Estrogen receptor α protects pancreatic β-cells from apoptosis by preserving mitochondrial function and suppressing endoplasmic reticulum stress

Estrogen receptor α (ERα) action plays an important role in pancreatic β-cell function and survival; thus, it is considered a potential therapeutic target for the treatment of type 2 diabetes in women. However, the mechanisms underlying the protective effects of ERα remain unclear. Because ERα regulates mitochondrial metabolism in other cell types, we hypothesized that ERα may act to preserve insulin secretion and promote β-cell survival by regulating mitochondrial-endoplasmic reticulum (EndoRetic) function. We tested this hypothesis using pancreatic islet-specific ERα knockout (PERαKO) mice and Min6 β-cells in culture with Esr1 knockdown (KD). We found that Esr1-KD promoted reactive oxygen species production that associated with reduced fission/fusion dynamics and impaired mitophagy. Electron microscopy showed mitochondrial enlargement and a pro-fusion phenotype. Mitochondrial cristae and endoplasmic reticulum were dilated in Esr1-KD compared with ERα replete Min6 β-cells. Increased expression of Oma1 and Chop was paralleled by increased oxygen consumption and apoptosis susceptibility in ERα-KD cells. In contrast, ERα overexpression and ligand activation reduced both Chop and Oma1 expression, likely by ERα binding to consensus estrogen-response element sites in the Oma1 and Chop promoters. Together, our findings suggest that ERα promotes β-cell survival and insulin secretion through maintenance of mitochondrial fission/fusion-mitophagy dynamics and EndoRetic function, in part by Oma1 and Chop repression.

Type 2 diabetes manifests in humans when the pancreatic ␤-cells can no longer compensate to provide sufficient insulin to overcome peripheral insulin resistance. This pancreatic decompensation phase is typically associated with a marked decline in ␤-cell mass and insulin secretory capacity (1). It is hypothesized that a variety of factors, genetic and environmental, contribute to cellular toxicity, organelle dysfunction, and ␤-cell apoptosis (2). Thus, novel approaches or strategies to preserve ␤-cell function and mass are critical for the prevention of diabetes. Considering that there is a reduced prevalence of diabetes in premenopausal women compared with agematched men (3,4) and that female versus male ZDF rats are protected against ␤-cell failure (5) suggests that harnessing the protective effects of estradiol/ER␣ 4 may prove efficacious in the treatment or prevention of type 2 diabetes in women. Moreover, because pancreata biopsies from healthy female subjects showed a significant increase in ␤-cell number compared with males (6), there is additional evidence of a sex-specific effect enhancing ␤-cell viability that might serve to improve islet transplantation outcomes for type 1 diabetic subjects.
Through the use of mouse genetics and the development of a pancreatic islet-specific estrogen receptor ␣ knockout (PER␣KO) mouse model, researchers have highlighted the mechanistic importance of ER␣ in islets/␤-cells as a regulator of insulin synthesis (7) and protector of ␤-cell health even during extreme tissue stress (8,9). Although the protective actions of estrogens/ER␣ in pancreatic ␤-cells are well documented, the mechanistic underpinnings remain incompletely understood.
Considering the high secretory burden of the islet, efficient protein synthesis and folding are critical for ␤-cell function and health. The EndoRetic is an organelle in which secretory proteins are folded to their native conformations; however, when the organelle becomes overwhelmed by unfolded or misfolded proteins, a classical stress response is triggered (10,11). Endoplasmic reticulum stress activates the unfolded protein response (UPR) signaling network that can, depending upon the severity and duration of stress, initiate cytochrome c release from the mitochondria to trigger ␤-cell apoptosis. There is a strong collaborative requirement between mitochondria and the endoplasmic reticulum especially in ␤-cells where proteostasis imposes a high energy demand on the cell. Thus, we hypothesize that chronic mitochondrial-EndoRetic stress as a consequence of impaired ER␣ action may contribute to apoptosis susceptibility and diminished insulin secretory capacity that underlies glucose intolerance and type 2 diabetes conversion as pathology progresses.
Herein, we show that pancreatic islet expression of ER␣ promotes ␤-cell survival by maintaining mitochondrial health while suppressing EndoRetic stress. ER␣-deficient Min6 ␤-cells showed imbalanced mitochondrial fission/fusion-mitophagy dynamics, increased expression of the mitochondrial stress gene Oma1, and excessive production of reactive oxygen species (ROS). We hypothesize that these alterations in mitochondrial function exacerbated EndoRetic stress, including the induction of Chop expression, and this was a primary factor promoting ␤-cell apoptosis.
Collectively, our findings support the notion that ER␣ preserves islet ␤-cell mass and protects against oxidative and EndoRetic stress through Oma1 and Chop repression. These data highlight the important actions of ER␣ in ␤-cells and support a potential therapeutic role for ER␣ in combating the onset of type 2 diabetes in women.

ER␣ knockout impairs glucose-stimulated insulin secretion and promotes inflammation
Previous studies by Mauvais-Jarvis and co-workers (8) have shown that islets from PER␣KO mice are more susceptible to lipid-and streptozotocin-induced damage. Herein, we sought to identify ER␣-regulated pathways central in the control of ␤-cell health. First we confirmed that Esr1 (the gene that encodes ER␣) knockdown (KD) reduced glucose-stimulated insulin secretion in Min6 ␤-cells (Fig. 1, A and B). Consistent with previous findings by Mauvais-Jarvis and co-workers (8) in pancreatic samples obtained from f/f and PER␣KO mice (9), we observed no difference in total insulin content between the genotypes of Min6 ␤-cells (Fig. 1C). Thus, the reduction in glucose-stimulated insulin secretion appears mediated by defects in secretory function and not total insulin content per se. Indeed, the reduction in insulin secretion was associated with a marked increase in inflammation as reflected by elevated expression of IL6, IL1␤, and monocyte chemoattractant (Mcp) 1 (Fig. 1D), consistent with previous findings showing defective glucosestimulated insulin secretion from ␤-cells under stress (8,9).
Because mitochondria and the endoplasmic reticulum are important in regulating ATP production, cellular redox, pH, Ca 2ϩ signaling, and cell death (12), maintenance of a healthy mitochondrial network is essential for cellular function. Thus, macroautophagy and more specifically mitophagy (the elimination of damaged or dysfunctional mitochondria) are elemental processes critical for maintaining an optimal mitochondrial reticulum and important for apoptosis prevention. Considering that we have previously shown impaired autophagy signaling in ER␣-deficient muscle cells and because autophagy is induced by cellular stress and impaired by elevated lipids, we studied the impact of ER␣ on the mitochondrial-EndoRetic-autophagy axis in mice and cells.

