Estrogen Regulates Snail and Slug in the Down-Regulation of E-Cadherin and Induces Metastatic Potential of Ovarian Cancer Cells through Estrogen Receptor α

Abstract

Tumorigenesis is a multistep process involving dysregulated cell growth and metastasis. Considerable evidence implicates a mitogenic action of estrogen in early ovarian carcinogenesis. In contrast, its influence in the metastatic cascade of ovarian tumor cells remains obscure. In the present study, we showed that 17β-estradiol (E2) increased the metastatic potential of human epithelial ovarian cancer cell lines. E2 treatment led to clear morphological changes characteristic of epithelial-mesenchymal transition (EMT) and an enhanced cell migratory propensity. These morphological and functional alterations were associated with changes in the abundance of EMT-related genes. Upon E2 stimulation, expression and promoter activity of the epithelial marker E-cadherin were strikingly suppressed, whereas EMT-associated transcription factors, Snail and Slug, were significantly up-regulated. This up-regulation was attributed to the increase in gene transcription activated by E2. Depletion of endogenous Snail or Slug using small interfering RNA (siRNA) attenuated E2-mediated decrease in E-cadherin. In addition, E2-induced cell migration was also neutralized by the siRNAs, suggesting that both transcription factors are indispensable for the prometastatic actions of E2. More importantly, by using selective estrogen receptor (ER) agonists, forced expression, and siRNA approaches, we identified that E2 triggered the metastatic behaviors exclusively through an ERα-dependent pathway. We also showed that ERβ had an opposing action on ERα because the presence of ERβ completely inhibited the EMT and down-regulation of E-cadherin induced by ERα. Collectively, this study provides a compelling argument that estrogen can potentiate tumor progression by EMT induction and highlights the crucial role of ERα in ovarian tumorigenesis.

OVARIAN CANCER is the leading cause of death from gynecological tumors and is the fourth most frequent cause of death from cancers in females. The disease is highly lethal because of its aggressive metastasis within the peritoneal cavity and the fact that it is often diagnosed at an advanced stage. Epithelial ovarian cancers originating from the ovarian surface epithelium constitute approximately 90% of human ovarian cancers (1).

Epithelial-mesenchymal transition (EMT) is a developmental process of crucial importance in cell differentiation, morphogenesis, and growth. This process allows cells to dissociate from the epithelial tissue and adopt a motile phenotype. During this process, epithelial cells become elongated, lose their polarity and cellular junctions, and show mesenchymal, fibroblast-like properties. Accumulating evidence has also proposed that similar processes occur during cancer progression (2). Because of its clinical importance, intensive research has been undertaken to identify reliable molecular markers and key mediators of EMT. They include proteins that actively control cell junction formation, cytoskeletal remodeling, and cell motility such as cadherins, matrix metalloproteinases, and several transcription factors (3, 4).

Among all members in the cadherin superfamily, E-cadherin represents one of the best-characterized markers of EMT and it is regarded as a tumor suppressor. It mediates calcium-dependent intercellular junctions, and abolishment of its function induces mesenchymal phenotypes. It also influences intracellular signaling pathways through its association with catenins, which are cytoplasmic counterparts of cadherins and are implicated in numerous pathological states including malignancy (5). Mutation or decreased expression of E-cadherin is often correlated with advanced-stage carcinomas (6, 7, 8, 9). E-cadherin can be silenced or functionally inactivated by several mechanisms, including down-regulation of gene expression, gene mutations, and posttranslational regulation. Down-regulation of gene expression can occur by promoter hypermethylation, histone deacetylation, and transcriptional regulation (10, 11, 12).

Transcriptional repression is emerging as a predominant mechanism controlling the expression of E-cadherin in most carcinomas. Several transcriptional repressors, which include Snail (13), Slug (14), SIP-1/ZEB-2 (15), E12/E47 (16), and Twist (17), have been characterized. They are found to interact with the proximal E-boxes of the E-cadherin promoter. Snail and Slug are zinc finger proteins that belong to the Snail superfamily (18). An inverse correlation between E-cadherin and Snail or Slug has been described in many different cell systems (13, 19, 20). Additionally, Snail and Slug have been detected at sites of EMT in vertebrates (18). Ectopic expression of Snail and Slug in epithelial cells caused EMT by acquisition of migratory and invasive behaviors, which were also evident in ovarian cancer (13, 21). Thus, these two transcription factors are considered as promising markers for malignancy, and identifying their upstream regulators is necessary.

Risk factors associating with ovarian cancers have been proposed in many epidemiological studies. Among them, age, family history, and infertility are known to increase the risk of ovarian cancer, whereas parity, oral contraceptive use, breast feeding, hysterectomy, or tubal ligation decreases the risk (22, 23). Although it is still under debate, there is a growing body of evidence that supports the contribution of estrogen in some of these risk factors. First, ovary is a main source and target tissue of estrogen. Second, estrogen receptors are found in human ovarian surface epithelium and more than 60% of ovarian cancer cells (24). Finally, a positive correlation between ovarian cancer incidence and estrogen has been established by many experimental and epidemiological studies (25, 26, 27, 28, 29, 30). Recent meta-analysis of cohort studies, case-control studies, randomized controlled trials and cancer registry studies has shown that estrogen therapy is a critical risk factor for ovarian cancer (31). The mitogenic effect of estrogen may provide one possible explanation for the increased risk. However, whether estrogen may play an additional role in ovarian cancer cell metastasis is poorly understood.

