The female sex steroid hormone oestrogen stimulates both cell proliferation and cell differentiation in target tissues. These biological actions are mediated primarily through nuclear oestrogen receptors (ERs). The ligand‐dependent transactivation of ERs requires several nuclear co‐regulator complexes; however, the cell‐cycle‐dependent associations of these complexes are poorly understood. By using a synchronization system, we found that the transactivation function of ERα at G2/M was lowered. Biochemical approaches showed that ERα associated with two discrete classes of ATP‐dependent chromatin‐remodelling complex in a cell‐cycle‐dependent manner. The components of the NuRD‐type complex were identified as G2/M‐phase‐specific ERα co‐repressors. Thus, our results indicate that the transactivation function of ERα is cell‐cycle dependent and is coupled with a cell‐cycle‐dependent association of chromatin‐remodelling complexes.
Sex steroid hormones exert their biological actions by directly binding to and activating their cognate nuclear receptors (Mangelsdorf et al, 1995). Steroid receptors are members of the nuclear receptor superfamily that function as sequence‐specific and hormone‐inducible transcriptional regulators in target gene promoters. Distinct classes of co‐regulators and multiprotein co‐regulator complexes are indispensable for the chromatin reorganization that occurs during the hormone‐dependent transcriptional control by nuclear receptors (Mellor, 2005; Rosenfeld et al, 2006). These complexes seem to modify chromatin configuration in a highly regulated manner by controlling nucleosomal rearrangement and enzyme‐catalysed modifications of histone tails (Workman, 2006; Li et al, 2007).
Two main classes of chromatin‐modifying complexes that co‐regulate nuclear receptors have been well characterized (Rosenfeld et al, 2006). One class is a histone‐modifying complex (Bannister & Kouzarides, 2005; Klose & Zhang, 2007; Lee & Workman, 2007) and the other class is an ATP‐dependent chromatin‐remodelling complex (Berger, 2007; Kouzarides, 2007). The latter complex uses ATP hydrolysis to rearrange nucleosomal arrays in a non‐covalent manner to facilitate or to prevent the access of nuclear receptors to nucleosomal DNA. These ATP‐dependent chromatin‐remodelling complexes have been classified into three subfamilies depending on which core catalytic components have DNA‐dependent ATPase activity. Brg‐1/Brm is a core component of the SWI/SNF‐type complex, SNF2h is a component of the ISWI‐type complex and Mi2 is a component of the NuRD‐type complex (Kitagawa et al, 2003; Fujiki et al, 2005; de la Serna et al, 2006; Denslow & Wade, 2007). The combination of the catalytic ATPase subunits with the components of the complex defines the role of each complex in transcriptional control. The SWI/SNF‐ and ISWI‐type complexes potently co‐activate the function of nuclear receptors (Belandia & Parker, 2003), whereas the NuRD‐type complex is assumed to co‐repress the function of nuclear receptors, owing to the presence of histone deacetylase (HDAC) in the complex (Denslow & Wade, 2007; Manavathi et al, 2007).
Dynamic structural changes of the chromosome are evident during the cell cycle and control global gene regulation (Martinez & Danielsen, 2002; Swedlow & Hirano, 2003; Narayanan et al, 2005; Groth et al, 2007; Takezawa et al, 2007). The role of chromatin‐remodelling complexes in the alteration of chromatin structure over much of the cell cycle has been studied (Baumann et al, 2007; Groth et al, 2007). However, the role of chromatin‐remodelling complexes in the cell‐cycle‐dependent transcriptional function of nuclear receptors during the S–M cell‐cycle stages is unknown.
Oestrogen stimulates both cell proliferation and cell differentiation of target cells (Deroo & Korach, 2006), presumably by means of cell‐cycle‐dependent co‐regulator function. Therefore, the transactivation functions of oestrogen receptors (ERs) and also their co‐regulators at specific cell‐cycle stages are an area of considerable interest for further research. Here, the cell‐cycle‐dependent switching of chromatin‐remodelling complexes is described.
