BMP-2 and insulin-like growth factor-I mediate Osterix (Osx) expression in human mesenchymal stem cells via the MAPK and protein kinase D signaling pathways.

Genetic studies place the transcription factor Osterix (Osx) downstream of Runx2, but limited information is available about Osx regulation during osteoblastic differentiation. An important role for bone morphogenetic protein-2 (BMP-2) and insulin-like growth factor-I (IGF-I) on Osx expression and the requirement for p38 for the BMP-2-mediated effect was reported previously by our group. In this study, we continued to investigate the molecular mechanisms by which BMP-2 and IGF-1 regulate Osx expression during osteoblast lineage progression. IGF-I-mediated Osx expression required all three MAPK components (Erk, p38, and JNK), whereas BMP-2 required p38 and JNK signaling. As a common mediator of growth factor signaling, we also investigated the involvement of protein kinase C/D (PKC/D) signaling. BMP-2- and IGF-I-mediated Osx expression was blocked in response to a PKD inhibitor. A selective inhibitor of conventional PKCs had no effect on the BMP-2-mediated Osx expression. BMP-2 and IGF-I induced a selective phosphorylation of PKD, and PKD was required for mineralization. PKC/D and MAPK signaling also mediate Runx2 activity. Therefore, to document the implication for Runx2 in Osx regulation, we blocked Runx2 activity using a dominant negative Runx2 construct and an ubiquitination mediator for Runx2 degradation. We showed that blocking Runx2 activity inhibited the BMP-2-mediated induction of Osx. These studies implicated that multiple signaling pathways mediate Osx, a critical gene for osteoblast differentiation and bone formation. In addition to Runx2, other signaling components may be necessary to regulate Osx during osteoblast lineage progression.

Osterix (Osx) 1 is a zinc finger-containing transcription factor expressed in osteoblasts of all endochondral and intramembra-nous bones (1). Genetic studies have indicated that formation of cortical bone and bone trabeculae are abolished in the absence of Osx, thus indicating its significance in osteoblast differentiation (2). One interesting aspect of the studies in the osteogenic cells of Osx-null mutants was that Runx2 expression levels were comparable with those in wild-type osteoblasts. On the other hand, Runx2-deficient embryos did not express Osx, suggesting that Osx acts downstream of Runx2 (1). Currently, there is no evidence to indicate whether Runx2 and Osx functionally or physically interact.
BMP-2 has been reported to induce Osx expression in mouse progenitor cells and chondrocytes (1,3). An extracellular matrix protein, Phosphophoryn, mediated Osx expression in NIH3T3 cells but not in human mesenchymal stem cells (hMSC) or MC3T3.E1 (4). We previously identified the involvement of BMP-2 and IGF-I in mediating Osx expression in hMSC (5). The BMP-2-induced effect on Osx expression was mediated via p38 but not via Erk. Under osteogenic culture conditions, both Erk and p38 were involved in mediating Osx expression (5).
Recent reports suggest that during osteoblast differentiation, BMP-2 activates JNK and p38 via protein kinase D (PKD), independent of protein kinase C (PKC) activity (6). The PKC signaling pathway has been reported as a stimulator of JNK and Erk MAPK components (7). PKC signaling is activated by the second messenger diacylglycerol in response to lipid-derived signal transduction across the plasma membrane (8,9). Multiple isoforms of PKCs are classified into three major categories based on their biochemical properties and sequence homologies: the conventional PKCs, the novel PKCs, and the atypical PKCs (8,10). The serine protein kinase PKC-(PKD) contains a cysteine-rich motif longer than that of conventional or other novel PKCs (8,9,11), thus establishing it as a novel PKC. PKD is ubiquitously expressed and is involved in diverse cellular activities (12)(13)(14).
In osteoblastic cell differentiation, PKD is activated in response to BMP-2 independent of PKC activity and cross-talks to MAPK components (6,15). Activation of PKD activates JNK and p38 in response to BMP-2 stimuli in MC3T3.E1 cells (6). PKD regulation via PKC may determine the modulation of Erk and JNK signaling activation in HEK 293 cells (16). Therefore, PKC/D and MAPK signaling downstream of BMP-2 may be regulated in a different manner for specific cell types and cellular activities. MAPK and PKD also mediate the IGF-Iinduced effect on various cellular activities (17,18); however, no reports indicate their involvement as a mediator for the IGF-I-induced effect on Osx expression in hMSC.
