Identification of the NC1 Domain of α3 Chain as Critical for α3α4α5 Type IV Collagen Network Assembly

The network organization of type IV collagen consisting of α3, α4, and α5 chains in the glomerular basement membrane (GBM) is speculated to involve interactions of the triple helical and NC1 domain of individual α-chains, but in vivo evidence is lacking. To specifically address the contribution of the NC1 domain in the GBM collagen network organization, we generated a mouse with specific loss of α3NC1 domain while keeping the triple helical α3 chain intact by connecting it to the human α5NC1 domain. The absence of α3NC1 domain leads to the complete loss of the α4 chain. The α3 collagenous domain is incapable of incorporating the α5 chain, resulting in the impaired organization of the α3α4α5 chain-containing network. Although the α5 chain can assemble with the α1, α2, and α6 chains, such assembly is incapable of functionally replacing the α3α4α5 protomer. This novel approach to explore the assembly type IV collagen in vivo offers novel insights in the specific role of the NC1 domain in the assembly and function of GBM during health and disease.

The network organization of type IV collagen consisting of ␣3, ␣4, and ␣5 chains in the glomerular basement membrane (GBM) is speculated to involve interactions of the triple helical and NC1 domain of individual ␣-chains, but in vivo evidence is lacking. To specifically address the contribution of the NC1 domain in the GBM collagen network organization, we generated a mouse with specific loss of ␣3NC1 domain while keeping the triple helical ␣3 chain intact by connecting it to the human ␣5NC1 domain. The absence of ␣3NC1 domain leads to the complete loss of the ␣4 chain. The ␣3 collagenous domain is incapable of incorporating the ␣5 chain, resulting in the impaired organization of the ␣3␣4␣5 chain-containing network. Although the ␣5 chain can assemble with the ␣1, ␣2, and ␣6 chains, such assembly is incapable of functionally replacing the ␣3␣4␣5 protomer. This novel approach to explore the assembly type IV collagen in vivo offers novel insights in the specific role of the NC1 domain in the assembly and function of GBM during health and disease.
Type IV collagen genes encode six distinct ␣-chains, ␣1 to ␣6 chains, which are expressed in development and tissue-specific patterns to allow for distinct type IV collagen composition in basement membranes (1)(2)(3)(4). The distinct supramolecular organization of type IV collagen networks likely provides a unique structural stability and distinct biochemical properties to the basement membranes they compose. Although genetically distinct, all chains of type IV collagen are highly homologous and assemble to form an irregular polyglonal scaffold that interacts with other extracellular proteins (2). The unique primary and secondary structure of type IV collagen chains likely allow for their intrinsic ability to assemble into specific networks. Each chain is composed of an N-terminal or 7S domain and a C-terminal or non-collagenous (NC1) domain composed of ϳ230 amino acids, separated by a major collagenous domain of ϳ1400 amino acid residues composed of Gly-X-Y repeats, in which X is proline or lysine, and Y is hydroxyproline or hydroxylysine. In theory, six chains of type IV collagen could assemble into 56 possible protomers combinations, yet in mammals only three protomers, ␣1␣2␣1, ␣3␣4␣5, and ␣5␣6␣5, have been speculated using in vitro assays (5,6).
The highest degree of sequence divergence between chains is observed in the NC1 domain, and in vitro studies suggest a role for the NC1 domain as the recognition and nucleation center for the folding of three chains of type IV collagen into protomers (7)(8)(9)(10)(11)(12)(13). In vitro self-assembly studies of type IV collagen NC1 domains (7,12,14), kinetics (15,16) and crystal structure analyses (15,17), and rotary shadowing microscopy of heatdenatured chains (18) suggest that the NC1 domain offers recognition sequences for chain selection (16). Immunoprecipitation studies on purified type IV collagen hexamers extracted mainly from bovine tissues indicate a strict chain composition of type IV collagen networks (14,16,19). At the protomer level, in vitro studies suggest a role for the NC1 domain in chain selection; that is, assembly of the purified ␣1 and ␣2 NC1 domains into ␣1␣2␣1 NC1 trimers and differential affinity of ␣2 NC1 domain as a mechanism for chain discrimination in ␣1␣2␣1 NC1 trimers in vitro assembly (15). Mutant NC1 domains of the type IV collagen ␣5 chain also identify NC1 sites for the in vitro assembly of recombinant NC1 monomers into ␣3␣4␣5 NC1 trimers (20). These studies, however, resort to in vitro self-assembly analyses using extracted and highly processed type IV collagen NC1 domains or expression of mutant proteins in cell culture, which might incorrectly recapitulate assembly dynamics of full-length chains in vivo. The underlying mechanism for ␣3␣4␣5 protomer assembly and network formation remains unknown, and it remains unclear whether the NC1 domain plays a definite role in protomer chain selection in vivo.
Despite insights gained from in vitro studies and genetic knock-out mouse model studies, the mechanism for chain selection in protomer network assembly and the role of the NC1 domain in these processes remain to be determined. To gain insights on the mechanism of type IV collagen assembly in vivo in the GBM, 2 we engineered a mouse in which the mouse ␣3NC1 domain was replaced with the human ␣5NC1 domain, keeping the 7 S-and collagenous domains of ␣3 chain intact. This NC1 domain genetic swapping strategy allows for the biological and biochemical analyses of type IV collagen in network assembly.