Here, our findings demonstrated that 17β-estradiol (E2) exposure led to acquisition of both phenotypic and molecular attributes typical of EMT in ovarian epithelial cancer cells BG-1. E2 promoted a spindle-like and mesenchymal phenotype and significantly enhanced cell motility. These pronounced changes were associated with inhibition of E-cadherin expression. Furthermore, we identified estrogen receptor (ER)α as a critical mechanistic link in mediating the promotion of EMT, whereas in the presence of ERβ, the pro-metastatic effects of ERα were inhibited. This is the first demonstration of the distinct roles of ERα and ERβ in ovarian tumor metastasis through the differential regulation of EMT-related genes.

RESULTS

E2 Stimulates EMT and Migration of BG-1 Cells

First, we aimed to elucidate the functional role of E2 in EMT of BG-1 cells, an estrogen-responsive ovarian cancer cell line. BG-1 cells underwent apparent changes in morphology compatible with EMT after 48 h of E2 (10⁻⁷ m) treatment (Fig. 1A). The phenotypic changes observed include conversion from cuboidal, epithelial morphology to spindle-shaped morphology, increased intercellular separation, and scattered and increased formation of pseudopodia. Epithelial cells that undergo EMT are characterized by increased motility; we therefore asked whether E2 could increase migratory ability of BG-1 by using a wound healing assay. E2 was administered 24 h before the wound was generated, and the rate of cell migration was subsequently measured at 0, 4, and 24 h after treatment. The wounded area of BG-1 cell monolayers healed slowly in vehicle-treated cells (Fig. 1B). On the contrary, in the presence of E2, the closure of the wounded gap was significantly accelerated at 24 h (Fig. 1B). The percentage of cell-free area at the indicated time points compared with that at time zero was determined (Fig. 1B, lower panel). In E2-treated cells, cell-free area was decreased by more than 2-fold compared with control cells at 24 h (P < 0.05). During the incubation period, cells moved forward and closed the gap independently of cell division. There was no significant increase in cell growth within 48 h under the same treatment conditions as revealed by cell proliferation assay (data not shown) and was consistent with previous studies (32, 33). We conclude that E2 promotes EMT, which leads to increased migration of BG-1 cells.

Fig. 1.

BG-1 Cells Are Induced by E2 to Undergo EMT

A, BG-1 cells were treated with vehicle (Cont) or 10⁻⁷ m E2 for 48 h. Morphological changes were observed by phase-contrast microscope, and representative photographs of at least three independent experiments are shown. Original magnification, ×200. B, Confluent monolayers of BG-1 cells were wounded with a uniform scratch, washed to remove cell debris, and cultured for indicated times in the presence of vehicle (Cont) or 10⁻⁷ m E2. Images of cell cultures were captured at 0, 4, and 24 h after scratching; representative pictures are shown in upper panel. The arrow indicates wound edge. The amount of wound repair was expressed as uncovered area at the indicated time compared with initial uncovered area of vehicle-treated control at time zero (lower panel). Values are the mean ± sd of three separate experiments. *, P < 0.05 compared with control. Cont, Control.

E2 Suppresses E-Cadherin Expression in Dose- and Time-Dependent Manner

Decrease of E-cadherin, a widely accepted characteristic associated with EMT, is often correlated to higher mobility of tumor cells (34). Given the role of E2 in controlling EMT as demonstrated in Fig. 1, we investigated the effect of E2 on the expression of this epithelial marker by Western blot analysis. Dose-response studies showed that lower concentrations of E2 (10⁻¹¹ m to 10⁻¹⁵ m) did not affect E-cadherin expression. The down-regulation effect became apparent with doses of 10⁻⁹ m and 10⁻⁷ m of E2 and was maximal at 10⁻⁷ m (Fig. 2A) (P < 0.05). Densitometric analyses of three independent experiments confirmed this finding (Fig. 2A, lower panel). Cells were also treated with 10⁻⁷ m E2 for different time courses. As shown in a representative blot in Fig. 2B, protein levels of E-cadherin were significantly decreased after E2 treatment. Maximal inhibition was seen at 48 h (a 70% decline) (P < 0.05), and the effect was still detectable after 72 h. To determine whether ER activation was involved, the effect of E2 was examined in the presence of the pure ER antagonist, ICI 182,780 (ICI). As shown in Fig. 2B, the down-regulation was prevented by 10⁻⁷ m ICI. These results suggest ER is a mediator of the suppression effects on E-cadherin by E2.

Fig. 2.

E2 Down-Regulates E-Cadherin Expression

A, BG-1 cells were treated with vehicle (Cont) or increasing concentrations of E2 (10⁻¹⁵ to 10⁻⁷ m) for 48 h. B, BG-1 cells were treated with either vehicle (Cont), 10⁻⁷ m E2 or in combination with 10⁻⁷ m ICI for 24, 48, and 72 h. Whole-cell lysates was extracted, and E-cadherin protein (E-Cad) was detected by Western blot. β-Actin was used as a loading control. The immunoblot shown is representative of three independent experiments. Cumulative results for quantitative densitometry of the three experiments are shown in lower panels. Mean ± sd values are depicted for protein abundance expressed as percentage in control cells. *, P < 0.05 compared with control. ⁺, P < 0.05 compared with the corresponding time-matched vehicle control. Cont, Control.

