On the TRAIL to successful cancer therapy? Predicting and counteracting resistance against TRAIL-based therapeutics

Abstract

TRAIL and agonistic antibodies against TRAIL death receptors kill tumor cells while causing virtually no damage to normal cells. Several novel drugs targeting TRAIL receptors are currently in clinical trials. However, TRAIL resistance is a common obstacle in TRAIL based therapy and limits the efficiency of these drugs. In this review article we discuss different mechanisms of TRAIL resistance and how they can be predicted and therapeutically circumvented. In addition, we provide a brief overview of all TRAIL based clinical trials conducted so far. It is apparent that although the effects of TRAIL therapy are disappointingly modest overall, a small subset of patients responds very well to TRAIL. We argue that the true potential of targeting TRAIL death receptors in cancer can only be reached when we find efficient ways to select for those patients that are most likely to benefit from the treatment. To achieve this, it is crucial to identify biomarkers that can help us predict TRAIL sensitivity.

Introduction

The holy grail of cancer therapy is to find drugs that will specifically and efficiently kill cancer cells while having little to no effect on normal cells. The variability between and within different kinds of cancer and the cancer cells’ inherent ability to adapt are obstacles in obtaining this goal. Thus, there is a significant need to define those individuals that will benefit from a specific therapy while experiencing few side effects.

Since the Tumor necrosis factor-related apoptosis inducing ligand (TRAIL) (also known as APO2 ligand, APO2L) signaling pathway was initially discovered (1), (2), the plausibility of exploiting it in cancer therapy has been under debate. Initial promising studies demonstrated a remarkable specificity for inducing apoptosis in tumor cell lines but not in normal cells. While clinical trials using TRAIL therapies have shown low toxicity in patients, disappointingly small therapeutic effects have been observed when TRAIL agonists are used as a monotherapy. It is becoming increasingly apparent that TRAIL therapy may indeed be very beneficial, but perhaps only for a small subset of patients. Therefore, it is crucial to identify biomarkers that can predict patient response and to maximize the therapeutic efficacy through drug combinations that not only synergize with TRAIL but that can also overcome resistance as it arises. This review covers some of the mechanisms of TRAIL resistance that have been reported and presents an overview of all the TRAIL-based clinical trials performed to date. We argue that lessons learned from preclinical research should be much more integrated into clinical trial design as a way to select the patients most likely to respond to therapy. Only then can we truly evaluate the efficacy of this drug and see the extensive research already done in this field come to fruition in the form of increased cancer patient survival.

TRAIL signaling

TRAIL is a member of the death receptor ligand family, a subclass of the tumor necrosis factor family. The TRAIL protein is expressed on the membrane of a limited number of immune cells and is also present in a soluble form. It binds to at least five receptors. Two of these, Death Receptor (DR) 4 (also known as TRAIL receptor 1, TRAIL-R1) and DR5 (TRAIL-R2), are transmembrane receptors with a cytoplasmic death domain (DD) that transmits apoptotic signals into the cells. Two decoy receptors (DcR), DcR1 (TRAIL-R3) and DcR2 (TRAIL-R4), do not have functional DD and do not enable apoptosis activation (3). TRAIL also binds weakly to a fifth receptor, osteoprotegerin (OPG). Several pro-apoptotic receptor agonists (PARAs) that can trigger TRAIL signaling have been developed, including recombinant human TRAIL ligand and agonistic antibodies against DR4 and DR5, as discussed further below.

TRAIL signaling induces apoptosis mainly through the extrinsic, or death receptor mediated pathway. When TRAIL binds to DR4 or DR5, the receptors homotrimerize, enabling the receptors’ DD to recruit the adaptor protein Fas Associated Death Domain (FADD) and the inactive, uncleaved form of caspase 8, pro-caspase 8. The receptors, FADD, and pro-caspase 8 or pro-caspase 10 together form the Death Inducing Signaling Complex, (DISC). At the DISC pro-caspase 8 is activated, a process found to be dependent on both dimerization and cleavage (4). Activated caspase 8 then cleaves downstream substrates resulting in, ultimately, the cleavage and activation of effector caspase 3. In some cell types, called Type I cells, this activation of the extrinsic pathway is sufficient to induce apoptosis. However, in other cell types, type II cells, activation of the intrinsic (mitochondrial) apoptosis pathway is required as well. The intrinsic pathway is typically triggered by DNA damage or other cell stressors, but it can also be activated through caspase 8 or caspase 10-mediated cleavage of the pro-apoptotic BCL-2 family protein BID. When cleaved, the activated, truncated form of BID can translocate to the mitochondrial membrane where it interacts with pro-apoptotic Bcl-2 family members BAX and BAK, enabling these proteins to induce permeabilization of the mitochondrial membrane. The pro-apoptotic proteins cytochrome c and SMAC/DIABLO are then released from the mitochondria. Analogous to the DISC, cytochrome c forms a protein complex, the apoptosome, with SMAC/DIABLO, APAF-1 and procaspase 9, enabling cleavage of procasapase 9 into active caspase 9. Caspase 9 cleaves downstream effector caspases such as caspase 3, thus converging with and amplifying death receptor-mediated caspase activation (Fig. 1).

