IgG autoantibodies and immune complexes have long been recognized as potent pathologic triggers of inflammatory responses. Early hypotheses regarding the pathogenesis of rheumatoid arthritis (RA) were molded by experimental models of immune complex disease. IgG aggregates are abundant in RA synovial fluids, and can trigger complement activation. Hence, it seemed logical that tissue damage in RA was attributable to the local deposition of immune complexes.

Rheumatoid factors (RFs), which are autoantibodies specific for the constant regions of IgG, are detectable in more than 80% of RA patients, and may also be present in a 'hidden' or complexed form in the synovial fluids of some seronegative patients. RFs efficiently fix and activate complement in vitro by the classic pathway 1. In vivo turnover studies of radiotagged complement proteins in seropositive RA patients demonstrated that, compared with control individuals, complement consumption was greatly accelerated, especially at the extravascular sites of inflammation 2. Consistent with the notion that immune complex formation was maximal at the synovial sites of inflammation, complement activation was shown to be much greater in RA synovial fluid than in blood 234. Levels of C4 fragments at these sites also correlated with titers of IgM RFs.

Infiltrating leukocytes are known to be recruited by the downstream products of complement activation, especially the soluble anaphylatoxin C5a, with subsequent enlistment of other components of the membrane attack complex 56. Flares of clinical activity in RA correlate with increased levels of RF secreting cells, which are especially prevalent in the bone marrow and synovial fluid of RA patients 7.

IgM RFs have been reported to account for more than 10% of local plasma cells in RA synovia 89. However, infusions of RF into healthy individuals causes neither sustained nor transient synovitis 10, indicating that RF autoantibodies by themselves are not pathogenic. Nevertheless, IgM-RF-containing immune complexes may also include IgG antibodies and unidentified peptides, which could derive from self or exogenous antigens. In addition, the recently resolved crystallographic structure of a human IgM RF–Fc co-complex revealed that contacts with IgG antigen involved only the periphery of the antigen-binding cleft of the autoantibody 11. These findings may indicate that RA RF is capable of binding IgG as well as another self or foreign antigen. Thus, although RF alone is not proinflammatory, RF associated with immune complexes can enhance local inflammatory processes, and there is compelling clinical evidence that RFs contribute to extra-articular disease.

In addition to being the precursors of antibody-secreting plasma cells, the B cells in RA can play a critical role in the afferent arm of the immune response (Table 1). Thus, B cells can act as highly efficient antigen-presenting cells (APCs), supporting the activation of autoreactive T cells. In fact, by virtue of the high affinity of a specific membrane-associated immunoglobulin for antigen, an antigen-specific B cell can take up, process, and present peptides from nominal antigen with 1000-fold or greater efficiency than a 'professional' APC. In addition, activated B cells can synthesize cytokines and membrane-associated molecules that provide nonspecific help to adjacent T cells.

Strong experimental support for a central role of B cells in RA pathogenesis, independent of antibody formation, came from studies of human synovium/SCID mouse chimeras 12. Those studies confirmed that the T cell activation is B cell dependent, in that targeted deletion of B cells impaired local T cell responsiveness, and APCs other than B cells could not substitute for the maintenance of T cell activation.

Following the development of methods for introduction of transgenes and for targeted gene disruption, in vivo murine models have been developed that have enabled a re-evaluation of the role of B cells in arthritis. The KRN/NOD murine model has been particularly revealing, because in these mice there is complete penetrance of a genetically determined disease process that results in severe distal joint inflammation, which emulates key features of RA 13. Moreover, development of the disease involves the coordinated functions of both B cells and T cells, because infusion of nondepleting antibody to CD4 blocks disease, and mice devoid of B cells also do not develop disease 13. A pathogenic role for IgG autoantibodies to glucose-6-phosphate isomerase (GPI) has been documented in the KRN/NOD model; it was found that infusions of these autoantibodies into otherwise healthy mice rapidly led to the development of synovitis 14. However, arthritis was not induced by introduction of preformed anti-GPI antibody–antigen complexes, which suggested that joint disease can result when autoantibodies, produced either locally or systemically, interact with antigens on synovial surfaces to develop immune complexes in situ. These locally formed immune complexes were shown to be proinflammatory only if they involve antibodies to two or more epitopes, and the more diverse the antibody response, the more efficient the recruitment of proinflammatory factors 15, suggesting that a complex IgG autoantigen lattice is required for the recruitment of downstream inflammatory effectors.

