|Home | About | Journals | Submit | Contact Us | Français|
TNF is a pleiotropic cytokine with important functions in homeostasis and disease pathogenesis. Recent discoveries have provided insights into TNF biology that introduce new concepts for the development of therapeutics for TNF-mediated diseases. The model of TNF receptor signalling has been extended to include linear ubiquitination and the formation of distinct signalling complexes that are linked with different functional outcomes, such as inflammation, apoptosis and necroptosis. Our understanding of TNF-induced gene expression has been enriched by the discovery of epigenetic mechanisms and concepts related to cellular priming, tolerization and induction of ‘short-term transcriptional memory’. Identification of distinct homeostatic or pathogenic TNF-induced signalling pathways has introduced the concept of selectively inhibiting the deleterious effects of TNF while preserving its homeostatic bioactivities for therapeutic purposes. In this Review, we present molecular mechanisms underlying the roles of TNF in homeostasis and inflammatory disease pathogenesis, and discuss novel strategies to advance therapeutic paradigms for the treatment of TNF-mediated diseases.
Forty years have passed since the description of a serum factor inducing tumour necrosis, 30 years since the cloning and purification of TNF, and almost 20 years since the approval of the first drug that targets TNF1. The initial concept of TNF as a potential drug for the treatment of cancer was followed by the opposite concept of TNF as a drug-target for inflammatory diseases2,3. Currently, five biologic agents targeting TNF are approved for the treatment of rheumatoid arthritis (RA), inflammatory bowel disease (IBD; for example, Crohn disease and ulcerative colitis), psoriasis, psoriatic arthritis, ankylosing spondylitis, juvenile idiopathic arthritis (JIA) and, most recently, hidradenitis suppurativa4,5 (TABLE 1). Notably, lower-cost biosimilar TNF-inhibitors have already been developed and introduced in the clinic6. In addition to the approved indications, TNF-blockade is also used, off-label, in Behçet disease, non-infectious ocular inflammation, and pyoderma gangrenosum, as well as in patients with TNF-receptor associated periodic fever syndrome (TRAPS), adult-onset Still disease and systemic-onset JIA7. In this Review, we focus on the latest discoveries about the biology of TNF, and outline new concepts that have been introduced in therapeutics for TNF-mediated diseases.
Newly synthesized TNF is expressed initially as a transmembrane protein, which requires proteolytic cleavage by TNFα-converting enzyme (TACE, also named ADAM17) to release soluble TNF8. TACE-dependent release of soluble TNF has been implicated in TNF-mediated inflammatory pathology in disease models9. TNF exerts versatile bioactivities via binding to, and activation of, two distinct receptors: TNF receptor 1 (TNFR1) and TNFR2 (REF. 10). TNFR1 is expressed ubiquitously, bears conserved death-domain motifs, and is activated by both soluble and transmembrane TNF. Expression of TNFR2 is restricted to specific cell types, such as neurons, immune cells and endothelial cells. TNFR2 lacks a death domain and thus is unable to induce programmed cell death directly; this receptor has been proposed to be activated primarily by transmembrane TNF11. One model of TNFR signalling proposes that TNFR1 primarily promotes inflammation and tissue degeneration, whereas TNFR2 mediates local homeostatic effects, such as cell survival and tissue regeneration12 (FIG. 1). This model suggests that selective therapeutic blockade of TNFR1 would keep homeostatic TNFR2 signalling intact; indeed, such a strategy is under development for the treatment of TNF-mediated diseases12.
Homotrimers of TNF bind to homotrimeric TNFRs to induce signalling10. Ligand-binding to TNFR1 results in the recruitment of the adaptor molecule TNFR1-associated death domain protein (TRADD) and the assembly of distinct signalling complexes, termed complexes I, IIa, IIb and IIc, which lead to distinct functional outcomes13,14 (FIG. 1a).