Fig 1.

The TRAIL signaling pathway. The TRAIL ligand binds to functional receptors DR4 and DR5 and non-signaling receptors OPG (not shown), DcR1 and DcR2. Binding of TRAIL ligand or receptor-specific agonistic antibodies to DR4 and DR5 induces trimerization of the receptors. The cytoplasmic part of the DR4 and DR5 receptors contain death domains (DD) that enable recruitment of Fas associated protein with death domain (FADD) and procaspase 8 (proCASP8), enabling cleavage and activation of proCASP8 to its active form caspase 8 (CASP8). CASP8 activates downstream effector caspases both directly and, in some cells, through activation of the mitochondrial apoptosis pathway through BID cleavage. Once activated, effector caspases cleave downstream substrates and induce DNA fragmentation, ultimately leading to apoptosis. For additional details, please refer to the text.

One important distinction between TRAIL-induced apoptosis and apoptosis induced by conventional chemotherapy and radiotherapy is that the latter is largely dependent on cellular damage recognition by, for example, the p53 tumor-suppressor protein. The dependence on p53 to elicit an apoptotic response poses a problem in cancer therapy since loss of p53 occurs in more than half of all cancers cells due to inactivating mutations (5). Although the TRAIL pathway is enhanced by p53 activation through upregulation of DR4, it can induce apoptosis in cells in which p53 is deleted (6). Therefore, TRAIL therapeutics may be an alternative way of eradicating tumors that are resistant to a wide range of conventional therapies.

In addition to the classical apoptosis-inducing TRAIL pathway, it is becoming increasingly evident that TRAIL signaling can also activate other pathways including the NFkB, PKB/AKT and MAPK pathways, through the formation of signaling complexes secondary to the DISC consisting of different proteins including receptor-interacting protein (RIP), FADD, caspase 8, and TNF-receptor associated protein (TRAF-2) (reviewed in (7). Some of these pathways promote survival and proliferation and importantly, in cells that are resistant to apoptosis induced by TRAIL, these pathways may still be intact. For example, in TRAIL-resistant primary leukemia cells, TRAIL induces proliferation in previously resting cells, accelerates the doubling time of proliferating cells, and reduces spontaneous cell death in a NF-κB-dependent manner (8). Furthermore, TRAIL promotes metastasis from TRAIL-resistant human pancreatic ductal carcinoma xenografts (9). Thus, if patients that have TRAIL resistant tumors are treated with TRAIL, it may actually lead to increased tumor burden and metastasis. This highlights the critical importance of identifying patients whose tumors will be responsive to TRAIL-induced death, and to develop means to counteract TRAIL resistance through combinations with other drugs.

The physiological role of TRAIL

The physiological function of TRAIL is reported to be in immune surveillance and immune mediated tumor suppression. TRAIL is exclusively expressed on immune cells and can be triggered by antigens such as lipopolysacharides (LPS) and cytokines such as interferons (IFNs), suggesting that TRAIL plays a regulatory role in the immune system (10–12) TRAIL deficient (TRAIL −/−) mice are viable, developmentally normal, and fertile (13,14). However, TRAIL−/− mice have a decreased level of lymphoid and myeloid cell death and are less sensitive to infection by Listeria monocytogenes (15). In addition, mice that lack the functional mouse TRAIL receptor mDR5 exhibit an enhanced innate immune response and are more resistant to murine cytomegalovirus (16), demonstrating a role of TRAIL in the negative regulation of both innate and adaptive immunity. The first evidence for a role of TRAIL in tumor suppression came from a study by Walczak and colleagues that demonstrated a reduction in the size of human mammary adenocarcinoma xenografts in mice administered soluble recombinant TRAIL(17,6). Accordingly, in both TRAIL −/− mice and mice treated with a neutralizing anti-TRAIL antibody, inoculation with renal carcinoma and mammary carcinoma cells results in faster growing and more metastatic tumors than in wild type mice (13), and aged TRAIL −/− mice have an increased incidence of lymphoma (18).

