Patients and rodents with Goodpasture’s syndrome (GPS) develop severe autoimmune crescentic glomerulonephritis, kidney failure, and lung hemorrhage due to binding of pathogenic autoantibodies to the NC1 domain of the α3 chain of type IV collagen. Target epitopes are cryptic, normally hidden from circulating Abs by protein-protein interactions and the highly tissue-restricted expression of the α3(IV) collagen chain. Based on this limited Ag exposure, it has been suggested that target epitopes are not available as B cell tolerogens. To determine how pathogenic anti-GPS autoantibody responses are regulated, we generated an Ig transgenic (Tg) mouse model that expresses an Ig that binds α3(IV)NC1 collagen epitopes recognized by serum IgG of patients with GPS. Phenotypic analysis reveals B cell depletion and L chain editing in Tg mice. To determine the default tolerance phenotype in the absence of receptor editing and endogenous lymphocyte populations, we crossed Tg mice two generations with mice deficient in Rag. Resulting Tg Rag-deficient mice have central B cell deletion. Thus, development of Tg anti-α3(IV)NC1 collagen B cells is halted in the bone marrow, at which point the cells are deleted unless rescued by a Rag enzyme-dependent process, such as editing. The central tolerance phenotype implies that tolerizing self-Ag is expressed in bone marrow.
Patients and rodents with Goodpasture syndrome (GPS)3 develop rapidly progressive renal failure due to necrotizing crescentic glomerulonephritis and potentially fatal pulmonary hemorrhage due to alveolitis. In a subset of patients injury is restricted to the kidneys, referred to as anti-glomerular basement membrane (GBM) nephritis. The pathologic and diagnostic hallmarks are the presence of circulating anti-GBM IgG autoantibodies and linear deposition of IgG along the glomerular and alveolar capillary basement membranes. The mainstay of therapy is an aggressive regimen of corticosteroids, cyclophosphamide, and plasmapheresis to broadly suppress the immune system. Disease-specific mechanism-based and less toxic interventions are desirable, the rational design of which requires better understanding of the origins and regulation of anti-GBM autoreactivity and of the genetic and environmental factors that subvert those mechanisms in disease.
Pathogenic autoantibodies from GPS patients bind the C-terminal globular noncollagenous domain 1 (NC1) of the α3-chain of type IV collagen, α3(IV)NC1 (1, 2). Similar epitopes are targeted by alloantibodies in a subset of patients with Alport nephritis who receive renal allografts (1, 2, 3). The native kidneys of Alport patients lack the collagen epitopes targeted in GPS due to mutations in one of the genes encoding the α-chains (α3, α4, or α5) of type IV collagen, such that introduction of the epitopes in the transplanted kidney leads to an antiforeign (antiallotypic) immune response. The sites of organ injury are in a large part explained by the highly tissue-restricted expression of the α3(IV) collagen chain, which is limited to basement membranes of the glomerulus, renal tubules, alveolus, cochlea, anterior lens capsule, Descemet’s membrane, ovary, and testis (4). The exact mechanism by which GPS Ig deposits in lung and kidney remains under investigation. The epitopes recognized by pathogenic Abs are conformational and normally buried within the native NC1 hexamer, such that patient sera IgG preferentially binds to dissociated, as compared with native, Ag (5, 6, 7, 8). It is proposed that in diseased individuals Ig binding is facilitated by limited NC1 hexamer cross-linking and environmental toxins that promote in vivo exposure of normally cryptic epitopes (9, 10, 11).
It is currently unclear whether potential tolerizing epitopes are accessible to developing and circulating lymphocytes in healthy individuals, or whether epitope exposure is essential for lymphocyte activation and disease initiation in patients. Low levels of anti-α3(IV)NC1 IgG are reported in normal serum (12), and CD4+ and CD8+ T cells reactive with α3(IV)NC1 can be isolated from healthy individuals as well as patients (13, 14, 15), the latter despite the presence of Ag in the thymus (15, 16). These findings have been interpreted to support a model wherein T and B cell epitopes are cryptic, such that anti-α3(IV)NC1 B and T cells are not tolerized in vivo. Disease is presumed to result from activation of “ignorant” lymphocytes by an unknown stimulant coupled with Ag exposure and end organ susceptibility.
