Structural Heterogeneity of the Noncollagenous Domain of Basement Membrane Collagen *

The noncollagenous domain of collagen from three different basement membranes of bovine origin (glomerular, lens capsule, and placental) was excised with bacterial collagenase, purified under nondenaturing conditions, and characterized. In each case the domain existed as a hexamer comprised of four distinct subunits (al(IV)NCl, aZ(IV)NCl, M2*, and M3). Each subunit exists in both monomeric and dimeric (disulfide-cross-linked) forms.  Certain  dimers  also exist which contain nonreducible cross-links. The hexamers from the three membranes differ with respect to stoichiometry of subunits and subunit isoforms and to the degree of cross-linking of monomers into dimers. The minor subunits, M2* and M3, vary in quantity over a 20-fold range relative to the major ones among the three hexamers. The results indicate that: 1) at least two populations of triple-helical collagen molecules, differing in chain composition, exist in each membrane and that their relative proportions are tissue-specific; and 2) the chemical nature of the noncollagenous domain of these populations is tissue-specific with regard to subunit isoforms and relative proportion of reducible and nonreducible cross-links in dimers. A novel structural feature of the noncollagenous domain of basement membrane collagen was also evinced from these studies. Namely, that each of the four monomeric subunits exists in charge isoforms.

amounts (4,5) and chemical nature of their known macromolecular constituents (4-ll), features which may be of fundamental importance in conferring the diverse functions.
The molecular properties of collagen IV, the major constituent of mammalian basement membranes, are of interest from the standpoints of structure-function relationships and their role in diseases. Collagen IV interacts with laminin, heparan sulfate proteoglycan, and fibronectin and it is proposed to serve as a scaffold for the proper organization of these constituents in basement membranes (12,13). The noncollagenous (NC1) domain of collagen IV is of particular importance because it is a critical site for cross-linking two triple-chain collagen molecules (14)(15)(16)(17) and it seems to be important for the lateral assembly of these molecules to form networks (18). Moreover, it contains the structural epitope which reacts with autoantibodies from patients with Goodpasture syndrome (19).
The NC1 domain is released from collagen IV as a hexamer, composed of monomeric and dimeric subunits, upon digestion of membrane with bacterial collagenase (19)(20)(21). In GBM,' three different subunits (Ml, M2*, and M3) were identified by chemical and immunochemical techniques (22). Each occurs in monomer and disulfide-linked forms. The GP epitope is exclusively localized to M2* and is sequestered under nondenaturing conditions.
The collagen chain origins of these subunits from LBM were recently determined (23). M1 comprises two polypeptides, designated al(1V)NCl and aS(IV)NCl, which correspond to the noncollagenous segments of the a l and a2 chains of collagen IV, respectively. M2* and M3 have physicochemical properties remarkably similar to those of al(1V)NCl and a2(IV)NC1 but their amino acid sequences differ. Each have Gly-X-Y triplets and hydroxyproline at their amino terminus, reflecting that each has a collagen chain origin, designated a 3 and a4, respectively. These new chains may be variants of the al(1V) and a2(IV) chains in which the NC1 segments are modified, or they may be entirely new chains with distinctive collagenous and noncollagenous sequences (23).
Earlier studies (24) suggested that the absolute amount of the GP antigen, now designated as M2*, varies among basement membrane preparations from different tissues (glomerulus, lung, and placenta). Such differences could reflect: 1) the presence of other contaminating connective-tissue elements in preparations from lung and placenta as compared to glomerulus; or 2) a stoichiometric difference in the subunit The abbreviations used are: GBM, glomerular basement membrane; LBM, anterior lens capsule basement membrane; PBM, placenta basement membrane; BM, basement membrane, GP, Goodpasture; SDS-PAGE, sodium dodecyl sulfate-polyacrylamide gel electrophoresis; HEPES, 4-(2-hydroxyethyl)-l-piperazineethanesulfonic acid; HPLC, high pressure liquid chromatography; ELISA, enzymelinked immunosorbent assay. composition of the hexamer form of the noncollagenous domain of basement membrane collagen.
