The Forkhead Box M1 Protein Regulates the Transcription of the Estrogen Receptor α in Breast Cancer Cells*

In this study, we have identified the Forkhead transcription factor FoxM1 as a physiological regulator of estrogen receptor α (ERα) expression in breast carcinoma cells. Our survey of a panel of 16 different breast cell lines showed a good correlation (13/16) between FoxM1 expression and expression of ERα at both protein and mRNA levels. We have also demonstrated that ectopic expression of FoxM1 in two different estrogen receptor-positive breast cancer cell lines, MCF-7 and ZR-75–30, led to up-regulation of ERα expression at protein and transcript levels. Furthermore, treatment of MCF-7 cells with the MEK inhibitor U0126, which blocks ERK1/2-dependent activation of FoxM1, also repressed ERα expression. Consistent with this, silencing of FoxM1 expression in MCF-7 cells using small interfering RNA resulted in the almost complete abrogation of ERα expression. We also went on to show that FoxM1 can activate the transcriptional activity of human ERα promoter primarily through two closely located Forkhead response elements located at the proximal region of the ERα promoter. Chromatin immunoprecipitation and biotinylated oligonucleotide pulldown assays have allowed us to confirm these Forkhead response elements as important for FoxM1 binding. Further co-immunoprecipitation experiments showed that FoxO3a and FoxM1 interact in vivo. Together with the chromatin immunoprecipitation and biotinylated oligonucleotide pulldown data, the co-immunoprecipitation results also suggest the possibility that FoxM1 and FoxO3a cooperate to regulate ERα gene transcription.

In this study, we have identified the Forkhead transcription factor FoxM1 as a physiological regulator of estrogen receptor ␣ (ER␣) expression in breast carcinoma cells. Our survey of a panel of 16 different breast cell lines showed a good correlation (13/16) between FoxM1 expression and expression of ER␣ at both protein and mRNA levels. We have also demonstrated that ectopic expression of FoxM1 in two different estrogen receptorpositive breast cancer cell lines, MCF-7 and ZR-75-30, led to up-regulation of ER␣ expression at protein and transcript levels. Furthermore, treatment of MCF-7 cells with the MEK inhibitor U0126, which blocks ERK1/2-dependent activation of FoxM1, also repressed ER␣ expression. Consistent with this, silencing of FoxM1 expression in MCF-7 cells using small interfering RNA resulted in the almost complete abrogation of ER␣ expression. We also went on to show that FoxM1 can activate the transcriptional activity of human ER␣ promoter primarily through two closely located Forkhead response elements located at the proximal region of the ER␣ promoter. Chromatin immunoprecipitation and biotinylated oligonucleotide pulldown assays have allowed us to confirm these Forkhead response elements as important for FoxM1 binding. Further co-immunoprecipitation experiments showed that FoxO3a and FoxM1 interact in vivo. Together with the chromatin immunoprecipitation and biotinylated oligonucleotide pulldown data, the co-immunoprecipitation results also suggest the possibility that FoxM1 and FoxO3a cooperate to regulate ER␣ gene transcription.
The biological effects of estrogen are primarily mediated through two nuclear steroid receptors, estrogen receptors ␣ and ␤ (ER␣ and ER␤) 2 (1)(2)(3). Estrogen receptors play a major role in regulating the growth, survival, and differentiation of normal and malignant breast epithelial cells. In normal mammary epithelium, ERs are rarely expressed in a large proportion of cells or at high levels (4,5). Approximately 60 -80% of all breast cancers overexpress ER␣, and ϳ70% of those respond to endocrine treatment, with further increased risk of breast cancer development in benign mammary tissue. These findings suggest that ER␣ plays a major role in breast cancer initiation and progression (4). ER␤ appears to have an opposing function to ER␣ in tumor growth (6), and low levels of ER␣ predict resistance to Tamoxifen therapy in breast cancer (7).
