Smad4 deficiency impairs chondrocyte hypertrophy via the Runx2 transcription factor in mouse skeletal development

Chondrocyte hypertrophy is the terminal step in chondrocyte differentiation and is crucial for endochondral bone formation. How signaling pathways regulate chondrocyte hypertrophic differentiation remains incompletely understood. In this study, using a Tbx18:Cre (Tbx18Cre/+) gene-deletion approach, we selectively deleted the gene for the signaling protein SMAD family member 4 (Smad4f/f) in the limbs of mice. We found that the Smad4-deficient mice develop a prominent shortened limb, with decreased expression of chondrocyte differentiation markers, including Col2a1 and Acan, in the humerus at mid-to-late gestation. The most striking defects in these mice were the absence of stylopod elements and failure of chondrocyte hypertrophy in the humerus. Moreover, expression levels of the chondrocyte hypertrophy–related markers Col10a1 and Panx3 were significantly decreased. Of note, we also observed that the expression of runt-related transcription factor 2 (Runx2), a critical mediator of chondrocyte hypertrophy, was also down-regulated in Smad4-deficient limbs. To determine how the skeletal defects arose in the mouse mutants, we performed RNA-Seq with ChIP-Seq analyses and found that Smad4 directly binds to regulatory elements in the Runx2 promoter. Our results suggest a new mechanism whereby Smad4 controls chondrocyte hypertrophy by up-regulating Runx2 expression during skeletal development. The regulatory mechanism involving Smad4-mediated Runx2 activation uncovered here provides critical insights into bone development and pathogenesis of chondrodysplasia.

