Collagen X Chains Harboring Schmid Metaphyseal Chondrodysplasia NC1 Domain Mutations Are Selectively Retained and Degraded in Stably Transfected Cells*

Collagen X is a short chain, homotrimeric collagen expressed specifically by hypertrophic chondrocytes during endochondral bone formation and growth. Although the exact role of collagen X remains unresolved, mutations in theCOL10A1 gene disrupt growth plate function and result in Schmid metaphyseal chondrodysplasia (SMCD). With the exception of two mutations that impair signal peptide cleavage during α1(X) chain biosynthesis, SMCD mutations are clustered within the carboxyl-terminal NC1 domain. The formation of stable NC1 domain trimers is a critical stage in collagen X assembly, suggesting that mutations within this domain may result in subunit mis-folding or reduce trimer stability. When expressed in transiently transfected cells, α1(X) chains containing SMCD mutations were unstable and presumed to be degraded intracellularly. More recently, in vitrostudies have shown that certain missense mutations may exert a dominant negative effect on α1(X) chain assembly by formation of mutant homotrimers and normal-mutant heterotrimers. In contrast, analysis of cartilage tissue from two SMCD patients revealed that the truncated mutant message was fully degraded, resulting in 50% reduction of functional collagen X within the growth plate. Therefore, in the absence of data that conclusively demonstrates the full cellular response to mutant collagen X chains, the molecular mechanisms underlying SMCD remain controversial. To address this, we closely examined the effect of two NC1 domain mutations, one frameshift mutation (1963del10) and one missense mutation (Y598D), using both semi-permeabilized cell and stable cell transfection expression systems. Although able to assemble to a limited extent in both systems, we show that, in intact cells, collagen X chains harboring both SMCD mutations did not evade quality control mechanisms within the secretory pathway and were degraded intracellularly. Furthermore, co-expression of wild-type and mutant chains in stable transfected cells demonstrated that, although wild-type chains were secreted, mutant chains were largely excluded from hetero-trimer formation. Our data indicate, therefore, that the predominant effect of the NC1 mutations Y598D and 1963del10 is a reduction in the amount of functional collagen X within the growth cartilage extracellular matrix.

Collagen X is a short-chain collagen expressed specifically at sites of endochondral ossification during normal skeletal development and under conditions that involve new bone growth, such as fracture repair or osteoarthritis (1)(2)(3). Although the precise role of collagen X remains ill-defined, it is proposed to provide structural support within the extracellular matrix of the epiphyseal growth plate during the transition from cartilage to bone (4). Collagen X is a homotrimer of three ␣1(X) chains (M r 59,000), each composed of a collagenous domain (COL1) 1 flanked at the carboxyl and amino termini by noncollagenous extensions, the NC1 and NC2 domains. Unlike the fibrillar collagens, the NC1 and NC2 domains of collagen X are not removed by proteolytic processing and are thought to direct supramolecular assembly of collagen X molecules into an hexagonal lattice that forms mat-like structures within the matrix (5). Collagen X has also been observed as fine pericellular filaments associated with collagen II fibrils within the hypertrophic cartilage matrix (6) leading to the hypothesis that functional roles for collagen X may differ according to the type of aggregate formed (7).
Given the restricted localization of collagen X to growth plate cartilage, it is not surprising that COL10A1 mutations result in the cartilage defect, Schmid metaphyseal chondrodysplasia (SMCD). However, the precise consequences of mutations on collagen X biosynthesis, secretion, and supramolecular assembly and their impact on SMCD molecular pathology have not been fully elucidated and remain controversial. SMCD is an autosomal dominant disorder of the osseous skeleton resulting in a relatively mild phenotype associated with growth plate abnormalities (8 -10). SMCD mutations characterized to date include single base substitutions, nonsense mutations, and deletions resulting in premature termination (for review, see Ref. 10). Apart from two missense mutations, which affect the putative signal peptide cleavage site (11), the remaining SMCD mutations identified are localized within the NC1 domain. The association of carboxyl-terminal domains is an important step in assembly of fibrillar procollagens and a potential site for NC1⅐NC1 interaction was identified based upon alignment of collagen X and fibrillar procollagen carboxyl-terminal amino acid sequences (12). This conserved aromatic motif (amino acids 589 -601 2 ) is indeed critical for the interaction of NC1 domains in vitro (13).
Based on the observation that SMCD mutations prevented the formation of SDS-stable trimers, it was initially proposed that SMCD is caused by exclusion of mutant chains from ␣1(X) assembly, resulting in a 50% reduction in functional collagen X within the growth plate (14). In two cases, analysis of SMCD patient cartilage has shown that the presence of a premature termination codon results in nonsense-mediated decay of the mRNA derived from the mutant allele, supporting the haploinsufficiency model (15). 3 Missense mutations, however, are unlikely to reduce mRNA stability, and studies have shown that ␣1(X) chains harboring certain point mutations can form stable homotrimers and co-assemble with the normal chain when expressed in semi-permeabilized cells (16). An in vitro competition-based assay that is not dependent on the detection of SDS-stable trimers also showed that SMCD mutant chains could reduce the efficiency of normal ␣1(X) chain assembly (13). Thus the model emerging from these studies is that SMCD missense mutations exert a dominant negative effect on collagen X assembly. However, in vitro expression systems may not accurately reproduce the full cellular response to mutant collagen X mRNA or mutant ␣1(X) chains. Indeed, expression of collagen X in transiently transfected mammalian cells demonstrated that wild-type ␣1(X) chains were secreted efficiently whereas SMCD mutant ␣1(X) chains were present in only trace amounts within the media (17). It is not clear whether lack of mutant chain secretion was due to poor expression levels of the mutant ␣1(X) constructs or rapid intracellular degradation of ␣1(X) chains.
