The Goodpasture Autoantigen

The Goodpasture (GP) autoantigen has been identified as the α3(IV) collagen chain, one of six homologous chains designated α1–α6 that comprise type IV collagen (Hudson, B. G., Reeders, S. T., and Tryggvason, K. (1993) J. Biol. Chem. 268, 26033–26036). In this study, chimeric proteins were used to map the location of the major conformational, disulfide bond-dependent GP autoepitope(s) that has been previously localized to the noncollagenous (NC1) domain of α3(IV) chain. Fourteen α1/α3 NC1 chimeras were constructed by substituting one or more short sequences of α3(IV)NC1 at the corresponding positions in the non-immunoreactive α1(IV)NC1 domain and expressed in mammalian cells for proper folding. The interaction between the chimeras and eight GP sera was assessed by both direct and inhibition enzyme-linked immunosorbent assay. Two chimeras, C2 containing residues 17–31 of α3(IV)NC1 and C6 containing residues 127–141 of α3(IV)NC1, bound autoantibodies, as did combination chimeras containing these regions. The epitope(s) that encompasses these sequences is immunodominant, showing strong reactivity with all GP sera and accounting for 50–90% of the autoantibody reactivity toward α3(IV)NC1. The conformational nature of the epitope(s) in the C2 and C6 chimeras was established by reduction of the disulfide bonds and by PEPSCAN analysis of overlapping 12-mer peptides derived from α1- and α3(IV)NC1 sequences. The amino acid sequences 17–31 and 127–141 in α3(IV)NC1 have thus been shown to contain the critical residues of one or two disulfide bond-dependent conformational autoepitopes that bind GP autoantibodies.

Goodpasture (GP) 1 autoimmune disease is characterized by pulmonary hemorrhage and/or rapidly progressing glomerulo-nephritis (1). Tissue injury is mediated by anti-basement membrane antibodies that bind alveolar and glomerular basement membranes. The target autoantigen of basement membranes has been identified as the ␣3(IV) collagen chain, one of six homologous chains designated ␣1-␣6 that comprise type IV collagen (2). In the glomerular basement membrane, the ␣3(IV) chain exists in a supramolecular network along with the ␣4(IV) and ␣5(IV) chains (3). The ␣3(IV) chain is composed of a long collagenous domain of 1410 amino acids and a non-collagenous (NC1) domain of 232 residues at the carboxyl terminus (4).
The GP autoepitope(s) has been localized to the NC1 domain of the ␣3(IV) chain (5,6). Antibodies that bind to the NC1 domain of other ␣(IV) chains may be found in some Goodpasture patients (7,8), but they only account for about 10% of autoreactivity (9). The autoepitope(s) in the ␣3(IV)NC1 domain appears to be conformational, because reduction of disulfide bonds abolishes most of the binding (9 -11). The identification of the precise amino acid residues that constitute this epitope(s) is important for understanding the etiology and pathogenesis of the GP disease and for the development of diagnostic and therapeutic agents. Several groups have attempted to map the location of the autoepitope(s) by using short linear peptides (9,(11)(12)(13)(14) or by site-directed mutagenesis of the ␣3(IV)NC1 domain expressed in Escherichia coli (15). Although linear sequences have been identified that bind GP antibodies, these findings are at variance with each other. Moreover, prior studies have not addressed whether these linear sequences constitute the major conformational, disulfide bond-dependent epitope(s).
The aim of this study was to identify the ␣3(IV)NC1 amino acid sequences that form the thus far elusive conformational GP epitope(s). To circumvent the limitations of previous approaches, we pursued an epitope mapping strategy based on chimeric proteins. This approach has been specifically developed and successfully used to map conformational epitopes (16) or autoepitopes (17). We hypothesized that the ␣3(IV)NC1 regions most likely to form the autoepitope(s) are those most divergent from the other homologous ␣(IV) chains. A series of chimeric ␣1/␣3(IV)NC1 domains were constructed in which these candidate ␣3(IV)NC1 sequences replaced the corresponding sequences in the non-immunoreactive ␣1(IV)NC1. The chimeras were expressed in mammalian cells for correct protein folding and disulfide bond formation. We report that two specific sequences, ␣3(IV)NC1 residues 17-31 and 127-141, contain the critical residues of one or two disulfide bond-dependent conformational GP autoepitopes within the ␣3(IV)NC1 domain.