ER␣ functions as a classical transcription factor as well as a signal transducer. Estrogen binding activates ER␣ through phosphorylation, dissociates it from chaperonin proteins, and alters its conformation. Several kinases in the mitogenic signaling pathways can also phosphorylate and consequently activate ER␣ through a ligand-independent manner. Hormone-bound ERs dimerize and bind to estrogen response elements present in the promoter regions of target genes, including c-fos, c-jun, c-myc, TGF-␣, retinoic acid receptor ␣1, and progesterone receptor A, etc., to activate gene transcription (8 -12). ER␣ has also been shown to modulate gene transcription through alternative regulatory DNA sequences, such as AP-1, NF-B, and Sp-1 binding sites (13). Many of these ER-regulated genes, including IGFR1, cyclin D1, c-myc, and the anti-apoptotic gene bcl-2, are important for cell proliferation and survival (14 -17). Recent studies also suggest that a pool of ERs are located in the plasma membrane and cytoplasm. Direct binding of ER␣ to a diversity of membrane/cytoplasmic signaling molecules have been observed, and they include the p85 regulatory subunit of class I phosphoinositide 3-kinase, Src tyrosine kinase, and the insulin-like growth factor 1 (13). Activation of these pathways by estrogen through the ER initiates cell survival and proliferation signals via phosphorylation and activation of Akt and MAPK. Additionally, these signaling molecules are able to phosphorylate the ER and its co-regulators to augment nuclear ER signaling (13). In summary, the genomic and non-genomic actions of ER␣ play a crucial role in breast epithelial cell proliferation and survival.
Endocrine therapies aimed at blocking the action of estrogens have been the most effective and widely used methods for treating ER␣-positive breast cancers. These therapeutic approaches involve blocking estrogen binding to ER using antiestrogens such as tamoxifen, inhibiting estrogen synthesis using aromatase inhibitors such as exemestane, and reducing ER protein levels using "pure" anti-estrogens such as fluvestrant (ICI 182,780 or Faslodex) (13). Nevertheless, a proportion of ER␣-positive tumors do not respond to hormone treatment at all (de novo resistance), and the majority of those that initially respond eventually become resistant (acquired resistance). Most resistant tumors remain ER␣-positive, indicating a continued role for ER␣ in breast cancer cell survival and proliferation (18,19).
A recent study suggested that ER␣ transcription is regulated by FoxO3a (Forkhead box class O, 3a), a member of the Forkhead family of transcription factors (20). FoxO3a activity is negatively regulated by the phosphoinositide 3-kinase signaling pathway. Activation of phosphoinositide 3-kinase by growth factors leads to phosphorylation and inactivation of FoxO3a by Akt (also termed PKB (protein kinase B)) (21). When activated, FoxO3a up-regulates p27 Kip1 , p130, and bim and down-regulates cyclin D1/2 and bcl-XL expression to mediate cell cycle arrest and apoptosis (22)(23)(24)(25). Because ER␣ expression is associated with breast cancer initiation and progression and FoxO3a activation with cell cycle arrest and/or apoptosis, it is implausible that FoxO3a would be the principal positive regulator of ER␣ transcription. Because all members of the Forkhead transcription factors share a similar DNA binding domain, we reasoned that other family member(s) could be the main physiological regulator(s) of ER␣ transcription.
Forkhead box (Fox) M1 is a transcription factor ubiquitously expressed in proliferating cells and a key regulator of both G 1 /S and G 2 /M phases of the cell cycle (26 -30). FoxM1 is localized mainly in the cytoplasm in late G 1 and S phases; nuclear translocation occurs during entry into the G 2 /M phase and is associated with FoxM1 phosphorylation by ERK1/2 (31). In this study, we have examined the possibility that FoxM1 regulates ER␣ expression and explored the mechanism involved. supplemented with 10% fetal bovine serum, 2 mM glutamine, and 100 units/ml penicillin/streptomycin in a humidified incubator in an atmosphere of 10% CO 2 at 37°C. MCF-7 Tet-on FoxM1 cells were selected and maintained in the presence of 500 g/ml of zeocin (Invitrogen).

