Induction of interferon-α by scleroderma sera containing autoantibodies to topoisomerase I: Association of higher interferon-α activity with lung fibrosis

Abstract

Objective

Peripheral blood cells (PBMCs) from some patients with systemic sclerosis (SSc) express an interferon-α (IFNα) signature. The aim of this study was to determine whether SSc patient sera could induce IFNα and whether IFNα induction was associated with specific autoantibodies and/or clinical features of the disease.

Methods

SSc sera containing autoantibodies against either topoisomerase I (anti–topo I; n = 12), nucleolar protein (ANoA; n = 12), or centromeric protein (ACA; n = 13) were cultured with a HeLa nuclear extract and normal PBMCs. In some experiments, different cell extracts or inhibitors of plasmacytoid dendritic cell (DC) activation, Fcγ receptor II (FcγRII), endocytosis, or nucleases were used. IFNα was measured by enzyme-linked immunosorbent assay.

Results

Topo I–containing sera induced significantly higher levels of IFNα as compared with all other groups. IFNα induction was inhibited by anti–blood dendritic cell antigen 2 (90%), anti-CD32 (76%), bafilomycin (99%), and RNase (82%). In contrast, ACAs induced low levels of IFNα even when necrotic, apoptotic, or demethylated extracts were used, despite the fact that CENP-B–binding oligonucleotide containing 2 CpG motifs effectively stimulated IFNα. IFNα production was significantly higher in patients with diffuse SSc (mean ± SEM 641 ± 174 pg/ml) than in those with limited SSc (215 ± 66 pg/ml) as well as in patients with lung fibrosis than in those without.

Conclusion

Autoantibody subsets in SSc sera differentially induce IFNα and may explain the IFNα signature observed in SSc. IFNα is induced by plasmacytoid DCs and required uptake of immune complexes through FcγRII, endosomal transport, and the presence of RNA, presumably for interaction with Toll-like receptor 7. The higher IFNα induction in sera from patients with diffuse SSc than in those with limited SSc as well as in sera from patients with lung fibrosis suggests that IFNα may contribute to tissue injury.

Systemic sclerosis (SSc) is a complex autoimmune disease characterized by inflammation, immune activation, excessive extracellular matrix deposition, fibrosis, and vascular obliteration in the skin and internal organs. SSc can be divided into 2 major subsets, limited cutaneous SSc (lcSSc) and diffuse cutaneous SSc (dcSSc), according to the extent to which the skin is affected. The lcSSc form is defined by the presence of skin thickening in areas solely distal to the elbows and knees, with or without involvement of the face. Patients with lcSSc present with various clinical features of CREST syndrome (calcinosis, Raynaud's phenomenon, esophageal dysmotility, sclerodactyly, and telangiectasias), the name by which lcSSc was previously known (1, 2). Patients with dcSSc have skin involvement proximal to the elbows and knees, tend to have earlier and more rapid involvement of the skin, and are at increased risk of cardiac disease, interstitial lung disease, renal crisis, and early death.

While the initial event leading to the development of SSc is still poorly understood, one characteristic feature of the disease is the production of a variety of autoantibodies to nuclear antigens. Specific autoantibodies associated with SSc are anticentromere antibodies (ACAs), anti–topoisomerase I antibodies (anti–topo I; also known as anti–Scl-70), and antinucleolar antibodies (ANoA) (3–5). ACAs are typically associated with lcSSc, whereas anti–topo I and ANoA are mostly observed in dcSSc, although some specificities, such as anti-Th/To, may be observed in lcSSc (3–5). The presence of these specific autoantibodies is further correlated with disease severity: ACAs have been associated with a reduced risk of pulmonary fibrosis, whereas anti–topo I have been associated with interstitial lung disease, more severe skin progression, and early death (3, 6, 7).

Although autoantibodies to (ribo)nucleoprotein particles have frequently been considered to be epiphenomena, it has recently been shown in systemic lupus erythematosus (SLE) that following binding to their cognate antigens and incubation with peripheral blood mononuclear cells (PBMCs) in vitro, these autoantibodies induce high concentrations of interferon-α (IFNα), a cytokine that is strongly implicated in disease pathogenesis (for review, see ref.8). Furthermore, evidence of type I IFN stimulation of PBMCs has recently been demonstrated by expression arrays in several other systemic autoimmune disorders (Sjögren's syndrome, polymyositis, and SSc) (9). In SSc, increased expression of type I IFN genes was found to be increased in PBMCs (10), and we recently reported similar findings in monocytes from SSc patients, as well as direct evidence of IFNα production in the skin (11). IFNα is a potent immune adjuvant that functions through activation of antigen-presenting cells as well as B and T lymphocytes (for review, see ref.12).

