XBP1S, a BMP2-inducible transcription factor, accelerates endochondral bone growth by activating GEP growth factor

We previously reported that transcription factor XBP1S binds to RUNX2 and enhances chondrocyte hypertrophy through acting as a cofactor of RUNX2. Herein, we report that XBP1S is a key downstream molecule of BMP2 and is required for BMP2-mediated chondrocyte differentiation. XBP1S is up-regulated during chondrocyte differentiation and demonstrates the temporal and spatial expression pattern during skeletal development. XBP1S stimulates chondrocyte differentiation from mesenchymal stem cells in vitro and endochondral ossification ex vivo. In addition, XBP1S activates granulin-epithelin precursor (GEP), a growth factor known to stimulate chondrogenesis, and endogenous GEP is required, at least in part, for XBP1S-stimulated chondrocyte hypertrophy, mineralization and endochondral bone formation. Furthermore, XBP1S enhances GEP-stimulated chondrogenesis and endochondral bone formation. Collectively, these findings demonstrate that XBP1S, a BMP2-inducible transcription factor, positively regulates endochondral bone formation by activating GEP chondrogenic growth factor.


Introduction
During foetal development of the mammalian skeletal system, the majority of bones form through a process of endochondral ossification. Chondrocytes in the primary centre of ossification begin to grow. Elaborate chondrogenesis is controlled exquisitely by cellular interactions with the growth factors, surrounding matrix proteins and other environmental factors that mediate cellular signalling pathways and transcription of specific genes in a temporal-spatial manner [1][2][3]. Production of and response to different growth factors are observed at all times, such as transforming growth factor-b (TGF-b) superfamily and bone morphogenic protein (BMP) subfamily. BMP2 is one of the most important cytokines and plays several important roles in a variety of cellular functions ranging from embryogenesis, cell growth, and differentiation to bone development and the repair of bone fractures [4,5]. Jang et al. [6] reported that BMP2 activates UPR transducers, such as PERK (PKR-like ER-resistant kinase), OASIS and ATF6 (activating transcription factor 6). BMP2 induces osteoblast differentiation through Runx2-dependent ATF6 expression, which directly regulates osteocalcin transcription. OASIS [7], a member of the CREB/ATF family, activates the transcription of Col1a1 through an unfolded protein response element (UPRE)-like sequence in the osteoblast-specific Col1a1 promoter region. The expression of OASIS in osteoblasts is induced by BMP2, the signalling of which is required for bone formation.
Human XBP1 (X-box-binding protein 1) is a signalling molecule downstream of IRE1 in the IRE1-XBP1 pathway of the UPR and participates in IRE1a-mediated UPR signal transmission. In eukaryotic cell, IRE1 is activated by ER stress and subsequently processes XBP1 mRNA to generate the spliced form of XBP1 protein (XBP1S). XBP1 exists in two forms: XBP1S and XBP1U (XBP1 unspliced form) isoforms [8][9][10]. Tohmonda [11] reported that inositol-requiring protein 1a (IRE1a), one of the most crucial UPR mediators, and its target transcription factor XBP1 is essential for BMP2-induced osteoblast differentiation. Osterix (Osx, a transcription factor that is indispensible for bone formation) is a target gene of XBP1. The IRE1a-XBP1 pathway is involved in osteoblast differentiation through promoting Osterix transcription by XBP1. Although there is some evidence that XBP1 plays an important role in the control of cell proliferation and the differentiation of numerous types of cells and tissues, including adipogenesis, myelomapathogenesis, skeletal muscle myotubes and dendritic cells in ER stress [12][13][14][15], little is known about the modulation and physiological significance of XBP1S in chondrocyte development and bone formation. Specifically, the molecular mechanism by which XBP1S regulates chondrogenesis also remains unknown.
Granulin-epithelin precursor (GEP), also referred to as pro-granulin, acrogranin, was first purified as a growth factor from conditioned tissue culture media. Granulin-epithelin precursor is a 593 amino acid secreted glycoprotein with an apparent molecular weight of 80 kD [16][17][18]. Granulin-epithelin precursor is secreted in an intact form and undergoes proteolysis, leading to the release of its constituent peptides, the granulins [19,20]. Granulin-epithelin precursor contains 7.5 repeats of a cysteine-rich motif (CX 5-6 CX 5 CCX 8 CCX 6 CCXD X 2 HCCPX 4 CX 5-6 C) in the order P-G-F-B-A-C-D-E, where A-G are full repeats and P is a half motif. Granulin-epithelin precursor is remarkably expressed in rapidly cycling epithelial cells, in chondrocytes [21][22][23], in the immune system cells, in neurons and in some human cancers [24][25][26][27]. Increasing evidence has implicated GEP in the regulation of differentiation, development and pathological processes. It has been isolated as a differentially expressed gene from macrophage development [28], skeletal muscle differentiation [29] and synovium (morphopathogenesis) in rheumatoid arthritis and osteoarthritis [22,30]. Granulin-epithelin precursor was also shown to be a critical mediator of wound response and tissue repair [31,32]. We previously reported that GEP regulates chondrocyte differentiation and endochondral bone formation, and cartilage repair through Erk1/2 signalling and its target gene, including JunB transcription factor [21].
In this study, we attempt to determine whether XBP1S is essential for skeletal development using both in vitro and in vivo approaches. Second, we studied its upstream and downstream molecules during chondrogenesis, as well as its molecular mechanisms by which XBP1S regulates chondrogenesis. Our results support a novel role of XBP1S, a key downstream molecule of BMP2 in the control of chondrogenesis and endochondral bone growth through activating GEP growth factor.
To generate XBP1S small interfering RNA (siRNA) expression constructs, siRNA corresponding to the coding sequence of the XBP1S gene (5 0 -ATGCCAATGAACTCTTT CCCTTTT-3 0 ) was cloned into a pSES-HUS vector (an adenoviral shuttle vector expressing siRNA) according to the manufacturer's instructions. Briefly, equimolar amounts of complementary sense and antisense strands were separately mixed, annealed and slowly cooled to 10°C in a 50-ll reaction buffer (100 mM NaCl and 50 mM HEPES, pH 7.4). The annealed oligonucleotides were inserted into the SfiI sites of pSES-HUS vector. All constructs were verified by nucleic acid sequencing; subsequent analysis was performed with BLAST software (National Institutes of Health, Bethesda, MD, USA).

