The tumor necrosis factor-related apoptosis-inducing ligand (TRAIL or Apo2L) is a potent inducer of death of cancer but not normal cells,which suggests its potential use as a tumor-specific antineoplastic agent. TRAIL binds to the proapoptotic death receptors DR4 and the p53-regulated proapoptotic KILLER/DR5 as well as to the decoy receptors TRID and TRUNDD. In the present studies, we identified a subgroup of TRAIL-resistant cancer cell lines characterized by low or absent basal DR4 or high expression of the caspase activation inhibitor FLIP. Four of five TRAIL-sensitive cell lines expressed high levels of DR4 mRNA and protein, whereas six of six TRAIL-resistant cell lines expressed low or undetectable levels of DR4 (χ2; P < 0.01). FLIP expression appeared elevated in five of six (83%) TRAIL-resistant cell lines and only one of five(20%) TRAIL-sensitive cells (χ2; P < 0.05). Two TRAIL-resistant lines that expressed DR4 contained an A-to-G alteration in the death domain encoding arginine instead of lysine at codon 441. The K441R polymorphism is present in 20% of the normal population and can inhibit DR4-mediated cell killing in a dominant-negative fashion. The expression level of KILLER/DR5, TRID,TRUNDD or TRID, and TRUNDD did not correlate with TRAIL sensitivity(P > 0.05). These results suggest that the major determinants for TRAIL sensitivity may be the expression level of DR4 and FLIP. TRAIL-resistant cells became susceptible to TRAIL-mediated apoptosis in the presence of doxorubicin. In TRAIL-sensitive cells,caspases 8, 9, and 3 were activated after TRAIL treatment, but in TRAIL-resistant cells, they were activated only by the combination of TRAIL and doxorubicin. Our results suggest: (a)evaluation of tumor DR4 and FLIP expression and host DR4 codon 441 status could be potentially useful predictors of TRAIL sensitivity, and(b) doxorubicin, in combination with TRAIL, may effectively promote caspase activation in TRAIL-resistant tumors.
TRAIL3, a member of the TNF cytokine family and a type II membrane protein, was initially identified by homology to the C-terminal extracellular domain of other TNF family members, such as Fas ligand(FasL), TNF-α, and lymphotoxin α (1). TRAIL is a potent inducer of apoptosis in a variety of transformed or cancer cells of human and mouse origin but not normal cells (1, 2).
The therapeutic use of the Fas/FasL or the TNF-α/TNFR1 system in cancer treatment has been hampered by severe side effects (3). The systemic administration of TNF causes a septic shock-like response possibly mediated by nuclear factor-κB activation, and the injection of agonist Ab to Fas can be lethal (3, 4). Compared to TNF-α or Fas, TRAIL may be a safer alternative because normal cells appear to be resistant, and it activates nuclear factor-κB only weakly (5). Recently, evidence for the safety and potential efficacy of TRAIL therapy against breast and colon cancer was obtained in a severe combined immunodeficiency mouse model (6, 7). Additionally, in cell culture, the human leucine zipper (LZ)-TRAIL had no cytotoxic effects on normal cells, including human mammary epithelial cells, human renal proximal tubule epithelial cells, human lung fibroblasts, and human skeletal muscle cells but was toxic toward mammary adenocarcinoma cells (6). The in vivoexperiments showed that the systemic administration of LZ-TRAIL into mice inoculated with breast cancer cells prolonged survival. These studies suggest that TRAIL may have a potential use for cancer treatment.
TRAIL can modulate an apoptotic response by binding to one of four cell-surface receptors: Death receptor (DR) 4 (TRAIL-R1; Ref. 8), KILLER/DR5 (TRAIL-R2, TRICK2; Refs. 9, 10, 11, 12), TRID (DcR1, TRAIL-R3, or LIT; Refs. 5, 10, 13, and 14),and TRUNDD (DcR2 or TRAIL-R4; Refs. 15, 16, 17). DR4 and KILLER/DR5 have two cysteine-rich extracellular ligand-binding domains and a cytoplasmic death domain that signals downstream caspase activation (2, 18). KILLER/DR5 was identified as a candidate p53 target gene, linking DNA damage signaling from p53 to downstream caspase activation and cell death (9). The extracellular domain of TRID shares a homology with DR4 and KILLER/DR5,but it does not have a cytoplasmic death domain, and it is anchored to the membrane through a glycosyl phosphatidyl inositol linkage. TRUNDD has a substantially truncated cytoplasmic death domain. These two decoy receptors have been reported to protect cells from TRAIL-mediated apoptosis by competing with DR4 and KILLER/DR5 for binding to TRAIL (10).
