Mapping of Conformational Epitopes on Human Proteinase 3, the Autoantigen of Wegener’s Granulomatosis

Wegener’s granulomatosis (WG) is an autoimmune small vessel vasculitis with granulomas affecting the upper and lower respiratory tract and kidneys in its generalized form (1). Anti-neutrophil cytoplasmic autoantibodies (cANCAs) recognizing conformational epitopes on proteinase 3 (PR3) are a disease-specific diagnostic marker for WG and are able to activate cytokine-primed neutrophils in vitro by binding to membrane-bound PR3 (2–4). Although the vast majority of PR3 is stored intracellularly in primary granules together with other proteases, small amounts of PR3 prestored in as yet poorly characterized small vesicles are rapidly mobilized during priming. This small fraction of PR3 is externalized together with NB1 (CD177), a glycolipid-anchored membrane protein (5, 6), and is presented on the neutrophil surface as a complex with NB1 (7). Simultaneous interaction of cANCAs with membrane-bound PR3 and FcγR is believed to trigger full activation and premature degranulation of extravasating neutrophils around small vessels in patients with active disease (3).

Although the cause and autoimmune pathogenesis of WG are not yet known, interactions of certain PR3–cANCA subpopulations with PR3 on the surface of neutrophils are believed to contribute to clinical progression and to the propagation of small vessel destruction. This hypothesis, however, has been challenged in view of patients with WG who are constantly cANCA negative, by observations that cANCAs are often absent in early or localized stages of the disease, and by whole-blood analyses indicating that circulating neutrophils are not carrying ANCAs on their surfaces even in active disease (8). As purified PR3-cANCA IgG preparations, but not cANCAs in whole blood, were able to bind to PR3 on the neutrophil surface, conformational alterations or steric inaccessibility of surface-bound PR3 was inferred to explain the low-affinity interaction between PR3 and cANCA. Although cANCA measurements are generally accepted as a valuable diagnostic tool and disease-specific biomarker, their utility for the monitoring of WG patients appears to be unreliable (9). Titer decreases and increases were found to be weakly correlated with remissions and relapses, respectively, and disease activity in general. In addition, absence of cANCA-PR3 deposits in inflammatory lesions and coexistence of normal neutrophil numbers and PR3 autoantibodies in peripheral blood favored the view that cANCAs are rather a consequence than a cause of the disease.

Several attempts were made to determine cANCA binding specificities using PR3-derived linear peptides (10–12) or chimeric molecules composed of human PR3 (hPR3) and human neutrophil elastase or mouse PR3 (13). The differential binding patterns of mAbs and cANCA sera, however, did not lead to the identification of distinct surface regions that were targeted by the Abs. Nonspecific binding of unrelated autoimmune sera to PR3 peptides turned out to be an insurmountable obstacle for meaningful conclusions (14). As these mapping results were of dubious value, biosensor-based competition analyses between different mAbs, on one hand, and cANCA sera from different patients, on the other hand, were performed. With biosensor technology, 13 mAbs to hPR3 were characterized and grouped into four major subsets of Abs that recognized different surface regions of hPR3 (15, 16). mAbs within the same group inhibited binding of each other and hence were assumed to target similar epitope areas on the PR3 surface. mAb 4A5, for example, representing group 3, partially or completely interfered with the binding of 50% of PR3-cANCA sera (15), suggesting that a fraction of PR3-cANCAs and murine mAbs interact with similar surface patches.

In a clinical study with a small number of patient sera, cANCA heterogeneity between patients and changes of binding properties in the same patient have been observed (3). PR3-cANCAs were shown to recognize a surface area on PR3 that is targeted by most, if not all, cANCA sera at the time of active disease presentation (3). At subsequent stages of the disease, PR3-cANCAs changed their epitope specificity and interacted with larger or smaller surface areas on PR3. Such differences in cANCA properties may be linked to their variable pathogenic potential but cannot be recognized by the currently available diagnostic techniques.

To distinguish cANCA subpopulations and Ab titers to individual nonoverlapping epitopes, we aimed at the construction of catalytically active PR3 Ag variants that differ minimally from the human Ag, yet lack one or more major epitopes. To this end, we investigated the natural surface diversity of PR3 homologs from closely related primates. As PR3-cANCAs from WG patients appeared to show little to no cross-reactivity with murine PR3 (17), we assumed that PR3 epitopes are conserved only in homologs of more closely related primate species. Natural residue substitutions in PR3 homologs of other primates should therefore reduce or abolish the binding affinity of PR3-cANCA sera, whereas cleavage and inhibitor specificity as well as overall surface properties should be very similar and most likely identical to hPR3.

To locate a distinct immunogenic subregion within the two subdomains of PR3, we adopted a very conservative approach for the design of unnatural PR3 variants and recombined only the entire N-terminal and C-terminal β-barrels from the human and gibbon species. These two subdomains of chymotrypsin-type serine proteases are homologous to each other and can be reassembled into catalytically active hybrid molecules that retain the functional properties of the parental subdomains (18). To narrow down the surface location of distinct epitopes further, we substituted gibbon-specific residues with those of hPR3 to rebuild human epitopes, and we substituted other residues with those from the more distantly related opossum homolog (Ensembl ID ENSMODP00000008115) to eliminate a third epitope that was common to both gibbon PR3 (gPR3)and hPR3.