TRAIL as a selective killer of cancer cells

Interest in TRAIL as an anti-cancer agent was initiated by early studies in which TRAIL-induced killing could be demonstrated in a wide variety of tumor cells in vitro (1) (2) and in vivo (17, 6, 19), whereas normal cells were unaffected by TRAIL treatment. Early on, an alarming study indicated that TRAIL was toxic to normal human hepatocytes, seemingly disqualifying TRAIL for clinical applications because of the risk of hepatotoxicity or fulminant hepatic failure (20). The group attributed the discrepancy to a difference in model systems, but the toxicity was later shown to be an artifact caused by the particular preparation of TRAIL. The recombinant proteins used in this study were tagged with polyhistidine and also had a suboptimal level of zinc, changing the tertiary structure of the protein and causing it to over-aggregate. The tagged preparation was not only toxic to normal hepatocytes but was also less efficient at inducing apoptosis in cancer cells (21). Untagged trimeric proteins with optimal zinc content showed an absence of hepatocyte toxicity both in primary human hepatocyte culture (22) and in systemic administration to non-human primates (6, 21)

The physiological role of TRAIL signaling in tumor suppression and its specificity for cancer cells versus normal cells has important implications. Since TRAIL signaling is a physiological means of preventing cancer, it is plausible that induction of this pathway would be an efficient way to combat existing tumors in a clinical setting. Furthermore, TRAIL-induced killing should be specific to cancer cells, sparing normal cells. However, since TRAIL-induced killing is likely part of the body’s first-line defense against cancer, successful tumors may need to acquire resistance against TRAIL-induced apoptosis. Indeed, a large proportion of tumor cells that metastasize are resistant to TRAIL, and TRAIL resistance may also arise during the course of cancer treatment. Thus we have a potential “Catch-22” situation, where the attributes that make TRAIL receptor targeted drugs attractive (their tumor selectivity and the fact that they are boosting a physiological tumor suppression mechanism) are likely to be selected against during tumor evolution. Therefore one would expect that the ability to predict and circumvent TRAIL resistance will be essential for the successful clinical use of these drugs.

TRAIL resistance

Multiple mechanisms of TRAIL resistance have been identified. Theoretically, resistance mechanisms could target any of the events starting from TRAIL ligand binding through to the endpoint of apoptosis by affecting the expression and/or function of TRAIL pathway components (Fig.2). Some of the major mechanisms are discussed below.

Fig 2.

TRAIL signaling pathway and some points of therapeutic intervention. The functional TRAIL receptors DR4 and DR5 can be activated by proapoptotic receptor agonists such as Mapatumumab (DR4), Lexatumumab (DR5) and Dulanermin (both DR4 and DR5). Some drugs, such as HDAC inhibitors and proteasome inhibitors, augment apoptosis by upregulation of receptor expression. Caspase 8 is activated downstream of receptor activation. This activation is prevented by FLIP, which in turn is inhibited by AS-PTO and quercetin. The mitochondrial activation through Bid cleavage is counteracted by pro-survival BCL2 protein family members such as BCL2 and MCL1, which can be downregulated by HA 14-1 and ABT-737. Further downstream, caspase 9 activation is prevented by survivin, which can be pharmacologically inhibited by cisplatin and SM-164. For additional details, please refer to the text.

