Estrogen Signalling and the Metabolic Syndrome: Targeting the Hepatic Estrogen Receptor Alpha Action

An increasing body of evidence now links estrogenic signalling with the metabolic syndrome (MS). Despite the beneficial estrogenic effects in reversing some of the MS symptoms, the underlying mechanisms remain largely undiscovered. We have previously shown that total estrogen receptor alpha (ERα) knockout (KO) mice exhibit hepatic insulin resistance. To determine whether liver-selective ablation of ERα recapitulates metabolic phenotypes of ERKO mice we generated a liver-selective ERαKO mouse model, LERKO. We demonstrate that LERKO mice have efficient reduction of ERα selectively within the liver. However, LERKO and wild type control mice do not differ in body weight, and have a comparable hormone profile as well as insulin and glucose response, even when challenged with a high fat diet. Furthermore, LERKO mice display very minor changes in their hepatic transcript profile. Collectively, our findings indicate that hepatic ERα action may not be the responsible factor for the previously identified hepatic insulin resistance in ERαKO mice.


Introduction
The metabolic syndrome (MS) refers to a group of interrelated metabolic abnormalities including insulin resistance, increased body weight and abdominal fat accumulation, mild dyslipidemia and hypertension [1,2,3,4]. Individuals with MS are at increased risk of cardiovascular disease and Type 2 diabetes (T2D) [5]. MS prevalence is on the rise worldwide, and has been correlated with an increased incidence of obesity resulting from a combination of a sedentary lifestyle and high energy diets [2]. Since a unifying mechanism underpinning the complex pathways leading to MS abnormalities remains undiscovered, present treatment regimes target MS symptoms through therapeutic intervention and lifestyle changes. Thus a better understanding of the underlying MS mechanisms is likely to provide a basis for development of more effective therapeutic strategies in MS treatment.
A growing body of evidence now demonstrates that estrogenic signalling can have an important role in MS development. Studies in both humans and rodents suggest that altered levels of estrogen or its receptors can lead to MS symptoms. For example, postmenopausal women, experiencing naturally decreased estrogen levels, are three times more likely to develop MS abnormalities than premenopausal women [6]. Furthermore, estrogen/progestin based hormone replacement therapy in postmenopausal women has been shown to lower visceral adipose tissue, fasting serum glucose and insulin levels [7]. Clinical observations in an estrogen receptor alpha (ERa) deficient male noted the development of hyperinsulinemia, impaired glucose tolerance (IGT), insulin resistance (IR) and increased body weight [8]. Additional cases show that men with decreased levels of aromatase, the principal enzyme of estrogen production, develop abdominal obesity, elevated blood lipids and IR, reviewed in [9,10]. Rodent studies demonstrate that whole body ERa knockout (KO) (ERaKO) models have increased body weight, IGT, and IR [11,12]. Aromatase KO mice diplay IR, IGT and increased abdominal fat, which are reversible by 17b-estradiol (E2) treatment [13,14]. Ovariectomy, resulting in low estrogen levels, leads to increased body weight, increased basal blood glucose and IGT which are reversible by reintroduction of estrogen [15,16,17,18]. The beneficial effect of estrogen in relation to normalising body weight and glucose homeostasis is further evidenced in ob/ob and high fat diet (HFD) fed mice, models of obesity and T2D. In both models, estrogen treatment improves glucose tolerance and insulin sensitivity [4,19,20], in addition to having a weight lowering effect in HFD-fed mice [20]. Collectively, these studies firmly establish a role for estrogenic signalling in the development of MS. However, these observations are derived from models with altered estrogenic action throughout multiple organs/tissues. This makes it difficult to correlate the sequences of events and tissue-specific contributions of the underlying estrogenic mechanisms to the observed phenotypes.
Estrogen signalling can be mediated by multiple receptors. Most of the known estrogenic effects are mediated via direct interaction of estrogen with the DNA-binding transcription factors, ERa and estrogen receptor beta (ERb) [21,22]. The resulting mechanism supports a ligand-modulated, ER-mediated, transcriptional gene regulation. Studies of whole-body ERbKO mice have shown that they do not exhibit altered insulin sensitivity and/or alterations in body weight [12]. However, some evidence exists that ERb might still contribute to the development of MS in older mice and/or under specific metabolic conditions [23]. In contrast, ERaselective signalling has clearly been associated with the MS. In addition to the observations from ERaKO mice [11,12], selective ablations of ERa in the hypothalamic brain region or the hematopoietic/myeloid cells have both been reported to give rise to an increase in body weight and attenuated glucose tolerance [24,25,26]. Furthermore, treatment of ob/ob mice with the ERaselective agonist propyl pyrazole triol (PPT) improved glucose tolerance and insulin sensitivity, supporting the importance of ERa action in the control of glucose and insulin function. In addition, estrogenic signalling has also been shown to occur via a membrane-bound G protein-coupled receptor (GPR) 30 [27] which has been implicated as an important factor in insulin production and glucose homeostasis [28].
Previously, using the euglycaemic-hyperinsulinaemic clamp, we showed that ERaKO mice exhibit defective insulin-mediated suppression of endogenous glucose production (EGP) [12]. Since the liver is the principal organ of EGP [29,30,31], we proposed that hepatic IR contributes to the observed IGT and IR in ERaKO mice. To further investigate the role of hepatic estrogenic action in the maintenance of hepatic glucose homeostasis, we now report the generation and characterisation of a liver-selective ERaKO mouse model, LERKO. We demonstrate that LERKO mice display efficient down-regulation of ERa expression selectively within the liver. LERKO body weight, hormone profiles as well as the glucose and insulin response are comparable to those of control (CT) animals even when challenged with HFD and/or aging. In addition comparative analysis of the hepatic transcriptional profile in LERKO animals with that of ERaKO animals showed that LERKO mice do not exhibit the changes observed in ERaKO mice. We henceforth speculate that the previously observed ERa-mediated hepatic insulin resistance in ERaKO mice occurs as a secondary effect in the development of MS abnormalities.

