Sox9 sustains chondrocyte survival and hypertrophy in part through Pik3ca-Akt pathways

During development, the limb skeleton is created through endochondral bone formation. Mesenchymal cells initially undergo condensation, which is followed by the differentiation of prechondrogenic cells within these condensations into round chondrocytes, which proliferate and produce cartilage extracellular matrix to form cartilage primordia. Proliferating chondrocytes in the central region of the cartilage then exit the cell cycle and differentiate into prehypertrophic and, subsequently, hypertrophic chondrocytes. The proliferating chondrocytes closest to the prehypertrophic chondrocytes flatten out and form orderly columns of flat chondrocytes that are still proliferating. Finally, hypertrophic chondrocytes progress to terminal maturation to express matrix metalloproteinase 13 (Mmp13). Terminally mature chondrocytes undergo apoptosis. Blood vessels along with osteoblasts, osteoclasts and hematopoietic cells then invade and form primary ossification centers (Lefebvre and Smits, 2005). Cartilage at both ends of each skeletal component remains as articular cartilage and sustains joint movement.

Sox9 is a member of the SOX (Sry-related high mobility group box) family of transcription factors that share the high mobility group (HMG) DNA-binding motif with the mammalian testis-determining factor Sry. Heterozygous mutations in the human SOX9 gene cause the skeletal malformation syndrome campomelic dysplasia (Foster et al., 1994; Wagner et al., 1994). Sox9 is expressed in progenitor cells in various organs (Akiyama et al., 2005), including chondroprogenitors, osteoprogenitors and preadipocytes (Wang and Sul, 2009), but is not expressed in most differentiated somatic cells such as osteoblasts and adipocytes (Wang and Sul, 2009), with the exception of chondrocytes. During endochondral bone formation, Sox9 expression starts in mesenchymal progenitor cells. Sox9 remains highly expressed in chondrocytes, and its expression ceases in prehypertrophic chondrocytes (Ng et al., 1997; Zhao et al., 1997). SOX9 expression continues in articular cartilage and decreases in osteoarthritic cartilage, a major cartilage disease caused by degeneration (Brew et al., 2010). That Sox9 plays a crucial role in mesenchymal progenitor cells has been established by analysis of Sox9 knockout chimeras and conditional knockout mice. In mouse chimeras, Sox9^−/− cells are excluded from cartilage primordia throughout embryonic development (Bi et al., 1999). Furthermore, mesenchymal condensation and subsequent cartilage formation are absent in the limbs of Prx1-Cre; Sox9^flox/flox conditional knockout mice, in which the Sox9 gene is inactivated in early mesenchymal limb bud cells before mesenchymal condensation occurs (Akiyama et al., 2002). In addition, the cartilage is very hypoplastic in Col2a1-Cre; Sox9^flox/flox conditional knockout mice, in which Sox9 expression is lost in condensed mesenchymal cells before differentiation into chondrocytes. In Col2a1-Cre; Sox9^flox/flox mice, Sox9^{floxdel/floxdel} cells remain as condensed mesenchymal cells and do not differentiate into chondrocytes (Akiyama et al., 2002).

The function of Sox9 expression post-differentiation into chondrocytes has not been determined. The functions of Sox9 in chondrocytes cannot be determined from previously reported chimeras or conditional knockout mice because they lack chondrocytes. Emerging evidence suggests that Sox9 inhibits chondrocyte hypertrophy; Sox9^+/− embryos show premature mineralization of cartilage and expanded hypertrophic zones (Bi et al., 2001). Mice that overexpress Sox9 under the control of Col2a1 regulatory elements exhibit delayed cartilage mineralization (Akiyama et al., 2004). In these genetically modified mice, however, the nature of the chondrocytes and matrix properties are altered before differentiation into chondrocytes owing to the manipulation of Sox9 expression from the stage of mesenchymal progenitor cells. A recent study has shown that misexpression of Sox9 in hypertrophic chondrocytes results in a lack of bone marrow, and that Sox9 is a major negative regulator of cartilage vascularization (Hattori et al., 2010).

We recently generated two types of transgenic mice in which Cre is expressed at different steps during chondrocyte differentiation (Iwai et al., 2008). In 11Enh-Cre transgenic mice, Cre recombinase activities are controlled by the Col11a2 promoter and enhancer and begin during the round chondrocyte stage. 11Enh-Cre directs recombination at a later stage (both developmental and within the chondrocyte differentiation pathway) than Col2a1-Cre. Cre recombinase activities in 11Prom-Cre transgenic mice are controlled by the Col11a2 promoter alone and begin in the flat chondrocyte stage. In the present study, we examined the function of Sox9 in chondrocytes by generating 11Enh-Cre; Sox9^flox/flox and 11Prom-Cre; Sox9^flox/flox conditional knockout mice.

To generate Sox9 conditional knockout mice, 11Enh-Cre transgenic mice, 11Prom-Cre transgenic mice (Iwai et al., 2008) and Sox9^flox/flox mice (Akiyama et al., 2002) were prepared and mated. To generate Sox9; Pten double conditional knockout mice, Pten^flox/flox mice (Suzuki et al., 2001) were prepared. For genotyping, genomic DNA was isolated from tail tips or embryonic skin and subjected to PCR analysis according to the methods previously described for the Cre transgene (Iwai et al., 2008), Sox9 allele (Akiyama et al., 2002) and Pten allele (Suzuki et al., 2001).

