Metaphyseal Chondrodysplasia Type Schmid Mutations Are Predicted to Occur in Two Distinct Three-dimensional Clusters within Type X Collagen NC1 Domains That Retain the Ability to Trimerize*

Metaphyseal chondrodysplasia type Schmid (MCDS) is caused by mutations in COL10A1 that are clustered in the carboxyl-terminal non-collagenous (NC1) encoding domain. This domain is responsible for initiating trimerization of type X collagen during biosynthesis. We have built a molecular model of the NC1 domain trimer based on the crystal structure coordinates of the highly homologous trimeric domain of ACRP30 (adipocyte complement-related protein of 30 kDa or AdipoQ). Mapping of the MCDS mutations onto the structure reveals two specific clusters of residues as follows: one on the surface of the monomer which forms a tunnel through the center of the assembled trimer and the other on a patch exposed to solvent on the exterior surface of each monomeric unit within the assembled trimer. Biochemical studies on recombinant trimeric NC1 domain show that the trimer has an unusually high stability not exhibited by the closely related ACRP30. The high thermal stability of the trimeric NC1 domain, in comparison with ACRP30, appears to be the result of a number of factors including the 17% greater total buried solvent-accessible surface and the increased numbers of hydrophobic contacts formed upon trimerization. The 27 amino acid sequence present at the amino terminus of the NC1 domain, which has no counterpart in ACRP30, also contributes to the stability of the trimer. We have also shown that NC1 domains containing the MCDS mutations Y598D and S600P retain the ability to homotrimerize and heterotrimerize with wild type NC1 domain, although the trimeric complexes formed are less stable than those of the wild type molecule. These studies suggest strongly that the predominant mechanism causing MCDS involves a dominant interference of mutant chains on wild type chain assembly.


Metaphyseal chondrodysplasia type Schmid (MCDS)
is caused by mutations in COL10A1 that are clustered in the carboxyl-terminal non-collagenous (NC1) encoding domain. This domain is responsible for initiating trimerization of type X collagen during biosynthesis. We have built a molecular model of the NC1 domain trimer based on the crystal structure coordinates of the highly homologous trimeric domain of ACRP30 (adipocyte complement-related protein of 30 kDa or AdipoQ). Mapping of the MCDS mutations onto the structure reveals two specific clusters of residues as follows: one on the surface of the monomer which forms a tunnel through the center of the assembled trimer and the other on a patch exposed to solvent on the exterior surface of each monomeric unit within the assembled trimer. Biochemical studies on recombinant trimeric NC1 domain show that the trimer has an unusually high stability not exhibited by the closely related ACRP30. The high thermal stability of the trimeric NC1 domain, in comparison with ACRP30, appears to be the result of a number of factors including the 17% greater total buried solvent-accessible surface and the increased numbers of hydrophobic contacts formed upon trimerization. The 27 amino acid sequence present at the amino terminus of the NC1 domain, which has no counterpart in ACRP30, also contributes to the stability of the trimer. We have also shown that NC1 domains containing the MCDS mutations Y598D and S600P retain the ability to homotrimerize and heterotrimerize with wild type NC1 domain, although the trimeric complexes formed are less stable than those of the wild type molecule. These studies suggest strongly that the predominant mechanism causing MCDS involves a dominant interference of mutant chains on wild type chain assembly.
Type X collagen is a short chain, homotrimeric collagen (␣1(X) 3 ) expressed specifically by hypertrophic chondrocytes in the endochondral growth plate (1). The expression of type X collagen is also re-activated during fracture repair and in osteoarthritis (2)(3)(4). The human ␣1(X) collagen chain consists of a short amino-terminal non-collagenous domain 2 of 37 amino acids followed by a triple helix-forming collagenous domain with 154 Gly-X-Y repeats and a carboxyl-terminal non-collagenous domain (NC1) 1 of 161 amino acids (5). The ␣1(X) 3 molecule is thought to assemble into a hexagonal lattice within the extracellular matrix in a fashion similar to that of type VIII collagen (6,7). Type X collagen is part of a family of collagenlike proteins sharing a condensed gene structure, a collagen triple helical domain, and in particular, a highly conserved carboxyl-terminal non-collagenous (NC1-like) domain (8,9). This family includes collagen types X and VIII, C1q component of complement, hibernation proteins (10), cerebellin (11,12), and ACRP30, an abundant serum protein implicated in energy homeostasis and obesity (13).