Mice
All animal studies were performed in accordance with institutional guidelines and approval by the Institutional Animal Care and Use Committee of Chongqing Medical University. The GEP-knockout (GEP À/À ) mice were bought from Jackson Laboratories (Bar Harbor, ME, USA), the generation and genotyping of GEP À/À mice on basis of Jackson Laboratory's protocol were used for these experiments (http://jaxmice.jax.org/query/).

Isolation and culture of mouse bone marrow stromal cells (BMSCs)
Mouse bone marrow was isolated by flushing the femurs and tibiae of 8to 12-week-old female GEP À/À knockout (GEP KO) mice with 0.6 ml of improved minimal essential medium (Sigma-Aldrich, St. Louis, MO, USA), supplemented with 20% foetal bovine serum (FBS), 100 units/ml penicillin, 100 lg/ml streptomycin (Invitrogen) and 2 mM glutamine (Invitrogen, Carlsbad, CA, USA), and then it was filtered through a cell strainer (Falcon, BD Biosciences, Franklin Lakes, NJ, USA). Cells were centrifuged for 10 min. at 260 9 g, washed by the addition of fresh medium, centrifuged again, resuspended and plated out in improved minimal essential medium supplemented with 20% FBS, 100 units/ml penicillin, 100 lg/ml streptomycin and 2 mM glutamine at a density of 2 9 10 6 cells/cm 2 in 25-cm 2 plastic culture dishes. The cells were incubated at 37°C in 5% CO 2 . After 72 hrs, non-adherent cells and debris were removed, and the adherent cells were cultured continuously. Cells were grown to confluence, washed with PBS and lifted by incubation with 0.25% trypsin, 2 mM ethylenediaminetetraacetic acid (Invitrogen) for 5 min. Nondetached cells were discarded, and the remaining cells were regarded as passage 1 of the BMSC culture. Confluent BMSCs were passaged and plated out at 1:2-1:3 dilutions. At passage 3, cells were transferred to DMEM (Invitrogen) supplemented with 10% FBS for differentiation studies.

Cell culture
The micromass culture was performed as described previously [46]. Briefly, trypsinized C3H10T1/2 cells were resuspended in DMEM with 10% FBS at a concentration of 10 6 cells/ml, and six drops of 100 ll of cells were placed in a 60-mm tissue culture dish (BD Biosciences). After a 2-hr incubation at 37°C, 1 ml of DMEM containing 10% FBS and BMP2 protein (300 ng/ml) was added. The medium was replaced approximately every 2-3 days. To test the effect of overexpression of XBP1S protein on chondrogenesis, C3H10T1/2 cells were infected with XBP1S expression adenovirus or control GFP adenovirus before micromass culture.
To test the effect of knocking down XBP1S on chondrogenesis, C3H10T1/2 cells were infected with Ad-XBP1S siRNA or control RFP adenovirus before micromass culture. Mouse chondrogenic ATDC5 cells were maintained in a medium consisting of a 1:1 mixture of DMEM and Ham's F-12 medium (Flow Laboratories, Irvine, UK) containing 5% FBS (Invitrogen), 10 mg/ml of human transferrin (Roche Applied Science, Penzberg, Germany) and 30 nM of sodium selenite (Sigma-Aldrich) at 37°C in a humidified atmosphere of 5% CO 2 in air. The ATDC5 cells were seeded at a density of 3 9 10 5 cells/well in 6-well cell culture plates (Corning Life Sciences, Edison, NJ, USA). The medium was replaced every other day. For adenovirus (Ad-XBP1S or Ad-GFP) infection and Ad-XBP1S siRNA and Ad-RFP infection, the same protocol as used with C3H10T1/2 cells was followed.

