Metabolic Actions of Estrogen Receptor Beta (ERβ) are Mediated by a Negative Cross-Talk with PPARγ

Estrogen receptors (ER) are important regulators of metabolic diseases such as obesity and insulin resistance (IR). While ERα seems to have a protective role in such diseases, the function of ERβ is not clear. To characterize the metabolic function of ERβ, we investigated its molecular interaction with a master regulator of insulin signaling/glucose metabolism, the PPARγ, in vitro and in high-fat diet (HFD)-fed ERβ -/- mice (βERKO) mice. Our in vitro experiments showed that ERβ inhibits ligand-mediated PPARγ-transcriptional activity. That resulted in a blockade of PPARγ-induced adipocytic gene expression and in decreased adipogenesis. Overexpression of nuclear coactivators such as SRC1 and TIF2 prevented the ERβ-mediated inhibition of PPARγ activity. Consistent with the in vitro data, we observed increased PPARγ activity in gonadal fat from HFD-fed βERKO mice. In consonance with enhanced PPARγ activation, HFD-fed βERKO mice showed increased body weight gain and fat mass in the presence of improved insulin sensitivity. To directly demonstrate the role of PPARγ in HFD-fed βERKO mice, PPARγ signaling was disrupted by PPARγ antisense oligonucleotide (ASO). Blockade of adipose PPARγ by ASO reversed the phenotype of βERKO mice with an impairment of insulin sensitization and glucose tolerance. Finally, binding of SRC1 and TIF2 to the PPARγ-regulated adiponectin promoter was enhanced in gonadal fat from βERKO mice indicating that the absence of ERβ in adipose tissue results in exaggerated coactivator binding to a PPARγ target promoter. Collectively, our data provide the first evidence that ERβ-deficiency protects against diet-induced IR and glucose intolerance which involves an augmented PPARγ signaling in adipose tissue. Moreover, our data suggest that the coactivators SRC1 and TIF2 are involved in this interaction. Impairment of insulin and glucose metabolism by ERβ may have significant implications for our understanding of hormone receptor-dependent pathophysiology of metabolic diseases, and may be essential for the development of new ERβ-selective agonists.


Introduction
The estrogen receptors (ERs) are members of the nuclear hormone receptor family (NHR) which act as eukaryotic liganddependent transcription factors. ERs are involved in the regulation of embryonic development, homeostasis and reproduction. Two major estrogen receptors, alpha and beta (ERa and ERb), convey the physiological signaling of estrogens (17b-estradiol, E2) [1]. Additionally, ERs are activated by specific synthetic ligands such as raloxifene, tamoxifen, the ERb-specific ligand diarylpropionitrile (DPN), and the ERb-specific agonist propylpyrazole-triol (PPT), which belong to the group of selective estrogen receptor modulators (SERMS) [2][3][4].
The prevalence of metabolic diseases such as obesity, insulin resistance and type 2 diabetes has increased dramatically during the recent ten years [5]. Gender differences in the pathophysiology of obesity and metabolic disorders are well established [6][7][8].
However, the molecular mechanisms of sexual dimorphism in metabolic diseases are largely unknown. In addition, lack of ER activation has been implicated in postmenopausal impairment of glucose and lipid metabolism, resulting in visceral fat distribution, insulin resistance and increased cardiovascular risk after menopause [9]. In this context the investigation of ER-signaling and its role in metabolic disorders has gained increasing attention [4,8].
To identify the ER subtype involved in the regulation of metabolic disorders, studies have been carried out in ER-deficient mice. ERa-deficient (aERKO) mice have profound insulin resistance and impaired glucose tolerance [10][11][12][13]. These studies indicate that ERa has a protective role in metabolic disorders by improving insulin sensitivity and glucose tolerance. The metabolic function of ERb is not clear. ERb knockout mice (bERKO) have a similar body weight and equal fat distribution in comparison to wild type littermates. Additionally, bERKO and wild-type (wt) mice exhibit similar insulin and lipid levels [14]. However, previous studies in bERKO mice were only carried out under low fat diet, which may have concealed a phenotype relevant for human obesity normally induced by high-energy/fat diet.
The peroxisome proliferator-activated receptor gamma (PPARc) belongs to the NHR family and is a major regulator of glucose and lipid metabolism by modulating energy homeostasis in adipose tissue, skeletal muscle and liver [15][16][17]. Glitazones or thiazolidinediones (TZDs) are high-affinity PPARc agonists, and act as insulin sensitizers. TZDs induce adipogenesis and adipose tissue remodeling followed by an improvement of glucose tolerance [18]. The role of PPARc in the control of glucose homeostasis expands beyond its primary action in adipose tissue, and involves the regulation of adipocytokine production such as adiponectin, leptin, and resistin [19][20][21]. Consistently, reduced PPARc activity has important metabolic and cardiovascular pathophysiological consequences leading to insulin resistance, diabetes and end organ damage [15].
The molecular mechanisms underlying PPARc function are similar to those of ER-signaling. In a basal state, PPARc, similar to ERs, is bound to corepressor proteins such as nuclear receptor corepressor (NCoR) or silencing mediator of retinoic acid and thyroid hormone receptor (SMRT) [22]. After binding within the ligand binding domain (LBD), PPARc ligands induce its heterodimerization with retinoid x receptor alpha (RXRa), and its subsequent interaction with co-activators like steroid receptor coactivators (SRCs) followed by binding to PPARc response elements (PPREs) within target gene promoters [23]. Importantly, PPARc is sharing a similar pool of cofactors with ERb which provides a platform for mutual interactions between these two NHRs [23,24].
To study the crosstalk between ERb and PPARc, we investigated the regulation of PPARc-mediated transcriptional activity by ERb. Our in-vitro experiments in 3T3-L1 preadipocytes showed that ERb inhibits ligand-mediated PPARc-transcriptional activity. That resulted in the blockade of PPARcinduced adipocytic gene expression and in decreased adipogenesis. Overexpression of nuclear coactivators such as steroid receptor coactivator 1 (SRC1) and transcriptional intermediary factor 2 (TIF2) prevented the ERb-mediated inhibition of PPARc activity, whereas the presence of vitamin D receptor (VDR)-interacting protein 205 (DRIP205) or PPARc coactivator-1alpha (PGC1a) had no effect indicating a role for distinct nuclear coactivators for ERb-PPARc interaction in-vitro. High fat diet (HFD)-fed bERKO mice showed increased body weight and fat mass. In contrast, triglyceride content in liver and muscle was decreased in bERKO mice, which was associated with a marked improvement of hepatic and muscular insulin signaling. Compared to wt, bERKO mice demonstrated improved systemic insulin sensitivity and glucose tolerance. In consonance with the metabolic phenotype and with the in-vitro data, bERKO mice exhibited augmented PPARc signaling in adipose tissue corresponding to increased food efficiency and significantly elevated RQ (respiratory quotient). Blockade of adipose PPARc signaling in bERKO mice by PPARc antisense oligonucleotide injection resulted in a reversal of the bERKO phenotype including body weight reduction and impairment of insulin sensitivity.
In summary, the present data demonstrate that ERb impairs insulin and glucose metabolism which may, at least in part, result from a negative cross-talk with adipose PPARc.

