Research ArticlePhysiology

Estrogen Reduces Lipid Content in the Liver Exclusively from Membrane Receptor Signaling

See allHide authors and affiliations

Science Signaling  21 May 2013:
Vol. 6, Issue 276, pp. ra36
DOI: 10.1126/scisignal.2004013

Abstract

Estrogen induces signal transduction through estrogen receptor α (ERα), which localizes to both the plasma membrane and nucleus. Using wild-type mice, ERα knockout (ERKO) mice, or transgenic mice expressing only the ligand-binding domain of ERα exclusively at the plasma membrane (MOER), we compared the transcriptional profiles of liver tissue extracts after mice were injected with the ERα agonist propyl-pyrazole-triol (PPT). The expression of many lipid synthesis–related genes was comparably decreased in livers from MOER or wild-type mice but was not suppressed in ERKO mice, indicating that only membrane-localized ERα was necessary for their suppression. Cholesterol, triglyceride, and fatty acid content was decreased only in livers from wild-type and MOER mice exposed to PPT, but not in the livers from the ERKO mice, validating the membrane-driven signaling pathway on a physiological level. PPT-triggered activation of ERα at the membrane induced adenosine monophosphate–activated protein kinase to phosphorylate sterol regulatory element–binding factor 1 (Srebf1), preventing its association with and therefore its proteolytic cleavage by site-1 protease. Consequently, Srebf1 was sequestered in the cytoplasm, preventing the expression of cholesterol synthesis–associated genes. Thus, we showed that inhibition of gene expression mediated by membrane-localized ERα caused a metabolic phenotype that did not require nuclear ERα.

Introduction

Estrogens regulate organ development and many complex functions through nuclear estrogen receptor α (ERα)–mediated transactivation of gene expression (1). In addition, estrogen binds to ER pools outside of the nucleus. In particular, a membrane-localized ER pool, which functions as an atypical G protein (heterotrimeric guanine nucleotide–binding protein)–coupled receptor lacking the seven-transmembrane structure, rapidly stimulates cyclic nucleotide production, calcium flux, and kinase activation in response to estrogen (2). Rapid signaling can either activate or inhibit nuclear ER function (3) and can cause nongenomic effects through kinase-induced phosphorylation of substrate proteins, regulating their cell localization and activity (4). Whether membrane-localized ER can regulate transcription independently from nuclear ER and whether this affects organ function are unknown.

One estrogen-responsive organ is the liver. We previously developed a mouse (MOER, for “membrane-only ERα”) that expresses the ligand-binding E domain of ERα exclusively at the plasma membrane of cells in many organs, but which lacks all other ERα pools (5). This E domain–only and membrane-localized ERα binds ligands, such as estrogen, but lacks the domains necessary for DNA binding or activation of RNA polymerase. MOER and ERα-knockout (ERKO) mice lack normal ovarian, uterine, and mammary gland development and function, indicating that nuclear ERα function is required for these processes. In contrast, rapid signal transduction by estrogen in vivo and in cells isolated from many organs, including the liver, is similar between wild-type and MOER mice (5), indicating that nuclear ERα may not be necessary for all organ functions. Furthermore, liver development appears normal in wild-type and MOER mice.

We therefore compared the in vivo effects of an ERα agonist, propyl-pyrazole-triol (PPT), in these mouse models, paying particular attention to genes that exhibited comparable changes in expression in the livers of wild-type and MOER mice but were not changed in ERKO mice. Because MOER mice lack nuclear ERα, and assuming that PPT specifically activated the E domain–only and membrane-localized ERα in these mice, regulation of the expression of those genes would be expected to be controlled by the membrane-localized ER. We found that the transcription of many genes encoding proteins involved in the cholesterol and other lipid synthesis pathways was comparably repressed by PPT administration in wild-type and MOER mice but not in ERKO mice. This finding correlated with a PPT-mediated reduction in hepatic lipid content and the suppression of both cholesterol synthesis and insulin-induced gene expression in cultured hepatocytes from wild-type and MOER mice. Membrane-localized ERα signaling through adenosine monophosphate (AMP)–activated protein kinase (AMPK) inhibited the nuclear localization of a key transcription factor controlling lipid synthesis in the liver, sterol regulatory element–binding factor 1 (Srebf1, also known as Srebp1) (6), thereby suppressing Srebf target gene expression. Thus, membrane-localized ERα signaling altered gene expression to produce a metabolic phenotype in the liver.

Results

ERα reduces the abundance of mRNAs in lipid synthesis pathways

Because estrogen [17β-estradiol (E2)]–induced signal transduction is similar in acutely isolated hepatocytes from wild-type and MOER mice (5), we used the liver to investigate the possible functions of membrane-localized ERα. Ovariectomized female wild-type, MOER, and ERKO mice were injected with either PPT or an oil vehicle for 3 days (fig. S1). The livers were removed and processed for mRNA microarray analysis and subsequent studies. We focused on genes with expression profiles, assessed by mRNA abundance, that were comparably regulated (activated or repressed) by PPT in both the MOER and wild-type mouse livers but which showed no change in livers from ERKO mice, indicating that regulation of these genes was mediated by membrane-localized ERα and was independent of nuclear ERα. Although most PPT-repressed gene expression required nuclear ERα (as seen in the wild-type mice), 30 genes (4%) were significantly and comparably inhibited in both MOER and wild-type, but not ERKO, mouse livers (Fig. 1A and Table 1). In addition, 18 genes were significantly stimulated by PPT in wild-type and MOER, but not ERKO, mouse livers (Table 1). By comparing expression patterns, we observed clustering of replicates and separation between the different treatments, which indicated similar responses to experimental treatment of select genes in wild-type and MOER mice, but which had no change in ERKO mice (fig. S2). Overall, the greatest separation occurred between oil- and PPT-treated wild-type mice. Scatter plots of the different groups demonstrate a high degree of correlation among all groups, which is expected because PPT is an ERα-specific agonist and, therefore, only a small number of genes out of the 28,000 probed on the Affymetrix chips will be differentially regulated (fig. S2).