The precise function of type X collagen remains to be determined (14 -16), but mutations in the COL10A1 gene cause metaphyseal chondrodysplasia type Schmid (MCDS), an autosomal dominant form of human skeletal dysplasia (17,18). An intriguing finding is that virtually all of the mutations causing MCDS occur within the carboxyl-terminal non-collagenous NC1 domain of type X collagen (Ref. 19 and references therein). The only two MCDS mutations not found in the NC1 domain affect the putative signal peptide cleavage site upstream of the non-collagenous domain 2 in the molecule (20). The role of the carboxyl-terminal non-collagenous domain of most collagens (including type X) in initiating intracellular ␣-chain selection, assembly, and helix formation is well established (Refs. 8 and 21 and references therein). Assembly studies based on cell-free translation of recombinant RNA encoding wild type and MCDS transcripts of type X collagen suggest that mutant chains do not interfere with the trimerization of the wild type protein based on SDS-PAGE assays (22). These findings have led to the suggestion that the phenotype of MCDS is best explained by haplo-insufficiency. In support of this hypothesis, in one individual with MCDS, only wild type mRNA for type X collagen could be detected in the growth plate, suggesting that the mutant transcript is unstable and rapidly degraded (23). However, haplo-insufficiency is in discord with the clustering of both point and frameshift mutations in the NC1 domain of type X collagen; in haplo-insufficiency one would expect to find MCDS-causing frameshift mutations to be randomly distributed through the gene (19). The non-random clustering of MCDS mutations in the NC1 domain is more consistent with a mechanism involving dominant interference in which the mutant chains retain the ability to trimerize. It is possible that previous investigations have failed to detect trimerization of chains containing MCDS mutations due to the harsh assay conditions, such as SDS-PAGE, that have been employed.
The crystal structure of the NC1-like domain of ACRP30, a protein closely related to type X collagen, has been recently solved at 2.1-Å resolution (24). We have therefore performed molecular modeling of the NC1 domain, based on the ACRP30 crystal structure coordinates, to gain greater insight into the structural basis of MCDS. The model reveals that the MCDS mutations are localized to two specific regions of the folded monomeric NC1 domain and that many of the MCDS mutations may not totally abolish the ability of affected chains to trimerize. Mutagenesis experiments, based on information obtained from the model, demonstrate that NC1 domains containing specific MCDS mutations retain the ability to form trimers and provide an explanation for the unusually high thermal stability exhibited by the assembled NC1 trimer.

EXPERIMENTAL PROCEDURES
Alignments-The amino acid sequence of the NC1 domain of human type X collagen was used to probe EMBL, GenBank TM , SwissProt and PIR data bases as well as data base updates using the advanced BLAST 2.0 search (25). We used a cut-off expectancy value (E) of 1 (such that no more than 1 match is expected to be found merely by chance, according to the stochastic model of Karlin and Altschul (26)) and a gapped alignment using the Blosum 62 matrix (27). Multalin (28) was used to re-align the resulting 25 sequences using the Blosum 62 matrix using a penalty of 12 for gap opening and 2 for gap extending. None of the resulting aligned sequences have three-dimensional structures in the Brookhaven data base.
Molecular Modeling-All molecular modeling was performed on an R5000 O2 Silicon graphics workstation using QUANTA® and CHARMm 23.1® programs. The three-dimensional model of the NC1 domain of type X collagen was built based on the coordinates of human ACRP30. 2 The file containing the coordinates was rewritten manually to make a Protein Data Base file format that could be imported by QUANTA®. A homology model was built by copying the coordinates of the backbone of the ACRP30 trimer and the coordinates of identical residues in the type X collagen NC1 domain. The remaining side chains were built in the Protein Design module using the Ponder and Richards' rotamer library. Disordered loops in the crystal structure of ACRP30 (loop G-H in M1 and loops G-H and A-AЈ in M2) were modeled where possible by overlaying the intact loop of a corresponding part of one of the other two monomers. This left a gap between Glu 658 to Gly 661 (corresponding to Asp 224 to Gly 228 in ACRP30) in each monomer.
