Filamin B represses chondrocyte hypertrophy in a Runx2/Smad3-dependent manner

Skeletogenesis occurs through two mechanisms: intramembranous ossification and endochondral ossification (Karsenty and Wagner, 2002). During intramembranous ossification, which occurs in some cranial bones and the clavicle, mesenchymal cells differentiate directly into osteoblasts without a chondrocyte intermediary step. In endochondral bone formation, mesenchymal cells first condense to shape the future skeletal elements. The mesenchymal cells within these condensations differentiate into chondrocytes, which express α1(II) collagen, whereas the α1(I) collagen–expressing undifferentiated mesenchymal cells at the periphery of the condensations form the perichondrium. After the formation of the cartilaginous templates, chondrocytes further differentiate into prehypertrophic chondrocytes and then into hypertrophic chondrocytes, which express α1(X) collagen. Terminally differentiated hypertrophic chondrocytes also express matrix metalloproteinase 13 (Mmp13; also called collagenase 3). This hypertrophic chondrocyte differentiation step does not occur in skeletal elements that are destined to be permanent cartilage. Finally, the osteoblasts in the perichondrium invade the hypertrophic chondrocyte area and deposit bone matrix.

Skeletogenesis is tightly controlled by transcription factors. Among them, Runx2, a transcription factor containing a runt DNA binding domain, is a key gene for osteoblast differentiation (Ducy et al., 1997). In addition, Runx2 plays multiple and opposite roles during chondrogenesis (Takeda et al., 2001; Komori, 2002; Hinoi et al., 2006). On one hand, Runx2 is transiently expressed in prehypertrophic chondrocytes, where it, in conjunction with Runx3 for some skeletal elements, is necessary to initiate chondrocyte hypertrophy (Yoshida et al., 2004), the ultimate event of chondrogenesis. Accordingly, Runx2 constitutive expression in nonhypertrophic chondrocytes leads to premature and ectopic chondrocyte hypertrophy and, thereby, bone formation (Takeda et al., 2001). On the other hand, through its perichondrial expression, Runx2 prevents chondrocyte maturation (Hinoi et al., 2006). To date, histone deacetylase 4 (HDAC4) is the only protein known to regulate Runx2 function during chondrogenesis (Vega et al., 2004). This contrasts sharply with the larger number of molecules regulating Runx2 function during osteogenesis (Bialek et al., 2004; Kronenberg, 2004; Komori, 2006) and suggests that additional regulators of Runx2's ability to favor chondrocyte hypertrophy may exist.

Filamins are a family of large cytoplasmic actin binding proteins that cross-link filamentous actin into a three-dimensional network (Stossel et al., 2001). The family consists of three members in mammals, Filamin A, B, and C (Flna, Flnb, and Flnc). All three members share well-conserved actin binding domains and a rodlike domain of 24 repeats, which is interrupted by two poorly conserved hinge domains (Popowicz et al., 2006). All three members are widely expressed, although Flnc is more restricted to skeletal and cardiac muscle (Feng and Walsh, 2004). Studies in different model organisms indicate that Filamins may interact with >30 proteins, thus implicating them in many cellular functions ranging from mechanical stability, to cell–cell and cell–matrix interactions, to integrators of signal transductions (Stossel et al., 2001; van der Flier and Sonnenberg, 2001; Popowicz et al., 2006).

Genetic evidence indicates that Filamins are essential for the development of multiple organs in human, including skeleton (Robertson et al., 2003; Krakow et al., 2004). For instance, mutations in Flnb are found in five human skeletal disorders: spondylocarpotarsal synostosis syndrome (SSS), Boomerang dysplasia, atelosteogenesis I, atelosteogenesis III, and Larsen syndrome (Krakow et al., 2004). The recessive SSS, characterized by short stature and fusions of vertebral, carpal, and tarsal bones, is caused by stop codon mutations within rod domain repeats, whereas the other dominant disorders are associated with missense mutations or small in-frame deletions. Patients with Boomerang dysplasia (Bicknell et al., 2005) have underossification of the limb bones and vertebrae. Atelosteogenesis I and III (Farrington-Rock et al., 2006) patients have vertebral abnormalities, disharmonious skeletal maturation, hypoplastic long bones, and joint dislocations. Patients with Larsen syndrome (Zhang et al., 2006) have multiple joint dislocations, craniofacial abnormalities, and accessory carpal bones. The molecular mechanisms whereby mutations in Flnb result in skeletal abnormalities remain unknown.

