How important are T cells in chronic rheumatoid synovitis?: II. T cell-independent mechanisms from beginning to end



A little over a decade ago, we reviewed studies on the role of T cells in rheumatoid arthritis (RA) and concluded that the prevailing paradigm was not consistent with the data (1). At the time, the primacy of T cells in the pathogenesis of autoimmune disease was undisputed. Based on a series of studies demonstrating a surprising paucity of T cell cytokines in the joints of our patients and the dominance of macrophage and fibroblast products (2–4), we proposed that the perpetuation of RA was not necessarily an antigen-specific process and that T cells were only one of several important players. Recently, the potential role of T cells in the pathogenesis of RA was reviewed, and their ambiguous contribution was emphasized (5).

Now, 12 years later, we think it reasonable to revisit these issues. It is important, though, to frame the questions appropriately. Our original hypothesis did not suggest that T cell products were absent from the rheumatoid joint. Rather, we argued that the presence or absence of T cell products in the joint is not as important as their relative amounts in RA compared with known antigen-specific T cell-mediated disease. Indeed, after several years of debate, the notion that T cell cytokines are relatively low (but not absent!) in RA is now ingrained in the literature (6, 7). Moreover, it is apparent that viewing RA as simply a T cell-mediated or T cell-independent disease is a narrow perspective.

Our original hypothesis needs to be placed in a larger context. RA could be considered as three separate, albeit interrelated, processes: disease initiation, perpetuation, and terminal destruction. In addition, we propose that antigen-independent mechanisms might play an integral role during all stages of RA.

Initiation of rheumatoid arthritis: What comes before T cells?

While attention has quite properly focused on lymphocyte accumulation in the joints of patients with RA, much less notice has been given to events that precede their arrival from the bloodstream. Almost nothing is known about how and when RA begins, although some recent commentaries have reflected on these questions (8). What is certain, however, is that the normal joint contains little more than blood vessels and the cells of the synovial lining. Therefore, it follows that the earliest change in inflammatory joint disease must involve synovial lining cells in preparation for any immune response that follows. These events are analogous to the “innate” immune response that comes before “acquired” (adaptive) immunity.

How innate immunity initiates arthritis in animal models. The idea that abnormalities in the blood and joints of patients with RA can be detected prior to, or simultaneously with, the appearance of articular inflammation has been the subject of a number of articles (9–12) and editorials (13, 14). For practical and ethical reasons, however, the demonstration of an antecedent initiation, or “induction phase,” of RA is difficult. On the other hand, it is possible to examine from the earliest time the progression of immunologically mediated arthritis in experimental animals. For instance, 2–4 weeks following immunization of appropriate strains of mice or rats with heterologous type II collagen and Freund's complete adjuvant (CFA), an inflammatory, destructive polyarthritis develops, with features reminiscent of RA (15). The latency in the development of this collagen-induced arthritis (CIA) is usually ascribed to the time needed for the development of a robust cellular and humoral immune response.

Analysis of articular tissues at earlier times provides an alternative explanation. Abnormalities in the synovium that antedate clinical arthritis were first observed by light and electron microscopy in the knee joints of rats immunized with collagen (16). Two distinct stages in the development of CIA were recognized, starting with an initial stage beginning around day 12. This stage was limited to fibrin deposition in the joints and synovial hyperplasia and was seen whether or not the animals subsequently got arthritis. Some animals later developed a severe inflammatory synovial reaction, with bone and cartilage destruction.

Similarly, examination of the synovial tissues of mice immunized with type II collagen and CFA many days before any clinical evidence of joint swelling or tenderness showed hyperplasia of the lining, vasodilation, and mesenchymal-appearing cells, but no infiltration with lymphocytes or leukocytes (17). If the immunization was limited to CFA without the collagen antigen, then arthritis did not develop, but the early synovial tissue changes were still seen, implying that there was an antigen-independent induction phase.

