Drs. Chowdhry and Kowal contributed equally to this work.
Autoantibodies that bind glomeruli: Cross-reactivity with bacterial antigen
Version of Record online: 28 JUL 2005
Copyright © 2005 by the American College of Rheumatology
Arthritis & Rheumatism
Volume 52, Issue 8, pages 2403–2410, August 2005
How to Cite
Chowdhry, I. A., Kowal, C., Hardin, J., Zhou, Z. and Diamond, B. (2005), Autoantibodies that bind glomeruli: Cross-reactivity with bacterial antigen. Arthritis & Rheumatism, 52: 2403–2410. doi: 10.1002/art.21143
- Issue online: 28 JUL 2005
- Version of Record online: 28 JUL 2005
- Manuscript Accepted: 8 APR 2005
- Manuscript Received: 30 JUN 2004
- National Institute of Arthritis and Musculoskeletal and Skin Diseases
- New York Chapter of the Arthritis Foundation
Systemic lupus erythematosus (SLE) is characterized by the production of multiple autoantibodies. Anti-DNA antibodies are associated with glomerulonephritis in SLE. It has been shown that anti-DNA antibodies cross-react with bacterial polysaccharide and, thus, might be elicited by microbial exposure. Non–DNA-binding antibodies also contribute significantly to the pathogenesis of lupus nephritis. The goal of this study was to characterize non–DNA-binding, kidney-binding antibodies.
We generated a combinatorial library derived from spleen cells of a patient with SLE who had just previously received pneumococcal vaccine. The phage library was used in an in vivo biopanning technique to identify non–DNA-binding, kidney-binding antibodies. Antibodies were then analyzed for binding to bacterial polysaccharide and to renal antigens.
Eight antibodies were characterized that bound glomeruli, but not DNA. All antibodies isolated by this protocol were IgG class, suggesting that there is affinity maturation for glomerular binding. Four of the antibodies cross-reacted with pneumococcal polysaccharide. Six of the antibodies bound to renal antigens that have previously been reported to be cross-reactive with DNA; the other 2 bound to histone.
This study suggests that both DNA-binding and non–DNA-binding antibodies in SLE may be elicited by the same bacterial antigens.
Systemic lupus erythematosus (SLE) is characterized by the production of autoantibodies. Anti–double-stranded DNA (anti-dsDNA) antibodies are present in the serum of most SLE patients, and fluctuations in titers correlate with disease activity, especially glomerulonephritis (1, 2). The association of nephritis with anti-DNA reactivity is strengthened by the demonstration that anti-DNA antibodies can be eluted from diseased kidneys (3, 4).
The origin of DNA-reactive antibodies is not known. Several recent studies have highlighted the potential for chromatin to act as the eliciting antigen (for review, see ref. 5), but other studies have demonstrated cross-reactivity with bacterial antigens (for review, see ref. 6). We have previously shown in a murine study that anti-DNA antibodies that are cross-reactive with antipolysaccharide antibodies routinely arise during the immune response to bacterial antigen but undergo negative selection in a nonautoimmune host, demonstrating the potential for antibodies to arise in the course of a response to microbial infection (7).
In murine studies, anti-DNA antibodies have been found to be sequestered in the kidney, where they contribute to glomerulonephritis (8). Several lines of evidence suggest that these antibodies deposit in the kidney by virtue of their cross-reactivity with renal antigens. First, anti-DNA antibodies will not deposit in glomeruli when they are complexed with antigen, suggesting that the antigen-binding site of the antibody must be free to bind to a renal antigen. Second, in vitro studies have demonstrated that some anti-DNA antibodies will bind to glomerular preparations that have been treated with DNase (9). Third, several renal protein antigens, including fibronectin, laminin, α-actinin, and heparan sulfate, are bound directly by monoclonal anti-DNA antibodies (10–15). Fourth, clinical studies have established that non–DNA-binding antibodies can also be pathogenic in the kidney.
