text
string | predicted_class
string | confidence
float16 |
|---|---|---|
This regression parameter dataset is based on high-resolution T1 3D structural brain MR imaging data of 1919 healthy subjects aged 13–900 months [1–75 years]. Images were acquired at 1.5 and 3 T and were selected from four publicly available datasets (NIH, C-MIND, fCONN, and IXI). The dataset contains regression parameters from 6 DARTEL iterations for GM and WM each and can be used within the CerebroMatic toolbox to generate matched DARTEL/SHOOT templates for a researcher's own population.
|
study
| 100.0 |
The aim of this article is to describe a set of regression parameters which can be used within the CerebroMatic toolbox . The general approach of this toolbox is as follows: instead of simple averaging a large number of subject's brain MRI data (usually following tissue segmentation) to generate a reference brain/template, the data is instead analyzed statistically. The main advantage is that this approach is able to take into account the dominating demographic (such as age and gender ) and technical factors (such as field strength and data quality , ). As opposed to the previous application of this idea (and its implementation within the Template-O-Matic toolbox ), the CerebroMatic now uses a much more flexible statistical approach, namely multivariate adaptive regression splines . This allows modeling smooth trajectories of change with much higher flexibility and accuracy, especially in the context of an inhomogeneous group (see , Fig. 1, for an illustration). The result of this modeling is a regression parameter set for each voxel, and each tissue class. From these parameters, a synthetic tissue class can then be generated as the predicted values are linear combinations of the original response values. Hence, the resulting tissue class can be described based on (and thus, matched to) the demographics of a new and independent input population.Fig. 1Overview of the intermediate steps of the image data processing pipeline: each whole brain T1 3D datasets was first bias-corrected (1) and, using CAT12, segmented into GM (2) and WM (not shown). Then, an iterative non-linear registration (3–8) to the respective group mean tissue map was applied, resulting in ever crisper tissue maps (upper row) and corresponding deformation fields (lower row, illustrated here by their Jacobian determinants). This results in conventional DARTEL templates (see Fig. 2, Fig. 4, top rows). The tissue maps were also submitted to the CerebroMatic toolbox, resulting in synthetic DARTEL templates (see Fig. 2, Fig. 4, bottom rows).Fig. 1
|
study
| 100.0 |
Overview of the intermediate steps of the image data processing pipeline: each whole brain T1 3D datasets was first bias-corrected (1) and, using CAT12, segmented into GM (2) and WM (not shown). Then, an iterative non-linear registration (3–8) to the respective group mean tissue map was applied, resulting in ever crisper tissue maps (upper row) and corresponding deformation fields (lower row, illustrated here by their Jacobian determinants). This results in conventional DARTEL templates (see Fig. 2, Fig. 4, top rows). The tissue maps were also submitted to the CerebroMatic toolbox, resulting in synthetic DARTEL templates (see Fig. 2, Fig. 4, bottom rows).
|
study
| 100.0 |
A shortcoming of this approach, however, was that more current (and computationally intense) high-dimensional warping approaches such as the commonly-employed DARTEL- or SHOOT-approach use an iterative self-registration scheme. Hence, a single tissue prior is not appropriate anymore as these approaches require an increasingly crisp set of tissue priors to register to. The DARTEL-approach has shown great accuracy when compared with other non-linear spatial deformation approaches and was later refined .1 To generate such high-quality tissue maps, however, large populations are required which may not always be available, especially in the case of an “unusual” population such as children or elderly subjects. The here-described parameter set is the result of using the CerebroMatic toolbox to statistically generate such tissue prior sets for ultimate use within the DARTEL/SHOOT framework, based on a large population of healthy infants, children, and young as well as older adults.
|
study
| 99.94 |
For this data in brief article, the same initial datasets as already described in were used, all of which are available from public repositories. Four large datasets with rigorous quality control mechanisms were selected, two for imaging data from children (the National Institute of Health's Study of Normal Brain Development and the Cincinnati MR Imaging of Neurodevelopment study ) and two for imaging data from adults (The 1000 functional connectome study and The Information eXtraction from Images study ). Following additional local quality control, a total of 1919 high-resolution 3D T1 images could be included. See Table 1 for demographic and imaging details of all included subjects. Further details on all subjects as well as respective credits, sponsors, and disclaimers can be found in the Supplementary material S1.Table 1Demographic information about all 4 contributing and the full dataset; n=number; T=Tesla. Values are described as sums or mean±SD. See text for details.Table 1nAge [months]Voxel volume [µl]Image quality [%]1.5 T [n]3 T [n]NIH414122.41±52.150.99±.0876.47±10.394140C-MIND20699.87±55.80.99±.0679.23±9.270206fCONN757331.76±156.351.18±.3282.72±2.8715742IXI542571.84±187.541.05±.0383.81±1.98178364Full Dataset1919329.51±228.031.08±.2281.31±6.76071312
|
study
| 99.94 |
Image data preprocessing was described in detail in and is therefore only briefly summarized here. All data processing and analysis steps were performed in Matlab (Mathworks, Natick, MA), in part using functionality available within the spm12 software package (rev. 6906; University College London, UK). A 7th degree B-spline interpolation algorithm was used when writing images , but all other parameters were left at their default values unless specified otherwise. Initially, all images were reoriented and bias-corrected, using functionality provided within the unified segmentation framework . Tissue segmentation was then achieved using the cat12 toolbox (r1092 ) which is a priorless modification and extension of the SPM12 “new segment” approach . Tissue probability maps (for gray matter [GM] and white matter [WM] only) were spatially normalized using an affine registration scheme to allow for an initial overlap of large structures. We opted for an affine approach here (instead of the usually recommended rigid-body procedure ) as the overall size difference between the subjects included here (between infancy and old age , ) must be expected to otherwise pose insurmountable challenges for the ensuing non-linear deformation steps (see below). Visual quality control was also performed as previously described , using individual inspection of each map at the level of the basal ganglia and the cerebellum to identify overt failure of spatial normalization or tissue segmentation.
|
study
| 100.0 |
The DARTEL approach performs an iterative but highly integrated spatial normalization scheme, in that all images in a population initially contribute to a straight mean to which then again all images are iteratively registered to. The images resulting from this first round are then again used to create a second average image, to which the images are again registered, and so on. Hence, in a first step, the standard DARTEL procedure (SPM12 batch module “DARTEL, create template”) was applied to the full dataset, yielding an initial set of six conventional templates for GM and WM each. In a second step (SPM12 batch module “Run DARTEL with existing template”), all images were then iteratively registered to these initial templates. However, the intermediate steps (reflecting the registration of each individual image to the first, second, third… template from the first step) are only computed internally, iteratively building on the results from the previous step. In order to obtain these intermediate images, the second processing job was therefore split into six successive jobs. The settings used correspond to the defaults and are listed in Table 2. After completing each iteration, the resulting intermediate deformation fields were copied before they were updated in the next iteration. See Fig. 1 for an overview. This ensures that each iteration builds upon the results from the previous step, in line with the original DARTEL approach. These twelve sets of deformation fields (two tissue classes per subject, times six iterations) were then used to write out corresponding sets of increasingly crisp tissue probability maps, six sets for GM and six sets for WM. These twelve sets of 1919 images each were then submitted for data analysis.Table 2DARTEL processing options used for each iteration. The penalizing energy term (linear elastic energy), the number of inner iterations (3), the Levenberg-Marquardt regularization (0.01), the number of cycles for the full multi-grid matrix solver (3) and the number of relaxations in each multi-grid cycle (3) as well as the third regularization parameter id (0.000001) were kept constant over iterations.Table 2Time stepsRegularization parameter [µ]Regularization parameter [λ]Iteration 1141Iteration 2121Iteration 321.5Iteration 44.5.25Iteration 516.25.125Iteration 664.25.125
|
study
| 100.0 |
DARTEL processing options used for each iteration. The penalizing energy term (linear elastic energy), the number of inner iterations (3), the Levenberg-Marquardt regularization (0.01), the number of cycles for the full multi-grid matrix solver (3) and the number of relaxations in each multi-grid cycle (3) as well as the third regularization parameter id (0.000001) were kept constant over iterations.
|
other
| 99.75 |
Image data analysis was performed within the CerebroMatic toolbox which employs a multivariate adaptive regression spline approach as available within the ARESLab toolbox . The data analysis settings were left at the defaults described in . Due to their dominating influence, we used age and gender , , , as well as field strength and data quality as predictors. The latter was here described by the cat12 overall image quality measure, a combined parameter with contributions from spatial resolution, image noise, and image inhomogeneity . Processing each iteration required about 12 hours per tissue class on a current PC workstation. The resulting regression parameters can now be used to generate a set of six increasingly crisp tissue maps (see Fig. 2, Fig. 4 for an illustration of the tissue maps, and Fig. 3, Fig. 5 for an illustration of their respective differences), matched to the demographic and technical details of a population under study, with regard to age (in the range of 13–900 months [1–75 years]), gender (male or female), and field strength (1.5 or 3 T). Tissue quality will automatically be set to “best”. These tissue maps can then serve as appropriately matched targets for spatial normalization within the DARTEL/SHOOT framework even for smaller studies, or studies of “unusual” populations.Fig. 2Top row: Illustration of the conventionally generated DARTEL GM templates, from the whole dataset (n=1919). Bottom row: Illustration of synthetically-generated DARTEL GM templates, generated by the CerebroMatic toolbox based on the here-presented regression parameter set (settings: age=330 months, field strength=3 T, gender=male, data quality=best).Fig. 2Fig. 3Top row: difference image of the conventionally generated and the synthetic DARTEL GM templates (cf. Fig. 2), showing voxels where the intensity difference exceeds 5% (in red) or −5% (in blue). Note overall only minor and decreasing differences. Bottom row: boxplot of all voxelwise differences, with the mean voxelwise intensity difference listed at the bottom (in %).Fig. 3Fig. 4Top row: Illustration of the conventionally generated DARTEL WM templates, from the whole dataset (n=1919). Bottom row: Illustration of synthetically-generated DARTEL WM templates, generated by the CerebroMatic toolbox based on the here-presented regression parameter set (settings: age=330 months, field strength=3 T, gender=male, data quality=best).Fig. 4Fig. 5Top row: difference image of the conventionally generated and the synthetic DARTEL WM templates (cf. Fig. 4), showing voxels where the intensity difference exceeds 5% (in red) or −5% (in blue). Note overall only very minor and decreasing differences. Bottom row: boxplot of all voxelwise differences, with the mean voxelwise intensity difference listed at the bottom (in %).Fig. 5
|
study
| 100.0 |
Top row: Illustration of the conventionally generated DARTEL GM templates, from the whole dataset (n=1919). Bottom row: Illustration of synthetically-generated DARTEL GM templates, generated by the CerebroMatic toolbox based on the here-presented regression parameter set (settings: age=330 months, field strength=3 T, gender=male, data quality=best).
|
study
| 98.94 |
Top row: difference image of the conventionally generated and the synthetic DARTEL GM templates (cf. Fig. 2), showing voxels where the intensity difference exceeds 5% (in red) or −5% (in blue). Note overall only minor and decreasing differences. Bottom row: boxplot of all voxelwise differences, with the mean voxelwise intensity difference listed at the bottom (in %).
