Death Ligands Beyond TNF in Inflammation and Autoimmunity

Unfortunately, anti-TNF therapy does not help all patients suffering from one of the aforementioned disorders. In about half the patients suffering from RA, long-lasting benefits can be achieved through the inhibition of TNF, whereas the other half of RA patients do not benefit from this treatment. The same is true for ~35% of patients with psoriasis and between 20% and 40% of patients suffering from the different forms of IBD (Cho and Feldman 2015; Lopetuso et al. 2017). TNF inhibition also provided no therapeutic benefit in patients with the autoimmune disorders multiple sclerosis (MS) and amyotropic lateral sclerosis (ALS), despite a clear up-regulation of the TNF system in these disorders (Lenercept Multiple Sclerosis Study Group 1999). One interpretation of these failures is that TNF is not the apical regulator of inflammation in nonresponding patients, and inflammation is instead instigated by different means.

This view was challenged by the realization that TNF can mediate inflammation not only by activating gene expression, but also by inducing aberrant cell death (Fig. 2). For example, mutation of the Sharpin gene in mice causes chronic proliferative dermatitis (cpdm) (Seymour et al. 2007). In cells of these mice, TNF-induced gene activation is attenuated, whereas the induction of cell death is increased (Gerlach et al. 2011). At the time, this result provided a conundrum: how did the mouse develop an inflammatory disease if proinflammatory gene expression in response to TNF was compromised? One could argue that an inflammatory mediator other than TNF drove inflammation. However, gene activation by other inflammatory mediators, including CD40L, was also compromised in Sharpin-deficient cells (Gerlach et al. 2011). Given that TNF-induced cell death was aberrantly increased, we reasoned that genetic ablation of Tnf could address two questions: (1) does aberrant cell death cause the inflammatory syndrome in cpdm mice, and (2) is TNF the intrinsic agent provocateur of this cell death? Indeed, genetic deletion of Tnf, even heterozygous deletion of Tnf, prevented both cell death in the skin and inflammatory disease in cpdm mice. Therefore, we concluded that TNF-induced cell death caused inflammation in these mice (Gerlach et al. 2011). This discovery was genetically confirmed when it was shown that codeletion of Casp8 and Mlkl, or Fadd and Ripk3, also prevented cell death in the skin and inflammatory disease in cpdm mice (Kumari et al. 2014; Rickard et al. 2014; Peltzer et al. 2018). Other genetic mouse models of inflammatory disease have also been shown to be driven by excessive cell death (Welz et al. 2011; Dannappel et al. 2014; Kumari et al. 2014; Vlantis et al. 2016). Importantly, several of these genetic alterations have also been found in humans, although these alterations appear to be rare. These individuals often suffer from a combination of autoinflammation and immunodeficiency (Fusco et al. 2008; Cuchet-Lourenco et al. 2018; Oda et al. 2019).

Open in a separate window

Figure 2.

Aberrant cell death in chronic inflammation. (A) Cell death is important for the repair and regeneration programs of different types of tissues, such as the epithelium of the skin and the intestine. Dying cells release factors that trigger the activation of an inflammatory program, whose ultimate purpose is to restore tissue integrity. (B) Deregulated cell death determines the persistence of tissue repair programs, which in turn leads to chronic inflammation, epithelial cell hyperproliferation, and, ultimately, to an autoinflammatory or autoimmune disorder. (DC) Dendritic cell.

A further advance in the understanding of the etiology of inflammatory diseases came when it was shown that pharmacologic or genetic inhibition of RIPK1 prevented dermatitis in cpdm mice (Berger et al. 2014). Thus, by blocking the kinase activity of RIPK1, TNF-induced cell death and skin inflammation was prevented. These findings raised the possibility that inhibition of RIPK1 might suffice to block aberrant TNF-induced cell death and consequent inflammation. If this were the case, then TNF blockade could potentially be replaced by inhibition of RIPK1. Additional preclinical studies showed that inhibition of RIPK1 could also ameliorate TNF-induced septic shock, anticollagen antibody-induced arthritis, and inflammation and tissue injury caused by A20 deficiency, but it provided no benefit in other models, such as chemically induced pancreatitis (Newton et al. 2016; Patel et al. 2020).

