Molecular Properties of the Goodpasture Epitope*

Goodpasture disease fulfils all criteria for a classical autoimmune disease, where autoantibodies targeted against the non-collagenous domain of the α3-chain of collagen IV initiates an inflammatory destruction of the basement membrane in kidney glomeruli and lung alveoli. This leads to a rapidly progressive glomerulonephritis and severe pulmonary hemorrhage. Previous studies have indicated a limited epitope for the toxic antibodies in the N-terminal part of the non-collagenous domain. The epitope has been partially characterized by recreating the epitope in the non-reactive α1-chain by exchanging nine residues to the corresponding ones of α3. In this study we have investigated to what extent each of these amino acids contribute to the antibody binding in different patient sera. The results show that seven of the nine substitutions are enough to get an epitope that is recognized equally well as the native α3-chain by all sera from 20 clinically verified Goodpasture patients. Furthermore, the patient sera reactivity against the different recombinant chains used in the study are very similar, with some minor exceptions, strongly supporting a highly defined and restricted epitope. We are convinced that the restriction of the epitope is of significant importance for the understanding of the etiology of the disease. Thereby also making every step on the way to characterization of the epitope a crucial step on the way to specific therapy for the disease.

Goodpasture disease is known and characterized as a classic autoimmune disease. The disease is B cell and antibody mediated, with autoantibodies directed against proteins in the glomerular basement membrane and lung alveoli. When bound to self-structures in the kidney and lung, the antibodies initiate an inflammatory destruction of tissue by recruitment of complement leading to rapidly progressive glomerulonephritis often accompanied with severe and life threatening lung hemorrhage. The major self-epitope is located on the ␣3-NC1 domain of collagen IV. Collagen IV ␣3-chain has a limited distribution in the body and is only found in a few specialized basement membranes including the glomerular and alveolar basement membranes, thus explaining the specific organ involvement in Goodpasture disease. Goodpasture disease is indeed an anti-body mediated disease as proven by the transfer of disease to monkeys by injection of kidney bound antibodies from Goodpasture patients (1) and the therapeutic effect of treating patients with plasma exchange and immunosuppressive drugs to reduce the amount of circulating antibodies (2). The Goodpasture epitope is a conformational epitope, which is indicated by the loss of reactivity to autoantibodies when the tertiary protein structure is disrupted by reduction of disulfide bonds (3). The epitope is also known as a cryptotope, i.e. the epitope is hidden in the native protein structure and is fully exposed first when the protein is partially denatured (4).
Trying to map an epitope for a specific autoimmune disorder is very difficult, since in most cases autoantibodies against a variety of epitopes on the target structure are formed. For Goodpasture syndrome a limited epitope distribution was indicated when binding of autoantibodies to collagen IV was successfully blocked by one single monoclonal antibody (5,6). In our first attempt to map the Goodpasture epitope we used linear synthetic peptides of the ␣3(IV) NC1 domain to block the binding of autoantibodies to collagen IV (7). With this method we were unable to map any epitope on the ␣3(IV) NC1 domain although for one patient an epitope on the ␣1(IV) NC1 domain was found. In a study by Kalluri et al. (8), using linear peptides, they suggested the C-terminal part of the ␣3 (IV) NC1 domain to comprise the Goodpasture epitope. However this study as well as ours suffered from the disadvantage of using linear peptides to characterize a conformational epitope and the results have not been confirmed.
To avoid the problem with linear peptides and mal-folded recombinants expressed in bacterial systems, new mapping strategies has been initiated by several groups, where recombinant collagen IV is expressed in eukaryotic cell lines. By substitution of amino acid residues in the ␣3(IV) NC1 against the corresponding ones from the homologous but non-reactive ␣1(IV) chain, and expressing the constructs in an eukaryotic expression system correctly folded proteins with intact conformational epitopes are produced. All these studies have emphasized the N-terminal part of the ␣3 (IV) NC1 as the principal epitope region (9 -11). Furthermore, we found that only autoantibodies against the N-terminal third of the ␣3(IV) NC1 domain are pathologically significant (12). Following studies have revealed a small region within the N-terminal part of the ␣3 NC1 as the major target for the circulating antibodies. This epitope has then been recreated by substitution of a few a.a. residues in the non-reactive ␣1-chain to the corresponding ones from ␣3 (13)(14)(15). In our hands, all 20 patients in the cohort recognized this epitope (15).