MAPK and IGF-I are also reported to influence Runx2, a master gene of osteoblast lineage progression, which acts downstream of BMP-2/Smad signaling (19,20). Several studies implicate that alternative signaling pathways may act independent of, or in parallel to, Runx2 during osteoblast lineage progression (4,(21)(22)(23). Therefore, we investigated whether Osx expression in mesenchymal progenitor cells required Runx2 in the absence or presence of the BMP-2 stimuli.
Previous studies implicate that during osteoblast differentiation, the MAPK-induced signal was downstream of Ras, a substrate of growth factor signaling, which cross-talks to PKC/D signaling (5,10,24). As a common mediator of BMP-2 and IGF-I signaling, we wanted to delineate the role for PKC/D signaling in the transcriptional regulation of Osx. Stress-activated MAPK component JNK was also studied as a mediator of the growth factor-induced effect on osteoblast lineage progression. Our results suggested that BMP-2-and IGF-I-induced Osx up-regulation proceeds via PKD. Further, osteoblast differentiation was arrested due to PKD inhibition as the cells did not mineralize and Osx expression was abolished. The synergistic interactions displayed by BMP-2 and IGF-I were also disrupted due to the inhibition of PKD. Additionally, Runx2 was required but not sufficient for the BMP-2-mediated Osx induction. This is the first study to show that Osx regulation by BMP-2 and IGF-I proceeds via PKD and MAPK signaling in hMSC. Multiple signaling pathways converged on Osx regulation to mediate osteoblast lineage progression.
Stephen Miller (generated as described in Refs. 27 and 28). Total protein assay kits and protease inhibitors were from Pierce. Effectene transfection reagent, DNase I, and RNeasy kits were obtained from Qiagen (Valencia, CA). Dual-reporter luciferase system was purchased from Promega (Madison, WI). The Ribogreen kit was from Molecular Probes (Eugene, OR). All quantitative PCR reagents were purchased from Applied Biosystems (Foster City, CA).
Cell Culture-hMSC and C3H10T1/2 cells were plated in 35-mm culture wells and grown in basal media approximately to 70% confluence. The basal medium for C3H10T1/2 cells was Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin. Basal medium for hMSC was complete mesenchymal stem cell medium (Biowhittaker). For mineralization assays, cells were maintained in osteogenic stem cell medium supplemented with 50 g/ml L-ascorbic acid, 10 mM ␤-glycerol phosphate, and 100 nM dexamethasone as supplied by the manufacturer (Biowhittaker osteogenic kit). Medium was renewed every second day. hMSC were treated with recombinant human (rh)BMP-2 (100 ng/ml) and/or IGF-I (200 ng/ml) for 48 h in gene expression studies and for 1 h in Western blot analysis based on prior work, indicating that this time point is optimal. For MAPK and PKC/D inhibition experiments, hMSC were treated with the following chemical inhibitors: 40 M U0126 (Erk1/2 inhibitor), 10 M SB203589 (p38 inhibitor), 10 M JNKII (JNK inhibitor), 10 M Go6976 (PKD inhibitor), or Go6983 (PKC inhibitor) as indicated under "Results." Growth factor and chemical inhibitor doses reported here were optimal for signal transduction as reported previously (5). The specificity and the efficacy of the chemical inhibitors in hMSC have been confirmed with the use of genetic tools to block the specific signaling component as reported previously (5).
RNA Extraction and Quantification-Total RNA was extracted using the RNeasy kit and DNase I treatment according to the manufacturer's protocol. The amount of extracted RNA was quantified using the Ri-boGreen RNA quantification kit, and total RNA quantitation was conducted as described previously (5).
Real-time RT-PCR Analysis-Real-time RT-PCR reactions were performed using One-Step Taqman® RT-PCR master mix, primers, and probes (Applied Biosystems). 80 ng of total RNA was used per 10-l reaction with Taqman® one-step RT-PCR master mix, sequence-specific primers (50 -200 nM), and Taqman® probes (100 nM). Real-time RT-PCR assays were carried out in triplicate using an ABI Prism 7900 sequence detection system and analyzed by SDS 2.1 Software. Genespecific primers, probes, and real-time quantitative PCR conditions and analysis have been described previously (4,5).
