Estrogen Receptor α Mediates 17α-Ethynylestradiol Causing Hepatotoxicity*

Estrogens are known to cause hepatotoxicity such as intrahepatic cholestasis in susceptible women during pregnancy, after administration of oral contraceptives, or during postmenopausal replacement therapy. Enterohepatic nuclear receptors including farnesoid X receptor (FXR), pregnane X receptor (PXR), and constitutive active/androstane receptor (CAR) are important in maintaining bile acid homeostasis and protecting the liver from bile acid toxicity. However, no nuclear receptor has been implicated in the mechanism for estrogen-induced hepatotoxicity. Here Era–/–, Erb–/–, Fxr–/–, Pxr–/–, and Car–/– mice were employed to show that Era–/– mice were resistant to synthetic estrogen 17α-ethynylestradiol (EE2)-induced hepatotoxicity as indicated by the fact that the EE2-treated Era–/– mice developed none of the hepatotoxic phenotypes such as hepatomegaly, elevation in serum bile acids, increase of alkaline phosphatase activity, liver degeneration, and inflammation. Upon EE2 treatment, estrogen receptor α (ERα) repressed the expression of bile acid and cholesterol transporters (bile salt export pump (BSEP), Na+/taurocholate cotransporting polypeptide (NTCP), OATP1, OATP2, ABCG5, and ABCG8) in the liver. Consistently, biliary secretions of both bile acids and cholesterol were markedly decreased in EE2-treated wild-type mice but not in the EE2-treated Era–/– mice. In addition, ERα up-regulated the expression of CYP7B1 and down-regulated the CYP7A1 and CYP8B1, shifting bile acid synthesis toward the acidic pathway to increase the serum level of β-muricholic acid. ERβ, FXR, PXR, and CAR were not involved in regulating the expression of bile acid transporter and biosynthesis enzyme genes following EE2 exposure. Taken together, these results suggest that ERα-mediated repression of hepatic transporters and alterations of bile acid biosynthesis may contribute to development of the EE2-induced hepatotoxicity.

Estrogens have long been known to cause intrahepatic cholestasis during pregnancy in susceptible women, who are using oral contraceptives or who are on postmenopausal hormone replacement therapy (1). Intrahepatic cholestasis of pregnancy (ICP), 2 the most common hepatic disease during pregnancy, starts with modest itching associated with elevated levels of serum bile acids and can lead to spontaneous premature delivery and intrauterine fetal death (2,3). Experimental intrahepatic cholestasis induced by 17␣-ethynylestradiol (EE2) treatment in rodents is a widely used in vivo model to examine the mechanisms involved in estrogen-induced cholestasis (4). EE2 treatment decreases the ATP-dependent taurocholate transport in the hepatic canalicular membrane, which is thought to be due to impaired expression of the canalicular bile salt export pump (BSEP) (5). Moreover, treatment with EE2 also decreases sinusoidal uptake of bile acids by down-regulating the expression of the Na ϩ /taurocholate cotransporting polypeptide protein (NTCP) (6). These studies suggest that estrogens induce cholestasis by reducing both the influx and efflux of bile acid in hepatocytes, resulting in a decrease in bile flow. In addition, EE2 treatment is shown to alter bile acid composition, which is associated with cholestatic features (7). However, the molecular mechanism of these EE2-dependent alterations is still not fully understood.
Bile acid homeostasis is tightly regulated by multiple nuclear receptors including FXR, PXR, and CAR in physiological and/or pathological conditions (8 -10). It is well understood that bile acids repress bile acid biosynthesis by down-regulating transcription of the rate-limiting enzyme CYP7A1 through the FXR-SHP-LRH1 cascade (11)(12)(13). NTCP, responsible for bile acid uptake into the hepatocytes, is repressed by FXR activation (11). Simultaneously, FXR up-regulates the expression of BSEP, which increases bile acid efflux from the liver into the bile (11,14). Therefore, FXR regulates transport of bile acids and prevents their overaccumulation in hepatocytes. The activation of PXR inhibits production of additional bile acid by inhibiting CYP7A1 and inducing OATP2, which increases the uptake of bile acids from sinusoidal blood to the hepatocytes (15,16). CAR induces SULT2A1 and MRP4 to prevent toxic bile acid accumulation (17). Consistent with preventive roles of PXR and CAR in the cholestatic condition, the double null mouse lacking PXR and CAR has a more severe disruption of bile acids and cholesterol homeostasis (18 -20). However, no nuclear receptor has been implicated in estrogen-induced hepatotoxicity.
