The molecular chaperone Hsp47 is essential for cartilage and endochondral bone formation

Cartilage and bone are primarily composed of type II and type I collagen bundles, respectively, with covalently bound collagen fibers (Eyre et al., 1986; Wu et al., 1987; Wu et al., 1992; Knott and Bailey, 1998; Chung et al., 2008). Collagen fibers form fibrillar networks in cartilage and interact with various extracellular matrix (ECM) components, including aggrecan and hyaluronan (Yaeger et al., 1997; Cortes et al., 2009; Matsumoto et al., 2009). Bones in mammals, except for the cranial, clavicle and some facial bones, are formed by endochondral ossification, a dynamic process in which cartilage is newly synthesized by chondrocytes, followed by the formation of hypertrophic cartilage after the differentiation of chondrocytes into hypertrophic chondrocytes, and finally bone formation by ossification of the hypertrophic cartilage region (Stickens et al., 2004; Mackie et al., 2008).

As many as 28 types of mammalian collagen have been reported and classified on the basis of their supramolecular structures: fibrillar collagens including types I, II, III, V, XI, XXIV and XXVII; fibril-associated collagens with interrupted triple helices (FACIT) including types IX, XII, XIV, XVI, XIX, XX, XXI, XXII and XXVI; collagen with transmembrane domains, types XIII, XVII, XXIII and XXV; network-forming collagen composed of types VIII and X; basement membrane collagen of type IV; and other types of collagen (Van der Rest and Garrone, 1991; Prockop and Kivirikko, 1995; Hubert et al., 2009). Collagen is composed of three α-chains, each of which contains characteristic Gly-x-y triplet repeats in their amino acid sequence. The three α-chains assemble at the C-terminus in the endoplasmic reticulum (ER) and form a triple helix structure from the C-terminus to the N-terminus (Bachinger et al., 1980; Engel and Prockop, 1991; Bulleid et al., 1997). The majority of the proline residues at the Y position of the Gly-x-y triplet are hydroxylated by prolyl 4-hydroxylase (P4H), which stabilizes the triple helices of procollagen. Mutations in fibrillar collagen genes have been reported to cause osteodysplasias such as osteogenesis imperfecta (OI), which is mainly the result of mutations in the COL1A1 and COL1A2 genes. In addition, mutations in cartilage-associated protein (CRTAP), 65 kDa FK506-binding protein (FKBP65 also known as FKBP10; a member of the collagen chaperone complex) and prolyl-3-hydroxylase-1 (P3H1, encoded by the LEPRE1 gene) result in OI (Baldridge et al., 2008; Vranka et al., 2004; Alanay et al., 2010). Mutations in collagen genes, result in either quantitative deficiencies or sequence alterations that lead to structural defects of the triple helix, altered secretion and abnormal fibril formation and/or assembly (Rauch and Glorieux, 2004; Marini et al., 2007).

Cartilage collagen fibrils are composed of type II collagen and other quantitatively minor collagens, including types IX and XI (Eyre, 2002). Type II collagen is a homotrimer of α1(II) chains encoded by a single gene, COL2A1. Defects in COL2A1 have been identified in a phenotypic continuum of chondrodysplasias that includes achondrogenesis type II, hypochondrogenesis, spondyloepiphyseal dysplasia congenita, Kniest dysplasia and Stickler syndrome (Spranger, 1988; Cole, 1994; Chan et al., 1995). In addition, mutations in COL9A1 or COL11A2 have also been linked to Stickler syndrome (Kuivaniemi et al., 1997).

During protein biosynthesis, a number of molecular chaperones are involved in ensuring the correct folding of protein in the cell (Hendrick and Hartl, 1993; Frydman et al., 1994; Hartl and Hayer-Hartl, 2002). In the case of procollagen biosynthesis in the ER, these molecular chaperones include BiP (Grp78), Grp94, PDI, P4H and Hsp47 (Chessler and Byers, 1993; Ferreira et al., 1994; Lamande et al., 1995; Wilson et al., 1998; Walmsley et al., 1999; Nagata, 2003). Hsp47 was first identified as a collagen binding protein that resides in the ER and functions as a collagen-specific molecular chaperone (Nagata et al., 1986). Hsp47 transiently associates with procollagen in the ER and dissociates before reaching the cis-Golgi (Satoh et al., 1996).