Immunohistochemistry
Sections of post-coital day 12.5, 14.5, 15.5, 17.5 and 18.5 embryos and newborn mice were deparaffinized, rehydrated and placed in Tris buffer [10 mM Tris-HCl (pH 8.0), 150 mM NaCl]. Serum block was applied for 30 min. at room temperature before incubation of the primary antibody. Antimouse XBP1S (BioLegend, San Diego, CA, USA) was diluted 1:50, and sections were incubated at room temperature for 2 hrs. For detection, biotinylated secondary antibody and horseradish peroxidase (HRP)-streptavidin complex (Santa Cruz Biotechnology Inc., Santa Cruz, CA, USA) were used. Horseradish peroxidase substrate was used for visualization, and sections were then counterstained with Mayer's haematoxylin.

Quantitative real-time PCR
To examine the effects of chondrogenesis by BMP2 and XBP1S, C3H10T1/2 or ATDC5 cells were plated at a density of 3 9 10 5 cells/well in 6-well tissue culture plates. 300 ng/ml BMP2 or Ad-XBP1S (MOI 20) was then treated into these cells respectively. After day 3 or day 7, total RNAs were isolated using the RNeasy minikit (Qiagen, Hilden, Germany) and reverse transcribed into cDNA. Real-time PCR was performed with an ABI 7400 system using the TaqMan EZ RT-PCR kit according to the manufacturer's protocol. TaqMan primers and probes were derived from the commercially available TaqMan assay-on-demand gene expression products. We select GAPDH as the endogenous control for the real-time PCR relative quantification analysis. The following pair of oligonucleotides was used as internal controls: 3 0 -GTTTAGGCAAGTGTGGCTGGA-5 0 and 3 0 -ACTGGAGTTGATGTACCAGATGT-5 0 for mouse GAPDH; 3 0 -GTGG TGGAAGAACTACAGTA-5 0 and 3 0 -GTTCGAGTAAAGGACCAT CA-5 0 for human GAPDH. 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 sec. at 95°C for denaturation and 2 min. at 62°C for annealing and extension.

Reporter gene assays
Micromass culture of ATDC5 cells were plated at a density of 3 9 10 5 cells/well in 6-well tissue culture plates and transfected with XBP1-specific reporter plasmids (pGL3-XBP1-luc) and pCMV-gal (an internal control for transfection efficiency). Forty-eight hours after transfection, cells were harvested, and luciferase and b-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 analysed as described above.

Chromatin immunoprecipitation
Micromass culture of ATDC5 cells treated with BMP2 was fixed by 1% formaldehyde for 10 min. before cell lysis. Cell lysates were subsequently sonicated, followed by centrifugation. The input (1% of the supernatant) was used in PCR as a positive control. The supernatant was then pre-cleared using protein A-agarose/salmon sperm DNA for 30 min. at 4°C. After centrifugation, the supernatant was then used for immunoprecipitation using an anti-Smad4 antibody or the control IgG antibody and incubated overnight at 4°C. The protein-DNA complex was subsequently incubated with protein Aagarose/salmon sperm DNA for 1 hr at 4°C. The immune complex was collected by centrifugation and then washed five times with the following for 5 min. each: once with low salt immune complex wash buffer, once with high salt immune complex wash buffer, once with LiCl salt immune complex wash buffer and twice with TE buffer. The histone-DNA complex was eluted from the antibody using elution buffer (1% SDS, 0.1 M NaHCO 3 ), and 5 M NaCl was added to reverse the histone-DNA cross-link by heating for 4 hrs at 65°C. The DNA was then extracted with phenol/chloroform and precipitated with ethanol in the presence of glycogen (20 mg) as a carrier. The precipitate was used as a template for PCR amplification. For PCR of the XBP1S minimal promoter region using the chromatin-immunoprecipitated DNA, one-tenth of the DNA was PCR amplified using forward primer, 5 0 -CAATGGACG CCGAGCTCG-3 0 ; and reverse primer, 5 0 -CATAGCTCCAGACTAC GC-3 0 . Thirty-five cycles of PCR at 94°C for 30 sec., 55°C for 30 sec. and 72°C for 30 sec. were performed. PCR products were analysed by 1% agarose gel.