Since autoantibodies in SSc target a diverse population of nucleoprotein antigens that are associated with different clinical subsets of disease, we sought to determine whether differences in IFNα-inducing capacity were associated with different autoantibody populations and clinical manifestations in SSc. We observed that anti–topo I antibodies were more potent at inducing IFNα and that ACA-containing sera rarely did so.

PATIENTS AND METHODS

Patients.

Patients were diagnosed as having SSc according to the American College of Rheumatology (formerly, the American Rheumatism Association) criteria (13), and lcSSc or dcSSc was diagnosed according to the modified criteria of LeRoy et al (2, 14). Pulmonary artery hypertension was defined as a pulmonary artery systolic pressure of >35 mm Hg, as estimated by echocardiogram. Renal disease was defined as the presence of hypertension (either a diastolic pressure ≥85 mm Hg or a systolic pressure >145 mm Hg) in the presence of proteinuria. Lung fibrosis was determined by the presence of honeycombing or ground-glass appearance on high-resolution computed tomography images (15). Blood samples were obtained after the patients gave their informed consent.

Identification of autoantibodies.

Antinuclear antibodies (ANAs) were determined by indirect immunofluorescence on HEp-2 cells (Antibodies Inc., Davis, CA). Autoantibodies to topo I (Scl-70), Ro/SSA, La/SSB, Sm, and U1 RNP were determined by immunodiffusion against calf thymus extract (Inova Diagnostics, San Diego, CA), and antifibrillarin antibodies were determined by immunoprecipitation as described elsewhere (15). Titers of anti-Sm RNP and anti–topo I antibodies were also determined by enzyme-linked immunosorbent assays (ELISAs) using commercial antigens (Arotec Diagnostics, Wellington, New Zealand). In some cases, IgG was isolated from serum by protein A affinity chromatography.

Preparation and analysis of cell extracts.

HeLa cells were grown in Dulbecco's modified Eagle's medium (Hyclone, Logan, UT) containing L-glutamine (4 mM) and glucose (4.5 mg/ml), supplemented with 10% fetal bovine serum at 37°C. In selected experiments, the cells were grown in the presence of 5-aza-2′-deoxycytidine (5 μM; Sigma-Aldrich, St. Louis, MO) to induce demethylation of genomic DNA. Nuclear extracts were isolated from HeLa cells as described previously (16). Briefly, nuclei were extracted following dounce homogenization and centrifugation. Pelleted nuclei were resuspended in digestion buffer (Active Motif, Carlsbad, CA), and Enzymatic Shearing Cocktail (Active Motif) was added to digest the chromatin into 200–1,000-bp fragments using a proprietary enzyme. Smaller fragments of chromatin are presumably more readily accessible to autoantibodies and may facilitate entry of the immune complex into cells.

Apoptosis was induced by the addition of either camptothecin (5 μM; Sigma-Aldrich) or etoposide (50 μM; Sigma-Aldrich) for 24 hours. Necrotic cell extracts were prepared from the human monocyte line U937 by repeated cycles of freezing and thawing (17). Protein concentrations were determined by the Bradford assay (Bio-Rad, Hercules, CA), and the supernatants were stored at −70°C. Western blotting was performed as previously described (17), and extracts were analyzed by 1% agarose gel electrophoresis and stained with GelRed (Biotium, Hayward, CA).

Stimulation of PBMCs and synthesis of oligonucleotides (ODNs).

PBMCs were prepared from healthy human donors using Ficoll-Paque Plus (GE Healthcare, Piscataway, NJ). Cells were plated in 96-well round-bottomed plates at 5 × 10⁵/well with 500 units/ml of Universal Type I IFN (PBL Biomedical, Piscataway, NJ) and 2 ng/ml of granulocyte–macrophage colony-stimulating factor (Invitrogen, Carlsbad, CA) in RPMI 1640 (Hyclone) with supplements (18). After IFN priming for 4 hours, Toll-like receptor (TLR) agonists or test serum was added to PBMCs in the presence or absence of cell extracts, and supernatants were collected after 22–24 hours.