Recent studies have revealed that the SERPINH1 gene, which encodes human HSP47, is one of the genes responsible for recessive OI (Christiansen et al., 2010). Our previous studies showed that disruption of murine Hsp47 caused early embryonic lethality at embryonic day (E) 11.5 with impaired basement membrane formation (Nagai et al., 2000). Type I collagen secreted from Hsp47 null cells formed abnormal fibrillar structures with thin and branched fibrils (Ishida et al., 2006). Thus, Hsp47 was established as essential for development due to its role as a collagen-specific molecular chaperone in the ER, at least for type I and type IV collagen maturation.

Here we report that chondrocyte-specific disruption of Hsp47 caused severe chondrodysplasia and defective endochondral bone formation. Conditional null mutant mice, which died just before or after birth, exhibited a marked decrease in type II and type XI collagen accumulation in the cartilage. Thus, Hsp47 has an essential role as a molecular chaperone for type II collagen, a major component of cartilage.

Previous results revealed that gene disruption of murine Hsp47 caused embryonic lethality at E11.5 (Nagai et al., 2000). To examine the role of Hsp47 in the molecular maturation of collagens at later stages of mouse development, a Cre-loxP system was used to disrupt the Hsp47 gene in a tissue-specific manner (Gu et al., 1994). To accomplish this, a transgenic mouse harboring a floxed Hsp47 gene in which the exon-6-containing region was flanked by loxP sites was established (Fig. 1A). Exon 6 is the final exon and encodes approximately 24% of the amino acid sequence of Hsp47, including a region essential for binding to collagen (T. Homma et al., personal communication) and a poly(A) addition signal. Correct targeting by homologous recombination and germline transmission was confirmed by Southern blotting (Fig. 1B), PCR and genomic sequencing around the loxP site (data not shown). Heterozygous and homozygous mice were viable and fertile without noticeable phenotypic differences between littermates on a mixed genetic background (129SvEvTaq/C57BL6). Mice were used for experiments after backcrossing with C57BL/6 mice for at least two generations. Crossing the mice harboring the Hsp47 flox allele with mice harboring Cre recombinase in many cell types, including fibroblasts and osteoblasts (Col1a1–Cre), resulted in embryonic lethality of Hsp47^flox homozygous mice harboring Col1a1–Cre (data not shown), similar to Hsp47 null mice (Hsp47^–/–) (Nagai et al., 2000). These results indicated that Hsp47 floxed mice could be used for tissue-specific disruption of Hsp47.

To examine the role of Hsp47 during chondrogenesis, a conditional knockout of Hsp47 in chondrocytes was created by crossing Hsp47 floxed mice with Col2a1–Cre transgenic mice expressing Cre recombinase in chondrocytes (Ovchinnikov et al., 2000), resulting in heterozygous and homozygous Hsp47^flox mice carrying the Col2a1–Cre transgene. The heterozygous Hsp47^flox mice carrying Col2a1– Cre showed no phenotypic abnormalities, similar to Hsp47 heterozygous knockout (KO) mice (Nagai et al., 2000). However, Hsp47 conditional knockout mice (Hsp47 cKO: Hsp47^flox/flox; Col2a1–Cre) exhibited marked phenotypic changes. Approximately 8% of the Hsp47 cKO mice were embryonic lethal at approximately E18.5 because of severe amnionic bleeding (n=37). Conversely, 92% of Hsp47 cKO mice died either just before or within 2 hours after birth. This could have been because of their cleft palate and breathing difficulties, resulting from their abnormal rib cage, narrowed airway and abnormal bronchial cartilage (see supplementary material Fig. S1L, arrow). Hsp47 cKO neonates exhibited severely twisted and deformed limbs, some of which bled in the joint region (Fig. 2A, arrowheads). The body weights of cKO mice were ∼17% less than control (Hsp47^flox/flox) littermates.

