Mycoplasma infection induces a scleroderma-like centrosome autoantibody response in mice

Abstract

Development of autoantibodies to intracellular molecules is a universal feature of autoimmune diseases and parallels onset of chronic inflammatory pathology. Initiating antigens of disease-specific autoantibody responses are unknown. We previously showed that the major targets of autoantibodies in scleroderma are centrosomes, organelles involved in mitotic spindle organization. Here we show that centrosome autoantibodies are induced in mice by mycoplasma infection. The centrosome-specific antibody response involves class switching of preexisting IgM to IgG isotypes, suggesting a T cell-dependent mechanism. The antibody response spreads to include additional intracellular targets, with newly recruited autoantibody specificities arising as IgM isotypes. Antibiotic treatment of mice prevents autoantibody development. Centrosome autoantibodies may provide an aetiological link between infection and human autoimmunity and suggest novel therapeutic strategies in these disorders.

INTRODUCTION

Antibodies to self antigens are an early feature of most human autoimmune diseases [1,2]. However, little is known about events that initiate autoantibody development, mechanisms that sustain their production, and how specific organelles are targeted. Infectious agents have been implicated in the development of autoantibodies [3]. Infections can induce or exacerbate chronic autoimmune syndromes in humans and in animal models of human disease [4,5]. Autoimmunity is thought to result from the initial loss of tolerance to a single self-antigen, and the immune response amplifies to target multiple antigen molecules [6]. Identifying the antigen that initiated the autoimmune attack is of critical importance in understanding the early events that led to disease. While the induction of autoantibodies characteristic of some systemic autoimmune diseases correlates with natural [7] or experimental infection [8–10] with viruses, bacteria or parasites, the self-antigens first targeted during aberrant responses to nonself have not been identified.

Several mechanisms have been proposed to link infection to the development of autoantibodies. In the molecular mimicry model, infection induces antibodies to peptide antigens on pathogens that share homology to self-peptides [11,12]. Antibodies of IgM isotype, cross-reactive to self and nonself antigens, are readily produced by the peripheral B cell repertoire in normal individuals [13,14]. Upon recruitment in autoimmune responses, B cells of the same specificities aquire class switched, high affinity antigen receptors [15,16]. A definitive link between infection and autoimmunity via cross-reactive self and nonself epitopes has yet to be established [17]. In an alternative model, infection induces polyclonal activation of self-reactive B cells [10]. Infection can induce polyclonal autoantibody responses with fine specificities that are similar to those of human and murine lupus, but does not account for autoantibody specificities that develop characteristically in most of the remaining systemic autoimmune diseases [8–10].

Scleroderma is a human systemic autoimmune disease characterized by chronic inflammation of the skin and visceral organs, and development of autoantibodies that orientate diagnosis [18]. There is little evidence to support a role for infectious agents in the pathogenesis or autoantibody development in scleroderma. Some self-epitopes recognized by patient antibodies share homology to retrovirus or herpesvirus antigens, consistent with a role for molecular mimicry in autoantibody development [19,20]. These cross-reactive viral epitopes have not been shown to be immunodominant during responses to viral infection nor have they been shown to initiate an immune response to self-antigens during scleroderma pathogenesis. One limitation to this type of analysis is the lack of experimental systems to test the role of infection in the initiation of antibody responses to scleroderma autoantigens.

A subset of autoantibodies in scleroderma target centrosomes, organelles that organize microtubule spindles for chromosome segregation [21,22]. We recently found that centrosomes are major autoantibody targets in scleroderma [23]. This finding was striking given the low frequency of centrosome autoantibodies previously reported in autoimmune diseases [21]. The use of more sensitive and specific centrosome assays allowed unambiguous identification of centrosomes and suggested that these organelles may have important diagnostic potential in scleroderma. In this study, we show that centrosome-specific autoantibodies are induced when naive mice are infected experimentally with mycoplasma. The initial autoantibody response appears to be specific for centrosome proteins and subsequently spreads to include antigens on other cellular structures. The anticentrosome response in mycoplasma-infected mice can be prevented by pretreatment with antibiotics. These results demonstrate loss of immune tolerance to centrosomal protein after mycoplasma infection and suggest that centrosome autoantibodies may link infection with the development of autoimmunity.