Fig. 1 ERα suppresses lipid synthesis–related gene expression in the liver.

(A) Venn diagram of PPT-repressed genes in ovariectomized wild-type (WT), MOER, and ERKO mice. (B) qRT-PCR analysis of Srebf1 and Hmgcr expression in the livers of mice treated with either 0.1 ml of oil or 100 μg of PPT in 0.1 ml of oil each day for 3 days. Data are means ± SEM for each condition (n = 6 mice per condition and genotype from two independent experiments; *P < 0.05). (C) Western blot analysis of hepatic Srebf1 and Srebf2 abundance in mice treated with 100 μg of PPT or 0.1 ml of oil (n = 6 mice per condition and genotype from two independent experiments; *P < 0.05). (D) ChIP analysis of the recruitment of Srebf1 to the Srebf1c promoter in MOER-, WT-, or ERKO-derived hepatocytes treated with insulin (10 μg/ml), 10 nM PPT, or both (+PPT) (n = 3 independent experiments; comparing samples of the same genotype: *P < 0.05, against the control; +P < 0.05, against insulin alone). (E) Left: ChIP analysis of Srebf2 recruitment to the Srebf2 promoter in hepatocytes treated with PPT or oil and cultured in cholesterol-depleted or cholesterol-replete medium. Right: Srebf2 recruitment to the Hmgcr promoter in the livers from WT, MOER, and ERKO mice. Data are means ± SEM from three experiments (left) or from six mice per condition. **P < 0.05, statistically significant differences from the control (left) or oil-treated (right) samples of the same genotype.

Table 1 Genes significantly changed by PPT compared to oil injection in wild-type and MOER mice.

Genes that showed significantly and comparably increased or decreased expression after PPT treatment in the livers of MOER and wild-type (WT) but not ERKO mice are presented. Microarray data were analyzed using the Cyber T program at a confidence interval of 99.99%; a minimum of a twofold change was considered significant. DHEA, dehydroepiandrosterone.

View this table:

Gene ontology analysis indicated that many of the repressed genes encoded proteins that are involved in lipid and steroid synthesis or secondary metabolite biosynthesis, transport, and catabolism (Table 2). PPT suppressed multiple genes encoding proteins involved in the cholesterol biosynthesis pathway, including Srebf1, which encodes a transcription factor that stimulates many genes in the lipid synthesis pathways, and Hmgcr, which encodes 3-hydroxy-3-methylglutaryl coenzyme A (CoA) reductase, the rate-limiting enzyme for cholesterol synthesis (6). Additionally, expression of genes encoding proteins involved in the fatty acid and triglyceride synthesis pathways, such as Acot1 (encoding acyl-CoA thioesterase 1) and Aacs (encoding acetoacetyl-CoA synthetase), was reduced by PPT. We confirmed the suppression of some of these genes involved in lipid synthesis pathways by membrane-localized ERα in response to PPT by quantitative reverse transcription polymerase chain reaction (qRT-PCR) analysis of DNA from the livers of PPT-treated wild-type, MOER, and ERKO mice (Fig. 1B and table S1). Although Srebf2, which encodes another transcription factor involved in regulating lipid synthesis genes, was not found to be significantly changed by microarray analysis (Table 1), qRT-PCR analysis indicated that Srebf2 mRNA was significantly decreased after PPT administration in wild-type and MOER, but not ERKO, mice (table S1). Accordingly, compared with untreated mice, the abundance of both Srebf1 and Srebf2 proteins was decreased in the livers of PPT-treated MOER and wild-type mice compared with oil-treated controls, but their abundance was not diminished in PPT-treated ERKO mice (Fig. 1C).

Table 2 Gene ontology of genes changed by PPT in wild-type and MOER mice.

Gene ontology (GO) analysis was performed with the online bioinformatics resource DAVID. Gene lists were analyzed using the functional annotation chart resource, and the resulting enriched annotation terms were determined to be significant if P < 0.05 after Benjamini correction.

View this table:

To support a direct effect of PPT on the liver, we acutely isolated hepatocytes from all three mouse models and exposed the cells to insulin. Insulin stimulates cholesterol and other lipid production in the liver by activating liver X receptors (LXRs) (7, 8), which stimulates Srebf1 expression (9). Insulin also increases Srebf1 activity by stimulating the cleavage of the Srebf1 protein through a mechanism independent from LXR, resulting in the mature, nuclear form of Srebf1 protein (10). Nuclear Srebf1 occupies its own and other gene promoters to stimulate transcription of genes related to lipid synthesis (11). PPT significantly reduced insulin-stimulated recruitment of Srebf1 to the sterol response element DNA binding site in the Srebf1c promoter in hepatocytes from wild-type and MOER mice, but not ERKO mice (Fig. 1D). Incubating hepatocytes in cholesterol-depleted medium increased the abundance of Srebf2 at its gene promoter, which was prevented by PPT only in wild-type and MOER-derived cells (Fig. 1E). In contrast, hepatocytes from all three mice grown in cholesterol-replete medium showed a low abundance of Srebf2 at its own promoter, which was not further decreased by PPT, indicating that ERα predominantly suppressed stimulated cholesterol synthesis but not basal synthesis. Compared to those from vehicle-treated mice, the livers of PPT-treated wild-type and MOER mice, but not ERKO mice, had a reduced amount of Srebf2 at the Hmgcr promoter (Fig. 1E).

