Article Text


Cartilage destruction and bone erosion in arthritis: the role of tumour necrosis factor α
  1. Richard O Williams,
  2. Marc Feldmann,
  3. Ravinder N Maini
  1. Kennedy Institute of Rheumatology, 1 Aspenlea Road, London W6 8LH
  1. Dr Williams (r.williams{at}

Statistics from

Rheumatoid arthritis (RA) is a chronic disabling disease characterised by synovitis, destruction of cartilage and bone and, ultimately, loss of joint function. A great deal of research in rheumatology over the past two decades has focused on identifying cytokines and other mediators responsible for the inflammatory and degenerative processes in RA, with the aim of developing specific inhibitors or antagonists of therapeutic value. A key question in this research has been whether there is a degree of hierarchy in proinflammatory cytokine expression in RA, such that inhibiting the activity of a cytokine high up in the inflammatory cascade has an impact on the expression of downstream mediators of inflammation and joint damage.

Of the many cytokines thought to contribute to the inflammatory and degenerative changes that occur in RA, tumour necrosis factor α (TNFα) has emerged as being of major pathological significance. For example, TNFα was identified in the synovial membrane, and particularly at the cartilage-pannus junction, of patients with RA,1 and was found to be spontaneously produced by cultured synovial cells derived from RA patients.2 The in vitro properties of TNFα were also found to be consistent with a pathogenic role. Thus, TNFα promoted cartilage and bone resorption3 4 and induced the release of prostaglandin E2 and collagenase by synovial cells.5 In addition, TNFα was reported to play a part in fibrosis6and to facilitate inflammatory cell infiltration by promoting the adhesion of neutrophils and lymphocytes to endothelial cells.7 8 Evidence was also provided to suggest that TNFα induces the production of interleukin 1 (IL1), another potential mediator of joint damage in RA,9 as antibodies to TNFα were shown to diminish the production of IL1 by rheumatoid synovium derived mononuclear cells.2 Using a similar experimental system, it was subsequently shown that TNFα also induces the expression of GM-CSF.10 These findings imply that the proinflammatory activities of TNFα in vivo are likely to be greatly amplified because of the fact that TNFα triggers the production of additional proinflammatory mediators.

In this report experiments are described that are aimed at validating the concept of an in vivo role for TNFα in joint damage using the collagen induced arthritis (CIA) model. In addition, the role of T cells in the arthritic process is discussed and an analysis is made of the effect of a combined therapeutic strategy that targets not only TNFα, but also CD4+ T cells.

CIA as a model for RA

Animal models of arthritis have proved to be extremely useful in elucidating pathogenic mechanisms of relevance to RA and for evaluating the effect of new forms of treatment. The CIA model has been used extensively in studies of immunotherapy, mainly because of its pathological and immunological similarities to human RA.11Thus, both RA and CIA exhibit similar patterns of pannus formation, erosion of cartilage and bone, fibrosis, and ultimately, loss of articular function.12 Like RA, the pathogenesis of CIA is thought to involve both humoral and cell mediated immune mechanisms and susceptibility to both diseases is associated with specific MHC class II genes, suggesting that similar immunological mechanisms are operating in both diseases.11

CIA has been shown to occur in rats, mice and primates after immunisation with type II collagen.13-15 In mice, immunisation with heterologous collagen results in a relatively acute form of arthritis whereas immunisation with autologous collagen results in a more chronic, though milder, form of the disease.16 17

Effect of TNFα blockade in CIA

One way of assessing the importance of a particular mediator in disease is to block its activity using neutralising antibodies, and a number of studies have focused on the effect of TNFα blockade during the induction phase of CIA. These studies showed that treatment of mice with monoclonal or polyclonal anti-TNFα antibodies, or soluble TNF receptors, reduced the severity of arthritis when administered before the onset of clinical arthritis.18-20

Subsequently, we assessed the effect of anti-TNFα treatment in mice with established CIA.19 DBA/1 mice were immunised with type II collagen in complete Freund's adjuvant. The mice were inspected daily and each mouse that exhibited clinical signs of arthritis was randomly assigned to one of three treatment groups. The mice were then given twice weekly intraperitoneal injections of TN3-19.12 (anti-TNFα mAb), L2 (isotype control) or phosphate buffered saline (PBS) over a period of 14 days. The half life of TN3-19.12 in mice had been previously estimated to be around seven days.21 The doses of antibody used ranged from 50–500 μg per mouse. The results showed that that there was a dose dependent reduction in the severity of arthritis following treatment with anti-TNFα mAb (fig 1). At the end of the treatment period, arthritic paws were decalcified, sectioned and stained with haematoxylin and eosin. Individual joints were then graded according to the histopathological severity of arthritis. It was found that anti-TNFα treatment reduced the histological severity of arthritis and protected joints from erosive changes (fig2).

