βig-h3 Interacts Directly with Biglycan and Decorin, Promotes Collagen VI Aggregation, and Participates in Ternary Complexing with These Macromolecules*

Recombinant human βig-h3 was found to bind 125I-labeled small leucine-rich proteoglycans (SLRPs), biglycan, and decorin, in co-immunoprecipitation experiments. In each instance the binding could be blocked by an excess of the unlabeled proteoglycan, confirming the specificity of the interaction. Scatchard analysis showed that biglycan bound βig-h3 more avidly than decorin with Kd values estimated as 5.88 × 10–8 and 1.02 × 10–7 m, respectively. In reciprocal blocking experiments both proteoglycans inhibited the others binding to βig-h3 indicating that they may share the same binding site or that the two binding sites are in close proximity on the βig-h3 molecule. Since βig-h3 and the SLRPs are known to be associated with the amino-terminal region of collagen VI in tissue microfibrils, the effects of including collagen VI in the incubations were investigated. Co-immunoprecipitation of 125I-labeled biglycan incubated with equimolar mixtures of βig-h3 and pepsin-collagen VI was increased 6-fold over βig-h3 alone and 3-fold over collagen VI alone. Similar increases were also observed for decorin. The findings indicate that βig-h3 participates in a ternary complex with collagen VI and SLRPs. Static light scattering techniques were used to show that βig-h3 rapidly forms very high molecular weight complexes with both native and pepsin-collagen VI, either alone or with the SLRPs. Indeed βig-h3 was shown to form a complex with collagen VI and biglycan, which appeared to be much more extensive than that formed by βig-h3 with collagen VI and decorin or those formed between the collagen and βig-h3, biglycan, or decorin alone. Biglycan core protein was shown to inhibit the extent of complexing of βig-h3 with native and pepsin-collagen VI suggesting that the glycosaminoglycan side chains of the proteoglycan were important for the formation of the large ternary complexes. Further studies showed that the direct interaction between βig-h3 and biglycan and between biglycan and collagen VI were also important for the formation of these complexes. The globular domains of collagen VI also appeared to have an influence on the interaction of the three components. Overall the results indicate that βig-h3 can differentially modulate the aggregation of collagen VI with biglycan and decorin. Thus this interplay is likely to be important in tissues such as cornea where such complexes are considered to occur.

Transforming growth factor-␤ (TGF-␤) 2 -inducible gene-h3 (␤ig-h3) (also known variously as MP78/70 (1,2), RGD-CAP (3), and keratoepithelin (4)), is an extracellular matrix protein expressed in a wide variety of tissues including developing nuchal ligament, aorta, lung, kidney, and cartilage; and mature cornea, skin, bladder, and bone (5)(6)(7)(8)(9)(10)(11). The name ␤ig-h3 stems from its identification and cloning as a major TGF-␤-responsive gene in A549 lung adenocarcinoma cells (12,13). ␤ig-h3 protein is 76 -78 kDa in size and contains four repeat domains, 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 carboxyl-terminal end to yield a 68 -70-kDa isoform (13). ␤ig-h3 has been shown to bind in vitro to a number of other matrix components including fibronectin, laminin, and several collagen types (14,15). In addition, ␤ig-h3 has multiple cell-adhesion motifs within the fasciclinlike domains that can mediate interactions with a variety of cell types via integrins ␣3␤1 (16,17), ␣1␤1 (18), or ␣V␤5 (19). The precise functions of ␤ig-h3 are unknown but it has been proposed that it may act as a cell adhesion molecule (19) and as a multifunctional linker protein interconnecting different matrix molecules to each other and to cells (5,9). Recent evidence suggests that ␤ig-h3 may be particularly important for skeletal muscle cell adhesion at the myotendinous junction (18), and for the induction of keratinocyte differentiation (20). The protein also appears to be involved in endothelial cell-matrix interactions during vascular remodeling and angiogenesis (21) and as a negative regulator of mineralization during cartilage differentiation and osteogenesis (22)(23)(24). Mutations in the human ␤ig-h3 gene (TGFBI) have been linked to several autosomal dominant corneal dystrophies (4) characterized by severe visual impairment resulting from the progressive accumulation of ␤ig-h3-containing protein deposits in the corneal matrix (25,26).
To elucidate the function of ␤ig-h3 within the extracellular matrix, studies in our laboratory have focused on the localization, molecular forms, and matrix interactions of ␤ig-h3 within various tissues. 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 (5). The collagen VI microfibrils, 3-10 nm in diameter, exhibit a characteristic, double-beaded period of about 100 nm (27). In some tissues collagen VI appears to form additional structures including thicker cross-banded fibrils and hexagonal networks (28,29). The precise functions of collagen VI are unclear but * This work was supported by the National Health and Medical Research Council of Australia and a University of Adelaide Faculty of Health Sciences Research Development Award. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 To whom correspondence should be addressed. Tel.: 61-8-8303-5337; Fax: 61-8-8303-4408; E-mail: mark.gibson@adelaide.edu.au.
the protein is considered to be important for tissue architecture, interlinking structural components of the matrix, and for cell-matrix interactions (27). Mutations in collagen VI genes (COL6A1, COL6A2, and COL6A3) have recently been linked to the muscle wasting diseases, Bethlem myopathy and Ullrich dystrophy (30,31).
