Reactive Oxygen Species Stimulates Receptor Activator of NF-κB Ligand Expression in Osteoblast*

It has been established that reactive oxygen species (ROS) such as H2O2 or superoxide anion is involved in bone loss-related diseases by stimulating osteoclast differentiation and bone resorption and that receptor activator of NF-κB ligand (RANKL) is a critical osteoclastogenic factor expressed on stromal/osteoblastic cells. However, the roles of ROS in RANKL expression and signaling mechanisms through which ROS regulates RANKL genes are not known. Here we report that increased intracellular ROS levels by H2O2 or xanthine/xanthine oxidase-generated superoxide anion stimulated RANKL mRNA and protein expression in human osteoblast-like MG63 cell line and primary mouse bone marrow stromal cells and calvarial osteoblasts. Further analysis revealed that ROS promoted phosphorylation of cAMP response element-binding protein (CREB)/ATF2 and its binding to CRE-domain in the murine RANKL promoter region. Moreover, the results of protein kinase A (PKA) inhibitor KT5720 and CREB1 RNA interference transfection clearly showed that PKA-CREB signaling pathway was necessary for ROS stimulation of RANKL in mouse osteoblasts. In human MG63 cells, however, we found that ROS promoted heat shock factor 2 (HSF2) binding to heat shock element in human RANKL promoter region and that HSF2, but not PKA, was required for ROS up-regulation of RANKL as revealed by KT5720 and HSF2 RNA interference transfection. We also found that ROS stimulated phosphorylation of extracellular signal-regulated kinases (ERKs) and that PD98059, the inhibitor for ERKs suppressed ROS-induced RANKL expression either in mouse osteoblasts or in MG63 cells. These results demonstrate that ROS stimulates RANKL expression via ERKs and PKA-CREB pathway in mouse osteoblasts and via ERKs and HSF2 in human MG63 cells.

Bone remodeling depends on a delicate balance between bone formation and bone resorption, wherein bone-forming osteoblast and bone-resorbing osteoclasts play central roles (1,2). Indeed, tipping this balance in the favor of osteoclasts leads to pathological bone resorption, as seen in bone diseases such as osteoporosis and rheumatoid arthritis. The differentiation or lifespan of osteoblast and osteoclast is believed to be particularly important in pathogenesis of these bone diseases (3)(4)(5). Osteoclast formation from hematopoietic prcecursors requires factors that promote their differentiation and survival. Such factors are produced by stromal/osteoblastic cells that originate from mesenchymal progenitors residing in the bone marrow. Two such factors appear to be essential. One is macrophage colony stimulating factor (M-CSF), 1 which is necessary but not sufficient for osteoclast formation. The other is a membrane cytokine, receptor activator of NF-B ligand (RANKL). RANKL is essential and, together with M-CSF, is sufficient for osteoclast differentiation. RANKL acts by binding to its receptor, RANK, on the surface of hematopoietic precursors, stimulating their differentiation into mature osteoclasts. The action of RANKL is prevented by osteoprotegerin (OPG), a soluble decoy receptor that competes with RANK for binding to RANKL and is also expressed by osteoblasts (2,6).
Reactive oxygen species (ROS) such as superoxides anions, hydroxyl radicals, and H 2 O 2 can cause severe damage to DNA, protein, and lipids. High levels of oxidant produced during normal cellular metabolism (e.g. mitochondrial electron transport) or from environmental stimuli (e.g. cytokines, UV radiation) perturb the normal redox balance and shift cells into a state of oxidative stress (11). Oxidative stress is believed to contribute to the etiology of various degenerative diseases such as diabetes, atherosclerosis, arthritis, cancer, and the process of aging (11). At the cellular level, ROS may act as second messengers in various signal transduction and elicits a wide spectrum of responses ranging from proliferation to growth or differentiation arrest to senescence and cell death by activating numerous major signaling pathways including phosphoinositide 3-kinase (PI-3K), NF-B, phospholipase C-␥1 (PLC-␥1), p53, CREB, HSF, and mitogen-activated protein kinases (MAPKs), which may classify into: extracellular signal-regulated kinases (ERKs), c-Jun N-terminal kinase p38 MAPK. The magnitude and duration of the stress as well as the cell type involved are important factors in determining which pathways are activated, and the particular outcome reflects the balance between these pathways (12).
Recent evidence has shown that ROS may be involved in the pathogenesis of bone loss-related diseases. Marked decrease in plasma antioxidants were found in aged osteoporotic women (13). Osteoporosis has been noted in two mouse models of premature aging associated with oxidative damage in which osteopenia is presumed to be the consequence of oxidative damage (14,15). Estrogen deficiency induced by ovariectomy causes bone loss by lowering thiol antioxidants in osteoclast and increasing osteoclast differentiation (16). There is also a biochemical link between increased oxidative stress and reduced bone mass in aged men and women (17). Oxidative stress is able to inhibit the osteoblastic differentiation of bone cells by ERK and NF-B (18,19).