The overall aim of this study was to define the structural position of individual Ab epitopes and to construct PR3 templates that permit the measurement of titers of major cANCA subsets independently. Three nonoverlapping Ab binding areas have been found, and one of these binding areas clearly overlaps with the contact area buried by α1-protease inhibitor (α1-PI) after complexation.

Sera from 34 patients with diagnosed WG were collected. Sera from eight healthy blood donors were used as negative controls. This study was approved by the ethics committee of the Ludwig-Maximilians University (Munich, Germany).

The expression vector used in this study was based on pcDNA5/FRT (Invitrogen, Carlsbad, CA). An Igκ-chain secretion signal was integrated into pcDNA5/FRT, using the oligoduplex 5′-Pho-CTA GCCACCATGGAGACAGACACACTCCTGCTATGGGTACTGCTGCTCTGGGTACCAGGTTCCAC-3′ and 5′-Pho-GTGGAACCTGGTACCCAGAGCAGCAGTACCCATAGCAGGAGTGTGTCTGTCTCCATGGTGG-3′ (Metabion, Martinsried, Germany) to allow the secretion of the expressed protein into the cell culture medium followed by a S-peptide tag. Full-length cDNA for hPR3 was amplified with 5′-TGGACACGTGGATGACGACGACAAAATCGTGGGCGGTCACGAGGC-3′ and 5′-CCCCACCGGTTTTCAGGGTAGAACGGATCCA-3′ (Metabion). In this experiment, an enterokinase cleavage site (DDDDK) is introduced at the N terminus of PR3. The resulting PCR product was digested with PmlI/AgeI and cloned into the reading frame of the S-tag. The construct was named pcDNA5/FRT/hPR3-H6. Gibbon cDNA was amplified from granulocyte cDNA, using the primers 5′-TGCGGAGATCGTGGGCGG-3′ and 5′-GGAACGGATCCAGTCCACGTA-3′ (Metabion). The PCR fragment was reamplified with 5′-TGGACACGTGGATGACGACGACAAAATCGTGGGCGGTCACGAGGC-3′ and 5′-CCCCACCGGTTTT CAGGGTAGAACGGATCCA-3′ and cloned into pcDNA5/FRT as described for hPR3 and named pcDNA5/FRT/gPR3-H6. The construction of two human/gibbon chimeras was performed by digestion of pcDNA5/FRT/PR3-H6 with BlpI and AgeI (Fig. 1) and exchange of the respective cDNA segments. A chemically synthesized 390-bp hPR3 fragment was digested with PmlI/BlpI and cloned into the same sites of the pcDNA5/FRT/gPR3-H6 plasmid to reconstruct the human CLB12.8 epitope in gPR3 (gibbon proteinase 3 variant 1 [gPR3v1]). A further modification of this gPR3v1 was done by replacing the A116 to Q119 tetrapeptide with the respective opossum residues Q-V-A-S. To this end, a DNA fragment was chemically synthesized, digested with BlpI and BstEII, and subcloned into the same sites of the gPR3v1 plasmid. In addition, the two glycosylation sites Asn113 and Asn159 were mutated to lysine residues in the gPR3v2 construct. Each construct was verified by sequence analysis. Fig. 1 shows the schematic description of the proPR3 expression constructs with N- and C-terminal tags.

Flp-in HEK293 cells (Invitrogen) were cultured in FreeStyle 293 expression medium or DMEM (Life Technologies, Rockville, MD) supplemented with 10% FCS. For each construct, 1.4 × 10⁶ flp-in HEK293 cells were plated in 35-mm cell culture dishes 1 d prior to transfection. Cells incubating in 3 ml DMEM containing 10% FCS were transfected by the addition of 1.8 μg pOG44 (encoding the flp-in recombinase; Invitrogen) and 0.2 μg pcDNA5/FRT recombinant plasmids in 100 μl serum-free OptiMEM medium and 7 μl FuGENE HD transfection reagent (Roche, Penzberg, Germany). At 48 h after transfection, the medium was replaced with DMEM plus 10% FCS containing 75 μg/ml hygromycin B (Invitrogen) every 2 to 3 d. Two weeks posttransfection, visible circular hygromycin B-resistant colonies were present. Cells were pooled and cultured until a desired number of plates had reached confluence. Cell culture supernatant was tested for recombinant PR3 expression via Western blot, using S-protein conjugated to HRP (Novagen, Madison, WI). For protein expression, cells were cultured in DMEM supplemented with 5% FCS and 75 μg/ml hygromycin B for 8–10 d before supernatants were harvested.

A total of 500 ml of harvested cell culture supernatant was filtered through a 0.22-μm membrane (Millipore, Schwalbach, Germany), concentrated to 100 ml, and dialyzed against starting buffer (20 mM Na₂HPO₄; 500 mM NaCl; 50 mM imidazole, pH 7.5) at 4°C. 6xHis-tagged proPR3 was purified using affinity chromatography. The protein solution was applied to a HisTrap HP column (Amersham Biosciences, Munich, Germany) previously equilibrated in starting buffer. The column was washed with starting buffer, and bound proteins were eluted with a linear imidazole gradient from 50 mM to 1 M imidazole in 20 mM Na₂HPO₄, 500 mM NaCl, pH 7.5. Fractions were collected and analyzed for proPR3 by SDS-PAGE and Coomassie staining. The expected size is 32 kDa. Nearly pure 6xHis-tagged proPR3-containing fractions (>85%) were pooled and concentrated, and protein concentration was determined by measuring the absorbance at 280 nm (spectrophotometer; Eppendorf, Hamburg, Germany) and by the bicinchoninic acid assay.