Human chondrodysplasia is a heritable disorder affecting skeletal development, characterized by skeletal abnormalities with reduced length of the long bone and short stature. Several genetic mutations causing chondrodysplasia have been identified including type X collagen (Col10a1), runt-related transcription factor 2 (Runx2), 3 and cartilage-derived morphogenetic protein 1 (CDMP1) (1)(2)(3)(4). These genetic mutations affect chondrogenesis crucial for endochondral bone development. Understanding the molecular pathways underlying chondrogenesis during skeletal development may provide crucial insights into the therapeutic treatments for chondrodysplasia in humans.
Endochondral ossification is an imperative developmental process for skeletal formation in vertebrates. Endochondral bone development is initiated by aggregation and condensation of mesenchymal cells. The mesenchymal cells further differentiate into primary chondrocytes with secreting type II collagen (Col2a1) and aggrecan (Acan) to form cartilage elements. Primary chondrocytes then proliferate and develop into hypertrophic chondrocytes that secrete Col10a1. Subsequently, hypertrophic chondrocytes are invaded by blood vessels and are replaced by bone and bone marrow (5)(6)(7)(8). In this process, chondrocyte hypertrophy is a terminal step of chondrocyte differentiation, and is essential for endochondral bone formation. A few essential molecules that control chondrocyte hypertrophic differentiation have been found (9). Indian hedgehog (Ihh), a member of the hedgehog family, is expressed in prehypertrophic chondrocytes. Ihh is required for chondrocyte proliferation and chondrocyte hypertrophy (10,11). Bone morphogenetic protein (BMP) family members are expressed throughout limb development and play vital roles in regulating chondrogenesis (12,13). Deletion of Bmp2 results in defective chondrocyte hypertrophy and endochondral bone formation, indicating Bmp2 is required for chondrocyte hypertrophic differentiation and endochondral bone formation (14). In addi-This work was supported by National Institutes of Health Grants 1R01HL131735, 1R01HL095810, 1R56HL129807, and 1K02HL094688, American Heart Association Grants 15GRNT25710153 and 0855808D, and March of Dimes Foundation Grant 5-FY07-642 (to C. L. C.). The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This article contains Figs. S1 and S2 and Tables S1-S25. The ChIP-Seq and RNA-Seq data can be accessed through the NCBI Gene Expression Omnibus (GEO) Repository using accession numbers GSE114081 and GSE114079, respectively. 1  tion, Runx transcription factors Runx2 and Runx3 also regulate and coordinate chondrocyte hypertrophic differentiation and proliferation (15,16). Smad4 is a central mediator of canonical transforming growth factor (TGF)/BMP signaling (17) and it binds to receptor-activated Smads (R-Smads) and forms a complex with R-Smads. The Smad complex translocates into the nucleus and serves as a transcription factor to regulate the expression of targeting genes involved in a wide variety of development and homeostasis processes. During mouse embryogenesis, Smad4 is expressed in the developing limbs and is required for early mesenchymal cell aggregation, chondrocyte differentiation, osteoblast differentiation, and maturation (18 -21). The regulatory mechanism of the Smad4 underlying chondrocyte hypetrophic differentiation during long bone development is unknown, although studies on TGF/BMP pathway genes showed that Bmp2 regulates Runx2 expression through Smad in the C2C12 cell line (22), and Runx2 physically interacts with Smads in COS7 cells in vitro (23). In this study, we performed a genetic study in mice and found that loss of Smad4 within the limb resulted in the absence of chondrocyte hypertrophy and ossification in humerus of mice. In addition, Smad4 promotes expression of hypertrophy differentiation genes, including Runx2, Runx3, and Ihh. Further RNA-seq and ChIP-seq analyses revealed that Smad4 regulates a wide spectrum of genes associated with limb development, including direct activation of Runx2. These findings reveal an important mechanism by which Smad4 regulates chondrocyte hypertrophy during mouse skeletal development.
To delete Smad4 in the chondrocytes, Tbx18 Cre/ϩ mice were crossed to Smad4-flox (Smad4 f/f ) (27) to obtain Tbx18 Cre/ϩ ; Smad4 f/ϩ compound heterozygous animals. They were further crossed to Smad4 f/f mice to generate Tbx18 Cre/ϩ ;Smad4 f/f mutants (denoted as Smad4 CKO). We found mice with the Tbx18 Cre/ϩ ;Smad4 f/ϩ genotype are normal and therefore utilized as controls in this study. Immunostaining revealed that Smad4 is ubiquitously expressed in the WT and Tbx18 Cre/ϩ ; Smad4 f/ϩ control limb at E11.5-13.5 (Fig. 1A, data not shown for WT). In Smad4 CKO embryos, almost no Smad4 expression was detected in the chondrocytes and adjacent perichondrial cells (Fig. 1B), suggesting that Smad4 is removed in these regions. The mutant mice died shortly after birth with severe ureteral development defects (24). At E14.5, they displayed a prominently shortened limb (arrows in Fig. 1, C and D), and the defect became severe at E16.5 and E18.5 (Fig. 1, E-J). Further skeletal analysis by Alizarin red and Alcian blue staining showed that the limb shortening was mainly due to the absence of stylopod elements (humerus and femur, arrows in Fig. 1, K-R) that failed in ossification at E15.5-18.5, indicating the defect was associated with abnormal chondrocyte differentiation. Additionally, zeugopod elements of the mutant limb (radius/ulna and tibia/fibula, arrowheads in Fig. 1, K-R, Fig. S2) were shorter than the control littermates. Because the stylopod elements were the most severely affected in both forelimbs and hindlimbs, we focused on the stylopod elements (humerus) in the forelimbs.

Loss of Smad4 causes chondrodysplasia in limb development
To investigate the cause of the shortened limbs in Smad4 CKO mice, we examined younger stage embryos from E10.5 to determine when the defects first appear. No apparent abnormalities were observed at E12.5 or earlier stages in the mutant (data not shown), except that the cartilage element was slightly shorter and smaller at E12.5 (Fig. 2, A and B). The difference became significant at E13.5 (Fig. 2, C and D), and was severe at E14.5 (Fig. 2, E and F). We analyzed the expression of chondrocyte differentiation markers including Lect1, Col2a1, and Acan (5,28,29) to determine whether chondrocyte differentiation was affected in the limbs at E12.5-13.5. Compared with control humerus, lower expression levels of Lect1 and Col2a1 were detected in the mutant humerus by RNA in situ hybridization (Fig. 2, G-L). Quantitative RT-PCR (qRT-PCR) confirmed that Lect1, Col2a1, and Acan mRNA expression levels were significantly lower by 50, 55, and 58% at E12.5, and by 63, 60, and 65% in the mutant forelimbs at E13.5, respectively (Fig. 2, M-O). The down-regulated Lect1, Col2a1, and Acan expression suggests that Smad4 is essential for chondrocyte differentiation.