Given that several biotherapeutic inhibitors of TNF figure in the list of the world's top 10 best-selling drugs, it is not surprising that this concept garnered the attention of the biotech and pharmaceutical industries. Many companies are currently bringing RIPK1 kinase inhibitors into clinical testing. However, the first phase II clinical trials, which were performed in RA, psoriasis, and IBD patients, did not appear to reveal a benefit and the inhibitor has recently “moved back to research” (see gsk.com/media/5745/q3-2019-results-slides.pdf). Could it be that the notion of inflammatory disease being caused by aberrant TNF-induced RIPK1-kinase-dependent cell death is just too simple? Recent evidence suggests this might indeed be the case. For example, lethal dermatitis induced by keratinocyte-specific genetic deletion of Hoip or Hoil-1 (Hoip^E-KO or Hoil-1^E-KO) is considerably delayed by codeletion of Tnfr1, whereas inhibition of RIPK1 only delays lethal dermatitis by a few days (Taraborrelli et al. 2018). Intriguingly, however, when Hoip^E-KO;Tnfr1^−/− or Hoil-1^E-KO;Tnfr1^−/− mice are given a RIPK1 kinase inhibitor, lethal dermatitis is prevented (Taraborrelli et al. 2018). Thus, at least in this genetic model, TNFR1 ablation and RIPK1 kinase inhibition are synergistic at inhibiting cell death and consequent inflammation. Nevertheless, this discovery suggests that combining a TNF inhibitor and a RIPK1-specific inhibitor might benefit certain autoimmune or autoinflammatory disease patients. For patients with rare germline mutations impacting LUBAC components, this combination may represent an alternative treatment strategy if their current therapies do not work or if they develop resistance to them. An outstanding question, however, is what drives RIPK1-dependent cell death and disease when the TNF–TNFR1 system is absent.

The idea that TNF superfamily (TNFSF) cytokines beyond TNF itself could act individually or in a concerted manner to promote inflammation has been proposed, albeit with cell-death-independent mechanisms and disease etiologies in mind and, hence, without considering the coinhibition of other cell death–inducing ligands (Croft and Siegel 2017). At present, several TNFSF proteins are under evaluation in preclinical and clinical studies as potential targets in various rheumatic and chronic inflammatory diseases. Especially noteworthy in the context of this review are TRAIL and CD95L. The primary signaling output of these two TNFSF members is cell death, which is in contrast to that of TNF and TL1A (Fig. 1). Nevertheless, as described earlier, both the TRAIL–TRAIL-R1/R2 and CD95L–CD95 systems can induce gene activation (Hartwig et al. 2017; Henry and Martin 2017). The CD95L–CD95 system drives activation-induced T-cell death, which is instrumental in immune homeostasis (Alderson et al. 1995; Brunner et al. 1995; Dhein et al. 1995; Ju et al. 1995). Indeed, mutations in CD95 cause the accumulation of aberrant T cells, which ultimately results in autoimmunity (Watanabe-Fukunaga et al. 1992). The TRAIL–TRAIL-R system has mainly been characterized for its ability to selectively kill cancerous, but not essential normal cells. In addition to being expressed in cancer cells, TRAIL-R1 and TRAIL-R2 are also expressed on different subpopulations of T cells, whose apoptosis they can promote (Roberts et al. 2003; Zhang et al. 2003). Hence, the immunological “day job” of the CD95 and TRAIL systems appears to be the proper termination of an immune response and the prevention of autoimmunity. Indeed, the killing function of CD95 and TRAIL-R1/R2 in activated or autoreactive T cells has led to the idea that autoreactive T cells might be eliminated in the context of autoimmunity using CD95L, TRAIL, or other DR agonists. The severe hepatotoxicity of CD95 agonists (Ogasawara et al. 1993) excludes their use, but TRAIL and other agonists of the TRAIL DRs may yet be an option, although this awaits therapeutic validation in patients.

Given that caspase-8-dependent cell death causes lethal inflammation in Hoip^E-KO and Hoil-1^E-KO mice, as well as in Hoil-1^E-KO;Tnfr1^−/− and Hoip^-KO;Tnfr1^−/− mice, the role of other DRs in inflammation must be considered (Taraborrelli et al. 2018). Indeed, cells deficient for HOIL-1 or HOIP are more sensitive to killing by TNF, TRAIL, or CD95L. Although neither constitutive deletion of Trail-r nor keratinocyte-specific deletion of the DD of CD95 ameliorated dermatitis in Hoil-1^E-KO;Tnfr1 mice, eliminating both CD95 and TRAIL-R signaling significantly ameliorated dermatitis (Taraborrelli et al. 2018). Hence, cell death driven by CD95L and TRAIL also contributes to inflammation. The therapeutic potential this discovery may unleash is tremendous because it suggests that the prevention of disease-causing inflammation may require the simultaneous neutralization of cell death induction by multiple DR–ligand systems. Thus, for patients with a cell death–dependent disease etiology, blocking TNF may not be sufficient to achieve a lasting therapeutic benefit, whereas blocking TNF and these additional death ligands or their cell death–inducing signaling pathways, may well afford such benefit.

