Loss of Smad3-Mediated Negative Regulation of Runx2 Activity Leads to an Alteration in Cell Fate Determination

Abstract

Runx2 is required for osteoblast differentiation but is expressed in certain nonosteoblastic cells without activating the differentiation process, suggesting that its activity is suppressed through a lineage-specific mechanism. Here we report that primary mouse dermal fibroblasts lacking Smad3 can acquire an osteoblast-like phenotype, including activation of Runx2 activity, expression of osteoblast-specific genes, and calcium deposition. We further show that negative regulation of Runx2 activity by Smad3 in dermal fibroblasts is likely mediated by controlling the expression of Msx2, an antagonist of Runx2 in this cellular context. These data support the presence of a novel mechanism for controlling cell fate determination of mesenchymal lineages by preventing differentiation toward the osteoblastic lineage via negative regulation of Runx2 activity.

Runx2/Cbfa1/Aml3, a member of the core binding factors α/acute myelogenous leukemia family, is a transcription factor required for osteoblast differentiation (9, 12, 15) and function (8). Runx2-deficient mice fail to develop an ossified skeleton due to the lack of osteoblasts, demonstrating the essential role of Runx2 in the determination of an osteoblastic cell fate from mesenchymal stem/progenitor cells (MSCs) (12, 15). Furthermore, overexpression of Runx2 in mouse skin fibroblasts leads to the induced expression of osteoblastic markers such as osteocalcin and bone sialoprotein (9). However, endogenous Runx2 expression was also found in the dermis (12), suggesting that the activity of Runx2 is likely regulated by unknown mechanisms in addition to the control of its expression in a tissue-specific manner.

The multipotent cytokine transforming growth factor β (TGF-β) regulates the differentiation of a variety of mesenchymal cells, including adipocytes (5), myoblasts (13), and osteoblasts (3). TGF-β potently inhibits osteoblast differentiation, as indicated by the reductions in the expression of osteoblast differentiation markers, including alkaline phosphatase activity, osteocalcin expression, and calcium nodule formation (3, 10). The mechanism through which TGF-β inhibits osteoblast differentiation and affects mesenchymal differentiation in general has been investigated through the discovery and characterization of TGF-β signaling effectors, the Smads. In the TGF-β signal transduction cascade, extracellular TGF-β ligand binds to a serine/threonine receptor complex, which results in the phosphorylation of serines in the C terminus of Smad2 and Smad3. These phosphorylated receptor-Smads form a heteromeric complex with the co-Smad Smad4 and translocate to the nucleus to regulate the transcription of target genes. Smads can activate or repress transcription through direct binding to DNA or through association with other transcription factors (16).

In the cellular contexts of calvarial osteoblasts and various established cell lines, Smad3 was found to inhibit the expression of Runx2 and block its transcriptional activity through direct interaction between the two proteins (1). However, it is unclear whether the same mechanism is responsible for the postulated negative regulation of Runx2 activity as a part of the program underlying cell fate determination associated with the differentiation process of MSCs. Here, we use primary mouse dermal fibroblasts as a model system to examine the mechanism through which the activity of Runx2 may be regulated in nonosteoblastic cells. We show that loss of Smad3 leads to the activation of Runx2 DNA-binding activity, which is accompanied by increases in alkaline phosphatase activity, expression of osteoblast-specific genes, and calcium deposition—a process mimicking osteoblastic differentiation. This ability of Smad3-deficient dermal fibroblasts to acquire an osteoblast-like cell fate suggests that the negative regulation of Runx2 activity by other transcription factors such as Smad3 may be just as important for the control of osteoblastic differentiation from MSCs as the conventional mechanisms which rely on the tissue-specific expression of this important differentiation factor.

MATERIALS AND METHODS

Isolation and culture of primary dermal fibroblasts and osteoblasts.

Mice were maintained and used in accordance with Institutional Animal Care and Use Committee protocols. Pairs of mice heterozygous for targeted deletion of Smad3 in a mixed 129-C57B6 background were mated to produce the litters used in this study. For dermal fibroblasts, dissected skins were incubated in 0.25% trypsin at 4°C overnight. Mice were genotyped as previously described (6). The dermis from each wild-type and Smad3 null littermate was minced in Dulbecco modified Eagle medium with 10% heat-inactivated fetal bovine serum, nonessential amino acids, and penicillin-streptomycin, and cells that attached overnight were cultured as dermal fibroblasts. For the preparation of osteoblasts, calvaria were used for the isolation of osteoblasts as previously indicated (3). For differentiation experiments, dermal fibroblasts or osteoblast controls were plated in minimal essential medium containing 10% fetal bovine serum, nonessential amino acids, and penicillin-streptomycin. Twenty-four hours after plating, osteogenic differentiation supplementary medium (10 mM β-glycerophosphate, 50 μg/ml ascorbic acid) was added and that point was designated day zero.