EXPERIMENTAL PROCEDURES
Plasmids-The FoxO3a expression vector pLPC-FoxO3a(wt) and pLPC-FoxO3a(A3) and the FoxM1 expression vector pcDNA3-FoxM1 have previously been described (22,32). The ER␣ promoter constructs pGL3-proA and pGL3-proB were kindly provided by Prof. Shin-Ichi Hayashi (Saitama Cancer Center Research Institute, Saitama, Japan) (33). The psiRNA-FoxO3a expression vector was generated by cloning small synthetic oligonucleotides encoding two complementary sequences of 19 nucleotides, TCACTGCATAGTCGATTCA, into the Invivogen psiRNA plasmid (Autogen Bioclear, Wiltshire, UK). The pTER-FoxM1 siRNA expression vector was generated by cloning the two complementary hairpin sequences of CACGCAAGTAGTGGCCATC into the BglII and HindIII sites of the pTER vector provided by Dr. Marc van de Wetering (Department of Immunology, University Medical Center, Utrecht, The Netherlands) (34). The respective control RNA interference vectors were generated from nucleotide sequences with the middle 2 bases mutated (i.e. FoxM1-TCACTGCCCAGTCGATTCA and FoxO3a-TCACTGCAC-CGTCGATTCA). The pcDNA4-FoxM1 vector used for the establishment of MCF-7 Tet-on FoxM1 MCF-7 cells was generated by PCR amplification of the human FoxM1 cDNA (Gen-Bank TM accession number U83113) using a high-fidelity polymerase (Clontech) and subcloning into the HindIII/XbaI restriction sites of pcDNA4 TM 4T/O (Invitrogen) mammalian expression plasmid.
Immunoprecipitation Assay-Cells were washed in phosphate-buffered saline and lysed with 500 l of Nonidet P-40 lysis buffer (1% Nonidet P-40, 150 mM NaCl, 20 mM Tris, pH 7.4, 10 mM sodium molibdate, 1 mM sodium orthovanadate, 1 mM sodium fluoride, and Complete Mini protease inhibitors (Roche Applied Science)) for 15 min on ice. The cell lysates were precleared for 1 h with protein G-Sepharose, incubated with specific antibodies for 1 h, and then with 50% slurry of protein G-Sepharose for 1 h. The beads were washed five times with 500 l of lysis buffer and analyzed by Western blotting.
Transfection and Luciferase Reporter Promoter Assay-MCF-7 cells were cultured in 96-well plates until 60% confluent. The cells were transfected using FuGENE 6 reagent (Roche Applied Science) according to the manufacturer's instructions for 24 h, washed twice in phosphate-buffered saline, and then harvested for firefly/Renilla luciferase assays using the Dual-Luciferase reporter assay system (Promega, Southampton, UK).
Mutagenesis-Mutagenesis of putative Forkhead binding sites of the ER␣ promoter was carried out using the QuikChange sitedirected mutagenesis kit (Stratagene, Cambridge, UK). The sequences for primers used for creating point mutations and deletions are essentially those described for the DNA pulldown assays plus franking regions to give oligonucleotides of 42 bases.
Real-time Quantitative PCR (RT-qPCR)-Total RNA was isolated using the RNeasy kit (Qiagen, Crawley, UK). 1 g of total RNA was reverse-transcribed using the Superscript first strand synthesis system for RT-qPCR (Invitrogen), and the resulting first strand cDNA was used as the template in the real-time quantitative PCR analysis. All measurements were performed in triplicate. The mRNAs analyzed were FoxO3a, FoxM1, ER␣, and GAPDH, which served as an internal control and was used to normalize for variances in input cDNA. The following gene-specific primer pairs were designed using the ABI Primer Express software: FoxO3a-sense, 5Ј-TCTACGAGTGGATGGTGCGTT-3Ј and FoxO3a-antisense, 5Ј-CGACTATGCAGTGACAGGTTGTG-3Ј; FoxM1sense, 5Ј-TGCAGCTAGGGATGTGAATCTTC-3Ј and FoxM1antisense, 5Ј-GGAGCCCAGTCCATCAGAACT-3Ј; ER␣-sense, 5Ј-CAGATGGTCAGTGCCTTGTTGG-3Ј and ER␣-antisense, 5Ј-CCAAGAGCAAGTTAGGAGCAAACAG-3Ј; GAPDHsense, 5Ј-TCCCATCACCATCTTCCA-3Ј and GAPDH-antisense, 5Ј-CATCACGCCACAGTTTCC-3Ј. The specificity of each primer was determined using the NCBI BLAST software module. Detection of FoxO3a, FoxM1, and ER␣ transcription was performed with SYBR Green (Applied Biosystems, Brackley, UK) and an ABI PRISM 7700 sequence detection system (Applied Biosystems, UK) using the relative standard curve method.