Additional support for cellular activation in the earliest stage of murine CIA is the increased expression of the transcription factors activator protein 1 and nuclear factor κB (NF-κB) in joint tissue extracts 1–2 weeks in advance of clinical arthritis (18). In rat adjuvant arthritis, the activated form of NF-κB was detected in synovial lining cells 10 days before joint swelling (19). Clinically uninvolved paws in mice with CIA exhibited an increase in 5 of 13 cytokines measured (tumor necrosis factor α [TNFα], interleukin-1 [IL-1], IL-11, RANTES, and macrophage inflammatory protein 1) compared with controls. The other 8 cytokines were detected only after arthritis had begun (20).

Synoviocyte activation was observed in rats injected with CFA (containing mycobacteria), but not with Freund's incomplete adjuvant, days before the onset of arthritis. The authors prophetically concluded that synoviocytes played an active role in adjuvant-induced arthritis, and “not merely a responder role that is dependent on the infiltrating lymphocytes” (21). The epiphyseal bone marrow adjacent to affected joints shows activated inflammatory cells, formation of tartrate-resistant acid phosphatase-positive multinucleated giant cells, and osteoclast-inducing activity before synovitis in CIA, MRL/lpr arthritis, and adjuvant-induced arthritis, respectively (22, 23).

T cell independence has been demonstrated in genetic models of arthritis. For instance, BALB/c mice rendered deficient in the IL-1 receptor antagonist (IL-1Ra−/−) by gene-targeting developed a spontaneous, chronic, inflammatory, destructive polyarthritis. The disease appeared in >80% of mice by 8 weeks after birth, and by week 13, all mice were affected. Hyperimmunoglobulinemia, autoantibodies, and a marked increase in the expression of inflammatory cytokines in the joints of the IL-1Ra−/− mice were noted by 16 weeks. Interestingly, levels of expression of IL-1 were elevated 2–3-fold in the nonarthritic joints of animals at week 4, showing that IL-1Ra deficiency causes the augmentation of proinflammatory cytokines and inflammatory mediators prior to clinically apparent synovitis (24).

Innate immunity in RA: the calm before the T cell storm. Based on our understanding of incipient inflammatory arthritis in animals, we propose that a similar series of events occur in RA. While it may be attractive to consider antigen-specific immune responses as the primary driving force in the very earliest stages of RA, it is more likely that engagement of innate immune mechanisms focuses this process. As noted above, this has been documented in a broad range of arthritis models. The rapid innate response system prepares the synovium for infiltration by T cells and the subsequent immune events in the joint.

The establishment of a chronic inflammatory joint disease such as RA likely begins with engagement of synovial lining cells. These produce the chemokines, including stromal cell-derived factor 1, that call monocytes and lymphocytes into the joint. If they recognize an antigen(s) in the articular tissues, the lymphocytes are retained and inflammation ensues. Within this “toxic” environment, the T cell phenotype is altered. This hypothesis does not require that one specific etiologic factor is the trigger for the innate response; any pathogen that gains entry to the joint could be responsible in a given individual.

The innate response might elicit little or no overt inflammation and would pass unnoticed, but since innate immune recognition is mediated by germline receptors, there is a possibility for genetically determined variability in the response to the same environmental insult. For instance, common mutations in the TLR4 gene are associated with different responses to lipopolysaccharide in both mice and humans (25, 26). Similarly, CIA-resistant H-2b mice (C57BL/6 mice) were found to develop an arthritogenic antibody response and arthritis if the immunization procedure used 4–6-fold more CFA. That finding led the investigators to consider “the possibility of important contributions from non-MHC genes to the expression of autoimmune disease, particularly those which might be influenced by CFA” (27).

Another corollary of this hypothesis is self-evident, namely, sensitized T cells in the circulation do not produce disease. They are only pathogenic when they access tissues. Conversely, interference with their translocation from the circulation into the joint during the self-limited induction phase could prevent the establishment of chronic arthritis.

