Quaternary Organization of the Goodpasture Autoantigen, the α3(IV) Collagen Chain

Goodpasture's (GP) disease is caused by autoantibodies that target the α3(IV) collagen chain in the glomerular basement membrane (GBM). Goodpasture autoantibodies bind two conformational epitopes (EA and EB) located within the non-collagenous (NC1) domain of this chain, which are sequestered within the NC1 hexamer of the type IV collagen network containing the α3(IV), α4(IV), and α5(IV) chains. In this study, the quaternary organization of these chains and the molecular basis for the sequestration of the epitopes were investigated. This was accomplished by physicochemical and immunochemical characterization of the NC1 hexamers using chain-specific antibodies. The hexamers were found to have a molecular composition of (α3)2(α4)2(α5)2 and to contain cross-linked α3-α5 heterodimers and α4-α4 homodimers. Together with association studies of individual NC1 domains, these findings indicate that the α3, α4, and α5 chains occur together in the same triple-helical protomer. In the GBM, this protomer dimerizes through NC1-NC1 domain interactions such that the α3, α4, and α5 chains of one protomer connect with the α5, α4, and α3 chains of the opposite protomer, respectively. The immunodominant Goodpasture autoepitope, located within the EA region, is sequestered within the α3α4α5 protomer near the triple-helical junction, at the interface between the α3NC1 and α5NC1 domains, whereas the EB epitope is sequestered at the interface between the α3NC1 and α4NC1 domains. The results also reveal the network distribution of the six chains of collagen IV in the renal glomerulus and provide a molecular explanation for the absence of the α3, α4, α5, and α6 chains in Alport syndrome.

The glomerular basement membrane (GBM) 1 is a key component of the kidney ultrafiltration barrier. Type IV collagen, the main structural component of the GBM, has been implicated in several glomerular diseases. Goodpasture's (GP) disease is caused by anti-GBM autoantibodies that bind to the type IV collagen networks of the GBM, and, in some patients, of the alveolar basement membranes (1). The bound antibodies induce an inflammatory response causing rapidly progressive glomerulonephritis and pulmonary hemorrhage. The GP autoantibodies are specifically targeted to the non-collagenous domain (NC1) of the ␣3(IV) chain, the "Goodpasture autoantigen," one of the six chains that comprise the collagen IV family (2). Two conformational GP epitopes, designated E A and E B , have been localized within the ␣3(IV) NC1 domain at residues 17-31 and 127-141, respectively (3). The GP epitopes are sequestered in the type IV collagen network of native basement membranes by quaternary interactions among NC1 domains (4).
How the ␣3(IV) chain is organized in the collagen IV network of GBM is unknown. Collagen IV networks are composed of triple-helical protomers that are connected at the carboxyl termini by NC1-to-NC1 interactions, forming dimers of protomers, and at the amino termini by interactions involving the 7 S domain, forming tetramers (5,6). In the GBM, an ␣1⅐␣2(IV) network and a distinct ␣3⅐␣4⅐␣5(IV) network have been identified based on differential solubilization with pseudolysin (7) and analysis of collagenase-solubilized NC1 hexamers (8). The protomer and network organization of ␣1(IV) and ␣2(IV) chains has been well established and confirmed by the recent determination of the crystal structure of the [(␣1) 2 ␣2] 2 (IV) NC1 hexamer (9,10). In contrast, little is known about the organization of ␣3(IV), ␣4(IV), and ␣5(IV) chains, i.e. which combinations of three chains form protomers, what is the relative position of chains within protomers, and which kinds of protomers associate through NC1-to-NC1 interactions. This information is essential for understanding the molecular basis of GP disease.
