The role of CD95 and CD95 ligand in cancer.

CD95 (Fas/APO-1) and its ligand, CD95L, have long been viewed as a death receptor/death ligand system that mediates apoptosis induction to maintain immune homeostasis. In addition, these molecules are important in the immune elimination of virus-infected cells and cancer cells. CD95L was, therefore, considered to be useful for cancer therapy. However, major side effects have precluded its systemic use. During the last 10 years, it has been recognized that CD95 and CD95L have multiple cancer-relevant nonapoptotic and tumor-promoting activities. CD95 and CD95L were discovered to be critical survival factors for cancer cells, and were found to protect and promote cancer stem cells. We now discuss five different ways in which inhibiting or eliminating CD95L, rather than augmenting, may be beneficial for cancer therapy alone or in combination with standard chemotherapy or immune therapy.

Facts

CD95 is a surface receptor that has the capacity to mediate apoptosis induction in cancer cells.

To induce apoptosis, CD95 recruits a number of proapoptotic factors including caspase-8 to form the death-inducing signaling complex when stimulated by CD95 ligand (CD95L).

Immune cells (i.e., cytotoxic killer and natural killer cells) use CD95L as one mechanism to kill cancer cells or virus-infected cells.

Most cancer cells are resistant to CD95-mediated apoptosis.

CD95L can not be used systemically for cancer therapy because of the side effects due to apoptosis induction in hepatocytes.

Open Questions

Why do most if not all cancer cells express both CD95 and CD95L?

Why do cancer cells acquire mutations in CD95 usually only in one allele?

Why are the cancer cells that are sensitive to CD95-mediated apoptosis (at least in vitro) among the most sensitive of any cells?

CD95/CD95L in the Immune System

CD95 (Fas/APO-1/TNFRSF6), a cell surface protein that belongs to the tumor necrosis factor receptor family, can mediate apoptosis when bound to its natural ligand, CD95L (CD178/TNFSF6) or stimulated with agonistic antibodies. It is ubiquitously expressed in the body, but is particularly abundant in the thymus, liver, heart, and kidney. CD95L is predominantly expressed in activated T lymphocytes and natural killer cells, and is constitutively expressed in tissues of ‘immune-privilege sites' such as the testis and eye.[1] Experiments with mutant mice have demonstrated the importance of CD95-mediated apoptosis in the maintenance of cell homeostasis and in the deletion of useless or autoreactive T cells.[1, 2, 3] A mutation found in the lpr (lymphoproliferation) mouse strain causes defective expression of CD95. Lpr mice develop lymphadenopathy and suffer from a systemic lupus erythematosus-like autoimmune disease.[4] A second mouse strain (gld, generalized lymphoproliferative disease) expresses a mutant form of CD95L. Gld mice have an abnormal phenotype similar to lpr mice, which includes lpr and autoimmune disease.[5] Complete knockout mice lacking either CD95 or CD95L expression also show an autoimmune phenotype that is more pronounced than the one seen in lpr or gld mice.[6, 7, 8] A third mutant mouse strain with an lpr-like phenotype (lprcg) was found to have a point mutation (T to A) in the center of the CD95 cytoplasmic region. This mutation generates a receptor in which the ability to transduce an apoptotic signal is blocked.[9] In a related human condition, autoimmune lymphoproliferative syndrome (ALPS),[10] ALPS type Ia patients carry dominant-negative mutations in CD95 and type Ib patients have mutations in CD95L, resembling mice with lprcg and gld mutations, respectively.

Canonical Signaling of CD95 in Cancer

CD95 is predominantly located at the cell surface, where it has been shown to pre-associate in homotrimers.[11] Similar to all death receptors, CD95 carries a conserved stretch of 80 amino acids in its cytoplasmic tail, the death domain (DD), that is essential for apoptosis initiation.[1, 12, 13] Upon binding of CD95L, the CD95 DD assembles the death-inducing signaling complex (DISC) composed of CD95, the adaptor molecule FADD (Fas-associated with a death domain), procaspase-8, procaspase-10, and the caspase-8/10 regulator c-FLIP.[13] Activated caspase-8 then initiates the apoptotic program by cleaving various intracellular proteins resulting in the execution of apoptosis.[14] Likely, the most established proapoptotic activity of CD95 is to mediate the apoptotic death of either virus-infected or cancer cells when engaged by a CD8+ cytotoxic lymphocyte (CTL; Figure 1). In addition to the perforin/granzyme pathway[15] and some indirect mechanisms involving cytokines such as tumor necrosis factor-α (TNFα) and interferon-γ,[16, 17, 18, 19, 20, 21, 22, 23] CD95/CD95L is a direct major system that both CTLs as well as CD4+ cytolytic effector T cells use to eliminate neoplastically transformed cells.[24, 25, 26, 27, 28, 29] CD95 can also mediate receptor interacting protein (RIP)-1-dependent necroptosis under circumstances of caspase inhibition or knockdown of TRAF2.[30, 31] However, the physiological relevance of this activity for cancer has not been established. Expression of CD95 and CD95L by cancer cells implies that they are themselves resistant to CD95-mediated apoptosis. Indeed, most cancer cells are relatively resistant to CD95-induced apoptosis even with high levels of CD95 at the surface of the cells.[32] Cancer cells have multiple ways of becoming resistant to a possible apoptotic insult mediated by CD95. A common mechanism used by the cells is to regulate cell surface expression of the receptor.[33, 34] The CD95 apoptotic signal can also be inhibited at the level of the DISC via increased expression of cFLIP (cellular FLICE inhibitory protein), which can inhibit the interactions of caspase-8 and -10 with the DISC,[35] or via reduced expression of FADD[36] or caspase-8.[37, 38] Loss of apoptosis signaling through CD95 can also be the consequence of deregulation of the expression of the Bcl-2 family proteins or inhibitor of apoptosis proteins, thereby favoring tumor survival.[39]