Total RNA was extracted using the RNeasy Mini Kit (Qiagen). Total RNAs were digested with DNase to eliminate any contaminating genomic DNA. PCR amplification was with SYBR Premix ExTaq (Takara) on a 7900HT real-time PCR system (Applied Biosystems). RNA expression levels were normalized to that of Gapdh. The primers used are listed in Table S1 in the supplementary material.

Embryos were dissected, fixed in 100% ethanol overnight, and then stained with Alcian Blue followed by Alizarin Red S solution according to standard protocols (Peters, 1977).

Embryos were dissected under a stereomicroscope, fixed in 4% paraformaldehyde, processed and embedded in paraffin. For immunohistochemistry, sections were incubated with primary antibodies (Table 1). Immune complexes were detected using secondary antibodies conjugated to Alexa Fluor 488 (Table 1). RNA in situ hybridization was performed using ³⁵S-labeled antisense riboprobes as previously described (Pelton et al., 1990) or using a DIG RNA labeling kit (Boehringer Mannheim, Indianapolis, IN, USA) according to the manufacturer's instructions. In situ hybridization and immunohistochemistry were performed at least three times for each analysis.

Pregnant mice were intraperitoneally injected with BrdU labeling reagent (10 μl/g body weight; Zymed Laboratories, South San Francisco, CA, USA). Two hours later, the mice were sacrificed and embryos were dissected and sectioned. Incorporated BrdU was detected using a BrdU staining kit (Zymed Laboratories) to distinguish actively proliferating cells. The average number of BrdU-positive cells among total cells (± s.d.) was calculated.

TUNEL assays were performed on semi-serial sections using the DeadEnd Fluorometric TUNEL System (Promega) according to the manufacturer's protocol.

Images were acquired on an inverted microscope (Eclipse Ti, Nikon) equipped with cameras (DS-Fi1, Nikon; C4742-80-12AG, Hamamatsu Photonics) and NIS Elements software (Nikon).

Cell lysates were subjected to SDS-PAGE, electroblotted and immunostained with the antibodies listed in Table 1. Immunoblots were performed at least three times for each analysis.

SW1353 (ATCC #HTB94), HeLa (Riken #RBC0007) and Saos2 (Riken #RCB0428) cells were cultured in Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% streptomycin/penicillin. ATDC5 cells were maintained at 20-80% confluency as described previously (Shukunami et al., 1996). For SOX9 overexpression experiments, SW1353 cells were transfected with the plasmid CMV promoter-SOX9 using Amaxa nucleofection technologies. Cells were harvested for real-time RT-PCR and immunoblot analysis 24 hours after transfection.

Amaxa nucleofection technology was used to transfect 1×10⁶ SW1353, HeLa or Saos2 cells with 200 nM negative control siRNA (Stealth RNAi Negative Control Medium GC Duplex #2, 12935-112, Invitrogen) or SOX9-a, SOX9-b or SOX9-c siRNAs (Invitrogen Stealth RNAi; siRNA target sequences are listed in Table S2 in the supplementary material).

Six hours after transfection, cell lysates were collected for immunoblot analysis. For morphological evaluation, 1×10⁶ cells were seeded into 6-well plates and 6 hours after transfection cells were stained with Hoechst 33342 (Dojindo Laboratories, Japan). Caspase 3/7 activities were measured 6 hours after transfection using the Caspase-Glo3/7 Assay Kit (Promega) according to the manufacturer's instructions. Six hours after transfection, the level of apoptosis was determined using the Cell Death Detection ELISA^PLUS Kit (Roche Applied Science).

Various lengths of human PIK3CA promoter sequence (Hui et al., 2008) were prepared. Mutations were introduced using the GeneTailor Site-Directed Mutagenesis System (Invitrogen). Fragments were inserted into the pGL3-basic vector (Promega). Undifferentiated ATDC5 cells and SW1353 cells were co-transfected with 20 ng reporter construct, 10 ng phRL-TK (Promega) and a total of 200 ng SOX9 expression vector and mock vector using FuGENE (Roche). Cell lysates were collected 48 hours after transfection. Photinus luciferase activity levels were normalized to those of Renilla luciferase (phRL-TK).

A ChIP assay kit (Upstate #17-295) was used according to the manufacturer's protocol. The purified DNA was used as a template in PCR assays. The primers are listed in Table S3 in the supplementary material.

Data are shown as averages with standard deviations. Student's t-test was used to compare data. P<0.05 was considered statistically significant.

We initially intercrossed 11Enh-Cre mice (Iwai et al., 2008) with Sox9^flox/flox mice (Akiyama et al., 2002). Sox9^flox/+ heterozygotes that harbor 11Enh-Cre were recovered with the expected Mendelian frequency (see Fig. S1A in the supplementary material); these mice were fertile and developed dwarfism (see Fig. S1B in the supplementary material). Eighty-five percent of 11Enh-Cre; Sox9^flox/+ mice were viable beyond 6 months of age (see Fig. S1C in the supplementary material), suggesting that these mice have a much milder phenotype than Col2a1-Cre; Sox9^flox/+ mice, which have a 95% death rate by 10 days of age (Akiyama et al., 2002).

