Zbtb20 regulates the terminal differentiation of hypertrophic chondrocytes via repression of Sox9

The terminal differentiation of hypertrophic chondrocytes is essential for endochondral ossification. During endochondral bone development, chondrocytes proliferate and then differentiate into hypertrophic chondrocytes. Hypertrophic chondrocytes express specific extracellular matrix molecules, such as collagen type X (Col10a1) (Elima et al., 1993; Linsenmayer et al., 1991), and normally undergo a further maturation process. Terminal hypertrophic chondrocytes at the chondro-osseous junction actively express additional molecules, such as matrix metalloproteinase 13 (Mmp13) (Stickens et al., 2004), osteopontin (Opn; also known as Spp1) (Chen et al., 1993) and vascular endothelial growth factor (Vegfa) (Gerber et al., 1999), which are crucial for the invasion of blood vessels, osteoclasts and osteoblast precursors from the perichondrium. Defects in the terminal differentiation of hypertrophic chondrocytes result in severely delayed formation of the primary ossification center (POC) and secondary ossification center (SOC) (Hattori et al., 2010).

Hypertrophic conversion of proliferating chondrocytes during development of the POC is regulated by Indian hedgehog (Ihh) and parathyroid hormone related peptide (PTHrP; also known as Pthlh) (Vortkamp et al., 1996) and several transcription factors including Runx2 and Runx3 (Takeda et al., 2001; Yoshida et al., 2004; Zheng et al., 2003) and Mef2c (Arnold et al., 2007). However, the molecular mechanisms regulating the late and terminal differentiation of hypertrophic chondrocytes are still poorly understood.

Sox9 is a multifunctional factor that regulates multiple processes in endochondral ossification. It plays an essential role in early chondrogenesis (Akiyama et al., 2002), promotes the hypertrophy of prehypertrophic chondrocytes (Dy et al., 2012) and inhibits the subsequent terminal differentiation and Vegfa expression of hypertrophic chondrocytes (Hattori et al., 2010; Ikegami et al., 2011; Kim et al., 2011). Sox9 mRNA is highly expressed in chondrocytes of the proliferating and prehypertrophic zones but declines abruptly in the hypertrophic zone. Downregulation of Sox9 in the hypertrophic zone of the normal growth plate is essential to allow vascular invasion, bone marrow formation and endochondral ossification. However, to our knowledge, the mechanism underlying its downregulation in hypertrophic chondrocytes remains largely unknown.

Zinc finger and BTB domain-containing protein 20 (Zbtb20, also known as DPZF, Hof and Zfp288) is a member of a subfamily of zinc finger proteins containing C2H2 Krüppel-type zinc fingers and BTB/POZ domains (Mitchelmore et al., 2002; Zhang et al., 2001). We and others have reported that Zbtb20 functions primarily as a transcriptional repressor and plays an essential role in the specification of pyramidal neurons in the developing hippocampus (Nielsen et al., 2007; Xie et al., 2010,, 2008; Zhang et al., 2012). Zbtb20 null mice exhibit severe postnatal growth retardation, metabolic dysfunction and lethality, suggesting that Zbtb20 plays nonredundant roles in multiple organ systems (Sutherland et al., 2009). Furthermore, conditional gene targeting demonstrates that Zbtb20 regulates insulin secretion in pancreatic β cells (Zhang et al., 2012), promotes Toll-like receptor-mediated innate immune response in macrophages (Liu et al., 2013), and modulates pain sensation in nociceptive sensory neurons (Ren et al., 2014).

The growth retardation of Zbtb20 null mice prompted us to investigate its potential role in skeletal development. Here, we demonstrate previously undescribed expression of Zbtb20 in developing cartilage. Conditional deletion of Zbtb20 in developing cartilage resulted in delayed endochondral ossification. Molecular analysis showed that Sox9 mRNA and protein are ectopically upregulated in the Zbtb20-deleted late hypertrophic zone, which impeded the terminal differentiation of hypertrophic chondrocytes, subsequent growth plate vascularization and endochondral ossification. Thus, our findings suggest that Zbtb20 regulates the terminal differentiation of hypertrophic chondrocytes at least partially through repressing Sox9.

