The NC1 domain of collagen IV encodes a novel network composed of the alpha 1, alpha 2, alpha 5, and alpha 6 chains in smooth muscle basement membranes.

Type IV collagen, the major component of basement membranes (BMs), is a family of six homologous chains (alpha1-alpha6) that have a tissue-specific distribution. The chains assemble into supramolecular networks that differ in the chain composition. In this study, a novel network was identified and characterized in the smooth muscle BMs of aorta and bladder. The noncollagenous (NC1) hexamers solubilized by collagenase digestion were fractionated by affinity chromatography using monoclonal antibodies against the alpha5 and alpha6 NC1 domains and then characterized by two-dimensional gel electrophoresis and Western blotting. Both BMs were found to contain a novel alpha1.alpha2.alpha5.alpha6 network besides the classical alpha1.alpha2 network. The alpha1.alpha2.alpha5.alpha6 network represents a new arrangement in which a protomer (triple-helical isoform) containing the alpha5 and alpha6 chains is linked through NC1-NC1 interactions to an adjoining protomer composed of the alpha1 and alpha2 chains. Re-association studies revealed that the NC1 domains contain recognition sequences sufficient to encode the assembly of both networks. These findings, together with previous ones, indicate that the six chains of type IV collagen are distributed in three major networks (alpha1.alpha2, alpha3.alpha4.alpha5, and alpha1.alpha2.alpha5.alpha6) whose chain composition is encoded by the NC1 domains. The existence of the alpha1.alpha2.alpha5.alpha6 network provides a molecular explanation for the concomitant loss of alpha5 and alpha6 chains from the BMs of patients with X-linked Alport's syndrome.

Type IV collagen, the major component of basement membranes (BMs), is a family of six homologous chains (␣1-␣6) that have a tissue-specific distribution. The chains assemble into supramolecular networks that differ in the chain composition. In this study, a novel network was identified and characterized in the smooth muscle BMs of aorta and bladder. The noncollagenous (NC1) hexamers solubilized by collagenase digestion were fractionated by affinity chromatography using monoclonal antibodies against the ␣5 and ␣6 NC1 domains and then characterized by two-dimensional gel electrophoresis and Western blotting. Both BMs were found to contain a novel ␣1⅐␣2⅐␣5⅐␣6 network besides the classical ␣1⅐␣2 network. The ␣1⅐␣2⅐␣5⅐␣6 network represents a new arrangement in which a protomer (triplehelical isoform) containing the ␣5 and ␣6 chains is linked through NC1-NC1 interactions to an adjoining protomer composed of the ␣1 and ␣2 chains. Re-association studies revealed that the NC1 domains contain recognition sequences sufficient to encode the assembly of both networks. These findings, together with previous ones, indicate that the six chains of type IV collagen are distributed in three major networks (␣1⅐␣2, ␣3⅐␣4⅐␣5, and ␣1⅐␣2⅐␣5⅐␣6) whose chain composition is encoded by the NC1 domains. The existence of the ␣1⅐␣2⅐␣5⅐␣6 network provides a molecular explanation for the concomitant loss of ␣5 and ␣6 chains from the BMs of patients with X-linked Alport's syndrome.
The basement membrane (BM), 1 a continuous sheet of extracellular matrix, separates epithelial cells from the underlying stroma and plays important roles in normal biological functions (such as cell adhesion, growth, and differentiation; tissue re-pair; and molecular ultrafiltration) as well as in pathological events (such as cancer cell invasion and metastasis). Moreover, degradation and de novo synthesis of vascular BMs are critical events in the angiogenesis processes. BMs function is impaired in hereditary and acquired diseases in which type IV collagen is affected, including Alport's syndrome, a hereditary form of progressive renal disease; diffuse leiomyomatosis, a benign proliferation of smooth muscle cells; and Goodpasture syndrome, an anti-type IV collagen autoimmune disease (1).
Type IV collagen is the major structural component of the BM, and it consists of a family of six homologous ␣(IV) chains, designated ␣1-␣6 (1). Each chain is characterized by a long collagenous domain of ϳ1400 residues of Gly-X-Y repeats, interrupted by ϳ20 short noncollagenous sequences, and by a noncollagenous (NC1) domain of ϳ230 residues at the carboxyl terminus. Three ␣(IV) chains assemble into triple-helical molecules (protomers) that further associate to form supramolecular networks by dimerization at the carboxyl terminus through NC1 domains and by formation of tetramers at the amino terminus (2). The chain composition, and thus the properties of the type IV collagen networks are influenced by two factors. First, the chain composition of networks is limited by chain availability: the six chains show a tissue-specific expression pattern, with the ␣1 and ␣2 chains being ubiquitous and the ␣3-␣6 chains having a more restricted distribution. Second, the NC1 domain confers specificity to the chain-specific assembly of networks (3); thus, yet unidentified recognition sequences must exist within the NC1 domain that direct the selection of chains to form triple-helical protomers and of triple-helical protomers to form networks. For instance, although the ␣1-␣5 chains are coexpressed in glomerular BM, they segregate into an ␣1⅐␣2 network and a disulfide cross-linked ␣3⅐␣4⅐␣5 network, characterized by loops and supercoiling (4). The existence of the latter network provides an explanation for the glomerular BM defect in Alport's syndrome, in which mutations in the genes encoding any of the ␣3, ␣4, and ␣5 chains perturb the assembly of the ␣3⅐␣4⅐␣5 network with the consequent loss of all three chains and progressive renal failure (3,4).
