Transcriptional Regulation of the Osterix (Osx, Sp7) Promoter by Tumor Necrosis Factor Identifies Disparate Effects of Mitogen-activated Protein Kinase and NFκB Pathways*

Osteoblast (OB) differentiation is suppressed by tumor necrosis factor-α (TNF-α), an inflammatory stimulus that is elevated in arthritis and menopause. Because OB differentiation requires the expression of the transcription factor osterix (Osx), we investigated TNF effects on Osx. TNF inhibited Osx mRNA in pre-osteoblastic cells without affecting Osx mRNA half-life. Inhibition was independent of new protein synthesis. Analysis of the Osx promoter revealed two transcription start sites that direct the expression of an abundant mRNA (Osx1) and an alternatively spliced mRNA (Osx2). Promoter fragments driving the expression of luciferase were constructed to identify TNF regulatory sequences. Two independent promoters were identified upstream of each transcription start site. TNF potently inhibited transcription of both promoters. Deletion and mutational analysis identified a TNF-responsive region proximal to the Osx2 start site that retained responsiveness when inserted upstream of a heterologous promoter. The TNF response region was a major binding site for nuclear proteins, although TNF did not change binding at the site. The roles of MAPK and NFκB were investigated as signal mediators of TNF. Inhibitors of MEK1 and ERK1, but not of JNK or p38 kinase, abrogated TNF inhibition of Osx mRNA and promoter activity. TNF action was not prevented by blockade of NFκB nuclear entry. The forced expression of high levels of NFκB uncovered a proximal promoter enhancer; however, this site was not activated by TNF. The inhibitory effect of TNF on Osx expression may decrease OB differentiation in arthritis and osteoporosis.

Osteoblasts (OBs) 2 are derived from pluripotent precursor cells of mesenchymal origin that are capable of differentiation to chondrocytes, myocytes, adipocytes, or fibroblasts (1). Bone formation in the embryo and remodeling in the adult require that a sufficient number of precursor cells differentiate to functional OBs. New OBs are continuously required for the synthesis of bone matrix and replacement of cells becoming osteocytes or undergoing apoptosis. A coordinated expression of transcription factors determines the commitment of precursor cells toward the OB phenotype under the control of autocrine, paracrine, and hormonal stimuli. Two of these transcription factors, RUNX2 (Cbfa1/AML3/Pebp2␣A) and Osx, are required for differentiation to the OB lineage. In mice, RUNX2 gene knock-out causes a lethal mutation with a cartilaginous skeleton. RUNX2 is presumed to function as an organizer on promoters of skeletal-specific genes (2). A phenotype similar to the RUNX2 knock-out is observed with knock-out of Osx. Here the arrest in development occurs slightly later but also results in a cartilaginous skeleton (3). In addition, Osx induces OB differentiation of dispersed embryonic cells (4). RUNX2 is expressed in Osx knock-outs, suggesting that Osx functions downstream of RUNX2 in the differentiation pathway.
Differentiation of precursor cells to OBs in adult bone is impaired by inflammatory stimuli. In rheumatoid arthritis, estrogen deficiency, and aging, there is an increased expression of cytokines, including tumor necrosis factor-␣ (TNF-␣). In adult bone, inflammatory cytokines blunt the formation rate of new bone in the face of increased resorption, contributing to net bone loss (5)(6)(7)(8)(9)(10)(11). Cytokines could interfere with the expression of factors required for OB differentiation.
We have previously shown that TNF inhibits osteoblast differentiation at the stage of precursor cell commitment to the OB lineage (6). Osteoblast precursors, including fetal calvaria cells, murine marrow stromal cells, and the clonal pre-osteoblastic cell line MC3T3, fail to differentiate in the presence of TNF. These models of osteoblast progenitors uniformly show enhanced sensitivity to TNF blockade of differentiation at an early stage in culture when the key transcription factors RUNX2 and Osx are required. An inhibitory effect of TNF on the RUNX2 promoter that could contribute to decreased expression and differentiation of cells has been described previously (12). This effect of TNF is isoform-specific, inhibiting the osteoblastic MASNS RUNX2 isoform 50% and the more ubiquitously expressed MRIPV isoform Ͼ90%. These results suggest that TNF may have additional targets.