Transcriptional Repression of E-Cadherin by E2

To assess whether the reduction in E-cadherin expression was due to transcriptional regulation, E-cadherin mRNA levels after E2 (10⁻⁷ m) treatment were evaluated by real-time PCR. E-cadherin mRNA expression was decreased by approximately 50% as early as 12 h after E2 administration, with sustained decreases observed thereafter (Fig. 3A) (P < 0.05). This suggests that E-cadherin expression is suppressed in BG-1 cells at the transcriptional level. To confirm this, cells were transfected with reporter construct cloned with the E-cadherin proximal promoter and treated with E2 for periods up to 72 h. Concurrent with the loss of mRNA expression, treatment with E2 inhibited E-cadherin promoter activity (Fig. 3B). The promoter activity was significantly decreased by 25% after 24 h. This inhibitory effect persisted after 48–72 h, producing up to 50% decrease (P < 0.05). ICI was able to overcome this inhibitory effect because E2 and ICI in combination increased the transcription activity compared with E2-treated cells (Fig. 3B).

Fig. 3.

E2 Reduces E-Cadherin mRNA Synthesis through Transcriptional Repression

A, BG-1 cells were cultured in the presence of vehicle (Cont) or 10⁻⁷ m E2 for various time periods indicated. Total RNA was isolated and used for real-time PCR analysis of E-cadherin mRNA expression with gene-specific primers. GAPDH primers were used to normalize data, and results are expressed as fold change relative to time zero. B, BG-1 cells were transfected with 1 μg reporter construct of E-cadherin promoter. All cells were cotransfected with 0.5 μg pSV-βGal as an internal control for transcription efficiency. Cells were treated 8 h after transfection with vehicle (Cont), 10⁻⁷ m E2, or in combination with 10⁻⁷ m ICI, for an additional 24, 48, or 72 h. The luciferase activities were calculated relative to the promoterless vector (pGL3-Basic) and expressed as fold change relative to vehicle control at the corresponding time point. The data are shown as mean ± sd of three repeated experiments. *, P < 0.05 compared with the corresponding time-matched vehicle control. Cont, Control; E-Cad, E-cadherin.

E2 Enhances Expression and Promoter Activities of Snail and Slug

Transcription of E-cadherin is known to be regulated by the Snail and Slug repressors (13). To elucidate the mechanisms through which E2 suppresses E-cadherin, the protein levels of Snail and Slug after E2 treatment were investigated by Western blot. As shown in Fig. 4A, both Snail and Slug were significantly elevated in dose-dependent manner after 8 h of E2 treatment (P < 0.05). In comparison, analysis of Twist, which is another known E-cadherin repressor (17), showed no change in protein level with the same treatment (data not shown).

Fig. 4.

E2 Regulates Expression and Gene Promoter Activities of Snail and Slug

A, BG-1 cells were treated with vehicle (Cont) or increasing concentrations of E2 (10⁻¹⁵ to 10⁻⁷ m) for 8 h. Whole-cell lysates were extracted, and Snail and Slug were detected by Western blot. β-Actin was used as a loading control. The immunoblot shown is representative of three independent experiments. Cumulative results for quantitative densitometry of the three experiments are shown in left panel. B, BG-1 cells were cultured in the presence of vehicle (Cont) or 10⁻⁷ m E2 for the various time periods indicated and harvested for RNA extraction. C, Cells were treated with 10⁻⁷ m E2 or vehicle (Cont) for 1 h. Then, at time 0 h, the transcription inhibitor actinomycin D (5 μg/ml) was added to the medium and cells were harvested at the indicated time points. Total RNA was isolated and used for real-time PCR analysis of Snail or Slug mRNA expression with gene-specific primers. GAPDH primers were used to normalize data, and results are expressed as fold change relative to time zero. D, BG-1 cells were transfected with either human Snail or Slug promoter-reporter gene construct and pSV-βGal for normalization. The luciferase activities were determined in cell lysates 2–8 h after initiation of E2 exposure. The luciferase activities were calculated relative to the promoterless vector (pGL3-Basic) and expressed as fold change relative to vehicle control at corresponding time point. The data are shown as mean ± sd of three repeated experiments. *, P < 0.05 compared with control. +, P < 0.05 compared with the corresponding time-matched vehicle control. Cont, Control.

The time course effects of E2 on Snail and Slug mRNA levels were determined by real-time PCR as shown in Fig. 4B. E2 caused a marked induction of Snail mRNA at 1 h (∼2-fold increase), with significant increased mRNA levels being maintained within 1–4 h. Similarly, E2 treatment elevated Slug mRNA with a maximal stimulation reached in 1 h. However, the induction was transient, and the level of Slug mRNA returned to basal level at 4 h. Having established a correlation between E2 and Snail or Slug expression, we asked whether E2 stabilized their mRNA transcripts by actinomycin D chase experiment. The apparent half-lives of both Snail and Slug mRNA transcripts were approximately 1 h in control cells and were not changed upon E2 treatment (Fig. 4C). Thus, the induction of Snail and Slug in response to E2 was not due to an effect on message stability. We next investigated whether E2 can regulate their transcriptions. To this end, we performed luciferase promoter assays and examined the ability of E2 to transactivate luciferase reporter plasmids containing the human Snail or Slug promoter. Figure 4D showed that both Snail and Slug promoters were strongly activated by E2. These activations were already evident after 2 h of E2 treatment (P < 0.05). Induction of the activities peaked at 4 h (2- to 3-fold) and started to decrease at 8 h.