Additional distinct type IV collagen networks must also exist in BMs of other tissues that contain the ␣5 and ␣6 chains along with the ␣1 and ␣2 chains, but are devoid of the ␣3 and ␣4 chains. These include the smooth muscle BMs of bladder, uterus, stomach, esophagus, small intestine, vasculature as well as the BMs of skin and the Bowman's capsule of the kidney (5-8). The existence of a novel ␣5⅐␣6 network was suggested based on the immunohistochemical co-localization of these chains, but a link between ␣5 and ␣6 has not yet been biochemically proven. Mutations in the COL4A5 gene encoding the ␣5 chain of type IV collagen also cause the loss of the ␣6 chain from these BMs in the X-linked form of Alport's syndrome (9 -11), presumably reflecting the co-assembly of the ␣5 and ␣6 chains into a single triple-helical monomer that self-assembles into an ␣5⅐␣6 network. In rare instances of X-linked Alport's syndrome associated with diffuse leiomyomatosis, deletions occur in both the COL4A5 gene and the adjacent COL4A6 gene (encoding the ␣6 chain), leading to both renal failure and a benign proliferation of smooth muscle cells (12).
In this study, the network organization of the ␣5 and ␣6 chains of the smooth muscle BMs of aorta and bladder was investigated to acquire knowledge important for understanding their structure, function, and defects in disease. Surprisingly, the findings revealed the existence of a novel network composed of the ␣1, ␣2, ␣5, and ␣6 chains together, but not of the predicted network composed solely of the ␣5 and ␣6 chains. Evidence was obtained that supports our previous finding that the NC1 domains play a recognition role in the chain-specific assembly of networks. The linkage between the ␣5 and ␣6 chains in the ␣1⅐␣2⅐␣5⅐␣6 network provides a molecular explanation for the loss of both ␣5 and ␣6 chains in X-linked Alport's syndrome, in which the ␣5 chain is mutated.

EXPERIMENTAL PROCEDURES
Proteins-Frozen bovine aorta, bladder, and kidney were purchased from Pel-Freez Biological (Rogers, AR) and stored at Ϫ20°C. Aorta and bladder BMs were purified as previously described (13) and then digested with bacterial collagenase (Calbiochem), and the NC1 hexamers of type IV collagen were purified on DE52 and gel-filtration columns. Truncated type IV collagen protomers were produced by digestion of BMs with pseudolysin (Pseudomonas aeruginosa elastase; EC 3.4.24.26; purchased from Nagase Biomedical, Fukushiyama, Japan) (14). Aorta BM (80 g of wet ground tissue) was digested with 0.5% pseudolysin for 24 h at 4°C, then the insoluble fraction was further digested at 25°C for 24 h as described previously (15). The reaction was arrested by the addition of 20 mM EDTA, and then the truncated protomers solubilized by pseudolysin digestion were twice precipitated with 2 M NaCl and resolubilized in Tris-buffered saline (0.15 M NaCl and 50 mM Tris-HCl, pH 7.5) for electron microscopy. The chain composition of the truncated protomers was analyzed by collagenase digestion followed by Western blot analysis with NC1 domain-specific mAbs. Recombinant human ␣1-␣6(IV) NC1 domains were expressed in human kidney 293 cells and purified as described previously (16).
Antibodies-For detection of ␣1-␣5 NC1 domains of type IV collagen by Western blotting and immunohistochemistry, the H-series rat mAbs (H11, H22, H31, H43, and H52) were used (5). In addition, mAb5 to the ␣5 NC1 domain (from Wieslab AB, Lund, Sweden) and a rabbit polyclonal antibody to the ␣6 NC1 domain, raised against an ␣6 NC1 synthetic peptide (17), were also used as indicated. The following precipitating mAbs were used for immunoaffinity purification of NC1 hexamers according to the chain composition. mAb1 to the ␣1 NC1 domain and mAb3 to the ␣3 NC1 domain (18 -20) were purchased from Wieslab AB. Briefly, 5-10 g of NC1 hexamers were incubated with mAb supernatant (0.1-0.5 ml) for 1 h at room temperature. The immune complexes were collected on protein G-Sepharose by gentle mixing with 30 l of beads for 1 h at room temperature and then solubilized in sample buffer and analyzed by Western blotting. Precipitating mAbs B51 to the ␣5 NC1 domain and B66 to the ␣6 NC1 domain were produced in rat using bovine glomerular BM hexamers as antigen (6). The specificity of mAbs B51 and B66 was determined by ELISA against the recombinant ␣1-␣6 NC1 domains. Affinity columns containing immobilized mAbs B51 and B66 were prepared by coupling protein Gpurified mAb to Affi-Gel-10 columns (0.5-1 mg of IgG/0.5 ml of activated gel). For immunoaffinity purification, the NC1 hexamers from aorta or bladder BM (0.2-0.5 mg) were applied to the mAb B51 and B66 columns; the bound fraction was eluted with 5 M guanidinium chloride.