TRAIL receptors as regulators of TRAIL resistance

Early on, it was postulated that TRAIL resistance was mediated by the non-functional decoy receptors, DCR1 and DCR2, competing with the functional receptors DR4 and DR5 for TRAIL binding. The expression of DR4 and DR5 is detectable in most tissues but is lower in normal cells than in cancer cells (23). However, studies have failed to find a consistent correlation between the expression of functional receptors, decoy receptors, and TRAIL sensitivity in cancer cell lines and tumors (24–27). In contrast, downregulation of DR4 and/or DR5 by epigenetic changes such as silencing by hypermethylation (28) or by deletions or point mutations (29–31) is frequently associated with TRAIL resistance in cancer, suggesting that a functional receptor is, at least, necessary for activity. Despite the poor correlation between exact levels and sensitivity to TRAIL, upregulation of DR4 and/or DR5 appears to mediate enhanced TRAIL sensitivity by several drugs including the quercetin derivative LY303511(32), HDAC inhibitors such as trichostatin (33) and the proteasome inhibitor bortezomib (23). Importantly, the response to TRAIL-stimulating drugs can not generally be predicted from the expression of DR4 and DR5. Even in cells where both DR4 and DR5 expression are intact, there is often a preference for usage of one receptor over another. For example, in chronic lymphocytic leukemia and mantle cell lymphoma, both DR4 and DR5 are expressed and normal, but TRAIL signaling occurs almost exclusively through DR4 (34), since signaling through DR5 requires crosslinking of the agonistic antibody (35), which can be impaired in immuncompromised patients due to lack of endogenous Fc receptors. Intriguingly, a point mutation in DR5 was found to have a “dominant negative effect” in the sense that the mutated dysfunctional receptor competes with the functional DR4 receptor for binding of the TRAIL ligand (31). In light of these findings, it is clear that soluble TRAIL and agonists to the respective receptors may have different effects in different tumors depending on which of the DR4 and DR5 receptors is active. If, in a given tumor, the importance of one receptor over another can be assessed, it may be possible to enhance TRAIL-induced apoptosis by using receptor agonists that are cross-linked (35) or that have a high selective affinity for one receptor over the other (36). It is, however, unlikely that DR4 and DR5 expression per se is a good enough predictor of TRAIL sensitivity to be used clinically for patient pre-evaluation.

In addition to the receptor levels playing a role in TRAIL sensitivity, their distribution on the membrane can also affect activity. Cholesterol- and sphingolipid enriched microdomains in the plasma membrane, called lipid rafts, facilitate pre-clustering of receptors on the cell membrane, which enhances ligand binding and recruitment of receptor associated signaling molecules. Sensitization to TRAIL via formation of lipid rafts can be accomplished by treatment with quercetin (37) and cox-2 inhibitors (38). TRAIL resistance may also be acquired through inefficient transportation of the receptors to the cell membrane. This process involves glycosylation, since glycosylation inhibitors restore receptor transport and TRAIL sensitivity (39). Glycosylation also influences the clustering of receptors so that highly glycosylated receptors are more prone to dimerize and, consequently, form the signaling complex required for caspase activation. Interestingly, the protein GALNT14, which promotes receptor glycosylation and clustering, is associated with TRAIL sensitivity, as discussed in more detail below (40).

FLIP

The degree by which caspase 8/10 is regulated by FLIP is another determinant of TRAIL sensitivity. FLIP is structurally similar to caspases and exists in two shorter forms, FLIP short (FLIP_S) and FLIP Raji (FLIP_R) and a long form, FLIP long (FLIP_L). These forms of FLIP interfere with TRAIL signaling in distinct ways. FLIP_L has a caspase-like domain that allows it to dimerize with pro-caspase 8 and 10 at the DISC. It can have either a pro-apoptotic function or an anti-apoptotic function depending on the amount of FLIP_L, the degree of receptor stimulation, and the abundance of the shorter FLIP isoforms (41). In addition, FLIP_L plays a role in survival by forming a complex with pro-caspase 8 that inhibits RIPK3-dependent necrosis (42). Notably, the FLIP_L/ caspase 8 heterodimer results in a lower degree of activation which alters the substrate specificity, favoring proliferation and differentiation pathways rather than apoptosis pathways (43). FLIP_S and FLIP_R are truncated forms that lack the caspase-like domain and act by competing with the caspases at the DISC to prevent binding and activation (44, 45). A novel cleavage form of FLIP_L, p22, can augment anti-apoptotic signaling pathways by inducing NFKB activation (46). Overexpression of FLIP has been linked to TRAIL resistance in many cancers and conversely, downregulation of FLIP enhances TRAIL-induced apoptosis (reviewed in (47). Several TRAIL-sensitizing drugs, including withaferin A (48), quercetin (49), agonists of peroxisome proliferator-activated receptor gamma (PPARg) (50) and a novel EGFR-targeted diphteria toxin (51) all act at least in part by downregulating FLIP. Expression of MYC is directly correlated to TRAIL sensitivity and involves transcriptional repression of FLIP (52). In a recent study, an antisense phosphorothioate oligonucleotide (AS PTO) targeting FLIP sensitized several cancer cell lines but not a normal lung cell line to TRAIL-induced apoptosis. This FLIP-targeted AS PTO also efficiently enhanced apoptosis in xenograft models (53). These studies suggest that direct targeting of FLIP may be beneficial as therapy in combination with TRAIL to circumvent and prevent TRAIL resistance. However, because of the complexity of the functions of FLIP, where subtle changes in concentration of the FLIP proteins themselves as well as other DISC components may lead to opposing effects on apoptosis sensitivity, the usefulness of c-FLIP overexpression as a biomarker for TRAIL resistance may be limited.

