Peroxisome Proliferator-activated Receptor-γ Activation Augments the β-Cell Unfolded Protein Response and Rescues Early Glycemic Deterioration and β Cell Death in Non-obese Diabetic Mice*

Type 1 diabetes is an autoimmune disorder that is characterized by a failure of the unfolded protein response in islet β cells with subsequent endoplasmic reticulum stress and cellular death. Thiazolidinediones are insulin sensitizers that activate the nuclear receptor PPAR-γ and have been shown to partially ameliorate autoimmune type 1 diabetes in humans and non-obese diabetic (NOD) mice. We hypothesized that thiazolidinediones reduce β cell stress and death independently of insulin sensitivity. To test this hypothesis, female NOD mice were administered pioglitazone during the pre-diabetic phase and assessed for insulin sensitivity and β cell function relative to controls. Pioglitazone-treated mice showed identical weight gain, body fat distribution, and insulin sensitivity compared with controls. However, treated mice showed significantly improved glucose tolerance with enhanced serum insulin levels, reduced β cell death, and increased β cell mass. The effect of pioglitazone was independent of actions on T cells, as pancreatic lymph node T cell populations were unaltered and T cell proliferation was unaffected by pioglitazone. Isolated islets of treated mice showed a more robust unfolded protein response, with increases in Bip and ATF4 and reductions in spliced Xbp1 mRNA. The effect of pioglitazone appears to be a direct action on β cells, as islets from mice treated with pioglitazone showed reductions in PPAR-γ (Ser-273) phosphorylation. Our results demonstrate that PPAR-γ activation directly improves β cell function and survival in NOD mice by enhancing the unfolded protein response and suggest that blockade of PPAR-γ (Ser-273) phosphorylation may prevent type 1 diabetes.

Type 1 diabetes (T1D) 2 is characterized by the loss of insulin production, as ␤ cells succumb to targeted autoimmunity. The non-obese diabetic (NOD) mouse spontaneously develops T1D and is used as a model that closely mimics human disease (1). In this animal model insulitis (infiltration of the islet by cells of the immune system) is observed as early as 4 weeks of age, with the subsequent development of diabetes in many animals by 12-20 weeks of age (2). The concept that the only defect in T1D lies in the immune system has been revisited in recent years. Our group and others have demonstrated that endoplasmic reticulum (ER) stress and resultant insulin secretory defects in the ␤ cell precede the onset of T1D in mice (3,4). These findings raise the intriguing possibility that ER stress in the ␤ cell might contribute to the aberrant production of "neoantigens" that subsequently invites autoimmunity and insulitis (5)(6)(7). Dysfunction of the ␤ cell in pre-T1D has been observed in other studies in NOD mice (8), and both ␤ cell ER stress and dysfunction have been observed in humans with early T1D or destined to develop T1D (9,10). Therefore, therapies that directly enhance the ER stress-remediating response known as the unfolded protein response (UPR) may prove useful as adjunctive therapies to delay/prevent T1D.
Thiazolidinediones (TZDs) are activators of the nuclear transcription factor peroxisome proliferator-activated receptor-␥ (PPAR-␥) and have been traditionally viewed as insulin sensitizers, due to their effects on adipose tissue. The anti-diabetic action of TZDs in adipose tissue occurs, in part, by blocking the phosphorylation of PPAR-␥ at Ser-273 (11). Additionally, studies in humans with type 2 diabetes (T2D) and in rodent models of T2D suggest that TZDs may also directly enhance ␤ cell function (12,13). In addition to their potential effects on ␤ cells, TZDs have also been implicated in the reduction of inflammation and autoimmunity due to the effects on dendritic cells, macrophages, and T cells of the immune system (14 -16).
