|Home | About | Journals | Submit | Contact Us | Français|
Investigating the dynamics of pancreatic β-cell mass is critical for developing strategies to treat both type 1 and type 2 diabetes. p53, a key regulator of the cell cycle and apoptosis, has mostly been a focus of investigation as a tumor suppressor. Although p53 alternative transcripts can modulate p53 activity, their functions are not fully understood. We hypothesized that β-cell proliferation and glucose homeostasis were controlled by Δ40p53, a p53 isoform lacking the transactivation domain of the full-length protein that modulates total p53 activity and regulates organ size and life span in mice.
We phenotyped metabolic parameters in Δ40p53 transgenic (p44tg) mice and used quantitative RT-PCR, Western blotting, and immunohistochemistry to examine β-cell proliferation.
Transgenic mice with an ectopic p53 gene encoding Δ40p53 developed hypoinsulinemia and glucose intolerance by 3 months of age, which worsened in older mice and led to overt diabetes and premature death from ~14 months of age. Consistent with a dramatic decrease in β-cell mass and reduced β-cell proliferation, lower expression of cyclin D2 and pancreatic duodenal homeobox-1, two key regulators of proliferation, was observed, whereas expression of the cell cycle inhibitor p21, a p53 target gene, was increased.
These data indicate a significant and novel role for Δ40p53 in β-cell proliferation with implications for the development of age-dependent diabetes.
The maintenance of adult β-cell mass is critical for the appropriate regulation of glucose homeostasis in both humans and rodents. β-Cell replication is generally accepted as essential for β-cell expansion after birth but remains controversial in adult β-cells (1–7). Nevertheless, a recent study highlighted the ability of human β-cells to replicate in response to combined expression of two cell cycle regulators (8).
Cell replication is regulated by a balance between the active cyclin-cyclin dependent kinase (CDK) complexes and CDK inhibitors, which regulate phosphorylation of retinoblastoma protein, S-phase entry, and cell cycle progression (4). Among the cell cycle proteins regulating the G1 to S phase transition, CDK4 and cyclin D2 are particularly important. Mice deficient in CDK4 develop β-cell hypoplasia and diabetes (9,10); conversely, overexpression of CDK4 in transgenic mice causes islet hyperplasia without hypoglycemia (9). Cyclin D2, the most abundant cyclin in the islet, has been reported to be critical for postnatal pancreatic β-cell growth (5,11). Global loss of cyclin D2 in mice causes loss of β-cell mass, glucose intolerance, and age-dependent diabetes (5,11), whereas absence of cyclin D2 prevents the expansion of β-cell mass in response to a spectrum of insulin resistance (12). Furthermore, transplantation of human islets transduced with cyclin D1 and CDK6 (expressed in humans but not rodent islets) into diabetic mice improved blood glucose (8). Other reports indicate that Akt is important for the regulation of cyclin D1, cyclin D2, and CDK4 (13), and confirm a link between cell cycle progression and insulin/IGF-1 signaling, a pathway critical for β-cell growth and function (14–17).
The tumor suppressor p53 is a transcription factor and a global suppressor of cell cycle genes, such as cyclins and CDKs (18), and also can actively block cell cycle progression by transactivating its main target gene p21 (19). Mice deficient in p53 are developmentally normal, but show an increased risk of developing spontaneous tumors in diverse tissues as they age (20,21), but they do not develop insulinomas. Although most studies have focused on p53 action in the context of tumor formation and apoptosis, the key role of p53 in cell growth suggests that the protein may participate in β-cell cycle control and consequently contribute to glucose homeostasis.
The current study investigated the importance of p53 in β-cell growth using mice carrying an extra copy of p53 that codes for Δ40p53, a 44 kD isoform of p53, also described as ΔNp53 or p44 (mouse)/p47(human), that lacks the first 40 amino acids of the full-length protein (22). Indeed, up to 10 p53 isoforms have been recently identified indicating the complex regulation of p53 (23). Inappropriate expression of Δ40p53 alters the balance that normally exists between the isoforms and stabilizes full-length p53. Stabilization of p53 in p44 transgenic (p44tg) mice confers increased tumor suppression, but also results in accelerated aging and is associated with abnormal insulin/IGF-1 signaling (22) and impaired cell proliferation (24). Because aging is an important risk factor for the development of type 2 diabetes (25,26), we used this model of accelerated aging to investigate the role of p53 in islet/β-cell growth. We report that misexpression of Δ40p53 in p44tg mice is associated with early hypoinsulinemia and glucose intolerance that worsens with age and leads to overt hyperglycemia and a dramatic decrease in β-cell proliferation. These data suggest that the coordinated activity of p53 isoforms, in particular that of Δ40p53 and full-length p53, is necessary for the maintenance of β-cell mass during aging, a finding that could have implications for the prevention of age-dependent diabetes.
