Site-directed mutagenesis of human type X collagen. Expression of alpha1(X) NC1, NC2, and helical mutations in vitro and in transfected cells.

Type X collagen is a short chain collagen expressed in the hypertrophic zone of calcifying cartilage during skeletal development and bone growth. The α1(X) homotrimer consists of three protein domains, a short triple helix (COL1) flanked by nonhelical amino-terminal (NC2) and carboxyl-terminal (NC1) domains. While mutations of the NC1 domain result in Schmid metaphyseal chondrodysplasia, which suggests a critical role for this protein domain, little biochemical detail is known about type X collagen synthesis, secretion, and the mechanisms of molecular assembly. To study these processes, a range of mutations were produced in human α1(X) cDNA and the biochemical consequences determined by in vitro expression, using T7-driven coupled transcription and translation, and by transient transfection of cells. Three NC1 mutants, which were designed to be analogous to Schmid mutations (1952delC, 1963del10, and Y598D), were unable to assemble into type X collagen homotrimers in vitro, but the mutant chains did not associate with, or interfere with, the efficiency of normal chain assembly in co-translations with a normal construct. Expression in transiently transfected cells confirmed that mutant type X collagen assembly was also compromised in vivo. The mutant chains were not secreted from the cells but did not accumulate intracellularly, suggesting that the unassociated mutant chains were rapidly degraded. In-frame deletions within the helix (amino acid residues 72-354) and the NC2 domain (amino acid residues 21-54) were also produced. In contrast to the NC1 mutations, these mutations did not prevent assembly. Mutant homotrimers and mutant-normal heterotrimers were formed in vitro, and the mutant homotrimers formed in transiently transfected cells had assembled into pepsin-stable triple helical molecules which were secreted.

consequence of joint degeneration in osteoarthritis (6). The type X collagen molecule is a homotrimer of three ␣1(X) chains encoded by a condensed gene (COL10A1) of three exons, one of which (exon 3) codes for the majority of the polypeptide chain including the entire triple helical domain (7,8). The ␣1(X) homotrimer consists of three distinct protein domains. The short triple helical domain, COL1 (amino acids 57-519), containing eight imperfections in the Gly-X-Y triplet repeat sequence is flanked by a small nonhelical globular NC2 domain at the amino terminus (amino acids 19 -56) and a larger more conserved nonhelical carboxyl-terminal NC1 domain (amino acids 520 -680) (7). However, the specific restricted distribution and transient synthesis of type X collagen has meant that, relative to the more abundant and well studied collagen types, little is known about its synthesis, secretion, and mechanisms of molecular assembly. The chains are synthesized with an amino-terminal signal peptide which is proteolytically removed from the pre␣1(X) chains during synthesis (7,9). This appears to be the only post-translational proteolytic processing event, since both the NC2 and NC1 domains, which are thought to be important for intermolecular interactions and the formation of extracellular supramolecular assemblies, are retained on extracellular type X collagen (10).
Sequence comparisons of the carboxyl-terminal NC1 domain of type X collagen with the C-propeptides of fibrillar collagens demonstrated a conserved cluster of aromatic residues within a 130-amino acid domain with a marked hydrophobicity profile similarity (11). The role of the carboxyl-terminal propeptide domain of interstitial collagens in initiating intracellular ␣-chain selection, assembly, and helix formation is well established, and thus by analogy the ␣1(X) NC1 domain is likely to play a similar crucial role in type X assembly. Further evidence for the importance of the NC1 domain comes from studies on patients with Schmid metaphyseal chondrodysplasia (SMCD, 1 MIM 156500), a mild autosomal disorder of the osseous skeleton resulting from cartilage growth plate abnormalities, where a range of type X collagen mutations have now been identified. All of the mutations so far detected are localized in the NC1 domain and include amino acid substitution mutations C591R (12), G595E, Y598H (13), Y598D, L614P (14), N617K (13), G618V (9), L644R and D648G (13); frameshift mutations 1952delC (12), 1952delCC (15), 1952del13 (16), 1963del10 (17), 2004delT (13), and 2088delCT (12); premature termination mutations Y628X and W651X (15).
