Arthritis Provoked by Linked T and B Cell Recognition of a Glycolytic Enzyme

See allHide authors and affiliations

Science  26 Nov 1999:
Vol. 286, Issue 5445, pp. 1732-1735
DOI: 10.1126/science.286.5445.1732


The hallmark of rheumatoid arthritis (RA) is specific destruction of the synovial joints. In a mouse line that spontaneously develops a disorder with many of the features of human RA, disease is initiated by T cell recognition of a ubiquitously expressed self-antigen; once initiated, pathology is driven almost entirely by immunoglobulins. In this study, the target of both the initiating T cells and pathogenic immunoglobulins was identified as glucose-6-phosphate isomerase, a glycolytic enzyme. Thus, some forms of RA or related arthritides may develop by a mechanism fundamentally different from the currently popular paradigm of a joint-specific T cell response.

The joint degradation characteristic of RA is thought to result from an autoimmune response, although the precise roles and targets of self-reactive B and T cells remain controversial (1–4). Given the disease's articular localization, it is tempting to infer that the response is elicited by an antigen (Ag) specifically expressed in the joints. However, there is currently no compelling evidence for or against this, so identification of a self-Ag that can spontaneously provoke arthritis would be an important advance.

The K/B×N T cell receptor (TCR) transgenic (tg) mouse line spontaneously develops a joint disorder with many of the clinical, histological, and immunological features of RA in humans (5,6). This line was generated by crossing the KRN/C57Bl/6 TCR tg strain with the NOD strain. By chance, the KRN TCR sees an unknown peptide bound to NOD-derived Ag7 molecules on all major histocompatibility complex (MHC) class II–positive cells. Thus, in the F1 animals, KRN T lymphocytes emerge into an environment of systemic self-reactivity. There is tolerance induction at multiple levels, but it is only partial. B lymphocytes are also critical for the K/B×N disease (5), engaging in a critical Ag7:TCR-mediated interaction with KRN T cells and ultimately acting by secreting arthritogenic immunoglobulins (Igs) (7). Injection of as little as 100 μl of serum from sick K/B×N mice into healthy animals, even lymphocyte-deficient ones, provokes arthritis within days. The arthritogenic activity in K/B×N serum resides in the immunoglobulin G (IgG) fraction and depends on particular Ig specificities, not rheumatoid factor [RF; antibody to IgG (anti-IgG)] or anti–collagen-II.

As a first step in identifying the Ag recognized by the arthritogenic Igs, we prepared different types of extracts (for example, NP-40 and GuHCl) from various mouse tissues and analyzed them by protein immunoblotting with K/B×N or control littermate serum, revealing bound Igs with an anti–mouse IgG reagent (8). With K/B×N serum, several bands showed up, but one band from the NP-40 extracts appeared dominant, corresponding to a protein of about 60 kD. When a large number of sera from K/B×N and control mice were screened, the band at 60 kD was always observed and always dominant with the former, but not with the latter; it was not specific to ankle extracts, also being present in spleen and kidney extracts (for example, Fig. 1A). The 60-kD protein was further highlighted in immunoprecipitation analyses: Precipitates of [35S]methionine-labeled ankle extracts with K/B×N, but not control, sera yielded a single 60-kD protein (Fig. 1B).

Figure 1

Identification of GPI. (A) Protein immunoblot analysis of NP-40 extracts from ankle (An), spleen (Sp), or kidney (Ki) of Ig-deficient mice, with serum from a 35-day-old K/B×N mouse or normal littermates (28). Ctl, control. Numbers at left give molecular mass in kilodaltons. (B) Immunoprecipitation (IPP) with similar sera from [35S]methionine-labeled ankle extracts (3 hours of metabolic labeling in vitro). (C) Immunopurification of kidney extracts on affinity matrices coupled with IgG from arthritic or control sera. (D) Peptides sequenced from the immunopurified 60-kD protein. Single-letter abbreviations for the amino acid residues are as follows: A, Ala; D, Asp; E, Glu; F, Phe; G, Gly; H, His; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; W, Trp; and Y, Tyr.

