The effects of 17β-estradiol (E2) and ICI 182,780 (ICI) on activity of a BMP-6 promoter were compared in osteoblast-like and breast cancer cells transiently transfected with ERα. E2 but not ICI stimulated BMP-6 reporter activity in breast cancer cells, whereas the opposite was observed in osteoblast-like cells, associated with lack of AF-2 dependence of the response, and absent intranuclear localization of ERα, suggesting the involvement of a distinct ERα-dependent response mechanism in osteoblasts.
Introduction: Previous studies suggest that the tissue-selective effect of antiestrogens on bone reflects the ability of these compounds to target certain osteoblast regulatory genes. To explore this hypothesis, we examined whether antiestrogens preferentially stimulate the bone morphogenetic protein 6 (BMP-6) promoter in bone cells, and if so, whether this activity is associated with a distinct estrogen receptor (ER)α-dependent response mechanism to that in other cell types.
Materials and Methods: We compared the effects of 17β-estradiol (E2) and ICI 182,780 (ICI) on activity of a 4.3-kb BMP-6 reporter construct in osteoblast-like cells (human MG63 and SaOS-2 cells and rat ROS 17/2.8 cells), human MCF-7 and T47-D breast cancer cell lines, and HepG2 hepatoma cells, after transient transfection with ERα, ERβ, and mutant ER constructs.
Results: E2, but not ICI, stimulated BMP-6 reporter activity by approximately 100% in MCF-7, T47-D cells, and HepG2 cells when transfected with ERα. In contrast, in ERα-transfected osteoblast-like cells, an increase in reporter activity of approximately 75% was observed after treatment with ICI but not E2. The response of MG63 cells to ICI and MCF-7 cells to E2 both required ERα as opposed to ERβ and the ERα activation function (AF)-1 activation domain. However, whereas the AF-2 domain was also required for E2 to stimulate reporter activity in MCF-7 cells, the response to ICI in MG63 cells was AF-2 independent. In further studies where we compared the intracellular distribution of ERα associated with these responses, E2-dependent stimulation of the BMP-6 reporter in MCF-7 cells was associated with intranuclear localization of ERα, whereas extranuclear localization was seen in rat osteosarcoma cells (ROS) cells treated with ICI.
Conclusions: Antiestrogens selectively stimulate BMP-6 reporter activity in osteoblast-like cells through a distinct ERα-dependent mechanism characterized by independence of the AF-2 domain and extranuclear localization of ERα.
SEVERAL ANTIESTROGENS HAVE been identified that antagonize estrogen's stimulatory effects in reproductive tissues but act as partial estrogen agonists at other sites such as the skeleton.(1–4) The term selective estrogen receptor modulator (SERM) has been used to describe this tissue selective mode of action, which has been exploited in the treatment of postmenopausal osteoporosis.(5) In terms of the mechanisms by which SERMs exert their tissue-selective effects, it is recognized that their binding to the estrogen receptor (ER) induces conformational changes distinct from that induced by estrogen(6) and can, under certain situations, lead to significant stimulation of promoter activity.(7) However, although the latter study reported that SERMs such as raloxifene are more active than estrogen at stimulating gene transcription at AP-1 sites in the presence of ERβ, neither nonclassical response mechanisms nor ERβ expression are specific to bone, and the precise mechanisms that enable SERMs to act as estrogen antagonists or partial agonists in different tissues remains unclear.