Alizarin Red Staining-hMSC were seeded at a density of 300,000 cells/cm 2 and treated with MAPK or PKC/D inhibitors in complete osteogenic stem cell medium for 28 days. The analogs for the inhibitors and the control (Me 2 SO) treatment gave the same results. Consequently, results are reported in comparison with the Me 2 SO (control) treatment. To stain for calcium deposition, cells were fixed with 70% ethanol, rinsed five times with deionized water, treated 10 min with 40 mM Alizarin red stain at pH 4.2, and then washed with 1ϫ phosphatebuffered saline for 15 min with gentle agitation.
Transient Transfections-C3H10T1/2 cells were plated in 12-well plates at an initial density of 4 ϫ 10 5 cells/well in complete medium the day before transfection. The cells were transfected with 1 g of the reporter plasmid pGL3(pro)-6XOSE2-Luc along with empty vector or AML1/ETO or Smurf1 plasmid at various doses (as indicated under "Results") using Effectene transfection reagent according to manufacturer's description. In each case, 10 ng of pRL-Luc was co-transfected to provide a means of normalizing the assays for transfection efficiency. Cell lysates were collected in lysis buffer 48 h after transfection, and Renilla and luciferase activities were measured with the use of a Dualreporter luciferase assay system in a Tecan SpectraFluorPlus platereader (Plastic Surgery Laboratory, University of Pittsburgh).
Statistical Analysis-For quantitative assays, the coefficient of variation was calculated from three assay replicates and did not exceed 3% for all treatment groups. Intra-day variation did not exceed 5%. Treatment groups within experiments were performed in triplicate and reported as means Ϯ S.E. Each experiment was repeated three times. Statistical analysis was performed using Statview software (Cary, NC) to determine significance among treatment groups. Analysis of variance followed by Tukey-Kramer post hoc test was performed for the experiments. Statistical significance was established at p Յ 0.05.

RESULTS
Within 1 h of treatment, rhBMP-2 activated PKD and JNK in hMSC (Fig. 1, A and B). For PKD activation, we used two different PKD antibodies to detect Ser 744/748 and Ser 916 phosphorylation in hMSC. In response to BMP-2, Ser 916 was phosphorylated; however, no change in phosphorylation of Ser 744/748 was detected. We previously reported that MAPK components Erk and p38 were also activated in response to BMP-2 stimuli (5). To study the implication of stress-activated MAPK component JNK and protein kinase D (PKC/D), we used selective inhibitors. PKC inhibitors Go6976 and Go6983 have different PKC inhibitory specificities; although Go6976 inhibits Ca 2ϩdependent isoforms and PKD, Go6983 does not inhibit PKD and serves as a general PKC inhibitor (6, 29).
Using quantitative real-time PCR analysis, we determined that Osx expression was inhibited (p Ͻ 0.0001) due to Go6976 or JNKII treatment in the presence of BMP-2, whereas Go6983 did not have a significant effect (p ϭ 0.9755) ( Fig. 2A). BMP-2-mediated Alp expression was inhibited in response to PKC/D or JNK inhibitors (p Ͻ 0.0001) (Fig. 2B). Furthermore, mineralization of hMSC is blocked due to the inhibition of PKD with Go6976 (Fig. 2C). In the presence of JNK inhibitor or the general PKC inhibitor, the cells exhibited mineralization. Tumor phorbol ester, 13-myristate acetate (1 nM 12-O-tetradecanoylphorbol-13-acetate), a known inducer of PKC signaling via diacylglycerol, was included as a positive control for the mineralization assay (Fig. 2C).
Among various growth factors implicated in bone formation, BMP-2 and IGF-I up-regulated Osx expression during early osteoblast differentiation (5). Here, we analyzed the effect of IGF-I on MAPK and PKD signaling. IGF-I exhibited a significant but mild effect on p38 and JNK phosphorylation and a dramatic effect on Erk1/2 (Fig. 3A). IGF-I also activated PKD-Ser 916 (Fig. 3B). Osx expression mediated by IGF-I required MAPK components as well as PKC/D (Fig. 4A). IGF-I is known to act on several different cellular substrates to mediate its effects, and the effects may be elicited via MAPK and PKC components.