Smad4 is essential for chondrocyte hypertrophy
Failure of endochondral ossification in stylopod elements is the most striking defect in Smad4 CKO limbs at E15.5 (arrows in Fig. 1, K and L). Chondrocyte hypertrophy is a crucial developmental step for endochondral ossification. Lack of hypertrophic chondrocytes could cause failure of endochondral ossification (5). We then try to determine whether chondrocyte hypertrophy is perturbed in Smad4 CKO limb. For Tbx18 Cre/ϩ ; Smad4 f/ϩ and WT limb at E13.5. H&E staining showed that chondrocytes were present in the center of humerus with hypertrophic differentiation (Fig. 3A). These hypertrophic characteristics were unable to be detected in the Smad4 CKO humerus (Fig. 3B). At E14.5, typical hypertrophic cells were present in the control humerus (Fig. 3C), and they were absent in the mutants (Fig. 3D). Alcian blue staining further confirmed the absence of hypertrophic chondrocytes in the mutant humerus (Fig. 3, E-H). We next performed RNA in situ hybridization to detect Col10a1, a specific marker for hypertrophic chondrocytes (30,31), and Panx3, a marker for prehypertrophic and hypertrophic chondrocytes (32,33). Both Col10a1 and

Smad4 in chondrocyte hypertrophy
Panx3 were found in the center of humerus as early as E12.5 in the control (Fig. 3, I and O), but were undetectable in the mutant (Fig. 3, J and P). At E13.5 and E14.5, robust Col10a1 and Panx3 expression was detected in the hypertrophic or prehypertrophic region on the control mice (Fig. 3, K, M, Q, and S), and they were not seen in the mutant humerus (Fig. 3, L, N, R, and T). Further qRT-PCR analysis showed that Col10a1 and Panx3 mRNA expression was decreased by 70 and 90% at E12.5, and by 76 and 87% at E13.5 in the mutant forelimbs, respectively (Fig. 3, U and V). Taken together, these observations indicate that the primary chondrocytes of Smad4 CKO humerus failed to undergo hypertrophic differentiation, and Smad4 is critical for chondrocyte hypertrophy development during limb formation.

Smad4 deficiency results in decreased chondrocyte proliferation
Chondrocyte proliferation is important for cartilage growth prior to hypertrophic differentiation. Given the small humerus in the mutant limb at E12.5 and onward, we hypothesized that Smad4 signals may be essential for chondrocyte growth. To confirm this, EdU was injected into pregnant mice intraperitoneally at E12.5 and E13.5, and forelimbs were collected after 4 h of injection for analysis. The proliferating cells in the cartilage were analyzed by immunofluorescence to determine whether chondrocyte proliferation was affected in the mutants. The proliferation rate was calculated by the ratio of EdU/Sox9 double positive cells divided by the total of chondrocyte cells (Sox9 positive) in humerus. It showed that the mutant chondrocyte proliferation was significantly reduced more than that of the control, with 18.4% in the control versus 14.2% in the mutant at E12.5, and 17.8% in the control versus 13.1% in the mutant at E13.5 (Fig. 4). These results uncover that Smad4 signaling is required for chondrocyte proliferation.

Smad4 is required for Runx2 expression in the limb
To further elucidate the molecular mechanism by which Smad4 deficiency impairs chondrocyte hypertrophy, we per- A3 and B3 are overlay images for A1/2 and B1/2, respectively. A4 and B4 are high magnification images in the areas outlined in A3 and B3. C-J, whole mount view of Smad4 CKO mice at E14. 5-18.5. Arrows indicate control forelimb (C, E, and G) and hindlimb (I), or mutant forelimb (D, F, and H) and hindlimb (J). K-R, skeletons of E15.5-18.5 embryos stained with Alizarin red (calcified tissue) and Alcian blue (cartilage). K, M, O, and Q are controls and L, N, P, and R are mutants. In K-P, arrows, arrowheads, and unnotched arrows indicate humerus, radius, and ulna, respectively. In Q and R, arrows, arrowheads, and unnotched arrows indicate femur, fibula, and tibia, respectively.

