Assembly of Type IV Collagen INSIGHTS FROM α3(IV) COLLAGEN-DEFICIENT MICE

Type IV collagen includes six genetically distinct polypeptides named α1(IV) through α6(IV). These isoforms are speculated to organize themselves into unique networks providing mammalian basement membranes specificity and inequality. Recent studies using bovine and human glomerular and testis basement membranes have shown that unique networks of collagen comprising either α1 and α2 chains or α3, α4, and α5 chains can be identified. These studies have suggested that assembly of α5 chain into type IV collagen network is dependent on α3 expression where both chains are normally present in the tissue. In the present study, we show that in the lens and inner ear of normal mice, expression of α1, α2, α3, α4, and α5 chains of type IV collagen can be detected using α chain-specific antibodies. In the α3(IV) collagen-deficient mice, only the expression of α1, α2, and α5 chains of type IV collagen was detectable. The non-collagenous 1 domain of α5 chain was associated with α1 in the non-collagenous 1 domain hexamer structure, suggesting that network incorporation of α5 is possible in the absence of the α3 chain in these tissues. The present study proves that expression of α5 is not dependent on the expression of α3 chain in these tissues and that α5 chain can assemble into basement membranes in the absence of α3 chain. These findings support the notion that type IV collagen assembly may be regulated by tissue-specific factors.

Type IV collagen is a family of complex polypeptides and a major constituent of mammalian basement membranes (1,2). The ␣1(IV) and ␣2(IV) chains are products of distinct genes located pairwise in a head-to-head fashion on chromosome 13 in humans (3). The ␣3(IV) and ␣4(IV) chains are present in the same orientation on chromosome 2, and the ␣5(IV) and ␣6(IV) chains are located on the X chromosome in humans (4). The type IV collagen protomer (trimer) consists of three ␣ chains that come together through associations among their NC1 1 globular domains followed by folding of the collagenous domains into triple helices through covalent and non-covalent interactions (1,2,4,5). The protomer is divided into three different domains: an NH 2 -terminal 7 S domain, a middle triple-helical domain, and a COOH-terminal NC1 globular domain (1,4,5). With six different ␣ chains known at present, 56 different combinations of triple-helical protomers are possible (1, 2, 4 -6). Transplantable, basement membrane-producing mouse Engelbreth-Holm-Swarm sarcoma (EHS) tumor is a plentiful source of type IV collagen (2). Type IV collagen isolated from EHS tumor contains only ␣1 and ␣2 chains (7). These chains were identified to exist as an ␣1/␣2 type IV collagen protomer in a 2:1 ratio (1, 2, 4 -6). By use of rotary shadow electron microscopy it could be shown that four protomers of type IV collagen are connected by association of the 7 S domain forming a spider-shaped structure (1, 2, 4 -6). Each protomer is bonded with another protomer by its NC1 domain to form interlocking hexamers (6). The combination of these two types of interaction along with extensive side-by-side (lateral) associations within the collagenous triple helix allows the formation of a network that serves as a scaffold for the basement membrane (6). Type IV collagen protomer compositions in the human basement membranes are still largely unknown.
Most of our current understanding of the structure and selfaggregating properties of type IV collagen is derived from studies using mouse EHS tumor, which contains only ␣1 and ␣2 chains (1, 2, 4 -6). Recently, all six chains of type IV collagen were identified by protein biochemistry techniques using bovine testis and kidney GBM (8,9). In these studies, pseudolysin (Pseudomonas elastase) at two different temperatures was shown to extract two different populations of type IV collagen selectively (8,9): one population that just contains ␣1 and ␣2 chains, like the EHS tumor, and another that contains ␣1, ␣3, ␣4, ␣5, and ␣6 chains (8,9). These reports speculate that type IV collagen networks involving the ␣3, ␣4, and ␣5 chains may play a unique role in the specialized basement membranes such as seminiferous tubule basement membrane (STBM) and GBM (8,9). Additionally, recent experiments have shown that in most cases, ␣5 chain mutations in Alport syndrome (an Xlinked hereditary kidney disease) lead to an absence of ␣3 chain of type IV collagen in the GBM of these patients (10 -17). These studies, along with immunoprecipitation experiments using human GBM hexamers, have led to the suggestion that ␣5 and ␣3 are regulated tightly in their expression and are dependent on each other to form protomeric network of type IV collagen (9,17,18).