Upon TNF binding to TNFR, complex I is assembled at TNFR cyto plasmic domains at the plasma membrane and comprises TRADD, receptor-interacting serine/threonine-protein kinase 1 (RIPK1), TNFR-associated factor 2 (TRAF2), cellular inhibitor of apoptosis protein 1 (cIAP1) or cIAP2, and linear ubiquitin chain assembly complex (LUBAC)14 (FIG. 2). LUBAC consists of haem-oxidized IRP2 ubiquitin ligase-1 (HOIL-1), HOIP (HOIL-1 interacting protein, also known as E3 ubiquitin-protein ligase RNF31) and Sharpin. The current TNFR signalling model postulates the building of a scaffolding ubiquitin network in a step-wise process15. Initially, TRAF2 and cIAP1 or cIAP2 mediate Lys63-linked ubiquitination of complex I components. Subsequently, the LUBAC complex adds Met1-linked linear ubiquitin chains that strengthen the ubiquitin network, resulting in the stabilization of complex I and amplification of signalling16,17. Additional ‘atypical’ ubiquitin chains, such as Lys11-linked chains, can also facilitate signal transduction through TNFRs18.
The scaffolding ubiquitin network enables the recruitment and activation of two signalling complexes: the transforming growth factor (TGF)-β-activated kinase 1 (TAK1) complex, comprising TAK1, TAK1-binding protein 2 (TAB2) and TAB3, and the inhibitor of κB (IκB) kinase (IKK) complex, comprising NFκB essential modulator (NEMO, also known as IKKγ), IKK subunit-α (IKKα) and IKKβ14 (FIG. 2). TAK1 triggers mitogen-activated kinase (MAPK) signalling cascades that lead to activation of downstream JUN N-terminal kinase (JNK) and p38, as well as AP1 transcription factors, whereas IKKβ activates the canonical nuclear factor κB (NFκB) pathway (FIG. 2). Thus, induction of signalling complex I leads to expression of NFκB and AP1 target genes that are important in inflammation, host defence, and cell proliferation and survival1,14.
In contrast to complex I, complexes IIa, IIb and IIc are assembled in the cytoplasm and have distinct signalling and functional outcomes13,14,19–21 (FIG. 1a). Complexes IIa and IIb lead to activation of a caspase cascade that results in TNF-induced cell death via apoptosis21,22. A highly controlled process, apoptosis is implicated in the turnover of cells during development and organogenesis, epi thelial homeostasis, inflammation, immunity, and disease pathogenesis. Apoptotic cells remain intact and are rapidly phagocytized by macrophages, a process which has suppressive effects and diminishes inflammatory cytokine production14.
Complex IIc activates the necroptosis effector mixed lineage kinase domain-like protein (MLKL) by a RIPK3-dependent mechanism13,19,23. In contrast to apoptosis, necroptosis results in plasma-membrane rupture, which releases intracellular contents and triggers local inflammation13. TNF-induced necroptosis of cells at barrier surfaces such as the skin and intestinal mucosa can compromise barrier function and thereby contribute to inflammation24–26. The implications of TNF-induced necroptosis in the pathogenesis of inflammatory diseases are the focus of active investigations19, and the RIPK1 and RIPK3 kinases that control these processes are considered potential therapeutic targets25–27.
TNF-induced complex I signalling requires precise control to prevent the development of deleterious excessive or chronic inflammation. Several studies support the concept that negative regulators are used to fine-tune the magnitude, amplitude and kinetics of TNFR signalling. Additionally, it has been suggested that specific negative modulators might set cell-type-specific thresholds for TNFR activation. These ‘signalling brakes’ operate by targeting receptor-proximal signalling events, by inhibiting IKK-complex activity, or by regulating downstream checkpoints. The mech anisms utilized by these negative regulators include cleavage or internalization of TNFRs10, destabilization of signalling complexes28, displacement of signalling molecules from complexes by antagonists29–34, as well as degradation (proteasomal or lysosomal)31,35–37, inactivation (for example, dephosphory lation)38–41, or seques tration of signalling factors42–44. Although the expression and activity of negative regulators can be constitutive, induction by TNF is consistent with a model of ‘applying the brakes’ as part of inducible inhibitory-feedback loops42. The three major classes of negative regulators of TNF-induced complex I signalling include regulators of ubiquitination, phosphatases, and IκBs (summarized in TABLE 2).
