Covalent and non-covalent interactions of betaig-h3 with collagen VI. Beta ig-h3 is covalently attached to the amino-terminal region of collagen VI in tissue microfibrils.

Transforming growth factor-beta induced gene-h3 (betaig-h3) was found to co-purify with collagen VI microfibrils, extracted from developing fetal ligament, after equilibrium density gradient centrifugation under both nondenaturing and denaturing conditions. Analysis of the collagen VI fraction from the non-denaturing gradient by gel electrophoresis under non-reducing conditions revealed the present of a single high molecular weight band that immunostained for both collagen VI and betaig-h3. When the fraction was analyzed under reducing conditions, collagen VI alpha chains and betaig-h3 were the only species evident. The results indicated that betaig-h3 is associated with collagen VI in tissues by reducible covalent bonding, presumably disulfide bridges. Rotary shadowing and immunogold staining of the collagen VI microfibrils and isolated tetramers indicated that betaig-h3 was specifically and periodically associated with the double-beaded region of many of the microfibrils and that this covalent binding site was located in or near the amino-terminal globular domain of the collagen VI molecule. Using solid phase and co-immunoprecipitation assays, recombinant betaig-h3 was found to bind both native and pepsin-treated collagen VI but not individual pepsin-collagen VI alpha chains. Blocking experiments indicated that the major in vitro betaig-h3 binding site was located in the pepsin-resistant region of collagen VI. In contrast to the tissue situation, the in vitro interaction had the characteristics of a reversible non-covalent interaction, and the Kd was measured as 1.63 x 10(-8) m. Rotary shadowing of immunogold-labeled complexes of recombinant betaig-h3 and pepsin-collagen VI indicated that the in vitro betaig-h3 binding site was located close to the amino-terminal end of the collagen VI triple helix. The evidence indicates that collagen VI may contain distinct covalent and non-covalent binding sites for betaig-h3, although the possibility that both interactions use the same binding region is discussed. Overall the study supports the concept that betaig-h3 is extensively associated with collagen VI in some tissues and that it plays an important modulating role in collagen VI microfibril function.

been stimulated with transforming growth factor-␤1 (1,2). ␤ig-h3 is also known variously as MP78/70 (3,4), RGD-CAP (5), and keratoepithelin (6). ␤ig-h3 has been established as a extracellular matrix protein in a wide variety of tissues including developing nuchal ligament, aorta, lung, and kidney and mature cornea, skin, bladder, and bone (7)(8)(9)(10)(11)(12). ␤ig-h3 is a 76 -78-kDa protein containing 4 repeat regions, with homology to the insect protein fasciclin, and 11 cysteine residues, most of which are clustered in a distinct amino-terminal region. The ␤ig-h3 molecule appears to undergo partial processing at the carboxylterminal end to yield a 68 -70 kDa isoform (2). ␤ig-h3 has been shown to bind in vitro to a number of other matrix components including fibronectin, laminin, and several collagen types (13,14). In addition, ␤ig-h3 has multiple cell adhesion motifs within the fasciclin-like domains that can mediate interactions with a variety of cell types via integrins ␣ 3 ␤ 1 (15,16), ␣ 1 ␤ 1 (17), or ␣ V ␤ 5 (18). The precise functions of ␤ig-h3 are unknown, but it has been proposed that it may act as a cell adhesion molecule (18) and as a bifunctional linker protein interconnecting different matrix molecules to each other and to cells (7,11). Recent evidence suggests that ␤ig-h3 may be important in endothelial cell-matrix interactions during vascular remodeling and angiogenesis (19) and that the protein functions as a negative regulator of mineralization during cartilage differentiation and osteogenesis (20,21). Mutations in the human ␤ig-h3 gene have been linked to several autosomal dominant corneal dystrophies (6) characterized by severe visual impairment resulting from the progressive accumulation of ␤ig-h3-containing protein deposits in the corneal matrix (22).
Ultrastructural localization studies on developing tissues showed that in most instances ␤ig-h3 was loosely associated with collagen fibers, although in developing kidney labeling was also observed close to the tubular and capsular basement membranes. Double immunolabeling experiments with antibodies to ␤ig-h3 and collagen VI indicated that much of the ␤ig-h3 was associated with collagen VI microfibrils rather than the collagen fibers themselves (7).
Collagen VI is present in the extracellular matrix of a wide range of tissues, usually as fine microfibrils 3-10 nm in diameter that exhibit a characteristic, double-beaded period of about 100 nm (23). In some tissues collagen VI appears to form additional structures including thicker cross-banded fibrils and hexagonal networks (24,25). Collagen VI consists of monomers containing three distinct polypeptides, ␣1(VI), ␣2(VI), and ␣3(VI), which form a triple helical collagenous central domain flanked by globular amino-and carboxyl-terminal domains. The monomers aggregate intracellularly, first into anti-parallel dimers with a 30-nm stagger and then into tetramers stabilized by intermolecular disulfide bonds before their secretion into the extracellular matrix (26). Ultrastructurally, the tetramers are symmetrical with double beads at each end of an extended central rod-like region. The outer beads represent the aminoterminal globular domains, and the inner beads correspond to the carboxyl-terminal globular domains (27). Extracellularly, the amino-terminal domains of the tetramers overlap end to end to form the collagen VI microfibrils (24,28). The microfibrils are resistant to digestion with bacterial collagenase (29,30), and treatment with pepsin removes the bulk of the globular domains from the tetramers but does not completely dissociate the microfibrils (28). However, the microfibrils can readily be dissociated into tetrameric subunits by treatment with a chaotropic agent (27) or low pH (31), indicating that the subunits are not covalently linked together within the microfibrils. The precise functions of collagen VI are unclear, but the protein is considered to be important for tissue architecture, interconnecting structural components of the matrix with each other and with cells (23). Mutations in collagen VI genes have recently been linked to the muscle-wasting disease, Bethlem myopathy (32).