To define the relationship of ␤ig-h3 with collagen VI 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 (32). The binding site is located close to the amino-terminal end of the collagen VI molecule. Additional binding assays have demonstrated that ␤ig-h3 binds in vitro to collagen VI, but in a non-covalent manner (32). The ␤ig-h3 attachment site on collagen VI appears to be close to those documented for the small leucinerich proteoglycans (SLRP), decorin and biglycan, and matrilins (33)(34)(35). SLRPs, in particular biglycan, have been shown to influence the aggregation and organization of collagen VI into networks in vitro, mimicking those found in some tissues, particularly cornea (28,34).
In the present study we investigated the influence of r␤ig-h3 on collagen VI aggregation and its interactions with decorin and biglycan in vitro. We have shown that r␤ig-h3 directly binds to biglycan and decorin and in turn forms ternary complexes with collagen VI and these SLRPs. Furthermore, r␤ig-h3 promotes the rapid aggregation of collagen VI tetramers into very large assemblies in vitro and differentially influences the aggregation of the collagen with the two SLRPs. The findings indicate that ␤ig-h3 is likely to be involved in the modulation of collagen VI-proteoglycan interactions during development of a range of tissues including cornea. It is possible that disruption of the normal interactions of ␤ig-h3 with collagen VI and/or SLRPs may be important for the development of ␤ig-h3-linked corneal dystrophies.

EXPERIMENTAL PROCEDURES
Materials-Biglycan and decorin were purified from the nuchal ligaments of 230-day-old fetal calves as described previously (36). Human r␤ig-h3 and pepsin-treated collagen VI were prepared and purified as described previously (32,37). Purified r␤ig-h3 was stored at 4°C in 50 mM Tris buffer, pH 7.4, containing 500 mM NaCl to prevent self-aggregation prior to aggregation experiments. Native collagen VI was purified from fetal bovine skeletal muscle. The frozen tissue was crushed and extracted at 4°C for 18 h in 5 volumes of TBS, pH 7.4, containing proteinase inhibitors and 0.1% Nonidet P-40. The non-solubilized material was then extracted extensively over 48 h with 0.6 M KCl, containing inhibitors (5 ϫ 10 volumes) to solubilize cytoskeletal proteins. The residue was rinsed with collagenase buffer (32) and digested sequentially with highly purified bacterial collagenase (7500 units/ml) at 37°C for 24 h and then hyaluronidase (200 units/ml) for a further 24 h. The supernatant was subjected at 4°C to CsCl equilibrium density gradient centrifugation at 30,000 rpm (g ϭ 193,000), with initial density of 1.325 g/ml, for 72 h in a fixed angle head (70.1 Ti) in a Beckman L7-55 ultracentrifuge. The fractions containing collagen VI microfibrils were pooled and purified further on a second CsCl gradient with initial density of 1.27 g/ml. Fractions containing collagen VI microfibrils were dialyzed into TBS, pH 7.4, containing 400 mM NaCl, and final purification was performed by fast protein liquid chromatography on a Superose 6 column (1.5 ϫ 25 cm). Proteinase inhibitors were included throughout the purification process. Purified native and pepsin-collagen VI were depolymerized by dialysis into 100 mM sodium citrate buffer, pH 4.0 (34). Any remaining aggregates were removed from the tetramers by centrifugation at 15,000 ϫ g for 10 min. Specific polyclonal rabbit antibodies to ␤ig-h3 and type VI collagen have been described previously (32,37). Anti-His 5 antibody was purchased from Qiagen.
Iodination of Biglycan and Decorin-Biglycan and decorin (100 g) were radiolabeled with Na 125 I (Amersham Biosciences) using IODO-BEADS (Pierce). Each proteoglycan was reacted with an IODO-BEAD and 0.75 mCi of 125 I for 5 min in 50 mM Tris buffer, pH 7.4, containing 0.3 M guanidinium chloride. Bound and free radiolabels were separated by gel filtration through Sephadex G-10. The specific activities of biglycan and decorin were both 3.4 ϫ 10 6 dpm/g of core protein unless stated otherwise. Core proteins of biglycan and decorin were prepared by digestion of the 125 I-labeled proteoglycans with chondroitinase ABC as described previously (36).
Control incubations contained no test protein. Specific rabbit polyclonal antiserum (10 l) raised to the test protein was then added and incubation was continued for 18 h at 4°C with gentle shaking. To ensure full recovery of the immunoprecipitate, protein A-Sepharose (30 l) was added and incubation was continued for 1 h at room temperature with gentle shaking. The immunoprecipitate-protein A-Sepharose complex was recovered by centrifugation (3,000 ϫ g for 10 min) and resuspended in 100 l of TBS containing 0.05% Tween 20. The complex was washed by centrifugation three times through 200 l of 1 M sucrose in TBS/Tween. The bound 125 I-proteoglycan was eluted from the complex by resuspension in non-reduced electrophoresis buffer (30 l) and heating to 100°C for 2 min. Each sample was centrifuged to remove the Sepharose beads and then ␤-mercaptoethanol was added to a final concentration of 2%. The samples were reheated for 2 min to reduce disulfide bonds and then analyzed for 125 I-labeled proteoglycan content by ␥ counting and SDS-PAGE/autoradiography as described previously (36). Each co-immunoprecipitation was performed in triplicate. The protocol was slightly modified for estimation of K d values, see Fig. 2 for details.