T cells and the perpetuation of RA

T cells and cytokine networks in RA: what we said 10 years ago. The purpose of our original 1990 editorial was to describe the cytokine milieu in RA joints and to propose a novel hypothesis that would explain its composition. One should recall that in the late 1980s, the data on intraarticular cytokines were limited and divergent. A number of studies putatively showed substantial amounts of T cell cytokines in the rheumatoid joint, including IL-2 and interferon-γ (IFNγ) (28, 29). However, this work relied almost exclusively on relatively nonspecific biologic assays. Using newly available selective reagents, we demonstrated in 1987 that only small amounts of IFNγ were present in RA synovial fluids. Very little was spontaneously produced by RA synovial tissue cells, although synovial T cells were fully capable of secreting the cytokine when stimulated. In addition, one of the hallmarks of the rheumatoid synovium, i.e., high expression of HLA-DR, was shown to be independent of IFNγ (3).

Subsequent studies revealed that the IL-2-like biologic activity in RA synovial fluid was not due to IL-2 itself and that the HLA-DR-inducing factor was actually granulocyte-macrophage colony-stimulating factor (GM-CSF) (30). Using in situ hybridization, we provided evidence that monocyte- and fibroblast-derived cytokine genes such as TNFα and IL-1β are overexpressed in RA synovium (4). Studying isolated populations of cells, even cytokines that could potentially be produced by T cells (e.g., TNFα, GM-CSF, IL-6) were shown to be derived from the macrophages and fibroblasts of the synovium. IFNγ messenger RNA was detected, but only at very low levels. Hence, the dominant cytokines in RA are produced by macrophages and fibroblasts. While initially a subject of controversy, the relative paucity of T cell factors, including IL-4 and TNFβ was subsequently confirmed by others (5, 6).

It is important to emphasize that we never claimed that T cell products are absent from the joint, but rather, that their concentrations are lower than expected for a chronic T cell-mediated disease. For instance, IFNγ concentrations are up to 100 times higher in pleural fluid from patients with chronic tuberculous pleuritis than in synovial fluid from patients with RA (31). Allergen-induced asthma is also marked by a plethora of T cell factors. Using immunohistochemistry, Tak and colleagues (32), among others, have elegantly shown that IFNγ protein can be detected in RA synovium, but it is present in far lower amounts than in other chronically inflamed tissues (33). Hence, the message was not that T cell products are absent from the RA joint, but that T cell-independent mechanisms must be important in rheumatoid synovitis.

The perpetuation of chronic RA synovitis: T cell-independent pathways. If T cells and their products do not drive synovitis, what does? Applying the data from the cytokine profile, we proposed that paracrine and autocrine cytokine networks contributed to chronic joint inflammation following the initiation of RA.

For example, IL-1 and TNFα (synovial macrophage products) stimulate fibroblast proliferation and increase the secretion of synoviocyte products, such as IL-6, GM-CSF, and collagenase. GM-CSF, which is made by synovial macrophages and by IL-1β– or TNFα-stimulated fibroblast-like synoviocytes (FLS), is a potent inducer of IL-1 secretion and increases the expression of class II MHC molecules on macrophages. Cytokines produced by macrophages and FLS could also affect T cells and contribute to the modest degree of T cell activation that is observed, while B cell activation and rheumatoid factor production might result from T cell-independent mechanisms or by undefined processes. For instance, some aspects of B cell function, including isotype switching from IgM to IgG, suggest a T cell contribution. Finally, stimulation of the vasculature by cytokines such as TNFα would induce angiogenesis, increase the synthesis of adhesion molecules, and contribute to the accumulation of mononuclear cells in the synovium.

Notably, this original hypothesis did not rule out a role for T cells; it merely proposed that multiple pathways, both antigen-specific and antigen-independent, could be involved. These are not mutually exclusive, and different mechanisms probably dominate at various phases of the disease.

What we know now that we didn't know then. Much has been learned since those early studies of T cell cytokines. It is now known that T helper cells can be divided into different subsets based on their cytokine profiles (34). Th1 cells produce IFNγ, IL-2, and IL-17, but not IL-4, IL-5, IL-10, or IL-13; Th2 cells express the opposite profile. Th1 cells generally regulate delayed-type hypersensitivity, and Th2 cells are involved with allergic responses and antibody isotype switching (35). Furthermore, Th1 and Th2 cells seem to be mutually antagonistic, and each can suppress the activity of the other.

