Increased Expression of the Diabetes Gene SOX4 Reduces Insulin Secretion by Impaired Fusion Pore Expansion

Reduced glucose-induced insulin secretion (GIIS) is a hallmark of type 2 diabetes (1). This disease results from a complex crosstalk between lifestyle factors (e.g., body weight, age) and genetics (1). Insulin is released by Ca²⁺-dependent exocytosis of insulin-containing secretory granules. Exocytosis involves the fusion of the granular membrane with the plasma membrane, a process initiated by the establishment of a narrow fusion pore that connects the granule lumen with the extracellular space (2). Effective release of insulin requires the rapid expansion of the fusion pore so that the secretory granules integrate with the plasma membrane.

Genome-wide association studies have led to the discovery of >100 loci associated with increased type 2 diabetes risk (3,4), many of which are believed to act through reduced β-cell mass or interference with insulin secretion (5). For most of these loci, gene annotation is only tentative and based on the proximity of the single nucleotide polymorphism (SNP) to a certain gene (4). An intronic SNP in the CDKAL1 gene (rs7756992) is associated with a 50% increase in the risk of type 2 diabetes (6–9). However, data indicate that CDKAL1 is not the causative gene (9) and that its effect is instead mediated by altered expression of a nearby gene encoding the transcription factor SOX4 (10).

We show that increased SOX4 expression is associated with reduced GIIS and elevated plasma glucose and that increased SOX4 expression impedes the delivery of insulin into the extracellular space through increased expression of the exocytosis-regulating protein STXBP6 (amisyn). Overexpression of amisyn promotes the stabilization of the fusion pore and locks it in a partially expanded state (2–3 nm) (11). This prevents the exit of insulin from the granule lumen into the extracellular space and thus impairs insulin secretion.

Animal work was approved by the local and national authorities. Mice used were described previously (12) but were heterozygous for the Sox4 mutation and were Insr^+/+. Mice were killed by cervical dislocation, and islets were isolated by liberase digestion and handpicking as previously described (13). Static and dynamic measurements of insulin secretion were performed as described previously (12). Insulin was measured by mouse insulin ELISA kit (Millipore, Hertfordshire, U.K.).

Islets were isolated from 22-week-old male wild-type and mutant mice (four animals per genotype), and RNA was extracted with an RNeasy Mini Kit (QIAGEN) and validated with an Agilent 2100 Bioanalyzer (Agilent Technologies). Labeled and fragmented complementary RNA (cRNA) was hybridized to the Affymetrix 430 2.0 whole-mouse genome microarray and processed on an GeneChip Fluidics Station 450 and Scanner 3000 (Affymetrix).

The constructs coding for green fluorescent protein–tagged mouse and human Sox4 and Stxbp6 were purchased from OriGene Technologies (Rockville, MD). The Y123C mutation was introduced into mouse Sox4 by using the QuikChange protocol (Agilent Genomics). INS-1 832/13 cells were transfected with Lipofectamine RNAiMAX reagent (Life Technologies, Paisley, U.K.). The following small interfering RNAs (siRNAs) were used: ON-TARGETplus siRNA SMARTpool for Stxbp6 gene and ON-TARGETplus Non-targeting Control siRNAs (Thermo Scientific, Hemel Hempstead, U.K.). After 24 h, the cells were cotransfected with human growth hormone (hGH) and either Discosoma species red fluorescent protein (DsRed)–or green fluorescent protein–tagged mouse Sox4 or Stxbp6 by using GeneJuice Transfection Reagent (Merck, Nottingham, U.K.). Supernatants and cell pellets were collected, and the amount of growth hormone was measured by using an hGH ELISA kit (Roche Diagnostics, West Sussex, U.K.). EndoC-βH2 cells (14) were transfected with human-specific ON-TARGETplus siRNA SMARTpools for STXBP6 and SOX4 used in siRNA knockdown experiments. siRNA and transfections were performed as described for INS-1 832/13 cells. Quantitative analysis of gene expression was performed using QuantiFast SYBR Green PCR kit and gene-specific QuantiTect Primer Assays (QIAGEN). Expression was calculated using ΔCt method, with GAPDH as a reference gene.