We then intercrossed 11Enh-Cre; Sox9^flox/+ mice with Sox9^flox/flox mice (see Fig. S1D in the supplementary material). The cartilage and bone of 11Enh-Cre; Sox9^flox/flox embryos at 16.5 days post-coitum (dpc) were very hypoplastic, whereas the calvarium, which undergoes membranous ossification and is formed without cartilage templates, appeared to be well formed (see Fig. S1E-G in the supplementary material).

Histological analysis showed that 11Enh-Cre; Sox9^flox/flox conditional knockout embryos exhibited normal mesenchymal condensation with normal Sox9 expression patterns in paws at 12.5 dpc (Fig. 1A). At 12.5 dpc, cells in the central region of mesenchymal condensation at the humerus differentiated into round chondrocytes, as indicated by round or polygonal cell morphologies and slight staining of the surrounding matrix with Safranin O in control Sox9^flox/+ embryos (Fig. 1B, left column). Round chondrocytes in the central region of condensation in 11Enh-Cre; Sox9^flox/flox humerus showed decreased staining of the matrix with Safranin O and lost Sox9 expression (Fig. 1B, right column). These results suggest that 11Enh-Cre initiated the direct recombination of floxed Sox9 genes in round chondrocytes. At 13.5 dpc, the primordial cartilage of control Sox9^flox/+ mouse humerus was formed, as indicated by Safranin O staining (Fig. 1C). Immunohistochemistry with anti-Sox9 antibodies showed that proliferative chondrocytes in cartilage expressed Sox9. By contrast, primordial cartilage in 11Enh-Cre; Sox9^flox/flox humerus was disorganized at 13.5 dpc, with weak Safranin O staining intensities, and the chondrocytes in the central region of cartilage had lost Sox9 expression. Cells in the periphery of the primordial cartilage still expressed Sox9. Col2a1 expression was lost in accordance with the loss of Sox9. This in vivo result supports the notion that Sox9 is needed for the maintenance of Col2a1 transcription in chondrocytes. Hypertrophic chondrocytes were absent and Col10a1 expression was lost. At 14.5 dpc, 11Enh-Cre; Sox9^flox/flox humerus showed only a small amount of cartilage, which was disorganized and lacked hypertrophic chondrocytes (Fig. 1D). Ihh expression was almost completely lost and expression of Col10a1 and Mmp13 was lost. Runx2 expression was maintained in bone collars. These results suggest that Sox9 deletion in round chondrocytes abolishes the subsequent differentiation of chondrocytes.

Together with an existing report that the loss of Sox9 is associated with apoptosis in neural crest cells (Cheung et al., 2005), the presence of just a small amount of disorganized cartilage, despite the initial mesenchymal condensation, raised the possibility that chondrocytes underwent apoptosis. We found that cells were TUNEL negative and did not show immunoreactivity for cleaved caspase 3 in either 11Enh-Cre; Sox9^flox/flox or Sox9^flox/+ humerus at 12.5 dpc (see Fig. S2A in the supplementary material). Cells in the epiphyseal region within the disorganized cartilage in 11Enh-Cre; Sox9^flox/flox humerus showed TUNEL staining and immunoreactivity for cleaved caspase 3, whereas chondrocytes in the control Sox9^flox/+ humerus did not at 13.5 dpc (Fig. 1E). As for positive controls, TUNEL-positive cells were detected in the interdigital regions of paws of both control Sox9^flox/+ and 11Enh-Cre; Sox9^flox/flox embryos at 13.5 dpc (see Fig. S2B in the supplementary material). These results suggest that epiphyseal chondrocytes undergo apoptosis ~1 day after deletion of Sox9 in round chondrocytes.

Disorganized primary ossification centers were formed in 11Enh-Cre; Sox9^flox/flox humerus at 15.5 dpc (see Fig. S2C in the supplementary material). Col2a1 expression was almost completely lost and expression of Ihh and Col10a1 was lost. Expression of Mmp13 was lost in chondrocytes but maintained in the primary ossification center. Bone in the primary ossification center could have been formed by endochondral ossification through abnormal chondrocytes that survived after Sox9 deletion or by membranous ossification from the bone collar.

We intercrossed 11Prom-Cre mice (Iwai et al., 2008) with Sox9^flox/flox mice (Akiyama et al., 2002) to examine Sox9 function in chondrocytes at a later stage of differentiation. 11Prom-Cre; Sox9^flox/+ mice were recovered with the expected Mendelian frequency (see Fig. S3A in the supplementary material), developed normally and showed similar crown-rump lengths as control Sox9^flox/+ mice 3 weeks after birth (see Fig. S3B in the supplementary material). 11Prom-Cre; Sox9^flox/flox embryos were recovered with the expected Mendelian frequency from 12.5-16.5 dpc (see Fig. S3C in the supplementary material). The skeleton was very hypoplastic, whereas the calvarium appeared to be well formed (see Fig. S3D-G in the supplementary material). The cartilage of the limb buds and vertebral bodies was very hypoplastic.

