Erythropoietin protects against diabetes through direct effects on pancreatic β cells

EPO-R messenger RNA (mRNA) and protein are present in isolated islets, pancreas, brain, liver, and kidney of mice as shown by RT-PCR and Western blot, respectively (Fig. S1, A and B). We also confirmed the expression of EPO-R by quantitative real-time PCR in FACS-sorted pancreatic β cells (Fig. S1 C).

To assess the effects of rHuEPO on diabetes protection, we first administered rHuEPO in an MLDS model of type 1 diabetes (Like and Rossini, 1976). rHuEPO or saline was administered to 6–8-wk-old C57BL/6 mice three times weekly for 4 wk starting on the first day of STZ treatment. In contrast to the saline-treated controls that developed hyperglycemia within 1 wk after the start of STZ injections, the rHuEPO-treated group maintained euglycemia throughout the treatment period (Fig. 1 A). Fasting blood glucose levels were also significantly reduced in the rHuEPO-treated group compared with the saline-injected control group (Fig. 1 B). Theoretically, the lower blood glucose levels in the rHuEPO treatment group could be attributed to either increased peripheral insulin sensitivity or beneficial effects of rHuEPO directly on the β cells. The peripheral insulin sensitivity was not increased, as assessed by insulin tolerance tests in the rHuEPO-treated mice (Fig. 1 C). Therefore, these results suggest that rHuEPO may have a direct role on the β cells to provide protection against STZ-induced diabetes.

Indeed, we observed that rHuEPO treatment in the MLDS model of diabetes led to an increase in β-cell area (Fig. 2 A). To investigate the cellular mechanism responsible for the protective role of rHuEPO against STZ-induced diabetes, we examined for β-cell proliferation and apoptosis by Ki67 immunostaining and terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL), respectively. There was an increase in Ki67-positive β cells (Fig. 2 B and Fig. S2 A) and a reduction in TUNEL-positive cells in the islets of the rHuEPO-treated group compared with saline-treated mice (Fig. 2 C and Fig. S2 B). EPO has also been shown to provide cytoprotection through angiogenesis. To assess the vasculature, we examined for the presence of factor VIII (Fig. 2 D) and CD31 (Fig. 2 E and Fig. S2 C), both endothelial cell markers which were present in higher concentrations within the islets of mice in the rHuEPO-treated group compared with saline-injected controls. Quantitative real-time PCR analyses showed enhanced bcl-xL and vascular endothelial growth factor (vegf) mRNA levels in isolated islets of rHuEPO-treated mice compared with saline-treated controls in the STZ-treated cohort (Fig. 2 F). These results are in keeping with the inhibition of apoptosis and enhanced angiogenesis observed from histological analyses of the islets of rHuEPO-treated mice. We also observed a significant attenuation in the induction of IL-1β, TNF, IL-6, and MCP-1 (monocyte chemotactic protein-1) transcript levels in islets of mice in response to STZ in the rHuEPO-treated compared with the saline-treated group (Figs. S2, D–G). These results suggest that rHuEPO also has the role of dampening the inflammatory response at the level of the islet in the context of STZ-induced diabetes. Finally, analysis of insulin and glucagon double immunostaining on pancreatic sections showed intact islet architecture in both the rHuEPO- and saline-treated mice (Fig. 2 G).

We assessed whether rHuEPO treatment can reverse established diabetes by starting rHuEPO treatment 2 wk after the start of STZ injections. We observed that rHuEPO was able to reverse STZ-induced diabetes (Fig. S2 H). 2 wk after the start of STZ injections, we observed a significant decrease in β-cell area in the STZ-treated mice compared with control mice treated with citrate buffer, as assessed by insulin-immunostained pancreatic sections (Fig. S2 I). Mice treated with rHuEPO 2 wk after STZ, after development of diabetes, had enhanced β-cell area compared with saline-treated mice after 4 wk of treatment (Fig. S2 J), which indicates that rHuEPO treatment is able to reverse STZ-mediated β-cell destruction.