A limited number of previous studies have examined the effect of SERMs on putative target genes in bone. For example, SERMs have been reported to stimulate activity of transforming growth factor (TGF)β3 and bone morphogenetic protein (BMP)-4 reporters in osteoblast-like cells.(8, 9) TGFβ and BMPs are expressed at relatively high levels in bone, are powerful stimulators of osteoblast function, and have been suggested to play an important role in mediating the effects of systemic hormones like estrogen on bone.(10, 11) Interestingly, stimulation of BMP-4 reporter activity by SERMs was not observed in MCF-7 breast cancer cells or endometrial carcinoma cells, and they required the presence of ERα as opposed to ERβ.(9) The latter observation is in line with our recent findings that ERα rather than ERβ mediates the stimulatory activity of estrogenic pathways on osteoblast activity as assessed in vivo.(12, 13)
The above studies raise the possibility that SERMs preferentially activate certain target genes in osteoblasts, which may reflect the role of a distinct ERα-dependent response mechanism in bone compared with other tissues. In this study, we addressed this hypothesis by examining whether SERMs preferentially stimulate the BMP-6 promoter in bone cells, and if so, whether this activity is associated with a different ERα-dependent response pathway to that in other cell types. The BMP-6 gene was analyzed in these studies, based on our previous findings that suggest that BMP-6 is involved in estrogen-dependent osteogenesis in vivo.(14, 15)
MATERIALS AND METHODS
17β-estradiol (E2), 4-hydroxytamoxifen, 16α-hydroxyestrone (16α-OHE1), 4-hydroxyestrone (4-OHE1), genistein, 4-hydroxytamoxifen (4-OHT), and 5α-dihydrotestosterone (DHT) were purchased from Sigma (Poole, Dorset, UK), ICI 182,780 (ICI) were purchased from Tocris (Bristol, Avon, UK), and raloxifene was purchased from Eli Lilly (Basingstoke, Hampshire, UK).
Plasmids expressing wildtype human ER (pRST7-ERα, pRST7-ERβ), mutant ER (pRST7-ERα-179C, pRST7-ERα-535-stop, pRST7-ERα/β), and an estrogen response element (ERE)-luc reporter consisting of three adjacent consensus EREs, were kindly provided by D McDonnell (Duke University, Durham, NC, USA).(16) Control pCR3.1 plasmid was obtained from Invitrogen (Paisley, UK). In addition, a plasmid expressing a constitutive β-galactosidase reporter (pCMV-β-Gal; Clontech, Palo Alto, CA, USA) was used to correct results for transfection efficiency. For localization studies, we used a red-fluorescent protein-tagged ERα expression construct provided by D McDonnell (Duke University.
We used a 4.3-kb BMP-6 reporter construct controlling the luciferase gene, based on the 1.2-kb portion of the BMP-6 gene immediately upstream of the transcription start site, which has previously been reported to contain promoter activity.(17) The larger 4.3-kb BMP-6 promoter fragment, which includes additional putative AP-1 domains, begins 4295 bases 5′ of the transcriptional start site, which is a further 179 bases 5′ of the gene's start codon and was cloned from a human genomic DNA library using the following PCR primers: forward, 5′-TCTTACTAGGTTGCCCAGCCTGATCTCGAG-3′ and reverse, 5′-AGAGGTCATCGCTAGTAAGAATGATCTGAC-3′. The PCR product was cloned into a TA-vector (TA-BMP-6), and the fragment from a KpnI/XbaI digest was subcloned into pGL3-Basic vector (Promega). The final 4.3-kb BMP-6-Luc reporter was created by performing a SalI/StuI digest of both the pGL3-TA-BMP-6 plasmid (vector) and the 1.2-kb BMP-6-Luc reporter (insert) and ligating the resulting fragments.
Cell culture and transient transfection assays
Rat osteosarcoma cells (ROS 17/2.8; kindly provided by L Lanyon, Royal Veterinary College, London, UK), human SaOS-2 osteosarcoma cells (American Type Culture Collection, Manassas, VA, USA), human MG63 osteosarcoma cells, human MCF-7 breast cancer cells, human T47-D breast cancer cells, and human HepG2 hepatoma cells (European Collection of Cell Cultures, Porton Down, Wiltshire, UK) were maintained at 37°C in a humidified 5% CO2 incubator. Phenol red-free DMEM (Life Technologies, Paisley, UK) was used containing 10% FCS(Life Technologies), supplemented with 2 mM L-glutamine (Life Technologies Ltd.). Cells were seeded at a density of 30,000 cells/well in 24-well plates 24 h before transfection with plasmids described above. Cells were washed once with PBS and transfected using Lipofectin reagent as described by the manufacturer (Life Technologies). Each transfection was performed in triplicate. The total amount of DNA transfected was kept constant at 1 μg per well after the addition of varying amounts of pCR3.1 control plasmid as necessary.