CpG ODN 2216 with a phosphorothioate (PT) backbone was synthesized by Invitrogen (sequence 5′-GGGGGACGATCGTCGGGGGG-3′) as the TLR-9 agonist. The following 21-bp sense and antisense sequences, which include the CENP-B box (19), were synthesized with either a phosphodiester (PDE) or a PT backbone: 5′-GCCTTCGTTGGAAACGGGATT-3′ (sense) 5′-AATCCCGTTTCCAACGAAGGC-3′ (antisense). Double-stranded DNA (dsDNA) was prepared from the sense and antisense DNA by heating to 95°C for 5 minutes, then cooling slowly to room temperature, and the slower-migrating duplex was verified by gel electrophoresis. Irrelevant PT single-stranded ODN control (sequence 5′-GCCGTCATTGGGATT-3′) and topo dsDNA constructed from the PT ODNs 5′-AAAAAGACTTAGAAAAATTTTT-3′ (sense) and 5′-AAAAATTTTTCTAAGTCTTTTT-3′ (antisense) were used as control ODNs in some experiments. Calf thymus DNA obtained from EMD Biosciences (San Diego, CA) and DOTAP liposomal transfection reagent obtained from Roche Applied Science (Indianapolis, IN) were used for transfection, as described previously (20).

Inhibitors of IFNα induction.

Briefly, anti-CD32 (AbD Serotec, Raleigh, NC), blood dendritic cell antigen 2 (BDCA-2; Miltenyi Biotec, Bergisch Gladbach, Germany), or mouse IgG1 isotype control (R&D Systems, Minneapolis, MN) was added to PBMCs 1 hour prior to the addition of test samples. In separate experiments, bafilomycin A1 (100 nM; Sigma-Aldrich) was added to PBMC cultures 1 hour prior to the addition of test samples. Cell extracts were treated with 8 μg/ml of DNase-free RNase (Roche, Basel, Switzerland) or 500 units/ml of RNase-free DNase (Sigma-Aldrich) for 3 hours at 37°C or for 15 minutes at room temperature, respectively, prior to incubation with PBMCs and test samples. Potential toxicity of the added components was monitored using the Rapid Cell Proliferation kit (EMD Biosciences), which measures the activity of cellular mitochondrial dehydrogenases.

Immunoassay for IFNα.

IFNα levels in supernatants were assessed using an in-house ELISA developed with commercially available antibodies and human IFNα standard (PBL Biomedical). Type I IFN used for priming PBMCs/plasmacytoid DCs did not react with the antibodies used for the ELISA. The sensitivity of the ELISA was 9 pg/ml.

Statistical analysis.

Data were first analyzed for normal distribution and were then compared by either parametric or nonparametric tests using SigmaStat software (SPSS, Chicago, IL). Correlation was performed by linear regression analysis.

RESULTS

Induction of IFNα by SSc sera containing anti–topo I autoantibodies, but not ACAs.

It was recently reported that PBMCs obtained from SSc patients demonstrate a messenger RNA (mRNA) expression “IFNα signature” (10), but the source of IFNα and the clinical significance of this observation remain uncertain. Since the IFNα mRNA expression signature in SLE is thought to be induced, at least in part, by autoantibody binding to nucleoprotein antigens and by stimulation of TLRs in plasmacytoid DCs, we sought to determine whether SSc autoantibodies could also induce IFNα following exposure to self antigens. Since autoantibody specificities in SSc are usually exclusive, in that anti–topo I antibodies, ACA, and ANoA occur infrequently in the same individual, we could also examine whether there were differences in IFNα induction between these 3 main specificities.

To determine whether autoantibodies in SSc sera had interferogenic activity, we performed serial 10-fold dilutions of well-characterized SSc sera and incubated them with and without a source of nuclear antigens in the presence of type I IFN–primed PBMCs. At a serum dilution of 10%, none of the sera from either healthy individuals or SSc patients produced amounts of IFNα that were >100 pg/ml in the absence of added antigen (data not shown). However, following the addition of a source of nuclear antigen, several ANoA-containing sera and most of the anti–topo I–containing sera induced the production of IFNα in cultures of PBMCs (Figure 1A). At a serum dilution of 1%, the difference in IFNα-inducing activity by anti–topo I–containing sera was even more apparent, with 11 of the 12 anti–topo I sera and 2 of the 9 ANoA sera, but none of the normal or ACA sera, inducing >218 pg/ml (mean + 2SD of the normal control value) (Figure 1B). Of interest, many sera induced higher amounts of IFNα at higher serum dilutions.