EXPERIMENTAL PROCEDURES
Cloning of the Mouse ␣3NC1 Knock-out/Human ␣5NC1 Knock-in Targeting Vector-This construct was generated with the aim of removing the NC1 domain from the mouse type IV collagen ␣3 chain and replacing it with the NC1 domain of the human type IV collagen ␣5 chain. A BAC clone (404A6, Invitrogen) containing the end of the mouse ␣3 type IV collagen gene was used to generate the long (4.5 kb) and short (2.1 kb) arms of the construct by PCR. The primers 5Ј-CAG TGC GGC CGC  ATA ATT CTC CCA AAA TAC TTC-3Ј and 5Ј-GAA GCC  GCG GAT TCT TGT ACC AGT GGC CGG CG-3Ј (long arm)  and 5Ј-TCT TAG GTA CCA AAG TCA TGA CTT AGA ACA  TG-3Ј and 5Ј-TGT GCG AAT TCA GTG GAG AAC ATG  AGA GGA TGA TG-3Ј (short arm) were used for the PCR of the long and short arms, respectively. Novel NotI and SacII as well as KpnI and EcoRI restriction sites were included in the primers, thus, generating restriction sites used during subsequent subcloning. The human ␣5 NC1 domain (1.8 kb) was sequenced from a human kidney cDNA library (Stratagene) using the primers 5Ј-CTG TTG CAC CGC GGT TTC TTA  TTA CAC GCC ACA GCC-3Ј and 5Ј-ATC CTA GGA ACA  TAT ATC TTT AAT TAA ATT TAT ATT G-3Ј. The primers contain novel SacII and BamHI restriction sites used for further subcloning purposes. All fragments were subcloned using the TOPO TA Cloning System (Invitrogen) and sequenced to ensure the correct amplification. The SacII restriction site in the 4.5-kb genomic fragment from the mouse ␣3 gene and in the human ␣5 cDNA fragment makes it possible to join the mouse and human sequences and maintain the correct reading frame. At the junction site there is a change from a methioninearginine-aspartic acid amino acid sequence to an isoleucinearginine-glycine sequence before the start of the actual NC1 domain. The human ␣5 cDNA fragment contains an translation termination site and the polyadenylation sequence for the hybrid mouse ␣3NC1/human ␣5NC1 chain. To ensure no transcription beyond the cDNA fragment, which would lead to the addition of mouse ␣3 sequences 3Ј of the hybrid chain, we added transcription termination signals in all three reading frames at the end of the human ␣5 cDNA fragment. The two genomic fragments as well as the human ␣5 cDNA fragment were subsequently cloned into the Triple LoxP vector generat-ing the final construct used for gene targeting (Fig. 1A). After targeting the COL4A3 gene, the NC1 domain of the mouse ␣3 gene was replaced by the novel hybrid allele. The Triple LoxP vector contains Neo and TK cassettes used as selection marker during gene targeting.
Generation and Genotyping of the Mouse a3NC1 Knock-out/ Human a5NC1 Knock-in Mice (A3 m A5 h )-The targeting vector was linearized, purified, and electroporated into embryonic stem cells at the Brigham and Women's Hospital transgenic facility, Boston, MA. G418-resistant embryonic stem cells clones were screened by Southern blot analysis using a 5Ј external probe. This probe was generated by PCR from the 404A6 BAC using the primers 5Ј-GGG AGC ACC AGG TAC CCC CGG TCT TCC CGG3-Ј and 5Ј-TAC AGT AAG GGT ACC CTT AGT GAT GCA GGG CTG-3Ј and subcloned into the Topo vector. A correctly targeted clone will be identified by a 2.1-kb band as compared with the 4.4-kb wild-type allele. 175 colonies were screened, and 3 correctly targeted colonies were identified. Two clones were injected into blastocysts and implanted into pseudopregnant mice, and the resulting founders were bred to C57BL/6J mice to generate inbred mutant strains. Mice were subsequently genotyped by PCR using the forward primer 5Ј-GTC TGG ATG GAT TGC CAG GCT-3Ј and reverse primer 5Ј-CAT CGG CTA ATT GCT GTC CTC-3Ј, which amplify a 600-bp fragment from the mutant allele (knock-in (KI)) containing both mouse ␣3 and human ␣5 sequence. The wild-type (WT) 300-bp fragments was amplified using forward primer 5Ј-CAG CTG CCT GCA GCG ATT CAC-3Ј and reverse primer 5Ј-GGG ACA CGG AGG GAT ACA GTA-3Ј. PCR reactions were carried out under the following conditions: 35 times (95°C for 1 min, 60°C for 1 min, 72°C for 1 min).
Animal Care-The generation and characterization of the renal disease progression in A3KO mice were previously described (25). The mice were backcrossed (10ϩ generations) from 129Sv to C57BL/6J genetic background. A3 m A5 h founder mice were back-crossed to C57BL/6J wild-type mice purchased from Charles River. The resultant F1 progeny with germ line transmission of the A3 m A5 h KI allele were derived from a single line of founder and mated with C57BL/6J wild-type mice and A3 m A5 h heterozygote to obtain A3 m A5 h KI mice. All mice in this study were housed under standard conditions at the Beth Israel Deaconess Medical Center animal facility. All animal studies were reviewed and approved by the Animal Care and Use Committee of the Beth Israel Deaconess Medical Center.