Approximately 15% of SLE patients with renal disease have no detectable serum titers of anti-DNA antibodies (16). Furthermore, only half of the Ig eluted from the kidneys of lupus patients binds to DNA (12, 17, 18).
To gain a better understanding of which types of antibodies can bind directly to renal antigen, we developed a method to isolate renal-binding antibodies. Our method is based on biopanning for bacteriophages that express recombinant Ig V-region genes derived from the B cells of a lupus patient. The phage library was generated from IgM and IgG heavy-chain sequences and κ light-chain sequences from the spleen cells of a lupus patient who received a pneumococcal vaccination prior to splenectomy. In the present study, we focused on analyzing autoantibody specificities that bind to renal glomeruli but not DNA. Characterization of these antibodies showed that they are uniquely IgG, display antigenic specificities that closely resemble nephritogenic anti-DNA antibodies, and many of them cross-react with bacterial polysaccharide.
MATERIALS AND METHODS
Construction of combinatorial library
We obtained a portion of a spleen from an SLE patient who had nephritis and idiopathic thrombocytopenia and had been immunized with pneumococcal polysaccharide (Pneumovax-23; Merck, West Point, PA) prior to therapeutic splenectomy. The preparation of total cellular RNA and the generation of the combinatorial library with IgM and IgG heavy-chain sequences and κ light-chain sequences have previously been described (19).
In vivo biopanning
Twenty microliters of the original combinatorial library was amplified, and 1 × 1012 freshly prepared phage particles in 200 μl of 1% bovine serum albumin (BSA)–phosphate buffered saline (PBS) were injected intravenously into BALB/c SCID mice. The mice were killed 24 hours after injection, and their kidneys were pooled and homogenized in a Leffert's buffer. Phage particles were eluted from the kidney preparations using 0.1M HCl buffer (pH 2.2) containing 1 mg/ml of BSA and amplified before injection into a second group of mice. Three cycles of in vivo biopanning were performed using 5 mice for each cycle.
Phage enzyme-linked immunosorbent assays (ELISAs)
All phage ELISAs were performed on Costar 3690 plates (Corning, Corning, NY). All incubations were performed for 1 hour at room temperature. Clones containing both heavy and light chains, as determined by hybridization, were assayed for DNA specificity. The plates were coated overnight at 37°C with 50 μg/ml of calf thymus dsDNA, allowed to dry, and then blocked with 1% BSA–PBS. Phage particles were incubated at the concentrations indicated below and the incubated with phage-specific biotinylated sheep anti-M13 antibody (5 Prime 3 Prime, Boulder, CO) at a 1:1,000 dilution in 0.2% BSA–PBS. Streptavidin-conjugated alkaline phosphatase (Southern Biotechnology, Birmingham, AL) was added at a 1:1,000 dilution in 0.2% BSA–PBS.
Non–DNA-binding phage clones were further tested for specificity to Pneumovax and to chicken α-actinin, mouse laminin, bovine fibronectin, mouse Matrigel, bovine histone H1, and bovine DNase I. Pneumovax was purchased from Merck (West Point, PA), calf thymus histone H1 was from Upstate Biotechnology (Lake Placid, NY), and all other antigens were purchased from Sigma (St. Louis, MO). Pneumovax was dry-coated at 25 μg/ml in PBS overnight at 37°C. The remaining antigens were adhered to microtiter wells overnight at 4°C. Laminin and α-actinin were diluted in PBS to 20 μg/ml and 10 μg/ml, respectively, fibronectin and Matrigel were diluted in PBS to 1 μg/ml, and histone H1 and DNase I were diluted to 10 μg/ml in a sodium carbonate buffer (0.035M NaHCO3, 0.015M Na2CO3, pH 9.6). Assays were subsequently performed as described for the DNA ELISA. Antibodies that bound to histone were detected with horseradish peroxidase–labeled anti-M13 antibody (Amersham Biosciences, Piscataway, NJ) and developed using ABTS (Zymed, San Francisco, CA).