In the present study we have isolated collagen VI microfibrils from collagenase-treated nuchal ligament and demonstrated that ␤ig-h3 is covalently attached to collagen VI at regular intervals along at least some of the microfibrils. The binding site is located close to the amino-terminal end of the collagen VI molecule. Additional binding assays showed that r␤ig-h3 binds in vitro to collagen VI but in a non-covalent manner. The results indicate that direct ␤ig-h3/collagen VI interactions occur in vivo, suggesting that they are likely to be important for the normal development and morphology of a range of tissues including cornea. Moreover these interactions appear to involve two distinct mechanisms of binding.

EXPERIMENTAL PROCEDURES
Materials-Pepsin-treated collagen VI was prepared as described previously (33). Individual ␣ chains were prepared from the pepsincollagen VI by high performance liquid chromatography. Briefly, reduced and alkylated pepsin-collagen VI was chromatographed on a Protein Pak DEAE-5PW column (Waters) in 50 mM Tris buffer, pH 8.4, containing 6 M urea using a linear gradient of 0 -0.2 M NaCl gradient (50 ml). The polypeptides eluted in the order pepsin-␣3 (VI), pepsin-␣2 (VI), and pepsin-␣1 (VI) (data not shown). The polypeptides were dialyzed into TBS before use in the binding assay.
Affinity-purified polyclonal antibodies to pepsin-treated collagen VI and to ␤ig-h3 peptide TQLYTDRTEKLRPEMEG have been described previously (7). Highly purified bacterial collagenase (type VII) and hyaluronidase from Streptomyces hyalurolyticus were purchased respectively from Sigma and Seikagaku Corp. (Tokyo, Japan).
Isolation of Collagen VI Microfibrils-Collagen VI microfibrils were purified using a method based on that of Kielty et al. (30). Briefly, nuchal ligament tissue (9 g) from a bovine fetus of 200 days of gestation was finely diced with a razor blade, rinsed several times with TBS, and resuspended in 18 ml of collagenase digestion buffer (50 mM Tris-HCl, pH 7.4, containing 0.4 M NaCl, 10 mM CaCl 2 , 2 mM N-ethylmaleimide, and 1 mM phenylmethylsulfonyl fluoride). The suspension was incubated with 1.8 mg of collagenase (specific activity ϭ 1980 units/mg) for 8 h at 37°C then centrifuged at 10,000 ϫ g for 20 min. The supernatant was chromatographed in 3 batches on a column of Sepharose CL-2B (1.6 ϫ 60 cm) equilibrated in Tris/NaCl buffer (composition as above but lacking CaCl 2 ). Flow rate was 30 ml/h, and 3-ml fractions were collected. Fractions containing ␤ig-h3 and collagen VI were identified by direct dot blotting on nylon membranes with specific antibodies. The void volume peaks from the 3 runs were combined and treated with 5 units of hyaluronidase for 24 h at 4°C. Samples of this digest were then analyzed by CsCl density gradient centrifugation in Tris/NaCl buffer (native conditions) or in Tris/NaCl buffer containing 4 M GdnHCl (denaturing conditions). Each sample was adjusted to a density of 1.30 g/ml with CsCl and then ultracentrifuged in a 70.1 Ti head (Beckman) at 30,000 rpm (62,000 ϫ g) at 15°C for 72 h. Each gradient was divided into 15 fractions, and those containing ␤ig-h3 and collagen VI were identified by dot blotting. Those fractions containing collagen VI micro-fibrils were analyzed by rotary shadowing (see below in "Experimental Procedures") and by SDS-PAGE and immunoblotting as described previously (34) except that PVDF-Plus membranes (Osmonics, Westborough, MA) were used. In some instances, samples were pretreated with 10 mM cysteine followed by 40 mM iodoacetamide using a previously described method (33).
Rotary Shadowing and Immunogold Labeling-Samples (5 l) from selected the CsCl gradient fractions were applied for 10 s to Formvarcoated nickel grids either directly or after adjustment to 4 M GdnHCl to depolymerize the collagen VI microfibrils into tetramer subunits. The grids were then rinsed with 3 changes of TBS and incubated with a 1:20 dilution of anti-(␤ig-h3 peptide) antibody (or a control antibody) in TBS plus Tween (0.05%) for 30 min. Grids were then rinsed again with TBS and incubated further with anti-(rabbit IgG) antibody conjugated to 10-nm gold particles (Auroprobe EM GAR G10, Amersham Biosciences) following the manufacturer's instructions. After further rinsing with TBS, the grids were dried, and rotary shadowed with platinum at an angle of 8°under a vacuum of 2 ϫ 10 Ϫ6 torr in a Cressington CFE50 Freeze Etch Unit. Each grid was observed using a Philips CM100 transmission electron microscope at 80 kV.