Transactivation function of ERα was lowered at G2/M
To investigate whether the transactivation function of activated ERα varies across the cell cycle, we tested the transactivation function of endogenous ERα activated by 17β‐oestradiol (E2) in synchronized MCF7 (ERα‐positive) or MDA‐MB231 cells (ERα‐negative breast carcinoma cell line; Fig 1A; supplementary Fig S1 online). The synchronization of each phase was confirmed by DNA content analysis using flow cytometry, as reported previously (Takezawa et al, 2007). In a luciferase assay with an ERE‐TATA reporter (a PEST sequence was tagged to luciferase for rapid protein degradation (Hill et al, 2001)), the transcriptional activity of ERα seemed to be unaltered at G1/S (Fig 1B, compare lane 4 with lane 2 in asynchronous cells), whereas at G2/M, its activity was significantly lower in MCF7 cells (Fig 1B, compare lane 2 with lane 6). As expected, the ERα activity was undetectable in MDA‐MB231 cells.
To explore the molecular basis of cell‐cycle‐dependent alteration of ERα activity, we generated a HeLa S3 stable transformant cell line (HeLa‐ERα) expressing human ERα tagged with Flag and haemagglutinin epitopes at the amino termini (Fig 1C). The cell‐cycle‐dependent transcriptional activity of ERα was also confirmed in these cells (supplementary Fig S2A online).
Given that cell‐cycle‐related proteins such as cyclins are regulated by both their cellular localization and their relative abundance over the cell cycle (Moore et al, 2003), we examined the expression levels and localization of ERα in synchronized HeLa cells. The expression levels of ERα were slightly reduced at G1/S and were higher at G2/M (supplementary Fig S2B online, left panel). Immunofluorescence using an ERα antibody showed that ERα was localized mainly in the nucleus and seemed to associate with chromosomes at G2/M, as well as G1/S, regardless of stimulation with E2 (supplementary Fig S2B online, right panel). Certain classes of sequence‐specific regulators, however, dissociated from mitotic chromosomes (Martinez‐Balbas et al, 1995; Nuthall et al, 2002; Prasanth et al, 2003).
As ERα is localized in chromosomal areas even at G2/M, we suggested that ERα‐interacting proteins define the transactivation function of ERα in each cell‐cycle phase. To test this, we biochemically purified ERα interactants from synchronized cells in the presence of E2 (Fig 1D). Nuclear extracts of asynchronous or synchronized HeLa stable transformants were subjected to sequential affinity column purification by using an anti‐Flag M2 affinity resin column. The interactants were separated by SDS–polyacrylamide gel electrophoresis and silver stained. By comparing the interactants of the cells at G1/S versus G2/M, we found that the association of ERα with several interactants was cell‐cycle stage dependent (Fig 1E, left panel, arrowheads). As ERα activity was lowered at G2/M, we focused on cell‐cycle‐stage‐specific interactants and identified them by using mass fingerprinting. One of them, Mi2, was a G2/M‐phase‐specific interactant; Mi2 is a core ATPase component of the NuRD‐type chromatin‐remodelling complex (Sif, 2004; Denslow & Wade, 2007). Western blotting verified known components of the NuRD complex in the interactants only at G2/M, and detected components of the SWI/SNF complex in asynchronous and G1/S‐phase cells (Fig 1E, right panel).
