The specificity of the autoantibody response in different autoimmune diseases makes autoantibodies useful for diagnostic purposes. It also focuses attention on tissue- and event-specific circumstances that may select unique molecules for an autoimmune response in specific diseases. Defining additional phenotype-specific autoantibodies may identify such circumstances. This study was undertaken to investigate the disease specificity of PMS1, an autoantigen previously identified in some sera from patients with myositis.
We used immunoprecipitation analysis to determine the frequency of autoantibodies to PMS1 in sera from patients with myositis, systemic lupus erythematosus, or scleroderma and from healthy controls. Additional antigens recognized by PMS1-positive sera were further characterized in terms of their susceptibility to cleavage by apoptotic proteases.
PMS1, a DNA mismatch repair enzyme, was identified as a myositis-specific autoantigen. Autoantibodies to PMS1 were found in 4 of 53 patients with autoimmune myositis (7.5%), but in no sera from 94 patients with other systemic autoimmune diseases (P = 0.016). Additional mismatch repair enzymes (PMS2, MLH1) were targeted, apparently independently. Sera recognizing PMS1 also recognized several other proteins involved in DNA repair and remodeling, including poly(ADP-ribose) polymerase, DNA-dependent protein kinase, and Mi-2. All of these autoantigens were efficiently cleaved by granzyme B, generating unique fragments not observed during other forms of cell death.
PMS1 autoantibodies are myositis specific. The striking correlation between an immune response to a group of granzyme B substrates (functioning in DNA repair and remodeling) and the myositis phenotype strongly implies that tissue- and event-specific biochemical events play a role in selecting these molecules for an autoimmune response. Understanding the role of granzyme B cleavage in this response is an important priority.
Autoantibodies are valuable tools with which to explore the initiating perturbed state in systemic autoimmune diseases (1). Several components of the DNA repair and nucleosome remodeling machinery are targets of the autoimmune response in this spectrum of diseases. Examples include poly(ADP-ribose) polymerase (PARP) (2), several components of DNA-dependent protein kinase (DNA-PK) (including Ku-70, Ku-86, and the catalytic subunit of DNA-PK [DNA-PKcs] [3–5]), and Mi-2 (a component of the nucleosome remodeling and deacetylase machinery) (6–8). Interestingly, in addition to their functioning in a similar process, most of these autoantigens are unified by the fact that they are cleaved by granzyme B and/or caspases during apoptosis (9). In a recent screen of a HeLa cell complementary DNA (cDNA) expression library with a serum from a patient with myositis, autoantibodies that recognize PMS1, a DNA mismatch repair protein which is mutated in some patients with hereditary nonpolyposis colorectal cancer, were identified (10). Like many other components of the DNA repair and remodeling machinery, both PMS1 and its close homolog PMS2 are efficiently cleaved by granzyme B during cytotoxic lymphocyte granule–induced apoptosis (9). In the present studies, we investigated the frequency and disease specificity of autoantibodies to the DNA mismatch repair proteins PMS1 and PMS2 in patients with systemic autoimmune diseases.
PATIENTS AND METHODS
Sera were obtained from 2 index patients with autoimmune myositis (meeting the Bohan and Peter criteria for dermatomyositis [serum PK] or polymyositis [serum AG]) (11) seen at the Johns Hopkins Medical Institutions, as well as from 51 randomly selected patients with biopsy-proven autoimmune myositis from the National Institutes of Health (NIH) myositis cohort. The NIH patients had dermatomyositis (47%) or polymyositis (45%) which satisfied the Bohan and Peter criteria (11), or eosinophilic myositis (6%) or inclusion body myositis (2%). PMS1-positive sera did not recognize any of the aminoacyl–transfer RNA (aminoacyl-tRNA) synthetases by immunoprecipitation (using reference sera to histidyl-, threonyl-, alanyl-, glycyl-, and isoleucyl-aminoacyl tRNA synthetases as standards). These PMS1-positive patients did not have clinical features of the “synthetase syndrome.” Sera from 50 patients with systemic lupus erythematosus (SLE) meeting the American College of Rheumatology (ACR; formerly, the American Rheumatism Association) diagnostic criteria (12) were randomly selected from the Hopkins Lupus Cohort, a prospective study of outcomes in SLE. Sera from 44 patients with scleroderma were randomly selected from stored samples. Thirty-seven of the 44 met ACR criteria for a diagnosis of scleroderma (13), and the remaining 7 had 3 of the 5 features of the CREST syndrome (calcinosis, Raynaud's phenomenon, esophageal dysmotility, sclerodactyly, telangiectasias). Equal numbers of patients with the diffuse and limited forms of scleroderma were included. A summary of demographic characteristics of the patients is presented in Table 1.