Static Light Scattering Analysis of Collagen VI Aggregation-An argon laser (Lexel model 95, Lexel Laser Inc.) was operated at a wavelength of 488 nm with an output power of 300 milliwatts. Light was focused into a sample cell containing a solution of the test macromolecules. The sample cell was maintained at 15°C by a temperature controller (Endocal model RTE-5DD, Neslab Instruments). The light scattered by the sample was detected at a scattering angle of 90°. Scattered light was detected by the single photon-counting photomultiplier tube (model EMI 9863B/350) with an aperture setting of 800 m. The output pulses from the photomultiplier were passed through an amplifier discriminator to a digital correlator (model BI-2030AT, Brookhaven Instruments Inc.). The correlator was set to run continuously with a sample time of 1 s and displayed the average count rate that was recorded every 30 s. Each experiment was commenced by mixing native or pepsin-collagen VI tetramers (120 nM) with r␤ig-h3 (120 nM) and/or a proteoglycan (biglycan or decorin (480 nM)) in TBS, directly in the cell.
Analysis of Collagen VI-containing Aggregates by SDS-PAGE-Pepsin-treated and native collagen VI samples were incubated in TBS containing ovalbumin (120 nM) at 4°C for 4 h with r␤ig-h3 only or with r␤ig-h3 plus biglycan or decorin at the concentrations indicated above. The samples were centrifuged at 18,000 ϫ g for 20 min and the supernatants were removed. The pellets were rinsed with TBS and dissolved in electrophoresis buffer. The proteins in the supernatants were recov-␤ig-h3 Interactions with Biglycan, Decorin, and Collagen VI MARCH 24, 2006 • VOLUME 281 • NUMBER 12 ered by acetone precipitation and dissolved in electrophoresis buffer using previously described methods (37). The precipitated and nonprecipitated protein fractions were analyzed by SDS-PAGE on 8% gels and Coomassie Blue staining.

RESULTS
␤ig-h3 Binds to Biglycan and Decorin via Their Core Proteins-Coimmunoprecipitation experiments were performed using 125 I-labeled biglycan and decorin to determine whether ␤ig-h3 directly binds one or both of these proteoglycans. Both biglycan and decorin were found to specifically co-precipitate with r␤ig-h3 indicating molecular interactions with the protein (Fig. 1). When the binding experiments were repeated with isolated core proteins from the two proteoglycans, each core protein co-precipitated showing that the ␤ig-h3 binding sites were contained within the core proteins and not in the GAG side chains. Gel and autoradiographic analysis of the co-precipitates confirmed that the precipitated radioactivity was associated with intact proteoglycans or, where appropriate, the isolated cores. Because the specific activities of the two proteoglycans and their cores were matched, comparison between the relative extent of binding could be made. Biglycan was precipitated to a greater extent than decorin suggesting that the former may have a higher affinity for ␤ig-h3. In contrast, the two core proteins were precipitated to an equal extent suggesting that their affinities for ␤ig-h3 were of similar magnitude. Because both cores were precipitated to a greater extent than the intact proteoglycans the results suggest that the interactions were inhibited to some degree by the GAG side chains. However, this could not explain the differences in the extent of ␤ig-h3 binding between the intact proteoglycans, because biglycan has a greater glycosaminoglycan content with two GAG side chains compared with one in decorin. Further immunoprecipitation experiments were performed over a range of proteoglycan concentrations that allowed K d values to be estimated for the interactions of ␤ig-h3 with decorin and biglycan (Fig. 2). Scatchard analyses estimated the K d values for ␤ig-h3 binding to decorin and biglycan as 1.02 ϫ 10 Ϫ7 and 5.88 ϫ 10 Ϫ8 M, respectively. This result confirms that biglycan binds ␤ig-h3 more strongly than decorin.
Blocking experiments were performed to confirm the specificity of the interactions and to determine whether the two proteoglycans compete for binding to ␤ig-h3 (Fig. 3, A and B). Preincubation of r␤ig-h3 with excess unlabeled biglycan completely blocked the subsequent binding of radiolabeled biglycan and decorin. Similarly unlabeled decorin blocked the binding of radiolabeled decorin and biglycan to the r␤ig-h3. The result indicates that biglycan and decorin either share the same binding site on ␤ig-h3 or that the sites are in close proximity such that the binding of one proteoglycan prevents the subsequent binding of the other. The binding activities were also completely blocked by both biglycan and decorin core proteins, indicating that the GAG side chains of the proteoglycans did not contribute measurably to the interaction of the intact proteoglycans with ␤ig-h3 (Fig. 3, C and D).