The potential utility of TZDs in T1D has been limited to only a few studies. In humans, treatment of established T1D with the TZD pioglitazone resulted in modest improvements in glycemic control (17)(18)(19), raising the prospect that intervention with the drug in the pre-diabetic phase might have greater impact. In accordance with this possibility, studies in NOD mice have shown that administration of the TZDs rosiglitazone and troglitazone at the time of weaning significantly reduced the incidence of T1D (20,21). These studies in NOD mice did not address the possibility that direct effects of TZDs on the ␤ cell may have led to reduced ␤ cell stress early in the pathogenesis of T1D. To address this possibility, we studied the effects of pioglitazone in the pre-diabetic phase in NOD mice. Our results were consistent with a ␤ cell sparing effect of pioglitazone independent of effects on insulin sensitivity or body fat distribution and suggest that reductions in early ␤ cell stress in NOD mice may underlie the reduction in T1D incidence caused by TZDs.

TZD Therapy Improves Glycemic Control in Pre-diabetic
Female NOD Mice-Pre-diabetic female NOD mice show progressively worsening ␤ cell function, failing UPR, and increasing ER stress in pancreatic islets as they age from 6 weeks onward (3,4,8). To determine if TZD therapy can restore the UPR and ameliorate ␤ cell stress in the pre-diabetic phase, 6-week-old female NOD mice were fed a standard chow diet containing pioglitazone (0.01 wt%) until 10 weeks of age. Control mice were fed an identical diet not containing pioglitazone. By design, none of the animals developed overt diabetes (defined as blood glucoses on two consecutive measurements Ͼ250 mg/dl) during the course of the study. Whereas pioglitazone is known to increase body weight and fat mass in mouse models of T2D (22), during this study pioglitazone-treated NOD mice gained weight no differently than controls (Fig. 1A) and demonstrated no differences in body fat composition as determined by dual-energy x-ray absorptiometry scanning at the end of the study (Fig. 1, B, C, and D). Nevertheless, at the end of the study, pioglitazone-treated mice exhibited a significant improvement in glycemic control compared with control mice, as judged by an intraperitoneal glucose tolerance test (GTT) and corresponding area under the curve (AUC) analysis (Fig. 1, E and F). To assess if the improved glycemic control in pioglitazone-treated mice was caused by altered insulin sensitivity, an intraperitoneal insulin tolerance test was performed, revealing no differences in insulin sensitivity between the two groups ( Fig. 1, G and H). Additionally, phosphorylation of Akt in liver and muscle after acute intraperitoneal insulin administration revealed no differences between control and pioglitazone-treated mice (Fig. 1I). Taken together, these results suggest that pioglitazone improves glucose tolerance in NOD mice without significant changes in tissue insulin sensitivity.
Pioglitazone Treatment Reduces Insulitis in NOD Mice but Does Not Affect T Cell Proliferation in Vitro-To determine if the improved glycemic profile in pioglitazone-treated mice emanated from alterations to the immune response, we next assessed insulitis in histological sections of pancreas from treated and control mice. Fig. 2, A and B, shows that pioglitazone treatment reduced the incidence and severity of insulitis compared with controls. To test whether the reduction in insulitis might have been caused by an effect of pioglitazone on T cell populations, we isolated pancreatic lymph nodes from mice and assessed CD4ϩ T cell populations in these tissues. Notably, pioglitazone had no significant effect on Th1, Th17 (CD4ϩ IL17Aϩ), or Treg (CD4ϩCD25ϩFoxp3ϩ) cell populations in the pancreatic lymph node (Fig. 2, C-E).
The reductions in insulitis without effects on specific CD4ϩ T cell populations led us to consider if pioglitazone might have caused a generalized reduction in T cell proliferation. Unfractionated splenocytes from NOD mice were loaded with the dye carboxyfluorescein succinimidyl ester (CFSE) and subjected to polyclonal stimulation using a combination of anti-CD3 antibody, anti-CD28 antibody, and IL-2 in vitro to mimic antigendependent and -independent signals as seen in T1D (23). After 4 days stimulation in the presence or absence of 1 or 10 M pioglitazone, cells were gated for CD4 positivity and analyzed for CFSE dilution by flow cytometry. Fig. 3A shows representative histograms demonstrating dilution of CFSE upon stimulation with anti-CD3/anti-CD28/IL-2, findings indicative of T cell proliferation. No differences in CFSE dilution were observed with either 1 M or 10 M pioglitazone (Fig. 3B), indicating that pioglitazone at these concentrations does not alter the proliferative rate of T cells. These data suggest that the reduction in insulitis seen in treated mice is not likely an effect of pioglitazone to directly alter T cell populations or suppress T cell proliferation.