Intracellular calcium concentration ([Ca²⁺]_i) was assessed in freshly isolated intact islets by using a dual-wavelength fluorescence microscopy system (Photon Technology International, Monmouth Junction, NJ) fitted on an inverted Zeiss microscope to allow ratiometric measurements with the probe Fura-2, AM (Invitrogen, Paisley, U.K.) as described previously (15).

For patch-clamp measurements, islets were dissociated into single β-cells. In whole-cell measurements, insulin-secreting β-cells were identified on the basis of their larger size (>5.5 picofarads) and complete inactivation of the Na⁺ current at −70 mV (16). Exocytosis and whole-cell Ca²⁺ currents were recorded with an Elektronik patch clamp 10 amplifier and Pulse software (HEKA Electronik, Lambrecht [Pfalz], Germany) as described previously (15). Single exocytotic events and fusion pore expansion were detected in the cell-attached configuration. The standard extracellular solution described in Zhang et al. (16) supplemented with 20 mmol/L glucose was used as the pipette filling medium. Patchmaster software (HEKA Electronik) and the Elektronik patch clamp 10 amplifier together implement an internal calibration and automatically correct for phase shifts and frequency-dependent attenuation when a sinusoidal voltage command of 25 kHz is generated (17). The scaled apparent capacitance (Im/ω) and conductance (Re) are then calculated online by the software.

P2X₂ currents were recorded in the whole-cell configuration in identified β-cells as described previously (19,20). Experiments in EndoC-βH2 cells were performed after transfecting cells with both STXBP6 as described previously and P2X₂ receptors (as described in Karanauskaite et al. [20]). To evoke exocytosis, the cells were infused with an intracellular medium containing 0–9 mmol/L CaCl₂/10 mmol/L EGTA mixture (calculated free [Ca²⁺]_i 0 or 2 μmol/L), 3 mmol/L Mg-ATP, and 0.1 mmol/L cAMP. The current spikes reflecting vesicular release of ATP were analyzed using Mini Analysis Program version 6.0.3 (Synaptosoft, Decatur, GA) to determine the charge (integrated current [Q]), rise times (t_10–90%), and half-widths (HWs) of the individual events (20).

Isolated islets were washed in PBS and fixed in cold 2.5% glutaraldehyde in PBS for 1 h at 4°C and washed twice in PBS. Islets were incubated for 1 h in 1% osmium tetroxide and washed twice in PBS. Islets were dehydrated through an ethanol series for 5–10 min each and then embedded in Agar 100 resin mix (Agar Scientific) pending sectioning and electron microscopy.

Islets from cadaveric donors of European ancestry were provided by the Nordic Islet Transplantation Program (http://www.nordicislets.org). All procedures were approved by the local ethics committee at Lund University. The islets were cultured in supplemented CMRL 1066 (MP Biomedicals) for 1–9 days before RNA preparation. Total RNA was isolated with the AllPrep DNA/RNA Mini Kit (QIAGEN). RNA quality and concentration were measured with an Agilent 2100 bioanalyzer and ND-1000 spectrophotometer (NanoDrop Technologies). Total RNA was converted into biotin-targeted cRNA, and the biotin-labeled cRNA was fragmented into strands with 35–200 nucleotides. This was hybridized onto a GeneChip Human Gene 1.0 ST Array overnight in a GeneChip Hybridization Oven 640 (Affymetrix). The arrays were washed and stained in a GeneChip Fluidics Station 450. Scanning was carried out with a GeneChip Scanner 3000, and image analysis was done with GeneChip Operating Software. Data normalization was performed by using Robust Multiarray Analysis. Insulin secretion in human islets was measured as described previously (21). Microarray expression data have been deposited at Gene Expression Omnibus under accession number GSE50398.