Indirect Immunofluorescence-Fresh bovine aorta and bladder were obtained at the local slaughterhouse. Fresh frozen sections (4 m) were fixed with acetone for 10 min at room temperature; treated with 6 M urea in 0.1 M glycine HCl buffer, pH 3.2, for 20 min at room temperature to expose the reactive epitopes (this treatment was not necessary for mAb B66); and then washed with phosphate-buffered saline and blocked with 10% skim milk. The sections were incubated with chainspecific mAbs (not diluted) for 60 min at room, washed with phosphatebuffered saline, and then incubated with fluorescein isothiocyanatelabeled goat anti-rat IgG antiserum (diluted 1:80) for 60 min. After a final wash, the samples were mounted and observed.
In Vitro Hexamer Assembly-Dissociation and in vitro re-association of hexamers were performed as previously described (3). Native NC1 hexamers from aorta BM were dissociated by dilution (Ͻ50 g/ml) into a solution of 50 mM formic acid buffered at pH 3.0 with Tris. Complete dissociation into NC1 monomers and dimers was verified by HPLC gel filtration using a Bio-Sil TSK 250 column (Bio-Rad). Reassembly of the dissociated NC1 subunits was performed by changing the buffer to Tris-buffered saline by repeated dilution-concentration cycles in Centricon-10 concentrators (Millipore Corp.) and then incubating the mixture (ϳ1 mg/ml) for 24 h at room temperature. The reaction products were separated according to their molecular mass by gel filtration on a Superdex 200 fast protein liquid chromatography column (Amersham Pharmacia Biotech). The composition of the NC1 hexamers re-associated in vitro was analyzed as described for the native NC1 hexamers.
Rotary Shadowing Electron Microscopy-The protein samples (25 g/ml) were diluted into a solution of 10 mM ammonium bicarbonate containing 60% glycerol, nebulized onto mica discs, evacuated in a Bal-Tec BAF500K freeze-etch-replica system to 1.7 ϫ 10 Ϫ9 millitorrs, coated at an angle (rotation rate of 60 rpm) with 0.9 nm of platinum at 120 K, and then backed with 8 nm of carbon. The replicas were viewed with a JEM 100CX II electron microscope (Jeol Ltd., Tokyo, Japan) and photographed at a magnification of ϫ29,000.

Chain Composition of Type IV Collagen of Bovine Smooth Muscle BM
Previous studies have shown that the ␣5 and ␣6 chains, along with the ␣1 and ␣2 chains, are co-localized in the smooth muscle BMs of murine, canine, and human tissues (5)(6)(7)(8). In this study, bovine aorta and bladder BMs were chosen as prototypes for the smooth muscle BMs of the vasculature and viscera because bovine tissues are amenable for isolation of sufficient quantities of type IV collagen for biochemical analyses (3,4,17,23). Immunostaining with chain-specific mAbs ( Fig. 1) showed that bovine aorta (a-g) and bladder (h-n) BMs contained the ␣1, ␣2, ␣5, and ␣6 chains, but were devoid of the ␣3 and ␣4 chains. The staining pattern for the aorta showed co-localization of the ␣1, ␣2, ␣5, and ␣6 chains in the smooth muscle BMs of the media. An arteriole, recognized as a circle of blue-colored nuclei and characterized by a thick wall of smooth muscle cells, also stained for all four chains in adventitia. In contrast, subendothelial BM, recognized as a thin layer of intima lining the upper end of the arterial wall, contained only the ␣1 and ␣2 chains. In bladder, the ␣1, ␣2, ␣5, and ␣6 chains were co-localized throughout the smooth muscle BMs of the lamina propria and media as well as in epithelial cell BM. The capillaries, recognized as rings with a thin membrane, contained only the ␣1 and ␣2 chains. These results show that the smooth muscle BMs of both bovine aorta and bladder are composed of ␣1, ␣2, ␣5, and ␣6 chains. These four chains can be arranged in several possible networks: separate ␣1⅐␣2 and ␣5⅐␣6 networks; a composite ␣1⅐␣2⅐␣5⅐␣6 network; or a mixture of networks with chain compositions of ␣1⅐␣2⅐␣5, ␣1⅐␣2⅐␣6, ␣1⅐␣5⅐␣6, and ␣2⅐␣5⅐␣6. Because all chains were co-localized, immunohistochemical analysis alone could not distinguish among these possibilities.

