Repression of the Estrogen Receptor-α Transcriptional Activity by the Rho/Megakaryoblastic Leukemia 1 Signaling Pathway*

Although involved in processes leading to the emergence and development of hormone-dependent breast cancers, the estrogen receptor α (ERα) also prevents transformed cells from progressing toward a more aggressive phenotype. The transcriptional activity of ERα is mediated through two transactivation functions, called activation function 1 and 2, whose respective involvement varies in a cell-specific manner. Here, we identify the Rho/megakaryoblastic leukemia 1 (MKL1) signaling pathway as a main actor in controlling the cell-specific activity of both transactivation functions of ERα. Notably, we show that, when the coregulator MKL1 is sequestered in an inactive form by unpolymerized actin, the transcriptional activity of ERα mainly relies on the activation function 1. The activation of MKL1, which results from its dissociation from unpolymerized actin, promoted by the ability of Rho to support polymeric actin accumulation, silences the activation function 1 of ERα and allows the receptor to mainly act through its activation function 2. Importantly, this switch in the respective contribution exerted by both transactivation functions is correlated with an impaired ability of ERα to efficiently transactivate estrogen-regulated reporter genes. MKL1 is further shown to be present on estrogen-responsive genes in vivo. Interestingly, the Rho/MKL1 signaling pathway is activated during the epithelial-mesenchymal transition. A reduced transactivation efficiency of ERα, resulting from the activation of this pathway, may therefore suppress the protective role exerted by ERα toward tumor progression and invasiveness.

The estrogen receptor ␣ (ER␣) 2 belongs to a class of structurally conserved transcription factors, the nuclear receptor superfamily. ER␣ regulates the transcription of specific genes upon binding to regulatory sequences, either directly through the association of its DNA-binding domain to specific sequences called estrogen-responsive elements (EREs), or indirectly through protein-protein interactions with activator protein 1 or Sp1 factors (1). ER␣ comprises an N-terminal transcriptional activation function, the activation function 1 (AF1), and a ligand-dependent transactivation function, the AF2, which is part of the C-terminal ligand-binding domain. Hormone binding to ER␣ triggers conformational changes that make these AFs available for interaction with transcriptional coregulators that, in turn, allow the regulation of target genes transcriptional activity. Several subclasses of NR coactivators have been identified, including members of the p160 family, the integrators CBP and p300, arginine methyltransferases, or components of the Mediator complex (2,3).
Mediating estrogenic signaling, ER␣ plays critical roles in many physiological processes such as the development and function of the reproductive system, bone metabolism, and cardiovascular activity (4). It is also associated with major human pathologies, and notably estrogen-dependent cancers developed from reproductive organs. For instance, ER␣ mediates the mitogenic effects that estradiol (E 2 ) exerts on ϳ70% of primary breast tumors. However, contrasting with this harmful promotion of tumor growth during initial stages of carcinogenesis, ER␣ also has protective effects. ER␣-positive breast tumors are indeed generally more differentiated and less invasive than ER␣-negative cancers and, therefore, display a better prognosis (5). Unfortunately, ER␣-positive tumors frequently acquire a resistance to steroid hormones-positive influence, which is accompanied with an amplification of growth factor receptor signaling. Notably, the overexpression of the human epidermal growth factor receptor HER2 in ER␣-positive breast cancer cells results in a down-regulation of ER␣ transcriptional activity and expression, both in vitro and in vivo (6). The molecular basis of the loss of sensitivity of ER␣-positive breast cancer cells to estrogens still remains elusive.
The relative contribution exerted by AF1 and AF2 on the transcriptional activity of ER␣ varies depending upon the promoter, but also upon the cell considered (7,8). For instance, AF2 is the only active AF in cells that have achieved their epithelial-mesenchymal transition (9). Importantly, the loss of E-cadherin-mediated cell-cell junctions, a key process that occurs during the epithelial-mesenchymal transition, dramati-cally affects ER␣ functions by attenuating its transactivation efficiency. This occurs through the silencing of the activity of an AF1 subregion termed box 1, which forces the receptor to mainly act only via its AF2 (10). Disruption of cadherin-mediated intercellular adhesion, migration, and metastasis has been correlated with high activities of RhoA or of its target Rho kinase (or ROCK) (11). According to recent work, RhoA signaling influences the activity of the myocardin-related transcription factor MKL1 (megakaryoblastic leukemia 1, also termed MAL, BSAC, or MRTF-A), which is a coactivator of the serum response factor (SRF) (12,13). The activity of MKL1 depends upon the integrity of epithelial cell junctions (14,15) and is regulated by changes in actin dynamics. In serum-starved cells, MKL1 binds to unpolymerized actin ("globular" actin or G-actin) through its RPEL motifs and is thus sequestered in an inactive form. Upon Rho activation, accumulation of fibrillar actin ("fibrillar" actin or F-actin) leads to a G-actin depletion and release of MKL1, which then coactivate SRF transcriptional activities (12,16).
