GDF11 Attenuates Development of Type 2 Diabetes via Improvement of Islet β-Cell Function and Survival

Type 2 diabetes (T2D) is characterized by insulin resistance and insulinopenia caused by β-cell failure and decreases in β-cell mass (1). As T2D develops, glycemic control gradually deteriorates over time, which is believed to be linked to the progressive loss of β-cell function and mass (2–4). Therefore, approaches to preserve pancreatic β-cell function and/or to arrest β-cell apoptosis would clearly be valuable therapeutic modalities for diabetes.

Growth differentiation factor 11 (GDF11) plays a pleiotropic role throughout mammalian development (5–8). Recently, GDF11 has attracted considerable attention for its contradictory relationship with aging. Some studies have shown that systemic restoration of GDF11 in old mice reverses age-related phenotypes in multiple organs, including the heart, skeletal muscle, and the cerebral vasculature (9–11). Strikingly, recently published data argued that GDF11 repletion inhibited muscle regeneration and failed to rejuvenate cardiac pathologies in mice (12,13). Very importantly, Poggioli et al. (14) confirmed that administration of recombinant GDF11 (rGDF11) reverses cardiac hypertrophy in both young and old mice. Furthermore, GDF11 promotes the recovery of renal and cardiac function in old mice (15,16).

Information regarding the role of GDF11 in pancreatic islets is scarce, however. Early studies showed that mice lacking GDF11 exhibited markedly reduced β-cell numbers and arrested β-cell development (6). In addition, we previously demonstrated that replenishment of GDF11 in mice fed a high-fat diet (HFD) improves glucose tolerance (17). We therefore hypothesized that GDF11 may play an important role in pancreatic β-cells under metabolic stress. In this study, we investigated the effects of GDF11 on adult islet biology using in vivo and in vitro experiments under diabetic conditions and investigated the possible mechanisms involved.

Mouse procedures were conducted in compliance with National Institutes of Health policies on the use of laboratory animals and were approved by the Animal Ethics Committee of the Wuhan General Hospital of Guangzhou Command. All animals were housed at 21 ± 2°C with a 12-h light-dark cycle and free access to water and food.

To determine the optimal dosage for the rGDF11 administration, we first performed a dose-response study, as shown in Supplementary Fig. 1A. Thirty male C57BL/6 mice aged 6 weeks were fed an HFD (60% of calories from fat; Research Diets, #D 12492) and randomized into five groups that received daily intraperitoneal (i.p.) injections of rGDF11 (0.1 mg/kg, 0.2 mg/kg, 0.3 mg/kg, or 0.5 mg/kg) (PeproTech, Rocky Hill, NJ) or an equivalent volume of citrate buffer for 14 days before an i.p. injection of 100 mg/kg streptozotocin (STZ; Sigma-Aldrich, St. Louis, MO). These mice were then maintained on their initial protocols for an additional 14 days of observation (n = 6 per group).

Seventy-six male C57BL/6 mice aged 6 weeks (for the nongenetic mouse model) were fed an HFD (40 mice for the diabetic group) or normal chow (36 mice for the nondiabetic group). After 4 weeks, HFD mice were injected i.p. with a single dose of STZ (HFD/STZ; 100 mg/kg). Mice fed normal chow received citrate buffer alone and were processed in parallel with the diabetic mice. All mice were maintained on their respective diets until the end of the study.

Blood samples were obtained by tail-tip bleeding, and blood glucose levels were measured using a OneTouch Ultra glucometer (LifeScan, Milpitas, CA) 14 days after the STZ or citrate buffer injection. The animals were considered diabetic when nonfasting blood glucose levels exceeded 13.9 mmol/L for at least 2 consecutive days (18). A total of 37 mice were considered diabetic from the HFD group (36 mice were used in the experiments). Diabetic and nondiabetic mice were further randomized to receive vehicle or rGDF11 (diabetic + vehicle, diabetic + rGDF11, nondiabetic + vehicle, or nondiabetic + rGDF11; n = 10 per group) for 6 weeks, or isotype control antibody (isotype), or GDF11 antibody (Ab) (diabetic + isotype; diabetic + GDF11 Ab; nondiabetic + isotype; or nondiabetic + GDF11 Ab; n = 8 per group) for 6 weeks, as shown in Fig. 1A and Fig. 4A.