BCL2 family proteins

The BCL2 family consists of proteins that either promote or inhibit mitochondria dependent apoptosis by influencing the permeability of the mitochondrial membrane and/or by regulating the activity of each other. In TRAIL signaling, pro-apoptotic BID activates BAX and BAK, inducing permeabilization of the mitochondrial membrane and release of cytochrome c. Loss of pro-apoptotic BAX (54) and/or BAK (55) as well as overexpression of anti-apoptotic BCL2 (56,57), BCLXL (58) and MCL1 (59–61) all confer resistance to TRAIL. Pro-survival BCL2 family proteins can be targeted clinically using BH3 mimetics (reviewed in (62)) or small molecule inhibitors such as the BCL2 inhibitor HA14-1 (57). The BH3 mimetic ABT-737 enhances TRAIL killing in multiple cancer cell types, regardless of MCL1 status, but this effect requires expression of BAX (63). ABT-737 induces upregulation of DR5, suggesting that induction of apoptosis by TRAIL is augmented through both the intrinsic and extrinsic pathways by this drug (63). Another BH3 mimetic, the BCLXL inhibitor BH3I-2’, synergizes with TRAIL in prostate carcinoma cells by increasing the amount of BAX/BAK available to be activated by BID (64). Despite the encouraging pre-clinical data, the complexity and redundancy of the BCL2 family proteins makes pre-evaluation of patients based on the expression of these proteins cumbersome, and thus may not be practically feasible.

The inhibitor of apoptosis proteins (IAPs)

The extrinsic and the intrinsic pathway converge at the level of activation of effector caspases such as caspase 3, 7 and 9. The IAPs are a family of proteins that can inhibit apoptosis by directly inhibiting effector caspases, thus influencing apoptosis signaling from both pathways of apoptosis induction. The IAPs are antagonized by the protein Smac/Diablo. The most studied IAP, XIAP, mediates TRAIL resistance when overexpressed (65), providing a rationale for the inhibition of XIAP as a strategy to enhance TRAIL-induced apoptosis. Indeed, TRAIL resistance in prostate cancer cells can be overcome by addition of a zinc chelator which downregulates XIAP (66) and a small-molecule XIAP antagonist synergizes with TRAIL in vitro and in vivo (67). In a study by Xu and colleagues, sensitization to TRAIL by cisplatin involved downregulation of the IAP survivin (68). Similarly, overexpression of SMAC/DIABLO sensitizes prostate cancer cells to TRAIL coinciding with a reduction in XIAP, cIAP1 and cIAP2 (69). In a recent, more clinically relevant approach, the non-peptide SMAC mimetic SM-164 was highly synergistic with TRAIL in breast, prostate and colon cancer cells in vitro, and in breast cancer xenografts in vivo (70). Interestingly, SM-164 did not only act by inhibiting IAPs, but also by enhancing DISC formation and recruitment of caspase 8 in a RIP1-dependent manner.

O-Glycosylation and TRAIL sensitivity

In an attempt to identify determinants of TRAIL sensitivity, Wagner and co-workers investigated TRAIL sensitivity in 119 human cancer cell lines and performed whole genome microarray profiling to detect differences in gene expression between TRAIL-sensitive and TRAIL-resistant cell lines. The most prominent molecule found to correlate with TRAIL response in this study was the O-glycosylation initiating enzyme N-acetylgalctosaminyltransferase-14 (GALNT14). Indeed, presence of the enzyme predicted sensitivity to TRAIL 61% of the time, whereas absence of the enzyme predicted resistance in 88% of all cases. The authors found that GALNT14 enhanced recruitment and activation of caspase 8 by promoting ligand-induced clustering of the DR4 and DR5 receptors, but not of Fas or TNFR1, demonstrating a specific role in sensitization to TRAIL (40). The same group subsequently developed an immunohistochemical assay that measured the levels of GALNT14 and of fucosyltransferase (FUT) 3/6, other enzymes involved in O-glycosylation, in formalin-fixed, paraffin-embedded human tumor tissues and in human cell lines. These assays are currently being evaluated in phase II clinical trials as diagnostic tools to predict sensitivity to the TRAIL receptor agonists dulanermin (recombinant TRAIL) and drozitumab (agonistic anti-DR5 antibody) (71) as discussed below.

