Filamin-interacting proteins, Cfm1 and Cfm2, are essential for the formation of cartilaginous skeletal elements

MutationsofFilamingenes,whichencodeactin-bindingproteins,causeawiderangeofcongenitaldevelopmen-tal malformations in humans, mainly skeletal abnormalities. However, the molecular mechanisms underlying Filamin functions in skeletal system formation remain elusive. In our screen to identify skeletal development molecules, we found that Cfm ( Fam101 ) genes, Cfm1 ( Fam101b ) and Cfm2 ( Fam101a ), are predominantly co-expressed in developing cartilage and intervertebral discs (IVDs). To investigate the functional role of Cfm genes in skeletal development, we generated single knockout mice for Cfm1 and Cfm2 , as well as Cfm1/Cfm2 double-knockout ( Cfm DKO) mice, by targeted gene disruption. Mice with loss of a single Cfm gene displayed no overt phenotype, whereas Cfm DKO mice showed skeletal malformations including spinal curvatures, vertebral fusions and impairment of bone growth, showing that the phenotypes of Cfm DKO mice resemble those of Filamin B (Flnb) -deﬁcient mice. The number of cartilaginous cells in IVDs is remarkably reduced, and chondrocytes are moderately reduced in Cfm DKO mice. We observed increased apoptosis and decreased proliferation in Cfm DKO cartilaginous cells. In addition to direct interaction between Cfm and Filamin proteins in developing chondrocytes, we showed that Cfm is required for the interaction between Flnb and Smad3, which was reported to regulate Runx2 expression. Furthermore, we found that Cfm DKO primary chondrocytes showed decreased cellularsizeand fewer actinbundlescompared with thoseofwild-typechondrocytes. Theseresults suggestthat Cfms areessentialpartnermoleculesofFlnbinregulatingdifferentiationandproliferation ofchondryocytesand actin dynamics.


INTRODUCTION
Chondrocytes provide an essential function in the skeletal system by producing and maintaining the cartilaginous matrix. In addition, chondrocytes play critical roles at several stages of endochondral ossification in skeletal development. Chondrocyte development is regulated by multiple cell-extrinsic and cellintrinsic factors, including growth factors and transcription factors (1,2). Interestingly, various studies have shown the importance of actin dynamics for chondrocyte differentiation (3). * To whom correspondence should be addressed at: Laboratory for Molecular and Developmental Biology, Institute for Protein Research, Osaka University, 3-2 Yamadaoka, Suita, Osaka, 565-0871, Japan. Tel: +81 668798631; Fax: +81 668798633; Email: takahisa.furukawa@protein.osaka-u.ac.jp Inhibitors of actin polymerization stimulate chondrocyte differentiation in cultured mesenchymal cells and murine embryonic stem cells (4,5). Importantly, abnormalities of the actin cytoskeletal system lead to various human chondrodysplasias. Actin cytoskeleton organization, arranged and rearranged by the assembly and disassembly of actins, is regulated by a large number of actin-binding proteins.
Filamin, an actin-binding protein, plays an important role in skeletogenesis. Filamin functions in stabilization of the actin cytoskeleton, linkage of the actin network with cellular membranes and mediation of the interactions between actin and transmembrane receptors (6). In addition to filamentous actin, Filamins have been reported to interact directly with .90 other binding partners with great functional diversity (7). In mammals, the Filamin family consists of three members: Filamin A (Flna), Filamin B (Flnb) and Filamin C (Flnc). The amino acid sequences of Filamins are very similar to each other, and share well-conserved actin-binding domains (8). Flna and Flnb are ubiquitously expressed, although Flnc is restricted to heart and skeletal muscles (9). Filamins are critical for the development of multiple human organs (10,11). Mutations in Flnb were found in five cases of human skeletal malformations including autosomal recessive spondylocarpotarsal syndrome (OMIM:#272460), autosomal dominant boomerang dysplasia (OMIM:#112310), Larsen syndrome (OMIM:#150250), atelosteogenesis type I (OMIM:#108720) and atelosteogenesis type III (OMIM:#108721) phenotypes (11)(12)(13)(14). Consistent with abnormalities in human patients with Flnb mutation, defects of Flnb in mice produce skeletal malformations, including vertebral fusions, scoliosis, kyphosis and shortening of the distal appendages (15)(16)(17)(18). Clustered missense mutations in Flna have been identified in a diverse spectrum of congenital malformations in humans, including otopalatodigital syndrome type 1 (OMIM:#311300), otopalatodigital syndrome type 2 (OMIM:#304120), frontometaphyseal dysplasia (OMIM:#305620) and Melnick-Needles syndrome (OMIM:#309350) (10,19,20). Similar to the abnormalities in human patients with Flna mutations, mice without Flna showed cardiac malformations and skeletal defects of the sternum and palate (21). Based on these observations, the importance of Filamin in skeletal development is well recognized. However, the precise molecular mechanisms underlying Filamin function in skeletal development are largely unknown.
To identify important molecules involved in skeletal system development, we performed a comparative microarray analysis between undifferentiated and differentiation-induced ATDC5 cells, a murine chondroprogenitor cell line (22). We identified multiple uncharacterized genes whose expression significantly increased in differentiated ATDC5 cells. In our previous report, we showed that Obif (Osteoblast induction factor, Tmem119), one of these uncharacterized genes, is essential for bone formation in association with osteoblastogenesis in mice (23). In the same screen, we also identified Cfm2 (Fam101a, RefilinA), which encodes a 204 amino acid protein of an unknown function in vivo. A Cfm2 paralog, Cfm1 (Fam101b, RefilinB), is expressed broadly in the developing central nervous system, including the forebrain and midbrain, however, Cfm1 expression in the skeletal system has not been analyzed. Cfm1 homozygous mutant mice (Cfm1 2/2 mice) have been shown to be viable and without any apparent developmental defects (24), probably due to functional redundancies between Cfm1 and Cfm2. To elucidate Cfm functions in vivo, we generated Cfm2 mutant mice by targeted gene disruption. A single gene knockout of Cfm2 showed no obvious phenotype, whereas Cfm1/Cfm2 double-knockout (Cfm DKO) mice displayed severe malformations due to premature ossification. Interestingly, these cytoskeletal defects of Cfm DKO mice were similar to those of Filamin mutant mice. In addition, we found that Cfm proteins co-localize with Flna along actin bundle-like structures in ATDC5 cells. These results led to the idea that Cfm proteins control Filamin functions and suggest that loss of Cfm proteins modify actin dynamics in developing cartilaginous cells.
Our results suggest that Cfm genes regulate cell survival and cell proliferation, and modulate the cyto-and nucleoskeleton through regulating actin dynamics by interacting with Filamins in cartilaginous cells. The current study shows that Cfm genes play an essential role in the formation of cartilaginous skeletal elements, leading to clarification of the molecular basis for Filamin functions in the skeletal system formation.