MATERIALS AND METHODS

Mouse rearing, infection and antibiotic treatment

Balb/c.SvJ, C57BL/6 J and 129/J mice were purchased from the Jackson Laboratories (Bar Harbor, Maine, USA), Charles River Laboratories (Wilmington, MA, USA) and Taconic Farms (Germantown, NY, USA) and housed pathogen-free or in biocontainment (UMass Medical School or University of Connecticut Health Center). Six week old male or female mice were prebled by venous puncture, anaesthetized (1 mg/mouse Nembutal, Abbott Laboratories, Chicago, IL, USA) and infected intranasally (or orally, intravenously, intraperitoneally) with 10³−10⁹ pfu (plaque-forming units)/ml of 4x plaque-purified inoculum of mycoplasma, that has been renografin gradient-enriched (Squibb Diagnostics, New Brunswick, NJ, USA) or sedimented on 70% sucrose in PBS. Control mice received supernatant from uninfected cultures, that was renografin-enriched or sedimented on a sucrose cushion. Balb/c.SvJ mice were used in all experiments; similar results were observed in C57BL/6 J and 129/J mice. Ciprofloxacin (20 µg/g body weight) was delivered orally by gavage once per day.

Antibodies

In this study we used a monoclonal antibody to Mycoplasma hyorhinis variable lipoprotein (gift of Dr Wise, University of Missouri), and affinity-purified polyclonal antibodies to Mycoplasma hyorhinis (Cortex Biochem, San Leandro, CA, USA), pericentrin [24], calnexin (Stressgen, Victoria, BC Canada) and BiP (BD Biosciences, Franklin Lakes, NJ, USA).

Mycoplasma stocks

Porcine Mycoplasma hyorhinis was obtained from ATCC (Rockville, MD, USA) and grown as indicated (Mycoplasma 243 media, ATTC). Our laboratory stock was produced as follows. Supernatants from mycoplasma-infected Vero cells were harvested 7 days after inoculation, spun (1075× g) at 4°C, 20 min (minutes) and loaded into tubes (25 × 89 mm, Beckman Instruments, Palo Alto, CA) over 70% renografin in PBS. Tubes were centrifuged (100 000 g, 4°C, 90 min) and the renografin–media interface was exposed to a 20–70% continuous renografin gradient as above for 12 h. Refractive indices of fractions was determined and mycoplasma stocks were maintained in sucrose, renografin or DMEM at −80°C.

Cell lines

We used mycoplasma-free NIH3T3, SP2, Vero and 293T cells (ATCC). Cells were propagated in DMEM with 100 U/ml of penicillin G, 100 µg/ml streptomycin sulphate, 2 mM l-glutamine and 10% heat inactivated (56°C, 30 min), mycoplasma-free fetal bovine serum (Gibco, Grand Island, NY, USA). All tissue culture reagents were tested for mycoplasma contamination (Stratagene, La Jolla, CA, USA).

Plaque assay

Mouse spleens in 1 ml of cold DMEM were homogenized and used to inoculate Vero monolayers in Eagles MEM (BioWhittacker, Walkersville, MD, USA supplemented as above) for 90 min, 37°C. Monolayers were overlayed EMEM with 0·5% agarose at 37°C (SeaKem, BMA Rockland, Maine) and six days later 2 ml of 0·5% agarose in EMEM containing 0·3 mg Neutral Red (Sigma, St. Louis, MO, USA) was added. Plaques were counted one day later. Ciprofloxacin (0·5 µg/ml, ICN Biomedicals, Irvine, CA, USA) or doxicycline (1 µg/ml, Sigma) were added to cultures for 1–7 days and supernatants examined by plaque assay.

