Glomerular Basement Membrane

Glomerular basement membrane (GBM) plays a crucial function in the ultrafiltration of blood plasma by the kidney. This function is impaired in Alport syndrome, a hereditary disorder that is caused by mutations in the gene encoding type IV collagen, but it is not known how the mutations lead to a defective GBM. In the present study, the supramolecular organization of type IV collagen of GBM was investigated. This was accomplished by using pseudolysin (EC3.4.24.26) digestion to excise truncated triple-helical protomers for structural studies. Two distinct sets of truncated protomers were solubilized, one at 4 °C and the other at 25 °C, and their chain composition was determined by use of monoclonal antibodies. The 4 °C protomers comprise the α1(IV) and α2(IV) chains, whereas the 25 °C protomers comprised mainly α3(IV), α4(IV), and α5(IV) chains along with some α1(IV) and α2(IV) chains. The structure of the 25 °C protomers was examined by electron microscopy and was found to be characterized by a network containing loops and supercoiled triple helices, which are stabilized by disulfide cross-links between α3(IV), α4(IV), and α5(IV) chains. These results establish a conceptual framework to explain several features of the GBM abnormalities of Alport syndrome. In particular, the α3(IV)·α4(IV)·α5(IV) network, involving a covalent linkage between these chains, suggests a molecular basis for the conundrum in which mutations in the gene encoding the α5(IV) chain cause defective assembly of not only α5(IV) chain but also the α3(IV) and α4(IV) chains in the GBM of patients with Alport syndrome.

Basement membranes are specialized extracellular matrices that compartmentalize tissues, provide substrata for cells, and provide signals for differentiation, maintenance and remodeling of tissues. Prominent basement membranes are the lens capsule (LBM) 1 of eye, seminiferous tubule basement mem-brane (STBM) of testis, and glomerular basement membrane (GBM) of kidney glomerulus. GBM plays a crucial function in the normal purification of blood plasma, and its function is impaired in at least three prominent renal diseases: diabetic nephropathy, Goodpasture syndrome, and Alport syndrome. In Goodpasture syndrome, a rapidly progressive autoimmune disorder, the target autoantigen is type IV collagen protein, whereas in Alport syndrome, a hereditary form of progressive renal disease, the underlying molecular defects are mutations in the genes encoding type IV collagen (1).
In Alport syndrome, mutations occur in the COL4A3, COL4A4, and COL4A5 genes encoding the ␣3(IV), ␣4(IV), and ␣5(IV) chains. About 80% of the affected families exhibit Xlinked inheritance of mutations in the COL4A5 gene (13,14), while some of the remainder inherit autosomal recessive mutations in ␣3(IV) and ␣4(IV) chains (15,16). To date, over 200 mutations have been found in the COL4A5 gene (13,14,17,18). However, the molecular mechanisms by which gene mutations in type IV collagen cause defects in GBM are unknown. The abnormal GBM in most patients with X-linked Alport syndrome contains only the ␣1(IV) and ␣2(IV) chains and is devoid of the ␣3(IV), ␣4(IV), and ␣5(IV) chains (5, 8, 19 -21). A recent study has revealed that this phenomenon reflects an arrest of an early developmental switch, wherein the ␣1(IV) and ␣2(IV) chains persist and are not replaced by the ␣3(IV), ␣4(IV), and ␣5(IV) chains that are necessary to form a mature GBM (22). How mutations in the ␣5(IV) chain cause defective assembly of not only the ␣5(IV) chain, but also the ␣3(IV) and ␣4(IV) chains, remains a conundrum. The mechanism linking the assembly of these three chains could be at the protein level in which the ␣3(IV), ␣4(IV), and ␣5(IV) chains are structurally linked into a supramolecular network (5,22), either at the level of triple-helical protomers or at the linkages between triplehelical protomers. The linkage mechanism may also involve the transcriptional/translational level, because the mRNA expression of the ␣3(IV), ␣4(IV), and ␣5(IV) chains appears to be coordinated (22).
The purpose of the present study was to characterize the supramolecular organization of type IV collagen chains in GBM. This was accomplished using pseudolysin cleavage to excise soluble triple-helical molecules for determination of their chain organization. The results revealed that the ␣3(IV), ␣4(IV), and ␣5(IV) chains exist in a novel supramolecular network that is cross-linked by disulfide bonds between triple helices. Moreover, the findings establish a structural linkage between the ␣3(IV), ␣4(IV), and ␣5(IV) chains in GBM, suggesting a molecular basis for the conundrum in which COL4A5 gene mutations in Alport syndrome cause defective assembly of not only the ␣5(IV) chain, but also the ␣3(IV) and ␣4(IV) chains.