Smad4 in chondrocyte hypertrophy
formed high-throughput RNA-seq to examine genes with altered expression when Smad4 is removed. Because Smad4 CKO limb defects were first detected at E12.5, and chondrocytes start the hypertrophic differentiation process at E13.5 and are largely accomplished at E14.5 (Fig. 3, A, C, E, and G), we isolated RNA at E12.5, E13.5, and E14.5 from forelimbs of mutants and controls for large-scale sequencing. Furthermore, to examine whether any molecular changes occur before E12.5, we also performed RNA-seq with E11.5 forelimb samples although no morphological abnormalities were detected.
Approximately 30 million raw reads per sample were obtained and the reads with good quality were aligned to mouse reference sequence (RefSeq) transcripts using BWA (34). In Smad4 CKO mice, 551 genes at E12.5 (Table S3) and 622 genes at E13.5 (Table S4) were down-regulated (p Ͻ 0.0001) using the DEGseq program (35), and the number increased to 2581 at E14.5 (Table S5) (Fig. 5A). As parallel, the number of up-regulated genes (p Ͻ 0.0001) was increased from 651, 775 to 3249 at E12.5 (Table S3), E13.5 (Table S4), and E14.5 (Table S5), respectively (Fig. 5B). At E11.5, very few genes (less than 10) were changed (Tables S7 and S8) and they seemed not associated with chondrocyte/limb development according to Gene Ontology (GO) analysis using DAVID 6.7 (36). At E12.5-14.5, the number of significantly changed genes appeared to reflect the degree of abnormalities between the mutant and control at corresponding stages (Fig. 2, A-F). A larger number of the overlapped genes were found between E13.5 and E14.5 (Tables S11 and S12) than E12.5 and E14.5 (Tables S13 and S14) or E12.5 and E13.5 (Tables S9 and S10) (Fig. 5, A and B). In addition, 123 genes were down-regulated (Table S15, Fig. 5A) and 30 genes were up-regulated (Table S16, Fig. 5B) at all three stages. At E12.5, the down-regulated genes are highly associated with skeletal, cartilage, and bone development (asterisks in Fig. 5C, and Tables S17 and S18). At E13.5, skeletal, cartilage, and limb/ bone development are still on the top lists of term according to GO (asterisks in Fig. 5D, Tables S19 and S20). It is of importance to note that although many genes were changed at E14.5, the limb/bone, skeletal, and cartilage development are not the primary terms (Tables S21 and S22). This may suggest that the genes changed at E14.5 are secondary or may not be directly regulated by Smad4.

Smad4 directly binds to Runx2 regulatory elements
RNA-seq analysis of Smad4 CKO forelimb revealed that most genes remain normal at E11.5, and the genes with altered expression at E14.5 are likely secondary and may not reflect direct regulation by Smad4 during limb development. We next performed ChIP-seq with E12.5 and E13.5 forelimbs to determine Smad4 direct downstream targets in chondrocyte hypertrophy development. Chromatin DNA was prepared from the WT forelimb tissues at E12.5 and E13.5 using anti-Smad4 antibody. The purified DNA was amplified for library construction and large-scale sequencing. Smad4 genome-wide occupancy was identified using MACS1.4 as described (41). Peaks within 5

Smad4 in chondrocyte hypertrophy
kb of the transcription start site (TSS) were assigned to the nearest genes. Regulatory regions in 1213 genes were identified and potentially bound by Smad4 (Table S6). We performed GO analysis on these genes and found that regulation of chondrocyte differentiation was on the top of the list (asterisk in Fig. 7B, Table S23).
To determine whether Smad4 directly regulated genes are associated with chondrocyte differentiation, we combined RNA-seq and ChIP-seq data (E12.5-13.5) for further integrative analysis. It revealed that 6 down-regulated genes, including Runx2 and Fgfr3, are related to cartilage development (Fig. 7A, see full list of the genes in Table S24). In contrast, the up-regulated 4 genes seem not to be associated with chondrocyte dif-ferentiation (Fig. 7A, Table S25). Consensus core sequences 5Ј-GTCT-3Ј were called and are consistent with previously reported Smad-binding elements (42,43). Because the Smad4 binding motif is already known in HOMER database (http:// homer.ucsd.edu/homer/motif/motifDatabase.html) 4 (Fig. 7, C-1), we applied HOMER motif discovery software (44) to screen for all enriched regions, and found the Smad4-binding motif (5Ј-TTTTGTCTGC-3Ј) is in the list of enrichment peaks (Fig. 7, C-2). The Smad4-enrichment peak in Runx2 is ϳ4 kb upstream of TSS (two Smad-binding elements within the