The purpose of the present study was to determine whether the subunit composition and other properties of the hexamer varies among different basement membranes. This was accomplished by purifying the hexamer under nondenaturing conditions from three different basement membranes of bovine origin and comparing their molecular properties. The results indicate that the chemical nature of the noncollagenous domain and the relative proportions of at least two distinct populations of triple-helical collagen molecules are specific for a basement membrane of a given tissue.

EXPERIMENTAL PROCEDURES
Materials-Bovine kidneys were collected as described previously (25). Bovine lenses were obtained from Pel-Freeze Biologicals. Bovine placenta (5-6 months of gestation) was obtained from a local slaughterhouse, transported on ice, and immediately used. The following fine biochemical products are specifically mentioned together with the supplier: bacterial collagenase (CLSPA) from Worthington, DE-52 cellulose from Whatman, Sephacryl S-200 and S-300 from Pharmacia LKB Biotechnology Inc. Cle columns (201TP, 10-micron) for reverse-phase HPLC from Vydac, and a TSK SW 3000 column (600 mm length) for gel filtration HPLC from Varian.
Basement Membrane Preparation-Basement membrane preparations were carried out in the presence of protease inhibitors and at 0-4 "C if not further specified. GBM was prepared as described previously (25). Bovine anterior LBM was prepared as described by Peczon et al. (26), using sonication in the presence of 1 M NaCl and protease inhibitors for cell removal. To prepare PBM, bovine placenta was freed from vessels, ground in a meat grinder, and washed until free from blood with 0.05 M Tris-HC1, pH 7.5. The insoluble material was then extracted at 37 "C for 24 h in 6 M guanidine HCl, 0.05 M Tris-HC1, pH 7.5. The residue was enriched in basement membrane and used as the PBM sample.
Collagenase Digestion-To solubilize the noncollagenous domain of collagen IV, basement membrane preparations were digested with bacterial collagenase at 37 "C for 20 h in the following digestion buffer: 0.05 M HEPES, pH 7.5,O.Ol M CaC12, 4 mM N-ethylmaleimide, 1 mM phenylmethanesulfonyl fluoride, 5 mM benzamidine HCl, 25 mM 6-aminohexanoic acid. For digestion of GBM, 1 g of dry membrane was thoroughly dispersed in 100 ml of digestion buffer with a Polytron tissue disrupter, and then 2 mg of collagenase was added. LBM from 100 anterior lens capsules was directly incubated in 20 ml of digestion buffer with 0.5 mg of collagenase. PBM was digested in 500 ml of digestion buffer with 2.5 mg of collagenase.
Chromatographic Procedures-Following treatment with collagenase, purification of the noncollagenous domain (hexamer) was carried out under associative conditions essentially as described before (21) using anion-exchange chromatography on DEAE-cellulose at pH 7.5 and 9 and gel filtration, consecutively, with columns of Sephacryl S-300 and S-200. In the case of LBM, the chromatography with DEAEcellulose at pH 9 and the gel filtration with Sephacryl S-200 were omitted in later experiments because this apparently led to the same final product as observed with the different biochemical tests used for characterizing the hexamer.
Purified hexamers were studied under dissociative conditions on gel filtration columns of TSK SW 3000 and Sephacryl S-200, using 6 M guanidine HCl, 0.05 M Tris-HC1, pH 7.5, for elution. To separate subunits by their hydrophobic properties, reverse-phase HPLC was performed on a Cla column as described previously (22). The amount of protein was estimated by absorbance measurements at 280 and forms of d(IV)NCl, a2(IV)NCl, M2*, and M3 were purified from 230 nm. For identifying spots in two-dimensional gels, monomeric LBM as described (23). Electron Microscopy-Electron microscopy using the rotary-shadowing technique was performed as described by Shotton et al. (27). Samples from gel filtration columns in 0.05 M Tris-HC1, pH 7.5, were diluted with 0.2 M ammonium bicarbonate, pH 7.5, to 50 pg/ml, mixed with an equal volume of 100% glycerol, then sprayed onto freshly cleaved mica sheets and rotary-shadowed with platinum at 9" followed by carbon at 90". Replicas were examined as described previously (21). For comparable measurements, samples were shadowed together on the rotary table to obtain similar amounts of platinum deposition on the proteins. A carbon replica with waffle pattern (Pelco) was used to determine actual magnification. To compensate for a possible polarity in spreading of globules that were oblong, particle diameters were measured in two fixed directions at right angles to each other.
