Evidence for Separate Networks of Classical and Novel Basement Membrane Collagen CHARACTERIZATION OF a3(IV)-ALPORT ANTIGEN HETERODIMER*

The COOH-terminal non-collagenous domains (NC1) of type IV collagen from glomerular basement membranes (GBM), lens capsule basement membranes, and Descemet’s membrane varied in the distribution of their NC1 subunits. All of these basement membranes (BMs) contained both classical (al(1V) and a2(IV)) and novel collagen chains (a3(IV), a4(IV) and the Alport antigen). Whereas GBM had a predominance of disul-fide-bonded subunits, the lens capsule and Descemet’s membrane were primarily monomeric, differences that are likely related to the functional and structural di-versity of collagen in various tissues. A heterodimer formed from monomeric subunits of a3(IV) and the Alport antigen exists in human and bovine GBM. This dimer represents an important cross-link of the NC1 domain of novel collagen. Additionally, immunoaffin- ity methodology showed that the novel BM collagen hexamers segregate into populations containing only novel BM subunits without the participation of the classical subunits (al(1V) and a2(IV)). These data pro- vided evidence for the presence of two separate networks of BM collagen: one containing al(1V) and a2(IV), and the other consisting of the novel collagen chains.

The COOH-terminal non-collagenous domains (NC1) of type IV collagen from glomerular basement membranes (GBM), lens capsule basement membranes, and Descemet's membrane varied in the distribution of their NC1 subunits. All of these basement membranes (BMs) contained both classical (al(1V) and a2(IV)) and novel collagen chains (a3(IV), a4(IV) and the Alport antigen). Whereas GBM had a predominance of disulfide-bonded subunits, the lens capsule and Descemet's membrane were primarily monomeric, differences that are likely related to the functional and structural diversity of collagen in various tissues. A heterodimer formed from monomeric subunits of a3(IV) and the Alport antigen exists in human and bovine GBM. This dimer represents an important cross-link of the NC1 domain of novel collagen. Additionally, immunoaffinity methodology showed that the novel BM collagen hexamers segregate into populations containing only novel BM subunits without the participation of the classical subunits (al(1V) and a2(IV)). These data provided evidence for the presence of two separate networks of BM collagen: one containing al(1V) and a2(IV), and the other consisting of the novel collagen chains.
Type IV collagen provides the structural framework for specialized sheets of extracellular matrix known as basement membranes (BM).' Intact al(1V) and a2(IV) molecules of 185 and 170 kDa, respectively, form heterotrimers of helical collagen in a ratio of 2:l (1). The collagen IV triple helix is joined to three additional helices at their NH2 termini (2), and the COOH terminus interacts with a second collagen helix at its COOH terminus (3), thus forming a network of type IV collagen similar to chicken wire. At the COOH terminus of each al(1V) and a2(IV) chain are non-collagenous domains (NC1) of 26 and 24 kDa, respectively. Homodimerization of * This work was supported in part by National Institutes of Health Grant AI10704, the Vikings Children Fund, The American Heart Association, and the Juvenile Diabetes Foundation. 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.
7 To whom correspondence should be addressed Dept. of Pediatrics, Box 491 UMHC, University of Minnesota, Minneapolis, MN 55455.
In a number of specialized BMs, including the glomerular basement membrane (GBM), this simplistic model of a type IV collagen network is complicated by the recognition in collagenase digests of two 28-kDa NCl monomers derived from additional collagen chains (6, 7). Two-dimensional gel analysis of human GBM reveals a charge distinction between the two peptides (M283+, pH > 9.0, M28+, pH 7.0) whereas the two bovine GBM peptides are closer in charge (pH 7.5 and 7.0, respectively) (6). Amino-terminal sequences of purified bovine monomers established the identity of all four monomers and revealed certain amino acid similarities between al(IV), a2(IV), and the two 28-kDa peptides (8). This has led to the designation of the peptides M2€i3+ and M28+ as derived from a3(IV) and a4(IV), respectively (9, 10). The aS(1V) NC1 is the primary target of Goodpasture autoantibodies (11) although the sera from patients with this disease react with all NC1 peptides on Western blots (6). Monoclonal and polyclonal antibodies with specificity for NC1 of al(IV), a2(IV), a3(IV), and a4(IV) identify homodimers of al(IV), a2(IV), and a4(IV) (12). Dimers which react with a3(IV) antibodies are not as cationic as the reactive monomer. Recently, the cDNA sequence of the a3(IV) NC1 has been determined and comparison with that of al(1V) and a2(IV) shows a 71 and 61% amino acid homology, respectively, with perfect alignment of cysteine residues characteristic of type IV collagen NC1 (13).