Glomerular binding assay and tissue staining
Rat glomeruli were extracted as described previously (9), adhered to glass slides (Superfrost Plus; Fisher Scientific, Pittsburgh, PA), and stored at –20°C.
Prior to use, the slides were thawed for 2 minutes at room temperature, washed with PBS–Tween, and treated with 200 μg/ml of DNase I (Sigma) at 37°C until complete removal of DNA was verified by propidium iodide staining. The slides were then blocked with 10% goat serum in PBS for 1 hour at room temperature, followed by incubation of phage particles at 1 × 1012 per section for 1 hour at room temperature or overnight at 4°C. The slides were then washed with PBS–Tween. Fluorescein isothiocyanate (FITC)–labeled goat anti-human κ secondary antibody (Fisher Scientific) at a 1:1,000 dilution was added in 0.2% BSA–PBS for 1 hour at room temperature, since the library contained Fab with only κ light chains. The slides were washed with PBS–Tween, air dried, treated with Aqua Poly/Mount mounting solution (Polysciences, Warrington, PA), and the intensity of fluorescence was observed under the fluorescent microscope (Zeiss, Thornwood, NY) using AxioVision software (Zeiss). A pComb vector phage was used as a negative control.
Human kidney was obtained from a non-SLE patient who underwent therapeutic nephrectomy. Histologically normal glomeruli were extracted as described previously (9), adhered to glass slides (Superfrost Plus), and stored at –80°C. Staining of human glomeruli was performed in a manner similar to that for rat glomeruli staining, with a few modifications. The slides were blocked with 3% fetal calf serum and 10% goat serum in PBS for 1 hour at 37°C, followed by incubation of phage particles at 1 × 1012 per slide overnight at 4°C. FITC-labeled goat anti-human κ secondary antibody at a 1:1,000 dilution was added in 0.2% BSA–PBS for 1 hour at 37°C.
In addition, sections of human skin and myometrium were adhered to slides and stained as described above.
Ig gene sequencing
DNA was prepared using reagents from Qiagen (Valencia, CA). The sequences were analyzed using the BLAST search tool at the National Center for Biotechnology Information (National Institutes of Health, Bethesda, MD; online at www.ncbi.nlm.nih.gov). Sequences have been submitted to GenBank.
Isolation of glomerulotropic antibodies. Phage-expressing human Fab fragments derived from spleen cells obtained from a patient with SLE were injected into naive SCID mice, and those that sequestered in the kidney were isolated. Three rounds of in vivo biopanning were performed. Fab-expressing clones were identified by hybridization after the second and third cycles of biopanning, and clonal repeats were excluded by sequencing. Sequence analysis revealed reduced diversity after the third panning, which was most likely due to the growth advantage and increased affinity for kidney antigens of particular clones; therefore, clones from both the second and third pannings were included in further studies. We screened a random population of clones from the second biopanning cycle for binding to isolated rodent and human glomeruli. Approximately 80% of the tested clones bound to both rodent and human glomeruli in vitro (Figures 1a and b). From the third cycle, all clones bound glomeruli. The selection strategy, therefore, was very efficient, with a great enrichment for specific antibodies.
There was selection against phages expressing a μ heavy chain, and enrichment for phages expressing a γ heavy chain. While 20% of the original library expressed a μ heavy chain, after 2 rounds of selection, only 3% of the phages expressed a μ heavy chain, and after 3 rounds, none expressed a μ heavy chain. Conversely, 40% of the original library expressed a γ heavy chain, and 60% of the phages expressed a γ heavy chain after 2 rounds of selection.
Approximately 50% of bioselected phages bound dsDNA. We focused further analysis on phages that did not bind dsDNA. To confirm the tissue specificity of selected phage antibodies that did not bind DNA, human skin and myometrium were incubated with phages. There was no binding to skin or smooth muscle (data not shown); thus, binding was specific for renal structures. We also tested the phages by ELISA for reactivity to DNase I, since it was possible that some DNase I remained attached to the glomerular preparations. Only one phage (clone 2-2) displayed reactivity with DNase I (data not shown).
