Runx2 Integrates Estrogen Activity in Osteoblasts*

Steroids significantly effect skeletal integrity. For example, bone mass decreases with glucocorticoid excess or with estrogen depletion after menopause. Glucocorticoid suppresses gene expression by an essential skeletal tissue transcription factor, Runx2, in rat osteoblasts. We now report that estrogen enhances Runx2 activity in dose- and estrogen receptor-dependent ways independently of changes in Runx2 levels or its DNA binding potential. Estrogen receptor and Runx2 can be collected by co-immunoprecipitation. By two-hybrid gene expression analysis, high affinity complex formation involves portions of Runx2 outside of its own DNA binding domain and the DNA binding domain of the estrogen receptor. Consistent with this interaction, the stimulatory effect of estrogen on Runx2 activity is lost when the DNA binding domain of the estrogen receptor is eliminated. Unlike the stimulatory effect of estrogen and the inhibitory effect of glucocorticoid, androgen fails to increase Runx2 activity, whereas Runx2 potently suppresses gene expression induced by all three steroids. Finally, estrogen increases gene transcription by the transforming growth factor-β type I receptor gene promoter, which contains several Runx binding sequences, and enhances Smad dependent gene expression by transforming growth factor-β in osteoblasts. These results reveal that Runx2 can integrate complex effects on gene transcription in hormone-, growth factor-, and tissue-restricted ways.

Bone is a dynamic tissue that forms and remodels throughout life. Osteoblast-dependent bone formation varies with age, with certain metabolic disease states, or with pharmacological intervention. When osteoclast-dependent bone loss predominates, the structural quality of bone is diminished. This is accompanied by an increase in fracture and pain and a decrease in mobility and function. Loss of bone integrity and its subsequent pathology may therefore follow disparities in growth regulators and molecular events that define the normally balanced bone remodeling cycle (1)(2)(3).
Runx2, a member of the Runx family of nuclear transcription factors, first termed PEBP2␣, CBF␣, AML, or OSF, contains a Runt domain homologous to a transcription factor involved in body segmentation, sex determination, and neurogenesis in Drosophila melanogaster (4 -6). Runx2 levels increase greatly during osteoblast differentiation, whereas homozygous gene deletion of Runx2 limits bone formation, and few, if any, differentiated osteoblasts are found in Runx2-deficient mice (7). Importantly, the Runt domain contains sequences essential for DNA binding, and this region seems to be required for heterodimerization with several other nuclear transcription factors. The carboxyl-terminal region of Runx contains a transactivation domain as well as a domain that targets its binding to the nuclear matrix. In many instances, Runx transcription factors are thought to increase gene expression through Runxsensitive response elements (RE) 1 , although in some cases, Runx2 may also be inhibitory (6,8).
Steroid hormones also have complex stimulatory and inhibitory effects on gene expression. This is controlled in part by the way hormone-activated steroid receptors complex with themselves or with other binding partners to form competent or inactive transcription factors. Like other transcription factors, the steroid hormone receptors contain domains important for DNA binding, dimerization, and gene transactivation. Some effects by steroid hormone require direct genomic interactions, and others occur through indirect effects on cellular signaling (9 -14).
Several studies suggest that transient exposure to or a low level of the adrenal steroid glucocorticoid permissively alters the expression of some genes associated with osteoblast function (15,16). In contrast, high levels of glucocorticoid, such as those commonly used therapeutically to suppress inflammatory disease or tissue rejection, cause generalized bone loss leading to fractures as well as gastrointestinal and immunological problems (17,18). Although in some instances glucocorticoid may enhance Runx2 mRNA expression, in vitro glucocorticoid in excess rapidly depletes the amount of functional Runx2 nuclear protein, with an accompanying inhibitory effect on Runx2-dependent gene expression (19,20). This seems to reprise the downstream events that occur with glucocorticoid therapy or with Runx2 gene deletion in vivo (7) and would essentially limit osteoblast activity in the adult remodeling skeleton. Bone loss also occurs after menopause or with sex steroid depletion. In ovariectomized animals, bone resorption rates increase, but the balance in bone remodeling can then be restored by hormone replacement therapy. This is believed to occur in part by way of an estrogen-dependent reduction in the expression of lymphokines associated with the development of bone-resorbing osteoclasts through inhibitory effects on tran-scription factor C/EBP␤ (also termed NF-IL6) (21,22). Estrogen also regulates factors associated with bone formation. For example, estrogen suppresses IGF-I expression in osteoblasts stimulated with hormones that increase cAMP generation and activate transcription factor C/EBP␦ by decreasing C/EBP␦-dependent activation of the IGF-I gene promoter (23). However, estrogen may also enhance bone formation through inhibitory effects on osteoblast apoptosis or stimulatory effects on other osteoblast transcription factors (24 -27).