Smad4 in chondrocyte hypertrophy
enriched peak), and the motif is 76 bp away from the peak midpoint. The Integrative Genomics Viewer (45) peak upstream of Runx2 TSS was shown in Fig. 7C. To verify direct binding of the Runx2 regulatory element by Smad4, we further carried out ChIP-PCR around this region with chromatin DNA immunoprecipitated by Smad4 antibody in mouse forelimbs (E12.5-13.5), with DNA ChIPed by IgG as control. The Runx2 regulatory region containing the Smad4-binding site was amplified by PCR (Fig. 7D), confirming that Smad4 directly binds to this Runx2 regulatory element during mouse skeletal development at E12.5-13.5. Furthermore, luciferase reporter assays revealed that Smad4 activates the Runx2 promoter fragment containing the Smad4-binding site (5.0 kb), but not fragments without this site (4.2 and 3.0 kb) (Fig. 7E). Taken together, these data suggest that Smad4 directly binds to Runx2 regulatory elements during mouse limb development.

Discussion
Chondrocyte hypertrophic differentiation is a precisely regulated morphogenetic process in which the chondrocytes differentiate into hypertrophic chondrocytes during skeletal development (Fig. 8). Smad4 has been shown to play important roles in the process of chondrocyte differentiation from mesenchymal cells and chondrocyte proliferation (19,20). In this study, we showed that Smad4 is essential for chondrocyte hypertrophic differentiation. Smad4 CKO mice display limb growth defects with chondrodysplasia characterized by low chondrocyte proliferation and lack of hypertrophy in humerus. We further demonstrated that Smad4 acts as a direct transcriptional activator for Runx2 in chondrocyte differentiation.

Smad4 regulates chondrocyte differentiation and hypertrophy
Previous studies showed that ablation of Smad4 with Prx1-Cre or Col2a1-Cre deleter mice resulted in defective limb formation (19,20). Specifically, Prx1-Cre;Smad4 f/f mutant mice displayed a complete loss of limb skeletal elements at E14.5 due to perturbation of mesenchymal cell aggregation, the first step of chondrocyte differentiation (20). Col2a1-Cre;Smad4 f/f mutant mice displayed a mild shortened limb at postnatal day 10 with defective proliferation and differentiation of chondrocytes (19). In contrast to defects of these two mutants, limb shortening of Tbx18 Cre/ϩ ;Smad4 f/f mice resulted from the absence of stylopod elements because of chondrocyte hypertrophy failure, the terminal step of chondrocyte differentiation. Phenotype variations of these mutants could be explained by location and timing of distinct Cre-mediated genetic deletions in the developing limbs: Prx1-Cre is expressed in the early limb mesenchyme at E9.5 and it introduces genetic deletions in both chondrocytes and perichondrium (46). Col2a1-Cre is expressed in the differentiating chondrocytes in limbs (47). Tbx18 Cre/ϩ is detected in the limb bud as early as E9.5 and mediates recombination in chondrocytes within the proximal domain of the limb (Fig. S1, C-F). The divergent defects of Smad4 mutants suggest multiple and temporospatially distinct functions of Smad4 in limb formation.
Chondrocyte hypertrophy is critical for endochondral ossification. Disruption of Smad4 in Tbx18 Cre/ϩ lineage results in complete loss of chondrocyte hypertrophy (Fig. 3, A-H) and failure of endochondral ossification in humerus (Fig. 1, K and  L), suggesting an essential role of Smad4 for chondrocyte hypertrophy during long bone development. This notion was supported by the observation of the absent hypertrophic chondrocytes and missing expression of Col10a1 (specific marker for hypertrophic chondrocytes (30,31)) and Panx3 (specific marker for pre-hypertrophic chondrocytes and hypertrophic chondrocytes (32,33)) in Smad4 CKO humerus (Fig. 3, I-V).

