GDF11 Improves Angiogenic Function of EPCs in Diabetic Limb Ischemia

Diabetes is a chronic metabolic disease with considerable morbidity and mortality (1–3). One devastating complication of diabetes is peripheral artery disease (PAD). Patients with diabetes with PAD always experience more severe disease, such as critical limb ischemia (CLI), than patients without diabetes. CLI, the most severe form of PAD, is a leading cause of incurable ulceration, gangrene, and even lower-extremity amputation (4). Many individuals with diabetes with PAD are not candidates for currently available surgical or endovascular procedures because of diffuse vascular disease (5). Consequently, therapeutic interventions aimed at enhancing angiogenesis and restoring blood flow in diabetic CLI are essential.

Ischemia-induced vascularization is strongly associated with endothelial progenitor cells (EPCs), which are mobilized from the bone marrow and then home to sites of vascular damage where they contribute to neovascularization and recovery (6). However, the levels and functions of EPCs are dramatically reduced in patients with diabetes (7).

Growth differentiation factor 11 (GDF11) belongs to the transforming growth factor-β (TGF-β) family and plays an essential role in mammalian development (8). GDF11 signals through the activin type II receptors and activates the canonical Smad2/3 signaling pathway (9). Some reports have revealed that GDF11 appears to decline with age and that exogenous administration of GDF11 reverses age-related defects in the heart, skeletal muscle, and cerebra (10–12). However, other groups have argued that systemic restoration of GDF11 fails to rejuvenate cardiac pathologies and inhibits skeletal muscle regeneration (13,14). GDF11 and GDF8 are highly homologous at the protein level, and many key reagents that recognize GDF11 also recognize GDF8 (14). More recently, Poggioli et al. (15) confirmed that exogenous GDF11 reduces cardiomyocyte size in mice. Importantly, GDF11 exhibits antisenescent effects in lung cells and promotes the recovery of renal and cardiac function in mice (16–18).

Of note, our early study showed that GDF11 can protect against endothelial cell injury (19). Further research has reported that GDF11 improves cardiac function and reduces infarct size in mice after ischemia/reperfusion injury (16). However, data regarding the effect of GDF11 on vascularization of EPCs are scarce. In the current study, we aimed to investigate whether GDF11 could rescue EPCs from damage induced by diabetes and improve neovascularization in diabetic limb ischemia and to reveal the possible mechanisms involved.

Male Sprague-Dawley rats weighing 200–220 g were used in this study. The model of diabetic ischemic limb was generated as described previously (20,21). As shown in Supplementary Fig. 1A, nondiabetic rats served as normal control and further randomized to receive vehicle (citrate buffer) or recombinant GDF11 (rGDF11) (the control and control + rGDF11 groups, respectively; n = 10 per group). Correspondingly, diabetic rats also were randomized to receive vehicle or rGDF11 (the vehicle and rGDF11 groups, respectively; n = 10 per group).

For rGDF11 treatment, rats were intraperitoneally injected daily with 0.1 mg/kg rGDF11 or an equivalent volume of vehicle for 2 weeks. Two weeks after treatment, the animals were euthanized for blood tests and tissue examinations. At the beginning and end of the treatments, glucose tolerance tests, insulin tolerance tests, blood pressure, and blood biochemical analyses were performed as in our previous studies (19,22).

To determine the optimal dosage for the GDF11 antibody (Ab) administration, we first performed a dose-response study. The specificity of the Ab was verified in our previous report (22). Fifteen diabetic rats were randomized into five groups that received intravenous injections of GDF11 Ab (0, 100, 150, 200, or 250 μg) twice a week or an equivalent volume of IgG Ab for 2 weeks (n = 3 per group). Accordingly, to evaluate the effects of GDF11 Ab in diabetic rats with hindlimb ischemia, rats were injected intravenously with GDF11 Ab or IgG Ab twice a week for 2 weeks (n = 10 per group).

