Insulin Storage and Glucose Homeostasis in Mice Null for the Granule Zinc Transporter ZnT8 and Studies of the Type 2 Diabetes–Associated Variants

The strategy used is described in results. Briefly, Slc30a8 knockout (ZnT8^−/−) mice were generated by GenOway (Lyon, France). This involved the insertion of a LoxP site together with an FRT (Flippase Recognition Target) flanked neomycin selection cassette within intron 1 and a single distal LoxP within the upstream exon 1 containing the translation initiation codon.

After homologous recombination in SV129-derived ES cells at the slc30a8/ZnT8 locus and injection into C57BL/6J blastocytes, the resulting flox'd mice were mated with CMV-Cre–expressing C57BL/6J mice. These were back-crossed three times (Toronto) or twice (London, Grenoble) onto a C57BL6 background. Separate colonies were maintained on normal rat diet with the fat/protein/fiber following contents: London, Grenoble, 2.7/14.4/4.7%; Toronto, 5.0/18.0/5.0%. Animals were continuously interbred to provide all three genotypes. ZnT8^−/− mice were viable and obtained with the expected Mendelian frequency. Comparisons were made between sex-matched littermates unless otherwise indicated.

Mouse pancreata were extracted and fixed in 10% neutral balanced formalin (Sigma) at 4°C for 18 h before dehydration and wax embedding and processing to obtain 5-μm slices (Carleton's Histological Technique, Drury and Wallington). Nuclei and cytoplasm were stained with hematoxylin and eosin, respectively. Sections were labeled with anti-insulin (1:300 dilution, Secondary-Alexa 488 1:1,000) and antiglucagon (1:200 dilution, secondary-Alexa 568 1:1,000) and sealed using Vector Shield Antifade Hard Set reagent (Vector Laboratories).

Brightfield and combined fluorescence and darkfield images were captured using a Zeiss Axiovert 200 inverted microscope with 2.5× air objective and a 1,300 × 1,000 pixel digital camera. Single islets within the slices were analyzed by confocal microscopy using a Zeiss Axiovert-200 microscope with an Improvision/Nokigawa spinning disc system running Volocity 4.0 (Improvision, Coventry, U.K.) software (24).

Human pancreata, fixed in 10% (v/v) formalin, were embedded in paraffin and sectioned. Sections (4 μm thick) were processed for immunohistochemistry (25) and labeled using mouse anti-insulin (Sigma, 1:2000), mouse antiglucagon (Sigma, 1:2000), and rabbit anti-ZnT8 (Mellitech, 1:200) and revealed with fluorescein anti-rabbit and Texas-red anti-mouse secondary antibodies (Jackson Immunoresearch, 1:200). Images were captured using a fluorescence microscope (Leitz DMRB, Leica) and digitized using a cooled CCD (C7780, Hamamatsu).

Islets were isolated from cadaveric donors according to ref. 26.

α- and β-cell populations were extracted from transgenic mice expressing the yellow fluorescent protein derivative Venus as described (27) (see supplemental Methods, available in an online appendix at http://diabetes.diabetesjournals.org/cgi/content/full/db09-0551/DC1).

Microarray analysis was performed as described in supplemental Methods. Quantitative RT-PCR (qPCR) for ZnT and ZiP isoforms was performed using SYBR Green on a 7900HT RealTime PCR system (Applied Biosystems) using three separate cell preparations (28).

Total islet RNA was isolated using TRIzol Reagent (Invitrogen, San Diego, CA) and treated with rDNase I (Ambion, Austin, TX). Isolated RNA (1 μg) was reverse transcribed using Moloney murine leukemia virus reverse transcriptase (Invitrogen). Details of qPCR primer design and reactions are given in supplemental Methods. Data were normalized to mouse β-actin mRNA.