In this study we have used two expression systems to evaluate the effect of the SMCD NC1 point mutation, Y598D (9), and a frameshift mutation (18) resulting from a 10-bp deletion at position 1963, 4 1963del10 (NC1⌬10) on the assembly of human collagen X. Firstly, we carried out in vitro analysis of ␣1(X) chain assembly using a semi-permeabilized cell translation system (19). Previous results have shown that bovine Y598D ␣1(X) chains, expressed in semi-permeabilized HT 1080 cells, assemble with similar efficiency to the normal chain when analyzed under low SDS concentrations (16). However, in vitro translated human Y598D trimers have markedly reduced stability in comparison to the wild-type NC1 trimers, when analyzed under similar conditions (20). Here we find that human ␣1(X) chains carrying the Y598D and NC1⌬10 mutations assemble less efficiently than the wild-type chain when expressed in semi-permeabilized cells. A similar pattern emerged when Y598D and NC1⌬10 ␣1(X) constructs were stably transfected into human bone (SaOS-2) cells. Although expressed at the mRNA level, mutant chains were retained within the cell and degraded whereas normal chains were secreted. To assess whether mutant chains could be incorporated into stable heterotrimers, we generated co-transfected cell lines expressing wild-type ␣1(X) chains together with either Y598D or NC1⌬10 ␣1(X) chains. Our results show that only trace amounts of mutant collagen X chains could be detected in the media, indicating that the predominant effect of these mutations is a 50% reduction in functional collagen X secretion.

Construction of Wild-type and Mutant Collagen X Recombinant Plasmids for in Vitro Expression-
The production of full-length human ␣1(X) cDNA (pTM1-h10wt) and engineering of the SMCD mutants used in this study have been described previously. The wild-type ␣1(X) sequence was cloned into the T7-driven expression vector, pTM1, to enable transcription of mRNA in vitro (14). Two SMCD mutations, 1963del10 (18) and Y598D (9), were generated by strand overlap exten-sion PCR using specific oligonucleotide primer sets as described previously (17). The 10-bp deletion at position 1963 (NC1⌬10) introduces nonsense sequence from amino acid residue 623 resulting in premature termination at residue 671. The missense mutation Y598D generates a single amino acid (tyrosine to aspartate) substitution at residue 598 in the NC1 domain.
In Vitro Transcription-Generation of full-length mRNA transcripts encoding wild-type and SMCD mutant ␣1(X) chains was carried out as previously described (21) using 10 g of plasmid cDNA linearized with SalI. Transcription reactions were performed for 4 h at 30°C using 60 units of T7 RNA polymerase (Promega). RNA was purified using an RNA extraction kit (Qiagen) and stored at Ϫ70°C in 10-l aliquots containing 1 unit/l RNasin (Promega) and 1 mM dithiothreitol.
In Vitro Translation in Semi-permeabilized HT 1080 Cells-HT 1080 fibroblasts (ATCC-CCL121, American Type Culture Collection) were maintained in Dulbecco's modified Eagle's medium (DMEM) containing 10% (v/v) fetal calf serum. HT 1080 cells were treated with digitonin (Calbiochem) at a final concentration of 40 g/ml to generate semipermeabilized cells, which were used to supplement rabbit reticulocyte lysate-based in vitro translation as previously described (22). Each translation contained 17.5 l of Flexi-lysate (Promega), 1 l of transcribed mRNA, 0.5 l of 1 mM amino acids (minus methionine), 0.5 l of 2.5 M KCl, 1 l of L-[ 35 S]methionine (1.5 mCi/ml, ICN Pharmaceuticals Inc.), 0.25 l of 25 mM sodium ascorbate (Sigma), and 4 l of semipermeabilized cells. After translation for 1 h at 30°C, semi-permeabilized cells were isolated by centrifugation and prepared for electrophoresis or treatment with chymotrypsin and trypsin.
Construction of Wild-type and SMCD ␣1(X) Expression Constructs for Stable Transfection-For the stable expression of His 6 -tagged collagen X in SaOS-2 cells we generated three expression constructs (pCEP4wt-his, pCEP4-Y598D-his, and pCEP4-NC1⌬10-his) encoding ␣1(X) sequences flanked at the amino termini by the BM-40 signal peptide, a 6-histidine tag, and enterokinase cleavage site. Wild-type, Y598D, and NC1⌬10 collagen X sequences in the plasmid pGEM11 were amplified using primers HX-SP (5Ј-GCGGCCGCGTGTTTTACGCTGAACGATA-3Ј) and HX6 (5Ј-CGCCGGCGCTTTTCAGCCTACCTCCATA-3Ј) corresponding to nucleotides 56 -75 and 2235-2216. PCR amplification was performed at 60°C for 30 cycles, using 0.5 unit of Pfu polymerase (Stratagene), to generate a 2-kb fragment corresponding to the entire collagen X coding sequence lacking only the signal peptide. Additional sequences encoding NotI restriction enzyme sites were added to primers (sequence underlined) to enable subsequent cloning into the plasmid pCEP4-BM40-hisEK (23), kindly donated by Dr. Anders Aspberg, Lund University, Sweden. This plasmid is derived from pCEP4 (Invitrogen), which encodes EBNA1, a gene that allows the plasmid to be maintained episomally, and the hygromycin resistance gene, which allows selection of positive transfectants.