Membrane ERα suppresses lipid synthesis in murine liver cells

We found that liver cholesterol content was significantly reduced after PPT treatment in wild-type and MOER mice compared with those treated with vehicle, but this was not observed in ERKO mice (Fig. 2A). In addition, triglyceride and fatty acid (palmitate) content in the liver was comparably reduced by PPT only in wild-type and MOER mice (Fig. 2A). These metabolic pathways are transcriptionally activated mainly by Srebf1 (6).

Fig. 2 PPT inhibits lipid synthesis in the liver and hepatocytes.

(A) Cholesterol triglyceride and fatty acid (palmitate) content in livers from WT, MOER, and ERKO mice after treatment with either 100 μg of PPT or 0.1 ml of oil (n = 6 mice per condition and genotype; *P < 0.05, between oil- and PPT-treated samples within each genotype). (B) Cholesterol content in cultured hepatocytes treated with insulin (10 μg/ml), 10 nM PPT, or both. Data are means ± SEM from three independent experiments (*P < 0.05, against control; +P < 0.05, against insulin). (C) Cholesterol and fatty acid content in hepatocytes depleted of Srebf1 or Srebf2 and treated with insulin (10 μg/ml) (n = 3 experiments; *P < 0.05, against control; +P < 0.05, against insulin alone). (D) Efficiency of knockdown in hepatocytes. (E) Insulin-stimulated cholesterol production in hepatocytes treated with E2, PPT, or the ERβ agonist DPN (n = 3 experiments; *P < 0.05, against control; +P < 0.05, against insulin).

Further supporting the direct effects of membrane-localized ERα signaling transduction in the liver, insulin-stimulated cholesterol synthesis in cultured hepatocytes from wild-type and MOER, but not in ERKO, mice was significantly inhibited by PPT (Fig. 2B). Knockdown with small interfering RNA (siRNA) indicated that insulin-stimulated cholesterol production depended on the action of Srebf1, but not Srebf2 (Fig. 2, C and D). Furthermore, insulin-stimulated fatty acid synthesis in wild-type and MOER-derived hepatocytes was significantly decreased only in Srebf1-depleted cells (Fig. 2C), indicating that Srebf2 is not important for this specific action. Srebf2 stimulates cholesterol synthesis in liver and fat cells (6), but our data indicated that Srebf2 does not mediate cholesterol synthesis in the liver in response to insulin. Whereas PPT and E2 both inhibited insulin-induced cholesterol synthesis in liver cells from wild-type and MOER, but not ERKO, mice, insulin-induced cholesterol synthesis was unaffected by the ERβ agonist diarylpropionitrile (DPN) (Fig. 2E). These results implicate a specialized function for membrane-localized ERα in inhibiting lipid metabolism, and the expression data suggest that this may be mediated by suppressing the expression of genes involved in lipid biosynthesis.

Membrane ERα stimulates a protein kinase A–liver kinase B1–AMPK pathway

To decipher the mechanism by which ERα activation at the membrane inhibited cholesterol synthesis, we used pharmacological inhibitors of mitogen-activated or extracellular signal–regulated protein kinase kinase (MEK) (PD98059), AMPK (compound C), phosphatidylinositol 3-kinase (PI3K; LY290042), or the tyrosine kinase Src (PP2) to identify kinases and analyzed the effect of the inhibitors on cholesterol content (Fig. 3A) and on Srebf1 and Hmgcr expression (Fig. 3B) in hepatocytes from wild-type and MOER mice exposed to insulin and PPT (Fig. 3A). PPT inhibited insulin-stimulated cholesterol synthesis (Fig. 3A) and reduced the abundance Srebf1 and Hmgcr transcripts (Fig. 3B) and protein (fig. S3). We did not analyze the effect of PPT on Srebf2 expression in insulin-treated hepatocytes because Srebf2 was not found to mediate insulin-stimulated cholesterol synthesis in this organ.

Fig. 3 Membrane-localized ERα signal transduction activates AMPK and ERK to repress cholesterol synthesis.

(A to C) Cholesterol content (A), Srebf1 expression (B), and Hmgcr expression (C) in hepatocytes treated with insulin (10 μg/ml), 10 nM PPT, both, or in combination with inhibitors of AMPK [compound C (CC)], MEK [PD98059 (PD)], Src (PP2), or PI3K [LY290042 (LY)] (*P < 0.05, against control; +P < 0.05, against insulin; ++P < 0.05, against insulin + PPT; n = 3 experiments). (C) Western blot analysis of phosphorylated AMPK (Thr172) or ERK (Thr202 and Tyr204) in lysates of WT and MOER hepatocytes treated with 10 nM PPT (left) or PPT and compound C (right). Blots are representative of and data are means ± SEM from three independent experiments (*P < 0.05, against control; +P < 0.05, against PPT). (D) Western blot analysis of phosphorylated AMPK and ACC in hepatocytes treated with AICAR (0.5 mM), PPT (10 nM), or either in combination with compound C (20 μM). Data are means ± SEM from three independent experiments (*P < 0.05, against control; +P < 0.05, against AICAR).