Western blotting.

For nuclear lysates, confluent cells were washed in phosphate-buffered saline and then gently harvested in 10 mM HEPES (pH 8.0)- 1.5 mM MgCl₂-10 mM KCl-300 mM sucrose-0.1% NP-40 with freshly added 0.5 mM dithiothreitol (DTT), 25 mM NaF, 25 mM β-glycerophosphate, 1 mM Na₃VO₄, 1 mM phenylmethylsulfonyl fluoride (PMSF), and protease inhibitors. The nuclear pellet was rinsed and then incubated for 30 min on ice with 50 mM HEPES (pH 7.9)-250 mM KCl-0.1 mM EDTA-0.1 mM EGTA-0.1% NP- 40-0.1% glycerol with freshly added 1 mM DTT, 25 mM NaF, 25 mM β-glycerophosphate, 1 mM Na₃VO₄, 1 mM PMSF, and protease inhibitors. Cells were centrifuged at 4°C at maximum speed for 10 min to isolate the supernatant. Equivalent amounts of nuclear extract were separated by sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis. Western analysis was performed with anti-Runx2 or anti-Msx2 (Santa Cruz Biotechnology, Santa Cruz, CA).

Electrophoretic mobility shift assay (EMSA).

Nuclear extracts prepared from dermal fibroblasts or osteoblasts of the indicated genotype were incubated with a ³²P-radiolabeled probe containing Runx2 DNA-binding elements from the osteocalcin or bone sialoprotein (BSP) promoter (9) and utilized for EMSA as previously described (21). For in vitro binding assays, the indicated proteins were cloned into pGEX-KG to produce glutathione S-transferase (GST) fusion constructs and the resulting GST fusion proteins were then purified from BL21 bacteria with GST-Sepharose.

Luciferase assay.

Cells were plated at 150,000/well in six-well plates, allowed to recover overnight, and transfected with the indicated constructs by using Lipofectamine according to the manufacturer's instructions (Invitrogen, Carlsbad, CA). Harvesting for the luciferase assay was done 48 h after transfection as previously described (22).

Alkaline phosphatase staining.

Cells were plated at 150,000/well in six-well plates and grown in osteogenic differentiation medium until the indicated time, when cells were stained according to the manufacturer's instructions (Sigma Diagnostics, St. Louis, MO).

Alizarin red staining.

Cell cultures were plated at 150,000/well in six-well plates and grown in osteogenic differentiation medium until the indicated time, when cells were fixed in a solution of 10% formaldehyde, methanol, and distilled water at 4°C for 20 min. Cells were stained in a 2% solution of alizarin red (Sigma Chemical Co., St. Louis, MO) for 30 min at room temperature and washed five times with distilled water.

Reverse transcription (RT)-PCR.

RNA was isolated from cells with the RNeasy Mini Kit (QIAGEN, Valencia, CA) according to the manufacturer's instructions. RNA was treated with Amplification Grade DNase I (Invitrogen, Carlsbad, CA) in accordance with standard protocols to eliminate contaminating genomic DNA. DNase-treated RNA was reverse transcribed with MultiScribe reverse transcriptase, oligo d(T)₁₆ primers, and reagents from the GeneAmp Gold RNA PCR reagent kit (PE Biosystems, Foster City, CA). The resulting cDNA was subsequently used for PCR.

Plasmid construction and virus production.

