Article Text


The rationale for the current boom in anti-TNFα treatment. Is there an effective means to define therapeutic targets for drugs that provide all the benefits of anti-TNFα and minimise hazards?
  1. Marc Feldmann,
  2. Jan Bondeson,
  3. Fionula M Brennan,
  4. Brian M J Foxwell,
  5. Ravinder N Maini
  1. Kennedy Institute of Rheumatology, 1 Aspenlea Road Hammersmith London W6 8LH
  1. Professor M Feldmann.


Progress in understanding mechanisms of disease are necessary to usher in major changes in treatment. A new era in rheumatoid arthritis (RA) and related chronic autoimmune/inflammatory diseases is now beginning, with a variety of anti-TNFα treatments licensed for use in both RA and Crohn's disease. The rationale for this new treatment lies in an understanding that cytokines are critical, rate limiting molecules lying at the heart of the chronic autoimmune/inflammatory disease process. This understanding was developed from the critical evaluation of a hypothesis that was proposed linking cytokines, antigen presentation and autoimmunity in 1983. Detailed analysis focusing on the major site of the disease, the rheumatoid synovium was essential to developing indications that blockade of TNFα might be efficacious. This clue was validated using anti-TNFα treatment of an animal model of RA, murine collagen induced arthritis, and by immunohistochemical demonstration of upregulated TNF and TNF-R expression in the synovium. With this three pronged rationale, the authors were able to convince Centocor, Inc, which had developed a chimaeric anti-TNFα antibody for use in sepsis, to work with them to test the concept that TNFα blockade would be beneficial in RA. With the success of that first trial, other companies have subsequently tested their anti-TNF strategies successfully. Current interests extend to understanding the processes that regulate TNF production in the rheumatoid joint. Progress in this area is discussed, using adenoviruses to infect normal macrophages and rheumatoid synovium.

  • rheumatoid arthritis
  • anti-TNFα

Statistics from

Cytokines are protein mediators that are intimately involved in all biological processes, including cell growth, differentiation, inflammation, immunity, etc. Their biochemical characterisation began in the 1970s and culminated in the cloning of the cDNAs for interferons α, β, γ, IL2, TNFα, and LT between 1981 and 1984.1-3 This provided new tools to evaluate cytokine expression and regulation in rheumatoid synovium.

Rationale for anti-TNFα treatment (table 1)

We rapidly found that essentially all the cytokines were consistently expressed in RA tissue, although in different quantities, regardless of the stage of disease, or its treatment (reviewed by Feldmann et al 4). This suggested to us that a key aspect of a chronic disease such as RA might be that, unlike experimentally induced cytokine production (for example, with mitogens or antigen), which is transient, diseased induced cytokine production is continuous. To test this idea, we cultured dissociated rheumatoid synovial cultures without extrinsic stimulation, and found that cytokine production in this in vitro model of RA was indeed spontaneous and prolonged, for over six days.5 Moreover, we had discovered a system for asking questions about what drives the continuous endogenous cytokine production in RA synovium.

Table 1

Rationale for anti-TNFα treatment in rheumatoid arthritis

Our approach to this question, as immunologists, was to use neutralising antibodies. Based on the observations that IL1 induced the destruction of cartilage and bone, we studied the regulation of IL1 in the RA synovial cultures. At least half a dozen signals present in the RA synovium were known to regulate IL1, for example, IFNγ, GM-CSF, TNFα, immune complexes as well as IL1 itself, and so it was a surprise when we found that blocking TNFα abolished IL1 bioactivity (fig 1).6 This was our first clue that blocking TNFα might be beneficial in RA, as it would also downregulate the equally pro-inflammatory cytokine, IL1. It led us to investigate the ability of anti-TNRα antibody to downregulate the production of other pro-inflammatory cytokines, and we found that GM-CSF,7IL6, IL88 were also downregulated. This has led us to the concept that the pro-inflammatory cytokines, in a diseased tissue, were not independently regulated, but were coordinately regulated. One of the key coordinators was TNFα, to a lesser extent IL1.

Figure 1

Inhibition of TNFα in rheumatoid synovium downregulated IL1: the first clue that TNFα was a therapeutic target. Modified and reprinted with permission from Brennan et al. Lancet 1989,ii:244–7.