Reports from several studies of complex I signalling kinetics have shown that TNF induces oscillations of NFκB activity that dampen over time42,45–47. This observation suggests that the coordinated function of signalling brakes does not simply terminate TNF-induced signalling, but progressively diminishes the amplitude of waves of NFκB dimers that enter and exit the cell nucleus. The fraction of responding cells and the number of waves are related to the concentration of TNF47. However, the oscillatory behaviour of TNFR signalling can depend on cell type and context. For example, a single pulse with a saturating dose of TNF induces sustained activation of NFκB in fibroblast-like synoviocytes (FLS) from patients with RA48. Although the concentration of TNF gradually diminished over time in culture, the activity of IKK complexes, as well as the nuclear localization and binding of target genes by NFκB subunit p65 (encoded by RELA), were maintained for several days after the single TNF-pulse48. Further studies are necessary to investigate whether the sustained NFκB activity in RA FLS is attributable to functional incompetence of signalling brakes.
TNF and TNFR1 are at the crossroads of inflammation, survival, apoptosis and necroptosis13,14. The factors that determine whether TNFR1 engagement induces complex I, leading to inflammatory responses, or complexes IIa, IIb and IIc, leading to cell death, are not well understood. The current model of this ‘live or die’ decision-making suggests that the expression levels of antiapoptotic molecules such as the long form of cellular FLICE-inhibitory protein (c-FLIPL), and the degree of RIPK1 ubiquitination are critical elements in determining which pathway is activated. c-FLIPL binds to caspase-8 and blocks specifically its proapoptotic effector functions14. Regarding the ubiquitination status of RIPK1, the model proposes that ubiquitinated RIPK1 is tethered to complex I, whereas stripping of RIPK1 from the ubiquitin cloak results in its cytoplasmic trafficking and interaction with programmed-cell-death-inducing machinery49. Studies published in the past 5 years have revealed the roles of Sharpin50–54, A20 (REF. 55) and ubiquitin carboxyl-terminal hydrolase CYLD49 in the decision to undergo cell death. Other factors related to cell differentiation and activation state, as well as additional environmental signals, are likely to affect this decision; thus, the outcome of TNFR1 signalling is context-dependent. Similarly, regulation of the balance of complex IIa/IIb versus complex IIc formation is incompletely understood, but RIPK1 and RIPK3 seem to have an important role in this process27,56–58.
At the cellular level, several hundred genes are regulated (induced or suppressed) by TNF in a cell-type-specific manner. For instance, in FLS, TNF induces copious amounts of IL-6, with rapid and sustained production48. In macrophages, lower levels of IL-6 are induced more transiently48, whereas neutrophils have limited capacity to produce IL-6 in response to TNF59. Cell-type specificity of gene expression results, at least in part, from the effects of cell differentiation and cell-state transitions on the chromatin landscape60.
The chromatin landscape refers to the genome-wide pattern of open chromatin sites that demarcate regulatory elements (promoters and enhancers) where transcription factors can readily bind to regulate gene expression60. During cell differentiation, lineage-specific transcription factors, termed ‘pioneer factors’ or ‘master transcription factors’, penetrate the barrier of histone-containing nucleosomes that maintain a ‘closed’ or inaccessible chromatin conformation to create lineage-specific topologies of open chromatin. According to this model, cellular patterns of gene expression in response to stimuli such as TNF are predetermined chiefly by the chromatin landscape. ‘Signalling transcription factors’, such as NFκB and AP1, which have limited ability to remodel chromatin, will predominantly bind to and activate genes that exhibit open chromatin at their promoters and enhancers60.
Studies in the past 3 years have modified this concept by showing that lineage-determining transcription factors can cooperate with various environmentally activated transcription factors to remodel the enhancer repertoire of a cell, and thus alter patterns of gene expression. For example, stimulation of macrophages with diverse agonists (including Toll-like receptor (TLR) ligands or cytokines) induced cooperative recruitment of master transcription factors and stimulus-specific transcription factors in new genomic topologies, generating novel enhancers termed ‘latent enhancers’ (REF. 61). In the same study, acquisition of latent enhancers was shown to alter responses to subsequent stimulation. Along these lines, another study showed that TLR8 agonists remodel the chromatin at the IL6 locus in neutro phils and uncover a latent enhancer, converting IL6 in neutrophils into a TNF-inducible gene59.