To study how Flnb mutations could lead to skeletal disorders, we generated mice deficient in full-length Flnb protein and found that the mutant mice display a phenotype similar to the recessive SSS. Cellular and histological analysis showed that these defects are due to premature and/or ectopic chondrocyte hypertrophy in skeletal elements that should be permanent cartilage. Molecular studies showed that these abnormalities are due to an increase in Runx2 activity. Here, we also show that Flnb does not interact with either Runx2 or HDAC4 but can bind to Smad3, a protein expressed in prehypertrophic chondrocytes and able to interact with Runx2 (Sakou et al., 1999; Kang et al., 2005). These results demonstrate for the first time that Flnb can influence the function of a transcription factor, identify a new mechanism whereby Runx2's function during chondrogenesis is regulated, and provide a molecular basis for SSS.

The Flnb gene trap embryonic stem (ES) cell line contains a β-galactosidase neomycin insertion within the 20th intron of Flnb (Fig. 1 A; see Materials and methods). The 20th exon of Flnb splices into the insertion to generate a fusion transcript, but it does not produce any wild-type transcript (Fig. 1 C and Fig. S1). The mutant transcript is predicted to encode a truncated Flnb protein of 1,624 of the full-length 2,602 amino acids and to be fused with β-galactosidase (Fig. 1 D). Intercrosses of Flnb heterozygotes in the 129/Ola-C57BL/6 mixed genetic background generate homozygous Flnb mice at Mendelian ratios at embryonic day (E) 18.5 (Fig. 1 F). However, at postnatal day (P) 5 or P21, the number of homozygous mutants is 50 and 90% decreased, respectively. The few surviving homozygous mutants are runted and have abnormal postures, with x-ray analysis showing severe kyphosis (Fig. 1 E). All these features are reminiscent of the clinical manifestation of SSS (Krakow et al., 2004).

Staining of skeletal preparations with Alizarin red for mineralized ECM and Alcian blue for unmineralized cartilaginous ECM reveals multiple skeletal defects in the Flnb mutants. For instance, similar to what is seen in SSS patients, homozygous mice have both cervical vertebrae and carpal bone fusions. Indeed, although in wild-type P30 mice, unmineralized cartilaginous ECM separates the seven cervical vertebrae, the intervertebral ECM of the mutant mice was stained by Alizarin red, indicating that this ECM was mineralized, causing vertebral fusions (Fig. 2 A). In P0 mutant mice, mineralization of the cartilage ECM, although detectable, is less extensive, suggesting that the fusion of cervical vertebrae develops progressively. The mineralization of intervertebral ECM was first detectable at E17.5, whereas no difference in cervical vertebrae staining was seen between wild type and mutant at E16.5, suggesting that the cartilage develops properly in the homozygous mutants but becomes mineralized (Fig. 2 A). In addition, the cartilage that would normally separate carpal bones in the wild-type mice is ectopically mineralized in the mutant mice. At P0, blue staining cartilaginous ECM separates the carpal bones, whereas by P15, this ECM is becoming mineralized, leading to fusions of carpal bones (Fig. 2 B).

Homozygous Flnb mutant mice also display abnormal ossification in the chondrocostal cartilage of the rib cage and sterna. In wild-type mice at all stages examined, chondrocostal cartilage and cartilage separating the sternebrae never ossify and are stained by Alcian blue. In contrast, in homozygous mutant mice, although the chondrocostal cartilage is stained with Alcian blue at P0, it stains progressively with Alizarin red at P15 and P30, indicating that this ECM is becoming prematurely mineralized (Fig. 3 A). Similarly, cartilaginous areas between sternebrae become mineralized in the mutant mice. Collectively, these ectopic ECM mineralization processes observed in cervical, carpal, and rib cage suggest that one of the Flnb functions is to regulate chondrocyte differentiation.