Human Sample-Normal human kidney tissue was obtained from rejected kidney grafts and was provided by Dr. C. Shield III under appropriate patient consent and institutional approval.
RT-PCR and in Situ Hybridization-RNA from kidneys of wild-type and A3 m A5 h KI mice was extracted using TRIzol according to the manufacturer's instructions. After DNase I treatment, cDNA was synthesized using High Capacity cDNA Reverse Transcription kit from AB Biosystems. Forward primer 5Ј-GGT TCA GAA CTG GCA CCC GAG-3Ј and reverse primer 5Ј-CCC GGT GAC CAT GTT CTT AGG C-3Ј were used to amplify the human specific ␣5 NC1 200pb PCR product, and forward primer 5Ј-CGT GGG CCG CCC TAG GCA CCA-3Ј and reverse primer 5Ј-TTG GCC TTA GGG TTC 2 The abbreviations used are: GBM, glomerular basement membrane; KI, knock-in; ECM, extracellular matrix; IP, immunoprecipitation.

NC1 Chain Selection for Type IV Collagen Network Assembly
AGG GGG G-3Ј were used to amplify the 250-bp mouse ␤-actin PCR product, 45ϫ (95°C for 30 s, 60°C for 30 s, 72°C for 30 s). These primers were used as probe for in situ hybridization. The probes were labeled using the Digoxigenin oligonucleotide 3Ј-end labeling kit, second generation (Roche Applied Science). Kidney frozen sections (15 m) on 3-aminopropyltriethoxysilane-coated slides were fixed with 4% paraformaldehyde in PBS for 15 min. The slides were then washed and treated with 0.2 N HCl for 15 min at room temperature and incubated with prewarmed proteinase K (20 g/ml) for 15 min.
The sections were then fixed in 4% paraformaldehyde in PBS, washed, pre-hybridized with hybridization buffer (50% formamide, 2ϫ SSC, 50 mM phosphate buffer, pH 7.0, 1ϫ Denhardt's, 5% dextran) and subsequently hybridized with Digoxigenin labeled probe in 1 g/ml hybridization buffer overnight in 37°C. The slides were then washed in wash buffer (100 mM Tris HCl, pH 7.5, 150 mM NaCl) for 10 min followed by 30-min incubations with 1ϫ blocking buffer (Roche Applied Science) and incubated with 1:5000 diluted anti-Digoxigenin antibody (Roche Applied Science) for 30 min, washed in wash buffer twice for 15 min, and equilibrated with detection buffer (100 mM Tris-HCl, pH 9.5, 100 mM NaCl) for 3 min. The sections were then incubated with color substrate (Roche Applied Science) in detection buffer, and Tris-EDTA buffer was used to stop the reaction.

Light Microscopy Staining and Morphometric Analyses-Kidneys
were harvested and fixed in formalin. The tissues were then embedded in paraffin, and paraffin sections were used for periodic acid-Schiff and Masson-Trichrome staining under standard conditions (Histology Core Facility, Beth Israel Deaconess Medical Center). Morphometric analyses for the histological assessment of renal injury, glomerular sclerosis, tubular atrophy, and interstitial fibrosis, were performed as previously described (26,27).
Immunohistochemistry-Sagittal sections of kidney were embedded in Optimal Cutting Temperature compound and snap-frozen in liquid nitrogen. A mouse-on-mouse (Vector Laboratory) immunodetection kit was used to immunolabel 4 -10-m frozen kidney sections with the monoclonal antibody against human ␣5 NC1 domain (dilution 1:200, Wieslab AB) following the instructions provided by the kit's manufacturer, omitting the avidin/biotin blocking step. These slides were mounted with Vectashield Mounting Media with DAPI (Vector Laboratory) and glass coverslips. Kidney sections were immunolabeled using standard immunostaining methods for the detection of podocin (generated by Dr. Peter Mundel, University of Miami, FL) and ␣1-␣6NC1 domains. Briefly, frozen kidney sections (5 m) were fixed in 0.2 N HCl 20 min at room  DECEMBER 31, 2010 • VOLUME 285 • NUMBER 53 temperature. Slides were then washed twice with phosphatebuffered saline (PBS) and blocked in 2% bovine albumin serum (BSA) in PBS for 1 h at room temperature. The slides were incubated with primary antibody against type IV collagen chains (dilution 1:200, a gift from Dr. Cosgrove, Boys Town National Research Center and were purchased from Dr. Sado, Shigei Medical Research Institute) in primary antibody diluent buffer (1% BSA, 0.1% porcine gelatin, 0.01 M PBS, pH 7.2) for 1 h at room temperature. The slide were then washed twice with PBS and incubated with anti-rabbit FITC-conjugated secondary antibody (dilution 1:1000 in PBS, Jackson ImmunoResearch Laboratories) for 1 h at room temperature. The slides were mounted with Vectashield Mounting Medium with DAPI (Vector Laboratory) and glass coverslips. All slides were analyzed using the Axioskop 2 fluorescent microscope, AxioCam HRC camera, and the Axiovision 4.3 software.