To determine whether and how Goodpasture-like anti-collagen B cells are regulated in vivo and how they escape regulation to cause disease, we developed a novel anti-α3(IV)NC1 collagen Ig transgenic (Tg) model. The Ig Tg model is useful to study B cell tolerance because it greatly increases the frequency of B cells of the defined specificity and permits determination of their fate in vivo. Our strategy involved identification of a murine anti-α3(IV)NC1 mAb that bound a pathogenic epitope, cloning of the Ab H and L chain genes and regulatory sequences, generation of Ig H+L Tg mice, and evaluation of the resulting B cell phenotype. The Tg BCR in this model targets a collagen epitope recognized by pathogenic IgG in the sera of GPS patients. Analysis of the fate of Tg B cells indicates that collagen-reactive cells are tolerized and not immunologically ignorant. Although autoantibodies are present at low levels in serum and recoverable by fusion of Tg mouse spleens, Tg mice have significantly fewer splenic B cells than non-Tg littermates and have evidence of L chain editing. Elimination of endogenous (non-Tg) Ab and T cells by establishing Tg mice deficient in Rag enzyme leads to near-complete Ig and B cell depletion that is evident in bone marrow as well as spleen. Collectively these findings indicate that anti-α3(IV)NC1 B cells are regulated in vivo, suggesting that the reactive epitopes are exposed and that either α3(IV)NC1 collagen or a cross-reactive self-Ag is tolerogenic. The central tolerance in Rag-deficient Ig Tg mice further indicates that tolerogen is encountered in bone marrow.
Materials and Methods
Immunization and generation of hybridomas
Immunization of SJL and C57BL/6 (B6) mice with 10 μg of recombinant human α3(IV)NC1 Ag (17) in Freund’s adjuvant, followed by 2–8 boosts with recombinant human or purified bovine Ag (Wieslab) in IFA, and generation of anti-α3(IV) NC1 hybridomas by fusion of unmanipulated splenocytes with murine Sp2 myeloma cells are described elsewhere (18). Hybridoma culture supernatants were screened for secreted Ig, IgM allotype, and binding to purified α3(IV)NC1 collagen by solid-phase ELISA as described below. Described below is selection of a prototypic Goodpasture mAb based on inhibition of Ag binding by multiple human Goodpasture patients’ sera IgG. Hybridomas producing anti-α3(IV)NC1 collagen Ab and hybridomas subsequently generated from Tg mice and producing Tg H chain IgMa allotype were subcloned either by limiting dilution or by cell sorting at the Duke University flow cytometry facility.
Production of constructs containing the anti-α3(IV)NC1 mAb VDJ or VkJk cDNAs and regulatory sequences
To identify the mAb-rearranged variable region sequences of our prototypic anti-α3(IV)NC1 collagen Ig, cDNA was generated from hybridoma RNA and rearranged VDJ and VkJk genes amplified by PCR using promiscuous upstream V gene primers and downstream IgM and κ constant region primers, as described (19). Complete H and L chain variable region sequences are described in detail elsewhere (18).
To produce a DNA construct encoding the anti-α3(IV)NC1 H chain V region with appropriate tissue-specific expression of secreted and transmembrane IgM, we took advantage of existing genomic DNA fragments containing Ig regulatory elements and signal sequences (20) and a strategy previously described in our laboratory to generate the 238H H chain Tg (21), with minor modifications. First, the mAb VDJ was amplified from cDNA using forward and reverse primers based on mouse germline J558.3 (VH) and JH2 sequences (GenBank accessions AF303834 and X63167) and designed to insert unique SfuI (AsuII) and SauI (Bsu36I) restriction sites at the signal peptide-VH junction and within JH2, respectively: 5′VH-SFU (5′-CGT TCG AAG TCC AGC TGC AAC AGT CTG GAC CTG-3′) and 3′JH2-SAU (5′-TCA CCT GAG GAG ACT GTG AGA GTG GTG CC-3′). We next used a 3-kb construct containing 5′ regulatory sequences (700 bp), 238H VDJ (320 bp), and the JH germline genes with Igμ enhancer (2 kb), previously generated in our laboratory and carried in plasmid Bluescript (Stratagene; Ref. 21). The 238H VDJ was replaced with the anti-α3(IV)NC1 VDJ by a two-step cloning method, taking advantage of the 5′ Sfu I site to link mAb VDJ in frame with 5′ regulatory sequences, and then using SauI digestion to link the 1-kb promoter-VDJ fragment to a 5-kb fragment containing JH genes and Cμ enhancer in plasmid Bluescript. A 3-kb NotI/EcoRI fragment from this vector was ligated to a 12-kb EcoRI/NotI fragment containing the 9-kb IgMa constant region (gifted by D. Nemazee, The Scripps Research Institute, La Jolla, CA) to generate the final 15-kb VDJ-Cμ construct. This construct was linearized with NotI and purified for transfection experiments.