Preparation of 4°C Truncated Protomers from Bovine GBM-One gram of GBM was suspended in 100 ml of 150 mM NaCl, 50 mM Tris-HCl, 2 mM CaCl 2 , pH 7.5. To this, 10 mg of pseudolysin was added and digestion was carried out at 4°C for 36 h. The suspension was then centrifuged at 10,000 rpm (16,000 ϫ g) in a Sorvall HB-4 rotor to remove un-solubilized GBM. The supernatant contained most of the ␣1(IV) and ␣2(IV) chains and the pellet contained primarily ␣3(IV), ␣4(IV), and ␣5(IV) chains. The supernatant solution was brought to 20 mM EDTA, and the digested collagen IV was precipitated by increasing the NaCl concentration to 2 M (23). After 4 h, the precipitate was sedimented at 10,000 rpm (16,000 ϫ g) in a Sorvall HB-4 rotor for 25 min. The pellet was dissolved in 75 ml of 150 mM NaCl, 50 mM Tris-HCl, pH 7.5. The residual pseudolysin was removed by filtration through a Sephacryl S-1000 column that had been equilibrated against 150 mM NaCl, 50 mM Tris-HCl, pH 7.5. The fractions free of pseudolysin were pooled and used for further experiments. The 4°C protomers accounted for 20% of the weight of GBM, based upon absorbance measurements at 280 nm of Sephacryl S-1000 fractions, using an absorbance of 0.38 for a 1 mg/ml in a cell of 1 cm pathlength. The value 0.38 was calculated from the amino acid composition of type IV collagen (24). GBM contains ϳ40% by weight of type IV collagen (25); thus, the 4°C protomers comprise ϳ50% of the type IV collagen of GBM.
Preparation of 25°C Truncated Protomers from Bovine GBM-The pellet from the 4°C pseudolysin digestion was washed three times with cold water, once with 150 mM NaCl, 50 mM Tris-HCl, pH 7.5, and then suspended in 50 ml of the same buffer. The suspension was made 0.5% in pseudolysin and 2 mM in CaCl 2 , and digestion was carried for 24 h at room temperature (25°C). The reaction was arrested by making the digest 20 mM in EDTA. The suspension was centrifuged for 25 min at 10,000 rpm (16,000 ϫ g). Collagen IV fragments were precipitated by adding NaCl to the cold (4°C) supernatant solution (final concentration 2 M). After 4 h the precipitate was collected by centrifugation. The pellet was dissolved in 150 mM NaCl, 50 mM Tris-HCl, pH 7.5, and the collagen IV fragments precipitated once again with NaCl and centrifuged as described above. The pellet was suspended in 150 mM NaCl, 50 mM Tris-HCl, pH 7.5, 2 mM EDTA and filtered through Sephacryl S-1000 to remove residual pseudolysin. The fractions free of pseudolysin were pooled and used for subsequent experiments. The solubilized 25°C protomers accounted for 18% of the weight of GBM, and ϳ45% of the type IV collagen of GBM, based upon absorbance measurements at 280 nm of Sephacryl S-1000 fractions, using an absorbance of 0.38 for a 1 mg/ml in a cell of 1 cm pathlength. These protomers required pseudolysin at 25°C for solubilization, as they remained insoluble after increasing the temperature to 25°C in the absence of pseudolysin. The protein that remained insoluble even after the 25°C pseudolysin digestion represented 5% of the total GBM, as determined by dry weight measurements, and had a type IV ␣-chain composition similar to that of the 25°C protomers.
Fractionation of 25°C Truncated Protomers under Denaturing Conditions-Truncated protomers obtained by 25°C pseudolysin digestion of GBM and fractionated through the Sephacryl S-1000 column were dialyzed against 6 M guanidine-HCl for 36 -48 h and then fractionated on a Sephacryl S-400 column that was equilibrated in the same solution. The fractions were pooled, concentrated using Amicon (M r cut-off, 30,000) and used for further experiments.
Analysis of NC1 Domains from 4°C and 25°C Truncated Protomers-The NC1 domains were released from truncated protomers by digestion with bacterial collagenase, and the NC1 hexamers were purified on either a Bio-Sil TSK-250 HPLC gel filtration column or a Sephacryl S-300 column (1.6 ϫ 50 cm) as described previously (26,27). Fifty mM Tris-HCl buffer, pH 7.5, containing 0.15 M NaCl and 0.05% NaN 3 , was used for equilibration and elution of the column. The NC1 hexamers were then analyzed by HPLC to quantitate the NC1 domains and by two-dimensional electrophoresis and SDS-PAGE to identify their ␣(IV) origins using previously established techniques (7).
