Type VI Collagen Anchors Endothelial Basement Membranes by Interacting with Type IV Collagen*

Type VI collagen filaments are found associated with interstitial collagen fibers, around cells, and in contact with endothelial basement membranes. To identify type VI collagen binding proteins, the amino-terminal domains of the α1(VI) and α2(VI) chains and a part of the carboxyl-terminal domain of the α3(VI) chain were used as bait in a yeast two-hybrid system to screen a human placenta library. Eight persistently positive clones were identified, two coding the known matrix proteins fibronectin and basement membrane type IV collagen and the rest coding new proteins. The amino-terminal domain of α1(VI) was shown to interact with the carboxyl-terminal globular domain of type IV collagen. The specificity of this interaction was further studied using the yeast two-hybrid system in a one-on-one format and confirmed by using isolated protein domains in immunoprecipitation, affinity blots, and enzyme-linked immunosorbent assay-based binding studies. Co-distribution of type VI and type IV collagens in human muscle was demonstrated using double labeling immunofluorescent microscopy and immunoelectron microscopy. The strong interaction of type VI collagen filaments with basement membrane collagen provided a possible molecular pathogenesis for the heritable disorder Bethlem myopathy.

Type VI collagen filaments are ubiquitous. They are present in all connective tissues that contain type I and type III collagen fibers and in cartilage, a tissue that contains predominantly type II collagen. The major functions that have been suggested for type VI collagen filamentous networks are as a substrate for cell attachment and as an anchoring meshwork that connects collagen fibers, nerves, and blood vessels to the surrounding matrix (1,2). This implies that not only is there an interaction, either direct or indirect, with the type I/III collagen fibers but also that there is an interaction with components of endothelial basement membranes.
Matrix components that have been shown to interact with type VI collagen in vitro include proteoglycans, collagens, hyaluronan, heparin, and integrins.
Proteoglycans appear to be particularly important, since cell surface-, basement membrane-, and collagen fibril-associated proteoglycans have all been reported to bind to various forms of type VI collagen. Decorin is a small dermatan sulfate proteoglycan that binds to fibrillar collagens (3). It was shown that the leucine-rich module of the core protein bound to type VI and that the binding could be inhibited by the core proteins of fibromodulin and biglycan (4). The cell surface-associated membrane chondroitin sulfate proteoglycan NG2 was originally detected in cells from the rodent central nervous system but subsequently was also detected in blood vessels and cartilaginous structures of the head, neck, and spine. It interacts via its core protein with type VI collagen and is thought to provide machinery for transmembrane signaling (5). Since ␣1␤1 and ␣2␤1 integrins also bind type VI collagen (6,7), the cell signaling potential of this molecule would appear to be significant. The interaction with the basement membrane proteoglycan perlecan appeared to be via the type VI collagen triple helix, since pepsin-solubilized type VI collagen, but not recombinant ␣1(VI) and ␣2(VI) globular domains, showed binding to perlecan (8). However, unlike decorin and NG2, perlecan interacts via its heparan sulfate side chains.
Recombinant segments of the amino-terminal globular domain of the ␣3 chain, which contain a series of von Willebrand factor A domains, were used by two groups. One group reported no binding to types I, III, and IV collagens or fibronectin (9), while the second reported specific binding to fibrillar type I collagen (2). Other experiments using pepsin-solubilized type VI collagen, which only contains a small remnant of the globular domains, showed binding to fibronectin (8), while others found no binding to fibronectin or type I and III collagens but found binding to type II (4) and XIV collagens (10). In vivo studies of GP140 (␣1(VI) ϩ ␣2(VI)) indicated that there was an interaction of type VI collagen with fibronectin in the pericellular matrix of fibroblasts (11,12).
Data concerning the interaction of type VI with hyaluronan are also inconsistent. Intact type VI isolated from bovine cartilage (13), fetal bovine skin (14), and recombinant ␣3 chain A domains (9) were reported to bind hyaluronan. Furthermore, Kielty and co-workers (15) presented evidence for hyaluronan being required as a template for the assembly of type VI collagen tetramers into filaments and for the stability of the filaments once they are formed (15). The results of Specks et al. (9) were later withdrawn (16), and their new data and our own data indicate that hyaluronan is not involved in the assembly or stability of type VI filaments (17).