To examine skeletal abnormalities in cKO mice, micro-computed tomography (μ-CT) examination was performed on neonatal mice. The maxilla, mandible, basilar bone, rib, humerus, femur and all the vertebral bodies were visible in newborn control mice by μ-CT scanning (Fig. 2B, left). By contrast, only the maxilla, mandible and part of the cervical spine were visible in Hsp47 cKO mice by μ-CT scanning (Fig. 2B, right), suggesting the impairment of normal bone formation.

Skeletal staining of E15.0 and E18.5 Hsp47 cKO mouse embryos revealed severe systemic chondrodysplasia and deformities of the long bones (Fig. 3B,D). The vertebral bodies of the thoracic, lumbar, sacral and caudal vertebrae of cKO mice were not ossified, but ectopic ossification were observed in cervical vertebrae (Fig. 3F, arrow; supplementary material Fig. S1). And the sacral vertebral arches of cKO mice failed to fuse (Fig. 3H, arrowhead), whereas the counterparts in control mice were connected by blue-stained cartilaginous bridges (Fig. 3G). The rib cartilage, humerus, radius, ulna, femur, tibia and fibula were considerably shorter and twisted (Fig. 3). The ossified region of the humerus of cKO skeletal preparations were shorter by ∼84% compared with E18.5 control embryos. In addition, cKO embryos had short snouts and partially uncalcified foreleg phalanges. However, the size of the cranial bone was similar to that of the control and appeared to be normally mineralized (Fig. 3K). Taken together, endochondral ossification in Hsp47 cKO mice was substantially impaired. Tissue sections of E14.5–18.5 and neonates revealed abnormal curvature of the spines in Hsp47 cKO mice; in particular, the cervical, lumbar and sacral spine were severely twisted, leading to kyphosis (Fig. 4H). In Hsp47 cKO mice, the notochords remained as rod-like axial structures at E14.5 (Fig. 4B), whereas they formed nucleus pulposus and disappeared in the vertebral bodies of sacral vertebrae in the control mice (Fig. 4A).

In control mice, the elbow joints contained distinct acellular joint cavities (Fig. 5A), whereas elbow joint cavities were not observed in Hsp47 cKO mice (Fig. 5B). Similar phenotypes were also observed in other joints of Hsp47 cKO mice. The long bones, such as the humerus, of Hsp47 cKO mice were also severely disorganized.

To examine the expression level of collagen and Hsp47 mRNAs, we performed in situ hybridization using antisense riboprobes of Col1a1, Col2a1 and Hsp47 on frozen sections of E14.5 lumbar vertebrae. The absence of Hsp47 mRNA from most chondrocytes of E14.5 Hsp47 cKO mice was clear, and a small number of Hsp47-positive cells was observed inside of perichondrium (Fig. 6A). Col2a1 mRNA was strongly expressed in the chondrocytes of control and Hsp47 cKO mice. Interestingly, Col1a1 mRNA was strongly upregulated only in the vertebral column of Hsp47 cKO mice. Additionally, the results of a TUNEL assay revealed that apoptosis was induced in the chondrocytes of Hsp47 cKO mice (Fig. 6B).

Immunohistological staining was performed on frozen sagittal sections of vertebrae from E18.5 embryos. Analysis of these sections revealed severe chondrodysplasia in Hsp47 cKO mice. Type II and type XI collagen were robustly stained in the annulus fibrosus, IVDs and vertebral bodies of control mice at E18.5, and thick collagenous fibrotic structures were well developed in these tissues (Fig. 7C,G). Strong Hsp47 staining was also observed in chondrocytes of the IVDs (Fig. 7A,E). By contrast, the accumulation of type II and type XI collagen in the cartilage of Hsp47 cKO mice was substantially reduced (Fig. 7D,H).

Extracellular accumulation of type II and type XI collagen was impaired in the chondrocytes of Hsp47 cKO mice, and previous results suggested that these collagens accumulated in the ER (Ishida et al., 2009). This was confirmed by measuring the induction of the molecular chaperone BiP, an indicator of ER stress (Haze et al., 1999; Bertolotti et al., 2000), in the IVDs of Hsp47 cKO mice (Fig. 4J).