Mice harboring a point mutation (Y123C) in Sox4 (Sox4mt mice) have previously been reported to be glucose intolerant and hypoinsulinemic (12), but the precise mechanism by which this mutation results in diabetes is elusive. As shown in Fig. 1A and B, islets from Sox4mt mice have reduced (−50%) GIIS over a range of glucose concentrations, despite similar insulin content, compared with wild-type (Sox4wt) islets. In dynamic (perifusion) measurements, both Sox4wt and Sox4mt islets responded to an elevation of glucose with a biphasic stimulation of insulin secretion (Fig. 1C), but GIIS was equally reduced at all times, and the secretion defect could not be corrected by the K_ATP channel blocker tolbutamide (12).

Insulin secretion is secondary to electrical activity and the associated elevation of cytoplasmic-free Ca²⁺ ([Ca²⁺]_i) (22). We measured β-cell [Ca²⁺]_i in intact Sox4wt and Sox4mt islets (Fig. 2A). Both Sox4wt and Sox4mt islets responded to high glucose with a marked, sustained, and reversible increase in [Ca²⁺]_i. Tolbutamide was equally effective at increasing [Ca²⁺]_i, regardless of the genotype.

Because the reduction of insulin secretion in Sox4mt islets was neither corrected by tolbutamide nor correlated with reduced [Ca²⁺]_i, we hypothesized that the mutation instead interferes with insulin exocytosis (i.e., beyond the increase in [Ca²⁺]_i). Thus, we compared insulin exocytosis in Sox4wt and Sox4mt β-cells by whole-cell capacitance measurements (23). Exocytosis was elicited by a train of depolarizations roughly emulating the bursts of action potentials that normally trigger insulin secretion (24). The exocytotic responses surprisingly were not reduced in Sox4mt β-cells (Fig. 2B). The magnitude of the voltage-gated Ca²⁺ currents in Sox4mt β-cells was the same as in wild-type cells (Fig. 2C), making it unlikely that an exocytotic defect is obscured by a compensatory increase Ca²⁺ entry.

Ultrastructurally, Sox4mt β-cells appeared normal (Fig. 3A and B), but the granule density was increased (Fig. 3C), whereas the dense core areas and granule diameters were marginally reduced (Fig. 3D and E).

The secretory granules of pancreatic β-cells store ATP together with insulin (25). We monitored single-granule release of ATP by infecting β-cells with an adenovirus encoding ionotropic purinergic P2X₂ receptors (26) (Fig. 4A and B). Exocytosis was triggered by intracellular application of high [Ca²⁺]_i (2 μmol/L). Quantal (single-granule) release of ATP activates the P2X₂ receptors near the release sites and, thus, gives rise to transient membrane currents (19). The frequency of the exocytotic events thus detected was not different in Sox4wt and Sox4mt β-cells (0.24 ± 0.04 and 0.24 ± 0.04 events/s [data not shown]), which agreed with the whole-cell capacitance measurements of exocytosis (Fig. 2C). Importantly, many of the current transients recorded in Sox4mt β-cells exhibited slower kinetics than those observed in Sox4wt cells (Fig. 4C). The size and shape of these transients reflect the amount and speed of ATP release (25). The charge (Q) and kinetics (rise times and HWs) were measured as illustrated in Supplementary Fig. 1. The cumulative histograms of the rise times in Sox4wt and Sox4mt β-cells are compared in Fig. 4D. Whereas 70% of the events in Sox4wt β-cells had a rise time of <5 ms, 70% of the events in Sox4mt β-cells had rise times >5 ms. The events recorded in Sox4mt β-cells were also longer, and the distribution of the HWs was biphasic (Fig. 4E); ∼20% of the events had a duration of 50–200 ms (highlighted by arrow). However, no differences in the distribution of the charge (reflecting the total amount of ATP released) were detected (Fig. 4F).