The interaction spectrum of type VI collagen is clearly very complex. The results described here indicate a newly discovered direct interaction of intact native type VI collagen with type IV collagen, a major component of basement membranes. This interaction is significant because it is the first type VI collagen protein-protein interaction with a basement membrane component that has been characterized, and it provides an plausible explanation for the Bethlem myopathy phenotype, a genetic disease caused by mutations in type VI collagen (18).

Materials
Yeast Matchmaker two-hybrid kit, multiple-tissue Northern blots, and a human Matchmaker placenta cDNA library were purchased from CLONTECH Laboratories, Inc. Superscript preamplification system for first strand cDNA synthesis, polymerase chain reaction (PCR) 1 kits, T4DNA ligase, several restriction enzymes including EcoRI, SalI, and PstI, and 123-base pair standards were obtained from Life Technologies, Inc. A random primed DNA labeling kit was from Boehringer Mannheim. QIAprep spin plasmid kits and Wizard PCR preparation kits were obtained from Qiagen and Promega, respectively. Streptavidin-linked agarose, NHS-LC-biotin, and immunopure biotin LC-hydrazide were purchased from Pierce. Protein G-Sepharose 4B and Sephacryl S400 were from Pharmacia Biotech Inc. Avidin-linked horseradish peroxidase was from Bio-Rad. Anti-mouse or anti-rabbit IgG (whole molecule) fluorescein isothiocyanate conjugate, anti-mouse IgG (whole molecule) Texas Red isothiocyanate conjugate, extravidin alkaline phosphatase conjugate, and Sigma 104 phosphatase substrate were purchased from Sigma. Anti-rabbit IgG Texas Red conjugate was from Vector Laboratories. Polyclonal and monoclonal (M3F7) type IV collagen antibodies were from Southern Biotechnology Associates, Inc. and the Developmental Studies Hybridoma Bank (University of Iowa), respectively. Goat anti-mouse 5-nm colloidal gold conjugate was from Amersham Corp. Polyclonal antibodies against ␣1(IV) NC1 and ␣2(IV) NC1 were gifts from Klaus Kü hn (Martinsried, Munich) (19). The type VI collagen monoclonal antibodies N3-VI-2, C1-VI-1, and C12-VI-2 have been described previously (7). Type VI collagen mouse monoclonal antibody 5C6 (1) was a kind gift from Eva Engvall (La Jolla Cancer Research Foundation, CA).
Construction of Recombinant Plasmids-Recombinant plasmid constructs for the "bait" consisting of the inserts described above cloned into the vector pGBT9 supplied with the Matchmaker two-hybrid system (CLONTECH) were prepared using standard methods (20). Restriction enzyme pairs used to construct the recombinant plasmids were EcoRI and PstI for ␣1(VI), EcoRI and SalI for ␣2(VI), and SmaI and SalI for ␣3(VI).
Yeast Two-hybrid Library Screening-The Matchmaker two-hybrid system (CLONTECH) was used according to the supplied protocol, with the amino-terminal domains of the ␣1(VI) and ␣2(VI) chains, and the carboxyl-terminal domain of the ␣3(VI) chain fused to the GAL4 binding domain as bait, to screen a human placenta cDNA library fused to the GAL4 activation domain. Vector pGBT9 containing the bait and the amplified library in vector pGAD10 were sequentially transformed into yeast strain HF7C and grown in synthetic media without tryptophan, leucine, and histidine. Approximately 1.8 ϫ 10 6 clones, one library equivalent, were screened with each of the bait constructs. Colonies that grew successfully in the selective media were further screened for reporter gene ␤-galactosidase expression. The ␤-galactosidase assays were done using replicas of colonies on filter papers. Colonies that were positive for ␤-galactosidase activity in less than 30 h. were selected for further analysis. The positive colonies were restreaked several times on selective media, and individual isolated colonies were rescreened for ␤-galactosidase activity, eliminating most of the false positives that initially passed through the dual selection. Persistent positive clones were grown on selective media that segregates the plasmids, thereby isolating the activation domain ("target") plasmid. Additional selection against false positives was done by performing ␤-galactosidase assays on yeast clones that only contain the target plasmid (similar assays were done with the bait constructs), thus eliminating any clones that have activation function in the absence of an interacting partner. Also, the bait and positive target vectors were switched so that the bait was in pGAD10 and the target in pGBT9, and these constructs were again subjected to the dual selection for interactions as an additional check for false positives. The target and bait constructs were also assayed for activation function in the presence of the partner vector without an insert. To avoid unnecessarily analyzing duplicate positive clones, they were further characterized by determining insert size using PCR amplification of the pGAD10 inserts with vector-specific primers (CLON-TECH). The PCR products were digested with restriction enzymes, and the restriction digestion patterns were compared. To determine the identity of the interacting clones, the PCR-amplified pGAD10 inserts were sequenced using standard techniques. The identities of the inserts were determined by comparing the cDNA sequences with GenBank TM , release 97, using the computer programs of the Wisconsin Package, version 8 (Genetics Computer Group, Madison, WI).