For one of these NC1 mutations (G618V), in vitro expression was used to demonstrate that the mutation prevented in vitro type X collagen assembly (9), providing direct evidence for the role of the NC1 domain in assembly. To further study the biochemical role of the NC1 domain, and to address the question whether the common molecular defect in SMCD is compromised type X collagen molecular assembly due to NC1 mutations, we have produced three NC1 mutations, duplicating the spectrum of mutations defined in patients, 1952delC (NC1⌬C), 1963del10 (NC1⌬10), and Y598D, and expressed these in vitro and by transient transfection of cells.
The biochemical consequences of mutations in the type X collagen helix and in the NC2 are also not known since naturally occurring mutations in these protein domains have not yet been defined. Transgenic mice have been generated expressing a chick type X collagen with a helical deletion (18). These express a more severe spondylometaphyseal dysplasia phenotype which was attributed to the dominant negative effect of the mutant chick ␣1(X) associating with endogenous mouse ␣1(X) chains although the association of mutant:normal heterotrimers necessary for this dominant effect was not directly demonstrated. In this study an in-frame helix deletion (helix⌬), removing 283 helical amino acid residues, and a deletion mutation, removing the entire NC2 domain (NC2⌬), were constructed and expressed in vitro and in vivo. These site-directed mutagenesis studies were designed to determine the contribution of the NC1, NC2, and helical domains on type X collagen homotrimer and on mutant:normal heterotrimer molecular assembly and secretion.

EXPERIMENTAL PROCEDURES
Construction of a Full-length Human Type X Collagen cDNA Containing the 3Ј-Untranslated Region-The production of a full-length ␣1(X) cDNA construct, pTM1-h10, in a T7-driven expression vector suitable both for in vitro transcription-translation and transient expression in cells using the vaccinia/T7 phage system (19) has been described previously (9). The 3Ј-untranslated region and a polyadenylation site was amplified using a reverse transcription-PCR kit (Perkin-Elmer) from oligo(dT)-primed cDNA produced from hypertrophic chondrocyte mRNA (9). A 1546-bp fragment was amplified using primers (BX1) 5Ј-CAGGGGGTAACAGGAATGCC-3Ј (1726 -1745) 2 and (HX14) 5Ј-TT-GTGTCGACTGAAAAGCCTTGAAA-3Ј (3256 -3242). An additional sequence containing a SalI restriction enzyme site (sequence underlined) was added to primer HX14 for subsequent cloning purposes. The amplified fragment was digested with SalI, purified by agarose electrophoresis, recovered by electroelution, and cloned into the SmaI and SalI site of pUC19. A 1046-bp region of the 3Ј-untranslated region (2211-3256) 2 was released from a positive clone by digestion with unique restriction enzymes NsiI and SalI and cloned into pTM1-h10 (9) to produce a full-length human type X cDNA (pTM1-h10wt) containing the 3Ј-untranslated region including the first polyadenylation site.
Construction of an In-frame Helix Deletion of Type X Collagen-An in-frame helix deletion, 310del849 (helix⌬) was constructed within the COL1 domain in plasmid pTM1-h10wt. Unique restriction sites XhoI and NcoI were utilized to remove a 853-bp fragment. The digested plasmid was purified by agarose electrophoresis, recovered by electroelution, the resultant overhang sequences were filled in with Klenow, and blunt ends were ligated to produce the plasmid pTM1-helix⌬. This procedure removed the amino acid residues 72-354, 3 a total of 283 amino acids within the amino-terminal region of the triple helical domain of human type X collagen (Fig. 1). Within the normal helical domain, there are eight interruptions of the Gly-X-Y repeats, and this helical deletion removed seven, but introduced one new Gly-X-Y interruption resulting from the blunt ligation of codons for Arg 71 to His 355 .