Treatment with inhibitors of MEK to block the extracellular signal–regulated kinase (ERK)–mitogen-activated protein kinase (MAPK) pathway or AMPK, but not with the PI3K inhibitor to block the PI3K-Akt pathway or the Src inhibitor, prevented PPT from inhibiting insulin-stimulated cholesterol synthesis (Fig. 3A) or induction of Srebf1 and Hmgcr (Fig. 3B). PPT significantly induced the phosphorylation of both AMPK and ERK in cultured wild-type and MOER hepatocytes (Fig. 3C), indicating that PPT stimulated the activation of these kinases. AMPK activation was upstream of the activation of ERK because the AMPK inhibitor compound C blocked PPT-induced activation of ERK (Fig. 3C). Consistent with PPT stimulating AMPK activity, PPT stimulated the phosphorylation of the AMPK substrate acetyl-CoA carboxylase (ACC), which was also stimulated by the control 5-aminoimidazole-4-carboxamide-1-β-d-ribofuranoside (AICAR), an AMPK activator, and compound C prevented the stimulation of phosphorylation of both AMPK and ACC by either PPT or AICAR in wild-type and MOER-derived hepatocytes (Fig. 3D). Thus, AMPK and ERK mediate membrane ERα signaling that inhibits insulin-stimulated Srebf1 and Hmgcr production.

Liver kinase B1 (LKB1), which is activated by protein kinase A (PKA), stimulates AMPK activity. Knockdown of LKB1 decreased the PPT-induced phosphorylation of AMPK in wild-type and MOER-derived hepatocytes (Fig. 4A). Inhibition of PKA with H-89 significantly suppressed PPT-induced activation of both LKB1 and AMPK (Fig. 4B), and blocking cyclic AMP (cAMP) binding sites in PKA with Rp-8-bromo-cAMP or depletion of PKA prevented PPT-induced activation of AMPK and LKB1 (Fig. 4C). PPT also increased the generation of cAMP and the activation of PKA in wild-type and MOER-derived hepatocytes (Fig. 4D). These pharmacological studies suggest that membrane ERα activated a signaling pathway composed of PKA-LKB1-AMPK to suppress the expression of genes involved in hepatic lipid synthesis.

Fig. 4 Involvement of cAMP, PKA, and LKB1 in AMPK activation by membrane-localized ERα.

(A) Western blot analysis of phosphorylated AMPK in hepatocytes depleted of LKB1 (siLKB1) and treated with PPT (10 nM) (n = 3 experiments; *P < 0.05, against control; +P < 0.05, against PPT). Control and PPT (alone)–treated cells were transfected with control siRNA. Efficiency of knockdown by Western blotting is shown. (B) Western blot analysis of phosphorylated LKB1 (Ser431) and AMPK (Thr172) in hepatocytes treated with PPT (10 nM), PKA inhibitor H-89 (1 μM), or both (n = 3 experiments; *P < 0.05, against control; +P < 0.05, against PPT). (C) Western blot analysis of AMPK and LKB1 phosphorylation in hepatocytes treated with 10 nM PPT alone or in combination with either the cAMP analog Rp-8-bromo-cAMP (Rp, 100 μM) or siRNA against PKA (siPKA) (n = 3 experiments; *P < 0.05, against control; +P < 0.05, against PPT). Efficiency of PKA knockdown by Western blotting is shown. (D) Western blot analysis of PKA (Thr197), and cAMP generation after 5 min in 10 nM PPT (n = 3 experiments; *P < 0.05).

Signaling from membrane ERα inhibits Srebf1 processing

Srebf1c is the primary Srebf1 isoform in the liver and is activated by proteolytic cleavage to transcribe genes in the lipid pathways (12). The N-terminal, cleaved form of this transcription factor translocates to the nucleus and binds sterol response elements in the promoters of its target genes. PPT blocked the insulin-induced processing of endogenous Srebf1 to the mature 64-kD form, and this was prevented by inhibiting AMPK or ERK (Fig. 5A), and PPT prevented the insulin-induced translocation of Srebf1 to the nucleus in hepatocytes (Fig. 5B).

Fig. 5 Srebf1 protein processing is inhibited by membrane ERα.

(A) Western blot analysis of full-length endogenous Srebf1 (P) to N-truncated Srebf1 (N) in whole lysates from WT- and MOER-derived hepatocytes treated with insulin (10 μg/ml) and 10 nM PPT, AMPK inhibitor compound C (CC, 20 μM), ERK inhibitor (PD, 1 μM), or a combination thereof. Blots are representative of and data are means ± SEM from three independent experiments (*P < 0.05, against control; +P < 0.05, against insulin; ++P < 0.05, against insulin + PPT). (B) Intracellular localization of Srebf1 in PPT- and insulin-treated hepatocytes. White arrow indicates cytoplasmic Srebf1; yellow arrow indicates nuclear Srebf1. (C) The insulin-stimulated processing of myc-tagged WT or AMPK-site mutant (Mut) Srebf1 in hepatocytes treated with insulin, PPT, compound C, or a combination thereof. Samples were immunoprecipitated (IP) with the antibody recognizing the myc tag and blotted (IB) with the myc antibody or an antibody recognizing Srebf1. Blot shows the cleaved form of Srebf1, and data are means ± SEM from three independent experiments (*P < 0.05, against control; +P < 0.05, against insulin; ++P < 0.05, against insulin + PPT).

Unlike PPT, incubation of the hepatocytes with actinomycin D, which suppresses general transcription, did not impede the ability of insulin to stimulate cleavage of endogenous Srebf1 in hepatocytes (fig. S4). Thus, PPT inhibited both the insulin-induced expression of the gene (Figs. 1 and 3) and the posttranslational cleavage (Fig. 5A) and nuclear translocation of Srebf1 (Fig. 5B).