For the stable expression of c-Myc-tagged collagen X in SaOS-2 cells we used an expression construct (pRc/CMV-wt-myc) encoding wild-type collagen X sequence flanked at the amino terminus by the BM-40 signal sequence and the c-Myc epitope. To introduce the c-Myc tag (amino acid sequence EQKLISEEDL), we performed strand overlap extension PCR using the plasmid pGEM11 containing the wild-type collagen X coding sequence lacking signal peptide as template and specific primer sets incorporating the c-myc coding sequence (underlined). The primary rounds of PCR were carried out using primer sets MYC-F1 (5Ј-AAGC-TTCTGCCTGCCGCCTG-3Ј) and MYC-R1 (5Ј-CAGATCCTCTTCAGA-GATGAGTTTCTGCTCGCGGCCGCTAGCTAGCGGGG-3Ј) or MYC-F2 (5Ј-GAGCAGAAACTCATCTCTGAAGAGGATCTGGTGTTTTACGCT-GAAGGATA-3Ј) and MYC-R2 (5Ј-CACCCAGGTCCTTCTGG-3Ј) to generate independent fragments with overlapping sequences. Second round PCR reactions were carried out with primers MYC-F1 and MYC-R2 using 5 ng of the primary PCR products as template. All PCR amplifications were performed at 62°C for 35 cycles, using 0.5 unit of DeepVent polymerase (New England BioLabs). The recombinant PCR product was digested with NheI and XhoI. The resulting 220-bp fragment was purified using a Geneclean kit (Qiagen) and used to replace the corresponding sequence in a pGEM7 construct encoding the fulllength wild-type collagen X sequence. Finally, this plasmid was digested with NheI and XbaI to release the entire 2.2-kb collagen X sequence, which was introduced into pRc/CMV (Invitrogen), an expression vector that encodes the neomycin phosphotransferase gene conferring resistance to the antibiotic G418. Constructs were sequenced (AmpliCycle, PerkinElmer Life Sciences) to verify insertion of the c-myc sequence and to ensure insertion of the recombinant fragment in the correct reading frame.
Stable Transfection of SaOS-2 Cells-The human osteosarcoma cell line, SaOS-2 (ATCC HTB-85, American Type Culture Collection), was maintained in DMEM (Invitrogen) containing 10% (v/v) fetal calf serum. SaOS-2 cells were grown to ϳ70% confluence and transfected with the ␣1(X) expression constructs using FuGENE 6 transfection reagent (Roche Molecular Biochemicals) according to the manufacturer's protocol. Cells transfected with pCEP4-wt-his, pCEP4-Y598D-his, and pCEP4-NC1⌬10-his were selected and maintained in growth medium containing 250 g/ml hygromycin B (Roche Molecular Biochemicals). Cells transfected with pRc/CMV-wt-myc were selected in growth medium containing 500 g/ml G418 (Invitrogen) and individual G418resistant colonies were isolated and expanded into cell lines. One cell line expressing wt-myc collagen X mRNA, as shown by Northern analysis, was used for protein expression analysis and subsequent co-expression studies. Co-transfection of wt-myc cells with pCEP4-wt-his, pCEP4-Y598D-his, and pCEP4-NC1⌬10-his was performed using the transfection protocol described above. Co-transfected cells were maintained in growth medium containing 500 g/ml G418 and 250 g/ml hygromycin B.
Northern Blot and Hybridization Analysis-RNA was prepared from transfected and untransfected cells cultured in individual 35-mm diameter dishes at ϳ80% confluence, using the RNeasy extraction kit (Qiagen). Heat-denatured RNA (3 g/sample) was electrophoresed on a 1% (w/v) agarose gel containing 7% (v/v) formaldehyde, followed by capillary blotting onto a nitrocellulose filter for 16 h. The filter was air-dried, baked under vacuum at 80°C for 1 h, and pre-hybridized for 2 h in 25% (v/v) deionized formamide, sheared herring sperm DNA (100 g/ml final concentration), Denhardt's solution, 4 mM EDTA, and 0.1% (w/v) SDS. A 32 P-labeled probe was prepared using a 2-kb NotI-digest product of pTM1-h10wt (14), purified using the Geneclean kit (Qiagen). Hybridization of the probe was carried out at 42°C overnight, and the washed filter was exposed to autoradiography film at Ϫ70°C for 2 weeks.
Biosynthetic Labeling and Immunoprecipitation-Biosynthetic labeling was performed by incubation of transfected cells grown to confluence in 6-well plates with 1 ml of methionine-free, serum-free DMEM (Invitrogen) containing 0.25 mM sodium ascorbate and 100 Ci of L-[ 35 S]methionine (Tran 35 S-label, 1032 Ci/mmol, ICN Pharmaceuticals Inc.). When appropriate, sodium ascorbate was replaced with 0.3 mM ␣␣Ј dipyridyl (Calbiochem). In pulse-labeling experiments, cells were pre-treated for 1 h with 1 ml of methionine-free, serum-free DMEM, prior to addition of 100 Ci of L-[ 35 S]methionine. After 1 h in labeling media, cells were chased for a further hour in serum-free DMEM containing excess unlabeled methionine. Where indicated, media was supplemented with 5 M clasto-lactacystin ␤-lactone (Calbiochem) or 1 g/ml brefeldin A (Roche Molecular Biochemicals). After removal of the media, cells were lysed on ice using 1 ml of ice-cold immunoprecipitation buffer (50 mM Tris/HCl, pH 7.5, 5 mM EDTA, 150 mM NaCl, and 1% (v/v) Triton X-100) containing 1 mM 4-(2-aminoethyl)-benzenesulfonyl fluoride hydrochloride (AEBSF) (Roche Molecular Biochemicals) and 2 mM N-ethylmaleimide (Sigma Chemical Co.). Cell lysates and medium samples were clarified by centrifugation (13,000 ϫ g for 20 min) at 4°C, incubated for 1 h at 4°C with 100 l of protein A-Sepharose (20% (w/v) in PBS) and centrifuged (13,000 ϫ g for 2 min) to remove protein A-Sepharose-binding components. Immunoprecipitation of ␣1(X) chains was performed overnight at 4°C in the presence of 10 l of the appropriate antiserum and 100 l of protein A-Sepharose (20% (w/v) in PBS). The rabbit polyclonal serum, raised to the NC1 domain of human collagen X, was a gift from Dr. Olena Jacenko (15). Mouse monoclonal antibodies against the c-Myc epitope (clone 9E10), and His 6 epitope tags were from Roche Molecular Biochemicals and used at a final concentration of 1 g/ml. Immunoprecipitates were washed three times with immunoprecipitation buffer and either prepared for SDS-PAGE analysis or treatment with chymotrypsin and trypsin.