We also examined the protective effects of rHuEPO against diabetes in the db/db (leptin receptor mutant) mouse model of type 2 diabetes. 5-wk-old male db/db mice were treated with rHuEPO, at which time severe diabetes was already present (>25 mmol/l). Blood glucose levels in the rHuEPO treatment group decreased promptly and remained lower compared with those in the saline-treated control group throughout the treatment period (Fig. 3 A). Fasting blood glucose levels were also significantly reduced in the rHuEPO-treated group compared with controls (Fig. 3 B). Similar to the experiment in the MLDS model, the improved glucose homeostasis in the rHuEPO-treated db/db mice was not caused by an increase in insulin sensitivity, as indicated by the results of insulin tolerance tests (Fig. 3 C). We then examined for β-cell area in the two treatment groups in the db/db cohort. There was a statistically significant increase in β-cell area per pancreatic area in the rHuEPO-treated mice compared with saline-treated controls (Fig. 4 A). Assessment of Ki67-positive cells showed no differences between the rHuEPO- and saline-treated db/db mice (Fig. 4 B). This may be because of the fact that islet proliferation in db/db mice was already enhanced in response to severe peripheral insulin resistance (DeFronzo, 1997), such that EPO may not have provided a significant additional proliferative stimulus. However, we did observe a decrease in TUNEL-positive β cells on pancreatic sections of the rHuEPO-treated db/db mice (Fig. 4 C), which is suggestive of decreased apoptosis to account for the increased β-cell mass in the rHuEPO-treated db/db mice. In addition, rHuEPO treatment in the db/db mice enhanced blood vessel density in the islets as assessed by factor VIII and CD31 staining on pancreatic sections, which is suggestive of enhanced islet oxygenation (Fig. 4, D and E). Furthermore, there was a significant increase in bcl-xL and vegf mRNA levels in the islets of rHuEPO-treated db/db mice compared with saline-treated controls, as assessed by quantitative real-time PCR (Fig. 4 F). Interestingly, glucose-stimulated insulin secretion was increased in the rHuEPO-treated db/db mice compared with controls (Fig. 3 D), which is suggestive of a role for EPO in enhancing β-cell function, perhaps through promoting β-cell viability in db/db mice. Finally, islet morphology was normal in both rHuEPO- and saline-treated db/db mice staining (Fig. 4 G).

Having demonstrated the protective effects of EPO in both type 1 and type 2 diabetes models, we next assessed whether rHuEPO had direct effects on pancreatic islets under basal physiological conditions. Interestingly, rHuEPO had glucose-lowering effects even in nondiabetic conditions, as assessed by random glucose levels (Fig. S3 A). Importantly, the rHuEPO-treated mice did not develop hypoglycemia, which suggests that rHuEPO is able to enhance β-cell secretion while maintaining glucose responsiveness. Their fasting blood glucose levels were also lower after 4 wk of rHuEPO treatment (Fig. S3 B). To assess for changes in peripheral insulin sensitivity, we performed insulin tolerance tests (Fig. S3 C). These tests revealed that the glucose-lowering effects of rHuEPO were not caused by an increase in whole-body insulin sensitivity but likely caused by biological effects of rHuEPO on the pancreatic β cells. These observations further suggest that the improved glucose homeostasis in the rHuEPO-treated mice is primarily caused by the direct biological effects of rHuEPO on the pancreatic β cells.

To further examine whether rHuEPO plays a role in the dynamic process of β-cell homeostasis under basal conditions, we examined the pancreata of C57BL/6 mice treated with rHuEPO for 4 wk. β-cell area per pancreatic area was increased in the rHuEPO-treated mice compared with the saline-treated group (Fig. S3 D), likely because of the increase in the percentage of Ki67-positive β cells observed in rHuEPO-treated mice (Fig. S3 E). Analyses of factor VIII and CD31 immunostaining showed enhanced vasculature in the islets of the rHuEPO-treated mice compared with control mice (Fig. S3, F and G). rHuEPO treatment did not appear to have an effect on islet morphology (Fig. S3 H). In addition, we observed a significant increase in the hematocrit in the rHuEPO-treated mice compared with controls, which reflects the expected biological effect of EPO on erythropoiesis (Fig. S4).