Evidence suggests that chronic pathological states are associated with disease-specific stable changes in gene expression that are consistent with epigenetic mechanisms. For instance, FLS derived from patients with RA have a DNA methylome that is different from that of osteoarthritis (OA) FLS62, and FLS derived from patients with early and longstanding RA have distinct DNA methylomes63. Notably, TNF alters chromatin states and induces higher levels of inflammatory cytokines and chemokines in RA FLS compared with OA FLS48. We hypothesize that, in the context of RA, unremitting inflammation induces disease-associated chromatin modifications, potentially altering cellular gene-induction responses to TNF. Consistently, inflammatory cytokines have been shown to modulate DNA hydroxyl-methylation in FLS and chondrocytes64,65. Overall, accumulating evidence suggests that epi genetic mechanisms, such as chromatin modifications or DNA methylation, established during differentiation or acquired in response to homeostatic or pathological environmental stimuli, contribute to the tissue-specific and disease-associated effects of TNF.
The microenvironment can also condition cellu lar responses to TNF independently of epigenetic mechanisms. Simultaneous engagement of TLRs and prostaglandin receptors cooperates with TNF to induce transcriptomes in monocytes, macrophages and dendritic cells that resemble chronic inflammatory states66. In addition, RA FLS modulate expression of approximately one-third of TNF-regulated genes in macrophages, including Myc-dependent, growth-factor-inducible and interferon (IFN)-inducible genes67. These findings suggest that signal integration and intercellular functional coupling will shape responses to TNF in complex inflammatory environments such as RA synovitis.
Three distinct patterns of induction kinetics for TNF-inducible genes have been identified: early, inter mediate and late68–70. The accessibility of chromatin is a critical factor in determining the expression kinetics of TNF-inducible genes. Genes with accessible chromatin are induced more rapidly by TNF (FIG. 3a), compared with genes that require chromatin remodelling (FIG. 3b). Genes requiring de novo protein synthesis for their induction are also usually expressed in a delayed manner. For instance, cell-type-specific clusters of genes exist that require the prior synthesis of IFN-β or the transcription factor IκBζ for optimal induction by TNF59,68 (FIG. 3c,3d). Regarding the IFN-β-dependent genes, TNF first induces NFκB activation and IFN-β synthesis68,71 (FIG. 3c). Subsequently, newly synthesized IFN-β operates in an autocrine manner, activating the Janus kinase (JAK) and signal transducer and activator of transcription (STAT) signalling pathway, which cooperates with NFκB and activates IFN-stimulated genes (ISGs) such as CXCL9 (encoding MIG) and CXCL10 (encoding IP-10)68. Regarding the IκBζ-dependent genes, TNF synergizes with TLR8 agonists or cytokines to induce IκBζ59, which binds to and promotes activation of a distinct subset of gene promoters by inducing an activating histone mark, histone H3 trimethylation at lysine 4 (H3K4me3). The need for prior histone modification results in delayed transcription of NFκB-dependent genes in a cell-type-specific manner59. These results suggest a model whereby TNF induces a cascade of transcription factors that sustains late-phase gene expression. This model is supported by evidence that TNF induces a cascade mediated by initial induction of AP1 that eventually leads to expression of NFATc1, which promotes osteoclastogenesis that can contribute to the bone-resorptive properties of TNF70. Discovery of additional transcriptional modules that mediate novel aspects of late-phase TNF-induced gene expression and function represents an important area for future research.
A currently accepted model suggests that chromatin barriers operate as rheostats to regulate the amplitude and timing of expression of TNF-inducible genes60. Genes induced with a delay before peak expression are typically associated with a nonpermissive chromatin environment, such as nucleosomes that occlude regulatory elements, repressive histone marks (for example, H3K27me3 and H3K9me3) and co-repressor complexes60 (FIG. 3). The nuclear receptor co-repressor 1 (NCoR) complex represses a cluster of TNF-inducible genes by recruiting the enzyme histone deacetylase 3 (HDAC3)72,73. Methyl-CpG-binding protein 2 (MeCP2), which binds to methylated CpGs, recruits the NCoR complex to gene promoters, suppressing their transcription74 (FIG. 3e). In macrophages, lack of MeCP2 resulted in increased expression of a cluster of inflammatory genes upon TNF-stimulation74. Overall, the need to overcome nonpermissive chromatin states before genes can be transcribed contributes to the delayed gene-induction observed after TNF stimulation.