In the present study, we investigated whether nuclear receptor could be involved in the pathogenesis of estrogen-induced hepatotoxicity, using Era Ϫ/Ϫ , Erb Ϫ/Ϫ , Fxr Ϫ/Ϫ , Pxr Ϫ/Ϫ , and Car Ϫ/Ϫ mice. We provide direct in vivo evidence that synthetic estrogen EE2 exposure induces liver damage by activating the ER␣ signaling pathways, leading to an alteration of bile acids biosynthesis and repression of multiple bile acid and cholesterol transporters.
Animal Treatment-All protocols and procedures were approved by the National Institutes of Health Animal Care and Use Committee and were in accordance with National Institutes of Health guidelines. Era Ϫ/Ϫ , Erb Ϫ/Ϫ , Pxr Ϫ/Ϫ , Fxr Ϫ/Ϫ , and Car Ϫ/Ϫ mice were generated and characterized previously (10,11,21,22). Era Ϫ/Ϫ and Erb Ϫ/Ϫ mice on a background of C57BL/6 were obtained from Taconic Farms (Germantown, NY). Pxr Ϫ/Ϫ , Fxr Ϫ/Ϫ , and Car Ϫ/Ϫ mice used in these studies were maintained in 129/Sv and C57BL/6 mixed genetic background. All animals were housed in a temperature-controlled environment with 12-h light/dark cycles with access to standard chow and water ad libitum. Age-matched groups of 8 -12-week-old mice were used for all the experiments. Four to eight mice were used for each treatment group. Adult mice received subcutaneous injections of EE2 (10 mg/kg) or vehicle (80% 1,2-propanediol with 0.15% NaCl) once daily for 5 successive days. Twenty hours after the last injection, the mice were fasted for 4 h before harvesting blood from the retroorbital plexus for subsequent serum analyses and livers for RNA isolation and histology.
Histology and Mitosis Measurements-Liver samples from each mouse were fixed in 10% neutral buffered formalin. Slides were stained with hematoxylin and eosin using standard protocols and examined microscopically for structural changes. Hepatocyte proliferation was evaluated by immunohistochemical staining for Ki67 using rat antimouse Ki67 (TEC3) antibody (Dako Corporation). Immunoreactivity was visualized with the Vectastain Elite ABC kit (Vector Laboratories), and slides were counterstained with hematoxylin. To assess proliferative responses to liver lesions, mitotic nuclei were counted in 20 randomly selected fields under ϫ200 magnification, and the mitotic index was calculated by dividing the number of mitotic cells by the total number of hepatocytes.
Measurement of Serum Chemistries-Serum alkaline phosphatase (ALP) and total bile acids were measured using reagents and controls from Diagnostic Chemicals Ltd. and the Cobas Mira plus CC analyzer (Roche Diagnostics). Individual bile acid concentrations were measured by LC-MS/MS using a PE SCIEX API2000 ESI triple-quadrupole mass spectrometer (PerkinElmer Life Sciences) controlled by Analyst software as described previously (23).
Analyses of Biliary Lipids-Biliary lipid secretion was measured following surgery for bile collection. Animals were fasted for 4 h before surgery following EE2 (10 mg/kg) or vehicle injection for 5 successive days. Bile collection started between 9:30 and 11:30 a.m. to minimize influence of circadian variations. Animals were anesthetized with a single dose of pentobarbital (50 mg/kg body weight) and maintained under this condition throughout the experiments. The gallbladder was cannulated with PE-10 tubing after ligation of the common bile duct and secured with a silk suture. Bile collection was measured gravimetrically, assuming a bile density of 1.0 g/ml. The initial 15 min of biliary secretion that contained concentrated gallbladder bile was not used for analysis. Bile collection continued for up to 1 h and was constant over this period. Body temperature was maintained with a heat pad to prevent hypothermic alterations of bile flow. The biliary concentration of bile acids, cholesterol, and phospholipids were measured using the total bile acid test, cholesterol E test, and phospholipids B test (Wako Chemicals). The biliary excretion rates were calculated as the product of the bile flow and the biliary concentration.
Reporter Assays-HepG2 cells, grown in phenol red-free minimum Eagle's medium supplemented with 5% charcoal dextranstripped fetal bovine serum, were transfected with a Gal-responsive luciferase reporter and Gal-DNA binding domain fused ER␣-LDB, ER␤-LDB, FXR-LDB, PXR-LDB, and CAR-LDB (24), using the calcium phosphate method (Amersham Biosciences). At 16 h after transfection, cells were treated with 10 nM 17␤-estradiol, 10 M 17␣-ethynylestradiol, or Me 2 SO for 24 h and assayed for luciferase activity with the dual-luciferase reporter assay system (Promega). Luciferase activity was normalized for transfection efficiency using the phRL-TK vector as an internal control.