Site-directed Mutagenesis of the NC1 Domain-Three human type X collagen NC1 Schmid metaphyseal chondrodysplasia mutations ( Fig. 1) were produced using splicing by overlap extension PCR (20). The mutant primer sets and their relative positions are shown in Table I. Primer set NC1⌬C-A and NC1⌬C-B introduce a single base deletion of cytosine 1952, 1952delC (12); primer set NC1⌬10-A and NC1⌬10-B introduce a 10-base pair deletion, 1963del10 (17); while primer set Y598D-A and Y598D-B introduce a single base substitution of T 1988 for G and changed the amino acid tyrosine 598 to aspartate, Y598D (14). The PCRs were carried out in 20 l of 10 mM Tris/HCl, pH 8.0, containing 3 mM MgCl 2 , 50 mM KCl, 0.2 mM dNTPs, and 0.75 M of each of the primers. The polymerase chain reactions were carried out in capillaries using the FTS-1 thermal sequencer from Corbett Research (Melbourne, Australia) over 26 cycles. Cycle one was performed at 96°C for 2 min, 62°C for 1 min, and 72°C for 1 min, then followed by 25 cycles at 96°C for 10 s, 62°C for 15 s, and 72°C for 20 s. The reaction was terminated at 72°C for 1 min. Two independent PCR products were first produced using BX1 (sense) with primer B (antisense) and primer A (sense) with HX6 (antisense) of each of the mutant primer sets in the primary round of PCR using pTM1-h10wt (5 ng) as a template. The amplified fragments were purified by agarose electrophoresis and recovered using Geneclean (BIO 101 Inc., Vista, CA) and subjected to a second round of overlapping PCR with primers BX1 and HX6. The recombinant mutant PCR fragments were purified, cloned, and sequenced to ensure that the correct mutations were introduced and that there were no PCR errors. A PflMI and NsiI fragment containing the mutations were first cloned into a genomic type X clone pMC2N (9), and then introduced into pTM1-h10wt via NcoI and NsiI sites.
Construction of an In-frame Deletion of the NC2 Domain-Splicing by overlap extension PCR was also used to produce a NC2 157del102 mutation (NC2⌬). This deletion removes most of the NC2 domain except the first two amino acids (residues 19 and 20) and the last two amino acids (residues 55 and 56) of the NC2 domain (Fig. 1). The signal peptide (residues 1-18) remains unaltered. The plasmid pTM1-h10wt (5 ng) was used as a template for the primary rounds of PCR with primer pairs pTM1-1 (sense)/NC2⌬B (antisense) and NC2⌬A (sense)/ HX10 (antisense) ( Table I) to generate the two independent fragments with overlapping sequences. The second round PCR reaction was carried out with primers pTM1-1 and HX10 and 5 ng of each of the purified fragments from the primary PCR reactions. The recombinant fragment was purified and cloned into Bluescript SK II(ϩ) and sequenced to confirm the deletion. A 362-bp fragment containing the NC2 deletion was released from the Bluescript SK II(ϩ) cloned by digestion with KpnI and XhoI and ligated to a 957-bp XhoI-KpnI fragment (305-1262) 2 from pTM1-h10wt. After digestion with KpnI, the expected 1319-bp fragment was purified and reintroduced into pTM1-h10wt via KpnI sites.
In Vitro Transcription and Translation-Plasmids for cell-free translation were purified using ethidium bromide/CsCl 2 centrifugation. The ethidium bromide was removed by several extractions with isopropanol equilibrated with NaCl-saturated 10 mM Tris-HCl, pH 8.0, 1.0 mM EDTA and dialyzed against milli-Q water. The plasmids were translated using the TNT T7 polymerase coupled transcription and translation system (Promega) as described previously in a reaction volume of 12.5 l (9). Routinely, 0.25 g of plasmid was used per reaction, and 2.5 l of the final reaction are required for gel electrophoresis. In some experiments, canine microsomal membranes (Promega) were also FIG. 1. Schematic representation of the protein products of normal (wt) and mutant human type X collagen cDNA constructs. Site-directed mutations were created in the NC2, COL1, and NC1 domains of type X collagen (see "Experimental Procedures" for details). The location of the mutations and the expected protein products are shown relative to the normal (wt) pre␣1(X) chain. NC1⌬C is a single base deletion of cytosine 1952 introducing a frameshift at amino acid residue 620 and a premature termination at 621; NC1⌬10 is a 10-base pair deletion, introducing a nonsense amino acid sequence (filled box) from residue 623 and a premature termination at 671; Y598D and G618V are amino acid substitutions of tyrosine 598 to aspartate and glycine 618 to valine, respectively; helix⌬ is an in-frame helix deletion of 283 amino acids of the COL1 domain, and NC2⌬ removes the NC2 domain. The signal peptide (open box) is unaffected in all mutations generated.
added to promote collagen assembly (9,21). For heterotrimer assembly experiments, 0.25 g of each of the normal and mutant plasmids were co-translated. Type X collagen chains were analyzed on 7.5% (w/v) SDS-polyacrylamide gels as described previously (9). The samples were routinely denatured at 60°C for 10 min prior to electrophoresis. In some instances, samples were heated at 100°C for 5 min to denature assemblies prior to electrophoresis with or without 10 mM dithiothreitol. Analysis of the NC1 and NC2 domains by bacterial collagenase digestion was performed as described previously (9).
