ATF6a, a Runx2-activable transcription factor, is a new regulator of chondrocyte hypertrophy

In eukaryotic cells, endoplasmic reticulum (ER) stress is known to initiate a signaling pathway called the unfolded protein response (UPR) (Hetz, 2012; Zhang and Kaufman, 2004; Ron and Walter, 2007). Three signaling pathways are involved in the UPR, and these are mediated by inositol-requiring enzyme-1 (IRE1, also known as ERN1), PKR-like ER resistant kinase (PERK, also known as EIF2AK3) and activation transcription factor 6 (ATF6), which act in concert to limit new protein synthesis and to increase the levels of chaperones. These three different pathways are triggered by ER stress to reduce misfolded protein levels (Kaufman, 2002,, 1999; Kim et al., 2006). In addition to induction of autophosphorylation and activation of IRE1 and PERK, ER stress also causes ATF6 transit through the Golgi complex, where it is activated through proteolytic cleavage by the enzymes S1P and S2P (also known as MBTPS1 and MBTPS2, respectively). The cleaved N-terminal cytoplasmic domain of ATF6 (known as ATF6a) is released from the Golgi complex, followed by translocation to the nucleus (Li et al., 2000; Nakanishi et al., 2005).

Chondrogenesis is an essential process for cartilage development and endochondral bone growth (Lui et al., 2010; Lefebvre and Smits, 2005). Bone morphogenetic proteins, in particular BMP2, are believed to play important roles in mediating chondrocyte differentiation, proliferation and function (Chen et al., 2004; Yu et al., 2012). BMP2 is known to activate UPR signaling molecules, including PERK, C/EBP homologous protein (CHOP, also known as DDIT3) and IRE1α. In addition, the unfolded protein response in ER stress has been reported to mediate BMP2 signaling in osteoblasts (Lai and Cheng, 2002; Murakami et al., 2009).

Previously, we have reported that the spliced isoform of XBP1 (XBP1s) is a crucial inducer in the BMP2 signal pathway and is involved in BMP2-triggered chondrogenesis and bone formation. XBP1s stimulates chondrocyte hypertrophy, maturation and bone growth through the growth factor granulin and epithelin precursor (GEP) (Guo et al., 2014a,b). Furthermore, BMP2 is known to activate ER stress sensors, including ATF6 and OASIS (also known as CREB3L1). BMP2 can stimulate osteoblast differentiation and extracellular matrix mineralization through Runx2-induced ATF6 expression, which enhances osteocalcin transcription and expression (Jang et al., 2012). In this study, we examined the expression of ATF6 in the course of chondrogenesis and the regulation of chondrocyte differentiation (especially hypertrophy) by ATF6 and its cleaved ATF6a form, as well as the molecular mechanisms involved.

We first examined ATF6 expression during cartilage development using immunohistochemistry at various time points [embryonic day (E)12.5, E14.5, E15.5, E17.5, newborn and postnatal day 10]. As shown in Fig. 1A, ATF6 was undetectable at E12.5, and became detectable in the center and surrounding of the condensation part of the cartilage at E14.5. It was highly expressed in prehypertrophic chondrocytes at E15.5 and E17.5. ATF6 was clearly expressed all over the entire growth plate in newborn and postnatal day 10 mice. This expression profiling of ATF6 in growth plates during development suggests that ATF6 is involved in the whole chondrogenic process.

We next assessed the expression profiles of ATF6a and ATF6 during chondrocyte differentiation using the C3H10T1/2 cell line, a pluripotent murine stem cell line widely used for in vitro chondrogenic studies (Zhang et al., 2008; Johnson et al., 2008; Meirelles Lda and Nardi, 2003). Firstly, BMP2 (300 ng/ml) was used to induce chondrocyte differentiation in micromass cultures of C3H10T1/2 cells, then these cells were collected at different time points followed by real-time PCR for measurement of ATF6a (i.e. ATF6), box B-binding factor-2 human homolog on chromosome 7 (BBF2H7; also known as CREB3L2) and type collagen X (Col X, also known as COL2A1) mRNA levels (Fig. 1B). As shown in Fig. 1B, the ATF6a mRNA level was comparably low until day 5, then it remained at a high level during the stages when differentiation was occurring. BBF2H7, the ER stress transducer, was expressed at a relatively low level at day 1, markedly increased at day 3 and showed a subsequent reduction. BBF2H7 is known to be induced in resting chondrocytes and proliferating chondrocytes (Saito et al., 2009). It has been reported that BBF2H7 is preferentially expressed in chondrocytes of developing cartilage, and both BBF2H7 and collagen II (ColII) are targets of Sox9 (Hino et al., 2014). The secreted BBF2H7 C-terminus can activate Hedgehog signaling and promote chondrocyte proliferation (Saito et al., 2014). Conversely, Col X, a hypertrophic chondrocyte marker, was obviously upregulated at day 7 to day 9. Interestingly, the expression of ATF6a increased substantially at day 5, whereas that of BBF2H7 was clearly reduced, and the peak level of ATF6a came 2 days earlier than that of collagen X. These results indicate that ATF6a might regulate Col X expression during BMP2-induced chondrocyte differentiation. Additionally, a similar expression profiling was also observed in the timecourse of chondrogenesis of primary bone marrow stromal cells (BMSCs) (Fig. 1C).