Here we present evidence that TNF is a potent inhibitor of Osx expression. In addition, we have evaluated the structure and regulation of the Osx promoter and report transcriptional regulation by TNF at a discrete site via a mitogen-activated protein kinase (MAPK) signal.

MATERIALS AND METHODS
Reagents-MC3T3-E1 (clone 14) mouse pre-osteoblast cells were provided by Dr. Renny Franceschi (University of Michigan). C3H10T1/2 cells were obtained from the America Type Culture Collection (Manassas, VA). Human TNF-␣ was purchased from Pepro-Tech (Rocky Hill, NJ). Real-time PCR was done using the Bio-Rad ICycler. SYBR Green was obtained from Bio-Rad. MAPK inhibitors PD98059 and SB203580 were obtained from Calbiochem. SP600125 * This work was supported by National Institutes of Health Grant AR046452 and a Veterans Affairs Merit Review grant (to M. S. N.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. was purchased from Tocris Cookson (Ellisville, MO). Minimal essential medium was purchased from Invitrogen and fetal bovine serum from Hyclone (Logan, UT). Other reagents were obtained from Sigma. Cell Treatment and RNA Harvest-MC3T3-E1 cells were plated on day 0 at 7.4 ϫ 10 6 cells/150-mm plate in minimal essential medium ϩ 10% fetal bovine serum (medium). On day 1, medium was replaced with differentiating medium (minimal essential medium ϩ 10% fetal bovine serum ϩ 50 g/ml L-ascorbate). On day 2, TNF-␣ was added in the doses indicated for each experiment. The half-life of Osx mRNA was measured in MC3T3-E1 cells plated on day 0 at 4.4 ϫ 10 5 cells/well in 6-well plates in medium. Actinomycin D (0.5 g/ml) was added 2 h prior to TNF-␣, and RNA was obtained at the time points indicated under "Results." For experiments using cycloheximide, the same protocol was followed as for actinomycin D, except 5 g/ml cycloheximide was used. RNA was prepared using the RNeasy Mini Kit (Qiagen, Valencia, CA). The addition of MAPK chemical inhibitors was done 2 h prior to TNF treatment. Transient transfection of dominant negative MEK1 or ERK1 was done with the Osx promoter reporter construct followed by TNF 24 h later and cell lysates for luciferase assay after an additional 24 h. The addition of the siRNA-p65 expression plasmid or the siRNA control (Santa Cruz Biotechnology, Santa Cruz, CA) was done 48 h prior to TNF and transfection of the N terminus-deleted IB 24 h prior to TNF treatment.
Real-time RT-PCR-Quantitation of Osx mRNA in total cell RNA samples was carried out using standard RT and PCR procedures, setting up duplicate or triplicate reactions. The primers used were OsxQRTf, 5Ј-CCT CTC GAC CCG ACT GCA GAT C-3Ј and OsxQRTr, 5Ј-AGC TGC AAG CTC TCT GTA ACC ATG AC-3Ј. The PCR reaction conditions were as follows: 94°C for 30 s, 58°C for 30 s, and 72°C for 20 s for 40 cycles.
Transfection-The C3H10T1/2 cells (n ϭ 3 wells/group) were plated at a density of 7 ϫ 10 4 cells/well in 12-well tissue culture plates (Corning, NY). After 24 h, the cells were transfected with a mixture of Superfect transfection reagent (Qiagen, Valencia, CA), medium, promoter reporter, and pRL-TK control vector (Promega, Madison, WI). 48 h after transfection, cells were harvested and assayed using firefly luciferase and Renilla luciferase substrates in the Dual Luciferase assay system (Promega, Madison, WI). Luciferase values were normalized to Renilla luciferase data to correct for variation in transfection efficiency.