Sequential Immunoprecipitations-Cell lysates and media fractions were prepared for immunoprecipitation as described above. Primary immunoprecipitations were carried out at 4°C for 5 h using anti-c-Myc or anti-His 6 antibodies to isolate epitope-tagged ␣1(X) chains. Where indicated, samples were analyzed directly without performing a secondary immunoprecipitation. Otherwise, samples were resuspended in 50 l of immunoprecipitation buffer containing 2% (w/v) SDS and heated to 65°C for 10 min to elute ␣1(X) chains. Samples were centrifuged briefly (13,000 ϫ g for 30 s), and the supernatants were carefully removed into fresh tubes and diluted to 1 ml in immunoprecipitation buffer. Secondary immunoprecipitation was performed overnight at 4°C using 100 l of 20% (w/v) protein A-Sepharose in the presence or absence of anti-Myc or anti-His 6 secondary antibodies, used at a final concentration of 2 g/ml.
Chymotrypsin/Trypsin Digestion-Protease treatment of semi-permeabilized cell translation products was carried out in 50 l of CT/T digestion buffer (50 mM Tris-HCl, pH 7.4, containing 150 mM NaCl and 10 mM EDTA) supplemented with 0.5% (v/v) Triton X-100 to solubilize cells. Immunoprecipitates were resuspended in 50 l of digestion buffer without Triton X-100. Samples were incubated with a combination of trypsin (100 g/ml) and chymotrypsin (250 g/ml) for 2 min at 25°C. Digests were stopped by addition of soybean trypsin inhibitor (Sigma) at a final concentration of 5 g/ml and two volumes of boiling SDS-PAGE sample buffer.
SDS-PAGE and Immunoblotting-Samples were prepared for electrophoresis by boiling for 5 min in an equal volume of SDS-PAGE sample buffer (50 mM Tris/HCl, pH 6.8, 2% (w/v) SDS, 20% glycerol, and 0.025% (w/v) bromphenol blue) unless otherwise stated. All samples were resolved through 7.5% SDS-PAGE gels, and labeled proteins were visualized by autoradiography, fluorography, or phosphorimaging (Molecular Dynamics, Storm) where indicated (see "Results"). Samples for immunoblotting were transferred to nitrocellulose and incubated for 1 h in PBS containing 0.1% Tween 20 in the presence of 5% (w/v) milk protein. The filter was probed with anti-KDEL mouse monoclonal antibodies (Stressgen Biotechnologies Corp.) at a final concentration of 1 g/ml, and primary antibodies were detected using rabbit anti-mouse IgG-horseradish peroxidase conjugate. Blots were developed by chemiluminescence (ECL, Amersham Biosciences, Inc.) according to the manufacturer's protocol.

Human Collagen X Chains Harboring SMCD Mutations Y598D and NC1⌬10 Assemble Less Efficiently Than Wild-type
Chains When Expressed in Semi-permeabilized Cells-In this study we used an in vitro translation system supplemented with semi-permeabilized human fibroblasts (HT 1080 cells) as a starting point to analyze the effect of SMCD mutations on assembly of the full-length human ␣1(X) chain. The ␣1(X) cDNAs encoding one SMCD point mutation, Y598D, and one frameshift mutation, NC1⌬10, were transcribed in vitro and translated in the presence of rabbit reticulocyte lysate supplemented with semi-permeabilized HT 1080 cells. Translation products were prepared for electrophoresis without heating and using 0.5% (w/v) SDS to detect trimers, which may dissociate under more harsh conditions (Fig. 1). A high proportion of wild-type ␣1(X) chains formed SDS-stable trimers, whereas both NC1 Y598D and NC1⌬10 mutant chains migrated essentially as monomers (lanes 1-3). This result indicates that the  1 and 4) or SMCD mutant collagen X (lanes 2, 3, 5, and 6) was translated in rabbit reticulocyte lysate supplemented with semi-permeabilized HT 1080 cells for 60 min at 30°C. Following translation, semi-permeabilized cells were separated from the mixture by centrifugation, and translated products were either incubated in 0.5% (w/v) SDS for 5 min at 25°C (lanes 1-3) or treated with chymotrypsin and trypsin (CT/T; lanes 4 -6) followed by separation through a 7.5% SDS-PAGE gel. Proteins were visualized by fluorography, and quantitation was performed by phosphorimaging analysis. The migration positions of helical domains (␣1(X) h ), monomeric (␣1(X)), and trimeric (␣1(X) 3 ) chains are indicated. ability of the NC1 domains to form SDS-stable trimers is severely compromised by both SMCD mutations. Despite this apparent lack of electrophoretic stability, weak or transient association of mutant ␣1(X) chains within the endoplasmic reticulum may result in triple-helix nucleation and subsequent propagation. Therefore, normal and mutant ␣1(X) chains were synthesized in semi-permeabilized cells in the presence of ascorbate, and translation products were treated with chymotrypsin and trypsin to probe formation of correctly folded triple helices. As expected, a high proportion of wild-type ␣1(X) chains assembled into correctly aligned triple helices as evidenced by a protease-resistant band corresponding to the collagenous domain (lane 4). In contrast, a much smaller fraction of the Y598D and NC1⌬10 chains synthesized was protease resistant (lanes 5 and 6). Together, these results demonstrate, using established assays, that the SMCD-causing mutations Y598D and NC1⌬10 significantly reduce the efficiency of human collagen X assembly in semi-permeabilized cells.
Mutant Collagen X Chains in Stably Transfected SaOS-2 Cells Are Expressed at the mRNA Level but the Protein Is Not Secreted-The chaperones and enzymes required for correct post-translational processing of collagens are present within the endoplasmic reticulum of semi-permeabilized HT1080 cells, allowing the initial stages of collagen folding and assembly to be reconstituted in vitro (16,24,25). However, the full cellular response to mutant ␣1(X) chains can only be assessed within the context of intracellular transport and secretion quality control. Data concerning the secretion of mutant collagen X homo-or heterotrimers is lacking due to the low expression level of SMCD mutants in transiently transfected cells (17). To overcome this, we generated stably transfected SaOS-2 cells expressing the wild-type human ␣1(X) chain and the NC1 SMCD mutants Y598D and NC1⌬10. As part of our strategy for analysis of normal and mutant ␣1(X) heterotrimer formation, we cloned the wild-type ␣1(X) cDNA and mutant ␣1(X) cDNAs into different vectors containing different antibiotic resistance markers, allowing us to select and culture cells co-expressing both wild-type and mutant chains. Furthermore, we used the c-Myc epitope (9E10) to tag normal ␣1(X) and a His 6 motif to tag NC1 Y598D and NC1⌬10 chains, thus allowing discrimination of wild-type and mutant chains.
