Estrogens Attenuate Oxidative Stress and the Differentiation and Apoptosis of Osteoblasts by DNA-Binding-Independent Actions of the ERα

Estrogens diminish oxidative stress in bone and bone marrow, attenuate the generation of osteoblasts, and decrease the prevalence of mature osteoblast apoptosis. We have searched for the molecular mechanism of these effects using as tools a mouse model bearing an estrogen receptor α (ERα) knock-in mutation that prevents binding to DNA (ERαNERKI/−) and several osteoblast progenitor cell models expressing the wild-type ERα or the ERαNERKI/−. We report that the ability of estrogens to diminish the generation of reactive oxygen species, stimulate the activity of glutathione reductase, and decrease the phosphorylation of p66shc, as well as osteoblastogenesis and osteoblast number and apoptosis, were fully preserved in ERαNERKI/− mice, indicating that the DNA-binding function of the ERα is dispensable for all these effects. Consistent with the attenuation of osteoblastogenesis in this animal model, 17β-estradiol attenuated bone morphogenetic protein 2 (BMP-2)–induced gene transcription and osteoblast commitment and differentiation in murine and human osteoblastic cell lines. Moreover, 17β-estradiol attenuated BMP-2-induced differentiation of primary cultures of calvaria- or bone marrow–derived osteoblastic cells from ERαNERKI/− mice as effectively as in cells from wild-type littermates. The inhibitory effect of the hormone on BMP-2 signaling resulted from an ERα-mediated activation of ERKs and the phosphorylation of Smad1 at the linker region of the protein, which leads to proteasomal degradation. These results illustrate that the effects of estrogens on oxidative stress and the birth and death of osteoblasts do not require the binding of ERα to DNA response elements, but instead they result from the activation of cytoplasmic kinases. © 2010 American Society for Bone and Mineral Research


Introduction
W ork from our group during the last 20 years has elucidated that estrogens protect the adult skeleton against bone loss by slowing the rate of bone remodeling and by maintaining a focal balance between bone formation and resorption. (1)(2)(3) Slowing of bone remodeling results from the attenuating effects of sex steroids on the birth rate of osteoclast and osteoblast progenitors. (4,5) Maintenance of a focal balance between formation and resorption results from opposite effects on the lifespan of osteoclasts and osteoblasts/osteocytes: a proapoptotic effect on osteoclasts and an antiapoptotic effect on osteoblasts and osteocytes. (6)(7)(8) The effects of estrogens on osteoclast and osteoblast apoptosis are exerted by a mechanism that is distinct from that requiring direct interaction of their receptors with DNA (hormone-response element) or proteinprotein interaction between the receptor and other transcription factors. Instead, the effect of estrogens on the apoptosis of either cell type is the result of an extranuclear action of the classical receptors that cause activation of cytoplasmic kinases, including extracellular signal-regulated kinases (ERKs) and kinase-dependent changes in the activity of transcription factors. (6,8,9) The mechanistic basis for the divergence of estrogens' effect on the survival of the two cell types downstream from ERKs is evidently dependent on the kinetics of ERK phosphorylation and the length of time that phospho-ERKs are retained in the nucleus, perhaps by determining the activation of a distinct set of transcription factors. (9) growth factors from the bone matrix) is not required for the increase in osteoblast precursors. Therefore, estrogens must suppress osteoblastogenesis by direct actions on early osteoblast precursors. Further, we have shown that most CFU-OBs are early transit-amplifying progenitors (i.e., dividing cells with limited self-renewal capacity) and that their replication is indeed attenuated by estrogens. (5) We and others also have shown previously that estrogens attenuate the transcription of bone morphogenetic protein 2 (BMP-2) target genes. (11)(12)(13) BMPs are members of the transforming growth factor b (TGF-b) superfamily and play an essential role in skeletal development and repair. (14,15) Specifically, BMPs promote embryonic and postnatal osteogenesis by inducing the commitment of mesenchymal cells to the osteoblastic lineage and promoting osteoblast differentiation. (16,17) Binding of BMPs to their receptor serine/threonine kinases results in the phosphorylation of Smads 1, 5, and 8 (18) at the carboxy terminus and translocation into the nucleus after heterodimerization with Smad4. In the nucleus, the complex either binds to DNA sequences directly or can interact with several transcription factors to control the activity of hundreds of downstream target genes. (19,20) The Smad proteins consist of two globular domains (MH1 and MH2 domains) connected by a linker region. In Smad 1, 5, and 8, the latter contains four MAPK phosphorylation sites and two putative GSK-3b sites. (19,21) Importantly, MAPK phosphorylation of the linker region inhibits Smad function and therefore BMP-induced transcription both in vitro and in vivo. (22)(23)(24) More recently, we and others have obtained evidence that the protective effects of estrogens on bone result from their ability to attenuate oxidative stress and that loss of estrogens accelerates the effects of aging. Specifically, we have shown that C57BL/6 mice lose bone strength and mass progressively between the ages of 4 and 31 months (25) and that these changes are temporally associated with decreased osteoblast and osteoclast numbers and decreased bone-formation rate as well as increased osteoblast and osteocyte apoptosis. These changes are also temporally linked with increased reactive oxygen species (ROS) levels and decreased glutathione reductase (GSR) activity in the bone marrow, as well as a corresponding increase in the phosphorylation of p66 shc -an adapter protein that serves as a key component of a signaling cascade that is activated by ROS and influences apoptosis and lifespan in invertebrates and mammals. (26) Indeed, proapoptotic signals, including ROS, release p66 shc from an inhibitory complex, and active p66 shc serves as a redox enzyme that catalyzes reduction of O 2 to H 2 O 2 through electron transfer from cytochrome c. H 2 O 2 , in turn, causes opening of the mitochondrial permeability transition pore, swelling, and apoptosis. An increase in ROS production and p66 shc phosphorylation, as well as decreased GSR activity, was reproduced acutely in our previous work by gonadectomy in either female or male C57BL/6 mice and prevented by the antioxidant N-acetyl-cysteine (NAC). (25) In agreement with our in vivo findings, results of in vitro experiments demonstrated that the effects of estrogens on osteoclastogenesis and osteoclast and osteoblast apoptosis result from cell autonomous antioxidant actions of the hormone on the respective cell types, are dependent on the estrogen receptor (ER), and are mediated via ERKs. Moreover, we have shown that estrogens attenuate the phosphorylation of p66 shc in osteoblastic cells and that this effect is also mediated via ERKs. Based on these findings, we have proposed that loss of estrogens accelerates the effects of aging on bone by decreasing the defense against oxidative stress.