ERb Inhibits PPARc Activity in a Ligand-Independent Manner
In order to demonstrate a molecular interaction between PPARc and ERb in a metabolically relevant cell system, we first investigated ligand-dependent PPARc activity in the presence of ERb in 3T3-L1 preadipocytes. Cells were treated with the PPARc-agonist pioglitazone (10 mM), with or without additional E2 stimulation, and PPARc activation was measured using pGal4-hPPARcDEF/pG5TkGL3 luciferase assay [25]. Upon pioglitazone stimulation, 3T3-L1 preadipocytes showed pronounced PPARc activation (bar 1+2, Figure 1A). This activation was not affected by co-treatment with ligands for ERb such as E2 (bar 2 vs. 3, Figure 1A) or DPN (data not shown). Overexpression of ERb led to a marked inhibition of ligand-dependent PPARc activity (bar 2 vs. 4+6+8, Figure 1A) which was also corroborated in a PPARc response element (PPRE) luciferase assay ( Figure S1). This inhibition was E2 (bar 4+6+8 vs. 5+7+9, Figure 1A) and DPN independent (data not shown). The inhibitory effect of ERb seemed to be isoform specific, since ERa overexpression resulted in no inhibition of PPARc activity (bar 11, Figure 1A and Figure  S2). To further explore the regulation of PPARc by ERb, we performed additional experiments coexpressing an activation function 1 domain (AF-1) deleted-ERb construct in 3T3-L1 cells. Overexpression of this truncated form of ERb which still contains a functional ligand binding domain (LBD) did not reduce PPARc activity indicating that ERb AF-1 is necessary for regulation of PPARc by ERb (bar 10, Figure 1A). To assure adequate overexpression and function of ERb in our system, 3T3-L1 preadipocytes were transiently transfected with ERb followed by Western blot analysis and transactivation assays using ER response elements (ERE)-luciferase system ( Figure 1B, C). Both assays confirmed adequate expression and function of ERb.

ERb Inhibits PPARc-Dependent Adipocyte Differentiation and Target Gene Expression
While our data implicated a negative regulation of ligandmediated PPARc transcription by ERb, we next investigated the regulation of PPARc-dependent gene expression during 3T3-L1 preadipocyte differentiation. The preadipocytes were transfected

Author Summary
In the present study, we demonstrate for the first time a pro-diabetogenic function of the ERb. Our experiments indicate that ERb impairs insulin sensitivity and glucose tolerance in mice challenged with a high fat diet (HFD). Loss of ERb, studied in ERb -/-mice (bERKO mice), results in increased body weight gain and fat deposition under HFDtreatment. Conversely, absence of ERb averted accumulation of triglycerides and preserved regular insulin signaling in liver and skeletal muscle. This observation was associated with improved whole-body insulin sensitivity and glucose tolerance. Increased adipose tissue mass in the presence of improved insulin sensitivity and glucose tolerance is usually observed under chronic stimulation of the nuclear hormone receptor PPARc. In consonance, we show that activation of PPARc was markedly induced in gonadal fat from bERKO mice and blockade of adipose PPARc signaling by antisense oligonucleotide injection reversed the metabolic phenotype. Moreover, our cell culture experiments indicate that ERb is a negative regulator of ligand-induced PPARc activity in vitro. Finally, we identify SRC1 and TIF2 as key players in the ERb-PPARc interaction. In summary, the present study demonstrates that ERb impairs insulin and glucose metabolism, which may, at least in part, result from a negative cross-talk with adipose PPARc.
with indicated plasmids and differentiated for 3 days using standard differentiation medium [25]. As the full differentiation procedure requires 7-10 days of treatment, the observed effect on fat droplet accumulation and expression pattern are typical for early phase of adipocyte differentiation. The 3T3-L1 cells transfected with ERb and differentiated for 3 days showed reduced adipogenesis visualized by fat droplet accumulation in comparison to control cells (Figure 2A). Low levels of ERb could also be detected in untransfected 3T3-L1 cells and its expression was slightly elevated during differentiation (data not shown) underlining the physiological importance of our findings. Overexpression of the ERa isoform in these cells did not show any inhibitory effect on preadipocyte differentiation (Figure 2A). The adipocyte protein 2 (aP2) gene belongs to the classical PPARcregulated genes involved in the early phase of adipogenesis [26]. The expression level of aP2 measured by real-time PCR was significantly elevated in the differentiated control cells (bar 2 vs. 1, Figure 2B). Overexpression of ERb-but not ERa-in these cells led to a significant reduction of aP2 expression (bar 2 vs. 4 and 6, Figure 2B) indicating that endogenous PPARc activation in 3T3-L1 cells was inhibited by ERb.
Furthermore pioglitazone (10 mM) treatment of 3T3-L1 cells overexpressing PPARc/RXRa showed increased adipogenesis, an effect that was markedly inhibited by coexpression of ERb ( Figure 2C). aP2 expression level was also significantly reduced in cells co-expressing ERb together with PPARc/RXRa (bar 2 vs. 3, Figure 2D). These data indicate that ERb inhibits PPARctranscriptional activity resulting in the blockade of PPARcinduced adipocytic target gene expression and amelioration of adipogenesis.