While most T cell cytokines are present in only small amounts in the joint, more recent studies using very sensitive techniques, such as reverse transcription-polymerase chain reaction or in situ hybridization, suggest a Th1 bias in the synovium (36, 37). This is similar to the situation in several animal models of autoimmunity, where Th1 cells are responsible for disease initiation, while Th2 cell activation correlates with disease regression (38). The very low levels of some Th2 cytokines, including IL-4 and IL-10 (known to have antiinflammatory activities), might theoretically account for disease perpetuation by a failure to suppress the production of proinflammatory cytokines, prostaglandins, and proteases. The disease duration and the degree of clinically apparent arthritis do not seem to influence the relative production of Th1 and Th2 cytokines: almost the same amounts of IFNγ are expressed in joints with very early RA or even in asymptomatic joints as in joints with longstanding disease (11). This suggests that a lack of some T cell factors and/or an insufficient antigen-driven Th2 cell activation might contribute to rheumatoid synovitis.

One mechanism that probably contributes to the muted T cell responses in the rheumatoid joint is local inhibition of cell activation. IL-1Ra and TGFβ are T cell inhibitors, and both have been identified in synovial effusions (39, 40). Nonspecific components of joint effusions, such as hyaluronate, are also toxic to cells and can indirectly suppress T cell activation. The inflamed joint is relatively hypoxic, and the hyporesponsiveness of synovial fluid T cells correlates with a significant decrease in the intracellular redox state that regulates glutathione (41, 42).

Diminished T cell activation could also be related to abnormalities in T cell receptor (TCR) signaling. Articular T cells in RA have decreased tyrosine phosphorylation of proteins after stimulation, especially in the p38 mitogen-activated protein kinase pathway. Furthermore, tyrosine phosphorylation of the TCR ζ chain, an early event in TCR signaling, is defective compared with that in peripheral blood T cells (41). Decreased levels of the ζ protein are also observed, suggesting that the TCR apparatus is abnormal in RA. It remains to be determined whether these are primary or secondary defects. In any case, gene expression profiles of RA T cells are more consistent with antigen-induced anergy than with activation (43).

We now know a great deal more about cytokines, and much of that information is relevant to RA. New cytokines, their cellular sources, and their redundancies may play a role in the perpetuation of disease. One example is IL-13, a product of Th2 cells and monocytes, which has structural similarities to IL-4 and shares pleiotropic functions with IL-4 on T cells, B cells, and natural killer cells. Like IL-4, IL-13 can reduce IL-1 and TNFα production by monocytes. Some studies suggest that IL-13 is present in RA (44), but most show that it is essentially absent (45). Another product of activated macrophages, IL-15, shares common receptor signaling elements with IL-2 and has similar in vitro biologic activities (46). IL-15 activates natural killer cell proliferation, cytotoxicity, and cytokine production, and it also indirectly enhances TNFα production by T cells after IL-1 exposure. The IL-15 gene and protein are readily detected in RA synovial fluids and tissues (47).

The Th1 cytokine IL-17 is also found in RA synovial samples (48). This cytokine has many of the same functions as IL-1 and TNFα and can enhance cytokine and metalloproteinase production by FLSs (49). IL-17 can synergize with IL-1 and TNFα, thereby increasing the local inflammatory reaction. IL-12, which is produced by synovial macrophages, has been detected in RA and is known to bias T cell differentiation toward the Th1 phenotype (50). Local production of IL-12 could play a key role by directing the expression of Th1 cytokines in the rheumatoid joint. IL-18, which is a member of the IL-1 family and is also present in the RA joint, can synergize with IL-12 and enhance Th1 cell differentiation (51). In addition to enhancing IFNγ production, IL-18 has IFN-independent proinflammatory functions that enhance macrophage activation.