Cloning, PCR, sequencing and sequence alignments

Genomic DNA extracted from gradient-purified supernatant of Vero cell cultures inoculated with spleen-derived murine mycoplasma, or Mycoplasma hyorhinis (ATCC) was cleaved with EcoRI or HindIII (N.E.B., Beverly, MA, USA), gel-purified, cloned into pUC18 vector, purified, sequenced and used to search NCBI databases for homologous sequences using BLASTX (http://www.ncbi.nlm.nih.gov/blast). Primers designed to gene sequences of highest homology were used to PCR-amplify sequences from mycoplasmas. Primer sequences: 16 s, 5′-GGTT AAGTCCTGCAACGAGC-3′ and 5′-GTTAACTCACCGACTT TGGG-3′; tuf, 5′-GGCTTGGTGCTGCTCAAATGGA-3′ and 5′-CCTACAGTTCTACCACCTTCACGG-3′; methylase, 5′-GA TAATACAAGAAGTGGTTTATTGC- 3′ and 5′-AAAACTTT CCAACTCGAGTT-ATATCC-3′; dehydrogenase, 5′-TGAAGA AACTTTAGATGTTTCAACAACTCC-3′ and 5′-TCCTGTTG ATTTTTCTACATTC-3′; permease, 5′-CCAGTTTTTGTAGAT ATTAAAGAAATCG-3′ and 5′-CTGTAGCTGCAAAAAAT CC-3′; tpi, 5′-ATTTGGATTTTGCAATTGC-3′ and 5′-TTTTCT TGCGAAACTGAGCCAACC-3′.

PCR reactions were performed in a laminar flow PCR hood (AirClean, Raleigh, NC, USA) using 50 pmol primer, 5 U of HotStarTaq polymerase (Qiagen Inc., Valencia, CA, USA), 100 mm of each deoxynucleotide triphosphate, in 67 mm Tris buffer (pH 8·8), 4 mmµgCl2, 16 mm (NH₄)₂SO₄, 10 mm 2-mercaptoethanol and 100 µg/ml bovine serum albumin (BSA). PCR cycles (95°C, 5 min × 1, 94°C, 1 min, 50°C, 2 min, 72°C, 1 min × 40 and 72°C, 5 min × 1) were carried out in an MJ Thermocycler (MJ Research, Watertown, MA, USA).

Immunoblotting

Immunoblotting was performed as described previously [23] using a multiwell miniblotter (Integrated Separation Systems, MA, USA).

Immunofluorescence

Immunofluorescence assays were performed as described [23] using NIH/3T3 or Vero cells fixed in methanol at −20°C or in 4% paraformaldehide (Electron Microscopy Science, Ft. Washington, PA, USA). Primary antisera were detected with fluorophore-tagged secondary antibodies to IgG (H + L), IgM, Fcγ (IgG isotype) (Jackson ImmunoResearch Laboratories, West Grove, PA, USA), IgG1, IgG2a, IgG2b (Pharmingen, San Diego, CA, USA) and IgG1 and IgG2b (Zymed, San Francisco, CA, USA). Slides were examined on an Axiophot microscope (Zeiss, Germany).

Electron microscopy

Electron microscopy was performed as described [24] using monolayers or mycoplasma fractions fixed in 4% paraformaldehyde (Electron Microscopy Science) in PBS at RT, 30 min and thin sections were viewed in a JOEL electron microscope.

Statistical analysis

Statistical analysis was performed using Epi Info 6·1 software and the Mann–Whitney test. Student t-test, two-tailed and unequal variances were also used.

RESULTS

We recently showed that centrosome autoantibodies represent the most prevalent autoreactivity in human scleroderma [23]. This makes them an important diagnostic indicator and suggests that they may play a role in disease pathogenesis [25]. To understand more about centrosome autoantibody induction, we examined the phenomenon in a murine model system. Using independent morphological and biochemical assays to detect centrosomes and centrosome autoantigens [23], we observed sporadic centrosome autoreactivity in several academic and commercial mouse colonies. Up to 35% (n = 55/154) of animals in some colonies had centrosome autoreactivity while autoreactivity was undetectable in other colonies (n = 0/22).

We hypothesized that the sporadic occurrence of centrosome autoreactivity in mice, with restriction to some colonies but absent in others, was consistent with autoantibodies being induced following cryptic infection. To test this hypothesis, we asked whether autoantibody reactivity could be initiated through contact with infected animals. Surprisingly, naive mice housed with sero-positive but not sero-negative mice rapidly developed a specific autoantibody response to centrosomes and centrosome antigens (Fig. 1).