To identify the 60-kD protein, we purified IgGs from a large pool of control or K/B×N sera and attached them to a matrix (9). Pooled NP-40 kidney extracts were passed over these matrices, the columns were washed extensively, and bound proteins were eluted. Upon electrophoresis and staining, only weak bands appeared with the control column eluate, but a dominant band showed up at 60 kD with the eluate from the K/B×N IgG column (Fig. 1C). This band was excised, the protein in the gel slice was digested by trypsin, and the resulting peptides were resolved by high-performance liquid chromatography (HPLC), allowing the isolation of several of them. Three peptides were sequenced by automatic Edman degradation (Fig. 1D), and the sequences were compared with public data banks with the BLAST program on the Swissprot database. All three peptide sequences resided within the glucose-6-phosphate isomerase (GPI; E.C. protein sequence. The molecular mass of GPI (62.636 kD) coincides well with that of the protein recognized by K/B×N serum (∼60 kD), as estimated from the protein immunoblots.

To confirm that GPI is indeed the protein seen by K/B×N serum, we produced recombinant mouse GPI in Escherichia coli(10). The recombinant protein, a fusion product with glutathione S-transferase (GST), was purified on a glutathione affinity column, electrophoresed, blotted, and probed with sera from K/B×N or control mice. All K/B×N sera reacted strongly with the recombinant protein, whereas control sera did not (Fig. 2A). Enzyme-linked immunosorbent assays (ELISA) with recombinant GPI/GST indicated that K/B×N sera also saw GPI (Fig. 2B). Reactive antibodies (Abs) were detected with K/B×N sera up to dilutions of 1/106. [Sometimes, at far lesser dilutions (for example, 1/100), trace reactivity could also be found with control sera.] Thus, GPI is the 60-kD protein seen by serum Igs from arthritic K/B×N mice.

Figure 2

GPI is recognized by Igs from arthritic K/B×N mice. Sera from arthritic or control mice were tested for reactivity to recombinant GPI/GST (rGPI) protein by (A) protein immunoblots (sera at 1/100 dilution) or (B) ELISA (each curve represents an individual mouse). Control GST protein gave no detectable signal in either assay.

To address the question of whether anti-GPI Igs are really the pathogenic Abs, we made GST or GPI/GST affinity columns and passed a pool of sera from arthritic K/B×N mice sequentially over them (11). The bound proteins were eluted and injected into RAG−/− mice, as were portions of the starting material and the flow-through fraction (12). All arthritogenic activity resided in the fraction bound to the GPI/GST column; none appeared in the flow-through fraction, even though it contained a large portion of the Igs (Fig. 3). Therefore, antibodies to GPI are the pathogenic Igs contained in serum from arthritic K/B×N mice, and no other arthritogenic Abs can be detected. GPI-specific Abs were first detected in the blood at low concentrations around the third week after birth, corresponding to the onset of clinical arthritis, and steadily increased until 8 to 10 weeks (13).

Figure 3

GPI can adsorb arthritogenic Igs. Whole serum from K/B×N mice was passed sequentially over GST- or GPI/GST- coupled affinity columns, and portions of the fractions were tested by injection into RAG−/− mice. Arthritis was read out as clinical index (12) or ankle thickness (millimeters).

Given precedents that B cells present Ags to T cells particularly effectively when they can internalize them through their Ig receptors (14), we wondered whether KRN T cells might not see the same Ag as the arthritogenic Igs. Indeed, they do: Recombinant GPI, when offered by NOD antigen-presenting cells (APCs), could stimulate either the original KRN T cell hybridoma or KRN T cells from tg mice (15) (Fig. 4). Stimulation was efficient, comparable with that by H-2k APCs presenting bovine ribonuclease [the original specificity of the KRN TCR (16)], and was specific because non–TCR tg cells did not respond.

Figure 4

(A and B) GPI is recognized by KRN T cells. The KRN T cell hybridoma or primary T cells from KRN TCR tg mice (on the neutral C57Bl/6 background) were stimulated with APCs from NOD or B10.BR mice and varying doses of Ag (fivefold serial dilutions; 41 to 61 peptide from bovine ribonuclease, maximum concentration, 10 μg/ml; recombinant GPI/GST or GST control protein, maximum concentration, 1 mg/ml). K, response of lymph node cells from nontransgenic control littermates. The bovine ribonuclease peptide presented by Ak represents the original specificity of the KRN receptor (16).

