 1.95 with standard error = 0.34; p = 0.010 for testing slope = 1 vs. slope \>1. Red line has slope = 1. ::: ![](1465-9921-6-41-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### RSA with Fractal Ventilation (FV) vs. RSA with Conventional Ventilation (CV) following oleic acid injury. Regressions through the origin (blue line) post oleic acid injury, using weighted least squares with weights proportional to 1/(RSA with CV). Estimated slope = 1.90, standard error = 0.46 and p = 0.044. Red line is slope = 1. ::: ![](1465-9921-6-41-5) ::: Discussion ========== RSA is a complex interaction between 3 nonlinear oscillators -- the oscillatory pattern seen with breathing, the interface with discharge in the brain stem and the oscillation of the heartbeat. By flipping a switch on a mechanical ventilator to initiate a fractal-breathing pattern with its physiological or 1/f noise, an enhanced communication is evident between these three oscillators. A physiologically important entrainment has occurred. Hayano and Yasuma have advanced the hypothesis that RSA is an intrinsic resting function of the cardiorespiratory system\[[@B13]\]. In this broader context, RSA is postulated to \"save cardiac and respiratory energy by suppressing unnecessary heartbeats during expiration and ineffective ventilation during waning phases of perfusion.\" By switching to FV, RSA has increased, thereby enhancing this intrinsic resting function of the cardiorespiratory system. This study shows that RSA decreases with severe lung injury following oleic lung injury, but fractal breathing remains associated with greater RSA. Heart rate variability is greater both at baseline and following lung injury with FV. Greater heart rate variability has been identified as an index of health \[[@B4],[@B14]\]. Recent work by Amaral and colleagues\[[@B15]\] provides a mathematical model for why FV could enhance RSA in the context of emerging complex dynamics in signalling networks. They demonstrate that complex dynamics such as 1/f scaling of power spectra (as in fractal time sequences for heart rate or respiratory rate) can be generated when simple systems meet two requirements i) they manifest \"small world\" topology and ii) demonstrate noisy input. When signal input has these two characteristics and individual units in a network react to their input with physiological responses, the system demonstrates robustness to degradation. In their model, if long-range connectivity is low, model stability can be sustained by increased noise or vice versa. If the two conditions described by Amaral are met with FV, an important additional explanation is offered for how this noisy life support device may be an improvement over a conventional ventilator. The first condition is clearly met as we have added extraneous fractal noise with the computer-controlled variability file. The second condition is less obvious. However, the \"small world\" conditions for information transmission -- the connections over long range -- may be provided by the neural connections themselves. Such may be physically representative of the vagal afferents with lung stretch and efferents to the sinus node of the heart as linked through the brain stem where cardiorespiratory neurons are locally interconnected. As Amaral et al.\[[@B15]\] suggest, \"the model may also provide a robust way to generate fluctuations that closely resemble physiological signals, which could be implemented in medical devices such as mechanical ventilators\...\" Programming a mechanical ventilator with normal physiological signals as with FV provides a means to generate the noisy fluctuations that in this circumstance enhances RSA; a complex dynamic between the respiratory, brain stem and cardiac oscillators. Thus, adding physiological noise with FV appears a working example of the emergence of complex dynamics as modelled mathematically\[[@B15]\]. Previous work has demonstrated that FV can improve gas exchange and respiratory mechanics by increased surfactant phospholipid\[[@B11]\] and through stochastic resonance -- the addition of noise to an input signal to enhance an output in a nonlinear system\[[@B12]\]. Suki and colleagues\[[@B16]\] analysed the variable inflation pressures seen with FV and showed how this noisy signal can enhance recruitment of collapsed alveoli with their nonlinear opening characteristics. Findings in this study define a new mechanism whereby FV differs from conventional mechanical ventilation -- by enhancing RSA -- both in health and following lung injury to simulate acute respiratory distress syndrome. Unlike the work by Hayano et al.\[[@B5]\], we did not measure shunt fraction, or dead space ventilation in these experiments. Changes in either variable were not expected given that the data measurement periods were limited to 4 -- 5 minutes to determine RSA in each 30 min experimental period. Previous work from the laboratory suggests that advantageous changes in either dead space or shunt fraction occur over a minimum time course of 90 -- 120 minutes following oleic acid injury with FV. However, this experiment showed that changes in RSA were measurable over a very short time period when switching between modes of ventilation. The relationship between positive pressure ventilation and RSA is controversial \[[@B17]-[@B20]\]. Some studies indicate RSA remains intact while others show a reversal of the pattern. Importantly, Taha et al.\[[@B20]\] suggest that vagal feedback from pulmonary stretch receptors is obligatory for the generation of neurally mediated RSA. Positive RSA with spontaneous ventilation (negative pressure breathing) couples increased venous return to the inspiratory phase when oxygen tension is maximized. Increased venous return does not occur during inspiration with positive pressure ventilation. This change decouples augmented venous return and maximal oxygenation. However, the timing of the positive RSA in relation to where in the expiratory cycle the longest R-R interval occurs could still have physiological relevance. Thus, if the first beat in the expiratory cycle has the longest R-R interval, alveolar oxygen levels would be higher than if the positive RSA was associated with a beat later in the expiratory cycle. Such seems to be the case with our experimental conditions. By way of example, in 5 experiments during baseline measurement of RSA while in CV mode, there were only 2 heartbeats in each expiratory cycle over the measurement period comprising approximately 100 beats. Positive RSA would be recorded for a given R-R interval irregardless of the beat order if either of the two expiratory R-R intervals was greater than the inspiratory R-R interval, but in 82 ± 8 % of cases the longest R-R interval occurred with the first expiratory beat. As a further example, examination of Figure [3](#F3){ref-type="fig"} during FV indicates that the greatest R-R interval occurs with the first expiratory heartbeat when 4 heartbeats are initiated within this expiratory cycle. Anaesthesia, both type and depth can influence RSA \[[@B21],[@B22]\]. However, the depth of anaesthesia was a controlled variable in our experiment as we employed a continuous infusion of propofol/ketamine with a cross-over design, thereby minimizing time effects. More elegant means to measure RSA have been published. One such example is from recent work by Giardino et al.\[[@B23]\], which describes a sophisticated spectral analysis for determination of RSA in human subjects. They had patients breath rhythmically by following a respiratory-pacing stimulus displayed on a computer monitor. If for example, the patient was breathing at 12 breaths/min, RSA was measured as the amplitude of a 0.20-Hz sine wave fitted to the heart rate series. Such an analysis could be undertaken to assess RSA with CV where breathing frequency is monotonous, but not for FV where the breathing rate is variable over time. Thus, we chose a simpler approach to determine RSA. Conclusion ========== Our findings may have important implications for the design of life support devices used clinically\[[@B3]\]. Buchman has suggested that multiple organ dysfunction, often seen in patients requiring life support in the intensive care unit, is a consequence of lost coupling between \"communicating\" organ systems\[[@B24]\]. The inability to re-couple following the perturbation of critical illness may be an important reason why multi-organ system dysfunction is so lethal\[[@B6]\]. Fractal ventilation, by augmenting respiratory sinus arrhythmia, may be one approach to enhance organ system re-coupling. Competing interests =================== Dr. Mutch is co-founder of Biovar Life Support Inc., which has developed the mechanical ventilator described in this paper. Worldwide exclusive rights to this ventilator have been licensed to Respironics Inc. To date no ventilators have been sold clinically. In the event of sales of this ventilator, Dr. Mutch and the University of Manitoba would stand to gain financially. None of the other authors have a financial interest in the ventilator. Authors\' contributions ======================= WACM conceived the study, helped analyse and interpret the data and helped write the paper. MRG helped conduct the study, analyse and interpret the data and write the paper. LGG conducted the study, collated data, and helped prepare the figures. JFB analysed the data, conducted the statistical analysis, helped prepare the figures and helped write the paper. Acknowledgements ================ Biovar Life Support Inc. provided financial support. Respironics Inc. provided the software development for the fractal ventilator and configured it to their Esprit^®^ ventilator.	PubMed Central	2024-06-05T03:55:57.407102	2005-5-9	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1142342/", "journal": "Respir Res. 2005 May 9; 6(1):41", "authors": [ { "first": "W Alan C", "last": "Mutch" }, { "first": "M Ruth", "last": "Graham" }, { "first": "Linda G", "last": "Girling" }, { "first": "John F", "last": "Brewster" } ] }
PMC1142343	Background ========== Bases in DNA suffer damage both from normal cellular functions such as metabolism and from external oxidative stress and radiation. Naturally the cell has several lines of defense against direct lesions and ensuing mutagenic mispairings \[[@B1]-[@B3]\]. Oxidation of the base guanine (G) often results in the formation of 8-oxo-G (7,8-dihydro-8-oxo-2\'-deoxyguanosine) \[[@B4]\]. During replication, 8-oxo-G can pair both with cytosine (C) and adenine (A) \[[@B5]\]. Following another round of replication, the 8-oxo-G:A pairs give rise to G:C to T:A mutations (if not corrected). In Escherichia coli, 8-oxo-G:C pairs are detected by the DNA glycosylase MutM (formamidopyrimidine glycosylase), which subsequently excises the 8-oxo-G from the DNA leaving an abasic site where the strand is nicked at both the 3\' and 5\' ends \[[@B6]\]. The abasic site is further processed by the base excision pathway (BER), eventually leading to the insertion of a G opposite the remaining C. The action of MutM brings the number of adenines A misincorporated opposite 8-oxo-G during replication down to around one per one million bases, even in cells challenged by H~2~O~2~\[[@B7],[@B8]\]. In E. colithe 8-oxo-G:A pairs are detected by another DNA glycosylase, MutY \[[@B9],[@B10]\], which excises the mispaired adenine A leaving an abasic site. The abasic site opposite the unpaired 8-oxo-G is further processed by the BER pathway, resulting in an 8-oxo-G:C pair. If on the other hand a G in the dGTP pool is initially oxidized and subsequently incorporated opposite an A during replication, the action of MutY increases the mutagenic conversion rate of T:A to G:C. Experimentally, this is seen in strains lacking the MutT enzyme \[[@B11]\] responsible for the hydrolysis of 8-oxo-dGTP to 8-oxo-dGMP \[[@B12]\]. Both the biochemical and mechanistic functions of the excision process and the specific recognition of the base to be excised have been unraveled for many DNA glycosylases \[[@B7],[@B13],[@B14]\]. The main step is flipping the base to be excised out of DNA and into a cleft in the DNA glycosylase. This extra-helical state is associated with a kinking of the DNA through an angle of 60°--80° depending on the particular DNA glycosylase. Even though questions still remain to be answered in this area, the main challenge is to understand how the mismatched oxidized base pair is located among all the normal, correctly paired ones \[[@B13]\]. Direct visualization using atomic force microscopy (AFM) reveals that the human 8-oxo-G DNA glycosylase hOGGl and the E. coliDNA glycosylase AlkA kink error-free DNA in the same way they kink DNA during the excision of a damaged base \[[@B15]\]. It is thus likely that some DNA glycosylases also flip correctly-paired bases into the active site cleft during their search for excision targets \[[@B16]\]. Furthermore, in vitro studies indicate that some DNA glycosylases including MutY move along the DNA while scanning its integrity \[[@B17]\]. Until recently it was more or less implicitly assumed \[[@B16]\] that the individual DNA glycosylases locate damaged DNA sites independently of each other. However, a bold new hypothesis suggests a certain sub-class of DNA glycosylases might cooperate in order to locate the damaged sites more quickly \[[@B18]\]. This sub-class is defined by the presence of an evolutionarily well-conserved \[4Fe-4S\]^2+^cluster and includes MutY and endonuclease III, but not MutM or AlkA \[[@B1]\]. Endonuclease III recognizes oxidized and ring-fragmented pyrimidines, while AlkA recognizes a wide spectrum of alkanated base adducts (both alkanated pyrimidines and purines). Thus the \[4Fe-4S\] cluster is not obviously associated with the recognition of specific substrates. Initial investigations suggested that the cluster is not redox active under physiological conditions \[[@B19]\]. This led to the speculation that the \[4Fe-4S\]^2+^cluster might be a rare example of a metal cluster with a purely structural role \[[@B20]\]. However, it was recently shown in vitro that upon binding of MutY to DNA, an electron is injected into the DNA and the \[4Fe-4S\] cluster is involved in this redox reaction, presumably changing its oxidation level from 2+ to 3+ \[[@B18]\]. The authors then went on to hypothesize that the MutY enzymes communicate through currents in the DNA and in this way accelerate error the location process. An error-free stretch of DNA is a good conductor, while a defective base pair introduces a huge resistance \[[@B21]\]; if a MutY enzyme receives an electron from an upstream MutY enzyme, the stretch of DNA ahead of it is thus error-free. Presumably the electron received destabilizes the binding of MutY to this error-free stretch of DNA by changing the oxidation level of the \[4Fe-4S\]^3+^cluster back to 2+. Thus, the net-effect of the charge exchange is rapid detachment of the MutY molecules from error-free DNA, followed by binding and scanning elsewhere. Intuitively, this fast detachment of MutY enzymes from error-free stretches of DNA speeds up the location of damaged base pairs. It should perhaps be emphasized here that the proposed mechanism is speculative and has not yet been firmly verified experimentally. Nevertheless it is of interest to estimate the extent of the potential speed-up and to consider whether there are other biologically relevant and experimentally testable consequences of the proposed charge exchange mechanism. As discussed in detail below there are two relevant time scales in the proposed process. The first, τ, is the time it takes to realize that a stretch of DNA is error-free, i.e. τis the time from attachment of a MutY enzyme attach to an error-free piece of DNA until detachment and binding to another site. The second, T, is the average time it takes to locate a damaged base pair by slowly scanning the DNA, without utilizing the charge exchange mechanism. In this paper, I show that the time it takes for MutY to locate a damaged base pair is roughly ![](1742-4682-2-15-i1.gif), corresponding to a speed-up of ![](1742-4682-2-15-i2.gif). This expression for the speed-up remains valid in the presence of many other kinds of charge exchanging repair enzymes. However, in this case, Tis the average time that it takes for anyrepair enzyme to locate the error by scanning alone. Secondly, I also point out that the charge exchange mechanism alters the response to over-expressed repair enzymes. As the total number of repair enzymes is increased the efficiency of the charge exchange mechanism decreases. In this way, doubling the number of repair enzymes only shortens the location time by 30%, not by 50% as in the independent scanning scenario. The gain relative to the independent scanning scenario is thus smaller. Results ======= Model ----- The model is presented in Figure [1](#F1){ref-type="fig"}. The repair enzyme MutY contains an evolutionarily well-conserved \[4Fe-4S\] cluster that is suspected to change its charge configuration from 2+ to 3+ upon binding to DNA \[[@B18]\]. Binding is thus associated with the emission of an electron into the DNA, while upon receipt of an electron from DNA the MutY-DNA binding complex is destabilized. As only error-free stretches of DNA are able to transport the electron from a MutY enzyme to a neighboring one \[[@B21]\], this charge exchange enables MutY to liberate scanning resources quickly from error-free stretches of DNA \[[@B18]\]. To make the argument and calculations as transparent as possible, I first consider the scenario where only MutY enzymes participate in the charge exchange. However, in real cells, many different kinds of repair enzymes each are expected to participate in the charge exchange, each specialized for fixing a specific kind of damage. This more general scenario is the focus of the next section. Finally, the effect of a finite scan length before MutY spontaneously detaches from the DNA is considered. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### The model. a) The left MutY repair enzyme is bound to DNA and slowly progress to the right while it scans the integrity of base pairing. The \[4Fe-4S\] cluster in MutY is in a 3+ charge configuration when bound to DNA, but a 2+ configuration when not bound (right MutY). b) Upon binding to DNA the right MutY enzyme emits an electron into the DNA and changes the charge of its \[4Fe-4S\] cluster from 2+ to 3+. c) If the DNA is free of errors, the emitted electron travels along the DNA until it reaches the left MutY enzyme. Here the electron changes the charge of the \[4Fe-4S\] cluster to 2+ and thus destabilizes the DNA binding of this MutY enzyme. The left MutY enzyme then attaches to and scans a different section of DNA that is more likely to contain an error, d) If on the other hand the DNA segment between the two MutY enzymes contains an error, the electron never reaches the left MutY enzyme, which then keeps scanning the DNA until it reaches and fixes the error. The charge exchange thus selectively frees up resources from error free patches of DNA. The model is also described in \[18,24\]. ::: ![](1742-4682-2-15-1) ::: ### Only MutY participates in the charge exchange How long is the time t~location~that elapses between damage to a base pair and detection of the error by MutY? The faulty base pair is either located by a MutY enzyme that happened to be bound to DNA downstream of the error at the time of damage or by one that subsequently binds to DNA downstream of the error and then scans the DNA until it finds the damaged base pair. In the regime where the charge exchange mechanism markedly accelerates the error location, the second mechanism dominates. Let t~location~denote the typical location time and vthe scanning velocity of MutY. The rate at which a MutY enzyme randomly docks onto a specific base pair and starts scanning is denoted by k. The probability that a MutY enzyme lands within a distance of vt~location~of the error in the time interval t~location~can be estimated as ![](1742-4682-2-15-i3.gif). This probability is of order 1, since in the time t~location~a MutY enzyme typically arrives at the faulty base pair. Thus ![](1742-4682-2-15-i4.gif) A more detailed derivation yields the same result apart from a factor of 1.3 (See additional file: MutY\_detailed\_derivation.pdf). However this factor is not reliable as no model fully incorporates all biological processes. Consequently I have made no attempt to keep track of such factors in the following argument. The average docking rate kcan be expressed as ![](1742-4682-2-15-i5.gif) where τis the time between two successive binding events for a single MutY enzyme. N^MutY^is the total number of MutY enzymes. Lis the total number of base pairs in DNA. T= L/v/N^MutY^is the time it takes for the MutY enzymes to scan all the bases of DNA once. It is here assumed that all the MutY enzymes belong to a single freely-exchanging pool and that MutY is equally able to bind to all Lbase pairs of DNA. Considering that DNA is folded into chromatin superstructures, this is probably not true, but as a first rough estimate it suffices. In terms of Tand τthe location time is (combining Eqs. (1) and (2)) ![](1742-4682-2-15-i6.gif) In the traditional scenario where the MutY enzymes scan the DNA independently to locate the mispaired sites, 1/vis the time it takes a single MutY enzyme to check the integrity of one base pair. According to standard Poisson statistics the location time without charge exchange mediated cooperation between the MutY enzymes is L/v/N^MutY^= T. Cooperation thus gives a speed-up of approximately ![](1742-4682-2-15-i2.gif). ### Many different kinds of repair enzymes The functionally central \[4Fe-4S\] cluster is also present in other repair enzymes e.g. endonuclease III. Very likely these repair enzymes are also are able to inject charges into DNA and participate in electrical scanning. Consequently these charge sensitive repair enzymes are also \'attracted\' to the damaged DNA pair in exactly the same way as MutY. Thus in the above model and calculation, \'MutY\' can be replaced by \'any repair enzyme participating in DNA mediated charge transport\' (repair enzyme). Likewise the calculated location time t~location~is the time before the first repair enzyme locates the damaged site and Tis the average time it takes for any repair enzyme to find the site without using currents. Here I have implicitly assumed that both the scan velocity vand the time τbetween successive binding events are of the same orders of magnitude for all repair enzymes i.e. MutY is a typical repair enzyme. Biologically, the time t~location~is not the most relevant one as the first repair enzyme that arrives at the damaged base pair is probably unable to fix the damage. On average the first MutY enzyme is the N/N^MutY^repair enzyme to arrive at the damaged site. Thus the MutY location time ![](1742-4682-2-15-i7.gif) Here Nis the total number of repair enzymes. N/N^MutY^can also be expressed as T^MutY^/T, where T^MutY^is the time it takes for the MutY enzymes to locate the site by scanning alone. Using Eq. (3) the MutY location time is ![](1742-4682-2-15-i8.gif) The speed-up relative to the independent scanning of the genome is thus again ![](1742-4682-2-15-i2.gif). However this time, T, is the time it takes for any repair enzyme to locate the damage. ### Finite scan length MutY is known to detach from DNA spontaneously after scanning in the order of 100 base pairs (bp) \[[@B17]\]. In order to estimate the resulting effect, if any, on the MutY location time, Eq. (5) is derived in a slightly different manner. The MutY enzyme that eventually locates the damage typically docks on to DNA within a distance Δ from the faulty base pair. Δ fulfills two constrains. First it is less than 100 bp in order to avoid spontaneous detachment of the MutY from the DNA before it has scanned the damaged site. Secondly it is so small that the probability that another repair enzyme will dock on to the DNA in front of MutY is less than 1. As the distance from MutY to the error is roughly Δ and the time it takes to scan the Δ intervening bases is Δ/v, the latter probability is approximately kΔΔ/v. Thus Δ ≤ ![](1742-4682-2-15-i9.gif). In terms of Δ, the MutY location time ![](1742-4682-2-15-i10.gif) is determined as above by setting the probability that a MutY enzyme docks within a distance Δ in the time interval ![](1742-4682-2-15-i10.gif) equal to 1 i.e. k^MutY^![](1742-4682-2-15-i10.gif) Δ = 1 or ![](1742-4682-2-15-i11.gif) T^MutY^= N^MutY^/ L/v= (k^MutY^vτ)^-1^is the location time in the scenario, where the MutY enzymes act independently of each other. The length ![](1742-4682-2-15-i12.gif) is the distance over which the charge exchange typically takes place. In the section \'Many different repair enzymes\', lwas the average distance between two repair enzymes vT. However, in vivo, other factors might limit land the expression ![](1742-4682-2-15-i13.gif) is the most general expression for the reduction in location time due to the charge exchange mechanism: T^MutY^/ ![](1742-4682-2-15-i10.gif). Estimating order of magnitude ----------------------------- Since no experimental data exist for τand l, the efficiency of the charge exchange mechanism, Eq. (8), must be estimated. The numerator Δ is the smallest of the maximal scan length 100 bp and the docking distance ![](1742-4682-2-15-i14.gif). I assume ![](1742-4682-2-15-i14.gif) ≤ 100 bp, with equality as the most likely option, as anything else seems inefficient. The distance lis estimated as the average distance between the repair enzymes vT= L/N. With these approximations the reduction is ![](1742-4682-2-15-i15.gif) ≥ vT/100 bp = 5·10^4^/N, where Nis the total number of repair enzymes with a charge exchange mechanism similar to MutY. I have assumed that the length of E. coli\'sDNA, L, is 5·10^6^base pairs. Unfortunately Nis unknown. The numbers of the two \[4Fe-4S\]^2+^-containing repair enzymes, MutY and endonuclease III, are estimated to be 30 and 500 respectively and the number of MutM repair enzymes is estimated at 400 \[[@B22]\]. The primary target of MutM, 8-oxo-G, is estimated to constitute 5% of all adducts due to oxidative damage \[[@B4]\]. All in all it seems reasonable that the total number of repair enzymes participating in the charge exchange mechanism is significantly smaller than 50000, and that a speed-up of order 10 is realistic. Notice this would correspond to a typical scan length that is 10 times smaller than the maximal one (100 bp) and that l≈ 1000 bp. Discussion ========== The implications of a proposed charge exchange mediated cooperation between repair enzymes in locating defects in single base pairs have been considered. From the theoretical point of view taken here, this mechanism is likely to speed up location by a factor of order 10 compared to the traditional scenarios in which the repair enzymes scan the genome for errors independently. In this paper the speed-up was quantified in terms of the time it takes to locate a damaged base pair t~location~. t~location~has to be considerably shorter than the replication time, which in E. coliis in the order of one hour. To be concrete, assume t~location~is 20 minutes. For the 30 MutY enzymes in E. colithe calculated efficiency of the charge exchange mechanism translates into a reduction in the necessary scan velocity from 125 bp/s to 13 bp/s. For comparison, the scan velocity for RNA polymerase is 50 bp/s, while for DNA polymerase it is 1000 bp/s. In the traditional independent-scanning scenario the location time Tis inversely proportional to the number of repair enzymes. Upon over-expression of all the repair enzymes the effective distance over which the charge exchange takes place, l, is reduced and the efficiency of cooperation is reduced (Eq. 8). Thus the decrease in location time ![](1742-4682-2-15-i10.gif) is smaller in the charge exchange scenario than in the traditional independent-scanning scenario, but t~location~remains shorter than T. Note that if only a small subclass, such as MutY, is over-expressed, the location time is still inversely related to the number of molecules. Assuming that the typical scan length vτremains constant during over-expression the location time is inversely proportional to the square root of the total number of repair enzymes. The important point is not the exact square root behavior but the relative insensitivity to simultaneous over-expression of all the repair enzymes. Physiologically, oxidative and radiative environments may result in an increased expression of repair enzymes \[[@B23]\], so the relative insensitivity of the location time and the coupling of the effectiveness of different kinds of repair enzymes are potentially of huge importance for mutation rates in these kinds of stress full environments. Conclusion ========== I have demonstrated that the charge transport mechanism indeed offers great potential benefit for the cell. However, only further experimentation can finally confirm the charge transport mechanism, the current status of which must be dubbed speculative. Furthermore, I have pointed out that the charge transport hypothesis, if valid, has consequences for the cellular response to stress-ful environments. In addition, the model is a simple model of protein cooperativity and one might wonder if the principles underlying it could be of practical use in apparently unrelated engineering problems. Competing interests =================== The author(s) declare that they have no competing interests. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Contains a more detailed derivation of Eq. (1), keeping track of all the numerical factors. 1 page. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ Kasper Astrup Eriksen acknowledges support from both the Danish Natural Science Research Council grant number 21-03-0284 and the Bio+IT program under the Øresund Science Region and Øforsk.	PubMed Central	2024-06-05T03:55:57.408887	2005-4-8	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1142343/", "journal": "Theor Biol Med Model. 2005 Apr 8; 2:15", "authors": [ { "first": "Kasper Astrup", "last": "Eriksen" } ] }
PMC1142344	Introduction ============ Many attempts have been made to simulate molecular processes in cellular systems. Perhaps the most active area of cellular simulation is the kinetics of metabolic pathways. Various software packages that quantitatively simulate cellular processes and are based on numerical integration of rate equations have been developed. These include GEPASI \[[@B1]\], which calculates steady states as well as reaction time behavior; V-Cell \[[@B2]\], a solver of non-linear PDE/ODE/Algebraic systems that can represent the cellular geometry; and DBsolve \[[@B3]\], which combines continuation and bifurcation analysis. The E-Cell project \[[@B4],[@B5]\], which aims to model and simulate various cellular systems, was launched in 1996 at Keio University. The first version of the E-Cell simulation system, a generic software package for cell modeling, was completed in 2001. E-Cell version2, which is a Windows version of the first E-Cell system, is now also available \[[@B6]\]. E-Cell version 3, which enables multi-algorithm simulation, is the latest version \[[@B7]\]. The E-Cell system allows the user to define spatially discrete compartments such as membranes, chromosomes and the cytoplasm. The collections of molecules in all cellular compartments are represented as numbers of molecules, which can be converted to concentrations, and these can be monitored and/or manipulated by employing the various graphical user interfaces. In addition, the E-Cell system enables the user to model not only deterministic metabolic pathways but also other higher-order cellular processes, including stochastic processes such as gene expression, within the same framework. By using the E-Cell system, a virtual cell with 127 genes that are sufficient for \"self-support\" \[[@B4]\] was developed. This gene set was selected from information about Mycoplasma genitaliumgenomic sequences and includes genes for transcription, translation, the glycolysis pathway for energy production, membrane transport, and the phospholipid biosynthesis pathway for membrane production. On the basis of existing models of single pathways and enzymes, various in silico models of human red blood cell (RBC) metabolism were first developed by Joshi and Palsson \[[@B8]-[@B11]\]. Subsequently, other groups developed RBC models \[[@B12]-[@B15]\]. The RBC is thought to be a good target for biosimulation because extensive studies over the last three decades have generated extensive biochemical data on its enzymes and metabolites. Moreover, the RBCs of many species, including humans, do not contain a nucleus or carry genes. This means that gene expression can be excluded from the model, which greatly simplifies the biosimulation. RBCs take up glucose from the plasma and process it by glycolysis, which generates the ATP molecules that are used in other cellular metabolic processes. The ATP molecules are mostly consumed by the ion transport systems that maintain the osmotic balance of the cell. Here we describe our computer model of the human RBC, which we developed on the basis of previous models \[[@B8]-[@B13]\]. Our prototype model of the human RBC consisted only of glycolysis, the pentose phosphate pathway, nucleotide metabolism and simple membrane transport systems such as the Na^+^/K^+^antiport channel. Here, we have employed this prototype model to reproduce the pathological condition of glucose-6-phosphate dehydrogenase (G6PD) deficiency. This is the most common hereditary enzyme deficiency in RBCs; it causes anemia, and more than 400 varieties of G6PD deficiency have been identified \[[@B16]\]. The deficiency is known to exert only mild effects as it does not cause clinically significant problems in most cases, except upon exposure to medications and foods that cause hemolysis. Computer simulations for analyzing this deficiency have been reported \[[@B17]-[@B19]\], but these simulation models consisted only of glycolysis and the pentose phosphate pathway. We found that including the glutathione (GSH) biosynthesis pathway and the glutathione disulfide (GSSG) export system, which are involved in suppressing oxidative stress, improved the ability of the model to reflect the real diseased RBC. This suggests that these pathways may compensate for the consequences of G6PD deficiency in human RBCs. Methods ======= Development of the prototype model and simulation experiments ------------------------------------------------------------- The E-Cell system version 1.1 was used as the simulation platform in this work. The software can be downloaded from <http://www.e-cell.org/>. Our prototype model of the RBC was developed on the basis of the whole-cell model of Joshi and Palsson \[[@B8]-[@B11]\] with slight modifications (Figure [1](#F1){ref-type="fig"}). We modified the model to represent the oxidant-induced decrease of hexokinase and pyruvate kinase, and the maximum activity of these enzymes was allowed to change according to the ratio of GSH and GSSG. The equations and parameters used are derived from the literature \[[@B17]\]. The parameters and kinetic equations in the original model of Joshi and Palsson were replaced with those obtained from the literature \[[@B17],[@B20],[@B21]\] (Table [1](#T1){ref-type="table"}) in order to fit the model to the measured concentrations during the calculation of the steady state. The steady state obtained had concentrations of many metabolites that were very close to those in real RBCs (Table [2](#T2){ref-type="table"}). However, the concentrations of several metabolites, namely adenosine, hypoxanthine, inosine, 5-phosphoribosyl 1-phosphate and ribose 1-phosphate, differed from the experimental values. These differences were due to the kinetic parameters and equations used, and because the nucleotide metabolism in the original model was represented as simple first-order kinetics or equilibrium. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Metabolic map of the prototype RBC model. The circles are metabolic intermediates and ions. These molecular species are defined as \"Substance\" in the E-Cell system. The boxes are enzymes and reaction processes. Their rate expressions are defined as \"Reactor\" whereas the enzyme molecules are defined as \"Substance\". ::: ![](1742-4682-2-18-1) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Enzymes and rate equations of the prototype model ::: Enzymes Abbreviation Group Reaction mechanism Reference --------------------------------------------- ------------------ ------------ ---------------------------- --------------- Glutathione turnover GSHox PPP Chemical reaction 24 Glutathione reductase (NADPH) GSSGR PPP Ordered Bi Ter mechanism 24 Glutathione reductase (NADH) GSSGR2 PPP Michaelis Menten mechanism 24 Glucose 6-phosphate dehydrogenase G6PD PPP Ordered Bi Bi mechanism 17 6-Phosphogluconolactonase 6PGLase PPP Michaelis Menten mechanism 17 6-Phosphogluconate dehydrogenase 6PGLDH PPP Ordered Bi Ter mechanism 24 Ribose 5-phosphate isomerase R5PI PPP Uni Uni mechanism 25 Xylulose 5-phosphate isomerase X5PI PPP Uni Uni mechanism 25 Transketolase I TK1 PPP Ping-Pong Bi Bi mechanism 25 Transketolase II TK2 PPP Ping-Pong Bi Bi mechanism 25 Transaldolase TA PPP Ping-Pong Bi Bi mechanism 25 Hexokinase HK Glycolysis 26 Phosphoglucoisomerase PGI Glycolysis Uni Uni mechanism 25 Phosphofructokinase PFK Glycolysis 27 Aldolase ALD Glycolysis Ordered Uni Bi mechanism 25 Triose phosphate isomerase TPI Glycolysis Uni Uni mechanism 25 Glyceraldehyde phosphate dehydrogenase GAPDH Glycolysis Chemical reaction 20 Phosphoglycerate kinase PGK Glycolysis Chemical reaction 20 Diphosphoglycerate mutase DPGM Glycolysis Michaelis Menten mechanism 20 Diphosphoglycerate phosphatase DPGase Glycolysis Michaelis Menten mechanism 20 Phosphoglyceromutase PGM Glycolysis Chemical reaction 20 Enolase EN Glycolysis Chemical reaction 20 Pyruvate kinase PK Glycolysis 28 Pyruvate transport process PYRtr Transport Michaelis Menten mechanism 22 Lactate dehydrogenase LDH Glycolysis Chemical reaction 20 Lactate transport process LACtr Transport Michaelis Menten mechanism 22 Leak of Potassium K\_Leak Transport 9 Leak of Sodium Na\_Leak Transport 9 Sodium/potassium pump Pump Transport 9 Adenosine transport process ADEtr Transport Chemical reaction 13 AMP phosphohydrolase AMPase NM Chemical reaction 20 Adenosine deaminase ADA NM Michaelis Menten mechanism 20 Adenosine kinase AK NM Michaelis Menten mechanism 20 Adenylate kinase APK NM Chemical reaction 20 Adenosine triphosphate phosphohydrolase ATPase NM Chemical reaction 8 Adenosine monophosphate deaminase AMPDA NM Michaelis Menten mechanism 20 Inosine monophosphatase IMPase NM Michaelis Menten mechanism 8 Purine nucleotide phosphorylase PNPase NM Chemical reaction 23 Phosphoribosyl pyrophosphate synthetase PRPPsyn NM 8 Adenine phosphoribosyl transferase ADPRT NM Michaelis Menten mechanism 8 Hypoxanthine-guanine phosphoryl transferase HGPRT NM Michaelis Menten mechanism 8 Hypoxanthine transport process HXtr NM 29 PPP, Pentose phosphate pathway; NM, Nucleotide metabolism. ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Steady state of the RBC model. ::: Concentration (mM) ----------------------------------- --------- -------------------- --------------------- 1,3-Diphosphoglycerate 13DPG 1.83E-04 4.00E-04 2-Phosphoglycerate 2PG 4.16E-03 1.40E-02 ± 5.00E-03 3-Phosphoglycerate 3PG 4.62E-02 4.50E-02 Adenosine ADO 8.93E-06 1.20E-03 ± 3.00E-04 Dihydroxy acetone phosphate DHAP 1.35E-01 1.40E-01 ± 8.00E-02 Erythrose 4-phosphate E4P 1.17E+00 \- Fructose 6-phosphate F6P 6.39E-02 1.60E-02 ± 3.00E-03 Fructose 1,6-diphosphate FDP 1.14E-02 7.60E-03 ± 4.00E-03 Glucose 6-phosphate G6P 1.96E-01 3.80E-02 ± 1.20E-02 Glyceraldehyde 3-phosphate GA3P 6.24E-03 6.70E-03 ± 1.00E-03 Gluconolactone 6-phosphate GL6P 7.62E-06 \- Gluconate 6-phosphate GO6P 2.72E+00 \- Glutathione GSH 3.21E+00 3.21E+00 ± 1.50E+00 Glutathione GSSG 1.03E-04 \- Hypoxanthine HXi 9.32E-06 2.00E-03 Inosine monophosphate IMP 5.03E-03 1.00E-02 Inosine INO 3.32E-08 1.00E-03 Potassium Ki 1.26E+02 1.35E+02 ± 1.00E+01 Lactate LACi 1.20E+00 1.10E+00 ± 5.00E-01 Nicotinamide adenine dinucleotide NAD 8.87E-02^d^ \- Nicotinamide adenine dinucleotide NADH 3.13E-04^d^ \- Nicotinamide adenine phosphate NADP 8.06E-05^d^ \- Nicotinamide adenine phosphate NADPH 6.58E-02^d^ 6.58E-02 Sodium Nai 2.27E+01 1.00E+01 ± 6.00E+00 Phosphoenolpyruvate PEP 1.89E-02 1.70E-02 ± 2.00E-03 5-Phosphoribosyl 1-phosphate PRPP 6.91E-05 5.00E-03 ± 1.00E-03 Pyruvate PYRi 6.00E-02 7.70E-02 ± 5.00E-02 Inorganic phosphate Pi 1.30E-01 1.00E+00 Ribose 1-phosphate R1P 2.12E-05 6.00E-02 Ribose 5-phosphate R5P 2.81E-04 \- Ribulose 5-phosphate RU5P 1.48E-04 \- Sedoheptulose 7-phosphate S7P 7.49E-02 \- Xylulose 5-phosphate X5P 4.30E-04 \- 2,3-Diphosphoglycerate 2,3-DPG 4.21E+00 4.50E+00 ± 5.00E-01 Adenosine diphosphate ADP 2.20E-01 2.70E-01 ± 1.20E-01 Adenosine monophosphate AMP 2.42E-02 8.00E-02 ± 9.00E-03 Adenosine triphosphate ATP 1.57E+00 1.54E-00 ± 2.50E-01 The values are given in scientific notation; E-01 denotes multiplication by 10^-1^. ^a^The initial data set was from experimental data in the literature and from predictions of previous simulation models \[12\]. ^b^The simulation was run for more than 1,000,000 seconds in simulation time until the model reached steady state. ^c^Biochemical experimental data taken from the literature and reported in Joshi and Palsson \[11\]. ^d^NAD(H) and NADP(H) pools are kept constant. ::: The parameters from the work of Jacobasch et al. \[[@B30]\] were used in the experiments simulating G6PD deficiency (Table [3](#T3){ref-type="table"}). Since the rate equation of G6PD deficiency is the same as that in the normal cell, the parameters were simply replaced in the deficiency experiment. We adopted the We.G variant of G6PD deficiency because its parameters are well described in the literature and its phenotype is rather severe. As with the original model, the oxidative load is represented as the conversion of GSH to GSSG, and the equation is expressed as a simple first-order kinetics. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Parameters for normal and deficient enzymes ::: t/2 (day) Vmax (mkat/l cells) KmG6P KmNADP (mM) KiNADPH KiATP Ki2,3DPG -------- ----------- --------------------- ------- ------------- --------- ------- ---------- Normal 27 575 67 3.7 3.1 749 2289 We.G. 2.5 10 152 3.8 0.62 180 520 These values are based on experimental data taken from the literature \[10\] ::: Expansion of the prototype model and simulation experiments ----------------------------------------------------------- The de novoGSH synthesis and GSSG export pathways (Figure [3](#F3){ref-type="fig"}) were added to the prototype model. The kinetic equations and parameters of these pathways were obtained from the literature \[[@B31]-[@B33]\] (Table [4](#T4){ref-type="table"}). Since these pathways have very low activity in normal cells, the concentrations of metabolites at the steady state were almost unchanged in the expanded model. The concentrations listed in Table [2](#T2){ref-type="table"} were used as the steady state concentrations. The conditions employed to simulate G6PD deficiency using this expanded model were the same as those of the prototype model. It is known that multidrug resistance-associated proteins (MRP1) and the cystic fibrosis transmembrane conductance regulator (CFTR) are expressed in human RBC and involved in GSH and/or GSH conjugates transport \[[@B35]\]. However, their rate equations and parameters are unavailable, so these proteins were not included in this model. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### *Pathway for the de novoof GSH and the GSSG export system. γ-GCS, γ-glutamyl cysteine synthetase; γ-CS, γ-glutamyl cysteine. ::: ![](1742-4682-2-18-2) ::: ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Rate equations and parameters of GSH synthesis and GSSG export that were used in the expanded model. ::: Rate equation for γ-glutamyl cysteine synthetase ---------------------------------------------------- ------------- ----------- Parameters for γ-glutamyl cysteine synthetase Parameter Value Reference Vmax 141.57 mM/h 31, 32 α* 0.2 31 Kmglu 1.8 mM 31 Kmcys 0.1 mM 31 KiGSH 3.4 mM 31 KmATP 0.4 mM 31 Rate equation for glutathione synthetase Parameters for glutathione synthetase Parameter Value Reference Km~γ\_GC~ 0.99 mM 33 Km~Gly~ 1,37 mM 33 Km~ATP~ 0,23 mM 33 α 2.6 33 Vmax 88.4 mM/h 33 Rate equation for GSSG export Parameters for GSSG export Parameter Value Reference Km~GSSG1~ 0.1 mM 34 Km~ATP~ 0.63 mM 34 Vm~1~ 20 μM/h 34 ::: Results and Discussion ====================== Simulation of G6PD deficiency using the prototype model ------------------------------------------------------- The prototype model was used to simulate the effects of G6PD deficiency. G6PD is a key enzyme in the pentose phosphate pathway that converts glucose 6-phosphate into gluconolactone 6-phosphate (GL6P); this simultaneously generates NADPH. The metabolic intermediate GL6P is then metabolized into ribulose 5-phosphate (Ru5P) acid viagluconate 6-phosphate (GO6P). This process also generates NADPH. This reduction power is employed by various other intracellular processes, in particular the reduction of GSSG. A major function of GSH in the RBC is to eliminate superoxide anions and organic hydroperoxides. Peroxides are eliminated through the action of glutathione peroxidase, which yields GSSG. The simulation experiments were carried out with steady state concentrations corresponding to those in the normal RBC. Sequential changes in the quantities of NADPH, GSH and ATP were observed (Figure [2](#F2){ref-type="fig"}). There is a negative peak in ATP concentration before 10 h. This was due to the shutting down of the pentose phosphate pathway. The Ru5P produced was mainly converted to fructose 6-phosphate (F6P), and this metabolite consumed ATP to make fructose 1,6-diphosphate (FDP). The FDP production led to an accumulation of dihydroxy acetone phosphate (DHAP), and the metabolite was not used to provide ATP. The high GO6P concentration could sustain normal levels of GSH concentration at the first stage of the simulation, but after the depletion of GO6P the rate of Ru5P production was drastically reduced. This decrease in Ru5P concentration led to decreased F6P concentrations. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Computer simulation time-course of metabolic intermediates. Changes in the concentrations of ATP (A), GO6P (B), GSH (C), GSSG (D), NADP (E) and NADPH (F) during the RBC simulation. The simulation was run for 200,000 seconds (Approx. 55 h) in simulation time. Concentrations change when G6PD kinetic parameters are shifted from the normal to pathological values (Table 3). ATP became depleted at around 20 h. ::: ![](1742-4682-2-18-3) ::: At around 20 h, ATP was rapidly consumed and depleted. Since ATP concentrations less than half the normal concentration have never been observed in enzyme deficiencies \[[@B36]\], cells in this condition will probably be destroyed. Although the half-life of the real G6PD-deficient We.G type RBC is known to be 2.5 days \[[@B30]\], the longevity of our computer model turned out to be much shorter (Table [3](#T3){ref-type="table"}). Since data on the concentration of metabolites in RBCs with G6PD deficiency are not available, it was not possible to determine whether the metabolite concentrations arising in our simulation experiments reflected those observed in real cells. Simulation of G6PD deficiency using the expanded model ------------------------------------------------------ It is obvious that decreased pentose phosphate pathway activity leads to faster cell death, and that the difference between the simulated cell and the real cell regarding the timing of cell death could be caused by the lack of a pathway producing GSH. This pathway may compensate for the decrease in GSH. A mature RBC normally contains 2 mM GSH but contains only several μM GSSG. Although GSSG reductase plays a prominent role in maintaining a stable GSH/GSSG ratio, other processes, including de novoGSH synthesis and GSSG export pathways, may generate GSH in the G6PD-deficient cell. After the expansion of the prototype model to include de novoGSH synthesis and GSSG export, the ATP levels were maintained at 80% of normal and the cell was longer lived (Figure [4](#F4){ref-type="fig"}). In addition, the GSH/GSSG ratio was higher (Figure [5](#F5){ref-type="fig"}). This indicates that the de novoGSH synthesis pathway can partially compensate for the lowered GSH concentrations resulting from G6PD deficiency, and that the concentration of GSSG can be kept at a very low level due to the active export system. These observations suggest that these reactions could alleviate the anemia resulting from G6PD deficiency. It is known that people with this deficiency are not normally anemic and display no evidence of the disease until the RBCs are exposed to oxidant stress. The compensatory effect of the de novoGSH synthesis and GSSG export pathways may thus help to explain why many varieties of G6PD deficiency have no evident phenotype. Moreover, it has been proposed that the high frequency of G6PD deficiency may be due to its ability to protect against malaria. Our observations suggest that the compensatory mechanism we have elucidated may have aided this spread of G6PD deficiency, as it counterbalances the worst effects of the deficiency, thus decreasing its severity and promoting the propagation of the disease during evolution. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Simulation of G6PD deficiency using the expanded model. Changes in the concentrations of ATP (A), GO6P (B), GSH (C), GSSG (D), NADP (E) and NADPH (F) during RBC simulation. Broken lines are the results of the prototype model, while solid lines are the results of the expanded model during the same parameter shift as described in Figure 2. The simulation was run for 200,000 seconds (Approx. 55 h) in simulation time. ::: ![](1742-4682-2-18-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### The GSH/GSSG ratio of the prototype and expanded models. The prototype model (A) and the expanded model (B). ::: ![](1742-4682-2-18-5) ::: Determination of a range of metabolic pathways for modeling ----------------------------------------------------------- These results showed that the de novoGSH synthesis pathway and the GSSG export system are essential for accurate simulation of G6PD deficiency in human RBCs. Previous simulations of this deficiency have not included these pathways \[[@B17]\] and the results they generated were similar to those obtained using our prototype model (Figure [2](#F2){ref-type="fig"}). Our prototype model and the previous models developed by others contain only three metabolic pathways, namely, the glycolysis pathway, the pentose phosphate pathway and the nucleotide metabolic pathway. Although these models are sufficient for representing the normal state of the human RBC, they are not adequate for simulating irregular conditions such as deficiencies, because they lack alternative pathways that may normally not be particularly active but can compensate for the deficiency to some extent. Indeed, our results indicate that all the metabolic pathways in the cell will be needed to develop a general purpose model that can be used to simulate any condition. However, dynamic simulation based on kinetic equations requires a large variety of rate equations and kinetic parameters, and unfortunately, such data are rarely available as a complete set. Recently, our laboratory proposed a novel simulation method that reduces the need for this kind of information \[[@B37]\]. This hybrid dynamic/static simulation method combines dynamic rate equations with a flux-based approach and as a result reduces the numbers of rate equations and parameters that are needed by up to 70--80%. It may solve the problems associated with developing a model that simulates all the cellular metabolic pathways. The mathematical steady state may not be the normal state of real cells ----------------------------------------------------------------------- During this simulation analysis, we realized that the longevity of enzymes should be considered in long-term simulation experiments. While in our model the activities of enzymes are decreased by oxidants, enzymes also generally become degraded over time. This natural decrease is not included in our model. As shown in this work, the prototype model was able to achieve a steady state. However, this mathematical steady state, which is when the rates of the production and consumption of all metabolic intermediates become equal, may not exactly represent the condition of the RBCs in the human body. Such a \"mathematical steady state\" never occurs in living organisms, especially in higher multicellular organisms. Rather, homeostasis in multicellular organisms is maintained by replacing the loss of disposable cells with additional cells. It is possible that these disposable cells never reach a mathematical steady state. Thus, a model that can tolerate long-term simulation for analyzing the pathology of human diseases should not approximate the \"mathematical steady state\". Moreover, in the case where the system reaches a steady state with a certain oscillation, it is impossible to obtain a mathematical steady state using an accurate model. It is known, for example, that some key enzymes in glycolysis bind to the Band III protein, an abundant membrane protein in the human RBC \[[@B38]-[@B40]\]. The interaction between glycolytic enzymes and Band III varies depending on the ratio of oxyhemoglobin to deoxyhemoglobin, and it is believed that this interaction is responsible for some oscillations in metabolic pathways in the human RBC. Conclusion ========== We developed a computer model of the human RBC that is based on a previous model but was expanded by introducing a GSH synthesis pathway and a GSSG export system. With this expansion, the model maintained high ATP concentrations in G6PD deficiency. This suggests that these pathways may play an important role in alleviating the consequences of G6PD deficiency. It also indicates that sub-pathways that are normally not particularly highly activated may play important roles in abnormal conditions such as deficiencies. Authors\' contributions ======================= Nakayama contributed mostly to the model development, Kinoshita contributed to the analysis, and Tomita developed the basic ideas and directed the project. Competing interests =================== The author(s) declare that they have no competing interests. Acknowledgements ================ We thank Ryo Matsushima and Kazunari Kaizu for providing technical advice. This work was supported in part by a grant-in-aid from the Ministry of Education, Culture, Sports, Science and Technology (the leading project for biosimulation and the 21st Century Center of Excellence (COE) Program: Understanding and Control of Life\'s Function via Systems Biology), and in part by a grant from New Energy and Industrial Technology Development and Organization (NEDO) of the Ministry of Economy, Trade and Industry of Japan (Development of a Technological Infrastructure for Industrial Bioprocess Project).	PubMed Central	2024-06-05T03:55:57.410910	2005-5-9	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1142344/", "journal": "Theor Biol Med Model. 2005 May 9; 2:18", "authors": [ { "first": "Yoichi", "last": "Nakayama" }, { "first": "Ayako", "last": "Kinoshita" }, { "first": "Masaru", "last": "Tomita" } ] }
PMC1142514	Background ========== The use of microarray technology to assess gene expression levels is now widespread in biology and, particularly in the clinical setting, the applicability of the methodology is likely to broaden as the technology evolves, data analysis procedures improve, and costs decline \[[@B1]-[@B3]\]. Two distinct microarray platforms, cDNA and oligonucleotide, are currently in general use \[[@B4]\]. While the relative merits of the two systems continue to be discussed \[[@B5]\], the validation of microarray results using independent mRNA quantitation techniques, including Northern blotting, ribonuclease protection, in situhybridization, or quantitative real-time reverse transcription-polymerase chain reaction (qRT-PCR) remains a critical element of any microarray experiment \[[@B6],[@B7]\]. Despite this, there have been few systematic validation studies of cDNA, or more noticeably, oligonucleotide microarray data using these independent approaches. For researchers to be confident with the interpretation of microarray results and for the establishment of consistent validation procedures in the microarray community for the purpose of data comparison, it is important that this issue be addressed. We have undertaken an extensive series of experiments examining gene expression profiles in pediatric cancer specimens and normal tissues using oligonucleotide microarrays. For these studies, we used HG-U133A GeneChips (Affymetrix) which contain 22,283 probe sets representing approximately 14,500 human genes. To determine the preferred methodology for the analysis of our microarray data we compared the correlation between microarray expression scores obtained using two different data normalization procedures -- Affymetrix MAS 5.0 \[[@B8]\], and robust multi-array analysis (RMA)\[[@B9]\] -- with the expression levels obtained from follow-up verification experiments using qRT-PCR \[[@B10]-[@B12]\]. We found that the correlation between qRT-PCR and microarray expression data is generally strong. While our results highlight the complementarity of oligonucleotide microarray and qRT-PCR technologies for validation of gene expression measurements, the poor correlations that we observed for 13--16% of genes emphasizes the importance and continuing requirement for caution in interpreting gene expression data. Results ======= We have assessed the degree of correlation between microarray expression scores obtained for 48 genes using HG-U133A GeneChips with expression levels measured for the same genes using qRT-PCR. The genes that we assessed were identified as part of a larger study underway in the laboratory examining differential gene expression in pediatric leukemias and brain tumor specimens. The 48 genes were targeted for validation either on the basis of their differential expression between our subsets of interest (e.g. brain tumour vs normal brain specimens, leukemia specimens vs normal CD34+ stem cells) as determined by microarray analysis, or because they mapped to chromosomal regions of interest. In those cases where there were multiple microarray probe sets for particular genes, only data from those that showed evidence of differential expression were chosen for validation. For genes that were selected from chromosomal regions of interest and not necessarily on the basis of differential expression, correlations were carried out using data from the probe set deemed most specific for the gene of interest by the Affymetrix software (e.g. microarray probe sets designated -atare considered more specific than -s-atand -x-atprobe sets). In total, 889 specimen/gene combinations were assayed by qRT-PCR and microarray in this study. Overall, statistically significant correlations (p \< 0.05) were observed between qRT-PCR and RMA normalized data for 33/48 (69%) genes, and between qRT-PCR and MAS 5.0 normalized data for 32/48 (67%) genes (Tables [1](#T1){ref-type="table"} and [2](#T2){ref-type="table"}, genes in bold). Typical data for a gene with a good correlation is presented in Figure [1](#F1){ref-type="fig"}. The correlation between the qRT-PCR data and microarray data normalized using either of the two methods was not significant (p \> 0.05) for 14/48 (29%) genes (Tables [1](#T1){ref-type="table"} and [2](#T2){ref-type="table"}, genes non-bold). Two genes, FLJ20003and RB, showed significant correlations by RMA but not by MAS 5.0 analysis, while one gene, GCLC, had a significant correlation by MAS 5.0 but not by RMA. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### A comparison of average qRT-PCR, RMA, and MAS 5.0 scores and the corresponding correlation values for the 31 transcript-concordant genes assayed in this study for which the Affymetrix microarray probesets (Affy IDs) were deemed likely to recognize identical transcripts to qRT-PCR probes. Genes are ranked from lowest to highest average log~2~RMA scores. Genes with significant correlations (p \< 0.05) obtained by either normalization procedure are highlighted in bold. The number of specimens tested for each gene is included (n). Expression levels are shown as log~2~\>. ::: GENE EXPRESSION CORRELATION ---------------- --------------- ---------------- ----------------- ------- ------- ---------- ---------- *LCE* 204256\_at 22 4.79 7.01 0.27 0.81 0.70 *ALDH1A1* 212224\_at 22 4.93 6.66 -2.93 0.89 0.88 *CFLAR* 211317\_s\_at 13 5.61 7.92 -0.12 0.65 0.75 *REL* 206036\_s\_at 13 5.63 8.36 0.53 0.76 0.77 *ABCC4* 203196\_at 22 5.84 7.54 0.09 0.78 0.89 *FOXO1A* 202724\_s\_at 19 5.91 7.43 -2.05 0.85 0.90 *NOTCH2* 212377\_s\_at 13 6.19 8.24 -0.43 0.77 0.82 *TNFRSF21* 214581\_x\_at 13 6.22 8.03 1.98 0.83 0.97 *MADH9* 206320\_s\_at 19 6.24 5.47 -2.29 0.87 0.74 *PPM1D* 204566\_at 30 6.29 8.80 0.50 0.73 0.72 *MAP7* 202889\_x\_at 22 6.42 6.27 -3.51 0.85 0.87 DMBT1 208250\_s\_at 19 6.49 7.25 -6.49 0.20 -0.11 *SNIP1* 219409\_at 13 6.57 8.34 1.08 0.69 0.77 *OSF2* 210809\_s\_at 19 6.59 7.68 -1.26 0.80 0.77 *ATBF1* 208033\_s\_at 19 6.64 7.14 0.27 0.81 0.84 *KIT* 205051\_s\_at 22 6.70 7.51 -2.73 0.86 0.87 P53 201746\_at 19 7.00 8.50 -3.44 0.41 0.11 *BAG3* 217911\_s\_at 19 7.04 8.61 -0.98 0.79 0.82 *RB* 203132\_at 19 7.04 9.14 -2.82 0.45 0.38 *WBP4* 203599\_s\_at 19 7.28 8.97 -0.24 0.62 0.74 *BNIP2* 209308\_s\_at 13 7.58 9.79 0.56 0.68 0.69 UMPCMPK 217870\_s\_at 13 8.17 10.98 1.10 0.37 0.12 *DCAMKL1* 205399\_at 19 8.18 9.23 -3.57 0.76 0.89 *OAZIN* 201772\_at 30 8.22 10.36 -0.36 0.72 0.77 *LHFP* 218656\_s\_at 19 8.37 9.27 -0.46 0.89 0.90 *BTG3* 205548\_s\_at 13 8.47 10.54 0.83 0.86 0.90 *DCX* 204850\_s\_at 19 8.81 10.08 0.62 0.87 0.88 TERF2 203611\_at 19 9.05 10.04 -0.14 0.32 0.31 *GADD45A* 203725\_at 19 9.17 9.80 -0.12 0.96 0.94 *PRSS11* 201185\_at 19 9.22 9.85 -3.54 0.63 0.64 *RAP1* 201174\_s\_at 19 10.34 11.59 -0.82 0.83 0.84 ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### A comparison of average qRT-PCR, RMA, and MAS 5.0 scores and the corresponding correlation values for the 17 genes assayed in this study for which the Affymetrix microarray probesets (Affy IDs) may not recognize the exact same transcript subsets recognized by qRT-PCR probes. Genes are ranked from lowest to highest average log~2~RMA scores. Genes with significant correlations (p \< 0.05) obtained by either normalization procedure are highlighted in bold. The number of specimens tested for each gene is included (n). Expression levels are shown as log~2~. ::: GENE EXPRESSION CORRELATION ---------------- --------------- ---------------- ----------------- ------- ------- ---------- ---------- CDC14A 210742\_at 13 5.77 7.64 -0.67 0.31 0.26 P125 209175\_at 19 6.61 8.43 0.54 0.11 -0.11 *GCLC* 202922\_at 13 6.65 9.20 0.33 0.46 0.56 MAP3K7 206853\_s\_at 13 6.65 8.72 1.19 0.11 -0.10 TIAL1 202405\_at 19 6.68 7.86 0.30 0.32 0.17 *FLJ20003* 219067\_s\_at 19 6.71 8.54 0.64 0.64 0.34 RUNX1 210365\_at 13 6.95 9.24 1.47 0.29 0.28 PLEKHA1 219024\_at 19 6.99 8.18 -2.88 -0.40 -0.28 FLJ12661 218420\_s\_at 19 7.35 8.50 0.57 -0.08 -0.17 *RGC32* 218723\_s\_at 19 7.36 8.11 -3.20 0.85 0.96 WDR11 218090\_s\_at 19 7.96 9.04 0.60 0.12 0.01 *RFC3* 204127\_at 19 8.10 9.77 1.20 0.62 0.64 ASAH1 213702\_x\_at 22 8.30 10.29 1.46 0.29 0.27 *P38IP* 220408\_x\_at 19 8.35 9.55 0.76 0.73 0.65 *BUB3* 201456\_s\_at 19 8.41 9.70 0.60 0.64 0.61 SAC2 203607\_at 19 8.86 10.21 0.35 0.22 0.12 *TSC22* 215111\_s\_at 19 10.56 11.87 -1.34 0.83 0.82 ::: ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Examples of Pearson\'s correlations between gene expression levels determined by qRT-PCR and oligonucleotide microarray for one gene assessed in this study. The mRNA levels for the gene GADD45Awere determined by qRT-PCR and correlated with microarray expression scores determined after data processing using MAS 5.0 software (A) or RMA (B). All data are shown as log~2~. ::: ![](1471-2164-6-59-1) ::: By careful analysis of the relevant databases (see Methods) we identified a subset of 31 genes for which the microarray probe-sets were deemed to recognize the exact same transcript or subset of transcripts as the qRT-PCR probes (Table [1](#T1){ref-type="table"}). When we assessed the levels of correlation for this group of 31 transcript-concordant genes a higher proportion of significantly correlating scores was observed; 84% (26/31) for MAS 5.0 normalized data and 87% (27/31) for RMA normalized data (Table [1](#T1){ref-type="table"}, genes in bold). In addition, the average correlations between the MAS 5.0 or RMA data and the qRT-PCR data for this subset of genes were very similar (0.71 and 0.72, respectively). In contrast, for the remaining 17 genes for which the Affymetrix microarray probe-sets may not recognize the same subset of transcript(s) recognized by qRT-PCR probes, significant correlations were observed for only 41% (7/17) genes by either MAS 5.0 and RMA (Table [2](#T2){ref-type="table"}). All genes with poor correlations were tested on the same numbers of samples as those genes that did correlate, and there was no relationship between sample type and whether or not correlation was significant. Separate genes were targeted for each sample type. Using a two sample t-test, the average correlations between RMA-qRT-PCR scores and MAS-qRT-PCR scores for the transcript concordant genes in Table [1](#T1){ref-type="table"} were significantly higher than the average of the equivalent correlations for the non-concordant genes in Table [2](#T2){ref-type="table"} (RMA-qRT-PCR Table [1](#T1){ref-type="table"} vs 2, p = 0.0005; MAS-qRT-PCR Table [1](#T1){ref-type="table"} vs 2, p = 0.0003). Determining fold-changes in gene expression levels between subsets of interest is often a major aim of microarray studies. To address this issue, we analyzed fold-change in average gene expression levels between our subsets of interest (e. g. tumor vs normal) by both qRT-PCR and RMA or MAS 5.0 microarray scores for the same genes. Only the 31 transcript-concordant genes were considered in this analysis (Table [1](#T1){ref-type="table"}). From a total of 587 specimen/gene combinations we found a significant and strong correlation in mean fold-change using both RMA (r = 0.89, p \< 0.05) and MAS 5.0 (r = 0.92, p \< 0.05) (Figure [2a, b](#F2){ref-type="fig"}). Interestingly, we noticed a trend towards poorer correlation for genes that exhibited fold-change differences of \<1.5 between subsets of interest based on microarray expression scores compared to those with fold-change differences of \>1.5 (data not shown). The slopes of the two regression lines in Fig. [2](#F2){ref-type="fig"} are significantly greater than one \[RMA vs qRT-PCR = 1.49 (95%CI = 1.20, 1.77); MAS vs qRT-PCR = 1.23 (95% CI = 1.03, 1.42)\]. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Pearson\'s correlations between fold-change in average gene expression levels between subsets of interest assessed by qRT-PCR and either MAS 5.0 software (A) or RMA (B) for the 31 transcript-concordant genes (see Table 1). All data are shown as log~2~. ::: ![](1471-2164-6-59-2) ::: Discussion ========== Microarray expression analysis has revolutionized many facets of biology and will continue to be applied widely. However, significant questions remain with regard to the generation, analysis, and in particular, interpretation of microarray data. Although the validation of microarray expression results obtained for specific genes using independent techniques is still considered a desirable component of any microarray experiment, the genes selected for validation a priori, are usually identified from the microarray data. The selection is based on the implicit assumption that there is a good correlation between the microarray data and actual mRNA levels in the cells under investigation. One fundamental issue that has not been adequately addressed is how well microarray expression scores reflect actual mRNA levels in the sample being examined. To facilitate data comparison between research groups it is important that the microarray community moves to adopt consistent validation methodologies. This is especially important if microarray technology is to play a role in the clinical setting \[[@B13]\]. However, the choice of validation methodology remains a contentious issue \[[@B14]\]. To date, qRT-PCR is the method of validation that has been used in the majority of published microarray studies, presumably because it is a rapid, sensitive, high throughput procedure that requires minimal amounts of test material compared to techniques such as Northern blotting or ribonuclease protection assays. As is the case for many studies, including ours, qRT-PCR is often the only feasible approach when rare or unique tissues are investigated. For these reasons, it would appear likely that qRT-PCR will continue to be used extensively for the validation of microarray expression data \[[@B15]\]. To our knowledge, this study is the most extensive and practical examination of mammalian cells that focuses on the degree of correlation between expression level measurements obtained by oligonucleotide microarray analysis and qRT-PCR. We observed strong correlations (p \< 0.05) for the majority (\>87%) of the 31 transcript-concordant genes that we examined in this study. In addition, although the MAS 5.0 software and RMA use different algorithms for the normalization of microarray data \[[@B8],[@B9]\] we found that the degree of correlation between microarray and qRT-PCR results was very similar irrespective of the normalization procedure employed. Our data clearly demonstrate that similar microarray scores for different genes do not necessarily mean that similar qRT-PCR scores will be obtained. For example, ATBF1, OSF2, and SNIP1yielded similar average log~2~RMA scores (\~6.6) but the average log~2~qRT-PCR scores for the same genes were substantially different (0.27, -1.26, and 1.08, respectively). Similarly, KITand ABCC4exhibited identical average log~2~MAS 5.0 scores (\~7.5), while the corresponding average log~2~qRT-PCR scores were -2.73 and 0.09, respectively. The finding that genes with similar microarray expression scores were unlikely to have similar qRT-PCR results presumably reflects the different hybridization kinetics of the probe sets for each gene. This observation has the major implication that on the basis of the qRT-PCR data that we obtained, it is generally not feasible to predict the true expression level of one gene based on the microarray expression score of another. In addition, we observed significant correlations for many genes with microarray expression scores, at least by RMA, of less than 100 (\~log~2~100 = 6.64), which is at the lower end of the range of microarray scores we obtained in this study (range 6--23000). This finding indicates that the exclusion of genes with low microarray expression scores (e.g. \<100) from further analysis, as has been adopted by some research groups in early microarray studies, may not be justified. Determining fold-changes in gene expression levels between subsets of interest is often a critical aim of microarray studies. We found a significant and strong correlation using RMA (r = 0.89, p \< 0.05) and MAS 5.0 (r = 0.92, p \< 0.05). These data indicate that the direction of change of gene expression levels (i.e. either up or down regulation) between subsets of interest is accurately predicted by comparison of average microarray expression scores. Again, the fold-change correlations we observed were very similar irrespective of the normalization procedure we employed. Consistent with the results of Yuen et al (2001)\[[@B16]\], fold change results determined by qRT-PCR were significantly greater than fold change assessed for the same genes by microarray analysis. A recent study addressing gene expression profiles in Arabidopsisreported a good correlation between oligonucleotide microarray and SYBR green qRT-PCR data when ratios of gene expression in shoot tissue versus root tissue were compared for highly expressed genes. However, the correlations between shoot versus root ratios were generally poor for genes expressed at low levels \[[@B17]\]. We observed a similar trend towards poorer correlation for genes that exhibited fold-change differences of \<1.5 between subsets of interest based on microarray expression scores compared to those with fold-change differences of \>1.5. It is likely that this trend relates to the fact that small variations in mRNA levels (\<2-fold) can be accurately detected by qRT-PCR, while the smaller dynamic range of microarrays means that the same changes may not be accurately reflected by microarray expression scores, especially for genes expressed at low levels (\<1.5 pM or approximately 3.5 copies/cell) \[[@B18],[@B19]\]. This latter point is a likely explanation for the poor correlation observed for one gene, DMBT1, which is expressed at very low levels according to our qRT-PCR data. Etienne et al., 2004 \[[@B20]\] observed a lower overall correlation between microarrray and semi-quantitative RT-PCR data compared to our study. These authors hypothesized that in addition to genes with low expression levels, those with very high expression levels or a greater percentage of absent calls, may show lower levels of correlation between Affymetrix expression scores and semi-quantitative RT-PCR data. We considered these issues in relation to the other poorly correlating genes in our study and found that none were expressed at levels that approach the fluorescence ceiling for the Affymetrix scanner (\~50000). In addition, the absolute number or percentage of absent calls did not correlate significantly (p \> 0.05) with the level of correlation between qRT-PCR results and microarray data (data not shown). It is possible that the differences between our results and those of Etienne and co-workers are related to the particular semi-quantitative RT-PCR methodology employed by these researchers, which may not be as sensitive as qRT-PCR, and as the authors point out, may not detect certain low level transcripts. In addition to DMBT1mentioned above, we identified 13 other poorly correlating genes from the 48 genes we assessed. Careful analysis of the alternative transcript data available through the LocusLink database <http://www.ncbi.nih.nlm/LocusLink> indicated that for 10 of these 13 genes, different subsets of alternative transcripts may be recognized by microarray probe sets and qRT-PCR probes. Hence, this may be the explanation for the poor correlations observed for these genes. Possible explanations for the poor correlations that were observed for the three remaining genes (p53, UMPCMPK, and TERF2), all of which were transcript-concordant, include the existence of alternative cross-hybridising transcripts differentially recognized by the oligonucleotide probe sets and qRT-PCR probes, gene specific variation related to the different hybridization kinetics associated with the two technologies, and misleading results associated with errors in GenBank sequence data and/or probe set annotations \[[@B21]\]. Additional experimental data will be required to address these possibilities. It is important to note that in our hands the reproducibility of both the qRT-PCR and oligonucleotide microarray methods is very high \[[@B22],[@B23]\]. Hence, it is unlikely that poor correlations observed in our study are associated with issues of experimental precision. Interestingly, the microarray and qRT-PCR expression data correlated well for five genes for which the microarray probe sets were deemed unlikely to recognize the same transcripts as the qRT-PCR probes. These data suggest that despite the possibility of differential transcript recognition, identical transcripts were being detected by both assays in the particular tissues involved. Conclusion ========== Our data indicate that correlations between qRT-PCR and microarray data are generally strong; a result that is particularly encouraging for those researchers with access to only very limited amounts of rare or unique test specimens. Our data also emphasize the importance of ensuring that qRT-PCR probes recognize the same transcript(s) as the microarray probe set. Finally, the 13--16% non-concordance that we observed indicates that independent validation of expression data continues to be an important consideration. Methods ======= Specimens --------- Informed consent for the use of tissues for research purposes was obtained for all individuals involved in this study according to hospital and Australian National Health and Medical Research (NHMRC) guidelines. We extracted total RNA from 64 specimens, including 13 primary pediatric brain tumors, six pediatric brain tumor cell lines, two normal adult brain cortices, and one fetal brain germinal matrix. We also obtained total RNA from fetal brain pooled from multiple individuals (Clontech). In addition, total RNA was extracted from 36 pediatric acute lymphoblastic leukemia bone marrow specimens and from CD34^+^hematopoietic stem cells isolated from the bone marrows of 5 normal individuals. Ficoll-hypaque purified leukemia cells or cryopreserved bone marrow specimens were snap frozen and stored in liquid nitrogen until required. Total RNA was extracted from \~1 × 10^6^-- 2 × 10^7^live cells. Primary brain tumour specimens (10 -- 150 mg) were either wrapped in foil or placed in RNAlater (Ambion) immediately after resection and stored at -80°C. Brain tumour cell lines were processed directly from tissue culture. RNA extraction, preparation of target cRNA and hybridization to HG-U133A GeneChips ---------------------------------------------------------------------------------- Total RNA was extracted from all specimens using a combination of TRIZOL reagent (Invitrogen), RNeasy Mini kit (Qiagen) and ethanol precipitation. Following the TRIZOL reagent procedure, 0.53 volumes of 100% ethanol were added drop-wise to the aqueous phase and the mixture applied to RNeasy mini columns according to the manufacturer\'s instructions. Further purification and concentration was achieved through an additional ethanol precipitation. The integrity of the RNA preparation was assessed using agarose gel electrophoresis and analysis on an Agilent 2100 Bioanalyzer (Agilent Technologies). Biotinylated cRNAs for hybridization were prepared from total RNA according to Affymetrix protocols. Agarose gel electrophoresis was used to confirm the integrity of labelled cRNA and to assess its fragmentation products. Biotinylated cRNA preparations (15 μg) were hybridized to HG-U133A arrays, which were subsequently washed, stained, and scanned using a GeneArray Scanner (Agilent Technologies) according to the Affymetrix protocol. Processing and statistical analysis of microarray data ------------------------------------------------------ Array images were reduced to intensity values for each probe (celfiles) using Affymetrix MAS 5.0 software and only those microarrays meeting acceptable Affymetrix quality control criteria were considered for further analysis. Celfiles were then processed using either the MAS 5.0 software \[[@B8]\] or RMA (Bioconductor release 1.2) \[[@B9]\], an alternative algorithm that is publicly available at <http://www.bioconductor.org>. The MAS 5.0 algorithm uses a scalar normalization technique taking into account perfect match (PM) and mismatch (MM) probe pairs to correct for non-specific hybridization, while RMA is based on a quantile normalization approach which ignores MM values. All microarrays processed using the MAS 5.0 software were scaled to a standard target intensity of 500. For comparison purposes, all microarray and qRT-PCR data are presented as log~2~and absent/present calls generated by the MAS 5.0 software were not taken into account. Pearson\'s correlations were used for the comparison of qRT-PCR and microarray data and p-values were obtained using Fisher\'s z-transformation. Correlations were considered significant at p \< 0.05. Bioinformatics -------------- To determine whether transcripts recognized by microarray probe sets \[[@B24]\] were likely to be identical to those detected by qRT-PCR probes, alternative splicing patterns for each gene were thoroughly reviewed using LocusLink <http://www.ncbi.nlm.nih.gov> and Ensembl <http://www.ensembl.org>. Any full-length human mRNA or cDNA sequences demonstrating alternative splicing, in addition to NCBI-reviewed Reference Sequences (RefSeq), were considered as potential isoforms for each gene. Using BLAST alignments <http://www.ncbi.nlm.nih.gov> of probe and cDNA sequences, the members of each isoform \'family\' that could be targeted by either qRT-PCR or microarray were identified (typically multiple isoforms for each gene). The potential number of isoforms recognised by each technology were then compared. Probes which targeted exactly the same isoform subsets for each gene were considered \'transcript-concordant\' and placed in Table [1](#T1){ref-type="table"}; those for which at least one of the targeted isoforms differed (regardless of the number of matching isoforms) were considered \'non transcript-concordant\' and placed into Table [2](#T2){ref-type="table"}. qRT-PCR ------- All qRT-PCR assays were carried out using primer and probe sets from Applied Biosystems (ABI Assays on Demand, <http://www.appliedbiosystems.com/>). Each assay was designed using ABI\'s primer/probe selection algorithm and bionformatics pipeline which includes access to both public and Celera DNA sequence databases. The combination of gene specific primers and a gene specific probe ensures a high degree of specificity. Aliquots of total RNA extracted for microarray analysis as described above were used for qRT-PCR experiments according to the manufacturer\'s protocols (ABI). All ABI Assays on Demand are designed to generate amplicons of 50--150 bp and are carried out using identical cycling conditions. 1--2 ug total RNA (quantitated by spectrophotometer at OD~260~) was used for each RT reaction. Three RT reactions were pooled and all qRT-PCR reactions were carried out using aliquots from the pool. We did not detect DNA contamination in any of our total RNA preparations after qualitative assessment using an Agilent Bioanalyzer. All qRT-PCR assays for a particular gene were undertaken at the same time under identical conditions and carried out in duplicate. All qRT-PCR experiments were run on an ABI 7700 sequence detector. For all qRT-PCR assays the expression levels of target genes were normalised to the levels of the ACTBhousekeeping gene utilising a standard curve method for quantitation as described previously \[[@B25]\]. Serial dilutions of cDNAs generated from selected cell lines that expressed target genes at a suitable level were used to generate a standard curve for each target gene and ACTB. The standard curves were then used to determine expression values (expressed as ng cDNA template) for each target gene after qRT-PCR analysis of each test specimen. Relative expression values for each target gene were expressed as a ratio of target gene expression level to ACTBexpression level in the same specimen. These ratios were then correlated with the microarray data. Authors\' contributions ======================= PBD and NGG contributed equally to this work and were responsible for designing the study, analysing, collating, and interpreting the data, and preparing the manuscript. MJF carried out the statistical analysis, AHB and KF assisted with data analysis, experimental design, and data interpretation. PAT, JRF, JMB, AJC and NGG carried out the microarray and qRT-PCR experiments. URK supervised all aspects of the study and preparation of the manuscript. Acknowledgements ================ This study was supported by funds from NHMRC project grants 254595 and 254596, NCI/NIH grant 95475, the Three Boys Legacy, and Variety Club of Western Australia. We would like to thank Nigel Swanson and Violet Peeva at the Lotterywest State MicroArray Facility, Perth, Western Australia. Thanks also to Reinete Orr for secretarial assistance. NGG was supported by a National Childhood Cancer Foundation Laura and Greg Norman Fellowship.	PubMed Central	2024-06-05T03:55:57.414570	2005-4-27	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1142514/", "journal": "BMC Genomics. 2005 Apr 27; 6:59", "authors": [ { "first": "Peter B", "last": "Dallas" }, { "first": "Nicholas G", "last": "Gottardo" }, { "first": "Martin J", "last": "Firth" }, { "first": "Alex H", "last": "Beesley" }, { "first": "Katrin", "last": "Hoffmann" }, { "first": "Philippa A", "last": "Terry" }, { "first": "Joseph R", "last": "Freitas" }, { "first": "Joanne M", "last": "Boag" }, { "first": "Aaron J", "last": "Cummings" }, { "first": "Ursula R", "last": "Kees" } ] }
PMC1142515	Background ========== Age related macular degeneration (ARMD) is the commonest cause of severe loss of central vision in people aged over 50 in the Western world \[[@B1]\] and accounts for almost 50% of those registered as blind or partially sighted in the UK \[[@B2]\]. One form of the disease, neovascular ARMD, is more aggressive and accounts for up to 15% of cases \[[@B1]\]. Evidence from two randomised trials indicates that photodynamic therapy (PDT), administered five to six times over a two year period, reduces the relative risk of losing three or more lines of visual acuity over two years \[[@B3],[@B4]\]. In particular, a subgroup analysis has shown a statistically significant benefit in the prevention of visual loss in people with wholly or predominantly classic choroidal neovascularisation (CNV). These findings contributed to pressure upon the UK National Health Service (NHS) from both patient and professional groups to make this therapy routinely available \[[@B5],[@B6]\]. However, systematic reviews have highlighted reservations with respect to the reliance placed on the subgroup analysis \[[@B7]-[@B9]\]. Concern about the cost-effectiveness of PDT \[[@B10]\] prompted the National Institute for Clinical Excellence (NICE) to undertake a technology appraisal of PDT. Yet, during the protracted course of the technology appraisal and subsequent appeals, there was evidence that ophthalmology units were establishing services or referring patients for PDT. When it was released in September 2003, the NICE guidance specified that only the subgroup of patients with 100% classic CNV should be eligible for PDT. Ideally, strategies to promote effective practice should be tailored according to identified local needs and barriers \[[@B11]\]. The rate of uptake of the NICE guidance may be influenced by several factors, such as ophthalmologists\' perceptions of treatment benefit and time required to establish new services. Groups such as local colleagues or professional bodies may themselves influence clinicians\' beliefs about treatment benefit. Hence the promotion of the NICE guidance may be more effective if it takes account of such professional norms. We have been conducting a series of \'tracker\' surveys to follow trends in the uptake of PDT. Our fourth survey, coincidently conducted as the final NICE guidance was being released \[[@B12]\], addressed whether provision of PDT had continued to expand despite the absence of the guidance and whether perceived professional norms represent a barrier to its implementation. Methods ======= Questionnaire design and administration were similar to those used in the previous three surveys \[[@B13]-[@B15]\]. We sought information about local treatment and referral policies for neovascular ARMD and the threshold of clinical benefit considered sufficient to justify the use of PDT. Benefit was rated on the basis of the number of patients that the respondent would be willing to treat to prevent the loss of three lines of visual acuity for two years for one patient at a given fixed cost of treatment of £8000 per patient (known as the number needed to treat (NNT)). This cost was estimated from the TAP study treatment protocol \[[@B1]\]. Six options for the NNTs were presented, four based upon the point estimate of effect and upper and lower limits of the 95% confidence intervals reported in the Cochrane Review \[[@B7]\], and two further categories of 1 in 50 and 1 in 100 for comparative purposes. We used two constructs from the Theory of Planned Behaviour \[[@B16]\] to measure attitudes and subjective norms (which we shall now refer to as perceived social pressure) concerning the use of PDT. The questionnaire incorporated previously recommended scales and items to measure attitudes and perceived social pressure for a specific action: the treatment of patients with predominantly (more than 50%) classic sub-foveal CNV using photodynamic therapy \[[@B17]\]. Attitudes towards this use of PDT were measured using 1--7 Likert scales for four items (beneficial/harmful to patients; an excellent/poor use of resources; clinically/not clinically effective; good/bad practice). Ophthalmologists might feel under pressure from different groups. Four perceived social pressure items asked how much each of the following would approve or disapprove of the use of PDT: local colleagues, the Royal College of Ophthalmologists, NICE, and other ophthalmologists. Local colleagues might encompass both (mainly) ophthalmologists and others; whilst \'other ophthalmologists\' would refer to any others encountered via formal and informal networks. The Royal College of Ophthalmologists and NICE refer to (perceived) official policy from each organisation. Following pre-testing of the additional theory-based items, the questionnaire was posted to all clinical directors or lead consultants in NHS ophthalmology units within the UK in the autumn of 2003. The units were identified using the Royal College of Ophthalmologists\' database. One reminder was sent to non-respondents. Changes in service provision and beliefs were analysed using the χ^2^-test, and the normal z-test to compare two proportions. P values are presented. Associations between attitudes (the combined means of the four items) and each of the four perceived social pressure items were measured using multiple regression analysis. Differences between levels of perceived approval from the four potential sources of social pressure were tested using the repeated measures ANOVA. Results ======= Out of 150 questionnaires, 111 (73%) were returned completed, a slightly lower response rate than those for the previous three surveys (82%, 79% and 80% respectively). No variations in response rate were detected by health region either overall or in the individual surveys. The proportion of units reporting routine provision of PDT for patients with more than 50% classic sub-foveal CNV significantly increased over four years from 8.5% to 41% (p \< 0.001), making this the most common reported policy for the first time. The proportion of units referring or treating no patients fell from 35% to 5% (Figure [1](#F1){ref-type="fig"}, p \< 0.001). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Trends in the reported provision of photodynamic therapy over 2000--2003. ::: ![](1472-6963-5-34-1) ::: There was a significant change in beliefs of what constituted a worthwhile clinical benefit over 2000--1 (p = 0.01), but no significant change was detected over 2001--2 (p = 0.97), or 2002--3 (p = 0.13) (Table [1](#T1){ref-type="table"}) The proportion of respondents requiring further evidence before supporting the use of PDT fell from 33% to 11% over the four years. Lower thresholds supporting the use of PDT were associated with greater reported provision over the four surveys (χ^2^= 20.4, df = 8, p = 0.01; χ^2^= 25.2, df = 8, p = 0.003; χ^2^= 21.2, df = 8, p = 0.007, χ^2^= 17.0, df = 8, p = 0.03 respectively). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Beliefs about clinical benefit and evidence of effectiveness over 2000--2003. ::: Threshold of clinical benefit that would make offering PDT as a treatment worthwhile Survey year ------------------------------------------------------------------------------------------ ------------------------------------------ ---------- ---------- ---------- ---------- To prevent the loss of 3 lines of visual acuity over 2 years At least In 1 person for every 7 treated 22 (19%) 31 (27%) 30 (28%) 36 (37%) In 1 person for every 4 treated 35 (30%) 37 (32%) 38 (35%) 34 (35%) In 1 person for every 2 treated 21 (18%) 27 (23%) 20 (19%) 16 (17%) Further evidence of effectiveness required 39 (33%) 20 (19%) 20 (17%) 11 (11%) Total 117 108 115 97 ::: The mean score for attitude was 4.86 (sd = 1.07), based on the 1--7 scale where higher scores represented more positive attitudes towards PDT for the treatment of patients with predominantly (more than 50%) classic sub-foveal CNV. Internal consistency was acceptable for the four attitude items (Cronbach\'s α = 0.69). The mean scores for perceived social pressure ranged from 2.86 for local colleagues to 3.82 for NICE, where lower scores represent greater levels of perceived approval (Table [2](#T2){ref-type="table"}). A repeated measures ANOVA showed that perceived approval differed significantly as a function of the source of perceived social pressure (F(3,91) = 15.58, p \< 0.001). NICE was seen as most disapproving of PDT use for the given indication; this was significantly different from scores for the other three sources of pressure (p \< 0.001). ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Mean perceived social pressure for the treatment of patients with predominantly (more than 50%) classic sub-foveal CNV using photodynamic therapy where possible scores range from 1 (strongly approve) to 7 (strongly disapprove). ::: Source of perceived social pressure Mean Standard deviation --------------------------------------------------- ---------- ------------------------ Local colleagues 2.86 1.34 Royal College of Ophthalmologists 3.01 1.23 National Institute for Clinical Excellence (NICE) 3.82 1.49 Other ophthalmologists 3.00 1.41 ::: In the multiple regression analysis, scores for the four sources of social pressure significantly predicted attitude scores, R = 0.51, F(4,93) = 7.67, p \< 0.001 (Table [3](#T3){ref-type="table"}). Inspection of individual standardised regression coefficients (beta weights) showed that only perceived social pressure from local colleagues was a significant predictor of attitudes (p \< 0.01). ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Multiple regression of scores for perceived social pressure on attitude scores. (Beta weights are negative because perceived social pressure and attitude scores were scored in opposite directions.) ::: Dependent variable Independent variables β R R^2^ Adjusted R^2^ ------------------------------------ --------------------------------------------- ----------- ------------ ------ --------------- Attitude concerning the use of PDT Social pressure from local colleagues -0.38\\ Social pressure from the RCO -0.03 Social pressure from NICE -0.05 Social pressure from other ophthalmologists -0.14 0.51\\\* 0.26 0.22 ::: Discussion ========== By the time that the NICE guidance was released, almost half of the surveyed ophthalmology units were already routinely providing PDT. The proportion of respondents who believed that further proof of effectiveness was required has also declined despite the absence of any new substantial evidence. Attitudes towards providing PDT were positive, on average, and were more strongly associated with perceived social pressure from local colleagues than from other sources. Local colleagues were seen as being most approving of PDT. There were several limitations to our methods. First, survey respondents may have held stronger views, or otherwise, about the merits of PDT compared with non-respondents. Second, additional variation in our findings over time may have been introduced by differences in respondents over each survey, although a sub-analysis provision of PDT for the units that responded to all four surveys (n= 76 (51%)) indicated similar findings as for the overall survey. Third, reported policies may differ from those used in practice. Fourth, the issue of NICE guidance around the time of the last survey may have altered responses. This seems unlikely to be a major factor, given the time taken to establish new services or referral policies. We also found no significant difference in attitudes towards PDT between those respondents who said they had read the NICE guidance and those who had not. Fifth, it would be wrong to assume the direction of causality was such that perceived social pressure influenced attitudes; more positive attitudes towards PDT might have preceded perceptions about which source approved or disapproved of PDT use. Finally, we enquired only about a limited number of sources of social pressure. Other sources might have included patients or the commercial sector, although social desirability bias might have led to an underestimation of effects of the latter. In a classic case of \'technology creep\' \[[@B18]\], PDT has become an established treatment in the NHS prior to national guidance and in the absence of new supporting evidence. This raises the question as to whether the routine application of new health technologies should be subject to a strict national moratorium. Proponents of PDT would argue that such a moratorium would stifle local service innovation, especially in light of the time taken to develop and issue final national guidance. However, we have previously suggested that the long time scale required to issue more controversial technology assessments may work in favour of the advocates of new technologies by forcing the hand of policy-makers \[[@B15]\]. There is also a risk of \'indication creep\', whereby the use of an intervention expands beyond its recommended indication and results in less cost-effective use of health care resources. The NICE guidance specifies that only patients with 100% -- as opposed to at least 50% -- classic CNV should be eligible for PDT. The uptake of this guidance will partly depend upon the perceived credibility of the messenger. Our findings suggest that ophthalmologists are currently more likely to be influenced by local colleagues than by NICE, thereby posing a potential threat to consistent adherence to the guidance. All of this means that those responsible for implementing the NICE guidance need to address ophthalmologists\' beliefs about the evidence of effectiveness for PDT anddraw upon supportive local individuals or networks to enhance the credibility of the guidance. The implementation of national guidance requires monitoring. We plan to conduct one further survey to audit the reported uptake of the NICE guidance and measure the extent of any \'indication-creep\'. Abreviations ============ PDT -- Photodynamic Therapy ARMD -- Age related Macular degeneration NICE -- National Institute for Clinical Excellence CNV -- choroidal neovascularisation NHS -- UK National Health Service Authors\' Contributions ======================= RF participated in study design, interpreted the findings and wrote the first draft. BF conceived the idea for the study and conducted the analysis of trends. JF conducted the analysis and helped with the interpretation for the theory-based questions. All authors participated in the design of the study, were involved in interpretation of the findings, drafted or critically revised the article, and gave final approval of the version to be published. Competing interests =================== The author(s) declare that they have no competing interests. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1472-6963/5/34/prepub> Acknowledgements ================ We are grateful to all those who participated in the survey.	PubMed Central	2024-06-05T03:55:57.418918	2005-5-10	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1142515/", "journal": "BMC Health Serv Res. 2005 May 10; 5:34", "authors": [ { "first": "Robbie C", "last": "Foy" }, { "first": "Barny", "last": "Foot" }, { "first": "Jill", "last": "Francis" }, { "first": "Usha", "last": "Chakravarthy" }, { "first": "Richard PL", "last": "Wormald" } ] }
PMC1142516	Background ========== With a frequency of up to 1%, congenital heart disease represents one of the most common major congenital anomalies \[[@B1]-[@B3]\]. Transposition of the great arteries (TGA) accounts for 5% of all congenital heart defects \[[@B4]\]. TGA manifests during the early fifth week of development affecting the septation of the common outflow tract into aorta and pulmonary arteries, and has been suggested to represent a laterality defect of the heart \[[@B5]\]. The more common dTGA (dextro-looped TGA) represents a complete inversion of the great vessels (atrioventricular concordance and ventriculoarterial discordance). In the less common lTGA (laevo-looped TGA), both atrioventricular and ventriculoarterial discordance is present. Despite the high prevalence and clinical importance of TGA, we are just beginning to unravel the etiology of this heterogeneous disease. Up to now, three genes have been suggested to be involved in the etiology of dTGA in humans: PROSIT240, a novel TRAP240-like gene, has been recently isolated and several mutations are suggested to be responsible for a subset of TGA patients \[[@B6]\]. Isolated mutations in ZIC3\[[@B7]\] and CFC1(human CRYPTIC gene) \[[@B8],[@B9]\] have also been detected in patients with TGA. ZIC3and CFC1have been shown before to be involved in laterality defects in humans \[[@B8],[@B10]\]. However the total number of mutations detected so far within these three genes is not sufficient to explain the high incidence of dTGA and point towards strong heterogeneity. As cardiac neural crest cells contribute to the formation of the outflow septum that divides the common outflow tract, an association between neural crest disturbance and TGA has been suggested. Extirpation experiments in chick could show that neural crest cells contribute to normal aorticopulmonary septation. Deletion of those cells causes malformation of the aorticopulmonary septum resulting in common arterial outflow channels or transposition of the great arteries \[[@B11],[@B12]\]. Pitx2, a bicoid-related homeodomain transcription factor involved in eye, heart and craniofacial development and establishment of left-right asymmetry, is expressed in several tissues of the developing mouse embryo including neural crest derived organs \[[@B13]\]. In humans, PITX2haploinsufficiency causes Axenfeld-Rieger Syndrome (ARS), an autosomal dominant disorder involving ocular, dental and umbilical defects \[[@B14]\] and, in some patients with unknown mutations, also cardiac defects \[[@B15],[@B16]\]. Most interestingly, Pitx2 loss of function experiments in mice cause severe cardiovascular defects including transposition of the great arteries \[[@B17]-[@B20]\]. Kioussi et al. reported that Pitx2^-/-^mice, that survive up to E15, invariantly exhibit major cardiac outflow tract abnormalities, amongst which 30% show incomplete septation of the great arteries, that may develop with double outlet right ventricle (DORV) or transposition of the great arteries \[[@B20]\]. Deletion of the Dvl2gene \[[@B21]\], which is regulated by the same pathway as Pitx2, leads to the same severe outflow tract malformations, indicating a strong implication of this pathway in the outflow tract phenotype. These lines of evidence prompted us to investigate whether PITX2mutations in humans can also contribute to the etiology of TGA. Methods ======= Human subjects and genomic DNA ------------------------------ Peripheral-blood samples were taken from healthy individuals and patients with simple dTGA after informed consent had been obtained, after approval by the institutional review board of ethics of the Medical Department of the University of Heidelberg and the Newcastle and North Tyneside Health Authority Joint Ethics Committee. Genomic DNA was prepared using the Puregene DNA Isolation Kit (Gentra, Inc., USA). PCR and mutation screening -------------------------- Amplifications were performed using the High Fidelity System (Roche) according to the manufacturer\'s protocol. Primers were designed according to the PITX2sequence gene bank accession number AF238048 and respective sequences are given in table [1](#T1){ref-type="table"}. Mutation screening was performed using denaturing high performance liquid chromatography (DHPLC). A WAVE DNA-Fragment Analysis System (Transgenomic Inc., Cheshire) was used. Sequencing ---------- Sequencing was performed on a MegaBACE sequencer (Amersham Bioscience, Piscataway) using the DYEnamic™ ET terminator Cycle Sequencing Kit following the manufacturer\'s protocol. Sequencing reactions were performed on both DNA strands. Sequences were analyzed using the Clustal program (German Cancer Research Center, Biocomputing Facility HUSAR, Heidelberg). Results and Discussion ====================== 96 patients with dTGA were analyzed for mutations in PITX2by DHPLC and direct sequencing. All coding exons of PITX2(exon 2 to 6, including both alternatively spliced exons 4a and b) were amplified by intron-specific exon-flanking primers to screen exon-intron junctions (table [1](#T1){ref-type="table"}, figure [1](#F1){ref-type="fig"}). Non-coding regions (exon 1 and the 3\'part of exon 6), intronic regions beyond the intronic sequences covered by the amplification, and promoter elements were not examined. We identified seven intronic and UTR variants, two silent mutations and two polymorphisms within the coding region. Most of these variants were found also in similar frequencies in 100 control individuals and are therefore unlikely to be of functional relevance. The missense mutation detected in exon 4b (204C\>A, P65T) most likely represents a polymorphic variant compared to the sequence in the database, as the heterozygous form was invariantly detectable in all tested patients and controls. This finding also excludes large deletions in the patients affecting the whole gene locus. Three intronic (IVS2+7A\>G, IVS3+11G\>T, IVS4a-62C\>A) and one silent mutations (30G\>C Ser10Ser) were not detectable in 100 controls. One variant in the 5\'UTR (2--40T\>C) and one missense mutation (30C\>T S27F) were only found once in control individuals (table [2](#T2){ref-type="table"}). We report on the mutation screening of PITX2, as we considered it to be an interesting candidate gene for TGA due to its role in regulating asymmetric cardiac morphogenesis \[[@B22]\] and interesting data from mouse studies. Impaired Pitx2 function in mice leads to severe cardiac malformations \[[@B17]-[@B20]\]. It has been suggested that altered PITX2expression in the outflow tract could underlie either TGA or DORV \[[@B22]\]. PITX2comprises three major isoforms, formed by differential splicing or alternative promotor usage: PITX2a, b, c, as well as one minor isoform PITX2d(Fig [1](#F1){ref-type="fig"}). We have included all coding exons in our screening as all forms exhibit a differential expression pattern \[[@B18],[@B19]\]. Pitx2cis of special interest, as only this isoform is asymmetrically expressed within the lateral plate mesoderm and the heart and governs asymmetric organ morphogenesis in a dose-dependent manner \[[@B23],[@B19]\]. Furthermore, the newly identified minor isoform, PITX2d, that in fact does not bind to DNA, was included in the study since it may influence expression levels of the other splice variants and also regulate the transcriptional activity of the major isoforms on protein level \[[@B24]\]. As only low amounts of PITX2 are required for normal cardiac development and as the different isoforms can possibly compensate for each other in some cell populations, it might require a combination of different sequence variants within different isoforms of the gene to dramatically reduce PITX2 function and therefore manifest a cardiac phenotype. Conclusion ========== To address whether altered PITX2 function could also contribute to the formation of dTGA in humans, we screened the coding regions as well as exon-intron boundaries of the PITX2gene for mutations in 96 patients with dTGA. The majority of detected variants, however, were also found in controls with comparable frequency. Three intronic and one silent mutation could not be detected in 100 controls. As they were only found once in the cohort of 96 patients and as none of the variants was found within the evolutionary conserved homeodomain, we consider them to be rare polymorphisms rather than functional mutations, although we cannot totally exclude the latter possibility. Further investigations will have to evaluate whether these sequence variants might change splicing processes. Due to the study design we can also not exclude mutations in the very 5\'and 3\' UTRs and within introns as well as the promoter regions of the gene. Nevertheless, we conclude that the detected mutations in PITX2are not a common cause of dTGA. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= NM designed the study, BN participated in the mutation analysis and drafted the manuscript, RR and KS performed PCRs and sequencing reactions. HJR, JG and EG provided patient care and collected blood samples. GR was involved in study design and supervision and finalized 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/1471-2350/6/20/prepub> Acknowledgements ================ We would like to thank D. Driscoll, Children\'s Hospital of Philadelphia, for support in collecting DNA samples. N.M. was supported by the Landesgraduiertenförderung, Baden-Württemberg, Germany. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Schematic diagram of PITX2isoforms a, b, c and d. The genomic organization of the PITX2gene is given on top, exons are numbered, 5\' and 3\' UTRs of the different possible transcripts are indicated by striped boxes and the homeodomain is shaded in dark grey. Modified from Cox et al, 2002 \[24\]. The regions of the PITX2gene included in the mutational screening are indicated as orange bars at the very top of the scheme. ::: ![](1471-2350-6-20-1) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Primer pairs used for mutation analysis, covering the coding region of PITX2. ::: ------------------------------------------------------------------------------------------------------- exon primer name sequence 5\'\> 3\' T~A~°C reference ---------- ------------------- ---------------------------------- ------------ ------------------------ 2 PITX2-exon2for:\ tag tct cat ctg agc cct gc\ 60 Ref: \[25\] this paper PITX2-exon2rev: gcg att tgg ttc tga ttt cct 3 PITX2-exon3bfor:\ ttg ctc ttt gtc cct ctt tct cct\ 60 this paper PITX2-exon3brev: cgg agt gtc taa gtt caa gca gca 4a PITX2-exon4afor:\ ccg cct ctg gtt tta aga tg\ 60 this paper PITX2-exon4arev: gca aag acc ccc ttc ttc tc 4b PITX2-exon4bfor:\ ctt gac act tct ctg tca gg\ 60/56/52\* Ref: \[25\] PITX2-exon4brev: aag cgg gaa tgt ctg cag g 5 PITX2-exon5for:\ cag ctc ttc cac ggc ttc t\ 60 Ref: \[25\] PITX2-exon5rev: ttc tct cct ggt cta ctt gg 6 PITX2-exon6for:\ gta atc tgc act gtg gca tc\ 65 Ref: \[25\] PITX2-exon6rev: agt ctt tca agg gcg gag tt ------------------------------------------------------------------------------------------------------- T~A~: Annealing temperature \* step down PCR was performed with three temperatures for 10/10/15 cycles. ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Summary of PITX2sequence variations in the dTGA study cohort ::: ---------------------------------------------------------------------------------------------------------------------------------------------- patients (n = 96) controls ---------------------------------------- ---------------------------- --------------------------- ------------------------ ------------------- type of variation: specific variation variant frequency (%) number of controls frequency (%) intronic/UTR variations: 2--40T\>C (5\'UTR exon 2)\ 10 (10.4%)\ 100\ 12 (12%)\ 2--18T\>C (5\'UTR exon 2)\ 0 (0%)\ 100\ 1 (1%)\ IVS2+7A\>G (intron 2)\ 1 (1.04%)\ 100\ 0 (0%)\ IVS2-106C\>A (intron 2)\ 17 (17.7%)\ 100\ 20 (20%)\ IVS3+11G\>T (intron 3)\ 1 (1.04%)\ 100\ 0 (0%)\ IVS4a+11G (intron 4a)\ 30 (31,25%)\ 100\ 39 (39%)\ IVS4a-62C\>A (intron 4a) 1 (1.04%) 100 0 (0%) silent mutations: 30G\>C (S10S) (exon 2)\ 1 (1.04%)\ 100\ 0 (0%)\ 63C\>T (A21A) (exon 4b) 1 (1.04%) 100 2 (2%) polymorphism within coding region: 30C\>T (S27F) (exon 3)\ 0 (0%)\ 100\ 1 (1%)\ 204C\>A (P65T) (exon 4b) 96 (100%) 100 100 (100%) ---------------------------------------------------------------------------------------------------------------------------------------------- UTR: untranslated region :::	PubMed Central	2024-06-05T03:55:57.420485	2005-5-12	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1142516/", "journal": "BMC Med Genet. 2005 May 12; 6:20", "authors": [ { "first": "Nadja", "last": "Muncke" }, { "first": "Beate", "last": "Niesler" }, { "first": "Ralph", "last": "Roeth" }, { "first": "Karin", "last": "Schön" }, { "first": "Heinz-Juergen", "last": "Rüdiger" }, { "first": "Elizabeth", "last": "Goldmuntz" }, { "first": "Judith", "last": "Goodship" }, { "first": "Gudrun", "last": "Rappold" } ] }
PMC1142517	Background ========== Positron Emission Tomography (PET) is a technique based on tracing of molecules labelled with positron-emitting radionuclides to image metabolism, physiology, and functionality in vivo in organs and tissues. PET has become an important, non-invasive technique for providing functional information about specific organs and areas of disease, and is used increasingly in clinical diagnosis, medical research and drug development. One of the most important properties of PET is its ability to supply quantitative values derived from functional images \[[@B1]\]. PET images are usually reconstructed either analytically by filtered back projection (FBP) or iteratively by ordered subsets expectation maximisation (OSEM), a much faster variation of maximum likelihood expectation maximisation (ML-EM) \[[@B2]\]. The former method is based on dividing the raw data into a number of subsets of projections (OS level) followed by applying a standard EM algorithm \[[@B2],[@B3]\]. FBP utilizes the 2D distribution from multi-angular projections \[[@B4]\], projects back these projections after applying 1D convolution with a specific high pass filter \[[@B5]\] to a common image plane. The filter used in FBP consists of a ramp filter, which is used to remove the blurring induced by the back projection. However, the filter also amplifies high-frequency noise. Therefore the filter is usually combined with a low-pass filter, such as a Hanning filter, to reduce noise. The properties of reconstructed images, using various filter functions, have been exploited \[[@B6]\], indicating how different degrees of filtering leads to different appearance and pattern of noise in the images. FBP is a relatively fast process but it has the drawbacks that the generated images contain more noise and are more sensitive to disturbing factors, such as patient movements during and between transmission and emission scans. Different algorithms \[[@B7]\] have been evaluated based on Fourier analysis to shorten reconstruction times and improve signal-to-noise ratio at equivalent resolution. In traditional PET scanners, with rotating ^68^Ge/^68^Ga transmission sources, the main sources of noise are in decreasing order of magnitude: emission, transmission and blank scans \[[@B8]\]. With newer attenuation correction modes, e.g. CT, the noise from emission is clearly dominating. Detectors and the recording system in the tomograph affect two characteristics of noise: magnitude and texture. The detector system affects only the noise magnitude, whereas the recording system affects both the noise magnitude and texture \[[@B9]\]. The choice of reconstruction algorithm and type of convolution kernel used in the reconstruction algorithm significantly affects the magnitude and correlation of noise \[[@B10]\]. Another important factor related to noise that affects the quality in PET images, and even more the potential to estimate precision in a measurement, is the correlation of noise between pixels. Image quality, in its simplest form characterized by pixel signal-to-noise ratio \[[@B11]\], becomes an inadequate measurement when different types of noise correlation exist between the pixels within the images. It has been shown that 3D PET images contain strong correlation between the values in adjacent pixels and the correlation is found to be a complex function \[[@B12]\]. The correlation of each element influences one or two pixels of the nearest neighbours \[[@B13]\]. The noise properties of emission tomographic images reconstructed by ML-EM and FBP have been compared \[[@B14]\]. The comparisons were based on the covariance matrix and noise properties as a function of iteration for ML-EM and as a function of noise apodization filter for FBP. The covariance matrix gave information about noise magnitude and texture, and indicated differences in noise pattern depending on applied reconstruction algorithm. Further studies have indicated that low intensity regions of images reconstructed by iterative algorithms tend to have low noise or a local noise pattern. In contrast, images reconstructed by FBP tend to have a much more globally distributed noise pattern \[[@B15],[@B16]\]. Comparisons have been made between images acquired in both 2D and 3D modes and in synthetic images. Due to differences in axial resolution and noise correlation between the two modes we compared images acquired in both modes. The decrease in axial resolution in 3D is attributed to the retraction of septa leading to increase in the crystal solid angles with a broadening of the point spread function and the smoothing effect introduced by 3D reconstruction. The observed 3D images are noisier near the centre of the FOV yet the axial correlation is lower in 2D images \[[@B17]\]. An essential aspect of PET is its ability to obtain quantitative values for regions of interest (ROIs) within the images. These values by themselves can have diagnostic value, which can give insights into physiology of normal and diseased tissues or can give important information about drug distribution or interaction with target systems. Since PET images are inherently noisy, the standard method to reduce noise for the quantitative estimates is to take averages over several pixels within an ROI. This is an adequate method, but because of the correlation between the pixels, it is not trivial to assign a precision value to these averages. Moreover an understanding of the noise properties in PET images is essential for adequate use of parametric images with statistical criteria included, such as in studies of blood flow changes during behavioural paradigms. Furthermore, it has been shown that variable noise levels in the different PET images, dramatically affect the subsequent principal component analysis, unless they are properly handled \[[@B18]\]. Although different aspects of noise have been covered extensively in the literature, we still feel that one aspect has not been adequately covered: the angular dependence of noise correlation in cases when the investigated object is asymmetrical. With asymmetric objects, the count rates will be different in the different acquisition angles and the relative magnitude of the noise is therefore different. It is possible that this angular dependent noise would during the image reconstruction propagate to the images and there generate a noise correlation which is non-isotropic. Here our main focus was to demonstrate the relationship between the shape of the experimental object and the properties of noise, especially the angular dependence of correlation in the PET images. Comparisons were made between images acquired in both 2D and 3D modes reconstructed using FBP and OSEM. We also compared results from autocorrelation function and the results using a covariance matrix applied to synthetic images. Methods ======= All experiments were performed on an ECAT Exact HR+ \[[@B19]\]. This scanner contains 32 detector rings separated by removable septa and is capable of performing 2D and 3D data acquisition. The total number of bismuth germanate detectors is 18 432 generating 63 contiguous image planes in the form of \[128 × 128\] matrices with an axial field of view (FOV) of 155 mm. These experiments were performed using either a 20-cm-diameter, 20-cm-long water-filled cylindrical phantom \[[@B20],[@B21]\], an elliptical torso phantom (long axis 30 cm, short axis 20 cm), or a combination of the 20-cm phantom with two adjacent 5-cm diameter, 20-cm-long water-filled cylinders positioned on opposite sides of the larger phantom (satellite phantom) to create a variety of noise textures in the images. Two different radionuclides, ^18^F and ^68^Ga, with 110 min and 68 min half-life, respectively, were used. ^18^F was produced using a Scanditronix MC-17 cyclotron (Scanditronix AB, Uppsala, Sweden). ^68^Ga was obtained from a ^68^Ge generator \[[@B22]\]. Prior to each experiment, a blank scan with rotating ^68^Ge /^68^Ga rod sources was acquired. The phantom was filled with 60 MBq of either ^18^F or ^68^Ga, placed at the centre of the FOV, and 30-min emission scans were made in both 2D and 3D mode. Finally to avoid artefacts in the images caused by movement of the object between transmission and emission scan, a 10-min post-injection (\'hot\') transmission scan was performed. A segmentation technique, as included in the ECAT 7.2 software (CTI, Knoxville, Tennessee) was applied on the hot transmission data before it was used for attenuation correction in the reconstruction process. The radioactivity concentration and mode of attenuation correction correlated reasonably well with the conditions used in clinical scans. For reconstruction of raw PET data, both FBP and OSEM which were included in the scanner software were used for reconstructing the data. The 3D data is rebinned to a 2D dataset using FORE followed by 2D reconstruction using either FBP or OSEM. Different types of low pass filters were used with each of the reconstruction methods, e.g. 4 mm (FWHM) Hanning and 6 mm (FWHM) Gauss filter. For FBP the filtering was made as part of the reconstruction ramp filter in the projections. For the iterative reconstruction it was made as post-reconstruction smoothing filters. The use of the same filter with FBP and OSEM ensured similar spatial resolution. In a simulation study on images reconstructed using FBP, a program was developed. For generating the synthetic PET images, Matlab (The Mathworks, Natick, Massachusetts) was used in which sinograms of a cylindrical phantom were calculated by forward projection of noise-free synthetic images of this phantom. An additive, Poisson noise with different magnitude in different angles was added to the sinograms, and images were reconstructed with FBP using Matlab\'s \"iradon.m\" routine. Another program was developed, using Matlab to calculate the ACF of the reconstructed PET images, performed both in the frequency and spatial domain. Papoulis \[[@B23]\] describes the autocorrelation function as a function often used in exploring similarity between images or image parts. The autocorrelation function A~ff~of a random process fis defined as a mean of the product of the random variables f(x~1~, y~1~, w~i~), and f(x~2~, y~2~, w~i~), A~ff~(x~1~, y~1~, w~i~, x~2~, y~2~, w~i~) = E{f(x~1~, y~1~, w~i~) f(x~2~, y~2~, w~i~)} (1) where Eis the mathematical expectation operator, f(x, y, w~i~) is a stochastic or random process, and w~i~is an element for a set of all events. Here we created the ACF based on the following equation by Kay \[[@B24]\]. r~yy~\[t, l\] = r~xx~\[t\] r~zz~\[l\] (2) where r~xx~\[t\] and r~zz~\[l\] are valid 1D ACFs. The spatial equation is based on 2D cross-correlation of the matrix a~i,\ j~with resolution of i× jwith itself using the lags k and l ![](1471-2342-5-3-i1.gif) where k and l refer to lags of the function and max(1, 1 - k) ≤ i≤ min(m, m- k) and max(1, 1 - l) ≤ j≤ min(n, n- l) A number of central slices of the image set were read to avoid effects at the edges of the FOV. The level of the noise is slightly higher near the centre of the FOV for 3D but not 2D mode. Subsequently, a matrix containing 25 × 25 pixels from the central part of the image was selected as a mask for the ACF. After subtraction of the average over this matrix, an ACF image was generated in which each pixel was set to the product of the mask matrix and a matrix in the original image, centred around the selected pixel position. The ACF image was then normalized by dividing each pixel value by the maximum pixel value within the ACF image. The results from this procedure applied to images from all experiments were studied and compared. The aim of this application was to study the noise correlation between the pixels within each image. The method was to analyse the shape of the 2D autocorrelation function in the images. The program results in images that can then be used to visualize and compare the ACF from PET images obtained with different reconstruction algorithms and different acquisition modes. 1D vertical and horizontal profiles through the ACF images were plotted to illustrate noise correlation. Third program was developed to calculate the covariance matrices based on a matrix with 25 × 25 pixels from the central part of synthetic reconstructed images using the FBP method, with the level of noise in projections angularly variable. This study was performed on 1500 synthetic images repeated 10 times, with independent random generated noise and the result as covariance matrices averaged. 1D covariance values were calculated using the equation, ![](1471-2342-5-3-i2.gif) where ![](1471-2342-5-3-i3.gif) N is the number of 2D synthetic images, which are 1500 in each attempt and ![](1471-2342-5-3-i4.gif) refers to the mean value of column vector containing all centre pixels x~i~of all synthetic images. ![](1471-2342-5-3-i5.gif) refers to the mean value of the column vector containing all j^th^adjacent pixels from the centre pixel x~i~of all synthetic images. The aim of this application was to study whether the autocorrelation and covariance methods gave identical results for noise magnitude and texture. To simplify this comparison of noise properties, we plotted a 1D profile through the middle section of the result after applying ACF and the result from calculated covariance values between same pixels. Results ======= In the 2D study on the NEMA phantom (Figures [1](#F1){ref-type="fig"}, [2](#F2){ref-type="fig"}, [3](#F3){ref-type="fig"}, [4](#F4){ref-type="fig"}), the results indicate an identical and isotropic form with a similar pattern of noise texture independent of applied reconstruction methodology and used filter (6 mm Gaussian and 4 mm Hanning). In the 3D study the same behaviour was noticed independent of applied reconstruction method and used filter. Image noise, propagating into the ACF images, generates low frequency oscillations. In both the 2D and 3D study on the torso phantom (Figures [5](#F5){ref-type="fig"}, [6](#F6){ref-type="fig"}, [7](#F7){ref-type="fig"}, [8](#F8){ref-type="fig"}), the results indicate non-isotropic behaviour of the noise with a dissimilar pattern of noise texture on images reconstructed using FBP, independent of used filter (6 mm Gaussian and 4 mm Hanning). However, the results from images reconstructed using OSEM indicate an identical and isotropic form with a similar pattern of noise texture in both 2D and 3D and independent of used filter. The main reason for performing the satellite phantom study was to observe how FBP and OSEM would handle an image with a stronger variation of the noise-texture. In the study performed on the satellite phantom, similar results were observed as in the study with the elliptical phantom. Hence the FBP generated a non-isotropic noise correlation whereas OSEM generated much more rotational symmetric pattern, although not fully for the 3D acquisition (Figures [9](#F9){ref-type="fig"}, [10](#F10){ref-type="fig"}, [11](#F11){ref-type="fig"}, [12](#F12){ref-type="fig"}). The ACF was applied to study noise correlation in synthetic images in which Poisson noise with different magnitude was added to the different projections (Figure [13](#F13){ref-type="fig"}). 1D profiles through the ACF image indicate a non-isotropic behaviour of the noise correlation (Figure [14](#F14){ref-type="fig"}). This study shows that FBP generates images in which the noise correlation is non-isotropic when the noise magnitude is angular dependent, such as in objects with asymmetric radioactivity distribution. Figure [15](#F15){ref-type="fig"} illustrates the result comparing a 1D profile using the 2D ACF and averaged covariance matrix from the study in synthetic images. The results are almost identical concerning noise correlation. In this case a higher resolution ramp filter was used. Discussion ========== The aim of the present study was to explore the properties of noise correlation in PET images. We expected that there could be differences in noise correlation in different directions when the noise level differed in different angular projections during acquisition. This would be the case with elliptical phantoms and under conditions where the radioactivity concentration was non-uniform. We therefore generated three sets of phantoms for the experimental determination of noise correlations. To explore noise correlation and to illustrate its features, we developed a program for the generation of auto-correlation images. Applying the auto-correlation function on the reconstructed PET images shows a very good similarity vs. dissimilarity concerning noise behaviour between two of the most commonly used reconstruction methodologies. ACF images from PET images reconstructed by FBP showed that the correlation pattern becomes asymmetrical when a study is performed on a non-circular phantom. This is illustrated in Fig. [5](#F5){ref-type="fig"} and [6](#F6){ref-type="fig"} for the elliptic phantom and Fig. [9](#F9){ref-type="fig"} and [10](#F10){ref-type="fig"} for the \"satellite phantom\". This is due to the way noise is handled in the FBP algorithm where the noise reduction or amplification by the filtering is the same in all projections. The back projection will then distribute different noise levels in different angular directions. On the other hand, the ACF image obtained from PET data reconstructed using OSEM did not show such a tendency. The pattern of correlation was shown to be isotropic and independent of the shape of the studied phantom. One possible explanation for this feature of iterative reconstruction is that the technique inherently attempts to iterate to similar deviations for each angular projection, a process that tends to equalise noise for the different projections. These results suggest the need for further studies on OSEM. We also expected that the result from applying autocorrelation function on PET data would provide the same information as calculating covariance matrix. Using ACF and covariance matrix to study synthetic images gave almost identical results for noise correlation and its behaviour. An advantage of using ACF instead of covariance or correlation is that by applying ACF on an image one can study and explore how adjacent, surrounding pixels affect the middle pixel, and illustrate the result in an image which can be visually inspected and used for further conclusions. E.g. artefacts due to detector or electronics mismatch might be easily revealed. Another advantage of applying ACF is that the function can be applied on one or a small number of images. A limitation of using ACF for a noise correlation description is that the method is only applicable on images without structural information, i.e., images obtained from uniform objects. It is not feasible to obtain reliable results from images with structural information, e.g. an image of human brain, because the data from the structural part affect the results. Another limitation of using ACF is that the function is sensitive to data with non-stationary statistics, e.g. variance, and consequently the mask used for performing ACF cannot be large. It is well known that the noise variance is variable over the image field but this aspect can be minimised by using a small mask. Here we selected to use a central mask of 25\*25 pixels, knowing that in this central region the noise amplitude is perhaps constant. Conclusion ========== We conclude that the noise correlation in FBP is angular and object dependent, and therefore it is e.g. not possible to apply general statistical methods to estimate precision in an average over a region of interest. With iterative reconstruction, the noise correlations seem to be more symmetric and vary less with the object. It might therefore be possible to apply generalised statistical methods but with due consideration to the fact that there is a significant correlation between pixels. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= Authors PR and MB helped with the design of the study. They created the method for applying ACF and performed the image and data analysis and drafted the manuscript. Author ML helped run the camera, acquire and reconstruct the data and write this paper. Authors HS, EB and BL helped with some of the practical approaches and the writing of the paper. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2342/5/3/prepub> Acknowledgements ================ The authors wish to thank the staff at Uppsala Imanet AB, especially Mr. Lars Lindsjö for his assistance with the measurements and the staff of the chemistry department for the radionuclide production. The authors thank Dr. Jan Axelsson, Dr. Gunnar Blomqvist and Felix Wehrmann for beneficial scientific discussions. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Results of the cylindrical NEMA phantom study. 2D acquisition image (upper left) and 3D acquisition image (upper right) reconstructed using FBP with applied 6 mm (FWHM) Gaussian filter. Corresponding ACF images of 2D (lower left) and 3D image (lower right). ::: ![](1471-2342-5-3-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Results of the cylindrical NEMA phantom study. Vertical (dash point) and horizontal profile (dash star) through the centre of the ACF image from 2D (upper) and from 3D acquisition image (lower) reconstructed using FBP with 6 mm Gaussian filter. ::: ![](1471-2342-5-3-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Results of the cylindrical NEMA phantom study. 2D image (upper left) and 3D image (upper right) reconstructed using OSEM with applied 6 mm (FWHM) Gaussian filter. Corresponding ACF images of 2D (lower left) and 3D image (lower right). ::: ![](1471-2342-5-3-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Results of the cylindrical NEMA phantom study. Vertical (dash point) and horizontal profile (dash star) through the centre of the ACF image from 2D (upper) and 3D acquisitions (lower) reconstructed using OSEM with 6 mm Gaussian filter. ::: ![](1471-2342-5-3-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Results of the elliptical torso phantom study. 2D image (upper left) and 3D image (upper right) reconstructed using FBP with applied 6 mm (FWHM) Gaussian filter. Corresponding ACF images of 2D (lower left) and 3D image (lower right). ::: ![](1471-2342-5-3-5) ::: ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Results of the elliptical torso phantom study. Vertical (dash point) and horizontal profile (dash star) through the centre of the ACF image obtained with 2D (upper) and 3D acquisition (lower) reconstructed using FBP with 6 mm Gaussian filter. ::: ![](1471-2342-5-3-6) ::: ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Results of the elliptical torso phantom study. 2D image (upper left) and 3D image (upper right) reconstructed using OSEM with applied 6 mm (FWHM) Gaussian filter. Corresponding ACF images of 2D (lower left) and 3D image (lower right). ::: ![](1471-2342-5-3-7) ::: ::: {#F8 .fig} Figure 8 ::: {.caption} ###### Results of the elliptical torso phantom study. Vertical (dash point) and horizontal profile (dash star) through the centre of the ACF image obtained with 2D (upper) or 3D acquisition (lower) reconstructed using OSEM with 6 mm Gaussian filter. ::: ![](1471-2342-5-3-8) ::: ::: {#F9 .fig} Figure 9 ::: {.caption} ###### Results of the satellite phantom study (NEMA phantom with adjacent hot cylinders). 2D image (upper left) and 3D image (upper right) reconstructed using FBP with applied 6 mm (FWHM) Gaussian filter. Corresponding ACF images of 2D (lower left) and 3D image (lower right). ::: ![](1471-2342-5-3-9) ::: ::: {#F10 .fig} Figure 10 ::: {.caption} ###### Results of the satellite phantom study. Vertical (dash point) and horizontal profile (dash star) through the centre of the ACF image obtained with 2D (upper) and 3D acquisition (lower) reconstructed using FBP with 6 mm Gaussian filter. ::: ![](1471-2342-5-3-10) ::: ::: {#F11 .fig} Figure 11 ::: {.caption} ###### Results of the satellite phantom study (NEMA phantom with adjacent hot cylinders). 2D image (upper left) and 3D image (upper right) reconstructed using OSEM with applied 6 mm (FWHM) Gaussian filter. Corresponding ACF images of 2D (lower left) and 3D image (lower right). ::: ![](1471-2342-5-3-11) ::: ::: {#F12 .fig} Figure 12 ::: {.caption} ###### Results of the satellite phantom study. Vertical (dash point) and horizontal profile (dash star) through the centre of the ACF image obtained with 2D (upper) and 3D acquisitions (lower) reconstructed using OSEM with 6 mm Gaussian filter. ::: ![](1471-2342-5-3-12) ::: ::: {#F13 .fig} Figure 13 ::: {.caption} ###### Example of synthetic image with additional Poisson noise with an angular dependent magnitude (left) and the result of applying ACF on the images (right). ::: ![](1471-2342-5-3-13) ::: ::: {#F14 .fig} Figure 14 ::: {.caption} ###### Vertical (dash point) and horizontal profile (dash star) through the centre of the ACF image obtained from the synthetic PET image ::: ![](1471-2342-5-3-14) ::: ::: {#F15 .fig} Figure 15 ::: {.caption} ###### Shows 1D profile through the centre of ACF (dash dot curve) and covariance matrix (solid curve) from the study using synthetic images. ::: ![](1471-2342-5-3-15) :::	PubMed Central	2024-06-05T03:55:57.422049	2005-5-13	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1142517/", "journal": "BMC Med Imaging. 2005 May 13; 5:3", "authors": [ { "first": "Pasha", "last": "Razifar" }, { "first": "Mark", "last": "Lubberink" }, { "first": "Harald", "last": "Schneider" }, { "first": "Bengt", "last": "Långström" }, { "first": "Ewert", "last": "Bengtsson" }, { "first": "Mats", "last": "Bergström" } ] }
PMC1142518	Introduction ============ Functional magnetic resonance imaging (fMRI) has become a vital tool for human brain function studies and medical diagnosis. Magnetic field strengths are expected to rise from the current 1.5 to 7 tesla (T) or further to achieve higher spatial resolution and signal-to-noise ratio, although biological effects of magnetic fields and their safety criteria for human subjects are still unclear. Biological effects of magnetic fields depend mainly on 1) the field strength and its gradient, 2) area and duration time of exposure to the field, and 3) the static or dynamic properties of magnetic field \[[@B1]\]. The threshold of time-varying magnetic fields to human exposure is a frequent theme and examined experimentally and theoretically \[[@B1],[@B2]\], because they might produce cardiac and peripheral nerve stimulation, heating, and magneto-phosphene during exposure. Furthermore, movement of subjects or patients within a static magnetic field produces sensory effects such as vertigo, nausea, and a peculiar metallic taste \[[@B3]\]. On the other hand, the static magnetic field has a potential to affect the blood flow through the following three mechanisms: 1) magnetohydrodynamic action, which is theoretically predicted to decrease the aortic blood flow velocity by 10% under homogeneous magnetic fields at about 5 T \[[@B4]\], 2) diamagnetic anisotropic interaction, which modifies the orientation of sickled and normal erythrocytes at homogeneous magnetic field of 0.35 T\[[@B5]\] and 4 T \[[@B6]\], respectively, and 3) paramagnetic interaction under a strong spatial gradient of magnetic field, which was applied to separate paramagnetic erythrocyte from whole blood \[[@B7]\]. Discussion ========== Paramagnetic interaction between gradient magnetic fields and flowing erythrocytes ---------------------------------------------------------------------------------- To estimate maximal effects of the gradient magnetic field, which is applied to the human in MR measurements, on the cerebral and pulmonary microcirculation, we examined the effects of gradient magnetic fields on erythrocyte flow in a model of branched blood vessels with Reynold\'s number as small as that for blood flow in capillaries and venules (see \'Supplementary Method 1\' in Additional file[1](#S1){ref-type="supplementary-material"}), which are major contributors to signal changes in fMRI. An inhomogeneous magnetic field (field strength and spatial gradient product: up to 60 T^2^/m) was transversally applied to the blood flow using an electromagnet whose direction of magnetic field is perpendicular to the vessel axis (see Fig. [1](#F1){ref-type="fig"} of Ref. \[[@B8]\]), which covered the gradient field of 10-T MRI system (see Fig. [1A](#F1){ref-type="fig"}). We observed that the spatial distribution of 70% oxygenated erythrocytes flowing in the branched vessel deviated in proportion to the product of magnetic field strength and its gradient up to 20 T^2^/m, above which the deviation saturated. On the other hand, fully oxygenated erythrocytes flow showed no deviation. The deviation of spatial distribution of 70% oxygenated erythrocytes can be attributed to paramagnetic attraction by an inhomogeneous magnetic field \[[@B8],[@B9]\]. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Paramagnetic effects of gradient magnetic fields on the pulmonary circulation during human brain mapping. (A) Magnetic field strength (Bz), and calculated field gradient (-dBz/dz) and force product (-Bz·dBz/dz) along the centerline of the magnet (see \'Supplementary Method 2\' in Additional file 1). Maximal values of -dBz/dz at Bz = 1.5, 3.0, 4.0, 7.0 and 10 T were 1.03, 2.06, 2.75, 4.81 and 6.87 T/m, respectively (at z = 1.47 m). Maximal values of -Bz·dBz/dz at Bz = 1.5, 3.0, 4.0, 7.0 and 10 T were 1.21, 4.84, 8.60, 26.40 and 53.75 T^2^/m, respectively (at z = 1.14 m), which are well covered by the range of our experiment shown in Fig. 1B. The head and lung of a healthy volunteer, whose height is 1.75 m, head length is 0.25 m, length of epigastric fossa from top of head is 0.5 m, were positioned at the magnet\'s center (A: z = 0 m) and at maximal of -Bz·dBz/dz (B: z = 1.14 m), respectively. (B) Maximally estimated redistribution of erythrocytes estimated in the pulmonary circulation at position B (see \'Supplementary Method 3\' in Additional file). ::: ![](1476-5918-4-5-1) ::: Effect of gradient magnetic fields on the pulmonary circulation --------------------------------------------------------------- First, we estimated the maximal effects of the deviation of erythrocyte distribution on the pulmonary blood flow, which could cause a ventilation perfusion mismatch (Fig. [1B](#F1){ref-type="fig"}). When a volunteer was positioned under the maximal inhomogeneous magnetic field (position B in Fig. [1A](#F1){ref-type="fig"}), the deviation of erythrocyte distribution in the pulmonary blood flow increased with increases in the product of field strength and its gradient but it saturated above \\|Bz·dBz/dz\\| \> 40 T^2^/m (Fig. [1B](#F1){ref-type="fig"}). However, the estimated value of deviation (maximally ca. 1.3%) was small and its effect on the ventilation perfusion mismatch is negligible even under the maximal inhomogeneous magnetic field generated by the 10-T system. Effect of gradient magnetic fields on the cerebral microcirculation ------------------------------------------------------------------- Next, we estimated maximal effects of such deviation of erythrocyte distribution on signal changes in functional neuroimaging based on the commonly used blood oxygenation-level dependent (BOLD) (Fig. [2A](#F2){ref-type="fig"}) and flow sensitive alternating inversion recovery (FAIR) (Fig. [2B](#F2){ref-type="fig"}) techniques. The gradient magnetic fields produced underestimation of BOLD and overestimation of FAIR signals almost linearly with the increases in the magnetic field strength. However, the values (0.024% for BOLD and 0.07% for FAIR) were quite small and these artifacts are negligible even for the 10-T system. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Effects of magnetic field under gradient echo echo planar imaging (GE-EPI) on the cerebral microcirculation. (A) Maximal effects of magnetic field under GE-EPI on BOLD signal at position A during functional brain mapping at slice gradient = 40 mT/m for all MR systems (see \'Supplementary Method 4\' in Additional file 1). (B) Maximal effects of magnetic field under GE-EPI on FAIR signal at position A during functional brain mapping at slice gradient = 40 mT/m for all MR systems (see \'Supplementary Method 5\' in Additional file). ::: ![](1476-5918-4-5-2) ::: Conclusion ========== Our results suggest that paramagnetic artifacts in the functional neuroimaging and disturbance in the pulmonary microcirculation during magnetic resonance imaging are quite small using the MR system with field strength up to 10 T. Authors\' contributions ======================= AS conceived, designed and coordinated the study, carried out the experiments, and drafted the manuscript. JS participated in the design and coordination of the study. MI participated in the experiments. TY participated in the design and coordination of the study, and directed it. All authors read and approved the final manuscript. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Supplementary Method 1: Experimental Setup. Supplementary Method 2: Estimation of magnetic fields. Supplementary Method 3: Estimation of redistribution of erythrocytes in the pulmonary circulation. Supplementary Method 4: Estimation of paramagnetic artifacts in BOLD signal. Supplementary Method 5: Estimation of paramagnetic artifacts in FAIR signal ::: ::: {.caption} ###### Click here for file :::	PubMed Central	2024-06-05T03:55:57.424125	2005-5-7	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1142518/", "journal": "Dyn Med. 2005 May 7; 4:5", "authors": [ { "first": "Akitoshi", "last": "Seiyama" }, { "first": "Junji", "last": "Seki" }, { "first": "Mari", "last": "Iwamoto" }, { "first": "Toshio", "last": "Yanagida" } ] }
PMC1142519	Cancer immunotherapy: transition from nonspecific to specific immunotherapy =========================================================================== There is broad appeal for the concept of treating cancer with the immune system. Early anecdotal experiences over 100 years ago suggested that induction of generalized immune activation, by a bacterial infection, could induce regression of solid human cancers in a small subset of patients \[[@B1],[@B2]\]. However, efforts to generalize this finding by treating patients with bacterial agents (e.g.: Bacille Calmette-Guerin, BCG) were disappointing \[[@B3]\]. Subsequent efforts were to vaccinate with cancer cell preparations to induce immune responses more specifically against cancer antigens that had not yet been defined. These included whole cell vaccines, cancer cell lysates, and cultured cell supernatants \[[@B4]-[@B8]\]. The molecular identity of cancer-specific antigens was sought over several decades, with most of the work focusing on melanoma. Initially, numerous cell surface antigens were identified by serologic methods in mice \[[@B9]\]. Vaccination against those antigens can induce specific antibodies \[[@B10]\]. However, a recent clinical trial of vaccination against one such antigen (the ganglioside GM2) had a negative result in terms of clinical outcome \[[@B11]\]. The potential of vaccines for induction of anti-tumor antibodies has not been fully explored, and deserves further investigation. However, in recent years, substantial effort has been directed at defining antigenic targets for CD8+ cytotoxic T lymphocytes (CTL), leading to new vaccine strategies designed to induce antigen-specific CTL using these antigens. Preclinical models of tumor vaccines: role of CD8+ and CD4+ T cells in tumor protection ======================================================================================= In murine studies, cell-based tumor vaccines can protect against cancer progression and can lead to regression of early established tumors. The protective immunity induced by syngeneic tumor vaccines appears to be mediated most directly by T-cells, and in many studies, depletion of CD8^+^T cells abrogates the protective effect of syngeneic tumor cell vaccines \[[@B12]\], suggesting cytotoxic T-cells are critical to that protective immunity. In some studies, however, depletion of CD4^+^T-cells also abrogates all or part of the protective immune response to vaccines \[[@B13]\]. Furthermore, adoptive therapy with CD4^+^T-cells can induce tumor protection in some model systems \[[@B14]\]. Thus, the protective immunity induced by syngeneic tumor cell vaccines appears to be mediated both by CD8^+^T-cells and by CD4^+^T-cells. These findings directed efforts toward identifying the molecular nature of tumor antigens recognized by CD8 and CD4 T cells. It was only in the last 1--2 decades that the nature of these antigens became known \[[@B15]\]. It was discovered that short peptides from cellular proteins were presented in association with cell-surface MHC molecules, and that these peptides represented epitopes for these T cells. Molecular definition of tumor antigens recognized by T-cells ============================================================ In the late 1980s, it was found that melanomas expressed shared antigens recognized by CD8^+^cytotoxic T lymphocytes (CTL) \[[@B16]\]. Subsequent studies beginning in the 1990s defined the molecular nature of some of these antigens \[[@B17]-[@B22]\]. The peptides recognized by cytotoxic (CD8^+^) T-cells are typically 8--10 amino acids long and are presented in association with Class I MHC molecules. The peptides recognized by helper (CD4^+^) T-cells are usually longer (generally 13--18 amino acids in length, although peptide elution studies have indicated no apparent restriction on peptide length) and are presented in association with Class II MHC molecules. For melanoma, the melanocytic differentiation proteins (MDPs) and the cancer-testis antigens (CTAs) are the most common source proteins for these defined shared peptide antigens. Now, a large number of peptide epitopes recognized by melanoma-reactive human CTL and helper T-cells are known (reviewed in \[[@B23],[@B24]\], making it possible to design vaccines using these antigens. At least as importantly, evaluation of T-cell responses to these defined antigens is now possible, and may permit evaluation of the immune responses induced by vaccine strategies, and to dissect the immune response. As outlined below, it has become clear that this approach can aid in optimizing vaccines. Peptide vaccines provide the unique opportunity to evaluate the T cell responses specifically to defined immunogens. Application of defined antigens to tumor vaccines ================================================= Peptide epitopes for melanoma-reactive cytotoxic T-cells were first identified in 1991, and epitopes for melanoma-reactive helper T cells have been identified in recent years. Some of these agents have been employed in experimental melanoma vaccines over the past 10 years or less. Peptide vaccines have theoretical and practical appeal, but also have certain drawbacks, as summarized in Tables [1](#T1){ref-type="table"} and [2](#T2){ref-type="table"}. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Practical and Theoretical Advantages of Peptide vaccines for cancer ::: Characteristic Detail Advantage vs -------------------- -------------------------------------------------------------------------------------------------- ------------------ ------- Pure Avoid tolerizing cellular antigens; exclude normal protein, avoid autoimmunity. X Processed Avoid effects of immunoproteasome X Cheap Feasible to study without corporate support X X Easier Lower regulatory hurdles X X Evaluable Excellent cancer vaccine model, allowing direct evaluation of response to the specific immunogen X Modifiable Create synthetic peptides better than native peptides X Immunogenic Induce T cell responses in patients Combinable Multipeptide vaccines may mimic immune effects of whole cell vaccines. ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Limitations of peptide vaccines ::: ------------------------------------------------------------------------------------------- • Limited by MHC restriction. • Unique individual tumor-specific antigens difficult to include. • Rapid degradation in vivo. • Heterogeneity of tumor antigen expression. • Ignorance. We don\'t yet know how best to vaccinate with them. \* • Clinical responses have been rare in most series (with peptide or any vaccine alone).\* ------------------------------------------------------------------------------------------- \* The last two points apply equally to practically all T cell vaccines, not just peptide vaccines. ::: With peptide vaccines, it has been possible to generate antigen-specific T cells at frequencies of 0.1% to greater than 2% percent of circulating CD8 T cells in many individuals \[\[[@B25]-[@B29]\], and unpublished results\]. However, when vaccines contain only single peptides, or small numbers of peptides, targeting CD8 T cells responses only, low clinical response rates have been observed \[[@B30]\]. In reality, that should not be surprising, especially in the setting of advanced tumor burden. Antigenic heterogeneity is the rule in tumor deposits. Adoptive therapy with T cell clones specific for a single antigen has led to eradication of melanoma cells expressing that antigen, but the tumors have not regressed, because of the persistence of antigen-loss variants \[[@B31]\]. Furthermore, T cells infiltrating tumor deposits are commonly found to be anergic or poorly responsive to antigenic stimulation, leading to the perception that the tumor microenvironment is hostile to the T cell response \[[@B32]\]. Effective immune therapy will require induction of T cell responses to multiple antigens simultaneously, and promotion of T cell activity in tumor tissue. Additional approaches to block immunoregulatory mechanisms may well also be needed for immune therapy to be successful. Is adoptive immunotherapy more fashionable than cancer vaccines? ---------------------------------------------------------------- Recent clinical successes in one study with adoptive T cell therapy in patients with metastatic melanoma have heightened enthusiasm for adoptive therapy \[[@B33]\]. In the wake of this renewed enthusiasm for adoptive T cell therapy, it has been stated that current peptide vaccines have failed \[[@B30]\]. Furthermore, a corollary argument is surfacing, that peptide vaccines (or other active specific immunotherapy for cancer) may not be worthy of continued investigation. This could not be more wrong. Perhaps the greatest failure of the tumor immunology research community is its reliance on fashion. Historically, encouraging early results with various immune therapies have induced great enthusiasm, followed soon thereafter by dashed hopes as the therapy proves not to be as effective as originally hoped. A lesson can be learned from the failures and successes of immunotherapy with monoclonal antibodies. In the 1970s and early 1980s, monoclonal antibodies were popularly considered magic bullets, and antibody therapy was in high fashion. Subsequently, in the 1980s, numerous therapeutic clinical trials with monoclonal antibodies led to very disappointing results. Consequently, monoclonal antibody therapy fell out of fashion. However, some persistent investigators focused on studying antibodies with certain specificities and on learning how to overcome HAMA reactions by humanizing monoclonal antibodies. The result has become common knowledge: multiple monoclonal antibodies are now used for several FDA approved therapies against cancer, such as herceptin (anti-Her-2/neu) and Rituximab (anti-CD20), and more recently Avastin (anti-VEGF). These successes took decades, but they now have firmly established immune therapy as a standard treatment option for multiple cancers. The lesson from this history is that one should persist in developing therapeutic approaches as long as they are promising and are built on continuous progress in the understanding of pathophysiological mechanisms. T cell immunotherapy of solid tumors is still in its experimental phase. Investigators in this field can and will bring together innovative tools and scientific reasoning in order to maximize the likelihood that the next generation of cancer vaccines will have therapeutic value. Similarly, adoptive therapy approaches have been studied for many decades, with many false starts and failures prior to the current exciting results. The recent successes with adoptive therapy are welcome and offer promise for further development. However, as with cancer vaccines, there remains much work to optimize adoptive therapy. The particular adoptive therapy study cited above is a modification of prior adoptive therapy approaches. Early enthusiasm for adoptive therapy with lymphokine-activated killer (LAK) cells in the 1980s was based on similar successes at the NCI, but subsequent multicenter investigations suggested that all or most of the therapeutic effect associated with LAK cell therapy could be mimicked by systemic therapy with high-dose IL-2 alone \[[@B34]-[@B37]\]. Subsequent studies with adoptive transfer of tumor-infiltrating lymphocytes (TIL) expanded ex vivoin IL-2 were associated with clinical regressions in 55% of patients in early studies \[[@B38]\], but this has largely been abandoned due to failure to maintain response rates that were convincingly better than that expected from high dose IL-2 alone \[[@B39],[@B40]\]. The new approach to adoptive therapy at the NCI involves peripheral lymphoablation followed by adoptive transfer of TIL expanded ex vivoafter selection for tumor lytic potential \[[@B33],[@B41],[@B42]\]. It is currently unclear whether the improved results with this combination therapy are due primarily to the lymphoablation, the adoptive transfer, or the type of T cells expanded for the adoptive transfer. Also, the high rate of objective clinical regressions in the current NCI experience (51%) is very similar to the high rate reported in prior NCI studies, which were not maintained in subsequent experience (Table [3](#T3){ref-type="table"}). ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Rates of clinical tumor regression in studies of adoptive transfer of tumor-reactive lymphocytes ::: Type of therapy Initial rate of objective responses Subsequent rate of objective responses Conclusion ---------------------------------------------------- ----------------------------------------- -------------------------------------------- ------------------------------------------------------------------------------------------------------------------------------------------------- LAK cell therapy + high-dose (HD) IL2 44% (11/25) \[ref 34\] 22% (23/106) \[ref 35\] Response rate not better than HD IL2 alone (28 vs 22%), but trend toward improved survival with LAK+IL2 for melanoma (p = 0.064) \[refs 36,37\] TIL therapy + HD IL2 55% (11/20) \[ref 38\] 22% (9/41) \[ref 39\] Not better than HD IL2 alone \[ref 39\]. Median duration of partial responses 4 months \[ref 40\]. Selected TIL therapy after lymphoablation + HD IL2 51% (18/35) \[ref 33\] Pending Results preliminary ::: Is it all about quantity or about quality? ------------------------------------------ One of the major arguments for use of adoptive cellular immune therapy for cancer is that it can achieve much higher numbers of circulating CD8 cells with anti-tumor specificity. Certainly it is true that patients treated with lymphoablation and adoptive TIL therapy plus high dose IL-2 have had extremely high numbers (and frequencies) of tumor-antigen specific T cells in circulation, with over 90% of circulating CD8 cells reacting to the immunodominant HLA-A2 restricted MART-1/MelanA antigen in one patient, and with a large proportion of patients having more than 10% of circulating CD8 cells with anti-tumor specificity \[[@B42]\]. A major observation is that the generation of high numbers of circulating anti-tumor CD8 T cells is insufficient to induce clinical tumor regressions in about half of patients, and is often insufficient to control melanoma completely in the large majority of patients. It can safely be concluded, thus, that factors other than the number of anti-tumor CD8 T cells affect immune control of cancer. These factors are being elucidated gradually, and they are the primary obstacles against which the next 5--10 years of translational and clinical research in immune therapy need to be targeted. These obstacles to success of adoptive transfer therapies are the same that interfere with the clinical efficacy of cancer vaccines. Some of the obstacles to immunologic control of tumor progression are listed in Table [4](#T4){ref-type="table"}. It is far more important for investigators in immunotherapy and cancer immunology to join forces in identifying and overcoming these factors than for us to argue whether peptide vaccines, viral vaccines, adoptive transfer, or other immunotherapy approaches are superior or inferior to others. ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Known or possible obstacles to immunologic control of tumor progression, which impact on both active immunotherapy (cancer vaccines) and adoptive immunotherapy. ::: ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------- 1\) Expression of tumor antigens in the absence of costimulatory molecules on tumor cells, leading to tolerance 2\) Chronic antigen exposure, leading to upregulation of immuno-regulatory mechanisms a) CTLA4 expression b) Accumulation of regulatory T cells in the tumor microenvironment 3\) Downregulation of MHC molecule expression by tumor cells 4\) Downregulation of tumor antigen expression by tumor cells 5\) Secretion of anti-inflammatory cytokines by tumor cells or tumor-associated stroma a) IL-10 b) TGF-β c) Others 6\) Expression of enzymes in the tumor microenvironment that interfere with T cell function a) Arginase b) Indoleamine 2,3-dioxygenase (IDO) 7\) Propogation of a tumor microenvironment that is hostile to T cell activation a) Immunoregulatory function of dendritic cells b) Anergic tumor-infiltrating lymphocytes 8\) Tumor-associated VEGF and other neovascularity-enhancing mechanisms may have immunoregulatory properties as well. 9\) Homeostatic mechanisms in the host may limit expansion of tumor-specific T cell responses, and may limit expansion and persistence of tumor-specific T cell responses. 10\) Resistance of tumor cells to apoptosis 11\) Elaboration of compounds associated with tumor necrosis, that inhibit anti-tumor immunity locally ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ::: Some patients enrolled in peptide vaccine studies have had marked expansion of antigen-reactive CD8+ T cells, with 5--10% of circulating CD8 cells reactive to antigen in some cases, and over 1% reactive to antigen in many cases \[[@B25]-[@B29]\]. While it is worthwhile to induce further expansion of T cells after cancer vaccines, it is likely that the quality of the immune response, rather than simply its magnitude, is critical to the success of immune therapy. Several approaches for improving immunotherapy with cancer vaccines need to be pursued, as listed in Table [5](#T5){ref-type="table"}. ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Potential avenues for improving therapeutic value of cancer vaccines ::: Obstacle Potential solution Status ----------------------------------------------------------------------------------------------------------------------- --------------------------------------------------------------------------------------------------------------------------------------------------------------------------- --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Heterogeneity of antigen expression Multi-antigen vaccines 12 peptide vaccine induces T cell responses in 100% of patients. Peptide competition for MHC binding does not inhibit immunogenicity \[ref 43\] MHC downregulation on tumor cells Targeting peptides associated with multiple MHC molecules Being investigated in many centers Failure of T cells induced in the periphery with vaccines to expand in the tumor microenvironment (inadequate memory) Addition of melanoma (or other cancer) associated helper peptides in vaccines \[refs 24, 44\] Early data inadequate to address the question refs \[45--47\]. Data in the HIV setting supports this approach \[ref 48.\] ECOG 1602 trial will address the questions with a cocktail of 6 melanoma helper peptides. Increased regulatory T cells in patients with advanced cancer, and in tumor microenvironment Inhibition of T reg function (anti-CTLA4 antibody); specific depletion of CD25+ regulatory T cells (Ontak); depletion of regulatory cells with chemotherapy (eg: cytoxan) Clinical trials with all of these agents are underway. Limited expansion of antigen-specific T cells after vaccination Pre-vaccine lymphodepletion to allow vaccination in the setting of naturally induced cytokines supporting homeostatic proliferation (eg IL7 and IL15) Studies are being designed to address this approach T cells induced by vaccination may not be activated effector cells Increase adjuvant function, perhaps by use of Toll-like receptor agonists CpGs and other TLR agonists being investigated as adjuvants \[29\]. Randomized phase II trials with immunologic endpoints needed. ::: Proof of principle of vaccines for cancer ----------------------------------------- The current manuscript is focused primarily on vaccine therapy, especially peptide vaccines, for solid tumors such as melanoma. However, a very important paradigm of cancer immunotherapy should be mentioned in this discussion. For those cancers whose primary etiologic factor is a known viral infection, vaccination against infection with that virus promises to have significant oncologic value. Specific examples are listed in Table [6](#T6){ref-type="table"}. ::: {#T6 .table-wrap} Table 6 ::: {.caption} ###### Virally-induced cancers subject to control by vaccines. ::: Cancer histology Etiologic virus Vaccine strategy Current use Clinical value -------------------------------------------- ----------------------- -------------------------------------------- ----------------------------------------------------------- --------------------------------------------------------------------------------------------------------------------------------------------- Hepatoma Hepatitis B Protein subunit vaccine In common use for high-risk populations. Protection against Hepatitis B infection is prolonged after three vaccines. Worldwide protection against hepatoma may have dramatic impact. Cervical adenocarcinoma Human Papilloma Virus Viral and other vaccines against E6 and E7 Strong evidence for efficacy in certain populations Likely will protect against cancer, especially for patients without access to Pap smears Burkitt\'s lymphoma, Nasopharyngeal cancer Epstein-Barr Virus Some T cell antigens identified Vaccines would have to be administered very early in life Untested. ::: There are differences between these clinical settings, where vaccines may prevent cancer by preventing the causative viral infection, and the more common scenario where vaccines are being considered to treat patients already diagnosed with cancer. The latter clinical setting represents chronic (vs. acute) antigen exposure and the reality that a cancer that progresses clinically has likely developed one or more mechanisms of immune escape or tolerance. Also, cancer progression commonly is associated with antigenic heterogeneity, which complicates the development of successful multi-antigen immunotherapy. However, the clinical and immunologic successes of anti-idiotype vaccines for some B cell lymphomas show that vaccines can induce protective immunity against a defined tumor-specific antigen, even in the setting of prior chronic antigen exposure \[[@B49],[@B50]\]. Where the antigen is integral to the malignant cell, targeting that antigen can have encouraging clinical results. Ultimately, the ideal cancer vaccine will be effective at inducing protective immunity, and will be safe enough to administer early in life before the initial carcinogenic events. Cancer vaccines, but not adoptive cellular therapy, hold out the prospect of being useful for cancer prevention on a wide scale. Objective clinical responses in patients enrolled in experimental melanoma vaccine trials ========================================================================================= In numerous published clinical trial results with cancer vaccines, one or more objective clinical tumor regressions have been observed. Though the overall objective response rate is low \[[@B30]\], even these infrequent clinical responses are proof of principle of cancer vaccines. Most current vaccines target only one or a few cancer antigens, restricted usually by just one MHC molecule. Since antigenic heterogeneity is the hallmark of cancer, it is most remarkable that these simple vaccines can lead to clinical regressions in any patients. The majority of current vaccines also target only CD8+ T cells and largely ignore CD4+ T cell responses, and responses of the innate immune system. Again, considering how simplistic the early peptide-based vaccines are, it is remarkable, and even encouraging, that they have been associated with any clinical tumor regressions. Some published studies have reported the proportion of patients with regressions of even just one lesion, and thus describe a higher proportion of clinical tumor regression than would be reported using RECIST criteria. However, a summary review of the NCI experience with vaccines and of the global experience with antigen-specific cancer vaccines, reveals that objective clinical response rates globally are in the range of 3--4% with recent cancer vaccines \[[@B30]\]. While this is certainly low, it is relevant that reported response rates with approved systemic therapy are only 12% for DTIC (dacarbazine), 11% for CVD (cisplatin, vincristine, dacarbazine), 16% for high-dose interleukin 2, and 17% for biochemotherapy \[[@B51]-[@B53]\]. Considering the low toxicity of peptide vaccines, an argument can be made that even current cancer vaccines have a prospect of clinical benefit for patients that rivals that of approved therapies, when one considers the risk:benefit ratio. Monitoring ---------- One of the arguable values of adoptive therapy is the ability to enrich or to deplete the cellular reagents and to define the specificity of the T cells used for therapy. With vaccines, it is not possible to select particular lymphocyte populations from the patient directly. However, the compartments of the immune system are natural environments for optimal expansion of T cells and for the complex interplay among innate and adaptive immune mechanisms. It is presumptuous to believe that our understanding of this complexity and our technologies are adequate to allow us to recreate optimal immune effectors in vitroand to expect them to perform as we desire upon re-infusion. However, it is possible in patients on clinical trials, to enrich for specific effectors by vaccination with defined antigens, and to measure their responses to each antigen simultaneously, in various compartments (eg: lymph node, blood, and tumor) \[[@B27],[@B32],[@B43],[@B54]\]. Furthermore, manipulations can be performed in vivo, to enrich or to deplete certain T cell subsets. Reagents exist for depletion of regulatory T cells (Ontak), for depletion of T cells (OKT3) or B cells (Rituximab), and there is increasing evidence that numerous cytotoxic chemotherapy agents have immunomodulatory effects that may be useful for augmentation of immunotherapy. Our challenge is to characterize these agents and their effects on development of protective immunity in patients treated with cancer vaccines. All such studies require careful immune monitoring, both to assess the effects of immune modulations over time, and to determine whether such changes are useful and evaluable. We would like to point out that surrogate endpoints for vaccine efficacy should be re-emphasized, despite some current sentiment to the contrary. For the development of new generation vaccines, we must rely on knowledge derived from basic research. In infectious diseases, it is well established that antigen specific lymphocytes must be activated substantially for successful (i.e. protective) vaccination. Consequently, assessing responses of antigen specific lymphocytes is an important step in the evaluation of novel vaccines. There are a number of new techniques permitting investigators to dissect T cell responses ex vivo. It is now possible to determine molecular features of human T cell responses in great detail, going much beyond what is usually done to assess T cells in animal models \[[@B55]-[@B57]\]. Economical and ethical considerations require that one takes maximal advantage by studying each patient in depth. Moreover, many issues in modern vaccinology must be assessed specifically in humans, since species differences do not allow to draw direct conclusions from experimental models. It is generally accepted that a protective T cell response includes T cells with high avidity T cell receptors, with expression of effector molecules and function, and with appropriate homing capability. Such features can and need to be determined by analyzing patients\' T cells ex vivobefore and after vaccine therapy, allowing evaluation of the potential value of a given vaccine. Many new vaccine candidates are being proposed to treat cancer patients. The scientific community is well advised to use biological readouts extensively in order to assess thoroughly the T cells from study patients. By doing so, one can rapidly eliminate useless approaches and promote good vaccine components for further development. Summary ------- Immune therapy of cancer may take many forms, specific or non-specific, adoptive or active, and may target antibody, T cell, and innate immune mechanisms. Each of these approaches has proven or potential value, and the complexity of the host: tumor relationship is such that a narrow focus on a single immunotherapy strategy is likely to fail. Adoptive T cell immunotherapy studies have provided strong proof of principle that antigen-specific CD8+ T cell responses to cancer can mediate dramatic cancer regressions. However, adoptive therapy is cumbersome and expensive, and difficult in the current regulatory environment. Vaccines, on the other hand, are more readily adaptable for therapy outside of highly specialized centers. In particular, peptide vaccines are easily produced, standardized, and administered. The current appeal of adoptive therapy is that antigen-specific T cells can be expanded and activated at high numbers ex vivo, more readily than they can be expanded in vivoin cancer patients. However, we argue that the lesion in current approaches to cancer vaccine therapy is our poor understanding of the mechanisms that limit expansion, activation, and effector function of tumor-antigen specific T cells. Bypassing this process by use of adoptive therapy is a reasonable short-term effort, but ultimately to advance the field of tumor immunology and immunotherapy it will be critical to elucidate the immunobiology of the host-tumor relationship. Appropriate design of cancer vaccines using multiple antigens should be combined with careful monitoring of T cell expansion and T cell function. Optimally, immune monitoring should be performed in multiple compartments (peripheral blood, tumor tissue, lymph nodes). The next wave of investigation in cancer immunotherapy has begun, and will include combination therapies designed to activate innate and adaptive immunity simultaneously and to down-modulate tumor-associated immune regulation. Vaccines with defined antigens are ideal for investigations of this type.	PubMed Central	2024-06-05T03:55:57.425088	2005-4-29	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1142519/", "journal": "J Transl Med. 2005 Apr 29; 3:18", "authors": [ { "first": "Craig L", "last": "Slingluff" }, { "first": "Daniel E", "last": "Speiser" } ] }
PMC1142520	Background ========== The concept of therapeutic implantation of autologous cells capable of replicating and differentiating into specialized progeny has generated great interest. This approach is being intensively investigated especially in cardiovascular medicine for the regeneration of the infarcted heart \[[@B1]\]. Skeletal muscle has an outstanding ability to respond to the requirements of growth, remodeling, physical activity, and injury. Since adult myofibers are terminally differentiated, the regeneration of skeletal muscle is mostly dependent on a population of resident cells termed satellite cells (SC). These cells lie beneath the basal lamina, but above the plasma membrane of myofibers \[[@B2]\]. When physiologically required, SC become activated, proliferate, differentiate and eventually fuse into myotubes that mature into myofibers \[[@B2]\]. The process of skeletal muscle regeneration is influenced by the nature of injury (e.g., traumatic, chemical, ischemic), the composition of the extracellular matrix, the availability of growth factors produced by inflammatory cells, vascular cells, and motor neurons \[[@B3]\]. SC have a smaller nuclear size compared with the adjacent nucleus of the myotube, and an increased amount of nuclear heterochromatin compared with that of the myonucleus \[[@B4]\]. Quiescent SC show little transcriptional activity. Following activation, they appear as a swelling on the myofiber with cytoplasmic processes that extend from one or both poles of the cell. In adult skeletal muscles, SC have been estimated to represent 1% to 12% of cells in adult muscle. Activation of SC during muscle regeneration requires upregulation of Myf5and MyoD, transcriptional activators of the myogenic regulatory factor family (MRF). Expression of these factors determines the transition of myogenic cells into myoblasts \[[@B3]\]. Expression of Pax3, Pax7, Sox15, MNF, cyclooxygenases and other regulatory factors is also needed for SC activation \[[@B5]-[@B7]\]. Expression of the late MRFs (myogenin and MRF4) precedes the production of muscle-specific proteins (e.g., myosin heavy chain and muscle creatine kinase). Expression of cytoskeletal (desmin) and surface markers (M-cadherin, Syndecan-3 and -4, VCAM-1, CD56, Glycoprotein Leu-19, CD34) as evidenced by immunostaining is often exploited to follow myogenic differentiation \[[@B3],[@B8],[@B9]\]. Over the past decade, experimental evidence has been accumulating that the implant of cultured myoblasts can represent an effective approach for repairing damaged myocardium \[[@B1],[@B10]-[@B12]\]. The functional benefit of intramyocardially transplanted autologous or heterologous skeletal myoblasts has been established in different animal models (rat, hamster, sheep) in which heart damage was determined primarily by ischemia \[[@B13],[@B14]\]. In humans, autologous skeletal myoblast implantation has been practiced only in cases of left ventricular dysfunction produced by ischemic damage \[[@B15],[@B16]\]. Recently, the implant of autologous myoblasts has been proposed also in the setting of dilated cardiomyopathy \[[@B17]\], urinary incontinence \[[@B18],[@B19]\], traumatic muscle damage \[[@B20]\], congenital muscle disease \[[@B21]\]. As compared to other attractive cell types, skeletal myoblasts have several clinically appealing properties, including autologous origin, ease of procurement, the ability of sufficient expansion in vitro, and the ability of adequate differentiation upon in vivo implant. The present study was initiated to determine if currently available cell culture methods were suitable for obtaining sufficient numbers of differentiated cells from muscle biopsies of human donors aged over 40 years. The latter condition was set since, with the exception of traumatic injury, most of the above listed clinical conditions arise in patients older than 50 years. In this context, it should be borne in mind that, for cellular therapy, substantial numbers of implantable cells (i.e., ≥10^8^) are needed within 1--2 months of disease onset. In addition, cells expanded in vitro need to be adequately differentiated along myogenic lineages, free of demonstrable genetic alterations, and free of contaminant macromolecules \[[@B22]\] and microorganisms. Methods ======= Cell donors and culture of skeletal muscle cells ------------------------------------------------ Except where specified, chemical reagents, tissue culture media, enzymes and human recombinant growth factors were obtained from Sigma-Aldrich (St. Louis, MO). Plasticware was obtained from Falcon (Becton Dickinson, Milan, IT). Certified double-filtered fetal bovine serum (FBS) produced in a TSE-free country was obtained from HyClone (Logan, UT). Five patients of 41 to 91 years of age gave informed consent for muscle biopsy in the course of programmed heart or orthopedic surgery. Muscle biopsies of at least 1 cm in length were obtained from either the abdominal rectus or femoral quadriceps. Specimens were stripped of visible connective and fat tissue and weighted. Each biopsy provided 0.6 to 1.9 g of muscle tissue. The tissue was immersed in a small aliquot of Hepes-buffered Ham\'s F12 medium supplemented with imipenem, vancomycin and amphotericin B (at the individual concentration of 10 mg/L), minced with sharp scissors to obtain fragments of approximately 1 mm in diameter. Enzymatic digestion was carried out in Ham\'s F12 medium containing 1.5 mg/ml Pronase E (nonspecific protease from Streptomyces griseus) plus 0.03% EDTA at 37°C for 30 min with occasional shaking. After 5 min sedimentation at 1 g, the supernatant was collected and stored at room temperature. Pelleted tissue debris were subjected to 2--3 further cycles of the above digestion process. Collected supernatants were filtered through a 100 μm nylon mesh (Becton Dickinson) and centrifuged at 400 g for 8 min. The pellet was gently resuspended in proliferation medium (PM). Viable cells were counted by the trypan blue exclusion method. PM consisted of a 1:1 mixture of Dulbecco\'s modified Eagle\'s medium and F12 medium (DMEM/F12) containing L-glutamine (2 mM), penicillin (50 units/ml), gentamicin (50 mg/L), and supplemented with: 18% heat-inactivated FBS, gamma-irradiated bovine fetuin (50 μg/ml), recombinant Hu EGF (10 ng/ml), recombinant Hu bFGF (1 ng/ml), recombinant Hu insulin (10 μg/ml), dexamethasone (0.4 μg/ml). Differentiation medium (DM) consisted of DMEM/F12 supplemented with antibiotics, 10% FBS and recombinant Hu insulin (10 μg/ml). Cell suspensions obtained from each biopsy were initially plated in two or more T-75 flasks (i.e., ≥150 square cm) depending on the weight of processed tissue. Cultures were incubated at 37°C in a humidified atmosphere containing 5% CO~2~. Thirty-six hr post-plating, adherent cells were gently washed and the growth medium was changed. Cell passages were performed by light treatment with trypsin-EDTA at the appropriate times indicated by microscopic observation (i.e., 70--80% monolayer confluency). The first passage was carried out 6 to 10 days post-plating. To minimize myogenic differentiation at higher myoblast density, subsequent passages were performed every 3--4 days. Cell proliferation ------------------ Before each passage, cultures were examined with an inverted microscope equipped with a digital camera. Images were acquired with a 10× objective from 8--10 random fields per culture and recorded. Counts of adherent cells were obtained with a computerized image analysis system (Image DB; Amplimedical, Mira, IT) and expressed as number of cells per square millimeter. After trypsinization, numbers of viable cells were counted by the trypan blue exclusion method. The two methods gave concordant results with an agreement of within 12%. Results are expressed as total numbers of viable cells at different times after culture initiation. Immunostaining of muscle cell monolayers ---------------------------------------- Monolayers of cells that had been expanded for 35 to 40 days in PM were trypsinized and plated with PM in two-well chamber slides (Nunc; PBI, Milan, IT). Two-three days later, after gently removing the medium, cell monolayers were fixed with cold 2.5% paraformaldehyde in phosphate-buffered saline (PBS) for 15 min at 4°C, washed with PBS, permeabilized with 0.1% Triton X-100 in PBS for 15 min, washed again with PBS and blocked for 60 min in TRIS-buffered saline (50 mM Tris, 0.15 M NaCl, pH7.6) containing 0.2% BSA, 0.2% FBS, 0.01% Tween 20, 0.01% sodium azide \[immunofluorescence assay (IFA) buffer\]. A mouse monoclonal antibody (mAb) to the intermediate filament desmin (clone D33) that is expressed during muscle development was obtained from DakoCytomation (Milan, IT). mAbs to skeletal myosin heavy chain type I slow (clone NOQ7.5.4D; specific for slow myosin HC of skeletal and cardiac muscle), skeletal myosin heavy chain type II fast (clone MY-32; specific for fast myosin HC of skeletal but not cardiac or smooth muscle), A-sarcomeric actin (clone C5C; specific for alpha-skeletal and alpha-cardiac actin), A-sarcomeric actinin (clone EA-53; specific for alpha-skeletal and alpha-cardiac actinin localized at the Z band) were obtained from Sigma-Aldrich. mAbs to spectrin (clone RBC2/3D5; specific for erythrocytes and muscle was used as a marker of membrane integrity) and to the mid rod domain of dystrophin (clone Dy4/6D3; specific for the skeletal, cardiac and smooth muscle forms of dystrophin that anchors the cytoskeleton to the plasma membrane) were obtained from Novocastra Laboratories (Newcastle upon Tyne, UK). Cell monolayers were washed with IFA buffer and incubated 60 min with primary antibody at room temperature. After three washings with IFA buffer, slides were incubated with FITC-labeled sheep anti-mouse IgG (diluted 1:800 in blue Evans-containing buffer) for 60 min in the dark. After three further washings, slides were mounted with an anti-fade containing medium (Vector Laboratories, Burlingame, CA) and examined with a fluorescence microscope (BX60; Olympus, Tokyo, Japan) equipped with a digital camera (Nikon, Tokyo, Japan). Cells showing cytoplasmic staining were counted at 200--400× magnification. Results are expressed as mean percentage of positive cells over the number of examined cells (at least 500 cells per slide were counted; experiments were run in duplicate). Cytogenetic analysis -------------------- When over 3 × 10^8^viable cells were obtained by in vitro culture (i.e., over day-35 post-plating), an aliquot of each culture was trypsinized and plated in two different chamber slides. Cells were grown in PM, treated with colchicine for 30 min, subjected to hypotonic treatment, fixed in methanol:acetic acid (3:1), and processed according to standard methods \[[@B23]\]. Fifty metaphases were analyzed by QFQ-banding with an automated cytogenetics system (Genikon; Nikon) following the rules of the International System for Human Cytogenetic Nomenclature \[[@B24]\]. Screening of important human viruses ------------------------------------ At the time of muscle biopsy, blood samples from each patient were tested for HBV, HCV, and HIV markers by conventional serology (DiaSorin, Saluggia, IT) and by molecular tests (Cobas Amplicor; Roche Diagnostics, Monza, IT). When over 3 × 10^8^viable cells had been obtained by in vitro expansion, an aliquot of each culture was tested by gene amplification methods for the presence of important human viruses capable of causing persistent infections. DNA was extracted from a frozen 0.5 ml aliquot of PM containing 3 × 10^6^cells using a commercial kit (QIAamp DNA blood mini kit; Qiagen, Milan, IT). PCR was used to detect the genome of DNA human viruses (CMV, EBV, HBV, parvovirus B19). Total RNA was extracted from 3 × 10^6^cells by the guanidinium thiocyanate method (Life Technologies, Gaithersburg, MD). cDNA was obtained from 2 μg of RNA with Mo-MLV reverse transcriptase in conjunction with random hexamer primers (Clontech, Palo Alto, CA). RT-PCR was used to detect HIV and HCV genomes. Routine Cobas Amplicor methods were used for HBV, HCV and HIV. Published primers were used to detect CMV, EBV, and parvovirus B19 genomes \[[@B25]\]. The sensitivity of the employed methods was ≤100 genomic equivalents/reaction. CMV, EBV, and parvovirus B19 amplicons were analyzed on 2% agarose gel using ethidium bromide staining and were photographed on a transilluminator with the help of a digital camera (Kodak Image Station 440CF; Celbio, Pero, IT). Clinical samples positive for EBV, CMV, HBV, Parvovirus B19, HCV, or HIV-1 were used as positive controls. Amplification of beta-globin (PCR) or GAPDH (RT-PCR) was used to normalize the data. Results ======= Growth of primary muscle cell cultures -------------------------------------- As shown in Table [1](#T1){ref-type="table"}, the age of muscle cell donors ranged from 41 to 91 years. Tissue obtained from each biopsy weighted from 0.6 to 1.9 g. The kinetics of cell growth using the described culture conditions is illustrated in Figure [1](#F1){ref-type="fig"}. Thirty-six hr post-plating, the numbers of adherent cells present in primary cultures of the 5 investigated patients ranged from 0.16 to 1.58 millions. Peak cell expansion occurred between day-42 and day-49 of culture. At these times, the total numbers of cultured viable cells ranged from 446 to 1,739 millions. From days 42 to 49 post-plating, the total number of viable cells tended to a plateau. Interestingly, a direct relationship existed between the initial weight of bioptic tissue and the maximal numbers of in vitro expanded cells. This is apparent by comparing tissue weights reported in Table [1](#T1){ref-type="table"} with the kinetic data of Figure [1](#F1){ref-type="fig"}. The average number of cell population doublings that was required to reach peak cell numbers, was 11.24 ± 0.73 (mean ± SD of 5 different primary cultures). Calculated mean cell doubling times of investigated primary cultures ranged from 3.56 to 4.60 days. Overall, the mean cell doubling time was 3.98 ± 0.38 days (mean ± SD, n = 5). From these data, no relationship was apparent between the age of the cell donor and the mean cell doubling time in culture (Table [1](#T1){ref-type="table"}). Under the employed culture conditions, the average fold increase of the number of adherent cells counted on day-1.5 post-plating was 2,490 (range 1,096- to 3,631-fold). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Patients, weight of muscle biopsy, growth parameters of primary cultures. ::: Patient \# (gender) Age (yr) Muscle biopsy site Weight of biopsy specimen (g) Mean cell population doubling time (days) Peak number of viable cells (×10^-6^) obtained at the indicated time of culture (day) ------------------------- -------------- ------------------------ ----------------------------------- ----------------------------------------------- ------------------------------------------------------------------------------------------- 1 (F) 91 Femoral quadriceps 1.0 3.88 1,148 (46) 2 (M) 61 Abdominal rectus 0.6 4.60 446 (49) 3 (M) 41 Abdominal rectus 1.3 3.56 562 (42) 4 (F)^1^ 56 Abdominal rectus 1.9 4.15 1,739 (42) 5 (M) 75 Abdominal rectus 1.8 3.74 1,585 (49) 1\. Patient showing HCV viremia at the time of muscle biopsy (53,000 genome equivalents/ml). ::: ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Growth kinetics of primary skeletal muscle cultured in proliferation medium. At different time points, adherent cells were counted as described in the Materials and Methods section. Kinetics of cultures derived from five different patients are reported: (▲) patient \#1; (■) patient \#2; (□) patient \#3; (○) patient \#4; (●) patient \#5. ::: ![](1479-5876-3-21-1) ::: Morphologic aspects of primary human skeletal muscle cultures are shown in Figure [2](#F2){ref-type="fig"}. The morphology of an adherent SC is shown in Figure [2A](#F2){ref-type="fig"} (day-5 post-plating). Figures [2B](#F2){ref-type="fig"} and [2C](#F2){ref-type="fig"} show cell monolayers cultured in PM for 7 and 10 days, respectively. Experience with muscle cultures from adult donors shows that changing the environmental conditions of cultured myoblasts from PM to DM is followed by cell fusion within a few days. Fusion was only used to confirm the presence of myoblasts. Confluent myoblasts in the process of fusing to form myotubes are shown in Figure [2D](#F2){ref-type="fig"}. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Morphologic aspects of primary human skeletal muscle cultured in vitro. Adherent satellite cell 5 days post-plating in proliferation medium (A; 40×). Semiconfluent monolayers of myoblasts cultured in proliferation medium for 7 (B; 20×) and 10 days (C; 10×). Confluent myoblasts cultured in differentiation medium are in the process of fusing to form myotubes (D; 10×). ::: ![](1479-5876-3-21-2) ::: Expression of differentiation markers by cultured muscle cells -------------------------------------------------------------- The expression of selected skeletal muscle markers was investigated by indirect immunofluorescence in fixed monolayers of primary cultures expanded in vitro for 35 to 40 days. Table [2](#T2){ref-type="table"} shows the average reactivity of primary cultures with different anti-muscle antibodies. The overall positivity with the anti-desmin mAb (an intermediate filament marker that stains muscle cells, but not fibroblastoid cells) was 49.6% (range 40 to 53%), indicating that approximately half of the in vitro-expanded cell population was differentiated along the myogenic lineage. As shown in Table [2](#T2){ref-type="table"} and Figure [3](#F3){ref-type="fig"}, antibodies to the core myofibrillar proteins myosin type I slow and A-sarcomeric actin produced banded filament images and stained approximately one-third of cultured myoblasts. Myosin type II fast (specific for skeletal but not cardiac muscle cells) and A-sarcomeric actinin antibodies stained filaments producing a more continuous pattern in 20--30% of cells. The staining pattern produced by spectrin and dystrophin antibodies was remarkably different since was essentially limited to the periphery of cells and to membrane patches in 30--40% of cells. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Cytoplasmic markers expressed by primary cultures of skeletal muscle cells grown in vitro for 35 to 40 days^1^. ::: Monoclonal antibody Percentage of positive cells (mean ± SD)^2^ Fluorescence pattern ------------------------------------------------------------------------- ------------------------------------------------- -------------------------------------- Desmin (skeletal muscle) 49.6 + 6.4 Cytoplasmic Skeletal myosin type I slow heavy chain (skeletal and cardiac muscle) 33.4 + 3.9 Cytoplasmic filaments A-sarcomeric actin (skeletal and cardiac muscle) 28.6 + 6.1 Cytoplasmic filaments Skeletal myosin type II fast heavy chain (skeletal muscle, not cardiac) 29.4 + 6.2 Cytoplasmic filaments A-sarcomeric actinin (skeletal and cardiac muscle) 19.0 + 3.7 Cytoplasmic filaments Spectrin (erythrocyte and muscle) 32.6 + 7.5 Peripheral rim and membrane staining Dystrophin (skeletal, cardiac and smooth muscle) 38.4 + 9.8 Peripheral rim and membrane staining 1\. Indirect immunofluorescence assay on monolayers of cells grown using proliferation medium. 2\. Positive cells in primary cultures derived from five different tissue donors. ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Indirect immunofluorescence to detect the expression of specific markers in primary muscle cells cultured for 35 to 40 days in proliferation medium. Staining of core myofibrillar proteins \[myosin type I slow (A) and alpha-sarcomeric actin (B)\] produced banded filament images. Staining of myosin type II fast (C) and alpha-sarcomeric actinin (D) produced a continuous filament pattern. The staining pattern produced by spectrin (E) and dystrophin (F) was limited to the periphery of cells and to membrane patches. FITC labeling with Evans blue counterstaining. Original magnification, 60×. ::: ![](1479-5876-3-21-3) ::: Karyotype of cultured muscle cells and screening of human viruses ----------------------------------------------------------------- Cytogenetic analysis and the search for human viruses were performed in muscle cell cultures grown in PM for 35 to 40 days. Normal diploid karyotypes were obtained from muscle cultures of all investigated patients. Figure [4](#F4){ref-type="fig"} shows a representative metaphase and the normal male karyotype of patient \#5. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### \(A) Q banding of a metaphase from the primary muscle cell culture of patient \#5 cultured for 35 days in proliferation medium (original magnification, 100×). (B) Normal diploid male cell karyotype of the same patient. ::: ![](1479-5876-3-21-4) ::: At the time when cultures were subjected to cytogenetic analysis, samples were also processed for detecting human viral pathogens. After DNA and RNA extraction, PCR and RT-PCR methods were used for detecting the genome of CMV, EBV, HBV, parvovirus B19, HIV, and HCV. All cultures gave negative results with virus-specific primers. Amplification for beta-globin (PCR), GAPDH (RT-PCR), and virus-positive controls consistently gave the expected results (data not shown). At the time of muscle biopsy, RT-PCR had shown that patient \#4 was HCV-positive both in blood (53,000 genome equivalents/ml) and in the muscle biopsy specimen. Interestingly, HCV genome was not detected in cultured muscle cells obtained from her biopsy at passage numbers 3, 6, and 9, indicating that HCV failed to replicate in muscle cell cultures. Discussion ========== Currently available tissue culture techniques allow to process skeletal muscle biopsies of patients aged over 40 as to obtain large numbers of isolated (not fused) cells that are differentiated along myogenic lineages. Local myogenesis, neoangiogenesis and homing of progenitor cells from the bone marrow appear to contribute to the repair of acutely infarcted myocardium \[[@B16],[@B26]\]. Implantation of skeletal myoblasts \[[@B27],[@B13],[@B28]\], angiogenetic and protective factors \[[@B29]-[@B31]\], hematopoietic stem cells \[[@B10]\] has been associated with improved contractile function in different animal models of myocardial infarction. Implantation of skeletal myoblasts has also been proposed for nonischemic cardiomyopathy \[[@B17]\]. Cautionary notes on the use of bone-marrow derived stem cells for heart regeneration derive from recent experiments in mice in which autologous hematopoietic cells were not capable of differentiating into cardiomyocytes and gave no functional benefits over sham-treated control animals \[[@B32],[@B33]\]. In humans, phase I clinical studies begin to demonstrate the clinical benefits of autologous myoblast transplantation \[[@B22],[@B12],[@B34],[@B14],[@B15],[@B11]\] and, to a minor extent, of autologous mesenchymal/hematopoietic cells \[[@B16],[@B35],[@B36]\]. In order to make clinical applications possible on a larger scale, conditions for the reproducible and safe in vitro expansion of human skeletal muscle need to be set. To this end, five points are of particular relevance: 1) processing of bioptic tissue as to obtain appropriate muscle stem cells; 2) use of culture media free of non-human components; 3) methods to evaluate the differentiation of cultured cells along myogenic lineages; 4) methods to evidence genetic alterations of cells to be implanted; 5) methods to assure the absence of pathogenic microorganisms (viruses causing persistent infections and TSE agents). In the present study, several of these conditions have been satisfied. Bioptic specimens have been processed with a widely available bacterial protease with pleasing results. Preliminary selection of muscle stem cells by fluorescence-activated sorting and/or antibodies has not been attempted in this study, but seems promising \[[@B37]\]. The medium employed has been supplemented with carefully selected human recombinant growth factors. Addition of FBS and fetuin remained however indispensable. Recently, media supplemented with autologous human serum have been proposed and appear to be associated with superior clinical results \[[@B22]\]. In particular, no malignant arrhythmias were reported among 20 patients and the post-infarctual LV ejection fraction was significantly improved. Myogenic differentiation of human myoblasts obtained with the proposed technique has been comparable to what reported by others. Using desmin as a cytoplastic marker, about 50% of cells cultured from adult donors were differentiated along myogenic lineages. This result is comparable to what previously reported using either the desmin marker \[[@B9]\] or the CD56 surface marker \[[@B15]\]. Detailed characterization showed that expanded muscle cell cultures maintained the ability to express proteins proper of skeletal muscle (myosin type I and II, actin, actinin, spectrin and dystrophin). Since the regulation of myosin heavy chain gene expression is strongly regulated by transcriptional events and by physical exercise \[[@B38]\], and since in old subjects muscle fibers co-expressing myosin type I and myosin type IIA are more frequent than in young subjects \[[@B39]\], immunostaining of in vitro cultured cells should not be expected to strictly reproduce what observed in tissue sections. Of particular interest for clinical applications is that cytogenetic alterations were not detected in cultured cells that had undergone at least 10 population doublings. This was particularly reassuring since the investigated samples were derived from adult/old donors and chromosomal alterations are known to occur at an increased frequency in tissues of adult/old peoples \[[@B40]\]. In future studies, the new CGH technology could offer superior sensitivity for detecting minor cytogenetic changes \[[@B41]\]. Finally, molecular methods for common human viruses are widely available and need to be employed in human clinical trials to validate myoblast cultures prior to implant. Of interest, is the chance observation that HCV failed to replicate in one muscle culture derived from an HCV-infected donor. This result is in agreement with experimental observations showing that HCV has no effects on liver myofibroblasts \[[@B42]\]. Conclusion ========== The results indicate that in about one month it is possible to produce in vitro approximately one billion of adequately differentiated skeletal muscle cells from human donors, independently of age. Clinical experience indicates that approximately 300 millions of autologous skeletal muscle cells are sufficient for the cellular therapy of infarcted heart \[[@B15],[@B14]\]. Thus, an early intervention may be possible by processing a muscle biopsy of about 2 grams. The proposed tissue culture methods may also represent a basis on which to envisage applications in the urologic \[[@B18],[@B19]\] and orthopedic fields \[[@B20],[@B21]\]. List of abbreviations used ========================== BFGF basic fibroblast growth factor DM differentitation medium DMEM/F12 Dulbecco\'s modified Eagle\'s medium plus Ham\'s F12 medium EGF epidermal growth factor FBS fetal bovine serum FITC fluoresceine isothiocyanate HCV hepatitis C virus IFA immunofluorescence assay PM proliferation medium QFQ quinacrine chromosome banding SC satellite cells Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= Design and conception of the study (AQT, AS, PC); Development of methods for tissue culture and virus detection (AQT, AB, AAB); Genetic studies of cultured cells (RB); Manuscript preparation (AQT, AB, AS, PC). All authors have read and approved the final manuscript. Acknowledgements ================ Work supported by: Banca del Monte di Lombardia (BML, Milan, Italy) and Neuroscience Center, University of Insubria (Varese, Italy). A. Baj is a Ph.D. student of the University of Pavia, Italy. A.A. Bettaccini is supported by a fellowship from the Medical School of the University of Insubria (Varese, Italy).	PubMed Central	2024-06-05T03:55:57.428331	2005-5-12	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1142520/", "journal": "J Transl Med. 2005 May 12; 3:21", "authors": [ { "first": "Andreina", "last": "Baj" }, { "first": "Alessia A", "last": "Bettaccini" }, { "first": "Rosario", "last": "Casalone" }, { "first": "Andrea", "last": "Sala" }, { "first": "Paolo", "last": "Cherubino" }, { "first": "Antonio Q", "last": "Toniolo" } ] }
PMC1143555	Introduction ============ The role played by T cell mediated immunity in the control of tumour growth has been established over recent years. As a result, most immunization strategies adopted in clinical trials of cancer treatments have aimed at enhancing tumour antigen (TA)-specific cellular immunity. The induction and expansion of TA-specific T cells requires optimal antigen presentation and T-cell co-stimulation. Dendritic cells (DCs) are specialized antigen-presenting cells with a remarkable ability to stimulate naïve T lymphocytes and generate memory T lymphocytes \[[@B1]\]. However, objective response rates to vaccine or DC trials in cancer remain low \[[@B2]\]. Differentiation and maturation of DCs are important to their protective activity against tumour development \[[@B3]\]. Exposure to necrotic tumor cells can induce maturation of immunostimulatory DCs \[[@B4]\] but the involved mechanisms are still unresolved \[[@B5]\]. Cytotoxic T lymphocytes (CTLs) directed against tumour cells can be amplified in vitrowith the use of DCs pulsed with treated tumour cells (DC-Tu) \[[@B6]\]. When assays were done with cells from healthy donors, DC-Tu stimulation repeatedly increased the final cytolytic activity of T cells more than twofold. However, we observed that a similar procedure applied to cells from cancer patients enhanced the final cytotoxic activity against autologous tumour only in half of the assays \[[@B6]\]. We noticed in these experiments that the final yield and phenotype of blood-derived myeloid immature DCs was heterogeneous in cancer patients \[[@B6]\]. These findings could be related to a relationship between immune suppression instilled during tumour development, as previously described by Kusmartsev and Gabrilovich \[[@B7]\], and increased production of immature myeloid cells in patients with advanced cancers \[[@B8]\]. Our aim in the present study was to detail the differences in characteristics of DCs between patients with cancer and healthy donors. We investigated blood cells from patients with breast, colorectal, or renal carcinoma and compared them, using the same assays, with cells from healthy donors. DCs were obtained from peripheral blood \[[@B9]\] and matured using various cocktails combining proinflammatory cytokines and danger or co-stimulating signals that are known for their ability to induce a T-helper-1 phenotype \[[@B3],[@B10]\]. Tumour cells were from the M74 melanoma cell line in all of the assays. Treatment of tumour cells was done for induction of late apoptosis (postapoptotic necrotic tumour cells) \[[@B11]\]. Necrotic cells were chosen for DC pulsing, in accordance with previous reports \[[@B5],[@B12],[@B13]\] and preliminary experiments by our group that demonstrated that processing and cross-presentation of TA led to specific CTL responses in DCs pulsed under these conditions (Gervais A, unpublished data). The ultimate mechanisms by which DC deficiency is established are not understood. The tumour microenvironment is rich in growth factors and molecules that are able to modulate the immune response of the host. Polyamines, which are conducive to proliferation and metastatic invasion, are synthesized in large amounts by tumour cells \[[@B14]\]. A therapeutic strategy combining inhibition of all cellular and exogenous sources of polyamines has been evaluated in several murine tumour models, with positive findings \[[@B15]\]. However, the role played by polyamines in immune processes is poorly understood \[[@B16],[@B17]\]. Nevertheless, our group showed that polyamine deprivation can prevent the development of in vivotumour-induced immunosuppression \[[@B18]\]. In the present study, the hypothesis that putrescine is involved in immunodeficiency was tested by investigating the effects of putrescine on functional activity of DCs from donors. This treatment was able to mimic the abnormalities observed in DCs from patients with breast cancer. Materials and methods ===================== Patients -------- Thirty-four patients with histologically confirmed cancer were enroled in the study. Seventeen (age 47--76 years) had breast cancer: 13 had infiltrating ductal carcinoma (grade I-III; Elston Ellis grading); one was invasive lobular carcinoma (grade III); one was mixed ductal-lobular carcinoma (grade I); and two were in situ ductal carcinomas (low and high grade). Seven (age 33--86 years) had colorectal cancer (stage 2--4 adenocarcinoma) and 10 (age 42--78 years) had renal cell carcinoma (Fuhrman grade III clear cell carcinoma). Patients were newly diagnosed and peripheral blood samples were collected at the time of initial surgery, with no prior therapy. The study was approved by the regional ethics committee (Comité Consultatif de Protection des Personnes dans la Recherche Biomédicale de Rennes, Rennes, France). Eleven healthy volunteers (HLA-A2) served as control individuals (Etablissement Français du Sang, Rennes, France). Isolation of cells from peripheral blood ---------------------------------------- Cells were centrifuged by applying a density gradient (UNISEP^®^; Novamed, Jerusalem, Israel). Mononuclear cells (MNCs) were frozen in human serum albumin and 10% dimethyl sulphoxide until use for DC and lymphocyte preparation. Tumour cell treatment --------------------- The HLA-A2 MelanA-Mart1 expressing M74 melanoma cell line was used for both antigen DC pulsing and as a target for evaluation of specific cytotoxic activity. This cell line and the K562 natural killer (NK) cell-sensitive erythroleukaemia cell line were maintained in RPMI 1640 medium (Eurobio, Les Ulis, France) containing 10% foetal calf serum (Gibco Invitrogen Life Technologies, Cergy Pontoise, France), 1 mmol/l L-glutamine, 50 μg/ml streptomycin and 50 IU/ml penicillin (ICN Biomedicals, Aurora, OH, USA). M74 cells were used for TA DC pulsing following necrosis-inducing treatment. The method was adapted from that presented by Lennon and coworkers \[[@B11]\]. Briefly, cells were treated with hydrogen peroxide 10 μmol/l (Sigma-Aldrich, Saint Quentin-Fallavier, France) for 3 consecutive days. Supernatant cells were collected each day, pooled and kept at 4°C. Collected cells (M74 per) were used for DC pulsing (DC-Tu~per~). Treated tumour cells were examined for degree of apoptosis and secondary necrosis using a standard fluorescence-activated cell sorting assay (Annexin V-FITC detection kit; Immunotech, Marseille, France), which detects binding of annexin V (A) and propidium iodide inclusion/exclusion. With 10 μmol/l peroxide, 16% of the collected cells were apoptotic (annexin V positive/propidium iodide negative) and 50% were in a state of postapoptotic necrosis (annexin V positive/propidium iodide positive). Culture of dendritic cells -------------------------- DCs were prepared from MNCs (from patients or healthy donors) in accordance with the method described by Sallusto and Lanzavecchia \[[@B9]\]. Briefly, 10 × 10^6^MNCs were seeded in 5 ml serum-free X-Vivo 10 medium (Biowhittaker, Walkersville, MD, USA) in a 25 cm^2^culture flask (Cellstar^®^; Greiner Labortechnik, Frickenhausen, Germany). Nonadherent cells were collected after 2 hours for lymphocyte culture. The remaining adherent cells were cultured in DC medium: serum free X-Vivo 10 medium supplemented with 10% AB serum (EFS de Rennes, Rennes, France), 10 μg/ml steptomycin and 100 IU/ml penicillin. Granulocyte--macrophage colony-stimulating factor (GM-CSF) 1000 IU/ml (Leucomax 400™; Novartis/Shering Plough, Huningue, France, Switzerland) and 400 IU/ml IL-4 (Promokine; Promocell, Heidelberg, Germany) were added on days 0, 2 and 5 of culture. After 7 days nonadherent cells (immature DCs) were collected and added to peroxide-treated M74 cells (ratio 1:10) for antigen processing. After 18 hours of contact, supernatant cells (DC-Tu~per~) were added to autologous lymphocytes (DC:lymphocyte ratio 1:100) for lymphocyte stimulation. DCs were phenotypically characterized before lymphocyte stimulation (day 8). For maturation assays, immature DCs were seeded in DC medium (density 10^6^cells/ml) in 24-well plates (Falcon^®^; Becton Dickinson, Franklin Lakes, NJ, USA) and maturating agents were immediately added. Three different maturation cocktails were evaluated. The first (cocktail A) was a combination of tumour necrosis factor-α (25 ng/ml; Pharmingen, San Diego, CA, USA), lipopolysaccharide (10 μg/ml; Sigma-Aldrich) and CD40L (0.4 μg/ml; Alexis Biochemicals, QBiogene, Illkirch, France). The second (cocktail B) was a combination of IL-1β (10 ng/ml), IL-6 (1000 U/ml), tumour necrosis factor-α (10 ng/ml; R&D systems, Lille, France) and prostaglandin E~2~(1 μg/ml; Sigma-Aldrich). The third (cocktail C) was a combination of bacterial extracts (1 μg/ml; Ribomunyl^®^; Pierre Fabre Médicament, Boulogne, France) and IFN-γ (1000 U/ml; Imukin^®^, Boehringer Ingelheim, Reims, France). After 18 hours of contact, supernatant cells (matured DCs) were collected and phenotypically characterized (day 8). Phagocytic activity of dendritic cells -------------------------------------- Phagocytic activity was evaluated using FITC-labelled opsonized bacteria (Escherichia coli; Phago Test^®^; OrpeGen Pharma, Heidelberg, Germany). Immature DCs from patients or donors and opsonized bacteria were co-cultured for 2 hours at 37°C and their internalization was evaluated by flow cytometry, in accordance with the manufacturer\'s recommendations. Controls were run at +4°C. IL-10 and IL-12 production -------------------------- Measurements of IL-10 and IL-12 were done in supernatants of immature DCs and mature DCs after 72 hours in culture at 37°C. Assays were done using enzyme-linked immunosorbent assay methods according to the manufacturer\'s instructions (Ready-set-Go^®^; eBioscience, San Diego, CA, USA) and were performed in duplicate. Dendritic cell treatments ------------------------- Immature DCs were treated for 18 hours with 10 mmol/l putrescine (1,4-diaminobutane dihydrochloride; Sigma-Aldrich) before phenotype or functional activity were measured. Because it was reported that all-transretinoic acid (ATRA) can reduce the number of immature myeloid cells and favour their differentiation into DCs \[[@B8]\], putrescine-treated immature DCs were added with ATRA (Sigma-Aldrich); 1 μmol/l ATRA was added each day for 5 days, in accordance with the procedure described by Almand and coworkers \[[@B8]\]. Phenotype and functional analyses were conducted after 5 days with or without ATRA treatment. Lymphocyte culture ------------------ Lymphocytes were cultured from MNCs in lymphocyte medium: RPMI 1640 containing 10% AB serum, 1 mmol/l L-glutamine, 2% pyruvate, 1% nonessential amino acids (Bioproducts, Gagny, France), 100 μg/ml streptomycin, 100 IU/ml penicillin and 150 IU/ml IL-2 (Proleukin^®^; Chiron, Suresnes, France). After 8 days in culture (density 10^6^cells/ml), lymphocytes were stimulated with DC-Tu~per~. Number, phenotypic and functional characteristics of lymphocytes were evaluated 7 days after DC-Tu~per~stimulation. Viability was evaluated using the trypan blue exclusion test. Controls were performed with nonstimulated lymphocytes. Flow cytometry analysis ----------------------- Cells (10^5^) were suspended in phosphate-buffered saline supplemented with 0.5% bovine serum albumin and labelled for characterization of lymphocyte or DC phenotype by incubation at 4°C for 30 min with the following PE-, FITC-, or PC5-conjugated antibodies and corresponding isotypes: anti-CD3 (clone UCTH1), anti-CD4 (13B8.2), anti-CD8 (B9 11), anti-CD25 (B1.49.9), anti-CD40 (mAb 89), anti-CD56 (NKH-1), anti-αβ-TCR (BMA 031), anti-γδ-TCR (immu 510), anti-CD80 (MAB 104), anti-CD83 (HB15A) and anti-CD152 (CTLA-4; after saponin permeabilization) from Immunotech (Marseille, France); anti-CD11c (S-HCL-3), anti-HLA-DR (L243) and Lin1 cocktail (anti-CD3, anti-CD14, anti-CD16, anti-CD19, anti-CD20 and anti-CD56) from Becton Dickinson/Pharmingen (CA, USA); and CD86 (BU63) from Immunotech (Oxford, UK). Cells were washed and suspended in 250 μl phosphate-buffered saline added with 0.3% formol. CD4^+^CD25^+^CTLA4^+^was considered to be the T regulatory cell phenotype, in accordance with the findings of Jonuleit and coworkers \[[@B19]\]. Data analysis was performed using a FACScan flow cytometer (Becton Dickinson). Cytotoxicity assays ------------------- T-cell mediated cytotoxicity was tested in triplicate with a standard ^51^Cr release assay. The assays were conducted in U-bottomed microtitre plates. Depending on the assays, target cells were M74 tumour cell line or K562 cells pulsed with ^51^Cr (\[^51^Cr\]sodium chromate; Amersham Life Sciences, Buckinghamshire, England) for 1 hour. A total of 5000 target cells/well were mixed with effector cells (ratio of effector to target cells 50:1) and incubated for 4 hours. Chromium release was assessed in culture supernatants using a γ-counter (Topcount, Packard Instrument, Rungis, France). Specific release was calculated as follows: (\[mean experimental counts/min - mean spontaneous counts/min\]/ \[mean maximum counts/min - mean spontaneous counts/min\]) × 100. IFN-γ production ---------------- Responder cells were evaluated for their production of IFN-γ in response to contact with antigenic cells. Analyses were performed 8 days after stimulation with DC-Tu~per~. Briefly 2 × 10^5^M74 cells were seeded in 24-well plates for 12 hours. The supernatant was discarded before adding 10^5^lymphocytes in a final volume of 500 μl of medium without IL-2. The plates were then incubated at 37°C for 72 hours and IFN-γ was measured in supernatant using enzyme-linked immunosorbent assay methods, in accordance with the manufacturer\'s instructions (Ready-set-Go^®^; eBioscience). Duplicate wells were run for each assay. Statistical analysis -------------------- Each assay was repeated with at least three different donors or patients. The nonparametric Mann--Whitney rank test was used for statistical analysis. Results ======= Yield and characteristics of dendritic cells -------------------------------------------- After 7 days in culture with GM-CSF and IL-4, a mean of 12.5% of the blood MNCs from healthy donors differentiated into immature DCs (Table [1](#T1){ref-type="table"}). These cells were predominantly HLA-DR^+^CD11c^+^(94 ± 4%) and CD11c^+^Lin^-^(87 ± 13 %), which is characteristic of myeloid DCs. The yield of immature DCs was reduced after the same procedure was conducted in MNCs from patients with cancer (Table [1](#T1){ref-type="table"}), and this reduction was significant for MNCs from patients with breast cancer. These patients had normal blood monocyte counts (0.44 ± 0.09 Giga/l). Furthermore, immature DCs prepared from peripheral blood MNCs from breast cancer patients expressed high levels of CD11c (79 ± 14% CD11c^+^Lin^-^), but large individual differences in HLA-DR expression were recorded. The percentages of HLA-DR^+^Lin^-^cells were significantly reduced and HLA-DR^-^Lin^-^significantly increased in patients with breast cancer (Fig. [1](#F1){ref-type="fig"}). This latter observation led us to focus our DC investigations on breast cancer patients. On comparing the expressions of CD40, CD83 and CD86 on immature DCs between healthy donors (n= 6) and patients with breast cancer (n= 10), no significant differences were observed (CD40: 87 ± 7% versus 78.5 ± 16%, respectively; CD83: 6 ± 6% versus 10 ± 10%; and CD86: 82 ± 14% versus 69 ± 22%). For patients with colorectal cancer or renal cell carcinoma, percentages of HLA-DR^-^Lin^-^cells in DCs were not significantly enhanced when compared with those for healthy donors (Fig. [1](#F1){ref-type="fig"}). The phagocytic capacity of immature DCs from breast cancer patients was similar to that of immature DCs from healthy donors (respectively; at 37°C: 42 ± 9% and 44 ± 15%; and at 4°C: 5 ± 1% and 2 ± 1%). Mean fluorescence intensity after 2 hours of co-culture with FITC-labelled bacteria was 653 ± 56 and 656 ± 30 for patients and healthy donors, respectively. Maturation of dendritic cells ----------------------------- For donors and patients, the three maturation cocktails induced significant increases in expression of CD80 and CD83 markers (Table [2](#T2){ref-type="table"}, Fig. [2](#F2){ref-type="fig"}). However, the level of maturation reached by DCs was weaker for patients than for donors; whatever the combination of maturating agents used, we observed lesser expression of mature DC markers in patients (Fig. [2](#F2){ref-type="fig"}). For donors, high percentages of CD86- and CD40-expressing cells were similarly observed in immature and mature DCs (Table [2](#T2){ref-type="table"}). These markers were heterogeneously expressed in patients. IL-10 and IL-12 production by immature DCs was similar in cells from donors and patients (Table [2](#T2){ref-type="table"}). Interestingly, maturation induced by Ribomunyl^®^/Imukin^®^stimulated IL-12 production more for DCs from patients than for DCs from donors (Table [2](#T2){ref-type="table"}). Dendritic cell mediated T-cell stimulation ------------------------------------------ When lymphocytes were subjected to DC-Tu~per~stimulation, expansion was observed. The Expanding Index was not significantly greater in healthy donors (7.5 ± 2; n= 7) than in cancer patients (5 ± 2.5; n= 5). However, contrary to our observations in healthy donors, the cytolytic activity of lymphocytes against the M74 cell line was not significantly enhanced after DC-Tu~per~stimulation for breast cancer patients (Fig. [3](#F3){ref-type="fig"}), which indicates that DC-mediated T-cell stimulation was unsuccessful in the patients. In addition, the basic cytotoxic activity of lymphocytes against the M74 cell line was significantly less for cancer patients than for healthy donors (Fig. [3](#F3){ref-type="fig"}). The differences persisted after DC-Tu~per~stimulation. In contrast, nonspecific lysis of the natural killer (NK) cell sensitive K562 cell line remained unchanged after DC-Tu~per~stimulation both for donors and for patients (respectively: from 51 ± 38% to 51 ± 25% lysis and from 19 ± 18% to 23 ± 23% lysis). Taken together, these observations suggest that TA-specific T cells were induced in donors but not in all of the breast cancer patients. No correlation could be established between reduced cytotoxic activity of lymphocytes from patients (against M74 or NK-sensitive cell lines) and percentage of regulatory T cells in the bulk (0.17 ± 0.14% CD4^+^CD25^+^CTLA4^+^cells; n= 11). Lymphocyte phenotype and IFN-γ production ----------------------------------------- Basic IFN-γ production in response to contact with M74 cells was similarly heterogeneous for lymphocytes from patients and those from donors. After DC-Tu~per~stimulation, enhancement in IFN-γ production -- a marker of T-helper-1 response -- was consistently observed in lymphocytes from healthy donors (Table [3](#T3){ref-type="table"}); this was in contrast to patients, for whom enhancement was seen only in two out of six assays. Considerable reduction in IFN-γ production was seen in lymphocytes from patient S137, indicating that autologous DCs were not immunogenic in the assay (Table [3](#T3){ref-type="table"}). Phenotypic characterization revealed 71% HLA-DR^-^Lin^-^cells in DCs from patient S137. Lymphocytes from donors and cancer patients were of similar phenotype after 15 days in culture with 150 UI IL-2 (Table [4](#T4){ref-type="table"}). Of the cells, 70% were T lymphocytes and more than 50% were CD8^+^T cells. Single stimulation with DC-Tu~per~did not changed the respective percentages of CD4^+^T cell, CD8^+^T cell, or γδ T cell subpopulations. In addition, the percentages of regulatory T cells remained similar after DC-Tu~per~stimulation, at 0--0.5% of cells. Influence of putrescine treatment on dendritic cell phenotype ------------------------------------------------------------- As shown in Fig. [1](#F1){ref-type="fig"}, of cells prepared from MNCs from donors according to the classic procedure for preparing immature DCs, a mean of only 4.6% had the HLA-DR^-^Lin^-^phenotype. An 18-hour treatment with 10 mmol/l putrescine increased this percentage to 29.5% (Fig. [4](#F4){ref-type="fig"}). Expression of other surface markers (CD40, CD80, CD83 and CD86) was not changed by putrescine treatment (data not shown). Putrescine was internalized by DCs because intracellular putrescine concentrations were dramatically enhanced after treatment (data not shown). When DCs were treated daily for 5 days with 1 μmol/l ATRA, the phenotypic change induced by putrescine was reversed (Fig. [4](#F4){ref-type="fig"}). Putrescine-treated dendritic cells are defective in their ability to stimulate T cells -------------------------------------------------------------------------------------- When DCs from donors were treated with putrescine, their ability to stimulate autologous T cells was significantly reduced. [Following the DC-Tu~per~stimulation procedure]{.underline}, the Expanding Index of T cells declined by a mean 30 ± 11% when DCs were treated with putrescine (3.3 ± 1.9 versus 4.6 ± 2.5). In addition, specific cytolytic activity of DC-Tu~per~-stimulated lymphocytes was decreased when DCs were treated with putrescine (Fig. [5](#F5){ref-type="fig"}). This reduction was consistently observed in all donors (n= 6). Treatment with ATRA reversed this putrescine-induced deficiency in DCs and restored cytolytic activity against M74 cells to normal values. A similar increase was repeatedly observed for all donors. These changes were not observed when the K562 target cells were used for nonspecific NK-type cytolytic activity (data not shown). Discussion ========== In recent years several groups have described defective immune function in tumour-bearing animals \[[@B18],[@B20],[@B21]\] and in cancer patients \[[@B8],[@B22],[@B23]\]. Of note, it was reported that factors produced by tumour cells could influence differentiation of DCs from CD34^+^progenitors, and that low concentrations of IL-4 could reverse the inhibitory effect of cancer cell conditioned medium, at least in terms of phenotype and some functional differentiation of DCs \[[@B22]\]. We show here that, even in the presence of IL-4, differences in differentiation of circulating monocytes into DCs persisted in cancer patients as compared with healthy donors. Using the classic procedure of blood monocyte derived DC culture (in the presence of IL-4 and GM-CSF), the ex vivoyield of DCs was found to be significantly reduced in patients with cancer, particularly in those with breast cancer. Furthermore, the phenotype of collected cells using this procedure was different in patients with breast cancer. Expression of MHC class II (HLA-DR^+^Lin^-^cells) was found to be lower and the percentage of HLA-DR^-^Lin^-^to be higher than in donors. In contrast, these subpopulations were not significantly modified in patients with colon or renal cell carcinoma. Whatever combination of maturating agents was used, significantly lower expressions of mature DC markers were observed in patients with breast cancer. Maturation induced by Ribomunyl^®^/Imukin^®^resulted in lower expressions of CD80 and CD86 in patients than in donors, but, interestingly, it also resulted in greater production of IL-12. Other groups have reported that, in breast cancer patients, monocyte-derived DCs have substantially lower level of expression of HLA-DR than do DCs isolated from control donors, leading to a reduced ability to stimulate allogenic and Flu-specific T-cell responses \[[@B8]\]. We confirm here that DCs from such patients not only exhibit low expression of MHC class II but they also have reduced ability to cross-prime exogenous antigens. Stimulation of CTLs by pulsed DCs was less efficient in patients than in donors. In a similar procedure for lymphocyte stimulation, using the same antigen preparation (peroxide-treated tumour cells) and tumour target (M74 cell line), we repeatedly observed defective stimulation when DCs were from patients with breast cancer. In general, the natural cytolytic activity of lymphocytes against the M74 or NK target cell line was found to be lower in patients than in donors. Unlike donors, patients were not selected for their expression of HLA-A2 class I molecules. This could represent an advantage in terms of CTL activation, but the opposite was observed. Cytolytic activity was enhanced by up to 40% when DCs were from donors but only up to 10% when they were from patients. IFN-γ production after DC-Tu stimulation was repeatedly found to be enhanced in donors. In contrast, nonspecific lysis of the NK-sensitive K562 cell line was the same after DC-Tu~per~stimulation, clearly indicating that in donors DC-Tu~per~stimulation induced TA-specific T cells. Induction of TA-specific T cells did not occur in all cancer patients. Blockade of DCs at an early stage in differentiation could be responsible for this inconsistency between patients. For example, in patient S137 a high percentage of HLA-DR^-^Lin^-^cells was observed in DCs, and concomitantly IFN-γ lymphocyte production was reduced twofold after DC-Tu stimulation, indicating that DCs were not only nonimmunogenic but were actually tolerogenic in this patient. However, the percentage of regulatory T cells was not changed after DC-Tu stimulation (\<0.1% in S137). Correlation could not be demonstrated in this study between clinical grade of disease and HLA-DR^-^Lin^-^DC phenotype. Defective function and poor ability of immature DCs to mature in some patients could represent an additional reason why DC cell therapy in cancer patients has, contrary to expectations, not yet yielded significant clinical responses \[[@B2]\]. Defective DC function can be mimicked by adding putrescine to the culture medium of DCs from healthy donors. The percentage of cells with HLA-DR^-^Lin^-^phenotype was found to be enhanced after putrescine treatment. In addition, expansion and final cytolytic activity of lymphocytes was reduced following the DC-Tu~per~stimulation procedure, leading us to conclude that adding putrescine to the microenvironment of antigen-presenting cells blocks their ability to cross-prime exogenous antigens efficiently, indicating a reduction in their immunogenic function. It was reported by other authors that spermine, another polyamine, is responsible for severe inhibition in proinflammatoty cytokine synthesis when added to cultures of human peripheral blood MNCs stimulated with lipopolysaccharide \[[@B24]\]. Our group previously established that polyamine deprivation leads to significant reduction in tumour growth in murine experimental models. Consistent with that effect, an enhancement in CD8^+^T lymphocytes was observed in the spleens of the animals \[[@B18]\]. With similar experimental tumours we observed that combining polyamine deprivation with cyclophosphamide, which is known to downregulate regulatory T cells \[[@B25]\], enhances macrophage tumouricidal activity, indicating that the two treatments have synergistic effects \[[@B26]\]. In addition, breast cancer tissues are characterized by high polyamine levels. In a study including 174 patients with invasive breast cancer \[[@B27]\], a correlation was established between enhancement of putrescine and spermidine levels and tumour aggressiveness. Taken together, these observations led to the conclusion that putrescine release by tumour cells may be involved in the defective DC function observed in breast cancer patients. Interestingly, we showed in the present study that in vitrotreatment of DCs with ATRA could reverse the putrescine-induced deficiency in DC function. ATRA and retinoic derivatives are known to influence DC differentiation, favouring a T-helper-1 response \[[@B28]\]. Further investigations are needed to detail the mechanism underlying the reversal in putrescine-induced deficiency in DC function. Nevertheless, use of ATRA treatment to initiate TA-specific CTL expansion in cancer patients could be of particular interest. Conclusion ========== Taken together, our findings are in agreement with those from Gabrilovitch and coworkers \[[@B7]\] on the contribution of immature myeloid DCs to cancer-induced immunosuppression -- a mechanism that is involved in the escape of tumours from immune system control. Breast cancers are known frequently to over-express several TAs, such as carcinoembryonic antigen, MUC1, HER2/neu, P53 and members of the MAGE family, but little is known about detection of pre-existing T-cell responses, and the rationale for initiating vaccination strategies remains to be fully established. Nevertheless, a phase I clinical trial using vaccine prepared by fusing autologous tumor and DCs (32 patients included) \[[@B29]\] found that two patients with metastatic breast cancer exhibited disease regression. Our opinion is that future vaccination strategies could be improved in view of the present data. Procedures (established with cells from donors) must be adapted to the characteristics of the patient\'s DCs. One simple treatment would be use ATRA to reverse blockade of DC function. The Ribomunyl^®^/Imukin^®^combination has demonstrated ability to induce DC maturation. Abbreviations ============= ATRA = all-transretinoic acid; CTL = cytolytic T lymphocyte; DC = dendritic cell; DC-Tu = DCs pulsed with treated tumour cells; GM-CSF = granulocyte--macrophage colony-stimulating factor; IFN = interferon; IL = interleukin; MHC = major histocompatibility complex; MNC = mononuclear cell; NK = natural killer; TA = tumour antigen. Competing interests =================== The author(s) declare that they have no competing interests. Authors' contributions ====================== AG carried out the DC-Tu preparation, lymphocyte stimulation procedures and measurements of functional activities, and participated in writing the manuscript. JL selected the patients with breast cancer and took biological samples. FB-T participated in the design and coordination of the study. FB carried out anatomo-pathological examinations. TL participated in drawing blood samples in breast patients. LS selected the patients with colonic carcinoma and took biological samples. J-JP selected the patients with renal cell carcinoma and took biological samples. NG carried out cytometric analyses. VC-Q conceived the study, participated in its design and coordination, and wrote the manuscript. Acknowledgements ================ We thank C Thomas de La Pintière for technical assistance. This work was supported by grants from the Comité Grand Ouest de la Ligue Contre le Cancer. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Phenotype of cells collected after immature dendritic cell preparation procedure in cancer patients. Peripheral blood mononuclear cells from healthy donors (n= 11) or patients with colorectal cancer (n= 7), renal cell carcinoma (n= 10), or breast cancer (n= 15) were depleted of lymphocytes (2-hour adherence step) and cultured for 7 days in the presence of granulocyte--macrophage colony-stimulating factor and IL-4. Data are expressed as the percentage of cells (with standard error) expressing the HLA-DR^+^Lin^-^and HLA-DR^-^Lin^-^phenotype. \*P\< 0.01 versus healthy donors. ::: ![](bcr1001-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Maturation of dendritic cells (DCs) from healthy donors or from breast cancer patients. Data are expressed as the percentage of the cells (with standard error) expressing the CD80, CD83 and CD86 surface markers after treatment of immature DCs with a combination of maturating agents: (a)tumour necrosis factor (TNF)-α/lipopolysaccharide (LPS)/CD40L (n= 3); (b)IL-1β/IL-6/TNF-α/prostaglandin (PG)E~2~(n= 4--5); and (c)Ribomunyl^®^/Imukin^®^(n= 3). ^a^Different from corresponding donors in each individual assay. ::: ![](bcr1001-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Cytolytic activity of lymphocytes from healthy donors or breast cancer patients against the M74 cell line. Lymphocytes are from donors (n= 5) or from cancer patients (n= 6), and were stimulated with autologous immature dendritic cells (DCs) pulsed with peroxide-treated M74 cells (DC-Tu~per~). Controls are nonstimulated lymphocytes (NSL). Values are expressed as cytolytic activity (with standard error) against M74 target cell line. \*P\< 0.03 versus NSL; ^†^P\< 0.05 versus corresponding donors. ::: ![](bcr1001-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Effect of putrescine and all-transretinoic acid (ATRA) on immature dendritic cell (DC) phenotype. Cells were collected after immature DC preparation procedure (imm DC; n= 11) and treated with 10 mmol/l putrescine (Put; n= 10). To putrescine-treated DCs was added 1 μmol/l ATRA (Put + ATRA; n= 5). Data are expressed as percentage of cells (with standard error) expressing the HLA-DR^+^Lin^-^and HLA-DR^-^Lin^-^phenotypes. \*P\< 0.01 versus imm DCs; \\P\< 0.02 versus putrescine-treated imm DCs. ::: ![](bcr1001-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Cytolytic activity of lymphocytes stimulated with putrescine and all-transretinoic acid (ATRA) treated dendritic cells (DCs). Immature DCs were from healthy donors and were treated with putrescine (Put) with or without ATRA before DCs pulsed with treated tumour cells (DC-Tu~per~) preparation. Autologous lymphocytes were stimulated with DC-Tu~per~, and data are expressed as cytolytic activity (with standard error) against M74 target cell line. Presented data are from seven different donors. Controls are nonstimulated lymphocytes (NSL). Decrease in M74 lysis was repeatedly observed for each of the donors in DC + Put compared with DC, and increased in DC + Put + ATRA as compared with DC + Put. \*P\< 0.05 versus NSL. ::: ![](bcr1001-5) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Dendritic cell yield in patients with cancer and healthy donors ::: DC source DC yield (%) ------------------------------------------ -------------- Healthy donors (n= 8) 12.5 ± 5.0 Patients with colorectal cancer (n= 4) 5.9 ± 2.6 Patients with breast cancer (n= 6) 3.1 ± 1.2\* Peripheric blood mononuclear cells (PBMCs) were cultured with granulocyte--macrophage colony-stimulating factor and IL-4. Data presented are the dendritic cell (DC) yield after 7 days: number of CD11c^+^Lin^-^cells/number of PBMCs at day 0. DCs were prepared from blood of healthy donors, patients with colorectal cancer, or patients with breast cancer. \*P\< 0.01 versus donors. ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Cell surface phenotype of immature and mature dendritic cells from breast cancer patients ::: Donors/patients Immature DCs/mature DCs ----------------------------- ------------------------- ---------- ---------- Healthy donors Donor M13 N14 015 CD40 79/95 80/97 84/96 CD80 0.1/90 1/97 18/91 CD83 2/52 8/48 19/40 CD86 85/98 68/98 57/97 IL-10 43/46 47/214 89/112 IL-12 5/400 7/29 5/182 Patients with breast cancer Patient S219 S221 S222 CD40 91/93 42/86 77/86 CD80 3/84 4/76 4.5/72 CD83 8/17 14/66 14/72 CD86 87/89 25/86 68/89 IL-10 46/0 57/132 30/63 IL-12 bdl/2154 bdl/1280 bdl/1261 Data are expressed as the percentage of HLA-DR^+^cells expressing CD40, CD80, CD83 and CD86, and IL-10 and IL-12 production by dendritic cells (DCs) before and after maturation with cocktail C: Ribomunyl^®^/Imukin^®^. Data are individual values from patients S219, S221 and S222. Controls are from three different donors (M13, N14, O15). bdl, below the detection limit; DC, dendritic cell. ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Lymphocyte IFN-γ production after DC-Tu~per~stimulation ::: Donors/patients IFN-γ (pg/ml) ------------------------ --------------- ------ Donors C3 516 1581 G7 52 257 H8 17 133 B2 0 8 Breast cancer patients S97 18 420 S101 12 908 S137 339 165 S108 418 176 S126 13 13 S94 13 10 IFN-γ production (pg/ml) was measured in response to tumour cells. Controls were nonstimulated lymphocytes (NSL). Data are individual values from four different donors and six breast cancer patients. DC-Tu, dendritic cells pulsed with treated tumour cells. ::: ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Phenotype of DC-Tu~per~stimulated lymphocytes ::: Phenotype (%) Donor Breast cancer --------------- --------- --------------- --------- --------- CD3^+^CD56^-^ 73 ± 30 76 ± 24 75 ± 17 73 ± 15 CD3^-^CD56^+^ 6 + 8 5 ± 6 10 ± 13 5 ± 5 CD3^+^CD56^+^ 18 ± 21 18 ± 18 11 ± 5 20 ± 13 CD4^+^T cell 28 ± 30 24 ± 27 29 ± 21 31 ± 22 CD8^+^T cell 51 ± 21 57 ± 18 51 ± 19 59 ± 18 TCR α/β 75 ± 23 77 ± 23 70 ± 20 83 ± 12 TCR γ/δ 14 ± 15 13 ± 18 14 ± 11 11 ± 12 Cells are from healthy donors (n= 3) or breast cancer patients (n = 6). The percentage of positive cells for lymphocyte markers was measured in a 99% CD45^+^population. Data are expressed as mean ± standard deviation. DC-Tu, dendritic cells pulsed with treated tumour cells; NSL, nonstimulated lymphocytes. :::	PubMed Central	2024-06-05T03:55:57.430818	2005-2-25	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143555/", "journal": "Breast Cancer Res. 2005 Feb 25; 7(3):R326-R335", "authors": [ { "first": "Alban", "last": "Gervais" }, { "first": "Jean", "last": "Levêque" }, { "first": "Françoise", "last": "Bouet-Toussaint" }, { "first": "Florence", "last": "Burtin" }, { "first": "Thierry", "last": "Lesimple" }, { "first": "Laurent", "last": "Sulpice" }, { "first": "Jean-Jacques", "last": "Patard" }, { "first": "Noelle", "last": "Genetet" }, { "first": "Véronique", "last": "Catros-Quemener" } ] }
PMC1143556	Introduction ============ The proto-oncogene human epidermal growth factor receptor 2 (HER2/neu, also called c-erbB-2) belongs to a family of receptors involved in the tyrosine kinase-mediated regulation of normal breast tissue growth and development \[[@B1]\]. HER2amplification or overexpression is fairly common -- present in 20 to 30% of human breast cancers -- and is a significant predictor of response to therapy, prognosis, and overall survival \[[@B1]\]. HER2is also a target for therapy. Antibody therapy with trastuzumab, which binds the extracellular portion of HER2, has been associated with improved patient outcomes including survival \[[@B2]\]. Because HER2 clearly has an important role in prognosis after a diagnosis of breast cancer, the gene encoding it is a natural target for investigation regarding polymorphisms that might indicate resistance or susceptibility for breast cancer development. One single-nucleotide polymorphism (SNP) at codon 655 indicates a guanine-to-adenine substitution (Ile655Val) in the transmembrane domain-coding region of the HER2gene \[[@B3]\]. This SNP has been evaluated in a variety of populations; studies show that the prevalence of the Val/Val genotype ranges from 3% to 7% in control women \[[@B4]-[@B6]\], although this genotype may be less common or unobserved in people with Asian or African descent \[[@B7]-[@B9]\]. Epidemiologic studies of the association between the Ile655Val polymorphism and breast cancer risk have generally shown null associations, with risk estimates below unity \[[@B4],[@B5],[@B10],[@B11]\] and above unity \[[@B6],[@B8],[@B12]-[@B14]\]. Subgroup analysis in several studies suggested that, among women who were younger \[[@B7],[@B8],[@B14]\], physically inactive \[[@B7]\], had greater body mass \[[@B7]\], or had a positive family history of breast cancer \[[@B6],[@B8]\], the Val/Val genotype was associated with an increased risk of breast cancer in comparison with the Ile/Ile genotype. Further study of this SNP has been supported because of the concern that subgroups of identifiable women might be especially susceptible to breast cancer \[[@B6],[@B10]\]. In the present study we evaluated the association between the HER2Ile655Val polymorphism and breast cancer risk in a population-based case-control study of midwestern United States women. Materials and methods ===================== Study subjects -------------- As part of a continuing epidemiologic study, we recruited population-based cases of incident invasive breast cancer as well as community controls across Wisconsin in accordance with a protocol approved by the University of Wisconsin Health Sciences Human Subjects Committee. Invasive breast cancer cases (excluding carcinoma in situ) aged 20 to 69 years were identified though the Wisconsin statewide tumor registry. Controls were randomly sampled from driver\'s license files (ages 20 to 64 years) and Medicare beneficiary lists (ages 65 to 69 years); controls were frequency-matched in 5-year intervals to have a similar age distribution to that of the cases. All participants were required to have an available telephone number, and controls who self-reported a personal history of breast cancer were not eligible. Before April 2003, when changes in federal law affected the willingness of physicians to acknowledge their care of our eligible participants, physicians (identified on the tumor registry reports) were contacted before case enrollment to obtain information that might contraindicate study participation, such as dementia. All cases and controls were contacted by mail before receiving an interviewer\'s call. The 35-minute structured telephone interview elicited complete reproductive and menstrual histories, exogenous hormone use, smoking history, recent alcohol use and recreational physical activity, lifetime occupational and residential history, and exposure to indoor and outdoor chemicals. Information regarding the women\'s personal and family history of cancer was obtained at the end of the interview to maintain interviewer blinding. During April 2001 to January 2004, 77% of eligible cases (N= 1,884) and 70% of eligible controls (N= 2,146) participated in the telephone interview. The major reasons for nonparticipation were refusal (15% of cases, 23% of controls), death before the interview (2% of cases, 1% of controls), and inability to locate (3% of cases, 6% of controls). Before April 2003, physicians refused participation for 2% of cases. At the conclusion of the telephone interview, all cases and controls were asked to provide a mouthwash rinse. Those agreeing were mailed a kit containing a 44 ml bottle of Scope mouthwash, consent forms, prepaid return mailing supplies, and other all materials needed for producing the sample. During April 2001 to January 2004, samples were obtained from 1,482 cases (79%) and 1,727 controls (81%). Genomic DNA was extracted by using the Gentra Systems DNA extraction reagents and protocol. DNA was resuspended in sterile water. Samples contained an average yield of 29.3 μg of DNA. Genotyping ---------- The laboratory staff were blinded to the identity and disease status of the subjects. Samples were genotyped for the HER2Ile655Val polymorphism with the Applied Biosystems allelic discrimination assay-by-design (no. 185078430). The primers and labeled oligonucleotide probes for this reaction were as follows: forward, 5\'-CCTGACCCTGGCTTCCG-3\' ; reverse, 5\'-ACCAGCAGAATGCCAACCA-3\' ; VIC probe (detects T), 5\'-ACGTCCATCATCTC-3\' ; FAM probe (detects C), 5\'-CCATCGTCTCTGCG-3\'. Samples were cycled with conditions recommended by ABI. Fluorescence was detected with the ABI 7700 and genotypes were called manually with the detection software for this instrument. Genotyping failed for 45 subjects (2%). For quality control, DNA from 79 subjects who had submitted two independent samples were genotyped; 100% (79 of 79) had identical genotypes for the two samples. HER2genotype was obtained for the 1,098 invasive breast cancer cases and 991 controls with European descent who had mailed their mouthwash samples to study staff by 30 June 2003. Because of the small number of women with non-European descent (46 cases, 55 controls) and the low prevalence of the HER2Val/Val genotype in Asian and African populations, these women were not genotyped. Statistical analysis -------------------- Only exposure status before an assigned reference date was used in this analysis. For cases, this was the date of breast cancer diagnosis. For comparability, control subjects were assigned a reference date corresponding to the average time from diagnosis to interview for the case group (about 1 year). The reference age was defined as the age at the reference date. Menopausal status was defined as postmenopausal if the subject reported natural menopause or bilateral oophorectomy before the reference date. Women reporting hysterectomy alone were classified as postmenopausal if their reference age was greater than or equal to the 90th centile of age at natural menopause for the control group (54 years for smokers and 56 years for nonsmokers). Menopausal status was considered to be unknown for women with hysterectomy without bilateral oophorectomy if their reference age was between 42 and 54 years (or 56 years for nonsmokers). Adjusted odds ratios (ORs) and 95% confidence intervals (CIs) were obtained from multivariable conditional logistic regression models stratified on age. Covariates for the models were chosen by forward stepwise regression (P~entry~= 0.20, P~removal~= 0.30). After forward stepwise regression had been performed, covariates remaining in the model were: family history of breast cancer in a mother, daughter, or sister (yes, no, unknown), recent alcohol consumption (four categories), parity (four categories), menopausal status and age at menopause (four categories of age at menopause, premenopausal, unknown), hormone replacement therapy use (never, former, current), age at menarche (five categories of age, plus unknown), height at age 25 years (continuous), weight at age 18 years (continuous) and weight change since age 18 years (five categories). Covariates that did not remain in the final model included age at first birth, education, and income. Women with unknown recent alcohol consumption, hormone replacement therapy use, or height at age 25 years were not included in the analysis (4 cases, 15 controls), so that 1,094 cases and 976 controls remained in the analysis. Interactions with genotype in relation to breast cancer risk were evaluated by including a cross-product term in the regression model and measuring the change in the log-likelihood. Results ======= Breast cancer cases were more likely than controls to report a positive family history of breast cancer, to drink modest amounts of alcohol, to have lower parity, to report menopause at later ages, to have a younger age at menarche, and to report taller adult height (Table [1](#T1){ref-type="table"}). Cases in this HER2analysis were slightly less likely to have non-localized breast cancer at diagnosis; 32% of cases who contributed buccal samples that were included in the HER2analysis had regional or distant-staged disease at diagnosis, whereas 37% of cases who refused to contribute a sample had non-localized disease (N= 356, P= 0.07 by Fisher\'s exact test). Participants in this analysis were similar to nonparticipants in body mass index (P= 0.14 for cases, P= 0.29 for controls by t-test; N= 442 nonparticipant controls) and family history of breast cancer (P= 0.27 for cases, P= 0.36 for controls by Fisher\'s exact test), although control participants were somewhat older (55 versus 53 years, P= 0.02 by t-test) and more likely to have attended college than nonparticipant controls (56% versus 50%, P= 0.05 by Fisher\'s exact test). Among cases, participants in this analysis did not differ significantly from nonparticipants in age (54 versus 53 years, P= 0.73) but were slightly more likely to have attended college (57% versus 52%, P= 0.09). The Ile allele frequency was similar for cases and controls (cases 76.3%, 95% CI 74.5 to 78.1%; controls 74.7%, 95% CI 72.8 to 76.6%), and the Val allele frequency was about 25% (cases 23.7%, 95% CI 21.9 to 25.5%; controls 25.3%, 95% CI 23.4 to 27.2%); 58.2% of cases and 56.5% of controls were homozygous for the Ile allele, 36.2% of cases and 36.5% of controls were heterozygous, and 5.6% of cases and 7.1% of controls were homozygous for the Val allele (Table [2](#T2){ref-type="table"}). Both the case group (P= 0.96) and the control group (P= 0.28) were consistent with Hardy-Weinberg equilibrium. After multivariable adjustment, the combined Ile/Val and Val/Val genotypes were not significantly associated with a risk of breast cancer relative to two copies of the Ile allele (OR 0.90, 95% CI 0.75 to 1.09; Table [2](#T2){ref-type="table"}). The presence of two copies of the Val allele was associated with a 37% reduced risk of breast cancer compared with the Ile/Ile genotype (OR 0.63, 95% CI 0.42 to 0.92). Whereas this inverse association was suggested for cases diagnosed with localized breast cancer (OR 0.69, 95% CI 0.45 to 1.06), the OR was significantly reduced for cases diagnosed with regional or distant metastasis (OR 0.49, 95% CI 0.26 to 0.90). Although no interactions between the HER2polymorphism and common risk factors were statistically significant, the inverse association with breast cancer risk was strongest in some subgroups (Table [3](#T3){ref-type="table"}). In particular, ORs were significantly reduced for women at older ages (more than 55 years), without a family history of breast cancer, with older age at menarche, currently using postmenopausal hormones, with greater recent body mass index, and women with greater weight gain since age 18 years. In addition, we could not find evidence to support heterogeneity in the association between the HER2Ile655Val polymorphism and breast cancer risk according to recent physical activity (P= 0.45), cigarette smoking status (P= 0.66), adult height (P= 0.78), recent alcohol intake (P= 0.83), parity (P= 0.81), or age at menopause (P= 0.41) (data not shown). Discussion ========== We observed a 40 to 50% decreased risk of breast cancer associated with the inheritance of two HER2valine alleles at codon 655 for some subgroups of women, including women older than 55 years of age and women without a family history of breast cancer. Three other studies -- one study of Asian women \[[@B11]\] and two studies of women with European descent \[[@B4],[@B5],[@B10]\] -- have also reported decreased risk estimates of breast cancer associated with inheritance of the HER2Val allele, although the estimates from these three other studies were not statistically significant. Our null results for younger women and women with a positive family history of breast cancer do not concur with findings by Montgomery and colleagues \[[@B14]\], which showed a threefold increased risk among Australian women less than 40 years of age. Wang-Gohrke and Chang-Claude \[[@B6]\] reported a twofold increased risk among German Caucasians with a first-degree family history of breast cancer. Similarly, Millikan and colleagues \[[@B8]\] reported a twofold increased risk of breast cancer associated with the Val/Val or Val/Ile genotype (compared with the Ile/Ile genotype) among women living in North Carolina (United States) who were both less than 45 years of age and reported a positive family history of breast cancer (OR 2.3, 95% CI 1.0 to 5.3). We were limited in our ability to examine the HER2polymorphism in younger women because of small numbers. Only 4 controls and 12 cases in our study were 45 years of age or younger, reported a positive family history of breast cancer, and also had the Val/Val or Val/Ile genotype (OR 1.44, 95% CI 0.21 to 9.79, with Ile/Ile as the reference category; data not shown). The first study of the HER2Ile655Val polymorphism in relation to breast cancer risk found a very high risk (OR 14.1, 95% CI 1.8 to 113.4) of the Val/Val versus Ile/Ile genotype \[[@B7]\]. In that study, the Val/Val genotype was detected in only 11 cases and 1 control. Risk estimates in subsequent studies have been much more modest, ranging from 0.3 to 2.8, and our results clearly fall within this (wide) range. Although risk estimates have suggested both inverse and positive associations with breast cancer risk, prevalence of the Val/Val genotype has consistently been 3 to 8% in breast cancer cases and 3 to 7% in controls in women with European descent. Allele frequencies for case and control women corresponding to the Val/Val genotype in our study are very similar to frequencies reported in three other studies of white women in North Carolina, southeast England, and Germany -- ranging from 23% to 25% -- and slightly higher than frequencies for control women in two other studies conducted in Australia and New York City (18.7% and 16%, respectively) \[[@B5],[@B6],[@B8],[@B13],[@B14]\]. Most studies of the HER2Ile655Val polymorphism have used a case-control design. Only one study population was a prospective cohort \[[@B12]\]. Two other published reports used a kin-cohort approach \[[@B15],[@B16]\]. Using this novel design with a study of 1,560 volunteers living in Washington DC and Israel, Rutter and colleagues \[[@B16]\] reported that the HER2valine allele might be associated with a twofold to eightfold increased risk of breast cancer. As with the Millikan study \[[@B8]\], these increased risks were confined to younger women with a family history of breast cancer. Many studies of the HER2Ile655Val polymorphism had insufficient power to evaluate interactions between the SNP and subgroups according to risk factors such as age and family history of breast cancer. Limited power is a common problem in studies of genetic polymorphisms. Sample size for only one other study was larger than the case and control enrollment in our own study \[[@B8]\]. Prevalence of the Ile655Val polymorphism clearly varies according to racial descent -- it is rare or unobserved in Asian and African populations \[[@B9],[@B17]\] -- further limiting statistical power to evaluate the significance or relevance of this SNP in different populations. Stratified analysis of the HER2Ile655Val genotype according to racial descent is warranted. Potential limitations might have influenced our findings. Although participation in our study was excellent for a population-based case-control study, certain subgroups might have been under-represented because participation probably declines with increasing age, decreasing attained education, and other factors. However, genetic inheritance with the HER2gene is probably not confounded with the variables that might influence a woman\'s participation in our epidemiologic study \[[@B18]\]. The distribution of the HER2polymorphism in our case and control groups was consistent with Hardy-Weinberg equilibrium, which suggests that any genotyping errors were not substantial. Duplicate genotyping of 79 samples was also reassuring, achieving 100% concordance. The mechanism through which this SNP might influence breast cancer risk is unclear, although studies in transgenic mice have demonstrated that activation or overexpression of the HER2gene leads to the development of mammary adenocarcinomas \[[@B19]-[@B21]\]. The transmembrane domain of the HER2 protein might be especially important, given the discovery of an activating mutation in codon 664 in the rat \[[@B22]-[@B25]\]. In humans, the Ile655Val amino acid substitution might alter the formation of active HER2 dimers, which would then alter the activity of the protein \[[@B26]\]. Conclusion ========== These data from our sample population of white women from the midwestern United States suggest that the Val/Val genotype of the HER2Ile655Val polymorphism is associated with a reduced risk of breast cancer in comparison with the Ile/Ile genotype for some women. Although the sample size in our study was relatively large compared with other studies published so far, the inconsistency of the findings across all studies argues against a strong relation with breast cancer risk. Future large studies of the HER2polymorphism might clarify this putative gene-environment interaction. However, given the promise of innovative and more comprehensive approaches to genomic and proteomic studies of breast cancer risk, focusing on this SNP without consideration of the role of other genes and polymorphisms may not be warranted. Abbreviations ============= CI = confidence interval; OR = odds ratio; SNP = single-nucleotide polymorphism. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= SEN performed the processing and genotyping of the samples and drafted the manuscript. MNG and ATD jointly conceived of and designed the study, obtained funding, and drafted the manuscript. JMH performed the statistical analysis and assisted with manuscript preparation. All authors read and approved the final manuscript. Acknowledgements ================ We thank Dr Patrick Remington, Dr Henry Anderson, Dr Polly Newcomb, and Dr Jane McElroy for support throughout this project; Laura Stephenson and the staff of the Wisconsin Cancer Reporting System; Katie Nelson, Jill Haag, Don Wigington, and Yu-Rong Wang for laboratory assistance; Susan Carlson, Lisa Sieczkowski, Emogene Dodsworth, Betty Granda, Liz Mannering, Kathy Peck, Christina Kantor, and Jan Langdon for data collection; and Amy Sapp, Mary Pankratz, Jerry Phipps, Jeff Pearson, and Lene Dotzler for technical support on this project. This project was supported in part by National Cancer Institute grants CA82004, CA28954, and CA77494, Department of Defense grant DAMD17-01-1-0459, and a gift from the Fraternal Order of Eagles Arie no. 1502 to the University of Wisconsin Comprehensive Cancer Center. Figures and Tables ================== ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Characteristics of invasive breast cancer cases and population controls, Wisconsin, 2001 to 2003 ::: Characteristic Cases (N= 1,094) Controls (N= 976) OR^†^ 95% CI^†^ ------------------------------------ -------------------- --------------------- ------- ----------- --------------- ------------ Family history of breast cancer Absent 855 78.2 829 85.3 1 (Reference) Present 231 21.1 132 13.1 1.69 1.33--2.16 Unknown 8 0.7 15 1.7 0.44 0.18--1.05 Recent alcohol consumption None 153 14.0 166 16.8 1 (Reference) 1 drink/week 424 38.8 418 42.5 1.10 0.84--1.45 2--6 drinks/week 385 35.2 270 28.2 1.57 1.19--2.09 7 or more drinks/week 132 12.1 122 12.4 1.23 0.87--1.74 Parity 0--1 262 23.9 217 23.4 1 (Reference) 2 392 35.8 275 29.7 1.22 0.95--1.55 3 253 23.1 209 21.3 0.99 0.76--1.30 4 or more 187 17.1 275 25.6 0.58 0.44--0.77 Menopausal status Postmenopausal 597 54.6 592 55.1 1 (Reference) Premenopausal 415 37.9 313 36.8 1.25 0.92--1.70 Unknown 82 7.5 71 8.1 1.06 0.71--1.58 Age at menopause (years)^‡^ \<45 138 23.1 175 30.1 1 (Reference) 45--49 113 18.9 127 21.4 1.15 0.80--1.65 50--54 205 34.3 172 29.0 1.61 1.16--2.23 55+ 68 11.4 65 10.1 1.42 0.91--2.20 Unknown 73 12.2 53 9.4 1.76 1.14--2.77 Body mass index (kg/m^2^)^‡^ \<22.6 129 21.6 135 23.1 1 (Reference) 22.6--25.2 144 24.1 136 22.8 1.18 0.82--1.69 25.3--28.9 166 27.8 161 26.8 1.09 0.77--1.54 29.0+ 153 25.6 156 26.6 0.97 0.69--1.38 Weight change since age 18 (kg)^‡^ Lost 5 or more 14 2.3 20 3.5 0.92 0.42--2.03 Lost 5 to gained 4 95 15.9 109 18.9 1 (Reference) Gained 5 to 11 148 24.8 140 23.3 1.42 0.97--2.09 Gained 12 to 21 179 30.0 172 28.7 1.30 0.90--1.88 Gained 22 or more 154 25.8 143 24.3 1.32 0.91--1.93 HRT use^‡^ Never 184 30.8 233 37.9 1 (Reference) Former 45 7.5 77 12.6 0.65 0.42--1.01 Current 368 61.6 282 49.4 1.49 1.14--1.95 Age at menarche (years) \< 12 228 20.8 176 18.1 1 (Reference) 12 282 25.8 253 26.4 0.83 0.64--1.09 13 297 27.1 244 24.4 0.94 0.72--1.23 14 183 16.7 172 17.6 0.81 0.61--1.10 15+ 98 9.0 119 12.4 0.63 0.45--0.89 Height at age 25 (m) \< 1.60 198 18.1 218 21.9 1 (Reference) 1.60--1.64 295 27.0 263 27.0 1.25 0.96--1.62 1.65--1.67 299 27.3 264 26.9 1.26 0.97--1.64 1.68+ 302 27.6 231 24.1 1.47 1.12--1.92 \Control percentages are age-adjusted to the distribution of cases; ^†^logistic regression models conditional on age; ^‡^among postmenopausal women. CI, confidence interval; HRT, hormone replacement therapy; OR, odds ratio. ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Risk of invasive breast cancer according to the HER2Ile655Val polymorphism ::: Polymorphism Cases Controls OR\ 95% CI\* OR^†^ 95% CI^†^ ----------------------------------- ------- ---------- ------ ---------- --------------- --------------- ------ ------------ All subjects Ile/Ile 637 58.2 551 56.5 1 (Reference) 1 (Reference) Ile/Val or Val/Val 457 41.8 425 43.5 0.92 0.76--1.10 0.90 0.75--1.09 Ile/Val 396 36.2 356 36.5 0.96 0.79--1.16 0.97 0.79--1.18 Val/Val 61 5.6 69 7.1 0.71 0.49--1.04 0.63 0.42--0.92 Localized disease^‡^ Ile/Ile 425 58.5 551 56.5 1 (Reference) 1 (Reference) Ile/Val or Val/Val 301 41.5 425 43.5 0.91 0.74--1.12 0.90 0.73--1.12 Ile/Val 257 35.4 356 36.5 0.94 0.76--1.16 0.95 0.76--1.19 Val/Val 44 6.1 69 7.1 0.78 0.52--1.18 0.69 0.45--1.06 Regional or distant metastasis^‡^ Ile/Ile 195 57.5 551 56.5 1 (Reference) 1 (Reference) Ile/Val or Val/Val 144 42.5 425 43.5 0.96 0.74--1.25 0.96 0.73--1.27 Ile/Val 128 37.8 356 36.5 1.03 0.79--1.36 1.08 0.81--1.44 Val/Val 16 4.7 69 7.1 0.60 0.34--1.09 0.49 0.26--0.90 \Logistic regression models conditional on age; ^†^logistic regression models conditional on age and adjusted for family history of breast cancer, recent alcohol consumption, parity, menopausal status, age at menopause, hormone replacement therapy use, age at menarche, height at age 25 years, weight at age 18 years, and weight change since age 18 years; ^‡^for cases at diagnosis. CI, confidence interval; OR, odds ratio. ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Risk of invasive breast cancer according to the HER2Ile655Val polymorphism and common risk factors ::: Risk factor\ HER2polymorphism ------------------------------------------ -------------------- ------- --------------- ------------------- ------ Age (years) 0.29 \<55 317/256 32/29 1 (reference) 0.78 (0.44--1.37) 55+ 320/295 29/40 1 (reference) 0.51 (0.29--0.89) Family history of breast cancer 0.24 None 505/461 41/59 1 (reference) 0.54 (0.35--0.84) Any 128/79 19/10 1 (reference) 0.92 (0.32--2.62) Age at menarche (years) 0.14 \<13 288/249 32/35 1 (reference) 0.88 (0.50--1.54) ≥ 13 346/297 29/33 1 (reference) 0.47 (0.27--0.84) HRT use^§^ 0.19 Never/former 149/169 12/10 1 (reference) 1.24 (0.46--3.39) Current 212/163 18/23 1 (reference) 0.46 (0.22--0.97) Recent body mass index^§^ 0.07 \<25.3 kg/m^2^ 159/166 14/10 1 (reference) 1.01 (0.39--2.64) ≥ 25.3 kg/m^2^ 196/163 14/25 1 (reference) 0.43 (0.20--0.91) Weight change since age 18 years (kg)^§^ 0.12 Lost 5 to gained 11 142/148 14/11 1 (reference) 1.04 (0.41--2.67) Gained 12 or more 204/164 14/22 1 (reference) 0.44 (0.20--0.98) \Risk factor cut-points based on the approximate median values for the controls; ^†^logistic regression models conditional on age and, as appropriate, adjusted for family history of breast cancer, recent alcohol consumption, parity, menopausal status, age at menopause, hormone replacement therapy use, age at menarche, height at age 25 years, weight at age 18 years, and weight change since age 18 years; ^‡^P*interaction using the likelihood ratio test and assuming a multiplicative model (risk factors parameterized as dichotomous variables as shown in the table for purposes of the interaction tests); ^§^postmenopausal women only. CI, confidence interval; HRT, hormone replacement therapy; OR, odds ratio. :::	PubMed Central	2024-06-05T03:55:57.433947	2005-3-11	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143556/", "journal": "Breast Cancer Res. 2005 Mar 11; 7(3):R357-R364", "authors": [ { "first": "Stephanie E", "last": "Nelson" }, { "first": "Michael N", "last": "Gould" }, { "first": "John M", "last": "Hampton" }, { "first": "Amy", "last": "Trentham-Dietz" } ] }
