Proteolysis Breaks Tolerance toward Intact α345(IV) Collagen, Eliciting Novel Anti–Glomerular Basement Membrane Autoantibodies Specific for α345NC1 Hexamers

Glomerulonephritis (GN), an important cause of kidney injury leading to end-stage renal disease, is often due to autoimmune responses against components of the glomerular filtration barrier. A prominent target is α345(IV) collagen, the major component of glomerular basement membranes (GBM). Formation of supramolecular collagen IV networks is driven by specific interactions among the noncollagenous domains, which direct the assembly of chains into trimeric molecules and then associate end-to-end forming noncollagenous domain 1 (NC1) hexamers. The NC1 domains of α345(IV) collagen are the targets of Abs associated with the most aggressive forms of rapidly progressive GN (1, 2). In anti–GBM Ab GN, including Goodpasture disease, IgG autoantibodies mainly target two conformational epitopes within α3NC1 monomers (3–6). The immunodominant α3NC1 autoepitopes abut the α4NC1 and α5NC1 subunits within native α345NC1 hexamers, which prevents the binding of autoantibodies until hexamers dissociate (7, 8). An alloimmune form of anti–GBM Ab GN, Alport posttransplant nephritis (APTN), occurs in some patients with Alport syndrome after a kidney transplant. These patients typically have mutations in COL4A3, COL4A4, or COL4A5 genes that prevent the normal assembly of α345(IV) collagen. APTN is mediated by anti–GBM alloantibodies that bind to the kidney allograft, targeting α3NC1 and/or α5NC1 epitopes accessible within α345NC1 hexamers (6, 9–12).

In anti–GBM Ab GN, the hallmark findings on the renal biopsy are smooth linear GBM deposition of IgG (mostly IgG1 and IgG4) and often complement C3, with crescentic and necrotizing injury as a result of complement- and FcγR-mediated inflammation. Clinically, the disease usually manifests as rapidly progressive crescentic GN, sometimes accompanied by life-threatening pulmonary hemorrhage due to autoantibodies binding to alveolar basement membranes (13). Prompt diagnosis and treatment with immunosuppressive drugs and plasma exchange are critical because of the fulminant course of the disease, with mortality >50% in untreated patients (14). However, some patients exhibit only mild glomerular involvement with normal renal function despite biopsy-proven linear GBM IgG autoantibodies (15–18). The molecular basis of this atypical presentation is not fully understood.

Deciphering the mechanisms uncoupling the anti–GBM autoantibodies from rapidly progressive GN can provide insights into the pathogenesis of Ab-mediated disease and may identify new targets for therapeutic intervention. By characterizing serum autoantibodies from a patient with strong linear GBM deposition of IgG but atypically mild nephritis, we identified a novel type of IgG4 subclass–restricted anti–GBM autoantibodies that selectively targeted α345NC1 hexamers but did not bind to α3NC1 monomers, unlike typical Goodpasture autoantibodies. The mechanisms eliciting the novel anti–GBM autoantibodies were studied in mouse models recapitulating this phenotype. Our results showed that tolerance toward autologous intact α345(IV) collagen is established in hosts expressing this Ag, even though autoreactive B cells specific for α345NC1 hexamers are not purged from the repertoire. Proteolysis selectively breaches this tolerance by generating autoimmunogenic α345NC1 hexamers. This provides a mechanism eliciting autoantibodies specific for quaternary epitopes within α345NC1 hexamers, which are restricted to noninflammatory IgG subclasses and are nonnephritogenic.

Recombinant NC1 monomers of human collagen IV chains were expressed in HEK293 cells and purified as described; recombinant α345NC1 hexamers assembled in vitro from NC1 monomers were used in some experiments (19). NC1 hexamers of GBM collagen IV, comprising a mixture of α345NC1 and α12NC1 hexamers, were solubilized by collagenase digestion of human and mouse kidney cortex basement membranes and purified by an ion-exchange column and gel-filtration chromatography (20). The α12NC1 hexamers were separated from α345NC1 hexamers by affinity chromatography using immobilized anti–NC1 mAbs (20).