Next, we detected the expression of ATF6a and ATF6 at the protein level during chondrocyte differentiation. As revealed in Fig. 1D,E, the expression of ATF6 was reduced whereas that of ATF6a (cleaved ATF6) was increased with the prolonged BMP2 treatment; ATF6a protein has no obvious expression until day 5 or 7 in the BMP2-stimulated chondrocyte differentiation of BMSCs and the BMP2-stimulated C3H10T1/2 cells. These results suggest that BMP2 induces ATF6 cleavage and produces ATF6a during chondrogenesis. Furthermore, ATF6a expression appeared earlier than that of Col X, which was immunopositive at day 5 in BMSCs or day 7 in C3H10T1/2 cells. Thus, we demonstrated that the expression of ATF6a is specific for the hypertrophic and prehypertrophic chondrocytes.

We then sought to clarify the molecular mechanism modulating hypertrophic chondrocyte expression of ATF6. A previous sequence analysis has shown that there are three Runx2-binding elements (RBEs) and three Sox6-binding elements (SBEs) in the promoter of ATF6 gene (Cantu’ et al., 2011; Roca et al., 2005). It has been reported that both Runx2 and Sox6 are transcriptional factors that regulate hypertrophic chondrocyte formation (Cantu’ et al., 2011; Yoshida et al., 2002). These findings pushed us to detect whether Runx2 and/or Sox6 interacts with the promoter of ATF6 gene. We firstly determined whether Runx2 and Sox6 interacted with ATF6 promoter using electrophoretic mobility shift assays (EMSAs) (Fig. 2A).

Two digoxigenin-labeled probes targeting the first Runx2-binding site (Fig. 2A) and the first Sox6-binding site (Fig. 2B) were incubated with the nuclear proteins from C3H10T1/2 cells infected with adenovirus encoding Runx2 or Sox6 (Ad-Runx2 and Ad-Sox6, respectively), which resulted in a particular Runx2–DNA complex (Fig. 2A, lane 3 and 4) or Sox6–DNA complex (Fig. 2B, lane 3 and 4). Addition of anti-Runx2 or -Sox6 antibodies supershifted the Runx2–DNA and Sox6–DNA bands, respectively (Fig. 2A,B, lane 4). In addition, the binding of corresponding probes to Runx2 (Fig. 2A, lane 2) or Sox6 (Fig. 2B, lane 2) in vitro was totally eliminated by addition of excess unlabeled probes. Thus, we demonstrate that the binding of Runx2 or Sox6 to their corresponding binding motifs is sequence specific.

Next, to examine whether Runx2 and Sox6 also interacted with the ATF6 promoter in vivo, chromatin immunoprecipitation (ChIP) assays were performed in C3H10T1/2 cells infected with Ad-Runx2 or Ad-Sox6. First, cell lysates were immunoprecipitated with control IgG (negative control), anti-Runx2 or anti-Sox6 antibodies after crosslinking with formaldehyde, then the DNA purified from this immunoprecipitation was analyzed by PCR with PCR primers that spanned the first Runx2-binding site (Fig. 2C) and Sox6-binding site (Fig. 2D), respectively, in the ATF6 promoter. We found that a clear PCR product was observed in the DNA isolated from immunoprecipitated complexes with anti-Runx2 (Fig. 2C, lane 3) or anti-Sox6 (Fig. 2D, lane 3) antibodies but not with control IgG (Fig. 2C, lane 4; Fig. 2D, lane 4). Thus, we demonstrate that both Runx2 and Sox6 interact with their corresponding elements in the ATF6 promoter in the infected living cells. To further testify their binding under physiological condition, we also collected the cell lysate from micromass cultures of C3H10T1/2 cells treated with BMP2. As revealed in Fig. 2E,F, both Runx2 and Sox6 can particularly bind to the promoter of ATF6 during the course of chondrocyte differentiation.