Previous studies have shown that ROS stimulates osteoclast differentiation and bone resorption in mouse bone in vitro and in vivo (16,20,21), and ROS produced by osteoclast or its precursor is an important local factor involved in the activation and differentiation of osteoclast (22,23). But mechanisms involved in ROS stimulating osteoclast differentiation and bone resorption remain unclear. The role of ROS in expression of factors essential for osteoclast differentiation such as RANKL and M-CSF and the signaling mechanisms through which ROS regulates these genes have also not been studied. In this paper, we report that H 2 O 2 or xanthine/ xanthine oxidase (XXO)-generated superoxide anion stimulates RANKL expression via ERKs and HSF2 in human osteoblast-like cell line MG63 and via ERKs and CREB signaling in primary mouse bone marrow stromal cells (BM-SCs) and osteoblast, respectively.
RNAi Assay-RNAi assay was performed by using pSilencer TM puro RNAi expression vector kit purchased from Ambion Tech (catalog number 5762) according to the manufacturer's instruction. Briefly, mouse CREB1 RNAi and human HSF2 RNAi was constructed by cloning the target sequence 5Ј-aagagaatgtcgtagaaagaa-3Ј and 5Ј-gcagcttatccagtatacc-3Ј, respectively, into the pSilencer TM puro plasmid. This sequence was determined to be unique to the mouse CREB1 and human HSF2 gene by BLAST search of the GenBank TM data base. MG63 cells and mouse osetoblast transfection were performed using SiRPORT TM XP-1 DNA transfection reagent (Ambion catalog number 4504) following the manufacturer's recommendations. After 24 h, the transfected cells were selected with the puromycin to enrich the culture for cells that were successfully transfected for the following 2 days. Then the population was analyzed for the expression of CREB and HSF2 by reverse transcription (RT)-PCR and Western blotting.
Flow Cytometric Determination of ROS-2,7-DCF-DA is a cell-permeable dye that becomes fluorescent upon reaction with ROS such as hydroxyl radical, hydrogen peroxide, or peroxynitrite. 2,7-DCF-DA was dissolved in Me 2 SO and stored as 50 mM stock. Cells were loaded with 10 M 2,7-DCF-DA for 30 min and then treated with different concentrations of H 2 O 2 or xanthine/xanthine oxidase in the presence or absence of their inhibitors, 500 units/ml catalase and 20 M oxypurinol. The cells were harvested at the indicated time points after incubation, washed three times with phosphate-buffered saline, and then immediately analyzed by flow cytometry using FACSort (Becton-Dickinson, Rutherford, NJ) with a 488-nm excitation beam. The signals were obtained using a 530-nm bandpass filter (FL-1 channel) for DCF. Each determination is based on the mean fluorescence intensity of 5,000 cells.
Osteoclastogenesis in Vitro and Resorption Assay-Osteoclast differentiation in vitro was performed by coculture of mouse calvarial osteoblasts and spleen cells. Briefly, mouse calvarial osteoblasts and spleen cells were plated together at a density of 5 ϫ 10 4 and 1 ϫ 10 6 cells/well, respectively, in 24-well culture plates and cultured for 7 days in ␣-minimum essential medium supplemented with 10% fetal bovine serum. During the culture periods, the cells were exposed to 100 M H 2 O 2 or 20 M xanthine and 20 milliunits/ml XXO in the presence or absence of their inhibitor, 500 units/ml catalase or 20 M oxypurinol, or exposed to 100 nM 1,25-(OH) 2 D 3 . At the end of culture, the cells were fixed with ethanol-acetone (50:50, v/v) and were incubated for 20 min in acetate buffer (0.1 M sodium acetate, pH 5.0) containing naphtol AS-MX phosphate, fast red violet LB salt, and 20 mM sodium tartrate. The numbers of tartrate-resistant acid phosphatasepositive multinucleated osteoclast-like cells that contain three or more nuclei were counted under a light microscope.
For resorption assay, mouse calvarial osteoblasts and spleen cells were plated on bovine bone slices and treated with different reagents as described above. After 14 days of culture, the cells were completely removed from the bone slices by abrasion with a cotton tip, and the slices were stained with 1% toluidine blue, allowing the visualization of resorption lacunae or pits. The photographs were taken under a light microscope at 40ϫ magnification, and areas of resorbed pits were analyzed by the Image Pro-Plus program version 4.0 (Media Cybernetics, Silver Spring, MD). Western Blotting-The cells were washed with cold phosphate-buffered saline and lysed in Laemmli buffer (62.5 mM Tris-HCl, pH 6.8, 2% SDS, 10% glycerol, 50 mM dithiothreitol, 0.01% bromphenol blue) for 5 min at 95°C. Cell lysates were analyzed by SDS/PAGE and transferred electrophoretically to polyvinylidene difluoride membrane (Bio-Rad). The blots were probed with antibodies specific for phosphorylated-p38 MAPK, phosphorylated-PLC-␥1, and phosphorylated-CREB purchased form Cell Signaling Technology and antibodies for RANKL, HSF2, ␤-actin, CREB, NF-B inhibitor protein ␣ (IB␣), phosphorylated-ERK1/2, phosphorylated IB␣, and cathepsin K from Santa Cruz (Santa Cruz, CA), and the immunoreactive proteins were revealed by an ECL kit.