Cre-mediated recombination patterns of 11Prom-Cre were analyzed by mating 11Prom-Cre transgenic mice with CAG promoter-flox-CAT-flox-lacZ transgenic tester mice (Sakai and Miyazaki, 1997). No recombination was detected in limbs at 12.5 dpc (Fig. 2A). lacZ activity was detected in several layers of chondrocytes in the humerus at 13.25 dpc, which is when prehypertrophy starts at the center of the primordial cartilage. In situ hybridization analysis of sections from different samples at the same stage of 13.25 dpc showed that recombination started to occur in Ihh-expressing cells (Fig. 2B). At later stages, cells closest to the prehypertrophic chondrocytes subsequently flattened out and formed orderly columns of flat chondrocytes in which 11Prom-Cre directs recombination (Iwai et al., 2008).

Histological analysis showed that 11Prom-Cre; Sox9^flox/flox conditional knockout mice form primordial cartilage normally in the humerus and a normal Sox9 expression pattern is present at 13.0 dpc (Fig. 2C). At 13.5 dpc, the humerus of Sox9^flox/+ control mice showed prehypertrophy and hypertrophy of chondrocytes in the central region of cartilage primordia. Proliferative chondrocyte-specific expression of Sox9 abruptly stopped when chondrocytes entered into the prehypertrophic stage in control mice. The length (proximodistal direction) of the zone of proliferative chondrocytes expressing Sox9 in 11Prom-Cre; Sox9^flox/flox humerus was 25% shorter than that in Sox9^flox/+ control humerus (Fig. 2D, arrows). These results suggest that Sox9 expression was lost in several layers of flat chondrocytes closest to the prehypertrophic chondrocytes in 11Prom-Cre; Sox9^flox/flox humerus.

The width (transverse direction) of the central region of the primordial cartilage was decreased in 11Prom-Cre; Sox9^flox/flox mouse humerus at 13.5 dpc, and the humerus had a dumb-bell-shaped appearance (Fig. 2D,E). The dumb-bell shape was confirmed by Alcian Blue staining of the skeleton at 14.5 dpc (Fig. 2F). The zone of hypertrophic chondrocytes and the zone of terminally mature chondrocytes were formed in the center of Sox9^flox/+ control humerus at 14.5 dpc. By contrast, the hypertrophic chondrocytes were absent and cells that resembled terminally mature chondrocytes were present in the center of the 11Prom-Cre; Sox9^flox/flox humerus (Fig. 2G).

We analyzed the expression of marker genes at 13.5 dpc. The expression pattern of Col2a1 mRNAs (Fig. 3A) corresponded almost exactly to that of Sox9 proteins in 11Prom-Cre; Sox9^flox/flox humerus (Fig. 2D). Ppr (receptor for parathyroid hormone and parathyroid hormone-related peptides; Pth1r – Mouse Genome Informatics) mRNAs were detected in 11Prom-Cre; Sox9^flox/flox humerus (Fig. 3A). Hypertrophic chondrocyte marker type X collagen (Col10a1) mRNA was detected in the center of Sox9^flox/+ cartilage, but not in 11Prom-Cre; Sox9^flox/flox cartilage. Runx2 mRNA was strongly detected in bone collars of primordial cartilage and weakly detected in chondrocytes in Sox9^flox/+ humerus. Runx2 expression in chondrocytes appeared to increase in 11Prom-Cre; Sox9^flox/flox humerus. The bone collar was thickened and type I collagen expression in the bone collar increased in 11Prom-Cre; Sox9^flox/flox humerus. The distances (Fig. 3B, black arrows) between the distal articular surface and prehypertrophic zone indicated by Ihh expression were reduced in 11Prom-Cre; Sox9^flox/flox humerus (Fig. 3B, bar chart). The distance (Fig. 3B, white arrows) between the two zones of Ihh-positive cells was increased in 11Prom-Cre; Sox9^flox/flox humerus. These results, together with the Sox9 expression patterns (Fig. 2D), suggest that the premature shutdown of Sox9 in several cell layers of flat chondrocytes caused a shift of Ihh expression toward the articular surfaces of primordial cartilage and an absence of Col10a1 expression.

The abnormalities found in 11Prom-Cre; Sox9^flox/flox cartilage at 13.5 dpc were enhanced at 14.5 dpc. In control Sox9^flox/+ humerus, flat chondrocytes, which had emerged next to prehypertrophic chondrocytes at 13.5 dpc, dramatically increased in number in the metaphyseal zones at 14.5 dpc. The number of chondrocytes in the metaphyseal zone of 11Prom-Cre; Sox9^flox/flox humerus was decreased as compared with the control Sox9^flox/+ humerus (see Fig. S4A in the supplementary material, top row, outlined region). Sox9 expression was only detected in epiphyseal round chondrocytes in 11Prom-Cre; Sox9^flox/flox humerus, whereas Sox9 was expressed both in epiphyseal round and metaphyseal flat chondrocytes in the control Sox9^flox/+ humerus. These results indicate that metaphyseal proliferative chondrocytes in 11Prom-Cre; Sox9^flox/flox humerus are Sox9^{floxdel/floxdel} proliferative chondrocytes (see Fig. S4A in the supplementary material, right panels, outlined region). The expression patterns of Col2a1 and Col11a2 corresponded to those of Sox9 proteins. Ppr was expressed, but Col10a1 expression was lost in 11Prom-Cre; Sox9^flox/flox cartilage. Real-time RT-PCR analysis of RNA from the humerus confirmed a significant decrease in Col10a1 expression (Fig. 3C).