To assess whether wild-type and mutant ␣1(X) chains are secreted, transfected cells and control (untransfected) SaOS-2 cells were biosynthetically labeled with [ 35 S]methionine. Wildtype (wt-Myc) and mutant (Y598D-His and NC1⌬10-His) ␣1(X) chains were recovered from cell and media fractions by immunoprecipitation using anti-c-Myc and anti-His 6 antibodies (Fig.  4). The cell and media fractions of untransfected cells were incubated with antibodies raised to the human ␣1(X) NC1 domain. Collagen X chains were detected within the cell fractions of the three transfected cell lines but absent from untransfected SaOS-2 cells, consistent with the mRNA expression pattern (Fig. 3). As samples were denatured prior to electrophoresis, collagen X trimers were not observed. Strikingly, although normal ␣1(X) chains were secreted (lane 6), mutant chains were not detected in the media fraction (lanes 7 and 8). Similar results were obtained when anti-collagen X antibodies were used instead of anti-c-Myc or anti-His 6 antibodies, indicating that detection of normal and mutant ␣1(X) chains was not biased by the use of two different epitope tags (data not shown). Considering the high level of Y598D and NC1⌬10 mRNA expression relative to the wild-type collagen X mRNA (Fig. 3) and the lack of accumulated intracellular material, these data strongly suggest that mutant ␣1(X) chains are degraded via an intracellular pathway.
Mutant ␣1(X) Chains Are Degraded via Multiple Intracellular Pathways-To investigate the mechanism by which mutant ␣1(X) chains are degraded, cells expressing either wild-type or mutant collagen X chains were biosynthetically labeled in the presence of the proteasomal inhibitor clasto-lactacystin ␤-lactone or brefeldin A, a reagent that prevents intracellular traffic to the lysosomal/endosomal compartments. Cells were pre-FIG. 2. Schematic representation of the epitope-tagged normal and SMCD mutant ␣1(X) chain cDNAs used to transfect SaOS-2 bone cells. The c-myc-tagged wild-type ␣1(X) cDNA and his 6 -tagged mutant ␣1(X) cDNAs were cloned into pRc/CMV-BM40 and pCEP4-BM40-hisEK vectors, respectively (see "Experimental Procedures"). The location of the mutations within the NC1 domain (filled gray boxes) of the expected protein products are shown. Y598D is a single tyrosine to aspartate substitution at amino acid position 598; NC1⌬10 is a 10-bp deletion introducing a nonsense amino acid sequence from residue 623 with premature termination at residue 671. The positions of the c-Myc (diagonal hatched box) and His 6 (cross-hatched box) tags at the amino termini are indicated.

FIG. 3. Northern analysis of collagen X mRNA expression in transfected and untransfected SaOS-2 cells. Total RNA was isolated from untransfected SaOS-2 cells (lane 4) or SaOS-2 cells trans-
fected with either c-myc-tagged wild-type collagen X cDNA (lane 1) or his 6 -tagged mutant collagen X cDNA (lanes 2 and 3). RNA was separated on a denaturing 1% (w/v) agarose gel containing 7% (v/v) formaldehyde and stained with ethidium bromide. The migration positions of 18 and 28 S rRNAs are indicated (a). Subsequently, RNA was transferred to nitrocellulose and hybridized to a 32 P-labeled collagen X-specific probe (b, see "Experimental Procedures" for details). treated for 1 h with or without inhibitors, pulse-labeled for 1 h then incubated in chase medium (see "Experimental Procedures") for a further hour (Fig. 5). In the absence of both inhibitors, wild-type (lane 1) but not mutant (lanes 5 and 9) ␣1(X) chains were detected in the media fraction. As expected, inclusion of brefeldin A abolished secretion of wild-type ␣1(X) chains (lanes 3 and 4), resulting in accumulation of material in the intracellular fraction. Significantly higher levels of intracellular Y598D and NC1⌬10 chains were detected when cells were treated with either clasto-lactacystin ␤-lactone (lanes 6 and 10, respectively) or brefeldin A (lanes 7 and 11, respectively). Furthermore, incubation with a combination of both inhibitors resulted in enhanced protection of mutant ␣1(X) chains in the cellular fraction (lanes 8 and 12). These data suggest that mutant chains are degraded via both proteasomal and vesicular transport-dependent pathways.
Mutant ␣1(X) Chains That Are Not Degraded Can Form Correctly Folded Helical Domains-Our results demonstrate that Y598D and NC1⌬10 ␣1(X) chains are not secreted but are degraded via lactacystin-and brefeldin A-sensitive pathways. However, at the end of overnight labeling (Fig. 4) and pulse labeling (Fig. 5) a minor fraction of the retained mutant chains remain undegraded. To analyze the folding status of these chains, transfected SaOS-2 cells were metabolically labeled overnight either in the presence of 0.25 mM ascorbate to permit hydroxylation, or in the presence of 0.3 mM ␣␣Ј dipyridyl, which effectively inhibits hydroxylation and triple helix formation (26). Cell lysates were incubated with antibodies raised to collagen X and immunoprecipitates were analyzed either di-rectly or after digestion with chymotrypsin and trypsin to probe the formation of correctly aligned triple helices. Collagen X chains synthesized in the absence of ␣␣Ј dipyridyl (Fig. 6a,  lanes 1, 3, and 5) had reduced electrophoretic mobility compared with ␣1(X) chains synthesized in the presence of ␣␣Ј dipyridyl (lanes 2, 4, and 6), evidence that ␣␣Ј dipyridyl prevented ␣1(X) chain hydroxylation. Accordingly, chains synthesized in the absence of ␣␣Ј dipyridyl had formed proteaseresistant helical domains (Fig. 6b, lanes 1, 3, and 5) whereas unhydroxylated chains were degraded by protease treatment (lanes 2, 4, and 6). As shown with the semi-permeabilized cell system, Y598D and NC1⌬10 chains are capable of triple helix formation (Fig. 1). However, in intact cells it is clear that only wild-type chains are secretion-competent (Fig. 6c, lane 1), indicating that mutant chains are recognized as non-native within the endoplasmic reticulum.