In this study we have investigated the molecular actions of ERa on osteoblasts using as a tool a mouse model bearing an ERa knock-in mutation that prevents binding to DNA (ERa NERKI/À ). (27) We previously determined that ERa NERKI/À mice have an atrophic uterus despite normal estrogen levels and that estrogen replacement does not restore it in ovariectomized (OVX) ERa NERKI/À mice, but it does induce the activation of ERKs and the ERK-mediated phosphorylation of the transcription factor Elk-1 in vertebrae. (13) In addition, in this study we have investigated in vitro the signaling cascades downstream from ERa that are responsible for its effects on osteoblast commitment and differentiation. We show that the effects of estrogens on oxidative stress and the birth and death of osteoblasts are fully preserved in ERa NERKI/À mice. Consistent with the attenuation of osteoblastogenesis in the ERa NERKI/À mice, 17b-estradiol (E 2 ) attenuates BMP-2-induced gene transcription and differentiation of preosteoblastic cell lines as well as primary cultures of calvaria-or bone marrow-derived osteoblastic cells from ERa NERKI/À mice as effectively as in cells from wild-type littermates. This effect is due to an ERa-mediated activation of ERKs and the phosphorylation of Smad1 at the linker region of the protein, which leads to proteasomal degradation.

Chemicals, reagents, and plasmids
Etoposide, H 2 O 2 , and E 2 were purchased from Sigma-Aldrich (St. Louis, MO, USA). BMP-2 and fibroblast growth factor 2 (FGF2) recombinant proteins were purchased from R&D Systems (Minneapolis, MN, USA). PD98059 was purchased from Cell Signaling Technology, Inc. (Danvers, MA, USA). The BMPresponsive luciferase reporter construct (BRE)-Luc was obtained from Peter ten Dijke (Leiden University Medical Center, Leiden, The Netherlands). (28) Animal experimentation Mice heterozygous for an ERa knock-in mutant in a 129SvJ background were provided by J Larry Jameson (Northwestern University, Chicago, IL, USA). (27) Mice harboring an inactivating mutation in the ERa locus (ERa þ/À ) in a C57BL/6 background were provided by Andree Krust and Pierre Chambon (Institute for Genetics and Cellular and Molecular Biology, Strasbourg, France). (29) ERa NERKI/þ mice were crossed with heterozygous ERa þ/À female mice to produce animals carrying only one NERKI allele (ERa NERKI/À ). Five-month-old female ERa NERKI/À mice of the F 1 generation and their ERa NERKI/þ , ERa þ/À , and ERa þ/þ littermates were electronically tagged (Biomedic Data System, Inc., Maywood, NJ, USA), and bone mineral density (BMD) measurements were performed on each mouse. Animals then were sham-operated or ovariectomized (OVX). The following day, OVX animals were subcutaneously injected with vehicle or with replacement doses of E 2 (30 ng/g/day; n ¼ 12 per group), and sham-operated animals were injected with vehicle. After 6 weeks of treatment, animals were sacrificed and the tissues dissected for further analyses. BMD, vertebral dimensions, osteoblast number, and osteoblast apoptosis were obtained as described previously. (7,30,31) ERa À/À and corresponding wildtype (WT) littermate control mice were generated by crossing heterozygous ERa þ/À mice.

Western blot analysis
The phosphorylation status of p66 shc was analyzed by immunoblotting in fifth lumbar vertebra lysates, as described previously, (8) using a mouse monoclonal antibody recognizing Ser36 phosphorylated p66 shc (Calbiochem, San Diego, CA, USA). Protein levels of p66 shc were analyzed using a rabbit polyclonal antibody recognizing p66 shc (BD Biosciences, Palo Alto, CA, USA). The antibody recognizing p-Smad1/5/8 was purchased from Cell Signaling. The antibody recognizing phospho-Smad1 (Ser214) was kindly provided by EM De Robertis (Howard Hughes Medical Institute and University of California, Los Angeles, CA, USA). (32) The antibodies against Smad4, p-ERK1/2, and ERK1/2 were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Quantification of the intensity of the bands in the autoradiograms was performed using a VersaDoc imaging system (Bio-Rad Laboratories, Hercules, CA, USA).