Thus, the molecular target of both self-reactive KRN T cells and the arthritogenic Igs in K/B×N serum is GPI, a critical enzyme of the glycolytic pathway, expressed in essentially all tissues from the earliest stages of embryogenesis until death, although with some quantitative variations (17–19). It normally resides in the cytoplasm, but soluble GPI has been detected in human (20) and murine (13) serum. This finding substantiates our earlier suggestion that a joint-specific disease can result from systemic self-reactivity (5). However, arthritis is not an inevitable consequence of widespread autoreactivity in TCR tg mice: KRN T cells also recognize an unknown self-peptide presented by a variant Ak complex, but when this class II molecule was expressed in KRN TCR tg mice, there were no signs of joint pathology (21). No doubt, a strict set of conditions must be met for arthritis to develop. The key to pathology is probably linked T/B cell stimulation by a self-protein that resides in the cytoplasm of most cells and also occurs at a very low concentration in the blood. All MHC class II–positive cells synthesize GPI, and certain of these, in particular dendritic cells, should be capable of stimulating T cells that can recognize it (5, 13). However, most cells, including the bulk of B cells, will not present GPI very effectively because they are poor at displaying peptides derived from cytoplasmic proteins and are inept at taking up and displaying peptides derived from low-concentration, blood-borne proteins. Only those rare B cells expressing high-affinity GPI-specific Igs on their surface will be capable of extracting it from the blood and efficiently presenting derivative peptides (14). These cells will be activated and helped by their T cell cognates in a manner that is qualitatively different from any influence on the bulk of B cells, resulting in high-level production of a very narrow slice of the Ig repertoire.

A critical outstanding issue is how this systemic self-reactivity provokes arthritis in the absence of other autoimmune manifestations. It is possible that special features of GPI expression in synovial tissues are key—concentration, subcellular partitioning, and secretion—especially because joints are characterized by high metabolic activity. Or it is possible that the pathogenic Igs cross- react with another protein specifically expressed in joints. However, we did not find evidence of any such pathogenic cross reactivity: K/B×N serum did “light up” a few bands in GuHCl extracts of ankle, but not other tissue, but these specificities partitioned in the nonarthritogenic flow-through fraction when K/B×N sera were passed over GPI affinity columns (13). We favor the notion that some unusual physiological feature of joints is responsible for focusing the autoimmune destruction, perhaps related to their being rich in proteoglycans, devoid of a basement membrane, hypoxic, and subject to reduced blood flow. Some aspect of the following pathway might thereby be promoted: Immune complexes, formed inside or outside the joint, activate the complement cascade and engage Fc receptors (13), setting off a chain of destructive events, including recruitment and activation of inflammatory cells and synoviocytes, overproduction of growth factors and cytokines, and synthesis of degradative enzymes. A likely possibility is enhancement of the production or retention of immune complexes, but effects on downstream events, for example, the recruitment of inflammatory cells, are also possible. The development of arthritis in lymphocyte-deficient mice systemically overproducing tumor necrosis factor (TNF)–α (22) may reflect convergence somewhere along such a pathway.

Finally, how do these results relate to human arthridites? Whether or not RA patients' arthritis is initiated by a joint-specific Ag has been debated for years, without a satisfying resolution. The view that it is has been espoused by many ever since injection of collagen-II into rodents was found to provoke arthritis (23). However, there is no compelling evidence that collagen-II or any other joint-specific Ag is a central player in RA—indeed, attempts to modulate disease by inducing tolerance with a joint-specific Ag administered orally (by means of a bystander effect) have been disappointing (24). Whether the identification of GPI as the Ag that initiates arthritis in K/B×N mice bears on this issue obviously depends on how relevant this model is to human RA. As has been discussed at length (5,6), it does show marked similarities but is certainly not a perfect mimic. It remains possible that the K/B×N model is most relevant to a related form of arthritis (for example, psoriatic arthritis) or a subset of RA (for example, juvenile RA, which often shares a lack of RF and a similar slightly atypical pattern of joint involvement). We suggest that a scenario similar to that proposed for the K/B×N model might unfold in at least some arthritis patients. GPI could be the inciting Ag, but equally plausible is that other self-Ags with similar properties are responsible. Certain aspects of this scenario are reminiscent of disease mechanisms proposed in the 1970s (25, 26), never refuted but eventually disregarded because a paradigm focusing on a joint-specific T cell response became the dominant one. This route to autoimmune disease may not be restricted to arthritis as Igs recognizing ubiquitously expressed proteins have been observed in other organ-specific autoimmune diseases, although their pathological relevance remains to be established (27).

  • * Present address: Joslin Diabetes Center, Harvard Medical School, One Joslin Place, Boston, MA 02215, USA.


Stay Connected to Science

Navigate This Article