PPARc Target Gene Expression and PPARc Activity Are Increased in bERKO Mice
To investigate ERb's action on PPARc in vivo, we studied PPARc activity and PPARc target genes in HFD-fed bERKO and wt mice. bERKO mice and their wt littermates were fed HFD containing 60% calories from fat for 12 weeks followed by the analysis of PPARc-dependent gene expression in gonadal fat tissue. Adipose mRNA expression of PPARc target genes involved in triglycerides (TG) synthesis such as lipoprotein lipase (Lpl), phosphoenolpyruvate carboxykinase (PEPCK) and CD36 was significantly upregulated in bERKO mice ( Figure 3 A-C). Key mediators of insulin and glucose metabolism such as the retinolbinding protein 4 (RBP4) were also regulated in bERKO mice ( Figure 3D). Consistently with these findings, adiponectin mRNA expression and adiponectin serum levels were elevated in bERKO mice ( Figure 3E, F). No difference of PPARc target gene regulation between bERKO and wt mice was observed in liver (data not shown).
Positive regulation of a series of adipose PPARc target genes in bERKO mice suggested a general induction of PPARc transcription in bERKO mice. To prove this, we performed EMSA assays in gonadal fat from bERKO and wt mice after 12 weeks on HFD. Nuclear fractions isolated from adipose tissues from bERKO mice showed an increased binding/activation of endogenous PPARc in comparison to wt mice (line 4-7 vs. 1-3, Figure 3G) in the presence of similar PPARc expression levels, as shown by real-time RT-PCR analysis and Western Blot ( Figure 3G). Increased adipose PPARc target gene expression and PPARc-DNA binding confirmed an augmented PPARc signaling in adipose tissue from bERKO mice.

bERKO Mice Exhibit Enhanced PPARc Signaling under Pioglitazone Treatment
To exclude the possibility that the augmented expression of PPARc target genes measured in HFD-fed bERKO is the result of increased adipose tissue mass, we performed experiments using exvivo fat pads isolated from wt and bERKO mice, treated for 24h with 10 mM pioglitazone or vehicle-control, followed by analysis of PPARc target gene expression using real-time RT-PCR. In this system augmented ligand-induced PPARc target gene expression mainly results from enhanced PPARc transcriptional activity and Figure 1. ERb inhibits PPARc activity in a ligand-independent manner. A) 3T3-L1 preadipocytes were transfected with the indicated plasmids together with pGal4-hPPARcDEF, pG5TkGL3 and renilla, followed by treatment with 10 mM pioglitazone, 100 nM E2, or in combination as indicated. # p,0.05 vs. pSG5+veh; * p,0.05 vs. pSG5+Pio, ns: not significant vs. pSG5+Pio, ns a : not significant vs. ERb+Pio. B) 3T3-L1 preadipocytes were transfected with ERb (as indicated), and protein level of ERb was analysed by Western blot. C) 3T3-L1 preadipocytes were transfected with the indicated plasmids together with pERE-TkGL3, and cells were treated with 100 nM E2, as indicated. * p,0.05 vs. pSG5+E2; ns: not significant vs. pSG5+veh. doi:10.1371/journal.pgen.1000108.g001

ERb-PPARc Interaction In Vivo Is Ligand Independent
To further characterize ERb ligand dependency for its interaction with PPARc in the mouse model, additional in-vivo studies were performed in estrogen-depleted, ovariectomized wt mice treated with the ERb-ligand DPN. Analysis of PPARc target genes (Lpl, PEPCK, CD36 and adiponectin) in gonadal fat isolated from these mice revealed no significant differences in the expression level between vehicle and DPN-treated rodents indicating ligand independency ( Figure 4C). These data are consistent with the in-vitro study in 3T3-L1 preadipocytes, where PPARc activation was not affected by co-treatment with ligands for ERb such as E2 (bar 2 vs. 3, Figure 1A) or DPN (data not shown).