Even more interesting than a direct contribution by soluble lymphokines is the potential contribution of cell-cell interactions that are independent of antigen-specific responses. Direct contact between T cells and fibroblasts or between T cells and macrophages can enhance cytokine, metalloproteinase, and prostanoid production (52, 53). This type of interaction does not require viable T cells, because isolated cell membranes can serve as the stimulus. Such activities depend on surface proteins, including membrane-bound cytokines (such as TNFα) and adhesion molecules (such as lymphocyte function-associated antigen 1). IL-15 causes T cells to enhance the production of TNFα by macrophages in a contact-dependent reaction (54).

The evidence for cell contact-mediated lymphocyte function suggests a role for T cells in RA that is unlike the role envisioned 10 years ago (see Table 1). The traditional paradigm assumed that T cells in the joint respond to a specific antigen(s). However, cell-cell contact models do not presuppose any specific response and depend only on the localization of memory T cells adjacent to synoviocytes or macrophages. As discussed in our original editorial, T cells with a memory phenotype are drawn into the joint by locally produced chemokines and because of the interaction between adhesion molecules expressed by circulating lymphocytes and their counterreceptors on the synovial vascular endothelium. While the number of chemokines and adhesion molecules is better defined today (and the number has increased exponentially), the same conclusions apply. For instance, chemokine receptors CCR5 and CXCR3 are preferentially expressed by Th1 cells, and their respective ligands are displayed by the rheumatoid synovium (55, 56).

Table 1. T cell participation in rheumatoid arthritis
Proinflammatory cytokine production
 Th1 activation (e.g., interleukin-17)
 Lack of Th2 activation (e.g., interleukin-4)
Osteoclast activation and bone erosion
 Receptor activator of nuclear factor κB ligand
Tumor necrosis factor α induction
 Response to macrophage-derived interleukin-15
Cell-cell contact
 Matrix metalloproteinase induction
 Proinflammatory cytokine induction
Antigen-specific responses
 Th1 activation
 Antigen-presenting cell activation

Thus, nonspecific influences draw certain kinds of T cells into the joint, where they can then either respond in an antigen-specific manner or serve as an antigen-independent stimulus for local inflammation. What is not clear is how cell-cell interactions can play an important role if the synovial intimal lining, the source of most cytokines and metalloproteinases, is essentially devoid of T cells. Lymphocytes must pass through the lining en route to the synovial fluid, but histologic and ultrastructural evaluations of synovium show very little evidence of contact between T cells and synovial lining cells. Although transient cell-cell interactions might occur while the lymphocytes migrate into the joint cavity, their functional consequences are still undefined. These cell contacts are probably more important in the sublining space and serve to activate infiltrating macrophages.

Reinterpretation of the cytokine network model: how cytokines (and T cells) perpetuate RA. How should we now view the cytokine network hypothesis? We believe that many of our original conclusions still hold true today. T cell cytokines, especially Th2 cytokines, are nearly absent in the rheumatoid joint, whereas Th1 cytokines appear to be produced in relatively low amounts compared with other chronically inflamed tissues. Many of the T cell influences on the cytokine network are antigen-independent and may not involve the traditional signaling pathways implicated in immune-mediated diseases. On the other hand, unquestionable evidence of T cell involvement has been documented, including the influences of IL-17 and direct cell-cell interactions. Whether these are primary or secondary remains to be determined. Perhaps a better title for the original editorial would have been, “How important are antigen-specific T cells in chronic rheumatoid synovitis?”

Much of what we said then about macrophages and FLS is still consistent with recent data. Although the cytokine network is increasingly complex, the general rules regarding their cellular sources in the rheumatoid joint and their paracrine and autocrine interactions remain the same. In fact, the efficacy of anticytokine therapy in the treatment of RA provides convincing proof that cytokine networks are critically involved in the pathogenesis of RA. However, clinical experience with TNF inhibitors also indicates that considerable heterogeneity exists among patients and that no single cytokine is likely to be the key. Since only about one-third of patients treated with TNF inhibitors have dramatic responses (i.e., American College of Rheumatology criteria for 50% improvement or better) (57), additional cytokines must influence the disease, and their relative contribution might vary between patients and, perhaps, between different stages of disease. Finally, the observation that cytokine blockade does not cure RA and that continuous treatment is necessary implies that the cytokine networks are not autonomous and require additional exogenous stimulation.