To evaluate the effects of GDF11 in the genetic diabetic mouse model, 40 male 6-week-old db/db mice and 10 age-matched db/m mice (The Jackson Laboratory, Bar Harbor, ME) were used in the experiments. As shown in Fig. 2A, after acclimation for 2 weeks, db/m mice did not receive treatment and were used as nondiabetic controls. The db/db mice were randomized into four groups: the vehicle, rGDF11, adenoassociated virus (AAV)-lacZ, and AAV-GDF11 treatment groups (n = 10 per group).

For rGDF11 treatment, mice were i.p. injected daily with 0.3 mg/kg rGDF11 or an equivalent volume of vehicle (citrate buffer) for 6 weeks. For GDF11 Ab treatment, mice were injected intravenously (i.v.) with 100 μg GDF11 Ab (clone #743833; R&D Systems, Minneapolis, MN) or 100 μg mouse IgG1 isotype Ab twice a week for 6 weeks. For AAV treatment, mice were injected i.v. once with AAV-GDF11 or AAV-lacZ at a dose of 1 × 10¹² viral genomes per mouse. The construction and efficiency of AAV-GDF11 have been verified as we previously reported (17).

Body weight, food intake, and 6-h fasting glucose levels were monitored weekly. At the end of the experiment, the mice were anesthetized by i.p. administration of pentobarbital sodium (60 mg/kg) and euthanized for blood tests, hormone measurements, and histological assays.

Glucose tolerance tests (GTTs) and insulin tolerance tests (ITTs) were done at the beginning and end of the studies based on a previous study (19). Blood samples were obtained at the indicated times for the measurement of plasma insulin levels during the GTT assays. The area under the curve (AUC) was calculated by applying the trapezoidal rule. Measurements of GDF11/8, HbA_1c, insulin, glucagon, total cholesterol, triglycerides, and free fatty acids were performed as previously described (17).

Pancreatic tissues were weighed, divided longitudinally, and collected for histological examination or measurement of hormone content (20). For histological examination, pancreatic tissues were fixed and processed as previously reported (21). The antibodies used for immunofluorescence staining were as follows: guinea pig anti-insulin (1:100; Abcam), goat anti-guinea pig IgG/fluorescein isothiocyanate (FITC; 1:100; Abcam), rabbit anti-glucagon (1:400, Cell Signaling Technology), and Cy3-conjugated goat anti-rabbit antibody (1:100; Boster Bioengineering). The β-cell mass and α-cell mass were calculated as the total insulin-positive/glucagon-positive area divided by the total area and multiplied by the weight of the pancreas. To calculate average β-cell size, the β-cell area was divided by the number of β-cell nuclei in the covered area. For β-cell proliferation, pancreatic sections were immunostained for insulin and Ki67 (1:200; Abcam). To assess apoptotic β-cells, TUNEL was performed using the In Situ Cell Death Detection Kit (Roche, Mannheim, Germany). Nuclei were counterstained with DAPI. The frequency of β-cell proliferation and apoptosis was calculated by dividing the number of Ki67⁺ or TUNEL⁺ β-cells by the total β-cell number.

Islets were isolated by collagenase V (Sigma-Aldrich) and Ficoll gradient separation (22). Islets were then handpicked under a stereomicroscope. For RNA or protein extraction, freshly isolated islets were immediately frozen until analyzed. Islets for in vitro secretion assays were cultured in RPMI containing 10% FBS for secretion studies.

MIN6 cells were cultured as previously described (23). For p-SMAD assays, MIN6 cells were fixed and processed as previously described (17).

Secretion assays in MIN6 cells (5 × 10⁵ cells per well) or size-matched islets (15 islets per tube) were performed as previously described (24). Results were presented as secreted hormone concentrations that were normalized to the total hormone content as analyzed by cold acid/ethanol extraction and ELISA (Millipore).

MIN6 cell proliferation was measured by the Click-iT EdU Kit (Invitrogen, Carlsbad, CA), and apoptosis was determined using flow cytometry (BD, Franklin Lakes, NJ), after double staining with Annexin V-FITC (BD Pharmingen) and propidium iodide.

Total RNA samples were prepared and measured as previously described (25). Primer sequences are listed in Supplementary Table 1.