Studies in the KRN/NOD model have also enabled a closer dissection of the role of complement in the development of immune complex mediated synovitis. Although anti-GPI antibodies are prevalent in the circulation, disease is limited to the joints, probably because only articular immune complexes are efficient at fixing complement 14. C4-deficient KRN/NOD mice develop arthritis of the same severity as do wild-type mice, whereas animals deficient in factor B of the alternative complement pathway develop attenuated disease or no arthritis at all 1617. In addition, a partial dependence on C3 was also demonstrated 16, which is consistent with the known role of C3 in the stabilization of immune complexes. These findings have led to a revision of the earlier notion that IgG immune complexes trigger inflammation primarily by engagement of the classical pathway of complement activation, because these murine studies indicated that disease initiation instead may involve the alternative pathway.

Although interactions between immune complexes and cellular receptors for the Fc regions of IgG (FcγR) were not considered in earlier clinical investigations of RA, the recent characterization of these receptors and the availability of relevant murine model systems has permitted a thorough examination of their impact on pathogenesis. The three classes of Fcγ cell surface receptors, isolated based on their capacity to bind IgG-containing immune complex, are heterogeneous in their binding specificities for IgG₁–IgG₄, and their intracellular signaling motifs that can either activate or inhibit cellular effector functions. Although mice with deficiencies in complement components typically exhibit attenuated immune complex mediated disease in relevant murine models, the loss of activating FcγRIII completely ablates arthritis development 18. Connecting these model systems to clinical disease, in vitro blockade of FcγRIIIA on human macrophages was recently shown to prevent the release of tumor necrosis factor (TNF)-α and induction of IL-1α in human macrophages 19. FcγRIII may be an especially important mediator of immune complex induced tissue damage in RA, because of the size and composition of immune complexes that arise at sites of disease (discussed by Edwards and coworkers 20). In more recent studies, C5a, acting through the C5a receptor, was shown to exacerbate immune complex induced inflammatory disease by altering the ratio of activating to inhibitory FcγR on macrophages 21. These studies demonstrated a direct link between the C5a chemoattractant and FcγR related mechanisms responsible for immune complex triggered inflammatory responses.

These investigations of murine models of RA have led to a revision of our understanding of the role of secreted autoantibodies in RA. Complement alone may be inadequate to sustain chronic inflammatory responses to immune complex deposition, because stimulatory and inhibitory FcγR may in fact be the primary regulators of subsequent immune responses (for review 21). The host response to immune complex functions in the joints is critically dependent on interactions with cell surface FcγR on B cells and macrophages.

Mast cells are prominent in the synovial infiltrates of many RA patients 22, and these tissue associated infiltrating cells express membrane associated FcR for IgE and IgG, which enable triggering resulting from antigen-specific (and perhaps nonspecific) immunoglobulin cross-linking. In murine models of immune complex disease, the mast cell has been shown to play a central role in immune complex mediated joint inflammation 23. Mast cells triggered by immune complexes produce proinflammatory cytokines and proteolytic enzymes at sites of cartilage erosions 24. The mast cells are also major sources of the vasoactive and chemotactic factors that facilitate the recruitment of other leukocytes to the synovial tissues. KRN/NOD mice that lack mast cells are resistant to inflammatory and erosive arthritis induced by arthritogenic serum 25. Along with macrophages, mast cells can release TNF-α and IL-1. For these reasons, the mast cell is now appreciated to be a cellular link between B cells, autoantibodies, complement, and other inflammatory mediators that contribute to erosive arthritis.

Based on the unexpected discovery that antibodies to GPI can induce synovitis in healthy mice, several groups have looked for a role for these autoantibodies in clinical disease. GPI is a ubiquitous enzyme that is essential for glucose metabolism in all cell types, and it therefore cannot represent a tissue-specific antigen recognized as part of an autoimmune disease that is limited to the joints. A more likely scenario is that increased local cell turnover results in the deposition of GPI along joint surfaces 14, where it may be recognized by autoantibodies that enter from the circulation. Although Schaller and coworkers 26 reported increased anti-GPI antibody titers in 64% of 69 RA patients, but not in patients with Lyme disease or Sjögren's syndrome, these findings were not confirmed in subsequent reports. In a recent flurry of studies, only a small minority of RA patients had detectable levels of autoantibodies to GPI, and there were no significant differences from patients with other types of joint diseases 27282930. However, it remains a possibility that anti-GPI antibodies may be more common in RA patients with extra-articular disease, and especially in patients with Felty's syndrome, which were highly represented in the first reported clinical survey 26.