Six1

Recently, the homeoprotein Six1 was identified as a novel mediator of TRAIL resistance (72). Six1 is aberrantly expressed in many forms of cancer and contributes to the malignant phenotype by promoting tumor initiation, progression and metastasis (73, 74) (75, 76). Overexpression of Six1 specifically induces TRAIL resistance in TRAIL sensitive ovarian cancer cells, and knockdown of Six1 sensitizes previously resistant ovarian cancer cells to TRAIL (72). In concordance with this finding, expression of microRNA-185 overcomes TRAIL resistance in ovarian cancer cells specifically through down-regulation of Six1 (77). Importantly, high Six1 expression occurs in more than 60% of women with metastatic ovarian cancer, and is strongly associated with worse survival in these patients (72). These data suggest that many ovarian cancers may be TRAIL resistant, and thus refractory to multiple different therapeutic regimens. Indeed, as Six1 correlates with adverse outcomes, including metastasis, in many different tumor types, it is tempting to speculate that this correlation is at least in part due to an acquired resistance to TRAIL, and that high Six1 expression in tumors may be a negative predictor of TRAIL sensitivity. Because Six1 overexpression occurs in a large percentage of ovarian and breast cancers, there is significant potential for screening of this biomarker to aid in patient selection and, ideally, in applying more educated treatment regimes. Thus, we are currently evaluating the feasibility of using Six1 as a biomarker for TRAIL resistance, and are attempting to find ways to define and circumvent this resistance.

Human Clinical Trials to Date

Several proapoptotic receptor agonists (PARAs), including the recombinant human TRAIL ligand (dulanermin), which targets both DR4 and DR5, and agonistic antibodies against the functional receptors DR4 (mapatumumab) and DR5 (lexatumumab, drozitumab, conatumumab, tigatuzumab and LBY-135) have been assessed within clinical trials for cancer treatment. At the time of this review there were a total of 27 human clinical trials published on pro-apoptotic receptor agonist therapies (see table 1 and references therein). Several of these clinical investigational studies have been published and summarized in prior reviews (78, 79). Table 1 provides an updated summary of all published human clinical trials to date of both monotherapy and combination therapy utilized in clinical trials. Some trials have been tumor type focused, while others, particularly the first-in-man studies; have explored the PARAs in general advanced cancer populations. No trials to date have involved any degree of molecular preselection.

Agonist	Site	Drug	Response	Reference
Conatumumab AMG 655 (monoclonal antibody DR5 agonist)	NSCLC	Paclitaxel and carboplatin	1CR, 3PR	(101)
	Sarcoma	Doxorubicin	2PR	(92)
	CRC	mFOLFOX and bevacizumab	6PR	(93)
	Pancreatic	Gemcitabine	2PR	(98)
	CRC	Panitumumab	None	(93)
	CRC	Panitumumab	None	(102)
	Advanced solid tumors	Phase 1: single agent	1 PR	(85)
	Advanced solid tumors	Phase 1: single agent	1 PR	(91)
CS-1008 (humanized monoclonal antibody DR5 agonist)	Advanced solid tumors	Phase 1: single agent	none	(86)
	NSCLC	Randomized phase 2: paclitaxel/carboplatin	—	—
	CRC	Randomized phase 2: irinotecan	—	—
	Ovarian	Single arm phase 2: paclitaxel/carboplatin	—	—
	Pancreatic cancer	Single arm phase 2: gemcitabine	—	—
Dulanermin rhTRAIL (Proapoptotic receptor agonist)	NHL	Rituximab	2CR, 1PR	(95)
	CRC	Phase 1: irinotecan/cetuximab or FOLFIRI± bevacizumab	3 PR	(94)
	ADV TUMORS	None	2PR	(80)
	NSCLC	Paclitaxel and carboplatin ± bevacizumab	1CR, 13PR	(99)
	NSCLC	Ramdomized Phase 2: Paclitaxel and carboplatin ± bevacizumab	—	(99)
Lexatumumab (monoclonal antibody DR5 agonist)	Advanced solid tumors	Gemcitabine, pemetrexed, doxorubicin, or FOLFIRI	PRs reported	(100)
	Advanced solid tumors	Phase 1: single agent	none	(81)
	Advanced solid tumors	Phase 1: single agent	none	(87)
PRO95780 (fully human monoclonal antibody DR5 agonist)	Advanced solid tumors	Phase 1: single agent	none	(82)
	CRC	Phase 1: cetuximab/irinotecan or FOLFIRI/bevacizumab	—	(82)
	CRC	Phase 1b: FOLFOX/bevacizumab	—	(82)
	NSCLC	Randomized phase 2: paclitaxel/carboplatin/bevacizumab	—	(82)
	NHL	Single arm phase 2: rituximab	—	(82)
	NSCLC	Randomized Phase 2: paclitaxel/carboplatin/bevacizumab		(106)
Mapatumumab (monoclonal antibody DR4 agonist)	Advanced solid tumors	Gemcitabine and cisplatin	12PR	(103)
	Advanced solid tumors	Paclitaxel and carboplatin	5PR	(97)
	Advanced solid tumors	None	19 had stable diesease	(83)
	Advanced solid tumors	Phase 1: single agent	none	(84)
		Hepatocellular carcinoma	Phase 1: sorafenib
			Single arm phase 2: sorafenib
	NHL	Phase 2: single agent	1 CR, 2 PR	(90)
	Multiple myeloma	Randomized phase 2: bortezomib	—	—
	CRC	Phase 2: single agent	None	(89)
	NSCLC	Phase 2: single agent	None	(88)
	NSCLC	Randomized phase 2: paclitaxel/carboplatin	—	(105)