To examine whether Runx2 regulates the transcription of the ATF6 promoter, we used two reporter gene constructs, ATF6-3R3SBE-luc [pGL3-ATF6(-2103)-luc] and ATF6-1R1SBE-luc [pGL3-ATF6(-384)-luc], in which segments with RBEs and SBEs were placed upstream of a gene encoding luciferase in the pGL3 vector (Fig. 3A,B). Co-transfection of C3H10T1/2 cells and ATDC5 cells with these two reporter plasmids and a Runx2 expression plasmid strongly increased reporter gene expression in a dose-dependent manner (Fig. 3C,E). In the case of the Sox6 transcription factor, Sox6 inhibited both pATF6-specific reporter gene constructs, and this inhibition was also dose dependent (Fig. 3D,F).

To dissect the importance of the Runx2-binding site and the Sox6-binding site in the 5′-flanking region of ATF6 gene, several point mutation reporter constructs of pGL3-ATF6-1R1SBE-luc were produced and their transcriptional activity by Runx2 or Sox6 was tested (Fig. 3G). The mutations of the six nucleotides (AGTGTG to TCCCCA) in the RBE resulted in an obvious reduction in reporter gene activities (Fig. 3H), and the alteration of five nucleotides (CCAGC to TTGAT) in the SBE clearly reduced the inhibitory effect of the reporter gene by Sox6 (Fig. 3I). These data confirm that both the RBE and the SBE are responsible for driving ATF6 expression mediated by Runx2 and Sox6.

We then assessed whether Runx2 and Sox6 were involved in the expression of the endogenous ATF6 gene. As shown in Fig. 4A, after Ad-Runx2 infection for 48 h, the ATF6 gene mRNA level was remarkably increased (approximately three- to four-fold). However, infection of Ad-Sox6 resulted in a nearly 50% reduction in the ATF6 mRNA level (Fig. 4B). These data were also verified at protein level; the expression of ATF6 and the cleaved ATF6 (ATF6a) were prominently enhanced after Ad-Runx2 infection, and remarkably reduced after Ad-Sox6 infection (Fig. 4C,D). These results clearly suggest that Runx2 and Sox6 can modulate the endogenous ATF6 gene expression and cleavage.

Next, we determined whether ATF6a and ATF6 can influence chondrogenesis. We first explored the effect of ATF6a, ATF6 and BMP2 in chondrocyte differentiation in high-density micromass cultures of C3H10T1/2 and BMSCs, which possess the ability to differentiate into various lineages of tissue cells (i.e. are multipotent) (Liu et al., 2004; Atkinson et al., 1997). In brief, high-density cultures of C3H10T1/2 or BMSC cells were treated with BMP2 (positive control, 300 ng/ml), or infected with Ad-ATF6a, Ad-ATF6 or adenovirus encoding control GFP (Ad-GFP) for 3 days for early markers (aggrecan and ColII), and for 7 days for late markers (MMP13 and ColX)., respectively, then, RNA was extracted for quantitative RT-PCR. As shown in Fig. 5A,B, chondrogenesis was monitored by analyzing the expression of marker genes specific for chondrocyte maturation, including Col II (also known as COL2A1), aggrecan, MMP13 and Col X (Canalis et al., 2003; Welch et al., 1998; Colnot, 2005). As for BMP2, ATF6a and ATF6 remarkably promoted the expression of Col II, aggrecan, MMP13 and Col X. This suggests that both ATF6a and ATF6 are favorable mediators of hypertrophic chondrocyte differentiation.

Using a long-term culture system (3 weeks) of high-density human mesenchymal stem cell (hMSC) pellets, we demonstrated that ATF6a and ATF6, like BMP2, induced chondrogenesis, as reflected by positive staining with Safranin O (pink color, left panel, Fig. 5C) and immunostaining for Col II and Col X (right panels, Fig. 5C). Then, 15-day-old mouse fetus metatarsal bones were cultured, and the effects of ATF6a and ATF6 on endochondral bone growth were assessed. These explants included undifferentiated cartilage. During a 5-day explantation culture of Ad-ATF6a and Ad-ATF6 stimulation, these explants went through all phases of endochondral bone growth and formation. As revealed in Fig. 5D, both ATF6a and ATF6 remarkably promoted chondrocyte hypertrophy, mineral conversion and bone length increase.

Based on our observation that ATF6a augments chondrogenesis, we next sought to examine whether suppression of its expression though a small interfering RNA (siRNA) approach would also affect hypertrophic chondrocyte differentiation. Infection with adenoviruses encoding ATF6a siRNA (siATF6a) efficiently inhibited endogenous ATF6a expression in C3H10T1/2 cells (77%, Fig. 6A) and BMSC cells (73%, Fig. 6F). High-cell-density cultures of C3H10T1/2 cells or BMSC cells were stimulated with BMP2 for various times, then infected with either control adenovirus [siRFP, multiplicity of infection (MOI) 50] or siATF6a adenovirus (MOI 50). We found that knockdown of ATF6a largely blocked the Col II (Fig. 6B,G), aggrecan (Fig. 6C,H), Col X (Fig. 6D,I) and MMP13 (Fig. 6E,J) mRNA expression in the course of chondrogenesis. This outcome obviously demonstrates that endogenous ATF6a is necessary for chondrocyte hypertrophy and chondrogenesis.

