Loss of HGF/c-Met Signaling in Pancreatic β-Cells Leads to Incomplete Maternal β-Cell Adaptation and Gestational Diabetes Mellitus

PancMet KO mice were generated by crossing mice homozygous for the floxed c-Met allele (18) with Pdx1-Cre transgenic mice (19) as previously described (15). Eight- to ten-week-old female c-Met^lox/lox;Pdx1-Cre (PancMet KO) mice and their wild-type littermates c-Met^lox/lox without Pdx1-Cre transgene were used in these studies. Cre expression driven by the Pdx-1 promoter does not lead to detrimental effects in β-cells of adult mice (15,20,21). Pregnant PancMet KO and wild-type mice were generated by timed mating. Nonpregnant female littermates were used as controls. All the studies were performed with the approval of, and in accordance with, guidelines established by the University of Pittsburgh Institutional Animal Care and Use Committee.

Blood obtained by retroorbital bleed was analyzed for glucose by a portable glucometer (Medisense, Bedford, MA) and plasma insulin by radioimmunoassay (Linco Research, St. Louis, MO) (11). Intraperitoneal glucose tolerance test was performed in mice fasted 16–18 h injected with 2 g glucose/kg body wt i.p., and insulin tolerance test was performed in random-fed mice injected with 1.5 units human insulin/kg body wt i.p. (Humulin; Lilly, Indianapolis, IN) (12).

Insulin release from 10 islet equivalents (IEs) (1 IE = 125 μm diameter) was measured after incubation in 5 or 20 mmol/L glucose for 30 min (12). Insulin secretion results are expressed as percentage of total islet insulin content. Insulin content was measured in islets homogenized in 0.18 mol/L HCl in 70% ethanol and extracted at −20°C for 24 h. Tubes were centrifuged at 2,500 rpm for 10 min at 4°C, and the supernatant was stored at −20°C until insulin determination by radioimmunoassay (11).

Paraffin-embedded pancreatic sections were immunostained for insulin and 5-bromo-2-deoxyuridine (BrdU) (Amersham, Piscataway, NJ) as previously described (11,14). β-Cell mass was measured in three insulin-stained pancreas sections per mouse using ImageJ software (National Institutes of Health, Bethesda, MD) (11,14). Islet number was quantified manually in these three insulin-stained sections in a blinded fashion using an Olympus CH2 upright microscope with an intraocular calibrated grid (22).

BrdU incorporation in β-cells was measured in pancreas sections of mice injected intraperitoneally with BrdU, killed 6 h later, and stained for insulin and BrdU (11). β-Cell death was determined in pancreas sections stained for insulin and using the terminal deoxynucleotidyl transferase–mediated dUTP nick-end labeling (TUNEL) method (Promega, Madison, WI) (15). Images of pancreatic islets were obtained in a Confocal Olympus Fluoview 1000 (Center for Biological Imaging, University of Pittsburgh). β-Cell size was determined as previously described (22). Immunostainings for STAT5 (Cell Signaling, Danvers, MA) and p27 (BD Pharmingen, San Jose, CA) were performed using methods previously described (15).

Mouse islets were isolated following injection of collagenase P through the pancreatic duct as previously described (11). Human islets were provided by the Integrated Islet Distribution Program and Juvenile Diabetes Research Foundation Basic Science Islet Distribution programs. Analysis of c-Met, forkhead box M1 (FoxM1), insulin, GLUT2, glucokinase, and Pdx-1 mRNA expression was performed by real-time PCR using specific primers (primer sequences available on request) (15).

HGF levels were measured in serum obtained from pregnant and nonpregnant wild-type mice using ELISA (Abcam, Cambridge, MA) and following the manufacturer’s instructions.

Mouse islet extracts were separated on 7.5–12% SDS-PAGE, transferred to an Immobilon-P membrane (Millipore, Bedford, MA), blocked in 5% nonfat dry milk, and then incubated with primary antibodies against c-Met and prolactin (PRL) receptor (PRLR) (Santa Cruz Biotechnology, Santa Cruz, CA), p27 (BD Pharmingen), tubulin (Calbiochem, La Jolla, CA), and HGF (R&D Systems, Minneapolis, MN). After several washes, blots were incubated with peroxidase-conjugated secondary antibodies followed by chemiluminescence detection (Amersham) (15).