Analyses of Gene Expressions-Total RNAs were isolated from mouse hepatic tissue using the TRIzol method (Invitrogen), and cDNAs were synthesized from total RNAs with the SuperScript First-Strand Synthesis System (Invitrogen) and random hexamer primers. Real time PCR measurements of individual cDNAs were performed with the ABI Prism 7700 sequence detection system. Gene-specific primers and probes were designed using Prime Express software (PE Applied Biosystems) or purchased as the predesigned TaqMan gene expression assays genespecific probe and primer mixture (PE Applied Biosystems). The assay identification number of predesigned TaqMan gene expression assays (gene, assay ID number) and sequences of the probes and primers (gene, probe/forward primer/reverse primer, 5Ј to 3Ј) used in this study are as follows: The TaqMan rodent glyceraldehyde-3-phosphate dehydrogenase control reagent (PE Applied Biosystems) was used as an internal control. All real time PCR data were obtained using RNA isolated from tissues of individual animals.
Statistical Analyses-Values were reported as mean Ϯ S.E. Statistical differences were determined by a Student's t test. p values less than 0.05 were considered to be statistically significant. Statistical significances relative to the corresponding vehicle-treated control are displayed as number signs in Figs. 1, 2, 4, 5, and 6.
These results clearly demonstrate that ER␣ is required for the development of EE2-induced hepatotoxicity.
The livers of EE2-treated Wt ϩ/ϩ and Era Ϫ/Ϫ mice were examined histologically. Liver sections of vehicle-treated mice showed normal morphology ( Fig. 2A), whereas EE2-treated Wt ϩ/ϩ mice exhibited hepatocyte degeneration and infiltration of inflammatory cells in hepatic sinusoids. However, EE2-treated Era Ϫ/Ϫ mice did not show any liver damage. Accordingly, livers from the EE2-treated Wt ϩ/ϩ mice but not the EE2-treated Era Ϫ/Ϫ mice contained many Ki67-positive proliferating hepatocytes. To further quantify the response to liver damage, we counted the number of mitotic nuclei in these samples. The percentage of mitotic nuclei was markedly increased with EE2 treatment in Wt ϩ/ϩ mice. However, no such increase was observed in EE2-treated Era Ϫ/Ϫ mice (Fig. 2B).
To confirm the specificity of EE2 in receptor activation, we performed gene reporter assays by cotransfecting a Gal-responsive luciferase reporter gene construct with Gal-fused ER␣, ER␤, FXR, PXR, and CAR-LBD in HepG2 cells (Fig. 3). As shown in previous reports, both ER␣ and ER␤ were exclusively activated by EE2 as well as by the endogenous ligand 17␤-estradiol (E2) (26). Unexpectedly, EE2 treatment activated FXR and PXR ϳ2-fold, although there was no activation by E2. CAR was slightly activated by E2 treatment (24) but not by EE2 treatment. Both ER␣ and ER␤ are predominantly expressed in reproductive organs. However, ER␣ is significantly expressed in liver, whereas the level of ER␤ mRNA in liver is very low (27). Although E2 and EE2 activate both ER␣ and ER␤, ER␤ probably does not play a significant role in hepatocytes because of its low expression level, which is in agreement with the hepatotoxic phenotypes in Erb Ϫ/Ϫ mice following EE2 treatment.
Alteration of Bile Acid Biosynthesis-We examined the expression of genes involved in bile acid biosynthesis: Cyp7a1, Cyp8b1, Cyp7b1, Cyp27, Fxr, and Shp (Fig. 4A). The biosynthesis of the primary bile acids, cholic acid, and ␤-muricholic acid is mediated by either the neutral or acidic pathway (28,29). The rate-limiting reaction in the neutral pathway is catalyzed by CYP7A1. The expression of CYP7A1 was decreased in EE2-treated Wt ϩ/ϩ , Erb Ϫ/Ϫ , Fxr Ϫ/Ϫ , Pxr Ϫ/Ϫ , and Car Ϫ/Ϫ mice but was not altered in the EE2-treated Era Ϫ/Ϫ mice. Although basal expression of CYP7A1 remained unchanged in Era Ϫ/Ϫ mice compared with that of Wt ϩ/ϩ mice (supplemental Fig. 2), the repression of CYP7A1 reflected the response activity of ER␣ to EE2. Similarly, Erb Ϫ/Ϫ and Pxr Ϫ/Ϫ mice did not change basal expression of CYP7A1. Fxr Ϫ/Ϫ and Car Ϫ/Ϫ mice showed an increase in basal expression of CYP7A1 mRNA of 4-and 1.6-fold, respectively. However, FXR and CAR did not contribute to the repression of CYP7A1 by EE2 treatment. The expression of CYP8B1, another enzyme involved in the neutral pathway, was also   decreased in all EE2-treated mouse lines, except that the Era Ϫ/Ϫ mice retained the original level of CYP8B1. In contrast, the expression of CYP7B1, an important enzyme in the acidic pathway, was markedly induced in all EE2-treated mouse lines but not in the Era Ϫ/Ϫ mice. In addition, the expression of the gene encoding sterol 27-hydroxylase CYP27 was not changed in any of the mouse lines treated with EE2. Furthermore, no differences were found in the expression of a key regulator of bile acid homeostasis FXR and SHP, suggesting that the FXR-SHP negative feedback was not involved in the regulation of these biosynthesis enzymes by EE2 treatment.