RT-PCR Analysis-Total RNA was extracted from cells using an RNA extraction kit (Promega, Madison, WI). Synthesis of cDNA from mRNA transcripts was performed using the following method: total RNA (10 g), dNTP (250 M), oligo(dT) (5.0 g), avian myeloblastosis virus reverse transcriptase (40 units; Takara) in a total volume of 200 l and incubated at 42°C for 1 h. RT-PCR was performed using 5 l of the cDNA solution and 35 cycles. The semi-quantitative method of RT-PCR was validated in preliminary experiments. PCR cycle number was optimized for each experimental condition and primer set. Representative samples were run at different cycle numbers (22-38 cycles), and the optimal cycle number was selected in the region of linearity between cycle number and PCR product intensity. To confirm a linear relationship between template and PCR product intensity at the optimal cycle number, PCR was run at different cDNA concentrations from representative samples. All of the reactions included a negative control in which cDNA was omitted from the PCR. To prove sample equality, amplification of the housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (G3PDH) was used for internal control. 8 l of each PCR product was run on 1.5% agarose gels containing 400 ng/ml ethidium bromide and visualized on a UV transilluminator. The range of amplified DNA product that was linear in relation to the input RNA was used for data presentation.
Electrophoretic Mobility Shift Assay (EMSA)-Nuclear extracts of stimulated cells were prepared with the NE-PER nuclear extraction reagent (Pierce). The protein concentration in nuclear extracts was determined by BCA assay. Biotin end-labeled double-stranded oligonucleotide probes for CRE-like domain in mouse RANKL promoter sequence (Ϫ945 to Ϫ926, sense, 5Ј-biotin-TGA GTT TGA GGT CAG CCT GG-3Ј) and HSE in human RANKL gene promoter sequence (Ϫ1717 to Ϫ1750, sense, 5Ј-biotin-ATA AGA AAG AAG AAA TAT GGA ATT ATT TCC TGA-3Ј) as well as their corresponding nonlabeled oligonucleotide probe were synthesized by Sangon Biotech (Sangon, China). The binding reactions contained 5 g of nuclear extract protein, buffer (10 mM Tris, pH 7.5, 50 mM KCl, 5 mM MgCl 2 , 1 mM dithiothreitol, 0.05% Nonidet P-40, and 2.5% glycerol), 1 g of poly(dI-dC), and 2 nM of biotin-labeled DNA. The reactions were incubated at 23°C for 20 min. The competition reactions were performed by adding 200-fold excess unlabeled double-stranded consensus oligonucleotide to the reaction mixture. The reactions were electrophoresed on a 6% Tris borate-EDTA gel at 100 V for 1 h in a 100 mM Tris borate-EDTA buffer. The reactions were transferred to a nylon membrane, and the biotin-labeled DNA was detected with a LightShift chemiluminescent EMSA kit (Pierce).
Statistical Analysis-Computer-assisted statistical analyses were performed using the Student's two-tailed t test. A p value of Ͻ0.05 was considered to be statistically significant.