Transient Expression of Normal and Mutant Type X Collagen in Mammalian Cells-Type X collagen cDNA constructs were transiently expressed in a rat osteogenic sarcoma cell line, UMR 106-01 (American Type Culture Collection, ATCC-CRL 1661), a mouse fibroblast line, 3T3-A31 (ATCC-CCL163), and a monkey COS cell line, CV-1 (ATCC-CCL70) using the vaccinia driven T7 bacteriophage expression system developed by Moss and colleagues (22). This system employs initial infection of the cells with a recombinant vaccinia virus (vTF7-3) in which the T7 RNA polymerase gene is under control of the early/late vaccinia P7.5 promoter so that it is continuously expressed resulting in high cytoplasmic levels of T7 polymerase. Subsequent transfection of the cells with cDNA constructs under the control of the T7 promoter in the vector pTM1 allows cytoplasmic transcription of the transfected gene.
In brief, 5 ϫ 10 5 UMR 106-01, 1.8 ϫ 10 5 CV1 or 3T3-A31 cells were cultured for 24 h in DMEM containing 10% (v/v) fetal calf serum and 0.25 mM sodium ascorbate. Cells were infected with a T7 polymeraseexpressing vaccinia virus vTF7-3 (30 plaque-forming units/cell) for 30 min in 0.5 ml serum-free DMEM, followed by transfection with 2 g of plasmid which had been complexed with 15 l of Lipofectamine (Life Technologies, Inc.) for 30 min at room temperature. Transfections were carried out in 1 ml of serum-free DMEM for 6 h at 37°C in a 5% CO 2 incubator. The transfection medium was removed and replaced with 1.0 ml of serum-free DMEM, containing 0.25 mM sodium ascorbate and 10 Ci of L-[2,3-3 H]proline and incubated for 8 or 18 h at 37°C. At the appropriate time point, the medium fractions were removed, and the cell layers were scraped into 1 ml of 50 mM Tris/HCl, pH 7.5, containing 0.15 mM NaCl, 5 mM EDTA, 10 mM N-ethylmaleimide, and 0.1 mM phenylmethylsulfonyl fluoride. The cells were lysed by repeated freezing and thawing. The labeled proteins were recovered from both the medium and cell fractions by 75% (v/v) ethanol precipitation and analyzed by 7.5% (w/v) SDS-polyacrylamide gel electrophoresis. In some experiments, type X collagen was detected by Western blotting and probed with a type X collagen specific antibody (23), generously provided by Dr. Gary Gibson (Henry Ford Hospital, Detroit, MI).

In Vitro Transcription and Cell-free Translation of Normal and Mutant Plasmids-In vitro transcription and cell-free
translation demonstrated that all the mutant plasmids were translated into pre␣1(X) chains of molecular weight consistent with the introduced mutations (Fig. 2, lanes 1-6). While no molecular weight change was observed for the Y598D amino acid substitution (Fig. 2, lane 4), both the single base deletion of cytosine 1952 (NC1⌬C) and the 10-base pair deletion (NC1⌬10) of the NC1 domain resulted in shortened pre␣1(X) chains (Fig. 2, lanes 2 and 3) when compared with the normal (wt) pre␣1(X) chain of 680 amino acids (Fig. 2, lane 1). Both these mutations produced codon frameshifts causing premature chain terminations at amino acid 621 for NC1⌬C and 673 for NC1⌬10. Similarly, the electrophoretic migrations of proteins translated from the in-frame deletion mutations, helix⌬ and NC2⌬, were consistent with the expected sizes of 397 and 646 amino acids, respectively (Fig. 2, lanes 5 and 6).