Identification of new chains of type IV collagen with their restricted tissue distribution and the presence of separate net-* This study was supported by National Institutes of Health Grants DK-51711 (to R. K.), DK-55001 (to R. K.), DK-55000 (to D. C.), and PO1-DC01813 (to D. C.), the 1998 American Society of Nephrology Carl Gottschalk award and the 1998 NKF Murray award (to R. K.), and research funds from Beth Israel Deaconess Medical Center. DK55001 (to R. K.) and DK-55000 (to D. C.) are part of NIDDK-funded Interactive Research Project Grant (IRPG). 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.
works raises a number of unanswered biological questions pertaining to the assembly of these polypeptides (3). It is important to determine the organization of newly discovered human type IV collagen ␣ chains in forming protomers (trimers) and to determine how these new molecules might contribute to the structural and functional diversity of basement membranes. In the present study, we isolated type IV collagen from the ␣3 collagen-deficient mice and their wild type (WT) littermates and evaluated their type IV collagen ␣ chain composition and NC1 hexamer organization.

MATERIALS AND METHODS
All chemical reagents, horseradish peroxidase, and alkaline phosphatase-conjugated antibodies toward mouse, rabbit, and human IgG were purchased from Sigma and Fisher Scientific. Type IV collagenspecific antibodies were generated and specificity determined as described previously in other publications (9, 19 -23).
The ␣3 collagen-deficient mice were generated as described previously (24,25). Kidneys, lungs, lens, and inner ears were isolated form the ␣3KO and heterozygote mice immediately after the sacrifice of the mice, and the tissue was snap frozen immediately in liquid nitrogen for type IV collagen extraction. For studies involving immunofluorescence, the tissue was processed as described previously (24). SDS-PAGE, Western blotting, and direct ELISA were performed as described previously (19 -24, 26 -30). For Western blotting, biotin-labeled anti-rabbit IgG was used as the second (detection) antibody.
Immunohistochemical Staining-Immunohistochemical detection of type IV collagen chains was performed as described previously (24). These antisera were produced using synthetic peptides corresponding to regions of the mouse NC1 domains homologous to the sequence used (31) to produce type IV collagen ␣ chain-specific antibodies for the human chains. Antisera were tested by ELISA using plates precoated with either the synthetic peptide or recombinant NC1 domains. All antisera were specific for the type IV collagen ␣ chains to which they were raised (24).
Fresh frozen tissues (with exception of the cochlea) were sectioned at 5 M and collected onto poly-L-lysine-coated slides. Cochlea were perfused with Carnoy's fixative, decalcified, embedded, and cut at 8 M as described previously (24). Slides were postfixed by immersion in cold 95% ethanol for 15 min and allowed to air dry. Prior to staining, sections were first denatured by immersion in 6.0 M urea and 0.1 M glycine (pH 3.5) for 1 h at 50°C. Following three 10-min washes in phosphate-buffered saline, primary antibodies were applied in 7% w/v non-fat dry milk (Bio-Rad) and allowed to react for 2 h at room temperature in a humidified chamber. After four 10-min washes with phosphate-buffered saline, fluorescein isothiocyanate-conjugated secondary antibodies were applied (Vector Laboratories) and allowed to react for 1 h at room temperature. Slides were then washed in phos-phate-buffered saline four times at room temperature, and a small amount of Vectashield mounting medium was applied (Vector Laboratories) before sealing the specimen under glass coverslips using clear nail polish. Immunofluorescence was visualized using an Olympus BH-2 fluorescence microscope configured with Cytovision Ultra (Applied Imaging, Inc.) image analysis software.