Figure 1

The canonical apoptosis-inducing function of the CD95/CD95L system in killing cancer cells. Cancer cells that are recognized by CTLs in an antigen-specific way are being attacked by direct mechanisms: release of perforin/granzyme (shown as GrB) or use of CD95L to engage CD95 on the surface of cancer cells. Alternatively, indirect mechanisms are activated that result in upregulation of cytokines such as TNFα and INFγ, which in turn cause upregulation of CD95 and MHC-I (by IFNγ) or induction of cell death through cancer-expressed TNF receptors (by TNFα). Ag, antigen; CTL, cytotoxic lymphocyte; GrB, granzyme B; IFNγ, interferon γ; MHC-I, major histocompatibility complex I; TCR, T-cell receptor; TNFα, tumor necrosis factor α

Other Activities of the Apoptosis-inducing Receptor CD95

In addition to the activities of CD95 and CD95L in mediating apoptosis induction, mostly in the contest of an immune response,[1, 2, 3] it is now established that CD95 has multiple nonapoptotic activities.[40, 41, 42, 43] For example, CD95 is required for efficient liver regeneration following partial hepatectomy;[44, 45] CD95 activation stimulates renal tubular epithelial cell migration by a β8 integrin-dependent mechanism,[46] and CD95 provides a mitogenic signal in quiescent hepatic stellate cells through activating epidermal growth factor receptor (EGFR).[47] CD95 is also important for neurite outgrowth.[48, 49] CD95 and CD95L have additional, cancer-relevant, activities. We have identified at least five cancer-relevant activities of CD95 that could be targeted for cancer therapy, and one (apoptosis induction through CD95) that should not be (Figure 2).

Figure 2

Graphical summary of the role of CD95/CD95L in cancer. Together with the tumor-suppressing ability to trigger apoptosis in (apoptosis sensitive) cancer cells (1), CD95L has a range of tumor-promoting activities, some of which are indirect, such as the suppression of the immune response in the cancer micro-environment by either tumor-generated CD95L (2) or by CD95L expressed by endothelial cells (3), and some of which are direct, such as the promotion of tumor growth and invasiveness (4) or the acquisition of a CSC phenotype (5). Importantly, a low-baseline level of CD95/CD95L signaling is required for survival of cancer cells. Elimination of CD95/CD95L signaling leads to an irreversible and effective type of cell death, DICE, which predominantly affects CSCs (6). CSC, cancer stem cell, CTL; cytotoxic T lymphocyte; IL-10, interleukin 10; EMT, epithelial-to-mesenchymal transition; PGE2, prostaglandin E2; TIL, tumor-infiltrating lymphocyte; VEGF-A, vascular endothelial growth factor A. Stippled arrows indicate hypothetical interactions

Apoptosis induction through CD95

Apoptosis induction is the most well-established activity of CD95, documented by thousands of publications and summarized in numerous review articles (e.g., Nagata,[1] Peter and Krammer,[13] and Nagata[50]). In the context of cancer, it is relevant that CD95L is one of only a few molecules that immune cells use to activate apoptosis to kill cancer cells (Figure 1).[51] Apoptosis induction as a cancer cell killing strategy is presumed to be accomplished by tumor-infiltrating lymphocytes expressing CD95L (Figure 2-1, apoptosis). Apoptosis induction in cancer cells through CD95 is the only scenario in which recombinant CD95L could be used for cancer therapy. However, given the fact that almost all established cancers express CD95, and the fact that most cancer cells are resistant to apoptosis induction, we would suggest that stimulating CD95 on cancer cells may not be an effective approach to killing cancer cells. In addition, stimulation of CD95 could never be used therapeutically because of major side effects such as massive apoptosis induction in the liver.[52] Based on recent data, we propose that inhibiting the activity of CD95L or targeting CD95L mRNA may be more effective for cancer therapy than using CD95L to induce apoptosis in cancer cells:

The tumor strikes back

It has been demonstrated a number of times that expression of CD95L by apoptosis-resistant tumor cells enables a powerful ‘counterattack' against antitumor immune effector cells, such as cytotoxic killer cells, many of which are themselves sensitive to CD95L-mediated apoptosis[53, 54, 55] (Figure 2-2, tumor counterattack). However, while there is some evidence for the occurrence of this counterattack, its existence remains controversial.[56] The reported increased concentration of soluble CD95L (sCD95L) in the serum of many cancer patients was often interpreted in the context of the CD95L counterattack theory (Table 1). Upregulation of CD95L in patient sera would suggest a possible immunosuppressive role for this molecule. However, the generalized immune suppression that would be expected from this situation could not be confirmed in cancer patients; thus, it may be that the increase in CD95L expression in tumor tissues has a more direct role in tumor progression.