Reduction and Alkylation of 25°C Truncated Protomers under Native Conditions-A sample from the pseudolysin digest that had been fractionated by filtration through a Sephacryl S-1000 column was dissolved in 0.5 M NaCl, 0.1 M Tris-HCl, 0.5% 2-mercaptoethanol (v/v), pH 8.3, at approximately 1 mg/ml protein concentration. The mixture was stirred for 24 h at room temperature. A 2-3-fold molar excess of sodium iodoacetate was then added, and the mixture was stirred for 4 h in the dark (28). The mixture was then centrifuged, and the supernatant was used for further experiments. The protomer contains three ␣(IV) chains and is triple-helical over most of its length. There are six genetically distinct ␣(IV) chains. The number of protomers that can be made in vivo is not known. Each chain contains a carboxyl-terminal noncollagenous NC1 and an amino-terminal 7 S domain; the latter contains a large Asn-linked oligosaccharide (Y-shaped structure). Protomers associate through NC1 domains and 7 S domains to form suprastructures. Digestion of basement membranes with pseudolysin yields truncated protomers that contain the NC1 domains and part of the triple helix.
Pepsin Digestion of 25°C Truncated Protomers-Pepsin digestion was performed as described by Miller and Rhodes (29). The pooled 25°C protomers from the Sephacryl S-1000 column were precipitated by making the solution 2 M in NaCl. The precipitate was collected by centrifugation, dissolved in 0.5 M acetic acid (5 mg of protein/ml and then dialyzed against 0.5 M acetic acid at 4°C. Pepsin was added at 1:10 pepsin-to-substrate ratio, and digestion was performed at 4°C for 24 h. The collagen IV products were purified by two precipitations with 2 M NaCl, 0.5 M acetic acid. The second precipitate was solubilized in 50 mM Tris-HCl, pH 7.5, dialyzed against 50 mM Tris-HCl, pH 7.5, and stored at 4°C until further use. Electrophoretic Techniques-Two-dimensional electrophoresis was performed as described previously but using 6 -22% gradient gels in the second dimension (27). In some cases, either 4 -22% or 5-15% gels were used for hexamer or large fragments, respectively. The approximate sizes of the fragments were determined as described previously (30). To determine approximate sizes of complex protomeric molecules, 1.5% agarose gel electrophoresis was performed as described previously (30) but using slab gels instead of cylindrical gels. Ascaris suum basement membrane collagen (31), calf skin collagen, and rabbit skeletal muscle myosin (30) were used as molecular weight standards.
Immunochemical Techniques-To prepare Western blots, the electrophoretically separated proteins were transferred to nitrocellulose, blocked with 2% bovine serum albumin, reacted with primary antibody, and stained with nitro blue tetrazolium and 5-bromo-4-chloro-3-indolyl phosphate substrates (32). Alkaline phosphatase conjugate was used as secondary antibody. Primary monoclonal antibodies directed against human ␣1(IV)-␣6(IV)NC1 domains were used for Western blots. The specificity of the monoclonal antibodies was previously established for both human and bovine NC1 domains (8,33) and they were used to determine the tissue distribution of ␣(IV) chains. About 5 g of protein was loaded in each lane in SDS-PAGE and 10 g for two-dimensional electrophoresis. Two-dimensional gels were stained with silver as described by Morrissey (34).
Chemical Analysis-Protein and hydroxyproline concentrations were determined by published procedures (35,36).
Electron Microscopy-Rotary-shadowing electron microscopy studies were performed on 4°C and 25°C truncated protomers using conditions (26) adapted from Shotton et al. (37). Samples (20 g/ml), in 25% glycerol containing either 150 mM ammonium bicarbonate or 50 mM acetic acid, were sprayed onto freshly cleaved mica sheets, and were rotary-shadowed with platinum at 9°followed by carbon at 90°. Replicas were examined with a JEOL JEM-100CX II electron microscope. The contour length of molecules was measured with a Summagraphics Data Tablet (model MM1201).