We then examined whether ATF6a interacts with Runx2 in chondrogenesis by performing a CoIP assay. Micromass cultures of C3H10T1/2 cells were induced with BMP2 for 5 days, then the treated cell extracts were incubated with control IgG or anti-Runx2 antibodies, and the complexes were detected with anti-ATF6a antibody. A specific ATF6a band was immunoprecipitated by anti-Runx2 from the cell lysates (Fig. 7A, lane 2), but not by control antibodies. Note that ATF6a specifically interacted with Runx2 during BMP2-triggered chondrogenesis. Control IgG and cell lysates were used as negative and positive controls.

Next, to detect whether ATF6a also interacts with Runx2 in BMP2-induced BMSCs, micromass cultures of BMSCs cells were stimulated with BMP2 for 5 days, then the treated cell extracts were incubated with the IgG control (Fig. 7B, lane 3) or the anti-Runx2 antibody (Fig. 7B, lane 2), and complexes were detected with anti-ATF6a antibody. We found that ATF6a bound to Runx2 during BMP2-induced chondrogenesis in BMSCs.

Then, we detected whether ATF6 interacted with Runx2 without BMP2. Extracts from C3H10T1/2 cells were incubated with the IgG control (Fig. 7C, lane 3) or the anti-Runx2 antibody (Fig. 7C, lane 2), and the complexes were examined with the anti-ATF6 antibody. The result showed that the anti-Runx2 antibody could not bring down ATF6 protein (Fig. 7C), indicating that ATF6 cannot interact with Runx2 without BMP2. Taken together, ATF6a and Runx2 can from a new protein complex during BMP2-triggered chondrocyte hypertrophy and chondrogenesis. Without BMP2 stimulation, this complex is not formed.

It has been reported that Runx2 is necessary for chondrocyte hypertrophy through Col X, IHH and MMP13 activation (Zheng et al., 2003). To test this, high-density cultures of C3H10T1/2 cells were infected with adenovirus encoding ATF6a (Ad-ATF6a), ATF6 (Ad-ATF6), Runx2 (Ad-Runx2) or a combination, then RNA was extracted for real-time PCR. As shown in Fig. 7D–F, the mRNA expression of Col X, MMP13 and IHH were remarkably increased in cells infected with both Ad-ATF6a and Ad-Runx2, or both Ad-ATF6 and Ad-Runx2 compared with those in Ad-Runx2-infected cells. This suggests that both ATF6a and ATF6 significantly augment the expression of Col X, MMP13 and IHH induced by Runx2, and the former (ATF6a) enhances that expression more obviously. These results illustrate that ATF6a is a cofactor of Runx2 in regulating chondrocyte hypertrophy and chondrogenesis.

It is well known that IHH and PTHrP cooperatively control the process of hypertrophic chondrocyte differentiation. We then determined whether ATF6a influences this signaling pathway in hypertrophic chondrocyte differentiation. First, high density micromass cultures of BMSCs (Fig. 8A,C) or C3H10T1/2 cells (Fig. 8B,D) infected with Ad-ATF6a or control (CTR) were subjected to BMP2 stimulation for different time points, then the real-time PCR was performed. The result indicates that ATF6a increased the IHH expression (Fig. 8A,B), whereas it obviously inhibited the PTHrP mRNA level (Fig. 8C,D) compared with the control. By contrast, lowering expression of ATF6a through siRNA completely suppressed IHH mRNA expression and augmented PTHrP expression during hypertrophic chondrocyte differentiation of both BMSCs (Fig. 8E,G) and C3H10T1/2 cells (Fig. 8F,H). Taken together, these results demonstrate that ATF6a can influence chondrogenesis through IHH and PTHrP, an important signaling pathway in hypertrophic chondrocyte differentiation.

The ER is a cellular organelle responsible for the folding and post-translational modifications of proteins, and its homeostasis is disturbed when ER stress occurs. ER stress triggers a conserved response, referred to as the UPR, which mitigates ER stress (Bernales et al., 2006; Urano et al., 2000). Accumulated evidence has shown that factors influencing cell fate and/or differentiation are activated during ER stress (Liu et al., 2000; Korennykh et al., 2009; Lin et al., 2007), but whether and how such factors affect chondrocyte differentiation remains largely unknown. Therefore, in this study we examined the role of ATF6 and ATF6a, one branch of the UPR, in hypertrophic chondrocyte differentiation and the molecular mechanism of this role.