In this study, we show novel interactions between Runx2 and the estrogen receptor ␣ (ER␣). This interaction has potent regulatory effects that can control Runx, steroid hormone, and growth factor activity in osteoblasts and may have important implications for therapies that use steroid hormones or steroid hormone analogues.

EXPERIMENTAL PROCEDURES
Cell Cultures-Primary osteoblast-enriched cultures were prepared from parietal bones of 22-day-old Sprague-Dawley rat fetuses (Charles River Breeding Laboratories, Portage, MI) by methods approved by the Yale Institutional Animal Care and Use Committee. Bone sutures were dissected, and cells were released from the bone fragments by five sequential collagenase digestions. Cells pooled from the last three digestions express high levels of nuclear factor Runx2, parathyroid hormone receptor, type I collagen synthesis, and alkaline phosphatase activity (28 -30). They exhibit an increase in osteocalcin expression in response to vitamin D 3 , differential sensitivity to TGF-␤, bone morphogenetic protein 2, and various prostaglandins, and form mineralized nodules in vitro (31)(32)(33)(34)(35). Cells were plated at 4000/cm 2 in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. COS-7 cells (American Type Culture Collection) were cultured in identical medium and passaged at subconfluence.
Transfection Constructs-We created reporter plasmid 4XARE with four consensus androgen RE (ARE) within a minimal Rous sarcoma virus promoter plasmid. Plasmids SXN1C, containing two copies of a Runx RE from the rat TGF-␤ receptor I (T␤RI) gene promoter within pGL3-promoter (19,28), pEN1.0, containing a 1.0-kb fragment of the rat T␤RI gene promoter within pGL3-Basic (35,36), and antisense Runx2 expression plasmid, containing a 2.25-kb restriction fragment of murine Runx2 in reversed orientation within vector pSV7d (37,38), were described previously. Our studies also benefited from the generous gifts of reporter and expression plasmids from other investigators. Dr. Stuart Adler (Washington University, St. Louis, MO) provided an expression plasmid for human ER␣ and a reporter plasmid driven by estrogen RE. Dr. Chawnshang Chang (University of Rochester, Rochester, NY) provided an expression plasmid for rat androgen receptor (AR). Dr. Ronald M. Evans (Scripps Research Clinic, La Jolla, CA) provided MMTV-Luc, a reporter plasmid driven by glucocorticoid RE (GRE). Dr. Joan Massague (Sloan-Kettering Cancer Center, New York, NY) provided 3TP-Lux, a reporter plasmid with multiple AP-1 related RE and a minimal fragment of the plasminogen activator inhibitor 1 gene promoter. Dr. Bert Vogelstein (Johns Hopkins University, Baltimore, MD) provided reporter plasmid SBE4, driven by multiple Smad RE. Dr. Yoshiaki Ito (Kyoto University, Kyoto, Japan) provided an expression plasmid for murine Runx2 (originally termed PEBP2-␣A1). Dr. Ivan J. Sadowski (University of British Columbia, Canada) provided cloning vectors M1 with the GAL4 DNA binding domain (DBD) and MVN1 with the gene transactivation domain of Herpesvirus protein VP16. We used these vectors to create expression plasmids encoding either full-length ER␣ or Runx2, or specific fragments of these proteins, as indicated under "Results." Dr. Richard A. Maurer (Oregon Health Sciences University, Portland, OR) provided reporter plasmid 5XGAL4-E1b-Luciferase, with five GAL4 RE (5XGAL4).
Transfections-Promoter-reporter constructs, gene expression plasmids, or empty parental vectors, were pre-titrated for optimal expression efficiency and transfected with reagent LT1 (Mirus). Cultures at 50 -70% confluence were exposed to an optimal amount of expression plasmid (10 -45 ng/cm 2 ) or reporter plasmid (20 -50 ng/cm 2 ) in medium supplemented with 0.8% fetal bovine serum for 16 h, and then supplemented to obtain a final concentration of 5% serum. Cells were cultured for another 48 h, treated in serum-free medium, rinsed, and lysed. Nuclear-free supernatants were analyzed for reporter gene activity and corrected for protein content. To account for competition among plasmids for limiting transcriptional components, control cultures were transfected with a compensating amount of empty vector. Transfection efficiency was assessed in parallel in cells transfected with positive and negative reporter plasmids as described previously (16,23,(35)(36)(37).