Mouse Msx2 cDNA was generated by RT-PCR, cloned into pcDNA3 (Invitrogen, Carlsbad, CA), and sequenced to confirm authenticity. Mutations corresponding to those present in human bone-related diseases were generated with the GeneTailor site-directed mutagenesis system (Invitrogen, Carlsbad, CA) and then transferred to the pBabe-puro vector (gift of Chris Counter, Duke University, Durham, NC) or the pGEX-KG vector after verification by DNA sequencing. Hemagglutinin (HA)-tagged wild-type Msx2 and mutant Msx2 P148H were generated by transferring the appropriate sequences from the pBabe-puro constructs to a pcDNA3.1 vector with an N-terminal HA tag. To produce retrovirus, 293T cells were transfected with pBabe-bleo and pCl10A with Fugene (Roche, Indianapolis, IN) according to the manufacturer's instructions. Cell supernatants with the viruses were harvested after 24 and 48 h and filtered, and titers were determined before used for infection of the primary cells. The Flag-Runx2 and Flag-Runt constructs were gifts from G. Karsenty, Baylor College of Medicine. For the creation of retroviral constructs containing short hairpin RNA (shRNA) directed against Msx2, sequences were inserted into the pSuperRetro vector targeting the following sequences: Msx2 shRNA1 (GCAGCATCCATATACGGCG), Msx2 shRNA2 (AGTCATGGCTTCTCCGACTAA), and scrambled control (ACT GTGACGTACAGAGCGT).

Coimmunoprecipitation.

Full-length Flag-Runx2 (4 mg) or Flag-Runt (3 mg) was transfected into 293T cells (6-cm plates) with 4 mg of pcDNA3 vector, wild-type HA-Msx2, or P148H mutant HA-Msx2, as indicated, by the calcium phosphate precipitation method. Twenty-four hours after transfection, cells were harvested in ULB+ (50 mM Tris, 150 mM NaCl, 50 mM NaF, 0.5% NP-40, 1 mM PMSF, 1 mM DTT, 1 mM Na₃VO₄, protease inhibitors, pH 7.5) and cell lysates were subjected to anti-HA (Y-11; Santa Cruz) immunoprecipitation at 4°C for 3 h. After four washes with ULB+, the immunocomplex was boiled in SDS sample buffer and resolved by 12% SDS-polyacrylamide gel electrophoresis. The proteins were then transferred to a PVDF membrane (Immobilon) and detected with anti-Flag (M2; Sigma) or anti-HA (F-7; Santa Cruz) antibody.

RESULTS AND DISCUSSION

Expression of Runx2 in dermal fibroblasts is Smad3 independent.

The presence of Runx2 has been closely associated with osteoblast differentiation and mineralization (reviewed in reference 7), although its expression has been found in nonossifying tissues such as the dermis (12). To evaluate whether cultured primary mouse dermal fibroblasts could be used as a model system to explore the mechanism for the negative regulation of Runx2 transcriptional activity, we first determined the pattern of Runx2 expression in these cells. As shown in Fig. 1A and B, wild-type dermal fibroblasts express Runx2 at both the RNA and protein levels. Importantly, the expression levels of Runx2 in the dermal fibroblasts are comparable to those detected in primary osteoblasts isolated from the same animals (Fig. 1A and B).

FIG. 1.

Runx2 is expressed in mouse primary dermal fibroblasts and osteoblasts in a TGF-β/Smad3-independent manner. (A) RT-PCR products generated by Runx2-specific primers from mRNA samples isolated from wild-type (WT) or Smad3-deficient (Null) primary dermal fibroblasts and osteoblasts were analyzed by agarose gel electrophoresis. Lanes 1 to 4 represent samples from dermal fibroblasts derived from four separate littermates, and lanes 5 and 6 represent samples of osteoblasts prepared from the animals whose dermal fibroblast samples are in lanes 1 and 2. The duplicates of all the numbers represent samples derived from two individual animals with identical genotypes in the same litter. GAPDH, glyceraldehyde-3-phosphate dehydrogenase. (B) The expression patterns of Runx2, Smad2, and Smad3 were determined by Western blot (immunoblot [IB]) analysis. The time course of TGF-β treatment is indicated in hours (Hrs). β-Catenin was used as a loading control. (C) Differential response of Runx2 expression to the treatment of TGF-β by two established osteoblastic cell lines.

In the context of osteoblast development, one previously published report indicated that TGF-β represses Runx2 mRNA expression in a Smad3-dependent fashion (1). However, loss of Smad3 or treatment with TGF-β did not have a significant effect on Runx2 expression at the RNA or protein level in primary dermal fibroblasts or osteoblast controls (Fig. 1A and B). In addition, the expression levels of the predominant form of Smad2 were not elevated and its splicing variant that does not contain sequences encoded by exon 3 of Smad2 was undetectable (Fig. 1B), suggesting that there was no compensatory upregulation of Smad2 in Smad3-deficient dermal fibroblasts. To explore this question further, we examined Runx2 protein expression in two established osteoblast-like cell lines in which TGF-β treatment was reported to reduce Runx2 mRNA levels (1). We found that TGF-β treatment reduces Runx2 expression in MC3T3 mouse osteoblast-like cells but not in ROS rat osteosarcoma cells (Fig. 1C). These data suggest that while TGF-β may suppress levels of Runx2 protein in certain cellular contexts, it does not regulate Runx2 expression in primary dermal fibroblasts and osteoblasts.