Expression and Purification of r␤ig-h3-Total RNA was purified as described previously from cultured osteoblast-like cells derived from a trabecular bone biopsy (12). First strand cDNA was synthesized using total RNA (1 g), random hexamer primers (50 ng), and Superscript II reverse transcriptase (200 units) (Invitrogen) following the manufacturer's instructions. The ␤ig-h3 cDNA was then amplified by PCR using forward primer 5Ј-CCCGCTCCATGGCGCTCTTCGTG-3Ј, reverse primer 5Ј-GCCGTGGTGCATTCCTCCTGTAGT-3Ј, and high fidelity Pfx DNA polymerase (Invitrogen). PCR was conducted as described previously for 28 cycles with an annealing temperature of 60°C (12). The PCR product (2092 bp) was purified by gel electrophoresis, A-tailed by incubation with dATP, and platinum Taq DNA polymerase (Invitrogen) at 70°C for 30 min, and cloned into pGEM T-easy vector (Promega, Madison, WI) following the manufacturer's instructions. An authentic ␤ig-h3 cDNA clone was selected, and a sequence encoding a His 6 tag was inserted into the construct using site-directed mutagenesis with a complementary pair of primers (5Ј-GGGTCCCGCCAAGCATCACCAC-CATCACCATTCGCCCTACCAGCTGGTG-3Ј and 5Ј-CACCAGCTGGT-AGGGCGAATGGTGATGGTGGTGATGCTTGGCGGGACCC-3Ј) and the QuikChange kit (Stratagene, La Jolla, CA) following the manufacturer's instructions. Again an authentic cDNA clone was selected, excised with NotI from the T-easy vector, and subcloned into the NotI site of the mammalian expression vector, pCEP4 (Invitrogen). Several clones were fully sequenced to check for errors, and the selected construct (pCEP/BH18) was transfected into 293-EBNA cells using the LipofectAMINE 2000 system (Invitrogen) following the manufacturer's instructions. Stably transfected cells were established in Dulbecco's modified Eagle's medium containing fetal calf serum (10%), penicillin, streptomycin, and nonessential amino acids and selected with hygromycin (50 g/ml). Recombinant ␤ig-h3 was harvested from the conditioned medium of confluent cells that had been incubated for 4 days in the above medium without serum. The medium was adjusted to binding conditions using an 8ϫ stock of binding buffer (10 mM phosphate, pH 7.60, containing 0.5 M NaCl and 10 mM imidazole) and loaded onto a nickel-agarose column (chelating-Sepharose fast flow, Amersham Biosciences). The column was washed with 10 volumes of binding buffer, and r␤ig-h3 was then eluted using the same buffer containing 500 mM imidazole. The quality and yield of r␤ig-h3 was determined by SDS-PAGE and immunoblotting using anti-(penta-His) antibody (Qiagen, Valencia, CA) and anti-(␤ig-h3 peptide) antibody.
Binding Assays-Solid phase binding assays were conducted as described previously (35). Briefly, microtiter plates (Immuno Maxisorp modules; Nunc, Roskilde, Denmark) were coated with collagen VI at 4°C for 18 h. Control wells were coated with BSA at the same concentration. After rinsing with TBS the wells were blocked in 3% nonfat dried milk in TBS for 1 h. After further rinsing, the wells were incubated either with unlabeled r␤ig-h3 or with 35 S-labeled r␤ig-h3 for 3 h at 37°C. See Figs. 6 and 5, respectively, for details. The wells were extensively rinsed with 0.05% Tween 20 in TBS. Binding of unlabeled r␤ig-h3 was measured using anti-␤ig-h3 antibody and a peroxide enzyme-linked immunosorbent assay technique (35), and that of 35 Slabeled r␤ig-h3 was determined by liquid scintillation counting. In some binding experiments, samples of r␤ig-h3 were pretreated with 40 mM iodoacetamide, 10 mM cysteine, or cysteine followed by iodoacetamide using a previously described method (33) and then dialyzed into TBS before use in the binding assay. The dissociation constant (K d ) for the interaction of ␤ig-h3 with collagen VI was calculated from non-linear regression analysis of the specific binding curve using the Prism program version 3 (GraphPad software, San Diego, CA).
Co-immunoprecipitation assays were performed following a previously described method (36) using affinity-purified collagen VI antibody (1 l) followed by protein A-Sepharose (20 l). See Fig. 5 for details. Co-immunoprecipitated 35 S-labeled r␤ig-h3 was measured by liquid scintillation counting.
For rotary shadowing analysis of binding interactions, pepsin-collagen VI (0.04 nmol) and r␤ig-h3 (0.02 nmol) were incubated together for 1 h in TBS, coated onto grids, and immunogold-labeled for ␤ig-h3 as described above. For a negative control, pepsin-collagen VI was incubated without r␤ig-h3.