The L chain construct was generated using PCR of hybridoma cell DNA to amplify a 2.2-kb genomic fragment containing mAb VkJk, 5′ regulatory sequences, and germline Jκ genes. The forward primer, 5′-AAAAAATTGTATTTAAGAAGGGTCCTTTGA-3′, lies 567 nucleotides upstream of the Vk gene and includes the regulatory elements. This primer was designed based on the mouse germline Vk sequence most closely matching the anti-α3(IV)NC1 mAb L chain sequence, using mouse genomic BLAST (http://www.ncbi.nlm.nih.gov/genome/seq/MmBlast.html). The reverse primer, 5′-ATAGTCGACAGACCACGCTACCTGCAGTCAGAC-3′, lies 498 bp downstream of Jk5 (22) and includes a unique SalI restriction site (underlined). This 2.3-kb gene fragment was then ligated to a 4.5-kb genomic fragment containing the κ enhancer and Cκ gene, provided by Dr. K. Rajewsky, Harvard University, Boston, MA (22), via Dr. T. Imanishi-Kari, Tufts-New England Medical Center, Boston, MA.
Ex vivo expression of the anti-α3(IV)NC1 H and L chains
H chain-loss variant myeloma (J558L) cells carrying the λ1 L chain were cotransfected with the coselectable marker, pSV2-neo, encoding aminoglycoside phosphotransferase and with either NotI-linearized H chain construct or with both linearized H and L chain constructs, using electroporation with a Gene Pulsar apparatus (Bio-Rad) in HT medium (84% RPMI 1640, 10% FBS; HyClone, Thermo Fisher Scientific) and 1× nonessential amino acids, hypoxanthine-thymidine, sodium pyruvate, HEPES buffer solution, l-glutamine, penicillin-streptomycin, respectively; all reagents were obtained from Invitrogen unless otherwise noted. Transfected cells were selected in medium containing 400 μg/ml aminoglycoside G-418 (Invitrogen), screened for secreted IgM,κ Ig by ELISA, and subcloned by limiting dilution.
Tg mice carrying both the H and L chain constructs were generated via standard techniques by the Duke Comprehensive Cancer Center Transgenic Mouse Facility, Duke University. Founders were generated in hybrid strain B6SJLF1/J obtained from The Jackson Laboratory. Offspring were genotyped by PCR using transgene-specific forward and reverse primers and Tg lines established on the B6 background. The experiments described here were conducted on mice of either sex hemizygous for the introduced transgenes and reared under conventional specific pathogen-free conditions. B6 breeders, BALB/c (IgMa allotype) controls, and mice carrying a targeted mutation in Rag1 on the B6 background (RAG-KO) were obtained from The Jackson Laboratory. The care and use of all experimental animals were in accordance with institutional guidelines, and all studies and procedures were approved by the Animal Care and Use Committees of Duke University and the Durham Veterans Affairs Medical Center.
Cell and tissue staining
Ab isotype, allotype, and binding specificity
Ig concentrations and isotype- and allotype-specific binding in serum and in culture supernatants from B cells, transfectants, and hybridomas were determined by ELISA as described (23). Collagen binding activity was determined by ELISA. In brief, Immulon II plates were coated overnight at 4°C with bovine α3(IV)NC1 collagen (Wieslab) diluted in 6M guanidine-HCl. After blocking with 3% BSA in PBS, serum or supernatant was incubated at 50 μl per well at room temperature for 1 h, washed once with PBS/0.05% Tween and twice with PBS, followed by alkaline phosphatase-conjugated anti-mouse IgM or IgG (Boehringer Mannheim) diluted in PBS/0.1% BSA. Binding was detected using p-nitrophenyl phosphate substrate and assays were monitored at 405 nm. Control Ig included anti-α3(IV)NC1 IgG mAb (Wieslab) and transfectant IgM. Results were recorded as the mean sample OD after subtraction of mean OD on diluent-coated plates and subtraction of OD blank (diluent without Ig). Ag binding by transgene-encoded Ig was confirmed using biotin-labeled anti-allotype (IgM-a)-specific second-step reagents (BD Biosciences) detected with avidin-alkaline phosphatase (Southern Biotechnology Associates). For assays of the inhibition of Ag binding by competitive ELISA, the mAb concentration (5 microg/ml) or Tg mouse serum dilution that gave 50% maximal binding to α3(IV)NC1 collagen was determined by direct-binding ELISA. This dilution was then preincubated with varying concentrations of inhibitor (Goodpasture patient or normal human serum) for 1 h at 37°C before incubation with Ag-coated wells. Bound mAb or serum Tg-encoded IgMa were detected with 1/1000 dilution alkaline phosphatase-labeled goat anti-mouse Ig (Southern Biotechnology Associates) or 1/500 dilution biotin-labeled anti-allotype (IgM-a)-Ig (BD Biosciences), followed by avidin-alkaline phosphatase and second-step reagents, respectively. Archived coded Goodpasture patient sera were provided by Dr. D. Howell, Duke University Medical Center and Durham Veterans Affairs Medical Center (Durham, NC) and purchased from Wieslab.