NC1 Chain Selection for Type IV Collagen Network Assembly
Extracellular Matrix (ECM) Protein Extraction and Western Blot Analysis of Type IV Collagen Chain Expression-Mouse kidney, lung, and testes ECM proteins were prepared as previously described (27). Briefly, kidneys were homogenized in PBS with proteinase inhibitors before DNase I digestion in 1 M NaCl. The proteins were then incubated in 2% sodium deoxycholate and collagenase (CLSPA, Worthington Biochemical Corp.)-digested. Type IV collagen hexamers in the supernatant of the collagenase digest were precipitated with 95% ethanol. This technique allows for type IV collagen matrix extraction with nearly no intracellular protein contamination. Before Western blotting analyses, protein concentrations were normalized using BCA assay (Thermo Scientific, according to the manufacturer's directions). The proteins are reduced and denatured in SDS-Laemmli buffer supplemented with ␤-mercaptoethanol, and after SDS-PAGE and transfer onto PVDF membranes, equivalent protein loading was controlled by Coomassie staining (data not shown). Western blot analysis was performed using standard methods and using anti-mouse ␣1-␣6NC1 antibodies (dilution 1:10,000, a gift from Dr. Cosgrove, Boys Town National Research Center and purchased from Dr. Sado, Shigei Medical Research Institute at a 1:500 dilution, as previously described (28)) as well as anti-human ␣5NC1 antibody (dilution 1:10,000, Wieslab AB) in 5% milk in Tris-buffered saline with Tween 20 (TBS-T). Anti-rabbit horseradish peroxidase secondary conjugated antibody was used at a 1:10,000 dilution (Sigma) also in 5% milk in TBS-T. 5% BSA in TBS-T was used to block blots when using the antibodies purchased from Dr. Sado, Shigei Medical Research Institute and diluted in 3% BSA in TBS-T.
Immunoprecipitation-100 g of collagenase-digested ECM proteins from A3 m A5 h KI and human kidneys was resuspended in 1 ml of immunoprecipitation (IP buffer: 50 mM Tris, 150 mM NaCl pH 7.5, 1% Triton-X, and proteinase inhibitors) and incubated overnight with anti-human ␣5 NC1 antibody. 100 l of pre-equilibrated protein A/G PLUS-agarose beads (Santa Cruz) were then added, and samples were incubated for 4 h at 4°C. Control samples were incubated with beads only. The beads were washed 5 times with IP buffer and boiled for 10 min in 100 l of SDS-Laemmli buffer supplemented with ␤-mercaptoethanol before Western blot analyses.
Urine Albumin/Creatinine-Mouse urine samples were collected at 8, 12, and 22 weeks of age. Creatinine concentration was measured using the colorimetric assay Quantichrome (DICT-500) from BioAssays (Hayward, CA) according to the manufacturer's directions. Albumin concentrations were measured using the Mouse Albuminuria ELISA (Bethyl Laboratories, Montgomery, TX) according to the manufacturer's directions.
Electron Microscopy-1 mm 3 of kidney sections were fixed overnight (in 2% glutaraldehyde in 0.1 M cacodylic acid, pH 7.4). Tissue specimens were post-fixed with 1% osmium tetroxide, dehydrated, and embedded in Embed 812 (Electron Microscopy Sciences, Hatfield, PA). The blocks were sectioned at 75 nm on a Leica UCT ultramicrotome (Bannockburn, IL) using a Diatome diamond knife (Electron Microscopy Sciences). Images were taken on a Technia G12 Biotwin TEM (FEI, Hillsboro, OR) equipped with an AMT CCD camera (Advanced Microscopy Techniques, Danvers, MA). All processing, sectioning, and scope work was done in the Electron Microscopy Center, Department of Anatomy and Cell Biology at the Indiana University School of Medicine.
Statistical Analyses-S.E. were calculated, and t test and analysis of variance were used to determine statistical differences. A level of p Ͻ 0.05 was considered statistically significant (*, p Ͻ 0.05).
Molecular Modeling-A model of the mouse ␣1␣2␣1 hexamer was created by mutating the protein chains in the crystal structure of the human placental ␣1␣2␣1NC1 hexamer (PDB ID 1LI1) (29) to the corresponding mouse sequences. Recent work identified specific interactions between the N-terminal 58 amino acids from the human ␣3NC1 and the ␣5NC1 domain. The C-terminal region of the ␣5NC1 domain from amino acids 188 to 227 was also shown to interact specifically with the ␣3NC1 domain (20). Using this information, we generated a model of the mouse ␣3␣4␣5 protomer by mutating the human ␣1 chains A and D to the mouse ␣3 sequence and chain B and E to the mouse ␣5 sequence. The mouse ␣4, chains C and F, were generated by homology modeling using the program MODELLER (30). The three-dimensional structure of the bovine lens capsule ␣2NC1 domain (PDB ID 1T60) (17), which shares 74% identity with the mouse ␣4NC1, was used as the template. The ␣4␣5 h ␣5(IV) model was constructed from the mouse ␣3␣4␣5 model by mutating the sequence of the mouse ␣3 NC1 chains to the human ␣5 NC1 sequence. All side chain conformations were adjusted to minimize inter-and intramolecular clashes using the program COOT (31). The models were superimposed by secondary structure matching (32). . Immunoprecipitation studies. A, the chimeric mouse (ms) ␣3/human (hu) ␣5 chain (composed of the mouse ␣3 7S and collagenous domain and human ␣5NC1 domain)-containing hexamers were subjected to immunoprecipitation using a specific antibody to human ␣5NC1 domain. B-F, the bound and unbound proteins were analyzed by Western blot using antibodies anti-human ␣5NC1 (B), ␣1NC1 (C), ␣2NC1 (D), mouse ␣5NC1 (E), and mouse ␣6NC1 (F). The supernatant of the immunoprecipitation reaction (unbound) was also analyzed, and total kidney ECM from human and KI kidneys was used for positive control. The blots were loaded as follows: 1, human kidney ECM, immunoprecipitated using anti-human ␣5NC1 antibody; 2, human kidney ECM, immunoprecipitation beads only; 3, supernatant from human kidney ECM IP using anti-human ␣5NC1 antibody; 4, supernatant from human kidney ECM IP with beads only; 5, KI kidney ECM immunoprecipitation using anti-human ␣5NC1 antibody; 6, KI kidney ECM IP with beads only; 7, supernatant from KI kidney ECM IP using anti-human ␣5NC1 antibody; 8, supernatant from KI kidney ECM IP with beads only; 9, human kidney ECM; 10, KI kidney ECM. D, dimer; M, monomer; ⌬, degradation product. G, schematic depiction of type IV collagen protomer assembly and type IV collagen network assembly; lack of ␣3NC1 domain (chimeric chain) precludes ␣3␣4␣5 protomer assembly, and the chimeric chain does not assemble into a ␣1␣2␣1/␣5␣6␣5 network assembly but forms a unique type IV collagen network incorporating the chimeric chain, thus, highlighting a role for the 7 S and collagenous domains in network assembly.