Micro-Computed Tomography (mCT)
mCT analysis of the sixth lumbar vertebrae was done after the bones were dissected, cleaned, fixed in 10% Millonig's formalin, transferred to ethanol, loaded into 12.3 mm diameter scanning tubes, and imaged (mCT40, Scanco Medical, Basserdorf, Switzerland). Scans were integrated into 3D voxel images (1024 Â 1024 pixel matrices for each individual planar stack), and a Gaussian filter (s ¼ 0.8, support ¼ 1) was used to reduce signal noise. A threshold of 200 was applied to all analyzed scans. Scans were done at medium resolution (E ¼ 55 kVp, I ¼ 145 mA, integration time ¼ 200 ms). The entire vertebral body was scanned with a transverse orientation excluding the pedicles and articular processes. Manual analysis excluded the cortical bone and the primary spongiosa from the analysis. All trabecular measurements were made by manually drawing contours every 10 to 20 slices and using voxel counting for bone volume per tissue volume and sphere-filling distance-transformation indices without assumptions about the bone shape as a rod or plate for trabecular microarchitecture. Cortical thickness was measured at the tibial mid-diaphysis.
Cell culture, transfections, and luciferase activity Osteoblastic cells derived from mouse calvaria or bone marrow were obtained and cultured as described previously, (33) and during exposure to E 2 , the cultures were maintained in 2% charcoal-stripped serum. Osteoblast differentiation was analyzed using freshly isolated cells cultured in 12 well tissue culture plates at 5 Â 10 6 cells per well in a modified essential medium (a-MEM) containing 10% fetal bovine serum (FBS) for 10 days. Half the medium was replaced every 5 days. FBS then was reduced to 2%, and 10 À8 M E 2 was added in the presence or absence of 25 ng/ mL BMP-2. Two days later, 10 mM b-glycerophosphate was added to the medium, and the cultures were maintained for an additional 2 weeks. The mineralized matrix was stained with 40 mM alizarin red, pH 4.2. Alizarin red was quantified after extraction with 10 mM sodium phosphate, 10% cetylpyridinium chloride, pH 7, and absorbance determination at 562 nm against a known alizarin red standard. For assay of caspase 3 activity, the medium was changed to serum-free prior to the addition of the different compounds. Colony-forming units-fibroblast (CFU-F) and CFU-OB number were determined as described previously, (5) using guinea pig feeder cells, (34) 15% FBS, and 1 mM ascorbate-2phosphate. Half the medium was replaced every 5 days. CFU-Fs were enumerated at 10 days of culture after staining for alkaline phosphatase, and CFU-OBs were enumerated at 25 days of culture after von Kossa staining. Colonies containing more than 50 fibroblastic cells were enumerated and plotted as a function of the number of cells seeded. C2C12 and 2T3 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% FBS and 1% each of penicillin, streptomycin, and glutamine and 1% sodium pyruvate. U2OS cells stably expressing tetracycline-inducible ERa (U2OS-ERa) were kindly provided by DC Leitman (University of California, San Francisco, CA, USA). (35) U2OS-ERa cells were maintained in phenol red-free McCoy medium supplemented with 10% FBS and 1% each of penicillin, streptomycin, and glutamine. Cells were incubated for 24 hours with or without doxycycline (1 mg/mL) and serum starved for another 16 hours previous to the addition of BMP-2 or E 2 . Mouse embryonic fibroblasts (MEFs) from WT or Smad1 L/L mice, kindly provided by P Soriano (Fred Hutchinson Cancer Research Center, Seattle, WA, USA), (24) were cultured in DMEM supplemented with 10% FBS. Plasmid constructs were introduced into cells by transient transfection using Lipofectamine Plus (Invitrogen, Carlsbad, CA. USA). Cells were plated in 48 well plates and transfected 16 hours later with a total of 0.4 mg of DNA. Luciferase activity assays were performed as described previously. (36) Alkaline phosphatase (AP) activity and osteocalcin production C2C12 or 2T3 cells were seeded at a density of 2 Â 10 4 /cm 2 in medium containing 10% FBS. The following day, before treatment, the medium was replaced with 5% serum-containing medium. Cells were lysed in 100 mM glycine, 1 mM MgCl 2 , and 1% Triton X-100 at pH 10. AP activity in the cell lysate was determined using a buffer containing 2-amino-2-methylpropanol and p-nitrophenylphosphate (Sigma-Aldrich. Inc.). The amount of osteocalcin secreted in the medium was determined using an ELISA kit (Biomedical Technologies, Inc., Stoughton, MA, USA). Both activities were normalized for total protein concentration, determined using a Bio-Rad DC protein assay kit.

Mineralization assay
Freshly isolated murine bone marrow cells pooled from three mice were seeded on 12 well tissue culture plates at 5 Â 10 6 cells per well in standard culture medium and cultured for 10 days. Half the medium was replaced every 5 days. Calvaria cells isolated from adult mice were seeded at 0.02 Â 10 6 cells per well and cultured for 3 days. FBS then was reduced to 2%, and 50 ng/ mL BMP-2 was added in the presence or absence of 10 À8 M E 2 in both types of cells. Two days later, 10 mM b-glycerolphosphate was added to the medium. The mineralization matrix was stained with 40 mM alizarin red solution 2 weeks later.