Yeast Two-hybrid One-on-one Interactions-To demonstrate the specificity of the ␣1(VI)-␣1(IV) interaction, one-on-one two-hybrid analyses were done to compare the interaction potential of ␣1(VI) versus ␣2(VI) with ␣1(IV). The vectors containing the yeast GAL4 binding domain (pGBT9), with or without the insert for the amino-terminal domain of the ␣1(VI) and ␣2(VI) chains, were individually cotransformed into yeast HF7C with a vector containing the GAL4 activation domain (pGAD424) fused with the carboxyl-terminal domain of ␣1(IV). Positive interactions were detected and analyzed as described above.
Northern Blot Analyses-A Northern blot with 2 g of poly(A) ϩ RNA from eight adult human tissues (CLONTECH) and an additional Northern blot with 2 g of poly(A) ϩ RNA from human skin fibroblasts were hybridized with probes made from purified PCR products from the activation domain inserts of positive clones. For the fibroblast Northern blot, human skin fibroblast poly(A) ϩ RNA was isolated and blotted using standard techniques (20). Probes were labeled with [␣-32 P]dCTP, and the resulting high specific activity probes (Ͼ1 ϫ 10 9 dpm/g) were hybridized to the blots in a neutral hybridization solution at about 1 ϫ 10 6 dpm/ml at 42°C for 16 -20 h. Following hybridization, the blots were washed under stringent conditions to remove unbound probe (0.1 ϫ SSC, 0.1% SDS, 68°C). The resulting signals were visualized by standard autoradiography.
Protein Extraction and Purification-Type VI collagen tetramers and globular domains were prepared as described previously (17). Type IV collagen native hexameric globular domains were isolated from human amniotic membranes using a method based upon a previously described procedure (21). Placental membranes were collected from approximately 60 full-term placentas. The amnion was stripped from the chorion, rinsed briefly with water, and stored at 4°C in water containing 2 mM N-ethylmaleimide, 1 mM phenylmethylsulfonyl fluoride, 5 mM benzamidine until used. The amnion was homogenized into ice-cold water containing the above inhibitors and centrifuged, and the tissue pellet was washed once more in 6 liters of the same solution. Finally, the tissue was washed in 6 liters of cold 10% NaCl solution containing the same inhibitors. One kilogram of washed tissue was suspended into 3 liters of 0.1 M Tris-HCl buffer, pH 7.5, containing the inhibitors and 20 mM CaCl 2 and 0.5 M NaCl. Bacterial collagenase (12,500 units CLSPA, Worthington) was added to the stirred tissue suspension, which was incubated at 4°C for 24 h and then at room temperature for a further 24 h. The digest was centrifuged, and the supernatant was fractionally precipitated at NaCl concentrations of 1. Biotin-labeled type VI collagen tetramers and globular domains were used as soluble ligands for affinity blots (17). Type IV collagen samples were separated on 8 -12% SDS-PAGE gels (23) and electrotransferred onto polyvinylidene difluoride membranes. The blotted membranes were incubated with 4% BSA or 1% nonfat dried milk in PBS for 1 h to block any nonspecific binding and then with biotinylated ligands for 3 h followed by a 2-h incubation with avidin-horseradish peroxidase. The membranes were developed with color reagent as for standard Western blots. Control membranes were developed only as Western blots using chain-specific antibodies N3-VI-2 and C1-VI-1 to type IV collagen.