The cytokine cascade concept is illustrated in figure 2, cytokine equilibrium in figure 3.

Figure 2

Cytokine cascade. By a multitude of blocking experiments with anti-TNFα and IL1 receptor antagonist, it was shown that in rheumatoid synovium both the pro-inflammatory and anti-inflammatory mediators are regulated by TNFα and IL1. Modified and reprinted with permission from Feldmann, Brennan and Maini. Cell 1996;85:307–10.

Figure 3

Cytokine disequilibrium. Modified and reprinted with permission from Feldmann, Brennan and Maini. Cell 1996;85:307–10.

Before the above studies, based on the presence and overlap in properties of the multitude of pro-inflammatory mediators, commonly termed “redundancy”, it was thought that cytokines were poor therapeutic targets, based on the premise that if only one was blocked, the others would be sufficient to maintain the pro-inflammatory activities. Clearly it was impractical to neutralise all the pro-inflammatory cytokines simultaneously.

To validate the concept that TNFα was a prime therapeutic target in RA, two further types of studies were performed. One was to verify by immunohistology that in flash frozen biopsy specimens of RA synovium, TNF and TNF receptors were upregulated.9-11 The second was to evaluate anti-TNFα treatment in an animal model of RA, collagen induced arthritis. We and others found that anti-TNFα reduced joint inflammation and joint destruction, even if given in a therapeutic mode, that is after disease onset.12-14

What regulates TNFα synthesis by macrophages and rheumatoid synovium?

The work we pioneered in the rationale for anti-TNFα treatment and its validation in both open 15 and randomised placebo 16 controlled trials with the chimaeric anti-TNFα antibody of Centocor (cA2), now termed Infliximab or Remicade, has led to multiple confirmatory studies with other antibodies (for example, CDP571, Celltech/Bayer17) TNF-R Fc fusion proteins (Etanercept, Enbrel p75TNF-R Fc, Immunex/AHP18 or lenercept, p55 TNF-R Fc, Roche19), and more recently another antibody D2E7 (Knoll/CAT).20 Thus the principle of TNF as an effective therapeutic target is amply confirmed in clinical studies.

However TNF, as a multifunctional cytokine is needed in small amounts at times for normal immune function. Hence TNF knockouts are at risk of infection with intracellular microorganisms such as listeria.21 Is this the case in anti-TNFα treated patients? Clinical trial data so far suggest that with both Remicade and Enbrel the risk of infection is not higher than in the placebo treated controls. In theory this may not be the case for patients treated in the longer term. It is conceivable that the mechanism of TNFα production in RA tissue may be different from that operating in the immune response to infections, and if so, it would be an intriguing possibility to block the “pathological” TNFα production in RA, but not “physiological” TNFα in an immune response.

Another important reason for seeking to understand what drives TNFα production in macrophages and RA synovium is the fact that most orally available drugs act on intracellular signalling pathways. A key problem is how to study signalling in normal macrophages. These cells, like most normal cells do not transfect easily. Viruses are adapted to inserting their genes into cells, and getting them expressed, and thus we sought to use viruses to delivery inhibitory genes to macrophages.

By using adenoviruses, we have developed conditions that permit us to infect >95% of normal macrophages.22 The TNFα promoter has numerous NF-κB sites, and so we used an adenovirus overexpressing IκBα, a natural inhibitor of NF-κB to evaluate the role of NF-κB in macrophage cytokine production. The results were interesting. For this single cytokine, for example, TNFα, the role of NF-κB varied markedly depending on the stimulus. LPS, PMA and UV light were markedly NF-κB dependent (that is, blocked), but in contrast zymosan and anti-CD45 induced TNFα were not blocked23 (fig 4). As NF-κB has been reported to prevent apoptosis,24 the above results with zymosan and anti-CD45 verify that the mechanism of NF-κB inhibition of TNF production was not by cell death.24 Of course this has been verified in other ways also, by MTT propidium iodide and dye exclusion assays.22 23

Figure 4

The mechanism of TNFα production in macrophages depends on the stimulus. LPS is NF-κB dependent, zymosan is not. Modified from Bondeson et al. J Immunol 1999;162:2939–45.

