The N-terminal N5 Subdomain of the α3(VI) Chain Is Important for Collagen VI Microfibril Formation*

Collagen VI assembly is unique within the collagen superfamily in that the α1(VI), α2(VI), and α3(VI) chains associate intracellularly to form triple helical monomers, and then dimers and tetramers, which are secreted from the cell. Secreted tetramers associate end-to-end to form the distinctive extracellular microfibrils that are found in virtually all connective tissues. Although the precise protein interactions involved in this process are unknown, the N-terminal globular regions, which are composed of multiple copies of von Willebrand factor type A-like domains, are likely to play a critical role in microfibril formation, because they are exposed at both ends of the tetramers. To explore the role of these subdomains in collagen VI intracellular and extracellular assembly, α3(VI) cDNA expression constructs with sequential N-terminal deletions were stably transfected into SaOS-2 cells, producing cell lines that express α3(VI) chains with N-terminal globular domains containing modules N9-N1, N6-N1, N5-N1, N4-N1, N3-N1, or N1, as well as the complete triple helix and C-terminal globular domain (C1-C5). All of these transfected α3(VI) chains were able to associate with endogenous α1(VI) and α2(VI) to form collagen VI monomers, dimers, and tetramers, which were secreted. Importantly, cells that expressed α3(VI) chains containing the N5 subdomain, α3(VI) N9-C5, N6-C5, and N5-C5, formed microfibrils and deposited a collagen VI matrix. In contrast, cells that expressed the shorter α3(VI) chains, N4-C5, N3-C5, and N1-C5, were severely compromised in their ability to form end-to-end tetramer assemblies and failed to deposit a collagen VI matrix. These data demonstrate that the α3(VI) N5 module is critical for microfibril formation, thus identifying a functional role for a specific type A subdomain in collagen VI assembly.

Although collagen VI is expressed in virtually all connective tissues, where it forms a complex and extensive microfibrillar network (see Refs. 1 and 2 for reviews), a definition of the precise functional roles of the collagen VI microfibrils in development and matrix architecture has been elusive. Even though the cell adhesion and extracellular matrix protein binding capabilities of collagen VI have suggested that it may play an important role in the interconnection between the cell and the structural scaffolding of the extracellular matrix, it is only recently that one such critical tissue-specific role in muscle has been demonstrated by the detection of mutations in collagen VI genes in Bethlem myopathy (3)(4)(5)(6) and by the generation of a myopathy by targeted inactivation of Col6a1 in mice (7).
Collagen VI is composed of three genetically distinct ␣-chain subunits, ␣1(VI), ␣2(VI), and ␣3(VI), each of which contain a relatively short triple helix and N-and C-terminal globular regions. Recent studies have demonstrated that initial association of all three chains into a collagen VI triple helical heterotrimer is a prerequisite for further intracellular assembly, stability and secretion (8). The ␣1(VI) and ␣2(VI) chains are similar in size (1009 and 998 amino acids, respectively) and contain one N-and two C-terminal subdomains of approximately 200 amino acids that show homology to the type A domains of von Willebrand factor (vWF) 1 (9). In contrast, the ␣3(VI) chain is much larger (up to 3149 amino acids), containing 6 to 10 N-terminal vWF type A-like subdomains (depending on alternative splicing), two C-terminal vWF type A-like subdomains, and subdomains similar to type III fibronectin repeats, and Kunitz protease inhibitors (10,11).
The precise protein interactions initiating and regulating formation of the unique collagen VI microfibril supramolecular assemblies are not known, but elegant structural studies have identified the organization of assembly intermediates, providing a model for a hierarchical assembly process unlike that of any other members of the collagen family (12,13). Following heterotrimeric assembly of the ␣1(VI), ␣2(VI), and ␣3(VI) chains, these collagen VI monomers form higher order structures intracellularly by aligning in an anti-parallel manner with a 30-nm stagger, first to form dimers and then tetramers by lateral association of dimers. After secretion, the tetramers associate end-to-end to form microfibrils with a distinctive 100-nm periodicity, comprising beaded globular domains separated by short triple helical regions (12,13). However, the critically important questions of what protein structures and interactions drive these complex and specific associations during collagen VI assembly have not been addressed in detail.