Smad3 represses Runx2 activity through control of DNA-binding activity.

After finding that Runx2 is expressed in both wild-type and Smad3 null dermal fibroblasts, we determined the functional status of Runx2 in these cells. To evaluate the transcriptional activity of Runx2, we created a reporter construct containing four consensus Runx2 DNA-binding elements, termed osteoblast-specific elements (OSEs), upstream of luciferase (Fig. 2A). As expected, the activity of this reporter construct is dramatically upregulated by the cotransfection of Runx2 into ROS cells (Fig. 2B). We next evaluated the ability of the endogenous Runx2 in dermal fibroblasts to transactivate the OSE-driven promoter. As shown in Fig. 2C, the wild-type cells exhibited a level of basal activity from the reporter construct that is higher than the activity of the control ERE reporter construct, indicating that there is a detectable level of Runx2 activity in wild-type dermal fibroblasts. Importantly, loss of Smad3 led to a significant increase in the level of OSE-driven promoter activity, suggesting that the equivalent amounts of Runx2 in wild-type and Smad3 null dermal fibroblasts display differential levels of activity as measured by OSE reporter activation. These data support the hypothesis that a Smad3-dependent mechanism negatively regulates Runx2 activity in wild-type dermal fibroblasts.

FIG. 2.

Runx2 transcription activity is elevated in Smad3 null dermal fibroblasts due to increased DNA-binding activity. (A) Sequence of the Runx2 DNA-binding elements (OSEs). Four OSE sequences were concatemerized and inserted into the pGL3 basic vector to create the 4×OSE reporter. (B) Coexpression of Runx2 and the 4×OSE reporter in ROS cells demonstrates the responsiveness of the reporter to the transactivation activity of Runx2. (C) The transcriptional activity of the OSE reporter was tested in wild-type (WT) and Smad3-deficient (Null) dermal fibroblasts. ERE represents a similarly designed construct as the 4×OSE except the OSE was replaced with the estrogen response elements. (D) EMSA with nuclear extracts isolated from primary dermal fibroblasts (DF) and osteoblasts (OB). The Smad3 status of the cells and the treatment of TGF-β are indicated. The probe was the OSE derived from the mouse osteocalcin promoter. The specificity of the electrophoretic mobility shift for Runx2 was demonstrated by the inhibition of DNA-binding activity upon the addition of anti-Runx2 antibody or the indicated amount of cold (unlabeled) DNA probe as a binding competitor.

To explore the potential mechanism of increased Runx2 activity in the absence of Smad3, we next evaluated the ability of Runx2 to bind to its DNA target sequence in wild-type and Smad3 null dermal fibroblasts. By using an EMSA with OSEs derived from the promoters of two Runx2 target genes, osteocalcin and BSP (9), we found that the Runx2 present in nuclear extracts from wild-type dermal fibroblasts displayed a minimal ability to interact with the DNA probes (Fig. 2D and data not shown). In contrast, the Runx2 derived from Smad3 null dermal fibroblasts strongly binds to the OSEs, at a level comparable to that found with Runx2 derived from primary osteoblasts (Fig. 2D). Interestingly, TGF-β treatment of the wild-type dermal fibroblasts and osteoblasts did not alter the DNA-binding pattern of Runx2 (Fig. 2D), suggesting that an increase in nuclear Smad3 levels upon TGF-β stimulation does not block the binding of Runx2 to its target sequences. Together, these data suggest that the presence of Smad3 in dermal fibroblasts inhibits Runx2 transcriptional activity by preventing Runx2 from binding to its DNA targets, most likely through an indirect mechanism.

Dermal fibroblasts acquire an osteoblast-like phenotype in the absence of Smad3.