ER␣ knockdown impairs autophagic flux in Min6 ␤-cells
To investigate the mechanisms by which ER␣ regulates autophagy in pancreatic ␤-cells, we performed Esr1 knockdown (KD) in Min6 ␤-cells. Shaw and co-workers and Guan and co-workers (13)(14)(15) have consistently shown that AMPK is an important cellular energy sensor that regulates both mitochondrial dynamics and macroautophagy. Because it was previously shown that AMPK activity can be driven by estrogen action, we hypothesized that AMPK signaling may be diminished in Esr1-KD cells. Indeed, total AMPK␣ protein was significantly diminished as was Thr-172 phosphorylation status and AMPK␣ catalytic subunit mRNA (Prkaa2) in ER␣-deficient Min6 ␤-cells (Fig. 2, A-C). No change was observed in the expression level of Prkaa1, the gene encoding the AMPK ␣1 catalytic subunit. To confirm the regulation of Prkaa2 expression by ER␣, we treated Min6 ␤-cells with an ER␣ agonist for 12 and 24 h (Fig. 2D). ER␣ activation by 4,4Ј,4Љ-(4-propyl-[1H]pyrazole-1,3,5-triyl) trisphenol (PPT; 10 nM) led to a 2-3-fold increase in Prkaa2 expression, and this finding is consistent with observations in other cell types showing a role for ER␣ in the activation of AMPK (16,17). In ER␣-deficient Min6 ␤-cells, decreased AMPK␣ signaling was associated with a reduction in the phosphorylation of Ulk1 Ser-467 (upstream activator of macroautophagy) (Fig. 2, A, E and F).
Similar to our findings for expression of autophagy proteins and transcripts in islets from PER␣KO mice, autophagy-related ER␣ protects ␤-cells from apoptosis proteins, including Atg3, -5, -7 as well as Beclin-1 total protein, were reduced in ER␣ knockdown Min6 ␤-cells (Fig. 2G). In contrast to PER␣KO islets, however, mRNA expression levels of Atg3 and -5 were unchanged, whereas expression of Atg7 and -12 was elevated in Esr1-KD versus control (Scr) Min6 ␤-cells (Fig. 2H). Thus, the differential expression of transcript and protein observed in Esr1-KD cells suggests an alteration in protein translation or turnover.
The processing of LC3B, including conversion of LC3BI to II (LC3B cleavage and lipidation), is often used as a surrogate marker of autophagic flux. Our findings show a significant reduction in both LC3B I and II protein levels as well as abundance of p62 (a polyubiquitin binding protein responsible for recruiting LC3 to autophagosomes) (Fig. 2I). Reduced LC3B and p62 protein levels were likely explained by a reduction in transcription and not accelerated protein turnover, as gene expression levels for both Sqstm1 and Map1lc3b were reduced in Esr1-KD cells versus control (Scr) (Fig. 2J). A similar but more robust reduction in Map1lc3b expression was observed in islets from PER␣KO mice compared with ER␣ replete WT animals (Fig. 2K) suggesting that the impairment in autophagy signaling was conserved between ␤-cells in culture and islets in mice.
To confirm the impairment in autophagic flux in Esr1-KD Min6 ␤-cells, we transfected cells with tandem RFP-GFP-labeled LC3B. Because GFP fluorescence is quenched in the  (2-20 mM), and insulin secretion into the media was assessed by ELISA (performed in triplicate). C, total insulin content was not different between the genotypes (n ϭ 3/genotype in triplicate). D, Esr1-KD promotes inflammation in Min6 ␤-cells as reflected by a 2-4-fold induction of IL6, IL1␤, and MCP1 gene expression (n ϭ 6/genotype). E and F, islet-specific deletion of ER␣ in female mice confirmed a reduction in expression of ER␣ and known ER␣ target genes. G and H, reduced expression and protein abundance of key autophagy-related signaling factors (n ϭ 5-6/genotype). Values are mean Ϯ S.E., and significant differences between Esr1-KD and control (Scr) as well as PER␣KO and control f/f were detected by Student's t test and one-way ANOVA where appropriate, significance ϭ *, p Ͻ 0.05. AU, arbitrary units.

ER␣ protects ␤-cells from apoptosis
acidic pH of the lysosome, the GFP-LC3B signal reflects autophagosomes prior to lysosomal degradation. Because RFP is pH-resistant and continues to fluoresce, the RFP-LC3B signal can be used to quantify both autophagosomes and autolysosomes (18,19). The apparent reduction in red fluorescence punctae observed in ER␣-deficient versus control (Scr) Min6

ER␣ protects ␤-cells from apoptosis
␤-cells is fully consistent with reduced autophagosome turnover, i.e. autophagic flux (Fig. 2L). Collectively, these data indicate that inactivation of ER␣ suppresses autophagic signaling and autophagosome turnover in part by reduced abundance of autophagy-related proteins and impaired AMPK signaling.