The GP epitopes are sequestered within the NC1 hexamer structure of the ␣3⅐␣4⅐␣5(IV) network and inaccessible for autoantibody binding unless the hexamer dissociates (4,11). Unmasking the GP epitopes is thought to be critical for etiology and pathogenesis of GP disease, but the molecular basis for the epitope sequestration is not known (12). Based on the identification of hydrophobic residues in the epitope of the immunodominant GP A autoantibodies, it has been proposed that the epitope is buried at the interface between interacting NC1 domains (13). An analysis of the homologous ␣1⅐␣2(IV) NC1 hexamer structure revealed that the GP epitopes are located adjacent to the triple-helical domain and to the interfaces between NC1 monomers within a protomer (9,10).
In the present study, the quaternary organization of the ␣3(IV) chain in relation to the ␣4(IV) and ␣5(IV) chains in the network was investigated to determine: (a) which chains coexist and their relative position in the triple-helical protomers, (b) how the protomers are connected through the NC1 domains, and (c) how the GP epitopes are sequestered. The results unambiguously identify a single quaternary organization of the ␣3⅐␣4⅐␣5 NC1 hexamer revealing that the ␣3, ␣4, and ␣5 chains coexist in a single protomer that self-associates through NC1-NC1 domain interactions to form a network. Moreover, the immunodominant GP epitope, located within the E A region, is sequestered at the ␣3-␣5 NC1 interface within the protomer, which partially buries residues required for autoantibody binding, whereas the E B epitope is sequestered at the ␣3-␣4 NC1 interface.
Based on these findings, we propose that the designation of Goodpasture autoantigen should henceforth refer to the ␣3⅐␣4⅐␣5(IV) triple-helical protomer, rather than to the ␣3(IV) collagen chain, as this molecules reflects both the organization and the cryptic property of the autoantigen in the GBM.

EXPERIMENTAL PROCEDURES
Proteins-Human glomeruli isolated from human kidneys by differential sieving were used to purify human GBM, which was solubilized by digestion with bacterial collagenase (Calbiochem), and then fractionated by ion-exchange and gel filtration to purify the NC1 hexamers of type IV collagen (14). Recombinant (r-) human ␣3, ␣4, and ␣5 NC1 domains with NH 2 -terminal FLAG epitope were expressed in human kidney 293 cells and purified by affinity chromatography using immobilized anti-FLAG M2 antibody (Sigma), as described previously (15). The recombinant NC1 domains were previously shown to be correctly folded and able to interact with each other (8).
Monoclonal Antibodies and Immunoaffinity Fractionation of NC1 Hexamers-For detection of human ␣1-␣6 NC1 domains in Western blot and ELISA, as well as for immunoprecipitation of dissociated NC1 hexamers (see below), rat monoclonal antibodies (mAbs) H11 (to ␣1), H22 (to ␣2), H31 (to ␣3), H43 (to ␣4), H52 (to ␣5), and H63 (to ␣6), previously described (16), were used, along with mouse Mab1 (to ␣1), Mab3 (to ␣3), and Mab5 (to ␣5), which were purchased from Wieslab AB (Lund, Sweden). A new monoclonal antibody specific for the E B region of ␣3 NC1 domain, designated mAb EB3, was produced by immunizing mice with human r-␣3 NC1 monomers. Its specificity was determined by ELISA against a panel of ␣1/␣3 NC1 chimeras (3); mAb EB3 bound only the C6 chimera that contained the E B region (residues 127-141) of ␣3 NC1 within an ␣1 NC1 scaffold. Several properties of mAb EB3 resemble those of human GP B autoantibodies: (a) its epitope is encompassed by the E B region, (b) it is cryptic in the NC1 hexamer, and (c) it is conformational. A full description of this antibody will be presented elsewhere. 2 Several NC1 hexamer-binding mAbs were used for the immunoaffinity fractionation of native NC1 hexamers from human GBM. Mab1 and Mab3 were used for immunoprecipitation of ␣1and ␣3-containing native NC1 hexamers, and immobilized rat mAbs B51 (to ␣5) and B66 (to ␣6) were used for affinity chromatography of hexamers containing the ␣5 and ␣6 chains, respectively, as described (17). In addition, several novel hexamer-binding mAbs specific for the ␣4 NC1 domain were produced by the rat lymph node method (18) in Wistar-Kyoto rats immunized with either human r-␣4 NC1 domain (mAbs RH42 and RH45) or with total human GBM hexamers (mAb N42). The specificities of these mAbs were determined by ELISA using human ␣1-␣6 NC1 domains and all were found to be specific for human r-␣4 NC1 monomers. Moreover, in indirect immunofluorescence of human kidney sections, mAbs RH42, RH45, and N42 exhibited a glomerular staining pattern identical to that of anti-␣4 mAb H43. However, whereas mAb H43 recognized cryptic epitopes and required pretreatment of tissue sections with acid urea, the newly produced mAbs stained native sec-tions, indicating that they bind to epitopes exposed in the NC1 hexamer. For further use in the affinity fractionation of NC1 hexamers, RH45 IgG was purified on protein G-Sepharose (Amersham Biosciences) and immobilized to Affi-Gel-10 (Bio-Rad), as described for mAbs B51 and B66 (17).