Runx2 acts as a crucial mediator of Smad4 signaling in chondrocyte hypertrophy
In this study, we applied RNA-seq and ChIP-seq to decipher how Smad4 controls chondrocyte hypertrophy during mouse limb skeletogenesis. GO analysis of the RNA-seq data showed the down-regulated genes are highly associated with skeletal, cartilage, and bone development at E12.5 and E13.5 (Fig. 5, C  and D). RNA-seq revealed that the down-regulated genes at
The integrated RNA-seq and ChIP-seq data analyses identified potential targets of Smad4. Six down-regulated genes highly related to bone formation, including Runx2 and Fgfr3, were uncovered (55, 61). Fgfr3 is a negative regulator for bone growth (61). Fgfr3 was shown to restrain chondrocyte proliferation and limit osteogenesis. Fgfr3 Ϫ/Ϫ mutant mice displayed prolonged endochondral bone growth accompanied by expansion of proliferating and hypertrophic chondrocytes in the cartilaginous growth plate. Given that Fgfr3 controls endochondral ossification by a negative regulatory mechanism, and Fgfr3 Ϫ/Ϫ bones display prolonged defects compared with

Smad4 in chondrocyte hypertrophy
Smad4 CKO mice (Fig. 1, K-R), the down-regulated Fgfr3 may not be the main cause of Smad4 CKO limb defects, although Smad4/Fgfr3 signaling cascade could play some unknown essential roles for limb skeletal development. In contrast, Runx2 positively regulates chondrocyte hypertrophy. Runx2 overexpression in chondrocytes restores chondrocyte hypertrophy in Runx2-deficient mice (56,57). Taken together, our data support that Runx2 acts as a key downstream gene of Smad4 given the critical role of Runx2 in the skeletal hypertrophic development.
Runx2 has been suggested as a common downstream gene of TGF-␤/BMP signaling in previous studies, although the mechanisms underlying this process were largely unknown before (22,62). In this regard, our study first demonstrated that Smad4 transcriptionally activates Runx2 expression by directly binding to Runx2 regulatory elements during chondrocyte hypertrophy, suggesting an important role of Runx2 in cartilage growth and bone formation.

Smad4 is required for chondrocyte proliferation
Chondrocyte proliferation is essential for cartilage growth (11,63,64). BMP signaling was shown to be important for chondrocyte proliferation (14,65). Reduced osteoblast and chondrocyte proliferation was detected in Smad4 mutant mice in the previous studies (18,19). Here we found that ablation of Smad4 with Tbx18 Cre/ϩ also led to a dramatic reduction in chondrocyte proliferation as early as E12.5 in the forelimb (humerus). The reduced chondrocyte proliferation should contribute to the short cartilage elements in Smad4 CKO limbs (Fig. 2, A-F). Smad4 promotes Ihh expression (Figs. 5E and 6,  E-H and Q), and Ihh plays a critical role in regulating chondrocyte proliferation during skeletal development (11,64,66). Ihhdeficient mice display severe reduction in skeletal growth and decreased chondrocyte proliferation (11). Ectopic activation of Ihh signaling promotes chondrocyte proliferation during cartilage development (64). Given that Ihh was significantly downregulated in Smad4 CKO humerus, it is potential that Ihh is one of the key downstream effectors of Smad4 signaling pathways in regulating chondrocyte proliferation.
In summary, our study uncovered divergent, crucial roles of Smad4 in chondrocyte differentiation, especially in hypertrophic differentiation during limb development. We demonstrated that Smad4 controls chondrocyte hypertrophic differentiation through direct regulation of Runx2. Ablation of Smad4 within the proximal domain of limbs results in chondrodysplasia due to reduced chondrocyte proliferation, impaired chondrocyte differentiation, lack of chondrocyte hypertrophy, and failure of ossification in humerus. As the study did not use hypertrophic chondrocyte-specific Cre (e.g. Col10a-Cre) to delete Smad4 from hypertrophic chondrocytes in the limb, it is uncertain whether the impaired chondrogenesis also affects chondrocyte hypertrophy in Smad4 CKO mice. ChIP-seq assay uncovered a previously unknown critical role of Runx2 as a direct downstream target of Smad4 signaling in cartilage and bone formation. The Smad4 -Runx2 regulatory pathways provide important mechanisms to understand hypertrophic differentiation of limb development and etiology of chondrodysplasia in mammals.

