Autocrine Action of IGF2 Regulates Adult β-Cell Mass and Function

Glucose homeostasis depends on the balance between insulin secretion by pancreatic β-cells and insulin action on peripheral tissues (1). In response to the development of insulin resistance in muscle, liver, and fat, pancreatic β-cells increase their insulin secretion capacity in order to maintain normoglycemia. This compensatory response depends not only on an enhanced secretion capacity of individual β-cells but also, at least in rodents, on an increase in their number (2). In adult life, this plasticity is essential to maintain normoglycemia in insulin resistance conditions associated with obesity, pregnancy, and aging (3–5). Failure of this β-cell compensatory response leads to the onset of type 2 diabetes.

The mechanisms by which insulin resistance in peripheral tissues induces compensatory insulin secretion capacity are incompletely understood. Their identification is, however, of highest interest for the development of novel therapies for diabetes. Evidence suggests that both circulating and nervous signals are involved. Glucose was one of the first signals identified to induce β-cell proliferation (6–8) through a signaling pathway that requires glucose metabolism, insulin secretion, and activation of the insulin receptor (IR)/Akt pathway (9,10). Incompletely characterized soluble factors, distinct from glucose, are produced by insulin-resistant hepatocytes to increase β-cell mass (11,12), whereas bile acids can increase β-cell secretion capacity, independent of a change in β-cell number (13). Nerve connections between the liver and the islets can also potently stimulate β-cell proliferation (14).

Increased β-cell proliferation and secretion capacity is a hallmark of pregnancy both in rodents and humans (3,15). This adaptive response is largely controlled by prolactin and placental lactogen acting through the prolactin receptor (16) and involves the synthesis, secretion, and autocrine action of serotonin binding to the 5HTR2B receptor to control proliferation (17) and to the ionotropic 5HT3 receptor to increase glucose-stimulated insulin secretion (GSIS) (18). Estrogens also contribute to β-cell proliferation, particularly through estradiol (E2) activation of G-protein–coupled receptor 30 (19).

β-Cell proliferation and function are also regulated by IR and insulin-like growth factor 1 receptor (IGF1R) signaling (20,21), as well as by the gluco-incretin hormones glucose-dependent insulinotropic polypeptide (GIP) and glucagon-like peptide 1 (GLP-1), which also protect β-cells against apoptosis (22,23). In recent years, we have explored the link between gluco-incretin signaling and IGF1R signaling in β-cells. These studies originated from comparative transcriptomic analysis of islets from control (Ctrl) mice and mice with genetic inactivation of the GIP and GLP-1 receptors (24), which display reduced GSIS despite normal insulin content, and increased susceptibility to cytokine-induced apoptosis. These studies demonstrated that gluco-incretin action in β-cells is mediated by activation of IGF1R expression and signaling. They further showed that IGF1R signaling, but not IR signaling, was dependent on insulin-like growth factor 2 (IGF2) synthesis and secretion by the β-cells. This autocrine mechanism mediates all the trophic actions of gluco-incretins on β-cells (25,26). Indeed, the effects of GLP-1 on β-cell proliferation and protection against cytokine-induced apoptosis were suppressed by small interfering RNA–mediated silencing of Igf2 in β-cells by immunoneutralization of secreted IGF2 or by β-cell–specific inactivation of the Igf1r gene. In these experimental conditions, GSIS was also markedly reduced, indicating an important role of this autocrine loop in the controlling β-cell secretion capacity (25,26). In further support of an important role of IGF2 in β-cell biology is its high level of expression in mouse and human β-cells, and its absence from non–β-cells (26,27), as well as the acute regulation of its biosynthesis and secretion by nutrients, in particular glutamine (28). Thus, the IGF2/IGF1R autocrine loop integrates different nutrient-related cues, glutamine and gluco-incretins, in controlling β-cell mass and function. The in vivo physiological importance of the autocrine action of IGF2 is, however, so far untested.

Therefore, here we examined the physiological role of the autocrine action of IGF2 on the β-cell adaptation in aging and pregnancy and in response to acute induction of insulin resistance using mice with β-cell–specific inactivation of igf2. We show a sex-dependent role of β-cell IGF2 in aging and upon metabolic stress; an essential role in β-cell mass adaptation during pregnancy and a significant contribution of IGF2 to β-cell expansion in response to pharmacologic induction of insulin resistance.