ROS Production in Response to H 2 O 2 or XXO-induced Oxidative Stress in Mouse Osteoblasts and Human MG63
Cells-To determine the effect of ROS on osteoblasts, we treated human osteoblast-like cell line MG63 and primary mouse osteoblasts with H 2 O 2 or XXO, a reaction in which xanthine oxidase converts xanthine to uric acid and generates superoxide anion. Intracellular ROS production by 10 -100 milliunits/ml XXO and by 50 -500 M H 2 O 2 was measured by DCF fluorescence over the course of 2 h. In response to H 2 O 2 , DCF fluorescence increased in both cell types compared with control in a concentration-dependent (Fig. 1A To determine the effect of ROS on factors essential for oste-oclast differentiation that are expressed by stromal/osteoblast cells, we examined the expression of M-CSF, RANKL, and OPG mRNA in mouse osteoblast after treatment with H 2 O 2 and XXO as shown in Fig. 2. H 2 O 2 or XXO increased intracellular ROS ( Fig. 2A) and induced RANKL mRNA expression in a dose-dependent manner (Fig. 2B). Time course analysis revealed that RANKL mRNA expression reached the peak at 4 h (Fig. 2C). Catalase or oxypurinol, specific inhibitors for H 2 O 2 and XXO, respectively, inhibited H 2 O 2 -or XXO-induced intracellular ROS production ( Fig. 2A) and RANKL mRNA expression completely (Fig. 2B). In contrast, the expression levels of M-CSF, OPG, and G3PDH mRNA were not changed by H 2 O 2 or XXO treatment (Fig. 2B). To confirm this result, we performed Western blot analysis using an anti-RANKL antibody. Consistent with the previous results, H 2 O 2 or XXO induced RANKL protein expression in a dose-dependent manner in mouse osteoblast. Catalase or oxypurinol inhibited H 2 O 2 -or XXO-stimulated RANKL protein expression (Fig. 2B). Under this condition, ␤-actin protein expression was not changed by H 2 O 2 or XXO treatment.  Fig. 2A, and the ROS levels were then determined by FACSort. *, p Ͻ 0.01 versus column H 2 O 2 (t test). **, p Ͻ 0.01 versus column XXO (t test). B, mouse BMSCs were treated as described for Fig. 2B, and expression of RANKL, OPG, M-CSF, or G3PDH mRNA were analyzed by RT-PCR. The cell lysates were subjected to Western blotting analysis with an anti-RANKL or an anti-␤-actin antibody. C, mouse osteoblasts and spleen cells were plated together at a density of 5 ϫ 10 4 and 1 ϫ 10 6 cells/well, respectively, in 24-well culture plates and cultured for 7 days. During the culture periods, the cells were exposed to 100 M H 2 O 2 or 20 M xanthine and 20 milliunits/ml XXO in the presence or absence of their inhibitor, catalase and oxypurinol, or exposed to 100 nM 1,25-(OH) 2 D 3 . At the end of culture, the cells were subjected to tartrate-resistant acid phosphatase staining as described under "Experimental Procedures." D, mouse calvarial osteoblasts and spleen cells were plated on bovine bone slices and treated with different reagents as described above. After 14 days of culture, the cells were removed from the slices, and resorbed pits were visualized by staining with toluidine blue. The percentages of resorbed areas were determined with an image analysis program. E, osteoblasts and spleen cells plated in 6-well culture plates at a density of 3 ϫ 10 5 and 1 ϫ 10 7 cells/well, respectively, were treated with different reagents as above mentioned for 10 days, and then total RNA and protein were extracted and subjected to RT-PCR or Western blot analysis of cathepsin K and matrix metalloproteinase 9 expression. The values are the means Ϯ S.D. for six determinations/group. *, p Ͻ 0.01 versus control (t test). CAT, catalase; OXY, oxypurinol; Con, control. RANKL mRNA expression in BMSCs. Consistent with the mouse osteoblasts, intracellular ROS stimulated by H 2 O 2 or XXO (Fig. 3A) enhanced expression of RANKL mRNA in a dose-dependent manner (Fig. 3B). Catalase or oxypurinol inhibited H 2 O 2 or XXO-induced intracellular ROS production (Fig. 3A) and RANKL mRNA expression (Fig. 3B). In contrast, the expression levels of M-CSF and OPG were not changed by H 2 O 2 or XXO treatment (data not shown). We also performed Western blotting analysis with an anti-RANKL antibody and found that ROS-induced RANKL protein expression in BMSCs. Catalase or oxypurinol inhibited H 2 O 2 or XXO up-regulation of RANKL protein expression (Fig. 3B). These results indicate that ROS induces RANKL expression in mouse BMSCs.

ROS Stimulates RANKL Expression in Mouse
We further examined the effect of ROS-induced RANKL in osteoclast differentiation by coculture of mouse osteoblasts and spleens cells. As shown in Fig. 3 (C and D), the number of tartrate-resistant acid phosphatase-positive multinucleated osteoclast-like cells and areas of bone resorption increased significantly (p Ͻ 0.05) with the treatment of 100 M H 2 O 2 or 20 milliunits/ml XXO, similar to the effect of osteoclasts stimulating hormone 1,25-(OH) 2 D 3 . Catalase or oxypurinol reduced H 2 O 2 or XXO up-regulation of osteoclastogenesis and bone resorption (Fig. 3, C and D). Consistent with this, expression of cathepsin K and matrix metalloproteinase 9, other osteoclast differentiation markers, were also increased by H 2 O 2 or XXO and reversed by their inhibitors (Fig. 3E). These results suggest that ROS stimulates mouse osteoclasts differentiation by inducing RANKL expression.