Solubilization of Truncated Protomers of Type IV Collagen by Pseudolysin Digestion of GBM
In our previous studies of STBM and LBM (7,38), digestion with pseudolysin at 4°C was established as a procedure to solubilize truncated protomers of type IV collagen ( Fig. 1) that retain a portion of the triple-helical domain and the complete NC1 carboxyl domain. Such truncated molecules can provide information about the chain composition of protomers and the organization and linkages of protomers in supramolecular networks. The retention of the NC1 domains on the truncated chains allows for the identification of the kind (␣1 to ␣6) of ␣(IV) chain using NC1-specific monoclonal antibodies. In the present study, truncated protomers were obtained under two separate conditions of pseudolysin digestion. First, the digestion was performed at 4°C, which solubilized a set of truncated protomers, and then a second digestion of the residual insoluble fraction was performed at 25°C, which solubilized another set of truncated protomers. The two populations of truncated protomers, designated 4°C and 25°C protomers, were characterized (see below) with respect to chain composition and supramolecular organization. The chain compositions were determined using the same chain-specific monoclonal antibodies that were used to determine the distribution of chains in renal tissue (8,33).
Gel Filtration Analysis-The elution profiles of the digestion mixtures of truncated protomers that were filtered through Sephacryl S-1000 are shown in Fig. 2. For both 4°C and 25°C protomers, the elution profiles are broad. However, in each case, SDS-PAGE analysis (data not shown) revealed that most of the fractions contained fragments of the same sizes upon reduction of disulfide bonds, indicating a large heterogeneity in size of the constituents of the disulfide-linked complexes. The major result of the fractionation was the removal of pseudolysin. Fractions 20 -120 were recombined, and the 4°C and 25°C protomers were analyzed by SDS-PAGE to determine the size and distribution of fragments that were dissociated by SDS (Fig. 2, inset). In the 4°C protomers, some fragments were retained in the stacking gel and those that were resolved by the gel exhibited apparent M r values of Ͻ16,000 to Ͼ200,000. The apparent M r of the most intense fragment was 160,000. In contrast, those of the 25°C protomers were of larger size, with a major fragment with an apparent M r Ͼ200,000.
The NC1 domains of the 4°C and 25°C protomers were also analyzed by HPLC and two-dimensional electrophoresis to identify and quantitate the kinds of chains (Fig. 4). The HPLC profile and two-dimensional electrophoresis pattern of whole GBM (Fig. 4, A and D, respectively) were similar to previous results (7). The composition of the 4°C and 25°C protomers, based on HPLC analysis, are presented in Table I. The ratios are consistent with the two-dimensional electrophoresis patterns which show that the NC1 domains of the 4°C protomers are mainly derived (Ͼ90%) from the ␣1(IV) and ␣2(IV) chains, and the NC1 domains (monomer and dimer) of the 25°C protomers are mainly derived from the ␣3(IV), ␣4(IV), and ␣5(IV) chains, with minor amounts from the ␣1(IV) and ␣2(IV) chains. Further experiments (data not shown) demonstrated that pseudolysin digestion at 25°C did not completely remove ␣1(IV) and ␣2(IV) chains from GBM, even when the digestion was extended to Ͼ50 h and with the addition of fresh pseudolysin. Overall, these results indicate that the 4°C protomers mainly comprise (Ͼ90%) the ␣1(IV) and ␣2(IV) chains, whereas the 25°C protomers mainly comprise (Ͼ90%) the ␣3(IV), ␣4(IV), and ␣5(IV) chains, along with a small amount (Ͻ10%) of ␣1(IV) and ␣2(IV) chains.

Supramolecular Organization of 4°C and 25°C Protomers
Rotary Shadowing Electron Microscopy-The 4°C truncated protomers, comprising the ␣1 and ␣2(IV) chains, were analyzed by rotary shadowing electron microscopy (Fig. 5). In Fig. 5A, they are shown to have a triple-helical (rodlike) segment, 300 Ϯ 28 nm in length, linked to a globular NC1 domain and are dimerized through NC1-NC1 interactions, forming molecules with lengths of about 600 nm. These molecules have a very similar length to the ␣1(IV)⅐␣2(IV) protomers (triple-helical domain ϭ 287 nm) that were solubilized from STBM by pseudolysin at 4°C (7), but they have a length twice that of the ␣1(IV)⅐␣2(IV) protomers solubilized from LBM (38). About 5-10% of the GBM protomers were not connected by NC1-NC1 interactions and thus existed as monomers (data not shown). A small percentage had triple helices of contour length of only 140 nm and these occurred as dimers (Fig. 5B)  The NC1 domains were prepared by collagenase digestion of samples and were analyzed by HPLC (panels A-C) and two-dimensional electrophoresis (panels D-F). The first dimension was run under non-reducing conditions in the presence of 6 M urea and the second dimension was run under non-reducing conditions in the presence of 1% SDS as described previously (27). Panels A and D correspond to whole GBM, B and E correspond to 4°C protomers, and C and F to 25°C protomers. The NC1 domains were resolved into fractions I to IV by HPLC (panels A-C). The composition of fractions is as follows: fraction I ϭ ␣1(IV), ␣2(IV), and ␣5(IV) monomers; fraction II ϭ ␣1(IV) and ␣2(IV) dimers; fraction III ϭ ␣4(IV) monomers and dimers; and fraction IV ϭ ␣3(IV) monomers and dimers (7). The arrowheads designate pI markers (panels D-F), and the location of the various NC1 domains was established previously (7). Panel E shows that the NC1 domain of the 4°C protein corresponds to monomers and dimers of ␣1(IV) and ␣2(IV) NC1 domains, according to the migration positions identified previously (63). In contrast, panel F shows that NC1 domains of the 25°C protomers correspond to monomers and dimers of the ␣3(IV), ␣4(IV), and ␣5(IV) NC1 domains.