Anti–GBM autoantibodies were analyzed in serum from a patient with mild nonprogressive GN, which tested negative for anti–GBM Abs using a commercial clinical immunoassay despite strong linear GBM staining for IgG on the renal biopsy. A detailed case report will be presented elsewhere (D.B. Borza, X. Geiger, A. Wasiluk, and A.B. Fogo, manuscript in preparation). Positive controls comprised sera containing previously characterized anti–GBM Abs (8, 9, 21). Normal human sera (Innovative Research, Novi, MI) were used as negative controls.

ELISA was performed in MaxiSorp plastic plates coated overnight with NC1 monomers (100 ng/well) or NC1 hexamers (300 ng/well) and then blocked with 1% BSA. Sera were diluted 1:50 unless otherwise indicated. Secondary Abs were alkaline phosphatase–conjugated anti–human IgG (American Qualex, San Clemente, CA) and peroxidase-labeled mAbs HP6069, HP6014, HP6047 (Zymed, San Franscisco, CA), and HP6023 (SouthernBiotech, Birmingham, AL) specific for human IgG1, IgG2, IgG3, and IgG4 subclasses, respectively. Human IgG4 was also detected using biotin-labeled mAbs HP6025 (Sigma-Aldrich, St. Louis, MO). Subclass-specific secondary Abs were diluted to yield approximately equal absorbance readings when purified human myeloma IgG1–IgG4 (100 ng/well; Sigma-Aldrich) were used as standards. After addition of substrate for phosphatase (Sigma-Aldrich) or peroxidase (Bio-Rad, Hercules, CA), color development was measured with a SpectraMax Plus 384 ELISA plate reader (Molecular Devices, Sunnyvale, CA).

DBA/1J mice were purchased from The Jackson Laboratory. Fcgr2b^−/− mice backcrossed onto the DBA/1 background for >12 generations (D1.Fcgr2b^−/−) were obtained from Dr. Sandra Kleinau. B6.Col4a3^−/− mice (22) were obtained from Dr. Jeffrey Miner. Col4a3^−/− mice expressing a human COL4A3/COL4A4 YAC transgene (23) on a mixed genetic background were obtained from Dr. Laurence Heidet. The genotype of mutant mice was verified by PCR. Mice were housed in a specific pathogen-free facility with free access to food and water. All procedures were approved by the Institutional Animal Care and Use Committee and conducted in accordance with the Guidelines for Animal Care and Use Program of Vanderbilt University.

Mice of either sex, between 6 and 10 wk old, were used for experiments. For immunizations, intact collagen IV from mouse GBM (5 mg) or purified NC1 hexamers from mouse GBM (100 μg) in 50 μl PBS were emulsified in an equal volume of CFA (Sigma-Aldrich), then injected s.c. at two sites on the back. Mice were boosted 3 wk later with Ag in IFA (Sigma-Aldrich). In control mice, the Ag was replaced by PBS. Mice were regularly checked for signs of disease. Blood was collected every 2 wk from the saphenous vein. Spot urine was collected every 2 wk. Mice were killed at 10 wk postimmunization. Blood and kidneys were collected for further analyses.

Urinary albumin excretion was measured in spot urine samples by capture ELISA using a mouse albumin quantitation kit (Bethyl Laboratories, Montgomery, TX). Urine creatinine and blood urea nitrogen (BUN) were measured using Infinity creatinine and urea liquid stable reagents (Thermo Fisher Scientific, Middletown, VA) according to the manufacturer’s protocols. Albuminuria was expressed as the urinary albumin/creatinine ratio (ACR). Portions of mouse kidneys or lungs were fixed in 10% buffered formalin, dehydrated through graded ethanols, embedded in paraffin, and kidney sections (2 μm thick) were stained with periodic acid-Schiff. At least 50 glomeruli from each mouse were observed to assess lesions. For transmission electron microscopy (EM), kidney cortex was fixed in 4% paraformaldehyde in 0.1 M cacodylate buffer (pH 7.4), postfixed in aqueous 1.25% osmium tetroxide, dehydrated through an ethanol series, embedded in plastic, sectioned with a diamond knife, and stained with 4% uranyl acetate and lead citrate.