The resulting trimer of the NC1 domain of type X collagen was energy-minimized (using the steepest descents followed by Newton-Raphson algorithm) to gradient convergence (Ͻ0.01 root mean square) removing bad steric and electrostatic contacts. The water molecules from the crystal structure were overlaid, and those internal to the trimer were incorporated into the model. The Protein Health module in QUANTA was used to check the integrity of the model using a Ramachandran map and to identify buried hydrophilic or exposed hydrophobic residues.
The solvent-accessible surfaces of monomers and assembled trimers were calculated to estimate the surface area buried upon trimerization. The local environment of each residue on the interfaces was examined to locate those residues important for the trimer structure. The interfaces are not symmetric, because of a slight twist and stagger to the arrangement of the ␤-sandwich monomers (24). Particular attention was given to the position of those residues in the NC1 model trimer that differed in ACRP30 and related proteins (see Fig. 1).
Preparation of the Constructs-DNA constructs were generated by PCR from the lambda genomic COL10A1 clone, HX3 (5). DNA encoding the entire human type X collagen NC1 domain was amplified using the primers 5Ј-CCAGGTCAAGCACATATGCCTGAGGGT-3Ј (primer A, sense nucleotides 1550 -1563), which incorporates an NdeI site (C (A/T)ATG) in-frame with the Met 521 codon of the NC1 domain, and 5Ј-GGGGTGTACTCACATTGGAGCCAC-3Ј (primer B, antisense nucleotides 2028 -2052), which incorporates the stop codon and 9 base pairs of the 3Ј-untranslated region. To generate the extended reporter construct, DNA encoding the entire NC1 domain and upstream DNA encoding the 11 most carboxyl-terminal Gly-X-Y repeats of the collagenous domain was generated by PCR. A sense oligonucleotide of sequence 5Ј-CCAGCTCATATGGCAACTAAGGGCCTC-3Ј (nucleotides 1429 -1455), which incorporates an NdeI site coding for an in-frame Met codon at the amino terminus of the extended NC1 coding region, was used together with the antisense primer B. To produce the truncated NC1 translation construct, DNA encoding a portion of the NC1 domain from the Met 549 codon to the stop codon was generated by PCR using the sense primer 5Ј-GGGGTAACACATATGCCTGTGTCT-3Ј (nucleotides 1632-1656) together with the antisense primer B. To create constructs containing MCDS and related point mutations, site-directed mutagenesis was carried out using the PCR-based single overlap extension procedure described elsewhere (29) using oligonucleotides A and B (described above) and mutagenic oligonucleotides (nucleotides 1875-1899) sense, 5Ј-CCAGGAATATACTTTTTTTCATAC-3Ј and antisense, 5Ј-GTATGAAAAAAAGTATATTCCTGG-3Ј for Y598F; sense, 5Ј-CCAGGAATATACGCTTTTTCATAC-3Ј and antisense, 5Ј-GTATGA-AAAAGCGTATATTCCTGG-3Ј for Y598A; 5Ј-CCAGGAATATACGATT-TTTCATAC-3Ј and antisense, 5Ј-GTATGAAAAATCGTATATTCCTG-G-3Ј for Y598D; and (nucleotides 1881-2004) sense, 5Ј-ATATACGCTT-TTCCATACCACGTG-3Ј and antisense, 5Ј-CACGTGGTATGGAAAAG-CGTATAT-3Ј for S600P. The resulting PCR fragments were cloned into the T/A cloning vector pCR 2.1 (Invitrogen). Constructs generating sense RNA upon transcription from the T7 promoter were identified by restriction mapping and sequenced.