PMC1143557	Introduction ============ Malignant tumors contain macrophages (Mps) as a major component of the host leukocytic infiltrate, and the role of Mps in tumor progression has generated contradictory evidence \[[@B1]\]. It has been recognized that Mps can act either as negative regulators by achieving tumor cytotoxicity or as positive regulators by promoting tumor growth. Neovascularization, an essential step in tumor progression and metastasis development, can be modulated by the presence of Mps in the tumor microenvironment. Angiogenic stimuli can proceed from tumor cells and/or immune cells such as lymphocytes and Mps. We have previously demonstrated the ability Mps from tumor-bearing mice to exacerbate the angiogenic response elicited by LMM3 tumor cells (derived from a murine mammary adenocarcinoma), confirmed by CD31 positivity at the angiogenic site \[[@B2]\]. There are several molecules, such as nitrogen metabolites, prostaglandins, vascular endothelial growth factor (VEGF), fibroblast growth factor and placental growth factor, that exert proangiogenic functions \[[@B3]\]. Less knowledge is available about the autonomic regulation of tumor neovascularization. Here we investigate the role of the parasympathetic nervous system on the angiogenic activity exerted by peritoneal Mps from 7-day LMM3 mammary-tumor-bearing mice (TMps) by studying the expression and function of muscarinic acetylcholine receptors (mAchRs) in new blood vessel formation induced by TMps. Materials and methods ===================== Animals and tumor cell line --------------------------- BALB/c mice (females 8 to 12 weeks old) from our Animal Care Division were used. Animal care was provided in accordance with the procedure outlined in the Guide for Care and Use of Laboratory Animals(NIH, 1986 edition). The tumor cell line LMM3 had previously been obtained from a spontaneous syngeneic mammary adenocarcinoma MM3 \[[@B4]\]. LMM3 cells were maintained as monolayers at 37°C in 5% CO~2~in MEM supplemented with 5% FCS. Cells were detached with trypsin; only cell suspensions with more than 90% viability (assessed by Trypan blue) were used. Tumor-bearing mice were obtained by subcutaneous inoculation into the flank of 4 × 10^5^LMM3 cells. Purification of peritoneal macrophages -------------------------------------- Resident peritoneal cells from normal mice and tumor-bearing mice were obtained by washing the peritoneal cavity previously inoculated with 5 ml of MEM supplemented with 10% FCS. The adherent Mps population from normal mice (NMps) and from 7-day tumor-bearing mice (TMps) were purified by adhesion to plastic for 2 hours. After being washed twice with PBS, adherent cells were scraped and resuspended in culture medium. Cell viability was assessed by the Trypan blue exclusion test; only suspensions with more than 95% viability were used. Angiogenesis assay ------------------ Mps and tumor cell-induced angiogenesis was quantified with an in vivobioassay described previously \[[@B5]\]. In brief, tumor cell suspensions were prepared by detaching and washing LMM3 cells twice with fresh MEM. Cell concentration was adjusted to 3 × 10^6^cells/ml and female normal mice were inoculated intradermally in both flanks with 3 × 10^5^LMM3 cells, NMps or TMps in a total volume of 0.1 ml of MEM with a drop of Trypan blue to localize the site of inoculation. Before inoculation, Mps were treated for 1 hour with carbachol (100 nM) in the absence or presence of 1 μM atropine, 1 μM pirenzepine, 1 μM methoctramine, 1 μM 4-diphenylacetoxy-N-methylpiperidine (4-DAMP), 100 μM N^ω^-hydroxy-L-arginine (NOHA), 1 μM indomethacin or 10 μM NS-398. Cells were washed before inoculation. On day 5, animals were killed with ether, the skin was carefully separated from the underlying tissues and the vascular response was observed with a dissecting microscope (Wild) at × 6.4 magnification. The inoculated sites were photographed and the slides were projected on a reticular screen to count the number of vessels per mm^2^of skin. Angiogenesis was quantified as vessel density (δ), calculated as the total number of vessels divided by the total number of squares. Detection of muscarinic acetylcholine receptor subtypes by Western blotting =========================================================================== Purified Mps (10^6^cells) were lysed at 4°C with 0.5 ml of 0.5% Nonidet P40, 10 mM Tris-HCl, pH 7.4, 5 mM MgCl~2~, 50 mM NaCl, 1 mM EDTA, 1 mM EGTA, 50 mM NaF, 0.1 mM orthovanadate and the following protease inhibitors: 10 μg/ml aprotinin, 10 μg/ml leupeptin, 5 mM PMSF and 50 μg/ml soybean trypsin inhibitor. Lysates were sonicated for 30 s and centrifuged at 3,000 r.p.m. for 10 min at 4°C. Supernatants were centrifuged at 10,000 r.p.m. for 20 min at 4°C. The resulting supernatants were stored at -80°C. Protein concentration was determined by the Lowry method \[[@B6]\]. Samples were subjected to 7.5% SDS-PAGE minigel electrophoresis, with 30 μg of protein in each lane. Standards of known molecular masses were also seeded. After electrophoresis, proteins were transferred to a nitrocellulose membrane (Bio-Rad) and washed with distilled water. The nitrocellulose strips were blocked in buffer (20 mM Tris-HCl, 500 mM NaCl, 0.05% Tween 20 (TBST) with 5% skimmed milk) for 1 hour at 20 to -25°C and subsequently incubated overnight with goat anti-M~1~, anti-M~2~and anti-M~3~polyclonal antibodies (Santa Cruz Biotechnology) diluted 1:100 in TBST. After several rinses with TBST, strips were incubated with the second antibody (goat anti-mouse IgG conjugated with alkaline phosphatase, diluted 1:4,000 in TBST) at 37°C for 1 hour. Bands were revealed with a mixture of nitro blue tetrazolium chloride and 5-bromo-4-chloroindol-3-yl phosphate p-toluidine salt (NBT/BCIP) \[[@B7]\]. Quantification of the bands was performed with a computerized densitometer connected to an image analyzer (Bio-Rad GS700) and is expressed in optical density units per mm^2^. Arginase activity assay ======================= Arginase activity was determined in cell lysates in accordance with methods described previously \[[@B8]\]. In brief, 10^5^cells were treated or not with 100 nM carbachol in the absence or presence of 100 μM NOHA, 1 μM atropine, 1 μM pirenzepine, 1 μM methoctramine or 1 μM 4-DAMP. After being washed, cells were lysed with 0.5 ml of 0.1% Triton X-100, 25 mM Tris-HCl, pH 7.4, containing 5 mM MnCl~2~. The enzyme was then activated by being heated at 56°C for 10 min. Arginine hydrolysis was performed by incubating 25 μl of the activated lysate with 25 μl of 0.5 M arginine, pH 9.7, at 37°C for 60 min. The reaction was stopped in acid medium. Urea concentration was measured at 540 nm with a microplate reader. Results are expressed as micromoles of urea per hour per million cells. Detection of arginase isoforms by Western blotting -------------------------------------------------- Mps were rinsed twice with ice-cold PBS and then scraped into 300 μl lysis buffer (50 mM Tris-HCl, pH 7.5, 0.1 mM EDTA, 0.1 mM EGTA, 1 μg/ml leupeptin, 1 μg/ml aprotinin and 0.1 mM PMSF). Lysis was completed by sonication. Samples (25 μg) were subjected to 10% SDS-PAGE as described previously \[[@B9]-[@B11]\]. Nitrocellulose membranes were incubated overnight with a monoclonal anti-mouse arginase I antibody (BD Transduction Laboratories) or with a rabbit anti-arginase II antibody (a gift from Dr Masataka Mori). The secondary antibody anti-mouse or anti-rabbit IgG conjugated with alkaline phosphatase was added for 1 hour at 37°C. Proteins were revealed with NBT/BCIP and quantified by a densitometric analysis. Prostaglandin E~2~assay ----------------------- Prostaglandin E~2~(PGE~2~) production by Mps was determined by RIA as described previously \[[@B12]\]. Purified Mps (2 × 10^6^cells per sample) were incubated for 1 hour at 37°C in a Dubnoff bath with carbogen in 1 ml of MEM with or without 100 nM carbachol in the absence or presence of 1 μM atropine, 1 μM methoctramine or 1 μM 4-DAMP, 1 μM indomethacin or 10 μM NS-398. After incubation, cells were centrifuged for 10 min at 200 gand supernatants were frozen at -80°C until the assay was performed. For PGE~2~RIA, 100 μl samples or standards were incubated for 30 min with 500 μl of rabbit anti-PGE~2~antiserum (Sigma) at 4°C. Then 5 pg of \[^3^H\]PGE~2~(specific radioactivity 154 Ci/mmol; New England Nuclear) was added to each tube. All dilutions were performed in 0.01 M PBS, pH 7.4, containing 0.1% BSA and 0.1% sodium azide. After incubation, a dextran-coated charcoal suspension was added to separate the bound and free fractions. The supernatants were removed from each tube and scintillation solution (Optiphase Hisafe 3; Wallac) was added to determine the amount of radioactivity present. Results are expressed in picograms per 10^5^cells. Detection of cyclo-oxygenase (COX) isoforms by Western blotting --------------------------------------------------------------- Purified Mps were washed twice in cold PBS and then resuspended in 300 μl of lysis buffer (20 mM Tris-HCl, 1 mM EDTA, 10 μg/ml leupeptin, 2 μg/ml aprotinin, 10 μg/ml dithiotreitol, 100 μg/ml soybean trypsin inhibitor, 1 mg/ml benzamidine). After 1 hour, lysates were centrifuged at 5,000 r.p.m. for 10 min. The resulting supernatants were stored at -80°C. Protein concentration was determined by the Lowry method \[[@B6]\]. Samples were subjected to 7.5% SDS-PAGE minigel electrophoresis, with 30 μg of protein in each lane. Standards of known molecular masses were also seeded. After electrophoresis, proteins were transferred to a nitrocellulose membrane (Bio-Rad) at 4°C for 18 hours. Membranes were then washed with distilled water and incubated with blocking solution (5% skimmed milk in TBST) for 1 hour at 20 to -25°C. Membranes were incubated with rabbit polyclonal anti-COX-1 or anti-COX-2 antibodies (Cayman Chemical) in Tris-buffered saline for 90 min at room temperature. Then secondary anti-rabbit IgG antibody conjugated with alkaline phosphatase was added for 1 hour at 37°C. Proteins were revealed with NBT/BCIP and quantified by a densitometric analysis \[[@B13]\]. Detection of VEGF by Western blotting ------------------------------------- Production of VEGF was measured in lysates from untreated TMps or TMps treated with 100 nM carbachol for 1 hour in the absence or presence of 1 μM atropine, 1 μM pirenzepine, 1 μM methoctramine or 1 μM 4-DAMP or the enzyme inhibitors 100 μM NOHA, 1 μM indomethacin or 10 μM NS-398. Cells were then cultured without FCS at 37°C for 24 hours in 100 mm Petri dishes. After being washed twice with cold PBS, TMps were lysed in 10 mM Tris-HCl, pH 8, 1% Triton X-100, 100 mM NaCl, 10 mM EGTA, 10 mM EDTA, with protease inhibitors (1 mM PMSF, 10 μg/ml leupeptin, 10 μg/ml aprotinin). After 1 hour in an ice bath, lysates were centrifuged at 10,000 r.p.m. for 10 min at 4°C. Samples were subjected to 10% SDS-PAGE electrophoresis. Proteins were transferred to nitrocellulose membranes and, after several rinses with doubly distilled water, were blocked with 5% skimmed fat milk in TBST buffer. The primary antibody (goat polyclonal anti-VEGF; Santa Cruz Biotechnology) was added for 18 hours, and the secondary antibody anti-goat IgG conjugated with alkaline phosphatase was added for 1 hour at 37°C. Proteins were detected with NBT/BCIP and quantified by densitometric analysis \[[@B13],[@B14]\]. Drugs ----- All drugs were purchased from Sigma-Aldrich unless otherwise stated. Solutions were prepared fresh daily. Statistics ---------- Results are given as means ± SEM for at least three independent experiments. The statistical significance of differences between groups was analyzed by analysis of variance, Tukey\'s modified t-test or the Mann--Whitney test, using the STAT PRIMER program; P\< 0.05 was considered to be statistically significant. Results ======= Carbachol stimulates angiogenesis induced by TMps ------------------------------------------------- We have reported previously that a small number of TMps (2 × 10^3^to 2 × 10^4^cells) from 7-day LMM3 tumor-bearing mice were unable to induce an angiogenic response in syngeneic mice. The vessel density (δ) of mice inoculated with TMps (δ = 1.70 ± 0.15) was not significantly different from that observed in normal skin (δ = 1.65 ± 0.20). In larger numbers (3 × 10^5^), TMps elicit positive angiogenesis (δ = 2.84 ± 0.06), similar to that observed with the same number of LMM3 tumor cells (δ = 2.91 ± 0.38). The addition of 100 nM carbachol increased the neovascularization induced by TMps by 112% (Table [1](#T1){ref-type="table"}). The participation of mAchR was confirmed by blunting the carbachol action with 1 μM atropine. Peritoneal NMps were unable to induce angiogenesis: the neovascular response (δ = 1.86 ± 0.55) was similar to that observed in normal skin. The addition of carbachol, at the same dose, did not modify the density of blood vessels (δ = 1.80 ± 0.46). We then investigated the participation of mAchR subtypes in neovascularization induced by TMps. As shown in Table [1](#T1){ref-type="table"}, blockade of M~1~or M~3~receptors with 1 μM pirenzepine or 4-DAMP, respectively, completely abolished the stimulatory effect of carbachol on angiogenesis, whereas preincubation with 1 μM methoctramine (an M~2~antagonist) partly prevented the carbachol action. Arginase and COX products are involved in angiogenesis induced by TMps ---------------------------------------------------------------------- Table [2](#T2){ref-type="table"} summarizes results indicating that preincubating TMps with 100 μM NOHA, 1 μM indomethacin or 10 μM NS-398 significantly blunted carbachol-stimulated angiogenesis, demonstrating the participation of arginase and COX in this effect. We have previously reported that arginase I and II were constitutively expressed in Mps, and their expression and activity were upregulated in TMps in comparison with NMps. Our present results indicate that carbachol, at the same dose that triggers neovascularization, increases urea formation (Fig. [1](#F1){ref-type="fig"}). This overproduction was completely reversed by NOHA (100 μM), an enzyme-specific inhibitor of arginase. mAchR activation was also blunted by 1 μM atropine or with 1 μM pirenzepine, indicating a relation between M~1~receptor subtype activation and arginase as its effector enzyme. It is also shown in Fig. [1](#F1){ref-type="fig"} that the M~2~selective antagonist methoctramine partly blunted the action of carbachol. Preincubation of TMps with 4-DAMP did not modify the action of carbachol on urea formation (78 ± 7 μmol/h per 10^6^cells). Western blotting therefore shows the presence of M~1~and M~2~receptor proteins in the membrane-enriched fraction of TMps (Fig. [1](#F1){ref-type="fig"}). Prostaglandins are important mediators in tumor progression because they promote tumor growth and immunosuppress tumor hosts. Here we show that 100 nM carbachol markedly increased the liberation of PGE~2~by TMps (Fig. [2](#F2){ref-type="fig"}). This stimulatory action was inhibited by preincubating cells with 1 μM indomethacin or with 10 μM NS-398, a non-selective COX inhibitor and a COX-2-selective inhibitor, respectively. In addition, we observed that M~3~receptor subtype is involved in carbachol-induced PGE~2~liberation: not only did 1 μM atropine blunt the agonist action, but a 1 μM 4-DAMP blockade was also effective (Fig. [2](#F2){ref-type="fig"}). Neither pirenzepine (2,879 ± 181 pg of PGE~2~/10^6^cells) nor methoctramine (2,799 ± 197 pg PGE~2~/10^6^cells) modified carbachol-induced PGE~2~liberation. We also detected M~3~receptor subtype expression in the TMps membrane-enriched fraction by Western blotting (Fig. [2](#F2){ref-type="fig"}). Participation of VEGF in angiogenesis induced by TMps ----------------------------------------------------- Several growth factors and/or cytokines are involved in tumor angiogenesis. VEGF is the most extensively studied. Here we measured VEGF expression in lysates of Mps. Western blot experiments indicate that TMps express larger amounts of VEGF than NMps do (Fig. [3a](#F3){ref-type="fig"}). In addition, carbachol significantly increases VEGF derived from TMps. Preincubation of cells with 1 μM methoctramine or pirenzepine blocked the action of carbachol on VEGF expression, whereas 1 μM 4-DAMP was ineffective in preventing agonist action. When TMps were treated with 100 μM NOHA, 10 μM NS-398 or 1 μM indomethacin, VEGF protein expression was decreased by almost 30% (Fig. [3b](#F3){ref-type="fig"}). Discussion ========== Mps perform multiple functions that are essential in tissue remodeling, wound healing, inflammation and immunity. These cells form the major component of the mononuclear leukocyte population of some solid tumors \[[@B1],[@B15]\]. In the 1980s, Polverini and Leibovich demonstrated that tumor-associated Mps isolated from 3-methycholanthrene-induced rat fibrosarcoma were potent stimulators of in vivoneovascularization and bovine endothelial cell proliferation; depletion of Mps from tumor cell suspensions significantly decreased their angiogenic potential, suggesting that neovascularization was mediated in part by Mps \[[@B16]\]. Taking into account the fact that murine mammary adenocarcinomas arising spontaneously in BABL/c mice in our laboratory are poorly infiltrated by Mps, we showed that peritoneal Mps from 7-day tumor-bearing mice, when present at low concentrations, contribute to the enhancement of LMM3 angiogenesis by providing polyamine precursors to tumor cells \[[@B2]\]. Although the origin of tumor-infiltrating Mps has been discussed extensively, evidence supports both recruitment from the circulating pool of monocytes and the proliferation of the local Mps population, and it has recently been discussed that Mps could become angiogenic in the presence of diverse stimuli such as growth factors or low oxygen tension as well as soluble tumor antigens \[[@B17],[@B18]\]. Here we show that peritoneal TMps, when inoculated in a number equal to that of LMM3 tumor cells, themselves elicit a potent angiogenic response. In contrast, \'unstimulated\' NMps did not promote angiogenesis in our model. Further investigation is required to determine whether TMps activation occurs in the host-tumor interface or can be triggered at distance by soluble cytokines and/or tumor antigens. Other authors have shown that the levels and functions of lymphocytes, granulocytes, Mps and natural killer cells are under the regulation of the autonomic nervous system \[[@B19]\]. We showed that the activation of mAchR in TMps by the cholinergic agonist carbachol increases their angiogenic ability. The participation of muscarinic receptors was demonstrated by preincubating cells with the non-selective muscarinic antagonist atropine. Angiogenesis is now considered an important step during inflammation and cancer, and it might be necessary as a local, protective response against invasion by pathogens and the proliferation of transformed cells. It is also important in tumor growth and metastasis. The nervous system reflexively regulates the inflammatory response and it has been recently documented that acetylcholine, the principal vagal neurotransmitter, significantly attenuates the release of cytokines (tumor necrosis factor, IL-1, IL-6 and IL-18, but not the anti-inflammatory cytokine IL-10) in lipopolysaccharide-stimulated human macrophage cultures \[[@B20]\]. These anti-inflammatory actions are generally related to nicotinic receptor stimulation \[[@B21]\]. In our model, mAchR stimulation seems to be promoting pro-inflammatory actions by stimulating angiogenesis induced by TMps. It remains to be tested whether the activation of nicotinic receptors in TMps might be exerting anti-inflammatory actions. M~1~, M~2~and M~3~antagonists decreased the carbachol stimulation of neovascularization induced by TMps, showing a collaborative activation of different mAchR subtypes in the neovascular response. It has recently been documented that different mAchR activation controls different functions in distinct systems simultaneously. The activation of M~1~and M~3~receptors by carbachol induces pigment granule dispersion in isolated retinal pigment epithelium from bluegill. Carbachol-induced pigment granule dispersion is blocked by the muscarinic antagonist atropine, by the M~1~antagonist pirenzepine and by the M~3~antagonist 4-DAMP \[[@B22]\]. We also showed that the activation of M~1~, M~2~and M~3~receptors by carbachol is involved in the proliferation of two different murine mammary adenocarcinoma cell lines, LM3 and LM2 \[[@B23]\]. The carbachol stimulation of angiogenesis induced by TMps occurs by the signaling of M~1~to arginase, because pirenzepine totally blocked the carbachol stimulation of urea production. Arginase I and II are overexpressed in TMps in comparison with NMps and are involved in the positive modulation by TMps of angiogenesis induced by LMM3 mammary tumor cells \[[@B2]\]. We were the first to report that carbachol was able to stimulate the proliferation of tumor cells by arginine metabolism through arginase linked to M~1~receptors in LM2 cells, derived from M2 murine mammary adenocarcinoma \[[@B23]\]. We also observed that methoctramine partly blunted carbachol-stimulated vascularization and urea formation, indicating that M~2~receptors are also involved in this effect. We have previously documented a collaborative action of M~2~and M~3~receptor activation by carbachol, which increases amylase secretion in lipopolysaccharide-inflamed salivary glands by stimulating PGE~2~liberation \[[@B7]\]. We are therefore reporting that the expression and function of M~1~and M~2~receptors are involved in the control of angiogenesis induced by TMps, by stimulating polyamine synthesis in these cells. The tumor microenvironment is rich in inflammatory cytokines, growth factors and chemokines, but generally poor in cytokines associated with a sustained immune antitumor response. It is now accepted that tumor-associated Mps produce soluble mediators that contribute to tumor progression. Our results indicate that the parasympathetic nervous system positively modulates neovascularization induced by TMps by stimulating M~3~receptors and PGE~2~liberation. Because indomethacin and NS-398 blunted the carbachol action on PGE~2~synthesis, the COX-1 and COX-2 isoenzymes are involved in angiogenesis induced by TMps. In particular, COX-2 protein expression is highly upregulated in TMps in comparison with NMps (data not shown). Several authors have stated that there is a role of COX-2 expression and function not only in tumors but also in immune cells from the host \[[@B14],[@B18],[@B25]\]. The overproduction of this prostanoid could be responsible for an autocrine loop that also promotes immunosuppression of the host. Previous results indicate that activation of G-protein-coupled receptors encoded by Kaposi\'s sarcoma-associated herpesvirus could be increasing VEGF expression and promoting an angiogenic response that characterizes Kaposi\'s sarcoma lesions \[[@B26]\]. In support of this view, we observed that stimulation of mAchR in TMps by carbachol increased the 45 kDa isoform of VEGF. This effect is linked to activation of the M~2~and M~1~receptors, which in turn promote the arginine metabolic pathway through arginase. We have previously observed that the arginase pathway is involved in the angiogenic response induced by LMM3 cells derived from a murine mammary adenocarcinoma. These cells, which exert a potent angiogenic response quantitatively similar to that induced by TMps, also produce significant amounts of VEGF \[[@B13]\]. Our results show that VEGF production by TMps depends partly on arginase metabolism because NOHA decreases VEGF production. Pretreatment of cells with COX inhibitors also diminished VEGF derived from TMps. In this way, the expression of COX-1 and COX-2 and their product PGE~2~has been shown to be promoters of angiogenesis by modulating the synthesis of various factors, including VEGF \[[@B13],[@B27]\]. It must be taken into account that the stimulation of VEGF expression by COX-derived PGE~2~in TMps is independent of M~3~receptor activation. Conclusion ========== Here we propose a new mechanism involved in angiogenesis induced by peritoneal TMps through mAchR activation sustained by the formation of arginase products and PGE~2~, which could be acting as promoters of the stimulation by VEGF of endothelial cell proliferation, vessel sprouting and organization during tumor progression or metastasis. Abbreviations ============= COX = cyclo-oxygenase; 4-DAMP = 4-diphenylacetoxy-N-methylpiperidine; FCS = fetal calf serum; IL = interleukin; kDa = kilodaltons; mAchR = muscarinic acetylcholine receptor; MEM = minimal essential medium; Mps = macrophages; NBT/BCIP = nitro blue tetrazolium/5-bromo-4-chloroindol-3-yl phosphate; NMps = peritoneal macrophages from normal mice; NOHA = N^ω^-hydroxy-L-arginine; PBS = phosphate-buffered saline; PGE~2~= prostaglandin E~2~; RIA = radio-immunoassay; TBST = Tris-buffered saline containing Tween 20; TMps = peritoneal macrophages from LMM3 murine mammary adenocarcinoma-bearing mice; VEGF = vascular endothelial growth factor. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= E de la T performed the Western blot assays, LD performed the in vivoangiogenesis experiments and the statistical analysis, MAJ made helpful criticism in discussion, TG developed the anti-arginase II antibody, ESL participated in the study design and coordination, and MES performed RIAs and conceived of the study, and participated in its design and coordination. All authors read and approved the final manuscript. Acknowledgements ================ This work was supported by the following grants: PEI 6335 from the National Research Council (CONICET), UBACYT M052 from the University of Buenos Aires and Premio a la Investigación en Bioquímica Molecular y Proteómica, Escuelas ORT, 2004. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Arginase and muscarinic receptors in TMps. Upper panel: macrophages from 7-day LMM3 tumor-bearing mice (TMps) were treated with carbachol (CARB) (100 nM) in the absence or presence of 100 μM N^ω^-hydroxy-L-arginine (NOHA), 1 μM atropine (AT), pirenzepine (PIR) or methoctramine (MET). Arginase activity was measured by urea production as μmol/h per 10^6^cells. Values are means ± SEM for five experiments. \\P\< 0.001, \*P\< 0.05 compared with basal; NS, not significantly different from basal by Tukey\'s modified t-test. Lower panel: Western blot assay to detect arginase isoforms and muscarinic receptors (M) in lysates of TMps. The molecular masses of the bands indicated on the left are coincident with arginase I, II and M~1~and M~2~proteins. One representative experiment of three is shown. ::: ![](bcr1005-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Cyclo-oxygenase (COX) muscarinic receptors in TMps. Upper panel: macrophages from 7-day LMM3 tumor-bearing mice (TMps) were treated with carbachol (CARB) (100 nM) in the absence or presence of 1 μM indomethacin (INDO) 10 μM NS-398, 1 μM atropine (AT) or 4-DAMP. COX activity was measured by prostaglandin E~2~(PGE~2~) liberation by TMps as pg/10^6^cells. Values are means ± SEM for three experiments. \*P\< 0.001 compared with basal; ^\#^P\< 0.001 compared with CARB; NS, not significantly different from basal by Tukey\'s modified t-test. Lower panel: Western blot assay to detect COX isoforms and muscarinic receptors (M) in lysates of TMps. Molecular masses of the bands indicated on the left are coincident with COX-1, COX-2 and M~3~proteins. One representative experiment of three is shown. ::: ![](bcr1005-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Western blot assay to detect vascular endothelial growth factor (VEGF) expression. (a)Expression in normal macrophages (NMps) and macrophages from 7-day LMM3 tumor-bearing mice (TMps). (b)TMps were treated with carbachol (CARB) (100 nM) in the absence or presence of 1 μM pirenzepine (PIR), methoctramine (MET) or 4-DAMP. Lane C, control (TMps without treatment). Bands were quantified in optical density units per square millimeter (OD/mm^2^). The molecular mass indicated at the left corresponds to the VEGF isoform detected. One representative experiment of three is shown. ::: ![](bcr1005-3) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Participation of mAchR subtypes in angiogenesis induced by TMps ::: Treatment Angiogenic response (vessels/mm^2^) n --------------- ------------------------------------- ----- None 2.33 ± 0.07 5 CARB 4.98 ± 0.40\\ 5 AT + CARB 2.57 ± 0.25 5 PIR + CARB 2.49 ± 0.63 4 MET + CARB 3.35 ± 0.17\* 5 4-DAMP + CARB 2.64 ± 0.48 6 Peritoneal macrophages (3 × 10^5^) from 7-day LMM3 tumor-bearing mice (TMps) untreated or stimulated with carbachol (CARB; 100 nM) in the absence or presence of 1 μM atropine (AT), pirenzepine (PIR), methoctramine (MET) or 4-diphenylacetoxy-N-methylpiperidine (4-DAMP) were inoculated intradermally into normal mice to evaluate the angiogenic response. \\P\< 0.0001, \*P\< 0.05 compared with untreated TMps (Mann--Whitney test). n, number of sites evaluated. Results are means ± SEM for three experiments. ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Participation of arginase and COX in angiogenesis induced by TMps ::: Treatment Angiogenic response (vessels/mm^2^) n --------------- ------------------------------------- ----- None 2.34 ± 0.06 5 CARB 4.96 ± 0.41\* 5 NOHA + CARB 2.4 ± 0.6 4 INDO + CARB 1.9 ± 0.92 8 NS-398 + CARB 1.3 ± 0.31 5 Peritoneal macrophages (3 × 10^5^) from 7-day LMM3 tumor-bearing mice (TMps) stimulated with carbachol (CARB; 100 nM) in the absence or presence of N^ω^-hydroxy-L-arginine (NOHA; 100 μM), indomethacin (INDO; 1 μM) or NS-398 (10 μM) were inoculated intradermally into normal mice to evaluate the angiogenic response. \*P\< 0.001 compared with control (without treatment) (Mann--Whitney test). n, number of sites evaluated. Results are means ± SEM for three experiments. :::	PubMed Central	2024-06-05T03:55:57.439362	2005-3-4	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143557/", "journal": "Breast Cancer Res. 2005 Mar 4; 7(3):R345-R352", "authors": [ { "first": "Eulalia", "last": "de la Torre" }, { "first": "Lilia", "last": "Davel" }, { "first": "María A", "last": "Jasnis" }, { "first": "Tomomi", "last": "Gotoh" }, { "first": "Eugenia Sacerdote ", "last": "de Lustig" }, { "first": "María E", "last": "Sales" } ] }
PMC1143558	Introduction ============ Deregulation of cell--cell contact, increased cell motility and invasion of carcinoma cells are key steps in the metastatic cascade \[[@B1]\], but the rate-limiting signalling steps that regulate this process in early-stage breast cancer have not yet been identified. One important molecule is the membrane cytoskeletal crosslinker protein ezrin, a member of the ezrin--radixin--moesin family, which is well documented to regulate several cytoskeletal-related functions, including cell adhesion, cell survival and cell motility \[[@B2]-[@B6]\]. There is also increasing evidence that ezrin regulates tumour progression \[[@B3]\]. Comparison of gene expression profiles in paired metastatic and nonmetastatic tumour cell lines and tissue samples revealed a strong increase in ezrin expression in metastases from rodent mammary and human pancreatic and colorectal carcinomas \[[@B7],[@B8]\]. Likewise, ezrin exhibited strong expression in a variety of invasive human cancers, including osteosarcomas, melanomas, astrocytic tumours, and pancreatic, lung and endometrial carcinomas \[[@B9]-[@B12]\]. Further studies have indicated that suppression of ezrin protein function abrogates pulmonary metastases of murine rhabdomyosarcoma \[[@B13]\] and osteosarcoma cells \[[@B14]\], suggesting that ezrin may be a key regulatory molecule in malignant disease. However, the role played by ezrin in breast cancer metastasis has not been delineated. Ezrin is regulated by an intramolecular association of its amino-terminal and carboxyl-terminal domains that masks their protein--protein binding sites \[[@B2]\]. Unfolding of the molecule into an active conformation occurs following binding to phosphoinositides and phosphorylation on the carboxyl-terminal threonine 567 \[[@B15]\]. The open molecule binds various membrane-associated adhesion molecules and ion exchangers to the amino-terminal region \[[@B2]\], and polymerized F-actin via the carboxyl-terminal domain \[[@B16]\]. Ezrin is involved in signal transduction pathways that depend on tyrosine kinases. Stimulation of cells with epidermal growth factor \[[@B17]\] or hepatocyte growth factor (HGF) \[[@B6]\] results in phosphorylation of ezrin primarily at two tyrosine residues (Tyr145 and Tyr353), which are important in regulating ezrin function. Phosphorylation of ezrin at these two tyrosine residues is required for tubulogenesis and motility \[[@B6]\], and Tyr353 regulates a phosphatidylinositol 3-kinase (PI3K)/Akt-dependent cell survival pathway through association with the p85 subunit of PI3K \[[@B5]\]. Our laboratory developed a mouse mammary carcinoma cell line, SP1, for studies of tumour progression and metastasis \[[@B18]\]. The parent SP1 cells form cadherin-based cell--cell contacts, exhibit oestrogen-dependent primary tumour growth following transplantation in vivo, and are poorly metastatic. Recently, we showed that ezrin acts cooperatively with activated c-Src in deregulating cadherin-based cell--cell contacts and scattering of SP1 cells \[[@B19]\]. We further showed that blocking ezrin function by overexpressing a truncated domain (amino-terminal amino acids 1--309) of ezrin, which has dominant-negative function \[[@B6]\], abrogates cell scattering and enhances cadherin-based cell--cell contacts in SP1 cells \[[@B19]\]. These findings prompted us to examine the role played by ezrin in cell invasion and metastasis of breast carcinoma cells. For this study, we used a highly metastatic variant cell line, namely AC2M2, selected from rare metastatic nodules of SP1 cells in vivo\[[@B18]\]. AC2M2 cells exhibit strong cytoplasmic localization of cadherins and extensive filopodia with weak cell--cell contacts. Our findings show that overexpression of the dominant negative amino-terminal ezrin mutant in AC2M2 cells abrogates in vivometastasis and inhibits cell motility and invasion in vitro. Furthermore, cells overexpressing the amino-terminal ezrin mutant showed marked reduction in PI3K/Akt, Erk1/2 and c-Src activities, indicating a possible role for these signalling molecules as downstream effectors of ezrin in the metastatic process. Materials and methods ===================== Antibodies and reagents ----------------------- Rabbit anti-sheep IgG conjugated with horseradish peroxidase was from Jackson ImmunoResearch Laboratories (West Grove, PA, USA). Mouse (monoclonal) anti-pan cadherin antibody was obtained from Sigma Immunochemicals (Oakville, Ontario, Canada). Alexa-488-conjugated goat anti-mouse IgG was obtained from ICN Biomedicals (Mississauga, Ontario, Canada). Mouse monoclonal antibody against the vesicular stomatitis virus glycoprotein (VSVG; clone P5D4) was obtained from Roche Diagnostics (Mississauga, Ontario, Canada). Rabbit anti-ezrin IgG (carboxyl-terminus specific) was prepared as described previously \[[@B6]\]. Antibodies against the phosphorylated forms of Akt pS473, Erk1/2 pT185/pY187 and c-Src pY418 (i.e. phospho-specific antibodies), and corresponding pan-Akt and pan-Erk1/2 antibodies were obtained from Medicorp (Montreal, Quebec, Canada). Pan c-Src antibody (Ab-1) was obtained from Oncogene Science (Cambridge, MA, USA), and Matrigel was obtained from Becton Dickinson Co. (Mississauga, Ontario, USA). The PI3K inhibitor LY294002, the c-Src inhibitor SU6656 and the mitogen-activated protein kinase (MAPK) kinase (MEK) inhibitor PD098059 were obtained from Calbiochem (San Diego, CA, USA). Cell lines and tissue culture ----------------------------- The SP1 tumour cell line was derived from a spontaneous, poorly metastatic murine mammary intraductal adenocarcinoma, isolated from a female CBA/J retired breeder \[[@B18]\]. AC2M2 cells are a highly metastatic variant selected from the SP1 cell line following three times serial passage of a lung metastatic nodule into the mammary fat pad of syngeneic mice, as described previously \[[@B18]\]. Cell lines were cultured in Dulbecco\'s modified Eagle medium (Invitrogen, Burlington, Ontario, Canada) supplemented with 7% foetal bovine serum. Cell transfection ----------------- The pCB6 vector containing cDNA encoding VSVG-tagged ezrin or the VSVG-tagged amino-terminal truncated domain (amino acids 1--309) of ezrin was previously described \[[@B6]\]. All transfections were carried out with Lipofectamine Plus reagent (Canadian Life Technology, Burlington, Ontario, Canada) in accordance with the manufacturer\'s instructions. Stable transfectants were selected with G418 (450 μg/ml; Sigma-Aldrich, Oakville, Ontario, Canada) and individual clones were isolated. Exogenous protein expression in each clone was confirmed using indirect immunofluorescence (data not shown) and semiquantitative western blot analysis. Indirect immunofluorescence --------------------------- Indirect immunofluorescence staining was conducted as previously described \[[@B19]\]. Briefly, cells were plated overnight on cover slips, fixed in 3% paraformaldehyde/phosphate-buffered saline (PBS), permeabilized with 0.2% Triton X-100 and blocked for 30 min with 3% bovine serum albumin. Cells were incubated with anti-cadherin antibody, followed by the appropriate secondary antibody. Preparations were observed using a Leica TCS SP2 confocal microscope (Leica Microsystems, Richmond Hill, Ontario, Canada) in the Queen\'s Cancer Research Institute and Protein Discovery and Function Facility. Image acquisitions were processed using Adobe Photoshop software. Western blotting ---------------- Cells were grown to 60% confluence in six-well tissue culture plates (NUNC, Mississauga, Ontario, Canada), rinsed with ice-cold PBS with 0.1 μmol/l CaCl~2~and 0.1 μmol/l MgCl~2~(PBS\), and lysed in 2× Laemmli buffer. For blotting with phospho-specific antibodies, cells were serum-starved overnight and plated on fibronectin-coated (10 mg/ml) plates for the times indicated. For analysis of c-Src, the cytoskeletal fraction was first extracted by a 1-min incubation with 250 μl of a Triton X-100 buffer (soluble fraction) that preserves cytosketal-associated material (csk buffer: 50 mmol/l MES, 3 mmol/l EGTA, 5 mmol/l MgCl~2~, 0.5% Triton X-100; pH 6.4). The remaining cellular material (insoluble fraction) was rinsed quickly with 500 μl csk buffer, and was further extracted with 250 μl 2× Laemmli buffer. Protein determination of cell lysates was performed using a DC protein assay kit (Biorad, Mississauga, Ontario, Canada). All cell lysates were subjected to 10% SDS-PAGE under reducing conditions (with 2.5% β~2~-mercaptoethanol) and transferred to PVP (polyvinylpyrrolidone) membranes. The membranes were blocked and probed with the appropriate primary and secondary antibodies, followed by chemiluminescence with the Northern Lightning™ reagent (Perkin Elmer Life Sciences Inc., Boston, MA, USA). Semiquantitation of exogenous versus endogenous ezrin expression was determined by western blotting of serial dilutions of total cell lysates (0.6--20 μg), as described previously \[[@B6],[@B19]\]. The fold increase in exogenous ezrin expression was determined by comparing the titration end-point of the corresponding ezrin band in each clone with that of cells transfected with empty pCB6 vector. For amino-terminal ezrin expressing clones, the ratio of VSVG--amino terminal ezrin to endogenous ezrin normalized to actin was calculated using densitometric analysis. Tumour transplantation and metastasis ------------------------------------- SP1 and AC2M2 cell lines were injected (7.5 × 10^3^cells in 10 μl/mouse) into the mammary fat pad of syngeneic mice, as described previously \[[@B18]\]. Primary tumour growth was monitored every 2--3 days, and metastasis in lung, viscera, liver and draining lymph nodes was assessed 6 weeks later. Histological analysis of tissue sections stained with haematoxylin and eosin was performed to confirm the presence of metastases in the various organs. Based on the gross and histological analyses, animals were assessed as positive or negative with respect to metastasis. At least eight mice were included in each group. Wound healing assay ------------------- Cells were plated onto 12-well tissue culture dishes (NUNC) at near confluence in complete tissue culture medium. Confluent cells were scored using a 20 μl Eppendorf micropipette tip. The medium was immediately replaced, and spontaneous cell migration was monitored using a Nikon inverted microscope for 18--24 hours, as indicated. For experiments comparing transfected clones, analysis at 18 hours is shown because maximal differences between groups were observed at this time. For experiments with pharmacological inhibitors, the assay was allowed to proceed for 24 hours to optimize the effect of the inhibitors. Viability of cells was maintained during this assay period. Phase contrast images were captured, and the distance of wound closure (compared with control at t = 0 hours) was measured in three independent wound sites per group. Relative cell motility was calculated as the wound width at t = 0 hours minus the wound width at t = 18--24 hours, as indicated. Values from at least three independent experiments were pooled and expressed as mean ± standard deviation (SD). Invasion assay -------------- Transfected AC2M2 cells were plated in 24-well transwell cultures (NUNC). Cells (5 × 10^4^) were overlayered in 200 μl of 0.5% foetal bovine serum/Dulbecco\'s modified Eagle medium on Matrigel-coated transwell membranes (8 μm pore size), and with 0.5 ml of complete medium in the lower chamber. After 36--48 hours (as indicated) the cells were fixed and stained with Harris\'s modified haematoxylin (Fisher Scientific, Nepean, Ontario, Canada), and noninvading cells on the top of the membrane were removed using a Q-tip. The membranes were then mounted on glass slides, and images corresponding to the entire membrane surface were captured using an Olympus inverted microscope equipped with a CCD camera (Apogee Instruments Inc., Auburn, CA, USA). The total numbers of cells invading through the membrane were quantitated using ImagePro software (Symbol Technologies, Mississauga, Ontario, Canada). Values were normalized to empty pCB6 vector group in each experiment, and the results from at least three independent experiments were pooled and expressed as mean relative cell invasion ± SD. Statistical analysis -------------------- Statistical significance among metastasis groups was determined using the two-sided Fisher\'s exact test. The day at which tumours reached 1 cm diameter was determined by linear regression analysis of growth curves from individual mice, and expressed as mean ± SD. Statistical significance between groups in the motility and invasion assays was assessed using a Fisher\'s two-tailed t-test with Microsoft Excel software. Results ======= Overexpression of amino-terminal ezrin inhibits metastasis of AC2M2 breast carcinoma cells ------------------------------------------------------------------------------------------ We previously showed that overexpression of a truncated amino-terminal domain of ezrin blocks HGF-induced migration and morphogenesis of epithelial cells \[[@B6]\], and reduces cell scattering in SP1 carcinoma cells expressing activated c-Src \[[@B19]\]. We therefore examined the effect of amino-terminal ezrin on invasion and dissemination of a highly metastatic mammary carcinoma variant cell line, namely AC2M2, which is derived from SP1 cells. We generated stable transfectants of AC2M2 cells expressing VSVG-tagged wild-type and amino-terminal ezrin in a pCB6 eukaryotic expression vector, as described previously \[[@B6]\]. Ezrin protein levels in clones transfected with pCB6 vector containing wild-type ezrin were found to be increased approximately 4-fold and 8-fold, respectively, in WTC4 and WTC6 cells compared with cells transfected with empty vector, as determined by semiquantitative western blotting (Fig. [1a](#F1){ref-type="fig"}). Expression of amino-terminal ezrin was increased 1.6-fold and 4.5 fold, respectively, in NTC6 and NTC7 cells compared with endogenous ezrin and normalized to actin, as determined by densitometric analysis (Fig. [1b](#F1){ref-type="fig"}). AC2M2 cells transfected with empty pCB6 vector, or overexpressing wild-type ezrin exhibited strong cytoplasmic expression of cadherins and filopodia extensions (Fig. [1c--e](#F1){ref-type="fig"}). In contrast, overexpression of amino-terminal ezrin expression caused strong membrane localization of cadherins with increased cell--cell contacts (Fig. [1f,g](#F1){ref-type="fig"}). To assess the role of ezrin function in metastasis, clones of AC2M2 cells overexpressing wild-type ezrin or amino-terminal ezrin were injected into the mammary fat pad of syngeneic female mice, and metastases were assessed 6 weeks after injection (Table [1](#T1){ref-type="table"}). No change in primary tumour growth rate was observed, as assessed by percentage primary tumour take and day of 1 cm tumour diameter, except for one amino-terminal expressing clone (NTC6), which showed reduced tumour growth rate. To compensate, mice in this group were killed approximately 1 week later to allow all tumours to grow to an equivalent size. Untransfected AC2M2 cells exhibited extensive pulmonary metastases (10/11), as compared with the poorly metastatic parent SP1 cells (3/13; P= 0.003). Pooled AC2M2 cells transfected with empty pCB6 vector (7/8) or two clones overexpressing wild-type ezrin (13/15 and 6/7) were also strongly metastatic. In contrast, expression of amino-terminal ezrin caused a marked reduction in metastases in two independent clones (0/8 and 3/8; P\< 0.0001 and P= 0.002, respectively). Similar results were obtained with an additional amino-terminal ezrin overexpressing clone (NTB8) from an independent transfection (0/5; P= 0.002). Analysis of pooled results showed that metastases in the three amino-terminal ezrin groups (3/21) were strongly reduced compared with the two wild-type ezrin groups (19/22; P\< 0.0001). Histological analysis of various organ sites in animals with tumours transfected with empty pCB6 vector or wild-type ezrin (WTC4, WTC6) revealed massive tumour nodules in the lung (Fig. [2a,b](#F2){ref-type="fig"}), as well as occasional metastases in the small intestine (data not shown). In contrast, the majority of mice injected with tumour cells overexpressing amino-terminal ezrin (NTC6, NTC7, NTB8) showed no metastatic lesions; the few metastases that did form (in the NTC7 group) were generally smaller and primarily localized to vascular channels (Fig. [2c,d](#F2){ref-type="fig"}). These findings suggest that ezrin function is necessary for metastasis in this breast cancer model. Overexpression of amino-terminal ezrin inhibits cell motility and invasion of AC2M2 cells ----------------------------------------------------------------------------------------- Metastasis is a multistep process involving intravasation, transport through the vasculature or lymphatics, and extravasation into target organs \[[@B20]\]. Previous studies indicated a role for ezrin in HGF-induced cell scattering and migration \[[@B6],[@B19]\]. We therefore examined the role played by ezrin in cell motility and invasion of metastatic AC2M2 cells. Wound healing assays were conducted using AC2M2 cells transfected with empty pCB6 vector or a vector encoding wild-type ezrin or amino-terminal ezrin. A wound was scored on a cell monolayer, and the wound closure was assessed at various times up to 24 hours. Our results show that expression of amino-terminal ezrin reduced the ability of AC2M2 cells to close the wound by approximately 2.5-fold compared with cells transfected with empty pCB6 vector or with wild-type ezrin (Fig. [3a](#F3){ref-type="fig"}). Invasion assays were carried out using Matrigel-coated transwell culture chambers, and invading cells were counted after 36--48 hours using image analysis. AC2M2 cells expressing wild-type ezrin showed increased cell invasion compared with cells transfected with empty pCB6 vector, whereas amino-terminal ezrin expressing clones exhibited markedly reduced cell invasion (Fig. [3b](#F3){ref-type="fig"}). PI3K and c-Src are required for ezrin-mediated cell motility and invasion of AC2M2 cells ---------------------------------------------------------------------------------------- PI3K, c-Src and MAPK pathways have been implicated in cell motility and invasion in many cell types \[[@B21]\]. As a first step in unravelling the signalling pathways involved in cell motility and invasion of AC2M2 cells overexpressing wild-type ezrin, we determined the effect of specific signal transduction inhibitors on these functions. The results showed that the PI3K inhibitor LY294002 markedly attenuated cell motility (3-fold) of two clones overexpressing wild-type ezrin as well as cells transfected with empty pCB6 vector (data not shown; Fig. [4a,b](#F4){ref-type="fig"}). The c-Src inhibitor SU6656 had a moderate (1.5-fold) blocking effect on cell motility. In contrast, the MEK inhibitor PD098059 had no detectable effect. In addition, cell invasion was dramatically inhibited by both PI3K and c-Src inhibitors, but not by the MEK inhibitor (Fig. [4c](#F4){ref-type="fig"}). All three inhibitors at the concentrations indicated were previously shown to block activity of the respective kinases, as determined by western blotting with the corresponding phospho-specific antibodies \[[@B22]\] (data not shown). Thus, PI3K and c-Src pathways, but not the MAPK pathway, are required for both cell motility and invasion of wild-type ezrin-expressing AC2M2 cells. Overexpression of amino-terminal ezrin abrogates signalling through PI3K/Akt, c-Src, and MAPK pathways in AC2M2 cells --------------------------------------------------------------------------------------------------------------------- The results shown in Fig. [4a--c](#F4){ref-type="fig"} raise the possibility that PI3K and c-Src are downstream of ezrin in the regulation of cell motility and invasion in AC2M2 cells. To investigate this notion, we examined the effect of amino-terminal ezrin on phosphorylation of Akt S473 (a downstream effector of PI3K \[[@B5]\]), c-Src Y418 (within the c-Src catalytic domain \[[@B1]\]) and Erk1/2 T185/Y187 (within the activation loop of Erk1/2; Fig. [4d](#F4){ref-type="fig"}). Serum-starved AC2M2 cells were plated on fibronectin for the times indicated, lysed, and subjected to western blotting with the appropriate phospho-specific antibodies. Interestingly, expression of Akt pS473 was increased in metastatic AC2M2 cells (empty vector) compared with the poorly metastatic parental SP1 cells. Overexpression of amino-terminal ezrin, compared with wild-type ezrin, markedly reduced the level of Akt pS473 in AC2M2 cells, indicating regulation by ezrin of the PI3K/Akt pathway in these cells. In parallel, the level of phospho-Erk1/2 (pT185/pY187) was sustained in AC2M2 cells transfected with empty pCB6 vector or wild-type ezrin, and was reduced in cells expressing amino-terminal ezrin. Because activated c-Src associates with its substrate at the focal adhesion complex \[[@B1],[@B23]\], we examined c-Src pY418 in both the Triton X-100 soluble and insoluble (cytoskeletal-associated) fractions in AC2M2 cells (Fig. [4e](#F4){ref-type="fig"}). Our results show that the levels of cytoskeletal-associated total c-Src and c-Src pY418 were increased in cells overexpressing wild-type ezrin, but were markedly reduced in cells expressing amino-terminal ezrin. In contrast, the level of c-Src pY418 in the soluble fraction (and in total cell lysates; data not shown) remained unchanged in all cell groups. Thus, ezrin plays a key role in stabilizing the activities of PI3K and c-Src, as well as Erk1/2. Discussion ========== In the present study we demonstrate for the first time that ezrin function is required for metastasis of breast carcinoma cells. Our results show that inactivating ezrin function by overexpressing a dominant-negative (amino-terminal) ezrin mutant blocks spontaneous pulmonary metastases of mammary carcinoma cells transplanted into the orthotopic site. We further show that overexpression of wild-type ezrin increases carcinoma cell invasion, whereas amino-terminal ezrin causes reduced cell scattering, motility and invasion, thus indicating a possible mechanism by which ezrin regulates progression to invasive cancer. Similar reports have shown that overexpressing ezrin antisense \[[@B13]\] or an ezrin T567A dominant-negative mutant \[[@B14]\] blocks both experimental and spontaneous metastasis of murine rhabdomyosarcoma and osteosarcoma cells, and in the latter report the rate-limiting effect was demonstrated to be on early survival of metastastic cells. Thus, ezrin may have multiple effects on the metastatic cascade. Moreover, we found that overexpression of wild-type ezrin does not augment metastasis of parental SP1 cells. Furthermore, no increase in expression of endogenous ezrin was observed in the metastatic (AC2M2) compared with the parent SP1 cell lines (data not shown). These findings imply that overexpression of ezrin alone is not sufficient to induce metastasis in this tumour model, suggesting that multiple pathways are involved in the metastatic cascade. However, it is difficult to relate quantitative changes in exogenous ezrin overexpression directly with dominant active or negative functional effects, because the signalling networks involved are complex and the functional assays are long term (18--36 hours for cell motility and invasion, and 5 weeks for metastasis). Our focus was therefore on the qualitative effects on breast cancer metastasis of blocking ezrin function using a dominant-negative amino-terminal ezrin mutant \[[@B6]\]. Our finding that Akt S473 phosphorylation is enhanced in AC2M2 cells compared with parental SP1 cells suggests a key role for the PI3K/Akt pathway in metastasis. Expression of amino-terminal ezrin reduced the levels of Akt S473 phosphorylation to that of SP1 cells, indicating a dominant regulatory effect of ezrin on PI3K/Akt signalling in AC2M2 cells. Furthermore, inhibition of PI3K blocked both cell motility and invasion in AC2M2 cells overexpressing wild-type ezrin, indicating that PI3K is a downstream effector of ezrin in these functions. In addition to its role in cell motility via PI3K, ezrin may also participate in metastasis by increasing cell survival. Indeed, we previously showed that ezrin signals cell survival by activating the PI3K/Akt pathway \[[@B5]\]. Interestingly, although Erk1/2 activation is also reduced in amino-terminal ezrin-expressing cells, inhibition of the MAPK pathway has no detectable effect on cell motility or invasion in this tumour model. However, previous reports have indicated that an activated MEK mutant can rescue early survival of metatastic osteosarcoma cells expressing ezrin antisense \[[@B14]\]. It is therefore possible that an ezrin-dependent MAPK pathway still plays a role in our breast metastasis model, as demonstrated by Khanna and coworkers \[[@B14]\]. We also observed a strong increase in cytoskeletal-associated c-Src pY418 in cells overexpressing wild-type ezrin, and this effect was abrogated in cells expressing amino-terminal ezrin. Furthermore, inhibition of c-Src activity partially blocks cell motility and completely abrogates invasion of ezrin-expressing AC2M2 cells. These findings are consistent with our previous demonstration \[[@B24]\] of a reciprocal relationship between c-Src and ezrin in phosphorylation/activation of these two proteins, and their role in regulating cell spreading and cell migration. The interactive role played by ezrin with c-Src in cell adhesion-dependent functions may provide an important mechanism by which integrin signals are amplified through the cytoskeleton. Previous findings from our laboratory \[[@B25]\] and others \[[@B26]\] have shown that inhibition of specific integrin function can block metastasis of breast carcinomas. The findings presented here raise the possibility that ezrin and c-Src are key regulators of integrin-dependent steps in cell invasion and metastasis. Because both PI3K and c-Src are key effectors downstream of ezrin in the cell motility and invasive phenotypes, these signalling pathways are likely to be rate limiting in the regulation by ezrin of metastatic progression in vivo. In addition, cooperativity between PI3K and c-Src may be important in regulating ezrin function in cell motility and invasion, for example through activation of Rho GTPases \[[@B4],[@B13],[@B21]\]. In addition, interaction of ezrin with other signalling molecules such as the Na^+^/H^+^exchanger regulatory factor (NHERF-1), recently described to be altered in breast cancer, may also be involved \[[@B27],[@B28]\]. Further investigation is required to assess the relevance of these downstream pathways in breast metastasis. Conclusion ========== In the present study we show for the first time that ezrin is required for invasion and metastasis of mammary carcinoma cells. We further show that PI3K and c-Src activities are modulated by ezrin and are required for ezrin-dependent cell invasion. Because we recently showed that c-Src is also upstream of ezrin \[[@B24]\], and acts cooperatively with ezrin in deregulating cell--cell contacts and cell scattering \[[@B19]\], we propose that coordinate upregulation of ezrin and c-Src activity may be a key regulatory step in metastatic breast disease. Together, our findings suggest that ezrin activation may represent an effective prognostic marker and a potential target for treatment of invasion and metastasis of human breast cancer. Abbreviations ============= HGF = hepatocyte growth factor; MAPK = mitogen-activated protein kinase; MEK = mitogen-activated protein kinase kinase; PBS = phosphate-buffered saline; PI3K = phosphatidylinositol 3-kinase; SD = standard deviation; VSVG = vesicular stomatitis virus glycoprotein. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= BEE carried out the tumour transplantation and metastasis studies and the cell motility, and wrote the manuscript. JAM performed invasion assays and western blotting studies, and assisted with the tumour transplantation studies. SKS performed the pathology on tissue sections. MA and DL participated in the design of the study, provided the wild-type and amino-terminal ezrin pCB6 expression vectors, and assisted in writing and editing the manuscript. All authors read and approved the final manuscript. Acknowledgements ================ We thank Jeff Mewburn and Eric Tremblay for technical assistance. Dr Jim Gerlach provided assistance with statistical analysis. This work was supported by funds from the Canadian Breast Cancer Research Alliance grant \#14315 (to BEE), the Canadian Institutes of Health Research grant \#3642 (to BEE), Cancer Care Ontario (to BEE), la Ligue Nationale contre le cancer (to MA) and l\'Association pour la Recherche sur le Cancer grant \#4601 (to MA). Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Localization of cadherins in metastatic mammary carcinoma cells overexpressing wild-type and amino-terminal ezrin. Metastatic AC2M2 cells were transfected with empty pCB6 vector, or a vector encoding wild-type or amino-terminal ezrin, as described in the text. (a)Serial dilutions of the total cell extracts (0.6--20 μg) were subjected to reduced 10% SDS-PAGE and transferred to PVP membranes. The membranes were probed with anti-ezrin and anti-actin antibodies, followed by the appropriate peroxidase-conjugated secondary antibodies, and developed with chemiluminescence. Lanes from left to right contained 10 μg of the following cell extracts: pooled AC2M2 cells transfected with empty pCB6 vector, and two clones transfected with wild-type (WT) ezrin. WTC4 and WTC6 exhibited 4-fold and 8-fold overexpression, respectively, of ezrin compared with vector control cells. (b)Membranes were probed with anti-vesicular stomatitis virus glycoprotein (VSVG), anti-ezrin and anti-actin antibodies. Lanes contained 15 μg of cell extracts from pooled pCB6-transfected cells, and two clones transfected with amino-terminal ezrin. Clones NTC6 and NTC7 exhibited 1.6-fold and 4.6-fold amino-terminal ezrin expression, respectively, compared with endogenous ezrin, as determined by densitometric analysis of VSVG and ezrin blots normalized to actin. (c-g)The above cell lines were immunostained with anti-pan cadherin antibody, as described the text. Representative confocal microscope images are shown. ::: ![](bcr1006-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Expression of amino-terminal ezrin significantly reduces lung metastases of carcinoma cells. Metastatic AC2M2 cells transfected with empty pCB6 vector or a vector encoding wild-type ezrin or amino-terminal ezrin were injected (7.5 × 10^3^/mouse) into the mammary fat pad of syngeneic mice, and metastasis was assessed 6 weeks later, as described in the text. Images are shown from tissue sections from the lungs of mice injected with AC2M2 cells transfected with (a)empty pCB6 vector, (b)wild-type ezrin (WTC4), or (c,d)amino-terminal ezrin, stained with haematoxylin and eosin. Normal lung tissue is indicated by \'NL\'. Tumour metastases are indicated by \'T\'. Arrows indicate endothelial lining of vascular channels with tumour emboli. Original magnifications: panels a--c, 200×; panel d, 400×. ::: ![](bcr1006-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Expression of amino-terminal ezrin inhibits motility and invasion of carcinoma cells. (a)Transfected AC2M2 clones (see Fig. 1) were grown to confluence in 12-well NUNC tissue culture plates with 2 ml of 10% foetal bovine serum (FBS)/Dulbecco\'s modified Eagle medium (DMEM). Cultures were wounded by streaking cell monolayers with a 20 μl Eppendorf micropipette tip and monitored up to 24 hours, as described in the text. Representative fields photographed after 18 hours are shown. Vertical bars indicate the wound distance. Bar, 50 μm. (b)Transfected AC2M2 cells were subjected to an invasion assay, as described in the text. Cells (5 × 10^4^) were over-layered in 200 μl of 0.5% FBS/DMEM medium on the transwell membranes (8 μm pore size), with 0.5 ml of complete medium in the lower chamber. After 36--48 hours, cells were fixed and stained with modified haematoxylin, and cells invading through the membrane were counted using ImagePro software, as described in the text. The numbers of invading cells were normalized to empty pCB6 vector group in each experiment. The results from at least three independent experiments were pooled, and expressed as the mean relative cell invasion ± standard deviation. Asterisks indicate (\\) a significant increase or (\) a significant reduction in cell invasion compared with vector control, using a two-sided Fisher\'s t-test. P= 0.03 for WTC4\\; P= 0.009 for NTC6\; and P= 0.02 for NTC7\. ::: ![](bcr1006-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Role of phosphatidylinositol 3-kinase (PI3K) and c-Src in ezrin-mediated cell motility and invasion. (a)In a wound healing experiment with WTC4 cells, the PI3K inhibitor LY294002 (10 μmol/l), the c-Src inhibitor SU6656 (10 μmol/l), or the mitogen-activated protein kinase kinase (MEK) inhibitor (PD098059; 30 μmol/l), or the solvent DMSO (dimethyl sulfoxide; 10 μl/culture) was added to cultures, and wound closure was monitored up to 24 hours, as described for Fig. 3a. Representative fields photographed after 24 hours are shown. (b)The histogram shows pooled results from clones WTC4 and WTC6 in three independent wound healing experiments with the above inhibitors. Significant reduction in motility was observed in groups treated with LY294002 (P= 0.002) and SU6656 (P= 0.03). (c)WTC4 cells were set up in transwell cultures with PI3K, c-Src, or MEK inhibitors at the concentrations indicated above, and cell invasion was assessed after 36 hours. Results are expressed as the mean cell invasion ± standard deviation of at least three independent experiments. Single asterisk (\) indicates a specific reduction in invasion compared with DMSO-treated cells (LY294002, P= 0.02; SU6656, P= 0.02). (d)For analysis of Akt and Erk1/2 activation, transfected AC2M2 cell lines (see Fig. 1a) were serum starved overnight and cultured on fibronectin substratum (10 μg/ml) for 45 min. Cells were then lysed, and equal protein amounts of each cell lysate were subjected to 10% SDS-PAGE under reduced conditions. Proteins were transferred to PVP membranes, and western blotting was carried out with antibodies against Akt pS473, pan-Akt, Erk1/2 pT185/pY187 and pan-Erk1/2. (e)For c-Src analysis, cells were plated for 2 hours on fibronectin substratum and cell lysates of Triton X-100 soluble and insoluble (cytoskeletal-associated) fractions were prepared, as described in the text. Blots were probed with antibodies against c-Src pY418 and pan c-Src. ::: ![](bcr1006-4) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Expression of amino-terminal ezrin inhibits metastasis of breast carcinoma cells ::: Cell line^a^ Transfected with Primary tumour takes (%) Day of 1 cm tumour diameter^b^ Day of sacrifice Metastasis (%)^c^ -------------- ------------------ -------------------------- -------------------------------- ------------------ ------------------- SP1 None 100% (13/13) 24 ± 7 35 23% (3/13) AC2M2 None 100% (11/11) 30 ± 5 41 90% (10/11)^†^ pCB6 Vector 100% (8/8) 33 ± 4 39 88% (7/8) WTC4 WT ezrin 100% (7/7) 28 ± 4 41 88% (6/7) WTC6 WT ezrin 100% (15/15) 26 ± 2 39 87% (13/15) NTC6 N-term ezrin 38% (3/8) 40 ± 10\ 47 0% (0/8)^‡^ NTC7 N-term ezrin 100% (8/8) 28 ± 1 39 38% (3/8)^‡^ NTB8^d^ N-term ezrin 100% (5/5) 29 ± 3 41 0% (0/5)^‡^ ^a^Poorly metastatic parental SP1 cells or highly metastatic variant AC2M2 cells alone, or transfected with empty pCB6 vector, or a vector encoding wild-type (WT) ezrin or amino-terminal (N-term) ezrin, were transplanted (7.5 × 10^3^cells) into the mammary fat pad of syngeneic mice (see text). ^b^Day to 1 cm tumour diameter was calculated by linear regression analysis of data from individual mice. Values are expressed as mean ± standard deviation. Clone NTC6 showed a significant increase (\) in the day of 1 cm tumour diameter compared with WTC4 and WTC6 (P= 0.012). Mice with NTC6 tumours were therefore killed approximately 1 week later to allow tumour growth to a comparable size. ^c^AC2M2 cells showed significantly more metastases than did the parental SP1 cells (^†^P= 0.003; Fisher\'s exact test). Pooled results from three N-term ezrin expressing clones showed a significant reduction in metastases compared with two WT ezrin expressing clones (^‡^P\< 0.0001). Individual P*values for NTC6, NTC7 and NTB8 are as follows (respectively): \<0.0001, 0.002 and 0.002. ^d^NTB8 is an N-term ezrin-expressing clone derived from an independent transfection of AC2M2 cells, and was transplanted as described above. :::	PubMed Central	2024-06-05T03:55:57.441339	2005-3-21	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143558/", "journal": "Breast Cancer Res. 2005 Mar 21; 7(3):R365-R373", "authors": [ { "first": "Bruce E", "last": "Elliott" }, { "first": "Jalna A", "last": "Meens" }, { "first": "Sandip K", "last": "SenGupta" }, { "first": "Daniel", "last": "Louvard" }, { "first": "Monique", "last": "Arpin" } ] }
PMC1143559	Introduction ============ Steroid hormones regulate the expression of proteins that are involved in breast cell proliferation and development, and coactivators that interact with steroid hormone receptors to modulate transcriptional activation have recently been described \[[@B1]\]. AIB1, located at 20q12, is a member of the steroid hormone coactivator family that interacts with oestrogen receptor-α, resulting in enhancement of oestrogen-dependent transcription \[[@B2]\]. AIB1is moderately expressed in the normal mammary epithelium and is required for female reproductive function and mammary gland development \[[@B3]\]. It is overexpressed in 64% of breast tumours, and the 20q12 region has been shown to be amplified in 5--10% of breast cancers and 7% of ovarian cancers \[[@B2],[@B4]\]. AIB1contains a glutamine repeat (CAG/CAA) polymorphism at its carboxyl-terminal region. Although its functional significance is currently unknown, it may alter the transcriptional activation of the receptor (as is the case for the androgen receptor CAG repeat polymorphism), and hence it may affect breast cancer susceptibility through altered sensitivity to hormones \[[@B5]\]. We evaluated this repeat polymorphism within the context of early-onset disease by conducting a case-control study of Australian women diagnosed with breast cancer before the age of 40 years and population-based control individauls \[[@B6],[@B7]\]. Materials and methods ===================== Participants ------------ The Australian Breast Cancer Family Study is a population-based, case-control family study conducted in Melbourne and Sydney \[[@B6],[@B8]\]. For this study, cases were women aged under 40 years at diagnosis of a first primary invasive breast cancer between 1992 and 1995, and were identified through the Victoria and New South Wales cancer registries. Controls were women without breast cancer selected via the electoral rolls (registration is compulsory) between 1993 and 1999 and were frequency matched for age. Cases and controls were administered the same questionnaire on risk factors, blood samples were collected from them at the time of interview, and a detailed family history was recorded for all first-degree and second-degree relatives, with verification sought for all reports of family cancers. To date 25 of these cases have been found to carry a deleterious germline mutation in BRCA1and 11 in BRCA2\[[@B9]\]. Written informed consent was obtained from all participants, and approval of the protocol was obtained from the relevant ethics committees. Genotype analysis ----------------- PCR amplifications were performed with a fluorescent labelled forward primer (5\'-GACAACAGAGGGTGGCTAT-3\') and an unlabelled reverse primer (5\'-AGGAGCTTGTGGCATTGTG-3\'). All PCRs were performed in 10 μl volumes containing 10--50 ng genomic DNA, 200 nmol/l dNTPs (Promega, Annandale, New South Wales, Australia), 25 ng of each primer, 1 × ReddyMix buffer (Abgene, Epsom, Surrey, UK) and 0.2 units of Thermoprime Plus DNA Polymerase (Abgene). PCR amplification cycle conditions involved an initial denaturation step at 94°C for 5 min, 40 cycles of denaturation at 94°C for 15 s, annealing at 55°C for 30 s, and extension at 72°C for 45 s. This was followed by a further extension step at 72°C for 7 min. The alleles were then separated on sequencing gels and analyzed using a scanning laser fluorescence imager (Bio-Rad FX Molecular Imager; Bio-Rad, Hercules, CA, USA). DNA was available for 410 (88%) out of 466 cases and for 441 (74%) out of 600 controls, and AIBIgenotyping was successful for all but 17 (4%) cases and 9 (2%) controls. Statistical analysis -------------------- Allele frequencies and genotypes were compared using Pearson\'s χ^2^test. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated using unconditional logistic regression, adjusting for reference age, study centre, country of birth, education level, marital status, number of live births, height, current oral contraceptive use status and reported first-degree family history. Routine model diagnostics procedures and goodness-of-fit were performed to check the adequacy of the logistic regression models. All tests were performed using Stata version 8.0 (Stata Corporation, College Station, TX, USA). Power calculations were performed using StatCalc module of Epi Info version 6 (Centers for Disease Control and Prevention \[CDC\], Atlanta, GA, USA). All statistical tests were two-sided. Results ======= For controls the allele frequencies for 26, 27, 28, 29, 30 and 31 repeats were 0.12, 0.002, 0.36, 0.50, 0.005 and 0.002, respectively. For cases they were 0.13, 0.001, 0.35, 0.51, 0.004 and 0, respectively. Table [1](#T1){ref-type="table"} shows that there was no difference overall between breast cancer cases and controls in allele frequencies (P= 0.7) or genotype frequencies defined by the number of repeats (P= 0.3). There were no differences in the genotype distribution of cases by family history (P= 0.1). Table [2](#T2){ref-type="table"} shows the estimates of breast cancer risk; we chose a cutoff of 26 repeats or fewer because that had been used in a previous study \[[@B1]\]. There was no evidence that women with one or two alleles had increased risk of breast cancer (OR = 1.03, 95% CI = 0.73--1.44) or that women with alleles of greater than 29 repeats were at increased risk. Exclusion of women who carried a BRCA1or BRCA2mutation (24 cases) and non-Caucasian women (44 cases) did not alter risk estimates or inferences. For example, the OR for the association with one or two alleles of 26 repeats or fewer became 1.09 (95% CI = 0.76--1.56). Discussion ========== In the present study of Australian women the allele frequencies and genotype distributions for the AIB1glutamine repeat (CAG/CAA) polymorphism were similar to those reported in previous studies \[[@B1],[@B7]\]. We found no evidence that the risk of early-onset breast cancer depended on AIB1genotype. We have presented the raw data, including that on mutation carriers, to allow pooling with other studies. The only other study of breast cancer risk and this polymorphism in noncarriers of a BRCA1or BRCA2mutation was conducted in women aged 43--69 years \[[@B1]\]; it found no association. Those investigators found a weak suggestion that premenopausal women who carried two short AIB1repeats were at decreased risk, but we did not find any support for this in women under the age of 40 years. Evidence for an increased risk in women with longer repeat lengths has been reported in BRCA1and BRCA2mutation carriers \[[@B3]\]. There was no indication from our data that mutation carriers, irrespective of the gene, were at increased risk if they carried longer repeat lengths, but because of the small numbers included we had little power to address this issue. Conclusion ========== We found no evidence that risk of breast cancer depends on AIB1CAG/CAA polymorphism status, even if affected women carry a mutation in BRCA1or BRCA2. Abbreviations ============= CI = confidence interval; OR = odds ratio; PCR = polymerase chain reaction. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= KGM and IGC conceived the study; participated in its design, concept and coordination; and drafted the manuscript. J-HC performed the statistical analysis. DMG, JLH, GGG and MRM participated in the concept and coordination, are the chief investigators and developed ABCFS, and helped to draft the manuscript. GSD contributed to the design and management of data. MCS contributed to the design and management of laboratory processes at the GEL to enable and provide the required biospecimens for genetic analysis. Figures and Tables ================== ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Distribution of genotypes defined by observed AIB1glutamine repeat alleles ::: Number of repeats in AIB1 Controls Cases Total Cases --------------------------- ------------ ------------ ------------ ----------- ----------- ------------- ------------- 26/26 5 (1.2) 9 (2.3) 14 (1.7) 4 (1.5) 5 (4.3) 0 (0) 0 (0) 26/28 39 (9.0) 34 (8.7) 73 (8.9) 24 (8.7) 10 (8.6) 2 (14.3) 0 (0) 26/29 58 (13.4) 52 (13.2) 110 (13.3) 40 (14.5) 12 (10.3) 0 (0) 1 (9.0) 27/28 1 (0.2) 0 (0) 1 (0.1) 0 (0) 0 (0) 0 (0) 0 (0) 27/29 1 (0.2) 1 (0.3) 2 (0.2) 1 (0.36) 0 (0) 0 (0) 0 (0) 28/28 56 (13.0) 64 (16.3) 120 (14.6) 52 (18.8) 12 (10.3) 5 (35.7) 2 (18.2) 28/29 160 (37.0) 116 (29.5) 276 (33.5) 81 (29.4) 35 (29.9) 3 (21.4) 3 (27.3) 28/30 2 (0.5) 1 (0.3) 3 (0.4) 1 (0.34 0 (0) 0 (0) 0 (0) 29/29 106 (24.5) 114 (29.0) 220 (26.7) 71 (25.7) 43 (36.8) 4 (28.6)^a^ 5 (45.5)^a^ 29/30 2 (0.5) 2 (0.5) 4 (0.5) 2 (0.7) 0 (0) 0 (0) 0 (0) 29/31 2 (0.50) 0 (0) 2 (0.2) 0 (0) 0 (0) 0 (0) 0 (0) Total 432 393 825 276 117 14 11 Shown is the distribution of genotypes defined by the observed AIB1glutamine repeat alleles (n\[%\]) in cases and controls, and in cases categorized by family history and by BRCA1and BRCA2germline mutation status. ^a^One case, with the 29/29 repeats AIB1genotype, had a germine mutation in BRCA1and a germline mutation in BRCA2\[9\]. ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### AIB1genotype and risk of breast cancer ::: Genotype^a^ Cases (n\[%\]) Controls (n\[%\]) OR (95% CI)^b^ ------------- ------------------ --------------------- ------------------- 0 298 (76) 330 (76) 1.00 (reference) 1 86 (22) 97 (23) 0.97 (0.68--1.38) 2 9 (2) 5 (1) 2.17 (0.69--6.85) 1 or 2 95 (24) 102 (24) 1.03 (0.73--1.44) ^a^Genotype defined as number of alleles with 26 or fewer repeats. ^b^Adjusted for study centre (Melbourne/Sydney), reference age (years), country of birth (Australia/other), education level (three levels), marital status (ever/never), number of live births, height (cm), current oral contraceptive use (yes/no) and affected first-degree relative (yes/no). CI, confidence interval; OR, odds ratio. :::	PubMed Central	2024-06-05T03:55:57.444000	2005-3-4	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143559/", "journal": "Breast Cancer Res. 2005 Mar 4; 7(3):R353-R356", "authors": [ { "first": "Karen G", "last": "Montgomery" }, { "first": "Jiun-Horng", "last": "Chang" }, { "first": "Dorota M", "last": "Gertig" }, { "first": "Gillian S", "last": "Dite" }, { "first": "Margaret R", "last": "McCredie" }, { "first": "Graham G", "last": "Giles" }, { "first": "Melissa C", "last": "Southey" }, { "first": "John L", "last": "Hopper" }, { "first": "Ian G", "last": "Campbell" } ] }
PMC1143560	Introduction ============ The protein expression status of c-erbB2 and, more recently, of topoisomerase IIα has been implicated in the prediction of clinical outcome and response to chemotherapy in breast cancer \[[@B1]-[@B5]\]. Gene amplification is the predominant but not exclusive mechanism causing abnormal expression in these tumors (reviewed in \[[@B6]\]). c-erbB2is localized on chromosome 17q12-21 and encodes for a transmembrane tyrosine kinase receptor protein. This highly glycosylated protein is a member of the epidermal growth factor receptor (EGFR; HER) family \[[@B7]\] and is expressed on most cells of epithelial origin. In vitro, overexpression of c-erbB2 in epithelial cells ultimately affects the regulation of cell proliferation, of apoptotic pathways, of motility, and of adhesion (overview in \[[@B8]\]). Accordingly, numerous studies have found that both c-erbB2amplification and c-erbB2 protein overexpression predicted disease outcome in patients with localized breast cancer (overview in \[[@B9]\]). The α isoform of topoisomerase is a key enzyme in DNA replication and also a target for various chemotherapeutic agents such as anthracyclines or epipodophyllotoxins. The gene is located in close proximity to the c-erbB2gene on chromosome 17q21 and encodes for a 170-kilodalton protein. The enzyme catalyzes the unwinding of the DNA to a partly uncoiled form by inducing single-stranded breaks on both DNA strands. These breaks allow the passage of double-stranded DNA through the gap \[[@B10]\]. Anthracyclines, one of the most widely used class of cytotoxic agents for the treatment of breast cancer, inhibit topoisomerase IIα by trapping the DNA strand break intermediates, leading to persistent DNA cleavage. An important role for the outcome of patients with breast cancer has been proposed both for the c-erbB2 and topoisomerase IIα protein \[[@B11]\]. Interestingly, a high percentage of primary breast tumors with c-erbB2amplification also show a gene copy number alteration of other genes such as topoisomerase IIαlocated near to the c-erbB2locus on chromosome 17 \[[@B12],[@B13]\]. Although the appearance of topoisomerase IIαgene alterations (amplification or deletion) was exclusively seen in c-erbB2-amplified breast tumors \[[@B3],[@B12]-[@B14]\], data on the protein expression status of these two genes in breast tumor samples are controversial. Jarvinen and colleagues report a high correlation of c-erbB2 and topoisomerase IIα protein expression \[[@B15]\]. In contrast, in a recently published study it was observed that in 14 of 33 topoisomerase IIα-amplified tumors, topoisomerase IIα protein was present in less than 10% of tumor cells \[[@B16]\]. Another study showed that topoisomerase IIα overexpression is present in 30% of c-erbB2non-amplified tumors \[[@B17]\]. From these studies it has to be concluded that gene amplification is only one of many mechanisms causing high intracellular levels of the topoisomerase IIα protein. This is in concordance with the fact that expression of topoisomerase IIα is regulated on multiple levels, including transcriptional, translational, and post-translational mechanisms \[[@B18]\]. These data suggest that the expression of topoisomerase IIα and c-erbB2 protein might independently predict the clinical outcome in breast cancer. We therefore analyzed the prognostic impact of the protein expression status of both c-erbB2 and topoisomerase IIα in a large cohort of patients with primary breast cancer. Materials and methods ===================== Patients and pathological data ------------------------------ Tumor tissue was analyzed for protein expression of c-erbB2 and topoisomerase IIα from 225 patients with primary invasive breast cancer who underwent surgery from 1986 to 1998 at the Department of Gynaecology, Robert Bosch Hospital, Stuttgart, Germany. The local ethics committee was informed and gave consent. The patient database was anonymized to guarantee privacy. The tissues were formalin-fixed and paraffin-embedded in accordance with standard methods. Histological classification was performed by following the recommendations of the World Health Organization \[[@B19]\]. Three histological types were discriminated: invasive ductal, invasive lobular, and all other specified tumor types such as tubular carcinoma. The pathological reports included tumor size, palpable nodes, metastasis, grading, estrogen receptor status, and progesterone receptor status. Immunohistochemical methods --------------------------- For immunocytochemistry, 3 μm sections were deparaffinized in xylene for 30 min and rinsed in 100%, 96% and 70% ethanol. Sections were then subjected to antigen retrieval by immersion in citrate buffer (pH 6.0) preheated to 99°C for 40 min. Endogenous peroxidase was blocked by incubation in 3% H~2~O~2~in methanol for 30 min, followed by rinsing in Tris-buffered saline containing Tween 20. Immunohistochemical staining was performed with the EnVision™+ System Kit (DakoCytomation, Glostrup, Denmark). Afterwards, the sections were incubated overnight in a humidity chamber with the monoclonal primary antibody against topoisomerase IIα (DakoCytomation; dilution 1:100) and c-erbB2 (clone CB11; Novocastra, Newcastle, UK; dilution 1:10) followed by incubation for 30 min with a dextran polymer conjugated with horseradish peroxidase enzyme and with goat anti-rabbit antibody. The antigen--antibody immunoreaction was revealed with 3,3\'-diaminobenzidinetetrahydrochloride as the chromogen, and the slides were counterstained with hematoxylin. Immunohistochemical analysis of c-erbB2 (HER-2/neu) protein was also performed with HercepTest™ (DakoCytomation) on an automated immunostaining system (Dako Autostainer), with the use of the manufacture\'s detection procedures. This procedure was part of the primary diagnostic process. For each case c-erbB2 was assessed twice: once by immunostaining with CB11 and once by herceptin staining. Estrogen and progesterone receptor statuses were assessed during the first years (up to 1990) by charcoal dextran method \[[@B20]\], samples examined after this time were analyzed by immunohistochemistry. Scoring interpretation ---------------------- The scoring system proposed by HercepTest was used for the interpretation of the immunoreactivity of both CB11 and HercepTest, distinguishing between no staining (0), weakly (1+), moderate (2+), and strong membrane staining (3+). Cytoplasmic staining was ignored. Only nuclear staining was considered for topoisomerase IIα. Immunostaining frequency of the tumor cells was scored subjectively on a scale of 1 to 4 (1, 0 to 5% positive tumor cells; 2, 6 to 25%; 3, 26 to 75%; 4, more than 75%), as proposed by Sandri and colleagues \[[@B21]\]. Finally, after cut-off analysis we stratified the results as negative for less than 25% tumor cells and positive for all cases in which more than 25% of tumor cells stained positive for topoisomerase IIα. For hormone receptor status we classified the tumor as positive if there was more than 15 fmol per mg of protein (charcoal dextran method) or with a score of more than 1 (immunohistochemistry). Statistical methods ------------------- Descriptive statistical analysis was performed with commercially available software packages (SPSS, version 11.1 (SPSS GmbH, Munich, Germany) and R, version 2.0 <http://www.r-project.org>). We used the Kaplan--Meier estimator for univariate statistical analysis and the Cox regression model for multivariate analysis. P\< 0.05 was considered to be significant. A log-rank test was applied for assessing statistical differences between survival curves. The χ^2^test was used to investigate the relationship between topoisomerase IIα expression and histological grading and to analyze the association between topoisomerase IIα and c-erbB2 overexpression with hormone receptor status or stage. Results ======= Patients -------- Tissue samples of 225 patients were analyzed for protein expression of c-erbB2 or topoisomerase IIα. The characteristics of these patients and tumors are summarized in Table [1](#T1){ref-type="table"}. The distribution of stage, tumor size, nodal status, histological grading, and receptor status is in conformity with that observed in published randomized clinical trials (Table [2](#T2){ref-type="table"}; reviewed in \[[@B22]\]). All patients had received appropriate local surgical treatment. Adjuvant medical therapy was given to 143 patients: 61 received hormone therapy; 82 were treated with adjuvant chemotherapy, either anthracycline-containing (46 patients) or anthracycline-free (36 patients; Table [3](#T3){ref-type="table"}) regimens. After a mean follow-up period of 67 months, 136 patients were alive and 89 deaths had been recorded. The death of 68 patients (30%) was documented to have been caused by breast cancer (Table [1](#T1){ref-type="table"}). Survival at 1 year was 95.73%, at 3 years 85.45%, at 5 years 74.26%, and at 10 years 63.69% for all patients. Again, this survival rate is comparable to that published for similar patient populations \[[@B22]\]. Prognostic impact of c-erbB2 overexpression ------------------------------------------- Tumors with an immunoreactive score exceeding 2 or 3 were considered to overexpress c-erbB2. Accordingly, 43 (19.1%) of the patients showed overexpression of this oncogene. Patients with c-erbB2 overexpression showed a similar distribution of age, stage, histological criteria, and receptor status to that of the c-erbB2 negative subgroup (data not shown). However, the former patients were characterized by a significantly inferior survival in a univariate statistical analysis (log rank 17.94; P\< 0.0001). The median overall survival time of patients with tumors overexpressing c-erbB2 was 55 months, with a 5-year survival rate of 46.0%. In contrast, median survival time was not reached in the c-erbB2 negative group. The 5-year survival in this group was 78.3% (Fig. [1a](#F1){ref-type="fig"}). Patients with c-erbB2-overexpressing tumors received no significant benefit from anthracycline-based adjuvant therapy, and even had the worst prognosis of all groups analyzed (log rank 10.17; P= 0.001; see below). Impact of topoisomerase IIα expression -------------------------------------- Topoisomerase IIα staining was strictly nuclear and highly variable between different tissue samples. To study the impact of different levels of overexpression of topoisomerase IIα protein, we analyzed the survival curves of patients with 0 to 5%, 6 to 25%, 26 to 75% and more than 75% positively stained cells (data not shown). These analyses demonstrated that the survival of patients was significantly inferior in cases with high topoisomerase IIα expression in more than 25% of the cells. Therefore, for further analysis overexpression was defined as more than 25% positively stained tumor cells. A subgroup of 77 patients (34.2%) had tumors overexpressing topoisomerase IIα. As shown in Fig. [1b](#F1){ref-type="fig"}, these patients had a significantly inferior survival (median 80 months; 5-year survival rate 54.4%) in a univariate Kaplan--Meier analysis compared with those without topoisomerase IIα overexpression (median not reached; 5-year survival rate 80.3%; log rank 15.59, P= 0.0001). The proportion of tumors expressing topoisomerase IIα increased with histological grading (χ^2^test, P\< 0.001). The fraction of patients with hormone-receptor-positive tumors was significantly higher in samples without topoisomerase IIα overexpression (χ^2^test, P= 0.004). The prognostic impact of topoisomerase IIα expression was dependent on stage: whereas stage IV patients had an identically poor outcome regardless of topoisomerase IIα expression (not shown), stage II and III patients had a significantly lower survival rate when topoisomerase IIα was highly expressed (log rank 9.35, P= 0.002 for WHO stage II; log rank 4.76, P= 0.029 for WHO stage III; Fig. [2a](#F2){ref-type="fig"}). In stage I patients a survival difference of similar magnitude was not statistically significant, probably because of the small sample size (Fig. [2a](#F2){ref-type="fig"}). Expression of topoisomerase IIα added prognostic information to histological grading. The analysis was restricted to grade 2 and 3 tumors, because only a few deaths were observed in patients with grade 1 disease. The survival of patients with tumors expressing topoisomerase IIα was inferior in both grade 2 tumors (log rank 5.08; P\< 0.05) and grade 3 tumors (log rank 7.86; P= 0.005; Fig. [2b](#F2){ref-type="fig"}). In addition, the prognostic impact of topoisomerase IIα expression was clearly dependent on the steroid hormone receptor status of the tumors: no significant difference was observed in patients with tumors negative for estrogen or progesterone receptor (log rank 0.94; P= 0.33; Fig. [3](#F3){ref-type="fig"}, right panel) whereas detection of topoisomerase IIα in more than 25% of tumor cells identified a subgroup with poor prognosis in receptor-positive breast cancer (log rank 12.0; P= 0.0005; Fig. [3](#F3){ref-type="fig"}, left panel). We further studied whether the prognostic impact of topoisomerase IIα expression was restricted to patients receiving either non-anthracycline-containing or anthracycline-containing regimens for adjuvant treatment. By analogy with the results obtained with c-erbB2-overexpressing tumors (Fig. [4b](#F4){ref-type="fig"}), treatment with anthracyclines was not capable of reversing the negative prognostic impact of topoisomerase IIα expression (log rank 4.74; P= 0.02; Fig. [4a](#F4){ref-type="fig"}). Coexpression of c-erbB2 and topoisomerase IIα --------------------------------------------- The subgroups overexpressing c-erbB2 and topoisomerase IIα were overlapping but not identical. One hundred and twenty-six tumors (56%) showed neither c-erbB2 nor topoisomerase IIα overexpression. Fifty-six (25%) patients had tumors overexpressing topoisomerase IIα only; 22 (9.8%) showed only c-erbB2 overexpression. Twenty-one (9.3%) tumors showed increased staining for both proteins. None of these groups differed significantly with regard to clinical stage (Table [4](#T4){ref-type="table"}). However, a statistically significantly higher proportion of hormone-receptor-negative cancers was observed in the group overexpressing one or both proteins studied (χ^2^test, P= 0.006; Table [4](#T4){ref-type="table"}). A survival analysis of these four groups demonstrated an independent negative prognostic impact of either overexpression. As shown in Fig. [5](#F5){ref-type="fig"}, patients with breast cancer lacking overexpression of both c-erbB2 and topoisomerase IIα had the best long-term prognosis (the median survival time was not reached; the 5-year survival rate was 84.4%). Overexpression of either c-erbB2 or topoisomerase IIα was associated with intermediate survival (for c-erbB2-overexpressing tumors the median survival time was 68 months and the 5-year survival rate was 57.7%; for topoisomerase IIα-overexpressing cases the median survival time was 104 months and the 5-year survival rate was 63.7%). Those patients overexpressing both proteins had the worst outcome, with a median survival of 45 months and a 5-year survival rate of 33.0% (log rank 29.71; P\< 0.0001; Fig. [5](#F5){ref-type="fig"}). In a multivariate Cox regression model (Tables [5](#T5){ref-type="table"} and [6](#T6){ref-type="table"}), tumor stage, estrogen receptor, topoisomerase IIα, and c-erbB2 all independently predicted disease-related death. Subsequent inclusion of grading was far beyond statistical significance (Table [5](#T5){ref-type="table"}). Adding the interaction term topoisomerase IIα by grading (topoisomerase IIα × grading; Table [5](#T5){ref-type="table"}) seems worth consideration. However, in a Wald statistics grading together with the last-mentioned interaction term failed to reach significance (P= 0.11). Models containing in addition interactions of c-erbB2 by grading (c-erbB2 × grading; Table [5](#T5){ref-type="table"}) and c-erbB2 by topoisomerase IIα (c-erbB2 × topoisomerase IIα; Table [5](#T5){ref-type="table"}) are not significantly superior. The last two interaction terms are therefore not included in the final model as presented in Table [6](#T6){ref-type="table"}. Discussion ========== Overexpression of c-erbB2 and topoisomerase IIα independently predicts poor survival in this retrospective series of patients. Topoisomerase IIα and c-erbB2 were found to be overexpressed in overlapping but distinct subgroups of patients. Moreover, the prognostic impact of topoisomerase IIα overexpression seems to be independent of other prognostic variables in a multivariate analysis and makes the prognosis significantly worse in both c-erbB2-positive and c-erbB2-negative patients. In addition, the prognostic impact of topoisomerase IIα overexpression is dependent on the steroid receptor status. The results of this study suggest that anthracycline treatment is not capable of reversing the negative prognostic influence of topoisomerase IIα or c-erbB2 expression. Recent reports have studied the interaction between topoisomerase IIα expression and pathological variables, chemotherapy response, and the proliferation rate of breast cancer. However, only very limited series reported a long-term outcome of patients with primary breast cancer depending on topoisomerase IIα. Di Leo and colleagues \[[@B2]\] studied patients participating in a randomized clinical trial comparing CMF (cyclophosphamide, methotrexate, and 5-flourouracil) with an anthracycline-containing regimen. The subgroup of patients with topoisomerase IIαamplification did not do obviously more badly than the remaining patients. However, the outcome was strongly dependent on the adjuvant therapy applied. Another series measuring topoisomerase IIα protein by immunohistochemistry showed a significantly adverse influence of topoisomerase IIα expression of a similar magnitude to that in our study \[[@B23]\]. Durbecq and colleagues \[[@B16]\] have shown that measurement of protein expression and gene amplification each identify different subsets of patients. These data suggest that detection of the protein might be better correlated with the biological properties of the tumor and therefore might predict the clinical outcome more precisely than genetic analysis. The data in our study support the view that protein expression of topoisomerase IIα is a relevant factor predicting long-term prognosis in patients with newly diagnosed breast cancer independent of c-erbB2. Treatment of cell lines with anthracyclines was less effective in cells with a low expression of topoisomerase IIα \[[@B24],[@B25]\]. A retrospective subgroup analysis of a randomized clinical trial suggests that the prognostic impact of topoisomerase IIαgene amplification is restricted to the patients not receiving anthracycline-based chemotherapy \[[@B2]\]. Furthermore, other studies that related tumor response to anthracyclines also showed some correlation between gene amplification status of topoisomerase IIαand clinical response \[[@B3]-[@B5]\]. This is in contrast with a recently published retrospective study in which amplification of c-erbB2and topoisomerase IIαwas not predictive of the response to anthracycline \[[@B26]\]. Our analyses show that the negative prognostic impact of topoisomerase IIα protein overexpression is observed both in patients who received an anthracycline-containing regimen for adjuvant chemotherapy and in those who did not. Moreover, the difference in survival between topoisomerase IIα-positive and topoisomerase IIα-negative patients exceeds the proportion of patients expected to be cured by chemotherapy. Protein overexpression has been related to a variety of other molecular markers predictive for high proliferation rate or high grading of malignancy such as Ki67 expression or aneuploidy \[[@B16],[@B23],[@B26],[@B27]\]. These findings support the view that overexpression of topoisomerase IIα protein indicates a poor prognosis irrespective of the therapy applied. Topoisomerase IIαgene amplification is much more closely associated with c-erbB2amplification than the protein expression status (overview in \[[@B6]\]). The predictive value of the gene amplification for clinical and long-term outcome might therefore be related to amplification of either or both of the c-erbB2or topoisomerase IIαgenes. As coamplification is the predominant mechanism of genetic alteration of these two genes in breast cancer, clinical observations separate the role of each gene in the clinical response to adjuvant treatment with anthracyclines. The biological role of topoisomerase IIα overexpression is unknown. Cells lacking topoisomerases II are not capable of finishing a normal cell cycle and should therefore not be viable \[[@B28],[@B29]\]. In addition, it has been shown that experimental overexpression of topoisomerase IIα in different human cell lines causes apoptosis \[[@B30]\]. From these observations one must conclude that cells staining with a low intensity are not cells with a complete lack of topoisomerase IIα function. This is in conformity with the observation that normal cells are also weakly positive for topoisomerase IIα. High expression might be related to a high proliferation rate. This view is supported by our finding that topoisomerase IIα expression is related to histological grading and by observations showing some correlation of topoisomerase IIα with Ki67 expression and other proliferation markers \[[@B23],[@B27]\]. However, viable cells with constitutive overexpression might also indirectly indicate defects in apoptotic pathways. Both hypotheses might provide a plausible explanation for the observation that topoisomerase IIα overexpression is related to a more aggressive tumor phenotype. Conclusion ========== Our data suggest that protein expression of topoisomerase IIα is a prognostic factor that is independent of c-erbB2, stage, and histological grading. In addition, the results of this exploratory study indicate that anthracycline treatment is not capable of reversing the negative prognostic influence of the expression of topoisomerase IIα or c-erbB2. Nevertheless, because of the small number of patients remaining in these subgroups, no firm conclusion can be made about the predictive value of topoisomerase IIα or c-erbB2 regarding sensitivity to anthracyclines. Our results support the view that studying the deregulation of either the topoisomerase IIαgene or topoisomerase IIα protein might yield different results depending on the method applied. This should be considered when planning prospective studies on the predictive and prognostic value of topoisomerase IIα. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= PF, WS, WEA, JD, and HvdK were responsible for generating the hypothesis and correcting the manuscript. PF, AG, and WS were responsible for surgery and for collecting the patient material. PF, CMC, and HvdK were responsible for immunostaining and for examination and interpretation of the results. PF, JD, and WEA performed the statistical analysis of the data. PF, HvdK, and WEA were responsible for writing the manuscript. All authors read and approved the final manuscript. Acknowledgements ================ We thank Monika McClellan, Susanne Gutzeit, and Kerstin Gawronski for technical assistance. This work was supported by a research grant from the Robert-Bosch foundation, project no. 11.5.8002.0004.1 (PF, CMC, WEA, HvdK). Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Overall survival of breast cancer patients with regard to c-erbB2 and topoisomerase IIα protein expression. (a)Patients with (solid line) or without (dashed line) c-erbB2 protein overexpression. The difference was significant in univariate statistical analysis (log rank 17.94; P\< 0.0001). (b)Significant difference of survival time of patients with tumors positive for topoisomerase IIα protein in more than 25% of tumor cells (solid line) versus patients with tumors with less than 25% topoisomerase IIα-positive cells (dashed line) (log rank 15.59; P= 0.0001). ::: ![](bcr1012-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Prognostic impact of topoisomerase IIα protein expression with regard to stage and grading. Patients with tumors positive for topoisomerase IIα in more than 25% of tumor cells (solid lines) were compared with those having tumors expressing topoisomerase IIα in less than 25% of cells (dashed lines) with regard to the stage (a)and the grading (b)of the tumor. The difference was significant in tumors of stage 2 and 3 (P\< 0.05) and in grade 2 and 3 tumors (P\< 0.05). ::: ![](bcr1012-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Prognostic impact of topoisomerase IIα protein expression with regard to hormone status. Patients with tumors positive for topoisomerase IIα in more than 25% of tumor cells (solid lines) were compared with those having tumors expressing topoisomerase IIα in less than 25% of cells (dashed lines) with regard to the hormone status. The difference was significant in tumors positive for estrogen receptor or progesterone receptor (log rank 12.0; P= 0.0005). ::: ![](bcr1012-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Survival of patients receiving either non-anthracycline-based therapy or anthracycline-based chemotherapy. (a)The difference between patients with less than 25% positive cells for topoisomerase IIα (dashed lines) and more than 25% positive cells (solid lines) was significant for both patient groups (P\< 0.05). (b)Significant difference between patients with (solid line) or without (dashed line) c-erbB2 overexpression in the group of patients receiving non-anthracycline-based adjuvant therapy (left panel; P= 0.002) or an anthracycline-containing regimen (right panel; P= 0.001) ::: ![](bcr1012-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Survival of patients with and without overexpression of c-erbB2 and topoisomerase IIα. The difference between patients without overexpression of c-erbB2 or topoisomerase (broken line), patients with tumors overexpressing one of the two proteins (c-erbB2, dot--dashed line; topoisomerase IIα, solid line), and patients with tumors overexpressing both c-erbB2 and topoisomerase IIα (dotted line) was significant (log rank 29.71; P\< 0.0001). ::: ![](bcr1012-5) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Characteristics of patients and tumors ::: Characteristic Value --------------------------------------------- ------------- Number of patients 225 Age, years (median (range)) 56 (29--88) Time to relapse, months (median (range))^a^ 25 (0--179) Alive without disease 123 Relapse 23 Disease status unknown 3 Alive (censored) 136 Death 89 By tumor 68 By unknown causes 8 Not tumor-related (censored) 13 ^a^Only cases with recurrence are included in this calculation. ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Clinical characteristics of study patients (N= 225) ::: Variable n(%) ----------------------- ------------ Stage (n= 214) I 40 (18.7) II 113 (52.8) III 53 (24.8) IV 8 (3.7) T (n= 222) T1 59 (26.6) T2 106 (47.7) T3 27 (12.2) T4 30 (13.5) N (n= 222) N0 94 (42.3) N1 112 (50.5) N2-3 16 (7.2) Grading (n= 220) G1/G2 151 (68.6) G3 69 (31.4) Histology (n= 225) Ductal 175 (77.8) Lobular 22 (9.8) All other subtypes 28 (12.4) Menopausal (n= 163) Premenopausal 43 (26.4) Postmenopausal 120 (73.6) ER (n= 212) Positive 135 (63.7) Negative 77 (36.3) PR (n= 211) Positive 127 (60.2) Negative 84 (39.8) Metastasis (n= 218) Negative 210 (96.3) Positive 8 (3.7) ER, estrogen receptor; N, palpable nodes; PR, progesterone receptor; T, tumor size. ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Medical treatment for primary breast cancer in the study patients (N= 198) ::: Therapy type Intervention n(%) -------------------------- ----------------------------------- ------------ Adjuvant hormone therapy No hormone therapy 137 (69.2) Tamoxifen 61 (30.8) Adjuvant chemotherapy Anthracycline-containing regimens 46 (23.2) Anthracycline-free regimens 36 (18.2) No chemotherapy 116 (58.6) No adjuvant therapy 55 (27.8) ::: ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Clinical characteristics of patients with and without overexpression of c-erbB2 and topoisomerase IIα ::: Parameter Value ------------------------- ----------- ----------- ----------- ----------- Stage, n(%) I 29 (24.2) 3 (15.0) 6 (11.1) 2 (10.0) II 59 (49.1) 7 (35.0) 36 (66.6) 11 (55.0) III 29 (24.2) 9 (45.0) 9 (16.7) 6 (30.0) IV 3 (2.5) 1 (5) 3 (5.6) 1 (5.0) Total^a^ 120 20 54 20 Receptor status, n(%) ER or PR positive 92 (79.3) 12 (57.1) 32 (58.2) 10 (52.6) PR and PR negative 24 (20.7) 9 (42.9) 23 (41.8) 9 (47.4) Total^b^ 116 21 55 19 ^a^214 patients with complete data were analyzed for stage; ^b^211 patients with complete data were analyzed for receptor status. ER, estrogen receptor; PR, progesterone receptor. ::: ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Choice of the Cox regression model (N= 196) ::: Stepwise inclusion of variable 2 × increase in partial log-likelihood (χ^2^distributed) df P -------------------------------- ---------------------------------------------------------- ---- ----------- Stage 23.43 3 \< 0.0001 Estrogen receptor 7.92 1 0.005 c-erbB2 7.34 1 0.01 Topoisomerase IIα 9.85 1 0.002 Grading 0.01 1 0.92 Topoisomerase IIα × grading 4.64 1 0.03 c-erbB2 × grading 1.74 1 0.19 c-erbB2 × topoisomerase IIα 0.17 1 0.68 Starting from the null model, successive adding of the variables mentioned above improves the Cox proportional hazard model in the first four steps significantly. In this sense the first four variables can be considered as independent prognostic factors. df, degrees of freedom. ::: ::: {#T6 .table-wrap} Table 6 ::: {.caption} ###### Coefficients of the linear predictor in the final Cox regression model (N= 196) ::: Variable B exp(B) P 95% CI for exp(B) -------------------------------------- ------- ---------- --------- --------------------- Stage II^a^ 0.72 2.06 0.14 0.79--5.36 III^a^ 1.69 5.4 0.001 1.97--14.74 IV^a^ 3.25 25.68 \<0.001 6.72--98.17 Estrogen receptor (pos. versus neg.) 0.58 1.79 0.039 1.03--3.11 c-erbB2 (pos. versus neg.) 0.82 2.26 0.0036 1.31--3.93 Topoisomerase IIα (pos. versus neg.) 0.39 1.48 0.005 0.77--2.83 Grading (3 versus {1,2}) t0.62 0.54 0.15 0.23--1.25 Topoisomerase IIα × grading^b^ 1.21 3.35 0.035 1.09--10.36 ^a^The hazard ratio exp(B) of the stage given is the relative risk of this stage with respect to stage I. ^b^The effect of the interaction term on survival can be described as follows: if grading equals G1 or G2, the hazard ratio of topoisomerase IIα positive versus negative is 1.48 (95% confidence interval (CI) 0.77 to 2.83); if grading equals G3, this hazard ratio is 4.96 (95% CI 2.00 to 12.29). :::	PubMed Central	2024-06-05T03:55:57.445587	2005-3-21	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143560/", "journal": "Breast Cancer Res. 2005 Mar 21; 7(3):R374-R384", "authors": [ { "first": "Peter", "last": "Fritz" }, { "first": "Cristina M", "last": "Cabrera" }, { "first": "Jürgen", "last": "Dippon" }, { "first": "Andreas", "last": "Gerteis" }, { "first": "Wolfgang", "last": "Simon" }, { "first": "Walter E", "last": "Aulitzky" }, { "first": "Heiko", "last": "van der Kuip" } ] }
PMC1143561	Introduction ============ The aromatic and heterocyclic amines are among the constituents of tobacco smoke that have been identified as carcinogens \[[@B1],[@B2]\]. These carcinogens require host-mediated metabolic activation to electrophiles, which readily bind nucleophilic DNA, to induce a mutation and, ultimately, cancer \[[@B3]\]. Two pathways metabolize aromatic amines \[[@B4]\]. First, aromatic amines can be N-acetylated in the liver by N-acetyltransferase-2 (NAT2) or N-acetyltransferase-1 (NAT1) \[[@B5]\]. This is a detoxifying pathway. Second, aromatic amines and heterocyclic amines can be N-oxidized by P450 enzymes in the liver or in extrahepatic tissues \[[@B4]\]. This oxidation competes with the hepatic N-acetylation for aromatic amines but not for heterocyclic amines. The product of the oxidation is then either O-acetylated by NAT2 or NAT1, a reaction that yields the activated electrophile, or detoxified by competing enzymatic pathways \[[@B4],[@B6]\]. The NAT2 enzyme therefore has a dual role: it detoxifies aromatic amines hepatically but may also play a role in activation of aromatic amines and heterocyclic amines in extrahepatic tissues such as the breast. NAT2 is a polymorphic enzyme in humans \[[@B5],[@B7]-[@B10]\]. Those who possess homozygous wild-type alleles are classified as fast acetylators -- because they have a higher rate of metabolic activation of aryl aromatic amines -- and those who possess certain polymorphisms are called slow acetylators because they have a lower rate of metabolic activation of these amines \[[@B9],[@B11]\]. Depending on which metabolic pathway predominates at critical junctures of exposure and tissue susceptibility, fast acetylators may be at higher or lower risk of smoking-induced breast carcinogenesis than slow acetylators. Postmenopausal women who smoke and have a reduced ability to detoxify by-products of tobacco smoke, as measured by their NAT2 genotype (slow acetylators), have an excess risk of breast cancer \[[@B12]-[@B14]\]. In one of these studies, this excess risk was found to be limited to women who had smoked for 20 years or more \[[@B14]\]. In another study, the postmenopausal women who were rapid acetylators were found to be at highest risk \[[@B15]\]. In a fifth study, the association between smoking and breast cancer showed little dependence on acetylation rate \[[@B16]\]. Most epidemiologic studies that have examined the relation between active cigarette smoking and breast cancer have found weak or null associations \[[@B17],[@B18]\]. A meta-analysis of the studies that excluded from the analysis those women who had been passively exposed reported that the risk of breast cancer for active smokers was more than twice as much as that for women never actively or passively exposed to tobacco smoke \[[@B19]\]. Studies comparing women who were passive smokers with women who had never been either active or passive smokers have also shown consistent elevations in breast cancer risk associated with smoking \[[@B20]-[@B23]\]. Recently, two case--control studies \[[@B24],[@B25]\] have reported effect modification by acetylation status for both active and passive smokers. Both studies found stronger associations between breast cancer risk and passive exposure to smoke among rapid acetylators. Though both studies also found an association between active smoking and breast cancer risk, the magnitude of the risk was greater among slow acetylators in the study by Chang-Claude and colleagues \[[@B24]\] and among fast acetylators in the study by Morabia and colleagues \[[@B25]\]. Inconsistent findings have prevented any meaningful conclusions from being drawn about the interaction of acetylation status and exposure to tobacco smoke in the etiology of breast cancer. We collected genetic and behavioral information from incident primary breast cancer cases arising in five different sites across the United States. We used a case-only design to examine the potential interaction between acetylation status -- as assigned by NAT2 genotype -- and self-reported active or passive smoking status. We hypothesized that slow acetylators would be at increased risk of breast cancer associated with both active and passive smoking, and that these risks would be more pronounced among women whose exposure began before their first pregnancy or at an early age. The case-only design is optimal for assessing multiplicative interaction when the genotype and environmental exposure are independent of one another. This investigation is the largest case-only study to examine the interaction between NAT2 acetylation status and history of tobacco exposure as it relates to the risk of breast cancer. Materials and methods ===================== Study population ---------------- The cases of female breast cancer included in this analysis were identified as parts of two study populations \[[@B26],[@B27]\]. The first population included women with pathologically confirmed incident invasive breast cancer diagnosed between 1987 and 1993 among residents of eight towns on Cape Cod, Massachusetts, and that were reported to the Massachusetts Cancer Registry. The second population included women with pathologically confirmed, incident stage I, stage II, or stage IIIa breast cancer that were diagnosed from December 1996 to September 1999 at hospitals in Los Angeles, California; Rhode Island; Minnesota; and North Carolina. Data collection --------------- ### Buccal cell samples for genotyping Introductory letters were mailed to breast cancer patients in 2001 and 2002. A trained interviewer followed the letter with a telephone call to answer questions and solicit participation. Patients who agreed to participate were sent an enrollment package containing an introductory letter, summary information about the study, an informed consent form, instructions for submitting a mouthwash sample, a safety-sealed sample of mouthwash, and a wide-mouth sample-collection bottle. Participants collected the sample and returned it in a postage-paid box along with their informed consent form. Buccal cells were precipitated by centrifugation and stored at -70°C until a batch of 90 samples had been collected. Batches were sent by overnight delivery on dry ice to Qiagen Genomics (Bothell, WA, USA) for DNA extraction and genotyping. Qiagen Genomics applied proprietary Masscode technology to measure Masscode tags, which are low-molecular-weight compounds linked to the DNA via a photocleavable linker. The tag is cleaved in flow into a mass spectrometer, and a Microsoft Access database converts the raw analytical data into statistically generated genotype calls. The assay has been validated in over one million genotypes. Existing primers were used to characterize NAT2 genotypes at ten single-nucleotide polymorphisms (SNPs) in each buccal cell sample. The Qiagen genotyping data characterized each participant as homozygous wild-type, heterozygous, or homozygous polymorphic at each SNP. Inferred haplotypes were estimated from the genotyping data using an expectation-maximization algorithm implemented in the software program SNPHAP <http://www-gene.cimr.cam.ac.uk/clayton/software/>, and the predicted haplotypes with the highest probability were used for the primary analyses. ### Interview data Patients who were included in the study were interviewed on the telephone by trained interviewers using a structured interview to obtain information on demographic characteristics, history of active and passive exposure to tobacco smoke, and known or suspected risk factors for breast cancer. Patients from the Cape Cod study population were interviewed between March 1997 and March 1998. Patients from the second study population were interviewed approximately 40 months after their date of diagnosis to gather the variables primarily used in this analysis. Analytic variables ------------------ ### NAT2 genotype The literature on the expression of specific SNPs in the NAT2 gene guided the phenotypic assignments for each haplotype used in this study \[[@B11],[@B28]-[@B34]\]. We considered a woman a \'rapid acetylator\' if she was homozygous for the NAT2\4a or NAT2\12 haplotype, an \'intermediate acetylator\' if she was heterozygous for the NAT2\4a or NAT2\12 haplotype, and a \'slow acetylator\' if she had any other combination of the NAT2 polymorphisms listed in Table [1](#T1){ref-type="table"}. ### Tobacco exposure We considered a woman an active smoker if she reported smoking 100 or more cigarettes in her lifetime, and a passive smoker if she was not herself a smoker but reported living with someone who was a smoker. Women who were neither active nor passive smokers were considered separately. For women who reported having smoked 100 or more cigarettes in their lifetime or who lived with someone who smoked, information on the duration, intensity, and timing of exposure to tobacco smoke (active or passive) was also collected. ### Covariates In addition to information about smoking, we collected information on health and behavioral risk factors, including alcohol use, body mass index (BMI), family history of breast cancer (yes or no), history of benign breast disease, and parity. BMI was calculated as weight divided by the square of height (kg/m^2^). A woman was considered to have a first-degree family history of breast cancer if she reported that her mother, sister(s), or daughter(s) had been diagnosed with breast cancer. We defined alcohol use according to the number of drinks a woman reported \'usually\' having: nondrinker, ≤ one drink/month, few drinks/month, few drinks/week, almost every day, and unknown. Analytic strategy ----------------- Ambrosone and colleagues \[[@B12]\] found that the rapid and intermediate arylamine N-acetyltranferase activity groups do not differ in their phenotypic expression (acetylation status). Based on this finding and others \[[@B35],[@B36]\], we collapsed rapid and intermediate acetylators into the group of rapid acetylators. We examined the interaction of acetylation status and exposure to tobacco smoke among the breast cancer cases available for analysis. We used logistic regression analysis in SAS \[[@B37]\] to quantify departure from multiplicativity. We generated odds ratios (ORs) and 95% confidence intervals (CIs) to estimate the departure from multiplicativity between smoking status and acetylation status (gene--environment interaction). We examined the ORs separately for active and passive smokers. Women who were fast acetylators and who had never been either active or passive smokers were the reference group for all analyses. We controlled for the influence of potential breast cancer risk factors including age at diagnosis of breast cancer, alcohol consumption, BMI, first-degree family history of breast cancer, geographic location (state where breast cancer diagnosis was made), and history of benign breast disease using multiple variable logistic regression. We also evaluated departure from multiplicativity for variables describing the duration, intensity, and timing of active smoking or exposure to passive smoking. For active smokers, we examined the ORs in categories of the number of packs of cigarettes smoked per day, duration of smoking, age at onset of smoking, when a woman began smoking in relation to the first live birth of a child, and time since cessation of smoking. For passive smokers, we examined the duration of passive exposure, age when first passive exposure began, and when this first passive exposure occurred in relation to the first live birth of a child. Results ======= Among the Cape Cod population, 330 of 483 eligible women agreed to receive a sample collection kit; the remainder refused or could not be contacted. Of the 330 who received a kit, 272 returned a sample and 269 samples yielded DNA that could be genotyped. Among the second study population (from California, Rhode Island, Minnesota, and North Carolina), 372 of 410 eligible women agreed to receive a sample collection kit and the remainder refused or were unable to be contacted. Of the 372 who received a kit, 321 returned a sample and 233 had samples that yielded DNA that could be genotyped and had the requisite interview data. In both studies, 56% of eligible participants were genotyped and included in the analysis. The proportion of smokers among nonparticipants was not significantly different from that among participants in either study population. The mean age was greater among nonparticipants than among participants (mean ages 66 years versus 61, respectively, in the Cape Cod population, P= 0.0001; and 74 versus 73 in the second study population, P= 0.03), reflecting greater losses to follow-up among older women. Age was not associated with genotype among the participants. The proportion of participants who were slow acetylators, active smokers, and passive smokers did not vary significantly with their site of enrollment. Among the genotyped controls in the Cape Cod study, the OR for association of acetylation status (fast versus slow) with exposure to tobacco smoke (women who had ever smoked actively versus all the others studied) was 1.06 (P= 0.90). This finding is consistent with those of earlier studies \[[@B12],[@B14]-[@B17],[@B24],[@B25]\], in which acetylation status and active smoking were not significantly associated among controls. Table [2](#T2){ref-type="table"} provides demographic and risk factor characteristics for the 502 breast cancer patients in the analytic sample according to acetylation status. The distribution of age, family history of breast cancer, and history of benign breast disease was nearly identical for fast and slow acetylators. There were small differences in alcohol consumption and BMI between fast and slow acetylators and a noticeable difference in the proportion of women who had had one to three live births (46% versus 61%, respectively). The great majority of the participants (97%) were white (data not shown). We observed no substantial departure from multiplicativity between acetylation status and history of ever having smoked actively (adjusted OR estimate of departure from multiplicativity (AOR) = 0.9) or of ever having experienced passive residential exposure to tobacco smoke (AOR = 0.7) (Table [3](#T3){ref-type="table"}, which also shows confidence intervals). The ratios of the upper limits of the intervals to their lower limits were about 3 and 3.7 for the crude and adjusted estimates of effect, respectively (Table [3](#T3){ref-type="table"}). These ratios measure the precision of the estimates of effect and indicate adequate precision about these estimates. Estimates for the departure from multiplicativity between acetylation status and the various measures of intensity, duration, and timing of active and passive tobacco exposure are presented in Table [4](#T4){ref-type="table"}. For active smokers, we found estimates lacking consistent directionality. The AOR estimates for women in the categories with the highest intensity (packs/day) and greatest length (in years) of smoking were in opposite directions. For example, the departure from multiplicativity was above the null for women who had smoked two or more packs per day (AOR = 1.8) but below the null for women who had smoked for 40 or more years (AOR = 0.7). For the variables describing the age at which a woman began smoking, when she began smoking in relation to her first live birth, and the time elapsed since she quit smoking, we observed estimates of departure from multiplicativity both above and below the null. The estimates for departure from multiplicativity between acetylation status and the measures of duration and timing of passive exposure to tobacco also lacked consistency. For the variable describing the duration of passive exposure to tobacco in the residence, we observed null and less than null associations with slow acetylation status: AOR = 1.0 for \<20 years, 0.6 for 20 to \<40 years, and 0.8 for 40+ years. We observed a departure from multiplicativity between slow acetylation status and passive exposure occurring exclusively before a woman\'s first live birth (AOR = 1.9), and a positive departure for women whose first passive exposure to tobacco smoke occurred between the ages of 12 and 20 (AOR = 2.4). Discussion ========== In this, the largest case-only study to evaluate the interaction between acetylation status and exposure to tobacco smoke, we found little evidence to support a departure from multiplicativity between acetylation status and a history of active smoking for women with breast cancer. There is some suggestion that women who were slow acetylators were at higher risk from passive exposure to tobacco smoke before their first live birth than women who had never been either passive or active smokers. A similar positive departure was observed for women who were first passively exposed between the ages of 12 and 20. The effect estimates observed in this study for measures of intensity, duration, and timing of exposure showed no consistent pattern and in some instances were statistically unstable. Our study is one of only a few to assess the interaction between exposure to tobacco smoke and acetylation status in relation to breast cancer risk. Hunter and colleagues \[[@B16]\], in addition to examining the association between slow acetylation status and the risk of breast cancer, for which they reported a null association, found no evidence of an interaction between recent smoking status and NAT2 acetylation status among 706 postmenopausal women (cases and controls). Recently, two studies that removed passive smokers from the analysis of the unexposed group found suggestions of an interaction between tobacco exposure and acetylation status. Both reported a greater breast cancer risk among passive smokers who were fast acetylators \[[@B24],[@B25]\]. The findings among active smokers were not consistent, however. Morabia and colleagues \[[@B25]\] found that active smokers who were fast acetylators were at greater risk, whereas Chang-Claude and colleagues \[[@B24]\] found the greater risk from active smoking among slow acetylators. By parsing their cases into contingency tables (genotype by smoking group) and applying a case-only analysis, we obtained estimates of departure from multiplicativity for both studies very similar to ours, but with wider CIs. In the only other case-only analysis, Ambrosone and colleagues \[[@B12]\] found a strong positive departure from multiplicativity between acetylation status and smoking at an early age (\<18) and for smoking 20 or more cigarettes 20 years previously. Overall, we found departures from multiplicativity between acetylation status in relation to both active and passive exposure to tobacco smoke below the null. In subanalyses, we did find a positive departure from multiplicativity between acetylation status and smoking initiation between 14 and 15 years of age, and, separately, for first passive exposure to tobacco smoke between the ages of 12 and 20. Both findings are consistent with the hypothesis that environmental insults to developing breast tissue may increase the tissue\'s susceptibility to carcinogenesis, and thus may increase a woman\'s risk of breast cancer \[[@B38],[@B39]\]. However, the lack of a consistent directionality to our estimates for the other age-at-initiation categories (≤ 13 and 16 to 17 years) suggests that these may be chance findings. To date, numerous polymorphisms on the NAT2 gene have been identified (Table [1](#T1){ref-type="table"}), which has furthered our understanding of NAT2 phenotypes and improved our ability to assign acetylation status to the breast cancer cases in this study. The genotyping procedures employed in this analysis are more accurate than the PCR-RFLP (PCR--restriction fragment length polymorphism) techniques used in previous studies \[[@B12],[@B16],[@B25]\]. Consequently, the rates of misclassification of acetylation status in this study should be less than in those studies. Misclassification of either the genetic or environmental variables involved in an assessment of interaction by case--control design can give rise to the appearance of interaction when, in fact, there is none \[[@B40]\]. Our analysis of interaction using case-only data provides greater control over the impact of potential misclassification errors, because there are only two variables that are susceptible to misclassification -- acetylation status and smoking status. If the misclassification rates are nondifferential, as one would expect, then the estimates of departure from multiplicativity will be biased towards the null \[[@B41]\]. As discussed above, in the previous case--control analyses, the impact of misclassification is less predictable. It is therefore possible that findings from previous studies evaluating the interaction of acetylation status and exposure to tobacco smoke in relation to breast cancer risk may have generated spurious estimates of interaction, even if the misclassification was nondifferential. As discussed above, case-only estimates derived from these studies were similar to ours. The attenuation of the interaction after reanalysis using the case-only design further suggests that the published case--control results may have been more susceptible to misclassification. By genotyping more SNPs with a more accurate method and by implementing a case-only design, our analysis provides a more valid assessment of the multiplicative interaction between NAT2 genotype and exposure to tobacco smoke in relation to breast cancer. Weighing against this advantage of the case-only design is the limitation that only departure from multiplicativity can be assessed. Many epidemiologists weigh departure from additive interaction more heavily, arguing that the additive scale corresponds better to the biologic meaning of synergistic effects \[[@B42]\]. A further limitation of the case-only design is its reliance on the assumption that the genetic polymorphisms and environmental exposure are independent of one another \[[@B43]\]. Violations of this assumption can substantially distort the estimates of interaction. However, NAT2 polymorphisms and smoking history were not associated among the genotyped controls in the Cape Cod study or among the controls in earlier studies \[[@B12],[@B14]-[@B17],[@B24],[@B25]\]. The absence of association supports the assumption of independence required to validly estimate departure from multiplicativity with the case-only design. These results must be interpreted with the following additional limitations in mind. First, only 56% of eligible cases were available for analysis. Participation was not related to smoking status and although participation was related to age, age was not related to genotype. We expect that the selection of participants introduced no substantial bias, although we acknowledge that our study of breast cancer survivors may have influenced the estimates of effect in ways that we are unable to anticipate. Second, haplotypes were inferred from genotyping data by assigning the haplotype with the maximum probability to each case. Forty-one percent of haplotype assignments had probabilities of 100% and 91% had probabilities of 80% or better. Less than 5% had probabilities of less than 50%. We expect that the procedure used to infer haplotypes introduced little error. Conclusion ========== This large case-only analysis is the first to be able to assign acetylation status on the basis of ten SNPs. No previous analysis assigned acetylation status on the basis of more than four. In addition, the study involved the largest number of breast cancer cases used to investigate the interaction between NAT2 acetylation status and exposure to tobacco smoke as related to breast cancer risk. The combination of the most complete genotyping data and the large case-only design provides important advantages, the results of which do not suggest any substantial interaction between acetylation status and exposure to tobacco smoke in the etiology of breast cancer. Weighing against the null result is the potential for an unanticipated bias towards the null to have arisen by selection of breast cancer survivors from among the incident cases. Abbreviations ============= AOR = adjusted odds ratio estimate of departure from multiplicativity; BMI = body mass index; CI = confidence interval; NAT = N-acetyltransferase; OR = odds ratio; PCR--RFLP = polymerase chain reaction--restriction fragment length polymorphism; SNPs = single-nucleotide polymorphisms. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= TL conceived of the study, collected genotyping samples, participated in data analysis, and drafted the manuscript. BB conducted data analysis and drafted the manuscript. JW inferred haplotypes from the genotyping data using SNPHAP and drafted the manuscript. AA collected interview data from the Cape Cod population and drafted the manuscript. All authors read and approved the final manuscript. Acknowledgements ================ This project was primarily supported by grant K07 CA87724 from the National Cancer Institute, National Institutes of Health. Interview data for the Cape Cod population was collected with support by grant number 2P42 ES07381 from the National Institute of Environmental Health Sciences, National Institutes of Health, with funds from the Environmental Protection Agency. Interview data for the second population were collected with support from grants R01 CA/AG70818 from the National Cancer Institute and National Institute on Aging, National Institutes of Health. The publication\'s contents are solely the responsibility of the authors and do not necessarily reflect the official views of the NIEHS, the NIH, or the EPA. Figures and Tables ================== ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Distribution of NAT2 haplotypes and their gene product acetylator phenotype in 502 breast cancer patients ::: Acetylator phenotype Haplotype References ---------------------- ----------- ------------ ------------------- -------------- Status No. \% Fast 24 4.8 NAT2\4/NAT2\4 31--34 Intermediate 158 31.5 All intermediate 14 148 93.7 NAT2\4/- 35 10 6.3 NAT2\12a/- Slow 320 63.7 All slow 14,33,36 81 24.6 NAT2\5b/NAT2\5b 139 43.4 NAT2\5b/NAT2\6a 37 11.2 NAT2\5b/- 14,31--33,36 43 13.1 NAT2\6a/NAT2\6a 18 6.1 NAT2\6a/- 14,36,37 2 3.3 NAT2\5c/- -, any other haplotype yielding heterozygosity; NAT, N-acetyltransferase. ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Distribution of breast cancer risk factors according to arylamine N-acetyltransferase activity (acetylation status) ::: NAT2 ------------------------------------------------- ---------- ---------- ----- ------------ Age (years) \<50^a^ 22 (12) 38 (12) 1.0 \- 50 to 59 15 (8) 28 (9) 1.1 0.5 to 2.4 60 to 69 59 (32) 103 (32) 1.0 0.5 to 1.9 70+ 86 (47) 151 (47) 1.0 0.6 to 1.8 Alcohol use Nondrinker^a^ 45 (14) 32 (18) 1.0 \- ≤ 1 drink/month 75 (23) 43 (24) 1.2 0.7 to 2.2 Few drinks/month 77 (24) 39 (21) 1.4 0.8 to 2.5 Few drinks/week 76 (24) 39 (21) 1.4 0.8 to 2.5 Almost every day 40 (12) 24 (13) 1.2 0.6 to 2.3 Unknown 7 (2) 5 (3) 1.0 0.3 to 3.4 Body mass index (BMI) (kg/m^2^) \<20.0^a^ 30 (17) 59 (19) 1.0 \- 20.0 to 24.9 114 (65) 196 (62) 0.9 0.5 to 1.4 25.0 to 29.9 24 (14) 50 (16) 1.1 0.5 to 2.0 30.0+ 8 (4) 11 (3) 0.7 0.3 to 1.9 First-degree family history of breast cancer^b^ No^a^ 140 (78) 240 (77) 1.0 \- Yes 39 (22) 72 (23) 1.1 0.7 to 1.7 Parity Nulliparous^a^ 72 (23) 27 (15) 1.0 \- 1 live birth 22 (7) 22 (12) 0.4 0.2 to 0.8 2 live births 56 (17) 43 (24) 0.5 0.3 to 0.9 3 live births 70 (22) 48 (25) 0.5 0.3 to 1.0 4 live births 46 (14) 21 (12) 0.8 0.4 to 1.6 5 or more live births 54 (17) 21 (12) 1.0 0.5 to 1.9 Geographical location Cape Cod, Massachusetts^a^ 91 (50) 178 (56) 1.0 \- Los Angeles, California 27 (15) 33 (10) 0.6 0.3 to 1.1 Rhode Island 17 (9) 38 (12) 1.1 0.6 to 2.1 Minnesota 28 (15) 34 (11) 0.6 0.3 to 1.1 North Carolina 19 (10) 37 (12) 1.0 0.5 to 1.8 History of benign breast disease No^a^ 116 (65) 206 (65) 1.0 \- Yes 63 (35) 110 (35) 1.0 0.7 to 1.4 ^a^Reference level. ^b^Includes a woman\'s mother, sister(s), and/or daughter(s). -, not calculated; CI, confidence interval; NAT, N-acetyltransferase; OR, odds ratio. ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Departure from multiplicativity between acetylation status and smoking status among patients with breast cancer ::: Acetylation status Crude statistic Adjusted statistic^a^ ---------------- -------------------- ----------------- ----------------------- -------------- ------ -------------- Active smoker 162 85 1.04 0.59 to 1.82 0.88 0.45 to 1.70 Passive smoker 114 73 0.85 0.48 to 1.52 0.75 0.39 to 1.45 Nonsmoker 44 24 1.0 \- 1.0 \- ^a^Controlling for age, alcohol consumption, body mass index, family history of breast cancer, family history of benign breast disease, parity, and geographical location. -, not calculated; CI, confidence interval; OR, odds ratio. ::: ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Departure from multiplicativity between acetylation status and smoking status among patients with breast cancer ::: Acetylation status --------------------------------------------- -------------------- ---- ------ ------ -------------- Nonsmokers* 44 24 1.0 1.0 \- *Active smokers* Packs (per day) \<1 96 47 1.11 0.97 0.48 to 1.95 1 to \<2 53 33 0.88 0.74 0.35 to 1.60 ≥ 2 12 2 3.27 1.80 0.33 to 9.81 Data missing 1 3 Duration (years) \<20 39 29 0.73 0.59 0.26 to 1.35 20 to \<40 79 25 1.72 1.32 0.62 to 2.81 40+ 43 30 0.78 0.74 0.33 to 1.63 Data missing 1 1 Duration (years) in relation to first birth Nulliparous 41 11 2.03 1.05 0.32 to 3.41 All before first 11 8 0.75 0.60 0.17 to 2.10 Before and after first 90 48 1.02 1.03 0.48 to 2.20 All after first 19 17 0.61 0.52 0.19 to 1.37 Data missing 1 1 Age started (years) ≤ 13 4 3 0.73 0.63 0.11 to 3.48 14 to 15 22 7 1.71 1.94 0.61 to 6.19 16 to 17 35 18 1.06 0.79 0.32 to 1.94 18 to 21 67 35 1.04 0.94 0.45 to 1.99 22 to 29 19 11 0.94 0.84 0.31 to 2.30 ≥ 30 14 10 0.76 0.60 0.21 to 1.71 Data missing 1 1 Quit before diagnosis date (years) Current or \<5 32 12 1.45 1.25 0.50 to 3.19 5 to 15 37 19 1.06 0.82 0.35 to 1.95 \>15 71 42 0.92 0.86 0.41 to 1.80 Data missing 22 12 *Passive smokers* Duration (years)^b^ \<20 34 18 1.03 0.99 0.43 to 2.33 20 to \<40 48 36 0.73 0.61 0.29 to 1.28 40+ 29 18 0.88 0.81 0.33 to 2.01 Data missing 3 1 Duration (years) in relation to first birth Nulliparous 20 14 0.78 0.52 0.16 to 1.69 All before first 20 8 1.36 1.85 0.59 to 5.83 Before and after first 56 34 0.90 0.91 0.39 to 2.10 All after first 6 5 0.65 0.39 0.09 to 1.73 Data missing 12 12 First lived with smoker (age in years) \<12 51 35 0.84 0.83 0.29 to 2.34 12 to 20 18 5 3.11 2.43 0.41 to 14.3 \>20 35 28 0.89 0.48 0.17 to 1.42 Data missing 10 5 ^a^Controlling for age at diagnosis, alcohol use, BMI, first-degree family history of breast cancer, history of benign breast disease, and parity. ^b^Length of time lived with a smoker. -, not calculated; CI, confidence interval; OR, odds ratio. :::	PubMed Central	2024-06-05T03:55:57.448707	2005-3-21	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143561/", "journal": "Breast Cancer Res. 2005 Mar 21; 7(3):R385-R393", "authors": [ { "first": "Timothy L", "last": "Lash" }, { "first": "Brian D", "last": "Bradbury" }, { "first": "Jemma B", "last": "Wilk" }, { "first": "Ann", "last": "Aschengrau" } ] }
PMC1143570	Introduction ============ Breast cancer is the most prevalent cancer in women worldwide, excluding nonmelanoma skin cancer, and is the second leading cause of cancer deaths in women (following lung cancer) \[[@B1]\]. Once metastasis has occurred, the survival rate is drastically reduced to a median of 2--3 years; therapy is then aimed at controlling symptoms, prolonging survival and improving quality of life \[[@B2]\]. Unfortunately, the diagnostic criteria currently used to stage breast cancer often yield inaccurate findings with regard to metastasis. Analyses of bone marrow samples (not a routine procedure) have revealed the presence of disseminated cells in up to 40% of primary breast cancer patients without any clinical or histopathologic signs of metastasis. Circulating breast cancer cells in bone marrow are indicative of metastasis to such sites as bone, lung, and liver \[[@B3]\]. Cancer cells form tumors and spread by degrading the extracellular matrix (ECM) through various matrix metalloproteinases (MMPs). The activity of these enzymes correlates with the aggressiveness of tumor growth and the invasiveness of the cancer. Rath and Pauling \[[@B4]\] postulated that nutrients such as lysine and ascorbic acid could act as natural inhibitors of ECM proteolysis, and as such they have the potential to inhibit tumor growth and expansion. These nutrients may exert their antitumor effect through inhibiting MMPs and strengthening the connective tissue surrounding cancer cells (a \'tumor encapsulating\' effect). Additionally, it has been suggested that, through inhibition of hyaluronidase, ascorbic acid can prevent metastases by preventing degradation of the ground substance surrounding the tumor. In a previous study \[[@B5]\] we demonstrated the antiproliferative and anti-invasive potential of a specific formulation (nutrient supplement \[NS\]) of lysine, ascorbic acid, proline, and green tea extract on human breast cancer (MDA-MB 231), colon cell cancer (HCT 116), and melanoma (A2058) cell lines. NS also suppressed the growth of these tumors, without any adverse effects, in nude mice. In the present study we investigated the inhibitory effect of NS in vivoon development of N-methyl-N-nitrosourea (MNU)-induced mammary tumors in rats. Methods ======= Animals ------- On arrival at our laboratory, 40-day-old pathogen free female Sprague--Dawley rats (Simonsen Laboratories, Gilroy, CA, USA) were housed in solid bottom cages with corncob bedding, at 22°C and 50% humidity, with a 12-hour light--dark cycle. The rats had free access to water and Purina rat chow diet. All animals were cared for in accordance with institutional guidelines for the care and use of experimental animals. Experiments ----------- At day 50, all rats (n= 20) received a single dose of MNU 50 mg/kg intraperitoneally. (MNU, reagent grade, was obtained from Sigma, St. Louis, MO, USA). Two weeks after MNU treatment, a time by which the animals had recovered from MNU-induced toxicity, the rats were divided into two groups. Rats in group 1 (n= 10) were fed a pellet Purina chow diet (Purina Mills, Richmond, IN, USA), whereas those in group 2 (n= 10) were fed a pellet diet custom prepared by Purina containing the same diet but supplemented with 0.5% of the NS. Body weight and diet consumption of the rats were monitored every week. Beginning 6 weeks after MNU administration, the rats were palpated every week for evidence of tumors. Dimensions (length × width) of the tumors were measured using a digital caliper, and the tumor burden was calculated using the following formula: 0.5 × length × width. Twenty-four weeks later the rats were killed by carbon dioxide asphyxiation and skinned; tumors were excised and a detailed necropsy was performed on each rat. Location, weight, and dimensions of excised mammary tumors were recorded. Tumors were processed for histologic examination, using criteria described by Russo and coworkers \[[@B6]\]. Briefly, histopathologic criteria used to determine malignancy were loss of tubular--alveolar pattern of the normal mammary gland; presence of large epithelial cells with increased nuclear--cytoplasmic ratio; stromal response by fibrosis and inflammatory cell infiltration; and necrosis and hemorrhage. Composition of the nutrient supplement -------------------------------------- Stock solution of the NS (total weight 4.2 g) is composed of the following: 700 mg vitamin C (as ascorbic acid and as magnesium, calcium, and palmitate ascorbate), 1000 mg L-lysine, 750 mg L-proline, 500 mg L-arginine, 200 mg N-acetyl cysteine, 1000 mg standardized green tea extract, 30 mg selenium, 2 mg copper, and 1 mg manganese. Green tea extract, derived from green tea leaves, was obtained from US Pharma Lab (Newark, NJ, USA). The certificate of analysis indicates the following characteristics: total polyphenol 80%, catechins 60%, epigallocatechin-3-gallate (EGCG) 35%, and caffeine 1.0%. Statistical analysis -------------------- Results are expressed as means ± standard deviation for the groups. Data were analyzed by independent samples t-test. Results ======= Tumor incidence and multiplicity -------------------------------- Of the 10 rats in the control group, nine developed at least one tumor; the total number of tumors in that group was 19. In contrast, five of the supplemented rats were completely free of tumors, and the total number of tumors in that group was six. As shown in Table [1](#T1){ref-type="table"}, treatment with nutrients significantly reduced the incidence of MNU-induced mammary tumors and the number of tumors per rat (tumor multiplicity) by 68.4%. Tumor burden ------------ Tumor burden (tumor length × width × 0.5) in MNU-induced mammary tumors was inhibited by nutrient synergy by 60.5% (P= 0.0001) as shown in Table [2](#T2){ref-type="table"}. The mean tumor burden per rat for the control group was 18.3 cm^2^, in contrast to a mean tumor burden of 7.22 cm^2^in the nutrient supplemented group (P\< 0.0001). Tumor weight ------------ The inhibitory effect of NS was also reflected in a decreased tumor weight (Table [3](#T3){ref-type="table"}). For example, tumor weight per rat and per group were decreased by 41% and 78%, respectively (P= 0.0001). The mean tumor weight per rat in the control group (4.34 g) was significantly greater (P= 0.002) than that in the supplemented group (0.97 g). The mean individual tumor weight also differed significantly between groups. Rat growth ---------- We observed no significant difference (P= 0.62) in growth between groups over the period of study (Table [4](#T4){ref-type="table"}). Tumor histology --------------- The tumors that developed in the control group rats were all adenocarcinomas (Fig. [1](#F1){ref-type="fig"}). The lesions are rather cellular and consist of a proliferation of epithelial and stromal components; the epithelial elements range from reactive to malignant in nature. The majority of the tumor is adenomatous and exhibits features of florid sclerosing adenosis admixed with low-grade ductal carcinoma in situ. Focal areas exhibit features diagnostic of adenocarcinoma, including increased mitotic index, atypical mitotic figures, moderate to severe cytologic atypia, coagulative tumor cell necrosis, and jagged infiltrating margins. The lesion is well circumscribed in areas and is associated with a brisk host response composed of reactive stromal cells and a mixed inflammatory infiltrate. Angiogenesis is much more prominent than in the adenomas in the supplemented rats. In contrast, the tumors that developed in the nutrient supplemented rats were all adenomas (Fig. [2](#F2){ref-type="fig"}). The lesion is moderately cellular and consists of epithelial and stromal components. The overall cytoarchitectural features are indicative of a fibroepithelial lesion, such as a fibroadenoma. The epithelial cells exhibit mild to moderate cytologic atypia and a low mitotic index. The lesion is well circumscribed and exhibits prominent papillary architecture. The stromal and vascular proliferation is much less than that seen in the adenocarcinoma. Discussion ========== We chose to study the effect of a mixture of nutrients on MNU-induced mammary tumors in the Sprague--Dawley rat model because the histologic structure of mammary gland tumors in this animal closely resembles that of human mammary tumors. Induction of mammary carcinomas by MNU in female rats is one of the most frequently used animal models for the investigation of breast carcinogenesis and mammary tumor treatment \[[@B7]-[@B9]\]. In contrast to mouse lesions, which are primarily alveolar, rat mammary tumors are predominantly ductal, as are human ones \[[@B9]\], and the most highly malignant rat tumors share some features with human intraductal and infiltrating ductal carcinomas \[[@B8]\]. It has been reported that the MNU model has several advantages, such as reliability of tumor induction, organ site specificity, tumor of ductal origin and predominantly carcinomatous histopathologic characterization, and the ability to examine tumor initiation and promotion processes \[[@B10]\]. Generally MNU-induced mammary carcinomas are aggressive and locally invasive. The results of the present study demonstrate significant inhibition of mammary tumor incidence and multiplicity in Sprague--Dawley female rats by supplementation with 0.5% of the nutrient mixture (which contains ascorbic acid, lysine, proline, and epigallocatechin gallate). Furthermore, rats that consumed the nutrient supplemented diet exhibited decreased growth of MNU-induced mammary tumors and tumor burden compared with rats fed control diets. The numerous tumors in the control rats not only were larger but also had characteristics diagnostic of adenocarcinoma, including increased mitotic index and prominent angiogenesis, in contrast to the few, small adenomas with low mitotic index found in the nutrient treated rats. Although the mechanism underlying the reduced tumor size in tumor bearing rats was not identified in this experiment, these findings are consistent with our previous in vitrostudies that demonstrated significant inhibition of angiogenic and invasive parameters in human breast cancer cell lines MDA MB-231 and MCF-7. Expression of vascular endothelial growth factor, MMP secretion, and matrix invasion by these breast cancer cells were dramatically inhibited in a dose-dependent manner by the combined effect of the nutrients in this mixture \[[@B11]\]. Matrix invasion can be controlled by inhibition of MMP expression, as well as by increasing connective tissue strength and stability, contributing to the \'encapsulation\' of the tumor. Optimization of synthesis and structure of collagen fibrils depends upon hydroxylation of proline and lysine residues in collagen fibers. It is well known that ascorbic acid is essential for the hydroxylation of these amino acids and that it regulates collagen synthesis at the transcriptional level. Inhibitory and chemopreventive effects in malignant cell lines of some of the individual nutrients composing the NS have been reported in both clinical and experimental studies. Ascorbic acid has been shown to have growth inhibitory and antineoplastic activities in human mammary tumor bearing mice \[[@B12]\]. In addition, low levels of ascorbic acid have been reported in cancer patients \[[@B13]-[@B15]\]. Green tea extract is another potent anticancer agent that has been reported to have a growth inhibitory effect against certain human cancer cell lines, especially breast cancer \[[@B16]-[@B18]\]. For example, both in vitroand animal studies of the effect of green tea extract on breast cancer revealed suppressed xenograft size and tumor vessel density and suppression of cell proliferation \[[@B19]\]. Furthermore, studies conducted before the clinical onset of breast cancer found that increased green tea consumption was associated with improved prognosis in stage I and II breast cancers, as well as decreased numbers of axillary lymph node metastases in premenopausal women, suggesting significant chemopreventative potential \[[@B20]\]. Utilizing the Sprague--Dawley rat model, researchers showed that EGCG reduced tumor burden and number of invasive tumors, and drastically increased the mean latency to the initial tumor in mammary tumor bearing rats \[[@B21]\]. Our previous in vitrostudies demonstrated that the anticancer effect of a mixture of ascorbic acid, proline, lysine, and EGCG on several cancer cell lines in tissue culture studies was greater than that of the individual nutrients \[[@B5]\]. Furthermore, in contrast to chemotherapy, which causes indiscriminate cellular and ECM damage, previous studies showed that cell morphology was not affected even at the highest concentrations of this nutrient mixture, demonstrating that this formulation is safe to cells. Conclusion ========== The results of the present study showed that the specific nutrient mixture of lysine, proline, arginine, ascorbic acid, and green tea extract tested significantly inhibited the incidence, as well as the growth, of MNU-induced mammary tumors. Although clinical trials are necessary to assess the antitumor ability of the tested nutrient mixture on cancer patients, the results of this study suggest strong potential for its use as a therapeutic regimen for inhibiting breast cancer development. Abbreviations ============= ECM = extracellular matrix; EGCG = epigallocatechin-3-gallate; MMP = matrix metalloproteinase; MNU = N-methyl-N-nitrosourea; NS = nutrient supplement. Competing interests =================== This research was funded by Matthias Rath Inc. Authors\' contributions ======================= MWR carried out tumor burden, tumor weight, and tumor multiplicity laboratory studies. NR assisted in laboratory studies. VI designed the study. TK drafted the manuscript and performed the statistical analysis. AN and MR conceived the study and participated in its coordination. All authors read and approved the final manuscript. Acknowledgements ================ We thank consulting pathologist Dr Kendall Price (Stanford University) for providing the pathology report on the mammary tumors. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Histological slides showing examples of N-methyl-N-nitrosourea induced mammary tumors in Sprague--Dawley rats fed the control diet (hematoxylin and eosin). Original magnifications: (a)100× and (b)200×. ::: ![](bcr989-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Histological slides showing examples of N-methyl-N-nitrosourea induced mammary tumors in Sprague--Dawley rats fed the supplemented diet (hematoxylin and eosin). Original magnifications: (a)100× and (b)200×. ::: ![](bcr989-2) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Tumor incidence and multiplicity for control and supplemented rats ::: Rat group 0 Tumors 1 Tumor 2 Tumors 3 Tumors 4 Tumors 5 Tumors Total number of Tumors ------------------------------ ---------- --------- ---------- ---------- ---------- ---------- ------------------------ Control group (n= 10) 1 4 3 0 1 1 19 Supplemented group (n= 10) 5 4 1 0 0 0 6 ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Mean tumor burden per rat and tumor burden per group ::: Rat group Total tumor burden per group Mean tumor burden per rat ------------------------------------------ ------------------------------ --------------------------- Control group (n= 10; 19 tumors) 183.2 cm^2^ 18.3 ± 1.3 cm^2^ Supplemented group (n= 10; six tumors) 72.2 cm^2^ 7.22 ± 1.8 cm^2^ The mean tumor burden per rat is expressed as mean ± standard deviation. The differences between groups were significant (P\< 0.0001). ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Mean tumor weight per rat and per group ::: Rat group Total tumor weight per group Mean tumor weight per rat Mean individual tumor weight ------------------------------------------ ------------------------------ --------------------------- ------------------------------ Control group (n= 10; 19 tumors) 43.38 g 4.34 ± 1.5 g 2.3 ± 0.8 g Supplemented group (n= 10; six tumors) 9.63 g 0.97 ± 2.6 g 0.73 ± 0.5 g Significance P= 0.002 P= 0.002 ::: ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Mean growth of control and supplemented rats ::: Rat group Mean initial weight Mean end weight ------------------------------ --------------------- ----------------- Control group (n= 10) 143 ± 4 g 422 ± 24 g Supplemented group (n= 10) 148 ± 5 g 415 ± 37 g :::	PubMed Central	2024-06-05T03:55:57.453152	2005-1-31	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143570/", "journal": "Breast Cancer Res. 2005 Jan 31; 7(3):R291-R295", "authors": [ { "first": "M Waheed", "last": "Roomi" }, { "first": "Nusrath W", "last": "Roomi" }, { "first": "Vadim", "last": "Ivanov" }, { "first": "Tatiana", "last": "Kalinovsky" }, { "first": "Aleksandra", "last": "Niedzwiecki" }, { "first": "Matthias", "last": "Rath" } ] }
PMC1143572	Introduction ============ The majority of breast tumours express oestrogen receptors (ERs). Several studies have shown that 5 years of tamoxifen therapy in breast cancer patients with receptor-positive tumours reduces the risk of recurrence and mortality \[[@B1]\]. However, about 30% of patients acquire tamoxifen resistance and relapse in the disease \[[@B1]\]. Several possible mechanisms for this have been suggested \[[@B2]-[@B4]\]. Tamoxifen and its metabolites compete with endogenous oestrogen for the ligand-binding domain of the ER. The complex formation between tamoxifen, or its active metabolites, and the ER inhibits recruitment of co-activator complexes necessary for transcription of oestrogen-responsive genes \[[@B5]\]. The biotransformation of tamoxifen is mediated by cytochrome P450 enzymes mainly through demethylation and hydroxylation to form several primary metabolites, principally 4-OH-tamoxifen, α-OH-tamoxifen, N-desmethyl-tamoxifen, and 4-OH-N-desmethyl-tamoxifen. 4-OH-tamoxifen is considered to be a more potent anti-oestrogen than the mother substance and is capable of binding the ER with greater affinity \[[@B6],[@B7]\]. From experimental studies it has been shown that the transformation of tamoxifen into 4-OH-tamoxifen is mainly catalysed by the liver enzyme CYP2D6 \[[@B8],[@B9]\]. A further step in the metabolism of tamoxifen is sulfate conjugation, catalysed by members of the sulfotransferase family, which generally increase the solubility and facilitate excretion of the drug. Sulfotransferase 1A1 (SULT1A1) is a major form of phenol sulfotransferase in the adult human liver, and it has been shown to be the primary sulfotransferase responsible for the sulfation of 4-OH-tamoxifen \[[@B10],[@B11]\]. Polymorphisms affecting the enzyme activity have been found in both cytochrome P450 2D6 (CYP2D6) and SULT1A1\[[@B12],[@B13]\]. Among Caucasians the most frequent inactivating polymorphism in CYP2D6is the CYP2D6\4allele, which generates a G → A transition at nucleotide 1934 leading to a disruption of the reading frame and to a truncated non-functional gene product \[[@B14]\]. The most common polymorphism in the SULT1A1gene is a G → A transition at nucleotide 638, resulting in an arginine to histidine substitution at the conserved amino acid 213. This allele, SULT1A1\2, is correlated with diminished capacity to sulfate SULT1A1 substrates \[[@B15]\]. The aim of the present study was to investigate the genotypes of CYP2D6and SULT1A1in breast cancer patients with and without tamoxifen treatment in order to validate the relation between the genotype and the benefit from tamoxifen therapy. Materials and methods ===================== Patients -------- The Stockholm Breast Cancer Group started a trial in 1976 to compare postoperative radiotherapy with adjuvant chemotherapy \[[@B16]\]. Both premenopausal patients and postmenopausal patients (age ≤ 70 years) with a unilateral, operable breast cancer were included. The patients were required to have either histological verified lymph node metastases or a tumour diameter exceeding 30 mm. All patients were treated with a modified radical mastectomy. Using a 2 × 2 factorial study design, the postmenopausal patients were then randomised to a comparison of adjuvant tamoxifen treatment or no endocrine treatment in a total of four treatment groups: adjuvant chemotherapy, adjuvant chemotherapy plus tamoxifen, radiotherapy, and radiotherapy plus tamoxifen. Tamoxifen was given postoperatively at a dose of 40 mg daily for 2 years and was initiated within 4--6 weeks of surgery. The mean follow-up time was 10.7 years (range, 0.24--18.6 years). Of the 679 postmenopausal breast cancer patients included in the trial, fresh frozen tumour tissues of 226 patients were available for the current investigation, of whom 112 had received tamoxifen therapy. The number of distant recurrences was 64 in the tamoxifen-treated group and 84 in the group not receiving tamoxifen. Furthermore, the fractions of lymph-node-positive and ER-positive tumours were 88/89 and 71/70, respectively, and the percentage of large tumours (\>20 mm) was 57/61 in the initial study and the current study. Polymerase chain reaction ------------------------- DNA was isolated from fresh-frozen tumour tissues using phenol, phenol/chloroform (1:1), and chloroform, was precipitated with ethanol and was re-dissolved in sterile water. The CYP2D6and SULT1A1genes were amplified with PCR in separate reactions using 30 ng DNA and 60 ng DNA, respectively. The primer sequences used in the PCR of CYP2D6and SULT1A1were adopted from Hanioka and colleagues \[[@B14]\] and Coughtrie and colleagues \[[@B17]\]. The following PCR reagents were added to a reaction volume of 20 μl: 2 mM MgCl~2~, 0.2 mM dNTPs, 0.5 units TaqDNA polymerase, and 1 μM each of forward and reverse primer in 1 × PCR buffer. The amplifications were carried out in a PTC-200 Peltier Thermal Cycler DNA Engine (MJ Research™ Inc, Waltham, MA, USA). An initial denaturation at 94°C for 3 min was followed by 40--43 cycles of 30 s at 94°C, 30 s of annealing at 63°C, and 40 s for extension at 72°C. An extension period of 5 min followed the final cycle. Restriction fragment length polymorphism ---------------------------------------- The CYP2D6and SULT1A1polymorphisms were detected using restriction enzymes. The MvaI enzyme distinguishes between the CYP2D6\4allele and other CYP2D6alleles. The polymorphic allele CYP2D6\4lacks the restriction site, and is thereby retained as one fragment. Alleles harbouring the MvaI restriction site generate two fragments and are classified as CYP2D6\1. SULT1A1\1(wild-type allele) has a restriction site recognised by the HaeII enzyme, while the polymorphic SULT1A1\2lacks this site. Ten units of MvaI (Fermentas, Stockholm, Sweden) and 1.5 μl R^+^Buffer (Fermentas) were added to each tube of CYP2D6PCR products and were incubated at 37°C for 2.5 hours. The SULT1A1PCR products were incubated with 5 units of the restriction enzyme HaeII (New England BioLabs, Beverly, MA, USA) in a 20 μl reaction mixture containing 1 × NE (50 mM potassium acetate, 20 mM Tris-acetate, 10 mM magnesium acetate, 1 mM dithiothreitol, pH 7.9) buffer (New England Bioloabs), supplemented with 100 μg/ml BSA. After digestion, fragments were resolved by electrophoresis on a 3% (w/v) agarose gel containing 1 × TBE (89 mM Tris, 89 mM Boric acid, 2 mM EDTA, pH 8.4) buffer and ethidium bromide (0.5 μg/μl). A 100-molecule weigh ladder was used as the base pair marker. The gel was finally processed in a UV detector (Spectromics Corporation, New York, USA). To confirm the reliability of the restriction fragment length polymorphism method, a number of randomly selected samples were DNA sequenced. No differences in genotype were obtained between the methods. Statistical analyses -------------------- Statistical analyses were performed with the Statistica 6.0 software program (Statsoft Inc., Tulsa, OK, USA). We compared distant recurrence-free survival by genotype and by endocrine treatment with the log-rank test. The relative risk (RR) of distant recurrences among ER-positive patients treated with and without tamoxifen was assessed using Cox proportional hazard regression, and adjustments for age, tumour size, and lymph node status were performed. Results ======= Information on tumour size, nodal involvement, ER status and tamoxifen therapy of 226 patients is presented in Table [1](#T1){ref-type="table"}. The patients were genotyped according to the CYP2D6\4and the SULT1A1\1or SULT1A1\2alleles. There were no significant differences in tumour characteristics between genotypes (Table [1](#T1){ref-type="table"}). The distributions of allele frequencies were 0.163 and 0.386 for CYP2D6\4and SULT1A1\2, respectively. Since the CYP2D6\4homozygotes were few, patients with at least one CYP2D6\4allele were combined in the statistical analyses. Similarly, patients carrying the SULT1A1\2allele were grouped together. To investigate whether the genotype had a prognostic value, in terms of distant recurrence-free survival, ER-positive and ER-negative patients homozygous for CYP2D6\1alleles were compared with carriers of the CYP2D6\4allele, and the patients homozygous for the SULT1A1\1allele were compared with carriers of the SULT1A1\2allele. No statistical differences in distant recurrences were found according to genotype (data not shown). To assess the benefit from tamoxifen treatment, distant recurrence-free survival was only calculated in ER-positive patients. Distant recurrence-free survival for CYP2D6\1homozygotes, for CYP2D6\4heterozygotes and homozygotes, for SULT1A1\1homozygotes, and for SULT1A1\2heterozygotes and homozygotes are shown in Figs [1a,b](#F1){ref-type="fig"} and [2a,b](#F2){ref-type="fig"}, respectively, and are presented in Table [2](#T2){ref-type="table"}. Patients possessing at least one CYP2D6\4allele had better survival when randomised to tamoxifen compared with those who were not randomised to tamoxifen (P= 0.0089), as also demonstrated by the significantly decreased relative risk (RR = 0.28, 95% confidence interval \[CI\] = 0.11--0.74). Among patients homozygous for the CYP2D6\1genotype, the outcome was approximately equal between tamoxifen-treated and non-tamoxifen-treated patients (P= 0.75). A tendency towards improved distant recurrence-free survival in SULT1A1\1homozygous patients treated with tamoxifen, compared with those receiving no tamoxifen, was found (P= 0.074, RR = 0.48, 95% CI = 0.21--1.12) (Fig. [2a](#F2){ref-type="fig"}). Finally, no influence of tamoxifen therapy on distant recurrence-free survival was found in carriers of the SULT1A1\2allele (P= 0.48) (Fig. [2b](#F2){ref-type="fig"}). The genotypes linked to the benefit from tamoxifen treatment are combined in Fig. [3](#F3){ref-type="fig"} as well as in Table [2](#T2){ref-type="table"}. In patients harbouring the combination with at least one CYP2D6\4allele and/or a homozygous SULT1A1\1, tamoxifen treatment significantly improved survival (P= 0.0041, RR = 0.38, 95% CI = 0.19--0.74). We also compared non-beneficial alleles (i.e. CYP2D6\1homozygotes and SULT1A1\2carriers), and no statistical difference was found in distant recurrence-free survival (P= 0.57, RR = 1.22, 95% CI = 0.61--2.4). A comparison of the RRs of distant recurrence, calculated for each combined genotype and adjusted for age, tumour size and lymph node status, demonstrated that the risk reduction with tamoxifen was significantly higher in patients harbouring the combination of CYP2D6\4and/or SULT1A1\1/\1(P= 0.018) (Fig. [3a,b](#F3){ref-type="fig"} and Table [2](#T2){ref-type="table"}). Discussion ========== We observed a significantly improved benefit from tamoxifen in patients carrying the CYP2D6\4allele and/or patients homozygous for SULT1A1\1(P= 0.018), compared with patients homozygous for the CYP2D6\1and carriers of the SULT1A1\2allele (Fig. [3](#F3){ref-type="fig"}). To our knowledge this is the first report of the influence of the CYP2D6genotype on tamoxifen therapy, while the influence of the SULT1A1\1allele has been investigated by Nowell and colleagues \[[@B13]\]. In agreement with the tendency found in the present report, Nowell and colleagues showed that the high-activity allele SULT1A1\1contributed significantly to tamoxifen response \[[@B13]\]. Those authors suggested that sulfation may affect bioavailability of 4-OH-tamoxifen by reduced clearance of the sulfated metabolite. This may provide a genotype-dependent reservoir of inactivated metabolite, which can be desulfated by steroid sulfatase expressed in breast tumours and can be recovered to the active 4-OH-tamoxifen, leading to a prolonged anti-oestrogen effect \[[@B18]\]. Coller and colleagues \[[@B19]\] and other workers \[[@B8],[@B9]\] have demonstrated in experimental studies that the CYP2D6genotype is a determinant of the ability to form 4-OH-tamoxifen. However, a clinical study by Stearns and colleagues \[[@B20]\] revealed that inhibition of CYP2D6 had no significant effect on 4-OH-tamoxifen concentration. We propose in the present study that genotypes of CYP2D6, which produce a large amount of the ER-active 4-OH-tamoxifen, would be beneficial for the tamoxifen-treated patients. As shown in the present study, patients with at least one CYP2D6\4allele demonstrated better response to tamoxifen treatment than patients homozygous for the CYP2D6\1allele. This is in contrast to the main hypotheses where the CYP2D6\1homozygous patients are supposed to generate the active metabolite 4-OH-tamoxifen more readily and thereby have improved response of tamoxifen. Our results were obtained from a small number of patients, and therefore the association of the genotype and the benefit of tamoxifen treatment may be a coincidence. An absent or decreased CYP2D6-dependent 4-hydroxylation is, however, compensated by CYP2C9 and CYP3A4 to the overall formation of 4-OH-tamoxifen, but the reaction proceeds at a lower rate \[[@B19],[@B21]\]. Interestingly, an additional active tamoxifen metabolite, 4-OH-N-desmethyl-tamoxifen (endoxifen), has been recently discovered by Stearns and colleagues \[[@B20]\]. Endoxifen may have clinical relevance since the metabolite inhibits MCF7 cell proliferation with equal potency as does 4-OH-tamoxifen, and it is present in higher plasma concentration in humans than 4-OH-tamoxifen. Endoxifen is mainly synthesised by CYP3A4-mediated N-demethylation of tamoxifen and a subsequent 4-hydroxylation by CYP2D6. In humans there are a large number of different polymorphic sites in the CYP2D6gene, and the vast majority is present in a very low frequency. In the present study we screened for the most common inactivating polymorphism in CYP2D6, the CYP2D6\4allele, which is present at a frequency of approximately 21--29%. Other less common inactive alleles are CYP2D6\3and CYP2D6\5, representing around 1% and 4%, respectively, of all CYP2D6alleles \[[@B22]\]. Among alleles with decreased enzyme activity the CYP2D6\41allele identifies a large proportion of the intermediate metabolisers \[[@B23]\]. The restriction fragment length polymorphism technique that we used identifies a restriction site not found in the CYP2D6\4allele but that is present in other CYP2D6alleles. This results in misclassification of the carriers of CYP2D6\3and CYP2D6\5alleles, which could occur in a few cases but would have a minor influence on the results. The definition of CYP2D6\1used in the present study mainly constitutes the normal activity alleles CYP2D6\1and CYP2D6\2, which represent a rather high frequency in a Caucasian population \[[@B22]\]. In the regression analysis we combined CYP2D6\4heterozygotes and homozygotes in one group since the number of homozygous CYP2D6\4patients was low. Some studies have shown that the hydroxylation ratios are significantly different between the homozygous and heterozygous genotypes, demonstrating intermediate hydroxylation ratios in heterozygous genotypes. There is also support, however, for the hypothesis that only CYP2D6\4homozygotes will demonstrate altered pharmacokinetics for a given drug \[[@B24]\]. Conclusion ========== The variability in distant recurrence-free survival found in endocrine-treated patients may be a result of differences in drug metabolism. The genotype of metabolic enzymes might thus be useful as a guide for adjuvant endocrine treatment of postmenopausal breast cancer patients. However, our results contradict the main hypotheses and the present sample size is relatively small. Our findings therefore need confirmation in a larger cohort. Abbreviations ============= BSA = bovine serum albumin; CI = confidence interval; CYP2D6 = cytochrome P450 2D6; ER = oestrogen receptor; PCR = polymerase chain reaction; RR = relative risk; SULT1A1 = sulfotransferase 1A1. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contribution ====================== PW carried out part of the laboratory work and drafted the manuscript. LV carried out part of the laboratory work. OS contributed with the coordination of tumour material and performed the statistical analyses. BN initiated the randomised clinical trial. LS and L-ER provided tumour material and clinical data. SW conceived the study and participated in its design and coordination. All authors read and approved the final version of the manuscript. Acknowledgements ================ This project was supported by grants from The Swedish Cancer and Allergy Society and from The Foundation of the National Board of Health and Welfare. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Distant recurrence-free survival among postmenopausal women with oestrogen-receptor-positive breast cancers, in relation to the CYP2D6genotype and adjuvant tamoxifen treatment. Solid line, patients receiving tamoxifen (Tam+); dashed line, patients who did not receive tamoxifen (Tam-). (a)Patients homozygous for the CYP2D6\1allele. The number of events for Tam+ and Tam- were 25 and 27, respectively. (b). Patients homozygous or heterozygous for the CYP2D6\4allele (null allele). The number of events for Tam+ and Tam- were 6 and 15, respectively. ::: ![](bcr993-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Distant recurrence-free survival of postmenopausal, oestrogen-receptor-positive breast cancer patients in relation to the SULT1A1genotype and adjuvant tamoxifen therapy. Solid line, patients receiving tamoxifen (Tam+); dashed line, patients who did not receive tamoxifen (Tam-). (a)Patients homozygous for the SULT1A1\1allele. The number of events for Tam+ and Tam- were 9 and 16, respectively. (b)Patients homozygous or heterozygous for the SULT1A1\2allele (low-activity allele). The number of events for Tam+ and Tam- were 24 and 26, respectively. ::: ![](bcr993-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Distant recurrence-free survival of postmenopausal, oestrogen-receptor-positive breast cancer patients with genotypes linked to the benefit from adjuvant tamoxifen therapy. Solid line, patients treated with tamoxifen (Tam+); dashed line, patients not receiving adjuvant tamoxifen therapy (Tam-). (a)Patients homozygous for the SULT1A1\1allele and/or homozygous or heterozygous for the CYP2D6\4allele. The number of events for Tam+ and Tam- were 15 and 27, respectively. (b)Patients homozygous for the CYP2D6\1allele and homozygous or heterozygous for the SULT1A1\2allele. The number of events for Tam+ and Tam- were 15 and 18, respectively. The relative risk for distant recurrence was calculated for each genotype; when compared, a significant decrease in relative risk was found for the beneficial genotypes (P= 0.018). ::: ![](bcr993-3) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Genotype, tumour characteristics and endocrine therapy of the total study population (n= 226), including both oestrogen receptor (ER)-positive and ER-negative patients ::: Characteristic Genotype \[n(%)\] ----------------------------------- --------------------- ----------- --------- ----------- ----------- ----------- Nodal involvement^a^, tumour size Node-, \>30 mm 19 (76.0) 5 (20.0) 1 (4.0) 6 (24.0) 11 (44.0) 8 (32.0) Node+, ≤ 20 mm 66 (74.2) 18 (20.2) 5 (5.6) 30 (33.7) 49 (55.1) 10 (11.2) Node+, \>20 mm 77 (68.8) 32 (28.6) 3 (2.7) 43 (38.4) 59 (52.7) 10 (8.9) Receptor status^b^ ER-negative 50 (74.6) 12 (17.9) 5 (7.5) 23 (34.3) 35 (52.2) 9 (13.4) ER-positive 109 (69.9) 43 (27.6) 4 (2.6) 56 (35.9) 82 (52.6) 18 (11.5) Endocrine therapy Tamoxifen 77 (68.8) 28 (25.0) 7 (6.3) 37 (33.0) 60 (53.6) 15 (13.4) No tamoxifen 85 (74.6) 27 (23.7) 2 (1.8) 42 (36.8) 59 (51.8) 13 (11.4) ^a^Nodal involvement: node+, node-positive; node-, node-negative. ^b^ER data from three patients were missing. ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Statistics of oestrogen-receptor-positive patients: association between tamoxifen therapy/no tamoxifen therapy (Tam+/Tam-) and distant recurrence rate, stratified according to genotype ::: Genotype Tamoxifen therapy Number of patients^a^ Number of recurrences Recurrence rate ratio (95% confidence interval) Pvalue --------------------------------------- ------------------- ----------------------- ----------------------- ------------------------------------------------- ---------- SULT1A1\1/\*1^†^ Tam- 29 16 1.0 Tam+ 26 9 0.48 (0.21 -- 1.12) 0.074 SULT1A1\*2^†^ Tam- 49 26 1.0 Tam+ 50 24 0.82 (0.47 -- 1.43) 0.48 CYP2D6\1/\1^‡^ Tam- 55 27 1.0 Tam+ 52 25 0.91 (0.53-- 1.57) 0.75 CYP2D6\4^‡^ Tam- 23 15 1.0 Tam+ 24 6 0.28 (0.11 -- 0.74) 0.0089 SULT1A1\1/\1and/or CYP2D6\4^§^ Tam- 45 27 1.0 Tam+ 43 15 0.38 (0.19-- 0.74) 0.0041 SULT1A1\2and CYP2D6\1/\1^§^ Tam- 33 18 1.0 Tam+ 33 15 1.22 (0.61-- 2.40) 0.57 The relative risks of distant recurrence, calculated for each combined genotype are adjusted for age, tumour size and lymph node status. ^a^Follow-up data of two patients were missing. ^†,‡,§^The risk ratio was first calculated separately for each genotype and genotype combination. Second, the test for interaction between the risk ratios was performed by Cox regression: ^†^P= 0.27, ^‡^P= 0.064, and ^§^P= 0.018. The risk ratio for patients not receiving tamoxifen (Tam-) is calculated as 1.0. :::	PubMed Central	2024-06-05T03:55:57.454674	2005-1-28	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143572/", "journal": "Breast Cancer Res. 2005 Jan 28; 7(3):R284-R290", "authors": [ { "first": "Pia", "last": "Wegman" }, { "first": "Linda", "last": "Vainikka" }, { "first": "Olle", "last": "Stål" }, { "first": "Bo", "last": "Nordenskjöld" }, { "first": "Lambert", "last": "Skoog" }, { "first": "Lars-Erik", "last": "Rutqvist" }, { "first": "Sten", "last": "Wingren" } ] }
PMC1143573	Introduction ============ The cyclical transformations and cellular kinetics of the mammary gland have long been the subject of intense investigation. Interest in alterations caused by steroid actions, specifically on epithelial proliferation, has risen sharply in the past decade owing to increasing numbers of breast carcinomas in the female population \[[@B1]\]. There is also an immediate need for more accurate predictors of breast cancer risk, particularly in light of the various chemoprevention trials under way. Understanding the factors and mechanisms that regulate hormone-related changes in the normal human breast is crucial, because alterations in breast structure and function during the menstrual cycle could predispose this tissue to malignant changes and hence to the development of breast cancer. The present study focuses on epithelial cells, the primary target for carcinogenesis, as demonstrated by various histological studies indicating that most breast tumors arise in the epithelial cell population \[[@B2]\]. Although several groups have examined the proliferative activity of epithelial cells in the normal breast throughout the menstrual cycle \[[@B3]-[@B8]\], an underlying limitation in these studies is the concurrent use of oral contraceptives by many of the patients studied. In addition, studies with the thymidine labeling index demonstrated that proliferative activity declines with age \[[@B7],[@B8]\], thus requiring standardization of the data to reduce variability. Finally, these studies do not provide a clear explanation for the hormone-related variations in programmed cell death (apoptosis) and mitosis (proliferation) observed during the menstrual cycle. It is well known that the mammary gland undergoes morphological modifications during the menstrual cycle. Epithelial proliferation is greatest during the luteal phase, suggesting a synergistic influence of estrogen and progesterone \[[@B9],[@B10]\]. Work by Ferguson and Anderson \[[@B5]\] and by Anderson and colleagues \[[@B6]\], using a linear regression model, suggests that there are cyclical variations in epithelial apoptosis levels, with maximal apoptotic expression on the 28th day of the menstrual cycle, approximately 3 days after the mitotic peak. This coincides with decreasing levels of estrogen and progesterone at the end of the menstrual cycle. Although small temporal differences were found between the two phenomena, the authors stated that it was not possible to assume a link between mitosis and apoptosis in those studies. Because the number of cells undergoing cell death was very low, apoptotic frequency was evaluated per mammary lobule (approximately 50 lobules per case). The few apoptotic cells observed seemed to be randomly distributed in the lobule, within central and peripheral ducts. The authors also included patients in whom neither apoptosis nor mitosis could be observed, requiring mathematical manipulation of the data obtained in all cases examined in order to derive mitotic and apoptotic frequencies. Indeed, in the study by Anderson and colleagues \[[@B6]\], it is still unclear whether the apoptotic activity described was more intense at the end of the luteal phase because it followed increased mitotic activity in the days before or because it was independently greater during that period of the menstrual cycle. To clarify the cyclical effects of gonadal steroids on breast homeostasis, we evaluated the apoptotic index (AI) in human mammary epithelial cells (by terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling (TUNEL) staining \[[@B11]\]) during the spontaneous (natural) menstrual cycle and correlated it with cellular proliferation as determined by Ki-67 expression \[[@B12]\] during the same period. We also correlated the progesterone levels with apoptosis and proliferation irrespective of the stage of the cycle. Materials and methods ===================== Patients -------- The series consisted of 42 patients (age 21.8 ± 6.1 years (mean ± standard deviation), range 14 to 38 years) undergoing excision of fibroadenomas, a condition thought not be associated with an increased risk of cancer, at the Department of Gyneology, Mastology Division, Federal University of São Paulo, Paulista Medical School. Institutional review board approval was obtained and patients signed an informed consent before enrollment in the study. Inclusion criteria for entry into the study were as follows: female sex, regular menstrual cycles (intervals of 28 ± 2 days) in the previous 6 months, and with a known date of last menstrual period. Exclusion criteria were: use of hormone therapy in the previous 12 months, pregnancy, females nursing in the previous 12 months, patients with any type of endocrine pathology, or patients taking any type of medication at the time that tissue for the study was obtained. Patients underwent a complete medical exam, ultrasonography, and fine needle aspiration with subsequent microscopic analysis to confirm that the mammary lesion was benign. All patients were randomized into two groups according to the phase of their menstrual cycle, which was determined by the date of the patient\'s previous as well as subsequent menstrual period: group A, follicular phase (age 23.5 ± 6.6 years (mean ± SD)); group B, luteal phase (age 20.0 ± 5.2 years). Progesterone levels were examined by radioimmunoassays, with values of 3.0 ng/ml or more considered compatible with adequate luteal activity \[[@B13]\]. The mammary tissue was obtained with a local anesthetic without vasoconstrictors (2% lidocaine was used), at about 1:00 pm to minimize the effects of circadian rhythms on hormone release \[[@B4],[@B14]\] and on cellular kinetics \[[@B15],[@B16]\]. The parenchymal mammary tissue removed (at least 1 cm from the lesion) was without any gross microscopic alterations \[[@B17]\]. Histopathology, TUNEL and immunohistochemistry ---------------------------------------------- After fixation in 10% formaldehyde and embedding in paraffin, tissues samples were cut into 4 μm sections, deparaffinized in xylene and hydrated in a series of ethanol washes, stained with hematoxylin and eosin, and examined for any pathological abnormalities. TUNEL staining was performed with the ApopTag Plus in situApoptosis Detection Kit (Oncor, Gaithersburg, MD, USA). Tissue sections were deparaffinized and hydrated as above, and stripped of proteins by incubation with 20 μg/ml proteinase K (42°C) for 15 min. The slices were then washed in distilled water and PBS and incubated in 3% H~2~O~2~to remove endogenous peroxidases. After equilibration, the sections were incubated at 37°C in terminal deoxynucleotidyl transferase (TdT) enzyme and digoxigenin-labeled substrate for 1 hour. Antidigoxigenin was then applied, and detection was accomplished with diaminobenzidine substrate solution. The sections were then counterstained with methyl green, cleared, and mounted. Sections treated with DNAse I enzyme were used as positive controls (Amersham, Cleveland, OH, USA), and sections from which TdT enzyme was omitted were used as negative controls. Adjacent sections were used for Ki-67 immunohistochemical staining. Sections were deparaffinized, hydrated, incubated with proteinase K, washed in distilled water and PBS, and incubated with 3% H~2~O~2~as above. Blocking serum was applied for 20 min and sections were then sequentially incubated in primary mouse anti-human Ki-67 antibody (1:100, clone Ki-S5; Dako, Carpinteria, CA, USA) for 1 hour, biotinylated rabbit anti-mouse IgG (Dako LSAB kit) for 12 min, and ABC reagent (avidin and biotinylated horseradish peroxidase complex) for 30 min (ABC kit; Vector, Burlingame, CA, USA). Detection was accomplished with diaminobenzidine substrate solution until the desired staining intensity was obtained (3 to 5 min). The sections were then counterstained with hematoxylin and eosin, cleared, and mounted. Quantitative evaluation of Ki-67 expression and TUNEL positivity was performed in epithelial cells from the anatomically normal mammary gland (without any fibrocystic changes), including central and peripheral ducts within the lobules; myoepithelial and lymphoid cells were excluded. Cells were counted as apoptotic only if they were TUNEL-positive and showed characteristic nuclear morphology typical of apoptosis (that is, cells containing pyknotic nuclei plus apoptotic bodies). Successive counts, performed by individuals blinded to the groups, were made until 1,000 cells per tissue sample had been examined. Two indices were thus obtained: the proliferation index (PI), defined as the number of Ki-67-positive nuclei per 1,000 epithelial cells, and the AI, defined as the number of TUNEL-positive cells per 1,000 epithelial cells counted. From these values the cell renewal index (CRI = PI/AI) was obtained. Statistical analysis -------------------- Statistical analyses were performed with analysis of variance for continuous data followed by Student\'s t-test, and with non-parametric tests (Mann--Whitney Utest) for non-continuous data (gestation and parity). Fisher\'s exact test was used to verify the homogeneity of the samples relative to nursing history. The variation in frequency of proliferation and apoptosis relative to variations in the menstrual cycle as well as the variation in frequency of proliferation and apoptosis relative to the progesterone levels were measured with linear regression models and fourth-degree polynomial curves (Polnom Program; University of Manchester, Manchester, UK) as indicated. All data are expressed as means ± SD; P\< 0.05 was considered significant. Results ======= There were no differences between patients in group A (follicular phase) and group B (luteal phase) in terms of age of menstruation onset (12.8 ± 1.3 and 12.7 ± 1.4, respectively), number of pregnancies, parity, and lactation history. The mean PI for group A was 13.5 ± 9.8, significantly smaller than in group B, which had a PI of 30.5 ± 22 (P= 0.003, by Student\'s t-test). Photomicrographs from representative patients showing Ki-67 expression for both groups are shown in Fig. [1a,b](#F1){ref-type="fig"}. The AI for both groups was very similar: group A (4.4 ± 1.8) and group B (5.2 ± 2.4; P= 0.21). TUNEL-positive cells from representative patients for each group are given in Fig. [1C--E](#F1){ref-type="fig"}; Fig. [1f](#F1){ref-type="fig"} shows a highly magnified TUNEL-positive cell displaying the characteristic features of chromatin condensation and apoptotic bodies. The CRI for group A (3.8 ± 3.4) was also significantly smaller than that for group B (7.4 ± 6.6; P= 0.03). The data show that, in a 28-day interval, the number of proliferative, apoptotic, and cell renewal events vary as a function of time. The linear regressions for each one of the indices are shown in Fig. [2a,c,e](#F2){ref-type="fig"}. The cyclical variability shows that the highest proliferation values occur near the end of the menstrual cycle, whereas the AI is greatest at the beginning and the end of the menstrual cycle. The CRI shows cyclical variations, with significantly greater values near the 28th day of the menstrual cycle (P= 0.033). The fourth-degree polynomial curves show that neither the PI (P= 0.6; Fig. [2b](#F2){ref-type="fig"}) nor the CRI (P= 0.25; Fig. [2f](#F2){ref-type="fig"}) is statistically different throughout the course of the menstrual cycle. In contrast, there is a statistically significant cyclical variability in the AI (P= 0.038; Fig. [2d](#F2){ref-type="fig"}). The mathematical equations that best represent the PI and AI are the linear regression and the fourth-degree polynomial curve, respectively. When these are superimposed, after mathematical adjustment of the median visible time of Ki-67 expression with that of apoptosis, it is clear that the maximum values for both indices (PI and AI) coincide at about the 24th day of the menstrual cycle. At this point, the linear regression for the PI is tangential to the fourth-degree polynomial (Fig. [2g](#F2){ref-type="fig"}). Furthermore, progesterone levels, irrespective of the stage of the cycle, correlate with proliferation. In apoptosis this is not so, because we found a decrease in apoptosis at progesterone levels higher than 15 ng/ml (Fig. [2h](#F2){ref-type="fig"}). It is important to note that, relative to the total number of cells present, the number of epithelial cells undergoing proliferation or apoptosis at any given time during the menstrual cycle was quite small, and the cells were distributed throughout the mammary lobules. Discussion ========== The mammary gland undergoes cyclical changes in response to the hormonal fluctuations of the menstrual cycle. The epithelium responds to these systemic hormonal changes by regional proliferation, differentiation, and programmed cell death, also known as apoptosis \[[@B18]\]. In the present study of normal mammary tissue we observed that this epithelial response is limited to a small fraction of the cells, suggesting that there is probably local regulation of cell survival and death in this tissue. Similar results were obtained by Andres and Strange \[[@B18]\], who also noted that cells undergoing proliferation or apoptosis were isolated, distributed in various mammary lobules, and positioned close to the lumen. Proliferation and apoptosis in normal breast tissue are influenced by several factors, including phase of the menstrual cycle, chronological age, breast age, use of oral contraceptives (especially if nulliparous) and recent parity \[[@B19]\]. It is therefore imperative to use strict criteria when selecting patients for studies of normal mammary tissue. Several previous studies have examined cell turnover in morphologically normal human mammary gland epithelium \[[@B4]-[@B8],[@B16],[@B20]-[@B22]\]. However, interpretation of the results of these studies is difficult for four reasons: first, the menstrual and parity status of the groups were not reported \[[@B23]\]; second, the studies included patients of perimenopausal and/or menopausal age \[[@B5],[@B8],[@B23]-[@B25]\]; third, several of the patients studied were using oral contraceptives \[[@B5],[@B8],[@B24],[@B26]\]; and fourth, there are no available data about, or there was no assessment of, the phase of the menstrual cycle with progesterone levels \[[@B5],[@B8],[@B23],[@B24],[@B26]\]. In the present study, all of the above parameters were evaluated and/or controlled for. As reviewed by Brown and Gatter \[[@B27]\], the PI of any tissue is determined by the growth fraction, which can be assessed by Ki-67 expression, and the time it takes the cell to complete the cell cycle. Thus, tissues in which many cells are in a very long cycle show extensive Ki-67 expression but not a very large PI. In contrast, tissues in which a few cells are in a very short cycle have a higher PI but few Ki-67-positive cells. Because Ki-67 is expressed throughout the cell cycle (G1, S, G2 and M phase), is quickly degraded and is not found in cells undergoing DNA repair \[[@B12]\], its expression provides information only about whether a cell is in the cycle, but not about the cycle length. In accordance with previous studies \[[@B5],[@B20]\], we find that the PI (the number of Ki-67-positive nuclei per 1,000 epithelial cells) of the human mammary epithelium is significantly greater in the luteal phase of the menstrual cycle, both when evaluated by the average of the means and when measured by linear regression. The mechanisms by which steroid hormones stimulate mammary epithelium growth are controversial and continue to be the source of intense investigation. Various studies suggest that the regulation of cell proliferation occurs via three main mechanisms, namely receptor-mediated \[[@B28]-[@B36]\], an autocrine/paracrine loop \[[@B37]\], and negative feedback \[[@B38],[@B39]\]. Among the specific hormone receptors involved, two estrogen receptors (ERs) are of particular interest: ER-α, the predominant subtype in the mammary gland, and ER-β \[[@B40]\]. In normal mammary tissue, the fraction of epithelial cells that express ER is small \[[@B41]\]. Recent studies have shown that the ER is a ligand-dependent transcription factor, which accounts for the latency of most estrogenic responses in target tissues \[[@B42]\]. The ER can regulate gene transcription either directly or indirectly and, depending on the patterns of co-regulator recruitment to the ligand--receptor--gene assembly, can elicit either the stimulation or inhibition of specific biological effects \[[@B43]\]. Additionally, binding of estrogen to the ER can be modulated by progesterone \[[@B44]\]. Acting through the progesterone receptor, progesterone is the physiological negative regulator of estrogen activation \[[@B45]\]. It is possible that steroid hormone fluctuations influence the activation of gene transcription in target mammary cells. These target cells express different types and levels of regulatory proteins in the first and second phases of the menstrual cycle. They require a latency period to initiate the transcription of regulatory factors that activate the cell cycle, culminating with greater proliferation in the second phase of the menstrual cycle. In the breast, proliferation and apoptosis of the mammary epithelium occur in response to estrogen and other hormones. One of the classical examples of apoptosis is the involution of the post-lactating mammary gland, in which the loss of lactogenic hormones results in collapse of the gland due, in large part, to a programmed elimination of differentiated mammary epithelial cells \[[@B46]\]. In the present study we observe that the apoptotic activity does not vary significantly between the luteal and follicular phases. Moreover, our results show that apoptosis reaches maximum values in the middle of the luteal phase (at about day 24 of the menstrual cycle), but there is also a small elevation in TUNEL-positive nuclei at about the third day of the cycle. This is in contrast to the study by Ferguson and Anderson \[[@B5]\], who found significant cyclical variations in the apoptotic frequency. The reason for the discrepancy in results is unclear but could be related to differences in methodology and patient selection criteria. It is also important to note that in the study by Ferguson and Anderson \[[@B5]\] the significance in cyclical variations is only evident in the logarithm of the transformed, or mathematically adjusted, values for the mitotic and apoptotic frequencies against day of the menstrual cycle. In the present study there is no need to transform any of the values obtained, and we find that the curve that best fits the results is the fourth-degree polynomial. It is also likely that cyclical variations take place in the rate with which mitotic and apoptotic events occur, thus affecting the time periods at which these phenomena are visible \[[@B47]\]. The outliers in the data may be related to this phenomenon. The number of events and the rate at which they occur have not been concomitantly evaluated. It is therefore possible that the increase in events during the luteal phase of the menstrual cycle is simply due to the fact that these events are visible for longer during that phase, and not because the absolute number of events is greater at that time relative to the follicular phase. Finally, the increase in events in the luteal phase may be due to a combination of an increased visible time and an increased number of events. Further studies with multi-parameter analysis, allowing us to differentiate between event numbers, duration, and rate of appearance, are needed to understand the cyclical effects of gonadal steroids on breast homeostasis. Mammary tissue is one of the best examples of the complexity of the biological relationship between cellular proliferation and cell death. The proliferation and apoptotic rates in the mammary gland vary numerous times throughout life depending on age, phase of the menstrual cycle, parity, puerperal involution, and menopause. To study the dual effect of proliferation and apoptosis in mammary epithelial cells, we use the CRI (calculated as PI/AI). We find that, although both proliferation and apoptosis occur throughout the menstrual cycle, the maximal CRI values occur at about the 28th day of the cycle. It is important to note that the isolated comparison of the apoptotic and proliferation indices has some limitations, because each index reflects both the number of times that each event takes place in unit time and the proportion of cells in the population capable of undergoing each event \[[@B48]\]. The CRI allows us to measure the frequency of occurrence of proliferation relative to apoptosis. This serves to reduce or eliminate the effects of differences between samples; cells that do not undergo cellular proliferation or apoptosis do not contribute to the calculation of the CRI \[[@B48]\]. In the present study we find that progesterone levels, irrespective of the stage of the menstrual cycle, are positively correlated with the PI, as seen in animal models \[[@B49]\]. However, the AI decreases with progesterone levels of more than 15 ng/ml, suggesting that mammary epithelial apoptosis is probably dependent on more than just alterations in progesterone levels. Aside from hormonal fluctuations, there are several factors that change during the menstrual cycle. For example, several Bcl-2 family (apoptosis-regulating) proteins are expressed in the normal mammary epithelium, including the anti-apoptotic proteins Bcl-2, Bcl-X, and MIC-1, and the pro-apoptotic Bax. It has been shown that the relative ratios of these various pro-apoptotic and anti-apoptotic members of the Bcl-2 family determine the ultimate sensitivity or resistance of cells to diverse apoptotic stimuli \[[@B46]\]. Ferrieres and colleagues \[[@B50]\] found a correlation between Bcl-2 expression and progesterone levels, with greater Bcl-2 expression in the follicular phase, diminishing as the menstrual cycle progresses. In contrast, Sabourin and colleagues \[[@B51]\] found that Bcl-2 was expressed preferentially in lobular epithelial cells, with maximal expression at the mid-cycle period and a sharp decrease at the end of the menstrual cycle. Despite the different patterns of Bcl-2 expression in the first part of the menstrual cycle, the results from those two studies suggest that the regulation of Bcl-2 expression in breast tissue is hormone-dependent. Furthermore, as discussed by Ferrieres and colleagues \[[@B50]\], these results suggest a loss in the control of Bcl-2 by progesterone in diseases originating from epithelial lobular components. Another protein that is also under hormonal control and undergoes cyclic variations throughout the menstrual cycle is epidermal growth factor receptor (EGFR). EGFR is involved in controlling proliferation and, probably, the differentiation of normal breast epithelial cells. Gompel and colleagues \[[@B52]\] have shown that EGFR expression in mammary epithelial cells is stronger in the luteal phase than in the follicular phase. This suggests an effect of progestins similar to that observed in breast cancer cell lines, although it is unclear whether high EGFR levels correlate with higher proliferation or with tissue differentiation. Conclusion ========== Our data indicate that the numbers of proliferative and apoptotic events vary during the menstrual cycle. The frequency with which these events occur is also subject to cyclical variations. The PI and the CRI are significantly higher in the luteal phase. The PI and the AI culminate on about the 24th day of the menstrual cycle, whereas the CRI peaks on the 28th day. Elucidating the cyclical effects of gonadal steroids on breast homeostasis is crucial if we are to understand how alterations in breast structure and function during the menstrual cycle can lead to breast cancer development. Abbreviations ============= AI = apoptotic index; CRI = cell renewal index; EGFR = epidermal growth factor receptor; ER = estrogen receptor; PBS = phosphate-buffered saline; PI = proliferation index; TdT = terminal deoxynucleotidyl transferase; TUNEL = TdT-mediated dUTP nick end labeling. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= MAHN, the principal investigator, conceived the study, was responsible for all the patient care and surgical procedures, and wrote the manuscript. CMM performed the TUNEL staining and immunohistochemistry, and aided in drafting of the manuscript. RF performed the pathological study. LGAQ helped with the statistical analyses. GRL and ECB provided valuable input into the manuscript. ACPN supervised the project. All authors read and approved the final manuscript. Acknowledgements ================ The authors thank Albert Maier for his expert advice and statistical analysis, Dr Pak H Chan for supporting the research efforts, and Elizabeth Hoyte for figure preparation. This work was supported by the Federal University of São Paulo (MAHN) and by the American Heart Association Postdoctoral Fellowship no. 0120142Y (CMM). Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Photomicrographs of human mammary tissue during the follicular and luteal phases of the natural menstrual cycle. (a,b)Ki-67 expression (arrows and/or deep brown color) in epithelial cells from normal mammary lobule of a representative patient in the follicular phase (a) and in the luteal phase (b). (c--f)TdT-mediated dUTP nick end labelling (TUNEL)-positive staining (arrows, deep brown color) indicating apoptotic epithelial cells from representative patients in the follicular phase (c,d) and in the luteal phase (e,f); (f) high magnification of a labeled cell displaying the characteristic features of chromatin condensation and apoptotic bodies. ::: ![](bcr994-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Linear regressions and polynomial curves for the proliferation index (a,b), apoptotic index (c,d)and cell renewal index (e,f). The graphs that best represent the proliferation and the apoptotic indexes were the linear regression (a,c,e) and the polynomial curve (b,d,f), respectively. (g)By superimposing the data after mathematically adjusting the mean visible time of Ki-67 expression \[53\] with that of the apoptotic index, one confirms that the maximum values for both indices coincide at about day 24 of the menstrual cycle (the point at which the linear regression and polynomial curves are tangential). (h)Correlation of progesterone levels with apoptosis and proliferation irrespective of the stage of the menstrual cycle. ::: ![](bcr994-2) :::	PubMed Central	2024-06-05T03:55:57.457210	2005-2-16	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143573/", "journal": "Breast Cancer Res. 2005 Feb 16; 7(3):R306-R313", "authors": [ { "first": "Maria Alicia H", "last": "Navarrete" }, { "first": "Carolina M", "last": "Maier" }, { "first": "Roberto", "last": "Falzoni" }, { "first": "Luiz Gerk de Azevedo", "last": "Quadros" }, { "first": "Geraldo R", "last": "Lima" }, { "first": "Edmund C", "last": "Baracat" }, { "first": "Afonso CP", "last": "Nazário" } ] }
PMC1143574	Introduction ============ There is now increasing evidence that stromal cells play a pivotal role in promoting the growth of most carcinomas, including breast cancer \[[@B1]-[@B4]\]. The myofibroblast is the predominant type of stromal cell found in most carcinomas \[[@B1],[@B5],[@B6]\]. Myofibroblasts are defined by their characteristic expression of α-smooth muscle actin as well as certain other markers such as vimentin and desmin \[[@B1]\]. Tumor-associated myofibroblasts are believed to originate from normal fibroblasts and are similar or identical to the myofibroblasts found in healing wounds \[[@B1]\]. They are largely responsible for the desmoplasia that is characteristically present in carcinomas because they secrete large amounts of collagen and other extracellular matrix proteins. A large body of work has been done to investigate the interactions between fibroblasts and carcinoma cells \[[@B1]-[@B4],[@B7]-[@B9]\]. Recently, an orthotopic xenograft model was created in mice in which both the stromal and epithelial components of the reconstructed mammary gland are of human origin \[[@B10]\]. This complex model again underscored the critical role played by heterotypic interactions in human breast carcinogenesis. Many other in vitroand in vivostudies have demonstrated that fibroblasts can promote the growth of cancer cells \[[@B11]-[@B16]\]. We are not aware, however, of any previous reports of studies investigating the ability of fibroblasts to promote the clonogenic growth of small numbers of primary breast cancer cells in vitro. To address this issue, we first co-cultured low numbers of a slow growing human breast cancer cell line (UACC-812) with a monolayer of normal human fibroblasts derived from human breast tissue. We selected this tumor cell line for our initial experiments because it was derived in 1991 from a ductal carcinoma (stage II, grade IV) and closely mimics the glandular morphology (Fig. [1a](#F1){ref-type="fig"}) and slow growth rate (doubling time \> 100 hours) of primary breast cancer \[[@B17]\]. In order to validate the potential clinical relevance of our experiments, we then co-cultured a monolayer of serum-activated fibroblasts with tumor cells obtained directly from human breast cancer samples. Gene expression profiling by microarrays and immunocytochemistry were used to characterize the serum-activated fibroblasts. Finally, we examined 25 randomly selected surgical specimens containing metastatic breast cancer of varying grades for the presence of fibroblasts expressing α-smooth muscle actin in close proximity to tumor cells. Here, we report the novel observation that serum-activated fibroblasts promote the clonogenic growth of small numbers of breast cancer cells obtained directly from human tumor samples. Moreover, we show that fibroblasts activated by serum in vitroshare many molecular features with the fibroblasts that are naturally and abundantly present within primary human breast cancers and metastases. Methods ======= Breast cancer cell line and fibroblasts --------------------------------------- The UACC-812 human breast cancer cell line (American Type Culture Collection \[ATCC\], Manassas, VA, USA) was passaged in Leibovitz\'s medium supplemented with 15% fetal calf serum prior to use. Normal fibroblasts (CCD-1068SK; ATCC) obtained from the breast of a 65-year-old female were passaged at 37°C in minimal essential medium (Eagle\'s) supplemented with 2 mmol/l L-glutamine, Earle\'s balanced salt solution (1.5 g/l), sodium bicarbonate, 0.1 mmol/l nonessential amino acids, 1 ml sodium pyruvate, and 10% fetal calf serum in a 5% carbon dioxide atmosphere. All cell culture reagents were obtained from ATCC. Our co-culture experiments utilized confluent monolayers of fibroblasts that had been passaged no more than 14 days. This precaution ensured that the fibroblasts were not senescent or transformed. Immunocytochemical assays of cultured cells ------------------------------------------- The presence of α-smooth muscle actin, epithelial membrane antigen (a marker of breast cancer cells), and syndecan-1 was detected by in situimmunostaining of methanol-fixed cells using prediluted murine monoclonal antibodies directed against α-smooth muscle actin, or syndecan-1, or epithelial membrane antigen (Chemicon International, Temecula, CA, USA). Bound antibody was then detected with an alkaline phosphatase procedure (IHC Select^®^AP/Fast Red; Chemicon International) in accordance with the manufacturer\'s directions. Clonogenic co-culture assay --------------------------- We seeded 100 UACC-812 breast cancer cells into individual wells of a 96-well cell culture plate containing a confluent monolayer of fibroblasts growing in fibroblast growth medium supplemented with 10% fetal calf serum. Controls included co-culture with fibroblasts in serum-free medium, co-culture with a monolayer of a nonfibroblast cell line (murine B16 melanoma), and culture in medium with fetal calf serum but without fibroblasts. In order to determine whether a soluble factor was responsible for any effect that we observed, we also co-cultured the fibroblasts and breast cancer cells separately on opposite sides of permeable transwell inserts with a 0.2 μm pore size (Nalge-Nunc International, Rochester, NY, USA). The inserts permitted diffusion of any soluble growth factors produced by the fibroblasts but prevented direct contact between the cells. All of the experimental and control combinations were performed in triplicate wells of 96-well culture plates. At intervals of 3--4 days, fresh medium was added. After 14 days the cells were fixed with 70% ethanol for 10 min prior to staining for 3 min with 0.1% toluidine blue. Using inverted microscopy, we counted the number of colonies containing 12 or more tumor cells in each well. A colony was defined as a cluster of epithelial cells in direct contact with one another. The means and standard deviations for each condition were compared statistically using pair-wise comparisons with the appropriate control in an unpaired, two-tailed t-test. Proliferation and apoptosis assays ---------------------------------- In order to determine whether serum-activated fibroblasts promote the growth of breast cancer cells in vitroeven at higher densities of tumor cells, we used flow cytometry to characterize the apoptotic and proliferative rates of breast cancer cells co-cultured with fibroblasts. Confluent monolayers of fibroblasts in T-75 culture flasks (Corning Inc., Corning, NY, USA) were seeded with 500,000 breast cancer cells and co-cultured for 1 week with a medium change at 4 days. On day 7 we separated the breast cancer cells from the fibroblasts using magnetic beads conjugated to epithelial membrane antigen (DYNAL MPC-S Kit; Dynal AS, Oslo, Norway) in accordance with the manufacturer\'s directions. The negative control consisted of 500,000 breast cancer cells cultured in the same medium but without fibroblasts. The purified breast cancer cells were then analyzed for their proportions of proliferating and apoptotic cells using the Absolute-S™ Kit for Measuring Cell Proliferation by Dual Color Flow Cytometry (Chemicon International) in accordance with the manufacturer\'s directions. The stained cells were analyzed on a Coulter^®^Epics^®^-MCL (Coulter, Hialeah, FL, USA) flow cytometer using appropriate positive and negative controls. Co-culture of primary breast cancer cells and fibroblasts --------------------------------------------------------- We aseptically dissected small pieces of tissue (2--3 mm in diameter) from three cases of invasive ductal carcinoma of the breast, three cases of ductal carcinoma in situ, and seven specimens of normal breast tissue. The tissue fragments were then mechanically disaggregated using Medicons™ (BD Biosciences, San Jose, CA, USA) with a 50 μm mesh size. This method routinely yielded between 1000 and 10,000 viable epithelial cells, as determined by trypan blue staining and expression of epithelial membrane antigen. We seeded about 100 epithelial cells into individual wells of 96-well cell culture plates containing a monolayer of fibroblasts. The control consisted of epithelial cells seeded into wells without fibroblasts. At the end of 14 days of co-culture, the cells were stained in situwith toluidine blue as described above, and the colonies of tumor cells (defined as 12 or more adjacent epithelial cells) in each well were counted using inverted microscopy. Target preparation/processing for GeneChip analysis --------------------------------------------------- We used gene expression analysis on Affymetrix U133A cDNA microarrays (Affymetrix Inc., Santa Clara, CA, USA) to compare the molecular signatures of serum-activated fibroblasts with those of serum-starved fibroblasts (experiment 1). We also compared the gene expression patterns of serum-activated fibroblasts cultured alone with those of serum-activated fibroblasts co-cultured with breast cancer cells for 3 days before being purified with magnetic beads (experiment 2). All assays were performed in duplicate at the UCI DNA Microarray Core Facility. Total RNA isolated from separate cultures of serum-activated, serum-starved, and purified, co-cultured fibroblasts was processed as recommended by Affymetrix Inc. \[[@B18]\]. The gene expression results were quantified and analyzed using GCOS 1.1.1 software (Affymetrix Inc.) and/or ArrayAssist\'s gcRMA (Iobion Informatics LLC, LaJolla, CA, USA) using default values. Gene expression in a sample was considered to be increased if the signal intensity was interpreted to be positive and at least fourfold higher (with P\< 0.001) than the baseline (defined as serum-starved fibroblasts in experiment 1 or serum-activated fibroblasts cultured alone in experiment 2). In vivometastasis study ------------------------- In order to determine whether fibroblasts expressing α-smooth muscle actin are in direct physical contact with metastases of breast cancer cells, we immunostained 25 randomly selected specimens of metastatic human breast cancer (ductal adenocarcinoma type) from the surgical pathology and autopsy archives at UCI Medical Center (Orange, CA, USA). These included four samples of well differentiated ductal adenocarcinoma, nine samples of moderately differentiated adenocarcinoma, and 12 samples of poorly differentiated carcinoma. Ten of the metastases were to lymph nodes, six were to bone marrow, five were to liver, two were to lung, one was to brain, and one was to pleura. The age range of the patients was 42--69 years. The study set included eight samples with micrometastases, which we defined as fewer than 100 visible tumor cells. The slides were stained with prediluted murine monoclonal antibody directed against α-smooth muscle actin (Ventana Medical Systems, Phoenix, AZ, USA). Bound antibody was detected using a biotin-streptavidin procedure employing horseradish peroxidase as the detecting enzyme. Appropriate positive and negative controls were included in each staining run. In order to confirm that cells expressing smooth muscle actin were myofibroblasts and not myoepithelial cells, we also stained serial sections of the same tissues with antibodies directed against CD10 and smooth muscle myosin heavy chain. These markers are preferentially expressed on myoepithelial cells \[[@B19]\] rather than myofibroblasts. Results ======= Microscopic appearance of co-cultured cells ------------------------------------------- When co-cultured with a monolayer of fibroblasts, the UACC-812 cell line consistently produced colonies of tumor cells that irregularly infiltrated the fibroblasts (Fig. [1b](#F1){ref-type="fig"}). Significantly, after just 14 days in co-culture the primary tumor cells derived directly from clinical specimens of ductal carcinoma in situ(Fig. [1c](#F1){ref-type="fig"}) or invasive ductal carcinoma (Fig. [1d](#F1){ref-type="fig"}) also produced multiple colonies in close proximity to fibroblasts. The epithelial cells in these colonies exhibited the usual cytologic features of malignancy, including prominent nucleoli, a high nuclear:cyoplasmic ratio, and marked cytologic atypia. Some of these colonies consisted of nearly 100 tumor cells (Fig. [1c](#F1){ref-type="fig"}), suggesting a doubling time of only 12 hours. It was also noteworthy that the fibroblasts adjacent to the colonies of tumor cells invariably expressed abundant α-smooth muscle actin (pink-staining fibroblasts in Fig. [1c,d](#F1){ref-type="fig"}). When the UACC-812 cell line was co-cultured on a monolayer of B16 melanoma cells, clonogenic growth was nearly absent (Fig. [1e](#F1){ref-type="fig"}). Moreover, UACC-812 cells cultured in the absence of fibroblasts survived only as individual cells or in very small colonies, generally of fewer than eight cells (Fig. [1f](#F1){ref-type="fig"}). The serum-activated fibroblasts in this model system uniformly expressed syndecan-1 (Fig. [1g](#F1){ref-type="fig"}). There was no evidence of overgrowth by fibroblasts during the course of the experiment. Clonogenic growth of breast cancer cells is promoted by direct contact with serum-activated fibroblasts ------------------------------------------------------------------------------------------------------- We performed co-culture experiments under a variety of conditions that are summarized in Fig. [2](#F2){ref-type="fig"}. Tumor cells co-cultured with serum-activated fibroblasts produced significantly (P\< 0.01) more colonies than did tumor cells cultured in serum-containing medium without fibroblasts or with fibroblasts in serum-free medium. Co-culture of the tumor cells and fibroblasts in separate chambers of transwells yielded low numbers of colonies, statistically identical to culture of tumor cells alone, as did co-culture on a monolayer of B16 cells. Fibroblast conditioned medium had no detectable effect on clonogenic growth of tumor cells grown in the absence of fibroblasts (data not shown). Serum-activated fibroblasts promote proliferation of breast cancer cells even at high densities of tumor cells -------------------------------------------------------------------------------------------------------------- A graphical summary of the proliferation and apotosis data from triplicate flow cytometry assays is presented in Fig. [3](#F3){ref-type="fig"}. These data confirmed that high densities of breast cancer cells co-cultured with fibroblasts had significantly higher proliferative rates and lower apoptotic rates than did breast cancer cells cultured in serum-supplemented medium without fibroblasts. Clonogenic growth of primary tumor cells derived from clinical samples ---------------------------------------------------------------------- The results of these experiments are shown in Table [1](#T1){ref-type="table"}. Each number in the table represents the total number of colonies of tumor cells obtained when the epithelial cells from one sample were cultured in the presence or absence of serum-activated fibroblasts. Notably, all of the clinical samples of ductal carcinoma in situand invasive carcinoma yielded substantially greater numbers of colonies when the epithelial cells were co-cultured with fibroblasts. In addition, the average size of the colonies produced by carcinoma cells co-cultured with fibroblasts (average 18 cells/colony) appeared to be substantially greater than the colonies produced by tumor cells cultured in the absence of fibroblasts (average eight cells/colony). The sample of invasive carcinoma that produced 33 colonies in the absence of exogenous, serum-activated fibroblasts was notable for the presence of numerous fibroblasts in the wells of the negative control, thereby suggesting contamination of the primary cancer cells by tumor-associated fibroblasts in that sample. The other clinical samples had no visible contaminating fibroblasts in the wells that were seeded with tumor cells alone. There was only minimal clonogenic growth of cells obtained from benign breast tissues, and these cells appeared cytologically benign. Gene expression profile of serum-activated fibroblasts ------------------------------------------------------ Because the gene expression profile of tumor-associated fibroblasts in situhas previously been described \[[@B20]-[@B22]\], we used cDNA microarrays to compare the gene expression patterns of the serum-activated fibroblasts used in our experiments with serum-starved fibroblasts, and to compare serum-activated fibroblasts with serum-activated fibroblasts that were co-cultured with tumor cells. The results, shown in Table [2](#T2){ref-type="table"}, confirm that serum-activated fibroblasts differentially upregulated genes related to myofibroblast differentiation (LOXL2, PLOD2, PLAUR, TAGLN, TPM2, MYL6, TFPI2) relative to serum-starved fibroblasts. In addition, there was upregulated expression of two genes that have previously been shown \[[@B20],[@B22]\] to be highly associated with tumor-associated fibroblasts in breast cancer (CXCL12and CLIC4) and upregulation of the COX2gene. Finally, we noted that serum-activated fibroblasts also upregulated genes related to activation and function of transforming growth factor (TGF)-β (TGFB1, LTBP2, SMAD3). The gene expression profile of serum-activated fibroblasts co-cultured with breast cancer cells was essentially the same as the profile of serum-activated fibroblasts cultured without tumor cells (data not shown). α-Smooth muscle actin-positive fibroblasts are in direct contact with tumor cells in metastases of breast cancer ---------------------------------------------------------------------------------------------------------------- There was intense staining for α-smooth muscle actin in all of the breast cancer metastases that we examined (Fig. [4](#F4){ref-type="fig"}). Significantly, the fibroblasts in these tissues closely enveloped the nests of tumor cells, even in the earliest micrometastases to lymph nodes (Fig. [4c](#F4){ref-type="fig"}). Our immunostaining results demonstrated that virtually all of the tumor cells were in direct contact or very close proximity to myofibroblasts, regardless of tumor grade or metastatic site. There was little or no staining of these same tissues with antibodies directed against CD10 or smooth muscle myosin heavy chain, thereby indicating that the cells expressing smooth muscle actin were myofibroblasts rather than myoepithelial cells. Discussion ========== We demonstrated that serum-activated human fibroblasts significantly enhance the clonogenic growth of small numbers of primary breast cancer cells obtained directly from clinical specimens. This effect requires direct physical contact between tumor cells and the serum-activated fibroblasts and cannot be replicated by fibroblast-conditioned medium, co-culture in transwells, or co-culture on a monolayer of another cell type (B16 cells). Moreover, we showed that serum-activated fibroblasts exhibit many molecular similarities to the fibroblasts that naturally infiltrate human breast cancer, including expression of syndecan-1, α-smooth muscle actin, and genes related to myofibroblast differentiation and activation by TGF-β. Finally, we demonstrated that fibroblasts expressing α-smooth muscle actin are ubiquitously present in metastases of human breast cancer and appear to be in close proximity or direct physical contact with nearly all of the tumor cells. On aggregate, these results provide further evidence of the essential role played by serum-activated fibroblasts in promoting the clonogenic growth of small numbers of breast cancer cells, both in vitroand in vivowithin metastases. It has long been known that feeder cell layers of fibroblasts can promote the growth of various cell lines in culture. Surprisingly few studies, however, have directly related this effect of feeder cells to the clonogenic growth of small numbers of primary breast cancer cells in vitro. In one such study investigators separated human breast cancer cell lines from various types of fibroblasts with a microporous membrane \[[@B7]\]. Under these conditions, tumor-associated fibroblasts did not influence the proliferation of the two breast cancer cell lines that were tested. This finding is consistent with our observation that fibroblast conditioned medium also does not promote the survival or clonogenic growth of small numbers of breast cancer cells, and suggests that a soluble growth factor or mediator is not involved in the growth enhancement effect produced by fibroblast feeder cell layers. In a later study, Brooks and coworkers \[[@B8]\] demonstrated that tumor-derived epithelial cells grew significantly better on stroma produced by mammary fibroblasts than on bone marrow stroma. Although it is generally difficult to culture malignant epithelial cells from primary human breast cancers, those investigators noted that a large proportion of their samples (78%) successfully grew when co-cultured with fibroblasts. They attributed this observation to their source of tumor cells (metastases to lymph nodes), but an equally plausible explanation is that the stromal cell layers that they used in their experiments promoted the growth of small numbers of tumor cells. Our study significantly extends the findings of Brooks and coworkers \[[@B8]\] by demonstrating that the ability of serum-activated fibroblasts to enhance the clonogenic growth of small numbers of breast cancer cells in vitrois not limited to cancer cells that have metastasized to lymph nodes. Although serum-activated fibroblasts promoted clonogenic growth of breast cancer cells in our experimental model, they did not appear to have a similar effect on benign breast epithelial cells. Specifically, cells derived from benign breast tissues yielded only small colonies of cells with benign cytologic features as compared with cells derived from malignant tissues. We attribute the slow growth of the benign epithelial cells from the normal breast samples to our use of a relatively un-enriched cell culture medium that had been optimized for growth of normal fibroblasts rather than epithelial cells. Benign mammary epithelial cells normally require more specialized additives such as epidermal growth factor and bovine pituitary extract for optimal growth. Hence, our clonogenic and cytologic findings clearly suggest that serum-activated fibroblasts have a more potent growth-promoting effect on malignant epithelial cells than on benign breast cells. Therefore, an important practical implication of our findings is that the well known difficulty in growing primary human breast cancer cells in vitromight be alleviated by simply co-culturing the tumor cells with a monolayer of serum-activated, nonsenescent and nontransformed fibroblasts. In this regard, it is interesting to note that virtually all current methods for growing primary human breast cancer cells rely on culturing purified epithelial cells that have been separated from the fibroblasts \[[@B10],[@B11]\] in order to prevent overgrowth by fibroblasts. Our data indicate that a monolayer of serum-activated fibroblasts provides an excellent substrate for growing small numbers of primary, human breast cancer cells in vitrowithout overgrowth by fibroblasts. In our model, the normal, nonsenescent fibroblasts merely formed a monolayer and then appeared to stop growing. We attribute this property to our use of normal, nonsenescent fibroblasts that exhibited contact inhibition of growth, which is characteristic of nontransformed cells. Our observations also raise several important questions. For example, how does serum activate fibroblasts? To what extent do serum-activated fibroblasts in vitroresemble their in vivocounterparts in human breast cancer? By what mechanism(s) do serum-activated fibroblasts promote the clonogenic growth of breast cancer cells? Serum -- the fluid component of clotted blood -- contains abundant coagulation-related substances, including thrombin, TGF-β released from degranulating platelets, and other serine proteases. The data from our gene expression studies raise the intriguing possibility that TGF-β released from degranulating platelets into clotting blood might be involved in activating the fibroblasts. Specifically, we observed high levels of expression of a chloride intracellular channel gene, CLIC4, that is differentially expressed in fibroblasts converted to myofibroblasts by TGF-β \[[@B20]\]. Morover, we observed that serum-activated fibroblasts expressed high levels of genes encoding induced TGF-β and latent TGF-β binding protein, which are expressed at high levels in cultured human myofibroblasts \[[@B23]\]. Inasmuch as TGF-β has previously been shown to be a powerful activator of fibroblasts, our experimental observations lead us to speculate that TGF-β released from degranulating platelets into serum could play a major role in the activation of tumor-associated fibroblasts by serum. Experiments using DNA microarrays have demonstrated that fibroblasts exhibit a temporal program of gene expression in response to serum exposure \[[@B24]\]. Notably, many features of the transcriptional program appeared to be related to the physiology of normal wound repair, including the acquisition of a myofibroblast expression profile \[[@B24]\]. In subsequent studies the gene expression signature of the fibroblast serum response was also found in tumors and in healing wounds \[[@B21]\], possibly because both processes involve localized blood clotting and production of serum. Our observations are consistent with and significantly extend these previous findings. Specifically, we showed that serum-activated fibroblasts express multiple genes that are characteristic of tumor-associated myofibroblasts, including the gene encoding CXC chemokine ligand-12, which was recently shown to be highly overexpressed in myofibroblasts associated with breast cancer \[[@B22]\]. In addition, we report the expression of genes related to TGF-β by serum-activated fibroblasts. This finding is important because a number of previous studies have demonstrated that primary breast cancer fibroblasts secrete abundant TGF-β \[[@B25],[@B26]\]. Finally, our immunocytochemical studies demonstrated that serum-activated fibroblasts uniformly express high levels of syndecan-1, a proteoglycan that is abundantly and specifically expressed by the stromal cells and myofibroblasts that are present in the connective tissue of human breast cancer \[[@B27],[@B28]\]. Hence, we conclude that the serum-activated fibroblasts used in our co-culture model have many important similarities to the tumor-associated myofibroblasts that are naturally found in breast cancer \[[@B13]\]. Interestingly, our gene expression data also suggest that serum may play a more important role than contact with tumor cells in activating fibroblasts because the gene expression profile of serum-activated fibroblasts co-cultured with tumor cells was essentially the same as the gene expression profile of serum-activated fibroblasts cultured alone. For example, the genes encoding cyclo-oxygenase-2 and myofibroblast-related proteins were upregulated relative to serum-starved fibroblasts in serum activated fibroblasts, regardless of whether the fibroblasts were co-cultured with breast cancer cells. This finding is consistent with the recent report of upregulation of the COX2gene in stromal fibroblasts co-cultured with pancreatic cells in a transwell cell culture system \[[@B29]\]. In that report, however, the authors attributed the COX2upregulation to a soluble factor produced by the pancreatic tumor cells. Our results clearly indicate that serum alone can also upregulate COX2gene expression. In our study we did not directly explore the mechanism by which serum-activated fibroblasts promote the clonogenic growth of breast cancer cells. Elenbaas and Weinberg \[[@B1]\] suggested that tumor-associated fibroblasts may promote the growth of cancer cells through production of extracellular matrix proteins and growth factors. A recent study conducted by Palmieri and coworkers \[[@B30]\] indicated that fibroblast growth factor 7 (keratinocyte growth factor) secreted by breast fibroblasts is a potent paracrine growth factor for human breast epithelial cells. The data from our transwell experiments, however, suggest that soluble growth factors by themselves are unlikely to promote clonogenic growth of human breast cancer. Instead, direct physical contact between fibroblasts and tumor cells appears to be necessary. A reasonable mechanism to account for the necessity of direct physical contact was recently proposed by Maeda and coworkers \[[@B31]\], who showed that syndecan-1 expression by mouse embryonic fibroblasts promotes proliferation of human breast cancer cell lines. Because the heparan sulfate proteoglycans of syndecan-1 are known to facilitate the assembly of signaling complexes between growth factors and their cognate receptors on breast cancer cells \[[@B32],[@B33]\], Maeda and coworkers concluded that stromal syndecan-1 on the surface of fibroblasts is a crucial factor regulating epithelial--stromal interactions in breast cancer. This model would explain the observation that syndecan-1 is required for Wnt-1-induced mammary tumorigenesis in mice \[[@B33]\] and would also account for our experimental findings. Conclusion ========== Our observations establish serum-activated fibroblasts as a critical promoter of clonogenic growth of small numbers of human breast cancer cells in vitroand suggest that similar activity might also be present in metastases of breast cancer in vivo. Further analysis of this interaction may lead to the development of novel therapeutic targets. Abbreviation ============ TGF = transforming growth factor. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= MKS conceived the study and drafted the manuscript. JT designed and performed the co-culture experiments. GC and JT jointly performed and analyzed the DNA microarray (gene expression) studies. Acknowledgement =============== This work was supported by funds from the Chao Comprehensive Cancer Center at the University of California, Irvine and the University of California Cancer Research Coordinating Committee. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Microscopic appearance of tumor cells and fibroblasts. (a)Numerous clumps of UACC-812 tumor cells resembling abortive mammary glands are evident. (b)UACC-812 breast cancer cells seeded at low density on a monolayer of serum-activated fibroblasts. Note the gland formation and the invasive infiltration by tumor cells into fibroblasts. Primary tumor cells derived from (c)human ductal carcinoma in situand from (d)invasive ductal carcinoma co-cultured with fibroblasts yielded clonogenic growth next to myofibroblasts that expressed α-smooth muscle actin (pink immunostain). Breast cancer cells grown (e)on a monolayer of B16 cells or (f)without a monolayer of fibroblasts generally grew as individual cells identifiable by their red staining for epithelial membrane antigen. (g)Serum-activated fibroblasts also uniformly expressed abundant syndecan-1, as detected by immunocytochemical staining. ::: ![](bcr995-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Results of the co-culture assay. Each bar represents the mean and standard deviation of triplicate measurements. Tumor cells seeded on a monolayer of serum-activated fibroblasts produced the highest average number of colonies. There were significantly fewer colonies if fibroblasts or serum were omitted. Co-culture with fibroblasts separated by permeable transwell inserts or on a monolayer of B16 cells yielded low numbers of colonies that were statistically indistinguishable from the control using tumor cells grown in serum-supplemented medium without fibroblasts. ::: ![](bcr995-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Results of flow cytometric measurements of proliferation and apoptosis. There was a distinct peak of (b)breast cancer cells with a high proliferative rate compared with (a)breast cancer cells cultured without fibroblasts. (c)Tumor cells grown on fibroblasts had a significantly higher proliferative rate and a lower apoptotic rate than did tumor cells cultured alone. ::: ![](bcr995-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Immunohistochemical staining for α-smooth muscle actin in metastases of human breast cancer to (a)bone marrow, (b)liver, (c)lymph node, and (d)pleura. In all of the cases that we examined, there was extensive infiltration of the tumor metastases by myofibroblasts (brown stain). Note that the myofibroblasts appeared to be in direct contact with nearly all of the tumor cells in these sections. Even very early micrometastases (defined as fewer than 100 tumor cells) of well differentiated tumors to lymph nodes (panel c) and larger metastases of poorly differentiated tumors (panel d) were in very close proximity to myofibroblasts. Serial sections of the same tissues did not stain for CD10 or smooth muscle myosin heavy chain (not shown), thereby confirming that the cells are myofibroblasts rather than myoepithelial cells. Hematoxylin counter-stain was used. ::: ![](bcr995-4) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Numbers of colonies of breast cancer cells cultured with or without fibroblasts ::: Ductal carcinoma in situ(n= 3) Invasive carcinoma (n= 3) ------------------------------------ ----------------------------- ------------------ --------------------- With fibroblasts Without fibroblasts With fibroblasts Without fibroblasts 6 2 1 0 1 0 130 33 65 12 21 0 ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Myofibroblast-related genes that are differentially upregulated in fibroblasts activated by serum compared with serum-starved fibroblasts ::: Gene symbol Gene name Fold increase in signal intensity in serum-activated fibroblasts versus serum depleted fibroblasts (P\< 0.001) ------------- ---------------------------------------------------------- ------------------------------------------------------------------------------------------------------------------ LOXL2 Lysyl oxidase-like 2 8.2 PLOD2 Procollagen-lysine, 2 oxoglutarate 5-dioxygenase 16 PLAUR Plasminogen activator urokinase receptor 5.3 TAGLN Transgelin 6.0 TPM2 Tropomysin 2 (beta) 4.6 MYL6 Myosin, light polypeptide 6, alkaline, smooth muscle 5.7 TFPI2 Tissue factor pathway inhibitor 2 4.8 SMAD3 Mothers against decapentaplegic homolog 3 (Drosophila) 4.0 CXCL12 Chemokine (C-X-C motif) ligand 12 4.5 TGFB1 TGF-β, induced 26.5 LTBP2 Latent TGF-β binding protein 2 30.2 COX2 Cyclo-oxygenase-2 16.4 CLIC4 Chloride intracellular channel 4 32 TGF, transforming growth factor. :::	PubMed Central	2024-06-05T03:55:57.459609	2005-1-28	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143574/", "journal": "Breast Cancer Res. 2005 Jan 28; 7(3):R274-R283", "authors": [ { "first": "Michael", "last": "Samoszuk" }, { "first": "Jenny", "last": "Tan" }, { "first": "Guillaume", "last": "Chorn" } ] }
PMC1143575	see related Commentary: <http://breast-cancer-research.com/content/7/3/83> Introduction ============ Body mass index (BMI) during adulthood is related to breast cancer incidence, although the direction of the association varies by menopausal status. More overweight women have a lower risk of breast cancer before menopause but a higher risk after menopause \[[@B1],[@B2]\]. In premenopausal women, greater adiposity may increase the frequency of anovulatory menstrual cycles, thus reducing exposure to ovarian hormones \[[@B3]\]. In contrast, greater adiposity in postmenopausal women increases both estrogen levels and breast cancer risk, which is probably due to the conversion of androgens to estrone in adipose tissue \[[@B3]\]. Despite the consistency of evidence for adult BMI, few studies have examined the role of body fatness during childhood and early adolescence, and results have been inconclusive. The majority have been case-control studies in which body fatness at young ages was recalled after diagnosis of breast cancer. Some \[[@B4]-[@B7]\] but not all \[[@B8]-[@B11]\] of these studies have shown inverse associations between body fatness at young ages and breast cancer risk. In a study that linked census records for residents of Hawaii to tumor registry data \[[@B12]\], a strong inverse association was seen between prospectively recorded body fatness from ages 10--14 years and incidence of premenopausal breast cancer; a study conducted in Finland that used height and weight measurements for ages 7--15 years from school health records \[[@B13]\] yielded similar findings. However, limited information on potential confounders was available in these studies. In an analysis that combined retrospective and prospective data within the earlier Nurses\' Health Study \[[@B14]\], inverse associations were observed for recalled body fatness at ages 5, 10, and 20 years in relation to incidence of both premenopausal and postmenopausal breast cancer, with the strongest inverse association of these three for body fatness at age 10 years. A prospective cohort study conducted in Norway and Sweden \[[@B15]\] found inverse associations of perceived body shape at age 7 years and BMI at age 18 years with incidence of premenopausal breast cancer, but these associations were attenuated and no longer significant when adjusted for BMI at enrollment. A recent Danish study that used information from school health records \[[@B16]\] also found a modest inverse association between BMI at age 14 years and breast cancer risk, although no data on adult BMI were available. A variety of evidence points to the importance of early life factors in the etiology of breast cancer. Studies of mammary gland development in rats have shown that developing breast tissue may be most vulnerable to carcinogens before the first birth, when undifferentiated cells are undergoing rapid proliferation \[[@B17],[@B18]\]. Epidemiologic studies conducted in humans and biomathematical models also suggest that the years between menarche and first birth may be a critical time period for breast carcinogenesis \[[@B19]\]. In light of these findings, it is plausible that body fatness at young ages could influence risk of breast cancer later in life. We examined prospectively the relation between body fatness during childhood and adolescence and incidence of breast cancer among premenopausal women in the Nurses\' Health Study II (NHS II). In addition, we investigated whether this relation is independent of characteristics of the menstrual cycle and BMI during adulthood. Materials and methods ===================== Study design and population --------------------------- The NHS II is a prospective cohort study that began in 1989, when 116,671 female registered nurses between the ages of 25 and 42 years completed a mailed, self-administered questionnaire about their health behaviors, lifestyle factors, and medical histories. Follow-up questionnaires have been sent to participants every 2 years to obtain updated information on risk factors and disease diagnoses, and the response rate for each biennial questionnaire has been greater than 90%. Deaths are reported by family members and the postal service, and regular searches of the computerized National Death Index are also conducted \[[@B20]\]. This analysis includes the 109,267 premenopausal women who provided information on their body fatness at ages 5, 10, and 20 years on the initial questionnaire in 1989 and who had no history of cancer (other than nonmelanoma skin cancer). The study was approved by human research committees at the Harvard School of Public Health and Brigham and Women\'s Hospital. Ascertainment of breast cancer cases ------------------------------------ On each of the biennial questionnaires between 1989 and 2001, participants were asked whether they had been diagnosed with breast cancer during the previous 2 years or since the last questionnaire they returned. Study physicians then confirmed the self-reported diagnoses by reviewing participants\' medical records and/or pathology reports. A total of 1318 cases of breast cancer (1044 invasive, 274 in situ) were reported and subsequently confirmed among eligible participants during 12 years of follow-up, from 1989 to 1 June 2001. Because epidemiologic studies have generally shown similar risk factors for in situand invasive breast cancer \[[@B21]-[@B23]\], both are included in our primary analyses, although secondary analyses in which the outcome was restricted to invasive disease were also conducted. Assessment of body fatness at young ages ---------------------------------------- In 1989, NHS II participants recalled their body fatness at ages 5, 10, and 20 years using a 9-level figure drawing (Fig. [1](#F1){ref-type="fig"}) originally developed by Stunkard and colleagues \[[@B24]\]. Participants who did not report their body fatness at one or more of these ages were excluded. To obtain estimates of childhood and adolescent body fatness, we averaged each participant\'s figures at ages 5 and 10 years (childhood) and at ages 10 and 20 years (adolescence); the goal of this approach was to reduce the effects of random error in the assessment of body fatness. Changes in body fatness between ages 5 and 10 years, between ages 10 and 20 years, and between ages 5 and 20 years were also calculated by subtracting each participant\'s figure (levels 1 through 9) at the younger age from that at the older age. Must and colleagues \[[@B25]\] evaluated the validity of remote recall of body fatness among 181 participants in the Third Harvard Growth Study, a longitudinal study of physical and mental growth in children that was conducted between 1922 and 1935 in the Boston area. Height and weight were measured as part of annual examinations during childhood and adolescence and were used to calculate BMI in kg/m^2^. In 1988 and 1989, when participants were between ages 71 and 76 years, they were interviewed again and asked to recall their body fatness at ages 5, 10, 15, and 20 years, using the same 9-level figure drawing as that on the 1989 NHS II questionnaire. Pearson correlations between recalled body fatness and BMI at approximately the same ages were 0.60 for age 5 years, 0.70 for age 10 years, 0.75 for age 15 years, and 0.66 for age 20 years. Other studies have yielded similar findings \[[@B26]-[@B29]\], indicating that these figure drawings can provide a reasonably accurate assessment of body fatness at young ages. Among participants in the Third Harvard Growth Study, which provides the best available \'gold standard\' for the interpretation of the figure drawing, the median values for BMI from measured height and weight at age 15 years according to recalled figure at age 15 years were 19, 19.5, 24, 25, 25.5, 30, and 32 kg/m^2^for levels 1 through 7, respectively \[[@B25]\]; no participants reported being greater than level 7. In the present study, the median values for BMI at age 18 years (based on self-reported data) according to recalled figure at age 20 years were 18.1 for level 1, 19.1 for level 2, 20.5 for level 3, 22.3 for level 4, 24.9 for level 5, 28.2 for level 6, 32.0 for level 7, 34.7 for level 8, and 37.8 kg/m^2^for level 9. Assessment of other risk factors -------------------------------- Information on other established and hypothesized breast cancer risk factors was collected at various points during the course of the study. Age, menopausal status, reproductive history, oral contraceptive use, smoking status, and diagnosis of benign breast disease were reported at baseline in 1989 and updated on each of the biennial questionnaires. History of breast cancer in a first-degree relative (mother or sister) was reported in 1989 and updated in 1997. Recent alcohol consumption was assessed on the initial questionnaire in 1989 and again in 1991, 1995, and 1999 from semiquantitative food frequency questionnaires, which also evaluated other dietary factors. Participation in physical activity was evaluated in 1989, 1991, and 1997. Height, alcohol consumption between ages 15 and 17 years and ages 18 and 22 years, and strenuous physical activity during high school and between ages 18 and 22 years were assessed in 1989. Other early life factors such as birthweight and being breastfed as an infant were reported on the 1991 questionnaire. A subset of participants (n= 43,317) provided information on dietary factors during adolescence by completing a supplementary questionnaire on high school diet in 1998. Weight at age 18 years was reported in 1989, and current weight was reported on each of the biennial questionnaires; these were used with height to calculate BMI at age 18 years and current BMI. Age at menarche, time from menarche until onset of regular menstrual cycles, and cycle regularity and length during high school and between ages 18 and 22 years were reported on the 1989 questionnaire, and recent menstrual cycle characteristics were reported in 1993. Analysis -------- Participants contributed person-time from the return date of the 1989 questionnaire until the report of breast cancer or other cancer (except nonmelanoma skin cancer), menopause, death, or the end of follow-up on 1 June 2001. Cases and person-time were assigned to the appropriate level of body fatness at each age and other risk factors. For time-varying covariates such as oral contraceptive use and parity, person-time was re-assigned every 2 years. Breast cancer incidence rates for each level of body fatness at ages 5, 10, and 20 years, and during childhood and adolescence were calculated as the number of breast cancer cases divided by the total number of person-years at each level. Few participants recalled their body fatness as greater than level 5 at ages 5 and 10 years; for example, for body fatness at age 10 years, 9.1% of participants recalled their figure as level 5, 2.6% recalled their figure as level 6, 0.4% recalled their figure as level 7, and only 0.1% recalled their figure as either level 8 or level 9. For this reason, figures 5 through 9 were combined into a single category for most analyses, and similar categories were created for childhood and adolescent body fatness. Relative risks (RRs) were calculated by taking the ratio of the incidence rate for each level compared with the lowest level, which was used as the referent category. Cox proportional hazards regression was used to estimate multivariate RRs and 95% confidence intervals (CIs) for body fatness at young ages while adjusting for age, time period, and other covariates. Body fatness at each age was examined in a separate Cox model, because recalled figures at these ages are highly correlated with one another. Indicator variables were used to obtain RRs for levels of body fatness at each age, and tests for linear trend were conducted by entering body fatness at each age into a Cox model as an ordinal variable with values 1 through 5. Changes in body fatness between ages 5, 10, and 20 years were also examined in Cox models, using the categories decreased, no change, increased 1 level, and increased 2 or more levels, with no change being the referent category. Because BMI during adulthood and menstrual cycle characteristics could be intermediate factors on the pathway from body fatness at young ages to breast cancer, these were considered separately from other covariates. To determine whether the association between body fatness at young ages and breast cancer is independent of later BMI, we adjusted for later BMI using several different variables, included in separate models: BMI at age 18 years, current BMI (updated every 2 years), and the cumulative average of BMI at age 18 years and all subsequent BMI reports up until the current time period. These variables were each divided into five categories (not including those who were missing) and included as continuous terms by assigning the median value of each category. Childhood body fatness and body fatness at age 20 years were also included in the same multivariate model, to examine the estimates mutually adjusted for one another. Height was included as a continuous term in all multivariate models, because it is positively associated with breast cancer incidence \[[@B1]\] and may reflect nutritional status during childhood \[[@B30]\]. To determine whether body fatness at young ages may have an impact on breast cancer risk by altering characteristics of the menstrual cycle, we first adjusted for age at menarche and years from menarche until the onset of regular menstrual cycles, because greater body fatness at young ages has been associated with earlier menarche but longer time until the establishment of regular cycles \[[@B31]\]. In addition, we considered adjustment for regularity and length of cycles during several different time periods: high school, ages 18--22 years, and recent (reported only once in 1993). With the exception of age at menarche, which was modeled as a continuous variable, indicator variables were used to represent categories of menstrual cycle characteristics. After conducting these analyses in the full sample, we then restricted them to certain subgroups of participants: women who had had at least one screening mammography during the follow-up period (n= 89,129; 1223 cases), women who had no reports of infertility (n= 82,892; 991 cases), and women who had no history of irregular menstrual cycles (n= 51,909; 666 cases). This was done to explore further whether differential screening or anovulation could account for associations between body fatness at young ages and premenopausal breast cancer incidence. The 1044 invasive cases were also examined in separate models and further stratified by tumor size (574 cases \<2 cm; 321 cases ≥ 2 cm) and hormone receptor status (604 estrogen receptor positive cases; 252 estrogen receptor negative cases; 571 progesterone receptor positive cases; 267 estrogen receptor negative cases) when this information was available. We also examined whether associations for childhood fatness varied according to family history of breast cancer (yes, no), age at menarche (\<12 years, ≥ 12 years), parity (nulliparous, parous), oral contraceptive use (never or past use \<4 years, past use ≥ 4 years, or current use), birthweight (\<7 lb, ≥ 7 lb), BMI at age 18 years (\<22 kg/m^2^, ≥ 22 kg/m^2^), or current BMI (\<25 kg/m^2^, ≥ 25 kg/m^2^). Separate models were constructed within each level of these factors to obtain stratum-specific estimates, and interaction terms were created by multiplying childhood body fatness as an ordinal variable by the level of each potential modifier. Wald tests were then used to evaluate whether the trends for body fatness were significantly different according to these factors. Results ======= The 109,267 premenopausal women in the analytic cohort contributed a total of 1,044,691 person-years of follow-up. Fatness levels at each of these ages were positively correlated with one another and with BMI at age 18 years and in 1989, although the correlations decreased with time. For example, the Spearman correlation between figures at ages 5 and 10 years was 0.81, whereas the correlation between figure at age 5 years and BMI in 1989 was only 0.25. Body fatness at young ages was also associated with other characteristics (Table [1](#T1){ref-type="table"}). Women who were fatter at age 10 years were heavier at birth, had earlier menarche, had higher caloric intake and were less likely to have participated in strenuous physical activity during adolescence, and in later life they were more likely to be nulliparous and to smoke. A greater proportion of participants in both the lowest and highest categories of body fatness at age 10 years reported 3 or more years from menarche until the onset of regular menstrual cycles, whereas those in the highest category were slightly more likely to have had irregular or long cycles between ages 18 and 22 years. Body fatness at ages 5, 10, and 20 years were each inversely associated with premenopausal breast cancer risk (Table [2](#T2){ref-type="table"}). The multivariate RRs for the most overweight (figure level ≥ 5) compared with the most lean (figure level 1) were 0.57 (95% CI 0.43--0.75; Pfor trend = 0.001) for age 5 years, 0.61 (95% CI 0.49--0.76; Pfor trend \< 0.0001) for age 10 years, and 0.70 (95% CI 0.52--0.94; Pfor trend \< 0.0001) for age 20 years, after adjustment for age and time period, birthweight, height, recent alcohol consumption, parity and age at first birth, recency and duration of oral contraceptive use, history of benign breast disease, and family history of breast cancer. The associations of average childhood and adolescent body fatness with breast cancer incidence were slightly stronger than the associations at individual ages (Table [3](#T3){ref-type="table"}), with multivariate RRs of 0.48 (95% CI 0.35--0.65) and 0.57 (95% CI 0.39--0.83) for the most overweight compared with the most lean in childhood and adolescence, respectively (Pfor trend \< 0.0001). Additional adjustment for adolescent intakes of animal fat, vegetable fat, and vitamin E \[[@B32],[@B33]\] among those participants who completed the high school diet questionnaire, physical activity during adolescence and adulthood, and recent smoking did not materially change the RRs. Increases in body fatness during childhood and adolescence were also inversely associated with breast cancer risk (Table [3](#T3){ref-type="table"}). Compared with participants who stayed at the same level from age 5 to age 20 years, the multivariate RR for those who increased 2 or more levels was 0.86 (95% CI 0.74--1.02). This association became stronger after adjustment for body fatness at age 5 years, with a multivariate RR of 0.77 (95% CI 0.65--0.91) for those who increased 2 or more levels. Similar patterns were observed for changes in body fatness from ages 5 to 10 years and from ages 10 to 20 years. We then evaluated whether the inverse association for childhood body fatness was independent of later BMI and menstrual cycle characteristics (Table [4](#T4){ref-type="table"}). When average childhood body fatness was adjusted for the cumulative average of BMI at age 18 years and subsequent BMI, the association was only slightly attenuated; the multivariate RR was 0.52 (95% CI 0.38--0.71) for the most overweight compared with the most lean (Pfor trend = 0.001). The results were similar when childhood body fatness was adjusted for BMI at age 18 years and current BMI separately. Adjustment for menstrual cycle characteristics had virtually no impact on the association for childhood body fatness. Furthermore, when the analyses were restricted to participants who reported having regular menstrual cycles during both adolescence and adulthood, and to participants with no reports of infertility, the multivariate RRs for childhood body fatness were almost identical to those in the full sample (data not shown). In separate models in which BMI at age 18 years and current BMI were adjusted for childhood body fatness, each remained inversely related to breast cancer incidence, although the associations were attenuated and less strong than the association for childhood body fatness. The multivariate RRs for the highest versus the lowest categories of BMI at age 18 years and current BMI, adjusted for childhood body fatness, were 0.84 (95% CI 0.66--1.07) and 0.86 (95% CI 0.71--1.04), respectively. When average childhood body fatness and body fatness at age 20 years were included in the same model as indicator variables, the multivariate RR for the most overweight compared with the most lean in childhood was 0.56 (95% CI 0.40--0.78; Pfor trend = 0.03), whereas the multivariate RR for the most overweight compared with the most lean at age 20 years was 0.82 (95% CI 0.59--1.14; Pfor trend = 0.02). The associations for childhood and adolescent body fatness were slightly stronger in the analyses including only invasive cases. The multivariate RR for the most overweight compared with the most lean during childhood was 0.45 (95% CI 0.31--0.64; Pfor trend \< 0.0001) and during adolescence was 0.52 (95% CI 0.33--0.80; Pfor trend \< 0.0001). The decreased risk was apparent both for small and large tumors and for hormone receptor positive and negative tumors (data not shown), although some of the estimates were imprecise because of small numbers of cases. The associations were also very similar among the subgroup of women who had had at least one screening mammography during the follow-up period (data not shown). The observed associations for childhood body fatness did not differ appreciably by family history of breast cancer, age at menarche, parity, oral contraceptive use, or birthweight (data not shown). The inverse trend across categories of childhood body fatness was somewhat more apparent among those who were heavy than among those who were lean at age 18 years, and the test for interaction was marginally significant (χ^2^~1~= 3.68; P= 0.06), although the multivariate RRs for the most overweight compared with the most lean in childhood were very similar in both categories of BMI at age 18 years (RR = 0.42 among those with BMI \<22 kg/m^2^at age 18 years, Pfor trend = 0.08; and RR = 0.49 among those with BMI ≥ 22 kg/m^2^at age 18 years, Pfor trend = 0.001). In contrast, the association for childhood body fatness did not differ according to current BMI (data not shown). Discussion ========== In this prospective study we observed a significant inverse association between body fatness during childhood and adolescence and incidence of breast cancer in premenopausal women, with approximately 50% lower risk for the most overweight compared with the most lean in childhood. The magnitude of the decrease in risk was greater for childhood body fatness than for body fatness at older ages. The inverse association was independent of later BMI and menstrual cycle characteristics, suggesting that body fatness at young ages may influence breast cancer risk through other biologic pathways. Our findings are consistent with the results of some other studies that have examined this relationship. Le Marchand and colleagues \[[@B12]\] linked prospectively recorded information on height and weight from census data for over 38,000 women to the Hawaii Tumor Registry, from which 607 cases of breast cancer were identified. In that study BMI at ages 5--9 years, 10--14 years, and 20--24 years were each inversely associated with breast cancer incidence, but the strongest association was for BMI from ages 10--14 years, with an odds ratio of 0.51 for the highest versus the lowest tertile. A second prospective study of 3447 women born at the University Hospital of Helsinki \[[@B13]\] obtained anthropometric measurements from birth and school health records and linked this information to the National Hospital Discharge Registry and the Cause of Death registry, identifying 177 incident cases of breast cancer. BMI at ages 7--15 years were inversely associated with breast cancer risk, with a RR of 0.83 for each 1 kg/m^2^increase in BMI at age 7 years. Both studies, however, lacked information on other important breast cancer risk factors that might confound these associations. In a prospective cohort study conducted in Norway and Sweden \[[@B15]\], the investigators observed inverse associations of perceived body shape at age 7 years and BMI at age 18 years with incidence of premenopausal breast cancer, with approximately 30% decreased risk for those who were fat or very fat at age 7 years compared with those who were average. In contrast to our findings, the association was strongest for adult BMI at enrollment, and the associations for perceived body shape at age 7 years and BMI at age 18 years were no longer significant after this adjustment. However, only 733 cases were included in that study, which could explain the lack of significance. A recent record linkage study conducted in Denmark that included over 3000 breast cancer cases \[[@B16]\] also observed a modest inverse association for BMI at age 14 years, based on information from school health records, but data on adult BMI were not available. Other epidemiologic studies have assessed body fatness at young ages through recall. Most of these utilized a case-control design in which women with breast cancer and cancer-free controls were asked to categorize their relative weight compared with other girls at specific ages. The majority found inverse associations for body fatness during the childhood and teenage years, observing a 30--50% decrease in risk for those who reported being heavier or much heavier compared with those who recalled being thin or average size \[[@B4],[@B5],[@B7],[@B10]\]. Several others \[[@B8],[@B9],[@B11]\], however, have not found such inverse associations. Two previous studies have assessed the relationship between body fatness at young ages and breast cancer risk using the same 9-level figure drawing as in our study. In a large Swedish population-based case-control study of 3345 cases of invasive breast cancer and 3454 controls between ages 50 and 74 years \[[@B6]\], body fatness at ages 7 and 18 years were both inversely associated with postmenopausal breast cancer risk, with a RR of 0.38 for figure level 7 or greater versus level 4 at age 7 years. In a largely retrospective analysis within the earlier Nurses\' Health Study \[[@B14]\], body fatness at ages 5, 10, and 20 years were each inversely associated with risk of premenopausal breast cancer, with the strongest association for age 10 years. When body fatness at these ages were mutually adjusted for one another, the RR for figure level 5 or greater compared with level 1 was 0.60 for body fatness at age 10 years. A similar pattern was also observed for postmenopausal breast cancer. Ours is one of the first prospective studies to examine the relation between childhood body fatness and breast cancer incidence, and we were able to control for a broad range of factors, both in early life and adulthood, that were not available in earlier studies. In addition, unlike most previous studies, we adjusted for later BMI using several different variables, including the cumulative average of BMI at age 18 years and all subsequent BMI reports to obtain the best long-term measure. Even with this adjustment, the inverse associations for body fatness during childhood and adolescence remained strong and statistically significant, suggesting that greater body fatness at early stages of life, perhaps even before puberty, may lower breast cancer risk. In addition, in the analyses stratified by current BMI, greater childhood body fatness was associated with reduced risk of breast cancer among both lean and heavy women, which indicates that greater childhood body fatness may confer a lasting protective effect. The biologic mechanisms that would explain this, however, are not well understood. One theory postulates that more overweight girls may experience slower pubertal growth and sexual maturation, despite their earlier menarche \[[@B6],[@B14]\]. In the Harvard Longitudinal Study of Child Health and Development \[[@B34]\], leaner body mass at age 10 years was predictive of more rapid adolescent growth, and in the Nurses\' Health Study \[[@B14]\] adolescents in the highest two quintiles of estimated growth rate had nearly 50% increased risk of premenopausal breast cancer. Rapid adolescent growth may increase breast cancer risk by increasing levels of growth hormones and epithelial proliferation in the breast or by decreasing the amount of time for repair of DNA damage \[[@B19]\]. The effect of body fatness at young ages may also be mediated through hormonal pathways. Obesity in pre-adolescent and adolescent girls is associated with higher basal insulin levels \[[@B35],[@B36]\], which can impair oocyte maturation and stimulate androgen production in the ovary \[[@B31],[@B37]\]. Hyperinsulinemia is also associated with decreased plasma levels of sex hormone binding globulin, leading to increases in free (unbound) testosterone and estradiol, and the aromatization of excess androgen to estrogen in adipose tissue may also increase estrogen levels \[[@B38]\]. Greater waist:hip ratio has been associated with higher serum concentrations of testosterone and estradiol in prepubertal and pubertal girls in some studies \[[@B31],[@B39]\] but not all studies \[[@B40],[@B41]\]. High levels of androgens in adolescent girls are associated with metabolic features of polycystic ovary syndrome \[[@B42]\], greater frequency of anovulatory cycles \[[@B37]\], and reduced fertility later in life \[[@B43]\]. In a previous study conducted among participants in this cohort \[[@B44]\], higher BMI at age 18 years was associated with increased risk of irregular and long menstrual cycles between ages 18 and 22 years as well as increased risk of ovulatory infertility in adulthood \[[@B44]\], and greater body fatness at age 10 years was also associated with moderately increased risk of menstrual cycle irregularities and nulliparity in the present analysis after adjustment for other factors. In addition, menstrual cycle regularity and length were related to breast cancer risk among NHS II participants during the first few years of follow up \[[@B45]\]. However, the observed associations for body fatness at young ages in the present study were nearly identical among participants with no history of irregular menstrual cycles or infertility, suggesting that these are not intermediate factors and that other mechanisms may be involved. High levels of sex hormones in overweight prepubertal and pubertal girls may also have a more direct protective effect on breast tissue. Several experiments have shown that neonatal, prepubertal, or pubertal administration of estrogen, prolactin, progesterone, or testosterone in rats leads to differentiation of cells of the mammary gland as well as a substantial reduction in the incidence of mammary tumors following exposure to chemical carcinogens \[[@B46]-[@B50]\]. Hence, some have recently hypothesized that the timing of exposure to estrogens and other hormones may determine their effects on breast tissue \[[@B51],[@B52]\]. Hilakivi-Clarke \[[@B51]\] has suggested that early estrogen exposure may reduce breast cancer risk by increasing the expression of tumor suppressor genes such as BRCA1, inducing differentiation of immature breast cells into more mature ductal structures in addition to stimulating epithelial growth. Higher levels of estrogens may be protective in the breasts of young girls, which are less likely to contain malignant cells, but harmful in older women, whose breasts are more likely to have acquired transformed cells. Of course, alternative explanations for our findings cannot be ruled out entirely. Although two well designed validation studies \[[@B25],[@B29]\] demonstrated that long-term recall of body fatness using this figure drawing has high correlations with BMI at the same ages, no participants in either of those studies recalled their figure as greater than level 7 at young ages; hence, the accuracy of recall at the highest levels of body fatness could not be assessed. Furthermore, the validation studies showed that women who were obese had a greater tendency to underestimate their body fatness at young ages than those who were lean, which could exaggerate the observed association for less extreme levels of body fatness. However, this would not explain the overall association, and the decreasing trend that we observed in age at menarche across all levels of body fatness at age 10 years is strong evidence of the validity of our assessment. We also repeated the analyses using the middle category of body fatness at each age and during childhood and adolescence as the referent group, to evaluate whether the observed inverse association could possibly be explained by higher risk among participants who were extremely lean at young ages. When we did this, we still observed significantly lower risk for the most overweight compared with the middle category. For example, for average childhood body fatness, the multivariate RR for the most overweight compared with the middle category (level 2.5--3) was 0.53 (95% CI 0.39--0.72), whereas the multivariate RR for the most lean was 1.11 (95% CI 0.95--1.31); this argues against elevated risk among the most lean as the major explanation for our findings. Other unmeasured factors, especially those during early life and childhood, could also confound the associations for body fatness at young ages, although a confounder would have to be very strong to account for an association of this magnitude. Detection bias is another possibility because women who are obese as adults may be less likely to get regular screening mammograms \[[@B53]\], which could delay or reduce the chance of detection. In this population, however, the probability of having a screening mammography was not appreciably related to childhood body fatness or current BMI among women in several age groups. For example, among women ages 50--54 years in 1999, 69.2% of those who were figure level 1 at age 10 years reported having had a screening mammogram within the preceding 2 years, as compared with 71.8% of those who were figure level 5 or greater at age 10 years. These percentages were similar when examined according to current BMI in 1999 among women ages 50--54 years, although a slightly greater proportion of those in the intermediate category of BMI (23--24.9 kg/m^2^) reported having had a recent screening mammogram compared with those in either of the two extreme categories (74.4% compared with 69.6% for those with BMI \<21 kg/m^2^and 69.0% for those with BMI ≥ 30 kg/m^2^). We still observed a strong inverse association between early body fatness and breast cancer risk among women who reported having at least one screening mammography. In addition, if easier detection in lean women were the main explanation for our findings, we would have expected the association to become weaker when in situcases were excluded, which is not what occurred. The results stratified by tumor size showed an inverse association for both large and small tumors, also arguing against a detection bias. Conclusion ========== These data indicate that greater body fatness during childhood and adolescence may reduce the incidence of premenopausal breast cancer, independently of adult BMI and menstrual cycle characteristics, and that this association may be stronger than the association for BMI at later ages. These findings should be interpreted cautiously, given that greater adiposity in adolescence has numerous adverse long-term health consequences \[[@B54]\] and that obesity in postmenopausal women increases breast cancer risk \[[@B3]\]. However, they could help to elucidate the biologic mechanisms that are involved in the etiology of breast cancer. In light of the strength of our findings and their consistency with those of previous studies, identifying causal pathways that would help explain this inverse association should be a priority for future research. Abbreviations ============= BMI = body mass index; CI = confidence interval; RR = relative risk. Competing interests =================== The author(s) declare that they have no competing interests. Acknowledgements ================ The work reported in this article was supported by Public Health Service grant CA50385 from the National Cancer Institute, National Institutes of Health, and Department of Health and Human Services. HJB was supported by Department of Defense grant DAMD17-00-1-0165 in Breast Cancer Epidemiology and Prevention, and GAC is supported in part by the American Cancer Society Cissy Hornung Clinical Research Professorship. The authors thank the participants of the NHS II for their dedication to this study and Sue Malspeis for technical support. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Figure drawing 24 used to assess body fatness at different ages among Nurses\' Health Study II participants. ::: ![](bcr998-1) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Characteristics of 109,267 premenopausal Nurses\' Health Study II participants in 1989 according to body fatness at age 10 years ::: Characteristic Figure at age 10 years ---------------------------------------------------------- ------------------------ --------------- --------------- --------------- --------------- Number participants (%) 20,554 (18.8) 33,512 (30.7) 24,666 (22.6) 17,274 (15.8) 13,261 (12.1) Mean Age (years) 34.6 34.0 34.1 34.3 34.7 Total caloric intake in high school (kilocalories)^a^ 2737 2746 2745 2746 2781 Animal fat intake in high school (% energy)^a^ 26.0 25.7 25.8 25.9 26.2 Vegetable fat intake in high school (% energy)^a^ 14.4 14.7 14.8 15.0 15.0 Age at menarche (years) 12.8 12.6 12.3 12.1 12.0 Height (inches) 64.9 64.9 64.8 64.8 65.0 BMI at age 18 years (kg/m^2^) 19.3 20.2 21.5 22.9 24.3 Current BMI (kg/m^2^) 22.0 22.7 24.4 26.1 27.1 Age at first birth (years)^b^ 25.4 25.5 25.6 25.6 25.4 Parity^b^ 2.1 2.1 2.1 2.1 2.0 Percentage Birthweight ≥ 8.5 lb 8.3 9.4 11.7 12.9 13.7 ≥ 3 years from menarche until regular menstrual cycles 19.5 18.5 17.7 17.9 19.6 Irregular menstrual cycles/no periods ages 18--22 years 9.7 9.0 8.8 10.2 11.2 Menstrual cycle length ≥ 40 days ages 18--22 years 7.1 7.1 7.3 8.2 8.6 Strenuous activity 10--12 months/year in high school 28.7 28.0 26.2 21.4 17.6 BMI at age 18 years ≥ 25 kg/m^2^ 1.8 3.1 8.9 19.5 32.2 Current BMI ≥ 30 kg/m^2^ 3.0 4.9 11.2 20.2 24.6 Nulliparous 29.6 29.2 29.3 31.3 35.0 Current oral contraceptive user 13.8 14.0 12.6 12.7 11.6 Current smoker 13.1 11.5 12.3 14.2 18.6 Alcohol consumption ≥ 10 g/day 9.7 9.0 9.1 9.1 9.9 First-degree family history of breast cancer 5.9 6.0 5.7 5.9 6.3 History of benign breast disease 30.3 28.1 27.6 27.5 28.0 All means and percentages refer to the 1989 time period unless otherwise noted. Participants who had a diagnosis of cancer (other than nonmelanoma skin cancer) and those for whom body fatness data were missing at ages 5, 10, or 20 years were excluded. ^a^Among participants who completed the high school diet questionnaire in 1998. ^b^Among participants who reported that they were parous in 1989. BMI, body mass index. ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Relative risks of breast cancer by body fatness at ages 5, 10, and 20 years, and during childhood and adolescence among 109,267 premenopausal Nurses\' Health Study II participants (1989--2001) ::: Cases^a^(n= 1318) Person-years Age-adjusted RR (95% CI) Multivariate RR (95% CI)^b^ ----------------------------------------------- --------------------- -------------- -------------------------- ----------------------------- Figure at age 5 years 1 350 252,779 1.00 (ref.) 1.00 (ref.) 2 431 335,974 0.97 (0.84--1.11) 0.96 (0.83--1.11) 3 304 251,625 0.89 (0.76--1.04) 0.88 (0.76--1.03) 4 173 133,574 0.94 (0.78--1.13) 0.93 (0.77--1.12) ≥ 5 60 70,740 0.59 (0.45--0.78) 0.57 (0.43--0.75) Pfor trend^c^ 0.002 0.001 Figure at age 10 years 1 277 191,720 1.00 (ref.) 1.00 (ref.) 2 423 323,197 0.95 (0.82--1.11) 0.96 (0.82--1.11) 3 307 239,185 0.92 (0.79--1.09) 0.93 (0.79--1.10) 4 197 165,957 0.84 (0.70--1.01) 0.84 (0.70--1.01) ≥ 5 114 124,633 0.62 (0.50--0.77) 0.61 (0.49--0.76) Pfor trend^c^ \<0.0001 \<0.0001 Figure at age 20 years 1 73 44,399 1.00 (ref.) 1.00 (ref.) 2 414 271,292 1.04 (0.81--1.33) 1.05 (0.82--1.35) 3 474 397,651 0.85 (0.67--1.09) 0.87 (0.68--1.11) 4 247 219,136 0.81 (0.62--1.05) 0.81 (0.63--1.06) ≥ 5 110 112,213 0.71 (0.53--0.95) 0.70 (0.52--0.94) Pfor trend^c^ \<0.0001 \<0.0001 Average childhood figure (ages 5--10 years) 1 262 177,512 1.00 (ref.) 1.00 (ref.) 1.5--2 416 322,714 0.92 (0.79--1.08) 0.93 (0.79--1.08) 2.5--3 339 265,111 0.90 (0.77--1.06) 0.90 (0.76--1.06) 3.5--4.5 253 215,175 0.82 (0.69--0.97) 0.81 (0.68--0.96) ≥ 5 48 64,180 0.49 (0.36--0.67) 0.48 (0.35--0.65) Pfor trend^c^ \<0.0001 \<0.0001 Average adolescent figure (ages 10--20 years) 1 56 34,780 1.00 (ref.) 1.00 (ref.) 1.5--2 398 267,086 1.07 (0.81--1.41) 1.08 (0.82--1.43) 2.5--3 449 366,516 0.92 (0.69--1.21) 0.93 (0.71--1.24) 3.5--4.5 363 312,070 0.85 (0.64--1.13) 0.86 (0.65--1.14) ≥ 5 52 64,240 0.58 (0.40--0.85) 0.57 (0.39--0.83) Pfor trend^c^ \<0.0001 \<0.0001 ^a^Including both invasive and in situcases. ^b^Adjusted for age (months), time period (6 periods), birthweight (\<5.5, 5.5--6.9, 7--8.4, 8.5--9.9, ≥ 10 lb), height (inches), recent alcohol consumption (0, 0.1--1.4, 1.5--4.9, 5.0--9.9, ≥ 10 g/day), parity and age at first birth (nulliparous, 1--2 pregnancies with age at first birth \<25 years, 1--2 pregnancies with age at first birth 25--29 years, 1--2 pregnancies with age at first birth ≥ 30 years, ≥ 3 pregnancies with age at first birth \<25 years, ≥ 3 pregnancies with age at first birth 25--29 years, ≥ 3 pregnancies with age at first birth ≥ 30 years), oral contraceptive use (never, past use \<4 years, past use ≥ 4 years, current use \<4 years, current use ≥ 4 years), history of benign breast disease (yes, no), and first-degree family history of breast cancer (yes, no). ^c^Wald test of coefficient for body fatness modeled as an ordinal variable. CI, confidence interval; RR, relative risk. ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Relative risks of breast cancer by changes in body fatness between ages 5, 10, and 20 years among 109,267 premenopausal Nurses\' Health Study II participants (1989--2001) ::: Cases^a^(n= 1318) Person-years Age-adjusted RR (95% CI) Multivariate RR (95% CI)^b^ Multivariate RR + starting figure^c^ --------------------------------- --------------------- -------------- -------------------------- ----------------------------- -------------------------------------- Change from ages 5 to 10 years Decreased 76 59,972 0.97 (0.77--1.23) 0.94 (0.74--1.19) 1.04 (0.82--1.32) No change 950 720,994 1.00 (ref.) 1.00 (ref.) 1.00 (ref.) Increased 1 level 218 194,725 0.87 (0.75--1.01) 0.88 (0.76--1.02) 0.88 (0.76--1.02) Increased 2 or more levels 74 69,001 0.79 (0.62--1.00) 0.79 (0.62--1.00) 0.77 (0.61--0.98) Change from ages 10 to 20 years Decreased 223 184,978 0.86 (0.74--1.01) 0.86 (0.74--1.01) 1.06 (0.89--1.27) No change 497 353,987 1.00 (ref.) 1.00 (ref.) 1.00 (ref.) Increased 1 level 463 388,462 0.92 (0.81--1.04) 0.92 (0.81--1.05) 0.85 (0.74--0.97) Increased 2 or more levels 135 117,264 0.93 (0.77--1.12) 0.92 (0.76--1.11) 0.82 (0.67--0.99) Change from ages 5 to 20 years Decreased 172 135,479 0.91 (0.76--1.08) 0.90 (0.75--1.07) 1.06 (0.88--1.29) No change 449 321,121 1.00 (ref.) 1.00 (ref.) 1.00 (ref.) Increased 1 level 474 384,088 0.94 (0.83--1.08) 0.95 (0.83--1.08) 0.88 (0.77--1.01) Increased 2 or more levels 223 204,003 0.87 (0.74--1.02) 0.86 (0.74--1.02) 0.77 (0.65--0.91) ^a^Including both invasive and in situcases. ^b^Adjusted for age (months), time period (6 periods), birthweight (\<5.5, 5.5--6.9, 7--8.4, 8.5--9.9, ≥ 10 lb), height (inches), recent alcohol consumption (0, 0.1--1.4, 1.5--4.9, 5.0--9.9, ≥ 10 g/day), parity and age at first birth (nulliparous, 1--2 pregnancies with age at first birth \<25 years, 1--2 pregnancies with age at first birth 25--29 years, 1--2 pregnancies with age at first birth ≥ 30 years, ≥ 3 pregnancies with age at first birth \<25 years, ≥ 3 pregnancies with age at first birth 25--29 years, ≥ 3 pregnancies with age at first birth ≥ 30 years), oral contraceptive use (never, past use \<4 years, past use ≥ 4 years, current use \<4 years, current use ≥ 4 years), history of benign breast disease (yes, no), and first-degree family history of breast cancer (yes, no). ^c^Adjusted for same factors as above, plus figure at age 5 years (for change from ages 5 to 10 years and from ages 5 to 20 years) or at age 10 years (for change from ages 10 to 20 years), each modeled as an ordinal variable. CI, confidence interval; RR, relative risk. ::: ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Relative risks for breast cancer by average childhood body fatness, with adjustment for later body mass index and menstrual cycle characteristics, among 109,267 premenopausal Nurses\' Health Study II participants (1989--2001) ::: Average childhood figure (ages 5--10 years) Pfor trend^d^ -------------------------------------------------------------------------- --------------------------------------------- ------------------- ------------------- ------------------- ------------------- ---------- Multivariate RR (95% CI)^a^ 1.00 0.93 (0.79--1.08) 0.90 (0.76--1.06) 0.81 (0.68--0.96) 0.48 (0.35--0.65) \<0.0001 Multivariate RR + later BMI^b^ BMI at age 18 years 1.00 0.95 (0.81--1.11) 0.95 (0.80--1.12) 0.87 (0.72--1.05) 0.53 (0.38--0.73) 0.004 Current BMI 1.00 0.93 (0.80--1.09) 0.92 (0.78--1.08) 0.84 (0.70--1.01) 0.51 (0.37--0.69) 0.0003 Cumulatively averaged BMI 1.00 0.94 (0.81--1.10) 0.93 (0.79--1.10) 0.86 (0.72--1.03) 0.52 (0.38--0.71) 0.001 Multivariate RR + age at menarche and menstrual cycle characteristics^c^ Age at menarche 1.00 0.93 (0.80--1.09) 0.89 (0.76--1.05) 0.80 (0.67--0.95) 0.47 (0.34--0.64) \<0.0001 Age at menarche, time until regular cycles 1.00 0.93 (0.80--1.09) 0.89 (0.76--1.05) 0.80 (0.67--0.95) 0.47 (0.34--0.64) \<0.0001 Age at menarche, regularity/length of cycles, ages 18--22 years 1.00 0.93 (0.80--1.09) 0.89 (0.76--1.05) 0.79 (0.67--0.95) 0.47 (0.34--0.64) \<0.0001 ^a^Adjusted for age (months), time period (6 periods), birthweight (\<5.5, 5.5--6.9, 7--8.4, 8.5--9.9, ≥ 10 lb), height (inches), recent alcohol consumption (0, 0.1--1.4, 1.5--4.9, 5.0--9.9, ≥ 10 g/day), parity and age at first birth (nulliparous, 1--2 pregnancies with age at first birth \<25 years, 1--2 pregnancies with age at first birth 25--29 years, 1--2 pregnancies with age at first birth ≥ 30 years, ≥ 3 pregnancies with age at first birth \<25 years, ≥ 3 pregnancies with age at first birth 25--29 years, ≥ 3 pregnancies with age at first birth ≥ 30 years), oral contraceptive use (never, past use \<4 years, past use ≥ 4 years, current use \<4 years, current use ≥ 4 years), history of benign breast disease (yes, no), and first-degree family history of breast cancer (yes, no). ^b^Adjusted for same factors as above, plus later body mass index (BMI) variables individually, each modeled as medians of the categories. Cumulatively-averaged BMI is the cumulative average of BMI at age 18 years and all subsequent BMI reports up until the current time period. ^c^Adjusted for same factors as above (not including later BMI), plus menstrual cycle characteristics. Age at menarche modeled as continuous. Time until onset of regular cycles (\<1 year, 1--2 years, ≥ 3 years, never) and cycle regularity and length from ages 18--22 years (regular \<26 days, regular 26--31 days, regular ≥ 32 days, irregular \<26 days, irregular 26--31 days, irregular ≥ 32 days) modeled as indicator variables. ^d^Wald test of coefficient for body fatness modeled as an ordinal variable. CI, confidence interval; RR, relative risk. :::	PubMed Central	2024-06-05T03:55:57.462738	2005-2-18	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143575/", "journal": "Breast Cancer Res. 2005 Feb 18; 7(3):R314-R325", "authors": [ { "first": "Heather J", "last": "Baer" }, { "first": "Graham A", "last": "Colditz" }, { "first": "Bernard", "last": "Rosner" }, { "first": "Karin B", "last": "Michels" }, { "first": "Janet W", "last": "Rich-Edwards" }, { "first": "David J", "last": "Hunter" }, { "first": "Walter C", "last": "Willett" } ] }