ER␣ regulates mitochondrial morphology and fission/fusion dynamics in Min6 ␤-cells
The process of selective mitochondrial degradation is dependent on critical signaling shifts in fission/fusion dynamics and Park family member targeting of damaged or unwanted mitochondria to autophagic turnover by the lysosome (20). Mitochondrial fission proteins, Drp1 (dynamin-related protein) and mitochondrial fission factor (Mff), as well as mitochondrial fusion proteins mitofusin (Mfn)1 and Mfn2, were markedly reduced in ER␣-deficient Min6 ␤-cells (Fig. 3A). OPA1 (optic atrophy type 1) protein was more highly cleaved in Esr1-KD cells, and this is a likely consequence of increased Oma1 (overlapping activity with m-AAA protease) gene and protein expression (Fig. 3, B and C). No difference in the expression of the OPA1-associated protease presenilin-associated rhomboid-like protein, Parl, was detected between the groups (Fig. 3, B and C). Interestingly, although we observed no difference in the mRNA expression of the inner mitochondrial membrane metalloprotease Yme1L1 between Scr versus KD, protein levels of YME1L1 were 50% of control in ER␣-deficient cells. Because OPA1 is coordinately controlled by the actions of YME1L1 versus OMA1, our findings are consistent with an increased influence of OMA1 on OPA, and a loss of YME1L1 action to yield mitochondrially-controlled apoptotic resistance, cristae morphogenesis, and cell proliferation as described previously (21)(22)(23).
Considering these robust differences in mitochondrial dynamics signaling between the genotypes, we explored the morphology of mitochondria using transmission electron microscopy. Micrographs of Min6 ␤-cells showed that ER␣ deletion promoted an enlarged, elongated, and fused mitochondrial phenotype (Fig. 3, D-F). Morphological quantification indicated a significant increase in the individual mitochondrion area and perimeter in Esr1-KD Min6 ␤-cells, but without alteration in overall mitochondrial number (Fig. 3, D-F). Of note, mitochondrial cristae were organized into longitudinal ribbons stretching in a head-to-tail configuration in Esr1-KD ␤-cells. This observation is in contrast to control (Scr) showing the characteristically compact mitochondrial cristae folding (Fig. 3D).
In parallel with this elongated mitochondrial phenotype, we found a reduced abundance of basal Pink1 and Parkin protein in ER␣-deficient Min6 ␤-cells (Fig. 3G). Although we observed no change in Park2 (Parkin) gene expression, Park6 (Pink1) was up-regulated 4-fold (Fig. 3H). Parkin-mediated ubiquitination of outer mitochondrial membrane substrates amplifies Pink1 phosphorylation of ubiquitin to promote mitophagy (24). Consistent with a reduction in Parkin protein levels in whole-cell lysates, we found a reduction in the mitochondrial localization of Parkin protein in ER␣-deficient Min6 ␤-cells (Fig. 3I). Of interest, we detected a small (ϳ8%) yet significant increase in mitochondrial DNA (mtDNA) copy number (Fig. 3J) that was paralleled by increased expression of mitochondrial biogenesis markers Pgc1a and Tfam (Fig. 3K) in ER␣-deficient Min6 ␤-cells versus control (Scr). Considering the observation of mitochondrial elongation/hyperfusion and the reduction in mitophagy markers in Esr1-KD cells (we previously observed similar findings in ER␣-deficient muscle cells (25)), we hypothesize that loss of ER␣ contributes to a reduction in mitochondrial turnover. We speculate that in the absence of ER␣, the cell maintains mtDNA content at the expense of mitochondrial health because mitochondrial health surveillance signaling through Pink1 and Parkin is markedly suppressed. Thus, these data show that ER␣ is critical in the regulation of ␤-cell mitochondrial dynamics and mitophagy signaling.

ER␣ deficiency alters mitochondrial bioenergetics in Min6 ␤-cells
Mitochondria are highly dynamic organelles that constantly undergo fission/fusion cycling to retain a healthy mitochondrial network allowing for the maintenance of oxidative metabolism. In light of the mitochondrial dysmorphology and reduction of fission/fusion dynamics signaling that we observed in ER␣-deficient ␤-cells, we hypothesized that mitochondrial respiration may also be altered in the absence of ER␣. Indeed, we found that cellular oxygen consumption and proton leak were modestly yet significantly elevated in ER␣-deficient Min6 ␤-cells (Fig. 4, A and B). Despite an elevated oxygen consumption rate for Esr1-KD cells, the % ATP-associated respiration, and thus the relative rate of ATP synthesis, was identical between the genotypes (Fig. 4C). Impaired mitochondrial homeostasis is often accompanied by a mitochondrial stress response. Indeed, the increased proton leak was paralleled by elevated levels of H 2 O 2 and O 2 Ϫ (Fig. 4, D and E) in Min6 ␤-cells lacking ER␣. Treatment of Min6 ␤-cells with the mitochondrial membrane depolarizing agent carbonyl cyanide 3-chlorophenylhydrazone (CCCP) exacerbated the increase in H 2 O 2 production in the context of ER␣-deficiency (Fig. 4, D and E). Despite the increase in proton leak in Esr1-KD, we observed no difference in expression of Ucp2 or Ucp3 between the genotypes (Fig. 4F). We have shown previously that scavenging of ROS is impaired in ER␣-deficient muscle (25), and indeed we find a similarly reduced expression and failed ER␣ ligand (PPT)-stimulated induction of Gpx3 in Esr1-KD Min6 ␤-cells total AMPK and phosphorylated (Ser-467) and total ULK1 from control (Scr) and Esr1-KD ␤-cells (n ϭ 6 -9/genotype). C and D, qPCR analysis of AMPK expression, Prkaa1 and Prkaa2, and Pkraa2 mRNA at 0, 12, and 24 h post-treatment with 10 nM PPT (n ϭ 3/condition in triplicate). G and H, immunoblots and densitometric analysis as well as qPCR of select autophagy-related proteins/genes (n ϭ 3/genotype in triplicate). I-K, immunoblots and densitometry of p62 and LC3B I and II as well as mRNA levels in Min6 ␤-cells with Esr1-KD and in PER␣KO islets. L, flow cytometry analyses of dually-labeled RFP-GFP LC3B expressed in control (Scr) and Esr1-KD ␤-cells. Bars depicting quantification of dual signals represent autophagosome abundance (closed bars); RFP signal represents autolysosome abundance (dotted bars); and cells showing loss of signal represent autophagosome turnover (gray bars). In some cases, immunoblots from different panels were derived from the same membrane of a specific experiment, and protein levels were normalized to either tubulin or ␤-actin. Values are mean Ϯ S.E., and significant differences between genotypes were detected by Student's t test, significance ϭ *, p Ͻ 0.05. AU, arbitrary units. (Fig. 4G). Collectively, in the absence of ER␣, there is an increase in cellular respiration and ROS production that is associated with aberrant mitochondrial cristae morphology and imbalanced OPA1 regulation by YME1L1 and OMA1. The increased oxygen consumption by ER␣-deficient ␤-cells is likely an attempt to maintain ATP production.  Fig. 2G; thus, the ␤-actin immunoblots were duplicated between the panels). B, Parl, Oma1, and Yme1L1 gene expression. C, immunoblots and densitometry quantifying protein abundance in control (Scr) and Esr1-KD Min6 ␤-cells (n ϭ 6 -9/genotype). D, transmission electron microscopy analysis of mitochondria in control (Scr) and Esr1-KD Min6 ␤-cells (magnification ϫ2500). E and F, ImageJ analysis of mitochondrial area, perimeter, number, and size frequency in control (Scr) and Esr1-KD Min6 ␤-cells (four slides/genotype). G, immunoblots of mitophagy-associated proteins in control (Scr) and Esr1-KD Min6 ␤-cells (n ϭ 6 -9/genotype). H, qPCR analyses of mRNA expression of Pink1 (Park6), Parkin (Park2), and DJ1 (Park7) in control (Scr) and Esr1-KD Min6 cells (n ϭ 6 -9/genotype in duplicate). I, immunoblots of Parkin in cytosolic and mitochondrial subcellular fractions from control (Scr) and Esr1-KD Min6 cells. J, mtDNA copy number normalized to nuclear DNA (n ϭ 3/genotype in triplicate); K, genes associated with mitochondrial biogenesis (Pgc1a and Tfam) were elevated in Esr1-KD versus control (Scr) Min6 ␤-cells. Values are mean Ϯ S.E., and mean differences between the groups were detected by Student's t test and ANOVA where appropriate, *, p Ͻ 0.05. AU, arbitrary units.