Glucose tolerance was assessed by intraperitoneal glucose injection. Mice were fasted for 5–15 h, weighed, and glucose (1.5 g/kg unless otherwise stated) administered via intraperitoneal injection. Glucose in tail vein blood was measured using a Glucometer (Accu-Check).

Mice were anesthetized with isoflurane (Abbott) and an indwelling catheter installed into the femoral vein (29). Insulin sensitivity was assessed 4 days after surgery (30,31).

cDNAs encoding human R325- or W325-ZnT8 were subcloned using primers bearing c-myc tags from hZnT8 and hZnT8-R325W EGFP fusion protein constructs (16) into plasmid pIRES2-EGFP (Clontech) or pIRES2-dsRed2 (Invitrogen). cDNA encoding a CD38-EGFP chimaera was generated as described under supplemental Methods.

Expression was analyzed in INS-1(832/13) β-cells transfected using Lipofectamine2000 (Invitrogen) according to the manufacturers' instructions. Cells were visualized on a Leica SP5 laser-scanning confocal microscope.

ZnT8-c-myc tagged-EGFP constructs were ligated into pAdtrack-CMV at KpnI and EcoRV sites. Adenoviruses were produced as described (32).

Islets were aseptically isolated by collagenase digestion of mouse pancreas (7) and insulin (24) and glucagon (7) secretion measured in batch incubations. Insulin processing was assessed by pulse-chase labeling with ³⁵S-methionine (33).

Plasma membrane potential (V_m) and whole-cell K_ATP channel conductance (G_KATP) of β-cells were recorded in perforated patch configuration (34). Electrical capacitance (C_m) and Ca²⁺ current (I_Ca) through the plasma membrane of β-cells were recorded in standard whole-cell configuration (24). See supplemental Methods for further details.

Isolated mouse islets were dispersed and infected with adenovirus expressing neuropeptide Y (NPY)-Venus (100 MOI) (35) before imaging (24). Release was stimulated using 50 mmol/l KCl, 1 μmol/l forskolin, and 25 μmol/l isobutyl-methylxanthine. Fusion events were detected by eye, and the release kinetics, defined by measuring changes in fluorescence within a 5-pixel diameter circle, analyzed using Microsoft Excel and fitted where appropriate to a first-order decay curve using Origin 7.5 software (OriginLab, Northhampton, MA).

Intracellular free Ca²⁺ concentration ([Ca²⁺]_i) was measured at 37°C using Fura-Red (Invitrogen) as previously described (7,24).

Isolated islets were fixed and analyzed as previously described (24).

MIN6 cells were transfected using Lipofectamine 2000 (Invitrogen) with scrambled siRNA or siRNA against ZnT8 (5′-CCGUCAUGAUCUUAAAAGAtt-3′) (Ambion) or cotransfected with mCherry subcloned into pcDNA3.1 plasmid (Invitrogen, a gift from Dr. H. Gaisano, Toronto) and ZnT8-R325 or -W325 p-IRES constructs. After 24–48 h, cells were processed by immunocytochemistry (see Fig. 8 A) or loaded with 3 μmol/l FluoZin-3 AM or Zinquin for 50 min at 37°C, then incubated for 10 min without dye, in perifusion buffer containing (mM) 130 NaCl, 5 KCl, 2 CaCl₂, 1 MgCl₂, 5 NaHCO₃, 1 glucose, 10 HEPES, pH 7.4 equilibrated with 5% CO₂/95% O₂. Coverslips were then transferred to an open chamber and continuously perifused with 5 μmol/l ZnSO₄ in the presence of 1 mmol/l glucose at 37°C. Membrane-permeable zinc ionophore pyrithione and zinc chelator N,N,N′,N′ tetrakis-(2pyridylmethyl) ethylene diamine were used as controls. Cells were imaged with or without continuous perifusion at 36–37°C using an Olympus BX51W1 fluorescent microscope controlled with Image Master3 software (PTI, Lawrencewille, NJ) with a high-speed monochromator (PTI). FluoZin-3 emission was monitored using 485 nm excitation, 505 nm beam splitter, and 525 nm band-pass filter. Zinquin emission was monitored using 365 nm excitation, 375 nm beam splitter, and 385 nm long-pass filter.