Transfection was allowed to proceed overnight, after which the transfection mixture was removed and replaced with phenol red-free DMEM supplemented with 10% charcoal-stripped fetal bovine serum (Hyclone Inc, UT, USA), containing 10−7M hormone unless otherwise specified, or an equivalent volume of ethanol vehicle. Twenty four hour later, cells were lysed and assessed for luciferase activity by chemical luminescence following the addition of luciferin substrate (Promega, Southampton, UK). β-galactosidase activity was used to correct luciferase activity for transfection efficiency, as measured following the addition of chlorophenol red-b-D-galactopyranoside substrate (Roche, East Sussex, UK).
Confocal laser scanning microscopy
Confocal laser scanning microscopy was used to characterize the cellular localization of a red fluorescent protein (RFP) tagged ERα (ds-red fluorochrome) within ROS 17/2.8 and MCF-7 cells. Cells were seeded onto glass coverslips in 6-well plates. ERα-RFP expression plasmid was transfected at 3 μg/well using Lipofectin (ROS 17/2.8) or FuGENE (MCF-7) cationic lipid transfection reagents according to the manufacturer's recommendations. Cells were treated with vehicle, 10−7 M E2, or ICI for 1 h before washing in PBS and fixing in 2% paraformaldehyde. Coverslips were washed again in PBS and mounted onto glass microscope slides using DAKO Fluorescent Mounting Medium (DAKO Corp., Carpinteria, CA, USA). Cells were observed using confocal laser scanning microscopy (Leica, Milton Keynes, Bucks, UK) with a TRITC filter and a 63× oil-immersion objective lens, and digital images were acquired using Leica TCSNT confocal software.
Results are expressed as mean ± SEM. Statistical analysis was by one-way ANOVA, which, where significant (i.e., p < 0.05), was followed by Fisher's posthoc least significant difference test.
We compared the effect of E2 and ICI on a luciferase reporter construct containing the 4.3-kb 5′-flanking region of the human BMP-6 gene promoter sequence in different cell lines, with or without cotransfection with ERα. In human MCF-7 and T47-D breast cancer and HepG2 hepatoma cell lines, reporter activity was increased by up to 100% in response to E2, but not ICI, after ERα transfection (Figs. 1A-1C). In contrast, in ROS 17/2.8 osteoblast-like cells cell lines and human SaOS-2 and MG63 osteoblast-like cells, an increase in reporter activity of ∼75% was observed after treatment with ICI but not E2 after transfection with ERα (Figs. 1D-1F). E2 and ICI had no effect on the 4.3-kb BMP-6 reporter in the absence of exogenous ERα in any cell type, and transfection with ERα did not itself influence reporter activity in the absence of ligand (results not shown).
ICI stimulated BMP-6 reporter activity in ROS cells in a dose-responsive manner across the range of 10−12-10−7 M, whereas E2 was without effect at any of these concentrations (Fig. 2A). In contrast, ICI caused dose-dependent suppression of an ERE-luc reporter in ROS cells. Stimulation of BMP-6 reporter activity by ICI 10−8 M was inhibited by 100-fold excess of E2, further confirming that this response is mediated by an estrogen receptor (Fig. 2B). To explore the ligand specificity of activation of BMP-6 reporter activity in osteoblast-like cells, we compared the effect of a range of estrogen-like compounds on ERE-luc and BMP-6 reporter activity in ROS cells. Compounds that demonstrated classical estrogen agonist activity, as reflected by stimulation of the ERE-luc reporter, had no effect on BMP-6 reporter activity (Fig. 3). Conversely, ICI, raloxifene, and 4-hydroxytamoxifen, which inhibited the ERE-luc reporter, significantly stimulated BMP-6 reporter activity. In further studies designed to examine the response to other classes of sex steroid, DHT 10−7 M was found to have no effect on BMP-6 reporter activity in ROS cells (results not shown).