p44tg mice were generated as described previously (22). Data from male mice are reported. Body and pancreas weight were recorded, and blood glucose (Glucometer Elite, Bayer, Terrytown, NJ), plasma insulin (enzyme-linked immunosorbent assay), and glucagon (radioimmunoassay) levels (Joslin DERC Specialized Assay Core, Boston, MA) were measured before experiments. Glucose tolerance tests, acute glucose-stimulated insulin secretion (GSIS) tests, and insulin tolerance tests were performed as previously reported (27). Pancreata were harvested for immunohistochemical analyses and islet morphometry, and islets were isolated for RNA extraction and in vitro GSIS on batches of 15 size-matched islets as reported previously (28).
Pancreas sections were used to assess β-cell mass and size as described previously (27). β-Cell proliferation was assessed in pancreas sections as described previously (28), using anti-insulin (guinea-pig antibody, DAKO, Carpinteria, CA), anti-BrdU (rabbit polyclonal antibody, DAKO), or antiphosphohistone H3 antibodies (rabbit polyclonal antibody, Upstate Biotechnology, Lake Placid, NY). Other antibodies used include β-catenin (rabbit polyclonal antibody, BD Biosciences, Oxford, U.K.), glucagon (mouse monoclonal antibody, Sigma-Aldrich, St. Louis, MO), somatostatin (rabbit polyclonal antibody, Abcam, Cambridge, U.K.), p21 (mouse monoclonal antibody, DAKO), insulin (mouse monoclonal antibody, Biogenex, San Ramon, CA), and DAPI (Sigma-Aldrich) for nuclear staining.
Quantitative real-time RT-PCR was performed on total RNA samples extracted from islets and processed as described previously (28). All primer sequences are available on request.
Western blotting was performed as described (3). Antibodies used include cyclin D2 (rabbit polyclonal antibody, Santa Cruz Biotechnology Inc., Santa Cruz, CA), FoxO1 (rabbit polyclonal antibody, Cell Signaling, Danvers, MA), pancreatic duodenal homeobox-1 (PDX-1) (gift from C.V. Wright, Vanderbilt University), p27 (mouse monoclonal antibody, BD Biosciences), p21 (mouse monoclonal antibody, DAKO), and tubulin (mouse monoclonal antibody, Abcam).
All data are presented as mean ± SEM and were analyzed using an ANOVA or unpaired, two-tailed Student t test as appropriate. A P value less than 0.05 was considered significant. A minimum of n = 3 was used for all studies unless indicated.
We compared the phenotype of mice at 3, 10, and 12–14 months of age. The phenotype of p44tg mice becomes evident at 3 months, and the animals begin to succumb at ~14 months of age (22). p44tg mice were lighter than their nontransgenic counterparts at all ages, as shown for 3 months of age in Fig. 1A. The animals had already developed hypoinsulinemia at this age, which persisted throughout life (Fig. 1C). Although the animals displayed normal glucose tolerance at 2 months (Fig. 2A), they had developed intolerance by 3 months (Fig. 2B) and were significantly hyperglycemic by 12–14 months (Fig. 1B and and2C),2C), suggesting overt diabetes that could contribute, in part, to death at this age. In contrast, all controls exhibited fed blood glucose levels <150 mg/dL at 12 months of age (Fig. 1B). Glucagon levels were not significantly different between groups (Fig. 1D). Insulin sensitivity was unchanged in p44tg mice from 3 months of age (Fig. 2E and F) and even mildly improved in younger p44tg mice (Fig. 2D). These data indicate that increased dosage of Δ40p53 leads to hypoinsulinemia and glucose intolerance as early as 3 months and worsens with age.