Figure 1

Effect of anti-TNFα mAb on clinical progression of established CIA. Arrows indicate time of injection. L2 is an isotype control mAb. (A) Clinical score. The scoring system was based on the following criteria: 0 = normal, 1 = slight swelling and/or erythema, 2 = pronounced oedematous swelling, 3 = ankylosis. Each limb was graded, giving a maximum score of 12 per mouse. (B) Paw swelling. Expressed as the percentage increment in paw width relative to the paw width before the onset of arthritis. Modified from Williams et al.19

Figure 2

Histopathological assessment of joints of arthritic DBA/1 mice treated with anti-TNFα. The scoring system was as follows. Mild, minimal synovitis, erosions limited to discrete foci, cartilage surface intact. Moderate, synovitis and erosions present but normal joint architecture intact. Severe, extensive erosions, joint architecture disrupted. Data taken from Williams et al.19

Soluble TNF receptors are understood to play an important physiological part in regulating the activity of TNFα, and it was subsequently shown that two soluble TNFR constructs were effective in established CIA. In the first study, a p75 TNFR-Fc fusion protein was found to reduce the severity of CIA whether given before or after the onset of the disease.22 In another study, we showed that a p55 TNFR-Ig fusion protein was effective in reducing both the clinical severity of established CIA.23 Furthermore, when the joints were examined by histology treatment with TNFR-Ig was found to have exerted a dose dependent protective effect on joint erosion (table1). The conclusion drawn from these studies was that TNFα is involved in the pathogenesis of CIA. In addition, the findings provided support for the testing of anti-TNFα antibody treatment in human RA.

Table 1

Significant reductions in the proportions of proximal interphalangeal (PIP) joints exhibiting erosion of cartilage and/or bone after treatment with TNFR-Ig fusion protein for a period of 10 days

Effect of TNFα blockade in other models of arthritis

Although CIA is probably the most widely used animal model of arthritis, many other models exist, all of which mimic human RA to a greater or lesser extent.24 Issekutzet al 25 carried out a study, using neutralising antibodies, into the respective roles of TNFα, IL1α and IL1β in inflammation and infiltration of polymorphonuclear leucocytes (PML) and T lymphocytes in established adjuvant arthritis in rats. Treatment with anti-IL1α and anti-IL1β on day 5 of arthritis did not significantly affect infiltration of PML or T cell infiltration into the joint. In contrast, anti-TNFα treatment reduced clinical scores, inhibited infiltration of PML by 40–50% and T lymphocytes by 30–50%. Combined treatment with anti-TNFα plus anti-IL1 produced an additive effect. It was concluded from this study that leucocyte infiltration in adjuvant arthritis is a strongly TNFα dependent disease with IL1 playing a relatively minor part. This contrasts with the situation in CIA, in which IL1 clearly plays an important part in both inflammatory and destructive processes.26 27

In antigen induced arthritis in rabbits neutralisation of TNFα was found to inhibit inflammatory changes in the joint during the acute phase of the disease but had little effect on the loss of proteoglycan from cartlilage in the long term.28 Similarly, anti-TNFα treatment was not found to prevent changes in cartilage proteoglycan synthesis or proteoglycan loss in antigen induced arthritis, zymosan induced arthritis, immune complex mediated arthritis or streptococcal cell wall induced arthritis in mice.29-31 It can concluded from these findings that arthritis models differ in terms of their response to TNFα blockade. Given that human RA is a relatively heterogenous disease, it seems probable that individual patients will also show different levels of response to anti-TNFα treatment.

Role of TNFα in arthritis: what can be learned from transgenic and knock out mice?

A further piece of evidence that helped to confirm the pathological role of TNFα in arthritis was the observation that huTNFα transgenic mice, which express human TNFα in a disregulated fashion, spontaneously develop arthritis that can be prevented by antihuman TNFα mAb.32 Histologically, huTNFα transgenic mice show synovial hyperplasia, accompanied by pannus formation and severe erosion of cartilage and bone. However, more recently it was shown that arthritis could also be prevented in huTNFα transgenic mice by the administration of an antagonistic antibody to the type I IL1R, which blocks the activity of both IL1α and IL1β.33 This observation indicates that the induction of arthritis by TNFα in this transgenic model is dependent on IL1. In addition, the finding indicates that TNFα and IL1 act in series, with TNFα inducing the expression of IL1. This conclusion is supported by the observation that cultured synovial cells from huTNFα transgenic mice, backcrossed onto a DBA/1 background, spontaneously produce increased levels of IL1β relative to non-transgenic littermates.34 This is also consistent with previously reported findings in human RA synovial cell cultures in which blockade of TNFα was found to inhibit IL1 production.2 This concept is also supported by the the results of an immunohistochemical study that we carried out into the kinetics of proinflammatory cytokine expression in CIA. In this study it was found that the expression of TNFα preceded, in a highly consistent fashion, the expression of IL1β during the early stages of arthritis.35