To determine if this ectopic mineralization was caused by ectopic chondrocyte hypertrophy, we analyzed the expression of molecular markers of proliferating, prehypertrophic, and hypertrophic chondrocytes in ribs by in situ hybridization. We used probes to α1(II) collagen, a gene expressed in proliferating and prehypertrophic chondrocytes; Indian hedgehog (Ihh) and parathyroid hormone–related peptide receptor (Ppr), two prehypertrophic chondrocyte markers; α1(X) collagen, a marker of hypertrophic chondrocytes; Mmp13, a terminally differentiated hypertrophic marker; and α1(I) collagen, which is expressed in osteoblasts and mesenchymal cells in the bone collar. At P0, although α1(II) collagen is expressed in the cartilage area between sternebrae 3/4 of wild-type mice, its expression cannot be detected in mutant mice (Fig. 3 B). Instead, the Flnb^mut cells express α1(X) collagen and Mmp13 (Fig. 3 C), indicating that the chondrocytes have differentiated into hypertrophic chondrocytes. Similarly, α1(II) collagen is expressed in the cartilaginous area between sternebrae 2/3 of P0 in wild-type mice, but its expression at this location is diminished in mutant mice. Instead, these cells express Ihh, Ppr, and α1(X) collagen. These results indicate that the chondrocytes within that area are undergoing hypertrophy in the homozygous mutant mice at P0 (Fig. 3 C). At P5, α1(II) collagen is no longer expressed in the cartilage area between sternebrae 2/3 and 3/4, but osteoblasts are present, as demonstrated by the expression of α1(I) collagen (Fig. 3 B). This observation explains the presence of ectopic bone formation in the sterna.

At P0, cells of both wild-type and Flnb^mut chondrocostal cartilage express α1(II) collagen but not α1(X) collagen, indicating that chondrocytes in the chondrocostal cartilage are not hypertrophic (Fig. 3 B). However, 5 d later, α1(X) collagen expression can be detected in the chondrocostal cartilage of the homozygous mutant mice, suggesting that hypertrophic chondrocyte differentiation has occurred (Figs. 3, B and D). Collectively, these results reveal that there is uncontrolled differentiation of chondrocytes into hypertrophic chondrocytes in the Flnb^mut chondrocostal cartilage, a cellular abnormality reminiscent of what was seen in mice overexpressing Runx2 in chondrocytes (α1[II]-Runx2 transgenic mice; Takeda et al., 2001).

The similarity between Flnb mutant and α1(II)-Runx2 phenotypes suggests that Flnb and Runx2 may interact genetically, with Flnb acting as a negative regulator of Runx2 activity. To test this hypothesis, we generated Flnb mutant mice on a Runx2 haploinsufficiency background and examined them for ectopic mineralization. Remarkably, Runx2^+/−; Flnb^mut mice have completely normal cervical vertebrae (Fig. 4 A), and the sterna phenotype observed in Flnb^mut mice is partially rescued (Fig. 4 B). Accordingly, normal expression of α1(II) and α1(X) was restored in Runx2^+/−; Flnb^mut mice (Fig. 4 C). The rescue, complete or partial, is always fully penetrant (Fig. 4 D). These results provide genetic evidence that Flnb and Runx2 interact and suggest that Flnb regulates chondrocyte hypertrophy by inhibiting Runx2 activity.

How does a cytoplasmic molecule like Flnb inhibit the function of a nuclear transcription factor like Runx2? We were unable to detect mislocalization of Runx2 and HDAC4 proteins in the Flnb mutants or to identify biochemical interactions between Flnb and Runx2 or HDAC4 (Fig. 5, A and B; and Fig. S2). Although negative and therefore to be interpreted cautiously, these observations suggest that Flnb must interact with another protein.

Filamin A, another member of the Filamin protein family, has been shown to bind CBFβ, a nuclear factor that can interact with Runx proteins. Through this interaction, Filamin A prevents CBFβ from entering the nucleus to activate the Runx1 transcription factor (Yoshida et al., 2005). Therefore, we examined whether Flnb also binds CBFβ. We failed, however, to detect any interaction between Flnb and CBFβ (Fig. 5 C), nor could we see abnormal intracellular localization of CBFβ in chondrocytes (Fig. S2). Thus, the Runx2 interaction with Flnb appears to be CBFβ independent under the conditions of these assays.