We performed on-cell measurements of single-granule exocytosis and fusion pore conductance within the piece of membrane enclosed by the recording electrode. This technique allows fusion pore conductances to be measured for pores up to 10 nm in diameter (27) (Fig. 5A). In both Sox4wt and Sox4mt β-cells, stepwise increases in membrane capacitance (reflecting exocytosis of single insulin granules) were observed. The average amplitude of these steps was 5.2 ± 0.5 (n = 119) and 5.3 ± 0.4 femtofarads (fF) (n = 119). Assuming spherical geometry and a specific membrane capacitance of 10 fF ⋅ μm⁻², these values correspond to a granule diameter of 0.4 μm, which is in reasonable agreement with the ultrastructural measurements (Fig. 3E). In Sox4wt β-cells, 88% of capacitance steps (Fig. 5B) were associated with either short-lived or no increases in membrane conductance (Fig. 5A, left), indicating rapid and irreversible expansion of the fusion pore and complete integration of the granule membrane with the plasmalemma (27). By contrast, in Sox4mt β-cells, 48% of capacitance steps often flickered between the baseline and an elevated level, as expected for kiss-and-run exocytosis (28). This difference between Sox4wt and Sox4mt β-cells is statistically significant (P = 8 × 10⁻⁷ by χ² test). During these flickery capacitance steps, persistent increases in membrane conductance were observed (Fig. 5A, right). The mean conductance increase averaged 340 ± 31 picosiemens (n = 58). Assuming a cylindrical pore length of 15 nm (i.e., 2 × 7.5 nm; 7.5 nm being the thickness of a single membrane layer) and a solution resistivity of 66.7Ω ⋅ cm (Eq. 11.1 in Hille [28]) (2); this value corresponds to a fusion pore diameter of 1.8 ± 0.2 nm (29). The cumulative distribution of the estimated fusion pore diameter is shown in Fig. 5C. In wild-type β-cells, the average fusion pore conductance and diameter were estimated to be 335 ± 42 picosiemens and 1.6 ± 0.3 nm (n = 14). The likelihood of kiss-and-run exocytosis has been reported to correlate with granule diameter (29). We examined this aspect in Sox4mt β-cells and found that capacitance increase associated with kiss-and-run exocytosis and full fusion averaged 6.3 ± 0.7 fF (n = 57) and 4.5 ± 0.4 fF (n = 62; P < 0.01), respectively. This is opposite to what is expected from Laplace’s law and argues that factors other than the diameter determines whether a granule undergoes kiss-and-run or full-fusion exocytosis.

To identify target genes of Sox4 in β-cells, we performed a gene expression microarray on Sox4wt and Sox4mt islets (Supplementary Table 2). Sixty genes were upregulated (1.5–2.9-fold), and 56 genes were downregulated (1.5–3.3-fold). Given the phenotype, genes involved in exocytosis are particularly interesting. These include Stxbp6 (also known as amisyn [upregulated 2.1-fold]), Doc2b (downregulated 1.8-fold), and Rab3c (downregulated twofold). Total ablation of Doc2b was found to have only a small (10%) inhibitory effect on GIIS (30). Expression of a GTPase-deficient mutant of Rab3c has a moderate effect on insulin secretion (31). Reduced expression of Doc2b and Rab3c are therefore unlikely to explain the reduction of GIIS in Sox4mt islets. However, overexpression of Stxbp6 has been reported to slow the expansion of the fusion pore in adrenal chromaffin cells (11). The latter effect is reminiscent of that we observed in Sox4mt β-cells.