Expression of Cfm1 and Cfm2
In the course of our microarray screening for genes specifically upregulated upon chondrocyte differentiation (22), we found that mouse Cfm2 transcripts markedly increased in ATDC5 cells upon differentiation. Mouse Cfm2 encodes a 204 amino acid protein with an unknown function in vivo. Hirano et al. identified Cfm1, using a screen to find genes that play roles in initial regionalization of anterior neuroectoderm (24). Mouse Cfm1 encodes a 216 amino acid protein with no known functional domains. There are two conserved domains, termed CR1 and CR2, in both Cfm1 and Cfm2 (24). The CR1 domain of mouse Cfm1 has 100% identity with that of mouse Cfm2. The CR2 domain of mouse Cfm1 has 43% identity with that of mouse Cfm2 (Fig. 1A).
To investigate Cfm2 expression during mouse skeletal development, we performed in situ hybridization analysis using a Cfm2 probe. Cfm2 expression was first detected in the marginal zone of the vertebral primordia at embryonic day 12.5 (E12.5), consistent with Collagen2 (Col2), a gene expressed in proliferating and prehypertrophic chondrocytes (  Fig. S1B). At E18.5, Cfm2 expression increased in the layers of proliferating and prehypertrophic chondrocytes (Supplementary Material, Fig. S1C). Furthermore, Cfm2 expression was also observed in intervertebral disc (IVD), including nucleus pulposus (NPs) and annulus fibrosus (AFs), during skeletal development ( Fig. 1H and I).
Taken together, these results demonstrated that Cfm genes are predominantly expressed in proliferating and prehypertrophic chondrocytes, and in nucleus pulposus cells.
Single Cfm gene mutants showed no obvious phenotype To examine a possible functional role of Cfm2 in development, we generated Cfm2-deficient mice by targeted gene disruption. We deleted the first exon that contains a start codon of the Cfm2 open reading frame (Supplementary Material, Fig. S3A). We confirmed that the 24 kDa full-length Cfm2 protein was undetectable by western blots using an antibody against the Cfm2 N-terminus (residues 1-110) in the femur from Cfm2 2/2 mice (Supplementary Material, Fig. S3B). Cfm2 2/2 mice were born in the expected Mendelian ratio and were indistinguishable in appearance from control littermates (Supplementary Material, Fig. S3C and D). Body weight, body length and tibial length at 12 weeks were unaltered between control and Cfm2 2/2 mice (Supplementary Material, Fig. S3E-G). There was no significant difference in survival between Cfm2 2/2 and wild-type mice (data not shown). Cfm1 2/2 mice were previously established, and it was reported that they exhibited no obvious phenotype in the developing brain (24). As is the case with Cfm2 2/2 mice, Cfm1 2/2 mice exhibited no overt phenotype (Supplementary Material, Fig. S3H -L), and survival was not significantly different between Cfm1 2/2 mice and wild-type mice (data not shown). The lumbar disks from Cfm2 2/2 and Cfm1 2/2 mice were unaltered when compared with those of wild-type mice (Supplementary Material, Fig. S3M).
To examine the compensatory expression of Cfm genes in mice development, we analyzed the expression level of Cfm1 in the lung, which expresses substantial levels of Filamins and Cfms, in Cfm2-deficient mice. Although the expression level of Cfm1 in Cfm2-deficient mice was slightly upregulated compared with that of wild-type mice, it was not statistically significant (Supplementary Material, Fig. S4A). Reciprocally, we analyzed the expression level of Cfm2 of the lung in Cfm1-deficient mice compared with that of wild-type mice. We found that the expression level of Cfm2 in Cfm1-deficient mice was also slightly upregulated compared with that of wild-type mice, but was not statistically significant (Supplementary Material, Fig. S4B).
Cfm DKO mice showed severe skeletal malformations in the spine Since we suspected functional redundancies between Cfm1 and Cfm2 for development of the skeletal system, we generated Cfm DKO mice to investigate Cfm function. Cfm DKO neonatal mice had an equal number of male and female mice, and the frequency of each genotype was close to the expected Mendelian ratio ( Fig. 2A).
We next examined the effect of the loss of Cfm genes during growth after birth. Newborn Cfm DKO mice showed slightly stunted growth compared with newborn wild-type mice, whereas there was no obvious difference in vertebral columns between control and Cfm DKO mice at P0 using Alcian blue and Alizarin red staining for cartilage (blue) and bones (red) (Fig. 2B). To confirm that the loss of Cfm genes affects skeletal formation, we analyzed Cfm DKO mice at P4 using a microcomputed tomography (mCT) instrument. We observed the small size vertebra, vertebral body osteopenia and shortening of the distance between vertebral bodies (Fig. 2C), and the pups of Cfm DKO mice showed kyphoscoliosis and decreased body weight, and height at P28 in Cfm DKO mice ( Fig. 2D and E). The fusions of vertebral bodies were observed throughout the vertebral column in the skeletal preparation at P28 (Fig. 2F). The crown-rump (CR) length, femoral length and width of Cfm DKO mice were significantly shorter than those of wild-type mice at 8 weeks (Fig. 2G). Taking into consideration these results and Cfm gene expression, we focused on the vertebral columns in Cfm DKO mice. In mCT analysis at 8 weeks, the spaces for IVD and intervertebral foramina were absent and the structures of spinous processes could not be identified in Cfm DKO mice (Fig. 2, H and I ′ ). Since individual vertebrae could not be distinguished, the vertebral column appeared as one unified block in Cfm DKO mice.
We further examined the process of abnormal vertebra formation at P14. Cfm DKO mice showed kyphosis, and the IVDs were almost absent (Supplementary Material, Fig. S5A and B). The cartilages of sternums in Cfm DKO mice prematurely mineralized at P14 (Supplementary Material, Fig. S5C). In the thoracic region, some of the transverse processes were absent or fused in Cfm DKO mice (Fig. 3A). The mineralization of vertebral bodies was advanced in Cfm DKO mice ( Fig. 3B and C). IVDs often showed notable shrinkage in the thoracic regions of Cfm DKO mice (Fig. 3B). The transverse processes were hypoplastic in the lumbar region, and IVDs displayed notable shrinkage throughout this region in Cfm DKO mice (Fig. 3C). The average of lumbar IVD height (L1 -L4) in Cfm DKO mice was significantly shorter than that in wild-type mice (Fig. 3D). Histological analysis revealed that NP cells in Cfm DKO mice decreased remarkably at 3 weeks (Fig. 3E). NPs and AFs of IVDs were diminished, and vertebral bodies were misshapen in Cfm DKO mice at 12 weeks (Fig. 3F). However, thoracic disks and lumbar disks of Cfm +/2 /Cfm2 2/2 mice and Cfm1 2/2 /Cfm2 +/2 mice were similar to those of wild-type mice at 4 weeks (Supplementary Material, Fig. S6).
To investigate abnormalities in other systems such as cardiovascular, central nervous and urogenital organs, we analyzed brains, hearts and kidneys in Cfm DKO mice at 4 weeks (Supplementary Material, Fig. S7A -F). As far as we examined, there was no substantial difference between wild-type and Cfm DKO mice (Supplementary Material, Fig. S7A, C and E). To further investigate tissue abnormalities, we analyzed sections of the brain, heart and kidney, in Cfm DKO mice, and confirmed that there was no substantial difference between wild-type and Cfm DKO mice (Supplementary Material, Fig. S7B, D and F). Moreover, there was no significant difference in survival between wild-type and Cfm DKO mice (data not shown).

