Different environmental factors (infection, pollution, UV …) involved in chronic inflammatory disorders and cancers affect the expression level and/or the interaction of different receptors and ligands, which in turn alter intracellular signaling pathways, subsequently leading to pathophysiological phenotypic changes. Death receptors (DR) are transmembrane receptors that can implement cell death signals via apoptosis, necroptosis, pyroptosis or ferroptosis. Ligands of the tumor necrosis factor (TNF) family and their receptors (TNF-R) are cytokines contributing to the induction of a caspase-dependent apoptotic death. Interestingly, these so-called “death receptors” can also trigger non-apoptotic signaling pathways involved in cell migration, differentiation, survival, and proliferation (1–5).

Six human death receptors (DRs) have been identified, TNF-R1 (6, 7), CD95 (Fas/APO-1/TNFRSF6) (8, 9), TRAIL-R1 (DR4) (10), TRAIL-R2 (DR5) (11, 12), DR3 (TRAMP) (13–16), and DR6 (also known as TNFRSF21 (17)). These death receptors are activated by TNF (18), CD95L (also known as FasL or CD178) (19), TRAIL (20), and TL1A, respectively (21), with the ligand for DR6 remaining to be confirmed even if amyloid precursor protein represents a solid option (22, 23). Apoptosis is finely regulated by these DRs, and mutations or expression deregulation of these receptors lead to various diseases (auto-immune, neurodegenerative, heart diseases or cancer) and development of chemoresistance (24).

CD95 is a ubiquitously expressed transmembrane receptor, which belongs to the TNF-R family (8). Its natural ligand, CD95L is a transmembrane protein involved in the induction of a caspase-dependent apoptotic signal (8, 25, 26). The CD95/CD95L pair contributes to immune homeostasis and surveillance, and different mutations mainly localized within the CD95 death domain (DD), an intracellular region involved in the recruitment of the adaptor protein Fas-Associated protein with Death Domain (FADD), have been associated with breakdown of self-tolerance in autoimmune lymphoproliferative syndrome (ALPS) patients (27, 28) and Lpr^Cg mice (29, 30). CD95 mutations have also been detected in lymphoma pushing the authors to classify CD95 as a tumor suppressor gene (31, 32). Although DD-localized CD95 mutations foster tumor progression by rendering tumor cells resistant to the apoptotic response (33), new and accumulating evidence support that this receptor exerts more complex biological functions, and might promote oncogenesis and inflammation/auto-immunity independently of its ability to trigger cell death (3, 34–36).

For CD95L, rare mutations have been reported in human and are associated with lupus (37) or ALPS type Ib (38, 39) pathologies. The gld (for generalized lymphoproliferative disease) mice also display a lupus-like phenotype and harbor a mutation in CD95L with the replacement of its phenylalanine 273 by a leucine (F273L). This mutation reduces the efficiency of CD95/CD95L interaction (40).

Interestingly, CD95L might also interact with another TNFR member, DR5 (41). The authors show that, although the CD95L affinity for DR5 was weaker than that for CD95 ((K_D was 1.23x10^-12 M for DR5–CD95L versus 6.01x10^-13 M for DR5–TRAIL), CD95L can compete TRAIL for DR5 binding, suggesting that both ligands share a similar interaction region in DR5 (41). More importantly, the CD95L/DR5 interaction has been suggested to promote arthritis severity in a mouse model (i.e., autoantibody-induced arthritis). Surprisingly, the K_D of CD95L for DR5 assessed in this study is far higher than that currently measured for CD95 (K_D comprised between and 7x10^-8 and 2x10^-9M (42–45), suggesting that CD95L would possess a stronger affinity for DR5 than for its own receptor. This conclusion remains to be strengthened with structural methods to definitively validate the CD95L/DR5 interaction.

At least, two main forms of CD95L exist. The transmembrane CD95L (m-CD95L) triggers cell death when it interacts with CD95-expressing cells, while metalloproteases can release soluble CD95L (s-CD95L) (46–48). Expressed by activated B and T-cells, m-CD95L contributes to the immune contraction (49) and its expression by myeloid cells participates in tissue inflammation by recruiting macrophage in damaged spinal cord (50). In this latter study, the role of m-CD95L and/or s-CD95L in the inflammatory process remains to be addressed. Contradicting studies exist on s-CD95L; while soluble CD95L can trigger apoptosis and promote lung damage in acute lung injury (ALI) (51, 52), it fails to induce cell death but rather stimulates inflammation in chronic autoimmune disorders such as systemic lupus erythematosus (SLE) (34, 48) and metastasis occurrence in cancers (53–57). Such a discrepancy might be ascribed to the stoichiometry of s-CD95L (43, 58), which seems to rely on the presence or absence of juxtamembrane N-terminal end (51, 59). In this respect, metalloproteases involved in the m-CD95L shedding as well as their preferential cleavage sites within the stalk region will directly impact the N-terminal length of s-CD95L ( Figures 1A –C ) end and thereby, its biological function as discussed below. It has been reported that m-CD95L can be shed close to its transmembrane domain releasing a s-CD95L encompassing a stalk region both in mouse (43, 60) and human (43, 61). This stalk region promotes the aggregation and the cytotoxic activity of s-CD95L. These observations point out that the presence or absence of certain metalloproteases involved in the CD95L shedding, might be responsible for the release of different ligands that either trigger cell death or aggravate inflammation or oncogenesis.