Portions of snap-frozen mouse kidneys or lung embedded in OCT were cryosectioned (5 μm), fixed in acetone for 10 min at −20°C, and blocked with 1% BSA. For direct immunofluorescence (IF), frozen sections were stained with FITC-conjugated goat anti–mouse IgG, IgG1, IgG2a, IgG2b (BD Biosciences/Pharmingen, San Jose, CA) or FITC-conjugated goat anti–mouse C3c (Nordic Immunology, Tilburg, The Netherlands). Secondary Abs were Alexa Fluor 488–conjugated goat anti–rabbit and anti–rat IgG (Invitrogen, Carlsbad, CA) or FITC-goat anti–rat IgG (BD Biosciences/Pharmingen). Sections were mounted with anti-fade reagent (Invitrogen) and examined under a Nikon Eclipse E800 epifluorescence microscope. Photomicrographs were recorded with a charge-coupled device digital camera, using the same exposure settings for each primary Ab.

Kidney-bound Abs were eluted from homogenized mouse kidney cortex with 0.1 M glycine (pH 2.8). Serum and kidney-eluted mouse IgG (mIgG) Abs were analyzed by ELISA, as described for human autoantibodies. Secondary Abs were alkaline phosphatase–conjugated goat anti–mouse IgG (Rockland Immunochemicals, Gilbertsville, PA) and HRP-conjugated goat anti–mouse IgG1, IgG2a, IgG2a^c (for C57Bl/6J mice), and IgG2b (Bethyl Laboratories). The dilutions of secondary Abs were chosen to yield comparable ELISA readings when standard amounts of murine IgG1, IgG2a, and IgG2b mAbs against mouse α345NC1 hexamers were used as positive controls in the same assay.

Data are shown as means ± SD. Statistical analyses were performed using GraphPad Prism version 5.01. The significance of differences among groups was evaluated by one-way ANOVA test, followed by post hoc tests for pairwise comparisons. A p value <0.05 was considered to be statistically significant.

Serum from a patient with mild proteinuria, microscopic hematuria, and elevated serum creatinine but stable renal function, whose renal biopsy showed linear GBM deposition of IgG but no crescents, tested negative for anti–GBM autoantibodies using a commercial clinical assay. Because serology implied the absence of IgG autoantibodies against α3NC1 monomers, a serum sample was analyzed for the presence of autoantibodies to other collagen IV NC1 domains. Indirect ELISA showed IgG autoantibodies selectively binding to human α345NC1 hexamers, but not to α3NC1, α4NC1, or α5NC1 monomers (Fig. 1A); typical autoantibodies from Goodpasture sera bound equally well to α3NC1 monomers and α345NC1 hexamers. Specificity for α345NC1 hexamers was corroborated by inhibition ELISA (Fig 1B). IgG binding to native NC1 hexamers from human GBM was abolished by hexamer dissociation (Fig. 1C). Analysis of subclasses of IgG binding to α345NC1 hexamers revealed that the patient’s autoantibodies were restricted to human IgG4, consistent with linear GBM staining for human IgG4 (hIgG4) in the renal biopsy, whereas Goodpasture sera also contained IgG1 besides IgG4 autoantibodies (Fig. 1D). IgG4 autoantibodies binding to α345NC1 hexamers were not found in normal human sera (Fig. 1E). Overall, these results identified a novel type of anti–GBM autoantibodies binding to α345NC1 hexamers but not α3NC1 monomers, restricted to hIgG4 subclass, and associated with mild GN.

We used mouse models to further investigate the mechanisms eliciting the novel anti–GBM autoantibodies and breaking tolerance toward α345(IV) collagen. Based on the specificity of autoantibodies, we hypothesized that the α345NC1 hexamer is the culprit autoantigen. In Col4a3^−/− mice, which cannot assemble α345(IV) collagen, immunization with native NC1 hexamers from normal mouse GBM elicits IgG Abs selectively targeting α345NC1 hexamers, but not α3NC1 monomers (10). However, it is not known whether mouse GBM NC1 hexamers can also induce IgG autoantibodies in mice with normal expression of α345(IV) collagen. This was addressed next.