Immunoprecipitation Experiments-The NC1 hexamer was prepared by bacterial collagenase and analyzed by SDS-PAGE and immunoblotting as described previously (19 -23, 26 -30, 32). Immunoprecipitation was performed using the polyclonal ␣1, ␣3, and ␣5 antibodies and the methods described by Kleppel et al. (18) and Johansson et al. (33), with some modifications. Briefly, 10 mg of collagenase digest of detergentextracted inner ear, lens, and kidney GBM from WT and knockout mice were incubated with antibodies to the ␣ chains, and the resultant mixture was precipitated with antibodies to rabbit IgG conjugated to protein A and analyzed by SDS-PAGE, immunoblotting, and ELISA.
Statistical Analysis-All values are expressed as mean Ϯ S.E. Analysis of variance with a one-tailed Student's t test was used to identify significant differences in multiple comparisons. A level of p Ͻ 0.05 was considered statistically significant.
However, in the anterior capsular material of the lens and the strial capillary basement membranes, the scenario is different. Here, in the ␣3KO mice, the ␣5(IV) chain continues to be assembled in the absence of ␣3(IV) and ␣4(IV) chains (Fig.  2). In the inner ear, mid-modiolar cross-sections reveal that the track of basement membrane running from the limbus, down the inner sulcus, across the basilar membrane, up the external sulcus to the spiral prominence, and radiating into the spiral ligament surrounding the root cell processes in tissue from the mutant mouse is devoid of collagen ␣3(IV), ␣4(IV), and ␣5(IV) chains (Fig. 2). In contrast, in the capillaries of the stria FIG. 1. Immunohistochemical detection of type IV collagen chains in renal and testis basement membranes. Frozen sections from either WT or ␣3KO mice were immunostained using antibodies specific for the indicated type IV collagen chains in the middle of the two panels. The antibodies were used at a dilution of 1:200 and detected with antirabbit IgG conjugated to fluorescein isothiocyanate. All primary antibodies were raised in rabbit against mouse peptide sequences. Magnification, ϫ 100. vascularis, the collagen ␣5(IV) chain is assembled into basement membrane (see structures denoted by arrows in Fig. 2, Inner ear WT and ␣3KO for ␣5 antibodies).
We extracted type IV collagen from detergent extracts of anterior lens capsule, cochlea, kidney cortex, and testis by bacterial collagenase from the control and ␣3KO mice. The collagenase-solubilized material analyzed by SDS-PAGE and Coomassie Blue staining revealed major sets of bands in the region of 46 kDa and 28 kDa (data not shown), consistent with the pattern seen for denatured NC1 hexamer from various tissues (32, 34 -37). This material was used (21,32,(35)(36)(37) in direct ELISA experiments using ␣ chain-specific antibodies. The ␣1, ␣2, ␣5, and ␣6 chains were present in all tissues analyzed (Fig. 3). The ␣3 and ␣4 chains were present in the control cochlea, anterior lens capsule, kidney cortex, and the testis, but they were absent in the tissues of the ␣3KO mice. These results are consistent with the results observed in the immunofluorescence experiments (Figs. 1 and 2). Immunoblotting experiments correlate well with the ELISA experiments. The ␣1/␣2 antibodies show binding to the dimeric (ϳ46 kDa) and monomeric (ϳ28 kDa) forms of the NC1 domain of ␣1 and ␣2 chains of type IV collagen in all the tissues analyzed (Fig. 4). The ␣5 chain was present in all the tissues analyzed except for the ␣3KO kidney and testis (Fig. 4). The ␣3 and ␣4 were detectable only in the WT anterior lens, cochlea, and the kidney and were absent in all tissue analyzed in the ␣3KO mice (data not shown). These results suggest that ␣5 expression is present in the inner ear cochlea and anterior lens capsule of the ␣3KO but is absent in the kidney cortex of the same mice.