Antisera and Immunochemical Technique-The antisera used were either from patients with GP syndrome, verified by immunofluorescence, or from rabbits immunized with the monomeric subunits of the globular domain of collagen IV from bovine GBM, namely M1, M2*, and M3, raised as described previously (22).
Competition ELISA was performed as described previously (19, 21,22). Coating of antigens was done overnight at 22 "C, either under associative conditions in 0.05 M Tris-HC1, pH 7.5, 0.15 M NaCl, or under dissociative conditions in 6 M guanidine HCl, 0.05 M Tris-HC1, pH 7.5. Samples analyzed under associative conditions were mixed with the antisera, using the incubation buffer (0.05 M phosphate, pH 7.5,0.15 M NaC1, 0.05% Tween 20, 0.2% bovine serum albumin) for dilution. Samples analyzed under dissociative conditions were first diluted in 6 M guanidine HCl, 0.05 M Tris-HC1, pH 7.5, followed by heating for 5 min in boiling water. The samples were then diluted 10 times or more directly with the incubation buffer containing the antisera. These sample-antibody mixtures were left to standovernight at 4 "C. The remaining steps were performed as described previously (21).
Incubations of Western blots with the antisera were carried out by the method previously used (22) employing a metal-enhanced diaminobenzidine reaction for visualization of anti-immunoglobulin-peroxidase conjugates (28).
Electrophoresis Techniques-SDS-PAGE was performed with 1.5mm thick slab gels of linear gradients of 6-22 or 10-22% polyacrylamide for one and two-dimensional analysis, respectively (23,29). The amount of hexamer applied on the gels was based on absorbance at 280 nm, considering an absorption coefficient of 2184 g" cm", in which the mass value was derived from amino acid analysis of the hexamer from GBM. The value is close to that for the hexamer from Engelbreth-Holm-Swarm tumor (20). Spectrophotometric scanning of Coomassie Blue-stained gels for quantitation of nonreducible dimers was carried out as described previously (30).
In two-dimensional gel electrophoresis, nonequilibrium pH gradient gel electrophoresis as the first dimension was conducted according to O'Farrell et al. (31) with the following modifications. Tube gels of 1.5 mm thickness and 11 cm length consisted of 4% polyacrylamide (5% cross-linker), 2 M urea, 2% Nonidet P-40, 20% glycerol, and 2% ampholine mixture (LKB, equal volumes of pH 5-8 and 7-9). Samples contained 5 pg of hexamer, 1 M urea, 20% glycerol, 2% ampholines as above, 25 mM @-alanine, and 25 mM 6-aminohexanoic acid, and marker proteins as reference for migration. After application, the samples were covered with overlay solution composed of 10% glycerol and 1% of the above ampholines. Electrophoresis was carried out at 8 "C for 3000 V-h. To check the pH gradient, 5-mm pieces were cut from a tube gel run without protein, incubated in 0.5 ml of degassed distilled water for 12 h, and then measured. Tube gels for the second dimension were incubated for 10-min periods, twice in 50% methanol and then twice in sample buffer (32). When layering a tube gel on top of the second dimension gel, an agarose plug containing molecular weight markers for the second dimension was added at each end. Marker proteins are useful in establishing the pH gradient developed in the first dimension because in nonequilibrium pH gradient gel electrophoresis, proteins, especially alkaline ones, do not reach their PI in this system. Marker proteins when run in the second dimension are helpful in establishing the quality of migration of proteins through the gel and aid in identifying the protein spots originating from the hexamers.
To prepare Western blots, proteins separated by SDS-PAGE or two-dimensional gel electrophoresis were electrophoretically transferred to nitrocellulose papers as described (33).