Deletion of a phosphorylation site on Smad3 causes ectopic chondrocyte hypertrophy, suggesting that TGF-β–Smad3 signals repress chondrocyte differentiation (Ferguson et al., 2000; Yang et al., 2001). Because phosphorylated Smad3 has been shown to recruit HDAC4 to inhibit Runx2 activity in osteoblasts by forming a HDAC4–Smad3–Runx2 complex (Kang et al., 2005), we tested whether Flnb could repress Runx2 activity by interacting with Smad3. Indeed, three lines of evidence support the notion that a direct interaction between Flnb and Smad3 exists. First, Smad3 can be immunoprecipitated with Flnb in an in vitro overexpression assay (Fig. 5 D). Notably, compared with Smad3, Smad2 and -5 show no detectable interaction with Flnb (Fig. 5, E and F), although an interaction with Smad4 and -6 is also detected (Fig. S2). Second, endogenous Smad3 protein can be immunoprecipitated by tagged Flnb protein (Fig. 5 G). Third, Flnb directly interacts with Smad3 in a pull-down assay (Fig. 5, H and I). Because the phosphorylation of Smad3 is important in chondrocyte differentiation, we further examined the amount of active Smad3 in mutant versus wild type. The amount of phosphorylated Smad3 is significantly increased in the rib chondrocytes of Flnb mutants by Western blot analysis (Fig. 5 J), although this could not be readily detected by immunostaining (Fig. S2). Meanwhile, the total amount of Smad3 is not changed in the mutants (Fig. 5 K). Thus, Flnb may regulate Smad3 to ensure the accumulation of an appropriate amount of activated Smad3 in the nucleus (Fig. 6 A). An excessive amount of phosphorylated Smad3 may titrate out HDAC4, leading to a failure to form the repressive HDAC4–Smad3–Runx2 complex. In this model, amounts of HDAC4 need not to be changed (Fig. S2). In support of this model, we noticed that Filb^mut rib chondrocytes contain more p-Smad3–HDAC4 complex than in wild-type cells because more HDAC4 can be coimmunoprecipitated by anti–p-Smad3 antibody (Fig. 5 L). The titration of HDAC4 proposed in this model may disrupt the repressive HDAC4–Smad3–Runx2 complex, resulting in the improper activation of Runx2 (Fig. 6 B).

Together, our data show that disruption of Flnb leads to abnormal differentiation of chondrocytes into hypertrophic chondrocytes, causing ectopic bone formation in cartilaginous elements of the cervical vertebrae, carpal bone, and rib cage. That this phenotype can be rescued by inactivating one Runx2 allele identifies an Flnb/Runx2 genetic cascade regulating chondrogenesis. Thus, the regulation of Runx2 function during chondrogenesis is more complicated than previously thought.

Our data indicate that the action of Flnb on Runx2 is mediated through the TGF-β–Smad3 pathway, likely by controlling phosphorylated Smad3. Interestingly, similar to the effect of the accumulation of phosphorylated Smad3, loss of TGF-β–Smad3 signals also causes excessive chondrocyte hypertrophy (Yang et al., 2001). These seemingly contradictory results led us to propose that the homeostasis of activated Smad3 protein is essential for suppressing Runx2 activity because the readout of TGF-β depends on the amount of HDAC4–Smad3–Runx2 triple protein complex in the nucleus (Fig. 6). Consequently, either excessive Smad3 or lack of Smad3 will result in Runx2 activation. This is consistent with both the role of Smad3 in preventing chondrocyte differentiation (Ferguson et al., 2000; Yang et al., 2001; Tchetina et al., 2006) and with reports suggesting that TGF-β can up-regulate Mmp13 production in human osteoarthritic chondrocytes (Tardif et al., 1999; Moldovan et al., 2000). We remain aware that further experiments are needed to test all aspects of this model. These results do not exclude the possibility that other mechanisms also contribute to the phenotype observed in the mutant mice.