Extracellular matrix (ECM) is composed of a large number of structural and functional components that includes enzymes, collagens and proteoglycans, which are secreted and self-assembled into the immediate cellular environment (82). Other non-proteoglycan matrix components include hyaluronic acid, fibronectin, elastin, and laminin. This ECM provides structural support to cells and an integral signaling network through the action of different cytokines and growth factors interacting with the matrix components (83–86). For instance, binding of the s-CD95L to ECM, and more specifically to fibronectin, transforms the non-apoptotic ligand into a potent death inducer (87) suggesting that immobilization and/or aggregation of the s-CD95L homotrimer can foster the induction of the apoptotic response. In agreement with this observation, although a soluble and homotrimeric CD95L fails to trigger apoptosis, its hexameric counterpart (58) can do it. We also observed that the more CD95L is aggregated, the more its ability to induce apoptosis is increased (88).

Most of the ECM protein components are processed by matrix metalloproteinases (MMPs). In human, this family of zinc-dependent endopeptidases englobes 23 members sharing structural domains (89, 90). These proteases are mainly secreted within the pericellular and extracellular space (61) but can also be anchored to the cell surface (91) or adopt an intracellular localization, that has been correlated in certain cases to non-proteolytic functions (90, 92). Except during specific stages of development involving tissue remodeling (e.g., embryogenesis) and wound healing processes, there is no constitutive expression of MMPs at homeostasis. Once secreted, these enzymes coexist within the extracellular space under latent forms (zymogens) and active forms, whose proteolytic activity is finely tuned by endogenous inhibitors such as tissue inhibitors of metalloproteases (TIMPs) or alpha-macroglobulin.

Recent N-terminomics and proteomics techniques have been used to profile hundreds of cleavage sites in proteomes associated with MMP activity, which reveal that more than two-third of MMP substrates are non-ECM proteins. Accordingly, far beyond their capacity to collectively cleave the ECM substrates, MMPs can process chemokines, cytokines, cell-surface receptors, growth factors, and nuclear proteins. Thus, MMPs are involved in inflammatory response, angiogenesis, cell-to-cell communication and cell proliferation, and the deregulation of their activity contributes to the progression of many diseases including cancer, chronic inflammatory disorders, vascular and central nervous system diseases (90).

MMPs are classified according to their linear sequence similarity, domain organization and substrate specificity (90). All the MMPs share a minimal N-terminal region, consisting in a signal peptide, a pro-domain and a metalloprotease/catalytic domain (90) ( Figure 2A ). Except for MMP-7, -26 and -23, all MMPs encompass a hemopexin-like C-terminal region, which is important in determining substrate specificity and interaction with tissue inhibitors of metalloproteinases (TIMPs). This C-terminal domain plays also an important role in cell migratory function of certain MMPs. Gelatinase-A (MMP-2) and gelatinase-B (MMP-9) contain fibronectin type-II inserts within their catalytic domain. These inserts confer the ability to bind and cleave gelatin and collagen.

Membrane-type MMPs (MT-MMPs) are embedded in the plasma membrane of the cells via a transmembrane domain or a glycosylphosphatidylinositol (GPI)-anchor ( Figure 2A ). This family includes the transmembrane proteins MMP-14, MMP-15, MMP-16, and MMP-24, and the GPI-anchored proteins MMP-17 and -25 (93). Some MMPs harbor a furin-like convertase cleavage site ( Figure 2A ), which is intracellularly cleaved to activate the protease and promote its distribution at the cell surface. MMP-23 is a unique MMP that contains a cysteine array and immunoglobulin-like domain, whose exact role remains elusive.

Within the extracellular space, a disintegrins and metalloproteinases (ADAMs) family can also exert a proteolytic activity (94–96). The main substrates for ADAMs are type I and II transmembrane proteins, which make them as shedding proteases. However, these proteases are also capable of processing cytokines and growth factors (95). Interestingly, in the case of transmembrane proteins, the cleavage consistently occurs between 10 and 15 amino acids from the plasma membrane. Like MMPs, ADAMs possess several domains, including a pro-domain, a metalloprotease/catalytic domain, a disintegrin domain, a cysteine rich domain, an EGF-like domain, a transmembrane domain and a C-terminal cytoplasmic tail ( Figure 2B ). All ADAMs contain a disintegrin domain, which can bind to integrins from adjacent cells, with potential consequences in cell adhesion and migration. These metalloproteases are implicated in different diseases including cancer (95), systemic inflammation (96), cardiovascular diseases and atherosclerosis (97, 98). A critical role in kidney pathologies (99) and in immunity (100) has also been documented.