Snail and Slug Are Responsible for E2-Induced Repression of E-Cadherin and Cell Migration

The ability of E2 to increase Snail and Slug expressions implicates that these two repressors might be involved in transcriptional repression of E-cadherin in BG-1 cells. To confirm this, we used small interfering RNAs (siRNAs) to knock down Snail and Slug expressions in BG-1 cells. Depletion of endogenous Snail and Slug with siRNAs (sequence 1) was verified by Western blot (Fig. 5A). We also showed that knockdown of either Snail or Slug significantly interrupted the E2-mediated down-regulation of E-cadherin (Fig. 5B), whereas nonspecific siRNA had no observable effect. Another set of siRNAs with different sequences to Snail and Slug (sequence 2) was used to rule out the off-target effects, and the same results were obtained (supplemental Fig. S1 published as supplemental data on The Endocrine Society’s Journals Online web site at http://mend.endojournals.org). Concomitantly, suppression of E-cadherin promoter by E2 was significantly abolished by transfection with either Snail or Slug siRNA, but not with nonspecific siRNA (Fig. 5C).

Fig. 5.

Snail and Slug Mediate the Regulation of E-Cadherin and Cell Migration by E2

A, BG-1 cells were transfected with either 250 nm of Snail siRNA (sequence no. 1), Slug siRNA (sequence no. 1) or nonspecific siRNA (NS siRNA) for 48 h. Efficacy of the siRNAs was determined by Western blot analysis. B, Eight hours after transfection with the siRNAs, BG-1 cells were treated with vehicle (Cont) or 10⁻⁷ m E2 for an additional 48 h. Whole-cell lysates was extracted and E-cadherin protein (E-Cad) was detected by Western blot. β-Actin was used as a loading control. C, BG-1 cells were transfected with the human E-cadherin promoter luciferase construct and pSV-βGal in combination with either one of the siRNAs. Eight hours after transfection, cells were incubated with vehicle (Cont) or 10⁻⁷ m E2 for an additional 48 h. The luciferase activities were calculated relative to the promoterless vector (pGL3-Basic) and expressed as fold change relative to vehicle control. The data are shown as mean ± sd of three repeated experiments. D, Scratch was made by scraping the monolayer 24 h after transfection of the siRNAs, and cells were incubated with medium containing vehicle (Cont) or 10⁻⁷ m E2. Wound closure was photographed after 24 h of cell migration. The amount of wound repair was expressed as uncovered area at the indicated time compared with initial uncovered area of vehicle-treated control at time zero. Values are the mean ± sd of three separate experiments. *, P < 0.05 compared with control. Cont, Control.

To further investigate the potential contribution of Snail and Slug in E2-induced cell migration, wound healing assay was performed with their corresponding siRNAs. As shown in Fig. 5D, cells treated with E2 efficiently migrated into the wound compared with control BG-1 cells or cells treated with control siRNA. In contrast, cells transfected with Snail or Slug siRNA reversed E2-induced migration into the wound. Together these findings indicate that Snail and Slug play pivotal roles in mediating EMT induction by E2.

ERα, But Not ERβ, Mediates E2-Induced EMT

To dissect which ER subtype plays a dominant role in the pro-metastatic effect of E2, BG-1 cells were treated with selective ER subtype ligands. The ERα-specific agonist, propyl pyrazole triol (PPT), is 410-fold more potent in binding to ERα than ERβ (35), whereas diarylpropionitrile (DPN) is an ERβ agonist with more than 70-fold higher binding affinity for ERβ than ERα (36). BG-1 cells were treated with either 10⁻⁷/10⁻⁸ m PPT or 10⁻⁷/10⁻⁸ m DPN for 48 h and harvested for E-cadherin detection by Western blot. PPT induced a comparable reduction in E-cadherin protein expression as E2 (Fig. 6A). In contrast, DPN induced an approximately 50% increase in the expression. Next, we confirmed this result by abrogating endogenous ERs in the cells. The efficiency of the siRNAs was confirmed by Western blot (supplemental Fig. S2A). Selective knockout of ERα dramatically reversed the down-regulation of E-cadherin expression and its promoter activity upon E2 administration (Fig. 6, B and C). Neither nonspecific siRNA nor ERβ-specific siRNA had any effect. Consistently, depletion of ERα also significantly abolished the cell morphological changes and cell migration induced by E2 (Fig. 6, D and E).

Fig. 6.

ERα Is Required for EMT Induction by E2

A, BG-1 cells were treated with either vehicle (Cont), 10⁻⁷ m E2, 10⁻⁷/10⁻⁸ m DPN, or PPT for 48 h. B, BG-1 cells were transfected with nonspecific siRNA (NS siRNA), ERα siRNA, or ERβ siRNA and subsequently treated with 10⁻⁷ m E2 for 48 h. E-cadherin protein (E-Cad) and β-actin were detected by Western blot. C, Cells were transfected with E-cadherin promoter construct and pSV-βGal in combination with the indicated siRNAs. Eight hours after transfection, cells were incubated with vehicle (Cont) or 10⁻⁷ m E2 for an additional 48 h. The luciferase activities were calculated relative to the promoterless vector (pGL3-Basic) and expressed as fold change relative to vehicle control. D, Morphological changes of cells were observed by phase-contrast microscope and representative photographs are shown. Original magnification, ×200. E, Scratch was made by scraping the monolayer 24 h after siRNA transfection, and cells were incubated with medium containing vehicle (Cont) or 10⁻⁷ m E2. Wound closure was photographed after 24 h. The amount of wound repair was expressed as uncovered area compared with initial uncovered area of vehicle-treated control at time zero. Values are the mean ± sd of three separate experiments. F, SKOV-3 cells were transfected with pCMV5-ERα or without plasmid DNA (mock) for 8 h and subsequently treated with 10⁻⁷ m E2 or vehicle (Cont) alone for another 48 h. E-cadherin (E-Cad), Snail, Slug, and β-actin were detected by Western blot. G, Morphological changes of SKOV-3 cells under the same treatment were also observed by phase-contrast microscope, and representative photographs are shown. Original magnification, ×200. The experiments were done in triplicate and repeated thrice. *, P < 0.05 compared with control. Numerical values below each lane of the immunoblots represent quantification of the relative protein level by densitometry (normalized to β-actin protein level). Cont, Control.