(data not shown). A small percentage existed as dimers in which one triple-helix had a length of 300 nm and the other 140 nm (Fig. 5C). Fragments that originated from the 7 S domain were also observed (Fig. 5D).
The 25°C protomers, mainly comprising the ␣3(IV), ␣4(IV), and ␣5(IV) chains, were also analyzed by electron microscopy. These exhibited a complex supramolecular structure characterized by looping and supercoiling of the triple-helical domains (Fig. 6A). Pepsin digestion removed the NC1 domains from the 25°C protomers but did not alter the supercoiling and loop structures of the triple-helical domain (Fig. 6B). Reduction and alkylation under non-denaturing conditions effected substantial conversion of the suprastructures into truncated protomers (monomers and NC1-NC1 linked dimers) (Fig. 7A). Reduction and alkylation under non-denaturing conditions of a pepsin digested sample resulted in the conversion of most of the complexes into relatively simple molecules that lacked NC1 domains (Fig. 7B). Overall, these results indicate that the suprastructures of the 25°C protomers are stabilized by disulfide cross-links between triple-helical domains and between NC1 domains of adjoining protomers.
Identification of a Disulfide-linked Complex of the ␣3(IV), ␣4(IV), and ␣5(IV) Chains-The 25°C protomers were further characterized by SDS-agarose gel electrophoresis (Fig. 8A) to estimate the size distribution of dissociated components with molecular weights Ͼ200,000. The components ranged in apparent molecular weight from ϳ300,000 to Ͼ1,000,000 (Fig. 8A, lane e). Reduction and alkylation under non-denaturing condi-tions (i.e. in 0.5 M NaCl, 0.1 M Tris-HCl, pH 8.3) produced a component with a M r 500,000 (Fig. 8A, lane f) indicating the presence of disulfide cross-links. This M r 500,000 component dissociated into multiple polypeptides (M r Ͻ160,000) upon complete reduction under denaturing conditions (see below), indicating that it is cross-linked by disulfide bonds. This M r 500,000 component was then treated with pepsin, which degrades the NC1 domains and cleaves at the noncollagenous interruptions of the collagenous domain. The peptic product retained an apparent M r 500,000 (Fig. 8A, lane g), which indicates that the disulfide cross-links were within the collagenous , was determined from the HPLC profile (area of fractions I ϩ II/area of fractions III ϩ IV). Fraction I corresponds to ␣1(IV) ϩ ␣2(IV) ϩ ␣5(IV)NC1 monomers, fraction II corresponds to ␣1(IV) ϩ ␣2(IV)NC1 dimers, fraction III corresponds to ␣4(IV)NC1 monomers and dimers, and fraction IV corresponds to ␣3(IV)NC1 monomers and dimers, as described previously (7,26). domain, as those within the NC1 domain would have been removed by the digestion.
The 25°C protomers were also denatured in 6 M guanidine-HCl and fractionated on a column of Sephacryl S-400 (Fig. 8B) to identify disulfide cross-linked components. Fraction I, the major fraction, corresponds to the high molecular weight component(s) (M r 500,000) observed by SDS-agarose electrophoresis (Fig. 8A, lane d (fraction I versus lane e (unfractionated)). When analyzed by SDS-PAGE, fraction I did not penetrate the stacking gel because of its large size, but upon complete reduction of disulfide bonds, it dissociated into multiple polypeptides with apparent M r Ͻ160,000 (Fig. 8C). Fractions II and III (Fig.  8B) consisted of components of lower molecular weight than fraction I, and both yielded components (M r Ͻ160,000) upon reduction (data not shown). Overall, these results indicate that fraction I is a high molecular weight component(s), composed of truncated ␣(IV) chains that are cross-linked by disulfide bonds between triple-helical domains.