Signaling Pathways Affected by ROS in Mouse Osteoblasts-PLC-␥1 plays an important role in regulation cell proliferation and differentiation. Recently, an anti-apoptotic role of PLC-␥1 phosphorylation and activation in oxidative stress was reported (25). NF-B transcription factor is known to be essential for osteoclastogenesis and oxidative stress is an activator of NF-B (11). Activation of NF-B occurs via phosphorylation of the inhibitory IB proteins, followed by protease-mediated degradation of IB, resulting in the release and nuclear translocation of active NF-B (12). ERK1/2 and p38 MAPK are involved in both oxidative stress and bone cell (osteoblast and osteoclast) differentiation (11,26). Phosphorylation and activation of CREB by its upstream sequence, PKA has been shown to be essential for hormone/ cytokine-induced RANKL expression in mouse osteoblasts (8,9). To determine whether these signaling pathways are affected by ROS in mouse osteoblast, the cells were treated with 100 M H 2 O 2 or 20 M xanthine and 20 milliunits/ml XXO for 10, 30, or 60 min, and the cell lysates were subjected to Western analysis with anti-phosphorylated PLC-␥1, IB␣, p38 MAPK, CREB, or ERK1/2 antibody and anti-IB␣, CREB antibody as described under "Experimental Procedures." We found that phosphorylation of PLC-␥1 and ERK1/2 increased markedly by H 2 O 2 and XXO from 10 to 60 min; IB␣ activation began to rise after treatment for 30 min, accompanied by significant degradation of IB␣; the treatment of mouse osteoblasts with H 2 O 2 and XXO induced phosphorylation of CREB/ATF2 from 10 -60 min without affecting CREB protein amount, but p38 MAPK was notably inhibited by H 2 O 2 and XXO treatment (Fig. 4). Our results demonstrate that PLC-␥1, ERK1/2, CREB, and NF-B signaling pathways are stimulated, whereas p38 MAPK is inhibited by intracellular ROS increased by H 2 O 2 or XXO in mouse osteoblasts.

Involvement of ERKs and PKA-CREB Signaling Pathways in ROS-induced RANKL Expression in Mouse Osteoblasts-To
identify signaling pathways involved in ROS regulation of RANKL, specific inhibitors for PLC-␥1, PI-3K, p38 MAPK, ERKs, PKA, and NF-B were added in mouse osteoblasts fol-lowed by exposure to 100 M H 2 O 2 or 20 M xanthine and 20 milliunits/ml XXO. Then RANKL mRNA and protein expression were detected by RT-PCR and Western blot analysis, respectively. It is found that 20 M PD98059 or 10 M KT5720, the specific inhibitor for ERK and PKA respectively, suppressed H 2 O 2 -induced RANKL mRNA and protein expression (Fig. 5A). 5 M U73122, 100 nM wortmannin, 50 M caffeic acid phenethyl ester, and 10 M SB203580, the specific inhibitor for PLC-␥1, PI-3K, NF-B, and p38 MAPK, respectively, however, had no notable effect on increase of RANKL expression elicited by H 2 O 2 (Fig. 5A). Experiments employing XXO get similar results (data not shown). We also found that these inhibitors had no significant effect of RANKL expression in the absence of H 2 O 2 and XXO (data not shown). Furthermore, when mouse osteoblasts were incubated with PKA activator forskolin, it induced RANKL mRNA expression in a time-dependent manner (Fig. 5B). These results suggest that activation of ERK and PKA is required for ROS up-regulation of RANKL expression in mouse osteoblasts.
Because ROS induced the phosphorylation of transcription factor CREB (Fig. 4), to confirm the involvement of PKA-CREB signaling pathway in ROS stimulated RANKL expression, we transfected pSilencer TM puro RNAi expression vector containing CREB1 RNAi target sequence into mouse osteoblasts. As shown in Fig. 5C, CREB mRNA and protein levels were significantly knocked down by overexpression of CREB1 small interfering RNA. Moreover, ROS-induced CREB phosphorylation and RANKL mRNA and protein expression were suppressed by transfecting CREB RNAi target sequence into mouse osteoblasts (Fig. 5C). These results strongly indicate that the PKA-CREB signaling pathway plays an important role in ROS-induced RANKL expression.
It has been reported that murine RANKL promoter contains a CRE-like domain (Ϫ939 to Ϫ932, TGAGGTCA) and that activation and binding of CREB to the CRE-like domains in mouse RANKL promoter sequence is essential for hormone/ cytokine-induced RANKL expression in mouse osteoblasts (8,9). To confirm interaction of CREB and CRE-like domain in RANKL promoter sequence in ROS-stimulated mouse osteo- blasts, we employed EMSA using double-stranded oligonucleotide probe containing 20 bp of mouse RANKL promoter sequence (Ϫ945 to Ϫ926), and nuclear extracts prepared from osteoblasts treated with or without PKA activator forskolin or H 2 O 2 and XXO in the presence or absence of their inhibitor catalase or oxypurinol (Fig. 6A). We observed the significantly increased levels of CREB binding to CRE-like domain when oligonucleotide probe was incubated with nuclear extracts from forskolin-, H 2 O 2 -, or XXO-treated mouse osteoblasts. The DNA-protein complex decreased notably when cells were treated with catalase or oxypurinol before H 2 O 2 or XXO treatment (Fig. 6A). Moreover, the formation of the DNA-protein complex was efficiently competed with 200-fold molar excess of unlabeled CRE consensus oligonucleotide. Thus, complex formation is mediated by the interaction with CREB/ATF transcription factors. We further determined whether ROS-induced phosphorylation and DNA binding activity of CREB is mediated by ERKs and PKA. We found that pretreatment with PKA inhibitor KT5720, but not ERK-specific inhibitor PD98059, suppressed H 2 O 2 -induced CREB phosphorylation and formation of CREB-CRE-like domain complex (Fig. 6B). Therefore, CREB/ ATF transcription factors might be involved in RANKL expression via PKA but not ERKs.