When laser light is focused on highly organized birefringent materials, such as correctly aligned type I collagen fibers or myosin filaments in animals, laser light with double the frequency is emitted from the material, a phenomenon called optical second-harmonic generation (SHG) (Williams et al., 2005). However, an SHG signal is barely detectable when scattered from nonlinear media such as heat-denatured or degraded collagen fibers (Theodossiou et al., 2006). Recently, SHG imaging has been used to study collagen fibers in cartilage, hard tissue and some tumors (Brown et al., 2003; Yeh et al., 2005).

There were two possible effects of the disruption of Hsp47 gene expression in chondrocytes on collagen secretion and/or accumulation in the extracellular matrix. One was a decrease in the accumulation of extracellular collagen as a result of the retention of misfolded procollagen in the ER, and the other was the disordered alignment of fibrillar collagen in the extracellular matrix as a result of the structurally abnormal collagen, as occurs with type I collagen (Ishida et al., 2006). SHG imaging was used to determine whether the fibrillar alignment of collagen in the bone and fibrocartilage was disordered. Frozen serial sections from the IVDs and spongy bones of control neonates generated strong SHG signals (Fig. 7M). However, the SHG signal emitted from the intervertebral disc (IVD) of Hsp47 cKO mice was ∼99% lower (Fig. 7N, see the red boxed region and bar graph) than that of the control with equal amounts of type I collagen (Fig. 7K,L). Thus, SHG imaging revealed that collagen molecules in the IVDs of Hsp47 cKO mice were severely misaligned and failed to form robust connective tissue. This clearly shows for the first time that disruption of Hsp47 impairs fibrinogenesis of collagens in fibrocartilage tissues during development.

The effects of Hsp47 disruption on the secretion of cartilage-specific collagen was investigated in cultured primary chondrocytes. First, immunohistological analysis of cultured cells was performed to identify changes in Hsp47 protein levels. No Hsp47 was detected in chondrocytes from Hsp47 cKO embryos, whereas type II collagen was observed (Fig. 8C). Localization of type II collagen was analyzed by double staining with antibodies against marker proteins. PDI and Golgi matrix protein GM130 were used as ER and Golgi markers, respectively. Type II collagen localized in the ER because it was well merged with PDI in the control and Hsp47 cKO chondrocytes (Fig. 8A), but only a small amount of type II collagen accumulated in the Golgi complex of control and Hsp47-null chondrocytes (Fig. 8B). Western blot analysis revealed that the secretion and extracellular accumulation of type II collagen was markedly less in Hsp47 cKO chondrocytes (Fig. 8D) than in the control, and that there was less intracellularly.

Next, ultrastructural analysis was performed to more closely examine phenotypic changes in the connective tissues. Cartilage and hypertrophic cartilage in E18.5 Hsp47^flox heterozygous embryos carrying Col2a1–Cre did not show any abnormalities compared with control (Hsp47^flox/flox) embryos. Normally, cartilage tissues in the vertebrae is composed of proteoglycans and long collagen fibers, including type II collagen (arrows in Fig. 9C), whereas the ECM proteins in Hsp47 cKO mice were non-uniform and lacked long collagenous fibers (Fig. 9D). Instead, dot-like proteoglycan structures were detected in both control and cKO mice. These results suggested that secreted cartilage collagens did not accumulate like normal collagen fibers in the ECM of the vertebrae of cKO mice. Chondrocytes in Hsp47 cKO embryos had dilated rough endoplasmic reticulum (rER) filled with granular material (Fig. 9B). Moreover, cytoplasmic projections were poorly developed in chondrocytes in cKO embryos (Fig. 9B), suggesting defects in the secretion of substrates and/or interaction between cells and the ECM matrix. In addition, normal type I collagen fibers with a 67 nm period banding pattern were observed in the cranial bones of control, Hsp47^flox/wt and Hsp47 cKO mice (supplementary material Fig. S2), indicating that the maturation of type I collagen secreted from osteoblasts of cranial bones might be normal even in Hsp47 cKO mice.