RESULTS
␤ig-h3 Co-purifies with Collagen VI Microfibrils-A collagenase-digested extract of nuchal ligament tissue was analyzed by gel permeation chromatography on Sepharose CL-2B (Fig.  1A). Dot blotting revealed that most of the collagen VI was present in the Vo peak (peak 1), consistent with the molecule being present as microfibrillar aggregates. Interestingly, a significant signal for ␤ig-h3 was also detected in the Vo peak. To isolate collagen VI microfibrils, the Vo peak was treated with hyaluronidase and analyzed by CsCl equilibrium density gradient centrifugation. The gradient was selected so that collagen VI banded in the middle of the gradient with a buoyant density of about 1.3 g/ml, whereas fibrillin-containing microfibrils migrated to the bottom of the tube and other proteins occupied the fractions at the top of the gradient (Fig. 1B). Dot blotting showed that the ␤ig-h3 in the sample co-distributed precisely with the collagen VI microfibrils under non-denaturing conditions (NaCl) and under denaturing conditions (4 M GdnHCl) even though the buoyant density of the denatured collagen VI had decreased slightly. Rotary shadowing of fraction 7 from the non-denatured gradient revealed that type VI collagen microfibrils were the only discernable structures present (data not shown).
␤ig-h3 Is Covalently Bound to Collagen VI Tetramers in Nuchal Ligament Tissue-SDS-PAGE analysis of fraction 7 under reducing conditions on a 12% gel revealed several bands around 200 kDa, a strongly staining band at 140 kDa, and a band at 66 kDa ( Fig. 2A, lane 1). No other protein bands were detected on the gel. Immunoblot analysis revealed that the 66-kDa band was ␤ig-h3 ( Fig. 2A, lane 2) and that the 200-and 140-kDa bands were collagen VI polypeptides corresponding, respectively, to isoforms of the ␣3 (VI) chain and a mixture of the ␣1 (VI) and ␣2 (VI) chains ( Fig. 2A, lane 3). When the sample was analyzed under non-reducing conditions no bands were detected on 12% gels (data not shown). This finding indicated that both ␤ig-h3 and collagen VI were present as high molecular weight aggregates. However, when the sample was analyzed on a 3-5% gradient gel under non-reducing conditions, a single slowly migrating band was detected (Fig. 2B, lane 2) similar in size to the tetrameric form of collagen VI (apparent M r ϳ2,000,000). The band stained with antibodies to ␤ig-h3 (Fig. 2B, lane 4) and collagen VI (Fig. 2B, lane 6), suggesting that the two proteins were present in the same complex. When analyzed under reducing conditions the complex was disrupted, and bands corresponding to ␤ig-h3 and collagen VI were again the only bands detected in the sample (Fig. 2B, lanes 1, 3, and 5). Colorimetric analysis of the Coomassie Blue staining of the bands indicated an approximate molecular ratio of one molecule of ␤ig-h3 to two collagen VI monomers assuming that both protein exhibit similar uptake of stain.
In a further experiment an aliquot of fraction 7 was treated with a mild reducing agent, 10 mM cysteine, followed by the alkylating agent iodoacetamide to irreversibly break intermolecular but not intramolecular disulfide bonds. When analyzed by SDS-PAGE on composite agarose-polyacrylamide gels under non-reducing conditions, bands corresponding to collagen VI tetramers, dimers, and monomers were obtained (Fig. 2C, lane  1), which immunoblotted with anti-(collagen VI) antibodies (Fig. 2C, lane 3). However, immunoblotting with anti-␤ig-h3 antibodies stained only the ␤ig-h3 polypeptide (migrating at the dye front on gels of this porosity) and not the collagen VI bands, indicating that the covalent bonds between ␤ig-h3 and collagen VI had been completely disrupted by the cysteine treatment (Fig. 2C, lane 2). Overall, the results indicate that a significant amount of ␤ig-h3 in tissues is covalently bound to FIG. 1. ␤ig-h3 co-purifies with collagen VI under non-denaturing and denaturing conditions. Nuchal ligament tissue was digested with purified bacterial collagenase, and solubilized material was fractionated by gel permeation chromatograph on Sepharose CL-2B (Fig. 1A). Fractions containing collagen VI and ␤ig-h3 were identified by dot-blotting with appropriate antibodies and pooled as indicated (Peak 1). Collagen VI was then purified from the pooled material by incubation with hyaluronidase and then centrifugation on CsCl density gradients in non-denaturing (0.4 M NaCl (solid squares)) or denaturing (4 M GdnHCl (open circles)) conditions (Fig. 1B). See "Experimental Procedures" for details. Note that dot-blotting showed co-fractionation of ␤ig-h3 with collagen VI under both non-denaturing (NaCl) and denaturing conditions (GdnHCl).

␤ig-h3 Interactions with Collagen VI
collagen VI microfibrils via reducible cross-links, most likely disulfide bonds.