ROS Up-regulates RANKL Expression in Human Osteoblastlike MG63 Cell Line-To determine whether human osteoblasts respond differently than mouse osteoblasts in vitro, we further studied the effect of ROS on M-CSF, OPG, and RANKL mRNA expression in MG63 osteosarcoma, a human cell line used widely for studies of hormone/cytokine/stress effects, closely related to secretory, matrix producing osteoblasts (27). Consistent with the mouse osteoblasts, intracellular ROS stimulated by H 2 O 2 or XXO (Fig. 7A)  . Then RT-PCR was carried out using murine RANKL or G3PDH primer sets, as described under "Experimental Procedures." The cell lysates were subjected to Western blot analysis with an anti-RANKL or an anti-␤-actin antibody. B, cells were serum-starved for 24 h. Then the cells were incubated in serum-free medium containing 10 M forskolin for the indicated time periods. RANKL and G3PDH mRNA was analyzed by RT-PCR as described above. C, mouse osteoblasts were transfected with Silencer TM puro RNAi expression vector containing negative control or CREB1 RNAi target sequence as described under "Experimental Procedures." After 24 h, the transfected cells were selected with the 0.8 g/ml puromycin for the following 2 days. Then the cells were incubated with or without 100 M H 2 O 2 for 4 h (RT-PCR analysis) or 8 h (Western blot analysis). The cells were harvested and lysed. RT-PCR was carried out using murine RANKL, CREB, or G3PDH primer sets. The cell lysates were subjected to Western blot analysis with anti-RANKL, anti-phosphorylated CREB, anti-CREB, or anti-␤-actin antibody.
FIG. 6. ROS induced binding of CREB to the CRE-like domains in mouse RANKL promoter sequence via PKA. A, nuclear extracts of mouse osteoblast that had been exposed to 10 M forskolin, 100 M H 2 O 2 , or 20 M xanthine and 20 milliunits/ml XXO in presence or absence of their inhibitors 500 units/ml catalase or 20 M oxypurinol for 30 min were incubated with biotin-labeled oligonucleotides probe for CRE-like domain in mouse RANKL gene promoter. DNA-protein complexes were fractionated by polyacrylamide gel electrophoresis and visualized by horseradish peroxidase-conjugated streptavidin as described under "Experimental Procedures." B, mouse osteoblasts were pretreated with 20 M PD98059 or 10 M KT5720 for 30 min followed by exposure to 100 M H 2 O 2 for 30 min, and then nuclear extracts were subjected to EMSA analysis as described above.
RANKL mRNA in a dose-dependent manner (Fig. 7B). Inhibitor for H 2 O 2 or XXO, catalase or oxypurinol inhibited H 2 O 2 or XXO-induced intracellular ROS production (Fig. 7A) and RANKL mRNA expression (Fig. 7B). In contrast, the expression levels of M-CSF, OPG, and G3PDH mRNA were not changed by H 2 O 2 or XXO treatment (Fig. 7B). Western blotting analysis also demonstrated that ROS induced RANKL protein expression in MG63. Catalase or oxypurinol inhibited H 2 O 2 or XXO up-regulation of RANKL protein expression (Fig. 7C). These results indicate that ROS induces RANKL expression in human osteoblasts.
HSF2 Is Required for ROS-induced RANKL Expression in Human MG63 Cells-We screened the drugs that could interfere with ROS-induced RANKL expression in human MG63 cells. In contrast to mouse osteoblasts, we found that only pretreatment of ERKs inhibitor PD98059, but not specific inhibitors for PLC-␥1, PI-3K, p38 MAPK, PKA, and NF-B, decreased H 2 O 2 -or XXO-induced RANKL mRNA and protein expression (data not shown). So unlike mouse osteoblasts, PKA-CREB may not be involved in ROS-induced RANKL expression in human MG63 cells.