Col1a1 mRNAs and proteins were ectopically expressed in the chondrocytes in the central region of 11Prom-Cre; Sox9^flox/flox humeral cartilage at 14.5 dpc (Fig. 3D). Together with the cell morphology (Fig. 2G, TC-L), the expression of Mmp13, bone sialoprotein (Bsp; Ibsp – Mouse Genome Informatics) mRNA and osteopontin (Op; Spp1 – Mouse Genome Informatics) mRNA (Fig. 3D) strongly suggest that these cells are terminally mature chondrocytes. The absence of Col10a1-expressing cells and the reduced number of Mmp13-, Bsp- and Op-expressing cells suggest that Sox9 deletion in flat chondrocytes severely inhibits subsequent hypertrophy and moderately inhibits terminal maturation. Mineralization of chondrocytes in the central region of the cartilage was increased. Cells were TUNEL negative and did not show immunoreactivity for cleaved caspase 3 in 11Prom-Cre; Sox9^flox/flox humerus at 13.5 dpc (see Fig. S4B in the supplementary material). Significant apoptosis associated with the expression of cleaved caspase 3 was detected in the zone of terminally mature chondrocytes (Fig. 3E) at 14.5 dpc. These results suggest that Sox9 deletion started in flat chondrocytes at 13.5 dpc and that chondrocytes lacking Sox9 undergo abnormal differentiation (severe inhibition of hypertrophy and moderate inhibition of terminal differentiation) into terminally mature chondrocytes, where excess apoptosis is detected at 14.5 dpc.

BrdU labeling analysis (see Fig. S4C in the supplementary material) revealed increased proliferation of perichondrial bone collar cells in 11Prom-Cre; Sox9^flox/flox humeral cartilage at 13.5 dpc; this increased proliferation was responsible for the thick bone collar (Fig. 2D). Proliferation rates in metaphyseal zone chondrocytes in 11Prom-Cre; Sox9^flox/flox humeral cartilage decreased at 14.5 dpc; this decreased proliferation caused a corresponding decrease in the length of the zone of metaphyseal chondrocytes (see Fig. S4A in the supplementary material).

Our results indicate that flat chondrocytes lacking Sox9 expression are characterized by the cessation of Col2a1 expression, decreased proliferation rates, a reduced cell population, a shift of prehypertrophy towards the articular end, and a lack of hypertrophy; these chondrocytes immediately enter into terminal maturation associated with ectopic type I collagen expression in the cartilage matrix, undergo increased apoptosis and stimulate the proliferation of bone collar cells. Thus, Sox9 is needed for subsequent chondrocyte hypertrophy. This phenotype is similar to those of Sox5^−/−; Sox6^−/− (Smits et al., 2001), Sox5^+/−;Sox6^−/− and Sox5^−/−;Sox6^+/− mice (Smits et al., 2004). Sox5^−/−; Sox6^−/− mice lack columnar chondrocytes and Col10a1 expression, but they do have prehypertrophic chondrocytes and express Mmp13. Because Sox5 and Sox6 facilitate the organization of transcription complexes (Lefebvre and Smits, 2005), the phenotypic similarities between 11Prom-Cre; Sox9^flox/flox mice and Sox5^−/−; Sox6^−/− mice suggest that the Sox9 conditional knockout phenotype is due to dysfunction of transcription complexes containing Sox5, Sox6 and Sox9. The observation that chondrocytes lacking Sox9 enter into terminal maturation without hypertrophy is consistent with findings that Sox9 misexpression in hypertrophic chondrocytes inhibits their terminal maturation (Hattori et al., 2010).

Next, we examined whether Sox9 is important for the survival of chondrosarcoma cells in vitro. To knockdown SOX9 mRNA, we transfected human SW1353 chondrosarcoma cells with small interfering RNA (siRNA) using the Amaxa nucleofection technique, which yields high transfection efficiencies (see Fig. S5A in the supplementary material), and harvested the cells 6 hours later. SW1353 cells transfected with SOX9-c siRNA showed a dramatic decrease in SOX9 protein, whereas SOX9-b siRNA yielded a moderate decrease and SOX9-a siRNA did not affect the level of SOX9 protein (Fig. 4A). Hoechst-stained SW1353 cells transfected with siRNA SOX9-b and siRNA SOX9-c exhibited typical apoptotic morphology characterized by chromatin condensation and DNA fragmentation (Fig. 4B). The decrease in SOX9 expression (Fig. 4A) correlated with the increase in caspase 3/7 activities (Fig. 4C) and the amount of cleaved DNA-histone complexes (nucleosomes) (Fig. 4D) in SW1353 cells. These results suggest that the cell survival mechanism that requires SOX9 is also active in SW1353 chondrosarcoma cells. A possible explanation for the rapid death of SW1353 cells is that these chondrosarcoma cells undergo rapid cell cycles, such that the outcome of loss of SOX9 is quickly realized. Transfection of human HeLa cells or osteoblastic Saos2 cells with SOX9 siRNAs did not cause apoptosis; this result is consistent with the lack of SOX9 expression in these cell types (see Fig. S5B-D in the supplementary material). Thus, Sox9 is specifically needed for the survival of chondrocyte lineage cells, including chondrocytes and chondrosarcoma cells.