TNF-inducible genes can be further subclassified as transient or sustained, on the basis of the duration of their expression48,69. TNF induces sustained expression of inflammatory mediators in RA FLS, potentially due to prolonged NFκB signalling and sustained chromatin remodelling48. For a subset of TNF-inducible genes, mRNA stability is an additional determinant of the duration of their expression69. The mechanisms that fine-tune the expression kinetics of TNF-inducible genes, including signalling brakes and rheostat control of chromatin, help to ensure a stepwise and sequential response to pathogens and inflammatory stimuli — optimal host defence relies on a rapid and transient inflammatory response, followed by a resolution phase to attain wound healing and tissue repair.
One important function of cytokines is to change the physiological state of cells and the way they respond to environmental stimuli. This feature has been referred to as ‘memory’ or ‘training’, and includes enhanced responses to subsequent challenges (priming) or refractoriness to stimulation (desensitization or tolerance). Accumulating evidence suggests that TNF modifies cellular responses to other stimuli. For instance, TNF displays a priming effect on RA FLS, enhancing their subsequent inflammatory response to IFNs75. Notably, this priming was gene-specific and resulted in sensitization of RA FLS even to suboptimal concentrations of IFNs. In macrophages, by contrast, TNF desensitized the cells to the effects of lipopolysaccharide (LPS) and protected mice from LPS-induced lethality76.
Molecular mechanisms underlying TNF-induced priming and tolerization have been revealed (summarized in FIG. 4). A common theme for both phenomena is the induction of gene-specific chromatin modifications by TNF, which alter cell responses to subsequent stimulations. In RA FLS, TNF induces an accessible chromatin state in the promoter region of CXCL10, enabling the immediate and robust induction of IP-10 mRNA upon re-stimulation with IFNs75 (FIG. 4a). A genome-wide study revealed a similar capacity of TNF and other cytokines to induce faster or augmented responses to subsequent challenges, owing to a gene-specific induction of latent enhancers61, although TNF can also suppress remodelling of chromatin at genes such as IL6 (REF. 76).
An additional mechanism that potentially contributes to TNF-induced cell priming or tolerization is the effect of TNF on signalling events upstream of chromatin. For instance, in RA FLS, prolonged exposure to TNF increases the intracellular reservoir of STAT1, amplifying the activation of STAT1 upon secondary challenge with interferons75 (FIG. 4a). Moreover, in macro phages, exposure to TNF induces the expression of signalling brakes, such as A20 and IκBα, which potentially restrict upstream signalling upon secondary stimulation with LPS76 (FIG. 4b).
An intriguing finding of these studies is the sustained effect of TNF on target cells, suggesting the establishment of ‘short-term transcriptional memory’. An attractive molecular explanation could be the induction by TNF of signalling molecules and chromatin modifications, which are not rapidly reversed77. Along these lines, TNF induces sustained expression of STAT1 (REF. 75), NFATc1 (REF. 70) and IκBδ43. Similarly, induction by TNF of the positive histone mark H3K4me1, which is associated with functional enhancers, is maintained for days in a large fraction of latent enhancers despite the absence of continuous stimulation61. Thus, short-term transcriptional memory is maintained by sequence-specific transcription factors and epigenetic mechanisms.
Ablation of the Tnf gene in mice revealed that TNF has homeostatic functions in addition to its immune and inflammatory roles78–80. TNF is required for optimal defence against pathogens, proper lymphoid-organ architecture and germinal-centre formation, development of granulomas, resolution of inflammation and induction of tissue repair. Homeostatic functions of TNF are also supported by studies in mice lacking TNFRs12 and by cell-based functional assays76,81,82 (summarized in BOX 1). Several reports have identified molecular mech anisms that could explain some aspects of these unexpected homeostatic roles of TNF. One study showed that TNF desensitizes macrophages to the deleterious effects of secondary inflammatory challenges (tolerization)76. Others have described homeostatic effects of TNF in tissue regeneration, such as neuronal remyelination83, cardiac remodelling84 and cartilage regeneration85. In addition, TNF has suppressive effects on adaptive immune processes in models of autoimmunity such as autoimmune encephalitis12. The homeostatic potential of TNF has been suggested to be mediated, at least in part, by TNFR2 (REFS 12,86).