It is likely that numerous motifs within the collagen VI ␣-chains contribute to the final microfibril structure and to the development of its heterotypic interactions with other matrix components. The type A subdomains present in all three collagen VI subunits, and found in a number of other extracellular proteins including integrins, matrilins, complement components, and collagens VII, XII, and XIV (14 -16), are likely to play an important role, as they have been shown to be commonly involved in specific molecular interactions (14). In particular, it seems probable that the N terminus of the ␣3(VI) chain may be the crucial domain involved in tetramer-tetramer association because its multiple type A modules would be exposed at both ends of the tetramers during microfibril formation.
To examine systematically the role of the ␣3(VI) N-terminal type A subdomains, N9 to N1, in collagen VI intracellular tetramer assembly and extracellular microfibril formation ␣3(VI) cDNA constructs with sequential deletions of the Nterminal type A modules were stably transfected into SaOS-2 cells, producing a range of cell lines that express ␣3(VI) chains with N-terminal domains containing modules N9-N1, N6-N1, N5-N1, N4-N1, N3-N1, or N1, as well as the complete triple helix and C-terminal globular domain. Our data show that ␣3(VI) subdomains N2 to N9 are not required for intracellular heterotrimer assembly or for the formation of disulfide-bonded dimers and tetramers. Ultrastructural examination of the secreted collagen VI demonstrated that tetramers were able to associate end-to-end to form microfibrils in cells expressing ␣3(VI) chains containing the N-terminal N9-N1, N6-N1, and N5-N1 modules. Significantly, however, the efficiency of tetramer-tetramer association was reduced in cells expressing ␣3(VI) chains with N-terminal domains lacking the N5 module. These data provide the first evidence that an ␣3(VI) N-terminal subdomain (N5) plays a critical role in the interactions between collagen VI tetramers leading to the formation of the microfibrillar network.

EXPERIMENTAL PROCEDURES
Production of ␣3(VI) cDNA Expression Constructs-All constructs were cloned in the mammalian expression vector pRc/CMV (Invitrogen), which contains the neomycin phosphotransferase gene conferring resistance to G418. Production and stable expression of the ␣3(VI) N6-C5 construct, encoding the BM40 signal sequence and the ␣3(VI) protein domains N6 to C5 (Fig. 1), have been described previously (8). Constructs encoding ␣3(VI) domains N5-C5, N4-C5, and N1-C5 were generated by PCR using the ␣3(VI) N6-C5 cDNA as a template. Forward primers were designed to anneal at the 5Ј end of subdomains N5, N4, and N1, beginning at nucleotides 2710 (domains N9-C5, ATG at base 256 (10)), 3317, and 5131, respectively, and each contained an NheI site at the 5Ј end to allow cloning of the PCR product into the NheI site at the 3Ј end of the BM40 signal sequence of pRc/CMV ␣3(VI) N6-C5 (8). The downstream primer in each reaction corresponded to nucleotides 5748 -5771 near the 5Ј end of the triple helix and included the SacII site at base 5757. PCR was performed using the proof-reading polymerase Pfu (Stratagene). The resultant fragments were digested with NheI and SacII and used to replace the corresponding fragment of the ␣3(VI) N6-C5 construct.
To facilitate production of the ␣3(VI) N9-C5 construct, an NheI site within the N9 subdomain cDNA was deleted using strand overlap extension PCR (18) to introduce a silent T to C substitution at position 747. Primers A and B were used to amplify bases 331-755 using the cDNA clone FO19 (10) as a template, and primers C and D to amplify bases 735-2097. Primers B and C overlapped, and both contained the single nucleotide substitution. To allow cloning, primer A contained a NheI site at the 5Ј end. These amplification products were gel-purified, and 100 ng of each used was as the template in a second round of overlapping PCR with primers A and D. The resulting PCR product was digested with NheI and cloned into the NheI site of ␣3(VI) N6-C5 to generate the ␣3(VI) N9-C5 construct.