After demonstrating increased Runx2 DNA-binding activity in Smad3-deficient dermal fibroblasts, we hypothesized that this elevated Runx2 activity could have a consequence in defining the cellular phenotype of these dermal fibroblasts by allowing them to acquire the ability to express osteoblastic markers and consequently an altered cell fate under permissive conditions. To test this postulation, the phenotypes of wild-type and Smad3 null dermal fibroblasts cultured in osteogenic differentiation medium were compared to those of osteoblast controls. One early marker for differentiation toward the osteoblastic lineage is alkaline phosphatase, an enzyme that plays an important role in the mineralization process by facilitating calcium phosphate precipitation (4, 11). Consistent with the requirement for alkaline phosphatase in bone development and its use as an osteoblastic marker, a significant amount of alkaline phosphatase activity was detected in wild-type calvarial osteoblasts at day 4 in culture, but this activity was undetectable in wild-type dermal fibroblasts at day 15 under the same culturing conditions (Fig. 3A). In stark contrast to the results obtained with wild-type dermal fibroblasts, Smad3-deficient dermal fibroblasts displayed a significant amount of alkaline phosphatase activity by day 7 in culture (Fig. 3A). By day 15, alkaline phosphatase staining in Smad3 null dermal fibroblasts was similar to that of the wild-type osteoblasts at day 4. The elevation of alkaline phosphatase activity upon loss of Smad3 clearly demonstrates a significant shift in the phenotype of dermal fibroblasts toward the osteoblast lineage.

FIG. 3.

Smad3 null dermal fibroblasts acquire an osteoblast-like phenotype. (A) Wild-type (WT) and Smad3-deficient (Null) dermal fibroblasts were cultured in osteogenic differentiation medium for the indicated times and stained for the expression of alkaline phosphatase. As a positive control, wild-type osteoblasts isolated from mice of the same litter as the wild-type dermal fibroblasts were stained at day 4. (B and C) Expression of osteocalcin and BSP by the Smad3-deficient dermal fibroblasts at culture days 10 and 15, respectively, as demonstrated by RT-PCR. (D) Ability of the Smad3-deficient dermal fibroblasts to deposit calcium as tested by alizarin red staining at the indicated times. The same staining of wild-type osteoblasts at day 14 was used as a positive control. GAPDH, glyceraldehyde-3-phosphate dehydrogenase.

As described above, two additional osteoblast markers, osteocalcin and BSP, have been implicated as direct targets of Runx2 (9). In this regard, we found that Smad3 null dermal fibroblasts expressed osteocalcin mRNA at day 10 (Fig. 3B) and BSP mRNA at day 15 (Fig. 3C), whereas neither osteoblast marker was expressed in wild-type dermal fibroblasts even when they were cultured in the same osteogenic differentiation medium for up to 30 days (data not shown). This induction of expression of additional genes in Smad3 null dermal fibroblasts appears to be restricted to osteoblast-specific genes, as these cells were found not to express type X collagen, a marker for hypertrophic chondrocytes (data not shown). These results clearly demonstrate that the expression of osteoblast-specific transcriptional targets of Runx2 is upregulated in Smad3 null dermal fibroblasts.

Differentiation toward the osteoblastic lineage is characterized by not only the expression of bone-specific markers but also the ability of differentiated cells to deposit calcium and mineralize the extracellular matrix. To determine if Smad3 null dermal fibroblasts could acquire this defining characteristic of osteoblasts, calcium nodule formation was determined at various time points in culture by alizarin red staining. As expected, the wild-type dermal fibroblasts were unable to form calcium nodules in comparison to the control osteoblasts, even when the dermal fibroblasts were cultured under the same conditions for a period of 30 days (Fig. 3D). In sharp contrast, Smad3-deficient dermal fibroblasts mineralized during the same period to a level similar to that of the osteoblast control at day 14 (Fig. 3D). This ability to form calcium nodules, a hallmark of mature osteoblasts, firmly establishes that loss of Smad3 enables dermal fibroblasts to acquire an osteoblast-like phenotype.

Significant reduction in Msx2 expression may mediate elevated Runx2 activity in Smad3 null dermal fibroblasts.

The adoption of an osteoblast-like phenotype by the Smad3-deficient dermal fibroblasts due to elevated Runx2 activity suggests that Smad3 is involved in the control of Runx2 transcriptional activity in these cells, probably through an indirect mechanism, as implied by results shown in Fig. 2D. To test this possibility, we explored whether wild-type dermal fibroblasts contain a Smad3-regulated gene product(s) that in turn acts to prevent Runx2 from binding to its DNA targets. One potential Smad target that fits the profile for such a candidate gene is the transcription factor Msx2.