Many RA patients have circulating antibodies to autoantigens other than IgG, including type II collagen, heat shock proteins, proteoglycans, cartilage link protein, and heavy chain binding proteins 31. In some cases T cell reactivity to the same antigens was demonstrated by in vitro proliferation assays 32. In certain instances experimental immunization with the antigens in complete Freund's adjuvant can cause arthritis in rodent models 33. However, immune responses to each of these antigens have been detected in only a minority of patients, and a causative role in clinical disease has not been firmly established. Therefore, at best, such joint-specific autoantibodies cannot account for chronic inflammatory arthritis in most RA patients.

More prominent in RA are IgG antibodies to citrulline-modified peptides and proteins (CCP). Citrullination represents a post-translational modification due to the enzymatic deamination of peptidyl arginine to peptidyl citrulline. It has been shown that citrulline is the essential antigen epitope recognized by anti-CCP antibodies, antiperinuclear antibodies, as well as antibodies to keratin, filaggrin, and Sa. In fact, one extensive study 34 concluded that although RF may be a more sensitive test for RA, detection of anti-CCP antibodies is more than 90% specific for the diagnosis of RA. Hence, the available evidence supports a role for serologic testing for anti-CCP antibodies as an aid in the diagnosis of RA, especially at early stages of disease 3536.

Although the biologic significance of anti-CCP antibodies is unclear, citrullination may be a byproduct of abnormal protein metabolism occurring at in vivo sites of increased or abnormal apoptosis 37. The B lymphocytes from RA patients are resistant to certain apoptotic stimuli 38, which may reflect prosurvival signals delivered from synovial or bone marrow stromal cells 39. Notably, increased production of citrullinated peptide autoantibodies has been demonstrated in two murine model systems of autoimmunity with abnormalities in B cell apoptosis, whereas anti-CCP responses were not associated with other autoimmune models that included the classical collagen-induced arthritis system 40. These findings may suggest that the generation of autoantibodies to citrulline related neoantigens could be closely linked to mechanisms responsible for impaired lymphocyte clonal regulation 34.

Citrullination of proteins may also have the potential to contribute directly to the autoimmune response 41. An unexpected mechanistic linkage was found in studies of mice transgenic for the human HLA-DRB1*0401 MHC class II molecule, which contains the 'shared epitope' – a well studied genetic susceptibility factor for RA. In vitro incubation of transgenic B cells and macrophages with citrulline-containing peptides resulted in enhanced peptide side chain interactions with the shared epitope that significantly increased peptide MHC affinity. Moreover, immunization of the HLA-DRB1*0401 transgenic mice led to greatly enhanced in vitro recall responses to synthetic citrullinated self peptides 41. These observations suggest a new mechanism by which the 'shared epitope' can predispose to RA.

About 60% or more of the synovial samples from RA patients have infiltrates of B and T lymphocytes. Three separate patterns of infiltrates have been described: diffuse lymphocytic infiltrates with interdigitating dendritic cells and variable amounts of B cells; aggregates of infiltrating B and T cells in more substantial numbers, associated with interdigitating dendritic cells in disorganized groupings; and T cells and B lymphocytes clustered in aggregates arrayed around interdigitating dendritic cells and associated with follicular dendritic cell networks. In this latter pattern, which is present in less than one-third of patients, the synovial cellular infiltrates appear to be organized into distinct B cell follicle-like structures in close spatial relationship to CD4⁺ T cells and CD8⁺T cells. In general, these histologic features are similar to the germinal center (GC) reactions that arise in peripheral lymphoid tissues during antigen-specific responses after immunization 42. Immunohistochemical analyses have also shown that within the lymphocytic infiltrates in the synovial tissue, plasma cells may be organized in concentric rings around large cellular clusters of T cells and CD20⁺ B cells, or as perivascular clusters 43. In addition, B cells and T cells in rheumatoid synovial samples may bear the proliferation associated cell marker Ki-67. This antigen has been detected primarily on lymphocytes in the GC-like pattern 44, but concurrent mitotic figures in RA synovial intima are very low, suggesting that local B or T cell proliferative expansions may be quite limited and/or that expression of this marker is not synonymous with proliferation. Moreover, B cell proliferation in ectopic lymphoid infiltrates has been seen mainly in the network of the follicular dendritic cell, whereas in human tonsilar GC reactions proliferation occurs in a separate dark zone 45. These findings suggest that ectopic GC-like structures have functional and organizational differences from their physiologic analogs.