To ensure that no errors had been introduced during PCR and cloning, all constructs were transcribed and translated in vitro (TNT ® , Promega), and the regions generated by PCR were sequenced (Ampli-Cycle, PerkinElmer Life Sciences) (data not shown).
Negative Staining Electron Microscopy-Confluent SaOS-2 cultures in 80-cm 2 flasks were incubated for 18 h in 10 ml of serum-free medium containing 0.25 mM sodium ascorbate. The medium was collected, the protease inhibitors described earlier were added, and sodium azide was included at a final concentration of 0.1% (w/v). To examine collagen VI that had been deposited into the extracellular matrix, confluent SaOS-2 cells were grown in the continuous presence of 0.25 mM sodium ascorbate for 15 days. Cell layers were washed with 50 mM Tris/HCl, pH 7.4, containing 0.15 M NaCl, scraped into 5 ml of this buffer, and then mildly disrupted using a Teflon homogenizer. Culture medium and homogenized matrix samples were clarified by centrifugation, adsorbed onto carbon-coated grids for 1 min, washed with water, and stained with 0.75% uranyl formate. The grids were rendered hydrophilic by glow discharge in air, and samples were observed in a Jeol 1200 EX electron microscope operated at 60 kV accelerating voltage.
Indirect Immunofluorescence-SaOS-2 cells were grown to confluence in 8-well chamber glass slides (Nunc) and then supplemented daily for 15 days with 0.25 mM sodium ascorbate. Cell layers were washed with PBS, fixed with 3.7% (v/v) formaldehyde at room temperature for 10 min and then air-dried. Slides were preincubated with 5% (v/v) fetal calf serum in PBS for 1 h at room temperature and then incubated with either the collagen VI antibody 3C4 in PBS or PBS alone for 1 h at room temperature. Bound antibody was detected using fluorescein isothiocyanate-conjugated sheep anti-mouse Ig (Silenus). Slides were mounted in FluorSave reagent (Calbiochem) and then viewed and photographed using a Zeiss fluorescence microscope.

Collagen VI Intracellular Assembly in Transfected SaOS-2
Cells-To examine the contribution of the ␣3(VI) N-terminal globular subdomains to collagen VI intracellular assembly, microfibril formation, and matrix deposition, we stably transfected SaOS-2 cells, which express ␣1(VI) and ␣2(VI) mRNA and protein but no ␣3(VI) mRNA (8), with six ␣3(VI) cDNA constructs from which the N-terminal globular domains N9 to N1 were progressively deleted. As a result of alternative splicing, more than 95% of the ␣3(VI) mRNAs found in cells and tissues lack the N10 subdomain (11); therefore, the ␣3(VI) N9-C5 construct was used as the "full-length" control in this study. Individual cell clones transfected with either the control construct (␣3(VI) N9-C5) or the shorter constructs (␣3(VI) N6-C5, N5-C5, N4-C5, N3-C5, and N1-C5) ( Fig. 1) were selected in medium containing G418 and screened for expression of ␣3(VI) mRNA by Northern blotting (data not shown). Cell lines were metabolically labeled with [ 35 S]methionine and cell layer and medium fractions immunoprecipitated with collagen VI antibodies. The immunoprecipitated material was electrophoresed on 5% polyacrylamide gels under reducing conditions to visualize the individual collagen VI chains. All transfected cell lines expressed ␣3(VI) protein that migrated with an apparent molecular mass in good agreement with the predicted size of each of the shorter chains (Fig. 2) when compared with the control ␣3(VI) N9-C5 (Fig. 2, lanes 11 and 12). Because the collagen VI antibodies used for immunoprecipitation were specific for the ␣3(VI) chain (N9, N6, N5, N4, and N3 constructs), or assemblies of all three chains (N1 construct), co-immunoprecipitation of ␣1(VI) and ␣2(VI) chains in all of the cell lines indicated that the shortened ␣3(VI) chains had formed collagen VI heterotrimers that were able to be secreted. Collagen VI found in the medium had a slower electrophoretic mobility compared with that isolated from the cell layer fraction. This is a consistent finding in both transfected SaOS-2 cells and primary human fibroblasts (5,6,8), and although the reason for this size difference is not known, it may reflect differences in glycosylation between secreted collagen VI and the more recently synthesized intracellular collagen VI.