Although both pancreas weight and number of islets exhibited an age-dependent increase in control mice, the total number of islets was less in p44tg mice even when normalized to total pancreas weight (Fig. 3A and Supplementary Fig. 1). Compared with 3 months of age, the differences became more marked at 12–14 months. The reduced pancreatic weight and islet number in p44tg mice were associated with a dramatic decrease in β-cell mass (Fig. 3B). Indeed, β-cell mass was significantly lower in the p44tg mice and ranged between 11% (young) and 14% (old) in the age-matched control animals. Quantification of β-cell size revealed a 30% decrease in the p44tg mice compared with controls at 3 months of age but showed the opposite in the older p44tg mice (Fig. 3C and D). This represented a twofold increase in β-cell size as the p44tg mice aged from 3 to 12 months. To explore the pathways that regulate β-cell size, we considered previous reports that proteins in the insulin/IGF-1 signaling pathways regulate β-cell size via Akt/mTOR/S6K1 (14–16). Quantitative real-time RT-PCR for these molecules did not reveal significant differences between groups at 3 months of age except for a decrease in insulin receptor (IR) expression in p44tg islets (Fig. 3E). However, in islets from older p44tg mice, we observed increased gene expression of IR, IRS1 (IR substrate-1), and especially Akt1, mTOR, and S6K1, suggesting an attempt at compensation by increasing the β-cell size in the older mutant mice. This was not observed in the liver or white adipose tissue of p44tg mice (Supplementary Fig. 5).
Examination of islet morphology by immunohistochemistry (Fig. 4A) revealed the typical pattern of a core of insulin-containing cells surrounded by non–β-cells in islets from controls and in a majority of islets from p44tg mice at 3 months of age (Fig. 4A, top). However, ~35% of islets in the normoglycemic and hypoinsulinemic p44tg groups exhibited infiltration of α- and δ-cells within the islet core and a relative decrease in the number of β-cells (quantified in Fig. 4B), which correlated with the hypoinsulinemia at this age (Fig. 1C). These abnormal features, which were not observed at 2 months of age (Supplementary Fig. 2), were evident in ~85% of the islets by 12 months of age in the p44tg group (Fig. 4A, bottom). In parallel, islet hormone gene expression (Fig. 4C) revealed a significant increase in somatostatin gene expression that was consistent with a relative increase in δ-cells in p44tg mice (Fig. 4A), whereas glucagon gene expression (Fig. 4C) and circulating glucagon levels (Fig. 1D) were not different between groups at all ages. A significant increase in the expression of insulin and GLUT 2 in 3-month-old transgenic mice suggested a compensatory effect at this younger age (Fig. 4C, top) that was not evident in the older hyperglycemic group (Fig. 4C, bottom). Because p44tg mice exhibit hypoinsulinemia and reduced β-cell numbers, we evaluated β-cell secretory function in vivo (Fig. 4D) and in vitro (Fig. 4E) by measuring insulin release in response to glucose stimulation. Although in vivo GSIS was impaired in p44tg mice (Fig. 4D) and likely contributed to the glucose intolerance (Fig. 2B), the GSIS in isolated islets in vitro was not different between groups (Fig. 4E), suggesting the absence of a cell autonomous insulin secretory defect.
Next, we determined whether the decrease in β-cell mass in p44tg mice (Fig. 3B) was due to changes in β-cell proliferation. Coimmunostaining of pancreas sections with antibodies against insulin and either one of two proliferation markers showed a decrease in the proliferation of β-cells with age in control islets (Fig. 5A and B) as reported previously (29). BrdU labeling demonstrated a virtual absence of replicating β-cells in young and old p44tg mice compared with controls (Fig. 5A). A decrease in pHH3 immunoreactive β-cells at 3 and 12–14 months of age in p44tg mice confirmed reduced number of cells in the M or late G2 phases of the cell cycle (Fig. 5B). Before 2 months of age, no significant difference was evident in β-cell replication between the two groups (Supplementary Fig. 2) and correlated with normal glucose tolerance (Fig. 2).
We next evaluated the effects of neogenesis (30) and apoptosis (31) in the maintenance of β-cell mass. Single and clusters (≤8) of insulin+ cells (markers of neogenesis) were virtually absent in pancreatic ducts in 3-month-old control and p44tg mice. Although some insulin+ cells/clusters were observed in 12- to 14-month-old mice, no significant difference was evident between groups (number/mm2 pancreas: 0.3 ± 0, control vs. 0.3 ± 0.2, p44tg; P = NS, n = 3). Similarly, no significant differences were observed in insulin+ cells in extraductal pancreatic tissue (3 month; 14 ± 0.3, control vs. 12.8 ± 3.4, p44tg; 12–14 month; 8.3 ± 0.2, control vs. 10.2 ± 2.4, p44tg; P = NS; n = 3), suggesting that neogenesis is not altered significantly when Δ40p53 is misexpressed. Although p53 has been studied in the context of apoptosis in multiple tissues, it has not been causally involved in β-cell death during diabetes (32). Nevertheless, we performed the transferase-mediated dUTP nick-end labeling (TUNEL) assay and found no significant differences in TUNEL+ β-cells between groups (Supplementary Fig. 3). These data provide strong evidence that the misexpression of Δ40p53 led to a marked reduction of β-cell proliferation, consistent with reports for other cell types (22,24).