The identity of ␣(IV) chains comprising fractions I, II, and III was determined from the identity of the NC1 domains that were released upon collagenase digestion. The NC1 domains were identified by SDS-PAGE with Western blotting, using chain-specific antibodies. Fraction I reacted with anti-␣3(IV), -␣4(IV), and -␣5(IV) antibodies and not with anti-␣1(IV), -␣2(IV), or -␣6(IV) antibodies (Fig. 8D). In contrast, fractions II and III reacted with ␣1(IV) to ␣5(IV) antibodies but not with ␣6(IV) antibodies (data not shown). The ␣4(IV) and ␣5(IV) antibodies reacted primarily with dimers in fraction I, but with monomers in fractions II and III. Six molar guanidine-HCl is a denaturing and dissociating agent, and its ability to dissociate the complexes shows that there are noncovalent interactions between ␣1(IV) and ␣2(IV) chain fragments and the ␣3(IV), ␣4(IV), and ␣5(IV) chain fragments. The failure of 6 M guanidine-HCl to dissociate fraction I into its constituent ␣3(IV), ␣4(IV), and ␣5(IV) chains, which were identified by SDS-PAGE after reduction (Fig. 8C), indicated that they exist as a high molecular weight complex in which they are cross-linked by disulfide bonds. Moreover, this ␣3(IV)⅐␣4(IV)⅐␣5(IV) complex serves as the target antigen for both GP autoantibodies and ALP alloantibodies (Fig. 8D). DISCUSSION Pseudolysin digestion of GBM at 4°C and then at 25°C solubilized two distinct sets of truncated protomers of type IV collagen. Those solubilized at 4°C were mainly comprising ␣1(IV) and ␣2(IV) chains. By electron microscopy, they exhibited structural characteristics of ␣1(IV)⅐␣2(IV) protomers that have been described for STBM and LBM (7, 38). These include  Fig. 7A, and lane g corresponds to the pepsin-digested protomers that were reduced and alkylated (Fig. 7B). Lane d corresponds to fraction I of Fig. 8B. In panel B, the 25°C protomers shown in Fig. 6A were dissociated in 6 M guanidine-HCl and fractionated on a Sephacryl S-400 column. Fraction I was pooled for further study. In panel C, fraction I was analyzed by SDS-PAGE before (NR) and after reduction (R) with 2-mercaptoethanol. Upon reduction the constituents penetrate the resolving gel. The constituents were identified by Western blotting with ␣(IV) chain-specific antibodies and their length in amino acid residues is shown. The M r of standard proteins is given in kDa. In panel D, the NC1 domains of fraction I were released by collagenase digestion and their ␣(IV) chain identities were determined by SDS-PAGE with Western blotting using ␣(IV) chain-specific antibodies. Fraction I reacted with ␣3(IV), ␣4(IV), and ␣5(IV) monoclonal antibodies and GP (patient LL) and ALP antibodies. CB indicates protein staining with Coomassie Blue. The mobility of molecular weight markers is shown. a long triple-helical segment that connects to the NC1 domain at the carboxyl terminus, and NC1-NC1 interactions, forming dimers of protomers. The protomer length is about two-thirds of the 450 nm length of the intact molecule (9), which reflects removal of a large region of the triple helix that includes the 7 S amino-terminal tetramerization domain. These truncated molecules, therefore, reflect the existence of a supramolecular network comprising the ␣1(IV) and ␣2(IV) chains in GBM. The existence of such a network is also supported by the finding that NC1 hexamers comprising ␣1(IV) and ␣2(IV) NC1 domains can be isolated from GBM (10,12), although the cited studies could not rule out the presence of ␣5(IV) and ␣6(IV) chains.