|
study
| 100.0 |
Top row: Illustration of the conventionally generated DARTEL WM templates, from the whole dataset (n=1919). Bottom row: Illustration of synthetically-generated DARTEL WM templates, generated by the CerebroMatic toolbox based on the here-presented regression parameter set (settings: age=330 months, field strength=3 T, gender=male, data quality=best).
|
study
| 99.75 |
Top row: difference image of the conventionally generated and the synthetic DARTEL WM templates (cf. Fig. 4), showing voxels where the intensity difference exceeds 5% (in red) or −5% (in blue). Note overall only very minor and decreasing differences. Bottom row: boxplot of all voxelwise differences, with the mean voxelwise intensity difference listed at the bottom (in %).
|
study
| 100.0 |
Theory of Mind (ToM) is traditionally characterized as the ability to represent mental states. Such a characterization leaves little room for studying individual differences in ToM – individuals either can, or cannot, represent mental states – and this binary classification cannot quantify the subtle individual differences observed in typical and atypical populations. In recognition of this problem, attempts have been made to provide a more detailed characterization of the constituent psychological processes which support the representation of mental states (Happé et al., 2017, Schaafsma et al., 2015), and the neurocomputational principles underpinning ToM (Koster-Hale & Saxe, 2013), in order to identify the source of individual differences. A recent model is of interest as it forwards the novel argument that interoception, perception of the internal state of the body, is a fundamental component of ToM (Ondobaka, Kilner, & Friston, 2017). Here we report the first test of the link between interoception and ToM.
|
study
| 99.9 |
Ondobaka, Kilner and Friston's model (Ondobaka et al., 2017) draws on the ‘Predictive Coding’ framework, in which the brain generates hypotheses about the world and tests their predictive validity against incoming sensory evidence. Several models within this framework argue for a role for interoception in emotion understanding (Seth, 2013), but Ondobaka and colleagues (Ondobaka et al., 2017) propose that, as emotional and other interoceptive states (e.g., hunger) constrain hypotheses about an individual's mental states, interoception plays a fundamental role in ToM. Strong and weak versions of this hypothesis can be constructed, where the weak version suggests that emotional and other interoceptive states provide evidence to form or evaluate hypotheses about another's mental state. The strong version of the hypothesis suggests interoceptive information is necessary for the representation of mental states – the defining feature of ToM. We therefore tested whether interoceptive accuracy predicted performance on the representation of mental states in general, or only in those situations where understanding emotion was crucial for accurate mental state representation.
|
study
| 99.94 |
Seventy-two participants completed a well-established measure of interoception in which they counted their heartbeats during intervals of varying duration (Supplemental Experimental Procedures). They were not allowed to monitor their pulse by any means other than “silently concentrating on their heartbeats”. Each participant's heartbeat signals were recorded and, through comparison with their count, interoceptive accuracy was computed [see (Garfinkel, Seth, Barrett, Suzuki, & Critchley, 2015)]. Performance on this task may be influenced by one's ability to estimate time or count, so this was controlled for by measuring participants' ability to estimate time intervals of varying duration (Supplemental Experimental Procedures – Interoception and Time Estimation). Participants completed the Movie for the Assessment of Social Cognition (MASC), a well-validated measure of ToM [(Dziobek et al., 2006); see Supplemental Experimental Procedures], which required them to watch a social event in which accurate mental state inferences are needed to understand the story (Fig. 1A). The video was interspersed with multiple-choice questions probing mental state understanding from which an overall percentage accuracy score was derived. Accuracy was also computed for a set of nonsocial control questions (e.g., “What was the weather like on that evening?”). Most importantly, performance was quantified separately for questions which required representation of another's emotion (e.g., “What is Sandra feeling?”), and for those which did not require the representation of emotional states (e.g., “What is Michael thinking?”).Fig. 1The link between interoception, emotion, and theory of mind. (A) The Movie for the Assessment of Social Cognition (MASC) was administered (Dziobek et al., 2006), in which participants watched a 15-min movie about a social interaction divided into short clips. After viewing each clip, they were presented with a multiple choice question requiring them to infer the mental state of one character. Only one of four answers was correct. Performance was quantified separately for emotional (e.g., “What is Sandra feeling?”) and non-emotional (e.g., “What is Michael thinking?”) questions. (B) Interoceptive accuracy was positively correlated with overall MASC score (rs = .31, P = .008, left panel), driven by a significant association between interoception and emotional items (rs = .41, P < .001, middle panel). However, there was no such association between interoception and the non-emotional items (rs = .03, P = .80, right panel) and the two correlations were significantly different (z = 2.38, P = .017).Fig. 1
|
study
| 100.0 |
The link between interoception, emotion, and theory of mind. (A) The Movie for the Assessment of Social Cognition (MASC) was administered (Dziobek et al., 2006), in which participants watched a 15-min movie about a social interaction divided into short clips. After viewing each clip, they were presented with a multiple choice question requiring them to infer the mental state of one character. Only one of four answers was correct. Performance was quantified separately for emotional (e.g., “What is Sandra feeling?”) and non-emotional (e.g., “What is Michael thinking?”) questions. (B) Interoceptive accuracy was positively correlated with overall MASC score (rs = .31, P = .008, left panel), driven by a significant association between interoception and emotional items (rs = .41, P < .001, middle panel). However, there was no such association between interoception and the non-emotional items (rs = .03, P = .80, right panel) and the two correlations were significantly different (z = 2.38, P = .017).
|
study
| 100.0 |
Greater interoceptive accuracy was associated with overall MASC score (rs = .31, P = .008). Importantly, however, there was only a significant association between interoception and performance on items requiring the representation of another's emotion (rs = .41, P < .001), not where representation of emotional states was not required (rs = .03, P = .80). The size of these correlations was significantly different (z = 2.38, P = .017). This pattern of results (Fig. 1B) was supported by a Bayesian analysis and held after controlling for participants' age, gender, task completion time, time estimation ability and their performance on control questions (Supplemental Tables S1–S5).
|
study
| 100.0 |
Considerable efforts have been made to understand the biological basis of ToM, culminating in a wealth of data. There is also on-going debate about whether human and non-human animals have evolved a domain-specific module to represent mental states, or whether this process may be underpinned by domain-general mechanisms (Heyes, 2014). As long as the psychological and neural mechanisms supporting ToM are still to be determined such debate will continue. Understanding the neurocomputational principles supporting ToM is likely to provide a step-change in our ability to address these issues, and Predictive Coding models suggesting that interoception plays a role in social abilities contribute to this endeavor (Happé et al., 2017, Koster-Hale and Saxe, 2013, Ondobaka et al., 2017, Seth, 2013). The current results suggest that interoception is not necessary for the representation of mental states per se, however it contributes to accurate representation of mental states in situations where this process is reliant upon emotional, or otherwise interoceptive, information. It was also notable that performance on emotional questions (M = 70.53, SD = 11.56) was significantly (t = 7.41, P < .001, d = .06) worse than on non-emotional questions (M = 81.21, SD = 9.76), which may be due to the fact that emotional ToM requires processing of additional interoceptive information.
|
study
| 99.94 |
The current results are supported by evidence that insular cortex, known to be critical for generating interoceptive predictions, is a reliable neural correlate of affective processing (Bernhardt and Singer, 2012, Seth, 2013, Zaki et al., 2012). The findings are also in accordance with recent work showing that alexithymia, a condition characterized by interoceptive atypicalities (Hogeveen et al., 2016, Livingston and Livingston, 2016, Shah et al., 2016a, Shah et al., 2016b), predicted performance on a task requiring emotional understanding but not on a task assessing non-emotional ToM, whereas Autism Spectrum Disorder, which is associated with ToM but not interoceptive deficits, predicted performance on tests of ToM but not emotion understanding (Oakley, Brewer, Bird, & Catmur, 2016). Nonetheless, we suggest that interoceptive training may have a beneficial impact in the real world, where an improved ability to represent the interoceptive/emotional states of oneself and of others is likely to result in more accurate mental state inferences, and benefit emotional understanding more generally.
|
study
| 99.9 |
In sum, this study reports the first empirical test of Predictive Coding models of the contribution of interoception to ToM, and thereby i) speaks to the psychological and computational underpinnings of ToM and ii) provides impetus for future research on the basis of (atypical) ToM and related social abilities.
|
study
| 99.94 |
the incidence of type 2 diabetes (T2D) is now reaching epidemic proportions across the globe, with deaths from the disease reaching 5.1 million and disease complications costing USD 548 billion in 2013 (30). These values are expected to continue to increase, with predictions of a further 205 million sufferers by 2035 (30). T2D is a complex metabolic disease involving hyperglycemia and dyslipidemia, which together conspire to cause serious secondary macro- and microvascular complications including cardiovascular disease, retinopathy, and neuropathy (11, 19). Although it is accepted that a loss of an appropriate balance between functioning pancreatic β-cell mass and insulin action in peripheral tissues leads to abnormal glucose homeostasis, the molecular basis of T2D onset and progression is still poorly understood (31, 57).
|
review
| 99.9 |
While environmental factors such as increasingly sedentary lifestyles and obesogenic diets have a substantial impact, genetic susceptibility also plays a significant role in T2D risk (57). Correspondingly, genome-wide association studies (GWAS) have identified ∼90 loci harboring single nucleotide polymorphisms (SNPs) that confer increased disease risk (13, 21, 23, 40, 57, 65, 86). Such studies have thus led to the discovery of novel genes involved in T2D, such as T cell factor 7-like 2 (TCF7L2) (21) and SLC30A8 (58, 65). Of note, the majority of the GWAS-identified loci affect insulin secretion rather than action, further emphasising the likely role in disease etiology of impaired insulin production.
|
review
| 99.7 |
The VPS13C/C2CD4A/C2CD4B locus was first associated with T2D and glycemic traits in GWAS published in 2010 (6, 15, 22, 29, 60). Subsequent studies identified further SNPs at this genomic location associated with poorer glycemic control and T2D (12, 67, 78). The above studies encompassed a range of distinct populations and age groups, thus providing confidence that SNPs in this locus, acting via either nearby or more remotely located genes, alter genetic susceptibility to T2D. SNPs within the VPS13C/C2CD4A/B locus have been linked to a range of glycemic parameters including higher fasting proinsulin (29, 67), higher 2-h glucose and lower 2-h insulin (60, 67), as well as increased fasting glucose (15, 22, 67) and increased waist circumference (22). Two studies also associated risk alleles with lower glucose-stimulated insulin secretion (GSIS) (6, 22, 67) and others with T2D (12, 67, 83). The “lead” (GWAS index) SNP in this locus, rs7172432, is in LD with a “functional” SNP, rs7163757, previously implicated by fine mapping as the most strongly associated (P = 3 × 10−19) SNP at this locus (61, 66). rs7163757 is located in an islet stretch enhancer (50, 61, 66), again suggesting that the disease-associated SNP acts on the expression of an effector gene(s) to alter diabetes risk.