Immunoprecipitation-The ligands used in each experiment were type IV collagen and type VI collagen in a bacterial collagenase digest of human amnion and purified NC1 domain of type IV collagen with purified type VI collagen globular domains. Type VI collagen polyclonal antibody 4104; type VI monoclonal antibody mixture containing N3-VI-2, C1-VI-1, and C12-VI-2; or type IV collagen polyclonal antibody (Southern Biotechnology) was used to precipitate the type VI-type IV complex. Preimmune serum was used to control for nonspecific precipitation. Mixtures of the proteins were incubated in PBS at room temperature for 2 h, and then antibody was added with gentle rotation for 2 h or at 4°C overnight. PBS-washed protein G-Sepharose (1:3, v/v) was added and mixed for another 2 h at room temperature. The antigenantibody-protein G-Sepharose complex was collected by centrifugation and washed with PBS, 0.05% Tween 20 three times (5 min each). The complex was eluted from the protein G with 0.5 M acetic acid. The identity of the interacting ligands was determined by Western blots analysis.
Binding Assay-Five micrograms of pepsin-solubilized and purified type I, II, V, and XI collagens, purified intact type VI collagen tetramers, type IV collagen NC1 domain, fibronectin, and BSA were coated onto microtiter plates at 4°C overnight. The plate was washed with PBS containing 0.05% Tween 20, incubated with 1% BSA/PBS for 30 min followed by a 3-h incubation at room temperature with varying amounts of biotinylated intact type VI collagen tetramers (0 -5 g). Bound type VI collagen was detected by further incubation with extravidin-alkaline phosphatase conjugate (100,000 dilution) for 2 h, three PBS/Tween 20 washes, and a 30-min incubation with Sigma 104 substrate. The absorbance was measured at 405 nm. All analyses were done in duplicate, and the values were averaged.
Indirect Immunofluorescent Staining-Human skeletal muscle was trimmed and quickly frozen in hexane with liquid nitrogen. The frozen muscle was sectioned in an ultracryotome, and 5-m sections were collected on glass slides, treated with cold (Ϫ20°C) acetone for 3 min, and allowed to air dry at room temperature. Double immunofluorescence staining was performed at room temperature on unfixed tissue. Tissue was briefly washed with PBS (5 min, twice) incubated with normal rabbit serum followed by PBS washes (3 ϫ 5 min), and incubated for 3 h in mixtures of type IV collagen polyclonal antibody and type VI collagen monoclonal antibody VI-1. After PBS washes, they were incubated with mixtures of anti-mouse IgG fluorescein isothiocyanate (50 ϫ dilution) and anti-rabbit IgG Texas Red (500 ϫ dilution) conjugates for 30 min followed by PBS washes. Sections were covered with 25% glycerol in PBS buffer and viewed under a Zeiss Axiophot fluorescent microscope.
Immunoelectron Microscopy-The method used in this study has been described previously (24). Briefly, en bloc immunolocalization of type VI collagen in human skeletal muscle from a 16-year-old individual was accomplished by immersing the tissue overnight at 4°C in monoclonal antibody 5C6 diluted 1:5 in PBS, rinsing extensively in PBS, and then incubating overnight at 4°C in goat anti-mouse 5-nm colloidal gold conjugate. Immunolabeled tissue was then rinsed, fixed, dehydrated, and embedded in Spurrs epoxy prior to ultramicrotomy and examination using a Philips 410 LS transmission electron microscope.

RESULTS
Screening with the amino-terminal domain of the ␣1(VI) chain identified four clones from the placenta library that coded protein domains that potentially interact with type VI collagen (Table I) Table I along with the approximate mRNA sizes. The tissue distributions are quite distinctive, ranging from 3C-30.1, with an apparently placentaspecific distribution, to 1N-5, which was in all tissues tested. All inserts except 3C-30.1 hybridized to fibroblast mRNA.