The principal goal for this study was to investigate if a reactive epitope could be created in the ␣1(IV) NC1 with fewer than the nine substitutions, previously reported, and to what extent each of these amino acids contribute to the antibody binding in different patient sera.

EXPERIMENTAL PROCEDURES
Patients and Sera-The sera of 20 Goodpasture patients with biopsyproven anti-glomerular basement membrane nephritis were obtained from the serum bank at the Department of Nephrology, Lund University Hospital. All patients showed crescentic glomerulonephritis with linear deposits of IgG at direct immunofluorescence. Seven of the patients had in addition overt lung hemorrhage. The mean age was 49 years ranging from 18 to 78 years, 10 males and 10 females. The mean serum creatinin at time of diagnosis was 920. Twelve of the patients were maintained on dialysis, four had died and four survived with native functional kidneys after 6 months of follow-up. Sera from five healthy blood donors were used as controls.
Antibodies-The alkaline phosphatase-conjugated antibodies, goat anti-human IgG (Fc part) and goat anti-rabbit IgG, were purchased from Sigma and the rabbit anti-mouse IgG from Dako, Glostrup, Denmark.
The polyclonal rabbit anti-collagen X antibodies were manufactured as sera at the laboratory by immunizing a rabbit with keyhole limpet hemocyanin-conjugated peptide (GYP GAK GER GSP GSD GKP GYP GKP GLD GC-(KLH)). It was diluted 1/200. The monoclonal antibody against the 6 ϫ His was purchased from Serotec (MCA1396, Serotec, Kidlington, Great Britain) and diluted 1/2000.
Expression Vector-In this study the expression vector pCEP4-BM40-HisEK was a kind gift from Dr. A. Aspberg, Cell and Molecular Biology, Lund University. 1 The vector is based on the pCEP4 vector (Invitrogen, Leek, The Netherlands) and adds a BM40 signal peptide followed by a hexahistidine tag and an enterokinase D cleavage site N-terminal of the recombinant protein.
Construction of DNA for Expression-The basic DNA construct used in this study is a cDNA coding for the human ␣1(IV) NC1. This was fused to cDNA coding for a stretch of around 100 base pairs of the triple helical part of collagen X, expressing the epitope for the anti-type X collagen antibody described above. Furthermore, in this construct called S1, five point mutations have been introduced causing substitutions of amino acids in five positions (P5-P9) in the ␣1(IV) NC1 to the corresponding amino acids of ␣3(IV) NC1 (Fig. 1b). Substitutions of these five amino acids have in a previous study been shown to be of vital importance for epitope recognition (15), although the substitution of these five amino acids was not enough to make the non-reactive ␣1 reactive. In the same study an immunoreactive epitope was created by the introduction of four additional amino acid substitutions, in positions P1-P4. Since the aim of this study is to investigate if a reactive epitope can be created with fewer than these four additional substitutions, P1-P4, we made recombinants with all possible combinations of the four additional substitutions, in all 14 constructs. For the introduction of mutations in the 14 different constructs, the overlap extension polymerase chain reaction technique (16) was applied using the primers 1 and 2 and 3-13 in Table I and the S1 construct as template (Fig. 1b). The 14 constructs were besides the five common substitutions substituted in one or more of the positions P1-P4. The constructs were named according to which positions that were substituted, e.g. a construct substituted in positions 1 and 3 is called R13 for "recombinant substituted in positions 1 and 3." The protein-coding cDNAs were then inserted in the pCEP4-BM40-HisEK vector (Fig. 1a).