Rapid Amplification of 5Ј-cDNA Ends-The 5Ј-end sequence of the alternatively spliced exon was obtained using BD SMART RACE cDNA amplification kit (BD Biosciences) following the manufacturer's instructions. The primer used was 5Ј-GAG CTT CTT CCT CAA GCA GAG AGG ACG CCA TCC TCG A-3Ј. The PCR products were cloned into the pCR-TOPO vector using the TOPO TA cloning kit (Invitrogen). Automated sequencing was done to determine the transcription start sites from the 5Ј-RACE products.
Primer Extension-Total RNA was extracted as described above and used for primer extension analysis. Reactions were performed by annealing 1 pmol of [ 32 P]-labeled primer, complementary to the Osx gene untranslated 5Ј-end (GGACT GGAGC CATAG TGAGC TTC), to 10 g of total RNA. Following annealing at 58°C for 20 min, extension was performed with 100 units of Moloney murine leukemia virus reverse transcriptase (Invitrogen) and deoxyribonucleotide triphosphates (0.5 mM) for 30 min at 42°C. cDNA products were denatured and analyzed on 6% (w/v) denaturing polyacrylamide gels. The gels were dried and analyzed in a phosphorimaging device.
Statistical Analysis-ANOVA was used to determine a statistical difference between multiple groups. Multiple comparisons between individual groups were done by the method of Tukey. Comparisons between any one group and a common control were done by the method of Dunnet. Analyses were done using Prism software, (Graph-Pad, San Diego, CA).

RESULTS
TNF Inhibits Osx Steady State mRNA-Osx expression was observed by day 2 of culture in MC3T3 cells. The effect of TNF was studied using doses previously shown to inhibit osteoblast differentiation in MC3T3, bone marrow stromal, and fetal rat calvaria pre-osteoblasts (6,14). Fig.  1A shows that treatment with 10 ng/ml TNF on day 2 of culture inhibited Osx steady state mRNA. TNF inhibited Osx mRNA 50% by 4 h, 85% by 8 h, and 95% by 24 h compared with levels in the control cultures. Fig.  1B shows that the inhibitory effect of TNF was dose-dependent with 50% inhibition of mRNA occurring at 0.75 ng/ml. This IC 50 is similar to that reported for TNF inhibition of osteoblast differentiation (6). To determine whether TNF decreased Osx by destabilization of mRNA, MC3T3 cells were treated with actinomycin D 2 h prior to the addition of TNF to stop RNA synthesis. The half-life of the Osx mRNA was then measured over 14 h using real-time RT-PCR. TNF did not decrease the brief (1.5 h) Osx mRNA half-life (Fig. 1C). MC3T3 cells were treated with cycloheximide to determine whether TNF action required new protein synthesis. Fig. 1D shows that cycloheximide alone decreased Osx mRNA, although TNF further inhibited the Osx mRNA level in the presence of cycloheximide treatment.
Cloning of the Osx Promoter and Determination of Transcriptional Start Sites-A 1357-bp fragment of the proximal promoter including a portion of 5Ј-untranslated RNA was amplified from a murine F-factorbased bacterial artificial chromosome containing genomic Osx, sequenced for confirmation, and cloned into the KpnI and XhoI sites of the pGL3 Basic-luciferase reporter. A map of homologous transcription factor binding sites was generated using MatInspector to assess potential regulatory sites (15). Fig. 2 shows that the promoter contains putative homologous binding sites for factors known to influence differentiation of pluripotent precursor cells to the osteoblast, chondrocyte, adipocyte, or myoblast lineage. These included MyoD, AML-1, RUNX2, cEBP, Msx, DLX, and NFB core consensus sequences. The Osx promoter fragment also included 14 contiguous repeats of a Myf5-binding site. The transcriptional start site of the Osx promoter was determined by 5Ј-RACE. Fig. 3A shows that 5Ј-RACE yielded two bands of 100 and 300 kb size, suggesting two potential start sites. (also identified in Fig. 2). These sites were confirmed by primer extension (not shown). Amplification using unique 5Ј primers and a common 3Ј primer yielded two mRNA species designated Osx1 and Osx2 (Fig. 3B). Sequencing revealed that the more abundant Osx1 corresponded to the expected mRNA as reported previously for murine Osx and the highly homologous human Osx (16,17). The less abundant Osx2 includes alternatively spliced exons 1 and 2. Fig. 3C maps the genomic Osx structure as deduced from these experiments.