Protein Extracts-Cells were rinsed, harvested by scraping and centrifugation, and lysed in hypotonic buffer supplemented with phosphatase and protease inhibitors and 1% Triton X-100 as described previously. Nuclei and cytoplasm were separated by centrifugation. Nuclei were extracted in hypertonic buffer with glycerol and phosphatase and protease inhibitors, and the nuclear proteins released in this way were separated from insoluble material by centrifugation (16,23,37).
Electrophoretic Mobility Gel Shift Assay-Double-stranded DNA probes were prepared by annealing complementary oligonucleotides. Overhangs were filled with dNTPs and radiolabeled with [␣-32 P]dCTP using the Klenow fragment of DNA polymerase I (New England Biolabs). Nuclear extract was incubated on ice with 32 P-labeled probe. In some samples, nuclear extract was pre-incubated with antiserum for 1 h before adding 32 P-labeled probe. Radioactive complexes were resolved on a 4% nondenaturing polyacrylamide gel and visualized by autoradiography (16,23,37).
Immunoprecipitations and Western Blots-Cell extracts were combined with pre-immune IgG or specific antibody and immunoprecipitates were collected with Protein A Sepharose (Pierce). Total cell extract or immunoprecipitates were fractionated by electrophoresis on a 12% SDS-polyacrylamide gel, blotted onto Immobilon P membranes (Millipore), probed with primary antibody, and visualized with secondary antibody and chemiluminescence (16,37).
TGF-␤ Activity-TGF-␤ sensitivity was determined with promoterreporter plasmids SBE4 and 3TP-Lux. SBE4 exhibits a high level of basal activity in fetal rat osteoblasts, and its activity is enhanced 50 -100% by TGF-␤1. In contrast, basal 3TP-Lux activity is minimal in osteoblasts and is highly enhanced by TGF-␤1 treatment (38). Osteoblasts were transfected, cultured to confluence, serum-deprived, and treated for 24 h with estradiol as indicated. Control or estradiol-treated cells were then challenged with maximally effective amount (120 pM) of TGF-␤1 (Bristol-Myers Squibb) for a second 24-h period (31).
Statistics-Statistical differences were assessed by one-way analysis of variance and Student-Newman-Keuls post hoc analysis, using Sig-maStat software (SPSS Inc., Chicago, IL). A significant difference was assumed by a p value of Ͻ0.05.

ER-dependent Gene Expression in Cultured
Osteoblasts-ER␣ expression parallels late stage osteoblast differentiation and mineralization in culture (39 -42). Consistent with this, primary cultures of fetal rat osteoblasts express little or no endogenous ER␣ at a time in culture when they are susceptible to transfection with plasmids encoding steroid dependent reporters or regulatory gene products (Fig. 1A). Nonetheless, they respond rapidly and potently to estradiol after transfection with an expression plasmid encoding ER␣ (Fig. 1B). Therefore, these easily obtained and well characterized osteoblast cultures can be made estrogen responsive in vitro, replicating the hormone-sensitive status of osteoblasts that occur in more mature organisms in vivo (43)(44)(45)(46). Importantly, this osteoblast cell culture model provides a sensitive system to examine specific aspects of ER␣ on osteoblast activity uncomplicated by the presence of endogenous ERs.
Estrogen Enhances Runx2 Activity-Earlier studies revealed that even in the presence of functional ER␣, estradiol does not alter basal IGF-I expression. However, it dose dependently decreases the effect of hormones that enhance IGF-I synthesis by way of a C/EBP RE in exon I of the IGF-I gene (23,37). In direct contrast to its inhibitory effect on C/EBP activity, estradiol increased Runx dependent gene expression in osteoblasts. Treatment with estradiol at 0.1-10 nM significantly enhanced gene expression driven by plasmid SXN1C, where two copies of a Runx RE from the T␤RI gene promoter were inserted into pGL3-promoter but had no effect on parental pGL3-promoter activity ( Fig. 2A). Stimulation seemed to be largely the result of an increase in endogenous Runx2 activation because estradiol caused an analogous effect when osteoblasts were transfected to express plasmid M1-Runx2, in which Runx2 was tagged with the GAL4 DBD. In this context, M1-Runx2 can drive gene expression by co-transfected reporter plasmid 5XGAL4-E1b-Luciferase, which contains five GAL4 RE (5XGAL4), by 4 -5fold relative to empty M1 vector (37,47). Estradiol failed to enhance gene expression by parental vector M1 in the presence of ER␣ or by M1-Runx2 in the absence of ER␣ (Fig. 2B). Therefore, estradiol enhanced Runx2 transcriptional activity, whether Runx2 used its own DBD to drive a native Runx RE or used the added GAL4 DBD to drive the 5XGAL4 RE system. These results are consistent with no obvious stimulatory effects by estradiol on total Runx2 levels in this short time frame or on the ability of Runx2 to associate with DNA that contains a consensus Runx RE (Fig. 2C).