Retained ␣1(X) Chains Are Associated with PDI/P4-H-Two endoplasmic reticulum resident proteins, Hsp47 and protein disulfide isomerase (PDI), have been identified as candidates to "chaperone" collagen X during biosynthesis (16,27). PDI has been proposed to prevent assembly of trimers into higherordered structures within the endoplasmic reticulum and also binds transiently to collagen X as the ␤ subunit of P4-H during the hydroxylation of ␣1(X) chains (16,27). Our results show clearly that mutant ␣1(X) chains are selectively retained within the endoplasmic reticulum of stable transfectants. Therefore, we attempted to identify interactions between collagen X and any endoplasmic reticulum resident proteins using co-immunoprecipitation. Control (untransfected) and SaOS-2 cells transfected with wild-type, Y598D, and NC1⌬10 constructs were incubated overnight in the presence of 0.25 mM ascorbate to allow ␣1(X) chain hydroxylation and folding. Cell lysates were incubated under native conditions, without addition of chemical cross-linkers, with anti-collagen X antibodies to isolate ␣1(X) chains and any stably associated proteins. Immunoprecipitates were denatured, resolved by SDS-PAGE, and proteins were transferred to nitrocellulose. The filter was probed with an anti-KDEL monoclonal antibody that recognizes proteins with the carboxyl-terminal KDEL endoplasmic reticulum retention motif. A protein band with the molecular weight of PDI was co-precipitated specifically with Y598D and NC1⌬10 ␣1(X) chains (Fig. 7, lanes 3 and 4) but not with wild-type ␣1(X) chains (lane 2). Previous studies using bovine ␣1(X) chains synthesized in semi-permeabilized cells showed FIG. 4. Analysis of collagen X secretion in SaOS-2 cells transfected with epitope-tagged wild-type and mutant collagen X cDNAs. Biosynthetic labeling of untransfected and transfected SaOS-2 cells was carried out as described (see "Experimental Procedures"). Wild-type and mutant collagen X chains were immunoprecipitated using anti-c-Myc antibodies (m) and anti-His 6 antibodies (h), respectively. As a control, [ 35 S]methionine-labeled proteins in the cell and media fractions of untransfected SaOS-2 cells were incubated with polyclonal antibodies raised to collagen X (x). Immunoprecipitates were denatured by boiling in SDS-PAGE sample buffer containing 2% (w/v) SDS prior to electrophoresis on a 7.5% gel. An unidentified protein band with the approximate molecular mass of trimeric collagen X was recovered from the media of Y598D-his and NC1⌬10-his cells (lanes 7 and 8, indicated by asterisk). However, this protein was also present in the cell and media fractions of untransfected SaOS-2 cells (lanes 1 and 5), which do not synthesize collagen X, and is therefore an endogenous product of this cell line. Wild-type and mutant ␣1(X) chains were isolated by immunoprecipitation using anti-c-Myc or anti-His 6 antibodies as appropriate. Immune complexes were denatured by boiling in SDS-PAGE sample buffer containing 2% (w/v) SDS, prior to separation on 7.5% SDS-polyacrylamide gels, and radiolabeled proteins were visualized by fluorography.
an equal level of binding of PDI to wild-type and mutant ␣1(X) chains (16). However, we have shown here that the formation of stable trimers and correctly aligned triple helices by human ␣1(X) chains is compromised by the Y598D and NC1⌬10 mutations. Our results, therefore, are consistent with a chaperone function for PDI, acting independently or as a subunit of P4-H.
When Co-expressed with Wild-type ␣1(X) Chains Only Trace Levels of Y598D Chains Are Secreted-When expressed alone, the SMCD mutants Y598D and NC1⌬10 are unable to assemble into secretion-competent homotrimers. However, as SMCD is a heterozygous disorder, it was important to assess whether wild-type and SMCD mutant chains could form stable heterotrimers that undergo secretion. To analyze normal-mutant ␣1 (X) chain heterotrimer formation we generated co-transfected cell lines expressing c-Myc-tagged wild-type ␣1(X) chains together with either His 6 -tagged Y598D or His 6 -tagged NC1⌬10 chains (wt-myc/Y598D-his cells and wt-myc/NC1⌬10-his cells, respectively; see "Experimental Procedures" for details). As a positive control for co-assembly of chains carrying the two epitope tags, we generated a third co-transfected cell line expressing both wt-Myc and wt-His tagged chains (wt-myc/wt-his cells). Co-transfected SaOS cells (wt-myc/wt-his) were biosynthetically labeled overnight, and cell and media fractions were incubated with either anti-c-Myc or anti-His 6 antibodies (Fig.  8). After washing, immune complexes were either analyzed directly (lanes 1, 2, 5, and 6) or denatured and subjected to a second round of immunoprecipitation using the reciprocal antibodies (lanes 3, 4, 7, and 8). As expected, wt chains carrying both His 6 and c-Myc epitopes were secreted into the medium (lanes 5 and 6). In cell and media fractions, sequential immunoprecipitation demonstrated co-assembly of wt-His and wt-Myc chains. Our results show that, after denaturation of immune complexes, the anti-c-Myc antibody was more efficient in the secondary immunoprecipitation (lanes 4 and 8) than the anti-His 6 antibody (lanes 3 and 7). Therefore, in subsequent co-assembly experiments, samples were incubated first with anti-His 6 antibodies to isolate mutant chains and secondly with anti-c-Myc antibodies. SaOS-2 cells transfected with wild-type (WT-myc) and mutant (Y598Dhis and NC1⌬10-his) collagen X and control (untransfected) cells were incubated overnight in DMEM containing 0.25 mM sodium ascorbate to allow ␣1(X) chain hydroxylation and triple-helix formation. Cell lysates were incubated with polyclonal antibodies raised to collagen X and immune complexes were denatured in SDS-PAGE sample buffer containing 25 mM dithiothreitol prior to separation on a 7.5% SDS-polyacrylamide gel. Proteins were transferred to a nitrocellulose filter, probed with an antibody that recognizes the endoplasmic reticulum chaperones Grp94, BiP, PDI, and Hsp47, and visualized by chemiluminescence (see "Experimental Procedures"). The migration positions of protein molecular mass standards are indicated. Immunoglobulin chains used in the immunoprecipitation that are detected nonspecifically by the horseradish peroxidase-secondary antibody conjugate are detected as a protein smear at ϳ50 kDa (indicated by asterisk). A protein band with an approximate molecular mass of 60 kDa is indicated (PDI). 6 -tagged wild-type ␣1(X) chains in co-transfected SaOS-2 cells. Co-transfected cells expressing wt-Myc and wt-His ␣1(X) chains (wt-myc/wt-his cells) were biosynthetically labeled overnight with L-[ 35 S]methionine as described (see "Experimental Procedures"). After labeling, media samples and cell lysates were incubated for 5 h with primary antibodies specific for the c-Myc (m) or His 6 (h) epitope tags. Immunoprecipitates were either analyzed directly without secondary immunoprecipitation (lanes 1, 2, 5, and 6) or after denaturation and subsequent incubation overnight with secondary antibodies as indicated (lanes 3, 4, 7, and 8). All immune complexes were denatured in SDS-PAGE sample buffer containing 2% (w/v) SDS, prior to electrophoresis on a 7.5% SDS-polyacrylamide gel, and radiolabeled proteins were visualized by fluorography.