Purified 6xHis-tagged proPR3 was dialyzed into 20 mM Tris-HCl, 50 mM NaCl, and 2 mM CaCl₂, pH 7.4, at 4°C. The N-terminal S-peptide tag was cleaved off by calf enterokinase (Roche) to generate native active 6xHis-tagged PR3. Enterokinase cleavage was carried out at an enzyme/substrate ratio of 1/20 for 20 h at room temperature. After S-peptide tag cleavage, PR3 was again analyzed using Coomassie-stained SDS-PAGE (reduction in molecular size), and the enzymatic activity was determined using a PR3 synthetic substrate to ascertain that the S-peptide tag was cleaved off. Activity was assessed using 1 mM Boc-Ala-Pro-Nva-4-chloro-SBzl (Bachem, Switzerland) and 0.5 mM 5′,5′-dithio-bis(2-nitrobenzoic acid) in 100 mM Tris-HCl, pH 8.1; 700 mM NaCl; and 1% Igepal (Sigma-Aldrich, Munich, Germany). Substrate hydrolysis was measured in a spectrofluorometer (Fluostar Optima, BMG Labtech, Offenburg, Germany). Inhibition of recombinant PR3 was performed by incubation of α1-PI (Sigma-Aldrich) at 10-fold M concentrations for 30 min at 37°C.

Proteins from the supernatants of degranulated neutrophils were separated by 15% NaDodSO₄-polyacrylamide gel electrophoresis (SDS-PAGE) under native conditions and transferred to a Hybond-ECL membrane (GE Healthcare, Munich, Germany), using a semidry electroblotter (Biometra, Göttingen, Germany). The membrane was blocked in PBS plus 0.05% Tween-20 (PBS-T) containing 5% milk for 1 h at room temperature. The membrane was then washed with PBS-T. S-protein-HRP (Merck, Darmstadt, Germany) was diluted 1 in 5000 and incubated overnight at 4°C. The membrane was washed again and developed with Super Signal West Chemiluminescence Signal (Thermo Fisher Scientific, Bonn, Germany).

Isolation of granulocyte proteins from human and monkey blood samples, as well as native SDS-PAGE followed by immunoblotting, was performed by Utecht & Lüdemann (Klausdorf/Schwentine, Germany). Natural granulocyte proteins from human and primates were separated by nonreducing SDS-PAGE (10% T, 2.5% C) of whole-cell extracts from isolated granulocytes using one wide slot per sample. After semidry transfer, nitrocellulose membranes were cut into 2.5-mm-wide strips, each representing 0.25 × 10⁶ granulocytes. Immunodetection with mAbs or patient sera was done by blocking the strips with milk proteins followed by 2 h of incubation with primary Abs and 1 h of incubation with species-specific alkaline phosphatase-conjugated secondary Abs. 5-Bromo-4-chloro-3-indolyl phosphate and nitro-blue tetrazolium were used for color development.

The 6xHis-tagged PR3 was used in a capture ELISA with precoated nickel-chelate microtiter 96-well plates (Thermo Fisher Scientific). Plates were incubated with purified 6xHis-tagged PR3 at a concentration of 1 μg/ml in PBS-T at room temperature overnight. The plates were washed three times with washing-buffer (Utecht & Lüdemann) and incubated for 1 h with mAbs against PR3 or human sera diluted in diluent buffer (Utecht & Lüdemann). mAbs to PR3 CLB12.8 (Sanquin, Amsterdam, The Netherlands); MCPR3-2 (19), 4A5, 6A6 (Wieslab AB, Malmö, Sweden), 1B10, 1F11, 1F10, and 2E1 (HyTest, Turku, Finland) were diluted 1 in 1000. Monoclonal anti-PR3 WGM2, WGM3, provided by Dr. W. Gross, and WG patient’s sera were diluted 1 to 50. Again the plate was washed three times with washing buffer and then incubated with a secondary anti–mouse AP (1:5000; Sigma-Aldrich) or anti–human AP Ab (1:50; Sigma-Aldrich), respectively. Finally, color was developed using 4-nitrophenyl phosphate disodium salt hexahydrate (pNPP) as a substrate. The OD values were measured at 405 nm. As a negative control, an irrelevant IgG1 mAb was used (control IgG1 Ab was purchased from BD Biosciences, Frankfurt, Germany). All tests were performed in duplicates.

ELISA assays with PR3 complexed to α1-PI was performed as described above. Briefly, PR3 (100 μg/ml) was incubated with a 10-fold M excess of α1-PI (1 mg/ml) for 1 h at 37°C. Completeness of PR3 inhibition by α1-PI was checked via activity measurements. In competition experiments, constant amounts of hPR3 and mAb CLB12.8 were incubated with various concentrations of α1-PI for 30 min at 37°C prior to immobilization of His-tagged PR3 to Ni-NTA plates. Bound CLB12.8 Ab was detected by ELISA using a secondary anti–mouse AP labeled Ab, followed by substrate development with pNPP, as described above.

