|Home | About | Journals | Submit | Contact Us | Français|
RUNX2, a key transcription factor for osteoblast differentiation, is regulated by ERK1/2 and p38 MAP kinase-mediated phosphorylation. However, the specific contribution of each kinase to RUNX2-dependent transcription is not known. Here we investigate ERK and p38 regulation of RUNX2 using a unique P-RUNX2-specific antibody. Both MAP kinases stimulated RUNX2 Ser319 phosphorylation and transcriptional activity. However, a clear preference for ERK1 versus p38α/β was seen when the ability of these MAPKs to phosphorylate and activate RUNX2 was compared. Similarly, ERK1 preferentially bound to a consensus MAPK binding site on RUNX2 that was essential for the activity of either kinase. To assess the relative contribution of ERK1/2 and p38 to osteoblast gene expression, MC3T3-E1 preosteoblast cells were grown in control or ascorbic acid (AA)-containing medium ± BMP2/7. AA-induced gene expression, which requires collagen matrix synthesis, was associated with parallel increases in P-ERK and RUNX2-S319-P in the absence of any changes in P-p38. This response was blocked by ERK, but not p38, inhibition. Significantly, in the presence of AA, BMP2/7 synergistically stimulated RUNX2 S319 phosphorylation and transcriptional activity without affecting total RUNX2 and this response was totally dependent on ERK/MAPK activity. In contrast, although p38 inhibition partially blocked BMP-dependent transcription, it did not affect RUNX2 S319 phosphorylation, suggesting the involvement of other phosphorylation sites and/or transcription factors in this response. Based on this work, we conclude that extracellular matrix and BMP regulation of RUNX2 phosphorylation and transcriptional activity in osteoblasts is predominantly mediated by ERK rather than p38 MAPKs.
Several transcription factors including DLX5/6, TWIST1/2, ATF4, RUNX2 and Osterix control bone cell lineages(1–6). Of these, RUNX2 and Osterix (OSX) have the most selective roles in osteogenesis. Deletion of either factor in mice results in severe defects in skeletal development. Runx2-null mice have a cartilaginous skeleton that contains no bone mineral or hypertrophic cartilage. In contrast, Osx, which is genetically downstream of Runx2, has a more selective role in osteoblast formation and skeletal mineralization(3–5). In addition to its essential role in bone development, Runx2 is necessary throughout life to promote the differentiation of new osteoblasts during bone remodeling(2).
Consistent with its fundamental role in bone formation, RUNX2 is tightly regulated. In addition to transcriptional control by factors such as bone morphogenetic proteins(7), RUNX2 activity is regulated both by its interaction with a number of accessory nuclear factors and by post-translational modifications, including phosphorylation. We have been particularly interested in this latter control mechanism and showed that ERK1/2 MAPK-dependent phosphorylation of RUNX2 is critical for osteoblast-specific gene expression and differentiation(8,9). This pathway mediates the response of bone cells to a variety of signals including hormone/growth factor stimulation(10,11), extracellular matrix binding/matrix tension (12–15) and mechanical loading(16,17). ERK1/2 phosphorylates four serine residues on RUNX2 (S43, S301, S319, S510 using the amino acid residue numbering for murine Type II RUNX2 isoform having N-terminal sequence MASN)(18). Of these, S301 and S319 are required for transcriptional activity since S to A mutations at these sites greatly reduces the ability of RUNX2 to stimulate osteoblast gene expression. ERK1 and ERK2 directly bind to RUNX2 via a consensus MAPK docking or “D” site in its C-terminal region distal to the runt domain. This RUNX2-ERK interaction also occurs on the chromatin of target genes in vivo and is necessary for activation of RUNX2 by the ERK/MAPK pathway(18,19).
Further evidence for the crucial role of ERK/MAPK signaling in osteogenesis is provided by transgenic mouse studies. Specifically, over-expression of constitutively active or dominant-negative forms of the MAPK intermediate, MEK1, in osteoblasts respectively stimulated or inhibited bone development as well as RUNX2 phosphorylation(20). Effects of MAPK on development are at least in part mediated by RUNX2, because the cleidocranial dyplasia phenotype of Runx2 heterozygous null mice can be partially rescued by crossing these animals with mice expressing constitutively active MEK1. In related studies, Prx1-cre mediated inactivation of Erk2 in osteochondroprogenitor cells of developing Erk1-null mice was shown to block osteoblast differentiation leading to ectopic cartilage formation in regions of the perichondrium that normally form bone. Furthermore, increased ERK/MAPK signaling in the same cell population increased osteoblast differentiation and inhibited chondrogenesis(21).
