All three TNF antagonists have been associated with increased incidence of TB. However, this incidence seems to be lower for Eta than for Ifx [12,13], and the median delay between treatment initiation and occurrence of TB was shorter with Ifx [11]. Membrane-anchored TNF (mTNF) is expressed by activated macrophages and T lymphocytes [14,15]. Although Ifx and Eta both neutralise soluble TNF, Ifx binds more efficiently to mTNF than does Eta. Thus, Ifx but not Eta induces apoptosis of activated monocytes and lamina propria T lymphocytes from patients with CD [15,16]. The mechanism by which TNF antagonists reactivate latent TB is not fully understood. In animal models, TNF plays a central role in the containment of mycobacterial infections, and T cell-derived soluble TNF as well as mTNF are essential in protecting against Mtb infection [17-22].

Detection of latent TB is crucial before starting treatment with TNF antagonists because it requires a preventive treatment for TB reactivation before TNF antagonist administration [23-25]. However, this detection is difficult, especially in individuals vaccinated with the Bacille de Calmette Guérin (BCG). Diagnosis of latent TB may benefit from new in vitro assays testing the immune response against proteins such as culture filtrate protein (CFP)-10 and early secreted antigenic target (ESAT)-6, which are encoded in the genome of Mtb and of a few other mycobacterial species (Mycobacterium kansasii, Mycobacterium szulgai, and Mycobacterium marinum) but not in that of BCG and other mycobacteria. Presence of an immune response against CFP-10 and ESAT-6 is a relatively specific indicator of Mtb infection and has allowed for precise diagnosis of active as well as latent TB in several studies of BCG-vaccinated individuals [26-32].

Assays based on the quantification of IFN-γ-releasing cells allow diagnosis of active TB [31], recent primary infection [30,32], and latent TB [28]. Our findings extend these previous reports in several aspects. We show that, in addition to IFN-γ release, proliferative responses induced by Mtb-specific antigens are witnesses of a prior contact with Mtb. We also show that in patients with IMIDs and before initiation of treatment with TNF antagonists, ex vivo evaluation of anti-Mtb immune responses accurately reflects previous or latent TB because our biological findings correlate well with the current investigation of previous or latent TB. Additional studies are in progress to determine which combination of mycobacterial antigens and assays is optimal to diagnose latent TB and whether it compares favorably with the TST.

Our results also show that the intensity of anti-Mtb immune responses was preserved in patients with IMIDs as compared with patients without. This was observed regardless of the IMID considered, suggesting that neither an IMID nor its classical treatment significantly affects anti-mycobacterial CD4⁺ T lymphocytes. Independently of TNF antagonist treatments, there are controversies concerning the impact of IMID on anti-mycobacterial immune responses. Berg et al. showed a decreased response to PPD in patients with RA [39]. These authors measured IFN-γ production by enzyme-linked immunosorbent assay 7 days after in vitro stimulation of PBMCs, an assay clearly different from those we used (proliferation and immediate IFN-γ release tested by ELISPOT). Likewise, a decreased intensity of TST in patients with RA has been noted [40], but this finding was not confirmed in two other recent studies [41,42].

TNF antagonists increase the incidence and the severity of TB. It was thus of interest to demonstrate that TNF antagonists act in vivo on anti-TB immune cells and to define the type of immune response targeted by these agents. Proliferation in response to mycobacterial antigens remained unaffected 14 weeks after initiation of treatment with TNF antagonists. This negative finding is significant because in patients with latent TB and receiving no anti-TB treatment, reactivation of TB peaks 12 weeks after initiation of Ifx treatment [11]. In contrast to the preservation of proliferative responses, immediate release of IFN-γ was affected by the administration of TNF antagonists. The number of lymphocytes releasing IFN-γ within 18 hours after challenge with mycobacterial antigens significantly decreased 14 weeks after initiation of treatment, as compared with pre-treatment values. In most patients with previous or latent TB (group II patients), an anti-TB treatment was associated with TNF antagonists, raising the hypothesis that anti-TB treatment rather than TNF antagonist treatment decreased the number of IFN-γ-releasing lymphocytes. In patients with active TB, anti-TB treatment is indeed associated with a rapid decline of the anti-TB immune response, as assessed by ELISPOT [28]. In this condition, the ELISPOT assay evaluates not only resting memory cells, but also T lymphocytes recently activated in vivo by TB antigens. Such activated lymphocytes presumably vanish while TB replication stops. In patients with latent TB, it is unknown whether the anti-TB treatment affects the number of IFN-γ-releasing cells. Although we cannot rule out a contribution of anti-TB treatment in the decrease of IFN-γ-releasing cells we observed in patients from group II, several arguments indicate that treatment with TNF antagonists itself largely explains this decline. First, the two patients from group II who did not receive any anti-TB treatment did not differ from the others in terms of evolution of the anti-Mtb immune response. Second, IFN-γ release in response to PPD in group I patients (who received no anti-TB treatment) also decreased during treatment with TNF antagonists.