PMC1143576	Introduction ============ There is growing evidence that combined estrogen and progestin therapy (EPT) increases the risk of breast cancer more than estrogen therapy (ET) alone \[[@B1]-[@B6]\]. One important question is whether we can identify subgroups of women who are at a particularly greater risk of developing breast cancer if they use EPT or ET. Mammographic percentage density (MPD) is a strong independent breast cancer risk factor \[[@B7]-[@B10]\] and increases when women commence EPT. On average the change is 4 to 5% \[[@B11]\]; however, a sub-group of about 25% (range 10 to 40%) of women starting EPT undergo a substantial increase in MPD of at least 10% or an upgrade in the four-level Wolfe classification \[[@B12]-[@B17]\]. Although it is not known which factors modify change in MPD in ET or EPT users, it is important to identify such factors because they might also modify the increase in risk of breast cancer associated with ET or EPT. Data from the Postmenopausal Estrogen and Progestin Interventions trial showed that the increase in serum estrone is a strong predictor of MPD increase in women randomized to EPT \[[@B18]\], suggesting that factors affecting the absorption or metabolism of EPT are important. As a next step, we decided to investigate whether known or suspected functional variants in genes involved in hormone metabolism would predict changes in MPD in women randomized to ET and EPT. We selected genes whose products are known to modulate important aspects in estrogen metabolism such as catechol-O-methyltransferase (COMT), cytochrome P450 1B1 (CYP1B1) and UDP-glucuronosyltransferase 1A1 (UGT1A1)), or in progesterone metabolism (aldo-keto reductase 1C4 (AKR1C4)). As far as we know, this is the first study to investigate genetic determinants of MPD changes in women randomized to ET, EPT or placebo. Materials and methods ===================== Study subjects -------------- Subjects were drawn from two randomized, double-blind, placebo-controlled studies \[[@B19],[@B20]\], conducted by the Atherosclerosis Research Unit at the Keck School of Medicine of the University of Southern California. The Estrogen in the Prevention of Atherosclerosis Trial (EPAT) \[[@B19]\] was a clinical trial conducted in postmenopausal women aged 45 years or older recruited from direct advertising. Eligible women had a serum estradiol level of less than 20 pg/ml and a fasting plasma low-density lipoprotein cholesterol of at least 130 mg/dl. Exclusion criteria were: use of postmenopausal hormone therapy for more than 10 years or within the previous month of the first screening visit; history of breast or gynecologic cancer; life-threatening disease with a prognosis of less than 5 years; fasting triglyceride level 400 mg/dl or more; high-density lipoprotein level less than 30 mg/dl; diastolic blood pressure more than 110 mmHg; current smoker; untreated thyroid disease; renal insufficiency (serum creatinine more than 2.5 mg/dl); fasting blood glucose more than 200 mg/dl. The 222 subjects enrolled in this study were randomized to receive either 1 mg/day of micronized 17β-estradiol (ET) or placebo over a period of 2 years. The Women\'s Estrogen--Progestin Lipid-Lowering Hormone Atherosclerosis Regression Trial (WELL-HART) \[[@B20]\] was conducted in postmenopausal women aged 50 to 75 years with angiographically demonstrable coronary artery disease. Participants were recruited from five cardiac catheterization laboratories in Los Angeles County that serve patients with diverse backgrounds. Other criteria for inclusion and exclusion were as in EPAT except that smokers of fewer than 15 cigarettes a day were not excluded from participation in WELL-HART. A total of 226 subjects were randomized to receive either 1 mg/day of micronized 17β-estradiol with medroxyprogesterone acetate (MPA) at 5 mg/day for days 19 to 30 each month (EPT), ET or matching placebo over a period of 3 years \[[@B20]\]. In the present study, we included all subjects who had participated in EPAT or WELL-HART for a minimum of 12 months, who had a current US telephone number and a mammogram within the 18 months before randomization that was at least 2 months after any previous episodes of postmenopausal hormone use based on patient-reported date of cessation or, if unavailable, patient response to screening interview question about use of any hormone therapy (HT) use in the previous month, and who did not have breast implants or a history of breast cancer between randomization and the follow-up mammogram. Potentially eligible women who were willing to participate in this mammography density sub-study signed a written informed consent form and provided a blood sample or, if they lived outside the greater Los Angeles area, a buccal cell sample. The study protocol was approved by the Institutional Review Board at the University of Southern California. Of the 222 subjects randomized in EPAT, 150 (68%) women were assessed as eligible for recruitment to the present study. The reasons why 72 women were not eligible for the current study were as follows: loss to follow-up (n= 27), death (n= 1), withdrawal from original trial (n= 42), breast implants (n= 1), and breast cancer diagnosed during the trial (n= 1). Of these 150 eligible women, we successfully contacted 149, and 146 (97%) consented and provided a blood (n= 131) or buccal (n= 15) specimen. An appropriately timed set of mammograms was available for 127 of these participants (85% of subjects contacted). Of the 226 subjects randomized in WELL-HART, 163 (72%) were eligible for recruitment. The reasons why 63 women were not eligible for the current study were as follows: loss to follow-up (n= 16), death (n= 10), withdrawal from original trial (n= 34), breast cancer diagnosed during the trial (n= 1), and not competent to sign informed consent (n= 2). We successfully contacted 155 of the 163 eligible women; 140 (89%) of these consented and provided a blood (n= 134) or buccal (n= 6) specimen. Mammograms were available for 105 of these women (68% of subjects contacted). Reasons for non-participation among eligible subjects were as follows: telephone contact unsuccessful (EPAT 1, WELL-HART 8), patient refusal (EPAT 3, WELL-HART 15), and specimen stored incorrectly (EPAT 1). Problems encountered in retrieving mammograms were as follows: baseline mammogram not eligible (EPAT 5, WELL-HART 25) and mammogram not located at the facility where it was taken and further tracking was unsuccessful (EPAT 10, WELL-HART 9). No follow-up mammogram was available for three EPAT subjects, and one WELL-HART subject was excluded because of technically poor mammographic images (see assessment score below). Data collection --------------- ### DNA samples and genotyping Genomic DNA was isolated from blood with a QIAamp 96 DNA Blood Kit (Qiagen, Valencia, CA). DNA isolation from buccal cells in mouthwash was performed with the Puregene DNA isolation method (Gentra Systems Inc., Minneapolis, MN). Genotyping for the following single nucleotide polymorphisms was performed using the fluorogenic 5\'-nuclease assay (TaqMan Assay) \[[@B21]\]: COMT(Val158Met), CYP1B1(Val432Leu), UGT1A1(\<7/≥ 7 TA repeats) and AKR1C4(Leu311Val). No signal or an indeterminate signal was recorded for one or two specimens (less than 1%) in assays for each polymorphism; these results are recorded as missing and are excluded from our analysis of that gene. Of the 5% blinded quality-control repeats, each result matched that of the corresponding specimen in all assays except one for which one of the controls gave no signal. ### Assessment of MPD change A baseline mammogram performed closest to, but before, the subject\'s randomization date and a mammogram obtained 1 year after randomization were used to measure MPD change. The mean time between the two mammograms was 14.4 months (range 9 to 31), and this was similar in all the treatment groups. In EPAT the mean time was 13.5 months in the placebo group and in the ET group 14.4 months (t-test P= 0.27); in WELL-HART the mean time was 14.9 months in the placebo group, 15.4 months in the ET group and 14.5 months in the EPT group (analysis of variance P= 0.64). All films were scanned at 150 dots per inch with a Cobrascan CX-812T scanner (Radiographic Digital Imaging Inc., Torrance, CA) with Adobe Image software. Mammographic density was determined with a computer-assisted validated method \[[@B22]\] that we have previously found strongly predicts breast cancer risk \[[@B10]\], and that we have shown identifies increases in MPD when women commence EPT \[[@B11]\]. One of us (GU) assessed all the mammograms for absolute density, and the breast area was measured by a research assistant trained by GU. MPD was calculated as the absolute dense area multiplied by 100 divided by the breast area. The correlation of repeated MPD readings performed on a subset of 14 mammograms at different times was 0.95. The density reader also assigned a \'difficulty assessment score\' for each mammogram ranging from 1 to 6, which indicated how difficult it was to read the scanned image; a score of 1 was \'normal\', and a score of 6 was \'impossible\' for technical reasons. Mammograms from one WELL-HART patient were scored as 6, and this patient was excluded from the analysis. Scores of 4 or 5 were recorded for 31 of 233 patients (13.3%). This number did not vary significantly by study group, timing of mammogram (before versus after treatment) or treatment arm (P\> 0.45 for all). Statistical analysis -------------------- We used general linear models to determine the least-square mean change in MPD for subjects in each treatment group overall and by genotype for each trial independently, and with the two trials combined. We adjusted for factors known or suspected to be associated with changes in mammographic density, namely race (White, African American, Latina, Asian/Pacific Islander), body mass index (BMI) at baseline (kg/m^2^), change in BMI on trial, age at baseline mammogram (years) and MPD at baseline and any past use of ET or EPT (ever/never). A multiplicative genotype × treatment interaction term was also included in the model to test for genotype differences in the treatment-related change in MPD. For our key findings we also report conservative adjustments of the significance levels by using the Bonferroni technique (36 comparisons, Bonferroni-adjusted α level = 0.0014). The laboratory personnel and the mammographic density assessor were blinded to treatment and study assignment. The SAS statistical software package (SAS Institute Inc., Cary, NC) was used for all statistical analyses. Results ======= Subject characteristics ----------------------- The distribution of baseline characteristics for the 232 subjects included in this study was similar to that of the parent trials, except that the WELL-HART participants included in this study were more likely to have had an education beyond high school (52%) than the WELL-HART participants not included (32%; P= 0.0003). However, education level was not associated with mammographic density at baseline or change in density. Within each trial, baseline characteristics including age, parity, BMI, family history of breast cancer, education and genotype were similar across treatment groups, except that within WELL-HART the racial distribution differed significantly by treatment assignment (P= 0.01). There were several differences between the two trials, reflecting the differences in the inclusion criteria and recruitment strategies. WELL-HART subjects were statistically significantly older, less educated, more obese, of higher parity, less likely to have used postmenopausal hormones and more racially diverse than EPAT subjects (Table [1](#T1){ref-type="table"}). Three of these factors -- age, parity and BMI -- are known to be independently and inversely associated with mammographic density \[[@B7]\] and, as expected from these differences, the mean MPD at baseline was significantly lower in WELL-HART subjects than in EPAT subjects (WELL-HART 10.6%; EPAT 20.3%; P= 0.0001). However, change in MPD did not vary with MPD at baseline (Pearson correlation coefficient -- 0.06, P= 0.33; Spearman correlation coefficient 0.03, P= 0.64). The allelic frequencies for all genes were in Hardy--Weinberg equilibrium across both trials with the exception of CYP1B1, which showed statistically significant variation by ethnicity. Because Hardy--Weinberg equilibrium was maintained within each ethnic group, a systemic genotyping problem with this locus is unlikely. Change in MPD by treatment arm ------------------------------ On average, women assigned to placebo did not exhibit any change in MPD from baseline (Table [2](#T2){ref-type="table"}). Women assigned to ET in each trial showed a similar 4 to 5% increase in MPD over placebo in each trial (EPAT P= 0.0001, WELL-HART P= 0.02). Women assigned to EPT in WELL-HART exhibited the greatest mean change in MPD (7.8%); this was statistically significantly greater than placebo (P= 0.0005) but not significantly different from the ET groups (P= 0.32). To determine whether this was due to a change in breast area or in the amount of dense tissue in the breast (absolute density), we examined the effect of treatment on changes in dense area and changes in the total breast area. The treatment effect was observed when the analysis was undertaken for change in mammographic absolute density. The adjusted mean change in absolute density was greatest in women assigned to EPT (10.5 cm^2^) in comparison with women assigned to placebo (- 0.27 cm^2^, P= 0.005) but not significantly different from women assigned to ET (9.4 cm^2^, P= 0.76). No treatment effect was observed on change in breast area (P= 0.31). The remaining analyses are therefore restricted to changes in MPD. Overall, one subject (1%) assigned to placebo showed a 10% increase in MPD over 12 months, compared with 17% of subjects assigned to ET and 32% assigned to EPT. Increases in MPD in the EPT arm were apparent within each ethnic group. Genetic determinants of MPD change ---------------------------------- Because there was no statistical significant heterogeneity in the ET effect between the two studies (MPD change 4.0% and 5.6% respectively, Pfor homogeneity of effect = 0.76), we combined the results of both studies to provide more power to investigate a treatment interaction between genotype and change in MPD. There was no statistically significant association between genotype and baseline MPD in either trial or in the trials combined (data not shown). Overall there was also no evidence for an association between change in MPD and genotype in women randomized to ET (Table [3](#T3){ref-type="table"}). However, in EPAT there was a statistically significant increase in MPD in women in the ET arm who possessed the COMTMet/Met genotype compared with those with the Val/Val genotype (P= 0.02, Pfor ET--genotype interaction = 0.17), but no such association was observed in the WELL-HART ET arm. Two of the genes studied modified the MPD changes associated with EPT use. Both the Val/Val and Leu/Val genotypes of the AKR1C4gene were associated with a statistically significant increase in density compared with the Leu/Leu genotype among women assigned to EPT (P= 0.0001, 0.0007, respectively; P= 0.004, 0.03, respectively, corrected for 36 multiple comparisons). However, there were only seven women heterozygous and one woman homozygous for the Val allele. The AKR1C4--treatment interaction was statistically significant (P= 0.001). When analyses were restricted to the WELL-HART study, the probabilities for the Val/Val and Leu/Val genotypes in comparison with the Leu/Leu genotypes were P= 0.001, 0.003, respectively, and the treatment interaction P= 0.05). There was also a statistically significant association between the CYP1B1genotype and MPD change among women taking EPT. The Leu/Leu genotype of CYP1B1was associated with a statistically significantly greater MPD change in women assigned to EPT than the Val/Val genotype (P= 0.03, not significant after correction for multiple comparisons). However, heterozygotes for this polymorphism showed the smallest increase in MPD. The interaction between ET/EPT and CYP1B1was statistically significant (P= 0.0004). The results were similar when analyses were restricted to the WELL-HART study; the probability for the Leu/Leu genotype was P= 0.09 and treatment interaction P= 0.006. The results for the COMTgene and the UGT1A1genes were also similar when analyses were restricted to the WELL-HART study (results not shown). Discussion ========== In this study we found that women randomized to ET and EPT had a statistically significant mean increase in MPD over 12 months compared with women assigned to placebo, with the women assigned to EPT having the greatest mean increase in MPD. These findings are consistent with those (using the same reader and same method) from the only published placebo-controlled randomized clinical trial with EPT \[[@B11]\]. Interestingly, there was a similar effect of ET in two diverse study populations in the current study despite significant differences in MPD at baseline and other potential confounders (age, BMI and parity) between these two populations. The lack of a statistically significant difference between the increase in MPD in the EPT arm and the ET arms might have been due to a small EPT sample size of 34 subjects. In data from the Postmenopausal Estrogen and Progestin Interventions (PEPI) trial, a greater increase in serum estrone level as a function of treatment was a significant predictor of MPD increase in women randomized to EPT, but not in women randomized to ET alone \[[@B18]\]. The PEPI study had no data on how serum progesterone or progestin levels changed. However, these results raised the possibility that factors associated with hormone absorption or metabolism are important determinants in how the breast tissue reacts to EPT. Medroxyprogesterone acetate has a similar structure and metabolic pathway to progesterone. After medroxyprogesterone is ingested it undergoes reduction and hydroxylation in the small intestine \[[@B23],[@B24]\]. After absorption it undergoes further metabolism in the liver, including 3α-hydroxylation (AKR1C4) \[[@B25]\]. The Leu311Val polymorphism on AKR1C4has been associated with a 66 to 80% decrease in the catalytic activity of the enzyme \[[@B26]\]. Consistent with this was our finding that subjects randomized to EPT who were heterozygotes or homozygotes for this low-activity allele showed significantly greater increases in MPD than homozygotes for the wild-type Leu/Leu allele. In addition, the one subject possessing two copies of the Val allele showed the greatest increase in MPD, suggesting a potential allelic dosage effect. A number of studies have investigated the role of estrogen metabolism on breast cancer risk. The major forms of estrogen, namely estrone and estradiol, are hydroxylated into 2-, 4- or 16-hydroxyestrogens. Initially, much research focused on the role of 2- and 16-hydroxy metabolites, with most of the later studies finding no protective effects of a high 2- to 16α-hydroxyestrone ratio \[[@B27]-[@B32]\]. In contrast, newer research suggests that the important question is how much estrogen is metabolized down the 4-hydroxy pathway, because the 4-hydroxy products are genotoxic \[[@B33]\]. An important enzyme involved in the 4-hydroxylation of estrogen is CYP1B1 \[[@B34]\]. After hydroxylation, these estrogens may further undergo sulfonation, glucuronidation (UGT1A1) or O-methylation (COMT), which increases the water solubility and therefore the excretion of these metabolites \[[@B35]\]. The CYP1B1Val432Leu polymorphism has been associated with breast cancer in an Asian study \[[@B36]\] but not in two studies of Caucasians \[[@B36]-[@B38]\]. A Swedish case-control study observed an increased risk of breast cancer in Leu/Leu carriers in comparison with Val/Val carriers in women who had used HT for longer than 4 years \[[@B39]\]. However, a cross-sectional study showed no association between CYP1B1genotype and mammographic density in women using HT \[[@B40]\]. In our study the CYP1B1Val432Leu polymorphism predicted the change in MPD in women randomized to EPT, although this finding was no longer statistically significant after adjustment for multiple comparisons. There was no consistent dose effect with Leu alleles because heterozygotes had the lowest density increase. Our findings are therefore consistent with the available data and we cannot exclude the possibility that this gene might have a role in gene--environment interactions in EPT users. The enzyme encoded by COMTis responsible for the conjugation and inactivation of catechol estrogen. A Val158Met polymorphism has been associated with lower activity of this enzyme \[[@B41]\] and is associated with increased plasma levels of 17β-estradiol in postmenopausal women taking ET \[[@B42]\]. In a recent cross-sectional study we reported a statistically significant association between the Met/Met allele and MPD in current users of HT (ET) \[[@B43]\]. In the present study, women assigned to ET in the EPAT study who possessed this high-risk variant showed a statistically significant increase in MPD compared with Val/Val homozygotes, but this effect was not observed in the ET arm of WELL-HART, in which the participants all had angiographically demonstrable coronary artery disease. We found no evidence that the UGT1A1polymorphism was associated with MPD increase in women assigned to ET or EPT. Strengths of our study included the randomized design, the use of a validated method and an experienced reader to assess mammographic density. However, there were several limitations. Study subjects represented only 57% and 47% of those originally randomized to EPAT and WELL-HART, respectively, because several of the participants in the parent trials had died or were lost to follow-up after the completion of the original trial. The small sample size, particularly in the EPT arm, limited our power to detect gene--environment interactions. However, it is unlikely that this could have biased our results and caused the apparent associations between genotype and change in MPD with treatment, because the most likely effect of this loss to follow-up would be to obscure a true association by a loss of statistical power. It is possible that some or all of the associations observed represent chance findings (false positives) due to multiple testing; however, our main findings of the effect of HT on MPD and the interaction with AKR1C4genotype are statistically significant after conservative correction with Bonferroni\'s technique. Another limitation is that women assigned to EPT and a subset of those on ET were drawn from a select study population with diagnosed cardiovascular disease and poor general health (the WELL-HART study). It is therefore unclear to what extent our findings can be generalized to populations with better health. The fact that we found that the COMTpolymorphism modified the effect of ET on MPD in the EPAT study but not in the WELL-HART study suggests that the women with angiographically detected heart disease in the WELL-HART study might have been different. However, similar MPD changes were observed in the ET arms of both the EPAT and the WELL-HART study. Further, our finding of the magnitude of the increase in MPD associated with EPT use was similar to that recently reported from a trial of EPT use \[[@B11]\]. Thus, although we cannot exclude the possibility that the observed modifying effects of the AKR1C4genotype were due to some other characteristic among these women with angiographically detectable heart disease, we find it unlikely. Conclusion ========== This is the first study to investigate genetic determinants of MPD changes in women randomized to ET, EPT or placebo. Although plausible, it is still unknown whether women with the greatest increase in MPD in response to EPT are at higher risk for breast cancer associated with EPT use than other women. Much research in this area remains to be done, but our findings from this pilot study suggest that the magnitude of the increase in MPD might be greater in women with a genetically determined lower activity of some enzymes that metabolize EPT. Abbreviations ============= AKR1C4 = aldo-keto reductase 1C4; BMI = body mass index; COMT = catechol-O-methyltransferase; CYP1B1 = cytochrome P450 1B1; EPAT = Estrogen in the Prevention of Atherosclerosis Trial; EPT = estrogen and progestin therapy; ET = estrogen therapy; HT = hormone therapy; MPD = mammographic percentage density; UGT1A1 = UDP-glucuronosyltransferase 1A1; WELL-HART = Women\'s Estrogen--Progestin Lipid-Lowering Hormone Atherosclerosis Regression Trial. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= SJL participated in the study design, data collection, conducted the statistical analyses and drafted the manuscript. WJM and HNH designed the original clinical trials and participated in the design of the study. DVB and WW performed the genetic analyses. SAI and CAH assisted in setting up the genetic analyses. YRP participated in study design and was responsible for obtaining the mammograms of study participants. MCP participated in the study design and drafting of the manuscript. GU 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 the research nurses of the Atherosclerosis Research Unit, Keck School of Medicine, University of Southern California, Los Angeles, for their dedicated efforts in contacting eligible subjects for recruitment to this study. This work was supported by a University of Southern California/Norris Comprehensive Cancer Center Support Grant P30 CA14089. EPAT was supported by a grant (R01-AG-18798) from the National Institute on Aging. WELL-HART was supported by a grant (U01-HL-49298) from the National Heart, Lung, and Blood Institute and by the Office of Research on Minority Health. Mead Johnson Laboratories supplied the micronized 17β-estradiol and placebo pills. Pharmacia & Upjohn Company supplied the medroxyprogesterone acetate and matching placebo pills. Figures and Tables ================== ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Baseline characteristics by study and treatment group in women from the EPAT and WELL-HART trials ::: Characteristic EPAT (n= 127) WELL-HART (n= 105) ----------------------------------------------------------------- ----------------- ------------ ---------------------- ------------ ------------ ------------ --------- -------- Age at baseline mammogram, years (mean ± SEM) 62.3 ± 1.0 60.2 ± 0.8 0.10 64.7 ± 1.0 61.8 ± 1.2 64.1 ± 1.1 0.15 0.005 Years since menopause at baseline (mean ± SEM) 15.1 ± 1.4 12.5 ± 1.0 0.12 18.6 ± 1.6 15.9 ± 1.4 19.5 ± 1.9 0.28 0.001 Number of deliveries (live and stillbirths) (mean ± SEM) 2.4 ± 0.2 2.3 ± 0.2 0.75 4.5 ± 0.6 3.6 ± 0.4 4.1 ± 0.5 0.47 0.0001 Race, n(%) White, non-latina 34 (59.7) 40 (57.1) 12 (32.4) 7 (20.6) 14 (41.2) Black, non-latina 6 (10.5) 10 (14.3) 4 (10.8) 10 (29.4) 5 (14.7) Latina 10 (17.5) 13 (18.6) 20 (54.1) 9 (26.5) 11 (32.4) Asian or Pacific Islander 7 (12.3) 7 (10.0) 0.91 1 (2.7) 8 (23.5) 4 (11.8) 0.01^‡^ 0.0005 Education, n(%) High school graduate or less 8 (14.0) 9 (12.9) 20 (54.1) 15 (44.1) 15 (44.1) Trade or business school/some college 29 (50.9) 33 (48.6) 11 (29.7) 12 (35.3) 13 (38.2) Bachelor\'s degree or more 20 (35.1) 27 (38.6) 0.92 6 (16.2) 7 (20.6) 6 (17.7) 0.90 0.0001 Family history of breast cancer (first-degree relative), n(%) 9 (15.8) 6 (8.6) 0.21 2 (5.4) 7 (20.6) 3 (8.8) 0.14^‡^ 0.91 Age at menarche, years (mean ± SEM) 12.7 ± 0.2 12.6 ± 0.2 0.86 12.7 ± 0.3 12.8 ± 0.3 13.3 ± 0.3 0.32 0.21 Ever used postmenopausal hormones, n(%) 36 (63.2) 42 (60.0) 0.72 13 (35.1) 15 (44.1) 21 (61.8) 0.08 0.02 BMI, kg/m^2^(mean ± SEM) 28.7 ± 0.7 28.7 ± 0.7 0.96 30.8 ± 1.0 31.4 ± 1.0 30.2 ± 1.1 0.73 0.007 Mammographic density at baseline, % (mean ± SEM) 17.8 ± 2.3 22.3 ± 1.9 0.13 7.9 ± 1.8 10.8 ± 2.1 13.3 ± 2.3 0.19 0.0001 \Comparison of characteristics by treatment group, χ^2^test for comparison of categorical variables, analysis of variance for comparison of means; ^†^comparison of characteristics between trials (EPAT versus WELL-HART), χ^2^test for comparison of categorical variables, analysis of variance for comparison of means; ^‡^Fisher\'s exact test. EPAT, Estrogen in the Prevention of Atherosclerosis Trial; ET, 1 mg/day micronized 17β-estradiol; EPT, 1 mg/day micronized 17β-estradiol with 5 mg/day medroxyprogesterone acetate for days 19 to 30 each month; WELL-HART, Women\'s Estrogen--Progestin Lipid-Lowering Hormone Atherosclerosis Regression Trial. ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Change in mammographic percentage density by treatment assignment ::: Treatment Change in mammographic density (%) --------------------------------------------------- ------------------------------------ -------- -------------- -------- EPAT Placebo (n= 57) \- 0.7 ± 0.8 Ref. \- 0.9 ± 0.9 Ref. ET (n= 70) 3.4 ± 0.7 0.003 4.0 ± 0.8 0.0001 WELL-HART Placebo (n= 37) 0.8 ± 1.4 Ref. 0.06 ± 1.7 Ref. ET (n= 34) 4.8 ± 1.5 0.06 5.6 ± 1.5 0.02 EPT (n= 34) 7.8 ± 1.5 0.001 7.8 ± 1.6 0.0005 Phomogeneity of treatment effect between trials 0.76 All women Placebo (n= 95) \- 0.02 ± 0.8 Ref. 0.02 ± 0.8 Ref. ET (n= 104) 3.9 ± 0.7 0.0003 4.6 ± 0.8 0.0001 EPT (n= 34) 7.8 ± 1.3 0.0001 7.2 ± 1.5 0.0001 \Adjusted for mammographic percentage density at baseline, race, age at baseline, years since menopause, past use of hormone therapy, body mass index (BMI) at baseline, change in BMI on trial and study group. EPAT, Estrogen in the Prevention of Atherosclerosis Trial; EPT, estrogen and progestin therapy; ET, estrogen therapy; WELL-HART, Women\'s Estrogen--Progestin Lipid-Lowering Hormone Atherosclerosis Regression Trial. ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Least-square mean change in mammographic percentage density by genotype and treatment arm, adjusted ::: Genotype Placebo ET EPT ---------------------- -------------- ----- ------ ---------- ----- ------ ----------- ----- -------- -------- COMTVal158Met Val/Val \- 0.1 (30) 1.4 Ref. 3.7 (30) 1.4 Ref. 8.0 (12) 2.3 Ref. Val/Met 0.5 (45) 1.2 0.74 4.3 (48) 1.1 0.72 6.2 (19) 1.9 0.54 Met/Met \- 0.6 (19) 1.8 0.82 6.0 (25) 1.6 0.27 9.8 (3) 4.5 0.72 0.77 CYP1B1Val432Leu Val/Val \- 0.7 (15) 1.9 Ref. 4.7 (23) 1.6 Ref. 6.9 (7) 2.9 Ref. Leu/Val \- 0.01 (42) 1.2 0.75 5.8 (44) 1.2 0.57 1.3 (13) 2.1 0.11 Leu/Leu 0.9 (36) 1.2 0.47 3.6 (37) 1.2 0.61 14.5 (13) 2.1 0.03 0.0004 UGT1A1 \<7/\<7 TA repeats 0.3 (37) 1.3 Ref. 5.7 (50) 1.1 Ref. 7.1 (18) 1.9 Ref. \<7/≥ 7 TA repeats \- 0.5 (43) 1.2 0.60 3.3 (39) 1.3 0.15 8.7 (12) 2.3 0.57 ≥ 7/≥ 7 TA repeats 0.6 (14) 2.1 0.92 4.4 (14) 2.1 0.58 5.4(3) 4.6 0.74 0.77 AKR1C4L311V Leu/Leu \- 0.3 (72) 0.9 Ref. 4.3 (79) 0.8 Ref. 4.0 (25) 1.6 Ref. Leu/Val 1.5 (21) 1.6 0.30 5.5 (25) 1.5 0.47 16.7 (7) 2.8 0.0001 Val/Val 0.6 (1) 0.90 (0) 29.3 (1) 0.0007 0.001 Figures are adjusted for race, age at baseline, years since menopause, body mass index (BMI) at baseline, change in BMI on trial, mammographic percentage density at baseline, past use of hormone therapy and study group. \Analysis of covariance Pfor comparison of means; ^†^analysis of covariance Pfor ET/EPT × genotype interaction. AKR1C4, aldo-keto reductase 1C4; COMT, catechol-O-methyltransferase; CYP1B1, cytochrome P450 1B1; EPT, estrogen and progestin therapy; ET, estrogen therapy; UGT1A1*, UDP-glucuronosyltransferase 1A1. :::	PubMed Central	2024-06-05T03:55:57.468642	2005-2-23	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143576/", "journal": "Breast Cancer Res. 2005 Feb 23; 7(3):R336-R344", "authors": [ { "first": "Sarah J", "last": "Lord" }, { "first": "Wendy J", "last": "Mack" }, { "first": "David", "last": "Van Den Berg" }, { "first": "Malcolm C", "last": "Pike" }, { "first": "Sue A", "last": "Ingles" }, { "first": "Christopher A", "last": "Haiman" }, { "first": "Wei", "last": "Wang" }, { "first": "Yuri R", "last": "Parisky" }, { "first": "Howard N", "last": "Hodis" }, { "first": "Giske", "last": "Ursin" } ] }
PMC1143774	Introduction ============ Behavioral and Brain Functions(BBF) is an Open Access, peer-reviewed, online journal considering original research, review, and modeling articles in all aspects of neurobiology or behavior, favoring research that relates to both domains. BBFis published by BioMed Central. The greatest challenge for empirical science is to understand human behavior, how human behavior arises from the myriad functions such as attention, language, memory and emotion, how these functions are reflected in the human brain, and how brain functions and behavior are altered in disease. Behavioral and cognitive neuroscience investigates the psychological, computational, and neuroscientific bases of normal and abnormal behavior. It is a field that receives a lot of attention through the Brain Awareness Weekin March every year. The \"Decade of the Brain\" (1990--2000) was also important for promoting the field. The interdisciplinary nature of the field covers developments in human and animal behavioral science, neuroscience, neuropsychology, cognitive psychology, neurobiology, linguistics, computer science, and philosophy. Behavioral and Brain Functionsis the first Open Access journal for basic research covering the entire area of behavioral and cognitive neuroscience -- an area where animal studies traditionally play a prominent role. Behavioral and Brain Functionsis published online, allowing unlimited space for figures, extensive datasets to allow readers to study the data for themselves, and moving pictures, which are important qualities assisting communication in modern science. Open Access =========== Behavioral and Brain Functions\' Open Access policy changes the way in which articles are published. First, all articles become freely and universally accessible online, and so an author\'s work can be read by anyone at no cost. Second, the authors hold copyright for their work and grant anyone the right to reproduce and disseminate the article, provided that it is correctly cited and no errors are introduced \[[@B1]\]. Third, a copy of the full text of each Open Access article is permanently archived in an online repository separate from the journal. Behavioral and Brain Functions\' articles are archived in PubMed Central \[[@B2]\], the US National Library of Medicine\'s full-text repository of life science literature, and also in repositories at the University of Potsdam \[[@B3]\] in Germany, at INIST \[[@B4]\] in France and in e-Depot \[[@B5]\], the National Library of the Netherlands\' digital archive of all electronic publications. Open Access has four broad benefits for science and the general public. First, authors are assured that their work is disseminated to the widest possible audience, given that there are no barriers to access their work. This is accentuated by the authors being free to reproduce and distribute their work, for example by placing it on their institution\'s website. It has been suggested that free online articles are more highly cited because of their easier availability \[[@B6]\]. Second, the information available to researchers will not be limited by their library\'s budget, and the widespread availability of articles will enhance literature searching \[[@B7]\]. Third, the results of publicly funded research will be accessible to all taxpayers and not just those with access to a library with a subscription. As such, Open Access could help to increase public interest in, and support of, research. Public accessibility may become a legal requirement \[[@B8]\]. Fourth, a country\'s economy will not influence its scientists\' ability to access articles because resource-poor countries (and institutions) will be able to read the same material as wealthier ones (although creating access to the internet is another matter \[[@B9]\]). Editorial Board =============== The editorial board covers a broad cross-section of brain and behavior research. Several members of the BBFEditorial Board are actively involved in promoting research and assisting researchers in poor and developing countries through The International Brain Research Organization(IBRO). Peer review policy ================== Publication of research articles in the broad area of behavioral and cognitive neuroscience is dependent on: relevance to the aims of the journal; scientific validity and excellence; and coherence with the research area as a whole, as judged by two independent reviewers. An Editorial Board oversees the peer review of manuscripts submitted to Behavioral and Brain Functions. Peer review is anonymous. A recommendation of acceptance or rejection is at the discretion of the Editors based on the reviewers\' recommendations. Articles will be published immediately upon acceptance and soon after they will be listed in PubMed and archived in PubMed Central as well as other national archives. Competing interests =================== The Editor-in-Chief and the Editorial Board of Behavioral and Brain Functionshave no financial incentive to accept manuscripts as they are not paid on the basis of the number of manuscripts accepted. In fact, they do not receive any financial remuneration for their involvement with the journal. We insist that decisions about a manuscript are based on the quality of the work, not on whether the authors\' can pay the article-processing charge. Authors will be asked to declare any competing interests that they have.	PubMed Central	2024-06-05T03:55:57.471802	2005-4-22	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143774/", "journal": "Behav Brain Funct. 2005 Apr 22; 1:1", "authors": [ { "first": "Terje", "last": "Sagvolden" } ] }
PMC1143775	Background ========== Methylphenidate (MPH) is a leading treatment for Attention Deficit Hyperactivity Disorder (ADHD). Although this compound has been used for decades, the neural mechanisms underlying MPH\'s therapeutic actions are still unknown. Recent advances in our understanding of the neurobiology of ADHD, and the identification of appropriate MPH doses for use in rodents, now allow the examination of therapeutic actions in animals. Converging evidence has demonstrated that ADHD symptoms arise from dysregulation of prefrontal cortical (PFC)/striatal and cerebellar circuits (reviewed in \[[@B1]\]. The PFC uses working memory to guide behavior and attention, inhibiting inappropriate responses and sustaining attention over long delays, particularly under conditions of interference from distractors \[[@B2],[@B3]\]. Deficits in PFC function lead to poor impulse control, distractibility, hyperactivity, forgetfulness and poor organization and planning \[[@B4]\]. There is general agreement that ADHD involves weakened PFC function e.g. \[[@B5]\], and speculation that medications might strengthen PFC abilities. Imaging studies have shown that MPH produces more efficient PFC function in both ADHD patients \[[@B6]\] and control subjects \[[@B7]\], consistent with this view. Many researchers have assumed that MPH acts by blocking dopamine (DA) transporters (reviewed in \[[@B8]\]). Indeed, elegant PET imaging studies of DA transporter occupancy in striatum have shown that MPH acts at this site \[[@B9]\]. However, the striatum contains very few noradrenergic (NE) transporters, and thus the important actions of MPH on the NE system have received far less attention. At the present time, imaging studies are unable to reliably visualize the low levels of NE and DA actions in cortex, although there is the suggestion that there may be fewer catecholamine terminals in the PFC of adults with ADHD \[[@B10]\]. Thus, animal studies are of particular importance for understanding MPH actions in PFC. Recent animal studies by Kuzcenski and Segal \[[@B11]\] have identified the low, oral doses of MPH which 1) produce plasma levels in rats similar to those observed in children taking MPH, and 2) decrease locomotor activity in rats just as they do in humans. Oral administration was key, as MPH administration by injection produces much higher blood and brain MPH levels \[[@B12]\]. Prior to appreciation of this research, MPH doses in rat studies were generally too high, and were usually administered by injection, producing kinetics and drug levels relevant to drug abuse but not to ADHD e.g. \[[@B13]-[@B15]\]. These injected, effectively higher doses produce locomotor [hyper]{.underline}activity with stimulant treatment e.g. \[[@B16],[@B17]\], and it has been assumed that there were species differences that would impede research. Thus, the identification of the appropriate dose regimen for MPH treatment in rats opens a new field of research that may more quickly elucidate MPH therapeutic mechanisms. Although previous research focused on MPH amplification of DA actions, more recent biochemical studies using low doses of MPH show more potent effects on hippocampal NE than on striatal DA \[[@B18]\], while increasing both DA and NE release in the PFC (\[[@B19]\] and C.W. Berridge, personal communication). Both NE and DA have a critical influence on PFC cognitive functioning. NE improves working memory, response inhibition and lessens distractibility through actions at post-synaptic α2A adrenoceptors in the PFC, while DA improves working memory through modest stimulation of D1 receptors in PFC (reviewed in \[[@B20],[@B21]\]). Although optimal levels of NE and DA are essential to proper PFC function, very high levels of NE and DA release, e.g. during stress, impair PFC function through α1, beta-1, D1 and possibly D4 receptors \[[@B22]\]. The current study characterized the effects of low, oral doses of MPH on PFC function in rats. Rats were tested on a working memory task, spatial delayed alternation, a classic test of PFC function in rodents \[[@B23]\]. MPH was found to have effects similar to those observed in patients: improving performance at moderate doses but producing perseverative errors at high doses. The second part of the study examined whether NE α2A adrenoceptor and/or DA D1 receptor actions contributed to the enhancing effects of MPH on PFC function. Results ======= MPH dose/response: Effects on delayed alternation performance ------------------------------------------------------------- The effects of an acute, oral dose of MPH were examined over the dose range found to produce drug plasma levels in rats similar to clinical use in ADHD (0.5, 1.0, 1.5, 2.0 and 3.0 mg/kg, oral administration 30 min before testing). MPH produced an inverted U dose response curve whereby the middle doses (1.0, 1.5, 2.0 mg/kg) generally improved performance, while higher doses (2.0, 3.0 mg/kg) often impaired performance. Representative dose/response curves are shown in Figure [1A](#F1){ref-type="fig"}. There were individual differences in MPH dose sensitivity that may result from differences in MPH absorption from the gastrointestinal tract, and/or variations in endogenous catecholamine levels in PFC circuits. For all animals, a dose was found between 1.0--2.0 mg/kg that significantly improved delayed alternation performance (Figure [1B](#F1){ref-type="fig"}; vehicle vs. MPH p = 0.002, df = 7). No change in response time was noted (mean ± SEM vehicle: 191.2 ± 39.1 sec; mean ± SEM MPH: 179.3 ± 30.3 sec; p \> 0.7, df = 7). These enhancing doses were used in subsequent experiments to examine the receptor actions contributing to MPH therapeutic actions. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### The effects of oral administration of methylphenidate (MPH) on delayed alternation performance in male rats. A.Representative dose/response curves from two individual rats. Results represent percent correct on the delayed alternation task following MPH administration. For most rats, a lower dose (1.0--2.0 mg/kg, p.o. 30 min) was found to improve performance, while higher doses often impaired performance (1.5--3.0 mg/kg). Rats showed individual differences in dose sensitivity. B.An optimal dose of MPH was found for all rats which significantly improved delayed alternation performance. Results represent mean ± S.E.M. percent correct on the delayed alternation task. VEH = cracker vehicle; MPH = optimal dose of methylphenidate (1.0--2.0 mg/kg); \\ significantly different from VEH p = 0.002. C.Higher doses of MPH impaired performance and produced a perseverative pattern of errors. Perseveration was assessed by the greatest number of consecutive entries into a single arm of the T maze. Results represent mean ± S.E.M. number of consecutive entries. VEH = cracker vehicle; MPH = impairing dose of methylphenidate (1.5--3.0 mg/kg); \* significantly different from VEH p = 0.046. ::: ![](1744-9081-1-2-1) ::: Six of the eight rats tested showed impairment in delayed alternation performance as the dose was raised (1.5--3 mg/kg, especially following 2--3 mg/kg). The number of perseverative errors significantly increased at these higher doses, as measured by the greatest number of consecutive entries into a single arm (Figure [1C](#F1){ref-type="fig"}; p \< 0.05, df = 5). A perseverative response pattern is consistent with PFC dysfunction. No consistent change in response time was observed following higher, impairing doses of MPH, although some animals were faster and some slower than usual (mean ± SEM vehicle: 201.4 ± 46.2 sec; mean ± SEM MPH: 194.2 ± 107.2 sec; range vehicle: 91--398 sec; range MPH: 50--675 sec). No stereotyped behaviors, common at much higher doses, were observed in these animals. Thus, cognitive choices, but not behavior per se, showed a perseverative profile. The role of α2 adrenoceptor mechanisms in the cognitive enhancing effects of MPH -------------------------------------------------------------------------------- The α2 adrenoceptor antagonist, idazoxan, was co-administered with MPH to test whether MPH enhances performance by facilitating endogenous NE stimulation of α2 adrenoceptors (n = 5). The optimal dose of MPH was selected for each animal; a dose of idazoxan was selected (0.1 mg/kg) that had no effects on its own. As shown in Figure [2](#F2){ref-type="fig"}, idazoxan significantly reversed the enhancing effects of MPH. Two way analysis of variance with repeated measures (2-ANOVA-R) showed a significant effect of MPH (F(1,4) = 18.45, p = 0.01), a trend toward significant effect of idazoxan (F(1,4) = 6.63, p = 0.06), and a significant interaction between the two drugs (F(1,4) = 27.2, p = 0.006). User defined contrasts showed that MPH+vehicle significantly improved performance compared to vehicle+vehicle (F(1,4) = 53.3, p = 0.002), while idazoxan+vehicle was not significantly different than vehicle+vehicle (F(1,4) = 0.02 p = 0.90). Most importantly, animals performed significantly lower on the delayed alternation task following MPH+idazoxan treatment than when they were administered MPH+vehicle (F(1,4) = 43.4, p = 0.0028), and were not significantly different in their performance from days in which they were administered vehicle+vehicle (F(1,4) = 0.005 p = 0.95). These results are consistent with α2 adrenoceptor actions contributing to the enhancing effects of oral MPH. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### The enhancing effects of methylphenidate were blocked by co-administration of the α2 adrenoceptor antagonist, idazoxan at a dose which had no effect on its own. Results represent mean ± S.E.M. percent correct on the delayed alternation task. VEH = cracker vehicle; MPH = optimal dose of methylphenidate (1.0--2.0 mg/kg); IDA = idazoxan (0.1 mg/kg); \\ significantly different from VEH, p = 0.002; † significantly different from MPH, p = 0.003. ::: ![](1744-9081-1-2-2) ::: The role of DA D1 receptor mechanisms in the cognitive enhancing effects of MPH ------------------------------------------------------------------------------- The DA D1 receptor antagonist, SCH23390, was co-administered with MPH to test whether MPH enhances performance by facilitating endogenous DA stimulation of D1 receptors (n = 7). A dose of 0.1 mg/kg SCH23390 was used, unless this dose produced impairment on its own. In these cases, the dose of SCH23390 was lowered to 0.01 mg/kg (n = 3). SCH23390 significantly reduced the enhancing effects of MPH on delayed alternation performance (Figure [3](#F3){ref-type="fig"}). 2-ANOVA-R showed a significant effect of MPH, a significant effect of SCH23390, and a significant interaction between the two drugs (effect of MPH: F(1,6) = 11.59, p = 0.014; effect of SCH23390: F(1,6) = 9.3, p = 0.023; interaction between MPH and SCH23390: F(1,6) = 9.3, p = 0.023). User defined contrasts showed that MPH+vehicle significantly improved performance compared to vehicle+vehicle (F(1,6) = 61.45, p = 0.0002), while SCH23390+vehicle was not significantly different than vehicle+vehicle (F(1,6) = 0.0, p = 1.0). Animals performed significantly worse on the delayed alternation task following MPH+SCH23390 treatment than when they were administered MPH+vehicle (F(1,6) = 15.0, p = 0.008). Although performance remained a bit above vehicle levels of response, performance following MPH+SCH23390 treatment was not significantly different than following vehicle+vehicle (F(1,6) = 0.84, p = 0.4). These results are consistent with D1 receptor actions contributing to the enhancing effects of oral MPH. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### The enhancing effects of methylphenidate were blocked by co-administration of the dopamine D1 receptor antagonist, SCH23390 at doses which had no effect on their own. Results represent mean ± S.E.M. percent correct on the delayed alternation task. VEH = cracker vehicle; MPH = optimal dose of methylphenidate (1.0--3.0 mg/kg); SCH = SCH23390 (0.01 or 0.1 mg/kg); \\ significantly different from VEH p = 0.0002; † significantly different from MPH, p = 0.008. ::: ![](1744-9081-1-2-3) ::: Discussion ========== The current study provides the first evidence that oral dosing of therapeutically relevant levels of MPH improves PFC cognitive function in rats. This same dose regimen has been found to decrease locomotor activity in rats, reinforcing the idea that rats can be used as an appropriate animal model for examining medications used to treat ADHD patients. Performance of the spatial delayed alternation task in a T maze is very relevant to many aspects of ADHD. Optimal performance of this task requires spatial working memory (remembering which side was most recently entered), response inhibition (inhibiting the tendency to return to the location where the animal was last rewarded) and the ability to sustain attention and suppress the distractions of being put into the start box. Thus, the delayed alternation task assesses many of the PFC operations that are known to be problematic in ADHD. The current study found that moderate doses of MPH produced a highly significant improvement in delayed alternation performance. This improvement likely reflects enhanced PFC cognitive performance, as there were no changes in response time characteristic of motor or motivational changes. Indeed, given that MPH [reduces]{.underline} eating, it is unlikely that simple changes in motivation for food reward could account for the improvement in performance. It is also noteworthy that higher doses of MPH impaired delayed alternation performance in a large subset of animals. This impairment manifested in a perseverative profile of errors in which the rats continued to choose the same incorrect arm of the maze. Perseveration is also seen with 1) PFC lesions \[[@B23]-[@B26]\], 2) infusion of a high dose of a DA D1 or NE α1 agonist into the PFC \[[@B27],[@B28]\], or 3) stress exposure, which causes high levels of NE and DA release in PFC \[[@B29]\]. Future studies will be needed to determine whether higher doses of MPH impair delayed alternation performance due to excessive catecholamine release in PFC, and if so, which receptor(s) underlie these impairing actions. Comparison to cognitive effects of MPH in humans ------------------------------------------------ The MPH profile observed in rats in the current study is very similar to that in seen in humans. Oral administration of clinical doses of MPH have been found to improve spatial working memory, response inhibition, set-shifting and other PFC cognitive functions in both \"normal\" college students \[[@B30]\] and in children and adults with ADHD \[[@B31]-[@B35]\]. Imaging studies have shown more efficient dorsolateral PFC activity (BOLD) following MPH doses that improve spatial working memory, consistent with improved PFC cognitive function \[[@B36]\]. Interestingly, in adults with ADHD, childhood ratings of ADHD (both self-reported and informant ratings) correlated with response to methylphenidate on the spatial working memory task \[[@B35]\]. Thus, studies of spatial working memory performance are likely very relevant to the therapeutic effects of ADHD medications. In the current study, higher doses of MPH impaired spatial working memory performance in a large number of animals. These findings are consistent with the original Lyon-Robbins analysis of stimulant actions in rodents that found increasingly perseverative responses with increasing dose of stimulant administration \[[@B37]\]. It is noteworthy that these \"higher\" doses are still much lower than those used in most other rodent studies, accentuating the fact that previous research in animals has often focused on MPH doses that are too high. Similar to our findings in rats, clinicians are often concerned that higher doses of MPH can induce perseverative thinking in patients e.g. \[[@B37],[@B38]\]. For example, higher doses of MPH (e.g. 1.0 mg/kg) can increase perseverative errors on the Wisconsin Card Sorting Task when the test is novel \[[@B39],[@B40]\]. Perseverative errors were not increased, and indeed were reduced by MPH, when the Wisconsin Card Sorting Task was given repeatedly \[[@B41]\]; however, this condition minimizes the need for flexible thinking, as the switching rule is \"discovered\" only during the first experience of the task. Douglas et al. also found no evidence of perseveration on other tasks, such as the Trails B, and concluded that doses below 0.9 mg/kg produced dose-related improvements in cognitive flexibility. As repeated daily doses can add together, they cautioned that doses above 0.6 mg/kg were not recommended. Thus, under optimal dosage conditions, MPH appears to improve flexible thinking in patients, but higher doses may produce a perseverative profile similar to that seen in rodents. Receptor mechanisms underlying PFC cognitive enhancing effects of MPH --------------------------------------------------------------------- The identification of an MPH dose regimen that improves PFC cognitive performance in rats provides the opportunity to examine the neural mechanisms underlying MPH therapeutic actions. The current study began by examining the role of NE α2 and DA D1 receptors, given the importance of these receptors to PFC cognitive function. The study found that both the α2 antagonist, idazoxan, and the D1 antagonist, SCH23390, reversed the cognitive-enhancing effects of MPH. Care was taken to use antagonist doses that did not impair performance on their own; thus, additive effects of drug treatment cannot account for the normalization of response. Rather, the data are consistent with MPH improving performance by increasing the availability of NE and DA, which in turn stimulate α2 and D1 receptors. It is interesting that either idazoxan or SCH23390 was fully effective in reversing the MPH response. These data suggest that there may be beneficial interactions between these receptors, an area that has received little investigation. Future studies will be needed to determine whether the enhancing effects of MPH occur in the PFC and/or elsewhere in the brain. Low, systemic doses of MPH and amphetamine are known to increase both NE and DA levels in the rat PFC while having more subtle effects in striatal regions (\[[@B19]\], and C.W. Berridge, personal communication). There are relatively low levels of DA transporters in the PFC \[[@B42]\]; thus the increase in both DA and NE levels likely occurs through blockade of NE transporters, which are thought to transport both NE and DA in the PFC \[[@B42]\]. The efficacy of idazoxan and SCH23390 in reversing the MPH response in the current study is not unexpected, given the importance of α2 and D1 receptor actions to PFC cognitive function. Catecholamine depletion in PFC produces working memory deficits as severe as ablation of the tissue itself \[[@B43]\], and D1 receptor blockade similarly weakens working memory regulation of behavior \[[@B44],[@B45]\]. In the current study, D1 blockade did not appear as completely effective as α2 receptor blockade, although both antagonists weakened the improvement such that it was statistically insignificant from vehicle. The potentially weaker effects with SCH23390 may be due to the lower dose used in some animals, and the difficulties in dealing with an inverted U dose response, where either too little or too much D1 receptor stimulation can impair performance. Under these conditions it is difficult to identify the correct dose of antagonist to perfectly normalize behavior. Individual variability in response to SCH23390 may arise from differences in endogenous D1 receptor stimulation under basal conditions. For example, D1 agonists have been shown to enhance attentional control in rats when infused into the PFC, but only in animals that were performing relatively poorly under basal conditions \[[@B46]\]. D1 receptors also play a key role in striatal function, and it is possible that these actions outside the PFC also contributed to the enhancing effects of MPH on the delayed alternation task. With some notable exceptions \[[@B47],[@B48]\], much of the ADHD field has focused on DA mechanisms in ADHD \[[@B49]\]. In turn, there has been intensive focus on the DA actions of MPH, with some researchers even referring to MPH as a selective DA transporter blocker. The current data, in addition to recent biochemical studies \[[@B11],[@B18],[@B50]\] caution that the NE actions of stimulants such as MPH are just as important as the DA effects. This point is accentuated by the findings that one can recreate the symptoms of ADHD- increased locomotor activity, poor impulse control and weakened working memory/distractibility- by blocking α2 adrenoceptors with yohimbine infusions in the monkey PFC \[[@B51]-[@B53]\], respectively. Yohimbine also reduces the delay-related activity of PFC neurons, the cellular measure of working memory and response inhibition \[[@B54]\]. Conversely, the α2 agonist, guanfacine, has been shown to strengthen working memory \[[@B55],[@B56]\], reduce distractibility \[[@B57]\], improve response inhibition \[[@B58]-[@B60]\], and increase regional cerebral blood flow in monkey PFC \[[@B61]\]. Most recently, guanfacine has been shown to reduce locomotor hyperactivity and improve attentional control in the spontaneously hypertensive rat, a rodent model of ADHD (T. Sagvolden, personal communication). Experiments are in progress to determine whether MPH loses efficacy in mice with a functional knockout of the α2A adrenoceptor. The current results with idazoxan indicate that at least some of the beneficial effects of MPH arise from NE stimulation of α2 adrenoceptors. Relevance to medications used to treat ADHD ------------------------------------------- Many of the medications used to treat ADHD increase endogenous NE stimulation of α2 adrenoceptors or mimic NE by directly stimulating these receptors. For example, like MPH, atomoxetine (Strattera) and amphetamine (Adderall) increase NE as well as DA in the PFC of rats \[[@B19],[@B42]\]. NE reuptake blockers have been shown to be very efficacious in treating ADHD symptoms, although their cardiac side effects have limited clinical utility in children \[[@B47],[@B62]\]. Future studies will need to examine whether low, oral doses of amphetamine and atomoxetine, like MPH, can improve delayed alternation performance in rats. Guanfacine mimics NE at α2 adrenoceptors, and is now in common use for the treatment of ADHD, especially in patients with tic disorders or drug abuse liability who cannot take stimulant medications \[[@B63]\]. Guanfacine has been shown to improve spatial working memory performance in mice \[[@B64]\], rats \[[@B65]\], monkeys \[[@B55],[@B66]\] and humans \[[@B67]\]. Thus, there is an excellent correspondance between drug effects in the laboratory and clinical efficacy in ADHD. The identification of an appropriate MPH dose regimen for use in rats should help in the development of safer and more effective medications for the treatment of ADHD. Conclusion ========== Low, orally-administered doses of MPH improve spatial working memory performance in rats, while higher doses often impair performance and induce perseverative errors. Both NE α2 adrenoceptor and DA D1 receptor stimulation contribute to the enhancing effects of MPH on working memory in rodents. Future studies with low, oral doses of MPH may continue to elucidate the mechanisms underlying the therapeutic actions of MPH in treating ADHD. Methods ======= Animals ------- Young adult (240--260 g) male rats were purchased from Taconic (Germantown, NY) and singly-housed in filter frame cages. Animals were kept on a 12 hr light/dark cycle, and experiments were conducted during the light phase. Rats were slowly habituated to a restricted diet (16 gm/day per rat) of autoclaved Purina (St. Louis, MO) rat chow during the first two weeks. Food was given immediately after behavioral testing and water was available ad libitum. Rats were weighed weekly to confirm normal weight gain. Food rewards during cognitive testing were highly palatable miniature chocolate chips. Rats were assigned a single experimenter who handled them extensively before behavioral testing. Cognitive assessment -------------------- Rats were habituated to a T-maze (dimensions, 90 × 65 cm) until they were readily eating chocolate chips placed at the end of each arm and were acclimated to handling. After habituation, rats were trained on the delayed alternation task. On the first trial, animals were rewarded for entering either arm. Thereafter, for a total of 10 trials per session, rats were rewarded only if they entered the maze arm that was not previously chosen. Between trials the choice point was wiped with alcohol to remove any olfactory clues. The delay between trials started at \"0\" sec (i.e. about 1.5 sec, minimum possible for delayed alternation) and was subsequently raised in 5 sec intervals as needed to maintain performance at about 70% correct. Animals were scored for accuracy (arm chosen) and response time for each trial. Drug administration ------------------- The experimenter testing the animal was unaware of drug treatment conditions. Given the need for oral administration of drug, rats were habituated to eating a small piece of cracker. All animals had learned to ingest the cracker rapidly and completely prior to the initiation of drug testing. MPH was acquired from the National Institute of Drug Abuse. Doses were based on those identified by Kuzcenski and Segal \[[@B11]\]. MPH was dissolved in tap water and injected onto a small piece of Saltine cracker that was fed to the rat 30 min before cognitive testing. The doses examined were: 0 (water only), 0.5, 1.0, 1.5, 2.0, and 3.0 mg/kg). For example, the 1.0 mg/kg dose was made by dissolving 1 mg MPH in1 ml water and injecting a volume equivalent to the weight of the rat, e.g. a 450 g rat would receive 0.45 ml injected onto the cracker. Animals rapidly ate the cracker once habituated to the procedure. Doses were administered in random order with the exception that no animal began with the 3.0 mg/kg dosage. Idazoxan was purchased from Sigma (St. Louis, MO) and administereed at a dose of 0.1 mg/kg. Idazoxan was dissolved in saline, and like MPH, was injected into the cracker for oral administration. SCH23390 also was purchased from Sigma and dissolved in saline. SCH23390 was initially administered in a dose of 0.1 mg/kg. Animals who were impaired by this dose were subsequently administered 0.01 mg/kg SCH23390 so that additive effects could not account for MPH reversal. Data analysis ------------- The dependent variables were percent correct on the delayed alternation task (accuracy), response time, and greatest number of consecutive entries into an incorrect arm (perseveration score). Statistical comparisons utilized within subjects designs; simple comparisons utilized paired (dependent) T tests. The effects of idazoxan or SCH23390 on the MPH response were analyzed with a two-way analysis of variance with repeated measures with factors of 1) MPH and 2) antagonist, and user defined contrasts to test pairwise comparisons. Authors\' contributions ======================= A. Arnsten designed the study, analyzed data, drew the figures and wrote the paper. A. Dudley, in collaboration with the laboratory technicians, made up drug, tested the animals on the spatial working memory task, and helped with data analysis and the writing of this paper. Acknowledgements ================ We would like to thank Tracy White Sadlon, Lisa Ciavarella, Sam Johnson and Jessica Thomas for their invaluable technical expertise, and Dr. Xavier Castellanos for his continued interest and support. This research was funded by R21 MH066393 and a research grant from Shire Pharmaceuticals.We also thank NIDA for the gift of methylphenidate	PubMed Central	2024-06-05T03:55:57.472609	2005-4-22	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143775/", "journal": "Behav Brain Funct. 2005 Apr 22; 1:2", "authors": [ { "first": "Amy FT", "last": "Arnsten" }, { "first": "Anne G", "last": "Dudley" } ] }
PMC1143776	Introduction ============ Several learning tasks are available to test spatial orientation abilities in mice. The most widely applied one is probably the water navigation task (also known as the \"Morris maze\"), which was developed originally for rats \[[@B1]\]. However, it has been noted that mice are animals living in dry habitats \[[@B2],[@B3]\], so that a swimming task may be less appropriate to them, because of the stress it may be expected to induce \[[@B4],[@B5]\]. A factor analysis of data from several thousand mice was carried out by Wolfer and Lipp \[[@B6]\], who reported that only the third and least important factor showed loadings of behavioral variables related to spatial orientation, explaining less than 20% of the observed variance in behavior. Indeed, mice with hippocampal lesions still can improve their performances in this task over time \[[@B7]\]. This does not necessarily mean that we should abandon the Morris water navigation task for use with mice, it just means that this test may reveal differences between groups of animals that may relate to different factors and therefore should not automatically be interpreted as differences in spatial learning ability. It would therefore appear that appropriately designed dry-land mazes might assess spatial learning capacities of mice more specifically than water mazes. Among the available mazes, the radial maze appears to be especially suitable \[[@B4]\]. Designing a radial maze ======================= As with many behavioral tasks, the radial maze was originally developed for use with rats. It consists of a central platform, with 4--17 arms radiating outwards. The configuration that is most frequently applied uses eight arms. A food reward may be present at the end of the arms, which food-deprived subjects (maintained at 85--90% of free-feeding body weight) must locate. In our studies, we have therefore attempted to adapt this device to mice, which are generally more anxious and more sensitive to stress than laboratory rats (which have probably been selected more strongly for docility than laboratory mice). To avoid elevation-induced anxiety, the maze we use is placed on the floor of the animal room and arms are enclosed. The arms have to be transparent, to enable animals to see extramaze visual cues, without which it would indeed be very difficult for the subjects to orient themselves in space. The fact that many commercially available radial mazes have metal walls, shows that this necessary condition is perhaps less obvious than might be thought at first sight. We also reduced the dimensions of the maze relative to those habitually used with rats: the central platform measured 22 cm in diameter, arms were 25 cm long, 6 cm wide, and 6 cm high. In addition, at the end of each arm a small compartment, separated from the rest of the arms by a perforated plate, always contained fresh food in order to saturate each arm with food odors, whereas a low barrier prevented animals from seeing the possible food reward hidden behind it. During the tests, animals would therefore have to remember which food rewards had already been eaten, because this procedure made it effectively impossible to smell or see the presence of a food reward in any particular arm while the animal is still standing on the central platform (Fig. [1](#F1){ref-type="fig"}). Using this apparatus, we carried out a number of studies investigating possible covariation between neuroanatomical variation in the mouse hippocampus and performance in different tasks, requiring either spatial or non-spatial modes of navigation. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Radial maze for use with mice. Plexiglas doors can be lowered to limit access to all arms simultaneously. From Crusio, 1999 \[4\], with permission. ::: ![](1744-9081-1-3-1) ::: Heritable neuroanatomical variation in the hippocampus ====================================================== Over a quarter century ago, the pioneering neurogeneticists Richard and Cynthia Wimer \[[@B8],[@B9]\] described large strain differences in the distribution of the hippocampal intra- and infrapyramidal mossy fiber projection (IIPMF; Fig. [2](#F2){ref-type="fig"}). Subsequently, Schwegler and Lipp showed the existence of quantitative differences as well \[[@B10]\]: They measured the surface of the IIPMF in consecutive sections as a measure of the total volume of this projection. Expressed as a percentage of the combined surfaces of the areas CA3 and CA4, the sizes of the IIPMF projections vary between 0.8 and 4.0% \[[@B11]\], that is, about a four- to five-fold variation in size. It should perhaps be noted here that these size variations are not pathological in nature. Between 35 and 53% of the differences between individuals can be attributed to genetic differences between them \[[@B12],[@B13]\]. Efforts are currently underway to identify some of the genes responsible for heritable differences in hippocampal neuroanatomy \[[@B14]\]. The idea that these variations in neuronal connectivity might have functional consequences was rather obvious and, indeed, only a few years after the Wimers\' discovery, Schwegler and Lipp reported a strong correlation between IIPMF sizes and two-way active avoidance learning in mice and rats \[[@B10],[@B15]\]: animals with larger IIPMF projections turned out to be poor learners in a shuttle-box task. The latter task is peculiar in the sense that, perhaps contra-intuitively, brain-damaged animals with lesions to the hippocampal formation perform better than intact animals do \[[@B16]\]. As such lesions generally impair spatial orientation abilities \[[@B16]\], we hypothesized that an opposite correlation would be found in spatial radial maze tasks. It should perhaps be noted here that other neuroanatomical features of the hippocampus also show heritable variations \[[@B12]\], but since no systematic correlations between these measures and behavior have been found, we concentrate here on the IIPMF ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Diagram of a Timm-stained cross-section of the hippocampus. The hippocampal subregion CA3-CA4 (the area of morphometry) is indicated in black, stippled and hatched areas. Black areas: suprapyramidal (SP), intra- and infrapyramidal (IIP) and hilar (CA4) mossy fiber terminal fields originating from the dentate gyrus. Stippled area: strata oriens (OR) and radiatum (RD). Hatched area: stratum lacunosum-moleculare (LM). CA1, subregion of the hippocampus without mossy fibers; FI, fimbria hippocampi; FD, fascia dentata; OL and ML, outer and middle molecular layers of the fascia dentata; SG, supragranular layer; GC, granular cells. ::: ![](1744-9081-1-3-2) ::: Simple tasks with all arms baited ================================= When we started our studies, only little information about radial-maze learning abilities in mice was available and one of the very few studies available reported an inability of mice to master this task \[[@B17]\]. We therefore decided to carry out a pilot experiment \[[@B11]\], using three males from each one of eight different inbred strains, and training them on the simplest task possible. In this first study, opaque PVC arms with clear covers were used, as at the time was being done in almost all maze studies. In order to orient themselves in space, animals would therefore have to look upward. To our own surprise, it turned out that mice learned this task extremely rapidly. Applying one trial per day, animals from the fastest learning strain, C3H/HeJ made in the mean only one error (defined as repeated entrances in a previously visited arm) in the third and last training session \[[@B11]\]. Because only few animals per strain were used in this pilot experiment, standard errors were quite large. Nevertheless, over the eight strains investigated, the numbers of errors committed correlated strongly with the IIPMF sizes (r~S~= -0.88, df = 6, P\< 0.01; see Fig. [3](#F3){ref-type="fig"}). ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Means ± SEM of the sizes of the hippocampal intra- and infrapyramidal mossy fiber projection (IIPMF) and the numbers of errors committed in a simple radial-maze task, with free access to arms and all 8 arms containing a food reward, each dot representing the mean of one inbred strain. Hippocampal data are based on 4 male mice per strain, behavioral data are from 3 males per strain. Data taken from Crusio et al., 1987 \[11\]. ::: ![](1744-9081-1-3-3) ::: At first sight, this result seemed to confirm our hypothesis: animals with larger IIPMF projections committed fewer errors, mastering the task more rapidly than animals with smaller IIPMF. However, matters were perhaps more complicated than that. Upon closer examination, it appeared that many animals used a kinesthetic strategy to solve the task, visiting adjacent arms in a clockwise or counter-clockwise fashion \[[@B11]\]. Whether such a strategy is based on spatial orientation capabilities or not, is not directly evident. Therefore, we decided to modify the radial maze task, as it was known from work with rats that confining subjects for 5 sec to the central platform in between subsequent arm choices interrupts this kind of chaining response \[[@B18]\]. In addition, observations made during behavioral testing suggested that mice are probably not using extra-maze cues when opaque arms are used as they rarely seemed to look upwards. We therefore replaced the PVC arms with arms made of clear Plexiglas and installed guillotine doors to enable the application of a confinement procedure. To facilitate the use of extra-maze cues for spatial orientation, we placed several objects close to the maze (of course, the experimentator is already a very visible cue in him/herself). This was done because the best-performing strain in the pilot experiment, C3H/HeJ, carries the Pde6b^rd1^mutation causing retinal degeneration \[[@B19]\]. Although these animals are not yet blind at the age that we use them (about 3 months, see \[[@B20]-[@B22]\]), their visual acuity obviously will be severely impaired, so we wanted to make this task as easy on their visual systems as possible. Six male mice from each one of nine different inbred strains were tested in this modified task. Because this task was expected to be more difficult than the previous one, we tested the animals for 5 days, again giving just one trial per day \[[@B23]\]. As in the previous experiment, animals from several strains mastered the task with surprising ease, whereas other strains did not improve their scores at all. Again, the IIPMF sizes correlated strongly with performance (r~S~= -0.92, df = 7, P\< 0.01; see Fig. [4](#F4){ref-type="fig"}). However, in contrast to the previous experiment, animals did not exhibit any obvious kinesthetic strategies any more. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Means ± SEM of the sizes of the hippocampal intra- and infrapyramidal mossy fiber projection (IIPMF) and the numbers of errors committed in a simple radial-maze task, with subjects confined to the central platform for 5 sec. in between subsequent arm choices and all 8 arms containing a food reward. Each dot represents the mean of one inbred strain. n= 6 male mice per strain. Data taken from Schwegler et al., 1990 \[23\]. ::: ![](1744-9081-1-3-4) ::: At the same time, we tested the same number of animals and strains in another radial-maze task that did not require any spatial orientation in order to be solved \[[@B23]\]. Here, opaque PVC arms were used and instead of manually operated guillotine doors we employed perforated aluminum plates fixed to the floor with adhesive tape. Subjects could easily open the doors, but as this took a few seconds, this procedure was also expected to disrupt any kinesthetic strategies. As with the previous spatial task, performance in this experiment ranged widely between different inbred strains and kinesthetic strategies were, indeed, absent. However, no correlation whatsoever with hippocampal mossy fibers became apparent (data not shown, see \[[@B23]\]). The results obtained were in accordance with our hypothesis that sizes of the IIPMF would correlate with spatial learning capacities, but not with nonspatial learning abilities. These data therefore provided support for the cognitive mapping theory of O\'Keefe and Nadel \[[@B16]\], which postulates that the hippocampus is uniquely involved in the regulation of spatial, allocentric memory. However, an alternative explanation was available, too. Olton \[[@B24]\] has hypothesized that the hippocampus regulates working memory, as opposed to other brain systems that would modulate reference memory, regardless of whether the information concerned was spatial or nonspatial in nature. Under this hypothesis, working memory stores information that is pertinent to one trial only (for instance, which arms have already been visited), but which has to be erased before the next trial to allow correct performance. Reference memory concerns information that is pertinent to all trials (for instance, the fact that food can be found at the end or an arm). Obviously, our spatial task had been a working memory task, whereas the nonspatial task was a reference memory task. Our results were therefore compatible with both competing theories, that of O\'Keefe and Nadel \[[@B16]\] and that of Olton \[[@B24]\]. We therefore modified our task yet again, to allow simultaneous measurement of working and reference memory in both spatial and nonspatial versions of the radial maze. More complex tasks dissociating working and reference memory ============================================================ Following Nadel and McDonald \[[@B25]\], we trained animals from the same nine inbred strains on a task in which only four out of the eight arms were systematically rewarded, the other four arms never containing any accessible food \[[@B26]\]. Two experiments were done. In one the task was spatial, using the radial maze with Plexiglas arms and guillotine doors as described above. In the other one, the task to be mastered was non-spatial, using the radial maze with opaque PVC arms, combined with guillotine doors. In both tasks, animals were confined to the central platform for 5 sec between subsequent arms choices. In the spatial version, mice were trained to locate four food rewards that were always placed in the same set of four arms. Each individual mouse had its own set of four rewarded arms. Following Olton\'s definition, entries into an arm that is never baited constitute a reference memory (RM) error, whereas repeat entries into an arm that has been visited previously constitute working memory (WM) errors. To prevent animals from using within-maze cues, the maze was rotated 45° at the end of each day (between subsequent trials), so that intra-maze and extra-maze cues were dissociated. This procedure prevented animals, e.g., from following hypothetical olfactory trails and forced them to use extramaze cues exclusively. In the nonspatial version, arms were marked by different black-white patterns on their floors. Food rewards were now associated with different sets of black/white patterns, each individual mouse again having its own combination of rewarded and non-rewarded patterns. As in the spatial task, RM and WM errors can now be defined. This experiment permitted the simultaneous measurement of WM and RM errors in tasks that either required spatial orientation abilities or not. O\'Keefe and Nadel\'s theory \[[@B16]\] would predict covariations between the IIPMF and both WM and RM performance in the spatial, but not in the non-spatial task. Olton\'s hypothesis \[[@B24]\] would predict covariations with WM in both the spatial and the nonspatial tasks, but not with RM in either task. As shown in Fig. [5](#F5){ref-type="fig"}, the results of this experiment were in complete agreement with the predictions of the cognitive mapping theory of O\'Keefe and Nadel. In addition, we found that both in the spatial and in the non-spatial tasks WM and RM were correlated very strongly, raising doubt as to whether the distinction between these two types of memory really is pertinent, at least for mice. Indeed, in those experiments where authors reported a dissociation between these two forms of memory (e.g., \[[@B27]\]), almost invariably two different tasks were used, one purported to be a WM task, the other one an RM task. Obviously, such different tasks differ for many more components, which might explain any dissociation at least as well. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Means ± SEM of the sizes of the hippocampal intra- and infrapyramidal mossy fiber projection (IIPMF) and the numbers of working-memory (WM) and reference memory (RM) errors committed in an 8-arm radial maze with only 4 arms containing a food reward. Animals were tested during 10 days, one trial per day. Numvbers of errors shown are cumulative error counts on days 3--10. Upper panels: Spatial task. Lower panels: Non-spatial task. Left panels: Working-memory errors. Right panels: Reference memory errors. Note the different scales in the upper and lower panels, the non-spatial task obviously being much easier for the subjects. ::: ![](1744-9081-1-3-5) ::: Using the radial maze to demonstrate mutational and pharmacological effects =========================================================================== Several other authors have also investigated strain differences in radial-maze learning tasks. They reported results that were sometimes rather different from ours. However, the experimental design and apparatus used differed strongly from ours, too. For instance, Roullet and Lassalle \[[@B28]\] used female instead of male mice, whereas their maze and the one used by Ammassari-Teule et al. \[[@B29]\] was elevated. Not surprisingly, given the different behavioral results, Roullet and Lassalle \[[@B28]\] did not find any correlations with the IIPMF. In contrast, we performed several experiments in the spatial task with all arms baited, investigating mutational \[[@B30]\], Y-chromosomal \[[@B31],[@B32]\], and mtDNA \[[@B33]\] effects on IIPMF distributions and learning behavior. When the results of all these studies are combined with those from our 1990 experiment (\[[@B23]\], see Fig. [6](#F6){ref-type="fig"}), we obtain a remarkably consistent picture. Despite the fact that these experiments took place over a period of about 10 years and were carried out by different people in different laboratories using different morphometrical methods, the overall correlation obtained is r~S~= 0.81 (df = 17, P\< 0.0001), which is only marginally lower than the correlation found in our 1990 study \[[@B23]\]. ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Means ± SEM of the sizes of the hippocampal intra- and infrapyramidal mossy fiber projection (IIPMF) and the numbers of errors committed in a simple radial-maze task, with free access to arms and all 8 arms containing a food reward, each dot representing the mean of one inbred strain. Data taken from Figure 4 (8 strain means) and from Refs. \[30-33\] (10 additional strain means). For clarity, SEMs have been omitted. ::: ![](1744-9081-1-3-6) ::: Of course, correlations between two variables need not indicate a causal relationship and the IIPMF-spatial learning correlation might be spurious. Hypothetically, a third, as yet unknown, neuronal variable might be the one causing the observed strain differences in learning. The IIPMF-learning correlation would then only appear because this hypothetical third variable itself would be correlated with the IIPMF. However, in the present case we believe that there are strong indications that this correlation is, indeed, causal. First, there is the remarkable consistency and strength of the correlations reported. If a third variable would be directly correlated with learning performance, the IIPMF correlation would be only secondary and the third variable would have to correlate with learning even stronger than the IIPMF do. This would be difficult to imagine. Second, a correlation between strain means differs in one important respect from ordinary correlations, estimated from individual values. Namely, such a correlation represents a genetic correlation, meaning that gene effects on one variable are correlated with gene effects on the other variable \[[@B34],[@B35]\]. Such a situation makes it highly likely that the statistical relationship found is, indeed, a causal one. Finally, we have also addressed this question in a pair of experiments in which newborn pups of a strain (DBA/2) known to possess scant IIPMF projections and feeble learning capacities in the radial maze were treated with thyroxin in the early postnatal period \[[@B36]-[@B38]\]. This treatment induces an increase in the size of the IIPMF in adults and we found that this increase was accompanied with a significant improvement in the spatial learning capabilities of these animals, both in a task in which all arms were baited as well as in a task in which only 4 of the arms were consistently baited. Other behaviors correlated with hippocampal neuroanatomy ======================================================== Although the present review concerns radial-maze learning, we would like to briefly mention some other behaviors that have been found to correlate with the IIPMF. The very first correlation that was reported, with two-way active avoidance learning, has already been mentioned above (for a review, see also \[[@B39]\]. Other correlations that have been found are with intermale aggression \[[@B40]-[@B42]\], paw preference \[[@B43]\], reversal learning in a water navigation task \[[@B44],[@B45]\], visual and tactile discrimination in a Y maze \[[@B46]\], and exploration \[[@B47]-[@B49]\] and habituation \[[@B50]\] in an open field. Conclusion ========== Taken together, we conclude that in inbred mice the hippocampal intra- and infrapyramidal mossy fiber projection plays an important role in the regulation and/or modulation of spatial orientation capabilities in the radial maze. Larger IIPMF projections go with better learning capabilities in the spatial radial-maze tasks described above. These correlations are specific to spatial learning, as no correlations were found in radial-maze tasks that could be solved using non-spatial cues or that employed elevated mazes, possibly inducing increased levels of anxiety in the experimental subjects. At this point it is not yet fully understood why variations in the size of the IIPMF have such drastic consequences for an animal\'s behavior. It has been shown that these variations are associated with differences in spontaneous bursting in region CA3 \[[@B51]\] and with differences in LTP \[[@B52]\]. LTP is generally regarded as the most promising physiological mechanism underlying learning and memory, although the extent of its implication in these processes remains controversial (see \[[@B53]\] for a critical discussion). However, it was recently found that blocking of mossy fiber LTP or LTD does not abolish spatial learning capabilities in mice \[[@B54]\]. Therefore, the IIPMF apparently modulate behavior by other mechanisms. One possibility may be that larger IIPMF would somehow facilitate LTP in the hippocampal CA1 region. This hypothesis would be consistent with our findings that larger IIPMF go with better spatial learning abilities and diminished two-way active avoidance learning, behaviors that are abolished, respectively enhanced, after a hippocampal lesion \[[@B16]\]. Further research is clearly needed to address these questions. Finally, among the different hippocampus-dependent radial maze tasks presented above, the simple spatial one (all eight arms rewarded, short confinement to the central platform between subsequent arm choices) appears to be the most useful task: It is rapid (in our protocol we use 5 daily trials only, each trial taking on a few minutes near the end of training) and we have used it successfully to investigate effects of pharmacological treatments \[[@B55],[@B56]\] or to detect subtle mutational effects \[[@B57]\]. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= The ideas presented in this manuscript are the results of many years of discussions between WEC and HS. The manuscript was drafted by WEC. Both authors read and approved the final manuscript. Acknowledgements ================ We would like to thank our colleagues and students who over the years have collaborated with us on the studies presented above: Ingrid Brust (Heidelberg, Germany), Pascale Guillot (London, UK), Laure Jamot (Marseille, France), Hans-Peter Lipp (Zurich, Switzerland), Charlotte Marican (Paris, France), and Frans Sluyter (London, U.K.).	PubMed Central	2024-06-05T03:55:57.475226	2005-4-22	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143776/", "journal": "Behav Brain Funct. 2005 Apr 22; 1:3", "authors": [ { "first": "Wim E", "last": "Crusio" }, { "first": "Herbert", "last": "Schwegler" } ] }
PMC1143777	Background ========== Habituation effects have been observed in a wide range of neural systems from simple sensory responses \[[@B1]\], to higher-order neural representations such as motion-sensitive populations in Area MT \[[@B2]\], and regions responding to written and spoken language \[[@B3]\]. We can take advantage of neural habituation to study the preattentive categorization of stimuli. By presenting a single speech stimulus repeatedly, we can observe habituation to that sound, then by comparing this condition to one in which a \"deviant\" stimulus occurs after a series of repeated \"standards,\" we can also determine which brain regions are sensitive to the change between the two stimuli (see review in \[[@B4]\]). Critically, habituation phenomena can be studied with passive paradigms, which have tremendous advantages in the study of speech perception, particularly for the study of populations in the process of acquiring language, or adult populations differing in their early language experience. Any overt task involves a range of decision processes that can act to obscure the processes underlying speech perception under more natural conditions. This is a particular problem for studying speech developmentally, because attention and decision processes develop very slowly \[[@B5],[@B6]\] which may cause us to underestimate children\'s ability to perceive phonetic contrasts. Furthermore, adults are often quite good at discriminating sounds in laboratory tasks that they do not perceive phonetically \[[@B7]\]. Even after extensive training on perception and production, it can be difficult to establish whether second language learners are using the same underlying mechanisms as monolinguals, even when many surface aspects of behavior are similar between the two groups (see, for example, \[[@B8]\]). Neuroimaging can help establish how stimuli are discriminated, for example by showing differential activity in regions specifically implicated in phonetic processing. The goal of the current study is to develop a short interval habituation trial paradigm optimized for event-related fMRI designs that builds on the strengths of currently available methods, and can be applied to a range of populations of interest. Developing an auditory habituation paradigm for fMRI ---------------------------------------------------- The mismatch negativity (MMN) response, as observed in EEG (and its equivalent mismatch field response in MEG) is a form of neural dishabituation that has been used as an index of the categorization of speech sounds \[[@B4]\]. For example, Naatanen et al. \[[@B9]\] presented stimuli from the partially overlapping vowel systems of Finnish and Estonian to native speakers of each language. They found smaller MMN responses for a deviant stimulus that was not a native-language vowel, even though it was acoustically more different from the standard than a non-native stimulus. Another major advantage of this technique is that the MMN can be observed in the absence of any attention-demanding task. Typically, subjects in such studies are reading or watching a film, but characteristic mismatch responses have also been observed in sleeping infants and comatose patients (see Cheour et al. \[[@B10]\], for review). There has been increasing interest in combining the advantages of the MMN paradigm with the higher spatial resolution available using fMRI. For example, in a study by Fiez and colleagues \[[@B11]\], spoken words were presented repeatedly and responses were observed to both repeated words and occasional deviants. A comparison between these conditions revealed responses in temporal regions involved in speech and auditory processing, as well as frontal and parietal regions implicated in attentional function. A number of technical challenges complicate this approach. A very real challenge to these studies is the noise created by the MR scanner itself, particularly at high field strengths, and when using scanning sequences optimized to provide higher signal to noise ratios \[[@B12]\]. Thus, there is a trade-off between the strength and resolution of the signal and the ability to present acoustic stimuli in relative quiet. Another trade-off exists between optimization of the stimulus presentation parameters for interpretation of the dependent measure (the blood oxygenation level dependent, or BOLD response) and paradigm optimization for the observation of change responses. The BOLD response evolves very slowly, making it difficult to observe \"baseline\" responses to rapidly repeated stimuli typical of MMN designs. When stimuli are presented at a constant rate, the BOLD response from any stimulus cannot easily be deconvolved from responses to previous stimuli. On the other hand, the temporal jittering of stimuli critical to fast event-related designs \[[@B13]\] is undesirable because mismatch responses are reduced when stimuli are not presented at a constant rate \[[@B14]\]. As we show below, directly comparing time course information from standard and deviant stimuli can provide novel insights into the role of particular brain regions in the perception of phonetic change. In the current study we address a number of these methodological challenges by presenting short interval habituation trials composed of trains of four stimuli. On \"standard\" (STD) trials, these consisted of four repetitions of the same speech sound, whereas on \"deviant\" (DEV) trials, three repetitions of one sound were followed by a different, dishabituating sound (see also \[[@B15],[@B16]\]). We used a clustered data acquisition protocol in order to present stimuli in silence, and combined a relatively short repeat time (i.e., whole brain volumes obtained every 3 seconds) with a long inter-trial interval to maximize our ability to derive time series from individual subjects\' data. The results demonstrate the feasibility of the methodology, and reveal responses to passively presented phonetic stimuli. In particular, activity in the left posterior superior temporal gyrus related to phonetic change is robust at the single-subject level and may serve as a signature for automatic categorization of speech sounds. Results ======= Two separate analyses were undertaken. In the first, we compared responses from trials on which speech was presented (collapsing across standard and deviant trials) to silence, in order to observe the BOLD signal for passive perception of speech. In the second analysis, we directly compared the deviant and standard trials, in order to study responses to phonetic change. We discuss each analysis in turn. Responses to speech versus silent baseline (SPCH \> SIL) -------------------------------------------------------- As shown in Figure [1](#F1){ref-type="fig"}, in comparisons between speech and silence, the STG is active bilaterally from its most posterior extent up to but not including the temporal pole (activity in the anterior portions of the temporal lobe is difficult to observe in fMRI, \[[@B17]\]). This is consistent with the well-established role these regions play in auditory processing, in particular for speech \[[@B11],[@B18]-[@B20]\]. This pattern of response was observed consistently across all subjects. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### SPCH \> SIL activations. Areas of greater activation for speech (SPCH) than silence (SIL); voxels thresholded at p \< .005 uncorrected. Note the lack of activity in Heschl\'s gyrus (top), likely the result of stimulation by the acoustic noise of the scanner. ::: ![](1744-9081-1-4-1) ::: As shown in Table [1](#T1){ref-type="table"}, we also observed cerebellar responses to speech stimuli bilaterally in the superior portion of the vermis. Numerous neuropsychology studies have linked damage in this area to disorders of speech production \[[@B21]\]. Interestingly, activity in the vermis has also been observed in auditory perception experiments in which stimuli are presented rhythmically, for example trains of tones or frequency modulated sweeps \[[@B22]\] or clicks \[[@B23]\] and pairs of words \[[@B24]\]. Thus, it is possible that the cerebellar responses in this task reflect rhythmic properties of the stimulus presentation paradigm. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Regions of significant activation in the SPCH \> SIL contrast ::: Talairach coordinates Area Z(voxel) cluster size ----------------------- ----------- ---------- -------------- -59, -33, 9 Left STG 5.50 1754 -51, -8, -3 4.45 -51, -21, 3 4.25 44, -25, 10 right STG 5.29 2512 53, -21, 3 5.14 50, -29, 5 5.07 0, -35, -2 vermis 3.15 53 -10, -33, 0, 2.79 Note: SPCH \> SIL = increased response to speech relative to silence; STG = superior temporal gyrus. Cluster size is based on a voxel-wise threshold of p \< 0.005 uncorrected. ::: Responses to phonetic change (DEV \> STD) ----------------------------------------- Responses to phonetic change were observed in a broad network of regions, including a portion of left STG known to play a role in phonological processing (see Table [2](#T2){ref-type="table"}). Activity was also observed in regions implicated in novelty responses, and right hemisphere regions potentially involved in processing of extra-phonetic aspects of the stimuli. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Regions of significant activation for the DEV \> STD contrast ::: Talairach coordinates Area Z(voxel) cluster size ----------------------- ------------------- ---------- -------------- Left -51, -36, 24 STG/SMG 3.84 122 -40, -34, 22 border 3.41 -42, -33, 29 3.31 -20, 57, 21 MFG, BA10 4.49 57 -20, 52, 27 2.71 -28, 27, 45 MFG 3.75 45 -30, 18, 47 2.91 -20, 18, 49 3.36 -24, -41, 43 Parietal 4.13 30 -24, -38, 50 2.95 -40, -72, 42 Precuneus (BA19) 3.44 36 -40, -64, 36 2.89 -40, -74, 33 2.78 Right 32, -9, -15 Hippocampus 4.65 40 4, -23, 14 Thalamus 3.40 64 4, -13, 10 (medial dorsal) 2.73 53, -17, 14 Postcentral (BA3) 3.95 52 46, -11, 17 3.50 55, -2, 6 R Precentral, BA6 4.52 35 28, -7, 17 Putamen 3.12 27 36, 0, 42 MFG (BA6) 3.43 27 44, 0, 44 2.80 Note: DEV \> STD = increased response to deviant trials relative to standard trials; STG = superior temporal gyrus; SMG = supramarginal gyrus; MFG = middle frontal gyrus; BA = Brodmann\'s Area. Cluster size is based on a voxel-wise threshold of p \< 0.005 uncorrected. ::: Dishabituation of the BOLD response in posterior left STG/SMG border -------------------------------------------------------------------- A restricted region of left posterior STG responds preferentially to trials containing a phonetic change (DEV) in comparison with trials containing no phonetic change (STD), as shown in Figure [2](#F2){ref-type="fig"}. This region is along the border with SMG and has been reported in a number of studies involving active phonetic change judgments \[[@B25],[@B26]\]. This region also shows a marginally significant response to the speech relative to silence contrast, voxelwise t(7) = 2.63, p \< .05. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### DEV \> STD activations. Areas of greater activation for deviant (DEV) than standard (STD); voxels thresholded at p \< .005 uncorrected. The perisylvian region (top) is actually located along the border of superior temporal and supramarginal gyri. ::: ![](1744-9081-1-4-2) ::: We observed a high degree of consistency across subjects in the pattern of activity in this region. The peak activations nearest this region are plotted on a translucent rendering of the MNI template brain in Figure [3A](#F3){ref-type="fig"}. Every individual had a peak BOLD response in the DEV \> STD contrast close to the peak in the random effects analysis (mean euclidean distance = 8.58, range = 4.47 -- 19.70, SD = 5.12). Time series for an 8 mm sphere around the mean peak (mean number of active voxels in sphere = 106.75, SD = 16.67) are plotted for each subject in Figure [3B](#F3){ref-type="fig"}. Note that for most subjects, this region shows subthreshold responses to speech overall, whereas in all subjects, responses to DEV trials are greater than responses to STD trials (Figure [3C](#F3){ref-type="fig"}). This can be seen in the mean time series (Figure [3D](#F3){ref-type="fig"}) for all 8 subjects. Critically, the pattern observed here is consistent with a dishabituation response: The region typically responds to speech, yet its response is attenuated when the same stimulus is presented repeatedly; when a novel stimulus is presented, a heightened response is observed. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Peak activations and time series from individual subjects for the DEV \> STD contrast. A. Peak activations for the DEV \> STD contrast in each subject nearest to the group mean plotted on a translucent rendering of grey matter. B. Time series for responses to all speech stimuli in each individual\'s peak., C. Mean percent signal change at scans 2 and 3 (approximately 2--7 seconds after the fourth syllable in each train) for STD and DEV trials. D. Mean time series for DEV and STD for all subjects. ::: ![](1744-9081-1-4-3) ::: We also observed timecourse differences between responses to speech and responses to phonetic change. Given the timing of scans relative to auditory stimulation (Figure [4](#F4){ref-type="fig"}), and the fact that the peak of the hemodynamic response occurs between 4 and 6 seconds \[[@B27]\], we would expect responses to the onset of speech stimuli to occur at around scan 2 and responses to the deviant stimulus to occur around scan 3. In fact, there is a difference in when the peak response to speech (median = scan 2) occurs and the peak difference between deviant and standard stimuli (median = scan 3), which is reliable according to a Wilcoxon signed-ranks test, W = 21, n~s/r~= 6, p \< .05. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Short interval habituation trials paradigm. Schematic diagram of the stimulus presentation paradigm. A. Timing of scans relative to stimulus presentation, taking as the stimulus onset either the onset of speech sounds (in red) or the fourth stimulus which distinguishes deviant from standard trials (in green). B. Spectrograms of standard (STD) and deviant (DEV) stimulus trains. The primary phonetic difference between the /la/ and ![](1744-9081-1-4-i1.gif) is the onset frequency of the third formant, circled in red. ::: ![](1744-9081-1-4-4) ::: Novelty responses ----------------- The hippocampus is known to be involved in novelty detection, in keeping with its role in the encoding of novel episodic memories \[[@B28]\]. In the current experiment, an anterior region of right hippocampus responded preferentially to deviant trials, consistent with this role. Two recent studies of neural responses to novel stimuli have found similar activations in hippocampus: Kiehl et al. \[[@B29]\] presented novel (low-probability) natural sounds in the context of an auditory target detection task with sine wave tones as the baseline stimuli. In addition to bilateral hippocampal activations, they observed novelty responses in regions of left superior temporal and superior frontal gyrus, right postcentral gyrus, and the medial dorsal nucleus of the thalamus similar to the current study. In a similar task using visual stimuli, Yamaguchi et al. \[[@B30]\] found hippocampal responses, as well as left superior frontal, right middle frontal and left parietal activations similar to the current study. Right hemisphere responses -------------------------- A number of right hemisphere regions also responded more strongly to deviant trials than to standards. Although our paradigm did not require any motor response, and did not result in any primary motor or somatosensory activity, the pattern of BOLD signal in the right precentral gyrus, postcentral gyrus and putamen, was similar to that observed in studies of disparate motor and proprioceptive responses \[[@B31]-[@B35]\]. One feature that the current experiment shares with other tasks in which these regions have been activated is the rhythmic presentation of acoustic stimuli. For example, in the Joliot et al. study \[[@B31]\], subjects were required to tap their fingers in time with a tone. In the current study, no motor response was required, but trains of stimuli were presented in a regular rhythm (see Figure [4](#F4){ref-type="fig"}). It is unclear, however, why these regions would respond preferentially to deviant trials whereas the vermis responds to rhythmic aspects of both STD and DEV conditions. One peak in right precentral gyrus (BA6, Tc = 55, -2, 6) appears from inspection of individual subjects to reflect activity in two different regions: a region of precentral gyrus that has been found in many of the same conditions as the post-central region discussed above, and a portion of right anterior STG. This may result from the fact that, using natural stimuli for which both steady-state and transitional portions differ between the standard and deviant stimuli, there are some unavoidable extraphonetic differences between the stimuli that may activate right superior temporal regions involved in the processing of tone timbre and amplitude \[[@B19],[@B22],[@B36]\]. Discussion ========== A central motivation for this study was to map out the spatial topography and timecourse of BOLD responses in regions generally sensitive to speech stimulation, and regions sensitive to changes in the speech signal. We also sought to test whether such responses might be collected under passive conditions similar to those in electrophysiology research using mismatch negativity designs, which have proven useful in developmental and cross linguistic studies in which explicit discrimination and labelling may introduce confounds. The use of short interval habituation trials affords the possibility of directly contrasting the topography and time course of responses to speech stimuli and phonetic habituation. Characterizing neural responses to speech stimuli and phonetic change --------------------------------------------------------------------- This study demonstrates that under passive presentation conditions, BOLD responses to short trains of syllables inserted within the 1.5 seconds of relative silence between volume acquisitions are reliably obtained in a broad network of superior temporal gyrus regions. This includes both early auditory areas involved in the processing of spectrally complex sounds \[[@B37],[@B38]\] and regions that are potentially specific to speech processing \[[@B19],[@B39]\]. We also found that passive presentation of a short interval of habituation, (i.e. four syllables in rapid succession), establishes sufficient context to generate dishabituation effects related to the change of single phoneme. A restricted region of posterior, left STG, along the border with supramarginal gyrus responded contrastively to information in the fourth syllable, such that a novel phonetic onset produced a greater response relative to the habituated one. This suggests a role for this region in phonetic processing, consistent with several lines of converging evidence. First, this region is active in explicit phonetic discrimination \[[@B25]\] and comparisons of passive listening to speech with other stimuli \[[@B40]\]. Furthermore, in a study of native Japanese speakers learning English, activity in this region is correlated with accuracy in discriminating English speech sounds \[[@B26]\]. These findings suggest that responses in such regions may be relatively specific to phonetic change. Although regions of the superior temporal gyrus have been implicated in mismatch negativity studies using simple sine-wave tones as stimuli, the regions observed in the current study are distinctly posterior and superior to the putative location of the mismatch negativity responses for those relatively simple stimuli \[[@B41],[@B42]\]. In a direct comparison of passive responses to stimulus change for speech stimuli and tones, Celsis et al. \[[@B20]\] found speech-specific responses in a region of left supramarginal gyrus contiguous with the extent of the activity observed along the STG/SMG border in the current study. Furthermore, a similar region is specifically activated by sine-wave stimuli when they are perceived as speech compared to the same stimuli when they are perceived as oscillating tones \[[@B43]\]. Finally, the posterior portion of left STG is involved in reading and seems in particular to be critical to aspects of decoding that require mapping of visual letters onto speech sounds \[[@B44]-[@B46]\]. Taken together, the data suggest a critical role for this region in the passive perception and categorization of speech sounds. Advantages and disadvantages of using short interval habituation trials in event related designs ------------------------------------------------------------------------------------------------ Short interval habituation trials provide a fast, passive paradigm, with which it is possible to observe robust data in small numbers of subjects. One critical advantage of short interval habituation trials design is that it allows us to examine relationships between responses to syllables and effects of phonetic habituation. Because the entire time course of activation is collected after each trial, it is possible to examine the response to speech stimulation in general and habituation specifically. For example, in Figure [3D](#F3){ref-type="fig"}, the BOLD signal for standard trials is consistent with general sensitivity to speech, but the response to deviant trials has a time course consistent with dishabituation to the fourth stimulus in a second-long train. The ability to extract a complete time series from each trial is an improvement over the silent paradigms with much longer repeat times used in earlier studies \[[@B20],[@B41]\]. In those paradigms, images are acquired every 10--12 seconds, so that the BOLD response to scanner noise returns to baseline between scans. This means that data are collected at a single discrete time point for each trial. While this makes it possible to isolate regions that respond more strongly to a run of stimuli containing deviants than a run containing only standards, it only provides time course information when stimuli are presented at multiple delays relative to the data acquisition. The main disadvantage of the current technique relative to slower designs is the lack of sensitivity to early auditory processing. For example, activity was not observed in left primary auditory cortex for either of the contrasts examined (speech \> silence, deviant \> standard). This may be due to the loud acoustic noise generated by the flipping of the gradients in the in-out spiral sequence which is likely to activate neurons in this region. If this region is activated by recurring scanner noise, the BOLD response may become saturated during the experiment, limiting the ability to observe speech-related responses in the current paradigm. Thus, there is a design tradeoff between the efficiency with which change-related responses to speech and early auditory processes can be observed. It may be possible to combine the current approach with a long inter-trial interval in order to observe the contribution of early auditory processing to speech perception. Belin et al. \[[@B37]\] developed a presentation paradigm in which individual stimuli are presented at different times relative to the onset of scanning on each trial. In this way, it is possible to reconstruct time course information by combining responses from multiple trials. Because the trains of stimuli used in the current design are quite brief (just over one second), it would be possible to present these trains in a similar manner, allowing us to observe the contribution of early auditory areas to phonetic change perception. Possible advantages of this technique may be outweighed by practical concerns. By using a short repeat time to sample time course information at 5 intervals for each short interval habituation trial, we collected full data sets for a two-condition contrast in a relatively short time (approximately 26 minutes of scanning). An experiment that took five times this long to collect data for a single pair of stimuli would not be practical in many cases. In particular, if one wanted to compare responses to native language stimuli with responses to non-native stimuli \[[@B25],[@B26]\] it would require multiple scanning sessions. Thus, while it has disadvantages for observing low-level auditory activity, the short interval habituation trial paradigm has a number of features that make it particularly applicable to studies of different populations of interest. The lack of any explicit task allows for investigation of responses to speech stimuli under processing conditions similar to well established procedures in mismatch negativity research. Furthermore, passive presentation of auditory stimuli while subjects are engaged in an unrelated visual activity (i.e. watching a video) reduces the influence of attentional and executive factors. This eschews difficulties with overt tasks where such factors may lead to an underestimation of children\'s performance, or to strategic attention to extra-phonetic cues that may lead to an overestimation of performance by adult second-language learners \[[@B7]\]. The granularity of the time series data is also critical in the analysis of data from developmental and cross-linguistic studies: In cases where no dishabituation effect is observed, establishing that some response to speech sounds is observed makes it less likely that the null result is a type I error. Finally, this study demonstrates robust data at the level of individual subjects, providing a strong basis for the study of individual variability in populations of interest. Conclusion ========== Using short interval habituation trials, we were able to isolate specific regions of superior temporal gyrus that are sensitive to changes in phonetic information in the absence of any explicit instructions or task in adult native English-speaking subjects. Many theoretical questions central to the investigation of speech processing revolve around comparisons of listeners of different ages and from different language backgrounds. The current approach can be extended to yield insights into the development and plasticity of the basic mechanisms that subserve phonetic perception. Methods ======= Subjects -------- Eight right-handed adult native English speakers (ages 23--38, mean = 26.8, SD = 4.6, 2 females) participated in the experiment. Subjects were paid for their participation. Stimuli ------- Natural speech sounds used in the experiments consist of recordings of an adult male native English speaker (JDZ) saying the syllables ![](1744-9081-1-4-i1.gif) and /la/ with a similar pitch and intonation pattern. Stimuli were digitized in 16-bit mono at 22050 Hz and cropped to 250 ms in duration using Praat \[[@B47]\] by deleting individual glottal pulses from the steady-state portion of the vowel. Recordings were made in a soundproof booth at the Speech and Hearing Research Center of the City University of New York Graduate Center. Stimulus presentation --------------------- On each trial, a train of four stimuli was presented. On \"standard\" trials the same stimulus was repeated four times; on \"deviant\" trials the final stimulus differed from the first three. Silent trials were also included to allow a baseline for comparison. The stimuli had a duration of 250 ms, and were presented with an inter-stimulus interval of 50 ms, so that the duration of an entire stimulus train was 1.15s. This allowed us to leave 175 ms of silence between both the onset and offset of the auditory stimuli and the spiral data acquisition time to prevent auditory masking. Stimuli were presented at approximately 70 dB. The IFIS system, combined with EPrime (both from Psychology Software Tools) software was used to synchronize stimulus presentation with the scan sequence. The experiment consisted of 12 pseudorandomized functional runs of 9 trials each, with an ITI of 12s. In each block, equal numbers of standard, deviant and silent trials were run. In the first six blocks, ![](1744-9081-1-4-i1.gif) served as the standard stimulus and /la/ was the deviant, and in the last six blocks this was reversed. Over the course of a full session, this provided 36 trials for each stimulus type (standard, deviant, silence). Each run began and ended with an extra 12 s silent trial in order to account for heterogeneity in the magnetic field at the beginning of scanning runs and to provide a full acquisition period for the final stimulus train in a block. Because of the high level of acoustic noise generated by the spiral sequence (\~120 dB), subjects were supplied with 30 db attenuating foam earplugs. In preliminary tests, this did not interfere with hearing the stimuli, but provided protection from the noise of the scanner. In addition to the earplugs, subjects were fitted with a large pair of padded piezo-electric headphones which provided additional protections from sound. We also used Tempurpedic pillows to fill in parts of the headcoil. This served two purposes: First, it aided in noise abatement. We have found that the headcoil itself acts to amplify acoustic noise by acting like a resonating body. By preventing the headphones from coming into direct contact with vibrating parts of the headcoil and filling in empty space directly around the subject\'s head this effect is reduced. The foam also helped subjects remain still. Finally, in order to make the experiment less tedious for subjects, nature films or cartoons were shown on the video monitor during stimulus presentation. Films were shown continuously throughout the session, providing a pattern of visual stimulation highly unlikely to be correlated with any experimental procedure. Data acquisition ---------------- After an initial three-plane localizer and a whole-head coronal localizer, a Fast Spin Echo sequence was taken in an axial-oblique plane prescribed to correct for head position in 3 dimensions (obviating the need for manual AC-PC alignment later in processing) TR = 3325 ms, TE = 68 ms, flip = 90°, FOV = 22, 5 mm slice thickness, 0 mm gap, matrix = 256 × 192, 20 slices, positioned to cover language and auditory processing regions. High-resolution T1-weighted images for normalization were taken using a 3D gradient echo SPGR sequence, axial plane, TR = 25 ms, TE = 5 ms, flip = 20°, FOV = 24 cm, 1.5 mm slice thickness, 0 mm gap, matrix = 256 × 256 × 160. Functional images were taken using the spiral in-out sequence developed by Glover and colleagues, and the same spatial prescription as the FSE, TR = 3000 ms, TE = 40 ms, matrix = 64 × 64. using a clustered acquisition sequence with a 3s TR and 1.5s TA. By using a clustered acquisition protocol, we are able to present stimuli in relative quiet, i.e., during 1.5s gaps during which no acoustic noise from the flipping of the gradients is present. This scanning sequence has been shown to have a very high signal to noise ratio \[[@B12]\]. Each functional run lasted 132s during which 44 volumes were collected. Data analysis ------------- We analyzed fMRI data using SPM2 in three major stages: pre-processing to retrieve the functional data and map all subjects into a common space; statistical parametric mapping to find regions with interesting patterns of activity and follow-up analyses using percent signal change estimates from regions of interest identified in the parametric maps. Pre-processing -------------- The first four volumes in each scanning session were deleted to allow the magnetic field to reach steady state. Slice-timing correction was then applied to account for the fact that slices are acquired in fixed order during a 1.5s TA for each 3s TR. Next, image realignment was applied to all functional images, generating a set of realignment parameters for each run and a mean functional image which was used to coregister functional scans to the FSE in-plane anatomical images. The FSE was then coregistered to the SPGR, and these parameters were applied to the functional scans. The SPGR was then normalized to MNI space resulting in oversampled voxels of 2 mm^3^. These parameters were applied to the realigned, smoothed functional images, and the normalized data smoothed using a FWHM kernel of 6 mm. Statistical Parametric Mapping ------------------------------ Statistical models were constructed by convolving the onsets of each trial type with a standard hemodynamic response function, including realignment parameters as covariates. These were used to generate first-level contrast images for each subject for two contrasts: 1) SPCH \> SIL, showing the pattern of positive correlation with the presence of any speech stimulus relative to silent baseline and 2) DEV \> STD showing the pattern of greater responses to deviant trials relative to standard trials. These contrast images served as the basis for random effects analyses. Results reported as significant exceed a voxel-wise threshold of p \< .005 and a spatial extent threshold of 25 contiguous voxels. This provides a conservative estimate of statistical significance \[[@B48]\]. Time series analyses -------------------- In order to examine the time series in posterior left STG for standard and deviant stimuli, a functional region of interest (ROI) was defined based on the mean image from all subjects in the DEV \> STD contrast. For each subject, an 8 mm sphere was drawn around this ROI and eigenvectors extracted using the VOI toolkit for SPM2. This provided a representative response for the region over time, which was then averaged for each stimulus type to generate a mean time series. Authors\' contributions ======================= The experiment was conceived, developed and reported collaboratively by both authors. JDZ was primarily responsible for data collection and analysis. Acknowledgements ================ We thank Gary Glover for providing us with the spiral sequence and reconstruction tools, Tor Wager for the visualization software used in Figure [3](#F3){ref-type="fig"}, and Oliver Tuescher for helpful discussions throughout the course of the study. This research was supported by NIH Kirchstein NRSA DC006352 to JDZ and a Merck Scholars Award to BDM.	PubMed Central	2024-06-05T03:55:57.477500	2005-4-22	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143777/", "journal": "Behav Brain Funct. 2005 Apr 22; 1:4", "authors": [ { "first": "Jason D", "last": "Zevin" }, { "first": "Bruce D", "last": "McCandliss" } ] }
PMC1143778	Background ========== It has long been known that prenatal exposure to alcohol can have devastating effects. In 1968, Lemoine et al. published a paper, in French, that described \"des anomalies dans les infants de parents alcooliques\" \[[@B1]\]. Five years later Jones and Smith \[[@B2]\] followed with their now classic paper on Fetal Alcohol Syndrome (FAS), which is the name given to a group of physical and mental birth defects that are the direct result of a woman\'s drinking alcohol during pregnancy. These defects may include mental retardation, growth deficiencies, central nervous system dysfunction, craniofacial abnormalities, and behavioral maladjustments (for an enumeration, see, among others, \[[@B3]\], and <http://www.acbr.com/fas/index.htm>). Since then literally thousands of studies have been done and substantial advances in the understanding of FAS have been made, not in the least because of animal models, among them the mouse \[[@B4],[@B5]\]. These models strongly parallel the physiological responses to alcohol in human development and have led to valuable insights into the susceptibility to alcohol of specific parts of the developing embryo. One of the more vulnerable brain regions, for instance, is the hippocampus, which is believed to be a primary target of prenatal alcohol and therefore responsible for many neurobehavioral abnormalities \[[@B6]\]. Thus, Riley et al. \[[@B7]\] found the effects of prenatal alcohol exposure on behavior to be similar to those of hippocampal lesions, while hippocampal dysfunction as a consequence of prenatal alcohol exposure has also been assessed in spatial learning tasks such as the Morris water navigation task (see, among others, \[[@B8]\]). However, little attention has been paid to the genetic vulnerability of the developing embryo to prenatal alcohol exposure, particularly genetically determined differences in the sensitivity of the developing hippocampus. These genetically determined differences might, at least partly, explain one of the major questions in the etiology of FAS: why only a small percentage of alcoholic women give birth to children with fetal FAS, whereas other alcoholic women who drink the same amount do not. The aim of this study was to determine whether different genotypes with varying hippocampal anatomy are oppositely affected by similar exposures to prenatal alcohol. To this end, pregnant females of three highly inbred mouse strains, C57BL/6J, BALB/cJ, and DBA/2J, which vary in hippocampal anatomy (e.g. \[[@B9],[@B10]\]), were exposed to a 12% alcohol solution as their only source of liquid throughout gestation. Following this, the male offspring was tested for their capacity to master an eight-arm radial maze, a spatial navigation task well known to depend on an intact hippocampus. Subsequently, the animals were sacrificed and the sizes of the hippocampal intra- and infrapyramidal mossy fiber (IIPMF) terminal fields were determined. Necessary control groups (untreated and pair-fed) were included. Results ======= Table [1](#T1){ref-type="table"} shows the numbers of breeding pairs used (pairings), the number of pairings resulting in pregnancies, and the number of pregnancies leading to live births in the nine different (3 strains × 3 groups) groups. Although ethanol exposed C57BL/6J females became pregnant at similar rates as animals from other treatment groups and strains, these pregnancies resulted in significantly lower live births than in the pair-fed and control groups (χ^2^= 9.5, df = 2; p \< 0.01). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Numbers of breeding pairs used (pairings), resulting pregnancies, and live births in the different groups ::: Strain/Treatment \# Pairings \# Pregnancies \# Births ------------------ ------------- ---------------- ----------- C57BL/6J Ethanol 16 13 1\* pair-fed 10 7 5 Control 7 7 4 BALB/cJ Ethanol 16 14 8 pair-fed 11 9 8 Control 6 6 5 DBA/2J ethanol 8 8 6 pair-fed 8 8 5 control 12 11 7 \* p \< 0.05 compared to the C57BL/6J pair-fed and control groups. See text for details. ::: Figure [1](#F1){ref-type="fig"} depicts the number of total errors (summed from day 3 up to day 5). Only the strain origin affected this variable (F~2,96~= 5.0; p \< 0.01). C57BL/6J made fewer errors than BALB/c (z= 3.2; p\< 0.01) and DBA/2 (z= 2.5; p\< 0.05). Treatment had no effect, neither alone, nor in interaction with strain. Figure [2](#F2){ref-type="fig"} shows the running speeds in the radial maze. Similar to the number of errors, only the strain origin affected this variable (F~2,96~= 13.9; p\< 0.001) with again C57BL/6 males differing from the other two strains (vs. BALB/c: z= 5.0; p\< 0.001; vs. DBA/2: z= 4.8; p\< 0.001). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Effects of prenatal alcohol exposure on total numbers of errors over the last three days of training in the radial-maze in male C57BL/6J, BALB/cJ, and DBA/2J mice. n= 11--17 animals per group, except for C57BL/6J prenatally-exposed to ethanol (n= 2) ::: ![](1744-9081-1-5-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Effects of prenatal alcohol exposure on running speeds (cm/s) in the radial-maze in male C57BL/6J, BALB/cJ, and DBA/2J mice. n= 11--17 animals per group, except for C57BL/6J prenatally-exposed to ethanol (n= 2) ::: ![](1744-9081-1-5-2) ::: Body and brain weights are presented in Figures [3](#F3){ref-type="fig"} and [4](#F4){ref-type="fig"}, respectively. The origin of the strain influenced both variables (body weight: F~2,96~= 104.3; p\< 0.001; brain weight: F~2,91~= 86.1; p\< 0.001). BALB/c weighed more than C57BL/6 (z= 3.1; p\< 0.01) and DBA/2 (z= 14.2; p\< 0.001). The latter strain, in turn, weighed less than C57BL/6 (z= 7.2; p\< 0.001). C57BL/6 males had higher brain weights than BALB/c (z= 5.0; p\< 0.001) and DBA/2 (z = 11.7; p \< 0.001). In addition, BALB/c had higher brain weights than DBA/2 (z = 9.1; p\< 0.001). Prenatal exposure to alcohol affected both variables. Body weights were affected independently of the origin of the genotype (F~2,96~= 6.6; p\< 0.01) with mice exposed to ethanol prenatally weighing less than control (z= 3.5; p\< 0.01) and pair-fed (z= 3.2; p\< 0.001) animals. The effect on brain weight depended on the background of the strain (strain × treatment: F~4,96~= 3.8; p\< 0.01). This interaction effect was mainly caused by the BALB/c strain where prenatal alcohol decreased brain weight (vs. pair-fed: z= 3.6; p\< 0.001; vs. control: z= 3.1; p\< 0.01), whereas no significant effects were seen in the other two strains. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Effects of prenatal alcohol exposure on body weight (g) in adult male C57BL/6J, BALB/cJ, and DBA/2J mice. a: significantly different from controls (P \< 0.05); b: Significantly different from controls (P \< 0.01) and pair-feds (P \< 0.001). n= 11--17 animals per group, except for C57BL/6J prenatally-exposed to ethanol (n= 2) ::: ![](1744-9081-1-5-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Effects of prenatal alcohol exposure on brain weight (mg) in adult male C57BL/6J, BALB/cJ, and DBA/2J mice. a: significantly different from controls (P \< 0.01); b: Significantly different from controls (P \< 0.01) and pair-feds (P \< 0.001). n= 11--17 animals per group, except for C57BL/6J prenatally-exposed to ethanol (n= 2) ::: ![](1744-9081-1-5-4) ::: The results of the hippocampal data are presented in Table [2](#T2){ref-type="table"} and Figure [5](#F5){ref-type="fig"}. Strain effects were observed for the following variables: regio inferior (F~2,73~= 8.0; p\< 0.001), stratum lacunosum-moleculare (F~2,73~= 18.9; p\< 0.001), stratum radiatum (F~2,73~= 5.6; p\< 0.01), suprapyramidal MF (F~2,73~= 6.0; p\< 0.01) and IIPMF (F~2,73~= 132.0; p\< 0.001). DBA/2 showed a smaller regio inferior than C57BL/6 (z= 3.7; p\< 0.001) and BALB/c (z= 2.9; p\< 0.001). C57BL/6 showed a larger stratum lacunosum-moleculare than DBA/2 (z= 6.1; p\< 0.001) and BALB/c (z= 4.3; p\< 0.001) while BALB/c males had a larger stratum lacunosum-moleculare than DBA/2 males (z= 2.4; p\< 0.05). C57BL/6 exhibited smaller stratum radiatum and suprapyramidal MF than DBA/2 (z= 3.3; p\< 0.01 and z= 3.4; p= 0.001) and BALB/c (z= 2.4; p\< 0.05 and z= 2.8; p\< 0.01). C57BL/6 showed larger IIPMF sizes than DBA/2 (z= 16.2; p\< 0.001) and BALB/c (z= 10.8; p\< 0.001). BALB/c had larger IIPMF sizes than DBA/2 (z= 7.0; p\< 0.001). ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Sizes of hippocampal fields in the nine different groups Regio inferior (hilus + CA3) in 10^3^μm^2^, other hippocampal fields as percentage of regio inferior. Values represent means ± SEM. For statistical details, see text. ::: Strain/Treatment Regio inferior Stratum oriens Stratum pyramidale Stratum radiatum Stratum lacunosum-moleculare Hilus Suprapyramidal Mossy Fibers ------------------ ---------------- ---------------- -------------------- ------------------ ------------------------------ ----------- ----------------------------- C57BL/6J Ethanol 753.9 ± 0.4 35.0 ± 0.05 15.2 ± 0.6 24.6 ± 0.1 8.1 ± 0.4 8.3 ± 0.9 8.7 ± 0.1 Pair-fed 775.4 ± 24.3 33.6 ± 0.8 14.7 ± 0.4 26.1 ± 0.4 7.9 ± 0.2 9.4 ± 0.2 8.5 ± 0.2 Control 747.8 ± 30.6 33.6 ± 0.5 14.8 ± 0.2 25.6 ± 0.6 8.3 ± 0.4 9.7 ± 0.3 8.3 ± 0.2 BALB/cJ Ethanol 726.9 ± 18.5 34.5 ± 0.6 14.7 ± 0.2 26.6 ± 0.2 6.5 ± 0.2 8.4 v 0.3 9.3 ± 0.4 Pair-fed 742.9 ± 27.7 33.6 ± 0.9 15.2 ± 0.3 26.1 ± 0.5 7.2 ± 0.3 7.9 ± 0.3 9.6 ± 0.4 Control 690.3 ± 23.9 34.3 ± 0.9 14.3 v 0.3 26.6 ± 0.2 6.8 ± 0.2 8.8 ± 0.3 9.3 v 0.3 DBA/2J Ethanol 638.0 ± 25.7 34.1 ± 0.5 15.0 ± 0.3 26.3 ± 0.3 6.6 ± 0.3 8.2 ± 0.3 9.7 ± 0.3 Pair-fed 649.9 ± 21.1 32.7 ± 0.6 15.3 ± 0.5 27.0 ± 0.3 6.0 ± 0.2 8.9 ± 0.4 10.0 ± 0.4 Control 699.6 ± 26.0 33.5 ± 0.5 15.2 ± 0.3 27.1 ± 0.5 6.4 ± 0.3 8.5 ± 0.2 9.4 ± 0.3 ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Effects of prenatal alcohol exposure on the sizes of the IIPMF terminal fields (percentage of regio inferior) in male C57BL/6J, BALB/cJ, and DBA/2J mice. For all groups n = 5, except for C57BL/6J prenatally-exposed to ethanol (n= 2). ::: ![](1744-9081-1-5-5) ::: Prenatal alcohol exposure did not affect any of the hippocampal variables, neither as main factor, nor in interaction with the background. Discussion ========== These data demonstrate that, in three inbred strains of mice (C57BL/6, BALB/c, and DBA/2), prenatal exposure to alcohol does neither affect spatial memory nor hippocampal neuroanatomy. Prenatal alcohol exposure did influence body and brain weight in some strains and dramatically reduced live birth rates in C57BL/6 animals. A first caveat is, of course, the observation that the low number (2) of C57BL/6 pups prenatally exposed to ethanol precludes any strong conclusions for that strain. We still decided to include these animals in the present report because simple visual inspection of the data shows that even for these two lone survivors there is not even a trend towards any differences in behavior or hippocampal morphology, just as is the case for the other two strains. Of course, sample sizes are much more adequate for strains BALB/c and DBA/2, so that the conclusions based on those strains are much stronger. In short, our results on both behavior and neuroanatomy appear not to be in line with those, for instance, summarized in Berman and Hannigan \[[@B11]\], who concluded that prenatal alcohol exposure consistently produced significant deficits in spatial learning and/or memory. A closer look, however, revealed that most animal studies on the effects of prenatal alcohol exposure have been performed in rats and not mice. For instance, a survey of the recent literature, using PubMed, revealed only few studies that used mice as experimental models for prenatal alcohol exposure, and, to our knowledge, no studies involved multiple (\>2) inbred mouse strains in the analysis of brain-behavior relations with regard to prenatal alcohol exposure. We could only find one study \[[@B12]\] that included two inbred strains (C57BL/10 and DBA/1), which reacted differently to early alcohol exposure. For instance, open field activity was decreased in C57BL/10, but not in DBA/1 mice, whereas aggression was more affected in DBA/1s. Most mouse studies, however, investigated the effects of prenatal alcohol exposure on various neurobehavioral aspects in only one strain, namely C57BL/6 mice. Thus, Opitz et al. \[[@B13]\] intubated pregnant C57BL/6 females with alcohol from gestational day 14--18 and investigated whether this affected radial maze performance in their offspring. As was the case in the present study, no effect of prenatal alcohol exposure on radial maze performance was found and it might, therefore, be concluded that radial maze performance is not affected by prenatal alcohol exposure in mice. One should keep in mind, though, that there might be other reasons why such an effect is not observed. One possibility is that blood alcohol levels were low in this study, an argument also raised by Opitz et al. in their discussion of their data \[[@B13]\]. Since we decided not to disturb pregnancies by taking blood samples (the rationale being that prenatal stress would then be a confounding factor), we were not able to determine blood ethanol concentrations (BAC). Hence, \'our\' BACs might have been too low to have an effect. However, brain weights as well as body weights were affected in these experiments by the exposure to prenatal alcohol, a finding reminiscent of Wainwright et al. \[[@B14]\]. In addition, live births were dramatically reduced in C57BL/6, but the two male pups that we did obtain were phenotypically completely normal. It should furthermore be noted that the alcohol concentration used in the present study was almost three times as high as those used by previous authors \[[@B15],[@B16]\], who reported significant effects on BACs (35-100mg/dl) in their female mice. A more recent study in which ethanol concentrations were slowly increased to 10% over pregnancy showed even higher BACs in the mother (varying from 50 to 150 mg/dl) \[[@B17]\]. Taken together these findings make it very unlikely that not sufficient ethanol reached the developing embryos to have significant effects. Another explanation for the lack of an effect on learning might be that this type of radial maze is not sufficiently demanding and that other spatial learning tests would be more appropriate to detect prenatal alcohol effects. However, Wainwright et al. \[[@B14]\] did not observe any apparent effects of prenatal alcohol exposure on any measures of performance in a water navigation task in an F2 cross between C57BL/6 and DBA/2 males. Hence, other types of learning tests, less spatial ones such as the puzzle box \[[@B18]\], might perhaps reveal prenatal alcohol effects. In this respect it is interesting to note that in adult C57BL/6 mice prenatal alcohol exposure weakens the efficacy of reinforcers \[[@B15]\], impairs the development of conditioned taste aversion \[[@B13]\], and enhances the sensitivity to amphetamine \[[@B16]\]. We would also like to point out that the radial-maze task used here is quite sensitive and has been used successfully in the past to show learning defects that were not, or only barely, detectable in a water navigation task \[[@B19]\]. An alternative explanation for the absence of effects could be species-specific characteristics with mice being less vulnerable to alcohol per se than rats. It should be noted, however, that other treatments affecting hippocampal mossy fiber projections, such as early postnatal hyperthyroidism, have similar neuroanatomical (and behavioral) effects in rats and mice \[[@B20],[@B21]\]. A final possibility is that effects depend on the age of testing. For instance, there might be a specific time window, in which the effects become visible. Either the effects might be transient or they might only appear at a later age (see \[[@B11]\] for a discussion). Another striking result of this study is the low delivery rate of C57BL/6J females. Out of 13 pregnancies only one female gave birth to a viable litter. The other twelve pregnancies resulted in still-born pups, or the mother died while giving birth, or the pups were eaten by their mothers immediately after birth. Whether this finding reflects a higher sensitivity in developing C57BL/6J embryos to alcohol or higher blood alcohol concentrations in the mother or a combination of both cannot be inferred from these experiments. Although the only litter born might not be a representative sample, males from this litter appeared to perform equally well in the radial maze as their pair-fed and control counterparts; neither were the sizes of the IIPMF terminal fields affected. Conclusion ========== Summarizing, in this experimental design, which used three distinct inbred strains of mice, prenatal alcohol exposure influenced neither radial maze performance nor the sizes of the IIPMF terminal fields. We believe that future research should be pointed either at different targets when using mouse models for FAS (e.g. more complicated behavioral paradigms, different hippocampal substructures, or other brain structures) or involve different animal models, including zebrafish \[[@B22]\] or guinea-pig \[[@B23]\]. Methods ======= Animals ------- Subjects were male mice from the inbred strains C57BL/6J, BALB/cJ, and DBA/2J that are known to differ in their sensitivity to ethanol (e.g. \[[@B24]\]). The mice were kept under controlled laboratory conditions: temperature 24 ± 2°C; 12:12 light-dark schedule with lights on at 8:00 AM; room; food (IU UAR) and tap water ad libitum; dust-free sawdust bedding. Animals were weaned at 29 ± 1 days and each male was housed with a female (preferably a littermate) in a Plexiglas cage (42 cm × 27 cm × 17 cm). The experiments were performed at the University René Descartes (Paris V; CNRS URA 1294, Génétique, Neurogénétique et Comportement). All experimental animals were born and raised in the Paris animal facilities, which were SPF and approved by the French Ministry of Agriculture. Prenatal Alcohol Exposure ------------------------- Eight days before mating, female mice of all three stains were habituated to alcohol solutions as their only source of liquid (4% ethanol the first 4 days, 8% ethanol the last 4 days). Females were then mated with experienced males from the same strain and, throughout gestation, exposed to a 12% ethanol solution as their only source of liquid. Two control groups were included: first, a pair-fed group which consumed an isocaloric solution of dextrin (Amisol) and a same amount of food as consumed by the alcohol-exposed dams and, second, a control group which received food and tap water ad libitum. Although food and liquid consumption were not specifically measured, no obvious differences in consumption were noticeable. It has been shown that voluntary drinking paradigms lead to significantly elevated blood alcohol levels in pregnant females \[[@B17]\]. Pregnancy was confirmed by visual inspection of females. After parturition offspring remained with their biological mother and tap water replaced the alcohol solution. Radial Maze Training -------------------- At the age of 3--4 months animals were tested in the radial maze. The 8-arm radial maze used in these experiments was similar to the original one used by Schwegler and Crusio (e.g. refs \[[@B25],[@B26]\]). The central part of the radial-maze measured 20 cm in diameter. Its arms (25 cm long, 6 cm high, 6 cm wide) were closed and made of transparent Plexiglas. At the end of each arm was a perforated partition behind which fresh food pellets were deposited. In this way, the animals could not smell the presence or absence of a reward. All arms were reinforced by placing a small food pellet behind a low barrier preventing the animal from seeing whether a specific arm was still baited or not. The maze was always oriented in space in the same way. Several extra-maze cues were provided close to the arms. A confinement procedure was used utilizing transparent guillotine doors at the entrance of each arm. The doors were lowered and kept closed for 5 seconds after animals returned to the central box. The radial maze was placed directly on the floor to avoid possible elevation-induced anxiety. Twenty-four hours prior to the experiment animals were moved to the test room. Animals were habituated for 1 day and subsequently trained for 5 days. The habituation consisted of a 15-min exploration trial with free access to all arms but without a food reward. Immediately afterwards companion females were removed from the home cages and all experimental animals, now single housed, were deprived of food. During the training sessions animals were weighed daily and kept at 80--90% of their original body weight. In between sessions the maze was cleaned with a dry cloth. On the first two days, trials were terminated after 15 minutes or after the animal had eaten all rewards, whichever came first. Thereafter, no time limit was imposed and trials were terminated when animals had found all 8 food rewards. The situation of animals not eating all rewards occurred frequently on the first two days, but never on days 3 to 5. For this reason, data from days 1 and 2 were not included in the analyses. Previous experiments have shown that significant learning occurs very rapidly and that strain or mutational effects can reliably be shown with this method \[[@B27]\] Two variables were sampled, one representing learning performance and the other running speed. Learning was measured by the number of errors while activity was depicted as the mean distance traveled (cm) per second. An error was counted if an animal entered an arm previously visited or did not eat the reward. Average running speed was estimated by dividing the distance traveled by the amount of time needed to complete a trial. Histology and Morphometry ------------------------- Histological treatments were performed as previously described by Schwegler and Lipp \[[@B28]\] see also \[[@B29]\]. Briefly, mice were deeply anesthetized and perfused intracardially with sodium sulfide and glutaraldehyde. This method allows a good fixation and preparation of the tissue for Timm\'s stain. Brains were removed, weighed, and post-fixed 24 hours in 3% glutaraldehyde with 20% sucrose and subsequently cut horizontally in 40 μm cryostat sections after which Timm\'s silver sulfide staining was applied. Methods used for visualization and measurements of the hippocampal terminal fields were similar to those described previously (e.g. \[[@B30]\]). Sampling started at the midseptotemporal level, directly below the most ventral extension of the septal pole of the fascia dentata. Taking every second section, 5 defined horizontal sections per animal were pseudo-randomly sampled, alternating between the left and right hippocampus. Areas of the strata oriens, pyramidale, radiatum, lacunosum-moleculare, and the mossy fiber terminal fields (hilus, suprapyramidal MF, and intra- and infrapyramidal MF; see Figure [6](#F6){ref-type="fig"}) were measured on an image analyzing system (Samba, Alcatel) and were expressed as a percentage of the whole regio inferior (hilus + CA3) to correct for possible slight variations in cutting plane or tissue shrinkage. This standardized method has been shown to yield reliable and replicable results \[[@B28]\]. ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Diagram of a Timm-stained cross-section of the hippocampus. The hippocampal subregion CA3-CA4 (the area of morphometry) is indicated in black, stippled, and hatched areas. Black areas: suprapyramidal (SP), intra- and infrapyramidal (IIP) and hilar (CA4) mossy fiber terminal fields originating from the dentate gyrus. Stippled area: strata oriens (OR) and radiatum (RD). Hatched area: stratum lacunosum-moleculare (LM). CA1, subregion of the hippocampus without mossy fibers; FI, fimbria hippocampi; FD, fascia dentata; OL and ML, outer and middle molecular layers of the fascia dentata; SG, supragranular layer; GC, granular cells. ::: ![](1744-9081-1-5-6) ::: Statistical analysis -------------------- Using χ^2^-tests the number of breeding pairs used (pairings), pregnancies, and births were compared for each strain. Pregnancies were compared relative to the number of pairings, and births relative to the number of pregnancies. The radial maze data of days 3 to 5 were analyzed using two-way repeated measures ANOVAs with strain and treatment as main factors. Both between subjects factors consisted of three levels (strain: C57BL/6J, BALB/cJ, and DBA/2J; treatment: alcohol exposed, pair-fed, and control). The hippocampal data were analyzed by means of two-way ANOVAs, both factors being identical to the above-mentioned analysis. When necessary, pair-wise comparisons were made using least square means. All ANOVAs were performed using the SAS GLM procedure. Authors\' Contributions ======================= FS participated in the interpretation of the data and drafted the manuscript, LJ carried out the radial maze tests and performed the histology, JYB participated in the design of the study and carried out the morphometrical analyses, and WEC conceived of the study, participated in its design, carried out the statistical analyses, and participated in the interpretation of data and drafting of the manuscript. All authors read and approved the final manuscript. Acknowledgements ================ Supported by a grant (no 92/02) from the Institut de Recherches Scientifiques sur les Boissons (IREB, Paris, France) to WEC.	PubMed Central	2024-06-05T03:55:57.480812	2005-4-22	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143778/", "journal": "Behav Brain Funct. 2005 Apr 22; 1:5", "authors": [ { "first": "Frans", "last": "Sluyter" }, { "first": "Laure", "last": "Jamot" }, { "first": "Jean-Yves", "last": "Bertholet" }, { "first": "Wim E", "last": "Crusio" } ] }
PMC1143779	I am pleased to introduce \'Globalization and Health\', a peer reviewed, open access (free to the end user) journal. In this, the début editorial, I will briefly outline the purpose and scope of this journal highlighting our intention to publish a balanced mixture of opinion on the subject. That the journal be \'Open Access\' is entirely appropriate. Knowledge, at its best utility, is a \'public good\' i.e. non-rival, non-excludable. While this journal will deal with the subject matter of creating \'global public goods for health\', it will also by virtue of its very existence, contribute toward that process. Globalization and Health\'s\'Open Access\' policy changes the way in which articles are published. First, all articles become freely and universally accessible online, and so an author\'s work can be read by anyone at no cost. Second, the authors hold copyright for their work and grant anyone the right to reproduce and disseminate the article, provided that it is correctly cited and no errors are introduced \[[@B1]\]. Third, a copy of the full text of each Open Access article is permanently archived in an online repository separate from the journal. Globalization and Health\'sarticles are archived in PubMed Central \[[@B2]\], the US National Library of Medicine\'s full-text repository of life science literature, and also in repositories at the University of Potsdam \[[@B3]\] in Germany, at INIST \[[@B4]\] in France and in e-Depot \[[@B5]\], the National Library of the Netherlands\' digital archive of all electronic publications. Importantly, the results of publicly funded research will be accessible to all taxpayers and not just those with access to a library with a subscription. As such, Open Access could help to increase public interest in, and support of, research. Note that this public accessibility may become a legal requirement in the USA if the proposed Public Access to Science Act is made law \[[@B6]\]. Added to this, a country\'s economy will not influence its scientists\' ability to access articles because resource-poor countries (and institutions) will be able to read the same material as wealthier ones (although creating access to the internet is another matter \[[@B7]\]). The rationale behind starting a journal dedicated to globalization and health is three fold: Firstly: Globalization is reshaping the social geography within which we might strive to create health or prevent disease. The determinants of health -- be they a SARS virus or a predilection for fatty foods -- have joined us in our global mobility. Driven by economic liberalization and changing technologies, the phenomenon of \'access\' is likely to dominate to an increasing extent the unfolding experience of human disease and wellbeing. Secondly: Understanding globalization as a subject matter itself needs certain benchmarks and barometers of its successes and failings. Health is one such barometer. It is a marker of social infrastructure and social welfare and as such can be used to either sound an alarm or give a victory cheer as our interconnectedness hurts and heals the populations we serve. And lastly: In as much as globalization can have an effect on health, it is also true that health and disease has an effect on globalization as exemplified by the existence of quarantine laws and the devastating economic effects of the AIDS pandemic. A balanced view would propose that the effects of globalization on health (and health systems) are neither universally good nor bad, but rather context specific. The extent to which individual states are able to engage the process of globalization on their own terms differs widely from one country to the next. Child mortality, for example, changes quickly in response to subtle changes in purchasing power in impoverished communities. In affluent communities however, a small change in income has little effect on utility in either direction. As we consider the effects of globalization on wellbeing it becomes apparent that we need to consider both the long term scenarios for populations as a whole, and the immediate effects for the more vulnerable within those populations who are dependent on fragile local economies. If the dialogue pertaining to globalization is to be directed or biased in any direction, then it must be this: that we consider the poor first. Competing interests =================== The author(s) declare that they have no competing interests.	PubMed Central	2024-06-05T03:55:57.483204	2005-4-22	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143779/", "journal": "Global Health. 2005 Apr 22; 1:1", "authors": [ { "first": "Greg", "last": "Martin" } ] }
PMC1143780	Editorial ========= Over the last 50 years there has been a steady convergence in the probability of survival and the causes of death between the wealthiest countries of the world and most living in low-middle income countries. This represents a triumph for global health and development. It is a demonstration of what should be possible for the 1 billion people living mainly in sub-Saharan Africa and South Asia. For them, convergence seems way off. And survival is seriously compromised by a combination of diseases classically associated with underdevelopment and the more recent emergence of HIV/Aids. For almost 5 billion people globalization has been associated with increased access to knowledge and technologies that improve life\'s prospects. The reduction of fatalism has been replaced by a growing global awareness that the \"right to health\" can become a reality when governments, nongovernmental organizations and industry play their part. Global awareness among people and governments in wealthiest countries about the need to reduce suffering for all has led to many new initiatives in just the last few years. These include new funding mechanisms to address HIV/Aids and vaccines, a global push for debt relief and better trade opportunities for the poorest countries, and recognition of how global norms that address food safety, infectious diseases and tobacco benefit all. This journal will encourage debate and dialogue about how progress can be accelerated so as to reduce health differences in survival and quality of life that are amenable to policy and operational interventions. It will also stimulate discussion about how new and continuing threats to all can be prevented through multi-sectoral and international actions. We will encourage an exchange of views on how the global architecture for health governance needs to changes in the light of global threats and opportunities. Over the relatively short period of the last 5 to 8 years, many new players have emerged on the global scene with additional funding and their own sets of priorities for investing in health and health research. Simultaneously, major United Nations players like the WHO, World Bank and UNICEF have changed their focus. We will stimulate discussion about how best to enhance the prospects of improving health at the local and national level through better global governance, and hope for an exchange of ideas that will be innovative and helpful to the process of improving global health. Competing interests =================== There are no competing interests related to this work. But in the interests of full disclosure, Derek Yach is funded at Yale to develop community based chronic disease prevention research models internationally by Novo Nordisk.	PubMed Central	2024-06-05T03:55:57.483917	2005-4-22	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143780/", "journal": "Global Health. 2005 Apr 22; 1:2", "authors": [ { "first": "Derek", "last": "Yach" } ] }
PMC1143781	Review ====== Introduction ------------ The global trading system today speeds unprecedented volumes of product to unparalleled numbers of markets throughout the world. Coincident with its growth has been the unwelcomed emergence and dissemination of new infections in human populations. This article will examine linkages between these two phenomena through the optic of \"stealth\" infections -- emergent infections with long, silent incubation periods. When such pathogens are carried in a globally traded product (vector product), spectacular geographic amplification can occur prior to the onset of clinically recognizable illness. Examples of these trade-related emerging infections caused by vector products include variant Cruetzfeldt Jacob disease (vCJD) and HIV/AIDS. Vector products can be defined as those products that carry pathogens. It should be noted, however, that \"vector\" in infectious disease parlance usually refers to living organisms, such as mosquitoes. The use of the term here with regards to product is expedient, albeit non-conforming to normative use. The drivers for broad marketing of the \"vector products\" in these two pandemics were 1) promotion and growth of global demand for beef and animal feed and 2) promotion and growth of global demand for human anticoagulant factors for medical treatment. These market drivers provoked a marked \"ramping up\" of production in pace and capacity for the vector product. The production of these vector products involved pooling of large amounts of biological material. In each case novel pathogens emerged and were disseminated over a broad geographic area prior to their clinical detection. Mad Cow Disease and vCJD ------------------------ Global trade trends for the products implicated in the transmission of HIV and vCJD reflect similar scales of increase in volume and dollar amounts during the decade prior to the detection of human diseases. The incubation period of vCJD is variable, estimated at up to 15--30 years \[[@B1]\]. The incubation period of BSE (Bovine Spongiform Encephalopathy), or mad cow disease, in cattle is also on the order of years, which has complicated culling interventions \[[@B2]\]. Mad cow disease was first detected in British herds in 1986. Wide-scale trade embargoes did not occur until 2000, although some products, such as bull semen, were embargoed as early as 1996, when BSE was first linked to human vCJD \[[@B3]\]. In the case of beef and bone meal, the suspected source of BSE, exports of the implicated product doubled in 1989, from approximately 15,000 tons in 1988 to more than 30,000 tons (Figure [1](#F1){ref-type="fig"}) \[[@B4]\]. The surge in exports was at least partly due to precautions taken as a result of concerns about potential disease risk. Meat and bone meal (MBM), a side product of animal slaughter, has long been appreciated as a source of protein for animal feed. In fact, it is speculated that BSE was introduced in cattle when scrapie-infected material from sheep was pulverized into cattle feed years ago \[[@B5]\]. Thus when BSE surfaced in the mid-1980s, the UK banned the use of ruminant-derived meat and bone meal in ruminant diets in 1988. However, the feed was still permitted for non-ruminants. The ban caused the price of the British MBM feed to drop, making it more attractive for international trade. Once the feedstuffs moved through Europe, they were used for a variety of purposes, including feed for ruminants; unlike in the UK, in continental Europe, bone meal was a legal commodity for all animal feeds. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### UK Exports of Meat-and-Bone Meal, in tonnes, 1985 -- 1991\. \ legal for swine and poultry feed at the time; above scenario also replicated in other countries. Source: Butler, 1996\[4\]. ::: ![](1744-8603-1-3-1) ::: Meanwhile, the value of UK beef exports more than doubled between 1985 and 1995 (Figure [2](#F2){ref-type="fig"}) \[[@B6]\]. A superficial glance at the data would suggest a ramping up of production; however, the UK beef industry, like most industries, is more complex. In fact, as of 1995, the production of beef in the UK had been static for two decades at about one million tons annually \[[@B7]\]. But consumption of beef in the domestic market was declining, partly associated with changing lifestyles. In 1980 Britons consumed 21.3 kg per capita; in 1993 this fell to 17.0 kg per capita. In response, consumer pricing for beef declined, while production and distribution costs remained largely static. Consequently, trade had, by 1995, become an important factor, representing 21 to 23 percent of the total market \[[@B7]\]. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### UK Beef Exports, in 1,000 U.S. Dollars, 1970 -- 1999. Source: Food and Agriculture Organization of the United Nations \[6\] ::: ![](1744-8603-1-3-2) ::: Prices fell both in response to declining domestic consumption and the implementation of the General Agreement on Tariffs and Trade (GATT) reforms, which proposed liberalization of agricultural trade through reduced border protection, domestic support payments and export subsidization \[[@B7]\]. The GATT agreements committed economies to reduce tariffs on imports, and in effect, there was price pressure to be competitive to keep their market share at home, to become more efficient. With margins in the supplying industries squeezed \[[@B7]\], there was market pressure for consolidation of the previously fragmented production system. Farms of 50 cows or more increased from 6,400 to 9,000 between 1986 and 1991. These changes in processing may have permitted prions to cross the species barrier from ruminant to human. Rendering had been a \"batch\" procedure in Europe and North America, relying on very high solvent extraction for the production of tallow. In the 1970\'s, a continuous system of rendering was introduced that relied on vacuum at relatively low temperatures and produced high quality tallow. This efficient process saved costs to the industry. After the BSE crisis, EU studies of abbattoir practices suggested that the processes in place in the United Kingdom in the 1990\'s were ineffective in deactivating prions \[[@B8]\]. Amplification and Transmission through Trade -------------------------------------------- The geographic distribution of BSE is broad. If in fact the emergence of this agent occurred only in Britain and bone meal and animal feed contamination occurred only from this source in the late 1980\'s it is impressive that OIE (World Organisation for Animal Health) lists only four countries currently confirmed as BSE-free \[[@B9]\]. By 2004, in addition to the UK, 22 countries had notified the OIE of cases of BSE in farmed cattle. Cases have been reported largely in Europe but also in North America and Asia \[[@B9]\]. The UK may not in fact have been the sole origin, given a recent report which suggests France had undetected BSE epidemics on a large scale during the same time period \[[@B10]\]. The BSE/vCJD emergence has also impacted the human blood supply. Based on the experience with HIV/AIDS, blood banks are concerned that an agent that cannot be screened for may enter the global blood supply, amplifying infectious transmission. This concern has been heightened by the occurrence of vCJD apparently related to blood transfusion \[[@B11]\]. Thus exclusion criteria in the United States and blood donation fractionation in the UK reflect measures to mitigate this risk. HIV/AIDS -------- By the end of 2003 UNAIDS estimated that almost 40 million people were living with HIV/AIDS \[[@B12]\]. While human sexual contact is the major global route of infection of HIV, the history of the pandemic also documents the wide-ranging infection due to infected blood products. Again, innovations in product in response to a global market were apparently correlated with amplification and dissemination of product. In the early 1980\'s, when AIDS was described in the first cluster of young gay men in San Francisco, ongoing concern about the risk of hepatitis in pooled blood for manufacture of blood factors had spurred producers to a search for a means of screening their product for hepatitis virus. However, viral deactivation processes were not generalized in the industry when HIV emerged; with a profitable international market, the new lyophilized blood product was globally traded. The commercial sector for the production of blood factors is most developed in the United States. Data from the U.S. Census Bureau attests to a growing trade in blood product from the mid-1970s to the late 1980\'s, from approximately 50 million USD in 1975 to 325 million USD in 1988. We now know that half of the hemophiliacs in the U.S. were infected with HIV through this route, and an untold number of hemophiliacs worldwide. In some countries, such as Japan, this was probably the primary route of entry into the population. By the end of the 1990\'s, there were 400 commercial centers for plasmapheresis operating in the U.S. These centers, which employ paid donors, provided 60% of the worldwide requirement for plasma \[[@B13]\]. By mid 1982 the possible link between AIDS and the blood supply was reported in the CDC\'s Morbidity and Mortality Weekly Report, and was widely known and accepted the following year with the occurrence of cases in hemophiliacs living in geographically dispersed areas. But retooling the fast-growing industry of factor production posed substantial difficulties, for several reasons. 1) The science of HIV/AIDS disease was still early in its evolution, and the data were not therefore clearcut 2) the plasmapheresis industry utilized largely compensated donors (who were often higher risk for HIV/AIDS) as a basis for obtaining plasma, 3) the most heavily impacted group, homosexual males, were wary of discrimination in measures adopted. The initial self-exclusion strategies, for example, did not ask donors about homosexual sexual practice for fear of discrimination. In early 1983 both the voluntary and commercial sectors had taken some measures for reducing the participation of high risk donors in plasmapheresis. But the Inquiry on the Blood System in Canada, published by Health Canada\'s Krever Commission in 1997, notes, \"There is evidence, however, that the voluntary sector refused to stop collecting in high risk areas, though its blood donor recruiting officials no longer targeted high risk individuals, and that the commercial sector also continued to operate in such areas\" \[[@B13]\]. Viral inactivation methods had been in development since the early 1970\'s to try to cut down on hepatitis transmission in blood. However, the industry leaders considered such steps proprietary information for production and so the work towards successful strategies was not shared across the corporate competitors. In 1984 the major producers had all been licensed to distribute heat-treated products to cut down on the threat of hepatitis and AIDS infection. As with any product, blood products can be subject to a recall, which is initiated when the public\'s health is at risk due to contamination of the product. This option was available to the US Food and Drug Administration as soon as the risk of HIV transmission in blood factors became known in March of 1983. An Institute of Medicine study convened years later to review the history was critical of the absence of a cogent, strong recall policy \[[@B14]\]. By 1985 the ELISA screening test for HIV came into use. This, coupled with heat treatment of the factor product, has markedly reduced the risk of HIV transmission in blood. However, global spread of the virus, in part facilitated by the global trade in factor VIII, had already occurred in the 1980\'s. This contributed to the profound global reach of the epidemic seen today. The experience of Japan illustrates this point. Japan, Factor VIII and HIV/AIDS ------------------------------- In 1994, the Tenth International AIDS Conference was held in Japan, focusing on subgroups, such as hemophiliacs, that had previously received less attention \[[@B15]\]. Although Japan had been characterized by some as reluctant to acknowledge the introduction of HIV to Japan via \"sexual tourism,\" the virus did initially hit Japan through the blood supply. According to the World Federation of Hemophilia, the majority of Japanese hemophiliacs are thought to have been affected during 1983--1985 by non-heat-treated factor concentrates imported from the US. In fact, by 1985 the US provided 90 percent of imported factor concentrates in Japan \[[@B13]\], and the Japanese AIDS Research Group reported that the causative agent of AIDS was found in factor VIII concentrates \[[@B13]\]. Of 5,000 hemophiliacs in Japan, approximately 2,000 had been infected with HIV/AIDS by 1997 according to the Krever Report. The occurrence of HIV/AIDS through factor concentrates became known in Japan as \"Yakugai\" AIDS (meaning \"AIDS due to the harmful effect of medicine\"). As factor concentrates were manufactured by pooling plasma from many donors, the entire pool could be contaminated by a single unit from an HIV-infected donor. Dr. Michael Rodell, former vice-president of the Armour Pharmaceutical Company, estimated that four infected persons could contaminate the entire world supply of factor VIII \[[@B13]\]. The Ministry of Health and Welfare in Japan regulates the national blood system by licensing the use of blood products. In 1983 the Japanese AIDS Research Group, the central technical group advising the Ministry on prevention and control policy, reportedly considered a proposal from the US fractionator Travenol to allow the importation of heat-treated (virus-inactivated) factor VIII. At that time, a new product could receive regulatory approval without undergoing clinical trials if it could be classified as a \"change to manufacturing method having no effect on the effective ingredients\" \[[@B13]\]. The Ministry rejected the idea, apparently mistrusting the reliability of the US FDA tests and fearing side-effects from heat-treated concentrates \[[@B13]\]. Thus, while US reports in 1983 showed that HIV could be contracted from contaminated factor concentrates, the Ministry continued the import of non-heat-treated factor concentrates, with the stipulation that they be accompanied by a certificate that they did not contain plasma from donors at high risk of contracting HIV/AIDS \[[@B13]\]. The use of imported non-heat-treated concentrates increased in 1984 and peaked in 1985 \[[@B13]\]. Although heat-treated factor concentrates were approved for use in 1985 -- because the Ministry recommended that physicians continue to prescribe non-heat-treated factor concentrates (and blood product manufacturers continued to distribute them without warning labels about the risk of AIDS) -- many hemophiliacs continued to use non-heat-treated concentrates through 1986 \[[@B13]\]. A further exacerbating problem was the fact that some health care providers neglected to inform their hemophiliac patients of their HIV infection, allowing their sexual partners to become infected \[[@B16]\]. This tragedy in Japan was due to a combination of factors in government, industries, and medical communities \[[@B17]\]. The former chief of the Office of Biologics at the Ministry of Health and Welfare, the head of the AIDS Research Group, and three Green Cross executives were indicted for misconduct and/or professional negligence. Still, the \"Yakugai AIDS\" scandal is not over, and trials continued through 2004. Conclusion ========== The following similarities are thus apparent for infections with very long incubation periods in the modern era of trade and travel: 1) global forces such as market demand or GATT provisions that favor increased exports also favor consolidation, which may set the stage for streamlining processing of product 2) such streamlining of production for efficiency and cost savings may have a role in emergence of new infections when biological materials are the basis for product formulation 3) the emergence of new pathogens that are disseminated through trade and travel creates a \"science gap\" where rapid research is imperative to find and implement agent-specific safety strategies before extensive infection occurs 4) once clinical disease is manifest widespread dissemination of infection has occurred and risk can be mapped using product specific trade data. This mapping may allow timely institution of surveillance. Among the most sobering lessons from these two emergent diseases are the consequences of the division between public and private science and the division between the world of health and the world of trade. When heat treatment processes were developed for factor VIII, they were not shared across the industry in an expeditious fashion. Urgent information on product risk and on safety innovations is crucial to public safety and timely public health intervention in disease. The \"sectors\" of global trade and global health have few working links. Information flow between the international agencies, the World Trade Organization and World Health Organization, while improving, is not routine. At the regional and national level, intersectoral communications between the trade and health communities is also troublesome. The European Union stands in contrast to this generalization, with extensive collaboration in disease surveillance among its members \[[@B18]\]. The regional initiatives on emerging infections of such regional trading groups as the Asia Pacific Economic Cooperation (APEC), ASEAN and others are encouraging. At the national level, formal collaboration between ministries of commerce and health on product investigation or tracing is sparse. British authorities were prescient in their implementation of the \"Offal Ban\" in 1988 \[[@B19]\]. While the risk of vCJD from BSE food animals was not established, they took precautionary measures to limit exposure which were, in retrospect, prudent. At the same time, the issue of potential \"opportunistic exporting\" of animal feed by the UK onto the international market, in response to the bans imposed domestically on such feed, has been raised. If precautionary recall of factor VIII had been fully implemented in 1986, it is clear that hundreds of cases of HIV may have been averted, and the geographic extent of human infection limited. Currently the costs and effectiveness of the \"precautionary principle\" in commerce are under discussion \[[@B20]\]. Accurate and complete description of these two global emergent diseases (through mathematical modeling and other research) will be important in informing those discussions. In summary, two novel infections with long incubation periods, HIV/AIDS and BSE/vCJD, are illustrative in describing the wide geographic dissemination and urgent consequences of emergent infections associated with globalized trade. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= Ann Marie Kimball and Yuzo Arima drafted and edited the manuscript. Jill Hodges reviewed and edited the manuscript. Acknowledgements ================ This research was sponsored by the John Simon Guggenheim Memorial Foundation. We are grateful to the interns at Field Epidemiology Training Program, National Institute of Infectious Diseases, Japan, for assistance with gathering information on HIV/AIDS in Japan.	PubMed Central	2024-06-05T03:55:57.484688	2005-4-22	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143781/", "journal": "Global Health. 2005 Apr 22; 1:3", "authors": [ { "first": "Ann Marie", "last": "Kimball" }, { "first": "Yuzo", "last": "Arima" }, { "first": "Jill R", "last": "Hodges" } ] }
PMC1143782	Review ====== As food production and distribution practices evolve to keep pace with rapidly diversifying consumer demand and international competition, new pathogens are emerging and long-known microbes are expanding their reach. Resilient bacteria such as Salmonella, Escherichia coli, Listeria monocytogenesand Cyclospora cayetanensisinsinuate themselves into fruit, vegetables, poultry, beef and dairy products as they circulate around the globe, generating \"trade-related infections\" \[[@B1]\] (see Table [1](#T1){ref-type="table"}). The pathogens may survive or multiply in foodstuffs and spread to humans and other vertebrates along the way. As soon as food safety measures address one problematic infection pathway, the microbes appear somewhere they have never been detected before. Food and waterborne diseases account for an estimated 2.1 million deaths annually in developing countries, and foodborne disease affects up to 30 percent of the population in industrialized countries \[[@B2]\]. Food safety is a \"farm to fork\" effort \[[@B3]\], and in the modern world of transnational integrated food supply and global trade, the distance between those two points has increased. International commerce has tripled over the last two decades \[[@B4]\]. In 2002, agricultural products accounted for about nine percent of international trade dollars -- worth some \$583 billion \[[@B5]\]. As the global exchange of agricultural products and the accompanying health risks proliferate, international efforts to control trans-border disease transmission are becoming increasingly important. This article will explore emergent and re-emergent foodborne infections that are coincident with the rise in global agricultural trade. Although the recent Bovine Spongiform Encephalopathy (BSE) and variant Creutzfeldt-Jakob Disease (vCJD) epidemics are compelling examples of this dynamic, this review focuses on enteric diseases and will not include prion disease. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### The nexus of global trade and foodborne pathogens. ::: Pathogen Origin Trade-related interaction --------------------------- ------------------------------------------------------------------------------------------------------------------------------------- -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Salmonella Described in the late 1880s in swine. Subsequently recognized in humans, poultry, cattle, rodents and exotic pets. Use of antimicrobials in livestock in response to heightened global competition has contributed to emergence of antimicrobial-resistant strains such as S. typhimuriumDT104 and S.Newport-MDRAmpC. Escherichia coliO157:H7 Identified as a pathogenic agent in humans in 1982. Hosts include cows, deer, sheep, horses, pigs and dogs. Intensified production and far-reaching distribution channels in the meat industry enable widespread dissemination in vehicles such as ground beef. Cyclospora cayetanensis First documented cases observed in humans 1977. Only known host is humans. Hardy oocysts are transported on produce exported to geographic regions where the parasite previously had been largely unknown. Listeria monocytogenes Detected in 1926 in rabbits and guinea pigs, identified as a source of human infection in 1929 and perinatal contamination in 1936. Increased popularity on the global market of raw milk cheeses and ready-to-eat products contributed to surge in listeriosis. ::: Salmonella: New Pathways and Strains ------------------------------------ While Salmonellais among the longest-known and most common foodborne pathogens, the Salmonellosis outbreaks stemming from the growing global exchange have revealed areas in which current knowledge is limited or outdated. Salmonella, traditionally linked primarily with poultry, eggs, raw meat, and dairy products, recently has been associated with a growing number of nuts, vegetables and fruits. Meanwhile, Salmonellahas demonstrated an unanticipated hardiness and increasingly is emerging in antimicrobial-resistant strains. Salmonellais found worldwide, with different serotypes prevalent in different regions -- making it possible to track the incursion of new strains that may be linked to international commerce. The pathogen readily reproduces in a variety of foods, especially milk, quickly reaching a high infectious dose if the food is not refrigerated. In the right conditions, Salmonellacan persist in the environment for weeks, even months. Infection commonly results from inadequately processed or undercooked eggs, poultry, dairy or meat. But humans also can transmit the bacterium through fecal-oral contact, and fruits and vegetables can be infected by contaminated water, work surfaces and utensils. Salmonellahas caused massive outbreaks of illness. In 1994, an estimated 224,000 people across the U.S. developed Salmonella enteritidisinfections after consuming ice cream apparently contaminated during transport in tankers that previously had carried nonpasteurized eggs \[[@B6]\]. The cross contamination highlights the tenacity of Salmonellaand the low doses required for infection. ### New Pathways The mechanisms of contamination of newly identified sources of Salmonellainfection are in some cases surprising, and in others, poorly understood. Among the previously unsuspected vehicles are seeds destined to become sprouts. During one of the many steps in their production, from growth to harvest to shipment, the seeds can be contaminated, primarily by contact with animal feces. The dry conditions under which the seeds are stored enable Salmonellaand other pathogens to survive for months. All the while, the seeds sheltering the bacteria appear unharmed. During sprouts\' warm, moist germination process, the microbes thrive. Sprouts have been linked to a series of outbreaks across North America, Western Europe and Japan \[[@B7]\]. A 1995 outbreak of SalmonellaStanley, for instance, spanned 17 U.S. states and Finland. The infection ultimately was traced to a distributor in the Netherlands who had obtained alfalfa seeds from Italy, Hungary and Pakistan; investigators were not able to determine which source was implicated in these outbreaks \[[@B8]\]. The contamination pathway in the first known case of an outbreak associated with imported mangoes is particularly paradoxical. In 1999, at least 78 people in 13 U.S. states became ill from a common strain of Salmonella enterica; 15 patients were hospitalized and two died \[[@B9]\]. Investigators traced the mangoes back to a farm in Brazil. They discovered that surprisingly, no Europeans who had consumed mangoes from the same farm had fallen ill. Investigators deduced that the mangoes destined for the U.S. likely had absorbed the microbe during a hot water treatment to repel fruit flies. The treatment was required to meet U.S. standards barring produce carrying the Mediterranean fruit fly -- standards the Europeans did not impose. ### Antimicrobial resistance While overall rates of Salmonellahave been dropping in the U.S. since 1996, the rates of drug-resistant strains have been on the rise. More than a quarter of Salmonellaisolates are resistant to at least one antimicrobial; a significant portion has multiple resistances\[[@B10]\]. Infections resulting from these strains are not only difficult to treat because of their resistance to drugs, they also can cause more serious illnesses and hospitalizations \[[@B11],[@B12]\]. Among the most prominent drug-resistant strains is Salmonella typhimuriumdefinitive phage type 104 (S. typhimuriumDT104), which began appearing with increasing frequency in the 1990s after fluoroquinolones were approved for use in food-producing animals. S. typhimuriumDT104 is unresponsive to a handful of common antimicrobials, including ampicillin, chloramphenicol, streptomycin, sulfonamides and tetracyclines\[[@B13]\]. First detected in the United Kingdom in 1984, it quickly became one of the most commonly reported strains of Salmonellain England and Wales, linked with consumption of chicken, pork sausage, meat cakes, and eventually, beef \[[@B14]\]. While isolated cases DT104 appeared in the U.S. in the early 1980s, by the mid-1990s, the pathogen had become widespread \[[@B15]\]. More recently, a drug-resistant strain of SalmonellaNewport, Newport-MDRAmpC, has emerged in the U.S. Newport-MDRAmpC is resistant or less susceptible to at least nine antimicrobials, including those on DT104\'s list and cephalosporins, which often are used to treat children with serious cases of salmonellosis\[[@B16]\]. Although most drug resistant infections in people have resulted from the use of antibiotics in human medicine, another contributing factor is antimicrobial applications in food producing animals. Once the bacteria in the animals develop resistance to the drugs, those new strains of resistant bacteria can in turn be transmitted to humans through contaminated meat, soil, and water \[[@B17]\]. Concerns over antimicrobial disease transmission have heightened as abattoirs and dairy operations are consolidated and more livestock are confined in closer quarters. The use of antibiotics as growth enhancers is particularly problematic because it entails applying low doses over long periods to large numbers of animals, potentially transforming the livestock into reservoirs for antibiotic resistant pathogens \[[@B18]\]. Mass Production and E. ColiO157:H7 ------------------------------------ Competitive pressures of the global market also have incited the consolidation of food production. Along with the efficiencies of intensified production come increased opportunities for cross contamination -- and significant challenges in tracing the original source of infections. Ground beef from a single cow may be mixed with that of hundreds of other cows at several different stops along the production process, from slaughter to processing to retail packaging. This consolidation is believed to have been a factor in the emergence and spread of diseases such as Escherichia coliO157:H7 \[[@B19]\]. In the winter of 1982, a series of outbreaks of enteric disease in Oregon and Michigan revealed the presence of Escherichia coliO157:H7, a serotype discovered in 1975 that had been identified in a human only once before \[[@B20],[@B21]\]. Investigators tracked the source of the infection to undercooked beef patties served at fast food restaurants and ultimately, to a lot in Michigan that had supplied the outlets with ground beef. Over the next decade, several E. coli0157:H7 outbreaks, mostly tied to ground beef, cropped up across western and Midwestern U.S. In December 1992, more than 500 people in Washington, Idaho, California and Nevada developed the trademark symptoms, and four people died. The E. coli0157:H7 outbreak was associated once again with undercooked, contaminated beef patties served at a fast food restaurant. In response, the restaurant chain recalled more than a million beef patties and recovered about 20 percent, preventing an estimated 800 additional cases\[[@B22]\]. A team from the Centers for Disease Control identified one Canadian and five U.S. slaughter plants as potential sources of the contaminated lots, but they were not able to definitively pinpoint the source. Today E. coli0157:H7 is a significant health concern in a growing number of regions around the world, particularly areas of Europe and North and South America; and South Africa and Japan. E-coliO157:H7 infections increasingly are linked with a range of meat and produce products, from salami to melons to lettuce. In most cases, produce is contaminated by water or soil containing feces, commonly from agricultural run-off. The bacteria can survive several months in standing water or frozen products. In February 2004, three cases of E. coli 0157:H7 infection in Okinawa, Japan were linked to hamburgers made from frozen ground beef purchased at a local U.S. military base. A subsequent investigation revealed the meat had been produced some six months earlier in the U.S. and also may have been responsible for several cases of infection that had been reported in California the previous August and September \[[@B23]\]. A 1991 outbreak traced to apple cider highlighted the microbe\'s ability to endure acidic conditions. Unwashed apples had been pressed in a mill, which then passed on the infection to subsequent batches. Investigators discovered the E-coliO157:H7 survived nearly three weeks in refrigerated, unpasteurized cider \[[@B24]\]. Conspiring to make E. coliO157:H7 an emerging threat in the international marketplace are its virulence and resilience, along with the relatively low doses required for infection, enabling ready transmission and enhancing opportunities for large outbreaks. Further contributing to the risk of spread is the growing list of contaminated produce. U.S.-grown radish sprout seeds, for example, were implicated in a massive outbreak in Japan. In the spring and summer of 1996, more than a dozen clusters of E. coli0157:H7 infection swept through central Japan, resulting in 10,000 cases -- 6,000 among school children and the rest among factory workers \[[@B25]\]. Uneven resources and Unknown Agents: Cyclospora ------------------------------------------------- While global commerce offers the promise of boosting struggling economies by enabling them to participate in the global marketplace, realizing that potential is a complicated proposition. One critical issue is determining whether a nation\'s resources and land are best invested in crops that serve and rely on external markets. Another concern is whether the small growers that traditionally have been the backbone of the agricultural system in developing countries have the basic sanitary infrastructure necessary to develop products that can compete internationally. The situation becomes even more complex when it involves little-known, and consequently unpredictable, microbes such as Cyclospora cayetanensis. When C. cayetanensiswas first detected in humans in the late 1970s, it appeared to be confined mostly to tropical and sub-tropical areas of the developing world, affecting primarily children and people with compromised immune systems. On the occasions that cases of cyclosporiasis appeared in developed countries, it struck travelers who had visited areas where the disease was known to exist -- and who presumably had been exposed to contaminated water. Humans are the only known host of the coccidian parasite, which is transmitted by oocysts excreted in feces that require at least several days outside the host to sporulate and become infective. It would become evident through a series of outbreaks in the 1990s that these properties enable C. cayetanensisto mature into an infectious agent while being transmitted on produce, apparently over long distances and several weeks. In the spring of 1996, the CDC received reports of nearly 1,500 cyclosporiasis cases in the U.S. and Canada \[[@B26]\]. Investigators examined patterns among the outbreaks and determined that virtually all were linked to events at which fresh raspberries had been served. The raspberries, in turn, were traced to Guatemala. Generally, the symptoms of cyclosporiasis don\'t surface for at least a week; consequently by the time the initial outbreaks had been recognized, it was too late to establish the precise source of the contaminated berries. But the far-reaching nature of the outbreak suggested that a common practice among several suppliers, rather than a single farm, was responsible for the contamination \[[@B27]\]. Investigators concluded the most likely source of the infection was contaminated water used in some step of the berry-growing process. In response to the findings, the Guatemalan Berry Commission implemented some measures targeting water and sanitation practices on the farms, and classified individual growers according to risk. But the efforts failed; the berry season in 1997 was a repeat of the year before -- the CDC received reports of 41 clusters involving more than 1,000 cases from 13 U.S. states, the District of Columbia, and one Canadian province \[[@B27]\]. The investigation once again lead to Guatemala, which then suspended raspberry exports, incurring an estimated \$10 million USD in lost income \[[@B28],[@B29]\]. For the 1998 spring season, the U.S. FDA banned imports of fresh raspberries from Guatemala. That year, Canada, which had not banned the raspberries, once again experienced a series of outbreaks affecting more than 300 people in the Ontario area, but the U.S. did not, further establishing the link to Guatemala as the source of the infections \[[@B30]\]. Along with raspberries, C. cayetanensisinfections have been associated with other produce, including basil, mesclun lettuce and snow peas \[[@B27],[@B31]\]. In only a couple of cases -- one involving basil, the other frozen raspberries -- has the microbe been identified on the produce suspected of causing the outbreak. Similarly, the modes of contamination have been even more elusive. Cyclosporiasis\' week-long incubation period, coupled with the fact that C. cayetanensisoften travels on fresh produce that is long gone by the time the infection is discovered can make it difficult to identify the source and manner of the infections. Adding to the challenge is the fact that very low doses of exposure apparently are required for infection -- consumption of just a couple raspberries can be sufficient\[[@B27]\]. Product Innovations and the Global Palate: Listeria --------------------------------------------------- Increasingly, appetites are bridging borders. Grocery stores feature \"ethnic food\" isles and delis are stocked with luxury imports from around the world, from foie gras to smoked duck. At the same time, food producers are developing \"ready-to-eat\" foods to meet consumers\' demands for convenience. In 2002, processed goods made up nearly half of all agricultural exports \[[@B5]\]. Left behind in the whirl of innovation are food safety regulations drafted before many of these foods became popular, providing an opening for bacteria such as Listeria monocytogenes. The L. monocytogenesbacterium, rare but also relatively dangerous, became a public heath concern in the \'80s, when the illness was linked definitively with foods such as deli meats, smoked fish, fresh soft cheeses and pâté. At the same time, ready-to-eat foods, attractive for their convenience as well as their profit margins, were growing in popularity. But the additional steps entailed in processing present more opportunities for pathogens to be introduced into products. L. monocytogeneswas first identified as a foodborne pathogen in 1953, when a woman who had consumed milk from an infected cow had stillborn twins. For the next few decades, however, it went largely unnoticed until several major outbreaks in the 1980s caught health officials\' attention. In 2000, it was the pathogen most commonly associated with hospitalization in the U.S., and accounted for a third of reported pathogen-related deaths \[[@B32]\]. A study examining the 54 L. monocytogenesoutbreaks reported around the world from 1970 to 2002 found that roughly one third occurred in the U.S. In more than 90 percent of the cases, contaminated meat or dairy products were identified as the source of the infections\[[@B32]\]. The challenges L. monocytogenesentails are particularly evident in the burgeoning world cheese market. In recent decades, the cheese varieties on offer have expanded from several dozen to several hundred, among them myriad soft cheeses and boutique artisanal cheeses -- many of them potential vehicles for L. monocytogenes. The pathogen, which can survive refrigeration, can invade early on in the process and endure in raw cheeses, or reinfect a cheese after pasteurization. In Switzerland, between 1983 and 1987, 122 infections and 34 deaths were linked to Vacherin Mont D\'Or cheese before officials discovered the microbe was lingering on the wooden shelves in the aging cellars, contaminating one batch after another \[[@B33]\]. In the Los Angeles area in 1985, 86 cases of L. monocytogenesinfection linked to raw milk cheese resulted in 29 deaths, including 13 stillbirths and eight newborns \[[@B34]\]. In that instance, the infections were linked to a commercial cheese. But often queso fresco -- a fresh, soft cheese made from unpasteurized or \"raw\" milk -- is produced in private homes, making it difficult for health officials to enforce sanitary regulations. Presently in the U.S., commercially manufactured cheeses must either be made from pasteurized milk or, if they are made from raw milk, be cured for a minimum of 60 days to outlast any remaining pathogens. However, the so-called \"60-day\" rule, developed in the 1950s before many of the cheeses it regulates existed, has come under scrutiny in recent years as it has become evident that a number of pathogens, including Salmonellaand E. coliO157:H7, can withstand the 60-day aging period. Investigators demonstrated that L. monocytogenescan endure for up to 434 days\[[@B35]\]. Further complicating the issue is the fact that the pathogens can survive outside the food product on equipment or storage facilities and contaminate cheeses via that route, rendering the aging process moot. Discussions about modifying the 60-day rule are under way. Meanwhile, regulations require that raw milk cheeses must be labeled as such. International Safety Systems ---------------------------- Clearly domestic food safety, tracing and surveillance systems play a key role in stemming foodborne outbreaks. But the cross-border nature of commerce and thus infections also requires an effective international response. To that end, the World Health Organization has established the Global Public Health Intelligence Network (GPHIN), a web-based system that monitors news reports of infectious disease outbreaks around the world; Salm-Surv, a global network linking laboratories tracking the incidence of Salmonellaand other foodborne diseases; the Global Outbreak Alert and Response Network (GOARN), which provides technical assistance within 24 hours to governments facing potential epidemics; and the International Food Safety Authorities Network (INFOSAN), which enables trans-border collaboration and assistance among food safety officials. While these networks are invaluable, ultimately their effectiveness relies substantially on individual nations\' surveillance and diagnostic capabilities. Meanwhile, the WHO\'s International Health Regulations currently only require notification of outbreaks of cholera, yellow fever and plague. The 50-year-old regulations are being revised to cover all outbreaks of public health significance, including those with the potential to spread beyond borders, such as foodborne diseases. Until they are revised, the regulations provide little protection against the spread of such diseases. The primary vehicles for addressing health matters as they relate to internationally traded goods are the World Trade Organization agreements. The Technical Barriers to Trade (TBT) and Sanitary and Phytosanitary (SPS) agreements address processes and standards for traded products. The SPS covers most potential vehicles for microbe \"hitchhiking\" -- that is, products from farms and fields. Both agreements aim to provide some measure of predictability and reduce discrimination among trading countries by applying common standards to all trading partners. Under the SPS, the recognized sources for these standards are international organizations addressing food, plant and animal safety -- respectively the Codex Alimentarius (\"Food Law\") Commission (run jointly by the World Health Organization and Food and Agriculture Organization of the UN), the International Plant Protection Commission (IPPC) and the Organisation for Animal Health (Office International des Épizooties, or OIE). While the standards are set forth as science based, some critics contend that the science is heavily influenced by the industry groups, such as those that attend Codex Alimentarius Commission meetings in large numbers \[[@B36]\]. Detractors maintain that the WTO and its consulting organizations are dominated by the major trading economies and generally serve the corporate interests of transnational companies rather than those of the public. While this can be debated, it is indisputable that the primary aims of the WTO agreements are to reduce -- not erect -- trade barriers. Conclusion ========== In response to the growing market pressures of global commerce, producers are scrambling to meet the challenge by making more diverse and better products. These constantly evolving dynamics of the global market are rendering existing safety systems outdated and in some cases simply impractical even as they are being adapted. The challenge of assuring the safety of the global food supply is a matter not only for public health but for private sector interests in the food industry. Nonetheless, the study and documentation of the complex changes taking place in food-related infections and the requisite health protections have fallen chiefly to the public sector. While the health and safety measures the World Health Organization and World Trade Organization provide are considerable, they are limited by the respective organizations\' resources and priorities. Global agricultural products trade in 2002 was a \$583 billion enterprise. The relevant industry science is geared toward discovery of new products or processing, quality assurance of specific products, and on occasion, creating an information base for use in standard setting for the industry. In view of conflict of interest, redirecting the scientific enterprise of industry towards investigating its own products seems an unlikely strategy. However, since the health and welfare of the consuming public is a common concern, additional investment of food industry proceeds in epidemiologic investigation, laboratory, and public health at local and global levels would seem a reasonable pathway. Networks such as INFOSAN will likely identify particular areas where arm\'s length investment by industry could help shore up public sector capacity in resource poor economies. Competing Interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= JH reviewed the literature and drafted the manuscript. AMK conceived of this review, surveyed the literature and helped to draft and edit the manuscript. Both authors read and approved the final manuscript. Acknowledgements ================ This research was sponsored by the John Simon Guggenheim Memorial Foundation.	PubMed Central	2024-06-05T03:55:57.487032	2005-4-22	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143782/", "journal": "Global Health. 2005 Apr 22; 1:4", "authors": [ { "first": "Jill R", "last": "Hodges" }, { "first": "Ann Marie", "last": "Kimball" } ] }
PMC1143783	Introduction ============ Whether it is through scientific curiosity, administrative edict or public health planning necessity, most countries have initiated some form of data collection and health surveillance/monitoring systems to provide information on health priorities. In some cases, such as the Bills of Mortality of the London Parishes, these attempts date back well over 300 years \[[@B1]\]. Cause of death statistics for the population of England and Wales have been collected for almost 200 years, and in most developed countries, for at least a century \[[@B2]\]. Further, many developed countries have instituted incidence registers for major diseases of public health importance, such as cancer, or routinely conduct health surveys to measure the prevalence of disease or risk factor exposures \[[@B3],[@B4]\]. In poorer countries, national registration and certification of all deaths is less common, due to the cost of establishing and maintaining such a system, and often the mortality data collected are incomplete and of poor quality \[[@B5]\]. \'Verbal autopsy\' procedures, using structured interviews with the family of the deceased, provide a history of symptoms experienced by the deceased, but translating these into reliable cause of death information for populations has only met with limited success \[[@B6]-[@B9]\]. Moreover, reliable information on the incidence and prevalence of diseases, injuries and risk factors is rarely available in developing countries, and what data are collected, particularly hospital records, are unlikely to reflect the true pattern of disease and injury in the community due to biases arising from the nature of conditions typically treated in hospitals and the ability of sectors of the population to afford tertiary care. As a result, while most countries have some information about prevalence, incidence and mortality from some diseases and injuries, and some information on population exposure to risk factors, it is generally fragmented, partial, incomparable and diagnostically uncertain. Setting health priorities, however, requires, or at least should, information that is comparable, reliable and comprehensive across a wide range of conditions and exposures that cause death or ill-health in a population. The importance of capturing disease burden from largely non-fatal, but prevalent conditions such as depression or musculoskeletal conditions is critical. Substantial resources are usually invested by society to reduce their impact in populations, yet they rank extremely low among causes of mortality, the traditional basis upon which health priorities have been considered. This paper describes a framework (the Global Burden of Disease Study\[[@B10]\]) for integrating, validating, analysing and disseminating fragmentary information on the health of populations so that it is truly useful for health policy and planning. Features of this framework include the incorporation of data on non-fatal health outcomes into summary measures of population health, the development of methods and approaches to estimate missing data and to assess the reliability of data, and the use of a common metric to summarise disease burden from both diagnostic categories of the International Classification of Disease and Injuries, and the major risk factors that cause those health outcomes. The approach has been widely adopted by countries and health development agencies alike as the standard for health accounting, as well as guiding the determination of health research priorities \[[@B11]-[@B14]\]. Global Burden of Disease 1990 Study =================================== The Global Burden of Disease(GBD) Study was commissioned by The World Bank in the early 1990s to provide a comprehensive assessment of disease burden in 1990 from over 100 diseases and injuries, and from 10 selected risk factors, for the world and 8 major World Bank regions \[[@B15]-[@B17]\]. The estimates were combined with research into the cost-effectiveness of intervention choices in different populations to develop recommended intervention packages for countries at different stages of development \[[@B18]\]. The methods and findings of the original (1990) GBD Study have been widely published \[[@B18]-[@B25]\], and have spawned numerous national disease burden exercises. The basic philosophy guiding the burden of disease approach is that there is likely to be information content in almost all sources of health data, provided they are carefully screened for plausibility and completeness; and that internally consistent estimates of the global descriptive epidemiology of major conditions are possible with appropriate tools, investigator commitment and expert opinion. To prepare estimates of the incidence, prevalence, duration and mortality from over 500 sequelae of more than 100 disease or injuries, a mathematical model, DISMOD, was developed for the 1990 GBD Study to convert partial, often non-specific data on disease/injury occurrence into a consistent age description of the basic epidemiological parameters in each Region \[[@B26]\]. To assess disease burden, a time-based metric which measured both premature mortality (years of life lost, or YLLs) and disability (years of life lived with a disability, weighted by the severity of the disability, or YLDs) was used. The sum of the two components, namely Disability-Adjusted Life Years, or DALYs, provides a measure of the future stream of healthy life (i.e. years expected to be lived in full health) lost as a result of the incidence of specific diseases and injuries in 1990. The effect of incident fatal cases (of disease or injury) is captured by YLLs, while the future health consequences, in terms of sequelae of diseases or injuries, of incident cases in 1990 that were not fatal, are measured by YLDs. A more complete account of the index, and the philosophy underlying parameter choices, is described elsewhere \[[@B27],[@B28]\]. DALYs are not unique to the Global Burden of Disease Study. A variant of DALYs was used by The World Bank in the seminal Health Sector Priorities Review study \[[@B29]\], and derive more generally from earlier work to develop time-based measures that better reflect the public health impact of death or illness at younger ages \[[@B30],[@B31]\]. DALYs are a particular (inverse) form of the more general concept of \"Quality-Adjusted Life Years\" or QALYs, proposed by Zeckhauser and Shepard in 1976 \[[@B32]\] and widely used in economic evaluations. Much of the comment and criticism of the GBD Study has focussed on the construction of DALYs \[[@B33]-[@B35]\], particularly the social choices around age-weights and severity scores for disabilities, and relatively little around the vast uncertainty of the basic descriptive epidemiology, especially in Africa, which is likely to be far more consequential for setting health priorities \[[@B36]\]. The results of the study confirmed what many health workers in mental health promotion and injury prevention had suspected for some time, namely that neuropsychiatric disorders on the one hand, and injuries on the other, were major causes of lost years of healthy life, as measured by DALYs. Table [1](#T1){ref-type="table"} summarises the major causes of disease burden worldwide in 1990 from among the 100 or so specific conditions quantified in the Study. The Table also lists the leading causes of premature mortality, as well as disability, as measured by YLLs and YLDs, respectively. Globally, in 1990, the leading causes of childhood diseases (lower respiratory diseases, diarrhoeal diseases, and perinatal causes such as birth asphyxia, birth traumas and low birth weight) were also the leading causes of disease burden, in part because of their concentration at younger ages. Interestingly, depression ranked fourth, ahead of ischaemic heart disease, cerebrovascular disease, tuberculosis and measles. Road traffic accidents also ranked in the top 10 causes of DALYs worldwide. Using more broad disease categories, non-communicable diseases, including neuropsychiatric disorders, were estimated to have caused 41% of the global burden of disease in 1990, only slightly less than communicable, maternal, perinatal and nutritional conditions combined (44%), with 15% due to injuries \[[@B10]\]. The class of infectious and parasitic diseases were the cause of more than one in five (23%) DALYs lost in 1990, followed by neuropsychiatric conditions (10.5%), cardiovascular diseases (9.7%), respiratory infections (8.5%), perinatal conditions (6.7%) and cancers (5.1%). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Leading causes of premature mortality, disability and disease burden, World, 1990 ::: Premature Mortality Disability Disease Burden --------------------- --------------------- ----------------- ------------------ -------------------- ----------------- ----------------- -------------------- ------------------ ---------------- Rank Disease/ injury YLLs (000s) Cumulative % Disease/injury YLDs (000s) Cumulative% Disease/injury DALYs (000s) % of Total 1 Lower res. inf. 108601 12.0 Depression 50810 10.7 Lower res. inf. 112898 8.2 2 Diarrhoeal dis. 94434 22.4 Iron def. anaem. 21987 15.4 Diarrhoeal dis. 99633 7.2 3 Perinatal cond. 82681 31.5 Falls 21949 20.0 Perinatal cond. 92313 6.7 4 Isch. heart dis. 41595 36.1 Alcohol use 15770 23.4 Depression 50810 3.7 5 Measles 36450 40.1 COPD^1^ 14692 26.5 Isch. heart dis. 46699 3.4 6 Tuberculosis 34304 43.9 Bipolar dis. 14141 29.5 Cerebrovas. dis. 38523 2.8 7 Cerebrovas. Dis. 32115 47.5 Congenital anom 13507 32.3 Tuberculosis 38426 2.8 8 Malaria 28038 50.5 Osteoarthritis 13275 35.1 Measles 36520 2.7 9 Road traffic acc. 26162 53.4 Schizophrenia 12183 37.7 Road traffic acc. 34317 2.5 10 Congenital anom. 19414 55.6 Obs.-comp dis^2^ 10213 39.9 Congenital anom. 32921 2.4 SourceMurray and Lopez (10) ^1^Chronic obstructive pulmonary disease ^2^Obsessive-compulsive disorders ::: By and large, the leading causes of years of potential life lost (YLLs) were similar, the major difference being that depression is not a major cause of premature mortality. It is, however, a major cause of non-fatal disease burden, causing more than 10% of all years lived with a disability (YLDs) worldwide, more than twice the contribution from the next leading cause, anaemia (4.7%). Indeed, as Table [1](#T1){ref-type="table"} shows, five of the top 10 leading causes of disability in 1990, as measured by YLDs, were neuropsychiatric conditions. For prevention, comparative estimates of the disease and injury burden caused by exposure to major risk factors is likely to be a much more useful guide to policy action and priorities than a \'league table\' of disease and injury burden alone. Over the past few decades, epidemiologists have attempted to quantify the impact of specific exposures, particularly tobacco, on mortality, either from major diseases such as cancer \[[@B37],[@B38]\], or across a group of countries using comparable methods \[[@B39],[@B40]\]. Specific country studies have examined the impact of several leading risk factors \[[@B41],[@B42]\], but prior to the GBD Study, there was no global assessment of the fatal and non-fatal disease and injury burden from exposure to major health hazards. Ten such hazards (see Table [2](#T2){ref-type="table"}) were quantified in the 1990 Study, based on information about causation, prevalence, exposure, and disease and injury outcomes available at the time. Almost one-sixth of the entire global burden of disease and injury that occurred in 1990 was attributed to malnutrition, another 7% or so to poor water and sanitation, and 2--3% from risks such as unsafe sex, tobacco, alcohol and occupational exposures. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Global burden of disease and injury attributable to selected risk factors, 1990 ::: Risk factor Deaths (thousands) As % of total deaths YLLs (thousands) As % of total YLLs YLDs (thousands) As % of total YLDs DALYs (thousands) As % of total DALYs ---------------------------------------------------------------- ------------------------ -------------------------- ---------------------- ------------------------ ---------------------- ------------------------ ----------------------- ------------------------- Malnutrition 5 881 11.7 199 486 22.0 20 089 4.2 219 575 15.9 Poor water supply sanitation and personal and domestic hygiene 2 668 5.3 85 520 9.4 7 872 1.7 93 392 6.8 Unsafe sex 1 095 2.2 27 602 3.0 21 100 4.5 48 702 3.5 Tobacco 3 038 6.0 26 217 2.9 9 965 2.1 36 182 2.6 Alcohol 774 1.5 19 287 2.1 28 400 6.0 47 687 3.5 Occupation 1 129 2.2 22 493 2.5 15 394 3.3 37 887 2.7 Hypertension 2 918 5.8 18 665 1.9 1 411 0.3 19 076 1.4 Physical inactivity 1 991 3.9 11 353 1.3 2 300 0.5 13 653 1.0 Illicit drugs 100 0.2 2 634 0.3 5 834 1.2 8 467 0.6 Air pollution 568 1.1 5 625 0.6 1 630 0.3 7 254 0.5 Source:Murray and Lopez (10) ::: Improving Comparative Quantification of Diseases, Injuries and Risk Factors: The Global Burden of Disease 2000 Study ==================================================================================================================== The initial Global Burden of Disease Studyrepresented a quantum leap in the global and regional quantification of the impact of diseases, injuries and risk factors on population health. The results of the study have been widely used by government and non-governmental agencies alike to argue for more strategic allocation of health resources to disease prevention and control programs that are likely to yield the greatest gains in population health. Following the publication of the initial study, several national applications of the methods have led to substantially more data on the descriptive epidemiology of diseases and injuries, as well as to improvements in analytical methods. Critiques of the approach, and particularly of the methods used to assess the severity weightings for disabling health states, have led to fundamental changes in the way that health state valuations are determined (population-based rather than expert opinion as used in the 1990 study), and to substantially better methods for improving the cross-national comparability of survey data on health status \[[@B43],[@B44]\]. Better methods for modelling the relationship between the level of mortality and the broad cause structure in populations, based on proportions rather than rates, have led to greater confidence in cause of death estimates for developing countries \[[@B45]\]. Improved population surveillance for some major diseases such as HIV/AIDS, and the wider availability of data from \'verbal autopsy\' methods, particularly in Africa, has lessened the dependence on models for cause of death estimates, although substantial uncertainty still remains in the use of such data. Perhaps the major methodological progress since the GBD 1990 Study has been with respect to risk-factor quantification. In the initial study, the population health effects of 10 risk factors were quantified, but there are serious concerns about the comparability of the estimates. Different risk factors have very different epidemiological traditions, particularly with regard to the definition of \"hazardous\" exposure, the strength of evidence on causality, and the availability of epidemiological research on exposure and outcomes. As a result, comparability across estimates of disease burden due to different risk factors is difficult to establish. Moreover, classical risk factor research has treated exposures as dichotomous, with individuals either exposed or non-exposed, with exposure defined according to some, often arbitrary, threshold value. Recent evidence for such continuous exposures as cholesterol, blood pressure and body mass index suggests that such arbitrarily defined thresholds are inappropriate, since hazard functions for these risks decline continuously across the entire range of measured exposure levels, with no obvious threshold \[[@B46],[@B47]\] For the GBD 2000 Study, a new framework for risk factor quantification was defined which, instead of the classical dichotomous approach, measured changes in disease burden that would be expected under different population distributions of exposure \[[@B48]\] Attributable fractions of disease due to a risk factor were then calculated based on a comparison of disease burden expected under the current (i.e. 2000) estimated distribution of exposure, by age, sex and Region, with that expected if a counterfactualdistribution of exposure had applied. The counterfactual distribution was defined for each risk factor as the population distribution of exposure that would lead to the lowest theoretical minimum levels of disease burden. Thus, for example, in the case of tobacco, the theoretical minimum distribution would be 100% of the population being life-long non-smokers; for BMI it would be 100% of the population having a BMI of 21 (SD1) kg/m^2^, and so on. The theoretical minima for each of the risk factors quantified in the WHO Comparative Risk Assessment (CRA)study (the risk factor arm of the GBD 2000 Study) were developed by expert groups for each risk factor and are described in more detail elsewhere \[[@B49],[@B50]\]. The main findings of the CRA Study are summarized in Table [3](#T3){ref-type="table"}. In all, 26 risk factors were quantified, each by age and sex, and within 14 WHO epidemiological Regions, as well as for the world. These regions were further grouped into \"developed\" \"low-mortality developing\" including China and much of Latin America, and \"high mortality developing\" including Sub-Saharan Africa, and many countries in Western and Southern Asia, including India, Bangladesh and Myanmar. As the table suggests, the world is currently experiencing a \"risk factor\" transition, with developed countries characterized by high disease burden from tobacco, sub-optimal blood pressure, alcohol, cholesterol and overweight. Disease burden in the poorest countries, on the other hand, is primarily caused by underweight, unsafe sex, unsafe water and sanitation, indoor air pollution and micronutrient deficiencies (zinc, iron, vitamin A). Interestingly, the risk factors which, on average, cause the greatest disease burden among the 2.4 billion people living in low-mortality developing countries are a mixture of both, led by alcohol, sub-optimal blood pressure and tobacco, followed by underweight and overweight. This juxtaposition of what might be termed \"new\" and \"old\" risk factors strongly suggests that health policy in developing countries must increasingly address risks such as tobacco and blood pressure that have often mistakenly been labelled, and treated, as conditions of affluence. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Leading risk factors for disease burden in 2000, by development category ::: Developing countries Developed countries -------------------------------------- ---------------------- -------------------------------- ---------------------- High mortality countries % of Total DALYs % of Total DALYs Underweight 14.9% Tobacco 12.2% Unsafe sex 10.2% Blood pressure 10.9% Unsafe water, sanitation and hygiene 5.5% Alcohol 9.2% Indoor smoke from solid fuels 3.6% Cholesterol 7.6% Zinc deficienty 3.2% Overweight 7.4% Iron deficiency 3.1% Low fruit and vegetable intake 3.9% Vitamin A deficiency 3.0% Physical inactivity 3.3% Blood pressure 2.5% Illicit drugs 1.8% Tobacco 2.0% Unsafe sex 0.8% Cholesterol 1.9% Iron deficiency 0.7% Low mortality countries % of Total DALYs Alcohol 6.2% Blood pressure 5.0% Tobacco 4.0% Underweight 3.1% Overweight 2.7% Cholesterol 2.1% Low fruit and vegetable intake 1.9% Indoor smoke from solid fuels 1.9% Iron deficiency 1,8% Unsafe water, sanitation and hygiene 1.8% Source: World Health Organization (46) ::: Improving Cross-Population Comparability of Disease Burden Assessments ====================================================================== While the first Global Burden of Disease Studyset new standards for measuring population health, the basic units of analysis for the study were the 8 World Bank Regions defined for the 1993 World Development Report. Designed to be geographically contiguous, these Regions were nonetheless extremely heterogenous with respect to health development. Other Asia and Islands (OAI) for example, included countries with such diverse epidemiological profiles as Singapore and Myanmar. This seriously limits their value for comparative epidemiological assessments. For the Global Burden of Disease 2000 Study, a more refined approach was followed. Estimates of disease and injury burden were first developed for each individual Member State of WHO (191 in 2000) using different methods for countries at different stages of health development, often largely determined by the availability of data \[[@B51]\]. For example, age-sex-specific death rates for countries were essentially determined using one of three standard approaches: routine life-table methods for countries with complete vital registration; application of standard demographic methods to correct for under-registration of deaths; or, where no vital registration data on adult mortality were available, application of model life tables \[[@B51],[@B52]\]. The detailed methodological approaches adopted for countries to estimate cause-specific mortality, and the descriptive epidemiology of non-fatal conditions in each country are described elsewhere \[[@B53]\]. This focus on individual countries as the unit of analysis, as well as the systematic application of standardized approaches for all countries in any given category of data availability, has vastly improved the cross- population comparability of disease and injury quantification, at least among countries at similar levels of health development. Caution is required, however, in inferring comparability of national disease burden assessments across countries at different levels of development. Estimates of mortality in countries where there is no functioning vital registration system for causes of death will always be substantially more uncertain than those derived from systems where all deaths are registered and medically certified, as is the case for developed countries. For example, in the United States, uncertainty around the mean life expectancy for males in 2000 (73.9 years) was ± 0.3 years, compared to ± 3.5 years in Uganda \[[@B51]\]. The same may be said for the quantification of disability due to various conditions, where the gap in data availability between rich and poor countries is likely to be even more extreme than for mortality. A major advance with the Global Burden of Disease 2000 Studyhas been the systematic attempt to quantify uncertainty in both national and global assessments of disease burden. This uncertainty must be taken into account when making cross-national comparisons, and needs to be carefully communicated and interpreted by epidemiologists and policy makers alike. To date, systematic national estimates of the burden of disease due to major risk factors, applying the standardized framework of the Comparative Risk Assessment Project, have not been attempted. Standardized approaches to measuring mortality attributable to some risk factors, such as tobacco, have been developed and applied to 50 or so developed countries \[[@B39]\], but more research is urgently needed to prepare comparative risk estimates, by country, using the broader, more comprehensive CRA framework. There is no a priorireason to expect that the uncertainties in cross-national comparisons for risk factors would be any greater than those for diseases and injuries that have already been quantified. Discussion and Conclusions ========================== The World Development Report 1993provided an enormous impetus to the development of global and regional quantification of disease and injury burden, and of what causes it. The vast exercise in global descriptive epidemiology that was required to develop estimates led to the first ever comprehensive estimates of the fatal and non-fatal burden for over 100 diseases and injuries, as well as for selected risk factors. The development and widespread application of a single summary measure of population health (DALYs) has greatly facilitated scientific and political assessments of the comparative importance of various diseases, injuries and risk factors, particularly for priority-setting in the health sector, and has led to strategic decisions by some agencies eg. WHO, to invest greater effort in program developments to address priority health concerns such as tobacco control and injury prevention. The subsequent Global Burden of Disease 2000 Study, and a plethora of country applications, have led to substantial improvements in both methods and data availability, as well as in the comparability of results. They have not, however, led to significant changes in the comparative magnitude of most conditions, the single exception being HIV/AIDs, largely as a result of the explosion of the epidemic during the 1990s in Southern Africa. Nor have these methodological advances adequately addressed the challenges that arise from new data sets becoming available. For example, better methods are needed to estimate adult mortality levels from survey data \[[@B54]\], to estimate biases in using hospital data to infer community-level cause of death patterns, and to more reliably quantify the joint effects of multiple risks acting in concert to produce disease outcomes. This relative stability in the outcomes of disease and risk factor quantification does not necessarily inspire greater confidence that the estimates are correct. Rather, it suggests that despite the progress of the past decade, the incremental gains in advancing our knowledge and understanding of global descriptive epidemiology have been modest. There is an urgent need for a globally-coordinated research and development effort to devise and implement cost-effective approaches to data collection and analysis in poor countries that is targeted to their health development needs, and that can routinely yield comparable information of sufficient quality to establish how disease and risk factor burden is changing in populations. Recent calls for the establishment of a global health monitoring Centre to continuously assess, using comparable methods, the impact of diseases, injuries and risk factors worldwide are a step in this direction \[[@B55]\], but much more needs to be done to assist countries with the development of minimal health information systems. It is lamentable how little is reliably known about the global impact of diseases, injuries and risk factors. It would be unconscionable if we were to be similarly ignorant 10 to 20 years hence.	PubMed Central	2024-06-05T03:55:57.489912	2005-4-22	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143783/", "journal": "Global Health. 2005 Apr 22; 1:5", "authors": [ { "first": "Alan D", "last": "Lopez" } ] }
PMC1143784	Global health policy actors =========================== The major actors in global health policy are changing. New actors are entering and old ones are losing power; the overall change has seen a shift from global nation-based health-policy-making structures towards more diversity that puts emphasis on private sector actors. In the 1980s and 1990s there was a shift in global health policy making from the UN agencies towards financial institutions. This shift has meant increasing attention being given to involving private actors in health policy \[[@B1]-[@B4]\]. Towards the end of the 20^th^century the UN increasingly collaborated with business, which subsequently increased the influence of private interests in the UN system. \[[@B5]-[@B8]\]. This development was partly due to the declining levels of development assistance of the OECD (Organisation for Economic Co-operation and Development) countries to the UN, which became particularly acute in the 1990s \[[@B9]\], and partly due to the fear that the UN would become marginalized if it did not increase its collaboration with the corporate sector, which had gained power in overall policy-making \[[@B10]\]. In the UN forums, civil society has become recognized as an important body of actors in global policy-making, as seen at the UN Conference for Environment and Development in 1992, and at the International Conference on Population and Development in 1994, where women\'s organisations were instrumental in shaping the Programme of Action. Regarding health matters, the not-for-profit sectors of the civil society have played an important role for much longer, most notably in the debates concerning essential drugs, breast milk substitutes, and weaning foods in the 1970s and 1980s. \[[@B11]\]. More recently the public health NGOs have been important, for example, in shaping pharmaceutical policies and emphasising the needs and rights of HIV-infected people. The emergence of new global health policy actors -- as a result of new global legally independent public-private entities such as the Global Alliance for Vaccines and Immunizations (GAVI), the Global Fund to Fight AIDS, Malaria and Tuberculosis (GFATM) and the Global Alliance for Improved Nutrition (GAIN) -- to address selected health issues at the turn of the century has further diversified the global health policy scene. Furthermore, new challenges in health research have been defined under the public-private partnership umbrella of the Global Forum for Health Research. Development aid to health has continued to grow substantially since 1992 despite the fall in total official development assistance (ODA) since that time. The USA provides about one third of the total bilateral aid to health. Other bilateral donors are substantially smaller. The multilateral agencies provide one third of the total official development assistance to health and of that assistance 80% comes from the International Development Association (IDA) \[[@B12]\]. As a new funding source, the Global Health programme of the Bill and Melinda Gates Foundation (BMGF) has become not only significant in size, but also in setting health policy. The funding from the USA, IDA and the BMGF are of about the same order. The US role in global health policy setting has increased in the 1990s. \[[@B13]\] Traditionally the US AID emphases have been on fostering goals such as privatization and economic liberation, and on ties to US exports and technical assistance \[[@B14]\]. During the past decade, the USA has been active in lifting global health issues in new forums, such as the G8. The USA was also instrumental in the creation of the GFATM, towards which the EU, for instance, was initially more critical. According to Kagan \[[@B15]\], the US foreign policy is less inclined to act through international institutions such as the UN and less inclined to work co-operatively with other nations to pursue common goals, while the European foreign policy emphasis is on multilateralism over unilateralism. Global health priorities ======================== Global health priorities have in recent years been defined through several processes and by several actors and at various forums. In 2000 and 2001, HIV/AIDS, tuberculosis and malaria came to be discussed in a variety of forums at the UN as well as outside the UN, and commitments to address the three diseases were made, for example, by the G8, the World Bank, the World Economic Forum and the European Commission \[[@B16],[@B17]\]. Millennium Development Goals (MDGs) \[[@B18]\] are a product of consultations between international agencies, but were also adopted by the United Nations (UN) General Assembly in September 2001 as part of the road map for implementing the substantially broader Millennium Declaration, which it had adopted in September 2000 \[[@B19]\]. The MDGs have eight goals, three of which are health-focussed, namely those on child mortality, maternal health, and HIV/AIDS, malaria and other diseases. The UN-led Millennium Project, directed by the economist Jeffrey Sachs, has the objective of ensuring that all developing countries meet the MDGs. The whole UN system has since been requested to adapt to addressing the MDGs, and to report to the Secretary General on their achievements in that direction. For health policies, this has meant, for example, pressures from some of the member states, such as the UK, for the WHO to refocus its work on the MDGs, most notably to the goal concerning HIV/AIDS, malaria and tuberculosis, while its wider mandate as the normative health organisation that sets norms and standards and promotes the building up a wider health systems would not be so emphasised \[[@B20]\]. The MDGs have become an important tool to steer both the UN system towards a narrower agenda with more emphasis on selected interventions and country presences, but more recently increased attention has been placed on the need for addressing development -- including health policy issues and systems -- more comprehensively \[[@B21]-[@B23]\]. Largely the same priorities for health emerged from the report of the Commission of Macroeconomics and Health (CMH) in December 2001 \[[@B24]\], which concluded that public health resources should be directed to the following priorities: communicable diseases; malnutrition, which exacerbates childhood infections; and maternal and perinatal mortality. Development aid for health is also largely steered towards tackling communicable infectious \[[@B25]\]. USAID has financed population programmes, including family planning, for three decades, while its emphasis on health issues is more recent. In 2002, the USAID population, health, and nutrition funding covered HIV/AIDS, family planning/reproductive health, child survival/maternal health, and infectious diseases \[[@B26]\]. The BMGF has provided strategic funding for the founding of new structures for global health policy making -- such as GAVI and GAIN -- and for the implementation of the recommendations derived from the CMH. Its Global Health programme focuses on infectious disease prevention, vaccine research and development, and reproductive and child health, with emphasis on the development and implementation of technologies, though recurrent costs or chronic conditions are not financed \[[@B28]\]. In GAVI, the substantial BMGF funding is targeted at new vaccines. Efforts have also been made to tackle health challenges through new health technology research and development funding under the Bill and Melinda Gates Foundation funded Grand Challenges in Global Health initiative \[[@B29]\]. According to global mortality and burden-of-disease calculations, the above-set priorities indeed represent the majority of deaths and ill-health in sub-Saharan Africa \[[@B27]\], but do not represent the majority of ill-health in any other region. They cover less that a third of the global ill-health \[[@B24],[@B27]\]. Today, non-communicable diseases are a cause of the majority of ill-health in developing countries, and their importance is increasing rapidly. They affect all socioeconomic groups and in many cases the risks are biggest in the poorest sections of the populations \[[@B25]\]. Kickbusch \[[@B13]\] argues that global unilateralism has linked the global health agenda to the US national interests, as well as created a systematic effort to respond to the challenge of the present US administration to show effectiveness. As a result, the four Es -- economics, effectiveness, efficiency, and evidence -- are now the new battle cries for the development community. Selected interventions to eradicate infectious diseases fit well with these premises. The lists of the current global health priorities can be seen as reflecting health-related problems in the developing countries that are perceived to threaten the vital interests of industrialised countries. Linking national interests to development aid is by no means new. In the 1970s, such concerns were central in, for example, the argumentation for population programme implementation \[[@B30],[@B31]\]. Nevertheless, it is noteworthy that since the mid-1990s the arguments for a greater US engagement in global health have been expressed increasingly in terms of national interests or enlightened self-interest \[[@B13],[@B16]\]. The joint strategic plan of the US Department of State and the US Agency for International Development (USAID) for the fiscal years 2004--2009 states that US foreign policy and development policy are fully aligned to advance the National Security Strategy. The strategy sets out its mission as being to create a more secure, democratic and prosperous world for the benefit of the American people and the international community. The purpose of the Strategy is to help American business succeed in foreign markets and help developing countries create conditions for investment and trade \[[@B32]\]. Added emphasis on the trade and industrial policies has been part of global development policies. The eighth MDG is to develop global partnerships for development, which includes developing an open trading and financial system that is rule-based and non-discriminatory in co-operation with both the pharmaceutical sector, for the purpose of providing access to affordable medicines, and in co-operation with the private sector in order to make available the benefits of new technologies. The CMH also argues for increased partnerships with business \[[@B24]\]. Approaches for improved global health ===================================== Health policy-making has become increasingly fragmented and verticalized, with the increasing emphases on selected interventions, the increasing number of partnerships and especially because of the founding of new entities for various health issues. Little emphasis has been put on comprehensive infrastructure building. These trends are in contrast to the stated aims of integrating health policy making with the broader development agenda or with comprehensive health sector planning. An emphasis on innovations and innovative approaches encourages the use of new technologies and the building of new structures. Problems of unsustainability and inequity have arisen with the high levels of funding required, an emphasis on fast results, and the construction of new structures both at global and national levels \[[@B2],[@B33]-[@B35]\]. In the initial faces of GAVI serious concerns were raised that those children that had been without basic vaccine coverage before GAVI funding would remain so and also be out of the reach of the new vaccines \[[@B33],[@B36]\]. The GAVI emphasis on new and more expensive vaccines have raised the costs of the immunizations programmes at country level making the future financing of the programmes highly vulnerable \[[@B37]\]. National priorities often differ from the global priorities, and the thinking around global public goods recognizes this as a starting point. Yamey \[[@B34]\] has argued that the increased emphasis on global programmes and global priority setting is problematic from the point of view of national sovereignty and empowerment. He furthermore states that partnerships rarely synchronise their activities with emerging processes within countries aimed at developing their national health systems. This observation has also been made in relation to GAVI country level action \[[@B38]\]. Partnerships are commonly defined as voluntary and collaborative relationships between state and non-state participants who agree to work together to achieve a common purpose or undertake a specific task, and to share risks, responsibilities, resources, competencies and benefits \[[@B39]\]. According to Richter \[[@B7]\] one of the most substantive losses resulting from the shift towards the partnership paradigm is the loss of distinction between different actors in the global health arena. UN agencies, governments, transnational corporations, their business associations and public interest NGOs are all called \'partner\'. The realisation that these actors have different and possibly conflicting mandates, goals and roles has been lost. The inclusion of business as an integral part of public policy making may weaken the vital role of the public sector in norm- and standard setting and monitoring, as the public sector has been made an equal partner with business, sharing a common purpose and tasks. The WHO collaboration with business has caused harm to the credibility of the WHO\'s normative functions \[[@B7],[@B40]-[@B43]\]. The legally independent global PPPs are structured so that public bodies with normative functions hold seats in the policy-making bodies together with business representatives both at global and national levels. This \'forced marriage\' within the legally independent PPPs may harm not only the credibility of the normative functions of the regulators, but also the normative functions as such. In GAIN and in the UNFPA private sector initiative, the normative bodies are directly requested for \'supportive environments\' as regards regulation, taxes and tariffs \[[@B6]\]. GAVI, GFATM and GAIN deal with essential health issues. Selected UN agencies (in the case of GAIN only one UN or other multilateral agency) that have mandates to deal with these health matters are invited to join their boards either as voting (GAVI and GAIN) or non-voting (GFATM) members, while industry and other private sector actors are included as full members at all levels of their structures \[[@B2],[@B6]\]. The marginalisation of the UN in the structures of the legally independent global PPPs did not happen accidentally. The cautious approach of the WHO to integrating private industry into its activities has been reported as one of the main reasons for GAVI\'s construction as an independent legal body. Problems were encountered, for example, when issues of intellectual property rights and profits arose \[[@B44]\]. According to Phillips \[[@B45]\], the USA opposed the running of GFATM by either the UN or the World Bank. The US also demanded that the fund set up a world-wide aid-delivery system instead of relying on established agencies, such as the UN and the World Bank. According to Stansfield et al. \[[@B46]\] many public sector leaders have raised the concern that in its eagerness to address market failures and pursue international public goods, PPPs are often structured so that the public sector absorbs the lion\'s share of the risks and costs, while the private sector absorbs a disproportionate share of the profit. On a more general note, a report by the International Monetary Fund has raised concerns over the inadequate risk-sharing in public-private partnerships \[[@B47]\]. This tendency can be demonstrated, for example, by the UNFPA private-sector initiative, which aimed at increasing access to reproductive health commodities. According to the initiative, governments were to give preferential tax and duty conditions and ease manufacturing and import regulations, as well as undertake and support market-related research, the donors were to provide support for marketing, advertising and marketing research, while the selected transnational contraceptive producers were requested to sell their products at affordable prices, and handle distribution and implement market-building activities. The initiative also suggested that the governments and the donors could improve the policy environment for private sector investment and security, and facilitate the building of an extensive distribution system so as to reduce the costs for the private sector. Transnational contraceptive producers were instrumental in the selection of the target developing countries, many of which had significant domestic contraceptive production \[48\]. Conclusion ========== While globalisation increases the risk that infectious diseases travel from South to North, it has also increased the risk that major risk factors for non-communicable diseases travel from North to South. Currently, global public health policies are concentrated on selected conditions around infectious diseases and on the technological solutions for them. Addressing infectious diseases in the South is important. However, other health matters are increasingly being left for private actors to deal with. Addressing the most important risk factors of non-communicable diseases, namely tobacco, alcohol and unhealthy foods, would benefit from normative actions, including restrictions on trade and marketing \[[@B25]\]. Simultaneously, global health policy making is increasingly aligned with industrial and trade policies, and is being done hand in hand with business, thus weakening the firewalls necessary for effective regulation and normative actions both at global and national levels. Acknowledgements ================ I would like to thank Mark Phillips for editing the language, as well as the editors and the anonymous reviewers for their comments on the earlier draft.	PubMed Central	2024-06-05T03:55:57.493421	2005-4-22	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143784/", "journal": "Global Health. 2005 Apr 22; 1:6", "authors": [ { "first": "Eeva", "last": "Ollila" } ] }
PMC1143785	Background ========== The National Health Service (NHS) has been suffering the effects of a nursing shortage for the past decade as fewer women train or remain in the nurse workforce, favoring improved job market opportunities in other sectors. The same is true of many other industrialised nations. Over the next 5--10 years, the nurse shortfall is predicted to be 275,000 in the US; 53,000 in the UK and 40,000 in Australia by 2010 \[[@B1]\]. By 2020, the US shortfall may be as high as 800,000 \[[@B2]\]. Efforts to address the shortage include return-to-work initiatives, improved pay, better working environment and flexible hours, and attracting more students to nurse training programs. Even so, the negative aspects of a nursing career discourage many people from training or remaining in the nurse workforce. Perceptions of the NHS as a poor employer are particularly acute and wages remain below other professions, even other jobs within the public sector \[[@B3],[@B4]\]. In this age of globalisation, many countries have turned to overseas recruitment to fill the vacancies caused by a limited or unwilling locally trained workforce. Foreign-trained nurses accounted for 23% of the nurse workforce in New Zealand in 2002; 6% in Canada (2001); 8% in Ireland (2002) and the UK (2001); and 4% in the US in 2000 \[[@B1]\]. And these numbers are likely to grow if domestic hiring continues to lag. In fact, in 2001, more than half of newly registered nurses were trained overseas \[[@B5]\]. Even doctors and pharmacists are being recruited from overseas \[[@B6],[@B7]\]. But foreign recruitment comes at a price, particularly for the source countries. The broadsheets are increasingly reporting on the \'global nursing crisis,\' the \'healthcare brain drain\' from developing countries and the effects the \'nurse exodus\' has on quality of care in poor countries. This is especially true of African nations which not only subsidise countries like England (by investing in the training of healthcare professionals who move to industrialised nations), but suffer from growing nursing shortages of their own, exacerbating problems of existing staff shortages, high infant and maternal mortality rates, the HIV epidemic, poor nutrition and a host of other public health concerns \[[@B8]-[@B10]\]. As one of the world\'s largest importers of healthcare professionals from developing countries, the UK is in need of adopting and enforcing standards of ethical conduct in recruitment. In 1999, the Department of Health issued guidelines on international recruitment intended to curtail poaching from poor countries already suffering from healthcare staffing shortages, such as South Africa and the Caribbean. Unless foreigners provide unsolicited applications or their governments have established programmes for professional development with the UK, the NHS was advised to avoid recruiting \[[@B11]\]. The guidelines were updated in 2001 to include recruitment agencies working for the NHS, but the guidelines do not apply to the independent sector and are not monitored or enforced \[[@B12]\]. Still, the recommendations help explain the strong presence of Filipino nurses who are explicitly encouraged to train overseas (and send income home) as part of the 2001--2004 Medium Term Philippines Development plan. Even the Philippines, however, is starting to feel the crunch of a nursing shortage and trying to find ways of ensuring that sufficient numbers of nurses, particularly nurse educators, are trained and retained in the future \[[@B1],[@B13]\]. As conditions worsen in source countries, many believe that it is neither realistic nor ethical for developed nations to continue relying on foreign recruitment from disadvantaged nations \[[@B8],[@B12],[@B14]-[@B16]\]. If overseas recruitment becomes more difficult, it will have a detrimental effect on vacancy rates in the UK in future, at least in the short-term. (In the long run, perhaps staff shortages even more extreme than today could provide the impetus for major changes that would attract British-trained nurses into the profession. The changing wage structure under Agenda for Change may also help, but it is far too soon to know.) In order to understand that future effect, it helps to study the current and historic impact of overseas recruitment on the NHS. Using data from the Nursing and Midwifery Council (NMC) register from April, 2004, this paper identifies where nurses reside, by country of origin, to see how foreign recruitment affects different Strategic Health Authorities (SHAs) across England. Hypothetical vacancy rates for 2004 are estimated (based on a worst case scenario) as if the UK had no overseas recruits to determine what the shortage would look like across the country if the UK could rely only on its own stock of nurses. Though hypothetical, these numbers help shed light on the extent to which the British workforce has been unwilling to enter or remain in the nursing profession, leading to reliance on foreign sources of labour. 1 Methods & Data ================ The Nursing and Midwifery Council (NMC) maintains the register of qualified nurses, midwives and health visitors for the UK. Because any qualified nurse wishing to work in the UK must register with the NMC, it represents the entire pool of potential nurses that could be recruited into the NHS (overseas residents or retired nurses still registered notwithstanding). The NMC register is updated on a daily basis, containing over 600,000 records. An extract of data from mid-April, 2004, was provided by the NMC, containing counts of all registered nurses by 5-digit postcode by country of qualificiation (where known). While there is certainly error in assigning postcode data to areas within England, the method produces aggregate numbers comparable to the NMC\'s own (unpublished) analysis by reported country of residence. The NMC estimated that, as of the end of March, 2004, roughly 632,000 (or 96%) of the 660,215 registrants resided in the UK, of whom roughly 509,000 were in England, 64,000 in Scotland, 32,000 in Wales and 22,000 in Northern Ireland (with an additional 4,000 in the UK unspecified). These are roughly comparable with the counts from April, 2004, presented here (Table [1](#T1){ref-type="table"}). They found almost 29,000 registrants either reside overseas or did not provide resident country or postcode, which is comparable to the estimated 31,000 unknown or foreign addresses found here and to previous estimates that roughly 5% of the register reside overseas \[[@B17]\]. These numbers are similar to a report on the 2002--2003 data \[[@B18]\]. Because the register is updated on a daily basis, published numbers will not match the current analysis exactly. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Nurse staffing, registration and vacancies by SHA, 2004^a^ ::: NHS Staffing NMC Register Vacancy ----------------------------------- ------------------ ------------------ ------------- ---------- ----- ------ Norfolk, Suffolk & ambridgeshire 12312 15936 21775 73 2.3 296 Bedfordshire & Hertfordshire 6818 8684 15416 56 6.4 441 Essex 6858 8637 13976 62 3.3 230 London North West London 12627 15556 16785 93 6.6 815 North Central London 10501 12553 11379 110^c^ 5.7 592 North East London 9416 11646 13632 85 3.8 332 South East London 11024 13930 16384 85 7.3 811 South West London 7503 9679 14437 67 2.3 170 Northumberland, Tyne & Wear 10365 12455 14588 85 1.1 113 County Durham & Tees Valley 7275 8536 12147 70 1.4 103 NE Yorkshire & N Lincolnshire 7951 9845 17842 55 2.1 174 West Yorkshire 14233 17926 20594 87 1.6 221 Cumbria & ancashire 11490 14226 21246 67 1.3 155 Greater Manchester 17422 20674 25279 82 1.6 289 Cheshire & Merseyside 16276 19730 26948 73 1.4 225 The Southeast Thames Valley 10621 13975 20044 70 2.9 295 Hampshire & Isle of Wight 9424 12564 18370 68 3.6 331 Kent & Medway 7261 9303 14562 64 2.7 195 Surrey & Sussex 12977 17258 28048 62 4.3 547 Avon, Gloucestershire & Wiltshire 12699 16836 23765 71 1.8 220 South West Peninsula 8888 10958 16902 65 0.9 77 Dorset & Somerset 6202 7945 12662 63 0.4 23 South Yorkshire 9138 10783 13355 81 0.9 85 Trent 14050 17333 26817 65 0.8 120 Leics, Northamptonshire & Rutland 7272 9041 15028 60 2.1 167 Shropshire & Staffordshire 7985 9862 16108 61 1.1 89 Birmingham & the Black Country 15288 18343 19620 94 1.9 296 West Midlands South^d^ 7460 9793 15813 62 1.2 86 All London SHAs 51071 63364 72617 87 5.1 2719 Southeast Region^e^ 40283 53100 81024 66 3.3 1368 Rest of England 199982 247543 349881 71 1.7 3421 England TOTAL 291336 364007 503522 73 2.6 7508 Wales 26300 \- 33281 \- 2.1 564 Scotland 39037 41270 66817 6.2 1.1 486 Northern Ireland \- \- 21645 \- \- \- ^a^Nurse staff based on the NHS Workforce Census conducted in September, 2003 by SHA of work (excluding staff of special health authorities or other statutory bodies besides SHAs) \[31\]. Registered nurse population based on April, 2004 NMC data by postcode of residence mapped to SHA. Vacancy data from the 2004 Vacancy Survey for England \[21\], ISD Scotland \[32\] and the Statistical Directorate of the National Assembly for Wales \[33\]. All figures relate to qualified nurses, midwives and health visitors. ^b^NHS headcount divided by \# registered nurses. ^c^Over 100% because fewer nurses reside in North Central London than work there (due to nurses who commute from other SHAs). ^d^Formerly named the Coventry, Warwickshire, Herefordshire & Worcestershire SHA. ^e^Includes SHAs of Thames Valley, Hampshire & Isle of Wight, Kent & Medway, and Surrey & Sussex. ::: To verify country of training, the NMC data was compared with a publication on London-based nurses. A 2002 analysis by Buchan, Finlayson and Gough found that roughly 12% of nurses reporting a London postcode were from overseas, similar to the 10% found here \[[@B19]\]. The lower count may be attributable to one of several reasons. First, a small number of registrants may have been misassigned to a country, or allocated to \'unknown postcode,\' due to partial or inaccurate postcode reporting (which is the only geographic identifier available in this analysis). These cases would not have been assigned to a London Strategic Health Authority. Furthermore, newly registered foreign recruits may list their recruitment agency address initially, introducing error in SHA assignment for newly registered immigrants (though this should disappear upon reregistration). Finally, in previous work done by Buchan analyzing the NMC data, all foreign recruits appeared to enter the UK after 1995, despite known foreign recruitment before then. The loss of data on country of training may be attributable to a change in computer systems used by the NMC to track nurse registrations. But whatever the explanation, it appears the NMC data undercount foreign-trained nurses, excluding those registering prior to 1996 and perhaps undercounting since then as well. So the current analysis focuses on recent recruitment, subject to the limitations of the NMC data itself. (Table [2](#T2){ref-type="table"} shows that of the 67,176 new non-UK admissions to the NMC register since 1993, only 7335, or 11%, registered prior to 1996. Some of them may not have entered the UK or could have left by now. But to the extent they remain in the UK, the majority of foreign recruits will have been captured and properly attributed to their source country in the 2004 data.) ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Newly registered nurses in the UK, 1993--2002^a^ ::: Year UK admissions Non-UK admissions UK as% of all admissions --------- --------------- ------------------- -------------------------- 1990/91 18980 \- \- 1991/92 18269 \- \- 1992/93 18064 \- \- 1993/94 17948 2121 89 1994/95 17411 2452 88 1995/96 16870 2762 86 1996/97 14210 3774 79 1997/98 12082 4300 74 1998/99 12974 4891 72 1999/00 14035 7383 65 2000/01 15433 9709 61 2001/02 14538 16155 47 2002/03 18048 13629 57 ^a^Counts of newly registered trained nurses and midwives on the NMC register (formerly the UKCC register) \[17\]. Data from 1990--1992 only available for UK admissions \[34\]. The year represents all nurses newly registered between April 1 of that year, and 31 March of the following year. ::: Postcode data was mapped to SHA code using the XYZ Digital Map Company\'s PostZon file, mapping 7-digit postcodes to SHAs (based on data supplied by Royal Mail) \[[@B20]\]. For NMC postcodes with multiple SHAs (particularly where only the first two postcode digits were reported in the register), nurse counts were split across SHAs according to the proportion of 7-digit postcodes within the NMC postcode assigned to an SHA. The Department of Health Vacancy Survey began in 1999 to the present, collecting data on vacancy rates in England by occupation code by Trust and Health Authority \[[@B21]\]. The survey asks respondents to report total number of positions that remained vacant for at least three months as of 31 March of each year. The vacancy rate is calculated as the number of openings that have remained vacant for 3 months or more, divided by the number of whole-time equivalent (wte) staff-in-post + number of vacancies. Using cross-sectional data from the NMC and Department of Health for Spring, 2004, hypothetical vacancy rates are calculated assuming all posts held by foreign recruits remained vacant. In other words, if the NHS had not been able to address labour shortages using foreign sources, how much worse might the vacancy problem be today? Assume: 1\. All overseas recruits work full-time within their SHA for the NHS (therefore, \#UK-trained NHS wte staff = \#NHS wte - \#foreign-trained within the SHA); 2\. There were no foreign-trained nurses working in the UK, and; 3\. Wages and job characteristics would be no different today without foreign recruits than they are with (notwithstanding the added stress of higher vacancies, which would probably further increase vacancies). Obviously these assumptions are severe and improbable, but they are needed to present a worst case scenario and to calculate the upper boundary of vacancies in the absence of foreign recruitment. 2 Results & Discussion ====================== Table [1](#T1){ref-type="table"} shows staffing levels (wte and headcounts), NMC register counts, proportion of the register working in the NHS (headcount divided by register count), and vacancy data by SHA in 2004, for all qualified nurses, midwives and health visitors. The register is based on postcode of residence, rather than postcode of work, so the estimated \'proportion working in NHS\' will be biased upwards by commuters coming to work in an SHA (the effect of which may be muted somewhat if resident nurses work in the private sector, commute out of the SHA or do not work, in comparable numbers to the incoming commuters). For example, in North Central London, obviously many nurses commute to and work in the SHA in addition to NHS nurses residing in the SHA boundaries, so more than 100% of resident nurses appear to work there. This commuting bias probably exists for much of the London area. It may also explain the lower proportion of registered nurses in the Southeast region who work within the Southeast SHAs if they travel to London to work (only 66% in the Southeast relative to the national average of 73% of registered nurses working in the NHS). Without data linking nurses to their postcode of work, it is impossible to know whether the low counts in the Southeast are caused by working in the private health sector; commuting to NHS jobs in London or other SHAs; working outside health altogether; or not working. About 25% of NMC-registered nurses are known to work outside the NHS \[[@B22]\], but neither the registry data nor other micro-datasets allow much detailed analysis of what fraction of trained nurses in an area choose to work (at all, or for the NHS, in particular). The job choices of nurses can be studied using data of trained or training nurses in the Quarterly Labour Force Survey from 1999--2003 \[[@B23]\]. Of roughly 1100 qualified or qualifying nurses (in each year of the data), about 64% worked as a nurse and another 17% did not work. Of those working as a nurse, 60% worked full-time, 83% of whom worked in the public sector. While the QLFS does allow this type of calculation at the national level, the sample size is insufficient to provide reliable breakdowns by SHA (with only about 40 nurses per Authority). Table [1](#T1){ref-type="table"} also provides aggregate vacancy rates by SHA. Obviously, vacancy rates are highest in London, which also sees a higher proportion of the registered nurse population working in the NHS (though this statistic is biased upwards by commuters from outside the London area). But these aggregate numbers mask the problem of encouraging British citizens to train and work in nursing, as vacancies are increasingly being filled by foreign-trained recruits. The next step is to determine the extent to which shortage areas rely on overseas recruitment. Table [2](#T2){ref-type="table"} shows a steady decline in nurses joining the NMC register during the 1990s, though it has steadily grown in the past six years. And as UK admissions (i.e. newly registered nurses who trained within the UK) fell, foreign-trained admissions rose. The past ten years have seen a tremendous change in international nurse mobility. This trend is similar for newly registered doctors over the 1990s as non-UK, non-EEA sources increased from 33% of newly registered doctors in 1994 to 44% by 2002 \[[@B8]\]. Table [3](#T3){ref-type="table"} shows the increase in newly registered nurses in the UK from 1998--2003 by source country grouping. The top panel shows foreign recruits from the top 25 non-EU countries, and the bottom panel provides counts of all newly registered nurses. Clearly the top 25 source countries have come to dominate international recruitment, representing 70% of newly recruited nurses in 1998, but 90% by 2003. Much of this increase is attributable to recruitment from the Philippines, which accounted for only 1% of newly registered overseas recruits in 1998, but over 40% by 2000. And while South Africa and India are contributing a growing number of nurses to the UK register, recruitment from the US, Canada, New Zealand and Australia fell significantly over this period. Of the 30,800 foreign-trained recruits known to reside in the UK and on the NMC register in mid-April, 2004, almost two-thirds (19,500) are from Asia; another 7000 from Africa and 2400 from major English-speaking countries (Canada, New Zealand, Australia and the United States). Only a handful of recruits are from Europe (1500) and fewer still from Latin America (300). But overseas recruits are not uniformly distributed across the UK once they arrive. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Newly registered overseas nurses^a^ ::: Country 98/99 99/00 00/01 01/02 02/03 --------------------------------- ------- ------- ------- ------- ------- Philippines 52 1052 3396 7235 5593 South Africa 599 1460 1086 2114 1368 India 30 96 289 994 1830 Australia 1335 1209 1046 1342 920 New Zealand 527 461 393 443 282 Canada 196 130 89 79 52 USA 139 168 147 122 88 West Indies 221 425 261 248 208 Pakistan 3 13 44 207 172 Malaysia 6 52 34 33 27 Singapore 13 47 48 43 25 Nigeria 179 208 347 432 509 Zimbabwe 52 221 382 473 485 Ghana 40 74 140 195 251 Kenya 19 29 50 155 152 Zambia 15 40 88 183 133 Mauritius 6 15 41 62 59 Malawi 1 15 45 75 57 Botswana 4 0 87 100 39 Other^b^ 0 0 0 0 131 Top25 Total 3437 5715 8013 14535 12381 Total foreign-trained 4891 7383 9709 16155 13629 Total UK-trained 12974 14035 15433 14538 18048 Total new registrants 17865 21418 25142 30693 31677 \%Overseas from Philippines 1 14 35 45 41 \%Overseas from Top25 70 77 83 90 91 \%New registrants from overseas 27 34 39 53 43 ^a^Counts of newly registered trained nurses and midwives provided by the Nursing and Midwifery Council based on their register of qualified nurses and midwives. Based on the top 25 non-EU foreign source countries only (top panel); middle and lower sections include counts of all newly registered nurses. Data obtained from NMC and \[5\]. ^b^Includes 23 each from Poland and Sri Lanka; 22 each from the Czech Republic and Saudi Arabia; 21 from Nepal and 20 from Japan in 2002/03. ::: Using the NMC mid-April data, broken down by country of training by postcode, an SHA of residence is assigned based on the reported postcode sector (the 5-digit postcode, assuming a valid or partial postcode is provided and is within the UK). Table [4](#T4){ref-type="table"} presents data on the NMC register, broken down by SHA of resident postcode, by training source (foreign or UK). Scotland has the lowest number of registered foreign recruits as a fraction of all registered nurses with only 1.3%, while England has the highest at 5.5%. SHAs outside of London and the Southeast experienced average foreign representation of only 4.3% (of all registered nurses within the SHA). SHAs with high fractions of foreign recruits (5.5% or more of all registered nurses), are either in London (over 10%) and the Southeast (6.2%) or contain some of the largest cities in England (namely Manchester, Birmingham and Bristol). Leeds (in the West Yorkshire SHA) has a slightly lower foreign presence (5%) with only Liverpool and Sheffield (the Cheshire&Merseyside and South Yorkshire SHAs) among Britain\'s seven largest cities with very low shares of foreign-trained nurses (less than 4.5% of their potential workforce). Unsurprisingly, the SHAs with higher proportions of overseas recruits are also the ones with the highest vacancy rates (Table [1](#T1){ref-type="table"}). Apparently, the worse the nursing shortage, the more active the foreign recruitment efforts (assuming high foreign representation does not drive vacancies, but vacancies drive foreign recruitment). ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Overseas and domestic registered nurses, April 2004^a^ ::: Training Source^b^ ----------------------------------- -------- ---------------------- ------- ------ ------ ------ Norfolk, Suffolk & ambridgeshire 21775 20364 1411 93.5 6.5 4.6 Bedfordshire & Hertfordshire 15416 14305 1111 92.8 7.2 3.6 Essex 13976 13200 776 94.5 5.6 2.5 London North West London 16785 14251 2534 84.9 15.1 8.2 North Central London 11379 10215 1164 89.8 10.2 3.8 North East London 13632 12330 1302 90.5 9.6 4.2 South East London 16384 15093 1291 92.1 7.9 4.2 South West London 14437 13260 1177 91.9 8.2 3.8 Northumberland, Tyne & Wear 14588 14080 508 96.5 3.5 1.7 County Durham & Tees Valley 12147 11867 280 97.7 2.3 0.9 NE Yorkshire & N Lincolnshire 17842 17496 346 98.1 1.9 1.1 West Yorkshire 20594 19569 1025 95.0 5.0 3.3 Cumbria & ancashire 21246 20647 599 97.2 2.8 1.9 Greater Manchester 25279 23882 1397 94.5 5.5 4.5 Cheshire & Merseyside 26948 25758 1190 95.6 4.4 3.9 The Southeast Thames Valley 20044 18461 1583 92.1 7.9 5.1 Hampshire & Isle of Wight 18370 17355 1015 94.5 5.5 3.3 Kent & Medway 14562 13930 632 95.7 4.3 2.1 Surrey & Sussex 28048 26274 1774 93.7 6.3 5.8 Avon, Gloucestershire & Wiltshire 23765 22178 1587 93.3 6.7 5.2 South West Peninsula 16902 16653 249 98.5 1.5 0.8 Dorset & Somerset 12662 12206 456 96.4 3.6 1.5 South Yorkshire 13355 12916 439 96.7 3.3 1.4 Trent 26817 26049 768 97.1 2.9 2.5 Leics, Northamptonshire & Rutland 15028 14546 482 96.8 3.2 1.6 Shropshire & Staffordshire 16108 15536 572 96.5 3.6 1.9 Birmingham & the Black Country 19620 18348 1272 93.5 6.5 4.1 West Midlands South 15813 15275 538 96.6 3.4 1.8 All London SHAs 72617 65149 7468 89.7 10.3 24.2 Southeast Region 81024 76020 5004 93.8 6.2 16.2 Rest of England 349881 334875 15006 95.7 4.3 48.7 England TOTAL 503522 476044 27478 94.5 5.5 89.2 Wales 33281 32022 1259 96.2 3.8 4.1 Scotland 66817 65935 882 98.7 1.3 2.9 Northern Ireland 21645 20591 1054 95.1 4.9 3.4 ^a^Qualified nurses, midwives and health visitors registered with the NMC in mid-April, 2004, known to reside in the UK. All those residing outside the UK or with unknown postcode are excluded (31154 cases). Counts of registered nurses in the Channel Islands or Isle of Mann are included in UK total, but not assigned to any SHAs or country totals within this table. Note: International refers to registered nurses who trained outside the UK. ^b^The \#UK(or foreign)-trained registrants divided by \#registrants residing in the SHA. ^c^Calculated as \#foreign-trained nurses in this SHA divided by total \#foreign-trained nurses on the NMC register known to reside in the UK (30,808). ::: Of the roughly 30,000 foreign-trained nurses residing in the UK, 24% are in the London area and another 16% in the SouthEast (with 49% in the rest of England and the remaining 11% divided among Wales, Scotland and Northern Ireland). Comparable numbers for British-trained registrants living in the UK are 11% in London, 13% in the Southeast and 56% in the rest of England. While there are problems with using NMC registration address to assign the SHA of work, the data can roughly determine the distribution of nurses across England (with some error for commuters and misassigned postcodes), and provide reasonable estimates of the pool of potential nurses located within SHA boundaries. Clearly a disproportionate share of foreign-trained nurses live (and probably work) in the London area, helping to lower London vacancy rates beyond what they would be were international recruitment not possible. Table [5](#T5){ref-type="table"} presents upper bound estimates for hypothetical vacancy rates in the absence of international recruitment. Without any foreign recruits, vacancy rates could be as high as 12% for England, but higher for individual SHAs, particularly in London where aggregate vacancies could rise above 20%. These vacancy rates are three to five times greater than current estimates. Even allowing for Filipino recruitment, vacancies would still be two to three times greater. Of course, given the assumption that all foreign-trained recruits work full-time for the NHS (rather than the independent sector that probably recruited them, for example), these numbers are an upperbound, with more reliable predicted vacancies lying somewhere between current rates and those in Table [5](#T5){ref-type="table"}. And few countries need worry about the existing stock of nurses from the Philippines as they were actively encouraged by their government to work overseas. In future, however, as their domestic shortage worsens, supply may fall, or ethical considerations may prevent continued (heavy) reliance on the Philippines. But for the UK, with or without Filipino nurse recruits, it is clear the labour shortage would be far worse today in the absence of foreign labour sources, particularly (and unsurprisingly) in London and the Southeast (and the Bedfordshire & Hertfordshire SHA). ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Hypothetical Vacancies without Foreign Recruits ^a^ ::: SHA name \%vac^b^ \%vacancyUK^c^ \#vacancyUK^d^ \%vacUK\_Filipino^e^ ----------------------------------- ------------ ------------------ ------------------ ------------------------ Norfolk, Suffolk & ambridgeshire 2.3 13.3 1707 7.4 Bedfordshire & Hertfordshire 6.4 22.5 1552 16.1 Essex 3.3 14.3 1006 9.6 London North West London 6.6 27.2 3349 20.0 North Central London 5.7 17.0 1756 14.0 North East London 3.8 18.7 1634 13.8 South East London 7.3 19.0 2102 15.1 South West London 2.3 18.6 1347 13.1 Northumberland, Tyne & Wear 1.1 5.9 621 2.7 County Durham & Tees Valley 1.4 5.2 383 2.9 NE Yorkshire & N Lincolnshire 2.1 6.4 520 5.0 West Yorkshire 1.6 9.2 1246 5.2 Cumbria & ancashire 1.3 6.4 754 4.5 Greater Manchester 1.6 9.5 1686 7.8 Cheshire & Merseyside 1.4 8.7 1415 6.2 The Southeast Thames Valley 2.9 18.5 1878 11.8 Hampshire & Isle of Wight 3.6 14.5 1346 7.6 Kent & Medway 2.7 11.3 827 7.3 Surrey & Sussex 4.3 18.3 2321 11.4 Avon, Gloucestershire & Wiltshire 1.8 14.4 1807 9.9 South West Peninsula 0.9 3.7 326 2.8 Dorset & Somerset 0.4 8.0 479 4.8 South Yorkshire 0.9 5.8 524 5.1 Trent 0.8 6.2 888 3.8 Leics, Northamptonshire & Rutland 2.1 8.2 649 6.2 Shropshire & Staffordshire 1.1 8.3 661 3.9 Birmingham & the Black Country 1.9 10.2 1568 5.9 West Midlands South 1.2 8.3 624 5.6 All London SHAs 5.1 18.9 10187 14.4 Southeast Region 3.3 15.3 6372 9.3 Rest of England 1.7 9.1 18427 6.0 England TOTAL 2.6 12.1 34986 8.3 Wales 2.1 6.8 1823 3.7 Scotland 1.1 3.1 1368 2.5 ^a^Assumes no foreign-trained nurses ever came to the UK and that existing NHS staffing counts include full employment of foreign-trained nurses (i.e. 100% of international recruits work full-time for the NHS within their resident SHA, and nurses working in the private health sector or not working at all are attributable entirely to the domestically trained population). While subject to measurement error and strong assumptions, numbers represent upper bound on possible vacancies without foreign recruitment (assuming wages and working conditions would not have improved more over the past few years to attract more locally-trained nurses). ^b^Actual 3-month vacancy rate reported by Department of Health. using wte count from March, 2004 (collected in the Vacancy Survey and used in the calculation of vacancy rates). ^d^\#vacUK = \#vacancy (regular, Table 1) + \# foreign-trained nurses. ::: 3 Conclusion ============ There is a growing literature concerning nurse labour shortages and foreign recruitment. This paper identifies where foreign recruits move in the UK and estimates that vacancy rates could be three to five times higher without such a strong foreign-trained presence. This is especially true of the highest vacancy areas, like London and the SouthEast, which could otherwise have double digit vacancy rates (excluding all international recruitment other than from the Philippines). An estimated 100,000 nurses on the NMC register were over the age of 54 and less than 12% (fewer than 66,000) are under the age of 30 \[[@B17]\]. This lopsided age distribution will cause further problems as the 55+ cohort retires over the next ten years. Without overseas recruits to rely on, it is not clear the domestic market can supply the necessary staff. So without major changes, the UK\'s reliance on foreign-trained nurses will continue (in the forseeable future), and nursing is likely to remain a safe career choice for foreigners hoping to emigrate to the UK. Reliance on foreign recruitment (not just in the UK, but the USA and other industrialised nations) poses two important questions for policymakers and researchers. The first of course is how to mitigate the impact of nurse emigration on source countries. Developing countries cannot hope to compete with the higher salaries and better working conditions offered by the UK or other developed economies, leaving their health services (especially in rural areas) with labour shortages of their own \[[@B24]\]. For example, an estimated two-thirds of the Jamaican nurse population has emigrated, leaving Jamaica to fill the void from Cuba \[[@B25]\]. And an estimated 18,000 nurses from Zimbabwe lived overseas in 2002 \[[@B8]\]. Of course some migration, particularly temporary, can be beneficial to source countries as their workforce gains additional skills and experience working overseas and, possibly, sends remittance income home, but what little evidence exists seems to suggest only a small proportion of nurses or other skilled migrants return to their home countries\[[@B8],[@B26]\]. Restricting emigration or taxing leavers (as was common in the 1970s) have a variety of problems and will not eliminate individual workers\' desire to leave \[[@B25]\]. Any long run solution should involve strategies to encourage workers to stay, through better working conditions, improved wages, or other positive inducements. These efforts could have the added benefit of encouraging even more people to train as healthcare providers in developing countries. To better understand and address these problems, further research is needed to quantify the effect of nurse migration on developing countries; estimate what fraction of the nurse workforce emigrates from each country; determine what fraction of emigrating nurses remain permanently overseas or return home (and in what timeframe); and study the effect of various policy options to encourage more nurses to stay in their home countries. So far, data from source countries is limited, and even industrialised nations have little information tracking skilled workforce migrants \[[@B24]\]. The second question to address is how to encourage more people in industrialised nations to train as nurses. Just as developing countries have difficulty competing with developed economies in the nurse labour market, the NHS cannot effectively compete with other employers (in healthcare or other sectors) in the domestic labour market. There are several reasons for the declining number of nurses in the UK, including working conditions, low wages, the cost of living, the changing nature of the job, feeling valued and of course, outside employment opportunities. The labour supply elasticity literature from the UK suggests nurses are relatively unresponsive to wage increases, with a 10% increase in wages leading to an estimated 4% increase in hours worked or 6% increase in the probability of working \[[@B27],[@B28]\]. However, survey data suggests pay does drive decisions (or intentions) to quit. And a comparison of nurses\' wages with those of other nonmanual female workers in the UK suggests nurse wages increased in real terms over the past 20 years, but fell relative to other workers (presumably making other careers more attractive by comparison). Furthermore, geographic variation in vacancy rates across the UK is partially driven by variation in housing costs. This also suggests low wages may be the culprit, since the relatively flat pay structure in the NHS does not adequately adjust wages in high cost areas (there is an adjustment, but it is small compared with London housing costs, for example), driving nurses away \[[@B29]\]. Another reason behind the nursing shortage in England is poor labour force planning in the early 1990s, during which the number of training posts was intentionally reduced. Since 1994, training positions have increased with the Department of Health and now, Strategic Health Authorities, setting the number of training positions needed. However, by failing to take into account demographic trends and increased competition from the private sector (in terms of rising demand from an ageing patient population, an ageing nurse workforce which will need to be replenished, and growth in private sector employment opportunities for nurses), the targets continued to fall short of demand, at least in the 1990s \[[@B30]\]. Ideally, future demand will be met from the domestic labour supply, but this requires better educational planning and improved working conditions and pay to train and retain appropriate numbers of nurses. Additional research is also needed on the labour supply of nurses in industrialised nations. The UK has a good starting point as the Nursing and Midwifery Council registry provides an invaluable resource containing geographic and basic training data for all nurses. With some effort, it could be used to track migration patterns of nurses within the UK as nurses update their information every few years (when they re-register). A long-term strategy might also entail adding employment information to the database, to determine where nurses work, whether in the NHS or private sector, and how far a commute they experience depending on their SHA of employment. And any of these analyses could be broken down by source country of training, to see whether foreigners choose different career paths or locations from the domestically-trained nurse population. Coupled with vacancy and turnover rate data from the Department of Health, more detailed information about the student nurse population (particularly dropout rates and job choice following graduation), and better information about working conditions and wages across Strategic Health Authorities, this data could help the NHS better understand what policies to develop and where to target them (demographically or geographically) in order to improve nurse recruitment within the UK. 4 Competing interests ===================== Financial support from Bristol-Myers Squibb is gratefully acknowledged. I have no competing interests. 5 Acknowledgements ================== I thank Jim Buchan for helpful comments, and members of the Department of Health and the Nursing and Midwifery Council for assistance with data.	PubMed Central	2024-06-05T03:55:57.495663	2005-4-22	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1143785/", "journal": "Global Health. 2005 Apr 22; 1:7", "authors": [ { "first": "Amber S", "last": "Batata" } ] }