Recent evidence also supports a role for the p38 MAPK pathway in osteoblast differentiation as part of downstream signaling from the TGF-β/BMP-responsive kinase, TAK1(22). Specifically, conditional deletion of TAK1 in osteoblasts using Osx-Cre results in a severe skeletal phenotype that includes low cortical and trabecular bone mass, clavicular hypoplasia and delayed fontanelle fusion. Although TAK1 stimulates ERK, JNK and p38 MAP kinases(23), its actions in osteoblasts were attributed to activation of p38 and subsequent RUNX2 phosphorylation. Consistent with this model, mice harboring deletions in the p38 MAPK intermediates, Mkk3, Mkk6, p38α or p38β, all had decreased bone mass(22). Three p38 phosphorylation sites on RUNX2 were identified (S31, S254 and S319) and their combined mutation was shown to reduce RUNX2 transcriptional activity.
Interestingly, one of the p38 phosphorylation sites on RUNX2 (S319) had previously been identified as a substrate for ERK1/2(18). This finding raises the intriguing possibility that ERK and p38 MAPKs have overlapping functional roles in the control of osteoblast gene expression and differentiation. In the present study, we explore this concept as well as compare the relative importance of these two MAPKs in directly controlling RUNX2 phosphorylation and transcriptional activity.
The reagents used in this study were obtained from the following sources: tissue culture medium and fetal bovine serum from Hyclone and Invitrogen; recombinant BMP2/7 from R&D; ITS, U0126 and SB203580 from Sigma; phospho-ERK, total-ERK, phospho-p38, total p38 and total-SMAD1/5/8 antibodies from Cell Signaling; phospho-SMAD1/5/8 antibody from Santa Cruz and RUNX2 antibody from MBL.
Convance (Princeton, New Jersey) using the following peptide as immunogen produced an antibody that specifically detects RUNX2 phosphorylated at S319: YPSYLSQIMTS(P)PSIHSTTPL. Peptide was conjugated to Keyhole limpet hemocyanin (KLH) and injected into rabbits at 4-week intervals for a total of three injections. After 12 weeks, the serum was harvested and affinity purified using a column containing non-phosphorylated peptide to remove antibody against total RUNX2 followed by a second affinity step using a phosphopeptide-containing column.
The 6OSE2-luc RUNX2 reporter and pCMV-RUNX2 expression vector were previously described(24,25). Constitutively active MEK1 (Meksp) and dominant negative MEK1 (MEKDN and kinase dead ERK1 (ERK1DN) expression vectors were obtained from Dr Kun Liang Guan (Univ of California San Diego)(26,27). Constitutively active MKK6, dominant negative MKK6, and dominant negative p38 expression vectors were described previously(28). Plasmids expressing constitutively active MEK1, wild type RUNX2, S301A,S319A mutant RUNX2 and RUNX2 containing a deletion of the MAPK-binding D site (Δ200-215) were developed in the project laboratory (18,29).
COS7 cells were obtained from the American Type Culture Collection and maintained in DMEM with 10% FBS and 1% antibiotics. MC3T3-E1 clone 42 cells (MC42 cells), which contain stably integrated copies of 1.3kb murine mOG2 promoter driving luciferase, were previous developed in this laboratory(30). To induce differentiation, cells were plated at a density of 5 × 104 cells/cm2 and cultured in α-MEM, 10%FBS containing 50μg/ml ascorbic acid (AA) for up to 14 days. Primary calvarial cells were isolated from newborn transgenic mice previously developed in this laboratory that use a 0.6 kb mOG2 promoter to drive dominant negative (TgMekdn) or constitutively active (TgMeksp) MEK1 in osteoblasts(20) and differentiation/mineralization was induced by growth in α-MEM, 10%FBS, 50μg/ml AA and 3 mM inorganic phosphate for 14 d. For calvarial bone organ cultures, calvaria were isolated from E18.5 C57BL6 mice and cultured in α-MEM with 1% ITS, 50μg/ml AA and 3 mM inorganic phosphate for up to 6 days. Medium was changed daily. Mineral was visualized using the either Alizarin Red or von Kossa staining as indicated. All studies with mice were performed in compliance with the University of Michigan Committee for the Use and Care of Animals.
COS7 cells were plate at a density of 5 × 104 cells/cm2 and transfected with the indicated plasmids using Lipofectamine (Invitrogen). For each dish, 0.05 μg of pRL-SV40 containing a cDNA for renilla reformis luciferase was used as a control for transfection efficiency. Cells were harvested 48 hours after transfection and assayed using a dual luciferase kit (Promega) with a Monolight 2010 luminometer (Pharmingen).
MC3T3-E1 clone 42 and transfected COS7 cells were grown on glass cover slips, fixed with 4% formaldehyde, and incubated overnight at 4°C with primary antibodies to total and phosphorylated RUNX2 (1:500 dilution). Alexa Fluor 488 conjugated donkey anti-mouse and Alexa Fluor 555 conjugated donkey anti-rabbit (Invitrogen) antibodies were used to visualize the localization of total and S319-P RUNX2. Fluorescence detection was accomplished using Nikon Eclipse 50i microscope and imaging system.