To explore the potential mechanism of GDF11, 40 diabetic rats were further randomized to four groups: rGDF11 + vehicle (citrate buffer), rGDF11 + SB431542 (an inhibitor of TGF-β type I receptor [TβRI]), rGDF11 + Ad-small interfering RNA (siRNA) control (Ad-siCon), and rGDF11 + Ad-siRNA hypoxia-inducible factor 1α (Ad-siHIF1α) (n = 10 per group). To determine the optimal dosage for the SB431542 administration, we performed a dose-response experiment. Twelve diabetic rats were randomized into four groups that received a single intraperitoneal injection of rGDF11 + SB431542 (5, 10, or 20 mg/kg) or rGDF11 + an equivalent volume of citrate buffer (n = 3 per group). On the next day, the rats were euthanized for phosphorylated (p)-SMAD2/3 expression examination. For Ad-siHIF1α treatment in vivo, rats received a single injection of Ad-siHIF1α at a dose of 2 × 10¹⁰ plaque-forming units through the tail vein. To assess transfection efficiency, we measured the mRNA and protein levels of HIF1α by RT-PCR and Western blot.

We synthesized HIF1α-specific siRNA (siHIF1α; GenBank accession number NM_024359) and scrambled control siRNA (siCon) oligonucleotides and cloned them into pSUPER vector. Generation, amplification, purification, titer determination, and transduction of adenovirus vectors were performed as previously described (23,24).

Blood flow was scanned using a laser Doppler perfusion image (LDPI) analyzer (Perimed, Stockholm, Sweden). Perfusion analyses were performed immediately after femoral artery excision and at postoperative days 7 and 14. The LDPI index was defined as the perfusion ratio of ischemic to nonischemic hindlimb. Ambulatory impairment was scored using the criteria described previously (25).

Rats were sacrificed 14 days after treatment, and adductor muscle samples were harvested for histological evaluation. To visualize capillary and arteriole density, the tissue sections were stained with CD31 (Abcam, Cambridge, U.K.) and α-smooth muscle actin (α-SMA; Boster Bio-Engineering) according to the procedures previously published (26). Capillary or arteriole density was assessed by expressing the data as CD31⁺ vessels/mm² or α-SMA⁺ vessels/mm².

The number of circulating EPCs was calculated by flow cytometry as previously described with a little adjustment (27). In brief, total mononuclear cells were isolated from rat peripheral blood by density gradient centrifugation with Histopaque-1083 (Sigma-Aldrich, St. Louis, MO) and then incubated with fluorescein isothiocyanate–labeled kinase insert domain receptor (FITC-KDR; Abcam) and phycoerythrin-labeled CD34 antibodies (PE-CD34; Abcam). To detect the homing population of EPCs, the adductor muscle sections from the ischemic limb were incubated with CD34 (R&D Systems) and KDR (R&D Systems) for immunofluorescent staining.

Bone marrow EPCs (BM-EPCs) were isolated, cultured, and identified according to methods described previously (7). For p-SMAD assays, EPCs were fixed and processed as previously described (19).

A matrigel tube formation and migration assay was performed to determine the angiogenic capacity of BM-EPCs (28).

Apoptosis was detected using flow cytometry (BD, Franklin Lakes, NJ) after double staining with Annexin V-FITC (BD Pharmingen) and propidium iodide (22). Western blot was performed as previously reported (29). The following antibodies were used: Smad2/3, p-Smad2/3, AMPK, p-AMPKα, AKT, p-AKT, ERK1/2, p-ERK1/2, HIF1α, vascular endothelial growth factor (VEGF), stromal cell-derived factor 1α (SDF1α), Bax, Bcl-2, cleaved caspase3 (all from Cell Signaling Technology), and β-actin (Boster Bio-Engineering). For real-time PCR, total RNA samples were prepared and measured as previously described (30). The following oligonucleotides served as primers: β-actin forward: 5′-ACACTGTGCCCATCTAGGAGG-3′, reverse: 5′-AGGGGCCGGACTCGTCATACT-3′; HIF1α forward: 5′-AAGTCTAGGGATGCAGCAC-3′, reverse: 5′-CAAGATCACCAGCATCTAG-3′.