Table 1

Tumor-promoting activities of CD95 and CD95L in clinical studies

Cancer type	Observation	Reference
Gynecological malignancies	aHigh serum CD95 was a negative prognostic marker for cervical, endometrial, and ovarian cancer	bKonno et al.[130]
B-CLL	aHigh serum CD95 was a negative prognostic marker	Osorio et al.[131]
Bladder cancer	CD95L expression (PCR) higher in cancer	bMuschen et al.[132]
	aSerum CD95L correlated with disease progression	Mizutani et al.[133]
	aSerum CD95 and CD95L were negative prognostic indicators	Mizutani et al.[134]
	A correlation existed between CD95L expression (IHC) and high tumor grade and stage	bChopin et al.[135]
	Urine CD95 level was significantly higher in cases with recurrent disease	Yang et al.[136]
Breast cancer	CD95L (IHC, PCR, WB) upregulated in cancer	bGutierrez et al.[137]
	aCD95L (IHC) correlated with lymph nodes metastasis and larger tumor size	Mottolese et al.[138]
	aCD95L/CD95 ratio >1 (PCR) correlated with worst prognosis	Reimer et al.[139]
	CD95L expression (IHC, PCR, WB) higher in cancer	bMullauer et al.[140]
	aSerum CD95L was a negative prognostic indicator	bBewick et al.[141]
	aCD95L (IHC) correlated with worse overall survival	Munakata et al.[142]
Cervical cancer	Serum levels of CD95L higher in cancer and correlating with CD95L in tumors	bSong et al.[143]
	CD95L expression (IHC) higher in cancer, correlating with disease progression	bWu et al.[144]
Cervical squamous cell carcinoma	aCD95L overproduction (IHC) was more frequent in advanced-stage tumors and was inversely related to survival	bLerma et al.[145]
Colon cancer	Metastasizing subpopulations of colorectal tumor cells express CD95L more frequently (RT-PCR) than the primary carcinomas	bMann et al.[146]
	A positive linear correlation was found between CD95L expression (IHC) and tumor progression throughout the colorectal adenoma–carcinoma sequence	Belucco et al.[147]
	High CD95L (IHC) expression correlated with lymph node involvement and distant metastases	bNozoe et al.[148]
	CD95L expression (IHC) higher in cancer, correlating with disease progression	bZhang et al.[149]
	Locally aggressive and metastatic human colon tumors express CD95L	Li et al.[150]
	aHigh serum CD95L levels were associated with poor survival	Hoogwater et al.[151]
	CD95L expression (IHC) correlated with disease progression	Kykalos et al.[152]
	CD95L expression (IHC) increased during cancer progression	Zheng et al.[112]
	aHigh CD95L expression (IHC) was significantly correlated with disease recurrence following neoadjuvant chemoradiotherapy	bSaigusa et al.[153]
Esophageal squamous cell	CD95L expression (IHC) correlated with metastases but had no impact on survival	bShibakita et al.[94]
carcinoma	aSerum CD95L was a negative prognostic indicator	bTsutsumi et al.[154]
	aLonger disease-free survival for CD95L (IHC)-negative tumors	Kase et al.[97]
Gastric cancer	Serum levels of CD95L higher in cancer	Ichikura et al.[155]
	aCD95L-positivity (IHC) correlated with lymph node metastases and poor outcome	bNagashima et al.[156]
	Upregulation of CD95L (IHC) correlated with the tumor progression	bOsaki et al.[157]
	CD95L expression was significantly correlated with tumor size, invasive depth, and metastasis	bZheng et al.[158]
	CD95L (IHC) upregulated in cancer	bNada et al.[159]
Hepatocellular carcinoma	Serum levels of CD95L higher in cancer	Tanaka et al.[160]
	Serum levels of CD95L higher in cancer	El Bassiouny et al.[161]
Large granular lymphocytic leukemia	Serum levels of CD95L higher in cancer	Tanaka et al.[160]
Natural killer cell lymphoma	Ascites-derived ovarian cancer cells secrete soluble CD95L (WB)	bAbrahams et al.[162]
Ovarian cancer	aCD95L (IHC)-positive cases showed a less favorable prognosis than those without CD95L expression	Munakata et al.[163]
	High CD95L expression (WB) is found in tumor-derived membrane fragments and in endometrial cancer correlates with the stage of the disease	bTaylor et al.[164]
	aPatients with a high post- and pre-operative CD95L serum expression ratio (ELISA) had worse prognosis to chemotherapy	Chaudhry et al.[165]
Ovarian and endometrial cancer	Serum soluble CD95 and CD95L correlated with disease progression	bBellone et al.[166]
Pancreatic cancer	Serum levels of CD95L higher in cancer	Hazar et al.[167]
	CD95 expression (gene array and IHC) correlated with stemness and EMT markers and was also in metastatic versus primary pancreatic ductal adenocarcinoma	Teodorczyk et al.[83]
Pediatric ALL, B-cell NHL	aHigh CD95 (IHC) associated with lymph node metastasis and worse survival	Macher-Goeppinger et al.[168]
Renal cancer	aHigh CD95/CD95L (IHC) neoplastic cells showed a more aggressive clinical behavior	Somma et al.[169]
	aHigh CD95 (PCR) correlated with worse overall survival	Sejima et al.[170]
Oral squamous cell carcinoma	aCD95-positive cancers (IHC) had a better response to chemotherapy and outcome	Muraki et al.[171]
	CD95L expression (IHC) increased in cancer especially in patients with lymph node metastases	bFang et al.[172]
Testicular germ cell cancer	CD95L expression (PCR) higher in cancer	Hara et al.[173]
	CD95L expression (IHC, PCR, WB) higher in cancer	bBaldini et al.[174]
Thyroid cancer	CD95L (IHC) upregulated in cancer	Rzeszutko et al.[175]
	aPatients with recurrence had higher levels of soluble CD95L expression	Owonikoko et al.[176]