To assess for potential molecular mechanisms responsible for the effect of EPO in pancreatic β cells, we isolated islets from rHuEPO-treated mice and measured their transcript levels of bcl-xL, a major target gene in response to STAT5 activation in providing antiapoptotic effects of EPO (Bittorf et al., 2000). bcl-xL mRNA levels were significantly increased in the islets of rHuEPO-treated mice (Fig. 5 A). Expression of c-myc, a STAT target gene which has potent proliferative effects, was increased in isolated islets in response to systemic rHuEPO treatment (Fig. 5 A). We also found an increase in mRNA transcript levels of c-kit in the isolated islets of rHuEPO-treated mice (Fig. 5 A). c-kit is a well-known stem cell marker that is present in regenerating pancreatic β cells after islet destruction (Tiemann et al., 2007). These findings suggest that rHuEPO may promote antiapoptotic and proliferative effects on β cells through these target genes. To analyze the role of EPO in angiogenesis, we measured mRNA levels of vegf, which was increased in response to rHuEPO.

EPO has been shown to use multiple intracellular signaling pathways in promoting tissue protection, including the JAK2–STAT5, PI3K–Akt, and mitogen-activated protein kinase pathways, although the JAK–STAT pathway is considered to be the principal signal downstream of EPO (Bittorf et al., 2000; for review see Brines and Cerami, 2006). To determine the intracellular signaling proteins in response to rHuEPO treatment in vivo, we treated C57BL/6 mice with rHuEPO for 1 wk and isolated islets to examine for proteins in the EPO-R signaling cascade. Phospho-JAK2, phospho-STAT5, and Bcl-xL expression was increased in the islets of the rHuEPO-treated group (Fig. 5 B), showing that EPO elicits its effects on the pancreatic β cells through the JAK2–STAT5–Bcl-xL pathway. The PI3K–Akt pathway has also been implicated in EPO-mediated tissue protection (Leverrier et al., 1999), and this pathway has been reported to play a major role in β-cell growth and survival (Bernal-Mizrachi et al., 2001, 2004; Nguyen et al., 2006; Stiles et al., 2006). We found that rHuEPO treatment resulted in increased Akt phosphorylation in β cells (Fig. 5 B). In keeping with the observation of enhanced insulin signaling, the islets of rHuEPO-treated mice showed greater expression of PDX-1 (pancreatic and duodenal homeobox 1), a downstream transcriptional target of insulin signaling which is important for β-cell growth and differentiation (Fig. 5 B; Kulkarni et al., 2004). Furthermore, GLUT2, a β-cell differentiation marker which is important for glucose sensing, was also increased in the islets of rHuEPO-treated mice compared with controls (Fig. 5 B).

To assess whether EPO may have direct effects on β-cell function, which may account for the improved glucose-stimulated insulin secretion in the rHuEPO-treated db/db mice, we performed islet perifusion experiments and examined for changes in membrane capacitance on individual β cells. Our results showed similar insulin secretion between the islets treated with rHuEPO or saline in response to high glucose (16.7 mM) or KCl during perifusion of individual islets (Fig. S5, A and B). By patch-clamp capacitance measurements on individual β cells, we assessed for changes in membrane capacitance (Fig. S5, C–E). The first two depolarization pulses, which determine the initial readily releasable pool of insulin granules, or first phase of insulin secretion, showed no difference in membrane capacitance of rHuEPO-treated β cells (6.89 ± 1.70 fF/pF) compared with control cells (6.47 ± 0.88 fF/pF). Similarly, the incremental and cumulative increases in cell membrane capacitance between the 3rd and 10th pulses, which measure the rate of insulin granule mobilization to the releasable pool, or the second phase of insulin secretion, also showed similar increments between the rHuEPO-treated and saline-treated β cells (20.13 ± 4.25 fF/pF in rHuEPO-treated β cells vs. 22.35 ± 1.79 fF/pF in saline-treated cells; Fig. S5 E). These results suggest that rHuEPO does not have direct enhancing effects on the insulin secretion machinery but rather provides diabetes protection through the enhancement of islet mass and viability.