After electroporation the cells were seeded on new plates, and after reaching confluence, about 48 h after transfection, selection with hygromycin was initiated. Transfected, hygromycin-resistant cells were cultured until a desired number of plates had reach confluence then collection of supernatants were initiated. During harvesting, the transfected cells were kept in fetal calf serum-free Dulbecco's modified Eagle's medium supplemented with 50 mg/liter ascorbate (17). The expected size of the recombinant protein is approximately 31 kDa.
Purification of Recombinant Protein-500 ml of serum-free conditioned medium were precipitated by adding ammonium sulfate to 50% saturation and then solved in 20 mM NaP i buffer, pH 7.8, with 250 mM NaCl and applied to 3 ml of ProBind d resin (Invitrogen) equilibrated in the same buffer. The column was then washed with a 20 mM NaP i buffer, pH 6.0, with 500 mM NaCl and eluted in a 20 mM NaP i buffer, pH 6.0, 500 mM NaCl, 500 mM imidazol. The eluted samples were concentrated and dialyzed against a phosphate-buffered saline buffer. The purified recombinant proteins were analyzed by silver stained SDSpolyacrylamide gel electrophoresis gels and by immunoblotting using antibodies against the type X collagen domain and antibodies against the His tag. Protein concentrations were determined using the BCA (Pierce) method and by measuring the absorbance at 280 nm.
Enzyme-linked Immunosorbent Assay (ELISA)2-The microtiter plates (Nunc Immunoplate, Roskilde, Denmark) were coated overnight at 4°C with 0.025 g/well of purified recombinant protein in coating buffer (50 mM Na 2 CO 3 , 0.05% NaN 3 , pH 9.6). The plates were washed three times with 0.15 M NaCl, 0.05% (v/v) Tween 20 and then incubated for 1 h at room temperature with 100 l/well of sera diluted 1/100 in phosphate-buffered saline-bovine serum albumin (1.5 mM KH 2 PO 4 , 8 mM Na 2 HPO 4 , 0.12 M NaCl, 2.5 mM KCl, 0.05% (w/v) NaN 3 containing 0.2% (w/v) bovine serum albumin, pH 7.3). After three new washes the plates were incubated with alkaline phosphatase-conjugated goat antihuman IgG for an additional hour. The amount of bound antibodies was detected by the use of p-nitrophenyl phosphate (Sigma) (1 mg/ml) in substrate buffer (1 M diethanolamine, 0.5 mM MgCl 2 , pH 9.8), as substrate. Color development was measured spectrophotometrically at 405 nm after 1-h incubation at room temperature. All assays were run in duplicate.
Inhibition ELISA-The inhibition ELISA was performed in the same way as described above, with the exception of the preceding overnight preincubation at 4°C of patient sera with recombinant protein in concentrations from 10 to 0.0001 g/ml. The plates were coated with recombinant ␣3-NC1 (0.025 g/well).
To find the optimal concentration of patient sera for inhibition, two dilution series were made for each of the six patient sera used in the inhibition ELISA. One of the dilution series was preincubated with recombinant ␣3(IV) NC1 (2 ng/well), and the analogous series was preincubated with an equal volume of phosphate-buffered saline-bovine serum albumin buffer. In this way two "dilution curves" were achieved, where the absorbance for the curve produced from the dilution series afflicted by inhibition from the ␣3(IV) NC1 declined earlier and reached background absorbance earlier than the non-inhibited dilution series. The serum dilution that gave the largest difference, in percentages, between the inhibited and non-inhibited dilution series was used in the inhibition ELISA assays for the different recombinants. Antibodies bound to the coat were detected with alkaline phosphatase-conjugated secondary antibodies as described above. FIG. 1. a, a schematic drawing of the cDNAs inserted into the expression vector (pCEP4-BM40-HisEK). This construct result in an expressed protein containing a 6 ϫ His tag followed by a 30-amino acid stretch of type X collagen and the NC1 domain from type IV collagen. b, in the basic construct, S1, amino acids in positions P5-P9 are substituted from the ␣1 ones (upper row) to the corresponding ones of ␣3(IV) NC1 (lower row). The positions for substitutions in the ␣1 protein sequence are indicated by arrows and numbers. For example, the R12 construct is substituted in positions P1 and P2 plus P5-P9.