Quantitative RT-PCR Total RNA was extracted using Ultraspec (Biotecx Laboratories, Houston, TX, USA) and reverse-transcribed using the High-Capacity cDNA Archive Kit (Applied Biosystems) according to the manufacturer's instructions. Taqman quantitative reverse-transcriptase polymerase chain reaction (RT-PCR) was performed as described previously. (36) The primers and probes for murine Smad6 and rRNA18S were manufactured by the TaqMan Gene Expression Assays service (Applied Biosystems). Gene expression was quantified by subtracting the rRNA18S threshold cycle (C t ) value from the C t value of the gene of interest and expressed as 2ÀDC t , as described by the protocol of the manufacturer.
Other assays Intracellular ROS were quantified with dichlorodihydrofluorescein diacetate dye, (37) using bone marrow cells flushed from femurs and washed with PBS. Glutathione reductase activity was assayed with a kit from Cayman Chemical Company (Ann Arbor, MI, USA). Apoptosis in cultured cells was determined by measuring caspase-3 activity by cleavage of the fluorogenic substrate Ac-DEVD-AFC (Biomol, Plymouth Meeting, PA, USA), as described previously. (38) Statistical analysis ANOVA was used to detect effects of various in vivo and in vitro treatments after establishing that the data were normally distributed and equivalency of variances. Bonferroni's method was used to perform appropriate pairwise comparisons of treatment groups. In cases where one or both of the requirements for performing ANOVA were not met, Kruskal-Wallis ANOVA on ranks was used, followed by Dunn's method, to perform pairwise comparisons of treatment groups. Unless otherwise stated, results are presented as mean AE SD and in vitro assays performed in triplicate and repeated at least one time.

Results
ERa NERKI is capable of mediating the antiapoptotic effects of E 2 on osteoblasts To investigate whether direct binding of the ERa to DNA is required for the ability of this receptor to mediate the effects of estrogens on the lifespan of bone cells in vivo, we first investigated whether the ERa NERKI was capable of mediating the known antiapoptotic effects of estrogens on osteoblasts in vitro and in vivo. In agreement with previous findings from our group, (6,25) we found that E 2 prevented apoptosis of osteoblasts induced by H 2 O 2 or the topoisomerase inhibitor etoposide as effectively in osteoblastic cells isolated from calvaria of ERa NERKI/À mice, as it did in cells from WT mice (Fig. 1A). In line with these in vitro effects, the prevalence of osteoblast apoptosis was increased in vivo in both WT and ERa NERKI/À mice 6 weeks following OVX, as determined by in situ end labeling of vertebral sections (see Fig. 1B). Moreover, E 2 replacement prevented the OVX-induced increase in osteoblast apoptosis in both WT and ERa NERKI/À mice.

Estrogens suppress oxidative stress in vivo independently of the DNA-binding function of the ERa
We next compared the effects of OVX and E 2 replacement on ROS levels and GSR activity in the bone marrow and on p66 shc phosphorylation in vertebral lysates in WT and ERa NERKI/À mice 6 weeks after the hormonal manipulation (see Fig. 1C, D). We also compared these changes in mice haploinsufficient for ERa (ERa þ/À ) and mice in which one copy of the ERa has been replaced by the NERKI mutant (ERa NERKI/þ ). As we had seen in our earlier work, ROS levels were increased (see Fig. 1C) and GSR activity was decreased (see Fig. 1D) following OVX, and these changes were prevented by E 2 replacement in WT animals. Similar results were obtained in ERa þ/À , ERa NERKI/þ , and ERa NERKI/À mice, indicating that the antioxidant properties of estrogens are independent of the DNA-binding function of the ERa.
We also observed an increase in p66 shc phosphorylation 6 weeks following OVX and its reversal by E 2 replacement in WT, ERa þ/À , ERa NERKI/þ , and ERa NERKI/À mice, demonstrating that the negative regulation of p66 shc phosphorylation by estrogens in vivo is mediated via a mechanism that does not require ERa binding to DNA (see Fig. 1E). The increase in p66 shc phosphorylation in vertebral lysates after OVX was reproduced in both WT and ERa NERKI/À mice in a second experiment in which the mice were sacrificed 5 days after OVX (data not shown).
ERa NERKI/À mice exhibit a decrease in osteoblastogenesis To ascertain the requirement or lack thereof of direct binding of the ERa to DNA for the suppressive effect of estrogens on osteoblastogenesis, we proceeded with a comparison of osteoblastogenesis in WT and ERa NERKI/À mice by quantifying the number of progenitors able to form CFU-Fs and CFU-OBs in ex vivo bone marrow cultures seeded at three different densities ( Fig.  2A). The number of CFU-Fs was similar among all four genotypes. However, the number of CFU-OBs was significantly decreased in the bone marrow cultures from both ERa NERKI/þ and ERa NERKI/À mice compared with the WT mice, independent of the initial seeding density. In addition to the decrease in numbers, CFU-OB colonies from ERa NERKI/þ and ERa NERKI/À mice were smaller and displayed irregular shapes compared with cultures from WT or ERa þ/À mice, suggesting a defective osteoblast differentiation process analogous to the situation we had described previously in SAMP6 mice. (39) (see Fig. 2B). To determine whether the DNAbinding domain of the ERa is required for normal development of osteoblast progenitors from mesenchymal stem cells, we examined CFU-OBs derived from ERa À/À mice. The shape and number of colonies from ERa À/À mice were indistinguishable from those of WT mice, indicating that normal CFU-OBs are formed in the absence of the ERa (see Fig. 2C).