Immunolocalization of ␤ig-h3 on Isolated Collagen VI Microfibrils-Isolated collagen VI microfibrils were immunogold-labeled with anti-␤ig-h3 antibody and subjected to rotary shadowing (Fig. 3A). The ␤ig-h3 was found to be regularly localized along many of the collagen VI microfibrils in association with the double bead structures at each end of the extended interbead regions. No other macromolecular structures were evident in the preparations, and background binding of the gold particles to the grids was very low. Control preparations treated with anti-LTBP-2 antibody in place of anti-␤ig-h3 antibody showed no immunogold binding to the microfibrils (data not shown). To determine more precisely the location of the ␤ig-h3, the experiment was repeated with microfibrils that had been dissociated into tetrameric subunits by treatment with 4 M GdnHCl (Fig. 3B). The immunogold label consistently labeled the tetramers close to the most peripheral beads at the ends of the molecules. This corresponds to the amino-terminal region of the collagen VI monomers, suggesting that the covalent ␤ig-h3 binding site on collagen VI was in one of the aminoterminal globular domains or close to the amino-terminal end of the triple-helix. Again, control preparations treated with anti-LTBP-2 antibody showed no immunogold labeling of the molecules (data not shown).
Recombinant ␤ig-h3 Binds to Collagen VI in Vitro-Recombinant ␤ig-h3 was produced to further investigate the interaction of ␤ig-h3 with type VI collagen. His 6 -tagged r␤ig-h3 (Fig.  4A) was expressed in human 293-EBNA cells and purified by nickel affinity chromatography (Fig. 4B). The purified recombinant protein migrated as a doublet of bands corresponding to apparent molecular masses of 66 and 70 kDa (lane 1), which stained with both anti-penta-His antibodies (lane 2) and anti-␤ig-h3 antibodies (lane 3). The yield of purified r␤ig-h3 was routinely about 1 g of protein/ml of medium.
The r␤ig-h3 was shown in a solid phase assay to bind to native collagen VI and pepsin-collagen VI but not to the individual ␣ chains of pepsin-collagen VI, each of latter gave a signal similar to that of the BSA control (Fig. 5A). The binding of r␤ig-h3 to pepsin-collagen VI was shown to be proportional to the amount of collagen coated on the well (Fig. 5B). Similar results were obtained in co-immunoprecipitation experiments where r␤ig-h3 was also found to bind to pepsin-collagen VI but not to the individual ␣ chains (Fig. 5C). In further solid phase experiments, preincubation of the r␤ig-h3 in the liquid phase with pepsin-collagen VI was found to almost completely block FIG. 2. Immunoblotting shows complexing of ␤ig-h3 with collagen VI. Fraction 7 from the non-denaturing CsCl gradient (see Fig. 1B) was analyzed by SDS-PAGE on 12% gels under strongly reducing conditions followed by immunoblotting (A). Lane 1, Coomassie Blue-stained gel; lane 2, immunoblot for ␤ig-h3; lane 3, immunoblot for collagen VI. The sample was also analyzed on 3-5% gradient gels (Fig. 2B) under strong reducing conditions (lanes 1, 3, and 5) and non-reducing conditions (lanes 2, 4, and 6). Lanes 1 and 2, Coomassie Blue-stained gels; lanes 3 and 4, immunoblots for ␤ig-h3; lanes 5 and 6, immunoblots for collagen VI. The migration under reducing conditions of ␤ig-h3 (white arrows) and collagen VI ␣ chains (black arrows) is indicated on both gels. Also indicated is the single band identified under non-reducing conditions, which stains with antibodies to both ␤ig-h3 and collagen VI (large arrowheads). C shows analysis of an aliquot of fraction 7 that had been pretreated under mild reducing conditions, 10 mM cysteine for 1 h followed by alkylation with 40 mM iodoacetamide. SDS-PAGE and immunoblotting were performed using composite agarose (0.4%), polyacrylamide (2.5%) gels under non-reducing conditions. Lane 1, Coomassie Blue-stained gel; lane 2, immunoblot with anti-␤ig-h3 antibodies; lane 3, immunoblot with anti-(collagen VI) antibodies. The bands, corresponding in descending order to collagen VI tetramers, dimers, and monomers, are indicated with black arrows, and ␤ig-h3, migrating at the dye front, is indicated with a white arrow.

FIG. 3.
Immunolocalization of ␤ig-h3 on isolated microfibrils and molecules of collagen VI. Samples of fraction 7 from the nondenaturing CsCl gradient (see Fig. 1B) were analyzed by rotary shadowing and immunogold labeling with anti-(␤ig-h3 peptide) antibodies. Samples were analyzed either directly (A) or after disaggregation of collagen VI microfibrils with 4 M GdnHCl (B) (see "Experimental Procedures"). In A the 105-nm periodicity between individual collagen VI molecules within microfibrils is highlighted (arrows). Note the periodic immunogold labeling corresponding to the bead regions between the extended triple-helical domains. In B, left panel, individual collagen VI "tetramers" are shown. The regions corresponding to the amino-terminal domains (black arrows) and carboxyl-terminal domains (white arrows) are indicated in the top electron micrograph. Note that immunogold particles specifically label the region corresponding to the aminoterminal domains of the collagen VI monomers, which have also been represented schematically in the right panel with arrows indicating the immunogold attachment region. Bar ϭ 200 nm.