An interaction between type VI and type IV collagens is likely to be physiologically significant, since type VI filaments and endothelial basement membranes are in close association in tissues. Therefore, this interaction was confirmed using isolated protein domains. The globular NC1 domain of type IV collagen was isolated from human amnion and purified by molecular sieve chromatography under native conditions (Fig.  1). The indicated fraction contained NC1, and the SDS-PAGE gel showed its purity and presence of both ␣1 and ␣2 chain dimers and monomers as previously reported (21).

TABLE I
Characteristics of the clone inserts isolated from a human placenta library using the yeast two-hybrid system In the clone names, 1N, 2N, and 3C denote that the ␣1(VI) amino-terminal, ␣2(VI) amino-terminal, and ␣3(VI) carboxyl-terminal domains, respectively, were used in the "bait" vector. H, B, Pl, Lu, Li, SKM, K, Pa, and Fib denote heart, brain, placenta, lung, liver, skeletal muscle, kidney, pancreas, and fibroblast, respectively. Parenthesis indicate tissues with the strongest signal. ND, not determined. For the interaction to be physiologically significant, the native amino-terminal globular domain of type VI collagen must have an affinity for the native carboxyl-terminal domain of type IV collagen. This was demonstrated by specifically immunoprecipitating a type VI-type IV complex from a bacterial collagenase digest of human amnion with either a polyclonal antibody specific to type IV collagen or a monoclonal antibody specific for the amino-terminal domain of the ␣3 chain of type VI collagen. In this digest, due to incomplete digestion by bacterial collagenase, both type VI tetramers and filaments are present together with type IV collagen in the form of a mixture of 7s domains (28) and NC1 domains with various lengths of the collagen triple helix still attached to them. The immunoprecipitate formed with type IV polyclonal antibodies contained type IV collagen and type VI collagen ( Fig. 2A). Type VI collagen in lane 3, blotted with a mixture of type VI collagen monoclonal antibodies, appears as a well defined band corresponding to the ␣1 and ␣2 chains and several light bands corresponding to the ␣3 chain. Type IV collagen in lane 1, blotted with a type IV collagen polyclonal antibody, appears as NC1 domains and larger chains due to incomplete removal of the triple helix. Lane 2 shows proteins that were nonspecifically precipitated by preimmune serum and is a negative control for lane 1. Note that only light staining of some of the bands corresponding to type IV or type VI chains are present. Repeating the immunoprecipitation with purified globular domains using a monoclonal antibody to type VI collagen reduced the nonspecific background and clearly demonstrated co-precipitation of both globular domains (Fig. 2B). This result also confirmed that the interaction is between the globular domains of the collagens as indicated by the two-hybrid interaction results.
Affinity blots using native biotinylated type VI globular domains as soluble ligand and type IV globular domain separated by SDS-PAGE showed that the interaction was specifically with the ␣1(IV) chain (Fig. 3). The biotinylated type VI collagen globular domains bound only to ␣1(IV) monomer and dimer chains.
To demonstrate the specificity of the yeast two-hybrid interaction, a one-on-one yeast two-hybrid interaction was used. The ␣1(IV) containing vector, when co-transfected into yeast with vectors containing either the ␣1(VI) or ␣2(VI) amino terminus, produced a positive response after 16 h only with the ␣1(VI)-containing vector (Fig. 4). The amino-terminal domain of the ␣2(VI) chain did not interact with ␣1(IV), although there  2. Coimmunoprecipitation of type VI and type IV collagens. A, a partially purified bacterial collagenase digest of human amnion was treated with a polyclonal antibody to type IV collagen followed by protein G-Sepharose. The immunoprecipitate was separated in three identical lanes on SDS-PAGE using a 7.5% gel and then electrotransferred to polyvinylidene difluoride. The membrane was developed as a Western blot with a polyclonal antibody to type IV collagen (lane 1), a preimmune serum (lane 2), and type VI collagen polyclonal antibodies (lane 3). Clearly, the immunoprecipitate contained both type IV and VI collagens. Arrowheads to the left identify type IV collagen-related bands, and arrowheads to the right show type VI collagen-related bands. B, a mixture of purified type VI and type IV globular domains were immunoprecipitated with a type VI collagen ␣3 chain aminoterminal domain-specific monoclonal antibody and analyzed as in A but using a 10% gel. The membrane was developed with a preimmune serum (lane 1), a polyclonal antibody against the type IV collagen NC1 domain (lane 2), and a type VI collagen ␣3 chain-specific monoclonal antibody (lane 3). Again the immunoprecipitate contained both type VI and type IV globular domains.