To determine the network organization of ␣5 chain in the ␣3KO mice, we performed immunoprecipitation experiments utilizing mouse ␣1 and ␣5 chain-specific polyclonal antibodies using NC1 hexamer preparations from the anterior lens capsule of the WT and ␣3KO mice. The specificity of the mouse ␣1 and ␣5 chain antibodies was established previously (9, 19 -23) and determined again by direct ELISA using recombinant human type IV collagen ␣1-␣6 NC1 domains (data not shown). Using collagenase-solubilized ␣3KO kidney renal basement membrane, the ␣1 antibodies immunoprecipitated a 160-kDa band corresponding to the molecular mass of NC1 hexamer (Fig. 5, ND). Under denaturing conditions this band resolved into a set of bands in the range of 46 and 28 kDa (Fig. 5, D). These bands were immunoreactive with ␣1, ␣2, and ␣5 antibodies (Fig. 5). The ␣3 and ␣4 antibodies did not reveal any significant binding (Fig. 5). These results suggest that ␣1 antibodies can precipitate the ␣2 and ␣5 NC1 domain in a hexamer form. The ␣2 antibodies only immunoprecipitate hexamers that contain ␣1 and ␣2 NC1 domains (data not shown), suggesting that ␣5 NC1 does not form hexamers with ␣2 NC1. The ␣5 antibodies immunoprecipitate NC1 hexamers that contain only ␣1 and ␣5 NC1 domains (Fig. 6). The coprecipitation of ␣1 and ␣5 with ␣5 antibodies suggests that these molecules exist in the same hexamer and hence participate in type IV collagen network formation independently of ␣2, ␣3, ␣4, and ␣6 chains. When ␣3 antibodies were used to immunoprecipitate WT lens capsule collagenase-digested material, the immunoprecipitate revealed predominant binding to ␣3 and ␣4 chains with minor binding to ␣1 antibodies (Fig. 7A). These results suggest that, at least in the lens capsule basement membrane, from control WT or ␣3KO mice was coated in each well for analysis with ␣ chain-specific antibodies. Each sample was analyzed for the presence of ␣1, ␣2, ␣3, ␣4, ␣5, and ␣6 chains. Detergent extraction and collagenase solubilization of the tissue were performed as described elsewhere (8,32,54,55). All antibodies were used at a dilution of 1:200. All values are average of triplicate readings from two separate experiments.

FIG. 4. Immunoblot analysis for type IV collagen isoforms in mice.
Panel A, ␣1/␣2 antibodies. Panel B, ␣5 antibodies. In both panels ϩ/ϩ indicates collagenase-solubilized WT tissue, and Ϫ/Ϫ denotes collagenase-solubilized ␣3KO tissue. Antibodies were used at a dilution of 1:100. All primary antibodies were raised in rabbit against mouse peptide sequences. D denotes dimers of NC1 domain of type IV collagen with a molecular mass of ϳ46 kDa, and M denotes monomers of NC1 domain of type IV collagen with a molecular mass of ϳ26 kDa. the ␣3 NC1 domain is not associated with the ␣5 NC1 domain in the NC1 hexamer structure. Immunoprecipitation of WT lens capsule with ␣5 antibodies precipitates ␣1 and ␣5 NC1 domains in a hexamer form, similar to the ␣3KO experiments (Fig. 7B). Additionally, as expected, immunoprecipitation of collagenase-solubilized GBM from ␣3KO mice with ␣1 antibodies precipitates NC1 hexamers containing ␣1 and ␣2 NC1 domains (data not shown). Collectively, these experiments support the notion that in the lens capsule basement membrane, ␣1 and ␣5 can form a NC1 hexamer and hence participate in type IV collagen protomer dimerization, independently of ␣3 chain.

DISCUSSION
Several reports in the last few years suggest that the expression of ␣3 chain of type IV collagen is dependent on the expression of ␣5 chain and vice versa (3, 8, 9, 11-17, 22, 38 -43). This suggestion is derived from biochemical and immunological studies using bovine and human kidneys and kidneys from patients with Alport syndrome 2 (3, 8, 9, 11-17, 22, 38 -46). Immunofluorescence experiments using human kidney sections from normal and Alport patients reveal an absence of ␣3, ␣4, and ␣5 chains of type IV collagen from the GBM of most Alport patients (3, 8, 9, 11-17, 22, 38 -45). In some of these X-linked Alport patients, mutations and deletions of ␣5 chain have been identified as the primary genetic defect (11-16, 45, 47). These studies led to the conclusion that assembly of ␣3, ␣4, and ␣5 chains to form networks of type IV collagen must be regulated tightly, and hence the absence of one of these chains (the ␣5 in the case of X-linked Alport syndrome) leads to the absence of the other two chains (8, 9, 11-14, 16, 38, 39). Coregulation of ␣3 and ␣5 protomer assembly gained further support when the mRNA levels for these proteins were shown to be unaltered in the human Alport kidneys (48).