Pioglitazone Inhibits PPAR-␥ (Ser-273) Phosphorylation in ␤ Cells and Islets-Pioglitazone and related TZDs interact with PPAR-␥ and block phosphorylation of Ser-273, an effect that promotes actions of the nuclear receptor (11). To confirm a direct effect of pioglitazone in islet ␤ cells, we pretreated the mouse ␤ cell-derived cell line MIN6 with 10 M pioglitazone or 10 M roscovitine (an inhibitor of cyclin-dependent kinases, which are known to promote Ser-273 phosphorylation) (24) then performed an immunoblot for phospho-PPAR-␥ (Ser-273). As shown in Fig. 4A, levels of phospho-PPAR-␥ (S273) was detectable in non-treated (control) cells. By contrast, phospho-PPAR-␥ (S273) levels were blunted in cells pretreated with 10 M pioglitazone and 10 M roscovitine (Fig. 4A). Islets of pioglitazone-treated NOD mice also showed reductions in phospho-PPAR-␥ (S273) (Fig. 4B), consistent with a direct effect of pioglitazone on islets.
TZD Therapy Augments Insulin Secretion and Reduces ␤ Cell ER Stress and Death in Pre-diabetic Female NOD Mice-To determine if the improved glycemic control in pioglitazonetreated NOD mice was a consequence of enhanced ␤ cell function, serum insulin levels were measured after an intraperitoneal glucose load. As shown in Fig. 5A, pioglitazone-treated mice displayed significantly higher levels of serum insulin compared with the control group at 10 min after the intraperitoneal glucose load. Isolated islets from treated mice also showed a lower basal insulin secretion at low glucose (2.5 mM) in vitro and a significantly more robust response to glucose stimulation (25 mM) compared with control islets (Fig. 5B). To assess if pioglitazone treatment affected ␤ cell stress and prohormone processing, we next measured random-fed insulin and proinsulin and the proinsulin:insulin ratio. Higher proinsulin:insulin ratios have been shown to correlate to ␤ cell ER stress in both humans and NOD mice (4,25,26). As shown in Fig. 5C, whereas the random-fed insulin was a significant 2-fold greater in the treatment group compared with controls, the proinsulin levels were not significantly different (Fig. 5D), resulting in a proinsulin:insulin ratio that trended lower (p ϭ 0.07) in the treatment group (Fig. 5E).
