The transcription factor regulatory factor X (RFX)-3 regulates the expression of genes required for the growth and function of cilia. We show here that mouse RFX3 is expressed in developing and mature pancreatic endocrine cells during embryogenesis and in adults. RFX3 expression already is evident in early Ngn3-positive progenitors and is maintained in all major pancreatic endocrine cell lineages throughout their development. Primary cilia of hitherto unknown function present on these cells consequently are reduced in number and severely stunted in Rfx3−/− mice. This ciliary abnormality is associated with a developmental defect leading to a uniquely altered cellular composition of the islets of Langerhans. Just before birth, Rfx3−/− islets contain considerably less insulin-, glucagon-, and ghrelin-producing cells, whereas pancreatic polypeptide–positive cells are markedly increased in number. In adult mice, the defect leads to small and disorganized islets, reduced insulin production, and impaired glucose tolerance. These findings suggest that RFX3 participates in the mechanisms that govern pancreatic endocrine cell differentiation and that the presence of primary cilia on islet cells may play a key role in this process.
Transcription factors belonging to the regulatory factor X (RFX) family are conserved in a wide range of species, including S. cerevisae, S. pompe, C. elegans, D. melanogaster, and mammals (1). They share a characteristic DNA-binding domain related to the winged-helix DNA-binding motif (2). Five RFX factors (RFX1 to RFX5) have been identified in humans and mice (1,3–6). It has been well established that RFX5 is a critical regulatory factor in the immune system (4,5). In contrast, clues concerning the in vivo functions of RFX1 to RFX4 have emerged only recently (7,8). The generation of Rfx3 knockout mice demonstrated that RFX3 plays a key role in controlling the expression of genes required for the formation of cilia (7). The orthologous factors DAF19 and dRFX also are essential for the expression of genes implicated in the formation, maintenance, and/or function of cilia in C. elegans and D. melanogaster (9,10). Regulation of cilia-related genes, thus, is an evolutionarily conserved function of RFX factors.
Cilia are cell surface protrusions that extend from a centriole-derived microtubule-organizing center called the basal body (11,12). The growth and maintenance of cilia is dependent on the bidirectional transport of proteins along their microtubular axoneme by a process called intraflagellar transport (IFT) (11,12). There are three structurally and functionally distinct types of cilia in mammals: nodal cilia, motile cilia, and primary cilia. Nodal cilia are found on cells of the embryonic node and play essential roles in establishing the left-right body axis (13). RFX3-deficient mice exhibit impaired growth of nodal cilia and consequently have frequent left-right asymmetry defects leading to embryonic lethality and situs inversus in the rare viable adults (7). Motile cilia are abundant on epithelial cells of the respiratory and reproductive tracts and ependymal cells of the brain, where they are involved in mucociliary clearance, the transport of gametes, and the movement of cerebrospinal fluid. Hydrocephalus, resulting from defects in specialized ependymal cells associated with alterations in the number of cilia, is a second characteristic of Rfx3−/− mice (14).
Primary cilia are solitary immotile organelles present on cells in many tissues (11). Their functions remain largely unknown, although there is growing evidence that they serve as sensory organelles involved in various biological processes (11,15). Immotile cilia situated in the embryonic node and primary cilia on renal epithelial cells elicit intracellular calcium mobilization in response to extracellular fluid flow (16,17). Primary cilia also are essential components of the vertebrate hedgehog (Hh) signal transduction pathway (18–20). Accordingly, mice carrying mutations in genes involved in IFT (Ift172, Ift88, Kif3a, Dnchc2, and Dync2li1) exhibit developmental phenotypes characteristic of severe defects in Hh signaling (21–23). Mutations in genes encoding cilia- or basal-body–related proteins are associated with human polycystic kidney diseases, nephronophthisis, and the Bardet-Biedl syndrome (11,15). Many pathological features of these diseases are reproduced in animal models in which genes coding for cilia-related proteins are mutated (24–27). Although the kidney is the major tissue affected in these diseases and animal models, the pathology frequently extends to other systems, particularly the liver and pancreas, suggesting that primary cilia play key roles in these organs as well (11,27–30).