Secretion of all three collagen VI chains by each of the transfected cell lines provided indirect evidence that heterotrimers containing the shortened ␣3(VI) chains were able to assemble further into tetramers, the normal secreted form of collagen VI (22). Tetramer formation was demonstrated di-rectly by composite agarose/acrylamide gel electrophoresis under nonreducing conditions (Fig. 3). The predominant form of collagen VI secreted by cells expressing control ␣3(VI) N9-C5 chains migrated with a molecular mass of approximately 2000 kDa (Fig. 3, lane 12) as expected for collagen VI tetramers. Comparable results were seen in cultures expressing each of the shortened ␣3(VI) chains (Fig. 3, lanes 1-10), clearly demonstrating that the ␣3(VI) N-terminal subdomains N9-N2 are not required for collagen VI monomer, dimer, and tetramer formation and secretion.
To formally exclude the possibility that the efficiency of tetramer aggregation was dependent on the concentration of tetramers in the medium, the levels of ␣3(VI) mRNA and protein in an ␣3(VI) N5-C5 cell line (able to form microfibrils) FIG. 1. Structure of the collagen VI subunits and ␣3(VI) expression constructs. A, diagram of collagen ␣1(VI), ␣2(VI), and ␣3(VI) chains showing the triple helix and numbered C-and N-terminal subdomains. The shaded type A subdomains, N10, N9, and N7 in ␣3(VI) and C2 in ␣2(VI), undergo alternative splicing. B, diagram of the ␣3(VI) cDNA expression constructs. Deletions were made from the N terminus leaving the triple helix and C-terminal domains intact. and two ␣3(VI) N3-C5 cell lines (not able to form microfibrils) were examined. Northern blot analysis demonstrated that the ␣3(VI) N5-C5 cell line (clone 28) contained much less ␣3(VI) mRNA than either of the two ␣3(VI) N3-C5 cell lines (clones 19 and 17) (Fig. 6A). The amount of collagen VI protein produced by these clones was consistent with the level of expression of  ␣3(VI) mRNA. Small amounts of collagen VI protein were present in the cell and medium fractions of the ␣3(VI) N5-C5 line (Fig. 6B, lanes 3 and 4), and substantially larger amounts were produced by the two ␣3(VI) N3-C5 clones (lanes 5-8). The faint bands in lanes 3 and 4 were clearly identified as collagen VI following a longer exposure. Because mRNA and protein expression is higher in the two ␣3(VI) N3-C5 cell lines than in the ␣3(VI) N5-C5 line, and end-to-end tetramer assembly is more efficient in the ␣3(VI) N5-C5 line than in either of the two ␣3(VI) N3-C5 cell lines (Fig. 5), it is clear that the efficiency of tetramer aggregation is not related to the level of collagen VI synthesis and the concentration of tetramers available for microfibril formation.