Coupled in Vitro Transcription and Translation-In vitro expression was carried out using the TNT T7 polymerase-coupled rabbit reticulocyte lysate system (Promega). 1 g of each purified construct was expressed in 40 l of TNT premix containing 2 l of [ 35 S]methionine (1000 Ci/mmol, NEN Life Science Products) made up to a final volume of 50 l with sterile distilled water. Reactions were incubated at 30°C for 90 min. 5-l aliquots of in vitro expression reaction were added to 10 l of sample buffer (50 mM Tris/HCl, pH 6.8 (room temperature), 50% (v/v) glycerol, 0.025% (w/v) bromphenol blue; the concentration of SDS in the sample buffer was either 10% (w/v) or 0.75% (w/v) as indicated under "Results"). Where heating is indicated, samples were overlaid with 20 l of mineral oil and incubated at the appropriate temperature for 5 min using Perkin-Elmer 480 thermal cycler prior to analysis by SDS-PAGE. All samples were resolved on standard 12 or 16% PAGE gels containing 0.1% SDS according to Laemmli (30). Gels were fixed (10% (v/v) acetic acid containing 10% (v/v) methanol) for 30 min and washed twice in 30% (v/v) methanol containing 3% (v/v) glycerol. Gels were dried and exposed to Kodak Biomax Film, or, if quantitation was required, phosphorimaged (Fuji-Bas).
Immunoprecipitation of NC1 Domains with the X34 Monoclonal Antibody-In vitro translations containing labeled NC1 domains were subjected to centrifugation at 109,000 ϫ g for 1 h to remove particulates. 50 l of the supernatant was then dialyzed against NET buffer overnight (100 mM Tris/HCl, pH 7.4, 100 mM NaCl, 1% (v/v) Tween 20, 1 mM EDTA). 5 l of hybridoma medium containing the X34 monoclonal antibody, which recognizes a conformation-dependent epitope in the NC1 domain of native type X collagen (31), was added to the dialyzed translation followed by incubation for 16 h at 4°C with gentle inversion. Following the incubation, 10 l of protein A-Sepharose (Amersham Pharmacia Biotech) was added, and the incubation was allowed to proceed for a further 4 h, after which the protein A-Sepharose was recovered by centrifugation at 12,000 ϫ g (control incubations were carried out in the absence of X34). The pelleted protein A-Sepharose was washed in 500 l of 1ϫ NET buffer three times followed by resuspension in sample buffer containing 1% (v/v) ␤-mercaptoethanol. Prior to SDS-PAGE analysis, the samples were heated for 2 min at 68°C.
revealed that those residues forming ␤-strands of ACRP30 were mostly conserved in the NC1 domain, whereas differences were predominantly localized to the loop regions (Fig. 2). Therefore, the secondary structural elements are highly conserved. Within the ␤-strands, those residues forming the hy-drophobic core of the monomer of ACRP30 (see below) are particularly conserved. This conservation strongly indicates that the tertiary fold of the NC1 domain is identical to the ␤-sandwich of ACRP30. In addition, the conservation of key interface residues between monomers (see below and indicates that the quaternary structures of these two proteins are similar. This pattern of conservation, based on alignments, also holds across the complete family of related proteins including collagen types X and VIII, C1q, ACRP30, hibernation protein and cerebellin (see Fig. 1).
Molecular Model of Trimeric NC1 Domain-A model of the trimeric NC1 domain (Fig. 3) was built using the three-dimensional coordinates of trimeric ACRP30 based on the alignment shown in Fig. 2. The amino-terminal 27 residues of the NC1 domain of type X collagen were not included in the model since FIG. 3. a, type X collagen non-collagenous carboxyl-terminal domain (Met 548 to Met 682 ) trimer in protein cartoon representation. Modeled using the crystal structure of ACRP30 trimer (24) in QUANTA®. Each monomer is a different color. Both the amino and carboxyl termini of this domain emerge at the base; the amino terminus has an additional 27 residues followed by the triple helical domain. The asymmetry can be seen particularly with the (blue) loop of M3 near the apex of the trimer. b, type X collagen NC1 trimer viewed from the base looking down the tunnel formed in the center by the association of the three monomers. The backbone is represented in pale green; hydrophobic residues at the base of the tunnel are shown in red licorice. c, Tyr 598 and Ser 600 shown from each monomer as viewed looking down the tunnel from the base. A putative network of hydrogen bonds is formed with water molecules and direct monomer-monomer contacts between these side chains. there is no equivalent sequence in ACRP30 (Fig. 2). The 8 loop regions in the NC1 domain were modeled on the ACRP30 structure since their lengths are conserved, and they are anchored spatially by the ␤-strands.