Abbreviations: ALL, acute lymphoblastic leukemia; CLL, chronic lymphocytic leukemia; ELISA, enzyme-linked immune assay; IHC, immunohistochemistry; NHL, non-Hodgkin's lymphoma; WB, western blotting

CD95 or CD95L served as prognostic marker

In these reports, upregulation of CD95L in cancer was solely discussed in the context of the CD95L counterattack model

The tumor endothelium expresses CD95L

Recently, the tumor strikes back concept was rediscovered in a different form. CD95L is expressed on the tumor endothelium in mice and humans[57, 58] (Figure 2-3, endothelial cell barrier). CD95L was reported to be expressed by tumor epithelium of various human solid cancers but not by normal endothelial cells.[59] Tumor cells were found to cause upregulation of membrane-bound (m)CD95L on endothelial cells through the action of interleukin 10, prostaglandin E2, and vascular endothelial growth factor A. Interestingly, mCD95L only induced apoptosis of effector killer T cells but not of regulatory T cells, which were found to be protected by expression of a number of antiapoptotic proteins including cFLIP, Bcl-2, and Bcl-xL. This finding was supported by a syngeneic in vivo mouse model of ovarian cancer, in which it was demonstrated that expression of CD95L on endothelial cells causes reduced CD8 T-cell infiltration into the tumor. Finally, it was shown that mice treated with a neutralizing anti-CD95L antibody show increased infiltration of adoptively transferred tumor vaccine-primed CD8 T cells.[59] These data suggest that inhibiting endothelial CD95L expression could be a new therapeutic strategy to enhance the potency of adoptive transfer of antitumor T cells.

The tumor-promoting activities of CD95

Although the concept of inducing apoptosis in cancer cells using death ligands such as CD95L was intriguing, it was unlikely that the only function of CD95 was to induce apoptosis. As early as 1993,[60] it was recognized that CD95 also induces proliferation in various cell types such as T cells, liver cells, and neurons.[45, 48, 49, 61, 62, 63] In 2004, we reported that stimulation of CD95 on 22 apoptosis-resistant cancer cell lines increases their motility and invasiveness in vitro.[64] In a study with cells from ALPS patients, as well as cellular and mouse model systems, we demonstrated that nonapoptotic signaling through CD95 involved activation of NF-κB and the three MAP kinases, Erk1/2, JNK1/2, and p38.[64, 65, 66, 67] In addition, we demonstrated in various cancer cell lines that CD95-mediated invasiveness requires activation of NF-κB and ERK, and involves active caspase-8 and urokinase plasminogen activator.[64] It is now widely accepted that once cancer cells acquire resistance to CD95-mediated apoptosis, further stimulation of CD95 is tumorigenic (Figure 2-4, invasiveness and growth).[64, 68, 69, 70, 71, 72, 73, 74, 75] CD95L is expressed in two flavors, a membrane-bound form and a soluble form that is generated through cleavage of mCD95L by metalloproteinases.[76, 77] mCD95L in vivo is essential for apoptosis induction, whereas sCD95L has nonapoptotic activities and may be the predominant tumor-promoting activity in vivo.[78] The concept that CD95 can be a tumor promoter has now gained wide acceptance, supported by a number of reports describing marked activities of CD95 in tumor growth and spread (Table 2).