|
review
| 98.06 |
The first identified member of the highly conserved VPS13 (vacuolar protein sorting 13) family of proteins was Soi1 (or Vps13) in Saccharomyces cerevisiae, where it plays an important role in membrane protein trafficking between the trans-Golgi network (TGN) and the prevacuolar compartment (7). Specifically, Vps13 is involved in trafficking the protease Kex2p, a protein involved in intracellular insulin processing after overexpression of the latter in yeast (85). Subsequently, a role for this protein was demonstrated in prospore formation in S. cerevisiae through the regulation of phosphatidylinositol 4-phosphate [PI(4)P] generation and membrane-bending activity (48, 49).
|
study
| 100.0 |
In both humans and mice, the VPS13 family comprises four members (A–D), with VPS13A and VPS13C showing the most similarity to the yeast homolog (73). All four proteins are large and have potential functions in membrane protein trafficking, Golgi structure, and/or phosphatidylinositol metabolism (37, 47, 53, 62, 63, 73). Mutations in VPS13A and VPS13B cause the genetic diseases chorea-acanthocytosis (ChAc) and Cohen syndrome, respectively (32, 53, 71), and a loss of VPS13C function has recently been linked to early-onset Parkinson's disease (35).
|
study
| 98.25 |
VPS13C is ubiquitously expressed in mammals, with particularly high levels in pancreatic islets and β-cells (60, 67). The observations above have thus led us to hypothesize that VPS13C may play a role in the intracellular trafficking of insulin or other aspects of pancreatic β-cell function. To explore this possibility, we first determined the relationship between the possession of T2D risk alleles in humans and the expression of VPS13C, C2CD4A (C2 calcium-dependent domain 4A), and C2CD4B in human islets. Subsequently, we developed mice inactivated for Vps13c highly selectively in the β-cell by using the recently developed Ins1Cre deleter strain (33, 69). The latter is a knock-in model that avoids the complications associated with earlier insulin 2 promoter-dependent Cre's including recombination in the brain (77) and coexpression of human growth hormone (8). This approach reveals roles for Vps13c in the control of whole body glucose homeostasis, insulin secretion in vivo, and glucose-induced Ca2+ signal generation in the β-cell but suggests that C2CD4A may also contribute to disease risk.
|
study
| 100.0 |
A custom polyclonal antibody against human VPS13C, based on amino acids 1582–1882 of human VPS13C isoform 2A (UniProtKB Q709C8-1; 84% identities, 92% positives with mouse VPS13C protein Q8BX70-1, positions 1580–1879) was raised in rabbits, as recently described (84).
|
study
| 99.94 |
All in vivo procedures were conducted in accordance with UK Home Office regulations [Animals (Scientific Procedures) Act of 1986, Home Office Project License number PPL 70/7349, Dr. Isabelle Leclerc]. Procedures were performed at the Central Biomedical Service at Imperial College, London. Isolation of islets from multiorgan donors was approved by the local ethics committee at the University of Pisa. Human pancreata were collected from brain-dead organ donors after informed consent was obtained in writing from family members. Use of human islets at Imperial College was approved by the local NRES Committee, Fulham; REC reference 07/H0711/114.
|
other
| 99.94 |
Human islet DNA samples obtained from 53 donors (see Supplementary Table 1 online for clinical characteristics), using the DNeasy Blood & Tissue Kit (QIAGEN, Hilden, Germany) as previously described (51), were genotyped for SNPs rs4502156, rs7172432, and rs7163757. The rs7172432 locus was amplified by semi-nested PCR using primers TAG GTA TCT TGG AGC TGA GG and CCA CAC TTC ACA GAA TCA GG for the first round amplification and then CAG GTC AAG TGA GCA CTT GC and CCA CAC TTC ACA GAA TCA GG for the second round. The amplicons were then digested with SspI and genotyped based on the resulting restriction fragment length polymorphism.
|
study
| 100.0 |
Islet RNA was isolated from hand-picked islets as described (39), using the Arcturus PicoPure RNA Isolation Kit (Applied Biosystems, Foster City, CA), according to the procedure recommended by the manufacturer for RNA extraction from cell pellets and was accordingly treated with DNase to remove the contamination with genomic DNA. Reverse transcription to cDNA was performed using a High Capacity cDNA Reverse Transcription Kit (Themofisher). The rs4502156 and rs7163757 SNPs were genotyped by qPCR using a commercial TaqMan assay (Applied Biosystems). VPS13C, C2CD4A, and C2CD4B mRNA abundances were measured relative to ACTB in corresponding RNA samples by qRT-PCR using commercial TaqMan assays (Applied Biosystems) and the ΔCT method. As a quality control step, samples with ΔCT standard deviation > 0.2 were excluded from the analysis. The association between VPS13C expression and genotype was tested using an ANCOVA model, controlling for age, sex, and BMI and implemented in R (52). The association of genotype with C2CD4A and C2CD4B was analyzed in the same manner. Linkage disequilibrium (LD) values for SNPs in the Tuscan population used here were obtained at: http://www.1000genomes.org/faq/which-populations-are-part-your-study.
|
study
| 100.0 |
To assess whether variants at rs7163757 might cause changes in the expression of nearby genes, two reporter constructs were generated. A 1.3-kb fragment of the genomic region flanking the SNP was amplified by PCR from a heterozygous donor by using Phusion High Fidelity DNA Polymerase (Thermo Scientific, Paisley, UK). The PCR product was subsequently cloned into CR™8/GW/TOPO (Thermofisher, Paisley, UK) according to the manufacturer's instructions. Plasmid DNA from clones was purified using a GenElute Plasmid Miniprep Kit (Sigma, Dorset, UK) and sent for sequencing to identify clones containing one of each allele. DNA fragments were then shuttled into the minimal promoter (DNA sequence: TAG AGG GTA TAT AAT GGA AGC TCG ACT TCC AG, containing a TATA box promoter element)-driven luciferase vector GL4.23-GW vector (76) using the Gateway LR Clonase II Enzyme Mix (Invitrogen, Paisley, UK). pGL4.23-GW is modified from pGL4.32 (Promega) with Gateway technology (Thermofisher) and has previously been used successfully for the analysis of enhancer activity (20).
|
study
| 100.0 |
The sequence and orientation of the insert was checked by restriction enzyme digest, and subsequently a QIAGEN Plasmid Maxi Kit (QIAGEN, Manchester, UK) was used to purify transfection grade DNA. HEK293, MIN6 (44), 1.1B4 (41), and EndoC-βH1 (54) cells were transfected using Lipofectamine 2000 (Invitrogen, Paisley, UK) in 48-well plates using 250 ng of each reporter construct and 1 ng of Renilla control vector. Each condition was repeated in six separate wells. Dual-Luciferase Reporter Assay (Promega, Southampton, UK) was used to measure Luciferase normalized against Renilla. All experiments were done in triplicate. The following cloning primers were used: CCA ACA AAT AGT AAG CAT TAT TAC C (rs7163757, forward) and CAA ATA GTT GTA GAT ATG TGG CAT T (rs7163757, reverse).
|
study
| 100.0 |
Generation of heterozygous embryos on a C57/BL6 background, carrying floxed alleles of Vps13c was conducted by GenOway (France). Vps13cfl/fl mice were crossed to mice expressing Cre recombinase under the control of the Ins1 promoter (33, 69) to generate mice in which exon 1 of the Vps13c gene was selectively excised in pancreatic β-cells. Mice were born at the expected Mendelian ratios without any obvious physical or behavioral defects. Mice were housed two to five per cage in a pathogen-free facility under a 12:12-h light-dark cycle and had ad libitum access to water and standard mouse chow diet (Research Diet, New Brunswick, NJ). High-fat diet (HFD, 60% wt/wt fat content; Research Diet) was introduced at 4 wk of age.
|
study
| 100.0 |
Genotyping was performed from ear biopsies using PCR. Knockout of Vps13c from pancreatic islets was assessed using both qPCR and immunoblotting, as described below. Mice were weighed weekly from 5 wk of age, and random, fed glycemia was tested fortnightly in the afternoon.
|
study
| 99.94 |
Mice were fasted for 15–16 h overnight prior to intraperitoneal (IPGTT) and oral glucose tolerance tests (OGTT) with free access to water. Blood samples were taken for glycemia measurement via venesection of the tail vein. Glycemia was measured using an Accu-Chek glucometer (Roche Diabetes Care, UK) and appropriate measurement strips. Fasting glycemia was first measured (time 0), and then glucose was administered via ip injection (1 g/kg body wt) or oral gavage (1.5 g/kg body wt). Glycemia measurements were then taken by injection at 15, 30, 45, 60, 90, and 120 min.
|
study
| 100.0 |
Mice were fasted overnight with free access to water. A fasting (time 0) blood sample (∼50 μl) was collected from the tail vein into a lithium-heparin-lined Microvette (Starstedt, Leicester, UK) before administering of glucose (3 g/kg body wt) via ip injection. Blood samples were then collected at 15 and 30 min after injection. Glycemia was also measured at these time points. Plasma was collected by centrifuging samples at 2,000 g for 10 min at 4°C. Plasma insulin was measured using an ultrasensitive mouse insulin ELISA (Crystal Chem, IL). For random-fed insulin/proinsulin ratio measurements, a blood sample was collected into a lithium-heparin-lined Microvette from the tail vein and the aorta immediately after culling via cervical dislocation. Samples were kept on ice at all times to prevent degradation of proinsulin, and plasma was collected as described above. Insulin was measured as described above, and proinsulin was measured using a Rat/Mouse Proinsulin ELISA (Mercodia, Uppsala, Sweden).
|
study
| 100.0 |
Mice were culled by cervical dislocation. Islets were isolated after pancreatic distension with collagenase essentially as previously described (55). Islets were allowed to recover from digestion for 24 h (RC-fed mice) or 48 h (HFD-fed mice) in RPMI medium (GIBCO) supplemented with 10% (vol/vol) fetal bovine serum, 1% (wt/vol) penicillin, 1% (wt/vol) streptomycin, 11.1 mM glucose, and 2 mM l-glutamine. Insulin secretion was measured from duplicate batches of 10 islets incubated in 0.5 ml of Krebs-Ringer medium [130 mM NaCl, 3.6 mM KCl, 1.5 mM CaCl2, 0.5 mM MgSO4, 0.5 mM NaH2PO4, 2 mM NaHCO3, 10 mM HEPES, and 0.1% (wt/vol) BSA, pH 7.4] containing 3 or 16.7 mM glucose or 20 mM KCl and 3 mM glucose as indicated, and shaking at 37°C for 30 min. Total insulin was extracted into 0.5 ml of acidified ethanol [75% (vol/vol) ethanol, 1.5% (vol/vol) 1 M HCl and 0.1% (vol/vol) Triton X-100]. For continuous measurements of secretion, insulin samples from 50 perifused islets were collected using a custom-built device and a perifusion rate of 500 μl/min at 37°C as described previously (10). Secreted and total insulin concentrations were measured using a homogeneous time-resolved fluorescence-based (HTRF) insulin assay (CisBio, Codolet, France) in a PHERAstar reader (BMG Labtech), according to the manufacturer's instructions.