FIG. 3. Affinity blots of type IV collagen with biotinylated type VI collagen.
A preparation of type IV collagen globular domains was separated on an 8 -12% SDS-PAGE gradient gel in nonreducing conditions. The protein was electrotransferred to a polyvinylidene difluoride membrane, and the membrane was cut into three identical strips. is 31% amino acid sequence identity between the two type VI amino-terminal domains.
Since interactions of type VI collagen with types I and II collagens have been reported, comparative solid phase binding studies were undertaken to determine the relative affinities of native type VI tetramers for various substrates. Binding to the NC1 domain of type IV collagen was found to be the strongest and was therefore fixed at 100% to normalize the results. On this scale, fibronectin was approximately 75%, and type VI binding to itself was at 65% (Fig. 5). Type V and XI collagens were at the level of BSA that was used as a control for nonspecific binding. Type I and type II collagen were about 10% above background and therefore weak compared with type IV collagen, fibronectin, and type VI collagen interactions. Binding of NC1 to type VI collagen was concentration-dependent and saturable as shown in Fig. 6, with half-maximal binding at 0.2-0.3 gtypeVIcollagen.BindingtoBSAwaslowandnotconcentrationdependent. Binding to type I collagen was low and concentrationdependent but did not appear to reach saturation under the conditions used, indicating a low affinity for type VI collagen.
Mutations in type VI collagen give rise to Bethlem myopathy, whose major clinical feature is weakening of the muscles (29,30). Fig. 7 shows the general co-distribution of type VI and type IV in human muscle using double immunofluorescent labeling of a section with Texas Red and fluorescein isothiocyanatelabeled secondary antibodies. The endomysium around the muscle fibers is stained, and the two patterns are identical. The immunoelectron micrograph in Fig. 8 shows the close juxtaposition of gold-labeled type VI collagen filaments to the basement membrane. The filaments appear to form a continuous network connecting the basement membrane with the interstitial collagen fiber network.

DISCUSSION
The yeast two-hybrid system has not previously been used to screen for interactions of extracellular matrix connective tissue protein domains. It has, however, been used successfully to study the interactions of the extracellular domains of transmembrane proteins (31,32). Numerous questions have been raised as to whether two-hybrid systems are applicable to the analysis of interactions between extracellular matrix proteins. Concerns include questions about proper folding of proteins, disulfide bond formation, targeting of extracellular matrix components outside the nucleus, and the absence of important post-translational modifications. These concerns can be appro-priately addressed by first recognizing that, although the yeast two-hybrid system is unlikely to be universally applicable to the analysis of extracellular matrix interactions, it is a valid and useful tool for some applications as demonstrated here. It is unknown whether the protein-protein interaction actually occurs inside or outside of the nucleus (33). Consequently, it is possible that the initial interaction does not take place in the nuclear environment, and yeast have been shown to be quite favorable to folding and disulfide bond formation in recombinant proteins and domains (34). However, although secondary and tertiary structure frequently contribute to protein recognition and subsequent interaction, the basis for protein-protein interaction can also be embedded in the linear amino acid sequence of the interacting domains. This is critical to the applicability and success of many well accepted in vitro experimental procedures including binding studies using recombinant proteins or synthetic peptides, peptide libraries, Western blot analyses, and production of peptide antibodies that recognize native proteins. Judicious selection of bait domains maximizes the potential for success. This includes using protein domains that can fold independently, avoiding sequences that would target the protein outside the nucleus or that contain multiple cysteine residues. The use of libraries with small random primed inserts that are most likely to code single domains of matrix proteins is probably beneficial. It is also important to note that the time required to detect expression of ␤-galactosidase activity for extracellular matrix protein-protein interactions is usually longer than 12 h, which reflects the FIG. 4. Yeast two-hybrid assay showing the type VI collagen chain specificity for binding to type IV collagen. The plasmids containing the cDNA insert from type IV collagen ␣1 carboxyl-terminal domain and two unidentified inserts were cotransformed back into HF7c yeast with plasmids containing the ␣1(VI) amino-terminal domain, the ␣2(VI) amino-terminal domain, or no insert (negative control). Only cotransfected yeast containing the ␣1(VI) insert showed galactosidase expression, demonstrating the specificity of the interaction for the ␣1 chain.