For Western blots, cell extracts were prepared by directly dissolving cells in 2 × SDS sample buffer (invitrogen). For immunoprecipitations, cells were harvested in IP buffer containing 20 mM Tris, pH7.6, 500 mM NaCl, 0.25% NP40, 5mM NaF, 1mM EGTA, 1% proteinase inhibitor and 1% phosphatase inhibitor (Sigma). The indicated antibodies were added and incubated overnight at 4°C with gentle rocking. Immunocomplexes were then collected by protein A/G agarose beads and released by boiling with SDS sample buffer. Samples were fractionated by SDS-PAGE using 6 to 20% gradient gels (Invitrogen). After transfer to nitrocellulose membranes, samples were incubated with the indicated antibody (1:500 dilution) overnight at 4°C. Sheep anti-mouse or donkey anti-rabbit second antibody conjugated with horseradish peroxidase (GE Healthcare, UK) was used at a 1:10000 dilution. Horseradish peroxidase activity was detected by ECL (Amersham).
All statistical analyses were performed using SPSS 16.0 Software. Unless indicated otherwise, each reported value is the mean ± S.D. of triplicate independent samples. Statistical significance was assessed using a one-way analysis of variance.
The ERK/MAPK pathway plays a important role in the control of osteoblast gene expression and differentiation by stimulating RUNX2 phosphorylation(8,14,18,20). S301 and S319 are of particular importance in that their mutation to alanine greatly reduces the ability of RUNX2 to stimulate osteoblast gene expression. P38 MAPK can also phosphorylate RUNX2 in response to TGF-β/BMP activation of TAK1(22). Interestingly, one of the ERK1/2 sites we described, S319, is also phosphorylated by p38α. Because it is a substrate for both kinases, we considered it likely that S319-P would be indicative of overall RUNX2 activation state and a good candidate for antibody production. The peptide containing S319-PO4, which was used to immunize rabbits, is conserved across mouse, rat and human species. A site homologous to S319 is also present in RUNX1 (AML1), although the flanking sequence is different from RUNX2 (Fig. 1A). A polyclonal antibody was generated to the RUNX2 phosphopeptide and affinity purified using positive and negative selection (see Methods). As shown in Figure 1B, the P-RUNX2 antibody selectively detects wild type RUNX2, but not RUNX2 containing S301A,S319A mutations. Furthermore, staining intensity increased when COS7 cells were transfected with a constitutively active form of MEK1 (Meksp). In contrast, total RUNX2, measured with a commercially available antibody, remained unchanged under these conditions. The phospho-RUNX2 antibody did not cross-react with phosphorylated RUNX1 (result not shown).
The S-319-P antibody also selectively detects phospho-Runx2 in intact cells when examined by immunofluorescence microscopy (Fig. 1C). Antibody clearly stains phosphorylated RUNX2 in COS7 cells transfected with wild type RUNX2, but gave no signal with the S301A,S319A mutant. Staining for both wild type and mutant RUNX2 revealed a predominantly nuclear localization. In summary, the RUNX2-S319-P antibody is able to discriminate between RUNX2 in different phosphorylation states and should be a powerful tool for monitoring activation of this transcription factor.
To compare the ability of ERK1/2 and p38 to stimulate RUNX2 phosphorylation and transcriptional activity, COS7 cells were transfected with a RUNX2 expression vector and increasing amounts (0.05–1.0 μg plasmid DNA) of constitutively active forms of MEK1 (Meksp; activates ERK1/2) or MKK6 (Mkk6sp; activates p38). RUNX2-dependent transcriptional activity was measured using a 6OSE2-luc-reporter gene (contains 6 copies of the RUNX2-specific enhancer, OSE2, driving luciferase; Fig 2-top) and changes in RUNX2 phosphorylation and MAP kinase activation were measured on Western blots (Fig 2-bottom). Meksp or Mkk6sp each dose-dependently increased P-ERK or P-p38, respectively, and stimulated RUNX2-S319 phosphorylation and transcriptional activity. Maximum stimulation by each MAP kinase kinase was achieved with 0.5 μg plasmid DNA, with no further increases noted at higher concentrations. This indicates that Meksp or Mkk6sp plasmids were in excess relative to their respective MAPK substrate (ERK1/2 or p38) at or above this level. However, large differences were observed in the magnitude of responses. Saturating amounts of Meksp stimulated transcriptional activity approximately 12-fold and strongly induced RUNX2 phosphorylation while Mkk6sp only stimulated transcription 2.9-fold and induced a barely detectable increase in RUNX2-P. Similarly, both kinases increased levels of nuclear RUNX2-S319-P in osteoblasts as measured by immunofluorescence, but the amount of the stimulation was much greater in the Meksp-treated group.