To investigate the potential role of Zbtb20 in skeletal development, we first characterized its expression pattern in the developing cartilage by immunostaining. Before embryonic day (E) 13.5, Zbtb20 protein was undetectable in the cartilage anlage of the long bones (supplementary material Fig. S1A). It was initially detected as early as E14.5, when the chondrocytes in the center of the cartilage anlage start to undergo hypertrophic differentiation (Fig. 1A). Zbtb20 protein was specifically expressed in the nuclei of hypertrophic chondrocytes (characterized by an increase in cell size and by expression of Col10a1), but was undetectable in the reserve and proliferating columnar chondrocytes as well as in Co1la1-positive osteoblasts in the perichondrium at embryonic stages. After birth, the reserve zone began to show expression of Zbtb20 protein (supplementary material Fig. S1B). The staining intensity was very weak at birth [postnatal day (P) 0.5] but rapidly increased thereafter. At P4, Zbtb20 protein was expressed at high levels in both the reserve and hypertrophic zone, and to a much lesser extent in the proliferating zone (Fig. 1B). This expression pattern of Zbtb20 in the growth plate was maintained at 3 weeks and 3 months of age (Fig. 1C; supplementary material Fig. S1C). These data suggest that Zbtb20 might play an important role in bone formation.

To explore the role of Zbtb20 in skeletal development in vivo, we deleted the gene specifically in the cartilaginous template of endochondral bones by mating mice with a floxed Zbtb20 allele (Xie et al., 2008) with Col2-Cre transgenic mice, a line that was reported to mediate loxP recombination specifically in cartilaginous structures at E13.5 (Yang et al., 2008; Zhang et al., 2005), when Zbtb20 protein expression is not detectable in chondrocytes. Immunohistochemical analysis showed that Zbtb20 protein is undetectable in chondrocytes of Zbtb20 chondrocyte-specific knockout embryos and postnatal mice (Zbtb20^flox/flox;Col2-Cre, hereafter referred to as Col-ZB20KO) (Fig. 1D,E), confirming the efficient disruption of the Zbtb20 gene in chondrocytes.

Born at the expected Mendelian ratios, Col-ZB20KO mice did not differ in size or weight from their littermate controls at birth (supplementary material Fig. S2A,B), with normal patterning of hindlimbs and forelimbs by visual inspection. From P14 onwards, Col-ZB20KO mice displayed significantly shortened tails, although they still had similar body weights to wild-type littermates (Fig. 2A,B; supplementary material Fig. S2B). From P21 onwards, the mutants showed a slight but significant decrease in the length of long bones (Fig. 2C,D). There was no significant difference between female and male mice in phenotype. Heterozygous mice were morphologically indistinguishable from wild-type littermates in the appearance and length of long bones (supplementary material Fig. S2C).

We then performed whole-mount Alcian Blue/Alizarin Red staining to examine skeletal development in detail. The mutant newborns revealed a remarkable defect in the ossification of several cartilage-based structures. In the axial skeleton, the defects in endochondral ossification were evident in the mutant vertebral column and sternum (Fig. 2E; supplementary material Fig. S2D). In the appendicular skeleton at the level of the hindlimbs, Col-ZB20KO newborns showed an impairment or absence of endochondral ossification centers in the calcaneus, the talus and middle phalanges (Fig. 2E). The ossification of calvaria, clavicle and mandible, however, was indistinguishable between mutant and wild-type control mice, suggesting that intramembranous ossification is unaffected in Col-ZB20KO mice (supplementary material Fig. S2E-G). At P14 and P21, the impairment of endochondral ossification was still apparent in the mutants and additionally manifested by a severe delay in the formation of the SOC (Fig. 2F; supplementary material Fig. S2H).