Chondrocyte maturation is accelerated in Cfm DKO mice
To precisely determine if abnormal mineralization results from the absence of Cfm genes, we analyzed the expression of skeletal

2956
Human Molecular Genetics, 2014, Vol. 23, No. 11 development markers in the sternums of Cfm DKO mice at P1 and P4 (Fig. 4Aand B). We used probes for Collagen2 (Col2, a marker of proliferating and prehypertrophic chondrocytes), Col-lagen10 (Col10, a gene expressed in hypertrophic chondrocytes), Osteopontin (OPN, a gene expressed in terminally differentiated hypertrophic chondrocytes and osteoblasts), Col-lagen1 (Col1, a marker of osteoblasts) and Runx2 (a gene expressed at high levels in prehypertrophic and hypertrophic chondrocytes and osteoblasts). At P1, Col2 is still expressed in chondrocytes between sternebrae 3 and 4 as well as in those between sternebrae 2 and 3 in wild-type mice; however, Col2 is not expressed in chondrocytes between sternebrae 3 and 4 in Cfm DKO mice. At this stage, chondrocytes in Cfm DKO mice are being terminally differentiated, as demonstrated by the fusion of Col10-expressing cell layers (Fig. 4A). At P4, Col10-expressing cell layers were also undergoing fusion between sternabrae 2 and 3 in Cfm DKO mice (Fig. 4B).
Col10-expressing cell layers disappeared in the area between sternabrae 3 and 4. OPN, Runx2 and Col1 were expressed in this area, showing its precocious ossification. (Fig. 4B and C). At P5, cartilage in this area completely disappeared (Fig. 4D). These findings suggest that chondrocyte maturation is accelerated in Cfm DKO mice, leading to precocious ossification.
Complete loss of Cfm genes affects chondrogenesis in the growth plate and bone growth Since Cfm genes were strongly expressed in growth plates of the longitudinal bones and vertebral bodies and Cfm DKO mice showed growth retardation (Fig. 1E, H, I, M, N and Fig. 2D-G), we examined a possible role of Cfm genes in chondrocytes and bone growth. The CR length and tibial length of Cfm DKO mice were significantly shorter than those of wild-type, Cfm +/2 /Cfm2 2/2 and Cfm1 2/2 /Cfm2 +/2 mice at 4 weeks (Fig. 5A). Histological analysis of proximal tibiae from 4-week-old mice revealed that a 20 -30% reduction in growth plate thickness in Cfm DKO mice ( Fig. 5B and C). Although the parameters of the resting zone (RZ) in Cfm DKO mice were unaltered from those of wild-type mice, the thickness of the proliferating zone (PZ) and the hypertrophic zone (HZ) were reduced in Cfm DKO mice ( Fig. 5B and C). Proliferating cells were properly oriented along the long axis of the bone, whereas the number of cells and cells per column were reduced in Cfm DKO mice ( Fig. 5B and D). A reduction of the number of hypertrophic cells was observed in Cfm DKO mice (Fig. 5D).

Cfm DKO mice exhibited decreased bone formation
Since the bone of vertebral bodies showed a moth-eaten appearance in Cfm DKO mice at P21 in mCT analysis (Supplementary Material, Fig. S8A), we investigated Cfm gene expression in the mouse bone tissue. We observed that Cfm genes were expressed at low level in the mouse calvariae using Q-PCR analysis (Supplementary Material, Fig. S2A and B). To assess the effect of Cfm gene loss for bone formation, we performed a quantitative mCT analysis. The bone volume/tissue volume was unaltered between wild-type mice and Cfm DKO mice. However, the trabecular number of Cfm DKO mice significantly decreased, whereas the trabecular separation of Cfm DKO mice significantly increased compared with that of wild-type mice (Supplementary Material, Fig. S8B and C). Therefore, Cfm1 and Cfm2 may also have some functions in osteoblasts or osteoclasts. Cfm genes are required for cartilaginous cell proliferation and apoptosis The reduced number of cartilaginous cells including NP cells, AF cells and chondrocytes suggested that Cfm1 and Cfm2 might protect against apoptosis. To assess this possibility, we performed TUNEL assays using the spines of Cfm DKO mice at P28. TUNEL-positive cells increased remarkably in the NP and AF, and moderately in the VBGP of Cfm DKO mice (Fig. 6A-G). These results suggest that increased apoptosis in Cfm DKO mice cause defects of the IVD and of growth plate development. We investigated the effect of Cfm gene loss on cell proliferation rates using primary chondrocytes from the rib growth plates of Cfm DKO mice at P24. The proliferation rate in Cfm DKO chondrocytes significantly decreased compared with that in control chondrocytes (Fig. 6H).

Cfms interact with Filamins in chondrocytes
To identify the mechanism underlying Cfm-mediated skeletogenesis, we performed a yeast two-hybrid screen using a cDNA library prepared from mouse whole embryo at E11 and a full-length Cfm2 construct as bait.  Human Molecular Genetics, 2014, Vol. 23, No. 11 2959 ( Fig. 7A and B ′ ). Next, to investigate the interaction between the endogenous Filamin and Cfm in chondrocytes, we performed an immunoprecipitation assay using primary chondrocytes isolated from E18.5 wild-type embryos. We detected specific interactions between Cfm1 or Cfm2 with Flna and Flnb (Fig. 7C). These results suggest that the conformation of Filamins is functionally controlled by Cfm, and loss of Cfm genes affects the actin cytoskeleton in chondrocytes.
To examine the expression level of Filamin in Cfm deficiency, we analyzed the lungs of Cfm1 2/2 , Cfm2 2/2 and Cfm DKO mice using Q-PCR. However, the expression level of Flna and Flnb was unaltered between wild-type and Cfm-deficient mice (Supplementary Material, Fig. S9A and B).