Analysis of the NC1 Dimers of Human GBM by Immunoprecipitation-Human GBM NC1 hexamers (2-10 g) were dissociated under various experimental conditions and immunoprecipitated with specific antibodies against ␣1-␣5 NC1 domains (10 -20 g). The NC1 domains bound to the precipitating antibodies were collected on protein G-Sepharose, separated by SDS-PAGE in replicate gels, and analyzed by Western blots for the presence of ␣1-␣5 NC1 domains. The choice of dissociation conditions was important, because complete dissociation of the NC1 hexamers into subunits had to be achieved without destroying the epitopes, to allow subsequent binding of the precipitating antibodies to the NC1 monomers and dimers. The optimal conditions were empirically determined to be heating with 0.5% SDS for 10 min at 60°C, followed by 10 -20-fold dilution with Tris-buffered saline containing 1 mg/ml bovine serum albumin. The small amount of SDS remaining did not interfere with the binding of anti-NC1 antibodies, but was sufficient to prevent further non-covalent interactions among the dissociated NC1 domains, as demonstrated by the identification of a single species of NC1 monomers in the immunoprecipitate.

Triple Helical and Network Organization of the ␣3(IV)
Chain, the Goodpasture Autoantigen-In the type IV collagen suprastructure, two triple-helical protomers interact at the carboxyl terminus via their NC1 domains, forming a stable NC1 hexamer complex that can be excised from basement membranes by collagenase digestion. Because of the special position of NC1 hexamers, connecting two adjoining protomers, the identity of monomer and dimer subunits of the hexamers reflects the chain composition of protomers, and the dimer subunits identify which chains are connected by covalent interactions between protomers. A strategy perfected for the analysis of NC1 hexamers from smooth muscle basement mem-branes that led to the discovery of a novel ␣1⅐␣2⅐␣5⅐␣6(IV) network (17) was used here to analyze the collagen IV network containing the ␣3, ␣4, and ␣5 chains.
The hexamer population from a preparation of human glomeruli contains all six chains of type IV collagen (Fig. 1, i). Immunohistochemical studies have established that the GBM contains the ␣1-␣5(IV) chains, Bowman's capsule basement membrane contains the ␣1, ␣2, ␣5 and ␣6(IV) chains, and the mesangial matrix comprises the ␣1 and ␣2(IV) chains (16,20). Hence, hexamers containing ␣3 and ␣4 (along with ␣5) origi-nate exclusively from the GBM, whereas those containing ␣6 originate exclusively from Bowman's capsule basement membrane. The ␣6-containing hexamers were removed from the mixture by absorption to an anti-␣6 affinity column (mAb B66). The B66-bound fraction, representing ϳ1% of the total hexamers, contained the ␣6 NC1 domain along with the ␣1, ␣2, and ␣5 NC1 domains (Fig. 1, ii). This composition, together with the mobility of dimers, indicate that the B66-bound hexamer population is identical to that identified in the smooth muscle basement membranes of aorta and bladder (17). Thus, the Bowman capsule contains the same ␣1⅐␣2⅐␣5⅐␣6(IV) network found in smooth muscle.