Table 1. Demographic characteristics of the patients
Serum PK was used to screen a λGT11 HeLa cell cDNA expression library. Several positive clones were processed to plaque purity, and the inserts were excised by restriction digestion with EcoRI. After subcloning into pBluescript II and sequencing, clone 1531 was found to encode the human DNA mismatch repair enzyme PMS1.
Generation of antibody raised against the C-terminus of human PMS1
A peptide corresponding to amino acid residues 903–922 at the C-terminus of human PMS1 (Boston Biomolecules, Woburn, MA) was used to immunize rabbits (Covance Research Products, Denver, PA). Serum from rabbit 3213 contained antibodies that detected PMS1. In some experiments, antibody sc-615 (Santa Cruz Biotechnology, Santa Cruz, CA) was used instead of serum 3213 to detect the C-terminus of PMS1.
Generation of 35S-methionine–labeled autoantigens and cleavage in vitro by granzyme B
35S-methionine–labeled PMS1, PMS2, and MLH1 were generated by coupled in vitro transcription/translation (IVTT) (14), using the appropriate full-length cDNAs (all generous gifts from Dr. B. Vogelstein, Johns Hopkins University). Radiolabeled PMS1 was diluted in buffer A (consisting of 10 mM HEPES [pH 7.4], 2 mM EDTA, and 1% Nonidet P40 [NP40]) and incubated in the absence or presence of 70 nM granzyme B (15) for 15 minutes at 37°C. Reactions were either terminated, by adding gel buffer and boiling, or were used for immunoprecipitations. In the latter case, the activity of granzyme B was stopped by adding 70 μM chymostatin and placing the reaction mixtures on ice.
Immunoprecipitation of IVTT substrates
Standard immunoprecipitations to test which sera immunoprecipitated radiolabeled substrates were performed as follows. IVTT substrates were diluted with buffer B (consisting of 1% NP40, 20 mM Tris [pH 7.4], 150 mM NaCl, 1 mM EDTA, and 0.2% bovine serum albumin) supplemented with a protease inhibitor cocktail. Serum (2 μl) was added and the mixture was rocked for 1 hour at 4°C, after which immobilized protein A–agarose (Pierce, Rockford, IL) was added. The immunoprecipitates were washed, electrophoresed on 10% sodium dodecyl sulfate (SDS)–polyacrylamide gels, and visualized by fluorography. In experiments in which immunoprecipitations were performed using granzyme B–cleaved PMS1, the reaction mixtures were either incubated on ice or SDS was added (0.7% final concentration) and the samples were boiled for 5 minutes. After boiling, the SDS concentration was decreased to 0.1% by adding buffer B. Immunoprecipitations were then performed as described above.
Immunoblotting to detect cleavage of endogenous HeLa substrates
HeLa cells were cultured and induced to become apoptotic by ultraviolet B irradiation as previously described (16, 17). For in vitro incubations in the absence or presence of purified granzyme B, HeLa lysate was prepared (9) and incubated with 42 nM granzyme B and 2 mM iodoacetamide for 1 hour at 37°C. Reactions were terminated by adding gel buffer and boiling, and samples were electrophoresed on 10% SDS-polyacrylamide gels. Proteins were transferred to nitrocellulose and immunoblotted as previously described (18). Blotted proteins were detected with horseradish peroxidase–labeled secondary antibody and chemiluminescence (Pierce). All experiments were performed on at least 2 separate occasions, using different batches of radiolabeled substrates or HeLa lysates.
The significance of the relationship of PMS1 targeting to systemic autoimmune diseases was investigated using Fisher's exact test. Statistical significance was defined as an α level of 0.05.
Detection of PMS1. Serum PK, obtained from a patient with dermatomyositis, was used to screen a λGT11 HeLa cell cDNA expression library. One of the positive clones had an insert of 0.7 kb, which was identical to the human DNA mismatch repair enzyme PMS1. We obtained a full-length cDNA for PMS1 and generated 35S-methionine–labeled PMS1 by coupled IVTT. The labeled 120-kd product was immunoprecipitated by serum PK and rabbit serum 3213 (raised against the C-terminus of PMS1), but not by normal human or preimmune serum from rabbit 3213 (Figure 1). Interestingly, although rabbit anti-PMS1 immunoblotted a molecule of 120 kd in control HeLa cell lysates, serum PK was not able to immunoblot a similar species, indicating that the epitope recognized by the autoantibody to PMS1 is likely conformational (results not shown).