Here, we show that RFX3 is specifically expressed in early endocrine cell progenitors as well as in all major endocrine cell lineages in the pancreas of mouse embryos and adults. Cilia on pancreatic endocrine cells consequently are reduced in number and severely stunted in Rfx3−/− mice. This ciliary abnormality is associated with a developmental defect that leads to an altered cellular composition of the islets of Langerhans. These results demonstrate that RFX3 is required for the development of pancreatic endocrine cells, raising the intriguing possibility that primary cilia may be implicated in this process.
RESEARCH DESIGN AND METHODS
All experiments were performed with littermates from crosses between Rfx3+/− mice. Adult mice were 8–14 weeks old. The stage of embryos was estimated from the gestational time, with day 0.5 being defined as the morning when a vaginal plug was detected. Adult mice and embryos were genotyped by PCR (7). Animal experimentation was performed with the permission of the federal and cantonal veterinary authorities.
Glucose tolerance tests.
Mice were given intraperitoneal injections of 15% glucose (1.5 mg glucose/g body wt) after a 16-h overnight fast. Blood was collected from tail bleedings, and glucose concentrations were measured using ACCU-CHEK Active (Roche Diagnostics, Indianapolis, IN).
Hormone quantification.
Pancreatic tissue was homogenized in ice-cold 0.18 N HCl and 70% EtOH, incubated overnight at 4°C, and centrifuged at 16,000g for 5 min, and supernatants were stored at −20°C. Insulin concentrations were determined by radioimmunoassay, as described (31), using 125I-labeled porcine insulin (SB-INS I-1; Sorin Biomedica, Saluggia, Italy) as tracer and rat insulin as standard. Glucagon and pancreatic polypeptide (PP) contents were determined using radioimmunoassay kits from Linco Research (St. Charles, MO). Somatostatin contents were determined using a radioimmunoassay kit from Phoenix Pharmaceuticals (Belmont, CA).
Histology and immunofluorescence.
Tissues were fixed in Bouin's solution and processed by standard procedures for hematoxylin/eosin staining and immunofluorescence labeling (32). Paraffin sections were washed in xylene and rehydrated with a series of ethanol washes. Frozen sections were prepared by standard methods. For ghrelin staining, slides were boiled for 15 min in 10 mmol/l citric acid (pH 6.0) and cooled for 10 min at room temperature. Slides were washed in PBS and blocked with PBS and 2% BSA for 30 min at room temperature. Primary antibodies were diluted in PBS, 0.1% BSA, 0.2% Triton X-100 (PBT) and incubated with the sections for 1 h at room temperature (insulin, glucagon, somatostatin, and PP antibodies) or overnight at 4°C (ghrelin, RFX3, detyrosinated, and acetylated α-tubulin antibodies). Sections were washed in PBS, incubated for 1 h at room temperature with secondary antibodies diluted in PBT, and washed in PBS. The antibodies and dilutions used are indicated in online appendix Table 1 (available at http://dx.doi.org/10.2337/db06-1187). β-Cell apoptosis and proliferation were measured for three mice of each genotype using a transferase-mediated dUPT nick-end labeling assay (Roche Diagnostics) and staining with antibodies against phospho-histone H3. For each parameter, 30–40 positive cells were scored.
Morphometry.
A total of 5- to 6-μm-thick pancreas sections spaced 200 μm apart were stained and visualized using an Axiophot I (Carl Zeiss MicroImaging, Zürich, Switzerland). Photographs were taken with ×4 (for measuring pancreas area), ×16 (for measuring islet area), or ×40 (for measuring endocrine cell area) objectives. Measurements were performed on 15–25 islets per section and on three to four sections covering the entire pancreas of each mouse. The total areas occupied by islets and specific endocrine cell populations were measured using Metamorph v6.2 software (Universal Imaging, Downington, PA). Volume densities (Vv) were obtained by expressing the areas occupied by hormone-positive cells as a percentage of total islet area. The fraction of islets occupied by hormone-positive cells in day postcoitum (dpc) 19 embryos is <100% (Fig. 6) because islets contain immature endocrine cells and other cell types and compartments not visualized in our analysis, notably endothelial cells and blood vessels.
Analysis of cilia.