bERKO Mice Exhibit Improved Hepatic and Muscular Insulin Signaling
Given the central role of PPARc in insulin and glucose metabolism, the metabolic phenotype of bERKO mice was assessed. No difference in fasting/fed blood glucose food intake, and mean arterial blood pressure was observed between bERKO and wt mice under HFD (Table 2). Body weight gain was significantly enhanced in bERKO mice, compared to wt mice (mean BW difference bERKO vs. wt mice after 12 week HFD: 3+/20.4 g, p,0.05, Figure 5A). Increased body weight in bERKO mice resulted from increased adipose tissue mass. MRIanalysis of body composition demonstrated significantly higher fat . bERKO mice exhibit enhanced PPARc signaling under pioglitazone treatment. A+B) Explanted gonadal fat pads isolated from wt-and bERKO mice were treated for 24h with 10 mM pioglitazone or vehicle control. Real-time quantitative RT-PCR studies on Lpl and PEPCK expression were carried out using total RNA (n = 4 per group), as indicated. For details, see Materials and Methods and supplemental data. *p,0.05 vs. wt+veh; # p,0.05 vs. wt+Pio. C) ERb-PPARc interaction in vivo is ligand independent. Analysis of Lpl, PEPCK, CD36, and adiponectin mRNA expression levels in gonadal fat from soy-free-fed and ovariectomized wt female mice, treated for 21 days with DPN (8 mg/Kg) or vehicle control (n = 4/group). Real-time quantitative RT-PCR studies were carried out using total RNA prepared from gonadal fat. For details, see Materials and Methods and supplemental data. ns: not significant vs. vehicle-treated mice. doi:10.1371/journal.pgen.1000108.g004 mass in bERKO mice compared to wt littermates ( Figure 5B), and fat pad weight from gonadal and perirenal depots was increased ( Table 1). In contrast, liver weight was significantly reduced in bERKO mice in comparison to wt control littermates (Table 1). Reduced hepatic weight likely resulted from decreased TGaccumulation assessed by H/E-staining of liver tissue sections ( Figure 5C), and by TG quantification in dried liver tissue ( Figure 5D). In accordance with reduced hepatic TG-content, hepatic insulin signaling was improved. After injection of insulin in the portal vein, liver tissue was dissected and proteins were isolated for Western blot analysis. Insulin-stimulated Akt phosphorylation was enhanced in bERKO mice ( Figure 5E and Figure S3). In parallel to decreased TG levels in liver, bERKO mice had decreased muscular TG-accumulation under HFD and improved insulin signaling ( Figure 5F, G, and Figure S3). Skeletal muscle and liver are the major insulin responsive tissues, and important sites of glucose metabolism in-vivo. An important mechanism of PPARc-mediated insulin sensitization involves adipose tissue remodeling and trapping of circulating triglycerides (TG) which protects the liver and skeletal muscle against TG overload. Increased adipose tissue mass in bERKO mice may protect these animals against TG-overload in liver and skeletal muscle resulting in an improvement of hepatic and muscular insulin sensitivity.

Systemic Insulin Sensitivity and Glucose Tolerance Are Improved in bERKO Mice
Next we investigated insulin and glucose metabolism in bERKO and wt mice. Whole body glucose disposal was assessed using an oral glucose tolerance test (OGTT) ( Figure 6A). Following an oral glucose challenge bERKO mice on HFD had moderately but significantly improved glucose tolerance compared to HFD-fed wt mice ( Figure 6A, B). In addition insulin sensitivity measured by an insulin tolerance test (ITT) was improved in comparison to wt mice ( Figure 6C, D). No difference in fasting and fed blood glucose was observed between bERKO and wt mice under HFD (Table 2). Despite an increased fat mass in bERKO mice, systemic insulin sensitivity and glucose tolerance were significantly improved under HFD when compared to wt-control. To further examine the enhanced weight gain and fat deposition in bERKO mice, we performed indirect calorimetry and monitored food consumption. Food intake did not differ between wt-control and bERKO mice (Table 2). However, deletion of ERb resulted in a marked increase of food efficiency (ratio of weight gain and food intake, Figure 6E). No significant difference in O 2 consumption ( Figure 6F), energy expenditure (Table 2), or locomotor activity ( Table 2) was detected between bERKO and wt mice. Low RQ values have previously been described for rodents under HFD and in diabetes [27]. Both wt and bERKO mice exhibited low RQ values. bERKO mice had a significantly higher RQ when compared to wt-controls which may be indicative for attenuated fatty acid (FA) oxidation promoting fat accumulation ( Figure 6G). These data show that bERKO mice are partially protected against HFD induced insulin resistance. Increased fat mass may likely result from increased food efficiency based on reduced oxidative utilization of fat and increased fat storage. The metabolic phenotype of bERKO mice including increased fat mass, reduced hepatic/muscular TG and improved systemic insulin sensitivity exhibits high similarity to augmented PPARc activation e.g. under thiazolidinedione (TZD) treatment [28,29].