EXPERIMENTAL PROCEDURES
cDNA Manipulation and Chimera Construction-A suitable vector ( Fig. 1) for the expression of recombinant proteins was based on pRc/ AC7, a derivative of pRc/CMV (Invitrogen) that contained an expression cassette consisting of the BM-40 5Ј-untranslated region, BM-40 signal peptide, and an ␣3 type VI collagen insert (18). By using a two-step inverse PCR with the appropriate primers (Table I), the original insert was replaced by a FLAG recognition sequence (Eastman Kodak Co.), and additional restriction sites (NheI, ClaI, HpaI, and SacII) were introduced further downstream. The resulting vector (pRc-X) was used for the expression of the chimeras (Fig. 1, middle). After cleavage of the signal peptide, secreted proteins would contain at the amino terminus a 14-residue fusion sequence (APLADYKDDDDKLA) that included the FLAG peptide (underlined) used for affinity purification.
The cDNA for the human ␣1(IV)NC1 domain was amplified from a human kidney cDNA library (Marathon-Ready, CLONTECH) by PCR using Klentaq polymerase (CLONTECH) and subcloned into pCRII vector by using a TA cloning kit (Invitrogen). The inserts with the correct sequence were subcloned into pRc-X. The resulting pRc/f␣1 construct was subsequently used for the construction of ␣1/␣3 chimeras (Fig. 2). Unless otherwise indicated, Pfu polymerase (Stratagene) was used in the PCR reactions for its low error rates. Restriction enzymes and ligase were purchased from New England Biolabs. The correct sequence of each construct was verified by sequencing.
The principle of the inverse PCR approach that was used for chimeras C2, C3, C5, and C6 is depicted inside the dashed circle in Fig. 1 (top). The primers (Table I) were designed in a back-to-back orientation, each containing (in 3Ј to 5Ј order) residues complementary to the ␣1(IV)NC1 template, residues complementary to a part of the replacement ␣3(IV)NC1 sequence, and the recognition site of the inwardcutting BbsI restriction enzyme (GAAGAC(N)2/6). PCR yielded 6.3kilobase pair amplicons that comprised the whole vector and insert. Digestion with BbsI removed the recognition site and created complementary ends inside the inserted ␣3(IV) sequence, and then ligation produced a circular expression vector containing a chimeric ␣1/ ␣3(IV)NC1 insert with no extraneous sequence.
Construction of C1 and C4 chimeras was based on a regular PCR strategy using pRC/f␣1 as a template and introducing ␣3 sequences by primers at the 5Ј and 3Ј ends of the NC1 insert, respectively. The PCR products were digested with restriction enzymes (Table I) and subcloned into the pRc-X vector for expression. The construction of C7 chimera followed a similar scheme, requiring C1 as template. In order to construct chimera C8, two collagenous Gly-X-Y triplets of the ␣1(IV) chain had first to be added to the 5Ј end of the ␣1(IV)NC1 sequence. An ␣1(IV)NC1 ϩ Gly-X-Y insert was amplified from a pRc/f␣1 template, digested with NheI/PpuMI, and ligated into a C3⅐4 vector preparation cut with the same enzymes. Inverse PCR with this template generated chimera C8.
Six combination chimeras were also constructed as follows: C1⅐2, C1⅐4, C2⅐6, C3⅐5, C1⅐2⅐5, and C7⅐8. C1⅐2 chimera was generated using primers for the C1 construct and C2 as template, digested with NheI and ClaI, and then subcloned into the pRc-X vector. The remaining chimeras were generated by subcloning the chimeric insert region of one chimera into a vector preparation of another chimera digested with the same restriction enzymes. C1⅐4 required subcloning of a NheI/ PpuMI C1 insert fragment into the C4 vector; likewise, C2⅐6 required an ApaI C6 insert in the C2 vector; C3⅐5 required a PpuMI/SacII C3 insert in the C5 vector; C1⅐2⅐5 required an ApaI C5 insert in the C1⅐2 CAGTGAAGACTCACGGTGGACGGAATAGGCTTTC BbsI a A, forward primer; B, reverse primer. b FLAG sequence underlined.