ER␣ Forms Dimers with Runx2-Runx transcription factors can form heterodimers with other transcription factors and modify gene expression in complex ways (6, 8, 37, 48 -51). We found heterodimers of ER␣ and Runx2 in extract from COS-7 cells co-transfected with plasmid MVN1-ER␣, where ER␣ was tagged with the transactivation domain of Herpesvirus protein VP16, and plasmid M1-Runx2. Protein that co-precipitated with anti-VP16 antibody was reactive with anti-Runx2 antibody by Western blot, but only precipitating antibody was detected in extracts from cells transfected to express MVN1-ER␣ and parental plasmid M1 (Fig. 3A). Moreover, by the two-hybrid dependent gene expression assay performed in COS-7 cells co-transfected also with reporter plasmid 5XGAL4, the complex formed by MVN1-ER␣ and M1-Runx2 increased reporter gene expression by 3-fold relative to MVN1-ER␣ and parental plasmid M1. MVN1-ER␣ also associated with amino-terminal and carboxyl-terminal fragments of Runx2, as indicated in Fig. 3B, whereas no complex was apparent with the Runt domain of Runx2 by itself. Functional complexes that together activate the two-hybrid gene expression system occurred even in the absence of estradiol, as shown here, inasmuch as MVN1-ER␣ acquires an ancil-lary nuclear localization domain encoded by the MVN1 transfection vector (47).
Importantly, in addition to COS-7 cells, similar complexes between MVN1-ER␣ and M1-Runx2 or its fragments occurred when the two-hybrid gene expression assay was performed in primary cultures of osteoblasts (Fig. 4A). To define the regions of ER␣ that associate with Runx2, MVN1 fusion proteins that contained fragments of ER␣, as indicated in Fig. 4B, were then co-transfected with full-length M1-Runx2. Functional complex formation with Runx2 occurred with only discrete fragments of ER␣. Notably, the C region or the DBD of ER␣ alone seemed to be a highly proficient binding partner by comparison with other ER␣ domains.
Combined but Independent Effects by Estrogen and Protein Kinase A Activation on Runx2-Like their stimulatory effects on C/EBP␦, prostaglandin E2 (PGE2) and parathyroid hormone enhance Runx2 activity in a protein kinase A-dependent manner (37,52). However, in direct contrast to the potent inhibitory effect of estrogen on protein kinase A-activated C/EBP␦, Runx2 activity increased even further in ER␣-transfected osteoblasts treated with both estradiol and PGE2 (Fig. 5A). Consistent with the results presented in Fig. 4, co-expression of MVN1-ER␣ increased M1-Runx2 activity in untreated cells, as defined by the two-hybrid gene expression system. However, treatment with estradiol and PGE2 further enhanced M1-Runx2 activity in complex with MVN1-ER␣, by comparison with either hormone alone (Fig. 5B, left).
To substantiate the importance of the ER␣ DBD for these effects, the C region was eliminated in plasmid MVN1-ER␣[-C]. In this context and without hormone treatment, M1-Runx2 activity (Fig. 5B, right) was reduced to approximately the level found in osteoblasts transfected with untagged ER␣ (Fig. 5A, right). Moreover, whereas the stimulatory effect of PGE2 on M1-Runx2 persisted in cells co-transfected with MVN1-ER␣[-C], estradiol failed to enhance M1-Runx2 activity without or with PGE2 (Fig. 5B, right). To address the C domain of ER␣ within the context of endogenous Runx2 activity, osteoblasts were co-transfected with plasmids containing the Runx-sensitive reporter plasmid SXN1C and either full-length ER␣ or ER␣[-C] subcloned into vector M1. In this way, the VP16 transactivation domain tag would not complicate activation of endogenous Runx2. Consistent with the results in Fig. 5B, SXN1C failed to respond to estradiol in cells transfected with M1-ER␣[-C] but remained sensitive to PGE2 (Fig. 5C). To confirm the loss of a functional C domain in ER␣[-C] plasmids, osteoblasts were co-transfected with M1-ER␣ or M1-ER␣[-C] and ERE reporter plasmid. M1-ER␣[-C] failed to increase estradiol-dependent promoter activity, inasmuch as gene expression in this system requires binding by the C domain of ER␣ to consensus ERE (Fig. 5D). However, the cells could clearly express plasmids containing ER␣[-C], because M1-ER␣[-C] was fully functional in osteoblasts co-transfected with 5XGAL4 reporter plasmid, where DNA binding and gene expression depend on the GAL4 DBD supplied by the M1 tag rather than the endogenous C domain of ER␣ (Fig. 5E). PGE2 did not further increase the stimulatory effect of estradiol on ER␣-dependent gene expression by way of an ERE-driven gene promoter (23).
