Mechanisms of Infliximab

The pro-inflammatory cytokine tumor necrosis factor-α (TNF-α) plays a central role in mucosal inflammation and is likely to be at the apex of the inflammatory cascade in Crohn's disease (CD). 1,–6 TNF-α is crucial for the recruitment 7,8 and activation of many different immune cells, such as neutrophils, monocytes, dendritic cells, and macrophages, as well as B and T lymphocytes. 9,–13 In addition, TNF-α augments interferon-γ (IFN-γ) production in mucosal T cells. 13 The combination of these two cytokines has been implicated in the apoptotic loss of gut epithelial cells and disruption of the epithelial barrier, 14,15 which may further exacerbate the disease. Given the importance of TNF-α in many inflammatory processes, the drug development strategy for antibodies or fusion proteins directed against TNF-α was simple and convincing. Elevated levels of TNF-α in blood, serum, or inflamed tissue have been found in most inflammatory diseases; therefore, the neutralization of soluble TNF-α by specific binding and subsequent removal of the resulting protein complexes should reduce inflammation. 11 One of the first and most prominent anti-TNF-α drugs is the chimeric monoclonal anti-TNF-α antibody infliximab (Remicade®, formerly cA2; Centocor, Malvern, PA), and its efficacy in rheumatoid arthritis (RA) 16 and CD 17,18 had originally been attributed to its blocking capacity for soluble TNF-α. Infliximab has a serum half-life of 9.5 days and is still detectable in serum 8 weeks after infusion. 19 However, some patients treated with infliximab showed clinical benefits for as long as 1 year after a single infusion, suggesting that the underlying immunologic patterns of CD may be altered beyond the immediate effect of TNF-α neutralization. This view is supported by the relationship between mucosal production of inflammatory signaling molecules in remission and clinical relapses. 3,20 Human TNF-α is a nonglycosylated polypeptide that exists as either a transmembrane or a soluble protein. 21,–23 TNF-α originates as a homotrimeric complex of membrane-anchored 26-kDa transmembrane TNF-α precursors (mTNF-α), which is cleaved to the secreted form (3×17 kDa) by TNF-α-converting enzyme (TACE), a member of the metalloproteinase-disintegrin family of membrane-anchored glycoproteins 24,25 (Figure 1). Secreted TNF-α can bind to two cell surface TNF receptors, the 55 kDa TNF receptor-1 (TNF-R1), and the 75 kDa TNF-receptor-2 (TNF-R2). TNF-R1 is expressed by a wide range of cells, while TNF-R2 is more or less restricted to cells of hematopoietic origin. However, membrane associated TNF (mTNF-α) on the surface of activated immune cells can also induce responses in target cells via cell–cell contact, probably mainly via TNF-R2. 26 TNF-R2 has previously been considered mainly a ligand-passing receptor, which binds TNF-α only transiently and quickly passes it over to TNF-R1. 27 Several recent studies have indicated that the mTNF-α/TNF-R system can pass information not only one way (from the ligand to the receptor), but in both directions; a process imagined like a “molecular handshake,” which is felt by both partners. 28,–30 The transfer of information from the receptor-bearing cell to the ligand (mTNF-α)-bearing cell is termed “reverse signaling,” and this phenomenon has been demonstrated for many members of the molecular superfamilies of TNF-α and its receptors. 31,–35 Reverse signaling, and not the neutralization of soluble TNF-α, is probably also the key to understanding the efficacy or therapeutic failure of infliximab and other TNF-α binding agents in CD. Schematic representation of inflimab-induced signal transduction. Transmembrane tumor necrosis factor-α (mTNF-α) is cleaved to the soluble form (sTNF-α) by the transmembrane metalloproteinase TNF-α-converting enzyme (TACE). Infliximab neutralizes sTNF-α, but its efficacy is probably based on its binding to mTNF-α and the reverse signaling triggered by this ligation. Via unknown upstream mediators, infliximab induces apoptosis involving caspase activation and mitochondrial release triggered by bak and bax, which are up-regulated by infliximab. Reverse signaling also activates p38α, leading to transcriptional induction of the TNF-α gene. p38α was activated in both responders and nonresponders to infliximab treatment, but the downstream effectors activating transcription factor-2 (ATF-2) and heat shock protein-27 (Hsp27) were activated only in the responders. In 1995, Scallon et al 36 showed that infliximab can bind mTNF-α and induce lysis of mTNF-α-expressing