We chose DBA/1 mice for experiments because this strain develops severe Ab-mediated glomerular disease after immunization with α3NC1 monomers (24–26). We also used congenic D1.Fcgr2b^−/− mice because the ablation of the inhibitory IgG Fc receptor FcγRIIB heightens humoral autoimmunity and Ab-mediated inflammation (27) and was required to induce GN in mice injected with human anti–GBM alloantibodies (21). The findings were essentially identical in DBA/1 wild-type and Fcgr2b^−/− mice. All mice immunized with autologous mouse GBM NC1 hexamers had normal renal function for 10 wk postimmunization, as judged from urinary albumin excretion and blood urea nitrogen levels, similar to those in control mice immunized with adjuvant alone (Fig. 2A). At 10 wk after immunization, kidney histology was normal, undistinguishable from that of control mice (Fig 2Ba–c). Direct IF staining revealed linear GBM staining for mouse IgG (Fig 2Bd–f), but no glomerular C3 deposition (Fig 2Bg–i). The GBM and podocytes appeared normal by transmission EM, whereas electron-dense deposits were completely absent (Fig. 2Bj, k). These findings contrast with those in α3NC1-immunized DBA/1 mice, which develop massive albuminuria, have granular deposits of IgG and C3 in the GBM, and exhibit GBM thickening with subepithelial immune complexes and podocyte foot process effacement by EM (24–26).

The specificity of circulating and kidney-bound mIgG autoantibodies in hexamer-immunized DBA/1 mice was analyzed by ELISA. IgG from sera (Fig. 3A) and kidney eluates (Fig. 3B) bound to mouse GBM NC1 hexamers; minimal binding to mouse α12NC1 hexamers implied that the autoantibodies were specific for α345NC1 hexamers. Unlike sera or eluates from α3NC1-immunized mice, those from hexamer-immunized mice did not bind to human α3NC1 monomers. As only human but not mouse α3NC1 monomers were available, and for comparison with human anti–GBM autoantibodies, we further verified that mouse IgG autoantibodies bind to native but not dissociated human GBM hexamers and to α345NC1 hexamers assembled in vitro (Fig. 3C).

Analysis of IgG subclasses of autoantibodies binding to mouse GBM NC1 hexamers from sera (Fig. 4A) and kidney eluates (Fig. 4B) revealed the presence of mIgG1 but not mIgG2a nor IgG2b autoantibodies. In corroboration of the ELISA results, analysis of IgG subclasses bound to GBM by direct IF staining showed linear GBM deposition of mIgG1 but not mIgG2a and mIgG2b subclasses (Fig. 4C). In contrast, small amounts of mIgG2a and mIgG2b accompany mIgG1 autoantibodies in the sera and kidneys of α3NC1-immunized DBA/1 mice (24). Thus, autologous GBM NC1 hexamers elicit autoantibodies restricted to mIgG1 subclass, which bind to mouse GBM in vivo but do not cause GN. These murine autoantibodies recapitulate the properties of human anti–GBM autoantibodies (Fig. 1) with respect to specificity, subclass restriction, and lack of nephritogenicity.

NC1 hexamers occur in tissues as parts of supramolecular collagen IV networks. We next evaluated whether intact collagen IV from mouse GBM also elicits autoimmune responses in DBA/1 and D1.Fcgr2b^−/− mice. Renal function (urine ACR and BUN) remained normal for 10 wk after immunization (Fig. 5A). Kidney histology was normal (Fig. 5Ba). Direct IF staining (Fig. 5Bb–f) revealed weak-to-moderate staining for total mIgG and mIgG1 (but not mIgG2a and mIgG2b) in a mesangial pattern, potentially due to Abs against Ags other than α345(IV) collagen present in the mouse GBM preparation. Glomerular deposition of C3 was absent. Neither sera nor kidney eluates contained mIgG1, mIgG2a, or mIgG2b autoantibodies binding to mouse GBM NC1 hexamers (Fig. 5Ca–c). The presence of mIgG anti–GBM autoantibodies in kidney eluates was further evaluated by indirect IF staining. IgG staining the GBM of normal mouse kidneys was only present in kidney eluates from mice immunized with GBM NC1 hexamers (Fig. 5D). These results indicate that DBA/1 and D1.Fcgr2b^−/− mice do not develop anti–GBM autoantibodies after immunization with intact collagen IV from mouse GBM.