We then investigated whether stimulation of the 4.3-kb BMP-6 reporter in human MG63 osteoblast-like and MCF-7 breast cancer cell lines by ICI and E2, respectively, requires ERα, but not ERβ, by co-transfecting cells with different isoforms (Fig. 4A). In MG63 cells, no response to ICI was observed when ERβ was transfected in place of ERα (Fig. 4B). This finding was not solely because of the fact that ERα has greater AF-1 activity, because use of a chimeric receptor in which the AF-1 domain of ERβ was replaced with that of ERα failed to restore the response. Equivalent results were obtained for the response of the 4.3-kb BMP-6 reporter to E2 in MCF-7 cells (Fig. 4C). In contrast, activation of ERE-luc reporter activity by E2 in MG63 cells seemed to be more efficient in the presence of ERβ compared with ERα (Fig. 4D). E2 significantly stimulated ERE-luc activity in MCF-7 cells in the absence of exogenous ER, reflecting the activity of endogenous ER in these cells, which was enhanced after transfection with ERα but not ERβ (Fig. 4E).
We used different ERα mutants to identify ERα domains required for activation of the 4.3-kb BMP-6 reporter in MG63 and MCF-7 cells (Fig. 5A). In MG63 cells, the response to ICI was prevented by use of the 179C AF-1-deficient mutant but not the 535-stop AF-2-deficient mutant (Fig. 5B). However, in MCF-7 cells, stimulation of BMP-6 reporter activity by E2 was inhibited when either the AF-1- or AF-2-deficient mutant were used, although in the case of the latter, a small but nonetheless significant response was observed relative to vehicle treatment (Fig. 5C). In contrast to findings in MG63 cells using the 4.3-kb BMP-6 reporter, use of the AF-1- and AF-2-deficient mutants, respectively, enhanced and inhibited stimulation of the ERE-luc reporter by E2 in these cells relative to cultures transfected with the wildtype ERα (Fig. 5D). Transfection of MCF-7 cells with intact ERα enhanced stimulation of ERE-luc reporter activity in response to E2 compared with cells transfected with the pCR3.1 control vector, whereas reporter activity in cells transfected with AF-1- or AF-2-deficient mutants were if anything reduced compared with the vector control (Fig. 5E).
To explore the basis for the different response of bone and breast cancer cells to estrogens and antiestrogens, we compared the cellular distribution of an ERα-RFP expression plasmid in ROS and MCF-7 cells by confocal microscopy. As reported previously,(18) ERα was restricted to the nucleus in MCF-7 cells in the presence of either vehicle or E2, whereas treatment with ICI induced limited but nonetheless detectable extranuclear translocation. In contrast, in ROS cells, ERα showed both cytoplasmic and nuclear expression in the absence of ligand (Fig. 6). Exposure to ICI resulted in rapid loss of nuclear expression, whereas treatment with E2 induced nuclear translocation of ERα.
This study shows that the BMP-6 promoter is regulated by ERα in a wide range of cell types, as assessed by reporter assay. However, the ligand specificity of this response seemed to differ between cells derived from different tissues, as did the mechanism involved in BMP-6 reporter activation: in breast- and liver-derived cells, the BMP-6 promoter was stimulated by E2 through a mechanism that required both AF-1 and AF-2 domains of ERα; in contrast, in osteoblast-like cells, BMP-6 promoter activity was stimulated by antiestrogens but not estrogens and required the presence of the AF-1 domain of ERα, but not the AF-2 domain.
This requirement for the AF-1 and AF-2 domains of ERα for E2 to stimulate the BMP-6 reporter in breast cancer- and hepatoma-derived cells was shared by the ERE-luc reporter in MCF-7 cells and is consistent with the possibility that ERα activates both the BMP-6 and ERE-luc reporters in these cells by recruitment of co-activators to the AF-2 domain, as occurs during “classical” estrogen responses.(19) In common with the majority of estrogen-responsive genes, analysis of the promoter sequence of the BMP-6 gene failed to identify any consensus EREs, whereas multiple domains involved in “non-classical” estrogen response pathways were identified, such as AP-1 sites,(19) suggesting that the precise mechanism by which E2 activates BMP-6 and ERE-luc reporters in MCF-7 cells is likely to differ significantly.