To further assess the direct effects of EPO on pancreatic β cells, we next took a genetic approach by assessing β-cell–specific EPO-R knockout mice (Fig. S6, A–C). We first assessed whether endogenous EPO action on β cells would have a direct impact in basal glucose homeostasis by measuring random and fasting blood glucose levels in the mice at 1 and 2 mo of age. Our results show that there were no differences in both measurements between the RIPcre⁺EpoR^+/+ and RIPcre⁺EpoR^fl/fl mice (Fig. S6, D and E). Furthermore, glucose and insulin tolerance tests revealed no significant differences between the genotypes (Fig. S6, F and G), suggesting that β-cell EPO-R does not play an essential role in the maintenance of glucose homeostasis in vivo. Similarly, there were no differences in β-cell mass and islet morphology between the RIPcre⁺EpoR^+/+ and RIPcre⁺EpoR^fl/fl mice at 2 mo of age (Fig. S6, H and I). These findings demonstrate that endogenous EPO signaling in β cells is not essential in regulating β-cell mass under homeostatic conditions.

We next assessed whether direct effects of EPO in pancreatic β cells were essential for the diabetes protection provided by exogenous EPO. RIPcre⁺EpoR^fl/fl mice and their control littermates were subjected to MLDS-induced diabetes while on either rHuEPO or saline treatment. In contrast to the RIPcre⁺EpoR^+/+ mice, which were protected against STZ-induced diabetes while on rHuEPO treatment (Figs. 1 A and 6 A), diabetes protection was not observed in rHuEPO-treated RIPcre⁺EpoR^fl/fl mice (Fig. 6 A). Similarly, pancreatic morphometry analyses showed enhanced β-cell area per pancreatic area in the rHuEPO-treated RIPcre⁺EpoR^+/+ mice, but this effect of rHuEPO on pancreatic β cells was abolished in the RIPcre⁺EpoR^fl/fl mice (Fig. 6, B and C). Furthermore, the increase in Ki67-positive β cells (Fig. 6 D) and increase in islet vasculature observed in the rHuEPO-treated RIPcre⁺EpoR^+/+ mice were also abolished in islets of RIPcre⁺EpoR^fl/fl mice (Fig. 6, E and F).

rHuEPO treatment in mice was shown to enhance the expression of bcl-xL, c-myc, c-kit, and vegf in isolated islets (Fig. 5). To determine whether the enhanced expression of these genes in isolated islets was caused by direct effects of rHuEPO on the pancreatic β cells, we treated RIPcre⁺EpoR^+/+ and RIPcre⁺EpoR^fl/fl mice with rHuEPO for 1 wk and analyzed the expression of these genes in the isolated islets. Our results by RT-PCR showed that the islets of RIPcre⁺EpoR^fl/fl mice treated with rHuEPO did not show an induction in these genes, and the gene levels were lower compared with rHuEPO-treated RIPcre⁺EpoR^+/+ mice (Fig. 7), which is indicative of the requirement of EPO-R in β cells for the transcriptional regulation by rHuEPO.

JAK2 has been identified as a canonical essential intracellular mediator of EPO action in erythrocyte precursors; however, EPO-mediated signaling events in pancreatic β cells have remained elusive. Therefore, we assessed β-cell–specific JAK2 knockout mice (Fig. S7, A–C) to determine its essential role in EPO signaling in β cells. To assess for changes in glucose homeostasis, we measured monthly fed and fasting blood glucose, which showed no differences between the RIPcre⁺Jak2^fl/fl mice and their littermate controls (Fig. S7, D and E). Glucose and insulin tolerance tests also showed similar results between the mutant mice and their littermate controls (Fig. S7, F and G).

JAK2 is described to function downstream of various growth factors, including growth hormone, prolactin, and placental lactogen, which have all been shown to play an essential role in pancreatic β-cell homeostasis (Sorenson and Stout, 1995; Cousin et al., 1999; Fujinaka et al., 2007). Surprisingly, analyses of insulin-immunostained pancreatic sections revealed no differences in β-cell area per pancreatic area and islet architecture in the RIPcre⁺Jak2^fl/fl mice compared with those of control RIPcre⁺Jak2^+/+ littermates (Fig. S7, H and I).