Circulating CD4⁺ T-lymphocyte numbers remain unaffected in patients treated with TNF antagonists [43,44]. We add functional findings to these previous studies. The preservation of proliferative responses we observed is consistent with the preservation of circulating CD4⁺ T-lymphocyte numbers. However, the decline of immediate IFN-γ release shows that treatment with TNF antagonists does not affect all aspects of the anti-mycobacterial immune response to the same extent. Memory T lymphocytes have been subdivided into effector/memory (T_EM) and central/memory (T_CM) subpopulations. Immediate release of cytokines is typical of T_EM, which proliferate poorly in response to antigens. In contrast, proliferation is a hallmark of T_CM (reviewed in [40]). This suggests that TNF antagonists affect anti-mycobacterial CD4⁺ T_EM while sparing T_CM. Expression of chemokine receptors also distinguishes T_EM from T_CM (CCR7 and CD62L) [45]. However, lack of defined combinations of Mtb peptides and HLA (human lymphocyte antigen) class II tetramers makes difficult the analysis of chemokine receptors on Mtb-specific CD4⁺ T lymphocytes.

The impaired IFN-γ release in patients treated with TNF antagonists indicates that ELISPOT assays should not be used to diagnose previous or latent TB in patients on treatment with TNF antagonists, whereas proliferative assays appear more reliable. Recently approved commercial assays for the in vitro detection of anti-TB immunity are based on IFN-γ release and thus, if they are used, should be performed before beginning anti-TNF. Interestingly, Ifx and Eta treatments decreased to the same extent the number of IFN-γ-releasing cells in response to mycobacterial antigens. This possibly contributes to the increased risk of TB infection in patients treated with either Ifx or Eta.

Ex vivo studies of anti-mycobacterial immune responses in patients treated with TNF antagonists are useful for evaluating the effect of these agents on circulating resting memory T lymphocytes, but they provide no clue for understanding their effect on lymphocytes recently activated by Mtb antigens. To address this issue, we determined whether the in vitro addition of TNF antagonists influenced anti-mycobacterial immune response. The results markedly differed from those evaluating the in vivo effect of TNF antagonists. In vitro addition of TNF antagonists resulted in a decreased proliferation in response to mycobacterial antigens. It also prevented the upregulation of mTNF expression by CD4⁺ T lymphocytes, another marker of activation induced by mycobacterial antigens. This shows that TNF antagonists either prevent activation of anti-mycobacterial memory T lymphocytes or directly affect activated cells. An effect of TNF antagonists on antigen-presenting cells could account for the first hypothesis. Both Ifx and Eta may induce apoptosis of monocytes/macrophages [46], which are involved in antigen presentation to T lymphocytes and IL-12-mediated activation of these cells. Elimination of monocytes/macrophages may participate in the inhibition of anti-mycobacterial responses. Although we cannot rule out this hypothesis, two findings supported a direct action of TNF antagonists on activated anti-mycobacterial T lymphocytes. First, delaying the addition of the TNF antagonist up to 24 to 48 hours before the end of the culture was sufficient to decrease both proliferation and the fraction of mTNF-expressing cells. Second, Ifx induced a rapid (within 4 hours) and active disappearance of mTNF from the surface of CD4⁺ T lymphocytes, possibly involving internalization or shedding of mTNF. Regardless of the mechanism(s) involved, a decreased mTNF expression by activated anti-mycobacterial CD4⁺ T lymphocytes may significantly alter their helper function, considering the role of mTNF on anti-mycobacterial immune responses [19]. In cultures performed with Ifx, apoptosis of activated CD4⁺ T lymphocytes, expressing mTNF, could be an additional mechanism of inhibition of anti-mycobacterial immune response. This apoptosis could occur through a direct effect of Ifx on either mTNF-expressing cells [47,48] or antibody-dependent cell cytotoxicity. Complement-mediated killing of mTNF-expressing cells is excluded in our experiments, performed in complement-free conditions.