Membrane ERα–induced AMPK inhibits the association of site-1 protease with Srebf1

Processing of full-length Srebf1 to the mature cleaved form is mediated by the sequential actions of the site-1 protease (S1P) and site-2 protease (S2P) in the Golgi (13). AMPK phosphorylates Srebf1 at Ser372 (14) and prevents Srebf1 processing. To investigate this mechanism, we expressed myc-tagged wild-type or an S372A mutant of Srebf1 in hepatocytes of wild-type or MOER mice. Insulin stimulated the processing of either wild-type or S372A Srebf1 (Fig. 5C). PPT inhibited the insulin-stimulated processing of the myc-tagged, wild-type Srebf1 in both wild-type and MOER hepatocytes in an AMPK-dependent manner. However, the insulin-stimulated processing of the S372A mutant was not affected by PPT (Fig. 5C), suggesting that ERα suppresses the transcription of lipid pathway–associated genes by stimulating the AMPK-dependent phosphorylation of Srebf1.

Because insulin should stimulate the physical association of Srebf1 and S1P to initiate processing, we postulated that the ERα-AMPK–mediated phosphorylation of Srebf1 disrupts this interaction. Coimmunoprecipitation experiments showed that insulin comparably stimulated the association of endogenous Srebf1 and S1P in hepatocytes from both wild-type and MOER mice, producing a maximal interaction by 60 min (Fig. 6A). The ERα agonist PPT inhibited this interaction in both wild-type and MOER hepatocytes, but not in ERKO hepatocytes (Fig. 6B). In hepatocytes expressing the myc-tagged wild-type Srebf1, but not those expressing the S372A mutant, PPT inhibited the insulin-stimulated interaction of Srebf1 with endogenous S1P (Fig. 6C). PPT-mediated inhibition of this insulin-stimulated interaction was blocked by inhibition of AMPK with compound C (Fig. 6C). Insulin stimulated a time-dependent interaction between Srebf1 and S2P (Fig. 6D) that was similar to the interaction with Srebf1 and S1P in that both peaked at about 60 min. PPT inhibited the interaction between wild-type Srebf1 and S2P in an AMPK-dependent manner. Thus, membrane-localized ERα suppresses lipid synthesis in the liver by inducing the AMPK-mediated phosphorylation of Srebf1 at Ser372, which prevents the association of the unprocessed protein with S1 and S2 proteases, preventing posttranslational processing and nuclear translocation of Srebf1.

Fig. 6 AMPK signaling inhibits the interaction between S1P and Srebf1.

(A) Time course of insulin-induced association of endogenous Srebf1 and S1P in hepatocytes. Blots are representative of and data are means ± SEM from three independent experiments (*P < 0.05, compared to time zero sample of the same genotype). (B) Immunoprecipitation (IP) and Western blot (IB) of endogenous Srebf1 and S1P in the presence of either insulin or insulin and PPT at 60 min in hepatocytes. n = 3 experiments; *P < 0.05, against control; +P < 0.05, against insulin. (C) Immunoprecipitation and Western blot of myc-tagged WT or myc-tagged AMPK-site mutant (Mut) Srebf1 with S1P. Blots are representative of and data are means ± SEM from three independent experiments (*P < 0.05, against control; +P < 0.05, against insulin; ++P < 0.05, against insulin + PPT). (D) Effect of insulin and PPT on the interaction between S2P and Srebf1. Top: Time course of insulin-induced association of endogenous Srebf1 and S2P in hepatocytes. Middle: Interaction between S2P and Srebf1 60 min after treatment with insulin, insulin and PPT, or insulin, PPT, and compound C (CC). The abundance of Srebf1 and S2P in the lysates is shown. Blots are representative of and data are means ± SEM from three independent experiments. *P < 0.05, against control; +P < 0.05, against insulin; ++P < 0.05, against insulin and PPT cotreated cells.

Discussion

Steroid hormones are historically considered to induce their effects by binding to their receptors in the nucleus, resulting in modulation of gene expression that affects the development and function of target organs (1). However, estrogen induces rapid signaling (2) that occurs by engaging extranuclear ER (3). Although increasing evidence supports the physiological roles for extranuclear receptors, the mechanisms of action that result from interaction with their ligand are not well understood. For example, previous studies show that estrogen suppresses cholesterol synthesis, but the mechanism mediating this effect and whether it was transduced through its membrane- or nuclear-localized receptors were not determined (15, 16).

Here, we showed that signaling from membrane-localized ERα suppressed the expression of a subset of genes in the livers of female wild-type and MOER transgenic mice but not in ERKO mice. These findings support the idea that gene regulation and a resulting metabolic phenotype can occur exclusively from membrane ERα signaling. Engagement of ERα by its pharmacological agonist PPT reduced cholesterol, triglyceride, and fatty acid content in the livers of wild-type and MOER mice but not in the ERKO mouse. Supporting its direct effects in this organ, PPT also inhibited insulin-induced cholesterol synthesis in cultured hepatocytes from wild-type and MOER mice; this effect was absent in cultured hepatocytes from ERKO mice. However, by comparing wild-type and MOER mice, we found that the PPT-induced suppression of most hepatic genes in vivo required nuclear ERα presence (Fig. 1A).

Our results suggest that some actions of steroid hormones occur from engaging membrane-localized receptors and do not require nuclear receptor binding or activation. In these regards, our findings should be considered alongside previous gene regulation studies in vitro (17) and functional studies in vivo (18) that used an estrogen dendrimeric compound that only binds to membrane ERα or ERβ receptors. In these models, signaling from the membrane receptor(s) could have affected nuclear ER function as an integrated result, even if the nuclear receptor was not bound by its ligand. An example is the resulting phosphorylation of nuclear ER that stimulates transcription (19). This mechanism is unlikely to be one that we describe, because MOER mice lack nuclear ERα, and the gene and lipid abundance data are similar in liver tissue and isolated hepatocytes from wild-type and MOER mice, yet unaffected in those from ERKO mice. Relevant to these considerations are studies using mice with either the deletion of the AF1 domain or a mutated AF2 domain of ERα that precludes DNA binding by the receptor. In these and wild-type mice, E2 comparably prevents acute vascular injury to the endothelium (19, 20) through a mechanism that is mediated by membrane-localized ER that induces nitric oxide production by already abundant cytoplasmic nitric oxide synthase but that does not require de novo gene transcription (21, 22), in contrast to our studies here. Overall, our findings suggest that rapid signaling from the membrane ER modulates both gene transcription (genomic) and protein activities (nongenomic), sometimes independently of nuclear ER function.