An additional NC1 peptide is detected using an alloantibody obtained from an Alport patient who developed anti-GBM nephritis post-transplantation (14). The reactive 26-kDa NC1 peptide (Alport antigen), more recently characterized using a monoclonal antibody (mAb A7) with identical specificity, is similar in size and charge to al(IV), but can be separated from al(1V) using immunoaffinity chromatography.* The Alport antigen, a3(IV) NC1, and a4(IV) NCl colocalize in the lamina densa of normal GBM using immunofluorescence microscopy and are absent from the GBM of male Alport patients (16, 17) whose GBM is laminated and split when analyzed by electron microscopy (18)(19)(20)(21). Like the Alport alloantibody, mAb A7 discriminates the X-linked mode of Alport antigen inheritance using immunofluorescence analysis of normal, homozygous, and heterozygous Alport GBM and epidermal basement membrane (16)

4137
A cDNA sequence with a 83% amino acid homology to al(1V) NC1, but with gene localization to the Alport locus on the X chromosome, has been designated as a5(IV) NC1 (22).
Mutations have been described in 3 out of 18 families tested by restriction fragment length polymorphism (23,24). A polyclonal antibody to a nonconsensus peptide of a5(IV) stains normal kidney in a distribution similar to that of a3(IV), a4(IV), and Alport antigen, but, in common with the latter three chains, is not detected in Alport kidney (22).* Because the a5(IV) NC1 and Alport antigen NC1 peptide identified by mAb A7 are similar in size and charge, we have made considerable efforts to establish the identity of the Alport antigen and its relationship to a5(IV).* Although the tissue distributions, absence from Alport GBM, and two-dimensional immunoblots of a5(IV) and the Alport antigen show identity, the necessary amino acid and nucleotide sequence data for each is unavailable at this time. At this time, we feel it is necessary to distinguish between a5(IV) and the Alport antigen identified by mAb A7 and the Alport alloantibody since the existence of two very homologous proteins has not been ruled out.
In this report, we have analyzed the subunit composition of three basement membranes: human and bovine GBM, bovine lens capsule (LC), and bovine Descemet's membrane (DM), all of which contain a3(IV), a4(IV), and the Alport antigen, as well as al(1V) and (u2(IV) (12). A heterodimer common to these BMs was formed by monomers of a3(IV) and the Alport antigen and did not contain al(1V) or a2(IV). Further, immunopurified hexamers segregated into disparate populations supporting the presence of two distinct networks of type IV collagen.

MATERIALS AND METHODS
Isolation of Basement Membranes-Fresh bovine renal cortices were diced and homogenized in 0.15 M NaCl using a Brinkmann Polytron (Brinkmann Instruments Co; Westbury, NY) at setting 4.5.
Glomeruli were retrieved by size-selective sieving (25). Anterior and posterior capsules were removed from lens of bovine eyes obtained from Pel-Freeze Biologicals (Rogers, AR). Descemet's membrane was peeled from the posterior corneal surface using blunt forceps after freeze/thawing of dissected bovine corneas. Human GBM was obtained as previously reported (6). The saline solution used in the isolation and extraction of the GBM, LC, and DM contained the protease inhibitors 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 1 mM N-ethylmaleimide, and 1 mM t-aminocaproic acid (inhibitor buffer). All stages of the preparation were kept on ice. Glomeruli were washed in PBS and disrupted by sonication in 1 M NaCl using 20-s bursts until whole glomeruli could not be found in the solution. GBM was retrieved by centrifugation, and washed extensively with PBS. LC and DM were also sonicated in 1 M NaCl to fragment the BM.