The expression of collagen correlated with that of Hsp47 (Clarke et al., 1993). During development, fibroblasts synthesize large amounts of Hsp47; amounts are much lower in aged individuals. The expression pattern of Hsp47 during development mirrored that of various types of collagen. Interestingly, in various pathological conditions where collagen synthesis is abnormally stimulated, such as fibrosis, cicatrization and keloid formation, the expression of Hsp47 is also upregulated (Naitoh et al., 2001). In the case of liver fibrosis, the synthesis of both type I collagen and Hsp47 is increased, particularly in the activated myofibroblastic Ito cells (Masuda et al., 1994). In addition to these correlative expressions of collagen and Hsp47 in collagen-related diseases, the downregulation of Hsp47 by treatment with antisense RNA or stellate cell-specific Hsp47 siRNA dramatically decreased the progression of fibrosis and represents a promising therapeutic strategy for fibrotic diseases (Sato et al., 2008). Recent studies also revealed that a mutation in Hsp47 resulted in the development of OI in humans and dogs (Drogemuller et al., 2009; Christiansen et al., 2010).

Previously, we reported that the disruption of Hsp47 in mice caused severe defects in procollagen maturation in the ER, resulting in embryonic lethality at E11.5 (Nagai et al., 2000). In Hsp47-null mice, type I collagen fiber formation in the mesenchyme and the formation of basement membranes consisting of type IV collagen are impaired, which clearly establishes that Hsp47 is a molecular chaperone necessary for the proper molecular maturation of type I and type IV procollagen in the ER. However, no information was available as to whether Hsp47 was also required for the maturation of other types of collagen. To address this question, we developed a strategy for the tissue-specific disruption of Hsp47 using the Cre–loxP system. By crossing Hsp47 floxed mice with mice harboring Cre gene driven by the regulatory regions of collagen II alpha 1, we were able to disrupt Hsp47 expression in a cartilage-specific manner.

Generally, chondrocytes have two major functions: endochondral ossification for proper skeletal development, and articular cartilage maintenance for joint movement. Chondrogenic cell-line-specific Hsp47 disruption caused perinatal lethality in mice, with severely defective cartilage and endochondral bone formation. In these mice, type II collagen secretion and accumulation in the ECM was impaired, indicating that Hsp47 is essential for the molecular maturation of type II procollagen in the ER, and thus for proper cartilage formation and endochondral bone formation.

A recent study showed that type II collagen has essential roles in the removal of notochord (Aszodi et al., 1998). In wild-type mice, notochord was removed at approximately E14.5. However, in mice deficient in type II collagen the notochord of sacral vertebrae did not disappear until P0.

In the case of Hsp47 cKO mice, Hsp47 was absent at E14.5, and the notochord was retained until this stage. The acellular notochordal tube was observed even at E18.5, which was not the case in control mice (Fig. 4). These results indicates that defects of production of type II collagen in Hsp47 cKO mice results in a delay in the developmental step of dismantling the notochord.

Chondrodysplasias and related phenotypes that arise from the misfolding and loss of procollagens in humans and mice have been shown to be caused by mutations in various collagen genes, including COL2A1, COL9A1–3, COL11A1 and COL11A2 (Spranger et al., 1994; Jackson et al., 2010; Li et al., 1995; Vikkula et al., 1995; Mortier et al., 2000). Many mutations result from the substitution of essential residues in the Gly–x–y triplet repeats that form the mandatory repetitive structure of the triple helical domains of α-chains (Mortier et al., 2000). In addition to mutations in collagen genes, the mutation of other genes can also result in the development of cartilage-related diseases. For example, multiple epiphyseal dysplasia is due to mutations in matrilin-3 (Chapman et al., 2001) or in COMP (Briggs et al., 1995). Achondroplasia is caused by mutations in the gene for fibroblast growth factor receptor 3 (Shiang et al., 1994), and metaphyseal chondrodysplasia is caused by mutations in the gene for the parathyroid hormone-related peptide receptor (Schipani et al., 1995). These connective tissue disorders can be divided into three groups: (1) structural defects or the incomplete maturation of collagen molecules; (2) defects in the production of ECM components other than collagen; and (3) defects in the differentiation of chondrocytes. Mutations in collagen genes or genes necessary for collagen modification lead to disorders in the first group.