L chain analysis from hybridomas or splenic cDNA
Rearranged endogenous VK-Jk genes were identified by sequence analysis of PCR-amplified cDNA, using a 3′ primer complementary to Cκ and 5′ primers designed to complement disparate Vκ families, essentially as described (24). Sequence analysis, alignment, and assignment of V families were performed using ClustalW (www.ebi.ac.uk/clustalw), IgBLAST (www.ncbi.nlm.nih.gov), and the international ImMunoGeneTics database (imgt.cines.fr; M.-P. Lefranc, Montpellier, France, initiator and coordinator).
All data are shown as median values and interquartile range (25th and 75th percentile) unless otherwise indicated. Comparisons between two groups were analyzed with Mann-Whitney U test and between three or more groups with the Kruskal-Wallis one-way ANOVA. Analyses were performed with Analyze-it software. A value of p < 0.05 was considered to be significant.
Identification of α3(IV)NC1 collagen mAb
Six SJL or B6 mice were immunized with human recombinant or bovine GBM α3(IV)NC1 collagen in Freund’s adjuvant. Mice developed high titer serum α3(IV) NC1-reactive IgG and linear renal basement membrane IgG deposits (Fig. 1,A). Splenic fusions from six immunized mice yielded five anti-α3(IV)NC1 Ig mAb as determined by ELISA. One B6-derived IgM, termed 1G6, met both criteria for reactivity with pathogenic GPS epitope(s); it bound strongly in ELISA to α3(IV)NC1 collagen coated in 6 M guanidine-HCl, which dissociates NC1 hexamers into monomers and dimers (25), but not to diluent coated wells, and mAb 1G6 binding to α3(IV)NC1 collagen was inhibited by serum IgG from multiple patients with GPS (Fig. 2). This mAb was subsequently designated the prototypic Goodpasture mAb.
Production of functional anti-NC1 H and L chain Ig DNA constructs and anti-NC1 Ig Tg mice
A mAb IgM H chain construct (Fig. 3,A) was generated by ligation of the rearranged VDJ gene, captured by PCR from hybridoma DNA of the prototypic mAb to existing genomic fragments containing regulatory elements. This construct was then ligated to the 9-kb IgMa constant region, as previously described for the 238H H chain Tg (21), with minor modifications. The L chain construct was generated using PCR of hybridoma genomic DNA to isolate a 2.2-kb fragment containing the 5′ promoter region, VkJk, and downstream elements including Jk5 (Fig. 3 B). The 2.2-kb gene segment was ligated to a 4.5-kb fragment containing the κ enhancer and Cκ gene. The integrity and successful expression of both constructs was confirmed by their cotransfection into λ L chain-expressing murine myeloma cells. Culture supernatants from cells transfected with the Ig H and L chain constructs, but not supernatants from mock transfected cells, contained IgM,κ Ig with anti-α3(IV)NC1 specificity (not shown).
Phenotypic characterization of mice expressing the anti-NC1 Ig H and L chain Tg
Ig Tg mice were established by coinjection of the H and L chain constructs into B6SJLF2 fertilized eggs at the Duke University Transgenic Mouse Core Facility and are maintained as hemizygotes. Four founders carrying both genes were identified by PCR genotyping and crossed with B6 breeders. Genotype analysis of back-cross progeny from these four lines is consistent with cosegregation of the Tg H and L chain constructs in each line.
Flow cytometric analysis of splenocytes and bone marrow lymphocytes confirmed in vivo expression of Tg IgM a-allotype on the B cell surface (Fig. 4, A and C) from progeny of each founder line. Tg expression was restricted to B220+ cells (B lymphocytes). Tg and non-Tg mice from each line were included in the analyses (Table I).
B cell numbers were decreased ∼4-fold in spleens of Tg progeny as compared with non-Tg littermates: 9.4 (6.9, 12.0) vs 37 (32.3, 42.3) million, Tg+ vs non-Tg littermates, p < 0.0001; Table I. In the bone marrow the frequency of IgM+ B220 cells was lower in Tg compared with non-Tg mice: 45.9 (43.4, 47.7) vs 52.8 (47.0, 54.5) percent, Tg vs non-Tg, p < 0.05. Collectively, these findings suggest that a large population of B cells were deleted in Tg mice.