To exclude the possibility that endogenous levels of ERK and p38 were limiting the magnitude of Meksp and Mkk6sp stimulation in the experiment shown in Figure 2, a second study was carried out where cells were transfected with increasing amounts of ERK1, p38α or p38β in the presence of excess Meksp or Mkk6sp plasmids, respectively (Fig 3AB). The amount of Meksp or Mkk6sp expression plasmid used (0.5 μg DNA each) was based on results in Figure 2 that showed maximal stimulation of ERK or p38 phosphorylation at or above this amount of plasmid. Studies focused on p38α and β because these isoforms were previously reported to phosphorylate RUNX2(22). In the absence of upstream activation, ERK1 or p38α/β over expression minimally affected RUNX2-dependent transcriptional activity (Fig 3A). However, a clear dose-dependence was seen for all three MAPKs when cells were also transfected with excess Meksp or Mkk6sp. Peak activation of RUNX2 phosphorylation and transcriptional activity was seen with 0.5 μg of each MAPK expression vector with a partial drop seen with 0.75 μg. However, large differences were seen in the magnitude of 6OSE2-luc reporter activation with ERK1 having the highest activity (Meksp/control = 16) followed by p38β (Mkk6sp/control = 6.9) and p38α (MKK6/control = 4.3). Large differences in peak luciferase activities were also seen. Normalized luciferase activities were as follows: ERK1, 6.0; p38β, 1.5; p38α, 0.65 (Ratios ERK1:p38β:p38α = 9.1:2.3:1.0). Similar differences were seen in the ability of each MAPK to stimulate RUNX2 phosphorylation with RUNX2 being more extensively phosphorylated by the ERK1/Meksp combination than by p38β/Mkk6sp or p38α/Mkk6sp (Fig 3B).
To better compare ERK1/Meksp and p38β/Mkk6sp stimulation of RUNX2 activity, COS7 cells were transfected with optimal amounts of each expression vector in the combinations indicated. Results are presented both in terms of absolute activity and fold-stimulation relative to control cells transfected with 6OSE2-luc and RUNX2 only (Fig 3C). Due to the increased availability of endogenous ERK1 in COS7 cells, over expression of Meksp alone induced a 12-fold stimulation of luciferase activity. This was further increased to 16-fold with ERK1 transfection. In contrast, MKK6 increased activity 2.9-fold and this was further increased to only 6.9-fold with p38β. Surprisingly, when optimal amounts of ERK1/Meksp and p38β/Mkk6sp were combined, neither an additive nor synergistic stimulation of transcriptional activity or RUNX2 phosphorylation was observed. Instead, RUNX2 phosphorylation and activity were actually lower than the ERK1/Meksp group (P<0.01).
Competition between ERK1 and p38β for a common docking site on RUNX2 is a possible explanation for this result. Our previous work showed that RUNX2 contains a conserved ERK1/2 docking motif or “D” site between amino acid residues 200 and 215(18). Since this motif is also recognized by JNK/SAPKs and p38 MAPKs when present in other transcription factors such as SAP-1/2, ATF-2 or ELK-1(31,32), it is possible that it could also serve as binding site for p38. In fact, p38 was previously shown to bind RUNX2 in co-immunoprecipitation assays although the actual binding region was not defined(22). In the study shown in Figure 4A, co-immunoprecipitation assays were used to demonstrate competition between P-ERK1/2 and P-p38 for binding sites on RUNX2 under conditions where either ERK or p38 was activated by increasing amounts of Meksp or Mkk6sp. For all groups, COS7 cells were transfected with FLAG-tagged RUNX2 and optimal amounts of either Mkk6sp or Meksp. M2 antibody (anti-FLAG) was then used to immunoprecipitate total RUNX2 and associated P-ERK or P-p38. As cells were transfected with increasing amounts of Meksp in the presence of optimal Mkk6sp, a concomitant increase in RUNX2-associated P-ERK1/2 was observed while the amount of bound P-p38 decreased. Similarly, when cells were transfected with increasing Mkk6sp, bound P-p38 increased at the expense of P-ERK.
To determine whether competition between ERK and p38 affected RUNX2 transcriptional activity and phosphorylation, COS7 cells were transfected with optimal amounts (based on results shown in Fig. 3) of either Meksp/ERK1 (Fig 4BD) or Mkk6sp/p38β (Fig 4CE) expression plasmids and increasing amounts of dominant-negative ERK1 or p38. Cells were then assayed for RUNX2-dependent transcriptional activity (Fig 4BC) and levels of RUNX2-S319-P, total ERK and total p38 protein (Western blots, Fig 4DE). Since they were epitope-tagged, transfected ERK1, ERK1(DN), p38β and p38β(DN) could be distinguished from endogenous kinases by their reduced electrophoretic mobility (see arrows, Panels DE, bottom 2 rows). For example, in the row probed for total ERK in Panel D, the upper band in lane 1 reflects levels of wild type ERK1 transfected into cells. Since transfected wild type and mutant ERK have the same mobility, the increased intensity of this band in the lanes 2–5 reflects increasing amounts of transfected ERK(DN). Similarly, the upper band in lanes 6–10 of the p38 row measures p38(DN). The same analysis applies to the results shown in Panel E, except in this case all cells were transfected with p38β instead of ERK1, so the ERK(DN) band is clearly visible (lanes 1–5) while p38(DN) is only distinguished from transfected wild type p38β by the increased intensity of the total p38 band in samples transfected with increasing amounts p38(DN) (lanes 6–10). As expected, ERK1(DN) strongly inhibited ERK1/Meksp-stimulated RUNX2 activity and phosphorylation. Under this condition, p38(DN) was only weakly inhibitory. On the other hand, when Mkk6sp/p38β activity was examined, ERK1(DN) and p38(DN) were both strongly inhibitory although ERK1(DN) was still somewhat more active. These are the results that would be expected if ERK1 had a higher affinity for the RUNX2 D site than p38.