Chondrocyte-specific Hsp47 cKO mice and other models of mammalian chondrodysplasia resulting from mutation of collagen genes show marked similarities. The skeletal defects in Hsp47 cKO mice revealed by μ-CT examination or direct staining of skeletal preparations were similar to those observed in human patients with achondrogenesis type II that harbor mutations in COL2A1, and other mouse models for chondrodysplasia that harbor abnormalities in the type II collagen gene causing short limbs, a small thorax and abnormal growth plates (Metsaranta et al., 1992; Chan et al., 1995; Li et al., 1995). Ultrastructural observation revealed other common features, including distended rERs in the chondrocytes of Hsp47 cKO mice and human patients with hypochondrogenesis or spondyloepiphyseal dysplasia caused by mutation of COL2A1 (Freisinger et al., 1994). It is worth noting that the disruption of collagen-specific molecular chaperones and mutation of the substrate gene for type II collagen, result in similar phenotypes, consistent with previous results demonstrating that Hsp47 is necessary for the correct folding and/or secretion of procollagens. Some patients and mice with chondrodysplasias exhibit deformed cartilage and compensatory upregulation of type I and III collagen in chondrocytes (Freisinger et al., 1994; Chan et al., 1995; Aszodi et al., 1998), leading to the abnormal fibrillation of cartilage. In the case of chondrocytes in Hsp47 cKO mouse, Col1a1 mRNA was upregulated in E14.5 embryos (Fig. 6), whereas the immunofluorescence was lower. Additionally, the SHG signal was dramatically decreased in fibrocartilage of Hsp47 cKO mice (Fig. 7). Taken together the ablation of Hsp47 might cause the compensatory production of type I collagen from chondrocytes although they do not make a fibrous structures.

Recently, a missense mutation (c.233TC.Leu78Pro) in SERPINH1 was reported to cause an autosomal dominant disorder affecting the OI condition (Christiansen et al., 2010). HSP47 expression in the fibroblasts of this patient was low, perhaps because of increased instability of the protein. The OI-like phenotype might have been caused by defective maturation of type I procollagen and possibly other types of procollagen, including types II, IX, X and/or XI. Therefore, examination of other types of collagen, such as type XI collagen, a minor species of collagen in cartilage, should be performed in Hsp47 cKO mice. Unfortunately, we were unable to succeed in addressing this issue because of the technical difficulty in detecting qualitative changes in FACIT collagens. Also, an issue to be addressed in the future is whether Hsp47 ablation causes the chondrocyte differentiation indirectly for example by global defects in collagen folding.

Previously, we reported that Hsp47 could bind to collagenous triple helical domains with consensus amino acid sequences containing the Gly-x-Arg triplet as the minimal requirement (Koide et al., 2002). This suggests that Hsp47 binds to many types of collagen, and we subsequently reported that Hsp47 could bind to collagen types I–V in vitro (Natsume et al., 1994). We also showed that Hsp47 has an essential role as a collagen-specific molecular chaperone in the ER, necessary for the correct folding and secretion of at least type I and type IV procollagen (Nagai et al., 2000; Matsuoka et al., 2004). However, it was unclear whether Hsp47 can act as a molecular chaperone for other types of procollagen in vivo.

In the current study, we clearly demonstrated that Hsp47 has an indispensable role as a molecular chaperone in the secretion of type II collagen in chondrocytes. From these studies, the general role of Hsp47 as molecular chaperone could apply to a wider spectrum of collagen species including major constituents of the ECM, fibrillar collagen (type I, II, III and XI) and basement membrane collagen (type IV), all of which contain a common triple helical structure. However, we have no direct evidence of whether Hsp47 is necessary for the molecular maturation of network-forming collagen, FACIT collagen and/or other types of collagen which have transmembrane domains.

The recently developed SHG imaging method is applicable for the examination of pathological conditions in which collagen synthesis is abnormally stimulated, such as the fibrillation that occurs during osteoarthritis or the development of certain tumors (Brown et al., 2003; Yeh et al., 2005).

To examine the fibrinogenesis of collagen in tissues, including fibrocartilage, we used SHG imaging and found that SHG signals from the fibrocartilage of Hsp47 cKO mice were much lower than in controls. Although type I collagen was observed in the IVDs, the accumulation of type I collagen in the fibrocartilage of Hsp47 cKO mice did not have well-aligned fibrillar structures that would provide a strong SHG signal. That was also verified by ultrastructural observation of ECM structures by electron microscopy. Thus, we were able to clearly show that inactivation of the Hsp47 gene in fibrocartilage results in considerable disruption of collagen fiber formation in the ECM. We also successfully demonstrated that SHG imaging could be useful for the detailed analysis of fibrogenesis in collagen-related disorders.