Table 2

Tumor-promoting activities of CD95 and CD95L in vitro and in mouse models

Cancer type	Observation	Reference
Multiple cancers	Stimulation of 22 breast, ovarian, lung, colon, renal, melanoma, or glioblastoma cancer cell lines through CD95 caused them to increase in motility and invasiveness by activating NF-κB and MAP kinase pathways and upregulation of uPA	Barnhart et al.[64]
	Knockdown of either CD95 or CD95L resulted in reduced growth of ovarian, liver, colon, and breast cancer cell lines in vitro, and of ovarian cancer cell lines in xenografted mice	Chen et al.[44]
	In lung cancer, GBM and hepatocellular carcinoma cell lines CD95L increased motility and cell growth through binding to c-Met	Lin et al.[177]
	Knockdown of either CD95 or CD95L resulted in induction of cell death in 12 cancer cell lines representing ovarian, liver, breast, cervical, colon, renal cancer, neuroblastoma, or glioblastoma	Hadji et al.[119]
	Stimulation of CD95 on breast, ovarian, renal, colon cancer, and glioblastoma cell lines increases cancer stemness	Ceppi et al.[115]
Breast cancer	Stimulation of CD95 on triple-negative breast cancer cells by soluble CD95L resulted in Yes/Orai1/EGFR/PI3K-mediated migration	Malleter et al.[82]
	Blockade of CD95 signaling in 4T1 cancer cells markedly reduced tumor growth, inhibited tumor metastasis in vivo, and prolonged survival of tumor-bearing mice	Liu et al.[178]
Colon cancer	Expression of CD95L on colon cancer cells greatly increased their local growth and ability to metastasize to the liver	Li et al.[150]
	CD95-driven liver metastasis of CD95-stimulated colon cancer cells is dependent on oncogenic Kras	Hoogwater et al.[70]
	Radiofrequency ablation of colorectal liver metastases induced hypoxia, which caused autocrine activation of CD95-promoting local invasion and accelerated metastasis outgrowth	Nijkamp et al.[75]
	CD95 triggering resulted in an increased metastatic ability and activation of EMT in cells resistant to oxaliplatin	Ametller et al.[125]
	CD95 stimulation induced phosphorylation of phospholipase C-γ1 through the platelet-derived growth factor receptor-β, resulting in phosphatidylinositol (4,5)-bisphosphate (PIP2) hydrolysis, liberating cofilin from the plasma membrane to initiate cortical actin remodeling in turn increasing tumor cell invasion	Steller et al.[81]
Gastrointestinal cancer	CD95 stimulation induced ERK1/2-driven EMT and motility	Zheng et al.[112]
Hepatocellular carcinoma	Neutralizing CD95L in a transgenic model of hepatocellular carcinogenesis reduced both inflammation and tumor formation	Nakamoto et al.[179]
	Mice with a point mutation in the CD95 DD expressed only on non-hematopoietic cells developed spontaneous liver cancer independent of the lack of apoptosis induction through CD95	Park et al.[180]
	Mice with tissue-specific deletion of CD95 in hepatocytes showed a 50% reduce occurrence of DEN-induced liver cancer	Chen et al.[44]
Histiocytic sarcoma	Cancer formed in the liver of mice engineered to express only soluble and lacking expression of membrane-bound CD95L	La et al.[78]
Lung cancer	CD95 overexpressing Lewis lung carcinoma (3LL) cells grew faster in vivo in syngeneic mice when compared with control-transfected cells	Lee et al.[68]
	CD95 ligation induced 3LL cells to produce the proinflammatory factor PGE2 by activating p38 contributing to CD95 ligation-induced chemoattraction of myeloid-derived suppressor cells	Zhang et al.[69]
	CD95-mediated activation of NF-κB was found to contribute to the resistance of lung cancer to a EGFR tyrosine kinase inhibitor	Bivona et al.[79]
Melanoma	Stimulation of B16 cells by exosome-derived CD95L in vitro activates NF-κB and ERK, and in vivo increases migration to the lung	Cai et al.[181]
Ovarian cancer	Mice lacking expression of CD95 in the surface epithelial cells of the ovaries barely developed cancer in a mouse model of endometrioid ovarian cancer driven by oncogenic Kras and deletion of pten	Chen et al.[44]
	Tissue-specific deletion of CD95 in the ovaries resulted in an increase in inflammation in the ovaries and reduced tumor development in a model of low-grade ovarian cancer driven by oncogenic Kras and deletion of pten. All outgrowing cancer cells still expressed at least one allele of wt CD95	Hadji et al.[119]
Pancreatic cancer	Stimulation of TRAF2 overexpressing cells resulted in increased invasiveness by activating NF-κB and AP-1 resulting in upregulated uPA	Trauzold et al.[73]
	Stimulation of CD95 on FADD knockdown cell lines mediated cell survival by recruiting calmodulin and Src resulting in activation of ERK	Yuan et al.[72]
	CD95 was identified as upregulated on cancer stem cells driving cell cycle progression by using Sck. Invasiveness and tumor growth could be inhibited in vivo by blocking CD95L	Teodorczyk et al.[83]
Thyroid cancer	Stimulation of CD95 induced cell growth through ERK, NF-κB, and AP-1	Mitsiades et al.[182]