We next investigated the molecular mechanism of Sox9-induced survival of chondrocyte lineage cells. We first examined the activation of signaling molecules in SW1353 chondrosarcoma cells by western blot analysis. SOX9 knockdown specifically decreased the phosphorylation level of AKT, but it did not affect the phosphorylation levels of p38 mitogen-activated protein kinase (p38 MAPK; MAPK14 – Human Genome Nomenclature Committee), extracellular signal-regulated kinase (ERK), Jun N-terminal kinase (JNK) and activating transcription factor 2 (ATF2) (Fig. 4E). Consistently, the overexpression of SOX9 specifically increased the phosphorylation level of AKT, but did not change the phosphorylation levels of other signaling molecules (Fig. 4F).

We then examined phosphorylation in vivo. Immunohistochemical analysis showed that phospho-Akt was below detectable levels in both the control Sox9^flox/+ and 11Enh-Cre; Sox9^flox/flox mice at 12.5 dpc (Fig. S5E in the supplementary material). Phospho-Akt was detected in the central part of primordial cartilage and not in the periphery, whereas Sox9 was detected both in the central and peripheral parts of primordial cartilage in the control Sox9^flox/+ mice at 13.5 dpc (Fig. 4G). The area of Sox9 deletion was limited to the central part of cartilage in 11Enh-Cre; Sox9^flox/flox mice and appeared to cover the area of phospho-Akt in the control Sox9^flox/+ mice at 13.5 dpc. Phospho-Akt was not detected in 11Enh-Cre; Sox9^flox/flox cartilage at 13.5 dpc (Fig. 4G), nor in Sox9^{floxdel/floxdel} proliferative chondrocytes in 11Prom-Cre; Sox9^flox/flox mice (Fig. 4H). There were no obvious differences in the levels of phosphorylated Atf2, Jnk or p38 Mapk, as detected by antibody, between Sox9^{floxdel/floxdel} proliferative chondrocytes in 11Prom-Cre; Sox9^flox/flox mice and proliferating chondrocytes in Sox9^flox/+ mice (see Fig. S5F in the supplementary material).

We next examined whether reduced phosphorylation of Akt is responsible for Sox9 deletion-mediated chondrocyte apoptosis. We prepared 11Prom-Cre; Sox9^flox/flox; Pten^flox/flox double conditional knockout mice. Pten (phosphatase and tensin homolog) is a lipid phosphatase, the major substrate of which is PtdIns(3,4,5)P₃. Thus, the deletion of Pten results in the accumulation of PtdIns(3,4,5)P₃, leading to forced activation of Akt. Col2a1-Cre; Pten^flox/flox mice show increased phosphorylation of Akt (Ford-Hutchinson et al., 2007; Yang et al., 2008). At 14.5 dpc, Alcian Blue staining of the skeleton showed that the degree of deformity of humeral cartilage in 11Prom-Cre; Sox9^flox/flox; Pten^flox/flox mice was milder than that in 11Prom-Cre; Sox9^flox/flox; Pten^+/+ mice (Fig. 5A). The constriction of central regions of the humerus was also significantly milder. Histological analysis showed that the decreased cell numbers in the center of the cartilage, the excess number of TUNEL-positive cells and the loss of Col10a1 expression in 11Prom-Cre; Sox9^flox/flox; Pten^+/+ mice were partly restored in 11Prom-Cre; Sox9^flox/flox; Pten^flox/flox mice (Fig. 5B,C). Pten expression was decreased in Sox9^{floxdel/floxdel}; Pten^{floxdel/floxdel} proliferative chondrocytes in 11Prom-Cre; Sox9^flox/flox; Pten^flox/flox mice (Fig. 5D and see Fig. S6 in the supplementary material). Phosphorylation levels of Akt in Sox9^{floxdel/floxdel}; Pten^{floxdel/floxdel} proliferative chondrocytes were higher than those of Sox9^{floxdel/floxdel}; Pten^+/+ proliferative chondrocytes. Thus, both Sox9 deletion and elevated Akt phosphorylation occurred in flat chondrocyte regions and the resulting modulation of apoptosis became apparent subsequently, when cells differentiated into terminally mature chondrocytes. These results suggest that decreased Col10a1 expression and increased apoptosis induced by Sox9 deletion are partly restored by the additional deletion of Pten, which restores Akt phosphorylation.

We used expression analysis and the candidate approach to investigate the molecular mechanism underlying the regulation of Akt phosphorylation by Sox9. We found that the expression of Pik3ca was regulated by Sox9. Pik3ca (also known as p110α) is one of three subunit proteins of phosphatidylinositol 3-kinase (PI3K). PI3K generates phosphatidylinositol (3,4,5)-trisphosphate [PtdIns(3,4,5)P₃] from PtdIns(4,5)P₂. PtdIns(3,4,5)P₃ causes phosphorylation of and activates Akt (Brazil et al., 2004). PI3K-Akt signaling regulates cell death (Cantley, 2002).

The regulation of PIK3CA expression by SOX9 was detected in SW1353 chondrosarcoma cells in vitro. Among genes encoding PI3K subunits, PIK3CA mRNA was induced by SOX9 overexpression (Fig. 6A). The expression levels of genes encoding the other PI3K subunits and AKT members were little affected by SOX9 overexpression. PIK3CG and PIK3CD (also known as p110γ and p110δ, respectively) mRNAs were not detected in SW1353 cells. Western blot analysis of SW1353 cell lysates showed that SOX9 overexpression increased PIK3CA protein levels but did not affect PIK3CB (p110β) or PIK3R1 (p85α)/PIK3R2 (p85β) protein levels (Fig. 6B). Consistently, SOX9 knockdown decreased PIK3CA protein levels but did not affect PIK3CB or PIK3R1/PIK3R2 protein levels (Fig. 6C).