Similarly, in the experiment shown in panel F, COS7 cells were transfected with increasing amounts of dominant-negative MKK6 (MKK6dn). This strongly inhibited basal P-p38 levels while dramatically increasing P-ERK, RUNX2 phosphorylation and transcriptional activity. Again, this would be the result expected if suppression of the less active p38 were allowing more RUNX2 docking sites to be available for P-ERK, leading to increased RUNX2 phosphorylation and transcriptional activity.
A final experiment in this section examined the requirement for an intact RUNX2 D site for ERK and p38-mediated RUNX2 phosphorylation/transcriptional activity (Fig 4G). COS7 cells were transfected with wild type RUNX2 or RUNX2 containing a D site deletion (Δ200-215) and stimulated with Meksp, Mkk6sp or both kinases. With wild type RUNX2, Meksp preferentially stimulated transcriptional activity and RUNX2-S319-P versus Mkk6sp, and, as expected, the Meksp/Mkk6sp combination was less active than Meksp alone. In contrast, neither kinase had any activity with the D site deletion mutant.
Pharmacological inhibitors were used to further explore the role of ERK and p38 MAPKs in bone cells. Studies shown in Figure 5 examined effects of inhibitors on transfected COS7 cells, MC3T3-E1 clone 42 preosteoblast cells and calvarial bone organ cultures. In all cases, cultures were titrated with the indicated inhibitor to give maximal suppression of p-ERK1/2 or P-p38, respectively (result not shown). Inhibitors were added 16 h before harvest to minimize the nonspecific toxic effects associated with prolonged exposures(14). In COS7 cells, the ERK/MAPK inhibitor, U0126, blocked RUNX2 activity (P<0.01) and inhibited ERK and RUNX2 S319 phosphorylation. In contrast, SB203580 (SB), a p38/MAPK inhibitor, while clearly blocking p38 phosphorylation, actually increased P-ERK, RUNX2 transcriptional activity and phosphorylation (panel A). This may explain why there was no decrease in overall RUNX2-dependent transcription under this condition. Addition of both inhibitors decreased RUNX2 activity and phosphorylation below levels seen with U0126 alone, which suggests that p38/MAPK also contributed to basal RUNX2 activity.
To assess the applicability of these findings to bone cells, inhibitor effects were also examined in MC42 preosteoblast cells (Fig 5B) and calvarial organ cultures (Fig 5CD). MC42 cells, which contain stably integrated copies of a 1.3 kb mOG2-luc reporter(30), were grown in AA-containing medium for 6 days to induce differentiation and mOG2-luc activity. In these cells, mOG2-luc activity is proportional to endogenous osteoblast gene expression and differentiation(30). Cells were then treated with U0126 and/or SB203580 for 16 hours immediately before harvest. As previously reported(14), U0126 dramatically inhibited mOG2-luc activity. This inhibitor also blocked ERK and endogenous RUNX2 phosphorylation while increasing P-p38. In contrast, SB actually mildly stimulated ERK and RUNX2 phosphorylation and transcription. However, when it was combined with U0126, SB further decreased mOG2-luc activity and RUNX2 phosphorylation.
For calvarial organ cultures, bones were isolated from E18.5 mouse embryos and cultured for 5 d with AA and 2 mM inorganic phosphate ± U0126 or/and SB203580 (Fig 5CD). Mineralization was measured using Alizarin red staining. During the 5 d culture period, sutures went from a patent state to nearly total fusion. U0126 or SB203580 each significantly inhibited mineralization although the ERK inhibitor was somewhat more active and an even greater inhibition was observed when both compounds were used. Similar to cell culture results, Western blot analysis of calvarial extracts showed that U0126 reduced P-ERK and RUNX2 S319 phosphorylation while mildly increasing P-p38. In contrast, P-p38 inhibition had no obvious effect on RUNX2 phosphorylation and increased P-ERK while combining U0126 and SB blocked both pathways and reduced RUNX2 phosphorylation below levels seen with U0126 alone.