Disruption of PPARc Signaling by Antisense Oligonucleotide Injection Reversed the Metabolic Phenotype of bERKO Mice
To directly demonstrate the role of PPARc in HFD-fed bERKO mice, PPARc signaling was disrupted by intraperitoneal (i.p.) injection of PPARc antisense oligonucleotide (ASO). HFDfed bERKO mice were injected twice a week for 6 weeks with either PPARc ASO or control oligonucleotides. PPARc expression was significantly reduced in liver of ASO-treated bERKO mice, similar to previously reported results in apoB/BATless mice (data not shown) [30]. However, suppression of hepatic PPARc by ASO injection is unlikely to play an important role in our model, since hepatic PPARc signaling did not differ between wt and bERKO mice, respectively. More importantly, i.p. application of PPARc ASO in bERKO mice resulted in 6364.8% (p,0.05) reduction of PPARc expression in gonadal adipose tissue compared to bERKO mice injected with control oligonucleotides ( Figure 7A). Accordingly, expression of the PPARc target genes Lpl, PEPCK, CD36, and adiponectin was markedly decreased in adipose tissue from PPARc ASO-injected bERKO mice, and adipocyte diameters were increased ( Figure 7A, G). These data corroborate a relevant reduction of adipose PPARc signaling by ASO intervention. Body weight gain and gonadal fat accumulation in HFD-fed-bERKO mice were significantly attenuated by PPARc-ASO injection ( Figure 7B, C). Finally, blockade of adipose PPARc by ASO led to reversal of the improved insulin response observed in bERKO mice, and to an impairment of insulin sensitivity and glucose tolerance ( Figure 7D-F). Together these data underline the importance of adipose PPARc signaling for the metabolic phenotype observed in bERKO mice.

ERb-Mediated Inhibition of PPARc Activity Involves SRC1 and TIF 2
Nuclear coactivators such as SRC1 and TIF2 are important mediators of ERb and PPARc-induced transcriptional activation. It has previously been shown that competition of distinct nuclear receptor (NR) for coactivator binding results in a negative crosstalk between NRs [31]. To prove whether common coactivators are involved in ERb-PPARc interactions, SRC1, TIF2, DRIP205 or PGC1a were co-expressed together with ERb and ligand induced PPARc activation was measured.   Figure 1G) and average food intake/ day (Table 2). Data are presented as x-fold over wt mice. ** p,0.01 vs. wt-control. F+G) O 2 consumption (VO 2 ), and respiratory quotient (RQ) from HFD-fed wt and bERKO mice. RQ was calculated as the ratio between CO 2 produced (VCO 2 ) and O 2 consumed (VO 2 ) using the calorimetry system described under methods. * p,0.05 vs. wt-control. doi:10.1371/journal.pgen.1000108.g006 Overexpression of SRC1 and TIF2 prevented the ERbmediated inhibition of PPARc activity ( Figure 8A, B) whereas the presence of DRIP205 ( Figure 8C) and PGC1a ( Figure S4) had no effect. To demonstrate that SRC1 and TIF2 are also involved in ERb-PPARc interaction in-vivo, we performed ChIP experi-ments with gonadal fat from HFD-fed bERKO and wt mice. The adiponectin promoter was selected as a PPARc-target promoter. Binding of SRC1 and TIF2 to the adiponectin promoter was enhanced in gonadal fat from bERKO mice ( Figure 8D), indicating that the absence of ERb in adipose tissue results in Figure 8. ERb inhibits PPARc activity in a ligand-independent manner involving SRC1+TIF2. A-C) 3T3-L1 preadipocytes were transfected with the indicated plasmids together with pGal4-hPPARcDEF, pG5TkGL3 and renilla followed by treatment with 10 mM pioglitazone as indicated. *p,0.05 vs. pSG5+veh; # p,0,05 vs. pSG5+Pio,^p,0.05 vs. ERb+Pio, ns: not significant vs. ERb+Pio. Values represent means6SEM of at least two independent experiments performed in triplicates. D) ChIP experiment with gonadal fat from HFD-fed wt mice and bERKO mice. IP was performed using Flag, RNA Pol II, SRC1 and TIF2 as indicated. (-no template control (NTC), + genomic DNA, input: 1% of the initial probe taken for IP). For details please see Materials and Methods. doi:10.1371/journal.pgen.1000108.g008 exaggerated coactivator binding to a PPARc target promoter. Together these data suggest that the coactivators SRC1 and TIF2 are involved in the negative regulation of PPARc by ERb in vitro and in vivo.