Preparation of Intact NC1 Hemmers-Washed BMs from GBM, LC, and DM were extracted for 24 h at 4 "C using 0.5 M NaCl in PBS (0.15 M NaCl, 0.01 M phosphate, pH 7.4), with protease inhibitors. Membranes were retrieved by centrifugation and a second 24-h extraction in 0.5 M NaCl, PBS was performed. Remaining membranes were then extracted for 72 h at 4 "C using 3% acetic acid containing 10 pg/ml pepstatin A, retrieved by centrifugation, dialyzed, and lyophilized (6). Extracted membranes were digested with type VI1 bacterial collagenase (Sigma) in 0.05 M Tris, 0.2 M NaCl, 0.002 M CaC12, pH 7.4, using 10 units/mg membranes. After 24 h of digestion at 37 'C, the insoluble material was removed by centrifugation at 10,000 X g, and the soluble GBM fraction was concentrated by evaporation and dialyzed into water and lyophilized.
Collagenase-digested GBM, LC, and DM were size separated on a Bio-Gel A-0.5 m (Bio-Rad) column (1.5 X 120 cm) run in 0.05 M Tris, 0.2 M NaCl, 0.1 M CaC12, 0.02% sodium azide, pH 7.4 as previously described (6). The NC1 pools were either concentrated by evaporation and dialyzed to equilibrium in TBS (0.05 M Tris, 0.15 M NaC1, pH 7.6) for immunoprecipitation or dialyzed against distilled water and lyophilized for gel application.
Preparation of Bovine NCl Dimers-Bovine GBM was extracted for 48 h in 6 M guanidine HC1, 0.05 M Tris, pH 7.4, retrieved by centrifugation, and washed extensively with 0.05 M Tris, 0.15 M NaCl, pH 7.4. The BMs were then dialyzed against distilled water and lyophilized. Membranes prepared in this way were then suspended in collagenase buffer and digested as described above.
Collagenase-digested bovine GBM was applied to a Sepharose-Q fast flow column (Pharmacia LKB Biotechnology Inc.) run in 0.5 M Tris, 4 M urea, pH 7.4. The unbound fraction containing the NC1 was pooled, dialyzed against distilled water, and lyophilized.
Unbound fractions from Sepharose-Q columns were separated into monomer and dimer pools by gel filtration using a S300 column (2.5 X 100 cm) (Pharmacia) run in 6 M guanidine HCl, 0.05 M Tris, pH 7.4. Pools of monomers and dimers were concentrated by evaporation and dialyzed into TBS for immunoprecipitation experiments. Dimer preparation has been described in detail elsewhere (26).
Gel Electrophoresis-To analyze NC1 subunit composition and immunoprecipitated proteins, 40 pg of bovine GBM, LC, and DM or the precipitation extract was applied to 8-18% SDS-polyacrylamide gel (PAGE) and electrophoresed using a discontinuous buffer system (27). 100-400 pg of human and bovine GBM, and bovine LC and DM NC1 were applied to each nonequilibrium pH gradient electrophoresis (NEPHGE) tube gel and electrophoresed for 2750 V h as described elsewhere (6,28). NEPHGE gels were separated in the second dimension by electrophoresis in 8-18% SDS-PAGE gels. Gels were either stained with Coomassie Blue dye or used for immunoblotting.