The interaction between Flnb and Smad3 in chondrocyte differentiation revealed in this study indicates that Filamins are involved in TGF-β–Smad signaling. Supporting this notion, Filamin A physically interacts with Smad2 and -5 and functionally promotes Smad2 phosphorylation, which is suggested by the lack of phosphorylation in Filamin A–defective human melanoma cells (Sasaki et al., 2001). Although our results indicate that Flnb normally prevents excessive Smad3 phosphorylation, we hypothesize that the action of Filamins on Smad proteins may vary in different cell types.

From a biomedical point of view, this study provides a mouse model for SSS, a disease associated with nonsense mutations in FLNB; more important, it uncovers a molecular mechanism explaining the vertebral and carpal bone fusions seen in these patients. In contrast to the ectopic ossification caused by nonsense mutations in FLNB, missense mutations in FLNB cause the absence or underossification of limb bones and vertebrae in Boomerang dysplasia (Bicknell et al., 2005) and atelosteogenesis I and III (Krakow et al., 2004). In these dominant disorders, we speculate that the mutant Flnb causes delayed chondrocyte hypertrophy or affects osteoblast differentiation. In fact, Flnb may also regulate Runx2 function in osteoblast differentiation, as indicated by the less developed occipital bone, which is formed through intramembranous ossification, in Runx2^+/−; Flnb^mut than in Runx2^+/−; Flnb^wt mice (Fig. S3). Examining more alleles of Flnb in the mouse will provide further insights into its additional roles in skeletal development. In summary, our study not only reveals a new mechanism for regulating chondrocyte differentiation but also provides a molecular basis for the skeletal dysplasia caused by FLNB nonsense mutations.

The Filamin B gene trap mice were generated with the 129/Ola ES cell line XD076 from BayGenomics (http://baygenomics.ucsf.edu/). The ES cell line was expanded and injected into C57BL/6-Albino blastocysts in the Darwin Transgenic Mouse Core facility at Baylor College of Medicine (http://imgen.bcm.tmc.edu/dtmc/). Chimeras were bred to C57BL/6 mice, and germline transmission was achieved. Mice were genotyped using Southern blotting after BglII digestion with a probe to the 5′ end of the β-geo insertion, which is amplified with the primer pair: filb-5southF (AACATGGCTTGCTGTGACTG) and filb-5southR (GGAGAGGGAAATCCGAAGTC). The Runx2 knockout was used previously (Takeda et al., 2001). The primers for genotyping the Runx2 mice are as follows: Cbfa1DB (CACGGAGCACAGGAAGTTGGG), Cbfa1DF (TGAGCGACGTGAGCCCGGTGG), and Neo3F (AAGATGGATTGCACGCAGGTTCTC). The Cbfa1DB/Neo3F primer pair amplifies the mutant allele, whereas the Cbfa1DB/Cbfa1DF primer pair amplifies the wild-type allele. For timed matings, the morning of the vaginal plug was designated E0.5. All animal protocols were approved by the Institutional Animal Care and Use Committee of Baylor College of Medicine.

Total RNA was prepared from liver or kidney dissected from P21 Flnb homozygous, heterozygous, and wild-type mice (n = 3 for each genotype) using RNAqueous-4PCR kit (Ambion). Transcripts were amplified using Superscript One-Step RT-PCR Systems (Invitrogen). The primers to amplify the wild-type Flnb transcript spanning the β-geo insertion are as follows: rtf1 (TCTTCCCACATACGATGCAA) and rtr (TCCACTACAAAGCCCACCTC). The primers to amplify the mutant transcript (fusion of Flnb and β-geo insertion) are as follows: rtf2 (CCTATATCCCTGATAAGACCGGACGC) and gal (GACAGTATCGGCCTCAGGAAGATCG).

Mice were dissected, fixed in 95% ethanol overnight, and stained in 0.015% alcian blue (Sigma-Aldrich) dye overnight. The carcasses were treated with 2% KOH for 1 to several days until most soft tissue disappeared. After removing the remaining soft tissues carefully with forceps, the mice were stained in 0.005% alizarin red (Sigma-Aldrich) solution for 30 min to several hours depending on the age of the mice. Finally, skeletons were cleared in a 20% glycerol/1% KOH solution. At least four mice of each genotype were analyzed for each stage.