Analysis of the Nonhelical NC1 and NC2 Domains by Bacterial Collagenase Digestion-The site-directed mutations in the NC1 and NC2 domains were confirmed by the analysis of these domains following bacterial collagenase digestion of pre␣1(X) chains synthesized in the absence of microsomes. The deletion of the NC2 domain in the NC2⌬ mutant was clearly demonstrated by its absence in the collagenase digest (Fig. 3, lane 6). The NC1 domains from the NC1⌬C and NC1⌬10 mutants were also markedly reduced in size as expected (Fig. 3, lanes 2 and  3), whereas the electrophoretic migrations of the NC1 and NC2 domains from the Y598D and helix⌬ mutants were unaffected (Fig. 3, lanes 1 and 5).   While the pre␣1(X) chains produced do not spontaneously assemble into multimers during in vitro translation, assembly could be induced by incubation in the Ca 2ϩ -containing buffer used for collagenase digestion (9). A significant proportion of the wt ␣1(X), helix⌬ ␣1(X), and NC2⌬ ␣1(X) chains could be induced to assemble into multimers, as evidenced by the presence of multimers of the collagenase-released NC1 domains (Fig. 3, lanes 4 -6), however, the mutant NC1 domains were present only as monomers (lanes 1-3). This is consistent with our previous report which demonstrated that a NC1 G618V mutation prevented in vitro multimer assembly (9).
In Vitro Chain Assembly Induced by Translation with Microsomes-The ability of the mutant type X collagen chains to associate into trimers in vitro was studied further by translation in the presence of canine microsomal membranes. Translocation of the normal and mutant pre␣1(X) chains into the microsomes was demonstrated by the removal of the signal peptide which resulted in the smaller ␣1(X) chains for all the plasmids translated (Fig. 2, lanes 7-12). As shown in previous studies (9), under these translation conditions the normal human type X collagen associated into a trimeric component (Fig.  2, lane 11). The helix⌬ ␣1(X) and NC2⌬ ␣1(X) chains also formed trimers in vitro (Fig. 2, lanes 12 and 7), whereas the NC1 mutants, Y598D (Fig. 2, lane 8), NC1⌬10 (Fig. 2, lane 10), and NC1⌬C (Fig. 2, lane 9) were unable to form trimers in vitro.
An important functional consideration is whether the mutant ␣1(X) chains are able to associate with normal ␣1(X) into heterotrimers. To determine this, normal and mutant plasmids were co-transcribed and translated in the presence of microsomes. Co-translation of the normal transcript and the helix⌬ transcript, which contained a helical deletion but a normal NC1 domain, showed that in addition to the trimeric components of each product, two additional multimer bands labeled as a and b were also observed (Fig. 4, lane 1). The relative electrophoretic migrations of these suggested that band a represented heterotrimers containing two normal ␣1(X) chains and one helix⌬ ␣1(X) chain, while band b was the heterotrimer containing one normal and two helix⌬ ␣1(X) chains. The efficient formation of helix⌬ heterotrimers with the full-length normal ␣1(X) chains demonstrated an important practical use of the helix⌬ construct as a protein reporter to assess the ability of the mutants to assemble into heterotrimers in further co-translation experiments. If the mutants can assemble with the helix⌬ chains, which are shorter but contain a normal NC1 domain, the stoichiometry of these heterotrimers can be readily determined by assessing the electrophoretic migration of the multimers.
Co-expression of the helix⌬ ␣1(X) reporter construct and the NC2⌬ ␣1(X) construct (Fig. 4, lane 8) showed that the NC2⌬ mutation allowed assembly of the mutant chain into homotrimers and heterotrimers (lane 8, labeled as c and d, respectively). Molecular weight analysis again indicated that band c represented heterotrimers containing two NC2⌬ and one he-lix⌬ chains, while band d was a heterotrimer containing one NC2⌬ and two helix⌬ chains.
In contrast, co-translation of normal ␣1(X) (wt) with NC1⌬C or NC1⌬10 transcripts showed no evidence of any heterotrimer formation (Fig. 4, lanes 4 and 5). Similarly, no heterotrimers were observed when these two NC1 mutant transcripts were co-translated with helix⌬ (Fig. 4, lanes 2 and 3). Mutant transcripts with single amino acid substitutions within the NC1 domain (G618V, Y598D) were also co-translated with the he-lix⌬ reporter to determine if association of normal and mutant NC1 domains occurred in vitro (Fig. 4, lanes 6 and 7). The predominant multimeric component in each co-translation was the helix⌬ ␣1(X) trimer and the two mutants were unable to form homotrimers in vitro. However, a faint band corresponding to species b (Fig. 4, lane 1) was evident with both the G618V and Y598D (Fig. 4, lanes 6 and 7) and was better demonstrated in an overexposure (Fig. 4, lanes 9 and 10). This species corresponds to a mutant:normal heterotrimer containing one mutant NC1 ␣1(X) chain and two reporter helix⌬ ␣1(X) chains. This species was a very minor product (Ͻ1%), demonstrating that heterotrimer formation was extremely inefficient. There was no evidence of trimers containing two or three mutant chains.