HeLa cells transfected with hZnT8-V5 and ZnT8-EGFP were lysed 48 h posttransfection, and immunoprecipitation was performed with Dynabeads Protein G (Invitrogen) coated with anti-V5 (Clontech) or anti-EGFP (Invitrogen) antibodies.

The W325 form of ZnT8 was modeled on the crystal structure of the Escherichia coli zinc transporter YiiP (PDB accession 2QFI; resolution 3.8Å) (36) using Modeler version 8.2 (37). Alignment of the ZnT8 and YiiP sequences, which exhibit only ∼19% sequence identity, was aided by comparison of the patterns of residue conservation in closer mammalian and bacterial homologues of these proteins, respectively, using the program ConSeq (38,39). One hundred models were generated, and the five of lowest energy were further analyzed using MolProbity (40). Of these shown in Fig. 6,C 93% had residues in the allowed regions of the Ramachandran plot (the corresponding figure for 2QFI was 80%). The R325 variant of ZnT8 was generated from the model in silico using PyMOL (DeLano, W.L. The PyMOL Molecular Graphics System, 2002, www.pymol.org), which was also used to create the images shown in Fig. 6 C.

Data were analyzed using Clampfit (Axon Instruments), CellR̂ (Olympus), and Excel (Microsoft) software. Statistical significance was estimated using Student's t test with appropriate Bonferroni correction. Pearson correlation statistics were analyzed with Simple PCI.

ZnT8 immunoreactivity was localized to both β- and non–β-cells in isolated human pancreatic slices (Fig. 1,A) and dispersed human islets (Fig. 1,B) consistent with recent findings in rat pancreas (18). ZnT8 immunoreactivity was absent from neighboring acinar cells (Fig. 1,A), in line with expression profiling (17,18). ZnT8 mRNA was absent in pancreata from neurogenin3-deleted (ngn3^−/−) mice, which lack an endocrine compartment, confirming restriction to the latter pancreatic cell types (supplementary Fig. 1 and see ref. 41). At the mRNA level, ZnT8 was the predominant ZnT family member in both FACS-purified α- and β-cells (Fig. 1,C) (27), assessed by microarray (Affymetrix) analysis or quantitative PCR. Notably, in both purified β-and α-cells, ZnT8 mRNA levels exceeded by >10-fold the level of any other ZnT/slc30a family member (Fig. 1 C).

To explore the potential in vivo role of ZnT8 we generated mice lacking the slc30a8 gene. Deletion of exon 1 (Fig. 2,A and B) led to essentially complete ablation of ZnT8 mRNA and protein expression (Fig. 2,C and D) in isolated islets. No significant alterations were observed in the expression of other ZnT family members or the most abundant β-cell ZiPs (supplementary Fig. 2), ZiP6 or -7 (Fig. 2,D). Examined at 12 weeks, β-cell mass was not significantly different between +/+ and −/− mice (Toronto colony: 3.1 ± 0.3 and 3.2 ± 0.2 islets/arbitrary area unit, respectively; London colony: 0.51 ± 0.05 and 0.68 ± 0.03% total pancreatic surface; Fig. 3,A). Moreover, the proportions of β- and α-cells did not differ between genotypes in either colony (Toronto: not shown; London, Fig. 3,B). Surprisingly, normal insulin crystallization was not required to maintain unaltered insulin content in the absence of ZnT8 (for isolated size-matched 50- to 100-μm islets from London mice: 21.4 ± 5.8, 18.5 ± 5.7, and 13.7 ± 3 ng/islet for +/+, +/−, and −/−, respectively, n = 5 mice/genotype). Whereas zinc content measured with FluoZin-3 (mainly cytosolic; see below, Fig. 8,B) was not reduced in ZnT8^−/− mice (Fig. 3,C), the zinc content of secretory granules and other organelles, estimated using zinquin (Fig. 3,C) (42) or dithizone (Fig. 3 D) (43) (Lemaire et al., unpublished data), was considerably (>70%) lowered in ZnT8^−/− islets from either colony.