The TRAIL-mediated biochemical signaling pathway leading to apoptosis is not yet clear. Previously, it was reported that the ectopic expression of FADD-DN (dominant-negative FADD, which blocks apoptotic signaling by the Fas/APO1 death receptor) does not efficiently block apoptosis triggered by TRAIL, and that overexpression of DR4 could induce apoptosis in FADD-deficient embryonic fibroblasts (19). These studies suggest that a FADD-independent pathway may link TRAIL to the caspase cascade (2, 19, 20). Moreover, it was shown that DR4 does not efficiently recruit FADD, TNF receptor-associated death domain (TRADD) protein, receptor interacting protein (RIP), or RIP-associated ICH-1/CED-3 homologous protein (RAIDD;Ref. 10). Although at present there is a missing link between TRAIL death receptors and caspase activation, it is clear that once TRAIL binds to its receptors, apoptosis ensues through the activation of caspases (5, 8, 10). Initiator caspases(caspases 8, 9, and 10) are composed of an N-terminal prodomain that contains the region for homotypic protein-protein interaction with adaptor molecules together with one large and one small subunit. When cells receive death-inducing signals, the prodomain is cleaved, and an active heterodimeric tetramer containing two small and two large subunits is formed. It was reported that caspases 3 and 8 became activated when HeLa cells were treated with TRAIL (21) and also that in TRAIL-sensitive breast cancer cell lines, caspase 3 cleavage was observed (22). In addition, a recent report that T lymphocytes that have catalytically inactive caspase 10 are TRAIL-resistant implicates caspase 10 in TRAIL-mediated apoptosis (23).
Although the efficacy and potential use of TRAIL in cancer treatment has been suggested, little is known about the factors that determine the sensitivity of cancer cells to killing by TRAIL. Recently, there were some reports on the determinants of TRAIL sensitivity in breast cancer cells (22), melanoma (24), and brain tumors (25, 26). The results have been somewhat controversial in that some reports showed no correlation between TRAIL sensitivity and the expression level of proapoptotic death receptors,whereas others demonstrated a correlation between them.
We investigated the expression level of various TRAIL receptor family members as determinants for TRAIL sensitivity and whether a DNA-damaging chemotherapeutic drug such as doxorubicin might have additive effects with TRAIL in killing cancer cells. We report here that the expression of the proapo-ptotic TRAIL receptors, in particular DR4, and the caspase activation inhibitor FLIP may be major determinants of TRAIL sensitivity. In addition to the expression level of DR4, a polymorphism found in the death domain region of DR4 prevents DR4-mediated cell killing in a dominant-negative fashion. Finally, we also report that a DNA damaging agent such as doxorubicin can sensitize cells to TRAIL-mediated cell killing. Our results provide essential preclinical information that may be useful in the design of clinical trials using recombinant TRAIL in the therapy of human cancer.
Materials and Methods
Human lung fibroblast WI38 and human foreskin fibroblast HS27 cells were obtained from the American Type Culture Collection (Rockville,MD). The human ovarian cancer cell line SKOV3, the human breast cancer cell line SKBr3, and the human nasopharyngeal squamous cancer cell line FADU were also obtained from the American Type Culture Collection. The human lung cancer cell lines H460 Neo/E6, the human colon cancer cell lines HCT116 Neo/E6, the human ovarian cancer cell lines PA1 Neo/E6,and the human colon cancer cell line SW480 were maintained as described previously (27). The J82 human bladder cancer cell line was a gift from T. McGarvey and B. Malkowicz (University of Pennsylvania, Philadelphia, PA), and the A875 human melanoma cell line was a gift from D. George (University of Pennsylvania, Philadelphia,PA).