In islets from Sox4mt mice, Sox4 expression is increased by 100% compared with wild-type islets (Supplemental Fig. 3 in Goldsworthy et al. [12]). We therefore hypothesized that the effects on GIIS in Sox4mt islets are mediated through increased Sox4 activity, which in turn leads to increased Stxbp6 expression and impaired fusion pore expansion. We compared the effects of similarly overexpressing Sox4wt and Sox4mt on Stxbp6 expression in INS-1 832/13 cells (Supplementary Fig. 2). Both Sox4wt and Sox4mt increased Stxbp6 expression, but the effect was stronger for the mutant (Fig. 6A). In secretion experiments, down- and upregulation of Stxbp6 stimulated and inhibited glucose-induced growth hormone release (used as a proxy for insulin secretion in transfected cells [32]) by 30% and 40%, respectively (Fig. 6B). Although overexpression of Sox4wt had only a small inhibitory effect (−15%) on hormone release, stronger inhibition (−40%) was observed when overexpressing Sox4mt (Fig. 6B). Silencing Stxbp6 largely reversed the inhibitory effect of overexpressing Sox4mt.

To extend these data to humans, we correlated SOX4 expression to donor HbA_1c levels (a surrogate measure of long-term plasma glucose control), STXBP6 expression, and GIIS in a large collection of human islet preparations. There was a positive association between SOX4 expression and HbA_1c levels (Fig. 7A) and STXBP6 expression (Fig. 7B) but a negative correlation between SOX4 and GIIS (Fig. 7C). In agreement with published data (33), there was a tendency toward higher SOX4 expression in islets from donors with type 2 diabetes (defined as established disease or HbA_1c >6%) (Supplementary Fig. 3A). A positive association also existed between CDKAL1 and SOX4 expression (Supplementary Fig. 3B).

These analyses suggest that increased SOX4 expression, through elevated STXBP6, reduces GIIS in human β-cells. We confirmed this by experimentally up- and downregulating SOX4 and STXBP6 in the glucose-responsive (Supplementary Fig. 3C) human β-cell line EndoC-βH2 (14). Moderate overexpression (+150%) of SOX4 in EndoC-βH2 cells was associated with a 250% increase in STXBP6 expression (Fig. 7D), an effect that was fully antagonized by knock down of STXBP6 by siRNA (si-STXBP6). By contrast, silencing of SOX4 did not reduce STXBP6 expression below the low basal level. Overexpression of SOX4 inhibited insulin secretion by 50%, an effect that was reversed by knock down of STXBP6 (Fig. 7E). Overexpression of SOX4 also increased expression of syntaxin-1a (STX1A) (Supplementary Fig. 3D), possibly to compensate for the binding of stxbp6 to syntaxin-1 (34), but this effect is unlikely to explain the suppression of insulin secretion because a positive correlation between STX1A expression and GIIS has been reported (35). Downregulation of SOX4 affected neither STXBP6 expression (Fig. 7D) nor hormone release (Fig. 7F).

To examine the effect of increased STXBP6 expression on granule emptying in human insulin-secreting cells, we expressed P2X₂ receptors in EndoC-βH2 cells and monitored single-granule ATP release (Fig. 7G). Overexpression of STXBP6 resulted in a marked decrease in ATP release (measured as a decrease in Q) (Fig. 7H). The mean charge of events was reduced by 90%. However, the frequency of the events was increased from 1.04 ± 0.38 to 2.8 ± 1.3 Hz (n = 9 or 10; not statistically different). Combining the two effects suggests that there is a 70% reduction in stimulated ATP release following overexpression of STXBP6 (Fig. 7I), which is in reasonable agreement with the 60% suppression of stimulated hGH release (Fig. 6B). Similar results were obtained in INS-1 832/13 cells. No ATP release events were observed when the EndoC-βH2 cells were infused with Ca²⁺-free medium (Fig. 7F, top trace).

Type 2 diabetes is the epitome of a polygenic disorder (1). Genome-wide association studies have led to the identification of ∼120 common gene variants (SNPs) with increased type 2 diabetes risk (36). For the majority of these gene variants, the precise cellular/molecular mechanisms underlying the increased disease risk remain obscure. The diabetes-associated SNP rs7756992 is commonly referred to CDKAL1, but it has been suggested that this locus increases disease risk through SOX4 rather than CDKAL1 (10). Of note, its expression is increased in diabetic islets (35), but whether increased expression of SOX4 is causally linked to type 2 diabetes and, if so, the underlying cell biological mechanism has not been studied.