Joint destruction in late rheumatoid arthritis: What comes after T cells?

Destruction is often considered to be the consequence of chronic synovial inflammation, presumably immune (T cell) mediated, acting on synovial lining cells. Thus, the mainstays of treatment for RA have been antiinflammatory therapies, ranging from cyclooxygenase inhibitors and corticosteroids to biologic agents, such as antibodies to cytokines or inhibitors of their receptors. Without a doubt, both processes—inflammation and destruction—are simultaneously present in the joint, but there are conflicting theories on whether in RA, they proceed sequentially or in tandem; whether the synoviocytes responsible for tissue injury have acquired a unique “transformed or aggressive” phenotype; whether all FLS are potentially injurious, or whether that property resides in a subpopulation; and finally, whether clinical studies can shed light on these important questions.

In RA, the destructive changes in cartilage and bone are associated with an aggressive granulation tissue (pannus), which is composed of proliferating FLS, new blood vessels, macrophages, and in the case of bone erosions, osteoclasts. Lymphocytes and granulocytes are conspicuously rare in these lesions, and this raises the possibility that synoviocyte-mediated invasion can be independent of T cell control.

In this section, we consider whether the late destructive stage of RA represents progression from a T cell-mediated process to one that is largely centered on autonomous FLS aggression. Considerable evidence both in animal models and in humans supports this model. If it is true, therapeutic interventions specific for traditional immune responses would require additional activity directed toward non-T cell elements of invasive synovium.

Synovial fibroblasts in rheumatoid arthritis: escape from T cell control. One of the essential tenets of this hypothesis is that synovial fibroblasts attain a degree of independence from T cell control in late destructive RA. While the origin of this model lies in careful observation of the disease in humans, animal models of chronic inflammatory polyarthritis indicate that FLS can be injurious in the absence of T cells. For instance, DBA/1 mice that are deficient in CD4 or CD8 lymphocytes develop CIA (albeit in a somewhat attenuated form), and recombination-activating gene 1 (RAG-1)-deficient DBA/1 mice that lack mature lymphocytes develop synovial hyperplasia and bone and cartilage destruction when immunized with collagen, but have only minimal articular inflammation (58, 59).

Additional evidence of autonomous activity of synovial fibroblasts comes from experiments in which a modified human TNFα transgene was introduced into mice, resulting in an overexpression of the cytokine. Shortly after birth, those animals developed a severe inflammatory arthritis with joint destruction that was not ameliorated by treatment with IL-4, IL-10, or IL-13 (60, 61).

Moreover, in vitro, isolated synoviocytes from mice with a deletion of the 3′ AU-rich regulatory element of the TNF gene spontaneously produced large amounts of the cytokine and matrix-degrading enzymes. Eliminating all adaptive immunity by breeding the TNF-mutant mice onto a RAG-1 background did not change the articular disease, implying that stromal elements and/or synovial fibroblasts are responsible (62). Kollias and colleagues believe that “these observations underscore the central role of synoviocytes as both target and effector cells capable of initiating and maintaining the arthritogenic response” (62).

Perhaps the most abundant cytokine in RA synovial fluid and tissues is IL-6, which is primarily an FLS product. Although IL-6 can possess pro- or antiinflammatory effects depending on the specific model examined, the majority of evidence suggests that it is deleterious in inflammatory arthritis and supports antigen-independent inflammation. For instance, systemically immunized mice develop a chronic inflammatory destructive arthritis after the introduction of antigen into the knee joint. Compared with wild-type mice, IL-6 knockout mice (IL-6−/−) developed acute joint inflammation, but a chronic destructive phase did not develop (63). Similarly, in zymosan-induced arthritis, a nonimmunologic joint disease, injection of zymosan into the knees of IL-6−/− mice caused acute inflammation, but not a chronic arthritis, suggesting an important nonimmunologic role for IL-6 in the propagation of arthritis (64).