p53 regulates the G1 to S phase progression through cell cycle inhibitors such as p21, but recent reports indicate that p53 also targets integral cell cycle proteins, such as cyclin E and CDK4 (18). We observed a significant decrease in the gene expression levels of cyclin D2, CDK2, and cyclin E1, and a significant increase in the level of cyclin E2 in 3-month-old p44tg mice (Fig. 6A). At 12 months of age, only cyclin E1 gene expression remained low; at the same time, however, CDK4 levels became significantly increased. The decrease in cyclin D2 gene expression at 3 months of age correlated with lower protein expression in isolated islets (Fig. 6B, left). In old mice, despite normal levels of gene expression, cyclin D2 protein level was significantly decreased (Fig. 6B, right).
We next examined the expression of FoxO1, which regulates β-cell proliferation (33), and two of its target proteins, PDX-1 (34) and p27 (35). Examination of gene and protein expression in whole islets revealed that FoxO1 expression was significantly lower at 3 months but unchanged at 10 months in p44tg mice (Fig. 7A and B). PDX-1 gene expression was consistently higher in p44tg mice (Fig. 7A); however, its protein level was consistently lower (Fig. 7B). This decrease was less pronounced in old mice and correlated with fewer β-cells in islets from the transgenic mice (Fig. 4B). p27 gene and protein expression remained unchanged (Fig. 7A and B). These data indicate that decreased β-cell mass and proliferation observed in p44tg mice (Figs. 3B and and5)5) are associated with decreased expression of cyclin D2 and PDX-1.
Finally, we evaluated the gene expression of p53 and its protein targets (19). To measure the level of transcription of both p53 (endogenous) and Δ40p53 (endogenous and transgene) genes, we designed two sets of primers for quantitative RT-PCR because the use of specific primers to distinguish both genes was not possible. Np53 primers targeted the 5′ end of the p53 transcript and amplified transcripts derived from the endogenous p53 gene only, whereas Cp53 primers targeted sequences at the 3′ end of RNA and amplified transcripts derived from both the endogenous p53 gene and the transgene. We also measured the expression of several well-known p53 target genes (19), such as p21, 14–3-3, Gadd45, Bax, MDM2, PTEN, and the IGF-1 receptor (IGF-1R). We observed a significant decrease in endogenous p53 gene expression (Np53 primers) that correlated with reduced Gadd45 gene in islets from 3-month-old p44tg mice (Fig. 8A, top). In islets from 12-month-old p44tg mice, we observed an increase in Δ40p53 expression (Cp53 primers) that was associated with an enhanced expression of p21 and MDM2 (Fig. 8A, bottom). Although Bax gene expression was significantly elevated in p44tg islets at both ages, we did not observe significance differences between groups in the TUNEL assay (Supplementary Fig. 3). The increase in gene expression of IGF-1R that was evident in young p44tg islets did not persist in older mice, and we did not observe alterations in PTEN gene expression at either age (Fig. 8A). However, we did find an increase in p21 protein by Western blot analysis of p44tg islets (Fig. 8B) and by immunohistochemistry in β-cell nuclei of p44tg islets (Fig. 8C). Thus, the strong inhibition of β-cell proliferation observed in p44tg islet cells correlated with a significant increase in p21, the main target of p53 in cell cycle regulation.