FIG. 5. Relative affinities of native type VI collagen tetramers to immobilized type I, II, V, VI, and XI collagens, fibronectin, and NC1. The matrix molecules indicated were plated onto microtiter plates (5 g), and biotinylated native type VI collagen tetramers were added. After washing, the bound type VI collagen was detected with an avidin-conjugated phosphatase and color reagent. BSA was used to assess nonspecific binding. Values represent means Ϯ S.D. (bars) obtained from duplicate experiments.
FIG. 6. Binding curve for type VI collagen tetramers binding to NC1. Microtiter plates were coated with 5 g of NC1, type I collagen, and BSA, and varying concentrations of biotinylated type VI collagen tetramers (0 -5 g) were added and incubated for 3 h at room temperature. After washing, the amount of bound type VI collagen was detected with an avidin-conjugated phosphatase and color reagent. Values represent means Ϯ S.D. (bars) obtained from duplicate experiments. lower affinities of these proteins (35) compared with many intracellular protein-protein interactions (33). However, reaction times should be less than 30 h, since false positives increase dramatically with longer times.
If the method were not applicable to extracellular matrix proteins, either no interactions or large numbers of nonspecific interactions would be detected. For the bait domains used here, a small number of positive clones were identified, which included two known extracellular matrix components. Furthermore, the selectivity of the interaction was demonstrated by showing that only the ␣1(VI) and not the ␣2(VI) amino-terminal domain interacted with ␣1(IV)-NC1 despite their high sequence similarity. Most importantly, the interaction was confirmed using isolated protein domains. The results of this study clearly demonstrate that two-hybrid analysis can be a powerful tool for identifying interacting domains of extracellular matrix structural proteins.
Our initial observations on the distribution of type VI collagen in skin led us to the conclusion that there was a much closer association with endothelial basement membranes around blood vessels and nerves than with the epithelial basement membranes (1). We speculated that perhaps endothelial basement membranes were anchored to the surrounding matrix by type VI collagen in much the same way that type VII anchoring filaments attach the epithelial basement membranes to the underlying dermis (36). More recent studies using immunogold electron microscopy co-localized type VI collagen and type IV collagen on the endothelial side of the glomerular basement membrane (37) and in basement membranes of the placenta (38). Here we show, by immunofluorescent labeling of muscle, a general co-localization of type VI with type IV collagen and show, by using immunoelectron microscopy, an intimate association of type VI collagen with basement membranes surrounding muscle fibers. These results further support the proposed general interaction between type VI and endothelial basement membranes.
Characterization of the interaction between type VI and type IV collagens identifies an important new physical link between basement membranes and the surrounding matrix. In a previous study, recombinant ␣(VI) chains did not bind to type IV collagen isolated from the EHS mouse tumor (8,9). The discrepancy, however, is understandable given that the two-hybrid system is capable of detecting weak interactions that cannot be detected using solid phase binding assays (33,39,40). In addition, the strong binding to type IV collagen in the solid phase binding assays described here may be due to the presence of the globular domains of all three chains in a single native conformation in the type VI collagen used, as opposed to recombinant single domains used in other studies. Binding of intact type VI collagen to type I fibrillar collagen has been previously reported by Bonaldo et al. (2), and pepsin-solubilized type VI collagen to type II collagen has been reported by Bidanset et al. (4). However, a comparison of the relative affinities of the various collagens has not been reported previously. The major ligands for type VI collagen appear to be type IV FIG. 7. Double immunofluorescent labeling of human skeletal muscle. The frozen muscle section (5 m) was incubated for 3 h in a mixture of type IV collagen rabbit polyclonal antibody (top) and type VI collagen mouse monoclonal antibody VI-1 (bottom). Secondary antibodies used were anti-mouse IgG fluorescein isothiocyanate and anti-rabbit IgG Texas Red conjugates. Note the intense staining of both type VI and type IV collagens around the muscle fibers, indicating a co-distribution of the molecules. Using a monoclonal antibody to type IV collagen and a polyclonal antibody to type VI collagen produced the same result. Magnification: ϫ 200.