In the present study, we show that increased SOX4 expression in human pancreatic islets correlates with reduced GIIS in vitro and increased HbA_1c levels in vivo. We also found a correlation between CDKAL1 and SOX4 expression in keeping with the idea that the CDKAL1 locus influences the expression of both CDKAL1 and SOX4 (10). Furthermore, we observed a tendency toward increased SOX4 expression in islets from donors with diabetes (Supplementary Fig. 3A), which agrees with previous reports based on microarray analysis of a β-cell–enriched fraction obtained my laser capture microdissection (35).

To address the mechanisms by which SOX4 may interfere with insulin secretion in humans, we made use of the Sox4mt mouse model in which islet Sox4 expression is twice that of wild-type islet (12). We found that GIIS was reduced by ∼50% in Sox4mt islets, a defect that was not corrected by the sulfonylurea tolbutamide. The reduction of insulin secretion could not be attributed to any impairment of key functional parameters, such as glucose-induced [Ca²⁺]_i, Ca²⁺ entry, or exocytosis (determined as depolarization-evoked increases in whole-cell membrane capacitance).

The delivery of insulin and other granule constituents, like ATP (25), into the extracellular space requires the establishment and expansion of a fusion pore connecting the granule lumen and the exterior (27). Normally, the initial opening of the fusion pore is followed by its rapid expansion, and following a short delay, the entire granule collapses into the plasma membrane, ensuring efficient delivery of the high-molecular-weight cargo into the extracellular space (full fusion). Occasionally, however, fusion pore expansion is halted following the initial opening and may eventually close (kiss-and-run exocytosis [27]).

By using on-cell capacitance measurements of individual exocytotic events and the associated fusion pore openings, we found that full fusion was the predominant (∼90%) form of exocytosis in wild-type β-cells, which agrees with previous reports (29). However, fusion pore expansion is impaired in Sox4mt β-cells, and the fraction kiss-and-run release events increased to ∼50% at the expense of full fusion. During kiss-and-run events, the fusion pore is locked in a partially expanded state with a diameter of 1–2 nm.

We determined the functional impact of this on the evacuation of the granule lumen by measurements of ATP release. The molecular dimensions of ATP are 1.6 × 1.1 × 0.5 nm. With the observed distribution of fusion pore diameters, we estimate that ∼20% of the total number of release events in Sox4mt β-cells (i.e., 43% of the 50% showing persistent fusion pore) have fusion pore diameters smaller or comparable to the narrowest cross-section of ATP (indicated by a dashed line in Fig. 5C). Thus, the fusion pore may function as a molecular sieve that restricts the exit of ATP by steric hindrance. This likely accounts for the slower rise times of ATP release and the component of ATP release events with a duration >20 ms (Fig. 4D). Larger molecules like insulin are likely to face an even greater problem with exiting through the narrow fusion pore. At concentrations >2 mmol/L, insulin exists as a hexamer with a hydrodynamic radius of 5.6 nm (37). Given that the intragranular insulin concentration is ∼100 mmol/L (22), we conclude that insulin is trapped within the secretory granules undergoing kiss-and-run exocytosis. Indeed, the observed decrease in full fusion from 88% to 52% is sufficient to account for the observed 40% reduction of GIIS in Sox4mt islets. It is tempting to speculate that the reduction in full fusion accounts for the increased granule density observed in Sox4mt β-cells. Although most fusion pores are too small to allow exit of insulin, 10% have a diameter >6 nm. A fusion pore diameter as large as this is sufficient to allow the exit of hexameric insulin, which might explain the slight reduction in dense core area of the insulin granules (Fig. 3D).