Examined by transmission electron microscopy, ZnT8^−/− β-cells displayed a significantly larger number of granules lacking a detectable dense core and an increase in granules containing atypical, rod-shaped cores (Fig. 3,E, supplementary Figs. 5–7), suggesting altered insulin crystal condensation (2). No difference in the number of granules morphologically docked at the plasma membrane was observed (supplementary Figs. 5F and 6F). The loss of dense core granules became more marked with age in Toronto mice (6 vs. 12 weeks; Fig. 3,E; supplementary Figs. 5C and 6C). Proinsulin processing to split proinsulins and the mature hormone were unaltered in ZnT8^−/− mice (London; Fig. 3 F).

Male ZnT8^−/− mice maintained on normal rat diet displayed unaltered changes in body weight with respect to control mice (London and Toronto colonies; not shown). In the Toronto colony, elevated fasting glucose levels and glucose intolerance during IPGTT were apparent in male ZnT8^−/− mice at 6 but not 12 weeks (Fig. 4). Female ZnT8^−/− mice in this colony displayed normal fasting glucose but were glucose intolerant at both ages (supplementary Fig. 4). Male null mice in the London colony were also glucose intolerant at both ages (supplementary Fig. 8) whereas normal glucose tolerance was seen in female null mice at 12 weeks (supplementary Fig. 8).

Whereas normal insulin sensitivity was observed in ZnT8^−/− mice (Toronto, not shown; London, supplementary Fig. 9), in vivo insulin release under IPGTT was significantly reduced (Fig. 4,A[b]). At 12 weeks, glucose-induced insulin secretion was significantly enhanced in ZnT8^−/− mouse islets (Toronto; Fig. 4,A[c] and B[b]), whereas glucose-inhibited glucagon secretion was unaffected (Fig. 4 A[d]). In islets from the London colony examined at 12 weeks, basal insulin release was significantly enhanced (ZnT8^−/− vs. wild-type) but glucose- or KCl-induced secretion did not differ between genotypes (supplementary Fig. 8[c]).

Examined in the Toronto colony, a high-fat diet led unexpectedly to a greater gain in body weight in ZnT8^−/− than wild-type mice (supplementary Fig. 3A and C). Under this regimen, null mice displayed elevated fasting blood glucose (supplementary Fig. 3D) and insulin (supplementary Fig. 3F) levels. Tendencies toward abnormal glucose tolerance, elevated glucagon levels, and decreased insulin sensitivity were also apparent (supplementary Figs. 3B, E, and G, respectively).

Examined in single β-cells from 12-week-old male ZnT8^−/− mice (London), no significant loss of glucose- or sulfonylurea-induced membrane depolarization (Fig. 5,A) or ATP-sensitive K⁺ (K_ATP) currents (not shown) were apparent, indicative of normal glucose metabolism. Similarly, depolarization-induced whole β-cell capacitance changes (Fig. 5,B) and Ca²⁺ currents (Fig. 5 C) remained unchanged, consistent with preserved exocytosis.