To investigate the mechanism underlying the delayed endochondral ossification in Col-ZB20KO mice, we first performed a histological examination of the growth plates at different stages. At E15.5, the central portion of the femur is composed of hypertrophic chondrocytes, with proliferative chondrocytes in each epiphyseal region. H&E staining showed that the hypertrophic domains in the wild-type and Col-ZB20KO femurs were similar, which was confirmed by the similar expression domains of Col10a1 and Opn (a marker for mature hypertrophic chondrocytes and osteoblasts) (McKee et al., 1992) (supplementary material Fig. S3A,B). This observation indicates normal timing in the initiation of chondrocyte hypertrophy during the formation of the POC. At E16, however, the mutants exhibited delayed vascular invasion in tibia [demonstrated by immunostaining for the endothelial marker CD31 (Pecam1)] (Cao et al., 2009), resulting in a delay in the formation of the POC when compared with wild-type littermates (Fig. 3A). By E17.5, the POCs in both control and mutant tibia had developed but the Col-ZB20KO tibia contained an expanded hypertrophic zone with a reduced domain of vascularization (Fig. 3B). At P2, the mutant hypertrophic zone was 74% longer than that in control mice, whereas the reserve and proliferating zones were indistinguishable between the two groups (Fig. 3C,D). This abnormality persisted, but to a lesser extent, during the growth of the bones (supplementary material Fig. S3C).

These findings suggest that the hypertrophic conversion of columnar chondrocytes remains largely normal in Zbtb20 mutants during the development of the POC, but subsequent vascular invasion and ossification are significantly delayed.

To determine whether the expansion of the hypertrophic zone was associated with increased chondrocyte proliferation in the Col-ZB20KO growth plate, we evaluated the proliferation rate of chondrocytes in vivo by a 2-h BrdU labeling approach. At P2, BrdU incorporation was comparable between Col-ZB20KO and control mice in both the reserve and proliferating zones (Fig. 4A,B). At P22, however, the mutant mice showed a marked decrease in BrdU incorporation in chondrocytes in the proliferating zone compared with control mice (Fig. 4A,B). The decreased proliferation rate of the mutant columnar chondrocytes was consistent with the observation of decreased long bone length in the mutant mice at P21.

To exclude the possibility that the expanded hypertrophic zone resulted from decreased cell death, we examined apoptosis in hypertrophic chondrocytes in the growth plate using the TUNEL assay. Very few apoptotic cells were observed at P22, and no difference was apparent between Col-ZB20KO and control mice (supplementary material Fig. S4). Together, these data suggest that the expanded hypertrophic zone in the mutant growth plate is not due to increased proliferation or decreased apoptosis of columnar chondrocytes. Therefore, we reason that the hypertrophic zone phenotype may be caused by the differentiation and/or ossification defect.

To determine whether the impaired vascularization in the mutant hypertrophic zone is associated with the alteration of chondrocyte biology, we analyzed the differentiation state of hypertrophic chondrocytes. As mentioned above, hypertrophic chondrocytes normally undergo a further maturation process. The late differentiated hypertrophic chondrocytes in the lower hypertrophic zone are capable of producing a mineralized matrix (Kirsch et al., 1997). von Kossa staining revealed an expanded mineralized hypertrophic zone in the mutant (Fig. 5A).

We next examined the expression of a set of differentiation markers in the Col-ZB20KO mice. At P4, the expression domain of Col10a1 mRNA was expanded in accordance with a lengthened hypertrophic zone in the tibia (Fig. 5B). The prehypertrophic zone, as characterized by high expression of Ihh mRNA (Vortkamp et al., 1996), appeared normal in the mutant (Fig. 5B). The expression domains and intensities of Mmp13 and Opn, which are markers for terminally differentiated hypertrophic chondrocytes at the chondro-osseous border (the lowest region of the growth plate) (Chen et al., 1993; Mark et al., 1988; Stickens et al., 2004), were comparable between mutant and control (Fig. 5C; supplementary material Fig. S5A). We then analyzed angiogenesis-related factors. Interestingly, in situ hybridization analysis revealed a marked decrease in Vegfa mRNA expression in the mutant hypertrophic zone at E16 and P4, which was confirmed by quantitative RT-PCR analysis in mutant primary hypertrophic chondrocytes at P5 (Fig. 5D-F). However, expression levels of Lect1, which encodes an angiogenesis-inhibiting factor (Hiraki et al., 1997), were comparable between control and mutant mice (supplementary material Fig. S5A). The downregulation of Vegfa was compatible with delayed cartilage vascularization in the mutant long bone at E16. Thus, Zbtb20 deletion impairs the terminal differentiation of hypertrophic chondrocytes.