Gene Silencing with Small Interfering RNAs-MCF-7 cells cultured in 100-mm plates were transfected with 15 g of pTER-FoxM1siRNA or psi-FoxO3a plasmid using FuGENE 6 reagent (Roche Applied Science). Cells were collected 72 h after transfection either for Western blot analysis or RT-qPCR.
Immunoprecipitation Assay-Cells were washed in phosphate-buffered saline and lysed with 500 l of Nonidet P-40 lysis buffer (1% Nonidet P-40, 150 mM NaCl, 20 mM Tris, pH 7.4, 10 mM sodium molibdate, 1 mM sodium orthovanadate, 1 mM sodium fluoride, and Complete Mini protease inhibitors (Roche Applied Science) for 15 min on ice. The cell lysates were precleared for 1 h with protein G-Sepharose and incubated with specific antibodies for 1 h and then with 50% slurry of protein G-Sepharose for 1 h. The beads were washed five times with 500 l of lysis buffer and analyzed by Western blotting.
Biotinylated Oligonucleotide Pulldown Assay-Nuclear extracts were prepared from cultured cells using the high-salt buffer (20 mM HEPES-KOH, pH 7.9, 1.5 mM MgCl 2 , 0.2 mM EDTA, 0.5 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride, 1 mM sodium orthovanadate, and Complete Mini protease inhibitors (Roche Applied Science) as described previously. After diluting with 2 volumes of low-salt lysis buffer without NaCl, 50 g of the cell extracts were incubated at 30°C for 10 min with either 0.1 nmol of the 5Ј-biotinylated double-stranded wild-type or mutant oligonucleotides (Invitrogen) previously coupled to streptavidin-agarose beads (Sigma) in the presence or absence of an appropriate amount of competitor non-biotinylated oligonucleotides for 1 h at 4°C and then for 10 min at 30°C. The beads were washed six times with phosphate-buffered saline and subjected to SDS-PAGE and Western blot analysis.

FoxM1, but Not FoxO3a, Expression Correlates with ER␣ Expression at the Transcriptional Level in Breast Cancer Cell
Lines-To explore the possibility that FoxM1 regulates ER␣ expression, we first analyzed the correlation between the levels of expression of FoxM1 or FoxO3a and ER␣ in a panel of 16 different breast cell lines by Western blotting. As shown in Fig.  1A, 13 of the 16 breast carcinoma cell lines showed good correlations between the levels of FoxM1 and ER␣ expression. In contrast, only 6 of the 16 cell lines demonstrated positive correlation between the levels of FoxO3a and ER␣ expression. We next examined the mRNA levels of ER␣ in these cells by RT-qPCR (Fig. 1B). The result showed that, although the expression level of FoxM1 protein correlated with that of ER␣ mRNA in 13 of the 16 cell lines, only 6 of the 16 cell lines analyzed showed significant correlations between FoxO3a and ER␣ FoxM1 Regulates ER␣ Transcription SEPTEMBER 1, 2006 • VOLUME 281 • NUMBER 35 mRNA. Because the activity of FoxO3a is negatively regulated by Akt phosphorylation, the phosphorylation status of Akt was also studied using specific antibodies against phosphorylated and total Akt. The result again failed to account for the lack of correlation between FoxO3a and ER␣ expression. These data suggest that FoxM1, rather than FoxO3a, is the physiological regulator of ER␣ transcription in breast cancer cells.
Ectopic Expression of FoxM1, but Not FoxO3a, Results in Upregulation of ER␣ Expression-To test the hypothesis that FoxM1, but not FoxO3a, regulates ER␣ expression, we transiently transfected the ER␣-positive breast cancer cell line ZR-75-30 with different amounts of expression plasmids encoding for FoxM1 and a constitutively active FoxO3a(A3) and analyzed the expression levels of ER␣ and known FoxM1 and FoxO3a targets at 48 h after transfection by Western blotting. These results showed that ectopic expression of FoxM1 increased the expression levels of the FoxM1 targets Plk and Cdc25c ( Fig. 2A). We also observed a moderate but significant increase in ER␣ expression in the ZR-75-30 cells overexpressing FoxM1 but not those overexpressing FoxO3a(A3). To further investigate whether FoxM1 was regulating the transcription of ER␣, we analyzed these cells by RT-qPCR. These results showed an ϳ2-fold increase in ER␣ transcription when FoxM1 was overexpressed, which was not observed when FoxO3a(A3) was overexpressed ( Fig. 2A).