ER␣ protects ␤-cells from apoptosis ER␣ knockdown increases EndoRetic stress and apoptosis in Min6 ␤-cells
Endoplasmic reticulum (EndoRetic) stress is often accompanied by oxidative stress especially in endocrine cell types responsible for high rates of protein synthesis (26,27). Excessive ROS production could both directly and indirectly impact EndoRetic homeostasis and protein folding thus contributing to increased susceptibility for cellular apoptosis. Electron micrograph images show a marked dilation of EndoRetic lumen in ER␣-deficient Min6 ␤-cells (Fig. 5A), a classic morphological feature of the UPR. These images led us to examine EndoRetic stress signaling in the context of ER␣ deficiency. Phosphorylation of PKR-like ER kinase (PERK) and eukaryotic eIF2␣ Ser-51 were markedly elevated in Esr1-KD cells indicating a strong response to reduce translation and protein synthesis (Fig. 5B). Indeed, mRNA expression and protein levels of C/EBP homo- Magnification: ϫ1500 (upper panels) and ϫ25,000 (lower panels). B, immunoblot analysis of EndoRetic stress markers in control (Scr) and Esr1-KD Min6 ␤-cells (n ϭ 6/group). C, qPCR analysis of gene expression in control (Scr) and Esr1-KD Min6 ␤-cells (n ϭ 6/group). D, ER␣ deletion in PER␣KO islets led to increased Chop and Gadd34 expression compared with islets from control f/f (n ϭ 3-4 mice/genotype in triplicate). E, immunoblot of cleaved caspase-3 in control (Scr) and Esr1-KD ␤-cells with corresponding densitometry (n ϭ 6/group). F and G, fluorescence microscopy analyses of control (Scr) and Esr1-KD Min6 ␤-cells with nuclei stained blue (DAPI) and apoptotic cells stained red; F, positive and negative control; G, control (Scr) and Esr1-KD. H, apoptosis, assessed by flow cytometry for FITC-BrdU in control (Scr) and Esr1-KD Min6 ␤-cells and expressed as a percentage of apoptotic cells (n ϭ 3/group in duplicate). In specific instances, immunoblots from different panels were derived from the same membrane of a specific experiment, and protein levels were normalized to either GAPDH or ␤-actin. Values are mean Ϯ S.E., *, p Ͻ 0.05, significant difference from control (Scr) cells or versus control f/f. AU, arbitrary units.

ER␣ protects ␤-cells from apoptosis
logous protein (CHOP), a master regulator of EndoRetic stress and apoptosis, were markedly elevated in Esr1-KD cells (Fig. 5, B and C). Of interest, Gadd34 expression was elevated while Atf4 expression was markedly reduced in Esr1-KD cells. Similarly, we observed a marked up-regulation of Chop and Gadd34 expression in islets from female PER␣KO mice versus control f/f (Fig. 5D). The differential regulation of Atf4 and Gadd34 may reflect an inadequate compensatory attempt to down-regulate EndoRetic stress by counteracting CHOP action (Fig. 5, C and D). The reduction in Atf4, a transcription factor known to induce Chop expression, suggests that Chop up-regulation is achieved independent of Atf4, and thus a plausible explanation could be by direct derepression of Chop by ER␣. No change in upstream stress signaling of p-XBP1 and XBP1 protein levels was observed, despite an ϳ80% reduction in spliced XBP1 expression in Esr1-KD versus control (Scr) Min6 ␤-cells (Fig. 5, B and C).
Our finding of increased CHOP expression is internally consistent with the increase in cleaved caspase levels in ER␣-deficient Min6 ␤-cells (Fig. 5E). To confirm the increase in cellular apoptosis, we performed TUNEL staining in control (Scr) and Esr1-KD Min6 ␤-cells (Fig. 5, F and G). Fluorescence microscopy images indicate an increased number of TUNEL-positive stained cells in ER␣-deficient Min6 ␤-cells (Fig. 5, F and G). Flow cytometry analysis showed nearly a 3-fold increase in the percentage of apoptotic cells in ER␣-deficient Min6 ␤-cells (Fig. 5H). These findings combined with increased cleaved caspase-3 levels are indicative of increased apoptosis susceptibility in ER␣-deficient Min6 ␤-cells. Moreover, because Atf4 is also thought to be involved in stress-induced autophagy gene expression, our findings of a reduction in Atf4 are consistent with findings in Figs. 1  and 2 showing impaired autophagic signaling.
We next determined whether elevated ROS, comparable with levels observed in Esr1-KD cells, alone could induce a similar degree of EndoRetic stress in "control" Min6 ␤-cells. EndoRetic stress signaling was quantified after exposure to exogenous H 2 O 2 , an insult shown to cause ␤-cell injury in the context of diabetes. In Min6 ␤-cells, H 2 O 2 treatment increased CHOP and cleaved caspase-3 levels (Fig. 6, A and B). These data prompted us to interrogate whether ER␣ activation protects ␤-cells against oxidative stress-induced Chop induction and caspase activation.