vector; and C7⅐8 required a PpuMI/SacII C8 insert in the C7 vector. Protein Expression and Purification-Recombinant ␣1/␣3 chimeras were expressed in human embryonic kidney 293 cells (ATCC 1573-CRL) grown in Dulbecco's modified Eagle's medium/F-12 medium (Sigma) supplemented with 5% fetal bovine serum (Sigma) and 50 g/ml ascorbic acid phosphate magnesium salt (Wako). Five to ten g of plasmid DNA were transfected by the calcium phosphate co-precipitation method (19) into 70% confluent 293 cells. After 2 days, transfected cells were selected with 250 g/ml G418 (Life Technologies, Inc.). Resistant cells were screened for expression of recombinant protein by Western blot using an anti-FLAG monoclonal antibody (M2, Kodak) and expanded for quantitative expression. The medium was collected from subconfluent cultures every 48 h, and the recombinant proteins were purified by affinity chromatography on anti-FLAG M2 affinity columns (Kodak) according to the manufacturer's instructions. Protein solutions were concentrated by ultrafiltration (Amicon) and stored at Ϫ70°C. The concentration of recombinant protein solutions was measured spectrophotometrically at 280 nm. An average extinction coefficient A of 1.6 Х 1 mg/ml was calculated from the amino acid composition of the six human ␣(IV)NC1 domains (20).
Sera-The plasmapheresis fluid or sera from eight patients diagnosed with Goodpasture syndrome (GP1-8) were used. The titer of GP autoantibodies was measured by direct ELISA in plates (Nunc) coated with ␣3(IV)NC1 (100 ng/well). Relative to the GP1 serum previously described (24), GP1-4 had about the same titer, GP5-6 had a titer about 10-fold lower, and GP7-8 had about 80-fold lower.
Direct and Inhibition Immunoassays-MaxiSorp polystyrene microtiter plates (Nunc, Denmark) were coated overnight at room temperature with antigen (50 -200 ng/well, as shown) in 50 mM carbonate buffer, pH 9.6, and then blocked with casein or BSA. In some experiments, the antigen was reduced prior to coating by treatment with 10% ␤-mercaptoethanol for 5 min at 100°C. GP sera and normal human sera (negative controls) were diluted in the incubation buffer (2% casein or 2 mg/ml BSA and 0.05% Tween 20 in Tris-buffered saline). Alkaline phosphatase-conjugated goat anti-human IgG (1:2000) was used as secondary antibody. p-Nitrophenol phosphate (1 mg/ml in 1 M diethanolamine buffer, pH 9.8, containing 0.5 mM zinc chloride) was used as substrate, and the development of color was monitored at 410 nm in a Dynatech MR4000 plate reader. For inhibition ELISA, the GP sera were incubated overnight at room temperature with various amounts of recombinant ␣(IV)NC1 domains or chimeras prior to addition to plates coated with ␣3(IV)NC1. The results shown are the averages of duplicate determinations.
PEPSCAN Analysis-Mapping of linear epitopes was performed using the "PEPSCAN" method (26). A complete set of solid phase overlapping 12-mer peptides was synthesized onto polyethylene pins following the published sequences of NC1 domains of ␣1(IV) (GenBank TM accession number P02462) and ␣3(IV) collagen (GenBank TM accession number X80031). The immunoscreening of these peptides was performed by ELISA. The pins were incubated for 1 h with GP serum (diluted 1:50) and then washed three times. The bound antibody was detected by the reaction with peroxidase-labeled secondary antibody for 30 min, followed by color development with 2,2Ј-azinobis-3-ethylbenzthiazolinesulfonic acid for another 30 min.

Design and Expression of ␣1/␣3(IV)NC1
Chimeras-In this study, ␣1/␣3 chimeras were used to identify the conformational epitope(s) of the GP autoantigen. This strategy relied on the high sequence homology between the NC1 domains of ␣1(IV) and ␣3(IV) collagen (71% sequence identity and six conserved disulfide bonds), which very likely adopt similar tertiary structures (27). In the chimeras, ␣1(IV)NC1 acted as an inert "carrier" and provided a three-dimensional scaffold for the substituted ␣3(IV) sequences.