Transient Expression of Normal and Mutant Type X Collagen in Transfected Cells-Transient transfection of the normal (wt), helix⌬ and NC2⌬ plasmids into UMR 106-01, which produce no endogenous type X collagen, demonstrated that these constructs (which contain normal NC1 domains) were expressed and secreted by these cells (Fig. 5, panel a). Similar results were obtained in transient transfections of CV1 and 3T3-A31 cells (data not shown). The normal and helix⌬ molecules were efficiently secreted by the cells within the labeling period. While the NC2⌬ molecules were also secreted, the efficiency of this secretion was consistently lower than that of the normal molecules, in three separate experiments. Furthermore, the molecules produced from these constructs (wt, helix⌬, and NC2⌬) were assembled into trimers which could be dissociated into their monomeric components by denaturation at 100°C for 5 min prior to electrophoresis under reducing (Fig. 5,  panel b) or nonreducing conditions (data not shown). Pepsin digestion resulted in pepsin resistant ␣-chains of 53 kDa for the normal and NC2⌬ molecules, while pepsin digestion of the helix⌬ molecules resulted in an ␣-chains of approximately 27 kDa (Fig. 5, panel c). The formation of these pepsin resistant molecules indicated that ␣1(X), helix⌬ ␣1(X), and NC2⌬ ␣1(X) chains had assembled into stable triple helical molecules. Additional pepsin resistant bands corresponding to endogenous ␣1(I) and ␣2(I) chains of type I collagen were also present in the UMR 106-01 samples, including the control untransfected cells that were infected only with the vTF7-3 virus, demonstrating that the infected cells continued to provide the necessary collagen assembly, post-translational modification, and secretion machinery.
A very different pattern emerged when the plasmids containing the NC1 mutations (G618V, Y598D, NC1⌬C, and NC1⌬10) were transiently expressed in UMR 106-01 cells (Fig. 6). The level of expression of the mutant proteins was consistently lower (Fig. 6, panel b) than the control, helix⌬, and NC1⌬C ␣1(X) chains expressed under identical conditions in the same cells (Fig. 5). The faint mutant ␣1(X) chain monomeric components were difficult to distinguish from other protein bands within the cell fraction (Fig. 6, panel b) but were identified by immunoblotting (Fig. 6, panel a). The most striking observation was the absence of type X collagen in the medium fraction (Fig.  6, panel a), indicating that the mutant collagen was not secreted by the UMR 106-01 cells. The intracellular mutant ␣1(X) chains were not resistant to pepsin digestion (Fig. 6, panel c) demonstrating that they were unable to form correctly assem-bled triple helical type X collagen. Intracellular mutant chains were more evident in short term labeling of the transfected UMR 106-01 cells. Extension of the labeling period from 8 to 18 h resulted in a decrease in the amount of mutant ␣1(X) chain detected, suggesting that the unassembled NC1 mutant ␣1(X) chains were unstable and degraded intracellularly (data not shown).
The defective assembly and secretion of the NC1 mutant ␣1(X) was best demonstrated in a transfection of CV1 cells with the normal and G618V mutant ␣1(X) (Fig. 7), and confirmed the results obtained in transfections of UMR 106-01 cells. The normal (wt) chains were fully assembled into trimers intracellularly, more than half of which were secreted into the medium during the labeling period (Fig. 7, panel a). Pepsin digestion demonstrated that the secreted type X trimer was stable to pepsin digestion and thus triple helical (Fig. 7, panel b). However, pepsin digestion of the intracellular trimeric type X collagen revealed that only a portion of this assembled collagen was helical. This may reflect delayed helix formation in these cells due to inadequate machinery to deal with the high levels of ␣1(X) from T7-driven cytoplasmic overexpression, or it may result from inherent differences between the rates of NC1driven assembly and subsequent helix formation of type X collagen. The expression of the G618V mutant in CV1 demonstrated that the mutant ␣1(X) chains were present as unassembled monomeric chains which were not secreted. However, pepsin digestion of cell and medium samples from transfected CV1 (Fig. 7, panel b) resulted in the detection of a very faint pepsin-resistant band in the cell fraction, with similar molecular weight to the normal chain, indicating that in this experiment a very small proportion of the intracellular G618V chains assembled into stable helical molecules. In other transfections of UMR 106-01 cells, overexposure of the fluorograms demonstrated the Y598D and NC1⌬10 mutant chains also form a minute amount of pepsin stable molecules, but this was even less efficient than the G618V mutant. No pepsin-resistant components were detected for the NC1⌬C chains in any transfection experiments (data not shown).