Results
The LERKO mouse model demonstrates liver-selective ERa ablation Successful liver-selective down-regulation of ERa was confirmed by evaluating the mRNA levels of ERa in muscle, liver, white adipose tissue (WAT), kidney, and uterus of LERKO and CT mice ( Figure 1A). Significant down-regulation (of approximately 90%) of ERa mRNA levels was observed exclusively in the LERKO livers (Figures 1 A and B). Livers from CT and LERKO mice were further assessed for levels of ERa protein (Figure 1 C). Uterus samples of control (CT) and ERaKO mice served as positive and negative controls, respectively. As expected, the liver and uterus of CT, but not the uterus of ERaKO animals showed the presence of a ,67 kDa ERa protein band. Densitometric analysis revealed that the protein band corresponding to ERa in male and female LERKO livers was ,1% that of controls (data not shown). To ensure that the observed differences were not due to varying total protein levels, we confirmed that the total protein levels across all samples were approximately equal ( Figure S1). Furthermore, actin levels were similar between LERKO and CT mice ( Figure 1C). ERa protein levels were also assessed by immunostaining, which indicated that ERa was predominantly localised in the hepatocyte nucleus, and was of weaker intensity in LERKO compared to CT animals ( Figure 2).
LERKO animals have normal liver histology, body weight, glucose and insulin response Analysis of gross liver tissue structure and lipid content in 6 month-old male and female LERKO and CT mice revealed no observable differences between LERKO mice and their respective controls ( Figure 2 and Figure S2). As expected, we observed increased levels of lipid droplets in female mouse livers, compared to males [32]. Additionally, 6 month old LERKO and CT mice had comparable body weight, glucose tolerance and insulin sensitivity (Figures 3 A-C). Furthermore, insulin-stimulated AKT phosphorylation in the liver was similar between CT and LERKO mice (Figure 3 D). We also examined hepatic transcript levels of SREBP-1c, a classic indicator of hepatic steatosis [33,34,35]. We observed no differences in SREBP-1c levels between CT and LERKO mice of respective genders ( Figure 4).
The global hepatic gene expression profile from LERKO mice does not reflect the profile observed in ERaKO mice In our previous study, we showed that livers from ERaKO mice demonstrated significant changes in the gene expression profile compared to CT mice [12]. To evaluate whether the hepatic transcriptional profile of LERKO animals shows similar changes to the previous observations in ERaKO animals, the LERKO hepatic gene expression was profiled by microarray expression analysis. Subsequently, the number of significantly regulated genes was evaluated and compared to the number of significantly regulated genes identified in ERaKO. Using a false discovery rate of 5%, we identified 3 significantly regulated genes (Table S1) in LERKO compared to 173 significantly regulated genes in ERaKO (Table S2). Of the 3 significantly regulated genes in LERKO mice, only the Esr1 (coding for ERa) gene was also significantly regulated in ERaKO mice. Furthermore, we specifically evaluated hepatic mRNA expression levels of glucose-6phosphatase (G6P), stearoyl-coenzyme A desaturase 1 (Scd1), ERb, GPR30 and the androgen receptor (AR) (Figure 4; ERb and GPR30 data not shown). We have previously shown that hepatic Scd1 and G6P expression levels are significantly affected by estrogenic signalling. In ERaKO mice, the hepatic Scd1 transcript is upregulated by ,5 fold [12], while in HFD mice, E2 treatment decreased the expression levels of both G6P and Scd1 within the liver [20]. In addition, a decrease in hepatic G6P mRNA levels is also observed in ob/ob mice treated with E2 or PPT [4]. Thus, we proposed that hepatic Scd1 and G6P are likely to be mediators of the observed effects of estrogen signalling on metabolic phenotypes. However, in LERKO livers, Scd1 and G6P did not It is notable that a second band is detected by the ERa antibody in the liver of male but not female mice. While it is difficult to identify the exact source of the second band, one possibility is that it represents male prominent ERa degradation products. In line with this, longer exposure reveals a double band also in one of the liver samples from female mice (data not shown). doi:10.1371/journal.pone.0057458.g001 demonstrate any significant change in mRNA expression levels compared to CT livers ( Figure 4).
Since both ERb and GPR30 can mediate estrogen signalling [27,28,36], we speculated that these signalling pathways could compensate for the reduced hepatic ERa signalling. However, in CT and LERKO mice, hepatic mRNA levels of ERb were undetectable while GPR30 mRNA levels were very low and of a similar level (data not shown) suggesting that ERb and GPR30 signalling did not have a compensatory role in the livers of LERKO mice. Additionally, AR signalling within the liver has recently been implicated in hepatic glucose and lipid homeostasis [37]. We measured the hepatic AR transcript to evaluate whether increased hepatic AR levels could be involved in the maintenance of insulin sensitivity in LERKO mice. However, LERKO and CT mice had comparable hepatic AR transcript levels ( Figure 4). Body weight, glucose response and hormone levels of LERKO mice are comparable to CT mice when challenged with a high fat diet and/or age To challenge the LERKO model, 8 month-old LERKO and CT mice were subjected to a 5 month HFD. As expected, the male and female HFD regimented mice showed a marked increase in body weight over the course of the diet, compared to age-matched standard chow-fed male and female mice (Figure 5 A). Importantly, there were no significant differences in body weights between male or female CT and LERKO mice (Figure 5 A). HFD-fed mice displayed pronounced reductions in glucose tolerance. However, no differences in glucose tolerance were observed between the CT and LERKO mice (Figure 5 B). CT and LERKO mice receiving the standard diet showed normal glucose tolerance (Figure 5 B). We determined levels of insulin and adiponectin as markers of insulin resistance, and of IGF-1 as a hormone sensitive to the diet. Analysis of circulating hormone levels revealed that CT and LERKO mice had comparable insulin, IGF-1 and adiponectin levels, within their respective sexes and for both diets (Figure 5 C). HFD-fed mice showed increased