cells via complement activation or antibody-dependent cellular cytotoxicity in vitro. When Baert et al 37 performed a histologic study in CD patients before and 4 weeks after infliximab treatment, they noted the virtual disappearance of neutrophils in the intestinal mucosa, a global reduction of CD4+ and CD8+ T lymphocytes as well as CD68+ monocytes, and a strong decrease in TNF-α+ lamina propria mononuclear cells. In the light of the findings of Scallon and colleagues, 36 these alterations in the mucosal cell populations were likely to be not only due to a general reduction of inflammation and thus the number of inflammatory cells in the mucosa, but also to a specific infliximab-induced death of immune cells. In 2001 and 2002, two studies showed that infliximab can indeed induce apoptosis in monocytes from patients with CD and from healthy individuals as well as in lamina propria T lymphocytes of CD patients by specifically binding to mTNF-α. 38,39 Infliximab-induced monocyte apoptosis was found to be mediated via caspase-8, -9 and -3, but was independent of CD95/Fas signaling. 38 Lügering et al 38 further investigated transcript levels of the bcl-2 family of pro- and anti-apoptotic proteins, which are key effectors in the mitochondrial pathway of apoptotis. Transcripts for the pro-apoptotic proteins bak and bax were induced in response to infliximab, and cytochrome c was released from mitochondria into the cytoplasm 38 (Figure 1). Ten Hove et al 39 chose a different cell system, namely lamina propria T cells and the Jurkat T cell line, but came to similar conclusions: CD3/CD28-stimulated Jurkat T cells showed an increase of the bax/bcl-2 ratio and subsequent apoptosis. According to the fact that most immune cells and cell lines (but not all, see below) express mTNF-α only upon activation, unstimulated Jurkat T cells showed no apoptotic response to infliximab. 39 Moreover, ten Hove et al presented the first data showing in situ apoptosis in patients with CD after infliximab treatment; a strong increase in apoptotic CD3+ lamina propria T lymphocytes was observed 24 hours after a single infusion of infliximab (5 mg/kg). However, infliximab treatment did not alter the expression of activation markers or intracellular bax/bcl-2 levels in peripheral T lymphocytes. 39 We have shown recently that infliximab treatment also induces apoptosis in some lamina propria macrophages and in most neutrophils, 40 accounting for the differential reduction of these cell populations observed by Baert et al 37 (see above). Taken together, these data suggest that infliximab selectively kills activated immune cells at sites of inflammation, but not resting or circulating cells. It is important to note that neither all cell populations nor all cells of one population are affected in the same degree, suggesting that mTNF-α-mediated apoptosis occurs only in certain cell subsets, which may be determined by the activation status but also by other determinants. Importantly, the complement-dependent lysis described by Scallon et al 36 appears to contribute to the cell-death induced by infliximab only on a small scale; peripheral blood mononuclear cell levels do not change significantly after infliximab treatment, 39 and infliximab is mostly well tolerated with no indications of a cell lysis syndrome. Therefore, it seems very plausible that the pro-apoptotic effects of infliximab are restricted to subpopulations of inflammatory cells that are crucial for the perpetuation of inflammation. 37,40 It is tempting to speculate that infliximab-induced reverse signaling via mTNF-α may break the barrier of apoptosis resistance in lamina propria T lymphocytes in CD. 39,41,42 Interestingly, the compound etanercept, a human soluble TNF-R2:Fc fusion protein that is successfully used in the treatment of RA, 43 is not effective in CD. 44 This surprising difference between the two diseases, which can both be treated succesfully with infliximab, underlines that it is not the neutralization of soluble TNF-α that accounts for the efficacy of infliximab in CD, since anti-TNF-α antibodies, such as infliximab or CDP571, 18,45 and etanercept have similar affinities to soluble TNF-α. However, consistent with the role of TNF-R2 in ligand-passing, 27 etanercept has a high “off rate,” which means that a considerable fraction of the bound TNF-α is quickly released again. Moreover, the binding rate of etanercept to mTNF-α on target cell surfaces does not exceed the affinity of a nonspecific control antibody. 