General Properties of the NCl Domain (Hexamer) of BM Col@en from Different
Tissues-Analysis of collagenase digests of GBM, PBM, and LBM by SDS-PAGE indicated that the subunit structure of the NC1 domains is very different with respect to relative proportions of subunits and quantity of GP epitope. This observation poses basic questions about variations in the structural organization of BM-collagen in relation to tissue location and function. Therefore, a more detailed study was undertaken to elucidate the structural differences using purified NCl domains (hexamers) obtained from two mature basement membranes (GBM and LBM), of vascular and of avascular origin, respectively, and from a developing basement membrane-rich tissue, placenta (PBM).
The hexamer from bovine GBM, LBM, and PBM was excised by bacterial collagenase and purified, under nondenaturing conditions, by sequential fractionation on columns of DE52, pH 7.5, DE52, pH 9.0, and Sephacryl S-300 as described for GBM (21). LBM and PBM gave identical results on the Sephacryl S-300 column to that of GBM. Namely, pool I contained 7s collagen, pool I1 contained hexamer with M, = 160,000, and pool I11 contained polypeptides with the same mobilities, on SDS-PAGE, as those in pool 11, indicating that they exist in a species of smaller size than that of the hexamer (data not shown). Pool I1 from GBM contained about 90% of the total material present in pools I1 and I11 as measured by absorbance at 280 nm, whereas for pool I1 from LBM and PBM this value amounted to 62 and 93%, respectively.
Comparison of the hexamers from LBM and PBM, by inhibition ELISA under nondenaturing conditions, revealed that the GP epitope is sequestered like that found in GBM (21). In each case, only very low levels of GP antibodies bind the GP epitope. Pretreatment of the hexamer with 6 M guanidine HC1 at 100 "C causes a 20-40-fold increase in binding of GP antibody (data not shown), while the amount of GP antigen in this pool from the three basement membranes varied in decreasing order of GBM, LBM, and PBM.
Electron micrographs also revealed similarities and distinct differences in size and shape of the hexamers from the three tissues (Fig. 1). The preparations from GBM and PBM appeared homogenous with respect to size and spherical character. However, the sample from LBM differed from both of these two aspects. Firstly, the range of particle sizes in the sample from LBM (8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19) was about twice that from GBM and PBM (12)(13)(14)(15)(16)(17)(18). The amount of particles in these ranges accounted for 95% of the total number in each population (Fig. 1, histograms). Secondly, two different shapes of particles were observed in the sample from LBM: spherical (diameter range: 8-17 nm) and ellipsoid (range of diameters of the long axis: [16][17][18][19][20][21][22]. The latter often appeared as closely associated pairs of spherical particles, possibly indicating partial dissociation. The diameter of the spherical particles when measured perpendicular to the long axis ranged from 8 to 10.6 nm. The dissociation phenomenon most likely occurs   Values are percentage of protein in dimer peak to total protein in monomer and dimer peak as obtained by gel filtration of hexamers on a TSK S W 3000 column as described in Fig. 4. Protein amounts are based on absorbance at 230 nm. Identical results were obtained when gel filtration was carried out on Sephacryl S-200 in the presence of 6 M guanidine HCI. Values were calculated from the relative areas presented in Fig.  7. resulted in a single, symmetrical peak with an elution position corresponding to M, = 160,000, and inhibition ELISA (see above) demonstrated that the GP epitope was sequestered.
The hexamers from GBM, LBM, and PBM are similar with respect to their banding pattern in the monomer (25-30 kDa) and dimer (43-53 kDa) regions, as shown by SDS-PAGE analyses, but they are dissimilar in their monomer/dimer ratios (Fig. 2). The hexamers from GBM and PBM consist mainly of dimer-size components, whereas, that of LBM consists primarily of monomer-size components. Also, the relative ratio of components in both regions differs among the three tissue sources. Most evident is the difference between components in the dimer-size region. In this region, GBM and PBM contain two intensely staining dimer bands (48.9 and 42.9 kDa) and different amounts of other weakly staining dimers. In LBM the staining of the dimer at 48.9 kDa is more intense than the one at 42.9 kDa, while other weakly staining dimers are also present.