Arrangement of Chains in Protomers and Networks
Analysis of the Type IV Collagen Networks from Aorta BM by Digestion with Pseudolysin-Our previous studies (4,23) showed that pseudolysin digestion solubilizes truncated protomers of collagen IV with retention of a portion of the triplehelical domain and the complete NC1-NC1 connection between protomers (Fig. 2a). Moreover, digestion of glomerular BM at 4°C followed by digestion at 25°C differentially solubilized two distinct networks, one composed of the ␣1 and ␣2 chains and the other composed of the ␣3, ␣4, and ␣5 chains (3,4). Thus, pseudolysin cleavage can reveal information about the protomer and network organization of chains. In the case of aorta BM, rotary shadowing electron microscopy revealed that the truncated protomers existed as dimers, connected through NC1-NC1 domain interactions (Fig. 2b). Unlike glomerular BM, no differential solubilization of networks was observed for aorta BM, as the relative quantities of the NC1 domains of the ␣1, ␣2, ␣5, and ␣6 chains solubilized at 4°C (Fig. 2c) and at 25°C (Fig. 2d) did not differ. These results indicate that the ␣1, ␣2, ␣5, and ␣6 chains are arranged in triple-helical protomers, connected through NC1-NC1 interactions, forming one or more networks that are equally susceptible to pseudolysin digestion.
Precipitating mAbs to ␣5 and ␣6 NC1 Domains-The arrangement of the ␣1, ␣2, ␣5, and ␣6 chains was investigated by characterization of the NC1 hexamers isolated from aorta and bladder BMs by collagenase digestion. The monomer and dimer subunits of an NC1 hexamer reflect the identity of the chains composing two triple-helical protomers that are adjoined by their NC1 domains forming a network (24,25). 2 In previous studies of glomerular BM, a strategy has been perfected to (a) separate NC1 hexamers of varying compositions using chainspecific mAbs that have the capacity to immunoprecipitate native hexamers and to (b) identify the chain identity of hexamer subunits by Western blot analysis using chain-specific monoclonal antibodies. In particular, mouse mAb1 and mAb3, targeted to the ␣1 and ␣3 NC1 domains, respectively, were used to separate two subpopulations of NC1 hexamers, an ␣1⅐␣2 NC1 and an ␣3⅐␣4⅐␣5 NC1 hexamer, establishing the existence of the ␣1⅐␣2(IV) and ␣3⅐␣4⅐␣5(IV) networks in glomerular BM (3,18,20).
In this study of the ␣5 and ␣6 chains, it was necessary to establish an additional set of mAbs that immunoprecipitate NC1 hexamers, but that are targeted to the ␣5 and ␣6 NC1 domains. In a previous study (6), mAbs B51 and B66 were produced using native NC1 hexamers from bovine glomerular BM as the immunogen. The specificity and precipitating properties of these mAbs were investigated herein. As shown in Fig.  3a, mAbs B51 and B66 are specific for the ␣5 and ␣6 NC1 domains, respectively, as determined by their reactivity against ␣1-␣6 NC1 domains. The capacity of mAbs B51 and B66 to bind native NC1 hexamers was demonstrated by (a) their capacity to stain tissue sections without acid/urea treatment (data not shown), (b) their capacity to coprecipitate the ␣5 and ␣6 NC1 domains along with other NC1 domains in immu-FIG. 1. Immunohistochemical localization of the ␣1-␣6 chains of type IV collagen in aorta and bladder BMs. Aorta (a-g) and bladder (h-n) sections were stained with hematoxylin-eosin (a and h) and with mAbs H11 for ␣1 (b and i), H22 for ␣2 (c and j), H31 for ␣3 (d and k), H43 for ␣4 (e and l), H52 for ␣5 (f and m), and B66 for ␣6 (g and n) NC1 domains. Aorta and bladder BMs stained positively for the ␣1, ␣2, ␣5, and ␣6 chains, but not for the ␣3 and ␣4 chains. In the aorta (a-g), all four chains were co-localized in the smooth muscle BMs of the media and vascular wall of an arteriole (art; thick arrows) in adventitia, whereas the subendothelial BM (thin arrows) of the intima contained only the ␣1 and ␣2 chains. In the bladder (h-n), the epithelium and smooth muscle layer contained all four chains, whereas the capillaries (cap; arrow) contained only the ␣1 and ␣2 chains.
FIG. 2. Analysis of pseudolysin-solubilized aorta BM by electron microscopy and Western blotting. Digestion with pseudolysin solubilized truncated protomers of type IV collagen that retained the NC1-NC1 junction between two protomers (the NC1 hexamer) and a long triple-helical fragment (a). Truncated protomers were solubilized from aorta BM for 24 h at 4°C and examined by rotary shadowing electron microscopy (b); the bar represents 100 nm. The chain compositions of truncated protomers solubilized from aorta BM by pseudolysin digestion at 4°C (c) and 25°C (d) were analyzed by collagenase digestion of the protomers, followed by Western blotting with mAbs H11 (␣1) and H22 (␣2), mAb5 (␣5), and rabbit anti-␣6 synthetic peptide (␣6). D and M indicate the positions of NC1 dimers and monomers on the Western blot, respectively. noprecipitation experiments or in affinity chromatography (see below), and (c) preferential reactivity with native rather than dissociated NC1 hexamers as measured by ELISA (Fig. 3b). Thus, mAbs B51 and B66 have the required properties for structural studies of NC1 hexamers, analogous to those of the well described mAb1 and mAb3 (see above). In contrast, the H-series of mAbs, previously prepared for the six distinct NC1 domains using short synthetic peptides rather than the NC1 hexamer (5), preferentially reacted with dissociated hexamer in ELISA (Fig. 3b) and were inactive toward native hexamers in the immunoprecipitation assay (data not shown), but they remain invaluable for Western blot analyses and tissue localization studies.