Biochemical studies involving immunoprecipitation with ␣ chain-specific antibodies using the NC1 hexamer from bovine and human GBM type IV collagen reveal an association of ␣3 and ␣5 NC1 domains in the same hexamer. These results validate further the notion that the expression of ␣3 and ␣5 is regulated tightly in basement membranes, especially the GBM (18). In a recent study, pseudolysin was used to extract selectively different networks of type IV collagen at different temperatures. At 4°C, the enzyme selectively extracted a complex of ␣1and ␣2-containing type IV collagen proteins. Interestingly, when the resultant pellet from the earlier digestion was re-extracted with pseudolysin at 25°C, a complex of proteins enriched in the ␣3, ␣4, and ␣5 was extracted. These results led the authors to propose the existence of two networks of type IV collagen: one that is composed of ␣1 and ␣2 (without interchain lateral disulfide bonds), and another that contains predomi-2 Alport syndrome is predominantly an X-linked hereditary kidney disease associated with progressive renal failure leading to end stage renal disease, sensorineural deafness, and ocular lesions. The genetic basis of the disease has been identified as mutations and deletion in the GBM-specific collagen isoform, the ␣5 chain.
In this report we present a mouse that lacks the expression of ␣3 chain at both the mRNA and protein level (24). This mouse provided the opportunity to test the hypothesis of whether assembly of ␣3 and ␣5 chains is always regulated tightly in basement membranes in which they are both normally present. In our analysis of various basement membranes in the ␣3KO mice, we found that in the kidney, a lack of ␣3 was associated with the absence of ␣4 and ␣5, whereas in the anterior lens capsule and the strial capillary basement membrane of the cochlea, the expression of ␣5 persisted despite the absence of ␣3 chain in these basement membranes. Our results show that the ␣5 NC1 colocalized with ␣1 chain in the same hexamer in the ␣3KO and WT mice. These results suggest that ␣1 and ␣5 can be present in the same network in the absence of ␣3 chain (Fig. 8). In the kidney, the expression of these two chains seems tightly coordinated and dependent on the expression of each other, but our study presents a possibility of tissue-specific type IV collagen assembly. Such tissue-specific ␣ chain assembly can be achieved presumably via factors such as chaperones. In this regard a heat shock protein (47 kDa) chaperone has been identified for collagen assembly, including the type IV collagen (50 -53). We speculate that assembly of type IV collagen ␣ chains in the formation of organ basement membranes is potentially dependent on tissue-specific assembly factors that are yet to be discovered. Such structural differences among organ basement membranes potentially contribute to important functional diversity. The ␣1and ␣5-containing hexamer, as shown for ␣3KO LBM, was also identified in the WT mice.

FIG. 7. Immunoprecipitation of collagenase-solubilized WT LBM by ␣3
and ␣5 antibodies. Panel A, anti-␣3 antibodies were used to immunoprecipitate WT collagenase-solubilized LBM (10 mg), and the immunoprecipitate was analyzed by direct ELISA. The ␣3 antibodies were used at a dilution of 1:200 for immunoprecipitation analysis. Panel B, anti-␣5 antibodies were used to immunoprecipitate WT collagenase-solubilized LBM (2 mg), and the immunoprecipitate was analyzed by direct ELISA. The ␣5 antibodies were used at a dilution of 1:200 for immunoprecipitation analysis. Immunoprecipitation was performed as described under "Materials and Methods." Direct ELISA is shown as a bar graph. ␣1, ␣2, ␣3, ␣4, and ␣5 denote antibodies against these chains.