Experimental models of chronic inflammatory arthritis are useful for testing hypotheses and therapies. But, rodents do not get RA, so it is critically important to demonstrate in vivo the pathogenic capabilities of isolated human synoviocytes. For instance, when cultured and passaged RA FLS, devoid of hematopoietic cells, were engrafted with articular cartilage beneath the kidney capsule of SCID mice, they formed a pannus that penetrated the cartilage, whereas neither normal nor osteoarthritis FLS and dermal fibroblasts were invasive (65). Similar results were seen in vitro when RA FLS were cocultured with articular cartilage explants or artificial matrices (66, 67).

Besides supporting the notion of autonomy of RA FLS, these studies are cited as evidence for the “transformed” phenotype of RA synoviocytes. Many other aspects of RA FLS biology support this notion (for review, see refs. 68 and 69), but as with almost all cellular abnormalities in RA, the question is whether the synoviocyte alterations are a consequence or a cause of the disease. Does the cytokine-rich milieu of the RA joint, including IL-1 and TNFα and growth factors such as basic fibroblast growth factor, platelet-derived growth factor, and TGFβ, favor the proliferation of FLS? Likewise, does the genotoxic articular environment with reactive oxygen and nitrogen species lead to somatic mutations in the p53 tumor suppressor gene and give a growth advantage to a minority population of FLS (70)? Whatever the reason, there is little question that RA FLS are aggressive and are different from conventional fibroblasts.

How synovial fibroblast autonomy might lead to joint destruction. If, as we suggest, years of rheumatoid inflammation lead to FLS autonomy, then RA and normal synoviocytes should behave differently in vitro. In fact, cultured RA FLS display a number of unique “transformed” features, namely, growth in suspension or semisolid medium, loss of contact inhibition, invasiveness, and expression of a number of early response genes (68). This phenotype is permanently imprinted and does not require continued exposure to T cells or their products. Not all synoviocytes in RA necessarily display this phenotype, and cells isolated from bone and cartilage erosions demonstrate behavior that is distinct from that of cells from other sites.

These observations suggest that considerable functional heterogeneity exists among FLS. For instance, cells isolated from RA pannus directly eroding cartilage have a distinctive morphology and features of both FLS and chondrocytes (“pannocytes”) (71, 72). Another study showed that FLS isolated from invasive pannus are oligoclonal, whereas nonerosion FLS are polyclonal (73). Cells from pannus lesions also produce more TGFβ and platelet-derived growth factor than do those from nonpannus areas, which possibly accounts for their growth advantage. This might also relate to genetic alterations that occur due to the genotoxic environment in the joint, including somatic mutations in the p53 and H-ras genes (74, 75).

Dedifferentiation of FLS and expansion of rare resident pluripotent mesenchymal cells appear to result from decades-long exposure to cytokines and reactive oxygen species in the rheumatoid synovium (76). Alterations in synoviocyte heterogeneity and gene expression patterns could have profound implications for matrix destruction, tissue remodeling, and repair. For instance, the receptor for bone morphogenetic protein is a surface marker that is expressed by primitive cells. Although <1% of normal FLS display this receptor, ∼10% of cultured RA FLS isolated from patients requiring joint replacement surgery are positive, and the cells are localized to the cartilage-pannus junction (77). The notion that mesenchymal stem cells can be isolated from joint tissue is now well established, and these cells can be induced to form bone, cartilage, and adipocytes (78, 79). Likewise, genes associated with mesenchymal cells in embryonic development, including Frizzled and Wingless, are expressed in RA, but not normal, synovial tissues (80). These genes, especially Wnt-5A, regulate IL-6 and IL-15 expression in RA FLS and contribute to the cytokine milieu. The dedifferentiated phenotype is maintained in culture long after T cells and macrophages have been removed from FLS cultures (81). Presumably, recapitulation of the embryonic gene patterning leads to T cell-independent remodeling of the joint architecture.