Figure 1. Stimulation of interferon-α (IFNα) production in peripheral blood mononuclear cells (PBMCs) in the presence of antigen by systemic sclerosis (SSc) patient sera or IgG containing anti–topoisomerase I (anti–topo I) antibodies. IFNα-primed PBMCs from healthy donors were incubated with SSc patient sera containing anticentromere antibodies (ACAs), antinucleolar antibodies (ANoA), or anti–topo I (A-Topo) at A, 10% volume/volume, B, 1% v/v, and C, 1% v/v and with 80 μg/ml of HeLa cell nuclear extract (NE), as described in Patients and Methods. Culture supernatants were collected after 22–24 hours, and the concentration of IFNα produced was quantified by enzyme-linked immunosorbent assay. Normal human serum (NHS) was used as a control. Bars show the mean. In C, IFNα induction was compared between serum and IgG (12.5 μg/ml) isolated from the same SSc patients. Results are representative of either >3 (A and B) or 2 (C) experiments. In A and B, P < 0.001 by Kruskal-Wallis analysis of variance, and in B, P < 0.05 by Dunn's pairwise multiple comparison for anti–topo I compared with each of the other specificities.

In a small subanalysis, antifibrillarin antibody–positive and –negative ANoA-containing sera did not induce significantly different levels of IFNα (mean ± SEM 247 ± 24 pg/ml in 8 antifibrillarin-positive sera and 291 ± 53 pg/ml in 6 antifibrillarin-negative sera).

To ensure that the results observed were due to immune complexes rather than other serum components, IgG was isolated from serum and tested side-by-side with diluted whole serum. As shown in Figure 1C, IFNα production was similar in serum and dose-adjusted IgG from the same individual patients. These results indicate that most anti–topo I sera, ∼25% of the ANoA sera, but none of the ACA sera generated IFNα in this in vitro bioassay. It should be noted that these values were generally of lower magnitude than those seen in SLE patient sera containing anti-Sm/RNP antibodies (see control in Figure 2 below) and that differences in the absolute amounts of IFNα induced by different PBMC donors were observed.

Figure 2. Generation of interferon-α (IFNα) by serum autoantibodies from systemic sclerosis (SSc) patients by plasmacytoid dendritic cells and mediation through Fcγ receptor II (FcγRII) engagement and endosomal trafficking of predominantly RNP antigens. A and B, Type I IFN–primed peripheral blood mononuclear cells (PBMCs) were preincubated for 1 hour with anti–blood dendritic cell antigen 2 (anti–BDCA-2; 0.2 μg/ml) (A), anti-CD32 (FcγRII; 5 μg/ml) (B), or mouse isotype control monoclonal antibody prior to the addition of SSc sera (1%) containing anti–topoisomerase I (anti–topo I) or normal human sera (NHS) and antigen. C, PBMCs were preincubated for 1 hour with bafilomycin A1 (Bafilo; 100 nM) before incubation with SSc sera (1%) containing anti–topo I or NHS and antigen. D, Nuclear extract was pretreated with RNase (8 μg/ml) for 3 hours at 37°C, with DNase (500 units/ml) for 15 minutes at room temperature, or with both prior to addition with SSc sera containing anti–topo I or antinucleolar antigen (ANoA) or NHS to PBMCs. IFNα was quantified by enzyme-linked immunosorbent assay. CpG 2216 stimulatory oligonucleotide (C) and a high-titer anti-Sm/RNP–containing systemic lupus erythematosus (SLE) patient serum (C and D) were used as positive controls. Results are representative of 2–3 independent experiments.

Although antibodies in SSc are generally exclusive (4), it is not possible to be certain that the IFNα induction in anti–topo I–positive serum was due to anti–topo I antibody. Whereas gel diffusion analyses did not reveal other common autoantibodies (results not shown), all samples were rescreened for anti-Sm/RNP by ELISA, and only 2 were found to be positive. When IgG anti–topo I levels and IFNα induction by ELISA were tested for a statistical correlation, a highly significant correlation (r = 0.678, P = 0.015, by Pearson's product-moment correlation; n = 12) was observed, which suggests, but does not prove, a direct involvement of anti–topo I in the induction of IFNα.

Dependence of IFNα induction by SSc autoantibodies on plasmacytoid DC activation, FcγRII engagement, endosomal transport, and the presence of RNA.

Activation of both TLR-9 (by DNA) and TLR-7 (by RNA) has been implicated in the stimulation of IFNα production by immune complexes in SLE, although recent studies in both human and murine lupus suggest that RNP complexes may be more relevant (21, 22). Following engagement by FcγRII, the immune complexes are taken up into endosomes and presumably engage TLRs in an endosomal/lysosomal compartment (9). Plasmacytoid DCs constitute <1% of total PBMCs, yet are by far the highest producers of type I IFN. To determine whether immune complexes formed following incubation of SSc serum with antigen-generated IFNα by plasmacytoid DCs, we used an agonistic monoclonal antibody (mAb) that binds to BDCA-2, a C-type lectin receptor that is present exclusively on plasmacytoid DCs and, upon ligation, suppresses IFNα production (23). Anti–BDCA-2 mAb, but not the isotype control mAb, almost completely inhibited IFNα production (Figure 2A), indicating that plasmacytoid DCs were the source of IFNα induced by anti–topo I antibodies.