Among residual splenic B cells L chain editing was common. λ L chains were expressed on the cell surface of up to 11% (median 8.8% (4.1, 10.9), n = 9) of splenic B cells in Tg mice as determined by flow cytometry, despite the presence of the productively rearranged κ-chain Tg. This exceeded surface λ expression in non-Tg littermates (median 5.4% (4.3, 5.8), p < 0.05). Whereas flow cytometry did not distinguish Tg and non-Tg (endogenous) κ-chains, κ-chain editing was revealed by two alternative approaches. Initial query of Tg mouse spleen cDNA by PCR using an upstream primer to detect the Tg L chain sequence (18) and a downstream Cκ primer revealed not only Tg L chain message, but also two κ-chains derived from productive secondary rearrangements at the endogenous loci and encoded by IGKV3–4*01-Jk1 and IGKV3–2*01-Jk2 genes. One of these two L chains was identified in spleen cDNA from each of four Tg mice, derived from three different Tg lineages. To further explore the extent of κ editing in this model, rearranged κ L chains were recovered from a cDNA library generated from B220+IgMa+ B cells sorted by flow cytometry from splenocytes of a fifth young H+L Tg mouse. PCR amplification used a universal 3′ Cκ primer and a panel of three upstream primers derived from disparate Vκ families, as previously described (24). Amplification products were cloned and three clones from each primer pair selected at random for sequencing. Three rearranged endogenous L chain sequences were recovered, each using a different Vκ family and different Jκ gene: IGKV1–133*01-Jκ4, IGKV19–120*01-Jκ5, and IGKV19–93*01-Jκ1. These findings confirm that L chain receptor editing contributes to regulation of B cells autoreactive with α3(IV)NC1.
Sera of non-Tg littermates contained only IgMb allotype, as expected in the B6 strain. IgMb was also the dominant allotype in serum of Tg mice, despite the paucity of IgMb+ B cells in bone marrow and peripheral lymphoid organs. Tg mice had limited spontaneous secretion of transgene-encoded IgMa. Serum total IgMa levels (range 0.64–22.06 μg/ml, median 9.18 μg/ml, n = 10, excluding one outlier with level 149.6 μg/ml) were lower than in concurrently measured healthy normal BALB/c (IgMa allotype) mouse serum (114.8 μg/ml). These levels are also ∼5- to 30-fold lower than reported in the non-self-reactive nontolerant anti-lysozyme Ig Tg model, but similar to levels in the tolerant anti-lysozyme Ig and neo-self-Ag double Tg model in which Tg B cells are regulated by anergy (range of serum Tg Ig ∼0.5–5.0 μg/ml; Ref. 26).
A subset of serum Tg Abs were autoreactive. ELISA revealed IgMa binding to α3(IV) NC1 Ag in the serum of most Tg mice (Fig. 5,A). Competition ELISA using GPS patient sera confirmed binding to pathogenic epitopes by a subset of Tg serum IgMa (Fig. 5,B). Immunohistochemical analysis of kidneys showed limited glomerular deposition of IgM in 4 of 10 Tg mice (Fig. 1, C–D). Glomerular scores for Ig fluorescence intensity did not correlate with serum Ag binding or serum IgMa levels (not shown).
T cell numbers were also decreased ∼33% in spleens of Tg progeny as compared with non-Tg littermates: 17.5 (11.5, 19.1) vs 27.1 (20.2, 28.4) million, Tg+ vs non-Tg littermates, p < 0.01; Table I. Both CD4+ and CD8+ T cell subsets were significantly decreased (not shown).
Characterization of mAb derived from H+L Tg mice
Autoreactive Tg Ig were also recovered as monoclonal Ig by fusion of both unmanipulated and in vitro LPS-stimulated Tg mouse splenocytes. Among 136 hybridoma recovered by fusion, 103 supernatants expressed Tg IgMa. Supernatants of the 25 clones with highest OD for IgMa were subsequently screened for binding to α3(IV) NC1 Ag: 18/25 (72%) of hybridoma, including 69% (11/16) of those derived from unmanipulated splenocytes and 78% (7/9) of those derived from endotoxin-stimulated splenocytes, bound Ag. PCR analysis of cDNA generated from mRNA of two subcloned anti-α3(IV)NC1 collagen Tg hybridoma revealed an unmutated Tg-encoded κ L chain transcript, confirming in vivo expression of the Ig Tg L chain.
Anti-NC1 Ig H+L chain Tg with superimposed Rag deficiency
To further dissect the contributions of editing to the survival of Tg B cells, we crossed H+L Tg mice with mice deficient in the Rag enzyme. Rag-1 and Rag-2 are required for variable region gene rearrangements that generate productive Ig and T cell receptors (27). Loss of either enzyme thus eliminates influences from endogenous (non-Tg) Ig, including receptor editing or revision, allelic inclusion, and follicular competition, as well as influences of immunoregulatory T cells (28, 29, 30, 31, 32, 33, 34, 35, 36, 37). Ig Tgs are expressed on the Rag-deficient background because the prerearranged VDJ and VJ genes bypass the requirement for Rag (38, 39).