These results indicated that the technique of adenoviral infection was suitable to use on RA synovial tissue, at least for analysing the macrophage TNRα production, which is the major, but not the only component of TNFα production. As we wished to study cytokine production in the RA joint in detail, we first evaluated the infectability of some of the critical subsets of cells. Overall, >90% of cells were infected, assessed using an adenovirus expressing β galactosidase and flow cytometry.25 When subsets of cells were analysed by two colour flow cytometry, the results were interesting. As expected CD14+CD3 cells, chiefly macrophages were infected >95%, CD14CD3+ cells, T lymphocytes were also infected at >80%. This was not predicted by prior reports, which reported the low infectability of T lymphocytes. CD14CD3 cells, chiefly fibroblasts, were as anticipated readily infected.

Having established that RA synovial cell mixtures were infectable by adenovirus, we set out to ascertain if the infection with AdvIκBα would change the cytokine profile produced. The results were clear cut. First to be assessed was TNFα, this was reduced by 70%, mean of five synovial samples (fig 5).22 Subsequently other pro-inflammatory mediators were assessed. The results varied. IL6 production was blocked 90%. However, IL1 production was different being reduced by only 40%, and IL8 production even more variably reduced perhaps 30−40%.25 These results emphasise that while NF-κB is an important transcription factor driving the inflammatory process, there are also other important pathways, the nature of which remain to be elucidated.

Figure 5

Rheumatoid synovial TNFα production is NF-κB dependent. Modified from Foxwell et al. Proc Natl Acad Sci 1998;95:8211–15.

Evaluation of the effects of AdvIκB on the production of anti-inflammatory mediators led to quite different results. Thus there was minimal effects of over-expressed IκBα on IL10, IL1 receptor antagonist or IL11. There was a moderate diminution of soluble TNF-R, which could be restored by adding inflammatory mediators.23 25

In view of the fact that it is the disequilibrium of pro-inflammatory and anti-inflammatory mediators that determines the “outcome” of an inflammatory process4 the effect of blocking NF-κB in RA joints would thus seem to be very beneficial, as it diminishes pro-inflammatory mediators without reducing the anti-inflammatory.

Does NF-κB influence joint destructive processes? This is an interesting question, which is not fully understood. The conventional wisdom, based on analysis of the promoters of metalloproteinases (MMPs) is interpreted to suggest that AP-1 is the critical regulator of matrix destructive MMPs.26 However, this concept has not been critically tested. We explored the effect of AdvIκBα on the expression of MMP-1, MMP-3 and MMP-13 and their inducible inhibitor from dissociated rheumatoid synovial cultures. These studies showed that the production of MMPs, assessed by immunoassay was markedly downregulated by excess IκBα. In contrast, TIMP-1 was upregulated.25 These results were verified by time course analysis over seven days and were stable. As pro-inflammatory cytokines upregulated MMPs these results by themselves do not establish that the effect is direct. It could be indirect, because of IκBα downregulation of pro-inflammatory mediators. Indirect effects can, in principle be overcome by excess proinflammatory mediators. These experiments showed that there seemed to be both direct and indirect effects.25


TNRα blockade provides considerable benefit in RA. The exact means of doing so may not matter, as multiple antibodies, first cA2 (Remicade, Centocor) then CDP571 (Celltech/Bayer) have all yielded significant results, as have TNF receptors, Ig fusion proteins TNF-R p75 Fc (Enbrel, Immunex/AHP), TNF-R p55 (Lenercept, Roche) or Amgen's monomeric TNF-R. These products, while all beneficial may yet differ, as the monoclonal antibodies are specific for TNFα, but the receptors can also block lymphotoxina.27 This may be more immunosuppressive.

We have reviewed here data that adenoviruses offer an efficient means of gene delivery and thus a powerful tool for defining what is really happening inside normal or pathological cells. The results indicate that blocking NF-κB in rheumatoid joints would deliver all the benefits of TNFα blockade. How to do it in a safe and effective manner remains a challenge.


The majority of the work at the Kennedy Institute of Rheumatology is funded by the Arthritis Research Campaign. The clinical studies were funded by Centocor, Inc. Professor Feldmann is a member of the Scientific Advisory Board of Centocor, Inc.


View Abstract

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.