FIG. 4. Analysis of the NC1 hexamers from aorta BM. Top, native NC1 hexamers were solubilized from aorta BMs by collagenase digestion, purified, and analyzed by Western blotting (a) for the presence of ␣1-␣6 NC1 domains. D and M indicate the positions of NC1 dimers and monomers on the Western blot, respectively. Center, the aorta NC1 hexamers were fractionated using mAbs B51 (b and c) and B66 (d and e) and mAb1 (f and g), and then the bound (b, d, and f) and unbound (c, e, and g) fractions were analyzed by Western blotting with chain-specific antibodies as described in the legend to Fig. 2. Bottom, the fractionation of the two NC1 hexamer populations of aorta BMs is shown schematically. NC1 heterohexamers. The permitted combinations remaining are hexamers of ␣1 only, ␣1⅐␣2, ␣1⅐␣5⅐␣6, and ␣1⅐␣2⅐␣5⅐␣6 NC1. The first two represent the allowable combinations in the hexamer fraction that did not bind to mAb B51 (or B66): the ␣1⅐␣2 NC1 hexamer must exist to account for the coprecipitation of ␣2 by mAb1, but the coexistence of an ␣1 homohexamer could not be ruled out because a precipitating mAb to ␣2 was not available. The last two represent the allowable combinations in the hexamer fraction bound to mAbs B51 and B66. The ␣1⅐␣2⅐␣5⅐␣6 NC1 hexamer must exist to account for the coprecipitation of ␣2 by mAbs B51 and B66. The coexistence of the ␣1⅐␣5⅐␣6 NC1 hexamer cannot be entirely ruled out based on these data alone, but it is inconsistent with the two-dimensional PAGE analysis (see below). Thus, aorta BMs consist of two NC1 hexamer populations, ␣1⅐␣2 and ␣1⅐␣2⅐␣5⅐␣6, representing 88 and 12% abundance, respectively (Fig. 4, bottom); the experimental evidence for the depicted organization of subunits in the NC1 hexamers is described below. Importantly, these results also exclude the existence of hexamers composed of ␣5 NC1 only, ␣6 NC1 only, and ␣5⅐␣6 NC1, which eliminates certain arrangements of the ␣5 and ␣6 chains in protomers (see below). These findings allow an interpretation of the sites in the aorta tissue for which the respective chains are localized. Hence, subendothelial BM contains the ␣1⅐␣2 NC1 hexamer, whereas smooth muscle BM contains both the ␣1⅐␣2 NC1 hexamer and the ␣1⅐␣2⅐␣5⅐␣6 NC1 hexamer.
Fractionation and Composition of NC1 Hexamers from Bladder BM-The NC1 hexamers of bladder BMs, chosen as a prototype of visceral smooth muscle BM, were analyzed in an analogous fashion to determine whether the composition of smooth muscle BMs is tissue-specific. Fractionation of the bladder BM hexamers using mAbs B51 (Fig. 5, b and c), B66 (d and e) and mAb1 (f and g) yielded the same qualitative results as the analysis of aorta hexamers. However, the ␣1⅐␣2⅐␣5⅐␣6(IV) hexamer population was more abundant in bladder BM (ϳ30%) than in aorta BM (ϳ18%). These results establish that the smooth muscle BM of the bladder contains hexamers composed of ␣1⅐␣2 and ␣1⅐␣2⅐␣5⅐␣6 NC1 domains, representing 70 and 30% abundance, respectively (Fig. 5, bottom). These results indicate that the basic organization of NC1 hexamers is the same in the smooth muscle BMs of two distinct tissues.
Chain Organization of Adjoining Protomers-The two populations of NC1 hexamers, ␣1⅐␣2 and ␣1⅐␣2⅐␣5⅐␣6, reflect several possible combinations of chains that exist within triple-helical protomers that are adjoined through their respective NC1 domains. The dimer subunits of these NC1 hexamers, observed under the dissociative conditions of SDS-PAGE, contain intermolecular disulfide bonds that connect NC1 domains of chains from each of two adjoining protomers (25-28). 2 Thus, the identification of dimers with respect to their chains of origin establishes which chains are cross-linked by disulfide bonds and therefore which chains exist in each of the two adjoining triplehelical protomers (Fig. 6b). The identities of NC1 dimers as well as monomers were determined by two-dimensional gel electrophoresis followed by Western blot analysis with chainspecific mAbs (Fig. 6a). To simplify the analysis, the ␣1⅐␣2 and ␣1⅐␣2⅐␣5⅐␣6 NC1 hexamers of bladder BM were separately studied.