Both MMPs and ADAMs belong to the superfamily of metzincin proteases. These metzincins share a conserved HEXXHXXGXXH motif within their metalloprotease/catalytic domain, where the three histidine residues bind to the catalytic zinc ion and the glutamate, as a general acid base, and activates a water molecule required for the peptide bond hydrolysis ( Figure 2C ).

CD95L can be cleaved by several metalloproteases, including MMPs and ADAMs, to release different soluble CD95Ls (s-CD95Ls), which have been reported to induce cell proliferation, migration, survival (36) but also cell death (51, 59). Rendering more complex to predict the biological outcome of s-CD95L, this ligand can also interact with other TNFR members, including as aforementioned, DR5 (41) or the soluble receptor DcR3 (44). Despite the complexity of the signaling pathways induced by the different forms of s-CD95L and their implication in the progression of different pathologies including chronic inflammatory disorders and cancers only a limited structural knowledge exists on these s-CD95Ls.

As aforementioned, CD95 can induce a broad range of signaling pathways, with different biological outcomes. This is related to a fine-tuned control of CD95 aggregation, conformation, distribution within plasma membrane sub-domains and post-translational modifications. These parameters rely on the quality of the CD95/CD95L interaction (141). MMPs and ADAMs are responsible for the generation of soluble CD95L, that might promote metastatic occurrence in cancers or stimulate trafficking/activation of immune cells in chronic inflammatory disorders and thus, inhibiting MMP or ADAM activity could represent an attractive therapeutic strategy in these pathologies ( Figure 3 ). In addition, inhibition of the non-apoptotic signaling pathways downstream s-CD95L/CD95 interaction might also represent an attractive option to treat certain cancers and chronic inflammatory disorders. Asunercept (initially called APG101) is a decoy receptor encompassing the extracellular region of CD95 fused to the Fc domain of human IgG1. APG101 interacts with CD95L, both transmembrane and soluble forms ( Figure 3 ), and abrogates all signals induced by these ligands. Asunercept in phase I/II clinical trials exhibits encouraging therapeutic effect on myelodysplastic syndromes (142) and glioblastoma (143, 144). In addition, the therapeutic value of this drug is also under evaluation (NCT04535674) in COVID-19 patients, in whom CD95L inhibition might protect against the macrophage/neutrophil-driven damage of epithelial cells (145). Although the clinical outcomes of these trials are motivating, it remains that APG101 blocks both apoptotic and non-apoptotic signals, rendering difficult to discriminate the role of each cellular response in the pathogenesis. We recently developed a drug (i.e., peptidomimetic) neutralizing in a selective fashion, the CD95 non-apoptotic pathway (78). This drug, designated DB550, disrupts the CD95/PLCγ1 interaction and the subsequent calcium signaling pathway, which is mandatory for cell migration (77). DB550 injection in SLE-prone mice prevents Th17 cell transmigration in inflamed kidneys and alleviates clinical symptoms (78). These findings support that the selective inhibition of CD95-mediated non apoptotic pathways might turn out sufficient to treat cancers and chronic autoimmune disorders in which s-CD95L is up-regulated (36).

Regarding m-CD95L shedding, another alternative to selectively block the CD95-mediated non-apoptotic signal is to prevent the generation of s-CD95L by inhibiting metalloproteases. Beyond the fact that metalloproteases are pleotropic enzymes, whose inhibition will engender clinical outcomes difficult to predict, an additional concern is the accumulation of membrane-bound CD95L that, might favor the elimination of certain cancer or immune cells, but might also engender undesired tissue damage ( Figure 3 ). Finally, another therapeutic approach for cancer patients could be to develop methods to extinguish the CD95 expression itself. Indeed, we recently observed that the elimination of CD95 in triple negative breast cancers induces a pro-inflammatory signal and promote the anti-tumor activity of NK cells (139, 140).

LD, NG, SB, AC, MJ and PL wrote the original draft. All authors contributed to the article and approved the submitted version.

This work was supported by INCa PLBIO (PLBIO 2018-132), ANR PRCE (ANR-17-CE15-0027), and with financial support from ITMO Cancer of Aviesan within the framework of the 2021-2030 Cancer Control Strategy, on funds administered by Inserm.

PL and MJ are involved in patents protecting the use of CD95 or CD95L in chronic inflammatory disorders and cancers WO2014118317; WO2015189236; WO2015158810; WO2015104284; WO2017149012; WO2018130679.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be constructed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.