Fig. 1.

Centrosome-specific autoantibodies develop in naïve mice following cohabitation with autoantibody-positive mice. (a) Immunoblots showing the recombinant centrosome proteins pericentrin (Pc) and centriolin (Cen), probed with sera from two mice (a,b) prior to exposure to centrosome-positive mice (0) and after two weeks of exposure (2). Strong autoreactivity develops to both centrosome proteins in both mice at 2 weeks. Each triplet shown under slanting bars represents 10-fold dilutions of sera. This result is representative of 2 experiments. (b,c) Immunofluorescence detection of centrosomes using sera from infected mice (b–b′) and uninfected mice (c–c′). Centrosomes in b and c were detected with pericentrin antibodies.

Characteristics of the infectious agent that induces anticentrosome autoantibodies

To isolate the putative infectious agent responsible for centrosome autoantibody development, we prepared suspension homogenates from mouse spleens and used them to inoculate cultured mammalian cells. Plaques formed in cell monolayers inoculated with homogenates from mice with centrosome autoantibodies, but not from centrosome autoantibody-negative animals. The infectious agent was plaque-purified and banded on renografin density gradients (Fig. 2a–c).

Fig. 2.

Isolation of an infectious agent from centrosome-autoantibody positive mice. (a,b) Plaques formed by infectious agent on monolayers of Vero cells. Inoculum used for this experiment was derived from plaques that formed following incubation with spleen cell preparations derived from autoantibody–positive mice (a) or from cell monolayers without plaques (b). (c) Graph showing banding pattern of infectious agent from culture supernatants on a renografin density gradient (at 1·19–1·20 g/cm³). Titers were determined by plaque assay and represented as the log of the titre in plaque forming units (pfu, diamonds). Refractive indices of renografin fractions are shown as squares. (d) Plaque formation induced by infectious agent in Vero cell monolayers (Inf, column 1) is inhibited in the presence of the bacteriostatic antibiotic doxicyclin (Dox, column 2) and ciprofloxacin (Cip, column 3), added in culture for 24 h or seven days, and is similar to that seen in cultures that received no mycoplasma (column 4) or no mycoplasma plus antibiotics (columns 5, 6). Assay sensitivity was 10 pfu/ml. Results are representative of four experiments. (e–h) Immunofluorescence detection of particles in cells at the margins of plaques using antibodies raised against renografin fractions (e,g). Low magnification images show staining of material at margins of plaques (e) defined by nuclei stained with DAPI (f). Note that cells surrounding plaques do not stain (e, corners). At higher magnification, spherical particles ∼0·4 µm in diameter were observed (g). Τhey colocalized with small DNA-containing particles labelled with DAPI (h, arrows) next to the host cell nucleus (N). No staining was seen with preimmune serum or with mice that received culture supernatant from uninfected cells (data not shown). Bars in f and h, 10 µm.

Plaques induced by the isolated infectious agent were indistinguishable from those observed with spleen homogenates and were atypical in many ways. Plaque formation was unusually prolonged (7–10 days postinfection compared to most viruses, 1–4 days) and was consistently observed in a wide range of host cells (fibroblasts, lymphocytes and epithelial cells from human, monkey and mouse). Infection of cultured cells occurred without significant cell lysis, except in agar plaque assays where cells lost could not be replaced by growth of the surrounding monolayer. Several antibiotics prevented plaque formation (doxicyclin, ciprofloxacin, actinomycin D, Fig. 2d) ruling out most viruses as infectious agents. The infectious agent appeared to be a membrane-bound particle as it was unable to induce plaque formation when treated with nonionic detergents. Immunofluorescence analysis of cells using antibodies generated against the infectious agent showed specific labelling of DNA-containing subcellular particles at the margins of plaques (Fig. 2e–h). Electron microscopic examination of cells in plaques (Fig. 3a) and pellets of peak renografin gradient fractions (Fig. 3b) revealed morphologically homogeneous structures ∼0·45 µm in diameter that lacked cell walls and were closely apposed to the plasma membrane of host cells [26]. Taken together, these features suggested that the agent was a mycoplasma.