RIPcre⁺Jak2^fl/fl and control mice were subjected to STZ-induced diabetes with or without rHuEPO treatment to assess for the essential direct role of JAK2 in EPO signaling in pancreatic β cells. rHuEPO provided protection in the RIPcre⁺Jak2^+/+ mice; however, diabetes protection was not observed in the RIPcre⁺Jak2^fl/fl mice, as assessed by measurements of postprandial blood glucose levels (Fig. S8 A). In keeping with the lack of diabetes protection, RIPcre⁺Jak2^fl/fl mice did not have a significant increase in β-cell area per pancreatic area in response to rHuEPO treatments, in contrast to RIPcre⁺Jak2^+/+ mice treated with rHuEPO, in which the increase was observed (Fig. S8, B and C). Similarly, the increase in Ki67-positive β cells (Fig. S8 D) and increase in islet vasculature present in the rHuEPO-treated RIPcre⁺Jak2^+/+ mice compared with saline-treated controls were absent in the islets of RIPcre⁺Jak2^fl/fl mice (Fig. S8, E and F). These data, which parallel those of EPO-R mutant mice, collectively indicate that the rHuEPO-mediated diabetes protection occurs through direct activation of EPO signaling in pancreatic β cells and that JAK2 is its essential mediator. Similarly, the islet gene modulation by rHuEPO appears to also be mediated through JAK2 in β cells. As such, bcl-xL, c-myc, c-kit, and vegf in the islets of rHuEPO-treated RIPcre⁺Jak2^fl/fl mice were lower than those of RIPcre⁺Jak2^+/+ mice (Fig. S8 G).

Central to both type 1 and type 2 diabetes is a pancreatic β-cell defect, as manifested by β-cell death and dysfunction. rHuEPO was previously shown to have cytoprotective effects in nonerythroid tissues (for review see Brines and Cerami, 2006); therefore, we investigated whether the cytoprotective effects of EPO extended to the pancreatic β cells in diabetes models. Intriguingly, we found that systemic rHuEPO administration confers diabetes protection by exerting a direct biological effect on β cells through its cognate receptor, EPO-R, and its immediate downstream kinase, JAK2, as we have demonstrated using β-cell–specific EPO-R and JAK2 knockout mice. Importantly, no other downstream kinase was able to compensate for JAK2 to mediate EPO-R signaling in β cells of the JAK2 knockout mice. Thus, our results show that pancreatic β cells may use similar EPO-mediated signaling machinery as that which is used during embryogenesis, as indicated by the similar phenotypes of embryonic lethality caused by fetal anemia observed in both JAK2- and EPO-R–null mice (Wu et al., 1995; Neubauer et al., 1998).

It appears that the protective effect of rHuEPO on the β cells under diabetes conditions occurs largely through the promotion of β-cell growth and survival rather than through direct effects on β-cell function. Indeed, the islet perifusion and patch-clamp capacitance measurements show that rHuEPO does not have a significant enhancing effect on β-cell function per se. Instead, our data show that rHuEPO treatment has antiapoptotic, proliferative, and angiogenic effects on the pancreatic islets.

The increase in bcl-xL expression induced by EPO may be the critical molecular mechanism through which EPO inhibits apoptosis in β cells during diabetes progression. A previous study has shown that increased bcl-xL expression is a key mediator of EPO action in maintaining erythrocyte mass in an organism, as bcl-xL–null mice die embryonically as the result of anemia and defects in erythropoiesis (Motoyama et al., 1995). Furthermore, EPO has been shown to up-regulate bcl-xL expression in various erythroleukemic cells lines (Silva et al., 1999).

We have also shown that rHuEPO treatment increases β-cell proliferation as well as enhances the transcription of c-myc and c-kit. EPO has been previously shown to induce the proliferation and expression of c-myc in normal human erythroid progenitors (Umemura et al., 1988). In addition, c-kit has been reported to be present on progenitor cells of the pancreas and play an important role during pancreatic development and during regeneration after STZ-induced diabetes (Tiemann et al., 2007). Furthermore, a recent study has also shown a physiological role of c-kit in the growth and maintenance of β-cell mass in mice (Krishnamurthy et al., 2007). Therefore, c-kit may be an important mediator through which effects of rHuEPO on pancreatic β cells are achieved.