Mouse and human islet cells were cultured as previously described (14) and incubated with different doses of dexamethasone with or without HGF for 24 h and then fixed in 2% paraformaldehyde. β-Cell death was determined by TUNEL assay and insulin and DAPI staining. At least 2,000 β-cells were counted per well. Images were obtained in the Confocal Olympus Fluoview 1000 microscope.

Data are presented as means ± SE. Statistical analysis was performed using unpaired two-tailed Student t test. P < 0.05 was considered statistically significant.

Islets from wild-type pregnant mice displayed significant HGF upregulation at GD15 (Fig. 1A). In addition, circulating HGF was also increased in pregnant wild-type mice at GD15 and -19 (Fig. 1B). The increase in systemic and islet HGF correlated with significant upregulation of islet c-Met mRNA and protein at GD15 (Fig. 1C and D), when maximal maternal β-cell proliferation occurs in rodents (23–27). To analyze whether pregnancy hormones regulate c-Met in islets, we determined the effect of 17β-estradiol, PRL, progesterone, and the synthetic glucocorticoid dexamethasone on islet c-Met mRNA levels by real-time PCR. As shown in Fig. 1E, 17β-estradiol significantly increased islet c-Met mRNA, suggesting a potential role for this hormone in potentiating HGF/c-Met signaling in islets during pregnancy. Taken together, these results indicate that both components of the HGF–c-Met axis are upregulated in the islet during pregnancy and raise the question of whether local or systemic HGF or both have a role in maternal β-cell adaptation during pregnancy.

Adult PancMet KO mice display normal glucose homeostasis and β-cell mass and proliferation under basal conditions (15). To address whether HGF/c-Met signaling is important for maternal β-cell expansion during pregnancy, we analyzed β-cell mass in 8- to 10-week-old pregnant and nonpregnant female PancMet KO mice and wild-type littermates. Body weight and the average number of progeny throughout pregnancy were not significantly different between both types of mice (Fig. 2A and B). β-Cell mass was also similar at GD11 and -15 and in nonpregnant conditions (Fig. 2C). In contrast, PancMet KO mice displayed significantly decreased β-cell mass at GD19, when maximal β-cell mass expansion occurs in normal pregnant mice (23–27) (Fig. 2C). This decrease in β-cell mass persisted at postpartum day (PPD)4 (Fig. 2C). Since recent studies indicate that islet number is increased in pregnant mice (27) and PancMet KO mice display decreased β-cell mass at GD19, we quantified this parameter in both types of pregnant mice. No alteration in islet number was observed between wild-type and PancMet KO mice at GD19 (Fig. 2D), suggesting that a decrease in islet size rather than total islet number might be responsible for the decrease in β-cell mass in pregnant PancMet KO mice.

We then explored whether alterations in β-cell size, proliferation, and apoptosis contributed to the impaired β-cell expansion observed in pregnant PancMet KO mice. At GD19, β-cell size was similar in PancMet KO and wild-type mice (Fig. 2E), indicating that a change in β-cell size was not responsible for the incomplete β-cell mass expansion observed in PancMet KO mice at this gestational age. β-Cell proliferation was not significantly different in nonpregnant or pregnant PancMet KO and wild-type littermates at GD11 and -19 (Fig. 3E). However, at GD15, when maximal β-cell proliferation occurs in mice (23–27), PancMet KO mice showed a marked decrease in β-cell replication compared with wild-type mice (Fig. 3A–E). These results indicate that HGF/c-Met signaling in the β-cell is required for adaptive β-cell replication and mass expansion during pregnancy. In addition, β-cell replication was also significantly decreased in PancMet KO mice at PPD4 (Fig. 3E), when β-cell mass was decreased in these mice (Fig. 2C).