Fig. 3.

The infectious agent shares morphological features with mycoplasmas and has a molecular composition similar to Mycoplasma hyorhinis. (a,b) Electron micrographs showing particles on cells in monolayers (a, Myc, arrows) and homogenous particles in renografin fractions (b). Mycoplasma features include variable shape, diameter of ∼0·4 micrometers, lack of a cell wall and close apposition to the host cell plasma membrane. (c) A monoclonal antibody raised against a unique Mycoplasma hyorhinis surface antigen stains cells surrounding a plaque formed by the infectious agent in Vero cell monolayers. (d) DNA staining shows position of cells. Note that cells outside area of plaques do not stain with the Mycoplasma hyorhinis-specific antibody (c, right). Bar in b, 0·5 µm for a and b, in d, 10 µm.

Identification of the infectious agent as mycoplasma

We used immunological and molecular biological methods to confirm the identity of the infectious agent as mycoplasma. Immunofluorescence microscopy using a monoclonal antibody specific for Mycoplasma hyorhinis variable lipoprotein (Vlp) [27] showed labelling of the margins of plaques containing the infectious agent (Figs 3c,d). Nucleotide sequences obtained by random cloning of DNA isolated from renografin fractions containing the infectious agent exhibited a high degree of homology to mycoplasma genes (34/46 sequences, 5 genes) including some with identity to the porcine Mycoplasma hyorhinis-specific variable lipoprotein gene. Additional mycoplasma sequences were obtained by PCR amplification of known mycoplasma genes (6 genes).

Mycoplasma induces centrosome autoantibodies in naive mice

We next tested whether introduction of our isolated mycoplasma fractions into naïve mice could induce centrosome autoantibody production. We also used mycoplasma fractions prepared commercially by more conventional isolation methods of colony-purification and broth culturing (Fig. 4, ATCC). In an attempt to mimic the mode of infection in vivo, we delivered the mycoplasmas into mice intranasally. We found that centrosome autoantibodies were produced in over half the mice that received either our M. hyorhinis preparation or the commercial preparation, but not in control mice (Fig. 4a–g).

Fig. 4.

Intranasal delivery of mycoplasma into mice induces anticentrosome autoantibodies and antimycoplasma antibodies. (a–d) Immunofluorescence detection of centrosome autoantibodies in mice. Antibodies in sera from mice infected with mycoplasma detect structures (a, red) that costain with antibodies to the centrosome protein pericentrin (b, centr/DNA, green). Note b is not a merged image with a, but shows position of centrosome (centr) and nucleus (DNA, blue). Mice that did not receive fractions containing mycoplasma did not stain centrosomes (c,d). (e) Mycoplasma-infected mice develop significantly higher titres of anticentrosome IgG autoantibodies 21 weeks postinfection in contrast to uninfected mice. (f) Anti-centrosome IgM autoantibody titres do not differ significantly in mycoplasma-infected and uninfected mice 21 weeks postinfection. In e, f, titres were determined by endpoint titration using the immunofluorescence assay and are expressed as the reciprocal of the serum dilution. (g) Mice infected with Mycoplasma hyorhinis obtained from ATCC as in e. For e-g, each point (diamond) represents the result for an individual mouse. Results are representative of at least 4 independent experiments. Horizontal bars represent mean values. (h,i) Mycoplasma fractions from Fig. 2(c) stained by immunofluorescence to show colocalization of antibodies from mycoplasma-infected mice (h, red) and antibodies specific for Mycoplasma hyorhinis[27] (h, green, overlap yellow); uninfected mice showed no mycoplasma staining (i). (j) Mycoplasma induces centrosome autoantibodies when introduced orally and intranasally but not intravenously or intraperitoneally. Immunoblots of recombinant pericentrin showing two dilutions (1/60, 1/300) of serum samples from four mice infected orally, intranasally, intravenously or intraperitoneally as indicated. h–j, results representative of 3 separate experiments.