Western blot was performed as previously reported (26). The following antibodies were used: Smad2, phosphorylated (p)-Smad2, Smad3, p-Smad3, AKT, p-AKT (Ser473), ribosomal protein S6 kinase (S6K), p-S6K (Thr389), FoxO1, p-FoxO1 (Ser256), Bax, Bcl-2, cleaved caspase-3 (all from Cell Signaling Technology); anti-GDF11 (R&D Systems), anti-myostatin (GDF8; Abcam), and β-actin (Boster Bioengineering).

All values are expressed as mean ± SEM. Statistical significance between groups was analyzed by unpaired Student t test or one-way ANOVA with a least significant difference test for multiple comparisons. Data were considered significant if P was <0.05. All statistical analyses were performed using SPSS 19.0 software (IBM Corp, Armonk, NY).

To evaluate the potential effect of GDF11 on β-cell function in diabetic mice, we first treated the nongenetic diabetic mice with rGDF11. First, the dose-response study showed that compared with vehicle-treated mice, rGDF11 treatment decreased blood glucose levels and increased serum insulin levels and β-cell mass in a dose-dependent manner. Notably, no significant differences were observed between mice injected with 0.3 or 0.5 mg/kg rGDF11 (Supplementary Fig. 1A–E), and thus, we chose 0.3 mg/kg as the optimum dose in the following animal study (Fig. 1A).

Second, the results from the nongenetic mouse model study indicated that:

1. Circulating GDF11/8 levels were markedly decreased in the HFD/STZ mice compared with nondiabetic mice at the end of the study. However, rGDF11 treatment elevated GDF11/8 levels in both diabetic and nondiabetic mice (Supplementary Table 2).

2. The diabetic groups did not significantly differ in body weight or food intake (Supplementary Table 2).

3. Although vehicle-treated HFD/STZ mice developed severe hyperglycemia and higher HbA_1c levels, rGDF11 treatment attenuated the progression of hyperglycemia and reduced HbA_1c levels (Fig. 1B and C).

4. The rGDF11-treated HFD/STZ mice had higher fasting serum insulin levels than vehicle-treated HFD/STZ mice (Fig. 1D).

5. Administration of rGDF11 to HFD/STZ mice mitigated serum total cholesterol, free fatty acids, and triglyceride levels (Supplementary Table 2), indicating that chronic treatment of rGDF11 alleviates hyperlipidemia beyond improved glycemic control.

6. Before these interventions, there was no significant difference in glucose and insulin tolerance between the diabetic groups (Supplementary Fig. 2A–D).

After the 6-week intervention, rGDF11-treated diabetic mice displayed a remarkable improvement in glucose and insulin tolerance (Fig. 1E and I). These were further corroborated by AUC analyses (Fig. 1F and J).

Third, we assessed the insulin secretory response after i.p. glucose administration. Indeed, fasting and glucose-induced insulin levels of rGDF11-treated HFD/STZ mice were both augmented, indicating that GDF11 enhanced β-cell secretory function in vivo (Fig. 1G). To further elucidate how GDF11 repletion might promote insulin secretion, we analyzed isolated islets for the expression of genes encoding insulin transcription. As a result, the expression of PDX-1, NKX6.1, MafA, and insulin2 was reduced in diabetic islets, but the mRNA levels were significantly upregulated by rGDF11 treatment (Fig. 1H). Notably, injection of rGDF11 in a group of nondiabetic mice did not affect any of the parameters examined in the study, and these animals were indistinguishable from vehicle-treated nondiabetic mice (Fig. 1A–J).

Next, we explored whether augmenting GDF11 levels could improve β-cell function in db/db mice, a genetic mouse model that recapitulates many of the features of T2D in humans (Fig. 2A). GDF11/8 levels were dramatically reduced in diabetic db/db mice, and injection of AAV-GDF11 and rGDF11 augmented the circulating GDF11/8 levels by the end of the study (Supplementary Table 3). As in the HFD/STZ models, we observed attenuated hyperglycemia and HbA_1c levels as well as improved plasma insulin levels and lipid profile in db/db mice after a 6-week treatment with rGDF11 or AAV-GDF11, an effect that seemed to be independent of changes in body weight and food intake (Fig. 2B–D and Supplementary Table 3). Although glucose tolerance and insulin sensitivity were comparable among db/db mice at the beginning of the study (Supplementary Fig. 3A–D), the rGDF11 and AAV-GDF11 interventions both attenuated glucose intolerance and insulin resistance at the end of study (Fig. 2E and I), which the AUC data confirmed (Fig. 2F and J). In addition, improved glucose-stimulated insulin secretion (GSIS) and the elevated expression of β-cell–specific genes were observed in rGDF11-treated and AAV-GDF11–treated db/db mice (Fig. 2G and H).