Msx2 is a homeobox transcription factor that has been recognized to play an important role in bone formation. Mutations in Msx2 are associated with craniosynostosis, a disease causing skull bones to prematurely fuse (14), and with parietal foramina, a disease associated with deficient skull ossification (19). Mechanistically, the development of these diseases has been attributed to changes in the ability of Msx2 to bind to its target DNA sequence. However, recent evidence suggests that Msx2 can also bind Runx2 and repress transactivation of the OSE luciferase promoter (17), perhaps through the inhibition of Runx2 DNA-binding activity (20). Taking into account a recent report that Msx2 expression is Smad4 dependent in the context of mouse embryonic stem cells and embryonic fibroblasts (18), we investigated the possibility that loss of Msx2 directly contributes to elevated Runx2 activity in Smad3 null dermal fibroblasts. As shown in Fig. 4A, we found that Msx2 is a direct transcriptional target for TGF-β in wild-type dermal fibroblasts, since its induction was not affected by the presence of the protein synthesis inhibitor cycloheximide. Importantly, the steady-state level of Msx2 expression is high in wild-type dermal fibroblasts, whereas it is significantly reduced with the absence of Smad3 (Fig. 4B), demonstrating a requirement for Smad3 in maintaining the expression of Msx2 in dermal fibroblasts. Interestingly, Msx2 is expressed at similar levels in both wild-type and Smad3-deficient osteoblasts (Fig. 4B), suggesting that the regulation of Msx2 is controlled in a cell-type-specific manner.

FIG.4.

Significant reduction in the expression of Msx2 in Smad3 null dermal fibroblasts contributes to elevated Runx2 activity. (A) TGF-β induces the expression of Msx2 in wild-type dermal fibroblasts. Dermal fibroblasts pretreated with or without cycloheximide were incubated in the presence or absence of 100 pM TGF-β1 for the indicated times. GAPDH, glyceraldehyde-3-phosphate dehydrogenase. (B) Expression patterns of Msx2 in wild-type (WT) and Smad3-deficient (Null) dermal fibroblasts or osteoblasts isolated from two mice for each genotype of the same litter as determined by Western blot (immunoblot [IB]) analysis. The levels of Runx2 expression were used as a control. (C) A vector control, Msx2, or a Smad3 expression construct was cotransfected with the 4×OSE reporter into Smad3 null dermal fibroblasts derived from two individual newborn mice, and the luciferase activity of the reporter was measured and normalized for transfection efficiency. (D) The ability of recombinant GST-Runx2 to bind its target DNA element derived from the osteocalcin promoter was tested by EMSA in the presence of increasing amounts of recombinant GST-Msx2 or GST-Msx2 mutants containing either the insertion of a stop codon (GST-Stop) or the replacement of a proline residue with histidine (GST-His). (E) Coimmunoprecipitation assay used to determine the interactions between full-length (FL) or Runt domain-only Runx2 and wild-type Msx2 (WT) or P148H mutant Msx2 (P->H). IP, immunoprecipitate; WCL, whole-cell lysate. (F) EMSA used to test the ability of wild-type Msx2, but not mutant Msx2 (His), to block the binding of Runx2 to its target DNA derived from the BSP promoter. Nuclear extracts were isolated from Smad3-deficient dermal fibroblasts (Null DF) or wild-type osteoblasts (OB). Cold Comp., unlabeled competitor. (G) Ability of the pSuper-Retro-Msx2 shRNA constructs to knock down the expression of Msx2 in 293T cells. 293T cells were transfected with 1 mg of pcDNA3-HA-Msx2 DNA and 6 mg of the indicated shRNA construct, and cell lysates were harvested 24 h after transfection. The level of Msx2 expression was determined by Western blotting with γ-tubulin as a loading control. scr, scrambled control sequence. (H) Effect of reduced Msx2 expression in wild-type dermal fibroblasts on the induction of osteogenic differentiation as measured by the levels of alkaline phosphatase expression. Cells were infected with retroviruses carrying the scrambled sequence (Control) or shRNA targeting two different regions of the mouse Msx2 gene (shRNA1 and shRNA2, respectively) and stained for alkaline phosphatase expression 7 days later. Smad3-deficient dermal fibroblasts (Null) infected with the control retrovirus were stained at the same time as a positive control.