Loss of Cfm genes affects actin bundle formation and cell morphology of chondrocytes
Our results demonstrated that Cfm1 and Cfm2 regulate chondrocyte survival and proliferation, and interact with Filamins. Cfms interacted with Filamins to organize perinuclear actin networks and regulated nuclear shape (25,27). Therefore, we examined whether Cfm1 and Cfm2 play intrinsic roles in actin filament networks and nuclear shape. We found that primary rib chondrocytes from wild-type mice often formed actin bundles, whereas those from Cfm DKO mice formed fewer actin bundles (Fig. 7D). We found that the cell surface area and the length of the nuclear long axis of chondrocytes from Cfm DKO mice significantly decreased compared with those of wild-type mice (Fig. 7E). These results suggest that Cfm promotes actin bundles and regulates cell shape through interaction with Filamins in chondrocytes.

Cfms are involved in the interaction of Flnb and Smad3
Since Flnb interacts with Smad3 and inhibits phosphorylation of Smad3, phospho-Smad3 increases in Flnb-deficient mice, leading to Runx2 hyperactivity (18). Thus, fewer phospo-Smad3-HDAC4 complexes bind Runx2, and abnormal differentiation of chondrocytes occurs in Flnb-deficient mice (18). This report and results in the current study raise the possibility that Cfm proteins associate with Flnb, influence the interaction between Flnb and Smad3, and regulate Runx2 function. To assess this hypothesis, we first performed an immunoprecipitation assay using primary chondrocytes isolated from ribs of P3 wild-type and Cfm DKO mice. We found that Cfm DKO rib chondrocytes contain less Flnb-Smad3 complex than those of wild-type chondrocytes (Supplementary Material, Fig. S10). This result suggests that Cfm is required for the interaction Cfm-deficient fibroblast had no effect on cellular migration during embryonic development Cell migration is an actin-dependent process. We tested whether the absence of Cfms would have an influence on cellular migration. The migration degree of primary fibroblasts isolated from E14.5 embryos was unaltered between wild-type mice and Cfm DKO mice using the Boyden chamber assay ( Supplementary Material, Fig. S11A). Morphology of Flna and actin filament of primary fibroblasts was unaltered between wild-type and Cfm DKO mice (Supplementary Material, Fig. S11B). These observations suggest that the absence of Cfm genes had no effect on fibroblast cellular migration during embryonic development.