Having established a difference in CFU-OB numbers under basal conditions among the four genotypes, in a second experiment, we went on to investigate the effect of estrogen manipulation in the different genotypes (see Fig. 2D). As in the preceding experiment, ERa NERKI/À mice exhibited strikingly fewer CFU-OBs than WT and ERa þ/À mice. In this experiment, CFU-OBs from ERa NERKI/À mice were discernibly decreased compared with ERa NERKI/þ mice. Nonetheless, all four genotypes exhibited an increase in CFU-OB numbers following OVX, and the OVXinduced increase was prevented in all four genotypes by E 2 replacement. In agreement with these results, the number of osteoblasts in vertebral cancellous bone also was decreased in sham operated ERa NERKI/À mice compared with WT shamoperated controls. More important, the number of osteoblasts from ERa NERKI/À mice was increased following OVX, and this was prevented by E 2 , just as was observed in WT mice (see Fig. 2E).
ERa NERKI/À mice have low basal BMD and lose cancellous bone following OVX ERa NERKI/À mice exhibited low bone mass at baseline in both femur and spine (Fig. 3A). These mice also exhibited a decrease in vertebral length and volume, as well as femoral bone area ( Fig. 3B). Interestingly, when comparing mice carrying two copies of the WT ERa (ERa þ/þ ) or one copy (ERa þ/À ) or the ERa NERKI mutant together with the WT ERa (ERa NERKI/þ ) or alone (ERa NERKI/À ), there was a decrease at baseline BMD in the femur and spine of ERa þ/À , ERa NERKI/þ , and ERa NERKI/À mice compared with ERa þ/þ mice. In agreement with the dual-energy X-ray absorptiometry (DXA) results, unstained longitudinal sections of bone viewed in dark field confirmed the decreased vertebral size and diminished cancellous and cortical bone in the ERa NERKI/À mice (see Fig. 3C). 3D BMD, cortical thickness, and trabecular thickness, as determined by mCT, also were decreased in the ERa NERKI/À mice (but trabecular number and trabecular separation were indistinguishable) compared with WT controls under basal conditions (Table 1). Moreover, similar to the WT controls, the ERa NERKI/À mice lost cancellous (see Fig. 3D) but not cortical (see Fig. 3E) bone following OVX. Nonetheless, E 2 replacement at the dose used in our study did not prevent the loss of cancellous bone. A similar phenomenon was observed in the studies of Syed and colleagues, (40) but cancellous bone loss in the ERa NERKI/À mice was prevented by a higher dose of E 2 replacement in that earlier study.

E 2 attenuates BMP-2-induced osteoblast differentiation and the expression of BMP-2 target genes via ERa
Prompted by the evidence from the ERa NERKI/À mouse suggesting that estrogens attenuate osteoblastogenesis independently of the ability of ERa to interact directly with DNA, we went on to investigate whether estrogens attenuate osteoblastogenesis via a cell autonomous mechanism and whether such an effect is exerted via an extranuclear action of the ERa mediated through the activation of cytoplasmic kinases. To establish the generality of such a mechanism, we used several cell models: two different established murine cell lines, a human osteosarcoma cell line with conditional expression of the ERa (U2OS-ERa), and primary cultures of osteoblast progenitors obtained from calvaria or the bone marrow of C57BL/mice or ERa NERKI/À mice. As shown in Fig. 4A, BMP-2 dose dependently stimulated the differentiation of the preosteoblast cell line 2T3, as determined by AP activity. This effect was noticeable as early as day 1, reached a peak at day 3, and decreased by day 5. Addition of E 2 attenuated the effect of BMP-2. Similarly, BMP-2 stimulated osteoblast differentiation in the uncommitted mesenchymal progenitor cell line C2C12 with a maximal effect at 3 days of culture, and E 2 attenuated the effect of BMP-2 (see Fig. 4B). Consistent with its suppressive effect on BMP-2-induced osteoblast differentiation, E 2 also attenuated the stimulating effect of BMP-2 on the secretion of osteocalcin, an osteoblast-specific biosynthetic product, at days 3 and 5 of the cultures (see Fig. 4B). Moreover, E 2 antagonized BMP-2-induced gene transcription in C2C12 cells, as determined by the expression of the BMP-2 target gene Smad6 (see Fig. 4C).
To establish that the effect of E 2 on BMP-2-induced transcription was indeed mediated via the ERa, we examined the effects of E 2 on BMP-2-induced Smad6 mRNA in the human osteoblast-like osteosarcoma cell line U2OS stably expressing doxycycline-inducible ERa. BMP-2 stimulated Smad6 transcription in U2OS cells both in the absence or presence of ERa. However, whereas E 2 attenuated the effect of BMP-2 in U2OS cells expressing the ERa, the effect of E 2 was abolished in cells lacking the ERa (see Fig. 4D).
The findings with the established cell lines were readily reproduced in primary cultures of bone marrow-derived stromal cells or primary cultures of calvaria-derived cells, models that more closely represent normal osteoblastic cells in vivo. As expected, addition of BMP-2 to these primary cultures strongly promoted osteoblast differentiation and maturation, as determined by osteocalcin secretion (Fig. 5A) and mineralization (Fig.  5B). In full agreement with the evidence from the cell lines, E 2 decreased BMP-2-induced AP and osteocalcin secretion as well as mineralization in both the bone marrow-and the calvariaderived primary osteoblastic cell cultures (see Fig. 5). Moreover, practically identical results were obtained in cultures of calvaria cells from WT (ERa þ/þ ) and the ERa NERKI/À mice (see Fig. 5C, D), establishing that the DNA-binding function of ERa is indeed dispensable for the attenuating effect of estrogens on osteoblast differentiation.