Interferon-γ, prolactin, and E2 were purchased from Sigma-Aldrich (St. Louis, MO); tumor necrosis factor-α was purchased from Millipore (Billerica, MA); interleukin-1β was purchased from Calbiochem (Billerica, MA); and exendin-4 was purchased from Bachem (Bubendorf, Switzerland).

Igf2 floxed mice were generated by homologous recombination in embryonic stem cells (Fig. 1A). The neo cassette was removed by crossing Igf2^lox/+ mice with Flp-deleter mice. Male Igf2^lox/+ mice were crossed with female Ins1-Cre mice (29) to generate Igf2^lox/+;Ins1-Cre (βIGF2KO) and Igf2^lox/+ (control [Ctrl]) mice. Mice with homozygous β-cell deletion of Igf2 were generated by crossing Igf2^lox/+;Ins1-Cre mice with Igf2^lox/+ mice to obtain Igf2^lox/lox;Ins1-Cre mice and Ctrl littermates (βIgf2^lox/lox). Mice were on a mixed 129S6/C57BL/6 background. Genotyping for the presence of the Igf2^lox allele was performed by PCR analysis with primers P1 and P2, and monitoring of exon 4 deletion with primers P1 and P3 (Fig. 1A). All studies were performed with littermates. Mice were fed a standard rodent chow (Diet 3436; Provimi Kliba AG) or a high-fat diet (HFD) (Purified Diet 235HF; Safe Diets, Augy, France).

Total RNA was isolated using RNeasy Plus Micro Kit or miRNeasy kit (Qiagen, Hombrechtikon, Switzerland). Mature microRNA (miR) levels were assessed by quantitative RT-PCR (qRT-PCR) using the miScript II RT kit with miScript HiFlex Buffer and the miScript SYBR Green PCR Kit (Qiagen), and normalized to the levels of U6 RNA (19); mRNAs levels were measured by qRT-PCR analysis and normalized to the levels of TATA binding protein (TBP) (28). Forward (F) and reverse (R) primers were as follows: IGF1R (F), 5′-TGGTGACCGGCTACGTGAAG-3′ and (R), 5′-CAAAGTACATCTTTCCGGACC-3′; IGF2 (F), 5′-TGGTGCTTCTCATCTCTTTGG-3′ and (R), 5′-GAACAGACAAACTGAAGCGTG-3′; Birc5 (F), 5′-GAATCCTGCGTTTGAGTCGT-3′ and (R), 5′-AAAACACTGGGCCAAATCAG-3′; tryptophan hydroxylase (Tph1) (F), 5′-TTCCAGGAGAATCATGTGAGC3′, and (R), 5′- CATAACGTCTTCCTTCGCAGT-3′; CyclinB1 (F), 5′-TGGCCTCACAAAGCACATGA-3′ and (R), 5′-GCTGTGCCAGCGTGCTAATC-3′; CyclinB2 (F), 5′GGCTGGTCAAGTCCATTCC-3′, and (R), 5′-GTCCATGATGCCAATGCACA-3′; and TBP (F), 5′-ATCCCAAGCGATTTGCTGC-3′, and (R), 5′- ACTCTTGGCTCCTGTGCACA-3′.

Islet isolation, insulin secretion experiments, and islet cultures on extracellular matrix–coated plates for apoptosis and proliferation assays were performed as described (25,26). For transfection with scrambled or anti–miR-338-3p islets cells were first dissociated by trypsinization, cultured overnight, plated on poly-l-lysine hydrobromide-coated glass coverslips (19). Assessment of Ki67-positive cells was performed as described previously (19).

Blood glucose and plasma insulin measurements, glucose tolerance tests (overnight-fasted mice, injection of glucose 2 g/kg i.p.), and insulin tolerance tests (5-h fasted mice, 0.3 units/kg insulin i.p.) were performed as described previously (30). IGF2 ELISA was purchased from R&D Systems (Minneapolis, MN).

Histomorphometric analysis was performed on 5-μm-thick sections of 4% paraformaldehyde fixed pancreas. Immunodetection of insulin and Ki67 and histomorphometric analysis of β-cell surface area using ImageJ software were performed as described previously (30). Five to six sections per pancreas and four to six pancreata were analyzed per group of mice, representing a total of 400–500 islets analyzed per condition. The cell mass was calculated based on individual pancreas weight.

All breeding and mouse experiments were approved by the Service Vétérinaire du Canton de Vaud.