PMC1145180	Background ========== The c-myconcogene is implicated in the genesis of many human cancers and accounts for about 70,000 US cancer deaths annually \[[@B1]-[@B3]\]. This oncogene produces the c-Myc transcription factor, which heterodimerizes with Max via the helix-loop-helix-leucine zipper (HLH-Zip) motif to bind specific target DNA sequences and regulate transcription \[[@B4]-[@B8]\]. The amino-terminal region of c-Myc, when tethered to the yeast GAL4 DNA binding domain, behaves as a potent transactivation domain (TAD) \[[@B9]\]. On the other hand, the transactivation potential of native c-Myc appears diminished when compared with other transcription factors, such as the HLH-Zip protein USF1 or to the GAL4 chimeric transactivators \[[@B10]\]. The basis for the diminished transactivation potential of c-Myc has remained elusive despite the discoveries that the Myc activation domain specifically binds to factors such as TRRAP \[[@B7],[@B8],[@B11]-[@B13]\]. TRRAP is a high molecular weight, multifaceted molecule that is capable of recruiting the histone acetyltransferase GCN5 \[[@B12]\]. The fact that c-Myc is able to transactivate in Saccharomyces cerevisiaeand that yeast Tra1 is similar to TRRAP suggest that Myc\'s ability to transactivate in yeast may involve Tra1 \[[@B14],[@B15]\]. The c-Myc TAD encompasses two conserved regions, termed Myc Box I and Myc Box II. Although Myc Box I does not appear to affect transactivation, Myc Box II is required for interaction with TRRAP. Although deletion of Myc Box II renders Myc defective in binding TRRAP, it does not affect the ability of Myc to transactivate specific promoter-reporter constructs and in particular it does not affect the ability of GAL4-Myc chimeric protein to transactivate \[[@B9]\]. In addition, deletion of Myc Box II appears to affect the induction of certain endogenous target genes but not others \[[@B16],[@B17]\]. These observations suggest that transcriptional regulation by Myc is likely to be manifold, involving chromatin modulation as well as direct interaction with components of the basal transcriptional machinery \[[@B18]-[@B20]\]. This spectrum of activities allows Myc to regulate subsets of genes that are more tightly controlled and susceptible to chromatin modulation, whereas other genes, such as the so-called \"housekeeping\" genes, may already exist in open chromatin configuration and hence may be regulated through recruitment of the basal transcriptional machinery. Searches for the interaction of c-Myc with components of the transcriptional machinery have uncovered an interaction with the coactivator CBP \[[@B21]\]. The C-terminal region of Myc has been found to interact with SWI/SNF5 and Miz-1, both implicated in transactivation and transrepression activities of Myc \[[@B22]-[@B27]\]. However, these activities could not account for the transactivation potential of the Myc N-terminal region (TAD). The interaction between Myc and the TATA binding protein (TBP) has been observed in diverse systems with evidence from intracellular chemical crosslinking, mammalian two-hybrid assay, yeast two-hybrid assay, and GST fusion protein pull-down assays \[[@B28]-[@B33]\]. In fact, two recent studies suggest that the Myc TAD is consisted of a structureless N-terminal (Myc~1-88~) portion connected by a linker followed by a C-terminal (Myc~92-167~) partly helical domain, such that the two domains are induced to adopt a specific conformation upon binding TBP \[[@B31],[@B32]\]. On the basis of these observations, we sought to determine in this study whether TBP is limiting for Myc transactivation in transient transfection experiments. We sought to characterize the functional interaction of the Myc TAD with TBP using the chimeric GAL4-Myc fusion proteins as well as full-length Myc with four model promoters (adenoviral major late promoter (AdML), lactate dehydrogenase A (LDHA), CDK4 and ornithine decarboxylase (ODC)) \[[@B34]-[@B37]\]. We found that addition of a TBP expression vector in the transactivation assays increases the transactivation by c-Myc from several fold to well over 30-fold \[[@B38]\]. By contrast, a GAL4-USF1 transactivator did not respond to increasing input TBP. We also observe different responses by promoters that contain initiator (InR) sequences versus promoters that only contain TATA elements \[[@B39]-[@B41]\]. Furthermore, a Myc Box II point mutation W135E does not affect the ability of GAL4-Myc fusions to synergize with TBP \[[@B42]\], but the deletion mutant D106-143 has a blunted effect with cotransfected TBP. Our findings not only support a functional interaction between c-Myc and TBP, but they also provide a means to improve transient transfection assays to study c-Myc target genes. Results ======= TBP is limiting for GAL4-Myc transactivation -------------------------------------------- To determine whether TBP is limiting, we titrated the GAL4-Myc transactivation assays with increasing amounts of TBP plasmid. Although the GAL4-Myc chimera has significant transactivation activity under the experimental conditions chosen, addition of increasing amounts of TBP plasmid resulted in a corresponding increase in transactivation of GAL4-luciferase reporter (G5TATALuc) from about 10 fold to over 60 fold with the highest amount of TBP plasmid used (Fig. [1](#F1){ref-type="fig"}). TBP itself does not affect the basal activity of G5TATALuc reporter, indicating that the effect of TBP requires the presence of the GAL4-Myc chimera. Further, the GAL4 DNA binding domain (GALO) with minimal transactivation activity was not affected by increasing TBP. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### TBP stimulates GAL4-Myc transactivation. The reporter G5TATALuc contains five GAL4 binding sites followed by a minimal TATA box and the luciferase cDNA. Activator alone: the reporter was cotransfected with GAL4(1-147) DNA binding domain (GALO), GAL4 fused to Myc residues 1-262 (GM1-262), GM1-262 with residues 106--143 deleted (GM1-262D106), or GM1-262 with a substitution of tryptophan 135 to glutamate (GMW135E) alone. TBP at increasing input plasmid amounts (μg indicated on the abscissa) was cotransfected into CHO cells with the reporter alone (TBP) or with the indicated GAL4 plasmids. Bars are shown as averages with standard deviation (n = 10). ::: ![](1471-2091-6-7-1) ::: Since Myc Box II is necessary for interaction with TRRAP, we sought to determine whether mutations in this region affect the response of the Myc TAD to TBP. While the deletion D106-143, which removes critical residues of Myc BoxII, activates the reporter better than wild-type Myc TAD, this deletion renders Myc TAD non-responsive to increasing input TBP (Fig. [1](#F1){ref-type="fig"}). A substitution mutation in Myc Box II, W135E, which was previously shown to have diminished interaction with TRRAP and diminished transformation activity \[[@B42],[@B43]\], has a robust response to increasing TBP. These results suggest that the entire region comprising residues 106--143 is required for synergy with TBP, whereas the transformation defective W135E mutant still responds to increasing input TBP. GAL4-Myc synergy with TBP is dependent on the TATA element ---------------------------------------------------------- Since TBP was shown to be limiting for TATA-dependent but not InR-dependent transcription \[[@B44]\], we sought to determine whether GAL4-Myc cooperation with TBP requires the TATA element. We compared the response of a GAL4 dependent reporter that is driven by an InR element (G5INRLuc) with one that is driven by a TATA element (G5TATALuc) (Fig. [2](#F2){ref-type="fig"}). In contrast to the TATA driven reporter, the InR-driven reporter responded more briskly to GAL4-Myc but the reporter activity did not further increase with TBP. This observation suggests that the Myc TAD activates InR dependent transcription, for which TBP is not limiting. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### TBP does not stimulate initiator driven luciferase reporter G5INRLuc that contains five GAL4 binding sites. Either G5INRLuc or G5TATALuc (see Fig. 1) was cotransfected into CHO cells with GM1-262 alone or with increasing amounts of TBP (μg indicated on the abscissa). For comparison, data points for G5TATALuc are the same as those shown in Fig. 1. Note that TBP does not stimulate GAL4Myc transactivation of G5INRLuc; there is a slight inhibition by TBP. Bars are shown as averages with standard deviation (n = 4). ::: ![](1471-2091-6-7-2) ::: TBP is not limiting for USF1 TAD -------------------------------- To determine whether the effect of TBP is selective, we compared the ability of the USF1 TAD with the Myc TAD to synergize with TBP. USF1 is a transcription factor that also belongs to the HLH-Zip family and either homodimerizes or heterodimerizes with USF2. The USF family members binds to DNA target sites (5\'-CACGTG-3\') that appear indistinguishable from Myc targets, though Myc is capable of binding additional non-canonical sites. To eliminate the effects of the USF1 and Myc DNA binding domains on transcription, we sought to study chimeras of GAL4 with USF1 or Myc TADs. Although the GAL4-USF1 chimera activates G5-TATA-Luc as effectively as GAL4-Myc, TBP did not further increase the activity of GAL4-USF1 (Fig. [3](#F3){ref-type="fig"}). This observation suggests that while USF1 and Myc may have overlapping target DNA binding sequences, their TADs appear distinctly different, particularly in response to the addition of TBP. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### TBP does not stimulate the activation of G5TATALuc by GAL4USF1. For comparison, the stimulation of G5TATALuc by GAL4Myc (GM1-262) and TBP is shown. By contrast increasing amounts of TBP (μg indicated on the abscissa) does not increase reporter activity that is stimulated by USF TAD. Bars are shown as averages with standard deviation (n = 4). Plasmids were transfected into CHO cells. ::: ![](1471-2091-6-7-3) ::: Effects of TBP on full-length wild-type and mutant c-Myc transactivation of the LDHA promoter --------------------------------------------------------------------------------------------- Having observed a TATA-dependent synergy between TBP and the Myc TAD in the chimeric GAL4 system, we sought to determine whether TBP could stimulate Myc transactivation of target gene promoters. We first chose the LDHA promoter that comprises two canonical E-boxes located about 100 and 200 bp upstream of the transcriptional start site \[[@B36]\]. We used an amount of input MLV-LTR-driven Myc expression vector that only minimally increased LDHA promoter activity to study the effects on increasing input TBP. TBP increased wild-type Myc activity from only about 20 % to about 800 % increase (Fig. [4](#F4){ref-type="fig"}). TBP alone, without added Myc, slightly increased reporter activity to about 2-fold. Note that the empty MLV LTR plasmid caused a slight increase in reporter activity, which was not accentuated by the addition of TBP. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Activation of the lactate dehydrogenase A promoter construct (LDHLuc) by Myc is increased by TBP. Activator alone: cotransfection with LDHLuc and MLV-LTR driven Myc expression vectors: MLV, empty expression vector; Myc, wild-type Myc; dHLH, helix-loop-helix deletion mutant D371-412; W135E, substitution mutant with glutamate replacing tryptophan 135. Increasing amounts of TBP (μg indicated on the abscissa) was cotransfected with Myc and reporter constructs. Bars are shown as averages with standard deviation (n = 10). Transfections were performed using NIH 3T3 cells. ::: ![](1471-2091-6-7-4) ::: We also studied two Myc mutants in the context of additional TBP (Fig. [4](#F4){ref-type="fig"}). The Myc dHLH mutant lacks the helix-loop-helix domain, and therefore neither dimerizes with Max nor binds DNA. The dHLH mutant minimally affects basal promoter activity and was not affected by increasing input TBP. The W135E mutant contains a substitution in the Myc Box II domain that renders Myc deficient in transformation \[[@B42],[@B43]\]. Intriguingly, W135E was active and fully responsive to increasing TBP. Because transient transfection reporter assays are confounded by many factors, we sought to assure that the synergistic transactivation of the LDHA promoter by Myc and TBP is dependent on Myc binding. We compared the response of the wild-type LDHA promoter (LDH Luc) with one in which both E-boxes are mutated (LDH DM Luc) so that Myc could not bind these sites (Fig. [5](#F5){ref-type="fig"}). The mutant LDH DM Luc displayed no increase in reporter activity with increasing amounts of TBP in contrast to the wild-type LDH Luc. This result confirms that the synergy between TBP and Myc is dependent on the Myc binding sites in the LDHA promoter. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Myc DNA binding sites in the LDHA promoter construct (LDHLuc) is required for TBP stimulation of Myc-mediated transactivation. Wild-type LDHLuc or a mutant promoter construct (LDHDMLuc), which has both Myc E-boxes mutated from 5\'-CACGTG-3\' to 5\'-CCCGGG-3\', were cotransfected with a constant amount of MLV-LTR driven wild-type c-mycplasmid and increasing amounts of TBP (μg indicated on the abscissa). Bars are shown as averages with standard deviation (n = 6). Transfections were performed using NIH 3T3 cells. ::: ![](1471-2091-6-7-5) ::: The effect of the initiator (InR) element on the synergy between Myc and TBP ---------------------------------------------------------------------------- We have previously studied the adenoviral major late (AdML) promoter as a model Myc responsive promoter that contains both InR and TATA elements \[[@B37]\]. Others and we have shown that c-Myc regulation of the AdML is biphasic with transactivation followed by transrepression at high levels of input Myc plasmids \[[@B37],[@B45]\]. The transactivation phase depends on E-boxes, whereas the transrepression phase depends on the InR. Here we chose the AdML promoter to determine the effect of the InR on the synergy between Myc and TBP. As compared with the LDHA promoter, which increased in activity with increasing input TBP, the AdML promoter briskly increased activity in response to Myc alone but did not further increase with TBP (Fig. [6](#F6){ref-type="fig"}). Mutation of the InR element in AdML promoter renders it much less responsive to Myc, but addition of TBP resulted in a marked increase in promoter activity. These observations suggest that the InR increases promoter response to Myc alone, but the InR renders the promoter independent of increasing TBP. Hence, it may be surmised that the combination of E-boxes and InR may be optimal for promoters that require sharp response to Myc regulation, since TBP would not be limiting. ::: {#F6 .fig} Figure 6 ::: {.caption} ###### The initiator element in the adenoviral major late promoter luciferase construct (pGLMLP Luc) increases the response to Myc but does not allow further stimulation by TBP. Compared with the reporter containing a mutant InR (pGLMLP dINR Luc), which displays a highly synergistic stimulation by Myc and TBP, the wild-type pGLMLP Luc construct briskly responds to Myc independent of increasing amounts of TBP (μg indicated on the abscissa). Bars are shown as averages with standard deviation (n = 10). Transfections were performed using NIH 3T3 cells. ::: ![](1471-2091-6-7-6) ::: Synergy of Myc and TBP with CDK4 and ODC responsive sequences ------------------------------------------------------------- We selected the CDK4 promoter, which contains four canonical E-boxes and no InR element, and the intronic ODC tandem E-box sequence to determine the extent of synergy between Myc and TBP. The CDK4 promoter responded to Myc as previously reported. Increasing input TBP further caused a 30-fold activation of the CDK4 promoter as compared to 5-fold activation with Myc alone (Fig. [7](#F7){ref-type="fig"}). The synergy between Myc and TBP is dependent on Myc binding, since the E box mutant CDK4 promoter is not responsive to the combination of Myc and TBP. ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Activation of the human CDK4 promoter (CDK Luc) by Myc is further stimulated by increasing amounts of TBP (μg indicated on the abscissa). The synergy between Myc and TBP is dependent on the Myc binding sites, which are mutated in the reporter CDK EMut Luc. Bars are shown as averages with standard deviation (n = 6). Transfections were performed using NIH 3T3 cells. ::: ![](1471-2091-6-7-7) ::: ODC, the prototypical Myc responsive gene, provides a different model that utilizes intronic Myc binding sites. The ODCLuc reporter comprises the ODC promoter and intronic E boxes with flanking sequences. Compared to the LDHA promoter, which displayed about an 8 fold induction, ODCLuc responded to TBP and Myc with a 35-fold induction (Fig. [8](#F8){ref-type="fig"}). With this robust response, we sought to determine the response of ODCLuc to Myc and TBP mutants (Fig. [9](#F9){ref-type="fig"}). In this experiment, ODCLuc was co-transfected with a constant amount of Myc and the maximal amount of input TBP. Whereas wild-type TBP stimulated ODCLuc, all TBP mutants studied had virtually no activity (Fig. [9](#F9){ref-type="fig"}). The lack of activity of TBP mutants defective in TATA binding or Pol II interactions was expected; however, we are intrigued by the lack of apparent activity of the Pol III defective TBP mutants. Further studies will be necessary to define the molecular basis for the lack of synergy between Myc and these various TBP mutants. We conclude however, that wild-type TBP is necessary for synergy with Myc in the transactivation of ODC Luc. ::: {#F8 .fig} Figure 8 ::: {.caption} ###### Activation of the ornithine decarboxylase (ODC) intronic E-boxes driven luciferase reporter (ODC Luc) by Myc is further stimulated by TBP (μg indicated on the abscissa). The relative marked stimulation of the ODC sequences by Myc and TBP is compared with the more diminished stimulation of the LDHA promoter (LDH Luc) previously shown in Fig. 4. Bars are shown as averages with standard deviation (n = 10). Transfections were performed using NIH 3T3 cells. ::: ![](1471-2091-6-7-8) ::: ::: {#F9 .fig} Figure 9 ::: {.caption} ###### Synergy between Myc and TBP to activate ODC Luc (see Fig. 8) is diminished by mutations in TBP that inhibits TATA box binding (TATA def), interaction with RNA polymerase II (pol2 def) or interaction with RNA polymerase III (pol3 def). Myc (1 μg) was cotransfected with ODC Luc and 4 μg of wild-type or mutant TBP constructs. Bars are shown as averages with standard deviation (n = 4). Transfections were performed using NIH 3T3 cells. ::: ![](1471-2091-6-7-9) ::: Discussion ========== Myc\'s dramatic biological effects, a plethora of well-characterized interactions between Myc and other proteins, an ever-expanding list of putative target genes and a seemingly weak transactivation potential characterize the enigma of c-Myc-mediated gene regulation \[[@B4],[@B46]\]. Compared with other more potent transactivators, especially in the same family of HLH-Zip proteins, c-Myc stimulates reporter constructs only 2- to 5-fold in an E-box dependent manner. The basis for this apparently weak transactivation is poorly understood. We report in this paper a strong synergy between Myc and TBP resulting in up to 35-fold induction of reporter plasmids. Our observations indicate that TBP is limiting for Myc transactivation and provide a means to enhance the characterization of Myc target genes. The weak transactivation potential of c-Myc may well be biologically significant when the degree of gene induction by c-Myc is considered \[[@B8]\]. The emergence of an increasing number of Myc target genes reveals several characteristics among the genes. With only a few exceptions, Myc induces endogenous genes by only a few fold above background. In multiple instances, it appears that the broad-based effect of inducing multiple genes in the same pathway by c-Myc may be more important than the marked induction of a few genes \[[@B7],[@B47]\]. Perhaps c-Myc globally affects gene expression through multiple mechanisms. The connection between c-Myc and histone acetylation has become more firmly established, suggesting a role for Myc to modulate chromatin \[[@B47]-[@B51]\]. Beyond chromatin modulation, the role of Myc in transcriptional initiation or elongation is less well understood. Searches for an interaction between Myc and members of the general transcription factors have revealed an interaction between the Myc transactivation domain and TBP \[[@B31]\]. In the work reported here, we provide evidence for a functional interaction between Myc and TBP in transient transfection reporter assays. Although the addition of TBP dramatically enhances these assays, the biological significance of this synergism is not delineated in our study. In particular, since many Myc target genes are induced only several fold in vivo, the role of TBP in modulating these target genes in vivois not at all clear. In response to TBP and Myc, promoters with an initiator element respond differently compared with those with a TATA element only \[[@B39]-[@B41]\]. With both the GAL4 chimeric proteins and full length Myc, InR driven promoters respond to the Myc TAD briskly. However, the increase in TBP did not further augment the activities of InR driven promoters. These observations are consistent with previous findings that TBP is limiting for TATA driven, but not InR driven promoters in Drosophila\[[@B44]\]. It is intriguing to note the initial brisk response of InR containing promoters to Myc, which at high levels can inhibit the same InR driven promoters. We surmise from these findings that promoters comprising both TATA and InR elements may control Myc responsive genes that require brisk increased expression within a narrow window of Myc levels independent of TBP. Such genes would be sharply induced by Myc, which in excess can inhibit the same genes through the InR \[[@B37],[@B45]\]. In contrast to InR containing promoters, promoters with TATA element only, such as CDK4 and LDHA, increase in activity with increasing TBP levels in the presence of a constant amount of Myc. These promoters may be regulated by the activity of TBP in vivo, although evidence for this is lacking. The observation that oncogenic Ras can augment TBP activity suggests that a subset of Myc target genes may also be further responsive to increased TBP through activated oncogenic Ras \[[@B52]\]. In fact, Myc and Ras can cooperate to regulated cdc2 \[[@B53]\]. Hence, it will be instructive to determine the set of Myc responsive genes versus the set of genes that are responsive to both Myc and Ras. Comparison of promoters or regulatory regions of these genes are likely to uncover a level of transcriptional complexity previously unappreciated. The fact that the synergy between TBP and Myc was observed with the GAL4 chimeric activator system and full length Myc suggests that the synergy is mediated through the Myc transactivation domain. Furthermore, the Myc Box II deletion mutant D106-143 was unresponsive to increasing TBP, indicating that this region of Myc is required for synergy with TBP. This observation is consistent with the previous finding that in vitro interaction between Myc and TBP requires Myc Box II \[[@B28]\]. Although we observed a significant synergy between Myc and TBP, none of the TBP mutants retained any synergistic activity. It is not surprising that both TATA box binding mutant and Pol II interaction defective TBP mutants were dysfunctional. Although it may seem surprising that Pol III interaction defective TBP mutants were also inactive, recent studies suggest a significant overlap between Pol II and Pol III interactions with TBP \[[@B38]\]. Although beyond the scope of the current study, it will be of significant interest to map the regions of TBP required for the interaction with Myc and correlate this with the ability for TBP mutants to synergize with Myc. Conclusion ========== In summary, we describe in this report a significant stimulation of Myc transactivation by TBP. However, the presence of an InR diminishes the promoter response to TBP. We surmise that these differences may be exploited in vivoto increase the complexity and range of gene regulation by Myc. Methods ======= Plasmids -------- GAL4 constructs were as described \[[@B9]\]. GAL4-MycW135E (GMW135E) is GAL4(1-262) in which the Pst1-Pst1 fragment was exchanged with a fragment containing the substitution W135E from full length c-mycin MLVMycW135E \[[@B43]\]. GAL4-USF1 was constructed by inserting a PCR-amplified, sequence-verified 560-bp USF1 fragment (corresponding to residues 1--180) into the NdeI-BstEII sites of pGALm. The USF1 cDNA template for PCR was a gift from M. Sawadogo and R. Roeder \[[@B10]\]. The Gal4 TATA-driven reporter G5TATALuc was constructed from G5TATA-CAT(gift from M. Green) by replacing CAT with luciferase (Luc) \[[@B54]\]. The Gal4 InR-driven G5INRLuc reporter (gift from J. Gralla) was as described \[[@B55]\]. Murine sarcoma virus long terminal repeat (MSV-LTR) promoter driven wild-type and mutant TBP expression vectors were gifts from A. Berk and are as described \[[@B56]\]. Expression vectors for wild-type and mutant c-mycare as described \[[@B43]\]. The reporter ornithine decarboxylase ODC-Luc is a gift from J. Cleveland \[[@B35]\]. The wild-type and mutant lactate dehydrogenase promoter LDH-Luc reporters were previously described \[[@B36]\]. Wild-type and mutant adenoviral major late promoter AdML-Luc constructs were previously reported \[[@B37]\]. Wild-type and mutant cyclin dependent kinase CDK4-Luc was described \[[@B34]\]. Cell culture and transfection ----------------------------- Chinese Hamster Ovary (CHO) cells were grown in 5% CO2 at 37°C in αMEM (LTI) supplemented with 10% fetal bovine serum (LTI) and antibiotics as described \[[@B9]\]. Cells were transiently transfected using DEAE-dextran (0.275 mg/ml). Two μg of GAL4 reporter luciferase plasmid, two μg of GAL4 chimeric activator plasmid and increasing amounts of pLTRTBP (0.5 to 4 μg) were cotransfected into 100 mm plates of 60% confluent CHO cells. DNA concentration was maintained at 8 μg by the addition of pLTR empty vector. Cells were incubated with the DNA in serum-free DEAE-dextran/MEM media overnight. Cells were harvested 48 hours after DMSO/ chloroquine shock and assayed for luciferase activity as described. NIH3T3 cells (gift from R. Eisenman) were grown in 5% CO2 at 37°C in DMEM (low glucose) (LTI) supplemented with 10% fetal bovine serum and antibiotics. Cells were transfected with Lipofectin (LTI). Lipofectin was added at 5 X the total concentration of DNA to serum-free Opti-MEM media (LTI) and incubated at room temperature for 45 min. Two μg of various promoter-reporter luciferase plasmid and one μg of full-length wild-type or mutant c-mycactivator plasmid were added with increasing amounts of pLTRTBP (0.5 to 4 μg) to the Opti-MEM/Lipofectin mixture. DNA concentrations were maintained at 7 μg by adding pLTR empty vector. The Opti-MEM/ Lipofectin/ DNA mixture was incubated at room temperature for 10 min. then added to 100 mm plates of 30% confluent NIH3T3 cells. Cells were incubated for 5 hours, aspirated and fed fresh media and harvested after 48 hours for luciferase activity. Luciferase assay ---------------- Luciferase activity was measured using the luciferase assay system according to manufacturer\'s instructions (Promega). Cells were washed in phosphate buffered saline (PBS), scraped using Cell Lysis Solution (Promega) and centrifuged for 2 min at 1000 rpm. Luciferin cocktail (80 μl) was added to 20 μl lysate and luciferase activity was measured in a luminometer. Samples were run in duplicate. Authors\' contributions ======================= J. Barrett and L.A. Lee performed experiments. All authors designed experiments, interpreted data, participated in writing the manuscript and approved the final version. Acknowledgements ================ We thank J. Cleveland, A. Berk, R. Eisenman, J. Gralla, M. Green, and G. Kato for reagents or comments. This work was supported in part by NIH grant CA 51497 and CA09159. C.V.D. is The Johns Hopkins Family Professor in Oncology Research.	PubMed Central	2024-06-05T03:55:57.502440	2005-5-5	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1145180/", "journal": "BMC Biochem. 2005 May 5; 6:7", "authors": [ { "first": "John F", "last": "Barrett" }, { "first": "Linda A", "last": "Lee" }, { "first": "Chi V", "last": "Dang" } ] }
PMC1145181	Background ========== Integrins are cell surface adhesion receptors which mediate cell-extracellular matrix (ECM), cell-cell and cell-pathogen interactions. Integrin receptors are structurally elaborate and composed of non-covalently associated α and β subunits. Integrins have a large extracellular domain responsible for binding extracellular ligands, a transmembrane domain and, a relatively small intracellular domain that interacts with the cytoskeleton and intracellular signaling pathways. Integrins integrate information from the extracellular and cytoplasmic environments by transducing signals bidirectionally across the plasma membrane. Hence, the binding of a specific ECM ligand by an integrin may elicit the activation of intracellular signaling pathways, cytoskeletal reorganisations and changes in cell adhesion or migration; and conversely, alterations in the intracellular environment and signaling can result in the activation or inhibition of ligand binding by the extracellular domain of an integrin \[[@B1],[@B2]\]. Consequently, integrins have fundamental roles in diverse physiological processes including: tissue morphogenesis and remodeling \[[@B3]\], immune and inflammatory responses \[[@B4]\], and regulation of cell growth, migration and differentiation \[[@B5]\]. Integrin ligands include ECM components such as laminins, fibronectin and collagens \[[@B2]\], cell surface intercellular adhesion molecules (ICAMs) and plasma proteins such as fibrinogen \[[@B4]\]. Since cell adhesion and the production of a collagen-based ECM are essential characteristics of metazoa, it is not surprising that integrins have been detected throughout the multicellular animal kingdom, from the simplest and most primitive phyla (sponge and cnidarians) \[[@B6]\] to higher vertebrates. In humans, 18 α and 8 β integrin subunits combine to form 24 functionally distinct heterodimers \[[@B2]\]. Within the complement of 24 human integrin receptors, distinct functional sub-divisions can be made on the basis of ligand specificity and tissue distribution. A previous study of integrin phylogeny \[[@B7]\] identified 5 α subunit clades with vertebrate chains occupying 4 of these, namely: laminin binding (α3, α6 and α7 -- PS1 clade), RGD tri-peptide binding (αV, αIIb, α5 and α8 -- PS2 clade), and 2 vertebrate-specific clusters consisting of a small clade comprising the α4 and α9 subunits and a large αA-domain containing clade including both collagen-binding (α1, α2, α10 and α11) and leukocyte-specific (αD, αE, αL, αM and αX) α subunits (I-DOM clade). The αA-domain is structurally homologous to the module identified originally in von Willebrand factor. The collagen-binding αA-domains contain an inserted α-helix which appears to contribute directly to ligand binding. α subunit integrin homologues from model invertebrates C. elegansand D. melanogasterclustered in the laminin clade (PS1) and the RGD binding (PS2) clade with remaining α chains forming a drosophila-specific PS3 clade \[[@B7]\]. The α chain αA-domain, shared by all members of the I-DOM clade, mediates ligand binding in a metal ion-dependant manner by way of a conserved, non-contiguous sequence termed the metal ion-dependent adhesion site (MIDAS) motif \[[@B8],[@B9]\]. It is somewhat surprising that no examples of collagen-binding (i.e. αA-domain-containing) integrins have been found in protostomes since basement membrane and fibrillar collagens are essential components of some of the most primitive invertebrates such as the cnidarian, Hydra vulgaris\[[@B10],[@B11]\]. The origin of the I-DOM clade remains to be fully determined although since the urochordate Halocynthia roretzihas at least one α integrin containing an αA-domain \[[@B12]\] the limited evidence available to date suggests that it may be a chordate invention. In contrast, all known β subunits have a conserved βA-domain (also known as the I-like domain) \[[@B13]\] that must therefore have been present in the prototypic metazoan β subunit. Vertebrate β subunits have been resolved into three branches by phylogenetic analysis, with a majority of sequences falling into two well-supported clades \[[@B7]\]. The two clades, termed β1 (β1, β2 and β7) and β3 (β3, β5, β6 and β8) included seven of the eight known subunits. In the β1 clade, the β2 and β7 subunits are leukocyte-specific, whereas their common ancestor, the β1 subunit, forms heterodimers with 12 of the 18 α subunits. Clade β3 subunits are all specific to RGD ligand binding integrins. The β4 subunit was positioned separately to other vertebrate clades and contains a unique extended cytoplasmic domain (\~1000 residues), which makes direct contact with intermediate filaments via type fibronectin III repeats \[[@B14]\]. Protostome β subunits from C. elegansand D. melanogasterwere not found to cluster with vertebrate sequences. Similarly, early deuterostome sequences (sea urchin) formed lineage-specific clusters with poor resolution amongst all invertebrate clades. A recent study by Miyazawa and Nonaka (2004) presented a phylogeny of integrin β subunits including novel sequences from the urochordate Halocynthia roretzi. These subunits are expressed on ascidian phagocytic blood cells (hemocytes) but phylogenetic analyses positioned them distal from vertebrate leukocyte β integrins in an ascidian-specific clade \[[@B15]\]. The recent sequencing of the genome of the ascidian Ciona intestinalis\[[@B16]\], a urochordate and one of the closest invertebrate relatives of vertebrates, provides a unique opportunity to gain insight into the complete set of integrins available in chordates prior to the large-scale or genome duplication events that many believe were associated with the early stages of vertebrate evolution \[[@B17]-[@B20]\]. An early preliminary analysis identified candidate integrin genes in the Cionagenome \[[@B21]\]. Here we have identified and refined the sequences of 11 α and 5 β integrin subunits from the Cionagenome. Eight of the α chains contain an αA-domain (also known as an I-Domain) that lacks the collagen-binding α helix and these chains form an ascidian-specific clade related to, but distinct from, the vertebrate I-DOM clade. The remaining 3 α chains are predicted to bind laminins and RGD-containing motifs. Two Cionaβ chains cluster with the vertebrate β1 clade and the remaining 3 form an ascidian-specific β clade. The majority of chains in ascidian-specific clades are expressed on cells in the blood and are likely to be involved in innate immunological processes. These novel data provide further insights into the mechanism of evolution of the vertebrate family of integrins, and specifically when and how specific clades of integrin chains first arose. The differences between the ascidian and vertebrate complements of integrins again emphasises how phyla and species mould their genomes by the amplification of specific subsets of genes as part of the process of acquiring a stable and successful phenotype. Results ======= Obtaining refined protein sequences for integrins encoded in the Ciona genome The C. intestinalisgenome database was searched for α and β integrin genes and the sequences listed in Fig. [1A](#F1){ref-type="fig"} were obtained. These sequences were already annotated and no additional genes were obtained during extensive BLAST searching of the Cionagenome using the 26 human integrin chain sequences. The 12 annotated α chain sequences recovered directly from genome database clearly represented fragments of genes, whereas the five β chain sequences represented a much more complete data set (Fig. [1A](#F1){ref-type="fig"}). In order to generate more complete α chain data, DNA flanking the predicted gene fragments were downloaded and searched manually for putative \"missing\" exons. The search involved the identification of ORFs based on their sequence being present in EST clones or conservation of their translated sequence in comparison with homologous genes. Eleven Cionaα chain genes were confirmed by this process (Fig. [1B](#F1){ref-type="fig"}) with two of the JGI predicted genes (ci0100152017 & ci0100152002 -- Fig. [1A](#F1){ref-type="fig"}) proving to represent different regions of a single gene (Fig. [2](#F2){ref-type="fig"}). The sequence refinement process was aided by the fact that four of the Cionaα chain genes were very highly conserved and had clearly radiated relatively recently by a process involving tandem duplication. The completion of one full-length gene sequence therefore considerably simplified the elucidation of the exon structures of the remaining three genes since these were well conserved (Table [1](#T1){ref-type="table"}), which was not the case when comparing the gene structures of more distantly related integrins (data not shown). The size and relative genomic loci of these four genes (ci100152017, ci0100130149, ci0100152615 & ci0100131399 -- subsequently referred to as α5--8) is presented in Fig. [2](#F2){ref-type="fig"}. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Comparison of sequences recovered directly from the JGI C. intestinalisdatabase before (A -- Recovered) and after (B -- Refined) sequence refinement. The annotated cartoons represent the domain structures of generic α & β integrin chains. Each subsequent row represents: (A -- Recovered) the domain structure encoded by a sequence as retrieved directly from the database together with its assigned database accession number; and (B -- Refined) the refined version of that gene after detailed analysis of the genomic sequence as described in the methods together with the name assigned during these analyses (e.g. Ci\_α1). Refined sequences are presented as alignments in Fig. 3-7 and are also included as amino acid residue sequence files (see [additional file 1](#S1){ref-type="supplementary-material"}). ::: ![](1471-2148-5-31-1) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Exon sizes (bp) of Ci\_α5--8 genes. ::: Scaffold 21 91 21 21 ------------ -------------- -------------- -------------- -------------- JGI acc no ci0100152002 ci0100131399 ci0100152615 ci0100130149 Exon no Ci\_a5 Ci\_a6 Ci\_a7 Ci\_a8 1 170 161 167 167 2 128 128 128 128 3 63 63 63 63 4 127 127 67 127 5 123 123 121 123 6 116 92 80 90 7 187 205 187 219 8 179 179 181 179 9 57 68 68 68 10 179 180 180 180 11 87 88 90 86 12 256 256 234 256 13 210 210 232 210 14 307 310 313 304 15 149 146 146 146 16 224 224 224 224 17 143 143 143 143 18 80 107 107 107 19 87 89 89 89 20 101 99 99 99 21 121 120 122 120 22 108 109 109 109 23 129 129 129 129 24 114 114 126 202 25 88 88 76 135 26 135 132 135 97 27 151 92 95 No Exon ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### The genomic locations and orientations of recovered and refined α integrin genes present on scaffolds 21 and 91. The genomic locations and orientations of the four very closely related A-domain containing α integrin genes identified after sequence refinement (Ci-α5--8, coloured red) together with the original five JGI-predicted gene fragments (blue) are indicated. ::: ![](1471-2148-5-31-2) ::: Sequence alignments and characterization ---------------------------------------- The sequences for the 11 Cionaα chains were aligned with human homologues. An annotated version of this alignment is presented in Fig. [3](#F3){ref-type="fig"},[4](#F4){ref-type="fig"},[5](#F5){ref-type="fig"}. Eight of the 11 Cionaα chains (α1--8) have a well conserved αA-domain including the essential residues that constitute the MIDAS motif (Fig. [3](#F3){ref-type="fig"}). The position of the αA-domain insertion in the human and Cionachains is identical, indicating that all these chains have arisen from a common progenitor. However, all 8 Cionachains lack 9--11 amino acid residues corresponding to the \'collagen binding\' α-helical domain present in the collagen binding (Hs\_α1 & α10) vertebrate α chains (Fig. [3](#F3){ref-type="fig"}). The Cionaα chains share all other major features with their vertebrate homologues across the alignment including the well-defined transmembrane and conserved intracellular interaction domains (Fig. [3](#F3){ref-type="fig"},[4](#F4){ref-type="fig"},[5](#F5){ref-type="fig"}). ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Alignment of the refined Cionaα chain sequences with representative human orthologues (residues 1-391 based on human α1 integrin chain). Protein domains and conserved motifs are annotated. Levels of sequence conservation are indicated (\>50% identical, red; conservative substitutions, blue). MIDAS and α C-helix within the inserted A-domain are highlighted, as are the β-propeller domains 1--3. ::: ![](1471-2148-5-31-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Alignment of the refined Cionaα chain sequences with representative human orthologues (residues 392-796 based on human α1 integrin chain). Protein domains and conserved motifs are annotated. Levels of sequence conservation are indicated (\>50% identical, red; conservative substitutions, blue). Ca^2+^-binding motifs in β-propeller repeats 5--7 are highlighted. ::: ![](1471-2148-5-31-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Alignment of the refined Cionaα chain sequences with representative human orthologues (residues 797 to C-terminus based on human α1 integrin chain). Protein domains and conserved motifs are annotated. Levels of sequence conservation are indicated (\>50% identical, red; conservative substitutions, blue). Transmembrane domain (TM) and cytoplasmic interaction motif are indicated. ::: ![](1471-2148-5-31-5) ::: The sequences of the five Cionaβ chains are also highly conserved with respect to their vertebrate orthologues including the MIDAS motif within the I-like domain, the four EGF domains, the transmembrane domain, and the intracellular interaction motifs and PTB-like domains (Fig. [6](#F6){ref-type="fig"} &[7](#F7){ref-type="fig"}). ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Alignment of the refined Cionaβ chain sequences with representative human orthologues (residues 1-542 based on the human β1 integrin chain). Protein domains and conserved motifs are annotated. Levels of sequence conservation are indicated (\>50% identical, red; conservative substitutions, blue). Adjacent to MIDAS (AMIDAS), ligand associated metal binding site (LIMBS) and MIDAS cation binding sites, and interaction motifs are highlighted as are the plexin/semaphorin/integrin (PSI), β-A domain (I-like) and epidermal growth factor (EGF) domains 1--2. ::: ![](1471-2148-5-31-6) ::: ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Alignment of the refined Cionaβ chain sequences with representative human orthologues (residues 543 to the C-terminus based on the human β1 integrin chain). Protein domains and conserved motifs are annotated. Levels of sequence conservation are indicated (\>50% identical, red; conservative substitutions, blue). EGF domains 2--4, transmembrane (TM) domain, interaction and phosphotyrosine binding (PTB) motifs are indicated. ::: ![](1471-2148-5-31-7) ::: Phylogenetic analyses --------------------- The α chain analysis is presented in Fig. [8](#F8){ref-type="fig"} in the form of a maximum likelihood tree with supporting data from 1,000 neighbor-joining bootstrap replicates and Bayesian analysis. Overall the inferred phylogenetic relationships are consistent with a previous phylogenetic reconstruction \[[@B7]\]. The clades identified by Hughes (PS1, PS2, and the vertebrate I-DOM and α4/9) are all present (Fig. [8](#F8){ref-type="fig"}). Note, the PS3 clade is not shown because it is specific to Drosophila. Cionaα9 and α10 cluster in the PS1 clade and their position, separating the protostome and vertebrate sequences, is as expected. In contrast, Cionaα11 clusters with its ascidian orthologue (Hr\_α2) in the PS2 clade but at an anomalous position distal to the protostome sequences. Only Neighbor Joining analysis (not shown) produced the anticipated branching in this region of the tree with protostome PS2 clade members being most distal. The remaining eight Cionaα chains (Ci\_α1 to α8) all have an αA-domain and form an ascidian specific clade that includes the αA-domain containing H. roretziα1 chain (I-DOM \[ascidian\] -- Fig. [8](#F8){ref-type="fig"}). ::: {#F8 .fig} Figure 8 ::: {.caption} ###### Phylogenetic relationship of Cionaα integrin chains with representative protostome and vertebrate orthologues. Maximum Likelihood tree is shown with supporting Neighbor Joining bootstrap replicates (red) and Bayesian clade credibility values (green). Horizontal scale is amino acid replacements per site. ::: ![](1471-2148-5-31-8) ::: The β chain analysis is presented in Fig. [9](#F9){ref-type="fig"}. Again, phylogenetic relationships are consistent with Hughes \[[@B7]\]. The vertebrate clades β1 and β4 include Cionaorthologues (Ci\_β1 and β5 respectively). The vertebrate β3 clade has no identified Cionaorthologue (Fig. [9](#F9){ref-type="fig"}). The remaining 3 Cionaβ chains (Ci\_2 to 4) form an ascidian specific clade including H. roretziβ1 & 2 chains (Fig. [9](#F9){ref-type="fig"}). ::: {#F9 .fig} Figure 9 ::: {.caption} ###### Phylogenetic relationship of Cionaβ integrin chains with representative deuterostome orthologues. Maximum Likelihood tree is shown with supporting Neighbor Joining bootstrap replicates (red) and Bayesian clade credibility values (green). Horizontal scale is amino acid replacements per site. ::: ![](1471-2148-5-31-9) ::: Discussion ========== The urochordate C. intestinalisoccupies a pivotal position in the animal kingdom for understanding the evolution of vertebrates. The Cionagenome provides an insight into the basic set of genes available at the very beginning of vertebrate evolution since the urochordates diverged just prior to the widespread gene duplication processes that are thought to have shaped and transformed the vertebrate genome \[[@B16]\]. The Cionagenome encodes 11 α and five β integrin chains. As expected, some Cionaα chains cluster in the PS1 laminin-binding clade (Ci\_α9 & α10), and in the PS2 RGD-binding clade (Ci\_α11 together with its ascidian orthologue Hr\_ α2 -- see Fig [8](#F8){ref-type="fig"}). It is unclear why the urochordate members of the PS2 clade cluster in an anomalous position distal to the protostome sequences although this suggests that the ancestral urochordate PS2 gene underwent significant and rapid sequence changes after divergence from the lineage leading to vertebrates. The remaining eight Cionaα chains all contain an αA-domain and, somewhat surprisingly, form an ascidian-specific clade related to, but distinct from, the vertebrate I-DOM clade. This phylogenetic relationship suggests the vertebrate and ascidian αA-domain containing clades arose from a common progenitor but that this gene radiated separately in both the ascidian and vertebrate lineages after their divergence. Data supporting this hypothesis includes: i) ascidian and vertebrate genes have the αA-domain inserted at the same location supporting the notion of both lineages having a common progenitor (Fig. [3](#F3){ref-type="fig"}); ii) all eight ascidian α chain αA-domains lack the same 9--11 amino acids encompassing and adjacent to the α-helical domain in the collagen-binding α chains (Fig. [3](#F3){ref-type="fig"}); and iii) at least four of the CionaαA-domain-containing α chains (α5--α8) appear to have arisen very recently as a result of tandem duplications within the ascidian genome based on retained similarities in exon size (Table [1](#T1){ref-type="table"}), their high level of sequence identity (Fig. [3](#F3){ref-type="fig"},[4](#F4){ref-type="fig"},[5](#F5){ref-type="fig"} and [8](#F8){ref-type="fig"}) and their genomic location (Fig. [2](#F2){ref-type="fig"}). The common progenitor gene presumably evolved in deuterostomes, possibly in the earliest chordates since the vertebrate and ascidian lineages of αA-domain α chains have radiated entirely separately and no protostome orthologues have been identified. The most likely function for the progenitor αA-domain-containing α chain involves the adhesion of blood cells to complement-like proteins or the extracellular matrix. In the urochordate H. roretzi, the Hr\_α1 αA-domain gene, together with the Hr\_β1 & β2 genes, are expressed on hemocytes and are thought to act as complement receptors \[[@B12]\]. A large number of EST\'s for Ciona αA-domain-containing α chains have been found in either blood cell or hemocyte cDNA libraries (Table [2](#T2){ref-type="table"}). Finally, the expression of half of the genes comprising the vertebrate I-DOM clade is leukocyte-specific. In addition, it would seem that the collagen-binding property exhibited by the other half of the I-DOM α chains was a late functional acquisition of this vertebrate clade, perhaps associated with an insertional mutagenic event creating the collagen binding α-helix. It is noteworthy that Cionaexpresses progenitor forms for all three clades of vertebrate fibrillar collagens \[[@B22]\]. It is therefore apparent that the early evolution of chordates did not require collagen-binding integrins. Functionally important interactions between integrins and collagen triple helices must have developed later in chordate evolution, possibly in the earliest vertebrates and co-incidental with the acquisition of the collagen-binding helix. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Expression profiles of A-domain containing α-integrins in Ciona intestinalis. Data has been obtained from the TIGR Gene Indices database <http://www.tigr.org/tdb/tgi/>. ::: Ciona Integrin JGI Acc Code TIGR cDNA Index Acc Code Tissue specific expression from EST data ---------------- -------------- -------------------------- ------------------------------------------ Ci\_a1 Ci0100131118 BW029582 Blood cells Ci\_a2 Ci0100149446 TC42900 Blood cells Ci\_a3 Ci0100130596 TC56905 Heart, neural complex, digestive gland Ci\_a4 Ci0100130838 TC66015 (Whole embryo only) Ci\_a5 Ci0100152002 TC6115, TC63051, TC63231 Blood cells, heart, hemocytes Ci\_a6 Ci0100131399 TC73566, TC69775 Blood cells, heart, neural complex Ci\_a7 Ci0100152615 TC59274 Blood cells, digestive gland, gonad Ci\_a8 Ci0100130149 TC75204 Blood cells, neural complex, gonad ::: The phylogenetic relationships of the Cionaand vertebrate β integrin chains (see Fig. [9](#F9){ref-type="fig"}) emphasizes the pivotal position invertebrate chordates occupy with respect to understanding how vertebrates and their genes evolved. Previous phylogenetic analysis has suggested that neither protostome nor early deuterostomes (echinoderm) β chain sequences cluster with their vertebrate orthologues \[[@B7]\]. The clustering of a Ciona(Ci\_β1) and a previously reported echinoderm sequence (Sp\_ βC) with the vertebrate β1 clade genes (Fig. [9](#F9){ref-type="fig"}) resolves more clearly how the promiscuous vertebrate β1 chain and its paralogues have evolved from a deuterostome-specific progenitor. In addition, the vertebrate β4 chain, which fails to cluster with any other vertebrate genes, has a Cionaorthologue (Ci\_ β5, Fig. [9](#F9){ref-type="fig"}) and must therefore have evolved prior to the divergence of ascidians. It was not possible to determine for certain whether the Ci\_ β5 chain has the extended intracellular C-terminal domain present in the vertebrate β4 chain using either EST analysis or the search for exons containing conserved ORFs. Nevertheless, direct translation of 10 kb of genomic sequence 3\' to the predicted C-terminus of the Cionagene revealed the presence of short ORFs homologous to a domain present in the vertebrate integrin β4 intracellular domain and shared by Na^+^-Ca^2+^exchangers (data not shown). The remaining three Cionaβ gene sequences formed an ascidian-specific clade together with two β chains from H. roretzi(Fig. [9](#F9){ref-type="fig"}). As phylogenetic relationships between novel α and β chains become defined, it is possible to start predicting likely interactions based on the known dimerisation partners of close relatives (Fig. [10](#F10){ref-type="fig"}). For instance, the vertebrate β1 chain dimerises with the laminin (PS1) and RGD (PS2) clade α chains. It is therefore probable that the Cionaβ1 clade orthologue (Ci\_β1) dimerises with the CionaPS1 and PS2 clade α chains (Ci\_α9, α10 & α11; see Fig. [10](#F10){ref-type="fig"}). Likewise, it has been established that the 2 β chains from H. roretzi(Hr\_ β1 & β2) dimerise with Hr\_α1 \[[@B15]\] and it is therefore probable that the related Cionaβ2--4 chains partner Cionaα chains in the ascidian αA-domain clade (Ci\_α1--8; Fig. [10](#F10){ref-type="fig"}). In H. roretzi, the β1, β2 and α1 chains are expressed on hemocytes \[[@B15]\] and it is noteworthy that the Cionaorthologues (Ci\_α1--8 & β2--4) are also expressed predominantly in blood tissues based on EST analysis (data not shown). ::: {#F10 .fig} Figure 10 ::: {.caption} ###### Prediction of dimerisation patterns for novel integrin chains based on a combination of known interactions and phylogeny. A. Schematic phylogenies of α and β chains with established heterodimer pairing indicated by adjoining solid colored lines. B. Heterodimer pairings predicted (dashed lines) on the basis of the data presented in A. The color coding in B related to the known parings in A used to make the prediction. ::: ![](1471-2148-5-31-10) ::: The phylogenetic relationships of integrin genes within the vertebrate and invertebrate branches of the chordate phylum provides new insights into the evolution of both of these divergent lineages. The integrin gene complement of the ascidian genome gives a strong indication of the numbers and classes of integrin chains available to organisms at the very start of vertebrate evolution. For α integrins, it would appear that there was a minimum of one laminin binding (PS1), one RGD binding (PS2) and one αA-domain containing chain (Fig. [8](#F8){ref-type="fig"}). The novel Cionadata therefore clearly indicate that radiation of vertebrate α chain genes took place after the divergence of urochordates. The β chain phylogeny (Fig. [9](#F9){ref-type="fig"}) indicates that the common progenitor of urochordates and vertebrates had a minimum of one β1-like gene (that subsequently radiated in vertebrates) and a single Hs\_β4/Ci\_ β5 progenitor that radiated in neither lineage (Fig. [9](#F9){ref-type="fig"}). The origins of the ascidian and vertebrate-specific β clades is not resolved. The ascidian lineage exhibits amplifications of subsets of integrin genes to produce ascidian-specific classes of novel integrins (Fig. [8](#F8){ref-type="fig"},[9](#F9){ref-type="fig"},[10](#F10){ref-type="fig"}). In particular, these novel integrins appear to be expressed in blood (see above) and may be involved in mechanisms associated with innate immunity. It has been proposed that the major metamorphic transformations between ascidian larval and adult body forms may be dependent upon innate immune responses \[[@B23]\]. This suggestion would provide a plausible explanation of the requirement for an expanded set of urochordate hemocyte integrins although more generic explanations, such as the preparation for a sessile life-style under attack by pathogens, are also possible. Methods ======= The complete sequences of the 26 known human integrin genes were used to probe the Ciona intestinalisgenome and TIGR cDNA gene index (<http://genome.jgi-psf.org/ciona4/ciona4.home.html> and <http://www.tigr.org/tdb/tgi/cingi> using TBLASTN and PSI-BLAST with cut-off expectancy values of E = 1) to identify homologous genes \[[@B16],[@B24]\]. Cionagene models were also detected using the orthologue detection program InParanoid using a keyword search using \'integrin\' as the query <http://abi.marseille.inserm.fr/cgi-bin/karine/inparanoid-para>\[[@B25]\]. To identify all the integrin genes, reciprocal BLAST searches of the Ciona, human and non-redundant databases were used. Frequently, EST for the Cionagenes contradicted the proposed gene models from JGI. In instances where an EST clearly demonstrated the misplacement of exons in the recovered JGI model, the protein sequence was corrected to reflect this. To detect missing exons not supported by EST data, genomic DNA flanking the sequence of interest was retrieved and analysed using the GENESCAN \[[@B26]\]<http://genes.mit.edu/GENSCAN.html> and GENEWISE \[[@B27]\]<http://www.ebi.ac.uk/Wise2> gene prediction programmes. Modified sequences were checked by aligning with respective human integrin profiles using CLUSTAL X \[[@B28]\] and corrected coding sequences used for subsequent analyses. Expression profiles for the Cionagenes were obtained from the TIGR database (see above). The α and β integrin sequences were aligned separately using CLUSTAL X. The variable domain structure amongst α integrins necessitated subdivision of the alignment groups based on the presence/absence of an αA-domain. Subgroups were aligned and then combined so that the final alignment contained all the α integrins with a 200-residue (approx) gap region corresponding to the αA-domain. For phylogenetic analysis, gap-containing sites were removed from each alignment and Maximum Likelihood trees were inferred using PROML from the PHYLIP package \[[@B29]\]. The JTT model of amino acid substitutions was used with and without global rearrangements and correction for rate heterogeneity (α value obtained from TREEPUZZLE \[[@B30]\]). The topologies of the trees were tested using two independent methods: Neighbour-joining bootstrap replicates and Bayesian tree inference using PHYLIP and Mr Bayes programmes respectively \[[@B31]\]. The accession numbers for protein sequences used in this study are presented in Tables [3](#T3){ref-type="table"} &[4](#T4){ref-type="table"}. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Summary of α integrins used in phylogenetic analysis. ::: Lineage Species Database Acc Code Gene ------------- ----------------- ------------- -------------- ------------ Chordate H. sapiens Swiss-Prot P56199 Hs\_α1 Chordate H. sapiens Swiss-Prot P17301 Hs\_α2 Chordate H. sapiens Swiss-Prot P08514 Hs\_αIIb Chordate H. sapiens Swiss-Prot P26006 Hs\_α3 Chordate H. sapiens Swiss-Prot P13612 Hs\_α4 Chordate H. sapiens Swiss-Prot P08648 Hs\_α5 Chordate H. sapiens Swiss-Prot P23229 Hs\_α6 Chordate H. sapiens Swiss-Prot Q13683 Hs\_α7 Chordate H. sapiens Swiss-Prot P53708 Hs\_α8 Chordate H. sapiens Swiss-Prot Q13797 Hs\_α9 Chordate H. sapiens Swiss-Prot O75578 Hs\_α10 Chordate H. sapiens Swiss-Prot Q9UKX5 Hs\_α11 Chordate H. sapiens Swiss-Prot Q13349 Hs\_αD Chordate H. sapiens Swiss-Prot P38570 Hs\_αE Chordate H. sapiens Swiss-Prot P20701 Hs\_αL Chordate H. sapiens Swiss-Prot P11215 Hs\_αM Chordate H. sapiens Swiss-Prot P06756 Hs\_αV Chordate H. sapiens Swiss-Prot P20702 Hs\_αX Urochordate C. intestinalis JGI Ci v1.0 See Fig. S1 Ci\_α1 Urochordate C. intestinalis JGI Ci v1.0 ci0100149446 Ci\_α2 Urochordate C. intestinalis JGI Ci v1.0 See Fig. S1 Ci\_α3 Urochordate C. intestinalis JGI Ci v1.0 See Fig. S1 Ci\_α4 Urochordate C. intestinalis JGI Ci v1.0 See Fig. S1 Ci\_α5 Urochordate C. intestinalis JGI Ci v1.0 See Fig. S1 Ci\_α6 Urochordate C. intestinalis JGI Ci v1.0 See Fig. S1 Ci\_α7 Urochordate C. intestinalis JGI Ci v1.0 See Fig. S1 Ci\_α8 Urochordate C. intestinalis JGI Ci v1.0 See Fig. S1 Ci\_α9 Urochordate C. intestinalis JGI Ci v1.0 See Fig. S1 Ci\_α10 Urochordate C. intestinalis JGI Ci v1.0 ci0100154687 Ci\_α11 Urochordate H. roretzi Genbank AB048261 αHr1 Urochordate H. roretzi Genbank AB048262 αHr2 Arthropod D. melanogaster Swiss-Prot Q24247 Dm\_aPS1 Arthropod D. melanogaster Swiss-Prot P12080 Dm\_aPS2 Arthropod D. melanogaster Swiss-Prot U76605 Dm\_aPS3 Arthropod D. melanogaster Swiss-Prot AAF58154 Dm\_a16827 Arthropod D. melanogaster Swiss-Prot AAF47029 Dm\_a5372 Nematode C. elegans Swiss-Prot P34446 Ce\_a1 Nematode C. elegans Swiss-Prot Q03600 Ce\_a2 Echinoderm S. purpuratus Swiss-Prot AF177914 Sp\_aP Porifera G. cydonium Swiss-Prot X97283 Gc\_alpha ::: ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Summary of β integrins used in phylogenetic analysis ::: Lineage Species Database Acc code Gene ------------- ----------------- ------------- -------------- ------------ Chordate H. sapiens Swiss-Prot P05556 Hs\_b1 Chordate H. sapiens Swiss-Prot P05107 Hs\_b2 Chordate H. sapiens Swiss-Prot P05106 Hs\_b3 Chordate H. sapiens Swiss-Prot P16144 Hs\_b4 Chordate H. sapiens Swiss-Prot P18084 Hs\_b5 Chordate H. sapiens Swiss-Prot P18564 Hs\_b6 Chordate H. sapiens Swiss-Prot P26010 Hs\_b7 Chordate H. sapiens Swiss-Prot P26012 Hs\_b8 Urochordate C. intestinalis JGI Ci v1.0 ci0100141446 Ci\_b1 Urochordate C. intestinalis JGI Ci v1.0 ci0100143908 Ci\_b2 Urochordate C. intestinalis JGI Ci v1.0 See Fig. S1 Ci\_b3 Urochordate C. intestinalis JGI Ci v1.0 ci0100131678 Ci\_b4 Urochordate C. intestinalis JGI Ci v1.0 ci0100143050 Ci\_b5 Urochordate H. roretzi Genbank AB154831 Hr\_b1 Urochordate H. roretzi Genbank AB154832 Hr\_b2 Echinoderm S. purpuratus NCBI AF0559607 Sp\_bC Echinoderm S. purpuratus NCBI NP\_999732 Sp\_bG Echinoderm S. purpuratus NCBI NP\_999731 Sp\_bL Arthropod D. melanogaster Swiss-Prot P11584 Dm\_bPS Arthropod D. melanogaster Swiss-Prot L13305 Dm\_bv Nematode C. elegans Swiss-Prot Q27874 Ce\_b-pat3 Mollusc B. glabraba Swiss-Prot AF060203 Bg\_beta Cnidaria A. millepora Swiss-Prot AF005356 Am\_beta Porifera G. cydonium Swiss-Prot O97189 Gc\_beta ::: Authors\' contributions ======================= RE and JHJ performed the database searches and analyses. RBH, DLR and MJH conceived of the study. All authors participated in the interpretation of data and in the writing of the manuscript. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Amino acid sequences of the refined Ciona intestinalisalpha and beta integrin chains (see Fig. [1](#F1){ref-type="fig"}) used to produce alignments in Figs [3](#F3){ref-type="fig"},[4](#F4){ref-type="fig"},[5](#F5){ref-type="fig"},[6](#F6){ref-type="fig"},[7](#F7){ref-type="fig"} inclusive. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ JHJ is supported by a BBSRC PhD studentship and MJH by The Wellcome Trust.	PubMed Central	2024-06-05T03:55:57.504742	2005-5-13	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1145181/", "journal": "BMC Evol Biol. 2005 May 13; 5:31", "authors": [ { "first": "Richard", "last": "Ewan" }, { "first": "Julie", "last": "Huxley-Jones" }, { "first": "A Paul", "last": "Mould" }, { "first": "Martin J", "last": "Humphries" }, { "first": "David L", "last": "Robertson" }, { "first": "Raymond P", "last": "Boot-Handford" } ] }
PMC1145182	Background ========== The absorption of biologically active compounds occurs via passive transcellular, paracellular and carrier-mediated transport mechanisms \[[@B1]\]. Analysis of the human genome sequence suggested the presence of 406 genes encoding ion channels and 883 genes encoding transporters \[[@B2]\]. Generally, these proteins establish the electrochemical gradient across membranes and provide the means for transporting electrolytes, amino acids, dipeptides, monosaccharides, monocarboxylic acids, organic cations, phosphates, nucleosides, and water-soluble vitamins \[[@B3],[@B4]\]. Frequently, transporters play a direct role in the absorption of bioactive compounds from the intestinal lumen. The bioavailability of some compounds can depend significantly on carrier-mediated systems and, thus, are sensitive to drug-drug and drug-food interactions; however, these interactions tend to be more relevant for bioactive molecules with low bioavailabilities \[[@B5]\]. The mRNA expression profiles of several functionally-defined transporter families have already been measured by real-time PCR \[[@B6],[@B7]\]; yet, for a majority of biologically active compounds it remains unknown which transporters play a role in their absorption. Although the human genome project has made the identification of most, if not all, genes encoding transporters and ion channels possible, only a few studies have focused on intestinal transporter expression using such genome-wide strategies \[[@B8]-[@B11]\]. Therefore, insight into the expression profiles of transporters along the intestinal tract enables a physiologically relevant assessment of their potential as drug- and nutrient-carriers. Even less is known about the expression of transporters along the crypt-villus axis \[[@B12]\]. The extensive automated literature data mining by Olsen et al. \[[@B12]\] revealed that more transporters are known to be expressed in the villi than in the crypts. Many transporters that are villus-specific have been implicated in absorption processes, such as the oligopeptide transporter Slc15a1, facilitated glucose transporter Slc2a10, and the sodium/glucose cotransporter Slc5a1 \[[@B13]\]; whereas the crypt-specific, basolaterally expressed Na+/HCO3 co-transporter (Slc4a4) is essential for intestinal anion secretion. Similarly, the mulitrdrug resistance protein 1 (Abcc1), is crypt-specific, basolaterally expressed and acts as a secretion pump for various compounds \[[@B14],[@B15]\]. On the other hand, P-glycoprotein (Abcb1) is villus-specific, expressed on the apical site and acts as well as a secretion pump for a variety of drugs \[[@B5],[@B16]\]. Although exceptions may be found, one can assume that villus-specific transporters might be more efficient as mediators of absorption as their surface availability is more extensive. Messenger RNA levels may not always correlate with the expression of encoded proteins \[[@B17]\]. Although use of proteomic techniques increasingly serves to resolve discrepancies between mRNA and protein levels, proteomics of integral membrane proteins still remains a challenge \[[@B18]\]. Also, protein levels do not necessarily correlate with protein activity. Various studies indicate that genomic profiling in combination with data mining of chemotoxicity databases can be an efficient strategy to identify new putative drug carriers \[[@B19],[@B20]\]. A first step in identifying genes relevant to drug absorption in the intestine is to obtain a molecular catalogue of all expressed mRNAs. In this study, we used the high-density oligomer microarray by Affymetrix to measure the mRNA expression levels of genes expressed in four intestinal regions (duodenum, jejunum, ileum, and colon) in the mouse. We identified all genes with transporter activity according to the Gene Ontology (GO) consortium system that are represented on the microarrays. Then, we examined specific transporter classes that are significantly changed along the intestine and compared our findings to publicly available human microarray data. Finally, the mRNA expression of some selected transporters was measured along the crypt-villus axis using laser capture microdissection. Results ======= \"Carriers\" and in particular \"symporters\" and \"antiporters\" are significantly changed along the A-P axis. However, he transporter class as a whole is not --------------------------------------------------------------------------------------------------------------------------------------------------------------- Even though 41 % of all genes on the microarray that belong to the GO class \"transporter activity\" were differentially expressed along the intestine (modified ANOVA, p ≤ 0.05), the class itself, however, was not found significantly changed according to a Fisher\'s exact test (p ≤ 0.05). In 8 of the 17 transporter classes represented on the microarrays (at depth 3 of the \"transporter activity\" mother class, cf. Figure [1](#F1){ref-type="fig"}) more than 50 % of the genes were differentially expressed along the intestine. But only the classes \"carriers\", \"electron\" and \"protein\" were significantly altered (cf. Table [1](#T1){ref-type="table"}) with \"carriers\" being the only class whose subclasses showed significant changes. In particular, genes with \"electrochemical potential-driven transporter activity\", and more specifically \"antiporters\" and \"symporters\" were affected along the intestine. Although the class \"channel/pore\" was not significantly affected, it was the only transporter class (depth 3) besides \"carriers\" that contained subclasses that varied significantly, namely \"chloride channel\" (depth 7, 9 genes out of 13 were changed) and its subclass \"voltage-gated chloride channel\" (depth 8, 7 genes out of 9 were changed). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Classification according to the Gene Ontology system of genes differentially regulated along the intestine. ::: GO description^a^ Total (1-/G-F-Flag)^b^ UGs on Chips^c^ DJ^b,d^ JI^b,d^ IC^b,d^ SI C^b,d^ -------------------------------------------------------- ---------------------------- --------------------- ------------- ------------- ------------- --------------- molecular function 2717 (1772/760/185) 6890 659 897 897 2055 transporter activity 350 (244/75/31) 853 95 115 126 263 amine/polyamine 18 (11/2/5) 35 8 7 11 14 auxiliary transport protein 2 (2/0/0) 2 0 0 1 1 carbohydrate 8 (5/1/2) 18 2 4 7 8 carrier activity 124 (91/10/24) 262 35 41 55 98 electrochemical potential-driven transporter 57 (43/9/5) 104 20 27 35 47 porter 56 (42/9/5) 103 20 26 35 47 antiporter 11 (10/0/1) 16 4 6 9 11 cation\\:cation antiporter 5 (5/0/0) 6 2 2 2 5 symporter 30 (18/7/5) 51 10 12 16 26 solute\\:cation symporter 19 (10/5/4) 31 5 8 9 17 solute\\:Na symporter 17 (8/5/4) 26 4 7 8 14 channel/pore class 64 (40/5/19) 211 15 19 22 47 electron 36 (272/7) 65 10 12 16 29 ion transporter 68 (49/5/14) 151 20 18 25 55 Lipid 6 (3/2/1) 16 2 3 5 4 neurotransmitter 8 (4/2/2) 12 2 3 5 7 nb/ns./nt/nucleic acid 2 (2/0/0) 4 1 0 1 1 Organic acid 22 (13/2/7) 43 8 9 11 16 Oxygen 2 (2/0/0) 5 0 1 0 0 Peptide 3 (3/0/0) 5 1 2 2 2 permease 1 (1/0/0) 5 0 0 0 2 Protein 78 (51/9/18) 163 18 17 13 56 Vitamin/cofactor 3 (1/2/0) 7 1 2 2 2 Water 1 (0/0/1) 2 0 0 1 1 ^a^Transporter activity classes (depth 3) for which no changes were observed are not listed (number of UniGene clusters represented on the microarrays): amide (0), boron (0), drug (1), group translocator (0), lactone (0), nitric oxide (0), organic alcohol (0), peptidoglycan (0), siderochrome (0), toxin (0) and xenobiotic (0). ^b^Listed is the number of differentially regulated genes along the whole intestinal tract and more specifically between adjoining intestinal segments (p ≤ 0.05). Grey cells indicate significant changes in the GO nodes (p ≤ 0.05). ^c^Total number of UniGene clusters being represented on the three Mu74v2 GeneChips. ^d^Sum of all differentially expressed UniGene clusters, i.e., from type (1), (F) and (G). ::: ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Selected branches of the classification system according to the Gene Ontology (GO) system. ::: ![](1471-2164-6-69-1) ::: Expression profiles of transporters change along the whole intestinal tract, but most significantly between the ileum and colon ------------------------------------------------------------------------------------------------------------------------------- The number of significantly (p ≤ 0.05) regulated genes was not different when comparing the jejunum and ileum and the ileum and colon (cf. Table [1](#T1){ref-type="table"}). However, comparisons of transporters revealed a trend that along the A-P axis the differences between the adjoining segments increased. Overall, this trend could also be observed in the transporter subclasses except from the \"amine/polyamine\", \"ion transporter\" and \"protein\" classes. For instance, the biggest subclass in absolute numbers of differentially changed transporters was the class \"carrier\". Along the intestine the difference between the adjoining segments increases clearly from 35 genes (duodenum vs. jejunum) to 41 (jejunum vs. ileum) to 55 (ileum vs. colon). The changes, however, are only significant (p ≤ 0.05) for the comparisons duodenum vs. jejunum and ileum vs. colon. Interestingly, no significant change was observed for the nodes of the jejunum vs. ileum comparison indicating that relative to all genes being differently expressed between the jejunum and the ileum the number of differently regulated transporters was not significantly higher. Like the mother class \"carriers\", the number of differentially regulated \"symporters\" and \"antiporters\" also increased along the intestine. All changes were significant for both types of porters except for \"symporters\" in the comparison duodenum vs. jejunum (cf. Table [1](#T1){ref-type="table"}). A similar number of transporters are over-expressed in the small and in the large intestine, but generally transporters are expressed at higher levels in the small intestine than in the colon ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- In order to identify transporters which could be interesting candidate carriers for bioactive compounds in the various segments, we examined expression profiles along the intestinal tract. Comparing the expression levels of transporter genes or genes involved in transporter activity in the small intestine and the colon revealed that similar numbers of genes were more highly expressed in either the small or the large intestine. However, small intestinal transporters were clearly more over-expressed (cf. Figure [2](#F2){ref-type="fig"}). The majority of differentially expressed transporters are members of the solute carrier super family (Tables [2](#T2){ref-type="table"} and [3](#T3){ref-type="table"}). The most pronounced change (fold changes \> 300), however, was observed for fatty acid binding protein 1 (Fabp1). Among the transporters being more than 20 times over-expressed in the small intestine were the neurotransmitter transporter Slc6a19, the sulfate transporter pendrin-like protein 1 (Slc26a6), L-type amino acid transporter 2 (Slc7a8), facilitated glucose transporter 2 (Slc2a2), the apolipoprotein C-II (Apoc2), the D11Ertd18e, the retinol binding protein 2 (Rbp2), sterolin 1 and 2 (Abcg5 and Abcg8) (genes are listed in decreasing order of the fold change). The Y(+)L-type amino acid transporter 1 (Slc7a7), the concentrative nucleoside transporter co-transporter (Slc28a3), the B(0,+)-type amino acid transporter 1 (Slc7a9) were more than 10 times over-expressed. In the colon, the most over-expressed transporter was the neurotransmitter transporter Slc6a14 which was more than 200 times higher expressed in the colon compared to the small intestine. Aquaporin 4 (Aqp4), serum amyloid A 3 (Saa3), the neurotransmitter transporter Slc6a14, the facilitated glucose transporter 1 (Slc2a1) and aquaporin 8 (Aqp8) were more than 10 times over-expressed. The concentrative nucleoside transporter 2 (Slc28a2, Cnt2) was the top five most over-expressed transporter in the small intestine compared to the colon, but its annotation quality was only classified as \"medium\". ::: {#F2 .fig} Figure 2 ::: {.caption} ###### A-D: M (log~2~of fold change) vs. A (log~2~of absolute average intensity) plots of the pair-wise comparisons between intestinal segments. All genes are plotted in orange, and transporters for which both a significant difference was measured between the two segments of interest and had a \"high\" quality annotation are depicted in blue (p ≤ 0.05). In black are the transporters highlighted which were measured in the crypts and the villi. Grey lines indicate thresholds for fold changes of four fold (M = ± 2) or five fold (M = ± 2.5), respectively. ::: ![](1471-2164-6-69-2) ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Differently regulated (p ≤ 0.05) solute carriers with a minimum absolute fold change of four in pair-wise comparisons of adjoining intestinal segments. Bold cases indicate fold changes of five or higher (Means ± SDs, log2 scale, n = 3 or 9, respectively.). ::: Gene Name D \> J D \< J J \> I J \< I I \> C I \< C SI \> C SI \< C Literature^a^ Gene aliases References --------------- --------------- --------------- ------------ --------------- --------------- --------------- --------------- ------------- ------------------------------------------------------------------------------------------ ---------------------------------------------------------- ---------------- Slc10a2 5.0 ± 1.2 Mouse and Human: Highly expressed in I ASBT; ISBT \[36, 37, 53\] Slc12a4 2.6 ± 0.5 2.9 ± 0.5 \- KCC1; RBCKCC1 Slc13a1 2.8 ± 0.4 2.2 ± 0.5 Mouse: D/J \~ C \< I; Human: only in kidney Nas1; NaSi-1 \[54, 55\] Slc14a1 2.2 ± 1.2 2.1 ± 1.2 Human: Expressed in C; Rat: expressed in C UT-B \[56, 57\] Slc17a6 2.2 ± 0.4 Rat: abundant in I DNPI, VGLUT2 \[58\] Slc17a7 2.2 ± 0.4 \- Vglut1 Slc19a3 3.8 ± 0.9 Human: D \> J \> C \> I ThTr2 \[59\] Slc1a1 2.4 ± 0.2 2.8 ± 0.2 Mouse: SI; rat: high in distal SI; human: intestine EAAC1, EAAC2, EAAT3, MEAAC1 \[60-62\] Slc20a1 2.8 ± 0.4 Mouse: Present in D Glvr-1, Glvr1 \[63\] Slc25a22 2.8 ± 0.3 2.7 ± 0.3 Human: SI \- \[64\] Slc26a2 2.5 ± 0.3 Human: C \> SI Dtd, ST-OB \[65\] Slc26a6 5.4 ± 0.2 6.1 ± 0.3 Mouse: D, J, I \> C; Human: D\>C\>J\~I, SI \> C CFEX; Pat1 \[66, 67\] Slc28a3 3.5 ± 1.2 3.5 ± 1.1 Human: D \> J \> I \> C Cnt3 \[68\] Slc2a1 3.9 ± 0.6 3.10 ± 0.3 Rat: Low expressed in SI; Human: expressed in colon carcinoma Glut1; Glut-1 \[69, 70\] Slc2a2 4.0 ± 0.4 2.6 ± 0.5 5.2 ± 0.8 Human: expressed in SI Glut2; Glut-2 \[71\] Slc2a5 2.6 ± 0.4 3.2 ± 0.7 Mouse: expressed in SI; Human: expressed in J Glut5; Slc5a \[72, 73\] Slc34a2 4.0 ± 0.3 3.7 ± 0.2 4.9 ± 0.6 Mouse: expressed in SI and colon; Human: SI specific Npt2b; NaPi-2b \[35, 74\] Slc35a1 2.1 ± 0.2 2.7 ± 0.3 \- \- Slc39a10 2.2 ± 0.3 \- \- Slc40a1 5.7 ± 0.3 2.5 ± 0.5 Mouse:high in D, not detected in I; Rat: expressed in D, C; Human: highly expressed in D MTP; Ol5; Pcm; Dusg; Fpn1; MTP1; IREG1; Slc11a3; Slc39a1 \[75-77\] Slc4a7 2.4 ± 0.2 2.5 ± 0.2 Human: expressed in SI and C NBC3; NBCn1 \[78\] Slc5a1 2.1 ± 0.1 2.7 ± 0.2 Mouse and Human: Specific for SI; Rat: possible expression in C Sglt1 \[79, 80\] Slc6a14 3.0 ± 1.9 6.0 ± 1.8 7.7 ± 0.7 Mouse: SI \< C; Human: weak in C, absent in SI ATB0plus; CATB0plus \[81, 82\] Slc6a19 6.7 ± 0.6 6.8 ± 0.6 \- B\<0\>AT1 Slc6a4 3.1 ± 0.1 3.1 ± 0.2 Mouse and human: Intestinal enterocytes Htt; Sert; 5-HTT \[83\] Slc7a7 3.6 ± 0.1 3.7 ± 0.2 Mouse: expressed in J, I; human: highly expressed in SI my+lat1 \[84, 85\] Slc7a8 2.1 ± 0.4 5.2 ± 0.7 5.7 ± 0.8 Mouse: expressed in SI LAT2 \[86\] Slc7a9 3.8 ± 0.2 3.4 ± 0.2 Mouse and human: expressed in SI CSNU3 \[85\] Slco1b2 2.1 ± 1.1 \- OATP2; Oatp4; lst-1; OATP-C; mlst-1; Slc21a6; Slc21a10 ^a^D = Duodenum, J = Jejunum, I = Ileum, C = Colon, SI = Small intestine, - = No data found in the literature ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Differently regulated (p ≤ 0.05) non-SLC transporters with a minimum absolute fold change of four in any pair-wise comparison of adjoining intestinal segments. Bold cases indicate fold changes of five or higher (Means ± SDs, log2 scale, n = 3 or 9, respectively.). ::: Gene Name D\>J D\<J J\>I J\<I I\>C I\<C SI\>C SI\<C Literature^a^ Gene aliases References --------------- --------------- ----------- ----------- --------------- --------------- ----------- --------------- --------------- ------------------------------------------------------------------------------------------ --------------------------------------------------- ---------------- Aldh9a1 2.4 ± 0.1 \- ESTM40; TMABA-DH \- Apoc2 3.7 ± 0.5 2.5 ± 0.5 4.9 ± 0.6 Mouse: expressed in intestine; Rabbit: expressed in J, D; Human: expressed in J \- \[87-89\] Aqp4 4.6 ± 0.8 4.3 ± 0.3 Mouse: present in SI in crypt cells, and C in epithelial surface cells; rat: SI, C MIWC; mMIWC \[90, 91\] Aqp5 2.9 ± 0.4 Rat: D only \- \[91\] Aqp8 4.8 ± 0.6 3.9 ± 0.4 Rat: SI and C \- \[91\] Atp2a3 2.1 ± 0.2 Human: C \> SI Serca3; SERCA3b \[92\] Atp7a 2.5 ± 0.5 4.0 ± 0.4 Human: Highly expressed in D Mo; br; Blo; I14; blotchy; mottled; MNK; brindled \[93\] Chrna1 2.5 ± 0.4 \- Acra; Achr-1 \- Chrne 2.0 ± 0.4 \- Acre \- Csng 2.0 ± 0.8 \- Csn1s2a; Csn1s2b \- D11Ertd18e 4.7 ± 0.5 4.7 ± 1.0 \- \- \- Fabp1 5.2 ± 1.1 4.7 ± 1.1 8.3 ± 0.10 Mouse: D, proximal J\>I, nothing in distal I; Human: highest in J, present in D, J, I, C Fabpl; L-FABP \[94\] Gfpt1 2.2 ± 0.4 2.4 ± 0.4 Mouse: expressed in SI and C; Human: C \> SI GFA; GFAT; Gfpt; GFAT1 \[95, 96\] Gfpt2 2.1 ± 0.4 Human: SI and C very low expressed \- \[96\] Gria1 2.2 ± 0.5 \- Glr1; Glr-1; GluRA; Glur1; HIPA1; GluR-A; Glur-1 \- Grm3 2.4 ± 1.2 2.3 ± 1.2 \- Gprc1c; mGluR3 \- Gsr 2.1 ± 0.3 \- Gr1; Gr-1 \- Kcna3 2.9 ± 0.6 2.0 ± 0.4 \- Mk-3; Kv1.3; Kca1-3 \- Kcnk6 Mouse: C \> SI Toss; Twik2; \[97\] Kctd3 2.7 ± 1.2 2.7 ± 1.3 \- NY-REN-45 \- Pkd2 2.4 ± 0.5 \- \- \- Pln 2.6 ± 0.5 \- PLB \- Rbp2 2.8 ± 0.2 3.0 ± 0.3 4.7 ± 0.5 \- Rbp-2; Crbp-2; CrbpII \- Rbp7 2.9 ± 0.3 2.7 ± 0.3 \- CRBP-III \- Saa3 3.5 ± 0.6 4.0 ± 0.4 Mouse and Human: Expressed in SI and C Saa-3 \[98\] Stard5 2.9 ± 1.0 \- D7Ertd152e \- Stx1b1 3.0 ± 1.2 \- Stx1bl \- Trpm7 2.0 ± 1.0 \- CHAK; PLIK CHAK1; Ltpr7; Ltrpc7; \- Xdh 2.8 ± 0.3 Mouse: Strong in SI Xor; Xox1; Xox-1 \[99\] ^a^D = Duodenum, J = Jejunum, I = Ileum, C = Colon, SI = Small intestine, - = No data found in the literature ::: There are particular transporters that are clearly segment-specific. However, most differentially regulated transporters are similarly expressed along the whole small intestine -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- A more detailed pair-wise comparison between the adjoining segments indicated that the differences between adjoining segments, regarding changes of expression levels of transporters, increased along the intestine (cf. Figure [2](#F2){ref-type="fig"}, Tables [2](#T2){ref-type="table"} and [3](#T3){ref-type="table"}). Most transporters that were differently expressed between the small intestine and the colon were similarly expressed along the length of the small intestine, but differently changed between the ileum and the colon; however, some transporters were clearly expressed in a region-specific manner. Ferroportin 1 (Slc40a1) was strongly over-expressed in the duodenum compared to the other segments. Besides Slc40a1, the intestinal phosphate transporter (Slc34a2) had the most pronounced fold change between jejunum and the duodenum. The expression level of Slc34a2 increased along the small intestine, but was again more lowly expressed in the colon than in the ileum. Fabp1 is similarly expressed in the duodenum and the jejunum, but significantly (p ≤ 0.05) decreased in the ileum and even more in the colon. Besides the Na+ dependent ileal bile acid transporter (Slc10a2) and Slc34a2, the sodium transporter Slc5a8, a tumor suppressor gene \[[@B21]\], was also more highly expressed in the ileum compared to the jejunum, but similarly expressed in the colon. On the other hand, Fabp1 is the most highly expressed gene involved in transport activity in the jejunum compared to the ileum. This is followed by D11Ertd18e, which has been associated to the sugar transporter super family based on its protein structure, and the facilitated glucose transporter member 2 (Slc2a2). Tables [2](#T2){ref-type="table"} and [3](#T3){ref-type="table"} show all transporters that were differently regulated along the intestine and whose fold changes were at least four-fold. Members of the ABC transporter family have not been integrated in the tables as they have been presented earlier in detail \[[@B10]\]. Most transporters are similarly expressed in mice and humans ------------------------------------------------------------ In order to assess if the expression levels in the mouse are a suitable estimator for the situation in humans our data were compared to publicly available human microarray data obtained with a custom-array. Overall, 20 % of all annotated transporters in mouse were compared. Previous studies have shown that there is a good correlation between data obtained with the Affymetrix platform and the data obtained with this custom-array \[[@B22]\]. The majority of common orthologous transporters were similarly expressed in the small intestine and the colon in both mice and humans. Some genes, however, were no less than four-fold up-regulated in at least one segment in the mouse, but were not identified as being differently regulated in humans except for low affinity Na-dependent glucose transporter Slc5a2 (cf. Table [4](#T4){ref-type="table"}). Other studies in humans not using microarrays, however, suggest that most of these genes have a similar expression profile in mice and humans (Table [4](#T4){ref-type="table"}) indicating that the fold changes measured with the custom-array may to a certain extent underestimate the true fold changes as shown earlier \[[@B22]\]. ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Comparison of the relative expression levels of transporters in mice and humans. Only genes with at least one four-fold difference in pair-wise comparisons are shown ::: Gene Name Human^a^ Mouse^a^ --------------- -------------- -------------- ------------ ------------ ------------ SI \> C SI \> C D \> C J \> C I \> C AQP5 x SLC19A3 x SLC26A6 x x x x SLC28A3 x x x x SLC2A2 x x x SLC34A2 x SLC37A4 x SLC6A4 x x x SLC7A7 x x x SLC7A8 x x x x SLC7A9 x x x C \> SI SI \< C D \< C J \< C I \< C AQP4 x x x x AQP8 x x x x PKD2 x SLC26A2^b^ x SLC10A2 x SLC2A1 x x x x SLC6A14 x x SLC35A1 x SLC40A1 x SLC5A1^c^ x ^a^Published expression data are listed in tables 2 and 3. D = Duodenum, J = Jejunum, I = Ileum, C = Colon, SI = Small intestine, - = No data found in the literature. ^b^According to Haila et al. \[65\] highly expressed in the human colon and low in the human small intestine. ^c^According to Wright et al. \[79\] specific for human and mouse small intestine. ::: The regulation of expression in whole intestinal tissues is a good indicator of the combined changes in epithelial crypt and villus cells ----------------------------------------------------------------------------------------------------------------------------------------- In order to assess whether the expression profiles of genes in whole tissue reflect their expression in the epithelium, we determined the expression profile of 12 genes in villus and crypt cells using a combined laser micro-dissection and RT-PCR approach. To validate this method we measured a marker for the villi (aminopeptidase N/Anpept), for crypts in the small intestine, i.e., Paneth cells (defensin related cryptdin 5/Defcr5) and crypts in the small intestine and colon (caudal type homeo box 1/Cdx1). Based on the expression levels of the three crypt/villus markers, RNA obtained from laser dissection of the small intestine seems to be highly enriched of region specific material, i.e. villi samples do not contain a notable portion of RNA originating from crypt cells (cf. Figure [3](#F3){ref-type="fig"}). ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Expression profiles of tissue-specific markers along the intestine. Blue, dash-dotted lines indicate the relative expression in the four intestinal segments (D, J, I, C) using microarrays (mean ± SD, n = 3 pools of 10 mice each). Black, continuous lines indicate the relative expression in the four intestinal segments in the crypts and villi using RT-PCR (mean ± SD, n = 5). Please note that connecting the black lines between crypts and villi are not logical connections and only for visual support. ::: ![](1471-2164-6-69-3) ::: In order to assess the expression profiles of transporters measured in whole intestinal tissue with microarrays, we verified the mRNA levels of nine transporters (Abca1, Abcc1, Abcc3, Abcg8, Slc10a2, Slc28a2, Slc2a1, Slc34a2 and Slc5a8) measured in the same RNA sample preparation by RT-PCR. The expression profiles measured with microarray and RT-PCR had a good concordance. In general, we observed that the profiles measured in the whole tissue in the four segments reflected the average expression in the crypts and the villi. Influence of the crypt-villus axis on the expression of transporters -------------------------------------------------------------------- To test if there is a relationship between the expression profile along the intestine and along the crypt-villus axis, we measured the expression of a subset of transporters in laser-dissected material. We selected transporters which were, based on microarray and RT-PCR results of the whole tissue (cf. Table [5](#T5){ref-type="table"}), not differentially expressed along the intestine (Abca1, Abcc1, Abcc3), small intestine specific (Abcg8), large intestine specific (Slc2a1), either specific for the anterior small intestine (Slc28a2) or posterior part of the whole intestine (Slc10a2, Slc34a2, Slc5a8). All nine transporters examined in laser-dissected material tended to conserve their tissue specificity along the intestine (cf. Table [5](#T5){ref-type="table"}, Figure [4](#F4){ref-type="fig"}). In other words, none of the examined transporters was crypt-specific in one segment and villus-specific in another. Abcc1 and Slc2a1 were crypt-specific, whereas Slc34a2 villus-specific in at least three segments. Abcg8 and Slc28a2 showed a villus-specificity in the posterior part of the small intestine, while Slc5a8 was only in the jejunum villus-specific and Slc10a2 in the ileum. ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Segment and tissue specificity of transporters along the A-P and crypt-villus axis. ::: Transporter Segment-Specificity^a,M^ Segment-Specificity^a,R^ Tissue-Specificity^b^ ----------------- ------------------------------ ------------------------------ ----------------------- ------------------ ------------------ ------------------ Duodenum Jejunum Ileum Colon Abca1 D=J=I=C J=I=C\>D Crypts\>Villi ND ND ND Abcc1 D=J=I=C I=C\>D=J Crypts\>Villi Crypts\>Villi Crypts\>Villi ND Abcc3 D=C\>J=C\>I=C I=C\>D=J ND Crypts\>Villi ND Crypts\>Villi Abcg8 J\>I=D\>C D=J=I\>C ND Villi\>Crypts Villi\>Crypts ND Slc10a2 I\>C\>D=J I\>C\>D=J ND ND Villi\>Crypts ND Slc28a2 D\>J\>I\>C D=J=I\>C ND Villi\>Crypts^x^ Villi\>Crypts^x^ ND Slc2a1 C\>D=J=I C\>I\>D=J ND Crypts\>Villi Crypts\>Villi Crypts\>Villi^x^ Slc34a2 I\>C\>J\>D I\>C=J\>D ND Villi\>Crypts Villi\>Crypts Villi\>Crypts Slc5a8 C\>I\>D=J C=I\>J\>D ND Villi\>Crypts ND ND ^a^Expression in the four intestinal segments (D, J, I, C) was measured in whole tissue using microarrays (M) and RT-PCR (R) (n = 3, p ≤ 0.05). ^b^Tissue specificity was determined in laser dissected crypts and villi (n = 5, where p ≤ 0.05 or if indicated with a superscript \"\\" p ≤ 0.10). \"ND\" indicates that no significant difference in expression levels were measured between the crypt and villus epithelial cells. ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Expression profile of selected transporters along the intestine. Blue, dash-dotted lines indicate the relative expression in the four intestinal segments (D, J, I, C) using microarrays and brown dashed lines using RT-PCR (mean ± SD, n = 3 pools of 10 mice each). The microarray results for the ABC family members are according to Mutch et al. \[10\]. Black, continuous lines indicate the relative expression in the villi (V) and crypts (C) as determined with laser dissected material and RT-PCR (mean ± SD, n = 5 individual mice). The black lines between crypts and villi are meant only for visual support. ::: ![](1471-2164-6-69-4) ::: Discussion ========== A crucial step in classifying genes into different molecular functions is the use of a consistent and universal classification system and a precise annotation of Affymetrix probe sets. Therefore, we included the following features in our analysis: i) Mapping of single probes to all transcripts referred to in the UniGene database and attributing of annotation tags to each probe set, and ii) assigning of flags to differentially regulated probe sets. Similar to Chalifa-Caspi et al. \[[@B23]\], we have seen that the annotation provided by Affymetrix (NetAffx) is not entirely accurate. In the Affymetrix annotation a probe set is represented by at most one UniGene cluster, while, conducting our own mapping of Affymetrix probes onto UniGene, we observed that in a number of cases multiple UniGene identifiers can be associated to a given probe set. Based on our quality criteria, 48 % of all probe sets in the murine genome were found to have a \"high\" quality annotation on the microarrays used in this study. Thus, selecting only \"high\" quality genes establishes a high degree of confidence regarding the correct annotation of probe sets. We observed that the genes on the microarrays are not represented by an equal number of probe sets. Hence, by counting each probe set as a unit in the GO classification system the way it is done by a majority of GO classification programs such as MappFinder, OntoExpress and the one provided by NetAffx, a bias may be introduced \[[@B24],[@B25]\]. Moreover, to our knowledge, none of the programs publicly available provides a confidence assessment regarding the significantly regulated functional classes. Therefore, we developed a software, the so-called ISREC Ontologizer (Io), that takes these factors into account. The use of the flags (i.e., 1, F, G) provided with the Io software permits such an assessment to be made. As indicated in Figure [2](#F2){ref-type="fig"}, a significant portion of the results have a flag \"F\" and were, therefore, considered ambiguous. Some of these \"flagged\" results, based on UniGene identifiers, could be due to the fact that some of the probe sets are associated with the same UniGene cluster, but are specific for different splice variants. However, we observed a similar portion of ambiguous results (i.e., flag \"F\") when doing the classification based on RefSeq identifiers and concluded that the ambiguity associated with flagged genes may stem from probe design rather than the differential expression of splice variants \[[@B26]\]. Some transporters are subject to post-translational modification and of course, one of the processing steps is translocation to the membrane. However, if one considers mRNA to be representative of function, then the majority of transporters can be expected to have similar activity along the length of the intestine. In this regard, it is interesting to note that a remarkable fraction of transporters annotated with \"carrier activity\" are differentially regulated along the A-P axis and could therefore significantly influence the absorption of carrier-mediated bioactive compounds. Our GO classification results indicate that, in the context of the whole transcriptome, transporters do not seem to be an especially dynamically regulated class, which differs slightly to the conclusions made by Bates and colleagues \[[@B9]\]. However, a direct comparison is difficult to make, as the classification system used by Bates et al. was not in accordance with the official GO classification system, and this may underlie the different biological conclusions made between the two studies. Furthermore, more importantly, while that study included only 4 % of all annotated transporters (6), this study covers 76 % and, provides the first extensive overview of the genomic profiles of transporters in the intestine. Similar to their study, the most pronounced difference between adjoining regions was observed for the ileum and the colon. Although, in general, the differentially expressed transporters were more highly expressed in the small intestine than in the colon, a surprisingly high number of transporters were identified as being more highly expressed in the colon than in the small intestine \[[@B13]\]. Many of these transporters have been described as highly expressed in the colon and involved in the bi-directional transport of electrolytes and fluids, which is the principal role of the colon \[[@B27]\]. However, the majority of known drug and nutrient transporters or carriers of bioactive compounds \[[@B28]\] are not differently expressed between the small intestine and the colon. The exploitation of transporters, which are highly expressed in the colon or at similarly high levels as in the small intestine is especially interesting for compounds such as proteins and peptides, which are susceptible to the high enzymatic activity present in the small intestine \[[@B29]\]. For example, Slc6a14, which transports neutral and amino acids, might be an interesting candidate to be targeted for carrier-mediated absorption. Generally, we observed that most of the differentially regulated \"carriers\" were not specifically over-expressed in any of the three small intestinal segments, but overall more highly expressed in the small intestine versus the colon. Thus, the ileum, which contains smaller concentrations of pancreatic enzymes, could be a suitable tissue for compounds that are sensitive to the activity of pancreatic enzymes. Genes classified as having \"carrier\", \"channel\", \"ion transport\" and \"electron\" and \"neurotransmitter\" activity are most likely to have different expression levels between the small intestine and colon and may, hence, play a more significant role in the difference of active drug and nutrient transport in these two anatomically distinct regions. The expression profiles of the most strongly regulated genes (i.e. absolute fold change of 4 between the segments) agreed with previous findings in mice (cf. Table [2](#T2){ref-type="table"}). Moreover, comparing transporter expression profiles in mice and humans suggests that gut transcriptional profiles obtained in mice are a valid estimator for the situation in humans (Table [2](#T2){ref-type="table"} and Table [3](#T3){ref-type="table"}). There are several methods that can be used to characterize the patterns of expression of a target gene. Traditionally, labeled antisense specific probes have been widely used for in situhybridization to detect specific gene expression in the intestinal epithelium \[[@B12],[@B30]\]. This technique, however, is very labor-intensive when several candidates are being tested. There are several techniques used to isolate the epithelial layer from the lamina propria of the intestine \[[@B31]\]. These methods use mechanical procedures and detergents and/or EDTA to detach the epithelial layer. Separation of crypts from villi epithelium can be achieved following adequate separation procedures. However, alterations of gene expression patterns depending on the procedure as well as cross-contamination from different tissue compartments are possible. Similarly, laser dissection microscopy has been developed to isolate single cells \[[@B32]\]. A key step in this procedure is the selection of an adequate fixation procedure that prevents RNA degradation and preserves tissue histological integrity \[[@B33],[@B34]\]. The protocols used in the present study allows the assessment of a minimally altered gene expression pattern that is concordant with in situhybridization experiments (M. Rumbo unpublished data). Within this context, our data suggest that the changes in expression profiles measured in whole intestinal tissue extracts are a suitable predictor for changes in epithelial cell-only gene expression along the crypts and the villi; however, it should be mentioned that all genes in our comparison were already known to be expressed in the epithelium. This good correlation could reflect the fact that these genes are solely expressed in the epithelium or that the epithelium contributes much more to the isolated RNA than underlying layers of intestinal tissues and, therefore dominates the genomic profile of the intestine. Thus, for genes which are not solely expressed in the epithelium the prediction based on measurements in whole tissue might not be appropriate. For some compounds, depending on the enzymatic stability and solubilization characteristics, targeting of carriers in distinct intestinal regions may selectively improve their absorption and bioavailability. Thus, we analyzed the expression profiles of known drug carriers along the intestine. The major part of inorganic phosphorus (P~i~) absorption from the small intestine occurs via a Na^+^-dependent phosphate co-transporter, NaP~i~-II~b~encoded by the gene SLC34A2. In vivo*pharmacokinetic studies in rats have indicated that foscarnet is transported across the enterocytes by the NaP~i~-II~b~system and thus acts as a competitive inhibitor of P~i~uptake \[[@B28]\]. Slc34a2 mRNA has been shown to be expressed in the small intestine and the colon \[[@B35]\], which is in accordance with our findings. Our studies further demonstrate that the expression significantly increases along the A-P axis of the small intestine. As a consequence, the most efficient transport rate is expected to arise in the ileum. In addition to the findings by Hilfiker et al. \[[@B35]\] that Slc34a2 protein is expressed on the apical membrane of mature enterocytes in the duodenum, we observed that Slc34a2 mRNA is villus-specific along the whole intestine except for the duodenum, reinforcing the concept that transporters expressed in the villi epithelium tend to be implicated more in absorptive processes. Slc10a2 (ASBT), the primary carrier for Na^+^-dependent bile salt uptake from the intestinal lumen by the ileum, has been shown to be present in the brush-border membrane of the terminal part of the ileum, but little is known concerning its regulation along the intestine \[[@B36],[@B37]\]. Various studies indicate that based on substrate examples ASBT could be an important target for increasing the bioavailability of various bile acid conjugates \[[@B28]\]. Our findings confirm that Slc10a2 is especially highly expressed in the ileum compared to proximal small intestinal regions; however a comparable level was measured in the colon. Interestingly, only in the ileum was Slc10a2 identified as being villus-specific which could be an indication for its specific role in the ileum. The four members of the ABC families, Abcg8, Abca1, Abcc1 and Abcc3, that we studied are known to act as secretion pumps in the intestine \[[@B14],[@B38],[@B39]\]. In accordance with earlier findings, Abcc3 expression increases from the anterior to posterior part of the intestine \[[@B10],[@B40]\]. On the other hand, our findings for Abcc3 did not indicate an increase in mRNA expression from the crypts to the villi as shown by Rost et al. \[[@B40]\] in rats. The mRNA levels of Abcc1 along the crypt-villus axis were in accordance with protein expression \[[@B14]\]. Generally, all four members may play a significant role along the whole small intestine as secretion pumps, whereas in the large intestine, Abcg8 is not expected to be involved in the secretion of compounds. The transporter Slc28a2 is known to be involved in the transport of nucleoside analogues such as some antiviral compounds \[[@B41]\]. Nucleosides are relatively hydrophilic molecules and their ability to be transported across cell membranes is a critical determinant of their metabolism \[[@B42]\]. Our data indicate that similar uptake rates can be expected along the whole intestine when targeting this carrier, as it is similarly expressed in the gut. On the other hand, Slc2a1, which was significantly (p ≤ 0.05) more highly expressed in the colon, might be an appropriate drug-carrier for compounds sensitive to pancreatic enzymatic degradation. It has to be noted though that the protein product of SLC2A1, GLUT1, is not detectable in healthy human colon tissue, but only in colon cancer serving as a marker for poor prognosis \[[@B43],[@B44]\]. Thus, it may be an interesting drug-target/drug carrier in individuals with colon cancers. In contrast to GLUT1, SLC5A8 has been shown to be highly expressed in the colon and to act as a tumor suppressor gene. SLC5A8 is silenced by methylation in most human colon tumors and reintroduction of this gene leads to growth suppression \[[@B21]\]. Recent studies have shown that SLC5A8 acts as a Na+-coupled transporter for short chain fatty acids and monocarboxylates. Its transport rate is strongly inhibited by various drugs such as ibuprofen. We have shown for the first time that this gene is expressed at similar levels in the distal regions of the gut and seems to be preferentially expressed in the villi of the small intestinal enterocytes. Summarizing, this Na+-coupled transporter may be involved in the active transport of known monocarboxylate drugs and serve as a valuable drug-carrier for new drugs along the whole intestine. Conclusion ========== In conclusion, we have characterized the mRNA expression profiles of 76 % of all currently known transporters along the A-P axis of the gut. We have identified various transporters that could serve as carriers for biologically active compounds. The significant regional specificities are likely to correspond to functional differences along the length of the intestinal tract. Finally, we have presented a comprehensive analysis of regional variations in gene expression using whole gut tissue that is sufficiently sensitive to provide a good assessment of relative changes in regional gene expression in epithelial cells; however, for the functionality of a given transporter to be unequivocally determined, it will be critical to examine these epithelial cells in a cell-type specific manner. Methods ======= Animals and tissue handling --------------------------- 8 week-old male Hsd:ICR(CD-1) mice (Harlan, Netherlands) were provided by AMS Biotechnology (Lugano, Switzerland). Thirty animals were divided into 3 pools. The small intestine was extracted and divided into three sections, where the first 2--3 cm after the stomach comprised the duodenum, the middle third the jejunum, and the section before the ileo-ceco-colic junction comprised the ileum. The colon was treated as a single intestinal section. Gene expression analysis using the murine Mu74v2 GeneChips ---------------------------------------------------------- Gene expression was measured as previously described \[[@B10]\]. Briefly, total RNA extracts were provided by AMS Biotechnology (Lugano, Switzerland) and RNA extraction was performed identically for each pool of mice. RNA was then re-purified, according to manufacturer\'s instructions using the Nucleospin kit and contaminating genomic DNA was removed by DNase1 treatment (Macherey-Nagel AG, Oensingen, Switzerland). The quality and quantity of RNA was determined using Agilent 2100 Bioanalyser (Agilent Biotechnologies, Germany) (1.6 -- 2.0 ratio for 28/18S and no significant amount of metabolized products). For each murine gut tissue section, 5 μg total RNA was used as the starting material for all individual samples. Labeling and fragmentation of cRNA, array hybridization and scanning was performed according to the protocol by Affymetrix. Fluorescence values from scanning were analyzed with Affymetrix Gene Expression Analysis Software (MAS 5.0). The complete data set is publicly available at <http://www.ncbi.nlm.nih.gov/geo/> through the accession number GSE849. Laser dissection microscopy --------------------------- The mRNA expression levels of selected transporters (Abca1, Abcc1, Abcc3, Abcg8, Slc10a2, Slc28a2, Slc2a1, Slc34a2 and Slc5a8) were measured in the crypts and villi of the intestinal mucosa of five 8 week-old male Hsd:ICR(CD-1) mice (Harlan, Netherlands). To assess the performance of the micro-dissection, specific markers for the crypt-villus axis were used (Anpept for villi, Cdx1 for crypts and Defcr5 for crypts in the small intestine, namely for Paneth cells) \[[@B45]-[@B48]\]. Immediately after sacrificing the animals, the intestinal tract was removed and regions classified similar to the microarray tissue sample definitions (cf. above). Sections of one cm length were cut and incubated overnight in zinc fixative/sucrose 30% solution (5g ZnCl~2~, 6g ZnAc~2~X2H~2~O, 0.1g CaAc~2~in 1L of 0.1M Tris pH 7.4). Afterwards, they were embedded in OCT and frozen by immersion in liquid nitrogen. 20 μm frozen sections were cut and mounted on Leica membranes for dissecting microscopy (Leica Microsystems, Wezlar, Germany), fixed in 96 % ethanol for 30 s and colored for an equivalent time with Mayer\'s hematoxylin solution. Membranes were then rinsed in water for 1 minute, transferred for 10 s to 70 % ethanol, followed by 96 % ethanol and air dried. Samples were processed using a laser dissecting microscope (Leica Microsystems), coupled to a CCD camera. Microdissected samples were collected in a tube cap placed below the sample holder. Samples were collected in 20 μL RNA lysis buffer. Total RNA was extracted using the total RNA extraction Nucleospin II kit by Machery-Nagel (Oensingen, Switzerland). Real-time PCR (RT-PCR) ---------------------- Reverse transcription (RT) was performed using Superscript II (Gibco BRL). RT-PCR amplification was performed using an ABI 5700 machine (Applied Biosystems, Foster City, CA, USA) with the following thermal cycling conditions: 2 min at 50°C, 10 min at 95°C, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. The quality of all RNA samples were examined with the Agilent 2100 Bioanalyser (Agilent Biotechnologies, Germany). All samples were standardized to equal RNA concentration using the RiboGreen RNA quantification kit (Molecular Probes, Leiden, Netherlands) and measured in duplicates. Cycle to cycle fluorescence emission was monitored and quantified using the GenAmp software provided by Applied Biosystem. Data was normalized to Gapdh. Primer probes were purchased from Applied Biosystems as Assays-on-Demand^SM^(Applied Biosystems, Foster City, CA, USA) Data analysis ------------- ### GEA analysis and ANOVA of microarray data Differential gene selection was determined using a Global Error Assessment (GEA) method of analysis \[[@B49]\]. Genes were considered as differentially regulated along the intestinal tract if the p-value = 0.05. GEA assumes that the variance of the absolute intensity of a given gene is a function of the absolute intensity. Therefore, rather than treating each gene on the microarray as a unique and unrelated element, neighboring genes are binned into groups of 200 based on similar intensity signals and the mean squared error is calculated for each bin. For tables [2](#T2){ref-type="table"} and [3](#T3){ref-type="table"} standard deviations were calculated for each gene independently. ### Analysis of RT-PCR data A one-way ANOVA and pair-wise comparisons based on Tukey\'s Honest Significant Difference method (p- values as indicated in the figures and tables) were used to confirm differences in gene expression. Annotation of probe sets and functional classification of genes --------------------------------------------------------------- Whereas in NetAffx the UniGene identifier is determined according to the probe set\'s representative sequence (i.e., the sequence used to design the probe set), we have independently mapped the probes and attributed a quality tag depending on the level of specificity \[[@B50]\]. In order to better assess the significance of the findings it is necessary to know if the different probe sets supposed to represent the same transcript give the same result and if all the probes in a probe set still represent one and the same transcript according to the most recent reconstruction of the mouse transcriptome. Therefore all probes of the Affymetrix Mu74v2 GeneChip were mapped to the latest release of UniGene clusters (Unigene build number 137, EST db, HTC db, RefSeq db and mRNA db) by an in-house developed tag-matching software as described earlier \[[@B51]\]. On this basis, every probe set was assigned to one of four probe set \"quality classes\". The top class \"high\", the only one used in this work, requires that all the probes of the same probe set have perfect sequence matches to at maximum two UniGene clusters, which was the case for 17625 probe sets, as defined elsewhere \[[@B51]\]. Structure of classification program ----------------------------------- Following the UniGene annotation, probe sets were associated to the nodes they represent in the gene function tree defined by the Gene Ontology (GO) project (<http://www.geneontology.org/>, version of April 2004). For probe sets mapping to two different UniGene clusters, both clusters were considered for GO annotation. A Java program called ISREC ontologizer (Io) was used to efficiently browse and analyze the data. Io is a general purpose Java program for classification of microarray results according to an annotation system <http://www.io.isb-sib.ch/>, which will be described in detail elsewhere. In short, for a given list of (differentially regulated) probe sets (features), Io shows their distribution over all GO classes subdivided by classes of probe set quality and evaluates the statistical significance of overrepresentation of a GO class. This analysis can be done with the features as individual entities or pooling in groups those that represent the same UniGene cluster to study the degree of agreement between probe sets of the same cluster. In this case, three counts are provided for each GO class: 1/F/G (\"one\"/\"flagged\"/\"good\"). The \"One\" counter is incremented when the group to which a feature belongs contains only that feature. For groups with more than one feature, the \"good\" counter is incremented when all features are present in the selected list of probe sets; otherwise the \"flagged\" counter is incremented and indicates an ambiguous experimental result. Significant changes (p ≤ 0.05) for each GO node were calculated using a Fisher\'s exact test using the \"1\" or \"F\" or \"G\" counts \[[@B52]\]. Comparison to human data ------------------------ The mouse microarray data were compared to publicly available microarray data (the data set is available at <http://www.ncbi.nlm.nih.gov/geo/> through the accession number GSE1368) of the human small intestine and the colon. For all \"high quality\" Affymetrix probe sets the corresponding human homologue genes were identified on the human chip using the LocusLink identifier and the Homologue mapping table <ftp://ftp.ncbi.nih.gov/pub/HomoloGene/>. 224 orthologous transporters were identified as being in common between the two platforms. Then, the relative fold changes observed in the mouse were compared to the human data. In case of multiple probes representing the same gene average expression values were used. In the case of the mouse data, relative fold changes were calculated by comparing the expression levels of each independent small intestinal segment to the colon and the average of all three segments to the colon as it was unclear which small intestinal segment was compared to the colon in the human study. Only genes four times or higher expressed in any segment compared to another segment were considered in this comparison. Authors\' contributions ======================= PA contributed to the design and coordination of the study, carried out the laser dissection, comparison of mouse and human microarray data and expression data in the literature, participated in the development of the ISREC ontologizer, was responsible for high level analysis of microarray data and drafted the manuscript. TS participated in the development of the ISREC ontologizer and the annotation of the probe sets, and implemented its methods in the Java environment. DMM contributed to the design of the study and participated in the editing of the manuscript. MR set up the protocols for the LDM and participated in the laser dissecting. VP was mainly responsible for the annotation of the probe sets. RM performed the global error assessment (GEA). MD participated in the development of the ISREC ontologizer and participated in the editing of the manuscript with respect to the statistical methods. GW contributed to the design and planning of the study, to the discussion and interpretation of results, and to the writing of the manuscript. MR contributed to the original conception, design and coordination of the study. Further contribution was made to choosingproject goals, experimental approaches, developing technical resources towards advancement of the project, and finally revising the manuscript prior to publication. Note ==== In the following, the term \"transporter\" will refer to all genes that are classified as genes with \"transporter activity\" according to the GO system. At the date of analysis the ATP-binding cassette (ABC) transporters were not classified as genes with \"transporter activity\". Thus, this class has not been integrated the GO analysis. However, in the analysis of expression levels on a gene by gene basis the ABC transporters have been integrated. Please note that according to the official human HUGO and mouse MGI system we will refer to human gene symbols in upper case, and to mouse genes with a first letter in capital and the subsequent ones in lower case. Acknowledgements ================ This work was supported by grants from the Nestlé Research Center in Lausanne and the National Center of Competence in Research (NCCR) Molecular Oncology, a research program of the Swiss National Science Foundation. We would like to thank Muriel Fiaux for technical assistance.	PubMed Central	2024-06-05T03:55:57.508356	2005-5-10	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1145182/", "journal": "BMC Genomics. 2005 May 10; 6:69", "authors": [ { "first": "Pascale", "last": "Anderle" }, { "first": "Thierry", "last": "Sengstag" }, { "first": "David M", "last": "Mutch" }, { "first": "Martin", "last": "Rumbo" }, { "first": "Viviane", "last": "Praz" }, { "first": "Robert", "last": "Mansourian" }, { "first": "Mauro", "last": "Delorenzi" }, { "first": "Gary", "last": "Williamson" }, { "first": "Matthew-Alan", "last": "Roberts" } ] }
PMC1145183	Background ========== Transfer back to the hospital is a common problem for patients admitted to skilled nursing facilities (SNF), with an estimated 18% of patients transferring to the hospital within the first 30 days of admission, and 38% within the first 90 days \[[@B1]\]. Avoiding hospital transfer in frail older individuals is desirable if care can be provided in the SNF because of the many adverse effects associated with hospitalization \[[@B2]\]. In addition, evidence suggests that SNF patients are often inappropriately hospitalized. One study which reviewed 100 unscheduled transfers to hospital found that 36% of transfers to the emergency room and 40% of hospital admissions from SNF were inappropriate, with poor quality of care in the SNF at least partially implicated \[[@B3]\]. Another study of processes and outcomes of care for Medicare SNF patients with acute heart failure found that if the patient\'s condition changed during the night shift, when staffing is generally lower, the odds of rehospitalization or emergency department evaluation were increased fourfold, suggesting an implicit connection between nurse staffing and rehospitalization rates among SNF patients \[[@B4]\]. This implicit relationship between nurse staffing and rates of hospital transfer was made explicit in the Center for Medicare and Medicaid Services (CMS) December 2001 report to Congress entitled \"Appropriateness of Minimum Nurse Staffing Ratios in Nursing Homes \[[@B5]\].\" This report argued that registered nurse, total licensed staff and nurse assistant staffing levels are related to rehospitalization rates in short-stay nursing home patients, and that certain staffing thresholds exist, below which quality problems are more likely. Since then, the Institute of Medicine, in a report entitled \"Keeping Patients Safe: Transforming the Work Environment of Nurses,\" supported the adoption of the minimum staffing ratios found in the report to Congress \[[@B6]\]. In addition, in at least one instance, state legislators have proposed minimum nurse staffing levels consistent with those recommended in the CMS \[[@B7]\]. To test the hypothesis that reductions in hospital transfer rates would offset increased labor costs associated with higher staffing, we created a simulation model to gauge the costs and effects of recommended minimum staffing ratios for short-stay nursing home patients. Methods ======= Model ----- We compared two scenarios for patients newly admitted to skilled nursing facilities from the hospital: 1) median staffing and 2) recommended staffing. The median staffing scenario used median staffing time for registered nurse (RN), licensed staff (RN plus licensed practical nurse (LPN)) and nurse assistant (NA) hours per patient day as defined in the CMS report to Congress. These median values per patient day were 2.02 hours for NAs, 1.02 hours for RNs plus LPNs, and 0.38 hours for RNs \[[@B5]\]. We used median staffing instead of mean staffing levels because the staffing data were highly skewed, and we considered the median staffing data more representative of most U.S. facilities. The recommended staffing scenario used weighted average threshold values for staffing below which there was an increased likelihood of transfer to the hospital based on the CMS report to Congress. These recommended values were 2.37 hours for NAs, 1.14 hours for RNs plus LPNs, and 0.55 hours for RNs \[[@B5]\]. Under recommended care, NA, RN plus LPN, and RN staffing levels were higher than median values, while LPN hours were slightly lower. We developed a Markov cohort simulation to track the costs and health outcomes for patients hypothetically assigned to one of the two staffing scenarios (median versus recommended). The Markov model was constructed using DATA Professional (TreeAge Software; Williamstown, MA). The model represents the typical flow of patients from their first day in the SNF to discharge, with a probability of being hospitalized for one of five conditions (congestive heart failure, electrolyte imbalance, respiratory infection, sepsis, and urinary tract infection) while in the SNF (Figure [1](#F1){ref-type="fig"}). These conditions were identified in the CMS report as being sensitive to licensed nurse and nurse assistant staffing (see section on transition probabilities). In the model, patients discharged from the hospital return to the SNF. Any patient completing 30 consecutive days in the SNF is then discharged from the SNF, which is not an overly restrictive assumption given an average length of stay of 22.9 days for SNF patients in the year 2000 \[[@B8]\]. Patients may also die at any point in the model, with probabilities of death that are higher while in hospital than while at the SNF or after discharge from the SNF. See Appendix A (in Additional File [1](#S1){ref-type="supplementary-material"}) for further details of model design. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Schematic diagram of Markov model.The model has a cycle length of one day. All patients begin in the bolded state entitled \"SNF day 1.\" From SNF day 1, patients can transition to a second SNF day, be hospitalized for any one of five conditions, or die. If patients spend 30 consecutive days in the SNF without being hospitalized, they transition to the \"Discharged\" state, where they remain until they die. If patients are hospitalized, they spend five or six days in the hospital, depending on the condition for which they are hospitalized, unless they die while in hospital. Upon completing their hospital course, patients then return to SNF day 1. Abbreviations: SNF, skilled nursing facility; CHF, congestive heart failure; EI, electrolyte imbalance; RI, respiratory infection; UTI, urinary tract infection. ::: ![](1472-6963-5-35-1) ::: Transition probabilities ------------------------ Transition probabilities for the Markov model are shown in Tables [1](#T1){ref-type="table"} and [2](#T2){ref-type="table"}. The likelihood of being transferred to the hospital from the SNF for patients both under the median staffing and recommended staffing scenarios was based on Chapter 2 of the CMS Report to Congress on appropriateness of minimum nurse staffing ratios in nursing homes \[[@B5]\]. The authors of the report identified threshold values for the amount of NA, RN plus LPN, and RN hours per patient day below which there was an increased likelihood for facilities to be in the highest 10% of the sample for hospital transfer rates for the five conditions that were posited to be sensitive to licensed nurse and nurse assistant staffing mentioned previously. The analysis included all hospital transfers within 30 days of admission to the SNF that had one of these five conditions as the primary or secondary diagnosis. Logistic regression analyses were conducted with the object of determining the relationship between facility-level staffing (regardless of the patient mix within the facility) and the likelihood of being in the highest decile for hospital transfer rates. Odds ratios for a facility being in the highest decile for hospital transfer rates ranged from 1.31 for respiratory infection among facilities that fell below 1.05 licensed staff (RN plus LPN) hours per patient day to 2.43 for sepsis among facilities that fell below 2.40 NA hours per patient day \[[@B5]\]. Point estimates for these odds ratios were used for our base case analysis, with 95% confidence intervals used for sensitivity analysis, using the procedures discussed below. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Parameters governing transition probabilities for the Markov model ::: Parameter Base Case Biased toward recommended staffing Biased against recommended staffing Ref. -------------------------------------------------------- --------------- ---------------------------------------- ----------------------------------------- ---------- 30-day hospitalization rates from SNF\* Congestive heart failure 3.9% \[7.8%\] \[2.0%\] \[5\] Electrolyte imbalance 4.4% \[8.8%\] \[2.2%\] \[5\] Respiratory infection 3.4% \[6.8%\] \[1.7%\] \[5\] Sepsis 1.4% \[2.8%\] \[0.7%\] \[5\] Urinary tract infection 3.0% \[6.0%\] \[1.5%\] \[5\] Case fatality rates at hospital\* Congestive heart failure 1.2% \[2.5%\] \[0.6%\] \[9\] Electrolyte imbalance 1.7% \[3.4%\] \[0.9%\] \[9\] Respiratory infection 1.3% \[2.6%\] \[0.7%\] \[9\] Sepsis 3.1% \[6.1%\] \[1.5%\] \[9\] Urinary tract infection 1.5% \[3.0%\] \[0.8%\] \[9\] Annual hazard of dying for SNF and discharged states 30.3% \[15.2%\] \[60.6%\] \[10\] Parameters governing transition probabilities for the Markov model are listed here. Brackets refer to assumptions made for sensitivity analyses where data were not available. Base case values represent the best available estimate of the parameter in question. Values listed under the column \"biased toward recommended staffing\" and \"biased against recommended staffing\" represent extreme values of the parameter that are most favorable and least favorable to the recommended staffing scenario when tested in sensitivity analyses. Abbreviations: SNF, skilled nursing facility; Ref., reference. \Probabilities of hospitalization for the five conditions mentioned were varied together in sensitivity analysis, as were case fatality rates. ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Relative risk reduction for hospitalization with recommended staffing ::: Parameter* Base Case Biased toward recommended staffing Biased against recommended staffing Ref. ------------------------------ --------------- ---------------------------------------- ----------------------------------------- ---------- Nurse assistant factors\* Congestive heart failure 3.4% 5.7% 0.2% \[5\] Electrolyte imbalance 3.2% 5.4% 0.2% \[5\] Sepsis 11.3% 14.5% 6.3% \[5\] Urinary tract infection 4.3% 7.2% 0.1% \[5\] Licensed staff factors\* Electrolyte imbalance 2.9% 4.9% 0.4% \[5\] Respiratory infection 2.8% 5.0% 0.1% \[5\] Sepsis 6.1% 10.3% 0.4% \[5\] Urinary tract infection 4.7% 6.9% 1.8% \[5\] Registered nurse factors\* Electrolyte imbalance 3.0% 5.2% 0.1% \[5\] Sepsis 5.7% 9.6% 0.4% \[5\] Urinary tract infection 3.9% 6.5% 0.4% \[5\] Parameters governing efficacy of recommended staffing levels at reducing hospital transfer rates are listed here. Not all staff were effective in preventing all five types of hospital conditions; only statistically significant reductions in hospital transfer were used in the analysis. Base case values represent the best available estimate of the parameter in question. Values listed under the column \"biased toward recommended staffing\" and \"biased against recommended staffing\" represent extreme values of the parameter that are most favorable and least favorable to the recommended staffing scenario when tested in sensitivity analyses, and were derived from 95 percent confidence intervals around efficacy parameters. Abbreviation: Ref., reference. \All efficacy factors were varied simultaneously and also by group (RN, Licensed staff, and NA). ::: To convert these odds ratios into meaningful information for the Markov cohort simulation, we used additional data available in the CMS report to estimate total rates of hospital transfer for hypothetical facilities with median and recommended staffing. We assumed that the median-staffed facility had mean rates of hospital transfer for each of the five conditions as defined within the CMS report. Because the five conditions were not mutually exclusive, we downward-adjusted hospital transfer rates for each condition so that the rates for all five conditions totaled 16%, the overall 30-day hospital transfer rate for any diagnosis \[[@B5]\]. This assumes that 100% of hospital transfers had one of the five selected conditions as either the primary or secondary diagnosis, an assumption favorable to the recommended staffing scenario. In addition to assuming mean hospital transfer rates for the median-staffed facility, we also assumed an average risk of being in the highest decile of hospital transfer rates for any given condition (namely, 10%). Using this 10% probability together with the odds ratios of hospital transfer for each condition and nurse staffing type (NA, RN plus LPN, and RN), we were able to estimate the (reduced) likelihood of being in the highest 10% of hospital transfer rates for the facility with recommended staffing (and conversely, the increased likelihood of being in the lowest 90%). Using percentile data on hospital transfer rates available in the CMS report \[[@B5]\], we were able to estimate bottom 90% and top 10% hospital transfer rates for each condition. Facilities with recommended staffing might have, for example, a 7% likelihood of being in the top 10% of hospital transfer rates, and therefore a 93% likelihood of being in the bottom 90%. Using these new weightings, we were able to estimate the relative reduction in hospital transfer rates for any of the five conditions (see Table [2](#T2){ref-type="table"}) under the recommended staffing scenario. These values were then inserted into the Markov model. For the five staffing-sensitive conditions combined, we estimated an 8% relative risk reduction in 30-day hospital transfer rates (from 16% to 14.7%) under recommended staffing levels. We assumed that meeting recommended staffing for NA hours, licensed staff (RN plus LPN) hours, and RN hours were independent effects, because the CMS report does not indicate to what extent staffing levels are correlated \[[@B5]\]. For further details on efficacy calculations, see Appendix B (in Additional File [1](#S1){ref-type="supplementary-material"}). Case fatality rates for hospitalized patients were taken from Medicare data \[[@B9]\]. While the actual case fatality rates for frail older SNF patients are likely higher than Medicare average rates, we did not have published estimates specifically for this population. Thus, we tested a range of fatality rates in sensitivity analyses. The likelihood of dying in the SNF or after discharge was based on a smoothed hazard from an eleven year follow up of nursing home patients in a single facility \[[@B10]\]. While this result may not be generalizable, it represents the longest published follow-up of patients likely to have similar levels of comorbidity as the target population. We tested a wide range of values (from 15.2% to 60.6% annually) for the hazard in sensitivity analyses given the uncertainty about this estimate. Health-related quality of life ------------------------------ The model measured health effects in terms of quality-adjusted life expectancy. The utility weights for the health states were derived from time-tradeoff scores collected from seriously ill patients (median age, 62.8 years) on day three of their hospital stay, and at two months post-hospitalization, as part of the SUPPORT study. These utility weights are shown in Table [3](#T3){ref-type="table"}\[[@B11]\]. The mean time trade-off score for this seriously ill sample was 0.73 at day three of their hospital stay and 0.79 at month two post-hospitalization, where 0 represents death and 1 represents perfect health. This means that at day three of their hospital stay, on average patients accepted one year of life in their current state of health as being equivalent to 0.73 years in perfect health, whereas by month two after hospital discharge, patients considered one year of life in their current state of health as equivalent to 0.79 years in perfect health. While the improvement in health-related quality of life may appear small (0.73 to 0.79), it should be noted that these are average values and there was wide inter-individual variation in scores \[[@B11]\]. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Utilities, costs and discount rate ::: Parameter Base Case* Biased toward recommended staffing Biased against recommended staffing Ref. --------------------------------------------------------------- --------------- ---------------------------------------- ----------------------------------------- ------------ Utilities Hospital 0.73 0.41 \[0.79\] \[11\] Discharge 0.79 1.00 \[0.73\] \[11\] Days to reach discharge utility Under recommended staffing\* \[30\] \[15\] \[30\] Hourly wages with benefits Nurse assistant \$13.28 \[\$10\] \[\$20\] \[13, 14\] Licensed practical nurse \$20.78 \[\$30\] \[\$15\] \[13, 14\] Registered nurse \$32.94 \[\$20\] \[\$40\] \[13, 14\] Other nursing facility costs \$193.5 \[\$400\] \[\$100\] \[15\] Hospitalization costs Congestive heart failure\\ \$4603 \[\$9206\] \[\$2301\] \[9\] Electrolyte imbalance\\ \$3913 \[\$7826\] \[\$1956\] \[9\] Respiratory infection\\ \$4722 \[\$9444\] \[\$2361\] \[9\] Sepsis\\ \$7142 \[\$14285\] \[\$3571\] \[9\] Urinary tract infection\\ \$3853 \[\$7707\] \[\$1927\] \[9\] Daily cost for discharged patients \[\$0\] \[\$0\] \$113 \[17\] Annual discount rate 3% 0% 5% \[25\] All costs in 2002 U.S. dollars. Brackets refer to assumptions made for sensitivity analyses where data were not available or did not provide an adequate range for the parameter. Base case values represent the best available estimate of the parameter in question. Values listed under the column \"biased toward recommended staffing\" and \"biased against recommended staffing\" represent extreme values of the parameter that are most favorable and least favorable to the recommended staffing scenario when tested in sensitivity analyses. \Number of days in skilled nursing facility required to go from utility at admission to discharge level of 0.79 under recommended staffing. Average staffed facilities were assumed to require 30 days to reach discharge utility level. \\Hospitalization costs were varied together. Hospitalization costs include Medicare reimbursement to the hospital plus initial and subsequent physician visits. ::: For the median-staffed scenario, we modeled the 30-day \"post-acute\" period as having a linear improvement in utility (health-related quality of life) from 0.73 on admission to 0.79 on discharge from the SNF. Preliminary analyses demonstrated that the cost-effectiveness ratio was sensitive to the differential in health-related quality of life between patients in facilities with median versus recommended staffing. Thus, we performed a sensitivity analysis in which health-related quality of life within the recommended-staffing scenario increased twice as rapidly in the nursing home as within the median-staffing scenario, with health-related quality of life leveling off at 0.79 by 15 days in the recommended-staffing scenario instead of at 30 days as in the median-staffing scenario. For further details on utilities, please see Appendix C (in Additional File [1](#S1){ref-type="supplementary-material"}). Costs ----- Our analysis adopts the cost perspective of Medicare, the typical payer for post-acute care. We report costs (shown in Table [3](#T3){ref-type="table"}) in 2002 U.S. dollars. Where data were not available for the year 2002, prices were inflated to 2002 values using the medical care component of the Consumer Price Index \[[@B12]\]. We discounted all costs and health effects at 3% per year for the base case, with sensitivity analyses performed at discount rates of 0% and 5%. Costs were grouped into two categories, SNF costs and hospital costs. To obtain nursing costs in the SNF, hourly wages for RNs, LPNs, and NAs were estimated by ascertaining the wages of each staff type \[[@B13]\] and multiplying by a factor to account for fringe benefits, which represented 26.2% of total compensation for nursing home employees \[[@B14]\]. The adjusted hourly wages for each staff type were then multiplied by the hours worked by that staff type per patient day, and summed together, for each staffing scenario, creating a total cost of nursing care per patient-day. Non-nurse staffing related costs were estimated by taking the average daily reimbursement by Medicare to skilled nursing facilities and subtracting an estimated margin of 5%, and the nurse staffing costs calculated for the median-staffed facility \[[@B15]\]. Patients were assumed to have one visit by a physician on the initial day of their first admission to the SNF at a cost of \$100 (CPT 99303) \[[@B16]\]. Hospital costs were estimated from the median Medicare reimbursement under Medicare Part A for each diagnosis \[[@B9]\]. Physician costs assumed an initial visit cost on day one (CPT 99223) and a subsequent visit cost (CPT 99233) on all days in which the patient remained in the hospital \[[@B16]\]. A range of total hospital costs (physician fees plus hospital costs), from half to double baseline values, was tested in the sensitivity analyses. We made no attempt to model costs after discharge from the SNF as this was not the focus of our analysis, so total daily costs per day after discharge were zero. However, we performed a sensitivity analysis on total daily costs per day for discharged patients to assess whether this would significantly affect our main findings. The upper limit of our sensitivity analysis on this parameter was \$113/day, which reflects the inflation-adjusted average per-diem rate paid by Medicaid for long-term care \[[@B17]\]. Results ======= Base-case analysis ------------------ Under the base case assumptions, staffing SNFs at recommended levels resulted in discounted medical costs of \$8941 per post-acute patient from admission to ultimate discharge, versus a cost of \$8767 for facilities with median staffing (incremental cost \$173.50 per patient). Discounted, quality-adjusted life expectancy was 2.28117 quality-adjusted life years (QALYs) in the facility with recommended staffing and 2.28063 QALYs in the median-staffed facility, for a difference of 0.00054 QALYs. The incremental cost-effectiveness ratio was \$321,000 per discounted QALY gained. Without quality adjustment, the cost-effectiveness ratio was \$271,000 per discounted year of life saved, and \$250,000 per undiscounted year of life saved. Undiscounted life expectancy was 3.13587 years in the recommended-staffing scenario and 3.13517 years in the median-staffing scenario, for a gain of 0.0007 life years per patient. Undiscounted costs were \$8780 for the median-staffing scenario and \$8953 for the recommended-staffing scenario, with an incremental cost of \$173.60. Sensitivity analysis -------------------- ### Factors affecting transitions between states We report results of sensitivity analyses in Table [4](#T4){ref-type="table"}. In one-way sensitivity analysis, the cost-effectiveness ratio was most sensitive to the rate of hospitalization for the five nursing-care-sensitive conditions: congestive heart failure, electrolyte imbalance, respiratory infection, sepsis, and urinary tract infection. When the hospitalization rates from the SNF for these conditions were simultaneously doubled from their base case values, then the cost-effectiveness ratio was \$36,000 per QALY. If hospitalization rates were simultaneously half their base case values, the cost-effectiveness ratio was \$896,000 per QALY. The cost-effectiveness ratio was not sensitive to changes in mortality rates for any of the health states. ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### One-way sensitivity analyses on selected parameters, expressed as dollars per quality-adjusted life year gained ::: Parameter* Range (Best, Worst) Biased toward recommended staffing Biased against recommended staffing ----------------------------------------------------------- ------------------------- ---------------------------------------- ----------------------------------------- Efficacy of optimizing nurse staffing See table 2 All staff \$124,000 \$2,508,000 NA \$225,000 \$556,000 Licensed staff (RN + LPN) \$230,000 \$510,000 RN \$289,000 \$369,000 Transition probabilities Hospitalization rate (double, half) \$36,000 \$896,000 In-hospital mortality rate (double, half) \$150,000 \$793,000 Annual mortality rate in SNF or when discharged (15%, 61%) \$181,000 \$760,000 Utilities of \"hospital\" state (0.41, 0.79) \$239,000 \$343,000 of \"discharged\" state (1.00, 0.73) \$218,000 \$371,000 Time to discharge utility in recommended staffing group (15 d., 30 d.) \$94,000 \$321,000 Costs Hospitalization cost (double, half) \$192,000 \$386,000 NA wage (\$10, \$20) \$256,000 \$455,000 LPN wage (\$30, \$15) \$292,000 \$339,000 RN wage (\$20, \$40) \$193,000 \$391,000 Non-nursing costs in SNF (\$400, \$100) \$243,000 \$356,000 Daily costs for discharged patients (\$0, \$113) \$321,000 \$428,000 Discount rate (0%, 5%) \$297,000 \$337,000 All costs in 2002 U.S. dollars. The column labeled \"range\" refers to the best-case and worst-case values tested in sensitivity analyses. \"Double\" and \"Half\" refer to double and half the base case values, respectively. SNF, skilled nursing facility; NA, nurse assistant; LPN, licensed practical nurse; RN, registered nurse. ::: The cost-effectiveness ratio was somewhat sensitive to the efficacy of nursing staff as a whole. When all staff (nurse assistants and licensed staff) were assumed to maximally reduce the risk of hospitalization for the conditions that were sensitive to their input (which we accomplished by setting the risk reduction for hospitalization to the favorable end of the 95% confidence intervals) the cost-effectiveness ratio was \$124,000 per QALY. When staff were assumed to be minimally effective under recommended staffing conditions (by using the least favorable end of the 95% confidence intervals for staff efficacy at reducing hospital transfer rates), the cost-effectiveness ratio was \$2,508,000 per QALY. The cost-effectiveness ratio was not sensitive to the contributions of individual staffing type (NAs, LPNs plus RNs, or RNs), nor was it sensitive to varying the efficacy of recommended staffing at preventing hospitalizations for a particular condition (e.g. sepsis). ### Health related quality-of-life and costs The cost-effectiveness ratio was sensitive to the rate of improvement of quality of life while in the SNF in the recommended-staffing scenario. The median-staffed group was assumed to require the entire 30-day stay in the SNF to reach the discharge level of health-related quality of life. If the group with recommended staffing reached the discharge level of health-related quality of life by 15 days, half-way through the expected stay and twice as quickly as the median-staffed group, the cost-effectiveness ratio was \$94,000 per QALY. If it took the entire 30 days to reach the discharge value for health-related quality of life for the recommended staffing group, then the cost-effectiveness ratio was the base case, or \$321,000 per QALY. The cost-effectiveness ratio was not sensitive to the utility of being in the \"hospital\" or \"discharged\" states. The cost-effectiveness ratio was not sensitive to any of the tested costs nor any of the tested discount rates. Discussion ========== To our knowledge, this is the first attempt to compare the cost and effectiveness of two different nurse staffing scenarios using a cost-utility framework. Because there are no randomized data on the efficacy of increasing nurse staffing in nursing homes to the minimums identified in the CMS report \[[@B5]\] and recommended by the Institute of Medicine \[[@B6]\], modeling can serve as a means of testing various hypotheses about efficacy of staffing interventions on different outcomes. Our simulation modeling results indicate that preventing hospital transfers alone is unlikely to make the recommended minimum-staffing ratios cost-effective by conventional medical standards, unless increases in staffing to the recommended levels are targeted to facilities with high hospital transfer rates. This result occurred because staffing effects on hospitalizations, though statistically significant as reported in the CMS study, were small in magnitude (we estimated a reduction in 30-day hospital transfer rates from 16% to 14.7% based on the CMS data). The cost-effectiveness ratio was sensitive to the rapidity of improvement in health-related quality of life in the SNF in the recommended staffing scenario compared to the median-staffing scenario, but only when the differential improvement was quite marked. However, the approach we used is likely to underestimate the true benefit of increased staffing on quality of life, since it does not capture non-health-related benefits, such as physical comfort, attentiveness of staff, and social interaction. In the long-term care arena, there is evidence that staffing affects non-health-related quality of life. Residents in homes staffed at the recommended minimum levels were more likely to be out of bed during the day, were more socially engaged, and were assisted more frequently with incontinence, repositioning, and eating assistance \[[@B18]\]. In addition, when asked, nursing home residents prefer more frequent help with basic activities of daily living than they would typically get under routine median staffing conditions \[[@B19]\]. Furthermore, families of nursing home residents placed a high financial value on the incontinence and exercise care activities that are associated with higher staffed homes, valuing these staffing-intensive activities more than private rooms, which are successfully marketed and expensive \[[@B20]\]. To what extent these findings, which come from the long-term care population, translate to patients in post-acute settings is not clear, but it seems plausible that the benefits of staffing to short-stay patients extend beyond preventing hospitalization for acute sickness episodes and also include non-medical aspects of care. The cost-effectiveness ratio was sensitive to the rate of hospital transfer from the nursing home, ranging from \$36,000 per QALY at two-fold the baseline rate, versus \$896,000 per QALY at half the baseline rate. This result occurred due to the greater absolute risk reduction in hospital transfers in the simulation model when the overall rates of hospital transfer are higher, as well as the downstream effects of rehospitalization, which include a higher mortality rate while in-hospital and additional days in the SNF after discharge from the hospital. Thus, our results predict that facilities with the highest baseline hospital transfer rates stand to benefit the most from meeting recommended nurse staffing levels. Compared to other medical interventions, the base case cost-effectiveness ratio of \$321,000/QALY was relatively expensive. For example, equipping suitable nursing home residents with hip protectors increases quality-adjusted life expectancy while saving society money \[[@B21],[@B22]\]. A practice-initiated quality improvement intervention to improve treatment for depression demonstrated cost-effectiveness ratios between \$9000 and \$36,000/QALY (1998 U.S. dollars) \[[@B23]\]. However, performing annual Papanicolaou smears, a common medical practice, cost \>\$1,600,000/QALY (1995 U.S. dollars) when compared to performing them every two years \[[@B24]\]. Our analysis has several limitations, the most noteworthy being the limitations of the data on which the analysis is based. The recommended staffing levels required to prevent increased rates of hospital transfer were estimated through a retrospective analysis of data collected for administrative purposes, and the exact thresholds were determined through a post-hoc, iterative process designed to isolate the recommended staffing levels for each staff type. All the limitations of retrospective data analysis, most notably the potential inability to adjust adequately for case mix, thus affect our best estimate of cost-effectiveness. Also, the CMS report estimates median and recommended nurse staffing at the facility level. Short-stay patients may occupy a different percentage of beds in each facility, and the actual intensity of staffing for those beds might vary systematically from the facility-wide estimate in different ways for facilities with median and recommended levels of staffing. To compensate for these limitations, we varied the efficacy parameters over their 95 percent confidence intervals simultaneously, thus testing a broad range of possibilities, and the results were comparable. This study focused on the costs and clinical benefits of recommended staffing, and not the costs of a staffing mandate. Thus, it did not include the costs for policy enforcement. A federal mandate to nursing facilities to staff at certain recommended levels requires an apparatus for accurate collection of staffing data and a mechanism for enforcement, and even then will not ensure full compliance. Predicting the consequences of a mandate was beyond the scope of our analysis, which focused on the immediate and downstream benefits of decreased hospital transfer rates under recommended staffing levels compared to median staffing levels. We focused solely on patients recently discharged from the hospital, often known as \"post-acute\" patients, where the Medicare Part A program is typically the payer. These patients generally enter a SNF for a limited time to recuperate from their acute hospital stay and receive skilled therapies (e.g. physical therapy, occupational therapy, speech therapy, or intravenous antibiotics). This analysis did not attempt to model the effect of improving staffing for the \"long-stay\" residents, which constitute a different patient population \[[@B5]\], so our findings are only generalizable only to short-stay patients. Strengths of the simulation model presented in this study include the clinical relevance of the model, which captures the situation faced by the post-acute care population. The model was quite robust to most tested variables, and plausibly sensitive to a few key variables. Conclusion ========== We conclude that an intervention to increase nurse staffing to recommended levels for short-stay patients would not be cost-effective on the basis of reduction in hospital transfer rates alone. Further research should quantify the non-health-related quality of life benefits experienced by short-stay patients as a function of recommended and median nurse staffing levels. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= DAG constructed the Markov model and drafted the manuscript. SFS and JFS contributed to the study design and critically revised 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/1472-6963/5/35/prepub> Supplementary Material ====================== ::: {.caption} ###### Additional File 1 APPENDICES A-C TABLES 5-7 FOR MANUSCRIPT: Cost-effectiveness of recommended nurse staffing levels for short-stay skilled nursing facility patients. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ The authors thank Emmett Keeler for his thoughtful review of previous versions of this manuscript, and Katherine Ward for advice on appropriate modeling of costs in the SNF. DAG thanks the Robert Wood Johnson Clinical Scholars Program and the UCLA Specialty Training and Advanced Research program for their support.	PubMed Central	2024-06-05T03:55:57.515593	2005-5-10	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1145183/", "journal": "BMC Health Serv Res. 2005 May 10; 5:35", "authors": [ { "first": "David A", "last": "Ganz" }, { "first": "Sandra F", "last": "Simmons" }, { "first": "John F", "last": "Schnelle" } ] }
PMC1145184	Background ========== The enjoyment of the highest attainable standard of health is one of the fundamental rights of every human being without distinction of race, religion, political belief, and economic or social condition \[[@B1]\]. Everyone has the right to a standard of living adequate for the health of himself and of his family, including food, clothing, housing, necessary social services, and medical care \[[@B2]\]. Governments and the international community have an obligation to see the right to health progressively realized which includes the responsibility for prevention, treatment and control of disease; and the creation of conditions to ensure access to health facilities, goods and services \[[@B3]\]. Access to goods and services include of course the provision of essential medicines necessary for the prevention and treatment of prevalent diseases \[[@B4]\]. In addition, access to essential medicines is fundamental to human rights \[[@B5]\]. Pharmaceuticals have made an important contribution to global reductions in morbidity and mortality \[[@B6]\]. In many developing countries, medicines represent the largest household health expenditure. To help save lives and improve health by closing huge gap between the potential that essential drugs have to offer and the reality that for millions of people medicines are unavailable, unaffordable, unsafe or improperly used, World Health Organization (WHO) is already working with a wide range of partners to achieve this aim by drugs and medicines, and by working with countries to ensure equity to access to drugs, drug quality and safety, and rational use of drugs \[[@B7]\]. The National Drug Policy (NDP) process brings all interested parties; legislation/regulation, quality control, local production, education of consumers, prescribers, dispensers, drug evaluation, selection and registration and rational use; together to focus political improvement and also policy guidance, management tools, and training materials, derived from successful drug list initiatives, do exist. Essential drugs are one of the tools for fighting ill health. By increasing access to essential drugs, their safety and their rational use, we could make the pharmaceutical potential to improve health and save development gains \[[@B8]\]. Essential drugs are high-value commodities. Their availability draws patients to health facilities, where they can also benefit from preventive services. Moreover, if drug procurement is efficient and transparent, the confidence of governments and ministry in the country\'s health system is increased, and provision of financial and other resources for health system development encouraged \[[@B7],[@B9],[@B10]\]. Since using appropriate drug list has a lot of benefits and impact in health and economic status, it seems that evaluation of this list could be useful for evaluation of country-adapted policies. Iran has already adapted a National Drug List (NDL). This list is selected by Iran Drug Selecting Committee (IDSC). All drug supply management including registration, procurement, inspection, quality control and post marketing control could be done for a drug that it is accepted to be in Iran drug list. Thinking the drug list is one very important element of a NDP, thus analysis and evaluation of the IDSC\'s decision-making is one necessary part of the process of evaluating the NDP. Therefore, evaluation of IDSC\'s decision-making during last 5 years may be an appropriate way to evaluate drug supply system in Iran. Actually, these kinds of studies are necessary to assess the organizational and political determinants of the policy process; to help explain the strengths and weaknesses identified by the indicators; and to assist in identifying and assessing strategies to improve pharmaceutical policy implementation. The main study questions discussed in the study are the followings: 1\. Are the existing basic characteristics of the pharmaceutical system as well structures and processes within it ensure achievement of the main NDP goals? 2\. Are the criteria used for drug selection by the IDSC for the purposes of NDL creation compatible with the achievement of main NDP goals? Methods ======= In this study, a descriptive, explanatory, and prescriptive objectives methodology has been designed to describe the consequences, stakeholders, interests, and networks involved in the NDP; to help explain how and why a particular decision was reached in the past; and to assist decision-makers in managing the politics of formulation or implementation. To identify strengths and weaknesses of pharmaceutical policy formation and implementation in Iran, four standard questionnaires that were designed by WHO (Action Program on Essential Drugs) was used (see table [1](#T1){ref-type="table"} for more information). Questionnaires contained four categories of drug policy indicators including background information, structural indicators, process indicators, and outcome indicators. The indicators serve two purposes in the research: assessment of the implementation of NDP by measuring progress in key components (structural and process indicators) and evaluation of the outcomes of NDP (outcome indicators) \[[@B11]\]. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### WHO standardized NDP indicators ::: Background information ---------------------------------------------------------------------- Population data Economic data Health status data Health system data Human resources Drug sector organization Structural and process indicators (quantitative and qualitative) Legislation and regulation Essential drugs selection and drug registration Drug allocation in the health budget/public sector financing policy Public sector procurement procedures Public sector distribution and logistics Pricing policy Information and continuing education on drug use Outcome indicators Availability of essential drugs Accessibility of essential drugs Quality of drugs Rational use of drugs Reference: Brudon--Jakobowicz P, Rainhorn JD, Reich MR: Indicators for monitoring national drug policies, a practical manual 2^nd^ed. Geneva: World Health Organization; 1999. ::: Background information provides data on the demographic, economic, health and pharmaceutical contexts in which drug policy is being implemented. The information is quantitative data, at a single point in time, which was readily available at the central level. Background information was obtained from Iran statistics center \[[@B12]\], Iran central bank \[[@B13]\], population report of UNICEF \[[@B14]\], health status report of Iran \[[@B15]\], Iran drug statistics annual report \[[@B16]\] and the ministry of health. Structural indicators provide qualitative information to assess the pharmaceutical system\'s capacity to achieve its policy objectives. Questions on structural indicators were answered as yes or no based on available information. Process indicators provide quantitative information on the processes by which a NDP is implemented. They assess the degree to which activities are being effectively implemented and the progress over time. Outcome indicators were used to measure the results achieved and the changes that can be attributed to the implementation of a NDP. They measure the effects of implementation on the overall objectives of NDP including availability and affordability of essential drugs, drug quality, and the rational use of drugs. Outcome indicators were obtained from available information. In this study, information on structural and process indicators of drug allocation in the health budget and public sector financing policy, public sector procurement procedures, public sector distribution and logistics indicators were collected from the ministry of health, Iran drug statistics annual report \[[@B16]\], drug regulatory affairs, and management and planning organization of Iran. The information of structural and process indicators of drug pricing policy in Iran were collected from ministry of health and drug regulatory affairs information resources. Because of epidemiologic transition and lack of exact prevalence of new pattern of diseases \[[@B15]\] and their standard treatment guidelines in Iran, calculation of the value of the basket of drugs was impossible. Structural and process indicators of drug allocation in information, continuing education on drug use and rational use of drugs were obtained from Iran drug and poison information center, and rational drug use (RDU)/prescribing auditing committee \[[@B17],[@B18]\]. Selection of drugs for adding to list is performed in IDSC by considering of efficacy and safety, rationality in use, accessibility, and affordability altogether. To assess the agreement between decisions of drug selecting committee and standardized NDP indicators in the last 5 years (1998--2002), an additional weighted questionnaire was designed based on the questions that should be answered about any drug during decision in IDSC. Questionnaire mostly included questions about the history and documentations on efficacy and safety of drugs and also pharmacoeconomics assessment of drugs that are usually provided by applicants. In addition, information related to above fields that are usually obtained from reliable and independent sources by expert team of the secretary of IDSC were included. Additional questions were about identification of applicants and history of clinical trials especially in Iran, approval record of drug by international agencies like FDA, existence of drug standard treatment guidelines (STGs), potential to be produced by local factories, the supportive coverage umbrella like subsidization and insurance, and also price of the nominated drug. This information was designed in a 29-question form with yes/no pattern. The ideal point for questionnaire could be acquiesced with all positive answers except one. Final decision for a new molecule to be added to NDL is performed in IDSC with ten members who are selected and authorized by minister of health. Therefore, questionnaire was weighted by nominal group technique and by 10 members of IDSC who had mostly participated during last 5 years in IDSC meetings. The relation between each question and four indicators of NDP including: \"efficacy and safety\", \"affordability\", \"availability and accessibility\" and \"rationality in use\" have been asked and in case of each positive relation, the member of IDSC was asked to give a score from 1--5 for that indicator and for negative answer we considered the score of zero for it. Related to the point which obtained from the IDSC\'s opinion, any \"yes\" answer for each question could acquire scores in the range of +1 to +5 and the answer \"no\" got score 0 for each indicator. Fifty-nine drugs during 1998--2002 have been approved by IDSC\'s members. Their weighted questionnaire was filled out for each medicine separately and the results were reported by means of percentage of agreement. Distribution of decision makers\' point of view is available in related scattergram for interested readers (see [additional file 1](#S1){ref-type="supplementary-material"}). Results ======= Demographic, economic, health and pharmaceutical information ------------------------------------------------------------ Background information about demographic, economic, health and pharmaceutical contexts in 2002 have been summarized in Table [2](#T2){ref-type="table"}. As shown in this table, the population of Iran is more than 66480000 with urbanization rate of 65% and 69 years old life expectancy \[[@B12]\]. GNP per capita in Iran is 1750 \$US \[[@B14]\]. Total health budget is 12000 billions Rials. Total health expenditure is 7596 billions Rials and 5684 billions Rials is total drug expenditure (8000 Rials≈1 \$US). Data in this table also show that the infant mortality rate is 32.1/100000 and maternal mortality rate is 37/100000. Total number of prescribers and pharmacists are 92548 and 10334 respectively. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Background information about the demographic, economic, health and pharmaceutical contexts in 2002 ::: Demographic data ------------------------------------------------------------------- ------------------------------------------ Total population 66480000 Average annual growth of the population 1.5% Rate of urbanization 65% Life Expectancy 69 years GNP per capita 1750 \$US Average annual rate of inflation 15.8% Infant mortality rate 32.1/100000 Maternal mortality rate 37/100000 Pharmaceutical related data Total number of prescribers 92548 Total number of pharmacists 10334 Total health budget 12000 billions Rials (1.5 billions \$US) Total health expenditure 7596 billions Rials (0.9 billions \$US) Total drug expenditure 5684 billion Rials (0.7 billions \$US) Total number of drugs in national drug list 1516 Total number of drugs under generic name sold in the country 1288 Total value of local production (numbers) 96.1% (19.55 billions/20.348 billions) Total value of drug imports (numbers) 3.9% (797.41 millions/20.348 billions) Total number of drug manufacturing units in the country 57 Total number of wholesalers in the country 132 Total number of private pharmacies in the country 6100 Total number of private pharmacies in the three major urban areas 1534 8000 Rials≈ 1 \$US ::: Table [2](#T2){ref-type="table"} indicates that 96.1% (19.55 billions/20.348 billions) of drugs in use are produced in local pharmaceutical factories and the total number of drug manufacturing units in the country is 57. The total number of dosage forms in NDL is 1516 for 933 INN names that 1288 of them are marketed by generic name in the country. There are 6100 private pharmacies all over the country. Drug supply management is performed centrally in Drug Regulatory Affairs of MOH including drug selection, registration, procurement, distribution, pricing and subsidization, GMP control, and rational use of medicines. Structural and process indicators of legislation and regulation, drug selection and registration ------------------------------------------------------------------------------------------------ Structural and process indicators of legislation and regulation, essential drug selection, and drug registration in Iran in 2002 have been summarized in Table [3](#T3){ref-type="table"}. As shown in this table, NDP has been established in Iran and updated during last 10 years. Regulations have been issued on the basis of drug legislation and there is a drug regulatory authority to control registration, licensing system, the sale and distribution of drugs. Pharmacists are legally entitled to substitute generic drugs for brand name products. There is a quality control section, which carries out required inspections in different stages of pharmaceuticals production. Data show that seventy percents of total number of drug outlets inspected was in contravention but no sanction has been implemented for them. Iran enjoys primary health care system (PHC). About 5000 public health centers are responsible to provide health services to rural and underdeveloped areas including dispensing of about 400 drugs freely. Value of drugs from NDL procured in the public sector, out of total value of drugs procured in the same sector was 27%. Of total number of drugs, 95% was prescribed and sold from NDL. Hundred percent of samples transferred from manufacturers to the laboratory of MOH are tested. Nevertheless, it should be noted that samples are not collected from the market; they are collected directly from manufacturer companies. Table [3](#T3){ref-type="table"} showed that only 910 out of 1516 dosage forms of NDL (60%) are produced locally. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Structural and process indicators of legislation and regulation, essential drug selection, and drug registration in Iran in 2002 ::: ------------------------------------------------------------------------------------------------------------------------------------------------------- ----------------- Is there an official national drug policy document updated in the past 10 years? \+ Is there drug legislation updated in the past 10 years? \+ Have regulations based on the drug legislation been issued? \+ Is there a drug regulatory authority whose mandate includes registration and inspection? \+ Is there a licensing system to regulate the sale of drugs (wholesalers, pharmacists,\...)? \+ Are pharmacists legally entitled to substitute generic drugs for brand name products? \+ Are there legal provisions for penal sanctions? \+ Is there a check-list for carrying out inspections in different types of pharmaceutical establishments? \+ Are there any institutions where quality control is carried out? \+ Is the WHO Certification Scheme on the Quality of Pharmaceutical Products Moving in International Commerce used systematically? \+ Are there controls on drug promotion based on regulations and consistent with the WHO Ethical Criteria for Medicinal Drug Promotion? \+ Is there a national essential drugs list (EDL)/formulary using INN officially adopted and distributed countrywide? \+ Is there an official drug committee whose duties include updating the national drugs list? \+ Has the national essential drugs list /formulary been updated and distributed countrywide in the past five years? \+ Do drug donations comply with the national drugs list? \+ Are there formal procedures for registering drugs? \+ Is there a drug registration committee? \+ Is drug registration renewal required at least every five years? \+ Number of drug outlets inspected, out of total number of drug outlets in the country. 100% (132/132) Number of drug outlets in violation, out of total number of drug outlets inspected. 70% (92/132) Number of sanctions and administrative measures implemented, out of total number of violations identified. 0% Number of advertisements in violation of regulations on the ethical promotion of drugs, out of total number of advertisements monitored. 10% (5/49) Number of sanctions implemented for advertisements in violation of regulations, out of total number of violations identified. 100% (5/5) Value of drugs from the national drugs list procured in the public sector, out of total value of drugs procured in the same sector. 27% (415/1516) Number of drugs from the national drugs list prescribed, out of total number of drugs prescribed. 95% (1440/1516) Number of drugs from the national drugs list sold, out of total number of drugs sold. 95% (1440/1516) Number of locally manufactured drugs sold in the country from the national drugs list, out of the total number of drugs from the national drugs list. 60% (910/1516) Number of combination drugs newly registered, out of total number of newly registered drugs. 0% (0/243) Number of registered drugs, which are banned in other countries, out of total number of registered drugs. 0% (0/243) ------------------------------------------------------------------------------------------------------------------------------------------------------- ----------------- ::: Structural and process indicators of financing policy, procurement, distribution, logistics, and pricing policy --------------------------------------------------------------------------------------------------------------- Structural and process indicators of drug allocation in the health budget and public sector financing policy, public sector procurement procedures, public sector distribution and logistics, and pricing policy in Iran in 2002 are summarized in Table [4](#T4){ref-type="table"}. Existence of international nonproprietary name (INN) system for listing the medicines, and their procurements, distribution, and selling in private pharmacies have been indicated in this table. Data indicate that drug procurement unit receives required foreign currency in less than 60 days from request to release and the average lead-time from drug order to receipt at central level is less than eight months. The total margin used by wholesalers and retailers is less than 35% of the CIF (cost, insurance, freight) price. There is a system for monitoring drug prices regulated in the private sector. The average lead-time required for a sample order in the last year, out of average lead-time during the past three years decreased about 80%. The number of drugs failed from quality control testing out of the number of drugs tested was 13%. The number of drugs beyond the expiry date out of the total number of drugs surveyed was 0.1%. ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Structural and process indicators of drug allocation in the health budget and public sector financing policy, public sector procurement procedures, public sector distribution and logistics, and pricing policy in Iran in 2002 ::: Budget and pricing --------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ---------------------- Is the drug budget spent per year more than 20% of the MOH operating budget spent per year for the last three years? \+ Is the drug budget spent per capita more than US\$1.00 per year for the last three years? \+ Is the drug budget spent for national hospitals less than 40% of the total drug budget spent for the last three years? \- Has the drug budget spent per capita increased in the last three years? \+ Are there any financing systems in addition to the drug budget that contribute to the provision of drugs in the public sector? \+ Are drug prices regulated in the private sector? \+ Is there at least one major incentive for selling drugs at low cost in the private sector? \+ Is the total margin used by wholesalers and retailers less than 35% of the CIF price? \+ Is there a system for monitoring drug prices? \+ Are essential drugs sold under INN or generic name in private drug outlets? \+ Average time period of payment for a sample of orders, out of average time period of payment stated in contract. 100% (7 days/7 days) Procurement Are drugs usually procured in the public sector through competitive tender? \- Is there a system for monitoring supplier performance? \+ Are procurements done under INN? \+ Does the procurement unit receive foreign currency in less than 60 days (from request to release)? \+ Is procurement in the public sector limited to drugs on the national drugs list? \+ Is the average lead time (from order to receipt at central level) less than eight months? \+ Is procurement based on a reliable quantification of drug needs? \+ Average lead time for a sample of orders in the last year, out of average lead time during the past three years. 80% (7/9) Number of drugs/batches tested, out of number of drugs/batches procured. 0% Number of drugs/batches that failed quality control testing, out of number of drugs/batches tested. 13% (147/1132) Average time between order and delivery from central store to remote facilities in the last year, out of average time between order and delivery in the past three years. \- Storage Are good storage practices observed in the central procurement/distribution unit and/or major regional warehouses? \+ Is the information recorded on the stock cards for drugs? \+ Are the stocks for drugs within their expiry dates in the central procurement/ distribution unit and/or major regional warehouses? \+ Have all incoming products been physically inspected the central procurement/distribution unit and/or in the major regional warehouses? \+ Are drugs which are not on the national drugs list in stock in the central procurement/distribution unit and/or in the major regional warehouses? \- Are 80% or more of the vehicles of the central procurement/distribution unit and/or major regional warehouses in working condition? \+ Are essential drugs sold under INN or generic name in private drug outlets? \+ Number of drugs beyond the expiry date, out of the total number of drugs surveyed 0.1% (2/1400) ::: Structural and process indicators of RDU ---------------------------------------- Structural and process indicators of drug allocation in information and continuing education on RDU in Iran in 2002 is presented in Table [5](#T5){ref-type="table"}. This table shows that the national drug information book has been regularly published and revised within the past five years. There is an official continuing education system on RDU for prescribers and dispensers. There are drug information centres that provide regular information on drugs to prescribers and dispensers. Data of table [5](#T5){ref-type="table"} also show that there is a national therapeutic guide with standardized treatments but the concept of essential drugs is not part of the curricula in the basic training of health personnel and there are not therapeutic committees in the major hospitals. There is at least one injection in 47% of total prescriptions surveyed with average number of 3.6 drugs prescribed for each patient. Antidiarrheal drugs have been prescribed for children fewer than five in 19% of cases for treatment of diarrhea. Average retail price of standard treatment of pneumonia out of the average retail price of a basket of food is 21%. No budget has been devoted for enlightening public on RDU. Of prescribers surveyed, only 7.6% had direct access to a national drug formulary. Only 6.9% of prescribers received independent drug bulletins sent by drug and poison information centres. Number of training sessions on RDU for prescribers in the last year out of the average number of training sessions organized in the past three years was 13%. Number of prescribers who attended at least one training session in the last year, out of total number of prescribers surveyed was 50%. ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Structural and process indicators of drug information, continuing education on drug use, and rational use of drugs in Iran in 2002 ::: -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------------ Is there a national publication (formulary/bulletin/manual, etc.), revised within the past five years, providing objective information on drug use? \+ Is there a national therapeutic guide with standardized treatments? \+ Is the concept of essential drugs part of the curricula in the basic training of health personnel? \- Is there an official continuing education system on rational use of drugs for prescribers and dispensers? \+ Is there a drug information unit/centre? \+ Does the drug information unit/centre (or another independent body) provide regular information on drugs to prescribers and dispensers? \+ Are there therapeutic committees in the major hospitals? \- Are there public education campaigns on drug use? \+ Is drug education included in the primary/secondary school curricula? \+ Number of prescribers having direct access to a (national) drug formulary, out of total number of prescribers surveyed. 7.6% (7034/92548) Number of training sessions on drug use for prescribers in the last year, out of average number of training sessions organized in the past three years. 13% (120/924) Number of prescribers who have attended at least one training session in the last year, out of total number of prescribers surveyed. 50% (15203/32524) Number of issues of independent drug bulletins published in the last year, out of average number of issues of independent drug bulletins published per year in the past three years. 0% (0/13) Average number of copies of independent drug bulletins sent to prescribers, out of total number of prescribers. 6.9% (6386/92548) Amount spent on public education campaigns on drug use, out of total amount spent on public health education campaigns. 0% (0/300000 \$US) Average retail price of standard treatment of pneumonia, out of the average retail price of a basket of food. 21% (2.9/13.75 \$US) Average number of drugs per prescription. 3.6 Number of prescriptions with at least one injection, out of the total number of prescriptions surveyed. 47.4% (692530/1461033) Number of children under five with diarrhea receiving antidiarrheal drugs, out of the total number of children under five with diarrhea surveyed. 19% (266/1400) -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------------ ::: IDSC\'s function ---------------- Estimation of NDP indicators including efficacy and safety, affordability, availability and accessibility, and RDU on approved drugs by IDSC during 1998--2002 is shown in Figure [1](#F1){ref-type="fig"}. The assessment of 59 dossiers of approved drugs for adding to NDL showed that IDSC\'s members pay more attention to efficacy, safety, and rationality in use of drugs rather than accessibility and affordability. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Estimation of NDP indicators including efficacy and safety, affordability, availability and accessibility, and rational use of drugs on approved drugs by IDSC during 1998--2002.Number of evaluated drug dossier is 59. ES means efficacy and safety. RIU means rationality in use. AC means accessibility and AF means affordability. ::: ![](1472-698X-5-5-1) ::: Discussion ========== Pharmaceuticals have made an important contribution to global reductions in morbidity and mortality \[[@B6]\]. Drug supply management has a great impact in health services, thus, evaluation, redeveloping and implementation of drug policy has an important role in health system. Policy guidance, management tools, and training materials are derived from a successful drug list initiative \[[@B7]\]. IDSC is responsible to establish NDL. In this study, Iran drug policy indicators were monitored by four standardized questionnaires of WHO \[[@B11]\]. Demographic, economic, health and pharmaceutical information ------------------------------------------------------------ As shown in table [2](#T2){ref-type="table"}, the population of Iran is more than 66480000 with urbanization rate of 65% and 69 years old life expectancy \[[@B12]\]. GNP per capita in Iran is 1750 \$US which 5.7% of that, is usually spent for health expenditures \[[@B13]\]. Related to demographic, economic, health, and pharmaceutical contexts, Iran is one of the developing countries \[[@B14]\]. The percentage of products in NDL which are produced locally measures a country\'s self sufficiency for supplying the most essential pharmaceutical products \[[@B10]\]. Table [2](#T2){ref-type="table"} indicates that 96.1% of drugs are produced in 57 local pharmaceutical factories. Production of drugs domestically decreases the final price of drugs and thus makes them more affordable and presents more feasible and reliable procurements. Legislation and regulation -------------------------- As shown in table [3](#T3){ref-type="table"}, NDP has been established in Iran and updated in the last 10 years with inclusion of necessary regulations and drug control legislations. By regulation of national medical council, only drugs exist in NDL should be prescribed \[[@B19]\]. However, this regulation has not been implemented absolutely. About 70% of total number of drug outlets inspected had problem of quality that may negatively affect objectives of NDP \[[@B7]\]. There was no special rule to sanction them at that duration of time but it has now being established. Drug control and regulations are measures of a government\'s capacity to implement beneficial policies and practices in pharmaceutical management. If a government does not enforce legislation and regulation, it means a situation where plans for pharmaceutical improvement exist only on paper and not in reality \[[@B10]\]. Moreover, even if drugs are available, weak drug regulation could mean that they are substandard, thus spending on these medicines seems to be the major source for poverty \[[@B8]\]. Almost all data have been feedback to policy makers and fortunately, improvements are observable in some parts. In addition, some items have been taken in the priority of many relevant committees to find reasonable resolutions. Access issues ------------- Access to drugs is a key priority. From the consumer\'s point of view, access means that drugs can be obtained with reasonable traveling distance from health facilities \[[@B7]\]. Physical access --------------- Appropriate distribution network could provide good accessibility for medicines within the country. Existence of INN system for listing the medicines, and their procurements, distribution, and selling in private pharmacies are the strengths of Iran NDP (table [4](#T4){ref-type="table"}) \[[@B6],[@B10],[@B11],[@B20]\]. Private pharmacies belong to pharmacists who are allowed to establish their pharmacies under district regulation and legislation of MOH. They are responsible to provide medicines of NDL for all patients with fixed price which announced by MOH. They have to procure their medicines from distribution centers. These centers work under supervision of government. They procure medicines from 57 local pharmaceutical factories and 3 important drug procurement centers. Procurement of drugs from local industries is in priority and other pharmaceuticals import from other countries. It should be noted that INN system listing existing in Iran is not against the availability of patent drugs. If the patent drug is approved in NDL, it would be subsidized too. A specific guideline has been established in Iran for availability of drugs that are not included in NDL. According to this guideline, the physician who prescribe a non-NDL drug should sign an agreement and accept the responsibility for any possible unwanted effects that might happen by use of that drug in that specific patient. The other point is that the physician should convince the drug regulatory affairs about not effectiveness of the existed similar drugs in the NDL. Finally, the drug will be imported for that patient without any subsidy or insurance coverage only in amount of prescription \[[@B19]\]. Pricing ------- Information in table [4](#T4){ref-type="table"} also indicate the existence of a good structure and process for financing in drug supply leading to better affordability. Potential for local production of drugs could result in constant accessibility of drugs with suitable prices \[[@B9],[@B10]\]. It is considerable to note that in Iran, there is a powerful subsidizing system for drugs. Insurance companies pay about 80% of drug price, which diminishes patient\'s out of pocket spending. Evaluation of rational drug use ------------------------------- The evaluation of RDU in Iran is presented in table [5](#T5){ref-type="table"}. RDU means the promotion of therapeutically sound and cost-effective use of medicines by health professionals and consumers \[[@B7],[@B8]\]. Iran was very successful in establishing of more than 20 drug and poison information centers around country since first 1997 and providing national independent drug information bulletins \[[@B17]\] to promote RDU. It is interesting to note that most of \"standard treatment guidelines\" about communicable diseases in Iran have been written \[[@B21]\]. Concerning epidemiologic transition \[[@B15]\], it is important to provide guidelines for non-communicable illnesses specially diabetes and cardiovascular diseases as well as infectious diseases. Lack of drug and treatment committee\'s in hospitals is indicated in Table [5](#T5){ref-type="table"}. To ensure proper use of drugs in therapeutically sounds and cost-effectiveness way, integrated approaches between medicines and treatment management are required \[[@B22]\]. There are evidences for irrational use of drugs in Iran such as existence of at least one injection in 47% of total prescriptions surveyed and average number of 3.6 drugs prescribed for each patient or receiving antidiarrheal drugs for children under five years old (20%) in treatment of diarrhea \[[@B10],[@B21],[@B23]\]. On the other hand, no budget has been devoted for cultivating public on RDU \[[@B6],[@B10],[@B11]\] that can be a deficit for drug policy makers. Function of IDSC ---------------- Essential medicines are perhaps the most cost-effective element of public health after immunization \[[@B20],[@B24]\]. Regarding figure [1](#F1){ref-type="fig"}, the most interested factor for IDCS members were efficacy and safety of drugs (61%). This item is mostly experience-based and could be found in many classic pharmacological and clinical references. However, the existence of safe and effective medicines could save millions of lives and prevent untold suffering all over the world but is not enough in term of RDU \[[@B7]\]. In decision making of IDCS, information obtained about experience of drug usage or approval in abroad has been very helpful. In addition, IDSC pay enough attention to viewpoints of health professional societies. Results of clinical trials have also good impact in decision-making by IDSC members. Worldly, the criteria for new drug approval have been developed from experience-based to evidence-based approach. In this regard, WHO has done many efforts to encourage countries to provide evidence-based NDL to implement RDU. Evidence used by WHO to add or remove a drug might provide some basis in country-level decision-making, but in some cases, local trials should be carried out. It should not be forgotten that existence of reliable and independent information on questioned drug \[[@B25]\] especially on its costs \[[@B26]\] are important elements for decision-making in IDSC. Accordingly, holding adequate training workshops for IDSC members to teach evidence-based medicine seems necessary \[[@B22]\]. RDU was the second frequently considered factor by decision-makers of IDSC. Existence of approved treatment guidelines for a drug before submitting the application to IDSC is very important. Based on the criteria examined in the present study, only 59% of drugs were selected according to therapeutic guidelines in the last 5 years. To promote RDU, it is necessary to provide updated STGs for new patterns of different diseases like cardiovascular or cancer. WHO emphasizes to provide essential life-saving drugs developed for leading non-communicable diseases as well as leading infectious ones \[[@B5],[@B7],[@B20]\]. Every one of these diseases impinges detracting from health gains and delaying progress in other areas such as education and economic development \[[@B7],[@B27]\]. Since 1999, WHO has recommended countries to design their NDL according to national STGs \[[@B25]\]. Accessibility and affordability with means of 51% and 49% respectively also support that IDSC\'s members have not paid enough attention in this respect. The price of pharmaceuticals seems a substantial barrier to access for governments and health insurers \[[@B24]\]. An important point is that cost-effectiveness has been almost a forgotten element in decision-making during the last 5 years (1998--2002). Considering financial patterns in providing pharmaceuticals is very important element, and if not, it will face countries with the problem of drugs procurement \[[@B28]\]. Recently two special workshops on pharmacoeconomics were held for IDSC members and stakeholders. This can be a starting point to show that importance. After this study and finding these interesting data, new questionnaires for drug selecting has been launched and implemented to improve the process of drug selection by considering standardized indicators of NDP. Finally it should be reminded that essential medicines are those that satisfy the priority health care needs of the population. They are selected with due regard to public health relevance, evidence on efficacy and safety, and comparative cost-effectiveness. Essential medicines are intended to be available within the context of functioning health systems at all times in adequate amounts, in the appropriate dosage forms, with assured quality and adequate information, and at a price that the individual and the community can afford. The implementation of the concept of essential medicines is intended to be flexible and adaptable to many different situations; exactly which medicines are regarded as essential remains a national responsibility \[[@B29]\]. Revision of drug system in Iran for implementation and improvement of the processes to achieve NDP\'s objectives is necessary to save public health. Clarification of NDP\'s objectives and their impact for IDSC\'s members will be helpful to improve the equity in access to pharmaceuticals. Conclusion ========== Finally, it is possible for us to answer to two main questions of this study as mentioned in introduction. Overall results of tables [2](#T2){ref-type="table"},[3](#T3){ref-type="table"},[4](#T4){ref-type="table"},[5](#T5){ref-type="table"} show that in Iran like most developing countries, the system, structures, and mechanisms are in place, however, they often do not function properly, which impeded implementation of strategies and policies and achievement of objectives \[[@B6]\]. In addition, data of figure [1](#F1){ref-type="fig"} shows that the criteria used for drug selection by IDSC are 50--60% compatible with achievement of main NDP goals. Collectively, it is concluded that drug system in Iran is in place but needs some revisions. Further studies are required to evaluate the exact impact of drug supply management in health in Iran. Competing interests =================== This study was the outcome of the MPH thesis of SN and was supported by a grant from Tehran University of Medical Sciences. Authors\' contributions ======================= SN carried out all parts of the study (designing, conducting and writing the manuscript). AK and RM participated in design and coordination of the study. MA participated in designing, and writing 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/1472-698X/5/5/prepub> Supplementary Material ====================== ::: {.caption} ###### Additional File 1 The scattergram of decision makers\' point of view. The relation between each question and four indicators of NDP including: \"Efficacy and Safety\", \"Affordability\", \"Availability and Accessibility\" and \"Rationality in use\" have been asked and in case of each positive relation, the member of IDSC was asked to give a score from 1--5 for that indicator and for negative answer we considered the score of zero for it. Related to the point which obtained from the IDSC\'s opinion, any \"yes\" answer for each question could acquire scores in the range of +1 to +5 and the answer \"no\" got score 0 for each indicator. Their weighted questionnaire was filled out for each member separately and the results were reported by means of percentage of agreement. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ Authors wish to thank very much the secretary office of Drug Selecting Committee and kind assistances of Dr. Akbar Abdollahi-Asl, Dr. Mohsen Khatibi, and Mrs. Fariba Aghasikhani.	PubMed Central	2024-06-05T03:55:57.519727	2005-5-10	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1145184/", "journal": "BMC Int Health Hum Rights. 2005 May 10; 5:5", "authors": [ { "first": "Shekoufeh", "last": "Nikfar" }, { "first": "Abbas", "last": "Kebriaeezadeh" }, { "first": "Reza", "last": "Majdzadeh" }, { "first": "Mohammad", "last": "Abdollahi" } ] }
PMC1145185	Background ========== Statistical tests performed to evaluate differences in the outcomes collected in randomized clinical trials (RCTs) are based on the assumption that all the data come from random variables that are independent of each other. However, in some trials, the patients are randomized more than once, usually twice. Investigators often pool these events and report data for all randomized patient episodes instead of episode per patient. Therefore, the independence assumption can be violated. Concerns about the re-randomization of patients on febrile neutropenia trials have been raised by some authors \[[@B1]\] due to a possible violation of the principle of independency. However, this practice is accepted by the Immunocompromised host society as a valid one \[[@B2]\] and has been used in some of the trials performed in neutropenic patients. We have recently encountered this problem during our conduct of meta-analysis (MA) on the role of colony-stimulating factors in the treatment of febrile neutropenia \[[@B3]\] where two out of the 12 trials reported analysis according to febrile neutropenia episodes instead of episodes per patients. In this paper, we address the issue whether the combined (pooled) risk difference (RD) after re-randomization is considerably different from the RD if the patients were not re-randomized. Methods ======= Assumptions ----------- Assuming that once an event (FN episode) occurred in a patient the probability of another occurrence is likely to be different from its initial value, we introduce a parameter dmeasuring this change: ![](1471-2288-5-17-i1.gif) Previously, Chouaid et al. \[[@B4]\] provided the data on the probability of FN after each randomization. Data were available from 39 patients, of which seven had an event; six of these patients were re-randomized and four of them had another occurrence of event. Thus, accordingly an estimate for dcan be obtained by ![](1471-2288-5-17-i2.gif). To avoid messy notation, we will omit the \'hat\' from the estimates of these parameters. This indicates that the patient who develops an FN episode after first cycle will have about 3.71 times higher likelihood of development of FN in the second chemotherapy cycle. Note that our calculations also work for values d\< 1. On the other hand, the patients who did not experience an episode of FN are usually less likely to experience it after re-randomization. The probability of an FN episode in a patient who didn\'t experience an episode of FN in the first round is ctimes smaller (or larger, if c \>1) in the second round of randomization. In this example, ![](1471-2288-5-17-i3.gif). To illustrate this study, we present a chronological tree diagram in figure [1](#F1){ref-type="fig"}. Although this is the only study we were able to find which provided data on re-randomization in both arms, it likely represents a typical scenario for the assumptions used in our model. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Chouaid et al. study ::: ![](1471-2288-5-17-1) ::: Results ======= Calculations ------------ Assume that Npatients in a clinical control trial are being randomized (for the first time) in such way that N~C~patients are randomized into the control group and N~A~patients are selected into the experimental (alternate) group. In order to simplify our calculations, we have set N~A~≈ N~C~≈ N/ 2. The percentage of people who were re-randomized (randomized for the second time) is x. After re-randomization, the total number of events is reported as E~CT~for the control group and E~AT~for the experimental group. The total number of events reported is given by E~T~= E~AT~+ E~CT~. Consider the following two variables: E~A~-- the number of events in experimental group in the initial randomization; E~C~-- the number of events in control group in the initial randomization; These two numbers are the most important in our calculations. We can then estimate the initial risk difference ![](1471-2288-5-17-i4.gif), compare it to the reported numbers and observe the relative error. In the following paragraphs, we will work toward estimating these two variables. E= E~A~+ E~C~-- the total number of events in the initial randomization; NE= N- E-- the number of patients without an event in the initial randomization; ![](1471-2288-5-17-i5.gif) -- the risk at which events occurred in the experimental group; ![](1471-2288-5-17-i6.gif) -- the risk at which events occurred in the control group; ![](1471-2288-5-17-i7.gif) -- the risk difference for the initial randomization; Assuming that re-randomization is approximately evenly divided (if not, we can change the fraction 1/2 with appropriate value), the re-randomized control group and re-randomized experimental group will have the following sizes: ![](1471-2288-5-17-i8.gif) -- the size of experimental re-randomized group; ![](1471-2288-5-17-i9.gif) -- the size of control re-randomized group; In Figure [2](#F2){ref-type="fig"}, we summarize the relationship between these variables in a general case. In order to do both, the analysis of difference in outcome between the experimental group and control group, as well as the analysis of repeated randomizations, all of the entries identified in the figure must be determined. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Tree diagram showing the relationships between the variables in the re-randomization process ::: ![](1471-2288-5-17-2) ::: Having in mind that the reoccurrence of an event is dtimes more likely (or less likely if d\< 1) if the patient comes from the E-- group (patients with events), and ctimes less likely (or more likely if c\> 1) if it comes from NE-group (patients without events), and that the events occur at the initial risks p~A~and p~C~for the experimental and control groups, respectively, we have: ![](1471-2288-5-17-i10.gif) -- the number of re-randomization events in the experimental group; ![](1471-2288-5-17-i11.gif) -- the number of re-randomization events in the control group; Therefore, the risk difference of the re-randomization alone, can be calculated as ![](1471-2288-5-17-i12.gif) The overall risk difference for both randomizations, the initial one and the re-randomization, can then be calculated as ![](1471-2288-5-17-i13.gif) or replacing the formula above ![](1471-2288-5-17-i14.gif) The relative error between the overall risk difference and the risk difference resulting from considering the patients only from the initial randomization is then estimated by the expression ![](1471-2288-5-17-i15.gif) However, the fraction ![](1471-2288-5-17-i16.gif) is usually not reported, and therefore has to be estimated. We will estimate this via the average risk ![](1471-2288-5-17-i17.gif) using the following equations ![](1471-2288-5-17-i18.gif) and ![](1471-2288-5-17-i19.gif) Adding these two equations, we get the quadratic equation in terms of p![](1471-2288-5-17-i20.gif), which can be solved as: ![](1471-2288-5-17-i21.gif), where E~T~= E~AT~+ E~CT~. Replacing this resulting estimate into the relative difference equation above, leads to ![](1471-2288-5-17-i22.gif) which after some simplifying becomes our main formula: ![](1471-2288-5-17-i23.gif) Therefore, the relative error of the reported risk difference and the actual risk difference before re-randomization can be estimated by the expression depending on x-- the percentage of the patients who were re-randomized, dand c-- multipliers for the probability of reoccurrence of the event, and the ratio ![](1471-2288-5-17-i24.gif) of the total events reported and the initial number of patients entering the trial. An Excel file performing all of these calculations and allowing the readers to enter their own data is enclosed as an Additional File [1](#S1){ref-type="supplementary-material"}. Figure [3](#F3){ref-type="fig"} describes our method graphically. The vertical axis is x-- the percent of the patients who were re-randomized. The horizontal axis is d-- the multiplier by which the probability of an event changed after an initial occurrence of an event. Figure helps to determine under which circumstances the error in risk difference is less than 5%. The curves represent 5% relative error level curves for different values of the ratio ![](1471-2288-5-17-i24.gif) of the total events reported and the initial number of patients entering the trial. The values of the ratio ![](1471-2288-5-17-i24.gif) are indicated at the top of the graph. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### The 0.05 level curves for the relative difference between the risk difference before and after the re-randomization. We assume that c= 0.40 in this figure. The horizontal axis represents d-- the multiplier for the probability of reoccurrence of the event after the re-randomization; The vertical axis represents x-- the percentage of the patients who were re-randomized; The curves represent 0.05 relative error level curves for different values of the ratio ![](1471-2288-5-17-i24.gif) of the total events reported and the initial number of patients entering the trial. The values of the ratio ![](1471-2288-5-17-i24.gif) are indicated at the top of the graph. For a chosen value of the ratio ![](1471-2288-5-17-i24.gif): if a point (d, x) is [between the two level curves with identical value (and color)]{.underline} -- the relative error of the risk differences is less than 5%; otherwise -- the relative error of the risk differences is more than 5%. ::: ![](1471-2288-5-17-3) ::: For a chosen value of the ratio ![](1471-2288-5-17-i24.gif): if a point (d, x) is between the two level curves with identical value -- the relative error of the risk differences is less than 5%; otherwise -- the relative error of the risk differences is more than 5%. Risk ratio ---------- The overall risk ratio after the re-randomization can be defined as ![](1471-2288-5-17-i25.gif), since we assumed that N~A~≈ (N/ 2) ≈ N~C~and ![](1471-2288-5-17-i26.gif). Then, using the definitions of p~A~and p~C~, we obtain the following equation: ![](1471-2288-5-17-i27.gif) which is exactly the risk ratio of the initial randomization. Therefore, the overall risk ratio after the re-randomization is equal to the initial risk ratio after the first randomization. Interestingly enough, the risk ratio is not affected by the re-randomization. Odds ratio ---------- The overall odds ratio after the re-randomization can be defined as ![](1471-2288-5-17-i28.gif), which can be simplified as ![](1471-2288-5-17-i29.gif) On the other hand, the odds ratio from the first randomization alone can be expressed as ![](1471-2288-5-17-i30.gif). Using the formulas ![](1471-2288-5-17-i31.gif) and ![](1471-2288-5-17-i32.gif), we get the expression ![](1471-2288-5-17-i33.gif). The relative difference between the odds ratio before the re-randomization and the overall odds ratio can therefore be expressed as ![](1471-2288-5-17-i34.gif) Data from CSF trials -------------------- To illustrate our method, we will use examples from two papers that studied use of CSFs in the treatment of chemotherapy-related febrile neutropenia (FN) and reported the number of patients who were hospitalized and remained in the hospital longer than 10 days \[[@B5],[@B6]\]. In both of these trials some of the patients were re-randomized after first FN episode (hospitalization). Mitchell et al\[[@B5]\] studied 112 patients in the first randomization for the effect of G-CSF in FN. Seventy-four of these patients were re-randomized (x= 66.07%). From a graph in their paper we deduced that 18% and 20% of the patients on GCSF and placebo, respectively, remained in the hospital longer than 10 days. This translates into 17 and 18 events in these two treatment arms ![](1471-2288-5-17-i35.gif). From the graph in the Figure [3](#F3){ref-type="fig"}, and assuming \[[@B4]\] that d= 3.71 and c= 0.40, we estimate the relative difference between the reported risk difference evaluated for all the patients, and the risk difference resulting from considering the patients only in the first initial randomization to be only 0.70%. The point corresponding to this example is marked by the purple circle in the Figure [3](#F3){ref-type="fig"}. In our case, the point is safely between the two identical ![](1471-2288-5-17-i37.gif) level curves (the first and the third curve from the right) indicating that the relative error is smaller than 5%. Anaissie et al. \[[@B6]\] studied effect of GM-CSF on FN in 92 patients, of which 15.9% were re-randomized (x= 0.159). Overall, 18 and 26 patients in GM-CSF and placebo arm remained in the hospital after 10 days ![](1471-2288-5-17-i38.gif). The difference is now 8.23%. The point corresponding to this example is indicated by the blue square in the Figure [3](#F3){ref-type="fig"}. Note that this time the point is not between two ![](1471-2288-5-17-i40.gif)-level curves, indicating that the relative error is larger than 5%. The odds ratio for this data also changes from OR= 0.554 before the re-randomization to the overall value after re-randomization OR= 0.535, a change of 3.49%. Estimating the risks in the initial randomization ------------------------------------------------- Meta-analysts would also like to know at what risks the events occur in the experimental and control groups before any re-randomization is performed. For example, while in the first example the relative difference between the reported risk difference and the risk difference before the re-randomization was not considerable; in the second example we found a considerable difference. Some meta-analysts might prefer to use the risks of the events from the initial randomization, i.e., before any re-randomization was performed. In this section we will try to estimate these risks. By estimating first the average risk ![](1471-2288-5-17-i17.gif) and solving the equations above for p, we have ![](1471-2288-5-17-i21.gif), where E~T~= E~AT~+ E~CT~. Using this parameter, the formulas for p~A~and p~C~are easily obtained as: ![](1471-2288-5-17-i41.gif) and ![](1471-2288-5-17-i42.gif) Using these estimates, we can get the estimates of E~A~and E~C~-- the number of events in the initial randomization in experimental and control arms of the trial, respectively. If the number of patients in each arm of the trial in the initial randomization is given, the events are estimated as E~A~= p~A~·N~A~and E~C~= p~C~·N~C~. If the only number we have is the total number of patients entering the initial randomization, N, the events must be estimated more crudely as ![](1471-2288-5-17-i43.gif) and ![](1471-2288-5-17-i44.gif). In the case of Anaissie et al. example, we obtain the following estimates for the number of events: E~A~= 14 and E~C~= 21, and since we are not given the number of patients in each arm, we have to estimate each to 46. Therefore, the numbers we could use for a meta-analysis from this particular trial would be Experimental 14 / 46 Control 21 / 46 Discussion ========== Our method has been developed using the example of febrile neutropenia (FN) in which the patients who once developed FN are considered to be at increased risk for subsequent event \[[@B7]\]. We assume that it is not possible to reliably predict which patients will develop the event of interest from the outset of the trial, but that once the event occurred the risk for subsequent event will be higher in those patients who had developed the first event. Because of this we adopted the values of d\> 1 and c\< 1. However, it is conceivable that biology of the process may differ in different circumstances. A meta-analyst is well advised to work with the content-specific experts to address this issue. The same formulas can be used to reflect any combination of the values for dand cparameters, except when c= d. A reader should note that in our model the parameters cand dwork in \"conditional probability\" framework, i.e., it represents a multiplicative rather than additive approach to asses the effect of previous occurrence of the event on the next one. This makes sense from purely clinical point of view. Error analysis -------------- A natural question to ask at this point is \"What is the extent of error in the estimation of the parameters cand din these formulas, and in particular, in the formula (1.1).\" We first attempted to determine confidence intervals for the parameters cand d. Unfortunately, since the proportions on top and bottom of the formulas for evaluation of cand dare obviously dependent random variables, none of the classical examples for evaluation of a confidence interval of a ratio of two proportions can be applied in this case. For example, Sutton et al \[[@B8]\] recommend that a confidence interval for a ratio of two risks ![](1471-2288-5-17-i45.gif) can be calculated using the formula ![](1471-2288-5-17-i46.gif), where ![](1471-2288-5-17-i47.gif). However, this formula, as well as most other formulas, are derived from Taylor\'s expansion of a transformation of random variables, and depend on the assumption that the ratios a/ band c/ dare independent (have zero covariance). In our case this assumption can not be applied. To determine this effect, we used the example of Anaissie et al. \[[@B6]\] and created Table [1](#T1){ref-type="table"}. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Error analysis using Anaissie et al. data ::: ≈ c ---------- ------- ------- ------- ------- ------- ------- ------- ------- ------- ------- ------ -- 1.0 -.286 -.272 -.259 -.245 -.232 -.197 -.162 -.128 -.096 -.064 .033 1.7 -.198 -.186 -.175 -.163 -.152 -.122 -.093 -.064 -.036 -.009 .018 2.4 -.131 -.121 -.111 -.101 -.091 -.065 -.039 -.014 .011 .035 .059 3.0 -.077 -.068 -.059 -.050 -.041 -.018 .005 .028 .050 .072 .093 3.7 -.032 -.024 -.016 -.008 .021 .042 .063 .083 .103 .123 5.6 .062 .069 .075 .082 .089 .106 .123 .140 .156 .173 .189 7.5 .131 .136 .142 .147 .153 .168 .182 .196 .211 .225 .238 9.4 .183 .188 .193 .198 .203 .229 .241 .254 .266 .278 11.2 .226 .231 .235 .239 .244 .255 .267 .278 .289 .300 .311 13.1 .262 .266 .270 .274 .278 .288 .299 .309 .319 .329 .339 15.0 .293 .296 .300 .303 .307 .317 .326 .335 .345 .354 .363 ::: The top row of Table [1](#T1){ref-type="table"} represents the estimate of the parameter cas it varies from 0 to 2, while the leftmost column indicates value of the estimate of the parameter das it varies from 1 to 15. The numbers in the table show the difference between the real relative difference from formula (1.1) and the relative difference calculated with these alternate values for cand d. The value .000 (boxed) in the table corresponds to the difference from the RD calculated with c= 0.4 and d= 3.7 in our example (Anaissie et al. \[[@B6]\]). However, if we assume that c= 0.7, and that d= 9.4, the value of RD (circled) will be larger by 21.6%, i.e., instead of 8.23%, the value of relative Risk Difference given by the formula (1.1) will be 29.83%. In this case we estimated that the original number of patients who underwent first randomization was 12/46 in experimental group, and 17/46 in the control group, respectively (see above). Note that the while some variation exists, the relative risk difference is not extraordinarily sensitive to the variation in the parameters cand d. Conclusion ========== In general, it is considered that meta-analyses represent the best method to provide reliable assessment of the effectiveness of health care interventions. However, reliability of meta-analyses depends on the validity of underlying methods used to combine data from the individual studies. One of the key assumptions for the valid MA is the independence of the data being analyzed. As illustrated here, in practice this assumption is often violated. The question for meta-analysts then becomes under which circumstances data can still be properly analyzed even if the independence axiom was violated. In the case of FN, for example, Immunocompromised Host Society issued a guideline in which they considered it acceptable to have multiple randomization of the same patients and reintroduce fully recovered patients into clinical trials \[[@B2],[@B9]\]. However, we believe that the answer regarding the acceptability of re-randomized data is not a simple categorical \"yes or no\" answer, but it will depend on the amount of error introduced into calculation. If the relative error of risk differences in the results with and without re-randomization is considered small, then such calculations can be safely accepted by meta-analysts. On the other hand, if such an error is considered large, then the analysts should not use the re-randomized data. Instead, the estimates of initial randomization from the method illustrated in the previous section should be used. It is very intriguing that if the summary effect measure selected is the risk ratio, under the assumptions of our model there is no difference in the final outcome before and after the re-randomization. Therefore, if a meta-analyst is concerned about the effect of re-randomization on the meta-analysis, one way to sidestep the issue and still obtain reliable results is to use risk ratio as the measure of interest. In the examples illustrated in this paper, we showed overall risk difference, RD (after second randomization) and RD after first randomization in Mitchell et al. \[[@B5]\] trial differed by relative value of 0.7%. In the case of the trial by Anaissie et al. \[[@B6]\] this difference was 8.23%. It is conceivable that this large relative error (of 8.23%) may affect the results of meta-analysis. Figures [4](#F4){ref-type="fig"} and [5](#F5){ref-type="fig"} show results of meta-analysis with and without re-randomized data. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Meta-analysis conducted with Review Manager using the reported data from Anaissie et al. with re-randomized data included. ::: ![](1471-2288-5-17-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Meta-analysis conducted with Review Manager using our estimates for the data from Anaissie et al. without including the re-randomized data. ::: ![](1471-2288-5-17-5) ::: In this particular example -- change in the input for the meta-analysis did not change the outcome and only marginally changed the pooled risk difference from -9% to -8%. In general, the effect of a given error on the results of meta-analysis can be approximated by multiplying the weight of a given study by the estimate of its relative error. In our example, weight was \~8% and relative error was also \~8% resulting in the change of the summary point estimate in meta-analysis less than 1%. Even when the relative error is high as 29.83% (see above), the summary point estimate would change only by 2.3%. However, should this study have contributed to the meta-analysis with the weight of 40%, the overall change in the summary point estimate would be about 12% (0.4 × 0.3 = 0.12). In general, however, it appears that if the study in question is of small size relative to the rest of the studies included in the meta-analysis, the error introduced by re-randomization will only minimally affect meta-analytic summary point estimate. Although we developed our method to help with meta-analyses, it should be stressed that our method is also relevant to the analysis of the original trials, and may help assess impact of re-randomization on the results of the trial. We strongly suggest that meta-analysts always compare overall results of their meta-analysis to results based on first randomizations only, as a general control of the reliability of the meta-analysis. In order to make that possible researchers conducting randomized clinical trials would have to report a much more detailed information on their methodology, specifying the numbers of patients being re-randomized, experiencing only one occurrence of an event, experiencing two occurrences of an event, or dropped out of the study. We believe that the method provided in this paper represents a valuable contribution to the improvement of interpretation of results of clinical trials and in particular their use in meta-analysis. Although the development of this method was stimulated by a concrete problem in the field of febrile neutropenia, practice of re-randomization is used in other areas of medicine such as testing of treatments for epilepsy, recurrent headaches and anti-emetics in cancer chemotherapy. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= IH developed most of the formulas for estimation of the risk differences. GHL found the relevant examples and references helping frame the problem in realistic environment. BD and OC conceived of the problem, participated in its framing and coordination, and helped IH to draft 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/1471-2288/5/17/prepub> Supplementary Material ====================== ::: {.caption} ###### Additional File 1 An Excel file performing all of these calculations and allowing the readers to enter their own data ::: ::: {.caption} ###### Click here for file :::	PubMed Central	2024-06-05T03:55:57.523496	2005-5-10	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1145185/", "journal": "BMC Med Res Methodol. 2005 May 10; 5:17", "authors": [ { "first": "Iztok", "last": "Hozo" }, { "first": "Benjamin", "last": "Djulbegovic" }, { "first": "Otavio", "last": "Clark" }, { "first": "Gary H", "last": "Lyman" } ] }
PMC1145186	Background ========== Benign prostate hyperplasia affects most men and often causes lower urinary tract symptoms. Prostate cancer leads in the number of estimated new cases diagnosed each year and is the second leading cause of cancer specific mortality in American men \[[@B1]\]. The development of the prostate, benign enlargement of the prostate and prostatic carcinogenesis and progression require androgenic stimulation \[[@B2]\]. Since androgen acts via androgen receptor (AR), AR expression in prostatic cells may play a central role in prostate pathology. AR belongs to the steroid receptor superfamily, which includes estrogen, progesterone, corticosteroids, vitamin D, thyroid, and retinoic acid receptors \[[@B3]\]. A study using radiolabeled natural and synthetic AR ligands showed the presence of AR in normal prostate, benign prostate and prostate cancer \[[@B4]\]. High levels of AR are associated with increased proliferation, markers of aggressive disease and are predictive of decreased biochemical recurrence-free survival independently, confirming the role of AR in tumor growth and progression in hormonally naive prostate cancer \[[@B5]\]. AR protein expression was assessed first using visual scoring of immunostained sections of prostate tissue. A visual scoring technique developed by Miyamoto et al. \[[@B6]\] assessed intensity of AR immunostaining on a scale of 0 (none) to intense (3+) for each nucleus. However, recognition of malignant nuclei requires a skilled pathologist or a highly trained technician and visual assignment of immunopositivity is subjective, tedious and poorly reproducible. Mean optical density (MOD) measured using image analysis was found more accurate and reproducible for measuring AR expression but object identification remained difficult. Investigators used a light pen to encircle \[[@B7]\] or a sampling window \[[@B8]\] to select areas of malignant nuclei for optical density measurement. However, these interactive image analysis methods are tedious and user bias influences the measurements. Tilley et al. \[[@B9]\] used an automated color video image analysis system to measure MOD of each positively stained area in visually marked malignant tissue. MOD was calculated as the total integrated optical density divided by the area. Such measurements tend to be less accurate when compared to MOD measured from individual nuclei. In addition, MOD depends on variability of immunostaining intensity among tissue sections and tissue thickness \[[@B7]-[@B9]\]. Segmentation of prostatic epithelial nuclei from complex histological images and classification of nuclei as malignant or benign posed a significant challenge to accurately quantifying AR expression by image analysis. This barrier was overcome by combining segmentation algorithms and nuclear morphometry. Qualitative nuclear morphometry was used to characterize the aggressiveness of prostate cancer \[[@B10],[@B11]\]. Kim et al \[[@B12]\] developed a semi-automated technique to identify individual prostatic epithelial nuclei and classify each nucleus as benign or malignant using nuclear shape descriptors. Segmentation of malignant prostatic nuclei allowed application of conventional image analysis algorithms to measure AR expression. A combination of CAS-200 image analysis system with Cell Sheet v2.0 and nuclear morphometric descriptors (NMD) has been used to develop quantitative nuclear grade (QNG), for making clinical, diagnostic and prognostic outcome predictions in prostate cancer \[[@B13]\]. Macros written in commercially available image analysis software such as Adobe Photoshop \[[@B14]\], Optimas \[[@B15]\] and Image Pro Plus \[[@B16]\] have also been used to perform semi-automated image analysis. User interaction is required for object (nuclei) selection, modification of object boundaries and/or selection of thresholds. Dedicated image analysis systems like CAS-200 \[[@B17]\], ACIS \[[@B18]\] and Autocyte \[[@B19]\] are employed in the determination of nuclear features from immunostained images. These systems use dedicated hardware and proprietary software and typically require manual interaction. Schnorrenberg et al. \[[@B20]\] developed an algorithm for automated analysis of breast cancer biopsies. The algorithm transforms the color image into bimodal distribution and identifies the presence and intensity of only the positively immunostained nuclei. Xu et al. \[[@B21]\] developed a software package for automated labeling of rat liver nuclei by integrating various commercial software using macros and Visual Basic. Loukas et al \[[@B22]\] used LabWindows libraries to develop an application for automated counting of nuclei in squamous cell carcinomas of the head and neck. Our image analysis system was developed on networked DEC 5000 workstations, and used programs written in C++, Windows-based Statgraphics 4.1 and Optimas 6.0 \[[@B23]-[@B25]\]. The C++ programs incorporated graphics and image processing libraries (IGLOO) \[[@B26]\]. The image analysis programs were later ported to a PC compatible system on a Linux platform. However, images were collected using Windows-based systems that required image transfer across platforms. The current image analysis software was developed on a Java platform and incorporates all features of previous versions. The software eliminates the need for Statgraphics and Optimas and can be deployed on a variety of platforms with several versions of the software running at the same time without conflict \[[@B27]\]. The program is associated with a Web browser and can be downloaded freely from our website \[[@B28]\]. Once downloaded, Web connection is no longer required. The image analysis software can be divided into four modules: a) detection of potential prostate nuclei; b) removal of artifacts; c) classification of prostate epithelial nuclei and d) measurement of AR expression. Human intervention is only required to create a set of classification parameters. These parameters are used to reduce the effect of local variations in slide preparation and image acquisition on nuclear measurements. The software identifies artifacts, distinguishes epithelial prostate nuclei from endothelial and stromal nuclei and inflammatory cells based on nuclear size and shape. Classification parameters are used to differentiate between immunopositive and immunonegative nuclei. AR expression is quantified by (a) percentage of immunopositive nuclei; (b) percentage of immunopositive nuclear area; and (c) intensity of AR expression among immunopositive nuclei or areas. The development and operation of an automated image analysis system are described. A preliminary study comparing AR expression in paired specimens of benign prostate and prostate cancer from 20 African and 20 Caucasian Americans is presented to demonstrate an application of this system. Hardware and Software ===================== 1. Hardware specifications -------------------------- The imaging system consists of a Leica DMRA2 microscope (Leica Microsystems Inc, Bannockburn, IL) with a Ludl motorized stage controller (Ludl Electronic Products Ltd, Hawthorne, NY), a Hamamatsu 3 Chip CCD camera with controller (Hamamatsu, Bridgewater, NJ) and a Flashpoint 3D image grabber card (Integral Technologies, Indianapolis, IN) in a Pentium IV based PC. The PC used is a Dell Optiplex GX240 (Dell Inc, Round Rock, Texas) with 1.8 GHz Pentium IV processor and 512 Mb RAM. Image Pro Plus 4.5 (Media Cybernetics Inc. Silver Spring, Maryland) software on Windows 2000 (Microsoft Corp, Redmond, Washington) platform was used for image acquisition and Linux based Dell Precision 530 and Windows based Dell Precision 340 were used for image analysis. 2. Software ----------- Java Runtime Environment (Sun Microsystems, Santa Clara, CA) provides the platform to run the image analysis software. Once Java has been installed, the image analysis software can be launched from our web page \[[@B25]\]. Once it is cached on the local computer, Web connection is not required. ### 2.1. Macro for obtaining RGB data A macro written in Visual Basic (Microsoft Corp, Redmond, WA) is used to obtain area and RBG color information from selected regions of the image. The macro was developed in Image Pro Plus 4.5. The macro prompts the user at every step of the process. The user is prompted to open an image and select immunopositive and immunonegative nuclei. The macro guides the user to select individual nuclei using tools provided by Image Pro. The macro calculates the area (in pixels) and the mean RGB color values (0--255) of the pixels enclosed in the selected region. The data is automatically transferred to Microsoft Excel. A numerical tag is attached to each set of values from immunopositive and immunonegative nuclei. ### 2.2. Overview of application software The application program is divided into two modules. #### A. Classification table Data obtained from the macro (Section 2.2.1) is input to the program under \'Parameter Training\'. The program checks the input data file format and displays it in a new window. The results of the analysis, \'Classification Table\', are shown in a new window. Once the user is satisfied and accepts the results, a new set of classification parameters is created and stored. #### B. Image analysis If a classification parameter dataset already exists, it is loaded into the program by clicking on \'Browse\' and locating the file. Parameters can also be entered manually by clicking on \'Modify\' (under Config File). The location of the output directory is set by clicking on \'Browse\' under Output Directory. Files are added and removed by clicking on \'Add\' and \'Delete\' buttons on Input Files frame. The output data files are stored in short (final results only) or long (results from individual nuclei) format with the same prefix as the input image file name. Images created during image analysis can be reviewed and stored by checking corresponding boxes under File in the main window. The graphical user interface (GUI) of the image analysis software is shown in figure [1](#F1){ref-type="fig"}. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Graphical user interface (GUI) of the image analysis software. ::: ![](1475-925X-4-31-1) ::: Methods ======= 1. Study specimens ------------------ Prostate tissue specimens were obtained from archived, paraffin-embedded blocks of radical prostatectomy specimens. The expression of AR in paired specimens of benign prostate and prostate cancer from 20 African and 20 Caucasian Americans was compared to demonstrate an application of this system. Clinical data from these specimens was obtained from prospectively maintained clinical databases. 2. Immunohistochemistry ----------------------- Immunohistochemistry allows for in situprotein localization and computer assisted image analysis of AR while preserving tissue architecture. Polyclonal or monoclonal antibodies target specific epitopes located within cellular structures to visualize epitopes of interest. Archival paraffin-embedded prostate specimens were cut into 6 μm sections and placed on ProbeOn Plus™ microscope slides (Fisher Scientific, Pittsburg, PA). After deparaffination and rehydration through graded alcohols (100%, 95%, 70%), tissue sections were subjected to antigen retrieval in Reveal Citra buffer (Biocare Medical, Walnut Creek, CA) using a pressurized antigen-decloaking chamber for 2 minutes at 120°C and 21 PSI. The sections were cooled to room temperature and blocked for non-specific staining with 2% normal horse serum for 15 minutes at 37°C. Endogenous peroxidase was blocked using 3% hydrogen peroxide diluted in methanol and endogenous biotin was blocked using an Avidin Biotin kit (Vector, Burlingham, CA). Sections were incubated using a capillary gap method with monoclonal antihuman AR antibody F39.4.1 (Biogenex, San Ramon, CA) diluted in Primary Antibody Diluting Buffer (Biomeda Corp., Foster City, CA) at 1:500 for 1 hour at 37°C in a humidified heating block. Sections were incubated with biotinylated anti-mouse IgG (Vector) 1:200 for 30 minutes at 37°C. The signal was then amplified using avidin-biotin complex (ABC) Vector and visualized using 3,3\'-diaminobenzidine (DAB) (Vector). Counterstaining was performed using hematoxylin (Fisher Scientific) for 15 seconds (diluted 1:3 in H~2~O). Slides were dehydrated through graded alcohol and mounted using Permount (Fisher Scientific). Benign and malignant tissues were immunostained in a single batch. 3. Image acquisition -------------------- The images were acquired using a 40:1 objective, N.A. 0.85 for a total magnification of 400×. Contrast and brightness were adjusted by manipulating the gain and exposure time of the camera. Illumination was adjusted to generate maximum contrast while avoiding over- and under-saturation of gray levels. A series of neutral filters were added to confirm the linearity of output in final optical settings. Temporal variation of light output was measured frequently and found insignificant (\<0.2%). Images were sampled randomly throughout histological sections, but areas of necrosis, artifacts and edges were avoided. Each image was captured under the same reproducible conditions. White and Black balance of the camera was performed to ensure the optimal use of the dynamic range of the camera. Ten images were collected from each tissue specimen. Each image consisted of 640 × 480 pixels collected in 24-bit color mode (16.7 million colors) and was stored in an uncompressed tagged image format file (TIFF). 4. Creation of classification parameters ---------------------------------------- Commercial reagents used in the immunohistological staining process are not standardized; thus immunostaining patterns differ between various research labs. When combined with local variations in image acquisition, the resulting automated analysis may produce significant errors. Classification parameters are used to calibrate the nuclear analysis software with each new dataset, thus making the automated image analysis software independent of the type of immunostaining or imaging system used. Figure [2](#F2){ref-type="fig"} shows the block diagram of the steps involved in the creation of classification parameters. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Steps involved in the creation of classification parameters. CP -- Classification Parameters. ::: ![](1475-925X-4-31-2) ::: ### 4.1 RGB dataset A minimum of 200 immunopositive and 200 immunonegative nuclei were randomly sampled from the acquired images. Red, Green and Blue information was extracted from the selected nuclei. A new column \'Class\' was added to the dataset. Objects identified as immunopositive are class 1 and objects identified as immunonegative are class 2. ### 4.2 Classification parameters The dataset was divided into two non-overlapping sets; (a) training set and (b) test set. Each consists of at least 100 immunopositive nuclei and 100 immunonegative nuclei. Classification coefficients were computed from the training set using either of the following methods. #### (a) Linear Discriminant Analysis Let x = {x~hue~, x~saturation~, x~intensity~} denote individual data structure present and x~1~= {x~11~, x~12~, x~13~\..., x~1n1~}, x~2~= {x~21~, x~22~, x~23~\..., x~2n2~} represent class 1 and class 2 datasets from the training set with μ~1~, μ~2~as their corresponding means \[[@B29]\]. The covariance of x~1~and x~2~is: ![](1475-925X-4-31-i1.gif) ![](1475-925X-4-31-i2.gif) The pooled covariance is given by: S~p~= (S~1~+ S~2~) / (n~1~+ n~2~- 2) (3) where n~1~, n~2~are the number of immunopositive and immunonegative nuclei in the training set, respectively. The classification coefficients λ= {λ~1~, λ~2~, λ~3~} and constant C are computed as ![](1475-925X-4-31-i3.gif) C = μλ (5) where S^-1^is the inverse of S and μis the mean of μ~1~and μ~2~. The classification function is of the form: G(x) = λ~1~x~hue~+ λ~2~x~saturation~+ λ~3~x~intensity~- Constant (6) where G(x) is the classification score. #### (b) Logistic Regression Let x = {x~hue~, x~saturation~, x~intensity~} denote individual data structure present, X = {x~1~, x~2~, x~3~\..., x~n~}, the training set and Y represent a column vector with class information of the test dataset. The probability of Y = 1 in a multiple logistic regression model \[[@B30]\] is given as p = 1 / (1 + e^-βX^) (7) where βis the coefficients vector. The equation can be rewritten as ln(p / (1-p)) = βX (8) Equation (8) represents the log of odds as a linear function of X. Since the values for log of odds is not available, a maximum likelihood function provides the solution. Each dataset can be considered as a Bernoulli trial. That is, it is a binomial with the total number of trials equal to 1. Consequently for the i^th^observation ![](1475-925X-4-31-i4.gif) Assuming all datasets are independent, the likelihood function is given by ![](1475-925X-4-31-i5.gif) The log of the likelihood function is given by ![](1475-925X-4-31-i6.gif) The parameter vector βare obtained by maximizing (11) using the efficient Newton-Raphson iterative technique. The classification function is of the form: G(x) = β~0~+ β~1~x~hue~+ β~2~x~saturation~+ β~3~x~intensity~ (12) where G(x) is the classification score. The classification function was tested on the test set. If z= {z~hue~, z~saturation~, z~intensity~} is an individual data structure in the test set, it is classified as class 1 if G(z) \> 0, otherwise, it is class 2. The classification scores were compared with actual scores and a classification table was constructed. If the percentage of class 1 nuclei and class 2 nuclei identified correctly is greater than 85, then the classification coefficients were used. If not, nuclei are randomly sampled again and the process was repeated. Limits for nuclear area are added to the parameter set to eliminate possible artifacts in the image. The lower limit and upper limits of nuclear area were calculated from the dataset Area ~upper~= Area ~Mean~+ 2SD (13) Area ~lower~= Area ~Mean~- 2SD (14) where SD is the standard deviation of the nuclear area measures. 5. Image analysis ----------------- A block diagram of the image analysis program is shown in Figure [3](#F3){ref-type="fig"}. Red, green and blue color information was extracted from the original uncompressed 24-bit color image and stored as 8-bit grayscale images. Discriminant analysis of grayscale histograms was used to determine optimal thresholds for automated segmentation of red, green and blue images \[[@B31]\]. The adaptive threshold was applied using an 80 × 80 pixel window. This window size was chosen because it is about four times the size of a typical nucleus (nuclear diameter \~20 pixels). The three segmented images were combined by a logical OR operation. The combined image was eroded and dilated twice using a 3 step erosion filter (3 × 3 cross, 1 × 3 horizontal and 3 × 1 vertical kernels). Erosion was used to shrink the detected nuclear boundaries and dilation was used to fill the nuclear areas. Artifacts were removed based on size and shape. The nuclear regions were then labeled in raster fashion to create a nuclear mask. Regions not labeled are regarded as background. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Block diagram of the image analysis program. CP -- Classification Parameters, MOD -- Mean Optical Density, NRF -- Nuclear Roundness Factor, OR -- Logical OR operation. ::: ![](1475-925X-4-31-3) ::: Use of red, green and blue images to separate immunopositive from immunonegative nuclei is problematic because the color of stain is mixed with the intensity of stain. An HSI color model was used because it decouples intensity information from color information \[[@B32]\]. The hue, saturation and intensity component images were multiplied by their corresponding discriminant coefficients from the parameter file and combined to form a single image. A nuclear mask was applied to the image and the resulting nuclear areas were classified as immunopositive or immunonegative depending on their classification score. Figure [4](#F4){ref-type="fig"} shows part of an image at different stages of image processing. The precise number of nuclei measured may be inaccurate due to the presence of nuclear overlap or clusters of nuclei. Addition of an upper limit for nuclear area measurement creates a reproducible error. Nuclear shape limits were also used to separate epithelial nuclei from artifacts, endothelial and stromal nuclei and inflammatory cells. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### (clockwise). A -- Part of the original image, B -- Intensity image, C -- Image after addition of R, G and B segmented images, D -- Image after erosion (twice), E -- Image after dilation (twice), F -- Nuclear mask with intensity information. ::: ![](1475-925X-4-31-4) ::: The nuclear mask was applied on the intensity image to obtain the intensity mask image. MOD of each nuclear area present in the image is calculated as: ![](1475-925X-4-31-i7.gif) where N is the total number of pixels in a nuclear mask, I~i~is the intensity level of the pixel i, and I~o~is the intensity level of the background measured in each field of view. NRF is the ratio of the radius of the circle the perimeter of which is equivalent to the measured perimeter to the radius of the circle of which is equivalent to the measured area. The NRF of each nuclear object is given by: ![](1475-925X-4-31-i8.gif) where A is the measured nuclear area and perimeter P is calculated using a chaining approximation, using weights (1, 4, 6, 4, 1). More information on MOD, area and perimeter calculations can be found in earlier publications \[[@B13],[@B23]\]. Results ======= The classification coefficients were computed using linear discriminant analysis and by logistic regression. There was no statistical difference in the results obtained by the two methods (Student\'s t test, P = 0.3896). The classification table is shown in Table [1](#T1){ref-type="table"}. The percent positive nuclei and percent nuclear area were similar by race in both benign prostate hyperplasia and prostate cancer (Table [2](#T2){ref-type="table"}). In prostate cancer epithelial nuclei, African Americans exhibited 38% higher levels of AR immunostaining than Caucasian Americans (two sided Student\'s t-tests; P \< 0.05). Intensity of AR immunostaining was similar between races in benign prostate. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Classification Table. AR -- Androgen Receptor. ::: Predicted ------------------------------------ --------------- -------------- Number of AR positive nuclei (106) 104 (98.11%) 2 (1.89%) Number of AR negative nuclei (136) 2 (1.47%) 134 (98.53%) ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Androgen receptor expression in benign prostate and prostate cancer tissue specimens from 20 African Americans and 20 Caucasian Americans. AA -- African American, CA -- Caucasian American, BP -- Benign Prostate, CaP -- Prostate Cancer, P value (two sample comparison). ::: MOD Percent Positive Area Percent Positive Nuclei -------- ------------- --------------------------- ----------------------------- ------------- ------------- ------------- AA 0.34 ± 0.09 0.32 ± 0.06 0.37 ± 0.20 0.45 ± 0.22 0.25 ± 0.15 0.28 ± 0.18 CA 0.25 ± 0.06 0.30 ± 0.08 0.38 ± 0.25 0.35 ± 0.18 0.26 ± 0.12 0.25 ± 0.11 P \< 0.05 P \> 0.05 P \> 0.05 P \> 0.05 P \> 0.05 P \> 0.05 ::: Discussion ========== The differences measured in the intensity of AR expression in prostate cancer were consistent with previous studies. Gaston et al. \[[@B33]\] reported higher AR expression in benign prostate hyperplasia and prostate cancer in African Americans in a comparison of 25 African Americans and 25 Caucasian Americans. Higher levels of AR expression were also found in malignant glands from a homogenous population of native African men when compared to similar tissues from Anglo-Saxon Caucasian men \[[@B34]\]. Small sampling size and non-matched African American and Caucasian American tissue specimens used in this pilot study may account for the similarity of AR expression in benign prostate hyperplasia. The research need for accurate measurement of AR expression in archival specimens to dissect the role of AR in prostate pathology led to a 19 year quest to develop a robust and automated image analysis system. The system has been used to measure cellular proliferation, apoptosis and tumor morphology. The system was also used to develop a sampling strategy for prostatic tissue microarray \[[@B35]\]. However extensive user training and computational knowledge of the operating system was required to operate the system. The current version of the system is platform independent and can be easily shared across research groups. The software GUI is intuitive and requires minimal user training. The system was also found independent of the imaging system as long as images were acquired using the same protocol. The constraint in the current version of the software is the image size. Image size is required to be a multiple of the segmentation window (80 × 80 pixels) for optimal performance. The option to set background to predetermined grayscale value has made it possible to differentiate regions of protein proliferation in dual-labeled, paraffin-embedded prostate tissue \[[@B36]\]. New features added to the system can be made available to all users by web-based transfer. Two methods to compute the classification coefficients are presented. The choice between the two methods depends on the training set data. If the data is found (or assumed) to come from multivariate normal distribution, linear discriminant analysis is computationally more efficient. If the multivariate assumption is violated, logistic regression should be used \[[@B37]\]. The program performs local segmentation to detect nuclei and then uses color information to classify nuclei. Linear discriminant analysis is employed at the outset to generate classification coefficients. This priori information is required due to non-standardized immunostaining and image collection methods across medical institutions and research laboratories. The addition of classification parameters to the program customizes the software for the specimen under study. Nuclear shape information measured from each nuclear object can be combined with other morphologic descriptors to predict clinical, diagnostic and prognostic outcomes in prostate cancer \[[@B13]\]. This new automated method analyzes an image and provides reproducible measurements in less than 10 seconds. The availability of a free, automated system creates new opportunities for testing, evaluation and use of this image analysis system by many research groups who study nuclear protein expression. Authors\' contributions ======================= All authors have contributed equally in the creation of this manuscript. Acknowledgements ================ Research Supported by National Institutes of Health (CA77739) and by Department of Defense Prostate Cancer Research Program (DOD-PC-012004 and DOD-PC-0340). The authors would like to acknowledge the assistance provided by Bahjat Qaqish, Ph.D., Department of Biostatistics, University of North Carolina at Chapel Hill, USA.	PubMed Central	2024-06-05T03:55:57.525837	2005-5-11	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1145186/", "journal": "Biomed Eng Online. 2005 May 11; 4:31", "authors": [ { "first": "Swaroop S", "last": "Singh" }, { "first": "Desok", "last": "Kim" }, { "first": "James L", "last": "Mohler" } ] }
PMC1145187	Introduction ============ COPD is a common disease, affecting 5--10% of the population of North America and Europe \[[@B1]-[@B3]\]. During the past two decades, death and disability from COPD have continued to increase worldwide \[[@B1],[@B4]\]. Although direct cigarette smoking is the major cause of COPD, up to two cases out of ten cannot be explained solely by direct smoking \[[@B5]\]. Environmental tobacco smoke (ETS) exposure, which appears to cause new cases of asthma, could also cause COPD \[[@B6]-[@B8]\]. Because it contains potent airway irritants, ETS could lead to chronic airway irritation, inflammation, and obstruction \[[@B6],[@B9]\]. The role of ETS exposure in causing COPD, however, has received limited attention in epidemiologic studies \[[@B10],[@B11]\]. Using data from a population-based sample of U.S. adults, we examined the association between lifetime ETS exposure and the risk of developing COPD. Methods ======= We used cross-sectional data from a cohort study of U.S. adults to elucidate the impact of lifetime ETS exposure on the risk of developing COPD. Initial recruitment and survey methods have been previously reported in detail \[[@B12]\]. The study was approved by the University of California, San Francisco Committee on Human Research. Briefly, 2,113 adults aged 55 to 75 years were recruited by random digit dialing among residents of the 48 contiguous U.S. states with random over-sampling of geographic areas that had the highest published COPD mortality rates \[[@B13]\]. The random \"hot spot\" sample was further enriched for additional subjects with COPD. The overall study participation rate was 53% among households with an eligible respondent present. Participants completed structured telephone interviews that included health history, work history, smoking, ETS exposure, and sociodemographic characteristics. We used the standard epidemiologic approach to define COPD based on a self-reported physician diagnosis of chronic bronchitis, emphysema, or COPD \[[@B1],[@B14],[@B15]\]. During the telephone interview, subjects were asked whether they had ever received a physician\'s diagnosis of any of several chronic respiratory conditions. Those who reported physician diagnoses of chronic bronchitis or emphysema were considered to have COPD, along with those who specifically reported a physician diagnosis of COPD. We included respondents with COPD who had concomitant asthma because they clinically resemble persons with COPD alone \[[@B16]\]. We obtained spirometry reports from the physicians of a subgroup of 47 participants with COPD. The majority (89%) had airflow obstruction, as indicated by a FEV~1~/FVC ratio of less than 0.70 or an FEV~1~less than 80% predicted. These physiologic data support the validity of our case definition for COPD. We ascertained lifetime cumulative ETS exposure at home and work. Prenatal exposure was evaluated using the following survey item: \"Did your mother smoke cigarettes when she was pregnant with you before you were born?\" Subsequent home ETS exposure was assessed during childhood and adolescence using the following survey item: \"Growing up until age 18, for how many years in total did you live in the same household with someone else who smoked tobacco products?\" In addition, we measured home ETS exposure during adulthood: \"Since age 18, for how many years in total have you lived in the same household with someone else who smoked tobacco products?\" Cumulative lifetime home ETS exposure was calculated using the sum of both home exposures in years. Moreover, we ascertained cumulative lifetime work ETS exposure using the following item: \"Thinking about all of the jobs you have had, for how many years of your employment have you been regularly exposed to another person\'s cigarette smoke inside your workplace?\" Direct personal cigarette smoking was evaluated using standard questions from the National Health Interview Survey \[[@B17]\]. Because workplace ETS exposure may be higher in occupations that involve exposure to other airway irritants and particulates, we also ascertained occupational exposure to vapors, gas, dusts, or fumes during the longest held job using a survey item developed for the European Community Respiratory Health Survey \[[@B18]\]. Statistical analysis -------------------- Statistical analysis was conducted using SAS 8.2 (Cary, NC). Bivariate analysis was conducted using the unpaired t-test for continuous variables and likelihood ratio chi-square test for dichotomous variables. We examined the impact of lifelong cumulative ETS exposure and the risk of COPD. Based on the distribution of cumulative ETS exposure at home and work, we defined quartiles of exposure. For lifetime work exposure, the first quartile was zero years, so the first and second quartiles were collapsed as the referent group (otherwise the first and second quartile groups would both include zero values). We used logistic regression analysis to examine the relationship between each measure of ETS exposure and the risk of self-reported COPD. We used multivariate logistic regression analysis to control for factors that could confound the relation between ETS exposure and COPD, including past smoking history, age, sex, race-ethnicity, educational attainment, marital status, and occupational exposure to VGDF \[[@B12],[@B19]\]. Consistent with the low prevalence of COPD among never smokers, there were too few never smokers with COPD (n = 75) to analyze this stratum separately. To address this issue, we controlled for the potential confounding effects of direct personal smoking in several ways. In the primary strategy, we included a history of ever smoking in the multivariate analysis. In an alternative analysis, we controlled for current smoking and ex-smoking as separate variables. There were no appreciable differences compared to primary analysis (this alternative analysis is not reported). We also restricted the multivariate analysis to subjects who reported no current smoking. There were no substantive differences compared to the primary analysis and these data are not reported. As an additional alternative analysis, we controlled for cumulative lifetime pack-years of smoking. Both home and workplace ETS exposure independently contribute to lifetime cumulative ETS exposure. Consequently, workplace ETS exposure is not a confounder in the putative pathway between home ETS exposure and the development of COPD. Similarly, home ETS exposure does not operate as a confounder in the relationship between workplace ETS exposure and COPD onset. Based on these considerations, we used separate logistic regression analysis models to examine cumulative lifetime home and workplace ETS exposure, rather than including both exposures in the same model. In a secondary analysis, we examined home and work exposure in a mutually adjusted model. Prenatal ETS exposure, which occurs via the placental circulation, may affect lung development and the subsequent risk of COPD by a different causal pathway than postnatal ETS exposure \[[@B21],[@B22]\]. In fact, prenatal ETS exposure might be associated with both postnatal ETS exposure and the risk of COPD, confounding the relationship between postnatal ETS exposure and COPD. To address this possibility, we conducted an additional analysis to examine the independent relation between the two measures of cumulative lifetime ETS exposure (home and work), taking prenatal ETS exposure into account. We also examined whether prenatal ETS exposure, occupational exposure to VGDF, or direct personal smoking modified the association between home and work ETS exposure and the risk of COPD. To accomplish this, interaction terms were evaluated in logistic regression models that included main effects for ETS exposure, the potential effect modifier, and personal smoking history. We evaluated significant statistical interactions for evidence of synergism or antagonism on an additive scale, which is more appropriate than a multiplicative scale for examining how two factors might biologically interact to produce disease. Using the odds ratio as an estimate of the relative risk, we used the formula OR-1 to calculate the relative excess risk \[[@B23],[@B24]\]. If the relative excess risk for exposure to both factors together (e.g., direct smoking [and]{.underline} ETS) is greater than the sum of the relative excess risks of each factor alone (e.g., smoking + ETS), a synergistic effect would be present. If the relative excess risk for exposure to both factors is less than the sum of the relative excess risks of each factor alone, antagonism would be present. As a sensitivity analysis, we repeated the analysis using a more restrictive definition of COPD that included only those who specifically reported a diagnosis of emphysema or COPD, excluding those with chronic bronchitis alone. We used the method of Greenland and Drescher to estimate the population attributable fraction from the multivariate logistic regression analysis controlling for personal smoking, sociodemographic factors, and occupational VGDF exposure \[[@B25]\]. This methodology provides a maximum likelihood estimator of the attributable fraction from the logistic model. Results ======= Subject characteristics ----------------------- The prevalence of COPD was 18% among the cohort of adults aged 55 to 75 years (95% CI 17 to 20%). Adults with COPD were more likely to be female and unmarried; they also had lower educational attainment than those without COPD (Table [1](#T1){ref-type="table"}). A greater proportion of adults with COPD indicated ever smoking cigarettes compared to other members of the general population (81% vs. 56%), with substantially more current smokers (33 vs. 16%). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Personal characteristics in a population-based sample of 2,113 U.S. adults aged 55 to 75 years ::: Characteristic COPD (n = 386) No COPD (n = 1,727) P value ----------------------------- ---------------- --------------------- ---------- Age (years) 64.3 (6.2) 63.9 (6.1) 0.24 Gender (female) 246 (64%) 961 (56%) 0.004 Race-ethnicity (white) 337 (87%) 1495 (87%) 0.70 Married or cohabitating 186 (48%) 1080 (63%) \<0.0001 Educational attainment \<0.0001 High school degree 207 (54%) 709 (41%) Some college 105 (27%) 529 (31%) College or graduate degree 74 (19%) 489 (28%) Cigarette smoking \<0.0001 Current smoker 127 (33%) 279 (16%) Past smoker 184 (48%) 685 (40%) Never smoker 75 (19%) 763 (44%) Proportions are column proportions (of those with and without COPD) P-values from unpaired t-test (age) and likelihood ratio chi-square test ::: ETS exposure ------------ The prevalence of previous prenatal ETS exposure was higher among adults with COPD than those without the condition (8.3% vs. 5.6%, p = 0.051) (Table [2](#T2){ref-type="table"}). Moreover, the lifetime prevalence of [any]{.underline} subsequent lifetime home or workplace ETS exposure was higher among those with COPD than among those without COPD (89% vs. 82%, p = 0.0004 and 67% vs. 60%, p = 0.012, respectively). Persons with COPD were also more likely to have a higher cumulative lifetime exposure to ETS, with a greater proportion of the COPD group having the highest quartile of home and workplace exposure than those without the condition (41% vs. 22% and 34% vs. 24%, respectively) (Table [2](#T2){ref-type="table"}). ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Lifetime cumulative ETS exposure in a population-based sample of adults aged 55--75 years ::: Source of ETS exposure COPD (n = 386) No COPD (n = 1,727) P value ------------------------------- ---------------- --------------------- ---------- Prenatal ETS\* 32 (8.3%) 96 (5.6%) 0.051 Cumulative lifetime home ETS \<0.0001 Quartile 1 (0--9 yrs) 70 (18%) 443 (26%) Quartile 2 (10--21 yrs) 72 (19%) 461 (27%) Quartile 3 (22--41 yrs) 86 (22%) 451 (26%) Quartile 4 (≥42 yrs) 158 (41%) 372 (22%) Cumulative lifetime work ETS 0.0002 Quartile 1 & 2 (0--5 yrs)^†^ 163 (42%) 881 (51%) Quartile 3 (6--22 yrs) 93 (24%) 440 (25%) Quartile 4 (≥23 yrs) 130 (34%) 406 (24%) Proportions are column proportions (of those with and without COPD) P-values from the likelihood ratio chi-square test \Mother smoked during pregnancy ^†^First and second quartile were both zero, so they were combined into one group. ::: ETS exposure and the risk of COPD --------------------------------- Maternal smoking during pregnancy was not statistically associated with a higher risk of COPD, after controlling for personal smoking and sociodemographic covariates (OR 1.41; 95% CI 0.90 to 2.21) (Table [3](#T3){ref-type="table"}). Higher cumulative lifetime home and work exposure were, however, related to a greater risk of developing COPD. The highest quartile of home ETS exposure was associated with a greater risk of COPD (OR 1.68; 95% CI 1.19 to 2.38). The highest quartile of workplace ETS exposure was also associated with a higher risk of COPD (OR 1.60; 95% CI 1.20 to 2.14). After controlling for workplace VGDF exposure, home and workplace ETS exposure remained associated with the risk of COPD (OR 1.55; 95% CI 1.09 to 2.21 and OR 1.36; 95% CI 1.002 to 1.84, respectively). Based on this most adjusted analysis, the population attributable fraction was 11% for the highest quartile of home ETS exposure and 7% for work exposure. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Lifetime cumulative ETS exposure and the risk of COPD in a population based sample of 2,113 U.S. adults aged 55 to 75 years ::: Source of exposure Level of exposure Risk of COPD (unadjusted) Risk of COPD, controlling for smoking Risk of COPD, controlling for smoking, sociodemographic indicators^†^ Risk of COPD, controlling for smoking, sociodemographic indicators, and workplace VGDF^†^ -------------------- --------------------------------- --------------------------- --------------------------------------- ----------------------------------------------------------------------- ------------------------------------------------------------------------------------------- Prenatal Mother smoked in pregnancy 1.54 (1.01 to 2.33) 1.32 (0.86 to 2.01) 1.41 (0.90 to 2.21) 1.36 (0.86 to 2.14) Home 1st quartile (0--9 yrs) 1.0 (referent) 1.0 (referent) 1.0 (referent) 1.0 (referent) 2nd quartile (10--21 yrs) 0.99 (0.69 to 1.41) 0.88 (0.61 to 1.25) 0.92 (0.64 to 1.34) 0.90 (0.62 to 1.31) 3rd quartile (22--41 yrs) 1.21 (0.86 to 1.70) 0.95 (0.67 to 1.35) 0.95 (0.66 to 1.37) 0.94 (0.65 to 1.36) 4th quartile (≥42 yrs) 2.69 (1.97 to 3.68) 1.88 (1.35 to 2.61) 1.68 (1.19 to 2.38) 1.55 (1.09 to 2.21) Work 1st & 2nd quartile (0--5 yrs)\ 1.0 (referent) 1.0 (referent) 1.0 (referent) 1.0 (referent) 3rd quartile (6--22 yrs) 1.14 (0.86 to 1.51) 0.97 (0.73 to 1.30) 1.02 (0.76 to 1.38) 0.91 (0.67 to 1.24) 4th quartile (≥23 yrs) 1.73 (1.34 to 2.24) 1.34 (1.02 to 1.75) 1.60 (1.20 to 2.14) 1.36 (1.002 to 1.84) VGDF = self-reported exposure to vapors, gas, dust, or fumes during longest held job \*First and second quartile were both zero, so they were combined as the referent group (otherwise both lower categories would include zero values) ^†^Multivariate logistic regression analysis controlling for age, sex, race, smoking history, educational attainment, and marital status. Each source of ETS exposure was evaluated in a separate logistic regression model. ::: We also examined the independent relationship between cumulative lifetime ETS exposure and COPD, controlling for prenatal ETS exposure. The highest quartiles of home and workplace ETS exposure were associated with a greater risk of COPD, controlling for personal smoking, sociodemographic factors, occupational VGDF exposure, and prenatal ETS exposure (OR 1.64; 95% CI 1.16 to 2.33 and OR 1.59; 95% CI 1.19 to 2.13, respectively). There was no evidence that prenatal ETS modified the relationship between ETS exposure and the risk of COPD (p values for interaction with home ETS = 0.59 and work ETS = 0.86). There was also no indication that occupational VGDF exposure modified the association between ETS exposure at home or work and the risk of COPD (p for interaction = 0.94 and 0.14, respectively). There was evidence of a statistical interaction between home ETS exposure and a history of ever smoking for the risk of COPD (p = 0.041), whereas there was no interaction for workplace ETS exposure (p = 0.17). The relative excess risk for the highest quartile of home ETS exposure among never smokers was 0.88. The relative excess risk for personal smoking among those with little or no home ETS exposure (i.e., first quartile) was 1.83. The relative excess risk for combined high level home ETS exposure and personal smoking was 4.32, which exceeded the sum of the relative excess risks for high level home ETS and personal smoking (i.e., 4.32 \> 0.88 + 1.83 or 4.32 \> 2.71). These results are consistent with a synergistic effect of home ETS and smoking for COPD risk. In addition, these results underscore the risk conferred by home ETS exposure among never smokers (i.e., excess relative risk of 0.88). ETS exposure and the risk of COPD -- secondary analyses ------------------------------------------------------- In a secondary analysis, we controlled for personal direct smoking history using cumulative lifetime pack-years of smoking. The highest quartile of home and workplace exposure were associated with a greater risk of COPD, controlling for pack-years of personal smoking and sociodemographic characteristics (OR 1.62; 95R CI 1.14 to 2.30 and OR 1.48; 95% CI 1.10 to 1.99, respectively). After controlling for workplace VGDF exposure, home and workplace ETS exposure were still related to a greater risk of COPD (OR 1.49; 95% CI 1.04 to 2.12 and OR 1.24; 95% CI 0.91 to 1.69, respectively). In the latter analysis, the confidence interval for workplace exposure widened and did not exclude no relationship. In other alternative analyses, we also examined lifetime cumulative home and work ETS exposure in the same model (mutually adjusted model). The highest quartile of home (1.55; 95% CI 1.09 to 2.21) and workplace exposure (OR 1.46; 95% CI 1.08 to 1.96) were associated with a greater risk of COPD, controlling for smoking history and sociodemographic characteristics. Emphysema or COPD subgroup -------------------------- In a sensitivity analysis, we examined the subgroup of subjects who reported emphysema or COPD (n = 189), excluding those who reported chronic bronchitis alone (n = 194). Greater cumulative lifetime home and work ETS exposure remained associated with a greater risk of COPD after controlling for all covariates (OR 2.38 for highest quartile; 95% CI 1.42 to 3.90 and OR 1.79; 95% CI 1.21 to 2.65). Discussion ========== ETS exposure, both at home and work, was associated with a greater risk of COPD in this population-based study of older adults, even after taking personal smoking history into account. On a population level, approximately 1 in 11 cases of COPD may be attributed, at least in part, to home ETS exposure; 1 in 15 cases may be attributable to workplace ETS exposure. The previous epidemiologic literature, albeit limited, supports an association between ETS exposure and COPD. A cross-sectional population-based study from Switzerland found a relationship between self-reported ETS exposure during the past 12 months and a higher risk of chronic bronchitis symptoms \[[@B26]\]. A case-control study demonstrated that self-reported ETS exposure was associated with obstructive respiratory disease, defined as asthma, chronic bronchitis, or emphysema \[[@B27]\]. Reports from the Adventist Health Study of Smog (AHSMOG) indicated a relationship between self-reported ETS exposure and a greater risk of \"airway obstructive disease\" (asthma, chronic bronchitis, or emphysema), chronic bronchitis symptoms, and airway obstruction by pulmonary function testing \[[@B28],[@B29]\]. These studies are limited by the lack of a comprehensive and specific definition of COPD (i.e., includes chronic bronchitis, emphysema, and COPD but not asthma), the absence of cumulative lifetime ETS exposure data, and the omission of other occupational exposures that could be correlated with ETS exposure. The results suggest that home ETS exposure and personal smoking may act synergistically to increase the risk of COPD. There are several possible biological mechanisms that could account for this synergistic action. ETS contains potent respiratory irritants, such as formaldehyde and acrolein, which could directly irritate the airways and exacerbate smoking-related airflow obstruction. Both ETS and direct smoking may increase airway permeability, causing increased IgE levels and enhanced allergic sensitization to airborne antigens \[[@B30],[@B31]\]. By this and other mechanisms, ETS and cigarette smoking could act to increase airway inflammation. Other possible mechanisms are combined effects of smoking and ETS on bronchial hyperresponsiveness \[[@B32]\]. Further experimental work will be necessary to elucidate the apparent synergy between ETS exposure and direct smoking. Our results suggest that the highest quartiles of home and work ETS exposure were associated with a greater risk of COPD. Is it therefore possible to conclude that lower levels of ETS exposure are \"safe\" in terms of obstructive lung disease? We believe that our data do not suggest a \"safe\" level of ETS exposure. Based on our results, the 95% confidence intervals for the lower exposure quartiles are compatible with a substantially increased risk of COPD. Moreover, we have previously shown that very low levels of ETS exposure can exacerbate adult asthma \[[@B33]\]. We have also shown that moderate levels of ETS exposure are associated with impaired pulmonary function \[[@B34]\]. Taken together, these results indicate that even low-to-moderate levels of ETS exposure may have deleterious effects on airway function and obstructive lung disease. We used the standard epidemiologic definition of COPD, based on a self-reported physician diagnosis of chronic bronchitis, emphysema, or COPD \[[@B1],[@B14],[@B15]\]. This survey-based approach enabled us to evaluate a population-based sample of adults who resided throughout the continental United States, which ensured generalizable results. On logistical grounds, conducting spirometry among subjects who reside thousands of miles apart would be highly difficult, if not impossible. The use of self-reported physician-diagnosis, however, may have resulted in some misclassification of disease status. Previous work indicated that a similar survey-based definition of COPD had a high positive predictive value (78%) when validated using a blinded medical record review that included spirometry and radiographic studies \[[@B35]\]. Other work confirmed that a self-reported history of COPD is a strong predictor of airflow obstruction \[[@B36]\]. In the subset of our participants with COPD who had available spirometry data, the prevalence of airflow obstruction was very high (89%). In addition, the high prevalence of lifetime smoking in our study, which was more than 80%, supports the diagnosis of COPD. The prevalence of COPD in our sample (18%) was also similar to that reported in two other population-based studies conducted in the United States \[[@B1]\]. Furthermore, reanalysis of our data using a more restrictive definition of COPD that excluded chronic bronchitis did not appreciably affect the results. In sum, misclassification of COPD is not likely to bias our results; if present, such bias would likely be non-differential with respect to ETS exposure and reduce effect estimates towards the null value. Lifetime cumulative ETS exposure was ascertained by self-report, which could have resulted in exposure misclassification. Previous studies have found moderate correlations between self-reported ETS exposure and biomarker levels (e.g., cotinine) or direct personal exposure monitoring (e.g., nicotine) \[[@B33],[@B37]-[@B42]\]. We cannot, however, exclude some systematic misclassification of ETS exposure. For example, persons with COPD, because they have respiratory symptoms, could be more likely to remember and report ETS exposure, upwardly biasing the effect estimates. Because our focus was on lifetime ETS exposure, there is no other available ETS exposure methodology. Cotinine level, the most widely used biomarker for ETS exposure, reflects exposure during the past 1--2 days \[[@B42]\]. Direct exposure monitoring, such as the personal nicotine badge, can only be used for brief periods of up to several weeks \[[@B33],[@B43]\]. Consequently, the only feasible method for lifetime ETS exposure is survey-based. Because smoking is the dominant risk factor for COPD, we cannot completely exclude some residual confounding by smoking. There were too few never smokers with COPD (n = 75) to restrict the overall analysis to never smokers. To address this issue, we controlled for personal smoking history in multivariate analysis, defined as ever smoking or current / past smoking. The multivariate analysis was also restricted to non-current smokers, yielding essentially the same results. We also controlled for cumulative lifetime pack-years of smoking in additional analyses, which continued to show highly significant results for home ETS exposure, but slightly attenuated findings for workplace ETS after VGDF exposure was also controlled. The interaction analysis also supported the elevated risk for home ETS exposure among never smokers. In sum, the results do not indicate that the results can be explained by residual confounding by direct personal smoking history. Conclusion ========== COPD is a leading cause of death and disability among middle-aged adults in developed nations \[[@B1],[@B4],[@B44]\]. Adults with COPD have a 10-fold higher risk of disability compared to members of the general population \[[@B16]\]. Although cigarette smoking is the dominant cause of COPD, other factors, such as occupational exposures, appear to contribute to disease causation \[12, 45\]. Based on our results, we believe that ETS exposure may also be an important cause of COPD. Consequently, public policies aimed at preventing public smoking may reduce the burden of COPD-related death and disability, both by reducing direct smoking and ETS exposure. Abbreviations ============= ETS = environmental tobacco smoke; COPD = chronic obstructive pulmonary disease Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= MDE conceived the study, performed the analysis, and wrote the paper; JB participated in the design and analysis and helped draft the manuscript; Laura Trupin participated in the analysis and drafting of the manuscript; PK participated in the design and drafting of the manuscript; EH participated in the analysis, interpretation of data, and drafting of the manuscript; PB initiated the cohort study, participated in the design, analysis, and drafting of the manuscript.	PubMed Central	2024-06-05T03:55:57.528631	2005-5-12	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1145187/", "journal": "Environ Health. 2005 May 12; 4:7", "authors": [ { "first": "Mark D", "last": "Eisner" }, { "first": "John", "last": "Balmes" }, { "first": "Patricia P", "last": "Katz" }, { "first": "Laura", "last": "Trupin" }, { "first": "Edward H", "last": "Yelin" }, { "first": "Paul D", "last": "Blanc" } ] }