Affinity Purification and Immunoprecipitation-mAb A7 and mAb A2 were purified using a goat anti-mouse IgGl Sepharose column (Sigma) and either linked to activated Affi-Gel HZ Hydrazine (Bio-Rad) or used directly. 200 pl of NC1 hexamers or dimers were precipitated with 9 volumes of ethanol and redissolved in 200 pl of immunobuffer (0.01 M Tris, 0.14 M NaCl, 0.0025 M EDTA, 0.05% SDS, 0.5% sodium deoxycholate, 0.5% Triton X-100, pH 7.6). NC1 hexamer preparations were allowed to react with 200 p1 of the mAb A2 or mAb A7-Affi-Gel HZ Hydrazine for 1.5 h at room temperature with rotation. The mixture was washed six times with immunobuffer with 15-s microcentrifugations. The supernatant was removed from the last wash and the Aff-Gel was extracted with SDS sample buffer for 15 min at room temperature with rotation.
For dimer immunoprecipitation, dimers were resuspended in immunobuffer and reacted with 200 pl of mAb A2, mAb A7, and mAb 10 in separate reactions for 2 h at 37 "C. Goat anti-mouse IgG1 (35 pl) was then added to form lattices and incubated for 45 min at room temperature. Antigen-antibody complexes were precipitated by incubation with activated SAC (formalin-fixed Staph A cells) from Bethesda Research Laboratories followed by centrifugation. SAC complexes were washed six times in immunobuffer and either extracted with SDS sample buffer for SDS-PAGE analysis or lysis buffer for NEPHGE analysis.

Subunit Composition of Bovine NCl from GBM, LC, and DM
Purification and Analysis of NCl Hexamers-Collagenase digests of bovine GBM, LC, and DM, and human GBM were separated by nondenaturing gel filtration chromatography. In all preparations, NCl hexamers (noncovalently associated NC1 dimers and monomers) eluted in a position typical for the 160-kDa complex. Portions of these bovine NC1 preparations were lyophilized and applied to a SDS-PAGE gel which was subsequently stained with Coomassie Blue dye (Fig. 1). The GBM had the highest proportion of NC1 disulfide-bonded dimeric subunits while the LC and DM contained primarily NC1 monomers. Two-dimensional Gel Analysis of NC1 from Human and Bovine GBM and Bovine LC and DM-The lyophilized NC1 was further analyzed using two-dimensional NEPHGE gels and immunoblotting with mAbs for subunit identification. In the first experiment, human GBM was compared with bovine GBM (Fig. 2). Anti-al(1V) identified similar monomeric and dimeric subunits of pH 8.0 in both human and bovine NC1. These subunits, and those of a2(IV), pH 6.0, are quite prominently stained by Coomassie Blue dye, whereas staining of the novel NC1 subunits was quite weak. As previously observed, the a3(IV) subunits had a cationic migration in human NC1 (pH > 9.0) while those of bovine NC1 where neutral (pH 7.0). The bovine a3(IV) NC1 dimers appeared to consist of six heterogeneous subunits while those of human NC1 con- sisted of three subunits. The subunits of the Alport antigen were more complicated than those of a3(IV). In human GBM, mAb A7 recognized the Alport antigen, a 26-kDa monomer of pH 8.0 and homodimers of this monomer, as well as cationic dimers which were also identified with antibody to a3(IV). A similar picture resulted from the staining of bovine GBM. However the dimers which also react with anti-a3(IV) antibodies were more difficult to distinguish because of overlapping Alport antigen homodimers in this pH range.
In two-dimensional blots of bovine LC and DM (Fig. 3), migration was similar to the NC1 components of GBM. Coomassie Blue staining of gels confirmed the presence of a2(IV) NC1 monomers and dimers. We observed rather weak reactivity of mAb 10 (anti-al(1V) NC1) with the monomer and homodimeric subunits of al(IV) NC1. The a4(IV) NC1 which was detected with mAb 85 was overwhelmingly monomeric in subunit composition. LC and DM both contained monomeric subunits of a3(IV) and a heterogeneous group of dimers similar to the bovine GBM, with three migrating beyond the monomers in the pH gradient. mAb A7 (the Alport antigen) was extremely reactive with the LC NC1, even at 50 pg of applied antigen and the excessive proportion of monomers made analysis of that region difficult. However, the dimers appeared identical to those in the bovine GBM with recognition of the a3(IV) containing dimers. Reactivities of mAbs with DM were identical to those with LC. More discreet banding facilitated identification of the dimers reactive with anti-aS(IV) (mAb A2) and anti-Alport antigen (mAb A7).