Results are reported as the mean ± SD. Statistical analysis used unpaired an Student t test and one-way ANOVA, which were followed by post hoc pairwise multiple-comparison procedures Tukey test (*P < 0.05, **P < 0.01, ***P < 0.001).

To generate mice with β-cell–specific inactivation of Igf2, exon 4 of the Igf2 gene was flanked by loxP sites (Fig. 1A), and inactivation of the Igf2 gene was achieved by crossing male Igf2^lox/+ mice with female Ins1Cre mice (29) to induce recombination only in islet β-cells (Fig. 1B). This led to suppression of Igf2 expression (Fig. 1C) since this gene is expressed only from the paternal allele (31) and expression of Igf2 was normal in other tissues, including the hypothalamus of βIGF2KO mice (Fig. 1D). We also generated mice with inactivation of both maternal and paternal Igf2 alleles (Igf2^lox/lox mice crossed with Igf2^lox/lox;Ins1Cre mice). This led to the same suppression of islet Igf2 mRNA expression, confirming its paternal origin.

We next assessed whether Igf2 inactivation would impact isolated β-cell function as predicted from our previous studies (25,26). GSIS by islets isolated from βIGF2KO mice was indeed significantly reduced compared with Ctrl islets but was normal at low glucose concentrations (Fig. 2A). Similarly, the protective effect of GLP-1 against cytokine-induced apoptosis was markedly reduced in β-cells from βIGF2KO mice (Fig. 2B) as was the proliferation induced by exendin-4 (Fig. 2C). As β-cell mass and insulin content were normal in the islets from young adult βIGF2KO mice (see below), these data are in agreement with results of our previous studies showing that autocrine secretion of IGF2 is required to maintain normal GSIS and for the effects of exendin-4 on proliferation and cytokine-induced apoptosis.

To assess glucose homeostasis in βIGF2KO mice, we generated cohorts of male and female Ctrl and βIGF2KO mice and fed them with a normal chow (NC) or an HFD from 6 weeks of age. Body weight gains of male and female Ctrl and βIGF2KO mice were the same when the mice were fed either diet (Supplementary Fig. 1A and B).

In NC-fed male mice, glucose tolerance, insulin sensitivity, and in vivo GSIS were the same in young (8- to 10-week-old; Supplementary Fig. 2A–C) and old (24-week-old; Supplementary Fig. 2D–F) Ctrl and βIGF2KO mice.

In young NC-fed female mice, no difference in glucose tolerance, insulin sensitivity, or GSIS was observed between Ctrl and βIGF2KO mice (Fig. 3A–C). However, in old female βIGF2KO mice, plasma insulin levels were lower than in Ctrl mice in the basal state and following intraperitoneal glucose injection. This was, however, compensated for by an increase in insulin sensitivity to maintain normal glucose tolerance (Fig. 3D–F). Circulating levels of IGF2 were below the detection limit of the ELISA (31.2 pg/mL) in both Ctrl and βIGF2KO mice.

When fed with an HFD from 6 weeks of age, 24-week-old male βIGF2KO mice developed the same level of glucose intolerance as Ctrl mice (Fig. 4A and Supplementary Fig. 2D). However, they displayed lower insulinemia in the basal state and following an intraperitoneal glucose challenge (Fig. 4B). An insulin tolerance test showed small but significantly higher insulin sensitivity in the mutant mice (Supplementary Fig. 3). Histomorphometric analysis showed no difference in β-cell mass relative to body weight in NC- or HFD-fed male Ctrl or βIGF2KO mice (Fig. 4C).

In female βIGF2KO mice, HFD feeding induced a higher glucose intolerance than in Ctrl mice (Fig. 4D), and this was associated with lower plasma insulin levels in basal and glucose-challenged conditions (Fig. 4E). Insulin tolerance test results showed a trend toward increased insulin sensitivity (Supplementary Fig. 3). HFD feeding led to reduced β-cell mass in βIGF2KO mice compared with Ctrl female mice (Fig. 4F). Here, β-cell mass is reported relative to body weight and is reduced by approximately twofold, whereas the increase in body weight due to HFD feeding is only ∼1.5-fold (from ∼23 g for NC-fed female mice to ∼35 g for HFD-fed female mice; Supplementary Fig. 1B), indicating a reduction in total β-cell mass.