To determine whether FcγRII (CD32a), which is known to be expressed by plasmacytoid DCs (24), was required for the induction of IFNα by SSc sera, a blocking antibody (clone AT10) was used. When preincubated with PBMCs, the anti-CD32 antibody inhibited IFNα production by SSc sera by an average of 76%, whereas the isotype control had no inhibitory effect (Figure 2B). To determine whether immune complexes entered a lysosomal pathway within the cell, PBMCs were incubated with bafilomycin A1, an inhibitor of the endosomal proton–translocating ATPase (25). As shown in Figure 2C, bafilomycin almost completely inhibited the induction of IFNα by SSc sera in this assay. This was not explained by toxicity, since a mitochondrial assay revealed no changes in activity (data not shown).

Intracellular TLRs are activated by nucleic acids. When cell extracts were preincubated with DNase, RNase, or the combination, we observed that both DNase and RNase reduced IFNα production (average of 48% and 82%, respectively) and the combination of enzymes reduced IFNα production to near baseline levels (91%) in anti–topo I–containing sera (Figure 2D). In contrast, RNase, but not DNase, attenuated IFNα production by an ANoA-containing serum sample. Anti–platelet-derived growth factor (anti-PDGF) antibodies that are present in SSc sera have been reported to induce reactive oxygen species and to activate a Ras MAPK pathway in transfected cell lines as well as normal fibroblasts (26). Since PDGF receptors are expressed at low levels on normal, unstimulated monocytes, it is unlikely that they influenced IFNα in the in vitro bioassay.

Lack of stimulation of IFNα after incubation of ACAs with cell extracts following apoptosis or necrosis.

We previously raised the hypothesis that all autoantibodies in systemic autoimmune diseases are stimulated through the immune complex activation of intracellular TLRs and T cell presentation of peptides derived from the protein component, which is known as the “Toll hypothesis” (27). While the production of IFNα following incubation of anti–topo I–containing SSc sera and some ANoA-containing SSc sera is consistent with this hypothesis, the lack of stimulation by ACAs argues against this hypothesis as a universal explanation for the generation of high-titer IgG autoantibodies. ACAs target the proteins CENP-A, CENP-B, and CENP-C, which bind to repetitive satellite DNA elements in the centromere. While the centromeric proteins can be detected in nuclear extracts (28), the signals that promote binding and release from the centromere are not known, although DNA methylation may play a role (19). More significantly, it is not known how α-satellite DNA and corresponding RNA may also be released following a cleavage event, for example during normal or abnormal cell death.

We therefore generated necrotic, apoptotic, and DNA-demethylated extracts and combinations thereof and examined the protein and DNA composition of these extracts. Western blot analysis confirmed the presence of the dominant autoantigens (28), CENP-B (M_r 80) (Figure 3A) and CENP-A (M_r 17) (results not shown) in the extracts that were generated by methods that included DNA digestion or shearing. Agarose gel electrophoresis also verified DNA release as either a “DNA smear” or a nucleosomal ladder from necrotic or apoptotic extracts, respectively (Figure 3B). Despite the presence of protein antigens and fragmented DNA in the extracts, no increase in IFNα-inducing activity was observed in these extracts (Figure 3C).

Figure 3. Lack of interferon-α (IFNα) induction by anticentromere antibodies (ACAs), as determined with extracts from apoptotic, necrotic, or DNA-demethylated cells. HeLa cells were cultured in conventional medium or in medium containing 5 μM 5-aza-2′-deoxycytidine (5AzdC), a demethylating agent. Cells were induced to undergo apoptosis by addition of 50 μM etoposide (ETP) or 5 μM camptothecin (CPT) or to undergo necrosis by repeated cycles of freezing and thawing. A, Western blots with normal human serum (NHS), a representative ACA-containing serum, and a representative anti–topoisomerase I (A-Topo)–containing serum. The blot of the ACA-containing serum shows CENP-B at ∼80 kd in nuclear extract from cells grown in conventional medium (M) and in extract from cells previously incubated with 5AzdC (5Az). B, Agarose gel electrophoresis of HeLa cell nuclear extract following a standard isolation procedure as described Patients and Methods (lanes 2 and 5), a standard procedure followed by brief sonication (lanes 3 and 6), and necrosis (lanes 4 and 7). Total extracts (lanes 2–4) and purified DNA (lanes 5–7) were loaded, and the gel was stained with GelRed (for nucleic acids). Note that the nucleosome ladders in lanes 5 and 6 are due to the use of a “shearing enzyme” during preparation of the nuclear extract (see Patients and Methods). Molecular weight markers are shown in lane 1. C, IFNα induction after incubation of 7 ACA-positive sera and an NHS (all at 1%) with apoptotic cell extracts induced by 5AzdC, CPT, or ETP or with nuclear extract (NE) alone. IFNα induction was determined by enzyme-linked immunosorbent assay.