Furthermore, rHuEPO administration increases vegf expression in the pancreatic islets, and EPO-mediated angiogenesis in the islets is dependent on β-cell EPO-R and JAK2. Islets are highly vascular, and previous work has shown that the dense vasculature is dependent on VEGF that is produced by the pancreatic β cells (Lammert et al., 2003). VEGF is known to play an essential role in maintaining the integrity of the microvasculature and in stimulating angiogenesis to aid in tissue healing (Ribatti et al., 2003; Galeano et al., 2004). VEGF has been shown to increase not only oxygenation and nutrient delivery to the islets but also to have a direct impact on pancreatic β-cell function by mechanisms that are not yet entirely clear (Heeschen et al., 2003). Given the potent effect of rHuEPO on the up-regulation of VEGF within islets, this important angiogenic factor may be another key mediator through which EPO promotes the survival and maintenance of function of pancreatic β cells.

Collectively, our in vivo data demonstrate potent protective effects of EPO against diabetes development by direct actions of EPO on β cells, thereby increasing β-cell mass through antiapoptotic, proliferative, and angiogenic mechanisms. In the context of the central role of β-cell insufficiency in the pathogenesis of diabetes, these biological effects of EPO reflect ideal properties for a potential new therapeutic agent for diabetes. Indeed, EPO is a safe, widely used hormone, and strategies to generate nonerythropoietic EPO-related agents are currently being actively investigated. Thus, further understanding of the biology of EPO may hold promise for the development of a potential novel strategy for diabetes prevention and treatment.

Experiments were performed on 6–8-wk-old C57BL/6 mice and 5-wk-old db/db mice in the C57BLKS/J background obtained from The Jackson Laboratory. RIPcre⁺ (Postic et al., 1999) and EpoR^fl/fl (Tsai et al., 2006) or Jak2^fl/fl (Wagner et al., 2004) mice were bred to generate RIPcre⁺EpoR^+/fl (or Jak2^+/fl) mice, which were then intercrossed to generate RIPcre⁻ and RIPcre⁺EpoR^+/+(or Jak2^+/+), RIPcre⁺EpoR^+/fl (or Jak2^+/fl), and RIPcre⁺EpoR^fl/fl (or Jak2^fl/fl) mice. These mice were examined at 8 wk of age and maintained on a 129J-C57BL/6-FVB/N mixed background, and RIPcre⁺EpoR^+/+ (or Jak2^+/+) littermates were used as controls. Mice were maintained on a 12-h light–dark cycle with free access to water and standard irradiated rodent chow (5% fat; Harlan Teklad) and housed in pathogen-free barrier facilities at the central animal facility in the Ontario Cancer Institute (Toronto, Ontario, Canada). All animal experiments were approved by the Ontario Cancer Institute Animal Care Facility.

rHuEPO (Epoetin-α; Eprex; Ortho-Biotech) was used. Mice were injected i.p. three times per week at a dose of 50 µg/kg body weight (Savino et al., 2006) between 9 a.m. and 10 a.m.

30 µl of tail vein blood was collected using EDTA-coated capillary tubes. Samples were kept at room temperature before analysis at the Toronto Centre for Phenogenomics (Toronto, Ontario, Canada).

MLDS were i.p. injected into mice (40 mg/kg body weight) for a consecutive 5 d (Liadis et al., 2005). Blood glucose was measured weekly after i.p. MLDS injections from tail vein blood with an automated glucose monitor (Precision Xtra; Abbott Laboratories).

Overnight fasts were 14–16 h in duration. Glucose tolerance tests were performed on overnight-fasted animals using a glucose dose of 1 g/kg body weight injected i.p., and measurements of blood glucose were taken at 0, 15, 30, 45, 60, and 120 min after the injection. Insulin tolerance tests were performed using human regular insulin (Novo Nordisk) at a dose of 1 U/kg body weight, and blood glucose levels were measured at 0, 15, 30, 45, 60, and 120 min after the injection of insulin. Glucose-stimulated insulin secretion in vivo was performed on overnight-fasted animals after an i.p. injection of glucose at a dose of 3 g/kg body weight, with tail vein blood collected at 0, 2, 10, and 30 min after the injection. Insulin levels were measured by an ELISA kit using a rat insulin standard (Crystal Chem Inc.).