Discussion
The present study demonstrates that ERb is a negative regulator of ligand-induced PPARc activity in-vitro. Consequently, data from bERKO mice suggest that ERb negatively regulates insulin and glucose metabolism which may, at least in part, result from an impairment of regular adipose tissue function based on a negative cross-talk between ERb and PPARc. Loss of ERb resulted in enhanced body weight gain and fat accumulation in HFD-fed mice. However, absence of ERb prevented hepatic/ muscular triglyceride overload, preserved regular insulin signaling in liver/ skeletal muscle, and improved whole-body insulin sensitivity and glucose tolerance under HFD. This metabolic phenotype strongly suggested augmented PPARc signaling in mice lacking ERb. And indeed, PPARc target genes and PPARc-DNA binding were markedly induced in gonadal fat from bERKO mice. Along this line, blockade of adipose PPARc signaling by PPARc ASO injection reversed the metabolic changes in bERKO mice.
A mutual signaling cross-talk between ERs and PPARc has been described previously. PPARc together with its heterodimeric partner RXRa has been shown to suppress ER-induced target gene expression through competitive binding to an ERE site in the vitellogenin A2 promoter [32]. In accordance with a bidirectional interaction, Wang and colleagues demonstrated that ERs are capable of inhibiting ligand-induced PPARc activation in two different breast cancer cell lines [33]. In contrast to our results, these authors show that basal and agonist-stimulated PPREactivity is also blocked by ERa. Transcriptional activity of PPARc differs markedly depending on the cell system and tissues. The highest level of PPARc-mediated transcription has been described in adipocytes and adipocytic cell lines, where molecular conditions such as cofactor availability seemed to be optimized [34]. Compared to adipocytes, breast cancer cells exhibit low PPARc expression and activity reflected by a less than 2-fold induction of PPRE-activity after ligand stimulation [33]. The presence of PPARc suppression by ERa in breast cancer cells might be a result of weak basal PPARc transcriptional activity in these cells. In contrast, the pronounced activation of the exogenous PPARc LBD in 3T3-L1 preadipocytes may require more potent inhibitory stimuli which could not be achieved by ERa overexpression in our system.
Suppression of PPARc-LBD activation by ERb did not depend on ERb ligands which is consistent with previous reports [33]. Also our in vivo studies in estrogen-depleted, ovariectomized wt mice treated with the ERb-ligand DPN indicate that PPARc-ERb interaction is ligand independent. More importantly, overexpression of a truncated form of ERb containing solely the ERb-LBD/ AF2 domain did not induce any inhibitory effect on PPARc suggesting an important role of ERb's NH 2 -terminal AF1 domain for ERb-PPARc interactions. Consistently, activity of the ER-AF1 domain is usually not dependent on ligand activation [35]. Furthermore, Tremblay and coworkers demonstrated that ERb-AF1 activation involves ligand-independent recruitment of SRC-1, a cofactor involved in ERb-PPARc interactions in our study [36]. These data corroborate our observation that PPARc suppression by ERb involves the AF1 domain and ligand-independent interactions with the coactivators SRC1 and TIF2. Repression of PPARc activity through ERb was reversed by titration of the p160 coactivators, SRC1 and TIF2, suggesting that the suppres-sive action of ERb is a result of p160 coactivator interaction with ERb thereby preventing the binding of PPARc to the same coactivators. Similar interactions have been described previously for ER interaction with the thyroid receptor [31].
The present study demonstrates for the first time that ERb impairs insulin sensitivity and glucose tolerance under HFD implicating pro-diabetogenic actions of this receptor. In consonance, we could recently demonstrate that ERb has a suppressive role on glucose transporter 4 (GLUT4) expression in skeletal muscle [8,37]. GLUT4 has been identified as the major mediator of insulin-induced glucose uptake in fat and skeletal muscle. In addition, removal of the E2-ERb signaling by ovariectomy in ERa-deficient mice improved glucose and insulin metabolism supporting the diabetogenic effect of ERb [12]. Loss of ERb resulted in a marked augmentation of adipose PPARc activity in our model indicating that ERb mediates its metabolic actions by a negative interaction with PPARc in adipose tissue. This concept is corroborated by a number of observations. HFD-fed bERKO mice exhibited increased adipose tissue mass in the presence of improved insulin sensitivity and glucose tolerance. These metabolic changes are usually observed under chronic PPARc stimulation [17]. PPARc has been identified as an essential regulator of whole-body insulin sensitivity. Two major mechanisms have been described: (1) Adipose PPARc protects nonadipose tissue against excessive lipid overload and maintains normal organ function and insulin responses (liver, skeletal muscle) by preserving regular adipose tissue function, and (2) Adipose PPARc guarantees a balanced and adequate production of adipocytokine secretion such as adiponectin from adipose tissue, factors which are important mediators of insulin action in peripheral tissues [38][39][40]. Both processes could be observed in bERKO mice. Further support of this notion comes from clinical actions of anti-diabetic PPARc agonists (TZD) [28,29]. Activation of PPARc by TZDs in diabetic patients resembles the phenotype of bERKO mice including improved insulin sensitization and glucose tolerance in the presence of weight gain. We also observed increased food efficiency and changes in nutrient partitioning reflected by an increased RQ in bERKO mice. Loss of ERb appears to result in attenuated fatty acid (FA) oxidation which may favor the storage of TGs in adipose tissue and increased fat accumulation, and may provide a possible explanation for the enhanced weight gain. Interestingly, treatment of obese mice with a synthetic PPARc agonist has been shown to mediate similar changes including an increase in food efficiency and higher RQ values [41]. Finally, blockade of PPARc signaling in adipose tissue of bERKO mice resulted in a reversal of the metabolic phenotype corroborating the importance of adipose PPARc in the present model. The observed suppression of hepatic PPARc activity by ASO injection is unlikely to play a major role since the initial metabolic characterization of untreated bERKO mice under HFD did not reveal any dysregulation of hepatic PPARc signaling. In summary, the metabolic phenotype of bERKO mice is mediated by an augmented adipose PPARc action, which implies that in the presence of ERb, PPARc activity might be partially suppressed.
The notion, that ERb-PPARc crosstalk requires receptor-p160 interaction, was underlined by our observations in WAT from bERKO mice. Binding of SRC1 and TIF2 to the PPARcregulated adiponectin promoter in WAT was enhanced in the absence of ERb. It has recently been demonstrated that p160 coactivators are important regulators of PPARc transcriptional activity in WAT [42]. In particular, TIF2 has been identified as a nuclear coactivator involved in the adipogenic actions of PPARc. Future experiments are required to define the functional relevance of TIF2 and SRC1 in our model. So far one may conclude that the metabolic phenotype of HFD-fed bERKO mice is, at least in part, explained by increased adipose PPARc activity as a result of exaggerated binding of p160 coactivators to PPARc-regulated target gene promoters. Diabetogenic actions of ERb are of major significance for the pharmaceutical development of new ERbselective agonists intended for use against a multitude of diseases such as rheumatoid arthritis or postmenopausal osteoporosis [43,44]. Despite the high tissue selectivity of such compounds, and despite the fact that the actions observed in our study were ligand-independent, one has to be aware of the potentially deleterious actions of ERb on insulin-and glucose metabolism. As a precautionary measure metabolic profiling of new ERb agonist should be performed.
Collectively, our data provide first evidence that ERb negatively regulates insulin signaling and glucose metabolism that involves an impairment of regular adipose PPARc function. Moreover our data suggest that the coactivators SRC1 and TIF2 are involved in this inhibition. In consonance, impairment of insulin and glucose metabolism by ERb has significant implications for our understanding of hormone receptor-dependent pathophysiology of metabolic diseases, and is essential for the development of new ERb-selective agonists.