Placental lactogen (PL)/PRL signaling through the PRLR is required for the maternal adaptive increase in β-cell proliferation during pregnancy (1,26). Since pregnant PancMet KO mice displayed decreased β-cell proliferation at GD15, we measured PRLR levels in islets from wild-type and PancMet KO mice at this gestational day. Islet PRLR mRNA was significantly downregulated both in nonpregnant and pregnant PancMet KO mice (Fig. 4A). Importantly, Western blot analysis also showed a decrease in PRLR in PancMet KO mouse islets at GD15, while the levels in nonpregnant mice were not significantly different (Fig. 4B). PRL increases β-cell proliferation through phosphorylation/activation and enhanced nuclear localization of STAT5 (28). Insulin and STAT5 coimmunostaining in pancreas sections from wild-type mice at GD15 clearly showed nuclear localization of this transcription factor in β-cells (Fig. 4C). However, PancMet KO mouse islets at GD15 displayed almost exclusive cytoplasmic STAT5 staining in β-cells (Fig. 4C), suggesting that the decrease in PRLR in pregnant PancMet KO mouse islets could result in diminished islet PL/PRL signaling and reduced STAT5 activation, leading to decreased maternal β-cell proliferation. The transcription factor FoxM1 is a PRL cell cycle target that regulates maternal β-cell expansion during pregnancy (24). Importantly, PancMet KO mouse islets at GD15 failed to upregulate FoxM1 mRNA expression compared with wild-type mouse islets (Fig. 4D). Furthermore, downregulation of the member of the Cip/Kip family of cyclin-dependent kinases inhibitors, p27, is associated with FoxM1 upregulation and enhanced β-cell proliferation during pregnancy (23,24). As previously shown, islets from wild-type mice at GD15 displayed significantly decreased p27 compared with nonpregnant mice (Fig. 4E) (23,24). This decrease was not observed in PancMet KO mice at GD15 (Fig. 4E). Insulin and p27 coimmunostaining of pancreatic sections from wild-type and PancMet KO mice at GD15 showed increased p27 expression in β-cells of PancMet KO mice compared with wild-type littermates (Fig. 4F). Collectively, these results indicate that proliferation, PRLR levels, STAT5 nuclear localization, and cell cycle regulation in β-cells during pregnancy require HGF/c-Met signaling.

β-Cell death occurs at the end of gestation and the early postpartum period to normalize the expanded maternal β-cell mass (1,29). Based on our previous observations that HGF/c-Met signaling deficiency in β-cells leads to enhanced sensitivity to cell death induced by streptozotocin and cytokines (15), we examined whether β-cell apoptosis was altered in PancMet KO mice during pregnancy. As shown in Fig. 5C, β-cell apoptosis was significantly increased in wild-type and PancMet KO mice at GD19 and PPD4 compared with nonpregnant littermates. However, apoptosis in PancMet KO β-cells was prematurely and significantly increased at GD15 compared with nonpregnant mice or pregnant wild-type mice at the same gestational age (Fig. 5A–C). This suggests that the prosurvival effects of HGF in the β-cell are required for proper maternal β-cell expansion during pregnancy.

Circulating steroid hormone levels rise during the last week of gestation in rodents and have been postulated to induce β-cell apoptosis and normalize β-cell mass in late gestation and postpartum (2,30–32). It could be possible that HGF/c-Met signaling deficiency renders β-cells more sensitive to apoptosis induced by low levels of steroid hormones, as we have previously observed with cytokines (15). Indeed, PancMet KO β-cells were significantly more sensitive to the cytotoxic effects of a low dose of the synthetic steroid hormone dexamethasone that did not induce β-cell death in wild-type islets (Fig. 5D–J). A higher dose of dexamethasone was equally cytotoxic in both PancMet KO and wild-type β-cells (Fig. 5J). Interestingly, dexamethasone-induced cytotoxicity was completely abrogated by HGF in human β-cells (Fig. 5K–O). These results indicate that HGF is a protective factor for the β-cell against cytotoxic doses of steroid hormones and that loss of HGF/c-Met signaling in β-cells can confer increased sensitivity to cytotoxicity induced by lower doses of these hormones.

Since loss of c-Met from β-cells leads to changes in maternal β-cell mass expansion, proliferation, and survival, we next analyzed glucose homeostasis in PancMet KO mice during pregnancy. Blood glucose was significantly higher in PancMet KO mice at GD19 compared with nonpregnant mice or pregnant wild-type littermates at this gestational day (Fig. 6A). This increase in blood glucose correlated with diminished plasma insulin (Fig. 6B). In addition, PancMet KO mice also displayed significantly increased fasting blood glucose compared with wild-type mice at GD19 (Fig. 6C).

Glucose tolerance was impaired in wild-type mice at GD19 compared with nonpregnant mice (Fig. 6D). Importantly, glucose tolerance was further impaired in PancMet KO mice at this gestational day (Fig. 6C and D), when β-cell mass is diminished in these mice (Fig. 2C). The enhanced impairment in glucose tolerance in pregnant PancMet KO mice was not related to changes in insulin sensitivity, since both types of pregnant mice responded similarly to insulin tolerance tests at GD18 (Fig. 6E). The impaired glucose tolerance in PancMet KO mice was maintained at PPD4 (Fig. 6D) and, surprisingly, was also present at GD15 (Fig. 6D), when no change in β-cell mass was observed (Fig. 2C). This suggests that the absence of c-Met from β-cells could also alter insulin secretion during pregnancy independent of its effects on β-cell mass and proliferation.