To assess more directly the possibility that pioglitazone enhanced the UPR and reduced ER stress, we next isolated islets from treated and control mice at the end of the study and measured both mRNA and protein markers of the UPR. The UPR is characterized by variable activation of three distinct pathways, IRE1␣, PERK, and ATF6. In islet ␤ cells, the IRE1␣ and PERK pathways predominate and are evident by increases in spliced Xbp1 mRNA levels and ATF4 protein levels (27). As shown in Fig. 6A, there was significant reduction in spliced Xbp1 (sXbp1) mRNA levels in islets from pioglitazone-treated mice compared with controls, demonstrating attenuation of the IRE1␣ pathway by pioglitazone treatment. By contrast, an increase in ATF4 protein levels was observed by immunoblot analysis (Fig.  6B). The elevation of ATF4 protein levels was confirmed by an increase in the mRNA encoding DNA damage-inducible protein 34 (Gadd34), a known ATF4 target gene ( Fig. 6C) (28). GADD34 is a phosphatase that dephosphorylates eIF2␣ to restore protein synthesis in the remediation phase of the UPR (28,29). As shown in Fig. 4B, phosphorylated eIF2␣ is reduced in islets of pioglitazone-treated mice. Consistent with the recovery of protein synthesis, a slight, but significant, increase in the levels of the protein folding chaperone BIP was observed in the islets of treated mice (Fig. 6C). FIGURE 1. Effect of pioglitazone treatment on metabolic parameters in pre-diabetic NOD mice. 6-Week-old pre-diabetic NOD mice (n ϭ 10 per group) were placed on either normal chow (Control) or chow containing 0.01 wt% pioglitazone (Pio). A, body weights during feeding. B, lean body mass as assessed by dual-energy x-ray absorptiometry after 4 weeks of feeding. C, fat mass as assessed by dual-energy x-ray absorptiometry after 4 weeks of feeding. D, percent body fat as assessed by dual-energy x-ray absorptiometry after 4 weeks of feeding. E, results of intraperitoneal GTT after 4 weeks of feeding. F, AUC analysis of GTT shown in panel E. G, results of intraperitoneal insulin tolerance test after 4 weeks of feeding. H, AUC analysis of corresponding ITT in panel G. I, immunoblot analysis of phospho-AKT (Ser-473) and total AKT in liver and muscle isolated from 4-week-fed mice after acute insulin injection. Immunoblots show representative data from three mice per group (Control and Pio). * indicates that the value is significantly different from control by two-tailed t test.
Failure of the UPR to adapt to the underlying stress leads to frank ER stress and to activation of the proapoptotic pathway mediated by CHOP (30). Concordant with the adaptive UPR in pioglitazone-treated mice, there was a reduction in Chop mRNA in islets (Fig. 6E) and a corresponding reduction in CHOP protein in ␤ cells by immunohistochemistry (Fig. 7A). Likewise, cleaved caspase 3 (CC3), a marker of apoptotic cells, was reduced in insulinϩ cells of pioglitazone-treated mice compared with control mice (Fig. 7B).
To clarify the underlying mechanism promoting an adaptive UPR, we evaluated pancreas tissue sections for evidence of oxidative stress, which is known to drive the development of ER stress (27). Fig. 7C shows that control NOD mice exhibited evidence of oxidative stress in islets, as assessed by immunostaining for 4-hydroxynonenal (4-HNE). By contrast, minimal to no 4-HNE staining was observed in islets of pioglitazone-treated mice. As a likely consequence of reduced oxidative stress and more robust UPR, ␤ cell area (as a percentage of total pancreatic area) was increased 2-fold upon pioglitazone treatment (Fig. 7D). This greater ␤ cell area percent in pioglitazonetreated mice resulted primarily from reduced ␤ cell death and not an increase in ␤ cell proliferation, as ␤ cell death parameters CC3 (Fig. 7C) and unmethylated preproinsulin DNA (Fig. 7E) (31) were reduced, but no changes in the proliferation marker proliferating cell nuclear antigen (PCNA) were observed (Fig.  7F).

Discussion
T1D develops as a disorder of the immune system in which ␤ cell autoantigens, released as a result of ␤ cell death, trigger the eventual selection and proliferation of autoreactive T cell clones (32). It has been speculated that the physiologic early turnover of ␤ cells seen in neonatal mice (and possibly humans, as well) might contribute sufficient antigen exposure to trigger autoimmunity (33), but this possibility has been challenged in studies of NOD mice (34). A second hypothesis suggests that stress responses intrinsic to the ␤ result in the pathologic formation of epitopes (as misfolded or postranslationally modified proteins) that are immunogenic (5,6,35). In this regard, ␤ cell ER stress has been shown to occur in the early pre-diabetic phase in NOD mice and in humans with new-onset T1D (4,10,25). The UPR is a coordinated response that is activated under a variety of stress conditions that put increased burden on the ER (ER stress). Two phases of the UPR have been recognized: an adaptive, stress-remediating phase, and a subsequent maladaptive, pro-apoptotic phase that is initiated when remediation is not possible (for a review, see Ref. 29).  The adaptation to ER stress by enhancing the UPR has been proposed as viable, ␤ cell-protective therapy for T1D (3). Because administration of TZDs has been shown to ameliorate ER stress (22), we asked if TZDs have the potential to directly impact ␤ cell function and survival in NOD mice.