A more detailed analysis of centrosome autoreactivity revealed that the mycoplasma-induced response was mainly of the IgG isotype (Fig. 4e). The Igs were of the IgG1 and IgG2b subclasses, they appeared as early as 2–3 weeks postinfection and were well established by 6–8 weeks in all animals. IgM centrosome autoantibodies were detected in naïve animals but they did not increase in titre following mycoplasma infection (Fig. 4f). This suggested that centrosome-reactive IgM were part of the natural antibody repertoire and class switched to high titre IgG during the autoimmune response [28,29]. Notably, high titre anticentrosome antibodies of IgG isotype are also a feature of scleroderma patients [23].

Anti-mycoplasma antibodies were detected in mice infected with mycoplasma fractions from our laboratory (11/12) or commercially produced mycoplasma (4/4 Fig. 4h, ATCC) but not in uninfected animals (0/9, Fig. 4i). Moreover, delivery of mycoplasma via routes consistent with in vivo infection and transmission (intranasal, oral) induced autoantibodies while delivery via nonphysiological routes (intraperitoneal, intravenous) did not induce detectable autoantibody production (Fig. 4j). These results suggest that an established and persistent mycoplasma infection of mucosal surfaces was required to induce and sustain centrosome autoreactivity.

Antibiotics prevent autoantibody production

We next tested whether mycoplasma-induced autoantibodies could be ameliorated by therapeutic intervention. We first attempted to prevent autoantibody production by pretreating mice with the bacteriostatic agent ciprofloxacin prior to the addition of mycoplasma. Autoantibody production in antibiotic-pretreated mice was effectively inhibited when assayed at various times after mycoplasma addition (3 and 6 weeks postinfection, Fig. 5a). Autoantibody titres were essentially indistinguishable from uninfected mice. We next tested the effect of antibiotic treatment on an established autoantibody response. M. hyorhinis-infected mice with centrosome autoantibodies showed significant reduction in autoantibody titres when treated for three weeks with ciprofloxacin, although this treatment was not as effective as pretreatment with antibiotics (Fig. 5b).

Fig. 5.

Antibiotic treatment prevents the onset of mycoplasma-induced centrosome autoantibodies and reduces established autoantibody titres. (a) Autoantibody titres in mice treated with the bacteriostatic antibiotic ciprofloxacin are dramatically lower than in mice that did not receive antibiotic and similar to uninfected mice (P = 0·6). (b) Antibiotic added 6 weeks after mycoplasma infection significantly reduces centrosome autoantibody titres. Each point represents data from a single mouse. Horizontal bars represent mean values. Cip, Ciprofloxacin, Myc, mycoplasma.

The centrosome-specific autoantibody response spreads to include additional intracellular epitopes

All mice that produced autoantibodies in response to mycoplasma showed exclusive autoreactivity to centrosomes and centrosome antigens (Fig. 4). In most cases tested, multiple centrosome components were targeted in this initial response (e.g. pericentrin, centriolin) suggesting that centrosome antigens are the initiating agents in the autoantibody response triggered by M. hyorhinis.

We next tested whether the initial centrosome autoantibody response could spread to include reactivity to other intracellular organelles and structures. This phenomenon, known as antigenic spreading, has been implicated in the generation of complex multitarget autoantibody responses in autoimmune diseases [28]. Beginning at approximately 12 weeks postinfection and 6–9 weeks after centrosome autoreactivity appeared, we detected autoantibodies to another intracellular structure. Immunofluorescence colocalization of mycoplasma-induced autoantibodies with antibodies to calnexin (Fig. 6) and Bip (binding protein of the endoplasmic reticulum, data not shown) [29,30], demonstrated that this structure was the endoplasmic reticulum (ER). Autoantibodies to the ER were exclusively of the IgM isotype (Figs 6d,e). In some animals, autoantibodies to cytoskeletal elements (intermediate filaments) also appeared several weeks after infection (>12, data not shown). These results provide support for the phenomenon of antigenic spreading where autoreactivity evolves from one structure to involve additional structures and antigens. While autoreactivity to the initial centrosome target involved isotype switching from pre-existing IgM to IgG, autoantibodies to additional targets involved generation of new IgM. This result provides insight into the mechanism of antigenic spreading. It suggests that recruitment of additional (secondary) autoantibody targets in this system and possibly during development of human autoimmune pathology is initially T-cell independent.