Deletion Analysis Reveals Independent Promoter Activities Upstream Of Osx1 and Osx2-The basal activity of the Osx promoter was analyzed by making successive deletions from the 3Ј-or 5Ј-end. These were inserted upstream of a promoterless pGL3basic luciferase reporter. Fig.  4A shows a diagram of the Ϫ1269/ϩ91 promoter with the locations of the Osx1 and Osx2 transcription start sites labeled for reference. Fig. 4B shows the effect of the deletions on promoter activity. The proximal promoter containing the Osx1, but not Osx2, start site had 40% of the activity of the full-length reporter (Fig. 4B, construct C versus A). Interestingly, deletion of the regions proximal to the Osx1 start site retained substantial activity (Fig. 4B, constructs D, E, and F versus A). Activity was lost with deletions upstream of Ϫ469 (Fig. 4B, constructs G, J, and K versus A). To determine whether there were independent promoter activities associated with the Osx1 and Osx2 start sites, Ϫ669/Ϫ469 and Ϫ269/ϩ91 fragments were cloned upstream of the promoterless pGLbasic. These reporters retained independent activity that was 40% of the full-length Ϫ1269/ϩ91 (Fig. 4B, constructs H and I versus A) and five times that of the pGLbasic control (H versus L). As previously noted, the region around Osx1 was also capable of independent promoter activity ( Two major signal pathways mediating TNF action are NFB and MAPK (11). The effect of NFB expression on the promoter activity was determined by transfection of C3H10T1/2 cells with NFB or its individual p50 or p65 subunits. Expression was done with a pef-Myc-nuc vector containing an independent nuclear localization signal and driven by a cytomegalovirus promoter (18). Fig. 6 shows the results for this transfection and also the effect of other stimuli for comparison, including RUNX2, Msx2, dexamethasone, 1,25(OH) 2 D3, or parathyroid hormone. Surprisingly, NFB caused a potent stimulation of the Osx promoter. This effect was mediated by the p65 subunit of NFB, which retained 40% of the activity of the intact dimer. The p50 subunit did not significantly increase Osx promoter activity. Msx2 expression increased Osx promoter activity 5.7-fold, but none of the other stimuli were effective. In a separate experiment, treatment with bone morphogenic protein-2 (BMP-2) did not stimulate Osx promoter activity (not shown).
Deletions of the promoter were studied to localize the regions conferring the inhibitory effect of TNF. Fig. 7 shows the effect of TNF as fold stimulation relative to the activity of the respective control constructs. Most of the TNF inhibition was localized to a region between Ϫ514 to Ϫ510, as seen by the effect of deletion of this region (Fig. 7, constructs F  and G versus A). This region was downstream of the Osx2 start site and embedded within the region of independent promoter activity shown in Fig. 4B. To further localize the TNF response element, small deletions or a four-base mutation were made within the Ϫ665/ϩ91 construct (Fig. 7,  constructs D and E). These localized TNF inhibition of the promoter to a region between Ϫ514 and Ϫ510. Confirmation of the localization was done by inserting three copies of the Ϫ520/Ϫ500 sequence upstream of the heterologous SV40 promoter (Fig. 7, construct H). TNF inhibited the activity of this promoter but had no activity on pSV40 alone (not shown).