TZDs are traditionally viewed as insulin sensitizers, but evidence in recent years suggests that they have direct effects on the ␤ cell. These effects include the stimulation of Pdx1 gene activity and increases in Pdx1 protein levels in isolated ␤ cells (36,37), stimulation of Serca2 gene activity and protein levels (22,38), and reductions in ␤ cell oxidative stress (39). Moreover, TZD administration was shown to delay the incidence of T1D in NOD mice (20,21) and to improve T1D glycemic control in humans (17)(18)(19), although it is unclear if these findings were a result of ␤ cell effects of TZDs. Notably, our studies showed that pioglitazone administration resulted in a reduction in phospho-PPAR-␥ (Ser-273) in islets and an improvement in the islet UPR, as evidenced by enhanced ATF4 and BIP protein levels and a trend to reduced proinsulin:insulin ratio. Although our study points to a PPAR-␥-dependent effect of pioglitazone, we should point out that pioglitazone was also shown to have an acute effect of reducing metabolic flux and insulin secretion in ␤ cells in a non-PPAR-␥-dependent fashion (40,41). This effect may have allowed for reduced ER load early in disease pathogenesis and subsequent improvements in ER function. Nevertheless, because our findings were not accompanied by changes in body weight, body fat distribution, or insulin sensitivity, we believe they suggest a direct effect of pioglitazone on islet ␤ cells.
It is noteworthy that we observed increased ATF4 levels despite a reduction in CHOP protein and mRNA in pioglitazone-treated mice. Whereas prior studies have linked ATF4 to activation of the gene encoding CHOP (42), it has become increasingly appreciated that other pathways and transcription factors may activate CHOP independently of ATF4 (43,44) and that of GADD34 may dissociate ATF4 from Chop activation (45). We, therefore, propose that the elevation of ATF4 in our study likely represents an adaptive, ameliorative effect on ER stress. A significant outcome in our studies was the increased area percent of ␤ cells in pioglitazone-treated animals compared with controls. Importantly, ␤ cell area was also higher in these animals when compared with 6-week-old NOD mice at the start of the study, suggesting that pioglitazone treatment either enhanced the rate of ␤ cell replication or reduced ␤ cell death or some combination of the two. This result is reminiscent of studies in type 2 diabetic db/db mice in which pioglitazone treatment resulted in increased ␤ cell area that was coincident with improved glycemic control (22). In our case, pioglitazonetreated mice showed a reduction in the serum unmethylation index, a sensitive biomarker of ␤ cell death (31), as well as CC3 staining in insulinϩ cells. In light of the reduction in ␤ cell death, it is noteworthy that we observed reductions in both oxidative stress (by 4-HNE immunostaining) and IRE1␣ activity (as judged by reduced spliced Xbp1 mRNA). IRE1␣ is an endoribonuclease that operates in a bifunctional manner, initially promoting the adaptive response of the UPR by leading to Xbp1 splicing and later, during prolonged ER stress, promoting the cleavage of other mRNAs and microRNAs and leading to cellular apoptosis (46 -48). In prior studies, our group demonstrated that NOD mice show significant elevations in spliced Xbp1 mRNA (and, by inference, IRE1␣ activity) compared with non-diabetes-prone controls at 10 -12 weeks of age (4). The studies of Kaufman et al. (27) suggest that oxidative stress is closely linked to ER stress. We suggest here that the reductions in IRE1␣ activity as a result of attenuated oxidative stress by pioglitazone likely reflect an augmented and more successful UPR.