An expanded hypertrophic zone could also result from a defect in osteoblast and/or osteoclast function. Therefore, we examined the expression of osteoblast markers, including Opn for immature osteoblasts and osteocalcin (also known as Bglap) for terminally differentiated osteoblasts, in the primary spongiosum (Mark et al., 1988). The mutant showed normal expression domains and intensities for these genes, as indicative of normal osteoblast development (supplementary material Fig. S5A). Similarly, the expression domain of Mmp9, a marker for chondroclast/osteoclast (Vu et al., 1998), and the number of TRAP⁺ multinuclear osteoclasts at the chondro-osseous junction (10.7±4.2 versus 9.4±1.5 TRAP⁺ cells/section; P=0.6, n=5 pairs) were comparable between control and mutant (supplementary material Fig. S5A-C).

Together, these data demonstrated that the absence of Zbtb20 causes an impairment in the final steps of chondrocyte terminal maturation but has little or no impact on the development of osteoblasts or osteoclasts.

To elucidate the mechanism by which Zbtb20 regulates the terminal differentiation and Vegfa expression of hypertrophic chondrocytes, we examined the expression of some key regulators of chondrocyte maturation. The expression domains of Runx2 and Mef2c mRNA, which are positive regulators of chondrocyte maturation that are normally expressed in the prehypertrophic and hypertrophic zones (Arnold et al., 2007; Zheng et al., 2003), were broadened in the mutant, similar to that of Col10a1, whereas the staining intensities of these genes were indistinguishable between the mutant and control growth plates (supplementary material Fig. S6).

Sox9 is required for hypertrophy but negatively regulates terminal differentiation (Dy et al., 2012; Hattori et al., 2010). Consistent with a previous report (Hattori et al., 2010), in the control growth plate Sox9 mRNA was highly expressed in chondrocytes of the proliferating and prehypertrophic zones but downregulated abruptly in the hypertrophic zone. In the mutant, however, the expression domain extended to the whole hypertrophic zone at P4 (Fig. 6A). This observation was confirmed by immunohistochemistry showing that the expression of Sox9 protein expanded to the chondro-osseous border in the mutant, whereas it was excluded from the late hypertrophic zone in control mice (Fig. 6B). Ectopic Sox9 expression in the late hypertrophic zone was also observed in the mutant embryos at E15.5 (Fig. 6C). The expression domain of Sox5 protein, which is reported to coexpress and cooperate with Sox9 in chondrogenesis (Han and Lefebvre, 2008), remained unchanged in the mutant (Fig. 6D). Thus, Zbtb20 is required specifically for Sox9 downregulation in the hypertrophic chondrocytes. As Sox9 is a direct suppressor of Vegfa expression (Hattori et al., 2010), the abnormal downregulation of Vegfa and upregulation of Sox9 in Zbtb20-deleted hypertrophic zones supports the hypothesis that Zbtb20 inhibits Sox9 expression, thereby positively regulating Vefga expression and chondrocyte terminal maturation.

To further examine the functional involvement of Sox9 in the regulation of chondrocyte terminal maturation by Zbtb20, we next investigated the effects of siRNA-mediated knockdown of Sox9 in wild-type and Zbtb20-deleted hypertrophic chondrocytes. Primary hypertrophic chondrocytes isolated from both wild-type and Col-ZB20KO mice expressed high levels of Col10a1 and alkaline phosphatase (Alpl) mRNA (Miao and Scutt, 2002), two widely used markers for hypertrophic chondrocytes, and their hypertrophic characteristics were maintained at 48 h post-seeding (supplementary material Fig. S7). The wild-type hypertrophic chondrocytes transfected with Sox9 siRNA failed to show a significant increase in Vegfa mRNA levels, probably owing to their relatively low basal level of Sox9 expression (Fig. 7). By contrast, the mutant hypertrophic chondrocytes treated with Sox9 siRNA exhibited a dramatic increase in Vegfa mRNA levels (Fig. 7). Taken together, these findings indicate that Zbtb20 positively regulates chondrocyte terminal maturation at least partly via downregulation of Sox9 expression.