We next examined whether GDF11 replenishment could prevent β-cell loss in the HFD/STZ mice. Histological examination showed that islets from rGDF11-treated diabetic mice exhibited normal distribution, with a large insulin cell core surrounded by a mantle of α-cells, which differed from the disorganized islet architecture observed in vehicle-treated diabetic mice (Fig. 3A). Measurement of β-cell mass of vehicle-treated HFD/STZ pancreatic sections revealed a 61% decrease compared with vehicle-treated nondiabetic mice (Fig. 3B). However, rGDF11 treatment elevated β-cell mass and pancreatic insulin content in HFD/STZ mice (Fig. 3B and C). We did not observe differences in islet β-cell size, suggesting that the increase in β-cell mass observed in the rGDF11 group was unlikely to be caused by enhanced β-cell size (Supplementary Fig. 4B). Regarding proliferation, animals injected with rGDF11 tended to have a higher percentage of Ki67⁺ β-cells than their diabetic littermates; however, this difference was not statistically significant (Fig. 3D). Quantification of TUNEL⁺ β-cells, an index for cell apoptosis, indicated that rGDF11 treatment decreased β-cell apoptosis (Fig. 3E).

In line with the HFD/STZ model, both rGDF11 and AAV-GDF11 treatment improved islet architecture as well as augmented β-cell mass and pancreatic insulin content in db/db mice (Fig. 3F and G and Supplementary Fig. 4A). Islet histology showed no significant differences in the frequency of β-cell proliferation and β-cell size among diabetic mice (Fig. 3H and Supplementary Fig. 4C). In contrast, the number of TUNEL⁺ β-cells in the rGDF11 and AAV-GDF11 groups was reduced by 33% and 51%, respectively, compared with the vehicle or AAV-lacZ groups (Fig. 3I). Overall, these data suggest that GDF11 repletion may primarily protect β-cell mass via an antiapoptotic mechanism rather than by an induction of β-cell proliferation.

To investigate whether GDF11 has an effect on α-cells, we assessed glucagon secretion and production as well as α-cell mass. Critically, both rGDF11 and AAV-GDF11 treatment decreased circulating glucagon levels in the two different mouse models of T2D (Supplementary Fig. 5A and D). Furthermore, compared with vehicle-treated islets, rGDF11 treatment reduced glucagon secretion at 1 (−22%, P < 0.05), 6 (−31%, P < 0.05), and 20 mmol/L glucose (−42%, P < 0.05) in vitro (Supplementary Fig. 5G). As expected, HFD/STZ mice and db/db mice exhibited a remarkable increase in α-cell mass compared with nondiabetic + vehicle mice or db/m mice. However, α-cell mass and pancreatic glucagon content in diabetic mice were unaffected by rGDF11 or AAV-GDF11 treatment (Supplementary Fig. 5B, C, E, and F).

Next, we sought to confirm the physiological role of GDF11 in β-cell function and mass. Thus, nondiabetic or HFD/STZ mice were treated with GDF11 Ab for 6 weeks (Fig. 4A). Given that GDF11 is highly related to GDF8 and shares 90% amino acid sequence identity in their mature active forms (27), we first verified the specificity of the antibody used in this study. Western blot analysis confirmed that this antibody specifically detects GDF11 without cross-reaction with GDF8 (Supplementary Fig. 6A–C). Furthermore, compared with the vehicle-treated group, the stimulatory effects of GDF11 on GSIS and insulin content were dose-dependently reversed by the GDF11 Ab treatment in vitro (Supplementary Fig. 6D–G).