␤ig-h3 Interactions with Collagen VI
the interaction r␤ig-h3 with both native-and pepsin-collagen VI coated on the wells (Fig. 6). This result suggests that the major binding site for ␤ig-h3 on collagen VI occurs in the pepsin-resistant region of the molecule.
The in Vitro Interaction between ␤ig-h3 and Collagen VI Is Non-covalent-To determine whether intermolecular disulfide bonding occurs between the ␤ig-h3 and collagen VI, the solid phase assay was repeated with r␤ig-h3, which had been pretreated with iodoacetamide to block any free sulfhydryl group(s) on the molecule. Because ␤ig-h3 has 11 cysteine residues, at least one of them must not be paired in intramolecular disulfide bonding and may be free to form intermolecular disulfide links. However the pretreatment with iodoacetamide had no effect on the solid phase binding of r␤ig-h3 to pepsincollagen VI (data not shown). In addition, the experiment was repeated with a sample of r␤ig-h3 that had been pretreated with the mild reducing agent, 10 mM cysteine, to dissociate any intermolecular disulfide bonds that may have formed in the r␤ig-h3 preparation and, thus, ensure that the unpaired cysteine(s) of ␤ig-h3 was free to interact with the collagen VI. The pretreatment with 10 mM cysteine or with cysteine followed by iodoacetamide had no effect on the binding of r␤ig-h3 to collagen VI (data not shown), indicating that the unpaired cysteine(s) of ␤ig-h3 was not involved in the in vitro interaction between the proteins. Indeed, the binding of ␤ig-h3 to collagen VI had the kinetics of a reversible non-covalent interaction, and the dissociation constant was calculated from the solid phase binding curve to be 1.63 ϫ 10 Ϫ8 M, assuming that there is one binding site per molecule (Fig. 7).
Rotary Shadowing of Molecular Complexes Formed in Vitro-To identify the binding site for r␤ig-h3 on pepsin-collagen VI, the two molecules were incubated together to form a complex, and the location of the r␤ig-h3 molecules was identified on the pepsin-collagen VI by rotary shadowing and immunogold labeling for ␤ig-h3 (Fig. 8). The immunogold particles specifically labeled the ends of the pepsin-collagen VI tetramers. Control preparations lacking r␤ig-h3 showed no significant immunogold labeling of the pepsin-collagen VI molecules (not shown). This finding indicates that for the in vitro interaction, the ␤ig-h3 binding site is located close to the aminoterminal end of the triple helical region of the collagen VI molecule. DISCUSSION The first evidence that ␤ig-h3 can be covalently associated with a poorly soluble element of the matrix came from tissue extraction studies on collagenase-digested nuchal ligament of fetal calves (3). ␤ig-h3 (then known as MP78/70) was shown to be resistant to chaotropic extraction from this tissue homoge-nate, whereas it could readily be solubilized with saline containing a strong reducing agent. At that time the fibrillincontaining microfibrils were incorrectly considered to be the insoluble structures binding the ␤ig-h3. However, it was documented that such homogenates were also rich in collagen VI (29). Subsequent immunoelectron microscopic studies in a range of tissues showed ␤ig-h3 to be associated with collagen VI microfibrils rather than fibrillin-containing microfibrils (7), and this prompted the current investigation into the molecular basis of the association.
In the present study ␤ig-h3 was found to co-purify with collagen VI microfibrils, extracted from fetal nuchal ligament, on CsCl density gradients even under denaturing conditions. Analysis of the molecular composition of the microfibrils by SDS-PAGE and immunoblotting showed that ␤ig-h3 and collagen VI co-migrated under non-reducing conditions as a single high molecular weight complex close in size to that predicted for the tetramer subunit of collagen VI. Analysis under reducing conditions dissociated the complex into individual collagen VI ␣ chains and the ␤ig-h3 polypeptide. Because no other proteins were detected in the complex it was evident that ␤ig-h3 was covalently bound directly to the collagen VI tetramers by reducible cross-links. The observation that the ␤ig-h3 was readily dissociated from the complex by treatment with the mild reducing agent, 10 mM cysteine, indicates that the crosslink is likely to be an intermolecular disulfide bond(s). The molecular ratio of ␤ig-h3 to collagen VI in the aggregate was estimated as two ␤ig-h3 molecules per tetramer. A minimum of four ␤ig-h3 molecules per collagen VI tetramer would be needed to saturate all the binding sites assuming one binding site per collagen VI monomer. Thus, it would appear that not all of the binding sites are occupied on the microfibrils within the nuchal ligament tissue. It is unclear from the SDS-PAGE if each collagen VI tetramer has attached ␤ig-h3, since the gel would be unable to resolve ␤ig-h3-free tetramers from those containing a small number of ␤ig-h3 molecules. Ultrastructural analysis of the purified collagen VI microfibrils showed that ␤ig-h3 was periodically associated with at least some of the microfibrils from the nuchal ligament tissue. The immunogold labeling showed that ␤ig-h3 was consistently localized to the double bead region of the microfibrils and to the outermost beads of isolated collagen VI tetramers, indicating that the covalent binding site was located in the amino-terminal region of the collagen VI molecule.