FIG. 8. Electron micrograph showing human muscle basement membrane with adjacent type VI collagen filaments labeled with 5-nm gold. Immunolocalization of type VI collagen in human skeletal muscle using monoclonal antibody 5C6 and a gold-labeled secondary antibody shows gold particles distributed in the endomysium among banded collagen fibrils and immediately adjacent to the basement membrane (BM). A Z-band of a muscle myofibril is labeled (Z). Scale bar, 100 nm. collagen, type VI collagen, and fibronectin. Although the interaction of type VI and fibronectin has not been studied in detail, they have been co-localized in structures referred to as strings in developing avian cornea (together with type IV collagen) (41) and in plaques in mouse articular cartilage (42). They also interact in the detergent-insoluble pericellular matrix around fibroblasts (11,12). The strong type VI/type VI interaction is probably important in the assembly and stability of beaded filaments in the extracellular space (17), so it is significant that the type VI/type IV interaction is even stronger. The interaction of type VI collagen with types I and II appears to be very low and probably not physiologically significant.
Type VI collagen has been specifically implicated in only one heritable disorder, called Bethlem myopathy (29,43). This is an autosomal dominant disease characterized by proximal muscle weakness with joint contractures and childhood onset. It is slowly progressive, with preservation of muscle function into late adulthood. In two different studies, disease loci have been linked to chromosome 21 in the region of the ␣1(VI) and ␣2(VI) chain genes (44) and to chromosome 2 in the region of the ␣3(VI) chain gene (45). This, together with the known distribution and properties of type VI collagen, made it a good candidate molecule. Mutations in the triple helical region of the ␣1(VI) chain in one family and the ␣2(VI) chain in two families have recently been identified (18), and the ␣3(VI) chain gene still remains a candidate for other families (46). In general, muscular dystrophies are caused by a disturbance of the attachment of muscle cells to their basement membrane (47). The perturbed attachment structure in muscle (Fig. 9) consists of intracellular dystrophin, a cell membrane dystroglycan complex, and merosin, an isoform of laminin, which is a basement membrane component (48). Disruption of this linkage, caused by mutations of the various proteins, results in different muscular dystrophy phenotypes (47). It was speculated that Bethlem myopathy may also be caused by a disturbance of a type VI/laminin interaction (18), but our results (not shown) and those of others show that there is no interaction between type VI and laminin (49). However, type VI collagen is associated with the external lamina of muscle cells in the myotendinous junction (50), and this study demonstrates a co-distribution of type VI and type IV collagens in human skeletal muscle. These studies have demonstrated a direct protein/protein interaction of type VI collagen with basement membrane type IV collagen, indicating that type VI filaments provide a physical link be-tween endothelial basement membranes and the surrounding matrix. Type VI collagen is therefore part of the muscle cell linkage complex that provides the important physical connection of muscle cells to their extracellular matrix. Mutations in the type VI collagen molecule in muscle may therefore disturb the anchoring of muscle basement membrane to the matrix, which could lead to a slow degeneration of muscle fibers as seen in Bethlem myopathy.
It appears that the interaction of type VI collagen with basement membranes involves multiple interaction sites, since the basement membrane proteoglycan perlecan also interacts with ␣2(VI) (8) and there may be other binding partners. It is therefore not possible to say exactly which interactions will be disturbed in Bethlem myopathy. If there are multiple binding sites on basement membranes for type VI collagen, this could also explain the mild phenotype, since the connection would not be completely broken, only weakened. FIG. 9. A schematic representation of the muscle-basement membrane interface. F-actin, dystrophin, and syntrophins are intracellular components of the muscle cell. The dystroglycan complex is a group of membrane proteins that interact with merosin in the basement membrane (BM). Type IV collagen, also in the basement membrane interacts with type VI collagen. This series of interactions provides a physical link between the extracellular matrix and intracellular filaments of the muscle cells.