RESULTS
Generation of the A3 m A5 h KI Mouse-We generated a DNA construct in which we substituted the genomic mouse ␣3NC1 domain with cDNA encoding the human ␣5NC1 domain (Fig.  1A). After screening of the embryonic stem cell clones by southern blotting and generation of F1 pups from backcrossing founder mice, we developed a genotyping protocol to ensure germ line transmission and genomic integration of the human ␣5NC1 sequence. WT, ␣3NC1 knock-out (A3KO, KO), ␣3NC1 knock-out/human ␣5NC1 knock-in heterozygous (A3 m A5 h Het), and ␣3NC1 knock-out/human ␣5NC1 knock-in homozygous (A3 m A5 h KI) mice tail DNA was purified and analyzed by PCR. The 350-bp DNA sequence (Fig. 1A) for the wild-type allele was amplified from the tail DNA of the WT and heterozygous mice but was not amplified in the homozygous KI and KO mice, confirming the specificity of the primers used to detect the wild-type allele (Fig. 1B). The 600-bp fragment (Fig. 1A) for the KI allele was amplified from the tail DNA of the heterozygous and KI mice but was not amplified in the WT and KO mice, likewise confirming the specificity of the primers used to detect the KI allele (Fig. 1B). The sequence identity of the 600bp fragment was confirmed by sequencing (data not shown).
To confirm expression of human ␣5NC1 in the KI mice, we designed human ␣5NC1 specific primers that allowed for the PCR amplification of human ␣5NC1 from genomic tail DNA of A3 m A5 h heterozygous mice (which contains the human ␣5NC1 cDNA sequence used in the gene construct, Fig. 1C). The expected 200-bp PCR product was obtained from cDNA generated from KI mouse kidney RNA but not from the cDNA generated from WT kidney RNA (Fig. 1C). Furthermore, human ␣5NC1 expression was detected using in situ hybridization in A3 m A5 h KI mouse glomeruli but not in the WT kidney (Fig. 1, D and  E). These results indicate that the integrated human ␣5NC1 sequence in the mouse genome is successfully transcribed in the A3 m A5 h KI mouse kidney. Immunolabeling of frozen kidney tissue sections reveals a specific human ␣5NC1 labeling strictly in the glomeruli of KI mice (Fig. 1, F-M), indicating successful translation of the human ␣5NC1 sequence in the KI mice. Double immunolabeling against podocin, a podocyte-specific antibody, and human ␣5NC1 reveal GBM labeling of the human ␣5NC1 ( Fig. 2A).
Insights into Type IV Collagen Protomer Assembly in the A3 m A5 h KI Mouse-Chain selection in protomer assembly is speculated to be influenced by interactions of NC1 domains between the ␣-chains. However, it is unclear whether the collagenous and 7 S domains play a role in chain selection. We evaluated mouse type IV collagen ␣-chain expression pattern in the glomeruli of WT, KO, and KI mice (Fig. 2, B-G). Our immunolabeling experiments revealed mesangial area labeling of mouse ␣1 and ␣2 chains in all mice, with an increase in ␣2 labeling in the KI glomeruli (Fig. 2, B and C), whereas in the GBM, mouse ␣3 and ␣4 chain labeling was restricted to the WT glomeruli, and ␣3 and ␣4 chain labeling in the KO and KI mice was not detected (Fig. 2, D and E). Using specific anti-mouse ␣5 and ␣6 chain antibodies (that do not cross-react with human chains; data not shown), mouse ␣5 and ␣6 chains were detected in the Bowman's capsule, and mesangium was detected in WT glomeruli and diffusely in the Bowman's capsule, GBM, and mesangium of the KO and KI glomeruli (Fig. 2F), with a marked increase in mesangium and GBM labeling for the ␣6 chain in the KO and KI glomeruli (Fig. 2G). Taken together, these results suggest a different pattern of ␣ chain expression in the glomeruli of the KI mice when compared with the WT and KO mice. The KI glomeruli, in contrast with the WT and KO glomeruli, presents with an increase in mesangial ␣2 and ␣5 chains labeling and a distinct ␣6 GBM immunolabeling pattern.