Immunofluorescence images of cilia stained with detyrosinated or acetylated α-tubulin were acquired in 10–20 Z-stack slices using a LSM510 Meta confocal laser scanning microscope (Carl Zeiss, Oberkochen, Germany). The Z-stack images were merged using Metamorph software to obtain composite images that were used to measure the number, fluorescence intensity, and length of cilia. The number of cilia was quantified relative to the number of DAPI-stained nuclei. The fluorescence intensity per cilium was calculated as the sum of values across the entire depth of the Z-stack. This intensity is directly proportional to ciliary length, independently of the orientation of the cilia relative to the confocal plane. Direct measurements of ciliary length on the composite images underestimate the length of cilia because they do not take into account the orientation of the cilia relative to the confocal plane. Three-dimensional depictions of cilia were reconstructed using IMARIS software (Bitplane) and rotated in all orientations to ensure that single cilia were visualized. To determine which endocrine cell types carry cilia, sections were costained with antibodies against insulin, glucagon, somatostatin, or PP.
Quantitative RT-PCR.
Total RNA was extracted using TRIzol (Invitrogen, Carlsbad, CA). cDNA was synthesized from 1 μg RNA using random hexamer primers and Superscript II (Invitrogen). PCR was performed using the iCycler iQ Real-Time PCR Detection System (Bio-Rad, Philadelphia, PA) and iQ SYBR green Supermix (Bio-Rad). Results were quantified relative to a standard curve generated with serial dilutions of a reference cDNA preparation and normalized using TATA-binding protein mRNA. All experiments were repeated at least three times. Primers are provided in online appendix Table 2.
Statistical analysis.
Statistical significance was evaluated using a t test for independent samples, one-way ANOVA followed by posttest comparisons, and the Kolmogorov-Smirnov or median nonparametric tests. P values of <0.05 were considered statistically significant.
RESULTS
RFX3 expression in the pancreas is restricted to endocrine cells.
Immunostaining of pancreas sections from adult mice demonstrated that RFX3 expression is restricted to the islets of Langerhans (Fig. 1A). Islet-specific RFX3 expression also was observed in embryos at dpc 19 (data not shown). As expected, RFX3 was localized in the nucleus of islet cells (Fig. 1A). Specificity was demonstrated by the absence of labeling in the islets of Rfx3−/− mice (Fig. 1A). Costaining with antibodies against insulin, glucagon, somatostatin, or PP demonstrated that all four major endocrine cell types express RFX3 (Fig. 1B).
We next examined RFX3 expression in the pancreas at various developmental stages. As of dpc 17.5, all four endocrine cell types express RFX3 (Fig. 1B). Insulin- and glucagon-positive cells, which appear before cells producing somatostatin and PP, are already positive at earlier stages. Thus, insulin-positive cells that have developed by dpc 15.5 already express RFX3 (Fig. 1C). Similarly, RFX3 is expressed in glucagon-positive cells that have appeared by dpc 13.5 (Fig. 1D). Finally, in dpc 13.5 and 15.5 embryos, RFX3 is expressed in cells that express Ngn3, which is the earliest known marker of endocrine progenitor cells (Fig. 1C and D). RFX3 thus is expressed throughout the development of the four major pancreatic endocrine cell lineages.
Cilia are defective on islet cells of Rfx3−/− mice.
Islet cells carry primary cilia (33). We examined these cilia in Rfx3+/+ and Rfx3−/− dpc 19 embryos by staining pancreas sections with antibodies against cilia-specific tubulin isoforms (Fig. 2). Cilia are readily detectable in the islets of wild-type embryos (Fig. 2A). In agreement with the expression pattern of RFX3 (Fig. 1), these cilia are found on cells that express insulin, glucagon, somatostatin, and PP (Fig. 2C). Cilia were evident on only a fraction of the cells (Fig. 2D). This fraction is likely to be underestimated because the average diameter of the endocrine cells exceeds the thickness of the sections (5–6 μm).
The number of cilia was reduced by >40% in the islets of Rfx3−/− mice compared with control littermates (Fig. 2D). The cilia remaining in the mutant mice were severely stunted (Fig. 2A, C, and D). The stunted cilia are retained on all four types of endocrine cells (Fig. 2C). Measurements made by two approaches demonstrated that cilia were up to fivefold shorter in the islets of Rfx3−/− embryos (Fig. 2D). In contrast, no alteration in length was observed for cilia found in the exocrine ducts of Rfx3−/− mice (Fig. 2B), where RFX3 is not expressed.
We next examined the expression of Ift88 and Dync2li1, two RFX target genes coding for proteins implicated in IFT (23–25,34,35). Both genes contain well-conserved RFX binding sites in their upstream regions (Fig. 3B). Ift88 and Dync2li1 mRNA expression was significantly reduced in the pancreas of dpc 19 Rfx3−/− embryos (Fig. 3A). No change was observed for the expression of two control IFT genes, Kif3A and Ift172 (Fig. 3A).