TNF induces inflammation, activates vascular endothelium, orchestrates the tissue recruitment of immune cells and promotes tissue destruction2 (the pathogenic functions of TNF are summarized in BOX 1). Uncontrolled production or function of TNF has been linked to the development of inflammatory diseases such as RA, IBD, psoriasis, PsA, ankylosing spondylitis and specific types of JIA. In contrast to the well-defined role of TNF in these diseases, its role in the pathogenesis of multiple sclerosis is a conundrum12. Although initial studies suggested a pathogenic function for TNF in multiple sclerosis, clinical trials of a global inhibitor of TNF were discontinued owing to unexpected aggravation of the disease87. In addition, patients receiving TNF blockade for other diseases sporadically developed lesions with demyelination, such as optic neuritis88. These observations suggest that TNF can suppress autoimmune processes and/or exert essential homeostatic functions within the micro environment of the central nervous system (CNS). Several studies suggest that the homeostatic and pathogenic activities of TNF are mediated by distinct molecular and cellular pathways12. TNFR2, which is expressed on regulatory T (TREG) cells, oligodendro cytes and astrocytes, mediates immunoregulation, neuronal survival and re-meyelination; TNFR1, by contrast, induces CNS inflammation and neuronal demyelination12. Selective inhibitors of the neurotoxic TNFR1 pathway, which preserve the neuroprotective TNFR2 pathway, are being developed and could be a new avenue in therapeutics for patients with multiple sclerosis.
TNF has been implicated in the hyperalgesia observed in the context of inflammation (inflammatory pain) and neuronal damage (neuropathic pain). In humans with RA and in several animal models of inflammation, neutralization of TNF induced rapid reduction of nociceptive neuronal activity in the afferent neurons, the spinal cord and the brain89–94. Notably, TNF blockade reverses hypernociception long before its effect on inflammation becomes obvious94. This observation suggests that the antinociceptive effects of TNF-blockade are distinct from their anti-inflammatory effects. At the cellular level, TNF seems to trigger hyperalgesia by inducing peripheral and central sensitization to mechanical stimulation93. Peripheral sensitization results from the direct action of TNF on C and Aδ nociceptive neurons, which innervate the site of inflammation91. Central sensitization is more complex, and potentially involves several cell types such as spinal cord nociceptive neurons, microglial cells and astrocytes93. At the molecular level, both peripheral and central sensitization are dependent on TNFR1-mediated signals90,93.
TNF was discovered as a serum factor that induces necrosis of tumour cells, in part by acting on tumour vasculature to compromise blood supply95. The cytolytic potential of TNF has been explained by its capacity to induce programmed cell death. Subsequent studies revealed that TNF also displays cytostatic effects on specific tumour cell lines81. In a 2013 report, TNF was shown to cooperate with IFN-γ to induce tumour-cell senescence, a cell-state characterized by permanent growth arrest82. The widespread use of TNF as a drug for cancer has been prevented by the systemic toxicity of TNF96.
Accumulating evidence suggests that TNF also has the potential for opposite effects on cancer, either by inducing carcinogenesis or promoting the progression of established tumours97. TNF could be genotoxic by inducing de novo mutagenesis and suppressing the DNA-repair machinery. In the context of established tumours, TNF has the potential to promote the survival and proliferation of malignant cells, either directly by activating NFκB in tumour cells, or indirectly by inducing the production of tumour-promoting cytokines such as IL-6. In addition, TNF contributes to the escape of tumour cells from immunosurveillance by promoting the immunosuppressive capacity of myeloid-derived suppressor cells and TREG cells98,99. Finally, TNF facilitates tumour metastasis by promoting epithelial–mesenchymal transition and by inducing the synthesis of matrix metalloproteinases (MMPs).
Several studies suggest that TNF is cardiotoxic for the healthy myocardium, and potentially cardioprotective for the failing myocardium. Cardiotoxicity is primarily attributable to TNF-induced cardiomyocyte apoptosis100, whereas cardioprotection results from TNF-induced ectopic expression of keratin 8 and keratin 18 in cardiomyocytes84. In addition to its role in myocardial diseases, TNF is potentially implicated in the pathogenesis of atherosclerosis by affecting lipid metabolism, activating endothelial cells and inducing vascular inflammation101.
A study published in 2013 revealed a pathogenic role for TNF during the course of Dupuytren disease102. Dupuytren nodules are infiltrated by classically activated macrophages that secrete TNF. Fibroblasts and myofibroblasts from Dupuytren lesions had increased expression of both TNFR1 and TNFR2, suggesting that the disease micro environment sensitizes local stromal cells to the effects of TNF. Neutralization of TNF reduced the contractile activity of myofibroblasts, suggesting that the local administration of TNF blockers might be beneficial for patients with Dupuytren disease102.