As conformation-dependent epitopes of hPR3 were not shared with murine PR3 (17), we used more closely related mammalian homologs for hPR3 (Fig. 1) to evaluate the extent of Ab cross-reactivities. Natural residue substitutions in PR3 homologs of primates should lead to a partial loss in binding affinity of cANCA sera, but should not interfere with the biological function and folding pathway of PR3. Fig. 2 shows the sequence alignment of several homologs of mature PR3 from humans, other primates, and mice. We assumed C-terminal trimming after R243 in all species and hence aligned the mammalian homologs only up to this structurally fixed position. With phylogenetic distance, the number of surface substitutions increases: chimpanzee, gibbon, rhesus monkey, and mouse PR3 carry 7, 16, 35, and 68 substitutions, respectively, compared with hPR3. Amino acid residues at positions 103, 119, and 120 are polymorphic in human populations, accounting for three different protein sequences (V103 or I103 in combination with A119 and T120, and I103 in combination with T119, S120). Evolutionary comparisons, however, suggest that V103, A119, and T120 (20–22) of hPR3 represent the ancestral allele (Fig. 2).

Using SDS-PAGE under nonreducing conditions followed by Western blotting, we were able to demonstrate the binding of murine mAbs and cANCAs from several WG patients to the structurally preserved hPR3 Ag on a nitrocellulose membrane (Fig. 3). To determine the influence of multiple natural amino acid variations on Ab recognition, we purified neutrophils from the blood of three primate species: chimpanzee (Pan troglodytes verus), gibbon (Hylobates pileatus), and macaque (Macaca mulatta), and extracted the granule proteins. PR3 from these supernatants, together with other granule proteins, was separated by native SDS-PAGE and blotted onto membranes. Detection of human myeloperoxidase, neutrophil elastase, azurocidin, lactoferrin, and lysozyme served as a control for granule proteins in the supernatants. All 11 WG sera and all mouse anti-PR3 mAbs tested by immunoblotting recognized hPR3 (Fig. 3). The very closely related chimpanzee PR3 was detected by 9 of 11 WG sera and by all anti-PR3 mAbs except for 1F11 (Fig. 3). Furthermore, the cANCA binding signal intensity compared with hPR3 was identical. In contrast, macaque PR3 showed nearly no reactivity to cANCAs. Only two WG sera and two mAbs, WGM1 and 4A5, gave weak signals (Fig. 3). gPR3 showed an intermediate cANCA binding pattern, whereas some anti-PR3 mAbs (e.g., 6A6) and some patient sera recognized the Ag, but the signal intensity was weaker compared with that of hPR3. A negative signal was observed for four WG sera and four anti-PR3 mAbs, namely CLB12.8, 1F10, 1F11, and MCPR3-2 (Fig. 3). Using the same approach, we were able to show an almost complete absence of cANCA binding to rhesus monkey PR3. We suggested that cross-reactivity of PR3-cANCAs with other PR3 homologs declined with increasing evolutionary distance between primates. The most useful mammalian PR3 homolog for epitope discrimination appeared to be gPR3, because it was recognized by some mAbs and most patient sera with significantly diminished signal intensity as compared with the human homolog. This favorable starting point prompted us to determine the complete cDNA and cDNA-derived protein sequence for mature gPR3 by PCR cloning. The natural substitutions found in gPR3 most likely accounted for the difference in cANCA reactivity. gPR3 appeared to be most suited for epitope mapping and further distinction of cANCA subpopulations.

gPR3 was cloned from PMNs of a freshly isolated gibbon (H. pileatus) blood sample and sequenced. It differs from its human homolog by only 16 residues (Fig. 2). Comparative structural analysis of hPR3 and gPR3 revealed the location of amino acid variations, which almost exclusively mapped to and modified the molecular surface. Fig. 4A shows the solvent-accessible surface of gPR3 based on the atom coordinates for hPR3 atoms (1FUJ) (23) (views from the front and the rear) after replacing the human residues by the respective gPR3 amino acids. The surface substitutions found on gPR3 (dyed green) are not equally distributed but are clustered on one side of the surface of the two β-barrels at some distance from the extended substrate binding cleft (Fig. 4A). The only exception is the very conservative Ile-208 substitution on the back side of the hPR3 molecule by a Val, which is one methyl group shorter. Active site residues K99, D61, and R143 implicated in the substrate binding of hPR3 and conferring its specificity are not substituted on gPR3 (24–26). Fig. 4A also illustrates the interface between the two densely packed subdomains, which runs as the substrate binding cleft from northwest to southeast. The homologous N-terminal subdomain carries the amino acid substitutions V35M, S38^AN, L39^BP, R60Q, Q63^AH, Q74R, and L90Q; the C-terminal subdomain substitutions are localized at positions 146, 166, 187, 208, 218, 219 and 223 (Figs. 2, 4A, residue numbering according to Brookhaven Protein Data Bank (PDB) entry 1FUJ).