Roccisana et al. (7) recently reported that binding of HSF2 to heat shock factor-responsive elements (HSEs) in the human RANKL gene promoter region (Ϫ1720 to Ϫ1731, AGAAAGAAGAAA) play an important role in modulating RANKL gene expression in human stromal/osteoblast cells (7). This observation, combined with the evidence that HSF could be activated by oxidative stress, prompted us to explore the role of HSF signaling in ROS-induced RANKL expression in human osteoblast. RT-PCR analysis for HSF1 and HSF2 mRNA expression in human MG63 cells detected only HSF-2 expression (Fig. 8A), consistent with the results of Roccisana et al. (7) with human stromal/osteoblasts. We further trans-fected pSilencer TM puro RNAi expression vector containing HSF2 RNAi target sequence into MG63 cells. As shown in Fig. 8B, HSF2 mRNA and protein levels were significantly knocked down by overexpression of HSF2 small interfering RNA. Moreover, H 2 O 2 -induced RANKL mRNA and protein expression were suppressed by transfecting HSF2 RNAi target sequence into MG63 cells (Fig. 8B). Furthermore, when MG63 cells incubated with MG132, a proteasome inhibitor and HSF2 activator, it induced RANKL mRNA expression (Fig. 8C). These results strongly suggest that activation of HSF2 is required for ROS up-regulation of RANKL expression in human osteoblasts.
To detect interaction of HSF2 and HSE in ROS-stimulated MG63 cells, we employed EMSA using probe containing HSE (Ϫ1717 to Ϫ1750) in human RANKL promoter region. We observed the notably increased levels of HSF2 binding to HSE when oligonucleotide probe was incubated with nuclear extracts from MG132-, H 2 O 2 -, or XXO-treated MG63 cells. The DNA-protein complex decreased when cells were treated with catalase or oxypurinol before H 2 O 2 or XXO treatment (Fig. 8D). We also found that the DNA-protein complex diminished in H 2 O 2 -treated MG63 cells transfected with pSilencer TM puro RNAi expression vector containing HSF2 RNAi target sequence (Fig. 8E). Moreover, the formation of the DNA-protein complex was efficiently competed with 200-fold molar excess of unlabeled HSE consensus oligonucleotide. We further determined whether ROS-induced DNA binding activity of HSF2 is mediated by ERKs. It is found that pretreatment with ERKspecific inhibitor PD98059 could not suppress H 2 O 2 -and XXOinduced HSF2-HSE complex (Fig. 8F). Therefore, ERKs is not upstream of HSF2 transcription factors in ROS-induced RANKL expression.

DISCUSSION
ROS may play its role in bone loss-related diseases by two ways: suppression of born formation and stimulation of bone resorption. Recent data from Mody et al. (18) and our laboratory showed that ROS such as H 2 O 2 -or XXO-generated superoxides anion is able to inhibit osteoblastic differentiation of mouse (18) and rabbit (19) BMSCs and calvarial osteoblasts by ERKs and NF-B. Although many reports have shown that ROS including H 2 O 2 and superoxides anion stimulate osteoclast differentiation and bone resorption (16, 20 -23, 28 -31), little is known about mechanisms involved in this process, and most investigations focus on osteoclast precursors or osteoclast itself. It is very important to identify the effect of ROS on stromal/osteoblast cells and their expression of RANKL, M-CSF, and OPG, factors essential for osteoclasts differentiation, survival, and activation. Our results in mouse BMSCs, osteoblasts, and human MG63 cells indicate for the first time that H 2 O 2 -or XXO-increased intracellular ROS induces RANKL expression in osteoblasts (Figs. 2, 3, and 7) and enhances osteoclasts formation in mouse osteoblast-spleen cell coculture (Fig. 3.). This provides a rational explanation for several recent reports. Thiol antioxidants fall substantially in rodent bone marrow after ovariectomy, and antioxidants prevent ovariectomy-induced estrogen deficiency bone loss, whereas reagent, which depletes thiol antioxidants, induces bone loss (16). Production of RANKL stimulated by ovariectomy results in increased osteoclasts formation and bone resorption in osteoporosis (32) or autoimmune arthritis (33) mice. Expression of RANKL and OPG correlates with age-related bone loss in mice (34). According to our results, increased ROS levels by aging or estrogen deficiency could contribute to enhanced osteoclast formation and bone resorption by stimulation of RANKL expression in stromal/osteoblast cells. But other identified signals, such as tumor necrosis factor ␣ (2, 35), interleukin-6, interleukin-11, OPG (36), interleukin-7 (37), and vascular epithelial growth factor (38), of course, may also be involved in this FIG. 8. Involvement HSF2 signaling in ROS-induced RANKL expression in human osteoblast-like MG63 cells. A, total RNA of MG63 cells and HeLa cells (HSF1-positive control) were extracted, and RT-PCR was carried out using human HSF1 or HSF2 primer sets. B, MG63 cells were transfected with Silencer TM puro RNAi expression vector containing negative control or HSF2 RNAi target sequence. After 24 h, the transfected cells were selected with 6 g/ml puromycin for the following 2 days. Then the cells were incubated with or without 100 M H 2 O 2 for process. It is unclear why ROS did not affect expression of M-CSF or OPG mRNA in our experiments.