Presence of CpG motifs with IFNα-inducing capacity in the CENP-B box.

It is of considerable interest that the CENP-B box contains CpG-rich sequences (Figure 4A) that, especially following demethylation (29), would be predicted to stimulate IFNα production. We therefore synthesized the sense and antisense 21-mer ODNs containing the CpG motifs with either a PDE backbone or a DNase-resistant PT backbone. These single-stranded ODNs as well as the double-stranded forms were first examined for their ability to generate IFNα production following incubation with PBMCs or following transfection into these cells with DOTAP. When transfected, both PDE ODNs and PT ODNs were potent inducers of IFNα, with the exception of the PDE dsDNA. In contrast, when the ODNs were incubated with PBMCs in the absence of a transfection reagent, the sense and double-stranded PT, but not the PDE ODNs, induced IFNα, although at not as high a level as with the potent activator CpG 2216 (Figure 4B). These findings indicate that the CpG-containing region of the CENP-B box has strong IFNα-inducing capacity but that it requires protection from DNase degradation as well as intracellular transport in order to exert this effect.

Figure 4. Presence of CpG motifs with interferon-α (IFNα)–inducing capacity in the CENP-B box. A, The 21-bp sense sequence containing the CENP-B box. The α-satellite DNA contains a CpG-rich (boxed areas) CENP-B box to which CENP-B binds. B, Induction of IFNα in peripheral blood mononuclear cells (PBMCs) after incubation with oligonucleotides (ODNs) containing CpG motifs with a phosphodiester (PDE) or a phosphorothioate (PT) backbone. Sense (S) and antisense (AS) ODNs containing either PT or PDE backbones were synthesized, and 1 μM of each was incubated with primed PBMCs as in Figure 1. The ODNs were also used to create double-stranded (ds) DNA, and both single-stranded and double-stranded DNA were transfected into PBMCs with the lipid reagent DOTAP, as described in Patients and Methods. IFNα was quantified by enzyme-linked immunosorbent assay. Controls used in this experiment were calf thymus DNA (CT; stimulatory), IRS (ODN control), topoisomerase double-stranded PT DNA (dsTopo), and CpG 2216 ODN (CpG). Calf thymus DNA is known to induce IFNα following transfection into cells (20). S (sense) and AS (antisense) are single-stranded and ds is double-stranded DNA.

Demethylation of DNA not only enhances its IFNα-inducing capacity, but it also leads to increased binding of CENP-B to the CENP-B box (19). However, when ACA-containing sera were incubated with an extract obtained from cells that had been incubated with 5-aza-2′-deoxycytidine or were induced to undergo apoptosis, no increase in IFNα was observed (Figure 3C). Thus, although CENP-B contains immunostimulatory CpG sequences that induce high concentrations of IFNα in vitro, we have been unable to generate IFNα with ACA-containing sera from patients with SSc under a multitude of different conditions.

Clinical associations of IFNα-inducing autoantibodies.

Anti–topo I antibodies are generally associated with diffuse SSc, but some patients with ANoA or ACAs may also develop systemic features of the disease (3). IFNα production was significantly higher in dcSSc sera than in lcSSc sera (mean ± SEM 641 ± 174 pg/ml versus 215 ± 66 pg/ml; P = 0.03). We also compared sera from 18 patients with lung fibrosis (10 with anti–topo I antibodies, 4 with ANoA, and 4 with ACAs) with sera from 13 patients without lung fibrosis (3 with anti–topo I antibodies, 5 with ANoA, and 5 with ACAs) for their IFNα-inducing activity. Sera from patients with lung fibrosis induced a median of 294 pg/ml of IFNα, whereas sera from patients without lung fibrosis induced a median of 115 pg/ml of IFNα (P = 0.033) (Table 1). Thus, there is an association between lung fibrosis and IFNα-inducing activity by patient sera. Although there was a trend toward shorter disease duration in the group without lung fibrosis, this difference was not statistically significant. Too few patients with renal disease (n = 1) and pulmonary hypertension (n = 3) were available for statistical comparisons to be made (Table 1). There was no association between the disease duration and the IFNα-inducing capacity of serum (R² = 0.03, P > 0.05).