Collectively, these data show that β-cell Igf2 is dispensable for the establishment of normal β-cell mass in young adult mice. However, the absence of autocrine/paracrine signaling by IGF2 leads to change in the balance between insulin secretion and insulin sensitivity and to impaired control of β-cell mass in old mice fed with NC; these effects are, however, only seen in female mice. In contrast, β-cell–specific Igf2 inactivation leads to lower insulin secretion in both male and female mice that were fed an HFD.

The rate of β-cell proliferation was the same in Ctrl and βIGF2KO mice before pregnancy, but at day 14 of pregnancy, which corresponds to the peak of β-cell proliferation (32), this was increased by approximately threefold in Ctrl mice but only ∼1.5-fold in βIGF2KO mice (Fig. 5A). This led to an ∼50% increase in β-cell mass in Ctrl mice at day 19 of gestation but no significant increase in βIGF2KO mice (Fig. 5B and C).

β-Cell expansion in pregnancy is under multihormonal control. Prolactin and placental lactogen activate proliferation through prolactin receptor/JAK/STAT signaling. To determine whether this proliferation signal was impaired in the absence of Igf2, islets from Ctrl and βIGF2KO mice were incubated for 48 h in the presence of prolactin. Figure 5D shows that the proliferation of β-cells was increased to a similar extent by prolactin in both types of islets. Proliferation of β-cells is also stimulated by E2, which reduces the expression of miR-338-3p to increase IGF1R expression (19), suggesting a mechanistic basis for an autocrine/paracrine effect of IGF2 on β-cell mass expansion in pregnancy. To test this possibility, we first exposed islets from Ctrl and βIGF2KO mice to E2 for 48 h and assessed β-cell proliferation. As shown in Fig. 5E, E2 significantly increased the proliferation of Ctrl β-cells but not of βIGF2KO β-cells. In agreement with this result, suppressing miR-338-3p expression with a specific anti-miR also increased β-cell proliferation in islets from Ctrl mice but not from βIGF2KO mice (Fig. 5F). In addition, the analysis of islet gene expression at day 14 of pregnancy showed suppressed expression of Igf2 in islets from βIGF2KO mice (Fig. 5G), and a similar reduction in miR-338-3p expression and increased Igf1r expression in islets from Ctrl and βIGF2KO mice (Fig. 5G and H). Although these observations suggest that Igf2 could be required for the effect of E2 on β-cells during pregnancy, E2 treatment or miR-338-3p silencing failed to induce Igf1R expression in isolated Ctrl or βIGF2KO mouse islets (not shown), in contrast to what has been reported previously in rat and human islets (19). Thus, during pregnancy, although Igf2 expression is required for β-cell expansion, probably also acting through Igf1R, increased expression of this receptor may not be under the sole control of E2.

Gene expression analysis also showed that induction during pregnancy of the prosurvival gene Birc5 and of the cyclinB1 and cyclinB2 genes, which are at least in part controlled by E2 (19), was lower in islets from βIGF2KO mice than in those from Ctrl mice (Fig. 5J–L). However, Tph1, whose expression is controlled by prolactin receptor–dependent JAK/STAT signaling, was similarly induced in islets from both types of mice (Fig. 5M), suggesting identical induction of serotonin production.

Impaired functional adaptation of β-cells during pregnancy because of the knockout of Prlr, Men1, Tph1, or the serotonin receptor Ht3r (17,18,33,34) leads to glucose intolerance. Here, we measured glycemia, insulinemia, and glucose tolerance in Ctrl and βIGF2KO mice before and at day 14 of pregnancy. No difference could be observed in any of the measured parameters (Supplementary Fig. 4A–C).

Finally, we assessed β-cell mass expansion in response to insulin resistance induced by the IR antagonist S961 administered by osmotic minipumps. As shown in Fig. 6A, hyperglycemia developed rapidly in both types of mice, leading to similar hyperinsulinemia (Fig. 6B). The proliferation of β-cells assessed by Ki67 staining was increased to a lower extent in βIGF2KO mice compared with Ctrl mice (Fig. 6C and D); this led to a 30% lower increase in β-cell mass in βIGF2KO mice than in Ctrl mice (∼2.5 vs. 1.8 mg in Ctrl vs. βIGF2KO mice, respectively). Thus, β-cell–specific expression of Igf2 is required for the full proliferative response of β-cells to insulin resistance induced by IR blockade.