DISCUSSION
In the current study, we found by microarray analysis that Cfm2 gene expression increased during chondrocyte differentiation and that Cfm2 is expressed predominantly in cartilaginous tissues, including the developing IVD and cartilage, by in situ hybridization. Mouse Cfm2 encodes a 204 amino acid protein without a known functional domain, showing high homology with Cfm1 protein. This indicates that Cfm1 and Cfm2 genes are paralogous. Both Cfm1 and Cfm2 are evolutionarily well conserved among humans, mice, chickens, Xenopus and zebrafish. There are two motifs termed CR1 and CR2, which are conserved throughout these species, in both Cfm1 and Cfm2 proteins (24).
To investigate the biological function of Cfm2 during skeletal development, we generated Cfm2-null mice by targeted gene disruption. Cfm2-null mice displayed no overt phenotype. It was previously reported that Cfm1 was uniquely expressed in the developing forebrain and midbrain, however, Cfm1 expression in the skeletal system has not been analyzed (24). In the current study, we found that Cfm1 is also expressed in cartilaginous tissues in the vertebral cartilage and IVD during skeletal development. Since we assumed functional redundancies between Cfm1 and Cfm2, we generated Cfm double-mutant mice. Cfm DKO mice exhibited severe skeletal malformations, as characterized by scoliosis, kyphosis, IVD defects, vertebral fusion in the spine and longitudinal bone growth retardation. We showed that these malformations are due to accelerated chondrocyte maturation, reduced chondrocyte proliferation and increased apoptosis in chondrocytes.
How do Cfm proteins play a role in skeletal formation? We found that Cfm proteins interact with Filamin, consistent with data published recently (25). Filamin is a ubiquitously expressed actin-binding protein that has been implicated in many cellular processes including cell proliferation, cell migration and signaling pathways that mediate organogenesis in multiple tissues (8,28). We examined the phenotypic similarity between Cfm DKO mice and Filamin-deficient mice. We observed that Cfm DKO mice displayed skeletal abnormalities similar to the malformations in Flnb-deficient mice related to scoliosis, kyphosis, vertebral fusion in the spine and shortening of the distal appendages (15)(16)(17)(18). In addition, the malformations in Cfm DKO mice are similar to those in mutational congenital anomalies of Flnb including autosomal recessive spondylocarpotarsal syndrome, autosomal dominant boomerang dysplasia, Larsen syndrome and atelosteogenesis I and III phenotypes (11)(12)(13)(14)29,30). These results suggest the possibility that Cfm and Flnb complex function mainly in skeletogenesis. Contrarily, there are phenotypic similarities and differences between Filamin-deficient mice and Cfm DKO mice. Flna mutant mice showed cardiac and skeletal malformations, including sternum and palate defects (21). Multiple lines of evidence showed that Flna deficiency in humans disrupts not only skeletal development but also development of other tissues including the neuronal and cardiac systems. Missense mutations in human Flna produced the severe thoracic hypoplasia, irregular ribs and scoliosis known as Melnick-Needles syndrome (19). Null mutations in human Flna cause X-chromosome-linked brain abnormalities known as periventricular nodular heterotopia (PVNH; OMIM:#300049). Neurons in PVNH patients fail to undergo radial migration from the ventricular zone to form the six-layered neocortex during fetal development (26,31). In contrast, we observed that Cfm DKO mice exhibited only skeletal malformation. Thus, these data suggest that Cfm proteins function with Flnb proteins predominantly in the developing skeletal system. Transforming growth factor-beta (TGF-b) is one of the key regulators in skeletal development, and Smad2 and Smad3 mediate intracellular signaling of TGF-b (32). It has been reported that Flnb binds to diverse proteins, including Integrin b1 and Smad3 (16,18). Flnb interacts with Smad3, inhibiting phosphorylation of Smad3, and leading to suppression of chondrogenesis (18). This previous study indicated that Flnb normally prevents excessive Smad3 phosphorylation. In the current study, we examined the possibility that Smad3 is unable to interact with Flnb in Cfm-deficient chondrocytes; however, the Smad3 phosphorylation level in Cfm-deficient chondrocytes was unaltered as compared with that in wild-type chondrocytes.