Estrogen Specificity-To assess steroid specificity within the context of Runx2 activation, osteoblasts were then treated with androgen or glucocorticoid. Analogous to our findings with estradiol as shown in Fig. 1, fetal rat osteoblasts exhibit maximal sensitivity to androgen when transfected to express AR and reporter plasmid driven by consensus ARE (Fig. 6A) or   FIG. 1. ER-dependent gene expression in osteoblasts. A, nuclear extract from untransfected osteoblasts (Control) or from cells transfected (Tfx'd) to express recombinant ER␣ were examined by electrophoretic mobility shift assay with a 32 P-labeled probe encoding a consensus ERE. Extracts were incubated with either non-immune IgG, or ER␣ antibodies MC-20 (anti-mouse carboxyl terminus; mouse, human, and rat reactive) or H-184 (anti-human amino terminus; mouse, human, and rat reactive) (Santa Cruz Biotechnology). The identity of the invariant, non-reactive complexes in extracts from osteoblasts that lack ER␣ expression is unknown. B, osteoblasts were co-transfected with empty vector (V) or ER␣ and reporter plasmid containing consensus ERE and then treated for 24 h with 10 nM estradiol. Reporter gene expression was measured in cytoplasmic extracts and corrected for protein content. Estradiol significantly increased gene expression by ERE in ER␣-transfected cells. Data represent results from a minimum of nine replicate samples and three separate studies with different culture preparations.
MMTV-Luc. 2 However, cortisol potently activates GREdependent gene expression in these osteoblast cultures through endogenous glucocorticoid receptor (GR) (19,53) (Fig. 6A) and is not activated further by transfection with GR expression vector. 3 Unlike estradiol, androgen failed to activate Runx2 in osteoblasts transfected to express plasmid SXN1C, and consistent with earlier studies (19), glucocorticoid was inhibitory (Fig.  6B). Androgen also failed to activate M1-Runx2 by way of the heterologous 5XGAL4 reporter. However, expression of M1-Runx2 in the context of the 5XGAL4 reporter rescued osteoblasts from the loss of endogenous Runx2 that occurs with glucocorticoid treatment (19) (Fig. 6C).
Runx2 Suppresses Estrogen Activity-Because complex formation with Runx2 involves the DBD of ER␣ (Figs. 4 and 5), Runx2 could possibly counter-regulate estrogen activity. Indeed, forced expression of Runx2 dose-dependently decreased the effect of estradiol on reporter gene expression driven by ERE (Fig. 7A). Consistent with this, and the normal presence of Runx2 in osteoblasts (5,28,35), transfection with Runx2 antisense to limit endogenous Runx2 levels (37,38) further increased ERE reporter gene expression by 2.5-3-fold (Fig. 7B). Analogous inhibitory or stimulatory effects by forced expression of Runx2 or Runx2 antisense occurred on androgen-and glucocorticoid-dependent gene expression (Fig. 7, A and B). Therefore, Runx2 suppresses the activity of estradiol, glucocorticoid, and androgen, whereas these same steroid hormones independently increase, decrease, or have no effect on Runx2 activity. This suggests that control and counter-control between Runx2 and each of these steroids is focused in very different ways.
Estrogen Focuses TGF-␤ Activity-To determine whether an estrogen-dependent increase in Runx2 activity has downstream effects, osteoblasts were co-transfected with ER␣ and reporter plasmid driven by full-length T␤RI gene promoter that contains four Runx binding elements. Consistent with evidence that Runx2 accounts for ϳ50% of T␤RI gene expression in osteoblasts (28), estradiol activated the T␤RI gene promoter by 40 -50% relative to the 40 -50% decrease that occurs with glucocorticoid treatment (19) (Fig. 8A) and caused a corresponding  FIG. 2. Estrogen enhances Runx2 activity. A, osteoblasts were co-transfected with ER␣ and either parental plasmid pGL3-promoter or SXN1C, where pGL3-promoter was supplemented with two copies of Runx RE. Cells were then treated for 24 h with vehicle (0) or 0.1-10 nM estradiol. B, osteoblasts were co-transfected with empty expression vector (V) or ER␣, empty expression vector M1 or M1-Runx2, in which Runx2 was fused to the GAL4 DBD, and reporter plasmid 5XGAL4, containing five copies of a GAL4 RE. Cells were then treated for 24 h with vehicle (Ϫ) or 10 nM estradiol (E). A and B, reporter gene expression was measured in cytoplasmic extracts and corrected for protein content. Estradiol significantly increased reporter gene activity by SXN1C and by 5XGAL4 in combination with M1-Runx2 in ER␣-transfected cells. Data represent results from a minimum of nine replicate samples and three separate studies with different culture preparations. C, left, nuclear extract from ER␣-transfected cells treated for 24 h with vehicle (Ϫ) or 10 nM E were examined by electrophoretic mobility shift assay (EMSA) with a 32 P-labeled probe encoding a consensus Runx RE. C, right, nuclear extracts were fractionated on a polyacrylamide gel and visualized by Western blot (WB) with antibody to Runx2 (Santa Cruz Biotechnology).