Recognition of the C-terminus of PMS1 by PMS1 autoantibodies. We have previously observed that PMS1 is efficiently cleaved by granzyme B, generating fragments of 55 kd and 48 kd (9) (Figure 2, lanes 1 and 2). We initially used rabbit antipeptide antibodies raised against the N- or C-terminus of PMS1 to immunoprecipitate these fragments from granzyme B–cleaved 35S-methionine–labeled PMS1 generated by IVTT. The 48-kd fragment was immunoprecipitated by the anti–C-terminal antibody (Figure 2, lane 4), while the 55-kd fragment was recognized by an antibody raised against the N-terminus of PMS1 (results not shown). Interestingly, the anti–C-terminal antibody also immunoprecipitated a small amount of the 55-kd N-terminal fragment (Figure 2, lane 3), suggesting that the N- and C-terminal fragments of PMS1 remain associated even after granzyme B–mediated cleavage. When the cleavage reaction mixture was boiled in SDS (which was subsequently diluted) prior to immunoprecipitation, recognition of the intact protein and the C-terminal 48-kd fragment was improved, but the 55-kd fragment was no longer detected (Figure 2, lane 4). These data demonstrate that the N- and C-terminal fragments of PMS1 remain at least partially associated after cleavage by granzyme B, unless separated by denaturation.
Similar immunoprecipitation experiments were performed using human autoantibodies recognizing PMS1 (sera PK [Figure 2] and 1088 [results not shown]). In the absence of boiling, both the 55-kd and the 48-kd fragments were immunoprecipitated by both human sera. In contrast, after boiling of the reaction mixture in SDS to disrupt complexes and diluting prior to immunoprecipitation, the 48-kd C-terminal fragment alone was immunoprecipitated by human sera. These findings confirm that human autoantibodies to PMS1 directly recognize the C-terminal granzyme B fragment, rather than an associated component.
Frequency and specificity of autoantibodies to PMS1. To investigate the frequency of targeting of PMS1 in systemic autoimmune diseases, sera from healthy controls or patients with myositis, SLE, or scleroderma were screened in an immunoprecipitation assay, using 35S-methionine–labeled PMS1 generated by IVTT. Four of 53 sera from patients with autoimmune myositis (7.5%) recognized PMS1 (Figure 3). In contrast, no sera from 94 other patients with systemic autoimmune diseases (50 from patients with SLE and 44 from patients with scleroderma) immunoprecipitated PMS1 (P = 0.016 by Fisher's exact test). In addition, none of the 39 healthy control sera immunoprecipitated PMS1. Interestingly, serum from 1 subject who did not have a systemic autoimmune disease, but who had an active zoster eruption, was also positive for antibodies to PMS1 in the immunoprecipitation assay (results not shown). Thus, the occurrence of antibodies to PMS1 was specific for autoimmune myositis as compared with SLE and scleroderma, and the difference was statistically significant. The relatively low frequency of this autoantibody response in myositis patients is in a range similar to that found with several other targeted protein autoantigens in this spectrum of disease (e.g., threonyl-, alanyl-, glycyl-, and isoleucyl-aminoacyl tRNA synthetases [each <2–3%], signal recognition particle [SRP] [∼4%], Mi-2 [<8%], and PM-Scl [<8%]) (19).
Recognition of several components of the DNA repair machinery by sera from patients with autoimmune myositis. Sera that recognized PMS1 by immunoprecipitation (sera PK, AG, 1088, and 946) were immunoblotted against lysates of HeLa cells. Although rabbit antibodies against PMS1 recognized their cognate antigen by immunoblotting in HeLa cell lysates, PMS1 was not detected by any of the human sera by immunoblotting. Interestingly, each of these human myositis sera recognized at least 1 other antigen by blotting and/or precipitation (Table 2). Serum PK immunoprecipitated PMS1 and an unidentified 68-kd protein, and immunoblotted Mi-2 (Figure 4, lane 1). The identity of this Mi-2 antibody in serum PK was confirmed by comigration in immunoblots and immunoprecipitation with a standard Mi-2–positive serum (ref. 9 and results not shown). Serum AG immunoblotted several proteins, including DNA-PKcs, PARP, and unidentified antigens of 190 kd and 130 kd (Figure 4, lane 5). The identities of the DNA-PKcs and PARP antibodies in serum AG were confirmed by immunoblotting and documented in an earlier study (16). Serum 946 recognized, in addition to PMS1, an unidentified 125-kd protein (p125) by immunoblotting (Figure 4, lane 9). Serum 1088 also recognized in vitro–translated PMS2 and MLH1 by immunoprecipitation (see below), and a single 85-kd protein (p85) by immunoblotting (Figure 4, lane 13).