To further address the involvement of the ER subtypes, we restored ERα in SKOV-3 cells by transient transfection of full-length ERα (pCMV5-ERα). This ovarian cancer cell line is ERβ positive but ERα negative due to inactivating mutation, which renders the cells insensitive to estrogen in terms of cell proliferation and gene induction (24, 37). Overexpression of ERα was confirmed by Western blot, and it was comparable to that detected in ERα-expressing BG-1, indicating that the expression was within normal levels (supplemental Fig. S2B). The receptor functionality was verified by profound activation of an estrogen-responsive reporter (thymidine kinase-ERE-Luc) in the presence of E2 or PPT (data not shown). Mock-transfected cells were used as control. As shown, untransfected or mock-transfected SKOV-3 cells showed significant elevation of E-cadherin protein expression in response to E2 (Fig. 6F). In marked contrast, in ERα-overexpressing SKOV-3 cells, E2 decreased E-cadherin expression with parallel increases in both Snail and Slug expressions (Fig. 6F). In line with this, we also observed pronounced morphological changes consistent with EMT (Fig. 6G). These data, together with those obtained from BG-1 cells (Fig. 6, A–E), further illustrate that the acquisition of metastatic properties is ERα dependent.

ERβ Opposes the Pro-Metastasis Effects of ERα

The findings in Fig. 6, A and F, suggest that ERβ may have an opposite effect to ERα. To further examine the role of ERβ in the metastatic activities of ovarian cancer cells, we used siRNA to knock down ERβ in SKOV-3 cells. Figure 7A shows that ERβ siRNA specifically inhibited the expression of ERβ, but not ERα. More importantly, ERβ siRNA abolished the up-regulation of E-cadherin by E2 (Fig. 7B). The E2-mediated increase of cell-free area could also be efficiently blocked by ERβ siRNA (Fig. 7C).

Fig. 7.

ERβ Up-Regulates E-Cadherin and Inhibits Cell Migration

A, SKOV-3 cells were transfected with nonspecific siRNA (NS siRNA), ERα siRNA or ERβ siRNA. Efficiencies of the siRNA were confirmed by Western blot with ERβ antibody. B, SKOV-3 cells were transfected with the siRNAs and subsequently treated with 10⁻⁷ m E2 for 48 h. E-cadherin (E-Cad) and β-actin were detected by Western blot. C, Scratch was made by scraping the monolayer 24 h after siRNA transfection, and cells were incubated with medium containing vehicle (Cont) or 10⁻⁷ m E2. Wound closure was photographed after 24 h. The amount of wound repair was expressed as uncovered area compared with initial uncovered area of vehicle-treated control at time zero. *, P < 0.05 compared with control. Numerical values below each lane of the immunoblots represent quantification of the relative protein level by densitometry (normalized to β-actin protein level). Cont, Control.

As an alternative approach, we introduced ERβ (pRST7-ERβ) into BG-1 cells by transient transfection. The overexpression of ERβ was verified by Western blotting (Fig. 8A). Figure 8, B and C, shows that the expression of ERβ significantly inhibited the E2-mediated down-regulation of E-cadherin protein and gene transcription. As expected, reduced Snail and Slug protein expressions could be observed in the ERβ-transfected cells (Fig. 8B). ERβ overexpression also inhibited the EMT and migration induced by E2 (Fig. 8, D and E). These results suggest that ERβ may function as a negative modulator of ERα in ovarian cancer cell metastasis.

Fig. 8.

ERβ Opposes ERα-Induced EMT

A, BG-1 cells were transfected with pRST7-ERβ or without plasmid DNA (mock). Cell lysates were extracted for ERβ detection by Western blot. B, BG-1 cells were transfected for 8 h and subsequently treated with 10⁻⁷ m E2 or vehicle (Cont) for an additional 48 h. E-cadherin (E-Cad), Snail, Slug, and β-actin were detected by Western blot. C, Cells were transfected with pRST7-ERβ, E-cadherin promoter construct, and pSV-βGal. Eight hours after transfection, cells were incubated with vehicle (Cont) or 10⁻⁷ m E2 for an additional 48 h. The luciferase activities were calculated relative to the promoterless vector (pGL3-Basic) and expressed as fold change relative to vehicle control. D, Morphological changes of BG-1 cells transfected with pRST7-ERβ or without plasmid DNA (mock) were also observed by phase-contrast microscope, and representative photographs are shown. Original magnification, ×200. E, Scratch was made by scraping the monolayer 24 h after transfection, and cells were incubated with medium containing vehicle (Cont) or 10⁻⁷ m E2. Wound closure was photographed after 24 h. The amount of wound repair was expressed as uncovered area compared with initial uncovered area of vehicle-treated control at time zero. Values are the mean ± sd of three separate experiments. *, P < 0.05 compared with control. Numerical values below each lane of the immunoblots represent quantification of the relative protein level by densitometry (normalized to β-actin protein level). Cont, Control.