Induction of FoxM1 Expression in an MCF-7 Tet-on System
Results in an Increase in the Levels of ER␣ Expression/Transcription-To further confirm that FoxM1 regulates ER␣ expression, we generated an MCF-7 cell line (MCF-7 Tet-on FoxM1) stably transfected with the pTet-On (Clontech) and pcDNA4-FoxM1 plasmids in which the FoxM1 expression was inducible by the addition of doxycycline. As shown in Fig. 2B, treatment of this cell line with 2 g/ml doxycycline increased the expression of FoxM1 with time, and this was accompanied by an increase in the expression of ER␣ and known FoxM1 targets such as Cdc25c and Plk. There was a modest downregulation of FoxM1 expression at 24 h, and this was probably because of the negative cell proliferative effect of doxycycline. Nevertheless, this was accompanied by a similar decline in ER␣, Cdc25c, and Plk expression, further confirming our hypothesis that FoxM1 regulates ER␣ expression. Notably, the level of FoxO3a expression remained constant up to 48 h and declined moderately at 72 h after doxycycline treatment, suggesting that the induction of ER␣ by FoxM1 is not mediated through modulating FoxO3a expression. It is notable that the level of ER␣ induction was lower in the ZR-75-1 cells compared with the MCF7 cells. The reason is unclear, but it is likely to be due to the fact that not all of the ZR-75-1 cells would be expressing

Inhibition of FoxM1 Activity Leads to Down-regulation of ER␣ Expression-We
next studied whether inhibition of FoxM1 activity would lead to down-regulation of ER␣ expression. To this end, we treated MCF-7 cells with 10 M U0126 and performed a time course experiment as indicated in Fig. 3A. Western blot analysis showed down-regulation of the known FoxM1 targets Cdc25c and Cyclin B1 as early as 1 h after treatment. This was accompanied by a mobility shift and decrease in FoxM1 levels. The faster migration species of FoxM1 observed after treatment with U0126 probably reflects the dephosphorylated and inactive form of this protein. We also observed a similar down-regulation of ER␣ expression but only at 3 h after treatment. This probably highlights the fact that the half-life of ER␣ protein is longer than those of Cdc25c and cyclin B1. The levels of expression of FoxO3a and the FoxO3a targets Bim and p27 Kip1 were constant throughout the time course, indicating that there was no change in FoxO3a activity. These results provide further evidence that FoxM1 is a physiological regulator of ER␣ expression in breast cancer cells. To extend these observations, we next studied the effects of FoxM1 and FoxO3a silencing on the expression of ER␣ in MCF-7 cells. To this end, we transfected siRNA vectors targeting FoxM1 or FoxO3a into MCF-7 cells and analyzed the protein expression at 72 h post-transfection by Western blotting. The results showed that abolishing FoxM1 expression in MCF-7 cells led to the abrogation of expression of ER␣ as well as FoxM1 targets, such as Cyclin B1, Plk, Cdc25b, and Cdc25c (Fig. 3B). Silencing of FoxO3a in MCF-7 cells resulted in the down-regulation of the known FoxO3a targets Bim and p27 Kip1 . Surprisingly, we also detected a moderate down-regulation of ER␣ and FoxM1 expression in these cells when compared with mock and control siRNA-transfected MCF-7 cells. The down-regulation of ER␣ in cells transfected with FoxO3a siRNA could be mediated through down-regulation of FoxM1 expression. The results also indicated that there were no significant changes in the expression and activity of ERK1/2 and Akt, which are upstream regulators of FoxM1 and FoxO3a, respec- . Gene transcripts of these cells were analyzed by RT-qPCR using the ABI PRISM 7700 sequence detection system (lower panels). B, MCF-7 Tet-on FoxM1 cells in which FoxM1 expression is inducible by the addition of doxycycline were treated with 2 g/ml doxycycline for the times indicated and analyzed by Western blot using specific antibodies (top panel). The mRNA of these cells was analyzed by RT-qPCR using the ABI PRISM 7700 sequence detection system (lower panels). SEPTEMBER 1, 2006 • VOLUME 281 • NUMBER 35 tively. These results together established that ER␣ is a physiological transcriptional target of FoxM1.