Mice at P0 were dissected so that most tissues other than skeleton are removed. The skeletons were fixed in 0.2% glutaraldehyde for 1 h at RT followed by washing in PBS twice. Fixed specimens were then put in fresh staining solution overnight at RT. The staining solution was made by mixing 1 ml 10× stock solution, 250 μl 0.2 M EGTA, and 250 μl 40 mg/ml X-gal into 8.5 ml PBS. The 10× stock solution was made as follows: for 100 ml solution, add 1.646 g K₄Fe(CN)₆ Ferro, 2.112 g K₃Fe(CN)₆ Ferri, 2 ml 1 M MgCl₂, 2 ml 1% NP40, and 0.01 g sodium deoxycholate. X-gal stock solution was made by dissolving 100 mg X-gal in 2.5 ml dimethylformamide. All microscopy for skeletal preparation and X-gal staining of the skeleton are performed with Stemi SV11 microscopy (Carl Zeiss MicroImaging, Inc.) with 0.5× (Plan) and 1.0× (Planapo) objectives at RT. Pictures were taken with a color video camera (DXC 300 3CCD; Sony), captured through the Spot software (Diagnostic Instruments), and further processed with Photoshop (Adobe).

Tissues were fixed in 4% paraformaldehyde/PBS overnight at 4°C. Mice obtained after P0 were decalcified in TBD-2 solution (Thermo Shandon) overnight at 4°C. Specimens were dehydrated and embedded in paraffin and sectioned at 5 μm. For histological analysis, sections were stained with hematoxylin and eosin. In situ hybridization was performed using digoxigenin-labeled riboprobes (Roche). The α1(I), α1(II), and α1(X) probes were used previously (Takeda et al., 2001). Hybridizations were performed overnight at 60°C, and washes were performed at 65°C. Sections were then incubated for 2 h with alkaline phosphatase–linked anti-digoxigenin antibody (1:2,000; Roche) at RT followed by color reactions with BM purple substrate (Roche). Sections were finally counterstained with nuclear fast red (Vector Laboratories) and mounted with Permount. Three mice of each genotype were analyzed for each stage. The microscopy for all histology and in situ sections were performed using a microscope (AxioPlan2; Carl Zeiss MicroImaging, Inc.) with 5× (Plan-Neofluar), 10× (Plan-Apochromat), or 20× (Plan-Apochromat) objectives at RT. All images were taken with AxioCam (Carl Zeiss MicroImaging, Inc.), captured through the AxioVision software (Carl Zeiss MicroImaging, Inc.), and processed with Photoshop software (Adobe).

Smad-expressing mammalian plasmids were provided by K. Watanabe (National Institute for Longevity Sciences, Aichi, Japan; pF:Smad1, -2, -4, -5, and -6) and R. Derynck (University of California, San Francisco, San Francisco, CA; flag-Smad3; Sasaki et al., 2001; Kang et al., 2005). The CBFβ- expressing plasmid was obtained from T. Watanabe (Tohoku University, Sendai, Japan; Yoshida et al., 2005). To generate myc-tagged N terminus–truncated mammalian-expressing Filamin B, the C-terminal 231-amino-acid coding region was generated by PCR and inserted into the StuI–XbaI sites of the CS2+MT vector. To produce his-tagged Smad3 protein, PCR-amplified Smad3 gene was subcloned into NdeI and BamHI sites of a bacterial expression vector, pET-15b (Novagen). To produce GST fusion protein, the C terminus of Filamin B was PCR amplified and subcloned into NdeI and HindIII sites of a pGEX2TL.

BL21-pLysS was transformed with His-tagged Smad3 and GST-fused Filamin B bacterial expression vectors, respectively. After induction with IPTG, the bacterial pellets were lysed by sonication in the presence of Ni-column binding buffer (20 mM TrisCl, pH 7.9, 500 mM NaCl, and 5 mM Imidazole) and GST binding buffer (20 mM TrisCl, pH 7.9, 1 M NaCl, 0.2 mM EDTA, 2 mM DTT, and 10% Glycerol), respectively, and centrifugation was followed. From each supernatant, His-tagged Smad3 and GST-fused Filamin B proteins were isolated by Ni-NTA (GE Healthcare) chromatography and glutathione-coupled Sepharose (GE Healthcare) chromatography.