To identify the role of ␣3NC1 in chain selection in ␣3␣4␣5 protomer formation and network assembly, we examined the collagenase-digested total kidney ECM proteins from WT, KO, A3 m A5 h heterozygous, KI, and human kidneys. We analyzed the extracted NC1 hexamers to reflect on secreted protomers that can assemble into type IV collagen network. The hexamers were then denatured and reduced to obtain NC1 monomers and dimers (Fig. 3A). The WT, KO, KI, and human kidney ECM protein samples were blotted against all six chains of type IV collagen (Fig. 3B). Although the antibodies against mouse ␣1-, ␣2-, and ␣3NC1 cross-react with human chains, the antibodies against ␣4, ␣5, and ␣6NC1 are specific to mouse chains and did not label human kidney ECM (Fig. 3B). The ␣1 and ␣2 chains were detected in all samples, whereas the kidney hexamers from KO and KI mice lacked the ␣3 and ␣4 chains. The ␣5 and ␣6 chains were detected in all mouse kidneys, corroborating our immunolabeling results (Fig. 2,  B-G). Although the antibody against the ␣3 chain cross-reacts with human ␣3 chain (Fig. 3C), mouse ␣3NC1 was not detected in the KI kidney ECM in contrast with WT and heterozygote controls (Fig.  3C). The ECM protein preparations blotted with the antibody specific against the human ␣5NC1 domain. This antibody is highly specific to human ␣5NC1 and does not crossreact with any WT (Fig. 3D) and KO (data not shown) mouse kidney ECM proteins. Human ␣5NC1 dimers and monomers were only detected in the KI and human ECM protein samples (Fig. 3D). These results confirm that the KI mice lack the mouse ␣3NC1 and that the chimeric chain composed of the mouse ␣3 7S-and collagenous domains and the human ␣5NC1 domain assembles into protomers, which can be extracted from hexamers and, thus, clearly indicates assembly into the type IV collagen network in the kidney. The mouse ␣3/human  ␣5 chimeric chain was also detected in the lung and testes basement membrane of the KI mice (Fig. 3D). Importantly, mouse ␣4NC1 domain was not detected in the KI ECM of the kidney, thus indicating that without the ␣3NC1 domain, despite intact ␣3 7 S and collagenous domains in the KI kidneys, the ␣3␣4␣5 protomer does not assemble. The ␣3NC1 domain is, thus, essential for ␣3/␣4 chain incorporation into the ␣3␣4␣5 protomer.
Insights into Type IV Collagen Network Assembly-Our analyses of extracted kidney ECM indicate that the chimeric mouse ␣3/human ␣5 chain translates, organizes itself into protomers, and assembles into a type IV collagen network (Fig. 3, A and D). To characterize the chain composition of the networks containing the chimeric chain (Fig. 4A), we immunoprecipitated human ␣5NC1 containing hexamers and analyzed the chain composition (Fig. 4, B-F). Human and KI kidney ECM were incubated with anti-human ␣5NC1 antibody and protein A/G PLUS-agarose beads (Fig. 4, B-F, lanes 1 and 5) or beads only (Fig. 4, B-F, lanes 2 and 6), and supernatant from IP reactions were collected (Fig. 4, B-D), and compared with human kidney ECM (Fig. 4, B-D) or KI kidney ECM (Fig. 4, E and F). This analysis results in a highly specific pulldown of human ␣5NC1 containing complexes (dimers (D) and monomers (M), Fig. 4B). The chimeric chain, thus, binds to neither ␣1, ␣2, ␣5, nor ␣6 chains (Fig. 4, B-F), indicating that the chimeric chain likely assembles into a unique type IV collagen network containing just the chimeric chain (Fig. 4G). Altogether these results indicate that the GBM of the KI mice is composed of ␣1 and ␣2 chains, assembling into the ␣2␣1␣1/␣2␣1␣1 network, and this network lacks the hybrid mouse ␣3/human ␣5NC1 chain. Although other studies have demonstrated that the full-length human ␣5 chain can assemble into protomers containing mouse achains (8,33), the chimeric chain in the KI mice does not lend itself to ␣2␣1␣1/␣6␣5␣5 networks. Together, our data suggest a crucial role for the ␣3NC1 in the assembly of the ␣4␣3␣5/␣4␣5␣3 network, whereas the 7 S and collagenous domains of the chimeric chain (␣3) does not favor assembly of the chimeric chain into the ␣2␣1␣1/ ␣6␣5␣5 networks (Fig. 4G).
A3 m A5 h KI Mice Present with Progressive Glomerulonephritis Similar to A3KO Mice-The A3 m A5 h heterozygous mice breed normally and give birth to progeny in expected mendelian ratios. The KI mice are fertile and give birth to normal size litter. Their weight, gait, coat color, breathing rate, and activity level appear normal, and histological findings indicate that all organs studied (heart, lung, brain, eye, stomach, pancreas, liver, spleen, gastrointestinal tract, skeletal muscle, skin, and sex organs) except for the kidney appear normal. Kidneys of young mice (8 weeks old) appear normal (data not shown), with onset of glomerular sclerosis, tubular atrophy, and interstitial fibrosis with inflammatory infiltration emerging at 12 weeks of age (data not shown). At 22 weeks of age, kidneys reveal extensive glomerular sclerosis, tubular atrophy, and interstitial fibrosis (Fig. 5A). A3KO and A3 m A5 h KI kidneys present with a similar degree of glomerular sclerosis (Fig. 5B, Table 1), tubular atrophy (Fig. 5C, Table 1), and interstitial fibrosis (Fig. 5D, Table 1).