To determine if loss of Msx2 expression in Smad3 null dermal fibroblasts contributed to the observed increase in Runx2 activity, we first restored Msx2 expression and evaluated the activity of the OSE luciferase construct. Cotransfection of the 4×OSE reporter with the pcDNA3 vector, Smad3 expression construct, or Msx2 expression construct into Smad3 null dermal fibroblasts revealed that the presence of Smad3 or Msx2 leads to inhibition of Runx2 activity, as reflected by a significant decrease in OSE-driven promoter activity (Fig. 4C). In fact, with the coexpression of Msx2 or Smad3, the activity of the OSE luciferase reporter in Smad3 null dermal fibroblasts was reduced to a level similar to the basal activity in wild-type dermal fibroblasts (data not shown). Therefore, loss of Msx2 expression is at least partially responsible for the elevation of Runx2 activity observed in Smad3 null dermal fibroblasts. To further probe the mechanism through which Msx2 decreases Runx2 activity, we determined whether recombinant wild-type or mutant Msx2 could influence the ability of recombinant Runx2 to bind to its DNA target sequence. Mutants of Msx2 used in this assay included the proline 148-to-histidine (P148H) mutation associated with the development of craniosynostosis (14) and the stop codon insertion that truncates the C terminus and is associated with parietal foramina (19). We found that increasing amounts of recombinant Msx2 inhibited the ability of Runx2 to bind DNA (Fig. 4D) but that the two mutants of Msx2 were unable to prevent Runx2 from binding to DNA. This result was replicated with COS cells in which Runx2 was ectopically coexpressed with or without Msx2 or the mutants of Msx2 indicated above (data not shown). Consistent with the EMSA results shown in Fig. 2D, Smad3 overexpression in COS cells was not able to reduce the DNA-binding activity of Runx2 (data not shown), an observation reminiscent of previously published results (1).

We next sought to determine the mechanism responsible for the differences observed above in the ability of recombinant Msx2 or the P148H mutant of Msx2 to inhibit the DNA-binding activity of Runx2. Therefore, we evaluated by coimmunoprecipitation the ability of Msx2 or P148H mutant Msx2 to bind to full-length Runx2 or its DNA-binding domain (Runt domain) alone. As shown in Fig. 4E, we found that both Msx2 and P148H mutant Msx2 can bind to Runx2, consistent with previous results (17). On the other hand, while Msx2 is capable of binding to the Runt domain of Runx2, the P148H mutant Msx2 is unable to do so (Fig. 4E), thus providing a possible explanation for the inability of this mutant Msx2 to block Runx2 DNA binding in vitro (Fig. 4D) and a mechanism through which craniosynostosis associated with the P148H mutant Msx2 expression could develop (14). However, since the mutant Msx2 retains the ability to bind to full-length Runx2, the possibility remains that P148H Msx2 may affect Runx2 activity through as yet unknown mechanisms.

To further evaluate whether restoring Msx2 expression to Smad3 null dermal fibroblasts could decrease Runx2 DNA-binding activity, we utilized a retroviral delivery system to introduce Msx2 expression constructs into those cells due to the extremely low efficiency of the transient transfection approach. With nuclear extracts prepared from Smad3 null dermal fibroblasts infected with the recombinant retrovirus, an EMSA was performed with the OSE from the BSP promoter as a probe. We found that the Runx2 DNA-binding activity observed in Smad3 null dermal fibroblasts was reduced upon the ectopic expression of Msx2, but not that of the P148H mutant of Msx2 (Fig. 4F). Mechanistically, these data suggest that the presence of Msx2 is critical for the inactivation of Runx2 in dermal fibroblasts, consequently preventing the expression of genes associated with differentiation toward the osteoblastic lineage and maintaining the nonosteoblast phenotype of these dermal fibroblastic cells.