Since GP sera react preferentially with ␣3(IV)NC1, but not the other ␣(IV)NC1 domains, the autoepitope(s) must contain amino acids specific to ␣3(IV)NC1. Our recent comparative analysis of the sequences of ␣(IV)NC1 domains has now permitted the identification of six putative locations of the epitope(s) as short regions (less than 15 residues) in ␣3(IV)NC1 that are most divergent from other ␣(IV)NC1 domains and that are also predicted to be accessible to solvent (27). Accordingly, six chimeric NC1 domains (C1-C6) were constructed in which these ␣3(IV)NC1 sequences replaced the corresponding amino acids within the ␣1(IV)NC1 domain (Fig. 2). Five combination chimeras (C1⅐2, C1⅐4, C2⅐6, C3⅐5, C1⅐2⅐5) were also constructed to allow identification of non-contiguous GP epitopes.
To analyze two previously proposed GP epitopes (11,15,28) using this approach, three additional chimeras (C7, C8, and C7⅐8) were constructed. The 26 amino-terminal amino acids of C7 chimera, which included four collagenous Gly-X-Y triplets, were from the ␣3(IV) sequence. C8 contained the 36 carboxylterminal residues of ␣3(IV)NC1 and, in addition, had two ␣1(IV) Gly-X-Y triplets at the amino terminus. The additional Gly-X-Y sequences, also present in the native collagenase-digested ␣1and ␣3(IV)NC1 domains, were incorporated in the C7 and C8 chimeras to emulate the proteins previously used to map the autoepitope (15).
The recombinant chimeric proteins were expressed in the human embryonic kidney 293 cells and isolated from the culture medium as monomers with an apparent molecular mass of about 25-30 kDa by SDS-polyacrylamide gel electrophoresis (Fig. 3a). Unlike expression in E. coli (22), expression in the human kidney 293 cells yields properly folded recombinant NC1 domains that are indistinguishable by FT-IR or immunoassays from those prepared from native sources. 2 This cell line has been successfully used to express other proteins with native folding, including basement membrane proteins nidogen (29) and laminin (30).
Immunoreactivity of ␣1/␣3 Chimeras with GP Sera-The reactivity of the chimeric constructs with GP sera was analyzed by Western blotting as well as by direct and inhibition ELISA. The pattern of autoantibody binding obtained in Western blots with three sera show remarkable similarities (Fig. 3, b-d).
Only two chimeras, C2 and C6 (containing residues 17-31 and 127-141 of ␣3(IV)NC1, respectively), reacted strongly and consistently with GP antibodies, as did combination chimeras containing one or both these regions (C1⅐2, C2⅐6, C1⅐2⅐5). Some sera showed weak reactivity with other chimeras, but this appears to be due to the cross-reactivity with the ␣1 backbone, because it was accompanied by comparable binding to ␣1(IV)NC1. Remarkably, neither C7 nor C8 chimeras bound autoantibodies.
To confirm these findings, the binding of eight GP sera to immobilized chimeras was assessed in direct ELISA (Fig. 4). Sera were diluted proportionally to their titers to allow visualization of the specificity of the low titer sera side by side with the high titer sera. In general, the pattern of reactivity observed in the Western blots was also apparent in the ELISA. All sera reacted strongly with C2 chimera (which averaged 71% of the maximal reactivity, obtained with ␣3(IV)NC1), C1⅐2 (47%), C2⅐6 (70%), and C1⅐2⅐5 (64%). There was more variation in the reactivity toward C6 chimera (31% of the reactivity of ␣3(IV)NC1), which bound significantly only five out of eight sera. All but one low titer serum (GP7) bound more to C2 than to C6 chimera. Sera that showed cross-reactivity with ␣1(IV)NC1 bound all chimeras, producing a higher background.
The relative reactivity of any given serum toward recombinant proteins was not influenced by the dilution of the serum. This was apparent in the titration curves shown in Fig. 5, which yielded parallel lines for various immobilized proteins. Similar results were obtained with the other sera. All sera had the highest reactivity toward ␣3(IV)NC1, which was closely followed by C2⅐6 and C2 chimeras (less than a 2-fold difference in titers). The C6 chimera titers of the sera were more variable, between 2-and 10-fold lower than the ␣3(IV)NC1 titers, but always higher than those of ␣1(IV)NC1 and ␣2(IV)NC1 controls.