The amino acid composition of the hexamers from GBM, LBM, and PBM is comparable to those reported for monomeric and dimeric subunits (22). Therefore, although there are major differences in the relative amounts of monomer and dimer species which comprise these globules (results are presented below), there are no striking differences in their amino acid composition. Identification of Subunits-Immunoblotting after SDS-PAGE was used to determine whether the subunits M1, which comprises the al(1V)NCl and a2(IV)NCl domain (23), M2* (GP antigen), M3, and their corresponding dimers Dl, D2*, and D3 are constituents of the hexamer from LBM and PBM as was described previously for that from GBM. The results are summarized in Fig. 3, which shows blots of the three hexamers immunostained with antibodies to M1, M2*, and M3 and with GP serum.
In each hexamer, antibodies to subunits M1, M2*, and M3 reveal the presence of all three subunit species in both monomer and dimer forms. With anti-M1 antibodies, the amount of reactivity in the dimer region increases and at the same time in the monomer region decreases in the sequence LBM, GBM, and PBM (Fig. 3). Anti-M2* and anti-M3 antibodies react in the monomer and dimer regions of the GBM and LBM hexamers, with the highest staining in the dimer region in the case of GBM, and in the monomer region in the case of LBM. Anti-M3 antibodies, which had not been characterized previously, permitted the identification of D3. D3 consists of a set of polypeptides with mobilities distinct from Dl; however, it was not determined whether D3 polypeptides are distinct from D2*. Reactivity of the hexamer from PBM was weak with anti-M2* and anti-M3 antibodies. For controls, purified monomerss MI, M2*, and M3 were run in separate lanes; immunoblots further substantiated the identity of the subunits of the three hexamers and the absence of crossreactivity of the three antibodies (not shown).
The presence of the GP antigen in the hexamers of LBM and PBM, analogous to GBM, was revealed by Western blotting using patients' sera (Fig. 3). The staining patterns show that for each hexamer the GP epitope is contained in subunit M2*, which is present in both monomer and dimer forms. It is particularly noteworthy that the staining intensity with GP sera is analogous to that of anti-M2*, reflecting a higher concentration of the M2* chain in GBM than in LBM or PBM.
These results indicate that the hexamers from GBM, LBM, and PBM are composed of identical monomer and dimer constituents, although they greatly differ in the amounts of monomers relative to dimers for each of the three monomerdimer pairs, and in the absolute amounts of M2*, D2*, M3, and D3 subunits. The basis for these distinct differences was further explored by quantitative analysis.
Quuntitation of Subunits-To determine the relative amounts of monomers and dimers in the three hexamers, samples were heated to 90-95 "C for 10 min in the presence of 6 M guanidine HCl to obtain complete dissociation, followed by gel filtration, which separates monomers from dimers (Fig.  4). The results indicate that the percentage of subunits in dimer form varies over a 5-fold range among the three hexamers in the increasing order of LBM, GBM, and PBM (Table  I).
The relative amounts of subunits M1, M2*, and M3, present in both their monomer and dimer forms, were determined in each of the three hexamer preparations on a chemical basis by reverse-phase HPLC. As shown in Fig. 5, these species resolve in the following sequence: M1 in pool 1 and pool 2, M3 in pool 3, and M2* in pool 4. The relative amount of each subunit in the three tissuess was calculated from the relative B~e~e n t ~e~b r a~e Co~~agen areas from the elution profile, and the data are presented in Table I. The major subunit is M1 for each hexamer, while the total amounts of M2* and M3 vary over a 20-fold range among the three hexamers in the decreasing order of GBM, LBM, and PBM. Similar results were obtained using rabbit antibodies specific for the various subunits and GP serum in competitive ELISA (Fig. 6).
The relative amounts of dimers present in nonreducible and reducible forms were also determined for GBM, LBM, and PBM. Previous studies have shown that dimers are held together by both disulfide and nondisulfide cross-links (20,22). The purified dimers from each of these tissues were reduced and analyzed by SDS-PAGE. The amount of nonreducible dimer was calculated from the relative areas of the profile (Fig. 7), and the data are presented in Table I. As noted the amount of nonreducible dimer varies over a 2-fold range with the highest amount in PBM and the lowest in LBM.