In summary, the two-dimension gel analysis revealed monomers and homodimers of al(IV), a2(IV), a3(IV), and the Alport antigen. The bovine NC1 dimers containing a3(IV) were far more complex than their human counterparts and may contain a3((IV) homodimers in addition to the a3(IV)-Alport antigen heterodimers. Most importantly, this heterodimer of a3(IV) and the Alport antigen was present in all three bovine BMs and in the human GBM.
Heterodimer Analysis Using Immunoprecipitation-Bovine GBM NC1 dimers purified by denaturing chromatography, and concentrated by evaporation, were used in SAC immu- noprecipitations. 200 p1 of antigen were incubated with either anti-a3(IV) NC1, anti-Alport antigen, or anti-al(1V) NC1. Following a second incubation with secondary antibodies to facilitate lattice formation, the complexes were allowed to bind to Staphylococcus A cell coat protein A. The complexes were pelleted, washed, extracted with SDS-sample buffer, and separated on a SDS-PAGE gel. Colloidal gold staining was used to stain protein transferred to Immobilon by Western blotting (Fig. 4, left panels) and showed NC1 dimers precipitated by all three antibodies. Immunoblotting of the same precipitates with anti-Alport antigen indicated that dimers precipitated by either anti-Alport antigen or anti-a3(IV) NC1 contained subunits of the Alport antigen while those precipitated by anti-al(IV) NC1 antibody did not.
To further analyze the heterodimer of nS(IV)-Alport antigen, SAC precipitations using anti-Alport antigen were applied to NEPHGE gels and after electrophoresis in two dimensions were transferred to Immobilon and immunoblotted with anti-a3(IV) and anti-Alport antigen (Fig. 4, rightpanels). The dimers precipitated with the Alport antibody contained both the a3(IV)-Alport antigen heterodimers of pH 7.0-7.5 and Alport antigen homodimers of pH 7.5-8.0.

Immunopurification of Intact Bovine NCl Hexamers
Affinity purified anti-a3(IV) NC1 (mAb A2) and anti-Alport antigen (mAb A7) were used to construct Affi-Gel HZ hydrazine affinity matrices. Anti-nl(1V) NC1 (mAb 10) was also used, but this antibody failed to bind antigen when linked to the support system. NC1 from bovine GBM, LC, and DM were subjected to batch purification using these two affinity matrices. It is noteworthy that NC1 isolated for this purpose was not subjected to denaturing conditions during its purification procedure. When experiments were performed with NC1 which had been subjected to 4 M urea, the results indicated that hexamers dissociated in this buffer and reassociated at random. Use of nondenaturing buffers and deletion of the Q-Sepharose step were precautions to keep NC1 hexamers intact until the post-purification extraction step. The and centerpanels, bovine NC1 dimers were immunoprecipitated with each of the following, anti-nS(1V) NC1 (mAb A2), anti-Alport antigen (mAb AS), and anti-tul(1V) NC1 (mAb 10) and then suhjected to SDS-PAGE. Colloidal gold staining of the three preparations is shown in the left panel and the Western blots stained with anti-Alport (mAb AT) in the center panel. Note that dimers precipitated with anti-n3(IV) NC1 or with anti-Alport react with anti-Alport, hut that dimers precipitated with anti-nl(1V) NC1 do not (center panel). The arrows indicate the 54-kDa dimers. In the right panel, two-dimensional Western blots of dimers precipitated with a n t i -d ( I V ) NC1 (top, right) or anti-Alport (lower right) are stained with anti-Alport antigen mAb AS. The dimers precipitated with either antibody contain the cr3(IV) NC1-Alport antigen heterodimer at pH 7.5. presence of Alport antigen monomers in precipitated protein from both matrices confirmed that the noncovalent bond within the hexamers had not been dissociated. The Affi-Gel HZ antibody-antigen complex was thoroughly washed and extracted into SDS sample buffer for analysis by immunoblotting.