Discussion
We previously demonstrated that ERaKO mice have pronounced hepatic insulin resistance [12]. To determine whether liver-selective ablation of ERa recapitulates the metabolic phenotypes of ERKO mice, we generated a liver-selective ERaKO mouse model, LERKO. The resulting LERKO mice displayed an efficient ablation of ERa expression selectively within the liver (Figures 1 and 2), confirming the successful generation of a liver tissue selective ERa KO model. We have previously shown that ERaKO mice display an increase in body weight, impaired glucose tolerance and insulin resistance [12]. In contrast, compared to CT mice, LERKO mice maintained comparable body weights, and responded similarly during glucose and insulin tolerance tests compared to CT mice even when challenged with a HFD or age (Figures 3 and 5). There were no differences either in the basal or glucose-stimulated insulin responses during the GTT between CT and LERKO mice (data not shown). Thus LERKO mice do not secrete more insulin to maintain a glucose response similar to that of CT mice during the GTT. Furthermore, to make sure that we did not miss an early and transient phenotype, we also measured glucose tolerance at 3 months of age and observed similar glucose tolerance for LERKO and CT mice for males and females (data not shown). Additionally, a key mediator of insulin signalling, phosphorylation of Akt, was similar between CT and LERKO mice for the liver. Thus LERKO mice maintain normal liver insulin sensitivity.
Della Torre S et al. [38] showed a small decrease in circulating levels of IGF-1 between CT and LERKO mice under specific dietary conditions. We do not observe any difference in circulating IGF-1 levels between CT and LERKO mice on a standard chow diet or on a HFD. However feeding with HFD resulted in increased level of IGF-1 both in CT and LERKO mice.
Together, these observations indicate that selectively ablating ERa action in the liver is not sufficient to recapitulate the metabolic phenotype observed in mice with whole body disruption of ERa. Our western blot analyses indicate that the remaining ERa in the liver of LERKO mice corresponds to less than 1% of that in livers from CT mice (Figure 1). The most likely source of the remaining ERa are the non-parenchymal cell types, as these have been noted not to express the albumin promoter used to drive the cre expression in LERKO mice [32,33]. Therefore, it remains possible that ERa signalling in non-parenchymal cells might have an important role in the ERa-mediated hepatic insulin resistance observed in ERaKO mice. In support of this possibility, a recent study showed that Kupffer cells (a hepatic nonparenchymal cell type) mediated responses that contribute to the onset of HFD-induced hepatic insulin resistance [39]. However, whether Kupffer cells contribute to hepatic insulin resistance in the absence of HFD, as observed in ERaKO mice and whether the observed Kupffer cell mediated effect is dependent on ERamediated signalling remains to be elucidated.
Other studies have noted that the albumin promoter is not fully active until the mice are 6 weeks of age [40]. Since ERaKO mice are ERa-null during their entire development, it is possible that the onset of the observed phenotype in ERaKO mice is dependent on ERa signalling during an early developmental phase.