Multiple Charge Forms of Subunits-The hexamer from each of the three basement membranes displays a complex pattern on analysis by nonequilibrium pH gradient gel electrophoresis and SDS-PAGE in a two-dimensional gel system  -M3 (uM3), and serum from a patient with GP syndrome (GP). For coating, the hexamer from GBM was used. In the assays for M2*, M3, and GP the hexamer was coated in the presence of 6 M guanidine HCl, but in the assay for M1, associative conditions were used (see "Experimental Procedures"). For competition, the hexamers from GBM, LBM, and PBM were first heated for 5 min in the presence of 6 M guanidine HCl in a bath with boiling water and then diluted with the antisera, except for the assay with anti-Ml antibodies for which the antigens appeared already fully exposed without activation in guanidine HCI. found under their respective peaks. As a control, the relative amounts of the tu1 and a2 chains of type I collagen, which exist in a 2 to 1 ratio, respectively, were determined by this technique. The values obtained were 68 and 32%, respectively.
( Fig. 8). Multiple spots exist in both the monomer and dimer regions. There is a striking commona~ity among the three patterns, as depicted in Fig. 7 0 . In the monomer region (Mr = 25,000) there are six major and five minor spots in common, and in the dimer region there are at least eight major ones, distributed about the 48.9-and 42.9-kDa positions. The identity of the various spots in the monomer region was determined by two-dimensional gel analysis of purified subunits. This result is also depicted in Fig. 80. The three main spots at pH 7-9 correspond to al(1V)NCl monomer, and the three at pH 6-7 correspond to a2(IV)NC1 monomer, In addition, M2* and M3 occur at least as two and three spots, respectively. These observations were further confirmed with immunoblots (not shown) of two-dimensional gels of whole hexamers and the isolated monomers as well, using GP serum and specific antisera against M1, M2*, and M3, prepared as described previously (22). The multiple spots for each monomer are designated as charge isoforms, because the unresolved forms of each subunit yield a single amino terminus (23). Furthermore, this designation is substantiated by the finding that up to 5% bacterial collagenase (enzyme/ substrate ratio) for 24 or 48 h did not alter the two dimensional profile, which rules out incomplete digestion as a basis for multiple forms. Of particular note, subunits al(1V)NCl and a2(IV)NCl, which comprise M1, are resolved by the twodimensional system. Several distinct differences exist in the monomer region of the two-dimensional patterns of the three hexamers (Fig. 8,  A-C). The most alkaline of the three isoforms of the al(1V)NCl monomer is the most prominent one in GBM, whereas the most acidic one is prominent in PBM, and an intermediate distribution occurs in LBM. The relative intensity of the a2(IV)NC1 isoforms is similar for both GBM and LBM, but PBM is richer in the more acidic one. The ml(1V)NCl and a2(IV)NCl monomers are the predominant ones for each of the three hexamers. In comparison to these, the concentrations of M2* and M3 are largest in GBM and decrease in the order GBM, LBM, and PBM. In PBM these latter constituents are barely visible (Fig. 8C). These results confirm those presented above regarding the relative abundance of monomers. Several distinct differences also exist in the dimers at the region of pH 7-9. The dimers of GBM and PBM show similar intensities at the 48.9-and 42.9-kDa positions, but with LBM the ones at 48.9 kDa are more prominent. The most alkaline dimers are more enriched in GBM than in PBM.
In summary, each of the four monomer subunits exists in charge isoforms, and the relative abundance of isoforms for the respective monomer subunits varies among the membranes. Subunits al(IV)NCl, a2(IV)NC1, and M3 exist in at least three isoforms, and subunit M2* exists in at least two. This large diversity of monomer forms accounts for the multiplicity of cross-linked dimers that are observed with the two-dimensional analysis.