To establish whether the low immunogenicity of intact collagen IV from mouse GBM is due to host (immune tolerance) or Ag properties (supramolecular assembly, insolubility), we compared immune responses to this Ag in wild-type C57BL/6 mice and congenic Col4a3^−/− Alport mice. Circulating IgG Abs binding to mouse GBM NC1 hexamers were found in sera from Col4a3^−/− mice but not wild-type mice immunized with intact mouse GBM collagen IV (Fig. 6A, left). In contrast, mouse GBM NC1 hexamers elicited comparable Ab responses in both Col4a3^−/− and wild-type mice (Fig. 6A, right). Lack of IgG binding to mouse α1α2NC1 hexamers implies that anti–GBM Abs produced in Col4a3^−/− mice were specific for α345NC1 hexamers. This specificity was confirmed by indirect IF staining of kidneys from Col4a3^−/− mice expressing human COL4A3, showing IgG binding to the GBM (Fig. 6B). These results indicate that intact collagen IV from mouse GBM elicits Abs binding to the NC1 domains of α345(IV) collagen only in Col4a3^−/− mice but not in wild-type mice, implying that immune tolerance to intact α345(IV) collagen is acquired.

If the subclass restriction of autoantibodies specific for α345NC1 hexamers is solely determined by intrinsic properties of the Ag (e.g., its quaternary structure), it should not be influenced by the genetic background of the host. However, we found that Col4a3^−/− mice immunized with intact collagen IV (Fig. 7A) or NC1 hexamers from normal mouse GBM (Fig. 7B) produced not only mouse IgG1 but also IgG2a^c and IgG2b Abs specific for α345NC1 hexamers. This result implies that the production of proinflammatory IgG2a/c and IgG2b Abs against α345NC1 hexamers can occur in Alport mice lacking α345(IV) collagen, but it is selectively suppressed in mice expressing α345(IV) collagen.

The loss of mouse α3(IV) collagen in Col4a3^−/− mice can be compensated by transgenic expression of human α3(IV) collagen in COL4A3-humanized mice (23). This allows assembly of hybrid human/murine α345(IV) collagen, which is deposited in the GBM but not tubular basement membranes of transgenic mice. In COL4A3-humanized mice, NC1 hexamers from normal mouse GBM (a heterologous Ag in this setting) induced autoantibodies specific for α345NC1 hexamers (Fig. 8A), comprising mouse IgG1, IgG2a, and IgG2b subclasses (Fig. 8B). COL4A3 humanized mice immunized with mouse GBM NC1 hexamers developed significant albuminuria (Fig. 8C). Kidney histology showed crescentic GN and glomerular deposition of mouse IgG in a linear GBM pattern (Fig. 8D). This indicates that nephritogenic IgG subclasses of anti–GBM autoantibody can be induced by immunizing mice with heterologous NC1 hexamers.

Anti–GBM Ab GN, including Goodpasture disease, is clinically important because of its fulminant course, causing significant morbidity and mortality if not promptly treated. However, some patients with anti–GBM Abs develop atypically mild GN. The molecular mechanisms underlying this unusual presentation are not known. In this study, we identified a novel type of anti–GBM autoantibodies associated with mild nonprogressive GN, which were restricted to human IgG4 subclass and selectively targeted α345NC1 hexamers but not α3NC1 monomers. These autoantibodies targeted epitopes dependent on the native quaternary structure of α345NC1 hexamers, as hexamer dissociation abolished binding. In contrast, reactivity with α3NC1 monomers is a hallmark feature of Goodpasture autoantibodies, which preferentially bind to α3NC1 subunits of dissociated α345NC1 hexamers (3). Thus, the novel autoantibodies more closely resemble anti–GBM alloantibodies from patients with APTN, which bind to α3NC1 or α5NC1 epitopes accessible in native α345NC1 hexamers (9, 10), and may also target quaternary α345NC1 epitopes (28).