When NC1 hexamers of GBM LC and DM were purified using the anti-Alport HZ matrix they contained subunits reactive with both anti-Alport antigen (homodimers and monomers of the Alport antigen and the heterodimer of a3(IV)-Alport antigen) and anti-a3(IV) NC1 (monomer and the heterodimer of n3(IV)) (Fig. 5A). The trend in these reactions to favor purification of the Alport antigen subunits was purely stoicheometric, since there are more Alport antigen containing subunits in the NC1. Most importantly, these hexamers did not react with anti-al(1V) NC1 antibody.
A control (c) lane contains GBM hexamers stained with secondary antibody only. Immunoglobulin reactive hands (arrows on right) are present in C and other lanes. Note that monomers and dimers (arrouls on left) reactive wit.h both anti-Alport (lanes [1][2][3] and anti-d(IV) (lanes 4-6) but not with anti-tul(1V) (lanes 7-9) are present in GRM, LC, and DM. Lane 4 stained weakly through dimeric components are present. The bands observed in lanes 7-9, stained with anti-eul(IV) are the same as those on the control lane ( c ) , which was stained only with secondary antibody. These represent IgG fragmentation from the affinity matrix occurring in the SDS extraction buffer following immunoprecipitation. H , hexamers of bovine GRM (lanes I , 4, and 7) LC (lanes 2, 5 , and R ) , and DM (lanes 3 , 6, and 9 ) were isolated by binding to an affinity matrix containing anti-tu3(IV) NC1 (mAb A2). Following SDS-PAGE and Western blotting, filters were stained with anti-Alport (lanes I-3), anti-d(IV) (lanes 4-6), and anti-nl(IV) (lanes 7-9). Lane IO contains the starting material ( b N C I ) stained with anti-nl(1V) for reference. Note the presence of monomers and dimers of Alport antigen (lanes [1][2][3] and er3(IV) (lanes 4-6)  weak to illustrate. Analysis of a2(IV) was limited to visualization on silver-stained gels and colloidal gold-stained blots since an a2(IV) monoclonal antibody has not been generated. These analyses indicated that a2(IV) segregated with al(1V) when hexamers were precipitated with either antibodies to (~3(1V) or the Alport antigen. We conclude that hexamers composed of the novel collagen NC1 subunits do not contain NC1 subunits of classical type IV collagen al(1V) and a2(IV) chains.

DISCUSSION
These studies show clearly that the NC1 domains of the novel collagen chains, a3(IV) and Alport Antigen, interact with each other in the formation of dimeric and helical structure and are not linked to the classical type IV collagen chains, al(1V) and a2(IV). Immunohistochemical studies suggest that two different networks of BM collagen exist. In contrast to al(1V) and a2(IV), the new chains are restricted to specialized BMs, have been shown to be differentially expressed during ontogeny (30), and are spatially and temporally distinct in matrix during the evolution of certain renal diseases, diabetic nephropathy and membranous glomerulonephritis (31,32). Compelling evidence for a novel chain network is the absence of the novel chains in BMs from males patients with X-linked Alport syndrome and the presence of mosaicism in affected females, whereas al(1V) and aZ(1V) are normally present (14-17). Alport antigen is detected in all specialized BMs that contain a3(IV) and a4(IV), suggesting that the a3(IV)-Alport antigen heterodimer contributes to the assembly of the novel network. However, the reverse is not true in that certain BM (e.g. epidermal basement membrane) do not contain a3(IV) or a4(IV), suggesting the likelihood that the Alport chain interacts with itself as a homotrimer or participates with other chains as a heterotrimer.