FoxM1 Regulates ER␣ Transcription
FoxM1 Expression Activates the ER␣ Promoter-dependent Transcription Activity-To test whether FoxM1 regulates ER␣ at the gene promoter level, we next investigated the ability of FoxM1 and FoxO3a to activate the ER␣ promoter. To this end, the ER␣ promoters were transfected into COS and MCF-7 cells with various amounts of expression vectors for FoxM1, wildtype FoxO3a, or a constitutively active FoxO3a. The co-transfection experiments demonstrated that, although FoxM1 does not transactivate ER␣ promoter B (distal) efficiently, it effectively induced promoter A (proximal) activity in both COS and MCF-7 cells. Conversely, FoxO3a specifically transactivates ER␣ promoter B, but not promoter A, consistent with the previous report (20). However, it was also notable that the basal activity of promoter B was low when compared with A, suggesting that promoter A is the predominant promoter in these cell lines. Sequence analysis of the ER␣ promoters A and B identified a number of consensus Forkhead-responsive elements (FHREs), including sites S1, S2, and S4 on promoter B, previously identified by Guo and Sonenshein (20) to be inducible by FoxO3a (Fig. 4A). To then map the transcriptional elements responsible for mediating the induction by FoxM1, a series of deletion reporter constructs of the ER␣ promoters A and B were co-transfected with FoxM1 expression vector into COS cells and analyzed for transcriptional activity 24 h after transfection. Deletion of the region of the ER␣ promoter B (Ϫ4542 to Ϫ3038) containing the previously defined FHREs S1 and S2 reduced significantly the activation by FoxM1 from 1.3-to 1.1-fold. The average induction of the longest wild-type ER␣ promoter A (Ϫ1168/ ϩ190) after FoxM1 transfection was 3.3-fold, and the comparably low levels of induction by FoxM1 probably reflects the fact that the FoxM1 expression vector is not expressing a constitutively active form of FoxM1. Deletion of the 5Ј distal region (Ϫ1168/Ϫ513) only marginally reduced the ability of FoxM1 to induce the promoter activity. The fold of activation by FoxM1 reduced significantly to background level when the region (Ϫ513/ϩ190) of the promoter containing three putative FoxM1 binding sites (S6, S7, and S8) was deleted. Similarly, a decrease in the level of induction was also observed when any one of the FHREs of the full-length ER␣ promoter A was mutated, suggesting that these FHRE sites are responsible for FOXM1 transactivation. Mutation of all three putative FHREs rendered the ER␣ promoter insensitive to FoxM1 induction. However, our transfection result also indicated that S8 is a weak FHRE compared with S6 and S7. Together, these transfection results suggested that the proximal ER␣ promoter FHREs S6, S7, and S8 primarily mediate the response to FoxM1 of the promoter.
FoxM1 and FoxO3a Bind to the Same Responsive Elements on the ER␣ Promoter-To further characterize these putative FHREs identified by promoter analysis and to demonstrate physical interaction of FoxM1 and FoxO3a with these sites, we used biotinylated oligonucleotides coupled to streptavidin-agarose beads to "pulldown" proteins interacting with different FHREs in MCF-7 cells and then analyzed the bound proteins by Western blotting. The result showed that FoxM1 bound to the S1/S2, S4, S6/S7, and S8 FHREs containing oligonucleotides but not to similar mutant oligonucleotides with the FHREs mutated, consistent with the transfection results. Surprisingly, FoxO3a was also shown to be interacting with the same FHREs. Similar to the deletion analysis results, we failed to demonstrate binding of FoxM1 or FoxO3a to the S5 FHRE. These interactions depend on the FHREs, as both the FoxM1 and FoxO3a binding could be competed off using increasing amounts of oligonucleotides containing the FHRE but not similar oligonucleotides with the FHRE mutated (Fig. 5A). These results together suggest that both FoxM1 and FoxO3a are recruited directly to the FHREs.