Since SOX4 is a transcription factor, it is not surprising that the expression of >100 genes were affected in Sox4mt islets. However, guided by the functional data specifically implicating kiss-and-run exocytosis, we identified Stxbp6 (amisyn) as a likely mediator. Overexpression of Stxbp6/STXBP6 mimics the effects of Sox4/SOX4 on secretion in rat/human insulinoma cells. Conversely, silencing Stxbp6/STXBP6 reverses the inhibitory effect of overexpressing Sox4/SOX4. Unexpectedly, and unlike what has previously been observed in rodent insulinoma cells (12), silencing SOX4 in human insulinoma cells did not affect hormone release. The mechanism underlying this discrepancy remains unclear.

The present data suggest that increased expression/activity of SOX4, through increased expression of STXBP6 and impaired expansion of the fusion pore, plays a role in insulin secretion and diabetes etiology. This scenario agrees with the proposal that Stxbp6 forms nonfusogenic complexes with syntaxin and thereby contributes to the regulation of SNARE complex formation (34). This is the first example of a major disease that can be linked to defective fusion pore expansion. It is possible that the mechanisms we have uncovered here may become activated during long-term hyperglycemia, a condition reported to be associated with an increase in kiss-and-run exocytosis (38). It appears that the effect of increased STXBP6 expression is graded. Whereas a moderate increase (up to +100%) results in an increased occurrence of kiss-and-run exocytosis (as seen in the Sox4mt mice), stronger overexpression may result in exocytosis being aborted before granule emptying (Supplementary Fig. 4). In cells overexpressing STXBP6, we observed small and short events, possibly reflecting repetitive openings and closures of the fusion pore, releasing a puff of ATP every time.

If increased SOX4 expression is part of the disease etiology and if it acts by interference with fusion pore expansion, then it may seem paradoxical that type 2 diabetes is not also associated with defects of neurotransmission. There is in fact a link between type 2 diabetes and neuropsychiatric disorders (39). However, slowed or partial fusion pore expansion will only have a marginal effect on the release of low-molecular-weight neurotransmitters because they are small enough to pass through a constricted fusion pore.

This study illustrates how detailed cell physiological studies can help us to move from SNPs through the identification of the actual gene involved to a fuller understanding of the causal mechanisms. These findings raise the interesting possibility that pharmacological procedures promoting full fusion may correct the insulin secretion defect in type 2 diabetes.

Acknowledgments. The authors thank David Wiggins (Oxford Centre for Diabetes, Endocrinology & Metabolism) for technical assistance and the staff at the Mary Lyon Centre at MRC Harwell for maintaining the Sox4 mutant mouse colony. The authors also thank Patricia Muller (Leicester University, Leicester, U.K.) for advice on cell transfection and Jochen Lang (Université de Bordeaux, Bordeaux, France) and Raphael Scharfmann (Institut National de la Santé et de la Recherche Médicale [INSERM], Paris, France) for the provision of the INS-1 832/13 and EndoC-βH2 cells, respectively.

Funding. This study was supported by the Medical Research Council (MR/L020149/1 to B.H., R.C., and P.R.), the Wellcome Trust–supported training programme OXION (to J.G.), the Wellcome Trust (Senior Investigator Award 095531/Z/11/Z to P.R.), Swedish Research Council (VR, International Recruitment), and the Knut and Alice Wallenbergs Stiftelse (Wallenberg Scholars Programme).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. S.C.C. contributed to the study design, performance of experiments, data analysis, and drafting and revision of the manuscript. H.W.D., B.H., and A.H. contributed to the performance of experiments, data analysis, figure preparation, and drafting of the manuscript. J.A., M.V.C., J.G., M.God., S.L., M.Gol., A.S., A.I.T., and A.H.R. contributed to the performance of experiments and data analysis. R.C. and P.R. contributed to the study design and drafting and revision of the manuscript. S.C.C. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.