Culture of foetal mouse bone explants
Foetal mouse metatarsals were dissected from foetal GEP null mice (GEP À/À , 15-day-old embryos) and cultured in DMEM (Gibco, Carlsbad, CA, USA) containing 1% heat-inactivated foetal calf serum (Invitrogen) and 100 U penicillin-streptomycin per milliliter in the absence or presence of various stimuli for 5 days, as indicated in Figures 4,8 and 9. For alizarin red and alcian blue staining (staining for bone and cartilage), the explants were placed in 4% paraformaldehyde in phosphatebuffered saline for overnight fixation. Subsequently, explants were placed in staining solution (0.05% alizarin red, 0.015% alcian blue, 5% acetic acid in 70% ethanol) for 45-60 min. Digital images of stained bones were analysed. For safranin O-fast green staining, explants were fixed in 96% alcohol and processed for paraffin embedding. Sections were stained with 0.1% safranin O (orange stain) to evaluate cartilage matrices and with 0.03% fast green to evaluate morphological features as previously described [57].

Immunocytostaining of GEP induced by XBP1S
To reveal the induction of GEP by XBP1S, the C28I2 cells were transiently transfected with the pcDNA3.1(-) expression vector containing cDNA encoding XBP1S. Then, detection of the expression of GEP in the control and XBP1S-treated C28I2 cells was performed, the cells were washed, fixed with 100% methanol in the freezer compartment for 5 min., washed twice in 4°C phosphate-buffered saline for 5 min. and then incubated with 30% goat serum in phosphate-buffered saline for 30 min.; the cells were incubated with primary antibodies (i.e. mouse monoclonal anti-GEP antibodies) at room temperature for 1 hr. After being washed with phosphate-buffered saline, the coverslips were incubated with secondary antibodies (i.e. goat antimouse IgG conjugated with rhodamine; Santa Cruz Biotechnology; diluted 1:100) and goat antirabbit IgG conjugated with fluorescein isothiocyanate (Santa Cruz Biotechnology; diluted 1:100) for 1 hr. The specimens were observed under a fluorescence microscope with appropriate optical filters. Microscopic images were captured by using the Image-Pro programme (Media Cybernetics, Sarasota, FL, USA) and an Olympus microscope. Images were arranged using Adobe Photoshop.

Statistical analysis
The statistical analysis was performed with SPSS 10.0.1 software (Chicago, IL, USA) for Windows. Data were expressed as mean AE SD from at least three independent experiments. Data for multiple variable comparisons were analysed by one-way ANOVA. P < 0.05 was deemed statistically significant.

Results
Differential expression of XBP1S in the chondrogenesis of a micromass culture of ATDC5 and C3H10T1/2 cells It is reported that ER stress signal molecules were associated with chondrogenesis [33][34][35]. In this study, we sought to determine whether XBP1S, a vital transcription factor in ER stress, participates in cartilage development. We first studied XBP1S expression profiles during chondrocyte differentiation using the ATDC5 cell line and C3H10T1/2 cell line [36][37][38].
It differentiates specifically to the cartilage lineage at high yields when inoculated under high-cell-density micromass cultures as well as when exposed to chondroinductive factors such as a well-documented growth factor BMP2 [5,39]. Therefore, both ATDC5 and C3H10T1/2 cells have the potential to become chondrocytes, making them a valuable in vitro correlate for studying the mechanisms of chondrogenesis. To obtain XBP1S expression profiles during chondrocyte differentiation, micromass cultures of ATDC5 and C3H10T1/2 cells were incubated in the presence of 300 ng/ml of recombinant BMP2 for induction of chondrocyte differentiation. Cells were harvested at various time-points and then followed by real-time PCR for measures of XBP1S and collagen X (a specific marker for hypertrophic chondrocytes). As shown in Figure 1A and B, the level of XBP1S mRNA was relatively low until day 5; when it had doubled, and thereafter remained at high levels during the differential stage, representing terminal differentiation marked by the increase in collagen X expression. In addition, similar results were also observed in the course of chondrogenesis of C3H10T1/2 cells. It is noteworthy, that the peak level of XBP1S was 2 days earlier than that of collagen X, suggesting that XBP1S may regulate collagen X expression.
We next examined the level of XBP1S protein. Micromass culture of ATDC5 and C3H10T1/2 cells were harvested at various time-points, respectively, followed by Western blotting (Fig. 1C and D). XBP1S protein was markedly elevated at day 5 and thereafter, remained at high levels.