CD95 is coupled to multiple potentially tumorigenic signaling pathways. CD95 was identified in a small hairpin RNA (shRNA) screen as a modifier that renders human lung adenocarcinomas resistant to EGFR tyrosine kinase inhibitors through activation of NF-κB.[79] Others have demonstrated that CD95 mediates invasion via the Src/PI3K/GSK3β/MMP (matrix metalloproteinase) pathway;[74, 80] however, the transactivation of tyrosine kinases by CD95 is incompletely understood. In colon cancer, it was shown that activated CD95 promotes the formation of cell protrusions through a new signaling pathway involving platelet-derived growth factor receptor-beta mediated phospholipase C-γ activation and phosphatidylinositol (4,5)-bisphosphate hydrolysis.[81] The subsequent release of cofilin from the plasma membrane and the continued suppression of LIMK1 by Kras/RAF1 together allow robust activation of the cofilin pathway. Cofilin activation was shown to be required for CD95-stimulated formation of membrane protrusions and increased tumor cell invasion. Recently, metalloproteinase-cleaved CD95L was reported to trigger a motility-inducing signaling complex formation in triple-negative breast cancer cells.[82] Most recently, it was shown that CD95-mediated activation of Sck/Shc2 is indispensable for cell cycle progression of metastatic pancreatic ductal adenocarcinoma (PDAC).[83] These data suggest that CD95 is connected to a myriad of prosurvival and migratory signaling pathways.

We recently tested the relevance of these nonapoptotic functions of CD95 and CD95L for cancer cells. We knocked down either CD95 or CD95L in numerous cancer cell lines using multiple small interfering RNA (siRNAs) and shRNAs. This resulted in a profound reduction in growth of the cancer cells.[44] In addition, we generated tissue-specific knockout mice lacking CD95 expression in the liver or on the surface epithelial cells of the ovaries. Using appropriate tumor mouse models, we found a severe reduction in liver cancer in mice lacking CD95 in hepatocytes (diethylnitrosamine injection model), and mice lacking CD95 in the ovaries barely developed cancer at all (using the KrasD12G/pten−/− endometrioid ovarian cancer model[84]). Finally, it was shown that mice that only express soluble but not mCD95L suffer from large histiocytic sarcomas in the liver,[78] likely owing to a lack of apoptosis induction and a tumorigenic activity of CD95L.

A number of studies reported CD95 as a positive prognostic marker for cancer.[85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99] This is likely owing to the fact that CD95 is often downregulated during tumor progression because cancer cells need to lower the risk of undergoing apoptosis while benefiting from CD95's tumorigenic activities. Occasionally, CD95L was also described as a positive prognostic marker for cancer.[86, 100, 101] However, the vast majority of reports have shown that disease progression is associated with progressively increased expression of CD95L and sometimes also CD95, and expression of both CD95 and especially of CD95L in most cases act as negative prognostic markers for many cancers (Table 1). In summary, most studies suggest that CD95 and/or CD95L expression promotes tumor growth and favors the establishment of tumor metastases.

Maintenance of CSCs by CD95 and CD95L

The cancer stem cell (CSC) model is an attractive hypothesis that translates properties of normal stem cells into the cancer field, and explains some of the most lethal features of cancers. The CSC model proposes that the cells within a tumor are hierarchically organized, and it predicts the existence of a subpopulation of cells with high tumorigenicity that are able to both self–renew and to generate differentiated cells (non-CSCs).[102, 103] One of the most malignant features of cancer is the appearance of relapses, sometimes years after radiotherapy or chemotherapeutical intervention, and this has been related to the occurrence of cells with the CSC phenotype.[104] Therefore, elucidating the mechanisms of CSC maintenance is important for understanding tumor cell persistence and relapse, and may enable specific targeting of CSCs, a promising therapeutic strategy to stably eradicate cancer.[105, 106]

CD95 and CD95 signaling have been connected to normal stem cells.[45, 107] CD95 was, in fact, previously identified as a candidate stem cell marker (along with well-established stem cell markers such as Lin28, Oct4, Nanog, and Sox2, among others) in a serial analysis of gene expression profiling of human embryonic stem cells.[108] Functional evidence of a prosurvival function of CD95 and CD95L signaling in normal stem cells came from experiments that showed that the stimulation of CD95 signaling in neuronal stem cells did not cause death, but rather increased the survival of neuronal stem cells via a Src/PI3K/AKT/mTOR signaling pathway, while, conversely, deletion of CD95 resulted in reduced neurogenesis.[107] Because normal stem cells are often the origin of CSCs, these data were suggestive that CD95 may also have a nonapoptotic function in CSCs.