FoxM1 and FoxO3a Associate with the ER␣ Promoter in Vivo-We next performed ChIP assays to examine the in vivo occupancy of the ER␣ promoter by FoxM1 and FoxO3a. As demonstrated in Fig. 5B, both the anti-FoxM1 and anti-FoxO3a antibodies, but not the control antibodies (IgGs), precipitated the ER␣ pro-

FoxM1 Regulates ER␣ Transcription
moter regions containing the S1/S2, S4, and S6/S7 FHREs. Consistent with the oligo pulldown result, the ChIP assay also illustrated that the anti-FoxM1 and anti-FoxO3a antibodies failed to precipitate promoter fragments containing the S5 site. Interestingly, although the pulldown experiment showed that the S8 FHRE could bind FoxM1 and FoxO3a, the ChIP assay failed to show either FoxM1 or FoxO3a binding to the promoter region containing S8 in vivo (Fig. 6). In general, the ChIP result confirms the findings from the co-transfection and pulldown experiments.
FoxO3a Interacts with FoxM1 in Vivo-With the chromatin immunoprecipitation and biotinylated oligonucleotide pulldown data indicating that FoxM1 and FoxO3a bind to the same elements on the ER␣ promoter, we next investigated whether these two proteins could interact in vivo. To this end, we performed immunoprecipitation assays using antibodies specific for FoxM1 or FoxO3a. The co-precipitated proteins were then identified by Western blotting. These results showed that the endogenous FoxM1 and FoxO3a proteins coprecipitated with one another in MCF-7 cells, thus confirming in vivo interaction between FoxM1 and FoxO3a.

DISCUSSION
Estrogens are powerful mitogens essential for the initiation and progression of human breast and other gynecological cancers. Because ER␣ mediates the effects of estrogens, endocrine agents, such as Tamoxifen, Raloxifene, and Fulvestrant (ICI 182,780), are frequently used to block their synthesis or their activities in breast cancer treatment regimes. The successes of endocrine therapies for breast cancer treatments are often hampered by the development of resistance to hormonal treatments. Although enhanced growth factor signaling (which induces both genomic and non-genomic activities of ER␣) is the principal mechanism of acquired resistance, the major reason for de novo resistance to endocrine therapy is loss of ER␣ expression. Thus, unraveling the molecular mechanisms by which ER␣ transcription is regulated might provide vital information not only for the development of novel therapeutic strategies against endocrine-resistant ER␣-positive tumors but also for understanding the loss of ER␣ in de novo resistant diseases.
In this study, we have identified FoxM1 protein as a physio-   (20). Although we have confirmed in vitro and in vivo binding of FoxO3a to the ER␣ promoter and shown that overexpression of a constitutively active FoxO3a is able to transactivate region B of the ER␣ promoter in co-transfection studies, the results of our functional studies do not totally support the idea that FoxO3a is the primary physiological regulator of ER␣ transcription. We observed a poor correlation of 6/16 between FoxO3a expression levels and ER␣ expression in a larger panel of breast carcinoma cells compared with the previous study. Moreover, although ectopic expression of FoxM1 resulted in a significant increase in the level of ER␣ transcription in the ZR-75-30 breast cancer cell line, overexpression of the constitutively active FoxO3a did not lead to an up-regulation of ER␣ transcription in the same cell system. Although we did observe a moderate down-regulation of ER␣ expression in MCF-7 cells transfected with siRNA against FoxO3a, the decline in ER␣ expression could be a result of the down-regulation of FoxM1 expression as the level of FoxM1 expression was also significantly lowered in the same cells. Furthermore, given that FoxO transcription factors FIGURE 5. In vitro and in vivo binding of FoxM1 and FoxO3a to the putative Forkhead-responsive elements on the ER␣ promoter. Shown is a schematic representation of the human ER␣ promoter showing the putative FHREs as well as the regions corresponding to sequences used for DNA pulldown and chromatin immunoprecipitation PCR analyses (top panel). A, oligonucleotides containing different putative FHREs of the human ER␣ promoter were generated with 5Ј ends either free or conjugated to biotin molecules. DNA-binding nuclear extracts prepared from cycling MCF-7 cells were incubated with 1 g of biotin-conjugated wild-type (wt) and mutant (mut) oligonucleotides; see the first two lanes and wt in all other lanes). In lanes 3-5 and lanes 6 -8, increasing amounts (1, 5, and 10ϫ molar excess) of competing non-conjugated oligos were incubated with the mix. Proteins binding to the biotinylated oligonucleotides were pulled down using streptavidineagarose beads, and the washed lysates were analyzed using SDS-PAGE followed by immunoblotting using specific antibodies against FoxO3a and FoxM1. B, protein-DNA complexes from cycling MCF-7 cells were formaldehyde cross-linked in vivo. Chromatin fragments from these cells were subjected to immunoprecipitation with antibodies against IgG (nonspecific), FoxM1, or FoxO3a as indicated. After cross-link reversal, the co-immunoprecipitated DNA was amplified by PCR using the indicated primers and resolved in 2% agarose gels.