Switch of ERα‐bound chromatin‐remodelling complexes
From those purification results, we hypothesized that, at G2/M, association of ATP‐dependent chromatin‐remodelling complexes with ERα leads to suppression of the transactivation function of ERα in the ERα target gene promoters. To test this hypothesis, the association of ERα with the components of the NuRD‐type complex was examined by immunoprecipitation of ERα from the nuclear extracts of asynchronous or synchronized MCF7 cells (Fig 2A) and HeLa‐ERα cells (Fig 1E, right panel), treated with or without E2. The core components of the NuRD‐type complex were detected in the immunoprecipitants of endogenous ERα from the E2‐treated MCF7 cells only at G2/M (Fig 2A, lane 6). The core component of the SWI/SNF‐type chromatin‐remodelling complex, Brg‐1, was co‐immunoprecipitated with activated ERα in asynchronous cells as reported previously (Belandia et al, 2002) and in synchronized MCF7 cells at G1/S in a ligand‐independent manner (Fig 2A). However, the association of Brg‐1 with ERα disappeared at G2/M (Fig 2A). Thus, there is a cell‐cycle‐dependent switching of the ATP‐dependent chromatin‐remodelling complexes associated with ERα. Next we examined the recruitment of components of the complex and ERα to oestrogen response elements in the endogenous oestrogen receptor binding site ebag9 (Tsuchiya et al, 2001) and c‐fos gene (Ohtake et al, 2003) promoters by chromatin immunoprecipitation analysis of synchronized MCF7, MDA‐MB231 and HeLa‐ERα cells (Fig 2B; supplementary Fig S2C online). Consistent with co‐immunoprecipitation data with ERα from synchronized MCF7 cells at G2/M, the components of the NuRD‐type complex were recruited to both gene promoters in an E2‐dependent manner (Fig 2B, right panel, lanes 3,6). E2‐dependent recruitment of Brg‐1 was seen in both an asynchronous and G1/S‐phase‐specific manner (Fig 2B, left panel). Unexpectedly, at G2/M, ERα anchored to the promoter regions in an E2‐independent manner (Fig 2B, left panel).
NuRD complex inhibits ERα transcriptional activity at G2/M
To investigate whether a NuRD‐type chromatin‐remodelling complex co‐represses the function of ERα in cells at G2/M, we examined the cellular colocalizations of Mi2 with E2‐bound/unbound ERα by using immunofluorescence with antibodies against Brg‐1, Mi2 and ERα (Fig 3A). At G1/S, Brg‐1 and Mi2 were distributed primarily in the nucleus and ERα was partly colocalized with Brg‐1 but not Mi2 (Fig 3A, left panel). At G2/M, however, Mi2 was detected on the mitotic chromosomes and partly colocalized with ERα only in the presence of E2 (Fig 3A, right panel). Brg‐1 was localized in the cytosol apart from the mitotic chromosomes, as reported previously (Muchardt et al, 1996). Thus, ERα seemed to associate with a NuRD‐type chromatin‐remodelling complex on mitotic chromosomes at G2/M in the presence of E2.
Finally, the putative co‐repressive function of Mi2 for ERα was tested in MCF7 cells synchronized at G2/M (Fig 3B). The E2‐dependent transcriptional activity of ERα was lower at G2/M. Either knockdown of the known components (MTA1/2/3) of the NuRD‐type complex (supplementary Fig S2D online) or treatment with an HDAC inhibitor, trichostatin A (TSA), abrogated the G2/M‐phase‐specific suppression of the transactivation function of ERα (Fig 3B). The MTA family member proteins (MTA1/2/3) seemed to associate physically with ERα in a glutathione S‐transferase pull‐down assay (supplementary Fig S2E online). This was consistent with the function of a NuRD‐type complex at least partly mediating G2/M‐phase‐specific suppression (Fig 3B). Furthermore, either RNA interference of MTAs or TSA treatment restored the endogenous expression of ebag9 or c‐fos at G2/M phase (Fig 3C). Knockdown of MTAs resulted in a decrease in E2‐dependent proliferation of MCF7 cells (Yanagisawa et al, 2002; Fig 3D). These findings further support the hypothesis that a NuRD‐type complex at G2/M lowers the E2‐dependent transcriptional activity of ERα and mediates oestrogen‐induced cell proliferation (Mazumdar et al, 2001; Manavathi et al, 2007).