|
study
| 100.0 |
Isolated pancreata were fixed in 10% (vol/vol) formalin overnight at 4°C and embedded in paraffin wax. Sections (5 μm) were cut and fixed onto Superfrost slides. For staining, five sections per mouse, 25 μm apart, were incubated in Histochoice Clearing Agent and then submerged consecutively in 100, 95, and 70% ethanol to remove the wax. Following washes with water, sections were permeabilized by boiling in a citrate-based antigen unmasking solution (Vector Labs, Peterborough, UK), washed with PBS, and then incubated in PBS-Triton X-100 [PBST, 0.1% (vol/vol)] containing 2% (wt/vol) BSA and 2% (vol/vol) goat and donkey serum for 2 h at room temperature. Sections were then incubated in a humidified chamber at 4°C overnight with guinea pig anti-insulin (1:200; Dako, Ely, UK) and mouse anti-glucagon (1:1,000; Sigma, Dorset, UK). After washing three times in PBST [0.25% (vol/vol)] containing 0.25% (wt/vol) BSA, sections were incubated with Alexa fluor 488 and 568-conjugated secondary antibodies (1:1,000; Invitrogen, Paisley, UK) for 2 h at room temperature in the dark. Sections were then mounted using Vectashield antifade mounting medium containing DAPI (Vector Labs). Slices were imaged in the Imperial College facility for imaging by light microscopy (FILM) (http://www3.imperial.ac.uk/imagingfacility), using a Zeiss Axio Observer inverted widefield microscope with LED illumination. Images were captured with a Hamamatsu Flash 4.0 fast camera controlled by Zen software (Zeiss, Cambridge, UK). Image analysis was conducted using ImageJ software (1) and an in-house macro as described under supplementary methods (see online).
|
study
| 99.9 |
Total RNA was extracted from 50–200 islets isolated from three control and three βVps13cKO mice (for both males and females) using TRIzol (ThermoFisher Scientific) according to the manufacturer's instructions. RNA (100 ng) was reverse transcribed to produce cDNA by using the High Capacity Reverse Transcription Kit (Life Technologies, Paisley, UK) with random primers. qPCR was conducted using SYBER Green (life Technologies, Paisley, UK) on an ABI-Fast Prism 7500 machine and primers specific to murine Vps13c, C2cd4a, C2cd4b, or cyclophilin A. Primers were designed using Primer Express 3.0 (Applied Biosystems, CA), and sequences used were: Vps13c forward CAC AAG CAT TGA AGA TAG AAG CAA AA, reverse AGT GAT GGC ACA ATG TCT TGT TG; C2cd4a forward CGG GTT GGA AAA CCA TCT GA, reverse GTC TGA ACC CTG TGA TCC TGT TC; C2cd4b forward ACG TCA CCT GCT TCG TTC CT, reverse CAC GAG CGT CTT TTC TTC TTC A; cyclophilin A forward TAT CTG CAC TGC CAA GAC TG A, reverse CCA CAA TGC TCA TGC CTT CTT TCA. Whereas the VPS13C and C2CD4B primers spanned exon/exon junctions, the C2CD4A primers spanned intron 1.
|
study
| 100.0 |
Total protein was extracted from 50–500 islets isolated from two or three control or β-cell-specific VPS13C knockout (βVps13cKO) mice (males and females) in ice-cold RIPA buffer [1% (vol/vol) Triton X-100, 1% (wt/vol) sodium deoxycholate, 0.1% (wt/vol) SDS, 0.15 M NaCl, 50 mM Tris, pH 8.0] containing a 2× concentration of Complete, EDTA-free protease inhibitor cocktail (Roche, Burgess Hill, UK). The samples were incubated in RIPA on ice for 10 min and then freeze-thawed twice to ensure release of proteins. Samples were clarified by centrifuging at 16,000 g for 10 min at 4°C, and then total protein content was quantified using a BCA protein assay kit (Pierce, ThermoScientific). Total protein (5 μg) was added to SDS sample buffer [0.5 M Tris·HCl, pH 6.8, 2% (wt/vol) SDS, 5% (wt/vol) glycerol, 0.6 M DTT, and 0.2 mg bromophenol blue] and incubated at room temperature for 30 min. Samples were then electrophoresed on a 4–10% discontinuous gradient gel alongside a HiMark Protein Standard (Novex; ThermoScientific) and transferred onto a nitrocellulose membrane overnight. Membranes were blocked with 5% milk and incubated with primary antibodies (rabbit anti-VPS13C, 1:2,000, described above, or goat anti-EEA1, Santa Cruz, TX, 1:2,000) overnight with agitation at 4°C. Membranes were then washed three times in PBS-Tween 20 (0.2% vol/vol) and incubated with horseradish peroxidase-conjugated antibodies for 1 h at room temperature. Following three washes in PBS-Tween 20, proteins were visualized with ECL reagent and X-ray film (Amersham, GE Healthcare Life Sciences).
|
study
| 100.0 |
Human islets were dissociated by 10-min incubation in Hanks'-based enzyme-free cell dissociation buffer (GIBCO, Invitrogen) and gentle pipetting to generate small clusters of cells. Dissociated cells were plated onto 13/24-mm sterile coverslips and allowed to recover for 1–2 days. Cells were fixed in 4% paraformaldehyde and permeabilized in 0.1% Triton X-100. Primary cells were blocked in 10% fetal calf serum and subsequently incubated overnight with VPS13C 15-E antibody (Santa Cruz sc-104751, 1:50) with or without anti-insulin antibody (1:200; Dako, Ely, UK) followed by incubations with Alexa 488 and Alexa 568-conjugated secondary antibodies in sequential order. Coverslips were mounted using VectaShield with DAPI and imaged as described elsewhere (42). Samples were illuminated using steady-state 488- and 560-nm laser lines, and emission was collected through ET535/30 and ET620/60 emission filters (Chroma). Images were captured using a Hamamatsu EM CCD digital camera controlled by an Improvision/Nokigawa spinning disc system running Volocit (PerkinElmer, MA) software.
|
study
| 99.94 |
Islets were isolated as described above. Simultaneous imaging of Ca2+ of individual cells was performed by spinning disc confocal microscopy after loading intact islets with Fluo 2-AM (Cambridge Bioscience, Cambridge, UK). Images were captured with a Zeiss Axiovert 200M microscope fitted with a ×10 0.3–0.5 NA, EC Plan Neofluar, Zeiss objective and a ×1.5 Optivar attached to a Nokigawa spinning disc confocal head, as described (27). The microscope was controlled using Volocity software. Islets were continuously perifused in Krebs-Ringer buffer containing 3 mM glucose, equilibrated with 95% O2-5% CO2 at 34–36°C. Islets were stimulated at 210 s and 1,300 s by perifusion with Krebs-Ringer supplemented with up to 16.7 mM glucose or 20 mM KCl as indicated. Offline processing and analysis were conducted using ImageJ software (1) and an in-house macro as described under supplementary methods.
|
study
| 100.0 |
Data were analyzed using Microsoft Excel, GraphPad PRISM 6.0, and R. Significance was tested using an unpaired Student's two-tailed t-test with appropriate posttests for multiple comparisons, or two-way ANOVA, as indicated. P < 0.05 was considered significant, and errors signify means ± SE unless otherwise stated. Figures were constructed using Adobe Illustrator.
|
other
| 99.44 |
GWAS studies have implicated SNPs close to VPS13C, C2CD4A, and C2CD4B in altered T2D susceptibility. We tested the association between genotype at one of the previously identified SNPs rs4502156 and the likely causal SNP rs7163757 (61, 66) (r2 = 0.939, D′ = 0.979 with rs4502156) and VPS13C expression in human islet samples from 53 donors. Initial analysis for rs4502156 and rs7163757 including all samples showed no significant association. Interaction plots indicated a possible interaction between sex and genotype, which was tested by including the interaction term in the ANCOVA model (see materials and methods). This was found to be significant (P = 0.015, n = 53), so data were stratified by sex, and subsequently males and females were analyzed separately (Fig. 1, A–C). Analysis of females revealed a significant association between possession of the risk allele (C) at rs7163757 and lowered VPS13C expression (P = 0.041, n = 40; Fig. 1C). A similar sex interaction (P = 0.016, n = 53) was also observed for rs4502156, and likewise a significant association was detected between genotype at this locus and expression of VPS13C in females (P = 0.043, n = 40). An association was also detected between rs4502156 (not shown), as well as rs7163757 (P = 0.011, n = 40; Fig. 1, D–F), with C2CD4A mRNA levels in female donors but not with C2CD4B (Fig. 1, G–I). Subsequent functional studies in the present report focussed upon VPS13C.
|
study
| 100.0 |
eQTL (expression quantitative trait loci) analysis. Expression of VPS13C (A–C), C2CD4A (D–F), and C2CD4B (G–I) was quantified relative to ACTB in 53 human donor islet samples and compared with the genotype at rs7163757. ΔCT is plotted against genotype for all samples (A, D, G; n = 53) or just samples from male (B, E, H; n = 13) or female (C, F, I; n = 40) donors, along with the mean and standard error. Since higher ΔCT corresponds to lower expression, possession of the risk allele (C) is significantly associated with lower VPS13C expression in samples from female donors (P = 0.041).
|
study
| 100.0 |
To determine whether and how the possession of risk alleles at the VPS13C locus might affect the expression of nearby or remotely located genes, we used reporter-luciferase assays in non-β-cells (HEK293) and in β-cell lines from mice (MIN6) and humans (1.1B4 and EndoCβH1). As shown in Fig. 2, inclusion of the risk (C) allele at the previously implicated causal SNP rs7163757 significantly lowered the enhancer/promoter activity of reporter constructs bearing this variant vs. the presence of the protective (T) allele in HEK293, MIN6, and 1.1B4 cells. A similar tendency (P < 0.1) was observed in EndoCβH1 cells (Fig. 2). These data are thus consistent with an enhancer function for this region, whose activity is lowered in carriers of risk alleles.
|
study
| 100.0 |
Comparison of promoter/enhancer activities of variants at rs7163757 in the VPS13C locus. Luciferase reporter assay performed in 4 cell lines (HEK293, MIN6, 1.1B4, and EndoC-βH1). The risk single nucleotide polymorphisms (SNP) caused a significant reduction in enhancer activity in HEK293, MIN6, and 1.1B4 (*P < 0.05, **P < 0.01 calculated using ratio paired Student's t-tests). Error bars represent SE from either 3 (HEK293 and MIN6) or 4 (1.1B4 and EndoC-βH1) independent experiments. P = 0.1 for the effect of the risk allele in EndoC-βH1 cells.