DISCUSSION

The role of estrogen in ovarian tumorigenesis remains unclear and controversial. However, compelling evidence associates estrogen with increased ovarian cancer incidence (27, 28, 31). Clinical and epidemiological studies have shown that long-term use of estrogen replacement therapy increases the risk of ovarian cancer (38). Breastfeeding and use of oral contraceptives, which appear protective in a number of studies, are associated with reduced serum concentration of estradiol (22, 23). Experimental results also implicate a role of estrogen in ovarian carcinogenesis. Cell survival and proliferation could be enhanced by estrogen both in vivo and in vitro (25, 26, 29, 30). Given the importance of estrogen demonstrated by these studies, understanding the mechanistic basis through which it controls ovarian tumorigenesis will have profound biological and medical implications. In fact, the proliferative effects of estrogen were found to be mediated through regulation of genes involved in growth control, such as c-myc, bcl-2 and ezrin (25, 30, 39). In contrast to the role in regulating cell proliferation, very little is known about whether and how estrogen promotes cell metastasis. Moreover, previous data on the effects of cell invasiveness and motility by estrogen were inconsistent (30, 40). In the present study, our findings extend the understanding of the involvement of estrogen in EMT and regulation of EMT-associated proteins. We provided supporting evidence, to our knowledge for the first time, that E2 can induce EMT of ovarian cancer cells. We also demonstrated that E2 enhanced cellmigratory ability through transcriptional regulation of E-cadherin and its repressors, Snail and Slug. Moreover, our results clearly established ERα as an essential mechanistic link in mediating the induction of EMT by E2. Of importance, our data indicated that ERβ counteracts ERα signaling in mediating the metastatic responses to E2.

EMT is a prerequisite for tumor progression by which epithelial cancer cells modulate their phenotypes and acquire metastatic capabilities. The E2-induced phenotypic changes demonstrated in our data are consistent with changes seen during EMT (2). Changes in cell morphology, from a cobblestone-like appearance to more elongated shape, cell dissociation with reduced cell-cell contacts, and pseudopodia formation were evident. Exactly how E2 promotes EMT and cell motility remains to be elucidated. In this regard, our data implied that the E2-induced EMT is, at least in part, mediated through repression of E-cadherin. Reduced expression of E-cadherin has been found in advanced-stage ovarian tumors and ectopic expression of E-cadherin inhibited cell invasion in vitro (41, 42, 43). Interestingly, although unlike other types of epithelial malignancy, E-cadherin expression is frequently acquired in primary ovarian carcinomas for epithelial differentiation, it is repressed in late-stage tumors for metastasis (44, 45). However, the cause for this subsequent loss of expression remains unclear. In ovarian cancer cells, E-cadherin gene mutation is infrequent (11). Rather, gene methylation and up-regulation of E-cadherin transcriptional repressors have been reported (46, 47). From our present data, estrogen might represent one of the mechanisms that silence E-cadherin in ERα-positive ovarian cancer cells. In agreement with this, previous studies also suggest a role for estrogen in regulating E-cadherin expression, but the underlying mechanism has not been fully investigated (48, 49). Because numerous transcription repressors have been characterized in controlling E-cadherin transcription, it is likely that the repressors involved are dependent on cell type and cell context. Our data clearly revealed that two transcription factors, Snail and Slug, act in concert in the suppression of E-cadherin promoter activity by E2.

The Snail gene family of transcription factors is best known for its ability to trigger EMT through repression of E-cadherin (13, 14, 50). Snail and Slug are also involved in the regulation of tight junction proteins (51, 52), matrix metalloproteinases (53, 54), and cytokeratins (55). High expression of Snail and Slug is often found in metastatic ovarian tumor cells compared with primary tumors (56). Consistent with these observations, we demonstrated the importance of Snail and Slug in E2-induced cell migration by inhibiting transcription of E-cadherin in ovarian cancer cells. The identification of Snail and Slug as novel estrogen-responsive genes in this report helps to confirm the role of estrogen in ovarian carcinogenesis. According to Elloul et al. (56), it has been suggested that Snail is regulated at the posttranscriptional level in ovarian carcinoma, whereas Slug is probably regulated transcriptionally. Here we found that the up-regulation of both Snail and Slug by E2 was predominantly mediated through activation of their promoter activities. The ERs elicit gene transcription by interacting with the classical estrogen response element (ERE) or non-ERE elements that bind heterologous transcription factors, including activator protein 1 sites (57), Sp1 sites (58) and cAMP-response elements (59). Of particular note, putative ERE sites can be found in the Snail promoter region (60), but not in the Slug promoter. By sequence homology search, we found several potential Sp-1, activator protein 1, and cAMP-response element sites within the 5′-flanking regions of Snail and Slug. In addition, ER might also be recruited to the 3′-enhancer downstream of the coding region, constituting a combinatorial network for transcription regulation (61). Whether E2 activates the two promoters through 5′-ERE, non-ERE sites, or 3′-regulatory region has yet to be determined.