Animal Care and Treatment
Female estrogen receptor b -/-mice (bERKO) received from J.-A. Gustafsson (Karolinska Institutet, Huddinge, Sweden) and their wt littermates were housed in a temperature controlled (25uC) facility with a 12-h light/dark cycle and genotyped using genomic DNA isolation kit (Invitek) and PCR primers described elsewhere [45]. 4-5 week old mice were fed ad libitum with a high-fat diet (60% kcal from fat, [25]) for 12 weeks. Body weight and food intake were determined throughout the experiment. At start and end of treatment, body composition was determined by nuclear magnetic resonance imaging (Bruker's Minispec MQ10). After 12 weeks' treatment, blood samples were collected from overnight-fasted animals by retroorbital venous puncture under isoflurane anesthesia for analysis of serum adiponectin (mouseadiponectin ELISA; Linco Research) and glucose (colorimetric glucose test; Cypress Diagnostics). An OGTT using a dose of 2 g/kg body weight (BW) glucose and ITT with intraperitoneally injected 0.5 units/kg BW insulin (Actrapid; Novo Nordisk) were performed. Tail vein blood was used for glucose quantification with a glucometer (Precision Xtra; Abbott). Blood pressure was measured invasively in the abdominal aorta using a solid-state pressure transducer catheter (Micro-Tip 3F; Millar Instruments) under isoflurane anesthesia. Afterwards animals were killed and organs were dissected. For immunohistochemical studies organs were fixed in 4% formalin, embedded in paraffin and stained with Haematoxylin/Eosin (H&E); for RNA, Western blot analysis and measurement of TG content isolated organs were frozen in liquid nitrogen; for EMSA and Chromatin IP assays abdominal fat was stored in ice-cold PBS with proteinase inhibitors (Complete Mini, Roche), and immediately proceeded as described below.
For DPN-treatment, 10 week old female C57BL/6J mice were ovariectomized, and after 1 week recovery set on soy-free diet. Subsequently mice were treated for 21 days with DPN (8 mg/kg) or vehicle administered using subcutaneous pellets (Innovative Research of America). Afterwards animals were killed under isoflurane anesthesia and organs were dissected.
All animal procedures were in accordance with institutional guidelines and were approved.
Energy Expenditure, Locomotor Activity, and RQ After HFD feeding, bERKO mice and their wt littermates were analyzed for energy expenditure, RQ, and locomotor activity using a custom-made 4-cage calorimetry system (LabMaster; TSE Systems). The instrument consists of a combination of highly sensitive feeding and drinking sensors for automated online measurement. The calorimetry system is an open-circuit system that determines O 2 consumption, CO 2 production, and RQ. A photobeam-based activity monitoring system detects and records every ambulatory movement, including rearing and climbing movements, in every cage. All the parameters can be measured continuously. Mice (n = 7 per group) were placed in the calorimetry system cages for 24h.

Explanted Gonadal Fat Pads Experiments
Tissue samples from gonadal fat were prepared from female wt and bERKO mice. Explanted gonadal fat samples were washed 3 times with ice-cold Hanks Balanced Salt Solution (HBSS) and treated for 24h with 10 mM pioglitazone or vehicle in Dulbecco's modified Eagle's medium F2 (DMEM:F12, Invitrogen). Afterwards tissue samples were washed with ice-cold PBS and RNA extraction was performed using trizol (Invitrogen).

Cell Culture and Differentiation
3T3-L1 preadipocytes were purchased from the American Type Culture Collection. Preadipocytes were cultured in Dulbecco's modified Eagle's medium with 10% Fetal Bovine Serum (FBS) and 1% Pen-Strep (Invitrogen). For differentiation experiments preadipocytes were grown to confluence and after 12h culture medium was supplemented with methylisobutylxanthine (0.5 mM), dexamethasone (0.25 mM), and insulin (1 mg/ml) in DMEM containing 10% FBS for 72h [25]. Afterwards cells were washed with ice-cold PBS and RNA extraction was performed using trizol (Invitrogen) according to the manufacturer's instructions. For the staining procedure differentiated cells were washed twice with ice-cold PBS, fixed with 4% PFA, and stained for 1h at room temperature with Oil-red-O solution.

RNA and Protein Analysis
Total RNA from cultured preadipocytes, abdominal fat tissue and skeletal muscle was isolated using trizol (Invitrogen) according to the manufacturer's instructions. For real-time PCR analysis RNA samples were DNAse digested (Invitrogen), reverse transcribed using Superscript (Promega), RNasin (Promega), dNTPs (Invitrogen), according to the manufacturer's instructions, and used in quantitative PCR reactions in the presence of a fluorescent dye (Sybrgreen, BioRad). Relative abundance of mRNA was calculated after normalization to 18S ribosomal RNA. Primer sequences are provided in Table S1. For Western blot detection of ERb cells were grown on W10 cm plates and transfected with increasing amount of ERb plasmid or empty vector control. After 24h cells were harvested and WB analysis was performed as following: cells (and tissues for Akt analysis) were lysed in RIPA buffer (50 mM Tris pH 7.5, 150 mM NaCl, 5 mM MgCl 2 , 1% Nonidet P-40, 2.5% glycerol, 1 mM EGTA, 50 mM NaF, 1 mM Na 3 VO 4 , 10 mM Na 4 P 2 O 7 , 100 mM phenylmethylsulfonyl fluoride with proteinase inhibitors (Complete Mini, Roche). Lysates (tissues (30 mg) and cells (20 mg)) were analyzed by immunoblotting using antibody raised against ERb (H-150, Santa Cruz), antibody raised against pS473-Akt and total-Akt (Cell Signalling), and secondary horseradish-conjugated antibodies (Amersham). For PPARc immunoblotting, 20 mg of nuclear fractions used for EMSA were analyzed using antibody raised against PPARc (E-8, Santa Cruz). For detection, enhanced chemiluminescent substrate kit (Amersham) was used.