␤ig-h3 Enhances the Binding of Biglycan and Decorin to Collagen VI-The co-immunoprecipitation experiments were extended to include collagen VI in its subunit tetrameric form (27). The results are shown in Fig. 4. The co-immunoprecipitation of radiolabeled biglycan in the presence of equimolar mixtures of r␤ig-h3 and collagen VI was increased 6-fold over r␤ig-h3 alone (Fig. 4A) and 3-fold over collagen VI alone (Fig. 4B). Similarly co-immunoprecipitation of decorin with the r␤ig-h3/collagen VI mixture was increased 5-fold over r␤ig-h3 alone (Fig. 4C) and 3-fold over collagen VI alone (Fig.  4D). The amount of each proteoglycan precipitated with the r␤ig-h3/collagen VI mixture was similar when either anti-␤ig-h3 antibody and or anti-collagen VI antibody was used. Thus it appears that ␤ig-h3 enhances the complexing of both proteoglycans to collagen VI and that it participates in a ternary complex in which it interacts directly with both the collagen VI and proteoglycan molecules.
␤ig-h3 Rapidly Complexes with Collagen VI Tetramers in Vitro to Form Very Large Aggregates-Since ␤ig-h3 was found to enhance the interaction of collagen VI with biglycan and decorin, its effect on collagen VI aggregation was investigated in the absence and presence of each proteoglycan. The process of collagen VI aggregation was measured by static light scattering experiments. The direct effect of ␤ig-h3 on aggregation of collagen VI tetrameric subunits is shown in Fig. 5. ␤ig-h3 was shown to cause the rapid aggregation of the pepsin-collagen VI tetramers into very high molecular weight complexes (Fig. 5A). As measured by the light scattering, the process appeared to be essentially complete within 2 min. The degree of aggregation was found to be related to the concentration of r␤ig-h3 added with the maximum aggregation being observed at equimolar concentrations, and above, of r␤ig-h3 relative to the collagen. Note that no light scattering signal was obtained for r␤ig-h3 (not shown) or collagen VI alone, indicating that no detectable aggregation of the individual components was occurring within the time span of the experiment.
Similar levels of aggregation with r␤ig-h3 were detected when native collagen VI tetramers were used in place of the pepsin-collagen VI. As with pepsin-collagen VI, r␤ig-h3 was shown to promote the aggregation of native collagen VI in a concentration-dependent manner and close to maximum aggregation occurred with molar ratios of I:I (Fig. 5B). However, native collagen VI had less tendency than pepsin-collagen VI to aggregate with r␤ig-h3 at ratios below 2:1. The threshold light scattering FIGURE 1. Biglycan and decorin co-immunoprecipitate with ␤ig-h3. 125 I-Labeled biglycan (A), decorin (B), and their isolated core proteins (7 ϫ 10 4 dpm) in TBS (50 l) were incubated with 1 g of r␤ig-h3 or BSA control for 3 h at 37°C. Proteoglycan or core protein bound to ␤ig-h3 was then recovered by adding affinity purified anti-␤ig-h3 antibody (1 l of 400 g/ml), followed by protein A-Sepharose (20 l). Co-immunoprecipitated 125 I-labeled proteoglycan was measured by direct ␥ counting (top panels). Column 1, intact proteoglycan plus r␤ig-h3; column 2, intact proteoglycan plus BSA control; column 3, core protein plus r␤ig-h3; column 4, core protein plus BSA control. The mean Ϯ S.D. of triplicate determinations are shown. SDS-PAGE on 12% gels and autoradiography confirmed that the precipitated counts were associated with the proteoglycan or core. The results from one immunoprecipitation experiment are shown (bottom panels). Lane 1, intact proteoglycan plus r␤ig-h3; lane 2, intact proteoglycan plus BSA control; lane 3, core protein plus r␤ig-h3; lane 4, core protein plus BSA control. The apparent molecular masses of the biglycan (220 kDa), decorin (115 kDa), and their cores (46 kDa) are indicated.
␤ig-h3 Interactions with Biglycan, Decorin, and Collagen VI signal for native collagen VI aggregation required 30 nM r␤ig-h3, whereas some aggregation of pepsin-collagen VI was detected with as little as 2.4 nM r␤ig-h3. The results suggest that a globular domain(s) of collagen VI had an inhibitory effect on the aggregation process.