We next assessed whether neutralization of GDF11 in the blood impairs glucose homeostasis in mice. No differences were observed between the nondiabetic mice treated with GDF11 Ab or isotype Ab (Fig. 4B–H). However, GDF11 Ab treatment further elevated 6-h fasting blood glucose and HbA_1c levels and reduced serum insulin levels in diabetic mice (Fig. 4B–D), without changing their body weight or food intake (Supplementary Fig. 7A and B). Notably, glucose intolerance was comparable between the diabetic groups at baseline (Supplementary Fig. 7C and D) but was worse in the GDF11 Ab–treated diabetic mice after the 6-week intervention (Fig. 4E and F), which partially resulted from perturbed acute insulin response to glucose (Fig. 4I). Consistent with this, GDF11 Ab treatment reduced the expression of essential β-cell transcription factors (Fig. 4J). Insulin tolerance was also significantly decreased, induced by GDF11 Ab administration (Fig. 4G and H and Supplementary Fig. 7E and F). Furthermore, the GDF11 Ab–treated diabetic mice displayed significantly reduced β-cell mass, which was partly caused by the increased β-cell apoptosis (Fig. 4K–M and Supplementary Fig. 8).

To further corroborate the direct effects of GDF11 on β-cell function, we explored the effects of rGDF11 on GSIS and insulin content in MIN6 cells and in mouse islets. rGDF11 pretreatment stimulated a dose-dependent and time-dependent increase in insulin secretion and biosynthesis in MIN6 cells (Fig. 5A–D). Accordingly, we chose 100 ng/mL rGDF11 and 24 h as the optimum concentration and time conditions in the following study. As shown in Fig. 5E and F, exposure of mouse islets to high glucose (25 mmol/L) for 72 h impaired GSIS and insulin content, whereas preincubation with rGDF11 improved insulin secretion and storage. Similar to mouse islets, rGDF11 partially restored GSIS and insulin content in MIN6 cells under high glucose conditions (Fig. 5G–L), whereas both SB431542 (an inhibitor of transforming growth factor-β [TGF-β] type I receptor [TβRI]) and LY294002 (the phosphatidylinositol-4,5-bisphosphate 3-kinase [PI3K] inhibitor), but not the mechanistic target of rapamycin (mTOR), partially abolished the stimulatory effects of rGDF11 on insulin secretion and production in MIN6 cells (Fig. 5I–L). Furthermore, rGDF11 pretreatment augmented insulin secretion at 16.7 mmol/L glucose but did not affect basal (2.8 mmol/L) insulin release. In addition, mRNA levels of PDX-1, MafA, NKX6.1, and Insulin2 were upregulated by rGDF11 pretreatment in both primary islets and MIN6 cells (Fig. 5M and N).

We next investigated whether rGDF11 pretreatment could affect β-cell proliferation and apoptosis in vitro. First, exposure of MIN6 cells to rGDF11 showed a tendency to promote β-cell proliferation (Supplementary Fig. 9A and B). Next, to determine whether GDF11 has a direct antiapoptotic effect on β-cells, we preincubated MIN6 cells with rGDF11. As expected, hyperglycemia robustly increased the percentage of apoptotic MIN6 cells compared with controls. However, the effect was partially blocked by rGDF11 pretreatment (Fig. 6A and B). Consistently, rGDF11 treatment enhanced the expression of the antiapoptotic protein Bcl-2 level but suppressed the expression of the proapoptotic proteins Bax and cleaved-caspase3 (Fig. 6C and D). However, treatment of MIN6 cells with SB431542 or LY294002 partially abrogated the antiapoptotic effects of GDF11 (Fig. 6A–D).

To explore the possible pathways whereby GDF11 exerts protective effects on β-cells in vivo and in vitro, Western blot and immunofluorescence staining were performed. It is commonly believed that GDF11 transmits its signals via dual serine/threonine kinase receptors and transcription factors called Smads (28). By Western blotting, we detected decreased Smad2 phosphorylation and increased Smad3 phosphorylation in islets from vehicle-treated diabetic mice compared with vehicle-treated nondiabetic mice. Intriguingly, both rGDF11 and AAV-GDF11 treatment increased Smad2 phosphorylation but had no influence on Smad3 phosphorylation (Fig. 7A–D). In addition, immunofluorescence analyses showed that preincubation of MIN6 cells with rGDF11 restored Smad2 phosphorylation but that this activation was partially blocked by SB431542 (Fig. 7E and F). Indeed, rGDF11 pretreatment had no effect on p-Smad3 expression in vitro.