We next analyzed mRNA expression of β-cell function markers including glucokinase, GLUT2, insulin, and Pdx-1 in islets from these mice at GD19. No significant alterations in glucokinase mRNA levels were observed between pregnant PancMet KO and wild-type littermates (Fig. 6F). However, GLUT2, insulin, and Pdx-1 mRNA were markedly and significantly increased in wild-type mouse islets but not in islets from pregnant PancMet KO mice at GD19 (Fig. 6F). The increased insulin mRNA in pregnant wild-type mouse islets correlated with a significant increase in islet insulin content; however, this increase was not present in islets from pregnant PancMet KO mice (Fig. 6G). Since pregnant PancMet KO mice display increased blood glucose and decreased plasma insulin, glucose tolerance, and islet insulin content, we analyzed GSIS in isolated islets from these mice. Islets from nonpregnant PancMet KO and wild-type mice displayed similar increases in GSIS (Fig. 6H). Islets from pregnant wild-type mice at GD19 showed significantly enhanced GSIS compared with islets from nonpregnant mice (Fig. 6H). However, this further increase in GSIS was not present in islets from pregnant PancMet KO mice (Fig. 6H). Collectively, these results suggest that HGF/c-Met signaling is required for the functional adaptive response of maternal mouse β-cells to pregnancy.

The current study provides the first direct evidence that HGF/c-Met signaling is essential for complete maternal β-cell adaptation during pregnancy. Loss of c-Met in the pancreas diminishes maternal β-cell proliferation and enhances apoptosis, leading to diminished β-cell mass expansion, hyperglycemia, hypoinsulinemia, and glucose intolerance in pregnant mice: hallmarks of GDM. Therefore, these observations indicate that impairment/attenuation of HGF/c-Met signaling might be a pathogenic mechanism of GDM and could be a potential target for detection, prevention, and therapy in diabetes.

Previous studies from our group and others have shown that HGF is a mitogenic, prosurvival, and insulinotropic factor for the β-cell (11–16). Although c-Met loss from the adult β-cell or pancreas does not lead to marked alterations in glucose or β-cell homeostasis under basal conditions (15,33,34), recent studies have shown that c-Met absence facilitates cytokine-mediated β-cell death and accelerates the onset of diabetes in a mouse model of multiple low-dose streptozotocin administration (15). Taken together, these previous results indicate that HGF may have a physiological role in the β-cell in situations of stress-mediated hyperglycemia/diabetes. During pregnancy, insulin resistance develops and leads to an adaptive process in maternal β-cells that includes enhanced insulin secretion and β-cell mass expansion (1,2,35–40). When β-cell expansion or hyperfunction fails to compensate during pregnancy, hyperglycemia/diabetes occurs, suggesting that defective maternal β-cell adaptation can lead to GDM (2–5,23–27). Therefore, there is a growing interest in deciphering the extracellular and intracellular regulators of maternal β-cell mass expansion and hyperfunction to find therapeutic targets not only for preventing/correcting GDM but also for enhancing β-cell regeneration in other diabetic conditions.

HGF expression is markedly enhanced in islet endothelium at GD15, when both islet vascular growth and β-cell proliferation occur in rodents (17). HGF, produced by the villous mesenchyme of the placenta, is essential for placental organogenesis and maintenance of pregnancy (41,42). In addition, circulating HGF is increased in the third trimester of pregnancy in women (16). In the current study, we observed an increase in HGF in both the islet and the circulation at GD15 in mice when maximal β-cell proliferation occurs. More importantly, c-Met mRNA and protein are also upregulated at the same gestational day, suggesting increased HGF/c-Met signaling in islets during gestation. Several hormones such as progesterone, PL, estrogen, glucocorticoids, and cytokines such as tumor necrosis factor-α all present during pregnancy (1,2,30,43) are known to upregulate c-Met expression in ovary, breast, and endometrium (44,45). We have recently shown that cytokines enhance c-Met expression in islets (15), and here we show that 17β-estradiol enhances c-Met mRNA expression as well, suggesting that cytokines and hormones upregulated during gestation can modulate c-Met levels to enhance HGF signaling in β-cells.