Interestingly, preincubation of the individual proteins for 5 min prior to mixing effected the extent of the aggregation (Fig. 6). Allowing collagen VI to pre-equilibrate in the incubation buffer resulted in some inhibition of the extent of aggregation. However, preincubation of the ␤ig-h3 at physiological electrolyte concentrations enhanced the aggre-gation. r␤ig-h3 and collagen VI have been shown to self-aggregate slowly when adjusted into physiological buffer from high salt and low pH, respectively (data not shown). Thus the results suggest that homotypic molecular interactions of collagen VI and ␤ig-h3 monomers may influence the size and shape of the aggregate containing the two proteins, i.e. collagen/collagen VI complexing may inhibit the subsequent interaction of ␤ig-h3 with the collagen, whereas ␤ig-h3/␤ig-h3 interaction may enhance aggregate formation with collagen VI. It is possible that some aggregation of ␤ig-h3 to dimers or oligomers (too small to FIGURE 2. Calculation of K d values for the interactions of biglycan and decorin with ␤ig-h3. r␤ig-h3 (0.5 g) was co-immunoprecipitated with increasing concentrations of 125 I-labeled biglycan (A, specific activity, 7.0 ϫ 10 6 dpm/g), or decorin (C, specific activity, 3.1 ϫ 10 6 dpm/g) as described under "Experimental Procedures." Specific binding (triangles) was calculated as the amount of proteoglycan co-precipitated with ␤ig-h3 (circles) minus that precipitated in control reactions where ␤ig-h3 was replaced with BSA (squares). The mean Ϯ S.D. of triplicate determinations are shown. Scatchard analysis was conducted to determine K d values for ␤ig-h3 binding to biglycan (B) and decorin (D). Slope of line ϭ Ϫ1/K d . FIGURE 3. Biglycan and decorin inhibit the binding of each other to ␤igh3. r␤ig-h3 (1 g) was incubated in TBS with unlabeled BSA, biglycan, decorin, or core protein (2 g) for 2 h at 37°C. 125 I-Labeled biglycan (A and C) or decorin (B and D) (7 ϫ 10 4 dpm) in TBS was then added and the incubation was continued for 3 h. 125 I-Labeled proteoglycan bound to r␤ig-h3 was recovered by co-immunoprecipitation with anti-␤ig-h3 antiserum (1 l) and measured by direct ␥ counting. A and B, blocked with intact proteoglycans. Column 1, r␤ig-h3 plus BSA positive control; column 2, r␤ig-h3 plus biglycan; column 3, r␤ig-h3 plus decorin; column 4, BSA no ␤ig-h3 control; column 5, biglycan no ␤ig-h3 control; column 6, decorin no ␤ig-h3 control. C and D, similar results were obtained when the unlabeled proteoglycans were replaced with their core proteins. Column 1, r␤ig-h3 plus BSA positive control; column 2, r␤ig-h3 plus biglycan core; column 3, r␤ig-h3 plus decorin core; column 4, BSA no ␤ig-h3 control; column 5, biglycan core no ␤ig-h3 control; column 6, decorin core no ␤ig-h3 control. The mean Ϯ S.D. of triplicate determinations are shown.
␤ig-h3 Specifically Enhances the Extent of Biglycan Aggregation with Collagen VI-The ␤ig-h3-collagen VI aggregation experiment was repeated in the presence of 4-fold excess biglycan or decorin (Fig. 7A). First, decorin and biglycan were found to form high molecular weight aggregates with pepsin-collagen VI, in the absence of r␤ig-h3, within a few minutes of mixing. This result is consistent with the findings of Wiberg et al. (34). The light scattering signals were of similar strength to that obtained between ␤ig-h3 and the collagen. Little observable difference in signal was detected if r␤ig-h3 was also included in the reaction with decorin and the collagen. However, the light scattering intensity was increased 3-fold if r␤ig-h3 was included in the biglycan-collagen VI reaction. Thus ␤ig-h3 appeared to be differentially interacting with biglycan to form distinct aggregates with collagen VI, which were more extensive than those formed with r␤ig-h3, decorin, and collagen VI, and between biglycan and collagen VI only. Similar patterns were evident for the influence of r␤ig-h3 on the interactions of native collagen VI with the proteoglycans, although, surprisingly, no aggregation was detectable by static light scattering for the native collagen with biglycan or decorin in the absence of r␤ig-h3 (Fig. 7B).
In parallel experiments, ␤ig-h3 was mixed with either native or pepsin-collagen VI in the presence or absence of biglycan or decorin and the complexes were allowed to stabilize. Ovalbumin was also included in each mixture. The complexes were recovered by centrifugation and analyzed by SDS-PAGE. The complexes with pepsin-collagen VI contained substantially less ␤ig-h3 (Fig. 8A) than those with native collagen VI (Fig. 8B), suggesting that the amino-terminal globular domains on the collagen, removed by pepsin treatment, may be important enhancers for binding to ␤ig-h3. For mixtures of three components, each precipitated aggregate contained collagen VI, r␤ig-h3, and the proteoglycan but lacked any ovalbumin content, confirming that all three molecules were specific components of the complexes. Incubation of the individual molecules alone resulted in little or no precipitation (Fig.  8C). As anticipated, the collagen VI content of the aggregates did not correlate with the static light scattering signal. For instance, in the absence of ␤ig-h3, substantial aggregates of native collagen VI with biglycan and decorin were recovered by centrifugation but these aggregates did not register significant light scattering signals indicating that they were of relatively small particle size. The complexes formed in the presence of r␤ig-h3 contained similar amounts of precipitable native collagen VI as those formed in the absence of ␤ig-h3. However, r␤ig-h3 FIGURE 4. Ternary complexing of ␤ig-h3 and biglycan or decorin with collagen VI. Aliquots of r␤ig-h3 (1 g) and collagen VI (2 g) were incubated separately, and in combination, for 2 h at 37°C in TBS (50 l). 125 I-Labeled biglycan (A and B) or decorin (C and D) (7 ϫ 10 4 dpm) were added and incubation was continued for 3 h at 37°C. Bound 125 I-labeled proteoglycan was recovered by co-immunoprecipitation with 1 l of anti-␤ig-h3 antiserum (A and C) or anti-collagen VI antiserum (B and D), and measured by direct ␥ counting. Column 1, r␤ig-h3; columns 2 and 6, BSA controls; columns 3 and 7, r␤ig-h3 and collagen VI; column 4, collagen VI (control); column 5, collagen VI; column 8, r␤ig-h3 (control). The mean Ϯ S.D. of triplicate determinations are shown. Note that both proteoglycans bound much more extensively to the r␤ig-h3/collagen VI complex than to r␤ig-h3 or collagen VI alone. The mean Ϯ S.D. of triplicate determinations are shown. ␤ig-h3 Interactions with Biglycan, Decorin, and Collagen VI greatly increased the particular size of the aggregates, especially of the complex formed between the collagen and biglycan, as indicated by the strong light scattering signal.