To investigate the effect of altered CYPs expressions by EE2, we measured the composition of bile acids in male and female Wt ϩ/ϩ and Era Ϫ/Ϫ mice using LC-MS/MS. Indeed, consistent with the expression of CYP7A1, CYP7B1, and CYP8B1, the ratios of ␤-muricholic acid to cholic acid were significantly increased in Wt ϩ/ϩ mice following EE2 treatment with no alteration in Era Ϫ/Ϫ mice (Fig. 4B). These results suggest that alterations in bile acid biosynthesis may contribute to the development of hepatotoxicity by EE2 treatment.
Repression of Bile Acid Transporters and Cholesterol Transporters-To further understand the molecular basis of EE2-induced hepatotoxicity, we examined the expression of key hepatic transporters involved in bile acid, cholesterol, and phospholipid transport. We first examined the expression of canalicular transporters BSEP and MDR2 that have been identified as the genes responsible for familial intrahepatic cholestasis (PFIC2 and PFIC3), respectively (30). Expression of BSEP was decreased in EE2-treated Wt ϩ/ϩ , Erb Ϫ/Ϫ , Fxr Ϫ/Ϫ , Pxr Ϫ/Ϫ , and Car Ϫ/Ϫ mice but was not altered in the EE2-treated Era Ϫ/Ϫ mice (Fig. 5). In contrast, MDR2 levels were not significantly changed in any of the mouse lines treated with EE2. NTCP is responsible for bile acid uptake into hepatocytes, and OATP transporters are also involved in sodium-independent hepatic uptake of bile acids. A reduction in NTCP and OATP-C (the human homologue of rodent OATP2) is observed in patients with cholestasis (31). Indeed, NTCP expression was decreased in all EE2-treated mouse lines except the Era Ϫ/Ϫ mice. OATP1 and OATP2 levels were also decreased by EE2 treatment in all mouse lines except Era Ϫ/Ϫ mice. We also analyzed the expression of cholesterol efflux transporters ABCG5, ABCG8, and ABCA1. Expression of ABCG5 and ABCG8 mRNAs was significantly decreased by EE2 treatment in all mouse models except the Era Ϫ/Ϫ mice. In contrast, EE2 treatment had no effect on ABCA1 expression. ABCG5 and ABCG8 play a crucial role in cellular cholesterol efflux from hepatocytes to bile (32). EE2 decreased both ABCG5 and ABCG8 levels, suggesting abrogation of cholesterol homeostasis. To evaluate whether the repression of hepatic transporters reflected biliary lipids secretion, biliary bile acids, cholesterol, and phospholipid secretion rates were determined directly by cannulating the gallbladder after ligation of the common bile duct (Fig. 6). In agreement with the decreased expression of hepatic transporters, the secretion of bile acids and cholesterol was significantly decreased by EE2 in male and female Wt ϩ/ϩ mice but not in Era Ϫ/Ϫ mice. The secretion of biliary phospholipids was also decreased by EE2. Although the degree of repression of the major phospholipid transporter MDR2 was small, a significant decrease of phospholipids secretion was observed. These data are consistent with the hypothesis that biliary phospholipids secretion is coupled to biliary bile acids secretion. Thus, these data indicate that the activation of ER␣ is associated with a marked impairment in biliary secretion.