ER␣ protects against oxidative stress-induced CHOP induction in Min6 ␤-cells
Because our findings indicate that inactivation of ER␣ triggers ROS-induced EndoRetic stress and subsequent ␤-cell apoptosis, we next hypothesized that ER␣ overexpression could exert a protective role in Min6 ␤-cells. Moderate overexpression of ER␣ via plasmid transfection led to increased levels of total AMPK␣ protein, enhanced LC3B processing, and reduced CHOP protein expression (Fig. 7A). Moreover, we found an ER␣ dose-dependent reduction in Chop expression in adenovirus-transduced Min6 ␤-cells (Fig. 7B). Next, we treated GFP or HA-tagged ER␣-transfected cells with H 2 O 2 and observed that ER␣ blunted the induction of CHOP protein (Fig. 7C). Moreover, to determine whether pharmacological activation of ER␣ exerts a protective effect against oxidative stress, Min6 ␤-cells were treated with H 2 O 2 and the ER␣-specific agonist PPT alone or with the ER␣ antagonist ICI 182,780 (Fulvestrant TM ). As hypothesized, PPT treatment suppressed the H 2 O 2 -induced increase in CHOP protein, although in contrast, ICI 182,780 blunted the PPT reduction in CHOP protein (Fig.  7D). In addition to blocking H 2 O 2 -induced EndoRetic stress, ER␣ activation by PPT also reversed thapsigargin (Tg)-induced EndoRetic stress phosphorylation of eIF2␣ and diminished CHOP protein expression (Fig. 7, E-G). Our findings suggest a strong role of ER␣ action to protect against stimuli that promote EndoRetic stress (i.e. H 2 O 2 and Tg).

ER␣ represses Chop and Oma1 expression
Considering that ER␣ is a transcription factor shown to induce or repress target gene expression in a cell-specific context, and because Min6 ␤-cells treated with PPT (10 nM, 6 -24 h) show a robust reduction in Chop and Oma1 expression (Fig.  8, A and B), we next interrogated whether Oma1 and Chop (Ddit3) gene expression is directly regulated by ER␣. We performed chromatin immunoprecipitation studies to show that ER␣ robustly binds the Oma1 and Chop promoters in Min6 ␤-cells (Fig. 8C). Indeed, we found that compared with vehicle treatment, E 2 induced ER␣ binding to an ERE consensus sequence within the Chop promoter (Fig. 8C) and possibly two ERE sites within the Oma1 promoter. Thus, our data suggest that under basal conditions, ER␣ likely represses Oma1 and Chop expression by direct binding to an estrogen-response element within the respective promoter regions. In the context of Esr1-KD, we hypothesize that the deletion of ER␣ removes the molecular brake from the promoter thus inducing Oma1 and Chop gene expression by a de-repression mechanism. In addition to the direct action of ER␣, we also asked the question as to whether mitochondrial stress induced by Oma1 de-repression could cause activation of CHOP via mitochondrial-EndoRetic cross-talk. To answer this question, we performed gain-and loss-of-Oma1-function studies in Min6 ␤-cells using adenovirus and siRNA, respectively. We found that overexpression of HA-tagged OMA1 promoted PERK and eIF2␣ phosphorylation and increased CHOP protein markedly (Fig. 8, D and E). Conversely, Oma1 knockdown reduced CHOP protein and cleaved

ER␣ protects ␤-cells from apoptosis
caspase-3 even under basal conditions (Fig. 8, F and G). Our data strongly suggest that reduced ER␣ expression promotes the up-regulation of Oma1 and Chop to induce mitochondrialendoplasmic reticulum stress and heighten apoptosis susceptibility in ␤-cells (Fig. 8H).

Discussion
Diabetes and its associated comorbidities are a leading cause of death worldwide with both type 1 and type 2 diabetes characterized by progressive ␤-cell failure (1, 28). Two randomized trials have shown that estrogen replacement therapy promotes an ϳ30% reduction in the incidence of type 2 diabetes in postmenopausal women (29,30), although the mechanism(s) of protection remain incompletely understood. These clinical observations have fueled mechanistic studies aimed at determining the tissue contribution of estradiol action in the maintenance of ␤-cell mass and the protection against type 2 diabetes progression (9,(31)(32)(33)(34)(35)(36). Taken together with the previous work by Mauvais-Jarvis and co-workers (31,(37)(38)(39), our findings shed light on additional ER␣-regulated targets controlling ␤-cell apoptosis susceptibility and insulin secretory capacity.

ER␣ protects ␤-cells from apoptosis
Herein, we studied the protective mechanisms of ER␣ in pancreatic ␤-cells using a pancreatic-specific ER␣ knockout mouse model and ER␣-deficient Min6 ␤-cells. We found that ER␣ protects ␤-cells from apoptosis through regulation of mitochondrial function and morphology, autophagy, and EndoRetic stress.
Previous studies have shown increased liver X receptor (LXR)-stimulated lipogenesis in ␤-cells lacking ER␣. An additional mechanism contributing to increased triacylglycerol accumulation is a reduction in the oxidation of fatty acids by islet mitochondria. Indeed we have previously observed that ER␣ is critical for mitochondrial oxidative metabolism in skeletal muscle (25). Herein, we show a marginal but significant increase in oxygen consumption rate for ER␣-deficient cells. Despite this increase in oxygen consumption rate, % ATPlinked respiration was identical between the genotypes, and this was a likely consequence of increased proton leak and elevated production of reactive oxygen species, including H 2 O 2 and O 2 . F and G, confirmation studies using siRNA to achieve Oma1 knockdown confirm proof-ofconcept that OMA1 expression is linked with EndoRetic stress signaling and cleaved caspase-3 abundance via CHOP, even in the basal state (n ϭ 3/condition in duplicate). H, schematic showing that ER␣ promotes cellular stress and apoptosis protection by direct repression of Chop and Oma1 expression. Our findings also support the notion that the derepression of Oma1 could contribute to the induction of CHOP1 and, in part, underlie EndoRetic stress associated with impaired ER␣ action. AU, arbitrary units.