We demonstrate in this report that the activation of the Rho/ Actin/MKL1 signaling pathway impairs ER␣-dependent transcriptional activation, mainly through the inhibition of the activity of the AF1 box 1. Our data thus provide important and novel insights into how actin-dependent processes such as cell shape changes or adhesion may deeply influence estrogenic signaling.
Cell Culture and Transfection-HeLa, HepG2, MCF-7, or MDA-MB231 cells or MDA-MB231 stably expressing ER␣ (19) were maintained in Dulbecco's modified Eagle's medium (Sigma-Aldrich) supplemented with 5% or 10% (MCF-7) fetal calf serum (Biowest) and antibiotics (Sigma-Aldrich) at 37°C in 5% CO 2 . Transfections were carried out using FuGENE 6 TM reagent according to manufacturer's instructions (Roche Applied Science). One day before transfection, cells were plated in 24-well plates at 50% confluence. 1 h prior transfection, the medium was replaced with phenol red-free Dulbecco's modified Eagle's medium-F12 (Sigma-Aldrich) containing 2.5% charcoal-stripped fetal calf serum (Biowest). Transfection was carried out with 100 ng of reporter gene, 100 ng of CMV-␤Gal internal control, and appropriate combinations of expression vectors. Plasmid mix was made up to 500 ng of total DNA per well with empty vector. Following an incubation overnight, cells were treated for 24 h with ligands or ethanol (vehicle control), and required drugs. The C3 exoenzyme (125 ng/ml) was added to the FuGENE/medium mix during the transfection step. Cells were then harvested, and luciferase and ␤-galactosidase assays were performed as previously described (18). Luciferase reporter gene activity was normalized to the ␤-galactosidase.
Pulldown Assays-To determine Rho activity in cell lines, the Rho Activation Assay Kit (Upstate) was used in accordance with manufacturer's instructions.
Reverse Transcription-PCR, ChIP, and Sequential ChIP Assays-Reverse transcription-PCR, ChIP, and sequential ChIP assays were performed as previously described (21). The primers against promoter regions are available upon request.

Rho GTPases and Actin Dynamics Control ER␣ Transactivation
Potency-We previously demonstrated that the transcriptional activity of ER␣ depends upon the differentiation stage of the cell, with the highest activity detected in differentiated cells (9). De-differentiation of epithelial cells is correlated with significant changes in the activity of members of the Rho GTPases family such as RhoA, Rac1, and Cdc42 (11). The impact of the Rho/actin signaling pathway on ER␣ transactivation efficiency was analyzed by transient transfection experiments in HepG2 and HeLa cells, which are two epithelial-like cell lines exhibiting divergent phenotypes. HeLa cells, originating from a cervix carcinoma, present a poorly differentiated phenotype, do not express the calcium-dependent cell-cell adhesion molecule E-cadherin, and strongly express vimentin, an intermediate filament protein that expression is associated with increased invasive and metastatic potency. In contrast, the hepatocarcinoma HepG2 cell line appears more differentiated, producing a high level of E-cadherin but no vimentin (9). Moreover, Rho activity was much higher in HeLa than in HepG2 cell extracts, as indicated by glutathione S-transferase pulldown experiments using as a bait the Rho-binding domain of the human rhotekin protein, which has the ability to retain only active Rho (Fig. 1A). Although no signal was observed within the HepG2 extracts, we cannot exclude that some active Rho is present in these cells but at really low levels that could not be detected by the procedure. The transactivation efficiency of ER␣ on an ERE-tk-LUC reporter gene is also much higher in HepG2 than in HeLa cells following treatment with 10 Ϫ8 M E 2 (Fig. 1B), corroborating our previous data (9,10).