Single Agent Trials

There have been several trials (80–90) evaluating efficacy of Conatumumab AMG 655 (monoclonal antibody DR5 agonist), CS-1008 (humanized monoclonal antibody DR5 agonist), Dulanermin (rhApo2L/TRAIL Proapoptotic receptor agonist), Lexatumumab (monoclonal antibody DR5 agonist), PRO95780 (fully human monoclonal antibody DR5 agonist) and Mapatumumab (monoclonal antibody DR4 agonist). These agents clearly have monotherapy activity with isolated responses reported in follicular lymphoma (81) adenocarcinoma of the lung (85, 91) synovial sarcoma (80). While these responses appear rare, in some cases they could be very long lasting with one patient still on study at the time of publication receiving 106 doses over 4 years (91). The most commonly reported toxicity that could be attributed as a class effect, as it occurred across several different agents, was rare transaminitis that was reversible on cessation of dosing (82).

Combined Trials

Several trials have been carried out utilizing combined therapy to evaluate the efficacy of Conatumumab, Dulanermin, Lexatumumab, PRO95780 and Mapatumumab in combination with other anti cancer therapy (80, 82 92–103). Four separate trials assessed the effect of adding a PARA to standard 1^st line carboplatin and paclitaxel (alone or with bevacizumab) for advanced NSCLC (101, 104–106). Toxicity was minimally affected by the addition of their PARAs. However, none of these trials achieved their primary endpoints of improving response rate or progression free survival.

These results are not surprising given the rarity of monotherapy activity. Perhaps this indicates the rarity of a sensitive subpopulation. Moreover, we should not view the agents themselves as inactive but only inactive in unselected patients. Within several of the randomized phase II studies in NSCLC a prolonged tail in the PFS curves, amounting to approximately 15% of the treated population is apparent. While this is not enough to sway the median results (104, 105), it is enough that with patient selection, these drugs could replicate the rare but sustained responses seen in isolated patients in the monotherapy studies.

A large proportion of the drugs used in the combination trials prevent cell division and DNA replication, for instance by preventing DNA from unwinding (topoisomerase inhibitors such as irinotecan), by disruption of microtubule function (taxanes such as paclitaxel), or by inducing apoptosis or cell cycle arrest through several pathways (HDAC inhibitors such as vorinostat). Other drugs used in combination therapy are specific antibodies targeting growth factor signaling pathways relevant to different tumor types (for instance, bevacizumab, which targets VEGFA and cetuximab, which targets EGFR). Importantly, although many preclinical drugs synergize with TRAIL in cells that are already sensitive to TRAIL, a much smaller proportion of drugs can be expected to actually reverse resistance to TRAIL that is already present or that evolves during treatment. If we hypothesize that most cancers are actually TRAIL resistant, the results from these unselected clinical trials become almost predictable. In a recent study performed by our group, synergism with TRAIL was obtained with all 10 drugs tested, whereas only one, the proteasome inhibitor MG132, was able to overcome resistance in cell lines that had been made resistant to TRAIL through long term in vitro selection in increasing concentrations of Lexatumumab, through overexpression of Six1 or through decoy receptor overexpression (107).

The conclusion that we should draw from these data is not that TRAIL holds little promise as a cancer drug, but rather that since resistance to TRAIL is such a common occurrence, sub-optimal patient selection is likely to have had a negative effect on the accurate evaluation of these drugs.