ER␣ protects ␤-cells from apoptosis
due to alterations in mitochondrial morphology in Esr1-KD ␤-cells.
Moreover, impaired mitochondrial fission/fusion flux may be an important contributor to the altered bioenergetics of ER␣-deficient ␤-cells. Others have shown that imbalanced mitochondrial fission/fusion dynamics promotes the production of ROS. Indeed, we have shown impaired bioenergetic efficiency, increased ROS production, and impaired fission/fusion dynamics previously in ER␣-deficient muscle cells (25).
Because mitochondria are highly dynamic organelles that intermix via rapid fission/fusion processing, an imbalance in this process polarizes the organelle in a static state, and this inflexibility in remodeling is often associated with pathology. Reorganization of the mitochondrial network allows for faster and more efficient removal of specific mitochondrial contents or entire organelles via autophagic turnover or mitophagy (40). Immunoblotting analyses revealed a marked reduction in both mitochondrial fission and fusion signaling, and electron microscopy analysis revealed a markedly elongated mitochondrial architecture accompanied by altered cristae structure in ER␣-deficient Min6 ␤-cells. Although mitochondrial fragmentation has often been described in connection with induction of apoptosis (41,42), some have observed a hyperfused mitochondrial network in connection with increased apoptosis. Thus, resolution of mitochondrial phenotypes contributing to apoptotic susceptibility requires further investigation.
Considering the increase in OPA1 cleavage, the marked reduction in YME1L1, and the significant increase in OMA1 expression, our data indicate a potential increase in inner membrane fragmentation accompanied by enhanced outer membrane associations (23,43,44). The imbalance in OMA1/ YME1L1 is thought to increase cellular apoptosis susceptibility. It has recently been shown that OMA1 is activated by Bcl2 family members Bax and Bak to promote inner and outer mitochondrial membrane permeabilization and cytochrome c release; thus, OMA1 is now more recently viewed as a molecular switch to engage apoptotic signaling (45).
Previous studies in rodents demonstrated that autophagy plays an important role in the maintenance of pancreatic ␤-cell mass by regulating cell survival and proliferation (46 -48). We found that expression levels of several autophagy-related genes were markedly decreased in islets in the context of ER␣ deletion, and indeed reduced expression of Atg5 and Atg7 has been previously associated with ␤-cell dysfunction (49). Our studies in Min6 ␤-cells also revealed that ER␣ deficiency interrupts autophagic signaling potentially in part through the suppression of AMPK␣ signaling (15,50,51).
Although the expression of p62 as a marker of autophagy remains controversial, our findings showed that both mRNA and protein levels of p62 were decreased in ER␣-deficient ␤-cells. We believe that the reduction p62 observed in ER␣deficient Min6 ␤-cells was due to interrupted p62 mRNA expression and not as a consequence of increased p62 degradation by autophagy. The notion of reduced p62 turnover in the context of ER␣ deficiency is internally consistent with the reduction in Atg5 and -7, and our expression data were confirmed functionally by flow cytometry analyses showing diminished autophagosome degradation.
Mitophagy is an autophagy-dependent mitochondrial quality control mechanism that eliminates dysfunctional organelles through autophagosomal degradation by the lysosome (20,52). In parallel with diminished macroautophagy and fission/fusion dynamics, a reduction of mitophagy signaling by Pink1 and Parkin was also observed in ER␣-deficient Min6 ␤-cells (24,53). We hypothesize that reduced mitochondrial turnover coupled with enhanced biogenesis (driven by Pgc1␣ and Tfam) led to the increase in mtDNA copy number observed in Esr1-KD ␤-cells. Thus, similar to our findings in ER␣-deficient skeletal muscle, the cell maintains or increases mitochondrial DNA content/ mitochondrial number in an attempt to preserve metabolism at the expense of mitochondrial DNA health. The mitochondrial architecture observed in Esr1-KD cells supports this notion since mitochondria elongate and fuse to promote coupling efficiency, to escape mitophagic turnover, and to dilute damaged contents across the network; indeed, all of these actions appear engaged in Esr1-KD cells.
In an attempt to mimic the oxidative stress observed in ER␣deficient ␤-cells, we treated wildtype ER␣-replete cells with H 2 O 2 . Exposure to ROS increased CHOP and cleaved caspase-3 expression in ER␣-replete Min6 ␤-cells. In agreement with these findings, ER␣-deficient Min6 ␤-cells that basally produce increased ROS showed a similar increase in CHOP expression. Additionally, PERK phosphorylation was significantly increased 4-fold in H 2 O 2 -treated cells, and this was associated with a similar induction of cleaved caspase-3 as seen in Esr1-KD cells (54). Moreover, consistent with elevated CHOP, the EndoRetic lumen was noticeably dilated in ER␣-deficient Min6 ␤-cells, an observation consistent with marked EndoRetic stress. This increase in CHOP and caspase-3 expression was linked with increased cellular apoptosis.
In contrast to our findings in KD cells, overexpression of ER␣ and pharmacological treatment of Min6 ␤-cells with PPT inhibited basal expression of Oma1 and Chop and prevented H 2 O 2 -induced Chop induction. Furthermore, both the Chop and Oma1 promoters contain multiple consensus ERE sequences, and we detected direct ER␣ binding at a minimum of one site by chromatin immunoprecipitation analysis. Our findings strongly suggest that ER␣ protects ␤-cells against EndoRetic-mitochondrial stress through transcriptional repression of Chop and Oma1.
Importantly, previous studies show that Chop inhibition is sufficient to normalize cleaved caspase-3, restore autophagic flux, reduce ␤-cell apoptosis, and delay the onset of diabetes in human islet amyloid polypeptide (h-IAPP) transgenic mice (55,56). These data support our findings that in the context of ER␣ deficiency, activation of Chop by de-repression alone could be sufficient to drive ␤-cell apoptosis susceptibility. Similarly, Oma1 inhibition has also been shown to protect neuronal cells against oxidative stress-induced apoptosis, and now our findings appear to extend this paradigm to pancreatic ␤-cells (57). Moreover, we now provide evidence of a novel cross-talk mechanism by which OMA1 induction promotes EndoRetic stress signaling through PERK-CHOP.
Although previous studies have shown that estradiol protects pancreatic ␤-cells from oxidative injury and apoptosis, our findings point to a mitochondrial-endoplasmic reticulum ER␣ protects ␤-cells from apoptosis nexus and specific ER␣ regulatory networks (including Oma1 and Chop) involved in estradiol-mediated ␤-cell protection. Overall, our findings support the notion that ER␣ may be a target of therapeutic potential in preventing type 2 diabetes in women.

Animals
Pancreatic islet-specific ER␣ knockout mice (PER␣KO) were generated using the Cre-loxP system by mating a Pdx1-Cre transgenic line with the floxed ER␣ line (f/f) obtained as a generous gift from Dr. Kenneth Korach as described previously (7,9). Animals were bred by the Hevener Laboratory at UCLA, and islet protein and mRNA samples from PER␣KO and f/f mice were also provided by Dr. Mauvais-Jarvis such that two independent cohorts of mice were studied to ensure reproducibility of findings. Animals had free access to food (normal chow) and water and were housed using a 12-h light/dark cycle. All procedures were performed in accordance with the Guide for Care and Use of Laboratory Animals of the National Institutes of Health and were approved by the Animal Subjects Committee of UCLA.