Pancreatic tissue was fixed for 24 h in 4% paraformaldehyde in 0.1 M phosphate-buffered saline, pH 7.4. 7-µm-thick paraffin sections were obtained at 150-µm intervals on three levels and stained for insulin, glucagon, Ki67, factor VIII (Dako), and CD31 (Santa Cruz Biotechnology, Inc.). TUNEL (ID Labs) was performed by immunohistochemistry. Immunofluorescence was visualized using an inverted fluorescent microscope (Axio Observer; Carl Zeiss, Inc.). Insulin- and factor VIII–immunostained sections were scanned using a ScanScope ImageScope system and analyzed with ImageScope version 9.0.19.1516 software (Aperio Technologies). Factor VIII immunostaining was calculated per β-cell area by an automated positive pixel count algorithm. Total β-cell area per total pancreatic area was determined on insulin-stained sections.

Islets were isolated from mice as previously described (Liadis et al., 2005). In brief, 3 ml of 3 mg/ml collagenase (Sigma-Aldrich) was injected into the common bile duct for pancreatic digestion. Isolated islets were maintained in suspension culture in RPMI 1640 supplemented with 10% fetal bovine serum, 100 U/ml penicillin, and 100 µg/ml streptomycin. For β-cell sorting, freshly isolated islets were dispersed and subjected to flow cytometry based on high autofluorescence, as previously described (Allison et al., 2005; Liadis et al., 2007).

Islets were isolated from C57BL/6 mice as previously described (Liadis et al., 2005) and incubated overnight in either 100 U/ml rHuEPO or saline. The next day, batches of 50 islets, on average, per group were placed in 37°C perifusion chambers (with rHuEPO for the rHuEPO-treated islets) and perifused at a flow rate of 1 ml/min with a Krebs-Ringer bicarbonate buffer. Islets were initially equilibrated for 30 min in Krebs-Ringer bicarbonate Hepes buffer supplemented with 2.8 mmol/l glucose. They were then stimulated with 2.8 mmol/l glucose for 10 min, which was followed by stimulation with 16.7 mmol/l glucose for 40 min. After 40 min with 16.7 mmol/l glucose alone, the islets were subjected to an additional 25-min stimulation period with 30 mmol/l KCl. Fractions were collected for insulin determination using an RIA kit (Linco Research, Inc.). At the end of each perifusion, islets were collected to assess insulin content.

Isolated islets from C57BL/6 mice were dispersed into single β cells using 0.05% trypsin-EDTA (Invitrogen) for 10 min in a 37°C incubator and treated with rHuEPO or saline overnight. Cells were patch-clamped in conventional whole-cell configuration at 33–34°C. rHuEPO-treated cells were treated again with rHuEPO during the capacitance measurements. β cells were identified by cell size (>4 pF) and by their Na⁺ current inactivation properties. Capacitance measurements were performed using EPC-9 amplifier and PULSE software from HEKA. Patch pipettes had typical resistances of 3–6 MΩ when fire polished and filled with an intracellular solution containing 120 mM CsCl, 20 mM tetraethylammonium chloride, 1 mM MgCl₂, 0.05 mM EGTA, 10 mM Hepes, 0.1 mM cAMP, and 5 mM MgATP, pH 7.3, with CsOH. After whole-cell configuration was established, intracellular solution was dialyzed into the cells via patch pipette for 1 min, and then a train of 10 500-ms depolarization pulses from −80 to 0 mV was applied to determine the cell membrane capacitance change (ΔCm). Extracellular solutions contained 118 mM NaCl, 20 mM tetraethylammonium chloride, 2.6 mM CaCl₂, 5 mM Hepes, 5.6 mM KCl, 1.2 mM MgCl₂, and 3 mM glucose, pH 7.4, with NaOH. Patch-clamp data were analyzed with IGOR Pro 3.12 software (WaveMetrics).