Smad3-deficient mice showed malformations including kyphosis, and decreased proteoglycan and collagen content in the IVDs; however, those mice did not exhibit scoliosis and vertebral fusions, which are observed in both Flnb-deficient mice and Cfm DKO mice (15)(16)(17)(18)33). It was reported that TGF-b1 stimulation upregulated Cfm1 protein expression, but Cfm2 protein and mRNA were below detection level in an epithelial cell line (25). In contrast, our current study shows that both Cfm1 and Cfm2 proteins cooperatively function in chondrogenesis and IVD formation in vivo, therefore, loss of the Cfm genes perturb cartilaginous cell development. Further studies are necessary to determine the exact roles of Flnb, Smads and Cfms in the TGF-b pathway in skeletal development.
Cfm1 protein was implicated in the regulation of the perinuclear actin network and nuclear shape through interaction with Flna from a study using cultured mouse NIH 3T3 fibroblasts and human astrocytoma U373 cells (25,27,33). In the current study, the primary chondrocytes isolated from Cfm DKO mice showed fewer actin filament bundles, a decrease of the cell surface area, and shortening of the nuclear long axis compared with those of the primary chondrocytes isolated from wild-type mice. The data presented in this study and previous reports raise the possibility that Cfm proteins associate with actin filaments, and loss of Cfm proteins in cells reduces the level of filamentous actin in chondrocytes. Our results suggest that Cfm proteins
We observed abnormal skeletogenesis in Cfm DKO mice. How does loss of Cfm genes affect the development of the skeletal system? We observed that TUNEL-positive cells increased in the cartilaginous cells including NP cells, AF cells and chondrocytes in the Cfm DKO mice. These results are reminiscent of the increasing chondrocyte apoptosis in Flnb-deficient mice (16). In addition, we found that primary chondrocytes in Cfm DKO mice exhibit decreased proliferation rate. Our results suggest that the balance of proliferation and apoptosis was disrupted, followed by the perturbation of cartilaginous cell development in Cfm DKO mice.
Our results showed that in the absence of Cfm1 and Cfm2, the stabilization of actin networks, the regulation of cartilaginous cell proliferation and survival are impaired, and cartilaginous skeletal formation is retarded. Thus, both Cfm1 and Cfm2 play critical roles in the molecular function of Filamin complex formation and are essential for actin cytoskeleton formation in developing cartilaginous cells.
Deficiency or mutation of the Lnma gene encoding the nuclear lamin A/C leads to absence of an actin cap which shapes the nucleus of interphase cells. Lamin-deficient cells showed few actin filament bundles that can produce protrusions (34,35). Moreover, lamin A/C is connected to the cytoskeleton through linkers of the nucleoskeleton to the cytoskeleton (LINC) complexes (36,37), and potentially link the actin cytoskeleton to lamin A (38). Lnma mutant mice displayed skeletal defects consistent with Hutchinson-Gilford progeria syndrome (OMIM:#176670), including a marked reduction in growth rate, osteoporosis and kyphosis (39)(40)(41). These malformations are similar to Cfm DKO mice defects. Gay et al. proposed a hypothetical model of actin perinuclear structure stabilization by the Cfm/Flna/Actin/LINC complex (27). Our results suggest that the Cfm1 and Cfm2/Filamins/Actin/LINC/lamins complex plays an important role in the formation of nucleoskeleton during cartilaginous tissue formation and chondrogenesis. Thus, our current study elucidates the molecular mechanisms of Filamins in skeletal development in vivo. In future studies, it will be important to assess how Cfm proteins coordinate and maintain the actin-binding partners.