Immunodominance of Antibodies Binding to C2 and C6 Chimeras-It is well established that adsorption of proteins to plastic may cause denaturation, so that the antibody binding measured in direct ELISA may actually be to the denatured antigen. To rule out such artifacts, the interaction between the GP antibodies and antigen was studied in solution by inhibition ELISA in the presence of soluble chimeras and control ␣(IV)NC1 domains. The inhibition curves were determined for three GP sera and were found to be similar. Typical data for one serum are shown in Fig. 6 (top panel). The inhibitory capacity of the chimeras and the control proteins followed the same order as found in direct ELISA, ␣3(IV)NC1 Ͼ C2⅐6 Ͼ C2 Ͼ C6 Ͼ ␣1(IV)NC1, consistent with the results obtained with the latter technique. The effect of the chimeras was saturable, leveling off at the highest concentration used, where it produced 42-67% inhibition, a significant proportion of the autoantibody reactivity.
The ␣3(IV)NC1 domain could completely inhibit autoantibody binding and had an I 50 (the concentration of competitor at which half-maximal inhibition is achieved) of 0.27 Ϯ 0.03 g/ml (about 11 nM), in good agreement with the previously reported values of 0.5 (31) and 0.8 g/ml (9). At high concentrations, ␣1(IV)NC1 but not ␣2(IV)NC1 inhibited autoantibody binding to ␣3(IV)NC1 by about 24%. This effect can be attributed to cross-reactivity, since ␣3(IV)NC1 is more similar to the ␣1and ␣5(IV)NC1 domains than to ␣2-, ␣4-, or ␣6(IV)NC1 domains (27). An I 50 value could not be reliably calculated for chimeras because the inhibition curves they produced could not be fitted adequately to a simple inhibition model. Visual examination of these curves revealed a bi-phasic behavior. The steep inhibition below 2 g/ml (Fig. 6, top panel) is probably due to the specific ␣3(IV) sequence in the chimeras, whereas the shallower portion of the curves at higher chimera concentrations, which parallels the ␣1(IV)NC1 inhibition curve, is likely caused by cross-reactivity with the ␣1(IV)NC1 scaffolding.
To quantitate the binding specificity of eight GP sera, an inhibition ELISA was performed at a fixed concentration of soluble antigen, 10 g/ml (Fig. 6, bottom panel). This concentration was chosen to minimize cross-reactivity with ␣1(IV), while giving almost complete inhibition with ␣3(IV)NC1. Inhibition with C2⅐6 was 65 Ϯ 13%, compared with 85 Ϯ 7% for control ␣3(IV)NC1, demonstrating that this chimera contains the immunodominant autoepitope(s) of ␣3(IV)NC1. C2⅐6 chimera had a stronger effect than either C2 (46 Ϯ 8%) or C6 chimeras (23 Ϯ 18%). This indicates that the ␣3(IV)NC1 residues 17-31 (hereafter referred to as E A ) and 127-141 (hereafter referred to as E B ) form either two separate epitopes or a single, more complete one, but it appears to rule out significant cross-reactivity between the two homologous sequences.
The data were further analyzed to estimate the fraction of autoreactivity that could be attributed specifically to the ␣3(IV)NC1 sequences in the chimeras. For each serum, the inhibition produced by the ␣1(IV)NC1 domain (which averaged 7 Ϯ 4%) was subtracted from the total inhibition given by the chimeras to correct for the cross-reactivity due to the common scaffold, and then the results were normalized to the effect produced by ␣3(IV)NC1 (Table II). The effect of E A (present in C2 chimera) was strong and consistent (on average 47%, ranging between 27 and 64%) and predominated in seven out of eight sera. In contrast, E B (present in C6 chimera) produced variable inhibition with different sera (on average 18%, ranging between 3 and 56%) and was predominant only in GP7. Together, as in C2⅐6 chimera, these sequences accounted for most inhibition of GP sera (on average 68%, ranging between 52 and 88%). Only a small fraction of GP reactivity toward ␣3(IV)NC1 (on average 23%, ranging between 6 and 38%) could not be accounted for by E A , E B , or by cross-reactivity with ␣1(IV)NC1.