To determine which cell lines were able to incorporate collagen VI tetramers into an extracellular matrix, transfected SaOS-2 cells were grown for 15 days post-confluence in the presence of sodium ascorbate, stained with collagen VI antibodies, and examined by fluorescence microscopy. A collagen VI matrix was present in cells expressing ␣3(VI) N9-C5, N6-C5, and N5-C5 (Fig. 7, A-C), but in contrast, no collagen VI matrix was seen by immunofluorescence in cells expressing ␣3(VI) N4-C5 (Fig. 7D), N3-C5, and N1-C5, where only intracellular staining was present. Transmission electron microscopy of the accumulated extracellular matrix revealed an extensive branched microfibrillar network in cells expressing ␣3(VI) N9-C5 (Fig. 8B), similar to the highly branched open meshwork formed by collagen VI in skin (23,24). Branching points commonly connected 4 -5 tetramers in a star-like arrangement (Fig. 8B). This extensive microfibrillar network was absent from the matrix of cells expressing ␣3(VI) N1-C5 (Fig. 8A); the microfibrils seen in Fig. 8A were extremely rare and were present in only a small number of fields. To confirm that these were collagen VI microfibrils, the extracellular matrices were mildly homogenized in a neutral salt buffer and the solubilized proteins examined by negative staining electron microscopy. The branching microfibrils extracted from the matrix of cells expressing ␣3(VI) N9-C5 were clearly composed of collagen VI, with the characteristic globular beads at 100-nm intervals (Fig.  8C). Collagen VI microfibrils were not present in the matrix extracts of cells expressing ␣3(VI) N1-C5 (not shown). These findings, which are in agreement with the quantitative electron microscopy data, demonstrated that collagen VI tetramers containing ␣3(VI) chains with only the N4-N1, N3-N1, or N1 modules have lost the capacity to interact end-to-end to form collagen VI microfibrils. The minimum requirement for N5-N1 modules identifies the N5 subdomain as critical for collagen extracellular matrix formation.

DISCUSSION
In this study, we investigated the contribution of ␣3(VI) N-terminal vWF type A-like subdomains N2 to N9 to collagen VI intracellular assembly, secretion, and microfibril formation. We show that the ␣3(VI) subdomains N2 to N9 are not necessary for collagen VI intracellular triple helical heterotrimer assembly, the subsequent formation of disulfide-bonded dimers and tetramers, and secretion of the collagen VI tetramers from the cell. However, the N-terminal ␣3(VI) type A modules were found to play a critical extracellular role in the end-to-end association of tetramers to form collagen VI microfibrils. Cells that expressed ␣3(VI) chains containing the N5 subdomain, ␣3(VI) N5-C5, ␣3(VI) N6-C5, and ␣3(VI) N9-C5, had the ability to form collagen VI microfibrils and deposit a collagen VI ma-  #19 and #17, lanes 3 and 4) was separated on a denaturing agarose gel, transferred to nitrocellulose, and hybridized simultaneously with 32  trix. In contrast, cells that expressed ␣3(VI) chains without the N5 subdomain, ␣3(VI) N4-C5, ␣3(VI) N3-C5, and ␣3(VI) N1-C5, were severely compromised in their ability to form end-to-end tetramer assemblies, and failed to deposit a collagen VI matrix. These results indicate that the ␣3(VI) N5 module is important for stable extracellular end-to-end association of tetramers to form collagen VI microfibrils.
The mechanism of collagen VI heterotrimer association has not been elucidated, but our studies clearly exclude a role for ␣3(VI) N-terminal subdomains distal to N1 in intracellular chain selection and assembly. Our previous studies demonstrating that two collagen VI triple-helix mutations, a Bethlem myopathy exon 14-skipping mutation in COL6A1 and an engineered 202-base pair deletion in the ␣3(VI) helix, have unimpeded heterotrimer formation (5), indicated that the N-terminal region of the collagen VI triple helix is not important for initial chain assembly. These data suggest that the information for heterotrimer formation is present in the N-terminal N1 subdomain, the C-terminal region of the helix, or the C-terminal globular domains of ␣3(VI) and in the other two subunits, ␣1(VI) and ␣2(VI). By analogy with fibrillar collagen trimer assembly (25,26) and the molecular constraints of the direction of triple-helix propagation (27), it seems likely that the Cterminal globular domains play a decisive role in this first event of assembly.
The subsequent complex hierarchical process of collagen VI intracellular assembly, the formation of dimers and tetramers, does critically involve the participation of collagen VI triplehelix sequences. Ultrastructural and binding studies predict that a specific interaction between the C-terminal globular domain of one monomer and the triple helix of the adjacent overlapping antiparallel monomer is involved in dimer formation, an interaction that is stabilized by disulfide bonding (12,13,28). Tetramers are thought to be stabilized by disulfide bonds between the N-terminal ends of the triple helix of the dimers (12,13). Furthermore, collagen VI heterotrimers, containing deletions that disturb the structure of the helix, are unable to assemble further into disulfide-bonded dimers and tetramers, and molecules containing the mutant chain are not secreted (5).