Constitutively active Myc-tagged forms of RhoA (Q63L), Rac1 (Q61L), or Cdc42 (Q61L), as well as their respective inactive forms (RhoA T19N, Rac1 T17N, and Cdc42 T17N), were then used to study the contribution of these Rho GTPases in transcriptional regulations mediated by ER␣. In HepG2 cells, the transfection of increasing concentrations of plasmids encoding inactive RhoA, Rac1, and Cdc42 forms, which inhibit endogenous Rho GTPase activity, has little effect on ER␣-mediated induction of the ERE-tk-LUC reporter. In contrast, the expression of these mutated Rho GTPases in HeLa cells increases 2-to 3-fold the transcriptional response of the reporter gene to ER␣ (Fig. 1B). Importantly, in HepG2 cells, although the transfection of low amounts of plasmids encoding the three constitutive Rho GTPase forms has little impact on ER␣ transactivation potency, the latter is severely reduced at higher amounts of plasmids. In contrast, in HeLa cells, the constitutively active forms have no obvious effect on the ER␣-mediated stimulation of ERE-tk-LUC reporter, as expected (Fig.  1B). The use of the C3 exoenzyme, which is an inhibitor of the three Rho proteins (A, B, and C), restores the situation observed with the inactive form of RhoA (T19N) (Fig. 1B). Finally, we analyzed the expression level of ectopically expressed proteins in both cell lines through Western blots (Fig. 1C). These experiments showed no correlation between results from gene reporter assays and any variation in the amounts of ER␣ or Rho GTPase forms. Moreover, the increased amount of ER␣ detected within HepG2 cells transfected with the three constitutive Rho GTPase forms is inversely correlated with its activity. Altogether, these results indicate that ER␣ transcriptional efficiency is negatively regulated by active Rho GTPases.
Rho GTPases are mainly studied for their roles in regulating the actin cytoskeleton (22). Activation of RhoA induces the polymerization of actin, partly through its effector ROCK (23). We consequently evaluated the transcriptional activity of ER␣ in both cell lines following an inhibition of ROCK by the Y-27632 compound. In agreement with data presented before, the Y-27632 treatment enhances ER␣ transcriptional activity in HeLa cells only (Fig. 1B). Contribution of actin dynamics in ER␣ transcriptional regulation was then further evaluated by disrupting the intracellular levels of actin monomers (G-actin). For this purpose, cells were transfected with increasing concentrations of expression vectors encoding wild-type or unpolymerizable mutant (G13R) of ␤-actin, the overexpression of which enhances the G-actin level (24). When expressed in HeLa cells, both actin forms potentiate the induction of the ER␣-dependent reporter gene. In contrast, ER␣ activity in HepG2 cells was not really modified in their presence (Fig. 1B). Western blots indicate that ␤-actin levels are not obviously affected by the transfections of both actin forms in both cell lines (Fig. 1C). In contrast, ER␣ expression level is slightly reduced by the overexpression of the wild-type ␤-actin form, while it remains active. Together, these results indicate that actin dynamics influence ER␣ activity. To substantiate this observation, we next used cytochalasin D, an actin-binding drug that mimics the coactivation of SRF by Rho (25). As expected, cytochalasin D strongly inhibits ER␣ transcriptional activity in HepG2 cells, thereby mimicking the situation observed upon the overexpression of constitutively active forms of Rho GTPases (Fig. 1B). Altogether, these results strongly suggest that Rho GTPases control ER␣ transcriptional activity by modulating actin dynamics.
Rho/Actin Regulation of ER␣ Transactivation Potency Involves ER␣ AF1 Box 1 Activity-In epithelial cells, the attenuation of ER␣ activity observed following the loss of E-cadherin-mediated cell contacts partly results from a silencing of the ER␣ AF1 box 1 activity (10). We therefore evaluated the respective contribution exerted by ER␣ AFs toward the transcriptional activity of the receptor, upon changes of the Rho/ actin pathway activity. Cell permissiveness to either ER␣ AFs was determined by comparing the transcriptional activity of ER␣ with those of ER␣ ⌬79 (deletion of AF1 box 1) and ER␣ CF (additional deletion of AF1 box 2/3). Importantly, the ERE-tk promoter mainly used in our transient transfection experi-ments exhibits no intrinsic preference for a specific AF (9). In HepG2 cells, the main region involved in ER␣ transcriptional activity is the AF1 box 1 (ER␣ ⌬79 versus ER␣, Fig. 2B), the remaining activity depending upon the AF1 box 2/3, as expected (10). In contrast, the AF2 (ER␣ CF) represents ϳ60% of the whole ER␣ transcriptional potency in HeLa cells. The AF1 box 2/3 (ER␣ ⌬79 versus ER␣ CF) is responsible for the remaining 40% of the activity (Fig. 2B).