PMC1145188	Introduction ============ As the Acquired Immune Deficiency Syndrome (AIDS) pandemic continues to grow, the number of people affected by the neurological complications of human immunodeficiency virus (HIV) infection expands. Neurological complications, known collectively as neuroAIDS, affect approximately 30% of those infected with HIV \[[@B1]\]. Although our knowledge of the process by which HIV causes brain disease is constantly expanding, we still have only a limited understanding of the underlying pathogenic mechanism leading to disease in the central nervous system (CNS). It has been shown that an increase in the population of brain macrophages is a significant pathological correlate of neurological disease \[[@B2]\], and that most strains of HIV isolated from the brains of individuals with neurological disease are macrophage tropic and utilize the CCR5 co-receptor \[[@B3],[@B4]\]. Macrophages and microglia, related cells of monocytic lineage, are the only cell types consistently infected in the brains of HIV-infected individuals \[[@B5]\]. Damage to neurons is thus indirect, resulting from effects of viral proteins or products of infected macrophages. The ability of HIV to infect macrophages and microglia in vitrois predictive of its neuroinvasiveness \[[@B6]\] and infected monocytes/macrophages are thought to carry HIV into the brain as per the Trojan horse hypothesis \[[@B7],[@B8]\]. Simian immunodeficiency virus (SIV) is closely related to HIV \[[@B9],[@B10]\] and SIV infection of macaques can generate a neuroAIDS-like syndrome that mirrors neuroAIDS in humans, demonstrating the neuropathological hallmarks of neuroAIDS found in HIV-infected humans along with cognitive, motor, and neurophysiological impairments \[[@B11]-[@B15]\]. The similarities between HIV- and SIV-induced neurological disease in humans and macaques, in light of the ethical and practical limitations of performing neurological research in humans, make the rhesus macaque an excellent model for the study of neuroAIDS. There are a variety of strains and molecular clones of SIV that have been used to study aspects of AIDS pathogenesis, many of which are derived from the SIVmac251 strain \[[@B16]\]. Of the molecular clones, the most commonly used is SIVmac239, derived from the SIVmac251 strain by animal passage and tissue culture proviral DNA cloning \[[@B17]\]. SIVmac239 is highly pathogenic in vivoand displays a very high infectious capacity for T cells, but not macrophages, in vitro\[[@B17]\]. Unlike T-cell-tropic strains of HIV, which utilize the CXCR4 but not the CCR5 co-receptor, the T-cell tropism of SIVmac239 may be based on its inefficient use of the relatively low cell-surface CD4 density on rhesus macrophages, rather than co-receptor specificity \[[@B18]\]. Yet this may not fully explain SIVmac239\'s lack of productive macrophage infection, since many studies have found efficient entry, but post-reverse transcriptional blocks in the SIVmac239 life cycle in macrophages \[[@B19]-[@B21]\]. Other studies have examined the molecular aspects of virus recovered ex vivofrom macrophages late in infection, revealing specific nucleotide/amino acid changes in viral genes and their products, which are associated with high levels of infection of macrophages in vitro\[[@B22]-[@B25]\]. In studying SIV cellular tropism, another commonly used clone is SIVmac316, isolated from proviral DNA in lung macrophages of a macaque that died rapidly after infection with SIVmac239 \[[@B24]\]. Tropism studies with this clone and others like it essentially examine viral revertants, examining changes in the viral sequence in the context of the backbone of SIVmac239 \[[@B26]\]. We have taken an independent approach to examine viral properties of SIV in the CNS. Using the SIVmac251 stock, we performed a serial passage of cell-associated virus isolated from the CNS of infected monkeys, followed by production of a cell-free stock of virus from in vitroinfected microglia \[[@B27],[@B28]\]. In this manner, we utilized a forward selection of neuroinvasive variants that exist in, or arose from, the SIVmac251 stock. In this study we discuss the development and analysis of SIV clones derived from virus released by cultured microglia that were isolated from the brains of monkeys infected with microglia-passaged viral stock. Sequencing and characterization of viral tropism and infectious phenotype were then undertaken to analyze genomic and functional characteristics common to these brain-derived viruses. Results ======= Molecular Cloning of Microglia-Derived SIV ------------------------------------------ A total of 43 clones of the 3\' region of SIV were isolated from viral RNA found in the supernatant of microglia cultures derived from the brains of SIVmac182-infected macaques 225 and 321. Of these clones, 24 clones were from animal 225, and 19 clones from animal 321. A portion of gp41 was sequenced in each clone to insure the identity of the clones and to determine if any of the clones contained premature truncations due to stop codons in the gp41 region, a common finding in macrophage-tropic SIVmac239-derived clones. Sequence analysis confirmed that all of the clones were SIV, and that none had truncations in the gp41 region. Infectivity and Cytopathogenicity --------------------------------- Each of these 43 clones containing the 3\' region of SIV was ligated to the 5\' region of SIVmac239, and transfected into 174xCEM cells, a common indicator cell line for SIV infection. Cultures were observed daily for syncytia formation and monitored for infectious virus formation by p27Gag analysis of culture supernatants. Of the 43 viruses, 19 (13 from macaque 225 and 6 from macaque 321) induced syncytia formation in the cultures and/or tested positive for p27Gag production in the culture supernatant. In vitroparameters of cytopathogenicity were then tested, using cells transfected with SIVmac239 as a positive control. SIVmac239 led to a very robust infection in 174xCEM cells, rapidly producing high levels of p27Gag (1.5 ng/ml) and syncytia. Pronounced cytopathic effects ensued, and the cells in the SIVmac239-transfected cultures were all dead by day 11 post-transfection. The microglia-derived molecular clones could be divided into three groups. A first group of five clones: 109, 129, 141, 142, and 169, produced the most consistent and robust infections, with all clones in this group generating syncytia, consistently high levels of p27Gag (above 1.5 ng/ml) and high levels of cell death by day 15 (Table [2](#T2){ref-type="table"}). ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### In vitrosyncytia formation and viral antigen production. The molecular clones, derived from microglia of the indicated monkey, were tested by transfection and subsequent growth in 174xCEM cells. ::: Clone Monkey Longest time to Syncytia Formation Longest time to Detectable p27 ----------- ------------ ---------------------------------------- ------------------------------------ Group 1 109 225 7 days 7 days 129 225 7 days 7 days 141 225 7 days 7 days 142 225 7 days 7 days 169 321 7 days 7 days Group 2 108 225 12 days 12 days 122 225 12 days 12 days 144 225 12 days 12 days 146 225 15 days 18 days 153 321 9 days 10 days 159 321 12 days 12 days Group 3 104 225 18 days 14 days 115 225 18 days 14 days 116 225 18 days 18 days 134 225 18 days 18 days 143 225 Never 18 days 164 321 12 days 12 days 171 321 Never 15 days 173 321 Never 15 days Control SIVmac239 4 days 7 days ::: A second group of six clones: 108, 122, 144, 146, 153 and 159, also produced high levels of p27Gag and syncytia, although syncytia formation was slower than syncytia formation by the first group and these clones did not induce large amounts of cell death, with cells just beginning to die by 15--18 days post transfection. The remaining eight molecular clones that demonstrated signs of productive infection were 104, 115, 116, 134, 143, 164, 171 and 173. This group of clones 3 was considered least pathogenic of the three in vitrobecause they did not cause any detectable cell death and were unable to consistently generate syncytia and detectable p27Gag levels by day 18 experiments, although all of them did generate syncytia and high levels of p27Gag in at least one experiment, with the exception of 171 and 173, which did not generate syncytia despite p27Gag production. Sequence Analysis ----------------- The nine clones judged the most pathogenic in vitrowere chosen for complete sequence analysis. These clones included all five from the most pathogenic group; 109, 129, 141, 142 and 169, as well as clones 108, 122, 153 and 159 from the second group. Clones from the second group were picked because they generated the highest levels of p27Gag in that group. Of the clones chosen, 108, 109, 122, 129, 141, and 142 were isolated from macaque 225 and clones 153, 159 and 169 were isolated from macaque 321. We fully sequenced the envand nefgenes as well as the 3\' LTR of each of these molecular clones. These sequences were used to develop a consensus sequence for all nine of the molecularly derived clones to be used for further analysis. The nine clones showed a remarkable degree of similarity in the three gene products analyzed, with more differences in the TM portion of Env and in Nef than in the SU portion of Env. The nine clones differed from the consensus sequence by zero to six amino acids of 1,144 in amino acids in all three genes sequenced (Table [3](#T3){ref-type="table"}). Comparison with other common molecular clones of SIV that were also derived from the SIVmac251 stock showed marked differences from the consensus sequence of the brain-adapted viruses in these regions (Table [3](#T3){ref-type="table"}). The detail of these differences can be found in Table [4](#T4){ref-type="table"}, showing that the brain-adapted genotype lacks the commonly seen truncation in gp41, and possesses 18 unique amino acids across Env and Nef. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Comparison of encoded amino acids (AA) from clones described here (top) with other SIV molecular clones (bottom). The number (\#) and percent (%) changes (Δ) in the indicated regions of Env and Nef are given. ::: Clone Monkey Derived From \#AA Δ in gp120 \#AA Δ in gp41 \#AA Δ in Nef Δ consensus in Env & Nef ----------------- ------------ ---------------------------------------- --------------------- -------------------- ------------------- ------------------------------ 108 225 Viral RNA from Microglia supernatant 0 1 0 0.08% 109 225 Viral RNA from Microglia supernatant 1 0 0 0.08% 122 225 Viral RNA from Microglia supernatant 0 0 0 0.00% 129 225 Viral RNA from Microglia supernatant 1 1 0 0.17% 141 225 Viral RNA from Microglia supernatant 1 2 1 0.35% 142 225 Viral RNA from Microglia supernatant 1 0 1 0.17% 153 321 Viral RNA from Microglia supernatant 1 3 2 0.52% 159 321 Viral RNA from Microglia supernatant 0 2 2 0.35% 169 321 Viral RNA from Microglia supernatant 0 2 1 0.26% SIVmac1A11 251-79 Proviral DNA from Tissue culture cells 28 8\* 18 4.72% SIVmac32H (pJ5) 32H Proviral DNA from Tissue culture cells 24 14 19 4.98% SIVmac316 316-85 Proviral DNA from Tissue culture cells 19 8\* N/A 3.07%\\ SIV/17E-Fr 17E Proviral DNA from Brain & Macrophages 22 11\* 20 4.63% SIVmac239 239-82 Proviral DNA from Tissue culture cells 18 14 15 4.11% \truncated gp41, \\Env only. ::: ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Predicted amino acid residue at the indicated location in the SU region of Env. Bold indicates unique amino acids in clones 129 and 169. ::: Env -- gp120* 67 79 127 132 134 135 144 153 176 178 309 382 385 475 511 ------------------ -------- -------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- SIVmac239 V N I S T S M A K D M G D G D SIVmac316 M N I S T S M A E D M R D G D SIV17E-Cl M N I S T S M A N D I R N G D SIV17E/Fr M N I S T S M A N D I R D G D SIVmac32H L E L P A \- M T K D M R D G D SIVmac1A11 L E S A P \- M V K D I G D G N Clone 129 L D S S T P V V K G I R D R N Clone 169 L D S S T P V V K G I R D R N \sequence not available, -- no amino acid residue. ::: Additional sequence analysis was performed on the gp41 cytoplasmic tail regions of SIVmac251, SIVmac182 and cDNA derived from the supernatant of microglia from macaques 225 and 321. These reactions were performed on the cytoplasmic tail because of the variable sequence and frequent truncations found in this region and used a different set of primers than previous sequencing reactions in order to serve as independent confirmation of the observed amino acid changes. The brain-adapted viruses developed a unique sequence in this area, with 4 synonymous and 14 non-synonymous changes in the gp41 cytoplasmic tail regions of both 225 and 321 cDNA when compared with the original progenitor strain SIVmac251. There were also two synonymous and five non-synonymous changes found when comparing 225 and 321 cDNA with that of their immediate progenitor, SIVmac182. The synonymous and non-synonymous changes from both SIVmac251 and SIVmac182 were identical in uncloned PCR products from both 225 and 321 microglia supernatants, and the resulting amino acid changes in gp41 can be seen in Table [5](#T5){ref-type="table"}. ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### Predicted amino acid residue at the indicated location in the TM region of Env. Bold indicates unique amino acids in clones 129 and 169. ::: Env -- gp41* 573 631 676 713 734 737 741 751 752 760 764 767 785 802 821 850 855 ----------------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- SIVmac239 K K D M Q I P R D S W E S L T G T SIVmac316 T K D V Q I P G D S W Stop \- \- \- \- \- SIV17E-Cl K K N V Q \* \* \* \* \* \* \* \* \* \* \* \* SIV17E/Fr K K D M Q I P G D S Stop \- \- \- \- \- \- SIVmac251 K N D M Q I P G D S W E S L T G T SIVmac32H K D D M Q I P G D S W E S L T G T SIVmac1A11 K K D M Stop \- \- \- \- \- \- \- \- \- \- \- \- Clone 129 K D D M Q T Q G D R W E N F A R T Clone 169 K D D M Q T Q G G R W E N F A R A \sequence not available, -- no amino acid residue. ::: Macrophage Infection -------------------- It has previously been shown that a majority of viruses isolated from the brains of individuals with neurological disease are macrophage tropic, and the ability to infect macrophages is thought to be key in the induction of neurological disease. Because the molecularly cloned viruses were all isolated from the brains of rhesus macaques that suffered from encephalitis, we hypothesized that these viruses were macrophage tropic. To test this hypothesis, we isolated macrophages from rhesus macaque PBMC and inoculated them with six of the molecularly cloned viruses. Three of the viruses that were fully sequenced, clones 108, 122 and 142, were dropped from this analysis because their sequences were greater than 99% similar to another clone being used for these infections. In order to generate more uniform results between experiments, all inoculations were performed using spinoculation. Spinoculation effectively eliminates potential differences in viral infection resulting from viral attachment to the cell, because it moves viruses directly onto their cellular targets \[[@B29],[@B30]\]. Viruses derived from all six of the molecular clones replicated well in macrophages, and while the levels of p27Gag produced fluctuated between infections, the pattern of p27Gag production between the six viruses was remarkably consistent between experiments with the exception of the p27Gag production by clone 153, which induced strong p27Gag production on day 4 of this experiment, had reduced levels on day 10, and had inconsistent p27Gag production in subsequent experiments. Clones 109, 129, and 169 consistently produced the highest levels of p27Gag (Figure [1](#F1){ref-type="fig"}). Although strong p27Gag production was induced by clone153 on day 4 of this experiment, p27Gag levels were much reduced by day 10, and p27Gag production with this clone was inconsistent in subsequent experiments. As controls, the T-cell-tropic clones SIVmac239 and the molecular clone SIVmac251 were used and neither of these clones was able to produce detectable p27Gag after ten days, whereas the SIVmac251stock (the progenitor strain for SIVmac239, the SIVmac251 molecular clone and our microglia serial passage) successfully infected macrophages but produced relatively low levels of p27Gag (data not shown). Based on these results and the genetic similarity of the brain-adapted clones, subsequent experiments focused only on clones 129 and 169 as representatives of this particular genotype of SIV. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Viral replication in macrophages of six brain-adapted clones on days four (left) and ten (right) days post-inoculation. Cultures were inoculated with virus produced from the indicated clones. Culture media was replaced one day before collection at the indicated day and a 24-hour supernatant was then analyzed by ELISA to determine p27Gag levels. ::: ![](1743-422X-2-44-1) ::: Both the 129 and 169 molecular clones produced a similar infectious phenotype following spinoculation, producing p27Gag levels that peaked early after infection and then slowly declined (Figure [2](#F2){ref-type="fig"}). This particular infection phenotype, an early peak in p27 levels, was seen in all infections with either of these two viruses, although SIVmac129 consistently produced higher peak p27 levels than SIVmac169. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Daily SIV production in macrophage cultures. Macrophages from two different rhesus monkeys (a -- 359, b -- 420) were inoculated with virus produced from the indicated clones. Culture media was replaced each day and the removed supernatant was analyzed by ELISA to determine 24-hour p27Gag levels. This figure is representative of the infectious phenotype for these viruses in this cell type seen in four separate experiments. ::: ![](1743-422X-2-44-2) ::: Microglia Infection ------------------- Along with perivascular macrophages, microglia are the most commonly infected cells in the brain \[[@B31]\]. In order to determine if the molecularly cloned viruses were able to infect microglia, these cells were isolated from the brains of 3 animals (uninfected with SIV, but treated with methamphetamine for other studies). The microglia were spinoculated with virus prepared from clones129, 169, or SIVmac239. The molecular clone of SIVmac251 was also used to infect microglia from two of the three animals. Viruses from both clones 129 and 169 were able to productively infect microglia, producing very high levels of p27Gag within the first 5 days, and then slowly declining out to day ten (Figure [3](#F3){ref-type="fig"}). While the peak levels of p27Gag production were reached more slowly in microglia than in macrophages, taking between 4 and 6 days rather than 3 or 4, the pattern of infection was similar to that seen in macrophage infections with these viruses. SIVmac239 was unable to infect microglia, failing to produce detectable levels of p27Gag in any of the infections. The SIVmac251 molecular clone was only able to infect microglia at a very low level, producing detectable p27Gag only sporadically during the course of infection (Figure [3](#F3){ref-type="fig"}). ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Daily SIV production in microglia cultures. Microglia were inoculated with virus produced from the indicated clones. Culture media was replaced each day and the removed supernatant was analyzed by ELISA to determine 24-hour p27Gag levels in the supernatant on each day of infection. This figure is representative of the infectious phenotype of these viruses in this cell type in three separate experiments using microglia from independent monkeys. ::: ![](1743-422X-2-44-3) ::: Spread in Macrophage Infection ------------------------------ Because macrophage tropism is a common characteristic of viruses found in the brains of individuals with neuroAIDS, the spread of virus between macrophages may carry important implications for understanding disease progression in the CNS. To assess spread of infection through macrophages, we enumerated the number of infected cells in cultures of primary macrophages inoculated with the viruses prepared from molecular clones 129 and 169, in comparison to the parental SIVmac251stock. In order to account for donor-related differences in macrophage infection, we examined infection of monocyte-derived macrophages in fourteen experiments, utilizing cells from four different macaques. The percentages of infected macrophages varied between experiments (most likely due to host-cell differences or the variability inherent in working with primary cells), but each viral clone produced infection within 48 hours of inoculation. Viruses generated from clones 129 and 169 both produced separate, unique infection patterns in all animals (Figure [4](#F4){ref-type="fig"}). In particular, clone 129 followed a similar infection pattern found by measuring supernatant p27Gag (shown in Figure [2](#F2){ref-type="fig"}), showing the highest percent of infected cells early on (17.1% by day 4). In contrast, clone 169 manifested the highest percent of infected cells later (8.2% on day 6) (Figure [4](#F4){ref-type="fig"}). The SIVmac251stock showed a very low level of infection, with 3% or less cells infected each day. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Infected cell percentage in macrophage culture. The percentage of macrophages infected each in chamber slide culture of primary macaque macrophages infected with brain-adapted clones 129 and 169 and the SIVmac251stock. In fourteen separate experiments, slides with primary macrophages from four different macaques were inoculated with SIV and then fixed and stained with DAPI and p27Gag at the indicated times. Data for each day are the average from these experiments. ::: ![](1743-422X-2-44-4) ::: To further examine the behavior of this brain-adapted phenotype in terms of viral spread, we performed a second macrophage infection over a 10-day time course, again using clones 129 and 169 as well as the SIVmac239 and SIVmac316 viruses, and the parental SIVmac251stock. Due to the limited number of macrophages derived from each animal, infections were only analyzed by staining and p27Gag analysis on days 1, 4, 7 and 10 post-infection. Since macrophages isolated from different rhesus macaques vary in their in vitrosusceptibility to infection, in order to account for this variation, macrophages from macaques with different susceptibilities were used for each experiment. Although the percentages of infected cells (Figure [5A](#F5){ref-type="fig"}, [5C](#F5){ref-type="fig"}) and p27Gag levels (Figure [5B](#F5){ref-type="fig"}, [5D](#F5){ref-type="fig"}) varied between animals, the general infection pattern, with one notable exception, was remarkably similar. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Spread and production of five different isolates of SIV in primary macrophages. Data from two monkeys (A, B -- \#408; C, D -- \#411) are shown for infected cell percentage (A, C) and supernatant p27Gag (B, D) from triplicate cultures in chamber slides. ::: ![](1743-422X-2-44-5) ::: Macrophages from donor monkey 408 showed no significant infection by any virus on day one post-inoculation, with all chambers showing a percentage (\<3%) of infected cells and no detectable p27Gag levels in the supernatant (Figures [5A,B](#F5){ref-type="fig"}). Day four post-inoculation was much different, showing increases in percent of cells infected and p27Gag levels in chambers infected with SIVmac316 (71% of cells infected with p27Gag levels of 11.5 ng/ml) and clone 129 (30% and 6.1 ng/ml). Chambers inoculated with clone 169, SIVmac239 and SIVmac251stock had an extremely low infected cell percentage (\<1%) and no detectable p27Gag levels in the supernatant. On day seven post-inoculation, the SIVmac316-infected cell percentage remained constant (66.6%) while p27Gag levels dropped (4.6 ng/ml). The percentage of infected cells in chambers infected with clone 129 dropped more than two-fold (12.8%) and supernatant p27Gag levels in the supernatant also dropped (0.8 ng/ml). No change was seen in SIVmac239 and SIVmac251stock infections. Surprisingly, the infected cell percentage in cultures inoculated with clone 169 greatly increased (14.6%), as did p27Gag levels in the supernatant (1.9 ng/ml). These trends all continued on day 10, with a greater than 3-fold increase in infected cell percentage (58.4%) and supernatant p27Gag levels (6.2 ng/ml). SIVmac316-infected chambers had reduced infected cell percentage (42.8%), though p27Gag levels in the supernatant increased (6.5 ng/ml). Chambers infected with clone 129 showed both reduced infected cell percentage (8.1%) and slightly reduced p27Gag levels in the supernatant (0.7 ng/ml). There was no change in the chambers infected with SIVmac239. Chambers inoculated with the SIVmac251stock showed increased infected cell percentage (5.5%) and a large increase in p27Gag levels in the supernatant (2.7 ng/ml) at this last time point. Somewhat similar infection trends were seen on infection of macrophages from macaque 411 (Figure [5C,D](#F5){ref-type="fig"}). On day one post-inoculation, infected cell percentages in chambers inoculated with SIVmac316 (10.1%) and clone 129 (4.1%) were both higher than those seen in the 408 infections, but neither infection generated detectable p27Gag levels in the supernatant. Inoculation with SIVmac316 and clone 129 then followed the same general pattern found in the 408 infections above, with increases in infected cell percentage and p27Gag levels in the supernatant on day 4, followed by a decline on days 7 and 10. Supernatant p27Gag levels were in general lower than those found from the macrophages from macaque 408. However, in contrast to the results found in the other monkey\'s macrophages, here clone 169 did not lead to detectable infected cells or p27Gag levels in the supernatant at any point in the infection. Furthermore a low infected cell percentage was seen in SIVmac251stock and SIVmac239 infected chambers on days 4 and 10 respectively, but neither of these cultures had detectable p27Gag levels at any point during the infection. Discussion ========== To improve understanding of the viral factors that allow certain strains of HIV/SIV to induce brain disease, we analyzed molecular clones generated from SIVmac251stock through serial passage in infected microglia in vivo. After the final passage, several brain-adapted molecular clones were isolated from two macaques, 225 and 321, both of whom died with SIVE. Sequence analysis of the envand nefgenes of these viral clones showed remarkable genotypic homology, as the all the brain-adapted clones differed from their consensus sequence by less than 0.55%. Separate examination of the cytoplasmic tail of gp41 from uncloned viral sequence derived from the same microglia supernatant used to isolate the brain-adapted clones provided independent verification of these similarities. The uniqueness of this genotype is seen in comparison with other common SIVs like SIVmac239, SIVmac316 and SIVmac17E-Fr, as the brain-adapted genotype differs from the envand nefgene sequences of the other virus by three to five percent (Tables [3](#T3){ref-type="table"},[4](#T4){ref-type="table"},[5](#T5){ref-type="table"}). Furthermore, uncloned viral sequence derived from both 225 and 321 microglia showed a large number of identical non-synonymous mutations in the gp41 cytoplasmic tail when compared with both SIVmac251 and SIVmac182. Because of the way they were derived, the sequential differences from other SIVmac251 derived viruses and the exceptional homology between the separately isolated viral clones, the amino acid changes in these clones likely represent positive selection for adaptations beneficial towards survival and infection in the brain and CNS. The large number of identical non-synonymous mutations from the original SIVmac251 strain supports this idea, as non-synonymous mutations are only maintained if they are beneficial adaptations. When these brain-adapted clones are examined in light of their separate derivation from two different animals and the relative frequency of mutations during viral infection of macaques, the extraordinary homogeneity and uniqueness of the sequence of these brain-adapted clones, along with the identical and numerous non-synonymous mutations found in virus from both animals, strongly indicates that this genotype developed as a result of viral adaptation to the unique environment found in the CNS. Numerous studies using SIV have linked brain infection to macrophage tropism \[[@B32]-[@B34]\], and indeed, virus from all of the brain-derived clones were macrophage tropic. Representative clones derived from each macaque (clone 129 from macaque 225, and clone 169 from macaque 321), were further characterized, and found not only to be infectious in primary macaque macrophages, but also in primary macaque CD4+ T-cells and primary macaque microglia. In addition to characterizing the tropism of the brain-adapted clones, the macrophage and microglia infection experiments also demonstrated a distinct, reproducible infectious phenotype associated with this viral genotype. Numerous studies have analyzed macrophage-tropic viruses found in animals infected with the T-cell tropic clone SIVmac239, a phenomenon that is thought to be due to a series of amino acid changes in the envelope gene. Using site-directed mutagenesis, Mori and colleagues found five amino acid changes in the SIV envelope, V67M, K176E, G382R from the SU region and K573T, R751G from the TM region that increased p27Gag production in macrophage cultures \[[@B24]\]. Kodama and colleagues examined 10 viral clones derived from the brain of macaque 316-85, and found that all contained 9 amino acid changes in the envelope gene, including the V67M and G382R changes as well as seven additional changes in the SU region; T158A, D178N, P334L/R, D385N, V388A and P421S and R751G in the TM region \[[@B35]\]. As macrophage tropism is thought to be crucial to viral infection in the brain, the emergence of amino acid changes that contribute to this characteristic is not surprising in clones derived from the brain. Indeed, the brain-adapted clones from this study were found to contain numerous changes in Env, including the G382R and R751G changes mentioned above (see Table [3](#T3){ref-type="table"}). Macrophage tropism alone is not sufficient for induction of neurological illness \[[@B26]\], and many studies cite specific genes thought to be important in the induction of CNS disease. Mankowski and colleagues demonstrated the primacy of the envelope gene in neurovirulence in the development of SIV/17E-Fr, and examination of this clone by Flaherty and colleagues found macrophage tropism associated changes V67M, P334R and G382R in the envelope, along with several unique amino acid changes \[[@B23],[@B26]\]. These and other studies demonstrate that while there is a group of amino acid changes associated with the macrophage tropic aspect of brain adaptation, it is an additional set of amino acid changes that allow a virus to successfully adapt to the environment of the brain. The brain-adapted clones described in this paper are a perfect example of this, with several macrophage tropism associated changes in the envelope, along with a group of entirely unique amino acid changes; seven in gp120 and seven in gp41. However, other studies of brain adaptation in SIV find that specific Nef sequences are also important for infection and replication of virus in the brain, implying that similar neuroadaptive changes may also occur in the nefgene \[[@B26],[@B36],[@B37]\]. Barber and colleagues have noted five amino acid changes in Nef between SIVmac239 and SIV/17E-Fr, including two, P12S and E150K, that mediate distinct Nef/kinase associations and may be important in neuroadaptation \[[@B38]\]. Also a study of four pigtailed macaques infected with SHIV containing neffrom an SIV background demonstrated that the majority of nefgenes amplified from an animal with neurologic disease encoded two amino acid changes, T110A and A185T \[[@B39]\]. The brain-adapted genotype described in this paper does not contain any of the Nef changes seen in SIV/17E-Fr but it does contain the T110A residue, along with four other amino acid changes unique to this genotype among the viruses examined. It is clear from the number of common changes found in various brain derived SIV clones that certain amino acid residues in Env and Nef are important to the adaptation of SIV to the CNS environment, including, but not limited to, those changes contributing to macrophage tropism. As with many derivatives of SIVmac251, the brain-adapted viruses described here do match the amino acids for several of the reversions noted in SIVmac239 described above, notably G382R and R751G in Env and T110A in Nef. The brain-adapted genotype described here also contains some amino acid differences from SIVmac251 and SIVmac239 that match other neurovirulent viruses like SIV17E/Fr, although it lacks the commonly seen truncation in the cytoplasmic region of gp41 and contains several amino acid changes that are unique to this group of viruses. Unlike the macrophage-tropic and neurovirulent variants of SIVmac239 described above, the brain-adapted viruses isolated in this study were selected, or evolved from, a stock which could infect macrophages naturally in the course of infection, which was then preserved and selected for by subsequent passage through the brains of other animals. We had previously reported that analysis of brain proviral DNA for a portion of gp120 revealed selection of homogeneous sequences over the course of microglia passage \[[@B27]\]. Here, we have expanded these studies to the entire Env as well as Nef, examination of viral RNA, and characterization of infectious phenotypes in macrophage and microglia. Unlike many other studies with SIV, the changes found in the brain-adapted genotype described in this paper are an example of forward selection, rather than reversions that function largely in the context of the backbone of the non-macrophage-tropic non-brain-derived strain SIVmac239. It is interesting to note that three of the amino acid changes found in gp41 of the brain-adapted clones are not found in the same region of their immediate progenitor, SIVmac182, therefore they developed during the course of infection in each animal. The presence of identical amino acid changes in viruses derived from two separate animals indicates that the genotype described by these clones results from convergent evolution rather than random mutation, and therefore the particular changes found in the genomes of these clones may be important to the natural adaptation of the virus to the brain. It is worth noting that both clones 129 and 169 show a distinct, reproducible phenotype of infection characterized by an early peak in viral p27 production, usually in the first 2--4 days, followed by a gradual decline over the remainder of the experiment. While these clones have been shown to cause disease in vivo, the presence of this phenotype in vivois still uncertain. However, if this phenotype does occur during in vivoinfection, it could be a method by which brain-adapted SIV establishes residence in the brain, using an initial burst of virus to seed macrophages and microglia, which, once infected, lie low, allowing the neutralization sensitive macrophage-tropic virus to avoid immune detection until virus in the periphery has sufficiently weakened the immune system for successful virus replication in the brain. This approach might be particularly effective for this virus, given its ability to infect microglia and the low-turnover rate of that cell type, to establish a viral archive in the brain. However, this phenotype has only been shown to be distinct among the five viruses tested in these experiments and may also be an experimental artifact due to spinoculation or another aspect of our infection protocol, so the relevance of this particular phenotype to SIV remains uncertain. Based on the data presented above; the method of isolation, the uniqueness of the Env and Nef sequences and their derivation through forward evolution in the brain and unique phenotype and tropism, it is clear that the genotype represented by the clones described in this study represents a new, brain-adapted genotype of SIV with a unique phenotype of infection. Further study of this genotype and its unique phenotype, both alone and in comparison with other brain-adapted strains of SIV, will hopefully generate greater understanding for the viral basis of brain disease and dementia, providing new targets and avenues of research in order to more effectively combat this disease. Methods & Materials =================== Cell isolation and culture -------------------------- All cell culture media and components were obtained from Invitrogen (Carlsbad, California), fetal calf serum (FCS) from Hyclone (Logan, Utah) and recombinant human macrophage colony-stimulating factor (M-CSF) from Peprotech (Rocky Hill, New Jersey). 174xCEM cells were obtained from the NIH AIDS and Reference Reagent Program (Germantown, Maryland), and grown in RPMI-1640 containing 10% FCS, 10 mM Hepes, and 100 U/mL penicillin, 100 μg/mL streptomycin and 250 ng/mL fungizone (added as an antibiotic cocktail called PSF). Monocyte-derived macrophages were prepared from rhesus peripheral blood mononuclear cells (PBMC). PBMC were isolated by Histopaque 1077 (Sigma-Aldrich, St. Louis, Missouri) gradient centrifugation, and cultured at 2 × 10^6^cells/ml in complete macrophage media (RPMI-1640, 10% FCS, 5% autologous monkey serum, 10 mM HEPES, 10 ng/ml M-CSF, 1% PSF). Cultures were incubated briefly with PBS and then washed with serum-free RPMI-1640 on days 1, 3, and 5 post-isolation to remove non-adherent cells. After washing on day 5 post-isolation, cells were washed one additional time with PBS and then incubated in Versene (Invitrogen) on ice, agitating every 1--2 minutes. Cells were then shaken loose, enumerated in a Coulter Z2 counter (Beckmann-Coulter, Fullerton, California) and resuspended in complete macrophage media. Purity was ascertained by FACS analysis. To prepare microglia, the meninges, surface vessels, and choroid plexi were removed from brains taken from sterile phosphate buffered saline-perfused rhesus monkeys. Then brain tissue was homogenized with a Tenbroeck tissue homogenizer, and microglia purified by collagenase/DNase digestion and Percoll gradient centrifugation as described \[[@B28]\]. Cells were enumerated in a Coulter Z2 counter, resuspended in complete microglia media (RPMI-1640, 10% FCS, 20 mM HEPES, 50 ng/ml M-CSF), and plated in 48-well plates at a concentration of 7.5 × 10^5^cells/well. Purity was ascertained by FACS analysis. Non-adherent cells are removed by washing 1, 3 and 5 days after isolation, and microglia used for experiments on day 6 post-isolation. Animal infection ---------------- Two rhesus macaques were infected intravenously with cell-free stock of SIVmac182, obtained by 3 generations of in vivoserial passage of SIVmac251 (44, 47). Both of these animals developed neuroAIDS with SIV encephalitis. Macaque 225 had a natural, rapid course of disease, requiring sacrifice at 80 days post-viral inoculation. Monkey 321 was treated with an anti-CD8 antibody at the time of infection \[[@B40]\] and also had a rapid course of disease, requiring sacrifice at 108 days post-viral inoculation. Following sacrifice, microglia from both animals were cultured as described above and cell-free supernatant was drawn from these cultures on days 8 through 12 post-isolation. Molecular cloning ----------------- Nucleic acids were isolated from the supernatants of cultured microglia, isolated from macaques 225 and 321, using a QIAamp Viral RNA mini kit (QIAGEN, Valencia, California) and were used to generate viral cDNA using the primer SIV GSP (see Table [1](#T1){ref-type="table"} for all primer sequences). Nested PCR were then used to amplify the cDNA, using the primers 10505 and 6516 for the first round of amplification, and primers 10505 and Sph2 for the second round of amplification. The PCR product was analyzed by gel electrophoresis to insure it was the proper size and then excised using the crystal violet gel excision system (Invitrogen). Approximately 10 ng of cDNA was cloned into the TOPO-XL vector (Invitrogen) and transformed into Max Efficiency STBL2 cells (Stratagene, La Jolla, California). Plasmid DNA was then isolated, restriction mapped, and initially sequenced using primer For8, corresponding to a region of the gp41 gene. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Sequence of oligonucleotide primers used for reverse transcription, PCR, and sequencing of SIV. ::: Primer* Sequence --------------------------------- ------------------------------------ Reverse Transcription SIVGSP TGCTAGGGATTTTTCCTGCYTCGGTTT Nested PCR 6516 CTCGCTTGCTAACTGCA CTTCTAATCATATCTA Sph2 GCATGCTATAACACATGCTATTGTAAAAAGTGTT 10505 AAGCAGAAAGGGTCCTAACAGACCAGGGTCTTCA Molecular Clone Sequencing For 1 AACTCAGTGCCTACCAGATAA For 2 TGGCATGGTAGGGATAATAGGA For 3 ATAAAAGAGGGGTCTTTGTGCT For 4 AACTGCAGAACCTTGCTATCG For 5 GTTTGATCCAACTCTAGCCTACAC For 6 ATGACAGGGTTAAAAAGAGACAAGA For 7 GAATTGGTTTCTAAATTGGGTAGA For 8 GAGGCACAAATTCAACAAGAGAAG For 9 CATACAGAAAACAAAATATGGATGA For 10 TCCTGGTCCTGAGGTGTAATCCTG Rev 1 CGCAAGAGTCTCTGTCGCAGAT Rev 2 AGAGGGTGGGGAAGAGAACACTG Rev 3 ACTTCTCGATGGCAGTGACC Rev 4 CCAGACATAATGGAGACTGGTAA Rev 5 AGAGTACCAAGTTTCATTGTACTC Rev 6 AGGCAAATAAACATTTTTGCCTAC Rev 7 GAGCGAAATGCAGTGATATTTATACATCAAG Population PCR and Sequencing 8877For ATAGCTGGGATGTGTTTGGC 8534For GCTGGGATAGTGCAGCAACAGCAAC 8406For CTACTGGTGGCACCTCAAG 9452Rev CGAGTATCCATCTTCCAC 9625Rev CCTACCAAGTCATCATCTTCCTCA 9880Rev ATCCTCCTGTGCCTCATCTG 10203Rev ATCAAGAAAGTGGGCGTTCCCGACC ::: 174xCEM Transfection and Infections ----------------------------------- Plasmid DNA was prepared in STBL2 cells (Invitrogen) and then isolated by miniprep (QIAgen). Clones were prepared for ligation by digestion with SphI, phenol extraction and ethanol precipitation. Following this preparation, 0.4 μg of the plasmid containing the newly-derived 3\'-regions of the viral cDNA was ligated to 1.6 μg of an SphI/SalI restriction digest of the previously characterized 5\' fragment of the SIV genome, constituting the first 6516 bp of SIVmac239 (p239SpSp5, NIH AIDS Research and Reference Reagent Program), which had been subcloned into the Litmus38 vector (New England BioLabs, Beverly, Massachusetts). This process was also used to combine the 3\' regions of SIVmac239, and the molecular clone SIVmac251, both originally obtained from the NIH AIDS Research and Reference Reagent Program, with the 5\' region of SIVmac239. The ligation products were transfected into 174xCEM cells by DEAE-Dextran. Transfections were observed for syncytia for 18 days and supernatant was taken from each culture periodically for use in p27Gag ELISA (SIV Core Antigen Assay, Beckmann-Coulter) to test for the presence of infectious virus. Clones were monitored microscopically for the speed of syncytia formation and speed of cell death in culture, and the media was assayed by ELISA for p27Gag level in order to characterize the robustness of infection. All transfections were each performed independently at least twice, with an extra transfection being performed on clones with inconsistent results in the first two transfections. Viral stocks were produced in 174xCEM cells using either transfection, as above, or infection, for the uncloned SIVmac251 stock and a viral stock of the SIVmac239/316EM\/nef-open molecular clone. SIVmac239/316/Macropahge Variant is a macrophage tropic derivative of SIVmac239 (this clone is hereafter referred to as SIVmac316, kindly provided by Dr. R. Desrosiers, New England Primate Research Center, Southborough, Massachusetts). Following infection or transfection, when syncytia become prevalent throughout the infected culture, cells were washed and resuspended in fresh growth media every twenty-four hours, with the cell-free supernatant aliquoted and stored at -80°C. Each stock was run in triplicate p27Gag ELISA for quantification. Sequence Analysis ----------------- The 3\' regions of nineteen molecular clones were completely and bi-directionally sequenced using a battery of primers designed to cover the full length of that fragment of the genome. The group of primers used included M13 Forward and M13 Reverse as well as ten SIV-specific forward primers and seven SIV-specific reverse primers, listed on Table [1](#T1){ref-type="table"}. All sequencing reactions on the molecular clones were performed at the TSRI Nucleic Acids Core Facility. Additional sequencing reactions were performed on uncloned cDNA prepared from cell-free viral RNA isolated from SIVmac251 stock, SIVmac182, and from cultured microglia from macaques 225 and 321. RNA was converted to cDNA using the First Strand cDNA synthesis kit (Marligen Biosciences, Ijamsville, MD) and amplified by using the primers 8534For, 8406For, 9880Rev and 10203Rev, followed by purification of the PCR product and sequence analysis with 8777For, 9452Rev and 9625Rev in order to sequence the gp41 cytoplasmic tail region (bases 8750--9499). All direct sequencing reactions of PCR products were performed by Retrogen (San Diego, California). Macrophage and Microglia Infection ---------------------------------- Macrophages were purified as above, and plated in 48-well plates at 7.5 × 10^4^per well. The macrophages were grown in these plates for one day and then inoculated in triplicate with 4 ng (p27Gag) of the virus stock in complete macrophage media. Infection was facilitated by spinoculation (30). Briefly, virus was added to each well and plates were centrifuged at 1,200 × g for 2 hours at 25°C. Plates were then incubated at 37°C for 22 hours, washed twice with Hanks Balanced Salt Solution (Invitrogen), and fed with complete media. Every 24 hours, supernatant was removed and stored at -80°C, while cells were fed with fresh complete macrophage media. Microglia were infected following the above spinoculation protocol. Following spinoculation media was collected and replaced with fresh complete microglia media daily, similar to the collection protocol followed in macrophage infections. Macrophage chamber slide culture infection and analysis ------------------------------------------------------- PBMCs were isolated and enumerated as described above, and monocytes purified using the MiniMacs magnetic separation system (Miltenyi Biotech, Cologne, Germany) using paramagnetic anti-CD11b beads. The cells were then enumerated and resuspended in complete macrophage media and plated onto 8-well chamber slides (Fisher Scientific, Pittsburgh, Pennsylvania) at 10^5^cells per well. Media was changed on day 3 post-isolation, and on day 6 removed and replaced with the viral inoculum containing 6 ng (p27Gag) of virus stock diluted in complete macrophage media, performed in triplicate. Slides were incubated for 24 hours at 37°C, after which the supernatant was gently aspirated, cells washed, and fed with complete macrophage media. Day one post-inoculation and every three days following, one set of slides (corresponding to 3 wells for each infecting virus and one uninfected well for each virus) was processed for staining and the supernatant from each chamber was carefully aspirated and stored at -80°C. Each chamber was then washed once with phosphate buffered saline (PBS, Invitrogen) and fixed by incubating 20 minutes at room temperature in 3% paraformaldehyde/PBS. After fixing, chambers were gently washed with distilled water and incubated in 5% bovine serum albumin (BSA) in PBS for five minutes. Cells were incubated for 1 hour with a 1:100 dilution of FA2 (mouse anti-p27Gag antibody, originally obtained from the NIH AIDS Research and Reference Reagent Program) in 5% BSA in PBS. Following this incubation, chambers were washed once with 1% BSA in PBS, and incubated for 1 hour with a 1:500 dilution of rhodamine-conjugated goat-anti-mouse antibody (Molecular Probes, Eugene, Oregon) in 5% BSA in PBS. Cells were then washed once with PBS and incubated for 3 minutes with 4,6-diamidino-2-phenylindole dihydrochloride hydrate (DAPI, Sigma-Aldrich) diluted to 10 ug/mL in PBS. All of these reactions were performed at room temperature in the dark. Cells were then gently washed once with PBS, chamber divisions were removed and coverslips were attached using Vectashield mounting medium (Vector Labs, Burlingame, California) and sealed with nail polish. Slides were stored in the dark at 4°C. One day after staining, each set of slides was assessed by computer-assisted fluorescence microscopy using the Axiovision program (Carl Zeiss, AG, Germany). Three 10x fields were assessed from each chamber using rhodamine and DAPI filters, generating nine fields per virus per day examined. Three fields taken of an uninfected control well were used to control for background fluorescence. All observation and focusing was performed under DAPI fluorescence to avoid experimenter bias. The images were analyzed using OpenLab (Improvision, Lexington, MA) to elucidate the total number of cells and infected cells in each field. The total number of cells per field was determined by counting the DAPI-stained nuclei, while the number of infected cells in each field was determined using Boolean operations to overlay the FA2-Rhodamine stained viral gag with those nuclei. The number of positive-staining cells in the control wells was used to account for background fluorescence. Competing interests =================== The author(s) declare that they have no competing interests. ::: {#T6 .table-wrap} Table 6 ::: {.caption} ###### Predicted amino acid residue at the indicated location in Nef. Bold indicates unique amino acids in clones 129 and 169. ::: Nef* 7 12 39 43 49 53 75 93 110 112 119 150 201 206 ------------ ------- -------- -------- -------- -------- -------- -------- -------- --------- --------- --------- --------- --------- --------- SIVmac239 M P Y P G R E E T S M E S P SIV/17E-Fr M S Y P G L K Q T S I K S P SIVmac251 R P S L G L E E T S M E S P SIVmac32H R P S L G L E E T S M E S P SIVmac1A11 M P S L G L E E T S M E S P Clone 129 R P S L D R E E A T M E A S Clone 169 R P S L D R E E A T M E A S ::: ::: {#T7 .table-wrap} Table 7 ::: {.caption} ###### Predicted amino acid residues in the gp41 cytoplasmic tail that differ in the indicated viruses. Residues found only in the microglia-passage derived SIVmac182 stock, and/or in the resulting virus released from monkey 225 and 321 microglia, are in bold; italics further indicated residues found in the microglia virus but not SIVmac182. For amino acid 747, a mixed sequence was present in SIVmac182 capable of encoding either amino acid listed. Population-based sequencing was performed on the SIVmac251, SIVmac182, 225 microglia, and 321 microglia viral stocks. SIVmac239 is provided for reference. ::: 695 737 741 747 748 751 758 760 785 794 802 808 821 831 837 850 858 ------------------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- --------- SIVmac239 (clone) V I P E S R G S S V L T T H V G R SIVmac251 (stock) V I P E S G S S S A L A T Q G G G SIVmac182 (stock) I I Q G/E G G G R S V F T T Q G R R 225 microglia I T Q G G G G R N V F T A Q G R R 321 microglia I T Q G G G G R N V F T A Q G R R ::: Acknowledgements ================ This work was support by NIH grants R01 MH59468, R01 NS045534, and P30 MH62261. PJG was supported by NIH training grant T32 AI07606. 174xCEM cells, p239SpSp5, p239SpE\' nef Open, and SIVmac251 phage were provided by NIH AIDS Research and Reference Reagent Program. This is manuscript \#16804-NP from The Scripps Research Institute.	PubMed Central	2024-06-05T03:55:57.531088	2005-5-12	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1145188/", "journal": "Virol J. 2005 May 12; 2:44", "authors": [ { "first": "Peter J", "last": "Gaskill" }, { "first": "Debbie D", "last": "Watry" }, { "first": "Tricia H", "last": "Burdo" }, { "first": "Howard S", "last": "Fox" } ] }
PMC1149491	Over the past century and a half, we have made enormous progress in assembling a coherent picture of genetic evolution---that is, changes in the pools of genetic information possessed by populations, the genetic differentiation of populations (speciation) (see summaries in \[[@pbio-0030194-b1],[@pbio-0030194-b2]\]), and the application of that understanding to the physical evolution of Homo sapiens and its forebears (\[[@pbio-0030194-b3]\]; e.g., \[[@pbio-0030194-b4],[@pbio-0030194-b5]\]). But human beings, in addition to being products of biological evolution, are---vastly more than any other organisms---also products of a process of "cultural evolution." Cultural evolution consists of changes in the nongenetic information stored in brains, stories, songs, books, computer disks, and the like. Despite some important first steps, no integrated picture of the process of cultural evolution that has the explanatory power of the theory of genetic evolution has yet emerged. Much of the effort to examine cultural evolution has focused on interactions of the genetic and cultural processes (e.g., \[[@pbio-0030194-b6]\], see also references in \[[@pbio-0030194-b7]\]). This focus, however, provides a sometimes misleading perspective, since most of the behavior of our species that is of interest to policy makers is a product of the portion of cultural evolution \[[@pbio-0030194-b8]\] that occurs so rapidly that genetic change is irrelevant. There is a long-recognized need both to understand the process of human cultural evolution per se and to find ways of altering its course (an operation in which institutions as diverse as schools, prisons, and governments have long been engaged). In a world threatened by weapons of mass destruction and escalating environmental deterioration, the need to change our behavior to avoid a global collapse \[[@pbio-0030194-b9]\] has become urgent. A clear understanding of how cultural changes interact with individual actions is central to informing democratically and humanely guided efforts to influence cultural evolution. While most of the effort to understand that evolution has come from the social sciences, biologists have also struggled with the issue (e.g., p. 285 of \[[@pbio-0030194-b10]\], \[[@pbio-0030194-b11]\], and p. 62 of \[[@pbio-0030194-b17]\]). We argue that biologists and social scientists need one another and must collectively direct more of their attention to understanding how social norms develop and change. Therefore, we offer this review of the challenge in order to emphasize its multidisciplinary dimensions and thereby to recruit a broader mixture of scientists into a more integrated effort to develop a theory of change in social norms---and, eventually, cultural evolution as a whole. What Are the Relevant Units of Culture? {#s2} ======================================= Norms (within this paper understood to include conventions or customs) are representative or typical patterns and rules of behavior in a human group \[[@pbio-0030194-b18]\], often supported by legal or other sanctions. Those sanctions, norms in themselves, have been called "metanorms" when failure to enforce them is punished \[[@pbio-0030194-b17],[@pbio-0030194-b19],[@pbio-0030194-b20]\]. In our (liberal) usage, norms are standard or ideal behaviors "typical" of groups. Whether these indeed represent the average behaviors of individuals in the groups is an open question, and depends on levels of conformity. Conformity or nonconformity with these norms are attributes of individuals, and, of course, heterogeneity in those attributes is important to how norms evolve. Norms and metanorms provide a cultural "stickiness" (p. 10 of \[[@pbio-0030194-b21]\]) or viscosity that can help sustain adaptive behavior and retard detrimental changes, but that equally can inhibit the introduction and spread of beneficial ones. It is in altering normative attitudes that changes can be implemented. Here, we review the daunting problem of understanding how norms change, discuss some basic issues, argue that progress will depend on the development of a comprehensive quantitative theory of the initiation and spread of norms (and ultimately all elements of culture), and introduce some preliminary models that examine the spread of norms in space or on social networks. Most models of complex systems are meant to extract signal from noise, suppressing extraneous detail and thereby allowing an examination of the influence of the dominant forces that drive the dynamics of pattern and process. To this end, models necessarily introduce some extreme simplifying assumptions. Early attempts to model cultural evolution have searched for parallels of the population genetic models used to analyze genetic evolution. A popular analogy, both tempting and facile, has been that there are cultural analogues of genes, termed "memes" \[[@pbio-0030194-b22],[@pbio-0030194-b23]\], which function as replicable cultural units. Memes can be ideas, behaviors, patterns, units of information, and so on. But the differences between genes and memes makes the analogy inappropriate, and "memetics" has not led to real understanding of cultural evolution. Genes are relatively stable, mutating rarely, and those changes that do occur usually result in nonfunctional products. In contrast, memes are extremely mutable, often transforming considerably with each transmission. Among humans, genes can only pass unidirectionally from one generation to the next (vertically), normally through intimate contact. But ideas (or "memes") now regularly pass between individuals distant from each other in space and time, within generations, and even backwards through generations. Through mass media or the Internet, a single individual can influence millions of others within a very short period of time. Individuals have no choice in what genes they incorporate into their store of genetic information, and the storage is permanent. But we are constantly filtering what will be added to our stored cultural information, and our filters even differentiate according to the way the same idea is presented \[[@pbio-0030194-b24],[@pbio-0030194-b25]\]. People often deliberately reduce the store of data (for example, when computer disks are erased, old books and reprints discarded, etc.), or do so involuntarily, as when unreinforced names or telephone numbers are dropped from memory. Such qualitative differences, among others, ensure that simple models of cultural evolution based on the analogy to genetic evolution will fail to capture a great deal of the relevant dynamics. A model framework addressed to the specific challenges of cultural evolution is needed. In the models discussed below, the most basic assumption is that the spread (or not) of norms shares important characteristics with epidemic diseases. In particular, as with diseases, norms spread horizontally and obliquely \[[@pbio-0030194-b14]\], as well as vertically, through infectious transfer mediated by webs of contact and influence. As with infectious diseases, norms may wax and wane, just as the popularity of norms is subject to sudden transitions \[[@pbio-0030194-b3]\]. On the other hand, there are unique features of cultural transmission not adequately captured by disease models, in particular the issue of "self-constructed" knowledge, which has long been a source of interest, and the development of problem-solving models in psychology (\[[@pbio-0030194-b26], [@pbio-0030194-b27]\]; D. Prentice, personal communication). New syntheses are clearly required. Microscopic Dynamic {#s3} ======================= Substantial progress has been made toward the development of a mathematical theory of cultural transmission, most notably by Cavalli-Sforza and Feldman \[[@pbio-0030194-b14]\], and Boyd and Richerson \[[@pbio-0030194-b11]\]. Cavalli-Sforza and Feldman consider the interplay between heritable genetic change and cultural change. This is an important question, addressed to the longer time scale, with a view to understanding the genetic evolution of characteristics that predispose individuals to act in certain ways in specified situations. For many of the phenomena of interest, however, individual behaviors have not evolved specifically within the limited context of a single kind of challenge, but in response to a much more general class of problems. Efforts to provide genetic evolutionary explanations for human decisions today within the narrow contexts in which they occur may be frustrated because generalized responses to evolutionary forces in the distant past have lost optimality, or even adaptive value. Extant human behaviors for example may be the relics of adaptations to conditions in the distant past, when populations were smaller and technology less advanced. Attempts to understand them as adaptive in current contexts may therefore be futile. Thus, we prefer to take the genetic determinants of human behavior (that, for example, we react strongly to visual stimuli) as givens, and to ask rather how those initial conditions shape individual and social learning \[[@pbio-0030194-b3]\]. Similar efforts have been undertaken by others, such as Henrich and Boyd \[[@pbio-0030194-b28]\] and Kendal et al. \[[@pbio-0030194-b20]\]. The sorts of models put forth by Cavalli-Sforza and Feldman, Boyd and Richerson, and others are a beginning towards the examination of a colossal problem. To such approaches, we must add efforts to understand ideation (how an idea for a behavior that becomes a norm gets invented in first place), and filtering (which ideas are accepted and which are rejected). How many ideas just pop up in someone\'s brain like a mutation? How many are slowly assembled from diverse data in a single mind? How many are the result of group "brainstorming?" How, for example, did an idea like the existence of an afterlife first get generated? Why do ideas spread, and what facilitates or limits that spread? What determines which ideas make it through transmission filters? Why are broadly held norms, like religious observance, most often not universal (why, for instance, has atheism always existed \[[@pbio-0030194-b29],[@pbio-0030194-b30]\])? Ideas may be simply stated, or argued for, but transmission does not necessarily entail the reception or adoption of behaviors based on the idea, e.g., \[[@pbio-0030194-b31]\]. What we accept, and what gets stored in long-term memory, is but a tiny sample of a bombardment of candidate ideas, and understanding the nature and origin of filters is obviously one key to understanding the life spans of ideas and associated behaviors once generated. The Emergence of Higher-Level Structure: Some Simple Models {#s4} =========================================================== Our filters usually are themselves products of cultural evolution, just as degrees of resistance of organisms to epidemics are products of genetic evolution. Filters include the perceived opinions of others, especially those viewed as members of the same self-defined social group, which collectively attempt to limit deviance \[[@pbio-0030194-b32]\]. "Conformist transmission," defined as the tendency to imitate the most frequent behavior in the population, can help stabilize norms \[[@pbio-0030194-b28]\] and indeed can be the principal mechanism underlying the endogenous emergence of norms. The robustness of norms can arise either from the slow time scales on which group norms shift, or from the inherent resistance of individuals to changing their opinions. In the simplest exploration of this, Durrett and Levin (unpublished data) have examined the dynamics of the "threshold" voter model, in which individuals change their views if the proportion of neighbors with a different opinion exceeds a specified threshold. Where the threshold is low, individuals are continually changing their opinions, and groups cannot form ([Figure 1A](#pbio-0030194-g001){ref-type="fig"}). In contrast, at high thresholds, stickiness is high---opinions rarely change---and the system quickly becomes frozen ([Figure 1B](#pbio-0030194-g001){ref-type="fig"}). Again, groups cannot form. In between, however, at intermediate thresholds (pure conformist transmission), groups form and persist ([Figure 1C](#pbio-0030194-g001){ref-type="fig"}). In the simplest such models in two dimensions, unanimity of opinions will eventually occur, but only over much longer time periods than those of group formation (see also \[[@pbio-0030194-b20]\]). When the possibility of innovation (mutation) is introduced in a model that considers linkages among traits and group labels, and where individuals can shift groups when their views deviate from group norms sufficiently, multiple opinions and multiple groups can persist, essentially, indefinitely ([Figure 1D](#pbio-0030194-g001){ref-type="fig"}). ::: {#pbio-0030194-g001 .fig} Figure 1 ::: {.caption} ###### \(A) Long-term patterning in the dynamics of two opinions for the threshold voter model with a low threshold. \(B) Long-term patterning in the dynamics of two opinions for the threshold voter model with a high threshold. Note the existence of small, frozen clusters. \(C) Long-term patterning in the dynamics of two opinions for the threshold voter model with an intermediate threshold. Note the clear emergence of group structure. \(D) Long-term patterning in a model of social group formation, in which individuals imitate the opinions of others in their (two) groups, and others of similar opinions, and may switch groups when their views deviate from group norms. ::: ![](pbio.0030194.g001) ::: The formation of groups is the first step in the emergence of normative behavior; the work of Durrett and Levin shows that this can occur endogenously, caused by no more than a combination of ideation and imitation. The existence of a threshold helps to stabilize these groups, and to increase stickiness; furthermore, if threshold variation is permitted within populations, these thresholds can coevolve with group dynamics. What will the consequences be for the size distribution of groups, and for their persistence? Will group stability increase, while average size shrinks? What will be the consequences of allowing different individuals to have different thresholds, or of allowing everyone\'s thresholds to change with the size of the group? When payoffs reward individuals who adhere to group norms, and when individuals have different thresholds, will those thresholds evolve? The answers to such questions could provide deep insights into the mechanisms underlying the robustness of norms, and are ripe for investigation through such simple and transparent mathematical models. Modeling may also shed light on why some norms (like fashions) change so easily, while others (like foot binding in imperial China) persist over centuries, and more generally on how tastes and practices evolve in societies. Norms in art and music change rapidly and with little apparent effort at persuasion or coercion. But three-quarters of a century of communism barely dented the religious beliefs of many Russians, despite draconian attempts to suppress them \[[@pbio-0030194-b35]\], and several centuries of science have apparently not affected the belief of a large number of Americans in angels and creationism (e.g., \[[@pbio-0030194-b36],[@pbio-0030194-b37]\]). Then there are the near-universal norms, such as the rules against most types of physical assault or theft within groups that, although they vary in their specifics, are interpreted as necessary to preserve functional societies. Group-selection explanations for such phenomena (e.g., \[[@pbio-0030194-b12]\]) are, we argue, neither justified nor necessary (see also pp. 221--225 of \[[@pbio-0030194-b38]\], \[[@pbio-0030194-b39]\]). Such behaviors can emerge from individual-based models, simply involving rewards to individuals who belong to groups. There are degrees: the evolution of cooperation is facilitated by tight interactions, for example when individuals interact primarily with their nearest neighbors \[[@pbio-0030194-b40],[@pbio-0030194-b41]\], and the payoffs that come to individuals from such cooperation can enhance the tightness of interactions and the formation of groups. This easily explains why mutually destructive behaviors, like murder, are almost universally proscribed. Group benefits can emerge, and can enhance these effects, but it is neither necessary nor likely that group selection among groups for these behaviors overrides individual selection within groups when these groups are not composed of closely related individuals \[[@pbio-0030194-b42]\]. Simple models could address such things as the role of contagion in cultural evolution, recognized in one of the first works on psychology \[[@pbio-0030194-b43]\] in the context of religious revivals and belief, as what has been described as "pious contagion" (p. 10 of \[[@pbio-0030194-b30]\]). But models must also address issues such as the roles of authority or moral entrepreneurs (individuals engaged in changing a norm) \[[@pbio-0030194-b32]\], to say nothing of the impacts of advertising and the norm-changing efforts of the entertainment and other industries. In reality, we are intentioned agents who act with purpose. In maturing, we master the norms that have been evolved over a long period, but to which we may adapt in different ways and even (in the case of moral entrepreneurs) strive to change. For a moral entrepreneur, a group that is too small may have little influence and be not worth joining. But large groups may be too difficult to influence, so also may not be worth joining. For such individuals, there is likely an optimal group size, depending on the change the individual wants to effect. Groups also introduce ancillary benefits of membership that change the equation. Such considerations influence decisions such as whether to join a third party effort in a political campaign; understanding the interplay between individual decisions and the dynamics of party sizes is a deeply important and fascinating question, with strong ecological analogies. Groups, collectively, must also wrestle with the costs and benefits of increasing membership, thereby enhancing influence while potentially diminishing consensus and hence the perceived benefits to members. Innovation and Conservatism {#s5} =========================== Cultural evolution, like biological evolution, contains what we like to call the "paradox of viscosity." Evolving organisms must balance the need to change at an appropriate rate in response to varying environmental conditions against the need to maintain a functioning phenome. This trade-off between conservatism and adaptability, between stability and exploration, is one of the central problems in evolutionary theory. For example, how much change can there be in the genes required to maintain adaptation in a caterpillar without lethally affecting the structure and functioning of the butterfly (p. 303 of \[[@pbio-0030194-b44]\])? Conservatism in religion might be explained by the lack of empirical tests of religious ideas. But even in military technology and tactics, where empirical tests are superabundant, changes are slower than might be expected. For example, the British high command in World War I did not react rapidly to the realities of barbed wire, massed artillery, and machine guns \[[@pbio-0030194-b45]\]. Even so, the conservatism of the generals may be overrated \[[@pbio-0030194-b46]\]. Macroscopic Dynamics {#s6} ==================== We have thus far examined the evolution of norms in isolation---as how the views of individuals (and thus the constituents of a pool of nongenetic information) change through time. But everywhere in common discourse and technical literature, it is assumed that norms are bundled into more or less discrete packages we call cultures, and that those packages themselves evolve. Recall that everyday notions such as that American culture of the 1990s was very different from that of the 1960s, that Islamic culture did not undergo the sort of reformation that convulsed Christian culture (for example, \[[@pbio-0030194-b47]\]), and that Alexander the Great carried Greek culture throughout the Mediterranean and as far east as Persia. The problem of defining "cultures" in cultural evolution seems analogous to that of defining "species" (or other categories) in genetic evolution. There has been a long and largely fruitless argument among taxonomists over the latter \[[@pbio-0030194-b48]\], and an equally fruitless debate in anthropology (and biology) on the definition of culture \[[@pbio-0030194-b39], [@pbio-0030194-b49]\]. Again, we suggest that the parsing of the various influences that create and sustain norms and cultures are ripe for theoretical modeling, but it must begin to incorporate the full richness on multiple scales of space, time, and complexity. Durrett and Levin \[[@pbio-0030194-b3]\] develop a model integrating the dynamics of clusters of linked opinions and group membership; appropriate extensions would allow group characteristics to evolve as well, but on slower time scales. The oversimplicity of models of symmetric imitation on regular grids, as represented in our simple models, must give way to those that incorporate fitnesses and feedbacks, as well as asymmetries and power brokers, on more complex networks of interaction \[[@pbio-0030194-b58]\]. Challenges and Hypotheses {#s7} ========================= One of the major challenges for those interested in the evolution of norms is, at the most elementary level, defining a norm. This is related to another general problem of defining exactly what is changing in cultural evolution---which we might call the "meme dilemma" in honor of Dawkins\' regrettably infertile notion. A second major challenge is discovering the mechanism(s) by which truly novel ideas and behaviors are generated and spread. A third is discovering the most effective ways of changing norms. We\'ve got a long way to go before being able to meet those challenges. One place to start is to begin formulating hypotheses about the evolution of norms that can be tested with historical data, modeling, or even (in some cases) experiments. Some hypotheses we believe worth testing (and some of which may well be rejected) are given in [Box 1](#box1){ref-type="boxed-text"}. Box 1. Sample Hypotheses about the Evolution of Norms ----------------------------------------------------- Hypothesis 1. Evolution of technological norms will generally be more rapid than that of ethical norms. Technological changes are generally tested promptly against environmental conditions---a round wheel wins against a hexagonal one every time, and the advantages of adopting it are clear to all. Ethical systems, on the other hand cannot often be tested against one another, and the standards of success are not only generally undetermined, they often vary from observer to observer and are the subject of ongoing controversy among philosophers. Hypothesis 2. In societies with nonreligious art, the evolution of norms in art will be more rapid than those in religion. We hypothesize that art is less important to the average individual than his or her basic system of relating to the world, and conservatism in the latter would be culturally adaptive (leading to success within a culture). Hypothesis 3. Military norms will change more in defeated nations than victorious ones. Was the Maginot Line and the generally disastrous performance of the French army in 1940 an example of a more general rule? Does success generally breed conservatism? Hypothesis 4. The spread of a norm is not independent of the spread of others, but depends on the spread of other norms (norm clusters). Does, for example, empathy decrease with social stratification? Hypothesis 5. Susceptibility to the spread of norms is negatively correlated with level of education. Are the less educated generally more conformist, or does the spread of norms depend almost entirely on the character of the norm? Hypothesis 6. Horizontal transmission will show less stickiness than vertical transmission. This conjecture is based on anecdotal observations that norms like using hula hoops come and go and are primarily horizontally transmitted, and religious values and other high-viscosity points of view are mostly vertically transmitted (p. 129 of \[[@pbio-0030194-b14]\], \[[@pbio-0030194-b59]\]). In this essay we have tried to be provocative rather than exhaustive. There is a welter of issues we have not even attempted to address, including: (1) asymmetries of power in the spread of norms, (2) the role of networks, (3) the efficacy of persuasion as opposed to imitation, (4) the cause of thresholds in the change of norms, (5) the genesis of norms during child development, (6) the connection between attitudes and actions, (8) competition among norms from different cultures; and (9) the question, can norms exist "free of people" in institutions? Institutions certainly may emerge as independent structures, stabilized by laws and customs that are enforced to varying degrees through formal punishment or social pressure. Can such norms persist long even when adherence to them is disappearing? The interplay between the dynamics of individual behaviors and normative rules, operating on different time (and other) scales, may be the key, we argue, to understanding sudden phase transitions that can transform the cultural landscape. We hope that, by being provocative, we can interest more evolutionists, behavioral biologists, and ecologists in tackling the daunting but crucial problems of cultural evolution. Few issues in science would seem to be more pressing if civilization is to survive. We have received helpful critical comments from Kenneth Arrow, John Bonner, Samuel Bowles, Kai Chan, Gretchen Daily, Partha Dasgupta, Adrian deFroment, Anne Ehrlich, Marcus Feldman, Michelle Girvan, Ann Kinzig, Deborah Prentice, and Will Provine. Amy Bordvik provided invaluable assistance in preparing the manuscript for publication. Citation: Ehrlich PR, Levin SA (2005) The evolution of norms. PLoS Biol 3(6): e194. Paul R. Ehrlich is with the Department of Biological Sciences, Stanford University (Stanford, California, United States of America). Simon A. Levin is with the Department of Ecology and Evolutionary Biology, Princeton University (Princeton, New Jersey, United States of America).	PubMed Central	2024-06-05T03:55:57.538058	2005-6-14	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1149491/", "journal": "PLoS Biol. 2005 Jun 14; 3(6):e194", "authors": [ { "first": "Paul R", "last": "Ehrlich" }, { "first": "Simon A", "last": "Levin" } ] }
PMC1149492	Over the last 20 years, many cell biological studies have moved from the single-cell level to the tissue level, and even to whole animals. This progress has been led by developments in fluorescence microscopy that permit molecular observations from single cells within intact tissue or animals. Concurrent developments in fluorescent probes, especially the cloning of the Green Fluorescent Protein and its use in transgenic animals, have also fueled this movement. The key instrumental technology for this work is optical sectioning microscopy; in this technique, instead of fixing and physically sectioning a sample, the investigator obtains a 3-D dataset from an intact (and more importantly, live) specimen. The most common optical sectioning technique is confocal microscopy, where fluorescence is created throughout the sample and a confocal pinhole is placed in front of the detector so that only the in-focus fluorescence is recorded. For live samples, whose cells can be killed by the excitation light (via photo-toxicity, particularly of ultraviolet and blue wavelengths), confocal microscopy may not be an option. A more recently developed optical sectioning method is two-photon excitation microscopy (which also goes by the names multi-photon microscopy and nonlinear optical microscopy). As described below, two-photon excitation offers very significant advantages for the high-resolution imaging of thick living samples (as deep as 1 mm). Most importantly, two-photon imaging is now ready for prime time because of instrumental advances that have made it as easy to use as any other fluorescence microscopy technique. Fluorescence Excitation {#s2} ======================= To understand two-photon excitation and its advantages for imaging, it is helpful to understand a little bit about fluorescence. Fluorescence is the process of absorption and re-emission of light. Normally, a single light particle (photon) is absorbed by a fluorescent molecule, causing an excited state, which subsequently relaxes by emitting another photon. The excitation light is typically ultraviolet, blue, or green. Any time a photon that has the correct energy to cause the excited state comes in close contact with a fluorescent molecule, it may be absorbed. In contrast, two-photon excitation of fluorescence depends on the simultaneous absorption of two photons (each of which contains half the energy, typically red or infrared, needed to cause the excited state). For this simultaneous absorption to happen, the photons must be so crowded that there is a good chance two photons will simultaneously be at the same place as the fluorescent molecule. In a two-photon excitation microscope, the photons are crowded in both time and space. The photons are crowded in time through the use of short pulses of light, which are about 100 femtoseconds (one tenth of one millionth of one millionth of a second) in duration. This causes about a million times more photons to be present at the same time than would be present in a normal constant wave laser of the type commonly used in confocal microscopes. The photons are crowded in space by focusing through the microscope objective lens. As a single laser beam is focused in the microscope, the photons become more than a million times more crowded still. The combination of short pulses and focusing crowds the photons by a factor of over one trillion. Even with the high powers used, the only place that photons become crowded enough that two of them would be interacting with a single fluorescent molecule at the same time is in a small region at the focus of the microscope. This region, called the focal volume, is the only place that two-photon excitation occurs. This localization of two-photon excitation leads to the advantages for deep-tissue imaging. If standard fluorescence microscopy is like probing the contents of a house by shining a powerful spotlight into the house from outside, two-photon excitation is more like taking a flashlight around the inside of the house; all of the excitation is generated inside the sample. High Resolution, High Contrast {#s3} ============================== As stated above, the major advantage of two-photon excitation is its ability to permit high-resolution and high-contrast imaging from deep within intact living tissue. [Figure 1](#pbio-0030207-g001){ref-type="fig"} shows an intact shark choroid plexus that has been stained with fluorescein in the extracellular space. [Figure 1A](#pbio-0030207-g001){ref-type="fig"} shows a confocal image taken 70 µm into the sample, which exhibits minimal contrast between the bright cell borders and the dark intracellular spaces. [Figure 1B](#pbio-0030207-g001){ref-type="fig"} shows the same section acquired with two-photon excitation; the contrast is much greater. In fact, even with a much deeper section (140 µm into the sample) ([Figure 1C](#pbio-0030207-g001){ref-type="fig"}), two-photon excitation still provides similar contrast. ::: {#pbio-0030207-g001 .fig} Figure 1 ::: {.caption} ###### Images of a Shark Choroid Plexus Stained with Fluorescein \(A) and (B) were collected 70 µm into the sample, and (C) was collected 140 µm into the sample. The contrast of the confocal image (A) is significantly degraded at this depth, while two-photon excitation at the same focal plane (B) allows the collection of an image with excellent intensity contrast. Further, using two-photon excitation to image deeper into the sample (C) does not significantly degrade the image contrast. ::: ![](pbio.0030207.g001) ::: One question that is often asked is, how deep can this approach go? The answer, of course, depends on the specific type of tissue, but a good rule of thumb is that one can image 6-fold deeper with two-photon excitation than with confocal microscopy. There are two reasons for this deeper penetration. The first is that there is no out-of-focus absorption in a two-photon excitation microscope. Because the photons are only crowded enough for two-photon excitation at the microscope focus, they are not absorbed by fluorescent molecules as they pass through the sample. In confocal microscopy, the excitation photons can be absorbed anywhere in the sample. Thus, a higher percentage of excitation photons reach the focus in two-photon excitation, and this advantage grows as the focus moves deeper into the sample. Greater excitation leads to greater signal, and in turn to increased contrast in the image. The second reason for better depth penetration is that two-photon excitation imaging is less sensitive to scattering in the sample ([Figure 2](#pbio-0030207-g002){ref-type="fig"}). This concept has not been well understood, and has been incorrectly reported in many papers. It is often stated that because the red and infrared photons used in two-photon excitation are less scattered by tissue, these photons can penetrate more deeply. While it is true that the photons are scattered slightly less than blue or green photons, this difference is small compared to the differences in sensitivity to scattering between confocal and two-photon excitation microscopy. In confocal microscopy, excitation photons that are scattered in the sample can cause fluorescence anywhere in the sample. As the laser power is increased in an attempt to image deeper into the sample, fluorescence due to scattered excitation also increases. This leads to a background haze in the image that reduces contrast. ::: {#pbio-0030207-g002 .fig} Figure 2 ::: {.caption} ###### Effect of Scattering in Confocal Microscopy and Two-Photon Excitation Microscopy In confocal microscopy (shown on the left), blue excitation light reaches the focus, and green fluorescence from the focus is collected and passes through a pinhole. Scattering of the fluorescence causes it not to pass through the pinhole, thus reducing signal, while any scattering of the excitation beam can cause fluorescence, which adds background haze to the image. In two-photon excitation microscopy (shown on the right), because no pinhole is needed, the scattered fluorescence photons can still be collected, thus increasing the collected signal. Further, the scattering of a single red excitation photon does not cause background (and the chance of two photons scattering to the same place at the same time is zero). ::: ![](pbio.0030207.g002) ::: The emitted fluorescence photons can also be scattered as they come out of the sample. When a fluorescent photon is scattered, it will not pass through the confocal pinhole, and therefore, will not be detected. This lowers the signal, which in turn lowers the image contrast. Thus, both the scattering of excitation light and emitted fluorescence lead to decreased contrast in the confocal image. For two-photon excitation, neither scattering event is deleterious to the image. As for scattering of the excitation photons, there is really no chance of two photons scattering to the same place at the same time, so even in a highly scattering sample, it is possible to increase the excitation power without generating background haze. As for the emitted photons, a two-photon excitation microscope collects most of the scattered fluorescence, since there is no pinhole needed (the only place fluorescence is being generated is in the focal spot). These two reasons, combined, allow two-photon excitation imaging to provide high contrast images from deep within intact tissue, although limitations in available laser power usually limit the depth penetration to less than 1 mm into the tissue. Further details about the advantages of two-photon excitation imaging are presented elsewhere \[[@pbio-0030207-b1]\]. Looking Deeper {#s4} ============== The advantages of two-photon excitation microscopy are truly realized for deep tissue imaging. While the technique can be used to image thinner samples, such as single cells, it will generally not be better than using confocal or deconvolution microscopy. In fact, these other approaches are better suited to such thin samples and also offer better spatial resolution. Further, there may be additional problems associated with two-photon excitation because of the extreme crowding of photons needed. With these high intensities, it is possible to activate other nonlinear processes, which can lead to increased photobleaching and photodamage, possibly negating the advantages of two-photon excitation in thinner samples. As one might expect for such a complicated physical phenomenon, it was some time before two-photon excitation found its way into biological research. In fact, two-photon excitation was first predicted theoretically by Maria Goppert-Mayer in her 1931 PhD thesis at the University of Göttingen (Göttingen, Germany) \[[@pbio-0030207-b4]\], and was experimentally verified in a very early laser experiment by Kaiser and Garrett in 1961 \[[@pbio-0030207-b5]\]. It was not until the invention of powerful, ultrafast lasers that Denk et al. were able to bring two-photon into use for microscopy in 1990 \[[@pbio-0030207-b6]\]. Since that time, there has been considerable interest, and most major research institutions have made some effort to set up a two-photon excitation microscope. Despite the inherent advantages, though, two-photon excitation microscopes are sitting idle in many of these labs. There are a couple of reasons for this. First, the Ti:Sapphire lasers that have been available over the last 15 years are reliable and "hands-free" from a laser-jock perspective, but it has proven difficult for a typical biology lab to keep the lasers in optimal working condition. Second, many investigators did not have projects that were well-suited to the strengths of two-photon excitation microscopy. In these cases, the results were often no better than confocal microscopy, and thus the extra overhead to maintain the Ti:Sapphire laser was not well-justified. These days, neither of these problems applies. The newest available lasers are in a single box, fully hands-off, and computer controlled. This permits any researcher to use two-photon excitation. Further, problems that are well-suited to the application of two-photon excitation have finally found the use of this powerful approach. For example, as demonstrated by two papers in this issue \[[@pbio-0030207-b7],[@pbio-0030207-b8]\], researchers are now able to characterize the activities and motion of individual lymphocytes in intact lymph node \[[@pbio-0030207-b7]\] and thymus \[[@pbio-0030207-b8]\], making direct observations of phenomena that had only been inferred using other approaches. Coupled with the now-mature instrumentation, we should expect two-photon excitation imaging to play a key role in our future understanding of in vivo biological processes. Citation: Piston DW (2005) When two is better than one: Elements of intravital microscopy. PLoS Biol 3(6): e207. David W. Piston is with the Department of Molecular Physiology and Biophysics at the Vanderbilt University Medical Center, Nashville, Tennessee, United States of America. E-mail: <dave.piston@vanderbilt.edu>	PubMed Central	2024-06-05T03:55:57.540705	2005-6-14	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1149492/", "journal": "PLoS Biol. 2005 Jun 14; 3(6):e207", "authors": [ { "first": "David W", "last": "Piston" } ] }
PMC1149493	RNA molecules fulfill a diverse set of biological functions within cells, from the transfer of genetic information from DNA to protein, to enzymatic catalysis. Reflecting this range of roles, simple linear strings of RNA---made up of uracil, guanine, cytosine, and adenine---form a variety of complex three-dimensional structures. Just as proteins form distinct structural motifs such as zinc fingers and beta barrels, certain structures are also commonly adopted by RNA molecules. Among the most prevalent RNA structures is a motif known as the pseudoknot. First recognized in the turnip yellow mosaic virus \[[@pbio-0030213-b1]\], a pseudoknot is an RNA structure that is minimally composed of two helical segments connected by single-stranded regions or loops ([Figure 1](#pbio-0030213-g001){ref-type="fig"}). Although several distinct folding topologies of pseudoknots exist, the best characterized is the H type. In the H-type fold, the bases in the loop of a hairpin form intramolecular pairs with bases outside of the stem ([Figure 1](#pbio-0030213-g001){ref-type="fig"}A and [1](#pbio-0030213-g001){ref-type="fig"}B). This causes the formation of a second stem and loop, resulting in a pseudoknot with two stems and two loops ([Figure 1C](#pbio-0030213-g001){ref-type="fig"}). The two stems are able to stack on top of each other to form a quasi-continuous helix with one continuous and one discontinuous strand. The single-stranded loop regions often interact with the adjacent stems (loop 1--stem 2 or loop 2--stem 1) to form hydrogen bonds and to participate in the overall structure of the molecule. Hence, this relatively simple fold can yield very complex and stable RNA structures. Due to variation of the lengths of the loops and stems, as well as the types of interactions between them, pseudoknots represent a structurally diverse group. It is fitting that they play a variety of diverse roles in biology. These roles include forming the catalytic core of various ribozymes \[[@pbio-0030213-b2],[@pbio-0030213-b3]\], self-splicing introns \[[@pbio-0030213-b4]\], and telomerase \[[@pbio-0030213-b5]\]. Additionally, pseudoknots play critical roles in altering gene expression by inducing ribosomal frameshifting in many viruses \[[@pbio-0030213-b6]\]. ::: {#pbio-0030213-g001 .fig} Figure 1 ::: {.caption} ###### RNA Pseudoknot Architecture \(A) Linear arrangement of base-pairing elements within an H-type RNA pseudoknot. Base pairing is indicated with dashed lines. \(B) Formation of initial hairpin within pseudoknot sequence. Base pairings from loop to bases outside the hairpin are indicated with dashed lines. \(C) Classic H-type pseudoknot fold. \(D) Three-stemmed RNA pseudoknot fold from SARS-CoV. ::: ![](pbio.0030213.g001) ::: Catalytically Active Pseudoknots {#s2} ================================ Hepatitis delta virus (HDV) is a satellite virus of hepatitis B virus. Infection of humans by both HDV and hepatitis B virus is generally more severe than a hepatitis B virus infection alone \[[@pbio-0030213-b10]\]. HDV has a circular genome that is replicated by the host RNA polymerase II through a double-rolling-circle mechanism. This mechanism produces long strands of RNA that must be processed into unit lengths for viral replication. The processing of the viral RNA is achieved by the self-cleaving HDV ribozyme encoded in the RNA \[[@pbio-0030213-b11]\]. The HDV ribozyme folds into a double-pseudoknot conformation and self-cleaves, producing single-genome-length HDV RNAs. The HDV ribozyme is the fastest-known naturally occurring self-cleaving ribozyme, with a cleavage rate greater than one per second, and is active in vitro in the absence of any proteins \[[@pbio-0030213-b12]\]. The HDV ribozyme consists of five helical segments that form two coaxial stacks of two (stems P2 and P3) and three (stems P1, P1.1, and P4) helices each ([Figure 2A](#pbio-0030213-g002){ref-type="fig"}) \[[@pbio-0030213-b3],[@pbio-0030213-b13]\]. Two pseudoknots are formed, each with one helix from each coaxial stack (stems P1 and P2, and stems P3 and P1.1). These two pseudoknots stack on top of each other, forming a nested double-pseudoknot conformation \[[@pbio-0030213-b13]\]. ::: {#pbio-0030213-g002 .fig} Figure 2 ::: {.caption} ###### Sequences and Structures of RNA Pseudoknots Stems and loops are numbered sequentially, unless otherwise noted. Structure coordinates were obtained from the Protein Data Bank (<http://www.rcsb.org>), and structural representations were produced using MOLMOL software. \(A) HDV (1SJ3). Numbering of stems reflects standard nomenclature for HDV. The U1A RNA binding domain is colored gray and is not included in the three-dimensional structure. \(B) Diels-Alder ribozyme (DA-R) (1YLS). \(C) Human telomerase (hTR) (1YMO). \(D) MMTV (1RNK). \(E) Pea enation mosaic virus RNA1 (PEMV-1) (1KPZ). \(F) Simian retrovirus 1 (SRV-1) (1E95). ::: ![](pbio.0030213.g002) ::: The removal of introns from pre--messenger RNA (pre-mRNA) is fundamentally important for eukaryotic life. Most introns are removed by a ribonucleoprotein complex called the spliceosome. A subset of introns are self-cleaving, catalyzing their own removal from pre-mRNA without the aid of proteins \[[@pbio-0030213-b14]\]. One such class of introns are the group I self-splicing introns, with the most well-studied example being from the ciliate Tetrahymena. The structure of this ribozyme is made up of three helical domains, with many tertiary contacts between the domains \[[@pbio-0030213-b15]\]. The only portion of the RNA that spans all three helical domains is a pseudoknot belt that wraps around the molecule, base-pairing with all three helices \[[@pbio-0030213-b15]\]. The pseudoknot establishes the catalytic core of the group I self-splicing introns. Naturally occurring ribozymes appear to perform mainly hydrolysis and transesterification reactions \[[@pbio-0030213-b16]\]; however, in vitro selection has yielded RNAs capable of performing a wide variety of enzymatic reactions \[[@pbio-0030213-b17]\]. Recently the structure of an RNA capable of catalyzing carbon--carbon bond formation by the Diels-Alder reaction was solved ([Figure 2B](#pbio-0030213-g002){ref-type="fig"}) \[[@pbio-0030213-b18]\]. The RNA adopts a λ-shaped fold of its three helices in which stems 2 and 3 stack coaxially, with stem 1 abutting the active site, forming a pocket precisely complementary to the reaction product. The 5′ end of the RNA bridges helical stems 3 and 1, generating a complex nested pseudoknot topology. Although conformationally distinct from the HDV ribozyme \[[@pbio-0030213-b3]\], it is worthwhile to note that they are two of the fastest-known ribozymes, and both utilize a nested pseudoknot architecture \[[@pbio-0030213-b18]\]. Chromosomes possess protective ends known as telomeres to protect themselves from degradation due to successive rounds of DNA synthesis. Telomerase, the ribonucleoprotein complex responsible for the maintenance of the telomere ends \[[@pbio-0030213-b19]\], is upregulated in most cancers \[[@pbio-0030213-b20]\] and might play a role in aging \[[@pbio-0030213-b21]\]. Human telomerase is made up of a 451-nucleotide RNA, a reverse transcriptase, and other proteins \[[@pbio-0030213-b22]\]. At the 5′ end of the RNA is a highly conserved pseudoknot, required for activity, which lies at the core of telomerase. The structure of the human telomerase pseudoknot reveals a classic H-type pseudoknot fold with a slight bend between the stems ([Figure 2C](#pbio-0030213-g002){ref-type="fig"}) \[[@pbio-0030213-b5]\]. A triple-helix structure flanks the junction of the helices and extends into each stem. Mutations within the telomerase pseudoknot have been directly linked to the diseases autosomal dyskeratosis congenita \[[@pbio-0030213-b21]\] and aplastic anemia \[[@pbio-0030213-b23]\]. Frameshift-Inducing Pseudoknots {#s3} =============================== Not all pseudoknots with biological functions are catalytically active. In fact, one of the most common functions of pseudoknots is to induce ribosomes to slip into alternative reading frames, otherwise known as frameshifting. Ribosomes typically translate mRNA without shifting the translational reading frame \[[@pbio-0030213-b24]\]. However, a number of organisms have evolved mechanisms to cause site-specific or programmed frameshifting of the ribosome in either the +1 or −1 direction \[[@pbio-0030213-b25]\]. Programmed −1 ribosomal frameshifting is typically found in viruses and is required for the replication and proliferation of all retroviruses. Therefore, the pseudoknot structures involved in frameshifting are attractive targets for the development of antiviral drugs. The frameshift event is induced by two RNA elements within the mRNA: (i) a heptanucleotide slippery sequence X XXY YYZ (spaced triplets represent preframeshift codons) and (ii) a downstream RNA structure, typically a pseudoknot \[[@pbio-0030213-b26]\]. The mechanism behind how these elements promote −1 frameshifting is not fully understood. The current model posits that the ribosome encounters the downstream pseudoknot while the slippery sequence is being decoded by the ribosome. The pseudoknot structure likely causes the ribosome to pause, which is necessary but not sufficient for frameshifting to occur \[[@pbio-0030213-b27]\]. While paused on the slippery sequence, the ribosome slips back one nucleotide and subsequently continues translation in the −1 reading frame. The nuclear magnetic resonance (NMR) structure of the mouse mammary tumor virus (MMTV) frameshift-inducing pseudoknot was the first structure of a frameshift-inducing pseudoknot ([Figure 2D](#pbio-0030213-g002){ref-type="fig"}) \[[@pbio-0030213-b6]\]. The MMTV pseudoknot forms a compact structure of two guanine/cytosine-rich A-form helices. The MMTV pseudoknot has a bend of approximately 60° between the two helices, caused by an unpaired adenine that intercalates between the helices and may act as a hinge. Subsequent structural and functional studies of several variants of the MMTV pseudoknot reveal that the intercalated nucleotide and the resulting bend between stems 1 and 2 are required for efficient frameshifting \[[@pbio-0030213-b28]\]. In beet western yellow virus, pea enation mosaic virus, and other luteoviruses, an RNA pseudoknot also stimulates a −1 frameshift between the P1 and P2 genes \[[@pbio-0030213-b29]\]. These structures, solved by X-ray crystallography and NMR, respectively, revealed compact H-type pseudoknots with extensive loop--stem interactions ([Figure 2E](#pbio-0030213-g002){ref-type="fig"}) \[[@pbio-0030213-b7],[@pbio-0030213-b9]\]. Like that of MMTV, frameshift-inducing pseudoknots in both the beet western yellow virus and pea enation mosaic virus have an unpaired nucleotide at the junction of the stems; however, this nucleotide is displaced from the helix, not intercalated as in MMTV. The frameshift-inducing pseudoknot from simian retrovirus 1 contains a number of unique features ([Figure 2F](#pbio-0030213-g002){ref-type="fig"}) \[[@pbio-0030213-b8]\]. Although predicted to resemble that of MMTV, with an unpaired adenine between the helices, the structure revealed the formation of a uracil--adenine pair at the junction, allowing the two stems to stack directly on top of each other ([Figure 2F](#pbio-0030213-g002){ref-type="fig"}) \[[@pbio-0030213-b8]\]. The simian retrovirus 1 pseudoknot forms an extensive loop 2--stem 1 triplex, which contains a ribose zipper motif in addition to base--base and base--sugar interactions \[[@pbio-0030213-b8]\]. The severe acute respiratory syndrome coronavirus (SARS-CoV) genome contains two large genes, ORF 1a and ORF 1b, separated by a programmed −1 frameshift element required for ORF 1b expression \[[@pbio-0030213-b30]\]. Recent work has suggested that the SARS-CoV frameshift-inducing pseudoknot may be unique because it contains a third stem--loop \[[@pbio-0030213-b31],[@pbio-0030213-b32]\]. In this issue of PLoS Biology, bioinformatic, phylogenetic, and structural evidence is reported indicating that the SARS-CoV pseudoknot is indeed a three-stemmed RNA pseudoknot (see [Figure 1D](#pbio-0030213-g001){ref-type="fig"}) \[[@pbio-0030213-b33]\]. Dinman and co-workers report the potential for the formation of this three-stemmed pseudoknot in all coronaviruses in the GenBank database. NMR experiments confirmed the proposed three-stemmed pseudoknot structure in SARS-CoV. Although the atomic-resolution structure has not yet been determined, this study identifies a new secondary structure capable of promoting frameshifting that is structurally distinct from previously described pseudoknots (see [Figure 1D](#pbio-0030213-g001){ref-type="fig"}). RNA pseudoknots have been identified in nearly every organism and comprise functional domains within ribozymes, self-splicing introns, ribonucleoprotein complexes, viral genomes, and many other biological systems. It is clear that the pseudoknot topology can result in many different, complex structures. The pseudoknot, therefore, represents an important piece of RNA architecture, as it provides a means for a single RNA strand to fold upon itself to produce a globular structure capable of performing important biological functions. Citation: Staple DW, Butcher SE (2005) Pseudoknots: RNA structures with diverse functions. PLoS Biol 3(6): e213. David W. Staple and Samuel E. Butcher are in the Department of Biochemistry at the University of Wisconsin-Madison, Madison, Wisconsin, United States of America. HDV : hepatitis delta virus MMTV : mouse mammary tumor virus mRNA : messenger RNA NMR : nuclear magnetic resonance SARS-CoV : severe acute respiratory syndrome coronavirus	PubMed Central	2024-06-05T03:55:57.542114	2005-6-14	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1149493/", "journal": "PLoS Biol. 2005 Jun 14; 3(6):e213", "authors": [ { "first": "David W", "last": "Staple" }, { "first": "Samuel E", "last": "Butcher" } ] }
PMC1149494	::: {#pbio-0030214-g001 .fig} ![](pbio.0030214.g001) ::: "Marketeers prove...every scientific term you use represents two thousand readers putting down the magazine and turning on a rerun of I Love Lucy," says a world-weary editor in David Mitchell\'s Russian doll of a novel Cloud Atlas. The marketeers of Ian McEwan (if he has any) clearly don\'t agree. His latest novel, Saturday, is saturated with medical and scientific terms, many of them unexplained. The book is, indeed, a celebration of neurosurgery and neuroscience with almost all the climactic moments in the book hinging around injury to the brain. And reaching beyond neuroscience, McEwan continues the 19th century debate between T. H. Huxley and the Bishop of Oxford over science versus faith, placing himself firmly on the side of Huxley.[](#pbio-0030214-g001){ref-type="fig"} McEwan is widely considered to be Britain\'s pre-eminent novelist. His last book, Atonement, moved him ahead of the pack, and Saturday puts him further out front. Following the model of James Joyce\'s Ulysses, the book tells the story of one day in the life of a man---only without the stylistic fireworks and complexity of Joyce\'s book. McEwan\'s character is not a drunkard, although he loves good wine and good food, but a neurosurgeon, Henry Perowne, and his day is full of incident and drama (which I won\'t reveal). But nobody should fear that the book is a lecture on neurosurgery and neuroscience. It tells a compelling story and says much that is insightful about human relationships---and at the crucial moment it is poetry not science that saves the heroes. Indeed, I read the book before I was asked to review it, and it was only as I read it for the second time that I grasped that every few pages there were scientific references. The science underpins rather than undermines the story. Wisdom is imparted lightly. The brain is a worthy subject for a writer at the height of his powers. In the acknowledgments, McEwan thanks several doctors and scientists for their support and mentions spending two years watching neurosurgeon Neil Kitchen operate. David Lodge, another British writer, tackled consciousness in his book Thinks but produced a novel that did feel too much like a lecture on psychology, albeit an entertaining and understandable one. McEwan emphasises that the mechanisms of the brain---which like an expensive car is intricate but "mass produced nevertheless, with more than six billion in circulation"---are still largely unknown. We don\'t know "how it holds experiences, memories, dreams, and intentions", but Perowne is sure that the brain\'s fundamental secret will be laid open one day. The result will not be man exposed as a machine but rather that such understanding will be magnificent and uplifting. "The actual, not the magical, should be the challenge." This is the cry of science, particularly when the actual is so magical, and surely need not be incompatible with faith. The followers of the Bishop of Oxford take, nonetheless, something of a beating. Perowne, clearly a hero to McEwan (and perhaps too much so), believes that "the primitive thinking of the supernaturally inclined amounts to what his psychiatric colleagues call a problem...an inability to contemplate your own unimportance." People resort to the supernatural through "insufficient imagination". It\'s too easy and comfortable to believe that "an all-knowing supernatural force had allotted people to their stations in life". Indeed, it\'s "a form of anosognosia, a useful psychiatric term for a lack of awareness of one\'s own condition". To insult your opponent is rarely the best policy---but can be hard to resist. Perowne is inevitably for McEwan a materialist and even a determinist. "There is much in human affairs that can be accounted for at the level of the complex molecule." For one of the main characters his "misfortune lies within a single gene, in an excessive repeat of a single sequence---CAG. Here\'s biological determinism in its purest form." (Readers clever enough to know the result of that sequence repeated excessively will know the condition of one character, but many other conditions are described in equal detail. Indeed, a whole operating list is vividly described close to the beginning of the book.) McEwan is scornful of those who would cure genetic diseases. "It is written," he writes with the words in italic and in parody of religious texts. "No amount of love, drugs, Bible classes, or prison sentencing can cure \[the character with the genetic disease\]." Despite the attention to scientific detail, Saturday provides more insights into neurosurgery and neurosurgeons than it does into science. McEwan is fascinated that neurosurgery is essentially plumbing when its object is brilliant circuitry. Who would want a plumber working on their computer? Yet often it turns out well. Perowne does about 300 cases a year. "Some fail, a handful endure with their lights a little fogged, but most thrive, and many return to work in some form; work---the ultimate badge of health." I found myself wondering if the success rates are so good. I don\'t think that they are. My main memory of neurosurgery was a professor angrily pulling bits of skull from the brain of a patient with a severe head injury. The result was not good. McEwan attributes a different view of neurosurgery to psychiatrists: "The neurosurgeons are blundering arrogant fools with blunt instruments, bone-setters set loose on the most complex object in the known universe." (Mitchell in Cloud Atlas, which is more cynical but also funnier, says this: "To us \[surgeons\] people aren\'t sacred beings crafted in the Almighty\'s image, no, people are joints of meat; diseased, leathery meat, yes, but meat ready for the skewer and the spit.") For McEwan his neurosurgeon is something close to a genius. Perowne sees himself as artless and is amazed that he has fathered a published poet and one of England\'s most promising blues musicians. How did this happen? McEwan has Perowne express his idea of genius: "Work that you cannot begin to imagine achieving yourself, that displays a ruthless, nearly inhuman element of self-enclosed perfection." Perowne is thinking of Bach, Beethoven, Mozart, Schubert, Gil Evans, Miles Davis, John Coltrane, Cezanne, and Einstein, but to the reader Perowne fits his own definition of genius. This is not so odd: a friend of mine, Charlie Wilson, a neurosurgeon with particular skills in pituitary surgery, was described a few years ago by a magazine as a physical genius. He was compared with Yo Yo Ma. Perowne likes to operate to Bach---to the Goldberg Variations played not by Glenn Gould (showy and unorthodox) but by Angela Hewitt (wise and silky)---and McEwan interweaves descriptions of a difficult operation and the music. We observe Bach, Hewitt, and Perowne all working together, all geniuses. Whether or not they are geniuses, neurosurgeons---like all surgeons---love to operate. The rest of life---and particularly the outpatient clinic---is an anticlimax. I remember as a junior surgeon (one who was so junior that I did only one---unfortunately disastrous---operation alone) calling in the senior surgeon in the middle of the night to operate on a man with a burst abdominal aneurysm. We operated until dawn, and I never saw that saturnine surgeon so happy---although he didn\'t thank me for pointing it out. For Perowne, called in by his own junior in the middle of a dramatic night, "operating never wearies him---once busy...he experiences a superhuman capacity, more like a craving, for work." The operating theatre is "home from home. Though sometimes things go wrong, he can control outcomes here, he has resources, controlled conditions." His possible genius and his contentment in operating are in contrast to the mess and uncertainty of his non-surgical life, of all our lives. But if neurosurgery is close to genius it is also close to sex. Perowne met his wife when she was admitted with "pituitary apoplexy" (which is well explained in the book), and the text moves quickly from the intimacy of being inside her skull to sexual intimacy (with no suggestion of misconduct). After he finishes a difficult operation, which he did after being awake for nearly 24 hours and having had many adventures, "even his awareness of his own existence has vanished. He\'s been delivered into a pure present, free of the weight of the past or anxieties about the future...It\'s a little like sex, in that he feels himself in another medium...." This wonderful book might dramatically increase recruitment into neurosurgery. Citation: Smith R (2005) A hymn to neurosurgery and neuroscience. PLoS Biol 3(6): e214. Richard Smith is at UnitedHealth Europe, London, United Kingdom. E-mail: <richardswsmith@yahoo.co.uk>	PubMed Central	2024-06-05T03:55:57.543419	2005-6-14	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1149494/", "journal": "PLoS Biol. 2005 Jun 14; 3(6):e214", "authors": [ { "first": "Richard", "last": "Smith" } ] }
PMC1149495	The scientific enterprise reflects the willingness of scientists to collaborate implicitly with people they don\'t know personally: investigators routinely obtain antibodies, enzymes, sterile supplies, and other items from commercial suppliers without hardly blinking, and often without knowing anything about the companies other than what appears on the sales brochures. This system works because of basic warranties and expectations, as well as competition among companies to maintain high quality standards and easy, quick availability of specialized reagents. It is no exaggeration to say that vendors, as de facto scientific collaborators, are a major driving force in scientific productivity. In contrast, many scientists are far more reluctant to enter into explicit collaborations with other academic scientists, even those who are well respected and well established, unless they have a strong prior personal relationship with them. Why? There are many concerns that a scientist might have that could outweigh the potential advantages of collaboration with another academic. Unlike a commercial vendor, whose participation is passive and circumscribed, an academic collaborator is likely to argue passionately about the design of the experiments, may not agree with the underlying hypothesis, might engage in experiments that lack certain perceived controls or quality standards, or may simply lose momentum (e.g., if the lab loses grant support, or if a student in the lab leaves who was doing the experiments). Even after the basic experiments are finished, an academic collaborator might delay or even prevent publication by insisting on the need for further experiments, proper credit, a certain author name order, having intellectual property rights, submitting to a certain journal, and so on. Doing scientific research is a highly personal and subjective activity, so that it is difficult to put two investigators together at random and expect them to see eye-to-eye. Also, each member of the collaboration is privy to valuable unpublished data, facts, ideas, and hypotheses that they are loathe to disclose to the other without good reason. Making a commitment to collaborate fully with another laboratory is far from a trivial decision. Thus, collaborations tend to fall into one of two categories at opposite ends of the spectrum: on the one hand, the passive or one-sided vendor model, where a person supplies a reagent with minimal warranties and expectations; and on the other hand, the active collaborator model, where two or more investigators are fully engaged in a common pursuit with full sharing of ideas and credit (think Watson and Crick). This leaves an enormous set of potential opportunities in the middle, consisting of limited collaborations that could be mutually fruitful, but that often cannot get started or be sustained because of uncertainty and lack of trust. We suggest that it is possible to encourage more of these limited collaborations to proceed by providing a list of the key points to be considered at the onset of a collaboration. What we have in mind is not along the lines of a legal contract; rather, we envision a set of voluntary "boilerplate" guidelines that two academic biomedical investigators, who are potential collaborators, can refer to as a reference point for negotiation. This would not threaten the current informal way that scientists tend to enter collaborations. Nor would the use of these guidelines incur any obligations on the universities or other institutions that employ the scientists involved. Although minimal guidelines for sharing of reagents and data have been widely discussed (e.g., \[[@pbio-0030217-b1]\]), to our knowledge, no one has explicitly enumerated the points of collaboration that should be negotiated between two academic biomedical investigators. Here, we propose a set of possible guidelines ([Box 1](#box1){ref-type="boxed-text"}). We confine the terms to collaborations between a supplier (of reagents, expertise, specialized equipment, data, methods, or computer code) and a receiver, who is often the initiator of the collaboration. However, the guidelines could apply to collaborations generally, regardless of the nature or direction of sharing involved, and regardless of whether the collaboration is formalized (e.g., as in a contract for services) or is pursued as an informal verbal agreement. As negotiations of intellectual property issues generally involve additional parties such as host institutions and funding agencies and are not solely under the control of the two collaborators, a detailed consideration of such issues falls beyond the scope of the present guidelines. Box 1. Suggested Guidelines for Negotiating Scientific Collaboration ==================================================================== The guidelines here are discussed in terms of a collaboration involving a supplier (e.g., of reagents, expertise, specialized equipment, data, methods, or computer code) (Party 1) and a receiver (Party 2). The options in each section are listed in order of increasing involvement by Party 1 in the collaboration. Sharing of reagents and data.Minimal: Reagents or data will be provided that are essentially as described in published description, and Party 1 warrants that these will be prepared, stored, and shipped in a manner that will preserve their value. Party 2 will pay the costs of shipping. If the reagent is not readily available in the supplier\'s laboratory and there are no plans to prepare more, Party 2 may be asked to share reasonable costs of preparing the reagent for shipping.Option 1: Party 1 will give all available information, based on unpublished results, to help the receiver save time and use the reagents in an optimal fashion. (For example, when supplying an antibody for immunohistochemistry, the supplier will suggest a proven range of dilutions, buffers, and fixatives.)Option 2: Both parties will describe the nature of unpublished information related to ongoing experiments in their respective laboratories, in sufficient detail that the parties can decide whether it is mutually to their benefit to share unpublished information, in whole or in part.Design of experiments.Minimal: Party 1 will not be privy to details of the intended use of the reagents or data, beyond those necessary to prepare the reagents adequately.Option 1: Party 2 will describe the design of their experiments and invite comments from Party 1. Any suggestions that are implemented will be considered part of an active collaboration and credited appropriately.Option 2: Both parties will contribute actively to the design of experiments, including control experiments.Division of labor.Minimal: Party 1 provides reagents/expertise/equipment/data but does not participate in the experiments of Party 2.Option 1: Party 1 provides training and specialized expertise to personnel of Party 2, thus facilitating the work, but does not carry out the research directly.Option 2: Both parties participate in experiments, but each does the subset of experiments they are most experienced, knowledgeable, or comfortable with and the other group(s) do the same with nonoverlapping subsets of experiments.Option 3: Party 1 provides assistance with personnel and training to help with Party 2\'s experiments, forming an interlaboratory team that has ongoing communication during experiments and that shares the costs of doing the research.Publication of results stemming from the collaboration.Minimal: Party 2 will acknowledge the source of the reagent or data in their publications, and will cite the providers\' relevant publications that described the reagent and its use.Option 1: Party 1 will be given a copy of the manuscript prior to publication and given the option to be included in the list of authors, unless Party 1 disagrees materially with the paper, or fails to answer in a reasonable time frame.Option 2: Both parties participate in writing the paper, and the resulting publication spells out the contributions of each.Co-authorship order.Minimal: Party 2 will write the paper and choose the order of authorship that Party 2 feels is fair and appropriate.Option 1: At the time that a decision is made to offer co-authorship to Party 1, Party 2 will discuss the planned order of authors and the rationale with Party 1.Option 2: As the experiments are being designed and planned, both parties will be apprised of the relative roles of individuals in both laboratories, and a tentative co-author list will be discussed. Any changes will be discussed in advance of writing the paper.Access to unpublished data arising from the collaboration.Minimal: Party 1 has no right to access to any data obtained by Party 2.Option 1: Party 2 will give basic feedback on how the material supplied by Party 1 was used.Option 2: Party 2 shares the specific results obtained with Party 1, with the understanding that such information is confidential and cannot be used by Party 1 without the written consent of Party 2.Option 3: Party 1 has access to data originally intended for publication under joint authorship, whether or not the data are actually published. The unpublished data can be used by Party 1 in their grant submissions and for their own knowledge.Option 4: Both parties will discuss the types of experiments that are being conducted in the general area of the collaboration, and will discuss which activities (beyond the joint experiments) should be shared knowledge.Intellectual property issues. Negotiations of intellectual property issues fall beyond the scope of the present guidelines. However, potential collaborators should at least acknowledge the types of intellectual property issues that may arise in the course of the collaboration.Minimal: Party 2 does not share in any intellectual property related to the existing reagent or data.Optional: If the joint experiments or data analyses result in new information or uses having commercial value, both parties will negotiate shared intellectual property. These guidelines will be posted on the Science Commons Web site (<http://science.creativecommons.org/>) to provide a public forum encouraging scientific collaborators and others to provide feedback and suggestions, and allowing the guidelines to be modified and extended over time. This Human Brain Project/Neuroinformatics research is funded jointly by the National Library of Medicine and the National Institute of Mental Health (grant LM07292). We thank the many colleagues who gave comments and suggestions, including Carol Bickford, Larry Ozeran, Eike Kluge, Charles Safran, Blaise Cronin, and John Milbanks. This paper reflects the views of the Working Group on Ethics, Legal and Social Issues of the American Medical Informatics Association. Citation: Smalheiser NR, Perkins GA, Jones S (2005) Guidelines for negotiating scientific collaboration. PLoS Biol 3(6): e217. Neil R. Smalheiser is in the Department of Psychiatry, University of Illinois, Chicago, Illinois, United States of America. Guy A. Perkins is in the Department of Neurosciences, University of California, San Diego, California, United States of America. Steve Jones is in the Department of Communication, University of Illinois, Chicago, Illinois, United States of America.	PubMed Central	2024-06-05T03:55:57.544714	2005-6-14	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1149495/", "journal": "PLoS Biol. 2005 Jun 14; 3(6):e217", "authors": [ { "first": "Neil R", "last": "Smalheiser" }, { "first": "Guy A", "last": "Perkins" }, { "first": "Steve", "last": "Jones" } ] }
PMC1149496	Just a handful of species are the basis for a staggering amount of our biological knowledge. From the ever-popular mouse ( Mus musculus ) to the always fruitful fruit fly ( Drosophila melanogaster ), biologists have cultivated a cadre of model organisms to unravel the intricate mysteries of cell communication, genetics, and embryonic development. "Model system," however, is a tricky term to define. The biological superstars are seven genetic organisms---yeast ( Saccharomyces cerevisiae ), Escherichia coli, fruit fly, roundworm ( Caenorhabditis elegans ), mustard plant ( Arabidopsis thaliana ), zebrafish ( Danio rerio ), and mouse---for which "model organism" has become shorthand. Although a second tier of organisms can be cast as models when they facilitate study of a particular biological process, they are often merely supporting actors relegated to a bit part on the biological stage. For all the laboratory tales these seven model species have helped to tell, there remains a wealth of evolutionary and ecological questions still to be addressed. Understanding organisms\' responses to mutations and the environment in order to paint a more complete story of biological networks is the biggest challenge in biology today \[[@pbio-0030219-b1]\]. In addition to providing insight into ecology and evolutionary lineages, studies of nonmodel organisms are sure to reveal as-yet unknown biological mechanisms. Recently, the plot thickened when genomic data revealed the amazing degree to which genes are conserved throughout species. Surprisingly, it\'s not simply the genes, but their regulation, that gives rise to the remarkable diversity of creatures. New molecular techniques take advantage of these findings to reveal gene expression and function in an expanded cast of characters. Functional genomics makes nonmodel organism studies more robust---blurring the line between model and nonmodel species and setting the stage for synergistic discoveries in evolutionary biology and ecology. A Star Is Born {#s2} ============== Previously held assumptions are already being reconsidered in the face of growing amounts of genomic data. The sea anemone may seem an unlikely character to shake the phylogenic tree, but new genomic evidence suggests that the sea anemone is more closely related to the greater majority of more complex animals than its position above the lowly sponge had once presumed \[[@pbio-0030219-b2]\]. "Our ideas of evolution---who gave rise to what---are changing rapidly as we get more data from more animals," says Linda Holland, evolutionary biologist at University of California at San Diego. Holland is one of the few researchers to study amphioxus, a small, translucent, fish-like animal that is the closest living invertebrate relative of the vertebrates ([Figure 1](#pbio-0030219-g001){ref-type="fig"}). She likens the use of such simple species to studying architecture. "You want to determine the architecture of the simple church before examining the gargoyles that vertebrates erect on the system," she says. Its strategic position on the phylogenetic tree is enough to provide valuable insights---worthy of the years of work to perfect techniques already standard in the superstar model species. "Its genome is as close as you can get to the ancestral vertebrate genome," she says. ::: {#pbio-0030219-g001 .fig} Figure 1 ::: {.caption} ###### Amphioxus, a Small, Translucent, Fish-Like Animal That Is the Closest Living Invertebrate Relative of the Vertebrates Its strategic position on the phylogenetic tree is enough to provide valuable insights---worthy of the years of work to perfect techniques already standard in the superstar model species. (Image: Giovanni Maki) ::: ![](pbio.0030219.g001) ::: Amphioxus is a perfect example of grooming a starlet model organism using today\'s techniques. Holland started out cloning genes and looking for gene expression. Determining when a gene is expressed is a first step toward understanding its function. When she and colleagues documented expression of an amphioxus counterpart (a "homolog") of a crucial vertebrate development gene, called Hox3, they began to assemble genomic tools to exploit this organism\'s potential as a model for vertebrate development. Although only a few organisms have a sequenced genome to add to their reagent list, Douglas Crawford, evolutionary biologist at the University of Miami, argues that genomic tools, such as a cDNA library---cloned complementary DNA molecules synthesized from expressed RNA---are affordable enough to be accessible to any research group that has the considerable time it takes to sequence and, more importantly, annotate genes. With the library comes the ability to create microarrays with which to study gene expression. Microarrays are sets of DNA sequences affixed to glass slides that enable researchers to determine which of the thousands of genes in an organ, organism, or population are expressed at any given time. The microarray is a revolutionary tool to cast widely for clues to gene function. "Thus far, gene functions could be studied only with the help of those model systems allowing genetic analysis," says Daniel Chourrout, director of the Sars International Centre for Marine Molecular Biology, University of Bergen, Norway. However, many interesting behavioral, physiological or ecological traits and responses are poorly expressed or absent in the genetic model organisms. Those behavioral genes that are known to model species instantly become candidate genes to search for in nonmodel organisms. Comparative genomics has changed the playing field for understanding mechanisms and evolution of behavior---genes being the common currency. For example, once the social honeybee\'s genome sequence is complete, its gene-behavior relationships can be compared to the solitary fruit fly. "The biggest problem with model organisms is that they are inbred strains," Crawford says. Restricting an evolutionary view to lab strains could miss important signals, such as epistasis, control of a phenotype by two or more genes. The microarray enables researchers to probe those organisms most suited to answer a specific biological question \[[@pbio-0030219-b3]\]. Crawford developed microarrays to assess the fitness of Fundulus heteroclitus, a fish species that lives along a steep thermal gradient in the Atlantic Ocean---from the cold north to the warmer south. Highlighting the functional importance of DNA polymorphisms, he has been able to show a biologically relevant difference in the expression of metabolic genes among individuals in a population. Other organisms are similarly poised to answer ecological questions. Researchers studying the freshwater crustacean Daphnia ( Daphnia pulex )---a toxicologically sensitive organism that plays a key role in freshwater ecology---are undertaking similar strategies to become ingénue genetic and genomic systems ([Box 1](#box1){ref-type="boxed-text"}). Box 1. Today\'s Model Toolkit ----------------------------- Using Daphnia ([Figure 2](#pbio-0030219-g002){ref-type="fig"})---a species showing adaptive traits that keep re-emerging to cope with a variety of ecological conditions---researchers want to detail what ecology can tell us about genomics. Indeed, the Daphnia community is actively assembling the elements to achieve status as a model genetic system. "We\'re trying to not necessarily build tools for what is considered at this point a nonmodel---we\'re trying to create a model system," says John Colbourne, founding member of the Daphnia Genomics Consortium ([Figure 3](#pbio-0030219-g003){ref-type="fig"}). To do so, they must assemble the genetic tools that are the defining feature of traditional model genetic species as well as functional genomics approaches. The first act was to create a consortium of the relevant researchers. Those researchers are now assembling cDNA libraries to make microarrays, finding ways to transform the organism using knockout techniques, developing cell lines assembling genetic and physical maps, and creating a customized electronic database. To complete the kit---the genome is on its way. Such a toolkit for a "model" nonmodel species is getting easier to assemble, particularly given the ease of creating a cDNA library. But until a genome is in hand, the easiest way to achieve status can always benefit from a robust classical biological history, a vocal or influential research community, or a critically understudied position on the phylogenetic tree. Phenotype Casting {#s3} ================= Until systems such as Daphnia take hold, there are two ways to get the most information from ecologically relevant populations---take the model genetic system to a natural setting or apply new techniques to nonmodels \[[@pbio-0030219-b4]\]. While some inventive researchers are finding interesting results with the former method ([Box 2](#box2){ref-type="boxed-text"}), ecologists are increasingly interested in the genes that underlie native organism fitness in order to shed light on environmental modifications to gene expression \[[@pbio-0030219-b5]\]. Box 2. Reality Ecology ---------------------- A typical ecologist is not usually inclined to study mice. But Wayne Potts is not typical. He has designed a phenotron ([Figure 4](#pbio-0030219-g004){ref-type="fig"}), a man-made enclosure replete with three-dimensional complexities, otherwise known as a mouse barn. Using wild mice, he studies the effects of stressful situations found in the real-world social ecology of these animals. The findings are startling. He\'s been able to show that the impacts of inbreeding are far greater than previous studies detected using lab assays. Taking offspring from one generation of full sib inbreeding and then allowing them to compete against outbred controls in the phenotron, Potts found an additional five-fold reduction in male fitness. "If you mate with your sister, your sons are effectively dead," he says. Extrapolating from this finding, he cautions against assuming that gene knockouts with little or no phenotypic effect means there is negligible impact on the organism. "If you care about gene function, you\'ve got to test them under the competitive conditions in which the genes evolved," he says. "Ecology is about phenotypes, and our goal is to understand the phenotypic variation that matters in a natural context," says Thomas Mitchell-Olds, plant ecologist at the Max Planck Institute for Chemical Ecology in Jena, Germany. Mitchell-Olds combines both approaches---sticking to wild relatives of the best genetic plant model Arabidopsis in order to take advantage of the already established experimental methods that allow him to focus on hypotheses in undisturbed environments. Studying wild relatives has one additional advantage: candidate genes responsible for ecological variation, such as resistance to insects and pathogens, drought tolerance, and flowering, have already been identified in Arabidopsis. Mitchell-Olds can clone these genes in the wild relatives he studies to see if they have the same function. In addition to exploring natural phenomena, ecologists are using microarrays to determine the gene-expression changes related to exposure to existing and emerging contaminants, including pharmaceutical compounds, pesticides, and nitrogen inputs from agriculture---a "canary on a chip" capable of assessing environmental impacts on an organism\'s reproduction and fitness \[[@pbio-0030219-b6]\]. "These techniques provide information about genetic mechanisms pertaining to physiology and behavior of organisms and how environment influences phenotype, either as a result of natural variables or toxicology," says Rebecca Klaper, ecologist at the Great Lakes WATER Institute, at the University of Wisconsin-Milwaukee. Klaper studies the lake sturgeon ( Acipenser fulvescens ), a complex yet tragic character that has an estimated 250 chromosomes, some of which are very small. For long-lived, threatened, or endangered species, such as the lake sturgeon, comparing cDNA libraries from different tissues and time points to known databases allows them to identify differentially regulated genes depending on reproductive stage or exposure to toxins. Ultimately, they will combine these techniques with home-grown microarrays for this species that could never be raised in the lab. They hope to better understand how evolutionarily ancient sturgeon are affected by toxins, how their immune system functions, how sexual development and reproductive stage are determined, and what mechanisms are responsible for cueing this development. Evo-Devo-Tees {#s4} ============= Linking phenotype to genotype, or---broadly---form to function is the core of evolutionary biology. Gene function is crucial to understand how evolution developed new body forms. "To really understand the evolution of development, you have to sample pretty broadly," says Nipam Patel, evolutionary biologist at the University of California at Berkeley. "Between mice and flies, you see conservation of genes, but that doesn\'t tell how evolution changes body plans and morphologies," he adds. "Expression studies tell you about conservation of expression but nothing on the conservation of function," says Gregor Bucher, developmental biologist at Göttingen University in Germany. Indeed, verification of function is the key step, often accomplished via transgenesis---incorporating an introduced gene that can be transmitted to successive generations---which is not an option for most species. In lieu of transgenesis, evo-devo-tees increasingly favor reverse genetics---knocking out a specific known gene to look for a change in phenotype---as opposed to the more robust, traditional method of forward genetics, which relied on induced mutations. Two techniques, RNA interference (RNAi) and oligonucleotide morpholinos, have been used successfully to effectively knock out specific genes in nonmodel systems. RNAi degrades RNA, causing reduced expression, while morpholinos block translation of proteins. In addition to being ridiculously easy to deliver in some species, RNAi has one added benefit: injecting pregnant mothers of some species creates knock-down embryos. Indeed, knock-down embryos work well for the red flour beetle ( Tribolium castaneum ), which is a more representative species than the fruit fly to research arthropod head development and segmentation. The fruit fly\'s head forms all at once instead of in the anterior-to-posterior progression usual to most arthropods. Like many other evo-devo researchers, Thomas Kaufman, evolutionary biologist at Indiana University and fruit fly devotee, is now exploring nonmodel species such as the milkweed bug ( Oncopeltus fasciatus ). He used RNAi to show that genes controlling mandible mouthparts in the fruit fly produce specialized piercing-sucking mouthparts in the milkweed bug. Such seemingly subtle differences represent regulatory paradigms differentiating evolution between orders of insects. Unfortunately, RNAi doesn\'t work in every species, or even every gene. And "one has to be careful about interpretations of phenotype," Crawford says. Often, oligonucleotide morpholinos can serve as a stand-in for the popular RNAi. Once Holland exhausted the utility of gene-expression patterns to infer homologies of structures in amphioxus, her research group moved from gene-expression patterns to mechanistic studies using oligonucleotide morpholinos to knock down gene function. It took about five years to work out the techniques, particularly since amphioxus eggs are currently available only about 15 nights out of the year. But the amphioxus genome has only single copies of most development genes, providing a straightforward route to interpret functional knock-downs. Using morpholinos, they are starting to put together an account of developmental patterning that may serve as a model for vertebrate systems. For all the excitement surrounding these new techniques, good old-fashioned forward genetics has allowed the three-spined stickleback ( Gasterosteus aculeatus ) to quickly achieve supermodel status in recent years by detailing how complex traits evolve in vertebrates. David Kingsley, evolutionary biologist at Stanford University, and colleagues generated a genome-wide linkage map by crossing two different species. The resulting data have detailed that a single gene, rather than small changes in many genes, can have a major impact on features such as the armor of these isolated lake fish---altering the course of evolution \[[@pbio-0030219-b7]\]. Using the map, they can now identify the genes controlling variable morphologies and behavioral ecology related to reproduction and mate choice. Given the success, the stickleback ensured that its genome would be sequenced, which will be completed later this year. To really take a biological system down to the deepest mechanistic levels, Kingsley believes that researchers need all the types of methods that are routinely used in the most successful model organisms. "In the long run, the systems we are going to understand the best are the ones where you have not only arrays and RNAi, but also methods for crossing animals, mapping traits, cloning traits, doing sophisticated embryology, decreasing and increasing the function of particular pathways, and transferring specific genomic changes from one population into another," he adds. Encore {#s5} ====== All the world\'s a stage---especially for biologists. Until now, the few genetic superstar model systems delivered the bevy of biological information applicable to the cast of thousands. Scientists now have the tools to determine the roles played by some of the unique and interesting supporting characters. Functional genomics has added a plot twist, as well as an element of suspense for ecological and evolutionary discovery. Mitchell-Olds foresees rapid results from functional genomics approaches. "In the next five to ten years, I think it will be feasible to identify genes controlling ecological important variation, and understand their functional effects in field, ecological consequences, and the historical and evolutionary forces that have influenced genetic variation for ecologically-important traits," he says. Can nonmodel species replace the genetic model species? It\'s doubtful \[[@pbio-0030219-b8]\]. "One should not underestimate the critical mass effect, which gives classical model systems a permanent advantage," Chourrout notes, adding that forward genetic approaches used in model genetic organisms are a more efficient way to reveal unsuspected mechanisms. Couple that with the wealth of knowledge and large research communities, and it\'s easy to see that the genetic organisms will continue in biology\'s starring roles. But the new cast of characters will be able to tell a richer story. ::: {#pbio-0030219-g002 .fig} Figure 2 ::: {.caption} ###### Daphnia pulex, a Species Waiting in the Wings to Achieve "Model" Status (Photo: Paul Hebert) ::: ![](pbio.0030219.g002) ::: ::: {#pbio-0030219-g003 .fig} Figure 3 ::: {.caption} ###### The Daphnia Genomics Consortium Logo (Design: S. Lourido) ::: ![](pbio.0030219.g003) ::: ::: {#pbio-0030219-g004 .fig} Figure 4 ::: {.caption} ###### Setting the Stage in Ecology---The Phenotron (Image: Wayne Potts) ::: ![](pbio.0030219.g004) ::: Citation: Gewin V (2005) Functional genomics thickens the biological plot. PLoS Biol 3(6): e219. Virginia Gewin is a freelance writer based in Portland, Oregon, United States of America. E-mail: <gewin@nasw.org> RNAi : RNA interference	PubMed Central	2024-06-05T03:55:57.545920	2005-6-14	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1149496/", "journal": "PLoS Biol. 2005 Jun 14; 3(6):e219", "authors": [ { "first": "Virginia", "last": "Gewin" } ] }
PMC1149497	In PLoS Biology, volume 3, issue 4, DOI: [10.1371/journal.pbio.0030127](10.1371/journal.pbio.0030127) The figure credits for Figures 1 and 2 are missing from the PDF versions of the article. For Figure 1, images are courtesy of Langdon Quetin and Robin Ross, researchers at the Marine Science Institute, University of California, Santa Barbara, and funded by the Office of Polar Programs, National Science Foundation. For Figure 2, images are courtesy of Donna Fraser. This correction note may be found online at DOI: 10.1371/journal.pbio. 10.1371/journal.pbio.0030224. Published May 17, 2005 Citation: (2005) Correction: As the Antarctic ice pack recedes, a fragile ecosystem hangs in the balance. PLoS Biol 3(6): e224.	PubMed Central	2024-06-05T03:55:57.548322	2005-6-14	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1149497/", "journal": "PLoS Biol. 2005 Jun 14; 3(6):e224", "authors": [ { "first": "Liza", "last": "Gross" } ] }
PMC1149498	Almost exactly five years ago, in early June 2000, BMC Bioinformatics received its first submission. Five years on, it has received over a thousand submissions, and the journal is continuing to grow rapidly (Figure [1](#F1){ref-type="fig"}). In the past few months, developments have included a refreshed international [editorial board](http://www.biomedcentral.com/bmcbioinformatics/edboard), which now consists of over 50 leaders in the field, and a [Bioinformatics and Genomics gateway](http://www.biomedcentral.com/gateways/bioinformaticsgenomics/) that brings together relevant content from across BioMed Central\'s 130+ Open Access journals. And by the time you read this, BMC Bioinformatics should have its first official ISI Impact Factor. Impact factors certainly have their problems -- a previous editorial in this journal\[[@B1]\] discussed the arbitrariness of the process by which ISI selects journals for tracking, and the resulting unnecessary time delay before Impact Factors become available. One thing is clear though -- with BMC Bioinformatics having an Impact Factor, there are more reasons than ever to make it the first choice for your research. Five years in bioinformatics ============================ Looking back over the first 5 years of the journal, are any significant trends evident? One thing that is noticeable is the prevalence of the open-source model of software development. In fact more than 10% of all BMC Bioinformatics articles include the term \"open-source\". Hundreds of open-source bioinformatics projects are now hosted on sites such as [bioinformatics.org](http://bioinformatics.org) and [sourceforge.net](http://sourceforge.net/softwaremap/trove_list.php?form_cat=252). No doubt the similar philosophies of open-source software and Open Access publishing have been a factor in making BMC Bioinformatics one of BioMed Central\'s most successful journals. Two other emerging trends are, firstly, an increasing use of web service technology to connect disparate tools into analysis pipelines, and secondly, the development of systems to allow biological knowledge to be modelled and expressed in structured form. The linking factor between both these trends is that increasingly, as the data deluge continues, the \'users\' of bioinformatics tools and the \'readers\' of the biological literature, are likely to be computer systems rather than human beings. Web services and data analysis pipelines ======================================== As bioinformatics tools have proliferated, the complexity of data analysis has increased. Often, a sequence of analysis steps each using different tools must be carried out one after the other. This might be done manually or by using a monolithic system that is capable of carrying out multiple analyses, or, more flexibly, by writing special \'glue code\', often in Perl, to connect together multiple tools into a pipeline. The problem with the latter approach, though, is that in the absence of defined standards for the input and output of different tools, lots of glue code has to be written in order to create each new pipeline. Worse, systems built in this way tend to be fragile, since at any time one of the tools in the pipeline may change the format of its input or output (breaking the system), because there is no explicit \'contract\' between the various tools as to what input and output formats each will support. Web services \[[@B2]\], and more generally, \'Service Oriented Architectures\' \[[@B3]\] promise to provide a solution by providing a means for codifying standard interfaces that can be used to expose bioinformatics tools over the web. Projects such as MyGrid \[[@B4]\] have then built on these standards to provide biologists with graphical user interfaces that can be used to build new analysis pipelines interactively, without needing to write code. BMC Bioinformatics has published several articles on the use of Web Service technologies such as the Simple Object Access Protocol (SOAP) - if you are interested, try searching the journal for: [SOAP OR \"web services\"](http://www.biomedcentral.com/search/results.asp?drpField1=&txtSearch1=SOAP+OR+%22web+services%22&drpPhrase1=and&drpField2=%5BTI%5D&txtSearch2=&drpPhrase2=and&drpField3=%5BAU%5D&txtSearch3=&drpPhrase3=and&drpField4=%5BTIAB%5D&txtSearch4=&drpPhrase4) Text mining and biological semantics ==================================== Another growth area in bioinformatics has been the structured representation and modelling of biological knowledge. The Gene Ontology project \[[@B5]\] has provided an important foundation for much of this work, defining a set of controlled vocabularies that allow biological concepts and relationships to be expressed in a standard way. Much of the initial work on modelling biological knowledge has explored the use of text-mining techniques to automatically derive structured semantic information from the relatively unstructured text of scientific research articles. [BioMed Central\'s Open Access corpus](http://www.biomedcentral.com/info/about/datamining)\[[@B6]\] is now rapidly approaching 10,000 articles and provides ideal raw material for such research.. It is already being used by many researchers, both in industry and academia. BMC Bioinformatics publishes many papers on text-mining topics, including the recently published supplement \[[@B7]\], which consists of papers presented at last year\'s BioCreAtIvE text-mining workshop in Granada, Spain. Text mining has its limits, however. Imagine what could be achieved if articles, rather than consisting entirely of free-form natural language, contained explicit assertions about biological knowledge in unambiguous, machine-readable form. This is the oft-vaunted promise of the 'Semantic Web' \[[@B8]\], but it has proved to be very difficult to realize in practice. Some recent developments, however, suggest that progress is being made. For example, this editorial was created using [Publicon](http://www.biomedcentral.com/info/ifora/publicon)\[[@B9]\]- a new breed of scientific authoring tool developed by Wolfram Research with input from BioMed Central. Publicon is easy to use, but it is also a highly structured authoring environment. It can not only output BioMed Central\'s native article XML format, but also embed mathematical equations as \'islands\' of semantically-rich MathML \[[@B10]\]. This structured mathematical information is then preserved throughout the publication process, from the author\'s computer right through to the reader\'s desktop with no intermediate unstructured version along the way that might cause information to be lost. So, for example, if you are accessing this editorial online using a suitable browser, you should be able to cut and paste the equation below into any MathML-aware application, as a mathematically meaningful equation rather than an image. ( i ∇ \- m ) Φ e 2 \[ B , x \] = B ⁡ ( x ) Φ e 2 \[ B , x \] \+ i e 2 γ μ ∫ δ \+ ( s x 1 2 ) ( δΦ e 2 \[ B , x \] / δ B μ ( 1 ) ) ⅆ τ 1 In two accompanying Commentaries, the issues associated with capturing and representing biological knowledge are discussed further. Murray-Rust et al.\[[@B11]\] consider how chemical information can best be represented within scientific articles, and what bioinformaticists and chemists can learn from one another. Meanwhile, Mons \[[@B12]\] explores in more detail how smart authoring tools can enrich the scientific literature by allowing authors to express themselves unambiguously, avoiding the \'data burying\' that makes text mining necessary in the first place. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Number of submissions to BMC Bioinformatics. The figure for 2005 represents a conservative projection based on the rate of growth of submissions during the first half of the year. ::: ![](1471-2105-6-140-1) :::	PubMed Central	2024-06-05T03:55:57.548601	2005-6-7	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1149498/", "journal": "BMC Bioinformatics. 2005 Jun 7; 6:140", "authors": [ { "first": "Matthew J", "last": "Cockerill" } ] }
PMC1149499	Introduction ============ In \"Representation and Use of Chemistry in the Global Electronic Age" \[[@B1]\] we showed that new technology can provide great increases in the quantity and quality of aggregated chemical information published in the primary literature. We also argued the benefits of Open Access and Open Data. The current invited overview and a parallel technical article extends the same methodology to chemistry in bioinformatics to remove the loss and corruption of data that occurs in current publishing. We are pleased that this article is an Open Access publication, and we expect that bioinformatics, with its culture of Open Data, is more likely than mainstream chemistry to adopt new approaches. The benefits of open access include higher quality, greater availability, and development of the Biochemical Semantic Web where robots mine text and data as a basis for knowledge-driven science. We argue that funders, institutions, authors, editors, publishers and readers will all benefit. Biosciences now require large amounts of detailed chemical information, examples of which include the occurrence and role of small-molecules in biological processes; the mechanism of biochemical reactions and interactions; the structure and properties of biomolecules; reagents, protocols and classificatory tools for performing bioscience; chemistry in the ecosphere. Such information is only available in a dispersed manner in the primary literature and current mechanisms for its collection and dissemination do not meet the needs of bioscience. However if there is a communal will, modern chemical informatics technology can provide what is required. Several excellent models for the capture of macromolecular sequence and structure data (e.g. the protein data bank) inform our architecture. Data in published articles can include reference to chemical compounds (often in free text), details of their synthesis (in vivo and in vitro), proof of their structure (spectra and analytical data), methods and reagents in bioscience protocols, the physical and biological properties and reaction of compounds both in enzymes and enzyme-free systems. With the tools that we and others have developed, this information can now be automatically captured with high precision from primary publications, especially if structured authoring tools are widely used. Unlike bioinformaticists, who routinely apply data- and text-mining tools in their research, conventional chemists appear culturally more suspicious of robotic data extraction and continue to rely on manually curated secondary publications whose philosophy has barely altered over 120 years. Such sources are necessarily incomplete in time, coverage and coverage of information types. For example for 99+% of all newly synthesised compounds the papers report that Infra-red spectra have been recorded, but only a tiny fraction of this is available in electronic form. We argue that even modest improvement in such a data capture rate would make an enormous difference. Moreover the data would be of consistently higher quality than manually re-keyed data. The main challenge is a cultural one. Thus biosciences and crystallography have communally convinced authors and publishers of the value of author-based deposition of data, later aggregated in communally accessible databanks. This largely does not currently happen in chemistry, where information is manually extracted from the primarily literature, jealously guarded and sold back to the community. Mechanisms such as \"supporting or supplemental data deposition\" are not widely used, and when they are, little care is given to enabling its re-use. One of the major secondary publishers has recently criticised the bioscience community for aggregating chemical data. \"It \[PubChem\] would not only injure us significantly, it would put information for free in the hands of world scientists and do it all with taxpayer money,\" Massie \[CEO, Chemical Abstract Service\] said. \"For me to wake up one morning and find I have to compete with my own government is extraordinary."\[[@B2]\] The attitude in chemistry to modern informatics (XML, ontologies, RDF, text-mining, metadata, etc.) \[[@B3]\] is largely apathetic, with some data- or software-centric organisations actively opposing interoperability for commercial reasons. This problem extends to mainstream chemical software, where there are no Open standards and where algorithms are closed and obscure. We have argued that the large data aggregators produce vendor-centric access systems to meet their needs rather than the community\'s. Another problem is that access is often only allowable on a per-item basis rather than to the data collection as a whole. This monopolistic \"thought control\" in chemistry stifles innovation in data-led science. However the Opens (Data, Source, Access, Standards) are changing the practice of scientific informatics and chemistry is starting to be affected. We therefore look to bioscience to take a lead in helping realise the following vision. We now believe that there are already enough Open tools and Open resources which can make the vision attractive and cost-effective. A model for automatic capture of chemical information ----------------------------------------------------- Much chemical data is largely context-free in that it can be understood and recreated independently of the location or motivation. The primary data model, inspired by Konrad Beilstein in the 19th century, has three components: compound, properties and citations. A pure compound is described by an immutable structural formula and has precisely reproducible properties. Current thinking asserts that the biological action of a compound is, in principle, reproducible and predictable if the system is carefully enough replicated and the components understood. This is the central dogma of the chemically-based pharmaceutical industry and the chemical information industry on which it relies. Chemistry has a tradition of ensuring quality through reporting properties and analysis, so every new compound (and many re-synthesised ones) must have published measurements of properties to justify their identity and purity. These facts are available, in text form, in the primary literature in which over a million new compounds are published annually. Because structure predicts properties, and because drug discovery is so difficult, the pharmaceutical industry tests many compounds for biological activity. The data in these public publications is a major feedstock for the chemical information industry. The chemical bioscientist has almost all of the required information available in electronic form on their benchtop already! It could be deposited for the scientific community with virtually no human intervention. We believe that, with the help of forward looking publishers, a working protocol can be set up in bioscience, which will then inspire (or terrify) mainstream chemical informatics. Note that much of the information captured is additional to that which the current abstracters collect. We argue that the key components to automatically capture chemical information are already in place (and are discussed in more detail in an accompanying technical article). We envisage the chemistry which can be captured using such mechanisms includes (a) Chemical entities and names. Many compounds have no explicit structures and are mentioned only by name or identifiers. Where these relate to specific compounds it is valuable to link them to a precise identification, such as PubChem. (b) Molecular structure, expressed as a compositional formula (e.g. CHaOH for methanol) and a graphical structural formula (\"2D diagram\" or connection table). (c) Spectra and physical properties. Much such information is already in digital form when produced by instruments (whose manufacturers are starting to create Open approaches \[[@B4]\], but is largely destroyed by conventional publishing processes If a community-wide digital template for the submission of this information were available and encouraged by publishers it would be welcomed by many, would eliminate errors introduced by transcription, and enable machine-reviewing of data leading to a higher standard of published data. The basis of our model involves conversion of experimental data to XML and its merger with the conventional text (giving a "datument" \[[@B5]\]). The author uses a authoring tool which can manage structured XML documents and provide normal textual support (spellchecks, etc.). The resulting datument contains fine-grained markup of facts (molecules, measurements, properties, chemical names) and can automatically be used to create derivatives such as the \"full-text\" or the \"supplemental data\". The complete datument, if Open, or the \"data\" if not is then reposited for further harvesting. All compound/property data is available for datamining and computational re-use (e.g. for further in silico prediction. Realising the vision -------------------- ### Data repositing and maintenance The current dissemination of data through publishers is largely unsatisfactory. Thus although many publishers allow the deposition of factual \"supplementary data", our experience with most is that it is an unwelcome chore, poorly resourced and maintained. Moreover although reviewers are often do what they can to validate data, publishers themselves do not. We believe that many publishers would welcome a model where they were no longer involved in data repositing. A few publishers such as the International Union of Crystallography are more committed to the curation of data; others in the biosciences see the value of semantically enhanced data.The crystallographic experience has shown that expert computer programs can act as powerful reviewers complementing the human; automatic curation enhances, rather than lowers, data quality. Our model is based on the availability of repositories, primarily Institutional, that accept data as well as full text. Already some academic institutions and an increasing number of funders mandate that research output should be reposited and there are national initiatives to develop the infrastructure. The storage for XML-ised chemical data is modest (less than 1 mbyte per publication) and we have shown that large numbers of molecules can be deposited in our own institutional repository \[[@B6]\] and recovered by undirected search engines such as Google \[[@B7]\]. Chemical data has required no semantic maintenance (e.g. through changes in meaning or use) over many decades and we see this continuing, so that the maintenance costs are those general to any repository. Components in a repository have a unique handle with which, in principle, a Digital object or other identifier (DOI) \[[@B8]\] can be associated so that data can be cloned for access and preservation. The handle or DOI would be published in the \"full text\" and would bind the data to it more effectively that at present and hopefully indefinitely. ### Metadata Through the InChI (International Chemical Identifier)\[[@B8]\] and a controlled vocabulary of chemical properties, generic search engines can achieve a very high degree of recall. This means that discovery and aggregation can be built on maintenance-free generic technology and can be made completely automatic, Conventionalists would argue of course that human curation is essential for re-usable chemical data. In a similar vein to much bioinformatic, we argue that robots can discover patterns in data, compounds and authors which are at least as powerful as many current abstracting services. Where human evaluation is critical (e.g. in human medicine, patents, etc.) then the robots will provide the primary resources on which a judgement can be based. ### Rights We assume that most bioscience authors and publishers will agree that whether or not a paper is Open Access the facts (and thereby all \"supplemental data\") therein are not copyrightable. XML resolves differences of interpretation in that XML markup can be regarded as identifying factual information and this would be consistent with its re-use under (say) the Budapest Open Access Initiative. In this way all published chemical data can be made immediately, completely and clearly available for indefinite scientific re-use. ### Potential Because the chemical information is structured we now have a biocheminformatics \"cycle\" where, for the first time, large scale robotic data analysis can take place. The data in the research (laboratory, in silico, or both) are published in a lossless manner. Molecules and their properties have unique identifiers as described above and can be integrated into mainstream bioinformatics in the same manner as collections such as PubChem, the macromolecular stucture database (MSD at EBI), the Kyoto Encyclopedia of Genes and Genomes (KEGG) etc. They will bring the added value of consistently captured property data and spectra. We also expect that many in silico properties will then be systematically added. ### Author and publisher compliance The introduction of structured authoring tools (e.g. Publicon) \[[@B9]\] will help this process considerably. Templates can be created for the chemical components described above and where the information exists in XML (connection tables, spectra, properties) it should be as easy as for committed authors as using a semantically void tool (e.g. Word). Where information needs to be converted from legacy formats, an increasing number of open Web Services, which publishers (and authors) may clone and customise are becoming avilable. We expect authors to have a greater incentive (even if only through mandation) to reposit data and to disseminate research findings. This also raises the vision of changing the \"citation economy\" (which values market perception) to a \"reuse economy\" where a the data in a paper are valued by how often they are re-used.	PubMed Central	2024-06-05T03:55:57.549636	2005-6-7	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1149499/", "journal": "BMC Bioinformatics. 2005 Jun 7; 6:141", "authors": [ { "first": "Peter", "last": "Murray-Rust" }, { "first": "John BO", "last": "Mitchell" }, { "first": "Henry S", "last": "Rzepa" } ] }
PMC1150289	Introduction {#s1} ============ Expertise in detecting and recognizing objects in natural scenes, where targets are camouflaged by their backgrounds, is critical for many of our interactions in complex environments: from identifying predators or prey and recognizing poisonous foods, to diagnosing tumors on medical images and finding familiar faces in the crowd. As with many skills, learning has been shown to be a key facilitator in the detection and recognition of targets in cluttered scenes \[[@pbio-0030204-b01]\]. Previous neurophysiological \[[@pbio-0030204-b09]\] and imaging \[[@pbio-0030204-b16]\] studies on object learning have concentrated on the higher stages of visual (inferior temporal cortex) and cognitive processing (prefrontal cortex), providing evidence that the representations of shape features in these areas are modulated by learning. In contrast, computational approaches have proposed that associations between features that mediate the recognition of familiar objects may occur across different stages of visual analysis, from orientation detectors in the primary visual cortex to occipitotemporal neurons tuned to object parts and views \[[@pbio-0030204-b20]\]. However, the neural implementation of object learning mechanisms across stages of visual analysis is largely unknown, and the question of how the visual brain learns objects in natural cluttered scenes remains open. The aim of our study was 2-fold: (1) to investigate the neural plasticity mechanisms that mediate shape learning in cluttered scenes across stages of visual processing in the human visual cortex, and (2) to examine the effect of regularities present in natural scenes (i.e., grouping of similar features) that determine the distinctiveness of targets in noisy backgrounds (i.e., perceptual saliency) on this learning-dependent plasticity. To this end, we used human functional magnetic resonance imaging (fMRI) combined with psychophysics. To gain insight into the neural mechanisms that mediate shape-specific learning, we examined fMRI responses evoked when observers detected shapes that they had learned through training compared with responses evoked when observers detected shapes on which they had not been trained. To investigate the effects of learning in the detection of visual shapes in cluttered scenes, we manipulated the salience of the target shapes by altering their distinctiveness from the background ( [Figure 1](#pbio-0030204-g001){ref-type="fig"}). We compared behavioral performance and fMRI responses for low-salience shapes in noise (Experiment 1) and high-salience pop-out targets (Experiment 2). ::: {#pbio-0030204-g001 .fig} Figure 1 ::: {.caption} ###### Stimuli Examples of symmetrical and asymmetrical low- and high-salience stimuli. ::: ![](pbio.0030204.g001) ::: Our stimuli consisted of shapes defined by a closed contour of similarly oriented Gabor elements that were embedded in a background of Gabor elements. These stimuli (see [Figure 1](#pbio-0030204-g001){ref-type="fig"}) yield the perception of a global figure in a textured background rather than simple paths (i.e., open contours). These aligned contours have been shown to result from the integration of the similarly oriented elements into global configurations \[[@pbio-0030204-b23]\]. Previous work has shown that these stimuli involve processing in both early retinotopic and higher occipitotemporal regions \[[@pbio-0030204-b26]\]. In Experiment 1, observers were presented with low-salience stimuli in which shapes were embedded in a background of randomly positioned and oriented Gabors. In Experiment 2, high-salience stimuli were used in which shapes were embedded in a background of randomly positioned, but uniformly oriented Gabors. In both experiments, observers were required to decide which of two shapes presented on either side of the central fixation point was symmetrical. Initially, observers performed this task in the scanner with two sets of untrained stimuli. Observers were then trained in the laboratory with feedback on three consecutive days on one set of stimuli, and then tested again in the scanner with the trained set and the originally presented, untrained set of stimuli ( [Figure 1](#pbio-0030204-g001){ref-type="fig"}). Our findings suggest a link between shape-specific perceptual learning and neural plasticity mechanisms in the human visual cortex. Specifically, for low-salience shapes, improved behavioral performance was coupled with increased fMRI responses for trained shapes in early retinotopic areas (V1, V2, Vp, and V4v) and the lateral occipital complex (LOC), a region in the lateral occipital cortex extending anterior in the temporal cortex ( [Figure 2](#pbio-0030204-g002){ref-type="fig"}) that is thought to be involved in the analysis of object shape \[[@pbio-0030204-b27]\] and processes of object recognition \[[@pbio-0030204-b19], [@pbio-0030204-b27]\]. In contrast to the increased responses for trained low-salience shapes, when observers learned high-salience shapes, lower fMRI responses were evoked in the LOC, and no evidence for plasticity in early retinotopic visual areas was observed. These findings provide novel evidence for distributed neural plasticity mechanisms across stages of visual analysis that are adaptable to image regularities which determine the perceptual saliency of targets in cluttered scenes ( [Figure 2](#pbio-0030204-g002){ref-type="fig"}). ::: {#pbio-0030204-g002 .fig} Figure 2 ::: {.caption} ###### ROIs Functional activation maps for one subject showing the early retinotopic ventral (V1, V2, VP, and V4) and dorsal (V1, V2, V3, and V3a) areas and the LOC subregions. Functional activations (color-maps) are superimposed on flattened cortical surfaces of the right and left hemispheres. A, anterior; CoS, collateral sulcus; ITS, inferior temporal sulcus; OTS, occipitotemporal sulcus; P, posterior; STS, superior temporal sulcus. ::: ![](pbio.0030204.g002) ::: Results {#s2} ======= Experiment 1: Learning Low-Salience Shapes {#s2a} ------------------------------------------ ### Behavioral performance {#s2a1} In experiment 1 we examined behavioral performance (accuracy in detecting symmetrical shapes) and fMRI responses when observers were trained with low-salience shapes. [Figure 3](#pbio-0030204-g003){ref-type="fig"}A shows the behavioral performance of the observers for trained and untrained stimuli during the scanning sessions before and after training. Before training, observers\' performance was similar for the set of stimuli that would become familiar through training and for the set of stimuli that would remain untrained. This was expected because both sets of stimuli were novel before training. However, after training, there was a significant increase in performance for the trained, but not the untrained, stimuli. Specifically, a repeated measures analysis of variance (ANOVA) showed the main effects of test session (before and after training) (F~1,10~ = 15.81, p \< 0.01) and familiarity (trained and untrained) (F~1,10~ = 81.63, p \< 0.001) and a significant interaction between these variables (F~1,10~ = 66.76, p \< 0.001). ::: {#pbio-0030204-g003 .fig} Figure 3 ::: {.caption} ###### Results for Experiment 1 Psychophysical data (A) and fMRI responses obtained during the scanning sessions before (B) and after (C) training. Error bars indicate the SEM across subjects. Significant differences are indicated by asterisks. (A) Psychophysical data (percent correct) are shown for trained and untrained shapes in the tests before and after training. Normalized fMRI responses across subjects for trained and untrained shapes before (B) and after (C) training across the LOC subregions and the early ventral areas. Normalized fMRI responses were computed by subtracting the mean signal (percent signal change from fixation baseline) across conditions, sessions, and ROIs from the signal in each condition per subject and adding the overall average across conditions, sessions, ROIs, and subjects. These normalized fMRI responses indicate differences across conditions independent of the variability in the fMRI signal across subjects, scanning sessions, and ROIs. ::: ![](pbio.0030204.g003) ::: ### Pre-training and post-training fMRI data {#s2a2} For each individual subject, we identified the early visual areas, posterior (lateral occipital \[LO\]) and anterior (posterior fusiform sulcus \[pFs\]) subregions of the LOC (see [Figure 2](#pbio-0030204-g002){ref-type="fig"}), as cortical regions of interest (ROIs) in which we examined fMRI responses before and after training. Before training, fMRI responses in these regions were not different between the two sets of shapes (those to become trained vs. those to remain untrained) ( [Figure 3](#pbio-0030204-g003){ref-type="fig"}B). Specifically, a repeated measures ANOVA showed no effect of familiarity (F~1,50~ \< 1, p = 0.65). Again, this is not surprising because the subjects have not been trained on any shapes. However, after subjects had been trained on the low-salience shapes, significantly stronger fMRI responses were observed for trained than untrained shapes ( [Figure 3](#pbio-0030204-g003){ref-type="fig"}C). This was true in both the early visual areas (F~1,50~ = 7.02, p \< 0.05) and the LOC subregions (LO: (F~1,50~ = 10.92, p \< 0.001), pFs: (F~1,50~ = 5.85, p \< 0.01)). Further, comparison of the fMRI responses before and after training showed a significant (F~1,50~ = 3.06, p \< 0.05) interaction between test session (before and after training) and familiarity (trained and untrained). That is, we observed significantly stronger responses for the trained shapes (F~1,50~ = 3.58, p \< 0.05) after than before training, but no significant differences for the untrained shapes (F~1,50~ \< 1, p = 0.41) ( [Figure 3](#pbio-0030204-g003){ref-type="fig"}). Experiment 2: Learning High-Salience Shapes {#s2b} ------------------------------------------- ### Behavioral performance {#s2b1} In Experiment 2, we examined behavioral performance and fMRI responses when observers were trained with high-salience shapes. [Figure 4](#pbio-0030204-g004){ref-type="fig"}A shows the behavioral performance of the observers for trained and untrained stimuli during the scanning sessions before and after training. Before training on high-salience shapes, there was no difference in performance when testing with the different stimulus sets (those to become trained vs. those to remain untrained). However after training, observers showed a significant improvement for the trained shapes compared with the untrained shapes. Specifically, a repeated measures ANOVA showed main effects of test session (F~1,7~ = 80.98, p \< 0.001), familiarity (F~1,7~ = 37.70, p \< 0.001), and a significant interaction for test session and familiarity (F~1,7~ = 39.05, p \< 0.001). ::: {#pbio-0030204-g004 .fig} Figure 4 ::: {.caption} ###### Results for Experiment 2 Psychophysical data (A) and fMRI responses obtained during the scanning sessions before (B) and after (C) training. Error bars indicate the SEM. Significant differences are indicated by asterisks. (A) Psychophysical data (percent correct) for trained and untrained shapes in the tests before and after training. Normalized fMRI responses across subjects for trained and untrained shapes before (B) and after (C) training across the LOC subregions and early ventral areas. ::: ![](pbio.0030204.g004) ::: ### Pre-training and post-training fMRI data {#s2b2} As shown in [Figure 4](#pbio-0030204-g004){ref-type="fig"}B, and as expected, no differences were observed before training between the fMRI responses to shapes that would become trained and those that would remain untrained. That is, there was no significant effect for familiarity (F~1,35~ \< 1, p = 0.82) before training. In contrast to the results from Experiment 1, after training we found lower fMRI responses for trained compared to untrained stimuli for high-salience shapes ( [Figure 4](#pbio-0030204-g004){ref-type="fig"}C). A further difference from Experiment 1 was that this learning effect was evident in the LOC, but not in early visual areas. Specifically, a repeated measures ANOVA showed significantly stronger responses for untrained than trained shapes in the LOC subregions (LO: (F~1,35~ = 13.73, p \< 0.01), pFs: (F~1,35~ = 5.21, p \< 0.05)) but not in early visual areas (F~1,35~ \< 1, p = 0.38). Further, comparison of the fMRI responses before and after training showed a significant interaction between test session and ROI (F~1,35~ = 3.87, p \< 0.01), and familiarity and ROI (F~1,35~ = 3.11, p \< 0.05). That is, in the LOC we observed significantly lower responses for the trained shapes (F~1,35~ = 27.43, p \< 0.001) after than before training, but no significant differences for the untrained shapes (F~1,35~ = 1.58, p = 0.19). In the early visual areas, no significant differences were observed for trained (F~1,35~ = 2.92, p = 0.13) or untrained (F~1,35~ = 2.77, p = 0.14) shapes before and after training ( [Figure 4](#pbio-0030204-g004){ref-type="fig"}). Comparison across Experiments {#s2c} ----------------------------- In [Figure 5](#pbio-0030204-g005){ref-type="fig"}, we summarize the fMRI learning effects after training for low-salience (Experiment 1) and high-salience (Experiment 2) stimuli, by plotting the differences between fMRI responses for trained and untrained stimuli in the post-training session across visual areas. This analysis showed positive differences (stronger fMRI responses for trained than untrained stimuli) for low-salience shapes across visual areas, whereas negative differences (stronger fMRI responses for untrained than trained stimuli) for high-salience shapes in the LOC. ::: {#pbio-0030204-g005 .fig} Figure 5 ::: {.caption} ###### Summary of Results fMRI learning effects for low-salience (Experiment 1) and high-salience (Experiment 2) shapes indicated by subtracting the fMRI responses for untrained from those for trained stimuli in the post-training session in each experiment. Error bars are plus or minus the SEM. Positive values indicate stronger fMRI responses for trained stimuli, whereas negative values indicate stronger fMRI responses for untrained stimuli. ::: ![](pbio.0030204.g005) ::: To further quantify the relationship between the behavioral and fMRI learning effects, we conducted a regression analysis on the psychophysical and fMRI responses from individual subjects across visual areas for Experiments 1 and 2. This analysis provides additional evidence for a link between behavioral improvement and neuronal changes after training; that is, higher differences between trained and untrained stimuli were observed in both the behavioral and fMRI responses after than before training. As shown in [Figure 6](#pbio-0030204-g006){ref-type="fig"}, for low-salience shapes, this regression analysis was significant in early visual areas (V1 is shown as a representative area, but see figure caption for other areas) and the LOC subregions, whereas for high-salience shapes, the regression was significant only in the LOC subregions. The majority of positive points in the plots for low-salience shapes indicates stronger responses for trained than untrained shapes after training, whereas the majority of negative points for high-salience shapes indicates lower responses for trained than untrained shapes after training ( [Figure 6](#pbio-0030204-g006){ref-type="fig"}). ::: {#pbio-0030204-g006 .fig} Figure 6 ::: {.caption} ###### Relationship between Psychophysical and fMRI Learning Effects fMRI data and the corresponding psychophysical response for low-salience (A) and high-salience (B) shapes. For each individual subject, we plotted a behavioral learning index (percent correct for trained minus percent correct for untrained stimuli) and an fMRI learning index (percent signal change for trained minus percent signal change for untrained stimuli) after training. Positive values indicate stronger responses for trained than untrained shapes, whereas negative values indicate lower responses for trained than untrained shapes. For low-salience shapes the regression analysis was significant in early visual areas (V1: r = 0.62, F~1,21~ = 13.65, p = 0.001; V2: r = 0.42, F~1,21~ = 6.65, p \< 0.05; Vp: r = 0.50, F~1,21~ = 7.62, p = 0.01; V4: r = 0.50, F~1,21~ = 8.91, p \< 0.01) and the LOC subregions (LO: r = 0.57, F~1,21~ = 9.84, p \< 0.01; pFs: r = 0.51, F~1,21~ = 7.18, p = 0.01). For high-salience shapes the regression was significant only in the LOC subregions (LO: r = 0.56, F~1,15~ = 6.69, p \< 0.05; pFs: r = 0.61, F~1,15~ = 8.58, p = 0.01) but not in the early visual areas (V1: r = 0.17, F~1,15~ \< 1, p = 0.51; V2: r = 0.25, F~1,15~ \< 1, p = 0.34; Vp: r = 0.28, F~1,15~ = 1.27, p = 0.27; V4: r = 0.15, F~1,15~ \< 1, p = 0.56). ::: ![](pbio.0030204.g006) ::: Might the different fMRI learning effects in the LOC for low-salience (Experiment 1) and high-salience (Experiment 2) shapes be due to the subjects being less interested or paying less attention to the high- than the low-salience trained stimuli? We think that it is unlikely that the different fMRI learning effects for low- and high-salience shapes could be significantly confounded by these general attention/arousal differences across conditions for the following reasons. First, the similar behavioral learning effects for low- and high-salience shapes indicate that the observers were attentive in both tasks. Specifically, the difference in accuracy between trained and untrained shapes after training was 23.8% for low-salience shapes and 19.9% for high-salience shapes. Moreover, the observers performed the task even in the hardest condition, with untrained low-salience stimuli, as indicated by their accuracy in this condition being above chance (t~10~ = 4.23, p \< 0.01). Further, reaction times in this condition were the slowest ( [Figure 7](#pbio-0030204-g007){ref-type="fig"}) rather than very fast, as would be expected if the observers had given up and were simply guessing. These psychophysical data indicate that observers were engaged in the task and not responding randomly. Further, it is highly unlikely that observers could selectively choose to attend to particular conditions as trials were presented in quick succession and were randomly interleaved. Second, if the results in the LOC were simply due to task difficulty, the following pattern in the strength of fMRI responses would be expected (from high to low): untrained low saliency, untrained high saliency, trained low saliency, trained high saliency. However, the fMRI responses in the hardest (lowest accuracy) condition, untrained low-salience shapes, did not differ (F~1,7~ \< 1, p = 0.68) from the responses in the easiest (highest accuracy) condition, trained high-salience shapes. Third, the lack of differences in the activations for trained versus untrained high-salience shapes in the early visual areas suggests that the effects observed in the LOC were not simply due to differences in general alertness or arousal across conditions that could modulate responses across all visual areas \[[@pbio-0030204-b28], [@pbio-0030204-b29]\]. Fourth, comparison of the variances after training did not show any significant differences (Levene\'s test, p \> 0.05 for all ROIs) across experiments, suggesting that the different fMRI learning effects across experiments were unlikely to be due to variance differences. Finally, an additional control experiment ([Figure S1](#sg001){ref-type="supplementary-material"}) in which the observers performed a target-monitoring task \[[@pbio-0030204-b28], [@pbio-0030204-b29]\] that ensured that the observers attended similarly across conditions showed similar patterns of fMRI learning effects as those reported in Experiments 1 and 2. ::: {#pbio-0030204-g007 .fig} Figure 7 ::: {.caption} ###### Reaction Times Reaction times before and after training for Experiments 1 (A) and 2 (B). Error bars are plus or minus the SEM. ::: ![](pbio.0030204.g007) ::: Consistently, analysis of the reaction times (see [Figure 7](#pbio-0030204-g007){ref-type="fig"}) showed that the learning effects observed in the LOC could not be due simply to differences in the duration of stimulus processing across conditions. After training, observers were slower for the untrained than the trained shapes in both experiments (Experiment 1: F~1,7~ = 136.64, p \< 0.001); Experiment 2: (F~1,7~ = 202.35, p \< 0.001). This effect would predict higher fMRI responses for untrained than trained stimuli in both experiments and thus could not explain the differences in the activation patterns observed across experiments. Finally, it is not likely that our learning results could be significantly confounded by eye movements. Eye movement recordings showed that the subjects were able to fixate for long periods of time, and any saccades that occurred did not differ systematically in number, amplitude, or duration for trained and untrained shapes after training ([Figure S2](#sg002){ref-type="supplementary-material"}). Discussion {#s3} ========== Our experiments provide novel evidence suggesting (1) a link between behavioral improvement in shape-specific perceptual learning and neuronal plasticity in the human visual cortex, and (2) distributed plasticity mechanisms across cortical stages of visual analysis that are adaptable to natural image regularities (e.g., grouping of background elements that have the same orientation) which determine the salience of targets in cluttered scenes. In particular, the behavioral results suggest that training enhances the observers\' ability to detect shapes embedded in noisy backgrounds, providing evidence for shape-specific learning. The fMRI data suggest that these learning-dependent plasticity mechanisms, as measured by fMRI at the level of large neural populations, differ depending on the salience of the shapes. Specifically, when the shapes appeared camouflaged in cluttered backgrounds (low salience), fMRI responses were higher for trained than untrained shapes, suggesting enhanced representations of the trained shapes. However, when shapes popped out from the background (high salience), decreased fMRI responses were observed for trained shapes, suggesting sparser coding after training. Interestingly, this learning-dependent plasticity was distributed across early and higher visual areas for low-salience shapes, but was restricted to higher occipitotemporal areas for high-salience shapes. We now review these main findings in further detail. Behavioral Improvement and Learning-Dependent Plasticity {#s3a} -------------------------------------------------------- Several psychophysical studies have shown perceptual learning at different stages of visual analysis from features \[[@pbio-0030204-b30]\] to complex objects \[[@pbio-0030204-b03], [@pbio-0030204-b05], [@pbio-0030204-b37]\]. Further, several neurophysiological \[[@pbio-0030204-b09], [@pbio-0030204-b10], [@pbio-0030204-b12], [@pbio-0030204-b13], [@pbio-0030204-b15], [@pbio-0030204-b40]\] and some imaging \[[@pbio-0030204-b16], [@pbio-0030204-b48]\] studies provide evidence that behavioral improvement after training correlates with changes in neuronal sensitivity. Our findings extend our understanding of learning mechanisms, by directly testing the neural correlates of shape learning in the human visual cortex using combined psychophysical and fMRI measurements before and after training. To acquire these event-related fMRI data with high spatiotemporal resolution our investigations concentrated in occipitotemporal regions. Future studies on the cortical connectivity between these areas and prefrontal regions thought to be involved in perceptual learning \[[@pbio-0030204-b44]\] will provide further insights in the neural plasticity of the cortical circuits that mediate adaptive cognitive behaviors. Perceptual Learning Mechanisms and Shape Salience {#s3b} ------------------------------------------------- To investigate how the visual brain learns novel objects in cluttered scenes, we chose stimuli that resemble camouflage conditions in natural images where targets are hidden due to their feature similarity with the background. Recent studies suggest that regularities (e.g., orientation similarity for neighboring elements) are characteristic of natural scenes and the primate brain has developed a network of connections that mediate integration of features based on these correlations \[[@pbio-0030204-b52]\]. In our stimuli, the orientation similarity of the target elements facilitates their grouping into global shapes. Furthermore, the uniform orientation of the background elements in the high-salience stimuli enhances the segmentation and thus the salience of the target shapes. Our findings revealed that plasticity mechanisms underlying shape learning in cluttered scenes are adaptable to these natural regularities and modulated by the perceptual saliency of the target shapes. Although our stimuli are optimal for tapping into the processing of early visual areas, these plasticity mechanisms could contribute in general to the improved detection of more natural ambiguous or low-salience targets, consistent with recent physiological investigations \[[@pbio-0030204-b45]\]. In particular, our findings are consistent with the idea that training with low-salience targets in cluttered scenes increases neuronal sensitivity to the target features and facilitates the detection and integration of local features into global shapes. Specifically, the learning of low-salience target shapes resulted in stronger responses to trained than untrained shapes in both early and higher visual areas. This increased neuronal sensitivity during perceptual learning \[[@pbio-0030204-b10], [@pbio-0030204-b11], [@pbio-0030204-b43], [@pbio-0030204-b50]\] has been suggested to involve increased recruitment of neurons with enhanced responses to similar features of the trained stimuli. As a result, the signal-to-noise ratio in the neural responses is increased for trained compared to untrained shapes. This process may enhance the salience of the target features, facilitating their segmentation from the background and enhancing the global integration that is important for the detection and recognition of visual targets in noise. In contrast, when targets appear in uniform backgrounds, they are easily segmented and can be searched more efficiently \[[@pbio-0030204-b55], [@pbio-0030204-b56]\]. The lower fMRI responses observed for trained than untrained high-salience shapes are consistent with the idea that training with these pop-out targets engages smaller neural ensembles that increase their selectivity for features unique to the stimulus but most relevant for its discrimination in the context of a task. This mechanism results in sparser but more efficient representations \[[@pbio-0030204-b57]\] of the trained stimuli or features that are important for prompt and successful object categorization and recognition. Supporting evidence for such a mechanism comes from learning effects in the primary visual cortex after training on orientation discrimination tasks \[[@pbio-0030204-b42], [@pbio-0030204-b48]\], and the prefrontal cortex \[[@pbio-0030204-b44]\] where fewer neurons respond selectively to familiar than to novel objects, but they are more narrowly tuned. Interestingly, this dissociable pattern of fMRI learning effects for low-compared to high-salience shapes provides insights into the activation patterns observed across previous learning studies. Previous studies have suggested that learning results from active long-term training \[[@pbio-0030204-b58]\] or rapidly from single \[[@pbio-0030204-b59]\] or repetitive exposure \[[@pbio-0030204-b60]\] to a stimulus. In our study the observers had substantial training (1,200 trials: three sessions of 400 trials each) that resulted in high-accuracy performance. It is possible that single or multiple passive exposures to target stimuli without extensive training would result in similar learning effects as those observed in our study. Taken together, previous fMRI studies show similar effects for long-term and rapid learning that depend on the nature of the stimulus representation. In particular, consistent with the fMRI activations for our low-salience shapes, enhanced responses have been observed when learning engages processes necessary for the formation of new representations, as in the case of unfamiliar \[[@pbio-0030204-b17], [@pbio-0030204-b61], [@pbio-0030204-b62]\], degraded \[[@pbio-0030204-b16], [@pbio-0030204-b63], [@pbio-0030204-b64]\] masked unrecognizable \[[@pbio-0030204-b19], [@pbio-0030204-b65]\] or noise-embedded \[[@pbio-0030204-b45], [@pbio-0030204-b49], [@pbio-0030204-b50]\] targets. However, when the stimulus perception is unambiguous (e.g., familiar, undegraded, recognizable targets presented in isolation), similar to our high-salience shapes, training results in more efficient processing of the stimulus features indicated by attenuated neural responses \[[@pbio-0030204-b18], [@pbio-0030204-b48], [@pbio-0030204-b62], [@pbio-0030204-b65]\]. Importantly, these effects are evident in areas that encode the relevant stimulus features selectively, whereas opposite activation patterns may occur in other cortical areas implicated in the task performed by the observers \[[@pbio-0030204-b44], [@pbio-0030204-b45], [@pbio-0030204-b50], [@pbio-0030204-b66], [@pbio-0030204-b68]\]. Distributed Learning-Dependent Plasticity across Visual Areas {#s3c} ------------------------------------------------------------- Finally, the contribution of the different visual areas in shape learning appears to depend on the salience of the target shapes. We observed fMRI learning effects in both the early visual areas and the LOC for low-salience shapes but only in the LOC for the high-salience shapes. Previous neurophysiological \[[@pbio-0030204-b09], [@pbio-0030204-b43]\] and imaging \[[@pbio-0030204-b16]\] studies have implicated temporal and frontal areas in the learning of complex objects but have not investigated the contribution of early visual areas in shape learning. Early retinotopic areas have been proposed by several psychophysical studies to be involved in learning feature-discrimination tasks \[[@pbio-0030204-b30], [@pbio-0030204-b33], [@pbio-0030204-b69]\], consistent with the specificity of the learning effects for the stimulus position in the visual field \[[@pbio-0030204-b30], [@pbio-0030204-b36], [@pbio-0030204-b72]\] and the trained stimulus attribute \[[@pbio-0030204-b04], [@pbio-0030204-b33], [@pbio-0030204-b70], [@pbio-0030204-b71], [@pbio-0030204-b75]\]. However, neurophysiological evidence for the contribution of V1 in behavioral improvement after training on visual discrimination, remains controversial \[[@pbio-0030204-b41], [@pbio-0030204-b42]\]. There is some evidence for sharpening of orientation tuning \[[@pbio-0030204-b42]\] after training, but no evidence for changes in the size of the cortical representation or the receptive field properties \[[@pbio-0030204-b41], [@pbio-0030204-b76]\]. Our findings are in accordance with studies suggesting that learning is mediated by interactions between global shape-analysis mechanisms and local connections, and its neural locus could be modulated by the task context \[[@pbio-0030204-b07], [@pbio-0030204-b59], [@pbio-0030204-b70], [@pbio-0030204-b71], [@pbio-0030204-b73], [@pbio-0030204-b76]\]. In particular, the recognition of low-salience targets in cluttered scenes entails integration of features to global configurations and figure--ground segmentation. These processes are known to involve both early and higher visual areas \[[@pbio-0030204-b26]\]. The similar fMRI responses for low-salience shapes in the LOC and the early visual areas (F~5,50~ \< 1, p = 0.77) are consistent with the involvement of both early and higher visual areas in the detection of shapes in noise. Learning has been suggested to modulate neuronal sensitivity in these areas \[[@pbio-0030204-b82]\] either by modulating networks of lateral interactions in the early visual areas \[[@pbio-0030204-b06], [@pbio-0030204-b49], [@pbio-0030204-b76]\] or via feedback connections from higher visual areas \[[@pbio-0030204-b05], [@pbio-0030204-b59]\]. However, when a salient target is present in the scene, its segmentation is easily achieved and learning may contribute to the representation of the critical features for shape recognition. Thus, learning tunes the representations of global shapes that are known to involve higher occipitotemporal areas \[[@pbio-0030204-b27]\]. Our results showing stronger fMRI responses for high-salience shapes in the LOC than in the early visual areas (F~5,35~ = 1.91, p \< 0.05) are consistent with processing of salient global shapes in the LOC. Consistent with this evidence for distributed cortical plasticity, recent psychophysical studies \[[@pbio-0030204-b59], [@pbio-0030204-b83], [@pbio-0030204-b84]\] have proposed a reverse hierarchy theory (RHT) of perceptual learning. This theory proposes that learning begins at high-level areas for easy tasks and proceeds to early retinotopic areas that have higher resolution necessary for finer and more difficult discriminations. Although fMRI studies lack the temporal resolution necessary for testing this proposal, our findings are consistent with plasticity mechanisms in early visual areas that mediate learning in difficult and fine tasks (i.e., detection of low-salience rather than high-salience shapes) \[[@pbio-0030204-b55], [@pbio-0030204-b56], [@pbio-0030204-b59]\]. It is possible that these learning effects in early visual areas are the result of feedback from higher areas. As the discrimination of low-salience shapes improves with training \[[@pbio-0030204-b06], [@pbio-0030204-b78], [@pbio-0030204-b79]\], higher shape-related areas increase their responses and enhance the processing of the trained shape features in the early visual areas that have fine spatial resolution necessary for the detection of targets in noise. Finally, this theory makes interesting predictions for learning specificity to the trained features in easy tasks, in contrast to generalization across image changes in difficult tasks \[[@pbio-0030204-b85]\]. Testing these predictions for specificity, feedback triggered by single vs. repeated exposure, and long-lasting plasticity would be of interest in future studies. Conclusions {#s3d} ----------- In summary, our findings propose that the human brain learns novel objects in complex scenes by reorganizing shape processing across early and higher cortical stages of visual analysis. Interestingly, this learning-dependent plasticity is implemented by mechanisms that are adaptable to the target scene. That is, the visual brain appears to take advantage of natural image correlations that determine the target distinctiveness in a scene while learning novel object targets. Our study provides novel neuroimaging evidence that this opportunistic learning \[[@pbio-0030204-b05]\] of salient targets in natural scenes is mediated by sparser feature coding at higher stages of visual analysis, whereas learning of camouflaged targets is implemented by bootstrapped mechanisms \[[@pbio-0030204-b05]\] that enhance the segmentation and recognition of ambiguous targets in both early and higher visual areas. Materials and Methods {#s4} ===================== Subjects {#s4a} -------- Twenty-six students from the University of Tübingen participated in the experiment. Seven observers (two for Experiment 1 and five for Experiment 2) were excluded due to either excessive head movement or poor performance in the psychophysical task (accuracy was two standard deviations below the mean). All subjects provided informed consent. Stimuli and procedure {#s4b} --------------------- Thirty-two symmetrical and 32 asymmetrical shapes were rendered with collinear Gabor elements (0.55°) and embedded in backgrounds of randomly-positioned Gabors (0.55°), as described previously \[[@pbio-0030204-b26]\]. Each stimulus covered an area 14.35° × 14.35° (average shape area: 7.72° × 7.78°) and was presented 0.19° to the left or right of fixation. Two types of background were used: (1) randomly positioned and oriented Gabors (low salience) or (2) randomly positioned but uniformly oriented Gabors (high salience). A pilot psychophysical experiment showed that detection of symmetrical and asymmetrical shapes in these stimuli was of similar difficulty. To ensure that subjects learned the shapes and not simply the background configuration, the arrangement of the background elements differed on every trial and the position of the shape target elements was jittered. Each observer was trained on a unique set of four symmetrical and four asymmetrical shapes that were presented on all days of the experiment. Observers performed a two-alternative forced choice (2AFC) task. On every trial, one symmetrical and one asymmetrical stimulus were presented on either side of the fixation point. Observers indicated (by pressing a button) which side of the display contained the symmetrical shape while maintaining central fixation. This 2AFC task was chosen for two reasons: (1) to encourage the observers to compare the two stimuli presented in a trial and improve their performance by learning to discriminate between shapes, and (2) to avoid biases that are observed in detection tasks when a single stimulus is presented in a trial. One possible limitation of this task is that the observers could adopt a strategy in which they only paid attention to one side of the screen. Such a strategy would make it easy for the observers to perform the 2AFC task even before training. However, the observers\' poor performance in the 2AFC task before training suggests that it is unlikely that observers relied on such a strategy.Furthermore, limiting attention to one side of the fixation point could result in hemispheric asymmetries in the fMRI data. However, we did not observe any differences in the pattern of fMRI data between hemispheres. Pre-training scanning session (Day 1) {#s4c} ------------------------------------- On the first day of the experiment, observers performed the 2AFC task without feedback in the scanner. This scanning session consisted for four different event-related runs in which the observers were presented with low-salience (Experiment 1), or high-salience (Experiment 2) trained stimuli (i.e., that were to become trained) and untrained stimuli (i.e., that were to remain untrained). In particular, on each run observers viewed a set of trained (four symmetrical and four asymmetrical) stimuli and a set of untrained (four symmetrical and four asymmetrical) stimuli. Each run consisted of one epoch of experimental trials and two 8-s fixation epochs (one at start, one at end). Each run had 25 experimental trials for each condition (trained and untrained) and 25 fixation trials; that is, a total of 75 trials. A new trial began every 3 s and consisted of a stimulus image presented for 300 ms and a blank interval of 2,700 ms. As in previous studies \[[@pbio-0030204-b26]\], the order of presentation was counterbalanced so that trials from each condition, including the fixation condition, were preceded equally often by trials from each of the other conditions. In total, 100 trials were collected for each experimental condition across the four runs in each scanning session. Psychophysical training sessions (Days 2--4) {#s4d} -------------------------------------------- Observers were trained with four symmetrical and four asymmetrical shapes from their unique trained-stimulus set (one set per observer) in either the low-salience (Experiment 1) or the high-salience (Experiment 2) target condition for three consecutive days. Each session consisted of 400 training trials (total trials across sessions: 1,200) during which error feedback was given. At the end of each training session, observers performed a short test of 96 trials (no feedback) in which trained shapes were intermixed with untrained shapes. As shown in [Figure 8](#pbio-0030204-g008){ref-type="fig"}, the observers\' performance improved across training sessions for the trained, but not the untrained, stimuli, suggesting shape-specific learning rather than general improvement in the 2AFC task. ::: {#pbio-0030204-g008 .fig} Figure 8 ::: {.caption} ###### Behavioral Data during Training Sessions Psychophysical data (percent correct) during the three training sessions. Data are shown for trained and untrained shapes in which the observers were tested without feedback at the end of each training session. Statistical analysis of the data showed that the observers\' performance improved for trained, but not untrained, shapes across training sessions. \(A) In Experiment 1, no significant differences between trained and untrained stimuli were observed for session 1 (F~1,20~ = 1.66 , p = 0.21) but increasing differences were observed for sessions 2 (F~1,20~ = 19.57, p \< 0.001) and 3 (F~1,20~ = 47.79, p \< 0.001). \(B) Similarly, in Experiment 2, a significant effect (F~1,14~ = 27.30, p \< 0.01) of familiarity (trained vs. untrained shapes) was observed across training sessions. ::: ![](pbio.0030204.g008) ::: Post-training scanning session (Day 5) {#s4e} -------------------------------------- Observers were tested in the scanner on the same set of stimuli with which they were presented in the pre-training scanning session. The same procedure was followed as in the pre-training scanning session. Imaging {#s4f} ------- Observers were scanned in a 3T Siemens scanner at the University Clinic in Tübingen. Gradient echo pulse sequences were used (TR = 1 s, TE = 40 ms for event related runs; TR = 2 s, TE = 90 ms for localizer runs. Data were collected with a head coil from eleven axial slices (3 × 3 mm in-plane resolution, 5-mm thickness) that covered the occipitotemporal cortical regions. Data analysis {#s4g} ------------- Psychophysical data were analyzed with repeated measures ANOVA on test session (before training, after training), and familiarity (trained, untrained) for each experiment. Contrast analysis followed significant interactions between these factors. The fMRI data were analyzed with Brain Voyager, as described previously \[[@pbio-0030204-b26]\]. For each individual subject we identified the early visual areas and the LOC as cortical ROIs (see [Figure 2](#pbio-0030204-g002){ref-type="fig"}). The LOC was defined as the set of all voxels in the ventral occipitotemporal cortex that were activated more strongly (p \< 10^−4^) by intact than scrambled images of objects presented in two blocked-design runs \[[@pbio-0030204-b26]\]. Two subregions of the LOC were identified \[[@pbio-0030204-b19]\]: the LO at the posterior part of the inferior-temporal sulcus and the pFs in the posterior fusiform gyrus. Early ventral visual areas were identified using standard retinotopic mapping techniques \[[@pbio-0030204-b26]\]. For each individual subject we extracted time-course data from each ROI and for each condition ([Figures S3](#sg003){ref-type="supplementary-material"} and [S4](#sg004){ref-type="supplementary-material"}). Fitting the time course data with the hemodynamic response function and ANOVA analysis across time points indicated that peak time fMRI responses occurred at 4 and 5 s after trial onset ([Figures S5](#sg005){ref-type="supplementary-material"} and [S6](#sg006){ref-type="supplementary-material"}). For statistical analysis of differences between conditions in the average fMRI responses at these time points we used repeated-measures ANOVA on test session (before training and after training), familiarity (trained and untrained) and ROI (V1, V2, Vp, V4, LO, and pFs). Contrast analysis followed significant interactions between these factors. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### Attentional Control Experiment: Post-Training Test fMRI responses after training obtained when observers performed a target-monitoring task while being presented with either low-salience (A) (three observers) or high-salience (B) (three observers) shapes, as in Experiments 1 and 2. The observers were instructed to press a button when a prespecified target shape appeared in a trial. This task ensured that the observers paid attention across all conditions similarly, as the target\'s appearance was rare (∼15% of trials) but of similar frequency across conditions. Detection of this target was of comparable difficulty for low-salience (79%) and high-salience (81%) shapes. Similar patterns of fMRI data were observed in this control experiment as in Experiments 1 and 2. (257 KB PDF). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### Eye Movement Controls Eye movements of six subjects in Experiment 1 and Experiment 2 were recorded (Eye-Link video based system, 250-Hz sample rate) for the pre-training and post-training sessions. We compared eye position and saccades between trained and untrained shapes in each experiment. In the pre-training session, the average number of saccades (Experiment 1: horizontal F~2,4~ \< 1; p = 0.77, vertical F~2,4~ \< 1; p = 0.45; Experiment 2: horizontal F~2,4~ \< 1; p = 0.43, vertical F~2,4~ \< 1; p = 0.36) and amplitude (Experiment 1: horizontal F~2,4~ \< 1;p = 0.88, vertical F~2,4~ \< 1; p = 0.37; Experiment 2: horizontal F~2,4~ \< 1; p = 0.58, vertical F~2,4~ \< 1; p = 0.67) did not differ between experimental conditions and the fixation condition. Data are shown for the post-training session, in which psychophysical and fMRI differences between trained and untrained stimuli were observed. Panels A--B show that the histograms of the horizontal eye position for each condition and experiment were peaked and centered on the fixation at zero degrees. Similar histograms of the vertical eye position were centered on the fixation but less sharply peaked. (This was probably due to observed drift in the vertical position signal over the course of the recordings). No significant differences were observed in the mean eye position between fixation, trained and untrained conditions for low-salience shapes (x position F~2,4~ = 1.29; p = 0.36, y position F~2,4~ = 3.05; p = 0.15) and high-salience shapes (x position F~2,4~ = 1.23; p = 0.38, y position F~2,4~ \< 1; p = 0.95). Furthermore, the average number of saccades (panel C) was similar in both experiments and did not differ significantly for trained and untrained shapes (low salience F~1,2~ \< 1; p = 0.94; high salience F~1,2~ \< 1; p = 0.70). The amplitude (low salience F~1,2~ \< 1; p = 0.47; high salience F~1,2~ \< 1; p = 0.99) and duration (low salience F~1,2~ \< 1; p = 0.32; high salience F~1,2~ \< 1; p = 0.63) of these saccades did not differ significantly for trained and untrained shapes (panels D--G). (428 KB PDF). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### Time Course of the fMRI Responses I: Experiment 1 These figures illustrate the time course (0--10 s after trial onset) of the fMRI responses for all ROIs in Experiment 1 ([Figure S3](#sg003){ref-type="supplementary-material"}) and Experiment 2 ([Figure S4](#sg004){ref-type="supplementary-material"}) before (A) and after (B) training. Error bars are plus or minus the standard error of the mean (SEM). As previously described \[[@pbio-0030204-b26]\], for each event-related scan, the fMRI responses were extracted by averaging the data from all voxels within each subject\'s ROIs. We averaged the signal intensity across trials in each condition at each time point and converted these to percent signal change relative to fixation. We then averaged each condition\'s time course across scans for each subject and then across subjects. Because of the hemodynamic lag in the fMRI response, the peak in overall response and, therefore, the differences across conditions are expected to occur at a lag of several seconds after stimulus onset \[[@pbio-0030204-b86]-- [@pbio-0030204-b88]\]. In accordance with the hemodynamic response properties, an ANOVA between familiarity (trained, untrained) and time point (0--10 s after trial onset) for each ROI showed statistical differences for time point 4 (e.g., LOC: Experiment 1 (F~1,100~ = 8.28, p \< 0.01), Exp 2 (F~1,70~ = 5.31, p \< 0.05)), and time point 5 (e.g., LOC: Experiment 1 (F~1,100~ = 19.75, p \< 0.001), Experiment 2 (F~1,70~ = 6.01, p \< 0.05)), but not at trial onset, i.e., time point zero (e.g., LOC: Experiment 1 (F~1,100~ \< 1, p = 0.59), Experiment 2 (F~1,70~ = 2.64, p = 0.14)). Results were similar in the other ROIs in that no significant differences were observed at trial onset in early visual areas for low-salience (Experiment 1: F~1,100~ = 2.23, p = 0.15) or high-salience (Experiment 2: F~1,70~ = 1.63, p = 0.23) shapes. (711 KB PDF). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S4 ::: {.caption} ###### Time Course of the fMRI Responses II: Experiment 2 These figures illustrate the time course (0--10 s after trial onset) of the fMRI responses for all ROIs in Experiment 1 ([Figure S3](#sg003){ref-type="supplementary-material"}) and Experiment 2 ([Figure S4](#sg004){ref-type="supplementary-material"}) before (A) and after (B) training. Error bars are plus or minus the standard error of the mean (SEM). As previously described \[[@pbio-0030204-b26]\], for each event-related scan, the fMRI responses were extracted by averaging the data from all voxels within each subject\'s ROIs. We averaged the signal intensity across trials in each condition at each time point and converted these to percent signal change relative to fixation. We then averaged each condition\'s time course across scans for each subject and then across subjects. Because of the hemodynamic lag in the fMRI response, the peak in overall response and, therefore, the differences across conditions are expected to occur at a lag of several seconds after stimulus onset \[[@pbio-0030204-b86]-- [@pbio-0030204-b88]\]. In accordance with the hemodynamic response properties, an ANOVA between familiarity (trained, untrained) and time point (0--10 s after trial onset) for each ROI showed statistical differences for time point 4 (e.g., LOC: Experiment 1 (F~1,100~ = 8.28, p \< 0.01), Exp 2 (F~1,70~ = 5.31, p \< 0.05)), and time point 5 (e.g., LOC: Experiment 1 (F~1,100~ = 19.75, p \< 0.001), Experiment 2 (F~1,70~ = 6.01, p \< 0.05)), but not at trial onset, i.e., time point zero (e.g., LOC: Experiment 1 (F~1,100~ \< 1, p = 0.59), Experiment 2 (F~1,70~ = 2.64, p = 0.14)). Results were similar in the other ROIs in that no significant differences were observed at trial onset in early visual areas for low-salience (Experiment 1: F~1,100~ = 2.23, p = 0.15) or high-salience (Experiment 2: F~1,70~ = 1.63, p = 0.23) shapes. (711 KB PDF). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S5 ::: {.caption} ###### Fits to the Time Course Data I: Experiment 1 To confirm the peak points obtained from ANOVA analysis, we fit the data using two Gaussians (one for the initial response and one for the undershoot) and a baseline: from Kruggel and von Cramon \[[@pbio-0030204-b89]\] where the hemodynamic response function (h) over time (t) is modeled as the sum of two Gaussians, each of which depends on gain (γ), dispersion (δ), a temporal lag (λ), and a baseline parameter (k). Fits for the fMRI responses across all areas for trained and untrained stimuli are shown for Experiment 1 ([Figure S5](#sg005){ref-type="supplementary-material"}) and Experiment 2 ([Figure S6](#sg006){ref-type="supplementary-material"}) before (A) and after (B) training. These fits showed the following peak time points for each condition: Experiment 1 (before training: trained: 4.25, untrained: 4.23; after training: trained: 4.27, untrained: 4.24); Experiment 2 (before training: trained: 4.24, untrained: 4.29; after training: trained: 4.23, untrained: 4.21). This analysis confirmed the selection of time points 4 and 5 as the peak points of the fMRI time courses. Therefore the average response at these peak points was taken as the measure of response magnitude for each condition in subsequent analyses. Analysis of time points 2--6 s after stimulus onset or the area under the curve showed the same pattern of results as reported in the paper. (363 KB PDF). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S6 ::: {.caption} ###### Fits to the Time Course Data II: Experiment 2 To confirm the peak points obtained from ANOVA analysis, we fit the data using two Gaussians (one for the initial response and one for the undershoot) and a baseline: from Kruggel and von Cramon \[[@pbio-0030204-b89]\] where the hemodynamic response function (h) over time (t) is modeled as the sum of two Gaussians, each of which depends on gain (γ), dispersion (δ), a temporal lag (λ), and a baseline parameter (k). Fits for the fMRI responses across all areas for trained and untrained stimuli are shown for Experiment 1 ([Figure S5](#sg005){ref-type="supplementary-material"}) and Experiment 2 ([Figure S6](#sg006){ref-type="supplementary-material"}) before (A) and after (B) training. These fits showed the following peak time points for each condition: Experiment 1 (before training: trained: 4.25, untrained: 4.23; After training: trained: 4.27, untrained: 4.24); Experiment 2 (Before training: trained: 4.24, untrained: 4.29; after training: trained: 4.23, untrained: 4.21). This analysis confirmed the selection of time points 4 and 5 as the peak points of the fMRI time courses. Therefore the average response at these peak points was taken as the measure of response magnitude for each condition in subsequent analyses. Analysis of time points 2--6 s after stimulus onset or the area under the curve showed the same pattern of results as reported in the paper. (366 KB PDF). ::: ::: {.caption} ###### Click here for additional data file. ::: We would like to thank M. Kleiner and L. Montaser Kouhsari for their help with stimulus design. We would also like to thank M. Giese and M. Sigman for helpful comments and discussions. This work was supported by the Max Planck Society and a Deutsche Forschungsgemeinschaft grant (TH 812/1-1). Competing interests. The authors have declared that no competing interests exist. Author contributions. ZK conceived and designed the experiments. ZK, LRB, and PS performed the experiments. ZK, LRB, PS, and AEW analyzed the data. AEW contributed reagents/materials/analysis tools. ZK and AEW wrote the paper. Citation: Kourtzi Z, Betts LR, Sarkheil P, Welchman AE (2005) Distributed neural plasticity for shape learning in the human visual cortex. PLoS Biol 3(7): e204. 2AFC : two-alternative forced choice fMRI : functional magnetic resonance imaging LO : lateral occipital LOC : lateral occipital complex pFs : posterior fusiform sulcus ROI : region of interest SEM : standard error of the mean	PubMed Central	2024-06-05T03:55:57.551234	2005-6-7	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1150289/", "journal": "PLoS Biol. 2005 Jul 7; 3(7):e204", "authors": [ { "first": "Zoe", "last": "Kourtzi" }, { "first": "Lisa R", "last": "Betts" }, { "first": "Pegah", "last": "Sarkheil" }, { "first": "Andrew E", "last": "Welchman" } ] }
PMC1150290	Introduction {#s1} ============ It has long been known in the literature on imprinting \[[@pbio-0030208-b01],[@pbio-0030208-b02]\], and indeed in studies of mammals, including human infants \[[@pbio-0030208-b03]\], that moving objects are more likely to evoke a response than are stationary objects. What is unknown, however, is whether learning plays a part in the formation of this preference. Consider the case of filial imprinting. By looking at the ethological literature (review in \[[@pbio-0030208-b04]\]), one finds the general assertion that object motion facilitates imprinting. However, no one has checked whether all types of motion are identically effective or if animals are especially sensitive to particular types of motion. The problem, of course, is that it is difficult to disentangle the stationary visual characteristics of an object (its shape, texture, colour, and brightness) from the dynamic aspects (its motion). We used point-light displays to solve the problem. When a biological creature, such as a hen, travels about its environment, its limbs and torso move in characteristic synchrony. Johansson \[[@pbio-0030208-b05]\] first noted that an animation sequence consisting of just a few strategically positioned points of light is sufficient to create the impression in a human subject of an organism engaged in coordinated activity, such as walking. This ability to perceive biological motion has been extensively investigated, even from the perspective of development \[[@pbio-0030208-b06]\] and neurobiology \[[@pbio-0030208-b12]--[@pbio-0030208-b14]\]. Using conditioning procedures, several animal species have been shown to be able to discriminate between different point-light animation sequences \[[@pbio-0030208-b15]\]. Taking advantage of the learning process associated with the phenomenon of filial imprinting, Regolin et al. \[[@pbio-0030208-b19]\] exposed day-old domestic chicks to point-light animation sequences depicting either a walking hen or a rotating cylinder; on a subsequent free-choice test, the chicks approached the novel stimulus, irrespective of whether it was the hen or the cylinder sequence. This demonstrates that chicks, similar to other avian and mammalian species, can discriminate between point-light animation sequences. However, this tells us nothing about any possible natural predisposition of the animals to attend preferentially to biological motion stimuli. We tested naive, newly hatched chicks, lacking any previous visual experience, to investigate whether they showed a spontaneous preference to approach stimuli depicting biological rather than non-biological motion. The first point-light sequence represented a "walking hen" (13 points of light located on the digitalization of the video recording of a real animal; see [Figure 1](#pbio-0030208-g001){ref-type="fig"}A and [1](#pbio-0030208-g001){ref-type="fig"}B; see also [Video S1](#sv001){ref-type="supplementary-material"}, which reproduces a version of the original walking hen stimulus). Three other sequences were used as "foil sequences." ::: {#pbio-0030208-g001 .fig} Figure 1 ::: {.caption} ###### Point-Light Displays and Sample Frames from the Animation Sequences \(A) The walking hen point-light display (above) The filled circles indicate the location of each point of light. \(B) Six frames sampled from the walking hen animation (below) \(C) The walking cat point-light display (above). \(D) Six frames sampled from the walking cat animation (below). ::: ![](pbio.0030208.g001) ::: \(1) "Rigid motion" (see [Video S2](#sv002){ref-type="supplementary-material"} for a clip of this animation). To produce this sequence, a single frame (made of 13 points of light) from the walking hen animation sequence was randomly selected and was moved rigidly about the vertical axis so as to produce the motion of a rotating, rigid hen-like object; \(2) "Random motion" sequence (see [Video S3](#sv003){ref-type="supplementary-material"} for a clip of this animation). In this sequence, the same set of 13 points of light described and used for each display moved now in arbitrary directions (see [Materials and Methods](#s3){ref-type="sec"} for details about how this display was obtained). \(3) "Scrambled hen" sequence (see [Video S4](#sv004){ref-type="supplementary-material"} for a clip of this animation). It consisted of the same set of points of light as the walking hen and the same set of frames, only now the original position of each point was spatially displaced a fixed amount throughout the animation (see [Materials and Methods](#s3){ref-type="sec"} for more details). Every single point of light, although displaced from its original position in the walking hen animation, moved identically in this sequence to that of the walking hen. As a result, this last display no longer conveyed the perception of a hen to human observers, though it retained the appearance of biological motion of some kind of unidentified creature. Results/Discussion {#s2} ================== Each chick underwent a 6-min free-choice test between two different displays in a standard runway apparatus ([Figure 2](#pbio-0030208-g002){ref-type="fig"}). When presented with the walking hen and the rigid motion sequences in a free-choice test, chicks preferred to approach and stay close to the walking hen animation sequence; the same occurred when the walking hen was paired with the random motion sequence ([Figure 3](#pbio-0030208-g003){ref-type="fig"}A). On the other hand, no preferences emerged between the walking hen and the scrambled hen sequences ([Figure 3](#pbio-0030208-g003){ref-type="fig"}A). Analysis of variance (ANOVA) revealed a significant overall heterogeneity (F~2,279~ = 5.438, p \< 0.005). Paired comparisons by Scheffé test revealed significant differences between the walking hen versus the rigid motion and the walking hen versus the scrambled hen conditions (p \< 0.02), and between the walking hen versus random motion and the walking hen versus scrambled hen conditions (p \< 0.02). ::: {#pbio-0030208-g002 .fig} Figure 2 ::: {.caption} ###### Schematic Representation of the Test Apparatus ::: ![](pbio.0030208.g002) ::: ::: {#pbio-0030208-g003 .fig} Figure 3 ::: {.caption} ###### Point-Light Sequence Preferences \(A) Preferences (group means and the standard error of the mean) estimated as the percentage of time spent close to the walking hen. \(B) Preferences are shown as the percentage of time spent close to the scrambled hen. \(C) Preferences are shown as the percentage of time spent close to the walking cat. Asterisks indicate significant departures from chance level (i.e., 50%) estimated by one-sample two-tailed t tests (\*p \< 0.05; \\p \< 0.01; \\\*p \< 0.001). ::: ![](pbio.0030208.g003) ::: As shown in [Figure 3](#pbio-0030208-g003){ref-type="fig"}B, the scrambled hen sequence was compared with the rigid and the random motion sequences (preferences are shown as percentage of time spent close to the scrambled hen). ANOVA did not reveal any difference between the two testing conditions (F~1,191~ = 2.239, p = 0.136). The scrambled hen was clearly preferred to both rigid and random motion ([Figure 3](#pbio-0030208-g003){ref-type="fig"}B). The results show that naive chicks exhibit clear and consistent preferences in approaching certain types of movements. Intriguingly, however, chicks\' choices seemed to reflect a generic preference for the patterns of biological motion rather than a specific preference for the typical form of the hen motion. The walking hen sequence was chosen as often as the scrambled hen ([Figure 3](#pbio-0030208-g003){ref-type="fig"}A); and both the walking hen ([Figure 3](#pbio-0030208-g003){ref-type="fig"}A) and the scrambled hen ([Figure 3](#pbio-0030208-g003){ref-type="fig"}B) were preferred to the rigid and random motion sequences. These findings suggest that chicks preferentially approach semi-rigid motion, the type of motion that is exhibited by vertebrate animals. In semi-rigid motion, some points maintain a fixed distance from each other (e.g., two points placed close on the same limb), but can nonetheless vary their distance with respect to other points (e.g., with respect to points located on the torso). Such a pattern of semi-rigid motion is shared by the walking and the scrambled hen sequences, even though the latter does not match any existing biological creature. As a control for this hypothesis, we used the motion of a cat (see [Figure 1](#pbio-0030208-g001){ref-type="fig"}C and [1](#pbio-0030208-g001){ref-type="fig"}D; see also [Video S5](#sv005){ref-type="supplementary-material"} for a clip of the cat animation), a species that can predate on young chicks. (This animation was obtained from the video recording of a real cat, following the procedure described for obtaining the walking hen animation). As predicted, chicks did not exhibit any preference between the walking hen and a walking cat point-light sequence, though they did prefer the walking cat to the random and the rigid motion sequences ([Figure 3](#pbio-0030208-g003){ref-type="fig"}C). ANOVA revealed a significant overall heterogeneity (F~2,279~ =5.644, p = 0.004). Paired comparisons by Scheffé test revealed significant differences between the walking cat versus walking hen and the walking cat versus rigid motion conditions (p \< 0.05), and between the walking cat versus walking hen and the walking cat versus random motion conditions (p \< 0.05). Conclusion {#s2a} ---------- It is known that, as a result of exposure to a particular object early in life, many species of birds and mammals will form a strong and exclusive attachment to that object, a process dubbed "filial imprinting" \[[@pbio-0030208-b02],[@pbio-0030208-b20]\] (see also \[[@pbio-0030208-b24]\] for a discussion on the recent use of imprinting in order to investigate cognitive mechanisms in a comparative perspective). Motion of the object is known to facilitate the learning process \[[@pbio-0030208-b01],[@pbio-0030208-b25],[@pbio-0030208-b26]\]. However, it was not known whether any type of motion would be equally effective in eliciting approach or if specific predispositions exist for the type of motion that is most likely encountered in an animal\'s natural social environment. We found that visually inexperienced, newly hatched chicks, reared and hatched in darkness, at their first exposure to point-light animation sequences exhibit a spontaneous preference to approach biological motion patterns. It is likely that such a predisposition would affect the type of stimulus on which the animal is more likely to imprint on in a natural environment. Intriguingly, the preference was not specific for the motion of a hen, but extended to the pattern of motion of other vertebrates, even to that of a potential predator, such as a cat. The predisposition found in the present research for certain kinds of movements shares characteristics in common with the predisposition for aspects of form demonstrated earlier: Visually inexperienced chicks prefer the head and neck region of a hen to artificial objects \[[@pbio-0030208-b27]\]. Similar to this preference for form, the preference for movement is not species specific. Evolution seems to have equipped the visually inexperienced bird with a sophisticated set of detection systems (see \[[@pbio-0030208-b28]\] for an extension of this argument to the human species). The evidence of predispositions in the young chick for head and neck regions has stimulated a substantial body of work of a similar kind in our own species, concerning face recognition in the human infant (e.g., \[[@pbio-0030208-b29]\]). When considered together with our observations, these findings seem to fit a general scheme for cognitive development of recognition of the mother based on the interaction between two separate and independent systems \[[@pbio-0030208-b03],[@pbio-0030208-b27],[@pbio-0030208-b32]\]. The first of these systems directs the attention of the young animal toward the appropriate class of objects to learn about, in the absence of any prior specific experience (e.g., in the case of motion, toward those objects that move semi-rigidly). The second system is concerned with learning about the peculiar characteristics of the objects to which attention has been directed by the first system. Given that in a natural environment it is more likely that the newly hatched chick would encounter a mother hen rather than a cat, a developing predisposition to pay attention to objects showing the characteristic motion of vertebrates would assure highest probability to learn (by way of the imprinting mechanism) about the specific pattern of motion of the mother hen. The perception of biological motion has been hypothesized to be an intrinsic capacity of the vertebrate visual system \[[@pbio-0030208-b05]\]. However, the evidence obtained so far in the human species is inconclusive: Human infants exhibit a preference for biological motion patterns starting from about 4--6 months of age \[[@pbio-0030208-b35]\], and this can be accounted for by both innate (maturational) and learning mechanisms. Our results with newly hatched chicks suggest that a preference for biological motion may be predisposed in the brain of vertebrates. Materials and Methods {#s3} ===================== {#s3a} Eggs were incubated (using a MG 70/100 incubator) and hatched in total darkness in our laboratory. Overall, a number of 765 newly hatched chicks underwent the experiment. Each chick was tested once only for its spontaneous preference between two animation patterns: A set of 100 chicks was tested in the scrambled hen versus rigid motion comparison; we tested a set of 95 chicks for each of the other seven comparisons we investigated (i.e., walking hen versus either the random motion, solid motion, or scrambled hen sequence; scrambled hen versus the random motion sequence; and walking cat versus either the walking hen, rigid motion. or random motion sequence). Two hours after hatching, each chick was taken from the hatchery and placed in a dark room, on a treadmill (3.7 × 10^−3^ m/s) for 30 min. Previous work \[[@pbio-0030208-b27],[@pbio-0030208-b32]\] has shown that such motor activity is crucial for the development of innate predispositions in the chick. Thereafter, each chick was placed in the test apparatus, a runway measuring 80 × 20 × 20 cm (see [Figure 2](#pbio-0030208-g002){ref-type="fig"}). At each end of the runway, a different point-light motion display was presented. The ends of the runway consisted of a transparent glass sheet (not shown in the figure,) making it visible at each end a computer screen (located 16 cm away) on which one of the two stimuli to be compared was presented. For the purposes of this study each chick underwent the test once only (see Ethical Considerations below). The test lasted 6 min, during which time each bird could freely approach and stay by either stimulus. Using a computerized event recorder, we scored the time (in seconds) spent by each chick in either of the two 30-cm long compartments that were closer to one or the other of the two stimuli. Such raw data were thereafter computed as the overall time spent by the biological stimulus divided by the overall time spent by both the biological stimulus and the comparison stimulus combined. When the comparison was between the walking and the scrambled hen, the walking hen was arbitrarily chosen as the "biological stimulus." Similarly, when the comparison involved the walking hen and the walking cat, the latter was arbitrarily chosen as the "biological stimulus." Data were analyzed by ANOVA for differences between stimulus conditions; significant departures from chance level (50%) were estimated by one sample two-tailed t tests. All animation sequences (see [Figure 1](#pbio-0030208-g001){ref-type="fig"}A--D) were obtained with the use of the software program Macromedia Director (Version 6.0) and consisted of sets of 13 bright dots (95.71 candelas \[cd\]/m^2^) seen against a black background (0.03 cd/m^2^). Each dot was made by four pixels on a 640 × 480 pixel resolution screen; the actual visual angle measured 0° 21′ 29″ at a viewing distance of 16 cm. Animation sequences were matched for average velocity (54 pixels/s) of each of the 13 dots. Each set of points of light occupied a window of 119 × 108 pixels on the centre of the computer screen; the actual visual angle of the window measured 16° 2′ 23″ (height) and 17° 40′ 46″ (width) at a viewing distance of 16 cm. The walking hen animation was obtained by carefully locating, frame by frame, each of the 13 points of light on the main joints of the digitalized image of the video recording of a real animal. (The same procedure was also used to produce the walking cat animation.) Twenty-three frames were required to cover an animal\'s entire step sequence, then the digitalized sequence was looped and projected onto a computer screen after subtraction of translation components. As a result, the display was stationary in the central window of the screen described above, but moved as if the hen was walking on a treadmill. All the other foil sequences were also produced by looping a 23-frame animation. The scrambled hen display was obtained by consistently displacing each point of light in each frame of the walking hen sequence by 1 cm (i.e., by a visual angle of 3° 34′ 34″ at a viewing distance of 16 cm). Each point could be displaced either up, down, right, or left, at random. Although displaced compared to its position in the walking hen display, each single point of light in the scrambled hen animation retained the same motion characteristics (i.e., the same trajectory and velocity) exhibited by that point in the walking hen. As a result only the reciprocal positions of the 13 points of light differed between the walking and the scrambled hen animations. The scrambled hen display even occupied the same window on the screen as the walking hen. The random motion display was obtained through the function "random movement and rotation" of the software program Macromedia Director MX (Version 9.0). The overall characteristics of the motion matched those portrayed in the walking hen sequence in the sense that each of the 13 points of light was associated with a different velocity, corresponding to the average velocity of each of the 13 points of light of the hen animation. Moreover the points of light in this display could move randomly within a 119 × 108 pixel window (corresponding to the area of the walking hen display); within this window, the points of light could cross each others\' trajectories (which were not linear in principle of course, but being randomly determined, could assume a linear fashion for some time) and even overlap, but once they reached the edge of the defined window, they would not disappear but rather would turn around and head back. The random display, although comprising the same number of frames as the other displays, was not obtained by looping a fixed sequence of 23 frames, hence the movement in itself kept varying throughout the 6 min of presentation. Stimuli were presented on two identical 13.8\" Macintosh CRT screens with a refresh rate of 117 Hz. Apart from the light arising from the monitor screens, the room was maintained in complete darkness. (This, together with the high refresh rate of the screens, was aimed at preventing any flicker detection by the chicks). ### Ethical considerations {#s3a1} All of the experiments reported comply with current Italian and European laws on the ethical treatment of animals, all experimental procedures have been licensed by the responsible office of the Italian Government (Ministero della Salute--Dipartimento Alimenti, Nutrizione e Sanità Pubblica Veterinaria), and the present project has been classified as purely behavioural testing, involving no distress or discomfort to the animals at all. Moreover, all of the chicks that entered the experiment were, after the 6-min behavioural observations, immediately caged in social groups with food and water available ad libitum and, on the second day, were donated to local farmers who provided them with free-range conditions, as approved by our Animal House licence for observational experiments on chicks. Supporting Information {#s4} ====================== Video S1 ::: {.caption} ###### The Walking Hen A sample video clip of the animation employed as the walking hen stimulus. The hen is walking leftwards. This demonstration does not retain the quality of the original stimuli which were obtained in a different format. (549 KB AVI). ::: ::: {.caption} ###### Click here for additional data file. ::: Video S2 ::: {.caption} ###### The Rotating Solid The first frame of the walking hen was treated as a solid object and rotated rigidly anticlockwise. (13 KB AVI). ::: ::: {.caption} ###### Click here for additional data file. ::: Video S3 ::: {.caption} ###### The Random Motion A sample sequence of the stimulus employed as random motion. More details on how this stimulus was obtained can be found in the text. (13 KB AVI). ::: ::: {.caption} ###### Click here for additional data file. ::: Video S4 ::: {.caption} ###### The Scrambled Hen The scrambled hen animation was obtained by displacing the positions of the dots of the walking hen. More information about how this was obtained can be found the text. Such manipulation results in a motion that is still perceived as biological, although it does not belong to any particular known animal. (15 KB AVI). ::: ::: {.caption} ###### Click here for additional data file. ::: Video S5 ::: {.caption} ###### The Walking Cat A sample video clip of the animation employed as walking cat stimulus. The cat is heading to the left. (14 KB AVI). ::: ::: {.caption} ###### Click here for additional data file. ::: We thank Patrick P.G. Bateson, Gabriel Horn, Jacques Mehler, Lesley J. Rogers, Elizabeth S. Spelke, and Carlo A. Umiltà for reading and commenting on the manuscript. The research was supported by grants from the Ministero dell\'Università e della Ricerca Scientifica MIUR Cofin 2004, 2004070353\_002 "Intel-lat" and Ministero per le Politiche Agricole e Forestali MIPAF "Ben-o-lat" to GV via Dip. Sci. Zootecniche, Univ. Sassari. Competing interests. The authors have declared that no competing interests exist. Author contributions. GV and LR conceived and designed the experiments. FM performed the experiments. GV, LR, and FM analyzed the data. FM contributed reagents/materials/analysis tools. GV and LR wrote the paper. Citation: Vallortigara G, Regolin L, Marconato F (2005) Visually inexperienced chicks exhibit spontaneous preference for biological motion patterns. PLoS Biol 3(7): e208. ANOVA : analysis of variance cd : candela	PubMed Central	2024-06-05T03:55:57.555233	2005-6-7	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1150290/", "journal": "PLoS Biol. 2005 Jul 7; 3(7):e208", "authors": [ { "first": "Giorgio", "last": "Vallortigara" }, { "first": "Lucia", "last": "Regolin" }, { "first": "Fabio", "last": "Marconato" } ] }
PMC1150291	Introduction {#s1} ============ As changing global climate and anthropogenic modification of the biosphere increasingly threaten global biodiversity \[[@pbio-0030222-b01], [@pbio-0030222-b02]\], accurately predicting extinctions takes on new urgency \[[@pbio-0030222-b03]\]. Although stochastic models of population growth and decline are commonly used to forecast extinction, quantitative model predictions for population dynamics in randomly fluctuating (i.e., natural) environments have only been poorly supported by experiments \[[@pbio-0030222-b04], [@pbio-0030222-b06]\]. In contrast, validation with observational field data \[[@pbio-0030222-b05], [@pbio-0030222-b08]\] is beset by numerous difficulties, including low precision \[[@pbio-0030222-b11]\], low replication \[[@pbio-0030222-b12]\], small sample size \[[@pbio-0030222-b12], [@pbio-0030222-b13]\], poor data quality \[[@pbio-0030222-b12]\], and lack of independence \[[@pbio-0030222-b12], [@pbio-0030222-b14]\]. Without empirical validation, viability analyses for threatened and endangered species are subject to doubt, and identifying risk factors for population extinction is difficult. The results reported here support the further use of stochastic models for predicting extinction, but only where detailed information about variation in individual fitness is available for estimating the structure and magnitude of demographic stochasticity. Two components contribute to random variation in per capita population growth rates: demographic stochasticity and environmental variability \[[@pbio-0030222-b15]\]. Here, demographic stochasticity refers to variation in individual fitness, sampling effects in finite populations, and chance events that affect individuals independently \[[@pbio-0030222-b18]\]. Variation in environmental conditions can cause expected individual fitness to fluctuate, resulting in environmental stochasticity. Defining ΔN = N ~t+1~ − N ~t~, a model for per capita population growth rate with demographic and environmental stochasticity is where λ is the (possibly density-dependent) expected multiplicative per capita population growth rate depending on population size N in the previous time step and parameters ψ, W is a normal unit variate, and σ ~e~ (N) and σ ~d~ (N) are models for the (possibly density-dependent) effects of environmental and demographic stochasticity, respectively \[[@pbio-0030222-b18]\]. This is a general model, which can approximate or has as its special cases many common stochastic population growth models, including linear and nonlinear stochastic differential equations \[[@pbio-0030222-b21], [@pbio-0030222-b22]\], birth-death processes \[[@pbio-0030222-b17], [@pbio-0030222-b23], [@pbio-0030222-b24]\], and discrete and continuous-time branching process models \[[@pbio-0030222-b25]\], although it is unclear if these approximations reliably represent the dynamics of small populations \[[@pbio-0030222-b26]\]. Recent research suggests that demographic stochasticity might have a density-dependent component in the sense that the variance contributed to the population\'s growth rate by demographic stochasticity, represented by σ ~d~ in [Equation 1](#pbio-0030222-e001){ref-type="disp-formula"}, is a function of population size \[[@pbio-0030222-b18], [@pbio-0030222-b19], [@pbio-0030222-b27]\]. This phenomenon is independent of the well-known scaling of demographic stochasticity with population size, represented by the ratio σ ~d~ (N)/N in the model \[[@pbio-0030222-b15], [@pbio-0030222-b16], [@pbio-0030222-b18], [@pbio-0030222-b25]\]. While theory predicts that density-dependent demographic variation will occur in all populations in which vital rates are density-dependent \[[@pbio-0030222-b19]\], and will therefore be ubiquitous in both natural and experimental populations, the severity of density dependence in demographic variation has rarely been investigated, and thus rarely measured \[[@pbio-0030222-b18], [@pbio-0030222-b27]\]. Extinction forecasts that ignore density-dependent demographic stochasticity in populations in which it is present will be biased, possibly severely. To test for effects of environmental variation on persistence, 281 laboratory populations of water fleas (Daphnia magna) were maintained under experimentally controlled conditions for 104 d using three levels of variation in food availability as a source of environmental stochasticity \[[@pbio-0030222-b28]\]. Here, those data are used to select and test stochastic models of population dynamics. To ensure independence, the data were divided into subsets for model fitting and model testing. Then, as an exploratory analysis, deviations were plotted from expected population growth against population size for populations in the low-variability treatment to suggest how demographic variation might be structured. On the basis of this analysis, a set of models was developed for population growth in fluctuating and constant environments to test for effects of environmental variation. Finally, population growth trajectories with (i) density-dependent demographic stochasticity, and (ii) density-independent demographic stochasticity were simulated to obtain model predicted extinction rates. Simulated extinctions for the model with density-dependent demographic stochasticity were consistent with observed rates of extinction in the test dataset or were only slightly biased, while the model with density-independent demographic stochasticity considerably underestimated observed extinctions. These results suggest that biased estimates of extinction risk resulting from ignoring density-dependent demographic stochasticity may lead to unwarranted optimism concerning the eventual fates of endangered populations. Results {#s2} ======= Exploratory Graphical Analysis {#s2a} ------------------------------ After rescaling to isolate density dependence in σ ~d~ (see [Materials and Methods](#s4){ref-type="sec"}), the scatter plot of observed deviations from expected population growth rate suggested that variation from demographic stochasticity was dependent on population size in populations in the low-variability treatment ( [Figure 1](#pbio-0030222-g001){ref-type="fig"}). Nonparametric tests for correlation confirmed that the negative relationship between log~e~-transformed deviations and population size was highly significant (Spearman rank-order correlation: ρ = −0.23, p \< 0.0001; Kendall\'s τ: ρ = −0.17, p \< 0.0001; N = 342). The average deviation was fit to an exponential model (see [Materials and Methods](#s4){ref-type="sec"}) ( [Table 1](#pbio.0030222.t001){ref-type="table"}). ::: {#pbio-0030222-g001 .fig} Figure 1 ::: {.caption} ###### Demographic Stochasticity Is Strongly Density-Dependent in Experimental Populations of D. magna Deviations from expected population size were rescaled by multiplying the observed deviation by initial population size for the interval to isolate density dependence in σ ~d~ (see [Materials and Methods](#s4){ref-type="sec"}). Rescaled deviations are strongly dependent on population size ( p \< 0.0001). Because observed deviations overlap, obscuring the pattern, points have been jittered in the dimension of the x-axis by the addition of a small amount of random noise. ::: ![](pbio.0030222.g001) ::: ::: {#pbio.0030222.t001 .table-wrap} Table 1 ::: {.caption} ###### AIC Scores for Ricker Models of Population Growth ::: ![](pbio.0030222.t001) ::: Estimation of Extinction Rates in Experimental Populations {#s2b} ---------------------------------------------------------- Extinction rates of populations in experimental treatments with low and medium levels of variation were not significantly different, although these were different from the extinction rate of populations subject to a high level of environmental variation (see [Materials and Methods](#s4){ref-type="sec"}). Consequently, observations in the testing dataset from the low and medium variability treatments were pooled for remaining analyses, resulting in a more conservative test of model-predicted extinction rates. The maximum likelihood estimate of extinction rate for populations from the low- and medium-variability treatments in the model-testing dataset was 48.4% (95% confidence interval \[CI\], 38.0%--58.9%; [Figure 2](#pbio-0030222-g002){ref-type="fig"}). The maximum likelihood estimate of extinction rate in the high-variability treatment was 70.2% (95% CI, 55.1%--82.7%; [Figure 2](#pbio-0030222-g002){ref-type="fig"}). ::: {#pbio-0030222-g002 .fig} Figure 2 ::: {.caption} ###### Simple Models Accurately Predict Extinction in Experimental Populations Estimates of the extinction rate in populations of D. magna reserved for model testing at three levels of environmental variation were obtained from the likelihood function of the binomial distribution (crosses, 95% CI). Because there was no difference between extinction rates in the low- and medium-variability treatments (see [Materials and Methods](#s4){ref-type="sec"}), data were pooled to obtain a more precise estimate, resulting in a more conservative test (right of dashed line). Model-predicted estimates of extinction rate obtained from models of density-dependent population growth with density-dependent demographic stochasticity fit with independent data (triangles) accurately predicted extinction of populations in the low- and medium-variability treatments, but not the high-variability treatment. The addition of environmental stochasticity (square) improved the prediction, although the chance of obtaining the observed 33 extinctions (or more) out of 47 populations was only 3.3%. The standard model with constant demographic stochasticity (stars) fails to predict extinction in all treatments. ::: ![](pbio.0030222.g002) ::: Model Selection {#s2c} --------------- Model-predicted extinction rates were obtained by fitting a simple Ricker model for population growth, E\[λ\] = λ ~0~ e^−bN^, to data in the model-fitting dataset, although a θ-Ricker model was also considered (see [Materials and Methods](#s4){ref-type="sec"}). Moderate to considerable support was obtained for models with only demographic stochasticity compared to models with both demographic and environmental stochasticity, according to Akaike\'s information criterion (AIC; [Table 1](#pbio.0030222.t001){ref-type="table"}). However, there was overwhelming support for a model with density-dependent demographic stochasticity compared to a model with density-independent demographic stochasticity for populations in the low- and medium-variability treatments (ΔAIC = 37.0) and considerable support for populations in the high-variability treatment (ΔAIC = 4.2). These results confirm that density dependence in demographic variance strongly influenced realized population growth in experimental populations. Accuracy of Model-Predicted Extinction Rate {#s2d} ------------------------------------------- The accuracy of model predictions can be assessed by comparing model-predicted extinction rates with the 95% CI for the estimated extinction rate of populations in the model-testing half of the experimental dataset (see [Materials and Methods](#s4){ref-type="sec"}). Overall, models with density-dependent demographic stochasticity accurately predicted population extinction rates within estimable accuracy for populations with low and medium levels of environmental variation, while predictions for populations with a high level of environmental variation were slightly biased ( [Figure 2](#pbio-0030222-g002){ref-type="fig"}). Models which incorrectly assumed density-independent demographic stochasticity were severely biased and underpredicted observed extinction rates ( [Figure 2](#pbio-0030222-g002){ref-type="fig"}). Discussion {#s3} ========== A stochastic Ricker model of population growth with density-dependent demographic stochasticity accurately predicted the chance of extinction, within the power of this experiment to reject the null hypothesis of a difference, or was only slightly biased ([Figure 2](#pbio-0030222-g002){ref-type="fig"}). Adding environmental stochasticity to the model for populations in the high-variability treatment increased the predicted chance of extinction, consistent with current theory \[[@pbio-0030222-b15]\] and previous experiments \[[@pbio-0030222-b04]\]. A model in which demographic stochasticity was assumed to be constant failed to predict extinction rates in all experimental treatments, implying that independent data on the structure of individual variation in vital rates, including information about density dependence, is required to reliably forecast extinction. In this analysis, the accuracy of model predictions was improved by relaxing the usual assumption that demographic stochasticity is density-independent \[[@pbio-0030222-b21], [@pbio-0030222-b25]\]. Since individual fecundity is commonly altered in response to population size or density \[[@pbio-0030222-b29], [@pbio-0030222-b30]\], density-dependent demographic stochasticity resulting from demographic covariation is probably not exceptional in natural populations. Indeed, density-dependent demographic stochasticity will result from any density dependence in vital rates or interactions among individuals that affect demography \[[@pbio-0030222-b19]\]. Presently, this aspect of population biology is poorly understood (compare \[[@pbio-0030222-b18], [@pbio-0030222-b27]\]). Additionally, the accuracy and precision of these predictions were facilitated by a relatively large, high-quality dataset (n ~low~ = 342, n ~med~ = 300, and n ~high~ = 280). In general, field data will not be so abundant. Thus, an important goal for population biology is to develop methods for obtaining reliable predictions from sparse, low-quality datasets \[[@pbio-0030222-b31]\]. Planning for increasing threats to rare species from diverse sources, including climate change, resource extraction, habitat modification, and invasive species will require greater and more precise estimates of extinction risk than ever before. While the reliability of theoretical models for predicting extinction in natural ecosystems remains to be established, the results presented here show that accurate predictions of population extinction in variable environments are indeed possible. Materials and Methods {#s4} ===================== Experiment {#s4a} ---------- Experimental microcosms (n = 281) of D. magna were maintained on a food resource of Selanastrum sp. for 104 d. The daily food availability was experimentally varied at three levels (coefficient of variation = 0, 1, 2), which are referred to as "low," "medium," and "high," respectively, while the long-run average volume of food over time was kept constant across all replicates in all treatments. Extinctions were tabulated and populations were counted daily, although populations with ten or more individuals were simply marked "abundant." With the exception of these observations, sampling error is negligible. For more detailed methods, see \[[@pbio-0030222-b28]\]. Prior to this analysis, these data were filtered, retaining only observations from every seventh day for days 1 through 99, corresponding to the approximate generation time, resulting in observations of change in abundance over up to 14 intervals for each population. To achieve independence between data used for model fitting and data used for model testing, populations were assigned to separate datasets. Observations of ten or more individuals were excluded from the dataset for model selection and parameter estimation. Exploratory graphical analysis {#s4b} ------------------------------ Because populations in the low-variability treatment were not exposed to any experimentally induced environmental variation, most variation in observed growth rates in these populations should be attributable to demographic stochasticity and is not confounded with environmental stochasticity. Thus, only these data were used for exploratory analysis of density-dependent demographic stochasticity. Estimates of the pairwise multiplicative population growth rate λ\^(N~t~) between times t and t + τ(τ = 7) were obtained from the ratio N~t~ ~+τ~/N ~t~ for observations in the model-fitting dataset from the low-variability treatment. A scatter plot of ^λ^(N~t~) versus N ~t~ suggested that density dependence in expected population size is approximately linear on a logarithmic scale (unpublished data). To investigate the structure of deviation from this expectation, data from this treatment were fit to a Ricker model λ\^(N~t~) = using ordinary least squares and the squared residuals (δ ~j~ ^2^) were retained. To account for the scaling of demographic variance with population size \[[@pbio-0030222-b15], [@pbio-0030222-b16], [@pbio-0030222-b18]\], the residuals were multiplied by the population size at the start of the interval resulting in the rescaled residuals . A scatter plot of the rescaled residuals shows clear dependence on population size even after the usual scaling has been accounted for (see [Figure 1](#pbio-0030222-g001){ref-type="fig"}), which is strongly confirmed by two nonparametric tests for correlation (Spearman rank-order and Kendall\'s τ). Relatively constant variation around the mean and linear decrease on a logarithmic scale suggest using an exponential function to model variation from demographic stochasticity. Thus, for model selection and estimation, the function σ ~d~ ^2^ (N) = e ^−\ αN+β^ was used. Hyperbolic models of demographic stochasticity were also explored, but these were poorly supported by formal model selection criteria such as AIC, relative to the exponential model. Estimating extinction rates in experimental populations {#s4c} ------------------------------------------------------- Because populations were independent, each population represents a Bernoulli trial for which the possible outcomes were extinction or persistence. Thus, the chance of extinction for a population in treatment level i is binomial with parameter p ~i~. For n ~i~ populations in the model-testing dataset, of which x ~i~ were observed to go extinct, the maximum likelihood estimate of p ~i~ and 95% CIs on the chance of extinction were determined from the likelihood function for the binomial distribution. Initially, it was unclear if there was an effect of experimental treatment. Using regression on pooled observations from the entire experimental dataset, Drake and Lodge \[[@pbio-0030222-b28]\] reported no significant effect of environmental variation on the chance of extinction (p = 0.0877). Logistic regression was performed on the testing half of the dataset using treatment as a categorical covariate. The global null hypothesis was rejected by the likelihood ratio test (p = 0.0450) but not by Wald\'s test (p = 0.0523), with no coefficients being significantly different than 0 (intercept, p = 0.060; β ~0~, p = 0.174; β ~1~, p = 0.236). Visual inspection of the data suggests that there might be no difference between the low- and medium-variability treatments ([Figure 2](#pbio-0030222-g002){ref-type="fig"}), but a difference between these and the high-variability treatment. Repeating logistic regression after pooling observations from the low- and medium-variability treatments unambiguously showed an effect of high variation on the chance of extinction (likelihood ratio test, p = 0.013; Wald\'s test, p = 0.015). Therefore, to compare observed extinction rates with model-predicted extinction rates, observations from the low- and medium-variability treatments were pooled for both model fitting and model testing. Model selection {#s4d} --------------- The theoretical variance in λ is given by σ ~e~ ^2^ (N) + σ ~e~ ^2^ (N)/N,while the mean can be given by any familiar population dynamical model such as Ricker, θ-Ricker, θ-logistic, or Gompertz growth. Using the exponential model for demographic variation and approximating the unknown distribution of λ with the first two moments, the negative log-likelihood function for the θ-Ricker model of population growth is where λ ~0~ and θ can be interpreted as governing the intrinsic rate of increase and the severity of density dependence in the expected growth rate, respectively, and n ~N~ is the number of observed intervals beginning at population size N. The negative log-likelihood for this model with the addition of constant environmental stochasticity is where σ ~e~ is a parameter representing the average level of variation from environmental stochasticity. Models were fit to data in the model-fitting dataset by minimizing the negative log-likelihood function using the Nelder-Mead simplex. Goodness of fit was quantified using AIC = 2 NLL(ψ\^\\|y) + 2 k where ψ\^ is the vector of maximum likelihood estimates of all parameters, and k is the number of parameters estimated. The model with the lowest AIC score is the best fit after accounting for model complexity, while the relative support for a model with the lesser of two AIC scores differing by greater than two is substantial \[[@pbio-0030222-b32]\]. Overall, the inclusion of the parameter θ (which allows for flexibility in the severity of density dependence in expected population growth rate) did little to fit the model and was fixed (θ = 1) for the remainder of the analysis, reducing the number of parameters to be estimated by one. AIC scores for remaining models are shown in [Table 1](#pbio.0030222.t001){ref-type="table"}. Interestingly, the estimates of σ ~e~ for data from all experimental treatments were not significantly different from 0, even for the treatment with a high level of experimentally induced variation. Accuracy of model-predicted extinction rate {#s4e} ------------------------------------------- Predicted extinction rates were obtained by simulating 100,000 iterations of the Ricker model with density-dependent demographic stochasticity at maximum likelihood estimates of all parameters for populations in the pooled low- and medium-variability treatments and populations in the high-variability treatment separately using Euler\'s method \[[@pbio-0030222-b22]\]. Since predicted extinction rates for populations in the high-variability treatment were biased, and the model without environmental stochasticity was only weakly supported for these populations ( [Table 1](#pbio.0030222.t001){ref-type="table"}), 100,000 iterations of the Ricker model were also simulated with density-dependent demographic stochasticity and environmental stochasticity for populations in the high-variability treatment. The model for stochastic population growth considered here accounts for two factors commonly ignored when predicting the chance of population extinction: density-dependent changes in expected population size and density-dependent demographic stochasticity. Although the effect of density-dependence in E\[λ\] on population persistence has been documented \[[@pbio-0030222-b25], [@pbio-0030222-b33]\], the effects of density-dependent demographic stochasticity have not been as extensively studied \[[@pbio-0030222-b18], [@pbio-0030222-b27]\]. Therefore, the chance of extinction that would have been predicted had the incorrect assumption been made that demographic stochasticity was density-independentwas also determined. An estimate of density-independent demographic variance σ ~d~ ^2^ (N) = α was obtained by fitting data from the pooled low- and medium-variability and high-variability treatments to the Ricker model by minimizing the negative log-likelihood function As above, model predicted extinction rates were obtained by simulating 100,000 iterations of the population growth process at maximum likelihood estimates of all parameters. The author thanks R. Schwartz, P. Baggenstos, and J. Frentress for assistance with this experiment; and J. Hellman, T. Coulson, M. Vellend, and two anonymous referees for comments on the manuscript. This research was supported by a Great Lakes Fishery Commission grant (to principal investigator, D. Lodge), a scholarship from the Illinois-Indiana Sea Grant (to JMD), an Environmental Protection Agency Science to Achieve Results Graduate Research Fellowship (to JMD), and a University of Notre Dame Schmitt Graduate Student Research Fellowship (to JMD). It was also supported by a contribution from the Integrated Systems for Invasive Species project (principal investigator, D. Lodge), funded by the National Science Foundation (DEB 02--13698) and the University of Notre Dame. Competing interests. The authors have declared that no competing interests exist. Author contributions. JMD conceived and designed the experiments, performed the experiments, analyzed the data, contributed reagents/materials/analysis tools, and wrote the paper. ¤ Current address: National Center for Ecological Analysis and Synthesis, Santa Barbara, California, United States of America Citation: Drake JM (2005) Density-dependent demographic variation determines extinction rate of experimental populations. PLoS Biol 3(7): e222. AIC : Akaike\'s information criterion CI : confidence interval	PubMed Central	2024-06-05T03:55:57.557773	2005-6-7	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1150291/", "journal": "PLoS Biol. 2005 Jul 7; 3(7):e222", "authors": [ { "first": "John M", "last": "Drake" } ] }
PMC1150292	Anyone with a passing familiarity with animal behavior knows the classic photo of Konrad Lorenz trailed by a gaggle of goslings. Lorenz showed that geese hatched in an incubator identified with the first moving stimulus they saw within 36 hours after birth. When the first thing was Lorenz---or more accurately, his boots---the baby geese imprinted on him. But what can a newborn recognize? What elements in the visual world are their brains preprogrammed to process? In a new study, Giorgio Vallortigara, Lucia Regolin, and Fabio Marconato take advantage of the natural imprinting behavior of newly hatched chickens to study motion perception. It\'s clear that animals are more likely to respond to something that moves than to something that doesn\'t and that the motion of animate objects (biological motion) has uniquely identifiable characteristics. Is this identification innate or is it something an animal learns through experience? [](#pbio-0030251-g001){ref-type="fig"} ::: {#pbio-0030251-g001 .fig} ::: {.caption} ###### Are you my mom? "Point light animations" test chicks\' preference for biological motion ::: ![](pbio.0030251.g001) ::: To tackle these questions, the researchers presented chicks with artificially induced motion patterns. The chicks were immediately drawn to and stayed close to iconic patterns of animal motion, also known as biological motion. Interestingly, the chicks didn\'t distinguish between friend and foe: they were just as likely to approach a cat as a hen. To test the chicks\' preference for biological motion, the researchers took advantage of the fact that many vertebrates---whether they be geese, chickens, or humans---move in a distinctive, coordinated manner. By strategically positioning tiny lights at key points along an animal\'s torso and limbs, it\'s possible to strip a moving object of all extraneous traits like shape, texture, and color. People watching such "point light display" animations can easily recognize a person walking, discern the person\'s gender, and even identify a friend. Animals are similarly endowed with the ability to extract information from point light displays. To test the chicks\' preference for different types of motion, Vallortigara et al. created four animations from 13 points of light. One represented a walking hen, a second created the impression of a "rotating rigid hen-like object" (true animal movement is fluid, though constrained by the skeleton), a third moved in arbitrary directions, and a fourth, the "scrambled hen," conveyed biological motion, but of an unknown creature (as perceived by human observers). Chicks were hatched in darkness to make sure that the first thing they saw was one of the animations. The chicks consistently approached the walking and scrambled hen, showing far less affinity for the rigid and random motion, suggesting a predisposition toward the movement typical of vertebrates. As a control, the authors generated an animation of a walking cat. Sure enough, chicks approached the walking cat as often as they approached the walking hen. Luckily for chicks, encounters with cats are not normally likely to precede encounters with a mother hen. These results suggest that chicks have evolved a predisposition to notice objects that move like vertebrates, which may maximize the probability of imprinting on the object most likely to provide food and protection after birth. This predisposition likely guides the learning that occurs during the imprinting process, when the chick learns how to distinguish mom from other hens, and hens from cats. Since both birds and mammals (tested in four-month-old human infants) show a preference for biological motion, the authors conclude, these results suggest that this preference is hard-wired into the vertebrate brain. With this new model, researchers can investigate the motion-specific features that guide the chicks\' behavior and how the brain processes biological motion.	PubMed Central	2024-06-05T03:55:57.559675	2005-6-7	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1150292/", "journal": "PLoS Biol. 2005 Jul 7; 3(7):e251", "authors": [] }
PMC1150293	Aside from global climate change, loss of biodiversity poses one of the greatest threats to the planet. Last year, the World Conservation Union reported an unprecedented decline in biodiversity, with nearly 16,000 species facing extinction. The biggest threat to the vast majority of these species is loss of habitat. And as habitat loss and degradation proceed nearly unabated, the need to accurately predict the population dynamics and extinction risk of potentially endangered species has never been greater. In a new study, John Drake tests models traditionally used to estimate the likelihood of extinction and shows that because the models ignore a critical parameter in projecting risk, they underestimate extinction rates. Standard models for predicting extinction assume that population growth and decline are governed by random, or stochastic, variables. The models typically incorporate two major contributors to random variation in population growth rates: changes in environmental conditions and chance fluctuations in population size---caused by variations in individual fitness, random mating behavior, and events that affect just one individual---that are referred to as demographic stochasticity. But since few scientists have tested these models with empirical data, the question remained whether the models were accurately predicting population fluctuations and extinction risk. To test the reliability of standard stochastic models, Drake used data from experiments with water fleas. He found that the models could accurately predict extinction risk only when there was enough information about variation in individual fitness to account for demographic variability---a finding that undercuts the conventional wisdom that demographic stochasticity is unimportant. Some traditional models do not even include demographic stochasticity. It\'s generally assumed that fluctuating environments, a given in the natural world, increase a species\'s chance of extinction. Drake tested this notion in experiments by manipulating the available food sources in 281 populations of water fleas. The flea populations received either low, medium, or high amounts of food, and Drake kept daily tallies of population number and extinctions. When he tried to predict the extinctions using traditional models, he couldn\'t. [](#pbio-0030253-g001){ref-type="fig"} ::: {#pbio-0030253-g001 .fig} ::: {.caption} ###### Experiments with Daphnia magna, the water flea, show that traditional extinction models may be underestimating extinction risk ::: ![](pbio.0030253.g001) ::: To account for the discrepancy between model and data, Drake began to investigate a possibility raised by recent theoretical research that population density and individual interdependence might affect a major component of the model---demographic stochasticity. The idea is that if organisms interact in their environments---which of course they do---then these interactions will likely affect an individual\'s probability of dying or reproducing, which ultimately affects species survival. Drake calls this variable density-dependent demographic stochasticity. Drake used half of the experimental data generated from testing the effects of environmental variability on water flea survival to select his models and estimate the range of parameters that might affect extinction, and the other half to test the models\' reliability. From the estimated parameters, Drake wrote a computer program to simulate all the possible population outcomes and predict extinction rates. One set of simulations included a parameter for density-dependent random interactions and another did not. When Drake analyzed all the possible outcomes, it turned out that manipulating food supply didn\'t have as great an effect on extinctions as predicted---possibly because individual water fleas live too long compared to the frequency of the environmental fluctuations. Only when density-dependence was included did the models match the observed extinction rates in the flea experiments. When density dependence was not included, extinction rates were greatly underestimated. Drake\'s results underscore the importance of bolstering extinction models with empirical validation---and of accounting for population density---to accurately evaluate risk and enhance recovery programs for at-risk populations. As threats to endangered species continue to mount, biologists will need ever more robust methods to estimate extinction risk. Unfortunately, field biologists typically can\'t generate the large, high-quality datasets that led to the precise predictions reported here. Conservation efforts will depend on developing methods of generating reliable predictions with the limited data available from the field.	PubMed Central	2024-06-05T03:55:57.560417	2005-6-7	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1150293/", "journal": "PLoS Biol. 2005 Jul 7; 3(7):e253", "authors": [] }
PMC1150294	Most of us don\'t have much trouble recognizing what we see. Whether it is a face in a crowd, a bird in a tree, or papers on a desk, our brains expertly distinguish the target from the clutter. It is a simple skill most of us take for granted, but object recognition is not hard-wired. As we navigate our environment, the brain\'s visual centers continually reorganize themselves, classify novel features, and learn to pick out important objects from the background. Just how the human brain does this is not well understood, but new research by Zoe Kourtzi and colleagues may have uncovered some important clues. To investigate how the human brain learns to separate targets (signal) from noise Kourtzi et al. showed subjects pictures of novel shapes embedded in a cluttered background and asked the subjects to determine whether or not the shapes were symmetrical. The researchers recorded the subjects\' responses while using functional magnetic resonance imaging (fMRI) to measure neuronal activity in brain regions associated with visual processing. Each subject was tested using two sets of novel shapes: high-salience shapes (shapes easily distinguished from the background), and low-salience shapes (shapes camouflaged by the background). After the initial testing, the subjects were trained to recognize a subset of the new shapes from each group, and then re-tested. Visual input is thought to go through a hierarchy of processing centers that transform retinal images into complex objects and scenes. Kourtzi et al. recorded responses from both early (V1, V2, Vp, and V4) and late (lateral occipital cortex) stages of visual analysis in 26 subjects. The authors found that subjects demonstrated an increased number of correct responses for shapes they encountered during the training sessions, regardless of the type of background the shapes were presented on. By contrast, the fMRI responses differed dramatically, depending on whether the surroundings made the shapes easy or difficult to detect. Low-salience shapes triggered an increased fMRI response across all brain regions following training; high-salience shapes precipitated a decrease in fMRI response in the regions of the lateral occipital cortex, but produced no change in any of the early visual areas (V1, V2, Vp, and V4). [](#pbio-0030256-g001){ref-type="fig"} ::: {#pbio-0030256-g001 .fig} ::: {.caption} ###### The human brain learns to detect the contours of target objects in cluttered scenes by recruiting early and higher centers of visual analysis ::: ![](pbio.0030256.g001) ::: These results demonstrate that the ability to learn to detect novel shapes is independent of the degree of difficulty, but suggest that the brain employs different mechanisms of perceptual learning depending on whether the objects stand out from their surroundings, or are obscured by them. Learning to detect highly camouflaged shapes results in increased brain activity levels that are presumed to reflect an increase in signal processing at the level of both the early visual areas and higher levels of cortical analysis. On the other hand, the reduction of neural activity that occurs during learning of more distinctive shapes likely reflects efficient neural coding of the critical features for their recognition at later stages of visual analysis. According to Kourtzi and her colleagues, their results provide evidence that the visual brain is capable of tailoring the mechanism of perception to best suit the task. When the signal is weak---as in the case of viewing camouflaged targets---learning amplifies neural responses to the target shapes and drowns out the noise. But when the signal is strong---as in the case of viewing easily distinguishable, highly salient targets---neural activity in the visual cortex is reduced, possibly because training engages smaller populations of neurons that respond much more selectively to distinctive features of the stimulus. In other words, all visual stimuli are not treated equally, and with just cause: the brain\'s unique ability to treat ambiguous signals differently than robust ones likely allows it to optimize neural coding, and in doing so, learn to increase detection of a broad spectrum of visual signals.	PubMed Central	2024-06-05T03:55:57.561107	2005-6-7	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1150294/", "journal": "PLoS Biol. 2005 Jul 7; 3(7):e256", "authors": [] }
PMC1151593	Introduction ============ This editorial is designed to introduce the new online journal Clinical Practice and Epidemiology in Mental Health(CPEMH). Mental health issues today can rely on medical model based researches, which are capable of providing reliable results following the introduction of descriptive diagnoses in psychiatry. This revolution (implying the use of a common language) took place as recently as the 1970\'s \[[@B1]\], but is still often deemed inadequate due to the fact that, contrary to other areas of medicine, it is impossible to define psychiatric disorders as etiopathogenetic, anatomopathologic and clinically descriptive entities. Due to the current situation in psychiatric research, when providing a clinical definition for a disorder made up of a group of cases presenting a similar symptomatology and course, it may prove necessary to establish the degree of homogeneity revealed with regard to environmental and genetic risk factors. It would be of particular interest to evaluate whether the clinical homogeneity could also be extended to \"conditions\" identified in the general population receiving no psychiatric treatment. The association between clinical research and epidemiology is of considerable importance in the medical field, and at the current time is possibly even more so for the field of psychiatric research. Clinical Practice and Epidemiology in Mental Health will encompass all aspects of clinical and epidemiological research in psychiatry and mental health, and will aim to build a bridge between clinical and epidemiological research. CPEMHis aimed at clinicians and researchers focused on improving the knowledge base for the diagnosis, prognosis and treatment of mental health conditions; and improving the knowledge concerning frequencies and determinants of mental health conditions in the community and the populations at risk. The journal will also cover health services research and economic aspects of psychiatry, with special attention given to manuscripts presenting new results and methods in the important area of epidemiology of treatments in mental heath, particularly clinical epidemiologic investigation of psychotherapies and pharmaceutical agents. Open Access =========== CPEMH\'s Open Access policy supports the changes in the way in which articles in mental health can be published \[[@B2]\]. First, all articles are freely and universally accessible online as soon as they are published. Second, the authors hold copyright for their work and grant anyone the right to reproduce and disseminate the article, provided that it is correctly cited and no errors are introduced. Third, a copy of the full text of each Open Access article is permanently archived in an online repository separate from the journal. CPEMHarticles are archived in the leading databases (such as PubMed Central \[[@B3]\], the US National Library of Medicine\'s full-text repository of life science literature, and also in repositories at the University of Potsdam \[[@B4]\] in Germany, at INIST \[[@B5]\] in France and in e-Depot \[[@B6]\], the National Library of the Netherlands\' digital archive of all electronic publications). Open Access has four broad benefits for science and the general public. First, all articles become freely and universally accessible online, and so an author\'s work can be read by anyone at no cost, given that there are no barriers to access their work. This is accentuated by the authors being free to reproduce and distribute their work, for example by placing it on their institution\'s website. Furthermore, there is evidence that free online articles are more highly cited because of their easier availability \[[@B7]\]. Second, the information available to researchers will not be limited by their library\'s budget, and the widespread availability of articles will enhance literature searching \[[@B8]\]. Third, publicly funded research will become accessible to all taxpayers (not just those with access to a library with a subscription). As such, Open Access could help to increase public interest in, and support of, research. Note that this public accessibility may become a legal requirement in the USA if the proposed Public Access to Science Act is made law \[[@B9]\]. Similar calls for a move to Open Access of all scientific research have been made recently by the UK government \[[@B10]\]. Fourth, a country\'s economy will not influence its scientists\' ability to access articles because resource-poor countries (and institutions) will be able to read the same material as wealthier ones (although creating access to the Internet is another matter \[[@B11]\]). This is particularly relevant in mental health issues which represent one of the major health priorities throughout the world. Indeed, by the year 2020 depression will be the second leading cause of disability \[[@B12]\]. Mental health research focused on low- and middle-income (LAMI) countries is needed in view of the burden of neuropsychiatric diseases and the deficiency of mental health resources in these countries \[[@B13]\]. Numerous projects set up by international corporations such as the WHO are attempting to develop mental health programmes and researches in developing countries \[[@B14]\]. This is an exciting opportunity to disseminate our science in the new world wide medium of electronic publishing. Clinical Practice and Epidemiology in Mental Healthlooks forward to receiving your submissions. General Journal policy and Peer Review policy ============================================= Clinical Practice and Epidemiology in Mental Healthconsiders the following types of articles: Research: reports of data from original research. Reviews: comprehensive, authoritative, descriptions of any subject within the scope of CPEMH. These articles are usually written by opinion leaders that have been invited by the Editorial Board. Short reports: brief reports of data from original research. Commentaries: short, focused and opinionated articles on any subject within the scope of the journal. These articles are usually related to a contemporary issue, such as recent research findings, and are often written by opinion leaders invited by the Editorial Board. Case reports: reports of clinical cases that can be educational, describe a diagnostic or therapeutic dilemma, suggest an association, or present an important adverse reaction. All case report articles should be accompanied by written and signed consent to publish the information from the patients or their guardians. Methodology articles: present a new experimental method, test or procedure. The method described may either be completely new, or may offer a better version of an existing method. The article must describe a demonstrable advance on what is currently available. Debate articles: present an argument that is not essentially based on practical research. Debate articles can report on all aspects of the subject including sociological and ethical aspects. Hypotheses: short articles presenting an untested original hypothesis backed solely by previously published results rather than any new evidence. They should outline significant progress in thinking that would also be testable. Study protocols: describes proposed or ongoing research, providing a detailed account of the hypothesis, rationale, and methodology of the study. Book reviews: short summaries of the strengths and weaknesses of a book. They should evaluate its overall usefulness to the intended audience. Manuscripts submitted to the journal will be reviewed by expert reviewers in the field. CPEMHwill have an open peer review policy. Once the reviewers have provided their feedback, the Editor makes the final recommendation. The Editor will be available to assist authors with content and formatting issues not resolved during the review process. He will also assist the authors of review articles with integration of content with the ATV website (where appropriate). Articles will be published online immediately upon acceptance and soon after listed in PubMed. Based on the above characteristics and by means of a series of commentaries on the current state of practice of different aspects of mental health, the journal will seek to stimulate discussion of the main psychiatric topics facing clinical and epidemiological research across the scientific community. An editorial board of international renown of about 50 members has been established \[[@B15]\]. The members of the board will contribute to this series of commentaries. It is no mere chance that the first commentary by Carta and Angst \[[@B16]\] deals with bipolar disorders, arguably a leading issue in current psychiatric research. In spite of the importance of the topic, a number of researchers maintain that it is not adequately taken into account in the medical field, probably due to the marked inconsistency in the available epidemiological data. Conclusion ========== There are several outstanding mental heath journals covering all aspects of this dynamic field, but none of these journals is devoted to bridging clinical and epidemiological research. The Open Access online distribution of the journal and its inclusion in the leading databases (such as PubMed Central) will ensure widespread and ready visibility, which are indispensable given the demand for immediate debate and comparison of scientific findings. We welcome any advice and input. Competing interests =================== Critics of Open Access often suggest that Editors have a financial incentive to accept articles as more articles means more revenue. However, BioMed Central insists that decisions about a manuscript must be based on the quality of the work, not on whether the article-processing charge can be paid. This policy will certainly apply for CPEMHwhose authors and readers will benefit from learning about mental health in regions of the world and in particular field of research (e.g psychotherapies vs pharmacoterapies) with limited financial resources. No member of the editorial or advisory boards of CPEMHor their Institutions will receive any portion of the article-processing charge.	PubMed Central	2024-06-05T03:55:57.561664	2005-4-27	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1151593/", "journal": "Clin Pract Epidemiol Ment Health. 2005 Apr 27; 1:1", "authors": [ { "first": "Mauro Giovanni", "last": "Carta" }, { "first": "Maria Carolina", "last": "Hardoy" } ] }