Upon trimerization (Fig. 3a), the mode of packing between the NC1 monomer subunits produces a tight association contributed to by the loss of solvent-accessible surface and hence providing a free energy gain. The total buried solvent-accessible surface of the NC1 domain on trimerization is 6222 Å 2 , compared with 5324 Å 2 for ACRP30. The three monomers pack together forming a twisted tunnel that extends through the core of the structure. The hydrophobic contacts that produce the close packing between the NC1 monomers are identical or show only conservative changes with respect to ACRP30. These include Ala 553 , Ile 557 , Leu 575 , Ile 596 , Val 621 , Ile 641 , Phe 675 , and Ala 678 in the NC1 domain. However, in the NC1 domain additional residues (Pro 550 , Val 551 , Val 668 , and Pro 679 ) contribute to hydrophobic interfaces between monomers. The equivalent residues in ACRP30 are Tyr 114 , Arg 115 , Asn 233 and His 244 , respectively.
Pro 550 , Val 551 , Ile 641 , Phe 675 , and Pro 679 form a hydrophobic plug at the base of the NC1 trimer tunnel as viewed in Fig. 3b. Moving up from the base, the trimer tunnel surface, contributed by all three monomers, becomes more polar with Tyr 598 , Ser 600 , Ser 639 , and Ser 671 forming a dense network of hydrogen bonds with each other and water molecules (e.g. see Fig. 3c). These polar residues are conserved across the entire family of type X collagen-related proteins (Fig. 1).
Examining the monomer-monomer interfaces is complicated by the asymmetric association of the three monomers (designated M1-M3; see Fig. 3a). Significant monomer-monomer contacts include a backbone hydrogen bond between the carbonyl oxygen of Tyr 623 and the amide nitrogen of His 669 (between M2-M3 and M3-M1). M1-M2 has a different backbone contact involving the carbonyl oxygen of Gly 631 and the amide nitrogen of Leu 633 . Examination of the NC1 trimer model revealed that the hydrophobic side chains of Pro 568 , Pro 570 , Ile 588 , Trp 611 , and Trp 651 are exposed to solvent on the external surface of the trimer.
Location of MCDS Point Mutations in the NC1 Trimer-All NC1 domain residues that are affected by single amino acid substitutions causing MCDS were displayed simultaneously on the monomer (Fig. 3d) and trimer (Fig. 3e) models revealing a striking clustering of position. The substituted residues are localized to two distinct regions. The side chains of residues that are substituted in MCDS either form the polar and hydrogen-bonded region within the trimer tunnel (e.g. Tyr 598 and Ser 600 , see Fig. 3c) or form a patch on the external surface of the monomer/trimer assembled from part of loop D-E and part of ␤-strand G (Figs. 1 and 2). A detailed summary of the location and orientation of amino acid residues substituted within the NC1 domain in MCDS is presented in Table I.
Trimerization and Stability of Recombinant Human Type X Collagen NC1 Domain-The model was first used in an attempt to predict motifs or residues responsible for the unusual stability of the NC1 domain of type X collagen (see below and "Discussion"). Expression of the full-length NC1 domain in a reticulocyte lysate system resulted in detection of the monomeric protein (M r ϭ 20,000) and a complex of M r ϭ 45,000 that is stable on SDS-PAGE in reducing conditions (Fig. 4a, lane 4). Similarly, expression of an "extended" NC1 construct containing 11 Gly-X-Y repeats of the collagenous domain of type X collagen (Fig. 4a, lane 3) and a "truncated" construct lacking the amino-terminal 27 residues of the NC1 domain (Fig. 4a, lane 1) produced monomers of M r ϭ 28,000 (extended) and 17,000 (truncated) as well as complexes of M r ϭ 60,000 and 35,000 (respectively). Based on M r , the complexes formed appeared too large to be dimers but too small to be trimers. In order to resolve the molecular composition of the complexes, co-expression studies were performed. Co-expression of the truncated and extended constructs (Fig. 4a, lane 2) and fulllength and extended constructs (Fig. 4a, lane 5)   On ␤-strand G, on outside face of trimer, side chain partially exposed and points towards loop II Aromatic ring almost all exposed; could be involved in hydrophobic contact with other molecule or another trimer a Indicates residues located in a cluster on an interior patch. b Indicates residues located on a surface patch on the exterior of trimer.
bands were formed in each co-expression reaction indicates clearly that the complexes are trimeric associations of the monomers running slightly faster on SDS-PAGE than would be predicted. It should be noted that the data in Fig. 4a were produced adding the translation mixes to sample buffer containing 10% SDS (final SDS concentration 6.7%) and running on the SDS-PAGE gel without prior heating.