We further explored the noncanonical signaling cascade of GDF11. Western blotting of islet lysates showed that GDF11 enhanced AKT phosphorylation in HFD/STZ and db/db models (Fig. 8A–D). FoxO1 and S6K are evolutionarily conserved substrates of AKT (29). Compared with vehicle-treated diabetic mice, FoxO1 phosphorylation was increased in islet lysates from both the rGDF11- and AAV-GDF11–treated diabetic mice (Fig. 8A–D). Meanwhile, S6K phosphorylation was unchanged by rGDF11 treatment (Fig. 8A–D). To further evaluate this noncanonical signaling pathway in vitro, we treated MIN6 cells with rGDF11. High glucose reduced AKT and FoxO1 phosphorylation in MIN6 cells, but preincubation with rGDF11 partially rescued this defect. Notably, coincubation with LY294002, but not rapamycin, exerted a suppressive effect on rGDF11-mediated AKT and FoxO1 phosphorylation (Supplementary Fig. 10A and B). As expected, no significant difference in the p-S6K expression was observed among the groups, suggesting that the protective mechanism of GDF11 may not be dependent on the mTOR signaling pathway (Supplementary Fig. 10C).

The major findings of this study are the following: 1) GDF11 improves β-cell function both in vivo and in vitro; 2) GDF11 protects against β-cell apoptosis; 3) GDF11 inhibits inappropriate glucagon secretion; 4) neutralization of GDF11 impairs GSIS and reduces β-cell mass in diabetic mice; and 5) the molecular mechanisms underlying these beneficial effects of GDF11 may involve the activation of the canonical TGF-β/Smad2 and the noncanonical PI3K-AKT-FoxO1 pathways. To the best of our knowledge, our data show for the first time that GDF11 restoration may curtail the progression of diabetes.

A prominent feature of T2D is blunted insulin response caused by β-cell dysfunction (1,2,30). We showed that GDF11 promoted insulin biosynthesis and secretion. In particular, in vivo and in vitro GSIS experiments both indicated that GDF11 promoted insulin release in a glucose-dependent manner, which is of vital importance for the precise regulation of blood glucose levels. Recent studies have also shown that β-cell transcription factors are critical for maintaining β-cell function (31,32). Moreover, PDX-1, MafA, and NKX6.1 work synergistically to regulate insulin transcription and exocytosis, and their expression is strikingly decreased under diabetic conditions, which partially accounts for the β-cell dysfunction observed in individuals with diabetes (23,33). Very importantly, MafA expression is diminished in the islets of GDF11-knockout mice (6,34). Consistently, GDF11 treatment elevated the expression levels of these genes both in vivo and in vitro. In contrast, neutralization of endogenous GDF11 impaired glycemic control and GSIS as well as reduced the expression of key β-cell transcription factors. Together, these results provide evidence that GDF11 improves β-cell function, in part by maintaining the expression of key β-cell genes.

That β-cell mass is reduced in T2D mainly as a result of increased β-cell apoptosis has been established (4,35,36). Hence, remission of β-cell apoptosis would be valuable in the treatment of diabetes. In this study, GDF11 improved the islet architecture and increased β-cell numbers by protecting against β-cell apoptosis caused by metabolic stress. Moreover, GDF11 attenuated glucose-induced β-cell apoptosis in vitro, which supports the direct effects of GDF11 on β-cells. However, although GDF11 promotes proliferation across other cell types (9,15–17), here, we showed that β-cell proliferation, as measured by Ki67 immunostaining or 5-ethynyl-2'-deoxyuridine (EdU) incorporation, was unaffected by GDF11 repletion both in vitro and in vivo. The reason for these results may be that the functions of GDF11 are dependent on the cell type. However, neutralization of endogenous GDF11 promoted β-cell apoptosis and reduced β-cell mass. Overall, these data indicate that GDF11 prevents β-cell loss under diabetic conditions.