To further document the involvement of MAPK and PKD signaling downstream of the growth factor-induced effect, we inhibited MAPK and PKD components in the presence of BMP-2 and IGF-I. When compared with BMP-2, IGF-I had a moderate (but significant, p Ͻ 0.0001) effect on Osx expression in undifferentiated hMSC. BMP-2 and IGF-I induce Osx in hMSC (5), and their synergistic effect on Osx requires PKD (p Ͻ 0.0001) (Fig. 4B).
It has been reported that MAPK and PKC/D may be involved in regulating post-translational levels of Runx2 (19,30). Current studies are limited to Runx2-null mice and do not indicate whether Runx2 is sufficient for Osx expression. To determine the requirement for Runx2 in BMP-2-mediated and basal Osx expression in mesenchymal progenitor cells, we used mouse C3H10T1/2 cells. The cells were transfected with two different constructs to block Runx2 activity: an ubiquitin ligase mediator (Smad ubiquitin regulatory factor 1 (Smurf1)) and a dominant negative Runx2 (AML/ETO). Smurf1 is a member of the HECT family of E3 ubiquitin ligases and has been reported to interact with Runx2 and mediate its degradation via the 26 S proteasome (27,31). Therefore, to determine whether Smurf1-mediated Runx2 degradation causes functional changes in Runx2, we examined the ability of Runx2 to activate a specific reporter gene, 6XOSE2-Luc, in C3H10T1/2 cells. The pGL3-6XOSE2-Luc construct was transiently co-transfected into C3H10T1/2 cells with increasing amounts of Smurf1 expression plasmid. Reporter activity that was induced with endogenous Runx2 activity was reduced (Fig. 5A), and degradation of Runx2 protein was confirmed with Western blot analysis (Fig. 5B). Smurf1 mediated Runx2 degradation and caused a decrease in Runx2 activity.
We confirmed our results using a dominant negative Runx2 construct to block Runx2 activity. AML1/ETO does not contain the transactivation domain and the C-terminal domain of Runx2 (26,32). It contains only the DNA binding domain of Runx2 and acts in a dominant negative manner. Runx2 activity was also reduced due to overexpression of AML1/ETO in C3H10T1/2 cells (Fig. 5A).
To determine the negative effects of AML1/ETO and Smurf1 on Runx2 activity, we transfected cells with empty vector or AML1/ETO or Smurf1 constructs in the presence or absence of BMP-2. Gene expression analysis indicated that Runx2 plays an important role for both basal and BMP-2-mediated-Osx expression in C3H10T1/2 cells (Fig. 6A). Both Smurf1 and AML1/ETO overexpressing cells exhibited Osx expression in the presence of BMP-2 treatment. This outcome indicated that Runx2 activity is not entirely blocked (leaky expression) or that additional signaling components were still active to induce Osx expression downstream of BMP-2 signaling. Alp expression was also inhibited due to blocking Runx2 activity with AML/ ETO or Smurf1 (Fig. 6B). DISCUSSION Skeletal development requires the integration of multiple signaling pathways, and understanding the multitude of cellcell signals remains as a challenge. Bone formation and remodeling require the careful orchestration of a number of signaling components, which in turn regulate the activities of osteoblast lineage-specific master genes and their cellular substrates. Runx2 and Osx are key transcription factors that control the osteoblast fate. Although current literature provides a broad assessment of the regulatory mechanisms that control Runx2, our knowledge of Osx regulation is limited. We previously reported that BMP-2 and IGF-I induce Osx in hMSC (5). The MAPK component p38 is required for BMP-2 induced effect, whereas Ras and one of its downstream effectors Erk were required for mineralization (5).
Ras proteins act as molecular switches downstream of tyrosine kinases (33). Ras GTP-binding proteins activate Raf kinase activity (34,35), in addition to phosphatidylinositol 3-kinase (36), ralGDS (37), and ATF6 (38,39). Activated Raf initiates a signal transduction cascade that involves components of the MAPK family members. In addition to Ras, Raf protein kinase has also been reported to be activated via diacylglycerol-regulated PKC (24); PKC induces cross-talk between PKC, Raf, and MEK/ERK signaling pathways (40 -42). The signaling cascades downstream of growth factors, mediated by PKC and MAPK, affect cellular substrates that regulate osteoblast proliferation and differentiation. Here, we presented our studies to elucidate the possible involvement of PKC and MAPK components in eliciting the BMP-2-and IGF-Imediated effect on Osx transcriptional regulation in mesenchymal stem cells.