In metastatic ovarian cancer cells, only ERα is present and ERβ expression is often lost (62, 63, 64, 65). This implicates that the ER subtypes may play distinct roles in ovarian carcinogenesis. However, their functions and the molecular mechanisms in ovarian carcinogenesis have just begun to be unveiled. ERα promotes ovarian cancer cell growth, and several potential ERα-regulated genes have been identified by cDNA microarray (66). These genes include junctional and extracellular matrix proteins (e.g. cadherin 6, fibronectin, and vimentin), metastasis-related proteinases (e.g. matrix metalloproteineses and urokinase-type plasminogen activator), and regulators of cell cycle (e.g. cyclin B1) (66). Consistently, it has been shown in other cancer types that ERα promotes carcinogenesis through activation of genes linked to cell proliferation and invasion, such as cyclin D1 (67), c-myc (68), and fibulin-1 C (69). In contrast, overexpression of ERβ exerts a protective role against cancer development by inducing apoptosis and inhibiting migration/invasion (68, 70, 71). Therefore, overexpression of ERα may represent a more metastatic behavior of cancer cells. Accordingly, we provided several lines of evidence that induction of EMT by E2 is mediated specifically by ERα, but not by ERβ. First, ERα is predominant whereas ERβ is barely detectable in BG-1 cells (our unpublished data and Ref. 65), implicating ERα as the major mediator of estrogen signaling in the cells. Second, the involvement of ERα was confirmed by pharmacological agonists and siRNAs. Third, the transforming capabilities of E2 could be observed in SKOV-3 cells with ERα restored by overexpression. This cell line is ERα negative but ERβ positive due to mutation in the ERα transcript and is insensitive to estrogen with respect to cell growth (24, 37). In contrast to our observations, a previous study suggested an ERα-independent pathway in regulation of E-cadherin and Snail by E2 in ovarian cancer cells (48). There is no definite explanation for this discrepancy yet, and more extensive experiments are needed to confirm the mechanisms underlying E2 effects in those cells. It is also possible that, in some cell types, regulation of these genes by E2 might depend on an alternative pathway. ER-independent, nongenomic signaling of E2 has been proposed to occur through membrane-localized G protein-coupled receptor GPR30 (72, 73). Whether this signaling cascade is involved warrants further investigation. Intriguingly, another remarkable finding in this study is the demonstration of the tumor suppressor activity of ERβ by counteracting ERα signaling. ERβ induced E-cadherin expression and inhibited migration of the ERα-negative SKOV-3 cells. Moreover, introduction of ERβ potently blocked the pro-metastatic effects of E2 in BG-1 cells. It is interesting to note that ERβ negatively regulates the decrease of E-cadherin by ERα, and this counteracting effect of ERβ was mediated at the promoter level. Indeed, it has been proposed that modulation of ERα-targeted genes by ERβ may contribute to the contrasting actions between the two ER subtypes (67, 74). Our data also suggest that this might be the case in our paradigm because ERα is responsible for E2-induced Snail and Slug expression, which could be reversed by ERβ. The molecular mechanism underlying this differential regulation by each ER subtype is currently under investigation in our laboratory.

Given the indispensable role of EMT in promoting progression of many carcinomas, it is important to unravel the mechanisms that govern this process. Our findings herein showed that E2 can enhance ovarian tumor cell motility by triggering EMT through ERα. The induction of cell transformation in vitro and the identification of E-cadherin, Snail, and Slug as novel transcriptional targets of estrogen signaling confirm the carcinogenicity of the hormone. Our discovery that ERα and ERβ differentially mediate the EMT process provides a possible explanation for the observed increase in the ERα:ERβ expression ratio during ovarian carcinogenesis. It also carries clinical implications for selective targeting of the ERs in therapeutic and prevention strategies against ovarian cancer.

MATERIALS AND METHODS

Cell Culture

The human ovarian adenocarcinoma cell line BG-1 was kindly provided by Dr. K. S. Korach (National Institute of Environmental Health Sciences, NIH, Research Triangle Park, NC) (75). The SKOV-3 ovarian adenocarcinoma cell line was obtained from American Type Culture Collection (Manassas, VA). The cells were maintained in DMEM/F12 medium supplemented with 10% fetal bovine serum, 100 U/ml penicillin, and 100 mg/ml streptomycin (Life Technologies, Inc., Gaithersburg, MD) in a humidified atmosphere of 5% CO₂ at 37 C.

Treatments

Cells were stripped from endogenous steroids by changing the medium to phenol red-free DMEM/F12 containing 10% charcoal-dextran stripped fetal bovine serum (HyClone Laboratories, Inc., Logan, UT) 48 h before treatment. Cells were then incubated in fresh medium with E2 (Sigma, St. Louis, MO), ICI (ER antagonist) (Tocris, Ballwin, MO), alone or in combination for indicated times. In some cases, cells were treated with either ERα agonist PPT or ERβ agonist DPN for 48 h (Tocris). Control cultures received the same volume of ethanol. To study mRNA stability, cells were treated with ethanol or E2 for 1 h. Actinomycin D (5 μg/ml) (Sigma) was then added to the cultures, and total RNA was prepared at the times indicated for up to 8 h. All treatments were performed in duplicate or triplicate in each experiment.

Transient Transfection

Snail- and Slug-specific siRNA oligos (predesignated siRNA) were purchased from QIAGEN (Valencia, CA). Sense sequence of Snail no. 1: 5′-GCG AGC UGC AGG ACU CUA A-3′; Snail no. 2: 5′-GGU GUG ACU AAC UAU GCA A-3′; Slug no. 1: 5′-GGA CCA CAG UGG CUC AGA A-3′; and Slug no. 2: 5′-CUC CGA AGC CAA AUG ACA A-3′. ERα- and ERβ-specific siRNAs were purchased from Invitrogen (Burlington, Ontario, Canada). Sense sequence of ERα: 5′-GGG CUC UAC UUC AUC GCA U-3′; ERβ: 5′-GCA GAC CAC AAG CCC AAA U-3′. Cells cultured in six-well plates were transfected with siRNA duplexes using Lipofectamine 2000 (Invitrogen). Nonspecific siRNA duplexes were used as negative controls (QIAGEN). Cells were treated for 48 h to allow maximum knockdown.

Full-length human ERα (pCMV5-ERα) and ERβ (pRST7-ERβ) expression plasmids were kindly provided by Dr. B. S. Katzenellenbogen (Department of Molecular and Integrative Physiology, University of Illinois at Urbana Champaign, IL) and Dr. D. P. McDonnell (Department of Pharmacology and Cancer Biology, Duke University Medical Center, Durham, NC), respectively.