In contrast to the truncated protomers solubilized at 4°C, those solubilized at 25°C were mainly comprising the ␣3(IV), ␣4(IV), and ␣5(IV) chains, along with a small amount of ␣1(IV) and ␣2(IV) chains. These exhibited some structural features that are in common with the ␣1(IV)⅐␣2(IV) protomers and others that are distinctly different. The common features include a long triple-helical domain, a globular NC1 domain and NC1-NC1 interactions causing end-to-end association of protomers. Distinct differences are evident at the supramolecular level in which the ␣3(IV)⅐␣4(IV)⅐␣5(IV) protomers are characterized by loops and supercoiling of the triple helices (Fig. 6A) that are stabilized by inter-protomer disulfide bonds. At the chain level, the ␣3(IV), ␣4(IV), and ␣5(IV) chains of the complex are covalently linked by disulfide bonds, exclusive of ␣1(IV) and ␣2(IV) chains. Evidence for disulfide bonds is that the high molecular weight complex does not dissociate in 6 M guanidine-HCl or 1% SDS in the absence of a disulfide reducing agent, but in the presence of a reducing agent it dissociates into its constituent ␣3(IV), ␣4(IV), and ␣5(IV) chains. Fragments of the ␣1(IV) and ␣2(IV) chains could be separated from the ␣3(IV)⅐␣4(IV)⅐␣5(IV) complex by exposure to 6 M guanidine-HCl, indicating that they were not linked in the complex by disulfide bonds. The 25°C protomers, therefore, reflect the existence of a novel supramolecular network in GBM that comprises the ␣3(IV), ␣4(IV), and ␣5(IV) chains and which is cross-linked by disulfide bonds between triple helices. Such a network is also supported by the finding that NC1 hexamers, comprising the ␣3(IV) and ␣4(IV) NC1 domains, can be isolated from GBM (10), although the studies did not address the presence of the ␣5(IV) and ␣6(IV) chains. Moreover, an ␣3(IV)⅐␣4(IV)⅐␣5(IV) network is consistent with immunocytochemical results showing that the ␣3(IV), ␣4(IV), and ␣5(IV) chains, but not the ␣6(IV) chain, are present in GBM (8,33). A similar network that also contains the ␣6(IV) chain is present in STBM (7).
The presence of disulfide cross-links between the triple helices of the ␣3(IV)⅐␣4(IV)⅐␣5(IV) truncated protomers is a distinguishing feature not found in the ␣1(IV)⅐␣2(IV) truncated protomers. This can be rationalized by the presence of several cysteine residues in the collagenous domain of the ␣3(IV) and ␣4(IV) chains that are not found in the ␣1(IV) and ␣2(IV) chains. Comparison of the amino acid sequences of the six ␣(IV) chains (39-48) reveals the following. First, there are 18 conserved cysteine residues among all six human ␣(IV) chains: 4 in the 7 S domain, 12 in the NC1 domain, and 2 in triple-helical interruption IX (Fig. 9). Second, the ␣3(IV) chain contains four cysteine residues and the ␣4(IV) chain contains nine cysteine residues within the collagenous domain that are not found in the other four ␣(IV) chains (Fig. 9). Third, the truncated ␣3(IV) and ␣4(IV) chains that are constituents of the ␣3(IV)⅐␣4(IV)⅐␣5(IV) complex, fraction I (Fig. 6B), have a sufficient length (Ͼ1300) residues to contain these additional cysteine residues, indicating their participation in the inter-protomer disulfide cross-links that stabilize the ␣3(IV)⅐␣4(IV)⅐␣5(IV) complex shown in Fig. 7A. Fourth, the truncated ␣5(IV)chain of the ␣3(IV)⅐␣4(IV)⅐␣5(IV) complex has a length of ϳ1350 residues and contains an NC1 domain. This length is sufficient to contain a cysteine residue in helical interruption VIII that could participate in a disulfide cross-link with an ␣3(IV) or ␣4(IV) chain, either within a protomer or between protomers. The ␣5(IV) chain is also linked to the ␣3(IV) or ␣4(IV) chain through disulfide bonds between NC1 domains because the 1100-residue fragment of the ␣3(IV)⅐␣4(IV)⅐␣5(IV) complex (Fig. 8C) is not of sufficient length to contain a cysteine residue within the collagenous domain. Presumably, the disulfide cross-links within the collagenous domain stabilize the looping and supercoiling of the triple helices to confer a specialized function to the ␣3(IV)⅐␣4(IV)⅐␣5(IV) network.
The ␣3(IV)⅐␣4(IV)⅐␣5(IV) supramolecular structure (Fig. 6A) is similar to structures found in vitro and in situ by Yurchenco and co-workers (49 -53) in membranes known to contain ␣1(IV) and ␣2(IV) chains but which were studied before it was possible to detect ␣3(IV), ␣4(IV), and ␣5(IV). In one of these studies (49), it was shown that mouse type IV collagen protomers from the Engelbreth-Holm-Swarm tumor, containing only ␣1(IV) and ␣2(IV) chains, could be induced to form a polygonal network  (49). The ␣3(IV)⅐␣4(IV)⅐␣5(IV) supramolecular structure that we observed is stabilized by disulfide cross-links between the constituent ␣(IV) chains. Conceivably, the ␣1(IV)⅐␣2(IV) protomers we observed may have had a similar suprastructure in vivo, but it was not preserved because of the absence of disulfide bonds to stabilize it for viewing by electron microscopy. In addition, our electron microscopy studies were done at much lower concentrations than those needed to form polygonal networks in vitro (49).