FIG. 8. Co-expression and sequential immunoprecipitation of c-myc-and his
To assess normal-mutant ␣1(X) co-assembly, co-transfected SaOS cells (wt-myc/Y598D-his and wt-myc/NC1⌬10-his) were biosynthetically labeled overnight. Cell and media fractions were incubated with either anti-c-Myc or anti-His 6 antibodies and either analyzed directly or after secondary immunoprecipitation. Wild-type and mutant chains were detected in the intracellular fractions of both co-transfected cell lines (Fig. 9, a  and b; lanes 2 and 3). To analyze heterotrimer formation, immune complexes containing Y598D-His or NC1⌬10-His ␣1(X) chains were denatured and re-precipitated with anti-c-Myc antibodies. In contrast to the wt-myc/wt-his cells (Fig. 8), after primary immunoprecipitation with anti-His 6 antibodies, little or no wt-Myc chains could be recovered in the secondary immunoprecipitation (Fig. 9, a and b; lane 4). Analysis of the media showed that, in comparison to the level of wild-type chains, very little Y598D-His material was secreted (Fig. 9a,  lane 8) and NC1⌬10-His chains were barely detectable in the media fraction (Fig. 9b, lane 8). In wt-myc/Y598D-his cells, a fraction of the ␣1(X) chains recovered using anti-His 6 antibodies could be re-precipitated with anti-c-Myc antibodies (Fig. 9a,  lane 9) indicating that the trace amounts of Y598D-His chains present in the media are secreted as heterotrimers. In both experiments we did not observe any significant reduction in the level of wt-Myc chains secreted into the media by co-transfected cells, compared with the original cell line expressing only wt-Myc ␣1(X) chains (compare lanes 6 and 7, panels a and b). Together, these results demonstrate that mutant ␣1(X) chains are largely excluded from heterotrimer formation. However, a very minor fraction of Y598D chains are co-assembled with wild-type chains within the cell, resulting in secretion of trace levels of normal-mutant heterotrimers DISCUSSION In this study we have employed two expression systems to analyze the effect of SMCD mutations on the assembly and secretion of human collagen X. The ability of NC1 domains carrying point mutations to form stable homotrimers and interact with wild-type ␣1(X) chains has been demonstrated previously in several studies (13,16). Using the semi-permeabilized cell system, we expressed normal and SMCD ␣1(X) chains to evaluate the effect of one point mutation (Y598D) and one frameshift mutation (NC1⌬10) on homotrimer and triple helix formation. Interestingly, the NC1⌬10 chains assembled poorly compared with the Y598D chains, consistent with the proposal that frameshifts may cause a more severe effect on the protein⅐protein interactions involved in NC1 trimer assembly (17). Comparison of the data from our study and that of McLaughlin et al. (16), shows that both wild-type and Y598D human ␣1(X) chains formed a lower proportion of helical material than the respective bovine homologues. One explanation for this difference is the presence of two cysteine residues within the helical domain of the bovine ␣1(X) chain, which would contribute to the stability of the triple helix during folding. Another property of the human and bovine ␣1(X) chains synthesized in this system that differs significantly is the relative stability of the homotrimers in low concentrations of SDS. Although bovine Y598D and wild-type chains formed SDS-stable homotrimers with similar efficiency, we found that the stability of human ␣1(X) trimers was dramatically reduced by the Y598D mutation. Previous results have shown that the human NC1 domain carrying the Y598D mutation can form SDS-stable trimers (20), although these trimers were more  2, 3, 7, and 8) or after denaturation and subsequent incubation overnight with secondary antibodies as indicated (lanes 4, 5, 9, and 10). Duplicate samples were incubated under the same conditions in the absence of secondary antibody (-Ab). In addition to co-transfected cells, SaOS-2 cells expressing only wt-Myc ␣1(X) chains were biosynthetically labeled (WT-myc). Media samples and cell lysates were incubated for 5 h with antibodies specific for c-Myc (m) and immunoprecipitates were analyzed (lanes 1 and 6). All immune complexes were denatured in SDS-PAGE sample buffer containing 2% (w/v) SDS, prior to electrophoresis on a 7.5% SDS-polyacrylamide gel, and radiolabeled proteins were visualized by fluorography. sensitive to denaturation by 0.5% (w/v) SDS than the wild-type NC1 trimer. This indicates that the conformation of the NC1 domain of human collagen X is altered by the Y598D substitution, such that the resulting trimers are destabilized. In support of this, the Y598D-engineered human NC1 domain expressed in Escherichia coli showed a dramatic reduction in stability and solubility when compared with the wild-type NC1 domain (28).