SDS-Polyacrylamide Gel Electrophoresis and Immunoblotting-A
10% NuPAGE gel from Novex (San Diego, CA) was run in a MOPS buffer system according to the supplier's recommendations. The gels were either silver-stained or transferred to a polyvinylidene difluoride membrane using a semidry blotting procedure (18). The antibodies were diluted as described above, and the membranes were treated as described elsewhere (6).

RESULTS
In a previous study we were able to recreate an immunoreactive epitope in the non-reactive ␣1(IV) NC1 domain by substitution of single amino acids to the corresponding ones of ␣3 (15). In that study two principle constructs were made, one with five amino acid substitutions in positions known to be critical for epitope recognition (P5-P9), called the S1 construct, and one with four additional substitutions in non-conserved positions in the same region (P1-P4), called S2. It was shown that the five substitution recombinants only reacted weak with the antibodies, whereas the nine-substitution recombinant showed immunoreactivity to all tested sera. In this present study, we have investigated what impact each of these amino acids has on the affinity of autoantibodies from different patients. Thereby further characterizing the immune response in Goodpasture disease to this major epitope. To achieve this, all possible combinations of the four additional substitutions were made in a total of 14 new recombinants. The S1 construct was used as template, and then subsequent substitutions were introduced with site-directed mutagenesis using the primers listed in Table I.
Initial Screening of Reactivity-Initial tests of the 14 recombinants' immunoreactivity against all 20 patient sera were performed using ELISA and inhibition ELISA. Of the 14 tested recombinants only five were reactive with the patient sera, one of the seven substitutions (R12) and four of the eight substitutions (R123, R124, R134, and R234). The other recombinants did not show reactivity above the reactivity against the recombinant ␣1(IV) NC1 (data not shown) and therefore not further evaluated. The reactive recombinants R12, R123, R134, and R234 were carefully analyzed again to establish their reactivity compared with the nine-substitution recombinant (S2), recombinant ␣3, and recombinant ␣1. In Fig. 2, the immunoblotting results are shown.
ELISA-The recombinants were first analyzed using direct ELISA (Fig. 3). With this assay, we found no significant difference in reactivity for the R12 recombinant compared with the S2 protein. In line with these results, none of the eight substitution recombinants had a significantly lower reactivity than the ␣3 recombinant. Although it is noticeable that recombinant R134 that lack substitution 2 has a significantly lower reactivity than R12, probably indicating a greater importance for substitution 2. In fact, two of the samples did not react with this R134 recombinant, and another six of them showed low reactivity.
Inhibition ELISA-In contrast to the direct ELISA, the inhibition ELISA revealed differences in affinity to the different recombinants. The highest inhibitory effect was found using recombinant ␣3, S2, or R12, while all the eight substitution recombinants showed a significantly lower inhibitory capacity than the seven-substitution R12 recombinant. The inhibition curves from one of the samples are shown in Fig. 4a, and the mean normalized inhibition values for all samples are shown in Fig. 4b.
In conclusion, the ELISA studies have shown that the sevensubstitution recombinant R12 is as reactive as the native ␣3 and the nine substitution recombinant S2 described previously (15). DISCUSSION Circulating autoantibodies against different parts of the ␣3(IV) chain, as well as antibodies against other ␣(IV) chains, are detected in serum from patients with Goodpasture disease (6,19). However, the major epitope region is found to be the N-terminal part of the ␣3(IV) NC1 (9,11), and only antibodies against this part of the molecule correlate with disease progression (12,15). In this study we have further narrowed down and characterized the molecular properties of this epitope recognized by the pathogenic autoantibodies. A recombinant protein comprised of the ␣1(IV) NC1 domain with seven amino acids substituted to the corresponding ones from the ␣3 chain was constructed. This recombinant protein, R12, was recognized by the autoantibodies from all patients with Goodpasture  1-8, respectively, and immunoblotted against the anti-type X collagen antibody in A and against one representative Goodpasture serum in B. The arrow indicates the 31-kDa recombinant proteins that appear as double bands, probably due to proteolytic degradation. The control antibody recognizes all recombinants, while the patient serum bound the recombinant ␣3, S2, R12, R123, and R124. A faint staining is seen against the R134, and no binding was detected against the recombinant ␣1 or S1. disease, in both direct ELISA and in inhibition ELISA, to the same degree as recombinant ␣3. These results support the previous findings where we and others have localized the major epitope to the same region (13)(14)(15).