Skeletal analysis
Calcified tissue was stained with Alizarin red (Sigma) and cartilage was stained with Alcian blue (Sigma) with a standard procedure (68). For E15.5-18.5 embryos, the skin and internal organs were removed. Skeletons were fixed in 95% ethanol overnight and stained with 0.015% Alcian blue followed by digestion in 1% potassium chloride for 2 days, and then stained

Smad4 in chondrocyte hypertrophy
with 0.005% Alizarin red for 3 days. E12.5-14.5 embryos were fixed in ethanol overnight, stained with 0.009% Alcian blue for 3 days, and cleared by benzyl alcohol/benzyl benzoate. Images of the stained skeletons and cartilages were taken under a Leica stereomicroscope.

X-Gal staining
Embryos were fixed with 4% paraformaldehyde (PFA) for 30 min at 4°C. The fixed embryos were washed twice with PBS and then stained with X-Gal staining solution (5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, 2 mM MgCl 2 , 1 mg/ml X-Gal) for 12 h at room temperature.

H&E staining and Alcian blue staining
Mouse embryos were fixed with 4% PFA overnight at 4°C and then dehydrated in an ascending ethanol series followed by two changes of 100% xylene. The tissues were embedded in liquid paraffin and placed on a cold plate for solidification. Sections were cut into 6-m thickness on a microtome and stained with hematoxylin and eosin (H&E). For Alcian blue staining, the sections were stained in 1% Alcian blue solution (in 3% acetic acid) for 20 min and counterstained in nuclear fast red solution (Eng Scientific) for 10 min.

Cell proliferation assay
Click-iT EdU Cell Proliferation Assay Kit (Invitrogen) was used for the cell proliferation assay according to the manufac-turer's instructions. Pregnant mice were intraperitoneally injected with 10 mM EdU in PBS (5 mg/100 g body weight). Embryos were collected 4 h after EdU injection and fixed with 4% PFA for 30 min at 4°C and washed with PBS twice. Samples were embedded in Optimal Cutting Temperature compound and sectioned into 6 m pieces. Frozen sections were incubated with Click-iT Reaction mixture for 30 min at room temperature. Slides were analyzed under a fluorescence microscope.

qRT-PCR
Total RNA was isolated from the forelimbs of mouse embryos using TRIzol reagent (Invitrogen). First strand cDNA was synthesized with QuantiTect Reverse Transcription Kit (Qiagen). qRT-PCR was carried out using the StepOne Plus PCR system (Applied Biosystems) and SYBR Green (Qiagen). The mRNA expression level was normalized to ␤-actin. Statistical analysis was performed by t test and p Ͻ 0.05 was considered significant. Primer sequences for qRT-PCR are listed in Table S2.

RNA-Seq (RNA-seq)
Total RNA was prepared using TRIzol reagent (Invitrogen) and RNA quality was determined using an Agilent Bioanalyser. RNA-seq Sample Preparation Kit (Illumina) was used to prepare mRNA and cDNA following the manufacturer's protocol. RNA-Seq (RNA-seq) was performed as described previously using a HiSeq 2500 (Illumina) (69).

Smad4 in chondrocyte hypertrophy
and was amplified with an Illumina ChIP-seq DNA Sample Prep Kit (IP-102-1001). DNA libraries were sent for sequencing with an Illumina HiSeq 2500 sequencer. ChIP-PCR was performed as described previously (69). Primers used for ChIP-PCR are listed in Table S1.

Luciferase reporter assay
ATDC5 cells were transfected with a mixture of luciferase reporter plasmid containing Runx2 promoter fragments (200 ng), Smad4 expression vector (200 ng), and Renilla plasmid (5 ng) using Lipofectamine 3000 (Invitrogen) according to the manufacturer's instructions. After 48 h of incubation, cells were lysed and luciferase activity was measured using a dual luciferase reporter assay system (Promega) and normalized to Renilla activity. The pGL3-TK vector (Promega) was used as control. Primers used to generate the 3.0-, 4.2-, and 5.0-kb Runx2 promoter fragments are listed in Table S1.