In the nucleus, Srebf1 and Srebf2 bind to the promoters of many genes involved in lipid synthesis (23), sometimes in conjunction with additional coactivators such as nuclear receptors (LXR) (24) and transcription factors NF-Y (nuclear factor Y) and CREB (cAMP response element–binding protein) (23). Although Srebf2 has traditionally been reported to regulate cholesterol synthesis (13, 25), there are ample data indicating that at least in the liver, Srebf1 stimulates the relevant gene expression and synthesis of both cholesterol and triglycerides (6, 12, 23, 26). A genome-wide study of promoter occupancy by Srebf1 and its coactivators NF-Y and SP1 (specificity protein 1) identifies cholesterol synthesis genes as the top targets by gene ontology analysis (27). In vivo, overexpression of liver-specific Srebf1a produces a metabolic phenotype of increased lipid synthesis–associated gene expression and markedly enhanced cholesterol and triglyceride production (6). The Srebf1 germline KO mouse shows increased Srebf2 abundance and maintenance of cholesterol synthesis (6). Compensation by Srebf2 indicates that the activity of Srebf1 is important for cholesterol synthesis in vivo, inducing Srebf2 to maintain lipid homeostasis only in the complete absence of Srebf1, which may explain why knockdown of Srebf2 had no significant effect on insulin-induced lipid synthesis compared with Srebf1 (Fig. 2, E and F). Also, PPT reduced the abundance of Srebf1 in vivo but did not completely deplete it, perhaps explaining why the expected compensation from Srebf2 was not observed (Fig. 2, E and F). Indeed, we also show that PPT decreased the abundance of Srebf2 mRNA and protein in the livers of wild-type and MOER, but not ERKO, mice, suggesting that cholesterol synthesis in the liver could depend on Srebf2 when Srebf1 is dysfunctional or deleted.

What are the important signals generated from the engagement of membrane ERα by PPT that suppress gene expression and sterol synthesis? We defined here the directionality of the mechanism as ERα-cAMP-PKA-LKB1-AMPK-ERK. The AMPK and ERK effector kinases are activated by PPT through PKA activation to stimulate LKB1 kinase. PKA activation occurs when membrane ERα rapidly stimulates cAMP generation as we showed here, and inhibiting cAMP led to suppression of PPT-stimulated LKB1 and AMPK activity (Fig. 3G). From our previous studies in multiple cell types (28, 29), membrane ER stimulates a Gs protein subunit α (GαS)–dependent generation of cAMP and subsequent PKA activation, and this occurs because membrane ER functions as a G protein–coupled receptor, activating both Gα and Gβγ subunits (30, 31). Using the soluble PKA inhibitor (H-89) and siRNA to PKA, we also showed that this kinase was necessary for PPT-induced activation of LKB1 and AMPK. LKB1 activated AMPK, and AMPK activated ERK. It has been reported that PKA activates LKB1 in the liver (32), and earlier studies identified Ser431 as a target for PKA-mediated phosphorylation of LKB1 (33). PPT stimulated the phosphorylation of LKB1 at Ser431, and this was inhibited by knocking down PKA (with siRNA) or inhibiting its activity (with H-89). Estrogen treatment of ovariectomized female mice enhances the phosphorylation of AMPK at Thr172 (its active site) in skeletal muscle (34); however, the localization of the receptor involved, the downstream mechanism activating AMPK, and the impact on lipid synthesis were not determined. Our focus in particular on AMPK signaling is justified because this master regulator of multiple metabolic pathways inhibits the processing of Srebf1 to its cleaved, mature form that translocates to the nucleus (14, 35, 36), a mechanism supported by our results.

Here, we established that membrane ERα signaling through AMPK causes the phosphorylation of Srebf1 that inhibits its processing induced by insulin in hepatocytes. We report the mechanism whereby AMPK-mediated phosphorylation of Srebf1 at Ser372 prevented the physical interaction of this transcription factor with the S1 and S2 proteases, presumably in the Golgi, thereby precluding its posttranslational processing. It is possible that other steps in the trafficking of Srebf1 to the Golgi and subsequent processing or nuclear transport are also affected by AMPK or ERK. Regarding ERK, there are 19 highly potential ERK-mediated phosphorylation sites predicted for Srebf1 that require further investigation. A summary of membrane ER regulation of this pathway is shown in fig. S5.

The mechanism that suppresses lipid synthesis by membrane ERα signaling involves the posttranslational modification of the transcription factor that results in cytoplasmic sequestration. As a result, the Srebf1 protein does not translocate to the nucleus, and its target genes, including Srebf1 itself, are not transactivated (35). Our findings provide potentially important insight about how membrane ER signaling may suppress the expression of some genes and produce metabolic phenotypes, a mechanism that might be more pervasive—in multiple cell types and physiological processes—than previously realized.

In this regard, we have previously reported that E2-activated membrane ERβ signaling prevents cardiac hypertrophy through posttranslational modification of the NFAT (nuclear factor of activated T cells) family of transcription factors that sequester these proteins in the cytoplasm of cardiomyocytes (37). As a result, NFATs do not traffic to the nucleus where they would normally activate the hypertrophic and sarcomeric reorganization gene program upon appropriate stimulus. A similar principle underlies the membrane ERβ–mediated suppression of cardiac fibrosis, involving signaling to SMAD sequestration in the cytoplasm of cardiac fibroblasts (30). Therefore, the ability of ER to inhibit in vivo pathology can sometimes result from signaling by the membrane receptor to alter the localization of key transcription factors.