To explore the molecular mechanisms underlying the regulation of Sox9 expression by Zbtb20, we first performed chromatin immunoprecipitation (ChIP) analysis to determine whether Zbtb20 directly binds the Sox9 promoter in vivo. Using ChIP grade anti-Zbtb20 antibodies (Xie et al., 2008), we did not detect any significant occupancy by Zbtb20 in the 1 kb region upstream of the Sox9 transcription start site in isolated mouse primary chondrocytes (supplementary material Fig. S8A-C), whereas the positive control with anti-acetyl-histone H3 antibody showed robust enrichment of acetyl-H3 in the Sox9 promoter. A reporter assay using a luciferase reporter construct (Sox9-Luc) harboring the mouse Sox9 promoter region (−2196 to +27) showed that overexpression of Zbtb20 in HEK293T cells had no significant effects on the transcriptional activity of the Sox9 promoter (supplementary material Fig. S8D). Taken together, these data suggest that Zbtb20 regulates Sox9 expression via an indirect, as yet unidentified, mechanism.

In this study, we examined the role of Zbtb20 in chondrocytes. Our findings provide compelling evidence that Zbtb20 is a novel essential regulator of cartilage development. First, Zbtb20 is expressed in developing chondrocytes, with expression most abundant in hypertrophic chondrocytes. Second, and most importantly, conditional deletion of Zbtb20 in developing cartilage results in delayed endochondral ossification and postnatal growth retardation, with a delay in cartilage vascularization and an expansion of the hypertrophic zone associated with reduced expression of Vegfa in the hypertrophic zone. Lastly, Sox9, a direct suppressor of Vegfa expression, is ectopically upregulated in the late hypertrophic zone by the loss of Zbtb20, which at least partially impairs the terminal differentiation of hypertrophic chondrocytes and subsequent growth plate vascularization.

The initiation of chondrocyte hypertrophy during the development of the POC is regulated by genes including Ihh, Runx2/3 and Mef2c. However, the most terminally differentiated hypertrophic chondrocytes actively express genes that allow the invasion of blood vessels, osteoclasts and osteoblast precursors from the perichondrium, suggesting additional regulatory mechanisms at this late stage. Sox9 has previously been reported to promote hypertrophy but to inhibit terminal differentiation, indicating that its downregulation is a necessary event to allow this process to occur (Dy et al., 2012; Hattori et al., 2010). Interestingly, loss of Zbtb20 resulted in the ectopic expression of Sox9 in the late hypertrophic zone, suggesting that Zbtb20 positively regulates terminal maturation, at least partially, by inhibiting Sox9 expression. Although this fits the notion that Zbtb20 functions primarily as a transcriptional repressor (Liu et al., 2013; Xie et al., 2008; Zhang et al., 2012), ChIP and luciferase reporter assays did not support direct binding of Zbtb20 to the Sox9 promoter to regulate transcriptional activity. One possibility is that Zbtb20 regulates Sox9 transcription via a cis-acting element that is located outside of the sequence that we analyzed, such as a distal enhancer. Alternatively, Zbtb20 might regulate Sox9 expression via an indirect mechanism. These possibilities are currently under further investigation.