To determine whether the GAG side chains of the proteoglycans are important for their aggregation with ␤ig-h3 and collagen VI, the light scattering experiments were repeated with core proteins isolated from the proteoglycans by enzymatic digestion (Fig. 9, A and  B). Static light scattering analysis showed that the core proteins of decorin and biglycan caused only limited aggregation of the pepsincollagen VI (Fig. 9A). However, when r␤ig-h3 was added together with each core protein to pepsin-collagen VI, more aggregation was observed but in both instances the degree of aggregation was significantly less than that observed between r␤ig-h3 and collagen VI only (Fig. 9A). Similar results were obtained when the experiments were repeated with native collagen VI, although as anticipated from results with intact proteoglycans, no detectable aggregation of native collagen VI occurred with either core alone (Fig. 9B). The results indicate that both core proteins partially inhibited the aggregating effect of ␤ig-h3 on collagen VI. Unlike intact biglycan, its core did not form characteristically large aggregates with ␤ig-h3 and collagen VI. This finding suggests that the GAG side chains of the biglycan are critical for the formation of the very large complexes. When the light scattering experiments were repeated with 8-fold excess of proteoglycans no further increases in the light scattering signals were observed (data not shown), indicating that the contribution of each proteoglycan was already maximal when present in 4-fold excess.
The Globular Domains of Native Collagen VI Influence Ternary Complex Formation with ␤ig-h3 and Biglycan-To obtain clues as to which interactions were most important for the enhancement of collagen VIbiglycan aggregation by ␤ig-h3, the three components of the interaction were preincubated for 5 min in pairs before the third component was added (Fig. 10). Interestingly, if pepsin-collagen VI and biglycan were allowed to aggregate for 5 min before the addition of r␤ig-h3 to the   Complexes formed by collagen VI with r␤ig-h3 and/or biglycan (BGN) or decorin (DCN) were recovered by centrifugation, washed, and analyzed as described under "Experimental Procedures" by SDS-PAGE and Coomassie Blue staining. A, pepsin-collagen VI containing complexes; B, native collagen VI containing complexes; C, individual macromolecules incubated alone show little or no precipitation (control). The bands corresponding to each component of the complexes are indicated by arrows. p-colVI, pepsin-collagen VI; n-colVI, native collagen VI; DCN, decorin; BGN, biglycan. Note the absence of ovalbumin (ovalb) from each complex, which remains in the supernatants following centrifugation of the complexes. An example of a supernatant (S/N) is shown (pepsin-collagen VI incubated with r␤ig-h3 and decorin). MARCH 24, 2006 • VOLUME 281 • NUMBER 12 reaction mixture, no enhancement of the aggregation was evident. Similar results were obtained if pepsin-collagen VI was preincubated with r␤ig-h3 prior to the addition of biglycan. In contrast, if r␤ig-h3 and biglycan were preincubated for 5 min the subsequent addition of pep-sin-collagen VI caused an increase in signal close to that observed when all three components are directly mixed together. This finding points to the direct interaction of ␤ig-h3 with biglycan being a major factor in the enhancement of pepsin-collagen VI-biglycan aggregation by ␤ig-h3.

␤ig-h3 Interactions with Biglycan, Decorin, and Collagen VI
However, different results were obtained when the above experiments were repeated with native collagen VI (Fig. 10B). In contrast to the results using pepsin-collagen VI, preincubation of r␤ig-h3 with biglycan appeared to reduce the extent of aggregation of subsequently added native collagen VI. Preincubation of native collagen VI with ␤ig-h3 closely mimicked the aggregation obtained when all three components were added together simultaneously. However, the most dramatic increase in the light scattering signal was observed when native collagen VI was preincubated with biglycan prior to the addition of r␤ig-h3. The findings indicate that ␤ig-h3 may have a major positive influence on the polymerization of collagen VI in tissue situations particularly if biglycan is already bound to the collagen. This may be important in situations where high levels of ␤ig-h3 accumulate in the matrix, e.g. in certain congenital corneal dystrophies. In addition, a globular domain(s) of collagen VI (most likely the amino-terminal domain that is closest to the ␤ig-h3 and biglycan binding sites) appears to have a profound influence on ␤ig-h3-induced collagen VI aggregation, particularly if biglycan is already attached.