In the context of cancer, CD95 expression and CD95 signaling have been connected with the differentiation of cells. We reported this based on an analysis of the NCI-60 panel of cancer cells, which could be divided in two super-clusters with distinct differentiation stages that responded differently to CD95 stimulation.[109] Interestingly, expression of CD95 inversely correlated with expression of the stem cell-inhibiting members of the let-7 family of micro RNAs (miRNAs), and stimulation of CD95 caused a reduction in let-7 expression.[110] Moreover, and related to this, CD95 has been shown to be capable of inducing the epithelial-to-mesenchymal transition (EMT) differentiation program in gastrointestinal cancer[111, 112] (Figure 2-5, EMT and CSC maintenance). In these studies, the authors demonstrated that CD95 signaling inactivates GSK3β by ERK/mitogen-activated protein kinase signaling resulting in increased nuclear import and interaction between AP-1 and NFAT4. This increases their transcriptional activity leading to nuclear accumulation of Snail and β-catenin and miR-23a expression, and subsequently, downregulation of E-cadherin and upregulation of MMP9 and vimentin in vivo and in vitro.[111, 113] EMT has been previously connected with the generation of cells with the properties of CSCs.[114]

We recently demonstrated that CD95 is required for the survival of CSCs and allows new CSCs to emerge[115] (Figure 2-5, EMT and CSC maintenance). Stimulation of CD95 on multiple tumor cells induced a conversion from non-CSCs to CSCs. This reprogramming activity of CD95 was independent of its apoptosis-inducing function, as it was not blocked by the pan-caspase inhibitor zVAD-fmk; rather, it represents a mechanism of retro-differentiation. Strikingly, CSCs from highly apoptosis-sensitive HeyA8 ovarian cancer cells enriched in tumor spheres were found to be almost completely resistant to CD95-mediated apoptosis. For breast cancer, we could connect this novel function of CD95/CD95L to the activity of miR-200, a miRNA previously linked to both EMT and CSCs.[114, 116, 117] miR-200c expression increased the sensitivity of cancer cells to CD95-mediated apoptosis.[118] Stimulation of CD95 not only increased the number of cancer cells with stem cell traits but also prevented differentiation of CSCs, suggesting that CD95 expression on cancer cells maintains the CSC pool.[115] A connection between CD95 and CSCs was recently also reported for PDAC.[83] CD95 expression strongly correlated with stemness and EMT markers and blocking CD95L reduced tumor growth and metastasis in vivo.

Death induced by CD95R/L elimination

Following up on our finding that CD95 contributes to the proliferation of cancer cells,[44] we recently reported that the elimination of either CD95 or CD95L kills cancer cells (in vitro and in vivo) in a process we termed DICE (death induced by CD95 or CD95L elimination)[119] (Figure 2-6, DICE). This activity of CD95 as a survival factor seems to be mostly relevant to cancer cells, as none of the normal tissues during embryonic development in either CD95 or CD95L knockout mice showed a growth defect or signs of cell death.[6, 7, 8] Consistently, we found increased sensitivity to DICE in ovarian surface epithelial cells after they were immortalized by expression of hTERT.[119]

We found that all cancer cells tested (~40 lines tested to date) substantially die by DICE when either CD95 or CD95L is knocked down. We used 15 different non-overlapping si/shRNAs against either of the two genes, and all induce DICE. We generated Tet-inducible vectors (pTIP) to express the shRNAs. They kill all cancer cells when doxycycline is added. In two ovarian cancer mouse models and one mouse model of chemically induced liver cancer, tumor formation was severely reduced in the absence of CD95.[44, 119] In fact, a reanalysis of the tumor samples revealed that not a single cancer cell could be detected in any of the models that had deleted both alleles of CD95.[119] We reported that DICE has the following properties:[119]

DICE represents a necrotic form of mitotic catastrophe with signs of apoptosis, autophagy, and senescence (unpublished data).

DICE is characterized by cell swelling and production of reactive oxygen species followed by DNA damage and activation of caspase-2, resulting in mitochondrial outer membrane permeabilization. Cells eventually die by a RIP1/mixed lineage-like kinase-independent mechanism. Although multiple cell death pathways are activated, RIPK-dependent necroptosis does not seem to be critical, suggesting that DICE induction may not cause inflammation.

DICE could not be inhibited by any of 1200 tested drugs or by knockdown of any single gene in a genome-wide shRNA screen, suggesting that it is a robust cell death mechanism that is difficult to block.

We recently postulated that DICE is a fail-safe mechanism, a dead man's switch, that prevents the survival of cancer cells that are devoid of CD95, and, hence, would not be eliminated by the immune system through CD95L/CD95-mediated apoptosis.[120] Thus, DICE is a naturally occurring antitumor defense mechanism. The observation that in tumor cells both alleles of CD95 are rarely if ever mutated or deleted (reviewed in Peter et al.[41]) is consistent with this interpretation. Our recent data show that all cancer cells autonomously produce a small amount of CD95L, suggesting that the loss of either CD95 or CD95L induces DICE, which is consistent with our observation that cancer cells never delete both alleles of CD95.[119]

Our study of CSCs revealed a crucial role for CD95 signaling in regulating cancer differentiation, and indicated that the two death mechanisms, DICE and canonical CD95-mediated apoptosis, have opposing roles in eliminating CSCs and non-CSCs. Conversion of non-CSCs to CSCs resulted in a loss of sensitivity to CD95-mediated apoptosis and a concomitant increase in the sensitivity of the cells to DICE.[115] In fact, we found that DICE preferentially targets CSCs.[115] When DICE was induced in multiple cancer cell lines or primary breast cancer cells, they became depleted of CSCs. Cells lost typical CSC surface markers, formed spheres less efficiently, and lost expression of endogenous CSC markers while becoming enriched in the stem cell-controlling miRNA miR-200c.