The Sox9-mediated regulation of Pik3ca expression was also detected in vivo. Real-time RT-PCR analysis confirmed that Pik3ca expression was significantly decreased in 11Prom-Cre; Sox9^flox/flox humeral cartilage as compared with the expression level in control Sox9^flox/+ cartilage (Fig. 6D). Immunohistochemical analysis showed that Pik3ca immunoreactivity decreased in proliferative chondrocytes that did not express Sox9 (Sox9^{floxdel/floxdel}) as compared with Sox9-expressing chondrocytes in 11Prom-Cre; Sox9^flox/flox mice (Fig. 6E), whereas the immunoreactivities of Pik3cb and Pik3r1/Pik3r2 were not obviously changed (Fig. 6E).

We examined the possibility that SOX9 directly controls PIK3CA transcription. DNA sequence analysis of human PIK3CA revealed several putative SOX9 binding sites around the transcription start site that are conserved in the mouse and rat Pik3ca genes (Fig. 7A). Co-transfection of either undifferentiated mouse ATDC5 or human SW1353 cells with SOX9 stimulated the promoter activities of luciferase reporter constructs bearing 241 bp or longer PIK3CA promoters (Fig. 7B). SOX9 enhanced the activity of the 2122 bp promoter in a dose-dependent manner (Fig. 7C), whereas the 54 bp reporter constructs showed weak promoter activities regardless of the presence or absence of SOX9 (Fig. 7B). One consensus SOX9 binding site exists between −241 bp and −54 bp, at −70 bp (Fig. 7A). The stimulating effect of SOX9 on the 241 bp promoter was strongly impaired when a mutation was introduced into this consensus SOX9 binding site (Fig. 7B).

Evidence of SOX9 binding to the PIK3CA promoter in vivo was provided by chromatin immunoprecipitation (ChIP) of fragmented DNA from SW1353 chondrosarcoma cells using a primer pair that is specific for the −70 bp consensus SOX9 binding region (Fig. 7D). SOX9 binding was not detected when primer pairs specific for the other consensus sites were used. These results suggest that SOX9 binds to the PIK3CA promoter region to enhance its activities.

In the present study, immunohistochemical analysis with anti-Sox9 antibodies clearly demonstrated Sox9 deletion patterns in conditional knockout mice. Fig. 8A shows a schematic representation of stages when Sox9 genes are inactivated during the chondrocyte differentiation process in Sox9 conditional knockout mice bearing each Cre transgene.

In Col2a1-Cre; Sox9^flox/flox conditional knockout mice, most chondroprogenitor cells are arrested as condensed mesenchymal cells, whereas a few cells differentiate into chondrocytes and undergo hypertrophy (Akiyama et al., 2002), suggesting the co-existence of a small population of chondroprogenitor cells in which Cre-mediated deletion does not occur and thus the chondrocyte differentiation program runs. By contrast, there were no hypertrophic chondrocytes in 11Enh-Cre; Sox9^flox/flox and 11Prom-Cre; Sox9^flox/flox mice, suggesting that Sox9 genes were inactivated uniformly at specific stages of chondrocyte differentiation.

Along with cellular differentiation processes, chondrocytes are believed to undergo apoptosis after going through hypertrophy. Our results show that Sox9 inactivation in round chondrocytes caused premature apoptosis and that Sox9 inactivation in flat chondrocytes caused immediate terminal maturation without hypertrophy and with excessive apoptosis (Fig. 8A). SOX9 knockdown caused apoptosis of SW1353 chondrosarcoma cells. Sox9 is expressed in progenitor cells in various organs (Akiyama et al., 2005), whereas most differentiated cell types, including osteoblasts and adipocytes, do not express Sox9 and survive. These findings suggest that chondrocytes survive by a mechanism that is sustained by Sox9. An association between Sox9 and cell survival has been shown in other cell lineages and tumors. Sox9 is transiently expressed in premigratory neural crest cells during development, and the loss of Sox9 results in apoptosis of neural crest cells (Cheung et al., 2005). Sox9 sustains the survival of neurofibromatosis tumor cell lines (Miller et al., 2009). Sox9 is expressed in the prostate epithelia and promotes prostate tumor cell proliferation and cooperates with Pten loss to drive tumor formation (Thomsen et al., 2010). However, the mechanisms by which Sox9 promotes cell survival were unknown. Our results suggest that Sox9 directly binds to the promoter region of the PI3K subunit gene Pik3ca, enhancing the phosphorylation of Akt. It will be of interest to investigate the roles of Sox5 and Sox6 in the regulation of PI3K subunit expression, e.g. to analyze Pik3ca expression in chondrocytes of Sox5^−/−; Sox6^−/− mice. Forced phosphorylation of Akt by Pten inactivation partially restored cell survival of Sox9^{floxdel/floxdel} chondrocytes. Thus, the chondrocyte-specific anti-apoptotic mechanism includes PI3K-Akt pathways that are activated by Sox9 (Fig. 8B). Because the restoration was partial, other as yet unknown mechanisms also contribute to the Sox9-dependent survival of chondrocytes. Our results showing that SOX9 regulates PIK3CA transcription suggest that cell death occurs cell-autonomously in SW1353 cells, although we speculate that non-cell-autonomous pathways also contribute to chondrocyte death. Decreased Col2a1 expression in Sox9^{floxdel/floxdel} chondrocytes results in reduced extracellular matrix, which might cause non-cell-autonomous chondrocyte death, especially in vivo. It is also possible that increased apoptosis in the terminal mature chondrocytes of 11Prom-Cre mice might reflect an acceleration of the whole process of terminal maturation, including their removal via apoptosis. Because Sox9 does not appear to increase phospho-Akt levels in the prostate (Thomsen et al., 2010), the mechanisms by which Sox9 sustains cell survival might differ between tissues.