The identities of monomers and dimers of the ␣1⅐␣2 NC1 hexamers of bladder BM were investigated first (Fig. 6, c-e). The monomers had several charge isoforms, whereas the dimers had both charge and size isoforms, as previously Center, the bladder NC1 hexamers were fractionated using mAbs B51 (b and c) and B66 (d and e) and mAb1 (f and g), and then the bound (b, d, and f) and unbound (c, e, and g) fractions were analyzed by Western blotting with chain-specific antibodies as described in the legend to Fig. 2. Bottom, the two NC1 hexamer populations thus identified in bladder BMs are schematically shown.

FIG. 6. Organization of the chains in the NC1 hexamers analyzed by two-dimensional gel electrophoresis.
Top, the NC1 hexamers from bladder BM (10 g) were separated by two-dimensional PAGE, transferred to nitrocellulose, and stained for total protein with colloidal gold (a). Under these conditions, the NC1 hexamer was resolved into its subunits, NC1 monomers (M) and disulfide-linked dimers (D), whose relationship to the parent hexamer is illustrated (b). The assignment of individual protein spots was based on the pattern of staining with chain-specific mAbs, as detailed below. Center, the bladder ␣1⅐␣2 hexamers (ϳ2 g) were separated by two-dimensional PAGE, transferred onto nitrocellulose, and analyzed by Western blotting with mAbs H11 for ␣1 (c) and H22 for ␣2 (d). Based on the identification of ␣1/␣1 and ␣2/␣2 homodimers, a model of the ␣1⅐␣2 hexamer was proposed (e). Bottom, the bladder ␣1⅐␣2⅐␣5⅐␣6 hexamers (ϳ3 g) were separated by two-dimensional PAGE; transferred onto nitrocellulose; and analyzed by Western blotting with mAb H11 for ␣1 (f), mAb H22 for ␣2 (g), mAb5 for ␣5 (h), and mAb B66 for ␣6 (i). Based on the identification of ␣1/␣5 and ␣2/␣6 heterodimers, three models for the ␣1⅐␣2⅐␣5⅐␣6 hexamer were proposed (j-l).
Although all three allowed isoforms have the same overall composition, each would be derived from a different pair of triple-helical isoforms. The hexamer isoforms depicted in Fig. 6  (k and l) would each presuppose the existence of two novel triple-helical protomers that interact with each other, but not with the predominant (␣1) 2 ⅐␣2 protomer. For instance, the hexamer model in Fig. 6k would imply the existence of an (␣1) 2 ⅐␣6 protomer connected to an (␣5) 2 ⅐␣2 protomer. These protomers would be relatively scarce (ϳ15% each in bladder BM); and upon interaction with the more abundant (␣1) 2 ⅐␣2 protomer (70%), they would form a mixture of ␣1⅐␣2⅐␣5 and ␣1⅐␣2⅐␣6 hexamers, neither of which was experimentally observed (Figs. 4 and 5). A similar argument would apply to the model shown in Fig. 6l. In contrast, the arrangement shown in Fig. 6j implies the interaction between a novel (␣5) 2 ⅐␣6 protomer and the classical (␣1) 2 ⅐␣2 protomer (the stoichiometry of ␣5 to ␣6 chains must be 2:1 to match that of ␣1 to ␣2 since ␣1 pairs with ␣5 and ␣2 pairs with ␣6). Thus, this model is compatible with all experimental findings and represents the most likely organization of the ␣1⅐␣2⅐␣5⅐␣6 hexamer.
In Vitro Reassembly of NC1 Hexamers from Smooth Muscle BM-In a previous study, we found that the NC1 domains contain recognition sequences that are sufficient to encode the assembly of ␣1⅐␣2 and ␣3⅐␣4⅐␣5 NC1 hexamers of glomerular BM (3). Likewise, it was of interest to determine whether the NC1 hexamers of aorta BM, after dissociation into monomers and dimers, had the capacity to reassemble into the two distinct ␣1⅐␣2 and ␣1⅐␣2⅐␣5⅐␣6 NC1 hexamers. This was investigated by comparing the composition of native hexamers from aorta BM with that of in vitro reconstituted hexamers using the experimental approach outlined in Fig. 7. Complete dissociation into NC1 monomers and dimers was achieved by dilution at pH 3.0, as shown by HPLC gel-filtration analysis (Fig. 7b). Under re-associative conditions, most (Ͼ95%) NC1 domains reassembled into hexamers, with only ϳ3% forming higher molecular mass aggregates (Fig. 7c). The composition analysis of the re-associated hexamers using mAb1 and B55 antibodies clearly showed that the re-associated hexamers were distributed in the same two subpopulations as the native hexamers. mAb1 antibodies quantitatively precipitated the ␣1, ␣2, ␣5, and ␣6 NC1 domains (Fig. 7d), with none remaining in the unbound fraction (Fig. 7e). In contrast, mAb B55 bound all ␣5 and ␣6 and some ␣1 and ␣2 NC1 domains (Fig. 7f), but some ␣1 and ␣2 NC1 