In vitro studies have shown that purified proliferating islet endothelial cells stimulate β-cell proliferation through secretion of HGF (17). Since HGF expression is enhanced at GD15 in the islet endothelium, these results suggest a potential role of HGF in maternal β-cell proliferation and expansion. Indeed, our results clearly indicate that the absence of c-Met from β-cells leads to incomplete maternal β-cell mass expansion during pregnancy. This decrease is not due to changes in islet number or β-cell size. However, maternal β-cell proliferation is clearly diminished at GD15 in mice deficient for c-Met in β-cells, a time when maximal β-cell proliferation is observed in wild-type mice. Importantly, since β-cell mass is diminished in PancMet KO mice at GD19 but not at GD15, these results indicate that maternal β-cell growth adaptation is dependent on HGF/c-Met signaling at GD15 and subsequent GDs but is HGF/c-Met signaling independent prior to GD15. Downregulation of p27 potentially contributes to the increase in β-cell replication during pregnancy (23,24). In addition, FoxM1 upregulation is required for the enhanced β-cell proliferation, β-cell mass expansion, and p27 downregulation during gestation in mice (24). Importantly, islets lacking c-Met and with diminished β-cell proliferation at GD15 failed to increase FoxM1 mRNA expression and downregulate p27 levels, suggesting that HGF signaling might regulate these mitogenic intracellular signals. Indeed, HGF has been shown to downregulate p27 in skeletal muscle myoblasts, and this could occur in β-cells as well (46). However, whether HGF regulates FoxM1 in β-cells or any other tissue is unknown and warrants further studies. The transcription factor hepatocyte nuclear factor (HNF)-4α is required for maternal β-cell mass expansion and activation of the Ras/extracellular signal–related kinase (ERK) signaling cascade in islets (25). HGF strongly induces ERK activation in β-cells (14); however, whether HNF-4α is involved in HGF-induced ERK activation or whether HGF requires HNF-4α and ERK activation for β-cell replication during pregnancy is currently unknown and also warrants further studies.

We cannot rule out the possibility that HGF/c-Met plays an indirect role, with the major effect being that the embryonic absence of c-Met leads to alteration of downstream factor(s) primarily responsible for the defect observed in pregnant PancMet KO mice. Another possibility is that HGF regulates the action of hormones known to participate in maternal β-cell adaptation during pregnancy. Accumulating evidence supports the implication of PRL/PL in this gestational adaptive response (1,2,26,47). Interestingly, PancMet KO mouse islets displayed decreased PRLR levels at GD15, suggesting that PRL/PL signaling might be reduced during pregnancy, when circulating levels of these hormones are markedly high. It is important to note that heterozygous PRLR mice displaying a partial decrease in PRLR expression show normal glucose and β-cell homeostasis in basal conditions but impaired β-cell proliferation and expansion during pregnancy, leading to dysregulated glucose homeostasis (26). Recent evidence indicates that PRL/PL induces serotonin synthesis in islets which, in turn, functions in a paracrine-autocrine fashion to stimulate β-cell proliferation during pregnancy (27). Therefore, it is plausible that the decrease in PRLR levels in pregnant PancMet KO islets could affect serotonin synthesis and signaling in the β-cell of these mice during pregnancy, leading to incomplete maternal β-cell adaptation. PRL/PL induces β-cell proliferation through activation and nuclear localization of STAT5 (26,48). Indeed, although STAT5 levels appear unchanged by immunostaining, its intracellular localization is altered in β-cells, being nuclear in pregnant wild-type mice and more diffuse and cytoplasmic in PancMet KO mice, suggesting perhaps diminished PRL/PL signaling. However, c-Met has also been shown to activate STAT proteins and physically associate with STAT5 (49). It is possible that the decreased activation of STAT5 in pregnant PancMet KO mouse β-cells merely reflects the loss of c-Met signaling or a combined decrease in PRLR and c-Met signaling in these cells.

A less explored cellular process in β-cell adaptation during pregnancy is death/apoptosis. It has been shown that apoptosis contributes to the involution of β-cell mass at the end of gestation and the early postpartum period in rodents (29). However, the extracellular and intracellular signals that regulate survival/death in β-cell mass expansion and attrition during gestation/postpartum are unknown. Steroid hormones have been shown to counteract the mitogenic, prosurvival, and functional effects of PRL/PL in β-cells in vitro (31,32,50), suggesting that upregulation of steroid hormones at the end of pregnancy could be responsible for the increase in β-cell death and normalization of β-cell mass and function. In our studies, we have found that the absence of HGF/c-Met signaling in β-cells leads to increased β-cell death at GD15, when β-cell death is not altered in wild-type mice, suggesting enhanced sensitivity to mild increases in steroid hormone levels that could occur at this gestational day (51,52). Indeed, PancMet KO β-cells are more sensitive than wild-type cells to death induced by low doses of dexamethasone, and HGF protects human β-cells from dexamethasone-induced cell death in vitro, further reinforcing the possibility that c-Met deficiency might enhance the susceptibility of β-cells to the negative effects of mild hormonal changes (51,52).