DISCUSSION
We have provided direct evidence that ER␣ can become a risk factor for the development of hepatotoxicity following estrogen exposures. Using various nuclear receptor knock-out mice, we have now shown that ER␣ is the receptor responsible for repressing multiple hepatic bile acid and cholesterol transporters; repressing hepatic transporters reduces biliary lipids secretions resulting in liver damage (Fig. 7). In addition, the ER␣-mediated alteration of bile acids biosynthesis may have also contributed to the EE2-induced development of hepatotoxicity. None of the other nuclear receptors (ER␤, FXR, PXR, and CAR) were shown to be involved in developing hepatotoxicity. Our present study with Era Ϫ/Ϫ mice is the first to demonstrate directly the involvement of ER␣ in estrogen-induced hepatotoxicities, although it has been reported that treatment with ER␣ agonists causes adverse effects such as gallstone formation in humans and experimental animal models (33,34).
The molecular mechanism of how ER␣ regulates CYPs and transporters remains an open question at the present time. Some of the genes  Wild-type and ER␣-null mice were injected daily with vehicle or EE2 for 5 days. The common bile duct was ligated, the gallbladder was cannulated, and hepatic bile was measured gravimetrically for 60 min. The concentrations of biliary bile acids, cholesterol, and phospholipids were determined, and the secretion rate of each lipid was determined from measurements of bile flow. White bars indicate vehicle-treated mice, and black bars indicate EE2-treated mice. #, statistical significance between vehicle-and EE2-treated mice of same genotype (p Ͻ 0.05).
analyzed in our present study can also be regulated by other nuclear receptors such as LXR, LRH1, and HNF4 (9,35). Although expression of the known HNF4-regulated gene (Vlacsr) was found to be repressed in the EE2-treated Wt ϩ/ϩ mice but not in EE2-treated Era Ϫ/Ϫ mice (supplemental Fig. 3), the possibility remains that ER␣ indirectly repressed NTCP and OATP1 through antagonizing HNF4 activity. However, BSEP that was repressed by EE2 in our study was up-regulated in HNF4-null mice (35). Hence even an indirect mechanism such as HNF4-mediated repression may explain some but not all of ER␣mediated regulation. Neither LXR targets (SREBP-1 (sterol regulatory element-binding protein-1), FAS, and SPOT14) nor the LRH1 target, scavenger receptor BI, was affected by EE2 treatment, suggesting no involvement of these two nuclear receptors in EE2 actions. We have performed promoter analyses for both Cyp7b1 and Cyp8b1 genes to delineate ER␣ response elements but failed to identify these elements. The molecular mechanism of EE2-induced hepatotoxicity may be complex and could require both direct and indirect regulations.
ICP is the most common liver disease during pregnancy. Although numerous studies have investigated ICP, the molecular basis for its development remains unknown. Because hepatic transporters BSEP and MDR3 (the human homologue of rodent MDR2) are identified as the genes responsible for progressive familial intrahepatic cholestasis (PFIC2 and PFIC3, respectively) (30), several studies have investigated the association of these cholestatic genes with ICP. Mutations in the BSEP gene are not implicated in sporadic cases of ICP in Swiss patients (36). In addition, common mutations in MDR3 are not found in familial and sporadic cases of ICP in Finnish patients (37). Although mutations in these genes may raise the susceptibility to ICP, the results of these studies do not confirm a direct correlation between mutational events in these candidate genes and ICP. However, it has been reported that BSEP expression is decreased in livers of ICP patients (4,31). In addition, the deletion of BSEP causes intrahepatic cholestasis in mice (38). Our current finding that ER␣ plays a key role to repress multiple transporters in the development of cholestasis suggests that mutations or polymorphisms in the ER␣ gene may be implicated in some form of ICP in humans.
Primary biliary cirrhosis, a chronic liver disease that results in cholestasis, predominantly affects women and is characterized biochemi-cally by elevated serum ALP. Interestingly, tamoxifen, an ER␣ antagonist used to treat breast cancer, is found to dramatically decrease ALP levels in primary biliary cirrhosis patients, suggesting that tamoxifen can also be used to treat cholestasis (39,40). These clinical reports support the notion that the hepatic gene regulation through ER␣ plays a key role in estrogen-induced cholestasis. Our present findings that ER␣ regulates bile acid transporters including BSEP have provided the molecular basis of ER␣-mediated estrogen-induced hepatotoxicity. FXR is also known to regulate the same genes involved in bile acid homeostasis. Although treatment with the synthetic FXR agonist GW4064 is shown to be effective in improving cholestasis and reducing gallstone formation in experimental animal model (41,42), no FXR agonist such as GW4064 has been approved as a therapeutic drug. Treatment of cholestasis with ursodeoxycholic acid is widely used and reduces pruritus in ICP. However, ursodeoxycholic acid therapy is not completely effective in preventing fetal death or premature delivery associated with fetal distress (43). The development of liver-specific ER␣ antagonists could yield alternative therapeutics for treatment of chronic cholestasis patients as well as ICP patients.