Conformational Nature of the Epitope(s)-It has been previously shown that the reduction of disulfide bonds in ␣3(IV)NC1 impairs its ability to react with GP antibodies, indicative of a conformational epitope (9 -11). Quantitation of this effect with the eight sera used in the present work showed that only 6 Ϯ 5% of the original immunoreactivity remains after reduction of ␣3(IV)NC1. To evaluate whether E A and E B form a linear or a conformational epitope, the GP reactivity of the ␣1/␣3 chimeras was measured before and after reduction. As in ␣3(IV)NC1, the reduction of disulfide bonds also abolished binding of autoantibodies to the C2, C6, and C2⅐6 chimeras and even to ␣1(IV)NC1 (Fig. 7). Less than 10% of the original immunoreactivity remained in the reduced proteins, although they had the same or higher reactivity with monoclonal antibodies that do not require a conformational epitope, such as anti-FLAG (data not shown). Overall, these results demonstrate that only a small proportion of GP antibodies can recognize linear epitopes and that E A and E B belong to one or two conformational GP epitopes that are disulfide bond-dependent.
Comparison of Chimera-based Epitope Mapping with Previous Approaches-To compare the chimera-based epitope mapping strategy with approaches using linear peptides (9,11,13,14) and to evaluate whether the latter identify linear or conformational GP epitopes, a peptide scanning analysis was performed. A valid comparison between the chimera-based and peptide-based strategies required using the same GP sera. Two complete sets of overlapping 12-mers based on the ␣1(IV)and ␣3(IV)NC1 sequences (Fig. 8) were therefore synthesized and analyzed by the PEPSCAN procedure (26). A previous report using 20-mer peptides to map the GP epitope has indicated nonspecific binding of GP sera to homologous ␣1and ␣3(IV)NC1 peptides (9).
The PEPSCAN results demonstrated lack of strong specific binding and a high background, presumably due to nonspecific binding. Both ␣1and ␣3(IV)NC1 sequences produced a num- domains. Sera, diluted as described in Fig. 4, were incubated overnight with 10 g/ml antigen. ELISA was performed in plates coated with 50 ng/well ␣3(IV)NC1. Individual data for the eight sera are represented by symbols, and their average is shown as a horizontal line.
ber of peaks higher than two standard deviations above the median. However, the most reactive peptides (above three standard deviations) varied among the three GP sera tested. The most significant PEPSCAN peak was produced by peptides overlapping residues 94 -110 of ␣3(IV)NC1 with the GP2 serum. Much weaker reactivity was recorded in this region with the other two sera. This region corresponds to the C5 chimera that did not interact with GP sera in direct ELISA (Fig. 4), perhaps due the conformations of the 12-mer peptides on the pin being different from those adopted by the same amino acids in the NC1 domain. Some isolated ␣3-derived peptides that overlapped the E A region produced peaks in PEPSCAN. However, the interactions were not strong enough to allow unambiguous identification of these residues as part of a GP autoepitope.
Two epitopes previously found by mutagenesis of ␣3(IV)NC1 expressed in E. coli (15) were not observed in C7 and C8 chimeras, made in eukaryotic cells in the present work. Recom-binant ␣3(IV)NC1 expressed in E. coli was found about four times less reactive than the native human protein (Fig. 9), in agreement with earlier reports (22). In contrast, the recombinant ␣3(IV)NC1 used in this work, expressed in 293 kidney cells, was found as reactive as native human ␣3(IV)NC1 (Fig.  9). This suggests a folding difference whereby the full complement of conformational epitopes is not assembled in the E. coli-made protein. It is also possible that mutagenesis to alanine of residues from E. coli ␣3(IV)NC1 may have caused the reported loss of GP immunoreactivity by affecting the overall structure of the protein and not the epitope itself. DISCUSSION In the present study, a new strategy based on chimeric proteins was employed to map regions within ␣3(IV)NC1 that constitute the conformational epitope(s) for GP autoantibodies. This novel approach has two methodological improvements over previous work. Unlike in peptide-based epitope mapping, short ␣3(IV)NC1 candidate regions (Ͻ15 residues) were grafted onto an inert ␣1(IV)NC1 framework and expressed in mammalian cells to ensure native folding. The resulting chimeras were assayed for "gain-of-function," i.e. capacity to bind autoantibodies, in contrast with previous site-directed mutagenesis studies (15) that relied on a "loss-of-function" of the protein expressed in E. coli. The results from 14 different chimeras revealed two previously unidentified regions, designated E A and E B (residues 17-31 and 127-141 of ␣3(IV)NC1, respectively), that strongly bound autoantibodies from eight GP patients. Together, E A and E B accounted for 50 -90% (on average 68%) of autoreactivity to ␣3(IV)NC1.