E 2 stimulates Smad1 phosphorylation at the linker region and attenuates BMP-induced transcription via ERKs
Based on the finding that estrogens inhibit osteoblastogenesis in both WT and ERa NERKI/À mice and cells, we next tested the hypothesis that the attenuating effect of E 2 on BMP-2-induced transcription and osteoblast commitment/differentiation was Fig. 2. Osteoblastogenesis is decreased in ERa NERKI/À mice. (A) CFU-Fs or CFU-OBs in the bone marrow from femora of intact mice of the indicated genotypes. Cells from three mice were pooled and plated in duplicate at three different densities for each genotype. CFU-Fs were stained for alkaline phosphatase after 10 days, and CFU-OBs were stained with von Kossa to detect mineral after 25 days (left panel). The graphs on the right represent the quantification of CFUs depicted on the left. (B) Photomicrographs show representative CFU-OB colonies (50Â) obtained from WT (ERa þ/þ ) or ERa NERKI/À mice or (C) from WT or ERa À/À mice; þ/þ and WT refer to the respective littermate controls, as detailed in ''Materials and Methods.'' The graph on the right represents the quantification of CFU-OBs depicted on the right. (D) CFU-OBs obtained from femora of mice used in the experiment described in Fig. 1C. Cells from three mice were pooled and plated in triplicate at 10 6 cells per well. The graph on the bottom represents the quantification of CFU-OBs depicted on the top. (E) Osteoblast numbers on longitudinal undecalcified sections of L1-4 vertebrae from mice used in the experiment described in Fig. 1C (n ¼ 6 animals per group). Ã p < .05 versus OVX; y p < .05 versus þ/þ sham. mediated by the activation of cytoplasmic kinases, such as ERKs. As shown in Fig. 6A, E 2 attenuated BMP-2-induced phosphorylation of Smad1/5/8 in the C2C12 cell model, in agreement with earlier studies of ours. (13) In addition, E 2 attenuated the BMP-2induced transcriptional activation of the Smad6-luciferase construct (see Fig. 6B). Importantly, the specific MEK inhibitor PD98059 reversed the attenuating effect of E 2 on both BMP-2induced Smad1/5/8 phosphorylation and activation of transcription. However, PD98059 by itself had no effect on the BMP-2induced Smad1/5/8 phosphorylation or transcription, indicating that these effects of BMP-2 do not require ERK activation. Total Smad1 levels were not affected by any one of these treatments (data not shown).
Furthermore, E 2 as well as FGF2, used here as a positive control, stimulated the phosphorylation of ERKs as well as the phosphorylation of Smad1 at p-Serine 214 in its linker region (see Fig. 6C). Ser214 phosphorylation is the direct result of ERK activation (23,41) and triggers Smad1 proteasomal degradation leading to a decrease in Smad1 transcriptional activity. (21,32) In agreement with this evidence, PD98059 prevented the ability of E 2 to stimulate ERKs as well as Smad1 Ser214 phosphorylation. As expected, PD98059 also prevented FGF2-induced ERKs and Smad1 phosphorylation (see Fig. 6C). Finally, to verify that the MAPK sites in the linker region of Smad1 are important for the inhibitory actions of E 2 on BMP-2 signaling, we used MEFs from WT or Smad1 L / L mice that carry a Smad1 allele lacking all four MAPK sites in the linker region. (24) Addition of FGF2 to WT MEFs decreased BMP-2-dependent activation of a BMP transcriptional reporter construct, whereas addition of FGF2 to Smad1 L/L cells had no effect, as seen before by others. (21) Importantly, E 2 also failed to inhibit BMP-induced transcription in Smad1 L/L cells (see Fig. 6D).

Discussion
The evidence presented in this report illustrates that estrogens attenuate oxidative stress as well as the differentiation and apoptosis of osteoblasts by a nonclassical mechanism of ERa action. Specifically, our data reveal that the ability of estrogens to suppress oxidative stress and thereby attenuate the apoptosis of osteoblasts does not require binding of the ERa to DNA. The demonstration of estrogens' ability to suppress ROS levels and increase GSR activity in the bone marrow of ERa NERKI/À mice strongly suggests that the dispensability of ERa binding to DNA for the antioxidant properties of estrogens extends beyond osteoclasts and osteoblasts and therefore must be a common mechanism of this property of estrogens in all their other target tissues.
Using the ERa NERKI/À mouse model, we have obtained compelling evidence that binding of ERa to DNA is also dispensable for the attenuating effects of estrogens on osteoblastogenesis. Moreover, using a variety of osteoblastic models from mice and humans, including primary cultures of calvaria-and bone marrow-derived osteoblastic cells, we demonstrate herein that E 2 attenuates BMP-2-induced transcription and thereby osteoblastogenesis via ERK activation and downstream phosphorylation of Smad1 at the linker region of the protein, which leads to Smad1 proteasomal degradation. Consistent with this mode of action, E 2 attenuated BMP-2induced differentiation of primary cultures of calvaria-or bone marrow-derived osteoblastic cells from ERa NERKI/À mice as effectively as in cells from WT littermates, establishing that the DNA-binding function of the ERa is indeed dispensable for this effect. These mechanisms are summarized in the model depicted in Fig. 7.