FIG. 3. ER associates with Runx2.
In A, COS-7 cells were transfected with empty expression vector MVN1 (V) or MVN1-ER␣, where ER␣ was fused to the Herpesvirus VP16 transactivation domain, and M1-Runx2. Either total nuclear extract (left) or extract collected with anti-VP16 antibody (right), were examined by polyacrylamide gel fractionation and Western blot (WB) with antibody to Runx2 (Santa Cruz Biotechnology Inc.). B, COS-7 cells were co-transfected with MVN1-ER␣ and empty expression vector M1, M1-Runx2, or the indicated fragments of Runx2 fused to the GAL4 DBD, and reporter plasmid 5XGAL4. Reporter gene expression was measured in cytoplasmic extracts and corrected for protein content. MVN1-ER␣ significantly increased 5XGAL4 reporter gene activity by M1 fusion proteins that encompassed amino acids 1-513 (full-length M1-Runx2), 1-92, or 228 -513 of murine Runx2. Data represent results from a minimum of nine replicate samples and three separate studies with different culture preparations. increase in T␤RI protein (Fig. 8B). Furthermore, the stimulatory effect of estradiol on Runx2 activity and on T␤RI expression allowed an analogous increase in Smad-dependent gene expression in response to TGF-␤ in cells co-transfected with plasmid SBE4 (Fig. 8C). However, although estradiol enhanced T␤RI gene promoter activity and the TGF-␤-stimulated Smad pathway, it suppressed reporter gene expression by 3TP-Lux in TGF-␤-treated cells (Fig. 8D), perhaps by its previously noted inhibitory effect on AP-1 activity in some situations (54 -56). DISCUSSION Although many studies focus on estrogen as an inhibitor of bone resorption, others suggest additional effects on bone for-mation. Recent evidence predicts that estrogen may act in part on osteoblasts to inhibit apoptosis (24 -27) or to regulate AP-1-related transcription factors by specific kinase-dependent pathways (25,27,57). In this report, we show that estradiol increases gene transactivation by Runx2, an essential transcription factor for skeletal tissue development (5,6). The effect of estradiol requires a functional ER and occurs without changes in Runx2 expression or its affinity for DNA. Moreover, the stimulatory effect of estradiol on Runx2 activity is further enhanced by PGE2, a potent kinase activator in osteoblasts (34,58), although their primary mechanisms of action are independent.
It is important to note that many aspects of our current study involved exogenous gene transfection. This required the use of cells at a stage in culture at which they do not yet express high levels of ERs (39 -42). To correct this, we transiently transfected or co-transfected osteoblasts to express ER␣. Other investigators addressed this deficiency by creating osteoblastlike cell lines that were stably transfected to express ER␣ under selective pressure from antibiotics or transforming viral gene elements. We chose to avoid possible complications from antibiotic toxicity, unknown effects from stable gene integration, and phenotypic drift by cells in continuous culture with our approach. Even so, it is difficult to know how the level of ER␣ expression in any transfected cell model compares with that in adult bone because, even in vivo, levels of ER␣ vary considerably with age and with anatomical bone location (43)(44)(45)(46). However, expression of ER␣ by transient gene transfection seemed appropriate and predictable, because the ability of estrogen to driven reporter gene expression by ERE was steroid-FIG. 6. Estrogen specific effects on Runx2. A, osteoblasts were co-transfected with ER␣ and ERE, AR and ARE, or MMTV-Luc containing multiple GRE, and then treated for 24 h with the amounts of estradiol, androgen, or cortisol indicated. B, osteoblasts were co-transfected with appropriate steroid hormone receptor and either reporter plasmid pGL3-promoter or SXN1C. C, osteoblasts were co-transfected with appropriate steroid hormone receptor, vector M1 or M1-Runx2, and reporter plasmid 5XGAL4. B and C, cells were then treated for 24 h with vehicle (Ϫ) or a maximally effective amount of hormone determined from A: 10 nM estradiol (E), 1 nM androgen (A), or 100 nM cortisol (G). Reporter gene expression was measured in cytoplasmic extracts and corrected for protein content. A, estradiol at or above 0.1 nM significantly increased reporter gene expression by ERE; androgen (methytrienolone, PerkinElmer Life Sciences) at or above 0.1 nM significantly increased reporter gene expression by ARE; and glucocorticoid at or above 1 nM significantly increased reporter gene expression by GRE. B and C, estradiol significantly increased reporter gene activity by SXN1C and by 5XGAL4 in combination with M1-Runx2 in ER␣-transfected cells, and glucocorticoid significantly decreased increased reporter gene activity by SXN1C. Data represent results from a minimum of nine replicate samples and three separate studies with different culture preparations.