By using p44tg mice with an ectopic p53 gene that encodes Δ40p53 (36), we have identified a role for the tumor suppressor p53 in the maintenance of normal β-cell mass and glucose homeostasis. Although Δ40p53 has virtually no autonomous transcriptional activity, it can interact with full-length p53 to form tetramers that bind DNA and activate or suppress target gene expression (36,37). Expression of ectopic Δ40p53 in the mouse alters the balance between the full-length and short isoforms that normally exists and hyperactivates p53 (22,36,38). This increased activity of p53 can account for the higher levels of p21 transcripts, which we observed in β-cells from 10–12-month-old p44tg mice. The expression of Δ40p53 in p44tg mice can also account for the higher level of mRNA encoding the IGF-1R, whose transcription is differentially regulated by wild-type and mutant p53 (39). An increased gene expression of IGF-1R in liver and white adipose tissue from p44tg mice (Supplementary Fig. 5) suggests a non–tissue-specific effect. Similar to embryonic fibroblasts derived from p44tg mice (22), islets from young p44tg mice expressed significantly high levels of IGF-1R, indicating impaired trans-suppression activity. Both the increased trans-activation of p21 in old islets and the decreased trans-suppression of the IGF-1R in the young can be explained by differences in expression of the transgene encoding Δ40p53 with age, as detected by primers that amplify sequences encoding the COOH-terminal domain of the p53 protein. Younger p44tg mice expressed control levels of Δ40p53 and slightly lower levels of p53, whereas older mice expressed normal levels of p53 and higher levels of Δ40p53. Biochemical experiments have demonstrated that the stability and activity of the p53 tetramer are exquisitely sensitive to the dose of Δ40p53, with low doses of Δ40p53 activating and high doses inactivating p53 function (40,41). The differential effects of Δ40p53 on p53-dependent trans-activation and trans-repression with age could also explain how p21 levels increase in both age groups, but by different mechanisms. In old mice, the increase would be due to a direct effect of p53 on the p21 promoter, leading to increased transcription of the p21 gene. In young mice, on the other hand, increased p21 could be an indirect effect of higher IGF-1R expression and activation of IGF signaling due to impaired IGF-1R trans-repression. Stimulation of the IGF-1 signal transduction pathway can increase p21 (42), and further work is necessary to delineate the link among Δ40p53/p53, p21, and IGF-1R levels in β-cell proliferation.
β-Cell proliferation and glucose tolerance were impaired in 3-month-old p44tg animals and worsened to overt diabetes as the animals aged. Although random-fed blood glucose levels were normal in the transgenic mice at 3 and 10 months, the mutants displayed clear intolerance in response to a glucose challenge indicating deficiency in functional β-cells. Previous studies in rodents have reported normoglycemia even when β-cell mass is reduced. For example, Sreenan et al. (43) report a reduced β-cell mass before the onset of diabetes in the nonobese diabetic mouse, and Tavana et al. (44) describe the phenotypes of 1-month-old mice doubly mutant for p53 and nonhomologous end-joining deficiency that exhibit an ~50% decrease in β-cell mass and yet manifest blood glucose levels that are not significantly different from controls. Further, obese and nondiabetic humans have been reported to express a wide range of β-cell mass that is adequate to maintain euglycemia up to a specific threshold, and crossing the threshold correlates with fasting hyperglycemia and overt diabetes (45). In our study, it is possible that p44tg mice that are significantly leaner than controls at all ages only cross this “threshold” at ~12 months of age when they begin to exhibit overt hyperglycemia. Additional longitudinal studies are necessary to investigate these observations. Thus, our data suggest that the balance between full-length and Δ40p53 isoforms plays a critical role in the maintenance of β-cell proliferation and glucose homeostasis, and may be important in regulating the cell cycle during aging. Further support for a role for p53 in glucose homeostasis is evident from recent studies. For example, Minamino et al. (46) reported a role for adipose p53 in the regulation of insulin resistance. In our study, circulating proinflammatory cytokines, which are potential effectors in insulin resistance associated with diabetes, showed no significant differences between groups (Supplementary Fig. 4). The cytokine expression profile in liver and white adipose tissue was also similar between groups. These data, along with the normal insulin sensitivity in the global p44tg mice, indicate that the glucose intolerance is due to intrinsic effects of p53 on β-cell mass rather than to secondary effects of transgene expression in other insulin sensitive tissues. Indeed, Tavana et al. (44) reported that DNA double-strand breaks combined with an absence of p53-dependent apoptosis in mice leads to reduced β-cell replication and severe age-dependent diabetes, supporting our hypothesis that p53 plays a role in the regulation of β-cell proliferation.