In another study, the role of signalling via the p55 TNFR in the development of arthritis was investigated. Mice lacking a functional gene encoding the p55 TNFR were generated by gene targeting and crossed to DBA/1 mice. Upon immunisation with type II collagen, p55 TNFR deficient mice were found to be partially resistent to CIA—that is, the incidence of arthritis was low and the severity was mild. However, it was also observed that in affected joints, arthritis could progress to the same destructive end stage as that observed in wild type mice. These findings indicate that signalling via the p55 TNFR plays an important part in CIA but that erosive arthritis can develop in the absence of the p55 TNFR.36

Anti-TNFα treatment in RA

Early clinical trials of anti-TNFα treatment in human RA were made possible by the availability of a chimeric anti-TNFα antibody, cA2 (infliximab, Remicade). In an open label trial of anti-TNFα treatment in 20 patients with severe longstanding RA, infliximab was found to ameliorate clinical symptoms, such as pain, stiffness and joint swelling and, in addition, to reduce laboratory-based parameters, such as serum C reactive protein and serum IL6.37 The effectiveness of TNFα blockade was subsequently confirmed in a multicentre, placebo controlled double blind trial involving 73 patients. Patients were given a single infusion of infliximab (1 or 10 mg/kg) or a placebo infusion. After four weeks, 44% of patients given infliximab at 1 mg/kg and 79% of patients given infliximab at 10 mg/kg showed a response at Paulus 20% criteria.38 In contrast, 8% of patients showed a positive response to placebo infusions. These findings demonstrated the efficacy of TNFα blockade in RA and paved the way for further trials with different TNFα inhibitors. For example, CDP571 (Celltech), a humanised anti-TNFα mAb was shown to have efficacy in RA.39 Similarly, a soluble p75 TNFR-IgG fusion protein (etanercept, Enbrel) has shown very clear evidence of efficacy in RA.40-42

The duration of these early clinical trials of TNFα blockade has in general been insufficient to demonstrate protective effects on joint erosion. However, preliminary data are beginning to emerge from longer term clinical trials showing that anti-TNFα treatment slows radiographic disease progression in RA. For example, data have been presented suggesting that D2E7, a fully human anti-TNFα mAb, was capable of slowing disease progression over a 12 month period.43

What is the role of T cells in driving joint destruction in arthritis?

The studies described above clearly show that TNFα is involved in the pathogenesis of arthritis but an equally important question is what is driving the production of TNFα in vivo. It is generally assumed that in CIA it is the CD4+ T cell that is driving inflammatory processes in the joint, as anti-CD4 treatment around the time of immunisation prevents the induction of arthritis.44 There is also strong evidence to suggest that CD4+ T cells are involved in the pathogenesis of human RA. For example, RA synovial mononuclear cell infiltrates are found to be dominated by CD4+ T cells, bearing markers of activation.45 46 In addition, epidemiological studies have identified a strong association between specific MHC class II alleles and susceptibility to RA, as well as severity of disease.47 48

Given the weight of evidence supporting the involvement of T cells in RA, it is perhaps surprising that the majority of placebo controlled clinical trials of depleting anti-CD4 mAb therapy have failed to provide convincing evidence of efficacy.49 Similarly, depleting anti-CD4 antibody treatment is found to be relatively ineffective in established CIA.44 50 One possible interpretation of these findings is that T cells are responsible for initiating the inflammatory response in RA and CIA but their influence declines to some extent as this inflammatory response becomes established.

Combination treatment

As discussed above, depleting anti-CD4 treatment is relatively ineffective in established murine CIA as well as in human RA. A study was therefore carried out to evaluate the effect of a combined therapeutic strategy designed to target not only CD4+ T cells, but also TNFα. Anti-TNFα mAb (TN3-19.12) was given alone or in combination with a cocktail of two rat IgG2b antibodies, YTS 191.1.2 and YTA 52 This anti-CD4 cocktail was shown previously to be highly effective in depleting CD4+ T cells and inducing tolerance.53

DBA/1 mice were immunised with type II collagen. Mice with CIA were then treated with an optimal or a suboptimal dose of anti-TNFα alone, anti-CD4 alone, anti-TNFα plus anti-CD4 or isotype controls. Anti-CD4 alone was relatively ineffective whereas anti-TNFα alone was effective at the optimal dose, as shown previously.19However, the combination of anti-CD4 plus anti-TNFα caused a much greater reduction in the severity of arthritis than either treatment alone54 with the synergistic effects of anti-TNFα and anti-CD4 being most evident with the suboptimal dose of anti-TNFα. A histological analysis of treated mice was then carried out that showed that suboptimal anti-TNFα alone reduced erosions in the PIP joints by 20% and anti-CD4 alone reduced joint erosions by 22%. However, combined anti-TNFα/anti-CD4 treatment reduced erosions by 72% (table2).54 In another study a synergistic therapeutic effect was demonstrated between anti-CD4 and a recombinant p55 TNFR-IgG fusion protein.23