PMC1151594	Introduction ============ In recent decades, a large number of cross-sectional psychiatric epidemiological surveys have reported prevalence estimates for major depression. Prevalence data provide a clear perspective on the burden of major depression in the population, and these estimates are useful for health system priority setting and planning. Unfortunately, the implications for clinical practice are not as clearly evident. Sometimes, clinicians misinterpret prevalence estimates as estimates of risk, but this is an error because whereas incidence is a measure of risk, prevalence is influenced by episode duration (prognosis) and, to a lesser extent in the case of major depression, by mortality. Ideally, it would be possible to decompose associations that are observed in epidemiological prevalence data into their main determinants: incidence and episode duration. Incidence reflects the probability of development of new depressive episodes. This parameter is particularly important for prevention. In primary prevention, an understanding of the risk of new episodes in sub-groups within the population can support the targeting of preventive efforts towards those groups at highest risk. Therapeutic decisions, such as those concerning the need for acute and maintenance-phase pharmacological treatment, depend upon an understanding of the expected course of a disorder in a particular patient An ability to decompose period prevalence data into more meaningful statements about risk and prognosis may increase the extent to which epidemiological data can inform clinical practice. Some psychosocial factors that could plausibly be associated with major depression (e.g. education) have been found not to be associated with it in cross-sectional studies. In these instances, there continues to be some value in examining the ways in which incidence and prognosis combine to influence prevalence. A lack of association in cross-sectional data may be the end result of offsetting factors: an elevated incidence associated with improved prognosis, for example, may mask an important association in prevalence data. A psychosocial factor could influence prognosis in a clinically relevant way, even if that factor is not associated with major depression in cross-sectional epidemiological data. In Canada, a National Study of Mental Health and Wellbeing has recently been conducted. This study utilized the WHO Mental Health 2000 version of the Composite International Diagnostic Interview (WMH CIDI) \[[@B1]\], and collected data from a nationally representative sample of 36,984 subjects. An expected pattern of cross-sectional association was observed (see Table [1](#T1){ref-type="table"}). The prevalence was higher in women than in men, in young age categories, and in previously married subjects. There was no evidence of an association with education, a result that was also observed in the pan-European ESEMeD study \[[@B2]\]. Notably, an association with this variable was observed in the American National Comorbidity Survey \[[@B3]\] and its Replication \[[@B4]\]. Unfortunately, it is difficult to know whether the pattern of association with period prevalence represents differences in risk, or prognosis, or some intermixing of these factors. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Cross-sectional Associations of Demographic Factors with Major Depressive Episode Prevalence\. ::: 12-month prevalence (%) 95% C.I. ---------------- -------------------- ------------------------- ------------ Overall 4.8 4.5 -- 5.1 Sex Male 3.7 3.3 -- 4.1 Female 5.9 5.4 -- 6.4 Age 15--25 6.2 5.4 -- 7.0 26--45 5.6 5.0 -- 6.0 46--65 4.4 3.8 -- 5.0 \> 65 2.0 1.5 -- 2.4 Marital Status Wid/Sep/Div/Single 4.4 2.5 -- 3.1 Married/Single\\* 2.8 4.0 -- 4.8 Education Some \> HS 4.8 4.4 -- 5.3 ≤ High School 4.8 4.3 -- 5.2 \* Canadian Mental Health and Wellbeing Survey <http://www.statcan.ca>. \\ includes \"common law\" status. ::: In this paper, Markov modeling procedures are used to synthesize several sources of epidemiological data from Canada, with the objective of recasting the data in a way that is more clinically useful. Methods ======= 1. Markov modeling procedure ---------------------------- In a series of previous papers, a method for modeling the epidemiology of major depression has been described \[[@B5],[@B6]\]. The method uses Markov models that simulate major depression epidemiology over a series of one week stages. The model contains two health states, depressed and non-depressed. Incidence is the transition probability associated with a change from the non-depressed to depressed state. The recovery pattern for major depression is more difficult to represent because, in major depression, the probability of transition between depressed and non-depressed states depends upon the amount of time spent in the depressed state (episode duration). As episodes become more chronic, the probability of recovery in any one week becomes smaller \[[@B7],[@B8]\]. This reality makes the use of conventional epidemiological incidence-prevalence-mortality models such as the World Health Organization\'s DISMOD program \[[@B9]\] for major depression modeling somewhat questionable, although they do continue to be used for this purpose \[[@B10],[@B11]\]. For these reasons, a \"markov tunnel\" \[[@B12]\] was used to model recovery in the analyses presented in this paper. At the onset of an episode, a subject enters the first stage of the tunnel: the first stage (or week) of the depressive episode. At the next stage, the subject either makes a transition back to the non-depressed state, or can move to the next stage in the tunnel (this is the complementary probability, such that the probability of one or the other of these events is 1.0). This next stage represents the second week spent in the interval, and so on. Using this procedure, the transition probabilities for recovery can be altered depending upon episode duration. Preliminary work with this type of model has determined that the impact of mortality on major depression models is small \[[@B6]\]. In the current analyses, mortality was not considered and period prevalence is viewed as a composite measure reflecting incidence and prognosis. The transition probabilities for incidence and all of the weekly recovery probabilities (those transition probabilities associated with stages in the Markov tunnel) were linked to epidemiological data using Monte Carlo simulation \[[@B5]\]. Tracker variables were programmed to count the number of weeks spent in an episode, the proportion of simulated subjects developing episodes during intervals of time, etc. Possible values for the various transition probabilities were explored in order to find ones that predicted epidemiological parameter values similar to those estimated from suitable data sources. 2. Sources of data ------------------ This analysis utilized data from three different national survey projects. The first was a cross sectional study, the Canadian National Survey of Mental Health and Wellbeing <http://www.statcan.ca/Daily/English/030903/d030903a.htm>, which was a cross-sectional study that used the WHO Mental Health CIDI \[[@B1]\]. Incidence data derived from a national longitudinal study called the National Population Health Survey (NPHS) <http://www.statcan.ca/english/concepts/nphs/nphs.htm>. Episode duration data was derived from a large cross-sectional survey, the Canadian Community Health Survey (CCHS 1.1) <http://stcwww.statcan.ca/english/sdds/3226.htm>. The NPHS is a longitudinal study that follows a cohort of approximately 17,000 subjects with repeated interviews every two years. The study started in 1994 (interview process completed in 1995), and the interviews were repeated in 1996, 1998 and 2000. The CCHS 1.1 is a cross-sectional study with a sample size of n = 130,880 conducted in 2000. Both studies employed the CIDI Short Form for major depression (CIDI-SFMD) \[[@B13]\], which is a brief predictive interview that assesses 12-month prevalence of major depression. The positive predictive value of the CIDI-SFMD for CIDI-defined major depressive episode is approximately between 75% and 90% \[[@B13],[@B14]\]. Using the NPHS, it is possible to estimate an approximation of annual incidence: the proportion of the cohort that were CIDI-SFMD negative at one interview (e.g. 1994), who were positive at their next interview two years later (in this case, 1996). Both studies included an item for those positive on the CIDI-SFMD, asking about weeks depressed in the past year. Data from the CCHS 1.1 was used in the analysis of this variable because the larger sample size led to greater precision of estimation. While the incidence approximation in the NPHS has been used directly in some studies \[[@B15],[@B16]\], Markov models can be used to refine these estimates \[[@B5]\]. The CIDI-SFMD includes an item that asks subjects classified as having major depression to report the number of weeks that they spent in the depressed state during the preceding year. The Markov modeling technique is useful for translating this weeks depressed in the past year data into a set of estimates for the weekly recovery rates for inclusion in a Markov tunnel \[[@B5]\]. Tracker variables are programmed into the Markov model to represent the probability of an episode in the last 52 weeks of a 104 week simulation interval (the incidence approximation available from the NPHS) and another tracker variable is programmed to count the number of weeks in the 52 weeks preceding an interview that were spent in the depressed state, simulating the data collected in the survey. Using Monte Carlo simulation, it is then possible to identify incidence rates and Markov tunnel recovery probabilities that are most consistent with the observed data \[[@B5],[@B6]\]. In this project, these methods are employed in order to further delineate the cross-sectional associations observed in Table [1](#T1){ref-type="table"}, assisting with a determination of whether these associations are due to differences in incidence or episode duration. For one variable, education (which was not found to be associated with major depression prevalence), the objective was to explore whether off-setting incidence-duration effects could explain this lack of association. 3. Data synthesis ----------------- In order to explore the epidemiology of major depression in relation to the categorical variables listed in Table [1](#T1){ref-type="table"}, the NPHS and CCHS datasets were stratified by these variables. Marital status was divided into unmarried (divorced, widowed or separated, single) and married categories -- with the latter category including \"common law\" relationships. Similarly, the subjects were divided into education and employment status categories. Because of the value of exploring multiple age levels, and the possibility of age-sex interactions, logistic regression models were created to explore the pattern of approximate incidence in these groups. An estimate of incidence was obtained in this way within the various strata for each of the three available NPHS cycles: 1994--96, 1996--98 and 1998--2000. Next, episode duration data for subjects within the specified age categories and for men and women were extracted from the CCHS 1.1 dataset. A series of Monte Carlo simulations were then run using ranges of possible values for incidence and recovery probabilities. The software, Data \[[@B17]\] was used for simulation. These trials were guided by a least squares minimization procedure to identify the set of transition probabilities that best explained the incidence and episode duration data \[[@B5]\]. Results ======= The estimated incidence approximations, stratified by the potential explanatory variables are presented in Table [2](#T2){ref-type="table"}. Since there was no evidence of variation across the data collection cycles, an average was taken and used in the Markov modeling simulations. These averages are presented in the right-hand column of Table [2](#T2){ref-type="table"}. Generally, the incidence approximation follows a pattern resembling the prevalence pattern, except that marital status appears not to be associated with a difference in incidence. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Major Depressive Episode, Approximation of Annual Incidence. ::: Major Depressive Episode Incidence ---------------- ----------------------- ------------------------------------ ------------------ ------------------ ----- Sex Male 2.8 (1.9 -- 3.6) 2.5 (1.9 -- 3.1) 2.9 (2.2 -- 3.6) 2.7 Female 4.3 (3.7 -- 5.0) 4.7 (3.8 -- 5.5) 4.6 (3.8 -- 5.3) 4.5 Age 12--25 5.3 (3.8 -- 6.9) 5.2 (3.5 -- 6.8) 4.8 (3.3 -- 6.2) 5.1 26--45 3.6 (2.7 -- 4.5) 3.8 (3.1 -- 4.5) 4.1 (3.6 -- 5.2) 3.8 46--65 2.8 (2.0 -- 3.6) 2.8 (1.8 -- 3.7) 2.7 (2.0 -- 3.5) 2.8 \> 65 1.5 (0.5 -- 2.4) 1.6 (0.6 -- 2.6) 1.5 (0.6 -- 2.4) 1.5 Marital Status Wid/Sep/Div or Single 3.7 (2.4 -- 5.0) 3.5 (2.4 -- 4.5) 3.5 (2.5 -- 4.5) 3.6 Married\* 3.6 (3.0 -- 4.1) 3.5 (3.2 -- 4.4) 3.8 (3.2 -- 4.4) 3.6 Education Some \> HS 2.8 (2.2 -- 3.2) 3.9 (3.2 -- 4.7) 3.7 (3.1 -- 4.4) 3.5 ≤ High School 4.6 (3.6 -- 5.4) 2.9 (2.3 -- 3.6) 3.8 (2.9 -- 4.7) 3.8 \* includes \"common law\" status. ::: 1. Sex ------ The reported number of weeks depressed in the past year were found to be almost identical for men and women. The same Markov tunnel was therefore used to simulate the pattern of recovery. The transition probabilities (TP) associated with each stage of the tunnel were found to be described adequately using the formula: TP~stage~= 0.14\e^-.047\stage^. This formula suggests that the probability of recovery in the initial week of an episode (ie. recovery by week three after the two week DSM-IV \[[@B18]\] duration criterion is met) is very high, approximately 14%. The weekly probability of recovery then declines by approximately 5% with each passing week. The final Markov model is depicted in Figure [1](#F1){ref-type="fig"}. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Markov Model for Major Depression, Stratified by Sex. ::: ![](1745-0179-1-2-1) ::: Figure [2](#F2){ref-type="fig"} presents the observed weeks depressed in the past year data from the CCHS (as a cumulative probability of recovery by week), and simulated values from the Markov model. Only one simulated curve is presented in the graphic because the curves for men and women were nearly identical. The simulated incidence approximation (the proportion of Monte Carlo simulations without an episode at baseline whose tracker variables indicate an episode in the last 52 weeks of a 104 week simulation run) varied linearly in relation to the incidence transition probabilities. The observed incidences (from Table [2](#T2){ref-type="table"}) were predicted by the weekly transition probabilities depicted in Figure [1](#F1){ref-type="fig"}: 0.000446 in men and 0.000744 in women. The flattening of the curve with advancing weeks depicts the emergence of more chronic episodes in the sense that the recovery probabilities per week become quite small as an episode approaches one year in duration. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Observed and Simulated Episode Duration Data, by Sex. ::: ![](1745-0179-1-2-2) ::: 2. Age ------ Age was evaluated at four levels, less than or equal to 25, 26--45, 46--65 and older than 65. The models suggested a more complex scenario than occurred for sex. In order to simulate the observed episode duration data, it was necessary to define three Markov tunnels, one for the less than or equal to age 25 (RP~group~= 0.19\e^-0.0443\week^), one for the 26 to 45 group (RP~group~= 0.15\e^-0.0499\week^) and one for the two older age categories, in other words, those over 45 years of age (RP~group~= 0.12\e^-0.0507\week^). The weekly incidence transition probabilities also needed to be higher in the younger age groups in order to reflect the age stratified incidence approximations estimated directly from the epidemiological data. The exponential parameters in the Markov tunnels indicate that the probability of recovery per week declines more quickly as the subjects\' ages become larger. The final Markov model is depicted in Figure [3](#F3){ref-type="fig"}, and the observed and simulated duration data is presented in Figure [4](#F4){ref-type="fig"}. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Markov Model for Major Depression, Stratified by Age Group. ::: ![](1745-0179-1-2-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Observed and Simulated Episode Duration, by Age Group. ::: ![](1745-0179-1-2-4) ::: 3. Marital Status ----------------- The Markov model for marital status is not presented here, since its structure resembled that presented above for sex. As described above, marital status was analyzed at two levels. The estimated weekly incidence transition probability was similar in the married and single group (0.000537) and the previously married (0.000598) categories. However, different Markov tunnels were required to accurately simulate the episode duration data: RP~group~= 0.14\e^-0.0419\week^for the married or single group and RP~group~= 0.10\e^-0.0461\week^for the previously married group. A simulation of the episode duration data is presented in Figure [5](#F5){ref-type="fig"}. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Observed and Simulated Episode Duration Data, by Marital Status. ::: ![](1745-0179-1-2-5) ::: 4. Education ------------ As seen in Table [1](#T1){ref-type="table"}, major depression period prevalence was not associated with education level. It is possible that offsetting effects involving risk and prognosis could account for this. However, the Markov modeling did not suggest that this was the case. Simulation of the annual prevalence and episode duration data required only a single incidence probability and a single Markov tunnel for recovery. As such, the model did not differ from the unstratified model previously presented \[[@B5]\]. Conclusion ========== The state of knowledge about major depression epidemiology is now supported by a large international literature of studies. However, the application of these data has largely been restricted to advocacy purposes and to broad-based priority setting exercises such as the global burden of disease project \[[@B10]\] or health economic studies \[[@B11],[@B19],[@B20]\]. Integrating period prevalence estimates with clinical practice is challenging. Traditionally, psychiatric assessment includes a formulation of etiological factors, and factors affecting prognosis are important for treatment planning, yet neither concept is closely linked to period prevalence. Point prevalence is a more complex parameter than is often assumed, reflecting a steady state outcome of other factors. Period prevalence is even more complex. In this study, Markov models were used to synthesize various sources of epidemiological data, and to decompose these into estimates of parameters that are may be more useful to clinicians: those involving the risk of new episodes, episode prognosis, both of these factors or neither of them. In the past, the most common application of Markov modeling in psychiatry has been in cost effectiveness analysis. Markov models are more widely used elsewhere in medicine to support clinical decision making, see review \[[@B12]\]. The simple models presented here help to clarify and synthesize epidemiological data in a way that could be integrated into clinical decisions concerning questions as the need for maintenance therapy, or the preferred duration of antidepressant treatment. In order to extend the approaches described in this paper towards application in quantitative decision analysis, it will be necessary to incorporate more variables (e.g. income, employment status, past history, family history) and also to incorporate procedures that account for more than one variable simultaneously. Whenever modeling procedures are used to interpret empirical data, it is important to qualify the output based on the quality of the input data. Two of the three surveys used in this analysis incorporated only a predictive short form of the CIDI, and therefore may have been vulnerable to measurement bias. Measurement bias is probably the main threat to the validity of the data, as the samples were population based and good response rates were achieved (see survey documentation at <http://www.statcan.ca>). In addition to concerns that the CIDI-SFMD does not always agree with the full CIDI \[[@B14]\], the recent literature contains expressions of concern about the extent to which even the full CIDI can identify clinically significant episodes in the population \[[@B21],[@B22]\]. In response to concern that episodes identified by the CIDI may not be clinically significant, some authors have suggested that epidemiological interviews should consistently incorporate clinical significance probes, typically eliciting subjects\' descriptions of the severity and intrusiveness of their symptoms and their behavioral responses to their symptoms, e.g. seeking treatment \[[@B23]\]. Other authors have gone further, suggesting that interviewers in epidemiological studies should be trained to make relevant clinical judgments \[[@B21]\]. Brief instruments such as the CIDI-SFMD include neither clinical significance probes nor opportunities for clinical judgments to be made by the interviewers. However, to the extent that judgments about clinical significance are based upon assessment of factors affecting risk and prognosis, the approach described here offers certain empirical advantages. Rather than letting clinical judgment shape the collection of empirical data, they allow empirical clinical data an opportunity to shape and inform clinical judgment. The demographic associations explored in this analysis are consistent with the current literature, providing some sense of confidence in their validity. Several previous studies have addressed the issue of an elevated prevalence in women by attempting to evaluate whether this is due to an effect on incidence or prognosis. Analyses deriving from the National Comorbidity Survey \[[@B24]\], the Vantaa Depression Study \[[@B25]\], the National Institute of Mental Health Collaborative Program on the Psychobiology of Depression \[[@B26]\] and the Baltimore Epidemiological Catchment Area Follow-up \[[@B27]\] have been consistent in observing a similar prognosis in men and women. This implies that the well-documented prevalence difference is probably due to incidence, the latter in fact having been confirmed by the Baltimore study \[[@B27]\]. The findings reported in this study are consistent with this literature. Fewer studies have directly addressed the issue of age. The NCS did not include subjects over the age of 64. The Baltimore study, however, reported declining incidence with age \[[@B27]\], as reported here. In the Vantaa study, the univariate analysis showed a trend towards increasing time to full remission with age (p = 0.073). An analysis of data from the Epidemiological Catchment Area Study in the US also reported lower recovery rates after one year in older subjects \[[@B28]\], but the effect was largely confined to women. The Baltimore study was consistent with the current findings in reporting no effect of education on episode duration \[[@B27]\]. However, the Baltimore study also reported that marital status had no \"important\" effect. In distinction to this, the current study found that an effect of unmarried status on prevalence was due to an impact of this variable on episode duration. An effect of marital status on prognosis seems plausible since many studies have found that personality disorders \[[@B25]\], and neuroticism scores \[[@B7]\] which would be expected to interfere with role functioning in relationships, are associated with diminished prognosis in major depression. However, marital status is likely to be a very crude indicator of these and a variety of other factors. More focused studies are required in order to more fully elucidate such effects. In general, the pattern of recovery emerging from the Markov models developed in this analysis are comparable to the pattern reported by previous studies \[[@B7],[@B25],[@B29]\]. Markov models offer an interesting opportunity for integration of epidemiological data with clinical decision making. Competing Interests =================== The author(s) declare that they have no competing interests. Acknowledgements ================ This study was supported by a grant from the Canadian Institutes for Health Research. Dr. Patten is a Health Scholar with the Alberta Heritage Foundation for Medical Research and a Research Fellow with the Institute of Health Economics.	PubMed Central	2024-06-05T03:55:57.562968	2005-4-27	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1151594/", "journal": "Clin Pract Epidemiol Ment Health. 2005 Apr 27; 1:2", "authors": [ { "first": "Scott B", "last": "Patten" } ] }
PMC1151595	Introduction ============ Current concerns about community care of the mentally disordered focus on the level of risk of unacceptable behaviour, particularly inappropriate aggressive or sexual acts. It is clear that such behaviour occurs in patients suffering from mental disorders, and might be expected to be more common in psychotic states, when provocation from delusional and hallucinatory perceptions is present \[[@B1]-[@B7]\]. The presence of concurrent substance abuse, personality disorder or history of previous violent or antisocial behaviour increases the risk of people with psychotic disorders committing violent offences \[[@B1],[@B2],[@B8]-[@B10]\]. The methodology of most of theses studies on which these conclusions are based has been flawed, with selection bias, lack of diagnostic precision and results often based upon criminal records rather than clearly defined ratings of behaviours \[[@B3],[@B11]-[@B14]\]. The issue of unreported and non-criminal aggressive behaviour, including inappropriate sexual behaviour, particularly in the context of family or residential care has been studied less extensively \[[@B15],[@B16]\]. Furthermore the course of this misconduct and the effects of continued community treatment have been seldom examined. One excellent study compared the course of aggressive behaviour in a cohort of patients after discharge from acute psychiatric hospital care \[[@B17]\]. In one center a comparison group of people living in the same neighbourhoods indicated that after the initial 10 weeks of follow-up there was no difference in the prevalence of violence by patients and that of their neighbours, when substance abuse was absent in both groups. Violence in both groups was usually committed at home and targeted at family and friends. The context of this violence was not explored in detail, although it is evident that severely mentally disordered patients are often themselves the victims of violent behaviour \[[@B18],[@B19]\]. It is probable that a substantial proportion of aggressive acts may be provoked by aggression and coercion towards the patients themselves. Finally, although it is clear that psychoactive medication can induce a substantial reduction in aggressive behaviour \[[@B20],[@B21]\], and that non-adherence to community-based treatment is a predictor of severe violence in schizophrenia \[[@B10],[@B22]\] the effects of long-term optimal treatment programmes on the level of aggressive behaviour have seldom been reported. It may be concluded that the hypothesis that aggressive and sexually inappropriate behaviour is more common in people suffering severe schizophrenic disorders, particularly those characterised by persisting psychotic symptoms, with impaired decision-making abilities, including difficulties with adherence to community-based treatment programmes, has not been proven conclusively. There appears to be an increased risk of misconduct in cases when active psychotic experiences provoke anger and frustration, particularly when judgement is further clouded by drugs or alcohol \[[@B11]\]. There is an urgent need to provide reassurance to the community that efforts to prevent offensive behaviour by people with schizophrenia are realistic and that offending is managed according to the same standards of justice expected of all citizens. The present report describes the changes in a broad range of aggressive behaviour and sexual misconduct in a cohort of chronic patients who were receiving continued evidence-based biomedical and psychosocial treatment throughout a four-year period. It was hypothesized that continued treatment would contribute to a lower rate of these behaviours over time. Methods ======= Selection of cases ------------------ 51 consecutive cases of schizophrenia with a DSMIV diagnosis of schizophrenic disorder based on a SCID-IV interview who were resident in central Athens and receiving continued treatment at the Kessariani Community Mental Health Service. These cases were a cohort who consented to participate in a 5-year study of the clinical and social outcome of continued evidence-based treatment pharmacological and psychosocial strategies \[[@B23]\]. Assessments ----------- ### OTP Misconduct Checklist All aggressive behaviour and sexual misconduct was noted on a continual basis over a four-year period. Each year a standardised rating was made by independent assessors, who interviewed staff, patients and key caregivers to ascertain the frequency of nine varieties of misconduct: shouting, verbal and non-verbal threats, pushing, slapping, punching, use of a weapon, inappropriate sexual behaviour, sexual assault, and aggression towards property or animals. The frequency ratings varied from 0 = not at all; 1 = 1--10 times a year; 2 = more than 10 times a year; 3 = at least once a week. Where misconduct was reported the context of the behaviour was reviewed and a rating of provoking circumstances was made: 1 = when clearly provoked; 2 = when mildly provoked; 3 = unprovoked. Inter-rater reliability on 50 co-rated assessments ranged from ICC = 0.78 to 0.95 on the frequency items and 0.71 to 0.89 on the contextual ratings. Two items were not scored on any of the cases. These were use of a weapon and sexual assault. Background, social and clinical variables ----------------------------------------- A standardized structured assessment of all socio-demographic and clinical variables was made at 0, 12, 24, 36 and 48 months. This included age, gender, ethnicity, marital status, living situation, education, employment, duration of disorder, course of disorder, multi-axial diagnosis, clinical impairment, social disability, caregiver stress, social support, cooperation with treatment, justice system involvement, days in gaol, treatment received, and side effects of medication. Data analysis ------------- Frequency counts and percentages were used to present descriptive data. Non-parametric correlations were conducted with Spearman\'s Rho to examine associations between clinical and social measures and levels of misconduct at each assessment point. The Index of Misconductwas a weighted score that accounted for the nature of the misconduct, the degree of provocation as well as its frequency. Standardised scores ranged from 0--3 (0 = no misconduct reported;1 = mild, infrequent or clearly provoked verbal aggression; 2 = moderate; frequent or unprovoked verbal aggression or any act of physical or sexual aggression; 3 = severe:repeated physical or sexual aggression, often accompanied by verbal aggression). Results ======= The cohort is described in Table[1](#T1){ref-type="table"}. There were similar proportions of men and women aged between 20 and 50 years, almost all of whom were unmarried. Most (87%) were living with their parents of other family members. Thirty-one percent had not completed high school education, and two cases had mild mental retardation. The average duration of mental disorders was 14 years, with 2 cases experiencing their first episodes of schizophrenia. On average they had been admitted to hospital 3.3 times. More than half had persistent positive (55%) and negative (58%) symptoms at the beginning of the study. Nearly a third were considered to have a personality disorder, but very few used alcohol or drugs excessively. One fifth experienced moderate or severe side effects of their medication. Almost a quarter (23.5%) had some problems cooperating with their treatment programs. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Description of the cohort of 51 cases of schizophrenic disorders ::: Age: mean years (s.d.) 35.4 (6.9) -------------------------------------------- ------------ Gender: male (%) 25 (49) female (%) 26 (51) Marital status single (%) 46 (90.2) married (%) 1 (2.0) separated/divorced (%) 4 (8.8) Diagnosis:axis 1 DSM-IV paranoid (%) 21 (41.2) disorganized (%) 5 (9.8) catatonic (%) 1 (2.0) residual (%) 8 (15.7) undifferentiated (%) 5 (9.8) schizoaffective (%) 11 (21.6) Axis 2: personality disorder 16 (31.5) mild mental retardation 2 (3.9) Substance use none 36 (70.6) within cultural norms 5 (9.8) likely to cause health or social problems 4 (7.8) likely to cause mental health problems 6 (11.8) Side effects of medication none 25 (49.0) mild 13 (25.5) moderate 7 (13.7) severe 3 (5.9) ::: The proportions of subjects who behaved frequently (10 or more times a year) in an aggressive or sexually inappropriate manner are summarized in Table [2](#T2){ref-type="table"}. Shouting was the most common aggressive behaviour. At the beginning of the study this was reported in one-fifth of cases, followed by verbal threats in one-sixth. Physical and sexual aggression was displayed by less than 10% of cases and in no instance was a weapon of any sort used. Two thirds of cases showed no misconduct of any sort in the year preceding the initial assessment. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Frequent or Very Frequent verbal, physical and sexual aggressive behaviour in 51 outpatients with schizophrenic disorders in the year before and throughout 4 years of evidence-based treatment ::: REPORTED BEHAVIOUR Year 00 (%) Year 01 (%) Year 02 (%) Year 03 (%) Year 04 (%) -------------------------------------- ------------- ------------- ------------- ------------- ------------- Shouting 11 (21.5) 8 (15.7) 7 (13.7) 5 (9.8) 5 (9.8) Verbal threats 9 (17.6) 7 (13.7) 4 (7.9) 4 (7.8) 4 (7.8) Pushing 4 (7.8) 2 (3.9) 0 (0.0) 1 (2.0) 1 (2.0) Slapping 2 (3.9) 2 (3.9) 0 (0.0) 0 (0.0) 0 (0.0) Punching 1 (2.0) 2 (3.9) 0 (0.0) 0 (0.0) 0 (0.0) Use of weapons 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) Inappropriate sexual behaviour 2 (4.0) 1 (2.0) 1 (2.0) 0 (0.0) 0 (0.0) Aggressive sexual behaviour 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) Other antisocial acts 4 (7.9) 3 (5.9) 1 (2.0) 0 (0.0) 0 (0.0) Days in gaol 0 0 0 0 0 At least one aggressive behaviour 18 (35.3) 14 (27.5) 13 (25.5) 13 (25.5) 12 (23.5) Moderate to high index of misconduct 11 (21.6) 6 (11.8) 3 (5.9) 3 (5.9) 3 (5.9) ::: The Index of Misconduct that weighted the severity of aggressive behaviour, both the nature of the acts as well as the presence or absence of provoking circumstances, was at least moderate at the baseline assessment in one-fifth of cases. It is evident that the number of cases showing aggressive and sexual misconduct declined throughout the course of the 4 years of optimal treatment. At the end of this period the number of cases showing frequent aggressive behaviours had halved. More than three-quarters showed no aggressive or sexual misconduct at all, and the index of misconduct had declined from a mean of 0.67 (± 1.03) to a mean of 0.33 (± 0.71), with only 3 cases showing persistent pattern of aggressive behaviour after the first year of the program. Involvement with the criminal justice system was negligible. Predictors of Misconduct ------------------------ The social, demographic, symptom pattern, impairment, disability, caregiver stress, cooperation with the treatment program, medication side effects did not correlate with the index of misconduct at baseline or any other assessment points. The severity of disability (Spearman\'s Rho = .285, p = .042), and use of non-prescription substances (Rho = .280, p = .047) and cooperation with treatment (Rho = .313, p = .025) were associated with the misconduct index at baseline. The GAF score (Rho = -.295, p = .036) was the only variable that showed a significant correlation with misconduct in the first year of treatment. Caregiver stress was consistently, although not significantly, associated with the severity of misconduct over the 4 years of treatment with coefficients ranging from .256 (p = .076) in year one, .352 (P = .015) in year two, .273 (p = .064) in year 3 and Rho = .281 (p = .056) in the fourth year. In the fourth year GAF (Rho = -.315, p = .024) and the impairment rating (Rho = .283, p = .044) were significantly correlated with misconduct. Given the low number of subjects and the modest associations between variables no attempts were made to use multivariate analyses. However, there were no trends towards association between misconduct and gender, persisting positive or negative symptoms, or medication side effects. Close examination of the 21 cases with a paranoid syndrome diagnosis suggested that this symptom profile might predict moderate or severe misconduct. In the year before the study 33% (7/21) of cases with a paranoid syndrome fell into this category, compared to 13% (4/30) with non-paranoid syndromes. In subsequent years the number of cases with moderate/severe misconduct fell to minimal levels, but two of the three persistent cases had a paranoid symptom pattern. Personality disorders that have been associated with misconduct include antisocial, borderline, paranoid, narcissistic and histrionic. We compared this group with those with other personality disorders, or those who did not have a DSM-IV axis two syndrome. In the year before the study 67% (6/9) of cases with a diagnosis of these personality disorders fell into the moderate or severe misconduct category in contrast to 12% (5/42) of those who did not have these vulnerability features. Again, two of the three persistent cases were in this personality disorder subgroup, and the third had a very early onset schizo-affective disorder with a schizoid personality disorder and persisting anxiety. Discussion ========== The present study is the first that we are aware of that examined the pattern of aggressive and sexually misconduct of patients with a diagnosis of schizophrenia over a period of several years during which time all subjects were receiving evidence-based treatment, with high levels of adherence. Although the sample was small, complete data was obtained on all cases. The most remarkable finding was a low rate of physically violent behaviour. Only 7.9 per cent of the patients were reported as having slapped or punched another person within the year before the study, and after four years this had reduced to 3.9 per cent. None of these acts had been reported to the justice system. This rate of violent behaviour is contrasted with substantially higher rates of similar misconduct in the US, UK and Scandinavia \[[@B3],[@B13],[@B15]\] However in the absence of a matched comparison group of people with similar educational and social background suffering similar levels of disability and discrimination, but without schizophrenic disorders, the low rates of misconduct in this sample cannot be used to draw any specific conclusions. Similarly the lack of a matched treatment control group prevents us drawing definitive conclusions that may link the progressive reduction of misconduct to effective treatment. Despite these limitations, several issues are worthy of further discussion. First, it should be noted that serious violence or offensive behaviour among psychiatric patients seems rare in Greece. There is a complete absence of secure units or units specializing in criminal violence anywhere in Greece. There is also a widespread belief amongst psychiatrists that although some patients are violent this is seldom severe. Secondly, it may be noted that only 12% of the sample used any substances that might have exacerbated their mental symptoms and contributed to behavioural problems. This relatively small group included those drinking excessive coffee or other caffeinated drinks as well as use of nicotine in quantities that may have reduced the effectiveness of neuroleptic drugs, in addition to use of alcohol and illicit drugs. Again this is consistent with the relatively low use of drugs in the Greek population \[[@B24]\]. However, these low levels of misconduct may be contrasted by the high level of stigma for patients with schizophrenia that have been reported in recent surveys of the Greek population. Three-quarters of respondents to a survey on attitudes to schizophrenia considered that such patients were a danger to the public because of violent behaviour. While it is possible that media presentation of isolated unprovoked violence perpetrated by the small minority of persistently offensive patients may have contributed to such attitudes, it is also possible that the close-knit family ties foster high tolerance of misconduct. This would suggest that our data may be biased by the unwillingness of family members to report episodes of misconduct towards them or other close associates. Regardless of the absolute levels of misconduct and arguments about concealment and tolerance, the most striking finding we report is the substantial decrease in the levels of misconduct over the five years studied. At the last assessment only 3 (6%) patients showed persisting patterns of aggression, compared to 22% at the start of a project that ensured that optimal biomedical and psychosocial treatment was provided to the cohort. This treatment program was oriented to help patients and their caregivers improve the quality of their lives as well as attempting to eliminate all residual symptoms and problems, including substance abuse, lack of intimate relationships and aggressive behaviour. It seems reasonable to hypothesize that a comprehensive biopsychosocial treatment programme such as this may contribute to a lowered rate of misconduct. To date only the use of neuroleptic drugs in acute psychotic episodes has been associated with a reduction in aggressive behaviours \[[@B20],[@B21]\]. The one-year follow-up study of Steadman and colleagues \[[@B15]\] showed that patients\' levels of misconduct were elevated significantly more than their non-patient neighbours only in the first 10 weeks after discharge from hospital. Such a feature might suggest that community treatment was effective, but it may have also indicated that many patients were still highly symptomatic after relatively brief hospital treatment. The three cases of persistent misconduct were all young single males living with their parents, who were highly stressed by their care. One merely shouted and threatened frequently without physical aggression. He had persistent delusions and negative symptoms as well as a paranoid personality disorder. The second developed schizophrenia aged 13, did not continue schooling and spent much of his adolescence in hospital. He had schizo-affective features of persistent psychotic symptoms with minimal negative symptoms, persistent generalized anxiety and frequent depressive episodes. He smoked cigarettes and drank alcohol regularly. The third had paranoid schizophrenia with persistent delusions and an antisocial personality disorder. His schizophrenic symptoms, social disability and cooperation with treatment improved over the three years and levels of unprovoked violence became much less frequent. These cases are consistent with epidemiological findings that associate antisocial behaviour patterns with persisting psychotic symptoms, personality abnormalities and substance use. It has been hypothesized that intensive cognitive behavioural treatments of the kinds used throughout this project applied at an early stage, perhaps before the onset of overt psychotic symptoms, may prevent criminal misbehaviour in this select group of difficult to manage cases \[[@B17]\]. The present study does not allow us to draw any constructive conclusions about the management of this specific subgroup, although all three showed some improvements in their antisocial behaviours with time. The hypothesis that optimal treatment reduces levels of social misconduct including verbal, physical and sexual aggression is supported by our study. This finding is not new, but previous reports have focused largely on taking anti-psychotic medication on a regular basis. Optimal pharmacotherapy alone would fail to explain the continued improvement over 4 years of an intensive rehabilitation programme based on the implementation of evidence-based pharmacological and psychosocial treatment strategies in a well integrated goal-oriented system of continued care. Although cooperation with every aspect of the treatment programme was not always enthusiastic, there were no drop-outs over the four years of the project, even in cases where progress was minimal. In all cases the carers, usually family members, were included in the treatment team. Their personal concerns and goals were considered crucial, in particular when aggressive and antisocial behaviour by the patients affected their own well-being. Strategies for managing anger and frustration and the ability to communicate unpleasant feelings or desires in a non-threatening problem oriented way was at the basis of continued family interventions \[[@B25]\]. This as well as the introduction of innovative medications may have contributed to the decline in the severity and frequency of misconduct both at home and in the community. Further research is needed so that community care for the seriously mentally disordered can be supported in a rational manner. In the absence of clearer predictors of antisocial acts, there is a strong tendency for mental health services to take a conservative view and to provide custodial care for the majority of cases in order to prevent the unacceptable acts of a small minority. As a recent reviewer of the links between mental illness and violence concluded \"The challenge that lies ahead is one of further specifying the form of the relationship of mental illness and community violence and testing theories of how this relationship can differ across subgroups of mentally ill persons. Uncovering a broad relationship does not in itself promote sounder policy or more effective services. However, continued integration of research and service provision sensitive to what this relationship means in the lives of people with mental illness could move us toward this goal\" \[[@B26]\]. Conclusion ========== Aggressive and sexual misconduct is present in a minority of patients with schizophrenic disorders, and is commonly directed towards family members and friends. Comprehensive continued treatment with evidence-based biomedical and psychosocial treatment is associated with a reduction in this misconduct. However, a small group of persistent offenders characterised by continuous psychotic symptoms, personality disorders and substance use improve but may require more specialised or intensive interventions. The small (but complete) inner city cohort that was studied prevented us from conducting correlational analyses of predictors of persistent misconduct. A larger cohort from the collaborative Optimal Treatment Project may provide predictive data in future reports. Competing Interests =================== The author(s) declare that they have no competing interests.	PubMed Central	2024-06-05T03:55:57.565285	2005-4-28	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1151595/", "journal": "Clin Pract Epidemiol Ment Health. 2005 Apr 28; 1:3", "authors": [ { "first": "Marina", "last": "Economou" }, { "first": "Alexandra", "last": "Palli" }, { "first": "Ian RH", "last": "Falloon" } ] }
PMC1151596	To our readers ============== This commentary is the first of a series designed to introduce the new online journal Clinical Practice and Epidemiology in Mental Health. Based on the characteristics that were described in the first editorial and by means of a series of introductory commentaries, the journal will seek to stimulate discussion of the main psychiatric topics facing clinical and epidemiological research across the scientific community. An editorial board of international renown will both chair these debates and actively contribute to this series. . It is no mere chance that the first commentary deals with bipolar disorders, arguably a leading issue in current psychiatric research. In spite of the importance of the topic, a number of researchers maintain that it is not adequately taken into account in the medical field, probably due to the marked inconsistency in the available epidemiological data. Bipolar disorders today ======================= Are we right then in defining the field of bipolar disorders as a poorly recognized medical problem? Apparently not. The World Health Organization has defined bipolar disorders as one of the leading causes of disability throughout the world \[[@B1]\]. In the fields of clinical practice and prevention, it is often underlined that bipolar disorders represent a devastating risk factor for both suicide attempts and suicide itself \[[@B2]\]. It is a well-known fact that subjects affected by this disorder are grossly penalized in the area of employment \[[@B3]\] The high costs of bipolar disorders =================================== The financial implications of bipolar disorders are only just beginning to be taken into account. The first studies carried out on the costs of bipolar disorders indicate expenditures ranging between 24 and 30 billion US\$ in the United States over a one year period \[[@B4]\]. Data obtained from an American insurance company reveal how patients affected by bipolar disorders, i.e. 3% of subjects seeking medical assistance, account for 12% of total expenditure \[[@B5]\]. The author of the paper concludes that this is \"the most expensive behavioral health care diagnosis\". However, the same author reports that this expenditure is largely concerned with the cost of inpatient care of subjects diagnosed years after onset of the disorder, suggesting that the financial burden could therefore be lessened. On the other hand, very few data are available on indirect costs, although these, too, are estimated to be extremely high. A study performed in the United States in 1991 (published in 1995) reported that indirect costs represented 83% of the total expenditure \[[@B6]\]. Is the prevalence of bipolar disorder really low? ================================================= However, if we analyze data from surveys of large samples at both national and trans-national level, the prevalence rates of bipolar disorder are dramatically low. The ECA study reports a prevalence rate for bipolar disorders of 1.5% in the general population, of which only 0.3% are bipolar II. The National Comorbidity study indicates a lifetime prevalence for mania and hypomania of 1.6% \[[@B7]\]. More recent data from the Netherlands (NEMESIS Study) also suggested a low prevalence rate for bipolar disorder of 1.9% \[[@B8]\]. The as yet unpublished findings reported by the recent multi-center European study ESEMED reveal even lower frequencies, under 1%. Although several reviews of studies performed on smaller samples do not exactly confirm these findings, it is the larger studies which determine the opinions of the managers of health and research programs. Accordingly, bipolar disorders would not appear to have the same degree of impact on public health as major depressive disorders. Indeed, the leading international studies on major depressive disorders show a lifetime prevalence rate ranging from 3 to 17% in western societies, with an upward trend being evidenced in recent research projects \[[@B9]\]. A main question is whether the low prevalence rates of bipolar disorders are not an artefact of the over-diagnosis of depression and under-diagnosis of bipolar-II. The definition of bipolar spectrum disorders ============================================ The current definitions of mixed state, hypomania and bipolar disorder II are not, however, universally accepted. In clinical practice very few psychiatrists apply a diagnostic threshold for mixed states as high as that indicated by DSM-IV (full criteria for a depressive and manic episode). The impressions held by many clinicians are supported by the findings of a twenty-year longitudinal community study carried out in Zurich \[[@B2]\], which found that depressives with a subthreshold hypomanic syndrome were similar to bipolar II disorders in terms of positive family history for mania, course, comorbidity and treatment rates. Moreover, sub-threshold manic symptoms in adolescence appear to be highly predictive of the subsequent onset of manic episode \[[@B10]\]. Decisive sources of information and under-diagnosis of bipolar disorders ======================================================================== A further explanation for the apparently low prevalence reported by epidemiological studies may be that the methodological instruments used have led to an under-diagnosis of cases of bipolar disorder. Analysis of the clinician\'s logical inferential diagnostic process, particularly when diagnosing bipolar disorder, confirms that the patient does not represent the sole source of useful information. At times, the patient\'s spouse, a relative or a close friend will refer fundamental information, often because many patients do not experience hypomania as distress but rather as recovery from depression or as a period during which they felt truly well. One of the clinician\'s tasks is to cross check the statements of the patient and significant others and to act as a mediator between them. Accordingly, the clinician will attempt to convey relatives\' comments to the patient in order to create an awareness that others may view his behavior as pathological. Moreover, in the course of the logical procedure leading to diagnosis, it is not only what the patient says that should be taken into account, but also the way in which he or she says it and the coherence of his or her ideas. It is not the patient\'s views that determine the diagnosis, but rather the clinician\'s judgment. We are all well aware of how far the views of a patient in a hypomanic state may differ from the clinical judgment, particularly in the case of a patient who has never received any form of treatment and who therefore has not yet accepted the medical model of the illness. In the IOWA STUDY (a prospective investigation of the hereditary nature of schizophrenia and bipolar disorders) the prevalence of mania among relatives of patients affected by bipolar disorders, calculated solely on the basis of diagnostic interviews, was 1.9%; however, when additional sources of information such as clinical records and cross-interviewing of relatives were considered, the rate increased to 5.3% \[[@B11]\]. Interviews ========== Epidemiological data are derived from interviews carried out by lay staff, which only reflect the patient\'s point of view. These instruments may at times be so highly structured that they do not allow any interaction between the clinician\'s judgment and the diagnostic algorithm. The accuracy of diagnosis of the lay interviews is measured by comparison with the so-called \"clinical\" interviews, which are again based on the patient as the source of information. Less structured and more clinically-oriented tools such as SCAN \[[@B12]\] are rarely used for the validation of epidemiological interviews. Basically then, the instruments used to validate epidemiological interviews are semi-structured conversations carried out by clinicians, although the diagnostic algorithm continues to be based to a large degree on patient\'s response. The gold standard applied for validating the diagnosis of bipolar disorder is much poorer than the free clinical judgment would be. In view of all these limitations the degree of reliability remains low. The clinical monitoring study carried out alongside the ESEMED project found for the diagnosis of mood disorders, a Kappa agreement (versus a clinical interview) which ranged from 0.23 in Spain to 0.49 in France \[[@B13]\]. The researchers considered this to be satisfactory! If we consider exactly what a Kappa of 0.4 implies for a disorder with an \"identified\" prevalence rate of 2%, we discover that the prevalence rate may have been under-diagnosed approximately 1.5-fold, so 67% of cases may not have been identified and fifty percent of the identified cases may be false positives (table [1](#T1){ref-type="table"}). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Simulation of the agreement between two instruments where Kappa = 0.4 for a condition with a prevalence of 3% (SCID) identified as 2% ::: SCID + SCID - TOTAL -------- --------------- ------------------- ------- Bipolar cases Not bipolar cases CIDI + 1 1 2 CIDI - 2 96 98 TOTAL 3 97 100 ::: It is legitimate to surmise that the prevalence reported by recent (extremely costly) epidemiological surveys may be doubtful. The way ahead ============= Which direction should epidemiology take in dealing with the serious matter of bipolar disorders? Recently, some community surveys carried out in the USA using the Mood Disorder Questionnaire as a screening tool suggested a prevalence for bipolar disorders of slightly less than 4% \[[@B14]\]. In the ensuing debate, one side claimed that the instrument was scarcely accurate when used in the general population, gave rise to numerous false positives and that the high prevalence reported was therefore a mere artefact \[[@B15]\]. The other side defended the results reported by the research studies, on the basis that \"positive\" cases were homogeneous with regard to the high level of subjective distress, low social functioning and employment and with the high recourse to health care structures \[[@B16]\]. It is quite probable that the problem lies at the root of the matter, in the definition of the gold standard. In the present state of our knowledge on course and response to treatment, the current diagnostic thresholds applied for mixed states and hypomanic episodes seem to be unsatisfactory. If we accept the arguments put forward here, it is inconceivable that the diagnostic gold standard should be determined only on the basis of a structured interview of patients alone. But unless there is clinical consensus on the diagnostic threshold for hypomania and mixed states, there can be no consensus on the findings of epidemiological research. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= MGC and JA conceived of the manuscript, and drafted it. Both authors read and approved the final manuscript. Acknowledgements ================ To Elizabeth Castellani.	PubMed Central	2024-06-05T03:55:57.567901	2005-4-28	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1151596/", "journal": "Clin Pract Epidemiol Ment Health. 2005 Apr 28; 1:4", "authors": [ { "first": "Mauro Giovanni", "last": "Carta" }, { "first": "Jules", "last": "Angst" } ] }
PMC1151597	Introduction ============ From the point of view of terminology, European suicidology seems to be oriented towards defining parasuicide as all non-fatal suicidal behaviours, regardless of their intentional nature (which is often difficult to investigate a posteriori). In other words, parasuicide covers behaviours that may vary from an exclusively manipulatory attempt, to an intentionally suicidal gesture and a serious act that was not fatal by chance (including attempted and failed suicide). However, the current difficulty in unequivocally defining and officially classifying suicidal behaviours (suicide, attempted suicide, failed suicide, parasuicide) is demonstrated by the substantial absence of specific diagnostic criteria in the most widely used diagnostic and epidemiological manuals. The DSM \[[@B1]\] manual limits itself to including self-harming gestures among the diagnostic criteria of other psychiatric categories (major depression, borderline personality disorders), while ICD \[[@B2]\] classification is more specific and speaks of suicide and self-inflicted injury, including injures in suicide and attempted suicide, and self-inflicted injury specified as intentional. The incidence of parasuicidal behaviours is ubiquitously higher among females, with a female/male ratio of between 0.71:1 and 2.15:1, with a median of 1.5:1 \[[@B3]\]. In a study aimed at verifying the trend of parasuicides from drug self-poisoning in 1987--1988 and 1992--1993, Bialas et al.\[[@B4]\] confirmed the prevalence of such behaviours among females and a female/male ratio of 1.13:1. The peak rate of parasuicidal acts is recorded between the ages of 15 and 34 years, with the minimum incidence being reached after 55 years \[[@B3],[@B5]\]. Similarly, the incidence of drug self-poisoning attempts peaks at the age of 15--25 years \[[@B4],[@B6]\]. At least half of the patients making suicidal gestures do so using prescription drugs \[[@B7]\]. The must frequently involved are psychotropic medications, which are used in 80% of the cases of fatal deliberate self-poisoning and 68% of parasuicides \[[@B8],[@B9]\]. Over the last few decades, there has been a considerable change in the types of drugs more frequently responsible for self-poisoning: barbiturates were the principal cause in the 1960s but, since the 1980s, the benzodiazepines have been the most frequently involved \[[@B10]\]. The data of Mc Loone and Crombie \[[@B6]\] indicate an increase in the number of cases of paracetamol intoxication for parasuicidal purposes as from the 1980s, followed by analgesics, antirheumatics, antedepressants and antipsychotics. These data were confirmed by Bialas et al. \[[@B4]\], who found that paracetamol was used for parasuicidal acts in 43.3% of the cases in 1992--1993, as against 31.3% in 1987--1988. The involvement of antidepressants increased from 11.3% in 1987--1988 to 17.6% in 1992--1993 \[[@B4]\], whereas there has been a trend towards a decrease in the parasuicidal use of benzodiazepines, particularly among women \[[@B6],[@B11]\]. The data of Alsen et al. \[[@B9]\] indicate that there is no significant difference in the distribution of the drugs used for parasuicide and completed suicide but, according to Michel et al. \[[@B8]\], only barbiturates are significantly associated with fatal self-poisoning, whereas there is no significant difference in the case of tricyclic antidepressants, which are involved in 13% of deaths due to self-poisoning and 10% of parasuicidal acts \[[@B8]\]. The estimated annual incidence of parasuicides in Europe is between 300 and 800/100,000 inhabitants aged more than 15 years, with significant inter-country differences \[[@B12]\]. Meehan et al. \[[@B13]\] suggest that more than 75% of non-fatal suicidal gestures are not included in the official figures. Current knowledge of the real incidence of parasuicides in Italy is particularly limited by the lack of official statistics, which almost exclusively relate to fatal cases. Parasuicidal behaviour is one of the most significant risk factors for death due to suicide, given that 30--60% of such deaths are the outcome of a series of characteristically repeated attempts \[[@B14],[@B15]\]. The aim of this study was to describe the demographic and clinical variables of a sample of subjects admitted to the Poisoning Treatment Centre (CAV), Niguarda Hospital, Milan, following drug self-poisoning during the years 1999 and 2000. In addition to contributing to the epidemiological definition of parasuicide in our area, it also had the aim of identifying the risk factors correlated with such behaviours. Materials and methods ===================== The study included all of the patients attending the Emergency Department, Niguarda Hospital, Milan, in 1999 and 2000 who satisfied the following criteria: • A diagnosis of self-inflicted poisoning by drugs according to the ICD-9 criteria \[[@B2]\] • A stay of at least one day in the hospital\'s Poisoning Treatment Centre (CAV) Two hundred and one subjects satisfied these criteria (106 in 1999 and 95 in 2000), all of whom came from the Province of Milan, whose resident population at that time was 3,773,893 inhabitants with a slight prevalence of females (51.7%) \[[@B16]\]. Evaluation of the study subjects -------------------------------- At the time of admission, all of the patients underwent clinical and toxicological evaluations. Clinical evaluation: • Physical examination • Hemochromocytometric examination with leukocytic formula • Markers of renal and hepatic function • Plasma electrolytes, glycemia and cholesterolemia • Arterial pressure and ECG monitoring Toxicological evaluation: • Plasma levels of the taken drugs • Plasma levels of substances of abuse, if any • Alcoholemia: it needs to be specified that alcohol intoxication at the time of the gesture was assessed exclusively on the basis of the presence of alcoholemia regardless of plasma alcohol levels because the extreme variability in the time between the parasuicidal gesture and hospital admission makes it difficult to establish the amount of alcohol actually consumed. After the stabilization of their medical condition, all of the patients underwent a psychiatric evaluation during which the specialist collected the following information: • The circumstances related to the parasuicide • Previous parasuicidal episodes • Previous contacts with psychiatric services • Current and previous history of drug and alcohol abuse • Concomitant organic diseases having a clinically relevant psychopathological impact or affecting life expectancy (cardiovascular disturbances, oncological diseases, acquired immunodeficiency, diabetes mellitus). When appropriate, at the end of the clinical interview, the psychiatrist formulated a psychiatric diagnosis on the basis of the DSM IV criteria \[[@B1]\]. The data were statistically analyzed descriptively, using Student\'s T test to compare continuous variables and the chi-squared test to compare discrete variables. Results ======= The selected sample consisted of 201 patients: 73 males (36%) and 128 females (64%) with a mean age of 40 years (± SD 15.29; range 17--89). The collected data indicate that parasuicides from drug self-poisoning accounted for 59.8% of the total of 336 CAV admissions in the years 1999--2000 (CAV database). The mean time between the committal of the parasuicidal act and Emergency Department admission was 4.8 hours ± SD 6.03 (range 0.5--48.0). Drugs taken ----------- In making their parasuicidal act, 118 subjects (58.7%) used at least one benzodiazepine, 34 (16.9%) at least one classic neuroleptic, 9 (4.4%) at least one atypical antipsychotic agent, 26 (12.9%) at least one second-generation antidepressant (SSRI, SNRI, NASSA), 11 (5.4%) at least one tricyclic antidepressant, 12 (5.9%) at least one mood stabilizer, 4 (1.9%) at least one antiepileptic drug (barbiturates), 21 (10.4%) at least one non steroidal anti-inflammatory drug (NSAID), 15 (7.4%) at least one antihypertensive agent, 8 (3.9%) at least one antispastic drug, 6 (2.9%) at least one antiallergic drug, 4 (1.9%) at least one antibiotic, 3 (1.4%) at least one diuretic, 4 (1.9%) at least one hypoglycemic agent, 2 (0.9%) at least one antiemetic, 2 (0.9%) at least one antiarrhythmic agent and 11 (5.4%) drugs with other indications \[Table [1](#T1){ref-type="table"}\]. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Drug classes used for parasuicide purposes and their relative frequency of use. ::: Pharmacological class No. of subjects \% frequency ----------------------------------------- ----------------- -------------- Benzodiazepines 118 58.7 Classic neuroleptics 34 16.9 Atypical antipsychotics 9 4.4 SSRIs, SNRIs, NASSAs 26 12.9 TCAs 11 5.4 Mood stabilizers 12 5.9 Anticholinergics 4 1.9 Antiepileptics 4 1.9 NSAIDs 21 10.4 Antihypertensives 15 7.4 Antispastics, prokinetic antidiarrhoics 8 3.9 Antiallergics 6 2.9 Antibiotics 4 1.9 Diuretics 3 1.4 Hypoglycemic agents 4 1.9 Antiemetics 2 0.9 Antiarrhythmics 2 0.9 Alcohol 23 11.4 Substances of abuse and methadone 3 1.4 Miscellaneous 11 5.4 SSRIs: Serotonin Selective Re-uptake inhibitors; SNRIs: Serotinin Noradrenaline Re-uptake Inhibitors; NASSAs: Noradrenaline Selective Serotonin Antidepressants; TCAs: Tricyclic Antidepressants; NSAIDs: Non Steroidal Anti-Inflammatory Drugs ::: Twenty-three subjects (11.4%) had consumed alcohol at the time of their parasuicidal act, but in combination with drugs in all cases, and three (1.4%) had taken a substance of abuse (cocaine, heroin) or methadone, in all cases in combination with drugs for different indications \[Table [1](#T1){ref-type="table"}\]. Gender ------ The study population included significantly more females than males (64% vs36%), but their mean age was not significantly different (females 39 years ± SD 14.69 vs males 41 years ± SD 16.31). In comparison with the population of origin (subjects resident in the Province of Milan in that period), the proportion of females in the study sample was significantly higher (51% vs 64%; chi-square = 11.125; p \< 0.001) \[[@B16]\]. Age-group analysis revealed that the incidence of parasuicides peaked between the ages of 21 and 30 years among the females (32% of the sample) and between 31 and 40 years among the males (40% of the sample). Comparison of the age distribution of the residents in the Province of Milan with that of the study population showed that females aged 21--30 years were over-represented in the latter (32% vs 13%), as well as males aged 31--40 years (40%vs 18%) \[[@B16]\]. The results of the analysis of clinical variables by gender were as follows \[Table [2](#T2){ref-type="table"}\]: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Clinical-anamnestic variables by gender. ::: Males (%) Females (%) Significance --------------------------- ----------- ------------- --------------------- History of parasuicide 30.8 30.4 No Significance No history of parasuicide 69.2 69.6 Drug addiction 16.4 4.0 Chi-square = 10.423 Alcoholism 19.2 15.7 p = 0.005 No history of dependence 64.4 80.3 Mood disorders 42.0 64.6 Chi-square = 8.347 Anxiety disorders 6.0 1.3 p = 0.039 Schizophrenia 22.0 10.1 Other 30.0 24.0 ::: • Despite the prevalence of females in the study population, there was no significant difference between gender in terms of a history of previous parasuicidal gestures (about 30% in both groups). • The frequency of alcoholism and drug addiction was significantly lower among the females (p = 0.005). • The distribution of psychiatric diseases was significantly different in the two groups, with mood disorders being more prevalent among females (64% vs42%) and schizophrenia more prevalent among males (22% vs10%) (p = 0.039). Sociodemographic variables -------------------------- The sample consisted of 169 Italians (85%), six citizens of the EU (3%) and 24 citizens of non-EU countries (12%). Sixty percent of the subjects were unmarried; the remaining 40% were married or co-habiting. In terms of occupation, 55% of the sample were students or employed, 16% unemployed, 21% pensioners, and 8% housewives. Clinical characteristics ------------------------ Sixty-one subjects (30%) had a history of previous parasuicidal acts. Analysis of the clinical symptoms present at the time of admission to the Emergency Department showed that 24 subjects (12%) were free of symptoms, 153 (76%) had symptoms attributable to central nervous system (CNS) depression (psychomotor slowing, drowsiness) and 24 (12%) had symptons of CNS activation (anxiety, psychomotor agitation). At the time of admission, 40 subjects (20%) had altered laboratory test results (hemochrome with leukocytic formula, hepatic and renal function markers, plasma electrolytes, glycemia, cholesterolemia) and 23 (11.5%) electrocardiographic alterations. ### Psychiatric diagnoses Sixty-nine subjects (34%) did not have a history of psychiatric disorders. Of the other hand 132 (66%) subjects had a previous psychiatric diagnosis: 72 (35%) had a previous diagnosis of mood disorder, 30 (15%) of personality disorders and 19 (9%) of schizophrenia; the remaining cases (7%) had other psychiatric diagnoses, including anxiety and eating behaviour disorders. The mean age of the subjects without a previous psychiatric diagnosis was significantly lower (35.98 years ± SD 14.54 vs42.03 years ± SD 15.32; t = 2.724; p = 0.007). Among the subjects with a previous psychiatric diagnosis, those affected by personality disorders (30) had a mean age of 37.00 years ± SD 13.57, whereas the mean age of the other 102 was significantly higher (43.52 years ± SD 15.54; t = 2.235; p = 0.030). The frequency of previous parasuicidal acts was significantly higher among the patients with a previous psychiatric diagnosis (45% vs11.7%; Chi-square = 19.837; p = 0.000). ### Alcoholism and drug addiction Thirty-five subjects (17.5%) had a diagnostic history of alcoholism, and seven (8.5%) a diagnostic history of substance abuse. The time between the parasuicidal gesture and admission to the Emergency Department was significantly shorter for the alcohol abusers than the non-alcohol abusers (2.6 hours ± SD 3.8 vs4.9 hours ± SD 6.4; t = 2.772; p = 0.007). ### Organic diseases Fifty subjects (24.9%) had clinically relevant organic diseases (HIV, oncological or cardiovascular diseases). In comparison with the rest of the sample, these subjects were significantly older (49.40 years ± SD 16.66 vs36.84 years ± SD 13.46; t = 4.825; p = 0.000) and had previously a significantly higher number of parasuicidal gestures (46.5% vs29.9%; chi-square = 4.018; p = 0.045). Discussion ========== The results of this study indicate that parasuicidal acts involving drugs accounted for a considerable proportion (56.8%) of the admissions to CAV in the years 1999--2000. Furthermore, these cases represented the majority of the self-inflicetd poisoning received by the Centre (the rest of the cases were mainly due to accidental intoxication caused by carbon monoxide, phosphorus poisoning, detergents or the dietary intake of toxic agents). Nevertheless, it is necessary to stress the difficulty in evaluating a posteriorithe correlation between these cases and real suicidal intentions. Drugs taken ----------- The published data agree in indicating that the drugs most frequently used for parasuicidal acts are those used to treat psychiatric disorders \[[@B5],[@B9]\]. In line with these observations, 164 subjects in our sample (81.6%) used one or more psychoactive drugs, including 48 subjects (corresponding to 23.9% of the sample) who combined them with other pharmaceutical specialties. The most frequently used drug classes were the benzodiazepines, which were involved in the parasuicide of 118 patients (58.7% of the sample), followed by classic neuroleptics (16.9%), second-generation antidepressants (12.9%) and NSAIDs (10.4%). These findings confirm the previously observed large-scale use of the benzodiazepines for parasuicidal purposes \[[@B9],[@B17]\], and can be attributed to their widespread prescription and ready availability. However, the low fatality index of the benzodiazepines should not lead us to underestimate the self-harming intentionality of parasuicidal gestures because there may be little relationship between the two, except perhaps in the case of self-poisoning on the part of healthcare workers, who are aware of the degree of lethality of the means used. Furthermore, benzodiazepines are also the most frequently used drugs in cases of fatal suicidal gestures \[[@B9]\]. It is likely that the drugs used to make the parasuicidal gesture are those related the current or previous therapy of the underlying psychiatric disorder. This is confirmed by the fact that 132 subjects (66% of the sample) had a history of a previous psychiatric diagnosis for which they probably received a pharmacological prescription. Our data do not agree with some recent findings indicating NSAIDs (particularly paracetamol) as drugs frequently used for parasuicidal purposes \[[@B4],[@B6],[@B18]\]. A majority of the sample (114 subjects, 56.7%) combined two or more drugs in making their parasuicidal gesture, probably seeking a cumulative effect. The parasuicidal gesture of 63 subjects (31.4%) was related to the improper use of one or more drugs intended for the treatment of organic diseases, including 28 (13.9% of the sample) who used them in combination with psychoactive drugs. This finding is consistent with the fact that 24.9% of our sample were affected by organic diseases. Gender ------ Women accounted for 64% of our study sample, which is in line with the findings of numerous studies confirming the prevalence of females in the ambit of parasuicidal behaviours \[[@B19]-[@B22]\]. Some authors believe that the higher prevalence of females is due to the search for means having less corporeal impact, regardless of whether the intent is really suicidal or not \[[@B5],[@B23]\]. In line with the findings of Batt et al. \[[@B22]\], our results indicate that females are at peak parasuicidal risk at a younger age than males (21--30 vs31--40 years). Among the females, the prevalence of alcoholism and drug addiction was significantly less than among men. The distribution of psychiatric diseases was also different between the two groups: females had a higher prevalence of mood disturbances (64% vs42%) and males a higher prevalence of schizophrenia (22% vs10%). Sociodemographic variables -------------------------- Non-Italians accounted for 16% of our sample: 3% EU and 13% non-EU citizens. The official ISTAT census figures relating to the same period indicate that only 4.6% of the residents in the Province of Milan are not Italians, a difference that is probably related to an underestimate of the real number of foreigners in the Province, as well as to a greater risk of parasuicide among foreigners due to difficulties of adapting to the new environmental context. The high percentages of unmarried (60%) and unemployed subjects (16%) is in line with published data indicating these two conditions as risk factors for parasuicidal gestures \[[@B24],[@B25]\]. Clinical characteristics ------------------------ Twenty-five percent of the sample were affected by severe organic diseases, 66% by psychiatric diseases, and 26% by alcoholism or drug addiction. These data describe a population characterised by considerable somatic, psychic and probably social difficulties. Furthermore, 30% of the cases had a history of previous parasuicidal episodes, thus confirming the substantial to repeat such gestures \[[@B15]\]. Kreitman and Casey \[[@B26]\] point out that only 40--60% of the subjects receiving medical attention after a parasuicidal gesture do not have a previous history of such episodes. The dramatic nature of the acute presentation of the majority of parasuicidal gestures tends to distract attention from the repetitive pattern of these behaviours, but the recognition that they may be one of a series in a significant number of cases implies the need to extend patient monitoring considerably beyond the acute condition. ### Psychiatric diagnoses The majority (66%) of our patients had a history of a previously diagnosed psychiatric disorder: a mood disorder in 35% of cases, a personality disorder in 15%, schizophrenia in 9%, and others (including anxiety and eating behaviour disorders) in the remaining 7%. These findings are in line with the published data, according to which 70--90% of the subjects making parasuicidal gestures have a history of repeated contacts with psychiatric services \[[@B5],[@B21],[@B27]\]. Several studies \[[@B28]-[@B30]\] found that mood disorders, followed by disorders due to substance abuse, are the psychiatric diagnoses most frequently associated with suicidal and parasuicidal behaviours. However, a number of recent clinical observations suggest that the absolute preponderance attributed to these disorders in relation to the risk of suicide needs to be reviewed \[[@B31],[@B32]\]. Much of the published data agree that schizophrenia (particularly its paranoid form) is associated with a high risk of suicide \[[@B33],[@B34]\], and 9% of the patients in our sample were schizophrenic. Given that recent definitions of post-psychotic depression underline the presence of a subsequent affective disorder arising during the course of schizophrenia \[[@B35]-[@B37]\], it is necessary to reconsider the risk of suicide due to affective reasons in schizophrenic subjects. However, the data in our possession are insufficient to clarify whether their parasuicidal behaviours are secondary to the presence of a major affective component or related to the presence of psychotic symptoms and their inadequate treatment. ### Alcoholism and drug addiction The results of this study indicate that 21% of the subjects had drunk alcohol at the time of their parasuicidal gesture and that about 18% had a history of a disturbance due to alcoholism. These results seem to be consistent with those of published studies showing a clear association between alcohol dependence and parasuicidal and suicidal behaviours \[[@B38]\]. Suokas and Lonnqvist \[[@B39]\] indicate that 62% of parasuicides are related to alcohol abuse at the time of the gesture or immediately beforehand. About 40% of the subjects being treated for a disorder due to alcohol dependence have a history of parasuicidal episodes, and about 5% of the subjects with such a disorder commit suicide \[[@B40],[@B41]\]. It is also known that there is a significant association between alcoholism and depressive disorders, on the basis of which it is possible to hypothesise that the affective disorders may contribute towards determining the high incidence of suicidal behaviours among alcoholics \[[@B5]\]. Furthermore, various data indicate that substance abuse is also related to a high incidence of suicidal and parasuicidal behaviours \[[@B5],[@B42]\], and we found that 8.5% of the patients in our sample had a history of diagnosed substance abuse. Chronic substance abuse is often accompanied by a progressive loss of affective-relational ties, a worsening reduction in working function, and consequent social isolation. These conditions favour the onset of depressive experiences and a loss of self-esteem that increase the likelihood of adopting suicidal and parasuicidal behaviours \[[@B43]\]. Like alcohol, substances of abuse can facilitate parasuicidal gestures in various ways: they can be used because of their own self-injuring effects, as a means of self-disinhibition, or in order to increase the lethal nature of pharmaceutical drugs and alcohol. It is therefore useful to underline the need for a wider ranging clinical and anamnestic evaluation of the subjects who are dependent on alcohol or substances of abuse that includes an investigation into the presence of psychic and socio-familial problems, as well as suicidal ideation. ### Organic diseases There are numerous published data showing that the presence of clinically relevant organic diseases is associated with a high parasuicidal and suicidal risk \[[@B5],[@B44]\]. De Leo et al. \[[@B45]\] have pointed out that about half of the people who commit parasuicidal acts have a chronic organic disease. Forty-five percent of the subjects with a chronic disease consider it one of the factors precipitating their parasuicidal gesture, and 22% the determining factor \[[@B45]\]. In line with these observations, our results indicate that many (24.9%) of the subjects making parasuicidal gestures had chronic organic diseases (cardiovascular disturbances, oncological diseases, acquired immunodeficiency, diabetes mellitus) and, in comparison with the rest of the sample, were characterised by an older age (49.40 years ± SD 16.66) and a high frequency of previous parasuicidal episodes (46.5% of the cases). Any disease affecting life expectancy is accompanied by inevitable changes in self-perception and socio-familial role. Maintaining the ability to project oneself into the future is a mental construct that is essential for preventing the onset of the ideation of death and the implementation of suicidal acts \[[@B46]\]. Conclusion ========== In line with those of many previous studies, our results indicate that the patients at highest risk of parasuicide have a number of characteristic traits, including the presence of psychiatric disorders, organic diseases, alcohol/drug dependence/abuse, and significant psycho-relational disturbances. However, it is precisely these characteristics that often hinder the creation of an optimal therapeutic alliance, and make these patients much less compliant to both pharmacological and psychotherapeutic treatment strategies. In line with previous published data, we found that psychoactive drugs (particularly the benzodiazepines) are the most frequently used by people making parasuicidal gestures. Although not fatal (and regardless of the lethality of the drug used), parasuicidal behaviours are characterized by a poor prognosis insofar as they are frequently repetitive and lead to a high risk of death by suicide. It is therefore necessary to establish a therapeutic programme for such patients that covers both the acute situation and long-term prevention. In the case of patients who verbalize suicidal ideations and are being treated pharmacologically (particularly if they are receiving psychoactive drugs), it is essential to choose drugs that are less lethal if an excessive dose is taken.	PubMed Central	2024-06-05T03:55:57.569182	2005-4-28	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1151597/", "journal": "Clin Pract Epidemiol Ment Health. 2005 Apr 28; 1:5", "authors": [ { "first": "Massimo Carlo", "last": "Mauri" }, { "first": "Giancarlo", "last": "Cerveri" }, { "first": "Lucia Sara", "last": "Volonteri" }, { "first": "Alessio", "last": "Fiorentini" }, { "first": "Alessandro", "last": "Colasanti" }, { "first": "Sergio", "last": "Manfré" }, { "first": "Rossana", "last": "Borghini" }, { "first": "Emma", "last": "Pannacciulli" } ] }
PMC1151598	Introduction {#s1} ============ A central question in the biology of host--parasite interactions is how a balance between the costs and benefits to both is achieved. If the burden to the host is too high, the parasite will go extinct, and for this reason it is often postulated that parasites confer some benefit upon their hosts, thereby arriving at an equilibrium in the competition between parasite-free and infected host populations. Bacterial viruses are parasitic. Roughly speaking, they either invade and kill the host or they invade and lie dormant, either integrated into the host genome or outside it as extrachromosomal elements \[[@pbio-0030229-b01]\]. Phage that can exist in a dormant state are called temperate phage, and bacteria carrying temperate phage are said to be lysogenic. In the lysogenic state, viral functions needed for replication and packaging are shut down by a phage-encoded repressor. Occasionally, a temperate phage genome escapes repression, the virus begins to replicate, and soon the host cell lyses, producing a new generation of viral particles ([Figure 1](#pbio-0030229-g001){ref-type="fig"}A). The lysogenic state thus imposes a cost on the bacterium because every so often the viral genome (prophage) replicates and kills the host. This selective disadvantage is offset, however, in several important ways. Because prophage produce a repressor that keeps the prophage genes from being expressed, the host bacterium enjoys immunity from lytic infection by temperate family members. A second, and different, mechanism confers immunity to lytic phage, those phage that cannot exist in the host as prophage \[[@pbio-0030229-b02],[@pbio-0030229-b03]\]. Temperate phage may also confer fitness on a host by coding for genes that enhance host virulence and resistance to the immune system \[[@pbio-0030229-b04]\], of which there are dozens of examples: the Shiga toxin produced by some strains of Escherichia coli; the β toxin produced by Corynebacterium diptheriae, the causative agent of diphtheria; endotoxin production by Clostridium botulinum; staphylococcal endotoxins; and cholera toxin produced by Vibrio cholerae, to name a few. Then too, prophage often code for functions that allow the lysogen to successfully colonize the animal host \[[@pbio-0030229-b04]\], and in general, temperate phage increase horizontal gene flow in microbial populations \[[@pbio-0030229-b05]\]. Thus there are several advantages to being a lysogen, and at least one big disadvantage: Occasionally the prophage replicates and kills the host. ::: {#pbio-0030229-g001 .fig} Figure 1 ::: {.caption} ###### The λ Life Cycle and Gene Organization \(A) The lytic compared to the lysogenic life cycle. Invading phage can either replicate and lyse the host cell (1), or they can integrate into the host genome (2) where they replicate as part of the host genome. At a frequency of approximately 10^−6^ to 10^−5^ per cell division, the prophage escapes cI repression, replicates, and lyses the host (3) \(B) The genetic map showing the approximate organization of functional blocks and, below the map, the location of the genes up and down regulated as revealed in this study. The lom gene is embedded between two different tail-fiber genes. The genome length is 48.5 kb, approximately 75 genes. See \[[@pbio-0030229-b02],[@pbio-0030229-b29]\]. ::: ![](pbio.0030229.g001) ::: Of the many prophage that litter bacterial genomes, the best studied is λ. When it recombines into the E. coli chromosome, phage multiplication is shut down by the phage-encoded repressor, cI, a critical element of the genetic switch. The continuous low-level production of cI protects the host against further infection by extracellular λ, while also regulating the levels of cI synthesis intracellularly (reviewed in \[[@pbio-0030229-b06]\]). The phage also codes for genes required for replication, maintenance, integration, and escape from the host cytoplasm, as well as a series of genes not required for growth in the laboratory ([Figure 1](#pbio-0030229-g001){ref-type="fig"}B). Most λ genes are repressed by cI, but there are several whose transcription is constitutive, their expression either dependent or independent of cI control. Two, the products of the rexA and rexB genes, exclude productive infection by the unrelated lytic bacteriophage T2, T4, and T6 \[[@pbio-0030229-b02]\]. The Rex proteins have also been reported to increase the advantage of lysogens in competition experiments \[[@pbio-0030229-b07]\], but there are conflicting results in the literature on this point \[[@pbio-0030229-b08]\]. Two other proteins, Bor and Lom, are found in the host outer membrane, and bor lysogens are resistant to guinea pig serum \[[@pbio-0030229-b09],[@pbio-0030229-b10]\]. Each of the above examples illustrates how prophage-encoded λ genes increase lysogen fitness by coding for a protein that protects the host from invaders or from the humoral system. Here we ask if the phage repressor directly or indirectly regulates host genes. There is indirect evidence that a streptococcal temperate phage may regulate a bacterial gene that protects cells against phagocytosis \[[@pbio-0030229-b11]\], but there have been no systematic studies on this subject. By surveying both the host and phage genomes with microarrays, we have found several new and unexpected expression patterns in E. coli lysogens, in addition to those viral genes known to be expressed. One in particular, a host gene partly responsible for gluconeogenesis, pckA, is down-regulated many fold, and this leads to a growth disadvantage for the lysogen when grown on succinate, a common carbon and energy source. The DNA sequence lying upstream of the pckA coding region contains sequences homologous to the DNA-binding site for cI, and, surprisingly, for the cI homologs of other temperate phage. Thus it appears that down-regulation of pckA is part of an adaptive strategy for many different temperate phage. Results {#s2} ======= Expression Profiles {#s2a} ------------------- Our arrays were designed to assay the expression levels of 4,539 E. coli and 73 λ open reading frames during exponential growth in tryptone broth, standard growth conditions for experiments with λ lysogens \[[@pbio-0030229-b02]\]. The results are summarized in [Figure 2](#pbio-0030229-g002){ref-type="fig"} and [Table 1](#pbio-0030229-t001){ref-type="table"}, in which we restrict our analysis to changes in gene expression levels of approximately 2-fold (\\|log ~2~\\| levels of approximately one) or more. ::: {#pbio-0030229-g002 .fig} Figure 2 ::: {.caption} ###### Gene Expression Pattern in Lysogen versus Nonlysogen In this representation, the log of the odds (B) \[[@pbio-0030229-b30]\] is plotted against the log of the ratio of the change (M) for each gene. The larger the odds, the higher the confidence in the fold change. M is log~2~(R~lysogen~/G~nonlysogen~), where R is the signal in the red channel and G the signal in the green. Twelve samples from each of two exponentially growing lysogen and nonlysogen cultures were harvested at 6-min intervals and used to prepare cDNA probes as described in [Materials and Methods](#s4){ref-type="sec"}. cDNA samples from each time point were labeled separately with Cy3 and Cy5, mixed, and then used to probe the microarrays in duplicate. For each time point, Cy3 and Cy5 labels were reversed, and reversed labels were also used to probe the microarrays in duplicate. Thus each time point is the average of four datasets, and the data in [Table 1](#pbio-0030229-t001){ref-type="table"} and [Figure 2](#pbio-0030229-g002){ref-type="fig"} represent 48 arrays probed with cDNA from exponentially growing cells. The dataset on which [Table 1](#pbio-0030229-t001){ref-type="table"} is based may be found in supplemental [Table S1](#st001){ref-type="supplementary-material"}. ::: ![](pbio.0030229.g002) ::: ::: {#pbio-0030229-t001 .table-wrap} Table 1 ::: {.caption} ###### Host and Viral Genes Regulated in λ Lysogens ::: ![](pbio.0030229.t001) ::: Host Gene Expression in λ Lysogens: The pckA Gene {#s2b} --------------------------------------------------- The most striking change is in the expression level of pckA, the gene coding for phosphoenolpyruvate carboxykinase \[[@pbio-0030229-b12]\]. PckA in the presence of ATP converts oxaloacetate to phosphoenolpyruvate \[[@pbio-0030229-b13]\]. Mutant pckA strains grow less well on succinate and other carbon sources that feed into oxaloacetate \[[@pbio-0030229-b14]\], although phosphoenolpyruvate can also be synthesized from succinate via phosphoenolpyruvate synthase \[[@pbio-0030229-b13]\]. We expected that the down-regulation of pckA transcripts in lysogens might lead to a similar problem. [Figure 3](#pbio-0030229-g003){ref-type="fig"} compares the generation times of lysogens and nonlysogens of both λ and another temperate phage, λimm434, on glucose and on succinate as the sole carbon source (the λimm434 cI and operator sequences replace the λ homologs in λimm434; \[[@pbio-0030229-b15],[@pbio-0030229-b16]\]). There is very little difference between lysogen and nonlysogen growth rates on glucose, but the difference on succinate is striking. In glucose, the difference in doubling time is at most a few percent, whereas on succinate it is increased at least 30% in λ lysogens and 20% in λimm434 lysogens. These are substantial changes in fitness: The lysogen will be reduced by 90% in 17 generations, 99% in 34 generations, and 99.9% in 50 generations when the doubling time differs by 20%. ::: {#pbio-0030229-g003 .fig} Figure 3 ::: {.caption} ###### The Growth of Lysogens and Nonlysogens in Glucose and Succinate Media In this figure, pcI codes for cI, and pcI^0^ is the empty vector. See [Materials and Methods](#s4){ref-type="sec"} for the genotypes of these strains. Growth rates were obtained by least-square fit of the optical density versus time curves. Error bars for the growth rates were obtained by fitting two subsets of the data and using the standard deviation of the two estimates. ::: ![](pbio.0030229.g003) ::: In the E. coli genome there are many sequences that exhibit some homology to the λ operator \[[@pbio-0030229-b17],[@pbio-0030229-b18]\]. The highest homology is with a 16/17-base pair (bp) match to O~L~2, and it occurs within the open reading frame of rlpB, a gene coding for an inner membrane lipoprotein. There are four additional sequences with a 15/17-bp match, two to O~R~1 and two to O~L~2. They are in the coding regions of rlpB and pabC for O~R~1 homologs, and hmpA and yjgQ for O~L~2 homologies, none of whose expression pattern is altered by our criteria. By relaxing our search for homologies to 11-mers, we find 244 more sites, but in all of these cases there are gaps in the matching strings. Some of these map in possible regulatory regions of the genome. The strongest λ homologies, however, map to the 370-bp sequence upstream of pckA ([Figure 4](#pbio-0030229-g004){ref-type="fig"}). ::: {#pbio-0030229-g004 .fig} Figure 4 ::: {.caption} ###### The Upstream pckA Region Homologous to Known Temperate Bacteriophage Operator Sites O~R~3, O~L~2, O~R~1, and O~R~2 are sequences known to bind either cI or the cI homologs of the lysogenic bacteriophage H-19B, λ, 434, P22, P21, and 434 shown in brackets next to each operator sequence \[[@pbio-0030229-b06],[@pbio-0030229-b31],[@pbio-0030229-b32]\]. A second P21 O~L~2 site lies further upstream from the −35 region. ::: ![](pbio.0030229.g004) ::: There are homologies to other temperate phage operators in this region as well. Phages H-19B, 434, 21, and P22 are four such examples. The presence of these sequences suggests that the phenotypes of the two lysogens summarized in [Figure 3](#pbio-0030229-g003){ref-type="fig"} might be caused by cI binding to these related operator sites. We first examined this idea by comparing the growth rates of strains carrying a cI-expressing plasmid to the parental strain carrying an empty plasmid (see [Figure 3](#pbio-0030229-g003){ref-type="fig"}, DH5αPro(pcI) versus DH5αPro(pcI^0^)). There is a dramatic increase in doubling time when these strains are grown on succinate as the only carbon source (τ = 139 min vs. 278 min), and very little difference on glucose. These results support the idea that cI expression depresses growth rate, but the data do not distinguish between direct and indirect effects. To make this distinction, we used real-time PCR to measure lacZ message in a strain where lacZ was fused to a 370-bp pckA upstream sequence ([Table 2](#pbio-0030229-t002){ref-type="table"}). Four different strains were compared: W3350 and W3350 (λ) as baseline controls (lines 1--4) and two different strains carrying lacZ on a plasmid (lines 5--10). One of these, DH5αPro(placZpcI), does not have a promoter and serves as a control on lacZ transcript levels. The other, DH5αPro(pP~pckA~lacZpcI), has the lacZ gene fused to a 370-bp upstream pckA sequence. Both strains also carried either cI on a plasmid, or the same plasmid lacking the cI gene. ::: {#pbio-0030229-t002 .table-wrap} Table 2 ::: {.caption} ###### lacZ Is Regulated by cI When It Is Fused to the pckA Promoter Region ::: ![](pbio.0030229.t002) ::: The chromosomal pckA transcript is down-regulated about 3-fold in lysogens compared to wild type (lines 1--4). Apparently cI acts directly on the pckA upstream sequence, because cI expression regulates lacZ transcript levels in strains where lacZ is fused to the pckA upstream sequence (lines 5--10). cI message is undetectable in nonlysogens, as expected (line 2), and the pckA transcript level is down-regulated 3-fold in a lysogen (line 1 compared to line 3). When cI is produced from a plasmid, it depresses pckA message approximately 2-fold (lines 5 and 8). In the pckA promoter-lacZ fusion, although the fold decrease in lacZ transcript level is not as large as it is in the previous examples, it is nonetheless down-regulated 25% to 30%. Because lacZ transcript levels are quite high in strains in which the lacZ gene lacks a promoter altogether (line 7), these smaller changes occur on top of a high background, possibly due to the high copy number of this plasmid. If we correct for this high background using these data, then the difference is more substantial, an approximately 35% reduction in transcript level in the strain carrying upstream pckA sequences. We note that even though the change in transcript number when lacZ is controlled by the pckA promoter region is not large, it is reproducible, and moreover, it is of the same order of magnitude as the changes in growth rate documented in [Figure 2](#pbio-0030229-g002){ref-type="fig"}. Taken together, these data suggest that cI binds to the upstream pckA sequence and down-regulates transcription. To define this interaction further, we asked if purified cI could bind to the pckA upstream sequence in a band shift assay. In these experiments, a 370-bp upstream region ([Figure 4](#pbio-0030229-g004){ref-type="fig"}) was labeled with ^32^P using the polymerase chain reaction and incubated with affinity-purified cI protein. Samples were then analyzed by gel electrophoresis, and binding specificity was analyzed by competition with unlabeled DNA. [Figure 5](#pbio-0030229-g005){ref-type="fig"}A shows that cI binds to the pckA DNA. The radioactive probe is authentic because it can be competed away with unlabeled upstream pckA DNA, as expected ([Figure 5](#pbio-0030229-g005){ref-type="fig"}B). Moreover, unlabeled DNA containing λ operator sequences (O~R~λ) also competes, showing that our cI preparation responds to the authentic target DNA ([Figure 5](#pbio-0030229-g005){ref-type="fig"}C). Finally, in the same assay, increasing quantities of a scrambled λ O~R~1 sequence fail to compete with cI binding, the expected result for a specific interaction between the promoter region lying upstream of pckA and the cI protein ([Figure 5](#pbio-0030229-g005){ref-type="fig"}D). ::: {#pbio-0030229-g005 .fig} Figure 5 ::: {.caption} ###### cI Binds to the pckA Promoter Region In these experiments, affinity-purified cI was added in variable (0 μM, 1.15 μM, and 4.6 μM, lanes 1, 2, and 3, respectively) or constant (4.6 μM, lanes 4--12) amounts to ^32^P-labeled upstream pckA DNA (pPckA\, 0.83 nM each lane). In lanes 4--5, a 5-, 10-, and 30-fold excess of unlabeled pckA* DNA was added as competitor. In lanes 7--9, a 10-, 30-, and 60-fold excess of unlabeled λ O~R~ DNA was the competitor. In lanes 10--12, a 10-, 30-, and 60-fold excess of unlabeled random λ O~R~1 sequence (nonspecific) was added (see [Materials and Methods](#s4){ref-type="sec"}). ::: ![](pbio.0030229.g005) ::: Bacteriophage λ normally infects E. coli, and P22 infects Salmonella. Gluconeogenesis is an important metabolic pathway in most organisms \[[@pbio-0030229-b13]\], and so we have asked if the upstream pckA sequences are highly conserved in bacteria, or whether the cI--pckA interactions reported here are unique to the Enterobacteriaceae. Apparently, the pckA regulatory region is highly conserved in Salmonella, Shigella, and Escherichia species, but not elsewhere. These are all Enterobacteriaceae, suggesting that the cI--host interaction reported here is restricted to Gram negative bacteria (results not shown). Host Gene Expression in λ Lysogens: Other Genes {#s2c} ----------------------------------------------- In addition to pckA, seven other host genes appear to be regulated approximately 2-fold in the lysogen (see [Figure 2](#pbio-0030229-g002){ref-type="fig"}; [Table 1](#pbio-0030229-t001){ref-type="table"}). b0557 is a bor homolog contained in the DLP12 prophage sequence resident in these strains \[[@pbio-0030229-b19]\]. Because it is 91% homologous to λ bor, this apparent increase in expression of a host gene is probably due to hybridization between λ prophage bor transcripts and the DLP12 resident prophage. We have not explored the significance of the six other transcript profiles shown here, other than to note that two up-regulated genes are copper transporters, and, like bor and lom \[[@pbio-0030229-b20]\], membrane proteins \[[@pbio-0030229-b19]\]. Finally, one of the down-regulated genes, b2002, is also embedded in another defective prophage, CP4--44, although it has no sequence homology with λ and is therefore likely to be directly or indirectly regulated by cI. λ Gene Expression {#s2d} ------------------- In general our results are consistent with the literature on λ gene expression patterns in lysogens (summarized in \[[@pbio-0030229-b02]\]). The promoter regulating cI expression is known to regulate the levels of cI, rexA, and rexB, as noted earlier, and our data clearly show elevated transcript levels. Likewise, both lom and bor are known to be transcribed in lysogens \[[@pbio-0030229-b09],[@pbio-0030229-b20]\]. The Bor protein makes λ lysogens more resistant to guinea pig serum \[[@pbio-0030229-b09]\], and the Lom protein is involved in buccal cell adhesion in the gut \[[@pbio-0030229-b21]\]. Both are outer membrane proteins. int required for integration of the prophage into the genome, and xis, required for excision of the prophage from the genome when the prophage escapes repression, are expressed from the cI-independent promoters pL and pI and are active when the lytic/lysogenic decision is made and during early lytic multiplication \[[@pbio-0030229-b06]\]. Our data reveal, however, that int is also expressed in the lysogen, suggesting a possible strategy for stabilizing the prophage by kinetic means, in that elevated levels of Int protein might help stabilize the lysogenic state by shifting the equilibrium toward integration. This possibility would be in addition to the role that cI plays by repressing the functions needed for viral replication and packaging. The product of the E gene is a procapsid protein essential for viral packaging. It is encoded in a polycistronic late message (see [Figure 1](#pbio-0030229-g001){ref-type="fig"}), and is elevated at least 3-fold in the lysogen. E is part of a polycistronic message coding for genes needed for viral assembly, and thus finding elevated levels of E transcript is unexpected. However, in addition to cI repression of λ genes, it has been noted that there are strong Rho-independent transcription terminators flanking the E gene, and this has led to the idea that transcription terminators may add an extra level of prophage regulation, helping to silence gene expression by terminating aberrantly initiated polycistronic message production and translation \[[@pbio-0030229-b22]\]. Our finding of E up-regulation is consistent with this idea. In this context, it may also be significant that a similar termination sequence lies just downstream of int. Finally, our microarray analysis suggests that five other λ open reading frames are also expressed in the lysogen: orf-64, nin221, ea22, kil, andea47. Their transcripts are all elevated 2-fold or more. Whether or not there is a functional reason for this will require additional analysis. In the microarray experiments, it may seem surprising that genes transcribed in the lysogen are not elevated many hundred- or even thousand-fold, because we are comparing ratios of samples with and without the λ genome. One has to remember, however, that background fluorescence is always present, and thus the reading is never zero, even from blank regions of the array. Discussion {#s3} ========== Our results show that there is strong interaction between the host and parasite genome in this model system, and we can ask what role this interaction plays in the evolution of fitness in these populations. We note first that lysogen growth rates on succinate are dramatically reduced both for λ and λimm434 lysogens (see [Figure 3](#pbio-0030229-g003){ref-type="fig"}). These are very large fitness changes, in the 20% to 30% range per generation, a profound disadvantage for lysogens in a competitive environment in which succinate is the carbon source and, by extension, gluconeogenesis is important. The second clear result is that there are at least seven DNA sequences with varying degrees of homology to the λ operator sites upstream of the pckA gene. These sites are homologous to operator sequences used by other temperate phage, and they lie either close to the --10 to --35 polymerase binding site, or between this region and the ribosomal binding site (see [Figure 4](#pbio-0030229-g004){ref-type="fig"}). One, the O~L~2 homology of phage 21 may lie too far upstream to be considered part of the pckA regulatory domain. Were there only one or two λ-specific sequences in this region, we might conclude that they were there by chance alone. However, the many distinct potential cI binding sequences for H-19B, 434, 21, and P22 make the chance hypothesis unlikely. Note that these sites are dissimilar, reflecting the different target sequences for the different phage repressors---they are not simply differences within a single short consensus sequence. Although the λ O~L~2 sequence is at best a half operator, the others show substantial homology, and the 370-bp pckA sequence binds cI in vitro with an affinity at least as strong as the authentic λ consensus sequence ([Figure 5](#pbio-0030229-g005){ref-type="fig"}). The clustering of operator homologies suggests that there is strong selection pressure maintaining these sites and that they are an important aspect of lysogen fitness, one in which the regulation of a host gene, rather than the production of a phage product, confers increased fitness. There is currently insufficient knowledge about whole-genome E. coli expression patterns in different environments to speculate about what these selective pressures might be, but we suggest that lysogens may preferentially survive because of their lower growth rates in a glucose-poor environment---which would be the case, for example, if the immune system more effectively attacks rapidly growing cells in such an environment---or it could be that an aspect of lysogeny itself must be offset by down-regulating the pckA gene. Independent of the mechanism, our main conclusion is that the multiple potential operator sites lying upstream of the pckA gene suggest strong positive selection for this subtle host--parasite interaction, one that can be added to the known advantages of cI expression leading to immunity from superinfecting λ, protection from infection by T2 and T4 phage, and resistance to host serum factors. How do these results compare to other studies with viral-infected cells? Following animal virus infection, many host genes are turned on and off. For example, in one early study of cells infected with human cytomegalovirus, 1,400 of the 12,626 genes surveyed changed by a factor of four or more following infection, and several of these changes pointed to the central role played by the immune system in a productive infection \[[@pbio-0030229-b23]\]. Other studies in a wide variety of infected cells confirm these general results---many host genes are both up- and down-regulated---but to our knowledge these changes in gene expression have not been firmly tied to phenotypic changes in host--virus dynamics. Materials and Methods {#s4} ===================== {#s4a} ### Strains {#s4a1} W3350 is from the E. coli Genetic Stock Center (CGSC \#5976) (<http://cgsc.biology.yale.edu/>, galK2(Oc), galT22, LAM^-^, andIN(rrnD-rrnE)1. DH5αPro is from Clontech (Palo Alto, California, United States), deoR, endA1, gyrA96, hsdR17(r~k~^--^m~k~^+^), recA1, re/A1, supE44, thi-1, Δ(lacZYA-argF)U169, φ80δlacZΔM15, F^--^, λ^--^, P~N25~/tetR, and P~laci~^q^/laci, Sp^r^. λ wild type is λW2 (λ PaPa), a gift from Dr. R. Weisberg, and λimm434 is bio16, a gift from Dr. D. Kaiser. ### Microarray analysis {#s4a2} A total of 4,539 E. coli MG1655 open reading frames and 73 λ genes were amplified by PCR, purified by ethanol precipitation, and verified by gel electrophoresis. The E. coli MG1655 genome open reading frame primer set is from Sigma-Genosys (Woodlands, Texas, United States), and 73 primer pairs for λ genes were designed by using the Primer3 program (<http://workbench.sdsc.edu>) and synthesized by Integrated DNA Technologies (IDT; Coralville, Iowa, United States). A list of the genes and primers is available from the corresponding author. The purified PCR products were resuspended in 50% DMSO and printed onto Corning Ultra GAPS slides (Corning, New York, United States) using OmniGrid from Genomic Solutions (Ann Arbor, Michigan, United States). E. coli W3350 and W3350(λ) were grown in Tryptone medium with aeration, and 12 samples from each culture were taken about every 6 min between OD~600~ 0.1 and 0.6. RNA samples were extracted using the Qiagen RNeasy kit (Valencia, California, United States). RNA labeling and microarray hybridization were as described \[[@pbio-0030229-b24]\]. For each time point, W3350 cDNA samples were labeled with Cy3, and W3350(λ) cDNA samples were labeled with Cy5. The samples were mixed and used to probe the microarrays in duplicate. In addition, for each time point, Cy3- and Cy5-labeling schedules were reversed, and the reverse-labeled cDNA samples were also used to probe the microarrays in duplicate. Thus each time point is the average of four datasets, and the data in [Table 1](#pbio-0030229-t001){ref-type="table"} and [Figure 2](#pbio-0030229-g002){ref-type="fig"} represent 48 arrays probed with cDNA from exponentially growing cells. The hybridized arrays were scanned using an Axon 4000B scanner (Sunnyvale, California, United States), and the data were analyzed using limmaGUI (<http://bioinf.wehi.edu.au/limmaGUI/>) with Bioconductor packages (<http://www.bioconductor.org>). ### Bacterial doubling times {#s4a3} Strains were grown in M9 minimal media \[[@pbio-0030229-b25]\] supplemented with 1 μg/ml B12, and either 0.4% glucose or 0.4% succinate as the only carbon source. ### Band shift assay {#s4a4} λ cI protein was purified using the PRO Tet 6xHN Bacterial Expression System (Clontech) as described \[[@pbio-0030229-b26]\]. The PCR product of the pckA upstream sequence (pPckA), which is the 370-bp E. coli genomic region 3530464 to 3530833 covering the pckA promoter region, was amplified using the primers TGGTTATCCAGAATCAAAAGGTG and GCTCCTTAGCCAATATGTATTGC, and labeled using α-P^32^-dCTP by PCR. Purified λ cI protein and labeled DNA were mixed in Sauer buffer \[[@pbio-0030229-b27]\] with or without competitors (pPckA, λ O~R~, or randomized λ O~R~1 DNA), incubated on ice for 2 h, and run on a precast 10% TBE native polyacrylamide gel from Bio-Rad (Hercules, California, United States). Gels were dried and exposed to x-ray films. The competitor λ O~R~ was amplified by PCR using primers CGTCCTCAAGCTGCTCTTGT and GCGCATTGCATAATCTTTCA, and is 184 bp long. Another competitor, randomized λ O~R~1, was synthesized by IDT and is 17 bp long. ### Real-time PCR analysis {#s4a5} The λ cI expression vector pE133-cI ("pcI" in the text) was constructed by cloning the cI gene (coordinates 37227 to 37940 on the λ genome) into the expression vector pPROTet.E133 (Clontech). The E. coli pckA upstream sequence fused to the lacZα coding sequence, the lacZ promoter sequence followed by the lacZα coding sequence, or the lacZα coding sequence alone were cloned into the SacII/NheI sites of pACYC177. The pE133 vector or the pE133-cI vector and each of the three modified pACYC177 vectors were electroporated together into DH5αPro electro-competent cells (Clontech). The transformed strains were grown under different conditions, depending on the experiment. RNA samples were extracted using Qiagen RNeasy kits. cI, lacZα, and pckA message numbers were analyzed by real-time PCR, using the SYBR Green PCR Master Mix from Applied Biosystems (Foster City, California, United States). 16S ribosomal RNA was used as a reference to normalize message copy numbers, assuming 20,000 copies of the rRNA molecule per cell \[[@pbio-0030229-b28]\]. Primers TGCATCTAGAGGGCCCAATTC and CGGGCCTCTTCGCTATTACG were used to detect lacZα, primers GCATAA CGTCGCAAGACCAA and GCCTAGGTGAGCCGTTACCC were used to detect 16S rRNA, primers ATGCGCGTTAACAATGGTTTGA and TAGTTAACACCCCGCGCTCAT were used to detect pckA, and primers GTTGAAGGTAATTCCATGACC and ACTAGCGATAACTTTCCCCACA were used to detect cI. Supporting Information {#s5} ====================== Table S1 ::: {.caption} ###### Host and Viral Genes Regulated in λ Lysogens Columns A, B, C, and D are the spot coordinates; Column E: The E. coli gene ID number of the original ORFmer primer set; Column F: gene name (λ genes are prefixed with "LMD"); Column G: M = log~2~ of the expression level ratio of lysogen to nonlysogen (see also [Figure 2](#pbio-0030229-g002){ref-type="fig"} caption); Column H: A = log \[(channel 1)(channel 2)\]^1/2^, a measure of the brightness of the fluorescent signal; Columns I, J, and K: Statistic based on the pooled data from 48 individual measurements using 12 different samples taken from exponentially growing cells. The E. coli MG1655 primer set from Sigma-Genosys (Woodlands, Texas, United States) was designed to amplify 4,539 open reading frames and is therefore very heterogeneous in target length. The λ set of 73 primer pairs contained some duplicate targets, for example, LMD\_\_rexA\_\_L and LMD\_\_rexA in which the L set was approximately 300 nucleotides long and the other set contained approximately 70 nucleotides. The LMD primer set is available from the corresponding author (E-mail: <ecox@princeton.edu>. (764 KB XLS). ::: ::: {.caption} ###### Click here for additional data file. ::: Accession Numbers {#s5a} ----------------- The GenBank (<http://www.ncbi.nlm.nih.gov/Genbank/>) accession numbers for the genes and gene products discussed in this paper are b0557 (GeneID 948980), b2002 (GeneID 946514), cI (GeneID 1238683), E (GeneID 2703482), ea22 (GeneID 2703469), ea47 (GeneID 2703536), hmpA (GeneID 947018),int (GeneID 2703470),kil (GeneID 2703471), nin221 (GeneID 2703476), orf-64 (GeneID 2703478),pabC (GeneID 946647), pckA (GeneID 945667), rexA (GeneID 2703537), rexB (GeneID 2703517), rlpB (GeneID 946257), xis (GeneID 2703504), yjgQ (GeneID 1037318), and λ genes (NC 001416). The GenBank accession number for the E. coli MG1655 primer set is U00096. This research was supported by grants from The Defense Advanced Research Projects Agency (DARPA) and the National Institutes of Health. IG is a Lewis Thomas Fellow. We wish to thank Roger Hendrix, Donald Court, and Lynn Thomason for their generous help and suggestions during the course of this work. Competing interests. The authors have declared that no competing interests exist. Author contributions. YC, IG, and ECC conceived and designed the experiments. YC performed the experiments and analyzed the data. YC, IG, SS, and LG contributed reagents/materials/analysis tools. EEC wrote the paper. Citation: Chen Y, Golding I, Sawai S, Guo L, Cox EC (2005) Population fitness and the regulation of Escherichia coli genes by bacterial viruses. PLoS Biol 3(7): e229. bp : base pair	PubMed Central	2024-06-05T03:55:57.571536	2005-6-21	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1151598/", "journal": "PLoS Biol. 2005 Jul 21; 3(7):e229", "authors": [ { "first": "Ying", "last": "Chen" }, { "first": "Ido", "last": "Golding" }, { "first": "Satoshi", "last": "Sawai" }, { "first": "Ling", "last": "Guo" }, { "first": "Edward C", "last": "Cox" } ] }
PMC1151599	Introduction {#s1} ============ The bioinformatics and computational analyses of a large number of completely sequenced genomes suggest that over 30% of all genes encode membrane proteins \[[@pbio-0030230-b01]--[@pbio-0030230-b04]\]. These predicted membrane proteins, identified to contain at least one trans-membrane domain, participate in a plethora of vital cellular activities, including cell signaling, cell--cell interactions, cell adhesion, cell migration and movement, cytoskeleton organization, protein trafficking, viral fusion, secretory and synaptic activities, ion and metabolite transport, and respiratory and photosynthetic electron transport. However, many detailed structures of diverse membrane proteins remain elusive. Thus the discovery and design of new detergents are acutely needed to facilitate membrane protein crystallization and structural studies \[[@pbio-0030230-b04],[@pbio-0030230-b05]\]. Photosystem I (PS-I) is a membrane protein complex that is associated with electron transport and exists as a large, multisubunit complex with dozens of transmembrane-spanning domains \[[@pbio-0030230-b06],[@pbio-0030230-b07]\]. The composition and organization of PS-I complexes have presented a great challenge in the development of experimental conditions for efficient isolation, while maintaining the native organization of the proteins, associated cofactors, prosthetic groups, and their biological functions. The largest membrane complex isolated and crystallized to date is the plant photosynthetic reaction center, PS-I \[[@pbio-0030230-b07]\]. This plant PS-I complex has 16 subunits, containing 45 transmembrane domains and 167 chlorophyll molecules. Because of its antennae size, the production of a low potential reductant, and its nanoscale structure, PS-I has recently been shown to be an attractive choice for integration into solid-state biophotovoltaic devices \[[@pbio-0030230-b08]\]. In order to fabricate protein-based devices, it is crucial to immobilize and stabilize the selected proteins and enzymes for an extended time. Although some progress has been made \[[@pbio-0030230-b09]--[@pbio-0030230-b11]\], little has been reported on the successful immobilization of membrane proteins on surfaces \[[@pbio-0030230-b12]--[@pbio-0030230-b15]\]. Several investigators have reported that bacteriorhodopsin can be immobilized by entrapment in dried xerogel glass and retains partial activity for extended periods \[[@pbio-0030230-b16],[@pbio-0030230-b17]\]. The difficulties associated with solubilization and stabilization of membrane proteins are largely due to exposure of their hydrophobic domains, which are normally buried within the lipid bilayer. These hydrophobic transmembrane segments render the protein particularly difficult to stabilize after removal from the membrane. In particular, they are extremely susceptible to aggregation in aqueous media. In previous decades, a wealth of chemical detergents was synthesized for the study of membrane proteins (<http://www.anatrace.com/>). In some cases these detergents have partly facilitated selective solubilization of membrane proteins from their native environments. Recent evidence suggests, for example, that detergents with short alkyl chains are more effective at solubilization than those with long alkyl chains that often lead to protein denaturation \[[@pbio-0030230-b18],[@pbio-0030230-b19]\]. These new chemical detergents have advanced the study of membrane proteins. Some efforts have been made to design helical peptide- and protein-based detergents \[[@pbio-0030230-b20],[@pbio-0030230-b21]\]. Stroud and colleagues designed the first α-helix peptide detergent, called a peptitergent, which has a hydrophilic face and a hydrophobic face \[[@pbio-0030230-b20]\]. The peptide formed a four-helical bundle in solution when not interacting with membrane proteins. They showed that this peptitergent can solubilize 85% bacteriorhodopsin and approximately 60% rhodopsin for over 2 d \[[@pbio-0030230-b20]\]. Privé and colleagues also designed a monomeric α-helix coupled with two alkyl chains at both N- and C-termini. This hybrid molecule has two faces, one hydrophilic face from the charged residues on one side of the helix that interacts with water and a hydrophobic face on the other that interacts with transmembrane domains of membrane proteins. They showed that this type of hybrid peptide--alkyl detergent solubilized and stabilized several membrane proteins \[[@pbio-0030230-b21]\]. Several other designed helical peptide detergents have been shown to solubilize membrane proteins as well \[[@pbio-0030230-b22]--[@pbio-0030230-b24]\]. However, these peptides require efficient folding, are not easily accessible, are expensive to produce, and are difficult to make for large-scale productions and widespread studies. We have designed and studied a new class of short self-assembling peptide detergents, which are six to eight residues long, for their ability to maintain solubility of membrane proteins. The behavior of these peptide detergents in the absence of proteins has been previously described \[[@pbio-0030230-b25]--[@pbio-0030230-b28]\]. In the absence of proteins, these detergent peptides form stable nanotubes, nanovesicles, and micelles when they are above their critical aggregation concentration (CAC). The behavior of these peptide detergents is similar to that of lipids and other chemical detergents. It is plausible that the protein--peptide detergent interactions may have enhanced stabilizing capability over short-chain alkane compounds since these peptide detergents share similar chemical properties. This increased stability has recently proven useful in the incorporation of spinach PS-I into a solid-state photovoltaic device \[[@pbio-0030230-b08]\]. We here report the stabilization of spinach PS-I complex following chemical detergent isolation. We studied the ability of these peptide detergents to interact with and stabilize isolated spinach PS-I complexes that have been dried under N~2~ on glass slides. The complex stability and organization of the chlorophyll was monitored via the −196.15 ^°^C steady-state emission spectrum of PS-I. We showed that the PS-I complex alone became unstable during the drying process or in the presence of chemical detergents. However, we further demonstrate that one of the designed peptide detergents, acetyl- AAAAAAK (A~6~K), is not only able to preserve the original spectral properties of the PS-I complex, but also stabilizes PS-I in the dry form for a period of at least 3 wk. In light of the recently reported pea PS-I crystal structure \[[@pbio-0030230-b07]\], we also discuss how the chlorophyll pigment--protein complexes may be influenced during dehydration. Results {#s2} ======= Isolation of Spinach PS-I {#s2a} ------------------------- To evaluate the effect of the peptide detergents for stabilizing membrane proteins, we isolated the chlorophyll pigment protein from spinach chloroplast thylakoids. PS-I was isolated from spinach using a modified procedure of Mullet and Arntzen \[[@pbio-0030230-b29]\]. The typical results of the continuous sucrose gradient used for isolation are shown in [Figure 1](#pbio-0030230-g001){ref-type="fig"}. The large green band on the bottom 2.2-M sucrose cushion is collected using a needle syringe ([Figure 1](#pbio-0030230-g001){ref-type="fig"}A). The additional chlorophyll at the top of the gradient is from LHCPII and PS-II. This preparation has a typical chlorophyll/P700 ratio of approximately 200 and a chlorophyll a/b ratio of 4.75 (data not shown). The SDS-PAGE gel analysis ([Figure 1](#pbio-0030230-g001){ref-type="fig"}B) of the preparation showed that the sample is in agreement with literature. The chlorophyll is associated with the PsaA and PsaB polypeptides that run as a diffuse band at greater than 65 kDa and the set of LHCI polypeptides shown with molecular weight in the range of 20--26 kDa. The minor band at approximately 30 kDa is due to the LHCPII antennae from PS-II. ::: {#pbio-0030230-g001 .fig} Figure 1 ::: {.caption} ###### Isolation of PS-I by Detergent Solubilization and Density-Gradient Centrifugation \(A) The final centrifugation step of the isolation protocol, the linear 0.1--1M sucrose gradients contain 0.02% Triton X-100. The detergent-solubilized (at 0.8% v/v Triton X-100) PS-I particles selectively aggregate upon entering the low Triton X-100 concentration of the gradient and sediment quickly relative to the other thylakoid proteins that remain at the top of the gradient. \(B) SDS-PAGE 18% slab gel lane containing the PS-I harvested from the 1M/2M interface as seen in (A). The region containing the LHCI is indicated. ::: ![](pbio.0030230.g001) ::: PS-I Stability in Various Conditions {#s2b} ------------------------------------ For PS-I complex dried in the absence of peptide detergent, a significant change in the fluorescence spectrum was observed. This change consisted of a decrease in the overall intensity of the far-red fluorescing chlorophyll in the approximately 735-nm region and an associated blueshift of this peak. In addition, the fluorescence peak at approximately 680 nm demonstrated an overall increase in intensity after the drying process. In some cases, a blueshift of this peak was also observed. The emission spectra of dried and solution state samples in the present chemical detergents are shown in [Figure 2](#pbio-0030230-g002){ref-type="fig"}. ::: {#pbio-0030230-g002 .fig} Figure 2 ::: {.caption} ###### The −196.15 °C Steady-State Emission Spectra of PS-I \(A) The solution spectrum of PS-I shows characteristic far-red fluorescence at the approximately 735-nm region, while dried PS-I shows a peak at approximately 683 nm. \(B) Emission spectra of samples containing DM retained partially their far-red fluorescence, but still showed a significant increase in fluorescence at approximately 683 nm. Additionally, the far-red fluorescence demonstrated a significant blueshift. \(C) Comparable changes are evident in the spectra of samples containing OG as seen in DM (B). ::: ![](pbio.0030230.g002) ::: We used four widely known membrane protein-stabilizing detergents, Triton X-100, N, N-dimethyldodecylamine N-oxide, N-dodecyl-β-D-maltoside (DM), and N-octyl-β-D-glucoside (OG) for the PS-I complex. Triton X-100 and N, N-dimethyldodecylamine N-oxide did not demonstrate a stabilizing effect based on the emission spectrum of PS-I, but in most cases caused increased degradation of the complex (data not shown). On the other hand, OG and DM showed modest stabilizing effect at low concentrations. The spectra for OG and DM are shown in [Figure 2](#pbio-0030230-g002){ref-type="fig"}. At low concentration, both OG and DM resulted in a blueshift from the approximately 735-nm region, but at higher concentration, both reduced the fluorescent emission at approximately the 735-nm region, with an increasing emission at 680 nm, implying they reduced PS-I stability. Thus, these results suggest that although OG and DM are good detergents for solubilization and stabilization of some membrane proteins, they are not ideal stabilizing detergents for PS-I under our conditions. Stabilizing the PS-I Complex Using Peptide Detergents {#s2c} ----------------------------------------------------- We carried out systematic experiments to determine if the peptide detergents A~6~K (molecular weight = 615 Da, CAC = 1.6 mM in water, approximately 0.23 mM in phosphate-buffered saline \[PBS\]) and acetyl-VVVVVVD (V~6~D) (molecular weight = 783 Da, CAC = 0.7 mM in water, approximately 0.1 mM in PBS) could stabilize PS-I complex during the drying process. Samples were prepared by addition of A~6~K or V~6~D at concentrations of 0.016%, 0.031%, 0.062%, 0.125%, 0.25%, and 0.5% (A~6~K: 0.5% = 8 mM, or 34X CAC in PBS; V~6~D: 0.5% = 6.4 mM, or 64X CAC in PBS). The results are shown in [Figure 3](#pbio-0030230-g003){ref-type="fig"}. Under these conditions at lower concentrations of A~6~K or V~6~D, PS-I not only showed blueshift but also showed a reduced fluorescent emission after drying. On the other hand, at higher concentration, the PS-I complex appeared to be stabilized during and after drying. In particular, samples containing 0.5% A~6~K demonstrated essentially negligible change in their emission spectra in the approximately 735-nm region. V~6~D was less effective at stabilizing PS-I as shown in [Figure 3](#pbio-0030230-g003){ref-type="fig"}B, even at the highest concentration of 0.5%. ::: {#pbio-0030230-g003 .fig} Figure 3 ::: {.caption} ###### Emission Spectra of the Dried PS-I Complex Containing Various Concentrations of the Peptide Detergents A~6~K and V~6~D \(A) A~6~K showed significantly enhanced ability to stabilize PS-I at a concentration of 0.5% (w/v) and showed reduced ability to stabilize PS-I complex at decreasing concentration. \(B) V~6~D showed lesser ability to stabilize PS-I complex at the same concentration. ::: ![](pbio.0030230.g003) ::: The Peptide Detergent A~6~K Stabilizes the PS-I Complex for an Extended Time {#s2d} ---------------------------------------------------------------------------- We asked how long the peptide detergent A~6~K could stabilize the PS-I complex on a dry surface. The stability of dried PS-I with 0.5% A~6~K (34X CAC in PBS) on a glass slide was measured at days 1, 7, 14, and 21. The results are shown in [Figure 4](#pbio-0030230-g004){ref-type="fig"}. A noticeable blueshift in the approximately 735-nm region coupled with a decrease in the intensity is apparent. At the same time, a slight increase of intensity in the approximately 680-nm region is visible. Despite this shift, the overall change in shape in the spectrum is negligible in contrast to other drastic changes observed with other detergents. Our observations suggest that the A~6~K peptide detergent is able to stabilize the PS-I complex under the conditions used. ::: {#pbio-0030230-g004 .fig} Figure 4 ::: {.caption} ###### Extended Time of PS-I Stability on Dried Surface Fluorescence spectrum of a PS-I film containing 0.5% A~6~K was monitored for 3 wk. Negligible increase in fluorescence was observed at approximately 683 nm. A blueshift from approximately 735 nm to approximately 727 nm fluorescence was observed. Thus, it suggests that the PS-I complex retained most of its structural integrity under conditions examined. Such shifts have been noted under other conditions. ::: ![](pbio.0030230.g004) ::: This observation is consistent with a separate experiment of another membrane protein, rhodopsin, using other peptide detergents. In these experiments, rhodopsin was stabilized for an extended time when high concentrations of peptide detergents were used (X. Zhao et al., unpublished data). Discussion {#s3} ========== PS-I Spectroscopy Features {#s3a} -------------------------- Study of the low-temperature chlorophyll fluorescence in the approximately 735-nm region was originated over 25 y ago \[[@pbio-0030230-b30]\]. It was initially presumed to be due to a chlorophyll species C-705 that was only observable at low temperatures because of its ability to undergo rapid energy transfer to P700 at elevated temperatures. However, a careful analysis of PS-I, with its full antennae complement, has demonstrated the presence of at least three far-red chlorophyll spectral forms with emission maxima of 720 nm, 730 nm, and 742 nm \[[@pbio-0030230-b31]\]. Recently, the longest wavelength fluorescence band has been observed even at room temperature using selective excitation, and has been shown to be associated with the lowest-energy spectral form of LHCI antenna complex \[[@pbio-0030230-b32]\]. The ability to monitor the reversible dissociation of pigment--protein complexes and the consequent cessation of intercomplex energy transfer by the relative intensities of −196.15 °C fluorescence at 685 nm and 735 nm has long been realized \[[@pbio-0030230-b33]\]. Monitoring of the F685/F735 chlorophyll fluorescence ratio has been used to study many in vivo and in vitro biological processes, including (1) the effect of leaf temperature \[[@pbio-0030230-b34]\], (2) seasonally induced chlorophyll breakdown in trees \[[@pbio-0030230-b35]\], (3) herbicide phytotoxicity in algae \[[@pbio-0030230-b36]\], (4) high-light-induced photoinhibition \[[@pbio-0030230-b37]\], (5) iron deficiency in algae \[[@pbio-0030230-b38]\], and (6) PS-I stability and disassembly \[[@pbio-0030230-b39]\]. Studies have also been carried out on the −196.15 °C emission spectrum of the PS-I core. Based on these studies, in which there is significant fluorescence at approximately 715 nm, it is presumable that damage could be done to the PS-I during drying as well. It is interesting to note that similar results have been reported from adverse interactions with detergent \[[@pbio-0030230-b31]\]. In these studies, it is argued that the fluorescence change observed is the result of uncoupled chlorophyll molecules. Analyses of PS-I Structure and Function {#s3b} --------------------------------------- Much progress has been made in the study of PS-I, mostly through the combination of membrane fractionation studies, various types of quantitative spectroscopy, protein crystallography and structural analyses. The chlorophyll pigment proteins responsible for these various spectrally distinct species are finally being resolved \[[@pbio-0030230-b06],[@pbio-0030230-b07]\]. The −196.15 °C steady state emission spectrum of plant PS-I has been attributed to the fluorescence of the LHCI proteins \[[@pbio-0030230-b40]\] associated with the core complex, which themselves demonstrate an extremely redshifted chlorophyll fluorescence \[[@pbio-0030230-b41]--[@pbio-0030230-b43]\]. The spectral changes presented here could therefore be the result of changes in the LHCI peptides, or in their association with the PS-I core complex. Several experiments on LHCI peptides have given similar results to those presented here. For example, it was shown that heterodimerization was an essential feature of LHC-730 \[[@pbio-0030230-b44]\]. The fluorescence at 730 nm occurred only for the dimeric state. Furthermore, site-directed mutagenesis studies of lhca1--4 genes showed that the approximately 730-nm emission could be abolished with a commensurate increase in the fluorescence at approximately 680 nm \[[@pbio-0030230-b45]\] by altering the specific interaction between protein and chlorophyll. It has been suggested that the LHC-680 peptides show an emission maximum at approximately 680 nm only in the monomeric form \[[@pbio-0030230-b46],[@pbio-0030230-b47]\]. These studies further indicate that such spectral changes are the result of a disruption of how the LHCI peptides and their associated chlorophyll are "connected" to the PS-I core complex. The current understanding for chlorophyll distribution in native PS-I including the antennae LHCI is that the core complex has approximately 110 chlorophyll molecules associated with PsaA/B. The peripheral antennae contain approximately 84 chlorophyll molecules composed of four distinct subunits, LhcA1--4, that exist in the form of heterodimers. These heterodimers are isolated as two fractions: LHCI-A (composed of LhcA1 and LhcA4) and LHCI-B (composed of LhcA2 and LhcA3). From analysis of chlorophyll content of isolated complexes and reconstituted complexes, each LHCI monomer subunit appears to bind approximately 10 chlorophyll molecules. This would then suggest that the native complex with approximately 195 chlorophyll molecules must have four heterodimers associated with each core complex. Undoubtedly, the precise number of antenna LHCI subunits will vary with plant species, plant age, and environmental conditions. In fact, analysis of the recently solved crystal structure of pea PS-I, which contains only 167 chlorophyll molecules, reveals that there are only two heterodimers. These four LHCI subunits that are asymmetrically positioned on one side of the PS-I core complex form an "LHCI belt" ([Figure 5](#pbio-0030230-g005){ref-type="fig"}). ::: {#pbio-0030230-g005 .fig} Figure 5 ::: {.caption} ###### Crystal Structure of PS-I as Determined by Ben-Shem et al. \[[@pbio-0030230-b07],[@pbio-0030230-b53]\], Visualized in VMD \[[@pbio-0030230-b52]\], and Rendered with POV-Ray 3.6 The color scheme is adapted from Ben-Shem et al. \[[@pbio-0030230-b07],[@pbio-0030230-b53]\] and is interpreted as follows. Portions shown in gray are common subunits found in both PS-I reaction centers in cyanobacteria and higher plants, while the red subunits are unique to higher plants. The two LHCI heterodimers are shown in green. The two different orientations of selected PS-I complex are shown top-down (A) and in a side view (B). In (C) and (D) the chlorophyll molecules are colored by a putative function: the P700 core antennae are tan, the special pair (P700) and voyeur chlorophyll molecules are cyan, the linker chlorophyll between the LHCI heterodimers is in yellow, the antennae chlorophyll molecules complexed within an LHCI subunit are in blue, and the "gap" chlorophyll molecules are shown in orange. (A), (C), and (D) are views looking down on the stromal hump highlighting protein, chlorophyll, and composite, respectively. (B) faces the LHC belt from inside the plane of the thylakoid membrane; only protein is shown in the upper portion and only chlorophyll below. ::: ![](pbio.0030230.g005) ::: The large Mg^2+^--Mg^2+^ distances between the core complex chlorophyll molecules and the primary LHCI chlorophyll molecules would suggest that energy transfer between the LHCI belt and core complex is directed through specialized "gap" chlorophyll molecules at the junction of LhcA1 and subunit G ([Figure 5](#pbio-0030230-g005){ref-type="fig"}A). It is believed that this region in the native protein may coordinate several LHCI subunits. Apparently, the optimization of energy transfer between LHCI complexes and the core complex has been facilitated through the evolution of a select number of strategically placed chlorophyll-binding sites that permit coordination of "gap" chlorophyll molecules \[[@pbio-0030230-b07]\]. These "gap" chlorophyll molecules in orange are shown in [Figure 5](#pbio-0030230-g005){ref-type="fig"}C. Interestingly, these "gap" chlorophyll molecules are positioned in a relatively open part of the structure, which may be highly vulnerable to the conformational changes of the complex following either dehydration or excessive detergent solubilization. The stabilizing properties of the detergent peptides may stem from their appropriate dimensions, allowing insertion into the complex so that they favorably stabilize these "gap" chlorophyll molecules, thus maintaining their low-temperature fluorescence properties. In this work, the PS-I complex that was dried with 0.5% A~6~K demonstrated minimal changes in its emission spectra, suggesting that there has been minimal change in the chlorophyll associated with PS-I. The most noticeable difference was a slight blueshift in the emission from approximately 735 nm to approximately 727 nm, associated with the far-red fluorescing chlorophyll, and a negligible increase in the approximately 680-nm region fluorescence. This change has also been reported before, and it is considered the result of a slight decoupling of the LHCI peptides from the PS-I complex. We therefore propose that the stabilizing effect of A~6~K and V~6~D on PS-I can be determined by the ratio of the fluorescence at approximately 730 nm to the fluorescence at approximately 680 nm. Designed Peptide Detergents for the Study of Membrane Proteins {#s3c} -------------------------------------------------------------- PS-I is one of the first examples that clearly demonstrate how designed peptide detergents can stabilize membrane proteins. In collaboration with others, we have also stabilized bovine rhodopsin with the peptide detergents A~6~D and V~6~D for an extended time at elevated temperatures (X. Zhao et al., unpublished data). In collaboration with another investigator, using the peptide detergent V~6~D, a crystal of the integral membrane protein glycerol 3-phosphate dehydrogenase with its cofactor was obtained and was diffracted to approximately 7 Å using an in-house X-ray diffraction instrument (J. Yeh et al., unpublished data). It is plausible that peptide detergents, similar to other detergents, may directly interact with the hydrophobic domains of membrane proteins. It is likely that numerous peptide detergent molecules, approximately 2 nm in size, like lipids, can effectively surround the hydrophobic trans-membrane domains of membrane proteins, thus sequestering them from directly interacting with water molecules and preventing them from undergoing self-aggregation. In the PS-I complex, the peptide detergents not only sequestered the protein and filled the "gaps" but also maintained its overall structural integrity, thus retaining its biological function \[[@pbio-0030230-b08]\]. The applications of designed peptide detergents are still in an early stage of development. The peptide detergents presented here represent a new class of bioengineered material because they can be designed at the molecular level, using a combinatorial approach characteristic of peptide chemistry. However, we have yet to introduce the variability inherent in amino acids. For example, so far, we have concentrated on peptides containing one kind of hydrophobic tail. It is likely that as we explore combinations of short and long tails, as well as different head groups, the molecules will become more specialized, and we may find specific sequences that work best for stabilizing specific membrane proteins. Equally useful is the exploration of mixing several different detergents together as a cocktail. In essence, this could let the protein select the most appropriate mixture for stabilization and crystallization. Materials and Methods {#s4} ===================== PS-I isolation and electrophoresis {#s4a} ---------------------------------- PS-I was obtained from spinach leaves by homogenization in 400 mM sorbitol and 50 mM tricine-KOH (pH 7.8). The resulting homogenate was then passed through Miracloth (approximately 30 μm; CalBiochem, San Diego, California, United States), supported by a double layer of cheesecloth, and centrifuged at 1,000 × g for 5 min. The pellet, consisting mostly of chloroplasts and broken plastids, was resuspended and forced through five passes in a Dounce homogenizer in 50 mM sorbitol, 5 mM EDTA-NaOH, and 50 mM tricine-KOH (pH 7.8), which lysed any unbroken plastids. It was centrifuged again at 10,000 × g for 5 min; this step was repeated once. The resulting pellet was highly enriched in thylakoids, which were completely nonappressed after treatment with the chelator; this pellet was resuspended in a small volume of water, and the chlorophyll content was determined. The thylakoid suspension was then brought to a final concentration of 0.8 mg/ml chlorophyll and 0.8% (w/v) Anatrace (<http://www.anatrace.com/>) Anapoe Triton X-100. Gentle stirring began immediately at room temperature for 30 min. Following the incubation, the solution was spun at 42,000 × g for 30 min, and the pellet was discarded. Approximately 8 ml of the supernatant was loaded onto approximately 25 ml linear 0.1--1 M sucrose gradients containing 0.02% Triton X-100 with 2 ml of 2M sucrose (no Triton) as a cushion and spun at 100,000 × g in a Beckman SW-28 rotor overnight (15 h). The dark green band at the 1M--2M interface was harvested with a long needle and passed through a Dounce homogenizer, pooled, and frozen at −80 °C. The protocol was adapted from Mullet et al. \[[@pbio-0030230-b48]\]. Peptide synthesis and preparation {#s4b} --------------------------------- Peptides were obtained from the MIT Biopolymers Laboratory and Synpep Corporation ([www.synpep.com](www.synpep.com)). The peptide detergent stock solutions were made by dissolving these peptides at a concentration of 1% (w/v) in 50 mM Tris (pH 7.8); 25 mM NaCl; 0.02% Triton X-100, and then sonicating in a water bath sonicator for 20 min. Stock solutions of peptides were stored at room temperature for the duration of the experiment. PS-I treatment with detergents and peptides {#s4c} ------------------------------------------- This was carried out by diluting 25 μl of PS-I with 25 μl of serial dilutions of the peptide stock solutions in buffer. This resulted in a chlorophyll concentration of 50 μg/ml in all samples and peptide concentrations of 0.5%, 0.25%, 0.125%, 0.063%, 0.032%, and 0.016%. For most experiments, the peptide detergent was used at final 0.5% concentration. PS-I dehydration {#s4d} ---------------- All samples were flash frozen in liquid nitrogen until use. These samples were then thawed and dried on 1-cm^2^ glass slides for at least 15 h under N~2~. Low-temperature fluorescence measurements {#s4e} ----------------------------------------- Their steady-state emission spectra at −196.15 °C \[[@pbio-0030230-b31],[@pbio-0030230-b49]--[@pbio-0030230-b51]\] were then determined to investigate the stability of the PS-I in each sample. Each slide was cooled under liquid nitrogen in a cryostat, and fluorescence was excited optically using a 408-nm laser. Steady-state emission spectra were recorded using a CCD spectrometer with an optical fiber input oriented 90° from the laser and a slit width of 20 μm. The fluorescence intensity of each sample was normalized to the integral over the wavelength shown. Modeling of pea PS-I {#s4f} -------------------- The structure of pea PS-I was determined at 4.4 Å \[[@pbio-0030230-b07]\]. We highlighted in [Figure 5](#pbio-0030230-g005){ref-type="fig"} the components of this structure that may add to our understanding of the fluorescence phenomena reported in this work. Manipulations of the molecular coordinates were made with VMD 1.8.2 \[[@pbio-0030230-b52]\], and the images were rendered in POV-Ray 3.6 (<http://www.povray.org> /download/). Supporting Information {#s5} ====================== Accession Number {#s5a} ---------------- The Protein Data Bank (<http://www.rcsb.org/pdb/>) accession number for the crystal structure of pea PS-I is 1QZV. We thank Steve Yang, Alexander Rich, and members of Zhang\'s laboratory for helpful and stimulating discussions. This work is supported in part by grants from the Defense Advanced Research Projects Agency--Air Force Office of Scientific Research (AFOSR), the Multidisciplinary University Research Initiative--AFOSR, and the National Science Foundation--Center for Bits and Atoms at the Massachusetts Institute of Technology. Competing interests. The authors have declared that no competing interests exist. Author contributions. SZ conceived and designed the experiments. PK and MV performed the experiments. PK, XZ, MV, MAB, BDB, and SZ analyzed the data. XZ, MAB, and BDB contributed reagents/materials/analysis tools. PK, MV, BDB, and SZ wrote the paper. Citation: Kiley P, Zhao X, Vaughn M, Baldo MA, Bruce BD, et al. (2005) Self-assembling peptide detergents stabilize isolated photosystem I on a dry surface for an extended time. PLoS Biol 3(7): e230. A6K : acetyl- AAAAAAK CAC : critical aggregation concentration DM : N-dodecyl-β-D-maltoside OG : N-octyl-β-D-glucoside PBS : phosphate-buffered saline PS-I : photosystem I V6D : acetyl-VVVVVVD	PubMed Central	2024-06-05T03:55:57.574449	2005-6-21	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1151599/", "journal": "PLoS Biol. 2005 Jul 21; 3(7):e230", "authors": [ { "first": "Patrick", "last": "Kiley" }, { "first": "Xiaojun", "last": "Zhao" }, { "first": "Michael", "last": "Vaughn" }, { "first": "Marc A", "last": "Baldo" }, { "first": "Barry D", "last": "Bruce" }, { "first": "Shuguang", "last": "Zhang" } ] }
PMC1151600	Introduction {#s1} ============ In the classical clathrin-mediated endocytic pathway, the step from early to late endosomes is crucial for selective transport of cargo and membrane components to lysosomes for degradation. This step involves cargo sorting and segregation, and the transfer occurs either by carrier vesicles that detach from early endosomes, or by the maturation of early endosomes to late endosomes \[[@pbio-0030233-b01]\]. Whereas vesicle traffic from the plasma membrane to early endosomes and homotypic fusion of early endosomes is regulated by a small GTPase, Rab5 \[[@pbio-0030233-b06],[@pbio-0030233-b07]\], and recycling to the plasma membrane depends on Rab4 and Rab11 \[[@pbio-0030233-b08],[@pbio-0030233-b09]\], the factors that mediate cargo transfer from early to late endosomes remain largely unidentified. In vitro experiments with biochemically purified endosomes have suggested that this step is mediated by a GTPase, Arf1. Together with COP I, it is thought to drive the formation of endosomal carrier vesicles (ECVs) \[[@pbio-0030233-b10]\]. Low pH inside the endosomes was found to be essential for this process. There is also evidence that Rab7 is critically involved in early to late endosome traffic. Expression of dominant negative Rab7 mutants blocks the exit of certain cargo molecules from early endosomes to late endosomes \[[@pbio-0030233-b14]\]. However, since Rab7 has been localized to late endosomes and lysosomes, it has been proposed that it is not involved in the formation of carrier vesicles at the early endosome but in downstream events \[[@pbio-0030233-b17]\]. That Rab7 is also involved in a later step---in trafficking from the late endosome to the lysosome---is supported by observations in many systems \[[@pbio-0030233-b18]\]. In this study, we have analyzed the sorting of cargo from early to late endosomes using triple-colored, video-enhanced fluorescence microscopy in live cells expressing various Rab- and Arf-GTPases labeled with green, red, yellow, or cyan fluorescent protein (GFP, RFP, YFP or CFP). As endocytic cargo, we used fluorescently labeled Semliki forest virus (SFV)---a simple, enveloped RNA virus known to enter cells via the classical clathrin-mediated pathway and to be degraded in lysosomes \[[@pbio-0030233-b22]\]. The main advantage of SFV over most physiological ligands is that fluorescently labeled, individual particles can be visualized and tracked during entry as single identifiable fluorescent spots. Our results show that the transport of SFV occurs in Rab7-positive vesicles formed from Rab7-positive domains in early endosomes. Although present in early endosomes, Arf1 did not appear to be involved. Results {#s2} ======= Characterization of Labeled SFV {#s2a} ------------------------------- To visualize the entry of SFV in live cells by fluorescence microscopy, we labeled purified virus with different fluorophores: Cy5 (SFV-Cy5), AlexaFluor594 (SFV-AF594), and FITC (SFV-FITC). SDS-PAGE and fluorography of the labeled viruses showed that the fluorophores were coupled to glycoproteins E1 and E2 (shown for SFV-FITC in [Figure 1](#pbio-0030233-g001){ref-type="fig"}A). Together with the small glycopeptide E3, E1 and E2 form interconnected spike complexes in the envelope \[[@pbio-0030233-b25]\]. As expected from its location on the lumenal side of the viral envelope, the capsid protein C was not labeled. Absorbance spectra indicated that virions carried an average of 400 molecules of dye. The labeling process reduced infectivity by about half, from 7 × 10^7^ to 3 × 10^7^ plaque forming units (PFU) per microgram of viral protein. ::: {#pbio-0030233-g001 .fig} Figure 1 ::: {.caption} ###### Fluorophores Label Glycoproteins E1 and E2 of SFV \(A) Analysis of SFV-FITC by non-reducing SDS-PAGE with Coomassie blue staining in the left lane and fluorography in the right lane. E1 (49 kDa) and E2 (52 kDa) are the only proteins that are fluorescently labeled; capsid protein C and E3 are not. \(B) Confocal microscopy of Cy5-labeled SFV, showing individual spots of uniform size (approximately 0.4 μm). The relative fluorescence intensity distribution of two spots is shown. The distribution of the fluorescence fits to a Gaussian function. (Fit is calculated using the equation: y = e ^−x^2^.)\ Scale\ bar\ represents\ 1\ μm.M^ ::: ![](pbio.0030233.g001) ::: When viewed by fluorescence microscopy, the fluorescent virus particles were visible as spots with a diameter of 0.4--0.5 μm (shown for SFV-Cy5 in [Figure 1](#pbio-0030233-g001){ref-type="fig"}B). The distribution of the signal could be fit to a single Gaussian function (blue line in [Figure 1](#pbio-0030233-g001){ref-type="fig"}B) with R ^2^ = 0.96, suggesting that the light was emitted from single particles. Structural studies have shown that the actual diameter of a SFV particle is 70 nm \[[@pbio-0030233-b25]\]. When the labeled SFV were allowed to bind to the surface of Vero cells, the staining of the fluorophore-labeled viruses overlapped completely with the pattern observed by immunofluorescence using a polyclonal antibody against E1 and E2 ([Figure S1](#sg001){ref-type="supplementary-material"}). Internalization, Penetration, and Degradation {#s2b} --------------------------------------------- SFV particles were bound to the cell surface at 4 °C for 1 h at a multiplicity of infection (MOI) of 20, the unbound virus was washed away, and the cells were shifted to 37 °C. Under these conditions, the viruses are rapidly internalized by clathrin-coated vesicles and delivered to endosomes and subsequently to lysosomes, where proteolytic degradation of envelope proteins and unfused viruses occurs \[[@pbio-0030233-b22]\]. The progressive movement could be followed using the fluorescently labeled virus in Vero cells ([Figure 2A--2C](#pbio-0030233-g002){ref-type="fig"}). As previously reported \[[@pbio-0030233-b26]\], the viruses were first observed as individual spots on the cell surface. Confocal microscopy of fixed and immunostained cells 10 min after warming showed the majority of internalized viruses in compartments positive for early endosome antigen 1 (EEA1), a Rab5 effector and an early endosome marker \[[@pbio-0030233-b27],[@pbio-0030233-b28]\] ([Figure 2](#pbio-0030233-g002){ref-type="fig"}A). At this time, many virus particles were still on the cell surface ([Figure S2](#sg002){ref-type="supplementary-material"}). After 20 min of warming, many viruses were already present in compartments positive for Rab7 ([Figure 2](#pbio-0030233-g002){ref-type="fig"}B), a late endosomal marker \[[@pbio-0030233-b29]\]. Only after 30 min or longer could overlap be observed with lysosome associated membrane protein 1 (LAMP-1), a marker for lysosomes \[[@pbio-0030233-b30],[@pbio-0030233-b31]\] ([Figure 2](#pbio-0030233-g002){ref-type="fig"}C). ::: {#pbio-0030233-g002 .fig} Figure 2 ::: {.caption} ###### Internalization, Penetration, and Degradation SFV-Cy5 (at a MOI of 20) was allowed to bind to Vero cells for 60 min at 4 °C, followed by transfer to 37 °C for different time periods. \(A) The cells were fixed and permeabilized 10 min after warming, followed by immunostaining using antibodies to EEA1. SFV-Cy5 is present in EEA1-positive endosomes. \(B) The cells were fixed 20 min after warming and immunostained with anti-Rab7 antibodies. SFV-Cy5 is present in Rab7-positive endosomes. \(C) Non-labeled SFV was used, and the cells were fixed and permeabilized 30 min after warming and immunostained with anti-LAMP-1 and anti-E1/E2 antibodies. SFV is present in endosomes positive for LAMP-1. \(D) A FACS-based infection assay was used to determine the time course of SFV penetration during entry. Vero cells were infected with a MOI of one. NH~4~Cl was added at different time points to inhibit infection. Cells were further incubated at 37 °C for 5 h, then fixed and immunostained for newly synthesized glycoproteins (n = 3). \(E) Analysis of E1/E2 degradation was determined using immunoblotting. Virus (MOI of 50) was bound to Vero cells in the cold, and unbound virus was washed away. Cells were incubated for indicated times, and surface-associated viruses were removed by Proteinase K treatment. Cells were lysed, and after SDS-PAGE, immunoblotting was performed with an antibody against E1/E2. Note that contrary to non-reduced samples ([Figure 1](#pbio-0030233-g001){ref-type="fig"}A), E1 and E2 co-migrate in SDS-PAGE after reduction. Scale bars represent 5 μm. ::: ![](pbio.0030233.g002) ::: To determine the timing of the acid-activated penetration event leading to infection, SFV was allowed to bind to the cells in the cold at a MOI of one. At different times after warming, 20 mM NH~4~Cl was added. Like other lysosomotropic weak bases, NH~4~Cl raises the pH in acidic organelles almost instantaneously \[[@pbio-0030233-b32]\], and prevents further acid activation of incoming viruses. After 5 h, the fraction of infected cells was determined using an indirect FACS-based assay, in which newly synthesized viral proteins were detected with an anti-E1/E2 antibody. In agreement with results from other cell types \[[@pbio-0030233-b33]\], the acid-induced fusion events started between 2 and 3 min after warming and reached a half maximal level at 6 min ([Figure 2](#pbio-0030233-g002){ref-type="fig"}D). Following a lag phase, SFV endocytosis is known to result in efficient degradation of E1 and E2 proteins in lysosomes \[[@pbio-0030233-b34]\]. When degradation was analyzed in Vero cells by immunoblotting using anti-E1/E2 antibodies, it was found to start 30 min after warming ([Figure 2](#pbio-0030233-g002){ref-type="fig"}E). In this experiment, Proteinase K--mediated removal of surface-bound viruses showed that about half of the cell-associated virus particles (52%) were endocytosed. For the incoming virus, the course of events in Vero cells thus followed a program that involved (1) rapid internalization, (2) exposure to low pH in early endosomes (2--15 min), and (3) transfer to late endosomes and lysosomes (starting after 20 min). Through all these different compartments, the size and intensity of fluorescent spots representing individual virus particles remained roughly unaltered (data not shown). This meant that even when the viruses fused and E1 and E2 became part of the endosomal membrane, the glycoproteins did not diffuse away from each other. This behavior of SFV membranes has previously been observed after fusion with the plasma membrane \[[@pbio-0030233-b35]\]. Localization of Endocytic Markers {#s2c} --------------------------------- To study the distribution of markers in endosomal compartments, we performed double immunofluorescence experiments with untransfected cells. We used antibodies to EEA1 (an effector of Rab5) \[[@pbio-0030233-b27],[@pbio-0030233-b28]\], and antibodies to Rab7 \[[@pbio-0030233-b36]\]. When the distribution was compared, we observed, as have others, that these early and late endosomal markers were largely separated ([Figure 3](#pbio-0030233-g003){ref-type="fig"}A, arrowheads). However, closer inspection and quantification showed that 13.6% of the EEA1-positive organelles were also stained with anti-Rab7 antibodies (arrows). In CV-1 and HeLa cells the corresponding numbers were 13.1% and 12.8%, respectively ([Figures S3](#sg003){ref-type="supplementary-material"} and [S4](#sg004){ref-type="supplementary-material"}). ::: {#pbio-0030233-g003 .fig} Figure 3 ::: {.caption} ###### Endosomes Labeled with Two Markers Show Presence of Rab7 in Early Endosomes Vero cells were fixed and viewed by confocal microscopy. \(A) Confocal microscopy of immunolabeled cells using anti-EEA1 (green) and anti-Rab7 (red) antibodies, (B--F) The cells were transfected with fluorescent-protein-labeled constructs as follows: (B) GFP-Rab5 and RFP-Rab7, (C) GFP-Rab4 and RFP-Rab5, (D) Arf1-GFP and RFP-Rab5, (E) GFP-Rab4 and RFP-Rab7, and (F) Arf1-GFP and RFP-Rab7. Arrowheads show individual endosomes positive for one of the two markers, and arrows indicate endosomes positive for two markers. Scale bars represent 10 μm. ::: ![](pbio.0030233.g003) ::: To determine whether the overlap was coincidental, we quantified the extent of overlap between EEA1 and COP II, which do not associate with common organelles ([Figure S5](#sg005){ref-type="supplementary-material"}). The apparent overlap was 4.6% ([Figure S5](#sg005){ref-type="supplementary-material"}B and [S5](#sg005){ref-type="supplementary-material"}C). The measured overlap between EEA1 and caveolin-1, which is known to form a domain in some of the early endosomes \[[@pbio-0030233-b37]\], was 10.5% ([Figure S5](#sg005){ref-type="supplementary-material"}A). We concluded that the co-localization of EEA1 and Rab7 was real, and that about one in ten EEA1-positive endosomes contained detectable amounts of Rab7. The double-labeled endosomes often displayed distinct red and green regions, suggesting location of EEA1 and Rab7 in separate domains of the mosaic structure of the endosomal membrane ([Figures 3](#pbio-0030233-g003){ref-type="fig"}A, S3, and S4, arrows) \[[@pbio-0030233-b38],[@pbio-0030233-b39]\]. To analyze the endosomes in live cells, we expressed a variety of fluorescent-protein-labeled small GTPases known to occur in distinct endosomal compartments: Rab5 for early endosomes, Rab4 for early and recycling endosomes, and Rab7 for late endosomes \[[@pbio-0030233-b06],[@pbio-0030233-b08],[@pbio-0030233-b29]\]. We also expressed Arf1, which has been reported to be present on early endosomes and ECVs \[[@pbio-0030233-b12]\]. First, the localization of GFP-Rab5 and RFP-Rab7 was determined in fixed, transfected cells by confocal microscopy in the absence of virus ([Figure 3](#pbio-0030233-g003){ref-type="fig"}B). As expected, most of the Rab5- and Rab7-positive structures were localized in distinct organelles, at least when viewed in the peripheral regions of the cytoplasm where individual organelles were easily distinguished (arrowheads). However, again in every cell a population of GFP-Rab5-positive endosomes (14.6%) could be observed also positive for RFP-Rab7 (arrows). That these were indeed hybrid organelles was confirmed in live video recordings of unfixed cells by the coordinated movement of both fluorescent colors ([Videos S1--S3](#sv001){ref-type="supplementary-material"}). The distinct red and green regions (domains) in the endosomes were persistent features also in live cells. When GFP-Rab4 and RFP-Rab5 were co-expressed, 37.7% of the Rab5-positive endosomes contained Rab4, and when Arf1-GFP and RFP-Rab5 were expressed, 35.5% of the Rab5 endosomes contained the labeled Arf1 ([Figure 3](#pbio-0030233-g003){ref-type="fig"}C and [3](#pbio-0030233-g003){ref-type="fig"}D), confirming that early endosomes are heterogeneous and contain different small GTPases. When RFP-Rab7 and GFP-Rab4 were co-expressed, some double-stained vesicles (12.3%) could also be observed ([Figure 3](#pbio-0030233-g003){ref-type="fig"}E, arrows). The same was observed for Arf1-GFP and RFP-Rab7 (12.5%) ([Figure 3](#pbio-0030233-g003){ref-type="fig"}F). These findings were consistent with the presence of Rab7 in some early endosomal structures. As expected, immunofluorescence showed that Rab7 was present in LAMP-1-positive structures. We found that of Rab7-positive organelles, 46% contained LAMP-1 (data not shown). SFV in Rab7-Positive Early Endosomes {#s2d} ------------------------------------ To analyze the endosomes that were reached by the virus, SFV-Cy5 was allowed to bind to Vero cells in the cold. The particle to cell ratio was kept low (20 PFU/cell) so that individual particles could be followed, and endosomes would not receive multiple SFV particles. The cells were then washed, shifted to 37 °C for different times up to 180 min, fixed, immunolabeled with anti-EEA1 and anti-Rab7, and analyzed by confocal microscopy. As shown in [Figure 4](#pbio-0030233-g004){ref-type="fig"}A (panel a), at 0 min the SFV particles (blue) did not co-localize with either of the endosomal markers (red and green). After 10 min ([Figure 4](#pbio-0030233-g004){ref-type="fig"}A, panel b), most of the virus particles co-localized with EEA1 in early endosomes. After 20 min ([Figure 4](#pbio-0030233-g004){ref-type="fig"}A, panel c), many of the virus particles were located in the endosomes positive for both EEA1 and Rab7, and from 30 min onwards ([Figure 4](#pbio-0030233-g004){ref-type="fig"}A, panel d), an increasing fraction was seen in organelles stained by Rab7 alone. ::: {#pbio-0030233-g004 .fig} Figure 4 ::: {.caption} ###### Progression of SFV through Endosomes Labeled for EEA1 and Rab7 \(A) Confocal microscopy of Vero cells. SFV-Cy5 was allowed to bind to Vero cells (at a MOI of 20) for 60 min at 4 °C, followed by transfer to 37 °C and fixed after different times (a, 0 min; b, 10 min; c, 20 min; d, 30 min) and immunolabeled with EEA1 and anti-Rab7 antibodies. Shown are three viruses for each time point. Scale bars represent 10 μm. In panel a, SFV do not co-localize with any endosomal marker; in panel b, SFV are localized in EEA1-positive endosomes; in panel c, SFV are localized in EEA1- and Rab7-positive endosomes; and in panel d, SFV are localized in Rab7-positive endosomes. \(B) Quantification of SFV-Cy5 in endosomes that contain EEA1 but no Rab7 (green line), endosomes that contain both (blue line), and structures that contain only Rab7 (red line) over time. The values are normalized such that 100% represents the number of cell-associated viruses at each time point corrected for the degradation. ::: ![](pbio.0030233.g004) ::: To quantify these observations, we viewed 260 virus particles in more than ten cells for each time point, and scored how many co-localized with different endosomal markers. The results ([Figure 4](#pbio-0030233-g004){ref-type="fig"}B) were normalized so that 100% represented the number of cell-associated viruses at each time point corrected for the degradation. The results showed a progression of the internalized viruses through different endosome subpopulations. Co-localization with EEA1 without Rab7 peaked at 10 min ([Figure 4](#pbio-0030233-g004){ref-type="fig"}B, green line). It dropped off rapidly and was replaced by a population positive for both EEA1 and Rab7 (blue line), with a maximum at 20 min. This was followed by organelles positive for Rab7 but negative for EEA1 (red line) starting at 30 min and persisting until 180 min. What was new and unexpected in these observations was that the viruses passed through an endosomal organelle population that contained both EEA1 and Rab7. This occurred in the time period 10--60 min after warming. As shown above, such EEA1- and Rab7-positive organelles were present whether SFV was added or not. We modeled SFV transit through the different endosomal compartments using coupled first-order reactions (Kinetikit/GENESIS) \[[@pbio-0030233-b40]\] assuming unidirectional transport: \[plasma membrane\] → \[early EEA1-positive endosome\] → \[hybrid EEA1- and Rab7-positive endosome\] → \[late Rab7-positive endosome\] → \[degradation\]. A set of parameters was found that allowed a fit with the experimental data in [Figure 4](#pbio-0030233-g004){ref-type="fig"}B ([Figure S6](#sg006){ref-type="supplementary-material"}). These parameters suggested that all the viruses that ended up in late endosomes passed through the hybrid compartment. Moreover, the model allowed us to estimate the average residence times of SFV in the different endosomal compartments. They were 10 min for the early endosomal compartments, 16 min for the hybrid endosomes, and 52 min for the late endosomal compartments. When GFP-Rab5- and RFP-Rab7-transfected cells were allowed to internalize SFV-Cy5, the progression of the virus particles could be observed by confocal microscopy in live Vero cells ([Figure 5](#pbio-0030233-g005){ref-type="fig"}A); the majority of viruses were first located in organelles that were positive for GFP-Rab5, later in organelles that were positive for both GFP-Rab5 and RFP-Rab7, and finally in RFP-Rab7- positive organelles. A similar progression of virus through endosomal compartments was observed in fixed and live CV-1 ([Figure S7](#sg007){ref-type="supplementary-material"}A and [S7](#sg007){ref-type="supplementary-material"}B) and HeLa cells ([Figure S7](#sg007){ref-type="supplementary-material"}C and [S7](#sg007){ref-type="supplementary-material"}D), although progress through the pathway was less synchronous in these cells. It is noteworthy that the expression of GFP-Rab5 and/or RFP-Rab7 did not affect SFV infection of Vero cells, consistent with reports that these chimeric proteins have no adverse effect on the pathway \[[@pbio-0030233-b37]\]. ::: {#pbio-0030233-g005 .fig} Figure 5 ::: {.caption} ###### Sorting of Rab7 and SFV from Early Endosomes (A--C) Selected images obtained from a time series starting 20 min after warming. Vero cells were transfected with (A) GFP-Rab5 and RFP-Rab7, (B) GFP-Rab4 and RFP-Rab5, or (C) GFP-Rab4 and RFP-Rab7; SFV-Cy5 was then added. SFV-Cy5 is sorted away from Rab5- and Rab4-positive endosomes in a Rab7-positive vesicle (arrowheads). Scale bars represent 2.5 μm. \(D) Confocal microscopy of a Vero cell transfected with the dominant negative GFP-Rab7T22N and RFP-Rab5, followed by incubation with SFV-Cy5. Virus was internalized for 1 h at 37 °C. SFV-Cy5 is present in large Rab5-positive endosomes (arrowheads). Scale bar represents 10 μm. (E and F) Confocal microscopy of Vero cells transfected with the dominant negative GFP-Rab7T22N (insets), followed by incubation with SFV-Cy5. Virus was internalized for 1 h at 37 °C. Cells were fixed, permeabilized, and immunostained for (E) EEA1 or (F) LAMP-1. In all, 86.3% of SFV-Cy5 (red) is present in large EEA1-positive endosomes (green in \[E\]) and not in LAMP-1-positive structures (green in \[F\]) (arrowheads). Note that SFV-Cy5 is present in LAMP-1-positive structures in untransfected cells (arrows). Scale bars represent 10 μm. ::: ![](pbio.0030233.g005) ::: A closer look at the distribution of GFP-Rab5, RFP-Rab7, and SFV-Cy5 in endosomes showed a typical "mosaic" distribution of the Rabs \[[@pbio-0030233-b38]\]. GFP-Rab5 and RFP-Rab7 were as a rule distributed in distinct domains of the organelle ([Figure 5](#pbio-0030233-g005){ref-type="fig"}A). That they were part of a common structure, and that the virus was trapped in it, was evident when endosomes in live cells were viewed over time; the three labels often moved together in the cytoplasm as parts of the same organelle ([Videos S4](#sv004){ref-type="supplementary-material"} and [S5](#sv005){ref-type="supplementary-material"}). An uneven distribution of fluorescent markers was also seen in the endosomes of HeLa and CV-1 cells whether viewed after immunofluorescence staining for EEA1 and Rab7 in fixed, untransfected cells ([Figure S7](#sg007){ref-type="supplementary-material"}A and [S7](#sg007){ref-type="supplementary-material"}C), or viewed in live cells expressing the fluorescent-protein-labeled Rabs ([Figure S7](#sg007){ref-type="supplementary-material"}B and [S7](#sg007){ref-type="supplementary-material"}D). Sorting of Rab7 and SFV from Early Endosomes {#s2e} -------------------------------------------- When the fate of SFV-Cy5-containing endosomes was followed by video microscopy, separation of the RFP-Rab7-positive domain and the virus from the GFP-Rab5-labeled domains could frequently be seen. The hybrid endosome was split in two: one part contained the Rab5, and the other the Rab7 and the virus particle. In the examples shown in [Figure 5](#pbio-0030233-g005){ref-type="fig"}A and [Videos S4](#sv004){ref-type="supplementary-material"} and [S5](#sv005){ref-type="supplementary-material"}, the SFV-Cy5- and RFP-Rab7-positive particle moved away while the Rab5-containing remnant remained stationary. In 70% of the cases recorded (n = 23) this was the case. In the other 30% of cases, it seemed that both parts of the endosome moved away after fission, in opposite directions. The viral cargo was thus transferred from an early endosome that contained both Rab5 and Rab7 to a detached organelle that contained Rab7 but no Rab5. In all analyzed time series, the virus always left the Rab5/Rab7 compartment together with Rab7, and Rab5 was left behind. We never observed the virus leaving an early endosome compartment without Rab7. It was therefore possible that formation of the Rab7-positive carrier vesicle was triggered by the presence of the virus. That the virus was not necessary, however, was shown by the observation that Rab7-positive vesicles could be seen leaving Rab5- and Rab7-positive endosomes in cells to which no virus was added (see [Video S3](#sv003){ref-type="supplementary-material"}). To confirm that the organelles that SFV were leaving were in fact early endosomes, we transfected either RFP-Rab5 or RFP-Rab7 together with another early endosomal marker, GFP-Rab4. We then allowed SFV-Cy5 to be internalized for 20 min. Rab4 is known to be present in early and recycling endosomes \[[@pbio-0030233-b08],[@pbio-0030233-b09]\]. In live recordings, the SFV glycoproteins (blue) could be observed to sort away from both Rab4 and Rab5 ([Figure 5](#pbio-0030233-g005){ref-type="fig"}B; [Video S6](#sv006){ref-type="supplementary-material"}). In cells expressing GFP-Rab4 and RFP-Rab7 the departure occurred together with Rab7, leaving Rab4 quantitatively behind ([Figure 5](#pbio-0030233-g005){ref-type="fig"}C; [Video S7](#sv007){ref-type="supplementary-material"}). Furthermore, we used fluorescently labeled transferrin. We loaded RFP-Rab5- or RFP-Rab7-expressing cells with transferrin-AlexaFluor488 and added SFV-Cy5 ([Figure S8](#sg008){ref-type="supplementary-material"}A and [S8](#sg008){ref-type="supplementary-material"}B; [Videos S8](#sv008){ref-type="supplementary-material"} and [S9](#sv009){ref-type="supplementary-material"}). SFV was present in the transferrin- and Rab5-positive early endosomes. When the sorting of SFV and Rab7 took place 20 min after internalization, transferrin failed to be included in the carrier vesicles. When a constitutively inactive GFP-Rab7T22N mutant and RFP-Rab5 were expressed together, we saw that the Rab5-positive early endosomes were bigger and more rounded than in cells without the Rab7 mutant ([Figure 5](#pbio-0030233-g005){ref-type="fig"}D). The same was true for GFP-Rab7T22N-expressing cells stained for EEA1 ([Figure 5](#pbio-0030233-g005){ref-type="fig"}E). Rab7 was not associated with any membrane organelles. When SFV-Cy5 was allowed to enter these cells for 1 h, we found that it was internalized but failed to leave the enlarged Rab5/EEA1-positive compartments. Consistent with capture in early endosomes, almost none of the SFV localized to LAMP-1 structures ([Figure 5](#pbio-0030233-g005){ref-type="fig"}F, arrowheads), whereas in control cells many did (arrows). Quantification showed that 86.3% of the viruses remained co-localized with EEA1 and only 3.9% co-localized with LAMP-1-containing organelles. This indicated that forward transport of SFV was inhibited by GFP-Rab7T22N at the level of early endosomes \[[@pbio-0030233-b14]\]. Active Rab7 was evidently needed for the transfer of SFV from this compartment. The inhibition in transport to late endosomes was likely the reason why the early endosomes were enlarged in these cells. Sorting Is Arf1-Independent {#s2f} --------------------------- Transport of cargo from early to late endosomes has been reported to occur in so-called ECVs regulated by Arf1 \[[@pbio-0030233-b12]\]. To test whether Arf1 mediates this sorting step for SFV, we transfected cells with Arf1-GFP and either RFP-Rab5 ([Figure 6](#pbio-0030233-g006){ref-type="fig"}A; [Video S10](#sv010){ref-type="supplementary-material"}) or RFP-Rab7 ([Figure 6](#pbio-0030233-g006){ref-type="fig"}B; [Video S11](#sv011){ref-type="supplementary-material"}). We then followed the fate of SFV-Cy5 in these cells over time in confocal video recordings. ::: {#pbio-0030233-g006 .fig} Figure 6 ::: {.caption} ###### Sorting of SFV into Late Endosomes is Arf1- and pH-Independent (A and B) Selected images, obtained from time series recorded approximately 20 min after warming. Vero cells were transfected with (A) Arf1-GFP and RFP-Rab5, or (B) Arf1-GFP and RFP-Rab7, then SFV-Cy5 was added. Arf1 is present on Rab5-positive endosomes. SFV-Cy5 is sorted away from Arf1- and Rab5-positive endosomes in a Rab7-positive vesicle (arrowheads). Scale bars represent 2.5 μm. (C and D) Confocal microscopy of a Vero cell transfected with GFP-Rab7 (green in \[E\]) or GFP-Rab5 (green in \[D\]) and Arf1T31N-myc and incubated with SFV-Cy5 (red). Cells were fixed, permeabilized, and immunostained using an antibody against myc-tag (9E10) after 1 h of internalization. SFV-Cy5 is not present in Rab5 but is present in Rab7-positive endosomes (arrowheads), indicating that Arf1 is not needed for the sorting of SFV into Rab7-positive endosomes. Scale bars represent 10 μm. \(E) Wide field fluorescence image of Bafilomycin A~1~--treated (25 nM) Vero cells expressing YFP-Rab5 and CFP-Rab7, at 2 h after internalization of SFV-AF594. SFV is present in Rab7-positive endosomes (arrowheads), indicating that no acidic pH is needed for the sorting of SFV into Rab7-positive endosomes. Scale bar represents 5 μm. \(F) NH~4~Cl-treated (20 mM) Vero cells expressing YFP-Rab5 and CFP-Rab7 incubated with Dextran--Texas Red for 2 h. Dextran is present in Rab7-positive endosomes (arrowheads), indicating that no acidic pH is needed for the sorting of SFV and Dextran into Rab7-positive endosomes. Scale bar represents 5 μm. ::: ![](pbio.0030233.g006) ::: It was relatively easy to find endosomes labeled with Arf1-GFP, RFP-Rab5, and SFV-Cy5. The three colors did not overlap completely, but moved together through the cytoplasm, indicating that Arf1, Rab5, and the virus were present in distinct domains of the same endosome. [Figure 6](#pbio-0030233-g006){ref-type="fig"}A and [Video S10](#sv010){ref-type="supplementary-material"} show how the virus leaves such an endosome. From several video movies like this, we concluded that when the virus leaves, both Arf1-GFP and RFP-Rab5 stay behind. A similar conclusion was reached when cells expressing RFP-Rab7 and Arf1-GFP were viewed. In this case, the virus and RFP-Rab7 left the endosome together, with Arf1-GFP staying behind ([Figure 6](#pbio-0030233-g006){ref-type="fig"}B; [Video S11](#sv011){ref-type="supplementary-material"}). Furthermore, when we used the myc-tagged dominant negative mutant Arf1T31N, we found SFV-Cy5 in GFP-Rab7-positive endosomes and not in GFP-Rab5-positive endosomes after 1 h of warming ([Figure 6](#pbio-0030233-g006){ref-type="fig"}C and [6](#pbio-0030233-g006){ref-type="fig"}D). Quantification showed that 86.6% of the viruses entered a Rab7-positive endosome, whereas only 5.9% remained in a Rab5-positive endosome. It was clear that SFV was not using Arf1-containing domains or vesicles for transport from the early to late endosomes, indicating that Arf1 was not needed for sorting of SFV into Rab7 vesicles. It has been reported that the transport of certain cargo by ECVs from early to late endosomes is pH-dependent \[[@pbio-0030233-b12]\]. To test in live cells whether this is true for SFV, we used drugs (NH~4~Cl and Bafilomycin A~1~) to neutralize the pH in endocytic organelles. That these agents did indeed neutralize the pH as expected was shown by a 12-fold drop in SFV infectivity in a FACS-based assay (not shown). We found that when SFV-Cy5 was allowed to enter for 2 h in the presence of the drugs, it reached Rab7-positive compartments devoid of Rab5 ([Figure 6](#pbio-0030233-g006){ref-type="fig"}E). The same was observed for a fluid phase marker, Dextran--Texas Red ([Figure 6](#pbio-0030233-g006){ref-type="fig"}F). We concluded that sorting of SFV from early endosomes to late endosomes did not require an acidic endosomal pH. Furthermore, since the virus did not penetrate under these conditions, sorting was independent of the fusion of the viral envelope with the endosomal membrane. Evidently, SFV was sorted into late endosomes and finally into lysosomes even if it had not released its capsid into the cytoplasm. The Role of Microtubules {#s2g} ------------------------ It is well known that endosomes can move along microtubules \[[@pbio-0030233-b36],[@pbio-0030233-b41]\]. When recording in cells expressing tubulin-GFP, complex bidirectional movement of GFP-Rab7-labeled late endosomes and RFP-Rab5-labeled early endosomes along microtubules could be monitored ([Video S12](#sv012){ref-type="supplementary-material"}). When the microtubule-disrupting drug nocodazole was added and the cells were incubated for 30 min, we observed larger and more numerous endosomes ([Figure 7](#pbio-0030233-g007){ref-type="fig"}A). The majority of these were positive for both GFP-Rab5 and RFP-Rab7 (approximately 75%). Again the two Rabs were, as a rule, located in nonoverlapping domains ([Figure 7](#pbio-0030233-g007){ref-type="fig"}A, arrowheads). In untransfected cells that were immunostained for EEA1 and Rab7 the effect of nocodazole was not as dramatic, but still many of the EEA1-positive endosomes were positive for Rab7 (approximately 41%) ([Figure 7](#pbio-0030233-g007){ref-type="fig"}C). ::: {#pbio-0030233-g007 .fig} Figure 7 ::: {.caption} ###### Endosomes Move along Microtubules \(A) Confocal microscopy of a Vero cell transfected with GFP-Rab5 and RFP-Rab7 after treatment with 5 μM nocodazole. Endosomes are enlarged and many (approximately 75%) Rab5-positive endosomes are also positive for Rab7 (arrowheads). Scale bar represents 10 μm. \(B) Cells treated as in (A) were incubated with SFV-Cy5, and fixed 1 h after warming. SFV-Cy5 is present in bigger endosomes positive for Rab5 and Rab7 (arrowheads). Scale bar represents 5 μm. \(C) Confocal microscopy of a Vero cell treated with 5 μM nocodazole and incubated with SFV-Cy5. After 1 h of internalization at 37 °C, cells were fixed, permeabilized, and immunostained for EEA1 and Rab7. Endosomes positive for both markers can be seen (white rectangles) (41%), and SFV-Cy5 is captured in these structures (yellow rectangles). The protrusions of Rab5 shown in (A) are not seen in (C) and are probably caused by overexpression of the Rab proteins. Scale bar represents 10 μm. ::: ![](pbio.0030233.g007) ::: When SFV-Cy5 was allowed to enter cells transfected with GFP-Rab5 and RFP-Rab7 and treated with nocodazole for 1 h, SFV-Cy5 was found to co-localize with the GFP-Rab5- and RFP-Rab7-positive endosomes ([Figure 7](#pbio-0030233-g007){ref-type="fig"}B). We observed the same in untransfected cells immunostained for EEA1 and Rab7 ([Figure 7](#pbio-0030233-g007){ref-type="fig"}C). These observations were consistent with a role for microtubules and microtubule-dependent motors in the separation of early endosomes and Rab7-positive carrier vesicles. That nocodazole had little effect on SFV infection (not shown) was expected given that penetration occurs at the level of early endosomes devoid of Rab7. Accordingly, expression of the dominant negative Rab5S34N mutant inhibits SFV infection whereas expression of the corresponding Rab7 mutant (Rab7T22N) does not \[[@pbio-0030233-b44]\]. Discussion {#s3} ========== The classical endocytosis pathway from clathrin-coated pits to lysosomes is well studied but far from fully understood. One of the unresolved issues concerns the mechanisms of molecular sorting in early endosomes, and the mechanism of cargo transfer from early to late endosomes and lysosomes. It is unclear whether the process involves maturation of early endosomes to late ones by gradual modification of membrane components, or by the formation of carrier vesicles between permanent compartments \[[@pbio-0030233-b45]\]. As reported by others the limiting membrane of endosomes contains a mosaic of domains with different composition and distinct functions \[[@pbio-0030233-b17],[@pbio-0030233-b38],[@pbio-0030233-b39],[@pbio-0030233-b46],[@pbio-0030233-b47]\]. In early endosomes, the Rab5 domains are the most prominent and best analyzed. Rab5 regulates homotypic endosome fusion and incoming vesicle traffic \[[@pbio-0030233-b06],[@pbio-0030233-b07],[@pbio-0030233-b29],[@pbio-0030233-b37]\]. The Rab4 domains, which are also present in early endosomes, organize recycling to the plasma membrane \[[@pbio-0030233-b08],[@pbio-0030233-b09],[@pbio-0030233-b38]\]. In addition, early endosomes are known to have domains containing Arf1 and COP I, caveolin-1, clathrin, and Rab11 \[[@pbio-0030233-b12],[@pbio-0030233-b37],[@pbio-0030233-b38],[@pbio-0030233-b48]\]. As cargo, we made use of labeled SFV, which we could easily follow in live cells during its journey through the endosomal system. After binding to proteoglycans and internalization via coated pits, SFV particles are known to pass rapidly from early endosomes to late endosomes and lysosomes \[[@pbio-0030233-b22],[@pbio-0030233-b23],[@pbio-0030233-b26],[@pbio-0030233-b51]\]. Triggered by the mildly acidic pH (\<6.2.) in early endosomes, membrane fusion is induced by the E1 glycoprotein, and the capsid is released into the cytosol, where the viral RNA is rapidly uncoated \[[@pbio-0030233-b22],[@pbio-0030233-b54],[@pbio-0030233-b55]\]. After fusion with the endosomal membrane the viral glycoproteins stay together as a patch \[[@pbio-0030233-b35]\], and are transported to lysosomes to be degraded together with unfused virus particles \[[@pbio-0030233-b56]\]. The most important new observation presented in this work came from following sorting of the virus from early Rab5- and EEA1-positive endosomes to late Rab7-positive, Rab5-negative endosomes. Neutralization using NH~4~Cl showed that the infective fusion reactions occurred when the virus was still in Rab5-positive endosomes devoid of Rab7. Shortly thereafter, a large fraction of the virus-containing early endosomes became positive for Rab7. In video recordings of live cells, these Rab7-enriched domains could be seen to detach with the labeled virus from the rest of the endosome, leaving no detectable Rab7 behind. The Rab7-positive vesicles were then seen to carry the virus away along microtubules. Occasionally, fusion of such vesicles with larger Rab7-containing organelles could be seen ([Videos S13](#sv013){ref-type="supplementary-material"} and [S14](#sv014){ref-type="supplementary-material"}). Co-immunostaining of EEA1 and Rab7 showed that Rab7 occurred in about one out of ten early endosomes, also in cells that did not contain any virus and did not express any fluorescent-protein-tagged proteins ([Figure S5](#sg005){ref-type="supplementary-material"}C). On the basis of these results, we propose that Rab7 can associate transiently with early endosomes. Since we did not observe arrival of GFP-Rab7 in vesicles, we assume that it binds directly from the cytosol. When abundant enough, it forms an endosomal domain that excludes Rab5, Rab4, Arf1, and transferrin. SFV moves selectively into this domain. Given the inhibitory effect of Rab7T22N, we further propose that the formation and/or detachment of these domains as independent carrier vesicles is Rab7-dependent. The mechanism of carrier vesicle formation appears to be independent of Arf1, and unlike Arf1-dependent formation of ECVs it is not influenced by endosomal pH. The final separation of the Rab7- and Rab5-containing elements is accelerated by transport of the Rab7-containing vesicles along microtubules. Our model is shown schematically in [Figure 8](#pbio-0030233-g008){ref-type="fig"}. In brief, we propose that the virus uses a Rab7-regulated pathway from early to late endosomes, and that Rab7 domains in early endosomes are involved both in the sorting of the virus and in the formation of a transport vesicle. ::: {#pbio-0030233-g008 .fig} Figure 8 ::: {.caption} ###### Schematic Representation of the SFV Transport Pathway from the Plasma Membrane to Lysosomes After uncoating of clathrin, the primary endocytic vesicles deliver the virus to Rab5-positive early endosomes, which contain Rab5 and Arf1 domains but do not contain Rab7. Rab7 is then recruited to these endosomes, and it forms additional domains (blue) that exclude the other early endosome-associated small GTPases. The virus moves into the Rab7 domain, and when this domain leaves the endosome as a Rab7-containing transport vesicle, the virus also leaves the Rab5-positive organelle. The Rab7-containing carrier moves along microtubules and fuses with other carriers or late endosomes. Also shown is an alternative route from the early to the late endosome, involving Arf1 and the ECV, which is not used by SFV. CCP, clathrin-coated pit; CCV, clathrin-coated vesicle; EE, early endosome; MT, microtubule; RE, recycling endosome; LE, late endosome; Lys, lysosome. ::: ![](pbio.0030233.g008) ::: Rab7 has been shown to be enriched in late endosomes in mammalian cells and yeast, and it is commonly thought to play an important role in fusion and transport of cargo from late endosomes to lysosomes \[[@pbio-0030233-b18],[@pbio-0030233-b57]\]. However, a role has also been suggested in the early to late endosome transport step \[[@pbio-0030233-b14],[@pbio-0030233-b57],[@pbio-0030233-b58]\]. This is because overexpression of Rab7 has been shown to accelerate transfer of some cargo molecules to late endosomes and lysosomes \[[@pbio-0030233-b58],[@pbio-0030233-b59]\]. Furthermore, transport of mannose 6-phosphate receptor (CI-MPR), CXCR2, cathepsin D, VSV G-protein, SV5 hemagglutinin-neuraminidase, and the LDL receptor are, like SFV, blocked at the level of early endosomes after expression of dominant negative Rab7 \[[@pbio-0030233-b14]\]. That Rab7 is involved in the transfer of other viruses is suggested by the report that influenza A virus, which fuses at the level of late endosomes, does not infect cells expressing the constitutively inactive Rab7 mutant protein \[[@pbio-0030233-b44]\]. Co-localization of Rab5 and Rab7 has not been observed in most localization studies \[[@pbio-0030233-b19],[@pbio-0030233-b29]\]. However, consistent with our observations, partial co-localization was recently reported in HEK 293 cells \[[@pbio-0030233-b58]\]. One reason why the co-localization may have been missed in other studies is that only a rather small fraction (approximately 14%) of endosomes have both Rabs at any given time, and when they do, Rab7 and Rab5 (and its effector EEA1) occur in distinct domains. That the proteins are, in fact, part of the same structure is particularly apparent in video recordings where the Rab5- and Rab7-positive domains can be observed to move together as part of a common organelle. The separation of the various GTPases into distinct parts in an early endosome was not only visible when they were expressed as fluorescent-protein-tagged proteins but also when viewed in fixed control cells after immunofluorescence staining. Since our results indicate that Rab7 is directly involved in the formation of late-endosome-bound carriers in early endosomes, one must ask what the function of the Arf1 domains is. These were clearly visible in about every third Rab5-positive early endosome. Do they induce ECVs for a parallel pathway to late endosomes, or do they serve some other function? Whereas Rab7-positive vesicles were frequently seen to depart from Rab5- and Rab4-positive endosomes, we did not observe detachment of Arf1-positive carriers corresponding to the description of ECVs. Evidence for the ECV pathway and its sensitivity to elevated lumenal pH has been obtained by in vitro experiments using purified endosome fractions \[[@pbio-0030233-b12],[@pbio-0030233-b13]\]. However, it is possible that more than one pathway connects early and late endosomes, and that degradation of SFV makes use of the Rab7 pathway. Whether the recruitment of Rab7 is part of an automatic process of early endosome maturation or whether it depends on the presence of specific cargo remains to be determined. Clearly, the virus as such was not required because association of Rab7 and shedding of Rab7 vesicles from early endosomes could be readily observed also in the absence of SFV. It will be important in future work to follow the fate of other cargo molecules in live cells, and determine whether there are indeed multiple pathways and, if so, how they relate to the Rab7 pathway. The identification of upstream regulatory elements and effectors of Rab7 must also become a priority. Currently, only a few Rab7-interacting factors are known, and none of them in early endosomes. The association of endosomes and carrier vesicles with microtubules also deserves close attention because microtubules and microtubule motors may play a role not only in the transport of vesicles but also in their formation or detachment. Materials and Methods {#s4} ===================== Cells, virus, antibodies, and reagents {#s4a} -------------------------------------- Vero cells were maintained in MEM (plus Earle\'s plus GlutaMAXI) supplemented with 10% fetal calf serum and nonessential amino acids (Gibco BRL, San Diego, California, United States). During live microscopy 20 mM Hepes (pH 7.4) was added to the medium to render the medium CO~2~-independent. A prototype tissue-culture-adapted strain of SFV was grown in BHK-21 cells and purified as described \[[@pbio-0030233-b26],[@pbio-0030233-b60]\]. The virus was either fluorescently labeled immediately, or stored at −80 °C. Plaque assays were performed as described \[[@pbio-0030233-b61]\]. Monoclonal antibodies against human EEA1 were purchased from Transduction Laboratories (Lexington, Kentucky, United States). Rabbit polyclonal anti-Rab7 was kindly provided by Marino Zerial \[[@pbio-0030233-b36]\]; mouse monoclonal IgG anti-LAMP-1 (H4A3) was from Santa Cruz Biotechnology (Santa Cruz, California, United States); polyclonal antisera against E1 and E2 were raised in rabbit \[[@pbio-0030233-b62]\]. AF594-X and FITC-X secondary antibodies coupled to a fluorophore, and Dextran--Texas Red were from Molecular Probes (Leiden, the Netherlands). Cy5-X was from Amersham Biosciences (Little Chalfont, England), and nocodazole was from Sigma (Steinheim, Germany). All other reagents were from Sigma. Transfection {#s4b} ------------ Cells were transfected using Nucleofactor by Amaxa (Köln, Germany) according to the following programs: Vero cells, Kit V, program A24; CV-1 cells, Kit V, program A24; and HeLa cells, Kit R, program T20. Briefly, 1 × 10^6^ Vero cells were pelleted and resuspended in 100 μl of solution V or R, and electroporated with 1--2.5 μg of DNA. The electroporated cells were resuspended in 350 μl of complete medium. Of this solution, 100 μl was seeded on one 18-mm coverslip and incubated for at least 5 h, normally overnight (15 h), at 37 °C and 5% CO~2~. Preparation of fluorophore-labeled SFV {#s4c} -------------------------------------- Fluorescent labeling of SFV was essentially done as described \[[@pbio-0030233-b26]\]. Purified virus was dialyzed in 0.1 M carbonate buffer (pH 8.3), and treated for 1 h at room temperature with the different fluorophores (Cy5, FITC, or AF594) using a 1:4 molar excess of fluorophore over E1 and E2. The fluorophores react with free amines, resulting in a stable carboxamide bond with the viral proteins. The labeled virus was purified again on a 5%--60% sucrose gradient in TN buffer (50 mM Tris \[pH 7.4\], 100 mM NaCl). The extracted virus band was stored at 4 °C. Fluorescent fusion proteins {#s4d} --------------------------- pmRFP vector was kindly provided by Roger Y. Tsien \[[@pbio-0030233-b63]\]. The RFP gene was cloned into the pEGFP-C3 vector (Clonetech, Palo Alto, California, United States). pmRFP-Rab5a was cloned using the restriction enzymes HindIII and PstI. pmRFP-Rab7 was constructed by PCR amplification of the Rab7 DNA from pEGFP-Rab7 using the primers 5′- CCCGATCTCGAGATGACCTCTAGG and 5′- CCCAAGCTTATCGATTTAACAACTGC, followed by cloning of the XhoI-HindIII fragment from the PCR product into pmRFP-C3 expression vector. pEGFP-Rab5 was kindly provided by M. Zerial. pEGFP-Rab7 was kindly provided by J. Gruenberg. Arf1-GFP was kindly provided by J. Lippincott-Schwartz. Microscopy {#s4e} ---------- For live microscopy, transfected cells were seeded on 18-mm glass coverslips and before analysis transferred to custom-built metal microscope-slide chambers (Anton Lehmann, Workshop Biochemistry, Swiss Federal Institute of Technology, Zurich, Switzerland) in medium plus 20 mM Hepes (pH 7.4). The chamber was mounted on a heated stage and analyzed at 37 °C using wide field or confocal microscopy. For wide field microscopy, cells were analyzed using a Zeiss (Oberkochen, Germany) Axiovert microscope equipped with a 100× plan-apochromat lens NA 1.40. Images were collected using a cooled charge-coupled-device camera (ORCA-ER, Hamamatsu, Hamamatsu City, Japan), and the software used was OpenLab. Alternatively, an Olympus (Hamburg, Germany) IX71microscope was used with a PlanApo 60× NA 1.45 oil immersion objective and a polychrome II monochromator (TILL Photonics, Martinsried, Germany) with the TILLvisION software, version 4.0 (TILL Photonics). For confocal microscopy the Zeiss 510Meta was used with a 100× 1.4 NA objective and with an Argon laser (458, 477, 488, 514) 30 mW, HeNe laser (543 nm) 1 mW, or HeNe laser (633 nm) 5 mW. Images were processed using Image J (National Institutes of Health, Bethesda, Maryland, United States) and Adobe Photoshop (Adobe Systems, San Jose, California, United States). Immunofluorescence {#s4f} ------------------ Cells were fixed using medium containing 4% formaldehyde, quenched with 50 mM NH~4~Cl, permeabilized with 0.05% saponin, and then incubated with primary antibodies overnight at 4 °C. Cells were incubated with the appropriate secondary antibodies. Coverslips were mounted with ImmuMount from Thermo Shandon (Pittsburgh, Pennsylvania, United States). Virus binding and infection {#s4g} --------------------------- SFV was added to cells in MEM (EAGLE) containing nonessential amino acids, 5% BSA, and 20 mM Hepes (pH 7.4) and allowed to bind for 1 h at 4 °C \[[@pbio-0030233-b26]\]. Still on ice, the cells were washed with ice-cold PBS to remove the unbound virus. Then the cells were shifted to 37 °C for different times to start the synchronized infection. For microscopy 1--3 μg of virus was used per coverslip. The timing of SFV penetration was determined as previously described \[[@pbio-0030233-b26]\]. By adding 20 mM NH~4~Cl to cells at different times after warming, acid activation of additional incoming virus particles could be blocked. To determine the fraction of cells that had been infected, a FACS-based infection assay was used \[[@pbio-0030233-b64]\]. The cells were stained with anti-E1/E2 antibodies and as secondary antibodies AlexaFluor488-labeled (green) or AlexaFluor647-R-phycoerythrin-labeled goat anti-rabbit IgG (Molecular Probes) was used. Analysis was performed on a FACS-Calibur cytometer using CellQuest 3.1f software (Becton-Dickinson Immunocytometry Systems, San Jose, California, United States). More than 50,000 cells were analyzed for each sample, and at least two independent experiments were performed. Internalization of Dextran--Texas Red {#s4h} ------------------------------------- Dextran--Texas Red (10,000 MW) was added to cells on ice for at least 30 min, then the cells were shifted to 37 °C for 10 min, followed by extensive washing with warm PBS to remove the not yet internalized Dextran--Texas Red. Cells were further incubated for 2 h at 37 °C. Degradation of the glycoproteins E1 and E2 {#s4i} ------------------------------------------ After binding of SFV at a MOI of 50 and washing, the cells were moved to 37 °C. At the indicated times, cells were placed on ice and incubated for 45 min with 0.5 mg/ml Proteinase K in PBS \[[@pbio-0030233-b26]\]. Cells were removed, washed in PBS/1 mM PMSF (phenylmethylsulfonyl fluoride)/0.2% BSA, pelleted, and dissolved in 50 μl of 50 mM Tris (pH 8.0)/2% SDS/2 mM PMSF, followed by 10 min of heating at 99 °C and TCA precipitation. The precipitate was redissolved in sample buffer, heated, and subjected to SDS-PAGE. Immunoblotting was done with an antibody against E1/E2 \[[@pbio-0030233-b62]\]. Nocodazole treatment {#s4j} -------------------- To disrupt microtubules, 5 μM nocodazole was added to the cells. They were incubated on ice for 30 min to support disruption and then analyzed or infected with SFV and then analyzed. Quantification of co-localization of different endosomal markers {#s4k} ---------------------------------------------------------------- To quantify the co-localization of different endosomal markers, at least 16 9-μm squares were randomly placed in the cytosol of a cell. The total number of EEA1-positive endosomes and the number of structures labeled by both EEA1 and the endosomal marker of interest within those squares were counted. The endosomes labeled by both markers were expressed as percentage of EEA1-positive compartments. For each experiment at least three cells were counted. Using Kinetikit/GENESIS for kinetic modeling {#s4l} -------------------------------------------- A simple model of coupled first-order reactions was assumed and implemented using Kinetikit/GENESIS \[[@pbio-0030233-b40]\] (available at [http://www.ncbs.res.in/∼bhalla/kkit/index.html](http://www.ncbs.res.in/~bhalla/kkit/index.html)). Simulations were carried out on a PC running Linux. Parameters were varied until they fit the experimental data. Note that the existence of other models that comply with the requirements cannot be excluded. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### SFV-AF594 and Immunofluorescence of E1/E2 SFV-AF594 was bound to Vero cells in the cold. The cells were fixed and immunostained with a polyclonal antibody against E1/E2. SFV-AF594 co-localizes with antibodies against E1/E2. Scale bar represents 10 μm. (8.3 MB TIF). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### Not All Virus Particles Are Internalized after 10 min of Internalization SFV-Cy5 was bound to Vero cells for 1 h at 4 °C, and unbound virus was washed away. After 10 min of warming to 37 °C to allow internalization, the cells were fixed and incubated with antibodies against E1/E2 (SFV glycoproteins) without permeabilizing the cells. After that the cells were permeabilized and incubated with anti-EEA1. Finally, the cells were stained with secondary antibodies to E1/E2 (green) and EEA1 (red). Many of the viruses are seen to be located on the outside of the cell, as indicated by the green and blue fluorescence. The internalized viruses are only labeled with Cy5 (blue), and most of them localize with EEA1 (red). (9.4 MB TIF). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### Distribution of EEA1 and Rab7 in CV-1 Cells The cells were immunostained using anti-EEA1 (green) and anti-Rab7 (red) antibodies. Arrowheads show individual endosomes positive for one of the two markers, and arrows indicate endosomes positive for two markers. Scale bar represents 10 μm. (1.1 MB TIF). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S4 ::: {.caption} ###### Distribution of EEA1 and Rab7 in HeLa Cells The cells were immunostained using anti-EEA1 (green) and anti-Rab7 (red) antibodies. Arrowheads show individual endosomes positive for one of the two markers, and arrows indicate endosomes positive for two markers. Scale bar represents 10 μm. (1.1 MB TIF). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S5 ::: {.caption} ###### Quantification of Co-Localization of Different Markers (A and B) Confocal microscopy of Vero cells immunostained for (A) EEA1 (green) and caveolin-1 (red), or (B) EEA1 (green) and COP II (red). Shown are examples of cells, which have been quantified. \(C) Co-localization was quantified by viewing GFP-Rab5- or EEA1-containing endosomes, and determining how many of them also contained RFP-Rab7, Rab7, caveolin-1, or COP II. Scale bars represent 10 μm. (2.6 MB TIF). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S6 ::: {.caption} ###### Kinetic Model and Parameters We had to add a pool of viruses trapped in the EEA1-positive compartment with bidirectional transport to the "normal" EEA1-positive compartment in order to fit the model to the experimental data. The average residence times of SFV in the different compartments can be calculated with the different k-values and the given equation for t ~1/2~. (328 KB TIF). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S7 ::: {.caption} ###### SFV-Cy5 Is Sorted in CV-1 and HeLa Cells (A and C) Confocal microscopy of (A) CV-1 and (C) HeLa cells. SFV-Cy5 (at a MOI of 20) was allowed to bind to the cells for 60 min at 4°C, followed by transfer to 37 °C for different times (0, 10, 20, or 30 min) and immunolabeled with anti-EEA1 and anti-Rab7 antibodies. Shown are three viruses for each time point. Scale bars represent 10 μm. At 0 min, SFV do not co-localize with any endosomal marker. At 10 min, SFV are localized in EEA1-positive endosomes. At 20 min, SFV are localized in EEA1- and Rab7-positive endosomes. At 30 min, SFV are localized in Rab7-positive endosomes. (B and D) Selected images obtained from a time series approximately 20 min after warming. CV-1 (B) and HeLa (D) cells were transfected with GFP-Rab5 and RFP-Rab7, and SFV-Cy5 was added. SFV-Cy5 is sorted away from Rab5-positive endosomes in a Rab7-positive vesicle. Scale bars represent 2 μm. (4.7 MB TIF). ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S8 ::: {.caption} ###### SFV-Cy5 Is Sorted with RFP-Rab7 from Transferrin-AlexaFluor488 and Rab5 Selected images, obtained from time series recorded after approximately 20 min of warming. Vero cells were transfected with (A) RFP-Rab5 or (B) RFP-Rab7, loaded for 1 h with transferrin-AlexaFluor488, and SFV-Cy5 was added. SFV-Cy5 is sorted away from transferrin- and Rab5-positive endosomes in a Rab7-positive vesicle. Scale bars represent 2.5 μm. (487 KB TIF). ::: ::: {.caption} ###### Click here for additional data file. ::: Video S1 ::: {.caption} ###### GFP-Rab5 and RFP-Rab7 Are Localized in the Same Mobile Endosome Dual-color live fluorescence microscopy experiment recorded in Vero cells expressing GFP-Rab5 and RFP-Rab7. Recording, 0.02 Hz; playback, 2 Hz. Scale bar represents 2.5 μm. Video was recorded at 37 °C with the Zeiss 510Meta confocal microscope. (920 MB MOV). ::: ::: {.caption} ###### Click here for additional data file. ::: Video S2 ::: {.caption} ###### GFP-Rab5 and RFP-Rab7 Are Localized in the Same Mobile Endosome Dual-color live fluorescence microscopy experiment recorded in Vero cells expressing GFP-Rab5 and RFP-Rab7. Recording, 0.02 Hz; playback, 2 Hz. Scale bar represents 2.5 μm. Video was recorded at 37 °C with the Zeiss 510Meta confocal microscope. (2.1 MB MOV). ::: ::: {.caption} ###### Click here for additional data file. ::: Video S3 ::: {.caption} ###### RFP-Rab7 Leaves a GFP-Rab5 Endosome Dual-color live fluorescence microscopy experiment recorded in Vero cells expressing RFP-Rab5 and GFP-Rab7. Recording, 0.17 Hz; playback, 2 Hz. Scale bar represents 5 μm. Video was recorded at 37 °C with the Zeiss 510Meta confocal microscope. (839 KB MOV). ::: ::: {.caption} ###### Click here for additional data file. ::: Video S4 ::: {.caption} ###### SFV-AF594 and CFP-Rab7 Leave a YFP-Rab5-Positive Endosome Triple-color live fluorescence microscopy experiment recorded in Vero cells expressing YFP-Rab5 and CFP-Rab7 and infected with SFV-AF594. Recorded approximately 40 min after SFV addition. False colored: YFP-Rab5 (green), CFP-Rab7 (red), and SFV-AF594 (blue). Recording, 0.32 Hz; playback, 2 Hz. Scale bar represents 2.5 μm. Video was recorded with a Zeiss Axiovert microscope. (884 KB MOV). ::: ::: {.caption} ###### Click here for additional data file. ::: Video S5 ::: {.caption} ###### SFV-Cy5 and RFP-Rab7 Leave a GFP-Rab5-Positive Endosome Triple-color live fluorescence microscopy experiment recorded in Vero cells expressing GFP-Rab5 and RFP-Rab7 and infected with SFV-Cy5. Recorded approximately 20 min after SFV addition. Recording, 0.17 Hz; playback, 2 Hz. Scale bar represents 10 μm. Video was recorded at 37 °C with the Zeiss 510Meta confocal microscope. (3.6 MB MOV). ::: ::: {.caption} ###### Click here for additional data file. ::: Video S6 ::: {.caption} ###### SFV-Cy5 Leaves an Early Endosome Positive for GFP-Rab4 and RFP-Rab5 Triple-color live fluorescence microscopy experiment recorded in Vero cells expressing GFP-Rab4 and RFP-Rab5 and infected with SFV-Cy5. Recorded approximately 20 min after SFV addition. Recording, 0.17 Hz; playback, 2 Hz. Scale bar represents 10 μm. Video was recorded at 37 °C with the Zeiss 510Meta confocal microscope. (1.5 MB MOV). ::: ::: {.caption} ###### Click here for additional data file. ::: Video S7 ::: {.caption} ###### SFV-Cy5 and RFP-Rab7 Leave a GFP-Rab4-Positive Endosome Triple-color live fluorescence microscopy experiment recorded in Vero cells expressing GFP-Rab4 and RFP-Rab7 and infected with SFV-Cy5. Recorded approximately 20 min after SFV addition. Recording, 0.24 Hz; playback, 2 Hz. Scale bar represents 10 μm. Video was recorded with an Olympus IX71 microscope. (2.7 MB MOV). ::: ::: {.caption} ###### Click here for additional data file. ::: Video S8 ::: {.caption} ###### SFV-Cy5 and RFP-Rab7 Leave an Early Endosome Positive for Transferrin-AlexaFluor488 Triple-color live fluorescence microscopy experiment recorded in Vero cells expressing RFP-Rab5, loaded with transferrin-AlexaFluor488, and infected with SFV-Cy5. Recorded approximately 20 min after SFV addition. Recording, 0.17 Hz; playback, 2 Hz. Scale bar represents 2 μm. Video was recorded at 37 °C with the Zeiss 510Meta confocal microscope. (204 KB MOV). ::: ::: {.caption} ###### Click here for additional data file. ::: Video S9 ::: {.caption} ###### SFV-Cy5 Leaves an Early Endosome Positive for Transferrin-AlexaFluor488 and RFP-Rab7 Triple-color live fluorescence microscopy experiment recorded in Vero cells expressing RFP-Rab7, loaded with transferrin-AlexaFluor488, and infected with SFV-Cy5. Recorded approximately 20 min after SFV addition. Recording, 0.17 Hz; playback, 2 Hz. Scale bar represents 2.5 μm. Video was recorded at 37 °C with the Zeiss 510Meta confocal microscope. (272 KB MOV). ::: ::: {.caption} ###### Click here for additional data file. ::: Video S10 ::: {.caption} ###### SFV-Cy5 Leaves an Early Endosome Positive for Arf1-GFP and RFP-Rab5 Triple-color live fluorescence microscopy experiment recorded in Vero cells expressing Arf1-GFP and RFP-Rab5 and infected with SFV-Cy5. Recorded approximately 20 min after SFV addition. Recording, 0.09 Hz; playback, 2 Hz. Scale bar represents 10 μm. Video was recorded at 37 °C with the Zeiss 510Meta confocal microscope. (1.7 MB MOV). ::: ::: {.caption} ###### Click here for additional data file. ::: Video S11 ::: {.caption} ###### SFV-Cy5 and RFP-Rab7 Leave an Arf1-GFP-Positive Endosome Triple-color live fluorescence microscopy experiment recorded in Vero cells expressing Arf1-GFP and RFP-Rab7 and infected with SFV-Cy5. Recorded approximately 20 min after SFV addition. Recording, 0.17 Hz; playback, 2 Hz. Scale bar represents 5 μm. Video was recorded at 37 °C with the Zeiss 510Meta confocal microscope. (1 MB MOV). ::: ::: {.caption} ###### Click here for additional data file. ::: Video S12 ::: {.caption} ###### GFP-Rab7- and RFP-Rab5-Positive Endosomes Move along Microtubules (tub-YFP) Dual-color live fluorescence microscopy experiment recorded in Vero cells expressing tub-YFP, RFP-Rab5, and GFP-Rab7. Recording, 2 Hz; playback, 6 Hz. Scale bar represents 10 μm. Video was recorded at 37 °C with the Olympus IX71 microscope. (8.2 MB KB MOV). ::: ::: {.caption} ###### Click here for additional data file. ::: Video S13 ::: {.caption} ###### Late Endosomes Labeled with CFP-Rab7 and Filled with SFV-AF594 Fuse with Other Late Endosomes Triple-color live fluorescence microscopy experiment recorded in Vero cells expressing YFP-Rab5 and CFP-Rab7 and infected with SFV-AF594. Recorded approximately 40 min after SFV addition. False colored: YFP-Rab5 (green), CFP-Rab7 (red), and SFV-AF594 (blue). Recording, 0.32 Hz; playback, 6 Hz. Scale bar represents 2.5 μm. Video was recorded with a Zeiss Axiovert microscope. (1.8 MB MOV). ::: ::: {.caption} ###### Click here for additional data file. ::: Video S14 ::: {.caption} ###### Fusion of Late Endosomes Labeled with RFP-Rab7 and Filled with SFV-Cy5 Dual-color live fluorescence microscopy experiment recorded in Vero cells expressing RFP-Rab5 and infected with SFV-Cy5. Recorded approximately 30 min after SFV addition. Recording, 0.17 Hz; playback, 2 Hz. Scale bar represents 5 μm. Video was recorded at 37 °C with the Zeiss 510Meta confocal microscope. (1.3 MB MOV). ::: ::: {.caption} ###### Click here for additional data file. ::: Accession Numbers {#s5a} ----------------- The Entrez Gene (<http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=gene> ) accession numbers of the proteins used in this paper are Arf1 (ID 375), EEA1 (ID 8411), LAMP-1 (ID 16783), Rab4 (ID 5867), Rab5 (ID 5868), and Rab7 (ID 7879). We thank all members of the laboratory for discussions and suggestions throughout this work. We thank E. Damm, E. Frickel, A. Mezzacasa, M. Schelhaas, and A. Smith for critical reading of the manuscript, and special thanks to A. Hayer for his help and input for the modeling with Kinetikit/GENESIS. For reagents we thank M. Zerial for anti-Rab7, GFP-Rab5, and GFP-Rab4; R. Tsien for pmRFP-vector; J. Gruenberg for Arf1T31N-myc and pEGFP-Rab7; and J. Lippincott-Schwartz for Arf1-GFP. We also thank A. Mezzacasa and G. Csucs for help with microscopy and A. Smith for valuable advice. The work was supported by a Boehringer Ingelheim Fonds fellowship to AV and grants from the Swiss National Science Foundation and ETH Zurich (Swiss Federal Institute of Technology Zurich) to AH. Competing interests. The authors have declared that no competing interests exist. Author contributions. AV and AH conceived and designed the experiments. AV performed the experiments. AV and AH analyzed the data and wrote the paper. Citation: Vonderheit A, Helenius A (2005) Rab7 associates with early endosomes to mediate sorting and transport of Semliki forest virus to late endosomes. PLoS Biol 3(7): e233. AF594 : AlexaFluor594 CFP : cyan fluorescent protein ECV : endosomal carrier vesicle EEA1 : early endosome antigen 1 GFP : green fluorescent protein LAMP-1 : lysosome associated membrane protein 1 MOI : multiplicity of infection PFU : plaque forming units RFP : red fluorescent protein SFV : Semliki forest virus YFP : yellow fluorescent protein	PubMed Central	2024-06-05T03:55:57.577745	2005-6-21	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1151600/", "journal": "PLoS Biol. 2005 Jul 21; 3(7):e233", "authors": [ { "first": "Andreas", "last": "Vonderheit" }, { "first": "Ari", "last": "Helenius" } ] }
PMC1151601	Introduction {#s1} ============ The SOS genetic network \[[@pbio-0030238-b01]--[@pbio-0030238-b03]\] includes more than 30 genes in Escherichia coli \[[@pbio-0030238-b04],[@pbio-0030238-b05]\] that carry out diverse functions in response to DNA damage, including nucleotide excision repair, translesion DNA replication, homologous recombination, and cell division arrest. A vast amount of biochemical, genetic and structural data are available on the various components of the network and the interactions between them, which makes the system a paradigm for studying and modeling regulation of DNA repair \[[@pbio-0030238-b06]--[@pbio-0030238-b09]\]. The network is controlled by the LexA repressor, which downregulates itself and the expression of the other SOS genes. Following DNA damage, a RecA nucleoprotein filament is formed along stretches of single-stranded DNA (ssDNA) near arrested replication forks. This RecA filament promotes autocleavage of the LexA repressor, leading to induction of the response \[[@pbio-0030238-b06]\]. While the initial stages of the SOS response have been well characterized, the temporal coordination of events during the response and its shutoff are poorly understood. Previous experiments investigated the dynamics of the response at the population level, using DNA microarrays \[[@pbio-0030238-b05]\], or a fluorescent protein as a reporter for promoter activity (PA) \[[@pbio-0030238-b10]\]. Those studies revealed an increase in the level of transcripts and of PA of SOS genes, respectively, after induction by ultraviolet (UV) irradiation. This was followed by a decrease in the activation of these genes, presumably when DNA damage was repaired and the system was shut off. Such a single-peaked response is expected from a simple model of the network, in which repression of transcription by LexA is the only regulation mechanism. Measurements performed over a population of cells might be limited in their ability to accurately describe network responses in the case of a nonhomogenous population, or an unsynchronized response. Examples include systems that show an all-or-none response \[[@pbio-0030238-b11]\] that is averaged out at the population level, steep response curves \[[@pbio-0030238-b12]\], or asynchronous oscillations \[[@pbio-0030238-b13]\] that are smeared out in ensemble measurements. Thus, a full understanding of a network\'s responses and the ability to understand them using computational models require experimental knowledge about the dynamics in individual cells. In this study, the dynamics of SOS response was investigated at high temporal resolution in individual living cells, using the green fluorescent protein (GFP) as a reporter for PA. Contrary to the single-peaked response observed in population studies, our measurements reveal that the response is highly structured, with precise temporal modulations of gene expression levels. Results {#s2} ======= The SOS Response Exhibits Discrete Activation Peaks {#s2a} --------------------------------------------------- To measure the dynamics of the SOS response at the level of individual cells, the activity of LexA-repressed promoters (recA, lexA, and umuDC) was monitored using low-copy reporter plasmids in which the promoter under investigation was fused to a gfp gene whose product becomes fluorescent within minutes of transcription initiation (gfpmut2 \[[@pbio-0030238-b14]\]). The accumulation of GFP in a cell is proportional to the rate of transcript production from the promoter \[[@pbio-0030238-b10],[@pbio-0030238-b15]\]. We used time-lapse fluorescence microscopy to measure the fluorescence intensity and size of bacteria containing the reporter plasmids over 150 min following DNA-damaging UV irradiation, at a 2-min temporal resolution. Typical snapshots obtained at two times during the response of a number of cells after a 10 J/m^2^ UV dose are shown in [Figure 1](#pbio-0030238-g001){ref-type="fig"}A and [1](#pbio-0030238-g001){ref-type="fig"}B. ::: {#pbio-0030238-g001 .fig} Figure 1 ::: {.caption} ###### Dynamics of the SOS Response Observed in Individual Cells (A and B) Snapshots from a time-lapse movie monitoring the fluorescence of live AB1157 E. coli cells taken (A) 8 and (B) 70 min after irradiation with a UV dose of 10 J/m^2^. Cells are expressing GFP as a reporter for the recA PA \[[@pbio-0030238-b10]\]. Some cells grow and undergo cell division (e.g., cell \#1), while others exhibit filamentation (e.g., cell \#2), as a consequence of DNA damage. The exposure time corresponding to (A) is ten times that for (B). \(C) Total GFP produced from the recA promoter as a function of time, measured in an individual cell irradiated at 20 J/m^2^. The full line corresponds to the data after filtration, which is used to compute the PA. \(D) PA/PA~0~ as a function of time for the same cell as in (C). (E--G) Normalized recA promoter activity PA/PA~0~ as a function of time for cells irradiated at 20, 10, and 50 J/m^2^, respectively. \(H) Average recA PA over all 23 cells in an experiment at 20 J/m^2^. \(I) recA PA/PA~0~ for two noninducible LexA(Ind^−^) cells (empty circles), and for two isogenic LexA^+^ cells (full circles), all irradiated at 20 J/m^2^. \(J) recA PA/PA~0~ from an unirradiated cell (empty circles), and lacZ PA/PA~0~ from two cells irradiated at 20 J/m^2^ (full circles). ::: ![](pbio.0030238.g001) ::: The length, L(t), and average fluorescence intensity, I(t), of each cell in the field of view were measured for each image, and the product I(t)L(t), proportional to the total amount of GFP in a given cell at time t was calculated, as illustrated in [Figure 1](#pbio-0030238-g001){ref-type="fig"}C. The PA was then computed as the rate of change of fluorescence per unit cell size: (see [Materials and Methods](#s4){ref-type="sec"}). The normalized PA (PA/PA~0~) of the recA promoter as a function of time, in representative cells irradiated with different doses of UV radiation, is plotted in [Figure 1](#pbio-0030238-g001){ref-type="fig"}D--[1](#pbio-0030238-g001){ref-type="fig"}G (PA~0~ is the average PA of uninduced cells; see [Materials and Methods](#s4){ref-type="sec"}). We find that the measured response of individual cells is highly structured. Up to three peaks in PA are typically observed within the duration of the experiments, with the typical number of peaks increasing with damage level. In contrast to the modulations found in cells following DNA damage, unirradiated cells exhibited a constant PA, close to the uninduced level, as shown in [Figure 1](#pbio-0030238-g001){ref-type="fig"}J. A non-SOS promoter (lacZ) showed a low and constant or decreasing PA following UV damage ([Figure 1](#pbio-0030238-g001){ref-type="fig"}J). As another control, we monitored the response of noninducible LexA(Ind^--^) cells (KY703), which did not show any response after damage ([Figure 1](#pbio-0030238-g001){ref-type="fig"}I). The isogenic LexA+ strain (KY700; see [Materials and Methods](#s4){ref-type="sec"}) responded in a similar way to the AB1157 strain, as shown in [Figure 1](#pbio-0030238-g001){ref-type="fig"}I. Response Peaks Exhibit Precise Timing {#s2b} ------------------------------------- To evaluate the distribution of peak parameters among the cell population, we plot in [Figure 2](#pbio-0030238-g002){ref-type="fig"}A and [2](#pbio-0030238-g002){ref-type="fig"}B the amplitude and time of the recA PA peaks for each and every cell in the experiments, for two different UV doses. We find that the data form distinct clusters, characterized by narrow variability in peak timing but larger variability in PA peak values. Note that at 50 J/m^2^, the peaks appear at later times than at 20 J/m^2^. To further characterize the timing of the peaks, we plot in [Figure 2](#pbio-0030238-g002){ref-type="fig"}C and [2](#pbio-0030238-g002){ref-type="fig"}D the corresponding histograms of peak times. The histograms exhibit three narrow peaks (standard deviation/mean less than 10%; see also [Table 1](#pbio-0030238-t001){ref-type="table"}) showing that peak timing is quite accurate among different cells under the same UV dose. In contrast, the variability of peak amplitude is larger (standard deviation/mean greater than 25%; see also [Table 1](#pbio-0030238-t001){ref-type="table"}). Some of the measured variance in PA timing and amplitude stems from cell-to-cell variability in the copy number of the plasmid-borne reporters \[[@pbio-0030238-b16]\]. Thus, these values serve as an upper limit, and actual variance for the chromosomal promoter may by even lower. Discrete activation peaks were observed also in the PA of the lexA and umuDC promoters ([Figure 2](#pbio-0030238-g002){ref-type="fig"}E--[2](#pbio-0030238-g002){ref-type="fig"}H). While no significant difference in peak timing between the lexA and recA promoters was observed ([Figure 2](#pbio-0030238-g002){ref-type="fig"}E and [2](#pbio-0030238-g002){ref-type="fig"}F), peaks in the activity of the umuDC promoter were delayed by 7--10 min ([Figure 2](#pbio-0030238-g002){ref-type="fig"}G--[2](#pbio-0030238-g002){ref-type="fig"}H). Similar delays in the decay of activity of different promoters of the SOS network have been observed in studies of populations \[[@pbio-0030238-b10]\]. ::: {#pbio-0030238-g002 .fig} Figure 2 ::: {.caption} ###### Quantitative Analysis of the Oscillatory Behavior: Distributions of Peaks\' Amplitude and Time Normalized amplitudes of the peaks in recA PA are plotted as function of peak time for individual cells irradiated with a UV dose of (A) 20 J/m^2^, (B) 50 J/m^2^. Each point corresponds to one peak in an individual cell. The three clusters in (A) (total of 51 cells) are centered at T ~1~ = 29 ± 3 min, T ~2~ = 57 ± 5 min, and T ~3~ = 93 ± 4 min (mean ± standard deviation). The average normalized promoter activities corresponding to these three clusters are: PA~1~ = 15 ± 4, PA~2~ = 19 ± 5, and PA~3~ = 13 ± 4 in units of PA~0~. (C and D) Histograms of peak times corresponding to (A) and (B), ranking each peak by its order of appearance: red: first peak in a trace of PA(t) of an individual bacteria; green: a second peak in its trace; blue: a third peak in its trace. Black lines: fits to the histograms with a sum of Gaussians. (E) Peak lexA PA as function of peak time for individual cells and (F) its corresponding histogram. (G) Peak umuDC PA as function of peak time for individual cells and (H) its corresponding histogram. Cells in the experiments (E) and (G) were irradiated with a UV dose of 20 J/m^2^. ::: ![](pbio.0030238.g002) ::: ::: {#pbio-0030238-t001 .table-wrap} Table 1 ::: {.caption} ###### Mean Values and Standard Deviations for the Peaks\' Time and Amplitude ::: ![](pbio.0030238.t001) ::: Given the precision observed in peak timings, one may wonder why temporal modulations have not been observed in experiments over cell populations. To address this, we computed the PA averaged over all individual cells in an experimental run. As [Figure 1](#pbio-0030238-g001){ref-type="fig"}H illustrates, peaks are washed out due to their slight missynchronization, highlighting the importance of carrying out these experiments in individual cells. Measurements of the response in cell cultures \[[@pbio-0030238-b10]\] show a single activation peak at about 30 min after irradiation, except for high UV doses (greater than 40 J/m^2^), where another small peak was observed around 90 min. Peak Timing Correlates with the Cell Growth Rate but Not with the Stage in the Cell Cycle {#s2c} ----------------------------------------------------------------------------------------- Measuring the response in individual cells allows not only for the evaluation of distributions of response dynamics in a population, but also for the calculation of correlations between these dynamics and parameters related to cellular growth. We find that there is a lack of correlation between the peak time and the size of the cells at the time of irradiation, as is shown in [Figure 3](#pbio-0030238-g003){ref-type="fig"}A. This indicates that peak timing is not synchronized with the bacterial cell cycle. The peak time is however correlated with the cell\'s growth rate (1/T~D~) after irradiation, as shown in [Figure 3](#pbio-0030238-g003){ref-type="fig"}B (T~D~ is the time it takes for a cell\'s length, or for the sum of lengths of its daughter cells, to double). The linear relation that exists between 1/T ~1~ and 1/T~D~ (1/T ~1~ = 1/T~D~ + 1/τ) suggests that the peaks\' timing is governed by the effective lifetime of a factor that is diluted by cell growth at a rate 1/T~D~ and is degraded at a rate 1/τ = 1/68 min^--1^ \[[@pbio-0030238-b17]\], which is independent of cell growth rate and of UV dose. Note that the timings of the three peaks T ~1~, T ~2~, and T ~3~ themselves are positively correlated ([Figure 3](#pbio-0030238-g003){ref-type="fig"}C), even at the level of individual cells: cells in which T ~1~ is larger than the average, tend to have a larger than average T ~2~ and T ~3~ as well (T ~1~, T ~2~, T ~3~ correspond to the time of the first, second, and third peak, respectively, observed in the same bacteria). As for the peaks\' amplitudes, no correlation was observed between PA~i~/PA~0~ and either the cell size at the time of irradiation or the cell growth rate. ::: {#pbio-0030238-g003 .fig} Figure 3 ::: {.caption} ###### Correlations between the Peaks\' Time and Bacterial Growth Parameters \(A) Scatter plot of the time of the first peak, T ~1~, (normalized by the population average value \<T ~1~\>) vs. the length of the cell at the time of UV irradiation, L ~0~ (normalized by the population average \<L ~0~\>). All cells from all UV doses are included; each point represents an individual cell. No correlation between the two quantities is observed (correlation coefficient = 0.02, p = 0.82). \(B) Scatter plot of 1/T ~1~ as a function of the growth rate, 1/T~D~, of individual cells irrespective of dose. A significant correlation is observed (correlation coefficient = 0.66, p \< 10^--4^). A linear fit to the data yields a slope of 1.0 ± 0.1. Inset: Cell doubling time T~D~ grows monotonically as a function of UV dose. \(C) Time of the second (T ~2~) and third (T ~3~) peaks is plotted against the time of appearance of the first peak (T ~1~). Each point corresponds to an individual cell. The data for T ~2~ correspond to 10 J/m^2^ (green), 20 J/m^2^ (red), 35 J/m^2^ (blue), and 50 J/m^2^ (magenta). The data for T ~3~ are shown in black irrespective of radiation dose for the sake of clarity. Full lines, T ~2~ = 2T ~1~, and T ~3~ = 3T ~1~ are shown as a guide to the eye. Averaging over all experiments at all UV doses we obtain: \<T ~2~/T ~1~\> = 1.99 ± 0.02 (mean ± standard error, over 132 cells), whereas \<T ~3~/T ~1~\> = 2.99 ± 0.06 (over the 60 cells that show a third peak). Peak times in A, B, and C are from measurements performed with the recA promoter. ::: ![](pbio.0030238.g003) ::: The Number of Peaks Grows with UV Dose, Their Normalized Timing Is Constant, and Their Amplitude Saturates {#s2d} ---------------------------------------------------------------------------------------------------------- Next, we analyzed the dependence of the response parameters on the UV dose. The number of peaks observed increases with the amount of damage ([Figure 4](#pbio-0030238-g004){ref-type="fig"}A). At a UV dose of 10 J/m^2^, most cells (approximately 60%) show a single activation peak, while some show two and even three peaks (approximately 25% and 15%, respectively) with a relatively low amplitude (see below). At 50 J/m^2^, on the other hand, most cells (approximately 55%) show three peaks, whereas 35% show two peaks, and no more than 10% show only one peak. In contrast, the normalized timing of the peak maxima averaged over all cells, \<T~i~\>/T (i = 1, 2, 3; T = \[1/T~D~ + 1/τ\]^--1^ ), is constant over the 10--50 J/m^2^ dose range ([Figure 4](#pbio-0030238-g004){ref-type="fig"}B). The average amplitude of the peaks shows some dependence on the UV dose, but it becomes saturated at approximately 20 PA~0~ at doses above 20 J/m^2^ ([Figure 4](#pbio-0030238-g004){ref-type="fig"}C). These observations indicate that the cells respond to increasing damage levels by increasing the number of activation cycles, rather than by increasing the amplitude of the response. ::: {#pbio-0030238-g004 .fig} Figure 4 ::: {.caption} ###### Dependence of Response Parameters on UV Dose \(A) Percentage of cells exhibiting at least zero, one, two, or three peaks at different damage levels. (B) Dependence of the mean peak time \<T~i~\>, normalized by T = (1/T~D~ + 1/τ)^--1^ (τ = 68 min), on UV dose. (C) Dependence of the mean normalized peak height \<PA~i~\>/PA~0~ on UV dose. Wild type (full symbols), ΔumuDC (empty symbols). Same color and symbol convention as in (B): \<PA~1~\>/PA~0~ (red circles), \<PA~2~\>/PA~0~ (blue triangles), and \<PA~3~\>/PA~0~ (green diamonds). In wild-type cells, the amplitude of peaks saturates at approximately 20 PA~0~ for UV doses greater than 20 J/m^2^. ΔumuDC cells do not show this saturation and reach much higher PA levels. Parameters in (A), (B), and (C) are from measurements performed with the recA promoter. ::: ![](pbio.0030238.g004) ::: The umuDC Operon Is Involved in Maintaining the Pattern of Activity Peaks and Its Precision {#s2e} --------------------------------------------------------------------------------------------- Since the modulations are observed in the activity of all three promoters investigated, it is likely that they represent modulations in the level of the master transcriptional regulator LexA. Thus, the behavior described above reveals the existence of another level of regulation of the SOS response, beyond the transcriptional control by LexA. What are the mechanism(s) underlying the precise temporal modulations of PA? It has been recently proposed that the products of the umuDC operon may act as a prokaryotic DNA damage checkpoint, effecting a timed pause in DNA replication \[[@pbio-0030238-b18]\], in addition to their role as an error-prone DNA polymerase (PolV) \[[@pbio-0030238-b19]--[@pbio-0030238-b22]\]. This has motivated our hypothesis that the products of umuDC may be involved in the mechanism behind the observed modulations. To test this hypothesis, we repeated the measurements on a strain deleted for the umuDC operon (ΔumuDC) ([Figure 5](#pbio-0030238-g005){ref-type="fig"}). We find that a number of important changes are observed in the response of this mutant, compared to the wild-type strain: first, most cells do not show the well-defined peak of PA around 60 min after irradiation. Instead, the prevailing pattern is a peak of PA around 30 min, followed by another peak appearing between 70 and 110 min, with a very large variance of cell-to-cell timing and amplitude (see [Figure 5](#pbio-0030238-g005){ref-type="fig"}A and [5](#pbio-0030238-g005){ref-type="fig"}B; [Table 1](#pbio-0030238-t001){ref-type="table"}). Second, the amplitude of the first peak reaches higher levels relative to wild type (compare [Figure 5](#pbio-0030238-g005){ref-type="fig"}A with [Figure 2](#pbio-0030238-g002){ref-type="fig"}A) and does not show saturation with UV dose (see [Figure 4](#pbio-0030238-g004){ref-type="fig"}C). Moreover, the amplitude and timing of the first peak are more correlated than in wild-type cells ([Figure 5](#pbio-0030238-g005){ref-type="fig"}A). Thus if the first peak in a ΔumuDC cell appears later than the average, it will tend to have a higher amplitude, whereas in wild-type cells, no such correlation is observed. Third, peak time becomes independent of growth rate ([Figure 5](#pbio-0030238-g005){ref-type="fig"}E). ::: {#pbio-0030238-g005 .fig} Figure 5 ::: {.caption} ###### Effects of the umuDC Operon on the Temporal Modulation of Promoter Activity \(A) Peak PA of the recA promoter as a function of peak time for individual ΔumuDC cells and (B) its corresponding histogram of the peak times. The amplitude and timing of the first peak are more correlated (correlation coefficient = 0.37, p = 0.002) than in wild-type cells ([Figure 2](#pbio-0030238-g002){ref-type="fig"}A, correlation coefficient = 0.18, p = 0.13). (C) Peak PA of the recA promoter as a function of peak time for individual AB1157 cells transformed with a noncleavable umuD mutant gene (K97A) expressed from a plasmid, and (D) its corresponding histogram of the peak times. The experiments were carried out at 20 J/m^2^. (E) Scatter plot of 1/T ~1~ as a function of the growth rate, 1/T~D~, of individual ΔumuDC cells irrespective of UV dose. In contrast with wild-type behavior (see [Figure 3](#pbio-0030238-g003){ref-type="fig"}B), experiments with ΔumuDC mutants show that the timing of first peak maxima and cell growth rate 1/T~D~ are poorly correlated (correlation coefficient = 0.15, p = 0.06, compared with a correlation coefficient = 0.66, p \< 10^--4^ measured for AB1157 cells). ::: ![](pbio.0030238.g005) ::: A further test of the involvement of umuD in setting the precision of the peak timing and its effect on the second peak is furnished by experiments with a dominant-negative, noncleavable umuD mutant gene (K97A), expressed from a plasmid in AB1157 cells ([Figure 5](#pbio-0030238-g005){ref-type="fig"}C, [5](#pbio-0030238-g005){ref-type="fig"}D, and [Table 1](#pbio-0030238-t001){ref-type="table"}). As with ΔumuDC, the amplitude of the first peak does not saturate and increases considerably; the peak at 60 min completely disappears, and in its place a minimum in recA PA is observed. Furthermore, the peak centered at approximately 90 min appears with a high timing and amplitude variance among cells. These observations indicate that the cleaved form UmuD′ is required for the reactivation of SOS PA at around 60 min after irradiation. It is interesting to note that the timing of reinitiation of DNA replication after damage reported in Opperman et al. \[[@pbio-0030238-b18]\] is similar to the timing of the observed second peak of the SOS response, and both require the existence of UmuD′. Discussion {#s3} ========== The present findings show that the SOS response is highly structured, exhibiting discrete peaks in the PA of some of the genes in the network, peaks that appear with high temporal precision. Mutations in the umuDC genes affect the precision of these modulations. A number of features of the response are important when evaluating possible mechanisms underlying the observed modulations. First, the observed similarity of the temporal patterns of activity of the three promoters studied suggests that the patterns are caused by modulation of LexA levels. This suggests that posttranscriptional regulation plays a role in regulating SOS dynamics. Second, the peak times, normalized by the average doubling time are independent of UV dose, and there is a lack of correlation between the peak time and the length of the bacteria at the time of irradiation. Thus, the normalized timing and phase of the modulations do not depend on factors such as the number of chromosomes (and hence the mean number of DNA lesions) at the time of irradiation. Third, peak timings exhibit a very low variability between cells, in spite of cell--cell variations in network component numbers (e.g., proteins), amount of damage, and the different growth stages at the time of irradiation. Our results demonstrate that the products of the umuDC operon play an important role in generating the temporal modulation of activity, maintaining its temporal precision, and endowing the SOS response with "digital pulses" \[[@pbio-0030238-b23]\], whose number but not amplitude increases with damage. In the absence of umuDC the response is analog in nature in that (i) the amplitude of the first peak increases with damage level, whereas in wild-type cells it saturates, and (ii) the amplitude of the first peak and its timing are correlated, in contrast with the amplitude-independent peak timing characteristic of a digital response. However, neither deletion of the umuDC operon nor the K97A dominant negative, non-cleavable umuD protein fully eliminates the modulations. Both mutants show peaks of activity at around 30 and 90 min after irradiation at 20 J/m^2^. Thus, other factors in the network of interactions between components of the SOS response must also play a role. The active RecA\* nucleoprotein filament is known to interact with additional factors induced by the SOS response, which modulate its coprotease activity. These factors include LexA and products of the umuDC operon, as well as the RecX \[[@pbio-0030238-b24],[@pbio-0030238-b25]\] and DinI \[[@pbio-0030238-b26],[@pbio-0030238-b27]\] proteins and double-stranded DNA during homologous recombination \[[@pbio-0030238-b28]\]. These factors might play a role in the decrease of activity after the first and third peaks, which are less affected by the umuDC gene products. In addition, the level of ssDNA in the cell changes during the response due to the progression of repair processes and the detection of new damage sites by propagating replication forks. The observed peak at 60 min may be an example of such a process, where the delayed translesion synthesis activity of PolV \[[@pbio-0030238-b29]\] causes formation of ssDNA in newly detected damage sites and, thus, reinitiation of the response. This explanation is in accord with the observed elimination of this peak in the ΔUmuDC and uncleavable UmuD mutants. This process can be more accurately timed and synchronized by the previously proposed role of uncleaved UmuD as a checkpoint inhibiting DNA synthesis after damage, and the timed release of this checkpoint by UmuD cleavage \[[@pbio-0030238-b29]\]. Other changes in activity may result from the encounter of persistent lesions during the next round of replication from OriC. This scenario is not favored by the observed lack of correlation between peak timing and the length of cells at the time of irradiation, unless there is a timed pause in initiation of replication from OriC after SOS induction. Such a pause was observed only at high levels of UV irradiation (greater than 60 J/m^2^) \[[@pbio-0030238-b30]\]. The present findings show that the progress of the response is accurately timed, irrespective of the level of the damage. The induction of the SOS response after irradiation inhibits DNA replication \[[@pbio-0030238-b31],[@pbio-0030238-b32]\] and cell division \[[@pbio-0030238-b03]\], establishing a common reference time point for all cells, which are otherwise unsynchronized in their cell cycle. Thereafter, features of the SOS network, such as the different affinities of LexA to its binding sites on the different promoters, and critical events, such as UmuD cleavage, govern the temporal execution of the response and can lead to the synchronization. Another possible mechanism behind the temporal modulation of PA is the existence of one or more negative feedback loops in the network. As mentioned above, factors such as DinI, RecX, the products of the umuDC operon, and double-stranded DNA during homologous recombination modulate the stability and coprotease activity of the nucleoprotein RecA filament. The increase in the concentration of these factors after SOS induction may lead to a decrease in the rate of LexA degradation, and consequently to a decrease in SOS induction levels, leading to a negative feedback. Any of these possibilities is likely to play a role in the reduction of PA after peaks. This type of negative feedback loops, particularly those with delays, such as transcription delays, can give rise to oscillatory behavior \[[@pbio-0030238-b33],[@pbio-0030238-b34]\]. The nearly integer ratio (1:2:3) between the timing of the peaks T~1~:T~2~:T~3~ in our experiments may support this scenario (see [Figure 3](#pbio-0030238-g003){ref-type="fig"}C and [Table 1](#pbio-0030238-t001){ref-type="table"}). Further experiments with other mutants would be required in order to explore the full mechanism behind the observed modulations and to understand their role in the cellular response to DNA damage. Of particular interest are mutations of other genes that interact with RecA, such as DinI and RecX, and mutations in RecA that show selectivity for either LexA or UmuD cleavage \[[@pbio-0030238-b35]\]. Finally, we suggest that the SOS network displays a number of design features that may also occur in other repair systems. First, it accurately times and synchronizes the repair process, thus allowing cells to respond not only according to the current amount of damage but also according to the time elapsed since damage was detected. One such timing mechanism, namely UmuD cleavage and its role as a DNA replication checkpoint, allows for an interval during which precise repair is carried out, while DNA replication is arrested and the levels of other repair enzymes are rising. In addition, mutagenesis by PolV may be limited to a short time window, starting from UmuD cleavage and ending by UmuD′ proteolysis. Second, it affords differential temporal activation of various promoters by the modulations in the level of a common repressor \[[@pbio-0030238-b10]\]. Third, the SOS network includes mechanisms to limit the response level, thereby avoiding a response that is too high at early stages. The decrease of the response after each peak may allow the cell to evaluate whether damage still persists and permit rapid shutoff when repair has been accomplished. The different controls on LexA most probably play important roles in this context: while LexA cleavage followed by proteolysis allows for a fast induction of the response, its autoregulation by negative feedback enables a quick shutoff. The results presented here for the SOS response are in striking similarity to recent observations of the behavior of the p53-Mdm2 network in individual mammalian cells \[[@pbio-0030238-b23]\], both systems showing modulations in response to DNA damage. Remarkably, the bacterial system shows modulations that are more precise than the human system, a precision that is reminiscent of that found in developmental patterning \[[@pbio-0030238-b36]\] and in circadian clocks \[[@pbio-0030238-b37]\]. It would be interesting to test whether other stress response systems show similar properties and to investigate what may be the fitness advantage of such digital responses over analog ones. Such modulations are readily detected at the single-cell level, whereas slight timing differences smear them out on cell averages. Materials and Methods {#s4} ===================== Strains and plasmids {#s4a} -------------------- The experiments were carried out in strain AB1157 (argE3, his4, leuB6, proA2, thr1, ara14, galK2, lacY1, mtl1, xyl5, thi1, tsx33, rpsL31, and supE44) \[[@pbio-0030238-b38]\] unless otherwise noted. Other strains used were WBY100 \[[@pbio-0030238-b39]\] (same as AB1157, but also ΔumuDC::cat), KY700 \[[@pbio-0030238-b40]\] (Δ\[pro-lac\]5 thi ara met srlR::Tn10) and LexA(Ind^−^) KY703 \[[@pbio-0030238-b40]\] (same as KY700, but also lexA3 malE::Tn10). The pGW2115 (K97A) plasmid carries the umuDC operon, with a point mutation Lys → Ala in position 97 of umuD \[[@pbio-0030238-b41]\]. To create the reporter plasmids, the promoter regions of the recA, lexA, umuDC, and lacZ genes were amplified by PCR from the DNA of the strain MG1655 and cloned into the plasmid pUA66 carrying the pSC101 origin of replication, using XhoI and BamHI upstream of a promoterless GFPmut2 gene as described in \[[@pbio-0030238-b10],[@pbio-0030238-b15]\]. Previously, it has been verified that the pSC101 average plasmid copy number per cell does not change after UV irradiation \[[@pbio-0030238-b10]\]. The addition of about 10 LexA binding sites due to the reporter plasmids \[[@pbio-0030238-b16]\] is expected to have a small effect on LexA occupation and on its regulation of the response because LexA has about 40 natural binding sites in the chromosome. Experimental setup {#s4b} ------------------ Experiments were carried out in a home-built inverted microscope, whose temperature was controlled and set to 37 °C. Images were acquired with an intensified camera (Videoscope International, Dulles, Virginia, United States; ICCD-350F), with integration times ranging from 0.1 to 4 s, and stored in a computer for later analysis. The samples were illuminated with an argon laser (488 nm) only during the time of integration of the camera to reduce photobleaching. Optical filters used were 480AF30 for excitation, 530DF30 for emission, and 505DRLP dichroic mirror (Omega Optical, Brattleboro, Vermont, United States). The incident power on the back aperture of the objective was approximately 5 μW, and the illumination area had a radius of approximately 50 μm. At these illumination level and exposure times, photobleaching was only appreciable after 4 h of experiment. Sample preparation {#s4c} ------------------ Cells were grown overnight in an LB medium and diluted into a fresh medium used in the experiments (M9 + 20 amino acids except tryptophan, 50 mg/l; thiamine 20 mg/l; thymine 20 mg/l; biotin 1 mg/l; glucose 0.4% v/v). After reaching midlog phase (OD~600~ = 0.25--0.4), cells were placed on a preheated agarose slab (experimental medium + 2% agarose) and incubated for 15 min at 37 °C. Cells were then irradiated in situ with UV light (wavelength 254 nm), using a low-pressure mercury germicidal lamp at levels between 10 and 50 J/m^2^ (20 J/m^2^ corresponds to an irradiation time of 12 s). After UV irradiation, bacteria were covered with a coverslip and monitored using the fluorescence microscope. The thin layer of agar allowed for an efficient supply of oxygen and nutrients to reach the cells during the experiment. Our irradiation procedure avoids any inhomogeneities due to the considerable UV absorption by the liquid media. Data analysis {#s4d} ------------- The average intensity in a cell I(t) and its length L(t) were measured from fluorescence images after background subtraction, using MATLAB software (MathWorks, Natick, Massachusetts, United States). The product I(t)L(t) is proportional to the total amount of GFP within the cell at time t. Since GFP degradation and photobleaching are found to be negligible during the experiment, the time derivative of this amount corresponds to the GFP production rate, or PA: . The normalization by L(t) is analogous to the normalization by the optical density in measurements of cell populations \[[@pbio-0030238-b10]\]. To reduce noise in the PA calculation, which stems mainly from focus changes between consecutive images, we filtered I(t)L(t) with a digital Butterworth low-pass filter of order 4, and a cutoff frequency of 1/32 min^--1^. The filtration process was phase-preserving, to eliminate time delays. Filter parameters were chosen for maximal noise reduction while preserving the important dynamic features of the PA. The same filter parameters were used for all experiments. We verified that changing the filter\'s order or cutoff frequency by ± 50% did not significantly change the peaks\' parameters. Values of PA were normalized by PA~0~, which was determined from the steady-state solution to the following equation for uninduced cells: dI/dt = PA~0~ − (ln2/\<T~D~\>)I, where \<T~D~\> is the mean doubling time of uninduced cells. T~D~ is defined as the time it takes for a cell length, or for the sum of lengths of its daughter cells, to double, and it was extracted from the slope of exponential fits to L(t). The use of unstable GFP is not necessary since the high temporal resolution allows tracking of production changes. Control experiments in which high levels of GFP production were induced (fully induced lac promoter) neither showed modulations, nor affected the growth of the cells. Research was supported by the Clore Foundation. NF acknowledges support from a Weizmann Institute postdoctoral fellowship. SV was supported by a fellowship from the Center for Complexity Science. We thank Z. Livneh and J. Little for useful conversations, Z. Livneh for providing strains, G. C. Walker for plasmids, and G. Lahav for a careful reading of the manuscript. Competing interests. The authors have declared that no competing interests exist. Author contributions. UA and JS conceived and designed the experiments. NF and SV performed the experiments and analyzed the data. MR contributed reagents/materials/analysis tools. NF, SV, UA, and JS wrote the paper. ¤a Current address: Department of Chemistry and Chemical Biology, Harvard University, Cambridge, Massachusetts, United States of America ¤b Current address: Department of Molecular Pharmacology, Stanford University CCSR, Stanford, California, United States of America Citation: Friedman N, Vardi S, Ronen M, Alon U, Stavans J (2005) Precise temporal modulation in the response of the SOS DNA repair network in individual bacteria. PLoS Biol 3(7): e238. GFP : green fluorescent protein PA : promoter activity PA~0~ : average PA of uninduced cells PA/PA~0~ : normalized PA ssDNA : single-stranded DNA UV : ultraviolet	PubMed Central	2024-06-05T03:55:57.582577	2005-6-21	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1151601/", "journal": "PLoS Biol. 2005 Jul 21; 3(7):e238", "authors": [ { "first": "Nir", "last": "Friedman" }, { "first": "Shuki", "last": "Vardi" }, { "first": "Michal", "last": "Ronen" }, { "first": "Uri", "last": "Alon" }, { "first": "Joel", "last": "Stavans" } ] }
PMC1151602	Any cell that receives a dose of radiation is placed in a dangerous situation. The DNA damage resulting from exposure to such radiation (or any other mutagen) can cause massive rearrangements to genetic information and potentially kill the cell. Bacteria have learned to cope with this threat by activating genes that repair DNA damage and by preventing a cell from dividing before these repairs are completed. In the bacteria Escherichia coli, these repair genes form what is known as the SOS response. The E. coli SOS response has been used to study DNA repair for decades, and a great deal is known about how the more than 30 genes involved in the response function. Two proteins figure prominently in this response. The LexA protein acts as a repressor and inhibits the expression of SOS genes under normal conditions; in the event of DNA damage, the protein RecA inactivates the LexA repressor by enhancing its autocleavage into two fragments, which initiates the SOS response. While these initial stages are well understood, how all the SOS genes are coordinated, and ultimately turned off, is only beginning to be explored. In a new study, Joel Stavans, Uri Alon, and colleagues have closely followed the SOS response in individual E. coli cells to investigate its dynamics. Previous studies, which monitored the temporal pattern of activation of entire populations of cells, found that SOS genes turned on in one peak upon DNA damage. But Friedman et al. found that SOS genes in individual bacteria respond to DNA damage in three precisely timed phases. This observation reveals the importance of examining complex processes at the level of single cells, while furthering our understanding of how the SOS response is structured in E. coli. [](#pbio-0030239-g001){ref-type="fig"} ::: {#pbio-0030239-g001 .fig} ::: {.caption} ###### Genes involved in the SOS response to DNA damage are expressed in three precisely timed phases ::: ![](pbio.0030239.g001) ::: Friedman et al. monitored the SOS response by attaching a green fluorescent protein (GFP) to the promoters (the section of DNA responsible for activating a gene) of three SOS genes (lexA, recA, and umuDC). Bacteria expressing these promoter-GFP fusions became fluorescent within minutes of being exposed to UV radiation, visualized using time-lapse fluorescence microscopy. Since GFP fluorescence is directly correlated with the expression of each of the chosen genes (i.e., their promoter activity), the authors could gauge the SOS response rate upon DNA damage. To induce the SOS response, the authors exposed E. coli cells to UV radiation. By monitoring individual cells at two-minute intervals after this dose, Friedman et al. found up to three peaks of promoter activity at roughly 30, 60, and 100 minutes. Although the amount of this activity and the average number of peaks varied between cells, the timing was always similar in different cells, suggesting a highly structured, timed response. When the authors averaged this response over the population, it "washed out" into a single peak---which explains why the three peaks of expression were not previously detected. A deeper look into the dynamics of the SOS response in single E. coli cells showed that it did not correlate with cell size, suggesting the SOS response is not synchronized with the cell cycle. In addition, Friedman et al. repeated their experiments in a bacterial strain lacking the SOS response gene umuDC. The peak pattern was altered in this mutant strain, and the precision in the appearance of the peaks was reduced. By re-examining the SOS response in single cells, Friedman et al. have visualized an accurately timed and synchronized DNA repair process. Modulations in response to DNA damage have also been observed recently in individual mammalian cells. Future experiments in E. coli---one of the most genetically tractable model systems---should help explain how this timed response is related to the different pathways of DNA repair and shutoff of the response.	PubMed Central	2024-06-05T03:55:57.586055	2005-6-21	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1151602/", "journal": "PLoS Biol. 2005 Jul 21; 3(7):e239", "authors": [] }
PMC1151603	Life is hard for bacteria. Not only must they constantly compete against their comrades for resources and living space, they\'re also subject to infection by pathogens---viruses called bacteriophages---which can affect their ability to survive and prosper. Two types of bacteriophages threaten bacteria: lytic phages and lysogenic (or temperate) phages. Acquisition of a lytic phage (for example, T2, T4, or T6) is an immediate death sentence for the bacterium; upon infection, a lytic phage subverts the bacterium\'s biochemical machinery to make copy after copy of itself until the bacterium bursts, or lyses, from the burden. In contrast, a temperate phage (for example, λ phage) can lie dormant for many generations before it co-opts the bacterium\'s machinery to reproduce, but eventually it, too, lyses the bacterial cell as it releases a host of new phages. From the perspective of the bacterium, it is better to be infected by a temperate phage than a lytic phage because infection with a lytic phage means instant death, while a temperate phage may lie dormant long enough for the bacterium to reproduce. [](#pbio-0030258-g001){ref-type="fig"} ::: {#pbio-0030258-g001 .fig} ::: {.caption} ###### The bacterial virus λ integrates into the E. coli genome, where it shuts down the cell\'s ability to grow on poor carbon sources ::: ![](pbio.0030258.g001) ::: Temperate phages achieve dormancy by producing a phage gene product (in the case of λ phage, called cI) that represses the production of other phage genes; phage reproduction ceases as long as this repressor is produced. Once infected by a temperate phage, bacteria are protected from secondary infections by various other phages, because the temperate phage prevents the others from becoming established in the cell. But might temperate phage infection confer other advantages on bacterial survival? Edward Cox\'s group at Princeton University examined this question by looking for evidence that temperate phage infection triggers changes in bacterial behavior. Working with λ phages, the authors studied how phage infection affects the regulation of genes that might impact the bacterium\'s survival by comparing the constellation of genes expressed in uninfected E. coli bacteria to those in E. coli carrying a dormant λ phage. They found that λ phage caused reduced expression of the bacterial gene pckA, which codes for an enzyme that helps bacteria grow on carbon sources (fuels) other than glucose; without functioning pckA, bacteria grow normally in an environment containing glucose, but grow only slowly in an environment containing alternative carbon sources such as succinate. E. coli carrying λ phage fail to make the pckA gene product because the pckA gene is turned off by the virally encoded repressor cI. Interestingly, the researchers found evidence that the repressors made by other temperate phages may also be able to turn off pckA expression, and that the pckA genes of other bacteria related to E. coli might also be regulated by temperate phage repressors. The fact that this relationship between temperate phage repressors and regulation of the pckA gene is so well conserved argues that the ability to turn off this gene might be positively selected; therefore, pckA repression must confer some sort of survival benefit to the bacterium. It\'s not clear what this benefit might be, but one explanation is that slowing bacterial growth in glucose-poor environments might help the bacterium elude detection by the immune system of any animal it invades, increasing its chances of survival. Alternatively, slower bacterial growth might slow down the onset of viral reproduction and eventual lysis. Regardless, it is clear that there is a strong relationship between the temperate phages and the bacteria they colonize. These results have significant implications for the evolution of fitness in these bacterial populations.	PubMed Central	2024-06-05T03:55:57.586763	2005-6-21	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1151603/", "journal": "PLoS Biol. 2005 Jul 21; 3(7):e258", "authors": [] }
PMC1151604	Cell membranes are largely made of proteins, and membrane proteins account for about a third of all genes. Despite their importance, they are devilishly hard to isolate and stabilize, and therefore are hard to study. The problem lies in their structure: membrane proteins have at least one hydrophobic domain, composed of a stretch of water-repelling amino acids, which holds the protein snugly in the lipid membrane. Purifying such a protein in an aqueous medium makes the hydrophobic parts aggregate, destroying the protein\'s delicate three-dimensional structure and often disrupting its function. The alternative is to extract the protein with a detergent, a two-headed "Janus" molecule with both hydrophobic and hydrophilic ends. The protein remains surrounded by the hydrophobic ends, while water clusters at the hydrophilic ends, easing the protein out of the membrane and into solution, where it can be studied. To date, though, relatively few complex membrane proteins have been successfully purified with available detergents. In this issue, Shuguang Zhang and colleagues show that a simple amino acid--based detergent can successfully stabilize the dauntingly large protein complex photosystem I (PS-I), an integral part of the photosynthetic machinery. The molecule they made, abbreviated A~6~K, links six units of the hydrophobic amino acid alanine to one of the hydrophilic amino acid lysine. The authors used it to stabilize PS-I and then attached the detergent--protein complex to a glass slide, allowed it to dry, and examined the stability of PS-I by testing its fluorescence. Intact PS-I emits red light with a characteristic peak wavelength; as it degrades, this peak subsides and is replaced by another, bluer peak. Even the two best standard detergents did poorly at maintaining the red peak. In contrast, the spectrum after A~6~K extraction was almost a perfect match for the normal one, indicating the complex was largely intact after drying. Furthermore, the complex appeared to remain stable for up to three weeks on the glass slide. The potential applications of this work are severalfold. PS-I itself remains to be fully characterized, and this stabilization technique offers new means to explore its properties. In addition, an isolated and stabilized form of PS-I may hold some promise as an alternative energy source, since it generates an electric current in sunlight. Perhaps most importantly, the full potential of such simple amino acid--based detergents has only begun to be explored. It is likely that either this one, or others like it, can be used to isolate and stabilize hundreds of other membrane proteins, allowing them to be studied in detail for the first time. [](#pbio-0030259-g001){ref-type="fig"} ::: {#pbio-0030259-g001 .fig} ::: {.caption} ###### The designed short peptide (protein fragment) detergents look like matches and behave like lipids or oil molecules that repel water at one end but attract water at the other end ::: ![](pbio.0030259.g001) :::	PubMed Central	2024-06-05T03:55:57.587372	2005-6-21	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1151604/", "journal": "PLoS Biol. 2005 Jul 21; 3(7):e259", "authors": [] }
PMC1151605	When Robert Hooke first looked at cork bark with a light microscope in 1655, he saw small empty chambers, reminiscent of monastery cells. We now know that living cells are full of organelles---specialized subcompartments surrounded by membranes in which different cellular life functions occur. This complex organization raises major transport and sorting problems similar to those encountered in a large city in which trains and trucks carrying different cargos arrive at peripheral distribution centers. The cargos must be sorted and transported to individual factories where goods are made for delivery to other city destinations or for export. At the same time, the different areas of the city produce waste products that also need to be sorted and transported correctly. Somehow, thousands of cargos must end up in exactly the right place in both the city and the cell. One cellular system that sorts and transports cargos is the clathrin-mediated endocytic pathway. Endocytosis---the ingestion of materials into the cell---is important for the interaction of cells with the environment because it allows the uptake of nutrients (the equivalent of the raw materials brought into the city) and signaling molecules (the letters brought in by the mail service). In clathrin-mediated endocytosis, materials arriving at the outside surface of the cell are engulfed in special areas of membrane known as coated pits, which pinch off to form intracellular vesicles. These lose their clathrin coat and other molecules involved in their formation to become early endosomes, a specific sort of intracellular vesicle. The cargos are then transferred to late endosomes, which have different proteins and functions than early endosomes. From endosomes, cargo can go either to lysosomes, where they are degraded, or to the Golgi apparatus, which sends cargo back to the cell surface. Although many details of clathrin-mediated endocytosis have been uncovered, cell biologists still hotly debate whether early endosomes mature into late endosomes or whether transport vesicles take cargos from early to late endosomes. Unraveling such details will improve our understanding of normal cellular processes and should help in the design of intracellularly targeted drugs. Andreas Vonderheit and Ari Helenius now provide new insights into this controversy by examining how Semliki forest virus (SFV) is sorted and transported to late endosomes. [](#pbio-0030260-g001){ref-type="fig"} ::: {#pbio-0030260-g001 .fig} ::: {.caption} ###### Immunofluorescence showing intracellular compartments containing Rab7 (red), EEA1 (green), and Semliki Forest Virus (blue) ::: ![](pbio.0030260.g001) ::: Like many animal viruses, SFV enters its host cells using clathrin-mediated endocytosis. One well-established way to study this process is to attach a fluorescent tag to individual virus particles and observe their travels through the cell. Vonderheit and Helenius now track this journey in greater detail than ever before by attaching different colored fluorescent tags to SFV and to protein markers of early and late endosomes. They then use video-enhanced triple-color microscopy to follow all the markers as they move through living cells. This analysis reveals that the virus is initially present in endosomes containing only proteins associated with early endosomes. Then, Rab7, a late endosome marker that is involved in transport of cargo from early to late endosomes, appears in distinct domains of these early endosomes. Finally, the viral cargo is transferred to a detached organelle that contains Rab7 but no early endosome markers. The researchers show that SFV transport to late endosomes requires Rab7 and the presence of intact microtubules, which often serve as a highway network along which vesicles travel. The researchers conclude that, at least for SFV, the mechanism underlying sorting and transport from early to late endosomes falls somewhere in between the two existing models for clathrin-mediated endocytosis. Early endosomes, they postulate, have to acquire some characteristics of late endosomes before SFV can be transported to late endosomes in Rab7-positive vesicles. But other cargos, the authors point out, may follow different pathways through the cell.	PubMed Central	2024-06-05T03:55:57.587901	2005-6-21	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1151605/", "journal": "PLoS Biol. 2005 Jul 21; 3(7):e260", "authors": [] }
PMC1151649	Introduction ============ In 1959 Frederick George Roberts founded the Chiropractic and Osteopathic College of Australasia (COCA). The Melbourne based College graduated about two hundred chiropractic and osteopathic practitioners from the period 1959 to 1979. The College closed its undergraduate program in 1979 and the students transferred to the Preston Institute of Technology chiropractic program. This is now the Royal Melbourne Institute of Technology (RMIT University) chiropractic course. An osteopathic course commenced alongside the chiropractic course at Phillip Institute of technology (now also RMIT University) in 1986. Even though it has now closed its undergraduate operations, the College has maintained its company structure and acted as a repository for the records of its alma mater \[[@B1]\]. In 1990 another organisation the Chiropractors and Osteopaths Musculo-Skeletal Interest Group (COMSIG) commenced. Several years later and after steady growth, COMSIG underwent a name change and incorporated under the company structure and banner of the Chiropractic & Osteopathic College of Australasia. From this beginning COCA has grown into the leading provider of post-graduate vocational training for both professions in Australia \[[@B1]\]. In 1992 COMSIG started its own journal and this was known as COMSIG Review. In 1995 after incorporation under the COCA banner the journal changed its name to Australasian Chiropractic & Osteopathy. It is this journal that has changed from a print journal to the Open Access, online journal Chiropractic & Osteopathy. The Chiropractic and Osteopathic professions ============================================ Both chiropractic and osteopathy are over a century old. They are now regarded as complementary health professions having started their evolution as alternative health groups; this evolution is still underway. There is an imperative for both professions to research the principles and claims that underpin them, and Chiropractic & Osteopathyprovides a scientific forum for the publication of such research. For many years both professions were driven by ideology rather than science. This has gradually changed and the pursuit of science is accelerating. The make up of the Editorial Board of Chiropractic & Osteopathyreflects this change. The Editorial Board =================== The Editorial Board (<http://www.chiroandosteo.com/edboard>) is a mix of academics and researchers from a broad cross-section of professions. These include chiropractic, osteopathy, epidemiology, public health, orthopaedics, surgery, rheumatology, biomechanics, education, ergonomics, biostatistics, demography, sociology and radiology. The main Editorial team (Bruce F. Walker, Simon French, Melainie Cameron, John Jannese and Alan Ralph) take care of the day to day running of the journal. Journal content =============== Chiropractic & Osteopathywill encompass all aspects of evidence-based information that is clinically relevant to chiropractors, osteopaths and related health care professionals. Manuscripts submitted to Chiropractic & Osteopathywill initially be reviewed by the Editorial team and subsequently by two external reviewers. Reviewers will be asked to indicate whether the manuscript is scientifically sound, relevant and also to indicate the level of interest to chiropractic and osteopathy professionals. Chiropractic & Osteopathywill consider the following types of manuscripts: primary research, case reports, reviews (both systematic and narrative), commentaries, database articles, debate articles, hypotheses, methodology articles, short reports and study protocols. The journal has Australian ancestry but now has an international span. As an Open Access journal, Chiropractic and Osteopathywill reach a wider audience, enabling the sharing of knowledge with practitioners, researchers and clinicians worldwide. The aim of the journal is to advance the body of chiropractic and osteopathic knowledge. Chiropractic & Osteopathyis published by BioMed Central (<http://www.biomedcentral.com>), an independent publishing house committed to ensuring that peer-reviewed biomedical research is Open Access -- immediately and permanently available online without charge or any other barriers to access. We at Chiropractic & Osteopathylook forward to receiving your submissions. Researchers interested in submitting a manuscript should see the Instructions to Authors (<http://www.chiroandosteo.com/info/instructions/default.asp>). What is Open Access? ==================== Chiropractic and Osteopathy\'sOpen Access policy changes the way articles are published. First, all articles become freely and universally accessible online, and so an author\'s work can be read by anyone at no cost. Second, the authors hold copyright for their work and grant anyone the right to reproduce and disseminate the article, provided that it is correctly cited and no errors are introduced \[[@B2]\]. Third, a copy of the full text of each Open Access article is permanently archived in an online repository separate from the journal. Chiropractic and Osteopathy\'sarticles are archived in PubMed Central \[[@B3]\], the US National Library of Medicine\'s full-text repository of life science literature, and also in repositories at the University of Potsdam \[[@B4]\] in Germany, at INIST \[[@B5]\] in France and in e-Depot \[[@B6]\], the National Library of the Netherlands\' digital archive of all electronic publications. The benefits of Open Access =========================== Open Access has four broad benefits for science and the general public. First, authors are assured that their work is disseminated to the widest possible audience, given that there are no barriers to access their work. This is accentuated by the authors being free to reproduce and distribute their work, for example by placing it on their institution\'s website. It has been suggested that free online articles are more highly cited because of their easier availability \[[@B7]\]. Second, the information available to researchers will not be limited by their library\'s budget, and the widespread availability of articles will enhance literature searching \[[@B8]\]. Third, the results of publicly funded research will be accessible to all taxpayers and not just those with access to a library with a subscription. As such, Open Access could help to increase public interest in, and support of, research. Note that this public accessibility may become a legal requirement in the USA if the proposed Public Access to Science Act is made law \[[@B8]\]. Fourth, a country\'s economy will not influence its scientists\' ability to access articles because resource-poor countries (and institutions) will be able to read the same material as wealthier ones (although creating access to the internet is another matter \[[@B9]\]). The Editorial Team look forward to an exciting time ahead for Chiropractic & Osteopathy.	PubMed Central	2024-06-05T03:55:57.588461	2005-4-11	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1151649/", "journal": "Chiropr Osteopat. 2005 Apr 11; 13:1", "authors": [ { "first": "Bruce F", "last": "Walker" }, { "first": "Simon D", "last": "French" }, { "first": "Melainie", "last": "Cameron" } ] }
PMC1151650	Introduction ============ Obesity is a problem of epidemic proportion \[[@B1],[@B2]\]. Despite record rates of non-physician supervised dieting and the availability of numerous weight loss programs, the problem is not abating \[[@B3]\]. Complicating this, is that most primary care physicians do not treat obesity, citing a lack of time, resources, insurance reimbursement, and knowledge of effective interventions as significant barriers \[[@B4]\]. Musculoskeletal disorders including low back pain (LBP) represent a considerable public health problem and a common diagnosis creating absenteeism and the need for disability pensions \[[@B5]\]. It is estimated that about 80% of the United States and Canadian population will experience LBP during adulthood \[[@B6]\]. Most low back pain is self-limiting and will ultimately resolve in two weeks (50% of those affected) to six weeks (90% of those affected), however it remains an intriguing clinical problem \[[@B6]\]. It is widely noted that the economic cost of obesity and its related disorders are staggering, with lifestyle related conditions such as diabetes mellitus and coronary heart disease placing a large economic burden on the health care system \[[@B1]-[@B4]\]. However, low back pain also has a significant socioeconomic impact. Cost estimates range from US\$20 billion to \$50 billion annually, with 10% of the patients accounting for 85 to 90% of the costs \[[@B6]\]. In Australia, Walker et al estimated the cost of low back pain in 2001 alone to be AUD\$9.17 billion \[[@B7]\]. One question, which arises from the discussion concerning obesity, is whether obesity is a risk factor for low back pain. \"Buckwalter et al contended that a number of medical conditions including obesity, along with diabetes and hypertension, may influence the pathophysiology of diseases of the tendons and ligaments during the process of aging thus potentially leading to low back pain \[[@B8]\]. Along with low back pain, the conventional wisdom is that overweight persons are at risk of osteoarthritis in weight-bearing joints such as the knee, the hips, and feet \[[@B9]\]. To date, literature reviews have given conflicting views based on the available data and method of data retrieval. The purpose of this review is to establish, from recent research, if there is a causal link between obesity and the affliction of low back pain. A secondary purpose of this review is to present the concepts of evidence-based practice to aid the chiropractor or osteopath in looking for health-related evidence for their patients who present with obesity. Methods ======= A MEDLINE search, from the National Library of Medicine, was used to ascertain pertinent articles between the years 1990 and 2004. The use of keywords \"obesity\", \"body mass index\", \"BMI\" and \"low back pain\" was used to obtain relevant studies. The references of papers retrieved were also reviewed, as were key texts and references. This section is devoted to presenting the concepts of evidence-based practice (EBP) to demonstrate to the reader the discovery process for finding a possible link between obesity and low back pain. The EBP method used is shown in Table [1](#T1){ref-type="table"}. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Steps to asking the answerable question using EBP principles ::: ---------------------------------------- Step 1: Asking an answerable question Step 2: Selecting an evidence resource Step 3: Executing the search strategy Step 4: Examining the evidence summary Step 5: Application of the evidence ---------------------------------------- ::: The first step in this process is \"asking an answerable question.\" In this paper we assume a patient has asked whether being overweight can cause low back pain. Construction of an appropriate answerable question would possibly be \"Does an increased BMI cause low back pain?\" \"Does being overweight create osteoarthritis?\" In this way questions can be constructed to allow the practitioner to effectively answer a clinical concern. Once the answerable question has been constructed the next task is to find an adequate resource. Internet access to the U.S. National Library of Medicine\'s MEDLINE or PUBMED, these database systems are considered by many experts to be the most up to date data source on medically related topics. The next step is to determine keywords to place in the search engine. From the answerable question(s) it can be appreciated that the initial keywords will be \"low back pain\", \"BMI\" and \"osteoarthritis.\" This search constitutes the third step. In initiating the search, one should look for the search engines \"limits\" area. In this area can one designate an age group (ex: 45 to 60 years), date span of the literature search, (ex: 1998 to 2003), and to select either English language or articles in foreign languages. Once the search has been completed, the articles, which may answer the question, are isolated and can be read. Step four involves collating the evidence to answer the question. In searches one may find answers that were not known to exist, and information that may challenge an already preconceived notion. The evidence summary should list the main author and year the paper was published. This is in order to retrieve the data should anyone wish to examine the source. As an example of an evidence summary see Table [2](#T2){ref-type="table"}. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Recent evidence: Obesity and low back pain (chronological order) ::: Author, Year (Ref) N BMI LBP Association ----------------------------- ------------- ---------------------------- --------------- ------------------------ Melissas, 2003 \[14\] 50 \>40 58% direct Bener, 2003 \[10\] 802 (26.4 males/ 27.8 females) 56.1% males 73.8% females moderate Tsuritani, 2002 \[16\] 709 \-- 40.3% none Bowerman, 2001 \[4\] 252 \-- 29.2% none Kostova, 2001 \[12\] 898 \-- \-- increased risk Bayramoglu, 2001 \[15\] 25 \-- \-- direct Mortimer, 2001 \[13\] 475 30 (43.6%) 31--40 (28.8%) 40+ (1.3%) \-- increased risk Han, 1997 \[11\] 7018 women 5887 men NR \-- females increased risk N = number; BMI = body mass index; LBP = low back pain; NR = not reported ::: It is at this time that the clinician is ready for the final step of applying the evidence. In our example clinical data from the experimental literature may or may not indicate that there is a link between overweight and low back pain. Results ======= The literature search into obesity and joint pain revealed several studies pertinent to the debate. Table [2](#T2){ref-type="table"} reveals an overview of the studies selected. Several studies \[[@B4],[@B10]-[@B13]\] had large populations to draw from yet the data from these studies were not in agreement as to a cause or association. In fact, only two studies \[[@B14],[@B15]\] found a direct association for obesity as a risk factor while two \[[@B4],[@B16]\] studies found no association. Several of the studies reviewed were unable to clarify BMI to the satisfaction of the authors. Discussion ========== Interest in the association between obesity and low back pain has piqued researchers interest for many years. Intuitively, a burgeoning waistline and an increased lordotic lumbar spine led researchers to conclude that overweight people would be more prone to low back pain. Historically, Kellgren and Lawrence (1958) found that that the prevalence of disk degeneration with obesity was not significant \[[@B17]\]. However, it was not until the mid-1970\'s when several studies observed a possible association. Obesity was found to increase the prevalence of disk degeneration significantly in a study by Magora and Schwartz in 1976 \[[@B18]\]. Barton et al (1976), in a review of 144 cases, found that 70% of those who complained of low back pain had been classified as being overweight \[[@B19]\]. This basic research appeared to conclude what was already intuitively thought about low back pain and increased weight. Body mass index --------------- Before an in-depth discussion of low back pain and obesity can ensue, the concept of Body Mass Index (BMI) needs to be discussed. BMI is a measure of fatness and is calculated by dividing the patient\'s weight in kilograms by height in metres squared kg/M^2^\[[@B20]\]. It is widely accepted, easily measured, and predicts morbidity and mortality in many populations \[[@B15]\]. Obesity is generally defined as a BMI of 30 kg/m2 and higher \[[@B20],[@B21]\]. Overweight is defined as a BMI between 25 and 30 kg/m2 \[[@B19],[@B20]\]. Overweight tends to be more common in men with obesity being more prevalent among women \[[@B21]\]. When body weight is increased 20% above average, mortality rises to 20% for men and 10% for women \[[@B22]\]. (Table [3](#T3){ref-type="table"}) Overweight individuals demand more from their cardio-respiratory and musculoskeletal systems \[[@B22]\]. It is known that more than 50% of adult Americans have a BMI equal to or greater than 25 \[[@B23]\]. Although there are certain limitations to BMI i.e. large muscular athletes who are in good cardiovascular shape, the rationale behind these numbers is that, across large population groups, there is an increased prevalence of certain diseases in people with a BMI over 25, and a much greater risk of disease and death in those with a BMI over 30 \[[@B4]\]. Being overweight or obese substantially raises the risk of developing hypertension, coronary heart disease, type 2 diabetes, stroke, gallbladder disease, sleep apnea and other respiratory problems, prostate and colon cancers \[[@B4],[@B23]\]. Yet, the evidence to date linking it to low back pain is not as clear cut as it is with the previously stated pathologies. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Clinically relevant differentiation between obesity and overweight ::: Overweight Obesity ----------------------------- ----------------------------- BMI of 25.0 to 29.9 kg/m^2^ BMI greater than 30 kg/m^2^ ::: BMI calculation without benefit of BMI charts --------------------------------------------- Body Mass Index (BMI) charts and hand held scales are available for individual clinician use. It is, however, unknown to what degree chiropractors or osteopaths use such tools. The following section is designed to aid the clinician with calculating BMI without benefit of chart or hand held scales. As noted earlier, BMI is calculated as weight in kilograms divided by height in square metres \[[@B20],[@B24]\]. This method is often too difficult to calculate for most people. A simpler method for those using the imperial system of measures is to take body weight in pounds × 703/height in inches squared. For example, a person weighing 150 pounds at 6 foot tall would correspond to a BMI of 20.3. TABLE [4](#T4){ref-type="table"} ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Calculation of BMI ::: ----------------------------------------------------------- 150 × 703 = 105450 divided by 72 inches (6 foot) squared. 105450 divided by 5184 (72 × 72) = 20.3 BMI. ----------------------------------------------------------- ::: Additional research findings ---------------------------- Leboeuf-Yde concluded from a review of the literature that due to lack of evidence, body weight should be considered a possible weak risk indicator and suggested that there is insufficient data to assess if it is a true cause of LBP \[[@B25]\]. Kostova found that in men over 40, overweight, obesity and number of pack years of smoking, estimated by duration of smoking and daily cigarette consumption (more than 20 years and more than 20 cigarettes per day), increased the risk of developing back disorders \[[@B12]\]. Despite these two studies, Garzillo et al and Leboeuf-Yde et al have given conflicting opinions \[[@B26],[@B27]\]. Garzillo\'s review of the data revealed a possible association between obesity and low back pain only in the upper quintile of obesity, and no evidence of a temporal relationship between weight change and changes in low back pain \[[@B26]\]. Leboeuf-Yde concluded from a twin study that obesity is modestly positively associated with low back pain, in particular with chronic or recurrent low back pain \[[@B27]\]. What appears to be a main concern in linking obesity as a causal factor for low back pain is the numerous variables encountered in these subjects. For example, it is hypothesized that overweight adult females may have negative self-concepts and body images compounded by chronic low back pain and obesity, these may be confounding factors \[[@B28]\]. Other variables such as less activity and/or muscular weakness leading to obesity are also possible considerations. Obesity and low back pain-related conditions -------------------------------------------- Not only is there controversy in obesity and low back pain, but there exists conflicting views of obesity and low back pain-related conditions such as spondylosis, decreased physical activities and discal herniation. The studies demonstrating a positive association are many. O\'Neil et alnoted that increasing BMI is associated with more frequent findings of osteophytes (bone spurs) at both the thoracic and lumbar spines \[[@B29]\]. The correlation of osteophytes and increased BMI is highest at the thoracic level \[[@B29]\]. Biering-Strenson et alnoted absolute weight and BMI are significantly higher in persons 60 years of age with spondylosis \[[@B30]\]. Both men and women with BMI of 30 kg/m2 or higher were twice as likely to have difficulties in performing a range of basic daily physical activities \[[@B30]\]. Compared with women with BMI lower than 25 kg/m2, those with BMI of 30 kg/m2 or higher were 1.5 times more likely to have symptoms of intervertebral disk herniation \[[@B31]\]. Conversely, Luoma et alconcluded that disc degeneration is not related to body height, overweight, smoking, or the frequency of physical activity \[[@B32]\]. In addition, studies by Riihimaki, Symmons, and Kang have shown no association between BMI and low back related problems \[[@B33]-[@B35]\]. Confounding the data is that the mechanism by which excess body weight causes osteoarthritis is poorly understood \[[@B9]\]. It is believed that contributions from both local increased force across the joint and systemic factors play a role \[[@B9]\]. A discriminating factor between fit and unfit patients with back pain may be the fact that fit persons more frequently are still employed, and as such may be involved more in physical activity \[[@B36]\]. Table [5](#T5){ref-type="table"} indicates where the research currently exists for the link between low back pain and obesity along with obesity and osteoarthritis. ::: {#T5 .table-wrap} Table 5 ::: {.caption} ###### BMI-related risk of osteoarthritis and low back pain ::: If your BMI is then your risk based solely on BMI ---------------- ------------------------------------ \<25 minimal 25 -- \<27 minimal 27 -- \<30 minimal 30 -- \<35 moderate 35 -- \<40 moderate \>40 moderate to high ::: We conclude, based on the available evidence to date, that those individuals with a BMI of under 30 are at a minimal risk of developing low back pain while those persons whose BMI increases to over 30 are a moderate risk of developing low back pain. We also suggest, based on the findings of the Melissas study \[[@B14]\] of those patients who relieved their low back pain symptoms after obesity surgery, that patients with a BMI of greater than 40 are at a high risk of developing low back pain. Albeit controversial, Table [5](#T5){ref-type="table"} may lead to a further refinement of risk of osteoarthritis and low back pain based solely on BMI. Limitations of obesity as a risk factor for low back pain --------------------------------------------------------- A significant difficulty in ascertaining cause and effect between obesity and low back pain is undoubtedly the term \"low back pain\" itself. Low back pain is a symptom not a diagnosis. A specific diagnosis, instead of the generalized form of \"low back pain\" may help separate out the association between LBP and obesity. The Agency for Health Care Policy and Research (AHCPR) in their 1995 Acute Low Back Problems in Adultsnoted common diagnoses used to explain back problems \[37\] (Table [6](#T6){ref-type="table"}). Given these possible diagnoses one can readily appreciate the dilemma in attempting to link obesity with its specificity in measurement to a broad symptom such as low back pain. ::: {#T6 .table-wrap} Table 6 ::: {.caption} ###### Common diagnoses used to explain back symptoms ::: ----------------------------------------------------------------------------- ---------------------------------------------------------------- --------------------------- Annular tear Adult spondylolysis Myofascitis Fibromyalgia Disc syndrome Strain Spondylosis Lumbar disc disease Facet syndrome Degenerative joint disease Sprain Spinal OA Disc derangement/disruption Dislocation \Other potential causes of low back pain symptomology Failed Back Surgery Syndrome\ Osteoporosis\* Urinary tract infection\* Compression fracture\* Somato-visceral mimicry syndrome\* Organic pathology (tumor, rheumatoid, endometriosis, arthritic disorders)\* Leg length inequity\* Sacro-iliac dysfunction\* Hip disorder\* \\Disagreement in research as cause of low back symptomology Morbid obesity?\\ ----------------------------------------------------------------------------- ---------------------------------------------------------------- --------------------------- OA = osteoarthritis ::: Another problem is the hypothesis that a person who suffers with continuing bouts of low back pain may be predisposed, due to inactivity or inability to exercise, to gain weight thus increasing their BMI. This hypothesis to our knowledge, has yet to be fully discussed and investigated. Conclusion ========== The data for a link between obesity and low back pain appears to be controversial. Yet, this does not adequately address the appropriate therapeutic approach to the obese patient with low back pain. The studies chosen for this review fail to document a definitive causal link between obesity and low back pain. Further research and epidemiologic data is needed to continue the search for a definitive answer. Competing interests =================== The author(s) declare that they have no competing interests.	PubMed Central	2024-06-05T03:55:57.589450	2005-4-11	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1151650/", "journal": "Chiropr Osteopat. 2005 Apr 11; 13:2", "authors": [ { "first": "Timothy A", "last": "Mirtz" }, { "first": "Leon", "last": "Greene" } ] }