DISCUSSION
The present study reveals similarities and distinct differences in the structural features of the noncollagenous domain of BM collagen from three different basement membranes of bovine origin (GBM, LBM, and PBM). In each case, after excision by bacterial collagenase, the domain exists in the form of a hexamer under nondenaturing conditions, and it is comprised of four distinct subunits, each of which exists in both monomeric and dimeric forms. The hexamers from the three membranes differ with respect to the stoichiometry of subunits and subunit isoforms, the degree of cross-linking of monomers into dimers, and the relative proportion of nonreducible and reducible cross-links.
Specifically, the hexamers differ in the following ways: 1) the minor subunits, M2* and M3, vary in quantity over a 20fold range relative to the major ones, al(1V)NCl and aZ(IV)NCl, in the decreasing order of GBM, LBM, and PBM; 2) the distribution of charge isoforms of al(1V)NCl and a2(IV)NC1 varies among the membranes with PBM containing the greatest amount of acidic ones, whereas GBM is enriched in the alkaline form of al(1V)NCl; 3) the percentage  " A denotes a population of triple-helical molecules comprised exclusively of al(IV) and a2(IV) chains; B denotes a population(s) comprised exclusively of a3 and a 4 chains; and C denotes a population(s) comprised of al, a2, a3, or a4.
The range of the value of A is 100 -B to 100 -C.
of subunits in dimer form, reflecting the degree of interchain cross-linking, varies over a 5-fold range in the decreasing order of PBM, GBM, and LBM. The stoichiometry of hexamer subunits together with our recent identification of their collagen-chain origins (23) lead to the conclusions that at least two different populations of triple-helical collagen molecules, differing in chain composition, exist in each membrane and that their relative proportions are tissue-specific. The predominant subunits, al(1V)NCl and a2(IV)NCl, are derived from the a1 and a2 chains of the classical collagen IV molecule, which appears to have a chain composition of (a1)2,a2 (34-37), denoted herein as population A. Subunits M2* and M3, which occur in minor amounts, are derived from two novel chains, a 3 and a4, respectively (23), which could exclusively comprise a separate triple-helical molecule(s), denoted as population(s) B. Alternatively, the a 3 and a4 chains could substitute for either the a1 or a2 chain in the collagen IV triple-helical molecule, denoted as population(s) C. The theoretical proportions of these populations, computed from the stoichiometric data (Table I), are presented in TableII.These computations show that the relative proportions of populations are tissue-specific.
A novel structural feature of the noncollagenous domain was also evinced from these studies. Namely, that each of the four monomeric subunits exists in charge isoforms and that the relative proportions of isoforms are tissue-specific. Monomers al(IV)NCl, aB(IV)NCl, and M3 exist in at least three isoforms, and M2* exists in at least two. The presence of four distinct monomers and their respective isoforms accounts for the multiplicity of disulfide-cross-linked dimers that are observed with the two-dimensional analyses. The identification of isoforms also provides an explanation for the complex twodimensional gel patterns observed by others (38, 39). The different isoforms presumably reflect amino acid substitutions or posttranslational modifications, a feature which may be an important structural determinant for the linear and lateral assembly of collagen molecules in the formation of the matrix network.
It is especially noteworthy that the amount of both intermolecular disulfide and nonreducible cross-links is very low in the LBM hexamer in contrast to that of GBM, PBM, and the hexamer from mouse Engelbreth-Holm-Swarm tumor (20). This property may account for the presence of the ellipsoid-shaped particle in the LBM hexamer ( Fig. 1) in which the absence of such cross-links would destabilize the hexamer under the conditions used for electron microscopy. The low level of cross-linking in the domain, however, indicates that such bonding is not essential for stabilization of the collagen IV framework of LBM and poses questions regarding the role and pathways of biosynthesis of these crosslinks in other basement membranes.
The present study provides the first direct evidence of tissue specificity in the chemical nature of basement membrane collagen. This specificity provides an explanation for differences in staining among basement membranes of different tissues using either GP-sreum or antibodies to specific regions of collagen IV (40-44). Conceivably, certain of these structural differences may be of importance in conferring a specific function to a membrane.