Despite the apparent wealth of relevant literature, an understanding of the full effect of SMCD mutations on the assembly and secretion of ␣1(X) chains is lacking. To overcome the limitations of transient transfection, where expression levels are difficult to ascertain, or synthesis of mutant ␣1(X) chains in vitro, in the absence of the complete secretory pathway quality control, we generated stably transfected cells expressing wildtype and SMCD mutant ␣1(X) chains. Although previous transfection studies have confirmed degradation of the two specific SMCD mutations that prevent cleavage of the signal peptide (17), our studies are the first to demonstrate the effect of two NC1 domain mutations, Y598D and NC1⌬10, on the secretion of human ␣1(X) chains in a cellular context. Here we have shown that, although normal chains are secreted, mutant chains are selectively retained and are degraded via both proteasomal and lysosomal/endosomal pathways, despite evidently high expression levels at the mRNA level.
Although both SMCD mutations severely compromized ␣1(X) chain assembly and folding, we observed small amounts of triple-helical Y598D and NC1⌬10 chains in both semi-permeabilized cells and transfected cells. However, only proteaseresistant material formed by wild-type ␣1(X) chains could be detected in the media fraction. Studies on the secretion of procollagen from fibroblasts indicate that the rate-limiting step in secretion is triple-helix formation (29). Our data are therefore the first to demonstrate that mutant collagen chains with correctly aligned helices can be selectively retained. One interpretation of this result is that correct folding and assembly of the other domains, most likely the NC1 domain, is critical for secretion and that mutant chains are recognized by a highly selective quality control mechanism in intact cells (30,31).
The specific interaction between mutant ␣1(X) chains and PDI identifies one endoplasmic reticulum protein that could function as part of such a quality control mechanism. Based on cross-linking studies using bovine ␣1(X) chains, it has been shown that PDI does not differentiate between the conformation of normal and mutant trimers (16). As discussed above, it appears that SMCD mutations affect the assembly of human NC1 domains more severely than bovine NC1 domains. It has been demonstrated that mis-folding of the human Y598D NC1 domain expressed in E. coli results in specific association with the bacterial chaperone GroEL (28). We did not detect any interaction with BiP, the mammalian homologue of GroEL, which is consistent with published results (16). Our results suggest that the SMCD mutant ␣1(X) chains adopt a conformation recognized by PDI, resulting in retention within the endoplasmic reticulum. PDI is also intimately linked with collagen assembly as the ␤ subunit of P4-H. This enzyme functions as a catalyst of proline hydroxylation and as a chaperone, which mediates retention of incompletely folded procollagen chains within the endoplasmic reticulum (24,26,32). Our data show that triple-helix formation by NC1 mutants is inefficient compared with the wild-type chains. Therefore, further investigation will determine whether PDI acts independently or as a subunit of P4-H during the retention of mutant ␣1(X) chains.
The main objective of this study was to assess the extent to which co-assembly and secretion of normal-mutant ␣1(X) heterotrimers contributes to the SMCD disease mechanism. Crit-ical analysis of the data published by this group and others argues that the SMCD phenotype cannot be explained comprehensively by a single disease mechanism. In two individuals, the introduction of premature termination codons within the mutant mRNA results in functional haploinsufficiency due to nonsense-mediated decay of the mutant allele (15). 3 In another report, mRNA encoding the missense mutant Y598D was detected within the growth plate cartilage of one SMCD patient carrying this allele (33). It is not clear whether the NC1⌬10 mutant allele is stable in vivo. However, the resulting mRNA is truncated by only 27 bp and therefore may escape mutant mRNA surveillance and degradation. Although it was initially reported that several SMCD mutations abolished NC1 trimer formation, it has been demonstrated that, in in vitro systems, wild-type and SMCD mutant ␣1(X) chains are capable of heterotrimer formation, supporting a dominant negative disease mechanism. Such a mechanism where mutant chains are "fully" included in ␣1(X) trimer formation has two potential outcomes. One is that heterotrimers incorporating one or more mutant chains are unstable and are targeted for intracellular degradation, reducing the secretion of functional collagen X chains by nearly 90%. The other possibility is that SMCD mutations do not destabilize assembled trimers, resulting in secretion of mutant homotrimers and heterotrimers that interfere with collagen X assembly and interactions within the cartilage extracellular matrix. In the present study we demonstrated that, in intact cells, mutant chains are largely excluded from co-assembly with normal ␣1(X) chains and that only a very minor fraction of Y598D chains had formed stable heterotrimers in the endoplasmic reticulum. Furthermore, Y598D and NC1⌬10 chains were secreted in only trace amounts, indicating that ␣1(X) chains harboring these mutations are retained and degraded within the cell. It is uncertain the extent to which such a minor population of normal-mutant heterotrimers would contribute to matrix dysfunction within the developing cartilage of SMCD patients. We also show that, compared with cells expressing wild-type ␣1(X) chains alone, secretion of normal chains from co-transfected cells is not dramatically reduced.
In conclusion, our data strongly indicate that, in stably transfected cells, mutant ␣1(X) chains harboring the Y598D and NC1⌬10 SMCD mutations are excluded from trimer formation. Therefore, we propose that, in contrast to the dominant negative model described above, their expression principally results in functional haploinsufficiency. Importantly, further studies are required to examine the effects of other SMCD mutations within stably transfected cells. Mapping of SMCD mutations onto the NC1 trimer modeled on the crystal structure of ACRP30 (20) raises the possibility that other mutations can be tolerated within stable homo-and heterotrimers. Studies on how these mutations affect supramolecular assembly and interactions with other cartilage matrix components will provide further insight into the molecular pathology of SMCD and consequently the role of collagen X in growth plate cartilage function.