The overexpression of the constitutively active form of RhoA (Q63L) specifically abolishes the activity of ER␣ AF1 box 1 in HepG2 cells (ER␣ versus ER␣ ⌬79) without significantly affecting the activities of the other AFs. In contrast, in HeLa cells, RhoA Q63L does not change the relative contribution exerted by the different AFs, as indicated by the similar activity promoted by the ER␣, ER␣ ⌬79, and ER␣ CF proteins (Fig. 2B). Thus, in the presence of a constitutively active form of RhoA, the transcriptional activity of ER␣ is similar in both cell types. As expected, the overexpression of the T19N-inactive form of RhoA had no effect on the transcriptional properties of the different ER␣ proteins in an AF1 box 1 permissive context such as HepG2 cells (Fig. 2B). As a consequence, the overexpression of this mutant in HeLa cells would have been expected to allow an activity of the AF1 box 1, but this is not the case (Fig. 2B). Similar results were observed following the overexpression of dominant positive and negative forms of Rac1 and Cdc42 (data not shown). To confirm the involvement of ER␣ AF1 in the regulation of ER␣ transactivation potency by the Rho/Actin pathway, we then evaluated the relative contribution of the different AFs toward ER␣ activity following the inhibition of the RhoA effector ROCK by the Y-27632. Importantly, while this treatment had no effect in HepG2 cells, it allowed HeLa cells to become permissive to an AF1 box 1 activity, as revealed by the higher transcriptional activity of full-length ER␣ versus ER␣ ⌬79 (Fig. 2B). Importantly, this treatment did not affect the contribution of AF1 box 2/3 in HeLa cells, indicating that pathways initiated by RhoA were targeting the AF1 box 1 activity. Corroborating this conclusion, the overexpression of unpolymerizable ␤-actin (G13R) activated the AF1 box 1 activity in HeLa cells, as revealed by the different transcriptional potency of full-length ER␣ and ER␣ ⌬79 proteins. In contrast, cytochalasin D totally abolished the activity of the AF1 box 1 in HepG2 cells and dictated a strict AF2 permissiveness in HeLa cells. Finally, as assessed by Western blots (Fig. 2C), although some ER␣ variants exhibited different expression levels, they were not affected upon the specific treatments that modify ER␣ activity in both cell lines.
These data demonstrate that Rho/actin axis controls ER␣ transcriptional activity, at least through the modulation of the respective contribution of each AFs toward the activity of the receptor. Notably, the activation of RhoA and subsequent actin polymerization inhibited the activity of the AF1 box 1 of ER␣.
MKL1 Links Rho/Actin Pathway to ER␣ Transcriptional Activity-RhoA and changes in actin dynamic are integral components of a T cell factor-independent pathway leading to the activation of SRF. One mediator of this pathway is the coactivator MKL1 (12,13). The organization of MKL1 is shown in Fig.  3A. MKL1 was expressed in both cell lines (Fig. 3B), therefore, we sought to evaluate the influence of this protein on ER␣ tran-Rho/MKL1 Signaling Pathway Regulates ER␣ Activity scriptional activity. For this purpose, flagged versions of wildtype or deletion mutants of MKL1 were overexpressed in the cells. The MKL1 ⌬N200 protein was deleted from the N-terminal RPEL motifs (Fig. 3A), which are critical for actin binding, and acted as a constitutively active variant on SRF target genes such as c-fos. The MKL1 ⌬C301 protein was deleted from the C-terminal transactivation domain and behaves as a dominant negative form on c-fos regulation (12,17). The correct expression of these proteins, their cellular localization, and their appropriate behavior on a c-fos-LUC reporter gene were first established as shown in Fig. 3, C and D. The amount of the plasmids encoding MKL1 variants used was previously determined by a dose-response experiment (data not shown). Their impact on the ERE-tk-LUC reporter gene was then analyzed. Wild-type and mutated forms of MKL1 dramatically change the basal activity of the ERE-tk-LUC measured in the absence of ER␣, paralleling variations observed on the c-fos-LUC reporter (Fig. 3D). The constitutively active mutant MKL1 ⌬N200, and to a lesser extent MKL1, enhanced the basal activity of the ERE-tk-LUC, whereas the dominant negative form MKL1 ⌬C301 had an opposite effect, notably in HeLa cells (Fig. 3D). Similar results were observed on other ERE-driven reporter genes as well as on ERE-less reporters ( Fig. 3D and data not shown). Thus, besides exhibiting different amplitudes of their response to MKL1 variants, with the ERE-less pTAL reporter gene being the less sensitive, the responses of all reporters were similar, in the presence or absence of an ERE. Finally, the overexpression of MKL1 variants internally deleted of short basic (⌬B), glutamine-rich (⌬Q), or SAP (SAF-A/B, Acinus, and PIAS) (⌬SAP) regions, which are conserved within the myocardin-related transcription factor family (17), also modify the basal activity of the reporter genes in a cell-and promoter-dependent manner ( Fig.  3E and data not shown).