46,47 Accordingly, etanercept failed to induce apoptosis in peripheral blood lymphocytes stimulated by mixed-lymphocyte reaction or in lamina propria T cells from patients with CD. 48 These findings allow us to tentatively draw two interesting conclusions: (1) the mechanism of anti-TNF-α treatment is probably different in RA and CD, since neutralization of soluble TNF-α is sufficient for a satisfying drug effect in RA but not in CD, and (2) apoptosis of active immune cells in the lamina propria, which is triggered by antibody-induced reverse signaling via mTNF-α, may be decisive for a therapeutic benefit in CD. 49 However, binding of infliximab to mTNF-α cannot only result in apoptosis of target cells, but also in down-regulation of IFN-γ production in colonic T cell cultures 50 and in stimulated Jurkat T cells, 39 suggesting a complex signaling pattern. The local anti-inflammatory and immunomodulatory effects of infliximab are not accompanied by an equally strong suppression of systemic immune functions, 19 even though a reduction of the pro-inflammatory cytokines interleukin-1β and interleukin-6 in serum of infliximab-treated patients with RA has been demonstrated within 24 hours after infusion. 51,52 It has been shown recently that the mitogen-activated protein kinases (MAPKs) p38α, c-Jun N-terminal kinase-1 and -2 (JNK1/2), and extracellular signal-activated kinase-1 and -2 (ERK1/2) are activated in both CD and ulcerative colitis, 53 that inhibition of p38α by the compound SB203580 in biopsy cultures from CD patients significantly reduces TNF-α production in the inflamed tissue, 54 and that inhibition of the stress-activated MAPKs p38α and JNK1/2 by the guanylhydrazone compound CNI-1493 is beneficial in the treatment of CD. 55 Since TNF-α is both an agonist and a target of the p38 and JNK pathways, 56,–59 and since TNF-α is probably the crucial pro-inflammatory mediator in CD (see above), we investigated the reaction of the MAPKs p38α and JNK1/2 to the systemic inhibition of TNF-α by infliximab. Colonic biopsies were taken from a pilot cohort of 5 patients with CD that were treated for the first time with a single infusion of infliximab (5 mg/kg body weight): 2 responders to infliximab treatment, 2 non-responders and 1 short-responder were chosen from a well-characterized study population. 60 In all of these patients, infliximab treatment induced a highly significant, transient activation of p38α, but not of JNK1/2, during the first 48 hours after infusion. 54 In monocytes from healthy volunteers, infliximab did not only induce apoptosis, but also activated p38α in a time- and dose-dependent manner. 54 Since most immune cells and cell lines express mTNF-α only after activation, which also triggers the MAPK cascades, we used the myelomonocytic cell line THP-1, which constitutively expresses mTNF-α, 61,62 to determine whether the activation of p38α was dependent on the presence of mTNF-α on the cell membrane. In agreement with the findings by ten Hove et al, 39 infliximab induced apoptosis and activated p38α and its target activating transcription factor-2 (ATF-2) in resting THP-1 monocytes, but not in resting Jurkat T cells. 54 Interestingly, the infliximab-triggered p38α activation specifically increased TNF-α mRNA expression up to 6-fold, suggesting a steady-state feedback loop within this novel pathway. A strong increase in circulating TNF-α – most likely bound to infliximab – during the first days after treatment with a single infusion of infliximab (5 mg/kg) has been reported in patients with RA, 52 which could be due to the infliximab-induced TNF-α induction. Regarding a possible connection between p38α signaling and the apoptosis induced by infliximab, the differences in apoptosis rates in THP-1 cells treated with the p38α inhibitor SB203580 and stimulated with infliximab at a concentration of 5 μg/ml, which roughly corresponds to the therapeutic dose of 5 mg/kg, failed to reach statistical significance. 54 However, no study has yet investigated in detail the interconnections of the MAPK and apoptosis pathways and other putative pathways triggered by mTNF-α reverse signaling. It appears however clear that inhibition of MAPK leads to a modification of the cytokine profile that is dependent on the T cell subpopulation studied. In TH1 cells inhibition of p38 MAP kinase has almost no effect on the production of the pro-inflammatory TNF, but severely impairs production of the contra-inflammatory IL-10. 65 In summary, the data available so far at least allow to conclude that concomitant use of p38α inhibitors and infliximab in CD may result in undesired interference of the drug effects. For yet unknown genetic and physiologic reasons, the efficacy of infliximab differs between patients. 