Recombinant gibbon proPR3 and human proPR3 were expressed in HEK293 cells. Single stable integration of the cDNA constructs into the nucleus was achieved by homologous recombination using the flp-in technology. The flip-in system was faster and more efficient than the conventional transfection protocol (27), and expression from the same nuclear integration site resulted in reproducible stable levels of expression. Constitutively secreted recombinant human and gibbon proPR3 containing an N-terminal S-tag and a C-terminal 6xHis-tag was detected by Western blotting in the culture supernatant using S-protein-HRP (Fig. 5A). Subsequently, proPR3 precursors were purified by single-step chromatography on Ni-NTA Sepharose and activated by cleavage of the N-terminal S-tag propeptide with bovine enterokinase into a highly active mature enzyme, resulting in almost complete conversion. Purified proPR3 and active PR3 were then analyzed by SDS-PAGE. The glycosylated proPR3 variants had the expected relative molecular mass of ~32 kDa, whereas N-terminally trimmed activated PR3 ran faster with a relative molecular mass of ~29 kDa (Fig. 5B). The gPR3 variant 2, gPR3v2 (V35M, S38^AN, I38^BP, Q74R, N113K, N159K, 119-122QVAS) (Fig. 4D), lacks both glycosylation sites at positions 113 and 159 and therefore runs faster at a different position before and after conversion. Most importantly, all N-terminally processed PR3 variants showed enzymatic activity, thereby confirming correct formation of disulfide bridges and folding of the molecules. The highly specific FRET substrate previously developed for hPR3 [Abz-VADCADQ-EDDnp (24, 28)] was cleaved by both gPR3 and hPR3 with similar efficacy. Purified human α1-PI inhibited gPR3 and formed a covalently linked complex with it, as shown by SDS-PAGE. Likewise, gPR3 was inhibited by purified human elafin, a canonical low-m.w. inhibitor of hPR3 (not shown) confirming the strong conservation of its functional properties.

A capture ELISA was developed using the C-terminal 6xHis-tag for immobilization of PR3 to nickel-coated microtiter plates. This technique avoids the drawbacks of direct immobilization or capturing of the autoantigen with an immobilized murine Ab. The natural substitutions found on gPR3 prevented the interaction of mouse mAbs CLB12.8 and MCPR3-2 belonging, respectively, to Ab groups 1 and 4, as previously defined (16). The ELISA data with purified recombinant Ags agreed with Western blot results showing that CLB12.8 and MCPR3-2 could not bind to gPR3. mAb 6A6, however, gave a signal with gPR3, although clearly weaker than with hPR3 in Western blots, but did not bind to gPR3 on ELISA plates. In all ELISA experiments, recombinant hPR3 served as a positive control (Fig. 6A).

As the residue substitutions in gPR3 are distributed over the surfaces of both barrels, we constructed chimeric molecules between hPR3 and gPR3 by recombining the N- and C-terminal six-stranded β-barrels of both homologs. These compact subdomains represent the smallest folding module of a serine protease domain. As a result, we obtained four sequence variants of the autoantigen, the human, the gibbon, the g/h (N-terminal β-barrel of gibbon origin) and the h/g (N-terminal β-barrel of human origin, Fig 4B) variant. The mAbs CLB12.8 and MCPR3-2 bound to h/gPR3, but not to g/hPR3. A murine isotype control Ab did not show any reactivity to either hPR3 and gPR3 nor gibbon-human chimeras (Fig. 6A). These results clearly indicated that amino acid residues in the N-terminal subdomain of hPR3 were essential for the binding of CLB12.8 and MCPR3-2 to hPR3. The four residues, M35, N38^A, P38^B, and R74, are substituted in gPR3 and are locally clustered on the hPR3 surface, whereas R60, Q63^A, and L90 form a second nonoverlapping cluster of surface exposed residues.

Our next goal was to map the location of the two nonoverlapping epitopes recognized by CLB12.8 and MCPR3-2, respectively. In agreement with Rooney et al. (29), the His-tagged hPR3 in a complex with α1-PI was not recognized by CLB12.8 in the Ni-NTA capture ELISA (Fig. 6B). Structural examination and inspection of the related covalent α1-PI–pancreatic elastase complex (PDB 2D26) (30) showed that residues of the 143–149 loop and residues from the 186–190 stretch of pancreatic elastase were located within the contact area between porcine pancreatic elastase and α1-PI. On the basis of the known porcine pancreatic elastase α1-PI structure, we, moreover, suggested that the residues of the loop at positions 35, 38^A, 38^B, and 74 (PDB 1FUJ) were sterically shielded or distorted at least in part by α1-PI complexation. These four residues appeared to directly represent important structural determinants of the CLB12.8 epitope, but distortions and loss of structural order in the complexed hPR3 at more distant sites could also explain the loss of CLB12.8 binding. To determine which of the two hypotheses was correct, we replaced the residues at these four positions in gPR3 by the respective human residues V35M, S38^AN, L39^BP, and Q74R, resulting in the gPR3 variant 1 (gPR3v1) (Fig. 4C,). By changing only these four amino acid residues in gPR3, the binding site for mAb CLB12.8, but not that for MCPR3-2, was reconstituted (Fig. 6C). These data clearly indicate that two distinct epitopes are targeted by the two mAbs. Moreover, our experiments prove that the CLB12.8 epitope is critically shaped by the surface residues M35, N38^A, P38^B, and R74, whereas R60, Q63^A, and L90 are the major determinants of the MCPR3-2 epitope. Another mAb 6A6 that was previously shown to compete with the binding of CLB12.8 also displayed the same binding properties. First, 6A6 weakly interacted with uncomplexed gPR3 and hPR3–α1-PI complexes in the ELISA (Fig. 6B, 6C). Second, 6A6 bound to the gPR3v1 (Fig. 6D). These experiments clearly demonstrate that 6A6 recognizes similar structural determinants on hPR3 as CLB12.8. We also evaluated the Ab WGM3, which mutually inhibited the binding of MCPR3-2 but was previously not assigned to a single epitope group. Like MCPR3-2, WGM3 bound neither to gPR3 nor to the humanized gPR3v1 variant, demonstrating that these two mAbs recognize highly overlapping surface structures on hPR3 (not shown). As the MCPR3-2 and 12.8 epitopes were different, we examined whether the binding region of MCPR3-2 also overlapped with the interface of the hPR3–α1-PI complex and was, therefore, not accessible after complexation to α1-PI. Capture ELISA assays, however, showed that MCPR3-2 binding was not impaired or abolished by α1-PI complexation (Fig. 6B). These data indicate that MCPR3-2 is targeting a surface region of hPR3 that does not undergo conformational changes after complexation with α1-PI.