Insulin production and secretion is defective in adult Rfx3−/− mice.
Blood glucose homeostasis was examined in adult Rfx3−/− mice. After fasting, Rfx3−/− mice had slightly higher basal blood glucose levels than control littermates (Fig. 4A). Moreover, they showed a greater increase in blood glucose levels and a marked delay in the return to baseline values after glucose injection (Fig. 4B).
This impaired glucose tolerance prompted us to assess insulin production and secretion by pancreatic β-cells. Compared with wild-type littermates, the total pancreatic insulin content was threefold lower in adult Rfx3−/− mice (Fig. 4C). A reduction in insulin secretion also was observed in response to perfusion of the pancreas with glucose (data not shown). The reductions in insulin content and secretion were highly significant despite the small number of viable adult Rfx3−/− mice that were available for these studies.
Islets are small and disorganized in adult Rfx3−/− mice.
To determine whether the defects in insulin synthesis and secretion were associated with alterations of the islets, we examined pancreas sections stained with antibodies specific for insulin (Fig. 4F). The average size of islets was reduced by ∼50% in Rfx3−/− mice (Fig. 4D and E). The mutant islets also were disorganized with respect to the relative localization of α- and β-cells (online appendix Fig. 1). In agreement with the islet-specific expression pattern of RFX3, no defect was observed in the exocrine pancreas of Rfx3−/− mice (data not shown).
Hormone synthesis is altered in the pancreas of Rfx3−/− embryos.
Most Rfx3−/− mice die in utero and at birth from defects in left-right body patterning (7). We therefore continued our analysis with dpc 19 embryos, because ∼25% of the Rfx3−/− embryos survive until this stage (7). Pancreatic insulin and glucagon contents were reduced 10- to 15-fold in Rfx3−/− embryos (Fig. 5A). In sharp contrast, the PP content was increased 15-fold (Fig. 5A). No significant change was evident for somatostatin (Fig. 5A).
We next quantified the abundance of insulin, glucagon, somatostatin, ghrelin, and PP mRNAs in the pancreases of dpc 19 embryos (Fig. 5B). Insulin 1, insulin 2, and glucagon mRNA levels were reduced ∼10-fold in Rfx3−/− embryos. A dramatic 150-fold reduction was observed for ghrelin mRNA. In contrast, PP mRNA abundance was increased over sixfold. No change was observed for somatostatin mRNA.
Islet cell development is defective in Rfx3−/− embryos.
Morphometric analysis revealed that the volume density of islets was decreased by 30% in dpc 19 Rfx3−/− embryos (online appendix Fig. 2), although the overall size of islets was largely unaffected (Fig. 6A). The reduction in volume density is attributed to a preferential decrease in the number of small islets (online appendix Fig. 2). This could not be attributed to a reduced rate of proliferation or to an increase in apoptosis of β-cells (data not shown, see research design and methods). No abnormality was detected in the exocrine pancreas of Rfx3−/− embryos (data not shown).
Immunostained pancreas sections from dpc 19 embryos revealed the presence of cells expressing insulin (β-cells), glucagon (α-cells), somatostatin (δ-cells), PP, and ghrelin in the islets of both Rfx3+/+ and Rfx3−/− littermates, although their numbers clearly were different between the control and mutant mice (Fig. 6A). Morphometric analysis demonstrated that there is a strong reduction in the number of β- (4- to 5-fold), α- (4- to 8-fold), and ghrelin-positive (13-fold) cells, whereas the number of PP-positive cells is increased 2- to 3-fold in Rfx3−/− mice (Fig. 6B). The number of δ-cells was unchanged (Fig. 6B).
The β- and α-cells that remain in Rfx3−/− embryos appeared to be labeled less intensely with antibodies against insulin and glucagon than those of control littermates (Fig. 6A). Quantification confirmed that the staining intensities for insulin and glucagon were significantly reduced in Rfx3−/− embryos, indicating that the residual β- and α-cells express less insulin and glucagon (online appendix Fig. 3).