ATF6 is a multifunctional transcription factor that has been implicated in the regulation of cell cycle and cell differentiation of many types of cells, such as osteoblasts and skeletal muscle myotubes (Jang et al., 2012; Kim et al., 2014; Vekich et al., 2012; Glembotski, 2014; Howarth et al., 2014; Wu et al., 2011). ATF6a is the cleaved N-terminal cytoplasmic domain of ATF6, produced by protease-mediated cleavage of ATF6 during ER stress (Li et al., 2000; Nakanishi et al., 2005). Herein, we found that ATF6 was clearly expressed in the entire growth plate chondrocyte in vivo (Fig. 1A) and was induced during BMP2-triggered chondrogenesis in vitro (Fig. 1B,C). Interestingly, the expression of ATF6 was reduced whereas that of ATF6a (cleaved ATF6) was increased as BMP2 treatment was prolonged; prominent ATF6a protein was detected at day 5 or day 7 of BMP2-induced chondrocyte differentiation (Fig. 1D,E). The different expression profiles of ATF6 and ATF6a during chondrocyte differentiation demonstrate that post-transcription regulations might also be involved in the regulation of ATF6 and ATF6a expression during chondrogenesis.

ATF6 is an ER membrane-bound bZIP transcription factor. Under normal conditions, it is bound to the ER membrane through an interaction with the chaperone BiP (also known as GRP78 and HSPA5) (Murakami et al., 2009; Li et al., 2000; Nakanishi et al., 2005). Upon the accumulation of unfolded or misfolded proteins in ER stress, ATF6 is released from BiP and transits to the Golgi, where it is cleaved by proteolysis, which is followed by nuclear translocation of its N-terminal fragment ATF6a (Jang et al., 2012; Roca et al., 2005). In this study, we also found that ATF6 was differentially induced and cleaved during BMP2-mediated chondrocyte differentiation. The Runx2 transcription factor appears to activate, whereas Sox6 inhibits, the expression of the ATF6 gene during BMP-2 stimulated chondrogenesis (Figs 3–5). There are three Runx2-binding consensus sequences [RCCRC(A/T)] and three Sox6-binding motifs [(A/T)(A/T) CAA(A/T)G] (Cantu’ et al., 2011; Roca et al., 2005) in the ATF6 gene promoter (Fig. 2). Our assays, including both EMSAs and ChIP, indicated that Runx2 and Sox6 interact with their binding sites in the regulatory region of ATF6 gene and mediate the delicate expression of ATF6 gene during BMP-2-trigged chondrogenesis (Figs 2–4).

Our in vitro chondrogenesis assays indicate that ATF6a and ATF6 are positive regulators of chondrocyte differentiation, given that their overexpression enhances, whereas their suppression inhibits, BMP2-induced expressions of marker genes for chondrogenesis, including aggrecan, Col II, Col X and MMP13. The concept that ATF6a and ATF6 are the stimulators of chondrocyte differentiation was further supported by ex vivo and in vivo studies demonstrating that ATF6a and ATF6 stimulate mineralization and endochondral bone growth (Figs 5 and 6). Interestingly, ATF6a was found to associate with Runx2 and to act as a co-factor of Runx2 in the course of chondrogenesis. This finding is in accordance with numerous previous publications showing that a number of Runx2-binding proteins can modify its transcriptional function (Yoshida et al., 2004,, 2002; Drissi, et al., 2003; Cui et al., 2003; Gutierrez et al., 2002; Schroeder et al., 2004). Here, our data showed that ATF6a associates with Runx2 and enhances Runx2-induced chondrocyte hypertrophy. In addition, this interaction is BMP-2 dependent, given that this complex is undetectable without BMP2 stimulation (Fig. 7).