To facilitate the analysis of the influence of MKL1 forms on ER␣ transcriptional activity and minimize the interpretation bias created by the variations in the basal transcriptional activity of the ERE-tk-LUC observed following MKL1 expression, ER␣ transactivation potency were expressed, for each MKL1 forms, as the -fold increase above reporter gene activity measured in the absence of receptor. The results of these assays clearly show that the overexpression of MKL1 and MKL1 ⌬N200 dramatically reduced the transactivation efficiency of ER␣ in HepG2 cells (Fig. 3E), whereas the dominant negative MKL1 ⌬C301 almost doubled the activity of ER␣. The overexpression of MKL1 and MKL1 ⌬N200 proteins has a similar impact on ER␣ transcriptional potency in HeLa cells, but to a lesser extent. Additionally, as in HepG2 cells, MKL1 ⌬C301 also enhanced by 2-fold the transcriptional activity of ER␣ in HeLa cells (Fig. 3E). On the other hand, the internal deletions of MKL1 B, Q, or SAP regions influenced ER␣ activity in a cell-dependent manner (Fig. 3E).  DECEMBER 4, 2009 • VOLUME 284 • NUMBER 49

JOURNAL OF BIOLOGICAL CHEMISTRY 33733
The impact of MKL1 forms on cell permissiveness to either ER␣ AFs was then determined by comparing ER␣, ER␣ ⌬79, and ER␣ CF transcriptional activity. These experiments demonstrate that, in HepG2 cells, MKL1 and MKL1 ⌬N200 dramatically reduced (MKL1) or even abolished (⌬N200) the activity of the AF1 box 1 activity. This is revealed by the decrease of full-length ER␣ activity to levels similar to those induced by ER␣ ⌬79 (Fig. 4A). In HeLa cells, the reduction of ER␣ activity provoked by these two MKL1 proteins was apparently a consequence of an inhibition of the AF1 box 2/3, which generated a cell context permissive only to ER␣ AF2 (Fig. 4A). These results therefore parallel those observed following the overexpression of dominant positive RhoA (Q63L) and cytochalasin D treatments (Fig.  2B). Importantly, the stimulation of ER␣ activity observed following the overexpression of MKL1 ⌬C301 in both cell lines seems to be a consequence of the de-repression of an AF1 box 1 activity in HeLa cells, and of the stimulation of its contribution toward ER␣ activity in HepG2 cells (compare ER␣ with ER␣ ⌬79 in Fig. 4A). Lastly, the overexpression of MKL1 variants internally deleted of B, Q, or SAP regions abolished the AF1 box 1 activity of ER␣ in HepG2 cells (data not shown). In HeLa cells, these deleted forms of MKL1 had a similar impact on ER␣ AF1 box 2/3 activity as full-length MKL1 (data not shown).
To confirm the role exerted by MKL1 in controlling the transactivation potency of ER␣ AF1 box 1, we used another system that relies on the response of the human complement C3 gene (C3) promoter to the agonist activity of 4-OHT. This response indeed depends exclusively upon ER␣ AF1 box 1 activity (9,10,26). Importantly, the overexpression of the dominant negative form of MKL1 (⌬C301) renders HeLa cells permissive to the AF-1 agonistic activity of 4-OHT (Fig.  4B). Correspondingly, the dominant positive MKL1 ⌬N200 abolishes the C3 promoter response to 4-OHT in HepG2 cells (Fig. 4B). Interestingly, upon the overexpression of MKL1 ⌬N200 in HeLa cells, ER␣ bound to 4-OHT represses the reporter gene activity. These results ultimately highlight the role of MKL1 in regulating ER␣ AF1 box 1 activity. Altogether, these results complete the demonstration that the Rho/actin/MKL1 signaling pathway plays a major role in modulating ER␣ transactivation efficiency through the modification of the respective activities of ER␣ AF subregions.