To check that the trimeric NC1 domains had adopted a native conformation, translation reactions containing co-expressed full-length and extended NC1 domains were immunoprecipitated with the X34 monoclonal antibody which recognizes a conformation-dependent epitope in the native trimeric NC1 domain but not monomeric NC1 (31). SDS-PAGE analysis of the immunoprecipitated co-translation reactions revealed that the X34 monoclonal antibody recognized the trimeric but not monomeric NC1 domains present in the co-translation. The control immunoprecipitation carried out in the absence of X34 revealed that low levels of monomeric and trimeric NC1 domain bound nonspecifically to the protein A-Sepharose (Fig. 4b,  lane 3). Comparison of the control immunoprecipitation with that carried out in the presence of the X34 antibody demonstrated that in addition to adopting a native structure, all of the trimeric NC1 domain was stable in the conditions used for SDS-PAGE analyses.
In order to investigate the stability of the trimers formed in the system, translations of the full-length NC1 domain (Fig. 4c) were heated for 5 min in 6.7% SDS at the temperatures indicated prior to loading on the gel. The trimer formed is stable at 60°C and, while some denatures to monomer at 80°C, a significant proportion is still trimeric at 100°C (see also Fig. 5a). The full-length monomer appears unaffected by the exposure to increasing temperatures (Fig. 4c). ACRP30 lacks the aminoterminal 27 residues found in the NC1 domain of type X collagen (Fig. 2) and also does not form a trimer capable of withstanding SDS-PAGE (13). In order see whether the aminoterminal NC1 sequence plays a role in determining stability, it was deleted. In comparison with the full-length NC1 domain (Fig. 4c), the domain lacking the amino-terminal 27 residues responded in a different fashion in two critical ways. First, as the temperature increases above 40°C, the monomer is seen to aggregate into a high M r complex that does not enter the resolving gel (Fig. 4d). Second, the trimer, which is stable at 60°C and unaffected by the aggregation process exhibited by the monomer, is completely denatured at 80°C (Fig. 4d).
Assembly and Stability of NC1 Domains Containing MCDS and Related Mutations-In order to examine the role in determining trimer stability of Tyr 598 , which forms a network of hydrogen bonds in the trimer tunnel (Fig. 3c) and is mutated to Asp in one case of MCDS (Table I), the stabilities of NC1 domains containing Y598F, Y598A, and Y598D were examined (Fig. 5). Removal of the OH group (Y598F) produced an approximately 10°C drop in the melting temperature of the trimer, and a further 10°C decrease was found for the trimer containing the Y598A mutation. Under these conditions, in which the sample is placed in 6.7% SDS to prevent "geling" of the translation mix during heating prior to loading on the gel, a trimer could not be detected when expressing the construct containing the MCDS mutation Y598D (see Fig. 5b). Indeed, trimerization could not be detected for a second MCDS mutation, S600P, under the same conditions (Fig. 5b).