The structural models proposed for the supramolecular organization of collagen VI tetramers into microfibrils (9,12,13) provide us with a basis for understanding the importance of the type A domains in microfibril formation. The lateral association of dimers into tetramers generates a structure in which the type A domains from the N termini of all three ␣-chains are exposed at both ends of the tetramer. The model for the junctional complex further predicts that N-terminal domains of adjacent tetramers overlap in this region, such that they are available to interact with each other and with the C-domains or the helical domain of the adjacent tetramer, providing multiple potential sites for the noncovalent interactions that stabilize microfibrils (9,12,13). Although the N termini in this junctional complex contain multiple type A subdomains from the ␣1(VI), ␣2(VI), and ␣3(VI) chains, all of which could potentially function in these interactions, our data demonstrates that the ␣3(VI) N5 subdomain is critical for microfibril formation. The ␣3(VI) chain is expressed as a number of splice variants in vivo and in cultured cells in which specific N-terminal modules are deleted from the protein structure (10,11,29). It is of interest to note that the N5 module is present in all of these splice variants, 2 consistent with a crucial role for N5 in the formation of microfibril assembly competent collagen VI. Although it is clear that the N5 module is important for extracellular tetramer association, other subdomains C-terminal to N5, and/or those within ␣1(VI) or ␣2(VI) and/or the N-terminal region of the triple helix, could contribute to microfibril formation or stability by providing an interaction partner for N5. A minor role for other subdomains is also suggested by the small amount of end-to-end association that occurs between tetramers lacking N5.
There are several other examples of interactions between A domains, either within the same molecule or between different extracellular matrix components, suggesting that these protein modules may play an important role in matrix architecture and assembly. Matrilin 1 contains two vWF type A-like domains, which have been shown to be involved in the assembly of matrilin oligomers and in the formation of matrilin-1 filamentous networks, suggesting that interactions between matrilin-1 type A domains can occur (30). Interaction between type A domains from different matrix proteins have been demonstrated for the A1 and A3 domains of vWF, both of which associate with the globular regions of chicken collagen VI microfibrils to promote the adhesion of platelets to collagen VI at high shear forces (31).
Further evidence for the role of ␣3(VI) N-terminal subdomains in the formation of functional collagen VI comes from studies of a family with Bethlem myopathy in which the mutation is a glycine to glutamate change at amino acid 167 of the 2 Marie Dziadek, personal communication. ␣3(VI) N2 subdomain (4). Normal levels of all three chains were detected in the medium of fibroblasts from affected individuals, suggesting that monomer and tetramer formation was not impaired (4). This result is consistent with our data showing that the removal of the entire N2 subdomain, as with the ␣3(VI) N1-C5 construct, does not affect monomer and tetramer formation and secretion. The molecular mechanism of how this N2 mutation results in a clinical phenotype is not known, but it could perturb collagen VI function by structurally interfering with end-to-end association of tetramers or, possibly, by modifying interactions of the mutant collagen VI with other matrix components.
Our emerging understanding of collagen VI assembly is that the ␣3(VI) N-terminal domain plays at least two broad roles, in microfibril assembly and in collagen VI-matrix interactions. The N-terminal domain of the ␣3(VI) chain interacts with vWF and heparin in vitro. The A1 subdomain of vWF interacts with the chicken ␣3(VI) N8 (31), and heparin binding has been assigned to ␣3(VI) subdomains N3, N6, and N9 suggesting the potential for binding to heparan sulfate proteoglycans (32). It seems reasonable to speculate that the specific role of each A-type subdomain may be related to its location within the three-dimensional structure of the globular beads of the tetramer junctional complex. Our data would predict that the N5 domain must be positioned so that it is exposed on the interacting surface in intimate contact with the N-and/or C-terminal domains or the triple helix of the collagen VI in the adjacent tetramer. Other subdomains may be positioned so as to be available to participate in interactions with a variety of extracellular matrix proteins.