|
study
| 100.0 |
The observations above suggested that risk variants at the VPS13C locus may decrease the expression of nearby genes. To explore the potential impact of lowered VPS13C levels on insulin secretion, deletion of exon 1 (Fig. 3A) of the Vps13c gene was achieved throughout the pancreatic β-cell compartment in C57BL/6 mice from ∼E11.5 using the highly selective Ins1Cre deleter strain (69). As shown in Fig. 3, B and C, VPS13C was barely detected in islets isolated from βVps13cKO mice (Fig. 3B and Ci) and levels of Vps13c mRNA were significantly reduced (Fig. 3, Bii and Cii) by >80%. These findings are fully consistent with efficient (>94%) and exclusive (69) recombination in β-cells, which comprise 60–80% of the rodent islet (16), given that Vps13c mRNA is about twofold more abundant in β- than in α-cells (5), which comprise the majority of the islet non-β-cells. Expression of C2cd4a and C2cd4b in islets was variable between mice but was unaffected by Vps13c deletion. Changes in body weight gain (Fig. 3, D and E) and random-fed glycemia (Fig. 3, F and G) over time were not different between control (Ctrl) and βVps13cKO (KO) mice irrespective of sex or diet [regular chow (RC) vs. high-fat diet (HFD); see materials and methods for details].
|
study
| 100.0 |
Generation of VPS13Cfl/fl::Ins1.Cre+/− (βVps13cKO) mice. A: LoxP sites were inserted on either side of exon 1 to enable Cre-mediated inactivation of the Vps13c gene in pancreatic β-cells after breeding to Ins1.Cre mice. The resultant colony consisted of VPS13C-null mice (KO, βVps13cKO) and control mice (Ctrl) at the expected 50:50 ratio. B and C: islets were isolated from 2–3 male (B) and 3 female (C) Ctrl and KO mice for (i) immunoblotting or (ii) qPCR analysis. Both ΔCT (relative to cyclophilin A) and log2-transformed fold changes, normalized to control mice, are shown. Error bars represent standard deviation in (ii) top and 95% confidence intervals in (ii) bottom. *P < 0.05, **P < 0.01 analyzed with 2-way ANOVA with Sidak's multiple corrections. D–G: changes in weight (D and E) and random-fed glycemia (F and G) over time for Ctrl (black) and KO (dashed) mice fed regular chow (RC, circles or squares) or high-fat diet (HFD, triangles). Inset: area under the curve (AUC) analysis for female mice on HFD, assessed for significance using an unpaired Student's t-test; n = 11–15 mice, as indicated.
|
study
| 100.0 |
Examined in male mice, intraperitoneal glucose tolerance (IPGTT) was not different between control and βVps13cKO animals up to the age of 16 wk, whereas βVps13cKO mice became glucose intolerant at 20 wk of age (Fig. 4, A, C, E, G). Although glucose tolerance was lower at all ages examined compared with animals maintained on RC, no differences were observed between control and βVps13cKO males maintained for up to 16 wk (i.e., 20 wk old) on a HFD (Fig. 4, B, D, F, H).
|
study
| 100.0 |
Glucose tolerance in male βVps13cKO mice. A–H: intraperitoneal glucose tolerance (1 g/kg body wt) was measured in Ctrl (solid black line) and KO (dashed line) male littermates fed either RC (A, C, E, G) or HFD (B, D, F, H). IPGTTs were conducted at 8 (A, B), 12 (C, D), 16 (E, F), and 20 (G, H) wk. Inset: AUC. Numbers of animals (n) for each experiment are given in AUC bars. *P < 0.05, **P < 0.01, 2-way ANOVA with Fisher's LSD post hoc test (main graphs) or unpaired Student's t-test (AUC, insets).
|
study
| 100.0 |
By contrast, when maintained on RC, female mice (Fig. 5) displayed abnormal IPGTT at 12 wk of age (Fig. 5C). This resolved at 16 wk but was again apparent at 20 wk (Fig. 5G). Consistent with observations in males (Fig. 4), female βVps13cKO mice fed HFD similarly failed to show abnormalities in IPGTT up to 20 wk of age (Fig. 5, B, D, F, H).
|
study
| 100.0 |
Glucose tolerance in female βVps13cKO mice. A–H: intraperitoneal glucose tolerance (1 g/kg body wt) was measured in Ctrl (solid black line) and KO (dotted line) female littermates fed RC (A, C, E, G) or HFD (B, D, F, H). IPGTTs were conducted at 8 (A, B), 12 (C, D), 16 (E, F), and 20 (G, H) wk. Inset: AUC. Numbers of animals (n) for each experiment are given in the AUC bars. *P < 0.05, 2-way ANOVA with Fisher's LSD post hoc test (main graphs) or unpaired Student's t-test (AUC, insets).
|
study
| 100.0 |
Oral glucose tolerance in βVps13cKO and control mice. A and B: oral glucose tolerance (1.5 g/kg body wt) was measured in Ctrl (solid black line) and KO [dashed (male) or dotted (female) lines] littermates fed RC. OGTTs were conducted at ages indicated. Inset: AUC; n, numbers for each experiment are given in AUC bars. *P < 0.05, 2-way ANOVA with Fisher's LSD post hoc test.
|
study
| 100.0 |
Examined in mice aged 19–21 wk, glucose-induced excursions in plasma insulin were not different between βVps13cKO and control male and female mice (Fig. 7, A–D). Likewise, by analyzing fasting glucose and insulin levels, we observed no indication of a change in steady-state β-cell function (36) as assessed using homeostatic model assessment (HOMA2-%B; Fig. 7E), nor insulin in insulin sensitivity (HOMA2-%S; Fig. 7F) in males maintained on either RC or HFD. By contrast, a tendency toward a lower HOMA2-%B value (Fig. 7G), accompanied by a significant increase in HOMA2-%S, was apparent in female βVps13cKO mice vs. controls fed on RC, whereas these differences were not observed on HFD (Fig. 7, G and H).
|
study
| 100.0 |
Effect of Vps13C deletion on glucose-stimulated insulin secretion (GSIS) in vivo. A–D: plasma insulin concentration was measured following intraperitoneal administration of glucose (3 g/kg body wt) in Ctrl (solid black line) and KO [dashed (males) or dotted lines (females)] littermates. Blood was sampled for insulin measurements when mice were 21 or 19 wk old (RC or HFD, respectively). Inset, top: respective glycemia measurements. Inset, bottom: AUC calculated from the main graph, measuring total released plasma insulin; n = 6–9 mice per genotype, as detailed in the key. E–H: homeostatic model assessment analysis (HOMA2)-%B (E and G) and -%S (F and H) analysis using fasting glycemia values and corresponding plasma insulin concentrations, respectively. **P < 0.01, unpaired Student's t-test with Welch's correction (E–H).
|
study
| 100.0 |
Impairments in glucose tolerance and a tendency toward impaired β-cell function apparent in vivo in female βVps13cKO mice might reflect abnormal glucose- or depolarization-dependent insulin secretion from β-cells. To investigate this, we studied insulin release from batches of islets from mice 20–23 wk old, as shown in Fig. 8. Interestingly, both glucose (16.7 mM) and KCl (20 mM) -stimulated secretion tended to increase in βVps13cKO vs. control islets from males fed either RC (Fig. 8A) or HFD (Fig. 8B). Whereas a similar tendency was also apparent for islets from females maintained on a HFD (Fig. 8D), those from female βVps13cKO mice fed RC showed no change in insulin secretion vs. controls (Fig. 8C) when stimulated with 20 mM KCl. No differences between HFD-fed control and βVps13cKO mice were seen when the same experiment was conducted under perifusion (Fig. 8E).
|
study
| 100.0 |
Effect of Vps13c deletion on GSIS in vitro. A–D: insulin secretion from isolated islets from KO and Ctrl mice over 20 wk old maintained on RC (A and C) or HFD (B and D) was assessed by incubating 10 size-matched islets in Krebs-Ringer solution containing 3 mM glucose (3 Glu), 16.7 mM glucose (16.7 Glu), or 20 mM KCl for 30 min and measuring the amount of insulin secreted (see materials and methods). Islets were lysed to measure total insulin; results are presented as %total insulin. E: insulin secretion from islets continuously perifused with Krebs-Ringer solution containing 3 mM glucose and then stimulated with 16.7 mM glucose; n = 3–5 mice per genotype, as indicated. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, 2-way ANOVA with Sidak or Tukey's post hoc test where appropriate.
|
study
| 100.0 |
One explanation for the differences in glucose tolerance and insulin secretion seen in βVps13cKO mice may be an alteration in β-cell mass. To establish whether this was the case, we conducted immunohistochemical analyses on pancreatic sections from βVps13cKO and control mice fed RC and aged over 20 wk. Using antibodies against either insulin or glucagon, we observed no differences in %β- or α-cell surface normalized pancreatic surface (Fig. 9, A and B, females; C and D, males).
|
study
| 100.0 |
β-Cell mass in βVps13cKO mice. A and C: representative images from pancreatic slices from female (A) and male (B) Ctrl and KO mice (20–23 wk of age) fed a RC diet. Slices were stained with antibodies against insulin (green) and glucagon (red). Nuclei were stained with DAPI; scale bar represents 50 μm. B and D: percentage of β- (i) and α-cell (ii) surface area, normalized to whole pancreas surface area; (iii): β/α-cell ratio. Data are from n = 3 Ctrl and 3 KO females and 3 Ctrl and 5 KO males. No significant differences between genotypes were detected.
|
study
| 100.0 |
Alterations in glucose tolerance and tendency toward impaired insulin secretion, which were apparent in vivo, may reflect altered signal generation by glucose. We next used the fluorescent intracellular probe Fluo 2 (27) to monitor intracellular free Ca2+ dynamics in β-cells in situ within the intact islet (Figs. 10 and 11). Under the conditions used, glucose-induced changes in free Ca2+ were largely restricted to the β-cell population (10, 27). No genotype-dependent differences in the peaks of the Ca2+ response to either high glucose or KCl were apparent in islets from male mice (Fig. 10), although islets from male βVps13cKO mice fed RC did display a significantly delayed response to high glucose stimulation (Fig. 10A, ii and v). Increases in free Ca2+ in islets from female mice were respectively augmented (Fig. 11A, i and vi) and reduced (Fig. 11B, i and vi) in high-glucose-stimulated islets from βVps13cKO animals fed RC or HFD. A similar trend was seen after depolarization with KCl (Figs. 11A, i and viii, and B, i, iv, and viii). As was the case for male mice, the response to glucose in islets from HFD-fed female mice was slightly delayed, with those from RC mice showing no significant difference in the time of the glucose peak (Fig. 11A, i and v, and B, i and v).
|
study
| 100.0 |
Effect of Vps13c deletion on calcium signaling in vitro in male mouse islets. Isolated islets from male mice (20–23 wk of age), maintained on RC (A) or HFD (B) were loaded with Fluo 2 and incubated in Krebs-Ringer solution containing 3 mM glucose (3 mM Glu) for 45 min. Dye-loaded islets (3–7 per field of view) were imaged on a spinning disk confocal microscope for 2 min in 3 mM glucose, as described in materials and methods. A perifusion system was used to allow subsequent imaging of the islets in 16.7 mM Glu for 18 min, followed by 20 mM KCl for 5 min. Individual traces from each islet were then averaged to give one trace per islet, which was then pooled with the other islets. (i), mean free Ca2+ (normalized to initial fluorescence; F/F0); (ii), inset from (i), mean free Ca2+ measured between 300 and 500 s, showing the effect of stimulation with 16.7 mM glucose; (iii), AUC analysis for high glucose stimulation; (iv), AUC analysis for KCl stimulation; (v), time to maximum peak value from stimulation with glucose; (vi), maximum peak value (F/F0) from stimulation with glucose; (vii), time to maximum peak value from stimulation with KCl; (viii), maximum peak value (F/F0) from stimulation with KCl; n = 3–5 mice per genotype. Number of islets used: n = male RC 31–38 islets from 3 mice; male HFD mice n = 41–46 islets from 4 mice. *P < 0.5, **P < 0.01, ***P < 0.001, unpaired Student's t-test.