Table 1. Clinical features of SSc patients with and those without lung fibrosis*

	SSc patients with lung fibrosis (n = 18)	SSc patients without lung fibrosis (n = 13)
* See Patients and Methods for definitions of renal disease and pulmonary hypertension. Disease duration was scored on a scale of 0–3, where a duration of 1–5 years = 1, a duration of 5–10 years = 2, and a duration of >10 years = 3. SSc = systemic sclerosis; IFNα = interferon-α. † P = 0.033 versus patients without lung fibrosis. Between-group differences in the other characteristics were not statistically significant.
No. (%) with renal disease	1 (6)	0
No. (%) with pulmonary hypertension	1 (6)	2 (15)
Median level of IFNα induced, pg/ml	294†	115
Disease duration score	1.76	1.30

DISCUSSION

The main findings of this study are that SSc sera with autoantibody specificities associated with diffuse or systemic, but not limited scleroderma, had IFNα-inducing properties. As observed for autoantibodies in SLE (9), IFNα induction required ingestion through FcγR, endosomal transport, and stimulation of plasmacytoid DCs. These findings could, in part, explain the IFNα signature found in many patients with SSc (10), the increase in SIGLEC1, an IFN-induced gene on monocytes from SSc patients (30), as well as the IFNα mRNA detected in the skin of some SSc patients (31). Since evidence of immunoglobulin secretion by B cells has also been found in skin samples obtained from patients with SSc (31), local production of autoantibodies may contribute to immune complex–mediated cytokine production.

The highest IFNα-inducing activity was observed in anti–topo I–containing sera. Anti–topo I antibodies are found in ∼20% of patients with SSc and are much more frequent in those with dcSSc. These antibodies bind to topoisomerase I, an abundant nuclear protein that catalyzes DNA relaxation and SR protein phosphorylation (32). In addition to its well-known role in preventing supercoiling during DNA synthesis, topo I is also physically associated with 36 nuclear proteins, several of which are involved in either RNA splicing, processing, or metabolism (33). This observation likely explains the greater inhibitory effect of RNase compared with DNase on IFNα induction, although the combination of both enzymes reduced IFNα to baseline levels. We confirmed that IFNα induction was produced by the IgG fraction, rather than by other serum components, and there was a strong correlation between anti–topo I levels and IFNα production. However, we cannot formally exclude the possibility that other autoantibodies in the anti–topo I–containing sera contributed to IFNα stimulation.

ACAs are observed in 20–30% of SSc patients and bind 1 or more of the 6 centromeric proteins, CENP-A through CENP-F (3). The 80-kd CENP-B is the most common autoantigen of the 3 centromere proteins, and it binds to an α-satellite DNA that comprises tandem repeat units of 171 bp and functions as a cis element for centromere-specific nucleosome assembly (34, 35). Intriguingly, CENP-B binding within the α-satellite DNA is relatively specific for a 17-bp sequence called the CENP-B box, which contains 2 CpG dinucleotides (34). CENP-C also binds to α-satellite DNA (36) and has recently been shown to be a dual DNA/RNA binding protein that is implicated in assembly of the nucleolus during mitosis (37). In striking contrast to the anti–topo I–containing sera, sera containing ACAs rarely induced IFNα (or tumor necrosis factor α [results not shown]). Immunoblotting with ACAs confirmed the presence of CENP antigens in the soluble HeLa nuclear extracts; this was likely due to inclusion of a micrococcal nuclease-like enzyme during preparation of the extract. To further mimic situations in vivo, such as apoptosis or necrosis, that might allow the release of nucleosomes or random DNA fragments, respectively, cells were induced to undergo cell death by a variety of manipulations, and their extracts were tested for stimulatory activity with ACAs. No increased IFNα-inducing activity was observed.

The CENP-B box contains CpG motifs that may stimulate the production of IFNα by plasmacytoid DCs (for review, see ref.38). We synthesized a 21-mer ODN to which CENP-B is known to bind and tested both the naturally occurring PDE-linked ODN as well as the relatively DNase-resistant ODN with a PT backbone for their IFNα-inducing capacity. The sense PT–containing dsDNA induced IFNα without the requirement for a transfection reagent. Intriguingly, the PT dsDNA was much more stimulatory than the PDE dsDNA following transfection, suggesting that intracellular DNase is crucial for protection from IFNα stimulation. Since demethylation promotes the binding of CENP-B to the CENP-B box (19) and would also be expected to enhance the stimulatory activity of CpG-containing DNA, we treated cultures with a demethylating agent, but we were unable to show induction of IFNα with ACAs. Thus, although the CpG motifs within CENP-B–binding DNA can stimulate IFNα, especially when protected from DNase activity, we have been unable to reproduce in vitro a condition that stimulates IFNα production.