Regarding the epitope recognized by 4A5 and other Abs with similar target specificity, we suspected its location on the back side of PR3 around the linker connecting the two β-barrels. Naturally occurring striking deviations from the human linker sequence were not observed in primates, but in more distantly related mammals, in particular in the oppossum homolog for PR3. The starting template for further modifications was gPR3v1, which carried the reconstructed CLB12.8 epitope and differed from hPR3 at 12 positions in total. In this template, we modified the gibbon ATVQ linker sequence at four positions (119–122) to match exactly the opossum sequence QVAS (Fig. 4D). Moreover, the two N-linked glycosylation sites at positions 113 and 159 have been removed by converting the asparagines to lysines. This gibbon-based gPR3v2 variant was no longer able to interact with the mAbs 4A5 and WGM2 (Fig. 6D), but retained its binding to CLB12.8. As a complement to these findings, we observed tight binding of 4A5 and WGM2 to the hPR3–α1-PI complex (not shown). In summary, we identified the precise structural determinants of three groups of mAbs (represented by the mAbs MCPR3-2, CLB12.8, and 4A5, respectively), as defined by van der Geld and coworkers (16). In this way we also developed the tools to measure epitope-specific cANCA titers.

Sera from 34 patients with WG and sera from 6 healthy blood donors were analyzed by a 6xHis-capture ELISA with recombinant PR3 Ags. Results are shown in Fig. 7. All WG sera recognized the hPR3 Ag, whereas sera from healthy controls did not bind to any of the tested recombinant proteins (not shown). Reactivity of cANCAs with gPR3, human-gibbon chimeras, and gPR3 variants was determined as the percentage of the signal obtained with the recombinant hPR3 Ag. In this way we divided our set of WG ANCA sera into two groups. WG1 sera (n = 15) contained autoantibodies that bound to gPR3 and h/gPR3 with similar efficacy and showed at least 70% of the reactivity obtained with hPR3 (Fig. 7A). The second group, WG2 (n = 19), showed <70% binding to gPR3 (Fig. 7B). Binding of cANCAs to the h/gPR3 chimeras, however, was about the same as observed with hPR3 in the WG2 subgroup. To narrow the major target specificity of WG2 ANCAs further, we tested several sera from the WG2 subgroup (n = 10) for gPR3v1 reactivity (Fig. 7C). Again, PR3 binding to this PR3 mutant was reconstituted close to those levels obtained with the human Ag. These data indicate that major subpopulations of cANCAs in the WG2 subgroup are directed toward the N-terminal hPR3 subdomain and interact with similar conformational determinants as a major group of murine Abs (CLB12.8, 6A6) that were previously defined by their pattern of mutual binding competition.

To map the approximate location of additional cANCA binding sites, we examined the influence of α1-PI complexation on cANCA binding, using the Ni-NTA capture ELISA. With the exception of two serum samples, cANCA titers from both patient groups WG1 and WG2 were strongly reduced when the α1-PI–PR3 complexes were used as the target Ag (Fig. 8A). These results indicate that α1-PI obscures more binding sites than just the CLB12.8 epitope (Fig. 9). As the 4A5/WGM2 and MCPR3-2/WGM3 regions are still accessible in the α1-PI–hPR3 complex, we anticipate that these additional α1-PI–sensitive cANCA binding sites are located somewhere else, most likely in proximity with the substrate interacting region.

Autoantibody-induced neutrophil activation via membrane-bound Ag and IgG receptors is widely regarded as an important pathogenetic mechanism in small-vessel vasculitis (3). Under temporary conditions of low α1-PI levels and cANCAs with high affinity and fast association rates, cANCAs may well outcompete the initial substrate-like interaction between α1-PI and hPR3 (Fig. 10, red arrow, situation 2). To simulate this situation with purified components, we used the mAb CLB12.8, which did not bind to α1-PI–inhibited hPR3 complexes. We varied the α1-PI concentration and examined the extent of CLB12.8 binding to the His-tagged hPR3 at physiological and subphysiological plasma inhibitor concentrations (Fig. 8B). At physiological concentrations of α1-PI, but not at low concentrations, the serpin was able to form a complex with hPR3. As shown previously, hPR3–α1-PI complexes are no longer retained by CD177 or the lipid bilayer on cellular membranes. Hence our experiments illustrate why the neutrophil-activating potential of cANCAs can vary in patients, depending on the association rates, target specificities, and titers of Abs, as well as on local concentrations of hPR3 inhibitors.

cANCAs from WG patients are known to be directed against three-dimensional features of hPR3, though linear peptides were initially used to distinguish different antigenic regions with conflicting and ambiguous results (14). Peptide-coated pins display a high density of various peptide conformations and are, therefore, prone to capture irrelevant Abs from complex patients’ sera. By definition, so-called continuous, linear, or sequential epitopes can be mimicked by synthetic soluble peptides that adopt a huge number of different conformations in solutions or occasionally a partially ordered secondary structure. Long linear peptides have the advantage of mimicking secondary structures and can also form an epitope that consists of two discontinuous determinants on a long peptide. Hence longer 15-mer peptides derived from hPR3 have also been employed to screen the sera of WG patients (11). Four peptide areas consisting of several 15-mer peptides were better recognized by WG sera than by control sera, but not by mouse mAbs.