5 mg of purified His-tagged Smad3 protein was incubated with either 10 mg immobilized GST or GST-fused Filamin B on glutathione-coupled Sepharose in BC300 (20 mM Tris, pH 7.9, 0.2 mM EDTA, 300 mM KCl, 20% glycerol, 0.1% NP-40, and 1 mM PMSF) solution containing 2 mg/ml BSA for 6 h at 4°C. The immobilized GST or GST-fused Filamin B beads were washed extensively with BC300 and analyzed by immunoblot after SDS-PAGE.

For the immunoprecipitation and immunoblotting for the interaction between overexpressed Filamin B and overexpressed Smad3, 293T cells were cotransfected with myc-tagged Flnb and flag-tagged Runx2, HDAC4, CBFβ, and Smad proteins, respectively, using Lipofectamine 2000 (Invitrogen) according to the manufacturer's manual. 48 h later, cells were harvested and washed once with cold PBS. The cell pellet was suspended and lysed by adding 10 times the volume of hypotonic buffer. The hypotonic buffer is composed of 50 mM Tris-Cl, pH 7.5, 2 mM EDTA, 0.1% NP-40, 1 mM PMSF, and phosphatase inhibitors. The lysates were subjected to immunoprecipitation with anti-myc antibody (Santa Cruz Biotechnology, Inc.) and protein A agarose overnight, followed by Western blotting with anti-myc and anti-Flag (Sigma-Aldrich) antibodies.

For the immunoprecipitation and immunoblotting for the interaction between overexpressed Filamin B and endogenous Smad3, 293T cells transfected with myc-tagged Filamin B expression vector using Lipofectamine 2000 (Invitrogen) according to the manufacturer's manual. 48 h later, cells were harvested and washed once with PBS. The cell pellet was lysed by adding 10 times more hypotonic buffer than the pellet volume and resuspending the pellet. The hypotonic buffer is composed of 50 mM Tris-Cl, pH 7.5, 2 mM EDTA, 0.1% NP-40, 1 mM PMSF, and phosphatase inhibitors. The lysates were subjected to immunoprecipitation with anti-myc antibody and protein A agarose for overnight, followed by Western blotting with anti-myc and -Smad3 (Upstate Biotechnology) antibodies.

For the detection of p-Smad3 level and coimmunoprecipitation of p-Smad3 and HDAC4, cartilaginous tissues from Rib cages of wild-type and Filamin B mutant mice at P4 were homogenized in 500 ml lysis buffer using a tissue miser (Tekmar). The lysis buffer is composed of 50 mM Tris, pH 7.5, 2 mM EDTA, 150 mM NaCl, 1% Triton X-100, and protease and phosphatase inhibitor cocktail. These lysates were subjected to immunoprecipitation with anti–phospho-Smad2/3 antibody (Santa Cruz Biotechnology, Inc.) and protein A agarose for 6 h, followed by immunoblotting with anti-Smad3 and anti-HDAC4 (Santa Cruz Biotechnology, Inc.) antibody. As an input control of the immunoprecipitation, the lysates were subjected to immunoblotting with anti-tubulin antibody (Santa Cruz Biotechnology, Inc.).

Fig. S1 shows the expression of different portions of the Filamin B transcript in the mutant compared with the wild type. Fig. S2 shows the expression of CBFβ, HDAC4, p-Smad3, and Runx2 in the chondrocytes and interaction between Filamin B and Smad4 or -6. Fig. S3 shows the aggravated occipital bone phenotype by deleting Filamin B in the background of Runx2 haploinsufficiency.

We thank Drs. Patricia Ducy, Richard Behringer, and Brendan Lee for critically reviewing the manuscript.

This work was supported by Public Health Service grants P01 ES11253 and U01 HD39372, by a gift from the Kleberg Foundation to M.J. Justice, and by The National Institute of Arthritis and Musculoskeletal and Skin Diseases grant R01 AR 045548 to G. Karsenty.