Whereas stimulation of the BMP-6 reporter by E2 required transfection with additional ERα, untransfected MCF-7 cells seemed to express sufficient ERα to enable E2 to stimulate ERE-luc reporter activity. However, rather than being caused by differences in activation mechanisms between ERE-luc and BMP-6 reporters per se, this finding may reflect the fact that MCF-7 cells contain insufficient levels of endogenous ERα to stimulate native as opposed to synthetic promoters. Consistent with this possibility, the estrogen response of several other reporter constructs based on native gene promoters, including those containing consensus EREs, also requires transient transfection with ERα in MCF-7 cells.(20) Any requirement for higher levels of ERα to stimulate native genes in transient transfection assays may be a consequence of reduced transcriptional efficiency caused by the lack of chromatin remodeling that occurs in endogenous or stably transfected genes.
While the basis for selective stimulation of BMP-6 by E2 in breast cancer- and hepatoma-derived cells is unclear, this may reflect the role of tissue-specific co-activators, which has previously been suggested to explain tissue-selective responses to tamoxifen.(21) As for the biological significance of these findings, BMP-6 has previously been found to be expressed in breast cancer cell lines, including MCF-7 cells, as well as normal breast tissue.(22) Although the function of BMP-6 in breast tissue remains uncertain, BMP-6 may play an equivalent role in the pathogenesis of bone metastases in breast cancer to that previously suggested in prostate cancer.(23) Hence, the present finding that E2 enhances BMP-6 reporter activity in MCF-7 cells raises the possibility that ERα-dependent pathways enhance skeletal secondary formation in breast cancer, which is consistent with the previous observation that patients with ER positive breast tumors are more likely to develop skeletal metatases.(24)
Contrary to the results obtained in breast cancer- and hepatoma-derived cells, BMP-6 reporter activity in bone cells was preferentially stimulated by antiestrogens. This observation is consistent with previous findings based on other osteoblast regulatory genes such as BMP-4, TGFβ3, and Cbfa1.(8, 9, 25) As in this study, stimulation of BMP-4 reporter activity by antiestrogens was observed after transient transfection with ERα as opposed to ERβ. In addition, a similar pattern of tissue specificity was observed because antiestrogens stimulated BMP-4 reporter activity in osteoblast-like cells but not cells derived from other tissues such as breast cancer or endometrial carcinoma cells.(9) However, in contrast to this study, E2 was not reported to stimulate BMP-4 reporter activity in any of the cell types studied.
Although previous observations that antiestrogens such as raloxifene prevent bone loss have been interpreted as an estrogen-like effect, there are differences in how estrogen and raloxifene influence the skeleton.(26) While these may reflect the fact that agents like raloxifene act as partial estrogen agonists, they are also consistent with the suggestion from the present findings that antiestrogens activate a distinct set of target genes in bone to estrogen. In terms of the molecular mechanisms involved in this action, it is generally accepted that, whereas binding of ER to estrogen agonists activates AF-2 function followed by the recruitment of co-activators, antiestrogens are either neutral or inhibitory at this domain.(19) Thus, ligand activation of ER by antiestrogens is thought to involve other activation functions such as AF-1, which is consistent with the present finding that AF-1 but not AF-2 seems to be required for antiestrogens to stimulate BMP-6 reporter activity in osteoblast-like cells. Our finding that antiestrogen-induced stimulation of BMP-6 reporter activity is dependent on ERα as opposed to ERβ is also consistent with this conclusion, because the latter ER isoform has a truncated N-terminal deficient in AF-1 function.(19)
In contrast, stimulation of the ERE-luc reporter by E2 in MG63 cells required the presence of the AF-2 domain, as expected for a classical estrogen response mechanism. However, one unexpected observation was that, rather than being AF-1-dependent, activation of ERE-luc reporter activity by E2 in MG63 cells seemed to be inhibited by AF-1, because the response was found to be enhanced in the presence of the AF-1-deficient ERα mutant. Previous investigations indicate that the ability of this AF-1-deficient ERα mutant to activate ERE-containing promoters varies according to cell type.(27) However, the suggestion from this study that the AF-1 domain of ERα serves to inhibit ERE-dependent transcription in osteoblasts has, to our knowledge, not been described previously.