This composition could reflect 18 distinct combinations in which two protomers (trimers of chains), varying in both chain composition and relative position of chains, interact in different trimer-trimer orientations forming a cross-linked hexamer (Fig. 2). For example, in group I all protomers contain an ␣3, ␣4, and ␣5 chain, but the combinations differ with respect to the trimer-trimer orientations (as in A1, A2, and A3) and with respect to the orientation of the three chains in the protomer (as in A1, B1, and C1), for a total of nine combinations. In group II, two different protomers, differing in chain composition and different from group I, associate forming three additional com-FIG. 1. Analysis of type IV collagen from the human GBM by immunoaffinity purification the NC1 hexamers. The NC1 hexamers solubilized from human GBM were fractionated using Mab3, RH45, B51, and B66 mAbs (to ␣3-␣6 NC1 domains, respectively), then analyzed by Western blot for the presence of ␣1-␣6 NC1 domains, using mAbs H11 or Mab1 (␣1), H22 (␣2), H31 or Mab3 (␣3), H43 (␣4), H53 or Mab5 (␣5), and H63 (␣6). Each panel (i-vii) represents the Western blot analysis of a different hexamer fraction; after transfer, each of the six replicate lanes was stained with a different mAbs to detect the presence of ␣1-␣6 NC1 domains. The total GBM hexamer preparation (i) contained all six NC1 domains as a mixture of dimers (d) and monomers (m) forms. The hexamers bound to mAb B66 (ii) contained ␣1, ␣2, and ␣5 and ␣6 NC1 domains. The hexamers not bound to mAb B66 (iii), containing the ␣1-␣5 NC1 domains, were further immunofractionated using Mab3 (iv), RH45 (v), or B51 (vi); all three antibodies bound ␣3, ␣4, and ␣5 NC1 monomers and dimers. The NC1 hexamers not bound to Mab3 (vii) consisted of only ␣1 and ␣2 NC1 domains; similar results were obtained with mAbs RH45 and B51 (data not shown).
binations. Similarly, groups III and IV represent six additional combinations.
Analysis of Interprotomer Interactions-Many of these combinations can be excluded by the identification of NC1 monomers that are cross-linked forming dimers. In most basement membranes, including the GBM, the NC1 hexamer contains a large proportion of NC1 dimers (ϳ75%) cross-linked by a covalent bond that connects two adjoining protomers in the network. Previously, this cross-link was identified as a disulfide bond (21), but recent crystallographic studies suggested a cross-link between the conserved Met-93 and Lys-211 residues (10). Because each NC1 domain of a protomer can be crosslinked to the corresponding NC1 domain of the opposite protomer, each hexamer is characterized by three NC1 dimers that reflect specific trimer-trimer interactions. Thus, the identity of dimers that exist in the ␣3⅐␣4⅐␣5 heterohexamer reflects which chains are connected between protomers. Overall, the 18 hexamer isoforms shown in Fig. 2 contain six possible dimer combinations. These are: ␣3-␣3, ␣4-␣4, and ␣5-␣5 NC1 homodimers and ␣3-␣4, ␣3-␣5, and ␣4-␣5 NC1 heterodimers. The absence of some of these NC1 dimer combinations would rule out the existence of certain hexamers isoforms. Two independent methods were used for this determination.