Although the mechanisms that regulate the dynamics of β-cell turnover during aging are still being unraveled, several studies including our own support a role for replication as the major mechanism that underlies the compensatory growth response to insulin resistance (1–5,47). Two proteins that are important for this compensatory growth response are cyclin D2 (5,11,12) and PDX-1, the pancreatic duodenal homeobox domain protein (3). The lower expression of cyclin D2 in islets would potentially exacerbate the loss of β-cells and lead to glucose intolerance and overt diabetes. In this study, the expression of PDX-1 was also altered in p44tg β-cells. Although the gene expression of PDX-1 was increased in both young and old p44tg mice, the protein expression in islets was low, suggesting compensatory effects in gene expression or more dominant regulation of the transcription factor at the protein level. In old p44tg mice, however, there were fewer β-cell nuclei in which PDX-1 could be detected. Immunohistochemistry (data not shown) also revealed that the localization of PDX-1 correlated inversely with that of FoxO1, a transcription factor that regulates expression of this homeodomain protein (34). Because FoxO1 transactivates cell cycle inhibitors (p21 and p27) (35) and represses cell cycle activators (cyclin D1 and D2), nuclear localization of FoxO1 could have a major impact on cell cycle progression in aging β-cells. Further subcellular fractionation approaches are necessary to directly address this question. It is also possible that FoxO1 and p53 interact with deacetylase sirtuin 1 and function in a cooperative manner in an aging environment (48). Thus, the alterations in both PDX-1 and cyclin D2, in our model of accelerated aging, underscore the concept that these two proteins are involved in β-cell replication during aging.
Two CDKIs, namely, p16 and p19, have been reported to be involved in β-cell regenerative failure in diabetes (49). Our observations of enhanced expression of p16 and p19 genes confirm the premature aging phenotype of p44tg mice (Supplementary Fig. 6). However, the regulatory link between p53 and the two aging markers in the maintenance or amplification of the phenotype requires further investigation.
Maier et al. (22) previously described that young p44tg mice displayed signs of aging in bone. Further, the osteoblast-secreted molecule, osteocalcin, has recently been reported to modulate insulin secretion and insulin gene expression that correlated with expression of CREB and NeuroD genes (50). In our study, we observed a mild but significant decrease in osteocalcin levels only in old p44tg mice; however, gene expression of insulin, CREB, or NeuroD was unaltered (Supplementary Fig. 7).
Finally, the significance of the presence of p53 in non–β-cell fractions obtained from dispersed islet cells (29) remains unclear. Although the relative increase in α-cells in p44tg islets was not associated with an increase in glucagon gene expression or circulating glucagon levels, we observed a significant increase in the number of somatostatin-secreting δ-cells and gene expression in the mutant islets. The presence of a p53 response element in the somatostatin promoter (Genomatix) warrants further studies to evaluate whether the increase in somatostatin actually decreases insulin secretion or limits β-cell growth in a paracrine manner in the p44tg mice.
In summary, increased dosage of Δ40p53 in mice promotes hypoinsulinemia and glucose intolerance, ultimately leading to overt diabetes with age and early death. The suppression of β-cell proliferation secondary to changes in the expression of p21, cyclin D2, and PDX-1 in this mouse model of accelerated aging implicates impaired p53 function in the development of type 2 diabetes in the elderly.
This work was supported in part by National Institutes of Health (NIH) grants RO1-DK-67536 and R21-RR-024905 to R.N.K., RO1-AG-026094 to H.S., and RO1-DK-60581 to R.G.M. C.H. was supported by an NIH Training Grant (1RL9-EB-008539-01) (SysCODE). H.S. was also supported by a Senior Scholar in Aging award from the Ellison Medical Research Foundation. D.K. is the recipient of a Research Fellowship (Manpei Suzuki Diabetes Foundation, Japan) and a Juvenile Diabetes Research Foundation Postdoctoral Fellowship. C.W.L. is supported by 1-K99-DK-090210-01.
No potential conflicts of interest relevant to this article were reported.
C.H. conceived and designed the experiments, performed the experiments, analyzed data, and wrote the article. D.K., C.W.L., and B.M. performed the experiments; contributed reagents, material, and mice; and analyzed data. J.H. performed the experiments and analyzed data. S.R.K. contributed reagents and performed experiments. R.G.M. and H.S. contributed reagents, material, and mice; analyzed data; and wrote the article. R.N.K. conceived and designed the experiments, performed the experiments, analyzed data, and wrote the article.
The authors thank Lindsay Huse, Elizabeth Morgan, and Kellianne Parlee (Joslin Diabetes Center, Boston, MA) for assisting with preparation of the article and Dr. Hannah Welters (Joslin Diabetes Center, Boston, MA) for reading the manuscript.
This article contains Supplementary Data online at http://diabetes.diabetesjournals.org/lookup/suppl/doi:10.2337/db09-1379/-/DC1.