The approved anti-TNF agents have been a commercial success and a scientific breakthrough, improving the quality of life of millions of patients with TNF-mediated diseases. Despite these successes, the current therapeutic paradigm of global TNF blockade has several limitations, the three most important being low rates of disease remission, the development of adverse effects and the generation of antibodies against biologic TNF inhibitors103. Adverse effects include common and opportunistic infections, reactivation of latent tuberculosis, ‘paradoxical’ induction of autoantibodies, lupus-like symptoms, demyelination, psoriasis, sarcoidosis, as well as a potentially increased risk for specific malignancies, such as lymphomas. Development of fatal disseminated histoplasmosis and aggressive hepato splenic T-cell lymphomas in some patients undergoing treatment with TNF inhibitors resulted in the issue of ‘black box’ warnings for these drugs104,105. To overcome these limitations and improve efficacy and safety, alternative strategies are in development.
An attractive strategy to improve the rate of response to TNF blockade is to combine anti-TNF agents with other medications. Along these lines, combined inhibition of TNF and IL-17A with a bi-specific construct proved superior to single-cytokine blockade in inhibiting the production of inflammatory mediators in in vitro studies and in an arthritis model106. Notably, however, previous experience with combinations of biologic agents has shown that excessive immunosuppression can lead to an alarming rise in risk of infections107,108.
Combining TNF blockade with drugs that target pathogenic pathways or cells not implicated in host defence, to improve response rates without compromising safety, is a therapeutic approach worth considering. Several attractive candidate combinations have been suggested: targeting of TNF and angiogenesis (for example, with an antibody against vascular endothelial growth factor); targeting TNF and tissue destruction (for example, with anti-MMP14 or anti-RANKL (receptor activator of NFκB ligand) which suppresses osteoclastogenesis); and targeting TNF and FLS (for example, by inhibiting cadherin 11, a cell-surface molecule that regulates FLS adhesion and inflammatory functions)103.
Another alternative could be to shift the current therapeutic paradigm of global TNF inhibition to selective targeting of the pathogenic bioactivities of TNF, keeping the homeostatic functions of this cytokine intact. The proposed functional dichotomy between the TNFR1 pathway (primarily pathogenic) and the TNFR2 pathway (primarily homeostatic) might provide an approach to selective inhibition12. Next-generation TNF-blockers that selectively block the TNFR1 pathway have been developed and are described in TABLE 3 (REFS 109–117). Notably, the development of selective TNFR2 agonists provides an option for the future: to combine selective blockade of the TNFR1 axis with selective boosting of TNFR2 signalling118.
In 2011, progranulin was found to operate as a natural antagonist of TNF by binding to both TNFR1 and TNFR2 (REF. 119). Administration of recombinant progranulin or of an engineered progranulin analogue (Atsttrin) was beneficial in several animal models of inflammatory arthritis119. Notably, progranulin has anabolic effects in cartilage via TNFR2-mediated activation of MAPK1 and MAPK3 (REF. 85). This observation suggests that progranulin might alleviate arthritis via dual effects on TNFR-mediated pathways, by blocking deleterious TNFR1 signalling and by engaging TNFR2 to trigger MAPK-mediated anabolic effects in chondrocytes85.
The development of antibodies against TNF-blocking biologic agents results in allergic reactions and in progressive loss of drug effectiveness. This issue of drug immunogenicity has been only partially addressed with the development of humanized or fully human biologic agents. An alternative approach is to trigger the production of natural neutralizing anti-TNF antibodies via active immunization. This method seemed feasible in a limited 12-month phase II study in patients with RA or IBD120, as active immunization with TNF-kinoid (composed of recombinant human TNF conjugated to the carrier protein KLH) resulted in the generation of neutralizing polyclonal antibodies against TNF. The clinical efficacy of this approach remains to be determined, and needs to be weighed against the potentially catastrophic consequences of life-long ablation of TNF.
Downregulation of the inappropriate transcription of TNF could be effective in ameliorating TNF-driven pathologies. IκBβ has been shown to function as a gene-selective co-activator, driving prolonged transcription of TNF121. Mice lacking IκBβ have reduced transcription of Tnf and are protected from LPS-induced lethality and collagen-induced arthritis. Gene-selectivity was explained by the specific interaction of the newly synthesized IκBβ with the p65–c-Rel NFκB dimer, which binds to the κB2 site in the promoter of Tnf121. Hence, blocking IκBβ might be a promising strategy to selectively inhibit the chronic phase of TNF production and treat chronic inflammatory diseases, such as RA.