Resultant triple-mutant H+L Tg Rag-deficient mice had small spleens with few lymphocytes. Flow cytometric analysis detected no IgM-expressing B220+ B lymphocytes in either the spleen or bone marrow of Tg Rag-deficient mice (Fig. 4, B and D and Fig. 6,B), indicating complete central deletion. Spleen size and flow cytometric plots of splenocytes from Tg Rag-deficient and non-Tg Rag-deficient littermates were essentially identical. The spleens of both Tg and non-Tg Rag-deficient mice contain a small population of B220lowIgM−C19− cells, consistent with previous reports in which negative CD19 staining suggests that these are not B lineage cells (38, 39, 40). Consistent with these findings, ELISA analysis revealed no detectable Ig in serum of wild-type Rag-deficient mice and only trace quantities of Tg IgM/κ Ig in serum of young H+L Tg Rag-deficient mice (Fig. 6). Residual serum IgM recovered from Tg Rag-deficient mice reacted with α3(IV)NC1 collagen in ELISA (not shown).
To better understand the immunological events that precipitate organ destruction in patients with Goodpasture syndrome, we explored mechanisms regulating autoreactive B cells that bind the Goodpasture Ag. For this purpose we developed a novel H+L chain Ig Tg mouse model expressing an anti-α3(IV)NC1 collagen Ig that binds epitopes recognized by human patient serum IgG. When the H and L chain transgenes are coexpressed in vivo, transgene-expressing B cells are readily recovered from the spleen. However, peripheral total B cell numbers are significantly decreased in Tg mice despite L chain editing. Genetic introduction of Rag enzyme deficiency to preclude Ig editing and remove influences from endogenous lymphocyte populations results in elimination of Tg B cells. Thus development of Tg anti-α3(IV)NC1 collagen B cells is halted in the bone marrow, where cells are deleted unless rescued by a Rag-dependent process.
The central tolerance phenotype indicates that tolerizing self-Ag is expressed in the bone marrow. The nature of this putative tolerogen is unclear. The α3(IV) collagen chain targeted by GPS patients’ autoantibodies has a restricted tissue distribution, being synthesized primarily in specialized basement membranes of glomerulus, renal tubules, alveoli, cochlea, eye, and testis (4). Moreover, pathogenic epitopes are conformational and normally buried in the native α3(IV) NC1 hexamer, partially hidden from circulating Ab and B cells (5, 6, 7, 8). Epitope exposure, possibly facilitated by environmental factors and dynamic interactions of autoantibodies and collagen protein (9), is presumed key to disease onset. Thus it has been predicted that anti-α3(IV)NC1 collagen B cells are not tolerized in vivo under normal circumstances, but rather exist ignorant of Ag. Our finding of deletion and editing challenges this view, and rather suggests the presence of a potent tolerogen in the bone marrow. One candidate is tumstatin, an ∼232 amino acid, 30-kDa circulating form of α3(IV)NC1 cleaved from the carboxyl terminus of the α3(IV) collagen chain by the action of metalloproteinases (41). Fragments of collagen IV are reported in sera of healthy individuals, and Kalluri and colleagues detected circulating tumstatin in plasma of normal mice at a mean concentration of 336 ng/ml using a direct ELISA with anti-tumstatin Ab (4, 42). Alternatively, a cross-reactive epitope on an as yet unidentified endogenous protein may tolerize anti-α3(IV)NC1 collagen B cells. Our results raise the possibility of a therapeutic immunoregulatory role for tumstatin or related soluble Ag.
Despite a default tolerance phenotype of profound central deletion, Tg B cells are present in substantial numbers in the periphery of Tg Rag-sufficient mice. This Rag-dependent escape is due in part to L chain editing, in which persistent activation of Rag enzymes catalyzes secondary rearrangements to replace existing productive ones (28, 29, 30, 40, 43, 44). Editing to an innocuous specificity can be an effective mechanism to abrogate autoreactivity. Direct evidence of editing in the anti-α3(IV)NC1 collagen B cells includes expression of λ L chains in cells that already carry the rearranged κ Tg and PCR-based recovery from Tg spleens of transcripts from productively rearranged endogenous κ-chains. It is notable, however, that splenic B cell numbers in Rag-sufficient anti-α3(IV)NC1 collagen Tg mice are only one-fourth to one-fifth the number in wild-type littermates. Although editing is a common phenomenon, accounting for ∼25% of splenic B cells in healthy individuals (45), it is also highly efficient. Editing typically minimizes deletional cell loss such that normal or near normal numbers of splenic B cells are maintained within a diverse polyclonal B cell population (40).