Table 2. Antigens recognized by sera PK, AG, 946, and 1088*
Data were obtained from the immunoblots shown in Figure 4 and from immunoprecipitations using radiolabeled substrates generated by in vitro transcription/translation (Figures 1 and 3 and results not shown). DNA-PKcs = catalytic subunit of DNA-dependent protein kinase; PARP = poly(ADP-ribose) polymerase.
Unidentified antigens are referred to by molecular weight, as estimated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis.
Since many autoantigens are proteolytically cleaved during apoptosis (1, 20), we investigated whether the additional antigens recognized by PMS1-positive patient sera were cleaved in apoptotic cells (Figure 4). In every case, the molecules were cleaved in apoptotic cells (Figure 4, lanes 2, 6, 10, and 14), generating stable fragments that were not further processed. Recent studies have demonstrated that ∼80% of autoantigens across the spectrum of human autoimmune disease are susceptible to cleavage by granzyme B during cytotoxic lymphocyte granule–induced cell death, generating unique fragments not observed during any other form of apoptotic death (9). We addressed whether the additional antigens recognized by PMS1-positive sera were susceptible to cleavage by granzyme B by adding purified granzyme B to lysates of HeLa cells in which endogenous caspases had been inactivated by alkylating the active site cysteine. In every case, these antigens were efficiently cleaved by granzyme B (Figure 4, solid arrows), and the fragments generated were distinct from those observed in apoptotic cells (Figure 4, open arrows). The immunoblotted cleavage products shown in Figure 4 correspond to those previously reported for Mi-2 (lanes 2 and 4) and DNA-PK and PARP (lanes 6 and 8) (9, 15, 16).
Autoantibodies to other DNA mismatch repair enzymes. PMS1 belongs to the MutL family of DNA mismatch repair enzymes. All family members have a conserved region of ∼300 amino acids at their N-terminus and a very diverse C-terminal region of 300–500 residues that shows no significant homology between family members (Figure 5). We generated radiolabeled PMS2 by IVTT and tested the same series of sera for antibodies to PMS2 by immunoprecipitation (Figure 3 and Table 3). Sera from 2 of 53 patients with myositis (∼3.7%) recognized PMS2 by immunoprecipitation. Interestingly, 1 of the positive sera (serum 1088) recognized both PMS1 and PMS2, while the other recognized PMS2 but not PMS1 (Table 3). Although PMS2 antibodies were not found in controls (0 of 29) or scleroderma patients (0 of 44), 1 of 50 SLE patients did have this antibody. This patient did not have clinically apparent myositis either at the time of assay or previously.
Table 3. Clinical features of the myositis patients with antibodies directed against DNA mismatch repair enzymes*
Elevated muscle enzymes
EMG = electromyography; C = Caucasian; NA = not available; AA = African American; ND = not done; EM = eosinophilic myositis (see Table 2 for other definitions).
Meets Bohan and Peter criteria for definite dermatomyositis (DM).
Meets Bohan and Peter criteria for probable polymyositis (PM); patient also had serologic evidence of systemic lupus erythematosus.
All myositis sera were also screened in an immunoprecipitation assay against in vitro–translated MLH1. Both of the PMS2-positive myositis sera (sera 962 and 1088) also recognized MLH1. Only 1 of these sera (serum 1088) also immunoprecipitated PMS1. One myositis serum (serum 981) that was negative in the immunoprecipitation assay for PMS1 and PMS2 was positive for antibodies to MLH1. Thus, although there are several examples of myositis sera recognizing multiple MutL-like enzymes, there are also clear examples in which each serum recognizes only 1 of these molecules. These data demonstrate that all of the human MutL homologs may be independently targeted in patients with myositis. The clinical features of those myositis patients with antibodies directed against the DNA mismatch repair enzymes are summarized in Table 3.
In these studies, we demonstrated that autoantibodies to PMS1 occur in ∼7.5% of patients with autoimmune myositis, but not in patients with other systemic autoimmune diseases. The proportion of sera that are positive for PMS1 is significantly different between the 2 groups (P = 0.016). Interestingly, 2 patients with antibodies to PMS1 had dermatomyositis, 1 had polymyositis, and another had eosinophilic myositis with prominent inflammation of the fascia.