It has been reported that IHH and PTHrP signaling regulates the process of hypertrophic chondrocyte differentiation through a negative-feedback loop during growth plate endochondral ossification. PTHrP suppresses chondrocyte hypertrophy and negatively regulates endochondral bone development (Huang et al., 2001; Karp et al., 2000; Kronenberg, 2006). IHH is expressed in prehypertrophic and hypertrophic chondrocytes and stimulates PTHrP production (Hu et al., 2007; Vortkamp et al., 1996). PTHrP-null mice exhibit shortened zones of proliferative chondrocytes and premature hypertrophic differentiation, and have a severe disorder of endochondral bone formation (MacLean et al., 2004). In the mice lacking either PTHrP or the parathyroid hormone (PTH) and PTHrP receptor, the proliferating zones of growth plate of metatarsal bones are decreased and hypertrophic chondrocyte differentiation is premature (Amizuka et al., 1994; Lanske et al., 1996; Lee et al., 1996; Weir et al., 1996). In contrast, chondrocyte-specific PTHrP transgenic mice display delayed hypertrophied chondrocyte differentiation (Vortkamp et al., 1996). It is known that the primary binding sites for ATF6a include the cis-acting ER stress response elements (ERSE and ERSEII) and the UPR element on the basis of its interactions (Okada et al., 2002; Wang et al., 2000). Interestingly, the PTHrP gene promoter contains one UPR element sequence [TGACGT(T/G)] and five ERSEII sequences [CCAC (G/A)]. In addition, the IHH promoter also contains two UPR element sequences and four ERSEII sequences (Yamamoto et al., 2004; Yoshida et al., 2000). These results suggest that ATF6a, as a transcription factor, might modulate IHH and PTHrP transcription and expression through interacting with specific cis-elements in their promoter regions. Our studies have demonstrated that overexpressing ATF6a enhances the expression of IHH, whereas it inhibits PTHrP; however, knockdown of ATF6a reduces the expression of IHH and increases PTHrP (Fig. 8A–H). Whether regulation of PTHrP and IHH by ATF6a requires the direct binding of ATF6a to their promoters needs to be further delineated.

Based on the findings in this study, together with the literature (Ron and Walter, 2007; Murakami et al., 2009; Han et al., 2014; Cantu’ et al., 2011; Roca et al., 2005; Kronenberg, 2006; Yamamoto et al., 2004), as well as our previous reports (Guo et al., 2014b; Han et al.,2014; Feng et al., 2010), we propose a model for illustrating the role and regulation of ATF6a in the course of hypertrophic chondrocyte differentiation (Fig. 8I). In brief, Runx2 activates the expression of ATF6 and enhances ATF6 cleavage. ATF6a, derived from ATF6, interacts with Runx2, and functions as a cofactor of Runx2 in mediating the chondrocyte hypertrophy. ATF6a and Runx2 form a positive-feedback control loop. However, Sox6 prohibits ATF6 expression and cleavage. Furthermore, ATF6a also regulates the expressions of IHH and PTHrP, modulating the process of chondrocyte hypertrophy and chondrogenesis as an opposing feedback loop. In summary, results in this study testify that ATF6a is a previously unrecognized modulator of chondrocyte differentiation and that it exerts its function in mediating chondrogenesis through, at least in part, acting as the co-activator of Runx2, and by affecting IHH and PTHrP signaling.

To construct the wild-type and two mutant versions of the pGL3-ATF6-luc reporter plasmid, the target sequences were amplified using PCR with the following primers: 5′-GGTACCGCTGCAGTGAGCTGAGATGGCT-3′ and 5′-CTGGAGATCACCCGGTACTTCCCCAGTG-3′ for wild-type ATF6-luc (−2103 to +321); 5′-GGTACCGTTCTGAG ATAGCCACGCTGTGG-3′ and 5′-CTGGAGATCACCCGGTACTTCCCCAGTG-3′ for wild-type ATF6-luc (−384 to +321); 5′-CTTCCCCCGCCTAGTGTGTAAAACAGCGGGAC-3′ and 5′-GTGGGCGGAAGTAGGGAGGAAGA-3′ for mut1; and 5′-TATTTTTAATATTACCAGCAAACTTTGTTTAGTC-3′ and 5′-GTGGGCGGAAGTAGGGAGGAAGA-3′ for mut2 (the mutated nucleotides in the primers are underlined). PCR products were inserted into the pGL3 vector.

Adenovirus encoding ATF6a (Ad-ATF6a), adenovirus encoding ATF6a siRNA (siATF6a), adenovirus encoding ATF6 (Ad-ATF6), and adenovirus encoding Runx2 and Sox6 were constructed, respectively, with methods described previously (Guo et al., 2014a,b). All constructs were verified by nucleic acid sequencing; subsequent analysis was performed using BLAST software (National Institutes of Health).

Mouse bone marrow was isolated by flushing the femurs and tibiae of 8-week-old female BALB/c mice with 0.6 ml of improved minimal essential medium (Sigma-Aldrich), supplemented with 20% fetal bovine serum, 100 units/ml penicillin, 100 µg/ml streptomycin (Invitrogen) and 2 mM glutamine (Invitrogen), and then it was filtered through a cell strainer (Falcon, BD Biosciences). Cells were centrifuged, resuspended and plated out at a density of ∼2×10⁶ cells/cm². Then cells were incubated at 37°C under 5% CO₂. After 72 h, non-adherent cells and debris were removed, and the adherent cells were cultured continuously. Non-detached cells were discarded, and the remaining cells were regarded as passage 1 of the BMSC culture. All animal experiments were performed according to approved guidelines.