FIG. 7. Runx2 inhibits steroid activity.
A, osteoblasts were cotransfected with ER␣ and ERE, AR and ARE, or GRE, and the amounts of Runx2 expression plasmid (pRunx2) are indicated. B, osteoblasts were co-transfected with ER␣ and ERE, AR and ARE, or GRE, and a maximally effective level (200 ng (37,38)) of Runx2 antisense expression plasmid (aS Runx2). Cells were then treated for 24 h with vehicle (Ϫ) or a maximally effect level of 10 nM estradiol (E), 1 nM androgen (A), or 100 nM cortisol (G). A and B, total plasmid loading was held constant by complementation with empty expression vector. Reporter gene expression was measured in cytoplasmic extracts and corrected for protein content. A, pRunx2 significantly decreased reporter gene activity driven by each steroid hormone. B, aS Runx2 significantly increased reporter gene activity induced by each steroid hormone. Data represent results from a minimum of nine replicate samples and three separate studies with different culture preparations. specific 4 and varied by less than 10% in our studies.
Other reports suggest links between estrogen and Runx2 through an increase in the number of Runx2-expressing cells (59). Furthermore, selective ER modulators with modes of action that are often distinct from estrogen itself seem to enhance Runx2 gene promoter activity (60,61). We found no increase in Runx2 protein or its DNA binding potential after estrogen treatment. Important differences between estrogen and its mimetics may result in part from variations in ER conformation that occur after engagement with different ligands, or from their abilities to activate or suppress different genomic RE (62)(63)(64). Moreover, effects on Runx2 gene promoter activity or mRNA expression may not readily correspond to the level of functional Runx2 protein or to its gene activation potential (19,37,52,65,66). Our studies therefore identify a new pathway that may account in part for the anabolic effect of estrogen on skeletal tissue cells and a decrease in bone formation after estrogen withdrawal.
We found that ER␣ and Runx2 interact through the DBD of ER␣. Interactions have been reported between ER␣ and several other transcription factors, including retinoid receptors (67), forkhead transcription factors (68), Smads (69,70), and Stat5 (71). The interaction between ER␣ and Stat5 also maps to the ER␣ DBD (71). Other transcriptional regulators seem to control Runx transcription factors by an interaction that requires, in whole or in part, the Runt domain (4,6,8,37,72), whereas ER␣ does not seem to associate in an efficient way with the Runt domain of Runx2 by itself (current study). As modeled in Fig. 9, the focused interaction that we find with the DBD of ER␣ retains and expands Runx2 activity, although it limits ER␣-dependent gene expression through consensus ERE. This distinction between direct and indirect control of gene expression by ER␣ is reminiscent of mutations in GR␣ that prevent its ability to form homodimers. In this instance, a lack of gene expression through GRE is compatible with life, whereas overall gene deletion of GR␣ is lethal (73,74). Similarly, inactivating mutations in the ER␣ gene in mice cause steroidogenesis and selective defects in reproductive and mammary tissue (71,75). Thus, effects on genes driven by elements other than the consensus steroid RE also seem to be critical, if not essential, targets of steroid hormone action in bone-forming cells.