Another possible explanation for the absence of an observable metabolic phenotype in LERKO mice could be the presence of compensatory mechanisms. Previous studies utilising liver-selective AR KO mice have implicated AR as a positive factor in preventing the development of hepatic steatosis and insulin resistance [37]. In addition, studies in mice lacking G proteincoupled receptor (GPR) 30, a functional estrogen receptor, have established GPR30 as an important factor in insulin sensitivity and glucose homeostasis [28]. Furthermore, since ERb is known to respond to estrogens [41], and associate with ERa related gene targets [41], it is possible that ERb could supplement the ERa function when ERa is low. All three of these proteins are potential candidates in driving potential compensatory mechanisms. However, we did not find any significant differences in AR, GPR30 or ERb transcript levels between livers of LERKO and CT mice, suggesting that these proteins are not involved in compensatory mechanisms in LERKO livers.
Our previous study of ERaKO mice demonstrated a moderate decrease in insulin stimulated glucose uptake in the muscle in vitro, suggesting that the muscle contributes to the metabolic phenotypes observed in ERaKO mice. This is consistent with the data reported by Ribas et al. [42], which shows a contribution by muscle to glucose disposal in vivo. Further studies using mice with skeletal muscle-selective ablation of ERa will further shed light on the contribution of ERa in the muscle to the observed phenotypes in ERaKO mice.
While an unknown compensatory mechanism might still be responsible for maintaining normal body weight, and glucose homeostasis in the LERKO mouse model, it is important to note that our gene expression profiling analysis indicated only 3 genes  out of the estimated ,28000 genes detectable by the utilised microarray had altered transcript levels between LERKO and CT mice (Table S1). Importantly, the presence of Esr1 in this category confirmed the successful downregulation of the ERa transcript in the LERKO mouse model. The remaining 2 genes were not identified as significantly changed in the corresponding ERaKO analyses (Table S2), thus further investigation is needed to evaluate the relevance of ERa-mediated signalling for the regulation of these genes. We speculate that the absence of further differences between the LERKO and CT hepatic transcriptional profiles does not support the presence of compensatory actions, which are likely to be reliant on alteration of metabolic pathways to maintain homestatic balance. However, it still remains possible that compensatory mechanisms working through metabolic pathways identical to ERa could still be present.
The MS is a complex multifactorial syndrome, involving multiple organs [43]. The exact sequence of events and the role of estrogenic action in the development of the MS remain to be elucidated. While our results suggest that downregulation of hepatic ERa action does not induce the development of insulin resistance and obesity, the possibility remains that ERa signalling has crucial roles downstream of the initiating factors.
In this study, we pursued a logical continuation of our previous work in delimitating the functional contribution of hepatic ERa for the previously observed hepatic insulin resistance exhibited by total ERaKO mice. However, our results indicate that downregulating ERa in the liver does not recapitulate the previously observed ERaKO phenotype. Whether this lack of an ERaKOlike phenotype in LERKO mice could be due to unidentified compensatory mechanism/s, or whether hepatic insulin resistance occurs as a secondary/downstream effect upon ablation of estrogen signalling in other cell types, remains to be elucidated.