An Hsp47 clone including all exons was isolated from a genomic BAC library derived from the 129/Sv strain. This clone was modified with two loxP sequences (5′-ATAACTTCGTATAATGTATGCTATACGAAGTTAT-3′), a neomycin (Neo) selection cassette and a diphtheria toxin A (DTA) selection cassette using pGKNEO-F2L2DTA as a targeting vector (Fig. 1A). The targeting vector was electroporated into the K.Y1.1 mouse embryonic stem cell line. After neomycin resistance (Neo) positive selection with G418, stem cell clones that had undergone homologous recombination were screened by PCR, Southern blotting analysis and partial genome sequencing, including the loxP site. Chimeric mice were generated from four homologous recombinant clones by aggregation, as described previously (Kondoh et al., 1999), and the chimeras obtained were bred with C57BL/6 mice to generate Hsp47 floxed heterozygous mice. Hsp47 floxed heterozygous mice were then bred to homozygosity. The Col2a1–Cre strain was acquired from the Jackson Laboratory (Bar Harbor, ME) (Ovchinnikov et al., 2000). In an initial cross, Col2a1–Cre transgenic mice were mated with mice heterozygous for the Hsp47 floxed allele. The offspring harboring the Col2a1–Cre transgene and an Hsp47 floxed allele were then mated with Hsp47 floxed heterozygous mice or homozygous mice to obtain embryos harboring the Col2a1–Cre transgene with two Hsp47 floxed alleles.

Routine mouse genotyping was performed by PCR. The following primer pairs were used for floxed Hsp47 alleles and the Cre transgene: (5′-GAGTGGGCTGAGCCCTCTCAAGAAAATCC-3′) and (5′-CTTCGGTCAGGCCCAGTCCTGCCAGATG-3′); and (5′-TCCAATTTACTGACCGTACACCAAAATTTG-3′) and (5′-CCTGATCCTGGCAATTTCGGCTATAC-3′), respectively.

Micro-computed tomography (μ-CT) analysis was performed using CT (SMX-100CT-SV3-type, Shimazu, Kyoto, Japan), as described previously (Ikeguchi et al., 2006). Whole-mount staining of skeletons with Alcian Blue and Alizarin Red used to visualize cartilage and ossified bone was performed essentially as described previously (McLeod, 1980). For histological analysis, neonates were fixed in 4% paraformaldehyde and embedded in paraffin. Sections (4 μm thick) were used for Hematoxylin and Eosin staining, Safranin-O–Fast Green staining and Alcian Blue–Nuclear Fast Red staining, Von Kossa reaction and Nuclear Fast Red staining were performed using established protocols (Matsumoto et al., 2009).

For immunofluorescence analysis and SHG imaging, embryos were first fixed in 4% paraformaldehyde, which was changed to 30% sucrose in PBS after 20% EDTA decalcification at 4°C. Immunohistological analysis was performed using 16 μm thick frozen sections and cultured primary chondrocytes as described below. The following antibodies were used: mouse monoclonal anti-Hsp47 antibody (Stressgen Biotechnologies; 1:600), rabbit polyclonal anti-type I collagen (LSL, Rochdale, UK; 1:200), mouse monoclonal anti-type II collagen (Lab Vision Corporation; 1:400), rabbit polyclonal anti-type II collagen (Rockland, Gilbertsville, PA; 1:300), rabbit polyclonal anti-type XI collagen (LSL; 1:200), rabbit polyclonal anti-Grp78 (Affinity Bioreagent; 1:400), mouse monoclonal anti-PDI (Assay Designs, Mona Vale NSW, Australia; 1:400) and mouse monoclonal anti-GM130 (BD Biosciences; 1:400). Alexa-Fluor-488-conjugated anti-mouse or rabbit IgG and Alexa-Fluor-594-conjugated anti-mouse or rabbit IgG (Invitrogen; 1:400) were used as secondary antibodies. We checked the recognition specificity of the anti-type I and anti-type II antibodies using with the limb sections of control mice (see supplementary material Fig. S3). In addition, cell nuclei were stained with Hoechst 33342.