A decrease in β-cell mass and proliferation during pregnancy can lead to negative outcomes in glucose homeostasis and development of GDM (2–5,23–27). Although the decrease in β-cell mass in pregnant PancMet KO mice might contribute to the increased blood glucose, hypoinsulinemia, and glucose intolerance in these mice, alterations in β-cell function due to the absence of HGF/c-Met signaling could also be present. HGF overexpression in β-cells increases insulin secretion, upregulates GLUT2 and glucokinase mRNA, and enhances glucose transport and metabolism in β-cells (12). During pregnancy, it has been reported that upregulation of GLUT2, glucokinase, insulin, and GSIS occurs in rodent islets (36,37). We found that islets from wild-type mice at GD19 display upregulated GLUT2 and Pdx-1 mRNA, islet insulin content, and GSIS. However, these upregulations do not occur in PancMet KO islets. Taken together, these results suggest that HGF/c-Met signaling in β-cells might be required for the hyperfunctional β-cell adaptation during pregnancy. Whether this is a direct effect of HGF or indirect due to decreased PRL/PL signaling is unknown at present. It is important to note that our current studies used Pdx1-Cre transgenic mice to delete c-Met from the pancreas and the β-cell and these mice have been shown to display Cre expression in parts of the brain known to participate in the regulation of glucose homeostasis (53). This could lead to potential elimination of c-Met in different regions of the brain in PancMet KO mice, resulting in alterations in insulin sensitivity and glucose homeostasis. Therefore, β-cell nonautonomous effects could be responsible for the changes in glucose homeostasis in pregnant PancMet KO mice. However, adult PancMet KO mice display normal glucose homeostasis in basal conditions (15). Furthermore, body weight and insulin sensitivity are not different between pregnant wild-type and PancMet KO mice, suggesting that the alterations in blood glucose, plasma insulin, and glucose tolerance in pregnant PancMet KO mice are not related to changes in insulin sensitivity or body weight gains. In addition, ex vivo GSIS is decreased in isolated islets from pregnant PancMet KO mice compared with pregnant wild-type mouse islets. Therefore, these findings suggest a cell-autonomous effect of HGF/c-Met signaling on β-cell function during pregnancy.

In conclusion, these studies are the first to identify HGF as a regulator of maternal β-cell adaptation during pregnancy. Alterations in expression or the presence of mutations/polymorphisms that reduce HGF signaling in the β-cell might be important targets for detection, prevention, and treatment of GDM.

This study was supported by grants from the American Diabetes Association (1-10-BS-59) and the National Institutes of Health (NIH) (DK067351 and DK077096 to A.G.-O. and DK072264 to R.C.V.). C.D. was the recipient of a research fellowship from the Lawson Wilkins Pediatric Endocrine Society. S.E. was a research fellow partly supported by NIH Research Training Grant T32DK07052-32. Human islets were provided by the Juvenile Diabetes Research Foundation and the NIH/National Center for Research Resources– and National Institute of Diabetes and Digestive and Kidney Diseases–supported Integrated Islet Distribution Program.

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

C.D. and S.E. performed research, analyzed data, designed research, and contributed to discussion. J.C.A.-P. performed research, analyzed data, and contributed to discussion. T.R., S.V., V.S., and G.P.C. performed research. L.C.A. and R.C.V. designed research, analyzed data, and contributed to discussion. A.G.-O. designed research, analyzed data, contributed to discussion, and wrote the manuscript. A.G.-O. 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.

The authors are grateful to Dr. Douglas A. Melton of the Harvard Stem Cell Institute, Harvard University, and Dr. Snorri Thorgeirsson of the Laboratory of Experimental Carcinogenesis, National Cancer Institute, for providing Pdx-Cre and loxP-c-Met mice, respectively. The authors are also grateful to Andrew F. Stewart, Donald K. Scott, and Nathalie Fiaschi-Taesch from the Division of Endocrinology and Dorothy Becker from the Department of Pediatrics, University of Pittsburgh, for thoughtful discussions.