The evidence that estrogens attenuate osteoblast apoptosis and oxidative stress by an ERa-mediated mechanism that does not depend on the DNA-binding function of this receptor is consistent with extensive work of others showing that estrogens are indeed able to influence cells in their numerous reproductive and nonreproductive target organs, in part, by extranuclear actions of the ER involving kinase-mediated signaling. (42)(43)(44) In addition, the findings of this study are in agreement with earlier studies of ours showing that the suppressive effect of E 2 on H 2 O 2induced phosphorylation of p66 shc -a cumulative index of oxidative stress-is kinase-mediated and is inhibited by the ERKspecific inhibitor PD98059. (25) ERb is intact in ERa NERKI/À mice, and therefore, we cannot categorically exclude the possibility that some of the effects of E 2 on this model are mediated by ERb. However, ERb expression in murine bone is two to three orders of magnitude lower than ERa, (25) and osteoblast number and bone mass were unaffected in mice lacking ERb. (45) In addition, studies in mice with a genuine null mutation of ERb indicate that with the exception of impaired ovarian function, this isotype of the ER is not required in the mouse for the development and homeostasis of the major body systems. (46) In addition to regulating kinases, steroid receptors exert significant effects on gene expression without direct DNA binding, such as through transreppression of NF-kB or AP-1, leaving the possibility that the effects of estrogens on oxidative stress, osteoblastogenesis, and apoptosis may have resulted from protein-protein interaction of the ERa with either one of these transcription factors. Such an alternative scenario, however, is very unlikely in the case of NF-kB because this particular transcription factor inhibits BMP-2-induced osteoblastogenesis. (47) Under basal conditions, ERa NERKI/À as well as ERa NERKI/þ mice exhibited decreased numbers of CFU-OBs compared with WT controls, raising the possibility that the DNA-binding domain of ERa may be required for normal osteoblastogenesis. This clearly is not the case because CFU-OBs from mice lacking both ERa alleles were indistinguishable from WT controls. Consistent with the decreased osteoblastogenesis, cancellous osteoblast number was decreased in the ERa NERKI/À mice compared to the ERa þ/þ mice. In addition, DXA measurements showed a decrease in bone mineral content of the femur and spine of both ERa NERKI/À and ERa NERKI/þ mice. The mechanistic basis of the effects of the ERa NERKI/À mutant on basal osteoblastogenesis is unclear, and additional work, beyond the scope of this report, will be required to elucidate it. A possible explanation is that the ERa NERKI protein interferes with osteoblastogenesis via a function that is unique to this mutant protein, for example, binding and sequestering a protein normally required for the process. We also found that femoral and spinal BMD were decreased in the ERa þ/À as compared with ERa þ/þ mice, in contrast to the observations of Smith and colleagues that osteopenia is not a consequence of the haploinsufficiency of the ERa in humans, (48) but we cannot account for this discrepancy.
The decreased BMD of the ERa NERKI/À mice was confirmed by mCT measurements showing decreased BV/TV and trabecular thickness in the vertebrae, as well as decreased cortical thickness at the mid-diaphysis of the tibia compared with ERa þ/þ mice. In agreement with our findings, Syed and colleagues (40) have reported previously that the ERa NERKI/À mice have decreased cancellous BMD in multiple sites, but in their studies, cortical BMD was not different from that of the ERa þ/þ mice. Also in agreement with this earlier work of Syed and colleagues, we found that ERa NERKI/À mice lose bone mass (BV/TV) with OVX, but at the dose used in our study, estrogen replacement does not reverse this effect. However, in difference from the report of Syed and colleagues, we found that OVX had no discernible effect on cortical thickness, whereas they reported that cortical thickness increased with OVX. The reason for the difference is most likely because Syed and colleagues made their measurement 9 mm from the proximal end of the tibia. Using a fixed distance would place the measurement more distally in the ERa NERKI/À mice, which have decreased tibia length compared with ERa þ/þ mice. Furthermore, a more distal measurement would inexorably occur in the tibia-fibular junction and include the cortex of two bones, thus confounding interpretation of the measurement. (49) A molecular explanation of the inhibitory effect of estrogens on osteoblastogenesis in vivo in both WT and ERa NERKI/À mice has been provided in this report by the in vitro demonstration that E 2 attenuates BMP-2-induced osteoblast differentiation and the transcription of BMP-2 target genes. Importantly, the attenuating effect of E 2 on BMP-2-induced osteoblast differentiation in vitro was indistinguishable in cells from ERa þ/þ and ERa NERKI/À mice, strongly supporting the view that the DNAbinding function of the ERa is indeed dispensable for this effect. Specifically, we have obtained evidence that similar to earlier findings regarding the anti-and proapoptotic effects of estrogens on osteoblasts and osteoclasts, respectively, the attenuation of BMP-2-induced Smad1/5/8 phosphorylation is mediated via activation of ERKs and, more precisely, that E 2 inhibits BMP-2 signaling by phosphorylating MAPK sites in the Smad1 linker region.