Generation of Cfm2 mutant mice
We obtained Cfm2 genomic DNA clones from a screen of the 129S6 (129/SvEv Taconic) mouse genomic DNA library (Stratagene). We subcloned a 7.5 kb SphI-EcoRI fragment and a 6.0 kb SpeI-EcoRI fragment from Cfm2 genomic clones into a modified pPNT vector (44,45), and transfected the linearized targeting construct into the 129S6 embryonic stem cell line. The culture, electroporation and selection of 129S6 were carried out as previously described (46). Embryonic stem cells that were heterozygous for the targeted gene disruption were microinjected into C57BL/6 blastocysts to obtain chimeric mice.

Anti-Cfm2 antibody production
By using PCR, a cDNA encoding the N-terminal portion (residues 1 -110) of mouse Cfm2 (N-Cfm2) was amplified and subcloned into pGEX4T-1 (GE Healthcare). The fusion was expressed in Escherichia coli strain DH5a and purified with glutathione Sepharose 4B (GE Healthcare) according to the manufacturer's instructions. An antibody against N-Cfm2 was obtained by immunizing rabbits with the purified GST-N-Cfm2. The rabbit antisera against N-Cfm2 was pre-absorbed with GST-Sepharose and affinity purified with an immunizing fusion protein-bound Sepharose column.

Histological analysis
To prepare sections for toluidine blue (Sigma-Aldrich) and safranin-O (Nacalai Tesque) staining, we fixed mice with 4% paraformaldehyde in PBS buffer, decalcified vertebral columns and Human Molecular Genetics, 2014, Vol. 23, No. 11 2963 limbs in Morse's solution (47) and embedded in Tissue-Tec Optimum Cutting Temperature. Frozen sections (18-25 mm thick) were stained with toluidine blue, and subjected to histochemical analysis. Analysis of the growth plates was performed as described by previously (48). To prepare sections of the brain, heart and kidney, we fixed the tissues with 4% paraformaldehyde in PBS buffer and embedded in Tissue-Tec Optimum Cutting Temperature. Frozen sections (10 mm thick) were stained with toluidine blue and subjected to histochemical analysis.

Staining of bone and cartilage
After euthanasia, mice were skinned, and all tissues were eviscerated. Mice were fixed in 95% ethanol for 3 -5 days, and then stained with 0.15% Alcian blue (Sigma A3157) dissolved in glacial acid plus 75% ethanol and with 0.5% Alizarin red (Sigma A5533) dissolved in 2% KOH.

mCT analysis
Quantitative mCT analysis was conducted as described previously (43). Briefly, Quantitative mCT was performed with an mCT system (mCT-40; ScancoMedical). CT scans were performed at the distal metaphysis to calculate trabecular parameters, and at mid-shaft to calculate cortical thickness in 8-week-old mice. The craniocaudal scan lengths were 1.2 mm in 8-week-old mice in the distal metaphysis, and the scan length was 0.24 mm at the mid-shaft. The analysis of the vertebral columns was performed with an RmCT system (Rigaku). The exposure parameters were 90 kV and 100 mA, or 90 kV and 150 mA.

TUNEL assay
Frozen vertebral columns were sectioned to a thickness of 20-30 mm and fixed with 4% paraformaldehyde in PBS for 15 min. The TUNEL assay was performed using Click -iT w TUNEL Alexa Fluor w imaging Assay (Invitrogen) according to the manufacturer's protocol.

Primary cultured chondrocytes preparation and proliferation assay
Primary cultured chondrocytes were isolated from growth plates of the rib cartilage from P24 wild-type mice and Cfm DKO mice using a modification of the previously published protocol (50). Briefly, chondrocytes were isolated using 0.2% collagenase P (Roche). Isolated cells were maintained in RPI-1640 (Sigma-Aldrich) supplemented with ascorbic acid (50 mg/ml) and 10% FBS. We seeded them in a 12-well plate at a density of 8 × 10 2 cells per well for the proliferation assay. At each time point, cells were harvested and counted. Duplicate measurements were performed on three-independent wells for each time point.

Yeast two-hybrid screening
Yeast two-hybrid screening was performed as described previously (51). Briefly, the ORF of the mouse Cfm2 cDNA (full length) was inserted into the pGBKT7 bait vector and transformed into the AH109 yeast strain. We screened 3.5 × 10 6 transformants from a mouse E11 cDNA library (Takara Bio) according to manufacturer's instructions for the Matchmaker System 3 (Takara Bio).

Plasmid constructs
A plasmid-encoding full-length Cfm2 was amplified using a Fam101a clone (3110032G18Rik; GenBank accession No. NM028443) as a template and inserted into the N-terminal FLAG-tagged pCAGGSII expression vector to produce pCAG-FLAG-Cfm2. A plasmid-encoding full-length Cfm1 was amplified using a Fam101b clone (1500005K14Rik; GenBank accession No. NM029658) as a template and inserted into the N-terminal FLAG-tagged pCAGGSII expression vector to produce pCAG-FLAG-Cfm1. A plasmid-encoding fulllength Flna was amplified using an Flna clone (IMAGE clone 9087937; GenBank accession No. BC151024) as a template and inserted into the N-terminal HA-tagged pCAGGSII expression vector to produce pCAG-HA-Flna. The DNA sequences were confirmed using an ABI310 genetic analyzer.