Fig. 6.

Centrosome autoantibody response amplifies to involve endoplasmic reticulum-associated autoantigens. (a–c) Several weeks after the appearance of centrosome autoreactivity in mycoplasma-infected mice, autoantibodies to a second cytoplasmic structure were detected by immunofluorescence. The autoantibodies (a) colocalize with the endoplasmic reticulum protein calnexin (b) as shown in the merged image (c, yellow, nuclei, blue). (d) Titres of the ER-associated autoantibodies (ER) in mycoplasma-infected mice (Myc) are significantly higher than in uninfected mice, as are centrosome antibodies (c). Data were obtained using an antibody that recognizes all immunoglobulin isotypes and both heavy and light chains. Results are representative of 3 independent experiments. (e) Unlike centrosome autoantibodies, those to the ER-associated autoantigen are of IgM isotype and have significantly higher titres than uninfected mice. Myc, mycoplasma.

DISCUSSION

In this study, we show that centrosome-specific autoantibodies can be experimentally induced by infection of mice with mycoplasma. Autoantibodies to mitotic spindles and centrosomes have been previously, albeit indirectly associated with infections with Mycoplasma pneumoniae, EBV or HIV [31–35]. We show that centrosome autoreactivity is the first to appear. Centrosome antigens trigger an initial antibody response that amplifies by recruiting additional specificities. Such amplification, defined as epitope spreading has been observed frequently following peptide immunization [28,36]. As the expansion of the response to immunodominant peptide epitopes progresses, isotypes shift from IgM to IgG [16]. We also find that anticentrosome IgM are part of the peripheral antibody repertoire of normal mice. Acute mycoplasma infections trigger class switching to anticentrosome IgG. As the response to the initiating epitopes on centrosomes progresses to the IgG isotype, determinant spreading may involve epitopes on endoplasmic reticulum and cytoskeletal antigens by recruiting new autoantibody specificities of IgM isotypes. IgM autoantibody responses have been described following Mycoplasma pneumoniae infection [37].

The mechanism by which mycoplasma infections initiate autoantibody production is currently unknown. Mycoplasmas present a variety of structures capable of engaging both innate and adaptive components of the immune system. Some studies have suggested that mycoplasma superantigens may induce polyclonal activation of autoreactive B and T cells with multiple specificities. In addition, the mycoplasma macrophage-activating lipopeptide (MALP) may stimulate B lymphocytes expressing Toll-like 2 receptors and initiate polyclonal autoantibody responses [38–40]. In turn, centrosome-specific B cells were demonstrated to have less stringent activation requirements compared to other autoreactive B cells, and may be activated first during infection-induced, scleroderma-like autoantibody responses [41].

A role for centrosomes and centrosome-related structures in initiating complex autoantibody responses has recently been suggested for human autoimmune diseases. Autoantibodies to mitotic spindles and spindle poles (centrosomes) have been shown to occur early in the evolution of occupational scleroderma [42]. In other studies, centrosome-specific autoantibodies were identified in a low percentage of patients with the prodromal Raynaud's syndrome or early stages of scleroderma, again implicating these components in the initiation of the autoantibody response [24,43,44].

Autoreactivity to centrosomes in autoimmune diseases has long been underestimated probably due to the difficulty of detecting these small structures by immunohistochemical methods [23,25]. We demonstrate that autoantibodies induced in our experimental system target centrosomes, some of the most prevalent autoantigens in scleroderma. In the human patient, centrosome autoantibodies are frequently accompanied by additional antinuclear antibody specificities [23,25]. None of these antinuclear antibody specificities develop within the first 12 weeks of mycoplasma infection of wild-type mice, while disease-related autoreactivity is confined to centrosomes. Because mycoplasma infection is required to develop the centrosome autoantibody response, it is possible that mycoplasmas contribute to autoimmune pathology. Intriguing remissions of scleroderma have been described following antibiotic treatments known to affect mycoplasmas [45,46].