The erosion of bone in RA also results from the maturation and activation of osteoclasts (82). Production of this cell lineage and the progression of joint destruction in animal models can be supported by either T cells or synovial fibroblasts through the expression of receptor activator of NF-κB ligand (RANKL) (83). Osteoclasts arise from myelomonocytic precursors, generally in intimate association with pluripotent mesenchymal cells (84, 85), but unsupplemented cultures of a fraction of normal blood containing both mononuclear cells and circulating mesenchymal precursor cells develop osteoclasts after 2–3 weeks (76). RA synovial tissues contain osteoclasts at sites remote from bone and in the pannus adjacent to bone erosions. Monocytes and bone morphogenetic protein receptor-bearing cells also reside in the same locations (77). In vitro, long-term culture of freshly isolated RA synovial tissues or coculture of adherent RA FLS with peripheral blood mononuclear cells leads to the formation of bone-resorbing multinucleated cells, but neither osteoarthritis FLS nor dermal fibroblasts form similar cells (86–88).

Joint damage in RA is equated with radiographic evidence of bone erosion and cartilage loss. Both are often considered to be consequences of synovial inflammation. However, some clinical observations have challenged this dogma (89, 90), and two recent cautionary studies are noteworthy. Lewis rats immunized with CFA develop an inflammatory, destructive polyarthritis. When given osteoprotegerin, a decoy molecule that blocks the interaction of members of the TNF ligand and receptor families (TNF-related activation-induced cytokine/RANKL; osteoclast differentiation factor/osteoprotegerin ligand) on osteoblast and osteoclast precursors, respectively, there was no effect on joint inflammation, but cartilage and bone damage was completely eliminated (91). A similar result was observed in a study of anti-TNFα antibody treatment of RA; joint destruction (determined by radiographic assessment) did not progress over 54 weeks in the treated group, regardless of whether or not inflammatory measures improved (92). These data suggest that bone erosion and cartilage erosion are distinct entities that are independent of either the inflammatory response or a T cell-mediated intraarticular process.

An updated model: T cell independence and interdependence from beginning to end?

This review was designed to show that RA is not just about T cells; rather, it is a multicellular process with 3 distinct, juxtaposed phases (see Figure 1). The induction phase precedes clinically apparent joint inflammation, may be antigen independent, and involves normal joint constituents (i.e., synovial lining cells) and their cytokines and chemokines to bring blood cells into the joint, where they can initiate and perpetuate an inflammatory phase. The typical clinical picture of RA ensues, driven by lymphocytes, macrophages, and antigen-presenting cells. This aspect may involve specific antigens—either foreign or native and either integral to the joint or presented in the periphery—but the phlogistic byproducts of the immunocompetent cells extends the inflammation by engaging and activating leukocytes. The destructive phase can be without symptoms (at least in the beginning), may be antigen independent, and is supported by mesenchymal elements, such as fibroblasts and synoviocytes. Bone erosion and cartilage damage probably develop independently. The former is caused by osteoclasts in direct contiguity with bone, while cartilage dissolution results from proteolytic enzymes produced by synoviocytes in the pannus or released from granulocytes in the synovial fluid bathing the articular surface. Connections between the 3 phases and their interactions are still poorly understood, but eventually, the successful treatment of RA will depend on dealing with all the elements of this multicellular process.

Figure 1.

Schematic diagram of disease mechanisms that likely occur in various phases of rheumatoid arthritis. Innate immunity activates mesenchymal cells and macrophages (MF) in the earliest phases, which can focus a subsequent immune response to the synovium. Antigen-specific responses, although not proven, are probably most important in the inflammatory phases, while macrophage and fibroblast-like synoviocyte (FLS) cytokines dominate. Direct T cell contact can also activate other cells in an antigen-independent manner. In the latter phases of disease, many cell types activate osteoclasts (OC) through the receptor activator of nuclear factor κB (RANK)-RANK ligand system, although FLS likely provide the greatest stimulus. Autonomous activation of FLS might contribute to this process.