XBP1S expression patterns in chondrocytes during both embryonic and post-natal development stages
Next, we characterized the temporal and spatial expression pattern of XBP1S during skeletal development using an immunostaining assay at multiple time-points, including embryonic day 12.5 (E12.5; onset of chondrogenesis that begins with the proliferation and subsequent condensation of mesenchymal cells), E14.5 (right after cartilage formation but before endochondral bone formation) and E15.5 (onset of skeletal growth), as well as E17.5, E18.5 and newborn. As revealed in Figure 2, XBP1S is detected at E14.5, and its level is increased in the centre of the condensation and around it at E15.5. It demonstrates prominent expression in pre-hypertrophic chondrocytes at E15.5 and E17.5, E18.5 and in newborn mice. A high level of XBP1S throughout the whole growth plate is observed at E17.5, E18.5 and newborn mice, suggesting that the expression profile of XBP1S is closely linked to the entire chondrogenic period.
tion. We next sought to determine the role of XBP1S and BMP2 (300 ng/ml) during chondrogenesis in micromass cultures of prechondrogenic ATDC5 cells and BMSC cells, which are capable of differentiation into various lineages, including chondrocytes [36][37][38].
In brief, the high-density culture system was incubated in the absence (CTR) or presence of Ad-XBP1S or 300 ng/ml BMP2 (serving as a positive control) for 3 or 7 days. Chondrogenesis was monitored by analysing the expressions of marker genes specific for chondrocytes (Fig. 3). ATDC5 or BMSC cells were treated with BMP2, adenovirus encoding XBP1S (Ad-XBP1S), Ad-XBP1S+BMP2 and control GFP (Ad-GFP), respectively, then, RNA was extracted every other day for real-time PCR.
As revealed in Figure 3A-F, chondrocyte differentiation was monitored by examining the expression of collagen II, collagen X and RUNX2, three marker genes widely used for chondrocyte maturation and hypertrophy [4,5]. As for BMP2, XBP1S markedly induced the expression of collagen II, collagen X and RUNX2. Besides, clearly enhanced expressions of collagen II, collagen X and RUNX2 in Ad-XBP1S+BMP2-treated cells were observed compared with those in BMP2-treated or Ad-XBP1S-treated cells, suggesting that XBP1S can enhance BMP2-induced chondrogenesis, thus, XBP1S is a positive mediator for chondrocyte differentiation and hypertrophy.
The effect of XBP1S on endochondral bone formation was then studied in an ex vivo model of 15-day-old foetal mouse metatarsal bones. At the time of explantation, these explants consisted of undifferentiated cartilage. In a 5-day culture period of Ad-XBP1S (MOI 20), these explants underwent all sequential stages of endochondral bone formation. As shown in Figure 4, XBP1S significantly stimulated chondrocyte hypertrophy, mineralization and bone length.
To testify the result of genome-wide DNA chip analysis, we tested a few cytokines known to be important for chondrogenesis in the primary human chondrocytes. Our results showed that XBP1S mRNA is up-regulated threefold by BMP2 and 1.5-fold by TGF-b. IL-1b had no apparent effects on XBP1S expression (Fig. 5B).
Next, we tested whether XBP1S was required for BMP2-mediated chondrogenesis using the siRNA approach. As shown in Figure 5C and E, infection with siXBP1S adenovirus resulted in 81% and 72%  Figure 5D and F, our realtime PCR assay showed that reductions of the endogenous XBP1S by siXBP1S adenovirus sharply decrease chondrogenic responses induced by BMP2: 77% down of Sox9, 62% down of collagen II, 83% down of collagen X and 75% down of RUNX2 compared with the control group responses in ATDC5 cells (Fig. 5D); 67% down of Sox9, 60% down of collagen II, 78% down of collagen X and 65% down of RUNX2 compared with the control group responses in BMSC cells (Fig. 5F). These results support the concept that XBP1S is a key downstream molecule of BMP2 during chondrocyte development.