In summary, during the past decade, we learned that (1) TNF-driven inflammatory and autoimmune disorders can also stem from TNF-induced cell death and not only from TNF-induced gene activation; (2) failure of TNF inhibitors to provide therapeutic benefit in all patients with inflammation-associated disease does not necessarily mean that TNF is not a valuable target in nonresponders; it might simply mean that blocking TNF is not sufficient and other death ligands—or their downstream effectors—have to be blocked in addition; and (3) these other death ligands can be TRAIL and CD95L, and the downstream effector essential for mediating cell death by them and not TNF could be the kinase activity of RIPK1. This last point leaves us with an intriguing biochemical question because, if anything, TNF-induced cell death would have been expected to require RIPK1 activity but not cell death induced by TRAIL and CD95L. This unexpected result leaves us with the realization that there is more plasticity between cell death pathways triggered by the different DRs than previously thought. Extending this concept, two recent studies showed that when the proteolytic activity of caspase-8 is genetically ablated and necroptosis is inhibited, the system is rewired to activate yet a third modality of programmed cell death, caspase-1-dependent pyroptosis (Fritsch et al. 2019; Newton et al. 2019). This discovery exposes a previously unappreciated plasticity between the three major programmed cell death pathways. This rewiring likely represents an evolutionary necessity to survive the challenge of infection. The other, pathological side of the coin is something we are only beginning to unearth. It may have a major impact on our understanding of neuroinflammatory and neurodegenerative diseases (Ising et al. 2019), possibly guiding how we treat these diseases effectively in the future.

In conclusion, combining a TNF inhibitor with a RIPK1 kinase inhibitor or with inhibitors of CD95L, TRAIL, and/or other death ligands, might extend the reach of TNF-inhibitory therapies to patients who currently do not benefit from them. For example, patients with a disorder in which TNF inhibitors provide therapeutic benefit in many, but not all patients, such as RA, IBD (CD and UC), or psoriasis. However, equally, or perhaps even more, excitingly, this concept could extend to chronic inflammatory, autoimmune, and neuroinflammatory disorders in which TNF inhibition so far failed. Only future clinical trials testing the therapeutic concepts summarized here will reveal whether the insights gained from studying mouse models are also relevant to human patients.

The CD95 and TRAIL Systems in Cancer

Coincident with hope fading for TNF as a viable cancer therapeutic, two antibodies, anti-APO-1 and anti-Fas, drew attention owing to their ability to directly kill cancer cells (Trauth et al. 1989; Yonehara et al. 1989). Their targets, Fas and APO-1, were found to belong to the TNFRSF (Itoh et al. 1991; Oehm et al. 1992). Disappointingly, however, both antibody and ligand-derived CD95 agonists were highly toxic, causing massive hepatocyte death and fulminant hepatitis (Ogasawara et al. 1993).

Subsequently, the CD95L–CD95 system was shown to stimulate cancer cell proliferation, migration, and invasion (Owen-Schaub et al. 1993; Barnhart et al. 2004; Chen et al. 2010), including in glioblastoma (Kleber et al. 2008). Subsequent clinical testing of CD95-Fc (APG101/asunercept), a CD95L antagonist, showed that glioblastoma patients treated with this inhibitor plus radiotherapy benefited when compared with radiotherapy alone (Wick et al. 2014). Thus, based on our current understanding of the cancer biology of the CD95L–CD95 system, its inhibition rather than activation appears to provide benefit for cancer patients.

Soon after it was realized that CD95 agonists, like TNF, would not be a “magic bullet” for killing cancer cells, a new member of the TNFSF was identified. It most resembled CD95L, which was also known as FasL or APO-1L. Given that it could induce apoptosis similar to FasL, this new TNFSF member was named TRAIL or Apo2L (Wiley et al. 1995; Pitti et al. 1996). Importantly, TRAIL selectively killed cancer cells, without killing any essential normal cells, a unique combination of characteristics, which holds true both in vitro and in vivo (Walczak et al. 1997). This finding provided the scientific basis for developing TRAIL-receptor agonists (TRAs) as novel cancer therapeutics. However, TRAIL-R1- and TRAIL-R2-specific antibodies as well as a first recombinant form of TRAIL showed very limited anticancer activity in patients in clinical trials. As recently thoroughly reviewed elsewhere (von Karstedt et al. 2017), various strategies are currently underway to overcome the two issues which, together, are likely responsible for the failure of these first-generation TRAs, that is, (1) poor agonistic activity, and (2) resistance of most primary cancers to TRAIL-induced apoptosis.