To investigate whether the high level of SDS (6.7% final concentration) to which samples were exposed prior to gel FIG. 5. a, analysis of the thermal stability of normal, Y598F, and Y598A NC1 domain trimers. Samples were analyzed as described in Fig. 4, b and c, having been heated in 6.7% SDS prior to loading on the gel. The relative proportion of trimer: total labeled NC1 domain was quantified in each case by phosphorimaging the gel. The data plotted represents the mean Ϯ S.E. (n ϭ 3). The autoradiograph to the right of the graph shows radiolabeled normal (N), Y598F, and Y598A monomers (NC1 1 ) and trimers (NC1 3 ) as separated by SDS-PAGE after incubation for 5 min at 20°C in the above conditions. b, SDS-PAGE analysis of NC1 homotrimers harboring MCDS point mutations. 5-l aliquots of translation reactions containing 35 S-labeled normal NC1 domain (N) and NC1 domains harboring Y598D and S600P substitutions were mixed with sample buffer (6.7% SDS final concentration) and separated, without prior heating, by SDS-PAGE followed by autoradiography. The positions of monomers (NC1 1 ) and the normal trimer (NC1 3 ) are indicated. c, detection of NC1 homotrimers and heterotrimers harboring MCDS mutations in low SDS conditions. Constructs encoding NC1 domains with the Y598D and S600P substitutions were expressed alone and co-expressed with the construct encoding the extended normal NC1 domain. 5-l aliquots of translation reactions containing 35  electrophoresis was denaturing potential trimer complexes formed by MCDS constructs, the concentrations of SDS in the sample buffer were reduced. Translations separated under native conditions failed to resolve in the gel, as did samples taken up in a final concentration of 0.1% SDS and run on standard SDS-PAGE gels (data not shown). However, when samples were exposed to 0.5% SDS and (without heating) run on standard 0.1% SDS-PAGE gels, resolution of wild type monomers and trimers was apparent (Fig. 5c, lanes 1 and 2). Analysis under the same conditions of NC1 domains containing the Y598D and S600P MCDS substitutions (Fig. 5c, lanes 3 and 5) demonstrated that each formed a trimeric complex. When the same MCDS translations were exposed to 1% SDS prior to running on the gel, the levels of detectable trimer were reduced by more than 50% (results not shown). Co-translation of each MCDS NC1 mutant construct with the extended wild type construct led to the formation of heterotrimers (Fig. 5c, lanes 4  and 6).

Structure and Assembly of the Wild Type Human NC1 Domain-
The NC1 domain of type X collagen is a putative asymmetric trimer composed of three 10 ␤-strand jelly roll motifs based on modeling of the ACRP30 crystal structure, which is remarkably similar to the structure of the TNFs (24). In both TNF and C1q, the sequences involved in molecular recognition (e.g. receptor and IgG binding, respectively) have been shown to include loop D--E and loop A-AЈ (see Ref. 24 and references therein). These sequences map to two specific exposed patches on each monomer within the assembled trimer. Interestingly, the residues forming these loops differ in composition and in length across the whole family of type X collagen-like proteins as shown in Figs. 1 and 2. These data suggest strongly that the equivalent regions in type X collagen are involved in molecular recognition events such as supramolecular assembly (6) or interaction with cells.
Previous studies on type X collagen assembly have demonstrated that trimers generated by in vitro translation possess an unusually high thermal stability (see Ref. 22 and references therein). We demonstrate here that this thermal stability is a consequence of the stability of the NC1 domain (see Fig. 4). It is of interest to note that the intact wild type NC1 domain trimerizes spontaneously forming a native structure (Fig. 4, a  and b), whereas when translating the full-length type X collagen polypeptide in the same system, either Ca 2ϩ or microsomal membranes are required to achieve trimerization (22). We assume that the presence of Ca 2ϩ , or microsomal membranes (and associated chaperone proteins), prevents the unfolded collagenous domain from interfering with the folding and trimerization of the nascent NC1 domain. Furthermore, it is noteworthy that recombinant human NC1 domain expressed in bacteria also spontaneously forms a thermally stable native trimer that is specifically recognized by the conformation-dependent X34 monoclonal antibody (32). ACRP30 and other members of this family of related proteins, with the exception of type VIII collagen (34), do not appear to exhibit this level of thermal stability. The modeling of the NC1 domain has revealed a number of features that may help explain the differences in stability. First, the model revealed that the packing of the aromatic side chains within the hydrophobic core of the monomers of both NC1 and ACRP30 is organized in a "herring bone" pattern similar to that associated with hydrophobic cores of other stable proteins (34). However, the NC1 domain contains a more densely packed hydrophobic core than ACRP30 because of the presence of three additional hydrophobic side chains that may increase stability of the monomer and trimer (35). Second, on trimerization, the total buried solvent-accessi-ble surface of the NC1 domain is 17% greater than ACRP30, and the hydrophobic plug at the base of the NC1 domain has additional contributions from Pro 550 , Val 551 , and Ile 588 (corresponding residues in ACRP30 are Tyr, Arg, and Gly, respectively). Third, sequence alignments show that type X collagen has an additional 27 residues at the amino terminus of the NC1 domain that are not present in ACRP30 or other members of the family of proteins with the exception of the two chains of type VIII collagen (Figs. 1 and 2).