BMP2 and Smads activate XBP1S-specific reporter genes
To elucidate the molecular mechanism by which BMP2 activates XBP1S expression, firstly, four XBP1S-specific reporter gene plasmids, À2000?+133XBP1Sluc [labelled p1], À1311?+133 XBP1Sluc [p2], À407?+133XBP1Sluc [p3] and À2000?À407 XBP1Sluc [p4], were generated in which segments of the XBP1S promoter, with or without the NF-Y or NF-Y/ERSE binding site, were inserted upstream of the luciferase coding region of the pGL3 basic vector (Fig. 6). On the other hand, deletion of the region from À407 to +133 leads to the complete loss of the reporter activity, indicating that this region is probably the basic promoter of the XBP1S gene. The core sequence of XBP1S promoter is found from À407 to +133 bp. Applications of BMP2 were able to activate all XBP1S promoter constructs containing the region between À407 and +133. Furthermore, this basic promoter region directly responded to BMP2 (Fig. 6A).
Because BMP2 activates the cellular signalling through Smads, we then tested interaction of Smad4, a coregulatory Smad that binds to Smad1 or Smad5 for transducing BMP2 signalling, with XBP1S promoter regions (in particular, the region of À407 and +133) in vitro using the ChIP assay. As shown in Figure 6B, we observed a clear PCR product using DNA isolated from immunoprecipitated complexes with anti-Smad4 antibodies from BMP2-treated cells, but not from BMP2-untreated cells, suggesting that the Smad4 is recruited into this XBP1S promoter region after exposure to BMP2.
Next, we determined whether Smad transcription factors could directly activate the XBP1S at the transcription level. Cotransfection of the XBP1S luciferase plasmid (À407XBP1Sluc) with an expression plasmid encoding either Smad1, Smad4 and Smad5 (cDNA constructs kindly provided by Dr. Chuanju Liu, Department of Orthopaedic Surgery and Department of Cell Biology, New York University School of Medicine), or a combination of either Smad1/Smad4 or Smad5/Smad4, markedly increased the expression of the XBP1S reporter gene. Both of the combinations of Smad1/Smad4 and combinations of Smad4/Smad5 gave the higher value than the others (Fig. 6C). The above data support the notion that BMP2 controls XBP1S expression through Smad signalling.

XBP1S induces GEP expressions in C3H10T1/2 and ATDC5 cells
We have found that XBP1S is expressed throughout the whole growth plate at E17.5, E18.5 and newborn mice (Fig. 2) and positively regulates chondrocyte development (Figs 3 and 4). We previously reported that GEP is a key downstream molecule of BMP2, and it is required for BMP2-mediated chondrocyte differentiation. We next used C3H10T1/2 cells and chondroprogenitor ATDC5 cells to examine the relationship between GEP and XBP1S. Micromass cultures of both C3H10T1/2 and ATDC5 cells pre-treated with 300 ng/ml of BMP2 for 1 week were cultured with or without Ad-XBP1S for various timepoints, and the level of GEP mRNA was measured by using real-time PCR (Fig. 7A).
Granulin-epithelin precursor mRNA was increased to 2.0-fold at day 1 and to 2.3-fold by day 3 in the XBP1S-untreated control ATDC5 cells. XBP1S markedly enhanced the level of GEP mRNA to 4.2-fold at day 1 and to 5.7-fold by day 3 in the XBP1S-treated ATDC5 cells. In the case of C3H10T1/2 cells, GEP mRNA was slightly increased to 1.3-fold at day 1 and to 1.5-fold by day 3 in the XBP1S-untreated cells; And XBP1S significantly induced GEP to 3.6-fold at day 1 and to 3.5-fold by day 3 in the XBP1S-treated C3H10T1/2 cells. GEP mRNA in the XBP1S-treated cells was approximately twofold higher than the mRNA in the control at the same time-point. In addition, induction of the GEP protein level by XBP1S was also visualized by both immunofluorescent cell staining in C28I2 chondrocytes (Fig. 7C) and immunoblotting in ATDC5 cells (Fig. 7B). Taken together, these findings demonstrate that GEP is a XBP1S-inducible gene in the process of chondrogenesis.
In addition, the dependence on GEP of XBP1S-mediated endochondral bone formation was revealed by using cultures of 15-dayold foetal GEP null mice metatarsal bones (Fig. 8D). In line with a previous report [46], XBP1S potently enhanced chondrocyte hypertrophy; and the effect of XBP1S-induced chondrogenesis and endochondral bone formation was largely abolished in GEP À/À BMSC cells. These results indicated that XBP1S-mediated chondrocyte differentiation and endochondral bone growth depends, at least in part, on GEP. We next sought to determine whether XBP1S recovered the valid stimulating in growth plates of GEP À/À embryos rescued by GEP with safranin O-fast green staining. As shown in Figure 8D, disorganized GEP null growth plates, including reductive chondrocyte hypertrophy, cannot be changed by Ad-XBP1S, however, it can be largely corrected in the presence of Ad-XBP1S+Ad-GEP. XBP1S recovered the potent stimulating effect of chondrocyte differentiation, mineralization and endochondral bone growth in GEP null growth plates rescued by GEP.
Taken together, endogenous GEP is required for XBP1S-stimulated chondrocyte hypertrophy, mineralization and endochondral bone formation.