The novel anti–GBM autoantibodies to α345NC1 hexamers were restricted to IgG4 subclass, which best explains the atypically mild kidney disease. Human IgG4 are noninflammatory because they do not activate the classic complement pathway, do not bind IgG Fc receptors, and are functionally monovalent owing to an Fab arm exchange (29). Other patients with hIgG4-restricted anti–GBM autoantibodies have normal kidney function or mild GN (16, 18, 30). Sera from healthy people contain natural autoantibodies to α3NC1 monomers, restricted to noninflammatory IgG2 and IgG4 subclasses (31). In contrast, IgG1 autoantibodies are prevalent in patients with anti–GBM/Goodpasture disease, suggesting that proinflammatory IgG subclasses are required for development of severe GN (32–34).

We found that mice immunized with autologous NC1 hexamers from mouse GBM developed autoantibodies specific for α345NC1 hexamers, restricted to mIgG1 subclass, which did not cause GN despite binding to the GBM in vivo. Mouse IgG1 is functionally equivalent to human IgG4 because it is associated with a Th2 type (or an IL-4–secreting T follicular helper type) response and does not activate the classic complement pathway. Because mIgG1 has comparatively high affinity for the inhibitory FcγRIIB receptor and it only binds the activating IgG receptor FcγRIII but not FcγRIV, it is less inflammatory than mIgG2a or mIgG2b (35). However, mIgG1 autoantibodies can be pathogenic when inflammation is FcγRIII dependent but does not require FcγRIV. For instance, mIgG1 mAbs to collagen II induce arthritis in DBA/1 mice, which is abolished in FcγRIII^−/− mice and enhanced in FcγRIIB^−/− mice (36). In our study, the binding of mIgG1 autoantibodies to the GBM was not sufficient to cause murine GN, as found in other models (21, 37, 38). Thus, we established a mouse model recapitulating the specificity and subclass restriction of the novel human anti–GBM autoantibodies and also the mild kidney phenotype.

A novel finding in our study was that Fcgr2b^−/− mice were also resistant to GN induced by mIgG1-restricted anti–GBM autoantibodies. That ligation of FcγRIII by mIgG1 was insufficient to cause murine GN even when inhibitory signals from FcγRIIB were absent implies a critical role for FcγRIV. This is consistent with findings of another study, which showed that FcγRIII and FcγRIV are both required for induction of GN by mouse IgG2a or IgG2b Abs planted in the GBM (37). Besides activating Fcγ receptors, mIgG2a and mIgG2b may also promote the development of murine GN by activating complement (39). Although the ablation of FcγRIIB is critical for the development of murine GN after passive transfer of human anti–GBM Abs (21) or immunization with pepsin-solubilized collagen IV from bovine lens capsule (40), pathology in these models is likely mediated by mIgG2a and/or mIgG2b Abs. Indeed, production of mIgG2a and mIgG2b anti–GBM autoantibodies in COL4A3-humanized mice immunized with mouse GBM NC1 hexamers was associated with Ab-mediated GN. Overall, our results suggest that proinflammatory IgG subclasses of anti–GBM autoantibodies are necessary for the development of severe GN.