Data are expressed as mean ± SEM. Statistical differences were evaluated by Student t test or one-way ANOVA with a least significant difference test. P < 0.05 was considered significant.

Our previous report suggested that GDF11 restoration improves glycolipid homeostasis in diabetic mice (22). Thus, we further confirmed the influence of GDF11 on metabolic characteristics in diabetic rats. First, circulating GDF11/8 levels (our assay for serum GDF11 did not distinguish circulating GDF11 from GDF8) were decreased in the diabetic rats compared with the nondiabetic rats. As expected, rGDF11 treatment elevated GDF11/8 levels in both diabetic and nondiabetic rats at days 7 and 14 (P < 0.05) (Supplementary Fig. 1B and C). Second, at baseline, no significant differences were observed in glucose or insulin tolerance among the diabetic groups (Supplementary Fig. 2A–D). Although vehicle-treated diabetic rats developed severe hyperglycemia, GDF11 restoration attenuated the progression of hyperglycemia (Supplementary Fig. 2E). Consistently, the rGDF11 intervention abated HbA_1c levels (Supplementary Fig. 2F). After 2 weeks intervention, rGDF11 treatment attenuated glucose intolerance and insulin resistance in diabetic rats (Supplementary Fig. 2G–J). Third, after different treatments for 2 weeks, administration of rGDF11 improved plasma insulin levels and lipid profiles in diabetic rats but not in nondiabetic rats. Finally, no significant difference in blood pressure was found among nondiabetic and diabetic groups (Supplementary Table 1).

GDF11 has been shown to protect against ischemia/reperfusion insult in mouse heart (16). Therefore, we investigated the effects of GDF11 on perfusion recovery in nondiabetic and diabetic hindlimb ischemia. The results show that GDF11 had no effect on blood flow recovery in nondiabetic rats (Supplementary Fig. 3A–C). The perfusion recovery was worse in diabetic rats than in nondiabetic rats. At 7 and 14 days, recovery of limb perfusion was significantly increased in the rGDF11-treated diabetic rats compared with the vehicle-treated diabetic rats (Fig. 1A and B). Consequently, the rGDF11 group manifested less ambulatory impairment compared with the vehicle group (Fig. 1C). Overall, these results suggest that GDF11 repletion improves blood perfusion in the diabetic hindlimb ischemic rats.

We next questioned whether GDF11 has an effect on vascularization. Indeed, significantly increased numbers of CD31⁺ and α-SMA⁺ vessels were detected in the ischemic muscles of the rGDF11 group compared with those in the vehicle group (Fig. 2). However, GDF11 had no effects on capillary and arteriolar densities in nondiabetic limb ischemia (Supplementary Fig. 3D–F). The increased numbers of CD31⁺ and α-SMA⁺ vessels provide a histological basis for the increased blood flow and tissue viability in the diabetic ischemic limb.

Given that EPCs have been identified as an important regulator of vascularization, we investigated whether GDF11 exerts its neovascularization activities through increasing mobilization and recruitment of EPCs. The results show that GDF11 had no effect on the number of CD34⁺/KDR⁺ cells in nondiabetic rats (Supplementary Fig. 3G). The number of CD34⁺/KDR⁺ cells in diabetic rats was less than that in nondiabetic rats. However, the rGDF11 group showed increased numbers of CD34⁺/KDR⁺ cells in diabetic rats (Fig. 3A and B). To further confirm these observations, we explored the homing of EPCs to sites of ischemia. GDF11 had no effect on the number of CD34⁺/KDR⁺ cells in ischemic sites in nondiabetic rats (Supplementary Fig. 3H and I). As expected, an approximate twofold increase in the number of CD34⁺/KDR⁺ cells was observed in the rGDF11 group compared with the corresponding numbers in the vehicle group (Fig. 3C and D). These findings suggest that GDF11 is involved in EPC mobilization and recruitment in diabetic rats. In addition, the rGDF11 group had higher plasma VEGF and SDF1α concentrations than the vehicle group. Compared with the IgG Ab group, VEGF and SDF1α levels were lower in the GDF11 Ab group (Supplementary Table 1). These findings further suggest that GDF11 is essential for EPC mobilization and recruitment.