Our results also raise the possibility that differences in intracellular distribution of ERα may in part explain why antiestrogens stimulate BMP-6 reporter activity in osteoblasts as opposed to breast cancer cells. Contrary to the nuclear expression pattern observed in breast cancer cells, ERα showed significant extranuclear expression in ROS cells, which is consistent with our recent observation that osteoblasts in mouse long bones express significant levels of cytoplasmic ERα as assessed by immunohistochemistry.(12) In support of the possibility that cytoplasmic ERα contributes to stimulation of BMP-6 reporter activity by ICI in osteoblasts, ICI treatment almost completely abolished nuclear ERα expression in ROS cells within 1 h of exposure. The latter finding is consistent with previous reports that ICI induces extranuclear translocation of ERα,(18, 28) which was also detected in MCF-7 cells in this study. Based on these observations, it is tempting to speculate that, in osteoblasts, extranuclear ERα is able to stimulate target genes after exposure to antiestrogens such as ICI. This possibility is consistent with the recent report that extranuclear ER is capable of activating Src/Shc/ERK intacellular signaling pathways involved in regulating osteoblast apoptosis.(29) However, the ligand specificity of the latter response differed to that reported here, because this was not found to be activated by SERMs.(30)
Contrary to the present findings in which antiestrogens stimulated BMP-6 reporter activity and those from studies using reporters of other genes involved in regulating osteoblast differentiation, such as BMP-4, TGFβ3, and Cbfa1, endogenous BMP-6 expression was previously found to be stimulated by E2, but not by ICI, in human fetal osteoblasts stably transfected with ERα.(31) Differences in number and/or cellular distribution of ERα between osteoblast-like cells used in these studies may be responsible for these apparently discrepant findings. In addition, there were important differences in experimental design in that, in the study by Rickard et al., cells were pretreated with ICI. Alternatively, these differences may reflect the role of distant cis-regulatory DNA sequences in regulating BMP-6 gene transcription in osteoblasts, as is frequently found in eukaryotic genes.(32) Although we attempted to address the latter question in this study, in preliminary studies intended to provide a basis for analyzing endogenous BMP-6 expression in osteoblasts, no expression was detected by Western blot analysis performed on untreated, untransfected ROS cells. This result is consistent with our recent observation that trabecular osteoblasts do not express BMP-6 as assessed by immunohistochemistry performed on sections from mouse long bones.(14) In contrast, stromal-like cells within the adjacent bone marrow were found to express BMP-6, of which a minority also expressed Cbfa1. Therefore, under in vivo conditions, BMP-6 may predominantly be expressed by early osteogenic cells, as opposed to osteoblasts, as represented by osteoblast-like cell lines used in this and related studies.
Although osteoblast-like cells used in this study also possess low levels of endogenous ER, both ERα and ERβ have been previously detected in these cells.(33–35) The requirement that artificial levels of ER and reporter genes are needed before a response can be detected is relatively commonplace and generally reflects the fact that response are too small to detect in the presence of low expression levels of ER and/or target genes. However, this strategy might also lead to spurious responses, and further investigations are required to confirm the physiological relevance of our findings, preferably using bone cell populations found to express significant levels of ERα and BMP-6 under in vivo conditions. For example, we recently found that in mouse long bones, ERα is expressed, not only by trabecular osteoblasts, but also by a subpopulation of adjacent bone marrow cells,(12) which may include the BMP-6-positive bone marrow stromal cells discussed above.
In summary, we have found that, although ERα stimulates the BMP-6 promoter in several cell types, the ligand dependency of this response varies according to the tissue from which the cell is derived, with antiestrogens stimulating activity in osteoblast-like cells and E2 in breast cancer and hepatoma cells. These differences in ligand specificity seemed to reflect differences in the molecular mechanisms involved in BMP-6 stimulation, because the responses to E2 and ICI were, respectively, dependent and independent of AF-2 and associated with nuclear and extranuclear localization of ERα. Taken together, these findings justify the need for further studies intended to investigate whether the tissue-selective effects of antiestrogens on bone are a consequence of the fact that ERα stimulates target genes like BMP-6 through a distinct extranuclear activation pathway in osteoblasts.
We thank Malcolm Parker for useful discussions. This project was supported by a PhD studentship for DO provided by Hunter Fleming, Ltd. and a project grant from the United Bristol Hospitals Trust Special Trustees.