Similar to E 2 , other activators of MAPK, such as fibroblast growth factors (FGFs), have been shown to restrain BMP action during neural differentiation, limb development, and tooth formation. (50)(51)(52) Moreover, FGF or epidermal growth factor (EGF) inhibit BMP-induced gene expression and osteoblastogenesis in cell lines and primary human bone marrow-derived osteoblastic cells. (53)(54)(55) Phosphorylation of Smad1 by MAPK primes Smad1 for subsequent phosphorylation by GSK3b, which leads to Smad1 ubiquitination and degradation. Importantly, activation of Wnt signaling (i.e., inhibition of GSK3b) abrogates Smad1 degradation, thereby prolonging the duration of the BMP signal. (32) Thus Smad1 represents a site of convergence of both negative (e.g., MAPK and GSK3b) and positive (e.g., Wnt) regulatory signals of BMP-induced transcription. The convergence of multiple pathways on Smad1 underscores the importance of the modulation of the ERK-dependent phosphorylation of Smad1 by estrogens. We had shown previously that activation of kinase-mediated actions of the ERa with synthetic ligands that selectively activate kinases without stimulating transcriptional activation results in increased osteoblast differentiation, whereas E 2 did not. Interestingly, while both E 2 and the synthetic ligands phosphorylated ERKs, only the latter inactivated GSK3b and stimulated TCF-mediated transcription. (8) The evidence of this study that E 2 attenuates BMP-2-induced transcription is in agreement with our earlier observation that E 2 is unable to activate canonical Wnt signaling-an inhibitor of Smad1 degradation. On the other hand, the ability of the synthetic ligands to stimulate osteoblastogenesis, whereas E 2 could not, may be explained by the property of the former compounds to activate both ERK and Wnt signaling.
Different from the evidence reported herein, results of others from experiments with established cell lines (in some of which the level of ERa was artificially increased) or primary bone marrow stromal cell cultures have suggested that estrogens stimulate osteoblast differentiation, as evidenced by increased mineralization, alkaline phosphatase activity, and runx2 expression. (56)(57)(58) The discrepancy between the results of these earlier studies and ours may be due to the different experimental design. Indeed, we searched for the effects of estrogen on osteoblast differentiation in a setting in which primary bone marrow (or calvaria-derived cell) cultures and the BMP stimulus were used in combination. This combination was not used in those earlier in vitro studies. In support of this explanation, Usui and colleagues, in line with our findings, have observed an attenuating effect of estrogens on BMPinduced alkaline phosphatase activity in ROS17/2.8 osteoblastic cells. (11) More important, in agreement with the conclusions of this report, Usui and colleagues found that mice lacking Tob, an inhibitor of BMP, exhibit superenhancement of osteoblastic activity and augmentation of the bone- (C) Alkaline phosphatase activity in parallel cultures of calvaria-derived osteoblasts from WT controls (ERa þ/þ ) and ERa NERKI/À mice incubated with BMP-2 (25 ng/ mL), E 2 (10 À8 M), or E 2 and BMP-2 for 3 days. (D) Osteocalcin levels in the culture medium of same cells as in C treated as described for 10 days. Ã p < .05 versus BMP-2 alone. The same results were reproduced in a second experiment.
formation and mineral-apposition rates following loss of estrogens.
In summary, the composite evidence from the in vitro and in vivo experiments described herein demonstrates that estrogens exert cell autonomous effects on the differentiation and apoptosis of osteoblasts and provides a molecular explanation of the well-documented fact that following loss of estrogens in humans and mice, bone formation increases, albeit not in balance with the increased bone resorption, as does the prevalence of osteoblast and osteocyte apoptosis. In addition, the results of this report provide strong support of the view that the effects of estrogens on the birth and death of osteoblasts and osteocytes do not require the DNA-binding function of the ERa and result, in part, from antioxidant properties of these hormones.

Disclosures
All the authors state that they have no conflicts of interest. , or E 2 (10 À8 M) for 1 hour followed by vehicle or BMP-2 (50 ng/mL) for 1 hour. Bar graph represents the quantification of the intensity of the bands with an imaging system. (B) Luciferase activity in C2C12 cells transfected with a Smad6-Luc reporter construct and pretreated as in A, followed by treatment with vehicle or BMP-2 for 24 hours. (C) ERK1/2 and Smad1 (Ser214) phosphorylation by Western blot analysis in C2C12 cells pretreated with vehicle or PD98049 for 1 hour followed by E 2 or FGF2 (5 ng/ mL) for 15 minutes in serum-free medium. (D) Luciferase activity in MEFs transfected with a BMP response element reporter construct and pretreated for 1 hour with E 2 or FGF2, followed by treatment with vehicle or BMP-2 (25 ng/mL) for 24 hours. Ã p < .05 versus BMP-2 alone. Fig. 7. DNA-binding-independent actions of ERa on osteoblasts. Estrogens attenuate BMP-2-induced transcription by promoting the phosphorylation of Smad1 at its linker region, which, in turn, increases the proteasomal degradation of Smad1. This latter effect of the ERa results from the activation of ERKs. This mechanism contributes to the suppressive effect of estrogens on osteoblastogenesis and thereby osteoblast number and bone formation. A similar ERK-dependent decrease of ROS by estrogens is responsible for the antiapoptotic effect of these sex steroids on osteoblasts. The unleashing of this inhibitory effect, on loss of estrogens (e.g., menopause), is responsible for the increased osteoblast number and bone formation as well as the increase in osteoblast/osteocyte apoptosis that ensues following acute estrogen deficiency.