We previously demonstrated that administration of GDF11 Ab reverses the biological effects of GDF11 (22). Accordingly, the effects of GDF11 Ab on the progression of hindlimb ischemia recovery in diabetic rats were detected in this work. First, we have previously demonstrated that this Ab specifically detects GDF11 (22). Compared with the control group, the stimulatory effects of rGDF11 on migration were dose-dependently reversed by GDF11 Ab treatment in vitro (Supplementary Fig. 4A). Second, to determine the optimal dosage for GDF11 Ab administration, we performed a dose-response study. GDF11 has been shown to alleviate impaired endothelium-dependent relaxation (19). Therefore, we explored the optimal dosage of GDF11 Ab on the basis of the changes of endothelium-dependent relaxation in diabetic rats. In present study, the results show that compared with control group, GDF11 Ab treatment decreased vascular endothelium-dependent relaxation at doses between 200 and 250 μg (Supplementary Fig. 4B); thus, we chose 200 μg as the optimum dose in this animal study. In addition, vascular endothelium-independent relaxation in response to sodium nitroprusside did not differ among the five groups (Supplementary Fig. 4C).

We next assessed the effects of GDF11 Ab on metabolic characteristics in diabetic rats. As expected, glucose intolerance and insulin resistance were comparable between the diabetic groups at baseline but were worse in the GDF11 Ab group at the termination of the study (Supplementary Fig. 4D–G and J–M). As shown in Supplementary Fig. 4H and I, GDF11 Ab treatment further elevated the fasting glucose and HbA_1c levels in diabetic rats. Consistently, GDF11 Ab treatment decreased plasma insulin levels and exacerbated lipid metabolic disturbance (Supplementary Table 1).

Next, we examined whether neutralization of GDF11 reduces blood reperfusion of the ischemic limb. At day 7, the LDPI showed no significant difference between the GDF11 Ab and IgG Ab groups. However, the perfusion recovery of diabetic hindlimb ischemia in the GDF11 Ab group was lower than that in the IgG Ab group at day 14 (Fig. 4A and B). The GDF11 Ab and IgG Ab groups did not significantly differ in ambulatory impairment (Fig. 4C). Further histological analysis demonstrated that GDF11 Ab treatment results in a notable decrease in vessel density compared with IgG Ab treatment (Fig. 4D and E). Likewise, the GDF11 Ab group showed a diminished number of circulating EPCs, suggesting that GDF11 deficiency impairs mobilization of EPCs into the circulation in response to ischemia (Fig. 4F). Moreover, the homing of EPCs to the ischemic region was significantly impeded by GDF11 Ab treatment compared with IgG Ab treatment (Fig. 4G and H).

Several clinical studies have observed that the number of circulating EPCs is reduced in individuals with diabetes (31–33). Our animal data suggest that the circulating number of EPCs was decreased in diabetic rats, but GDF11 protected against EPC reduction (Fig. 3A and B). Therefore, we next investigated whether rGDF11 pretreatment could directly affect EPC proliferation and apoptosis in vitro. BM-EPCs were identified as Dil-ac-LDL and lectin double-positive cells under fluorescence microscopy (Supplementary Fig. 5). First, exposure of EPCs to GDF11 promoted EPC proliferation in a dose- and time-dependent manner (Supplementary Fig. 6A and B). Accordingly, we chose 50 ng/mL rGDF11 and 60 min as the optimum concentration and time conditions in the in vitro study. Second, flow cytometry analysis indicated that high-glucose (25 mmol/L)/hypoxia/serum-deprivation medium (HG) greatly increased the percentage of apoptotic EPCs (Supplementary Fig. 6C). However, the effect was attenuated by rGDF11 pretreatment (Fig. 5A and B). Consistently, the increased expression level of the antiapoptotic protein Bcl-2 and the decreased expression levels of the proapoptotic proteins Bax and cleaved-caspase3 further confirmed the results of the flow cytometry analysis (Fig. 5C and D).