Targeting CD95L to Kill Cancer Cells

The data summarized above suggest that CD95 and CD95L act as oncogenes once cancer cells have become resistant to the apoptosis-inducing activity of CD95. The data further seem to suggest that the reason that cancer cells die after removal of either CD95 or CD95L is that they are addicted to their oncogenic activities. However, for the following reasons, we would argue that DICE is not the result of a broken oncogene addiction: (1) CD95 and CD95L intrinsically have tumor-suppressive activities in the context of the immune system (see above). (2) Elimination of CD95 or CD95L can kill any cancer cell we have tested, not just cells that overexpress CD95 or CD95L. In fact, CD95L expression in most cancer cells is barely detectable, yet elimination of CD95L induces DICE more effectively in cells that express less CD95L, perhaps because CD95L becomes rate limiting more easily. CD95 and CD95L may be the first identified tumor-suppressive genes that are so important that their loss (which could occur as neoplastically transformed cells continue to acquire mutations) triggers a fail-safe program to kill such cells. An interesting aspect of this model is that, by definition, the DICE mechanism has not been triggered in any cancer cell found in a cancer patient, the implication being that cancer cells do not become resistant to DICE, but they become resistant to apoptosis and may evade DICE by retaining expression of CD95 and CD95L.

Because neither CD95 nor CD95L knockout mice are known to exhibit any defects in the proliferation of any tissue and exhibit no defects in stem cell compartments,[6, 7, 8] it is possible that CD95 or CD95L could be safely targeted for therapeutic purposes. Targeting CD95L systemically would block all the tumorigenic activities summarized in Figure 2.

Inducing DICE in Combination with Standard Chemotherapy

Although induction of DICE alone may be effective in killing cancer cells, the combination of induction of DICE with existing therapies and concepts may be beneficial in improving outcomes of cancer therapy. During our analysis of the role of CD95 in CSCs, we identified a strong synergy between DICE and CD95-mediated apoptosis in eradicating cancer.[115] The synergy is a direct consequence of the differential sensitivities of CSCs and non-CSCs to the two death mechanisms. Thus, a therapy that combined the two death mechanisms could be beneficial to cancer treatment by targeting two differentiation stages of cancer development. It has been reported multiple times that many forms of chemotherapy act by inducing, at least in part, apoptosis in cancer cells, sometimes through upregulation of CD95L.[121, 122] It is also established that cancer patients who become refractory to therapy have an increased CSC population,[123, 124] which we recently showed to be more sensitive to DICE than non-CSCs. Thus, a combination of low-dose chemotherapy coupled with targeting CD95 may be beneficial as it should target both non-CSCs and CSCs. Targeting of CD95L could also be a beneficial addition to chemotherapy because chemotherapy-induced upregulation of CD95L has been suggested to not only drive cancer cells into apoptosis but to promote growth of drug resistant tumor cells.[125]

Inducing DICE in Combination with Inhibition of Immune Checkpoint Receptors

An effective mechanism to treat certain cancers involves the mobilization of the immune system. Cancer cells have found ways to suppress the antitumor response mounted by the immune system, but recent successes of therapies that are aimed toward de-repressing the tumor-imposed block on the immune system are evidence of the power of these mechanisms. Anti-PDL1 and anti-PD-1 clinical trials have shown promising effects in melanoma, renal, colorectal, and non-small cell lung cancer patients, and, for the first time ever in the development of immune therapy, a sizeable fraction of patients were observed who had a durable response that increased their life span.[126, 127, 128, 129] Based on these early data, one can predict that success in cancer therapy will come from harnessing natural mechanisms that control cancer in general (e.g., an antitumor immune response) rather than from cancer-specific strategies. Empowering the immune system by targeting immune check point signaling and simultaneously attacking the cancer cells by inducing DICE may represent a viable combination of therapies both of which activate preexisting antitumor mechanisms.

Conclusions and Perspectives

Using CD95L for cancer therapy was never a viable option to treat cancer because of its devastating effects on the liver. Accumulating evidence now suggests that cancer cells can never lose CD95 or CD95L and if they do, they die. This provides an opportunity to use targeting either CD95 or CD95L to treat cancer. However, there are many open questions that need to be addressed first. Although excess of CD95L secreted by tumor cells may drive EMT and stemness and render tumor cells more motile and invasive, it is not clear whether targeting this secreted protein will be enough to block the tumor-promoting activities of CD95 and CD95L. In our hands, blocking the activity of extracellular CD95L has had no effect on cell viability.[119] In addition, recent data suggest that both CD95L and CD95 are located in intracellular stores,[119] hence they may exert their activity to protect cancer cells from DICE from within the cell. It may therefore be necessary to target the intracellular proteins or the mRNAs of CD95 and CD95L. This will require development of more efficient ways to deliver siRNAs to cells or the development of new technologies to eliminate genes from cells all together (i.e., using the clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9 system).