To further investigate the interaction of ␤ig-h3 with collagen VI, in vitro binding assays using r␤ig-h3 were performed. Recombinant ␤ig-h3 was found to bind to both native and pepsintreated collagen VI in vitro. Moreover, the pepsin-collagen VI was shown to block the binding of r␤ig-h3 to native collagen VI, indicating that the major binding site(s) was contained within or close to the triple helix. In contrast to the findings from tissue-extracted microfibrils, the in vitro binding of r␤ig-h3 to collagen VI did not appear to involve disulfide bonding between the two molecules since reduction and alkylation of unpaired cysteines in the r␤ig-h3 with 10 mM cysteine and/or iodoacetamide did not block the interaction between the two proteins.
Indeed, the in vitro binding characteristics between r␤ig-h3 and collagen VI were those of a strong but reversible, noncovalent interaction with a K d measured as 16.3 nM, assuming one binding site on each molecule. The interaction is comparable in strength to that between ␤ig-h3 and fibronectin, which has a K d measured as 72 nM (13). Immunogold staining of r␤ig-h3-pepsin VI collagen complexes indicated that the in vitro ␤ig-h3 binding site was at the amino-terminal end of the triple helix.
The evidence indicates that ␤ig-h3 possesses covalent and non-covalent binding activities with collagen VI. It remains to be established if the both types of interaction utilize the same or distinct binding sequences at the amino-terminal end of the collagen VI molecule. However, from the data presented herein, FIG. 5. Recombinant ␤ig-h3 interacts with triple-helical collagen VI but not individual ␣ chains. Panel A shows solid phase binding assays. Purified proteins (500 ng/well) were coated onto microtiter plates. After blocking, 7 ϫ 10 3 dpm of 35 S-labeled r␤ig-h3 (specific activity 3.5 ϫ 10 5 dpm/g) was added, and the wells were incubated at 37°C for 3 h. After washing, binding was measured by liquid scintillation counting. Column 1, native collagen VI microfibrils; column 2, pepsin-collagen VI; column 3, pepsin-␣ 1 (VI); column 4, pepsin-␣ 2 (VI); column 5, pepsin-␣ 3 (VI); column 6, BSA. Panel B shows that in the solid phase assay, the binding of r␤ig-h3 is proportional to the amount of pepsin-collagen VI on the well. Pepsin-VI collagen was coated onto microtiter wells in serial dilution and incubated with 35 S-labeled r␤ig-h3 as described in panel A. Panel C shows co-immunoprecipitation experiments. 35 S-Labeled r␤ig-h3 (1 ϫ 10 4 dpm) was incubated with 1 g of pepsin-collagen VI, individual ␣ chains, or BSA for 4 h at 37°C. Collagen VI species were then recovered by the addition of anti-collagen VI antibody followed by protein A-Sepharose. Co-immunoprecipitated 35 S-labeled r␤ig-h3 was measured by liquid scintillation counting. Column 1, pepsin-collagen VI; column 2, pepsin-␣ 1 (VI); column 3, pepsin-␣ 2 (VI); column 4, pepsin-␣ 3 (VI); column 5, BSA control. In all panels, the means Ϯ S.D. of triplicate determinations are shown.
FIG. 6. Pepsin-collagen VI blocks the binding of r␤ig-h3 to native collagen VI microfibrils. Pepsin-collagen VI (columns 1 and 2) and native collagen VI microfibrils (columns 3 and 4) were coated onto microtiter plates (0.67 pmol/well). The wells were then incubated with r␤ig-h3 (1.2 pmol/well) either untreated (columns 1 and 3) or preabsorbed (columns 2 and 4) with pepsin VI collagen (8 pmol/pmol of r␤ig-h3). Binding was measured using the peroxide enzyme-linked immunosorbent assay technique with anti-␤ig-h3 antibody and color development at 490 nm. The binding of pre-absorbed r␤ig-h3 is expressed as a percentage of the binding of untreated r␤ig-h3. The means Ϯ S.D. of triplicate determinations are shown.
FIG. 7. Calculation of the K d for the interaction of ␤ig-h3 with collagen VI. Microtiter plates coated with pepsin-collagen (100 ng/ well) or BSA control (45 ng/well) were incubated for 3 h at 37°C with 35 S-labeled r␤ig-h3 at concentrations of 0 -250 ng/100 l (specific activity, 4.8 ϫ 10 5 dpm/g). The liquid phase was then removed, the wells were washed, and the amount of bound and unbound radioactivity in each well was determined by liquid scintillation counting. Specific binding was calculated as the average amount of r␤ig-h3 bound to tropoelastin-coated wells minus the average amount bound to the corresponding BSA-coated wells from triplicate determinations (not shown). The specifically bound r␤ig-h3 was plotted against the concentration of r␤ig-h3 originally added to the well, and the K d was calculated using non-linear regression analysis of the curve (see "Experimental Procedures").
there is a strong possibility that collagen VI contains only one major binding site for ␤ig-h3, close to the amino-terminal end of the triple helix, which is shared by both types of interaction. Indirect evidence for a single binding region includes (a) immunogold labeling for ␤ig-h3 to the amino-terminal region of both tissue-extracted native collagen VI and pepsin-treated collagen VI, (b) the almost complete blocking of ␤ig-h3 binding to native collagen VI by the pepsin-resistant fragment of the molecule in in vitro binding assays, and (c) a complete absence of non-covalently bound ␤ig-h3 in association with tissue-extracted native collagen VI microfibrils.