Characterization of kidney-binding antibodies. The variable (V)–region genes were sequenced from selected phage clones. Eight unique Fab fragments were identified (Table 1). Two of them, clones 29 and 32, were derived from the third panning. The rest were derived from the second panning. These Fab fragments possessed γ1, γ2, and γ3 heavy chains. Five were found to be encoded by a member of the VH3 gene family; 3 of these 5 were encoded by VH3-23. One was encoded by VH4-4, another by VH6-1, and another by VH1-2. Four light chains were encoded by a Vκ3 gene, 3 by a Vκ1 gene, and 1 by a Vκ2 gene. V gene usage for non–DNA-binding, kidney-binding antibodies was heterogeneous, similar to V gene usage for DNA-binding antibodies (20, 21). VH or Vκ family usage did not distinguish kidney-binding antibodies that cross-reacted with DNA from those that did not.
|2-1||Pn, H||VH4-4||D5-18||JH5||γ1||Vκ3; A27||Jκ1|
|2-2||Pn, F, M||VH3-23||D3-9||JH4||γ1||Vκ3; L6||Jκ5|
|2-8||A, L||VH3-23||D5-12||JH6||γ2||Vκ3; A27||Jκ1|
|29||Pn, M||VH1-2||D6-19||JH5||γ3||Vκ3; L8||Jκ4|
|32||Pn, A, F, L, M||VH6-1||D6-19||JH3||γ3||Vκ1; L11||Jκ3|
Heavy-chain and light-chain V regions were compared with published germline sequences. Both heavy-chain and light-chain V regions have undergone somatic mutation, with the heavy chains displaying extensive mutation. In 6 of the antibodies, the mutations were predominantly replacement mutations, which strongly suggests antigen selection (Table 2). While it is possible that the antibodies are kidney-binding in the germline configuration, we believe this to be unlikely, given the absence of IgM antibodies in the selected population.
We next tested the phages for binding to antigens implicated in lupus nephritis. Six of the 8 antibodies displayed specificity for 1 or more known renal antigens, α-actinin, fibronectin, laminin, and Matrigel, which contains laminin, heparan sulfate, and type IV collagen (Table 1 and Figure 2). The binding was not nonspecific, since each phage clone, except for clone 32, was negative in at least 1 ELISA, and clone 32 was negative on tissue sections. Phage 2-2 with reactivity to DNase I also bound Matrigel and fibronectin. Two phages, clones 2-1 and 2-5, did not bind to any of these antigens. Because histones are a known autospecificity in lupus, and because chromatin can be bound to glomerular basement membrane, we tested these phages for binding to purified histone proteins. Both 2-1 and 2-5 bound to purified histone on Western blotting (data not shown) and by ELISA (Figure 2). This binding could be detected with either anti-human κ or anti-M13 phage-specific antibody. Vector alone did not exhibit this binding (Figure 2). Thus, all 8 Fab fragments had specificities that are known to characterize pathogenic anti-DNA antibodies. Six bound renal protein antigens and 2 bound histone, which, like chromatin, can be sequestered in renal glomeruli.
The patient from whom the library was derived had received Pneumovax prior to splenectomy, which permitted us to investigate whether the kidney-binding Fab fragments bound also to pneumococcal polysaccharide. Four Fab fragments displayed this microbial cross-reactivity. Thus, antipolysaccharide antibodies that may have been elicited by pneumococcal vaccination can bind glomerular proteins.