Persistent B cell depletion in the anti-α3(IV)NC1 collagen Tg model may be attributed in part to accelerated B cell development, imparted by the transgene, that limits the window for serial editing (46). The duration of the editing-susceptible period may be crucial to its effectiveness as a salvage mechanism. Its prolongation provides additional opportunity for secondary Ig gene rearrangements. In this regard, tolerance-induced slowing of immature B cell turnover and enforced bcl2 expression are credited with rescue of wild-type B cells reactive with a ubiquitous membrane Ag (47). In the anti-α3(IV)NC1 collagen Ig Tg model, the capacity of editing to restore normal B cell numbers may also be thwarted by Tg H chain dominance. If a single H chain generates autoreactivity with a variety of different L chains, many secondary rearrangements will fail to abolish autoreactivity. This possibility is supported by results of additional transfection experiments (not shown) in which introduction of the Tg H chain construct into a Vk8Jk5 L chain-only hybridoma cell line (gifted by Dr. M. Radic; Ref. 48) produced a novel IgM,κ Ig that bound α3(IV)NC1 collagen in ELISA. Regardless of mechanism, in the anti-α3(IV)NC1 collagen H+L Tg model, neither editing nor proliferation and expansion of peripheral B cell pools is sufficient to replenish splenic populations to normal levels.
Rag-dependent survival of a subset of Tg anti-α3(IV)NC1 collagen B cells may depend on secondary rearrangements at the endogenous Ig H chain loci. Although endogenous IgMb H chain is rarely detected on the surface of peripheral B cells in neonatal and young anti-α3(IV)NC1 collagen Ig Tg mice (Fig. 4), dual positive IgMa+IgMb+ and isolated IgMb+ B cell populations emerge in a subset of older Tg mice (not shown). This coexpression of an endogenous H chain may promote survival of autoreactive cells in a manner analogous to L chain editing, as is proposed for high affinity anti-dsDNA B cells in the R4A-γ2b H chain Ig Tg model (49). In nonautoimmune mice productive rearrangements at both H chain alleles are detected in ∼2–4% of B cells (50), although recovery of dual H chain-producing B cells in this setting is rare (51). H chain inclusion may function physiologically as a salvage mechanism to ensure formation of a functional pre-BCR (50). The extent of dual H chain expression is generally higher, although variable, in IgM Tg or Ig gene-targeted models (52, 53, 54, 55, 56, 57), and may be detected at the level of the cell population or individual cells (32, 46, 58). The contribution of H chain inclusion to autoreactive cell survival and autoimmunity appears to be minimal for at least some specificities. In anti-phosphocholine Ig Tg mice, endogenous H chain is coexpressed by 20% of Tg B cells; however, back-cross with mice lacking the IgM transmembrane exons (muMT mice), which eliminates endogenous H chains, fails to alter splenic B cell numbers or block autoantibody production (32).
Rag deficiency may influence anti-α3(IV)NC1 collagen Tg B cell fates by additional mechanisms. However, none seems likely to account for profound central deletion. Loss of endogenous B cells removes competitor populations that would otherwise restrict autoreactive cell entry into follicles and germinal centers in secondary lymphoid organs (33, 34). The absence of T cells can influence autoreactive B cell fate and survival in the periphery by a variety of mechanisms (34, 59, 60), including a permissive role in B cell anergy that appears to be Ag or model specific (39, 61). However, a significant role has not been defined for T cells or circulating T cell-derived cytokines in the bone marrow-specialized microenvironments that support early B cell development, survival, selection, and emigration (62). Congenitally athymic (nude) mice are not autoimmune, despite subtle abnormalities of B cell development and primary V gene repertoire (63), and introduction of the nude mutation does not discernibly alter outcomes in anti-phosphocholine Ig Tg mice (32). Similar to our finding in the Rag-deficient anti-α3(IV)NC1 collagen Ig Tg, central deletion is efficient in mice lacking Rag and bearing an Ig Tg reactive with the ubiquitous cell membrane self-Ag MHC H-2k, whereas abundant B cells are present in Rag-deficient Ig Tg mice lacking this deleting self-Ag (38).