To determine whether the kinetics of granule cargo discharge may be affected by abnormal insulin crystallization, total internal reflection fluorescence (TIRF) analysis of release events was performed in single islet cells infected with adenovirus encoding neuropeptide-Y fused to the enhanced and red-shifted variant of green fluorescent protein, Venus (NPY-Venus) (Fig. 5,D–F) (35). Exocytosis was stimulated by KCl in the presence of the cAMP-raising agents forskolin and isobutyl-methylxanthine. This protocol maximized the number of detectable events and ensured rapid fusion pore dilation, previously shown to lead to “full” events in which the entire granule cargo was released (44). In this way, crystal dissolution was expected to be rate limiting for the dispersal of the fluorescence after fusion. No differences were obtained in the rate of NPY-Venus diffusion (t_½ = 0.17 ± 0.02 s [+/+] vs. 0.17 ± 0.03 s [−/−]) suggesting that crystal dissolution was not affected by the absence of ZnT8 (Fig. 5,E and F). Moreover, there was no difference in the number of “full” events, leading to complete or near complete loss of granule fluorescence (Fig. 5,E and F), or “partial” release events in which granule fluorescence remained after the transient increase during fusion (Fig. 5 F) (35).

To assess the potential impact of replacement of the lower-risk with the higher-risk residue at position-325, we undertook homology modeling studies based on the bacterial zinc transporter and ZnT homologue, YiiP. A member of the cation diffusion facilitator transporter family (38), YiiP reportedly functions as a dimer (36), confirming similar behavior for mammalian ZnT8 from β-cell lines migrated on SDS-PAGE as bands with apparent molecular masses of ∼40 and ∼90 kDa and indicating the presence of monomers and possibly an SDS-resistant dimer (Fig. 6,A) (18). To confirm the existence of dimers in living cells, we coexpressed EGFP- or V5-tagged ZnT8 in epithelial HeLa cells and performed immunoprecipitation experiments. Efficient coprecipitation indicated the existence of the transporter as dimers (Fig. 6,B). Based on the structure of YiiP (36), the probable locations of Zn²⁺ ion binding sites in both the cytoplasmic and transmembrane domains of the ZnT8 model, the latter accessible from extracellular/intragranular side of the membrane, were inferred from the bacterial protein structure (Fig. 6,C). Residue-325, at the cytoplasmic “tip” of each monomer in the homology model, forms part of the dimer interface and in different mammalian species is occupied by lower or higher risk (Homo sapiens), Q (rat and mouse), or higher risk (macaque) (supplementary Fig. 10). The impact of replacing lower with higher risk was assessed by in silico mutagenesis and revealed minimal distortion in overall folding (Fig. 6 C). Although confirmation of the precise arrangement of this region of the protein awaits determination of the crystal structure, alignment of the ZnT8 sequence with the template structure in this region of the homology model was supported by analysis of the pattern of residue conservation (see research design and methods).

To identify potential physiological differences between β-cells expressing either the low-risk or increased-risk variants of ZnT8 we used plasmid-based transfection or adenovirus-mediated infection to introduce either form of the protein into β-cell lines. Each isoform displayed a similar expression level as evidenced by Western immunoblotting (Fig. 6,D) and immunocytochemical analysis of single cells transfected with pIRES2-based plasmids (Fig. 6,E). Moreover, close subcellular colocalization with insulin-containing granules, apparent for endogenous ZnT8 in human β-cells (Fig. 1,B) (17), was also apparent for each of the overexpressed c-myc–tagged variants (Pearson correlation coefficient 0.77 ± 0.01 and 0.78 ± 0.02 for higher- and lower-risk mutants, respectively; Fig. 6 E).

We next examined the effects of each ZnT8 allele on the regulation of insulin release. No significant differences in glucose-stimulated insulin secretion were apparent in INS-1(832/13) cells (Fig. 7 A) or MIN6 β-cells (not shown) between cells overexpressing either the higher- or the lower-risk variant.

Measured in the whole-cell perforated-patch configuration, resting membrane potential (V_m) was identical in MIN6 cells overexpressing either vector alone, R325-ZnT8, or W325-ZnT8 (Fig. 7,B), and the resting conductance of ATP-sensitive K⁺ channels (G_KATP) was also of the same order in each case (Fig. 7,B). No differences in glucose- or KCl-induced increases in apparent intracellular free Ca²⁺ ([Ca²⁺]_i) were apparent between cells overexpressing R- or W325-ZnT8 (Fig. 7,C). However, overexpression of either ZnT8 isoform significantly potentiated [Ca²⁺]_i increases in response to elevated glucose concentrations (Fig. 7 D), suggestive of an effect of ZnT8 overexpression on either Ca²⁺ channel numbers or via changes in intracellular cation handling.