The pattern of IgG subclasses of anti–α345NC1 autoantibodies (hIgG4 in the patient and mIgG1 in NC1 hexamer–immunized mice) implies that IL-4–secreting T cell help was involved. In renal disease, this has been attributed to an IL-4 Th2 type response (41), with class switching to IgG2a (in mice) being IFN-γ (Th1 or T follicular helper) mediated (42). Because restriction to mouse IgG1 subclass was observed when murine GBM NC1 hexamers were used as an autologous Ag (in wild-type mice) but not as a heterologous Ag (in Alport mice or COL4A3-humanized mice), the production of mIgG2a and mIgG2b autoantibodies against α345NC1 hexamers appears to be effectively suppressed in mice expressing α345(IV) collagen. This likely reflects the activity of a subset of regulatory T cells inhibiting Th1/IFN-γ–directed responses, although direct suppression of B cells is also possible (43).

Because autoantibodies elicited by α345NC1 hexamers bind to the GBM, they must cross-react with epitopes with native α345(IV) collagen, implying that autoreactive B cells specific for self α345(IV) collagen are not purged from the repertoire of wild-type mice. Nonetheless, tolerance toward intact α345(IV) collagen is established in mice expressing this Ag, because intact collagen IV from murine GBM was not immunogenic in wild-type mice (even though it elicited anti–α345NC1 Abs in Col4a3^−/− Alport mice). Whereas autoreactive B cells are clearly critical for the production of anti–GBM Ab, T cells are required to provide B cell help (44). Thus, tolerance to intact α345(IV) collagen is likely maintained in the T cell compartment. This tolerance is breached by proteolysis generating an autoimmunogenic fragment, the α345NC1 hexamer.

Proteolysis has been implicated in generation of other autoantigens, particularly in extracellular matrices. A fragment of rat α2(I) collagen generated by MMP-1 cleavage, but not intact rat α2(I) collagen, induces experimental autoimmune anterior uveitis in rats (45). Neoepitopes generated by proteolysis are often targets for pathogenic autoantibodies. Examples include desmoglein-1 and desmoglein-3 neoepitopes implicated in pemphigus foliaceus and pemphigus vulgaris (46), as well as ectodomain shedding generating neoepitopes on collagen XVII, the major autoantigen for bullous pemphigoid (47). Thus, proteolysis is emerging as a general mechanism breaching immune tolerance, along with other types of posttranslational modifications (48).

First, even though the immunogen used in our study comprised a mixture of α12NC1 and α345NC1 hexamers, mice did not produce autoantibodies to α12NC1 hexamers. The robust tolerance toward self α12(IV) collagen may be due to its ubiquitous presence in all basement membranes, including in primary lymphoid organs where autoreactive lymphocytes are negatively selected. Second, the breach of tolerance was limited to a subset of autoepitopes dependent on the native quaternary structure of α345NC1 hexamers. Goodpasture-like autoantibodies binding to α3NC1 monomers were not produced by immunization with native GBM NC1 hexamers. Goodpasture autoantibodies are thought to be elicited by structural alterations or conformational changes of α345NC1 hexamers that unmask cryptic α3NC1 epitopes (6, 8, 21, 49). In mice, this can be induced by hexamer dissociation (10) or oxidation (50). Autologous rat GBM NC1 hexamers are autoimmunogenic in both LEW and WKY rats, but they induce anti–α3NC1 autoantibodies and GN only in WKY rats (51, 52), suggesting that the host genetic background also affects the production of nephritogenic anti–α3NC1 autoantibodies.

Unlike wild-type mice, Col4a3^−/− mice immunized with intact collagen IV from normal mouse GBM readily developed IgG Abs against α345NC1 hexamers, including proinflammatory IgG subclasses. Because production of class-switched IgG Abs requires help from cognate CD4⁺ T cells, tolerance toward α345(IV) collagen in Alport syndrome must be absent in both B cell and T cell compartments. Despite immunosuppression, Alport patients exposed to allogeneic α345(IV) collagen in the kidney allograft are prone to develop nephritogenic anti–GBM alloantibodies and posttransplant nephritis because they do not have tolerance toward α345(IV) collagen or mechanisms suppressing the production of proinflammatory IgG subclasses of anti–α345NC1 Abs.

We thank Selene Colon for technical support and Dr. Elena Tchekneva for access to the mouse urine collection station (53) described in the patent application available at http://www.freepatentsonline.com/20110239953.pdf.

The authors have no financial conflicts of interest.