Determining the antigenic specificity of kidney-binding antibodies has been a challenge for many years. This study demonstrates that non–DNA-binding antibodies from a patient with lupus can bind both bacterial and renal antigens, and this study is the first to show that multiple autospecificities in SLE are cross-reactive with the same bacterial antigen. This observation is consistent with a model in which SLE is triggered by molecular mimicry of a foreign antigen. Consistent with this hypothesis, our laboratory has recently demonstrated in a murine model that a synthetic foreign antigen can induce antibodies that are cross-reactive with DNA (22). Thus, molecular mimicry can lead to an SLE-like illness in mice upon exposure to antigen. It was recently shown that antibody-mediated glomerulonephritis can occur in the absence of any antinuclear specificities (23). Thus, antibodies that cause a lupus-like nephritis can be induced by foreign antigen and, perhaps, by non-nuclear self antigens, as well as by chromatin (24–26). There is, therefore, heterogeneity in the mechanisms of the induction of autoantibodies that cause glomerulonephritis in mice. It is important to remember that the heterogeneity in lupus patients may be at least as great as the heterogeneity in mouse models.
Combinatorial libraries expressing Fab fragments have proven to be a useful way to analyze antibodies of a particular specificity. While the combinatorial library consists of a random association of heavy and light chains, and, in principle, might include combinations that do not exist in vivo, studies that have actually compared antibodies to a particular antigen produced by B cells with antibodies created by combinatorial technology have demonstrated stunning similarity (27–31). For example, high-affinity DNA binding was not displayed by phage clones from a combinatorial library generated from peripheral blood B cells obtained from a subject without autoimmune disease, while clones with this specificity were readily identified in a library from a patient with SLE (32).
We have developed a novel protocol that permits the identification of antibodies that are sequestered in the kidney. This protocol depends on the presence of shared antigenic determinants in human and mouse kidney. There was reason to believe that renal antigens would be shared across species (33, 34). Studies of established animal models of Goodpasture's syndrome, an autoimmune glomerulonephritis caused by anti– glomerular basement membrane (anti-GBM) antibodies, have demonstrated the conservation of mammalian GBM structures (33, 34). Moreover, sequence analysis of the Goodpasture's antigen of 9 mammalian species, including human and mouse, shows more than 90% identity spanning ∼200 amino acids (35). These studies encouraged us to establish a method by which to identify human kidney-binding antibodies relevant in lupus nephritis by screening in vivo on mouse kidney. The method proved to be successful, yielding monoclonal Fab fragments exhibiting binding to both rat and human isolated glomeruli, as well as binding to several renal antigens tested by ELISA.
In this study, approximately half of the antibodies selected for glomerular binding were also DNA binding. This result is remarkably consistent with the observation that only 50% of antibodies eluted from lupus glomeruli bind DNA (12, 17, 18). Thus, substantial numbers of non–DNA-binding antibodies may contribute to renal disease. The most striking feature of the antibodies is that all possess IgG heavy chains. The selection of phages expressing IgG Fab fragments appears to be the result of panning on kidney, since panning of the same library for an idiotype specificity yielded clones dominated by IgM heavy chains (19).
Clinically, individuals with high-affinity anti-DNA antibodies are more likely to have renal disease, whereas those with low-affinity antibodies may escape tissue injury (3, 12). Furthermore, in mice prone to the spontaneous development of lupus and in induced mouse models of lupus, the onset of disease correlates with the onset of heavy-chain class-switching of autoantibodies (36, 37). Because heavy-chain class-switching and somatic mutation, in general, occur concurrently (38), it has not been clear whether the preferential pathogenicity of IgG autoantibodies relates to a higher affinity of the antibodies following somatic mutation or to the presence of Fc receptors for IgG on mesangial cells increasing the tethering of the antibody to the kidney. The current data support the hypothesis that the preferential pathogenicity of IgG antibodies is due to their fine antigenic specificity and affinity, and demonstrate conclusively that the preferential deposition of IgG in the kidney is not exclusively a function of an interaction between IgG and Fc receptors since there is no Fc receptor binding domain in the portion of the constant region expressed in the phage Fab fragments. Because many phage will express 2 Fab fragments, the data suggest that 1) antibodies with sufficiently high affinity for renal antigens in their germline configuration to be sequestered in renal tissue are not present in the repertoire and 2) the requisite affinities are achieved by somatic mutation, concomitant with heavy-chain class-switching. Consistent with this hypothesis is the observation that all the variable regions of the phage Fab fragments, especially the heavy-chain variable regions, contain multiple mutations, with a strong preponderance of replacement mutations.