The animal data revealed that increased circulating GDF11 levels are associated with increases in vessel density and EPC mobilization. Thus, we next conducted a series of ex vivo or in vitro experiments to determine whether the effect of GDF11 on angiogenesis and perfusion in diabetic rats with hindlimb ischemia can be attributed, at least in part, to changes in the angiogenic activity of EPCs. First, the capacity of BM-EPCs for tube formation and migration was assessed by ex vivo experiments. BM-EPCs from the rGDF11 group exhibited significant augmentation of the tube-like structure formation and migration capacity. Conversely, the angiogenic activities of EPCs from the GDF11 Ab group were impaired compared with those of EPCs from the IgG Ab group (Fig. 6). Second, angiogenic functions were further examined in EPCs from Sprague-Dawley rats in vitro. As shown in Supplementary Fig. 6D–G, EPCs cultured in normal glucose (5 mmol/L)/hypoxia/serum-deprivation medium (NG) exhibited decreased tube formation and migration capacity compared with the normal control. Compared with NG treatment, EPCs treated with mannitol/hypoxia/serum-deprivation medium, which served as the osmotic control, exhibited no significant changes in tube formation and migration. However, exposure of cells to HG significantly decreased both capacities of tube formation and migration compared with NG. As expected, EPCs pretreated with rGDF11 demonstrated increased angiogenic capacities compared with EPCs subjected to HG treatment (Fig. 7). These findings indicate that improved tube formation and migration capacities of EPCs may be attributable to GDF11.

We next explored the possible pathways on which GDF11 exerts its positive effects on the recovery of ischemia. GDF11 transmits its signals through dual serine/threonine kinase receptors and transcription factors called Smads (34). We found that rGDF11 upregulates Smad2/3 phosphorylation in diabetic rats with hindlimb ischemia. However, the GDF11 Ab group exhibited decreased Smad2/3 phosphorylation compared with the IgG Ab group (Fig. 8A and B). We further confirmed the involvement of TGF-β signaling in vivo by injecting SB431542 intraperitoneally. We first tested various doses (5, 10, and 20 mg/kg) of SB431542 and observed that 10 mg/kg was effective at reducing GDF11-induced Smad2/3 phosphorylation (Supplementary Fig. 7A and B). Therefore, we chose a dose of 10 mg/kg to use in the experiments. As expected, the perfusion recovery of diabetic ischemic hindlimb was partially blocked in rats cotreated with rGDF11 and SB431542 (Supplementary Fig. 7C and D). In addition, the effects of rGDF11, including increased circulating EPCs and improved EPC recruitment, were partially abolished in rats cotreated with rGDF11 and SB431542 (Supplementary Fig. 7E–G). In in vitro experiments, cultured EPCs in HG conditions decreased Smad2/3 phosphorylation. Of note, treating EPCs with rGDF11 activated the canonical TGF-β/Smad signaling, revealed by an increase in the Smad2/3 phosphorylation cascade, although the phosphorylation could be blocked by coincubation with SB431542 (Fig. 8C and D). Moreover, SB431542 partially abolished the protective effects of GDF11 on EPCs, including improved tube formation, migration, and antiapoptosis (Figs. 5 and 7).

We further explored the noncanonical signaling cascade of GDF11. Activation of HIF1α-related pathways is known to be associated with changes in postischemic neovascularization (35). Therefore, we investigated whether activation of HIF1α-dependent signaling is required in mediating the proangiogenic effects of GDF11. The results show that rGDF11 significantly upregulated HIF1α protein levels in diabetic ischemic rats (Fig. 8E and F). HIF1α is known to regulate angiogenesis through mediators such as VEGF and SDF1α (36). Consistent with the increase of HIF1α, rGDF11 elevated the VEGF and SDF1α levels in ischemic muscle samples (Fig. 8E and F). However, neutralization of GDF11 reduced HIF1α, VEGF, and SDF1α expression levels (Fig. 8E and F). The activation of HIF1α can be regulated by several kinases (37,38). Thus, we examined changes in the levels of protein kinase B/AKT, AMPK, and extracellular regulated protein kinases 1/2 (ERK1/2). rGDF11 treatments increased AKT phosphorylation compared with that in the vehicle group, but they did not influence the expressions of AMPK and extracellular regulated protein kinases 1/2 (Supplementary Fig. 8A–D). We next verified this noncanonical signaling in cultured EPCs. HG conditions reduced AKT phosphorylation and HIF1α and VEGF expression in EPCs, but preincubation with rGDF11 partially rescued these defects (Supplementary Fig. 8E–H). Furthermore, coincubation with an AKT inhibitor exerted a suppressive effect on GDF11-mediated AKT phosphorylation and HIF1α and VEGF expression (Supplementary Fig. 8E–H).