Analysis of Interprotomer Interactions-Further discrimination among the remaining possibilities required an analysis of the intraprotomer interactions among ␣3, ␣4, and ␣5 NC1 domains. The recently solved crystal structure of the [(␣1) 2 ␣2] 2 NC1 hexamer (9, 10) provides a framework for this analysis. Due to the high sequence homology among the six NC1 domains (52-83% identity), the ␣3⅐␣4⅐␣5 NC1 hexamer must have a similar tertiary and quaternary structure. The NC1 hexamer is formed of two identical trimeric caps, each derived from one protomer, that interact through a large planar surface (one trimeric cap is shown in Fig. 5a). Within each trimeric cap, three NC1 domains (shown in red, green, and blue) are arranged radially around a pseudo-3-fold symmetry axis (Fig.  5b). Each NC1 domain is composed of two homologous subdomains, an amino-terminal subdomain A (lighter shade) and a carboxyl-terminal subdomain B (darker shade). As illustrated for the ␣3 NC1 domain, within the NC1 trimer, subdomain A of each monomer interacts with subdomain B of the next monomer. Thus, an ␣3 NC1 monomer has two distinct neighbors, "x" and "y," interacting with its subdomains A and B, respectively. The identity of x and y distinguishes among the four remaining possible isoforms of the ␣3⅐␣4⅐␣5 hexamer actually exist (Fig. 5c).
Their identities were determined using antibodies targeted to different regions of the ␣3 NC1 domain, which block the interactions of either x or y with the ␣3 interface. Our previous studies established that recombinant human ␣3, ␣4, ␣5 NC1 monomers associate with each other in vitro and that all three are required to form an NC1 hexamer (8). Here, we showed that several antibodies to the ␣3 NC1 domain could selectively block the interaction of ␣3 NC1 with either ␣4 or ␣5 NC1 domains, but not both. Because the epitopes of these antibodies are known and they are located in the proximity of either x or y NC1 domains (Fig. 6, top), this information could be used to identify x and y and, therefore, the relative orientation of ␣3, ␣4, and ␣5 chains within the protomer. GP A auto-Abs bind to the E A region (residues 17-31) (3), located proximal to the x, but not y, monomers. mAb EB3 specifically recognizes the E B region (residues 127-141), and the epitope of mAb H31 was mapped to residues 208 -214 (IPSTVKA) (18), both located proximal to the y, but not x, monomers. The epitope of Mab3 is jointly formed by the E A and E B regions (4). Binding of these antibodies to the ␣3 NC1 monomer would either allow or prevent the interactions of the ␣3 NC1 with the ␣4 and/or ␣5 NC1 monomers.
The ability of anti-␣3 NC1 antibodies to block the interaction between soluble ␣3 NC1 monomers and immobilized ␣4 and ␣5 NC1 monomers was studied by ELISA, using the appropriate secondary antibodies to detect the formation of a trimolecular complex. Control experiments showed that in the absence of ␣3 NC1, anti-␣3 antibodies did not react with immobilized ␣4 or ␣5 monomers (Fig. 6, solid bars), indicating that when an ␣3 NC1-antibody complex bound to the immobilized ␣4 and/or ␣5 NC1 domains (Fig. 6, gray bars), the interaction was mediated by the ␣3 NC1 domain. When in complex with Mab3, the ␣3 NC1 domain could interact with both ␣4 and ␣5 NC1 monomers FIG. 4. Analysis of the NC1 dimers from human GBM by immunoprecipitation. NC1 hexamers from human GBM were dissociated (top), then analyzed by immunoprecipitation with chain-specific antibodies. The NC1 hexamer subunits precipitated by antibodies against ␣1 (H11), ␣2 (H22), ␣3 (GP A ), ␣4 (H43), or ␣5 (H52) NC1 domains were analyzed by Western blot (middle panels). Each panel represents the Western blot analysis of NC1 domains precipitated by the antibody indicated on the right side; each of the five lanes within a panel was stained with a distinct specific mAb against the ␣1-␣5 NC1 domains (as shown in the top panel). The identities of the NC1 monomers (m) and dimers (d) bound to each antibody, as revealed by this analysis, are depicted on the right side of each panel. The identification of ␣3-␣5 and ␣4-␣4 NC1 dimers, along with the absence of ␣3-␣4 or ␣5-␣4 NC1 dimers (and the possible presence of ␣3-␣3 and ␣5-␣5 NC1 dimers), together rule out all but four isoforms of (␣3) 2 (␣4) 2 (␣5) 2 NC1 hexamers, corresponding to A1, B1, C1, and D1 in Fig. 2.  (Fig. 6a). This confirms the previous observation that ␣3 NC1 monomers can form a binary complex with either ␣4 and ␣5 NC1 monomers (8). Furthermore, Mab3 does not hinder either of these interactions, which is consistent with the accessibility of Mab3 epitope in the ␣3⅐␣4⅐␣5 hexamer (4,8,25).