Collectively, our results provide the impetus to develop and test ER isoform–specific agonists and antagonists that only engage the membrane receptor and avoid the proliferative effects of estrogen at nuclear ERα. In the liver, agonist engagement at the membrane ERα could contribute to favorable lipid homeostasis, including preventing excessive lipid content that can progress to cirrhosis. It must be emphasized that in most situations, estrogen action requires membrane ER signaling to integrate with nuclear ER action to produce a final cellular outcome. However, in others such as hepatic lipid synthesis presented here, stimulation or inhibition of one pool may be sufficient to alter steroid receptor–induced effects.

Materials and Methods

Materials

The ERα agonist PPT {4,4′,4′′-(4-propyl-[1H]-pyrazole-1,3,5-triyl)trisphenol} and the ERβ agonist DPN were used at 10 nM in vitro and purchased from Tocris Bioscience. E2 was used at 10 nM in vitro and purchased from Steraloids. H-89 (PKA inhibitor, used at 1 μM), PD98059 (MEK inhibitor, 1 μM), LY290042 (PI3K inhibitor, 10 μM), AICAR (0.5 mM), and compound C (20 μM) were also obtained from Tocris. Rp-8-bromo-cAMP was from Calbiochem. Insulin was from Sigma. Antibodies against phosphorylated AMPK (Thr172), total AMPKα, phosphorylated ERK (Thr202 and Tyr204), total ERK, phosphorylated ACC (Ser79), total ACC, and total S1P and S2P were obtained from Cell Signaling Technology. Antibodies against phosphorylated LKB1 (Ser431) and total LKB1 were from Santa Cruz Biotechnology. Antibody against Srebf1 was a gift from T. Osborne (Sanford-Burnham Institute). Antibody against phosphorylated PKA (Thr197) and two siRNAs each against Srebf1, Srebf2, and LKB1 mRNAs were obtained from Santa Cruz Biotechnology.

Animal studies

Three-month-old wild-type, MOER, and ERKO mice were ovariectomized and used per Institutional Review Board–approved protocols at the Long Beach VA Healthcare facility. After ovariectomy and recovery, the animals were injected intraperitoneally with PPT (100 μg) or oil vehicle (0.1 ml) daily for 3 days. Whole livers were removed upon euthanasia.

Liver cell culture

Primary hepatocytes were obtained from whole livers by enzymatic digestion at 37°C as previously described (5). The cells were seeded in collagen-coated flasks overnight and then grown in standard Dulbecco’s modified Eagle’s medium with nutrient mixture F12 (DMEM/F12) supplemented with 15% fetal calf serum and insulin (10 μg/ml). The next day, the cells were switched to insulin-free medium for experiments. The cells were incubated with insulin and either PPT (10 nM) or dimethyl sulfoxide vehicle for 24 hours for lipid measurements and transcription-related experiments or for 5 to 15 min for kinase activation assays. Kinase inhibitors were added 20 min before insulin and PPT addition. In some experiments, each of two different siRNAs against Srebf1 or Srebf2 or LKB1 mRNA was transfected individually into hepatocytes (3 μg of siRNA per well in six-well plates) with Oligofectamine. The cells were recovered for 36 hours and then incubated with insulin or PPT.

Microarray analysis

Snap-frozen dissected whole liver tissue was ground to a fine powder, and total RNA was extracted and used for qRT-PCR analysis and Affymetrix DNA microarray analysis. Complementary DNA (cDNA) was synthesized using about 500 ng of RNA and oligo(dT) primers with the ImProm-II reverse transcription system (Promega) for qRT-PCR or converted to complementary RNA for microarray. The RNA from three animals was pooled for each treatment type and each mouse type for microarray analysis; the two independent experiments were performed. For each pooled sample, mouse DNA microarray using the Affymetrix mouse gene 1.0ST chip was run at the Microarray Facility at the University of California at Irvine. The array data were analyzed with the Cyber T program at a confidence interval of 99.99%, and a minimum of a twofold change was considered significant. Hierarchical cluster analysis, dendrogram generation, and plotting were carried out with R version 2.15 cluster package and custom scripts (38). The online bioinformatics resource DAVID (39) was used to determine mRNA targets changed only after PPT treatment in both wild-type and MOER mice and not in ERKO. Gene ontology analysis was also performed with the resource DAVID. Briefly, gene lists were analyzed with the functional annotation chart resource, and the resulting enriched annotation terms were determined to be significant if P < 0.05 after Benjamini correction for false positives.

Quantitative RT-PCR

qRT-PCR was performed on cDNA from murine liver tissue and hepatocytes with a 2× SYBR Green mix reagent (Bio-Rad) and the Bio-Rad iCycler with a melting temperature of 60°C. Reactions were performed in duplicate for each cDNA sample and normalized against the expression of 18S ribosomal RNA. Primers were designed with the Primer3 design tool (http://frodo.wi.mit.edu/) and are listed in table S2. Relative mRNA fold changes were calculated with the Ct method.