Additionally, however, it recently emerged that the TRAIL–TRAIL-R system can be “hijacked” by certain cancers. Rather than killing cancer cells, recombinant TRAIL was found to act in a tumor-supportive manner (Trauzold et al. 2006), particularly in KRAS-mutated cancers (Hoogwater et al. 2010). It was then realized that endogenous TRAIL can be recruited by KRAS-mutated cancers to act in this manner. The mechanism whereby this is achieved is two-pronged: (1) by enhancing tumor cell proliferation, invasion, and metastasis in a TRAIL-R2 DD- and FADD-independent manner (von Karstedt et al. 2015), and (2) by creating a TRAIL-R2 DD- and FADD-dependent cytokine-rich micromilieu in which monocytes are polarized to become myeloid-derived suppressor cells (MDSCs) or alternatively activated (M2) macrophages (Fig. 3; Hartwig et al. 2017). Interestingly, the latter mechanism not only supports tumor growth, it also acts via immunosuppression. Therefore, TRAIL may represent a previously unrecognized immune checkpoint whose inhibition may unleash tumor immunity. It is interesting to note that TRAIL, albeit mostly in non-cancer-related contexts, has been shown to inhibit T helper type 1 (Th1) cells (Ikeda et al. 2010), promote regulatory T cells (Pillai et al. 2011), and suppress T-cell activation and proliferation by interfering with proximal T-cell receptor (TCR) signaling (Lehnert et al. 2014). It can also kill immature dendritic cells (Leverkus et al. 2000). Thus, TRAIL inhibition may act via tumor-cell-centered suppressive as well as immunity-enabling antitumor effects.

Open in a separate window

Figure 3.

Mechanisms of death receptor (DR)-mediated immunosuppression in cancer. (A) The tumor necrosis factor (TNF)-related apoptosis-inducing ligand (TRAIL)/TRAIL-R1/2 system contributes to immunosuppression via distinct mechanisms. The activation of TRAIL-R1/2 expressed by cancer cells induces the secretion of cytokines that polarize monocytes into myeloid-derived suppressor cells (MDSCs) and alternatively activated (M2) macrophages. These cell types in turn create an immunosuppressive tumor microenvironment that prevents CD8⁺ T-cell-mediated antitumor immune responses. In addition, TRAIL expressed by cancer cells or noncancer cells in the tumor microenvironment can directly kill TRAIL-R1/2-expressing immune cells, which are required for mounting an adaptive immune response, such as dendritic cells (DCs) or T cells, and thereby interfere with adaptive antitumor immunity. (B) Cancer cells have the potential to induce the expression of CD95L on different cell types of the tumor microenvironment, such as endothelial cells, MDSCs, and fibroblasts. By cross-linking CD95 on the surface of T cells, CD95L induces their demise by apoptosis, thereby facilitating tumor cell evasion from an adaptive immune attack.

Intriguingly, CD95L, arguably the best T-cell killer encoded by our genome and the physiological mediator of activation-induced T-cell death (Alderson et al. 1995; Brunner et al. 1995; Dhein et al. 1995; Ju et al. 1995), can also be expressed by various cells in the tumor microenvironment, including endothelial cells (Motz et al. 2014), MDSCs (Zhu et al. 2017), and cancer-associated fibroblasts (Lakins et al. 2018). Thus, CD95L inhibition may also act by a combination of tumor-cell-centered suppressive and immunity-enabling antitumor effects (Fig. 3).

In summary, both TRAIL and CD95L may represent the ultimate form of an immune checkpoint. Rather than suppressing tumor immunity in the subtle manner of CTLA4 and PD-1/PD-L1, they can remove cells required for adaptive antitumor immunity simply by killing them. Whether or not these different means are co-opted by a given cancer, or represent alternative ways by which particular cancers circumvent immune recognition, will determine whether TRAIL and/or CD95L inhibitors can serve as immune checkpoint blockers in their own right or whether they may (only) act in synergy with CTLA4- and/or PD-1/L1-targeting therapeutics. It will be exciting to answer these eminent questions in cancer immunotherapy.