domains remained unbound (Fig. 7g). These hexamer populations were identical to those found for native hexamers of aorta BM (see above), indicating that the NC1 domains specify the assembly of distinct hexamers. Similar results were obtained upon re-association of purified fractions containing NC1 monomers only or NC1 dimers only (data not FIG. 7. Analysis of NC1 hexamers re-associated in vitro from dissociated bovine aorta BM hexamers. HPLC gel filtration showed that the aorta BM hexamers (H) (a) were completely dissociated into NC1 monomers (M) and dimers (D) at pH 3 (b). After dissociation, the NC1 hexamers were re-associated in vitro and separated by HPLC gel filtration (c). The composition of hexamers re-associated in vitro was analyzed by fractionation with mAb1 (d and e) and mAb B51 (f and g), followed by Western blot analysis of the bound (d and f) and unbound fractions (e and g) fractions as described for the native hexamers of aorta BM in the legend to Fig. 4. shown), indicating that the reassembly of an ␣1⅐␣2⅐␣6⅐␣6 hexamer in vitro can proceed even in the absence of NC1 dimers, as previously found for the reassembly of ␣1⅐␣2 and ␣3⅐␣4⅐␣5 NC1 hexamers in vitro (3). DISCUSSION Random combinations of the six chains (␣1-␣6) of type IV collagen allows for 56 different protomers (triple-helical isoforms), which can further self-associate, forming a multiplicity of networks that differ with respect to which isoforms are connected through NC1-NC1 interactions (3). Studies to date indicate that only a few of these arrangements exist in BMs. All mammalian BMs appear to contain an ␣1⅐␣2 network, assembled from (␣1) 2 ⅐␣2 protomers that interact with each other at the carboxyl terminus through their NC1 domains (Fig. 8a) and at the amino terminus through their 7 S region. Certain BMs contain additional specialized networks such as the ␣3⅐␣4⅐␣5 network of renal glomerular BM (3,4), which self-assembles similarly from ␣3⅐␣4⅐␣5 protomers (Fig. 8b). 4 This study was designed to elucidate the organization of the remaining ␣6 chain, which is known to coexist with the ␣5 chain in smooth muscle BMs. Aorta and bladder BMs were chosen for study, as they could be isolated in sufficient quantities for biochemical analyses. The experimental strategy required the establishment of two new mAbs directed against the ␣5 and ␣6 NC1 domains (mAbs B51 and B66) for use in the immunoaffinity fractionation of NC1 hexamers.
In the smooth muscle BMs of both aorta and bladder, the ␣6 chain was found to assemble with the ␣5 chain, forming a novel (␣5) 2 ⅐␣6 triple-helical protomer, which further interacts with an (␣1) 2 ⅐␣2 protomer through the NC1 domains to yield an ␣1⅐␣2⅐␣5⅐␣6 hexamer (Fig. 8c). The "classical" network composed of the ␣1 and ␣2 chains was found to coexist with the ␣1⅐␣2⅐␣5⅐␣6 network in smooth muscle BMs, amounting to 70% in aorta BM and 82% in bladder BM. The ␣1⅐␣2 network also composed the subendothelial BM of the aorta, which is devoid of the ␣1⅐␣2⅐␣5⅐␣6 network (Fig. 9), as well as the corresponding BM of capillaries in the bladder.
We recently obtained evidence that the recognition mechanism governing the chain-specific assembly of the ␣1⅐␣2 and ␣3⅐␣4⅐␣5 networks of glomerular BM is encoded in the NC1 domains (3). Herein, a similar study was conducted with a mixture of ␣1⅐␣2 and ␣1⅐␣2⅐␣5⅐␣6 hexamers from smooth muscle BM. The native hexamers were dissociated into NC1 monomers and dimers, which were then reassembled in vitro. The organization of reconstituted hexamers duplicated that of native hexamers, demonstrating the specificity of interactions among NC1 domains. Thus, the NC1 domains of smooth muscle BM encode the chain-specific assembly of the ␣1⅐␣2 and ␣1⅐␣2⅐␣5⅐␣6 networks, as illustrated in Fig. 10. Additionally, the recognition mechanism must allow for the ␣5 chain to assemble into two different networks: the ␣3⅐␣4⅐␣5 and ␣1⅐␣2⅐␣5⅐␣6 networks. This may explain why, in various tissues, the expression of the ␣3 and ␣4 chains (in glomerular BM) is segregated from the expression of the ␣6 chain (in smooth muscle, skin, and Bowman's capsule BMs). Overall, these studies of glomerular and smooth muscle BMs indicate that the NC1 domains play a fundamental recognition function in the chain-specific assembly of networks: first, by specifying the selection of chains for protomer assembly, and second, by specifying the selection of protomers for network assembly.