|
study
| 100.0 |
Effect of Vps13c deletion on calcium signaling in vitro in female mouse islets. Isolated islets from female mice (20–23 wk of age) maintained on RC (A) or HFD (B) were loaded with Fluo 2 and analyzed as for male islets (Fig. 10). (i), mean free calcium (F/F0); (ii), inset from (i), mean free calcium measured between 300 and 500 s, showing stimulation with 16.7 mM glucose; (iii), AUC analysis for high glucose stimulation; (iv), AUC analysis for KCl stimulation; (v), time to maximum peak value from stimulation with glucose; (vi), maximum peak value (F/F0) from stimulation with glucose; (vii), time to maximum peak value from stimulation with KCl; (viii), maximum peak value (F/F0) from stimulation with KCl; n = 3–5 mice per genotype. Number of islets used: female RC, n = 47–48 islets from 4 mice; female HFD, n = 46–59 islets from 4 or 5 mice (KO vs. Ctrl, respectively). *P < 0.5; **P < 0.01; ***P < 0.001; unpaired Student's t-test.
|
study
| 100.0 |
To determine whether VPS13C might conceivably affect the properties (i.e., “fusogenecity”), or the distribution of secretory granules, we explored the localization of the protein with single human β-cells by confocal immunocytochemistry (Fig. 12). Close colocalization was observed between insulin and VPS13C-labeled structures, indicative of the presence of the latter on the limiting membrane of insulin-containing dense core granules.
|
study
| 100.0 |
Previous studies (17) have revealed that VPS13C expression in human islets is associated with HbA1c levels by massive parallel sequencing (RNA-seq) and microarray analysis at both nominal and permutation P values (P < 0.05), with lower mRNA levels observed in T2D subjects. We extend these findings here by showing that VPS13C mRNA levels were lower in carriers of risk alleles at rs4502156 and rs7163757 in female, but not male subjects. We note that in the present study a lower number of male vs. female samples may have limited our power to detect changes in the former. However, and arguing against this possibility, no tendency toward lowered VPS13C or C2CD4A expression with the interrogated SNPs was observed in males: rather, the trend was toward increased expression with risk alleles.
|
study
| 100.0 |
Using mouse genetics we provide evidence that VPS13C plays a role in the control of pancreatic β-cell function. It should be emphasized that the impact of deleting this gene selectively in the β-cell was relatively mild and indeed was not apparent in males until 20 wk of age. Evidence for deficiencies in β-cell function were, however, more apparent in females from an earlier age, in line with the human eQTL data. These included the transient appearance of glucose intolerance at 12 wk and its reemergence at 20 wk. Interestingly, the same phenomenon is also observed in a monogenic form of diabetes resulting from misexpression of the ZAC gene, termed transient neonatal diabetes mellitus (TNDM), and is apparent in mouse models with this disease (albeit in younger animals than observed here) (38). While the reasons for this transience are not known either in TNDM or in the case of Vps13c deletion, dynamic changes in the balance between islet function and insulin sensitivity may provide one explanation. Similarly, the emergence of glucose intolerance with age in βVps13cKO mice, which is reminiscent of changes seen after the inactivation of the T2D GWAS gene Tcf7l2 in mice (43, 82), seems to reflect, at least in part, increasing insulin resistance as well as impaired insulin output from the pancreas. Of note, recent studies report relatively preserved glucose sensing of isolated islets with age in both mice and humans (3, 26) but suggest a role for altered vascularization and fibrosis in impaired insulin secretion in vivo (3).
|
study
| 99.94 |
Strikingly, the impairments in glucose tolerance apparent in both male and female βVps13cKO mice vs. littermate controls at this age were abolished after maintenance on HFD. These findings demonstrate an interesting interaction between the inheritance of a genetic factor influencing risk, and age [as observed in human T2D, (87)] as well as sex and diet. The reasons for the difference in penetrance between the effects of Vps13c deletion observed here between male and female mice remain unknown but may reflect interactions with sex hormones at the level of the individual β-cell (46) or, alternatively, subtle differences in insulin sensitivity between the sexes that go on to influence the effect of perturbations in the β-cell on overall glucose homeostasis.
|
study
| 100.0 |
Examined in either males or females, β-cell mass was not different between control and knockout mice, indicating a possible defect in β-cell function as underlying the glucose dyshomeostasis reported above. Correspondingly, clear tendencies were apparent toward impaired β-cell function and lowered insulin levels when one combined fasting glycemia with corresponding insulin plasma concentration using HOMA2 analysis (particularly in females; Fig. 7, E and G). According to this analysis, insulin sensitivity was slightly but significantly increased in knockout females vs. littermate controls (Fig. 7H), again indicating that a defect in β-cell function is likely to underlie the mild glucose intolerance in knockout mice (and subject to caveats in extrapolating HOMA2 models from humans to rodents) (74).
|
study
| 100.0 |
On the other hand, we were unable to detect any impairment in glucose or depolarization-induced insulin secretion as assessed ex vivo in isolated islets (Fig. 8). Indeed, in islets isolated from animals maintained on either RC or HFD, we observed a tendency in male βVps13cKO mice toward enhanced insulin secretion in response to either high glucose or KCl and in female βVps13cKO in response to high glucose. By contrast, stimulated insulin secretion in response to KCl tended not to change in βVps13cKO vs. control islets from females fed RC or HFD. We are therefore unable at the present time to assign the changes in β-cell function and glucose homeostasis observed in vivo unambiguously to alterations in islet responses measureable in vitro. We would stress, however, that the mechanisms responsible for the stimulation of insulin secretion by elevated glucose in vivo, which are likely to be modulated by a multitude of humoral (e.g., circulating fatty acids, incretins, adipokines, etc.), neuronal (56), and other inputs into the islet, are unlikely to match perfectly those tested in vitro.
|
study
| 100.0 |
Nonetheless, detailed analysis of glucose- and KCl-induced Ca2+ dynamics did provide evidence for alterations at the level of secretory granule behavior, which may play a role to impair insulin secretion in vivo. Importantly, islets from male βVps13cKO mice maintained on RC or on HFD responded normally with insulin secretion in response to either glucose or high KCl (Fig. 8), consistent with mild, and late-onset, glucose intolerance in these animals. Glucose-induced Ca2+ increases were nonetheless significantly delayed in the KO animals (Fig. 10). By contrast, when fed on RC, islets from female βVps13cKO mice displayed a significant enhancement in glucose-stimulated Ca2+ increases vs. islets from control littermates (Fig. 11A), whereas GSIS was unaltered (and tended to be decreased in response to KCl), as mentioned above. These observations suggest that the Ca2+ responsiveness of the secretory machinery to intracellular Ca2+ increases may be diminished in female βVps13cKO mice, perhaps reflecting changes in the number of fusion-competent secretory granules, as reported after manipulation of the GWAS gene TCF7L2 (81) or the microRNA miR124 (4) and might suggest a common mode of action of genes affecting T2D risk. Finally, it is possible, given that proinsulin levels were elevated in carriers of risk alleles, that prohormone processing is altered after Vps13c inactivation (67). To investigate this hypothesis, we measured random-fed insulin and proinsulin concentrations in RC-fed mice. No differences were seen in either insulin or proinsulin plasma concentrations, nor was the insulin/proinsulin ratio different between controls and KO mice (Fig. 13).
|
study
| 100.0 |
Random-fed insulin and proinsulin plasma concentrations in RC-fed mice. Plasma insulin and proinsulin were measured following collection from the tail vein (before culling) and aorta (immediately post mortem) of RC-fed Ctrl (black) or KO [green, males (A and B); purple, females (C and D)] mice aged over 21 wk. The ratio plasma insulin/plasma proinsulin is shown in E; n = 7 Ctrl and 8 KO (males), 10 Ctrl and 9 KO (females).
|
study
| 100.0 |
Changes in glucose tolerance were not apparent after the maintenance of mice (either male of female) on a HFD, suggesting that the phenotype might be rescued by changes in response to high-fat feeding and insulin resistance. One possible explanation might be an increase in the expression of other VPS13 family members, which could be triggered by a HFD, thus compensating for the absence of VPS13C. According to a recent study of mouse islet cell transcriptomes (5), Vps13c mRNA levels are two to three times those of Vps13a, -b, and -d in the β-cell under conditions of normal feeding. Whether changes in the expression of any of these genes occur under the stress of a HFD has yet to be investigated. We note also that a recent eQTL study (9) did not report a significant association with VPS13C (or other genes at this locus) and T2D risk, although whether the latter study was adequately powered to detect small changes is uncertain.
|
study
| 99.94 |
How might VPS13C influence insulin secretion in vivo? Clues might be gleaned by comparisons with other members of the VPS13 family. BLAST analysis of the protein sequences of the four family members indicates that VPS13C is most similar to VPS13A, sharing 41% identity (73) and the two proteins possess several common domains and have similar NH2 and COOH termini, indicating that they may have similar functions (73). Both can attach to membranes, although VPS13C has intramolecular duplications vs. VPS13A, which may imply neofunctionalization (i.e., the acquiring of new roles) compared with VPS13A (73). As noted above, a loss of VPS13A (also called chorein) expression leads to the rare neurodegenerative disease chorea-acanthocytosis (ChAc) (53, 71). Symptoms include cognitive dysfunction, hyperkinetic movement disorder, and erythrocyte acanthocytosis (72), leading to significant disability and a reduced life expectancy. Since the discovery of the cause of ChAc, much work has been done to investigate the molecular function of VPS13A. The protein has been localized to endosomal structures in yeast and erythrocytes (14, 28, 64, 70) as well as to the Golgi, and cofractionates with dense-core vesicles in synaptosomes (25, 34). VPS13A is also implicated in a plethora of cellular processes in different settings, including regulation of the actin cytoskeleton (2, 18, 64), protein trafficking (7), membrane morphogenesis (48), autophagy (45), and phagocytosis (59). Cells depleted of VPS13A have decreased levels of PI(4)P and of phosphorylated PI3K (18, 47, 48). Importantly, further evidence for a role for VPS13A in the control of regulated exocytosis was provided recently by Hayashi et al. (25), who demonstrated that VPS13A is localized to neurites in dopaminergic PC12 cells. These findings are thus strongly reminiscent of our findings here of colocalization between VPS13C and insulin in human β-cells (Fig. 12). The role for VPS13A in phosphoinositide (PI) metabolism is a function that is conserved between yeast and human orthologs and a possible mechanism by which VPS13A can function in so many different cellular processes (18, 47–49). If VPS13C were to have similar functions in β-cells as VPS13A, we would hypothesize that the former might be involved in protein trafficking, potentially through the regulation of PI metabolism.