Assessment of Cell Viability.
Recombinant soluble human TRAIL was purchased from Kamiya Biomedical Co. (Seattle, WA), and the anti-FLAG M2 mAb was purchased from Sigma(Saint Louis, MI). Three thousand cells were seeded into each well of a 96-well plate. After 24 h, the cells were treated with TRAIL (200 ng/ml) and cross-linked with the anti-FLAG M2 mAb (2 μg/ml). Cell viability was measured by using the MTT assay at 16 h after treatment (28). When normal cells were treated with both doxorubicin and TRAIL, the cells were treated with increasing concentrations of chemotherapeutic drugs alone (doxorubicin, 0, 0.1, 1,10, and 100 μg/ml) or in combination with TRAIL (20 ng/ml)cross-linked with the anti-FLAG M2 Ab (2 μg/ml). To assess the long-term effect of TRAIL, a total of 5 ×104 of each cell line were seeded in triplicate into 24 wells, and at 24 h, cells were treated with TRAIL (50 ng/ml) and the anti-FLAG M2 Ab (2 μg/ml). The media containing TRAIL and Ab was changed every 48 h, and the culture was maintained for 7 days, at which time the remaining cells were stained with Coomassie Blue.
Total RNA was isolated from cell lines as described (29). cDNA was generated from 2 μg of total RNA in a final volume of 20μl using SuperScript II (Life Technologies, Inc., Gaithersburg, MD)and random primers. The sequences of specific primers used in this experiment were as follows: DR4 F, 5′-CGATGTGGTCAGAGCTGGTACAGC-3′; DR4 R, 5′-GGACACGGCAGAGCCTGTGC-CATC-3′; KILLER/DR5 F,5′-GGGAGCCGCTCATGAGGAAGTTG G-3′, KILLER/DR5 R,5′-GGCAAGTCTCTCTC-CCAGCGTCTC-3′; TRID F,5′-GTTTGTTTGAAAGACTT-CACTGTG-3′, TRID R,5′-GCAGGCGTTTCTGTCTGT-GGGAAC-3′; TRUNDD F,5′-CTTCAGGAAACCAGAGCTT-CCCTC-3′, TRUNDD R,5′-TTCTCCCGTTTGCTTATCA-CACGA-3′; GAPDH F,5′-ACCACAGTCCATGCCATCAC-3′, GAPDH R, 5′-TCCACCACCCTGTTGCTGTA-3′.
To analyze the expression level of the death receptors, 2 μl (out of 20 μl) of synthesized cDNA was amplified in a total volume of 50 μl containing 200 μm each of all four dNTPs, 2 μCiα-32P-dCTP (3000 Ci/mmol), 2 μmeach of death receptor-specific primer set along with 2μ m each of the GAPDH primers, and 1 unit of Taq DNA polymerase (Perkin-Elmer). The cycle numbers that showed linear growth of product were initially determined for each PCR product by analyzing a 10-μl sample from multiple identical amplification reactions (Fig. 2 A and data not shown). In the case of DR4 and KILLER/DR5, 23 cycles were chosen; for TRID and TRUNDD,24 cycles were chosen; and in the case of GAPDH, 18 cycles were chosen. During PCR, 10 μl of the reaction were remove at the indicated cycle numbers. PCR conditions were as follows: 1 cycle, 5 min/95°C; 23 or 24 cycles, 30 s/95°C, 30 s/55°C (for DR4, KILLER/DR5, and TRUNDD),52°C (for TRID), or 30 s/72°C. Nondenaturing PAGE (7%) was performed, and the gel was fixed, dried, and autoradiographed. Band intensities were quantitated by using a Phosphorimager Storm 840(Molecular Dynamics, Sunnyvale, CA).
Genomic DNA Isolation and Cycle Sequencing.