A final example of this reciprocal regulation is shown in Figure 6. In this study, calvarial osteoblasts were isolated from previously generated transgenic mice in which a 0.6 kb mOG2 promoter was used to drive dominant negative (TgMekdn) or constitutively-active (TgMeksp) MEK1 in osteoblasts (Fig 6A,B). In these animals, Mekdn delayed embryonic bone development while Meksp accelerated osteogenesis(20). Cells were prepared from newborn wild type mice and cultured under differentiating conditions for 14 days. As previously reported, TgMeksp cells exhibited increased mineralization relative to wild type while TgMekdn cells were under mineralized(20). Western blot analysis (panel B) revealed that endogenous ERK and RUNX2-S319 phosphorylation was decreased in TgMekdn cells while P-p38 was increased. In contrast, ERK and RUNX2 phosphorylation were increased and P-p38 was inhibited in TgMeksp cells. U0126 inhibited ERK and RUNX2 phosphorylation in WT and TgMekdn groups, but not in cells from TgMeksp mice that were resistant because of the constitutively active Mek1 transgene. U0126 also further increased the already high levels of phosphorylated p38 in TgMekdn cells. Inhibition of p38 lowered RUNX2 phosphorylation and mildly increased P-ERK in wild type and TgMekdn cells, but led to a dramatic increase in P-ERK, and RUNX2 phosphorylation in TgMeksp cells.
Taken together, these studies show that there is a reciprocal relationship between ERK1/2 and p38 MAPK pathways; inhibition of one leads to activation of the other. Furthermore, due to competition between the two kinases for the RUNX2 docking site, inhibition of p38 actually makes RUNX2 more available for ERK phosphorylation and activation.
Extracellular matrix (ECM) and BMP signals synergistically interact to stimulate osteoblast gene expression and differentiation(14). The basis for this interaction is not known. However, binding of type I collagen to α2β1 integrins or the discoidin receptor type 2 (DDR2) stimulates ERK/MAPK, RUNX2 phosphorylation and gene expression(13,33,34) while BMP activation of type I and II BMP receptors activates receptor SMAD proteins as well as TAK1. Since this kinase also stimulates ERK and p38 MAP kinases(22,23), the potential exists for extracellular matrix and BMP signals to interact at the level of kinase activation of RUNX2. To explore this possibility, we examined MAP kinase activation and RUNX2 S319 phosphorylation in MC42 cells during ascorbic acid (AA)-induced differentiation in the presence or absence of BMP2/7 heterodimer (Fig 7). Previous worked showed that actions of this vitamin on differentiation require collagen synthesis and binding to α2β1 integrins(13,35)
To first evaluate the AA-induced component of kinase signaling, cells were grown in control or AA-containing medium for 3, 7, 10 and 14 days and gene expression (1.3 kb mOG2-luc activity), kinase activation and RUNX2 phosphorylation were measured (Panels A,B). As previously reported, AA strongly induced osteocalcin promoter activity(13,30). Furthermore, time-dependent increases were seen in ERK1/2 and RUNX2 S319 phosphorylation only in cells grown in AA-containing medium. In contrast, little to no change was seen in the level of p38 phosphorylation. This suggests that AA-induced differentiation is mainly associated with ERK/MAPK activation/phosphorylation of RUNX2.
Studies shown in Figure 7CD examined combined effects of AA and BMP2/7 on transcriptional activity, kinase signaling and RUNX2-S319 phosphorylation. As previously reported(14), BMP treatment dose-dependently increased mOG2-luc activity and this response was synergistically stimulated when BMP2/7 was added in the presence of AA (Panel C). In the absence of AA, BMP2/7 also weakly stimulated RUNX2, ERK, p38 and Smad 1/5/8 phosphorylation without affecting total RUNX levels (Panel D). P-RUNX2 and P-ERK levels were reduced by U0126 while p38 inhibition did not affect promoter activity or Smad phosphorylation and slightly increased RUNX2 and ERK phosphorylation. Consistent with results shown in Fig 5B, treatment with AA alone increased promoter activity, RUNX2 and ERK phosphorylation, and all these responses were sensitive to ERK, but not p38 inhibition. Significantly, in the presence of AA, BMP2/7 dramatic stimulated mOG2-luc activity, RUNX2, ERK and p38 phosphorylation and modestly increased Smad1/5/8 phosphorylation without affecting total RUNX2. ERK/MAPK inhibition reduced promoter activity and ERK/RUNX2 phosphorylation to basal levels without affecting SMAD phosphorylation. Under these conditions, p38 inhibition also partially inhibited mOG2-luc activity, but, surprisingly, did not affect RUNX2 phosphorylation. Two important conclusions may be drawn from these studies: First, the BMP pathway can stimulate RUNX2 S319 phosphorylation and transcriptional activity, particularly in the presence of AA, and this stimulation is not associated with increased levels of RUNX2. Second, although BMPs via TAK1 are know to activate all 3 MAPKs, ERK1/2 is a major mediator of this response in cultured osteoblasts with p38 having a minor role via a mechanism unrelated to RUNX2 S319 phosphorylation.