Lipid assays

To quantify cholesterol and triglyceride in the liver, 10 mg of liver tissue from each condition and for each mouse genotype was extracted with 200 μl of chloroform/isopropanol/NP-40 (7:11:0.1) in a microhomogenizer. The extracts were centrifuged for 5 to 10 min at 15,000g, and the liquid organic phase was transferred to a new tube and air-dried at 50°C to remove the chloroform. The samples were then put under vacuum for 30 min to remove trace organic solvent. The dried lipids were dissolved in Triton X-100 in 200 μl of the specific lipid assay buffer with vortexing. Assays for quantification of either triglyceride or fatty acid (palmitate) were carried out according to the manufacturer’s instructions for the Free Fatty Acids and Triglycerides Quantification Kits (BioVision), and cholesterol was measured with the Cholesterol/Cholesterol Ester Kit from Abcam. For in vitro cholesterol synthesis, cells were seeded in six-well dishes to 70% confluency and allowed to attach for 24 hours in medium supplemented with insulin (10 μg/ml), then switched to insulin-free medium for an additional 24 hours. Hepatocytes were then treated with PPT, E2, or DPN (10 nM for 24 hours) with or without insulin or kinase inhibitors, and cholesterol extraction was performed as described above.

Kinase activity assay

As a measure of kinase activation, phosphorylation of active sites in each kinase was detected by Western blotting of hepatic cell lysates with antibodies described in Materials. For ERK activity assays, the cells were exposed to PPT for 5 and 10 min (22). PKA activity was determined in cells incubated with PPT for 10 min. cAMP generation was determined with a kit, as previously described (28, 29), after incubating the hepatocytes under various conditions with PPT for 5 min.

Cholesterol depletion

Primary hepatocyte cultures were incubated in serum-free minimum essential medium supplemented with 50 μM cerivastatin (Sigma) and 50 μM mevalonate (Sigma). Cell lines were washed, and medium was replaced with depletion medium containing 10% (v/v) fetal bovine lipoprotein-depleted serum (Gibco BRL) in place of fetal bovine serum along with the same concentrations of cerivastatin and mevalonate. Control cells were fed with replete medium composed of the depletion medium containing 25-OH-cholesterol (10 μg/ml; Sigma) (40). Additional pools of cells were incubated with serum-free medium as the depletion medium.

Chromatin immunoprecipitation assay

Chromatin immunoprecipitation (ChIP) assays were performed with the Active Motif ChIP-IT Kit. DNA was used for ChIP from the in vivo livers. Mouse hepatocytes were treated for 6 hours and cross-linked with 1% formaldehyde, and the reaction was stopped by adding glycine. Preparation and enzymatic fragmentation of chromatin was performed with the Enzymatic Shearing Kit. An aliquot was used to verify fragmentation on agarose gel and the amount of DNA. For the assays, 50 μg of digested chromatin was precleared for 1 hour with protein A–Sepharose beads, and supernatants were incubated overnight with 10 μg of antibody against Srebf1 (H160) or rabbit immunoglobulin G (Santa Cruz Biotechnology) or no antibody (mock). The immunoprecipitated complexes were washed in lysis buffer, LiCl buffer, and tris-EDTA buffer. Proteins were eliminated with proteinase K in 10% (w/v) SDS. Chromatin DNA was extracted, dissolved in water, and used for PCR. Primers for the human Srebf1c proximal (active) promoter were 5′-GCTCAGGGTGCCAGCGAACCAGTG-3′ (sense) and 5′-GGGTTACTAGCGGACGTCCGCC-3′ (antisense). Primer sets for the analysis of the Srebf response element in the distal region of Srebf1 (negative control, exons 4 and 5) were 5′-CCCACTTCATCAAGGCAGACTCGC-3′ (sense) and 5′-CGACCATGTGGACTGTTGCCAAGATG-3′ (antisense). For the sterol response element (−119 to −128) in the Srebf2 promoter, the following primers were used: 5′-CCATCTTCCCCTCTCTTTCC-3′ (sense) and 5′-AGGGAAGATCCTGGGAGAAA-3′ (antisense). For the sterol response element in the Hmgcr promoter, the following primers were used: 5′-GCTCGGAGACCAATAGGA-3′ (sense) and 5′-CCGCCAATAAGGAAGGAT-3′ (antisense). 5′-PCR amplification products were analyzed by ethidium bromide staining of 2.5% (w/v) agarose gels.

Statistical analysis

The results (means ± SEM) are shown as bar graphs on the basis of data from at least three experiments and were analyzed by analysis of variance (ANOVA) plus Scheffe’s post hoc test. Significance was taken at P < 0.05.

Supplementary Materials

www.sciencesignaling.org/cgi/content/full/6/276/ra36/DC1

Fig. S1. Experimental design of in vivo studies.

Fig. S2. PPT modulates selective gene expression comparably in wild-type and MOER mouse livers.

Fig. S3. AMPK and ERK mediate the PPT inhibition of insulin-induced abundance of Srebf1 and Hmgcr1.

Fig. S4. PPT-induced inhibition of Srebf1 processing is a posttranscriptional event.

Fig. S5. Summary of ER signaling from the membrane.

Table S1. Validation of gene modulation by PPT from qPCR using hepatic RNA.

Table S2. Primer sequences.

References and Notes

Funding: F.O. was co-funded by Marie Curie Actions, the Irish Higher Education Authority Programme for Third Level Institutions Cycle 4, and the Italian National Research Council. This work was supported by grants from the NIH (CA-100366) and the Department of Veterans Affairs Merit Review Program (E.R.L.). Author contributions: F.O., A.P., and M.R. designed and carried out the studies; B.J.H. and H.H. contributed to study design and data analysis; and E.R.L. proposed and developed the studies, analyzed the data, and wrote the paper. Competing interests: The authors declare that they have no competing interests. Data and materials availability: Microarray data are available from the Gene Expression Omnibus, National Center for Bio-Informatics, accession number GSE45346. The MOER mouse or ERα C451A plasmid can be obtained by contacting E.R.L.
View Abstract

Navigate This Article