We showed that BMP-2 requires PKD, but not the conventional PKCs, to stimulate Osx expression during osteoblast lineage progression (Fig. 2A). As an upstream regulator of p38 and JNK signaling (6), inhibition of PKC/D blocked BMP-2mediated Alp expression (Fig. 2B). PKD was also required for mineralization of hMSC (Fig. 2C). Within 1 h of BMP-2 treatment, we detected phosphorylation of the Ser 916 residue of PKD (Fig. 1A); however, we were unable to detect any changes in the phosphorylation of Ser 744/748 . A study by Lemonnier et al. (6) presented similar results in MC3T3.E1 cells. It has been reported that Ser 916 is activated in response to autophosphoryl- ation of PKD and serves to monitor PKD catalytic activity (43). Additionally, Ser 744 and Ser 748 are the activating residues of the PKD catalytic domain (44). The same group of researchers have also reported that PKC phosphorylates these residues, leading to activation of PKD (45). In addition to PKD, the requirement for JNK in Osx expression is evident, as indicated by gene expression analysis (Fig. 2B). As an activator of Smad signaling, which cooperates with AP-1, BMP-2 signaling may require JNK to activate the AP-1 complex. AP-1-dependent transcription of target genes can be achieved by phosphorylation of the c-Jun transactivation by JNK (46 -48). BMP-mediated activation of JNK and p38 proceeds via PKD in MC3T3.E1 cells (6), and a similar scenario may also be valid in hMSC.
MAPK and PKD signaling components may also mediate the IGF-I-induced effects on cell proliferation and differentiation. Although the IGF-I-mediated stimulation of Osx is not as dramatic as BMP-2, combined delivery of these growth factors gave a synergistic response in hMSC. Evidence from ST2 cells indicates that a decline of IGF-I may be necessary to permit terminal differentiation of mouse stromal cells; 2 however, at an early step in proliferation-differentiation, IGF-I may still have a role as a positive regulator, perhaps via cross-talk with BMP-2 signaling. IGF-I activates MAPK and PKD within 1 h of treatment in hMSC (Fig. 3, A and B). We detected a decrease in Osx stimulation when hMSC were treated with inhibitors of MAPK (Erk, p38, and JNK) as well as PKC/D components (Fig. 4A). MAPK and PKC/D also mediate the IGF-I-induced effect in other cellular systems. PKC is thought to bind the CCAAT/enhancer-binding protein site within the promoter region of human IGF-I to mediate its expression (49). Further, IGF-I is reported to activate JNK in promoting the survival of T lymphocytes (50) and Erk1/2 and JNK in promyelocytic cell lines (18). IGF-I also activates PKC isoforms in several different processes (17,51,52); however, our study is a first in showing the activation of PKD in response to IGF-I stimulation during osteoblast differentiation of hMSC. The PKD activation mediated by IGF-I may not necessarily proceed independent of PKC activity (Fig. 4A), a point that will be the focus of future studies.

FIG. 3. IGF-I activates MAPK and PKD activation in hMSC.