PMC1151651	Introduction ============ Chiropractic and Osteopathy are complementary health professions that enjoy Government imprimatur to the extent that they have Registration Boards in every jurisdiction and a National Uniform Code of Conduct \[[@B1]\]. Also, third party payers such as private health funds, workers compensation authorities and the Department of Veterans Affairs recognise both professions and fund treatment provided by approved chiropractors or osteopaths \[[@B2],[@B3]\]. The training of chiropractors is by way of degree courses at Macquarie University (New South Wales \[NSW\]), Royal Melbourne Institute of Technology-RMIT University (Victoria) and a new program at Murdoch University in Perth (Western Australia). For osteopaths there are undergraduate courses at Victoria University of Technology, RMIT University (Victoria) and the University of Western Sydney (NSW) \[[@B4]\]. Importantly, there is no undergraduate program for these professions in Queensland. After graduation the learning process usually continues by way of an experiential process and also (but not always) from continuing education. The provision of continuing education and other vocational services to chiropractors and osteopaths in Queensland between 1996 and 2002 is the subject of this paper. To maintain Registration, continuing education for chiropractors and osteopaths is not compulsory in Australia as it is the USA. However, it is compulsory to maintain provider status with the Department of Veterans Affairs (DVA) \[[@B2]\]. As DVA income represents only a relatively small percentage of revenue for chiropractors and osteopaths there is no real financial or statutory imperative to participate in continuing education. Before arguing for continuing education one must ask a preliminary question regarding its validity. To do this a MEDLINE literature search was conducted from 1980 until 2002, using the key indexing words of \"continuing medical education\", \"chiropractic\", \"osteopathy\", \"educational intervention\", \"clinical audits\", \"performance indicator\", \"competency\", \"patient outcome\" and \"validity\". A search was also conducted of the literature indexing system for chiropractors and osteopaths known as MANTIS \[[@B5]\]. From this search a key paper by Werth \[[@B6]\] identified a paucity of information on the subject of continuing education concerning chiropractors and osteopaths; this paper reviews and relies on the medical and para-medical literature to review continuing education from a number of different perspectives including continuing education definition, needs assessment, evaluation and compulsory participation. Two significant issues arising from this paper are practitioner performance and health care outcomes after the administration of continuing education. Davis et al assessed these issues in a review of 50 randomised controlled trials \[[@B7]\]. The conclusion reached by the authors was there is evidence for changes in practitioner performance from continuing education but very little for improved patient outcome. A later paper by the same authors further reviewed studies that met the following criteria: randomized controlled trials of education strategies or interventions that objectively assessed physician performance and/or health care outcomes \[[@B8]\]. These intervention strategies included (alone and in combination) educational materials, formal continuing medical education (CME) activities, outreach visits such as academic detailing, opinion leaders, patient-mediated strategies, audit with feedback, and reminders. They found 99 trials, containing 160 interventions. Almost two thirds of the interventions (101 of 160) displayed an improvement in at least one major outcome measure: 70% demonstrated a change in physician performance, and 48% of interventions aimed at health care outcomes produced a positive change. Effective change strategies included reminders, patient-mediated interventions, outreach visits, opinion leaders, and multifaceted activities. Audit with feedback and educational materials were less effective, and formal CME conferences or activities, without enabling or practice-reinforcing strategies, had relatively little impact. Langworthy, in the only published study of clinical audit in chiropractic, concluding that a voluntary national audit scheme succeeded in raising awareness and standards of clinical practice \[[@B9]\]. Mugford et al.\'s review of 36 studies of the use of statistical information from audit or practice reviews suggest that it is most likely to affect practice when the recipients have already agreed to review that practice \[[@B10]\]. Cantillon and Jones\'s review of CME in general practice found 18 evaluations of audits with educational interventions, of which 17 showed a positive influence on doctor behaviour \[[@B11]\]. A Cochrane review has concluded that audit and feedback may be effective in improving the practice of healthcare professionals, especially prescribing \[[@B12]\]. Therefore, it appears that evidence for continuing education in achieving a positive change in practitioner performance and health care outcomes is mixed with some evidence for specific styles of continuing education. On balance, it is plausible to argue that quality continuing education has a general beneficial effect. The advent of evidence-based health care has increased the demand on health care providers of all persuasions to base their decisions and actions on the best possible evidence. The ability to receive or track down, critically appraise (for its validity and utility), and incorporate a rapidly growing body of evidence into one\'s clinical practice has been the \"mantra\" of the past decade \[[@B13]\]. Within the State of Queensland prior to 1996 practising chiropractors and osteopaths had some opportunities to participate in scientific forums specifically designed to improve their information gathering, clinical, scientific appraisal and bio-ethical skills. Dr. Keith Charlton principally developed this with the advent of the Brisbane Spinal Studies Group. However, when Dr. Charlton left for the United Kingdom there was an apparent lull. Professional associations and their subsidiaries provided much of the continuing education in Queensland prior to 1996. The major associations then were the Chiropractors Association of Australia \[[@B14]\] (CAA) (and its antecedents) and the Australian Osteopathic Association \[[@B15]\] (AOA). It can be argued that these two National organisations are in many ways the equivalent of trade unions. They aim to represent the professions in every respect and to the best of their ability. In the three decades prior to 1996, important political issues such as Government Registration and third party payer acceptance privileges consumed much of these associations\' time. Skilled chiropractors and osteopaths who gained political experience \"on the job\" have generally led both associations. Apart from their political agenda both associations have historically provided annual and occasional conferences, educational seminars and also respective journals newsletters for their members. These have assisted knowledge advancement. Nevertheless, it was observed by the Chiropractic & Osteopathic College of Australasia (COCA) \[[@B16]\] Executive of the day that there was an educational and vocational hiatus for both professions. It was thought that this was because the associations were not providing enough \"best practice\" continuing education. COCA determined that if chiropractors and osteopaths were to progress in the information and evidence-based age they would need access to high level material prepared and delivered by the best mentors available. In the mid 1990\'s it appeared that the associations (CAA and AOA) were strongly interested in the political agenda of the day and accordingly there appeared to be room for another group or professional body to provide these educational and other vocational services to chiropractors and osteopaths in Queensland. The identification of this need led to the expansion of the COCA from a predominantly Victorian based organisation into the State of Queensland and later nationally. The objective of this expansion was to develop, provide and foster quality vocational and educational services for chiropractors and osteopaths in Queensland and other States, with the ultimate goal of improving the public health. Methods ======= The national Executive of COCA reviewed a draft plan to attain these objectives. This plan involved was formulated by the author and involved: 1. Commencing continuing education in Townsville (a regional city in North Queensland) and if successful expand to Brisbane the State\'s capital ------------------------------------------------------------------------------------------------------------------------------------------------ This required some degree of \"faith\" on behalf of the COCA Executive, as the proposal was to run at a financial deficit (loss leader) for 2 to 3 years in both locations to encourage participation and COCA membership. It was a contention that once seminar attendances and membership had grown to a critical level, more realistic fees could be charged. 2. Organising conferences and offering the best available speakers from a broad variety of health provider areas ---------------------------------------------------------------------------------------------------------------- This decision was based on the notion that there was a wide-spread intellectual isolation of chiropractors and osteopaths in practice. In addition there were few opportunities for chiropractors and osteopaths to practice in multi-disciplinary environments. There was also a widespread feeling within the Executive of the day that University under-graduate education did not expose students to any training in hospitals where they would be likely to interact with a wide group of health professionals and also see very unwell patients. Accordingly, it was identified that speakers from other health fields would provide this initial interaction missing from the chiropractors and osteopaths professional training and life. 3. Introducing speakers who were chiropractors and osteopaths undertaking research at a high level and expose participants to this research ------------------------------------------------------------------------------------------------------------------------------------------- The modus operandi here was not to expose registrants to just research results but also the rigours of the methodology involved in such research. It was a goal that this exposure would give participants a better understanding of the research process. 4. Introducing the concepts of Evidence-based practice ------------------------------------------------------ Patient care is often outmoded because health providers of all persuasions lack awareness about important advances in their particular discipline and or general medical knowledge. One method of keeping abreast of current knowledge is reading journals. It is a popular method of staying informed, and is particularly suited to Queensland because of the vast size of the State and its decentralised structure. However, it is one thing to read a journal article, it is another to critically review it. It was therefore considered important to continue the work of the Brisbane Spinal Studies Group by enhancing literature review skills for chiropractors and osteopaths in Queensland. Using this strategy COCA aimed at enhancing the efficiency and effectiveness of journal reading and provide sufficient skills to assess the relevance, validity, and clinical application of new knowledge. 5. Building COCA membership --------------------------- All chiropractors and osteopaths in Queensland were offered membership of COCA. COCA reasoned that in order to build membership it was necessary to develop \"ownership\" in the concepts of a vocational College. The Executive took the view that it should model COCA along the lines of the Royal Australian College of General Practitioners (RACGP) \[[@B17]\]. A College of this nature has many objectives other than providing continuing education. It was therefore important for COCA to define a Mission Statement and Objectives over time. Many other benefits potentially accrue to members and the public if a College of this nature flourishes, particularly if such an organisation is built on a foundation of knowledge advancement, science, ethics, and the public health. Another benefit of building membership in Queensland was to recruit other practitioners to assist with COCA\'s activities. 6. Assessment of outcomes between 1996 and 2002 ----------------------------------------------- COCA assessed its continuing education in several ways. Initially, by surveying registrants after seminars and conferences, and then using this data and other pertinent educational material to develop guidelines for continuing education. As the organization expanded examining growth in COCA membership also became an important outcome measure. Where appropriate descriptive statistics of these measures were generated. Another outcome measure was whether COCA could establish itself as a stakeholder representing both professions on issues where there was a synergy with its stated objectives. This meant lobbying Government and other bodies about such issues. Results ======= 1. Commencing in Townsville --------------------------- A financial commitment was given by COCA in 1996 to commence operations in Queensland. These operations were based in Townsville and handled by Past President of COCA, Dr. Bruce Walker. 2,3. Conferences and speakers ----------------------------- The first conference was held in Townsville in June 1996. This conference set the scene for COCA in Townsville; it was multi-disciplinary with two medical specialists, one General Practitioner, 7 chiropractors and one chiropractor/histo-pathologist. Later that year another multi-disciplinary seminar was held, this one being held in conjunction with the James Cook University Chiropractic Research Fund. In 1997 the main conference was \"An overview of orthopaedic surgery\" delivered by Surgical Registrar and chiropractor Dr. David de la Harpe. Dr de la Harpe later specialised in Spine Surgery and is currently COCA Patron. A \"meet and greet\" for new members was held for new members in October 1997. On this informal occasion the guest speaker was Dr. R Jackson, from the Tennessee Chiropractic Licensing Board who spoke about Chiropractic in the USA and in particular on compulsory continuing education. In 1998 buoyed by the success of the Townsville events, the first Brisbane Conference was approved by COCA and organised for March 1998. The seminar was strong on clinical science and in particular focussed on the reliability of clinical instruments of measurement. In the same month another Townsville seminar was held with a multi-disciplinary group of speakers. In July 1998 a follow up conference in Brisbane featuring Dr. de la Harpe was held and later that year in November another Brisbane conference was conducted. Thereafter there have been regular conferences and seminars in both Brisbane and Townsville. A one-off conference was also held on the Gold Coast. In 2000 COCA gained the assistance of Drs. Ken Lorme (KL) and Jo-Anne Maire (JM). These two chiropractors (both with post-graduate degrees) assisted by taking over the conduct of COCA conferences in Brisbane (KL) and preparing a clinical audit program for members (JM). There was a total of 25 conferences or seminars organised by COCA between June 1996 and June 2002. 4. Evidence-based health care ----------------------------- At the initial COCA conference in Townsville participants were introduced to the concept of evidence-based health care. This was achieved by posing a clinical question \"Does scoliosis predispose to back pain?\" Then 3 key papers were presented and critically reviewed (by the author) demonstrating how to derive an answer to such a clinical question. A workshop conducted at the July 1998 Brisbane Conference presented similar work and using a checklist participants reviewed a chiropractic paper on infantile colic. The conclusion reached (after critical review by the group) was that the journal authors\' conclusion was not supported by the study as published. It was for many their first systematic review of a journal article. COCA members are now regularly exposed to the concepts of evidence based health care. 5. Building COCA membership --------------------------- The membership of COCA in Queensland grew from 3 in June 1996 to 167 in 2002. This growth must be seen as an endorsement by practitioners on the ground in Queensland as there were only about 400 registered chiropractors and osteopaths domiciled and working in the State at the time. Queensland chiropractors and osteopaths often have geographical limitations placed on their continuing education (\"The tyranny of distance\"). As a consequence COCA felt it was important for these and other regional, rural and remote members to have access to both distance learning and a library. Distance learning is now an integral part of the benefits of COCA membership; further a special benefit for members is access to the RACGP library. Such access is crucial to Queensland members. Other tangible benefits for members include regular multi-disciplinary continuing education in a variety of formats including distance learning at an affordable price, access to professional indemnity insurance, a journal (Australasian Chiropractic and Osteopathy), a regular newsletter (COCA News) and belonging to an organisation that has knowledge advancement, science, ethics and public health as its main objectives \[[@B16]\]. 6. Other outcomes ----------------- The satisfaction of attendees at COCA conferences was measured using an optional \"Exit questionnaire\". In order to evaluate this variable a convenience sample of 181 was supplied by COCA secretariat. It appears from close review that some returns are missing. Also the response rate was not calculable because of a loss of COCA data at Secretariat level. Therefore the following results are only indicative for those returns processed and cannot be generalised to the entire population of attendees. The first 2 questions of the survey ask \"What were the best aspects of the seminar?\" and \"What were the worst aspects of the seminar?\" A review of the answers to these questions shows that for question 1, there was a mean of 2.4 out of a possible 3 (sum total = 431) and for question 2, the mean was 0.71 out of 3 (sum total 127). These results show a substantially greater number of \"best aspects\" than \"worst aspects\". The third question asks \"Was the seminar value for money?\", 175 (96.7%) answered \"yes\", 1 (0.6%) answered \"no\" and 5 were missing. The fourth question asked respondents to rate the seminar on a Likert scale out of 7. With 1 being \"very poor\" and 7 being \"very good\". The mean rating for all surveys from all conferences was 6.1 (Range 3 to 7), missing 3. Discussion ========== The development of a vocational body (COCA) for chiropractors and osteopaths in Queensland is now a reality. Registrants in the field have supported (\"with their feet\") the notion of a body that delivers continuing education at an affordable cost and also an organisation that concentrates on the vocational aspects of their practice with a scientific and ethical focus. COCA\'s objectives are set out and are on the public record \[[@B16]\] and all applicants for membership sign their acceptance of the COCA Code of Ethics \[[@B16]\]. COCA has dedicated materials and resources used in its activities, to serve Queensland members and non-members alike. This has included equipment, staff, volunteers, facilities and financial resources. Without these inputs achievement of the College\'s objectives would have been futile. The activities and processes underpinning the programmes delivered to the practitioners were designed to meet their needs and to potentially improve the public health. This has been attempted through teaching, distance learning, information exchange, seeking benefits for members such as indemnity insurance and also lobbying Government. COCA\'s educational and vocational outputs in Queensland can be measured by the number of seminars and conferences, the number of additional programs such as distance learning modules, the RACGP library access, the number of Journals and Newsletters supplied to members. This has been quite substantial for a relatively small organisation. Outcomes can also be measured by the number of practitioners who have become members, the satisfaction ratings of attendees at seminars and conferences, the attraction of senior practitioners from within the State to become involved at organisational level. Like all continuing education providers COCA aims (through its educational programs) to improve the participants knowledge and skills and although these are often considered to be rather short-term outcomes, they may also lead to positive changes in behaviours and then hopefully changes in values, conditions and improvements in the public health. Further research should concentrate on the objective measurement of these outcomes and in particular patient health outcomes. There have been limitations to COCA development in Queensland. Limited resources have prevented any attempt at objective measurement of practice outcomes from COCA\'s continuing education programs. COCA\'s first 6-years in Queensland have been more about development of and growing the organisation within the State. Nevertheless, the programs have been well received by surveyed participants and it should be noted that practical change by practitioners is unlikely to occur in an unhappy continuing education recipient. Another major limitation has been the limited number of chiropractic or osteopathic academics on the ground in Queensland; it is postulated that this may be due in part to the lack of an under-graduate program in the State. Conclusion ========== The chiropractors and osteopaths in Queensland are now better serviced in the area of continuing education, vocational programs and professional assistance. This has been achieved by the development and expansion of the Chiropractic & Osteopathic College of Australasia within the State. It is hoped that COCA\'s continuing emphasis on information transaction skills, multi-disciplinary fora, science and ethics will have a positive impact on public health. To continue to be successful COCA must constantly review its modus operandi in the State of Queensland. Acknowledgements ================ The author would like to thank Dr. Peter Werth, COCA National President for his review of this paper and also the suggestions of the two referees.	PubMed Central	2024-06-05T03:55:57.591277	2005-4-11	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1151651/", "journal": "Chiropr Osteopat. 2005 Apr 11; 13:3", "authors": [ { "first": "Bruce F", "last": "Walker" } ] }
PMC1151652	Introduction ============ Hamstring injuries are the most prevalent injury in Australian Rules football \[[@B1],[@B2]\]. This may be possibly due to the unique physical demands of the game requiring rapid acceleration, endurance and agility running, kicking and bending to pick up the ball. Hamstring injuries are not confined strictly to Australian Rules football but are also seen in soccer \[[@B3]\], athletics \[[@B4]\], hurling \[[@B5]\], cricket \[[@B6]\] and touch football \[[@B7]\]. This makes hamstring injuries the most prevalent muscle injury in sports consisting of rapid acceleration and maximum speed running. Such injuries can and do result in significant financial consequences to players and clubs alike. It is agreed that hamstring injuries have a complicated multi-factorial etiology, including muscle weakness and balance, lack of warm up, decreased flexibility, previous injury history and fatigue \[[@B8]\]. The only conclusive risk factors for future injury is a current hamstring injury or a previous history of hamstring injury \[[@B1],[@B9]\]. This makes prevention of the initial injury a primary focus in management efforts. The purpose of this paper is to present two cases of hamstring injury that were effectively managed with spinal manipulative therapy (SMT) and correction of lumbar-pelvic biomechanics. Prevention of re-injury may have been due to ongoing maintenance type care. Back related hamstring injury ----------------------------- Some authors have listed a separate category of hamstring injury known as a \'back related hamstring injury\' which is classified as having both local hamstring signs and positive lumbar signs \[[@B9],[@B10]\]. It is known that referred myotomal pain from lumbar-pelvic structures, the sciatic nerve and the gluteal or piriformis muscles can mimic hamstring strains \[[@B9]\]. The world\'s longest serving injury surveillance, performed by the elite Australian Football League (AFL) uses an umbrella term for hamstring injury which fails to differentiate the potential diagnoses. This means the true prevalence of back-related hamstring injuries in Australian Rules footballers is unknown. Using MRI to confirm the diagnosis of hamstring injury, 19% are without muscle damage \[[@B3]\], suggesting no local muscle pathology and injury to be related to altered functional biomechanics or pain referral that does not appear on cross sectional imaging. This type of injury would logically require different forms of treatment than simple muscular-tendon injuries. It has been postulated that hamstring injuries may have a biomechanical basis and therefore it is reasonable to suggest that assessment of hamstring injury should include a biomechanical evaluation, especially that of the lumbar spine, pelvis and sacrum \[[@B3]\]. There is a paucity of literature about the role of aberrant lumbar-pelvic biomechanics as an etiological factor predisposing to hamstring injury. It is tempting to speculate that this may explain why hamstring injuries have the highest recurrence rate of any injury in the AFL. Thirty three per cent of injured players are likely to re-injure their hamstring on return to competition and miss subsequent matches \[[@B1]\]. A significant risk of injury recurrence exists in the first few weeks following return to play, with the cumulative risk of recurrence for the remainder of the season being 30.6% \[[@B11]\]. No significant change in recurrence rates has been noted over the last 7 years, while players are missing more time on average due to injury \[[@B1],[@B12]\]. In contrast, other injuries in the AFL have noted a considerable improvement in decreased rates of recurrence over this time frame \[[@B12]\]. This suggests that players are being managed more conservatively with regards to return to competition from hamstring injuries and there appears to be no change in the treatment protocol if recurrence rates have yet to decline. This may suggest the possibility of a biomechanical factor that may require a differing approach that has yet to be introduced. No prevention effort will be successful without understanding the etiological factors predisposing hamstring injury and efforts to decrease recurrence rates for hamstring injuries will be unsuccessful if the possibility of a biomechanical factor is excluded in the etiology. Case Report 1 ------------- A 19-year-old male, semi-elite Australian Rules footballer presented with left sided hamstring pain that occurred during a game two weeks prior. The patient had not played a game or been able to train for two weeks since the injury. He had been treated with cryotherapy, NSAID\'s, slump stretching, lumbar spine mobilizations, ultrasound and massage to the hamstrings. He had a history of mild osteitis pubis 12 months previously that was treated with rest and modified activity. There had been no prior history of hamstring or low back injury. On physical examination the patient was standing with an apparent lumbar spine hyperlordosis, anterior pelvis tilt, flexed knees and increased thoracic kyphosis. There was tight (reduced range of motion) bilateral hip flexors (modified Thomas position\* +15°) and hamstring muscles (45° straight leg raise \[SLR\]), hypertonicity of the gluteii, hamstring, lumbar and psoas muscles and general mid thoracic and lumbar spine motion restriction, determined by inter-segmental motion palpation and observation of range of motion (ROM). (\Modified Thomas testing requires the patient to sit at the edge of the table and to bring one knee to their chest to firmly flatten their back. They then assume the supine position, allowing the testing leg to extend off the table. An angle is formed between the femur and a line drawn parallel to the tabletop. A positive angle means the femur is projecting upwards. A negative angle means the femur hangs downwards)*. There was weakness of the left hamstring and gluteus maximus graded 4/5. Hamstring tenderness could not be localized on palpation. Other physical examination findings, including Trendelenburg, valsalva, neurological, slump, extension leg raise and hip and sacroiliac joint motion palpation and orthopedic testing were unremarkable. The patient was given a working diagnosis of back-related hamstring injury as a result of lumbar-pelvic myofascial pain referral, mimicking a grade one hamstring strain. Differential diagnoses included pain referral from gluteal trigger points. Treatment involved long-lever SMT to the lumbar spine, short-lever SMT to the mid thoracic spine, drop piece knee manipulation, active release soft tissue massage techniques (ART) to the psoas, gluteal, lumbar and hamstring muscles and proprioceptive neuromuscular facilitation (PNF) stretching of the hamstring and psoas muscles. Post treatment, modified Thomas position bilaterally was +5°, SLR 60° bilaterally and muscle strength was graded 5/5. The patient received 3 treatment sessions that week and played a match the next week without re-injury. He was put on a maintenance program for the rest of the 12 weeks of the season including finals (one visit per week for a month, one visit per fortnight thereafter) which included the above treatment and strengthening and muscle activation work (to improve hip extension and abduction motor patterns) to the gluteus maximus and medius, multifidus, transversus abdominus and internal oblique muscles. Maximum medical improvement (MMI) was reached after 7 treatments. The patient finished the season without re-injury. Posture and muscle length changes continued to improve over this period (bilateral modified Thomas position -5°, SLR 85°). Case Report 2 ------------- A 25 year old male, semi-elite Australian Rules footballer felt a \'twinge\' in his right hamstring during a game. He presented to us the day after, complaining of tightness in his medial right hamstring and a stiff low back. He had no previous history of hamstring injury but had suffered episodic low back pain over a 5-year period. On physical examination, the right pelvis was low compared to the left in standing position; there were tight right (45° SLR) and left (55° SLR) hamstrings and tight left hip flexors (+10° modified Thomas position). There was palpable hypertonicity through the right hamstring, left psoas, lumbar and gluteal muscles, thoracolumbar spine and right sacro-iliac joint (SIJ) motion restriction, positive Gillett\'s (standing S-I joint motion palpation) and extension leg raise testing for the right SIJ and weakness of the right hamstring and gluteus maximus muscles rated 4/5. Hamstring tenderness could not be localized on palpation but mild discomfort was reproduced in resisted muscle testing. Other physical examination findings and testing procedures, including Trendelenburg, valsalva, neurological, slump, lumbar ROM and hip joint motion palpation and orthopedic testing were unremarkable. The patient was diagnosed with a back-related hamstring injury on the basis of his apparent right SIJ motion restriction and pain referral. Differential diagnoses included pain referral from lumbar-pelvic myofascial structures, gluteal trigger points or a grade 1 hamstring injury with concurrent lumbar-pelvic dysfunction. Treatment involved high velocity low amplitude (HVLA) SMT to the right SIJ and thoracolumbar spine, long axis manipulation to the right hip joint, ART to the right hamstring, left psoas, lumbar and gluteal muscles, PNF stretching of the right hamstring and left psoas and hamstring cryotherapy. Post treatment, modified Thomas position was 0° on the left, SLR 55° on the right and 65° on the left. Muscle strength was graded 5/5. The patient did not participate in training during the week and received 2 more treatment sessions. He played a match the next weekend without re-injury. He was seen twice the next week and put on a maintenance program for the 16 weeks remaining in the season (one visit per week for a month, one visit per fortnight for a month, one visit per month thereafter). This included the above treatment plus strengthening and muscle activation work (to improve hip extension motor patterns and running technique) to the gluteus maximus, multifidus, transversus abdominus and internal oblique muscles and home advice including flexibility and stability work. MMI was reached after 10 treatments. No re-injury occurred during this period and muscle length changes continued to improve (bilateral modified Thomas position 0°, SLR 75°). Discussion ========== In the sports medicine literature, spinal manipulation for the treatment or prevention of hamstring injury has not been documented, despite it being frequently used by chiropractors and other manual therapists. In fact, there is a lack of literature on the management of hamstring injuries in general. A recent review of the literature suggested that low back pain from the zygopophyseal joints at the levels of spinal nerve roots supplying the hamstrings may provoke local muscular responses such as increased muscle tension which may predispose injury \[[@B8]\]. However, this potential association with injury is yet to be scientifically validated. The only treatment methods that have been documented in randomized controlled trials are slump stretching \[[@B13],[@B14]\] and rehabilitation protocols \[[@B15]\]. Slump stretching involves maximal cervical, thoracic and lumbar flexion with full hip flexion, knee extension and ankle dorsiflexion with passive practitioner overpressure. These studies have had low subject numbers, making conclusions weak. The slump test is said to significantly differentiate referred posterior thigh pain from that due to muscular-tendon strain \[[@B13]\] and has been able to identify those with recurrent hamstring strains in a small study \[[@B14]\]. Slump stretching as a treatment procedure (when slump testing is positive) has been shown to be more beneficial in returning athletes to competition than standard physiotherapy treatment alone (ultrasound, massage, progressive flexibility and strengthening) \[[@B13]\]. The slump test has been proposed to be a measure of \'neural tension\' which is postulated to predispose hamstring injury \[[@B14]\]. However, the anatomical relationship of the hamstrings with the thoracolumbar fascia (TLF) system has been neglected. The tendon of bicep femoris is continuous with the sacrotuberous ligament, passing across the sacrum and attaching to the thoracolumbar TLF \[[@B16]\]. This functionally connects the hamstrings to the lumbar spine, upper torso, shoulder and occiput and casts doubt on reliability of the slump test as being able to measure neural tension \[[@B17]\]. Contracture of the muscular attachments of the TLF has been documented to cause TLF its displacement \[[@B16]\]. Therefore neural tension may only be an assumption and it may more likely be myo-fascial tension, or possibly a combination of the two giving a positive slump test. Postural changes such as forward weight bearing, as occurs during forward lean gait, will also cause hamstring tension and predispose hamstring injury. This suggests that the TLF system should be assessed during treatment of hamstring injuries. Australian Rules footballers with a previous back injury have been found to have a significant increased risk of hamstring injury \[[@B9]\]. A strong relationship between age and hamstring, calf and Achilles injuries (with a L5 and S1 nerve supply) also exists in AFL players \[[@B18]\]. The L4/5 and L5/S1 levels are the most common areas for spinal degeneration and athletes are susceptible to degenerative changes at an earlier age than the normal population \[[@B19]\]. Altered neural input from the levels that innervate hamstrings may be causing and prolonging hamstring injuries. Long term prospective studies are required to further investigate this finding. Significant excessive lumbar lordosis has been found retrospectively in athletes with previous hamstring injury when compared to a group with no injury history \[[@B20]\]. Prospectively, thigh injuries as a group (hamstring, quadriceps and adductor injuries) have been linked to postural defects, including increased lordosis, sway back and knee interspace measurements \[[@B21]\], while the incidence of muscle injuries in general has been linked to the existence of defective body mechanics associated with the site of injury \[[@B22]\]. This indirectly suggests that improving lumbar-pelvic biomechanics may play a role in treatment and prevention of hamstring injury. Of the other risk factors linked with hamstring injuries, low hamstring strength is a risk factor with some degree of clinical evidence \[[@B23]\]. Strength deficits have been found to exist in athletes with a history of recurrent hamstring strain \[[@B24]\]. This may have been the cause of the initial injury, be due to weakness from ineffective rehabilitation or from dysfunction in the lumbar spine, SIJ or pelvis. An association between altered pelvic function and hamstring injury is suggested by a past history of groin injury and osteitis pubis being significant risk factors for hamstring injury \[[@B9]\]. Although it is only an assumption that pelvic problems contribute to groin injuries through its kinematic chain relationship. One small randomized clinical trial has looked at the effectiveness of manipulation targeted at the SIJ for the treatment of hamstring injuries \[[@B25]\]. The manipulation group improved hamstring strength compared to the control group, suggesting SIJ dysfunction may be related to initial hamstring injury. We believe that the two cases we saw and treated were related to a lumbar-pelvic biomechanical aspect. In our two cases, there existed clinically either lumbar hyperlordosis, anterior pelvis tilt or lateral pelvis tilt. This is consistent with the findings of Hennessy and Watson (1993) and Watson (1995, 2001). Improvement of these biomechanical factors, including the use of SMT, resulted in successful treatment and prevention of the hamstring injuries. This leads us to hypothesize that inter-segmental and/or global lumbar-pelvic biomechanical dysfunction produced either referred pain or hamstring muscle insufficiency via the TLF as a cause of the hamstring injuries and possibly why these cases did not improve with previously used standard treatment modalities. There are limitations to this hypothesis including the reliability (or lack thereof) of the diagnosis of mechanical dysfunction of the low back and pelvic areas. To conclude that there is mechanical dysfunction in the low back particularly in the absence of pain also needs further research. Conclusion ========== Hamstring injuries have a complex multi-factorial etiology. Two forms of hamstring injury have been identified with potentially different pathogenesis, notionally requiring different treatment methods. From our case reports and evidence presented, it appears that spinal manipulation and improving lumbar-pelvic biomechanics and function may play a role in treatment and prevention of hamstring injury. This should be further investigated in prospective, randomly controlled trials with long-term follow up. Given that a recurrence rate exists for hamstring injuries, the possibility that a concomitant biomechanical aspect exists should be pursued.	PubMed Central	2024-06-05T03:55:57.593479	2005-4-12	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1151652/", "journal": "Chiropr Osteopat. 2005 Apr 12; 13:4", "authors": [ { "first": "Wayne T", "last": "Hoskins" }, { "first": "Henry P", "last": "Pollard" } ] }
PMC1151653	Method ====== A search through Medline, Meditext, PubMed, OVID, Science direct, Austats, CINAHL, Expanded Academic ASAP was performed using the key words: Biopsychosocial model, hypothyroidism, treatment, levothyroxine, thyroid. What is the Biopsychosocial model? ================================== The biopsychosocial model depicts a health care concept that has evolved in close association with current pain theory. It has sought coexistence with the dominant biomedical model of health care, which describes \'disease\' as a failure of or within the soma, resulting from infection, injury or inheritance \[[@B1]\]. The biomedical model has its roots in the Cartesian division between mind and body \[[@B2]\]. In 1977, Engel described a crisis that modern medicine and psychiatry were facing. Disease, from a biomedical perspective was described in somatic parameters alone, there was little or no room for psychological, social and behavioural dimensions of illness within this model. This made adherence to this framework very difficult. There were somatic and mental disorders that simply did not fit the biomedical model, and hence it was no longer sufficient for the scientific and social responsibilities of either medicine or psychiatry \[[@B2],[@B3]\]. Engel set out to develop a new framework that would account for the biological, psychological and social dimensions of illness and disease. It was essential that this new model provide a basis for the understanding and treatment of disease, whilst taking into account the patient, his/her social context and the impact of illness on that individual from a societal perspective \[[@B4],[@B5]\]. This represented the development of the biopsychosocial model \[[@B2]\]. The biopsychosocial model states that ill health and disease are the result of an interaction between biological, psychological and social factors. The biopsychosocial model makes the distinction between pathophysiological processes that cause disease and the client\'s perception of their health and the effects on it, called the illness \[[@B6]\]. It seeks to build upon the biomedical model. Biological indices are still held in high regard, however, they represent only one of the defining factors for the diagnosis and management of disease under a biopsychosocial framework \[[@B2]\]. The biopsychosocial model describes psychological and social effects of disease risk, prevention, treatment compliance, morbidity, quality of life and survival \[[@B4]\]. Situations paradoxically arise in medicine in which a person who feels well is described biochemically, as having \'disease\'. In contrast, a client\'s laboratory findings may reveal no \'disease\', however the client still feels unwell. The biopsychosocial model provides a conceptual framework for dealing with such situations \[[@B2]\]. Of late, a great deal of attention has been given to the factors involved in chronic pain and depression. It is research like this that has highlighted the need for a paradigm such as the biopsychosocial model in the management of conditions other than chronic pain. Following the success of various psychological and cognitive interventions in the reduction of somatic signs and symptoms associated with certain conditions such as irritable bowel syndrome (IBS), non-cardiac chest pain, fibromyalgia and rheumatoid arthritis (RA), research has set out to explore the role of psychosocial factors in the disease process \[[@B3],[@B7],[@B8]\]. The biopsychosocial model avoids a strong \'disease\' focus and seeks to address the client and his or her illness \[[@B9]\]. Clients are helped not only with biological disruptions, but also with their capacity to deal with being ill. It is proposed that this approach may be of benefit in reducing the frequency of clinic visits, hospitalisations, laboratory investigations and use of pharmacological agents. Changes in ones ability to cope, inherent belief systems as well as behavioural and social processes associated with being ill may also be improved through the implementation of this model \[[@B10]\]. The Evolution of the Biopsychosocial Model ========================================== The biopsychosocial model represents the evolution of the biomedical model, and aligns favourably with recent ideas in pain theory and pathophysiology \[[@B3]\]. Early theories concerning pain were consistent with the specificity theory, which described pain as the result of noxious stimuli or somatic pathology alone \[[@B11]\]. There was little or no place for the influence of psychological or social factors within these original theories \[[@B3]\]. The mid nineteen sixties saw the emergence of the gate control theory of pain \[[@B11],[@B12]\]. The theory proposed by Melzack and Wall, represented a more concise model that took into account the multidimensional nature of pain, allowing for physiological factors and the role of the brain in the processing of nociceptive stimuli. The gate control theory depicts three dimensions of pain, a sensory-physiologic, motivational-affective and a cognitive evaluative dimension. Scope for the influence of environmental factors on pain also exists within the body of this theory. The gate control model also provides a basis for understanding the depression and cognitive and motivational shifts witnessed in chronic pain situations \[[@B3],[@B11]\]. Melzack through his study of phantom limb pain (pain that is localised in the region of a deafferented body part, subsequent to the loss of a limb)\[[@B13]\] developed the neuromatrix model \[[@B3],[@B14]\] of pain. This model builds upon the gate control model and represents one of the more recent ideas in pain theory. The neuromatrix theory proposes that the various dimensions of pain experience are the result of a neural network program called the neuromatrix. This neuromatrix is influenced by genetic factors in conjunction with cognitive, sensory and affective experiences, which are individual specific. This model states that the unified pain experience depicts an aggregate of information from somatosensory, limbic and thalamocortical pathways. The end product is the cyclic processing of neural information into a characteristic pattern, known as the neurosignature. Melzack postulates that a neurosignature exists for all types of pain, mood and psychological states \[[@B3],[@B14]\]. The neuromatrix is also described as having both static and dynamic qualities, meaning that the neurosignature can be influenced/modified through learning and experience \[[@B14]-[@B17]\]. Put into context, this model suggests that neurosignatures for pain and depression exist throughout the neuromatrix. These signatures cannot be erased, but can be altered if changes are made to the entire network \[[@B3],[@B11],[@B14]\]. The biopsychosocial model has emerged over the past two decades and has sought to expand upon disease paradigms and complement pain models. It states that in order to understand and manage ill health, pain and disease, one must take into account the influence of biological, psychological and social factors \[[@B7]\]. Biological Factors ================== It is acknowledged that there are many conditions (eg osteoarthritis and rheumatoid arthritis)\[[@B3]\] in which the symptoms experienced by clients are strongly linked to peripheral factors such as, inflammation or cartilage damage. These represent examples of biological disruption under a biopsychosocial framework. Previously, the influence of psychological and social factors on these conditions was considered of little importance and treatment was directed to areas of somatic pathology or nociceptive input \[[@B3]\]. Historically, the underlying disease process could account for many if not all of the biological features of disease and ill health. In this instance a biomedical treatment strategy could be implemented with high efficacy. It is important to note that the biomedical model has not been replaced. There are a number of diseases that can be diagnosed and managed without any consideration of psychosocial factors (eg. legionnaires disease and toxic shock syndrome) \[[@B18]\]. In contrast, the existence of pain and discomfort in conditions such as fibromyalgia, irritable bowel syndrome (IBS) and noncardiac chest pain is strongly linked to central nervous system (CNS) disturbance (eg. altered central processing) and psychosocial factors \[[@B3],[@B7],[@B8]\]. From a biopsychosocial perspective, biological factors represent one determinant of ill health and disease. Biological disruption may exhibit inconsistent weighting amongst various conditions, but it represents an essential component of diagnosis and management under the biopsychosocial model \[[@B2]\]. Psychological Factors ===================== Research into chronic pain and depression has revealed a number of significant psychological factors \[[@B19]\]. Clients experiencing pain lasting for prolonged periods of time often display a series of maladaptive coping responses that can influence the pain experience \[[@B3],[@B10]\]. These responses include catastrophizing, perceived low self-efficacy and perceived helplessness. Clients presenting with the above responses, often report higher pain levels than those subjects that do not display these responses \[[@B3],[@B19],[@B20]\]. Psychological intervention (eg pain-coping strategies, cognitive behavioural therapy {CBT}) aimed at these maladaptive patterns and behaviours has demonstrated high rates of success in reducing symptomatology and disease progression in certain chronic pain conditions \[[@B3],[@B10],[@B20]\]. One area pain researchers are currently exploring is the differences in information processing and recall bias between chronic pain sufferers who present with comorbid depression and those non-depressed pain groups. It appears that clients with chronic pain demonstrate a propensity for recalling the negative information regarding their condition, in the presence of co-morbid depression. Clients without depression demonstrate a recall bias for positive illness related information. These studies suggest that depressed patients with chronic pain may be processing information differently to the non-depressed groups \[[@B3]\]. Pincus and Morley propose a model to account for the bias in information processing (specifically cognitive bias) demonstrated in chronic pain sufferers \[[@B21]\]. It is postulated that bias is the result of intersection of three schemas representing pain, illness and self. Within these schemas is a store of information that can interact and influence the processing of information. The painschema encompass sensory intensity, along with spatial and temporal features of pain. The schema accounts for the immediate properties of the pain experience. They depict the interruption of ongoing behaviour and the commencement of pain avoidance and recuperative behaviours. The illnessschema depict information relating to affective and behavioural consequences of illness (eg. goal attainment, both long and short term and quality of life etc.). The identity, timeframe, perceived causes and consequences along with control of a particular illness, make up the schema. The schema representing selfis described in a number of ways. The self can be explained as \'an organized cognitive structure within long-term memory, which may incorporate both general trait like information about the self, as well as specific behaviours\' \[[@B21]\]. The self can be viewed as temporally dynamic processing and assimilating information throughout life \[[@B21]\]. It is stated that contemporary individuals are striving to meet their positive goals and avoid unwanted outcomes \[[@B21]\]. This is determined by components of the self-schema and their projections as to what they might become. Illness, pain and other significant life events have the potential to disrupt aspects of the self. For example repeated pain whilst performing a specific task can interfere with or result in failure to complete that task. Variations in the state of a person may be explained by the interaction of these schema. Activation of elements from one particular schema has the potential to simultaneously activate components of another schema, in a process known as enmeshment. Consequently, the activation of one schema via a relatively innocuous stimulus following enmeshment, can elicit unwanted effects (eg illness behaviour associated with pain syndromes) \[[@B21]\]. Recent studies performed by Buer and Linton describe the importance of cognitions (fear-avoidance beliefs and catastrophizing) in chronic pain situations \[[@B22]\]. The model used by Buer and Linton describe fear-avoidance specifically, as a fear of movement /(re) injury, which predispose to catastrophizing and avoidance, which in turn could lead to disuse, disability and depressive symptoms \[[@B22]\]. Social Factors ============== There exists a considerable amount of data on the social determinants of ill health and disease. One of the contributors to this body of knowledge has been the Whitehall studies. The Whitehall I study initiated in the late 1960\'s, examined mortality rates over 10 years among male British Civil Servants aged 20--64 \[[@B23]\]. The Whitehall study sought to expand on the issue of the social class grouping, and counter some of the problems associated with this topic. Participants were physically examined and asked to complete questionnaires regarding their jobs, smoking habits in conjunction with personal and family medical histories. Certain participants were asked about car ownership, physical activity at work and general leisure activity, which completed the list of socio-economic outcome measures. The results of the study depicted an inverse association between grade (level) of employment and mortality from CHD and a range of other causes was observed \[[@B23]\]. The Whitehall II study was set up to investigate the effect of social gradient on morbidity and mortality, including determinants such as, work characteristics and social support \[[@B24]\]. Measures taken into consideration were, grade of employment, depressive symptoms, physical functioning, psychosocial work characteristics, life events and material problems and health related behaviour. The results revealed that some risk factors contribute jointly to the inequalities witnessed in mental and physical health. The incidence of secondary psychological stress associated with physical ill health is more prevalent in lower employment grades. Work was deemed most important in inequalities in depressive symptoms amongst men. Amongst female respondents, work and material disadvantage were equally salient in explaining inequalities in depressive symptoms \[[@B24]\]. There are various social factors that have gained recognition in recent years with respect to influences on depression among pain patients. Patients who report high family conflict and low family cohesion often display higher depression levels. Likewise, patients with low socio-economic status and lower levels of education also exhibit higher depression scores. Research into this area highlights a number of studies in which psychological intervention involving spouses or caregivers resulted in reduced pain and psychological distress \[[@B3]\]. In a recent review, Truchon describes a set of socio-demographic factors that have featured in chronic disability (related to low back pain) studies over the last decade \[[@B25]\]. These include; age, sex, education, ethnic background and financial compensation. Ringel et aldescribes a diagnostic protocol for irritable bowel syndrome (IBS), encompassing social factors such as, break up of a close relationship, early life experience, familial dysfunctions and family environment. It is postulated that these factors may influence the development of symptoms, clinical expression, course of the disorder and the utilization of health services \[[@B7]\]. Hypothyroidism ============== Hypothyroidism refers to any metabolic state that results from a decrease in the amount of circulating thyroid hormones in the body. Hypothyroidism can be classified based on its time of onset (congenital or acquired), severity (overt {clinical} or mild {subclinical}), and the degree of endocrine aberration (primary or secondary) \[[@B26]\]. Primary hypothyroidism follows a dysfunction of the thyroid gland itself, whereas secondary hypothyroidism results from the dysfunction of metabolic or messenger pathways associated with thyroid hormone production and metabolism \[[@B27]-[@B29]\]. Primary hypothyroidism is characterised by reduced free thyroxine (FT4) levels and elevated thyroid stimulating hormone (TSH) levels. Diagnosis of secondary hypothyroidism presents as a clinical challenge as TSH levels can be reduced, normal or slightly elevated. Evaluation of other pituitary hormones becomes necessary in this situation \[[@B27]\]. According to the 1995 report from the Australian Bureau of Statistics (ABS) 5.3/1000 males and 27.3/1000 females experience thyroid dysfunction in Australia \[[@B30]\]. This figure is higher in the elderly, postmenopausal women and various groups presenting with psychological dysfunction \[[@B31]-[@B33]\]. In the United States hypothyroidism is the second most common endocrine disorder and it has been estimated that 18/1000 members of the general population display decreased thyroid hormone levels \[[@B33]\]. Hypothyroidism and subclinical hypothyroidism are considered more common than their counterpart, hyperthyroidism \[[@B33],[@B34]\]. Hypothyroidism is the most common pathological hormone deficiency \[[@B26]\]. Aetiology ========= Hypothyroidism is caused by a variety of different states and conditions. Autoimmune and iatrogenic causes constitute the most common sources of reduced thyroid hormone levels \[[@B26]\] (see table [1](#T1){ref-type="table"}). ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Aetiology of Hypothyroidism ::: Category Cause --------------- ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Autoimmune Hashimoto\'s Thyroiditis, Reidels Disease, previous Graves disease, de Quervains thyroiditis, postpartum thyroiditis, Downs syndrome, family history of autoimmune disease or associated disorders (vitiligo, adrenal insufficiency, diabetes mellitus type-1, Sjogren\'s, coeliac disease), Turner\'s syndrome, Multiple Sclerosis, primary pulmonary hypertension Nutritional Iodine deficiency, excess intake of goitrogens, excessive iodine intake Environmental Radiation Exposure, exposure to polybrominated and polychlorinated biphenyls and resorcinol Iatrogenic Radioactive iodine therapy, medical radiation exposure, total or subtotal thyroidectomy, drugs impairing thyroid function (amiodarone, thalidomide, betaroxine, lithium carbonate, stavudine, aminoglutethimide) Hypothalamic Hypothalamic or suprasellar mass, history of hypothalamic radiotherapy or surgery, disorders causing hypothalamic dysfunction (eg sarcoidosis, heamochromatosis) Pituitary Known pituitary tumour, other elements of hypopituitarism, manifestations of sellar mass, history of pituitary surgery or radiotherapy, history of head trauma Other Postpartum status ::: Signs and Symptoms of Hypothyroidism ==================================== Hypothyroidism manifests in a variety of different forms \[[@B26]\]. Young infants and children born with deficiencies in thyroid hormones are at risk of brain damage and mental retardation \[[@B35],[@B36]\]. Technological advancement in hormone assays has revealed that hypothyroidism is a relative, rather than an absolute state \[[@B37]\]. In fact, a spectrum of thyroid hormone deficiency exists. This spectrum, in conjunction with individual differences, allows for a multitude of differing hypothyroid presentations \[[@B26]\]. Overt hypothyroidism refers to patients with elevated thyroid stimulating hormone (TSH) levels and low free thyroxine (T4) levels. The term \'subclinical hypothyroidism\' has featured repeatedly in the literature in recent years. Subclinical hypothyroidism is characterised by elevated TSH concentrations associated with normal thyroxine (T4) and triiodothyronine (T3) serum levels. Subclinical hypothyroidism is further categorized with respect to the following guidelines: \[[@B29],[@B38]\] (see table [2](#T2){ref-type="table"}). ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Subclinical Hypothyroidism Grading System ::: Grade TSH levels Thyroxine (T4) levels ----------- ---------------------------------------------- --------------------------- 1 TSH above normal limits of reference range Normal 2 TSH; 10.1--20 mU/L Normal 3 TSH \> 20 mU/L Normal ::: A relatively exhaustive index of signs and symptoms is listed below. However, it is important to keep in mind that the manifestation of hypothyroidism is often far from the textbook presentation. The symptoms of hypothyroidism can be subtle and are often confused with the signs of aging \[[@B39],[@B40]\] (see table [3](#T3){ref-type="table"}). ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Signs and Symptoms of Hypothyroidism ::: System Symptoms ------------------------ ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Central Nervous System Depression, fatigue, lethargy, forgetfulness, decreased concentration, memory deficit, slow thinking, cold intolerance, nerve entrapment syndromes, decreased sweating, ataxia Musculoskeletal Muscular weakness, cramps, myalgia, arthralgia, and delayed relaxation phase of reflexes Cardiovascular Bradycardia, diastolic hypertension, increased serum cholesterol, raised triglycerides, raised low-density lipoproteins, ascites, hyperhomocysteinaemia Gastrointestinal Constipation EENT Puffy eyes, enlarged tongue, hearing impairment, goitre, hoarseness of voice, dysphagia, sore throat Genito-Urinary Infertility, menstrual irregularities, heavy bleeding, impotence, galactorrhea, hyperprolactinemia General Weight gain, dry and coarse skin, brittle hair and nails Radiological Pericardial and pleural effusions, pituitary gland enlargement ::: Treatment ========= The current treatment of choice for individuals suffering from hypothyroidism is supplementation using the synthetic thyroid hormone, levothyroxine. Levothyroxine is an artificial version of the naturally occurring thyroid hormone, thyroxine (T4). Patients are required to take 50--150 μg of levothyroxine daily for the rest of their lives \[[@B31],[@B34]\]. Thyroid hormones are monitored every 6--12 months to ensure that hormone levels are being maintained within physiological norms \[[@B31],[@B32],[@B34],[@B41]\]. Hypothyroidism, Mood Disorders and Stress ========================================= Depression is one of the major symptoms associated with hypothyroidism. According to the DSM-IV, a person presenting with depression (major depressive episode) must either have a depressed mood or interest for two or more weeks \[[@B42]\]. This mood must represent a change from the person\'s normal mood; social, occupational, educational, or other important functioning must be negatively impaired by the change in mood. To make the diagnosis of major depressive episode a patient must exhibit a depressed mood or interest and four or more of the following symptoms; sleep increase/decrease, diminished interest in formerly compelling or pleasurable activities, guilt, low self esteem, poor energy, poor concentration, appetite increase/decrease, psychomotor agitation/retardation and suicidal ideation \[[@B42]\]. While depression is strongly associated with hypothyroidism, the exact mechanisms are not yet known \[[@B43]-[@B45]\]. Haggerty states that almost 100% of patients presenting with severe hypothyroidism, are found to have serious concurrent depression \[[@B37]\]. It is well established that patients presenting with a decreased thyroid status exhibit higher lifetime frequency of depression than euthyroid subjects \[[@B46]\]. Furthermore, patients with major depression have a poorer response to antidepressant medication if they are hypothyroid \[[@B47],[@B48]\]. Subclinical hypothyroidism may also reduce the threshold for the occurrence of major depression \[[@B46],[@B47]\]. Research has sought to explain the above findings via the concept of central and peripheral hypothyroidism \[[@B46]\]. It is postulated that central abnormalities in thyroid hormone economy will not necessarily manifest in static peripheral hormone assays \[[@B49]\]. Therefore, serum thyroid hormone levels may appear normal in the presence of a central deficiency. Gunnarsson et alsuggests that biological correlates may exist for some of the depressive symptoms of hypothyroidism \[[@B50]\]. These include CSF CCK-4 (an anxiogenic peptide) and trytophan (the precursor to serotonin), as well as serum thyroid hormone levels \[[@B50]\]. It has been suggested that central serotonergic activity is reduced in hypothyroid patients. It is also postulated that comparatively higher thyroid stimulating hormone (TSH) levels may be a predictor of lower serotonin mediated endocrine responses and the presence of clinical depression \[[@B44]\]. Depression is associated with a deficiency in brain serotonergic (5-HT) activity. Duval postulates that the changes witnessed in hypothalamic pituitary thyroid (HPT) axis hormones during major depressive episodes, may be regarded as compensatory changes in order to correct reduced central serotonergic activity \[[@B44]\]. Sullivan states that transthyretin, a thyroid carrier protein, exists in differing forms both centrally and peripherally \[[@B46]\]. This carrier protein appears reduced in the cerebrospinal fluid (CSF) of depressed patients, which prevents the transport of thyroid hormones to the brain \[[@B45],[@B46],[@B49]\]. Alterations in central nervous system (CNS) thyroid hormone levels have a major effect on the serotonergic, adrenergic and GABAergic systems. Sher states that the brain utilizes thyroid hormones differently to other organs, it appears especially sensitive to subtle thyroid insufficiency. This means that thyroid function may significantly affect mood, behaviour, and cognitive function \[[@B45]\]. Research into stress, immunity and the HPT axis has demonstrated some interesting findings. In a recent study investigating the effects of stress on rat brains, Friedman et alsuggested that acute stress may alter the levels of thyroid hormone T3, but not T4 in the rat brain \[[@B51]\]. Bauer et alresearched the effects of chronic stress on hormone secretion in human subjects \[[@B52]\]. Thyroid hormone concentrations were assessed in a group of 84 East German refugees suffering from various psychiatric disorders. The results demonstrated reductions in both TSH and thyroid hormone concentrations. It was postulated that these results reflect severe chronic stress as opposed to the effects of psychiatric illness or thyroid dysfunction \[[@B52]\]. Cremaschi et alanalysed the effects of chronic stress on the thyroid axis and its influence on the immune response in animals. The results showed that the thyroid hormones could be influenced by chronic stress, in particular T3 concentrations. Furthermore it was postulated that changes in thyroid axis function may play a role in regulation of the immune response \[[@B53]\]. Bauer et aldescribe a study in which a significant number of patients with prophylaxis resistant affective disorders (bipolar depression, unipolar depression, schizoaffective disorder) improved after being given supra-physiological doses of synthetic thyroid hormone \[[@B54]\]. It was suggested that this method of intervention may represent a useful and well-tolerated maintenance treatment for patients presenting with these subcategories of mood disorders \[[@B54]\]. There is a multitude of evidence highlighting the importance of the HPT axis in depression \[[@B37],[@B49]\]. Alterations in thyroid hormones witnessed in mood disorders are unquestionable and this increased understanding of the role of thyroid hormones has lead to improvements in the management of depression \[[@B48]\]. There is however, evidence to dispute the relationship between thyroid dysfunction and mood disorders. The state of the literature as a whole suggests an association between thyroid dysfunction and depressive disorder, however the mechanism is unclear \[[@B55]\]. Engum et aldescribe a study of a large randomly selected population group examining the risk of anxiety disorders or depression in individuals with thyroid dysfunction. The results of this study showed a higher prevalence of depression in groups with previously known thyroid disorders, with lower prevalence in those with more recently diagnosed thyroid dysfunction. This study demonstrated a weak association between thyroid disorder and symptoms of anxiety and depression. Based on the results of their study Engum et alsuggests that when depression and anxiety disorders are diagnosed in thyroid dysfunction sufferers, they should be treated as separate entities, and considered that these common conditions can occur and coexist without influencing each other \[[@B55]\]. Baldini et aldescribed similar findings in a study of psychopathology and subclinical hypothyroidism. It was stated that when interfering factors related to individual vulnerability to depression and perception of disease were excluded, there was no direct correlation between subclinical hypothyroidism and mood disorders \[[@B56]\]. Fountoulakis et alsuggests that the information available on hypothyroidism and depression does not demonstrate a causal relationship \[[@B57]\]. Stating that overt thyroid dysfunction is uncommon in depressed patients. Instead those authors suggest the presence of an underlying autoimmune process, which may affect thyroid function in depressed populations \[[@B57]\]. The Biopsychosocial Model and Hypothyroidism ============================================ The biopsychosocial model represents a health concept. It depicts a treatment paradigm that acknowledges the contribution of biological, psychological and social factors in the disease process \[[@B2]\]. Alonso states that the biopsychosocial model is gaining acceptance within academic and institutional contexts \[[@B1]\]. This change however is not necessarily being reflected in the practical areas of medicine. The biopsychosocial model has been used to obtain a better understanding of the disease process, but its acceptance and incorporation into medical practice is taking longer to transpire \[[@B1]\]. Hypothyroidism is a disease that has been treated relatively successfully using pharmacological agents for many years. However, the treatment and management of this disease has been approached purely from a biomedical standpoint. Thyroid hormone levels that are considered inappropriate are restored using pharmacological supplementation. This management requires a life long commitment to drug therapy. In addition to this, significant proportions of patients under the current management protocol continue to experience the plethora of signs and symptoms associated with hypothyroidism, even though their thyroid hormone levels are returned to normal \[[@B26],[@B27]\]. In a recent review by Roberts et alit was stated that approximately one fifth of hypothyroid patients are receiving an inadequate thyroxine dose and a fifth are given an excess of the synthetic thyroid hormone. Reasons for this are postulated in table [4](#T4){ref-type="table"}. ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Potential causes of thyrotropin elevation in patients (thyroxine treated) with primary hypothyroidism ::: --------------------------------------------------------------------- Suboptimal dosing, inadequate prescribed dosage, dispensing error Non compliance Progressive decrease in endogenous thyroxine production Drug and supplement interactions Co-morbidities Malabsorption disorders Autoimmune thyroiditis Previous thyroid irradiation Reduced thyroxine absorption Other systemic illnesses --------------------------------------------------------------------- ::: Another important issue that has been raised within the current literature is the suitability of modern treatment strategies for the elderly. Elderly patients often present with an extensive list of different medications that they are taking in combination. In prescribing an additional long-term pharmacological agent to the patient presenting with reduced thyroid hormone levels, practitioners must consider the issue of drug interaction. This debate is especially relevant as this group of patients makes up a substantial proportion of the hypothyroid population \[[@B33]\]. It is essential that thyroid hormone levels be monitored at prescribed intervals in those patients undergoing levothyroxine therapy \[[@B26]\]. Weetman states that overzealous supplementation can lead to an increased risk of osteoporosis in postmenopausal women and atrial fibrillation in the elderly \[[@B38]\]. Woeber states that in patients with pre-existing angina pectoris, treatment of hypothyroidism will result in an aggravation of symptoms in one fifth of cases. Patients with coronary heart disease run the risk of myocardial infarction some time after the initiation of levothyroxine treatment \[[@B34]\]. Research suggests that is it important to provide multidisciplinary care in chronic diseases (eg rheumatoid arthritis). It is further postulated that the implementation of programmes of this nature, may not only improve functioning, but may lead to improvement in disease activity \[[@B58]\]. The World Health Organisations (WHO) International Classification of Functioning, Disability and Health (ICF), describes a framework for understanding and structuring the impact of disease on individuals \[[@B59]\]. A person\'s functioning and disability is described as a dynamic interaction between health conditions and contextual, environmental and personal factors. Health conditions encompass disease, disorders, injuries and traumas. Contextual, environmental and personal factors describe the psychosocial elements of a person\'s life \[[@B60]\]. These guidelines align favourably with the biopsychosocial paradigm. It seems that while current treatment protocols for hypothyroidism are quick and relatively cost efficient, there is a strong influence in form, from the biomedical model. The reductionist thinking associated with this model may be the cause of the inherent problems associated with the current treatment. As the biopsychosocial model gains credibility, it seems plausible that there may be a place for this paradigm in the management of hypothyroidism. If this disease were to be approached from a more wholistic standpoint, solutions to current management problems may be revealed. It is also feasible that if this disease were approached using a different model, factors in addition to pure biological influences may be discovered. Research into thyroid hormones and mood disorders further highlights the need for investigation in this area. A framework such as the biopsychosocial model seems to be a plausible model of inquiry. Conclusion ========== The biopsychosocial model represents the latest ideas in chronic illness management and compliments recent ideas in pain theory. It states that in order to rationalise and contend with chronic conditions, one must take into account the influence of biological, psychological and social factors. The biopsychosocial model is gaining acceptance within educational institutions and medical fields and is proving very successful in the areas in which it is applied \[[@B1]\]. Hypothyroidism is one chronic condition that may benefit from the application of the biopsychosocial model. Application of biopsychosocial-based interventions/therapies may help mediate some of the signs and symptoms associated with hypothyroidism. If nothing else this model represents an adjunctive framework that may facilitate a more consistent management of this chronic disease.The authors plan a randomised control trial that will seek to investigate the effect of a biopsychosocial-based treatment regime on hypothyroidism Authors\' Contributions ======================= BTB wrote the manuscript. All authors took part in researching for, reading and approving the final manuscript.	PubMed Central	2024-06-05T03:55:57.595669	2005-4-12	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1151653/", "journal": "Chiropr Osteopat. 2005 Apr 12; 13:5", "authors": [ { "first": "Benjamin T", "last": "Brown" }, { "first": "Rod", "last": "Bonello" }, { "first": "Henry", "last": "Pollard" } ] }
PMC1151654	Introduction ============ There is an increasing interest and acceptance in psychosocial factors and their correlations to the onset and outcomes of acute pain episodes. This review will briefly review its evolution and summarize the past and present theoretical models in relation to low back pain (LBP). Psychlit, MEDLINEand medindexsearches were conducted to identify relevant articles with the search words \'psychological factors, chronic/persistent pain\'. Historical development ---------------------- The psychological and psychiatric aspects of pain had been infrequently noted by modern writers as early as 1768. For a comprehensive historical review see Merksy & Spear \[[@B1]\]. By the second half of the 19th Century, however, pain was considered sensorial and organic causes were offered to explain all pains, even those without an obvious basis in tissue damage or organic disease. The belief that all pain was a direct result of tissue damage was firmly entrenched by the early 20th Century \[[@B2]\]. By the late 1950\'s it became increasingly evident that sensory explanations failed to account for certain puzzling pain phenomena (e.g., relief from pain with placebo interventions, phantom limb pain). Around the mid-20^th^Century several different theories were developed from differing theoretical backgrounds to explain the observation that sensory input did not always correlate with pain. I have summarized these differing schools of thought by précising a comprehensive review by Gamsa \[[@B3],[@B4]\]. Psycholanalytic Formulations ---------------------------- Here intractable pain, which defies organic explanations, was seen as a defence against unconscious conflict. Emotional pain is displaced onto the body where it is more bearable. For example, conscious or unconscious guilt with pain serving as a form of atonement, or the development of pain to replace feelings of loss. Critics have raised serious methodological and conceptual concerns \[[@B5],[@B6]\]. For example; the ability to quantify and research the constructs of Id, ego and superego. Psychoanalytic thinking no longer forms a significant basis for research or source of current interventions. Behaviourist Models ------------------- Following the work of Skinner \[[@B7]\], behaviourists tried to show that all behaviour could be shaped, altered, weakened or strengthened as a direct of environmental manipulations. Fordyce et al. \[[@B8]\] were the first to apply the behaviour model to pain. It was thought that there was a simple causal connection between pain and its reinforcers. Respondent (acute) pain was seen as a reflexive response to antecedent stimulus (tissue damage). The respondent pain may eventually evolve into operant and persisting pain if the environment offers pain contingent reinforcement. Pain behaviour may also be learned by observing \"pain models\" i.e., individuals who exhibit such behaviour. More complex factors such as personal dynamics, emotional state, physical vulnerability, and numerous psychosocial variables were not addressed. It proposed that operant pain persists because the behaviour of others (family, friends and health care providers) during the acute pain stage reinforced that pain returned secondary gains, such as permission to avoid chores, or obtain otherwise unobtainable attention and care. Behaviour models have however contributed to the study of pain by the introduction of carefully designed control procedures and laboratory methods \[[@B4]\]. Cognitive Approaches -------------------- Cognitive approaches were inspired in part by Melzack and Wall\'s \[[@B9]\] gate control theory, which established a role for the cognitive-evaluative process in the modulation of pain. Since the mid 1970\'s proponents of cognitive theory studied the influence of the meaning of pain to patients, and examined the effect of coping styles on pain, for further review see Weisenberg \[[@B10]\]. Cognitive theory examines intervening variables such as attributions, expectations, beliefs, self-efficacy, personal control, attention to pain stimuli, problem solving, coping self-statements and imagery. Pain studies investigated the effects of these thought processes on the experience of pain and related problems. Cognitive theory has added an important dimension to psychological research into pain, but cognitive theorists themselves emphasise that they do not provide thesolution, in isolation from other aspects of the multidimensional problem of pain \[[@B4],[@B19]\]. The combination of cognitive and behavioural approaches has been employed extensively in pain programmes during the last 15--20 years with some reported success \[[@B11]\]. Psychophysiological Approaches ------------------------------ Examines the influence of mental events (thoughts memories and emotions) on physical changes which produce pain, for a comprehensive review see Flor and Turk \[[@B12]\]. For example, general arousal models propose that frequent or prolonged arousal of the Autonomic Nervous System (ANS) including prolonged muscular contractions, generate and perpetuate pain. Treatment, such as EMG, biofeedback, and relaxation techniques are designed to decrease the levels of muscular tension and ANS arousal and thereby decrease the pain. Studies have shown positive results from these interventions, but not necessarily more than other psychological techniques \[[@B3],[@B4]\]. In sum, psychological thought during the past half century has shifted from linear to multicausal models of pain. Methods of investigation have also improved. Current theoretical models ========================== A substantial number of acute painful musculoskeletal injuries do not resolve quickly and account for the majority of the associated costs \[[@B13]\]. Early intervention appears to result in improved outcomes \[[@B14]\]. Consequently, it is not surprising that the on-going evolution of the understanding of the non-physical aspects of pain has been applied to the areas of screening for, intervening in and predicting those at risk of developing into a chronic and disabling situation \[[@B15],[@B16],[@B33]\]. The recent New Zealand Government review into LBP, its subsequent published guidelines, and resultant growing acceptance of the \"Yellow Flags\" concept is a pertinent example \[[@B17]-[@B19]\]. Variables such as attitudes, beliefs, mood state, social factors and work appear to interact with pain behaviour, and are cumulatively referred to as psychosocial factors. However, to date there has not been developed a comprehensive, multivariate and empirically supported Integrated Biopsychosocial Risk-for-Disability Model. During a plenary session at the Forth International Forum on LBP Research in 2000 \[[@B20]\] Pincus et al amalgamated the Cognitive and behavioural thinking and proffered the closest structure yet to such a model. It has sought to incorporate many of these factors, and as such offers a structure from which to review these psychosocial factors. The cognitive-behavioural researchers in the late 20^th^century noted that acute pain was associated with a pattern of physiologic responses seen in anxiety attacks, whilst chronic back pain was characterized more effectively by habitation of autonomic responses and by a pattern of vegetative signs similar to those seen in depressive disorders. One of the prominent researchers, Waddell, noted that one of the striking findings was that \"fear of pain was more disabling than the pain itself\" \[[@B21]\]. As a result the notion that reduced ability to carry out daily tasks was merely a consequence of pain severity had to be reconsidered. Several studies have indicated that pain-related fear is one of the most potent predictors of observable performance and is highly correlated to self-reported disability levels in subacute and chronic pain \[[@B22],[@B23]\]. In the acute pain situation, \"avoidance\" behaviours, such as resting, are effective in allowing the healing process to occur \[[@B24]\]. In chronic pain patients, the pain and disability appear to persist beyond the expected healing time for such a complaint. The danger is that a protracted period of inactivity, as a strategy for coping with the persistent pain may lead to a disuse syndrome (see Figure [1](#F1){ref-type="fig"}). This is a detrimental condition. It is associated with physical deconditioning such as loss of mobility, muscle strength and lowered pain thresholds (allodynia). Consequently, the performance of daily physical activities may lead more easily to pain and physical discomfort. As a result, the avoidance of activity becomes increasing likely, as does the risk of chronicity. Cognitive-behavioural theorists have variously described this process that leads to chronicity stemming from pathological levels of fear / anxiety as \"Fear of pain\" \[[@B25]\], fear of physical activity and work \[[@B26],[@B27]\], avoiders and confronters \[[@B28]\], kinesiophobia \[[@B29]\] and anxiety sensitivity \[[@B30]\]. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### A cognitive-behavioural model of pain related fear \[43\]. ::: ![](1746-1340-13-6-1) ::: When a person experiences pain they experience varying degrees of psychological distress. A recent study suggests that as many as one third of people seeking care at physical therapists may have significant levels of distress \[[@B31]\]. Many dimensions of this process have been identified and their role posited in the development of chronicity. One such example is catastrophic thinking processes and is broadly described as an exaggerated orientation towards pain stimuli and pain experience \[[@B32]\]. Negative appraisals about pain and its consequences have been postulated to be a potential precursor to persistent pain. People who consider pain as a serious threat to their health are more likely to become fearful as compared with those who approach pain as a trivial annoyance \[[@B33]\]. Pain-related fear can also contribute to disability through interference with cognitive functions. Fearful patients will tend more to possible signals of threat (hyper-vigilance) and will be less able to shift attention away from pain related information at the expense of other tasks, including actively coping with problems of daily life \[[@B34]\]. Although these and other factors such as coping strategies \[[@B35]\], sense of control \[[@B36]\], personality type \[[@B37]\], faith and religious beliefs \[[@B38]\], have been reported in literature (for a comprehensive review see Keefe et al.\[[@B44]\], the most significant and reproducible factors have been mood / depression and to a lesser extent somatization / anxiety \[[@B16],[@B39]\]. Depression has been associated with decreased pain thresholds and tolerance levels, reduced ability, general withdrawal and mood disturbance such as irritability, anhedonia (loss of enjoyment of good things in life), frustration and reduced cognitive capacity. Somatization disorder is a chronic condition in which there are numerous physical complaints. It is perceived as very similar in nature to, and difficult to differentiate from an anxiety disorder \[[@B40]\]. The most common characteristic of a somatoform disorder is the appearance of physical symptoms or complaints for which there is no organic basis. Such dysfunctional symptoms tend to range from sensory or motor disability, and hypersensitivity to pain. This is a difficult and complex syndrome and is more fully dealt with elsewhere \[[@B41]\]. A mention should be made of occupational factors. Job dissatisfaction has repeatedly demonstrated itself to be a significant factor in disability / persistent pain studies. The most recent literature has implicated such factors as support from supervisors at work and low job control (i.e., inadequate power to make decisions and utilize one\'s skills) which can create distress, and, when perpetual, may result in ill health \[[@B42]\]. Conclusion ========== In sum, while this cognitive-behavioural model focused on fear / avoidance shows much promise; it has yet not been validated by the research to date \[[@B15]\]. There are studies in progress that may further our knowledge of identifying those at risk of progressing from acute to chronic \[[@B13]\]. Until the veracity of this model becomes further elucidated, depression and somatization / anxiety should be regarded as the central and dominant influencing psychological factors in the assessment for identification and intervention strategies. Competing interests =================== The author(s) declare that they have no competing interests.	PubMed Central	2024-06-05T03:55:57.599100	2005-4-27	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1151654/", "journal": "Chiropr Osteopat. 2005 Apr 27; 13:6", "authors": [ { "first": "Stanley I", "last": "Innes" } ] }
PMC1156864	Findings ======== The National Health Statistics in Cambodia \[[@B1]\] estimated 89.4% of pregnant women in Cambodia to have given birth outside of health facilities, and approximately 48.4% of pregnant women to have at least once attended antenatal care (ANC) at health centres in 2003, which is an entry point of prevention of mother-to-child transmission of HIV (PMTCT) services. Furthermore, most of ANC services are provided in health centres or fields as one of outreach activities, where laboratory services by technicians are not available. In May 2003, the ministry of health in Cambodia started expanding PMTCT services following pilot projects in urban cities along with training of VCCT counsellors and laboratory technicians. Those pregnant women in ANC wishing to receive PMTCT services are offered voluntary confidential counselling and testing (VCCT) consisting of pre-test counselling, HIV testing and post-test counselling. However, some blood samples withdrawn by VCCT counsellors or some pregnant women that accepted VCCT need to be transported to a nearest laboratory by some means. For this reasons, some VCCT counsellors could be expected to provide HIV testing as well as pre-test and post-test counselling services to expand PMTCT services to cites where laboratory services are not available, and to increase uptake of pre-test and post-test counselling \[[@B2]\] if the HIV testing was accurately performed by VCCT counsellors in health centres or fields. In addition, rapid HIV testing is useful to improve access to learn HIV status in populations at high risk of HIV infection \[[@B3]\]. The aim of this study is to assess our training on HIV testing by comparing HIV testing performances of trained non-laboratory health staffs and laboratory technicians using a rapid and simple test. Those VCCT counsellors involved in PMTCT in our centre were trained on HIV testing for half day using Determine (Abbot Laboratories) HIV1/2 test kits by experienced laboratory technicians. All the trained VCCT counsellors were midwives working for ANC and delivery ward in our centre, that hadn\'t have experience and knowledge on laboratory works. The contents of the training were how to use a pipette (30 minutes), how to process whole blood samples with chase buffer (90 minutes) including practical training, and how to read test result (60 minutes) according to the instructions provided by the manufacturer. The VCCT counsellors trained on the test withdrew blood sample with EDTA tubes from the clients\' vein after pre-test counselling following ANC if informed consent to participate in PMTCT services was obtained, and performed the rapid test with whole blood and chase buffer. The rest of the blood sample was sent to a laboratory in our centre and centrifuged. The plasma samples were stored at 4 degrees centigrade and tested by the laboratory technicians with the same test kit on the next day. The test results by laboratory technicians were reported to counsellors in individual envelopes to keep confidentiality according to their routine. The rest of the plasma was stored in a freezer at -20 degrees centigrade for further examination. Only code numbers were used for the identification of samples with confidentiality, and printed on stickers in advance, which were used to label tubes and reports to minimize human error such as miswriting. Two reports from counsellors and laboratory technicians were compared, and discordant reports in two groups were re-tested with the same test kit in front of two groups using the same blood sample kept in the freezer, and cause of the discordance was detected in discussion with both groups. Of 563 blood samples tested by six VCCT counsellors and three laboratory technicians, 11 samples (2.0%) were reported positive in each group, however four discordant reports (0.7%) between the groups were observed, in which two positive reports and two negative reports by the counsellors were negative and positive by the laboratory technicians, respectively (table [1](#T1){ref-type="table"}). Further examination using the same test kit and blood samples of the discordant reports confirmed that two positive reports and two negative reports by the counsellors were positive and negative, respectively. In discussion with the groups, human error in writing reports in the laboratory was identified as a cause of these discordant reports. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Comparison of reports of HIV test results between by VCCT counsellors and laboratory technicians. Further examination using the same test kit and blood samples confirmed that all the reports by the VCCT counsellors were correct, and that four discordant reports\* were caused by human errors in the laboratory. ::: HIV positive reports by VCCT counsellors HIV negative reports by VCCT counsellors Total ------------------------------------------------ ------------------------------------------ ------------------------------------------ ------- HIV positive reports by laboratory technicians 9 2\* 11 HIV negative reports by laboratory technicians 2\* 550 552 Total 11 552 563 ::: In our study, the accuracy of reports by the VCCT counsellors scored 100%, which was higher than laboratory technicians (99.3%) though the Determine rapid and simple test requires one more step to use chase buffer for testing whole blood samples. This result showed that the half-day training for VCCT counsellors was feasible enough to provide satisfactory proficiency for non-laboratory health staffs using the rapid and simple test in order to implement the PMTCT services where laboratory technicians were not available. A study from the United States used the OraQuick rapid test to evaluate how well untrained persons with no laboratory experience can perform the HIV test, and concluded that they could achieve a level of satisfactory proficiency however they could not reach 100% accuracy \[[@B4]\]. The guidelines by World Health Organisation and National Centre for HIV/AIDS, Dermatology and STIs in Cambodia recommend either parallel testing or serial testing using two different test kits for VCCT activities \[[@B5],[@B6]\]. The reason why our study, however, used only one test kit was that this study was the first step to assess our training and to evaluate accuracy of testing by the trained VCCT counsellors based on routine reports of HIV test results. Further study using two rapid test kits would be referred to follow the guidelines for the implementation of HIV testing by counsellors in Cambodia. Ziyambi, Z. et al. \[[@B7]\] reported that the combined sensitivity and specificity with two rapid tests by trained non-laboratory staff was 100% in Zimbabwe. However, our study was more practical by comparing two routine reports from laboratory technicians and the trained VCCT counsellors. These findings suggest that the rapid and simple testing by non-laboratory health staffs trained through the half-day training could be recommended to expand PMTCT services providing same-day results. VCCT services for PMTCT with same-day results could expect to increase the access to HIV prevention and care \[[@B8]\]. Furthermore, it was realized in our study that human error was more likely to occur in process of the laboratory, even though the laboratory technicians were more skilled and experienced than VCCT counsellors and made effort to reduce the error by using the stickers. This happened not because of their capasity to process samples and to read the results, but because laboratory technicians have more complex recording and reporting processes before giving reports to counsellors. HIV testing by VCCT counsellors could reduce this risk as well. Whoever performs the test, however, importance of having strict procedures for quality assurance in testing cannot be overstated \[[@B9]\], and linkages to high-quality reference laboratory facilities for confirmatory testing and supervision system need to be carefully considered to expand PMTCT services using rapid tests \[[@B10]\]. In conclusion, the half-day training was sufficient enough for non-laboratory health staffs to attain proficiency of HIV testing and its report with a rapid and simple test kit, and it could contribute to enhance and to expand PMTCT activities more efficiently where laboratory technicians are not available. Further study using two kinds of test kits is needed to evaluate the feasibility of the training more practically. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= K. Kanal and K Kakimoto carried out data analysis and drafted this manuscript. T. L. C. and Y. Mukoyama contributed to acquisition of data, and participated in coordination of study design to involve VCCT counsellors for PMTCT in this study. L. S. and Y. Morikawa participated in this study to organise the training on HIV testing for the VCCT counsellors, and contributed to acquisition of data from the laboratory. K. Kakimoto conceived of the design of this study with intellectual contribution. Acknowledgements ================ This study was carried out as a part of technical assistance by Japan International Cooperation Agency (JICA), and funded by JICA Maternal and Child Health Project in Cambodia and UNICEF. We thank Dr. Etienne Poirot and Ms. Chin Sedtha from UNICEF Cambodia office for their kind cooperation. Part of this study was presented at the 15th International AIDS Conference in Bangkok, Thailand, July 2004.	PubMed Central	2024-06-05T03:55:57.600522	2005-5-20	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1156864/", "journal": "AIDS Res Ther. 2005 May 20; 2:5", "authors": [ { "first": "Koum", "last": "Kanal" }, { "first": "Thai Leang", "last": "Chou" }, { "first": "Ly", "last": "Sovann" }, { "first": "Yasuo", "last": "Morikawa" }, { "first": "Yumi", "last": "Mukoyama" }, { "first": "Kazuhiro", "last": "Kakimoto" } ] }
PMC1156865	Background ========== Corynebacterium diphtheriaeis well known as an agent of localized respiratory tract disease potentially complicated by systemic effects of exotoxin \[[@B1],[@B2]\]. It is also able to cause cutaneous and wound infections. But nontoxigenic strains can produce atypical manifestations of disease. They are able to cause diseases such as mild diphtheria-like pharyngitis, cutaneous infections, septic arthritis, abscesses, septicemia and endocarditis \[[@B3]-[@B14]\]. The pathogenesis of infection caused by nontoxigenic C. diphtheriaeis unknown. We have described the first case of bacteremia and endocarditis due to nontoxigenic C. diphtheriaevar. gravisin Poland. Since 1996 there was no cases of C. diphtheriaeinfections in Poland area. Case presentation ================= In January 2004 previously healthy 38-year-old man was admitted to hospital with fever (40°C) and arthralgia of upper and lower extremities. He was not addicted to drugs or alcohol. He smoked about 20 cigarettes per day. Physical examination revealed mild tachycardia with normal cardiac sounds and normal blood pressure. Oral cavity examination showed poor dentition. He presented generalized arthralgia and edema and redness of involved joints. Also there was a 3 day history of papulomacular haemorrhagic rash on his lower extremities. The patient had no symptoms of respiratory tract infection and X-ray examination showed no evidence of pneumonia. Laboratory evaluation revealed elevated white blood cell count (17,8 × 10^3^/μl) with 34% of band neutrophils and 51% of granulocytes, decreased platelet count (29 × 10^3^/μl), slightly elevated liver enzymes (AST 52 U/L, ALT 41 U/L, LDH 589 U/L, alkaline phosphate 150 U/L), hyperbilirubinemia (2,69 mg%), elevetad creatinine concentration (1,51 mg%), mild proteinuria, leucocyturia and erythrocyturia. The initial diagnosis was septicemia. Three blood samples were drawn during two first days of hospitalization for microbiological evaluation and empirical therapy with ceftazidime and teicoplanin was started. Nontoxigenic C. diphtheriaebiotype graviswas isolated from all blood cultures. On the third day of treatment the antibiotics were changed to amikacin and ciprofloxacin according to antibiogram. Despite resolution of most symptoms and negative blood and throat swab cultures after eight days of treatment the patient was still febrile. For that reason ciprofloxacin was changed to clindamycin. Transthoracic and transoesopharyngeal echocardiography performed on the 13th day of treatment showed two vegetations attached to the mitral and aortal valves. The patient underwent surgery for replacement of both valves. The cultures from vegetations were negative. After the operation the patient recovered. Although the cultures were negative we have supposed that the vegetations were caused by C. diphtheriaebecause the patient had no cardiac troubles before bacteremia. Nontoxigenic C. diphtheriaebiotype gravisisolated from blood cultures was identified and biotyped with use of morphological and biochemical methods as described elsewhere. Toxin production was examined in vitroby the conventional \[[@B1]\] and modified Elek test \[[@B15]\]. Polymerase chain reaction (PCR) was used for the detection of diphtheria toxin gene \[[@B1],[@B16]\] and the PCR result was negative. The susceptibility of isolates to 13 antibiotics was determined by the disk diffusion method accordingly to the National Committee for Clinical Laboratory Standards \[[@B17]\] guidelines on Mueller-Hinton II blood agar (supplemented with 5% sheep blood). However NCCLS does not define breakpoints for Corynebacterium sp. For that reason interpretation was done comparatively as for Streptococcus spp. and Staphylococcus spp., because some breakpoints are different for that genera. The antimicrobial disks contained penicillin, cefaclor, cefuroxime axetil, cefazolin, ceftazidime, ceftriaxone, cefepime, amikacin, meropenem, azithromycin, trimethoprim-sulfamethoxazole, vancomycin and teicoplanin. Determination of MIC (results are shown in brackets) for ampicillin (0.5 mg/L), gentamicin (0.38 mg/L), ciprofloxacin (0.125 mg/L), clindamycin (0.25 mg/L), erythromycin (0.016 mg/L), chloramphenicol (2 mg/L) and tetracycline (0.5 mg/L) was done using E-test strips. Clindamycin and erythromycin MIC breakpoints for Streptococcus spp. are lower than for Staphylococcus spp. but both interpretations showed susceptibility of examined C. diphtheriaestrain. Ampicillin MIC breakpoints pointed to investigated strain as ampicillin resistant. The strain was also resistant to penicillin and ceftazidime and intermediate to cefuroxime axetil and ceftriaxone. Conclusion ========== C. diphtheriaecauses systemic disease sporadically. Only 58 cases of bacteremia infections due to that microorganism were described between 1893 and 2003 \[[@B3]-[@B9]\]. Forty four of them were caused by nontoxigenic strains. There has been no information of bacteremia of such etiology in Poland. In previously described cases infections were connected with low socioeconomic group of people. Most of them were intravenous drug users, alcohol addicts, unemployed and homelesses \[[@B4],[@B7]-[@B12],[@B18],[@B19]\]. Our patient does not belong to any of above mentioned risk groups. It is supposed that the predominant route for bacterial contamination and penetration into the bloodstream are various skin lesions such as cutaneous ulcers, bullous pemphigoid, scabies and open fractures \[[@B9],[@B12],[@B18],[@B19]\]. But in the case described here no connection between bacteremia and skin lesions was found. We have supposed that mass dental caries enabled bacteria penetration into the bloodstream. The strain isolated from the patient was resistant to penicillin and third generation cephalosporins. Penicillin G, erythromycin or amoxicillin is the reference treatment for C. diphtheriaeinfections \[[@B1],[@B13],[@B20]\] but resistance to these antibiotics was reported \[[@B1],[@B13],[@B14]\]. Our results are also in agreement with the observation of Patey et al. \[[@B7]\] that appreciable resistance to third generation cephalosporins exists among C. diphtheriaestrains. Diphtheria is still endemic in Eastern Europe and other regions of the world although it has virtually disappeared in developed countries following mass immunization in the 1940s. Current vaccine against diphtheria contains the toxoid so it protects only against the toxigenicity but not the invasiveness of C. diphtheriae\[[@B9]\]. The high rate of immunization with diphtheria toxoid may place selective pressure on the microorganism to develop other patogenicity factors. The bacterium could acquire exogenous DNA such as transposons, bacteriophages DNA or plasmids that contains single virulence genes or whole pathogenicity islands. Nontoxigenic C. diphtheriaecan produce atypical manifestations of disease. The pathogenesis of infection caused by nontoxigenic strains of C. diphtheriaeis unknown and requires investigation. The organism is capable of tissue invasion and causing fulminant disease and appears to have a predilection for cardiac valvular endothelium and synovium \[[@B8]\]. The occurrence of joint involvement connected with bacteremia due to C. diphtheriaewas also reported \[[@B7],[@B8]\]. The case presented in this article and cases described in other papers show the importance of identifying Corynebacteriumisolates from normally sterile sites at the species level. Determination of antimicrobial susceptibility has also fundamental role in success of treatment because resistance to some antimicrobial agents has been reported in C. diphtheriae\[[@B1],[@B13],[@B14]\]. It is worth to underlined that infections connected with C. diphtheriaecan occur in immunized population. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= AAZ carried out molecular and phenotypic tests of toxigenicity and participated in identification and antimicrobial susceptibility tests and drafted the manuscript, MZ participated in identification and antimicrobial susceptibility tests and helped to draft the manuscript, RBP and IS participated in diagnosis, observation and treatment of the case and helped to draft the manuscript. All authors read and approved the final manuscript. Acknowledgements ================ We thank Dr Piotr Litwinski and Dr Jan Chejbowicz from Department of Cardiosurgery, Institute of Cardiology, Warsaw for their surgical cooperation.	PubMed Central	2024-06-05T03:55:57.601570	2005-5-5	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1156865/", "journal": "Ann Clin Microbiol Antimicrob. 2005 May 5; 4:8", "authors": [ { "first": "Aleksandra Anna", "last": "Zasada" }, { "first": "Maria", "last": "Zaleska" }, { "first": "Regina Beata", "last": "Podlasin" }, { "first": "Ilona", "last": "Seferyńska" } ] }
PMC1156866	Background ========== Complexity combined with an imprecise terminology has hindered the understanding of biology. A formal and structured vocabulary is now being developed to address this imprecise biology terminology. This vocabulary or Gene Ontology (GO) is being developed by the Gene Ontology Consortium (GOC) \[[@B1]\] to standardize the descriptions of gene products. Ontologies define the basic terms and relations comprising the vocabulary of a topic area, as well as the rules for combining terms and relations to define extensions to the vocabulary \[[@B2]\]. Despite these efforts, the mechanism of representing these terms lacks a unifying architecture that can be applied to the annotation of a gene product. However, computer science has developed a well-defined process and methodology for the development of software models. Adapting this process and methodology can orchestrate the assembly of biological models with integrated gene ontologies. In doing so, a standardized terminology and object-oriented model is created that can facilitate communication between biologists and computer scientists. The Gene Ontology project is a collaborative effort that addresses the need for a controlled vocabulary that provides a consistent description of gene products in different databases \[[@B1]\]. The GO collaborators are developing three structured, controlled vocabularies that describe gene products, which have been classified into molecular function, biological process, and cellular component domains. GO terms are organized in structures called directed acyclic graphs (DAGs), which differ from hierarchies in that a \'child\' (more specialized term) can have many \'parents\' (less specialized terms). As part of these graphs, each component is given a GOid (unique identifier), and is associated with a GO definition. Collectively, these agreed upon terms are being developed to help explain various aspects of biology. When applied to a gene, that gene is annotated with a concise description of its molecular function, cellular location and associated biological processes. However, the GOC never intended to represent gene products or correlate ontological terms with these gene products \[[@B1]\]. To address this need, a Gene Ontology Annotation database \[[@B3]\] has been created to associate the GO terms with their gene product counterparts. With sustained effort, the descriptions of these gene products will ultimately be established. Still, much of the current bioinformatics work regarding GO has focused on constructing databases \[[@B4]-[@B7]\], applying it to other research areas \[[@B8]-[@B22]\], and building tools to mine the GO database. (For a description of some of these tools see \[[@B23]\].) In addition, there has been an ongoing discussion regarding the depth of information obtained from the Gene Ontology \[[@B24]\]. It has been noted that there remains a need for a unifying architecture that integrates all three GO domains as part of a gene product\'s annotation. Furthermore, to enhance the Gene Ontology and facilitate its use as a cross-disciplinary tool, several additional issues need to be addressed. First, relationships between the biological processes, molecular functions and cellular components are not readily apparent \[[@B25]-[@B28]\]. Second, GO terms lack details. For instance, when one looks at molecular function there is no indication of what is inputted or outputted. Finally, existing tools such as GO-DEV \[[@B29]\] only contain software used for tool development and information retrieval, not software modeled directly after the three domains of the Gene Ontology. However, these issues can be resolved by integrating the Gene Ontology into an object-oriented system. On a conceptual level, the Gene Ontology has features that support an object-oriented architecture. Consequently, the Gene Ontology can be applied and mapped to the fundamental concepts that form the object-oriented paradigm (i.e. class, object, inheritance, composition, polymorphism, and encapsulation) (Table [1](#T1){ref-type="table"}). Furthermore, in an object-oriented sense, biological process terms are equivalent to high-level concepts. However, GO biological process terms do not contain descriptive information about the dynamics or static interactions defined by the terms. By translating a biological process into an object-oriented model the dynamic and static events occurring within a process can be represented. Building a static and dynamic model of a biological process requires defining the components of the process as well as the functions and attributes contained within these components. These components are biological entities (bioentities) that may include individual gene products, whose processes, functions and cellular components are captured in the Gene Ontology, or other higher-level entities such as gene product complexes. The functions of gene products are the jobs or abilities that it has. In the GO terminology these are described in the molecular function domain. These are analogous to the operations that an object can perform in an object-oriented paradigm. Attributes, which define key properties of a component that when changed may alter the function of that component, may be defined by the cellular component and molecular function sections. For example, the cellular component domain can specify the place in a cell where a gene product is located. When there are multiple cellular components associated with a gene product, however, there is currently no mechanism to designate which cellular component represents the appropriate location. The unified modeling language has been used to capture various aspects of biology \[[@B30]-[@B32]\]. These examples highlight the utility of the unified modeling language as a tool for biological data integration, and indicate that it can be applied to construct large, complex biological models. Therefore, to demonstrate the feasibility of integrating the Gene Ontology into an object-oriented model we have created unified modeling language (UML) representations of a GO biological process, \"transforming growth factor beta (TGF-beta) receptor complex assembly\" (GO:0007181). The TGF-beta receptor pathway is involved in numerous cellular events including apoptosis, tumor development, differentiation, and development. These processes stem from the binding of TGF-beta to its cellular receptors. Briefly, dimerized TGF-beta 1 binds to TGF-beta receptor II (TβRII) and then TGF-beta receptor I (TβRI) complexes \[[@B33]\], causing their tetramerization (two type II receptors and two type I receptors) \[[@B34]-[@B36]\]. Constitutively activated type II receptor phosphorylates and activates type I receptor. Type I receptor propagates the signal by phosphorylating Smad 2, which is presented by the Smad Anchor for Receptor Activation (SARA) \[[@B37]\]. Phosphorylation of Smad destabilizes the Smad interaction with SARA, releasing it \[[@B38]\]. On TGF-beta stimulation, Smad 2 forms heterotrimeric complexes with Smad 4 and accumulates in the nucleus, binds DNA and remains for several hours \[[@B39]-[@B42]\]. Dephosphorylation allows Smad 2 to dissociate from Smad 4 and to be exported to the cytoplasm \[[@B43],[@B44]\]. If the receptors are no longer active, then the Smads accumulate over time in the cytoplasm \[[@B44]\]. Alternatively, activated Smad 2 is ubiquitinated in the nucleus and undergoes proteasome-mediated degradation \[[@B45]\]. To create a unified model using the Gene Ontology we have taken the biological process term, \"transforming growth factor beta (TGF-beta) receptor complex assembly\" (GO:0007181), and used object-oriented models to define its dynamic and static architecture. We also show that one can augment the biological process domain terms by using the ontological terms and gene products associated with this process, and integrating them into an object-oriented model. Furthermore, we show that the molecular function, and cellular component domains can serve as a basis for the generation of object functions and attributes to create a standardized, comprehensive, and integrated model encompassing all the Gene Ontology domains. Results ======= Converting GO directed acyclic graphs to object-oriented diagrams ----------------------------------------------------------------- The current DAG structure in which the Gene Ontology is represented is not readily amenable to transformation into software code. However, the architecture of directed acyclic graphs mimics that of an object-oriented class diagram. GO terms are presented in a parent-child hierarchy connected by \'is a\' (generalizations) and \'part of\' (composition) relationships. Read from top to bottom, the GO terms proceed from more specific to less specific. Directed acyclic graphs also allow the properties of multiple parent nodes to be inherited by child nodes, a form of multiple inheritance in object-oriented modeling. In figure [1A](#F1){ref-type="fig"}, cellular components related to the TGF-beta receptor complex are shown. One can create a UML diagram to mimic these relationships as shown in figure [1B](#F1){ref-type="fig"}. Since not all cellular components involved in the TGF-beta receptor complex assembly process are present in the current Gene Ontology, additional gene products based on literature searches were added to the object-oriented diagram (Figure [1B](#F1){ref-type="fig"}, shaded boxes). Relationships are captured as an object-oriented system through containment, composition and inheritance. Cellular components were decomposed into objects and connected via generalizations, which illustrate inheritance. Because these terms inherit all the attributes of their parents, only GO terms at the terminal nodes need be characterized in an object-oriented model. Relationships described as \'part of\' were also extrapolated into object-oriented terms as composition. The functions of gene products were also decomposed into object functions. The creation of object functions involved the transition from gene product functions to standardized GO molecular functions, and then to standardized, fully parameterized object functions. By applying formal ontological terms from the molecular function domain to gene products, object functions can be created with a consistent vocabulary. In table [2](#T2){ref-type="table"} we show the relationships between the function of a gene product defined in our model, and the GO molecular function term most closely corresponding to that cellular function. Here, we first compared ontology terms from the molecular function domain to those ascribed to individual gene products. Due to the incompleteness of the Gene Ontology, some gene product functions were extrapolated from the current literature, and then comparable GO molecular function terms were assigned to the gene products. Next, these molecular function terms were converted to object functions through reverse engineering. We identified the parameters that would normally be input into and output from a cellular reaction. In this way we defined the input and output parameters necessary for an object function. The object function itself was given the GOid that corresponds with its closest matching molecular function as defined by the GOid\'s definition. Together, object functions were created that are fully parameterized with inputs and outputs and that contain a standardized GO notation. We conclude that it is feasible to create standardized functions for objects based on the current literature and an approved ontology. Together, ontological terms can be integrated into an object-oriented model paralleling the relationships, capturing the inherited aspects of the GO terminology, and providing a compact architecture while maintaining a standardized notation. Sequence diagram generation --------------------------- The GO biological process term, TGF-beta receptor complex assembly (GO:0007181), contains both static and dynamic features. The events of the TGF-beta receptor complex assembly (GO:0007181) process include TGF-beta binding (GO:0050431) to its receptors and SMAD binding (GO:0046332) and activation (GO:0042301). To capture the dynamic nature of these actions as an object-oriented software system, sequence diagrams were created. The events leading to Smad 2 activation are reflected chronologically in a high-level sequence diagram in Figure [2](#F2){ref-type="fig"}. The creation of the sequence diagram first entails identifying gene products and their functions by literature searches. Simple or complex bioentities are modeled as objects, which are represented by rectangles with vertical lifelines. Ontology terms taken from the molecular function domain that best corresponded to these functions were incorporated as object functions, which represent the functions of these gene products. These functions are implemented by the methods contained within the objects. Furthermore, these methods allow an object to communicate and interact with other objects, thus capturing cellular activities. To capture interactions between objects, one object can call a method of another object by connecting object lifelines in the sequence diagram (Figure [2](#F2){ref-type="fig"}). This invocation of a function of one object by another is described as one object sending a message to another object. Alternatively, a message may be passed from an object to itself as in the case of self-checks or autoactivation signals. In this way, real world processes may be captured using an object-oriented approach. For instance, to capture the formation of the TGF-beta and TGF-beta RII complex a GOid that closely corresponds to this ability is chosen as the method name. In this way the method can be cross-referenced to a GO term. Specifically, the method \'GO:0046982 (in Dimerized\_TGF-beta, in Dimerized\_RII)\' references via the GOid, GO:0046982, \"protein heterodimerization activity\", and shows that a homodimer of TGF-beta and a homodimer of TGF-beta RII are needed to form the complex. Here, each dimer is thought of as a single entity, so the combination of these two entities is best represented as heterodimerization. A value of TGF-beta-TβRII\_Complex is returned upon completion of the method as indicated by the return arrow. In contrast, the function call \"GO:0042803 (in: SMAD2, in SMAD2),\" references a self-call. The GOid can be cross-referenced to \"protein homodimerization activity\", which requires two SMAD2 components to generate the SMAD2 homodimer, but the message is passed only within the SMAD2 object. Furthermore, a message need not accept any parameters, as in the \"translate()\" function, which only returns a Boolean value indicating whether the action has occurred. Additional events such as TGF-beta RI activation, and Smad homodimerization, binding and activation are also reflected in figure [2](#F2){ref-type="fig"}. Together, this diagram demonstrates that the sequence of events occurring in the biological process, TGF-beta receptor complex assembly (GO:0007181), can be represented using the Gene Ontology, and can be integrated as part of the dynamics of an object-oriented software system. Activity diagram generation --------------------------- Biological processes are created from a series of complex events. While there may be one main event scenario that most frequently leads to a specific outcome often, alternative scenarios that lead to a process conclusion exist. This is exemplified by the sequence of events found in the TGF-beta receptor complex assembly (GO:0007181). For instance, TGF-beta may initially bind to TGF-beta RII or TGF-beta RIII. To capture these alternative events as part of the dynamic architecture, an activity diagram was created to reflect the initial stages of TGF-beta signaling (Figure [3](#F3){ref-type="fig"}). Unlike the sequence diagram, which captures main scenario events, the action sequence or flow of the activity diagram can portray alternative outcomes. Taking the example above, if TGF-beta binds to the type III receptor then an alternative flow of events occurs for a time that then returns to the main flow of events. Other possible divergences that were modeled included whether to internalize the TGF-beta receptors via clathrin-dependent or lipid raft-dependent mechanisms. These pathways lead to either complex degradation or signal promotion. Because complex degradation is not specified in our use case, for simplicity, this event is routed to the final state. However, the main success scenario, signal promotion, continues until SMAD2 is released and TGF-beta complex assembly is finished. Together, the dynamic events occurring during the biological process, TGF-beta receptor complex assembly (GO:0007181) are captured. Class diagram generation ------------------------ The major components of a biological system are bioentities with functions and interactions. Likewise, the center of an object-oriented software system is objects. Complex bioentities formed from multiple gene products along with their relationships, are contained within the biological system encompassing the biological process term, TGF-beta receptor complex assembly (GO:0007181). To represent the components that execute the process, we captured these components as bioentities with functions, and their interactions. The events of the TGF-beta receptor complex assembly (GO:0007181) process include TGF-beta binding (GO:0050431) to its receptors, and SMAD binding (GO:0046332) and activation (GO:0042301). To capture this static architecture, class diagrams were generated that model the bioentities, operations, and interrelationships that occur between TGF-beta, its receptors, and Smad 2. Similarly to figure [1](#F1){ref-type="fig"}, figure [4](#F4){ref-type="fig"} captures the major components of the initial phases of TGF-beta signaling as objects with their associations, using an object-oriented representation. However, unlike figure [1](#F1){ref-type="fig"}, this object-oriented representation of the components of the main receptor complex is enhanced by the addition of attributes and functions. These objects were given attributes that describe important characteristics that if changed, might alter the function of a component. The functions of the objects, which parallel gene product functions, were generated from the sequence diagrams and were represented using Gene Ontology terms. These functions or operations are a declaration of the methods that an object may use. Together, the models generated using the described object-oriented methodology yield a software system representation of a biological process, TGF-beta receptor complex assembly, capturing both static and dynamic relationships annotated with Gene Ontology terms. In addition, the UML notation provides a mechanism to specify inheritance that may be used to indicate an object that is the foundation for other objects. For instance, a TGF-beta receptor object might be a generalization of the TGF-beta receptor I object (data not shown). These specific objects inherit the properties of the receptor object. In addition, binary associations containing cardinalities may indicate the number of objects interacting with another. For instance, TGF-beta can interact with one to many receptors, while a receptor can only interact with one TGF-beta at a time (Fig. [4](#F4){ref-type="fig"}). Cellular compartments where these gene products can be found are also shown. Here, guard conditions are added to distinguish conditions under which each gene product might be found in a particular cellular compartment. In this way, a spatial representation of the TGF-beta receptor complex components is also achieved. These class diagrams demonstrate that the static structure of a biological system can be represented as an object-oriented model with integrated Gene Ontology terms. Collectively, the models generated using the described object-oriented methodology yield a software system representation of a biological system, capturing both static and dynamic relationships annotated with integrated Gene Ontology terms. Discussion ========== We have utilized the Gene Ontology to construct an object-oriented representation of the initial steps of TGF-beta signaling, and the gene products contained therein. In doing so, we have provided a standardized framework for the integration of Gene Ontology terms into gene product descriptions. By capturing all of the relevant GO terms in one model, the disjointed GO vocabulary is assembled into a cohesive structure. This cohesive structure encompasses the fundamental concepts of the object-oriented paradigm. We proposed a solution to three unaddressed issues within the current Gene Ontology. First, while the Gene Ontology has helped to formalize the vocabulary that describes biological systems, it lacks a specific integration method. Currently, when applied to gene products, Gene Ontology terms are only categorically listed. Second, the Gene Ontology domains, biological process, molecular function and cellular component lack coherence. In particular, no association exists between domains. Finally, the current Gene Ontology defines GO terms, but gives no indication of what is necessary to accomplish a particular function, or process. To resolve these problems we defined an object-oriented methodology and architecture that provides a unifying framework to integrate all Gene Ontology domains. The central dogma of the object-oriented paradigm revolves around several key aspects. Specifically, an object-oriented framework should accommodate the class, object, inheritance, composition, encapsulation and polymorphism concepts. As shown in table [1](#T1){ref-type="table"}, gene products and other bioentities can be decomposed into objects, which are created based on template classes. These objects utilize inheritance to acquire the attributes and properties of more general objects. Complex classes can also be disassembled into subclasses using composition. Encapsulation allows the simplification of the model without sacrificing functionality. For instance, we do not need to know specific details regarding how a gene product is translated, just that a process that is encapsulated by the function \'translate()\' can create a protein. However, if we wished to delve deeper into the mechanics of the translation process the layered architecture of the object-oriented system would allow us to do so. It is also worth noting that the modular nature of the object-oriented system closely resembles the recently discovered modular structure of biological networks \[[@B46]-[@B48]\]. This resemblance further indicates that biological systems can be easily modeled as object-oriented systems. Finally, polymorphism allows one to describe shared functions among different gene products. In this way, a function that may be shared broadly with other gene products can be uniquely specified for a particular gene product. By applying object-oriented methodologies and concepts the various domains of the Gene Ontology can be coordinated into one model. Currently, the mechanisms in the biological process domain are veiled. There is no indication as to what gene products form the biological process, or what molecular functions are necessary to accomplish the process. Furthermore, the outcome of a specific process is not obvious. As in our example, a process such as TGF-beta receptor complex assembly (GO:0007181) does not give any indication of the components, dynamics or outcomes that occur during this process. However, by incorporating GO terms as attributes and functions we can discern relationships between the three domains. Likewise, the cellular components domain does not provide temporal or spatial clues when applied to gene products. For instance, GO terms \'extracellular\' and \'intracellular\' may both be associated with a particular gene product. However, the distinction between when a gene product is extracellular and when it is intracellular is not apparent. By applying object-oriented principles we can set extracellular and intracellular to Boolean values, and we can specify which location is the current (true) location of a gene product. In addition, by using object-oriented principles a GO molecular function term can be augmented with parameters and outcomes. For example, the function \"GO:0046982: protein heterodimerization activity\" has different input and output parameters depending on the particular protein that contains the function. This type of polymorphic behavior, where one function can be performed in multiple ways is not supported by the Gene Ontology. For example, protein A may heterodimerize with protein B, whereas protein C heterodimerizes with protein D. From the Gene Ontology it is not readily apparent as to what is being inputted into the dimerization function. However, by applying an object-oriented architecture to function \"GO:0046982: protein heterodimerization activity\" we get \"GO:0046982 (in: Protein A, in: Protein B): Protein AB\". This format is an improvement to the unparameterized GO term in that the function can be cross-referenced to protein heterodimerization activity via its GO term, and we also see that for protein A to heterodimerize we need both protein A and protein B. In addition, we now observe that a new entity called protein AB is created from this function. By capturing the above details in an object-oriented model the GO term becomes far more useful for both biologists and computer scientists. Using an object-oriented approach the Gene Ontology domains are integrated into one cohesive model. Integration of the Gene Ontology terms into an object-oriented representation offers several additional benefits. The object-oriented model provides additional levels of detail not found in the Gene Ontology. One of the strengths of object-oriented technology is the ability to capture the dynamics of a system. For example, sequence diagrams can chronologically order events in a biological process. Activity diagrams afford one the opportunity to envision different scenarios that might be occurring in a process. This additional level of detail significantly increases the depth of information that can be applied to the description of a biological process. State-transition diagrams also contribute to the realization of the full dynamics of a process by allowing the visualization of gene product states within a process. Furthermore, UML models can be translated into code, facilitating the creation of simulations. The standardization of biological system modeling and integration is growing rapidly. A widely accepted example of the drive toward standardization is the Systems Biology Markup Language (SBML) \[[@B49]\], which has been adopted by more than 70 software tools \[[@B50]\]. The Gene Ontology is another example. However, each of the technologies, the Gene Ontology, the object-oriented approach, and SBML, has strengths and weaknesses. The Gene Ontology provides a standardized vocabulary but contains disconnected domains with no details regarding terms. SBML was developed to communicate biological models, with an emphasis on mathematical modeling of biological systems, but does not specify how to construct these models. Object-oriented technologies, on the other hand, provide a well-defined process for model creation and visualization, but have not been standardized for biology. However, the Gene Ontology, object-oriented paradigm, and SBML can form a new synergism when jointly applied to a common biological system model. These technologies are steps toward a unified approach to biological information integration, and studying biological phenomena at the systems level. Together, this unified approach will make biological system integration and analysis consistent, manageable and controllable, which is essential in handling complex systems, as demonstrated by decades of software industry experience. While the described object-oriented approach can significantly enhance the annotation of gene products using the Gene Ontology, several challenges will need to be addressed. Specifically, object-orientation was not specifically designed for use in biological systems. Therefore, its use in capturing biological systems is not well defined. Furthermore, the Gene Ontology is still expanding and undergoing revisions. Consequently, in the near future it will still be necessary to do literature searches to define all the gene ontologies associated with a gene product. However, automated extraction of information for UML model generation and software implementation for simulations is under development, but is beyond the scope of this paper. Future systems may also be implemented as software libraries in object-oriented programming languages (C++ and Java) for computer scientists to construct software for various applications and can be distributed as part of the GO-DEV toolkit for Gene Ontology development \[[@B29]\]. In addition, reformatting gene products with Gene Ontology terms will require the cooperation of multiple groups of biologists and computer scientists. However, we must take into consideration that a primary issue with this approach is the lack of people with cross-disciplinary skills able to comprehend both the biology and the computer science. Nonetheless, our own experience has shown that with supervision one biologist without a formal computer science background can learn to model a biological system using UML in a matter of months. Furthermore, automation of some of the annotation process will significantly reduce the human effort, but not eliminate the need for human annotators. Additional standards for automation will also need to be developed to thoroughly specify the process of object-oriented biological system integration. Despite these challenges the ultimate goal of creating a library of UML objects or modules integrated with Gene Ontology attributes and functions is worthwhile. Through this endeavor, biological processes could be assembled from these libraries for the development of simulation tools that will increase the productivity of biologists through increased insight into disease pathways and mechanisms. Conclusion ========== Here, we have demonstrated that Gene Ontology terms can be integrated into an object-oriented model. Furthermore, the object-oriented technology and methodologies used for this integration should improve the usability of these terms, and increase the depth of information that they contain. This work also serves as a framework for reverse-engineering biological gene products as objects in an object-oriented system. Together, this should facilitate additional collaborations between biologists and computer scientists. Methods ======= UML representations of the TGF-beta receptor complex assembly process were created following a software engineering process consisting of phases of requirement-gathering, analysis and design. UML models were generated using Microsoft Visio Pro. AmiGO \[[@B51]\] was used to determine Gene Ontology links for TGF-beta gene products. Requirement-gathering phase --------------------------- ### Information collection To define the requirements and collect the information necessary for the generation of the models, two approaches were necessary. First, annotations of the TGF-beta signaling pathways were conducted during an extensive literature review. Second, gene ontologies and Uniprot entries were searched to assign Gene Ontology terms to gene products. The attributes and the interactions of the TGF-beta signaling components were captured using class-responsibility collaboration (CRC) cards as described previously \[[@B52]\] \[see Additional files [1](#S1){ref-type="supplementary-material"}, [2](#S2){ref-type="supplementary-material"}, [3](#S3){ref-type="supplementary-material"}, [4](#S4){ref-type="supplementary-material"}\]. ### Use case development Based on the gathered information, best-case and alternative scenarios were developed within a so-called \"use case\" to describe the TGF-beta receptor complex assembly process \[see [Additional file 5](#S5){ref-type="supplementary-material"}\]. The use case also serves to define the boundary and scope of the TGF-beta model. For demonstration purposes the boundary of the system was limited to the steps TGF-beta receptor complex assembly. Therefore, alternative events such as receptor ubiquitination and degradation, as well as the specifics of SMAD 2 mobility were not captured in the dynamic models (i.e. sequence diagram). Analysis phase -------------- ### Conceptual model generation To provide an overview of the system and its interrelationships a conceptual based on the information defined in the requirement-gathering phase was generated \[see [Additional file 6](#S6){ref-type="supplementary-material"}\]. This conceptual model integrated biological information, and represented TGF-beta and the cellular components involved in the complex assembly and their relationships in UML notation. By applying object-oriented analysis, the TGF-beta receptor complex assembly was decomposed into objects and component relationships were realized. However, information regarding component properties is hidden through encapsulation. This conceptual model defines the organization of the biological system and provides an overview of the components and their relationships. Design phase ------------ ### State diagram generation The dynamics of the system can also be captured using state diagrams, which can be used to describe the transitions and different states that a cellular component can exist \[see [Additional file 7](#S7){ref-type="supplementary-material"}\]. In addition, multiple concurrent states can be illustrated using this UML notation. ### Sequence, activity, class diagram generation Sequence, activity and class diagrams have been used as an example to demonstrate the feasibility of generating an object-oriented representation of the biological process described by the GO term TGF-beta receptor complex assembly (GO:0007181), with Gene Ontology terms applied to generate these diagrams. Objects representing corresponding gene products are created, and their essential attributes are captured. Interactions among objects are also identified. For each interaction, a corresponding method is generated. This method is matched to a Gene Ontology term. The nature of the interaction determines the method parameters. The sequence of events is captured, and used to generate sequence diagrams. Scenarios are also generated for object interactions, and used to generate activity diagrams. The information captured in the sequence diagram and activity diagrams are used, along with the gene products attributes, to generate class diagrams. Authors\' contributions ======================= DS drafted the manuscript, constructed the models and participated in the design of the study. WJZ was the principal investigator, conceived of the project and guided its development. All authors read and approved the final manuscript. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Class-responsibility-collaboration card for TGF-beta 1. Attributes, collaborators and responsibilities of the specified protein are given. The attributes section allows the ordered listing of information not easily captured by the UML notation. The collaborator section lists the cellular components that interact with TGF-beta 1. The responsibilities section specifies the consequence of TGF-beta 1 interacting with its collaborator. This card allows TGF-beta 1 to be decomposed into an object containing attributes, operations and interactions. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 2 Class-responsibility-collaboration card for TGF-beta receptor II. Attributes, collaborators and responsibilities of the specified protein are given. The attributes section allows the ordered listing of information not easily captured by the UML notation. The collaborator section lists the cellular components that interact with TGF-beta receptor II. The responsibilities section specifies the consequence of TGF-beta receptor II interacting with its collaborator. This card allows TGF-beta receptor II to be decomposed into an object containing attributes, operations and interactions. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 3 Class-responsibility-collaboration card for TGF-beta receptor I. Attributes, collaborators and responsibilities of the specified protein are given. The attributes section allows the ordered listing of information not easily captured by the UML notation. The collaborator section lists the cellular components that interact with TGF-beta receptor I. The responsibilities section specifies the consequence of TGF-beta receptor I interacting with its collaborator. This card allows TGF-beta receptor I to be decomposed into an object containing attributes, operations and interactions. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 4 Class-responsibility-collaboration card for SMAD2. Attributes, collaborators and responsibilities of the specified protein are given. The attributes section allows the ordered listing of information not easily captured by the UML notation. The collaborator section lists the cellular components that interact with SMAD2. The responsibilities section specifies the consequence of SMAD2 interacting with its collaborator. This card allows SMAD2 to be decomposed into an object containing attributes, operations and interactions. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 5 Use case describing the events leading to TGF-beta receptor complex assembly. This use case defines the boundaries of the system model. Here, the main success and alternative scenarios leading to the assembly of the TGF-beta receptor complex are described. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 6 Conceptual diagram of the TGF-beta receptor complex, and the proteins that associate with this complex. Gene products have been decomposed into objects. Object attributes and operations are hidden to reduce complexity. Gene products that comprise the receptor complex are shown in blue with their associated relationships in green. As the boundaries of the use case do not include them, other associated proteins that comprise the TGF-beta signaling pathway are not emphasized in further diagrams. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 7 State diagrams for the TGF-beta receptor complex components. State diagrams describe the possible states in which a bioentity can exist. Concurrent states are separated by vertical dashed lines. A)State diagram for TGF-beta. B)State diagram for TGF-beta receptor II. C)State diagram for TGF-beta receptor I. D)State diagram for SMAD2. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 8 An example of a class diagram with expanded function names showing the interactions between the components of the TGF-beta receptor complex (GO:0007181) (grayed). Cellular components containing these gene products are also shown. GO function terms are shown with their corresponding GO ids. This format demonstrates a more user-friendly interface for reading the diagrams, whereas figure [4](#F4){ref-type="fig"} is more suitable as a computer readable format. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ Daniel Shegogue is supported by NLM training grant 5-T15-LM007438-02. W. Jim Zheng is partly supported by a grant (DE-FG02-01ER63121) from the Department of Energy. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### The GO terms associated with the process, TGF-beta receptor complex assembly (GO: GO:0007181). Lines with solid diamonds (![](1471-2105-6-113-i1.gif)) at the end indicate composition. These are read from the diamond end, for example, as \'cell\' (GO:0005623) contains a \'membrane\' (GO:0016020). Lines with open triangles (![](1471-2105-6-113-i2.gif)) represent generalizations. These are read from the triangle end, as a \'membrane\' (GO:0016020) is a general type of \'plasma membrane\' (GO:0005886). A)Directed acyclic graph for the cellular component GO terms associated with TGF-beta receptor complex assembly (GO:0007181). B)Object-oriented representation of the DAG described in Figure 1A (white). Additional cellular components not represented by the current Gene Ontology, but essential to the TGF-beta receptor complex assembly process are shown in the gray boxes (i.e. TGF-beta, TGF-beta receptors, SMAD2) ::: ![](1471-2105-6-113-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### An example of the sequence diagram showing the TGF-beta receptor complex assembly (GO:0007181). The binding of TGF-beta to its receptor (GO:0050431), receptor heterotetramerization (RI and RII homodimers, heterodimerizing)(GO:0046982), translation (GO:0043037), transferase activity (GO:0016740), and Smad 2 binding (GO:0046332) and activation (GO:0042301) are shown. ::: ![](1471-2105-6-113-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### An example of an activity diagram showing the main and alternative flow of events occurring during TGF-beta receptor complex assembly (GO:0007181). ::: ![](1471-2105-6-113-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### An example of a class diagram showing the interactions between the components of the TGF-beta receptor complex (GO:0007181) (grayed). Cellular components containing these gene products are also shown. Due to space constraints functions are named using their corresponding GO ids. The class diagram was also constructed with a descriptive function name \[see [Additional file 8](#S8){ref-type="supplementary-material"}\]. The GOid reference list is: GO:0042803, protein homodimerization activity; GO:0005160, TGF beta receptor binding; GO:0046982, protein heterodimerization activity, GO:0050431, TGF beta binding, GO:0005524, ATP binding; GO:0016740, transferase activity, GO:0046332, Smad binding; GO:0042301, phosphate binding; GO:0003677, DNA binding. ::: ![](1471-2105-6-113-4) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### The use of object-oriented concepts in the integration of the Gene Ontology into an object-oriented model. Object-oriented terms, their definitions, and corresponding mechanisms of incorporating GO terms into an object-oriented model are shown. A specific example from the manuscript is also given. GO, Gene Ontology; DAG, directed acyclic graph; OOM, object-oriented model ::: Object-Oriented Term Object-Oriented Definition \* Object-Oriented use of the GO Example -------------------------- -------------------------------------------------------------------------------------------------------------------------------------------------- ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Class A class is a template from which object instances are created. It specifies the common characteristics that objects created from it will contain Classes are created from gene products whose characteristics are defined by the GO molecular function and cellular component terms The class Smad 2 is created based on the properties of the gene product Smad 2, which are defined by molecular functions such as \"protein homodimerization\' (GO:0042803) and \'ATP binding\' (GO:0042301) Object An instance of a class that contains unique properties Objects are created from the template classes, but may contain properties unique to a particular object Two different Smad 2 objects may be created, one of which is phosphorylated, and one which is not Inheritance Relationships between classes, whereby a more specific class inherits all the properties and methods of the classes they belong to Relationships defined by \'is a\' are generalizations in which child classes of the DAG inherit the properties of the parent class (if a child class has multiple parent classes, multiple inheritance applies) The cellular component \'plasma membrane\' (GO:0005886) inherits the properties of the general class cellular component \'membrane\' (GO:0016020) Composition Certain objects may be assembled from collections of other objects \'part\_of\' relationships defined in the GO DAG are rendered as composition relationships in an OOM The \'membrane\' (GO:0005623) and \'intracellular\' (GO:0005622) space are part of the \'cell\' (GO:0005623) Polymorphism The ability of an object to interpret messages differently when received by different objects GO functions may change for different proteins and be given different input and output values The function \'protein homodimerization activity\' (GO:0042803) in the context of SMAD2 accepts two SMAD2s and outputs a dimerized SMAD2, whereas in the context of TGF-beta receptor II it accepts two receptors and outputs a dimerized receptor Encapsulation Hiding the state and implementation of an object The exact mechanism by which an object is created is hidden in an OOM The details involved in the translation (GO: 0043037) of Smad 2 are hidden, but a Smad 2 molecule is still created \\[[@B53]\] ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### The gene product functions described herein are listed with their associated GO molecular functions and parameters. These gene product functions are mapped to corresponding Gene Ontology molecular functions. These GO functions are integrated into an object-oriented model by amending them with input and output parameters, thereby creating object functions. ::: -------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Gene Product* Gene Product Function Corresponding GO Term and GO ID Input Output Figure Location ------------------ --------------------------- -------------------------------------------------- ------------------ ------------------------------- --------------------- TGF-beta Dimerize protein homodimerization activity (GO:0042803) 2X TGF-beta Dimerized TGFβ Fig.4 bind TGF-beta receptor TGF-beta receptor binding (GO:0005160) TGFβ homodimer\ TGFβ-TGFβR complex Fig.4 TGFβR homodimer TβRII Dimerize protein homodimerization activity (GO:0042803) 2X TβRII Dimerized RII Fig.4 TGF-beta binding TGF-beta binding (GO:0060431) TβRII homodimer\ TGFβ-TβRII heterotetramer Fig.2, 4 TGFβ homodimer Heterotetramerize protein heterodimerization activity (GO:0046982) TβRI homodimer\ TβRI-TβRII heterotetramer Fig.2, 4 TβRII homodimer phosphorylate RI transferase activity (GO:0016740) ATP\ phosphorylated TβRI Fig.2, 4 TβRI homodimer TβRI Dimerize protein homodimerization activity (GO:0042803) 2X TβRI Dimerized RI Fig.4 Heterotetramerize protein heterodimerization activity (GO:0046982) TβRI homodimer\ TβRI-TβRII heterodimer Fig.2, 4 TβRII homodimer TβRI activation phosphate binding (GO:0042301) ATP\ phosphorylated TβRI Fig.4 TβRI bind SMAD2 Smad binding (GO:0046332) SMAD2 TGFβ-TβRII-TβRI-SMAD2 complex Fig.2, 4 phosphorylate Smad transferase activity (GO:0016740) ATP\ phosphorylated SMAD2 Fig.2, 4 SMAD2 SMAD2 bind TβRI TGF-beta receptor binding (GO:0005160) TGFβ homodimer\ TGFβ-TβRII-TβRI-SMAD2 complex Fig.4 TβRI homodimer SMAD2 activation phosphate binding (GO:0042301) ATP\ phosphorylated SMAD2 Fig.4 SMAD2 Trimerize protein heterodimerization activity (GO:0042803) SMAD2\ Trimerized SMAD2 Fig.4 SMAD2 homodimer activate transcription DNA binding (GO:0003677) DNA\ SMAD2-DNA complex Fig.4 SMAD2 -------------------------------------------------------------------------------------------------------------------------------------------------------------------------- :::	PubMed Central	2024-06-05T03:55:57.602617	2005-5-10	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1156866/", "journal": "BMC Bioinformatics. 2005 May 10; 6:113", "authors": [ { "first": "Daniel", "last": "Shegogue" }, { "first": "W Jim", "last": "Zheng" } ] }
PMC1156867	Background ========== Recent high-throughput and whole-genome experimental methods create new challenges in data analysis and visualization. Gene expression and protein microarrays output hundreds of thousands of data points that can be used for prediction of gene function over the entire genome. However, there are serious and fundamental challenges in the analysis of these data. Microarray data contain substantial experimental noise and as our knowledge of biology is incomplete, no perfect gold standard exists for verification of microarray analysis methods. In order to determine gene/protein relationships and functions from microarray data, methods must be robust to noise and must identify groups of genes that may be functionally related. Statistical methods, such as clustering, attempt to identify data patterns and group genes together based on various distance metrics and algorithms. The lack of a true gold standard makes it impossible to verify the absolute accuracy of any clustering method. Several statistical approaches have been presented for assessing cluster quality \[[@B1]-[@B4]\], but these are all either internal validation methods or methods that rely on incomplete external standards such as MIPS \[[@B5]\] or Gene Ontology \[[@B6]\] functional protein classifications. Further, these methods do not address the issue of identifying specific problems within clusters of microarray profiles or assessing the relationships between clusters of genes. Well designed visualization methods are capable of aiding in these tasks by helping to bridge the gap between raw data and the analysis of that data \[[@B7]\]. To perform more comprehensive cluster analysis, statistically integrative, dynamic, noise-robust data visualizations are required to complement purely analytical evaluation methods. Existing visualization tools do not include methods to statistically and dynamically evaluate clusterings of genes. Several tools can display expression data in various static ways suitable for publication \[[@B8]\] or provide useful dynamic views of tabular data \[[@B9]\], but are not specifically intended for cluster analysis. JavaTreeView \[[@B10]\] and the HierarchicalClusteringExplorer \[[@B11]\] dynamically display hierarchically clustered data for analysis and VxInsight \[[@B12]\] displays the result of a built-in clustering algorithm in an interactive 3D topology, but none are able to display results of other clustering methods for analysis. TreeMap \[[@B13]\] provides an innovative way to visualize hierarchically clustered data as well as data organized in the context of the GO hierarchy, but is not intended for cluster analysis. New tools such as GeneXplorer \[[@B14]\] provide an interactive method for visualization and analysis of microarray data on websites, but do not focus on the task of cluster analysis. Several tools, including the MultiExperimentViewer \[[@B15]\] and Genesis \[[@B16]\], provide multiple methods of performing clustering as well as some visualization methods to analyze the resulting clusters. Commercial tools, such as GeneSpring \[[@B17]\] and SpotFire \[[@B18]\], offer various statistical and visualization tools for general analysis, but neither offer visual methods specific to analyzing the results of clustering algorithms. Therefore, there is a need for a visualization-based methodology designed specifically to statistically and dynamically evaluate clusters produced by the variety of available algorithms and software tools. Here we present a suite of interactive microarray analysis methods that integrate relevant statistical information into visualizations for the purpose of assessing the quality and relationships of clusters in a noise-robust fashion. Our methodology is general and can be used to analyze the results of most clustering algorithms performed on either protein or gene expression microarray datasets. Results and discussion ====================== Noise robust visualization -------------------------- Microarray data contain a substantial amount of noise; therefore, visualizations must facilitate tasks like pattern identification and outlier detection in a noise-robust fashion. Microarray data span a rather large and noisy numerical range, so traditional microarray visualizations use a cutoff value that specifies where maximum saturation occurs. While this is necessary in order to see variation around zero, it obscures variation in highly over or under expressed areas (Fig. [1a--c](#F1){ref-type="fig"}). At a minimum this cutoff value should be dynamically controlled by the user so that they have the ability to see both types of variation. Several currently available tools include this ability, as does our method, but while the ability to change the cutoff value helps to increase dynamic range and decrease the effects of noise in visualizations, it fails to address the entire problem. Traditional visualization methods essentially display the Euclidean distance between gene expression profiles, a measure that is not robust to outliers. Distance metrics more robust to noise, such as a rank-based Spearman correlation coefficient, can be used for numerical analysis of microarray data. We propose a rank-based visualization method to serve as the complement to these noise robust distance metrics (Fig. [1d](#F1){ref-type="fig"}). Our method performs a rank transform on each gene by sorting the gene\'s expression levels, then ranking the experiment for each gene with the lowest expression 0, the next lowest 1, and so on to the highest expression which is ranked N-1, where Nis the number of experiments. Each experiment is then displayed as a grayscale percentage of rank/(N-1). In this display, the experiment with lowest expression for each gene is colored black, the experiment with the highest expression is colored white, and the intermediate experiments gradate between them in shades of gray. In addition to being more robust to noise, this rank-based visualization allows users to easily see patterns of shape/trend that are not apparent in traditional visualizations. Clustering algorithms that use a rank-based distance metric will group together genes based on their pattern of expression which can result in clusters that look very nonuniform when traditionally displayed (Fig. [2](#F2){ref-type="fig"}). However, in our rank-based visualization it is clear that these genes do belong together because they share expression profiles with the same shape/trend. While the example in Fig. [2](#F2){ref-type="fig"} is an extreme case, this rank-based visualization approach is useful in a variety of biological settings. For example, in many time series data sets it is useful to observe changes in expression over time in response to some process such as environmental changes, drug introduction, or cell cycle phase. In particular, a group of genes which all rise in expression over a period of samples in a cell cycle experiment, but whose absolute expression levels are not the same will appear heterogeneous when displayed traditionally. However, when displayed using our rank-based method, the pattern of expression is much clearer, which can aid users to identify biologically meaningful trends of expression (Fig. [3](#F3){ref-type="fig"}). Genes exhibiting a coherent progression of shape/trend over time may be co-regulated. Thus, it is important to identify trends and not just examine similarities of absolute expression level. Assessing cluster quality ------------------------- While multiple statistical methods have been developed for assessing the quality of clusters produced by different algorithms \[[@B1],[@B3],[@B4]\] the most appropriate clustering algorithm choice depends on the dataset, distance metric, and goal of the analysis \[[@B2]\]. Due to the limitations of these methods, it is important to effectively display clustered data in a manner that allows researchers to examine the variation and consistency of the results of different clustering algorithms. We propose two new visualization techniques that can be used to assess overall cluster quality, and also identify individual outliers and other anomalies in the data quickly and efficiently. First, to analyze the overall cohesion of each cluster, we developed a \"difference display\" method. For each cluster, we display the cluster average bar to show the general expression of the cluster as a whole. We calculate the vector of the cluster average ![](1471-2105-6-115-i1.gif) from the vectors of expression profiles of each gene, ![](1471-2105-6-115-i2.gif), for each cluster containing Mgenes with expressions measured over Nexperiments using the standard formula: ![](1471-2105-6-115-i3.gif) Each gene\'s expression is displayed as a difference, ![](1471-2105-6-115-i4.gif), from the cluster average, ![](1471-2105-6-115-i1.gif): ![](1471-2105-6-115-i5.gif) Thus if a gene is shaded green in an experiment, it is expressed lower than the cluster average for this experiment, and if shaded red it is expressed more in an experiment than the cluster average for that experiment. In this visualization a cluster that is relatively dark is more uniform since the genes are generally close to the average (Fig. [4a](#F4){ref-type="fig"}). Individual genes that differ from the average more than others will stand out as brighter than their neighbors, which allows for easy visual detection of outliers (Fig. [4b](#F4){ref-type="fig"}). Thus, this visualization allows researcher to easily identify genes that do not fit well with the cluster\'s expression profile, and thus may be functionally distinct from the rest of the cluster. Second, in addition to assessing overall cluster quality and identifying gene outliers, it is important to look at variation of individual experiments within each cluster. We calculate the standard deviation, s, of each experiment, j, within a cluster in the normal manner: ![](1471-2105-6-115-i6.gif) Where Mis the number of genes in the cluster, ![](1471-2105-6-115-i7.gif) is the cluster average for experiment j, and g~i,\ j~is the expression level of gene iin experiment j. We display the standard deviation of each experiment within the cluster below the cluster average bar. Here black indicates a standard deviation of zero and white indicates higher standard deviations, saturating at a user defined cutoff value. This allows a user to quickly identify high and low variation experiments on a per-cluster basis (Fig. [5](#F5){ref-type="fig"}). High variation experiments may imply that the genes in this cluster were less related under those particular experimental conditions. Visualizing clusters in this difference display method allows users to see variations in expression level that may be biologically significant that are not visible in traditional visualization methods. For example, the data shown in Fig. [5](#F5){ref-type="fig"} is the glycolysis cluster (2E) from \[[@B19]\]. When viewed traditionally this cluster appears very homogenous and consistent. However, when viewed as a difference from the cluster average, we can observe that in the region of highly under-expressed experiments some genes are more expressed than the average while others are less expressed than average (red and green boxes are shown in this area). This suggests that the cluster could be split into two smaller clusters that would be even more homogenous. In this example 8 of the 9 genes indicated by the red box, but only 3 of the 8 genes indicated by the green box are annotated to glycolysis. The genes in the green box are better categorized as more generally related to alcohol metabolism than to glycolysis in particular (see web supplement to Fig. [5](#F5){ref-type="fig"} for details, located at <http://function.princeton.edu/GeneVAnD>). Traditional visualization is unable to show this type of biologically meaningful variation in highly over or under expressed regions. Assessing cluster relationships ------------------------------- In addition to assessing the quality of clusters produced by an algorithm, it is also important to understand how the clusters and genes in different clusters relate to each other. Clusters with similar overall expression profiles may functionally interact with one another. One method to show high level cluster-to-cluster relationships is to calculate a hierarchical clustering using only the averages of each cluster. We can then hierarchically arrange the cluster averages and display the dendrogram relating the averages to each other (Fig. [6](#F6){ref-type="fig"}). As this method only creates a hierarchy for the cluster averages, rather than for individual genes as in the case of hierarchical clustering of the entire dataset, it allows us to show cluster relationships for arbitrary clustering algorithms. However, this dendrogram of averages fails to show the relationships between genes in different clusters. It is important to examine gene-to-gene and gene-to-cluster relationships to assess whether or not genes are included in the most appropriate cluster. In order to view the lower level relationships among genes in clusters we can project high dimensional microarray data into a lower dimensional space such that genes with similar expression profiles are spatially closer to each other than genes with different expression profiles. We use Principal Component Analysis (PCA) to define the axes of a three-dimensional space to project the genes and clusters onto. PCA has been used previously in microarray data analysis for dimensionality reduction to facilitate easier analysis and comparisons \[[@B4],[@B20]\] and to identify patterns of noise \[[@B21]\]. Our method is interactive and navigable which allows users to examine individual genes and view relationships between clusters as they separate out spatially. To perform PCA on the microarray datasets, we use Singular Value Decomposition (SVD). SVD decomposes an m× nmatrix of the full microarray data, X, into three additional matrices: ![](1471-2105-6-115-i8.gif) Where Mis the number of genes and corresponds to rows of the matrix, and Nin the number of experimental conditions and corresponds to the columns of the matrix. We use the eigengenes, or Principal Compenents (PCs), defined in the rows of V^T^as the axes for our PCA visualization. The position of each gene in that space is determined by the corresponding column of UΣ. The square of the singular values, contained on the diagonal of Σ, correspond to the variance included by each PC such that the percent of variation, p, captured by the k^th^PC is determined by: ![](1471-2105-6-115-i9.gif) In this formulation, the singular values are in decreasing order, meaning that the first PC includes more variation than the second, and so on. Thus, using the top 3 PCs includes the most variation possible in a three dimensional projection. We would expect that well-formed clusters would separate out the most when using the top PCs as the axes of projection. However, in some data sets the top PCs are not the most appropriate space for projection. For example, in the Spellman et al. cell cycle data set \[[@B22]\] using our tool we can see that the first PC does not show the \"banded\" pattern typical of ordered cell cycle data, which the second, third, and fourth PCs do display (Fig. [7a](#F7){ref-type="fig"}). Accordingly, a projection into the first two PCs does not separate out cell cycle regulated genes/clusters spatially (Fig. [7b](#F7){ref-type="fig"}). This is consistent with previous PCA analysis done by Alter et al. \[[@B21]\] which identified the first PC of this data as highly correlated to noise rather than meaningful information. Our method allows the user to dynamically specify which PCs define each axis, which allows exploration of which PCs are most appropriate for analysis and identification of potential noise-correlated patterns in the data. In the case of Spellman et al. cell cycle data, we can use the 2^nd^, 3^rd^, and 4^th^PCs for projection, which leads to much better spatial separation (Fig. [7c](#F7){ref-type="fig"}). In this projection, we can see that each phase of the cell cycle spatially separates in temporal order around the origin and that the G1 and M phases appear opposite each other, which is consistent with the underlying patterns of expression for cell cycle genes. Our projection of genes and clusters into a space defined by user selected PCs allows the user to view and analyze relationships on both a cluster-to-cluster basis and a gene-to-gene basis. Multiple simultaneous views and scaleable architecture ------------------------------------------------------ In our system each of the visualizations described above are dynamically linked to each other, so that selections, colorations, etc. are shared among views. This allows users to perform tasks in conjunction with one another. For example, using the difference image visualization and the PC projection, users can assess the quality of a clustering as well as the relationship between clusters very easily (Fig. [8](#F8){ref-type="fig"}). Our implementation of these methods is both modular and scalable. Although all of the visualizations share a common data structure for dynamic linking, each visualization is displayed in its own panel, allowing for easy addition or removal of new visualization components. Each of the panels is fully scalable for use on both desktop/laptop size displays as well as large display walls. The ability to use these visualizations on large, high-resolution displays facilitates collaboration among researchers and allows users to view greater portions of their datasets simultaneously (Fig. [9](#F9){ref-type="fig"}). Conclusion ========== Statistical clustering of microarray data is vital for identifying groups of genes that may be functionally related. However the high level of noise in microarray data and the lack of a gold-standard for comparison deeply complicate the evaluation of clustering algorithms. Here we have presented a set of visualization methods geared specifically toward evaluating clustering of microarray datasets. Our rank-based method allows for more noise-robust visualizations of expression levels, our difference display method facilitates visual assessments of general cluster quality as well as outlier detection, and our PC projection method allows for visual assessments of cluster relationships. Our methodology integrates meaningful statistics into an interactive and noise-robust data visualization package for use in analyzing the results of clustering algorithms. Through several examples we have demonstrated the effectiveness of these methods to aid researchers in the analysis of the results of clustering algorithms by facilitating noise-robust assessments of cluster quality and cluster relationships. We believe that more statistically integrative and targeted visualization methods can benefit not only cluster analysis, but many other important data analysis problems in genomics. Implementation ============== Our methodology has been implemented in GeneVAnD (Genomic Visual Analysis of Datasets). GeneVAnD is written in Java and is cross platform for use on Windows, Linux/Unix, and Macintosh operating systems. We use Java3D \[[@B23]\] to display the PC projections and Piccolo \[[@B24]\] to display the expression profiles. The JAva MAtrix Library (JAMA) \[[@B25]\] is used to perform the SVD calculation. The GeneVAnD package is designed in a modular way to allow future extensions and inclusion of additional information and visualizations. The executables and source code of GeneVAnD can be found at <http://function.princeton.edu/GeneVAnD>. Authors\' contributions ======================= MAH and NCD originally conceived the visualization techniques presented and were responsible for initial implementations. MAH created the final implementation of GeneVAnD and drafted the manuscript. KL provided advice and aided in the scalability of the methods to large scale displays and helped draft the manuscript. OGT provided advice and opinions key to the development of the methods and helped draft the manuscript. All authors read and approved the final manuscript. Acknowledgements ================ This work was funded in part by NSF grants EIA-0101247 and CNS-0406415 and by the Program in Integrative Information, Computer and Application Sciences (PICASso) which is funded by NSF grant DGE-9972930. We wish to thank Chad Myers and Grant Wallace for their help and support of this work. We also thank the Botstein laboratory members for their feedback on the early implementations. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Example of noise in microarray visualization. Four views of the same data displayed in different ways. (a-c) show a traditional display using different cutoff values. Note that in (a) variation in the highly over and under expressed regions cannot be seen due to saturation, while in (c) variation in the highly expressed regions can be seen, but variation near zero cannot. (d) uses our rank-based visualization method. In this rank-based view (d), the experiment with the lowest expression for each gene is colored black, the experiment with the highest expression is colored white, and the other experiments interpolate between in grayscale. Using this method, users can see the overall pattern of variation in the data, which makes it clear that heterogeneity in the traditional view is mostly the result of noise. (Data from \[26\]) ::: ![](1471-2105-6-115-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Rank-based visualization of synthetic data. Synthetic data displayed (a) traditionally and (b) using our rank-based method. This data was generated by creating a single sinusoidal expression profile and for each gene (row) randomly shifting that profile up or down and introducing small amounts of Gaussian random noise throughout. The result is that the genes generally follow the same shape/trend over experiments, but the shapes are shifted up/down from one another. Traditional view (a) masks the similarity between genes, but their relationship is clear in the rank-based view (b). ::: ![](1471-2105-6-115-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Rank-based visualization of time series data. Yeast cell cycle data displayed (a) traditionally and (b) using our rank-based method. In the traditional visualization the top 4 genes (within the purple box) appear to be very different from the rest of the genes in this cluster. However, using the rank-based method it becomes clear that these genes follow the same general pattern of the entire cluster, with initially low expression building up to highest expression in the central time points and then falling to roughly middle values. (Data from \[22\]) ::: ![](1471-2105-6-115-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Difference display visualization. Three clusters displayed traditionally on the left and in our difference image visualization on the right. In the difference display, the large top bar on each cluster shows the cluster average, each gene is displayed as its difference from that average (green indicates expressed less than the cluster average, red shows more expressed, and black means equally expressed with the cluster average). Cluster (a) is a coherent cluster of genes and appears very dark because of its homogeneity. Cluster (b) is another dark, uniform cluster, but it also contains one randomly inserted gene, which can be easily identified in our difference display. Cluster (c) contains a random selection of genes, and its randomness is clear from the brightness of the difference display. This difference display allows for quick assessment of overall cluster homogeneity and facilitates quick outlier detection. (Data and clusters a & b from \[19\]) ::: ![](1471-2105-6-115-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Experiment variation display. A cluster displayed traditionally on the left and in our difference image visualization on the right also showing the standard deviation within the cluster for each experiment. Black on the standard deviation bar indicates a standard deviation of zero, while white indicates a higher value. Purple arrows point to several experiments in this cluster that show high variance. In general, the high variance among some experiments may indicate that this cluster is unregulated under those conditions. In this example, we can inspect the differences from the cluster average in the high variance experiments and see that for these conditions the upper group of genes (indicated by a red box) is less under expressed than the lower group of genes (indicated by a green box) which suggests that the cluster could be split into two sub-clusters to reduce this variation. The biological function of these genes is consistent with such a split (see web supplement for details, <http://function.princeton.edu/GeneVAnD>. Data and cluster from \[19\]) ::: ![](1471-2105-6-115-5) ::: ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Dendrogram of averages. A dendrogram created from cluster averages with the genes in a cluster displayed below each average. The length of each branch of the tree is proportional to the distance between the averages. We create the hierarchy from the cluster averages, which allows us to show high level relationships between clusters generated by arbitrary clustering algorithms. (Data and clusters from \[19\]) ::: ![](1471-2105-6-115-6) ::: ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Principal component projection visualization. A projection of genes from a cell cycle data set into a 3D space defined by user selected Principal Components. Genes in each cluster are colored by phase (Red-G1, Green-S, Blue-G2, Yellow-M, and Cyan-M/G1). Cluster averages are displayed by larger solid spheres. The much larger transparent spheres show the region included by one standard deviation away from the average. (a) shows the top ten PCs of this data set and the percent of variance accounted for by each PC. (b) is a projection of cell cycle genes onto a space defined by the 1^st^and 2^nd^PCs. The separation is poor due to the first PC being highly correlated to noise in this data set. (c) shows the same data projected into a space defined by the 2^nd^, 3^rd^, and 4^th^PCs. These PCs are highlighted in (a) corresponding to the axis colors in (c). Notice that the cell cycle phases are separated in order around the origin, and that G1 and M phase genes are opposite each other, which is consistent with their opposing expression profiles. (Data and clusters from \[22\]). ::: ![](1471-2105-6-115-7) ::: ::: {#F8 .fig} Figure 8 ::: {.caption} ###### Multiple simultaneous views. A screenshot of GeneVAnD displaying clustered data. The panels shown are the expression level window on the left which can toggle between traditional, difference, and rank-based displays, and the PC projection window on the right. A selected gene is highlighted in blue in all views. ::: ![](1471-2105-6-115-8) ::: ::: {#F9 .fig} Figure 9 ::: {.caption} ###### Large scale display. GeneVAnD in use on a large-scale display wall. The high resolution enables display of more information simultaneously and the large scale creates an environment conducive for collaboration between multiple researchers. ::: ![](1471-2105-6-115-9) :::	PubMed Central	2024-06-05T03:55:57.607163	2005-5-12	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1156867/", "journal": "BMC Bioinformatics. 2005 May 12; 6:115", "authors": [ { "first": "Matthew A", "last": "Hibbs" }, { "first": "Nathaniel C", "last": "Dirksen" }, { "first": "Kai", "last": "Li" }, { "first": "Olga G", "last": "Troyanskaya" } ] }
PMC1156868	Background ========== Recent studies have found that many complex networks, ranging from the World-Wide Web \[[@B1]\] and the scientific collaboration network \[[@B2]\] to biological systems such as the yeast protein interaction network \[[@B3]\], are scale-free. The scale-free property states that the distribution of the connectivity k(number of links per node) in a network can be described by the power law, i.e., P(k) = ck^-γ^, c\> 0, γ\> 0. (1) A visual diagnosis of the scale-free behavior can be made through the log-log plot of the connectivity distribution, in which a straight line with slope -γis expected. In scale-free networks, the nodes are not randomly or evenly connected with some highly connected nodes (\"hubs\"). The ratio of the number of \"hubs\" to that of nodes in the rest of the network remains constant as the network changes in size. One attractive feature is that scale-free networks are more resistant to random failures compared with random networks due to the existence of a few highly connected \"hubs\" \[[@B4]\]. Remarkably, it has been observed that the scale parameter γvaried only in the narrow range of 2.1 -- 4 in the aforementioned real-world networks. All existing studies on scale-free networks assumed that the observed links represented the underlying structure of the network, but paid little attention to the fact that the observed links often involved errors, namely, false positives and false negatives. For example, Jeong et al. \[[@B3]\] considered the Saccharomyces cerevisiaeprotein interaction network inferred from yeast two-hybrid (Y2H) experiments. It is well-known that the Y2H system has many false positives as well as false negatives \[[@B5]\]. A natural question to ask is whether a scale-free network is still observed as scale-free in the presence of errors. And if it is, what are the possible underlying error mechanisms and how variable is the observed scale parameter γ? Answering these questions may lead to further insight to the scale-free property, better understanding and correct usage of the observed network data. For convenience, we will call networks observed with erroneous links as perturbed networks in the rest of this article. Results ======= In this article, we address the above questions by both theoretical derivations and simulation studies using the yeast protein interaction network as a prototype. However, the results apply to general scale-free networks. Connectivity distribution of scale-free networks with erroneous links under a simple model ------------------------------------------------------------------------------------------ We first study how the connectivity distribution of a scale-free network is affected when errors are present. Following previous studies on the reliability of protein interaction networks \[[@B6]\], we assume a simple error mechanism in which the false positive rate (r~FP~) and false negative rate (r~FN~) are the same for all node pairs, and false positives and false negatives are independently generated. The false positive rate and false negative rate of a node pair refer to the probability that the pair of nodes is observed as linked when they are actually not and the probability that the pair of nodes is observed as unlinked when they are actually linked. Under this assumption, every truly linked pair of nodes has a probability r~FN~to be observed as unlinked nodes, and every truly unlinked pair of nodes has a probability r~FP~to be observed as linked nodes. The above assumption is similar to the grand canonical ensembles of random networks in Chapter 4 of Dorogovtsev and Mendes \[[@B7]\], in which networks evolve by removing existing edges and adding new edges with certain probabilities. We can also view the perturbed network as obtained by removing edges (false negative) and adding edges (false positive) from the underlying network. The probability of adding an edge between two non-linked nodes is the false positive rate r~FP~, and the probability of removing the edge between two linked nodes is the false negative rate r~FN~. However, while Dorogovtsev and Mendes mostly discussed the connectivity distribution of equilibrium networks (networks obtained after infinite times edge adding and removing), we focus on the connectivity distribution of the observed network that are obtained by considering removing every existing edge and adding non-existing edges just once. ### Connectivity distribution of the perturbed network In the following, we will derive the distribution of the observed connectivities for a scale-free network of size nfor given values of r~FP~and r~FN~. Let N~P~and N~T~denote the observed and true connectivity of a node, respectively. Then the probability to observe a node with klinks is ![](1471-2105-6-119-i1.gif) The minimum and maximum connectivity of a node, T~min~and T~max~, are assumed to be the same for all the nodes in the network, and their values depend on the specific network. In general, we set T~min~= 0 and T~max~= n- 1 when expert knowledge is not available, where ndenotes the size of the network, i.e., the total number of nodes in the network. The following elucidates how to calculate (2) analytically. Let N~FP~, N~TP~, N~FN~, N~TN~, and N~N~be the numbers of false positive links (observed as linked but actually not), true positive links (observed as linked and actually linked), false negative links (observed as unlinked but actually linked), true negative (observed as unlinked and actually unlinked) and negative links (actually unlinked) associated with the node, respectively. Since the observed links of a node consist of both false positive and true positive ones, and the true links consist of true positive and false negative ones, we have N~P~= N~FP~+ N~TP~, N~T~= N~FN~+ N~TP~, N~N~= N~FP~+ N~TN~, and T~max~= N~T~+ N~N~. Furthermore, underour assumed error mechanism, following similar derivations as shown in \[[@B7]\], N~FP~and N~FN~follow the binomial distributions Bin(T~max~- N~T~, r~FP~) and Bin(N~T~, r~FN~), respectively, for a given value of N~T~. This implies that r~FP~= E(N~FP~)/(T~max~- N~T~) = E(N~FP~)/(N~FP~+ N~TN~) and r~FP~= E(N~FN~)/N~T~= E(N~FN~)/(N~TP~+ N~FN~), where E(X) denotes the expectation of random variable X. Then the conditional probability P(N~P~= k\\|N~T~= j) in (2) can be written as follows. ![](1471-2105-6-119-i2.gif) where dBin(k; p, n) = P(X= k) with X\~ Bin(n, p). Moreover, the power law of the scale-free network implies that P(N~T~= j) = cj^-γ^. Hence, the observed connectivity distribution can be calculated by ![](1471-2105-6-119-i3.gif) ### Simulations We next explore the impact of the erroneous links on the topology of the scale-free networks. With an emphasis on the yeast protein interaction network, we compute the distribution of the observed connectivity of scale-free networks with the false positive rate (r~FP~) and false negative rate (r~FN~) similar to the yeast protein interaction network under the assumption of the aforementioned simple error mechanism. We set the scale parameter γ= 3, the size of the network n= 1000 or 7000, and vary r~FP~from 0.0001 to 0.0003 and r~FN~from 0.1 to 0.9 on 9 equally spaced values. These ranges of r~FP~and r~FN~are based on Deng et al. \[[@B8]\], in which the authors estimated the false positive rate and false negative rate to be less than 0.000285 and greater than 0.64, respectively, based on the Y2H data. We consider a larger range of r~FP~to cover other data sources, such as the MIPS complex data, where false positives are less frequent. In the calculations, we use T~min~= 1 and T~max~= n- 1. In the log-log plot (Figures [1](#F1){ref-type="fig"} and [2](#F2){ref-type="fig"}) of the observed connectivity distribution of the perturbed networks when (r~FP~= 0.0001, r~FN~= 0.3) and (r~FP~= 0.00015, r~FN~= 0.8), it can be seen that the connectivity distribution after perturbation still maintains the scale-free property in the middle range of the connectivity, but deviates from the original linear pattern at both the small and large connectivity regions. The slope of the linear part is close to the true value -3 (see Tables A.1 and A.2 in [Additional file 1](#S1){ref-type="supplementary-material"}). The deviation is more significant in the large connectivity region than that in the small connectivity region. This deviation pattern is consistent across networks of different sizes considered in our calculations (data not shown). Comparisons among the observed connectivity distributions (figures not shown) of perturbed networks with different values of r~FP~and r~FN~suggest that the deviation depends little on r~FP~but largely on r~FN~. As r~FN~increases, the deviation of the tail probability becomes more significant. This deviation is also more obvious in a smaller network. ### Estimation of γ The connectivity distribution of the perturbed network suggests a cautious use of the observed link data, especially on estimating γ. The scaling parameter γ, an important characteristic measure of the scale-free network, is commonly estimated using the ordinary least squares (OLS) in the linear model from the log transformation of (1). log P(k) = log c- γlog k. (4) It is well known that the OLS estimator can be very sensitive to even a small number of outliers. For example, applying the OLS estimator in Figure 1(a) will not be able to capture the linear trend if the point at the last end is included in the estimation. Therefore, robust estimators, such as the M-estimator and the least trimmed squares (LTS) estimator \[[@B9]\] are more proper choices in such situations due to their resistance to outliers. Our simulations suggest that the LTS estimator can correctly capture the linear trend without visual diagnosis of the connectivity distribution, while the OLS and M-estimator often fail to estimate the slope of the linear part correctly. Therefore, we will use the LTS estimator in our following simulation studies. Exploring error mechanisms of yeast protein interaction networks by simulations ------------------------------------------------------------------------------- In the previous section, we found that the scale-free property can be conserved to a large extent under a simple error mechanism. However, the error mechanisms of the real data are often more complicated. For more complicated error mechanisms, theoretical derivations of the connectivity distribution of the perturbed networks are often intractable. But it is also important to know how the empirical connectivity distributions of real networks are affected by the erroneous links. Therefore, we conduct extensive simulation studies to investigate the finite-sample impact of the error mechanisms on the connectivity distribution. Our study focuses on the yeast protein-interaction network data. For real network data, no matter whether erroneous links are involved or not, the empirical connectivity distribution will not display a linear pattern as clear as the ones in Figure [1](#F1){ref-type="fig"} due to sampling variations and its discrete approximation to the tiny probability of nodes with large connectivities. For example, Figure [3](#F3){ref-type="fig"} shows the connectivity distribution of a simulated scale-free network Net~0~and Figure [4](#F4){ref-type="fig"} shows the connectivity distribution of Net~0~after perturbation by the simple error mechanism discussed above. In Figure [4](#F4){ref-type="fig"}, we observe a much larger curvature deviation from the linear trend at the small connectivity region than that in Figures [1](#F1){ref-type="fig"} and [2](#F2){ref-type="fig"}. It is not clear why the empirical distributions of the simulated networks are so different from the theoretical calculations, but this observation demonstrates that simulation studies are necessary to complement the findings from the theoretical calculations. In addition, simulation studies can also explore possible error mechanisms by comparing the connectivity distributions of simulated perturbed scale-free networks with the observed networks by assuming that their underlying structure are indeed scale-free. In the following, we investigate the error mechanisms of two real yeast protein interaction network data sets used in Jeong et al. \[[@B3]\] and Deng et al. \[[@B6]\] by comparing the connectivity distribution of these two networks with that of the simulated network perturbed by different error mechanisms. We assume that the true underlying topology of the yeast protein interaction network is scale-free \[[@B3]\]. Then if we perturb the simulated scale-free network by the error mechanisms similar to the ones of the real yeast protein interaction networks, the resulting connectivity distribution should be similar to the ones of the real networks. ### MIPS and Y2H yeast protein networks Jeong et al. derived the yeast protein network from combined, non-overlapping Y2H data \[[@B10],[@B11]\]. This network has 1,870 proteins as nodes, connected by 2,240 identified direct physical interactions \[[@B12]\]. The other network was obtained from the gold standard of yeast protein interactions based on the MIPS complex data \[[@B13]\]. This gold standard data set has 1,376 proteins and 2,876 interacting protein pairs, out of which 2,559 are also recorded in the Yeast Proteome Database (YPD) \[[@B14]\]. The YPD subset has 1,373 proteins. Estimates of γfrom the Y2H network, the gold standard data and the YPD subset are 2.396, 2.721 and 2.870, respectively. The connectivity distributions of these two networks are shown in Figure [5](#F5){ref-type="fig"} and Figure [6](#F6){ref-type="fig"}, respectively. ### Error mechanisms We consider different error mechanisms in terms of different types of false positive rates (p~ij~= P(x~i~and x~j~are observed linked\\|x~i~and x~j~are actually unlinked)) and false negative rates (q~ij~= P(x~i~and x~j~are observed unlinked\\|x~i~and x~j~are actually linked)) for node pair (x~i~, x~j~), i= 1,\..., n, j= 1,\..., n, i≠ j. Assume that the overall false positive rate and false negative rate are r~FP~and r~FN~, in the sense that the expected number of false positive links and false negative links are E(N~FP~) = r~FP~N~N~and E(N~FN~) = r~FN~N~P~. We consider nine different error mechanisms by letting p~ij~and q~ij~be one of the following three different types: 1\. constant: p~ij~= r~FP~and q~ij~= r~FN~for all (x~i~, x~j~); 2\. increasing (with connectivity): ![](1471-2105-6-119-i4.gif) 3\. decreasing (with connectivity): ![](1471-2105-6-119-i5.gif) where L(x) denotes the true connectivity of node x. For Net~0~, N~P~= 49, 007 and N~N~= 24, 503, 521. The combinations of different structures on false positive rates and false negative rates produce nine error mechanisms in Table [1](#T1){ref-type="table"}. ### Simulation studies We simulate a scale-free network Net~0~using the preferential attachment growth model \[[@B15],[@B16]\]. In this algorithm, we start from m~0~= 7 isolated nodes and add m= 7 links to the existing nodes with probability proportional to their connectivity in each of the T= 7, 000 evolving steps. Net~0~has L= 49, 007 links and n= 7, 008 nodes. The mean-field theory \[[@B15]\] suggests that the theoretical value of γfor Net~0~is 3, which agrees well with the estimates in Table [2](#T2){ref-type="table"}. We always assume that false positives and false negatives are independently generated. In the simulations, a link is added (false positive) between every two unlinked nodes (x~i~, x~j~) in Net~0~with probability p~ij~, and the link is removed (false negative) between two linked nodes (x~i~, x~j~) in Net~0~with probability q~ij~. We also consider these error mechanisms under high and low overall false positive (r~FP~) and false negative rates (r~FN~). The connectivity distributions of Net~0~after perturbation are shown in Figures [7](#F7){ref-type="fig"}, [8](#F8){ref-type="fig"}, [9](#F9){ref-type="fig"}, [10](#F10){ref-type="fig"} for different values of r~FP~and r~FN~: (0.00025, 0.5), (0.00025, 0.8), (0.00015, 0.5), (0.00025, 0.8). Under the nine different error mechanisms, the connectivity distribution of the perturbed Net~0~can be dramatically different. Under error mechanisms S2, S5, S6 and S9, the perturbed networks contain a small proportion of nodes with low connectivity, which differs greatly from the observed yeast protein interaction networks (Figures [5](#F5){ref-type="fig"} and [6](#F6){ref-type="fig"}). This finding suggests that these four mechanisms are far different from the true error structure, and we will not discuss them in the following. We also observe that changes in r~FP~render little impact on the connectivity distribution under all error mechanisms, but a higher value of r~FN~increases the probability of nodes with small connectivity under S1, S3 and S8. And mechanisms S4 and S7 are highly stable structures, that is, the connectivity distribution changes little in response to changes in r~FP~or r~FN~under these two error mechanisms. This suggests that scale-free networks with constant false negative rates can still provide very credible information about its topological structure. This finding is also confirmed by the fact that the estimates of γvary little when r~FN~changes (see Tables A.5 and A.6 in [Additional file 1](#S1){ref-type="supplementary-material"}). The estimated values of γvary only from 2.61 to 3.03 with a standard error of 0.125 under S4 and only from 2.56 to 3.31 with a standard error of 0.161 under S7, whereas the estimate of γclearly decreases as r~FN~increases under S3 and S8 (Tables A. 4 and A. 7 in [Additional file 1](#S1){ref-type="supplementary-material"}). Under S1, there is no clear pattern on the estimated γas r~FN~changes (Table A.3 in [Additional file 1](#S1){ref-type="supplementary-material"}), but the estimates of γvary in a much wider range (1.16 -- 4.35) compared with those under S3 and S8. It is worth noting that our conclusions are restricted to the particular range of r~FP~and r~FN~we have studied, however these ranges are believed to be reasonable to describe the Y2H systems. The simple error mechanism S1 with a high false negative rate produces patterns (Figures 8(a) and 10 (a)) similar to that of the gold standard data (Figure [6](#F6){ref-type="fig"}). For the Y2H yeast protein interaction network (Figure [5](#F5){ref-type="fig"}), S4 gives the best approximation, but still differs slightly in the probabilities of nodes with small connectivity. This suggests that the real error structure of the Y2H analyses may be more complicated than all the simple proposals we have considered. Conclusion ========== This article first investigates the impact of erroneous links on network topological inference. From our theoretical and simulation results, we find that, under a simple error mechanism, the scale-free property is preserved for moderate connectivities. But the linear pattern is distorted at both the small and large connectivity regions. Accordingly, we recommend to use robust estimators (e.g. LTS) that are more resistant to the outliers at both ends of the distribution to estimate the scale parameter γ. Moreover, we have also explored possible error mechanisms of the yeast protein interaction data by simulations considering nine different error mechanisms. The results suggest that changes in the overall false positive rates have little impact on the resulting connectivity distribution, but increasing the overall false negative rates can increase the probability of nodes with small connectivities under some error mechanisms, and hence decrease the scale parameter γ. The connectivity distribution can be very stable under several error mechanisms when the overall false positive rates and false negative rates change, which suggests that in certain situations the observed data can provide suffcient topological information on the underlying network structure even when the false negative rates are quite high. The simple error mechanism that assumes that the false positive rate and false negative rate of each protein pair are constants agrees well with the MIPS gold standard data when the false negative rate is high. A different error mechanism is suggested for the Y2H data, where more connected protein pairs tend to have higher false positive rates and lower false negative rates. As this error mechanism provides only a reasonable approximation to the Y2H data, more sophisticated mechanisms might be needed to better capture its error structure. Methods ======= Preferential attachment growth model ------------------------------------ In a series of papers \[[@B15],[@B16]\], Barabási et al. demonstrated that a scale-free network could be obtained by growing from a small number of isolated nodes by preferential attachment. The simulation scheme is defined in two steps: 1\. Growth: starting with a small number (m~0~) of nodes, add a new node at every time step and connect it to m(≤ m~0~) nodes already present in the system 2\. Preferential attachment: The new node is more likely to connect to nodes with larger connectivity. The probability Π~i~that a new node will be connected to node idepends on its connectivity k~i~, such that ![](1471-2105-6-119-i6.gif). Least Trimmed Squares (LTS) --------------------------- The basic idea of LTS estimation is to minimize the sum of hsmallest squared residuals instead of all squared residuals in the OLS to achieve robustness and also maintain good effciency. Please refer to \[[@B9]\] for more details of the algorithm, such as practical choices of h. In this article, the LTS estimation is performed using the lqs() function implemented in R \[[@B17]\]. Authors\' contributions ======================= HZ had the initial idea and initiated the study. NL conducted the data analyses, and created all tables and figures, under the supervision of HZ. Both authors read and approved the final manuscript. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 Tables of the estimates of the scale parameter γ. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ This work was supported in part by NSF grant DMS 0241160 and NIH grant R01 GM59507. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### *Connectivity distribution of the perturbed scale-free networks (r~FP~= 0.0001, r~FN~= 0.3). This picture shows the connectivity distribution of the the perturbed networks using (3) provided that r~FP~= 0.0001 and r~FN~= 0.3. Figure 1(b) and 1(d) are the linear parts of Figure 1(a) and 1(c), respectively, imposed with the regression lines fitted by the OLS. ::: ![](1471-2105-6-119-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Connectivity distribution of the perturbed scale-free networks (r~FP~= 0.00015, r~FN~= 0.8). This picture shows the connectivity distribution of the the perturbed networks using (3) provided that r~FP~= 0.00015 and r~FN~= 0.8. Figure 2(b) and 2(d) are the linear parts of Figure 2(a) and 2(c), respectively, imposed with the regression lines fitted by the OLS. ::: ![](1471-2105-6-119-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Connectivity distribution of Net~0~. This picture shows the connectivity distribution of the simulated scale-free random network Net~0~imposed with regression lines given by different methods (dashed line: OLS; dotted line: M-estimation; solid line: LTS). ::: ![](1471-2105-6-119-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Connectivity distribution of Net~0~after perturbation (r~FP~= 0.0002, r~FN~= 0.7). This picture shows the connectivity distribution of the simulated scale-free random network Net~0~perturbed by the simple error mechanism using r~FP~= 0.0002 and r~FN~= 0.7. Regression lines given by different methods are also imposed (dashed line: OLS; dotted line: M-estimation; solid line: LTS). ::: ![](1471-2105-6-119-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Connectivity distribution of the Y2H yeast protein interaction network. This picture shows the connectivity distribution of the protein interaction network in Jeong et al. \[3\] inferred from the Y2H data. The imposed regression line is fitted by the LTS method. ::: ![](1471-2105-6-119-5) ::: ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Connectivity distribution of the MIPS yeast protein interaction network. This picture shows the connectivity distribution of the protein interaction network in Deng et al. \[6\] inferred from the MIPS gold standard data. The imposed regression line is fitted by the LTS method. ::: ![](1471-2105-6-119-6) ::: ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Connectivity distribution of Net~0~perturbed by different error mechanisms (r~FP~= 0.00025, r~FN~= 0.5). ::: ![](1471-2105-6-119-7) ::: ::: {#F8 .fig} Figure 8 ::: {.caption} ###### Connectivity distribution of Net~0~perturbed by different error mechanisms (r~FP~= 0.00025, r~FN~= 0.8). ::: ![](1471-2105-6-119-8) ::: ::: {#F9 .fig} Figure 9 ::: {.caption} ###### Connectivity distribution of Net~0~perturbed by different error mechanisms (r~FP~= 0.00015, r~FN~= 0.5). ::: ![](1471-2105-6-119-9) ::: ::: {#F10 .fig} Figure 10 ::: {.caption} ###### Connectivity distribution of Net~0~perturbed by different error mechanisms (r~FP~= 0.00015, r~FN~= 0.8). ::: ![](1471-2105-6-119-10) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Nine error mechanisms. ::: Error mechanism p~ij~ q~ij~ ----------------- ------------ ------------ S1 constant constant S2 constant increasing S3 constant decreasing S4 increasing constant S5 increasing increasing S6 increasing decreasing S7 decreasing constant S8 decreasing increasing S9 decreasing decreasing ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Parameter estimates for Net~0~. ::: Parameter OLS M-estimation LTS ----------- -------- -------------- ------- log c* 1.4600 1.7846 4.008 γ 2.0918 2.1769 2.803 :::	PubMed Central	2024-06-05T03:55:57.609298	2005-5-16	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1156868/", "journal": "BMC Bioinformatics. 2005 May 16; 6:119", "authors": [ { "first": "Nan", "last": "Lin" }, { "first": "Hongyu", "last": "Zhao" } ] }
PMC1156869	Background ========== There are two major methodological problems that deal with the issue of stochastic dependence between gene expression signals in microarray data. The first arises naturally when adjustments for multiplicity of tests are made by pooling across genes (or tests)in an effort to find differentially expressed genes in two-sample comparisons. The empirical Bayes methodology in the nonparametric \[[@B1]-[@B3]\] and parametric formulations \[[@B4],[@B5]\], and closely related methods exploiting a two-component mixture model \[[@B6]-[@B8]\] represent typical examples. The common feature of such methods is that a test statistic (measure of differential expression) is first calculated for each gene to account for biological variability and then all the statistics (or the associated p-values) are pooled together and treated as a sample from which to estimate the sampling distribution of this statistic, the false discovery rate (FDR), q-values, etc. The same kind of pooling is typically used in maximum likelihood inference from microarray data \[[@B9],[@B10]\] and some other methods of testing for differential expression of genes. In all such approaches, the stochastic dependence between gene expression values or test statistics is a nuisance that hinders their application. The independence assumption is frequently invoked when building a theoretical foundation for a particular method of statistical inference. Some authors (e.g., \[[@B11]\]) allow for dependence between differentially expressed genes while assuming stochastic independence of those genes that do not change their expression between the two conditions under study. The biological rationale for such a hypothesis is unclear, because the normally functioning genes are involved in numerous biochemical pathways much like the altered ones. The stochastic dependence between expression levels and thus between the associated test statistics is really a serious problem. It may cause high variability of statistical estimators and even deteriorate their consistency. To obtain theoretical results it is frequently assumed that weak or almost sure convergence holds for an empirical distribution function constructed from the data pooled across genes (see, i.e. \[[@B12],[@B13]\]). However, this assumption is diffcult to validate biologically so that the required convergence to the true distribution function is always questionable; it may or may not be the case depending on the type and strength of stochastic dependence. Storey \[[@B12]\] advocates the assumption of weak dependence between test-statistics when discussing some concerns raised in the paper by Ge, Dudoit, and Speed (hereafter abbreviated by GDS) \[[@B14]\]. It is worth quoting his line of reasoning at length: \"I hypothesize that the most likely form of dependence between the genes encountered in DNA microarrays is weak dependence, and more specifically, \"clumpy dependence\"; that is, the measurements on the genes are dependent in small groups, each group being independent of the others. There are two reasons that make clumpy dependence likely. The first is that genes tend to work in pathways, that is, small groups of genes interact to produce some overall process. This can involve just a few to 50 or more genes. This would lead to a clumpy dependence in the pathway-specific noise in the data. The second reason is that there tends to be cross-hybridization in DNA microarrays. In other words, the signals between two genes can cross because of molecular similarity at the sequence level. Cross-hybridization would only occur in small groups, and each group would be independent of the others.\" This hypothesis does not seem plausible from a biological standpoint because of the pleiotropic character of gene function: one gene participates in multiple molecular pathways. However, the possibility that it may approximately be true for all practical purposes cannot be ruled out. There are two key words in the above quotation: \"small groups\" and \"weak dependence\". Whether or not such groups are small and stochastic dependence is suffciently weak can be deciphered only from real world data. To the best of our knowledge, no attempt has been made so far to systematically study dependence structures in microarray data using large data sets. In this connection we would like to continue quoting from \[[@B12]\]: \"Many assumptions that have been made for modeling microarray data have yet to be verified. Hopefully evidence either for or against these assumptions will emerge\... GDS have stressed the dependence between the genes\... I leave it as a challenge to them to provide evidence from real microarray data that the aforementioned assumptions do not hold. I have not been able to find it myself\". In the present paper, we take the first step in this direction by conducting an empirical study of the correlations between test statistics associated with different genes. The second research area where the dependence between gene expression levels plays a crucial role is the discovery (reverse engineering) of molecular pathways and networks from microarray data \[[@B15]\]. A popular approach to pathway reconstruction is based on the sample correlation coeffcient or mutual information measures that are deemed to characterize interactions between genes via their products. These measures of interaction are computed from gene expression values observed across various experimental conditions. The snag here is that strong correlations in the raw (not normalized but background corrected) expression data may be induced by an array-specific technological noise, thereby producing numerous false-positive edges in the corresponding graph representing the underlying structure of a given pathway or network. However, if the data are normalized before the analysis, then the correlation structure of expression signals may be partially destroyed by the normalization procedure so that many edges in the resultant graph may be missing. The same applies to clustering techniques that utilize information on pairwise dependencies between the genes. The problem is less pressing where causal inference is possible from gene perturbation experiments. Although the present paper does not have a direct bearing on such settings, our results suggest that associative networks built on microarray data alone may have little to do with biological reality. The problem merits careful investigation in order to make the reverse engineering of this type more credible. The present paper is focused on the correlations between test-statistics associated with expression signals produced by each gene and the effects of normalization procedures on these correlations. We limit our consideration to the t-statistic which is the most popular choice in microarray data analysis. Normalization is intended to mitigate the effect of technological noise that is inherent in microarray data. Normalization procedures tend to reduce the variability of original microarray data (Park et al, \[[@B16]\]), however no study has been carried out to assess the effect of such procedures on the correlation structure of microarray data in general and the correlation of t-statistics in particular. In a methodological study such as ours, it is a great advantage to have access to a large data set involving hundreds of arrays. We used the St. Jude Children\'s Research Hospital (SJCRH) Database on childhood leukemia which falls into this category. Computer simulations provide the necessary, albeit not very realistic, control where the actual model is known and arbitrarily large samples can be generated for testing various methodologies. Results ======= The design of our study is presented in the Methods section. This design allows us to compute the t-statistics (across arrays) for each gene and each pair of subsamples. This computation results in 15 values (corresponding to the 15 pairs of subsamples) of the t-statistic associated with each gene. Then we compute the sample correlation coeffcients between the t-statistics thus obtained for every pair of genes. The resulting coeffcients are summarized in the form of a histogram. We interpret such histograms as pertinent summary characteristics and not as estimators of some population distribution densities. We also look at pre-selected individual genes to determine the range of their correlation with all other genes. This range can be characterized by the number of gene pairs formed by a given gene with the correlation coeffcient exceeding some threshold level. We adopt the value of 0.5 as such a threshold. Using these tools we attempt to answer the following questions: • What is the (pairwise) correlation structure of the t-statistic in a large population of genes? • What is the impact of normalization procedures on this structure? • What is the impact of normalization procedures on the number of highly correlated pairs formed by a given gene? Figure [1A](#F1){ref-type="fig"} shows the distribution (histogram) of correlation coeffcients for the t-statistics estimated from the SJCRH leukemia data for all pairs of genes. It is clear that the distribution is heavily shifted towards high positive correlation between the genes. In particular, more than 36% of pairs have their sample correlation coeffcients higher than 0.75 and only 7.6% have the coeffcients smaller than 0.25. The proportion of gene pairs with correlation coeffcients greater than 0.5 is 76%. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### The histogram of correlation coeffcients for overlappingpairs of t-statistics associated with individual genes in the SJCRH data. A: data before normalization, B: GEO, C: RANK, D: QUANT, E: simulated set of data SIMU1. ::: ![](1471-2105-6-120-1) ::: The effects of three normalization procedures (GEO, RANK, and QUANT, as defined in the Methods section) are shown in Figures [1B--1D](#F1){ref-type="fig"}. Figure [1E](#F1){ref-type="fig"} presents an ideal case where the t-statistics were obtained from independent normally distributed data (see the Methods section for explanations) produced by simulations (SIMU1). In this case, the proportion of gene pairs with correlation coeffcients greater than 0.5 is only 1.5%. While the normalization procedure GEOdestroys a large proportion of correlation, the procedures RANKand QUANToutperform it as far as the reduction of between-gene dependence is concerned. The effects of the latter two procedures are largely similar. The procedure RANKreduces the proportion of correlation coeffcients greater than 0.5 to 4.3%, while the procedure QUANTreduces this proportion to 7.2%. For comparison, this indicator is equal to 14% for GEO. Thus the procedure RANKhas the strongest effect on the correlation structure. Figure [1](#F1){ref-type="fig"} in the Additional Material Files \[see the file [\"Additional File 1\"](#S1){ref-type="supplementary-material"}\] shows essentially the same effect for randomly selected non-overlapping pairs of genes. The effect of normalization on the between-gene correlations observed in the simulated data SIMU2N and SIMU2 is stronger than that in the case of biological data (the SJCRH leukemia data set). This can be seen in Figures [2](#F2){ref-type="fig"}, [3](#F3){ref-type="fig"}, where only the results for the quantile normalization are shown. Again, if we look at the proportion of gene pairs with correlation coeffcients greater than 0.5, this indicator equals 1.5% for SIMU2N and 1.5% for SIMU2. The effects of GEOand RANKare displayed in Figures [1](#F1){ref-type="fig"}, [2](#F2){ref-type="fig"} included in the Additional Material Files \[see the file [\"Additional File 2\"](#S2){ref-type="supplementary-material"}\]. The stronger effect of GEOon the correlation structure of the SIMU2N data as compared to the SJCRH data comes as no surprise because the noise is simulated as an array-specific random effect, for which a heuristic justification of the GEOprocedure is possible \[[@B20]\]. The normalization procedures exert their effect both on the correlation induced by the noise and on the true correlation that reflects interactions between gene products. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### The effect of the normalization procedure QUANTas applied to the SIMU2N data. A: data without noise (SIMU2), B: data with noise (SIMU2N), C: SIMU2 after normalization, D: SIMU2N after normalization. ::: ![](1471-2105-6-120-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### The behavior of the standard deviation of the sample mean as a function of the number of involved genes. 1. Raw biological data; 2. Quantile normalization; 3. Independent simulations (SIMU1). ::: ![](1471-2105-6-120-3) ::: The effect of the quantile normalization for the SIMU3N, shown in Figure [3](#F3){ref-type="fig"} in the Additional Material Files \[see [\"Additional File 2\"](#S2){ref-type="supplementary-material"}\], deserves special discussion. Recall that each gene in the data set SIMU3N correlates only with a distinct group of genes termed a clump. Even if the genes involved in the same clump are heavily correlated, the average (over all pairs of genes) correlation coeffcient may still be quite low. When a uniformly distributed multiplicative random noise is imposed on each array, the genes pertaining to different clumps become highly correlated. The noise strengthens the intra-clump correlation as well. Recall that the clumpy structure of simulated data serves as a simplistic model of gene interactions within distinct pathways. As seen in Figure [3](#F3){ref-type="fig"} \[see [\"Additional File 2\"](#S2){ref-type="supplementary-material"}\], the normalization procedure QUANTis not nearly as effective as in the case of the SIMU2N data. This procedure does not eradicate the overall correlation between genes in the SIMU3N data. In this sense, the effects of normalization seen in the SIMU3N and in real biological data look similar. Another way of studying such effects is to look at the number of pairs characterized by a relatively high correlation with a pre-selected gene. Tables [1](#T1){ref-type="table"}, [2](#T2){ref-type="table"}, [3](#T3){ref-type="table"} present the results for 20 genes that produce large numbers of highly correlated (with correlation coeffcients greater than 0.5). These initiator genes were identified through each of the data sets (simulated and biological) under study. The final column in every table gives the number of highly correlated pairs formed by a given gene before normalization. All the selected genes form such pairs with the overwhelming majority of genes. We term this type of dependence the long-range correlation. The number of highly correlated gene pairs remaining after a given normalization procedure serves as an indicator of its effciency. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Long-range correlation analysis for the SIMU2N data. ::: Gene Label GEO QUANT RANK SIMU2N ------------ ------- ------- ------- -------- 1 743 746 741 12558 2 754 750 756 12558 3 723 723 721 12558 4 705 698 718 12558 5 736 734 754 12558 6 751 763 765 12558 7 702 695 709 12558 8 667 665 679 12558 9 747 747 759 12558 10 728 730 736 12558 11 713 717 713 12558 12 696 699 685 12558 13 743 750 762 12558 14 725 721 733 12558 15 691 691 740 12558 16 789 789 799 12558 17 724 725 669 12558 18 716 712 722 12558 19 762 762 720 12558 20 676 673 708 12558 Mean 724.6 724.5 729.5 12558 STD 30.1 31.8 31.9 0 ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Long-range correlation analysis for the SIMU3N data. ::: Gene Label GEO QUANT RANK SIMU3N ------------ -------- -------- -------- --------- 1 483 520 512 12297 2 471 582 591 10656 3 436 523 614 12506 4 644 643 744 11031 5 677 739 765 11320 6 610 543 570 12413 7 612 863 788 12429 8 802 727 711 12077 9 1743 1406 1077 11898 10 975 895 920 12001 11 1352 1330 1543 12453 12 670 707 686 12480 13 1874 1849 1890 6913 14 1858 1765 1808 9371 15 1925 1790 1974 12469 16 1792 1718 1796 12520 17 1764 1526 1679 12499 18 1769 1684 1821 12509 19 1476 1300 1569 12514 20 2223 2307 2148 12507 Mean 1207.8 1170.9 1210.3 11743.2 STD 617.3 557.5 576.5 1402 ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Long-range correlation analysis for the SJCRH data. ::: Gene Label GEO QUANT RANK raw data ------------ ----------- -------- -------- ---------- 1 5644 462 494 12481 2 7330 3175 1431 12486 3 4189 1480 2062 12496 4 5218 2728 1548 12493 5 8169 1888 1064 12451 6 8140 956 1162 12482 7 323 1169 839 12480 8 6774 1479 839 12497 9 7676 1832 2140 12390 10 8234 794 1440 12384 11 7652 930 466 12498 12 8266 1329 708 12476 13 8197 1343 2045 12391 14 7422 2118 2513 12501 15 1588 1467 1011 12494 16 7861 1931 1133 12429 17 1292 1477 1445 12489 18 6389 2949 1456 12481 19 7359 490 514 12469 20 4384 970 787 12488 Mean 6105.4 1548.4 1254.9 12467.8 STD 2545 2512 756 589.5 38.2 ::: Consider first the results obtained with simulated data. Each of the twenty initiator genes selected from SIMU2N form exactly 12,558 highly correlated pairs. When applied to the SIMU2N data, the normalization procedures RANKand QUANTbring this number down to 700 on average (see Table [1](#T1){ref-type="table"}). The variability in the size of this set of genes is low. For example, the number of highly correlated genes ranges from 661 to 794 after the application of the QUANTprocedure. Both procedures indiscriminately reduce the true (intrinsic) correlation and its spurious (nuisance) counterpart. Although less effective, the procedure GEOdoes a similar job. The results for the SIMU3N data are different (see Table [2](#T2){ref-type="table"}). While the number of highly correlated gene pairs tends to decrease significantly for each of the twenty initiator genes, the size of this effect depends on the group of genes from which the initiator gene was chosen. This increases the variability of the number of highly correlated pairs remaining after normalization. For the QUANTmethod the range is from 526 to 2,368 showing that the remaining correlation extends far beyond the specified clumpy structure. We then selected 20 initiator genes in the SJCRH data set representing real biological data. The number of highly correlated pairs formed by these genes before normalization ranges from 12,384 to 12,501, which is a very narrow range indeed. As is seen in Table [3](#T3){ref-type="table"}, the procedure GEOdoes not destroy the correlation effectively; it leaves huge numbers (up to 8,266) of highly correlated gene pairs. The rank normalization results in much smaller numbers of highly correlated genes that range from the lowest of 494 to the highest of 2,513. The average is 1,255, which is about twice as much as we get from any normalized SIMU2N data. The variability is also very high, resembling a clumpy effect seen in the SIMU3N set. We do not consider this similarity as evidence for a clumpy structure of microarray data, but the results in Table [3](#T3){ref-type="table"} suggest that, if such a structure exists, an average clump should be expected to involve at least an order of magnitude more genes than the clump size postulated by Storey \[[@B12]\]. Another interesting finding in Table [3](#T3){ref-type="table"} is that the quantile normalization tends to leave more highly correlated genes in comparison to the rank normalization. This is contrary to our expectations based on the comparisons of correlation histograms reported above. The effect of the QUANTis also more variable than that of the RANK, which is another dissimilarity of practical importance. Leaving aside the fact that the RANKprocedure is applied to gene expressions, while the QUANTworks at the probe feature level, the difference between the two normalization methods is that we replace entries in an array by their ranks in the former case and by ![](1471-2105-6-120-i1.gif) in the latter. Recall that ![](1471-2105-6-120-i1.gif) is the average of entries having the same rank over all arrays. Obviously, the QUANTpreserves more quantitative information in the data than does the RANKprocedure. This explains why the result of the rank normalization is less variable. The effect of the normalization QUANTon the distribution of the t-statistics across the genes for the actual and simulated data is shown in Figures [1](#F1){ref-type="fig"}, [2](#F2){ref-type="fig"}, [3](#F3){ref-type="fig"}, 4 included in the Additional Material Files \[see the file [\"Additional File 3\"](#S3){ref-type="supplementary-material"}\]. From Figures [2](#F2){ref-type="fig"} and [3](#F3){ref-type="fig"}, it is clear that, when applied to the simulated data SIMU2 and SIMU2N, this procedure makes the distribution of t-statistics similar to that in the ideal case shown in Figure [1](#F1){ref-type="fig"} \[see [\"Additional File 3\"](#S3){ref-type="supplementary-material"}\]. However, the effect appears to be somewhat less satisfactory with real data, especially in the tail regions of the resultant distribution of t-statistics. The results shown in this section are obtained with a single initial random split of the pooled set of arrays into two groups. We have conducted several such splits in this study. All the above-described effects are highly reproducible, and reporting the results for other splits in the paper is not warranted. Discussion ========== It follows from our observations that normalization procedures are capable of destroying a significant part of correlations between gene expression signals and associated test-statistics. In doing so, they affect both the spurious correlation induced by the noise and the true correlation that reflects gene interactions. The clumpy structure (involving relatively large clumps of genes) of the SIMU3N data set is more resistant to this effect than the SIMU2N data. This is even more so for real biological data. The weaker effect of normalization seen in the SJCRH data indicates that the actual noise structure may be more complicated than assumed in the simulation studies (multiplicative array-specific random effect model). A clumpy structure of gene expression signals may also play a role in this phenomenon. This observation explains why it is so diffcult to remove correlations from the data. The destructive effect of normalization procedures on pairwise correlations in microarray data is good news for the methods of statistical inference that resort to \"pooling across genes\". However, it remains unclear whether or not the remaining correlation may still be substantial enough to invalidate such methods by affecting important properties of statistical estimators and tests. The problem invites further investigation. However, we would like to present an experiment specially designed to address the consistency question mentioned in the Background section. To this end, we applied the following algorithm to the SJRCH data: 1\. Select randomly 100 genes and compute the arithmetic (sample) mean of the t-statistics across these genes for each pair of subsamples. 2\. Compute the standard deviation of the sample mean across the 15 pairs of subsamples. 3\. Select randomly 100 from the remaining genes and compute the arithmetic mean for the 200 genes for each pair of subsamples. 4\. Compute the standard deviation from the sample means resulted from the previous step. 5\. Continue until the set of all genes is exhausted. 6\. Plot the estimated standard deviation of the sample mean as a function of the number of genes involved in each step of the algorithm. 7\. Repeat the procedure ktimes to generate ktrajectories of the standard deviation of the sample mean. The results of one such experiment are given in Figure [3](#F3){ref-type="fig"}. It is known that the sample mean is an unbiased and consistent estimator for the true mean value in the case of independent and identically distributed observations. This case is represented by Curve 3 generated by simulations. It is clear that the standard deviation decreases very rapidly and tends to zero with increasing the number of genes. However, the same is not true for the biological data. For the raw data, the standard deviation does not show a distinct tendency to decrease (Curve 1). When the data are normalized using the quantile normalization procedure, the standard deviation first drops and then stabilizes at an approximately constant level, no matter how many (up to 12,500) genes are involved in its estimation (see Curve 2). This is clearly the effect of (long-range) correlation between the t-statistics associated with different genes. The pattern seen in Figure [3](#F3){ref-type="fig"} was highly reproducible across k= 20 experiments with different random starts. If the standard deviation of an unbiased estimator tends to zero, this estimator is consistent. This is the case for Curve 3 but not quite so for Curve 2. While not a rigorous disproof of consistency of the sample mean in this case, the pattern seen in Curve 2 suggests that the estimator is likely to converge to a random variable (with the variance greater than zero) rather than to the true parameter to be estimated. This is definitely not a good sign for estimation procedures based on pooling across genes such as those built in the empirical Bayes methodology. The observed effect of normalization procedures is definitely bad news for the associative network reconstruction from gene expression data. Unless further technological advancements result in a significant reduction of the noise in microarray data, this kind of analysis will continue producing unreliable inferences. To normalize, or not to normalize: that is the question to which no scientifically sound answer is currently known as far as this kind of reverse engineering is concerned. Although limited to cell cultures, the causal inference from gene perturbation (disruption and over-expression) experiments seems to be the only solid alternative. From this standpoint the observations reported in the present paper add to the concerns expressed by several investigators regarding how much confidence to place in the thousands of papers already published using microarray technology \[[@B17]\]. Conclusion ========== The present paper provides quantitative insight into correlation between the t-statistics associated with different genes. This study leads us to conclude that: • There is a long-range correlation in microarray data manifesting itself in a huge number of genes that are heavily correlated with a given gene in terms of the associated t-statistics. • Using normalization of microarray data it is possible to significantly reduce correlation between the t-statistics computed for different genes. • Normalization procedures affect both the true correlation, stemming from gene interactions, and the spurious correlation induced by random noise. • It is likely that some noise effects represent non-monotone transformations of the underlying gene expression signals because even the rank normalization does not make the t-statistics independent when applied to the biological data. • Even the most effcient normalization procedures are unable to completely remove correlation between the t-statistics associated with different genes in biological data. Furthermore, the long-range correlation structure persists in normalized data. This remaining correlation may be strong enough to deteriorate consistency of statistical estimators built from measurements on the genes. Methods ======= Study design and biological data -------------------------------- There are 335 arrays (Affymetrix, Santa Clara, CA) in the SJCRH data set, each array representing N= 12, 558 genes. Each gene is represented in the data set by the logarithm of its expression level. The data are publicly available on the following website: \[[@B18]\]. The SJCRH data include the information on gene expression in normal blood and various types of childhood leukemia. The raw (background corrected but not normalized) expression data were generated by the output of the Bioconductor RMA (Robust Multi-Array Average) procedure when choosing the option: normalization = false. Since our focus was on purely methodological problems, we pooled all the available arrays together and shuffed the pooled sample. After randomly choosing and dropping 5 arrays (to make all subsamples of the same size), the pooled sample was randomly split into 30 parts, each containing 11 arrays. Then 15 pairs of the array samples were arranged and the corresponding 15 t-statistics were computed for each gene, thereby mimicking 15 two-sample comparisons under the null hypothesis of no differential expression. As a result, each gene was associated with 15 values of the t-statistic so that the Pearson correlation coeffcient between the t-statistics thus derived could be computed for any pair of genes. The output of the above-described series of procedures is a 12, 558 × 15 matrix of t-statistics and the associated vector of correlation coeffcients for all pairs of genes. We proceeded through the same sequence of operations when analyzing normalized and simulated data sets. In a separate experiment, we formed ![](1471-2105-6-120-i2.gif) non-overlapping pairs of genes to eliminate the spurious correlation between gene pairs due to multiple entries of the same gene in different pairs. Since this procedure begins with a randomly selected pair and proceeds through many steps of successive elimination of the previously selected pairs, it was repeated several times with different (random) starts and each output being analyzed separately. To study the long-range correlation, we picked 20 genes that produce large numbers of gene pairs with correlation coeffcients greater than 0.5. This experiment was designed to see how a given normalization procedure affects the number of such pairs associated with each of the twenty genes. A similar design was used with simulated data. Simulated data -------------- We simulated several sets of data to gain a better insight into the effects of normalization. All of them included the same numbers of arrays and genes as in the biological data described in the previous section. Specific characteristics of these data sets are given below. 1\. SIMU1: Every element x~ij~, i= 1, \..., 12, 558; j= 1, \..., 335 in SIMU1 represents log-intensity of expression of the ith gene from the jth array. The independent and identically distributed random variables x~ij~are generated from the standard normal distribution. This implies that the original expression signals are modeled as log-normally distributed random variables but we used their logarithms in our computations. This data set was used to illustrate the correlation analysis under independence of gene expression levels. 2\. SIMU2 is a 12, 558 × 335 random matrix that models an exchangeable correlation structure. The entries in this matrix are normal random variables with mean zero and unit variance. The entries from different columns are independent, while the correlation coeffcient between any two elements x~ij~of the same column is equal to 0.8. 3\. SIMU2N is a data set based on SIMU2. First we generate a 335-dimensional random vector A. The elements of Aare independent and identically distributed. The marginal distri- bution of every element a~ij~of Ais uniform over the interval \[[@B5],[@B10]\]. This random vector is used to model an array-specific noise. We then define y~ij~to be x~ij~+ a~j~, where x~ij~is the ij-th entry in the SIMU2 data set. The new matrix Y= {y~ij~} represents the data SIMU2N 4\. SIMU3 is a 12,550 × 335 matrix. The 12,550 rows (genes) are divided into ten groups of genes, each containing 1,255 rows. If two genes are both from the k-th group (gene numbers 100·(k-1)+1 through 100·k), for k= 1, 2, \..., 10, then the correlation coeffcient between them equals ![](1471-2105-6-120-i3.gif). Any two genes pertaining to different groups are stochastically independent. 5\. SIMU3N is the same as the SIMU3 data set but with an added noise. An array-specific multiplicative and uniformly distributed noise is modeled exactly as in the SIMU2N data. Normalization methods --------------------- Suppose there are Marrays of length N, and we represent the corresponding log-intensities as an N× Mmatrix Xsuch that each array is represented by a column in X. In this work, we used the following normalization methods: ### 1. Geometric mean normalization GEO If the array-specific random noise is multiplicative then a reasonable way to remove it from the expression values is to divide each element of the data matrix by the geometric mean over all gene expression signals on the array to which this element belongs. Szabo et al\[[@B20]\] discuss conditions under which this method is a valid one for testing two-sample hypotheses with microarray data. ### 2. Rank normalization RANK This method was proposed by Tsodikov et al\[[@B19]\] and discussed further in \[[@B20]\]. In accordance with their suggestion, we first obtain a vector X^sort^by arranging all gene expression signals for the same array in increasing order. Next we replace every entry in this array by its position (rank) in X^sort^counted from the smallest value. The idea behind this method is that ranks are invariant to any monotone transformation, implying a much more general model for the technological noise than the multiplicative array-specific random effect model. ### 3. Quantile normalization QUANT As discussed in \[[@B21],[@B22]\], this method is motivated by the idea that a quantile-quantile plot shows that the distribution of Mdata vectors is the same if the plot is a straight line in the direction of unit vector ![](1471-2105-6-120-i4.gif) but it is not the same otherwise. So we could make a set of data to have the same distribution if we projected the points of an M-dimensional quantile plot onto the diagonal. Much like as with the RANKmethod, this approach is applied to genes rather than arrays. We refer the reader to \[[@B21],[@B23]\] for more details. When working with the SJCRH data, this method was applied to probe feature level measurements. When working with simulated data, the method was applied to the levels of gene expression directly by processing them in exactly the same way. Authors\' contributions ======================= This work represents a truly collaborative endeavor. All members of the research team contributed equally to the design of this study, discussion of its technical issues and formulation of the net results. XQ was responsible for the computational component of the study. AB brought his biological expertise to the project. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 The effect of normalization with non-overlapping pairs of genes; ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 2 The effect of normalization procedures on the correlation structure of simulated data; ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 3 The effect of the quantile normalization on the distribution of the t-statistics across genes. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ We are grateful to anonymous reviewers whose comments have helped us improve the manuscript. We thank our colleague Cristine Brower for technical assistance. The research is supported in part by NIH Grant GM075299 and Czech Ministry of Education Grant MSM 113200008.	PubMed Central	2024-06-05T03:55:57.611451	2005-5-16	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1156869/", "journal": "BMC Bioinformatics. 2005 May 16; 6:120", "authors": [ { "first": "Xing", "last": "Qiu" }, { "first": "Andrew I", "last": "Brooks" }, { "first": "Lev", "last": "Klebanov" }, { "first": "Andrei", "last": "Yakovlev" } ] }
PMC1156870	Background ========== Phylogenetic footprinting is a computational approach to finding functional elements in DNA by comparing sequences between species. In an evolutionary context the \'footprint\' is the altered pattern of divergence resulting from a functional constraint. This is typically estimated as a reduced number of sequence changes. Phylogenetic footprinting was popularised by Tagle et al.\[[@B1]\] who demonstrated its utility in analysing the beta globin gene cluster. As larger amounts of non-coding DNA became available, easy to use tools such as Pipmaker were developed promoting wider adoption and acceptance of the technique\[[@B2],[@B3]\]. Interest in phylogenetic footprinting was increased when comparison of the mouse and human genomes showed a surprisingly high proportion of the genome could be aligned and that at least 1.5% of the genome was highly conserved non-repeat, non-protein coding DNA\[[@B4]\]. The conventional phylogenetic footprinting method relies on the comparison of a measure of evolutionary distance to determine conservation. Since the initial development of Pipmaker several groups have sought to improve upon the simple pair-wise percent identity measure of evolutionary distance. Chapman et al.\[[@B5]\] used alignment score from a multi species alignment. Margulies et al.\[[@B6]\] define a multi-species conserved site, using a parsimony and binomial probability method to score the importance of a match in each species. The binomial approach calculates the probability of observed or greater conservation with a reference sequence in a window given a neutral substitution rate (calculated from fourfold degenerate sites in codons) for that species. Equal levels of sequence conservation are given different scores, depending on their local neutral substitution rates. For example, a higher score is awarded where there is a higher neutral substitution rate. An averaging that halves the contribution of a species to the score at each node in the tree is then applied to reduce the weighting bias effect of non-symmetrical tree topologies. The parsimony model calculates a parsimony score for each column in the alignment and assigns a P-value based on simulations of parsimony scores for data generated under the HKY85 model calibrated with a previously determined phylogenetic tree and branch lengths. The methods of Margulies et al. and Chapman et al. both improve the sensitivity of detecting changes in divergence by including additional species in their calculation, reducing the probability of conservation due solely to chance. Likelihood methods for analysing sequence evolution have been widely adopted in the molecular evolution community for their advantages in consistency, sufficiency and the ability to naturally compare hypotheses. By definition, parsimony is minimum evolution, and will be biased towards underestimation, especially with longer branch lengths where the probability of multiple events at the same site increases. Central to the likelihood-based approach are continuous time Markov process models of substitution. The states in this Markov chain correspond to elements in the respective sequence alphabet, and will subsequently be referred to as motifs. The probability a motif changes (or remains the same) can be parameterised in many ways, e.g. according to biochemical attributes of the motifs involved. Motifs can be individual nucleotides or amino acids, biologically meaningful groups such as the triplets of nucleotides that make up a codon, or artificial groupings designed to capture dependencies such as dinucleotides. The Markov process for modelling motif changes is represented as a matrix of average relative rates of change and the matrix of substitution probabilities for a given time period is determined by a matrix exponentiation procedure. Details of these molecular evolution methods can be found elsewhere \[[@B7]-[@B9]\]. Methods using probabilistic evolutionary modelling for phylogenetic footprinting have been applied in two previous publications. Boffelli et al\[[@B10]\], who call their method phylogenetic shadowing, utilized fastDNAml\[[@B11]\] to determine mutation rates under the HKY85 model\[[@B12]\] for conserved and non-conserved regions from a training set of closely related primate data. These fitted models were then used for a likelihood ratio test to determine the probability of a given alignment column being in the conserved or non-conserved state and the likelihood ratios averaged over windows. The UCSC genome browser PhyloHMM track uses a combined HMM and probabilistic model of evolution to develop a HMM categoriser that can distinguish previously trained states (e.g. conserved vs. non-conserved). This model includes dependency of a nucleotide on the preceding site and implements a fully parameterised model analogous to a general time reversible model\[[@B13]\]. All of the previously discussed programs and strategies consider only measures of the expected number of substitutions per motif. This expected number of substitutions derives from the product of substitution rate and time, and reflects the combined influence of all mutational and selective processes. Although looking at low values of this statistic has proven an effective method for identifying regions that have a biological function that constrains their evolution, it does little to provide answers to the how and why of sequence divergence. By examining both the degree of conservation and individual components of the evolutionary processes, it should be possible to elucidate and infer the nature of evolutionary processes occurring at different sites in the genome. Using models that include terms for biologically relevant sequence changes, and tracking the change in estimates of the model terms along the sequence, regions where specific biological processes are predominating can be identified. The models can then be tailored to both the properties of the sequence being analysed (e.g. coding, or intergenic), and the effect of the biological process of interest. Vestige allows an examination of the temporal component of substitutions. Branches on a tree represent episodes of evolutionary time. Two regions may have experienced a similar amount of evolution when the entire tree is considered, but in very different ways. The extent to which such relative shifts in rate occur can be evaluated by examining the spatial distribution of a substitution statistic (such as length) for individual branches of a tree. Recent evidence of shifts in the evolutionary process between different mammalian lineages\[[@B14],[@B15]\] suggests that in order to accurately assess the spatial distribution of evolutionary processes, changes in the temporal distribution across the tree will also need to be taken into account. The combination of both the temporal and spatial partitioning therefore provides a useful tool with which to illuminate processes that may have occurred in a restricted region and stopped millions of years ago or continue to occur in current populations. Implementation ============== Vestige is implemented in the Python scripting language\[[@B16]\]. This allows rapid development and reuse of components by advanced users. The package utilises the PyEvolve toolkit\[[@B17]\], building upon its performance optimisations and capacity for flexible construction of existing and novel models of molecular evolution. The implementation of Vestige is as an extensible framework. Predefined scripts for footprinting (whole tree and per branch) of length, transition/transversion ratio, and Ka/Ks ratios are distributed with the package. These can be run as command line executable python scripts (unix or windows) or as GUI \"droplets\" under MacOSX. Command line arguments can be given to alter parameters such as the size of the analysis window and the step size the analysis window advances. These simple control scripts can be modified to alter the model of evolution, apply constraints on model terms and specify the terms from the model that are visualized, creating a new distributable and easy to use script. Vestige is implemented with parallelization at the per window level and will automatically use any available additional processors when run in an MPI environment. Two key sequence manipulations must be performed prior to analysis using Vestige. As alignment algorithms are a fast moving and specialized field, it was decided not to integrate any alignment algorithm into the Vestige package but require user-supplied alignments. Vestige also does not do any automated masking of repeats. Masking of repeats sequences such as transposable elements can be done by the user prior to footprinting by Vestige using programs such as RepeatMasker\[[@B18]\] to generate masked sequence prior to alignment. This may be desirable as repeats may not share the same phylogeny and may show a high level of conservation that can visually clutter the analysis. Alternatively, these can be annotated and their influence on the analysis assessed visually. Probabilistic phylogenetic modelling requires a phylogenetic tree. For reversible substitution models, an unrooted tree is used with the result that for two and three sequence cases only one unrooted tree is possible. In these cases, no phylogenetic inference is necessary and Vestige constructs the tree. In the case of more than three sequences the user can either supply a tree topology or Vestige will automatically construct a neighbour joining tree using the PyEvolve toolkit\[[@B19]\]. For some groups of species certain tree topologies are accepted by the community such as the Murphy et al.\[[@B20]\] mammal tree, and should be used as they represent a more robust reconstruction than is possible with a single sequence. Visualization is an important aspect of phylogenetic footprinting packages, as analysis often involves the integration of complex data from multiple sources. The drawing of Vestige results utilises the ReportLab\[[@B21]\] library, outputting PDF files that can be edited in many illustration packages and are suitable for high quality publications. Due to the familiarity of many potential users of the package with existing software that displays conserved regions as a peak (high Y value) on a graph, it was decided to mimic this format, even though the natural mapping of branch lengths would be to display short evolutionary conserved sequences as zero. As evolutionary distance varies from zero to (theoretically) positive infinity, a transformation that clearly displays all values and provides good discrimination between values close to zero was required. For this reason, distances are plotted as e^-length^. The standard statistics displayed in a Vestige run are the sum of the branch lengths in the tree for each window and the individual branch lengths for each branch and ancestral node in the tree. This allows examination of changes in the spatial pattern of divergence for each evolutionary episode, and visual assessment of the consistency of a signal between regions of the tree and the entire tree. In addition to the commonly used distance measurement, any parameter of the model of substitution can be estimated and plotted, either globally for the entire tree or individually for each branch, providing an indication of the spatial and temporal distribution of the process that is represented by that term. Although Vestige is implemented to allow the user maximum flexibility in model specification and parameter scoping, care must be exercised to avoid overreaching the capabilities of the model or data. Users should be wary of over parameterisation, which will result in estimates with very large variance. To some extent this problem can be addressed by using larger window sizes and/or global (whole tree) rather than local (branch specific) scope for parameters. Ultimately, there is a trade off between detecting individual short sequence elements and the accuracy of estimation. To aid in the interpretation of regions of conservation the GFF\[[@B22]\] and Genbank\[[@B23]\] annotation formats can be used to provide flexible multi-track decorations on the alignment to integrate data from multiple sources. The way that annotations are interpreted, grouped and displayed can readily be customized in user scripts. Annotation of biologically known features allows visual comparison between the level of divergence (or other parameters) of previously identified features and confirmed functional regions as a guide to importance of novel regions. The Vestige package is conceived as a data mining and hypothesis generation tool rather than a strict hypothesis testing framework. Correcting for multiple non-independent tests from a sliding window analysis is a difficult problem. Instead, we suggest an effective solution is visual inspection that draws on functional annotations as a reference coupled with an indication of the uncertainty of parameter estimates. From this perspective, novel predictions based on phylogenetic footprinting require validation. Accordingly, Vestige does not calculate test statistics for assessing the significance of conservation. Instead, the support for parameter estimates is provided in the form of a 95% confidence interval, estimated for branch specific parameters and global parameters. This is determined by fixing the values of all other parameters at their maximum likelihood estimates and calculating the likelihood for different values for the parameter until the point at which the likelihood ratio differs by the amount equivalent to a 5% probability in a one degree of freedom chi-squared test\[[@B24]\]. As is the case for all parametric statistical models, the accuracy of parameter estimates increases with increasing data (bigger window sizes). Users should remain mindful of this fact when inspecting all estimates, and that by chance one window in twenty are expected to lie outside this range. Further guidance to the importance of a conservation signal is given by indicating the top 5% of windows for the summed distance measure. The PyEvolve toolkit provides global and local optimisers for estimating model parameters. The global simulated annealing optimiser is capable of finding the global optima in complex functions with multiple optima, while the fast Powell optimiser is more prone to falling into local optima. To improve the speed of Vestige analysis, a global estimate of parameters is generated by simulated annealing optimisation of either the entire alignment or a random sample of columns drawn without replacement. This estimate is then used as a starting point by the Powell optimiser for rapid analysis or the more robust simulated annealing optimiser. To identify any windows that may suffer from poor optimisation a graph of the absolute value of the log-likelihood is included on the output. Abrupt discontinuities in the graph indicate optimisation problems. The user can then re-run the analysis with more robust optimisation settings. Vestige determines the frequency of the sequence motifs from the alignment. By default, the frequency of motifs in the entire alignment is used or the user can select to use the frequency in only the window being analysed. The ability to employ models of substitution with motif sizes greater than 1 raises the issues of generating invalid motifs, and inaccurate estimation of motif probabilities. Invalid motifs can be generated when the window is advanced into a different frame, and motifs are observed that were not in the original frame. For example, if the codon sequence (ATG) (AAG) is analysed in the second frame the (TGA) stop codon motif occurs, an invalid state for codon models of substitution\[[@B25]\]. Vestige addresses this problem using a flag that asserts that step should only be a multiple of the motif size. Inaccurate estimation of motif frequencies can occur when using models with motif sizes greater than one, as motif frequencies are normally counted only in the current frame. If a motif is rare in the frame used to calculate the motif frequency, but occurs frequently in the current analysis frame, the likelihood for windows in the current frame will be significantly decreased. Vestige therefore calculates an average motif frequency across all frames when using global motif probabilities for models with motif sizes greater than one. Results and discussion ====================== To demonstrate the functionality and broad utility of the Vestige package we present two analyses, the SCLlocus analysed with a dinucleotide model, and exon 11 of 5 primate BRCA1sequences \"footprinted\" for the ratio of Ka/Ks model terms using a codon model. The alignment of human, mouse, rat and dog totalling 128 kb from the SCLlocus and its annotations in GFF format are those used by Chapman et alallowing direct comparison with their results\[[@B5]\]. The BRCA1alignment is identical to that used by Huttley\[[@B26]\] with redundant gaps removed. The results for globally summed branch lengths across the SCLlocus provided similar results to those obtained by Chapman et al. All of the experimentally determined and biologically important conserved regions in this sample alignment were detected (figure [1](#F1){ref-type="fig"}). Like other phylogenetic footprinting methods, Vestige fails to find any conservation at the region 8--9 kb upstream of the SCLstart site. This site represents an altered chromatin structure in mouse and may be due to a non-DNA sequence specific change in chromatin. Graphing statistics for individual branches on the tree provides a mechanism for discriminating between artefactual and biological causes of spatial heterogeneity. Artefacts can arise due to properties of the data and properties of the statistical models and numerical algorithms. Missing data are addressed by calculating the likelihood over all possible states and gaps are conventionally treated as missing data in the likelihood framework. Both missing data in sequences and gaps in multiple alignments have a structure to their occurrence that impacts on parameter estimates: gaps occur in trees in a taxonomically structured way reflecting their evolutionary origin; and, both gap and missing data symbols tend to occur in patches. When the patch size is greater or equal to the window size, the extreme case, one or more lineages without any information result and the true total tree branch length is equivalent to that of a smaller number of taxa. Yet because the optimiser still attempts to estimate values for these parameters, optimiser behaviour can introduce a systematic error -- such as zero branch length creating a region of conservation -- whose pattern depends on the optimiser chosen. The introduction of a gap term into the model doesn\'t necessarily solve this issue as they tend to be rare and insertions reversing gaps are highly improbable, causing short branch lengths. As a solution to this problem, Vestige estimates and displays a 95% confidence interval for individual branch length estimates. The minimum and maximum of the 95% confidence interval are then displayed rather than the optimised value. Windows where a large proportion of N\'s creates a broad (or in the extreme case infinite) confidence interval will be displayed as white space. The user can look along the graph to identify either well supported regions of conservation (the lower graph) or well supported regions of divergence (the upper hanging graph). While this capacity for establishing artefacts as the basis for a spatial distribution is essential, more interesting is the situation in which there are lineage specific effects that are biological in origin. Biological origins for spatial heterogeneity can, in principal, originate from spatial fluctuations in a biological process that is common to all lineages or unique to a subset of lineages. To date there has been little consideration of the latter case but several studies demonstrate this is a real topic of interest. Direct evidence of differences in DNA repair between rodents and humans\[[@B27]\], and indirect evidence between rat and mouse\[[@B14]\] and between eutherian and non-eutherian mammals\[[@B15]\] establish the existence of differences in DNA mutation and repair between lineages. Furthermore, the existence of substitution rate heterogeneity across genomes\[[@B28],[@B29]\] illustrates the existence of spatially localised differences in mutation and repair, local effects that can change character over evolutionary time. Consequently, inference regarding the property of a single region based on a summary statistic calculated for the whole tree benefits from the ability to establish that the fundamental pattern is consistent across the majority of branches on the tree. Of course, inconsistency across the tree can also be of interest since they may indicate changes in local mutagenic environment providing insight into shifts in the boundaries of biological features. An example of one such change is given in figure [1](#F1){ref-type="fig"}, where a region that is exonic in mouse and rat is intronic in dog. This region is conserved in all lineages except that leading to the dog. The tight confidence intervals on these estimates for this region across all lineages, coupled with the strength of signal on the lineage leading to the common ancestor of mouse and rat, suggests loss of this exon occurred on the lineage leading to dog. These observations suggest the hypothesis that the open chromatin region observed in mouse will have been lost or significantly altered in dog. Inconsistency of the spatial pattern between branches or clades can therefore facilitate identifying key regions of biological differentiation. We have broadened the definition of a footprint to include spatial changes in estimates of any model term. These terms may be either global in scope or locally estimated for each branch. To demonstrate the utility of this broader definition we have footprinted primate BRCA1for signals of adaptive evolution. This was chosen as an example as it demonstrates the ability to work with different sequence data types and with biologically meaningful model terms other than branch length. Implementation of non-standard annotations requires some scripting and the python script for this example provides a template for other applications that may require fine control over the behaviour of Vestige (script included in distribution). The Ka/Ks statistic is the ratio of non-synonymous to synonymous substitution rates, and represents the impact of natural selection. A Ka/Ks ratio of less than one indicates purifying selection, a ratio equal to one indicates selective neutrality, and a ratio greater than one is evidence of positive or adaptive evolution. The Ka/Ks ratio is modelled by adding a term (omega) to the standard codon model that applies when an instantaneous substitution results in a change in the amino acid coded for by the codon\[[@B30]\]. Due to the requirement of a moderate sized data set for omega estimation a 100 codon (300 bp) window and global (whole tree) scoping of the omega parameter were used. The analysis of BRCA1indicates there are three regions that have Ka/Ks ratios greater than one, putatively indicative of adaptive evolution (figure [2](#F2){ref-type="fig"}). Two of these three regions have estimated 95% confidence intervals that do not contain the value 1, suggesting the hypothesis that adaptive modifications to BRCA1 function have occurred within them. Both regions fall within the RAD51binding domain and suggest that modulation of RAD51binding may be the driving force for BRCA1adaptive evolution in primates. This result is consistent with the inference of Huttley et al.\[[@B31]\] and variants in RAD51that modify breast cancer risk in BRCA1mutation carriers\[[@B32]\]. Conclusion ========== Vestige provides a flexible, open platform for phylogenetic footprinting that expands the range of model terms and hence biological processes that can be evaluated. The framework facilitates visualisation of results from the increasingly rich probabilistic models of molecular evolution aimed at detecting and categorizing regions of the genome. Availability and requirements ============================= Project name:Vestige Project home page:<http://cbis.anu.edu.au/software/> Operating system(s):Platform independent Programming language:Python Other requirements:Python 2.3 or higher, PyEvolve 0.89 or higher, Reportlab, and Numeric Python. License:GNU GPL Authors\' contributions ======================= MJW designed and implemented the software. PM wrote the drawing code, refactored the code during development and contributed to software and algorithmic design. GAH wrote a prototype (sans visualisation) that stimulated the current project, and provided guidance in the application of the underlying molecular evolutionary modelling methods and PyEvolve toolkit. All authors read and approved the final manuscript. Additional data files ===================== • File Name: Vestige.tar.gz • File Format: gzip compressed tar archive • Description: The python script files, documentation and data for the software described in this paper. • File Name: scl\_vestige\_dinucleotide.pdf • File Format: Adobe Portable Document Format • Description: The complete output of vestige analysis of the scl region, a subsection of which is presented in figure [1](#F1){ref-type="fig"}. See figure [1](#F1){ref-type="fig"} for details. • File Name: Vestige\_MacOX\_Droplets.tar.gz • File Format: gzip compressed tar archive • Description: Binary distribution of standalone MacOSX graphic user interface droplets, that can be used for running vestige by dragging an alignment and optionally a tree and gff file onto the application icon in the MacOSX finder. Requires MacOSX10.3 • File Name: Cogent0\_89.tar.gz • File Format: gzip compressed tar archive • Description: Source code distribution of the PyEvolve package Acknowledgements ================ Ms Stella Lee contributed to the implementation of an early prototype. MJW is supported by an Australian Research Council APD Fellowship. This research used facilities provided by the Australian Partnership for Advanced Computing. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Footprinting the SCL gene. Phylogenetic footprinting of the genomic region around the SCL gene. The alignment of Chapman et al.\[5\], with their experimentally determined regions of biological importance annotated, was footprinted in 100 bp windows with a 25 bp step using a dinucleotide model of evolution based on the HKY85 model\[12\]. This model contains terms for the frequency of each dinucleotide (taken from the complete 139 kb alignment) and for the transition/transversion ratio which is applied when the difference between the dinucleotides is a transition. The total branch length summed across the tree is plotted as e^-length^in red and the absolute value of the log likelihood (smaller is better) in blue. The yellow line indicates the level of conservation of the top 5% of windows for the entire 139 kb alignment. Local branch lengths are presented in 5 panels aligned with a stepped dendrogram representation of the phylogenetic tree. Annotations for each species are displayed below the graph, with the lower black lines representing sequence and white space gaps. Coloured annotations in the upper track are described below the mainplot. The fourth track is the derived ancestor of mouse and rat, and therefore has no sequence or annotation. The display of local branch lengths consists of a plot of the length at the lower bound of the 95% confidence estimate in salmon, and the upper bound of the 95% confidence estimate in green. The 95% confidence interval estimate for the branch length is represented by the white space between these graphs. Regions of high confidence conservation can be identified by looking for peaks in the lower salmon graph, and conversely regions of high confidence divergence can be identified by identifying hanging peaks in the upper green graph. Regions where no reliable branch length estimate can be given will appear white. Individual branch lengths can be compared to changes in annotation between branches. For example, the grey boxed region highlights a high confidence signal of divergence in dog between 75000 and 75300. This region correlates with part of an exon and a region of open chromatin in mouse, but is intronic in dog. This suggests that the open chromatin region will be altered or not present in dog, potentially altering regulation and function of this gene. Full analysis of the entire 139 kb region around the SCL gene is provided as a supplementary file. ::: ![](1471-2105-6-130-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Footprinting the Ka/Ks ratio in primate BRCA1. A DNA alignment of five partial primate sequences from exon 11 of BRCA1 footprinted for adaptive evolution. A codon model of evolution with a model parameter for replacement changes\[25\] was \"footprinted\" in 300 base (100 codon) windows and 30 base steps. Red lines indicate the 95% confidence interval for the omega replacement parameter that is an estimate of the Ka/Ks ratio. Ka/Ks \> 1 is indicative of adaptive evolution. Two regions (around 1850 and 2400 bp) have 95% confidence intervals for omega that do not include 1, suggesting adaptive evolution is occurring within them. Note that although plotted as single lines in the middle of the window range the 300 bp windows overlap and a single region or site can affect the parameter estimate for multiple adjacent windows. Annotations of protein-protein interaction domains (blue boxes) and phosphorylation sites (red diamonds) are derived from Deng\[33\]. Sequences: Human 961-3798 of NM007294, Chimpanzee 150-2987 of AF019075, Gorilla 150-2987 of AF019076, Orangutan 150-2987 of AF019077 and Rhesus 150-2984 of AF019078. Scale is in bases and refers to gapped alignment positions. ::: ![](1471-2105-6-130-2) :::	PubMed Central	2024-06-05T03:55:57.615248	2005-5-29	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1156870/", "journal": "BMC Bioinformatics. 2005 May 29; 6:130", "authors": [ { "first": "Matthew J", "last": "Wakefield" }, { "first": "Peter", "last": "Maxwell" }, { "first": "Gavin A", "last": "Huttley" } ] }