Table 2

Proportions of PIP joints showing significant erosion of cartilage and/or bone in mice treated with anti-CD4 alone, anti-TNF alone or anti-CD4 plus anti-TNF

Mechanisms of action of anti-CD4 and anti-TNFα

To gain a better understanding of the mechanisms of action of anti-CD4, anti-TNFα and combined anti-CD4/anti-TNFα treatment in CIA, an experiment was carried out in which the effects of the different treatments on cellular infiltration, adhesion molecule expression, proinflammatory cytokine expression and level of Th1 activity was compared.55

This study provided a number of potentially significant findings. For example, it was found that although depleting anti-CD4 antibody treatment alone almost completely eliminated CD4+ T cells from the circulation, it was relatively ineffective in eliminating CD4+ T cells from the joints of arthritic mice (table 3) and inhibiting the level of Th1 activity, as judged by IFNγ production in draining lymph node cells. This finding provides a possible explanation for the relative lack of efficacy of anti-CD4 treatment in established arthritis—that is, the treatment fails to eliminate pathogenic Th1 cells and fails to eliminate T cells from arthritic joints.

Table 3

Comparison of the effect of anti-CD4, anti-TNFα and combined anti-CD4/anti-TNFα treatment on the severity of arthritis and the numbers of T cells and macrophages in the joints of mice with CIA. Mice were treated on days 1 and 4 of arthritis and MTP joints were studied by immunohistochemistry on day 6

Blockade of TNFα was found to cause a significant reduction in the expression of adhesion molecules and proinflammatory cytokines in the joint. It was also found that anti-TNFα treatment caused a marked decrease in IFNγ production by draining lymph node cells, suggesting that TNFα is involved in the T cell response to type II collagen. Anti-TNFα treatment was also found to inhibit the expression of IL1β in the joint (fig 3), thereby supporting previous findings regarding the role of TNFα as an inducer of IL1 expression in arthritis.2 33 34 Combined anti-TNFα/anti-CD4 treatment was found to be highly effective in reducing the level of adhesion molecule expression, in eliminating T cells and macrophages from the joint and in reducing IFNγ production by draining lymph node cells. However, a particularly striking finding was the ability of combined anti-CD4/anti-TNFα treatment to abrogate IL1β expression in the joint (fig 3). The significance of this finding lies in the fact that IL1, and particular IL1β, is a major contributor to the pathogenesis of arthritis.26 27 56 57 Thus, a possible reason for the profound ameliorative effect of combined anti-CD4/anti-TNFα treatment is that it provides effective suppression of IL1 expression, as well as blockade of TNFα activity.

Figure 3

Effect of anti-CD4, anti-TNFα and combined anti-CD4/anti-TNFα treatment on the expression of IL1β in the joints of mice with CIA (day 6 of arthritis). Mice were treated on days 1 and 4 of clinical arthritis. On day 6 of arthritis mice were killed and sections of MTP joints were analysed by immunochistochemistry. Each bar represents seven mice, each sampled at four levels throughout the joint—that is, for each cell marker, 28 sections were assessed. The results are expressed as the mean (SEM) number of cells per high power field. Data taken from Marinova-Mutafchieva et al.55


TNFα has now been established as an important mediator of pathology in RA, with the majority of patients showing a positive response to anti-TNFα treatment. Definitive data concerning the ability of TNFα inhibitors to prevent cartilage and bone erosion are awaited, but if it is possible to extrapolate the findings from CIA in mice to RA in humans, then a joint protective effect can be expected. However, two points should not be overlooked. Firstly, although the majority of patients respond to anti-TNFα antibody, a minority do not, and this begs the question of whether there is a specific subset of RA patients that are unresponsive to anti-TNFα treatment. Secondly, while anti-TNFα antibody treatment is clearly effective, a relapse of symptoms is generally observed after cessation of treatment, suggesting that anti-TNFα treatment affects the pathology of RA, but not its underlying autoimmune processes. Future goals in this field of research should be to learn how to treat patients who fail to respond to TNFα inhibitors, and how to maximise the duration of therapeutic effect of anti-TNFα. Combination treatment, targeted not only at TNFα, but also at the underlying pathogenic T cell response, may offer the possibility of achieving both goals.


View Abstract


  • Funding: The Kennedy Institute of Rheumatology, London, is supported by the Arthritis Research Campaign of Great Britain.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.