It has been well established that PLC-␥1, PI-3K, NF-B, and p38 MAPK are oxidative stress-sensitive signals (12). They also play important signaling roles in differentiation and survival of osteoclasts (39 -45). Consistent with our previous studies in rabbit osteoblasts (19), H 2 O 2 (100 M) or XXO (20 milliunits/ ml) induced PLC-␥1 and IB␣ phosphorylation but suppressed p38 MAPK phosphorylation in mouse osteoblasts (Fig. 4). Evidence that specific inhibitors for PLC-␥1, PI-3K, NF-B, or p38 MAPK could not change ROS-increased (Fig. 5A) or native (data now shown) RANKL mRNA and protein levels indicates that these signaling pathways are not involved in ROS upregulation of RANKL expression in osteoblasts. They may play a role in osteoclastogenesis mainly by acting on osteoclast precursors but not on osteoblasts.
It has been suggested that parathyroid hormone or transforming growth factor-␤ stimulates RANKL expression via a PKA-CREB pathway and that CREB may be a central regulator of RANKL expression (8,9,46) in mouse stromal/osteoblast cells. Because ROS also stimulates PKA-CREB signaling in some cells (47), we then focused on the roles of PKA and CREB in ROS up-regulation of RANKL in mouse osteoblasts. We found that KT5720, a specific inhibitor of PKA, reversed ROSincreased RANKL expression (Fig. 5A). Moreover, the PKA activator forskolin induced RANKL mRNA expression in a time-dependent manner (Fig. 5B). The role of PKA is to phosphorylate and activate CREB, and we speculated that ROS activated CREB via PKA activation. Our results demonstrated that H 2 O 2 , XXO, or forskolin phosphorylated CREB/ATF2 (Fig.  4) and significantly increased levels of CREB binding to CRElike domain located in murine RANKL promoter region in mouse osteoblasts (Fig. 6A). The DNA-protein complex decreased notably when cells were treated with scavenger of H 2 O 2 or XXO or inhibitor of PKA before H 2 O 2 or XXO treatment (Fig. 6A). Furthermore, the results of RNAi CREB1 transfection (Fig. 5C) clearly showed that CREB mediates ROS up-regulation of RANKL expression in mouse osteoblasts.
Although immediate upstream sequences from the transcription start site (215 bp) of the mouse and human RANKL gene appear to be conserved (74% identity), there is no significant homology of sequences further upstream (48). It is suggested that regulatory mechanisms of mouse and human RANKL gene must be different. Recently, it was reported that binding of HSF2 to HSE in the human RANKL gene promoter region plays an important role in modulating RANKL gene expression in human stromal/osteoblast cells (7). Consistent with this, we found that PKA inhibitor KT5720 did not change ROS-induced RANKL expression. HSF2 activator MG132 stimulated RANKL expression in human osteoblast-like MG63 cells (Fig.  8C). H 2 O 2 , XXO, or MG132 significantly increased levels of HSF2 binding to HSE in human RANKL promoter region in MG63 cells (Fig. 8D). Moreover, the results of RNAi HSF2 transfection (Fig. 8, B and E) showed that HSF2 binding to HSE mediates ROS up-regulation of RANKL expression in human MG63 cells.
ERKs is another signaling pathway required for ROS-stimulated RANKL expression either in mouse osteoblasts or in human MG63 cells based on our results (Fig. 5A). ERKs have been shown to be activated by oxidative stress (12) and act as an upstream stimulator of CREB (49,50) or HSF (51) under some circumstances. It also plays an important role in osteoclasts differentiation by mediating RANK signaling in osteoclast precursors (2) and high extracellular Ca 2ϩ -induced RANKL in mouse osteoblast. Our results demonstrated that ROS stimulated ERKs phosphorylation, and ERKs inhibitor PD98059 reversed ROS-induced RANKL expression in mouse osteoblast (Figs. 4 and 5A) and human MG63 cells (data not shown). However, PD8059 neither suppressed ROS-increased CREB/ATF2 phosphorylation and CREB binding to CRE-domain in mouse RANKL promoter (Fig. 6B) nor inhibited HSF2 binding to HSE in human RANKL promoter elicited by ROS (Fig. 8F). Therefore, ERKs is not upstream of CREB or HSF2 transcription factors in ROS-induced RANKL expression.
In conclusion, this study provides evidence that ROS stimulates osteoclastogenesis via ERK-and PKA-mediated induction of RANKL in mouse stromal/osteoblast cells and that ROS stimulation of RANKL in mouse osteoblasts and human osteoblastic-like MG63 cell line involves CREB-or HSF-mediated transcription, respectively. Our results predict that bone lossrelated diseases should be prevented by therapies that increase oxidant defenses in bone and inhibit ROS-stimulated RANKL signaling in stromal/osteoblast cells.