Identification of Protein-DNA Binding in the TNF-responsive Region-EMSA was done using overlapping probes that spanned 200 bp around the TNF-responsive region to determine sites of nuclear  protein-DNA binding. Nuclear extract obtained from control and TNF-treated C3H10T1/2 cells was used for incubation with the probes. Five sites were found with strong protein-DNA interaction, one of which overlapped the TNF-responsive region (Fig. 8, probe 9, specific). Binding to probes 3 and 5 represented binding to the Myf5 consensus, whereas probe 6 spanned the Osx2 start site. Localization of the TNF response by deletion and mutational analysis led us to focus on a region within probe 9 (Ϫ520/Ϫ500); however, TNF did not change the pattern of binding at this site, although a small 2-fold increase in binding intensity was confirmed in repeated experiments. Incubation with antibodies to transcription factors was done using probe 9, but no supershifts were observed for p65, p50, RUNX2, RUNX1, JunD, RBP-J, Fra-1, or Fra-2 (not shown).
TNF Inhibition of Osx Is Mediated by MAPK Signaling-The TNFbound p55 receptor activates a MAPK cascade in addition to the NFB pathway. These pathways diverge downstream of activation of the cytosolic adapter protein TRAF2 in osteoblastic cells (11). We evaluated MAPK as a possible mediator of TNF inhibitory action, because NFB paradoxically stimulated Osx transcription. Cells were pre-incubated for 2 h with PD98059, SP600125, or SB203580, inhibitors of MEK1 (the immediate upstream activator of ERK1/2), cJun-N-terminal kinase (JNK), or p38 kinase, respectively. The effect of these treatments on TNF inhibition of the Osx promoter was measured in the Ϫ1269/ϩ91 Osx reporter. Fig. 9A shows that the MEK inhibitor reversed TNF inhibition of Osx transcription, whereas the inhibitors of JNK and p38 had no effect. The MEK inhibitor alone increased Osx promoter activity above the level of control. The effect of the MEK inhibitor was also observed using a reporter limited to the Ϫ669/Ϫ469 TNF-responsive region. The MEK inhibitor (PD98059) completely abrogated TNF inhibition of   this limited promoter and also increased basal activity (Fig. 9B). Fig.  9C shows that the MEK inhibitor also abrogated TNF inhibition of the heterologous SV40-LUC promoter bearing the 3ϫ(Ϫ520/Ϫ500) sequence, indicating that the TNF element was sufficient to confer MAPK responsiveness. To determine whether the MEK inhibitor blocked TNF inhibition of Osx mRNA, MC3T3 cells were treated with PD98059 for 5 h before the addition of TNF. RNA was harvested 18 h later for measurement of Osx mRNA by real-time RT-PCR. Fig.  9D shows that the MEK inhibitor completely abrogated TNF inhibition of Osx mRNA. PD98059 alone caused a significant increase in steady state Osx mRNA. Similar results were obtained for the Ϫ1269/ϩ91 Osx promoter using a dominant negative MEK1 or ERK1. Fig. 10 shows that these dominant negatives partially reversed TNF inhibition of the promoter, confirming the results obtained using PD98059.
Identification of the NFB Regulatory Element-The pattern of NFB stimulation of the promoter was determined by overexpression of the p65 subunit, as described for Fig. 6. NFB responsiveness was abolished with deletion of the proximal promoter between Ϫ270 and Ϫ70 (Fig. 11, construct J versus A or I). A promoter fragment containing the TNF response site, but not the proximal promoter sequence, was not stimulated by NFB (Fig. 7, construct H). Further deletions localized the NFB response to a region within 200 bp of the consensus p50 binding site indicated in Fig. 2 (-214/-217). Mutation of four bases representing the core binding motif abolished NFB stimulation of the Ϫ1269/ϩ91 promoter (Fig. 11K). These  1-11). B indicates background activity of the probe alone (no nuclear protein), C shows binding of control nuclear protein, and T represents binding of nuclear protein obtained from TNF-treated cells. There was no effect of TNF on the pattern of binding to the probes, although the intensity of binding is 2-fold higher in the TNF-specific band for probe 9. ns, nonspecific band; specific, bands observed uniquely for the probe; fp, free probe. Specific binding was seen for probes 3, 5, 9, 10, and 11.

FIGURE 9. An inhibitor of MEK1 abrogates TNF inhibition of the Osx promoter and Osx mRNA.