Consistent with a previous report (Belandia et al, 2002), ERα associated with components of the SWI/SNF‐type chromatin‐remodelling complex, and these complexes were recruited to ERα target gene promoters in asynchronous cells. Similarly, at G1/S, the components were also recruited (Fig 2B). Thus, the SWI/SNF‐type complex seems to support the ligand‐induced transactivation function of liganded ERα in a cell‐cycle‐stage‐specific manner (Fig 4, left panel).
In this study, we biochemically identified Mi2 as an ERα interactant in the cells at G2/M; associations with the other components of the NuRD‐type chromatin‐remodelling complex were also detected. Furthermore, recruitment of the core components of the NuRD‐type complex to the ERα target gene promoters was seen only at G2/M and not at the other stages (Fig 2). Given that the NuRD‐type complex contains HDAC1 (Denslow & Wade, 2007), which, in this study, interacted with ERα at G2/M, the NuRD‐type complex might suppress the transcriptional function of ERα at G2/M (Fig 4, right panel). As knockdown of MTAs lowered E2‐dependent cell proliferation, this association seems to be significant for the actions of E2. Although we cannot exclude the possibility that activated ERα induces expression of co‐repressor genes responsible for cell‐cycle regulation during G2/M, it is more likely that the expression of ERα target genes is suppressed through association with the NuRD‐type complex. ERα might be transcriptionally silent at G2/M, but it might still contribute to cell‐cycle progression by reorganizing chromatin structure through its association with the NuRD‐type complex. At present, the molecular basis of cell‐cycle‐dependent switching of the two ATP‐dependent chromatin‐remodelling complexes remains to be uncovered. A switching factor seems to associate, at least transiently, with these complexes and is presumably functional at specific cell‐cycle stages. The appearance of such a factor might be under cell‐cycle regulation, similar to a recently characterized cell‐cycle‐dependent co‐repressor (Ret‐CoR) that represses the constitutively transrepressive‐type nuclear receptor (Takezawa et al, 2007).
Cell‐cycle analysis. Cells were synchronized 24 h after spreading (for suspension culture, at a cell density of 2 × 106/ml) by treatment with 2.5 mM thymidine (Wako, Osaka, Japan) at G1/S phase or 10 ng/ml demecolcine (Wako) at G2/M phase under phenol red‐free DMEM containing 5% serum. Each treatment was for 24 h except the luciferase assay. Cell synchronization was determined by using flow cytometry, as reported previously (Takezawa et al, 2007).
Transfection. Transfection with Lipofectamine Plus (Invitrogen, Carlsbad, CA, USA) was performed according to the manufacturer's protocol, before treatment with synchronization reagents. E2 and/or TSA treatment was 5 h before cell collection.
Purification of the ERα interactants. Nuclear extracts were prepared as described previously (Yanagisawa et al, 2002; Kitagawa et al, 2003; Ohtake et al, 2007; Takezawa et al, 2007) from synchronized HeLa‐ERα cells treated with E2 for 1.5 h. Extracts (∼400 mg) were loaded onto an anti‐Flag M2 affinity resin column (approximately 200 μl bed volume) and were washed extensively with washing buffer containing 300 mM KCl. Bound proteins were eluted from the column by incubation with 5 × volume of Flag peptide (100 μg/ml; Sigma, St Louis, MO, USA) three times for 5 min. Each sample was applied onto a 2–15% gradient polyacrylamide gel for mass fingerprinting (Ohtake et al, 2007; Takezawa et al, 2007).
Supplementary information is available at EMBO reports online (http://www.emboreports.org).
Conflict of Interest
The authors declare that they have no conflict of interest.
We thank F. Ohtake, S. Sawatsubashi, R. Fujiki, M.‐s. Kim and A. Yokoyama for helpful discussion and plasmids; S. Fujiyama for technical assistance; H. Higuch, K. Motoi and H. Yamazaki for manuscript preparation; and H. Ito and A. Miyajima for flow cytometric analysis. This work was supported in part by a grant‐in‐aid for priority areas from the Ministry of Education, Culture, Sports, Science and Technology (to H.K. and S.K.).
- Copyright © 2008 European Molecular Biology Organization