XBP1S enhances the chondroinductive activity of GEP
Then, we examined whether XBP1S-mediated augment of chondrocyte hypertrophy and endochondral bone growth is exerted by activating GEP's chondroinductive activity. We previously reported that GEP is a novel growth factor increasing chondrocyte differentiation and endochondral bone formation, and cartilage repair [21,22]. For this purpose, we first examined whether XBP1S was able to increase GEP-stimulated chondrocyte hypertrophy using chondroprogenitor ATDC5 cells. As noted in Figure 9A, XBP1S contains a DNA binding domain (a domain) and a transactivating domain (b domain; Fig. 9A, top scheme). We generated XBP1S derivatives with mutations in the DNA binding domain (XBP1Smt-a), the transactivating domain (XBP1Smt-b) or both domains (XBP1S mt-a/b; Fig. 9A, three lower schemes).
Then, ATDC5 cells pre-treated with BMP2 for 1 week were cultured without or with Ad-GEP (MOI 20), Ad-XBP1S (MOI 20) or its point mutation Ad-XBP1Smt-a, Ad-XBP1Smt-b, Ad-XBP1Smt-a/-b or various combinations as indicated in Figure 9B. Realtime PCR results showed that GEP-stimulated Col X and MMP-13 expressions were remarkably increased by the addition of Ad-XBP1S. In BMP2-induced ATDC5 cells infected with Ad-XBP1S, GEP-induced Col X and MMP-13 expression were increased~2.2-fold, whereas XBP1S in which either of the DNA binding domain and the transactivating domain were mutated, failed to enhance the GEP-induced Col X and MMP-13 expression ( Figure 9B). This augment of GEP action depends on XBP1S activity, as the XBP1S point mutation failed to do so. These results indicate that both the DNA binding domain and the transactivating domain of XBP1S are needed for stimulating GEPdependent chondrogenesis.
We next determined whether XBP1S was also able to improve the GEP activity in regulating endochondral bone growth. As expected, GEP growth factor stimulated chondrocyte maturation, mineralization and bone growth; and GEP-mediated endochondral bone growth was clearly increased by the addition of Ad-XBP1S (Fig. 9C). These observations, together with the finding that GEP is required for the XBP1Sinduced chondrocyte differentiation and endochondral bone formation, suggested that XBP1S positively regulates chondrocyte hypertrophy and endochondral bone growth through stimulating with GEP and activating its chondrogenic activity.