Animals
C57BL/6J mice expressing the cre recombinase under the liver specific albumin promoter (B6.Cg-Tg(Alb-cre)21Mgn/J) were purchased from The Jackson Laboratory. The generation of floxed ERa mice has been described elsewhere [44]. To generate mice with a liver-specific knockout of ERa, the Alb-cre transgenic mice were crossed with ERa fl/fl animals to obtain Alb-cre/ERa fl/+ . These mice were subsequently crossed with ERa fl/fl animals, resulting in Alb-cre/ERa fl/fl mice (LERKO). ERa fl/fl animals served as controls. Genotyping for the Alb-cre transgene was performed with the following primers: F: 59-TAATGGGGTAG-GAACCAATG-3', R: 5'-GTTTCACTATCCAGGTTACGG-3'. Genotyping of the floxed-ERa locus was performed as described elsewhere [44]. Male and female LERKO mice were fertile with the females exhibiting a regular estrous cycle (data not shown). All animals were maintained on 12 h light-dark cycle, with food and water available ad libitum. From 8 to 13 months of age male and female LERKO mice, together with age-matched CT mice, were maintained on chow diet or on a HFD containing 34.9 g% fat, 26.2 g% protein and 26.3 g% carbohydrate (Research Diet, New Brunswick, NJ, USA). At the end of the experiment the mice were decapitated, blood was collected in heparinized tubes, centrifuged, and plasma was stored at 220C. Liver, adipose tissue, muscle, kidney and uterus were removed and stored at 280C.