To confirm the loss of β-cells, we monitored expression of the genes encoding the β-cell–specific markers islet amyloid polypeptide, Glut2, NeuroD1, and Pdx-1. mRNA levels for all four genes were reduced in Rfx3−/− embryos (online appendix Fig. 4). Taken together, the skewed cellular composition of the islets (Fig. 6), the altered pattern of pancreatic hormone production (Figs. 4 and 5; online appendix Fig. 3), and the reduced expression of β-cell markers (online appendix Fig. 4) indicate that Rfx3−/− mice exhibit major defects in the differentiation of pancreatic endocrine cells.
DISCUSSION
The differentiation of pancreatic endocrine progenitors into hormone-producing islet cells is regulated by specific transcription factors (36). Although the roles of several transcription factors have been defined by the analysis of mice carrying mutations in the corresponding genes (36), a great deal remains to be learned about the molecular and cellular mechanisms that govern cell lineage specification in the islets and differentiation of the different types of endocrine cells. Here, we show that RFX3 is a critical new player in this process. Rfx3−/− mice display a developmental defect leading to a markedly skewed composition of pancreatic islets. By birth, the islets of Rfx3−/− mice are characterized by a strong reduction in the numbers of cells producing insulin, glucagon, and ghrelin and a concomitant increase in the number of PP cells. The increase in PP cells partly compensates for the loss of the other cells, such that there only is a modest reduction in the volume of the pancreas occupied by islets and in the fraction of islets occupied by endocrine cells. The loss of β-, α-, and ghrelin-positive cells, together with the compensatory increase in PP cells, leads to a markedly altered pattern of pancreatic hormone production. The latter is evidenced by a decrease in total pancreatic insulin, glucagon, and ghrelin expression, whereas PP expression is increased.
In adult Rfx3−/− mice, altered islet development manifests itself by abnormal glucose homeostasis resulting from small disorganized islets and defects in insulin production and secretion. A partial effect was evident for several parameters in Rfx3+/− littermates, indicating that the haploinsufficiency of RFX3 is sufficient to perturb islet development (online appendix Table 2).
The pancreatic phenotype of Rfx3−/− mice is unprecedented, particularly with respect to the marked increase in PP cells, suggesting that RFX3 plays a unique role in pancreatic endocrine cell development. Two possible, nonmutually exclusive functions will be explored in future studies. Both would be consistent with the pattern of RFX3 expression observed during pancreatic endocrine cell development. First, since RFX3 already is expressed in Ngn3-positive cells, it could be required for one or more early cell fate decisions that favor the differentiation of endocrine progenitors into α-, β-, and ghrelin-positive cells at the expense of PP cells. Interestingly, cell ablation studies have suggested that α- and β-cells develop from precursors that express PP (36,37). A role of RFX3 in this early developmental decision would be consistent with the skewed cell composition we observed in the islets of Rfx3−/− mice. A second possibility is that RFX3 functions at later stages in cells that already are committed to specific endocrine cell lineages. RFX3 might be required for the maturation of α-, β-, and ghrelin-positive cells but inhibit the differentiation of PP cells. This would be consistent with the finding that RFX3 is expressed throughout the development of all pancreatic endocrine cell lineages. Our results provide some support for a function of RFX3 in mature β-cells. The residual β-cells in Rfx3−/− mice produce less insulin and feature impaired glucose-stimulated insulin release, pointing to defects in the terminal differentiation or function of β-cells.
RFX factors play evolutionarily conserved roles in regulating the expression of genes required for the growth, maintenance, and function of cilia (7,9,10,35,38). In agreement with this function, the phenotypes documented previously in Rfx3−/− mice can all be attributed to defects in cilia (7,14). Here, we show that primary cilia on endocrine cells in the islets of Langerhans are reduced in numbers and severely stunted in Rfx3−/− mice. This is attributed, at least in part, to a reduction in the expression of Dync2li1 and Ift88, two RFX3 target genes coding for proteins implicated in IFT. The function of primary cilia found on islet cells is unknown (33). However, there is growing evidence that primary cilia function as sensory organelles that relay signals affecting the development and physiological responses of the cells that carry them. The discovery that several kidney diseases are attributed to mutations in genes encoding cilia-related proteins, analysis of the function of these genes, and the study of mouse models of these diseases has provided strong support for the model that primary cilia on renal epithelial cells function as sensors that relay signals in response to extracellular fluid flow (11,16). A sensory role in perceiving extracellular signals also has been attributed to cilia situated in the embryonic node (13,17). Finally, the somatostatin receptor 3 and serotonin 5-HT6 receptor are localized in cilia of specific brain neurons, suggesting that these cilia have chemosensory functions (39,40). Given these findings, it is tempting to propose that cilia found on pancreatic endocrine cells could function as sensory organelles that relay responses to extracellular cues governing the differentiation, maintenance, and/or function of these cells. The defect we have documented in the pancreas of Rfx3−/− mice thus could be a consequence of defective cilia on islet cells.