Although the presence of lymphoid infiltrates generally correlates with clinical disease in RA, there is currently limited information regarding the natural history of these histologic patterns 46, and it is unknown whether a particular histologic pattern has implications for prognosis or treatment responsiveness. Furthermore, although the diseased joints in RA patients may have organized lymphoid infiltrates that emulate physiologic peripheral lymphoid tissues, it is still not clear that these changes are entirely specific for RA. The affected joints of patients with ankylosing spondylitis have also been shown at times to harbor GC-like aggregates 47. Even the synovia from osteoarthritic joints can occasionally contain infiltrates of activated B cells and plasma cells exhibiting clonally related antibody gene sequences 4849. In addition, lymphotoxin-β (LT-1β2), which is produced by activated T cells, may be a downstream effector that is required for the development of primary B cell follicles in the synovial infiltrates 505152. Rheumatoid synovial tissues have also unexpectedly been found to be rich sources of CXCL12 (also termed stromal cell derived factor [SDF]-1) 53. In fact, SDF-1 produced by fibroblast-like synoviocytes has been shown to contribute to the resistance of B cells to apoptosis, which supports an earlier hypothesis that specialized synovial 'nurse-like cells' peculiar to RA synovium mediate homing and survival of B cells 54. In recent studies plasma cells have also been shown to migrate toward gradients of SDF-1 55, CXCL9 (monokine induced by IFN-γ), CXCL10 (IFN-γ-inducible protein 10), and CXCL11 (IFN-inducible T cell α chemoattractant) 55. SDF-1, as well as IL-5, IL-6, TNF-α, and ligands for CD44, can also prolong the longevity of plasma cells 56. In addition, the TNF-α family member B lymphocyte stimulator (BLyS, also called Baff) has also recently been detected at high levels in rheumatoid synovial fluid, suggesting that this prosurvival factor can also be locally produced in inflamed joints 57.

The dysregulated expression of cytokines and costimulatory molecules explains the reported accumulation in one-third of RA samples of an unusual subpopulation of peripheral B cells with a restricted immunoglobulin variable region gene repertoire that coexpress conventional light chains and the surrogate light chain of pre-B cells 58. Such B lineage cells have been postulated to be promiscuous in antigen presentation, and thus could present diverse autoantigens. This topic is still highly controversial because other investigators have not found these B lineage cells in healthy tissues, and have challenged the authenticity of their reported phenotype 59.

The imprint of the potent chemoattractive activating and antiapoptotic factors described above may largely explain the skewed distribution of the B lymphoid cells that accumulate in the synovium. The local production of these factors at sites of inflammation in RA joints may serve as beacons to foster B cell accumulation, proliferation, and differentiation. An active (auto)antigen driven GC reaction may not be required to explain the development of the ectopic B lymphoid infiltrates in the joints of RA patients.

Recent reports have confirmed the importance of immune complexes in the pathogenesis of RA, and have elucidated additional critical roles for B cells and their immunoglobulin products in self-sustaining chronic inflammatory processes (Table 2). These findings have contributed to the rationale for the development of targeted therapies that delete B cells 60 or that attenuate the function of secreted and membrane associated factors that contribute to B cell accumulation and survival at sites of disease 61. However, an appreciation of these diverse potential roles may also predict that the targeted interference with just one B cell antigen or costimulatory molecule may not be sufficient to arrest ongoing disease in all RA patients.

None declared.

APC = antigen-presenting cell; CCP = citrulline-modified peptides and proteins; CXCL = CXC ligand; GC = germinal center; GPI = glucose-6-phosphate isomerase; IFN = interferon; IL = interleukin; RA = rheumatoid arthritis; RF = rheumatoid factor; SDF = stromal cell derived factor; TNF = tumor necrosis factor.