MKL1 Is Present on Estrogen-responsive Promoters in Vivo and Is a Potential ER␣-interacting Protein-We next explored the hypothesis that MKL1 could regulate ER␣ activity in cells endogenously expressing the receptor. The ER␣-positive breast carcinoma cell line MCF-7 was selected for this purpose. Firstly, the impact of transiently transfected forms of MKL1 on the transactivation efficiency of endogenous ER␣ was measured on the ERE-tk-LUC reporter. As shown in Fig. 5A, the constitutively active mutant MKL1 ⌬N200 and MKL1, to a lesser extent, enhance the basal activity of the reporter gene measured in the absence of E 2 . Importantly, the transcriptional activity of endogenous ER␣ is up-or downregulated by dominant negative or dominant positive forms of MKL1, respectively. These results are in total agreement with those previously observed for transfected ER␣ in HepG2 and HeLa cells. As expected from their differentiated phenotype (E-cadherin-positive and vimentin-negative) (9), MCF-7 cells exhibited a cell context that was mainly permissive to the ER␣ AF1 box 1 activity (compare ER␣ with ER␣ ⌬79 and ER␣ CF in Fig. 5B). The overexpression of MKL1 and MKL1 ⌬N200 totally abolished this permissiveness, whereas the dominant negative form of MKL1 (⌬C301) enhanced it (Fig. 5B). These results clearly indicated that MCF-7 cells provide a similar context to HepG2 cells, in terms of ER␣ transactivation efficiency on an ERE-driven reporter gene. In contrast, the dedifferentiated ER␣-negative breast cancer MDA-MB 231 cell line (MDA) was insensitive to AF1 box 1 activity, as HeLa cells (Fig. 5C). Importantly, the stable re-expression of ER␣ into these cells (MDA-ER␣), which induced a more differentiated phenotype, restored an AF-1 box 1-sensitive cell context (Fig. 5C).
Having shown that MKL1 regulates ER␣ transcriptional activity, we next investigated whether MKL1 is present on endogenous estrogen-responsive promoters. We therefore performed ChIP experiments and sequential ChIP in MCF-7 cells to determine the (co)occupancy of several gene promoters with both ER␣ and MKL1. As shown in Fig. 5E, MKL1 was recruited   DECEMBER 4, 2009 • VOLUME 284 • NUMBER 49

Rho/MKL1 Signaling Pathway Regulates ER␣ Activity
to the promoter of all tested E 2 -regulated genes (control reverse transcription-PCR within Fig. 5D), including pS2/TFF1 (trefoil factor 1), Wisp2 (WNT1-inducible signaling pathway protein 2), PR A (progesterone receptor, A isoform), and SDF-1 (stromal cell derived factor-1) in the absence or presence of E 2 . Further-more, with the exception of PR A , a co-recruitment of MKL1 and ER␣ may occur on the three other analyzed promoters, as revealed through sequential ChIP (Fig. 5E). However, the association of ER␣ and MKL1 on these three promoters seems to exhibit some differences respective to the treatment of the cells

Rho/MKL1 Signaling Pathway Regulates ER␣ Activity
with E 2 . For instance, MKL1 appears constitutively associated with pS2/TFF1 and Wisp2 promoters, in conjunction with ER␣ of which the recruitment is stimulated by E 2 . In contrast, the recruitment of MKL1 on the SDF-1 promoter seems enhanced by E 2 and appears associated with ER␣ only in the presence of E 2 . This reflects the strict E 2 dependence of ER␣ recruitment on this promoter. These results were partly confirmed on the MDA-ER␣ cells, which offer a different system to analyze the association of MKL1 with the promoter region E 2 -regulated genes. However, differing from the MCF-7 cells, the recruitment of MKL1 on the pS2/TFF1 promoter appears to be enhanced by E 2 . Among the tested genes, only SDF-1 was not expressed in these cells (Fig. 5D). Importantly, the lack of expression of SDF-1 in MDA-ER␣ cells is reflected by its inability to recruit MKL1 or ER␣, likely illustrating a closed chromatin conformation (Fig. 5E). This provides an interesting additional control of the specificity of the signals observed. In conclusion, these data indicate that ER␣ and MKL1 can be corecruited and are likely to cooperate on the promoter region of several E 2 -regulated genes.

DISCUSSION
Transactivation efficiency of ER␣ results in part from the respective contribution exerted by its two activation functions, whose activity are tightly regulated in a cell-specific manner (7). For instance, we and others frequently use HeLa and HepG2 cell lines, which represent strict AF2-and AF1-permissive contexts, respectively. The relative sensitivity of a cell line to AF1 and AF2 can be defined by a comparison of the transcriptional activity of transfected wild-type ER␣ protein with that of an AF1-deleted form (ER␣ CF). A similar activity of both receptors defines a strict AF2-permissive cell context, whereas a lower transcriptional activity of ER␣ CF is inherent to a cell context permissive to AF1. However, differences observed in the expression level of some of the variants used in a given cell line (for instance in HeLa cell line (Ref. 10 and Fig. 2C)) might weaken the interpretation of the results. Nevertheless, the use of a second approach, based on the strict dependence of the partial agonistic activity of 4-OHT on the human complement C3 promoter toward a cell context sensitive to AF1 (precisely to AF1 box 1 (Refs. 9, 10 and Fig. 4B)), allowed us to confirm the respective contributions exerted by the two activation functions in a given cell line.