Discussion

It is apparent that despite the extensive research on TRAIL resistance mechanisms, we have somehow failed to incorporate this preclinical data into the selection of patients and tailored treatment regimens in clinical trials. Because of this failure, we are still lacking an accurate picture of how efficient TRAIL treatment can be in the right setting. It is crucial that drug regimens that show no response, when used in studies with suboptimal selection of patients, are not disqualified, in the event that the treatment would be highly beneficial in another patient group. It is highly noteworthy that there is to date not a single clinical trial of TRAIL published that has attempted to preselect patients using specific biomarkers that may predict response to a specific treatment regimen (108). Because of this, the results of TRAIL clinical trials performed to date may greatly underestimate the potential of drugs that activate this pathway. Complicating the design of any such preselection study in the future is the fact that multiple different resistance mechanisms may occur, potentially influencing the optimal biomarkers to use depending on the specific receptor agonist and partner drug involed. For example, in a recent preclinical study evaluating the mechanisms of cisplatin and TRAIL in combination, cisplatin was shown to enhance TRAIL-induced apoptosis through survivin downregulation (68). Thus, the ideal future scenario might be that in a clinical trial involving cisplatin treatment, we should carefully select patients in which the prevailing TRAIL resistance mechanism exhibited in tumors is an overexpression of survivin. As we learn more about what governs the susceptibility to DR4 stimulation vs DR5 stimulation, it may even be possible to preselect an optimal TRAIL-targeting agent for some TRAIL sensitive patients.

Although the theoretical side of a biological approach to patient selection is very attractive, the practical issue of selecting and assaying appropriate biomarkers is a challenge. Optimally, we would like to be able to select patients that are already sensitive to TRAIL and/or to define TRAIL resistance markers that are already present or that may arise as a result of treatment. One exciting approach to this is a GALNT14/FUT3/6 assay which is currently being evaluated in phase II clinical trials as a diagnostic tool to predict sensitivity to the TRAIL receptor agonists, dulanermin and drozitumab (71).

However, if there is one thing we have learned from clinical trials it is that in patients, TRAIL sensitivity is the exception to the rule. In addition, if sensitivity can be predicted 61% of the time from expression of GALNT14 (71), it is likely that an additional biomarker would be necessary to increase the success rate. Given the large number of patients that express Six1, this gene may be a critical means by which a large percentage of tumors acquire TRAIL resistance, and may thus be a potent biomarker as a negative predictor of responsiveness to TRAIL therapy.

It must be emphasized that most resistance mechanisms have been discovered in cell lines and that the practicality of screening patients for the same resistance mechanisms and, even more so, inhibiting these mechanisms is not always a plausible route to take. Indeed, in a recent study by Menke et al, selective inhibition of TRAIL receptors reduced chemosensitivity in vivo but not in vitro (109). In addition, because of the complexity and the redundancy of the apoptosis machinery, too focused an approach may be as ineffective as no approach. One additional approach is to evaluate apoptosis susceptibility further downstream in the machinery, taking into account not only known players of the intrinsic as well as the extrinsic pathway, but also as yet unidentified resistance mechanisms. The principle behind this is that if a tumor cell is more primed for death, regardless of the mechanisms involved, it is more likely to undergo apoptosis in response to TRAIL. In a study by Zhang et al, the basal apoptotic rate (BAR) of tumors was measured by the cleavage of the apoptosis substrate keratin 18 before and after treatment with TRAIL. The results were encouraging, showing that an initial high BAR was more likely to lead to a total BAR increase after treatment (110).

Perhaps some combination of approaches will be required. For example, an initial estimation of the likelihood of a positive response to TRAIL therapy could involve a screening for positive markers of sensitivity such as high GALNT14 expression, the presence of at least one of the functional TRAIL receptors, and low or absent Six1 expression. This could be combined with a more general assessment of the tumor cells’ propensity to undergo apoptosis by using the BAR assay. When deciding on a combination treatment, we would ideally like to see a connection between known modes of action of drugs and patient expression profiles, i.e. if a drug enhances expression of a specific pro-apoptotic protein it may be more fruitful to use it in patients where the expression of that particular pro-apoptotic protein is initially low

In conclusion, we have greatly expanded our knowledge about TRAIL signaling and the factors that regulate it. At the same time, novel drugs that have the potential to specifically target these regulators have been developed. With these advancements in the field and with an understanding of the importance of individualized patient selection, we have an obligation to take the evaluation of TRAIL based treatments to the next level, where this cancer drug can finally receive a fair trial.