To further probe this postulation, we determined whether a reduction in the expression of Msx2 alone could promote the osteoblast-like phenotype in wild-type dermal fibroblasts by knocking down the expression of Msx2 through the RNA interference approach. After generating several pSuper-Retro Msx2 shRNA constructs targeting different regions of the mouse Msx2 gene, we tested their ability to knock down the expression of Msx2 in 293T cells cotransfected with the mouse pcDNA3-HA-Msx2 plasmid. We identified two of the shRNA constructs whose expression leads to a significant reduction in Msx2 protein levels as determined by Western blot analysis with an anti-Msx2 antibody (Fig. 4G). Subsequently, we infected wild-type dermal fibroblasts with retroviruses carrying the Msx2-specific shRNA or a scrambled control sequence and subjected the cells to drug selection for the stable expression of the shRNA constructs. As shown in Fig. 4H, wild-type dermal fibroblasts stably expressing either of the two different Msx2 shRNA constructs exhibit elevated alkaline phosphatase activity at day 14 in culture compared to the same type of cells expressing a control scrambled shRNA, supporting a role for Msx2 in the prevention of differentiation toward an osteoblast-like cell fate in dermal fibroblasts. In this case, the level of alkaline phosphatase activity in Msx2 shRNA-expressing wild-type dermal fibroblasts is less than that of Smad3 null dermal fibroblasts expressing the control scrambled shRNA, probably because of the incomplete elimination of endogenous Msx2 expression by the shRNA. This partial phenocopy by the Msx2 knockdown wild-type dermal fibroblasts of the Smad3-deficient cells may also be explained by the possibility that although Msx2 plays a critical role in mediating the effect of Smad3, in its absence Smad3 may prevent the differentiation of dermal fibroblasts to osteoblast-like cells through an unknown mechanism.

Conclusions and perspectives.

Our results support the presence of a novel mechanism through which the activity of Runx2, a critical transcription factor for osteoblastic differentiation from MSCs, is negatively regulated in the context of cell fate determination. Instead of controlling the expression of Runx2 in a tissue-specific manner, the activity of Runx2 in dermal fibroblasts is modulated by the presence of another transcription factor, Msx2. Msx2 expression is in turn controlled by Smad3, an effector of TGF-β that is known to play an important role in the regulation of cell fate determination and differentiation. Consistent with our finding that Msx2 plays a critical role in negatively regulating the activity of Runx2, a recent report demonstrated that reduction in Msx2 expression by the antisense strategy leads to activation of Runx2 and ossification of human ligament fibroblasts (23), suggesting that the pathway revealed in this study represents a common mechanism to negatively regulate the activity of Runx2 and prevent ossification of nonosteoblastic tissues. The importance of regulation of Runx2 inhibition is further supported by another recent study indicating that osteoblast differentiation requires loss of Twist-mediated repression of Runx2 (2).

We speculate that both the control of Runx2 activity and the control of Runx2 expression are important in the complex processes of MSC differentiation toward multiple lineages. Thus, the differentiation of mesenchymal progenitor cells toward an osteoblast lineage likely depends not only on the expression of Runx2 but also on the levels and activities of Runx2 inhibitors, such as Smad3 and Msx2. In supporting this notion, our preliminary data indicate that TGF-β can induce Msx2 expression in mouse mesenchymal progenitor cells derived from bone marrow and the loss of Smad3 in those cells promotes osteoblastic differentiation (unpublished observation).

We also anticipate that the role of Smad3 in mesenchymal differentiation toward the osteoblast lineage may involve multiple mechanisms including, but not limited to, the regulation of Msx2 expression. While our results demonstrated that the reduction in Msx2 expression in Smad3 null dermal fibroblasts contributes to development of the osteoblast-like phenotype, TGF-β/Smad3 could inhibit differentiation toward the osteoblast lineage through the transcriptional control of genes in addition to Msx2. The possibility also remains that TGF-β/Smad3 may indirectly inhibit mesenchymal differentiation toward the osteoblast lineage by sequestering the co-Smad Smad4 away from Smad1/Smad5/Smad8, which mediate bone morphogenetic protein signals that often stimulate osteogenic differentiation. Alternatively, it is possible that, at a specific stage of the mesenchymal differentiation process, TGF-β/Smad3 directly regulates Runx2 expression to control osteoblastic differentiation as observed in MC3T3 cells (1). Finally, the ability of Runx2 to bind to Smads, including Smad3 (1), may provide another mechanism through which Smad3 could regulate Runx2 activity to control cell fate determination. Thus, multiple potential direct and indirect mechanisms may be used by TGF-β/Smad3 to control osteoblastic cell fate determination through a differentiation stage- and cell-type-specific manner.

Finally, the presence of this novel mechanism in the specific cellular context of dermal fibroblasts may be generalized to support the notion that cell fate determination and complex differentiation processes can be regulated by both positive and negative forces. The expression of so-called “master genes” such as Runx2, MyoD, or PARP-γ during a specific step of the mesenchymal differentiation pathway may be countered by the presence of negative regulators, such as Smad3 and Msx2 in the case of Runx2. Thus, tight control of both master differentiation genes and their repressors could be necessary to generate proper tissues and cell types, as well as maintain the balance of homeostasis of adult tissues.