For histological examination, the sections of postcoital day 12.5, 14.5, 15.5, 17.5 embryos and newborn mice, and postnatal day 10 mice were immunostained for ATF6 after serum blocking for 30 min at 37°C. For detection, biotinylated secondary antibody and horseradish peroxidase (HRP)–streptavidin complex (Santa Cruz Biotechnology, CA) were used. A total of 0.5 mg/ml 3,3′-diaminobenzidine (DAB) substrate in 50 mM Tris-HCl was used for visualization, then sections were counterstained with Mayer's hematoxylin.

To examine the effects of Runx2 and Sox6 on the expression of the ATF6 gene, total RNA was extracted from C3H10T1/2 cells infected with Ad-Runx2, Ad-Sox6 or control Ad-GFP with the RNeasy kit (Qiagen, Hilden, Germany). 1 μg of total RNA per sample was reverse transcribed into cDNA. The following sequence-specific primers were synthesized: sense 3′-AACGAGAACGACGAGGTGGT-5′ and antisense 3′-AAAGGAGGCAGATGACAGGTGAC-5′ for collagen II; sense 3′-TACCACGTGCATGTGAAAGG-5′ and antisense 3′-GGAGCCACTAGGAATCCTGAG-5′ for collagen X; sense 3′-GAGTCCCCAAGAGCCACCCA-5′ and antisense 3′-TGGTGGGCTGATAGGTGGGC-5′ for IHH; and sense 3′-ATGCTGCGGAGGCTGGTTCA-5′ and antisense 3′-GCACGGAGTAGCTGAGCAGGAA-5′ for PTHrP. The following pair of oligonucleotides was used as internal controls: 5′-ACCACAGTCCATGCCATCAC-3′ and 5′-TCCACCACCCTGTTGCTGTA-3′ for GAPDH. Real-time PCR was performed with the SYBR Green PCR kit (Qiagen, Hilden, Germany) in a 96-well optical reaction plate formatted in the ABI 7400 system according to the manufacturer's protocol. The transcript of GAPDH mRNA was employed as an internal control for RNA quality. PCR cycling conditions were as follows: initial incubation step of 2 min at 50°C, reverse transcription of 60 min at 60°C and 94°C for 2 min, followed by 40 cycles of 15 s at 95°C for denaturation and 2 min at 62°C for annealing and extension. For each gene, three independent PCRs from the same reverse transcription sample were examined.

Total cell lysates of C3H10T1/2 treated with BMP2 (300 ng/ml) were subjected to SDS-PAGE and examined by western blotting with either the mouse monoclonal anti-ATF6 antibody (diluted 1:500; BioLegend) or rabbit polyclonal anti-Col-X (diluted 1:500; Santa Cruz Biotechnology) for 1 h, then followed by HRP-conjugated anti-mouse immunoglobulin G (IgG) and anti-rabbit-IgG antibodies at a 1:1000 dilution. After washing, the signals were detected using the enhanced chemiluminescence system (Amersham Biosciences).

Nuclear proteins from C3H10T1/2 cells infected with Ad-Runx2 or Ad-Sox6 were extracted as described previously, according to the manufacturer's protocol (Nuclear/Cytosol Fractionation Kit, Biovision, CA). After 48 h, the C3H10T1/2 cells were scraped with cold PBS from the plate to 10 ml tubes, and centrifuged at 1000 g for 10 min. The cell pellet was resuspended in 400 µl of cold buffer with protease inhibitors (Roche Applied Science) and placed on ice for 15 min. Then 25 µl of 10% Nonidet P-40 was added to the cell pellet, which was vortexed vigorously for 10 s and centrifuged for 1 min at 25,000 g, and the pellet was kept. The nuclear pellet was resuspended in 50 ml ice-cold buffer (same as above) and the tube was rocked for 30 min at 4°C. It was then centrifuged for 15 min at 25,000 g at 4°C. Oligonucleotides containing the first Runx2-binding site, from −240 to −233, and the first Sox6-binding site, from −294 to −285, within the 5′-flanking region of the ATF6 promoter were synthesized. The probes were labeled with digoxigenin (DIG)-11-ddUTP, and EMSAs were performed using a DIG gel shift kit (Roche Applied Science). Competition experiments were performed by preincubating nuclear extract with excess unlabeled probes before adding labeled oligonucleotides. In supershift assays, 5 µg of anti-Runx2 or anti-Sox6 antibody (Santa Cruz Biotechnology) were incubated with the reaction mixture for 15 min before the addition of the digoxigenin-labeled probe. Reaction mixtures were incubated for 20 min at room temperature. Samples were subjected to electrophoresis on a native 5% polyacrylamide gel run in 0.5× TBE for 2.5 h at 100 V. The signal was detected using a chemiluminescent detection system (Roche Applied Science).