Renal function tests indicate a progressive loss of renal function in the KI mice, with increased proteinuria (increased albumin/creatinine ratio) and serum creatinine levels (Fig. 5, E and F, Table 1). At 22 weeks of age, both KI and KO mice present with similar proteinuria levels (Fig. 5E). No statistical significant differences in serum creatinine measurements between KO and KI mice were observed at 8, 12, and 22 weeks of age, but serum creatinine was much higher in KI and KO than the WT mice (Fig. 5F). The KI mice die between 23 to 30 weeks of age (Fig. 5G). There was no significant difference in life span between the KO and the KI mice, suggesting that at 22 weeks of age both groups of mice present with a similar renal phenotype, leading to death from loss of renal functions. Electron microscopy analyses revealed normal GBM ultrastructure in kidneys from WT mice (Fig. 5H), with focal thinning and thickening of the glomerular basement membrane with podocyte foot process effacement in the KO (Fig. 5I) and KI (Fig. 5J) mice at 22 weeks of age. Taken together, these results indicate that the KI mice present with GBM defect-associated progressive glomerulonephritis leading to renal failure similar to the KO mice. Both KI and KO mice present a similar phenotype despite GBM type IV collagen chain composition in the KI mice differing from the KO GBM type IV collagen chain composition, with the chimeric chain expressed and assembled into type IV collagen network in the KI mice.
Modeling the Type IV Collagen Assembly in the GBM-Our biochemical analyses from KI mice kidneys provide new insights into type IV collagen assembly, and our results provide evidence for a chain selection mechanism employing a ␣3NC1 domain in the organization of ␣3␣4␣5 protomers. Similarly, our results indicate that the mouse ␣5 chain requires mouse ␣3NC1 to assemble into the ␣3␣4␣5 protomer. Critical amino acids in the NC1 domain likely play a role in the chain selection leading to specific assembly of the ␣3␣4␣5 protomer (20) (Fig.  6, A and B).
We generated molecular models for the mouse ␣1␣2␣1, ␣3␣4␣5, NC1 hexamers and the hypothetical mouse ␣5 h ␣4␣5 protomer containing the human ␣5sequence NC1 substituting the mouse ␣3 NC1 sequence. The mouse ␣1␣2␣1 NC1 hexamer model was made using the crystal structure of the human placental ␣1␣2␣1 trimer (PDB ID 1LI1) (29), which shares 95.2% sequence identity (96.8% similarity) with the mouse ␣1␣2␣1 NC1 trimer as a template. Given the highly similar sequence identity, we simply mutated the side chains to the corresponding mouse sequences. Recent work identified specific interactions between the N-terminal 58 amino acids of the human ␣3 NC1 domain and ␣5 NC1 domain. The C-terminal region of the ␣5 NC1 domain from amino acids 188 -227 was also shown to interact specifically with the ␣3 NC1 domain (20). Using this information we generated a mouse ␣3␣4␣5 model by mutation and homology modeling. The human ␣1 sequence shares 71.2% sequence identity (83.4% similarity) with the mouse ␣3 and 81.2% sequence identity (90.8% similarity) mouse ␣5 sequence; therefore, we mutated the human ␣1 chains A and D of the human placental ␣1␣2␣1 protomer x-ray structure to the mouse ␣3 sequence and ␣1 chains B and E to the mouse ␣5 sequence. The mouse ␣4 model was made by homology modeling using the program MODELLER (30) with the x-ray structure of the bovine lens capsule ␣2 NC1 domain, which shares 72.5% identity (82.5% similarity) as the template model (PDB ID 1T60) (17). Additionally, we analyzed the quality of the mouse ␣1␣2␣1 and ␣3␣4␣5 models using the program ProSA II. The energy profile diagrams for the mouse ␣1␣2␣1 and ␣3␣4␣5 models were plotted as a function of residue number for each chain and superimposed on the corresponding chains from the crystal structure of the human pla- Arg-1661 is an exposed amino acid. Mutation to a cysteine is not predicted to have any effect on the folding of ␣3 or its interaction with other chains in the protomer. It is possible, however, that mutation to a cysteine may lead to mismatched disulfide bond formation and protomer instability. ␣4 C1634S Cys-1634 is also absolutely conserved and forms a disulfide bridge with Cys-1641. The mutation C1634S would lead to instability in the ␣4 NC1 domain secondary structure. P1572L The P1572L mutation is likely to give rise to an altered path for the ␣4 main chain leading to misfolding of the protein.

R1511H
Undetermined significance. P1517T Pro-1517 is located at the N-terminal end of ␤6 and servers to redirect the main chain. Mutation to a threonine would be expected to change the orientation of the backbone and disrupt the secondary structure. W1538R,S Trp-1538 is in the ␤8 strand and stabilizes the hydrophobic interior of the ␣5 NC1 chain by packing against Phe-1513, Leu-1489, Leu-1510, and Ile-1561. The shorter hydrophilic side chain of serine would be insufficient to fill the void left by the tryptophan and have a destabilizing effect on the secondary structure. Mutation to arginine would introduce a hydrophilic side chain into a hydrophobic pocket and sterically clash with several hydrophobic side chains.