DISCUSSION
Recent evidence indicates that SLRPs, particularly decorin and biglycan may be complexed with collagen VI in tissues and that these proteoglycans can influence the rate, size, and shape of collagen VI aggregation in vitro (34,35). In particular, biglycan was shown to rapidly organize collagen VI monomers into a precise hexagonal network. Decorin was also reported to form similar complexes with the collagen but at a much slower rate. Most of these experiments were performed using pepsin-treated collagen VI, which lacks most of the globular amino-and carboxyl-terminal domains of the molecule. However, native collagen VI was reported to form similar aggregates, at least with biglycan (34). It was speculated that biglycan may be involved in the assembly of similar networks found in tissues such as cornea (28). The binding site for both decorin and biglycan has been located close to the amino-terminal end of the central triple helical region of the collagen VI molecule (33). In a recent study we have shown that ␤ig-h3 is covalently bound to collagen VI in some tissues and that the binding site is also close to the amino-terminal end of the triple helix (32). In the present study we have addressed the possibilities that ␤ig-h3 might inhibit or enhance biglycan and/or decorin interactions with collagen VI and also may have a positive or negative influence on collagen VI aggregation. For these experiments we have used both pepsin-treated and native collagen VI. The latter was purified from skeletal muscle as the collagen from this tissue was found to be substantially free of ␤ig-h3 in contrast to collagen VI sourced from ligament or cornea, which contains covalently bound ␤ig-h3 at a ratio of about one ␤ig-h3 molecule per tetrameric collagen molecule. To ensure that the native collagen VI was free of non-covalently bound contaminants it was purified by size exclusion chromatography and equilibrium density gradient centrifugation (36).
In co-immunoprecipitation experiments, r␤ig-h3 was found to directly interact with both biglycan and decorin, with K d values estimated as 5.88 ϫ 10 Ϫ8 and 1.02 ϫ 10 Ϫ7 M, respectively. These interactions appeared to be of the same order of magnitude but significantly weaker than the non-covalent interaction of ␤ig-h3 with collagen VI (K d ϭ 1.6 ϫ 10 Ϫ8 M (32)) and those of the proteoglycans with the col-  . The globular domains of collagen VI have a modulating influence on ternary complex formation with ␤ig-h3 and biglycan. The influence of individual intermolecular interactions on ternary complex formation was investigated by incubating two components together for 5 min before the introduction of the third component. Aggregation was monitored by static light scattering. A, pepsin-collagen VI, and B, native collagen VI. Solid triangles, collagen VI (120 nM), r␤ig-h3 (120 nM) and biglycan (480 nM) mixed simultaneously; squares, collagen VI preincubated with biglycan followed by addition of r␤ig-h3; circles, collagen VI preincubated with r␤ig-h3 followed by the addition of biglycan; open triangles, r␤ig-h3 preincubated with biglycan followed by addition of collagen VI. The mean Ϯ S.D. of triplicate determinations are shown.
␤ig-h3 Interactions with Biglycan, Decorin, and Collagen VI lagen (both K d values ϭ 3 ϫ 10 Ϫ8 M (34)). However, both proteoglycans appeared to bind to ␤ig-h3 more strongly than to the elastin precursor, tropoelastin, the latter interactions having been documented previously in the same solid phase system (36). The binding sites on ␤ig-h3 for the two proteoglycans appear to be shared or to occur in very close proximity to each other. Further experiments showed that ␤ig-h3 binds the proteoglycans via their protein cores. In addition, r␤ig-h3 enhanced, rather than inhibited, the interaction of each proteoglycan with collagen VI, resulting in ternary complex formation. The influence of ␤ig-h3 on collagen VI aggregation and complex formation with decorin and biglycan was investigated using a static light scattering technique. First, ␤ig-h3 was found to cause the rapid and extensive aggregation of both pepsin-and native collagen VI tetramers into large complexes. Interestingly, over 10-fold more r␤ig-h3 was required to initiate detectable aggregation of native collagen VI than of pepsin-collagen VI indicating that the globular domains of the collagen had a negative influence on the process. However, this was not because of inhibition of ␤ig-h3 binding because analysis of the aggregates by SDS-PAGE showed that much more ␤ig-h3 was complexed to the native collagen than to its pepsinized form. This finding also suggests that there is a direct interaction of ␤ig-h3 with the pepsin-sensitive amino-terminal globular region of collagen VI as well as the previously documented interaction with the pepsin-resistant domain (36).