Two distinct types of recognition sequences must specify the assembly of triple-helical isoforms and the interaction of protomers through NC1 domains, respectively. In the assembly of the ␣1⅐␣2 and ␣3⅐␣4⅐␣5 networks of glomerular BM, both types of interaction are specific, resulting in distinct hexamers and networks. In smooth muscle BM, specificity is manifested in the assembly of distinct (␣1) 2 ⅐␣2 and (␣5) 2 ⅐␣6 protomers; but subsequently, these protomers interact with each other to form a mixed network. This is presumably the result of higher homology between ␣1 and ␣5 NC1 domains in the ␣1-like subfamily (ϳ83% sequence identity compared with ϳ69% for ␣1-␣3 and ϳ68% for ␣3-␣5). We conjecture that residues divergent between the ␣1 and ␣5 NC1 domains (ϳ17%) encode the specific assembly of distinct protomers in the chain selection step, whereas residues identical in both the ␣1 and ␣5 NC1 domains but different in ␣3 (another ϳ18%) encode the protomer selection step. Thus, the (␣1) 2 ⅐␣2 protomers may interact with each other or with the more homologous (␣5) 2 ⅐␣6 protomers, but not with the ␣3⅐␣4⅐␣5 protomers. A similar reasoning would apply to the ␣2-like subfamily of NC1 domains, with the ␣2-␣6 pair having more homology (ϳ76% sequence identity) than the ␣2-␣4 (ϳ69%) and ␣4-␣6 (ϳ67%) pairs. FIG. 8. Triple-helical isoforms of type IV collagen and their interactions through the NC1 domains. Three kinds of NC1 hexamers with distinct compositions have been identified. The ␣1⅐␣2 hexamer (a) was first demonstrated chemically by studies with type IV collagen from the Engelbreth-Holm-Swarm tumor or human placenta (25,28,30) and was second demonstrated by immunoprecipitation with mAb1 (to the ␣1 NC1 domain) of hexamers from glomerular BM (3,20). The ␣3⅐␣4⅐␣5 hexamer (b) was identified by immunoprecipitation with mAb3 (to the ␣3 NC1 domain) of hexamers from glomerular BM (3,20). The existence of an ␣1⅐␣2⅐␣5⅐␣6 hexamer (c) was established in this study by immunoprecipitation with mAb1 and mAbs B51 and B66. These hexamers are derived from three basic kinds of triple-helical isoforms that differ in composition and stoichiometry of the six chains of type IV collagen: (␣1) 2 ⅐␣2, ␣3⅐␣4⅐␣5, 4 and (␣5) 2 ⅐␣6. The existence of an ␣1⅐␣2⅐␣5⅐␣6 network provides a molecular explanation for the loss of the ␣6 chain from BMs when the ␣5 chain is mutated in X-linked Alport's syndrome. It is well established that both the ␣5 and ␣6 chains are absent in the kidneys and skin of human patients with mutations in the COL4A5 gene and in multiple tissues (including skin, kidney, bladder, and lung) of a canine model of Alport's syndrome characterized by a COL4A5 nonsense mutation (7). Thus, the loss of the ␣6 chain is a consequence of mutation of the ␣5 chain, which causes failure of assembly of the ␣5⅐␣6 protomer and of the ␣1⅐␣2⅐␣5⅐␣6 network and/or degradation of the mutated network. This mechanism is analogous to that for the loss of the ␣3⅐␣4⅐␣5 network from glomerular BM when the ␣5 chain is mutated (3,4). The loss of the ␣3⅐␣4⅐␣5 network leads to deterioration of glomerular BM and progressive renal failure over 5-20 years, whereas the loss of the ␣1⅐␣2⅐␣5⅐␣6 network from smooth muscle tissues does not lead to apparent pathology, raising the issue of the biological significance of the latter network (7). The ␣3⅐␣4⅐␣5 network is required for long-term stability of glomerular BM and glomerular filtration function, possibly conveying resistance to the action of proteolytic enzymes (31,32). The long-term consequences of the loss of the ␣1⅐␣2⅐␣5⅐␣6 network on the function of smooth muscle in various tissues, if any, have not yet been reported.
The ␣6 chain is postulated to be critical for smooth muscle function based on the pathology associated with mutations in the COL4A6 gene. Mutations in both the COL4A5 and COL4A6 genes occur in patients with Alport's syndrome associated with diffuse leiomyomatosis (12,33). The combined mutations are associated not only with progressive renal failure, as seen in Alport's syndrome patients, but also with smooth muscle cell proliferation, forming benign tumors of the esophagus, tracheobronchial tree, and genital tract. However, recent studies with the canine model of Alport's syndrome have shown that smooth muscle cell proliferation is not a characteristic of pure Alport's syndrome, even though both the ␣5 and ␣6 chains are absent in the corresponding smooth muscle tissues of the canine model (7). Moreover, both the ␣5 and ␣6 chains are predicted to be absent in the smooth muscle BMs of human patients with Alport's syndrome, whether associated or not with diffuse leiomyomatosis, because of the existence of the ␣1⅐␣2⅐␣5⅐␣6 network. Thus, the absence of this network is not sufficient to cause diffuse leiomyomatosis; therefore, pathology of diffuse leiomyomatosis does not underscore the functional importance of the ␣6 chain. It is likely that diffuse leiomyomatosis is a result of mutations in a yet undiscovered gene near the locus of the COL4A6 gene, as postulated by others (34).