Our results therefore predict that Runx2 expression in growing or remodeling bone in part limits those aspects of estrogen activity that require binding by ER␣ directly to DNA but expands its gene targets in a tissue-restricted way. Runx2 also suppresses gene expression through ARE and GRE, even though androgen has no effect on Runx2 activity in osteoblasts (current study) and glucocorticoid is inhibitory (19). Consequently, changes in the balance of Runx2 and steroid hormone receptors in osteoblasts could integrate gene expression in complex but focused ways. For example, a high level of glucocorticoid could significantly limit development of the osteoblast phenotype in bone cells where Runx2 expression is evolving but 4 M. Centrella and T. L. McCarthy, unpublished observations. FIG. 8. Estrogen focuses TGF-␤ activity. A, osteoblasts were co-transfected with ER␣ and reporter plasmid driven by full-length promoter DNA from the rat T␤RI gene (35,36) and then treated for 24 h with vehicle (0), 10 nM estradiol (E), or 100 nM cortisol (G), as indicated. B, cell extracts were fractionated on a polyacrylamide gel and visualized by Western blot (WB) with antibody to T␤RI (Santa Cruz Biotechnology). C, osteoblasts were co-transfected with ER␣ and reporter plasmid SBE4, driven by multiple Smad RE. D, osteoblasts were co-transfected with ER␣ and reporter plasmid 3TP-Lux, driven by multiple AP-1 RE. C and D, the cells were pre-treated for 24 with vehicle (Ϫ) or 10 nM estradiol and then for 24 with vehicle or 120 pM TGF-␤1. Reporter gene expression was measured in cytoplasmic extracts and corrected for protein content. A, T␤RI gene promoter activity was significantly increased by estradiol and significantly decreased by cortisol. C, estradiol significantly increased the stimulatory effect of TGF-␤ on SBE4 reporter gene activity. D, estradiol significantly decreased the stimulatory effect of TGF-␤ on 3TP-Lux reporter gene activity. Data represent results from a minimum of nine replicate samples and three separate studies with different culture preparations.
FIG. 9. Runx2 integrates ER-dependent gene expression. A conventional model of estrogen-dependent effects on gene transcription indicates that estrogen associates with the ligand-binding domain (LBD) of ER␣, which can dimerize, bind consensus EREs, and drive gene expression in a way that requires the gene transactivation (TA) domain of ER␣. In the current article, we report evidence that ER␣ can heterodimerize with transcription factor Runx2, at least in part through the DBD of ER␣. This interaction enhances Runx2 activity on gene expression through Runx response elements (RRE) in an estrogendependent manner but blocks the stimulatory effect of ER␣ on gene expression through consensus ERE.
have negligible effects on Runx2-sensitive genes in more differentiated osteoblasts, where Runx2 levels are high. Consistent with this, the inhibitory effect of glucocorticoid was not evident in osteoblasts, in which Runx2 was sustained by gene transfection. Protein hybridization and domain mapping studies showed that Runx2 also interacts with the DBD of GR, 4 consistent with its ability to suppress gene expression at GRE. Again, the interaction between androgen and Runx2 in osteoblasts seems to be limited to inhibitory effects on AR activity through ARE. Our results are similar to those of earlier studies showing complex formation between full-length Runx2 fusion protein and either AR or GR. However, they also differ in part because those authors found that overexpression of Runx2 in monkey kidney cells failed to suppress the effect of androgen or glucocorticoid on gene expression driven by steroid hormone RE fused to a fragment of the thymidine kinase gene promoter (51). Nevertheless, in osteoblasts, Runx2 suppressed gene expression driven by consensus ERE, ARE, or MMTV-Luc, a well characterized androgen and glucocorticoid sensitive reporter plasmid. We have not yet mapped interacting domains between Runx2 and AR to learn whether this is similar to our current findings with ER and GR.
One of several target gene products for Runx2 in osteoblasts is T␤RI (28,35). Indeed, loss of Runx2 by glucocorticoid excess (19) or by transfection with Runx2 antisense expression plasmid (38) limits T␤RI expression and suppresses the stimulatory effect of TGF-␤ on collagen synthesis. This would therefore reduce the amount of the major organic component of the framework for mineral deposition in bone and subsequently limit bone formation or repair. Thus, anabolic effects by estrogen may occur in part through changes in the balance of T␤RI expression that maintains or enhances the effectiveness of local TGF-␤ in the bone environment. Unlike the inhibitory effects of estradiol on TGF-␤-sensitive and bone morphogenetic proteinsensitive Smads that occur in mesangial and breast cancer cells (69,70), estradiol enhances Smad-dependent gene expression in response to TGF-␤ in osteoblasts. Interactions between ERs and AP-1 transcription factors are also complex, because estrogen can either increase or decrease the expression of AP-1sensitive genes (56,57,71,76). We found that estrogen reduced TGF-␤ activation of the AP-1-sensitive promoter 3TP-Lux. Therefore, estrogen also seems to distinguish specific aspects of TGF-␤ activity that are downstream of Runx2 activation and T␤RI expression.
In summary, our studies present new evidence for anabolic effects by estrogen on bone-forming cells through an increase in gene activation by transcription factor Runx2. In contrast, Runx2 limits the so-called "classic" pathway of steroid hormone action by estrogen, androgen, and glucocorticoid. Therefore, changes in the relative synthesis of Runx2 and steroid hormone receptors may regulate the expression of "classic" and "nonclassic" steroid-sensitive gene products during specific phases of osteoblast development in different ways. We predict that similar important interactions may also occur with steroid hormones in those cells in which other Runx gene family members have critical effects on tissue-restricted gene expression.