Finally, we should comment on the significant reduction in insulitis we observed upon pioglitazone treatment. Despite the reduction in insulitis, we did not observe differences in relative proportions of CD4ϩ T cell subtypes in the draining pancreatic lymph node. These findings raise the possibility that pioglitazone may have had a primary effect on reducing proliferation of T cells. We tested this possibility by performing T cell activation assays in vitro wherein no changes in CD4ϩ cell proliferation or differentiation were observed in response to pioglitazone. Studies have shown stimulatory and inhibitory effects of

. Pioglitazone suppresses PPAR-␥ (Ser-273) phosphorylation in ␤ cells in vitro and in islets of NOD mice in vivo.
A, MIN6 ␤ cells were preincubated in vehicle, 10 M pioglitazone (Pio), or 10 M roscovitine overnight, then extracts were subjected to immunoblotting using anti-phospho-PPAR-␥ (S273) and anti-acetyl-histone H3 (Lys-14) as a loading control. Representative immunoblots are shown, and the bar graph below shows the quantitation of immunoblots (normalized to loading control) from three independent experiments. * indicates that the value is significantly different (p Ͻ 0.05) compared with vehicle-treated (control) cells. B, 6-week-old pre-diabetic NOD mice were placed on either normal chow (Control) or chow containing 0.01 wt% pioglitazone (Pio). After 4 of weeks feeding, islets from 2 mice per group were isolated and subjected to immunoblotting using anti-phospho-PPAR-␥ (S273), anti-phospho-eIF2␣, and anti-acetyl-histone H3 (Lys-14) as a loading control.
PPAR-␥ on Treg and Th17 cells, respectively (14,49), so we cannot rule out the possibility that differences in CD4ϩ T cell subpopulations occurred in the immediate vicinity of the islets themselves. In the absence of a direct effect of pioglitazone on T cell proliferation, it nevertheless remains possible that the augmentation of the ␤ cell UPR with pioglitazone may have diminished the stimulus for autoimmunity. In this regard recent studies have shown that chemical restoration of ␤ cell UPR in type 1 diabetic mice results in a reduction in insulitis (3).
Certain key limitations to our study must be recognized. First, notwithstanding some evidence to the contrary, we cannot rule out non-␤ cell effects of pioglitazone in our study, and it remains possible that ␤ cell-protective effects may still indirectly emanate from the effects on adipose tissue, muscle, or immune cells. Second, the implications of our findings to human T1D remain speculative, but recent observations suggesting elevations in ␤ cell ER stress parameters in human T1D (3,10,25) suggest promise for the use interventions that enhance the UPR. Although pioglitazone and similar TZDs have fallen out of favor in clinical use, our findings linking pioglitazone treatment to reductions in phospho-PPAR-␥ (Ser-273) open the exciting possibility that non-TZD approaches to blocking Ser-273 phosphorylation may be an alternative approach to enhancing the ␤ cell UPR. Recently, imatinib (Gleevec) was shown to block PPAR-␥ (Ser-273) phosphorylation (50) and as such represents a non-TZD with diabetes-ameliorating effects (50,51). In this respect, an ongoing clinical trial (NCT01781975) is testing the potential for imatinib as therapy for recent-onset T1D. Taken together, our data support the conclusion that TZDs and similar-acting agents may have protective effects to reduce ␤ cell stress and death in T1D and also support the notion that an aggressive approach to ␤ cell function represents an important adjunctive means to control progression or severity of T1D. . Effect of pioglitazone on insulin secretion in pre-diabetic NOD mice. 6-Week-old pre-diabetic NOD mice were placed on either normal chow (Control) or chow containing 0.01 wt% pioglitazone (Pio). A, serum insulin levels at the indicated time points after intraperitoneal glucose injection (2.0 g/kg body weight) after 4 weeks of feeding (n ϭ 12 mice per group). B, glucose-stimulated insulin secretion from isolated islets from control and pioglitazone-fed mice (n ϭ 9 mice per group). C, random-fed insulin levels after 4 weeks of feeding (n ϭ 8 mice per group). D, random-fed proinsulin levels after 4 weeks of feeding (n ϭ 8 mice per group). E, random-fed proinsulin:insulin ratio after 4 weeks of feeding (n ϭ 8 mice per group). * indicates value is significantly different for the comparisons shown by two-tailed t test. FIGURE 6. Effect of pioglitazone on ␤ cell ER stress in pre-diabetic NOD mice. 6-Week-old pre-diabetic NOD mice were placed on either normal chow (Control) or chow containing 0.01 wt% pioglitazone (Pio). After 4 weeks of feeding, pancreas was harvested (n ϭ 5) or islets were isolated and subjected to RT-PCR (n ϭ 3 per group) or immunoblot analysis (n ϭ 3 per group). A, spliced Xbp1 mRNA (relative to Actb mRNA). B, immunoblot analysis and quantitation for ATF4. C, Gadd34 mRNA (relative to Actb mRNA). D, immunoblot analysis and quantitation for Bip. E, Chop mRNA (relative to Actb mRNA). * indicates that the values are significantly different (p Ͻ 0.05) by two-tailed t test.