Interestingly, r␤ig-h3 did not bind to individual ␣ chains of pepsin-collagen VI, suggesting that the triple helical conformation of the molecule may be necessary for the non-covalent binding. However, separation of the three ␣ chains also involved disruption of interchain disulfide bonds at each end of the triple helix. Thus, it is possible that the integrity of pepsinresistant fragments of collagen VI N1 domains containing clusters of cysteines at the amino-terminal end of the helix is important for binding ␤ig-h3 rather than the triple-helix itself. Interestingly, the ␣1(VI), ␣2(VI), and ␣3(VI) chains have 3, 2, and 6 Cys residues, respectively, adjacent to the amino-terminal end of the triple helix. The odd number of Cys residues means that they cannot all be involved in intrachain disulfide bonding. It has been speculated that some may be involved in stabilizing interactions between adjacent N domains in collagen VI tetramers (23). However, it is possible that at least one of these cysteines may be involved in disulfide linkage with the covalently bound ␤ig-h3 extracted from tissues.
It remains to be established if the disulfide link between ␤ig-h3 and collagen VI is formed intracellularly before secretion of the proteins, the most likely possibility, or if it occurs extracellularly after non-covalent binding of the ␤ig-h3 to assembled collagen VI microfibrils. Recent evidence suggests that another binding partner of ␤ig-h3, fibronectin, has intrinsic protein disulfide isomerase activity that may be important for cross-linking fibronectin into extracellular matrix (37). It is possible that an as yet unidentified extracellular protein with disulfide isomerase activity is responsible for establishing disulfide bonding between ␤ig-h3 and collagen VI on the microfibrils. The cysteine residue(s) of ␤ig-h3 involved in the interaction remains to be identified. However, a prime candidate for intermolecular cross-linking is the lone cysteine residue (Cys-214) in the first fasciclin-like domain, which is predicted from x-ray diffraction studies of fasciclin to be exposed on the surface of the domain between two ␤-sheet structures (38).
The purpose of the ␤ig-h3 associated with collagen VI microfibrils remains to be elucidated. ␤ig-h3 interacts with integrin ␣ 3 ␤ 1 and ␣ V ␤ 5 cell surface receptors, and it appears to function as a cell adhesion molecule like its insect relative, fasciclin. Collagen VI microfibrils have also been shown to be associated with a variety of cell surfaces where they may bind via the proteoglycan NG2 (39) or via members of the ␤ 1 integrin family (40,41), including in cornea at least, the ␣ 3 ␤ 1 integrin (42). Therefore, the association of ␤ig-h3 with collagen VI microfibrils may be important for the modulation of their interaction with cell surfaces.
It is interesting that the non-covalent binding site for ␤ig-h3 on collagen VI is close to that for the small proteoglycans decorin and biglycan at the amino-terminal end of the triple helix (43). These proteoglycans, particularly biglycan, have recently been shown to organize collagen VI into network structures resembling those found in some tissues including cornea (44). Preliminary results from our laboratory indicate that in in vitro binding assays ␤ig-h3 binds to these proteoglycans and actually enhances their interactions with collagen VI. 2 Experiments are planned to determine the effects of ␤ig-h3 on collagen VI network formation. Collagen VI, ␤ig-h3, and dermatan sulfate proteoglycans have been immunolocalized with overlapping distributions within cornea (7,10,45), and ␤ig-h3 has been identified in collagen VI-enriched extracts of this tissue (46). Thus, it is possible that disruption of the normal interrelationship between these components contributes to the abundant deposition of abnormal ␤ig-h3 aggregates characteristic of ␤ig-h3-related genetic corneal dystrophies.
In conclusion, the finding that ␤ig-h3 can be covalently associated with collagen VI microfibrils in developing tissue supports the concept that it plays an important role in collagen VI biology, perhaps in cell-microfibril interactions, and in the organization and modulation of different collagen VI architectures to suit particular tissue environments.
FIG. 8. The ␤ig-h3 binding site on pepsin-collagen VI is close to the amino-terminal end of the triple helix. Pepsin-collagen VI was incubated in solution with r␤ig-h3, and the complex was analyzed by rotary shadowing and immunogold labeling with anti-(␤ig-h3 peptide) antibodies as described in the "Experimental Procedures." The left panel shows individual tetrameric (A-C) and double tetrameric (D) pepsin-collagen VI molecules. The right panel is a schematic representing each electron micrograph, with arrows indicating the immunogold attachment region. Note that in each case the immunogold particle(s) labels the outermost (amino-terminal) region of individual pepsin-collagen VI molecules. Bar ϭ 200 nm.