A, C3H10T1/2 cells were treated with inhibitors of MEK1 (PD98059), JNK (SP600125), or p38 (SB203580) 2 h prior to the addition of TNF (10 ng/ml). Cells were transfected with the Ϫ1269/ϩ91 Osx promoter reporter, and activity was measured 18 h later. PD98059 partially abrogated TNF inhibition and increased basal promoter activity. B, as described for A but using a reporter restricted to Ϫ469/Ϫ669 spanning the TNF-responsive region. C, as described for A using three copies of the Ϫ514/Ϫ500 TNF-responsive sequence upstream of a minimal SV40 promoter. D, TNF inhibition of Osx mRNA is prevented by pretreatment with a MEK1 inhibitor. C3H10T1/2 cells were treated with PD98059 5 h prior to the addition of TNF (10 ng/ml), and RNA was isolated 18 h later. Osx mRNA was measured by real-time RT-PCR. Mean Ϯ S.E. Bars labeled with different letters (a, b, and c) are significantly different from each other (p Ͻ 0.05 by ANOVA).
bases and their flanking sequence conferred NFB activation in a heterologous SV40 promoter, confirming independent NFB enhancer activity (Fig. 11L). The potency of the NFB stimulus to Osx transcription was predominant over the inhibitory effect of TNF in the full-length Ϫ1269/ϩ91 construct, because TNF treatment did not inhibit transcription when NFB was simultaneously overexpressed (not shown, C ϭ 1.0 Ϯ 0.1, p65 ϭ 13.1 Ϯ 0.6, TNF ϩ p65 ϭ 12.4 Ϯ 0.8). This paradox led to a more detailed assessment of the effects of these treatments on binding to the NFB element.
The NFB Element Is Not Activated by TNF-The nuclear protein binding characteristics of the NFB-responsive region were investigated by generating probes spanning a 231-bp sequence centered around the p50 binding motif. Fig. 12A shows a map of the promoter and indicates the sequence of probes used for EMSA analysis with recombinant NFB. Fig. 12B shows that probes 2 and 4 bound a recom-binant p50 monomer and a p50/p65 dimer but not the p65 monomer. The sequence bound was concordant with the functional sequence identified by deletion and mutational analysis of the promoter. This sequence is compared with the NFB consensus in Fig. 12C. Binding of recombinant p50 to this sequence was of very low affinity compared with binding to the NFB consensus sequence (Fig. 12D). Nuclear protein from TNF-treated C3H10T1/2 cells did not bind this low affinity site, demonstrating that the amount of NFB stimulated by TNF was insufficient to bind and activate the enhancer (Fig. 12E, lane 5).
To confirm that NFB activation was not modulating TNF inhibition of the Osx promoter, NFB activation was blocked by siRNA-p65 or expression of a degradation-resistant N terminus-deleted IB (⌬IB, (19)) in C3H10T1/2 cells. Fig. 13A shows that neither the siRNA-p65 nor the ⌬IB could prevent TNF inhibition of Osx promoter activity. Fig. 13B shows the efficacy of siRNA-p65 in blocking TNF stimulation of an NFB-dependent reporter.

DISCUSSION
Our results show that TNF regulates expression of Osx by inhibiting the transcriptional activity of its promoter. The low dose of TNF that inhibits Osx expression is similar to doses that inhibit the differentiation of OB (6,14). TNF inhibition of Osx was dose-and time-dependent and observed in two cell lines representative of the early stages of OB differentiation. We have previously shown (12) that TNF inhibits the expression of RUNX2, another critical transcription factor required for OB differentiation (12). Previous studies have shown that both RUNX2 and Osx are needed for OB differentiation. Thus, the effect of TNF can be attributed to suppression of both factors during the early events in OB differentiation.