Experimental Procedures
Cells, Animals, and Procedures-Mouse-derived MIN6 ␤ cells were cultured and maintained as previously described (52). Pioglitazone and roscovitine were dissolved in DMSO at 1000ϫ concentration and applied directly to cell culture medium at concentrations of 1 M or 10 M as indicated in the figures. Cells were preincubated with pioglitazone or roscovitine overnight before experimentation.
Female NOD/ShiLTJ (NOD) mice were purchased from The Jackson Laboratory (Bar Harbor, ME) and maintained at the Indiana University Laboratory Animal Resource Center under pathogen-free conditions according to protocols approved by the institutional animal care and use committee. Mice were allowed to acclimate for 1 week before experimentation. Mice were fed with either a standard chow diet or a standard chow diet containing 0.01 wt% pioglitazone (Harlan-Teklad Global). This concentration of pioglitazone was designed to deliver ϳ20 mg/kg body weight pioglitazone based on average daily food intake (22,39). Body weights and blood glucose levels were measured weekly. At 10 weeks of age, mice underwent a intraperitoneal GTT at 2 g/kg body weight of glucose as described previously (4). Serum during the GTT was collected for insulin measurements at 0 and 10 min post glucose injection. Lean body mass and fat mass was determined by dual-energy x-ray absorptiometry (DEXA) using a Lunar PIXImus2 densitometer (GE Medical Systems). Insulin tolerance tests were performed with 1.5 units/kg body weight of insulin injected intraperitoneally as described (22). At the end of the study, mice were euthanized, serum was collected, and pancreata harvested or islets isolated as described (53). Serum insulin was measured using an Ultra Sensitive Mouse Insulin ELISA kit (Crystal Chem), and serum proinsulin was measured using a Mouse Proinsulin ELISA kit (ALPCO). Some mice were stimulated with 10 units/kg insulin (Humulin; Lilly) 5 min before euthanasia, and liver and muscle were harvested.
Quantitative Real-time RT-PCR and Immunoblot Analysis-Total RNA from islets or MIN6 cells was isolated, reverse-transcribed, and subjected to quantitative real-time RT-PCR as described previously (54). Results of RT-PCR analysis were normalized to Actb levels. Primers for mouse spliced Xbp1, GADD34, Chop, and Actb were described previously (30,(55)(56)(57)(58). Measurement of methylated and unmethylated preproinsulin DNA in the serum by multiplex PCR was performed as we recently detailed (31).