18,60 However, some recent results have provided initial evidence for molecular differences between responders and nonresponders. In the small pilot cohort mentioned above, we investigated phosphorylation levels of ATF-2, a direct effector target of p38α, and the secondary target heat shock protein-27 (Hsp27), which is activated by p38α via one intermediate kinase. Twenty-four hours after infliximab treatment, ATF-2 and Hsp27 were only phosphorylated in the responders, but not in the nonresponders or in the short-responder. In contrast, the patient groups displayed no clear-cut differences in the protein expression of MAPKs and p38α targets. Despite the diversity of the patients, the inter-individual variability between the two responders and the two nonresponders, respectively, was very low with regard to the presence or absence of ATF-2 and Hsp27 phosphorylation after infliximab treatment. Even with the low patient numbers available, it can be concluded that CD patients can differ considerably in their reaction to infliximab on the primary and secondary target level of p38α while showing identical p38α activity. Since phosphorylation of ATF-2 by p38α after infliximab treatment is the “normal” cellular response, 54 we hypothesized that the functional p38α /ATF-2 axis in responders could enhance mTNF-α production and mTNF-α-mediated apoptosis of immune cells, 38,39 while non-responders could be defective in this regard. However, immunofluorescence stains for active caspase-3, a key marker for apoptosis, in biopsy sections showed similar apoptosis of lamina propria immune cells – mainly macrophages and neutrophils–24 hours after infliximab treatment in both responders and non-responders. Nevertheless, it must also be kept in mind that the pilot cohort consisted of only 5 patients, and that a systematic study would be warranted in the near future. Treatment of chronic active Crohn's disease by a single dose of a humanized monoclonal antibody directed against TNF-a (infliximab) induces complete clinical remission in 33% to 50% of patients. 18,60,66 Microsatellites at the TNF-a locus have been suggested as a tool to predict responders to anti-TNF treatment. 67 This finding could not be confirmed in another large study in 2 cohorts recruited from clinical trials with infliximab. 68 Recently, genetic variations in the NOD2 gene have been shown to be causatively involved in the etiology of a large fraction of patients with Crohn's disease. 69,–71 As these polymorphisms determine the subphenotype 72 and also alter NOD2 induced activation of NFκB, 70 that regulates expression of the TNF-gene, they were examined as pharmacogenetic determinants for the efficacy of an anti-TNF therapy with infliximab. However, polymorphisms in the NOD2 gene (SNP's 8,12 and 13) are also of no influence to response to anti-TNF therapy with infliximab in Crohn's disease. 73,74 The statistical construction of the testing process, which involved the sequential analysis of two large cohorts from GCP-level clinical trails, allows the exclusion of the tested polymorphisms in the TNF, TNF-R and NOD2 genes as pharmacogenetic determinants for the clinical response to infliximab. The negative finding also strongly supports the hypothesis that the mechanism of action of anti-TNF-a may involve additional effects to the neutralization of the cytokine. Therefore, response to anti-TNF treatment may be determined by variations in other genes in or outside the TNF signaling cascade. In view of the pivotal role of TNF-α in many defense responses, it is astounding how well systemic TNF-α inhibition is tolerated. One possibility is that in most cases, the anti-TNF-α agents merely reduce TNF-α activity to normal levels, thus leaving important protective functions of TNF-α intact. 63,64 The inhibition of signaling pathways which are (over) activated in a pathophysiologic context will rarely block these pathways completely, but rather restore a more physiologic level of activation. A lesson to be learned from our preliminary understanding of the mechanisms of infliximab action is that if a simple drug design has therapeutic efficacy, this does not necessarily mean it works in the way we devised it. Supported by public grants from the Competence Network Inflammatory Bowel Disease and the National Genome Research Network funded by the German Bundesministerium für Bildung und Forschung (BMBF) and the Deutsche Forschungsgemeinschaft (DFG).