Cell culture, reagents, and transfections
Min6 ␤-cells were maintained in high-glucose DMEM, 15% fetal bovine serum, 10 mM Hepes, 50 M 2-mercaptoethanol with penicillin/streptomycin. For ligand-specific studies, cells were cultured in phenol red-free DMEM with 15% charcoal-stripped fetal bovine serum, 10 mM Hepes, 50 M 2-mercaptoethanol with penicillin/streptomycin for 3 days prior to treatment with 10 -100 nM PPT (as indicated) and 1 M ICI 182,780. To mimic the oxidative and EndoRetic stress observed as a consequence of ER␣ deletion, Min6 ␤-cells were treated with either H 2 O 2 (100 M for 24 h) or thapsigargin (500 nM for 18 h; EndoRetic Ca 2ϩ -ATPase noncompetitive inhibitor). Adenoviral constructs containing GFP or ER␣ were generated using the ViraPower adenoviral expression system (Invitrogen) according to the manufacturer's suggestions. pcDNA 6.2-GFP plasmids were purchased from Invitrogen. pBS-HA-ER␣ plasmids were a gift from Dr. Peter Tontonoz. pcDNA-HA-Oma1 plasmids were a gift from Alexander M. van der Bliek. Scrambled siRNA (sc-37007) and Oma1 siRNA (sc-151297) were purchased from Santa Cruz Biotechnology. Seeded cells were cultured in 6-well plates for 24 h prior to transfection. Transfections were performed using Lipofectamine 2000 and PLUS reagent according to the manufacturer's instructions (Invitrogen).

Construction of lentiviruses
Mixed scramble shRNA or pLKO.1-Esr1 plasmids (Sigma) with packaging plasmid pCMV-dR8.2 (gift from Dr. Peter Tontonoz, UCLA) and envelope-expressing plasmid pMD2.G (Addgene, catalog no. 12259) were transfected into a 293T packaging cell line by calcium phosphate transfection. After 48 and 72 h of transfection, cell culture medium was collected and filtered through a 0.45-m filter.

ER␣ knockdown in Min6 ␤-cells
To achieve Esr1 knockdown, lentiviral particles (Sigma) carrying shRNA targeted to Esr1 or scramble shRNA were used to transduce Min6 cells.
After selecting positive transformants with puromycin (4 g/ml) for 4 weeks, the selected clones were expanded and analyzed for KD efficiency by qPCR and immunoblotting. The resulting cultures were then used for subsequent assays.

Insulin secretion and content
Min6 ␤-cells were seeded in 96-well plates and maintained in culture medium for 3 days. The cells were washed twice by glucose-free KRH buffer (129.4 mM NaCl, 3.7 mM KCl, 2.7 mM CaCl 2 , 1.3 mM KH 2 PO 4 , 1.3 mM MgSO 4 , 24.8 mM NaHCO 3 at pH 7.4) with 0.1% radioimmunoassay grade BSA and cultured for 1 h in KRH buffer with 1 mmol/liter glucose. After washing, cells were incubated with KRH buffer with 0, 2, 5, and 20 mM glucose for 1 h. Cell-free media were collected and diluted 10-fold with KRH buffer prior to assaying for insulin concentration using the ultra-sensitive mouse insulin ELISA kit (catalog no. 90082, Crystal Chem Inc.). To measure total insulin content, Min6 cells were cultured in 6-cm dishes for 3-5 days. The cells were trypsinized and washed twice with PBS and then divided into 2 equal aliquots. One aliquot was extracted with acid/ethanol (0.18 M HCl in 75% ethanol) overnight at Ϫ20°C, followed by centrifugation and supernatant collection. The insulin level in the supernatants was determined as indicated above. The other aliquot was lysed with RIPA buffer, and the protein concentration of the extract was determined as described above. The insulin content was expressed as the amount of insulin (g) per mg of protein.

Mitochondrial isolation
Mitochondrial isolation was performed using a kit for cultured cells (Thermo Fisher Scientific, Waltham, MA). Briefly, 2 ϫ 10 7 pelleted cells were vortexed at medium speed for 5 s in Reagent A and incubated on ice for 2 min. Next, cells were vortexed with 10 l of Reagent B for 5 s and incubated on ice for 5 min with intermittent vortexing at maximum speed every minute. After addition of Reagent C and mixing by inversion, cell homogenates were centrifuged at 700 ϫ g for 10 min at 4°C. Supernatants were further centrifuged at 12,000 ϫ g for 15 min at 4°C to separate cytosolic (supernatant) and mitochondrial (pellet) fractions. Washed mitochondrial pellets were lysed by RIPA buffer prior to immunoblotting.

ER␣ protects ␤-cells from apoptosis Mitochondrial DNA
Total DNA was extracted from cells using DNeasy blood and tissue kit (Qiagen, Valencia, CA). Expression levels of mtCO3 for mtDNA and 18S for nuclear DNA were assessed by realtime qPCR. The ratio of mtDNA (mtCO3) to nuDNA (18S) was used as an estimate for comparison of mtDNA content between the genotypes.

Immunoblot analyses
Mouse islets and Min6 ␤-cells were homogenized in RIPA lysis buffer containing complete EDTA-free protease (Roche Applied Science) and phosphatase inhibitor mixture 2 (Sigma). All lysates were centrifuged, clarified, and resolved by SDS-PAGE. Samples were transferred to polyvinylidene difluoride membranes and subsequently probed with the following antibodies for protein and phosphoprotein detection: ER␣ (Santa Cruz Biotechnology); Mfn1 (Neuromab, Davis, CA); Mfn 2, t-Ulk1 (Sigma); Mff (Abcam); LC3B (Novus, Saint Charles, MO); Fis1 (Genetex); OPA1 (BD Biosciences); Pink1 (Cayman Chemicals); DJ-1 (R&D Systems); p62 (Progen Biotechnik GmbH); pan-actin, Parkin, DRP1, phospho-AMPK␣ Thr-172, total AMPK, phospho-Ulk1 Ser-555 (catalog no. 5869), Atg3, Atg5, Atg7, Atg12, Phospho-eIF2␣ Ser-51 , and total eIF2␣ (Cell Signaling, Danvers, MA). ER␣ protein in Min6 ␤-cells was compared with a whole-cell lysate control sample of ovary to confirm protein molecular weight. The OMA1 antibody was a gift from Alexander M. van der Bliek, and the YME1L1 antibody was a gift from Carla Koehler. In several instances, membranes were sectioned, stripped (Restore Plus, Thermo Fisher Scientific), and re-probed with different primary antibodies so that the maximum amount of data could be obtained from a specific experiment. We have indicated in the figure legends where panel immunoblots from different figures were obtained from the same experiment and thus normalized to the same ␤-actin loading control. Densitometric analyses were performed using Quantity One image software (Bio-Rad).