Here, we show that the autocrine action of IGF2 contributes significantly to the adaptation of the β-cell secretion capacity and β-cell mass in aging and pregnancy and in response to pharmacological induction of insulin resistance. This action is sex-dependent with reduced insulin secretion capacity and impaired β-cell mass adaptation in response to metabolic stress seen mostly in old female βIGF2KO mice. The expression of igf2 in β-cells is, however, not required to attain normal adult β-cell mass.

Islets isolated from young adult βIGF2KO mice show reduced GSIS and impaired exendin-4–induced proliferation and protection against cytokine-induced apoptosis. This is in agreement with our previous studies showing that the trophic effects of GLP-1 were largely dependent on enhanced IGF1R expression and the autocrine action of IGF2 to increase Akt signaling (25,26). We also demonstrated that immunoneutralization of IGF2 reduced in vitro GSIS, indicating that this peptide has an acute modulatory role in insulin secretion. Here, however, βIGF2KO mice showed normal glucose tolerance and in vivo GSIS, suggesting that in living mice additional glucose-activated signals, such as increased parasympathetic tone, may contribute to the control of insulin secretion.

In old βIGF2KO female mice, however, GSIS was lower than in Ctrl mice, but this was not associated with the development of glucose intolerance because of a compensatory increase in whole-body insulin sensitivity. It is well known that glycemic control depends on the balance between insulin secretion and insulin sensitivity and that normoglycemia can be preserved over a wide range of insulin secretion capacity provided that insulin sensitivity is appropriately modulated or, vice versa, that insulin secretion can increase to compensate for the development of insulin resistance (35). Although it is more common to consider that β-cells adapt to the degree of insulin sensitivity of peripheral tissues, our data show that a decrease in insulin secretion capacity of βIGF2KO mouse islets leads to a compensatory increase in insulin sensitivity. This may be caused by a lower insulin-induced IR downregulation, leading to increased cell surface receptor expression and insulin sensitivity (36,37).

When fed with an HFD, both male and female βIGF2KO mice displayed lower basal and glucose-stimulated plasma insulin levels, but their glucose tolerance was not (males) or only slightly (females) different from that of Ctrl mice. This suggested increased insulin sensitivity compared with Ctrl mice. Insulin tolerance test results indeed showed slightly higher insulin sensitivity in males and only a tendency for higher sensitivity in female βIGF2KO mice. This again indicates that lower insulinemia is linked to higher insulin sensitivity. The slight, but significant, increase in glucose intolerance seen in HFD-fed female βIGF2KO mice was accompanied by and could be caused by the observed relative decrease in β-cell mass. Thus, in conditions of metabolic stress induced by HFD feeding, the absence of igf2 production by β-cells prevents the development of the same level of insulinemia as in Ctrl mice; in female mice, this leads to a relatively lower β-cell mass, which may result from a lower rate of β-cell proliferation, an increased rate of apoptosis, or a combination of both.

The molecular basis for the sex difference in the phenotype of old βIGF2KO mice is not understood. However, in many mouse models of diabetes, such as the streptozotocin diabetic mice, mice with β-cell overexpression of human islet amyloid polypeptide, or the Zucker diabetic rats, females are mostly protected from diabetes. However, ovariectomy sensitizes them to diabetes development, an effect that can be reversed by E2 treatment (38,39). Also, E2 treatment of postmenopausal women with type 2 diabetes improves hyperglycemia through increased GSIS (40). Thus, E2 has positive effects on β-cell proliferation, protection against apoptosis, and increases in GSIS (38) that are similar to those reported for GLP-1 and GIP (41). E2 effects in β-cells are mediated by ERα, ERβ, and G-protein–coupled receptor 30. The data presented here show that E2-induced β-cell proliferation is absent in βIGF2KO mice, suggesting that an interplay between estrogens and the IGF2/IGF1R autocrine loop contributes to the control of β-cell mass and function in female mice during aging.