Circulating hormone analyses
Plasma insulin levels were measured by RIA using 125 I-labeled porcine insulin, guinea pig anti-porcine serum and rat insulin as a standard (Novo Nordisk, Denmark). Plasma adiponectin levels were assessed with a double-antibody RIA technique in which 125 I-labeled murine adiponectin, multispecies adiponectin rabbit antiserum and mouse adiponectin standard were used (Millipore, Billerica, MA, USA). Plasma levels of IGF-1 were analysed by mouse/rat IGF-1 ELISA (Mediagnost, Germany).

Intraperitoneal glucose tolerance test (IPGTT)
In overnight-fasted mice, blood glucose concentrations were measured before and after (30, 60 and 120 min) the intraperitoneal injection of glucose at a dose of 2 g/kg. Blood glucose concentrations were analysed using the MediSence glucose analyser (Abbott Scandinavia AB, Solna, Sweden).

Intraperitoneal insulin tolerance test (IPIIT)
IPITT was performed in overnight fasted mice. Blood glucose concentrations were initially measured at the basal condition (0 min), then the animals were administered an intraperitoneal injection of insulin (0.25 U/kg) (Actrapid, Novo Nordisk) followed by an intraperitoneal injection of glucose (1 g/kg). Subsequently, blood glucose concentrations were measured at 15, 30, 60, 90 and 120 min after the glucose load.

Insulin signalling in vivo
Overnight-fasted animals were injected intraperitoneally with saline or human insulin (Actrapid, Novo Nordisk) at a dose 2 U/ kg. After 5 min, mice were sacrificed, tissues were harvested and stored at 280uC.

Assessment of AKT[pS473]
Liver protein extracts were prepared by homogenizing tissue in RIPA buffer, followed by centrifugations at 10 0006g for 10 min. Supernatants were transferred to new tubes and centrifuged again as previously. Finally, the supernatants were centrifuged at 14 0006g for 10 min. Protein concentrations of the extracts were determined using the BCA Protein Assay (Thermo Scientific, USA). Akt[pS473] was assessed by ELISA (BioSource, Belgium).

Immunohistochemistry
Paraffin-embedded tissue blocks were cut at 4mm thickness, deparaffinized, and rehydrated. Antigen retrieval was executed by microwaving the sections at 650 W in 10 mM citrate buffer (pH 7.0) for 15 min. Endogenous tissue peroxidase was then quenched by immersion in 0.5% H 2 O 2 /PBS for 30 min/RT, then 0.5% Triton X-100/PBS for 15 min. To minimise non-specific antibody binding, sections were treated with BlockAce (Dainippon Pharmaceutical, Japan) for 40 min/RT. The anti-ERa (Santa Cruz, MC-20: sc-542) primary antibody (1:250 dilution) was applied to the sections overnight/4uC in 10% BlockAce/PBS. Subsequently, the sections were washed and incubated for 1 h/RT with appropriate biotinylated secondary antibody (1:200 dilution). Visual staining was achieved with the avidin-biotin complex (ABC) method [45] with the Vectastain ABC kit (Vector). Peroxidase activity was visualized with 3,3'-diaminobenzidine (DAKO). Sections were lightly counterstained with hematoxylin. Negative controls were treated equally, without incubation with primary antibodies.

Oil Red O Staining
Fresh liver tissue were immersed in Tissue-Tek OCT compound (Sakura, Japan) and then frozen in isopentane cooled by liquid nitrogen. Samples were subsequently stored at -80uC. Stock Oil Red O solution was made by dissolving 300 mg of Oil red O powder (Sigma) in 100 mL of 99% isopropanol. Mixing, then filtering, 60 ml of the stock solution with 40 ml of distilled water produced the working solution, which was used within 1 h. Twelve micrometer frozen liver cryosections were air dried, incubated in Oil Red O working solution for 30 min then washed in distilled water. Sections were subsequently lightly counterstained with hematoxylin.

RNA extraction
Total RNA was isolated from frozen mouse tissues using the Trizol reagent (Invitrogen) as per the manufacturer's instructions. Isolated RNA was subsequently, purified with the RNeasy Plus Mini Kits (Qiagen), and quantitated using a NanoDrop 1000 spectrophotometer (Thermo Scientific) and the accompanying software.