While these data clearly demonstrated that in multiple experiments the assembly and secretion of the NC1 ␣1(X) was severely compromised in transiently transfected cells, some assembly and secretion was observed in one experiment. UMR  6. Expression of G618V, NC1⌬C, NC1⌬C, and Y598D type X collagen constructs in transiently transfected UMR 106-01 cells. 106-01 cells transiently transfected with NC1 mutants secreted trace amounts of pepsin resistant ␣1(X) G618V chains, and even smaller amounts of ␣1(X) Y598D and ␣1(X) NC1⌬10, after 18 h of labeling. There was no evidence of any secretion of the NC1⌬C mutant chains (data not shown).

DISCUSSION
Despite intense interest, relatively little is known about type X collagen biosynthesis and assembly, and its functional role in the hypertrophic chondrocyte extracellular matrix remains controversial. To gain insight into the protein domains which control type X collagen trimer assembly and helix formation, mutations were produced in the ␣1(X) noncollagenous aminoterminal NC2 domain, the carboxyl-terminal NC1 domain, and the collagenous triple helical COL1 domain using overlap extension PCR. To assess the effects of these mutations on type X collagen assembly and helix formation, the mutant constructs were expressed in vitro using T7-driven coupled transcription and translation, and in transiently transfected cells.
Consistent with our previous finding of a NC1 G618V substitution in a patient with SMCD (9), the three additional mutations of the NC1 domain which represent the spectrum of naturally occurring mutations in SMCD, similarly prevented the in vitro assembly of the mutant ␣1(X) chains into trimers. While no detectable homotrimer was evident from these NC1 mutants in vitro, it was of considerable importance to test whether the mutant chains can associate with normal NC1 domains and interfere with normal chain assembly, since the analogous NC1 mutations in SMCD are heterozygous. To test this, an ␣1(X) protein reporter chain construct was produced which contained a normal NC1 domain but an in-frame helix deletion (helix⌬) allowing discrimination of heterotrimers containing the reporter by a molecular weight shift on electrophoresis. Assembly analysis of these chains with the helical deletion (helix⌬) have resulted in a number of important observations. First, the efficient association of the helix⌬ chains into trimers during cell-free translation further demonstrates the crucial role of the NC1 domain in this in vitro assembly process which appears to be unaffected by the partial removal of helical sequences. Co-translation with normal ␣1(X) chains resulted in the formation of helix⌬ ␣1(X) and normal ␣1(X) heterotrimers of all possible chain configurations. The ability of the helix⌬ chain to freely associate with normal type X collagen ␣1(X) chain suggests that heterozygous helical mutations would result in the generation of aberrant type X collagen heterotrimers and exert a "dominant negative" effect in vivo, with the biochemical consequences of increased intracellular degradation and matrix disruption similar to that seen with type I collagen helix mutations in osteogenesis imperfecta (24). This suggestion is supported by results from a transgenic mouse model heterozygous for a deletion in the type X collagen helical domain (18). The transgenic mice expressed a more severe clinical phenotype with skeletal deformities consistent with a spondylometaphyseal dysplasia phenotype.
Co-expression of the reporter helix⌬ and the NC1 mutant constructs demonstrated that the NC1 mutations prevented in vitro heterotrimer formation. The data indicated that the NC1 deletion mutations, which either truncated (NC1⌬C) or added abnormal carboxyl-terminal sequences (NC1⌬10), totally prevented heterotrimer assembly. On the other hand, point mutations allowed the incorporation, albeit inefficiently, of one mutant NC1 chain into the trimeric assembly with two normal NC1 domains. These results suggest that amino acid substitution mutations compromise assembly by introducing more subtle changes to NC1 protein structure which can be partially accommodated in trimeric assemblies containing two normal NC1 domains.