C3H10T1/2 cells infected with Ad-Runx2 or Ad-Sox6 or treated with BMP2, were fixed by 1% formaldehyde for 10 min before cell lysis. Then cells were lysed, the chromatin was subsequently sonicated and protein–DNA complexes were immunoprecipitated using IgG (control) as well as anti-Runx2 antibody (1:500; Abcam). The DNA recovered from the immunoprecipitation was then amplified by PCR using primers that span the Runx2-binding site and the Sox6-binding site of the ATF6 gene promoter. The sequences of primers are as follows: sense 5′-GTTTAGTCGAATTGATGTCTGCG-3′ and antisense 5′-CAGAGACTAAGCAAATTTGA-3′ (for the first Runx2-binding site); and sense 5′-CGCTGTGGCATTAAGAAGGA-3′ and antisense 5′-GGGAAGACACGCAGACATCA-3′ (for the first Sox6-binding site). The input (1% of the supernatant) was used in PCR as a positive control. PCR was performed under the following conditions: 94°C for 5 min, 35 cycles at 94°C for 30 s, 56°C for 30 s and 72°C for 45 s.

C3H10T1/2 and ATDC5 cells were plated at a density of 3×10⁵ cells/well in six-well tissue culture plates, and were transfected with ATF6-specific reporter plasmids [pGL3-ATF6-luc, pGL3-ATF6(mut1)-luc, or pGL3-ATF6-(mut2)-luc], pcDNA3.1(-)-Runx2, pcDNA3.1(-)-Sox6 and pCMV-gal (an internal control for transfection efficiency). At 48 h after transfection, cells were harvested, and luciferase and β-galactosidase activity was measured using the Bioscan Mini-Lum luminometer. Relative transcriptional activity was expressed as a ratio of luciferase reporter gene activity from the experimental vector to that from the internal control vector. The cultures were processed and analyzed as described above.

Fetal mouse metatarsals were extracted from 15-day-old pregnant fetal FVB/N mice and cultured in Dulbecco's modified Eagle's medium (DMEM; Gibco) including 1% heat-inactivated fetal calf serum (FCS) and 100 U/ml penicillin-streptomycin per ml with or without various stimuli for 5 days, as shown in Fig. 5. For Safranin O and Fast Green staining, and Alizarin Red or Alcian Blue staining, the methods were as explained previously (Feng et al., 2010; Guo et al., 2014a,b).

Chondrogenic differentiation was induced by placing 2.5×10⁵ hMSCs into the defined chondrogenic medium [e.g. that described by Penick (Penick et al., 2005)]. and subjecting the cells to gentle centrifugation (800 g for 5 min) in a 15-ml conical polypropylene tube, after which the cap was loosened, and the tube was placed in the incubator, where the cells adhered to one another and consolidated into a cell pellet within 24 h. After 3 to 4 days, the hMSCs had formed a 1-mm ball in the bottom of the tube. The chondrogenic medium was replaced with medium containing fresh BMP2, Ad-ATF6 or Ad-ATF6a every 3–4 days, and the medium was changed by careful aspiration, because the cell pellets were free floating. After 3 weeks, for histological analysis, the pellets were fixed in 4% formaldehyde, paraffin-embedded, sectioned and analyzed by immunostaining for Col II or Col X. Sections were also be stained with 0.1% Safranin O for detection of proteoglycans (Feng et al., 2010).

Micromass cultures of C3H10T1/2 cells or BMSCs were treated with BMP2 (300 ng/ml). Cells were harvested after incubation in DMEM with 10% FCS for 5 days, and then they were lysed in lysis buffer (1% NP-40, 50 mmol/l Tris-HCl, pH 7.2, 0.15 mol/l NaCl, 0.01 mol/l sodium phosphate, pH 7.2, 1% Trasylol, and protein inhibitor cocktail, Sigma-Aldrich). Approximately 500 μg of protein samples were incubated with anti-Runx2 (25 μg/ml; Santa Cruz Biotechnology) or control rabbit IgG (25 μg/ml) antibodies for 1 h, followed by incubation with 30 μl of protein-A–agarose (PerkinElmer Life Sciences) at 4°C overnight. The immunoprecipitated complex was detected using western blotting with anti-ATF6a antibody.

Results are expressed as mean±s.d. from at least three independent experiments. One-way ANOVA was performed using R software to determine the significant differences (F>3.35, α<0.05) of the activity among different doses. In addition, Tukey's test was also used in conjunction with an analysis of variance to find significant differences (P<0.05; P<0.01) of the levels of genes of interest.