The potent stimulatory effects of IFNα on immune function are now well described and include enhanced antigen presentation, homing to lymphoid organs, activation of immune cells, and stimulation of a Th1 response (for review, see ref.12). Type I IFNs also stimulate the development of CD8+ T cells, and these cells have been implicated in pulmonary injury in patients with SSc (39). In addition, type I IFNs decrease the threshold for activation of B cells via the B cell receptor and enhance differentiation, antibody production, and immunoglobulin isotype class switching. This positive feedback loop may contribute to the relentless course of disease seen in many patients with dcSSc. It should also be noted that IFNα induction by SSc sera was not as high as that observed in SLE sera and that variations in the levels of IFNα induced by different PBMC donors was also observed. The source of variation is currently under study.

A second mechanism by which immune complex–mediated IFNα induction may cause pathologic changes is through vascular injury, a key component of SSc (40, 41). It has been suggested that vasospasm-induced ischemia (42) or antiendothelial autoantibodies (43) lead to apoptosis, thereby providing a source of nucleoprotein-containing antigens. Administration of IFNα has been associated with the temporal development of Raynaud's phenomenon (44) as well as SSc (45). IFNα is known to have an antiangiogenic effect, and it retards wound healing (46, 47). The antiangiogenic effects have previously been explained by IFNα-mediated antiproliferative activity or induction of apoptosis of endothelial cells, whereas more recent studies have shown that after prolonged exposure, IFNα induces replicative senescence of endothelial cells (48). Many other factors have been implicated in vascular injury in SSc, but our results should encourage further studies to determine whether prolonged local exposure of vessels to IFNα may be a contributory factor to the loss of small vessels that is characteristic of this disease.

We also observed higher IFNα-inducing activity in patients with lung fibrosis as compared with those without lung fibrosis. This observation is consistent with the higher frequency of anti–topo I–containing sera in the group with lung fibrosis (55%) as compared with the group without lung fibrosis (23%) (3). The association between IFNα-inducing activity and lung fibrosis is similar to the recently reported association between Jo-1 and lung fibrosis observed in polymyositis patients (49) and further supports a role of IFNα in the promotion of fibrosis. It is of considerable interest that a trend toward a decline in forced vital capacity and a reduction in diffusing capacity for carbon monoxide was observed in SSc patients receiving treatment with IFNα (50). Whether fibrosis is the end result of a vasculopathic effect, as discussed above, an exuberant immune response, or has some as-yet-unknown effect on profibrotic factors remains to be determined. Despite the fact that IFNα-inducing activity was higher in patients with lung fibrosis, sera from some patients with lung fibrosis did not induce IFNα in the bioassay, which suggests that other factors also contribute to lung fibrosis. Immune complexes may also engage FcγR on other cell types, releasing inflammatory cytokines.

We conclude that SSc sera containing anti–topo I induce IFNα by plasmacytoid DCs in vitro and likely contribute to IFNα production, as recently observed in ex vivo blood and tissue samples obtained from patients with SSc. The ability to induce IFNα is positively correlated with clinical disease, especially lung fibrosis. These results shed light on the immunologic differences between dcSSc and lcSSc and help to elucidate the molecular mechanisms responsible for the pathogenesis of SSc and other autoimmune diseases characterized by antibodies directed against autoantigens associated with nucleic acids. The nature and availability of antigen and the relationship between IFNα and fibrogenic stimuli, particularly in the lung, are key questions to address in future studies.

AUTHOR CONTRIBUTIONS

Dr. Elkon had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Kim, Peck, Santer, Patole, Schwartz, Elkon.

Acquisition of data. Kim, Peck, Patole, Molitor, Arnett.

Analysis and interpretation of data. Kim, Peck, Santer, Patole, Schwartz, Molitor, Arnett, Elkon.

Manuscript preparation. Kim, Peck, Santer, Schwartz, Molitor, Arnett, Elkon.

Statistical analysis. Kim, Peck, Elkon.

Acknowledgements

We thank Lars Ronnblom, Maija Eloranta, William Earnshaw, and Andy Choo for helpful discussions.