Quantitative RT-PCR
Mouse islets and Min6 ␤-cells were homogenized using TRIzol reagent, and RNA was isolated and further cleaned using RNeasy columns (Qiagen, Valencia, CA) with DNase digestion. Cells were lysed using RLT buffer (Qiagen, Valencia, CA) with 2-mercaptoethanol. For RNA isolation from lysed cells, the RNeasy Plus kit was used as per the manufacturer's instructions. cDNA synthesis was performed using 1 g of RNA with iSuperscript reverse transcriptase (Bio-Rad). PCRs were prepared using iQ SYBR Green Supermix (Bio-Rad). All PCRs are performed in a MyiQ real-time detection system (Bio-Rad). Quantification of a given gene, expressed as relative mRNA level compared with control, was calculated after normalization to a standard housekeeping gene (Ppia or 18S). Separate control experiments were performed to ensure that the efficiencies of target and reference amplification were equal, as described in User Bulletin 2 from Applied Biosystems. Primer pairs were designed using Primer Express 2.0 software (Applied Biosystems, Waltham, MA) or obtained as previously published sequences. Primer sets were selected spanning at least one exon-exon junction when possible and were confirmed for specificity using BLAST (Basic Local Alignment Search Tool, NCBI) and melting curve analysis to ensure that a single product with its characteristic melting temperature was obtained.

TUNEL assay
Cell proliferation and apoptosis were quantitatively determined using the APC BrdU flow kit (Pharmingen) according to the manufacturer's instructions. Briefly, both scramble and Esr1-KD Min6 ␤-cells were pulsed with 10 l/ml of BrdU solution for 1.5 h at 37°C with 5% CO 2 . Trypsinized cells were stained with BrdU and 7-aminoactinomycin D (7ADD). Cells were resuspended in FACS buffer (PBS 3% BSA) with DAPI (25 g/ml) and analyzed immediately by flow cytometry on an LSRII (BD Biosciences) instrument with FlowJo software (Treestar Inc).

In situ cell death detection
Programmed cell death in Min6 ␤-cells was detected by in situ cell death detection kit (Roche Applied Science). Briefly, air-dried cells were fixed with freshly prepared 4% paraformaldehyde and permeabilized on ice with a solution containing 0.1% Triton X-100 and 0.1% sodium citrate. PBS-washed cells were incubated with Tunel reaction mix (1: 50 mixture of enzyme solution and label solution) at 37°C for 1 h. Negative control cells were incubated with label solution only. For positive controls, cells were treated with 3 units/ml DNase I in 50 mM Tris and 1 mg/ml BSA buffer before being incubated with the Tunel reaction mix. Cells were mounted with Vectashield mounting medium for fluorescence detection with DAPI (H-1200, Vector). Images were acquired using a Leica LMD 7000 fluorescence microscope.

Reactive oxygen species production
Min6 ␤-cells were washed and incubated in low-glucose DMEM at 37°C, 5% CO 2 in the dark with either 25 M 2Ј,7Јdichlorodihydrofluorescein diacetate (carboxy-H 2 DCF-DA) (Molecular Probes, Eugene, OR) or 5 M mitoSOX for 15 min (Molecular Probes, Eugene, OR), with or without 5 M CCCP for 30 min. Cells were washed with PBS and quickly trypsinized, pelleted, and retained on ice. Samples were resuspended in FACS buffer (PBS 3% BSA) with DAPI (25 g/ml) and analyzed immediately by flow cytometry on an LSRII (BD Biosciences) with FlowJo software (Treestar Inc). Unstained and singlestained cells were used for establishing compensation and gates, and only live cells were analyzed by the software as described previously (25).

Mitochondrial respiration
Mitochondrial respiration in cultured Min6 ␤-cells was measured using an XF24 extracellular flux analyzer (Seahorse Biosciences, North Billerica, MA). Briefly, cells were plated to confluence on a 24-well plate. Measurements of oxygen consumption were made continuously, and cells were sequentially treated with oligomycin (ATP synthase inhibitor), carbonyl cyanide p-trifluoromethoxyphenylhydrazone (an uncoupling agent), and rotenone/myxothiazol (inhibitors of complex I/III of the electron transport chain) (10).

Flow cytometry
To detect autophagy, both control scramble (Scr) and Esr1-KD Min6 ␤-cells were incubated with RFP-GFP-LC3B (Tandem-LC3B) adenovirus generated using the ViraPower adenoviral expression system (Invitrogen) according to the manufacturer's instructions. Briefly, the coding sequence for Tandem-LC3B with flanking BP sites was amplified from the Addgene plasmid no. 21074 using High Fidelity Pfx50 DNA polymerase (Invitrogen). Entry clones for tandem-LC3B were obtained by recombination of the purified DNA fragment with pDONR221 vector using Gateway BP clonase II. Tandem-LC3B insert was transferred from the entry clones into the pAd/ CMV/V5-DEST vector (Invitrogen) using the Gateway LR clonase II enzyme mix. Recombinant adenovirus-purified plasmid was used to generate adenoviral particles according to the manufacturer's instructions (Invitrogen). Adenoviral titer was determined using the AdenoX rapid titer kit (Clontech). Autophagy was induced by nutrient starvation in Hanks' balanced salt solution (Gibco) for 4 or 8 h followed with or without autophagy inhibition by 25 nM bafilomycin A 1 (Sigma) for an additional 4 h. Following treatment, cells were washed twice with PBS, fixed in 4% paraformaldehyde for 15 min at room temperature, and resuspended in FACS buffer for analysis using an LSRII (BD Biosciences) cytometer with FlowJo software (Treestar Inc). Untreated cells and cells infected with adenovirus carrying EGFP-LC3B and transfected with mRFP-LC3B plasmid were used for establishing compensation and gates.

Statistics
Values presented are expressed as means Ϯ S.E. Statistical analyses were performed using Student's t test when compar-ing two groups of samples or one-way analysis of variance (ANOVA) with Tukey's post hoc comparison for identification of significance within and between groups using SPSS and GraphPad Prism 5 (GraphPad Software, San Diego). Significance was set a priori at p Ͻ 0.05.