Importantly, we used Ad-siHIF1α to further confirm the involvement of HIF1α-related pathways in vivo and in vitro. First, in Ad-siHIF1α–treated cells, we observed significant decreases in HIF1α mRNA and protein levels compared with those in Ad-siCon cells in the hypoxia condition (Supplementary Fig. 9A–C). Second, the results demonstrate that Ad-siHIF1α partially abolished the protective effects of GDF11 on EPCs, including improved tube formation and migration, and antiapoptosis (Figs. 5 and 7). Next, we evaluated the effect of HIF1α silencing on GDF11-induced neovascularization in vivo. We first evaluated in vivo silencing of endogenous HIF1α. At day 2 after interventions, Ad-siHIF1α decreased HIF1α mRNA and protein levels in both ischemic and nonischemic muscle compared with those in the Ad-siCon group (Supplementary Fig. 9D–F). As expected, blood perfusion was reduced in Ad-siHIF1α rats treated with rGDF11 compared with that in Ad-siCon rats treated with rGDF11, indicating that silencing of HIF1α partially abrogates the effect of rGDF11 (Supplementary Fig. 9G and H). In addition, the other positive effects of rGDF11, including increased circulating EPCs and improved EPC recruitment, were partially abolished in Ad-siHIF1α rats (Supplementary Fig. 9I–K).

The major findings of this study were that 1) GDF11 improves vascularization and thus accelerates the recovery of blood flow in diabetic limb ischemia; 2) GDF11 improves EPC angiogenic functions, including tube formation and migration; 3) GDF11 protects against EPC apoptosis; 4) neutralization of GDF11 impairs vascularization in diabetic hindlimb ischemia; and 5) the molecular mechanisms underlying these beneficial effects of GDF11 may involve the activation of the canonical TGF-β/Smad and the noncanonical AKT/HIF1α signaling pathways. The novel findings suggest that GDF11 restoration may rescue EPCs from damage induced by diabetes and cause greater improvement in vascularization in diabetic hindlimb ischemia.

It is well accepted that the evolution of ischemic damage in patients with diabetes is worsened owing to impairment of the reparative angiogenesis process (39,40). In this study, we have demonstrated for the first time in our knowledge that GDF11 treatment significantly increases vascular density in ischemic tissue in diabetic rats and subsequently markedly intensifies perfusion of ischemic limb. However, GDF11 had no angiogenic effect in nondiabetic rats. Conversely, neutralization of GDF11 impaired neovascularization and alleviated the recovery of blood flow. In addition, GDF11 replenishment improved glycolipid metabolism in diabetic rats as previously observed (22), which might partially be of benefit for angiogenesis. Taken together, these data indicate an unequivocal role for GDF11 in revascularization and recovery from hindlimb ischemia.