To examine whether the additional 27-residue sequence stabilizes the NC1 domain, we compared full-length and truncated NC1 domains. In the absence of the 27 residues, monomeric NC1 forms high molecular weight aggregates at temperatures above 40°C (Fig. 4d) implying that the monomer has a defined structure in SDS at temperatures below 40°C that becomes denatured at higher temperatures leading to aggregation. However, no aggregation is exhibited by the full-length NC1 monomer, even at 100°C (Fig. 4c), strongly suggesting that it has a folded structure that is resistant to denaturation. The trimer is still formed by the shortened NC1 protein although its stability is reduced (compare Fig. 4, c and d) presumably because of the decreased stability of the monomer.
Structure and Assembly of NC1 Domains Containing MCDS and Related Mutations-Mapping MCDS point mutations onto the NC1 model revealed a striking three-dimensional clustering of affected residues into two distinct domains (Fig. 3, d and  e). The first cluster of residues substituted in MCDS is located on an external patch that includes loop D-E and is known to play a role in molecular recognition in structurally related molecules such as TNF and C1q as described above. Since these residues are located on the external surface of the trimer, it seems unlikely that the associated MCDS mutations would prevent the trimerization process. The second cluster of mutations affect the polar region of the trimer tunnel. It is significant that not one of the MCDS point mutations (see Table I) affect the hydrophobic plug at the base of the trimer tunnel (Fig. 3b) or any of the monomer-monomer hydrophobic contacts that are likely to drive the trimerization process.
On the basis of the findings described above, we predicted that NC1 domains containing MCDS point mutations would trimerize but form less stable complexes that may be difficult to detect on SDS-PAGE gels. In order to test this hypothesis, the role of Tyr 598 in determining NC1 trimer stability was examined. Tyr 598 not only participates in a hydrogen bond network as shown in Fig. 3c but is also mutated to Asp in MCDS ( Table  I). Replacement of the Tyr 598 by either Phe or Ala did not prevent the formation of SDS-resistant trimers although their thermal stabilities were considerably reduced (Fig. 5a). Under equivalent conditions, in which the translation mix was dissolved in sample buffer containing 6.7% SDS, no trimers could be detected when expressing NC1 domains containing MCDS Y598D or S600P mutations (Fig. 5b). However, both Y598D and S600P mutant NC1 domains were shown to form homotrimers, and heterotrimers with wild type NC1 domains, that are detectable in low concentrations of SDS (Fig. 5c) but denature as the level of SDS is increased (see Fig. 5b). These data clearly demonstrate that the stabilities of trimers containing MCDS mutant chains (in terms of sensitivity to the concentration of SDS and temperature) are considerably reduced compared with that of the wild type. Nevertheless, the fact that noncovalently associated MCDS mutant trimers can be detected on SDS-PAGE indicates that the stabilities of these mutant complexes are still greater than equivalent non-covalent trimeric complexes formed during the assembly of collagens with the exception of type VIII (33). We also have preliminary evidence suggesting that NC1 domains containing frameshift mutations causing MCDS also retain the ability to trimerize. 3 The MCDS mutant NC1 domains examined here not only retain the ability to trimerize but several ␣1(X) chains containing similar MCDS substitutions have recently been shown to fold stable collagenous triple helices. 4 In conclusion, the study presented here has produced a three-dimensional model of the NC1 domain of type X collagen that reveals a clustering of MCDS mutations that is consistent with the mutant chains retaining the ability to participate in the trimerization process. In addition, we have demonstrated that the NC1 domain forms an unusually stable trimer and that NC1 domains containing MCDS mutations retain the ability to trimerize. The modeling and protein assembly studies suggest strongly that the predominant molecular mechanism causing MCDS involves a dominant interference of mutant chains on wild type assembly.