The finding of an ␣1(IV)⅐␣2(IV) network and an ␣3(IV)⅐ ␣4(IV)⅐␣5(IV) network provides a conceptual framework to explain several features of the GBM abnormality in Alport syndrome. First, recent studies reveal that normal glomerular development involves a developmental switch in which the early expression of the ␣1(IV) and ␣2(IV) chains is replaced by the expression of the ␣3(IV), ␣4(IV), and ␣5(IV) chains to form a mature GBM (20,54). In X-linked Alport syndrome, the switch is arrested, causing the persistence of the ␣1(IV) and ␣2(IV) chains and the absence of the ␣3(IV), ␣4(IV), and ␣5(IV) chains (20). Based upon the present results, the switch can now be defined at the supramolecular level in which the ␣1(IV)⅐␣2(IV) network is replaced by the ␣3(IV)⅐␣4(IV)⅐␣5(IV) network in normal glomerular development, but in Alport syndrome the switch is arrested. Second, it is well established that mutations in the ␣5(IV) chain cause defective assembly of not only the ␣5(IV) chain, but the ␣3(IV) and ␣4(IV) chains as well (5, 8, 19 -21). This conundrum indicates that the developmental switch contains a mechanism that links the assembly of all three ␣(IV) chains. The finding of an ␣3(IV)⅐␣4(IV)⅐␣5(IV) network, in which all three chains are covalently linked by disulfide cross-links, provides a possible linkage mechanism, wherein the ␣5(IV) chain is required for the assembly of the ␣3(IV) and ␣4(IV) chains. The dependence of one chain on the assembly of another chain is a well established mechanism for type I collagen in patients with osteogenesis imperfecta in which mutations in the ␣1(I) chain cause defective assembly of the ␣2(I) chain (55). The ␣5(IV) chain requirement could be at: (a) the protomer level in which an ␣5(IV) chain together with an ␣3(IV) and ␣4(IV) chain forms a triple-helical protomer or (b) the supramolecular level in which an ␣5(IV)-containing protomer is required for the incorporation of an ␣3(IV) or ␣4(IV) containing protomer. The linkage mechanism may also involve the transcriptional/translational level because the mRNA expression of the ␣3(IV), ␣4(IV), and ␣5(IV) chains appears to be coordinated (22). The protein assembly and the transcriptional/translational mechanism need not be mutually exclusive because the incorporation of an ␣5(IV) chain into extracellular matrix could be required to modulate the transcription of the ␣3(IV) and ␣4(IV) chains.
The fundamental importance of the ␣3(IV)⅐␣4(IV)⅐␣5(IV) network for normal glomerular ultrafiltration function is evident from the gene mutations that cause renal failure in Alport syndrome and from the ␣3(IV) knockout mouse models (18,57). The importance is underscored by the restricted distribution of this network to peripheral GBM within the nephron. Conceivably, the looping and supercoiling of the triple helices and the disulfide cross-linking of the network (Fig. 6A) confer long term stability to the GBM by protecting against proteolytic degradation (20). An increased susceptibility of the ␣1(IV)⅐␣2(IV) network to the action of proteases is supported by the observation that: (a) the level of 3-hydroxyproline, a constituent of type IV collagen, is elevated in the urine of 20 Alport patients in comparison to that of the other renal disorders (64), and (b) Alport renal basement membranes appears more susceptible to proteolysis than normal GBM (20). Moreover, the ␣1(IV)⅐␣2(IV) network is more easily excised from GBM by pseudolysin than the ␣3(IV)⅐␣4(IV)⅐␣5(IV) network as described herein, which further supports the hypothesis that the ␣3(IV)⅐␣4(IV)⅐␣5(IV) network may protect against proteolysis.
The present findings also establish the supramolecular structure of the target molecules in GBM that bind the pathogenic anti-GBM antibodies in patients with Goodpasture syndrome and in patients with Alport syndrome who develop posttransplant nephritis. Previous work has established that Goodpasture autoantibodies are also targeted to the NC1 domain of the ␣3(IV) chain (1). The Alport alloantibodies, produced in some patients in response to a renal transplant, are targeted to the NC1 domain of the ␣3(IV) chain in certain patients (59,60) and the ␣5(IV) chain in others (21,61,62). As shown in Fig. 8D, both GP autoantibodies (63) and Alport alloantibodies (59) bind to fraction I of the 25°C protomers. Thus, the ␣3(IV)⅐␣4(IV)⅐␣5(IV) network shown in Fig. 6A is the target for both kinds of antibodies that cause anti-GBM nephritis.