The value of peptide arrays in mapping the epitopes of native proteins, however, is limited. It is known that as many as 90% of all B cell epitopes on native proteins are conformational rather than linear. During an autoimmune response, B cells most likely interact with the folded stable autoantigen, not with partially degraded and unfolded peptide fragments. Such fragments would be removed and degraded very rapidly in biological fluids. The solvent-accessible surface area that is typically buried by Ab binding amounts to 700–1000 Ų (31), whereas the entire surface area of hPR3 is only 9800 Ų. As the typical contact area between a native protein Ag and an Ab is much larger than between a peptide or a hapten and an Ab, a conformational epitope of hPR3 is expected to comprise a number of residues from at least two different noncontinuous peptide segments that have close surface proximity. Crystallographic studies of Ag–Ab complexes revealed some structural characteristics of Ab binding sites on protein surfaces (32). Five to seven core residues (hot spot residues) of the protein Ag are clustered in the center of the interface and dominate the molecular interaction with the Ab (32). Up to 10 additional residues at the periphery of the interface shield this core from the bulk solvent, but do not contribute so much to the binding free energy of the complex (32). Core interacting residues are often located in highly flexible turn or surface loops that are not under strong evolutionary constraints. Binding site selection by B cells is, conversely, an evolutionary process that operates in an organism’s own body and exploits a highly diversified repertoire of Ab templates in the body. We hypothesized that favorable shape complementarities and optimized interactions are more easily obtained for flexible regions of an Ag, which are also known to be reshaped more often during phylogenetic evolution. PR3 homologs of the primate lineage were, therefore, considered as useful initial probes to evaluate the involvement of nonconserved residues in Ab binding. By further variations of the PR3 surface, we were able to distinguish three nonoverlapping epitope areas that are targeted by murine mAbs.

The first group of Abs, represented by CLB12.8, 6A6, and PR3G-2 (16), binds to a region on hPR3 that differs from gPR3 in two closely spaced surface loops, the so-called 37-loop and 70-loop. M35, N38^A, and P38^B are substituted by V35, S38^A, and I38^B, respectively, whereas R74 of the 70-loop is replaced by Q74 in gPR3 (Fig. 9) (33). Reversing these four gibbon substitutions with the respective human residues reconstitutes the CLB12.8 epitope, proving that these four residues are located within the contact area of CLB12.8. In addition, complex formation between hPR3 and α1-PI impairs the accessibility of both loops as predicted from a model of the pancreatic elastase α1-PI. In line with this prediction, binding of CLB12.8 to the hPR3–α1-PI complex was lost. By contrast to all other epitopes we identified, the CLB12.8 epitope is unique in that it is occluded or altered by α1-PI binding and is very close to the active site region. Elafin, another physiological peptide inhibitor of PR3, binds noncovalently in a canonical manner to PR3 (34). As it consists only of 57 residues, the anticipated substrate-like contacts between elafin and PR3 are fewer and are restricted to the S5 to S2′ pockets. According to the crystal structure of the homologous porcine elastase-elafin complex (PDB 1FLE) (35), the 37- and the 70-loops of porcine elastase are not rearranged after elafin binding, and hence the CLB12.8 epitope in the PR3-elafin complex is fully accessible, as demonstrated experimentally.

The second Ab binding region of the N-terminal subdomain is located to the north of the active site binding cleft and has been identified with the help of three natural variations of the gibbon homolog. The 60-loop of gPR3 displays two differences. The divergent residues in gPR3 are Q60 and H63^A, instead of R60 and Q63^A in hPR3. The third variation concerns the hydrophobic L90, which is Q in gPR3. This residue is at the beginning of the long 99-loop and in close proximity to the R60 and Q63^A residues (Fig. 9). We found that MCPR3-2 and WGM3, a mAb of the IgM class, bind to this surface area, previously designated as epitope region 4 (16). Both Abs can bind to PR3–α1-PI complexes and evidently do not extend into the α1-PI contact area. These findings are consistent with previous competition experiments showing that MCPR3-2 and WGM3 mutually inhibit their binding to immobilized PR3. Our attempts to map the mAb PR3G-4 were unsuccessful because of its low and unspecific binding. Hence no Ab could be assigned to the hypothetical epitope area 2 (16).

Another group of Abs belonging to the epitope group 3 (16) showed binding to both hPR3 and gPR3. The only larger surface completely shared between both species is found on the back side of PR3, opposite the active site region. The linker (residues S115 to H132) that connects the N- and C-terminal β-barrel runs across the back side of PR3 and has been reported as one of several linear regions that bind Abs from WG sera (10, 11). We have, therefore, replaced the residues ATVQ between positions 119 and 122 by the naturally occurring sequence QVAS, which is present in the opossum homolog of PR3 (Figs. 4D, 9). Indeed, these minor sequence changes led to the loss of the binding site for two mAbs we tested: 4A5 and WGM2.