The effects of overexpression of either isoform on intracellular free Zn²⁺ concentrations ([Zn²⁺]_i), were next explored using trappable intracellular Zn²⁺-selective dyes: FluoZin-3, Zinquin ethyl-ester, or RhodZin3-AM (45). Whereas FluoZin-3 (Fig. 8,B) and RhodZin-3 (not shown) were localized largely to the cytosol and nuclear compartments, Zinquin displayed marked punctate staining (Fig. 8,B) (46), consistent with accumulation into secretory granules (47). We monitored zinc uptake by incubation in the presence of a supraphysiological concentration (5–10 μmol/l) of zinc ions. Suggesting that ZnT8 may mediate uptake of Zn²⁺ into the cytosol as well as the secretory granules, rates of uptake monitored with FluoZin-3 were diminished when the endogenous transporter was silenced (Fig. 8,A). Initial rates of Zn²⁺ uptake were markedly elevated after overexpression of the W- but not the R325-ZnT8 variant as chimaeras with mCherry (Fig. 8,B) consistent with a higher Zn²⁺ transporting activity of the former. In contrast to the c-myc–tagged chimaera (Fig. 6,E), overexpression of either variant interestingly led to detectable labeling of the plasma membrane as well as granules (Fig. 8,C), suggesting that ZnT8 was able to mediate zinc transport across both membranes. Importantly, zinc accumulation into granules, assessed using zinquin (Fig. 8 B), was significantly higher for overexpressed W- versus R-ZnT8.

We show that ZnT8 is required for the normal accumulation of zinc by β-cell granules, for normal insulin crystallization and glucose tolerance. However, the impact of ZnT8 deletion appeared to depend subtly upon variations in genetic background, as well as age, sex, and diet. Unexpectedly, and for reasons that remain to be elucidated, we also show that ZnT8 is required for normal body weight homeostasis under metabolic (high-fat diet) stress. Whether human carriers of the at-risk R-allele are similarly affected is presently unknown.

Studied in male ZnT8^−/− mice on a substantially C57BL6 background (Toronto colony; see research design and methods), glucose tolerance improved with age (from 6–12 weeks). Interestingly, this change was associated with a decrease (from 35–16%) in the number of dysmorphic, “rod-like” granules but an increase (from 24–36%) in electron-lucent (“empty”) granules (Fig. 3,E), whereas glucose tolerance normalized over the same period (Fig. 4). By contrast, in the London colony, rod-like still predominated over electron-luscent granules in ZnT8^−/− mouse β-cells at 12 weeks: glucose-stimulated insulin secretion was not enhanced with respect to wild-type controls, and the mice remained glucose intolerant (supplementary Fig. 8). Such observations may suggest that a preponderance of atypical crystals, rather than their total absence, contributes to abnormalities in insulin release and hence glucose intolerance.

Surprisingly, we observed no abnormalities in insulin processing neither in ZnT8^−/− islets from the London colony nor in islets from mice where dense cores were essentially eliminated (Lemaire et al., unpublished data) questioning the requirement of insulin crystallization for this process. Likewise, the kinetics of insulin release as assessed by TIRF microscopy were unchanged under conditions favoring the formation of large fusion pores (35). However, basal insulin release from ZnT8^−/− islets was enhanced (supplementary Fig 8[c]), perhaps indicative of more efficient release of noncrystalline (mono/dimeric) insulin when fusion pores are small (35).