No pancreatic endocrine phenotype was observed in conditional knockout mice in which cilia formation was perturbed in the pancreas by inducing deletion of the Kif3a gene with a Pdx-cre transgene (41). This raises the possibility that the islet phenotype observed in Rfx3−/− mice might be because of a mechanism that is independent of cilia. An alternative possibility is that no pancreatic endocrine phenotype was observed in the conditional Kif3a knockout mice because the loss of cilia was incomplete or induced at the wrong time during development. Furthermore, the deletion of Rfx3 could have a more severe impact on the formation and function of cilia than the deletion of Kif3a because RFX3 is likely to control genes coding for numerous components of cilia and their associated basal bodies (9,10,42).
Independent support for a key function of cilia on pancreatic endocrine cells is inherent in several other observations. First, mice deficient in Dync2li1—one of the RFX3 target genes that is downregulated in Rfx3−/− mice—exhibit a loss of nodal cilia and a strong reduction in Foxa2 (Hnf3β) expression in the embryonic node (23). Early embryonic lethality, resulting from severe developmental defects, precluded an analysis of pancreas development in Dync2li1−/− mice (23). However, Foxa2 is known to be implicated in the development and function of pancreatic endocrine cells (43). Second, mice carrying a hypomorphic allele of the Ift88 gene—the second RFX3 target gene that is downregulated in Rfx3−/− mice—are hypoglycemic after fasting and show a marked delay in return to baseline blood glucose levels after glucose challenge (30). This altered glucose homeostasis could result from deficient glucagon and insulin production, although no overt islet defects were documented (30). Third, the transcription factor Hnf6 has been implicated in both endocrine cell differentiation and the regulation of primary cilia formation in the pancreas (44,45). Finally, diabetes is associated with certain diseases, such as Bardet-Biedl syndrome, resulting from mutations in cilia-related genes (15). Taken together, these observations constitute compelling evidence for an important function of cilia in the development and/or function of pancreatic endocrine cells.
Recent work (18–22) has established that cilia are essential components of the Hh signaling pathway, which plays multiple roles at different stages of pancreas development. During early development, Hh signaling blocks pancreas formation (46). Less is known about its role at later stages because the ablation of Hh signaling in mice leads to early embryonic lethality (18). Nevertheless, several lines of evidence suggest that Hh signaling is required for endocrine cell development and function. In zebra fish, Hh signaling is required for the specification of pancreatic endocrine cells (47). In the adult mouse pancreas, the expression of components of the Hh pathway is restricted to islets and ducts (48). Finally, Hh signaling enhances insulin production and secretion by β-cells by regulating the expression of Pdx1, a key activator of insulin gene transcription (49,50). The defects documented here in the islets of Rfx3−/− mice thus could be a consequence of perturbed Hh signaling resulting from defective cilia.
Published ahead of print at http://diabetes.diabetesjournals.org on 17 January 2007. DOI: 10.2337/db06-1187.
Additional information for this article can be found in an online appendix at http://dx.doi.org/10.2337/db06-1187.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Article Information
Work in the laboratory of W.R. was supported by grants from the Swiss National Science Foundation, the Roche Research Foundation, and the Fondation Romande du Diabète. Work in the laboratory of B.D. was supported by the Centre National de la Recherche Scientifique, the ACI Bio du Développement et Physiologie Intégrative, the ACI Jeune Chercheur, and the Région Rhône-Alpes. Work in the laboratory of P.M. was supported by grants from the Swiss National Science Foundation, the Juvenile Diabetes Research Foundation International, and the National Institutes of Health. A.A.-L. was supported by fellowships from the Association de Langue Française pour l'Etude du Diabète et des Maladies Métaboliques and the Jules Thorn Foundation. D.B. was supported by a fellowship from the Région Rhône-Alpes and a short-term European Molecular Biology Organization fellowship. C.B. was supported by a fellowship from the French Research Ministry.