Discussion
Growth and development of endochondral bones is regulated through the well-orchestrated proliferation and differentiation of growth plate chondrocytes. Chondrogenesis is a process that is important for cartilage remodelling both during embryogenesis and in adult life [47,48]. The IRE1/XBP1 branch of the UPR is known to be essential for normal development. XBP1S is required for the terminal differentiation of B cells, hepatocytes and pancreatic b cells. It is also important for myeloma cells to survive hypoxic stress [49,50]. Many studies have shown that factors influencing cell fate and/or differentiation are activated in ER stress [51,52], but how such changes impact differentiation programmes in chondrocytes is poorly understood. Therefore, to test a link between the IRE1/XBP1 branch of the UPR and chondrocyte differentiation, we focused on the role of XBP1S in chondrogenesis as well as the molecular mechanism involved. Our results showed that XBP1S protein was highly induced in the course of BMP2-stimulated chondrogenesis in vitro (Fig. 1) and also demonstrated prominent expression in the entire growth plate chondrocyte population in vivo (Fig. 2). Real-time PCR for measurements of XBP1S showed that the level of XBP1S mRNA was relatively low until day 5, and at day 7, it tripled and thereafter remained at high levels during the late differential stage (Fig. 1A and B). The different expression between the protein and mRNA of XBP1S during chondrogenesis suggests that post-transcription regulations, such as mRNA stability, translation and protein degradation, might be also important in the control of XBP1S expression during chondrogenesis. The in vitro, ex vivo and in vivo studies support the concept that XBP1S is a potent stimulator of chondrocyte differentiation, mineralization and endochondral bone growth (Figs 3 and 4).
Saito et al. [35] reported that BMP2 induced ER stress in osteoblasts, and ER stress-inducing agents activate the IRE1a/b proteins. IRE1a, a kind of ER type I transmembrane protein containing a serine/threonine kinase module and an endoribonuclease domain, executes site-specific cleavage of XBP1 mRNA to remove a 26-nucleotide intron during UPR. XBP1S is more potent as a transcriptional activator and more stable than XBP1U (unspliced). XBP1S activates the promoters of many genes, including those coding for enzymes necessary for the degradation of improperly folded ER proteins, and participates in cell proliferation and differentiation [53,54].
Firstly, our work also supports the concept that XBP1S is a key downstream molecule of BMP2 in chondrogenesis and endochondral . The explants were cultured for 5 days, and safranin O-fast green staining was observed by using low-power (a) or high-power (b) microphotography. (c) Alizarin red S and alcian blue staining of metatarsals. Metatarsals were cultured as described above and processed for alizarin red S and alcian blue staining; a representative photograph is presented. (d) Per cent increase in total and mineralization length of metatarsal bones. Metatarsals were cultured as described above, total or mineralization length was determined, and the per cent increase was calculated (per cent increase = [length at day 5 À length at day 0]/length at day 0). Asterisk indicates a significant difference from the control (P < 0.05). bone growth based on the following evidence. (i) Both BMP2 and XBP1S are potent in inducing in vitro chondrogenesis and induction of chondrogenetic markers such as collagen II, collagen X and RUNX2 (Fig. 3). The effect of XBP1S on chondrogenesis is similar to the effects of BMP2 and differs significantly from many factors that have opposite effects on collagen II and collagen X [40,47,48]; (ii) BMP2 induced XBP1S in chondrocytes, as shown in (Fig. 5); (iii) Notably, knockdown of XBP1S strongly inhibited BMP2-mediated chondrogenesis, as assayed by collagen II, SOX9, collagen X and RUNX2 expression in the course of chondrocyte differentiation (Fig. 5); (iv) Finally, BMP2 activates XBP1S specific reporter genes through Smad transcription factors (Fig. 6).
Granulin-epithelin precursor, as a growth factor, has been linked to development, tissue regeneration, tumourigenesis and inflammation [26,31,32,55,56]. We previously reported that GEP accelerates chondrocyte hypertrophy, mineralization and endochondral bone growth through Erk1/2 signalling and its target gene, including JunB transcription factor [21]. Herein, we present evidence showing that XBP1S induces GEP expression in the course of chondrocyte development (Fig. 7), and endogenous GEP is required for XBP1S-stimulated chondrocyte hypertrophy, mineralization and endochondral bone formation. Our previous report showing that (i) XBP1S cannot improve chondrocyte differentiation and endochondral bone formation in GEP À/À BMSC cells, and (ii) XBP1S recovered the potent stimulating effect of chondrocyte differentiation, mineralization and endochondral bone growth in GEP null growth plates rescued by GEP (Fig. 8). In addition, XBP1S increases GEP-mediated chondrocyte maturation, mineralization and endochondral bone growth. XBP1S enhances chondrocyte differentiation and endochondral bone formation through activating the chondrogenic activity of GEP (Fig. 9). (c) Alizarin red S and alcian blue staining of metatarsals. The explants were fixed and processed for alizarin red S and alcian blue staining; a representative photograph is presented. (d) Per cent increase in total and mineralization length of metatarsal bones. Metatarsals were cultured as described above, the total or mineralization length was determined and the per cent increase was calculated (per cent increase = [length at day 5 À length at day 0]/length at day 0). Asterisks indicate a significant difference from the control (P < 0.05).
Recently, we reported that ADAMTS-7 binds to and degrades COMP and that COMP interacts with GEP and potentiates GEP-stimulated chondrocyte functions, indicating that ADAMTS-7, GEP and COMP form an interaction and interplay network in regulating chondrocyte functions [57][58][59]. It remains to be determined how the interaction network among ADAMTS-7 metalloproteinase, GEP growth factor and COMP extracellular matrix molecule acts in concert in regulating chondrocyte differentiation and endochondral ossification. Our current study focuses on the relationship between XBP1S transcription factor, a UPR signal molecule, and GEP growth factor in chondrocyte development for the first time.
On the basis of the data in the literature [3,7,9,39], our earlier findings [21,22,46] and the results of this study, we propose a model for the role of XBP1S -specifically, its expression and function -in chondrocyte differentiation (Fig. 10). This study provides novel insights into the role of XBP1S, a novel mediator in the BMP2 pathway, in regulating chondrocyte differentiation and endochondral bone formation and sheds light on the molecular mechanism by which XBP1S positively regulates chondrogenesis; i.e. XBP1S, a key downstream molecule of BMP2, increases chondrocyte differentiation and endochondral bone formation through activating GEP growth factor, and endogenous GEP is required for XBP1S-stimulated chondrocyte hypertrophy, mineralization and endochondral bone growth. Our work supports a hypothesis that XBP1S, a transcription factor induced by BMP2, regulates chondrogenesis and endochondral bone formation through GEP growth factor. The elucidation of XBP1S's role and molecular events involved in chondrocyte differentiation will better our understanding of normal cartilage development and the pathogenesis of cartilage disease.