During pregnancy, insulin resistance develops to favor the transfer of glucose to the fetus. In the mother, this leads to β-cell expansion and increased insulin secretion, a response driven mostly by prolactin, placental lactogen, and estrogens (17,34,41,42). Our data show that, in vitro, β-cells from βIGF2KO mice were as sensitive to the proliferation effect of prolactin as Ctrl β-cells. In contrast, they failed to proliferate in response to E2 treatment or following knockdown of miR-338-3p. During pregnancy, miR-338-3p expression was equally suppressed and Igf1r expression was similarly induced in islets from βIGF2KO and Ctrl mice. As our in vitro experiments failed to show the induction of Igf1r expression in isolated islets from Ctrl or βIGF2KO mice, it is unclear whether the regulation of miR-338-3p and Igf1r expression during pregnancy is controlled by E2, by another hormone, or by the combined action of several hormones. Nevertheless, our data suggest that the autocrine/paracrine action of IGF2 is required for full β-cell expansion during pregnancy. Interestingly, the induction of Tph1 (which encodes Tph1, the rate-limiting enzyme in serotonin production) was normal in islets from pregnant βIGF2KO mice, indicating normal prolactin receptor activation. This was, however, not sufficient to fully induce cyclinB1, B2, and cell cycle progression. It is interesting to note that, despite the defect in β-cell mass expansion during pregnancy, this was not associated with impaired glucose tolerance, as described for mice with genetic inactivation of the Prlr or of Tph1 (17,33). This could be due to the increase in insulin sensitivity discussed in the context of aging mice. Another nonexclusive possibility is that insulin secretion in pregnancy is increased in large part by activation by serotonin of the ionotropic HT3R (18), a pathway controlled by the prolactin receptor; this mechanism may maintain sufficient insulin secretion activity for normal glycemic control despite the relatively low β-cell mass of βIGF2KO mice.

Finally, β-cell–specific expression of IGF2 was found to be important for the proliferation of these cells in response to insulin resistance induced by infusion of the IR antagonist S961. As previously reported (43,44), this treatment induces rapid and massive hyperglycemia and leads to an approximate doubling of β-cell mass within 7 days. Our data show that IGF2 contributes to ∼30% of the β-cell mass expansion. This also indicates that several signals must contribute to the observed increase in β-cell mass (45).

Collectively, our data show that IGF2 production by β-cells is important for the adaptation of β-cell mass and function in aging and pregnancy and in response to pharmacological induction of insulin resistance. We previously showed that IGF2 activates the IGF1R/Akt signaling pathway and that this autocrine loop mediates the trophic effects of GLP-1 on β-cells. In vitro, E2 action on β-cell proliferation is abolished in the absence of IGF2, indicating that it also acts, at least in part, through activation of the IGF2/IGF1R autocrine loop. How estrogen and IGF2 signaling interact to explain the sex-specific phenotype of βIGF2KO mice is still to be discovered, as are the factors that increase this autocrine loop following acute induction of insulin resistance. However, in a recent study (35) we showed that biosynthesis and secretion of IGF2 by β-cells are rapidly increased by glutamine to induce Akt phosphorylation and protect β-cells against cytokine-induced apoptosis. It is interesting to note that plasma glutamine levels are negatively correlated with insulin resistance and risks of diabetes (46–48). Thus, both arms of this autocrine loop, IGF2 biosynthesis and secretion as well as IGF1R expression, are controlled by and integrated with multiple hormone and nutrient-related cues to control adult β-cells mass and function.

Acknowledgments. The authors thank Dr. Raphaël Scharfmann for comments on the article. The authors also thank Dr. Lauge Schäffer (Novo Nordisk) for the gift of S961. In addition, the authors thank the personnel at Turku Center for Disease Modeling for skillful technical assistance.

Funding. This study was supported by grants from the Swiss National Science Foundation (310030-146138 to R.R. and 3100A0B-128657 to B.T.); a European Research Counciladvanced grant (INSIGHT to B.T.); Innovative Medicines Initiative Joint Undertaking grant no. 155005 (IMIDIA), resources of which are composed of financial contributions from the European Union Seventh Framework Programme (FP7/2007-2013) and European Federation of Pharmaceutical Industries and Associations companies in-kind contributions; and the European Union Seventh Framework Programme Collaborative Project BetaBat.

Duality of Interest. O.D.M. is an employee of Novo Nordisk. T.G. is an employee of AstraZeneca. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. H.M. designed and performed research, analyzed the data, and drafted the article. C.J., D.T., and T.G. performed the research and analyzed the data. S.M. performed research. O.D.M., F.-P.Z., P.R., M.P., S.N., and R.R. contributed new reagents and analytic tools. B.T. designed the research and drafted the article. B.T. 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.

Prior Presentation. Parts of this study were presented in abstract form at the 74th Scientific Sessions of the American Diabetes Association, San Francisco, CA, 13–17 June 2014.