In this study, a variety of approaches were used to examine ERK1/2 and p38 signaling in osteoblasts and relate kinase activities to phosphorylation and activation of the RUNX2 transcription factor. Although RUNX2 was a substrate for both kinases, it was preferentially phosphorylated/activated by ERK1 relative to p38α or β. Additional novel findings include the discovery that both ERK and p38 MAPKs compete for a common docking site on RUNX2, the observation that these two kinase classes are reciprocally regulated and the demonstration that RUNX2-S319 phosphorylation is strongly stimulated by BMP treatment, particularly in the presence of ascorbic acid. Lastly, ERK and p38 were shown to have separate, but overlapping roles in mediating the response of bone cells to AA and BMP signals.
Transgenic studies clearly established the requirement for both ERK1/2 and p38 in osteoblast differentiation and bone development(18,22). Furthermore, both pathways were shown to regulate osteoblast gene expression, at least in part, by controlling the level of RUNX2 phosphorylation at specific serine residues. However, the relative ability of ERK and p38 to phosphorylate/activate RUNX2 and potential interactions between these two kinases in osteoblasts had not previously been examined. To address this issue, we developed a unique antibody to monitor RUNX2 phosphorylation at serine 319, a site phosphorylated by both ERK1/2 and p38α(18,22). Under a variety of conditions, a good correlation was observed between the level of RUNX2 S319 phosphorylation measured on Western blots and RUNX2 transcriptional activity measured using either a RUNX2-specific reporter (6OSE2-luc) or a 1.3 kb murine osteocalcin gene 2 promoter-luciferase construct (1.3mOG2-luc). In spite of the fact that ERK1/2 and p38 can each phosphorylate RUNX2 at other sites, the level of S319 phosphorylation appears to be a good indicator of RUNX2 transcriptional activity (Figs.5–7). For this reason, this antibody should be a useful reagent for monitoring RUNX2 activation under different physiological conditions.
Although ERK1/2, p38α and p38β can each phosphorylate and activate RUNX2, large differences were noted when kinases were compared. When assayed under conditions where each MAPK was optimally stimulated by its respective MAPKK (constitutively-active MEK1 or MKK6), ERK1 was found to be four times more active than p38β and nine times more active than p38α in stimulating RUNX2-dependent transcription (Fig 3). As expected, differences in transcriptional activation were paralleled by differences in the magnitude of RUNX2-S319 phosphorylation. The observation that ERK1/2 and p38 can both phosphorylate RUNX2 at the same site is consistent with findings in other systems. For example, T581 of MSK1 is phosphorylated by ERK and p38 in response to oxidative stress(36,37) while both kinases phosphorylate the EGF receptor at T669, S1046 and S1047(38).
Phosphorylation/activation by either class of MAPKs required a consensus MAPK docking D site in RUNX2 at residues 200–215 (RGKSTLTITVFTNPP; consensus sequence Rxxxxθxθ where θ is L/I/V(39)). There are several examples of other transcription factors with D sites that bind more than one MAPK. These include Elk-1 (ERK and JNK), ATF-2 (p38 and JNK) and SAP-1/SAP-2 (ERK and p38)(31). The D site in RUNX2 has properties similar to those of other MAPK-responsive transcription factors in that it is immediately N-terminal to a transcription activation domain that contains MAPK phosphoacceptor sites necessary for transcriptional activation(31) (ERK sites at S301, S319, S510(18) and p38 sites at S254, S319(22)). Interestingly, ERK and p38 were shown to compete for binding to the RUNX2 D site (Fig 4). This competition was demonstrated in two ways: i) In co-immunoprecipitation experiments, increasing levels of P-ERK were able to disrupt RUNX2:P-p38 complexes and vice versa. ii) Expression of increasing amounts of DN ERK1 strongly inhibited ERK and p38-dependent RUNX2-S319 phosphorylation and transcriptional activity while DN p38β strongly inhibited p38 and partially inhibited ERK1. Results from both these studies support the conclusion that ERK1 has a relatively higher affinity for the D site when compared with p38. One consequence of this competition is that when constitutively-active MKK6/p38β was expressed in the presence of MEK1/ERK1, RUNX2 phosphorylation and transcriptional activity was actually lower than that observed with constitutively-active MEK1/ERK1 alone, a result likely due to displacement of ERK by the less active p38 (Fig 3C). Also, over-expression of DN MKK6, which blocked p38 activation, actually increased P-ERK, RUNX2-S-319-P and transcriptional activity, again consistent with the idea that endogenous p38 was initially competing with ERK on RUNX2 (Fig 4F).