C1564S
Removes a disulfide bond. C1567R Removes a disulfide bond. C1586F,R Removes a disulfide bond. W1590G The side chain of Trp-1590 fills a hydrophobic pocket occupied by Ile-1574, Val-1572, and Lys-1683 from the ␣5 chain. Glycine would be insufficient to occupy this space. Trp-1590 also makes a hydrogen bond from N⑀1 with the side chain Glu-1552 from the ␣3 chain, which would be lost on mutation to a glycine. G1596D Undetermined significance. L1649R Introduces a steric clash with Met-1600, Lsu-1614 of ␣5, and Leu-1510 of ␣3.

R1677P
This mutation likely leads to instability of the ␤-sheet secondary structure due to insertion of a proline residue in a ␤-strand.

C1678W
Removes a disulfide bond. C1681F Removes a disulfide bond. DECEMBER 31, 2010 • VOLUME 285 • NUMBER 53 cental ␣1␣2␣1 (supplemental Figs. 1 and 2). These data show that there are no major errors in the modeling. We identify in the mouse ␣1␣2␣1 protomer (Fig. 7A) important hydrogen bonds (Fig. 7B) and hydrophobic surface interactions (Fig. 7C) between mouse ␣1 chain and mouse ␣2 NC1 domains. The hydrogen bonding interactions are possible between the side chain NZ and Lys-56 in the mouse ␣2 chain and the main chain of Asp-124, Gln-23, and Ile-196. The carbonyl and amides of Arg-55 and Lys-56 also make hydrogen bonds with the side chain OE1 of Gln-123 (Fig. 7B). Favorable hydrophobic surface interactions are also possible between the mouse ␣1 a and mouse ␣1 b chains involving the side chains of Met-60 with Cys-65 in the ␣2(IV), Val-144 with Thr-147, and Ala-188 with Leu-193 (Fig. 7C). Based on this analysis, we identify specific amino acids interactions in that could prevent assembly of the ␣5 h ␣4␣5 protomer. In the mouse ␣3␣4␣5 protomer model (Fig. 7D), it is possible for Ile-144 from the ␣3 chain to insert into a shallow hydrophobic pocket composed of Leu-158 and Leu-193 from the ␣3 chains and Met-36, Leu-33, and His-34 from the ␣4 chain. In the hypothetical ␣5 h ␣4␣5 model, a serious hydrophobic clash between Met-114 of the ␣5 h chain and the side chains of Met-36 from the ␣4 chain can result in a destabilizing effect possibly averting the formation of this protomer combination.

NC1 Chain Selection for Type IV Collagen Network Assembly
Using the information on amino acid substitution mutations reported in patients with Alport syndrome (Fig. 6B, Table 2), we analyzed the structural changes in the NC1 domain folding and interaction in the KI mice. Using the contact interface information in the NC1 domains of the mouse ␣3␣4␣5 and mouse ␣1␣2␣1 protomers, we highlight specific amino acid residues that appear critical in stabilizing bonds between NC1. Analyses of the contact interfaces in human ␣5/mouse ␣4␣5 NC1 interactions reveal hydrophobic clashes, which would in part avert assembly of this chimeric protomer.

DISCUSSION
Type IV collagen network is a major component of basement membranes, and their specific chain composition and organization into unique networks are critical in determining specific molecular properties and specialized functions (1-3, 11, 21). The mechanism underlying the self-assembly and intrinsic chain selection of type IV collagen protomer assembly is unknown and highlights its critical importance in type IV collagen-associated pathologies such as Alport syndrome.
In this study we generated and characterized of a novel mouse that enables to study the functional role of the NC1 domain in chain selection of protomers. The A3 m A5 h KI mice possesses all normal ␣-chains of type IV collagen except for the ␣3 chain, in which the mouse ␣3NC1 domain is substituted with the human ␣5NC1 domain. However, unlike the ␣3KO mice, the KI mice express a chimeric mouse ␣3/human ␣5 chain that successfully assembles to form a protomer within the type IV collagen network. In agreement with previous studies demonstrating podocytes as the main source of the ␣3 chain in the glomeruli (36), the chimeric chain is observed in podocytes and the GBM of the KI mice. Our NC1 domain genetic swapping approach provides a definitive evidence for NC1 domainmediated chain selection in protomer formation. The ␣3NC1 domain is required for ␣4 chain selection in ␣3/␣4 chain assembly in ␣3␣4␣5 protomer and network formation. We, thus, speculate that the loss in the ␣4NC1 domain (with the knock-in of the human ␣5 NC1 domain, ␣4/␣5 hu chimeric chain) in the ␣4 chain is expected to provide a similar phenotype as observed in the ␣3/␣5 hu chimeric chain.
In conclusion, our results provide the first in vivo evidence for NC1-mediated chain selection in the type IV collagen protomer assembly. The loss of ␣3NC1 domain results in kidney disease with progressive glomerulonephritis leading to renal failure. This pathological finding underlines the type IV collagen assembly defect as the underlying mechanism for the GBM defect associated with Alport syndrome. Although the A3 m A5 h KI and A3KO GBM type IV collagen composition differs, the similar renal phenotype observed in both models underlines the critical and unique biological properties of the ␣3␣4␣5 protomer in sustaining glomerular filtration.