SLE patients and lupus-prone mice have been shown to have elevated serum titers of antibodies to laminin, fibronectin, collagen, heparan sulfate, and α actinin (10–15). Several non–DNA-binding antibodies we isolated shared cross-reactivity with these antibodies. Thus, binding to extracellular matrix components is common to non–DNA-reactive glomerulotropic antibodies.
Our data further suggest that non–DNA-binding antibodies are binding to the same renal antigens that are recognized by DNA-binding antibodies. Two Fab fragments bound histone. It is known that antihistone antibodies are present in murine and human SLE and that antihistone antibodies can bind to chromatin sequestered in glomeruli. The authors of a recent study describing lupus nephritis in mice that lack anti-DNA antibodies suggested that antihistone antibodies are present and may contribute to nephritis (23). The use of DNase I to remove DNA from the glomeruli will not remove histone proteins. Thus, the 2 Fab fragments that were not cross-reactive with known renal antigens bound to histone, another known target antigen in SLE that can mediate glomerular immunoglobulin deposition. At present, it is not clear whether the lack of focal localization of Fab fragments in the glomeruli-binding assay reflects the fact that the assay is performed by bathing isolated glomeruli in antibody or whether it represents a ubiquitous distribution of the renal antigens that are recognized by these monoclonal Fab fragments.
The observation that 50% of glomerulotropic antibodies do not bind DNA raises the question of the identity of the eliciting antigens. There has been much recent interest in chromatin as a critical antigen in SLE. Chromatin can trigger B cells through ligation of Toll-like receptor 9 (TLR-9) (39). Therefore, increased apoptosis or decreased clearance of apoptotic particles might activate anti-DNA B cells through crosslinking of TLR-9 and membrane Ig. Perhaps 2 of the Fab fragments arose through this pathway, but it is important to note that 1 of these Fab fragments also bound pneumococcal polysaccharide. An alternative hypothesis that we and others have proposed, is that foreign antigen, pneumococcal polysaccharide, or other microbial antigens trigger a response to renal antigens and to DNA through a process of somatic mutation (7). Consistent with this model is our observation that 4 of the antibodies bound pneumococcal polysaccharide, including 1 that was cross-reactive with histone. Cross-reactivity to this microbial antigen was also found to occur in a high percentage of anti-DNA antibodies isolated from this library, which was derived from a patient who had just been immunized with pneumococcal polysaccharide (19). Those data, together with the data from this study, support a hypothesis that microbial antigens may be a trigger for both DNA-binding and non–DNA-binding antibodies that recognize kidney antigens.
It would obviously be of interest to determine if such cross-reactive antibodies arise routinely in all individuals. Unfortunately, it is not possible to obtain spleen samples from a nonautoimmune individual following pneumococcal vaccination. Based on serum studies, however, we believe this cross-reactivity is specific to SLE patients, since studies from one of our laboratories have shown that nonautoimmune subjects do not develop autoantibodies following pneumococcal vaccination (40).
The identification of non–DNA-binding antibodies that bind to glomeruli will make possible the development of new assays to identify potentially pathogenic antibodies. Furthermore, the analysis of the antibodies will ultimately yield a more comprehensive picture of the spectrum of autoantigens that is targeted in SLE.
The methodology described in this paper may be used in any system that depends on the interaction of tissue antigens with a protein that can be displayed on phage particles.
We thank Dr. Matthew Scharff for critical reading of the manuscript. We also thank Dr. Chaim Putterman and Xiaoping Qing for the generous gift of rat glomeruli. We thank Olatilewa Awe for technical assistance and Sylvia Jones for assistance in the preparation of the manuscript.