EPCs are well known to be an important contributor to neovascularization (41). Circulating EPCs are decreased in patients with diabetes at an early stage and are further impaired in patients with macrovascular complications (32,42), and EPC reduction is related to the worsening of diabetic complications. Our previous studies demonstrated that GDF11 plays a critical role in the regulation of proliferation and survival of endothelial cells (19). In the current study, we show that GDF11 supplementation leads to more EPCs homing to sites of ischemia, which consequently may participate in the formation of new vessels. Decreased release of VEGF and SDF1α resulted in a defective recruitment process (43). In our study, ischemic tissue and plasma levels of VEGF and SDF1α were increased after GDF11 supplementation, which partially explains the effect of GDF11 on EPC recruitment. With this study, we provide evidence that GDF11 treatment improves impaired EPC functions both ex vivo and in vitro, including tube formation and migration. It has been shown that patients with diabetes with ischemic foot lesions as a result of end-stage PAD have a great reduction of EPCs compared with PAD but without foot lesions (32). These data highlight the importance of our finding that GDF11 treatment significantly increases the number of EPCs in diabetic rats with CLI.

Next, we explored the underlying pathways that may explain the positive effects of GDF11 on vascularization and recovery of blood flow. It is well documented that GDF11 activates the canonical Smad signaling pathway (9). Our data show that GDF11 elevated p-Smad2/3 expression. The TβRI inhibitor partially blocked the ability of GDF11 to rescue diabetic rats from hindlimb ischemia and EPCs from dysfunction and apoptosis.

In addition, GDF11 can activate several non-Smad signaling pathways in a context-dependent manner, including the AKT pathway, which can crosstalk with Smad signaling (44). Of note, one of the AKT targets that may be particularly important in the angiogenesis process is HIF1α (38). The HIF1α-related mechanism appears to involve an increase in vessel density and limb perfusion and a rise in the number of circulating EPCs (25). The current data support the notion that GDF11 exerts its biological activity through the AKT/HIF1α pathway. GDF11 increased AKT phosphorylation and HIF1α expression both in diabetic rats and in EPCs. Moreover, GDF11-induced neovascularization was blunted in Ad-siHIF1α rats, and Ad-siHIF1α partially abrogated GDF11-mediated protection against EPC dysfunction and demise in vitro, indicating that HIF1α downregulation attenuated the proangiogenic effect of GDF11. Cumulatively, we conclude that the GDF11-mediated beneficial effects may depend on the activation of the Smad and AKT/HIF1α cascades.

Our study had some limitations. First, GDF8 is a close structural homolog of GDF11, with 90% amino acid sequence identity shared in its mature active forms. Our assay for rat serum GDF11 does not distinguish circulating GDF11 from GDF8, and as a result, we did not accurately determine the GDF11 concentration in rats. Second, we did not explore the specific mechanisms of the effect of GDF11 on metabolism. Third, many other cells, such as mesenchymal stem cells and immunomodulatory cells, are known to be involved in neovascularization (45). However, we did not further investigate the effect of GDF11 on the other cells associated with vascularization in this study; some additional studies are needed.

In conclusion, the data clearly suggest a critical role for GDF11 in vascularization through increasing EPC number and improving its functions through the activation of the Smad2/3 and AKT/HIF1α pathways. These results imply the possible translational value of this research in the clinical care of patients with diabetes in the future. Considering the result that controlled elevation of GDF11 in the circulation has therapeutic benefits in treating CLI, treatment with GDF11 may be applied in patients with diabetes with ischemic vascular diseases, including myocardial infarction, cerebral stroke, and peripheral arterial occlusive diseases, and especially in patients with diabetic foot ulcer and gangrene. With regard to the administration routes, we envision that GDF11 treatment could work in patients through several approaches, such as inhalation delivery, wound external application, or intravenous, intramuscular, or subcutaneous injection of exogenous rGDF11 protein and analogs. Long-acting formulation of GDF11 (e.g., microspheres, nanoparticles, fusion protein that is based on IgG heavy chain constant regions [IgG Fc], adeno-associated virus–mediated stable gene expression) may be developed in the future for the sake of better medical compliance. However, there is still a long way to go to get this treatment to market because there are a lot of clinical trials and regulatory affairs to be finished and because the cost-effectiveness analysis of this treatment needs to be further clarified.

Funding. This work was supported by grants from the National Natural Science Foundation of China (81370896, 81570730), National Key Research and Development Program of China (2016YFC1305601), and Research Project of Hubei Health and Planning Commission (WJ2017H0031).

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

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