While the experiments expressing the protein by transcription and translation provided important information on the effect of these NC1 mutations on assembly in vitro, cellular collagen assembly mechanisms are infinitely more complex, with coordinated intracellular post-translational processing as well as possible interactions with accessory proteins which may regulate both normal assembly and secretion and the fate of the mutant ␣1(X) chains. To address these problems, normal and mutant ␣1(X) was expressed in transiently transfected cells. Expression studies in UMR 106-01 cells, a collagen-producing osteoblast-like cell line, indicated that chains with normal NC1 domains (wt ␣1(X), helix⌬ ␣1(X), and NC2⌬ ␣1(X)) can be assembled and secreted efficiently by the cells. The formation of pepsin resistant components indicated that these assembled trimers were post-translationally hydroxylated and assembled into stable triple helical molecules (25).
Transient expression of the mutant NC1 plasmids in transfected cells confirmed the results of the in vitro expression and assembly experiments, demonstrating that all the NC1 ␣1(X) mutations resulted in a biochemical phenotype where assembly and secretion were severely compromised. This result was consistent in collagen-producing cells (UMR 106-01 and 3T3-A31) and cells which produce little or no collagen (CV1). The NC1 region of type X collagen contains a sequence domain of approximately 130 amino acid residues including a conserved cluster of aromatic acids which has a hydrophobicity profile of marked similarity to that of the carboxyl-terminal propeptides of the fibrillar collagens. Mutations in the NC1 domain which disturb this aromatic cluster, or change NC1 protein folding such that hydrophobic surface interactions are impaired, would be expected to result in the observed failure to trimerize.
The mutant NC1 collagen retained within the cell was in an unassembled, pepsin-sensitive form. Taken together with the consistently lower levels of protein expression seen with the mutant constructs compared to control construct transfected in parallel, these results suggest that the mutant collagen is unstable and rapidly degraded intracellularly. Such an intracellular degradative "quality control" mechanism for type I collagen with structural mutations has been described (26 -28). While the vast majority of the NC1 mutant ␣1(X) chains were retained within the cell and degraded, in one experiment a trace level of mutant assembly and secretion was detected. While these preliminary studies are not totally conclusive, the data suggest that there may be subtle biochemical heterogeneities in NC1 mutations. In general, amino acid substitutions in the NC1 domain showed a limited ability to assemble, whereas mutations causing frameshifts and premature terminations had a more severe effect on the protein-protein interactions involved in assembly. These results have possible implications for the molecular pathology of type X collagen mutations and begin to address the apparent contradictions between the phenotypes of the type X null mice which have little or no abnormal skeletal phenotype (29) and the NC1 mutations that result in SMCD. Previous in vitro studies have suggested that the SMCD mutations completely prevent assembly and exclude the mutant allele ␣1(X) product leading to a haploinsufficiency (9), at apparent odds to the reported findings in the type X null transgenic mice. The data presented here suggest that while ␣1(X) haploinsufficiency due to compromised assembly, secretion, and intracellular breakdown is likely to be the major defect in SMCD, it is possible that some NC1 mutations may result in the assembly of trace amounts of mutant:normal heterotrimers and/or mutant:mutant homotrimers which may be secreted and further disrupt the type X collagen matrix.
These transient transfection studies provide the first direct information on how cells handle mutant type X collagen. It is important, however, to recognize some possible limitations of the vaccinia-based transfection protocol. Vaccinia virus infection leads to cell death, and it is possible that during these expression studies some cell lysis may have occurred resulting in the unphysiological release of some type X collagen. To clarify this important point and determine if mutant collagens are secreted from the cells, it is crucial that these cells are stably transfected with the mutant constructs and that assembly and secretion be studied in detailed kinetic experiments.
The role of the ␣1(X) amino-terminal NC2 domain in assembly was also explored by site-directed mutagenesis. Removal of the NC2 domain did not prevent NC2⌬ homotrimer formation or affect the ability of the mutant chains to associate with normal ␣1(X) into heterotrimers in cell-free translations or into homotrimers in transiently transfected cells. While NC2⌬ mutant containing homotrimers were secreted from the cells, the efficiency of this will need to be accurately quantified in pulsechase experiments of stable transfectants to determine if this domain plays a subtle role in regulating assembly and the efficiency of secretion. Recent data suggest that type X collagen can form several supramolecular assemblies, either as fine pericellular filaments, in association with type II collagen fibrils (30), or as a hexagonal lattice (10). The NC2 domain is likely to play an important role in the formation and stabilization of these structures, and mutations of the NC2 could be expected to have a severe effect on the structure and stability of these assemblies. It has also been suggested that the hexagonal lattice is stabilized, in part, by lateral overlapping of the type X collagen helical domains (10), and helical mutations should thus also compromise these supramolecular structures.