EMSA
Nuclear extracts were prepared by using a nonionic detergent method as described previously [46]. The inputs were normalized for protein contents, as ERb-deficient mice have increased fat tissue mass. Detection of PPARc was performed with a [ 32 P] cATP-labeled PPRE oligo (59-CAAAACTAGGTCAAAGGTCA-39 59-TGACCTTTGACCTAGTTTTG-39). The DNA binding reactions were performed with 40 ml of binding buffer (20 mg nuclear extracts, 2 mg of poly(dI-dC), 1 mg of bovine serum albumin (BSA), 5 mM dithiothreitol (DTT), 20 mM HEPES, pH 8.4, 60 mM KCl, and 10% glycerol) for 30 min at 37uC. For competition experiments, a cold oligonucleotide probe was used. The reaction products were analyzed via 5% polyacrylamide gel electrophoresis using 12.5 mM Tris, 12.5 mM boric acid, and 0.25 mM EDTA, pH 8.3. Gels were dried and exposed to Amersham TM film (Amersham Pharmacia Biotech) at 280uC using an intensifying screen.

Chromatin IP
Abdominal fat tissue (gonadal fat) isolated from wt and bERKO mice was washed in ice-cold PBS with proteinase inhibitors (Complete Mini, Roche), cut into small pieces, and incubated for 12h in 1% formaldehyde, PBS and proteinase inhibitors (Complete Mini, Roche) with rotation at 4uC. Formaldehyde was removed by intensive washing in ice-cold PBS and centrifugation. Samples were lysed in RIPA (with proteinase inhibitors, Complete Mini, Roche), sonicated on ice (Sonopuls HD 2070, 4 times 10s, 100%), and centrifuged. Samples from each group were pooled and protein content of clear phase lysates was measured using a Bradford assay (Amersham). For each immunoprecipitation (IP) 1.5 mg of protein was taken. The volume of the samples was kept constant by using dilution buffer (prepared according to Upstate protocol). For preclearance 90 ml of Protein A Sepharose slurry (Amersham) was added, and the samples were rotated for 1h in 4uC. After centrifugation beads were discarded, and 1% of supernatant volume per aliquot was used as an input control. The residual volume was incubated with 6 mg of appropriate antibodies (anti-Pol II (C-18, Santa Cruz), anti-Flag (Sigma), anti-SRC1 (M-20, Santa Cruz), anti-TIF2 (C-20, Santa Cruz)). The antibody-bound proteins were then precipitated using 300 ml Protein A Sepharose slurry (Amersham), washed and further processed according to the Upstate protocol.

Quantification of Hepatic/Muscular Triglycerides
Triglyceride-content in skeletal muscle and liver was measured as described previously [47]. Briefly, tissues were homogenized in liquid nitrogen and treated with ice-cold chloroform/methanol/ water mixture (2:1:0.8) for 2 min. After centrifugation the aqueous layer was removed and the chloroform layer was decanted. The mixture was incubated at 70uC for chloroform clearance, and the residues were dissolved in isopropanol, and assessed for the triglyceride content using an enzymatic-calorimetric test (Cypress diagnostics) according to the manufacturer's instructions.

Statistical Analysis
Results from real-time PCR of cell lines, transfections, and animal experiments were analyzed by ANOVA followed by multiple comparison testing or with paired/unpaired t tests, as appropriate. Data are expressed as mean6SEM or as indicated. Results were considered to be statistically significant at p,0.05. Figure S1 ERb inhibits PPARc activity in vitro. In order to demonstrate a molecular interaction between PPARc and ERb in a metabolically relevant cell system, we first investigated liganddependent PPARc activity in the presence of b in 3T3-L1 preadipocytes. Cells were transfected with 100 ng of PPARc, 50 ng of RXRa, 700 ng of PPRE-luc, 5 ng of renilla, and increasing amount of ERb, as indicated. Afterwards cells were treated with the PPARc-agonist pioglitazone (10 mM), and PPARc activation was measured using PPRE-luc luciferase assay. Upon pioglitazone stimulation, 3T3-L1 preadipocytes showed increased PPARc activation. Overexpression of ERb led to a marked inhibition of ligand-dependent PPARc activity (bar 1 vs. 2 and 3).  Figure S4 PGC1a overexpression does not affect ERb-mediated PPARc repression. 3T3-L1 preadipocytes were transfected with the PGC1a plasmids together with pGal4-hPPARcDEF, pG5TkGL3 and renilla and 500 ng ERb followed by treatment with 10 mM pioglitazone as indicated; *p,0.05 vs. pSG5+veh; # p,0,05 vs. pSG5+Pio. Found at: doi:10.1371/journal.pgen.1000108.s004 (0.26 MB TIF)