|
study
| 99.94 |
Correct regulation of PI metabolism is essential for efficient insulin secretion from β-cells (79). Indeed, PI(4)P is the main precursor to form PI(4,5)P2, which is rapidly turned over to form second messengers required for insulin secretion (68) in a Ca2+-dependent manner akin to the release of neurotransmitters from neurons. Interestingly, a distinct role for PI(4)P in signaling from the plasma membrane in the β-cell has been suggested (80), since PI(4)P displayed antisynchronous oscillations compared with PI(4,5)P2 when MIN6 β-cells were stimulated with glucose. A direct role in secretion has already been shown in yeast (24), and it is well known that PI(4)P is involved in membrane trafficking between the Golgi and the plasma membrane and other endosomal compartments (75). Hence, VPS13C could function in insulin secretion through regulation of PI metabolism, affecting intracellular insulin trafficking.
|
study
| 99.94 |
Interestingly, new work shows that VPS13C is involved in lipid droplet formation and regulation of galectin-12 and seems to function in adipogenesis (84). The latter findings indicate that VPS13C may play additional roles in T2D in extra-pancreatic tissues.
|
study
| 100.0 |
In conclusion, human islet expression data suggest that variations in the level of expression of VPS13C and C2CD4A in the β-cell may contribute to altered T2D susceptibility in risk allele carriers, at least in females. The relatively mild effects of Vps13c ablation on glucose homeostasis are consistent with the hypothesis that changes in the expression of both genes may contribute to overall risk. Future functional studies will be required to determine the role of C2CD4A in the control of insulin secretion and the possible contribution of indirect mechanisms resulting from changes in the expression of either gene in extrapancreatic tissues.
|
study
| 100.0 |
G. A. Rutter thanks the Medical Research Council (UK) for Programme grant MR/J0003042/1, the Biotechnology and Biological Sciences Research Council (UK) for a Project grant (BB/J015873/1), the Royal Society for a Wolfson Research Merit Award, and the Wellcome Trust for a Senior Investigator Award (WT098424AIA). T. J. Pullen was a Diabetes research and Wellness Foundation postdoctoral Fellow (SCA/01/F/12). The work leading to this publication has received support from the Innovative Medicines Initiative Joint Undertaking under Grant Agreement no. 155005 (IMIDIA), resources of which are composed of a financial contribution from the European Union's Seventh Framework Programme (FP7/2007-2013) and EFPIA companies' in-kind contribution (G. A. Rutter, P.M.). Additional support was obtained from a Wellcome Trust core grant (075491/Z/04) and from the Advocacy for Neuroacanthocytosis Patients (to A. P. Monaco and A. Velayos-Baeza).
|
other
| 99.94 |
Z.B.M., N.F., M.C.C., M.H., P.C., G.M., A.V.-B., A.P.M., L.M., and P.M. performed experiments; Z.B.M., N.F., T.J.P., M.C.C., M.H., P.C., G.M., A.P.M., and G.A.R. analyzed data; Z.B.M., N.F., T.J.P., and G.A.R. interpreted results of experiments; Z.B.M., T.J.P., and M.C.C. prepared figures; N.F., T.J.P., M.C.C., M.H., P.C., G.M., A.V.-B., A.P.M., L.M., and P.M. approved final version of manuscript; A.V.-B. and P.M. edited and revised manuscript; G.A.R. conception and design of research.
|
other
| 99.94 |
Telomeres are unique nucleoprotein complexes that occur at the termini of eukaryotic chromosomes and are essential for the maintenance of genome integrity and prevention of chromosome degradation and fusion (1). Telomeric DNA is double stranded, consisting of multiple repetitive guanine-rich units (TTAGGG in humans), terminated by ∼300 nucleotide (nt) single-stranded overhang at the 3′-end (1). In vitro, short guanine-rich oligonucleotide sequences can fold into non-canonical DNA secondary structures, called G-quadruplexes (G4s) (2). Over the last decade a large number of proteins and enzymes participating in the regulation of G4 formation (and disruption) have been identified both in vitro and in vivo (3,4). More recently, several groups reported evidences of G4 formation in the genome of eukaryotic cells (5,6). The formation of G4s from exogenous G4 forming DNA sequences was also observed in live Xenopus laevis oocytes (7,8). In addition, a novel high-throughput sequencing technology allowed identifying a vast number of G4 structures and mapping their distribution throughout a human genome (9). Taken together, these observations strongly support the occurrence and functional relevance of G4s in vivo. G4s have been proposed to play a regulatory role in a number of cellular processes such as telomere maintenance, DNA replication, recombination and gene expression (4). They have been found to be involved in several oncological (10) and neurological (11) diseases and are potential targets for pharmacological intervention, utilizing specific G4-binding ligands (10,11). G4 structures are also used in a number of applications for DNA nanotechnology (12).
|
review
| 99.9 |
Despite the simplicity of the short telomeric G4 forming sequences, G4 structures have an intrinsic conformational polymorphism, due to differing backbone orientations and type of internal loops (13,14). The prevalence of one particular conformation in solution is dictated by factors such as counter-ion type and concentration, solvent composition, as well as the length of the DNA sequence and the type of terminal flanking nucleotides (15). Depending on these factors several stable folded G4 structures have been identified for the human telomeric sequence (16–20). This multifactorial nature of G4s makes their folding highly complex, possibly involving a number of stable intermediate states and misfolded species and is beyond a simple two-state approximation (21). Several studies have found G4 folding to occur through a simple sequential pathway, involving up to two intermediate states (21–26), however recently more complex folding scenarios have started to emerge (27–30). One important bottleneck in uncovering the G4 folding pathway remains the structural identification of the different folded states appearing during the folding process.
|
study
| 55.88 |
To address this problem we have utilized single-molecule Förster Resonance Energy Transfer (FRET) microscopy in combination with molecular dynamics (MD) simulations. Single-molecule FRET microscopy allows real-time monitoring of the folding of single G4 molecules, omitting both time and population averaging (31) and thus provides a direct view of G4 conformational diversity and dynamics. Whereas, combining the single-molecule FRET data with MD simulations allowed obtaining the structural description of the experimentally observed G4 states. We find that the hybrid 1 conformation represents the predominant long-lived folded conformation of the human telomeric sequence in potassium containing solutions. Before reaching this conformation, several alternative off-pathway folding transitions occur into marginally stable anti-parallel chair, 2-tetrad anti-parallel basket and hybrid 2 conformations.
|
study
| 100.0 |
Thus, our work provides a comprehensive understanding of the folding process of the human telomeric G4s in vitro that may be essential for elucidating the microscopic details of regulation of G4 structures by G4-binding chaperone proteins and resolving enzymes, interactions of G4s with small-molecule ligands and their overall function.
|
study
| 100.0 |
All fluorescently labeled DNA oligonucleotides were purchased from IBA (Germany) as a HPLC and PAGE purified product (Supplementary Table S1). The G4 forming oligos include a 22 (or 23) nucleotides overhang that was terminally (3′) labeled with Cy3 via amidite coupling. In case of the hTelo-I14 sequence the last guanine of the overhang was labeled with Cy3 via NHS coupling to the amino-C6 dG. The complementary duplex forming oligo was labeled with Cy5 via NHS coupling to the amino-C6 dT. It also contained a biotin at the 3′-end used for surface immobilization.
|
study
| 99.94 |
The G4 forming and complementary oligos were mixed with a ratio of 1.2:1, respectively, to a final concentration of 5 μM in the annealing buffer (20 mM Tris–HCl buffer (pH 7.5), with 50 mM LiCl or KCl). The sample was thermally annealed in a water bath, by keeping it at 95°C for 5 min, followed by gradual cooling to room temperature over 24 h.
|
study
| 99.94 |
Single-molecule FRET experiments were performed on surface-immobilized molecules employing a prism-based total internal reflection microscope (Zeiss). Labeled molecules were immobilized inside a coverslide chamber (a pair of quartz and glass slides assembled together by Parafilm stripes) using a BSA-biotin and streptavidin anchoring. Stock solutions of the labeled sample with a concentration of ∼5 pM were used for immobilization. The excess of non-immobilized molecules was washed out by flushing the chamber with dilution buffer (20 mM Tris–HCl buffer (pH 7.5), with 25 mM LiCl or KCl). Detailed protocols of these experimental procedures are published elsewhere (32). Prior to imaging, the coverslide chamber was flushed with an imaging buffer consisting of the dilution buffer supplemented with an oxygen scavenging system composed of 2 mM Trolox (Sigma Aldrich), glucose oxidase (Sigma Aldrich, 17 U/ml), catalase (Sigma Aldrich, 260 U/ml) and glucose (Sigma Aldrich, 4.5 mg/ml). Fresh imaging buffer was flushed into the chamber every 20 min, to avoid pH gradients. The excitation of fluorophores was achieved using an alternating laser excitation scheme (33) with 514 nm Ar-ion and 630 nm dye lasers. Fluorescence from the donor and acceptor fluorophores was spatially separated onto the EMCCD camera (Andor, iXON 3) by a wedge mirror. Movies were recorded with a 200 ms integration time per frame with a total length of 600 s (10 min) and were analyzed using the home-built iSMS software (34). Only the trajectories with single-step donor and/or acceptor photobleaching, or showing clear anti-correlated donor–acceptor dynamics, were selected for further analysis. The FRET efficiencies were obtained from the donor and acceptor fluorescence intensities as: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}\begin{equation*}E = \frac{{{F_A}}}{{{F_A} + \gamma {F_D}}}\end{equation*}\end{document}
|
study
| 100.0 |
where FD and FA denote donor and acceptor fluorescence intensities after donor excitation, respectively, that were corrected for background signal, donor leakage (D = 0.13) and acceptor direct excitation (A = 0.05) contributions. The γ-factor (γ = 1.2) was used to account for differences in brightness and detection efficiency for the donor and acceptor fluorophores. The correction factors were determined for all samples and a single average value of those was used for correcting the transfer efficiencies for all datasets.
|
study
| 99.94 |
Single-molecule FRET histograms were built based on the data arising from molecules containing both active donor and acceptor fluorophores (>100 molecules, except for the histograms shown in Figure 1F corresponding to the 15–60 min times, which are based on 50–60 molecules). All frames of each FRET trajectory (prior the first fluorophore bleaching event) were used to make single-molecule FRET histograms, where each frame yields a count in the histograms.
|
study
| 99.94 |
Cation-induced folding dynamics of the human telomeric G4 in 25 mM KCl. (A) Schematic representation of the folding of hTelo into various G4 conformations. Green and red circles placed at the G4 structures indicate the Cy3 and Cy5 fluorophores, respectively. (B) Representative FRET trajectory of single G4 molecule revealing several E states observed upon G4 folding. Colored solid lines indicate the FRET efficiency of these states as obtained from the HMM analysis. The black dashed line indicates the time of K+ addition. (C) Transition density plot showing the probability of transitions between different E states. (D and E) Dwell time histogram of the E≈0.3 and E≈0.73 states fitted with a single- and double-exponential functions, respectively (black lines). (F) Representative transfer efficiency histograms of G4 prior addition and for different time periods after addition of 25 mM K+, showing the slow kinetics of G4 folding and distribution of several E states. The last histogram was obtained for a sample that was thermally annealed and equilibrated for 24 h. Histograms were fitted globally to a mix of four Gaussian peak functions with shared peak positions and widths for each state (black lines). The color code used for identification of the different states and underlying FRET distributions is the same as in Figure 1B.