Here, we have demonstrated that hexamers of NC1 from BM of three different bovine tissues (GBM, LC, and DM) can be separated into two populations by specific monoclonal antibodies. Hexamers containing NC1 of novel chains can be isolated from pooled NC1 free of al(1V) and a2(IV). Further, dimers precipitated with antibodies specific for a3(IV) or the Alport antigen do not contain subunits recognized by antial(1V) NC1. The reverse is also true since dimers precipitated with anti-al(1V) NC1 fail to react with antibodies to novel collagen. We conclude that al(1V) and a2(IV) do not appear to incorporate novel collagen chains into their helical structures and that classical and novel chain NC1 domains do not appear to interact to form COOH-terminal cross-links in the BM, supporting the concept of separate networks of classical and novel type IV collagen. Because precipitating antibodies to a2(IV) and a4(IV) are not available, we cannot rule out the possibility that unique hexamers exist which incorporate the a4(IV) chain with the classical chains of type IV collagen (al(1V) and a2(IV)). The function of the novel network in specialized BM is currently unknown, although the absence of these chains in Alport nephritis suggest an important role in maintenance of normal renal function and integrity. The mechanism of the progressive development of renal failure when these chains are missing has not been determined.
It is of interest that the BMs of the anterior cornea and lens form hexamers composed primarily of monomers whereas the hexamers from GBM are predominantly disulfide-bonded dimers. These structural differences may be related to varied rates of BM turnover or multiple functional roles of collagen that are tissue-specific. It is important to note that all dimers including the heterodimers are detected in the LC and DM, although in low abundance.
The a3(IV)-Alport antigen heterodimer is likely a critical assembly step in the formation of the novel collagen network and is the only heterodimer known to exist in the human NC1. This disulfide bonding may occur between similar helices composed of novel collagen heterotrimers or between helices of novel collagen heterotrimers and Alport antigen homotrimers. Since it will be difficult to prove how many novel networks exist, we will assume that tissue-specific combinations are limited by the cells synthesizing these chains. Nonetheless, formation of this heterodimer has far-reaching implications with respect to Alport familial nephritis. It would appear that most of the a3(IV) NC1 is engaged in a disulfide linkage with the Alport antigen. We have shown that this also occurs in cultured bovine corneal endothelial cells: a reaction that may drive formation of the COOH-terminal cross-link in the network. Absence of the Alport antigen would interfere with formation of hexamers containing the novel heterotrimers, and the related helix as well. Furthermore, in BMs like the GBM, LC, and DM, where a3(IV) is integral to the novel collagen network, absence of a3(IV) itself could result in the Alport phenotype.
What is the role of a5(IV) and how does it relate to the Alport antigen? We have shown that the polyclonal anti-a5(IV) and anti-Alport (mAb A7) antibodies stain the same renal BMs and that neither bind the Alport GBM, indicating complete concordance in tissue reactivity.' In addition, Western blotting of NC1 with anti-a5(IV) showed a monomeric and dimeric subunit structure similar to that observed using antibodies to the Alport antigen. However, without biochemical proof that a5(IV) and the Alport antigen are identical, we cannot rule out the presence of a sixth chain of type IV collagen which is very homologous to a5(IV). It is possible that the genes encoding Alport antigen and the a5(IV) are both located at the Alport locus on the X-chromosome, perhaps in a head to head, bi-directional promoter arrangement as was found with the classical collagen chains (34). The a3(IV) and a4 (IV) genes may be similarly structured on autosomal chromosome 2 (35).
With respect to BM collagen nomenclature, we wish to note that the designation of the novel collagens as chains of type IV collagen marks the first instance in which collagen molecules which do not form helices or known interactions together have been grouped in the same collagen family, but this information was not available when the chain nomenclature was designated. Homology of the known NC1 amino acid sequences range from 60 to 83% and cysteine residues are in perfect alignment, which is compelling evidence for their relatedness (22,36,37). However, the demonstrated difference between the classical and novel collagen IV molecules with respect to ontogeny, tissue localization, alterations in disease, and chain interactions, all provide strong support for two distinct BM collagen networks. Previous studies of the distribution of these components in the glomerulus suggest that each network is assembled independently by distinct and separate populations of cells (12,14,16,17,26,(30)(31)(32). In this regard, the classical collagen chains are formed by mesangial/ subendothelial cells whereas the novel network is derived from the visceral epithelium. These differences are amplified in both hereditary and acquired disease.