Islets, pancreas, brain, liver, muscle, hypothalamus, and kidney tissue were isolated, and protein lysates were obtained as previously described (Wijesekara et al., 2005). Lysates were separated by Na dodecyl sulfate–10% PAGE and immunoblotted with antibodies against JAK2 (Millipore), Akt, phospho-Akt (Ser473), phospho-JAK2 (Cell Signaling Technology), STAT5 (Invitrogen), phospho-STAT5 (BD), GLUT2, PDX-1 (Millipore), Bcl-xL, and EPO-R (Santa Cruz Biotechnology, Inc.). Western blot signal densities were analyzed using ImageJ software (National Institutes of Health). Protein levels were normalized to α-tubulin levels and expressed in arbitrary units relative to littermate control levels.

RNA from FACS-sorted β cells or isolated whole islets were extracted using the RNeasy mini plus RNA extraction kit (QIAGEN), and cDNA was synthesized by reaction with MMLV RT (Invitrogen) and random primers, according to standard protocols. bcl-xL, c-kit, c-myc, vegf, EpoR, IL-1β, TNF, IL-6, MCP-1, and 18S levels were quantified (ABI Prism 7900HT real-time PCR system; Applied Biosystems) according to the manufacturer’s instructions. Real-time PCR reactions were performed in a final volume of 10 µl containing 0.3 µM gene-specific primers and SYBR green master mix (Applied Biosystems). PCR cycles consisted of an initial denaturation step at 95°C for 10 min, followed by 40 cycles at 95°C for 15 s and at 60°C for 60 s. To confirm amplification specificity, PCR products from each primer pair were subjected to a melting curve analysis. A standard curve was generated from cDNA (from 100 ng to 10⁻² ng RNA) serial dilutions for each PCR reaction and was used to determine the relative gene expression level in each sample. The expression level of each unknown was normalized by the internal control 18S, GAPDH, or actin. Each sample was run in triplicate, and each experiment included three nontemplate control wells.

mRNA was extracted from isolated islets, pancreas, brain, liver, and kidney tissues by TRIZOL (Invitrogen) according to the manufacturer’s protocol and treated with RNase-free DNase (Invitrogen). Semiquantitative RT-PCR was performed with a one-step RT-PCR kit (Invitrogen). EpoR, bcl-xL, vegf, c-myc, c-kit, and β-actin cDNA were amplified by PCR using specific primers. mRNA levels were normalized to β-actin levels and expressed in arbitrary units relative to littermate control levels using ImageJ software.

Data are presented as means ± SEMs and were analyzed by one-sample t test, independent-samples t test, or one-way analysis of variance, as appropriate, using the statistical software SPSS (version 17.0 for Windows). P-values of ≤0.05 were accepted as statistically significant.

Fig. S1 shows the expression of EPO-R in pancreatic β cells and islets. Fig. S2 shows the effects of rHuEPO on STZ-induced diabetes. Fig. S3 demonstrates the improvement of glucose homeostasis as a result of rHuEPO treatment. Fig. S4 shows increased hematocrit in mice treated with rHuEPO compared with saline-treated mice. Fig. S5 demonstrates that rHuEPO does not have an effect on β-cell function. Fig. S6 shows that EPO-R is not required for glucose homeostasis. Fig. S7 shows that JAK2 is not essential for glucose homeostasis under basal conditions. Fig. S8 demonstrates that β-cell JAK2 is required for rHuEPO-mediated protection against STZ-induced diabetes.

This work was supported by a grant MOP-201188 to M. Woo from the Canadian Institutes of Health Research (CIHR) and grants to M. Woo and R. Retnakaran from the Physicians’ Services Incorporated Foundation and the Banting and Best Diabetes Centre. This work was also supported by CIHR grant MOP-64464 to H.Y. Gaisano and National Institutes of Health grant CA117930 to K-U. Wagner. We received partial funding from Ortho-Biotech in the form of a fellowship. D. Choi is supported by the CIHR Canada Graduate Scholarship Doctoral award.

The authors have no competing financial interests.