Microarray Analysis
All microarray experiments have been performed in accordance to the MIAME microarray experiment guidelines [32]. The gene expression profile of liver in ERaKO mice was determined using Affymetrix MOE430 A arrays as reported previously [12]. In the current study, Affymetrix Mouse Gene 1.1 ST arrays were used for analysis of liver gene expression in LERKO mice. To facilitate the comparison between the ERaKO and LERKO data sets all microarray data were analysed or re-analysed with related packages available from Bioconductor [46]. Normalization and calculation of gene expression was performed with the Robust Multichip Average (RMA) expression measure using the affy package and oligo packages respectively for the ERaKO and LERKO studies. Prior to further analysis, a nonspecific filter was applied to remove genes with expression signals below 100 across all samples. Significant differential expression between KO and CT groups was assessed with the limma package, mean fold changes were estimated, and a false discovery rate of 5% was employed. The data discussed in this publication have been deposited in NCBI's Gene Expression Omnibus [47] and are accessible through GEO Series accession number GSE36514 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc = GSE36514).

Quantitative real-time PCR
Individual cDNA samples were assessed for gene expression by quantitative real-time PCR using the 7500 Fast Real-Time PCR System (Applied Biosystems) with the Fast SYBR Green master mix (Applied Biosystems) according to the accompanying protocol. Melting curve analyses was applied to confirm system/primer specificity. Relative gene expression was evaluated using either the Standard Curve (Applied Biosystems, User Bulletin #2) or, where appropriate, the comparative CT method [48]. Gene expression was normalised to murine glyceraldehyde-3-phosphate dehydrogenase (mGAPDH). Targeted qPCR primers utilised were mERb (F:GCCAACCTCCTGATGCTTCT;  R:TCGTACACCGG-

Western blot analysis
Frozen tissue was homogenized in RIPA Buffer (Sigma) containing complete mini protease inhibitor cocktail (Roche). To ensure equal loading, protein concentrations were determined using the BCA Protein Assay Kit (Pierce). Protein extracts (100mg liver, 5mg uterus) were resolved on a 7.5% Mini-PROTEAN TGX Precast Gel (Biorad) then transferred to Amersham Hybond-LFP membranes (GE Healthcare). Membranes were blocked with 5% skimmed milk (1 h, RT), then incubated overnight at 4uC with the anti-ERa antibody [1:200 dilution] (Santa Cruz, MC-20: sc-542) or anti-Actin antibody [1:10000] (Sigma Aldrich) under gentle agitation. Subsequently, the membranes were washed (3615 min) in PBS containing 1% Tween-20 (Duchefa Biochemie), then incubated with the anti-rabbit [1:2000 dilution] (for ERa) or antimouse [1:5000 dilution](for Actin) secondary antibody for 1 h at room temperature under gentle agitation. Following a repetition of the washing steps, the antibody targeted proteins were visualized with the use of the Pierce SuperSignal West Pico Chemiluminescent Substrate kit (Thermo Scientific) followed by autoradiograph film exposure. To ensure the observed differences in band intensity were not due to differential protein concentration, membranes were checked for equal lane loading by Coomassie R-350 staining as previously described [49]. Densitometric analysis was performed with ImageJ [50], utilising coomassie staining for normalisation of ERa band intensity.

Statistical analysis
Unless otherwise stated, quantitative data are expressed as mean 6SD. Statistical significance was assessed using the two-tailed Student's t-test assuming unequal variance. Figures 5A and 5 C were analyzed using two-way repeated measurements ANOVA or two way ANOVA, respectively, followed by the Tukey post-hoc test. Significance was established at P#0.05 and represented as an asterix (*).

Analysis of hepatic lipid composition
Hepatic lipids were extracted as previously described [51,52]. In brief, ,100 mg of liver was extracted in chloroform-methanol (2:1, v/v), solubilised in 1% Trition X-100 solution, and total cholesterol and triglycerides were determined by enzymatic assays.
Cholesterol and triglyceride reagents were purchased from Roche Diagnostics (GmbH, Mannheim, Germany).