The PI3K-Akt pathway in chondrocytes has been studied extensively. Akt knockout mice show delayed calcification (Chen et al., 2001; Fukai et al., 2010; Peng et al., 2003), whereas calcification is increased in 11Prom-Cre; Sox9^flox/flox mice. Osteoblasts in 11Prom-Cre; Sox9^flox/flox mice are not affected. A possible explanation for this discrepancy is that delayed ossification in Akt knockout mice is caused by Akt deletion in osteoblasts or bone collar cells. Akt signaling enhances chondrocyte proliferation (Kita et al., 2008) and PI3K decreases apoptosis (Ulici et al., 2008), consistent with the phenotypes of 11Prom-Cre; Sox9^flox/flox mice. The effects on hypertrophy are controversial, as PI3K enhances hypertrophy (Fujita et al., 2004), whereas Akt signaling inhibits hypertrophy (Kita et al., 2008). The phenotypes of 11Prom-Cre; Sox9^flox/flox mice appear to be consistent with the former. Discrepancies might reflect the complex downstream pathways of Akt, as Akt regulates chondrocyte proliferation and differentiation through GSK3, mTOR and FoxOs differently (Rokutanda et al., 2009). Taken together with the finding that Akt-GSK3 pathways are relatively unimportant in vertebral bodies (Rokutanda et al., 2009), the very small primordial cartilage of the vertebral bodies and limbs in 11Prom-Cre; Sox9^flox/flox mice suggests that Akt-mTOR and Akt-FoxO pathways mediate the chondrocyte survival mechanism that is sustained by Sox9.

Previous in vivo and in vitro studies have implied that Sox9 inhibits the hypertrophy of chondrocytes, thus maintaining proliferating chondrocyte characteristics (Akiyama et al., 2004; Bi et al., 2001). By contrast, our results show that Sox9 deletion in flat chondrocytes results in an absence of hypertrophic chondrocytes and in immediate terminal differentiation. Sox9 interacts with Runx2 and represses its activities (Zhou et al., 2006). Our results revealed that the Runx2 expression level is increased in Sox9-deficient chondrocytes, indicating that Sox9 also directly or indirectly inhibits Runx2 at the level of transcription. The expression of Col1a1, but not Col10a1, was activated in Sox9-deficient chondrocytes, although both genes are direct targets of Runx2 (Ducy et al., 1997; Zheng et al., 2003). This observation suggests that flat chondrocytes lacking Sox9 lose other components necessary for subsequent Col10a1 expression and hypertrophy. It is also possible that decreased levels of phospho-Akt secondarily affect Col10a1 expression. It is interesting that, in 13.5 dpc humerus, the premature shutdown of Sox9 expression by only a few cell layers causes dramatic changes in subsequent chondrocyte differentiation and survival. This suggests that expression of Sox9 to the very last stage of flat chondrocytes, just prior to hypertrophy, is essential for chondrocyte hypertrophy and for proper subsequent endochondral bone formation.

11Enh-Cre; Sox9^flox/+ embryos exhibited a straight radius and ulna (see Fig. S1E,F in the supplementary material). By contrast, campomelic dysplasia, which is associated with mutations in a single allele of SOX9, is characterized by the bending of long bones (Foster et al., 1994; Wagner et al., 1994). Bowing and angulation of the radius and ulna are also seen in Sox9^+/− mouse embryos (Bi et al., 2001). These results suggest that skeletal bending is caused by Sox9 deficiency at the mesenchymal cell stage.

In summary, our results from 11Enh-Cre; Sox9^flox/flox and 11Prom-Cre; Sox9^flox/flox mice and from SOX9 knockdown in human chondrosarcoma cells suggest that the expression of Sox9 in differentiated chondrocyte lineage cells is essential for cell survival. Sox9 in differentiated chondrocytes is also needed for subsequent hypertrophy and proper endochondral bone formation in mice. Sox9 sustains cell survival at least partly through its binding to the Pik3ca promoter, inducing Akt phosphorylation. Understanding Sox9 function in differentiated chondrocytes will contribute both to maintaining cartilage homeostasis and treating chondrosarcoma.

We thank Junko Murai, Kunihiko Hiramatsu, Mari Shinkawa, Kanako Nakagawa, Mina Okamoto, Hidetatsu Outani and Satoru Sasagawa for assistance and helpful discussions; and Andreas Schedl for the preparation of Sox9^flox/flox mice. This study was supported in part by Scientific Research Grants 18390415, 19659378 and 21390421 from MEXT and JST, CREST.