Immunoblots were performed using 4 -20% polyacrylamide gel electrophoresis as previously described (59). Antibodies were used as follows: anti-phospho-AKT (Ser-473) (1:1000 dilution, Cell Signaling Technology #9271), anti-total AKT (1:1000 dilution, Cell Signaling Technology #2920), anti-BIP Immunohistochemistry, Immunofluorescence, ␤ Cell Area, and Insulitis Scoring-Pancreata were fixed, sectioned, and stained for insulin as described (22). Pancreata were stained for immunofluorescence for glucagon, CHOP, PCNA, cleaved caspase 3, or 4-HNE with insulin and counterstained with DAPI and quantified as described (4, 60). Images were captured using an AxioObserver Z1 equipped with a high resolution color FIGURE 7. Pioglitazone reduces oxidative stress, enhances ␤ cell mass, and suppresses ␤ cell death in pre-diabetic NOD mice. 6-Week-old prediabetic NOD mice were placed on either normal chow (Control) or chow containing 0.01 wt% pioglitazone (Pio). After 4 of weeks feeding, pancreas was harvested from animals (n ϭ 5). A, representative islet immunostaining from control and pioglitazone-treated mice for CHOP and insulin and corresponding quantitation of CHOPϩinsulinϩ cells. Arrows indicate cells that costain for insulin and CHOP. B, representative islet immunostaining for CC3 and insulin and corresponding quantitation of CC3ϩinsulinϩ cells. The arrow indicates cells that costain for insulin and CC3. C, representative islet immunostaining for 4-HNE and insulin and corresponding 4-HNE pixel intensity. D, ␤ cell area as a percentage of total pancreas area for animals at the start of the study (6 weeks of age) and for control and pioglitazone treated at the indicated ages. E, serum unmethylation index for control-and pioglitazonetreated mice (n ϭ 4 -5 per group). F, representative islet immunostaining from control and pioglitazone-treated mice for PCNA and insulin and corresponding quantitation of PCNAϩinsulinϩ cells. Arrows indicate cells that costain for insulin and PCNA. * indicates that the values are significantly different (p Ͻ 0.05) by two-tailed t test. camera or LSM 700 confocal (Carl Zeiss). ␤ cell area was calculated as described previously (61). For insulitis scoring, 3 pancreas sections at least 70 m apart from 5 animals per group were scored using the following grading scheme (62): grade 1, no islet-associated mononuclear cell infiltrates; grade 2, periinsulitis affecting Ͻ50% of the circumference of the islet without evidence of islet invasion; grade 3, peri-insulitis affecting Ͼ50% of the circumference of the islet with evidence of islet invasion; grade 4, islet invasion.
Flow Cytometric Analysis of T Cells-Single cell suspensions were prepared from harvested pancreatic lymph nodes from NOD mice (62,63). For Treg cell analyses, equal volumes of the single cell suspensions were stained using anti-CD4-FITC (eBioscience RM4 -5) and anti-CD25-APC (eBioscience PC61.5) antibodies and fixed overnight before being permeabilized and stained with anti-Foxp3-PE (eBioscience FJK-16s) antibody. For the Th1 and Th17 cell analyses, equal volumes of the single cell suspensions were first incubated with Cell Stimulation Mixture (eBioscience) for 4 h before staining with anti-CD4 antibody; cells were fixed overnight then permeabilized and stained for IL-17A (eBioscience eBio17B7) and IFN␥ (eBioscience XMG1.2) according to the manufacturer's instructions (eBioscience). Cells were analyzed using a BD FACSCalibur flow cytometer (BD Biosciences) and FlowJo software (TreeStar). For determination of cellular proliferation, single cell suspensions of unfractionated splenocytes from NOD mice were incubated in 5 M membrane dye carboxyfluorescein diacetate succinimidyl ester (CFSE) for 10 min at 37°C, diluted with 5 volumes of ice-cold media, and incubated for 5 min on ice. Samples were then washed with PBS three times and subjected to stimulation in vitro and subjected to flow cytometry as described above.
Statistical Analysis-All data are presented as the mean Ϯ S.E. One-way analysis of variance (with Dunnett's post-test) was used for comparisons involving more than two conditions, and a two-tailed Student's t test was used for comparisons involving two conditions. Prism 7 software (GraphPad) was used for all statistical analyses. Statistical significance was assumed at p Ͻ 0.05.