FoxM1 Regulates ER␣ Transcription
mediate cell cycle arrest and apoptosis whereas expression of ER␣ is associated with breast cancer cell proliferation, it is unlikely that FoxO3a is the primary physiological activator of ER␣ expression in breast cancer cells. However, paradoxically, a number of previous studies have shown that ER␣ and the proliferation marker Ki67 nuclear antigen are not usually expressed in the same breast tumor cells, leading to the suggestion that ER␣ regulates breast epithelial cell proliferation indirectly (35). Despite these earlier studies, more recent data demonstrates that ER␣ is expressed in proliferating breast epithelial cells and co-expresses with proliferative markers, including Ki67, MYC, cyclin D1, and stromal cell-derived factor (SDF)-1 (36). The failure in these earlier studies to demonstrate the coexpression of ER␣ and Ki67 in breast epithelial cells has been proposed to be due to the fact that the low mitotic index in these early samples and the short half-life and narrow expression window of Ki67 during the cell cycle have made the detection of transient co-expression of Ki67 and ER␣ difficult (36).
In exploring the mechanism by which FoxM1 regulates ER␣ expression, we performed mutation and deletion analyses of the ER␣ promoter and identified two FHREs located at the proximal region of the ER␣ promoter A to be primarily responsible for the transactivation of the ER␣ promoter by FoxM1. Further, our chromatin immunoprecipitation and biotinylated oligonucleotide pulldown assays indicate that FoxM1 binds directly to the FHREs of the ER␣ promoter to induce ER␣ transcription. Interestingly, these experiments also show that these FoxM1-bound FHREs are also occupied by FoxO3a. The significance of this observation is unclear, but it does suggest that FoxO3a could have a role in ER␣ transcription. One explanation is that FoxO3a could form a complex with FoxM1 to enhance FoxM1-dependent transcription activity. Supporting this hypothesis is the co-immunoprecipitation results showing that FoxO3a and FoxM1 interact in vivo. Indeed, different Forkhead subfamily members have been shown to form transcriptional complexes to regulate gene transcription. It has been demonstrated previously that FoxG1 can bind to FoxO3a in a Smad-containing complex to negatively regulate the cyclin-dependent kinase p21 Cip1 expression (37). This observation not only indicates that Forkhead transcription factors can interact but also that they can cooperate to control target gene transcription.
In summary, we have presented comprehensive experimen-tal evidence demonstrating that FoxM1 is a physiological regulator of ER␣ expression in breast cancer cells. We have also shown that FoxM1 regulates ER␣ expression at the transcriptional and promoter levels primarily through binding directly to FHREs located at the proximal region of the ER␣ promoter. Although we could detect binding of FoxO3a to the ER␣ promoter, we failed to demonstrate that FoxO3a can activate expression of endogenous ER␣ expression. However, our chromatin immunoprecipitation, biotinylated oligonucleotide pulldown, and co-immunoprecipitation results all suggest the possibility that FoxM1 and FoxO3a cooperate to regulate ER␣ gene transcription. Thus, despite an evident role of FoxM1 in ER␣ gene regulation, the function and mechanism of FoxO3a in regulating ER␣ expression in breast epithelial cells are still unclear and warrant further investigation.