Although no unique clinical features are yet apparent in this group of patients, we propose that PMS1 be added to the list of myositis-specific autoantibodies (21). The association of unique autoantibodies with specific clinical states is a common feature in autoimmune myositis (19). For example, antibodies to Mi-2 are associated with typical rash in dermatomyositis, and antibodies to aminoacyl-tRNA synthetases are associated with interstitial lung disease, arthritis, Raynaud's phenomenon, and “mechanic's hands” (referred to as the “synthetase syndrome”). It is possible that unique clinical features associated with antibodies to PMS1 or the other DNA mismatch repair enzymes might yet be found as additional patients with these autoantibodies are identified. Of interest, the 4 patients with antibodies to PMS1 described here did not have antibodies to any of the aminoacyl-tRNA synthetases, although their sera did target several additional autoantigens active in DNA repair and nucleosome remodeling (see below).
In humans, the MutL family of proteins consists of several homologous ATPase proteins, including PMS1, PMS2, and MLH1. These molecules play an essential role in DNA mismatch repair, and mutations in genes encoding these proteins cause cancer susceptibility, as evidenced by the association of these mutations with hereditary nonpolyposis colorectal cancer (10, 22, 23). Interestingly, antibodies to PMS1, PMS2, and MLH-1 are all found infrequently in myositis patients. Due to the very low frequency of targeting of PMS2 and MLH1, it is not yet clear if these antibodies have specificity for autoimmune myositis. All MutL homologs share a very similar N-terminal domain of ∼40 kd, which contains the ATPase function (24) (Figure 5). The function of the C-terminal region of 30–50 kd, which has a very diverse sequence among the various family members, appears to be responsible for dimerization of MutL family members (25), as well as for interactions between MutL and DNA helicase II (26). It is of particular interest that this diverse C-terminal region, rather than the conserved N-terminal region containing the catalytic ATPase, is the target of autoantibodies.
A striking observation made with the myositis sera studied here is that, in addition to recognizing components of the mismatch repair pathway, several sera also recognize other defined autoantigens which function in DNA repair and nucleosome remodeling, including DNA-PKcs (double-strand break repair) (27), PARP (28), and Mi-2 (nucleosome remodeling and deacetylation) (8, 29). Interestingly, both the mismatch repair machinery and the Mi-2 deacetylase are methyl directed and include components which recognize and bind specifically to methylated DNA at methyl-CpG (8, 30). Thus, MLH1 interacts directly with MBD-4 (a methyl-CpG–binding protein) (31). Similarly, MBD-3 shares this highly conserved methyl-CpG–binding domain and directs the major histone deacetylase complex, containing Mi-2 (a nucleosome-stimulated ATPase belonging to the SWI2/SNF2 family) to effect dynamic changes in chromatin structure associated with transcription, replication, recombination, and repair (8, 32). It will be important to define whether the expression and activity of these processes is altered in a muscle-specific manner during initiation and propagation of these diseases.
All of the proteins described above (PMS1, PMS2, MLH1, DNA-PKcs, PARP, and Mi-2), as well as several aminoacyl-tRNA synthetases, SRP-72, U1–70 kd, and PM-Scl, are efficiently cleaved by granzyme B. This feature therefore unifies the majority of proteins recognized by autoantibodies in myositis and demonstrates that they contain a conserved structural motif recognized by the protease. We have previously proposed that the structural motif recognized by granzyme B has an important functional role: granzyme B evolved to cleave this motif, thereby regulating protein function (9). It is of interest that the subset of granzyme B substrates that are targeted in different systemic autoimmune diseases differ, and correlate well with the clinical phenotype. For example, topoisomerase I, centromeric protein B, and fibrillarin are targeted in scleroderma, while La is targeted in SLE and Sjögren's syndrome. This striking correlation between an immune response to specific groups of granzyme B substrates (functioning in unique pathways) and the biologic phenotype strongly implies that tissue- and event-specific biochemical events may play a role in selecting molecules for an autoimmune response. The role of granzyme B sites within these molecules, or their cleavage during cytotoxic lymphocyte–induced death, in this immune response is presently under investigation.
The myositis sera targeting DNA mismatch repair enzymes described here also recognize other substrates that are cleaved during apoptosis and by granzyme B. Many of these additional components are cleaved by caspases during apoptosis, and are cleaved at unique sites by granzyme B, generating novel fragments. Although the identities of these proteins remain unknown, many have molecular sizes of previously defined components of the human DNA repair and remodeling machinery. Since autoantibodies have traditionally been extremely powerful probes of basic biochemical and cell biologic processes (33), it is likely that further study of myositis autoantibodies will provide additional insights into the important processes of DNA repair and chromatin remodeling in humans.
The authors thank Nina Lu for help with analysis of data on the myositis cohort.