TNF action on Osx is likely to be transcriptional. First, there was no effect of TNF on Osx mRNA stability. Second, treatment with cycloheximide, although inhibitory on its own, was unable to prevent a further inhibitory action of TNF. These data suggest that the effects of TNF are direct rather than requiring the indirect induction of a protein mediator. The requirement for protein synthesis has been suggested for the stimulation of Osx mRNA by BMP, which is prevented by cycloheximide treatment (20). Consistent with these reports, we did not observe stimulation of Osx promoter activity by BMP-2 (not shown). Finally, studies of the Osx promoter confirm regulation of activity in constructs driving expression of luciferase.
To evaluate TNF action on the Osx promoter, we cloned a 1269-bp fragment upstream of the luciferase reporter. 5Ј-RACE and primer extension analysis revealed two transcription start sites. One of these, termed Osx1, represented the start site for the more abundantly expressed and previously described mRNA species. Previous work (16,17) has shown that the human Osx mRNA has at least two isoforms. These include an abundant form highly homologous to the murine Osx and a scarcer alternatively spliced form. Using selective primers, we confirmed the expression of the two isoforms in MC3T3 cells and hypothesized that they arose from the two transcription start sites. Our data also show that the Osx promoter contains at least two regions capable of independent promoter function. This suggests that the two mRNA isoforms could be expressed under the regulation of the two different promoters, although full transcriptional activity requires both. Osx2 mRNA could function similarly to Osx1, have a unique function, or be expressed as an unstable species that is rapidly removed. Further work will be needed to determine the function of Osx2, its regulation, and whether two protein forms are translated. Osx1 appears to be the major form of Osx mRNA.
TNF inhibited the activity of the Osx promoter in a dose-and time-  dependent manner in both C3H10T1/2 and MC3T3 cells. These results were similar to the dose-and time-dependent regulation of Osx mRNA, with a small delay in TNF action that can be attributed to the longer half-life of luciferase mRNA. Deletion analysis of the promoter localized the inhibitory effect of TNF to Ϫ514/Ϫ510, a region proximal to the Osx2 start site. This region and its flanking sequence contain consensus sites for thyroid transcription factor (TTF1), pleomorphic adenoma gene 1 (PLAG1), AP-2, and the RBP-J site. Although TNF did not change the pattern of binding at this site, a small 2-fold increase in binding intensity was observed. Regulation at this site could occur through phosphorylation of one of the proteins in the bound complex, thereby changing the regulatory effect of the complex from an enhancer to a suppressor. Further investigation will be needed to identify the protein in this complex that is regulated by TNF. TNF signals through multiple intracellular pathways (11). We designed experiments to distinguish the roles of two major TNF pathways, MAPK and NFB, as mediators of TNF inhibition of Osx transcription. In osteoblasts, a TNF trimer binds two receptor forms, TNFSFR1 or TNFSFR2, of which only the TNFSFR1 (type 1, p55) mediates inhibition of OB differentiation (14). In a well established paradigm, the bound receptor activates a large cytosolic complex that includes TRAF2 and IB kinase isoforms. The IB kinase isoform then phosphorylates IB, an NFB binding protein that normally sequesters NFB in the cytoplasm. Phosphorylation of the IB N terminus leads to its degradation and liberation of NFB for nuclear entry and gene regulation. However, TRAF2 also stimulates the MAPK cascade with downstream activation of ERK1/2, p38, or JNK. Our data suggest that TNF inhibition of Osx expression is mediated via MAPK, because the MEK1 inhibitor PD98059 prevents TNF effects on Osx promoter activity and Osx mRNA expression. This blockade of TNF action was limited to the inhibitor of MEK and not observed with inhibitors of p38 or JNK. The MEK inhibitor alone stimulated basal Osx promoter activity and mRNA expression, suggesting that a tonic inhibitory influence of MAPK on the Osx promoter must be active. The MEK inhibitor PD98059 also abrogated TNF inhibition of a reporter containing the isolated Ϫ669/Ϫ460 region or the Ϫ520/Ϫ500 core response element cloned upstream of a heterologous promoter. Thus, our results on localization of TNF action and mediation of the TNF signal by MAPK are concordant. Although our results exclude a role for JNK or p38 in TNF regulation of Osx, PD98059 could still inhibit additional kinases. Data supporting MEK/ ERK as the responsible pathway for TNF action was also supported by the partial abrogation of TNF action by dominant negatives MEK1 and ERK1. Further work will be needed to define the MAPK utilized by TNF and the nuclear protein target of this phosphorylation cascade.