Through these approaches, we have shown that a high transactivation potency of ER␣ is correlated with an AF1-permissive cell context generally found in most differentiated epithelial-like cell lines (10). Accordingly, AF2 is the only active AF in cells that have achieved their epithelial-mesenchymal transition (9). A main step in epithelial-mesenchymal transition is the loss of E-cadherin-mediated cell contacts, which is also associated with tumor dedifferentiation and metastasis (27,28). We recently demonstrated that the existence of intercellular junctions is required for a cell line to be permissive to AF1 (10). Rho proteins and the actin/MKL1 effector pathway, which are also reported to be involved in the epithelial-mesenchymal transition (15,29,30), were shown in this study to have a dramatic impact on ER␣ transcriptional activity.
High activities of Rho or of its target ROCK have been correlated with the disruption of cadherin-mediated intercellular adhesion, cell migration, and metastasis (11,31). In accordance with previous work (32), the expression of constitutively active forms of RhoA, Rac1, and Cdc42 reduced ER␣ transactivation efficiency in a cell-specific manner, whereas dominant negative forms of the Rho GTPases enhanced ER␣ activity. Because the overexpression of such specific proteins might result in a multiplicity of effects that may indirectly influence ER␣ transcriptional activity, specific inhibitors of RhoA and ROCK were used to confirm the regulation. This effect involves a modulation of the activity of the AF1 box 1 activity, which was previously identified as the main domain involved in the cell context-dependent contribution of the AF1 (10). Rho GTPases control cellular processes that are connected with actin cytoskeleton, such as changes in cell shape, adhesion, and migration (22,23). A direct impact of such processes on gene expression has been recently illustrated by the control exerted by RhoA on SRFregulated genes, such as the immediate early gene c-fos. This process involves Rho GTPases promoting F-actin accumulation, thereby changing the cellular levels of actin monomers (G-actin) (25) and inducing the activity of the transcription factor MKL1 (12,16). We demonstrate here that this actin/ MKL1 pathway influences ER␣ transactivation efficiency by controlling the respective contributions of AF1 and AF2 toward ER␣ transcriptional activity (integrative model illustrated in Fig. 6). This was evidenced through (i) changing the amount of cellular G-actin through cell treatment with an actin-capping drug, or by overexpressing wild-type or unpolymerizable ␤-actin, and (ii) the use of dominant positive and negative forms of MKL1. In all conditions leading to MKL1 activation, the cellular context became refractory to ER␣ AF1. The physiological function of this AF is therefore likely to require low or no MKL1 activity. In B and C, cells were treated for 24 h with 10 nM E 2 . Transactivation efficiency corresponds to normalized luciferase activities expressed as -fold increase above values measured in the absence of ER␣ forms (empty pCR3.1). Values correspond to the mean Ϯ S.E. from at least three different experiments. *, ER␣ transactivation efficiency significantly differs from ER␣ ⌬79 transactivation efficiency (p Ͻ 0.005, as determined by ANOVA followed by a Fisher's post hoc test). D, reverse transcription-PCR assays were performed on MCF-7, MDA-MB-231, or MDA-ER␣ cells that were starved for 48 h in a steroid-free medium and then treated 8 h with 10 nM E 2 . Targeted mRNAs are indicated on the left side of the images. The mRNA encoding the acidic ribosomal protein PO was used as a control. E, chromatin prepared from MCF-7 or MDA-ER␣ cells treated as in D was used in ChIP and sequential ChIP (Re-ChIP) experiments. These assays were performed using anti-ER␣, anti-MKL1, or control (anti-HA) antibodies. PCR was performed on immunoprecipitated DNA, using primer pairs targeting the amplification of proximal promoter sequences of indicated genes. Experiments were repeated three times with identical results. According to our model, MKL1 adopts distinct functional states in HeLa and HepG2 cell lines, which exhibit a differential permissiveness for each of ER␣ AF. Therefore, how might MKL1 control the transcriptional activity of ER␣ and act to modify the relative contribution of AF1 and AF2? Differences in the expression of MKL1, or its subcellular localization, which is supposed to regulate its activity (12), were not observed between both cell lines. Post-translational modifications of MKL1 may thus provide additional cell-specific mechanisms controlling its activity, such as phosphorylation (12), or sumoylation (33). Results obtained through ChIP experiments show that MKL1 is associated with the promoter region of estrogenresponsive genes in the ER␣-positive MCF-7 cells and the MDA-MB 231 cells manipulated to stably express ER␣ (MDA-ER␣). Interestingly, both cell lines were demonstrated to be mainly permissive to AF1 transactivation. Following our model, we therefore hypothesize that