Cell culture and transfection
ATDC5 cells were maintained in F-12 with 5% FBS, 10 mg/ml human transferrin (Sigma-Aldrich) and 3 × 10 28 M sodium selenite (Sigma-Aldrich). Transfection of plasmid DNA was performed using Lipofectamine LTX (Invitrogen) for ATDC5 cells according to the manufacturer's instruction. Cells were cultured for 24 h. For immunostaining, cells were washed with PBS, fixed with 4% paraformaldehyde in PBS for 5 min at room temperature, and subsequently incubated with blocking solution for 1 h. Cells were immunostained with a primary antibody in blocking solution for 2 h at room temperature and subsequently incubated with a secondary antibody solution for 1 h at room temperature. Alexa Fluor 488 (1:500) or Cy3 (1:400)-conjugated IgG (Jackson ImmunoResearch Laboratories) were used as secondary antibodies.

2964
Human Molecular Genetics, 2014, Vol. 23, No. 11 Anti-Cfm1 antibody production An antiserum against the N-terminal portion (residues 39-142) of mouse Cfm1 (N-Cfm1) was raised by immunizing guinea pig with purified GST-N-Cfm1, then, the antisera were reacted with the fusion protein-coupled CNBr-activated Sepharose 4B (GE Healthcare). This sepharose 4B was washed with PBS and eluted with 0.1 M glycine buffer (pH 2.5) to obtain a purified antibody against N-Cfm1. The eluted antibody was neutralized by 20× PBS, and further dialyzed in PBS at 48C.

Immunoprecipitation and western blotting
For the immunoprecipitation and immunoblotting for the interaction between endogenous Cfm and Filamins proteins, protein samples from E18.5 chondrocytes in vertebral bodies were lysed in a lysis buffer (50 mM Tris-HCl at pH 7.5, 150 mM NaCl, 2 mM MgCl 2 , 1 mM EDTA, 10% glycerol, 1 mM PMSF, 1% NP40, one protease inhibitor cocktail tablet (Roche)). Lysates and antibodies were incubated for 16 h with protein G-Sepharose (GE Healthcare), washed four times with wash buffer (50 mM Tris-HCl at pH 7.5, 150 mM NaCl, 2 mM MgCl 2 , 1 mM EDTA, 10% glycerol, 1 mM PMSF, 1% NP40) and resolved by SDS -PAGE. Western blot analysis was performed using a semidry transfer cell (iBlot system; Invitrogen) with iBlot Gel Transfer Stack PVDF (Invitrogen). Signals were detected using ECL Plus Western Blotting Detection System (GE Healthcare). For the immunoprecipitation and immunoblot analysis of the interaction between endogenous Smad3 and Flnb proteins without Cfm proteins, primary chondrocytes from rib cages of wild-type and Cfm DKO mice at P3 were isolated using 0.2% collagenase P (Roche) and were seeded in a 6 cm dish. Analysis and detection procedures were mentioned above.

Morphometric measurement and immunofluorescence microscopy
Primary cultured chondrocytes were isolated from growth plates of the rib cartilage of wild-type and Cfm DKO mice at 8 weeks. Cells were spread on a 3.5 cm dish and incubated for 48 h. Immunocytostaining was performed as described above. To visualize actin filaments, cells were incubated with phalloidin Rhodamine (Cytoskeleton, Inc). DAPI (Sigma-Aldrich) was applied to stain nuclei. The specimens were observed under a laser confocal microscope (LSM710, Carl Zeiss). Cell surface area and nuclear long axis were measured using AxioVision rel.4.8 (Carl Zeiss).

Migration assay
Primary fibroblasts were isolated from Cfm DKO and wild-type embryos at E14.5. Migration assays were performed on cells in passages 2 -3. Cell migration was examined using BD Falcon Cell Culture Insert containing polycarbonate filters with 8 mm pore size (BD Sciences), essentially as described (15). 3 × 10 4 cells were seeded in the upper chamber in 2% FBS DMEM, and the chamber was lowered into a well containing 10% FBS DMEM, used as a chemoattractant. After 4 h of incubation at 378C, the filters were collected and cells migrating to the lower surface were fixed in methanol, stained with Giemsa's azure eosin methylene blue solution (MERCK) and counted by using a light microscope.

Statistics
Statistical analysis was performed using Student's t-test for comparisons between two groups and analysis of variance with Tukey-Kramer test for comparisons among three groups, unless otherwise described. All data are expressed mean + SEM.

Study approval
All procedures were approved by the Institutional Safety Committee on Recombinant DNA Experiments and the Animal Research Committee of Osaka Bioscience Institute, and by the Recombinant DNA Committee (3380-3) and the Animal Care Committee (24-05-1) of Osaka University.