Reverse Transcription and Real-Time Quantitative PCR

Total RNA was isolated using TRIzol reagent (Invitrogen). Single-stranded cDNA was synthesized from 2 μg of total RNA (Amersham Biosciences, Baie d'Urfé, Quebec, Canada). Quantitative real-time PCR was carried out using ABI Prism 7300 Sequence Detector System (Applied Biosystems, Foster City, CA) and SYBR green PCR master mix (Applied Biosystems). Gene-specific primers were designed using Applied Biosystems’ Primer Express 3.0 software. The following oligonucleotides (Invitrogen) were used for PCR amplification: E-cadherin, sense, 5′-ACA GCC CCG CCT TAT GAT T-3′; and antisense, 5′-TCG GAA CCG CTT CCT TCA-3′; Snail, sense, 5′-CCC CAA TCG GAA GCC TAA CT-3′; and antisense, 5′-GCT GGA AGG TAA ACT CTG GAT TAG A-3′; and Slug, sense, 5′-TTC GGA CCC ACA CAT TAC CT-3′; and antisense, 5′-GCA GTG AGG GCA AGA AAA AG-3′. The relative mRNA level was normalized by glyceraldehyde-3-phosphate dehydrogenase (GAPDH) levels with the following specific primers: 5′-CCT CCC GCT TCG CTC TCT-3′ (forward), 5′-TGG CGA CGC AAA AGA AGA T-3′ (reversed). The thermal profile consisted of 2 min at 50 C, 10 min at 95 C, followed by 40 cycles of 15 sec at 95 C and 1 min at 60 C. The melting temperature profiles of amplicons were determined to show the specificity of amplification.

Immunoblot Analysis

Cells were harvested in lysis buffer (1% Triton-X, 0.1% sodium dodecyl sulfate, 0.5% sodium deoxycholate) with protease inhibitors cocktail (Sigma) and protein concentrations determined by the Bradford protein assay (Bio-Rad Laboratories, Inc., Hercules, CA). Protein (20 μg) was subjected to SDS-PAGE followed by electrotransfer onto nitrocellulose membrane. Membranes were incubated overnight at 4 C with mouse monoclonal human E-cadherin antibody (1:2000) (BD Transduction Laboratories, Heidelberg, Germany), human Snail antibody (1:1000) (Abgent, San Diego, CA), human Slug antibody (1:1000) (Abgent), human ERα antibody (1:1000), and ERβ antibody (1:1000) (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). Blots were reprobed with anti-β-actin antibody (1:2000) (Santa Cruz Biotechnology) to confirm equal loading. The immunocomplex was detected with appropriate horseradish peroxidase-conjugated secondary antibodies (Amersham Biosciences) and visualized using enhanced chemiluminescence detection system (ECL) (Amersham Biosciences). All blots were repeated in at least three different experiments. Quantitative densitometry was performed using Scion Image software (Scion Corp., Frederick, MD).

Reporter Gene Assay

Luciferase reporter vector containing human E-cadherin promoter (−178 to +66) was kindly provided by Dr. Yoshito Ihara (Nagasaki University, Japan) (76). Plasmid containing the human Slug promoter (−3371/+1) was a kind gift of Dr. Kaname Kawajiri (Research Institute for Clinical Oncology, Saitama, Japan) (77). The human Snail promoter reporter plasmid (−869/+59) was a generous gift of Dr. Antonio Garcia de Herreros (Universitat Pompeu Fabra, Barcelona, Spain) (78). All plasmids were cloned in the promoterless pGL3-Basic vector. Cells were transfected with 1 μg of each firefly luciferase reporter construct together with 0.5 μg pSV-βGal plasmid as normalization reference for transfection efficiency. Transfections were performed with Lipofectamine 2000 (Invitrogen). Cells were treated with E2, ICI, or vehicle 6 h after transfection. Where indicated, cells were also cotransfected with Snail or Slug siRNA. Cells transfected with promoterless vector (pGL3-Basic) were used as control for background luciferase activity. Cells were then harvested in reporter lysis buffer (Promega Corp., Madison, WI) at designated time points. Luciferase activity and β-galactosidase enzyme activity were determined with luciferase reagent and the β-galactosidase enzyme assay system (Promega). Luciferase units were calculated as luciferase activity/β-galactosidase activity and are presented as the mean ± sd of three individual experiments with triplication. The fold change was calculated by comparison with the promoterless pGL3-Basic.

Scratch Assay

Cells were grown to confluence and formed a monolayer covering the surface of the entire plate. Cells were kept in medium with appropriate treatments for 24 h before scratching. A linear scratch (∼0.7 mm wide) was made gently with sterile pipette tip across the diameter of the well and rinsed with PBS to remove debris. Fresh media for continued drug treatment were then added. For each well, at least six pictures were taken with microscope at a magnification of ×10 at three time points, 0, 4, and 24 h after scratch. The percentage of nonrecovered wound area was calculated by dividing the nonrecovered area after treatments by the initial wound area at time zero.

Statistics

All assays were repeated at least three times in duplicate or triplicate, and the graphic data were presented as the mean ± sd. Statistical analysis was performed where appropriate using the Student’s t test or ANOVA followed by Tukey’s post hoc test (GraphPad Software, San Diego, CA). Differences with P < 0.05 were considered to be statistically significant.

NURSA Molecule Pages:

• Ligands: 17β-Estradiol | Fulvestrant;

• Nuclear Receptors: ERα | ERβ.