However, Col-ZB20KO mice showed less severe retardation in vascular invasion, when compared with transgenic mice misexpressing Sox9 in hypertrophic chondrocytes under the control of a BAC Col10a1 promoter (Hattori et al., 2010). Indeed, the expression of Vegfa was completely inhibited in the Sox9-overexpressing hypertrophic zones but only downregulated in corresponding Zbtb20-deleted areas. Similarly, expression domains of Mmp13 and Opn were decreased in the Sox9 transgenic hypertrophic zones but appeared unchanged in corresponding Col-ZB20KO zones. These discrepancies could result from differences in Sox9 expression levels and domain between the two animal models. In Zbtb20-deleted hypertrophic zones, Sox9 protein levels were increased but were similar to those in the proliferating zones, as shown by comparable immunostaining intensity in these two regions in the same section (Fig. 6B). In the Sox9 overexpression model, however, the Sox9 immunostaining intensity was stronger than that in the proliferating zone in the same section, suggesting a much higher dosage of Sox9 in the hypertrophic zones. Furthermore, in Col-ZB20KO mice upregulation of Sox9 was restricted to the late hypertrophic zones, whereas in the transgenic mice Sox9 was highly overexpressed by all chondrocytes expressing Col10a1, which spans from the prehypertrophic to the terminal hypertrophic zones; indeed, the prehypertrophic zone, as characterized by strong expression of Ihh and Runx2, was extended and hypertrophic chondrocytes were retained in a prehypertrophic/early hypertrophic state in the Sox9 transgenic cartilage (Hattori et al., 2010). In addition, the other reported hypertrophic chondrocyte-specific Sox9 overexpression mouse model, which was constructed by mating CAG-mRFP1^floxed-Sox9-EGFP transgenic mice and Col10a1-Cre mice, exhibited delayed chondrocyte terminal differentiation but did not show Mmp13 downregulation (Kim et al., 2011). These discrepancies between the different animal models indicate that some of the functions of Sox9 might be dosage dependent.

A microdeletion in human chromosome 3q13.31, which encompasses five RefSeq genes including ZBTB20, has recently been reported (Molin et al., 2012). Patients with this microdeletion share several major characteristics, including significant developmental delay. We thus speculate that ZBTB20 might play crucial roles in human skeletal development, similar to that in mice.

Mouse strains carrying floxed Zbtb20 were described previously (Xie et al., 2008). To achieve chondrocyte-specific deletion of floxed Zbtb20, mice were crossed with transgenic mice expressing Cre recombinase under the control of regulatory regions of the mouse Col2a1 gene (Zhang et al., 2005). All animal experimental procedures were preformed according to institutional guidelines of the Second Military Medical University, Shanghai, China. Sex-matched littermates carrying floxed Zbtb20 were used as controls.

Tissue was fixed with 4% paraformaldehyde (pH 7.4) and decalcified in 0.5 M EDTA from P5 onwards, then embedded in paraffin or OCT (Thermo Scientific, 22-110-617). Immunostaining for paraffin sections or cryosections (CD31) was performed using the protocol described previously (Xie et al., 2010). The following antibodies were used: Zbtb20 monoclonal antibody 9A10 (made in our lab, 1:1000) (Xie et al., 2008), CD31 (Abcam, ab28364, 1:2000), Sox5 and Sox9 (Santa Cruz, sc-20091, sc-20095; 1:2000), All secondary antibodies were from Vector Laboratories (PI-1000, PI-2000; 1:500).

Riboprobes labeled with digoxygenin-UTP were transcribed from cDNA clones (supplementary material Table S1). Tissue samples were fixed overnight in 4% paraformaldehyde (pH 9.0) and decalcified in 0.5 M EDTA from P5 onwards. In situ hybridization of cryosections was performed as previously described (Xie et al., 2010).

Fetal and postnatal skeletons were freed from adherent tissue, fixed in 95% ethanol and stained for cartilage with Alcian Blue and counterstained for bone with Alizarin Red as described previously (Ovchinnikov, 2009).

Tissue was fixed and decalcified as for immunohistochemistry. Paraffin sections (4 µm) were stained with Hematoxylin and Eosin (H&E), Alcian Blue or Safranin O as described (Schmitz et al., 2010). To measure growth plate length, five semi-serial sections (20 µm between each level) through the center of proximal tibia from each mouse were cut, stained with H&E, and photographed for measurement using Image-Pro Plus software (Media Cybernetics). For von Kossa staining, nondecalcified tissue was used. Sections were deparaffinized and immersed in 1% silver nitrate solution under ultraviolet light for 45 min. Slides were rinsed in distilled water and immersed in 3% sodium thiosulfate for 5 min. Slides were rinsed again in distilled water and counterstained with Nuclear Fast Red for 5 min. Osteoclast activity was detected by staining for tartrate-resistant acid phosphatase (TRAP) using a leukocyte acid phosphatase kit (Sigma-Aldrich). At least five semi-serial sections through the center of proximal tibia from each mouse were used to reveal mean TRAP⁺ cell counts at the chondro-osseous border.