Extensive data support the notion that inappropriate glucagon secretion is associated with excess hepatic glucose output and hyperglycemia in diabetes (20,24,37). Indeed, the inhibition of glucagon signaling has proven to be beneficial in various animal models and in humans (38). In this study, we observed that GDF11 restoration improved glucose-induced suppression of glucagon secretion in mice and in primary islets via a direct effect on α-cells. Of note, given that insulin secretion from β-cells directly inhibits glucagon secretion (39,40), the decreased glucagon levels in animals may be partially attributed to GDF11-induced insulin secretion. In addition, as reported previously (20,41), we observed increased α-cell mass and pancreatic glucagon content in diabetic animals. However, those remained unchanged after GDF11 repletion. These results suggest that the glucose-lowering effect of GDF11 may be dependent on the improvement of α-cell function but not α-cell expansion.

As observed previously (17), GDF11 replenishment improved lipid metabolism and insulin sensitivity, which might be partially responsible for the improvement in glucose homeostasis. There are two possible mechanisms by which GDF11 might attenuate insulin resistance. First, hyperglycemia and hyperlipidemia are critical determinants of insulin sensitivity (42), and thus, improvements in glucose and lipid metabolism might partially explain improvements in insulin sensitivity. Second, previous studies showed that GDF11 can promote skeletal muscle regeneration (9) and reduce adipose tissue (14), which may be partially responsible for the elevation in insulin sensitivity. In addition, the alleviated hyperlipidemia may be a result of the beneficial effects of GDF11 on glucose metabolism, insulin sensitivity, and adipose tissue (14). Notably, a previous study showed that the administration of 0.5 or 1.0 mg/kg rGDF11 caused a significant decrease in the body weight of mice (14). However, we found that injection of 0.3 mg/kg rGDF11 had no effect on body weight in normal or diabetic mice. This discrepancy may be partly caused by the dose-dependent effect of GDF11.

Next, we explored the molecular signaling pathways that may underlie the protective effect of GDF11 on islets. It is well documented that GDF11 exerts its function by interacting with heterotetrameric membrane receptors composed of activin type II and activin type I receptors to activate the canonical Smad signaling pathway (43). Furthermore, Smad2 is requisite for the maintenance of GSIS and β-cell mass (44,45). GDF11 may act through Smad2 to facilitate β-cell maturation during development, because the phenotype of mice that are heterozygous for Smad2 is strikingly reminiscent of that of GDF11-knockout mice (6). Here, we showed that GDF11 treatment elevated the p-Smad2 expression both in vivo and in vitro. The TβRI inhibitor partially blocked the ability of GDF11 to rescue β-cells from hyperglycemia-induced dysfunction and apoptosis. In addition, GDF11 can activate several non-Smad signaling pathways in a context-dependent manner, including the PI3K pathway, which can cross talk with the Smad pathways (46). Moreover, AKT-mediated phosphorylation of FoxO1 promotes its nuclear exclusion and mediates the increase of PDX-1, NKX6.1, and MafA expression and β-cell survival (47). Here, our data support the notion that GDF11 exerts its biological activity through the PI3K pathway. GDF11 increased AKT and FoxO1 phosphorylation both in diabetic mice and in MIN6 cells, and the PI3K inhibitor partially abrogated the GDF11-mediated protection against hyperglycemia-induced β-cell malfunction and demise. Cumulatively, we conclude that the GDF11-mediated beneficial effects may be dependent on the activation of the Smad2 and PI3K-AKT-FoxO1 cascades.

This study has some limitations. First, GDF8 is a close structural homolog of GDF11, with 90% amino acid sequence identity shared in their mature active forms (27). Our assay for mouse serum GDF11 does not distinguish between circulating GDF11 and GDF8, and as a result, we did not accurately determine the GDF11 concentration in mice. Second, an earlier study showed that GDF11 is endogenously expressed in mouse, rat, and human islets (48). However, we did not conditionally knock out GDF11 in mouse islets, and thus, we could not determine the regulatory effects of circulating GDF11 or local islet GDF11 on β-cells.

Taken together, our data clearly suggest a critical role for GDF11 in the regulation of β-cell function and mass via the activation of Smad2 and PI3K-AKT-FoxO1 pathways. Thus, the identification of GDF11 opens up a new avenue for the treatment of T2D.

Funding. This work was supported by grants from the National Natural Science Foundation of China (NSFC 81370896, 81570730).

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

Author Contributions. H.L. and J.Z. conducted the animal experiments. Y.L., Lin X., J.D., and M.L. performed the in vitro experiments. Ling.X. and B.Z. analyzed the data and wrote the manuscript. G.X. designed the research. G.X. 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.