The aggregates formed by pepsin-collagen VI with decorin or biglycan appeared to be of similar size to that formed between the collagen and ␤ig-h3. Surprisingly, no detectable aggregation of native collagen VI with either proteoglycan was observed over the time span of the experiment. However, in each case, a precipitable complex was formed that was detectable on SDS-PAGE, suggesting that these complexes were too small to be detectable by the light scattering technique. The findings contrast with those described by Wiberg et al. (34) who observed extensive formation of high molecular weight complexes containing native collagen VI and biglycan. Although there is no obvious explanation for this difference, the source and purity of the native collagen VI used in the experiments may be factors to consider. The native collagen VI used in our study is likely to be of higher purity than that used by Wiberg et al. (34), because of the absence of attached ␤ig-h3 in collagen prepared from skeletal muscle and removal of other contaminants by the extra density gradient step in our purification process. Density gradients have been found to remove strongly bound proteins such as latent TGF-␤binding proteins from fibrillin microfibrils (38,39) and the same is likely to be true for collagen VI microfibrils. It may be that additional proteins bound to corneal collagen VI, such as ␤ig-h3, influence the extent of aggregation stimulated by biglycan and decorin. Extensive aggregation of biglycan and decorin with native collagen VI was also observed when ␤ig-h3 was included in the initial mixture.
The ternary complexes formed with decorin were of a similar size to those formed between ␤ig-h3 and collagen VI only. In contrast, the ternary complexes of ␤ig-h3 and biglycan with both pepsin-collagen VI and native collagen VI appeared much more extensive as measured by light scattering. The results indicate a differential effect of ␤ig-h3 on the interaction of the two proteoglycans with the collagen. The ternary complexes formed between biglycan, r␤ig-h3, and both forms of collagen VI did not appear to contain more collagen VI than some of the other combinations. Therefore it appears likely that the high molecular weight complex formed between r␤ig-h3, biglycan, and collagen VI has a unique structural arrangement that yields a higher light scattering signal. However, despite extensive electron microscopic analyses of these aggregates conducted in our laboratory, we have been unable to identify any distinctive ultrastructural characteristics for this ternary complex. Additional experiments, where biglycan was replaced by its core protein, indicated that the GAG side chains of the biglycan were essential for the formation of the very large complexes. This latter observation is consistent with the findings of Wiberg et al. (34).
Further light scattering experiments analyzing the contribution of components of this ternary complex added sequentially, rather than simultaneously, showed differences between pepsin-collagen VI and native collagen VI, indicating that the globular domains had an important influence on the aggregation process. Interestingly, the most dramatic effect on the size of the ternary complex formed with native collagen VI was observed when the collagen was allowed first to bind biglycan prior to binding ␤ig-h3. Although no large complexes were formed between the collagen and the biglycan, the biglycan appeared to prime the collagen for dramatic aggregation on the addition of ␤ig-h3. In comparison, the later addition of biglycan to the native collagen VI-␤ig-h3 aggregate had little effect on the size of the aggregate. Thus at least in vitro, it would appear that the order of the protein binding events is a critical determinant for the size and organization of the ternary complex between collagen VI, ␤ig-h3, and biglycan. Because ␤ig-h3 is covalently attached, apparently stoichiometrically, by disulfide bonding to collagen VI in tissues such as ligament and cornea, determining when this attachment occurs during collagen VI synthesis may be critical to our understanding of how collagen VI supramolecular assembly is controlled in tissues such as ligament and cornea. If the ␤ig-h3 is attached pericellularly or intracellularly, which seems most likely, then subsequent exposure to biglycan may have little effect on the supramolecular structure of the collagen. If, however, biglycan can interact with the collagen before ␤ig-h3 is attached then a distinct collagen VI architecture may result.
It is clear from the experiments with pepsin-collagen VI that the direct interaction of ␤ig-h3 with biglycan can affect the extent of collagen VI aggregation in vitro. It raises the possibility that the binding of ␤ig-h3 to biglycan and decorin may also influence their molecular interaction with other matrix binding partners such as collagen I. ␤ig-h3 has been localized to interstitial collagen fibers as well as collagen VI microfibrils in a range of developing tissues (5). More recently, comprehensive analysis of ligament and cornea extracts has shown that, in addition to the form attached to collagen VI, ␤ig-h3 occurs as oligomeric homoaggregates. 3 It seems likely that this form of ␤ig-h3 has a function independent of collagen VI. Because ␤ig-h3 has been shown to bind collagen I in vitro (14) it is an interesting possibility that the homoaggregate form can directly associate with interstitial collagen fibers and that it may also influence interactions of biglycan and decorin with these structures.
Overall the results indicate that ␤ig-h3 can differentially modulate the interactions of biglycan and decorin with collagen VI in vitro and that it can influence the rate of formation, size, and shape of such aggregates. Because ␤ig-h3 is covalently bound to collagen VI in tissues such as ligament, and cornea where biglycan is also abundant it is likely that ternary complexes occur in vivo. Thus, ␤ig-h3 may be important for tailoring the organization and shape of these complexes in a range of tissue situations. The disruption of the normal inter-relationships between ␤ig-h3, collagen VI, and the SLRPs may contribute to the phenotype of ␤ig-h3-linked corneal dystrophies that are characterized by the development of abnormal ␤ig-h3-containing deposits in the corneal matrix (25,26). In contrast, the very low levels of ␤ig-h3 found in skeletal muscle make it unlikely that ␤ig-h3 contributes to the critical function of collagen VI in this tissue as illustrated by the phenotypic mani-