Previous studies provide conflicting evidence on the role of MAPK in OB differentiation that may be explained by specific actions of the individual pathways. Commitment of pluripotent precursors to the OB phenotype is stimulated by a variety of factors, including BMPs that increase expression of the key transcription factor RUNX2 or that modify Osx expression. The sensitivity to such stimuli may be increased or decreased by the different MAPK pathways. Stimulation of OB differentiation by BMP-2 has been shown to require p38 kinase (21)(22)(23)(24). MEK blockade, which would block activation of ERK1/2, is synergistic with BMP-2 action, consistent with this idea. In support of this concept, Higuchi et al. (32) establish that chronic suppression of MEK increases markers of differentiation, including expression of alkaline phosphatase and osteocalcin in MC3T3-E1 cells and of alkaline phosphatase in more primitive pluripotent C2C12 cells. Recently, Osyczka and Leboy (25) showed that inhibition of MEK using PD98059 permitted BMP stimulation of OB phenotypic gene expression in human marrow stromal cells, whereas a constitutively active MEK inhibited this response. In parallel with our findings on TNF, ERK blockade has been shown to prevent inhibition of OB differentiation by other factors, including EGF and FGF (26). These results suggest that some inflammatory signals via ERK inhibit OB differentiation, whereas others via p38 stimulate OB differentiation.
There are clear exceptions to an inhibitory effect of ERK. Cell contact with matrix is one signal that activates ERK and phosphorylates RUNX2, an obligatory step for RUNX2 organization of skeletal gene promoters (27)(28)(29)(30). More recently, prevention of pre-osteoblast apoptosis by wingless factors (wnt) was shown to be ERK-dependent (31). Thus, ERK effects may depend on whether the stimulus also activates other signal pathways or on the stage of cell differentiation (32). Finally, although the MAPK inhibitors used in the present study distinguish between ERK, p38, and JNK, there may still be other closely related kinases to MEK/ERK that could be inhibited by PD98059 that will have to be evaluated (33).
Because TNF also stimulates activation of NFB, we evaluated the effect of NFB expression on the Osx promoter. Our results indicate that NFB does not mediate the inhibitory effect of TNF, as blockade of NFB activation did not prevent TNF action. Surprisingly, targeted expression of NFB to the nucleus using a strong cytomegalovirus promoter containing an independent nuclear localization signal unmasked a potent enhancer function in the proximal Osx promoter. This enhancer was not activated by TNF, as TNF does not stimulate a sufficient elevation of nuclear NFB to bind the low affinity site. Nevertheless, the potency of transcriptional activation observed using the NFB expression vector suggests that the enhancer might be an important regulator of Osx expression and OB differentiation. Such an enhancer might be functional under other circumstances that increase higher levels of NFB. NFB is activated by a variety of stimuli other than TNF, including members of the TNF superfamily that have not been studied for effects on OB (34). In addition, the potency of NFB to stimulate Osx transcription could be modulated by phosphorylation of p65 on key serines and tyrosines shown to modulate p65 action at other targets (18). Such a modulation of NFB potency at its enhancer could serve to balance the negative inflammatory stimulus of TNF via MAPK. Additional information is needed to determine the role of NFB in Osx expression and OB differentiation.
Bone remodeling is a coupled process that balances the rate of resorption with formation. Continued recruitment of new OB from the precursor pool must occur to counteract the resorption stimulus associated with estrogen deficiency at menopause or in inflammatory arthritis. TNF impairs the recruitment of OB and thus blunts the magnitude of bone formation to further shift skeletal balance toward a catabolic state. The inhibition of Osx by TNF may contribute to the mechanism of suppressed OB differentiation.