MKL1 is recruited to the promoters of E 2 -responsive genes in AF1 contexts which, by definition, would require low or no MKL1 activity. In these conditions, MKL1 is expected to be co-recruited with ER␣, as demonstrated by our sequential ChIPs on given promoters. This may support a direct action of MKL1 on ER␣ activity by modulating the mobilization of coactivators and/or corepressors. For instance, one possible mechanism relies on the fact that the target of the Rho/actin/MKL1 pathway is the ER␣ AF1 box 1, located between amino acids 38 and 79, which is also involved in the control of ER␣ activity through E-cadherinmediated cell contacts (10). This AF1 subregion constitutes an interaction surface for cofactors of the p160 and p300 families and synergistically bridges AF1 and AF2 activities (18,34,35). Importantly, the overexpression of Rho guanine nucleotide dissociation inhibitor potentiates the action of CBP/p300 on ER␣ transcriptional activity (36). Therefore, upon activation by Rho, MKL1 might inhibit the coactivation of ER␣ by CBP/p300. Rather than having an impact on cofactor recruitment, MKL1 action could also involve corepressors. Indeed, in HeLa cells, the dominant positive form of MKL1 allowed 4-OHT-bound ER␣ to repress the activity of the C3-LUC reporter gene. A high activity of MKL1 might therefore allow ligand-bound ER␣ to recruit corepressors such as the nuclear receptor corepressor or the silencing mediator of retinoic acid and the thyroid hormone receptor, which are known actors of the antagonistic actions of selective estrogen receptor mediators such as 4-OHT (37). Finally, MKL1 might act alternatively through a mechanism that is independent of its interaction with ER␣. Such a hypothesis is strengthened by the fact that the association of MKL1 with the promoters of endogenous E 2 -target genes appears to be mainly constitutive. In the absence of estradiol, MKL1 is always recruited to the tested promoters, whereas this is not systematically the case for ER␣. Furthermore, the sequential ChIP experiments performed on the PR A promoter indicate that MKL1 is likely to be recruited onto a subset of regulatory elements independently of ER␣. Thus, MKL1 might regulate epigenetic marks such as histone acetylation or methylation, which indirectly could have an impact on ER␣ activity. Corroborating this hypothesis, we observed important variations in the basal activity of the luciferase reporter genes following overexpression of MKL1 and its deleted variants. Notably, ERE-driven (ERE-tk-LUC) and ERE-less reporter genes (c-fos-LUC and pTAL-LUC) behaved in a similar manner in response to MKL1 mutants while displaying totally different plasmid backbones. Only the amplitude of the responses differed. Accordingly, MKL1 might directly or indirectly affect the folding of the chromatin-like structure adopted by plasmids in the nucleus. Nevertheless, the correlation between a reduced transactivation potency of ER␣ and an increased basal activity of the reporter gene is not systematic. For instance, both ⌬Band ⌬Q-deleted MKL1 forms reduced ER␣ activity but exerted opposite effects on the basal activity of some reporter genes. Consequently, the mechanisms underlying the MKL1 influence on ER␣ activity are likely different from those responsible for the variations of the basal activity of reporter genes. Studying more extensively the combinatorial recruitments of coactivators and corepressors on natural ER␣ target promoters as well as epigenetic regulations, in conditions where MKL1 exhibits distinct functional state, is now awaited to obtain further insights into the cellspecific mechanisms engaged by MKL1 to influence ER␣ transcriptional activity.
In conclusion, we provide evidence that the respective contribution exerted by AF1 and AF2 toward ER␣ transcriptional activity is controlled by the Rho/actin/MKL1 pathway. This implies that the transcriptional activity of ER␣ is greatly modified during the epithelial-mesenchymal transition, through a switch from a dominant AF1-to an AF2-permissive cell context. Changes in actin-dependent cell process, which occur during cell transformation and tumorigenesis, are thus likely to FIGURE 6. Integrative model of the regulation of ER␣ transcriptional activity by MKL1. Differentiated epithelial cells express proteins involved in calcium-dependent contacts such as E-cadherins. In these cells, the establishment of such contacts leads to a low activity of Rho GTPases, which displaces the equilibrium between G-and F-actin toward the generation of an important pool of G-actin. This form of actin interacts with MKL1 which, in this case, promotes the transcriptional activity of ER␣ through the potentiation of an AF1 box 1 activity. In contrast, in undifferentiated cells that have accomplished their epithelial to mesenchymal transition, the high activity of Rho GTPases triggers the polymerization of actin, which reduces the amounts of G-actin present in these cells. Actin-unbound MKL1 then inhibits the activity of the AF1 box 1 and weakens the transactivation potency of ER␣.