We show here that ZnT8 is expressed in the majority of α-cells in both humans and rodent islets. Nonetheless, no differences in the morphology of α-cell granules in ZnT8^−/− mice were apparent (not shown). Moreover, we observed no alterations in the regulation by glucose of glucagon secretion from ZnT8^−/− islets (Fig. 4 A[d]), questioning the importance of zinc release from β-cells in the control of the latter process (7).

Although polymorphisms in the ZnT8/SLC30A8 gene that alter the risk of type 2 diabetes (20,48) are likely to have a milder effect on glucose homeostasis than that of whole body inactivation of slc30a8 in mice as described here, it seems likely they may alter β-cell Zn²⁺ content and hence the normal storage of insulin. We also confirm that this transporter is likely to be important for the uptake of Zn²⁺ into granules. However, the assay for ZnT8 transporter activity we have developed here unexpectedly revealed that the transporter may operate in the “reverse” direction, whereby low levels of ZnT8 on the plasma membrane also permit influx into the cell in the face of high Zn²⁺ concentrations. Similar bidirectional zinc transport has also been described for ZnT5 (49).

Importantly, we show here that the at-risk R-form of ZnT8 is less active as a zinc transporter. We also demonstrate by molecular modeling that R325 resides at the monomer interface, a site important for the transport activity of YiiP and other bacterial citation diffusion facilitator transporters (36,50). The resultant introduction of positive charge into this region of ZnT8, close to the predicted sites of bound structurally important Zn²⁺ ions, may be expected to affect the kinetics of Zn²⁺ transport and possibly the recruitment of zinc-metallochaperones (50). Carriers of this allele may therefore accumulate zinc, and thus package insulin, less efficiently into β-cell granules. Future studies, possibly using appropriately engineered mice expressing either allelic form of the transporter, will be needed to test this hypothesis.

The present report demonstrates that ZnT8 modulates glucose homeostasis in mice and shows that a “high risk” allele for type 2 diabetes displays decreased activity. Small molecule activators that target ZnT8 may thus represent an interesting new means to treat insulin secretory deficiency in this disease.

G.A.R. thanks the Wellcome Trust (Programme Grant 081958/Z/07/Z), the Medical Research Council (G0401641), National Institutes of Health (ROI DK-O71962-01), and the European Union FP6 (SaveBeta) for grant support. F.M.G. and F.R. received Clinical and Basic Science Senior Fellowships, respectively, from the Wellcome Trust. E.A.B. and T.J.N. received Divisional Studentships from Imperial College. S.A.B. thanks the Biotechnology and Biological Sciences Research Council (Grant number BBS/B/14418) and the University of Leeds for support. F.C.S. was supported by Gene Ontology Annotation 2004/11 (K.U.Leuven, EURODIA LSHM-CT-2006-518153; European Community) and Juvenile Diabetes Research Fund Grant 1-2006-182. Work in the R.S. laboratory was supported by Institut National de la Sante et de la Recherche Medicale (INSERM) and the Association Française des Diabetiques (AFD). M.B.W. was funded by an operating grant from the Canadian Institutes of Health Research (CIHR, MOP-49521) and N.W. by a CIHR doctoral award. The work in the R.Sk. laboratory was funded by the Canadian Institutes for Health Research.

F.C. and S.L. are employed by Mellitech. No other potential conflicts of interest relevant to this article were reported.

We thank M.T. Gage-Soufflot for assistance with immunocytochemistry, Gao Sun, Nasret Harun, and Sahar Vakshouri for able technical assistance. We thank Sabrina Mosaheb for preliminary electron microscopy experiments. We thank M. Nonnenmacher for technical expertise in coimmunoprecipitation experiments. We thank Christian Leibig and Martin Spitaler of the Facility for Live Cell Imaging and Microscopy (Sir Alexander Fleming Building, Imperial College London) and Lorraine Lawrence (NHLI) for valuable assistance with imaging and histology, respectively, and Sebastian Barg for advice on capacitance measurements.