In contrast to Mab3, the complex between GP A autoantibodies and ␣3 NC1 monomers could interact with the ␣4, but not with the ␣5 NC1 monomers (Fig. 6b). Thus, GP A antibodies block specifically the interaction of ␣3 NC1 with ␣5 NC1, but not with ␣4 NC1. Because their epitope, the E A region of ␣3 NC1, is proximal to the x, but not to y, NC1 domain, it follows that ␣5 is not allowed at position y and hence it must occupy position x, whereas ␣4 is allowed at position y (whether ␣4 could be x is undetermined). This rules out ␣3⅐␣4⅐␣5(IV) isoforms B1 and C1 (Fig. 5). Conversely, when in complex with mAb EB3 (Fig. 6c) or mAb H31 (Fig. 6d), the ␣3 NC1 monomer interacted with ␣5, but not with ␣4 NC1 monomers. This indicates that mAbs EB3 and H31 block the interaction of ␣3 NC1 with ␣4 NC1, but not with ␣5 NC1. Because their epitopes are proximal to the y NC1 domain, but not to x, it follows that ␣4 is not allowed at position x and hence must occupy position y, whereas ␣5 is allowed at position x. This further rules out ␣3⅐␣4⅐␣5(IV) isoforms B1, C1, and D1 (Fig. 5), leaving only one possible isoforms, A1. Together, these data indicate that the ␣3, ␣4, and ␣5 chains are arranged in the protomer such that subdomain A of the ␣3 NC1 domain interfaces with subdomain B of the ␣5 NC1 domain, whereas subdomain B of ␣3 NC1 interfaces with subdomain A of the ␣4 NC1 domain.

Organization of Chains within the ␣3⅐␣4⅐␣5(IV) Network-
The NC1 domains play a critical role at multiple stages in the assembly of collagen IV networks. First, interactions among NC1 domains initiated assembly of three chains into a triplehelical protomer (26). Second, NC1 domains mediate the association of two protomers head-to-head, forming at the junction an NC1 hexamer, which in turn is stabilized by cross-links (5). Third, the NC1 domains encode the specificity of interactions among the ␣1-␣6(IV) chains at two levels: (a) the selection of FIG. 5. Organization of the NC1 hexamers. The x-ray crystal structures of the [(␣1) 2 ␣2] 2 NC1 hexamers from the lens capsule (9) and placenta (10) basement membranes reveal that the NC1 hexamers are composed of two trimeric caps, each derived from one protomer, interacting through a planar interface (arrow). A lateral view (a) and a view from the trimer-trimer interface (b) show that within each trimeric cap, the three NC1 domains (shown in red, green, and blue, respectively) are arranged radially such that subdomain A (lighter color) of one chain interacts with subdomain B (darker color) of the next chain. The high homology between ␣1-␣6 chains indicates that the ␣3⅐␣4⅐␣5 NC1 hexamer must have tertiary and quaternary structures similar to the [(␣1) 2 ␣2] 2 NC1 hexamer, but the relative positions of the ␣3, ␣4, and ␣5 NC1 remained to be determined. The NC1 domains interacting with subdomains A and B of the ␣3NC1 were designated x and y, respectively. The four possible isoforms of the ␣3⅐␣4⅐␣5 NC1 hexamer differ in the identities of x and y; the prime symbol was used to distinguish between the two ␣3 NC1 monomers of the ␣3⅐␣4⅐␣5 hexamers (c).