As mentioned in this Review, precursor transmembrane TNF is cleaved by TACE, releasing soluble TNF. As TACE has important homeostatic functions by regulating EGFR-signalling in epithelial barriers, direct targeting of TACE raises safety concerns. However, inactive rhomboid protein 2 (iRhom2) has been identified as a myeloid-specific regulator of TACE that facilitates the shedding of membrane-bound TNF8. This observation suggests that therapeutic targeting of iRhom2 is a potentially attractive strategy to block the activity of TACE, specifically in macrophages, to prevent the release of pathogenic soluble TNF. This therapeutic strategy has the advantage of leaving intact the proposed homeostatic transmembrane TNF–TNFR2 pathway.
Small molecules targeting chromatin regulators have the capacity to block specific pathogenic pathways downstream of TNFRs. In a 2014 study, pharmacologic inhibition of bromodomain and extra-terminal domain (BET) proteins suppressed TNF-induced endothelial cell activation, attenuating leukocyte rolling, adhesion and transmigration122. Notably, at the transcriptional level, inflammatory genes associated with TNF-induced super-enhancers were preferentially downregulated122. This observation suggests that BET inhibitors selectively block branches of the TNF-induced transcriptional programs.
Kinase inhibitors have also been described to suppress intracellular responses to TNF. For instance, tofacitinib, a JAK inhibitor approved for use in the treatment of RA, suppressed the late phase of TNF-induced STAT activation and cytokine production in macrophages123.
Important advances in the understanding of TNF biology and pathogenic functions are summarized in BOX 2. These advances have important therapeutic implications as they identify therapeutic targets — such as LUBAC components, mediators of necroptosis, epigenetic regulators (for example, BET proteins124 and HDACs125) and novel transcription factors — whose inhibition might have more-selective effects on disease-specific pathogenic mechanisms,possibly achieving comparable efficacy without the toxic effects associated with global TNF blockade. Recognition of the homeostatic functions of TNF has led to new concepts for the treatment of rheumatic diseases, presenting the possibility of strengthening mechanisms of tolerization and tissue repair as a complement to suppression of inflammatory pathways.
Despite decades of research, however, many important challenges and gaps remain in our understanding of the basic science of TNF and in how to translate this knowledge into better treatments. Important directions for future basic research include a deeper understanding of the molecular underpinnings and functional consequences of the alternative outcomes of ‘life’ versus ‘death’ signalling, of the epigenetic regulation of TNF responses, of mechanisms that restrain TNF responses, and of homeostatic functions that could be manipulated to alter the pathogenic profile of TNF. An intriguing research direction related to epigenetic mechanisms is the effect of TNF on the expression of non-coding RNAs, including microRNAs126,127 and long non-coding RNAs128. Another future research priority is the harnessing of basic science and the power of new high-throughput sequencing technologies and genome-wide approaches to study patients before and after TNF blockade therapy to identify markers predictive of clinical response and mech anisms associated with resistance to TNF blockade129. The identification of such markers and mechanisms is a major unmet need that will enable replacement of the current trial-and-error therapeutic approach with a stratified and increasingly precise strategy, in which the decision to introduce TNF blockade will be based on validated biomarkers that predict response and not on empirical criteria130. Such translational studies also offer hope for the discovery of alternative pathogenic pathways that can be targeted in patients who are resistant to therapy with TNF inhibitors.
EMT, epithelial–mesenchymal transition; MDSC, myeloid-derived suppressor cell; MMP, matrix metalloproteinase; TREG cell, regulatory T cell.
LUBAC, linear ubiquitin chain assembly complex; MLKL, mixed lineage kinase domain-like protein; RIPK3, receptor-interacting serine/threonine-protein kinase 3; TNFR2, TNF receptor 2.
This work was supported by grants from the NIH (to L.B.I.) and the Feldstein Medical Foundation (to G.D.K.).
Both authors researched data for article, made substantial contributions to discussions of the content, wrote the article and reviewed and/or edited the manuscript before submission.
Competing interests statement
The authors declare no competing interests.