|
study
| 100.0 |
Single-molecule FRET trajectories showing conformational dynamics were analyzed using hidden Markov modeling (HMM) with the variational Bayesian expectation maximization technique (Figure 1B and Supplementary Figure S1) (34,35). The dwell times corresponding to each G4 conformational state were selected based on the plot of dwell times against transfer efficiency (Supplementary Figure S2). To avoid artefacts of the HMM analysis only the states with dwell times longer than two frames were included in further analysis.
|
study
| 100.0 |
For the kinetic analysis and estimation of folding/unfolding rate constants, the individual transitions were grouped based on the states before and after the transition. It should be noted that in this analysis the states that were cut by blinking, photobleaching or terminated by the end of the experiment were not included. The probability of transitions between different states is reported in Supplementary Table S2. Dwell time histograms were then built for each group and were fitted with single-exponential function to extract the rate constants for each transition (Supplementary Figure S3). The obtained rate constants are summarized in Supplementary Table S3. Detailed procedures for thermodynamic analysis are provided in the Supplementary Data and Table S4.
|
study
| 100.0 |
MD simulations were performed for the G4-duplex DNA constructs containing the wild-type and mutant chair (PDB ID: 2KM3 and see Supplementary Data for details), 2-tetrad basket (PDB ID: 2KF8), hybrid 1 (PDB ID: 2JSM) and hybrid 2 G4 conformations (PDB ID: 2JSL). All MD simulations were performed with generalized born implicit solvent (GBIS) in NAMD (36) version 2.9 with bonds to hydrogen atoms held rigid using an integration step of 2 fs. The simulation temperature of 298 K was maintained by the Langevin piston method with a damping coefficient of 1 ps−1. For the non-bonded interactions a switching function was used with a switching distance of 15 Å and with an 18 Å cutoff. The AMBER parm99bsc0 force field parameters with χOL4 corrections (37,38) in CHARMM format have been applied to DNA in all simulations. The DNA terminals were terminated according to standard AMBER topology. The solvent was described by the GBIS algorithm with a dielectric constant of 78.5 and implicit ions were included to a concentration of 0.15 M. In total 1.5–2.0 μs simulation was produced for each of G4 conformations. Sufficient sampling was achieved for all constructs (Supplementary Figure S4). Further details on MD simulations are provided in the Supplementary Data.
|
study
| 100.0 |
For the cluster analysis the G-quartet nucleobases were chosen for alignment. A cutoff of 8 Å for 10 clusters was used with the RMSD function in the VMD clustering plugin by L. Gracia (http://physiology.med.cornell.edu/faculty/hweinstein/vmdplugins/clustering). The results of the cluster analyses are shown in Supplementary Table S5.
|
study
| 99.94 |
The available volume (AV) approach (39,40) was used to quantify the transfer efficiencies for each construct. For the Cy3/Cy5 FRET pair used here the parameters were assessed by measuring the dye and linker dimensions in VMD and these are shown in Supplementary Table S6. The distances between AV determined fluorophore mean positions were found for all frames. Transfer efficiencies were calculated based on the fluorophore distances using four approaches with different frames of the trajectories as input. Supplementary Figure S5 shows the average distance and average efficiency for all four approaches.
|
study
| 100.0 |
For single-molecule FRET experiments we used a DNA construct consisting of a duplex DNA terminated by a G4 forming overhang AGGG(TTAGGG)3, corresponding to the human telomeric sequence (hTelo) that mimics the natural telomeric DNA. The DNA construct was labeled with Cy3 (donor) and Cy5 (acceptor) fluorophores for FRET detection (Figure 1A and Supplementary Table S1). Cation-induced folding of G4s and the subsequent conformational dynamics was monitored upon rapid addition of K+ (20 mM Tris–HCl and 25 mM KCl, pH 7.5), while imaging on the microscope (Figure 1B). The sample was prepared and initially imaged under non-folding conditions (20 mM Tris–HCl and 25 mM LiCl, pH 7.5) and an unfolded state with low transfer efficiency (E) state (E≈0.3) was initially populated. Upon addition of K+, we observe a rapid increase of transfer efficiency as a result of G4 formation (Figure 1B). Thereafter, the majority of FRET trajectories show multiple transitions between several E states, reflecting the highly dynamic nature of G4 folding (Figure 1B and C). Four distinct E states were identified through an HMM analysis of the FRET trajectories (Figure 1B and Supplementary Figure S1). The lowest E state (E≈0.3) represents the unfolded overhang and the three higher E states centered at E≈0.57, E≈0.73 and E≈0.88 correspond to different folded G4 conformations. These states are in dynamic interconversion with dwell times <10 s, except for the E≈0.73 state, which shows an additional long-lived subpopulation with a dwell time an order of magnitude larger (Figure 1D, E and Supplementary Figure S2). This long-lived state is also evident in FRET trajectories, where after initial conformational dynamics between several transient E states, G4s get stabilized in the long-lived E≈0.73 state (Figure 1B and Supplementary Figure S1). The long dwell time is also evidenced by an increased occurrence of static trajectories with E≈0.73 at later stages after folding initiation (Supplementary Figure S1). In addition to fast conformational dynamics, our single-molecule experiments allowed monitoring the G4 folding kinetics occurring on slower timescales (Figure 1F). Prior to the addition of K+, the G4 overhang is unfolded and appears as a single peak at low transfer efficiencies. After addition of K+ three high E peaks become apparent in the histograms. Interestingly, the distribution of these states varies notably within the first 60 minutes after folding initiation. At the early stages the E≈0.88 state is dominant, while after 60 min the E≈0.73 state becomes the major population. The FRET histogram obtained 60 minutes after KCl addition looks very similar to the one obtained in conditions where the population distribution has reached a complete equilibrium (Figure 1F, lowest panel). These results indicate that G4s reach thermodynamic equilibrium rather slowly, adopting the most stable (lowest energy) conformation within several hours after folding initiation in the presence of 25 mM KCl (Figure 1F).
|
study
| 100.0 |
But what are the actual conformations of the folded G4 states observed during the folding process? To address this question we used several variants of the hTelo sequence, with specific sequence modifications, that are less polymorphic and can potentially allow structural assignment of the observed E states.
|
study
| 99.9 |
First, we used a mutant telomeric sequence (hTelo-m) (AGGG(CTAGGG)3), which primarily forms an anti-parallel chair conformation in the presence of KCl (41). As expected, one major population is observed in the FRET histogram (Figure 2A). The peak is centered at E≈0.85, which is characteristic for the chair conformation, as previously found in similar buffer conditions (30). The highest FRET state (E≈0.88) of hTelo has a very similar FRET value, strongly indicating that this E state corresponds to the chair conformation.
|
study
| 100.0 |
Single-molecule transfer efficiency histograms of G4s formed from hTelo-m (A), hTelo - BrG9,14 (B), hTelo - BrG15 (C) and hTelo-I14 (D) sequences in 25 mM KCl. Measurements were performed on samples that were thermally annealed (A–C) or equilibrated for several hours (D). Histograms were fitted to one or a mix of three Gaussian functions (black lines). The color code used for identification of FRET peaks corresponding to different G4 conformations is the same as in Figure 1F. (E) Structure of one of the experimental DNA constructs with the hybrid 2 G4 structure (PDB ID: 2JSL) as obtained from molecular dynamics simulations. The positions of the donor and acceptor dyes were estimated using the available volume (AV) approach and are shown as green and red clouds, respectively. (F) Transfer efficiencies obtained from single-molecule FRET experiments (EsmFRET) and AV calculations based on the MD simulations (EAV). For each particular G4 conformation (see schematic structures) the EsmFRET values represent an average of mean transfer efficiencies obtained for hTelo and corresponding modified sequences. The EAV values are the average of transfer efficiencies for each G4 conformation estimated from four approaches of choosing input frames for the AV calculations (see Supplementary Figure S5 and ‘Materials and Methods’ section for details). The diagonal dashed line shows the identity line.
|
study
| 100.0 |
Next, we used variants of the hTelo sequence, where some guanines were substituted by 8-bromoguanine (BrG) derivatives, which preferentially adopt the syn orientation (42), allowing to achieve a stabilization of specific G4 conformations (18,43,44). The hybrid 2 structure is expected to be favored for a sequence containing two BrG substitutions at positions 9 and 14 (hTelo - BrG9,14) (44). Three E states were observed in the FRET histogram obtained for this sample (Figure 2B): a minor low E state representing an unfolded G4 and two high E states centered at E≈0.70 and E≈0.88, both corresponding to folded G4s. Taking into account the previous assignment of the E≈0.88 state to the chair conformation, the E≈0.70 state can be assigned to the hybrid 2 conformation. The hybrid 1 conformation is expected to be favored for the hTelo - BrG15 sequence (18). Three FRET peaks were also observed for this sample, with similar transfer efficiencies as for hTelo - BrG9,14 (Figure 2C). The highest FRET state is again attributed to the chair conformation and the dominant peak at E≈0.70 to the hybrid 1 conformation. The hybrid 1 and hybrid 2 conformations are structurally analogous (45) and would likely yield very similar FRET efficiencies, as observed here (Figure 2B and C). Thus, we conclude that the E≈0.73 state observed for the hTelo sequence is represented by a mixture of hybrid 1 and hybrid 2 conformations (Figure 1E). The hybrid 1 conformation is expected to be largely favored for the hTelo sequence (45). We therefore expect that the major long-lived subpopulation (E≈0.73) is likely to represent the hybrid 1 conformation, whereas the short-lived subpopulation can be attributed to the hybrid 2 conformation. Recent time-resolved NMR experiments suggest similar structural assignments for the G4s formed from an analogous telomeric sequence (29). It should be noted that the parallel G4 conformation appears at similar transfer efficiencies (E≈0.71) as the hybrid conformations (Supplementary Figure S6). Therefore, potentially it can also contribute to the E≈0.73 state of the hTelo sequence, however, the formation of a parallel conformation has been observed only in a crystal state or under dehydrating conditions and is unlikely to be observed under our experimental conditions (46,47).
|
study
| 100.0 |
The anti-parallel basket type G4 conformation containing only two tetrads has been previously observed to form from the human telomeric sequence (7). To check if this structure could contribute to the E≈0.57 state, we used a modified telomeric sequence where a guanine at position 14 was replaced by an inosine (hTelo-I14), which has been shown to favor the formation of the 2-tetrad anti-parallel basket conformation (48). Three E states were observed in the FRET histogram obtained for this sample (Figure 2D), where the low and high E states correspond to unfolded and chair conformations, respectively, whereas the intermediate peak appearing at E≈0.6 is expected to originate from the 2-tetrad basket conformation. The E≈0.57 state observed for the wild-type sequence (Figure 1F) is thus assigned to this conformation.
|
study
| 100.0 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.