Our proposed model for regulation of Tg anti-α3(IV)NC1 collagen B cells, dependent on central selection and editing, must also explain the autoreactivity recovered from Tg mice. Most animals have low levels of serum anti-α3(IV)NC1 collagen autoantibodies, and autoreactivity is recovered from hybridomas derived from Tg mouse spleens. This autoreactivity may depend on the same editing process that permits the cells to escape deletion. Editing that occurs in trans on a second L chain allele, as occurs both physiologically and in H and L chain Ig Tg mice, permits expression of both L chain proteins in the same B cell (40). Dual L chain expression is reported in wild-type mice and healthy humans (64, 65). In mouse spleens, an estimated 10% of mature B cells express both κ alleles (66) and 20% of λ-producing cells coexpress a κ-chain (64). Dual L chain expression is reported in a significant fraction of peripheral B cells in mice bearing the 3H9H-56R anti-DNA or the 3–83 anti-MHC class I Ig Tg (67, 68). It is notable that young 3H9H/56R anti-DNA mice do not secrete anti-DNA Ab in vivo (67). Nonetheless, analogous to the anti-α3(IV)NC1 collagen Ig Tg reported here, 3H9H/56R anti-DNA Tg B cells captured as hybridomas secrete both λ-encoded autoantibody and nonautoreactive κ editor L chain (67). Similarly, Ig Tg mice engineered to coexpress two L chains spontaneously produce the autoreactive Tg Ig; lack of or escape from tolerance in this model is attributed to receptor dilution (32). In this scenario the innocuous (nonautoreactive, editor) receptor presumably promotes cell survival, activation, and differentiation, ultimately supporting secretion of both editor and autoantibody (44). The extent to which similar mechanisms permit autoreactivity in the anti-α3(IV)NC1 collagen Ig Tg mice may be dissected by genetic exclusion of endogenous L chain loci. Nonetheless, the presence of spontaneous anti-α3(IV)NC1 collagen autoreactivity in the Tg mice, and at low levels in healthy humans (12), invites speculation as to whether this specificity is preserved because it confers some immunological advantage to the host, as proposed for anti-phosphocholine autoantibodies (32).
The presence of trace quantities of Tg Ig in a few Tg Rag-deficient mice in the absence of detectable IgM+ B cells in the periphery further suggests that rare anti-α3(IV)NC1 collagen B cells, possibly anergic cells with below-threshold levels of surface Tg expression, do escape censoring and can be activated in a T-independent manner. Our studies do not indicate significant B1 lineage development in the peritoneum, although the significance of the rare peritoneal IgM+ B220+ cells in a few Tg Rag-deficient mice (data not shown) is unclear.
It is notable that the majority of serum IgM in Rag-sufficient Tg mice derives from the relatively few endogenous (non-Tg IgMb) B cells, despite domination of the peripheral B cell repertoire by Tg IgMa-bearing B cells. The role of homeostatic mechanisms in supporting normal serum IgM levels in the setting of B cell lymphopenia or in Ig Tg models expressing a tolerance phenotype is unclear. Adult mouse serum IgM normally derives approximately equally from infrequent B1 lineage cells and from Ag-recruited B2 lineage cells (69). The low serum Tg IgMa levels suggest that many Tg B cells either have not been activated in vivo or are anergic.
Rag-sufficient Tg mice also demonstrate an unexplained significant T cell depletion. It is possible, and perhaps likely, that the restricted diversity and size of the endogenous B cell population influences the T cell repertoire. It is now well appreciated that B cells have important Ab-independent regulatory roles that include Ag presentation for T cell activation or tolerance induction, coordination of T cell migration and differentiation, regulation of T cell polarization and effector function, regulation of dendritic cell functions, and production of immunomodulatory cytokines and chemokines (70). Depletion of regulatory T cells could promote activation of Tg-expressing B cells and contribute to serum autoantibody levels. Conversely, naive Tg anti-α3(IV)NC1 collagen B cells may engage collagen-reactive T cells to promote their deletion, although it seems unlikely that such interactions would significantly deplete a diverse T cell repertoire.
In conclusion, our findings invite speculation about the origin of pathogenic autoantibodies in GPS in humans. Results from the Tg model suggest that anti-α3(IV)NC1 collagen B cells must either escape central deletion, possibly through editing with dual receptor expression, or they must arise de novo from somatic events in mature B cells in the periphery, such as hypermutation and secondary Ig rearrangements (71) and escape regulation at these more distal points. Tight regulation at multiple stages may explain the relative rarity of this deadly autoimmune disease.
Dr. Raghu Kalluri (Harvard University) is gratefully acknowledged for provision of recombinant human α3(IV)NC1 and Dr. David Howell (Duke University Medical Center and Durham Veterans Affairs Medical Center) for providing patient Goodpasture serum. We thank the Duke University Comprehensive Cancer Center Hybridoma, Cell Culture, and DNA sequencing facilities and Dr. Carol Wikstrand for expertise. We thank Erica Ungerwitter, Joanna Bradley, Jacquie Anderson, and Russell Williams for technical assistance.
The authors have no financial conflict of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported in part by the National Institutes of Health (T32AI07217 to D.B.H.; R01DK47424 to M.H.F.), a Veterans Affairs Medical Center Merit Award (to M.H.F.), the Durham Veterans Affairs Medical Center Research Service and Institute for Medical Research, Duke University Undergraduate Research Support Grants (to S.C.S.), and a grant from the American Society of Nephrology (to M.H.F.).
Abbreviations used in this paper: GPS, Goodpasture syndrome; GBM, glomerular basement membrane; Tg, transgenic; B6, C57BL/6.