In situ hybridization was performed using 7 μm thick frozen sections. Antisense and sense RNA probes for each gene to be examined were transcribed from their respective plasmids using a digoxigenin (DIG) RNA labeling kit (Roche Diagnostics, Penzberg, Germany). For mouse RNA probes, a 0.9-kb cDNA fragment of mouse Col1a1 were subcloned into pGEM4Z as previously reported (Masuda et al., 1994), a 1.6-kb cDNA fragment of mouse Col2a1 were subcloned into pBluesript2 KS(+) and a 0.5-kb cDNA fragment of mouse Hsp47 was subcloned into pBluescript2 SK(+); they were used for in situ hybridization following established protocols (Shukunami et al., 2008). TUNEL analysis was performed on frozen 7 μm sections using a fluorescein-conjugated in situ cell detection kit (Roche).

Frozen sections of E18.5 embryos were used for SHG imaging. Sample preparation was performed at 30°C to reduce the denaturation and degradation of collagen molecules as much as possible. Images were scanned using a custom-built multiphoton laser scanning microscope. Images of the SHG signal were obtained using a 120 femtosecond laser pulse centered at 840 nm excitation and 405–435 nm bandpass filters. Spectra were generated using a focal spectrum analyzer on a multiphoton laser scanning microscope. The short wavelength signal from the fibrillar structures in vertebrae was located at exactly half the excitation wavelength and had an extremely narrow bandwidth. The fact that it was located at half the excitation wavelength shifted was consistent with SHG as described previously (Brown et al., 2003).

For ultrastructural observation, sacral vertebrae and tibiae from embryos at E18.5 were fixed with 1% acrolein, 2% paraformaldehyde and 3% glutaraldehyde in 0.04 M cacodylate buffer (pH 7.4). After decalcification in 10% EDTA, the samples were post-fixed with 1% OsO₄ in 0.1 M cacodylate buffer and embedded in Epon 812 (TAAB, Aldermaston, UK). Ultrathin sections were examined by TEM (H-7600; Hitachi Co., Tokyo, Japan) at an accelerating voltage of 8 V.

Primary chondrocytes were prepared from the rib cartilage of E18 embryos using a modification of previously published protocols (Muramatsu et al., 2007). Chondrocytes were isolated using 0.15% collagenase (Sigma) for 1.5 hours at 37°C after adherent connective tissue had been removed with 0.1% trypsin (Nacalai, Kyoto, Japan) pretreatment for 40 minutes at 37°C. Isolated chondrocytes were maintained in DMEM–F12 (Gibco) supplemented with 10% FCS and ascorbic acid (50 μg/ml) and medium were changed every 24 hours. After 7 days in culture, immunofluorescence analysis was performed as described previously (Ishida et al., 2006) using the antibodies listed above.

Western blotting analysis was performed as described previously (Fernandes et al., 2007), using the following primary antibodies: mouse monoclonal anti-type II collagen alpha1 (Lab Vision Corporation; 1:1000) and mouse monoclonal anti-GAPDH (HyTest, Turku, Finland; 1:1000). Anti-mouse IgG alkaline phosphatase conjugate (Zymed; 1:1000), anti-rabbit IgG HRS conjugate (Biosource International, Camarillo, CA; 1:1000) and anti-goat IgG alkaline phosphatase conjugate (Biosource; 1:1000) were used as secondary antibodies.

Results are presented as means ± s.d. Statistical differences were determined using Student’s t-test. Significance was accepted at P-values less than 0.01.

We are very grateful to J. Takeda (Osaka University) for providing KY1.1 ES cells. We also thank T. Matsushita (Kyoto University), D. Okui (Kyoto University), J. Nakamura (Kyoto Sangyo University), K. Yasuda (Nagasaki University), C. Syukunami (Kyoto University), A. Kitamura (Hokkaido University) and Y. Ishida (Shionogi & Co., Ltd.) for professional help and discussion.