The binding of PR3-specific ANCAs to neutrophil cell surfaces presenting either constitutively expressed or degranulation-induced PR3 is the most attractive explanation for the contribution of these autoantibodies to the pathogenesis of WG. Locally primed neutrophils expose the autoantigen on their surface either in a complex with CD177 or in direct association with plasma membrane lipids transiently. Peripheral blood neutrophils from WG patients with ANCAs, however, do not carry surface-bound autoantibodies (8). Specific IgG could not be detected on neutrophils after their exposure to ANCA sera. Various explanations for the absence of PR3-ANCAs on neutrophils suspended in whole blood or plasma dilutions have been put forward, but the role of the α1-PI from plasma has so far been neglected. As shown previously, this abundant plasma inhibitor interacts with membrane-bound active PR3 and rapidly removes it as a covalent complex from neutrophil membranes (36). Certain Abs (e.g., those from rabbits or murine mAbs with fast association kinetics, high affinity, and the appropriate target specificity) may well be able to anticipate the action of α1-PI and to stabilize the membrane-bound PR3 by secondary interactions with FcγRs. When such PR3-autoantibodies are formed in WG patients and α1-PI levels are low, newly exposed membrane-associated PR3 could readily mediate neutrophil activation via FcγRs upon autoantibody binding (Fig. 10).

In support of this view, we mapped the target specificity for a group of mAbs (CLB12.8, 6A6, PR3G-2) that bind in close proximity to the active site region. One of these, CLB12.8, was previously shown to induce a respiratory burst, and this pathogenic effect was indeed suppressed by α1-PI at higher concentrations (29). Although these authors did not observe binding to PR3-inhibitor complexes, another group (37) reported successful quantification of PR3–α1-PI complexes in inflammatory fluids using the same CLB12.8 monoclonal as a capture Ab. We re-examined these conflicting findings and confirmed that the epitope of CLB12.8 was inaccessible in PR3–α1-PI complexes in agreement with the location of amino acid substitutions that abrogated CLB12.8 binding.

To evaluate the functional significance of the CLB12.8 epitope in WG, we compared the reactivity of WG sera with hPR3, gPR3, and variants thereof lacking the CLB12.8 epitope. In this way, we identified a major group of sera whose ANCAs were directed toward a similar conformational epitope as recognized by group 1 mAbs (CLB12.8, 6A6, PR3G-2). According to our observations, a significant number of cANCA sera appear to contain IgG populations that share the binding specificity with CLB12.8 and could thus stimulate neutrophils under certain conditions like CLB12.8 (“the bad ANCAs”).

As shown by several previous studies, purified cANCAs are able to interfere with the enzymatic activity of soluble PR3 (38–40), but not in all studies (10, 41–43). In view of our results with PR3 mutants and PR3-α1-PI complexes, some ANCA populations most likely recognize structural determinants around the active site region and could, therefore, prevent or delay the interaction between membrane-bound PR3 and α1-PI. These Abs would therefore have a high pathogenic potential, as these Ab-PR3 complexes cannot be cleared and stripped off from membranes by α1-PI (“the ugly ANCAs”). A third, so far theoretical, category of PR3 autoantibodies (“harmless ANCAs”) consists of those whose epitope is buried by CD177 or lipid interactions. Such Abs would not be able to activate neutrophils and hence would have the lowest pathogenic potential. So far, no Ab that prevents membrane binding of secreted PR3 has been described. Abs with this target specificity may be beneficial, as they prevent the association of PR3 with cellular membranes. As circulating PR3-cANCAs are not always pathogenic and can even persist during remission, target specificity, plasma levels, and association kinetics of autoantibodies with PR3 may be critical for their pathogenicity and should be taken into account. According to our concept, the most harmful PR3 autoantibodies react with PR3 on neutrophil surfaces faster than α1-PI and activate neutrophils before α1-PI can interfere (Fig. 10). Alternatively, a catalytically inactive proform of PR3 may be released by neutrophils in WG patients that cannot be trapped and removed by α1-PI. Besides that, α1-PI variants in the plasma of WG patients that poorly react with catalytically active PR3 may favor neutrophil activation by PR3-ANCAs. Certain α1-PI alleles are indeed correlated with an increased risk for WG (44). Although a number of observations underscore the potential relevance of epitope-specific ANCA discrimination, the predictive value and clinical impact of epitope-specific ANCA measurements remain to be determined. Larger studies with statistical power for deeper subgroup analyses are needed to develop new prognostic parameters for WG relapses and appropriateness of long-term therapeutic interventions.

We thank Dr. E. Csernok, Dr. W. Gross (both from the Department of Rheumatology, University of Lübeck, Lübeck, Germany), and Dr. Heeringa (Department of Pathology and Medical Biology, University Medical Center Groningen, Groningen, The Netherlands) for providing mouse monoclonal anti-human proteinase 3 Abs (WGM-2, WGM-3, and PR3G-4) and H. Kittel and E. Stegmann (both from the Max-Planck-Institute of Neurobiology, Martinsried, Germany) for skillful technical assistance.

Disclosures The authors have no financial conflicts of interest.