FIG. 6. Anti-␣3 NC1 antibodies selectively block binding of the ␣3 NC1 monomers to either ␣4 NC1 and ␣5 NC1 monomers. The epitopes of several anti-␣3 antibodies are located in the proximity of either x or y NC1 domains that interact with an ␣3 NC1 domain (top). The ability of these antibodies to block the interaction of ␣3 with ␣4 or ␣5 NC1 domains was measured by ELISA. The r-␣3, r-␣4, and r-␣5 NC1 monomers were immobilized overnight onto microtiter plates (500 ng/ well in 50 l of carbonate buffer). Binding of anti-␣3 NC1 antibodies Mab3 (a), GP A autoAbs (b), mAb EB3 (c), and mAb H31 (d), in complex with r-␣3 NC1 monomers (0.5 g/ml), was measured using appropriate secondary antibodies (gray bars). In control experiments, anti-␣3 antibodies alone did not interact with ␣4 or ␣5 NC1 domains (solid bars). In complex with Mab3, ␣3NC1 interacted with both ␣4 and ␣5 NC1 monomers. In complex with GP A auto-Abs, ␣3NC1 interacted with ␣4 but not ␣5 NC1 monomers. In complex with EB3 or H31, ␣3NC1 interacted with ␣5 but not ␣4 NC1 monomers. Together, these results indicate that x is the ␣5 NC1 domain, and y is the ␣4 NC1 domain. chains for protomer assembly and the relative orientation of the three chains within the protomer and (b) the selection and relative orientation of the two adjoined protomers (8,17). Thus, the quaternary structure of the NC1 hexamer reflects the molecular interactions governing the affinity, organization, and specificity of network assembly. Previous studies of NC1 hexamers have revealed the existence of three distinct collagen IV networks: an ubiquitous one containing the ␣1 and ␣2(IV) chains; another containing the ␣3, ␣4 and ␣5(IV) chains, found in the GBM, and other containing the ␣1, ␣2, ␣5, and ␣6(IV) chains, found in the smooth muscle basement membranes. The organization of chains at the protomer and network level has been determined for the ␣1⅐␣2(IV) and ␣1⅐␣2⅐␣5⅐␣6(IV) networks, but not for the ␣3⅐␣4⅐␣5(IV) network.
In the present study, the chain organization of the ␣3⅐␣4⅐␣5(IV) network of the GBM was determined to define the quaternary organization of the GP autoantigen, the ␣3(IV) chain. This was accomplished using a combination of techniques, including immunochemical analysis of native and dissociated NC1 hexamers, two-dimensional electrophoresis, and analysis of the specificity of interactions between NC1 monomers. The results establish that the ␣3(IV) chain, along with an ␣4(IV) and an ␣5(IV) chain, exist as a single triple-helical protomer, which self-associates forming a dimer through NC1to-NC1 interactions (Fig. 7A). At the protomer level, the aminoterminal subdomain A of the ␣3 NC1 domain interfaces with the carboxyl-terminal subdomain B of the ␣5 NC1 domain, and its carboxyl-terminal subdomain B interfaces with the N-terminal subdomain A of the ␣4 NC1 domain. At the network level, the NC1 trimer-trimer interface between two adjoining protomers is formed by interactions between the ␣3 NC1 domain of one protomer and the ␣5 NC1 domain of the other and between the ␣4 NC1 domains of each protomer; these connections are stabilized by cross-links.