Confluent hMSC in basal stem cell medium were treated for 1 h with 200 ng/ml rhIGF-I. Cell lysates were harvested and subjected to SDS-PAGE, and in A, probed for phospho-specific Erk or p38 or JNK antibodies for Western blot analysis. The membranes were stripped, and in B, reprobed for total Erk, p38, or JNK phospho-specific PKDSer 916 antibody for Western blot analysis. The membrane was stripped and reprobed with total PKD antibody. The membranes were stripped and reprobed with ␤-actin antibody to show equal loading of the protein samples in the membrane. In addition to MAPK and PKD, Runx2 mediates the BMP-2 signal and is regulated at different levels via MAPK and PKC. Runx2 is a critical gene for progression of osteoblast differentiation (28), and its association with Osx has been presented in genetic studies (1). Several studies implicate the presence of Runx2-independent mechanisms for ossification (21)(22)(23). These studies implicate that additional signaling pathways may act in parallel to, or independent of, Runx2 during osteoblast differentiation. Our studies in hMSC with the use of a protein translation inhibitor (cycloheximide) indicate that the BMP-2mediated induction of Osx requires de novo protein synthesis even in the presence of Runx2 expression. 3 Here, we have also presented evidence for the requirement for Runx2 in mediating the BMP-2-induced effect on Osx expression. In agreement with studies by Zhao et al. (27,53), we showed that Smurf1 blocks Runx2 and that Runx2 was required for BMP-2-mediated Osx expression (Fig. 6A). Smurf1 exhibits inhibitory effects on BMP signaling in cooperation with Smad6/7 in vivo (54). We confirmed our results with the use of a dominant negative Runx2 construct, which is reported to exhibit strong DNA binding activity but lack transactivation ability (32,55). Interestingly, overexpression of the Runx2 in-hibitory constructs significantly decreased, but did not abolish, Osx expression (Fig. 6A). This result indicated either that there was still leaky expression of Runx2 despite the presence of inhibitory factors or that additional factors (e.g. Dlx5) acting downstream of BMP-2 could induce Osx independent of the levels of Runx2 activity.
MAPK and PKC/D components may regulate Osx directly or indirectly via their effects on signaling components acting downstream of the growth factors. It remains to be determined whether the signaling molecules such as Dlx5, MAPK, and PKC/D regulate the Osx promoter. Milona et al. (56) have reported that there is an OSE2 element in the Sp7 regulatory region, so the Osx promoter may be a direct target of Runx2. Further analysis on the relationship between Runx2 and Osx will be the focus of future studies.
Growth factors and downstream signaling components may also exert their effects at the post-translational level. Although we and others have been unable to detect changes in Runx2 expression levels in response to IGF-I (5, 57), IGF-I may affect Runx2 or Osx activity at the post-translation level. It is possible that the BMP-2 mediated effect on Osx may require not only an adequate amount of Runx2 protein but also post-trans- 3 A. B. Celil and P. G. Campbell, unpublished results.
FIG. 5. Smurf1 and AML1/ETO regulate Runx2 biological activity. A, C3H10T1/2 cells were transfected with empty vector (pCMV) with or without Runx2 reporter gene, 6XOSE2-Luc, or increasing concentrations of AML1/ETO or Smurf1 constructs with 6XOSE2-Luc in the presence of 10 ng of pRL vector. Cell lysates were collected 48 h after transfection and subjected to luciferase assay with Dual-reporter assay system. The total amount of DNA used for transfection was equalized with pCMV. *, significant from empty vector with reporter gene (pCMVϩ 6XOSE2-Luc), p Ͻ 0.05. Results were reported as mean of three independent cultures ϮS.E. B, C3H10T1/2 cells were transfected with empty vector (pCMV) or Smurf1 (1.5 g) construct. Cell lysates were collected 48 h after transfection, and Western blot analysis was conducted to determine Runx2 protein levels. The membrane was stripped and reprobed with ␤-actin antibody to show equal loading of the protein samples.
FIG. 6. Runx2 is required for Osx transcriptional regulation. C3H10T1/2 cells were transfected with empty vector (pCMV) (1.0 g), AML1/ETO (1.0 g), or Smurf1 (1.5 g) in the presence or absence of rhBMP-2 (300 ng/ml). A, Osx gene expression was analyzed. B, Alp gene expression was analyzed with quantitative real-time PCR analysis. *, significant from control (pCMV) without BMP-2 treatment, p Ͻ 0.05; **, significant from control (pCMV) with BMP-2 treatment, p Ͻ 0.05. Results were reported as mean of three independent cultures ϮS.E. lational modification of Runx2. Further, a study by Qiao et al. (20) reveals that IGF-I regulates Runx2 binding via activation of phosphatidylinositol 3-kinase and Erk. Whether this effect is critical for Osx regulation remains unanswered.
In summary, we have shown that MAPK and PKD signaling pathways serve as points of convergence for mediating the BMP-2-and IGF-I-induced effects on Osx expression in mesenchymal stem cells. The integration of signaling pathways is complex in nature and may involve negative feedback mechanisms for tight control of osteoblast differentiation. In view of the importance of Osx as a pivotal regulator of osteoblast lineage progression and bone formation, this study further advanced our understanding of how multiple signaling pathways orchestrate their effects to regulate Osx and other bone-specific genes.