c The specificity not assigned represented the amount of binding that could be inhibited by ␣3(IV)NC1 but not by C2⅐6 chimera. gence between ␣3(IV)NC1 and the other NC1 domains confer antibody binding to the former. The four regions that were found non-reactive (i.e. those substituted in C1, C3, C4, and C5 chimeras) further distinguish E A and E B as the primary regions for the GP epitope. It is significant that the E A and E B regions are homologous (47% sequence identity) and are located at corresponding positions in the two homologous NC1 subdomains (27), but they are noncontiguous. E A and E B could represent two separate and distinct epitopes or a single epitope E AB , in which E A and E B are held in close proximity to each other by the disulfide bonds. In either case, the complete epitope(s) probably includes additional residues from other regions, less critical for binding. So far, the x-ray crystallographic structures of other protein-antibody complexes have revealed noncontiguous epitopes of 15-22 amino acids that belong to several surface loops (32,33).
Our results demonstrate that regions E A and E B reproduce very well the authentic GP epitopes in the ␣3(IV)NC1 domain. Most remarkably, E A and E B form conformational epitopes that require intact disulfide bonds to bind GP antibodies, as demonstrated by loss of GP immunoreactivity of the C2, C6, and C2⅐6 chimeras upon reduction (Fig. 7). The majority of GP autoantibodies appears to recognize conformational epitopes in ␣3(IV)NC1 (9 -11), but epitope mapping studies have not addressed until now the nature of the epitopes found (see below). Further demonstrating the good mimicry of the original epitope(s), the chimeras produced significant inhibition of GP sera at concentrations in the range of 10 Ϫ8 M, comparable with ␣3(IV)NC1 domain. In contrast, linear ␣3(IV)NC1 peptides produced a comparable effect in inhibition ELISA only at concentrations 100 -1000-fold higher (11,14).
The E A and E B regions have not been previously identified by peptide-based epitope mapping (9,(11)(12)(13)(14). As shown here, this was due to the inability of peptide scanning procedures to reliably identify the conformational GP epitope(s). An intrinsic tendency of peptide-based methods to identify sequential epitopes has already been noted (34). Thus, the ␣1(IV)NC1 framework of the chimeras is instrumental for adoption of the native conformation by E A and E B , and in addition, it may contribute auxiliary residues for binding. It is likely that the previous reports have largely identified linear GP epitopes, which constitute a minority (about 5% of the reactivity against ␣3(IV)NC1). Furthermore, various linear sequences were found reactive in different studies, suggesting heterogeneity of the linear epitopes. In contrast, the chimera-based approach has successfully identified the critical regions of one or two immunodominant, conformational GP epitope(s) that were consist-ently recognized by all autoimmune sera used in this work. 3 Region E A clearly represents an immunodominant epitope. It was recognized strongly and consistently by all sera analyzed, whereas E B reacted significantly (Ͼ10%) with only half of the sera. This may be due to the higher divergence of E A (eight distinct amino acids) compared with E B (five distinct amino acids). The existence of an immunodominant epitope explains the considerable cross-inhibition between GP sera from different patients or between GP sera and certain monoclonal antibodies (13,31,35). E A and E B may well be the counterpart of the shared structural determinants on the GP antibodies, found by using an anti-idiotype antibody against anti-␣3(IV) IgG (36).
In summary, two specific homologous sequences in ␣3(IV)NC1 have been identified for the first time to be part of one or two disulfide bond-dependent, conformational and immunodominant GP autoepitopes. This finding provides new knowledge to investigate further the pathogenesis of GP disease. It has recently been shown that ␣3(IV)NC1 but not ␣1(IV)NC1 can induce experimental GP disease in mice (21). A very important question, relevant for the identification of the nephritogenic epitope(s) in ␣3(IV)NC1, is whether any of the ␣1/␣3 chimeras can induce experimental GP syndrome. In myasthenia gravis, another autoimmune disease, the immunodominant epitope on the acetylcholine receptor (known as "MIR" or main immunogenic region) was also pathogenic (37). By providing a highly specific target, the new identification of an immunodominant GP epitope should be useful for the development of more specific therapeutic approaches, such as use of vaccines to induce tolerance or the manipulation of the idiotype network.