An additional important finding concerns the relationship between ERK and p38 activities. Regardless of the experimental system, we consistently found that suppression of one MAPK was associated with activation of the other. This was true regardless of whether we used pharmacological inhibitors to block ERK versus p38 signaling (Fig 5) or transgenic modification of MEK1 activity (Fig 6) and was seen in transfected COS7 cells, MC3T3-E1 preosteoblasts, calvarial organ cultures or primary osteoblasts from transgenic mice. This reciprocal regulation had some unexpected consequences. While the ERK1/2 inhibitor, U0126, consistently blocked RUNX2 phosphorylation, transcriptional activity and calvarial mineralization (Fig 5), p38 inhibition with SB203580 either had no effect or actually increased RUNX2 phosphorylation and transcriptional activity. Since p38 stimulated RUNX2 activity in the over-expression studies shown in Figures 2–4, one would have predicted its inhibition to reduce activity. This paradoxical result may be a consequence of the stimulation in ERK/MAPK activity seen after SB203580 treatment combined with a reduction of p38 at the RUNX2 D site thereby allowing increased docking and phosphorylation by ERK1/2. On-the-other-hand, when cells/organ cultures were treated with both ERK and p38 inhibitors, RUNX2 phosphorylation, transcriptional activity and mineralization (organ cultures only) were reduced to levels below those observed with the ERK inhibitor alone. This suggests that p38 is responsible for the residual RUNX2 activity seen in U0126-treated samples, but does not have a major role in controlling RUNX2 activity under conditions where there is a strong ERK/MAPK signal. Reciprocal regulation of ERK and p38 MAPKs was previously reported in other systems such as osteoclasts where inhibition of one MAPK increases the other(40) and in osteoblasts from subchondral bone that exhibit increased ERK and reduced p38 activity(41).
Although in vivo transgenic studies clearly showed that ERK and p38 MAPK pathways are both necessary for bone formation(20–22), based on our studies, it is unlikely that these two kinases function in a redundant manner to control RUNX2 activity and osteoblast differentiation. Some insight into their separate functions may be gleaned from the studies shown in Figure 7 that compared AA and BMP-induced osteoblast activities. Standard conditions for inducing osteoblast differentiation involve providing an environment conducive to collagen matrix secretion by inclusion of AA in tissue culture media. Numerous studies showed that the resulting ECM induces osteoblast differentiation, at least in part, by interacting with cells via collagen-binding integrins (α2β1 and α1β1) and the discoidin domain receptor 2 (DDR2), leading to activation of ERK/MAPK signaling(13,33). Although AA may also regulate gene expression via activation antioxidant-responsive elements in target genes(42), its long-term effects on osteoblast differentiation are most likely mediated by the ECM/integrin/MAPK pathway described above. The ability of AA to dramatically stimulate ERK/RUNX2 phosphorylation and osteoblast gene expression was shown in Fig 7B. Under these conditions, P-p38 levels remained essentially unchanged. AA-induced RUNX2 S319 phosphorylation and gene expression were highly sensitive to the ERK/MAPK inhibitor, but were refractory to p38 inhibition (Figs 5,,7D).7D). A different situation was encountered when the combined effects of AA and BMP treatment were examined. As previously reported(14,43), induction of osteoblast gene expression/differentiation by BMPs is synergistically stimulated by AA (Fig 7CD). The mild stimulation of gene expression by BMPs in the absence of AA was sensitive to ERK, but not p38 inhibition. In contrast, in the presence of AA, BMP dramatically stimulated P-ERK and RUNX2-S319 phosphorylation while synergistically stimulating mOG2-luc activity. This response was still completely blocked by ERK inhibition and partially reduced with the p38 inhibitor. However, while ERK inhibition completely blocked RUNX2 phosphorylation, p38 inhibition was without effect.
Several important conclusions may be drawn from these studies: First, the BMP pathway can stimulate RUNX2 S319 phosphorylation and transcriptional activity, particularly in the presence of AA, and this stimulation is not associated with increased levels of RUNX2. This provides an alternate route for BMP regulation of RUNX2 in addition to the previously described induction of Runx2 expression(44). Second, although BMPs via TAK1 are known to activate all 3 MAPKs, ERK1/2 is a major mediator of this response in osteoblasts with p38 having a minor role via a mechanism unrelated to RUNX2 S319 phosphorylation. Finally, the synergistic stimulation in gene expression observed with BMP and AA treatment was also partly dependent on p38. This p38 component distinguishes the AA/BMP response from the response in cells treated only with AA that were resistant to p38 inhibition. However, blocking p38 in this case did not noticeably affect RUNX2-S319-P. This suggests either that additional RUNX2 phosphorylation sites are involved or that p38 regulates osteoblast gene expression by controlling the activity of other transcription factors. In this regard, p38 is known to phosphorylate/activate OSX or DLX5 in response to BMP treatment(45,46).
This work was supported by NIH Grants DE11723 and DE12286 (to RTF), DE018290 (to KLK) and Core Facilities of the Michigan Diabetes Research and Training Center funded by NIDDK Grant DK020572.
All authors state that they have no conflicts of interest.
Authors’ roles: Study design: CG and RTF. Study conduct: CG, QY, GZ, HY. Data interpretation: CG, RTF, KLK. Drafting manuscript: CG, RTF. All authors approved final version of manuscript. CG and RTF take responsibility for the integrity of the data analysis.
University at Buffalo, South Campus
Buffalo, NY 14215