The very limited region recognized by all Goodpasture sera indicates a similar immunization and maturation process in all patients. Interestingly, an overlapping T-cell epitope is found by Phelps and co-workers (20,21). However, if this process is initiated by a foreign immunogen, i.e. molecular mimicry, or if it is a self-immunization with degraded type IV collagen, is yet unknown.
As shown in Fig. 5a, the epitope is localized to a small loop in the secondary sequence. The epitope seems to be dependent on correct folding of this loop, indicated both by the fact that disruption of the disulfide bridges (4, 7) as well as changes of the charge of certain residues disturb the immunoreactivity. This is shown by the substitutions in positions 3 and 4, of which one, but not both, of the resulting amino acids must have a negative charge for the epitope to be fully recognized. In R124 both residues are negatively charged, and in R123 both residues are uncharged, and in both, the changes result in a loss of affinity. The preservation of the positively charged lysine in position P2, seen in the S1 and R134 recombinants (Fig. 5b), dramatically reduce affinity to the recombinants. Surprisingly, the R12, as well as the S2, R123, and R124, actually reacted stronger than the recombinant ␣3(IV) in the direct ELISA. This could possibly be explained by a less rigid structure in the ␣1(IV) background that results in a more accessible epitope. This theory is supported by the fact that this difference in reactivity did not appear when the recombinants were analyzed using inhibition ELISA. An alternative explanation for the higher reactivity could be that the produced recombinant, e.g. R12, displays an epitope more similar to a hypothetical mimicry structure than the native ␣3(IV) NC1 does.
Although all samples recognize one very limited area on the ␣3(IV) NC1, there are some differences in recognition pattern between the different samples. As discussed above the effect of charge changes within the loop have different effect on antibodies from different sera (especially the R134). Furthermore the relative amount of antibodies against the epitope defined by the R12 constructs varies from serum to serum, ranging from 65 to 95%.
We believe that this study of the Goodpasture epitope adds new and important data that will help us to understand the underlying immunological mechanisms in Goodpasture disease in particular, but also for autoimmune diseases in general. By using the R12 recombinant protein in an assay instead of the complete ␣3(IV) NC1, a more specific diagnostic test could be developed that could distinguish between pathogenic antibodies and harmless autoantibodies. FIG. 4. a, one representative Goodpasture serum is inhibited by different amounts of recombinant proteins. Crosses indicate recombinant ␣1(IV) NC1, boxes recombinant ␣3(IV) NC1, triangles the S1 recombinant, closed circles the S2 and the R12 recombinants, open circles the R134 recombinant, and diamonds the R234, R123, and R124 recombinants. b, all recombinants were tested for their ability to inhibit the binding of antibodies from the Goodpasture patients to native ␣3. Equal amounts of recombinant protein were added, and the greatest inhibition was defined as 100% inhibition and used as reference for determining the inhibition for the other recombinants. Displayed in the figure is the mean inhibition of the sera in percentage for each recombinant. Bars indicate S.D. values. FIG. 5. a, a model (22) of the R12 NC1 domain from type IV collagen, where each ball represents one amino acid. The nine amino acid residues P1-P9 are indicated with a dark gray tone. The epitope region is enlarged, and the different residues are indicated with one-letter symbols. Here the mutated residues are indicated with a dark gray tone. The P1-P9 positions are indicated as well as the charge or polarity of these amino acids. b, a lineup showing the amino acid sequence for the analyzed recombinants covering the principal epitope region, including all positions for substitution except P9. The amino acid in the P9 position is common for the ␣3(IV) NC1 and the recombinants, but differs from the corresponding one of ␣1(IV) NC1.