The existence and distribution of these networks provide a molecular explanation for the absence of the ␣3, ␣4, ␣5, and ␣6(IV) chains from the glomerulus in Alport syndrome, a hereditary form of progressive renal disease caused by mutations in collagen IV genes. In the X-linked form of the disease, mutations in the COL4A5 gene encoding the ␣5(IV) chain cause the absence of the ␣3, ␣4, and ␣5(IV) chains from the GBM, despite the expression of the ␣3 and ␣4(IV) chains in the podocytes (28,29), and the absence of the ␣5 and ␣6(IV) chains from Bowman's capsule (30 -32). Thus, mutations in the ␣5(IV) chain could either prevent the assembly of ␣3␣4␣5 and (␣5) 2 ␣6 protomers altogether or result in defective protomers that are degraded or cannot self-assemble into networks, causing the loss of the ␣3-␣6(IV) chains. In comparison, mutations in the ␣3(IV) or ␣4(IV) chains in the autosomal form of the disease cause the absence of the ␣3, ␣4, and ␣5(IV) chains from the GBM, but do not affect the presence of the ␣5 and ␣6(IV) chains in Bowman's capsule (33,34). Hence, these mutations prevent the assembly of the ␣3␣4␣5 protomer and network in the GBM, but do not affect the assembly of the (␣5) 2 ␣6 protomer of Bowman's capsule.
Sequestration of Goodpasture Autoepitopes-Previous studies have established that the epitopes of GP autoantibodies are sequestered within the NC1 hexamers and hence remain inaccessible for autoantibody binding unless the hexamer dissociates (4,11,12). However, the molecular basis of this cryptic property has remained unknown. Based on the hydrophobic character of the amino acids that compose the GP A epitope (Ala, Val, Ile), it has been proposed that the epitopes are buried at the interfaces between interacting NC1 domains within a hexamer (13). Determination of the x-ray structure of the homologous [(␣1) 2 ␣2] 2 (IV) NC1 hexamer (9, 10) has revealed that the E A and E B regions, encompassing the epitopes of GP A and GP B autoantibodies, are located adjacent to the triple-helical domain and to the interfaces between NC1 monomers within a protomer, but distant from the NC1 trimer-trimer interface. However, which NC1 domains sequester the GP epitopes not been determined. Here, we showed that within the ␣3␣4␣5 protomer, the E A epitope region interfaces with the ␣5 NC1 domain, and the E B epitope region interfaces with the ␣4 NC1 domain. Thus, intraprotomer interactions of ␣3NC1 with ␣5NC1 appear to reduce the accessibility of certain residues within the E A epitope that are required for the binding of GP A autoantibodies (and interactions of ␣3NC1 with ␣4NC1 would likewise sequester certain residues in the E B epitope). Conversely, binding of GP A autoantibodies to the E A epitope prevents the interaction of ␣3NC1 with ␣5NC1, as experimentally determined.
Knowledge of the amino acids that comprise the epitope of GP A antibodies, along with the structural alignment of the ␣3, ␣4, and ␣5 NC1 sequences within the [(␣1) 2 ␣2] 2 NC1 hexamer structure, allowed an inference about which residues contribute to the sequestration the GP A epitope. Within the E A region of the ␣3 NC1 domain, four amino acids (Ala-18, Ile-19, Val-27, and Pro-28) were found critical for binding of GP A autoantibodies (13), and outside this region, Gln-57 may also be required (35). A comparative analysis of the solvent-accessible surfaces of the NC1 monomers and hexamers revealed no changes in the accessibility of the Ala-18, Ile-19, and Pro-28 residues. How-ever, the accessibility of Val-27 and Gln-57 side chains was significantly lower in the NC1 hexamer than in the monomer, indicating that partial burial of these residues likely contributes to the cryptic property of the GP A epitope. Only two other residues within the E A region, Leu-29 and Tyr-30, become buried in the NC1 hexamer and may potentially contribute to the epitope sequestration. Whether these residues are important for GP A antibody binding could not be determined in previous studies using homologue scanning mutagenesis, because Leu-29 and Tyr-30 are conserved between ␣1 and ␣3 NC1 domains. Further progress in understanding the molecular basis for the cryptic nature of the GP epitopes will be achieved by structural studies of the ␣3 NC1 domain in monomer form, in the NC1 hexamer complex, and in complex with (auto)antibodies against the E A or E B regions.