PMC1156871	Background ========== The ability to identify single nucleotide changes in DNA is a fundamental requirement in many fields of biological research. The identification and characterisation of naturally-occurring single nucleotide polymorphisms (SNPs) underlies a vast body of work on genetically-linked disorders, diagnosis and risk prediction \[[@B1]-[@B4]\] as well as being important in genomic mapping and population genetics \[[@B5]-[@B8]\]. Identification of point mutations is of equal importance to many researchers, for roles as diverse as identifying specific alterations caused by random mutagenesis screens \[[@B9],[@B10]\] to validation of the fidelity of sequences amplified by PCR. For labs studying SNPs or point mutations, identification of these can be a time-consuming and error-prone process, particularly if novel changes are being investigated. In some cases, software such as the Staden package \[[@B11],[@B12]\] or Sequencher \[[@B13]\] may provide a suitable solution. However these are sophisticated and multifunctional programs, and can prove overly complex for simple sequence comparisons. Consequently, many small-scale projects may rely solely on manual analysis, for example simply carrying out a direct text comparison of the processed sequence to a known reference. This manual approach is affected by variations in sequence quality and incorrect base calling, and may also miss heterozygous bases if, for example, the wild-type peak is higher that the additional peak. To address these issues and to provide a simple and efficient way to accurately identify sequence changes, we have developed a web-based application which compares DNA sequence chromatograms directly. SeqDoC (Sequence Difference of Chromatograms) is freely accessible, very easy to use, and highlights differences characteristic of single base changes, including heterozygous SNPs and insertions and deletions. Implementation ============== Read in chromatograms --------------------- Two ABI sequence chromatograms, one a reference and the other the test sequence, are the only user-supplied data. These are uploaded through a web form and the sequence traces and other relevant data extracted using the Perl ABI.pm module \[[@B14]\]. The sequence traces for each channel (i.e. A, C, G and T) are stored as individual arrays within the chromatogram object. Blank sequence at the beginning and end of each chromatogram (resulting for example from sequencing of a PCR product, when the trace continues past the end of the template) are removed by deleting the terminal values from the traces where all channel values are less than 50. In tests, we established that a filter value of as low as five resulted in the removal most blank terminal sequence, while a value of as high as 500 still retained virtually all genuine sequence; we therefore used 50 as an appropriate intermediate value. Normalize traces ---------------- Comparison of the test to the reference sequence is performed by subtraction of trace values, so it is necessary to normalize the trace values so that a sequence run with strong signal can be meaningfully compared to one with weaker signal. Normalisation is performed for each channel individually, and scales each data point so that the local mean value for that channel is 100 (local being defined as 500 data points prior to the point being scaled, the point itself and 499 points after). The mean value of those local points for the channel is calculated and divided by 100 to give a scaling factor, and the point being normalized is then scaled by being divided by this factor. Special exceptions are made for the initial 500 and final 499 values of the trace, where there are either not 500 values before, or not 499 after, the point being normalized. For these two cases, the mean is based on as many points as are present between the point and the end of the sequence on one side, while still using 500 on the other side. Align traces ------------ Due to variations arising both from the sample and the sequencing matrix, the start position of the sequence traces and the base separation within the traces may differ between the test and the reference sample. The software compensates for this by automatically calculating the optimal start alignment combined with continual fine adjustment throughout the length of the sequences to maintain alignment of the test and reference sample. To identify start point offsets, the software tests a range of initial alignments, from -200 to +200 data points, corresponding to approximately +/- 20 bases of sequence. The offset which results in the best alignment (the smallest total value of the absolute differences between the test and reference traces for all four channels) is used for subsequent analysis. This is implemented by the addition or deletion of \'spacer\' data points at the beginning of the test sequence. Ongoing fine adjustment of the sequences is achieved by the addition or removal of individual data points from the test sequence as required. The sequences are sampled every five data points, and difference scores calculated for the subsequent 30 data points. If that difference score is reduced by the insertion or deletion of a single data point, then the test traces are adjusted accordingly (by either duplication or removal of the data point at the test position). Calculate differences --------------------- Following normalisation and alignment of the sequences, a \'difference profile\' is calculated. The trace values of the test sequence are subtracted from the equivalent values for the reference sequence for all four channels, and the resulting values are passed through an algorithm which highlights changes characteristic of base substitutions, while reducing random noise. This is achieved by squaring the difference value and multiplying the result by the square root of the sum of the differences of other channels which change in the opposite direction. The overall outcome of this process is to enhance the signal given by large differences with at least one channel changing in the opposite direction, which is characteristic of base substitutions, while minimising the signal from small unidirectional changes (typical of signal noise). Difference profiles are calculated for all four channels and the results overlaid in the final output. Generate trace and difference images ------------------------------------ User output is provided in the form of three aligned images: aligned sequence chromatograms for the reference and test sequences, and a similarly aligned difference profile. These outputs are based on the normalized, aligned traces generated during earlier stages of the analysis. The difference trace is typically primarily flat, with a large bidirectional peak at the point of any base changes between the sequence. Other patterns are possible, depending on specific features of the test and reference sequences, and are discussed in more detail below. The three images are generated by the Perl GD::Graph module, and are returned to the user as a web page. Identification of the site of base substitutions is simply a matter of examining the difference trace for the bidirectional peaks mentioned above; by examining the aligned test and reference sequences at the point of these peaks, single base changes between the two sequences can be rapidly and simply identified. Staden comparison ----------------- The Staden 1.5.3 Windows installer was downloaded from SourceForge and installed on a PC running Windows XP Pro. Tracediff comparison was performed through Pregap4 using the following modules in order: General Configuration, Initialise Experiment Files, Reference Traces & Sequences, Trace Difference. All user-definable parameters were left at their default values (except that \'Write trace differences out to disk\' was selected). We used Gap4 to both align and view the initial and tracediff-generated traces as well as to carry out trace subtraction directly. The reference and test sequences (and difference trace where appropriate) were assembled into a contig, which was opened in the Trace display window. A subtracted trace was generated using the Diff button. Results and discussion ====================== All scaling factors, cutoff filters and algorithms described in the methods section were chosen after testing of multiple settings as giving the clearest identification of single base changes and best retention of genuine data while minimising the signal resulting from noise. The process was initially optimised using sequences covering known polymorphisms in different regions of the human melanocortin 1 receptor gene \[[@B15]\]. In all cases the polymorphisms were clearly visible in the difference trace. The software has since been successfully used to test for single base changes in several hundred sequence chromatograms. Extracts from typical output traces are shown in figures [1](#F1){ref-type="fig"} and [2](#F2){ref-type="fig"}, which identify homozygous and heterozygous polymorphisms respectively. The difference trace does not differentiate between these two different substitutions (although the size of the double peak is typically smaller for heterozygous sites). Instead it makes it rapidly obvious to an investigator where the sites of difference are, and allows the investigator to make the final decision about the nature of the substitution. SeqDoC is also able to highlight the occurrence of single base insertion or deletion events. Figure [3](#F3){ref-type="fig"} shows the typical pattern from a deletion; at the point of the deletion, there is a major difference between the two chromatograms, which is gradually eliminated by the alignment algorithm bringing the two chromatograms back into phase. Insertion events give similar patterns. As with substitutions, no interpretation of the pattern is calculated; the role of the software is to alert the investigator to the location of the change, not to characterise it. The automatic start alignment process means that it should not be necessary to use the same primer to sequence the reference and test sequences, providing that the two different primers produce sequences which start within approximately 20 nucleotides of each other. Alternatively, it would be possible to manually edit the raw chromatograms to bring the start points of the two sequences into approximate alignment. We do not believe that the comparison will work for samples sequenced in opposite directions, using the reverse complement of one or the other. Sequencing chemistry is such that peak heights are typically affected by the local sequence composition \[[@B16],[@B17]\], and while this is consistent for samples sequenced in the same direction from different primers, it is not true for those sequenced in opposite direction. Because of the normalisation and noise reduction algorithms built into the software, it is relatively resilient to poor quality sequence. Typically, if the sequence quality is sufficient for an investigator to unambiguously identify the base call, it is good enough for automatic identification of sequence differences. Most problems with sequence quality only occur at the end of the sequence run, although unincorporated dye terminators may cause \'dye blobs\' at the beginning of the sequence, which can partially mask base changes occurring at the same site. Examples of the output produced in these cases, along with full instructions on the use of SeqDoC, are provided on our website at <http://research.imb.uq.edu.au/seqdoc/>. The Staden software package \[[@B11]\] is an established, well-supported and widely used sequence analysis package, and has functions (such as \'tracediff\') for direct comparison of chromatogram traces analogous to those provided by SeqDoC. It can also display trace subtractions through the Gap4 program. We have therefore evaluated the performance of SeqDoC using Staden as a benchmark. Although the principle advantage of SeqDoC over Staden is that it is specifically designed and optimized for the single purpose of chromatogram comparison, and therefore provides a much simpler user interface, we also believe that it offers advantages in normalisation, alignment and, particularly, sensitivity. On the other hand, Staden can allow much higher throughput, since multiple sequences can be analysed at once; data can also be saved and analysed in detail with multiple additional functions. Figure [4](#F4){ref-type="fig"} shows a comparison of a weak and strong signal trace with Staden and SeqDoC respectively. The SeqDoC local normalisation algorithm means that the trace heights for both are very similar, and therefore more readily comparable. Although the Y-scale can be altered in the Staden trace display window to compensate for this, the problem is then shifted to the beginning of the trace where the sequence is proportionally much stronger. The effects of misalignment are observed earlier in the same comparison, where the Staden difference trace shows a characteristic cyclical pattern which is not observed in the SeqDoC alignment (figure [5](#F5){ref-type="fig"}). Although it is impossible to tell whether either of these factors significantly compromises the performance of tracediff, we suspect that alignment at least will have some influence, since any misalignments will introduce unnecessary noise into the difference trace. The main functional benefit of using SeqDoC over Staden is that of sensitivity, particularly for identifying heterozygous peaks or when using either weak or poor quality sequence. The heterozygous base shown in figure [2](#F2){ref-type="fig"} is identified by tracediff, but the output suggests that it is a direct replacement rather than a mixed base, while another (figure [6](#F6){ref-type="fig"}) is missed by tracediff altogether. The latter example occurs in a weak strength trace, which is compensated for by the SeqDoC normalisation. Similarly tracediff can miss differences in noisy sequence; SeqDoC is more robust, because calls are made by visual inspection and the difference profile is used only to draw the investigator\'s attention to areas of difference. For example, figure [7](#F7){ref-type="fig"} shows a comparison using poor quality reference sequence data. Although the difference trace is consequently noisy, it still highlights a heterozygous substitution in the test sequence. In summary, SeqDoC proves a lightweight but effective substitute to Staden for sequence trace comparisons. While Staden is a more appropriate choice for applications where high throughput is the main priority, SeqDoC provides a better solution when sensitivity, specificity or simplicity are more important considerations. Conclusion ========== SeqDoC is a very easy to use web-based application which rapidly highlights differences between ABI sequence chromatograms, including substitution and insertion/deletion events. It uses chromatograms directly, rather than extracted text-based sequence data, so eliminating errors introduced by base calling software and allowing identification of heterozygous substitutions which might otherwise be missed. No user intervention or adjustment is required for processing, with all normalisation, alignment and noise reduction being carried out automatically; on the other hand the ultimate decision on the specific change identified remains with the investigator. SeqDoC is free and requires no training to use, and is ideally suited for use by researchers carrying out small scale SNP analysis or mutagenesis experiments. It can also be used to rapidly screen PCR-amplified products for introduced mutations. Availability and requirements ============================= Program name: SeqDoC Project home page: <http://research.imb.uq.edu.au/seqdoc/> Source code: <http://research.imb.uq.edu.au/seqdoc/seqdoc.pl> or [additional file 1](#S1){ref-type="supplementary-material"}. Operating system(s): Platform independent Programming language: Perl CGI Other requirements: Requires Perl CGI, GD::Graph and ABI modules License: None for web access, GNU GPL for source code Any restrictions to use by non-academics: No restrictions Supplementary Material ====================== ::: {.caption} ###### Additional File 1 The perl source code for SeqDoC (seqdoc.pl) is available with the online version of this article ([additional file 1](#S1){ref-type="supplementary-material"}). Instructions for use of the program can be obtained using the \'perldoc\' command. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ This work was supported by the Australian Research Council Special Research Centre for Functional and Applied Genomics. We thank Dr R.A. Sturm for his initial conceptual suggestions, W. Chen for providing test data and Dr S.M. Grimmond for reviewing this article. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Difference pattern from a single base substitution. Replacement of one base by another (here a G for an A at position 251 in the reference sequence (top trace)) means a major decrease in the value of one channel and a similar increase in another. This causes a bi-directional peak in the difference profile. ::: ![](1471-2105-6-133-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Difference pattern from a heterozygous SNP. A multiple peak in the test sequence (bottom trace) is characterised by an increase in the value in one channel (in this case the A at position 195 in the test sequence), and typically a decrease in the original channel (G), therefore giving a bi-directional peak in the difference profile similar to a direct substitution. ::: ![](1471-2105-6-133-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Difference pattern from a single base deletion. When a base is deleted (in this case the G at position 254 in the reference sequence) the resulting phase shift in the test sequence will result in major differences between the traces until the software compensates by bringing them back into alignment. ::: ![](1471-2105-6-133-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Comparison of weak and strong sequence traces. Weak and strong sequence traces are successfully aligned by both Staden\'s Gap4 program (fig 4a) and SeqDoC (fig 4b). The local normalisation algorithms of SeqDoC mean that the two traces are displayed with comparable peak heights (weak trace at the top), and possibly results in a less noisy difference profile. Both figures show the same region of the same sequences. ::: ![](1471-2105-6-133-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Effects of sequence misalignment. A minor misalignment of test and reference sequences by Gap4 causes a cyclical pattern in the difference profile (fig 5a, bottom trace) which increases signal noise. The fine adjustment algorithm of SeqDoC ensures that sequences are properly aligned (fig 5b) and eliminates this noise signal. ::: ![](1471-2105-6-133-5) ::: ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Missed heterozygous base call. Tracediff does not identify the heterozygous base at position 191 in the test sequence (upper trace, fig 6a), possibly because of the weak signal strength and noise in the difference profile. The normalisation and difference profile optimisation algorithms built into SeqDoC give a very strong signal to noise ratio for this change in the difference profile (fig 6b) and make the substitution obvious. ::: ![](1471-2105-6-133-6) ::: ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Poor sequence quality. A poor quality sequence (in this case the reference sequence, top) unavoidably causes a noisy difference profile. However SeqDoC still successfully highlights nucleotides which differ between the sequences. In this case the C at position 390 in the reference sequence becomes a mixed T/C peak at position 398 in the test sequence. ::: ![](1471-2105-6-133-7) :::	PubMed Central	2024-06-05T03:55:57.618174	2005-5-31	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1156871/", "journal": "BMC Bioinformatics. 2005 May 31; 6:133", "authors": [ { "first": "Mark L", "last": "Crowe" } ] }
PMC1156872	Background ========== Colorectal cancer (CRC) is one of the leading causes of death in the world. This malignancy is also one of the most serious complications of inflammatory bowel disease (IBD), including ulcerative colitis (UC) and Crohn\'s disease \[[@B1]\]. Long-term UC patients have an increased risk of developing CRC compared with the general population \[[@B2]\]. The precise mechanisms of the IBD-related carcinogenesis process are largely unclear, although it is generally assumed that IBD-related carcinogenesis occurs as a result of chronic inflammation \[[@B3]\]. Several agents, such as folic acid, short chain fatty acid (butyrate), ursodeoxycholic acid, and 5-aminosalicylic acid, have been suggested to be useful for prevention of CRC in UC \[[@B4]\]. Epidemiological studies have shown that prolonged use of aspirin is associated with a reduced risk of CRC \[[@B5]\]. Consistent with these data, several non-steroidal anti-inflammatory drugs (NSAIDs), including cyclooxygenase (COX)-2 inhibitors, suppressed the development of chemically-induced colon carcinomas in rats \[[@B6]\] and intestinal polyps in Minmice with a nonsense mutation of the Apcgene \[[@B7]\]. In addition, clinical trials have demonstrated that a NSAID sulindac causes regression of adenomas in patients with familial adenomatous polyposis \[[@B8]\]. Nimesulide (4-nitro-2-phenoxymethanesulfonanilide), a selective inhibitor of COX-2, belonging to the sulfonamide class \[[@B9]\], is less ulcerogenic than other NSAIDs \[[@B10]\], and suppresses the formation of aberrant crypt foci (ACF), being putative precancerous lesions of the colon cancer, induced by a colon carcinogen, azoxymethane (AOM) in rats \[[@B11]\]. Moreover, this COX-2 inhibitor could effectively reduce the development of intestinal polyps in Minmice \[[@B12]\]. Peroxisome proliferator-activated receptors (PPARs?) are ligand-activated transcription factors, belonging to the nuclear hormone receptor superfamily \[[@B13]\]. Three PPAR isotypes, PPARα, PPARδ(β), and PPARγ, have been identified. PPARγ is highly expressed in fat tissue, and play important roles in adipocyte differentiation and lipid storage \[[@B14]\]. PPARγ is also expressed in a number of epithelial neoplasms, such as cancers in colon, breast, and prostate \[[@B13]\]. PPARγ ligands, including thiazolidinediones (troglitazone and rosiglitazone) and tyrosine analogue (GW7845), can induce apoptosis and adipogenic differentiation, and inhibit tumor growth both in vitroand in vivostudies \[[@B15]-[@B17]\]. We previously reported that pioglitazone, bezafibrate or troglitazone in diet are able to suppress ACF formation induced by dextran sodium sulfate (DSS)/AOM in the rat colon \[[@B18]\]. Osawa et al. \[[@B19]\] confirmed our findings by demonstrating that ligands for PPARγ (troglitazone, rosiglitazone, and pioglitazone) suppress the occurrence of colonic tumors in mice initiated with AOM. Niho et al. \[[@B20]\] also demonstrated that ligands for PPARα (bezafibrate) and PPARγ (pioglitazone) suppress intestinal polyp formation in Apc-deficient mice. Moreover, PPARγ could suppress β-catenin levels and colon carcinogenesis during the early steps of tumor formation \[[@B21]\]. On the other hand, high doses of troglitazone and rosiglitazone can promote polyp formation in the Minmouse colon \[[@B22],[@B23]\]. For understanding the pathogenesis of IBD and IBD-related CRC, several animal models have been established. Most used is a mouse model with DSS \[[@B24]\]. Modifying effects of several xenobiotics on IBD-related colon carcinogenesis were reported \[[@B25]\] in animal models of IBD. However, the colitis model using DSS with or without carcinogen needs to a long period and repeated administration of DSS to induce colitis and colitis-related CRC. Recently, an endogenous anti-inflammatory PPARγ pathway was suggested in the intestine, which was found in PPARγ-deficient mice \[[@B26],[@B27]\]. To search novel and effective chemopreventive agents against IBD-related CRC, we recently have developed a novel colitis-related CRC mouse model, in which large bowel adenocarcinomas occur within 20 weeks and their histology and biological characteristics are resemble to those found in human cased \[[@B28]\]. As a part of our search for a safer chemopreventive agent for colitis-related colon cancer, in the present study we examined the chemopreventive ability of nimesulide, troglitazone, and bezafibrate using our mouse colon carcinogenesis model for colitis-related colon carcinogenesis \[[@B28]\]. Methods ======= Animals, chemicals and diets ---------------------------- Female Crj: CD-1 (ICR) (Charles River Japan Inc., Tokyo, Japan) aged 5 weeks were used in this study. They were maintained at Kanazawa Medical University Animal Facility according to the Institutional Animal Care Guidelines. The mice were quarantined for the first 7 days then randomized by body weight into experimental and control groups. All animals were housed in plastic cages (five or six mice/cage) with free access to drinking water and a pelleted basal diet, CRF-1 (Oriental Yeast Co., Ltd., Tokyo, Japan), under controlled conditions of humidity (50 ± 10%), light (12/12 hour light/dark cycle) and temperature (23 ± 2°C). A colonic carcinogen AOM was purchased from Sigma Chemical Co. (St. Louis, MO, USA). DSS with a molecular weight of 40,000 was purchased from ICN Biochemicals, Inc. (Aurora, OH, USA). DSS for induction of colitis was dissolved in distilled water at a concentration of 2% (w/v). Nimesulide was purchased from Sigma Chemical Co. (St. Louis, MO, USA). Troglitazone and bezafibrate were kindly supplied by Sankyo Co. (Tokyo, Japan) and Kissei Pharmaceutical Co. (Matsumoto, Japan), respectively. Experimental diet containing nimesulide (0.04%, w/w), troglitazone (0.05%, w/w) or bezafibrate (0.05%, w/w) was prepared every week by mixing the respective compound in powdered basal diet CRF-1. The dose levels were determined on the basis of previous studies \[[@B18],[@B29]\]. Experimental procedure ---------------------- A total of 71 female ICR mice were divided into 7 experimental and control groups (Figure [1](#F1){ref-type="fig"}). Mice in groups 1 through 4 were given a single intraperitoneal injection of AOM (10 mg/kg body weight). Starting one week after the AOM injection, animals in group 1 were administered to 2%DSS in drinking water for 7 days, and then followed without any further treatment for 15 weeks. Mice in groups 2, 3, and 4 were fed the diets containing 0.04% nimesulide, 0.05% troglitazone, and 0.05% bezafibrate, respectively, for 14 weeks, starting 1 week after the stop of DSS administration. Animals in groups 5, 6, and 7 were respectively given the diets containing 0.04% nimesulide, 0.05% troglitazone, and 0.05% bezafibrate alone for 14 weeks. Group 8 was given a single dose of AOM. Group 9 was given 2% DSS for 7 days. Group 10 consisted of untreated mice. All animals were sacrificed at the end of the study (Week 17) by ether overdose. Their large bowels were flushed with saline, excised, measured their length (from ileocecal junction to the anal verge), cut open longitudinally along the main axis, and then washed with saline. The large bowels were macroscopically inspected, cut, and fixed in 10% buffered formalin for at least 24 hours. Histopathological examination was performed on paraffin-embedded sections after hematoxylin and eosin (H & E) staining. Colonic mucosal dysplasia (mild- and severe-graded) was diagnosed according to the criteria described by Ridell et al. \[[@B30]\] and Pascal \[[@B31]\]. Colonic neoplasms were diagnosed according to the description by Ward \[[@B32]\]. Immunohistochemistry -------------------- Immunohistochemistry for the proliferating cell nuclear antigen (PCNA), apoptotic nuclei, β-catenin, COX-2, inducible nitric oxide synthase (iNOS), and nitrotyrosine was performed on 4-μm-thick paraffin-embedded sections from colons of mice in each group by the labeled streptavidin biotin method using a LSAB KIT (DAKO Japan, Kyoto, Japan) with microwave accentuation. The paraffin-embedded sections were heated for 30 min at 65°C, deparaffinized in xylene, and rehydrated through graded ethanol at room temperature. A 0.05 M Tris HCl buffer (pH 7.6) was used to prepare solutions and for washes between various steps. Incubations were performed in a humidified chamber. For the determination of PCNA-incorporated nuclei, the PCNA-immunohistochemistry was performed according to the method described by Watanabe et al. \[[@B33]\]. Apoptotic index was also evaluated by immunohistochemistry for single stranded DNA (ssDNA) \[[@B33]\]. Sections were treated for 40 min at room temperature with 2% BSA and incubated overnight at 4°C with primary antibodies. Primary antibodies included anti-PCNA mouse monoclonal antibody (diluted 1:50; PC10, DAKO Japan), anti-ssDNA rabbit polyclonal antibody (diluted 1:300, DAKO Japan), anti-β-catenin mouse monoclonal antibody (diluted 1:1000, Transduction Laboratories, Lexington, KY, USA), anti-COX-2 rabbit polyclonal antibody (diluted 1:50, IBL Co., Ltd., Gunma, Japan), anti-iNOS rabbit polyclonal antibody (diluted 1:1000, Wako Pure Chemical Industries, Ltd., Osaka, Japan), and anti-nitrotyrosine rabbit polyclonal antibody (diluted 1:500, Upstate Biotechnology, Lake Placid, NY, USA). To reduce the non-specific staining of mouse tissue by the mouse antibodies, a Mouse On Mouse IgG blocking reagent (Vector Laboratories, Inc., Burlingame, CA, USA) was applied for 1 h. Horseradish peroxidase activity was visualized by treatment with H~2~O~2~and 3,3\'-diaminobenzidine for 5 min. At the last step, the sections were weakly counterstained with Mayer\'s hematoxylin (Merck Ltd., Tokyo, Japan). For each case, negative controls were performed on serial sections. On the control sections, incubation with the primary antibodies was omitted. Intensity and localization of immunoreactivities against all primary antibodies used were examined on all sections using a microscope (Olympus BX41, Olympus Optical Co., Ltd., Tokyo, Japan). The PCNA and apoptotic indices were determined by counting the number of positive cells among at least 200 cells in the lesion, and were indicated as percentages. Each slide for β-catenin, COX-2, iNOS, and nitrotyrosine immunohistochemistry was evaluated for intensity of immunoreactivity on a 0 to 4+ scale. The overall intensity of the staining reaction was scored with 0 indicating no immunoreactivity and no positive cells, 1+ weak immunoreactivity and \< 10% of positive cells, 2+ mild immunoreactivity and 10--30% of positive cells, 3+ moderate immunoreactivity and 31--60% of positive cells, and 4+ strong immunoreactivity and 61--100% of positive cells. Statistical analysis -------------------- All measurements were compared by Student\'s t-test, Welch\'s t-test or Fisher\'s exact probability test for paired samples. Results ======= General observation ------------------- Bloody stool was noted in a few mice received 2% DSS and their body weight gains were slightly decreased during the period of the treatment. However, thereafter no such clinical symptoms were observed. After Week 12, anal prolapsus due to the tumor development in the distal colon was found in a few mice treated with AOM and 2% DSS (group 1). The body weights and lengths of large bowel of mice in all groups at the end of the study are shown in Table [1](#T1){ref-type="table"}. The mean body weights of groups 2 (AOM/DSS/0.04% nimesulide, P\< 0.005), 3 (AOM/DSS/0.04% troglitazone, P\< 0.02), and 4 (AOM/DSS/0.05% bezafibrate, P\< 0.05), were significantly higher than that of group 1 (AOM/DSS). The mean body weight of group 9 (DSS alone, P\< 0.05) was significantly lower than that of group 10 (untreated). The mean liver weights of mice in groups 3 (P\< 0.05) and 4 (P\< 0.01) were significantly greater than that of group 1. However, there were no pathological alterations suggesting toxicity of test compounds in the liver, kidneys, lung, and heart of mice (data not shown). Pathological findings --------------------- Macroscopically, nodular, polypoid or caterpillar-like tumors were observed in the middle and distal colon of mice in groups 1 through 4. They were histologically tubular adenoma (Figure [2a](#F2){ref-type="fig"}) or well-/moderately-differentiated tubular adenocarcinoma (Figure [2b](#F2){ref-type="fig"}). Dysplasia (Figure [2c--e](#F2){ref-type="fig"}) was also developed in mice of these groups. Animals of groups 5--10 did not have large bowel neoplasms and dysplasia. The incidences and multiplicity of colon neoplasma are shown in Table [2](#T2){ref-type="table"}. Group 1 (AOM/DSS) induced 100% incidence of colon adenocarcinomas with a multiplicity of 3.0 ± 1.8. The incidences of colorectal adenocarcinomas in groups 2 (AOM/DSS/0.04% nimesulide), 3 (AOM/DSS/0.05% troglitazone), and 4 (AOM/DSS/0.05% bezafibrate) were significantly smaller than that of group 1 (P\< 0.01, P\< 0.01 and P\< 0.05, respectively). The multiplicities of colon adenocarcinomas in groups 2 and 3 were also significantly lower than that of group 1 (P\< 0.005 and P\< 0.05, respectively). While the multiplicity of colon adenocarcinoma of group 4 (AOM/DSS/0.05% bezafibrate) was smaller than group 1, the difference was insignificant. In this study, mucosal ulcer with or without focal dysplasia (Figure [2c--e](#F2){ref-type="fig"}) were also found in the distal colon of mice in groups 1 through 4. The incidences and multiplicity of colonic ulceration and dysplasia are shown in Table [3](#T3){ref-type="table"}. The incidences and multiplicities of colorectal mucosal ulcer and dysplasia of groups 2, 3, and 4 were smaller than group 1, but the differences did not reach to statistical significance. Immunohistochemistry for PCNA, ssDNA, β-catenin, COX-2, iNOS and nitrotyrosine in colonic adenocarcinoma -------------------------------------------------------------------------------------------------------- As summarized in Table [4](#T4){ref-type="table"}, PCNA-labeling index (Figure [3a](#F3){ref-type="fig"}) of colonic adenocarcinomas developed in groups 2, 3, and 4 was significantly smaller than group 1 (P\< 0.001). Apoptotic index measured by ssDNA immunohistochemistry (Figure [3b](#F3){ref-type="fig"}) in groups 2, 3, and 4 was significantly greater than group 1 (P\< 0.001). Strong β-catenin expression was seen in the nucleus and cytoplasm of adenocarcinoma cells (Figure [3c](#F3){ref-type="fig"}). Although the intensity was relatively weaker than carcinoma cells, adenoma cells showed positivity for β-catenin in their cytoplasm and cell membrane. β-catenin immunoreactivity was also found in the cell membrane and cytoplasm of dysplastic cells. Non-lesional cryptal cells showed weak positivity of β-catenin in their cell membrane. In the positive cases of COX-2 (Figure [3d](#F3){ref-type="fig"}), and iNOS (Figure [3e](#F3){ref-type="fig"}) expression in the dysplasia and adenocarcinoma, the staining pattern was granular and localized to cytoplasm and/or nuclei. Slight immunoreactivity for COX-2 and iNOS was observed in the superficial layers of the non-lesional colonic mucosa and in parts of basal layer in all groups. The expression pattern between COX-2 and iNOS of colorectal adenocarcinomas was well correlated. Furthermore, a positive staining for nitrotyrosine, a marker of nitrosative injury, was mainly observed in mononuclear cells infiltrated in the colonic mucosa (Figure [3f](#F3){ref-type="fig"}). Neoplastic cells also showed negative or very weakly positive immnoreactivity of nitrotyrosine. Scores for β-catenin, COX-2 and iNOS expression in colonic adenocarcinomas are given in Table [4](#T4){ref-type="table"}. β-Catenin expression scores of colorectal adenocarcinomas in groups 2 and 3 were significantly decreased when compared with that in group 1 (P\< 0.05 and P\< 0.05, respectively). Scores for COX-2 and iNOS expression of colorectal adenocarcinomas in groups 2, 3, and 4, were significantly smaller than those in group 1 (P\< 0.001). The scores of nitrotyrosine positivity in groups 2, 3, and 4 were significantly lower than group 1 (P\< 0.001). Discussion ========== The results of the present work clearly indicated that a COX-2 inhibitor nimesulide and a PPARγ ligand troglitazone effectively inhibited AOM/DSS-induced colitis-related colonic carcinogenesis in mice. Inhibitory effect of nimesulide was superior to that of troglitazone. Bezafibrate also reduced the occurrence of colonic adenocarcinoma, but the ability was relatively weaker than that of nimesulide and troglitazone. The suppressive effects of nimesulide, troglitazone and bezafibrate on the development of colonic adenocarcinoma was well correlated with the inhibition of cell proliferation activity, induction of apoptosis, and lowered immunoreactivity of β-catenin, COX-2, iNOS, and nitrotyrosine in the colonic malignancies. However, no differences on the frequency of dysplastic lesions could be observed among the groups. These data may suggest that the pharmacological classes tested under the present investigation slow down the time course of tumor development rather than completely preventing it. The pathogenesis of IBD-associated colorectal carcinogenesis is widely believed to involve a step-wise progression from inflamed and hyperplastic epithelia through flat dysplasia to finally adenocarcinoma \[[@B30]\]. IBD-associated colorectal carcinogenesis is probably promoted by chronic inflammation, but the mechanism is still unclear. However, mucosal inflammation may result in colonic carcinogenesis through several proposed mechanisms such as induction of genetic mutations, increased-cryptal cell proliferation, changes in crypt cell metabolism, changes in bile acid enterohepatic circulation, and alterations in bacteria flora \[[@B4],[@B34]\]. These events are considered to promote IBD-associated CRC development. Given the correlation between increased COX-2 expression and colonic carcinoma and/or inflammation, the chemopreventive effects of NSAIDs seem to be mediated, at least in part, by COX inhibition \[[@B35]\]. We \[[@B36]\] and others \[[@B29],[@B37]\] demonstrated that NSAIDs including nimesulide inhibited both colon tumorigenesis and colitis. In the current study, powerful chemopreventive ability of nimesulide was observed in our colitis-related mouse colon carcinogenesis model, suggesting that nimesulide can be applied as an effective chemopreventor of both sporadic and IBD-associated cololectral carcinogenesis. We previously demonstrated that dietary administration of PPARα and PPARγ ligands inhibits AOM and/or DSS-induced ACF in rodents \[[@B38]\]. In the present study, cancer chemopreventive ability of the PPARγ ligand, troglitazone, or the PPARα ligand, bezafibrate, was found in AOM/DSS-induced mouse colon carcinogenesis model, although their ability was lower than nimesulide. Inhibition of colonic inflammation and decrease in cell proliferation activity by these PPARs ligands might be responsible for their chemopreventive effects on colitis-associated colon carcinogenesis \[[@B38]\]. DNA damage caused by reactive oxygen and nitrogen species may contribute to colitis-related colon tumorigenesis \[[@B39]\]. Several NSAIDs can bind to PPARα and PPARγ \[[@B40]\]. Their anti-inflammatory activities might be mediated through inhibition of COX-1 and/or COX-2. PPARα could suppress COX-2 induction \[[@B41]\]. In addition, immunomodulation by the PPARs ligands might contribute to inhibition of colitis and colon carcinogenesis \[[@B42]\]. Expression and activity of iNOS is increased in colonic mucosa in patients with IBD \[[@B43]\] and colonic adenomas \[[@B44]\]. Several studies using experimental colon carcinogenesis models indicate that chemically induced colon tumors have higher expression and/or activity of iNOS compared with those in their adjacent non-tumorous tissues \[[@B12],[@B25]\]. Numerous iNOS-positive and nitrotyrosine-positive inflammatory cells are observed in non-cancerous colonic mucosa of mice treated with DSS \[[@B25]\]. PPARα \[[@B45]\] and PPARγ \[[@B46]\] involve in inflammation control, and can inhibit iNOS expression \[[@B47]\]. In addition, Rao et al. \[[@B48]\] showed that an iNOS-selective inhibitor suppresses AOM-induced colonic ACF development and iNOS activity. Furthermore, nitrotyrosine may originate from the reaction of iNOS generated NO with reactive oxygen species \[[@B49]\] or the myeloperoxidase-dependent pathway \[[@B50]\]. In the present study, we found a positive immunoreactivity for iNOS and nitrotyrosine in the inflamed colon, suggesting the formation of peroxynitrite and other NO-derived oxidants. These results may suggest that one of the mechanisms by which tested agents exert chemopreventive ability might be related to suppression of iNOS activity and/or expression. Cell proliferation plays an important role in multi-step carcinogenesis \[[@B51]\]. In the colon, the number of cryptal cells is strictly regulated by a balance between cell proliferation and cell death that maintains homeostasis \[[@B52]\]. Changes in cell proliferation and apoptosis are regarded as a common denominator in the pathogenesis of tumor formation \[[@B53]\]. Reduced tumor incidence is generally associated with decreases in cellular proliferation and/or increases in apoptosis \[[@B54]\]. An increased COX-2 expression in CRC \[[@B55],[@B56]\] may confer a survival advantage on cells by inhibition of apoptosis and a change in cellular adhesion to the extracellular matrix \[[@B57]\]. Cancer cells treated with PPARγ ligands induce cell differentiation and apoptosis \[[@B14],[@B17]\]. Recently, Tardieu et al. \[[@B58]\] demonstrated that nimesulide increases apoptosis in colonic mucosa of DSS-treated rats. Our findings that nimesulide and troglitazone inhibited cell proliferation activity and induced apoptosis in colorectal mucosa are in accordance with these findings. Thus, cellular responses like cell growth and/or apoptosis to nimesulide and troglitazone may contribute to chemopreventive effects against colon carcinogenesis processes. β-Catenin is a key regulator of the cadherin-mediated cell-cell adhesion system and an important element in the Wnt signal transduction pathway \[[@B59]\]. Accumulated β-catenin interacts with T-cell factor (Tcf) or lymphoid-enhancer factor (Lef) and translocates to the nucleus, in which it transactivates target genes including c-mycand cyclin D1that are the potentially oncogenic \[[@B47],[@B60]\] in the cytoplasm or nucleus as a consequence of mutant Apcor β-cateningenesis frequently observed in early stages of colorectal carcinogenesis \[[@B61],[@B62]\]. Recently, Girnun et al. \[[@B21]\] indicated that a ligand of PPARγ suppresses β-catenin levels and colon carcinogenesis in Pparγ^+/-^mice treated with AOM. Furthermore, COX-2 is regulated by nuclear β-catenin accumulation \[[@B63]\]. In the current study, treatment with nimesulide, troglitazone, and bezafibrate significantly suppressed β-catenin expression in colorectal adenocarcinomas. Thus, it seems likely that the preventive efficacies of the COX-2 inhibitor (nimesulide) or the PPAR ligands (troglitazone and bezafibrate) against AOM/DSS induced mouse colon carcinogenesis might be mediated, at least in part, by β-catenin down-regulation. Chemoprevention of cancer might be defined as the deliberate introduction of these selected non-toxic substances into the diet for the purpose of reducing cancer development. In the present study, the mean liver weights of mice in dietary feeding of troglitazone and bezafibrate were significantly increased. However, there were no pathological alterations suggesting toxicity of the drugs examined. It is known that administration of PPARs ligands to rodent exhibit hepatomegaly due to both cellular hypertrophy and hyperplasia \[[@B64]\]. Increase in liver weight by exposure of troglitazone and bezafibrate in this study may be caused by these pathological changes. The estimated daily intakes of nimesulide, troglitazone, and bezafibrate in mice given the diets containing 400 ppm and 500 ppm were approximately 160 mg/kg and 200 mg/kg in the present study. In a direct extrapolation to a 60 kg person, these doses are slightly lower than those of clinical trial \[[@B65]-[@B67]\]. These findings may suggest that the efficacy of these agents at dietary dose-levels has a direct practical and translational relevance to human. Conclusion ========== In conclusion, dietary administration of nimesulide, troglitazone, and bezafibrate could effectively suppress colon carcinogenesis induced by AOM and DSS in female ICR mice. Our on-going study on molecular profiles in colonic samples from the current experiment will provide precise molecular mechanisms involved in their inhibitory action in AOM/DSS-induced mouse colon carcinogenesis. Abbreviations ============= AOM, azoxymethane; ACF, aberrant crypt foci; CRC, Colorectal cancer; COX-2, cyclooxygenase-2; DSS, dextran sodium sulfate; FAP, familial adenomatous polyposis; H&E, hematoxylin and eosin; IBD, inflammatory bowel disease; iNOS, inducible nitric oxide synthase; NSAID, non-steroidal anti-inflammatory drug; PPAR, peroxisome proliferator-activated receptor; PCNA, proliferating cell nuclear antigen; ssDNA, single stranded DNA; UC, ulcerative colitis. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= HK participated in study design, performed the animal studies, and drafted the manuscript. RS carried out tissue collection and data analysis. SS participated in the histopathological and immunohistological analysis. TT participated in study design, coordination, and manuscript preparation. 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/1471-2407/5/46/prepub> Acknowledgements ================ Dr. Rikako Suzuki is a Research Fellow of the Japan Society for the Promotion of Science. This research was supported in part by the Grant-in-Aid (13--15) for Cancer Research from the Ministry of Health, Labour and Welfare of Japan; the Grant-in-Aid for the 3rd Term Comprehensive Control Research from the Ministry of Health, Labour and Welfare of Japan; the Grants-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan; and the grant (C2004-4) for the Collaborative Research from Kanazawa Medical University; and the grant (H2004-6) for the Project Research from the High-Technology Center of Kanazawa Medical University. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Experimental protocol. Arrows, AOM 10 mg/kg body weight, i.p. injection; densely cross-hatched bars, 2% dextran sodium sulfate (DSS) in drinking water; open bars, basal diet and tap water; crosses, death. ::: ![](1471-2407-5-46-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Histopathology of colonic lesions in mice of group 1. (a) adenoma, (b) adenocarcinoma, (c) mild dysplasia, and (d) severe dysplasia. Hematoxylin and eosin stain. Original magnifications, × 10. ::: ![](1471-2407-5-46-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Immunohistochemistry of PCNA, ssDNA, β-catenin COX-2, iNOS, and nitrotyrosine in mice of group 1. (a) PCNA immunohistochemistry, (b) ssDNA immunohistochemistry, (c) β-catenin immunohistochemistry, (d) COX-2 immunohistochemistry, (e) iNOS immunohistochemistry, and (f) nitrotyrosine immunohistochemistry. Original magnifications: a, d, e, f, × 10; c, × 20; b, × 40. Insets, immunohistochemistry staining for each antibody in the mouse colon from group 10 (Original magnifications, × 10). ::: ![](1471-2407-5-46-3) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Body, liver, relative liver weights, and length of colon. ::: Group no. Treatment No. of mice Body weight (g) Liver weight (g) Relative liver weight (g/100 g body weight) Length of colon (cm) ----------- ---------------------------- ------------- ----------------- ------------------ --------------------------------------------- ---------------------- 1 AOM/DSS 10 36.6 ± 5.6 ^a^ 2.0 ± 0.5 5.5 ± 0.7 11.5 ± 2.1 2 AOM/DSS/0.04% Nimesulide 10 44.8 ± 5.6 ^b^ 2.3 ± 0.3 5.3 ± 0.7 12.1 ± 1.3 3 AOM/DSS/0.05% Troglitazone 10 43.6 ± 6.3 ^c^ 2.5 ± 0.5 ^d^ 5.7 ± 1.0 11.4 ± 1.1 4 AOM/DSS/0.05% Bezafibrate 10 42.4 ± 6.4 ^d^ 2.8 ± 0.7^e^ 6.6 ± 1.5 12.1 ± 1.4 5 0\. 04% Nimesulide 5 39.9 ± 7.3 2.0 ± 0.3 5.0 ± 0.9 12.5 ± 0.7 6 0\. 05% Troglitazone 5 39.8 ± 4.0 2.1 ± 0.3 5.2 ± 0.7 11.9 ± 0.9 7 0\. 05% Bezafibrate 5 46.8 ± 9.6 2.7 ± 0.5 5.8 ± 0.9 12.6 ± 1.4 8 AOM 5 42.8 ± 2.6 1.9 ± 0.1 4.5 ± 0.4 12.3 ± 1.2 9 DSS 6 34.4 ± 2.8^f^ 1.8 ± 0.2 5.2 ± 0.7 12.3 ± 0.6 10 None 5 43.2 ± 6.5 2.0 ± 0.2 4.8 ± 0.8 12.0 ± 0.6 ^a^Mean ± SD. ^b-e^Significantly different from group 1 by Student\'s t-test (^b^P\< 0.005, ^c^P\< 0.02, ^d^P\< 0.05, and ^e^P\< 0.01). ^f^Significantly different from group 1 by Welch\'s t-test (^f^P\< 0.05). ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Incidence and multiplicity of colonic neoplasia. ::: Group no. Treatment No. of mice Incidence (no. of mice with neoplasms) Multiplicity (no. of tumors/mice, means ± SD) ----------- ---------------------------- ------------- ---------------------------------------- ----------------------------------------------- --------------- -------------- -------------- -------------- 1 AOM/DSS 10 10/10 (100%) 10/10 (100%) 10/10 (100%) 5.2 ± 3.0 2.1 ± 1.8 3.0 ± 1.8 2 AOM/DSS/0.04% Nimesulide 10 8/10 (80%) 6/10 (60%)^a^ 4/10 (40%)^b^ 1.8 ± 1.7^b^ 1.2 ± 1.3 0.6 ± 1.0^c^ 3 AOM/DSS/0.05% Troglitazone 10 9/10 (90%) 9/10 (90%) 4/10 (40%)^b^ 2.5 ± 1.8^a^ 1.6 ± 1.1 1.2 ± 2.5^a^ 4 AOM/DSS/0.05% Bezafibrate 10 8/10 (80%) 7/10 (70%) 6/10 (60%)^a^ 2.6 ± 2.5^a^ 1.1 ± 1.0^a^ 1.8 ± 2.6 5 0.04% Nimesulide 5 0/5 (0%) 0/5 (0%) 0/5 (0%) 0 0 0 6 0.05% Troglitazone 5 0/5 (0%) 0/5 (0%) 0/5 (0%) 0 0 0 7 0.05% Bezafibrate 5 0/5 (0%) 0/5 (0%) 0/5 (0%) 0 0 0 8 AOM 5 0/5 (0%) 0/5 (0%) 0/5 (0%) 0 0 0 9 DSS 6 0/6 (0%) 0/6 (0%) 0/6 (0%) 0 0 0 10 None 5 0/5 (0%) 0/5 (0%) 0/5 (0%) 0 0 0 ^a,b,c^Significantly different from group 1 by Fisher\'s exact probability test or Student\'s t-test (^a^P\< 0.05, ^b^P\< 0.01, and ^c^P\< 0.005). ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Incidence of multiplicity colonic mucosal ulcer and dysplasia. ::: Group no. Treatment Incidence (%) Multiplicity (no. of lesions / mouse, means ± SD) ----------- ---------------------------- --------------- --------------------------------------------------- ----- ----- ----------- ----------- ----------- ----------- 1 AOM/DSS 40% 90% 80% 50% 0.5 ± 0.7 3.2 ± 1.5 1.4 ± 1.0 1.1 ± 1.3 2 AOM/DSS/0.04% Nimesulide 10% 90% 80% 50% 0.1 ± 0.3 2.2 ± 2.3 1.2 ± 0.9 0.6 ± 0.7 3 AOM/DSS/0.05% Troglitazone 20% 90% 50% 30% 0.3 ± 0.7 2.1 ± 2.2 0.7 ± 0.8 0.8 ± 1.6 4 AOM/DSS/0.05% Bezafibrate 30% 80% 60% 20% 0.4 ± 0.7 1.9 ± 1.8 0.9 ± 1.0 0.4 ± 0.8 ::: ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### PCNA and apoptosis indices and scores of β-catenin, COX-2, iNOS and nitrotyrosine expression in colonic adenocarcinomas. ::: Group Treatment PCNA-labeling Apoptotic Scores for: ------- ---------------------------- --------------------- ------------------- ------------------ ------------------- ------------------- ------------------- 1 AOM/DSS 62.4 ± 13.7^a^(30) 4.1 ± 1.9 (30) 3.7 ± 0.7 (30) 3.3 ± 0.7 (30) 3.0 ± 1.0 (30) 3.1 ± 0.8 (30) 2 AOM/DSS/0.04% Nimesulide 38.3 ± 11.1 ^b^(6) 11.8 ± 2.9 ^b^(6) 2.3 ± 1.0 ^c^(6) 1.3 ± 0.5 ^b^(6) 1.3 ± 0.5 ^b^(6) 1.7 ± 0.5 ^b^(6) 3 AOM/DSS/0.05% Troglitazone 43.6 ± 9.0 ^b^(9) 10.0 ± 2.4 ^b^(9) 2.7 ± 0.7 ^c^(9) 1.8 ± 0.8 ^b^(9) 1.6 ± 0.5 ^b^(9) 1.8 ± 0.6 ^b^(9) 4 AOM/DSS/0.05% Bezafibrate 40.5 ± 12.7 ^b^(15) 9.7 ± 2.7 ^b^(15) 3.0 ± 0.8 (15) 1.8 ± 0.8 ^b^(15) 1.6 ± 0.6 ^b^(15) 2.2 ± 1.1 ^b^(15) Numbers in parentheses are numbers of lesions examined. ^a^: means ± SD. ^b,c^Significantly different from group 1 by Student\'s t-test (^b^P\< 0.001 and ^c^P\< 0.05). :::	PubMed Central	2024-06-05T03:55:57.620255	2005-5-16	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1156872/", "journal": "BMC Cancer. 2005 May 16; 5:46", "authors": [ { "first": "Hiroyuki", "last": "Kohno" }, { "first": "Rikako", "last": "Suzuki" }, { "first": "Shigeyuki", "last": "Sugie" }, { "first": "Takuji", "last": "Tanaka" } ] }
PMC1156873	Background ========== Maspin is a member of serpin family with unique tumor suppressing activity. Initially identified from normal mammary epithelial cells, maspin gene is not mutated nor deleted but is rather transcriptionally down-regulated or silenced by epigenetic changes in breast cancer \[[@B1]-[@B5]\]. Maspin protein made from E.coli, yeast, and insect inhibits breast tumor cell migration and invasion \[[@B6]\]. In animal models, maspin has been found to inhibit angiogenesis in rat cornea and xenograft models \[[@B7]\], and to inhibit mammary tumor progression and metastasis in bitransgenic mice \[[@B8]\]. In human breast tissue, maspin is expressed in both luminal and myoepithelial cells, and it has been suggested that the maspin-expressing myoepithelial cells form a defensive barrier for the progression from ductal carcinoma in situ to more invasive carcinoma \[[@B9]\]. Despite these findings, the molecular mechanism underlying maspin\'s inhibition is not well characterized. Previously, our laboratory showed that overexpression of maspin in normal mammary gland inhibited alveolar development during pregnancy through the induction of mammary cell apoptosis \[[@B10]\]. We also demonstrated that in mammary tumors isolated from WAP-SV40 TAg and WAP-maspin bitransgenic mice there was a strong correlation between maspin overexpression and increased apoptosis \[[@B8]\], suggesting that maspin might induce tumor cell apoptosis in vivo. Jiang et al also showed that maspin could sensitize MDA-435 mammary tumor cells to apoptosis induction with a chemical reagent \[[@B11]\]. Recently, we have established a new mammary tumor implantation model to further elucidate the mechanism of maspin-mediated tumor suppression \[[@B12]\]. We demonstrated that retrovirus infection of TM40D mammary tumor cells with maspin significantly blocked tumor growth and metastasis. Using the TM40D mammary tumor cells, we showed that maspin induced tumor cell apoptosis through translocation to mitochondria \[[@B13]\]. Here, we show that maspin-overexpressing tumor cells display a high rate of apoptosis when cells are treated with a chemical reagent staurosporine or under serum starvation. We have shown that the death signal does not involve the secreted maspin on cell surface but rather is mediated through the intracellular function of maspin. An intrinsic death signal pathway is induced which alters the protein level of Bcl-2 family members in maspin-expressing tumor cells. Methods ======= Cell line and cell culture -------------------------- Murine mammary tumor cell line TM40D cells were infected with retrovirus vector for establishing stable cell lines as described by Shi et al. \[[@B12]\]. Briefly, human maspin cDNA was cloned into pS2-GFP, a retroviral vector that was derived from the pS2 family of retroviral vectors. The plasmid constructs, pS2-maspin and pS2-blank vector were transfected into 293T package cells to produce infective viral particles. The viral supernatants were then allowed to infect TM40D cells in the presence of Polybrene. The transfected cells were then selected in the presence of 100 μg/ml of zeocin (Invitrogen Co., CA). Cells were seeded by limiting dilution in 96-well plates. Single clones of stably transfected cells were transferred to individual wells of 24-well plates and cultured in medium containing 100 μg/ml Zeocin. Individual clones were confirmed for the presence of human maspin cDNA by RT-PCR, immunobloting with maspin polyclonal antibody and immunofluorescence staining. Two maspin expression clones were named as TM40D-Mp (16) and TM40D-Mp (18), respectively. One TM40D cell line infected by pS2-vector was used as a negative control. All tumor subclones were maintained at 37°C in a humidified 95% O2-5% CO2 atmosphere in DMEM supplemented with 10% FBS and L-glutamine. Maspin immunostaining --------------------- For intracellular maspin immunostaining, cells grown on chamber slides were fixed in 4% paraformaldhyde solution for 1 hr and were permeabilized with 0.5% NP-40 in PBS for 30 min. The slides were blocked with 10% normal horse serum for 1 hour before they were treated with the primary antibody Abs4A at a dilution of 1:200. The secondary antibody (Texas red conjugated goat anti-rabbit antibody, Santa Cruz, CA) was used at a dilution of 1:1000 at room temperature for 1 hour. For cell surface staining, cells grown on slides were washed with PBS without Mg++ and Ca++ and blocked with 2% BSA for 1 hr at 4°C. First antibody was used at a dilution of 1:200. The secondary antibody (Texas red conjugated goat anti-rabbit antibody, Santa Cruz, CA) was used at a dilution of 1:500 at 4°C for 30 min. Slides were mounted and viewed under a Leica fluorescence microscope. Apoptosis induction ------------------- TM40D cells were plated on coverslips or 10 cm plates and cultured to 80% confluence. They were treated by one of the following procedures: 1) Serum starvation: TM40D cell lines were cultured under serum free condition for 24--72 hours in D-MEM medium with 0.1% BSA; 2) TNF-alpha treatment (modified according to Kulik\'s method) \[[@B14]\]. Briefly, the cells were incubated in serum-free DMEM medium for 12 hours, and then were treated with 100 ng/ml TNF-alpha (Sigma) for another 7 hours; 3) Staurosporine treatment: 1 μM staurosporine was added to the D-MEM medium with 2.5% FBS for another 4 hours. In conjunction with the above treatments, either the caspase-8 inhibitor II (Z-IETD-FMK) or caspase-9 inhibitor II (LEHD-CHO) were added into the media at a concentration of 20 nM. TUNEL assay ----------- After the various treatments, the cells cultured on coverslips were washed with ice-cold PBS and fixed in 4% paraformaldehyde for 1 hour at 4°C. The TUNEL (Terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling) assay was performed according to the manufacturer\'s specifications (Roche). Briefly, cell samples were fixed with freshly prepared 4% paraformaldehyde solution for 1 hour. Cells in the coverslips were rinsed with PBS twice and permeabilized with 0.1% Triton X-100 in 0.1% sodium citrate for 2 min on ice. Following the triton treatment, 50 μl TUNEL reaction mixture was added to each sample for 30 min at 37C. The slides were then rinsed 3 times with PBS and counterstained with DAPI. The mounted slides were analyzed using fluorescence microscopy. The quantitation of apoptosis was done by counting the number of apoptotic positive cells in four randomly selected fields with a 20 X objective. For antibody-blocking assay, TM40D-Mp (18) cells were cultured at 80% confluence. Cell apoptosis was induced using staurosporine (1 μM in DMEM) and the anti-maspin antibody (AbS4A) was added to the medium at a concentration of 1:200 or 1:400. Normal fetal bovine serum FBS (1:25) and rabbit IgG (1:200) were used as negative controls. After 7-hour incubation, cells were fixed and used for TUNEL assay as described above. Cytosolic protein extraction and cytochrome c immunoblot -------------------------------------------------------- Cytochrome c immunoblot was carried out using a modified protocol of Kluck et al \[[@B15]\]. Cultured cells were starved in serum-free DMEM medium for 48 hours and then collected with lysis buffer (1 mM CaCl2, 1 mM MgCl2, 1% NP-40, 1 μg/ml leupeptin, 1 μg/ml aprotinin, 1 μM PMSF, and 100 μM NaVO4). After the lysis mixtures were incubated on ice for 20 mins, they were centrifuged at 14,000 rpm for 15 min. The supernatants were collected and total cytosolic proteins were quantitated by a Bradford spectrometer. About 80 μg protein was loaded in a 10% SDS-PAGE gel and transferred to polyvinylidene difluoride (PVDF) membrane (Bio-rad Laboratories, Richmond, CA) at a constant current of 250 mA for 2 hours. The membrane was incubated with the primary cytochrome c polyclonal antibody (H-104, Santa Cruz) overnight at a dilution of 1:500. Horseradish peroxidase-conjugated anti-rabbit secondary antibody at a 1:1000 dilution was incubated with the membrane for another 1 hour at room temperature. The protein was detected with enhanced chemiluminescence (Pierce, Rockford, IL) for autoradiography (Hyperfilm-Pharmacia). Immunoblot and immunoprecipitation assay for Bcl-2 family proteins ------------------------------------------------------------------ To determine the changes in the expression of Bcl-2 family proteins in TM40D cells, cells induced under different procedures were collected and homogenized in lysis buffer and were sonicated for 4 times for 5 seconds each on ice. The lysis mixture was centrifuged at 14,000 rpm for 15 min. Sample protein concentration was measured using the method of Bradford (Bio-rad). Bax protein level was determined using conventional Western blot analysis. For the treatment of cycloheximide (CHX), TM40D-cont and TM40D-Mp (18) cells were first induced with STS for 1 hour before the addition of CHX (50 μM) at the time of 0, 2, 3, and 4 hrs. Vehicle treated tumor cells for 4 hr were used as control. Bcl-2 and Bcl-X~L~protein were determined by immunoprecipitation assay. Briefly, 1,000 μg of protein lysis was incubated with 1 μg of anti-Bcl-2 or anti-Bcl-X~L~antibodies overnight at 4°C, respectively. Twenty microliters of protein A agarose beads (Santa Cruz) were then added to the mixture for an additional 4 hours at 4°C. Immunoprecipitates were harvested by centrifugation at 2,500 rpm for 5 min at 4°C. The beads were washed 5 times in lysis buffer. Pellets were resuspended in 20 μl of 3× SDS samples buffer and heated for gel electrophoresis. For Western blot analysis of Bcl-2 family proteins, samples were run in 15% SDS polyacrylamide gels. The proteins were transferred onto a PVDF membrane (Bio-rad) and blotted with anti-Bax (1:1000), anti-Bcl-2 (1:1000) and anti-Bcl-X~L~(1:500). Horseradish peroxidase-conjugated anti-rabbit secondary antibody at a 1:1000 dilution was incubated with the membrane for another 1 hour at room temperature. The protein was detected by enhanced chemiluminescence (Pierce, Rockford, IL) and autoradiography (Hyperfilm-Pharmacia). Ribonuclease protection assay ----------------------------- For ribonuclease protection assays (RPA), total RNA was isolated from TM40D-cont and TM40D-Mp cells treated by staurosporine (1 μM) for 4 hrs. ^35^S labeled Bcl-2 family antisense probes were generated by transcription of Bcl-2 family genes using a RPA kit (BD PharMingen, Inc., CA) in accordance with the manufacturer\'s instruction. Protected bands were exposed to an X-ray film and quantified by phosphoimaging analysis. Relative RNA levels were obtained by dividing the quantitated volume of the protected Bcl-2 family gene bands by the volume of either GPADH or L32 control. Statistical analysis -------------------- The differences of means between groups were assessed by the Student\'s t-test. P \< 0.05 was considered significant (two tail analysis). All statistical analysis was performed using Microsoft office XP Software. Results ======= To examine whether maspin is involved in the induction of tumor cell apoptosis, stable clones with various levels of maspin expression were selected and analyzed using semi-quantitative RT-PCR, immunostaining, and Western blot analyses. As shown in Fig. [1](#F1){ref-type="fig"}, two maspin expression clones TM40D-Mp (16) (T16) and TM40D-Mp (18) (T18), and a negative control subclone TM40D-cont (TC) were selected for further study. We examined the localization of maspin in maspin expressing and control tumor cells using immunofluorescence staining with an anti-maspin antibody. T16 and T18 had detectable immunostaining signals (B, C, E, and F) while TC did not express maspin (A, D). Under high magnification, we observed that maspin immunostaining was abundantly located in the cytoplasm and on the cell surface (Fig. [1G](#F1){ref-type="fig"}). T18 cells had a higher level of maspin expression than T16 cells at both mRNA and protein levels (Fig. [1H](#F1){ref-type="fig"}). TC, T16, and T18 cells were treated with chemical reagent staurosporine (STS) or TNFα. As shown Fig. [2A](#F2){ref-type="fig"}, TC, T16, and T18 cells underwent different levels of apoptosis when they were treated with STS. Caspase 8 and caspase 9 inhibitors were added to TC and T18 cells treated with STS and TNFα \[[@B16],[@B17]\]. In both TC and T18 cells, the rate of apoptosis in STS-treated cells was inhibited when cells were treated with caspase 9 and 8 inhibitors, more so in T18 cells than that in TC cells (Fig. [2B](#F2){ref-type="fig"}). In contrast, when TC and T18 cells were induced to apoptosis with TNF-α, only caspase 8 inhibitor inhibited tumor cell apoptosis (Fig. [2C](#F2){ref-type="fig"}). However, there was no difference in the rate of apoptosis between T18 and control TC cells when they were induced with TNF-α (Fig. [2C](#F2){ref-type="fig"}). Serpins can regulate cellular apoptosis through both extrinsic and intrinsic pathways \[[@B18],[@B19]\]. If maspin is involved in an extrinsic pathway, it will require that the death signal be initiated by extracellular maspin protein. Since maspin is present in both cytoplasm and on the cell surface in TM40D-Mp cells (Fig. [1D](#F1){ref-type="fig"}), it is possible that maspin could be secreted and bound to the surface of tumor cells. One major function of maspin is the inhibition of tumor cell migration which is mediated through a cell surface event \[[@B20]\]. One maspin antibody Abs4A acts as function-blocking reagent in the migration assay \[[@B20]\]. To further determine whether maspin-mediated apoptosis is through secreted maspin protein, an anti-maspin AbS4A polyclonal antibody was incubated with T18 cells that were stimulated for apoptosis with staurosporine. As shown in Fig [3](#F3){ref-type="fig"}, AbS4A maspin antibody did not block STS-induced cell apoptosis at concentrations up to 1 μg/ml (1:200 dilution), even though the same antibody was effective at blocking maspin\'s effects on cell migration at a much lower dosage (data not shown). This suggests that secreted maspin from TM40D-Mp cells does not induce cell apoptosis through secreted maspin acting on the cell surface. The Bcl-2 family proteins consist of pro- and anti- apoptotic proteins that act through mitochondria to regulate apoptosis through the caspase 9 and cytochrome c pathway. Since we know that maspin-mediated apoptosis involves mitochondrial death pathway \[[@B13]\], we examined the protein level of three key apoptosis proteins in tumor cells treated under serum starvation. These proteins include Bax, a pro-apoptotic protein \[[@B21]\], and the anti-apoptotic Bcl-2 and Bcl-X~L~\[[@B21]-[@B23]\]. When these cells were serum starved to induce apoptosis, the levels of anti-apoptotic proteins such as Bcl-2 and Bcl-XL were too low to be detected by conventional Western blot analysis. Thus, we used the method of immunoprecipitation followed by Western blot analysis to analyze their expression levels. No difference was detected in Bcl-XL protein between the maspin expressing T16 and T18 cells, and the control TC cells (Fig. [4A](#F4){ref-type="fig"}). However, a decrease was consistently observed in Bcl-2 level in T16 and T18 cells compared to TC cells under serum starvation condition (Fig. [4A](#F4){ref-type="fig"}). The pro-apoptotic protein Bax could be easily detected when tumor cells were placed under serum starvation condition. In contrast to Bcl-2 protein, a dramatic increase in Bax protein level was observed in T16 and T18 cells compared to TC cells (Fig. [4B](#F4){ref-type="fig"}). Such increased level of Bax was not observed when cells were treated with TNF-alpha, which mediates cell apoptosis through death receptor rather than the mitochondrial pathway (data not shown). To determine whether maspin overexpression regulates the gene expression of Bcl-2 family during apoptosis, a quantitative ribonuclease protection assay (RPA) was carried out using radiolabeled antisense probes for Bcl-2 family genes (RPA kit, BD PharMingen, Inc., CA), and RNAs isolated from tumor cells under apoptotic condition. As shown in Fig. [5](#F5){ref-type="fig"}, the levels of Bax, Bcl-2, and several other Bcl-2 family genes were not changed significantly between control TC cells and maspin expressing T16 and T18 cells, suggesting the regulation is not at the transcription level. Bcl-2 proteins could be controlled at the level of protein stability. The decreased level of Bcl-2 in T16 and T18 cells (Fig. [4](#F4){ref-type="fig"}) is in line with a previous report demonstrating Bcl-2 protein was specifically targeted for degradation during apoptosis by an ubiquitin-dependent event \[[@B24]\]. Since T16 and T18 cells have higher level of apoptosis, more Bcl-2 proteins could be targeted for degradation in proteasome complex. The Bax level was higher in T16 and T18 cells compared to TC cells (Fig. [4](#F4){ref-type="fig"}). To determine whether the increased level of Bax protein in T16 and T18 cells is due to increased protein stability, cells were induced to apoptosis with STS along with cycloheximide (CHX), which prevents new protein synthesis. The level of Bax protein in TC cells was as stable as that in T18 cells after CHX treatment for 2, 3, and 4 hrs (Fig. [6A](#F6){ref-type="fig"}). However, Bax level was specifically reduced in TC cells untreated with CHX compared to TC cells that were treated with both STS and CHX for the same time period (4 hrs). Less Bax protein was degraded in vehicle treated T18 cells than that in vehicle treated TC cells. A change in the level of Bcl-2 family proteins in TC and T18 cells may increase the sensitivity of cells to apoptotic signal through their downstream components of mitochondrial death pathway. Cytochrome c is a critical component of the cascade. In normal condition, cytochrome c is located in the intermembrane space of mitochondria in the cell, where it functions as a transducer of electrons in the respiratory chain. When cells undergo apoptosis, cytochrome c is released into the cytoplasm, which initiates the mitochondrial death cascade \[[@B25]\]. To further determine whether maspin regulates the cytochrome c release from the mitochondria, TM40D-cont and TM40D-Mp cells were serum starved and the cells were harvested for both whole cell extracts and cellular fractionations without mitochondria. Both the whole cell and cytosolic fractions were collected for Western blot analysis using an anti-cytochrome c antibody. As shown in Fig. [6B](#F6){ref-type="fig"}, the total amount of cytochrome c in the whole cell extracts of TC and T18 cells remained at similar level. However, maspin expressing T18 cells had an increased release of cytochrome c to the cytosolic fraction compared to the TC cells. This suggests that maspin-mediated apoptosis is through the caspase 9 and cytochrome c death pathway. Discussion ========== We have shown in the past few years that maspin acts as a tumor suppressor, inhibiting both primary tumor growth and metastasis in cell culture and animal models. Here, we provide evidence that maspin is actively involved in the induction of tumor cell apoptosis. There are at least two general apoptosis pathways, the intrinsic and extrinsic pathway. Although maspin protein is present in cytosol and on the cell surface, the induction of apoptosis is not mediated by extracellular maspin. Treatment of maspin expressing T18 cells with STS and an anti-maspin antibody did not affect the apoptosis rate (Fig. [3](#F3){ref-type="fig"}). This data is in line with a previous report by Jiang et al. which showed that exogenous maspin failed to sensitize human breast tumors to chemical induced apoptosis \[[@B11]\]. This implicates that intracellular maspin is actively affecting cell apoptosis, which is supported by our report of translocation of intracellular maspin to the mitochondria \[[@B13]\]. Our data also confirmed that in maspin expressing T16 and T18 cells, the Bcl-2 level was decreased and the level of pro-apoptotic protein Bax was significantly increased compared to control TC cells. Such regulation was selectively controlled since another Bcl-2 family protein Bcl-Xl was not affected (Fig. [4A](#F4){ref-type="fig"}). The effect of maspin on cell apoptotic machinery was not preexisting since both TC and T18 cells had similar, low level of apoptosis in high serum medium. Second, the levels of Bax and caspase 3 were similar before the cells were treated under serum starvation or with STS (data not shown). However, their levels were changed under stress treatment, indicating that the change is dependent on cell death stimulation. The regulation of Bcl-2 genes by maspin was not at the transcriptional level (Fig. [5A](#F5){ref-type="fig"}). Rather, the stability of these proteins was affected by maspin. Others have shown that Bcl-2 protein was dephosphorylated and targeted to ubiquitin-dependent degradation during apoptosis in HUVEC cells \[[@B24]\]. Phosphorylation of Bcl-2 confers resistance against induction of apoptosis. Since Bcl-2 controls the activation of caspase cascade by its interaction with Bax and by participation in apoptosome ensemble, degradation of Bcl-2 may unleash the inhibitory function of Bcl-2 over the apoptosome \[[@B26]\]. The unexpected finding is that the pro-apoptotic Bax was stabilized in the presence of maspin through an unknown mechanism. When new protein synthesis was inhibited by cycloheximide, including those proteins control protein degradation, Bax protein was as stable in TC cells as in maspin expressing T18 cells. Such event of ordered degradation of proteins happens for the regulation of p53 protein during the cell cycle control \[[@B27],[@B28]\]. It is possible that when T16 and T18 cells were induced for apoptosis, more Bax protein might oligomerize and translocate to mitochondria \[[@B29],[@B30]\], which might make the protein less susceptible for degradation in the cytosol. However, in control TC cells treated under apoptotic stress, less Bax proteins might oligomerize and move to mitochondria and thus were more sensitive to degradation in the cytosol. A recent study from my laboratory showed that during maspin-mediated apoptosis, there was also an increased translocation of maspin from the cytoplasm to the mitochondria \[[@B13]\]. Whether there is any connection between Bax and maspin translocation during apoptosis remains to be tested. The finding that maspin-induced apoptosis was partially blocked by both caspase 8 and caspase 9 inhibitor is interesting. Several other reports also show that certain reagents including STS can induce caspase 3 activity and apoptosis through both the activation of caspase 8 and caspase 9 \[[@B31]-[@B35]\]. In addition, there are numerous reports demonstrating the cross-talk between caspase 8 and 9 \[[@B36],[@B37]\]. At the moment, no data suggest that maspin-mediated apoptosis is connected to the death receptor pathway. Rather, our previous report and this study demonstrate that at least the mitochondrial mediated apoptosis pathway is involved \[[@B13]\]. Finally, our data also indicate that the increased apoptosis in maspin expressing cells is due to changes in Bcl-2 family protein level. Due to the changed level of Bcl-2 proteins, the apoptotic pathway became more active in maspin-expressing cells and more cytochrome c proteins were released from mitochondria to the cytosol (Fig. [5B](#F5){ref-type="fig"}). In many cancers, the anti-apoptotic protein Bcl-X~L~and Bcl-2 are found to be over-expressed while the activity of pro-apoptotic Bax is counterbalanced by strong surviving signals \[[@B38]\]. Our finding suggests that one of the functions that maspin plays is to change the level of Bcl-2 family proteins in tumor cells. It is noted that the increased apoptosis mediated by maspin was observed under both serum starvation and chemical drug induction, confirming that the existence of a general mechanism of action through the intracellular, mitochondria death pathway. Such findings offer a new direction for cancer therapy. Identifying the molecular mechanism of maspin-mediated apoptosis will help us to design better reagents for maspin-based therapeutic interventions against breast cancer. Conclusion ========== Overexpression of maspin in mouse mammary tumor cells increased the rate of tumor cell apoptosis when they were treated with STS or serum starvation. Maspin-mediated apoptosis could be partially blocked by caspase 9 and caspase 8 inhibitors. The effect was not through extracellular maspin. Maspin overexpression in breast tumor cells regulated the level of Bcl-2 family proteins. Maspin-expressing T16 and T18 cells had a reduced level of anti-apoptotic protein Bcl-2 but an increased level of pro-apoptotic protein Bax compared to control TC tumor cells. Such regulation occurred at the level of protein stability. The change in Bcl-2 family proteins resulted in an increased release of cytochrome c from mitochondria, thus the increased apoptosis in maspin-expressing tumor cells. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= WZ Performed the experiments. YS Performed the experiments. MZ Conceived and coordinated the study. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2407/5/50/prepub> Acknowledgements ================ This work is supported by a NIH grant CA79736 and a DAMD 17-021-0294 to M.Z. Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Maspin expression in TM40D stable cell lines. Cells were stained with an AbS4A maspin primary antibody and followed by Texas-red conjugated secondary antibody. Immunofluorescence staining was carried out using TC (A, C), T16 (B, E), and T18 cells (C, F). G. High magnification of immunofluorescence staining of T18 (100× oil len) under non-permeabilized condition. Arrows indicate cell surface staining of maspin. Nuclei were stained with DAPI (D, E, F, and G). H. Quantitation of maspin expression in TC (lane 1), T16 (lane 2), and T18 (lane 3) by RT-PCR and Western blot analysis. L19 serves as a loading control for RT-PCR analysis. ::: ![](1471-2407-5-50-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Detection of an increased apoptosis in maspin expressing tumor cells in cell culture under the induction of STS and TNF-alpha. A. TC, T16, and T18 cells in monolayer culture were treated with staurosporine (1 μM) for 4 hrs, and apoptosis rate was detected using the TUNEL assay. Note the increased apoptosis in T16 and T18 cells. B, C, Maspin-mediated apoptosis was blocked by caspase 9 and caspase 8 inhibitors. B. TC and T18 cells were treated with staurosporine (1 μM, 4 hrs), caspase-8 inhibitor II (Z-IETD-FMK) and caspase-9 inhibitor II (LEHD-CHO) were used at a concentration of 20 nM. C. TC and T18 cells were treated with TNF-alpha (100 ng/ml, 7 hrs), caspase-8 inhibitor II (Z-IETD-FMK) and caspase-9 inhibitor II (LEHD-CHO) were used at a concentration of 20 nM. Apoptotic cells were assayed by TUNEL and counted in four randomly selected fields under 20 X objective. Bars were from four repeated experiments. ::: ![](1471-2407-5-50-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Maspin mediated apoptosis is not through a cell surface event. T18 cells in monolayer culture were treated with staurosporine (1 μM). Maspin blocking antibody AbS4A at two concentrations (from a dilution of 1:200 to 1:400) was added to the cell culture. After 7 hours incubation, cells were fixed for TUNEL staining and apoptotic cells were counted in four randomly selected fields under 20 X objective. Results were from three repeated experiments. ::: ![](1471-2407-5-50-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Changed levels of Bcl-2 family proteins in TC, T16, and T18 tumor cells. A. Analysis of anti-apoptotic Bcl-2 proteins in tumor cells. Cells were cultured for 48 hrs in serum starvation condition before they were lysed in cell lysis buffer. Cytosolic proteins (1 mg) were immunoprecipitated with Bcl-2 and Bcl-X~L~monoclonal antibodies overnight at 4°C. The level of Bcl-2 and Bcl-X~L~proteins was detected using a HRP-conjugated anti-mouse secondary antibody. IgG serves as an internal control. Right panel, relative level of Bcl-2 and Bcl-X~L~proteins in TC, T16, and T18 cells. The protein bands were quantitated using a densitometer and were normalized to IgG signals. B. Analysis of pro-apoptotic Bax protein in TM40D cells. Cells were induced to apoptosis with serum starvation. Cell extracts were run on SDS PAGE gel and immunobloted with anti-Bax and anti-beta-actin antibodies. Beta-actin serves as a loading control. Right panel, relative level of Bax protein in TC, T16, and T18 cells. The protein bands were quantitated using a densitometer and were normalized to beta-actin signal. ::: ![](1471-2407-5-50-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Expression profile of Bcl-2 family genes in TC, T16, and T18 tumor cells. A. Ribonuclease protection assay (RPA) for Bcl-2 family genes. Antisense radiolabeled multiprobes were prepared using a RPA kit for detection of Bcl-2 family genes. B. Relative level of Bcl-2 family genes quantitated using a densitometer. Results are from three RPA experiments. Error bars represent standard deviation. No significant difference (p \>\> 0.05, t-test) in the expression of any Bcl-2 family genes was observed for TM40D-cont (TC), and TM40D-Mp (T16, T18) cells. ::: ![](1471-2407-5-50-5) ::: ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Maspin regulates the stability of Bcl-2 family proteins and the release of cytochrome c from mitochondria. A. Bax stability in TM40D cells. TC and T18 cells were induced with STS (1 μM) for 1 hour before they were treated by CHX (50 μM) for 0, 2, 3, and 4 hours. DMSO was used as a vehicle control (-CHX, 4 hrs). β-actin was used as a loading control. B. Release of cytochrome c protein from mitochondria to cytosol. TC, T16, and T18 tumor cells were treated with serum starvation for 48 hrs before they were harvested for preparation of cell extract and for cellular fractionation. Mitochondria were removed from cytosolic fraction by centrifugation. Both whole cell extracts and cytosolic fractions were subjected to Western blot analysis using antibodies against cytochrome c and beta-actin. β-actin was used as a loading control. ::: ![](1471-2407-5-50-6) :::	PubMed Central	2024-06-05T03:55:57.623089	2005-5-20	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1156873/", "journal": "BMC Cancer. 2005 May 20; 5:50", "authors": [ { "first": "Weiguo", "last": "Zhang" }, { "first": "Heidi Y", "last": "Shi" }, { "first": "Ming", "last": "Zhang" } ] }
PMC1156874	Background ========== Cancer still appears to be a challenging disease to treat. According to most recent estimates, more than 10 million new cancer cases were reported in the year 2000 killing around 6 million people \[[@B1]\]. In addition, 10 % of all cancers appear to be the breast cancer. Being the most frequently diagnosed cancer type in women, the breast cancer claims about 370,000 deaths each year around the world \[[@B2]\]. Surgery, radiotherapy and chemotherapy are among the most widely used treatment methods for patients with breast cancer \[[@B3]-[@B5]\]. Still, these conventional treatment modalities did not improve the survival rate of patients with locally advanced or metastatic breast cancer. With standard therapy, locally advanced breast cancer has a five year survival rate of 55 % and a ten year survival rate of 35 % \[[@B6]\]. There is a 40 % recurrence rate after ten years following the diagnosis and removal of primary tumor in patients with breast cancer \[[@B7]\]. For all these reasons, novel treatment methods are needed for the treatment of patients with breast cancer. Induction of programmed cell death known as apoptosis \[[@B8]\], appears to be a viable alternative to currently employed treatment modalities in the fight against cancer \[[@B9]\]. In order for chemotherapy and radiotherapy treatment options to work as anticancer agents; tumor suppressor gene, p53, is required \[[@B10]\]. Unfortunately, p53 mutations are acquired during the progression of cancer in more than half of the human tumors \[[@B11],[@B12]\]. Therefore, the resistance to both chemotherapy and radiotherapy is almost unavoidable in tumors lacking p53 \[[@B13]\]. On the other hand, death ligands are capable of inducing apoptosis independently of p53 status of cells \[[@B14]\]. Because of this reason, death ligands are currently considered as anticancer agents \[[@B15]\]. Among the death ligands tested, Tumor Necrosis Factor (TNF) \[[@B16]-[@B18]\] and FasL \[[@B19]\] effectively induced apoptosis in cancer cells. However, due to their systemic toxicity, the application of these agents in cancer gene therapy is very limited. The discovery of a novel death ligand, TRAIL \[[@B20],[@B21]\], changed this view, since unlike other members of the TNF family, TRAIL selectively killed cancer cells without causing any harm to normal cells \[[@B22]\]. Thus, treating tumor cells with TRAIL ligand appeared as an invaluable way of inducing apoptosis specifically in tumor cells, as normal cells are protected against the death-inducing effects of TRAIL \[[@B23],[@B24]\]. However, the mechanism of TRAIL resistance in normal cells is not understood \[[@B25]\] and significant proportions of cancer cells \[[@B26]\] including those of breast \[[@B27],[@B28]\] appeared to be TRAIL resistant. Consequently, TRAIL resistance constitutes a barrier if one wishes to use TRAIL as a death ligand in any breast cancer gene therapy approach. Resistance to TRAIL-induced apoptosis in normal cells was initially considered to be caused by the presence of decoy receptors (TRAIL-R3 and TRAIL-R4), which compete with death receptors (TRAIL-R1 and TRAIL-R2) for binding to TRAIL \[[@B29],[@B30]\]. So far, no correlation between TRAIL sensitivity and the expression pattern of TRAIL receptors has been demonstrated in cancer cells yet \[[@B31]\]. The presence of intracellular apoptosis inhibitory substances (bcl-xL, c-FLIP, cIAP etc.) was also blamed to be responsible for TRAIL resistance \[[@B31]-[@B33]\]. Intriguingly, the engagement of both TRAIL death receptors and TRAIL-R4 decoy receptor also activated NF-kB pathway \[[@B24],[@B34],[@B35]\]. Because NF-kB activation is known to hamper the apoptotic pathways in cells by up-regulating the expression of various apoptosis inhibitory molecules such as cFLIP, bcl-xL, c-IAP and the decoy receptor TRAIL-R3 \[[@B34],[@B36],[@B37]\], high levels of NF-kB activation might be a strong factor responsible for blocking apoptotic processes in order to establish TRAIL resistance. For this reason, we analyzed both the TRAIL induced as well as endogenous NF-kB activities using Luciferase reporter gene assays in MCF7 breast cancer cells. Because TRAIL-R1, TRAIL-R2 and TRAIL-R4 induced NF-kB activation has been shown to be primarily mediated by TRAF2-NIK-IkappaB kinase alpha/beta signaling cascade \[[@B35]\], MCF7 breast cancer cells were coinfected with adenovirus vectors encoding a dominant negative mutant to IKKβ(AdIKKβKA) \[[@B38]\] and hTRAIL (Ad5hTRAIL) in order to test if TRAIL resistance in breast cancer cells is eliminated through the inhibition of IKK, a leading modulator of NF-kB. The molecular mechanism of TRAIL resistance in breast cancer cells (MCF7 and MDA-MB-231) was studied by novel Real Time RT-PCR assays and conventional flow cytometry in order to verify if there is any relationship between TRAIL resistance and the expression pattern of TRAIL receptors. Lastly, a DcR2 siRNA approach was utilized to knock down the expression of relevant TRAIL decoy receptor in order to reveal its connection to TRAIL resistance. Methods ======= Recombinant adenovirus vector production ---------------------------------------- Amplification of the vectors Ad5hTRAIL \[[@B39]\], AdIKKβKA \[[@B17]\], AdEGFP \[[@B18]\], AdCMVLacZ \[[@B40]\] and AdNFkBLuc \[[@B38]\] was performed as previously described \[[@B41]\]. Amplified vectors were stored at -80°C in 10 mM Tris with 20 % glycerol. AdIKKβKA expresses a dominant negative mutant of IKKβ, which interacts with other IKK subunits to form inactive IKK complexes. The particle titers of adenoviral stocks were in the range of 10^13^DNA particles/ml, whereas the typical particle/plaque forming unit ratio was equal to 50. Infection of breast cancer cells with first generation recombinant adenovirus vectors ------------------------------------------------------------------------------------- Breast cancer cell lines were cultured in RPMI 1640 medium supplemented with 10 % FBS, 2.2 g/l sodium bicarbonate, 1 mM L-glutamine, and 1 % penicillin-streptomycin mixture, at 37°C in a humidified 5 % CO~2~atmosphere. Experimental steps of transduction of breast cancer cells with adenoviral vectors can be summarized as follows: Breast cancer cells were infected with an increasing multiplicity of infection (MOI) of AdEGFP (vector expressing enhanced green fluorescent protein (EGFP) reporter gene) vector at 37°C in RPMI 1640 without FBS. Two hours following infection, equal volume of RPMI 1640 supplemented with 20 % FBS was added to increase the serum concentration in the media to 10 %. 48 hours after the infection, the level of transduction was detected by examining of the percentage of GFP (+) cells under a fluorescent microscopy and subsequently by flow cytometry. Propidium iodide exclusion technique was used to determine the cell viability. Overexpression of hTRAIL was provided by Ad5hTRAIL infection. Cells were coinfected with adenovirus vectors encoding IKKβ dominant negative mutant (AdIKKβKA) and Ad5hTRAIL in order to block IKK activity thereby NF-kB activation. NF-kB promoter based Luciferase assay system was utilized to conduct NF-kB transcription activation assays using AdNFkBLuc construct. AdCMVLacZ vector was used as a control. NF-kB directed transcription activation assays ---------------------------------------------- AdNFkBLuc construct was utilized in order to determine the NF-kB activation status of MCF7 cells. AdNFkBLuc vector \[[@B38]\] possesses four tandem copies of the NF-kB consensus sequence fused to a TATA-like promoter from the herpes simplex virus-thymidine kinase gene driving the expression of a Luciferase reporter. Transcriptional induction mediated by NF-kB in the presence or absence of TRAIL was measured according to the manufacturer\'s protocol using the Luciferase assay system with Reporter Lysis Buffer (Promega, Inc.). All measurements of Luciferase activity expressed as relative light units were normalized against the protein concentration. Cell viability assays --------------------- Discrimination of live cells from dead cells was performed using Live/Dead Cellular Viability/Cytotoxicity Kit from Molecular Probes (Eugene, OR). This assay is based on the use of Calsein AM and Ethidium homodimer-1 (EthD-1). Calsein AM is a fluorogenic substrate for intracellular calsein esterase. It is modified to a green fluorescent compound (calsein) by active esterase in live cells with intact membranes, thus serves as a marker for viable cells. Unharmed cell membranes do not allow EthD-1, a red fluorescent nucleic acid stain, to enter inside the cell. However, cells with damaged membrane uptake the dye and stain positive. Apoptosis detection by Annexin V binding ---------------------------------------- Annexin V conjugated to fluorochromes such as FITC has successfully been used as probes to detect cells undergoing apoptosis. Annexin V binding assays were carried out according to manufacturer\'s instructions (Alexis Biochemicals). For this purpose, a FITC conjugated mouse monoclonal antibody to human Annexin V (ALX-804-100F-T100) was employed to detect apoptotic cells via flow cytometry. The detection of TRAIL receptor expression profile by flow cytometry -------------------------------------------------------------------- Anti-TRAIL receptor flow cytometry set (Cat. ALX-850-273-KI01) was used to detect TRAIL receptor protein expression on cell surface. This kit contains 100μgs of MAb to TRAIL-R1 (clone HS101, Cat. 804-297A), -R2 (clone HS201, Cat.804-298A), -R3 (clone HS301, Cat. 804-344A) and -R4 (clone HS402, Cat. 804-299A). Primary antibodies were used at 5 μg/ml concentration. Biotinylated goat anti-mouse IgG1 (Cat. ALX-211-202) was used as a secondary antibody followed by streptavidin-PE (Cat. ANC-253-050) prior to flow cytometry. Flow analysis was performed according to manufacturer\'s protocols using BD FACSCALIBUR at the Akdeniz University Hospitals. Purified mouse IgG1 (MOPC 31C, Cat. ANC-278-010) served as an isotype control. Quantitative Real Time RT-PCR assay for human TRAIL receptors ------------------------------------------------------------- TRIzol reagent (Life Technologies, Gaithersburg, MD) was used to extract total RNA from breast cancer cells, according to the instructions from the manufacturer. Reverse transcription of 2 μg of total RNA was performed using TaqMan Reverse Transcription Reagents (Applied Biosystems Cat. N8080234). Despite the fact that the sequences for TRAIL-R1 and TRAIL-R2 primers and probes were recently described by our group \[[@B42]\], we had to design new probe sets for the decoy receptors. Following is the sequence information for TRAIL decoy receptor sets: TRAILR3-5\'CCC-TAA-AGT-TCG-TCG-TCG-TCA-T, TRAILR3-3\'GGG-CAG-TGG-TGG-CAG-AGT-A, TRAILR3 Probe:5\'6FAM-TCGCGGTCCTGCTGCCAGTCCTAGC-TAMRA 3\'; TRAILR4-5\'ACA-GAG-GCG-CAG-CCT-CAA, TRAILR4-3\'ACG-GGT-TAC-AGG-CTC-CAG-TAT-ATT, TRAILR4 Probe: 5\'6FAM-AGGAGGAGTGTCCAGCAGGATCTCATAGATC-TAMRA 3\'. rRNA was amplified as an internal control in the same reaction. Both the rRNA primers and probes were obtained from PE Applied Biosystems (Cat. 4308329). ΔΔCt method was used as described by Applied Biosystems to calculate the relative quantities of TRAIL receptors. The TaqMan PCR reaction was performed as described by the manufacturer (Applied Biosystems Cat. N8080228). A DcR2 siRNA approach targeting TRAIL-R4 expression --------------------------------------------------- Posttranscriptional silencing of gene expression became a very useful approach within the last couple of years in research. DcR2 siRNA experiments were performed using DcR2 siRNA (sc-35185), siRNA transfection medium (sc-36868) and siRNA transfection reagent (sc-29528) in MCF7 breast cancer cells as described by the manufacturer (Santa Cruz Biotechnology). Flow cytometry analysis was performed to assess any changes in TRAIL-R4 gene expression. MCF7 cells were infected with Ad5hTRAIL or AdCMVLacZ vectors at increasing doses 35 hours following the transfection. Molecular Probe\'s Live/Dead Cellular Viability/Cytotoxicity Kit was used to assess the amount of live cells 48 hours following the infection. Results ======= MCF7 breast carcinoma cells were efficiently transduced with recombinant adenoviruses ------------------------------------------------------------------------------------- In order to find out the efficacy of transduction of breast cancer cells by first generation adenoviral vectors, MCF7 cells were infected with increasing Multiplicity of Infection (MOI) of adenovirus encoding Enhanced Green Fluorescent Protein (AdEGFP). The transduction profiles were followed under fluorescent microscopy and the results were quantitatively analyzed by flow cytometry 48 hours following the infection (Figure [1](#F1){ref-type="fig"}). While an MOI of 5000 DNA particles/cell was sufficient to transduce more than 90 % of the cells, nearly 100 % of the cells were transduced with AdEGFP at an MOI of 10,000 DNA particles/cell. These assays were also pivotal in obtaining the optimum dose of adenovirus required for efficient transduction of MCF7 breast carcinoma cell line without observing deleterious cytotoxic effects. These results demonstrated that breast cancer cells were transduced successfully with recombinant adenoviral vectors. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### First generation adenoviral vectors efficiently transduced MCF7 breast cancer cells. MCF7 cells were infected with increasing MOIs of AdEGFP for 48 hours prior to analysis. The number of EGFP expressing cells was detected under fluorescent microscopy (Panel A), and analyzed by flow cytometry (Panel B). Numbers represent viral doses applied in MOI values as DNA particles/cell. ::: ![](1471-2407-5-54-1) ::: MCF7 breast cancer cells displayed complete resistance to TRAIL --------------------------------------------------------------- Although TRAIL appeared as a promising therapeutic ligand to treat cancer, a variety of tumor types were reported to be resistant to TRAIL-induced cell death. For this reason, we wanted to investigate if exogenous TRAIL expression delivered by adenovirus vectors would induce killing of breast cancer cells. To test this, MCF7 cells were infected with increasing titers of Ad5hTRAIL or AdCMVLacZ. Amount of viable cells were detected using Molecular Probe\'s Live/Dead Cellular Viability/Cytotoxicity Kit 48 hours following the infections (Figure [2](#F2){ref-type="fig"}). MCF7 cells displayed complete resistance to TRAIL, as no reduction in the level of viable cells was observed even at an MOI of 10,000 DNA particles/cell, at which almost all cells were infected. Thus, it was concluded that MCF7 breast cancer cells were completely resistant to adenovirus delivery of TRAIL. Similarly, AdCMVLacZ infection alone revealed no significant degree of cell death either (data not shown). ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Ad5hTRAIL or AdIKKβKA infection alone did not decrease the viability of MCF7 breast cancer cells. MCF7 cells were infected with increasing MOIs of either Ad5hTRAIL or AdIKKβKA construct. Cell viability was detected using Molecular Probe\'s Live/Dead Cellular Viability/Cytotoxicity Kit 48 hours following the infection. Numbers represent viral doses applied, in MOI values as DNA particles/cell. ::: ![](1471-2407-5-54-2) ::: Blocking IKK induced NF-kB activation pathway alone did not cause any reduction in the viability of MCF7 breast carcinoma cells ------------------------------------------------------------------------------------------------------------------------------- Because increased NF-kB activity was claimed to be responsible for the resistance to death ligand induced cytotoxicity in some tumors \[[@B36],[@B37]\], we wanted to test if the inhibition of IKK activity thereby NF-kB would reduce the viability of breast cancer cells. In order to block the intracellular anti-apoptotic NF-kB pathway, MCF7 cells were infected with increasing MOIs of adenoviral vectors encoding a dominant negative mutant of IKKβ(AdIKKβKA), a key molecule involved in the activation of NF-kB. Cell viability was examined 48 hours following the infection under fluorescent microscope (Figure [2](#F2){ref-type="fig"}). Interestingly, AdIKKβKA vector alone proved inefficient in reducing the viability of MCF7 cells, even at an MOI of 10,000 DNA particles/cell. Adenovirus delivery of IKKβKA gene expression sensitized MCF7 breast cancer cells to TRAIL-induced apoptosis ------------------------------------------------------------------------------------------------------------ Adenovirus-mediated delivery of IKKβ (Ad.IKKβKA) \[[@B17],[@B18]\] or IkBα (Ad.IkBαSR) \[[@B40],[@B43]\] dominant negative mutants have previously been demonstrated to sensitize lung cancer cells to TNF death ligand. Because most of the breast cancer cell lines tested appeared to be TRAIL resistant \[[@B27],[@B28]\], NF-kB targeting strategies involving IKK inhibition was employed to verify whether MCF7 breast carcinoma cells were sensitized to TRAIL under these circumstances. To accomplish this, MCF7 cells were coinfected with a constant MOI of Ad5hTRAIL construct and increasing doses of AdIKKβKA vector. In order to better assess the sensitization phenomenon, Ad5hTRAIL was infected at two different MOIs into MCF7 breast cancer cell lines. While a constant MOI of 1000 DNA particles/cell of Ad5hTRAIL was used in infection experiments depicted on Figure [3](#F3){ref-type="fig"}, infection experiments conducted at an MOI of 5000 DNA particles/cell are displayed in Figure [4](#F4){ref-type="fig"}. The amount of viable cells was detected 48 hours following the infections using Molecular Probe\'s Live/Dead Cellular Viability/Cytotoxicity Kit. Intriguingly, MCF7 cells were sensitized to TRAIL only when Ad5hTRAIL was coinfected with AdIKKβKA vector. For instance, nearly 55 % cell death was observed when cells were coinfected with 1000 MOI of Ad5hTRAIL and 5000 MOI of AdIKKβKA constructs (Figure [3](#F3){ref-type="fig"}). When MOI of Ad5hTRAIL was increased to 5000 as depicted on Figure [4](#F4){ref-type="fig"}, the death rate went up to 90 %. On the other hand, AdCMVLacZ infection instead of AdIKKβKA in breast cancer cells revealed no TRAIL sensitization (data not shown). These results suggested that IKKβKA expression via adenoviral vectors defeated TRAIL resistance observed in MCF7 breast cancer cells. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### IKKβKA expression via adenoviral vectors sensitized MCF7 cells to TRAIL-mediated apoptosis. MCF7 cells were infected with increasing doses of adenoviral vectors encoding dominant negative mutant of IKKβ (as shown below each panel), while simultaneous infection with Ad5hTRAIL (as shown above each panel) was performed at a constant MOI of 1000. Cell viability was detected using Molecular Probe\'s Live/Dead Cellular Viability/Cytotoxicity Kit 48 hours following infection. Numbers represent viral doses applied in MOI values as DNA particles/cell. Fluorescent micrographs are provided in Panel A; Panel B depicts quantitative analysis of such infections. Values represent the mean (± SEM) of three different experiments. ::: ![](1471-2407-5-54-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### AdIKKβKA infection defeated the resistance to TRAIL-induced apoptosis in MCF7 breast cancer cells. These cells were coinfected with a constant MOI of 5000 DNA particles/cell of Ad5hTRAIL (as shown above each panel) and increasing doses of AdIKKβKA (as shown below each panel). Live/Dead Cellular Viability/Cytotoxicity Kit from Molecular Probe was used to detect TRAIL cytotoxicity 48 hours following infection. Numbers represent viral doses applied, in MOI values as DNA particles/cell. Data represent the mean of (± SEM) six independent data points (n = 6). ::: ![](1471-2407-5-54-4) ::: Exogenous TRAIL overexpression elevated the basal NF-kB activity in MCF7 cells, whereas IKKβKA expression blocked both TRAIL-induced and basal NF-kB activities --------------------------------------------------------------------------------------------------------------------------------------------------------------- It is well known that different tumor cells display diverse levels of endogenous NF-kB activities. Furthermore, intracellular NF-kB activity in tumor cells is upregulated by both TRAIL death receptors (TRAIL-R1 and TRAIL-R2) \[[@B34],[@B44]\] as well as TRAIL decoy receptor TRAIL-R4 \[[@B45]\] upon ligand binding. Knowing the endogenous NF-kB status of cancer cells before the therapy is obviously crucial for TRAIL mediated gene therapy targeting to induce apoptosis in cancer cells. A coinfection experiment was performed using a recombinant adenovirus vector carrying NF-kB driven Luciferase reporter gene (AdNFkBLuc) and Ad5hTRAIL vector in order to study the extent of NF-kB activation as a result of TRAIL overexpression in MCF7 breast cancer cell line. NF-kB Luciferase assays were conducted 24 hours following the infection in order to determine cell\'s NF-kB activation status. As seen in Figure [5](#F5){ref-type="fig"}, Ad5hTRAIL at an MOI of 5000 DNA particles/cell (Panel B) but not at an MOI of 1000 DNA particles/cell (Panel A) stimulated NF-kB activation. In order to determine the magnitude of NF-kB inhibition, a triple coinfection experiment involving AdNFkBLuc, Ad5hTRAIL and AdIKKβKA or AdCMVLacZ was performed. While IKKβKA overexpression in MCF7 cells gradually reduced both the TRAIL-induced and basal NF-kB activities in MCF7 cells, no such NF-kB inhibiting effect was observed in cells upon super-infection with AdCMVLacZ virus as a control (Figure [5](#F5){ref-type="fig"}). ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Distinctive regulation of NF-kB activation in MCF7 breast cancer cells by Ad5hTRAIL and/or AdIKKβKA infections. MCF7 cells were simultaneously infected with AdNFkBLuc, Ad5hTRAIL and/or increasing doses of AdIKKβKA construct for 24 hours. AdCMVLacZ infection was also performed as a negative control. The types of constructs used in the infection are shown on the x axis. MOI values represent DNA particles/cell. Ad5hTRAIL vector was used at two different constant MOIs (MOI of 1000 and 5000) in order to avoid cell death complicating our assay result. Luciferase activity expressed in Relative Light Units per microgram protein is shown on y axis. Values represent the mean (± SEM) of six independent data points (n = 6). ::: ![](1471-2407-5-54-5) ::: Coinfection of Ad5hTRAIL and AdIKKβKA results in apoptotic cell death in MCF7 breast cancer cells ------------------------------------------------------------------------------------------------- To show that apoptosis is the mechanism of cell death mediated by TRAIL overexpression under the setting of IKK inhibition in MCF7 cells, Annexin V staining was performed using flow cytometry. For this purpose, MCF7 cells were infected with Ad5hTRAIL or AdIKKβKA vectors alone or in combination. Thirty-five hours following the infection, apoptotic cell death was analyzed by Annexin-V-FITC staining. As displayed in Figure [6](#F6){ref-type="fig"} Panel A, there was no substantial Annexin V binding generated by the expression of TRAIL or IKKβKA in MCF7 cells. However, considerable levels of Annexin V binding were observed in cells coinfected with Ad5hTRAIL and AdIKKβKA indicating apoptotic cell death (Figure [6](#F6){ref-type="fig"}, Panel B). As predicted, Ad5hTRAIL and AdCMVLacZ (negative control) coinfection did not yield any significant levels of Annexin V binding as MCF7 cells are resistant to TRAIL in the absence of IKK inhibition. These results suggested that the mechanism of cell death experienced by MCF7 cells is apoptosis following TRAIL stimulation under the setting of IKK inhibition. ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Ad5hTRAIL and AdIKKβKA coinfection induced apoptosis in MCF7 breast carcinoma cells. FITC conjugated Annexin V and Propidium Iodide (PI) staining were utilized using MCF7 cells infected with various combinations of adenovirus constructs as described in Methods prior to flow cytometry. Each histogram represents 10^4^gated MCF7 cells. Histograms were illustrated in two panels for clarity. Various treatment settings were provided in Panel A. MOI of 5000 DNA particles/cell was used for each viral construct unless stated otherwise in the Figure. Control line represents uninfected but FITC-Annexin V and PI stained MCF7 cells. Only one representative assay out of three independent assays was provided. ::: ![](1471-2407-5-54-6) ::: MCF7 breast cancer cell line displayed significant levels of TRAIL decoy receptor-4 expression ---------------------------------------------------------------------------------------------- So far no evidence of the connection between the expression pattern of TRAIL receptors and TRAIL sensitivity was found in cancer cells \[[@B31]\]. Part of the reason might have been the inability to screen all TRAIL receptors at once in breast cancer cells then \[[@B28]\]. In order to compensate this deficiency, quantitative novel Real Time RT-PCR assays were conducted using primer-probe sets specifically designed to detect each TRAIL receptor in MCF7 breast cancer cells (Figure [7](#F7){ref-type="fig"}, Panel A). According to our results, while all TRAIL receptors were expressed in MCF7 cells, TRAIL-R4 expression was the highest among the four. In addition, the level of TRAIL-R2 expression was much higher than that of TRAIL-R1. Lastly, TRAIL-R3 decoy receptor expression was the lowest. These results suggested that high levels of TRAIL-R4 decoy receptor expression correlated well with TRAIL resistance. However, as the gene expression detected inside the cell may not necessarily correlate with the receptor expression on cell surface, we decided to perform flow cytometry analysis using antibodies specific to four different TRAIL receptors. As shown in Figure [7](#F7){ref-type="fig"} Panel B, MCF7 cells expressed all TRAIL receptors excluding TRAIL-R3 on cell surface. While similar levels of TRAIL death receptors TRAIL-R1 and TRAIL-R2 were expressed, there were still considerable levels of TRAIL-R4 decoy receptor expression on the surface of MCF7 cells. ::: {#F7 .fig} Figure 7 ::: {.caption} ###### MCF7 breast carcinoma cell line displayed substantial levels of TRAIL-R4 decoy receptor expression. Quantitative Real Time RT-PCR of TRAIL receptors was performed as described in Methods (Panel A). TRAIL receptor levels per 25 pg of ribosomal cDNA are presented in the graph for clarity. Ribosomal RNA primers and probes were included in each TaqMan reaction as an internal control. Panel B depicts the surface TRAIL receptor expression pattern of MCF7 cells using flow cytometry. Experimental parameters are defined in colored lines. 10^4^cells were gated for each histogram. Only one representative assay for each experiment (independently repeated three times) is shown. ::: ![](1471-2407-5-54-7) ::: TRAIL sensitive MDA-MB-231 cells displayed very low levels of TRAIL-R4 decoy receptor expression on cell surface ---------------------------------------------------------------------------------------------------------------- In order to solidify the importance of TRAIL-R4 expression and its connection to TRAIL resistance, another breast cancer cell line, MDA-MB-231, was also analyzed in terms of TRAIL receptor expression profile. Real Time RT-PCR assays revealed that while TRAIL-R2 expression was the highest on transcript levels, TRAIL-R4 decoy receptor expression was the lowest TRAIL receptor expressed in MDA-MB-231 breast cancer cells (Figure [8](#F8){ref-type="fig"}, Panel A). Furthermore, flow cytometry analysis indicated that insignificant levels of TRAIL-R4 expression were detected on the surface of MDA-MB-231 breast cancer cells (Figure [8](#F8){ref-type="fig"}, Panel B). TRAIL-R3 decoy receptor expression, however, was not detectable using flow cytometry. Intriguingly, in contrast to what was observed with MCF7, adenovirus delivery of TRAIL alone killed significant proportions of MDA-MB-231 breast cancer cells (Figure [9](#F9){ref-type="fig"}). ::: {#F8 .fig} Figure 8 ::: {.caption} ###### MDA-MB-231 breast cancer cells displayed trivial levels of TRAIL-R4 decoy receptor expression on surface. TRAIL receptor composition of MDA-MB-231 breast cancer cells revealed by Real Time RT-PCR assay is displayed in Panel A. Panel B illustrates flow cytometry analysis showing the surface expression pattern of TRAIL receptors. 10^4^cells were gated for each histogram. Only one representative assay out of three is shown. ::: ![](1471-2407-5-54-8) ::: ::: {#F9 .fig} Figure 9 ::: {.caption} ###### MDA-MB-231 breast cancer cell line is sensitive to Ad5hTRAIL infection. MDA-MB-231 breast cancer cells were infected with increasing MOIs of Ad5hTRAIL construct. Molecular Probe\'s Live/Dead Cellular Viability/Cytotoxicity Kit was used to detect % viable cells 48 hours following the infection. Numbers represent viral doses applied in MOI values as DNA particles/cell. Values represent the mean (± SEM) of six independent data points (n = 6). ::: ![](1471-2407-5-54-9) ::: Lowering of TRAIL-R4 gene expression sensitized MCF7 breast cancer cells to TRAIL --------------------------------------------------------------------------------- In order to solidify the connection between TRAIL-R4 decoy receptor gene expression and TRAIL resistance, a DcR2 siRNA approach was executed in TRAIL resistant MCF7 breast cancer cells. Flow cytometry analysis conducted 35 hours following the transfection revealed that the level of TRAIL-R4 protein expression on surface went down drastically (Figure [10](#F10){ref-type="fig"}, Panel A). At this stage, MCF7 cells were further infected with either Ad5hTRAIL or AdCMVLacZ vector at increasing doses. Cell viability assays were conducted 48 hours following the infection. Only Ad5hTRAIL infected cells exhibited considerable amount of cell death following transfection (Figure [10](#F10){ref-type="fig"}, Panel B). No such effect was observed when cells were infected with AdCMVLacZ virus (data not shown). ::: {#F10 .fig} Figure 10 ::: {.caption} ###### Knocking down TRAIL-R4 expression sensitized MCF7 breast cancer cells to TRAIL. A DcR2 siRNA approach was administered as described in Methods using TRAIL resistant MCF7 breast cancer cell line. Panel A depicts a flow cytometry analysis confirming strong attenuation of TRAIL-R4 expression on cell surface. TRAIL-R2 death receptor expression was also detected as a control. Sensitization of MCF7 breast cancer cells to TRAIL following a DcR2 siRNA approach is provided in Panel B. MCF7 breast cancer cells were infected with increasing doses of Ad5hTRAIL alone following a DcR2 siRNA transfection. Cell death was detected 48 hours following the infection (Panel B). Data represent the mean (± SEM) of 6 independent data points. ::: ![](1471-2407-5-54-10) ::: Discussion ========== Although, conventional treatment modalities could not satisfactorily improve the survival rates of patients with locally advanced and metastatic disease, adenovirus delivery of death ligands represents a feasible choice for the treatment of patients with breast cancer. However, recent observations demonstrating that a considerable portion of human cancers including those of the breast \[[@B27],[@B28]\] were TRAIL resistant undermined the potential application of TRAIL against cancer. Accordingly, the understanding of the mechanism of TRAIL resistance is the key to resolve primary obstacles in TRAIL mediated gene therapy approach. Based on recent findings from our laboratory and others, we think that NF-kB signaling is one of the most crucial pathways involved in the constitution of TRAIL resistance \[[@B26]\]. Despite the fact that TRAIL-R1, TRAIL-R2 and TRAIL-R4 induced NF-kB activation has been shown to be primarily mediated by TRAF2-NIK-IkappaB kinase alpha/beta signaling cascade \[[@B35]\], there is some doubt on whether or not NF-kB activation can block TRAIL mediated apoptosis. For example, in one particular study it was reported that NF-kB inhibition by way of IkappaBalpha mutant expression sensitized MCF7 cells to TNF but not TRAIL-induced apoptosis \[[@B35]\]. Considering the fact that there are different ways to activate NF-kB pathway (IkB dependent and independent ways) \[[@B46]\] we decided to inhibit IKK activity rather than targeting IkappaBalpha itself to look for the possibility of sensitizing MCF7 breast cancer cells to TRAIL. First of all, in order to find out the efficacy of adenovirus transduction in breast cancer cells, MCF7 cells were infected with increasing MOIs of AdEGFP virus. The transduction profiles analyzed by flow cytometry showed that nearly 100 % of the cells were transduced with AdEGFP at an MOI of 10,000 DNA particles/cell (Figure [1](#F1){ref-type="fig"}). The efficacy of TRAIL in mediating apoptosis of MCF7 breast cancer cells was assessed using Ad5hTRAIL construct. Interestingly, MCF7 cells displayed complete resistance to TRAIL as no reduction in the level of viable cells was observed even at an MOI of 10,000 DNA particles/cell (Figure [2](#F2){ref-type="fig"}). IKK inhibiting strategy alone proved inefficient in reducing the viability of MCF7 cells suggesting that an apoptotic stimulus was required in order to induce cell killing (Figure [2](#F2){ref-type="fig"}). Interestingly, in order to break down TRAIL resistance and to induce cell death, a coinfection of MCF7 cells with Ad5hTRAIL and AdIKKβKA was required (Figures [3](#F3){ref-type="fig"} and [4](#F4){ref-type="fig"}). Luciferase assays confirmed that both the TRAIL induced and endogenous NF-kB activities were drastically reduced by the infection of MCF7 cells with AdIKKβKA virus (Figure [5](#F5){ref-type="fig"}). Moreover, IKKβKA sensitization of MCF7 breast carcinoma cells resulted in TRAIL induced apoptosis as revealed by Annexin V binding assays (Figure [6](#F6){ref-type="fig"}). These results suggested that NF-kB activation pathway has a hampering effect on TRAIL-induced cell death in MCF7 cells, and blocking this pathway is essential to sensitize breast cancer cells to TRAIL mediated apoptosis. So far, no correlation between TRAIL resistance and TRAIL decoy receptor gene expression has been reported. For example, analysis of breast cancer cell lines by just examining the expression levels of TRAIL death receptors (TRAIL-R1 and TRAIL-R2) and TRAIL-R3 decoy receptor using RNase protection assay did not reveal any connection between the expression pattern of TRAIL receptors and TRAIL resistance \[[@B28]\]. But whether or not TRAIL-R4 decoy receptor gene expression in any way contributes to TRAIL resistance in breast cancer cells remains to be tested yet. Quantitative Real Time RT-PCR assays were developed in order to assess the level of TRAIL receptor gene expression in breast carcinoma cells. While all TRAIL receptors were detectable in MCF7 breast carcinoma cell line, the level of TRAIL-R4 decoy receptor gene expression was the highest among the four (Figure [7](#F7){ref-type="fig"}, Panel A). This intriguing observation is consistent with a previous report suggesting that transient TRAIL-R4 overexpression protected target cells from TRAIL induced cytotoxicity \[[@B45]\]. TRAIL R4 is known to protect cells from apoptosis by acting both as a decoy receptor and an antiapoptotic signal provider. While Real Time PCR assay is useful in assessing the level of gene expression on mRNA levels, obviously this assay does not necessarily reflect TRAIL receptor composition on cell surface. For this reason, conventional flow cytometry analysis was carried out in order to determine the level of TRAIL receptor protein expression on cell surface. Despite the presence of TRAIL death receptors, substantial levels of TRAIL-R4 decoy receptor expression were detectable on the surface of MCF7 breast carcinoma cells (Figure [7](#F7){ref-type="fig"}, Panel B). On top of that, TRAIL sensitive MDA-MB-231 cell line (Figure [9](#F9){ref-type="fig"}) displayed very low levels of TRAIL-R4 decoy receptor expression on cell surface (Figure [8](#F8){ref-type="fig"}, Panel B). Neither of the cell lines expressed detectable levels of TRAIL-R3 decoy receptor on surface. Intriguingly, administration of a DcR2 siRNA approach lowered surface TRAIL-R4 expression and sensitized MCF7 breast cancer cells to TRAIL (Figure [10](#F10){ref-type="fig"}). Conclusion ========== Our results demonstrated that the expression of TRAIL-R4 decoy receptor but not TRAIL-R3 appeared to correlate well with TRAIL resistance phenotype observed in MCF7 breast cancer cells. Further screening of another breast cancer cell line, MDA-MB-231, revealed that low levels of TRAIL-R4 expression on surface were correlated with TRAIL sensitivity. These results strengthen our argument that TRAIL-R4 but not TRAIL-R3 is the decoy receptor which appeared to influence TRAIL sensitivity in breast cancer cells. This is further confirmed by a DcR2 siRNA assay which suggested that down regulation of TRAIL-R4 expression sensitized MCF7 breast cancer cells to TRAIL. In addition, the inhibition of IKK pathway thereby NF-kB sensitized MCF7 cells to TRAIL induced apoptosis despite the expression of TRAIL-R4 decoy receptor on cell surface. Consequently, this complementary gene therapy approach involving IKK inhibition might be necessary to breakdown TRAIL resistance encountered in patients with breast cancer. Abbreviations ============= TRAIL= Tumor Necrosis Factor (TNF)-Related Apoptosis-Inducing Ligand, EGFP= Enhanced Green Fluorescent Protein, MOI= Multiplicity of Infection, DcR2= Decoy receptor 2. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= ADS performed cell viability, Luciferase, Flow Cytometry, Real Time RT-PCR and siRNA assays, ED assisted ADS with adenovirus preparation, CA performed AdEGFP transduction assays, NE cultured breast cancer cells, SK optimized flow cytometry assays, SS participated in the coordination and execution 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/1471-2407/5/54/prepub> Acknowledgements ================ MDA-MB-231 cell line was kindly provided by Dr. Burhan Savas MD, PhD. This work is supported by grants from Akdeniz University Scientific Research Project Administration Division and the Health Science Institute (to SS).	PubMed Central	2024-06-05T03:55:57.626076	2005-5-25	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1156874/", "journal": "BMC Cancer. 2005 May 25; 5:54", "authors": [ { "first": "Ahter D", "last": "Sanlioglu" }, { "first": "Ercument", "last": "Dirice" }, { "first": "Cigdem", "last": "Aydin" }, { "first": "Nuray", "last": "Erin" }, { "first": "Sadi", "last": "Koksoy" }, { "first": "Salih", "last": "Sanlioglu" } ] }
PMC1156875	Background ========== Despite intensive multi-modality treatment including tumor resection, post-operative radiotherapy and frequently adjuvant chemotherapy, the prognosis of malignant glioma continues to be poor. Local recurrence, occuring almost exclusively inside the high-dose volume of post-operative radiotherapy \[[@B1]\], represents a major therapeutic challenge. Re-irradiation as a salvage treatment option is limited by the radiation tolerance of surrounding normal brain tissue. In recent years, single-dose radiosurgery, normofractionated and hypofractionated stereotactic radiotherapy have been investigated as a single modality or in combination with chemotherapy \[[@B2]-[@B8]\]. Re-irradiation appeared to be associated with acceptable toxicity when certain treatment volume and dose limits were respected. Median survival in these series was between 7 and 13 months from the time of salvage radiotherapy suggesting a therapeutic benefit in selected patient groups. Since published data on hypofractionated stereotactic radiotherapy as a sole modality in recurrent malignant glioma are limited \[[@B4],[@B5],[@B8]\], we now analyzed the results of such treatment in 19 consecutive patients treated at a single institution. Methods ======= Patients -------- Between 1997 and 2003, 19 patients were treated with hypofractionated stereotactic radiotherapy (HFSRT) for recurrent malignant glioma at the University of Würzburg, Germany. The patient characteristics are summarized in Table [1](#T1){ref-type="table"}. All patients had received previous involved-field radiotherapy with 45 to 61 Gy and 94% of patients had been pretreated with chemotherapy, most often nimustine (ACNU) and teniposide, a combination protocol favored by the Neuro-Oncology Working Group of the German Cancer Society \[[@B9]\]. Treatment technique ------------------- HFSRT was performed non-invasively using a commercially available relocatable mask system as previously described \[[@B10]\]. For planning CT, 3 mm scans of the brain were obtained after administration of an i. v. contrast agent. Treatment planning was performed using Helax TMS (Nucletron, Veenendal, Netherlands) software. For planning target volume (PTV) definition, a margin of 1 to 3 mm around the contrast-enhancing volume was used. Depending on the shape of the PTV, treatment plans were created containing multiple non-coplanar arcs, multiple non-coplanar fixed fields or combinations of both. The treatment plans were normalized to 100% at the isocenter and prescribed to a median isodose of 80% (range 70 to 90%), enclosing the PTV. Details of treatment planning and dose prescription are given in Table [2](#T2){ref-type="table"}. Immediately before the first treatment session, CT simulation was performed as described \[[@B10]\]. Briefly, the isocenter position was verified with regard to bony and parenchymal landmarks and to the contrast enhancing tumor and positioning errors were corrected after attaching the mask system to the treatment table. Treatment was delivered from a Philips SL 75/20 or Siemens Primus linear accelerator (6 MV or 8 MV photons, respectively) equipped with a manual micro-multileaf collimator. Subsequent CT simulations were performed at the discretion of the treating physician depending on the setup error determined at time of the first fraction. HFSRT treatment was performed five days per week. During the HFSRT series, patients were maintained on their previous corticosteroid dose or corticosteroids were started. Daily doses ranged between 32 mg of prednisolone and 40 mg of dexamethasone. Statistics ---------- The main endpoint of this study was overall survival (OS) from the time of HFSRT. Kaplan-Meier survival curves were calculated for these intervals considering death as an event and patients alive at last follow-up as censored. Subgroups were compared using the log-rank test. Statistical analysis was performed with Statistica 6.1 software (Tulsa, OK, USA). Overall survival from initial diagnosis and time to tumor progression were considered as secondary endpoints. For the analysis of time to recurrence, neuroradiological diagnoses of follow-up CT and MRI scans were used. Imaging was routinely reviewed by an interdisciplinary tumor board which based its recommendations on these studies. Results ======= HFSRT was well tolerated and no acute neurotoxicity or deterioration in the general health status was observed. Five of 19 patients were reoperated after HFSRT. In three of these, resection was performed for tumor progression on imaging. Histology showed tumor only in two of these cases and predominant tumor with necrosis in one case. In two patients, Ommaya reservoirs were implanted. Thirteen patients received further chemotherapy after HFSRT. Temozolomide was used in six patients, nimustine / teniposide in four and other multi-agent protocols in three. For the whole group of n = 19 patients with recurrent malignant glioma, the median overall survival from the time of HFSRT was 9.3 months (range 1.9 to 77.6+ months; Fig. [1](#F1){ref-type="fig"}). At the time of analysis, two patients were alive at 34.6 months and 77.6 months. One-year and two-year survival rates were 26% and 16%, respectively. Overall survival from time of HFSRT for subgroups is shown in Table [3](#T3){ref-type="table"}. WHO grading, both that determined at initial diagnosis of glioma and the most recent before HFSRT, had a significant impact on survival. Patients with a most recent histopathology of a grade III glioma had a median overall survival from the time of HFSRT of 15.4 (1.9 to 77.6) months, those with grade IV tumors of 7.9 (4.2 to 38.8) months (p = 0.029, log-rank test; Fig. [2](#F2){ref-type="fig"}). A trend toward a beneficial effect of higher total doses (30 Gy vs. \<30 Gy) on overall survival was observed (p = 0.051; Table [3](#T3){ref-type="table"}, Fig. [3](#F3){ref-type="fig"}). Median overall survival from the first histological diagnosis of glioma (of any grade) was 40.8 months (12.9 to 135 months). Respective median overall survival times from first diagnosis by initial histological grade were 134.7 (64.5 to 135) months for WHO grade II, 57 (30.8 to 119.4+) months for grade III and 26 (12.9 to 40.8) months for grade IV tumors. Information on follow-up imaging was available in 15 patients. Tumor progression was first diagnosed on MRI in 11 cases and on CT in four cases. The median interval between HFSRT and documented tumor progression was 4.9 months (1.3 to 37.3 months) in the overall group and 7.6 months (3.9 to 37.3 months) and 4.6 months (1.3 to 29.5 months) in the subgroups with a most recent histology of grade III astrocytoma and glioblastoma, respectively. Survival times from initial diagnosis for the two living patients were 49.7 and 119.4 months. The patient alive at 119.4 months (77.6 months from HFSRT) is a child first treated with HFSRT at the age of eleven for a fourth recurrence of an anaplastic astrocytoma after multiple resections, fractionated radiotherapy and intensive multi-agent chemotherapy. After further adjuvant chemotherapy, this patient recurred 37 months after HFSRT and has since been treated with another resection, a second series of HFSRT (49 months after the first HFSRT) to a volume adjacent to the initial HFSRT region (7 × 5 Gy, 80% isodose) and implantation of BCNU polymers. The most recent MRI in this patient showed no evidence of tumor recurrence. This patient was the only pediatric case included in the present analysis. The patient alive at 49.7 months from initial diagnosis (34.6 months from HFSRT) was first treated for anaplastic astrocytoma at 34 years with biopsy, involved-field radiotherapy and multi-agent chemotherapy. Twelve months after the first radiotherapy series, HFSRT and sequential temozolomide were given for tumor recurrence. Discussion ========== Local recurrence of malignant glioma pretreated with resection, post-operative radiotherapy and frequently with adjuvant chemotherapy is a common problem in clinical practice. Reoperation, re-irradiation and systemic or intratumoral chemotherapy are among the therapeutic options available in this situation. Only the implantation of BCNU polymers at the time of reoperation is supported by randomized trial data, having been shown to be superior to re-resection alone \[[@B11]\]. The benefit of other treatments needs to be evaluated based on phase II data and retrospective analyses which both are prone to bias by patient selection for salvage treatment. A recent large retrospective investigation analyzed the benefit of salvage treatment in patients with glioblastoma multiforme \[[@B12]\]. The authors found that the first, mostly chemotherapeutic, reintervention in this patient group was associated with a doubling in median overall survival from 26 to 61.5 weeks, although selection effects in this series can not be excluded. Prolongation of survival in recurrent malignant glioma has not been convincingly shown for either re-resection or re-irradiation by interstitial brachytherapy, single-dose radiosurgery or fractionated radiotherapy \[[@B13]\]. Hypofractionated stereotactic radiotherapy (HFSRT) has been proposed as a combination of high-precision treatment with small margins and maximum sparing of normal brain tissue, non-invasive technique and short treatment duration using single fraction doses of 3 Gy to 9 Gy (Table [4](#T4){ref-type="table"}). In the present single-institution series of patients with recurrent malignant glioma, median survival times from the time of HFSRT of 9.3 months (most recent histology grade III 15.4 months, grade IV 7.9 months) were achieved. In published series of HFSRT, applied either as a single modality or in combination with chemotherapy, median overall survival ranged between 7 and 12.7 months (Table [4](#T4){ref-type="table"}). Comparability between these series is limited by differences in patient selection and treatment concept. The median total dose of 30 Gy in 5-Gy fractions in the present series is close to the limit of 35 Gy which is thought to be applicable in pre-irradiated patients with acceptable toxicity \[[@B4]\]. However, a total dose of 20 Gy in 4-Gy or 5-Gy fractions, as used in about one third of patients, may be regarded as too low based on recent literature data \[[@B4]\]. Indeed, a trend toward longer survival with total doses of at least 30 Gy was observed (Table [3](#T3){ref-type="table"}, Fig. [1C](#F1){ref-type="fig"}). The finding that no patient in the present series had to be reoperated for symptomatic radiation necrosis also suggests that the dose schedules prescribed were rather conservative. Other potential prognostic factors such as age, Karnofsky performance score or tumor size (planning target volume) were not significant predictors of survival in the present series, possibly due to the limited patient number. It must be noted that in many patients HFSRT was one element of salvage therapy and further chemotherapy or surgical interventions were performed in 68% and 26% of patients, respectively. In the patients with available imaging information, the median time from HFSRT to tumor progression was approximately five months which may be regarded as an approximation of the lifetime gained by HFSRT in these highly selected patients. In the patient with the longest survival time of 77.6+ months after re-irradiation, a second HFSRT series was performed as well as multiple other treatments, highlighting the benefit of aggressive multi-modality treatment in individual patients. The longest survival time in glioblastoma was 38.8 months from HFSRT, indicating that even in grade IV tumors single patients may survive much longer than expected. Data from the present series and published reports suggest that efficacy and toxicity are favorable compared to single-dose radiosurgery and the invasive modality of interstitial brachytherapy. While radiosurgery resulted in median overall survival times of 26 to 50 weeks and reoperation rates of 0 to 22%, median overall survival of 47 weeks and reoperation in 41 to 44% (with about 5% of patients showing radionecrosis only) were reported for interstitial brachytherapy \[[@B13]\]. In a recent review of re-irradiation, HFSRT was favored as radiotherapy modality and the following criteria for the application of HFSRT were recommended \[[@B14]\]: good general status (WHO 0--1), at least one year disease-free interval, initial grade II or III histology and maximal tumor diameter 3 cm. While such criteria may identify patients with the greatest benefit from HFSRT, salvage treatment decisions in recurrent malignant glioma will remain highly individualized. Given the low toxicity of the method, HFSRT may be offered to patients in good general condition with tumor recurrences of limited volume. Conclusion ========== HFSRT with moderate total doses of no more than 30 Gy is safe and leads to similar OS times as more aggressive treatment schemes. In individual patients, HFSRT in combination with other salvage treatment modalities, can be associated with long-term survival. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= DV designed the analysis, reviewed patient data, performed statistical analysis and drafted the manuscript. OK treated the patients, reviewed patient data and revised the manuscript. KR treated the patients, reviewed patient data and revised the manuscript. GHV treated the patients, reviewed patient data and revised the manuscript. KB performed radiotherapy planning for the patients analyzed, reviewed radiotherapy details and revised the manuscript. MF treated the patients, reviewed patient data and revised the manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2407/5/55/prepub> Figures and Tables ================== ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Overall survival of n = 19 patients with recurrent malignant glioma treated with hypofractionated stereotactic radiotherapy (HFSRT), from the time of re-irradiation. ::: ![](1471-2407-5-55-1) ::: ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Overall survival from the time of re-irradiation of n = 19 patients with recurrent malignant glioma treated with hypofractionated stereotactic radiotherapy (HFSRT), by WHO grade (most recent histopathology before HFSRT). ::: ![](1471-2407-5-55-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Overall survival from the time of re-irradiation of n = 19 patients with recurrent malignant glioma treated with hypofractionated stereotactic radiotherapy (HFSRT), by total dose. ::: ![](1471-2407-5-55-3) ::: ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Clinical characteristics of n = 19 consecutive patients treated with hypofractionated stereotactic radiotherapy (HFSRT) for recurrent malignant glioma. ::: ------------------------------------------------------------- ----------------- ----------------- median age (range) at time of HFSRT 50 years (11--74 years) male / female 8 (42%)/11(58%) median Karnofsky performance score at time of HFSRT (range) 90 (60--90) initial histology astrocytoma II° 3 (16%) pure anaplastic astrocytoma III° 4 (21%) anaplastic oligoastrocytoma or oligodendroglioma III° 3 (16%) glioblastoma multiforme IV° 9 (47%) most recent histology before HFSRT anaplastic astrocytoma III° 5 (26%) glioblastoma multiforme IV° 14 (74%) initial surgical procedure resection 15 (79%) biopsy 4 (21%) post-operative radiotherapy 1.8--2.0 Gy daily (total dose 54 to 61 Gy) 12 (63%) 3 Gy daily (total dose 45 Gy) 2 (11%) 2 × 1.8 Gy daily (total dose 54 Gy) 5 (26%) salvage surgery before HFSRT 12 (63%) adjuvant or salvage chemotherapy before HFSRT nimustine / teniposide 12 (63%) temozolomide 5 (26%) other 1 (5%) median interval post-OP radiotherapy to HFSRT (range) 19 months (3--116 months) ------------------------------------------------------------- ----------------- ----------------- ::: ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Treatment details of hypofractionated stereotactic radiotherapy (HFSRT) in n = 19 consecutive patients with recurrent malignant glioma. ::: --------------------------------------- ---------- ------------- Median planning target volume (range) 15 ml (4--70 ml) Median prescription isodose (range) 80% (70--90%) Median single dose (range) 5 Gy (4--10 Gy) Median total dose (range) 30 Gy (20--30 Gy) Fractionation schedules 5 × 4 Gy 2 (11%) 4 × 5 Gy 5 (26%) 6 × 5 Gy 6 (32%) 5 × 6 Gy 5 (26%) 2 × 10 Gy 1 (5%) HFSRT technique multiple non-coplanar arcs 5 (26%) multiple non-coplanar fixed fields 11 (58%) combination arcs / fixed fields 3 (16%) median field no. (range) 5 (3--8) --------------------------------------- ---------- ------------- ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Median overall survival from the time of hypofractionated stereotactic radiotherapy (HFSRT) in subgroups of patients with recurrent malignant glioma. ::: n median (range) overall survival (months) p ---------------------------------------------- ---- ------------------------------------------ ------- WHO grade (initial histopathology) II or III 10 13.5 (1.9--77.6) 0.036 IV 9 7.4 (4.2--10.8) WHO grade (most recent before HFSRT) III 5 15.4 (1.9--77.6) 0.029 IV 14 7.9 (4.2--38.8) initial surgical procedure biopsy 15 9.3 (4.2--77.6) 0.89 resection 4 9.2 (1.9--34.6) chemotherapy before HFSRT yes 13 9.3 (1.9--77.6) 0.76 no 6 8.9 (7.3--15.4) reoperation before HFSRT yes 12 8.4 (1.9--77.6) 0.93 no 7 9.3 (5.6--34.6) age at time of HFSRT \< 50 years 11 11.9 (1.9--77.6) 0.14 \> 50 years 8 8.9 (5.0--11.1) Karnofsky performance score at time of HFSRT ≥ 90 14 7.9 (4.2--77.6) 0.80 \< 90 5 11.9 (1.9--15.4) planning target volume (ml) \< 15 ml 9 9.3 (5.2--38.8) 0.59 \> 15 ml 10 7.4 (1.9--77.6) interval first RT series to HFSRT \> 20 months 9 7.4 (1.9--77.6) 0.90 \< 20 months 10 9.3 (5.6--38.8) total HFSRT dose 30 Gy 11 11.1 (1.9--77.6) 0.051 \< 30 Gy 8 7.4 (4.2--11.9) chemotherapy after HFSRT yes 13 9.3 (1.9--77.6) 0.27 no 6 8.9 (4.2--11.1) ::: ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Overview of published results on hypofractionated stereotactic radiotherapy (HFSRT) in patients with recurrent malignant glioma (RT: radiotherapy). ::: authors patient number single dose (Gy) total dose (Gy) median tumor volume (ml) chemo-therapy overall survival (median) surgery for toxicity prognostic factors comment ----------------------------- ----------------------------------- ----------------------- --------------------- ------------------------------ --------------------------------------- ------------------------------- ------------------------------------------- ----------------------------------------------------------------- ------------------------------------------------------------------------------------------------------------ Shepherd et al., 1997 \[4\] 29 (\"high-grade astrocy-toma\") 5 (daily) 20--50 24 (3--93) \-- 10.7 months 6% initial low-grade histology associated with longer survival 36% steroid dependent toxicity (increased risk \>40 Gy) Glass et al., 1997 \[6\] 20 (7 grade III, 13 grade IV) 3.5--6 (twice / week) 35--42 14 (2--122) cisplatin 40 mg/m^2^weekly 12.7 months 15% pre-dominant necrosis on re-operation \-- 40% treated within 10 weeks of first RT series for (\"potential\") progression Hudes et al., 1999 \[5\] 19 (glio-blastoma) 3--3.5 (daily) 24--35 13 (1--48) \-- 10.5 months 0% \"minor response\" on imaging associated with ≥30 Gy and ≤20 ml recurrent or persistent tumors treated, median interval between completion of RT and HFSRT only 3.1 months Lederman et al., 2000 \[7\] 88 (glio-blastoma) 4--9 (weekly) 18--36 32.7 (2--150) paclitaxel 120 mg/m^2^(median) weekly 7 months 8% necrosis only at re-operation volume ≥30 ml associated with longer survival median time from diagnosis to HFSRT only 6.5 months Voynov et al., 2002 \[8\] 10 (5 WHO grade III, 5 grade IV) 5 30 34.7 \-- 10.1 months ? \-- stereotactic intensity-modulated radiotherapy (IMRT) used present series 19 (5 WHO grade III, 14 grade IV) 4--10 (daily) 20--30 15 (4--70) \-- 9.3 months 0% WHO grade IV and \<30 Gy associated with short survival median interval post-OP RT to HFSRT 19 months :::	PubMed Central	2024-06-05T03:55:57.628726	2005-5-30	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1156875/", "journal": "BMC Cancer. 2005 May 30; 5:55", "authors": [ { "first": "Dirk", "last": "Vordermark" }, { "first": "Oliver", "last": "Kölbl" }, { "first": "Klemens", "last": "Ruprecht" }, { "first": "Giles H", "last": "Vince" }, { "first": "Klaus", "last": "Bratengeier" }, { "first": "Michael", "last": "Flentje" } ] }
PMC1156876	Background ========== A chronic, low-grade inflammatory response plays a key role in atherosclerosis \[[@B1]\]. It has been suggested that chronic inflammatory stimuli originating from infectious agents may contribute to this low-grade inflammation and thus, might play a causal role in atherogenesis and disease progression in patients with manifest chronic coronary artery disease (CAD) \[[@B2]\]. The cytomegalovirus (CMV) infection has been discussed as potential culprit to cause atherosclerosis and especially to be involved in restenosis \[[@B3]\] as it, if once acquired, persists life long and may undergo periodic reactivation from latency \[[@B4]\]. Genes involved in the inflammatory response, such as the interleukin-1 receptor gene, which is polymorphic in nature, might be important in determining the response profile in patients with chronic inflammatory stimuli \[[@B5]\]. This gene is located on chromosome 2 in close prximity to genes coding for IL-1α and IL-1β. The main role of the IL-1 system is to mediate the early inflammatory reactions for protection against many different stimuli ranging from microbial colonization, infections, to malignant transformation. IL-1RA levels typically increases during the course of an inflammatory event so that an induced inflammation gets terminated. As persons homozygous for the allele 2 of the interleukin 1 receptor antagonist (IL-1RA) gene (Il-1RN\2) show a more prolonged and more severe immune response compared to persons with other allele constellations, subjects with the Il-1RN\2 allele and with evidence of chronic infection might be at particular high risk for cardiovascular events. However, data examining the combination of both factors are lacking. We investigated whether homozygousity for the allele 2 of the IL-1RA and CMV-seropositivity might be associated with an increased risk for the development of cardiovascular events in patients with prevalent CAD to determine its prognostic value in secondary disease prevention in a prospective cohort study and whether this constellation was associated with various inflammatory response markers. Methods ======= Study design and population --------------------------- ### Baseline examination Patients were recruited between October 1996 and November 1997 and were part of a case-control study investigating the role of infectious agents in primary CAD; details were reported elsewhere \[[@B6]\]. Briefly, the patient group consisted of 312 patients aged 40--68 years with clinically stable CAD who underwent elective coronary angiography in the Department of Cardiology at the University of Ulm Medical Centre during this period and who had a coronary stenosis of ≥ 50% of the luminal diameter of at least one major coronary artery. Patients with diagnosis of CAD older than 2 years, patients with acute coronary syndromes, and patients on anticoagulant therapy within the previous four weeks were excluded from the study. At baseline, all study participants underwent a standardized interview carried out by specially trained interviewers. Participation was voluntary and written informed consent was obtained from each subject upon entry into the study. The study was approved by the ethics committee of the University of Ulm. ### Follow-up examination A follow-up (FU) of all patients with CAD was conducted between October 2000 and April 2001. A personal interview was conducted in the medical clinic (96.5%) or -- if patients were not willing or able to follow the invitation -- by phone by the same trained medical staff. Cardiovascular events (CVE) were defined by assessing the first occurrence of the following events during FU: cardiovascular death, nonfatal myocardial infarction, ischemic cerebrovascular event, and the need for coronary revascularization. All CVEs were validated by chart review. Laboratory methods ------------------ Venous blood was drawn under standardized conditions directly before diagnostic coronary angiography at the baseline examination. A complete blood cell count was done automatically by a Coulter STKS counter (Coulter, Krefeld, Germany). The remaining blood was centrifuged at 3,000 g for 10 minutes within 30 min after venipuncture, immediately aliquoted and frozen at -70°C until further analysis. IgG antibodies against CMV were determined using a commercially available ELISA (CMV-IgG-ELISA PKS, medac, Wedel Germany) for the detection and quantitative determination of human IgG to cytomegalovirus according to the manufacturer\'s instructions (borderline zone 0.35--0.45 U/ml, only n = 3 (1%) of the samples had a borderline result). IL-1RA gene polymorphisms were done by PCR as described in a similar study (7). Subjects homozygous for allele 2 of the IL-1RA gene (IL-1RN\2) were grouped against all other alleles constellations in the statistical analysis. Additionally, the following markers of inflammation were determined by ELISA: Interleukin-6 (IL-6), and Tumor Necrosis Factor (TNF)-α (Quantikine, R&;D Systems, Wiesbaden, Germany), inter-cellular adhesion molecule (ICAM)-1 (Diaclone, Besancon, France). In addition, C- reactive protein (CRP) determinations were done by an immunoradiometric assay (range 0.05--10 mg/L) calibrated with the WHO reference standard 85/506. Serum amyloid A (SAA) was also determined by immunonephelometry (Dade Behring, Marburg, Germany). All laboratory analyses were done in a blinded fashion. Statistical analysis -------------------- Baseline demographic and clinical characteristics of cases were compared in a descriptive way. The association of anti-CMV IgG antibody titer (positive vs negative or borderline \[1%\]) and of the IL-1RA gene polymorphisms (IL-1RN\2 vs. others) with the occurrence of cardiovascular events during follow-up was analyzed by Chi-square (χ^2^) statistics. If expected cell frequencies were \< 5, Fisher\'s Exact Test was used. In addition the relation of CMV, of the IL-1RA gene polymorphisms and their combination with CVD events during follow-up was assessed by the Kaplan-Meier method and quantified by means of the log-rank test. Multivariate Cox regression analysis was performed to determine the hazard ratio (HR) and 95% confidence intervals (CI) for future cardiovascular events taking potential confounding factors into account (controlling for age (years), gender, body mass index (BMI, kg/m^2^), school education, cigarette smoking, alcohol consumption, history of Myocardial infarction, history of hypertension, history of diabetes, statin intake, intake of aspirin, and intake of diuretics). The proportional hazard assumption were checked graphically. A general linear regression method was employed to calculate gender and age adjusted mean values (arithmetic, if skewed geometric) associated with CMV-seropositivity and the IL-1RN\2 allele compared to all other constellations. All analyses were carried out with the SAS statistical software package (SAS Institute, Version 8, Cary, North Carolina: SAS Institute, Inc). Results ======= In this prospective study a total of 300 (96.2%) out of 312 patients with a mean age of 57.9 years were followed for a median of 38.2 months (maximum 53.8 months). Twelve subjects could not be included in the follow-up as they refused to participate or they moved out of Germany and could not be contacted during follow-up. During the follow-up, 11 fatal and 49 non-fatal CVE occurred (20%) (5 patients died of a non-cardiac cause). Among CAD patients, who subsequently developed a non-fatal CVE, four patients suffered a myocardial infarction, seven an ischemic cerebrovascular event, and coronary revascularization was performed in 38 subjects. Table [one](#T1){ref-type="table"} shows the main characteristics of the patients in patients with (n = 60) and without (n = 240) a CVE during follow-up. There were no statistically significant differences with respect to gender, age, mean body mass index, alcohol consumption habits and smoking status, and history of dyslipidemia, hypertension, diabetes and prior myocardial infarction. IN addition, coronary status and history of percutaneous coronary intervention (PCI) or coronary artery bypass graft (CABG) at baseline was also similar among the two groups. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### Baseline Characteristics of the Study Population (N = 300) ::: Patients With Cardiovascular Event (N = 60)* Patients Without Cardiovascular Event (N = 240) p-value ---------------------------------------- ------------------------------------------------- ----------------------------------------------------- ------------- Male, n (%) 51 (85) 206 (85.8) 0.9 Age(y) mean ± SD 57.5 ± 7.7 58.0 ± 7.2 0.7 Body mass index(kg/m^2^) mean ± SD 27.0 ± 5.7 27.2 ± 3.5 0.6 School education\<10 yrs, n (%) 46 (76.7) 165 (68.7) 0.8 Daily alcohol consumption, n (%) 18 (30) 71 (29.6) 0.9 Smoking status: Current, n (%) 1 (1.7) 25 (10.4) Past, n (%) 41 (68.3) 158 (65.8) Never, n (%) 18 (30) 57 (23.8) 0.07^a^ History of: Dyslipidemia, n (%) 37 (61.7) 167 (69.6) 0.2 Hypertension, n (%) 36 (60) 138 (57.5) 0.7 Diabetes, n (%) 9 (15) 32 (13.3) 0.7 Prior myocardial infarction, n (%) 40 (66.7) 144 (60.3) 0.4 ^a^= Fisher\'s Exact Test used ::: Table [two](#T1){ref-type="table"} shows the prevalence of the IL-1RN\2 allele, and the seroprevalence of CMV infection (results of the quantitative CMV analysis were not available in 5 patients). 8.3% and 10.4% (p = 0.6) of the patients with and without CVE respectively, were homozygous for the IL-1RN\2 allele (known to be associated with a more severe and prolonged inflammatory response). ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Prevalence of interleukin-1 receptor antagonist gene polymorphisms (IL-1RA) and serostatus of CMV and their combination in patients with and without cardiovascular events during follow-up ::: Patients with Cardiovascular Event (n = 60) Patients without Cardiovascular Event (n = 240) p-value -------------------------------- ------------------------------------------------- ----------------------------------------------------- ------------- IL-1 RA \- other^a^ 55 (91.7%) 215 (89.6%) \- Il-1RN\2 5 (8.3%) 25 (10.4%) 0.6 CMV-serostatusb* \- negative 29 (49.2%) 110 (46.8%) \- positive 30 (50.9%) 125 (53.2%) 0.7 IL-1 RA and CMV serostatus \- IL-other^a^and CMV-negative 27 (45.8%) 96 (40.9%) \- IL-other^a^and CMV-positive 27 (45.8%) 114 (48.5%) \- Il-1RN\2 and CMV-negative 2 (3.4%) 14 (6.0%) \- Il-1RN\2 and CMV-positive 3 (5.1%) 11 (4.7%) 0.8^c^ ^a^= all other alleles summarized (IL-1RN\4, IL-1RN\5, IL-1RN\3) except IL-RN\2 ^b^= in 6 patients quantitative determination of IgG was not available ^c^= Fisher exact test ::: Overall, 52.7% of the patients were seropositive for CMV, and the distribution was similar in patients with and without CVE during follow-up (50.9% and 53.2% respectively, p = 0.7). There was also no difference in distribution among the two groups if IL-1RA genotype and CMV-seropositivity were combined: the IL-1RN\2 allele and CMV-seropositivity occurred in 5.1% of patients with CVE and in 4.7% of patients without, respectively (p = 0.8). Table [three](#T1){ref-type="table"} shows the results of the multivariate analysis. Patients who had the IL-1RN\2 allele and patients seropositive for CMV (the latter independently also from titer values above the mean) showed no statistically significant increased risk for subsequent fatal or non-fatal CVE, in both, the partially as well as in the fully adjusted model. ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### Distribution of factors and relation to cardiovascular events during follow-up and partly and fully adjusted hazard ratios for cardiovascular events associated with Interleukin-1 receptor antagonist gene polymorphisms (IL-1RA) and serostatus of CMV ::: Factor CVD-events during follow-up,row % (^a^p-value) ^b^Partly-adjusted hazard ratio(95% confidence interval) ^c^Fully-adjusted hazard ratio(95% confidence interval) -------------------------------- ---------------------------------------------------- -------------------------------------------------------------- ------------------------------------------------------------- IL-1 RA \- other^d^ 20.4% 1^reference^ 1^reference^ \- Il-1RN\2 16.3% (0.73) 1.1 (0.80--1.66) 1.03 (0.60--1.76) CMV-serostatus* \- negative 20.7% 1^reference^ 1^reference^ \- positive 0.91 (0.55--1.52) 1.00 (0.59--1.72) - positive, titre \< median 16.7% 0.74 (0.38--1.43) 0.78 (0.40--1.54) - positive titre ≥ median 22.1%(0.54) 1.12 (0.61--2.05) 1.32 (0.69--2.52) IL-1 RA and CMV serostatus \- IL-other^d^and CMV-negative 21.9% 1^reference^ 1^reference^ \- IL-other^d^and CMV-positive 19.1% 0.86 (0.50--1.45) 0.94 (0.54--1.64) \- Il-1RN\2 and CMV-negative 12.5% 0.56 (0.13--2.36) 0.37 (0.08 -- 1.67) \- Il-1RN\2 and CMV-positive 21.4% (0.83) 0.96 (0.32--3.43) 0.95 (0.27 -- 3.38) ^a^= according to log-rank test ^b^= adjusted for age and gender ^c^= adjusted for age, gender, body mass index, school education, cigarette smoking, alcohol consumption, history of myocardial infarction, history of hypertension, history of diabetes, statin intake, intake of aspirin, intake of diuretics ^d^= all other alleles ::: In addition, if both factors were combined no increased risk for CVE during follow-up was obtained (HR = 0.96 (95% CI 0.32--3.43) in the gender and age adjusted model, and HR = 0.095 (95% CI 0.27--3.38) in the fully adjusted model compared to CMV seronegative patients not having the IL-1RN\2 allele). In addition, we found no statistically significant differences for mean values (geometric means with exception of ICAM-1 (arithmetic)) of CRP, SAA, Il-6, TNF-α, and ICAM-1 after adjustment for age and gender when CMV-seropositive patients with the IL-1RN\2 allele were compared to others (table [4](#T4){ref-type="table"}). ::: {#T4 .table-wrap} Table 4 ::: {.caption} ###### Mean concentrations^a,b^of various markers of inflammation in patients with Il-1RN\2 and positive CMV-serostatus compared to the others ::: IL-1 RA and CMV serostatus* ------------------------- -------------------------------- ------- ------ \- CRP \[mg/L\] ^c^ 0.92 1.58 0.5 \- SAA \[mg/L\] ^c^ 3.04 3.52 0.5 \- Il-6 \[pg/mL\] ^c^ 2.19 2.43 0.6 \- TNF-α \[pg/mL \] ^c^ 2.37 2.53 0.6 \- ICAM-1 \[ng/mL\] 456.6 537.1 0.08 ^a^= Adjusted for age and gender by general linear regression ^b^= Arithmetic or if skewed ^c^geometric means ::: Discussion ========== In this prospective cohort study involving 300 patients with angiographically defined CAD at baseline, homozygousity for allele 2 of the IL-1 RA and seropositivity to CMV alone and in combination were not associated with an increased risk for cardiovascular events during follow-up; in addition, combination of the CMV-seropositivity and IL-1RN\2 allele were not associated with a proinflammatory response. Therefore, these data do not support the hypothesis that IL-1RN\2 genotype in combination with CMV seropositivity might be a strong risk factor for secondary cardiovascular events in patients with already prevalent cardiovascular disease. These results are in contrast to several reports in the literature suggesting a positive association between CMV seropositivity and progression of atherosclerosis \[[@B8]\]. However, the earlier evidence was mostly based on small case-control studies with rather vague definitions of clinical endpoints and with little or no adjustment for potential confounders. Meanwhile, several large prospective seroepidemiological studies have shown no independent association for CMV and coronary heart disease \[[@B9],[@B10]\]. But clearly more carefully conducted prospective studies in different patient populations are needed before definite conclusions can be drawn \[[@B11],[@B12]\]. Evidence for a possible influence of the IL-1 RA polymorphism on CAD has been inconclusive so far \[[@B5]\]. One study reported a positive association with risk of CAD \[[@B13]\], another reported an positive association for the IL1RN\2 genotype with risk of restenosis in patients with CAD, if the population was restricted to a subgroup of patients with single vessel disease \[[@B7]\]. We found no association of CMV-serpositivity and IL1RN\2 allele with risk of CVE during follow-up, even if both factors were combined. It has been described that subjects with the IL1RN\2 genotype might be more resistant against some infections \[[@B5]\]. Patients with the IL1RN\2 genotype indeed showed a lower CMV-seroprevalence than others in our population (55.2% vs. 46.6%), although this difference was not statistically significant. Recently, it has been suggested that the detrimental effects of CMV-seroprevalence may be limited to subjects with an increased inflammatory response as characterized by high levels of CRP \[[@B14],[@B15]\] or high IL-6 levels \[[@B16]\]. We did not find a positive association of CMV-seroprevalence with the occurrence of CVE confined to subjects with high levels (above the median) of IL-6 or CRP, although baseline IL-6 and CRP values (the latter only tentatively) showed a positive predictive association with the occurrence of cardiovascular events during follow-up. However, this association was not modified by CMV-serostatus (data not shown). As CMV-seroprevalence may be related to various adverse factors associated with CAD itself, issues of confounding have to be considered and may explain part of the discrepancies in the literature so far. There may be indeed a complicated interplay among established risk factors and exposure to infectious agents as recently suggested by our group \[[@B17]\]. When looking at the results of this study the following limitations should be considered: seroprevalence may not be a good marker for recurrent active infection with CMV, or even more relevant in this context, reactivation of CMV; cytomegalovirus-IgG does not diagnose active virus infection after primary infection. Furthermore, the current study cannot exclude a weak association between seroprevalence to CMV and IL-1RN\2 genotype and risk of CVE. However, the study had a power of 80 % to detect an RR of at least 1.6 and more associated with CMV-seropositivity alone, and an RR of 2.8 and more associated with combined CMV-seropositivity and presence of the IL-1RN\2 allele, respectively. In addition, lack of association of the CMV-seropositivity and IL-1RN\2 allele with established inflammatory marker levels, which play a key role in atherogenesis, are further reassuring the essentially negative findings of this study. Conclusion ========== Despite its limitations, our study suggests that seropositivity to CMV and IL-1RA\2 genotype alone or in combination might not be a strong risk factor for recurrent cardiovascular events in patients with manifest CAD, and is not associated with levels of established inflammatory markers. List of abbreviations ===================== CABG = coronary artery bypass graft CAD = coronary artery disease CMV = Cytomegalovirus CRP = C-reactive protein CVE = cardiovascular events FU = follow-up ICAM-1 = intercellular adhesion molecule -- 1 Il-6 = interleukin 6 IL-1RA = interleukin 1 receptor antagonist OR = odds ratio PCI = percutaneous coronary intervention SAA = serum-amyloid- A TNF-α = tumor-necrosis factors \- Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= DR did the statistical analysis and wrote the initial draft. DR, HB, TM, AH, WK had the idea of the study and DR, HB, AH, WK did the study design and conduct. TM did the immunoassays. MMH carried out the molecular genetic part. All authors critically revised the MS and read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2261/5/10/prepub>	PubMed Central	2024-06-05T03:55:57.631555	2005-5-20	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1156876/", "journal": "BMC Cardiovasc Disord. 2005 May 20; 5:10", "authors": [ { "first": "Dietrich", "last": "Rothenbacher" }, { "first": "Hermann", "last": "Brenner" }, { "first": "Thomas", "last": "Mertens" }, { "first": "Michael M", "last": "Hoffmann" }, { "first": "Albrecht", "last": "Hoffmeister" }, { "first": "Wolfgang", "last": "Koenig" } ] }
PMC1156877	Background ========== Stent coating with immunosuppressive or cytostatic agents are valid advances in the struggle against restenosis following coronary intervention. However these therapies are hampered by high costs, especially in the case of multivessel disease. Moreover it is not entirely clear whether restenosis is merely delayed and not inhibited \[\[[@B1]\], review\]. Consequently the intense search for a systemic approach to inhibit restenosis is required. Restenosis is essentially characterized by migration and proliferation of smooth muscle cells and extracellular matrix accumulation. However, there is now increasing evidence for a role of inflammation in the development of restenosis. The underlying molecular mechanisms of restenosis are, in fact, most probably regulated by inflammatory mediators, such as cytokines \[\[[@B2]\], review\]. The situation resembles to a certain degree to the activation of the immune system during organ rejection. Therefore it is not surprising that immunosuppressive agents are potential candidates in the treatment of restenosis. Mycophenolate mofetil (MMF), the prodrug of mycophenolic acid (MPA), is a rationally designed immunosuppressive drug. The active metabolite MPA is a selective, non-competitive and reversible inhibitor of inosine monophosphate dehydrogenase (IMPDH) and of the type II isoform in particular \[[@B3]\]. The primary mechanism of action of MPA is presumed to be anti-lymphoproliferative, a result of inhibition of inosine 5\'-monophosphate dehydrogenase (IMPDH), which is required for the de novo synthesis of guanosine nucleotides which are necessary for DNA and RNA synthesis and for lymphocytes to proliferate maximally after stimulation \[[@B4]\]. Antiproliferative effects of MMF have been described in non-immune cells \[\[[@B5]\], review\]. In the past data of animal studies could not be transferred easily to the clinical situation due to species differences. In the current study a cascade of human in vitro models in combination with a porcine coronary ex vivo model is applied to investigate the effects of MMF on key pattern of restenosis. Unexpected or conflicting data can be analysed before large experimental studies are initiated. Furthermore attention is focused on the relation between significant inhibitory effects in vitro (SI) and maximal plasma levels in vivo (MPL), the SI/MPL-ratio \[[@B6]\]. Central part of the current study is a 3D human coronary transfilter co-culture model of leukocyte attack \[[@B7]\]. In this model the effect of MMF on monocyte (MC) adhesion and chemotaxis and reactive cell proliferation of co-cultured smooth muscle cells (SMC) are investigated. In order to obtain information on the effect of MMF on TNF-alpha induced expression of adhesion molecules, the effect on expression of ICAM-1 is studied in northern blot and cytoflow studies. Finally MMF is added in a porcine coronary organ culture model of restenosis for 1, 2, 3, 4, 5, 6, and 7 days and the effect on reactive cell proliferation and neointimal hyperplasia is investigated. Methods ======= Cell culture ------------ Human coronary endothelial cells (HCAEC) and human coronary smooth muscle cells (HCMSMC) were purchased at Cambrex Bio Science (Vervier, B). HCAEC were cultured in Endothelium Growth Medium (Cambrex) and identified by the typical \"cobble stone\" growth pattern and positive reaction against von Willebrand factor (Dakopatts). HCMSMC were grown in Smooth Muscle Cell Growth Medium (Cambrex). For identification of HCMSMC antibodies against smooth muscle α-actin (Renner, Darmstadt, D) were used. Human MC were isolated from the residual leukocytes of single donors using MACS cell-isolation kit (Milteny Biotec GmbH). Mycophenolate Mofetil --------------------- Mycophenolate mofetil (MMF): Cellcept^®^, Roche, Basel, CH, 0.005 -- 500 μg/mL, dilution: aqua ad inject., MPL: 34 μg/mL \[[@B8]\]. RNA extraction and Northern blot analysis ----------------------------------------- For Northern blot studies of the effect of MMF/TNF-α treatment on expression of ICAM-1, monocultures of HCAECs and HCMSMCs were incubated with MMF (50, 100, 150, 200, 250, and 300 μg/mL) for a period of 18 hrs. During the last 6 hrs of MMF incubation, expression of adhesion molecules was stimulated by adding of TNF-α (20 ng/mL). Total RNA (3 × 10^6^cells) was isolated with RNEasy Mini Kit (Qiagen), and 10 μg of RNA was used in standard Northern blot analysis with an ICAM-1 probe. A non radioactive labelling and detection system (Amersham Biosciences Europe GmbH, Freiburg, D) was used to detect the relative band density of ICAM-1 mRNA in comparison with TNF-α-stimulated cells. GAPDH was used as a control. Experiments were performed in triplicate. Flow cytometry -------------- For flow cytometry analysis of the expression of ICAM-1 in HCAEC and HCMSMC cells were trypsinized and seeded into 6-well dishes (5 × 10^4^cells). MMF (50, 100, 150, 200, 250, and 300 μg/mL) was added to the cultures for a period of 18 hrs. During the last 6 hrs of MMF incubation, the expression of adhesion molecules was stimulated by adding of TNF-α (20 ng/mL). After MMF/TNF-α treatment, cells were washed twice with phosphate-buffered saline (pH 7.2) and trypsinized. Cells were resuspended in 100 μL of a FITC-conjugated monoclonal antibody directed against ICAM-1 (clone 84H10, Dianova Immunotech; final concentration 10 μg/mL) and incubated for 20 min at 4°C. A total of 1 × 10^4^cells (100% gated) were analyzed immediately with a flowcytometer (BDFACsCalibur, Durchflußzytometer Becton Dickinson, Heidelberg, D). The effects of MMF (50 μg/mL -- 300 μg/mL) on vitality of HCAEC and HCMSMC were analyzed with propidium iodide (Sigma-Aldrich, Taufkirchen, D). The 3DLA model -------------- Three-dimensional human coronary units of leukocyte attack (3DLA units) mimic the inner layers of human coronary arteries \[[@B7]\]. The internal elastic membrane is represented by a polycarbonate filter with a thickness of 10 μm and a pore size of 5 μm (Whatman, Göttingen, D). Filters were fixed in a specially designed frame and inserted in a siliconized culture dish. On both sides of the filters cell cultures were established, direct contact of the cultures was made possible through the pores of the filter. HCMSMC were seeded on one side of the filter at a density of 2.5 × 10^4^cells / cm^2^. After 24 h cells had attached to the surface and frame and filters were turned upside down. HCAEC were seeded on the opposite side of the filter at a density of 2.5 × 10^4^cells / cm^2^. Both HCAEC- and HCMSMC-cultures were supplied with the appropriate culture medium and cultured for 14 days. At day 14, 3DLA-units were incubated with MMF (50 μg/mL) for 18 h. During the last 6 h of MMF incubation, the models were treated with TNF-α (20 ng/mL). For leukocyte attack, the required numbers of MC was calculated in relation to the relative concentration of MC in the full human blood \[[@B7]\]. 3 × 10^5^MC were seeded on the endothelial side of the 3DLA-units. The effect of MMF on MC-adhesion and MC-chemotaxis was studied at 30 min, 1, 2, 3, 4, 6, and 24 h after leukocyte attack; the effect on proliferation of co-cultured HCMSMC was investigated 24 h after leukocyte attack. Controls were performed without MMF-treatment. \[for detailed information: \[[@B7]\]\]. The porcine coronary organ culture model ---------------------------------------- Fresh hearts of 12 pigs, ranging in age from 3 to 5 months, weighing 100 to 120 kg, were obtained from a local slaughterhouse. In the laboratory, the left anterior descending coronary artery (LAD) was carefully prepared. Section were made at 4 mm intervals perpendicular to the vessel wall axis \[[@B9]\]. ### Ex vivo ballooning and adding of MMF For ex vivo angioplasty the prepared LAD segments were placed over a 3 mm balloon catheter (Medtronic 14K2030E, Medtronic, Kerkrade, North Carolina, USA) and were treated with 9 bar for a period of 60 s. After ex vivo ballooning MMF (50 μg/mL) was added to the cultures for a period of 1, 2, 3, 4, 5, 6, and 7 days. At each medium exchange the drug was renewed. ### Cultivation and fixation of coronary organ cultures After ballooning the segments were transferred to six-well plates (Tecnomara, Fernwald, D) and cultured in a mixture of Waymouth\'s MB 752/1 and Ham F12 nutrient mixture (1:1; vol/vol; Cambrex) supplemented with 15% fetal calf serum (Cambrex) at 37°C in 5% carbon dioxide. Organ cultures were cultured for 7 and 28 days, culture medium was exchanged every second or third day. Culture conditions for control groups were exactly the same as described for the angioplasty/MMF group. ### Analysis of reactive cell proliferation and neointimal thickening The effect of MMF (50 μg/mL) on reactive cell proliferation and neointimal thickening was studied at day 7 and day 28 after ballooning, controls were performed with and without ballooning \[for detailed information: \[[@B9]\]\]. Statistical analysis -------------------- Data of northern blot and flow cytometry studies were presented as mean ± S.D. Statistical significance of differences between controls and drug-treated cells was determined by paired Student\'s t-test. The Mann-Whitney rank-sum test was used to investigate the significance of differences in the 3DLA-model and the organ culture model. Statistical significance was accepted for P\< 0.05. Results ======= Identification of cells ----------------------- In monocultures of HCAECs cells were identified by a positive reaction with antibodies directed against von Willebrand factor and by the typical \"cobblestone\" growth pattern in culture. Monocultures of HCMSMC exhibited the \"hill and valley\" growth pattern and reacted positively with antibodies against smooth muscle α-actin. Effect of MMF on ICAM-1 mRNA levels: Northern blot studies ---------------------------------------------------------- After TNF-α stimulus, band density of mRNA ICAM-1 in HCAEC was increased 30-fold, which corresponds to a relative band density of 100% (Fig. [1](#F1){ref-type="fig"}). After incubation of HCAEC with MMF in concentrations of 50, 100, 150, 200, and 250 μg/ml expression of ICAM-1 was further increased by 23.92% (p = 0.01), 29.51% (n.s.), 75.87% (n.s.), 79% (n.s.), and 24.34% (n.s.). MMF in a concentration of 300 μg/mL caused an inhibition of ICAM-1 expression by 40.57% (n.s.). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Northern blots and relative band densities of TNF-α-induced expression of ICAM-1 mRNA after incubation of HCAECs and HCMSMCs with 50, 100, 150, 200, 250, and 300 μg/mL of MMF. ::: ![](1471-2261-5-9-1) ::: In HCMSMC, band density of mRNA ICAM-1 was increased 7.5-fold after TNF-α-stimulus, relative band density was increased from 13.4% to 100% (n.s.). Incubation with MMF in the concentration of 50 μg/mL caused a slight inhibitory effect on band density of mRNA ICAM-1 by 12.02% (n.s.). After incubation of HCMSMC with MMF in concentrations of 100, 150, 200, 250, and 300 μg/mL relative band density of ICAM-1 was increased by 59.9% (n.s.), 70.61% (n.s.), 68.7% (n.s.), 67.8% (n.s.), and 69.4% (n.s.). Both in HCAEC and HCMSMC, expression of GAPDH after adding of MMF in concentrations of 50, 100, 150, 200, and 250 μg/ml was identical with untreated controls. Effect of MMF on ICAM-1: Flow cytometry studies ----------------------------------------------- The effects of MMF (50, 100, 150, 200, 250, and 300 μg/mL) on the TNF-α induced expression of ICAM-1 are demonstrated in Figure [2](#F2){ref-type="fig"}. A dose dependent significant inhibition of ICAM-1 expression was detected in HCAEC. No significant effect was seen in HCMSMC. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Graphics and histograms (cytoflow data) of the effect of MMF (50, 100, 150, 200, 250, and 300 μg/mL) on TNF-α-induced expression of ICAM-1 in HCAECs (A, B) and HCMSMCs (C, D) after 18 h. Negative control (dotted line) and ICAM-1 expression in untreated cells (grey line) are included. ::: ![](1471-2261-5-9-2) ::: In HCAEC, treatment with TNF-α increased the mean fluorescence levels (%) of ICAM-1 expression 10.5-fold from 9.48% to 100.00%. Incubation of HCAEC with MMF caused a dose dependent inhibition of ICAM-1 expression. After incubation with MMF in concentrations of 50, 100, and 150 μg/mL expression of ICAM-1 was significantly decreased by 18.57%, 26.46%, and 37.92% (p \< 0.05, p \< 0.01, p \< 0.05). Incubation with 200, 250, and 300 μg/mL caused an decrease by 83.77%, 92.41%, and 93.17% (p \< 0.01, p \< 0.001, p \< 0.001). In HCMSMC a very weak inhibition of ICAM-1 expression was detected without statistical significance. After incubation of HCMSMC with MMF in concentrations of 50, 100, 150, and 200 μg/mL no inhibitory effect on ICAM-1 expression was detected. MMF in concentrations of 250 and 300 μg/mL caused a 9.03% (n.s.) and 16.68% (n.s.) inhibition of ICAM-1 expression. In HCAEC no toxic effects were detected after adding of MMF in concentrations of 50 μg/ml, 100 μg/ml, and 150 μg/ml, little toxic effects were found after adding of MMF in concentrations of 200 μg/ml, 250 μg/ml, and 300 μg/ml. In HCMSMC no toxic effects were detected after adding of MMF in concentration of 50 μg/ml -- 300 μg/ml. 3DLA-Model: Effect of MMF on monocyte adhesion, chemotaxis, and proliferation of human coronary smooth muscle cells ------------------------------------------------------------------------------------------------------------------- In 3DLA units the effect of MMF in a concentration of 50 μg/mL on monocyte adhesion, chemotaxis, and proliferation of HCMSMC was studied. 3DLA-units were successfully established (Fig. [3](#F3){ref-type="fig"}). On the endothelial side of the units one to two layers of cells were found. The superficial layer of these cells was composed of HCAEC, as identified by positive reaction with antibodies directed against von Willebrand factor. On the HCMSMC side of the units, three to five layers of cells with the typical hill and valley growth pattern were observed. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### \(A) Effect of MMF (50 μg/mL) on monocyte adhesion after leukocyte attack on the endothelial side of the human coronary transfilter co-culture model (3DLA units). (B) Effect of MMF (50 μg/mL) on monocyte chemotaxis after leukocyte attack on the endothelial side of the human coronary transfilter co-culture model (3DLA units). (C) Identification of monocytes by positive staining against CD68 (arrow). ::: ![](1471-2261-5-9-3) ::: Human monocytes were isolated from the residual leukocytes of single donors and identified by positive reaction with antibodies directed against CD68. A 93% purity of monocyte preparations was determined by flow cytometry. Adhesion of MC was slightly stimulated 0.5, 1, 2, 3, 4 and 6 h after leukocyte attack in MMF-treated 3DLA-units (Fig. [3A](#F3){ref-type="fig"} and [3C](#F3){ref-type="fig"}). In comparison to 3DLA-units without MMF treatment the stimulatory effect was 12.6% (n.s.), 31.5% (p \< 0.05), 29.2% (p \< 0.05), 33.2% (n.s.), 24.9% (n.s.), and 64,1% (p = 0.002). 24 h after leukocyte attack a very small inhibitory effect of MMF by 8.1% (n.s.) was detected. Chemotaxis of MC from the endothelial side of the model to the HCMSMC side of the model was very little, both with and without MMF-treatment (Fig. [3B](#F3){ref-type="fig"}). Proliferation of HCMSMC in the transfilter co-culture units was significantly inhibited by MMF (50 μg/ml) by more than 90% (Fig. [4A](#F4){ref-type="fig"} and [4B](#F4){ref-type="fig"}). Proliferation of HCMSMC in the 3DLA-units after TNF-α stimulus was significantly stimulated by 103% (p \< 0.001) in comparison to untreated controls. Adding of MMF in a concentration of 50 μg/mL significantly inhibited proliferation of HCMSMC by 95.5% (p \< 0.001). The inhibitory effect was not influenced by TNF-α. In the transfilter co-culture units without TNF-α the inhibitory effect in comparison to control was 93.4% (p \< 0.001). ::: {#F4 .fig} Figure 4 ::: {.caption} ###### \(A) Effect of MMF (50 μg/mL) on proliferative activity of co-cultured HCMSMC after leukocyte attack with monocytes on the endothelial side of TNF-α-stimulated 3DLA units. (B) Proliferation of smooth muscle cell demonstrated by positive staining of against BrdU (arrow). F = Filter, \\\* = p \< 0.001. ::: ![](1471-2261-5-9-4) ::: Coronary Organ Culture-Model: Effect of MMF on cell proliferation and neointimal proliferation after ex vivo ballooning ----------------------------------------------------------------------------------------------------------------------- The effects of ex vivo ballooning in the porcine organ culture model have been recently characterized by our group (9). Maximal reactive cell proliferation was detected at day 7, maximal reactive neointimal hyperplasia was found at day 28. In the current study the effect of a 1, 2, 3, 4, 5, 6, and 7 days incubation with MMF (50 μg/mL) on reactive cell proliferation (Fig. [5](#F5){ref-type="fig"}) and neointimal hyperplasia was studied 7 (Fig. [6A](#F6){ref-type="fig"}) and 28 days (Fig. [6B](#F6){ref-type="fig"}) after ex vivo ballooning. No clear inhibitory or stimulatory effect was detected. ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Effect of a 1--7 days treatment with MMF (50 μg/mL) on reactive cell proliferation in balloon-injured porcine coronary organ cultures at day 7 (A) and day 28 (B). C = control, CB = control ballooning. Positive staining against BrdU (arrow) in the ballooning control group (C). ::: ![](1471-2261-5-9-5) ::: ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Effect of a 1--7 days treatment with MMF (50 μg/mL) on reactive neointimal hyperplasia in balloon-injured porcine coronary organ cultures at day 7 and day day 28. C = control, CB = control ballooning. Elastica van Gieson-staining in the ballooning control group (C). ::: ![](1471-2261-5-9-6) ::: At day 7 neointimal cell proliferation was slightly increased (n.s.) in comparison to untreated controls. Adding of MMF for 1, 3, and 5 days did not exhibit an effect on cell proliferation. A stimulatory effect was found after adding of MMF for a period of 2 and 6 days, an inhibitory effect was seen after adding of MMF for a period of 4 and 7 days. 28 days after ballooning cell proliferation in the POC-model was very low, both in untreated controls and after treatment with MMF. Almost no cell proliferation was detected in the media. At day 7 after ex vivo ballooning neointimal hyperplasia was very low in comparison to controls. No neointimal hyperplasia was detected after adding of MMF for 1, 2, 4, and 6 days, an inhibitory effect on neointimal hyperplasia was seen after adding of MMF for a period of 3, 5, and 7 days. 28 days after ex vivo ballooning neointimal hyperplasia was increased by 809% (n.s.). Adding of MMF for a period of 1, 2, 3, 4, and 7 days caused an increase of neointimal thickening by 108.9%, 144%, 51.8%, 104.9%, and 4.5%. Adding of MMF for a period of 5 and 6 days decreased neointimal hyperplasia by 9.2% and 7.7%. Due to high standard deviations statistical significance was not achieved. Discussion ========== The present study employes a cascade of in vitro and ex vivo models to investigate the effect of MMF on key processes of coronary restenosis. Three basic conclusions of the study were determined. First, therapeutical concentrations of MMF exhibit a significant antiproliferative effect in HCMSMC. Second, this antiproliferative effect was not triggered via inhibition of adhesion and chemotaxis of MC or reduced expression of ICAM-1 on mRNA or protein levels. Third, a significant antiproliferative effect of MMF could not be reproduced in the porcine coronary ex vivo model. MPA, a product of Penicillium fungus, was originally isolated in 1896, and shown to have anti-neoplastic, anti-viral, anti-fungal and immunosuppressive activity. MMF is the semi-synthetic morpholinoethyl ester of MPA. After oral administration and absorbtion of MMF, the ester linkage is rapidly hydrolyzed by esterases to yield MPA, the active immunosuppressive agent. The bioavailability of oral MPA from MMF is 96% and the maximum of plasma concentration occurs about 2 h after administration \[\[[@B10]\], review\]. Over 10 years ago, MMF was first shown to prolong organ allograft survival \[[@B11]\]. Additional early studies demonstrated that MMF prevents acute rejection, reverses acute rejection and increases graft survival in several species and in different animal transplantation studies. Shortly thereafter, the first human renal allograft recipients were treated with MMF \[\[[@B12]\], review\]. Although many data exist on MMF in organ transplantation, data of the effect of MMF in experimental models of atherosclerosis and restenosis are very limited \[[@B13],[@B14]\]. The group of Fraser-Smith \[[@B13]\] demonstrated that orally administered MMF inhibits restenosis after carotid injury in a rat model and Raisanen-Sokolowski et al. \[[@B14]\] reported that MMF inhibits inflammation and SMC proliferation in a rat aortic allograft model. In these studies the relation between the significant in vitro respectively experimental effect (SI) and the maximal plasma level (MPL), the SI/MPL-ratio \[[@B6]\] was smaller than one, indicating an at least theoretical clinical relevance of the data. Recently the antiproliferative properties of MMF in non-immune cells have been summarized in a valuable review \[[@B5]\]. Gregory et al. \[[@B15]\] reported that human aortic smooth muscle cell proliferation was significantly reduced by MMF in the presence of angiotensin II or β-FGF and Mohacsi et al. \[[@B16]\] demonstrated that MMF potently inhibited rat and human aortic smooth muscle cell proliferation. In accordance with these reports the current study demonstrates a significant inhibition of HCMSMC-proliferation after incubation with 50 μg/mL of MMF in the 3DLA-model. The SI/MPL-ratio \[[@B6]\] of 1.47 indicates that this concentration is merely slightly above systemic plasma levels of MMF. The complexity of the 3DLA-model allows studies of HCMSMC-proliferation and MC-adhesion and MC-chemotaxis in one model. Although an inhibitory effect of MMF on CD4/CD8-lymphocyte proliferation is described in the literature and an inhibitory effect on MC-adhesion and chemotaxis might have been expected with reference to the clinical successes of the agent in the therapy of acute organ rejection \[[@B11]\], no inhibitory effect on MC-adhesion and MC-chemotaxis was detected in the 3DLA-model. An inhibitory effect on MC-adhesion and chemotaxis would have been of importance due to the fact that during the first 24 h after leukocyte attack monocytes seem to play a predominant role in comparison to CD4/CD8-lymphocyte attack, as recently demonstrated by our group \[[@B7]\]. The data indicate that the strong antiproliferative effect of MMF was not triggered via an inhibitory effect on MC.adhesion or chemotaxis. It has been reported that MMF inhibited the induced expression of adhesion molecules on endothelial cells measured by marked antibodies and scanning fluorimetry \[[@B17]\]. On the other hand the group of Raab et al. \[[@B18]\] demonstrated that MMF in a concentration 10 μM (corresponding to 3.2 μg/mL) does not inhibit TNF-α-induced stimulation of ICAM-1 in HUVEC. In the current study conflicting data are reported on mRNA and protein levels. Although in northern blot studies a stimulatory effect of MMF on TNF-α-induced expression of ICAM-1 was found, a dose dependent inhibitory effect was detected in flow cytometry studies. Due to the fact that littly toxic effects were found after incubation of HCAEC with MMF in concentrations of 200 μg/ml -- 300 μg/ml, the decrease of ICAM-1 protein levels in HCAEC may be partially explained by these toxic effects. Clinically relevant concentrations of MMF (50 μg/mL) however neighter stimulated nor inhibited expression of ICAM-1. If a concentration of 50 μg/mL of MMF is considered the data are in accordance with the reports of Raab et al. \[[@B18]\]. Due to the fact that MMF (50 μg/mL) did not inhibit expression of ICAM-1 on mRNA- and protein levels and exhibited no effect on MC-adhesion and MC-chemotaxis, the described antiproliferative effect of MMF eighter a direct one or it was triggered via other pathways not investigated in the current study. With the hypothesis of a direct antiproliferative effect MMF (50 μg/mL) was studied in the coronary porcine organ culture system of restenosis, the POC-model \[[@B9]\]. In the POC-model we have previously described a maximal reactive cell proliferation at day 7 and a maximal reactive neointimal hyperplasia at day 28. In order to get information on the period of time needed to treat MMF was added for a period of 1, 2, 3, 4, 5, 6, and 7 days, the time span between angioplasty and the peak of reactive cell proliferation. In the POC-model reactive cell proliferation was inhibited after adding of MMF for a period of 4 days and 7 days, no inhibitory effect on neointimal hyperplasia was detected. These results are in accordance with the data of the experimental models applied by Fraser-Smith et al. \[[@B13]\] and Raisanen-Sokolowski et al. \[[@B14]\], describing an antiproliferative effect of an about 5-times increased concentration of MMF. Surprisingly in the current study no antiproliferative effect was detected after adding MMF for a period of 5 days and 6 days. Due to the fact that the ester linkage of MMF is rapidly hydrolysed in the plasma to MPA \[[@B10]\], it can be excluded that an inactive form of MMF caused the missing inhibitory effect. However high standard deviations and the absence of perfusion in the model may have contributed to the negative effect. In the presented coronary ex vivo model of restenosis \[[@B9],[@B19]\] the solved drug gets into contact with the adventitial side, the endothelial side, and both frontal sides of the artery segment. Therefore our group has reported earlier that the model mimics a simultaneous intra/extravascular drug administration \[[@B19]\]. However due to the absence of perfusion the contact between the endothelial side of the artery segments and the culture medium is limited. Limited nutrition of this area may be critical because ballooning injury and reactive cell proliferation&neointimal hyperplasia are expected to occure predominantly in this region of the vessel wall. Conclusion ========== The current data demonstrate a significant antiproliferative effect of MMF in concentrations close to the systemic plasma level (SI/MPL-ratio: 1.47). The effect was not triggered via inhibitory effects on expression of ICAM-1 or via inhibitory effects on MC-adhesion and MC-chemotaxis. Eighter the effect was a direct antiproliferative effect or it was triggered via pathways not investigated in the present study. Probably due to technical limitations (as e.g. the missing of perfusion) the antiproliferative effect of MMF (50 μg/mL) could not be reproduced in the coronary organ culture model. A cascade of focused in vitro and ex vivo models may help to gather informations on drug effects before large experimental studies are initiated. Competing interests =================== The author(s) declare that they have no competing interests. Authors\' contributions ======================= RV, RB, and VH designed the study, RV wrote the manuscript. VK carried out the cytoflow studies, northern blot studies were done by RB and IGB. CMW carried out the studies with the 3DLA-model, coronary organ culture studies were done by SV. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2261/5/9/prepub>	PubMed Central	2024-06-05T03:55:57.634134	2005-5-12	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1156877/", "journal": "BMC Cardiovasc Disord. 2005 May 12; 5:9", "authors": [ { "first": "Rainer", "last": "Voisard" }, { "first": "Sandra", "last": "Viola" }, { "first": "Verena", "last": "Kaspar" }, { "first": "Christian M", "last": "Weber" }, { "first": "Lutz", "last": "von Müller" }, { "first": "Regine", "last": "Baur" }, { "first": "Iris", "last": "Gastrock-Balitsch" }, { "first": "Vinzenz", "last": "Hombach" } ] }
PMC1156878	Background ========== Activated macrophages are known to play an important role in the degradation process of normal and abnormal matrix. Upon differentiation, the response of macrophages to pathogens is markedly enhanced, allowing them to participate in the inflammatory and immune responses \[[@B1]\]. The differentiation process is a complex one and is controlled by the expression or the activation of several factors \[[@B2]\]. The activated inflammatory macrophage plays a crucial role in matrix destruction by producing matrix metalloproteinases (MMPs) both directly and indirectly \[[@B3]-[@B5]\]. Activated macrophages up-regulate the activity of MMPs, particularly that of MMP-9, MMP-2 (gelatinases), MMP-12 (metalloelastase) and MMP-7 (matrilysin) \[[@B6],[@B7]\]. Activated macrophages are also known to release collagenases (MMP-1 and MMP-13) under certain circumstances \[[@B8]\]. In co-culture assays, macrophages can stimulate myofibroblasts to release collagenases. In the studies of kidney disease, inflammatory macrophages of the glomeruli are found to have induced myofibroblast mesangial cells to produce stromelysin (MMP-3) \[[@B9]\]. MMPs are a family of structurally related endopeptidases that resorb macromolecules of the extracellular matrix. They participate both in physiologic connective tissue remodeling and in pathologic tissue destruction. However, the exact mechanism underlying the up-regulated expression of MMPs in macrophages is not yet clear. CD147, a 57-KD transmembrane glycoprotein, (also called extracellular matrix metalloproteinase inducer (EMMPRIN) and leukocyte activation-associated M6 antigen) is located on the surface of human tumor cells and normal keratinocytes \[[@B10],[@B11]\]. CD147, rich on the surface of most tumor cells, has been found to stimulate tumor cells and stromal cells to produce elevated levels of MMPs \[[@B12]-[@B16]\]. However, the involvement of CD147 in stimulating MMP release and activation during cellular maturation process remains unknown. This paper reports our study, using a phorbol myristate acetate (PMA)-induced cell differentiation model of the human monocytic cell line THP-1 \[[@B17]\], on the involvement of CD147 expression in the differentiation process of monocytes into macrophages, and on the correlation between CD147 expression and the secretion and activation of MMP-2 and MMP-9, and the correlation between CD147 expression and the invasive ability of mono-macrophage cells. Results ======= Alteration of morphology and phenotype on PMA-treatmented THP-1 cells --------------------------------------------------------------------- Untreated THP-1 cells are round in shape and do not adhere to the plastic surfaces of the culture plates. In the presence of 200 nM PMA for 24 h, the cells became flat and amoeboid in shape, and adhered to the dish bottom. Flow cytometry analysis revealed that these cells expressed higher levels of CD14 (51.19%), a macrophage-specific differentiation antigen, than those (2.55%) of the untreated THP-1 cells (p \< 0.05) (Figure [1a, 1b](#F1){ref-type="fig"}). The mean fluorescence intensity (MFI) of CD147 in the differentiated THP-1 cells (289.61 ± 31.63) was higher than that of the undifferentiated THP-1 cells (205.1 ± 19.25) (p \< 0.05) (Figure [1c, 1d](#F1){ref-type="fig"}). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Expression of CD14 and CD147 on the surface of differentiated and undifferentiated THP-1 cells. THP-1 cells were incubated for 24 h in the presence (b and d) or absence (a and c) of PMA (200 uM). Expression of CD14 and CD147 were measured by flow cytometry using FITC-labeled anti-CD14 mAb and anti-CD147 mAb. Analysis was conducted on a FACS™ brand flow cytometer. ::: ![](1471-2121-6-25-1) ::: MMPs release, activation and invasive ability of differentiated THP-1 cells --------------------------------------------------------------------------- As shown in Figure [2](#F2){ref-type="fig"}, SDS-polyacrylamide gelatin electrophoresis zymography showed that the secretion and activation of MMP-2 and MMP-9 were significantly enhanced in differentiated THP-1 cells compared with those observed in the undifferentiated THP-1 cells (p \< 0.05). RT-PCR showed that MMP-9 mRNA significantly increased after a 24-hour PMA treatment (Figure [3](#F3){ref-type="fig"}). As shown in Figure [4](#F4){ref-type="fig"}, a significantly increase was observed by ELISA in the release of MMP-9 in the differentiated THP-1 cells compared with that in the undifferentiated THP-1 cells. The differentiated THP-1 cells, after a 24-hour PMA treatment, were found to have significantly higher number of cells/filter to invade through transwell chambers (1626 ± 476.8) than that (282.3 ± 74.87) of undifferentiated THP-1 cells (p \< 0.05) (Figure [5](#F5){ref-type="fig"}). ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Gelatin zymography of culture medium conditioned by undifferentiated THP-1 cells and differentiated THP-1 cells. Cells were cultured in serum-free medium for 5--20 h, and the conditioned media were collected and analyzed for MMP activity by gelatin zymography. (a) image of gelatin zymography. Top two bands correspond to MMP-9 gelatinase, lower two bands to MMP-2 gelatinase. (b), (c), (d) Densitometry analysis of MMP activity. The value of MMP activity of undifferentiated THP-1 cells is shown as control (100%). Mean ± SE of at least 6 independent samples per group. p \< 0.05, Student\'s t test. ::: ![](1471-2121-6-25-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### RT-PCR analysis of MMP-9 mRNA expression in human macrophages after 24-hour PMA treatment. THP-1 cells were treated with 100 μg/ml PMA for 24 h, and total RNA was isolated by electrophoresis and transferred to a nylon membrane. The membrane was hybridized with MMP-9 specific cDNAs, and visualized by autoradiography. (a) nuclear run-on assay for newly transcribed undifferentiated and differentiated THP-1 cell mRNA. β-actin was used as the inner parameter; (b) Densitometric analysis of the bands in panel A. MMP-9 mRNA of undifferentiated THP-1 cells is shown as control (100%). Mean ± SE of at least 6 independent samples per group. p \< 0.05, Student\'s t test. ::: ![](1471-2121-6-25-3) ::: ::: {#F4 .fig} Figure 4 ::: {.caption} ###### ELISA assay MMP-9 secretion in undifferentiated and differentiated THP-1 cells. ::: ![](1471-2121-6-25-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### The invasive potential of undifferentiated and differentiated THP-1 cells. The invasive potential was evaluated in Transwell chambers as described in Materials and Methods. Briefly, cells were suspended in serum-free medium supplemented with or without PMA (100 uM) and seeded into the upper side of the Matrigel (5 μg/ml)-coated chambers. After incubation for 24 h at 37°C, the number of invaded cells was determined using a colorimetric crystal violet assay. Values are the means ± SE (n = 3\~4). ::: ![](1471-2121-6-25-5) ::: Effect of HAb18G/CD147 antagonistic peptide AP-9 on MMPs release and activation and Invasion processes ------------------------------------------------------------------------------------------------------ A significantly reduced release and activation of MMP-9 and MMP-2 from cells pre-incubated with HAb18G/CD147 antigonistic peptide AP-9 were found by gelatin zymography. The inhibitory rate of the secretion of proMMP-9 in the undifferentiated THP-1 cells was 45.07%. The inhibitory rate of the secretion of proMMP-9, the activated MMP-9 and proMMP-2 in differentiated THP-1 cells was 52.90%, 53.79% and 47.80%, respectively (p \< 0.05) (Figure [6a, 6b, 6c, 6d, 6e](#F6){ref-type="fig"}). ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Effects of AP-9 on MMPs release and activation in undifferentiated and differentiated THP-1 cells. (a) Image of gelatin zymography. Top two bands correspond to MMP-9 gelatinase, lower two bands to MMP-2 gelatinase. (b), (c), (d), (e) Densitometry analysis of MMP activity. The value of MMP activity of THP-1 cells without AP-9 is shown as control (100%). Mean ± SE of at least 6 independent samples per group. p \< 0.05, Student\'s t test. (b) Effects of AP-9 on pro-MMP-9 activity of undifferentiated THP-1 cells. (c) Effects of AP-9 on pro-MMP-9 activity of differentiated THP-1 cells. (d) Effects of AP-9 on MMP-9 activity of differentiated THP-1 cells. (e) Effects of AP-9 on MMP-2 activity of differentiated THP-1 cells. ::: ![](1471-2121-6-25-6) ::: Invasion assay showed that the amounts of cells invaded through Matrigel coated filter decreased after the treatment with AP-9 (200 ug/ml) for 24 h in both undifferentiated and differentiated THP-1 cells. The inhibitory rate of invasive potential in the undifferentiated cells and the differentiated THP-1 cells was 41.82 % and 25.15%, respectively (p \< 0.05) (Figure [7](#F7){ref-type="fig"}), control peptide SSP was not inhibitory. ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Inhibitory effects of AP-9 on invasive potential of undifferentiated and differentiated THP-1 cells to Matrigel. Cells were suspended in serum-free medium and seeded into the matrigel (5 ug/ml)-coated wells. After the incubation for 24 h at 37°C, the number of invaded cells was determined using a colorimetric crystal violet assay. Values are the means ± SE (n = 3\~4). ::: ![](1471-2121-6-25-7) ::: Discussion ========== In this study we demonstrate that the overexpression of CD147 enhances the release and the activation of MMPs (MMP-2 and MMP-9) and the invasive potential during the differentiation of monocyte THP-1 cells to macrophage cells. This is in accordance with our previous study in human hepatoma cells \[[@B18]\]. MMPs are a family of Zn^2+^-containing enzymes that cleave most of the components of extracellular matrix (ECM) and are involved in physiologic connective tissue remodeling and in pathologic tissue destruction. MMPs can be regulated by different factors. Popp et al reported that calpain/calpastatin system mediated MMP-mRNA expression of the leukemic THP-1 cells and, as a result, their invasiveness \[[@B19]\]. Worley et al identified that PPARγ and RXR agonists had complex effects on monocyte MMPs expression \[[@B20]\]. In our present study, we employed CD147 to promote MMPs. CD147 is a highly glycosylated transmembrane protein belonging to the immunoglobulin superfamily with two Ig domains. Previous studies have clearly demonstrated that CD147 is highly expressed on some tumor cells and is responsible for stimulating MMP production by stromal cells and/or other tumor cells, thereby leading to extracellular matrix degradation and elevated tumor growth and metastasis \[[@B21],[@B22]\]. CD147 has also been found to stimulate the secretion and the activation of MMPs, which are associated with tissue degradation and remodeling during inflammatory damage and wound healing. The up-regulation of the expression of CD147 in the synovial membrane of rheumatoid arthritis (RA) patients has been reported \[[@B23],[@B24]\]. Our recent studies have demonstrated that CD147 is highly expressed on monocytes in circulating blood and synovial fluid, and on macrophages/macrophages-like synovial cells in synovium from RA patients \[[@B25]\]. Similarly, the up-regulation of the expression of CD147 on monocytes/macrophages may induce MMPs produced by fibroblasts and other monocytes/macrophages, and this process may play an essential role in articular cartilage lesion development in RA. To explore the association of CD147 with the secretion and activation of MMPs, especially those in the cellular maturation, in our present study, human THP-1 monocytic cells were further stimulated to differentiate into a macrophage-like stage and some specific morphological and phenotypic alterations of PMA-treated THP-1 cells were detected. The high expression level of CD147 was observed in the differentiated process. Zymography, RT-PCR and ELISA results showed that the secretion and activation of MMP-2 and MMP-9 were significantly enhanced in the differentiated THP-1 cells. MMP-9 mRNA expression significantly increased in the differentiated THP-1 cells. Our results agree with the reports that the stimulation of the cells by PMA significantly augmented the release of MMP-9 \[[@B19],[@B20]\]. In view of the fact that CD147 may be required for the expression of gelatinases MMP-2 and MMP-9 and that the gelatinases MMP-2 and MMP-9 are expressed in leukemic cells \[[@B19],[@B20],[@B26]\], our study focused on the potential interaction between CD147 and MMPs, MMP-2 and MMP-9, in the differentiation process of monocytes into macrophages and in the invasion of monocytes/macrophages. We found that the overexpression of CD147 induced elevated levels of proMMP-2, pro-MMP-9 and their activated forms in differentiated THP-1 cells and the elevated levels of MMPs in turn enhanced the invasive ability of THP-1 cells. But the elevated expression and activation of MMPs and the enhanced invasive ability of THP-1 cells were blocked when HAb18G/CD147 antigonistic peptide AP-9 was added. These findings indicate that CD147 is involved in promoting the expression and activation of MMPs, which in turn increases the invasive potential of THP-1 cells. Conclusion ========== The results suggest that the expression of CD147 is upregulated during the differentiation of monocyte THP-1 cells to macrophage cells, and CD147 induces the secretion and the activation of MMP-2 and MMP-9 and which in turn enhance the invasive ability of THP-1 cells. The increased secretion and the activation of MMP-2 and MMP-9 and the enhanced invasive ability of THP-1 cells can be blocked by antagonistic peptide of CD147. The matured monocytes / macrophages, via their high expression of CD147, may play an important role in promoting the tissue repair or tissue damage during their inflammatory response. These findings, together with a better understanding of the possible mechanism and regulation of CD147 on MMPs production, will help the development of innovative therapeutic intervention for inflammatory diseases. Methods ======= Cell culture ------------ The human monocytic THP-1 cells (American Type Culture Collection, Manassas, Va.) were cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum (Gibco BRL, Gaithersburg, Md.), 1% penicillin/streptomycin and 2% L-glutamin at 37°C in a humidified atmosphere of 5% CO~2~. For the induction of cell differentiation, cells (5 × 10^5^to 10^6^per ml) were seeded in RPMI 1640 serum medium with 200 nM PMA for 24 h \[[@B27]\]. After incubation, nonattached cells were removed by aspiration, and the adherent cells were washed with RPMI 1640 three times. THP-1 cells in RPMI 1640 without PMA were used as control (undifferentiated) cells. Flow cytometry analysis ----------------------- Expression of CD14 and CD147 on the surface of the differentiated and undifferentiated THP-1 cells was determined by flow cytometry. Cells (5 × 10^5^) were washed 3 times with phosphate-buffered saline (PBS) and then were treated respectively with fluorescein isothiocyanate (FITC)-conjugated anti-CD14 antibody (Becton-Dickinson, USA), and FITC-conjugated anti-CD147 antibody (or FITC-conjugated Mouse IgG1, control, R&D, USA), for 20 minutes in a dark condition. Cells were washed with PBS and then analyzed on a FACS Calibur flow cytometry. Data were processed using the Cell Quest software (Becton-Dickinson, USA). Reverse transcription-polymerase chain reaction ----------------------------------------------- Using a Fast Track messenger RNA (mRNA) isolation kit (Invitrogen, San Diego, CA), RNA was extracted from the differentiated and undifferentiated cells using Trizol Reagent. The isolated total RNA (3 ug) was reverse transcribed to complementary DNA (cDNA) with a Ready to Go T-Primed First-Strand kit. The completed first-strand cDNA was amplified by PCR with primers specific for MMP9 (sense and reverse). β-actin mRNA was used as an inner parameter. Amplification was performed by adding 25 ul of RNase Free dH~2~O, 5 ul of dNTP mix, 5 ul 10× PCR buffer, and 1 ul of AmpliTaq DNA polymerase. The reaction tubes were heated to 94°C for 2 minutes and then used 28 cycles of 94°C for 30 seconds, 56°C for 60 seconds and 72°C for 90 seconds. The samples were then incubated at 72°C for 7 minutes. PCR products were electrophoresed on 1% agarose gels and visualized by ethidium bromide staining. Gel zymography analysis ----------------------- Zymography of gelatinases, the method not only reveals the activated form of the enzyme but also the zymogen, giving rise to a closely spaced doublet of digestion. When enzymes are electrophoresed in gels where their substrate is co-polymerized with polyacrylamide, they leave traces of their activity under suitable conditions (Gels were incubated first in a solution containing 2 % Tween 80 and 50 mM Tris, pH 7.5, and then in a solution containing additionally 5 mM CaCl~2~and 1 μM ZnCl~2~at 37°C in order to remove SDS from the gels and activate gelatinases). Gelatinase activity was revealed by negative staining with Coomassie Brilliant blue and quantified by densitometry \[[@B28]\]. The differentiated and undifferentiated THP-1 cells in serum-free medium were cultured for 5--20 h in the presence or absence of HAb18G/CD147 antagonist peptide AP-9, which was produced and characterized in our lab based on HAb18G. HAb18G is a new member of CD147 family and is abundantly expressed in human hepatoma tissues and on the cell surface of several hepatoma cell lines with a highly invasive potential \[[@B29]-[@B32]\]. Conditioned media were collected and the MMP activity was determined by SDS-polyacrylamide gel zymography. Media samples were centrifuged to remove cellular debris and the supernatant was collected and stored at -20°C. Each sample suspension (30 μl) was mixed with SDS sample buffer without reducing agent and loaded onto a 10% polyacrylamide gel containing 0.1% gelatin. After electrophoresis, gels were washed in 2.5% Triton X-100 and incubated in low salt collagenase buffer containing 50 mmol/l Tris, 0.2 mol/l NaCl, 5 mmol/l anhydrous CaCl~2~and 0.02% Brijdetergent at 37°C for 30 minutes. The gels were subsequently stained with 0.5% Comassie blue (R-250) and destained with buffer consisting of 20% methanol, 10% acetic acid and 70% distilled water for 30 minutes to visualize the zymogen bands. The zymography gels were scanned and analyzed using US National Institutes of Health Image 1.6 software. Invasion assay -------------- The chemotactic cell invasion assay was performed using 24-well transwells units (Costar, Cambridge NY, USA), each with an 8-um pore size polycarbonate filter coated with Matrigel (5 ug/ml in cold medium) to form a continuous thin layer. Prior to the addition of cell suspension of differentiated and undifferentiated THP-1 cells (3 × 10^5^) in serum-free medium in the presence or absence of HAb18G/CD147 antagonist peptide AP-9, Meantime an irrelevant peptide SSP, a synthetic epitope peptide of SARS-S-protein, FFSTFKCYGVSA, produced in our lab was used as control. The dried layer of Matrigel matrix was rehydrated with medium without fetal bovine serum (450 ul). The cells were then cultured for 24 h at 37°C in a CO~2~incubator. The cells remaining in the upper compartment were completely removed with gently swabbing. The number of cells invaded through the filter into the lower compartment was determined using colorimetric HE assay. Authors\' contributions ======================= JZ performed the studies and drafted the manuscript. PZ, ZNC, JLJ, conceived of the study, and participated in the design the manuscript. All authors read and approved the final manuscript. QZ participated in the writing assistance. ZBW, XYY, HT, NL, YY provided purely technical help. Acknowledgements ================ This work was supported by grants from the project of the National high-tech research and development of China (863 program, 2001AA215061).	PubMed Central	2024-06-05T03:55:57.636025	2005-5-17	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1156878/", "journal": "BMC Cell Biol. 2005 May 17; 6:25", "authors": [ { "first": "Jun", "last": "Zhou" }, { "first": "Ping", "last": "Zhu" }, { "first": "Jian Li", "last": "Jiang" }, { "first": "Qing", "last": "Zhang" }, { "first": "Zhen Biao", "last": "Wu" }, { "first": "Xi Ying", "last": "Yao" }, { "first": "Hao", "last": "Tang" }, { "first": "Ning", "last": "Lu" }, { "first": "Yong", "last": "Yang" }, { "first": "Zhi Nan", "last": "Chen" } ] }
PMC1156879	Background ========== Membrane traffic is needed for the synthesis and processing of proteins and lipids as well as the maintenance of the compartmentalization of the cell. Trafficking of intracellular membranes involves the budding of vesicles from the donor membrane and the fusion of vesicles with their respective target membranes. Several proteins are involved in membrane fusion events, including the N-ethylmaleimide (NEM)-sensitive factor (NSF), soluble NSF attachment proteins (SNAPs) and SNAP receptors (SNAREs). SNAREs are a super family of integral membrane proteins characterized by α-helical motif. The SNAREs that are functioning in neuronal exocytosis are best characterized. They include the vesicle SNARE synaptobrevin (also referred to as VAMP, vesicle-associated membrane protein) and the membrane proteins SNAP-25 and syntaxin 1 \[[@B1]\]. The pairing of target SNARE (t-SNARE) with the vesicle SNARE (v-SNARE) (trans complex) pulls the membranes together and this is possibly the driving force in the mixing of the lipid bilayers. SNAREs form bundles which contain four α-helices in a parallel arrangement \[[@B2]\]. In the middle of the hydrophobic bundle, there is a hydrophilic section which either contains three conserved glutamines (Q) or one conserved arginine (R). This led to the classification of SNAREs into Q-SNAREs and R-SNAREs \[[@B3]\]. For instance, SNAP-25 and syntaxins are Q-SNAREs and VAMP is an R-SNARE. Three helices of the helical bundle come from Q-SNAREs and one from an R-SNARE. Syntaxins and VAMP contain one helical SNARE motif but SNAP-25 contains two motifs \[[@B2]\]. The disassembly of the SNARE complexes that are formed is mediated by NSF attachment proteins, SNAPs, and the ATPase activity of NSF \[[@B1],[@B4]\]. A unique set of SNAREs is located in distinct intracellular compartments. Liposome fusion assay has demonstrated that SNAREs do show high specificity in forming complexes with each other \[[@B5]\]. The formation of functional trans complexes was mostly restricted to physiologically relevant SNARE combinations. The specificity of the complex formation resides in the SNARE motifs \[[@B6]\]. However, v-SNAREs are present in both anterograde and retrograde vesicles and therefore other proteins are needed to contribute to the specificity of vesicle targeting \[[@B7]\]. Those proteins include inter alia small Rab GTPases, Sec1 proteins, and complexins \[[@B8],[@B9]\]. Recently it has been reported that the formation of non-cognate SNARE complexes that are non-fusogenic might have a regulatory role. These inhibitory SNAREs have been suggested to increase the specificity of membrane targeting by inhibiting membrane fusion outside their specific compartments \[[@B10]\]. Syntaxins belong to a t-SNARE family of which over a dozen have already been cloned \[[@B11]\]. Over-expression studies have suggested that syntaxin 1, 2, 3, and 4 are located predominantly at the plasma membrane. Syntaxin 1 is mainly expressed in brain tissue and is thought to function specifically in neurotransmitter release, whereas syntaxin 2, 3, and 4 have a wider tissue distribution \[[@B12]\]. We have previously demonstrated that syntaxin 4 is localized, in addition to the plasma membrane, in intracellular vesicular structures as well \[[@B13]\]. These structures co-localized with rab11 staining. Treatment with NEM caused accumulation of syntaxin 4/rab11 positive labelling to actin filaments \[[@B13]\]. In this study, we investigated subcellular localization of endogenous syntaxin 2 and 3 in NRK cells. Similar to syntaxin 4, syntaxin 2 and 3 were found to localize in intracellular vesicular structures in addition to regions of the plasma membrane. In the case of syntaxins 2 and 3, NEM treatment resulted in the accumulation of these proteins in perinuclear membrane vesicles and the trans-Golgi network (TGN), respectively. Kinetic analysis in the presence of NEM suggested that both syntaxin 2 and 3 were redistributed to the perinuclear sites through actin containing structures. Results ======= Characterization of syntaxin 2 and 3 anti-sera ---------------------------------------------- Previous over-expression studies have suggested that syntaxins 1, 2, 3, and 4 are primarily localized at the plasma membrane \[[@B12],[@B14]\]. However, we have previously found that endogenous syntaxin 4 is localized in rab 11 positive intracellular membranes as well as at the plasma membrane \[[@B13]\]. Therefore in this study we investigated the localization of endogenous syntaxin 2 and 3. Anti-sera against syntaxin 2 and 3 were raised by immunization of rabbits with the amino terminal cytosolic domains of rat syntaxin 2 and 3. Since syntaxins 2, 3 and 4 are highly homologous to each other \[[@B12]\], we first determined the specificity of the anti-sera. Both syntaxin 2 and 3 anti-sera specifically recognized only their respective antigens and did not cross-react with the cytosolic domain of other syntaxins on Western blotts (Fig. [1A](#F1){ref-type="fig"}). Enriched membrane fraction of NRK cells was resolved with SDS-PAGE and blotted with syntaxin 2 or 3 anti-serum. Syntaxin 2 anti-serum blotted a band migrating close to the calculated molecular weight of syntaxin 2 monomer 34 kDa (Fig. [1B](#F1){ref-type="fig"}, lane 1). This band was abolished with pre-incubation of the syntaxin 2 anti-serum with syntaxin 2 GST-protein (Fig. [1B](#F1){ref-type="fig"}, lane 2). When the membrane fraction was blotted with syntaxin 3 anti-serum two bands appeared migrating with the apparent molecular masses of approximately 33 kDa and 40 kDa. Both of these bands were abolished by the pre-incubation of the syntaxin 3 anti-serum with recombinant syntaxin 3. These bands most likely represent two different forms of monomeric syntaxin 3. This conclusion is supported by previous observations, which indicate that mouse brain contains four different forms of syntaxin 3 and two forms of syntaxin 1 that migrate at different molecular weights \[[@B15],[@B16]\]. Our syntaxin 3 anti-serum has also been previously shown to blot syntaxin 3 at the apparent molecular weight of approximately 40 kDa in Caco-2 cell extract \[[@B17]\]. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Characterization of the syntaxin 2 and 3 anti-sera. (A) Equal amounts of syntaxin 2, 3, and 4 cytosolic domain GST fusion proteins (2 μg) were incubated with thrombin (0.3 U) for two hours to cleave off the GST (27 kDa, indicated with arrow) and analysed by SDS-PAGE, transferred to nitrocellulose filters and probed with syntaxin 2 or 3 anti-serum followed by alkaline phosphatase-conjugated secondary antibodies. (B)The Western blotting of enriched membrane fraction of NRK cells were probed with syntaxin 2 anti-serum (lane 1), syntaxin 2 anti-serum pre-incubated with syntaxin 2-GST protein (lane 2), syntaxin 3 anti-serum (lane 3) and syntaxin3 anti-serum preincubated with syntaxin 3-GST protein (lane 4) as well as horseradish peroxidase conjugated secondary antibodies. Each lane contains 25 μg protein. All the samples were boiled for 3 minutes in the presence of 2% SDS in Laemmli sample buffer. ::: ![](1471-2121-6-26-1) ::: Endogenous syntaxin 2 and 3 were found to be localized in intracellular compartments ------------------------------------------------------------------------------------ Syntaxin 2 and 3 anti-sera stained intracellular vesicular structures and regions of the plasma membrane in NRK cells (Fig. [2A](#F2){ref-type="fig"}). These stainings can be totally blocked with recombinant syntaxin 2 and 3 proteins (Fig. [2B](#F2){ref-type="fig"}). Since these syntaxins are thought to be located at the plasma membrane it was considered possible that the syntaxin 2 and 3 positive intracellular vesicles were constitutive exocytic vesicles carrying syntaxins as cargo. To study whether this is the case, we treated cells with cycloheximide for two hours to deplete newly synthesized proteins from the biosynthetic pathway membranes (Fig. [3](#F3){ref-type="fig"}). This treatment did not abolish the intracellular vesicular labelling, demonstrating that the syntaxin 2 and 3 positive membranes did not represent the newly synthesized syntaxins on their way to the plasma membrane. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Syntaxin 2 and 3 anti-sera stained intracellular vesicular structures and regions of the plasma membrane in NRK cells. The cells were fixed with 0.0 8 M lysine-0.01 M periodate-2% paraformaldehyde and permeabilized with 0.05% saponin. Syntaxin 2 and 3 were visualized using syntaxin 2 and 3 anti-sera and LRSC-conjugated goat anti-rabbit IgG. Conventional fluorescence images were viewed using an Olympus AX70 fluorescence microscope. Bars,10 μm. ::: ![](1471-2121-6-26-2) ::: ::: {#F3 .fig} Figure 3 ::: {.caption} ###### The intracellular syntaxin 2 or 3 antibody labelled structures do not represent newly synthesized syntaxin 2 or 3 proteins. The NRK cells were treated with 50 μg/ml cycloheximide for two hours. The cells were fixed with 0.08 M lysine-0.01 M periodate-2% paraformaldehyde and permeabilized with 0.05% saponin. Syntaxin 2 and 3 were visualized using syntaxin 2 and 3 anti-sera and LRSC-conjugated goat anti-rabbit IgG. Conventional fluorescence images were viewed using an Olympus AX70 fluorescence microscope. Bars,10 μm. ::: ![](1471-2121-6-26-3) ::: In the presence of NEM syntaxin 2 accumulates in perinuclear vesicles and syntaxin 3 in the TGN ----------------------------------------------------------------------------------------------- Next we investigated how the inhibition of the SNARE complex disassembly with NEM affects the localization of syntaxin 2 and 3. NEM is a sulphydryl alkylating reagent, which has been reported to inactivate the membrane fusion component NSF \[[@B18]\]. We have previously shown that NEM treatment stopped membrane transport from the TGN to the plasma membrane \[[@B19]\]. We have also found that NEM treatment caused the majority of the syntaxin 4/rab 11 positive staining to move from intracellular vesicular structures to actin filaments \[[@B13]\]. Since NEM treatment resulted in a dramatic redistribution of syntaxin 4, we studied the localization of syntaxin 2 and syntaxin 3 in similar experiments. Interestingly, NEM treatment affected the localization of syntaxin 2 and syntaxin 3 differently from syntaxin 4. In the presence of NEM endogenous syntaxin 2 and 3 accumulated in perinuclear membrane structures and very little staining of these syntaxins can be observed in vesicles or at the plasma membrane (Fig. [4](#F4){ref-type="fig"}). We used markers to study the site of accumulation of endogenous syntaxins. In the presence of NEM endogenous syntaxin 2 partly co-localized with transferrin receptor and did not co-localize with a late endosome marker, lyso-bis-phosphatidic acid (not shown) (Fig. [4](#F4){ref-type="fig"}). Some endosomal association for syntaxin 2A and 2B has been reported previously as well \[[@B20]\]. Endogenous syntaxin 2 was found to be partly co-localized with v-SNARE, cellubrevin in the perinuclear compartment. In the presence of NEM endogenous syntaxin 3 accumulated in perinuclear elements which co-localized with the TGN marker, TGN38 (Fig. [4](#F4){ref-type="fig"}). Similar redistribution of syntaxin 2 and 3 in the presence of NEM can also be observed when syntaxin 2 or 3 is transiently expressed from the cDNA in NRK cells (Fig [5](#F5){ref-type="fig"}). In control cells the syntaxins were seen both at the plasma membrane and in intracellular sites (Fig. [5A](#F5){ref-type="fig"}). After NEM- treatment the labelling of expressed syntaxins at the plasma membrane was greatly reduced and the syntaxins were seen accumulated in intracellular perinuclear elements similar to the endogenous syntaxin 2 or 3 (Fig. [5B](#F5){ref-type="fig"}). The different distribution of syntaxin 2, 3 and 4 in the presence of NEM suggests that these syntaxins are not only present at the plasma membrane but are also cycling between the plasma membrane and different intracellular compartments of the cell. According to the present results syntaxin 3 might be the SNARE responsible for the fusion of the exocytic carriers to the plasma membrane in NRK cells. This is supported by the previous observation that syntaxin 3 is present in zymogen granules of pancreatic acinar cells \[[@B21],[@B22]\]. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Syntaxin 2 localized in perinuclear membrane vesicles and syntaxin 3 localized in the TGN in NEM treated NRK cells. The NRK cells were incubated in the presence of 1 mM NEM for 15 minutes and then further incubated for two hours. Both syntaxin 2 and 3 accumulated into intracellular compartments in the presence of NEM (A,B). The cells were double stained with syntaxin 2 (A) or syntaxin 3 (B) anti-serum and LRSC-conjugated goat anti-rabbit IgG as well as antibodies against transferrin receptor (Ox26) (C) and TGN38 (D) mouse monoclonal antibodies and FITC-conjugated goat anti-mouse IgG. The yellow colour in merged images (E and F) reveals the co-localization. Confocal fluorescence images were viewed using a Leica SP1 microscope system. Bars,10 μm. ::: ![](1471-2121-6-26-4) ::: ::: {#F5 .fig} Figure 5 ::: {.caption} ###### The effect of NEM on the localization of expressed syntaxin 2 and 3 in NRK cells. Syntaxin 2 and 3 were expressed in NRK cells and stained with syntaxin 2 or 3 antiserum. (A) Control cells show localization of expressed syntaxins both at the plasma membrane and in intracellular sites. The control cells were treated with 50 μg/ml cycloheximide (controls). Also expressed syntaxin 2 and 3 accumulated into intracellular sites in the presence of NEM. (B) NEM-treated cells were double stained using syntaxin 2 or syntaxin 3 anti-serum and monoclonal antibodies against transferrin receptor (Ox26) or TGN38 as in Fig. 4. The yellow colour in merged images reveals the co-localization. Exposure in the pictures was adjusted so that only expressed syntaxins can be seen. Conventional fluorescence images were viewed using an Olympus AX70 fluorescence microscope (A). Confocal fluorescence images were viewed using a Leica SP1 microscope system (B). Bars, 10 μm. ::: ![](1471-2121-6-26-5) ::: Redistribution of syntaxin 2 and 3 takes place via actin containing structures ------------------------------------------------------------------------------ We have previously observed that in the presence of NEM syntaxin 4 accumulates on actin filaments and is directly associated with actin. In contrast to this, syntaxin 2 or 3 did not cosediment with polymerizing actin filaments suggesting that these syntaxins are not directly associated with actin \[[@B13]\]. However, syntaxin 2 and 3 positive vesicles were not redistributed when microtubules were depolymerised with cold treatment and repolymerization was prevented with nocodazole (not shown). Therefore we investigated the possible involvement of actin filaments in the transport of syntaxin 2 and 3 vesicles from the plasma membrane to the intracellular sites. Little is known about the role of actin cytoskeleton in membrane transport in animal cells. However, recently some evidence has emerged suggesting that in addition to microtubules actin cytoskeleton is involved in membrane/organelle transport. The actin cytoskeleton promotes internalization of ligands and vesicle trafficking along the endosomal pathway as well as the movement of membrane elements such as lysosomes \[[@B23]-[@B25]\]. We used an actin monomer binding protein cofilin/actin depolymerising factor, ADF, as a marker of dynamic region of actin cytoskeleton \[[@B26]\]. Both syntaxin 2 and 3 co-localized with ADF at the plasma membrane (Fig. [6](#F6){ref-type="fig"}). Also another actin marker, Oregon Green phalloidin, co-localized with syntaxin 2 and 3 in actin rich areas at the plasma membrane (Fig [7A--7D](#F7){ref-type="fig"}). When we studied the time course of the localization of syntyaxin 2 and 3 we observed that both syntaxin 2 and 3 were associated in filament-like structures after one hour\'s treatment with NEM (Fig. [7E--7H](#F7){ref-type="fig"}). These structures co-localized with actin. After two hours\' treatment with NEM syntaxin 2 and 3 accumulated at their perinuclear destinations without actin association (Fig [7I--7L](#F7){ref-type="fig"}). This result with syntaxin 2 and 3 is in contrast to our previous observation which indicated that syntaxin 4 stays associated with actin filaments even after two hours of NEM treatment \[[@B13]\]. ::: {#F6 .fig} Figure 6 ::: {.caption} ###### Syntaxin 2 and 3 staining coincides with actin at cortical regions. The NRK cells were fixed, permeabilized and then double stained with syntaxin 2 (A) or 3 (C) anti-serum, LRSC-conjugated goat anti-rabbit IgG as well as guinea pig anti-α-ADF (B,D) and FITC-conjugated donkey anti-guinea pig IgG. Conventional fluorescence images were viewed using an Olympus AX70 fluorescence microscope. Bars, 10 μm. ::: ![](1471-2121-6-26-6) ::: ::: {#F7 .fig} Figure 7 ::: {.caption} ###### Syntaxin 2 and 3 can be transported to their intracellular sites along actin filaments. The NRK cells were incubated in the presence of 1 mM NEM for 15 minutes and then further incubated for one or two hours. Syntaxin 2 and 3 were visualized using syntaxin 2 and 3 anti-sera and LRSC-conjugated goat anti-rabbit IgG; actin filaments were stained with Oregon Green phalloidin. Enlargements of the boxed areas are shown below. The arrows show the colocalization of syntaxin 2 and 3 staining with actin staining in filament-like structures after one hour of NEM-treatment. Conventional fluorescence images were viewed using an Olympus AX70 fluorescence microscope. Bars, 10 μm. ::: ![](1471-2121-6-26-7) ::: We have previously observed that the disruption of the actin cytoskeleton with cytochalasin D accumulated syntaxin 4 into actin containing aggregates \[[@B13]\]. However, when we investigated the effect of cytochalasin D on syntaxin 2 and 3 labelling no accumulation into actin aggregates was observed (Fig. [8](#F8){ref-type="fig"}). Both syntaxin 2 and 3 stayed in vesicular structures. ::: {#F8 .fig} Figure 8 ::: {.caption} ###### Syntaxin 2 or 3 staining does not accumulate into actin containing aggregates after depolymerization of actin filament. The NRK cells were treated with 10 μM cytochalasin D for 30 minutes to disrupt the actin filaments. The cells were fixed, permeabilized and then stained with syntaxin 2 or 3 anti-serum and LRSC-conjugated goat anti-rabbit IgG as well as Oregon Green phalloidin. Conventional fluorescence images were viewed using an Olympus AX70 fluorescence microscope. Bars, 10 μm. ::: ![](1471-2121-6-26-8) ::: Discussion ========== In the present study, we have shown that endogenous plasma membrane t-SNAREs syntaxin 2 and 3 are not exclusively localized at the plasma membrane. In addition to the plasma membrane localization, syntaxin 2 and 3 were found to localize in intracellular membrane compartments as well. Treatment with NEM caused syntaxin 2 to accumulate in perinuclear vesicular structures which partly co-localized with the transferrin receptor whereas syntaxin 3 accumulated in the TGN. It is therefore possible that syntaxin 2 might cycle between the plasma membrane and the perinuclear membrane vesicles, and syntaxin 3 between the plasma membrane and the TGN in NRK cells. Kinetic analysis suggested that actin cytoskeleton is involved in recycling of syntaxin 2 and 3 to the perinuclear sites. Our immunofluorescence microscopy studies indicate that endogenous syntaxin 2, 3 and 4 are located only in short sections of the plasma membrane and they are not dispersed all over of the plasma membrane. Syntaxin 2, 3 and 4 co-localize with ADF, a marker for highly dynamic regions of the actin cytoskeleton. This indicates that each of these syntaxins is present in the same section of the plasma membrane and these syntaxins cycle between this active section of the plasma membrane and different intracellular sites. In previous reports it has been suggested that syntaxin 2 is present both at the apical and basolateral portion of the plasma membrane and syntaxin 3 at the apical portion of the plasma membrane in polarized cells \[[@B21],[@B27],[@B28]\]. Therefore it is possible that syntaxins are targeted to the active sections of the plasma membrane in non-polarized cells as well. The cycling of syntaxins would make it possible to ensure that all the components of the fusion machinery are correctly targeted to an active site in an active form. We have previously observed that syntaxin 4 is directly associated with actin as examined by using a cosedimentation assay. Syntaxin 2 and 3, on the other hand, are not directly associated with actin \[[@B13]\]. Kinetic studies indicated that syntaxin 2 and 3 transiently co-localize with actin containing filaments when transported to the perinuclear sites, whereas syntaxin 4 accumulates into actin filament like structures in the presence of NEM \[[@B13]\]. These actin bundles are morphologically distinct from stress fibers and may result from altered assembly kinetics of actin filaments \[[@B29]\]. Disassembly of actin fibers with cytochalasin D causes the accumulation of actin containing aggregates. Syntaxin 4 accumulates with actin into these aggregates \[[@B13]\]. In contrast, syntaxin 2 and 3 stay in vesicular structures in the presence of cytochalasin D. This suggests that syntaxin 2 and 3 are not as tightly attached to actin as syntaxin 4. Interestingly, depolarization of Madin-Darby canine kidney epithelial cells (MDCK) caused relocalization of the apical and basolateral plasma membrane to functional apical and basolateral vacuoles, respectively. These vacuoles are associated with actin cytoskeleton \[[@B28]\]. In a few previous investigations endogenous intracellular syntaxin 2 or 3 labelling has also been observed. In those studies it has been suggested that syntaxin 2 and 3 have another role in intracellular membrane fusion processes besides the fusion process at the plasma membrane. It has been reported that syntaxin 2, 3 and 4 are present in phagosomal membranes and suggested that they are involved in phagosomal maturation \[[@B30]\]. Similarly, syntaxin 3 was found in granular membranes of the zymogenic cells and a role of granule-granule fusion was suggested \[[@B22]\]. Also in over-expression studies intracellular syntaxin labelling has been observed and it has been thought to be caused by over-expression or mislocalization \[[@B12]\]. We used lysine-periodate-paraformaldehyde fixation and saponin permeabilization in our studies to preserve the intracellular vesicular structures and therefore made them more visible than if a standard paraformaldehyde fixation had been used. Conclusion ========== The present study clearly indicates that syntaxin 2 and 3 are not solely localized at the plasma membrane but are also present in intracellular compartments and that they may cycle between these compartments and the plasma membrane through actin containing structures. This cycling of syntaxins is likely to have an important role in the regulation of t-SNARE function. Methods ======= Materials --------- Mouse monoclonal anti-rat TGN38 antibody was a gift from Dr. G. Banting (University of Bristol, Bristol, UK.) and mouse monoclonal anti-rat transferrin receptor Ox26 hybridoma cell line was obtained from Peninsula Laboratories, Inc. (San Carlos, CA). Actin monomer binding protein cofilin/actin depolymerizing factor (ADF) guinea pig affinity purified anti-serum \[[@B31]\] was a gift from Dr. P. Lappalainen (Institute of Biothechnology, Helsinki, Finland). Plasmids encoding full length mouse syntaxin 2A and rat syntaxin 3A cDNAs in pBK-CMV vectors (Stratagen, La Jolla, CA USA) were gifts from Dr. V. Olkkonen (National Public Health Institute, Helsinki, Finland). Lissamine rhodamine (LRSC)-conjugated, Rhodamine Red™-X-conjugated and fluorescein (FITC)-conjugated secondary antibodies were purchased from Jackson Immuno Research (West Grove, PA, USA). All other reagents were of analytical grade and were obtained from commercial sources. Production of fusion proteins and antibodies -------------------------------------------- The cytosolic domain of mouse syntaxin 2 (1--265), rat syntaxin 3 (1--263) and rat syntaxin 4 (1--272) in pGEX 2T vectors and syntaxin 2 and 3 antisera were gifts from Dr. V. Olkkonen (National Public Health Institute, Helsinki, Finland). The production and the purification of GST fusion proteins were performed according to the manufacturer\'s instructions (Amersham Pharmacia Biotech AB, Uppsala, Sweden). The anti-serum for syntaxin 4-GST proteins was produced in New Zealand White rabbits. Cell culture ------------ Normal rat kidney (NRK) cells were grown at 37°C in 5% CO~2~in DME supplemented with 2 mM L-glutamine, 100 U of penicillin, 10 mg/ml of streptomycin, and 10% (v/v) foetal calf serum (Biological Industries, Beit Haemek, Israel). NEM-treatment was performed by incubating cells in 1 mM NEM for 15 minutes and then further incubating in serum free DME medium. Western blotting ---------------- The samples were dissolved into SDS-Laemmli buffer, separated in SDS-PAGE, and transferred to Hybond ECL nitrocellulose membrane (Amersham Pharmacia Biotech AB, Uppsala, Sweden). Nonspecific binding of antibodies was blocked with 5% fat-free milk in TBST buffer (0.15 M NaCl, 0.05% Tween 20, 10 mM Tris-HCl pH 8.0). Secondary antibodies were conjugated with either alkaline phosphatase (Sigma, St. Louis, MO, USA) or horseradish peroxidase (Bio-Rad Laboratories, Hercules, California, USA). Alkaline phosphatase and ECL reactions were performed according to the manufacturer\'s instructions (Promega, Madison, WI, USA and Amersham Pharmacia Biotech AB, Uppsala, Sweden, respectively). The preparation of membrane fraction ------------------------------------ NRK cells were grown as confluent monolayers on 10 cm dishes. The cells were washed with hypotonic swelling buffer (10 mM KCl, 5 mM MgCl~2~, 10 mM Tris-HCl pH 7.2) and scraped in transport medium (115 mM Kacetate, 3.5 mM MgCl~2~, 1 mM EGTA, 1 mM DTT, 0.1 mM PMSF, 25 mM Hepes-KOH pH 7.4) in the presence of 1 mM phenylmethylsulfonyl fluoride, 10 μg/ml leupeptin and 2 μg/ml pepstatin A. The cells were then distrupted by repeated passage through a 23-gauge needle. The homogenate was first centrifuged at 5 000 g for 20 minutes and then the supernatant was further centrifuged at 100 000 g for one hour to precipitate membranes. Immunocytochemistry ------------------- NRK cells were grown as confluent monolayers on coverslips in DME. The cells were fixed with 0.08 M lysine-0.01 M periodate-2% paraformaldehyde \[[@B32]\] and permeabilized with 0.05 % saponin to maintain vesicular structures. Conventional fluorescence images were viewed using an Olympus AX70 fluorescence microscope with a SenSys CCD camera (Photometrics, Ltd., Munich, Germany). Images were converted using the Image-Pro Plus version 3.0 software (Media Cybernetics, Silver Spring, MD, USA). Confocal images were recorded using a laser scanning Leica SP1 confocal microscope. One layer was superimposed in the image. Abbreviations ============= ADF, actin monomer binding protein cofilin/actin depolymerizing factor; NEM, N-ethylmaleimide; NRK, normal rat kidney; NSF, N-ethylmaleimide-sensitive fusion factor; SNAP, soluble NSF attachment proteins; SNARE, SNAP receptor; stx, syntaxin; TGN, trans-Golgi network; VAMP, vesicle-associated membrane protein Authors\' contributions ======================= AMB carried out all the experiments and wrote the manuscript. EK gave advice on experiments and edited the manuscript. Both authors read and approved the final manuscript. Acknowledgements ================ We thank Anne Makkonen for skillful assistance. This study was supported by research grants from the Academy of Finland (E.K., A.M.B.).	PubMed Central	2024-06-05T03:55:57.637701	2005-5-19	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1156879/", "journal": "BMC Cell Biol. 2005 May 19; 6:26", "authors": [ { "first": "Arja M", "last": "Band" }, { "first": "Esa", "last": "Kuismanen" } ] }
PMC1156880	Background ========== Antimicrobial peptides have a critical role in the vertebrate innate immune defence against microbes. These peptides have potential as therapeutics and intelligent drug design relies on understanding how these molecules function. Defensins are peptides, which are generally cationic, are produced as prepropeptides and can be divided into subclasses based on the distribution of the six canonical cysteines that are located in the mature peptide. There are only two subclasses shared between mouse and human, the α and β-defensins. These molecules both have six canonical cysteine residues but differ in the spacing of these residues and the intramolecular disulphide bridges formed \[[@B1]\]. The antimicrobial activity of both α and β-defensins in vivois well established \[[@B2]\]. More recently β-defensins have been shown to act as a link between adaptive and innate immunity \[[@B3]\] and play important roles in cancer progression \[[@B4]\]. This has stimulated great interest in the function and evolution of β-defensins in primate lineages \[[@B5]\]. Genes that are involved in host defence often display high rates of genomic divergence and evidence for adaptive evolution. As seems to be the case with other proteins involved in the immune response, such as MHC molecules, immunoglobulins and α-defensins, this selection may be a response to the rapid evolution of pathogens \[[@B6]-[@B8]\]. In agreement with this, the four well-studied human β-defensins vary in their expression patterns as well as their antimicrobial and antiviral activities \[[@B5]\]. We have previously investigated the 8 functional human genes at the major 8p22-p23 β-defensin locus and 11 genes at the orthologous mouse locus. In both mouse and human, β-defensin paralogues show little sequence similarity in the mature peptide region and this divergence appears to have been driven by positive selection following duplication \[[@B9],[@B10]\]. These genes show an unusual pattern of evolution, with rapid divergence between second exon sequences that encode the mature peptides matched by relative stasis in the first exons that encode signal peptides \[[@B10]\]. However, these previous studies detected positive selection acting during the more distant evolutionary history of this locus to produce a diverse cluster of paralogous genes apparently established early in mammalian evolution \[[@B9],[@B10]\]. The evidence for positive selection acting within primate lineages since this paralogous cluster was established has been more equivocal and fragmentary. It has been reported that both DEFB1 \[[@B11]\] and DEFB103 (formerly DEFB3) \[[@B12]\] have evolved neutrally in primate lineages with no evidence for positive selection. In contrast there is circumstantial evidence to suggest that the evolution of primate DEFB4 (formerly DEFB2) genes has involved positive selection \[[@B13]\]. The selective forces operating on the other β-defensins at this locus in primate lineages have, until now, remained unknown. Many of the previous analyses of β-defensin genes depended upon statistical tests based on a traditional pairwise approach, calculating and comparing the rate of non-synonymous (dN) and synonymous (dS) substitution between two sequences averaged over all codons. However this approach is not appropriate for short sequences (such as β-defensins) since they do not provide sample sizes (numbers of sites and/or substitutions) that are large enough to give significant results. In addition, because such methods are based upon the dN/dS ratio across the whole length of the sequences under consideration, purifying selection at some sites could obscure the action of positive selection at other sites. A popular alternative strategy is to use likelihood ratio tests (LRTs) that allow one to estimate the dN/dS ratio (ω) at particular sites rather than averaging it over the whole molecule. This site-specific analysis has been successful at detecting positive selection in a variety of genes, particularly in gene families following expansion by duplication, and computer simulations have confirmed the power of the analysis \[[@B14]-[@B16]\]. However, Suzuki and Nei \[[@B17]-[@B19]\] found that positively selected amino acid sites are more reliably inferred by parsimony-based methods than by likelihood-based methods, with the latter prone to producing false positives. Recently a new sliding window approach based on maximum parsimony has been devised to conservatively predict the presence of selection from alignments, with special attention paid to reducing false positives \[[@B20]\]. Also Suzuki \[[@B21]\] has developed software that can detect positively and negatively selected sites based upon maximum parsimony, Bayesian or maximum likelihood methods. Given the known shortcomings of the methods available it seems most prudent to restrict attention to sites that are inferred to be subject to selection by multiple methods \[[@B22],[@B23]\]. We have used a combined strategy, complementing likelihood and parsimony-based approaches to give a comprehensive account of the selection acting at the main β-defensin locus in human, and the orthologous loci in a range of other primates and mouse. We provide statistically significant evidence for the action of both positive and negative selection in both rodent and primate lineages, and reveal the putatively functional sites within the peptide structures that have been subject to these forces at different times. Results ======= A neighbor-joining (NJ) tree was constructed from the 97 aligned mouse and primate amino acid sequences using p-distance estimates (Figure [1](#F1){ref-type="fig"}; full branch length and bootstrapping support annotation for this tree are available in Additional Files). This tree and an alignment of nucleotide sequences derived from the protein alignment (see supplementary material in Additional Files) were used in the analyses that follow. It should be noted that all 21 mouse genes analysed here are readily detectable on the orthologous rat chromosome 16 and that apparently mouse specific clades in Figure [1](#F1){ref-type="fig"} are therefore likely to be rodent specific. Orthologs of the genes within all rodent specific clades in the tree (Defb37/38/39/40, Defb2/9/10/11, Defb7/8, Defb3/5 and Defb6) were not detectable in searches of the whole genome shotgun sequence data for the dog genome. (Rat and dog genes were not analysed in detail due to the gapped and potentially misassembled nature of these draft genomes.) In contrast orthologs of all 8 primate genes were readily detectable in the dog genome sequence data (implying that they are more than 90 million years old), which supports the conclusion that the apparently rodent specific clades have indeed arisen more recently in the rodent lineage. ::: {#F1 .fig} Figure 1 ::: {.caption} ###### Phylogenetic tree relating primate and mouse β-defensin proteins constructed using neighbour-joining.Branches with less than 50% bootstrap support have been collapsed. Species and families of interest are coloured as follows: mouse sequences are red, human sequences are dark blue, Hominidae other than human are light blue, Cercopithecidae are purple the Hylobatidae are green and the Callitrichidae are yellow. Primate species names are abbreviated as detailed in Materials and Methods, mouse genes are in lower case. Branches labelled with letters show significant (P \< 0.0001) evidence of positive selection (see Figure 3 and Additional Files 3 and 4 details) and asterisks indicate branches showing significant (P \< 0.0001) evidence for negative selection. ::: ![](1471-2148-5-32-1) ::: Positive selection has acted in both primate and rodent lineages ---------------------------------------------------------------- Significant evidence of selection was sought using three different programs: PAML, ADAPTSITE and SWAPSC. Three pairs of PAML site-specific likelihood models were compared that assume variable selective pressure (as determined by the value of ω) among sites: M0 (one-ratio) and M3 (discrete), M1 (neutral) and M2 (selection), and M7 (beta) and M8 (beta&.). The M3 model (allowing variation in ω between two site classes) was a significantly better fit to the data than M0 (allowing no variation in ω) with the LRT statistic as follows: 2Δl = 2(-3468.13 -(-3260.32) = 415.62, P \< 0.0001 with 2 degrees of freedom (df). However subsequent Bayesian analysis failed to identify any sites under positive selection with greater than 95% confidence. The LRT between M1 and M2 failed to show a significant difference in fit to the data. Model M7 assumes a beta distribution for ω over 10 categories of sites. The beta distribution is limited to values between 0 and 1 providing the most flexible null hypothesis, and most stringent test, for testing positive selection. Model M8 adds another site class to the M7 model, within which ω is estimated from the data. The M8 model suggested that a small proportion of sites were under strong positive selection (ω = 21.82), but again no specific sites were implicated as under positive selection in the subsequent Bayesian analysis (95% threshold). An LRT showed that the M8 model allowing positive selection was a significantly better fit to the data than M7: 2Δl = 2(-3260.66 -(-3253.50) = 16.32, P \< 0.001 with 2 degrees of freedom (df). In addition, M8 was the best fit to the data of all 6 site-specific models tested. These somewhat equivocal results are not unexpected for this data set. It is known that these LRTs suffer from a lack of power to detect significant effects when divergence between sequences in the data set is low \[[@B15]\]. The levels of divergence between many of the primate sequences in this data set are often very low, and occasionally zero. Nevertheless the LRTs suggested that the best description of these data is a model incorporating many categories of variable ω including one showing positive selection. However, using this dataset, PAML could not confidently (i.e. at greater than 95% confidence) suggest the particular sites subject to positive selection. An earlier analysis of the 8 paralogous human β-defensins alone (and therefore based upon a set of sequences with higher average pairwise divergence), and using the same LRTs, demonstrated significant evidence for the operation of positive selection and 9 sites were nominated in more than one model \[[@B10]\]. The locations of these sites are shown in Figure [4](#F4){ref-type="fig"}. ::: {#F4 .fig} Figure 4 ::: {.caption} ###### Alignment of human and mouse β-defensin peptides showing sites of selection.Sites of significant positive and negative selection (according to SWAPSC, PAML and ADAPTSITE) in primate and mouse lineages are indicated by asterisks. Hashes indicate every tenth position in the alignment. Hominidae, Cercopithecidae, Hylobatidae and Callitrichidae are abbreviated to Hom, Cer, Hyl and Cal respectively. The positions of the beta strands and alpha helix in DEFB1 are indicated, the residues within the prepropeptide region are those before the alpha helix. The horizontal lines under the alignment denote the results of different analyses: SWAPSC results for primate and mouse lineages individually, SWAPSC sites under positive and negative selection in different branches within primates (Primate +/-) and mouse (Mouse +/-), PAML results for sites under positive selection, ADAPTSITE results for sites under positive an negative selection, sites implicated as under positive selection by two or more different programs for primate (Primate ++ and Primate \--) and mouse (Mouse ++ and Mouse\--) lineages. ::: ![](1471-2148-5-32-4) ::: To complement the analysis of Semple et al. \[[@B10]\] the same site-specific LRTs were applied to an alignment of the mouse genes alone. In each case the models consistent with the presence of sites under positive selection were significantly better fits to the data than the paired null models. M3 was a better fit to the data than M0 with an LRT statistic 2Δl = 2(-2716.01 -(-2563.54)) = 304.94 and P \< 0.001 with 2 df. M2 was a better fit to the data than M1: 2Δl = 2(-2603.43 -(-2569.69) = 67.50 and P \< 0.001 with 2 df. M8 was a better fit to the data than M7: 2Δl = 2(-2575.43 -(-2559.81) = 31.25, P \< 0.001 with 2 df. Again, M8 was the best fit to the data of all 6 site-specific models tested. In summary, these LRTs indicated that ω varies significantly between sites among these mouse genes, and in every LRT the parameters estimated suggested a substantial proportion of sites are under positive selection. The parameters estimated were fairly consistent, with ω estimated to be between 1.84 and 3.68 and with this positive selection acting upon 34--50% of sites. Seven particular sites were consistently implicated (in M2, M3 and M8 models) as being under positive selection (Figure [4](#F4){ref-type="fig"}) with greater than 95% confidence. The ADAPTSITE analysis of these data was in general agreement with the LRT results described above: all three approaches (maximum parsimony, distance-based and maximum likelihood) estimated a minority of sites (1--5% in primate sequences and 2--8% in mouse) where dN \> dS. However only the likelihood approach yielded statistically significant evidence for positively selected sites (1% in primate sequences and 8% in mouse). This may be explicable by the greater sensitivity of the likelihood approach \[[@B21]\]. All three ADAPTSITE approaches showed significant evidence for negative selection in a minority (3--5% in primate sequences and 0--8% in mouse) of sites. The general picture that emerges from the ADAPTSITE analysis suggests that positive selection has been more important than negative selection in mouse lineages and that the opposite is true of primate lineages. The SWAPSC sliding window analysis of all mouse and primate data also broadly reflected the LRT results: the data set was estimated to contain sites subject to a wide range of ω values, including a small number under positive selection. Specifically 0.77% of the sites were estimated to be subject to positive selection and 1.15% to negative selection. The branches identified as under positive and negative selection are indicated in Figure [1](#F1){ref-type="fig"} and reveal the dynamic evolutionary history of this locus. Of the 8 primate genes examined, positive selection has played a role in the evolution of 6 and negative selection has acted upon all 8. However the 21 genes at the orthologous rodent locus appear to have less turbulent histories, with 10 and 4 genes subject to positive and negative selection respectively. This leaves 7 mouse genes lacking significant evidence of either positive or negative selection. It is also notable that the majority (7/11) of mouse genes that have experienced detectable selection belong to apparently rodent specific clades in Figure [1](#F1){ref-type="fig"} (clades containing rodent genes not present in the human genome or in the whole genome shotgun sequence data for the dog genome). Selection in β-defensins varies spatially and temporally -------------------------------------------------------- The branches in the tree in Figure [1](#F1){ref-type="fig"} can be divided into three categories: (i) primate branches (i.e. those relating only primate orthologs that diverged \~5--40 million years ago); (ii) rodent branches (i.e. those among rodent genes that diverged \~12--24 million years ago (MYA) but absent from the human and dog genomes); (iii) more ancient branches (i.e. those leading to clades containing primate and rodent genes or those leading to primate clades that possess rodent or dog orthologs indicating events \~40--92 million years ago). Figure [2A](#F2){ref-type="fig"} provides an overview of the evolutionary dynamics within these three broad categories according to SWAPSC. It is clear that the selective episodes affecting primate genes have involved relatively low values of ω with many periods of negative selection while those affecting rodent genes have spanned a broader range of ω with few episodes of negative selection. More ancient branches seem to have involved the highest values of ω, which is consistent with the view that the early stages of duplication and diversification among mammalian β-defensin paralogs involved strong selection. The later stages of evolution within mammalian groups, and particularly primates, seem to have involved less innovation. Figure [2B](#F2){ref-type="fig"} shows that the focus of most positive selection in rodent and ancient branches but also of negative selection in primate branches has been the first \~120 bp of the alignment. These first 40 amino acids include the alpha helix and first beta strand of the mature defensin peptide \[[@B24]\]. ::: {#F2 .fig} Figure 2 ::: {.caption} ###### Selection across evolutionary time and sequence space.The graphs show data for primate (black diamonds), mouse (white squares) and more ancient branches (white triangles) for all significant positive and negative selection detected A: values of omega and dates. B: values of omega across the alignment in bp (midpoints of 3 codon windows). The positions of the beta strands and alpha helix in DEFB1 are indicated. ::: ![](1471-2148-5-32-2) ::: SWAPSC analysis suggests that in both primate and rodent lineages there have been episodes of selection in historically consecutive branches that have affected the same regions of these genes. For example in the primate lineage there was an episode of positive selection \~40--92 MYA (branch E in Figure [1](#F1){ref-type="fig"}) which was followed by an episode of negative selection within the past \~14.5 million years (MY), with both episodes affecting positions 109--117 in the alignment. Similarly, a burst of positive selection during the evolution of the common ancestor of primate and rodent DEFB1 (branch Q in Figure [1](#F1){ref-type="fig"}) was followed by negative selection at the same site (positions 94--102 of the alignment) within the past 8.09 MY of primate evolution. Primate evolution has also been marked by bouts of recurring negative selection at the same site, most clearly in the case of DEFB105. Here a site (positions 55--63 in the alignment) within the alpha helix has been subject to negative selection on four occasions during the divergence of old world monkeys (OWM) from H. sapiensand P. troglodytes. This contrasts with the rodent lineage where only consecutive bursts of positive selection are seen to affect the same sites. Within the rodent specific clade containing Defb38/39/40 two different sites are affected by positive selection, and each site has been targeted by selection at two points during their evolutionary past: positions 31--39 in the prepropeptide region at branches N and K; positions 109--120 in the second beta strand at branches L and M. Sites of ancient and relatively recent selection ------------------------------------------------ SWAPSC found statistically significant support for positive and negative selection at many sites. Figure [3](#F3){ref-type="fig"} shows the raw data graphed for two of the branches (the most recent and oldest in the tree) demonstrating significant evidence for positive selection (Figure [1](#F1){ref-type="fig"}). For each branch the Ka and Ks measured at successive 3 codon windows are shown. These graphs make clear that the sites of positive and negative selection identified as significant are likely to be a subset of those actually subject to these forces in reality. Figure [3](#F3){ref-type="fig"} is also typical of the results obtained for relatively recent and older branches (Figures for all other branches showing significant selection are available in Additional Files.) The recent branch shows the changes between the last common ancestor of C. aethiops(vervet monkey) DEFB106 and P. anubis(olive baboon) DEFB106 and DEFB106 in C. aethiops. Most regions of the molecule show little or no changes, as expected over \~9.62 MY but two consecutive windows (bp 40--48 and 43--51) demonstrate a significant excess of Ka over Ks. The older branch shows changes between the last common ancestor of all primate DEFB1 and mouse Defb1 sequences and the ancestral primate DEFB1 sequence. This older branch concerns events \~40--92 MYA and shows greater variation in Ka and Ks across the sequences, though only two regions show significant evidence for positive selection (bp 4--12 and 67--75) and a further two for negative selection (bp 28--42 and 160--168). There are many sites such as this where selection is detected similarly in all or most lineages, reflecting more ancient events in mammalian evolution. All except one of the conserved cysteine residues are implicated as being under negative selection in both primate and mouse lineages. Similarly, a small region at the extreme N terminal of the mature peptides (positions 2--4 in Figure [4](#F4){ref-type="fig"}) was found to be under positive selection in primate and mouse lineages. However certain regions of these molecules have experienced positive and negative selection in different lineages. Arguably it is these sites, where selection has at one time favoured a change but at another required stasis, that are likely to be most potent in altering the functions of these proteins. These sites mainly cluster at a central region of the mature peptides (positions 36--40 in Figure [4](#F4){ref-type="fig"}), although other sites, often those neighbouring cysteine residues (positions 13--14, 24--25, 33 and 55--56), appear to have been subject to such opposing selection. ::: {#F3 .fig} Figure 3 ::: {.caption} ###### Substitution rates and selection measured in two branches of the tree relating mammalian β-defensins.Each graph shows Ka (black circles), Ks (white circles) and significant positive selection across the sequence (midpoints of 3 codon windows) encoding the mature peptide. The graphs display the analysis for branches C and F in Figure 1, a relatively recent and an older branch demonstrating positive selection respectively. The positions of the beta strands and alpha helix in DEFB1 are indicated. ::: ![](1471-2148-5-32-3) ::: ADAPTSITE (using a likelihood approach) also nominated sites showing significant evidence of positive and negative selection and these can be compared with those nominated by SWAPSC and by PAML run separately on mouse and primate datasets (Figure [4](#F4){ref-type="fig"}). However there are no sites in primates and only three in mouse (positions 21, 33 and 37 in Figure [4](#F4){ref-type="fig"}) where all three programs show evidence for positive selection. This lack of agreement between all three programs is perhaps unsurprising considering the different restrictions of each analysis. SWAPSC can find lineage-specific events but relies on many simulated datasets to assess significance and lacks the single site resolution of the other two programs. PAML was restricted here to examining selective episodes between paralogs and does not test for significant evidence of negative selection. ADAPTSITE results are only reliable for alignments positions with more than 15 nucleotide differences \[[@B21]\]. In addition both ADAPTSITE and PAML do not consider gapped positions in the alignment, whereas SWAPSC will if the gap is absent from the lineage under examination. Perhaps most importantly both the PAML and ADAPTSITE analyses discussed here examine site-specific events across the whole alignment under scrutiny, effectively averaging over lineages. Figure [4](#F4){ref-type="fig"} shows the level of agreement between any pair of programs and appears to indicate only modest agreement between them on the location of sites under positive selection. Most of the positively selected primate sites (5/9) and mouse sites (7/7) implicated by PAML are also supported by either SWAPSC or ADAPTSITE. Similarly most of the positively selected primate sites (2/2) and mouse sites (14/18) implicated by ADAPTSITE are supported by at least one of the other programs. Following the logic of Podlaha and Zhang \[[@B22]\] such sites, supported by more than one independent analysis, are the most reliably inferred. However, this assertion assumes that the three methods used are similarly informative for the present dataset. Significant heterogeneity between the results of the three methods might indicate they are not. A heterogeneity G-test \[[@B23]\] was used to assess uniformity between the outcomes of the three tests (SWAPSC, PAML and ADAPTSITE) for positive selection. The numbers of sites predicted to be positively selected (per 10 residue interval across the alignment in Figure [4](#F4){ref-type="fig"}) were counted. It was necessary to consider all positively selected sites predicted for mouse and primate data together, and to collapse the first and final intervals to create intervals containing sufficient numbers of predicted sites (i.e. greater than zero). This calculation indicates that G~H~= 11.86 with 6 degrees of freedom which is not significant. Thus although the results of the three tests for positive selection used do not agree perfectly there is no significant heterogeneity between them. Structural implications of evolutionary history for β-defensin peptides ----------------------------------------------------------------------- It has been shown that primate and murine β-defensins share striking similarity at the level of secondary and tertiary structure, in spite of very low levels of sequence similarity \[[@B26]\]. The most reliably inferred sites of selection in Figure [4](#F4){ref-type="fig"} (those implicated by more than one different method) were mapped to the known structures of the human DEFB1 and the mouse Defb7 mature peptides to examine differences in the distribution of these sites between primate and rodent lineages (Figure [5](#F5){ref-type="fig"}). As discussed above, there are more sites demonstrating positive selection in the murine defensins as compared to the primate defensins (Figure [4](#F4){ref-type="fig"}). However some clear similarities between the positions of positively selected sites are evident on the murine and primate structures (Figure [5](#F5){ref-type="fig"}). It seems that sites within the triple beta-strand so characteristic of these peptides are largely unaffected by positive selection. The few exceptional sites subject to positive selection found in the triple stranded β-sheets that form the structural core of the β-defensins, may represent alterations in the oligomerisation of β-defensins (Figure [5](#F5){ref-type="fig"}). ::: {#F5 .fig} Figure 5 ::: {.caption} ###### Structural implications of primate and mouse sites under significant selection.Each site is implicated by two or more programs from the following three: PAML, ADAPTSITE and SWAPSC. Primate sites were mapped to the structure of the human DEFB1 mature peptide (A) and mouse sites to the structure of the mouse Defb7 mature peptide (B). Sites subject to selection are depicted as inflated regions of the structures coloured red to indicate positive selection. The particular residues in DEFB1 and Defb7 corresponding to positively selected sites are also indicated with arrows. Ala (marked with an asterisk) is subject to negative and positive selection in different primate lineages. ::: ![](1471-2148-5-32-5) ::: Both primate and rodent lineages show a large number of sites subject to positive selection within the N-terminal portion of the mature peptide. Two sites for both primate and murine, were located within a region which in DEFB1 and Defb7 forms an alpha helix. Since regions of proteins within membranes are often helical, with surfaces covered with hydrophobic resides, we speculate that the alpha helical section may be involved in anchoring the β-defensin to a bacterial cell wall. Thus the sites within the alpha helix under positive selection may be significant in the specificity of β-defensins, either with respect to their antimicrobial or chemoattractant properties. The longest loop region of these peptides (indicated to the right on Figure [5](#F5){ref-type="fig"}) also contains sites of positive selection. For the murine form this loop is almost exclusively subject to high selection, which suggests that this part of the structure has a key functional role in these small peptides. If, as shown for the β-defensin HNP3 \[[@B27]\] the second beta-strand is involved in oligerimisation, many of these sites would all be left exposed after dimer formation, suggesting an rapidly diverging exposed \'skirt\' around the peptide. This is confimed by the NMR data of Schibli et al. \[[@B25]\] whose structures of the human β-defensin DEFB3 suggest symmetrical dimer formation, through the beta strand 2 of the β-sheet. Some sites found to be subject to positive selection in rodents (see Figure [4](#F4){ref-type="fig"}) are not represented in Figure [5](#F5){ref-type="fig"} as they are part of the N-terminal prepropeptide region that is removed to produce the mature peptide. This contrasts with studies of α-defensins which have found an absence of positively selected sites in the prepropeptide region \[[@B24]\]. In primate lineages sites within the prepropeptide region have undergone negative selection (Figure [4](#F4){ref-type="fig"}). These observations strongly imply that the prepropeptide region is more important to β-defensin function than has previously been appreciated. Discussion ========== The present data demonstrated evidence for positive and negative selection in different branches but at overlapping regions of the same molecules. It is clear that in such a case the use of site specific LRTs (effectively averaging across the branches of a tree) may have little power to detect the sites involved. A combined approach, using such LRTs with an independent method (SWAPSC) examining each branch individually has previously been employed in an analysis of α-defensins \[[@B24]\]. Here we have extended this approach to examine ancient and more recent events in primate β-defensin evolution. It is thought that positive selection may play a major role in the divergence of paralogues from one another following a duplication event \[[@B28]\]. In agreement with this many of the more ancient selective episodes detected here appear to date back to the birth of these genes by duplication at an ancestral mammalian locus. However it is striking that later episodes of positive selection often focus on sites overlapping those subject to the more ancient events, particularly in the rodent lineage. This suggests that many of the same sites that originally conferred specificity of function upon these peptides were altered to provide novel functions many millions of years later. Although in the primate lineage it would appear that the original episodes of positive selection following duplication were usually followed by negative selection. Boniotto et al. \[[@B13]\] found some evidence for positive selection in human DEFB4 during the divergence of primate species. They reported several residues of interest in the mature peptide where substitutions were observed in primate groups other than the great apes. These N-terminal residues are concerned with oligomerisation in the human peptide and they suggested that a particular form of oligomerisation might have evolved in apes and humans. We detected no significant positive selection in DEFB4 since the divergence of the same primate groups, although our analysis is conservative. However the only statistical evidence presented by Boniotto et al. \[[@B13]\] to support their hypothesis that DEFB4 is subject to positive selection was a Z-test, which is not stringent enough for short sequences such as these. A consequence of the more rigorous statistical analyses to which we subject the sequences described here is that we may miss real episodes of positive selection. Antcheva et al. \[[@B29]\] synthesised variant molecules based on their observation that DEFB4 (formerly DEFB2) has been subject to positive selection during the divergence of various primate lineages. They synthesised the M. fascicularisDEFB4 orthologue (\"mfaBD2\") and a variant of the human peptide lacking Asp(4), (\"-D)hBD2\", which is characteristic only of the human/great ape peptides. hBD2 and mfaBD2 showed a significant difference in specificity, the former being more active towards Escherichia coliand the later towards Staphylococcus aureusand Candida albicans. Asp(4) in the human peptide appears to be important, as (-D)hBD2 was less structured and had a markedly lower antimicrobial activity but this site was not identified as being subject to positive selection here. A clear but unexpected result of the present analyses was that the preproregion has been subject to significant positive selection in rodents and negative selection in primates. This has not been observed previously. It is commonly assumed that the preproregion is cleaved as the mature peptide is secreted from the cell. We conclude that further investigations of cleavage and the functional consequences of sequence changes in this region are merited. Conclusions =========== We have used a combination of maximum likelihood based tests and a maximum parsimony based sliding window approach to give the most statistically rigorous and detailed view of the selective history at the major primate β-defensin locus. These data shed light on the evolution of human innate immunity but also have practical applications in the design of novel antibiotics. Sites within the active, mature peptides have been subject to repeated episodes of selection in different primate lineages, and by implication are important in functional diversity. Additional sites within the prepropeptide region, which is cleaved before secretion, were also subject to selection suggesting a previously unappreciated functional significance of this region. Methods ======= Sequence data ------------- In alignments and figures primate species names were abbreviated to two letters as follows: Cercopithecus preussi(Preuss\'s monkey) (cp), Cercopithecus aethiops(Vervet monkey) (ca), Cercopithecus erythrogaster(Red-bellied monkey) (ce), Presbytis cristata(Silvered langur) (pc), Presbytis obscurus(Spectacled langur) (po), Presbytis melalophos(Banded langur) (pm), Macaca mulatta(Rhesus Macaque) (mm), Macaca fascicularis(crab-eating macaque) (mf), Papio anubis(olive baboon) (pa), Hylobates lar(Lar gibbon) (hl), Hylobates moloch(Silvery gibbon) (hm), Hylobates concolor(crested gibbon) (hc), Callithrix jacchus(common marmoset) (cj), Saguinus oedipus(cotton-top tamarin) (so), Pan troglodytes(chimpanzee) (pt), Gorilla gorilla(gorilla) (gg), Pongo pygmaeus(orangutan) (pp), Homo sapiens(human) (hs). The Cercopithecidae are represented by cp, ca, ce, pc, po, pm, mm, mf and pa; the Hylobatidae by hl, hm and hc; the Callitrichidae by cj and so; and the Hominidae by pt, gg, pp and hs. Note that sequences from every species were not available for each primate gene (see Figure [1](#F1){ref-type="fig"} and Additional Files). All mouse sequences were recently published by Zaballos et al. \[[@B30]\]. The H. sapiensand P. anubissequences were as previously published \[[@B10]\]. Previously published sequence data for primate DEFB1 \[[@B11]\], DEFB4 \[[@B13]\] and DEFB103 \[[@B12]\] were combined with the following novel data. Whole genome shotgun reads from the M. mulattagenome representing DEFB104 (69840222, 73807150), DEFB105 (73492526, 74381588, 72060044) and DEFB108 (71259620, 72564100, 71889652, 72776644, RHQRA66TR) were identified using BLAST \[[@B31]\] from the Ensembl Trace server <http://www.ensembl.org/>. Genomic sequence assembly contigs from the P. troglodytesgenome were obtained from Ensembl in the same way for DEFB105 (AADA01159356) and DEFB107 (AADA01159356). The published Rattus norvegicus(rat) genome assembly \[[@B32]\] and the full Canis familiaris(dog) \~7.6X coverage whole genome shotgun data (downloaded September 2004 from the Ensembl Trace Server: <http://trace.ensembl.org/>) were searched using TBLASTN \[[@B31]\] with default settings. PCR of novel second exon sequences from P. anubis, C. aethiops, P. pygmaeus,C. jacchusDNA was achieved using primers designed to the human exon 2 flanking sequence. PCR programmes were used with a relaxed annealing temperature that revealed a single species by gel electrophoresis. PCR products were cloned and several clones were analysed for each PCR. At least two clones were sequenced in both directions. Primers were as follows with forward primer sequence preceding the reverse primer sequence for each gene. DEFB103: 5\'GTGCTGTTTTGTCATTGCAG, 5\'GATTTAAAAAAAAAAATCAAGCTC; DEFB104: 5\'CAGTGCCATATCCTGTTATCTAG, 5\'GCTGCTAGGCCGCAGGAAGG; DEFB105: 5\'GCAGCTCTTTCTTGGCAGAG, 5\'GCTGGTCTGGTTTGTCAGATC; DEFB106: 5\'TGGCTCCTTCCCTGTGTAG, 5\'CACTTGACAAACTGAGCAAAG; DEFB107: 5\'CTGCTTTCTTTACTTAGCCA, 5\'GTGCTTAGTTTTTAATGTTTCTTTC; DEFB108: 5\'CAATAACCCCTTCTGCATGTAG, 5\'CTCAATTCTTGGTTGATGCCC. Novel sequences were deposited in GenBank under accession numbers: AY831729, AY831730, AY831731, AY831732, AY831733, AY831734, AY831735, AY831736, AY831737, AY831738, AY831739, AY831740, AY831741, AY831742, AY831743, AY831744, AY831745, AY831746. Phylogenetic inference and evolutionary analyses ------------------------------------------------ Protein sequences were aligned using T-Coffee \[[@B33]\]. An alignment of coding sequences corresponding to the protein alignment was constructed using the tranalign program from the EMBOSS package \[[@B34]\]. A phylogenetic tree was constructed using MEGA (version 2.1 \[[@B35]\]) by the neighbour joining (NJ) method \[[@B36]\] using p-distance estimates, which is thought to be the most reliable method for constructing NJ trees of closely related sequences \[[@B37]\]. The tree was rooted with chicken gallinacin 1 (P46156) and the reliability of each node was assessed with 1000 bootstrap replications. All of the best supported branches (\>= 50% of replications) were also observed in an equivalent NJ tree constructed with the gamma distribution model implemented to account for rate heterogeneity among sites. The shape parameter of the gamma distribution (α) was estimated using baseml from the PAML package (version 3.13 \[[@B38]\]). The same alignment was used as the basis for trees constructed by maximum parsimony (using MEGA version 2.1 with default settings) and maximum likelihood (using PHYLIP version 3.6 \[[@B39]\] with default settings). In both cases the trees produced shared substantively similar topology with the NJ trees. (All trees are available on demand.) Nodes within the primate lineage were dated according to a widely accepted phylogenetic analysis \[[@B40]\]. The divergence of rat and mouse was taken to be 12--24 MYA \[[@B32]\] and the last common ancestor of mammals was assumed to be 92 MYA \[[@B41]\]. Likelihood ratio tests (LRTs) were performed using codeml with the site-specific models of the PAML package and the tree constructed as above. The six site-specific models recommended by Anisimova et al. \[[@B15]\] were tested: M0 (one-ratio), M1 (neutral), M2 (selection), M3 (discrete), M7 (beta), and M8 (beta+ω). These LRTs indicate whether the substitutions inferred from an alignment are best explained by one of two models of ω = dN/dS, where dN and dS are the nonsynonymous and synonymous substitution rates respectively. When parameter estimates under a model allowing positive selection suggest the presence of a number of sites with ω \> 1, Bayes theorem was used to calculate the posterior probability that a given site is one of those that are selected \[[@B38]\]. It is worth noting that PAML LRTs are reported to be conservative for short sequences, though the Bayesian prediction of sites under positive selection is largely unaffected by sequence length \[[@B15]\]. Concerns have been raised over the reliability of particular sites inferred to be subject to positive selection using PAML \[[@B19]\] and so corroborating evidence was sought from independent methods. Evidence for positively and negatively selected sites was sought using ADAPTSITE (version 1.3) according to the procedure recommended by Suzuki \[[@B21]\]. Specifically, equal equilibrium codon frequencies were assumed, an NJ tree based upon p-distance was used as above but was unrooted, the transition/transversion rate ratio was taken to be 1.02 for both primate and mouse datasets (estimated using MEGA) and the significance level for detecting selection was 5%. All three approaches accommodated within ADAPTSITE were employed: maximum parsimony, a distance-based Bayesian method and maximum likelihood. The maximum likelihood analysis was run with two different initial values of ω (0.00001 and 1) to ensure the results were robust to such differences, and only sites where the estimated number of nucleotide substitutions was greater than 15 were considered \[[@B21]\]. The selection operating in different regions of the sequences and within different branches of the phylogenetic tree under study were also estimated using SWAPSC with 1000 simulated data sets \[[@B20]\]. This program uses the differences between the estimated and expected numbers of synonymous and nonsynonymous substitutions to evaluate various hypotheses \[[@B42]\]. Firstly it seeks evidence for regions that have suffered the saturation of synonymous sites: where the number of synonymous substitutions is significantly smaller than expected. In addition it seeks mutational hotspots: regions where the number of synonymous and nonsynonymous nucleotide substitutions are greater than expected under neutrality. Remaining regions where the number of nonsynonymous nucleotide substitutions is smaller than expected (or where ω is significantly smaller than the mean ω estimated for the alignment) are identified as under negative selection. Positive selection is inferred where the estimated number of nonsynonymous nucleotide substitutions is greater than expected by chance and where ω is significantly greater than 1. Where regions have an estimated number of nonsynonymous substitutions greater than expected but ω \< 1 or where ω \> 1 but there is evidence for saturation of synonymous sites such regions are said to have accelerated rates of nonsynonymous substitutions. Thus SWAPSC seeks to avoid inferring positive selection where there is insufficient data to support it or where saturation may cause bias. A by-product of the SWAPSC analysis is substitution rate estimates for all branches of the tree under study, within overlapping 3 codon windows across the alignment. Each site identified by PAML or ADAPSITE as subject to selection was checked against the synonymous rate estimates made by SWAPSC to ensure that these sites were not saturated at any branch of the tree. Structural analysis ------------------- In order to establish the relevance of these finding to the solution structure of murine-defensins, we mapped the adaptive sites onto the nuclear magnetic resonance (NMR) structures of mouse Defb7 \[[@B26]\] and human DEFB1 \[[@B25]\]. For each structures were downloaded as PDB files from the Brookhaven protein databank, <http://www.rcsb.org/pdb>. The structures were viewed using VMD <http://www.ks.uiuc.edu/Research/vmd/>. Authors\' contributions ======================= CS performed the sequence analysis (in combination with PG) and statistical analyses, participated in experimental design and drafted the manuscript. AM and FK were responsible for the PCR experiments. HE and PB performed the structural analyses and helped to draft the manuscript. JD conceived of the study, and participated in its design and coordination and helped to draft the manuscript. All authors read and approved the final manuscript. Supplementary Material ====================== ::: {.caption} ###### Additional File 1 The full protein multiple sequence alignment for all primate and mouse sequences analysed, it is presented in FASTA format followed by interleaved (CLUSTALW) format. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 2 The full nucleotide multiple sequence alignment (derived form the corresponding protein alignment in [Additional file 1](#S1){ref-type="supplementary-material"}) for all primate and mouse sequences analysed, it is presented in FASTA format followed by interleaved (CLUSTALW) format. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 3 Primate substitution rates and selection measured in various branches of the tree relating mammalian β-defensins (see Figure [1](#F1){ref-type="fig"}). Each graph shows Ka (black circles), Ks (white circles) and significant selection (rectangles for positive selection and circles for negative selection) within sliding SWAPSC windows of 3 codons across the sequence encoding the mature peptide. Graphs A, B, C, D, E and F correspond to branches A, B, C, D, E and F respectively in Figure [1](#F1){ref-type="fig"}. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 4 M. musculussubstitution rates and selection measured in various branches of the tree relating mammalian β-defensins (see Figure [1](#F1){ref-type="fig"}). Each graph shows Ka (black circles), Ks (white circles) and significant selection (rectangles for positive selection and circles for negative selection) within sliding SWAPSC windows of 3 codons across the sequence encoding the mature peptide. Graphs G, H, I, J, K, L, M, N, O, P and Q correspond to branches G, H, I, J, K, L, M, N, O, P and Q respectively in Figure [1](#F1){ref-type="fig"}. ::: ::: {.caption} ###### Click here for file ::: ::: {.caption} ###### Additional File 5 Phylogenetic tree relating primate and mouse β-defensin proteins constructed using neighbour-joining. Identical to Figure [1](#F1){ref-type="fig"} but with the addition of bootstrapping support (above branches) and branch lengths (below branches). Primate species names are abbreviated as detailed in Materials and Methods, mouse genes are in lower case. ::: ::: {.caption} ###### Click here for file ::: Acknowledgements ================ We thank Drs Dominic Campopiano and Bob Hill for helpful discussions with this work. Research was supported by the UK MRC. The School of Chemistry and the EPSRC are to be thanked for funding the studentship of HE, in addition PB holds an Advanced Research Fellowship from the EPSRC.	PubMed Central	2024-06-05T03:55:57.640119	2005-5-18	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1156880/", "journal": "BMC Evol Biol. 2005 May 18; 5:32", "authors": [ { "first": "Colin AM", "last": "Semple" }, { "first": "Alison", "last": "Maxwell" }, { "first": "Philippe", "last": "Gautier" }, { "first": "Fiona M", "last": "Kilanowski" }, { "first": "Hayden", "last": "Eastwood" }, { "first": "Perdita E", "last": "Barran" }, { "first": "Julia R", "last": "Dorin" } ] }

Subsets and Splits

No community queries yet

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