Cell proliferation was analyzed using BrdU (5-bromo-2ʹ-deoxyuridine) incorporation. BrdU was administered by intraperitoneal injection (100 mg/kg) 2 h prior to sacrifice. BrdU immunoperoxidase staining of paraffin knee joint sections was performed using a similar protocol to that described previously (Xie et al., 2010), except that antigen retrieval used 95% (v/v) formamide. BrdU⁺ cells were counted in designated areas on five semi-serial sections through the center of proximal tibia or distal femur from each animal. Apoptotic chondrocytes were detected using the In Situ Cell Death Detection Kit (Roche Diagnostics).

Hypertrophic cartilage was microdissected from P5 long bones and ribs, and digested with collagenase as described (Belluoccio et al., 2010). Total RNA was extracted using TRIzol (Life Technologies) and digested with DNase I to eliminate any contaminating genomic DNA. cDNA was synthesized by reverse transcription and amplified in triplicate using the SYBR Green PCR assay (QuantiFast SYBR Green PCR Kit, Qiagen, 204057). RNA expression levels were normalized to that of internal control 36B4 (Rplp0). The primers used are listed in supplementary material Table S2.

Pooled primary resting and hypertrophic chondrocytes from P2 wild-type or P7 Col-ZB20KO mice were isolated as described above and subjected to further digestion with 1.0 mg/ml of pronase (Roche) for 3 h at 37°C with slow agitation. The resting chondrocytes were harvested, while the hypertrophic chondrocytes were washed twice in PBS and electroporated with 500 nM Sox9 small interfering RNA (siRNA; sc-36534, Santa Cruz) or control siRNA (sc-36869, Santa Cruz) diluted in BTXpress High Performance Electroporation Solution (Harvard Apparatus). Electroporation was performed in a 2 mm cuvette with settings of one pulse of 170 Volts for 15 ms using a BTX ECM830 electroporator. Chondrocytes were then plated and cultured in DMEM supplemented with 10% FBS (Hyclone), 25 μg/ml L-ascorbic acid 2-phosphate, 10 mM β-glycerophosphate and 1× penicillin/streptomycin/amphotericin B solution (Sigma). The cells were collected 48 h after electroporation and mRNA or total proteins were extracted. Sox9, Vegfa, Col10a1 and Alpl mRNA levels were analyzed by real-time RT-PCR, standardized to 36B4. Sox9 protein levels were assessed by western blotting. The intensity of the signals was evaluated by densitometry and semi-quantified using the ratio between the protein of interest and β-actin for each experiment.

The Sox9 luciferase reporter was constructed by cloning the mouse Sox9 promoter region (−2196 to +27) into the XhoI and BglII sites of the pGL4.10 vector (Promega). HEK293T cells were co-transfected with 1 µg reporter construct, 100 ng β-galactosidase expression vector as internal control and 1 µg either Zbtb20 expression vector or a mock vector. Luciferase and β-galactosidase were measured 48 h post-transfection as described previously (Xie et al., 2008).

Fragmented chromatin from mouse P5 primary chondrocytes was incubated overnight with specific antibodies: anti-acetylated histone H3 (Millipore, 06-942), anti-Zbtb20 (Xie et al., 2010), or normal mouse IgG (Upstate, 12-371), followed by incubation with Dynabeads Protein G (Invitrogen). Purified chromatin DNA was subjected to conventional and quantitative PCR with primers for core promoter regions of Sox9 (supplementary material Table S2). Two independent ChIP reactions were performed.

Results are expressed as mean±s.d. Statistical comparison between genotypes was performed with a two-tailed Student's t-test. Proliferative index was analyzed by one-way analysis of variance (ANOVA). Data from the RNA interference experiment were analyzed using a two-way ANOVA. P<0.05 was considered significant.

We thank Y. Zhang, Q Hao and L Gao for technical assistance.