Whole blood (20 ml) from 10 normal healthy volunteers was drawn, and genomic DNA was isolated using the Blood and Cell culture DNA maxi kit(QIAGEN Inc., Valencia, CA). The DNA (50 ng) was used as a template for the amplification of the DR4 death domain region spanning nucleotide 1322. Sequences of primers used in PCR are as follows: DR4 11,5′-CTCTGATGCTGTTCTTTGAC-3′, DR4 12, 5′-TCACTCCAAGGACACG-GCAGA-3′. After amplification, each PCR product was visualized and purified from an agarose gel using the QIAquick gel extraction kit (QIAGEN Inc.) and was then used as a DNA sequencing template. Cycle sequencing was performed using a SequiTherm cycle sequencing kit (Epicentre Technologies, Madison, WI) according to the manufacturer’s instructions.
Site-directed Mutagenesis and Sequencing.
Site-directed mutagenesis was performed using a Quick change site-directed mutagenesis kit (Stratagene, La Jolla, CA) according to the manufacturer’s instructions. To change a base in the death domain region of DR4 (A to G at nucleotide 1322 of DR4), plasmids that contained either the full-length DR4 (f/DR4 (A) in pCEP4, Invitrogen,Carlsbad, CA) or the cytoplasmic domain of DR4 (CD/DR4 (A) in pcDNA3.1-Myc, His; Invitrogen, Carlsbad, CA) were used as templates. The sequences of the primer pairs used for changing the base were as follows: DR4DDMUT F,5′-GGAAGAGAGACATGCAAGAGAGAAGATTCAGGA-CC-3′; DR4DD MUT R,5′-GGTCCTGAATCTTCTCTCTTGCATGTCTCTCTTCC-3′. The sequences of the mutagenized plasmids were confirmed. Sequencing of expression plasmids was performed using a T7 DNA sequencing kit (United States Biochemicals, Cleveland, OH) according to the manufacturer’s instructions.
The mutagenized f/DR4 or CD/DR4 was used for transfection into SW480 colon cancer cells as previously described (30). After 24 h of transfection, cell lysates were prepared from each transfectant followed by Western immunostaining for confirmation of expression after mutagenesis.
Evaluation of Cell Death Induced by Transfected DR4.
For cell death evaluation, cotransfection of the CMV-β-gal marker gene and the DR4 mutant constructs generated was performed as previously described (31). Briefly, 1 ×105 of SW480 cells were plated per well in 24-well plates and transfected with 2 μg of the corresponding parental vectors, f/DR4 (A), CD/DR4 (A), f/DR4 (G), or CD/DR4 (G), with CMV-β-gal at 10% of the total amount of DNA. At 24 or 48 h later, cells were fixed and stained with 5-bromo-4-chloro-3-indolyl-β-galactopyranoside to quantify the number of blue cells. To determine whether polymorphic DR4 has a dominant-negative effect on cell killing, SW480 cells were transfected with variable ratios of CD/DR4 (A) to CD/DR4 (G), f/DR4 (A) to CD/DR4(G), or f/DR4 (A) to f/DR4 (G) (4:1, 1:1, and 1:4) along with CMV-β-gal.
Abs and Western Blot Analysis.
Western blot analysis was carried out as previously described (32). Blotted membranes were immunostained with anti-PARP(1:2000; Boehringer Mannheim, Mannheim, Germany), anti-caspase 3(E-8, 1:500; Santa Cruz Biotechnologies, Inc., Santa Cruz, CA),anti-caspase 7 (1:500; PharMingen, San Diego, CA), anti-caspase 8(C-20, 1:500; Santa Cruz Biotechnologies, Inc.), anti-caspase 9(1:500; IIMGENEX, San Diego, CA), anti-caspase 10 (N-19, 1:500; Santa Cruz Biotechnologies, Inc.), anti-caspase 2 (H-19, 1:500; Santa Cruz Biotechnologies, Inc.), anti-DR4 (1:500, PharMingen), anti-DR5 (1:500;IMGENEX,) anti-FLIP (1:500; IMGENEX), anti-Myc (9E10, 1:500; Santa Cruz Biotechnologies, Inc.), or antiactin (I-19, 1:200; Santa Cruz Biotechnologies, Inc.).
The statistical correlation between the expression level of TRAIL death receptors and TRAIL-mediated apoptosis was performed using regression analysis and the correlation between the expression of FLIP and TRAIL sensitivity, or the expression of DR4 and TRAIL sensitivity was performed using the χ2 test.
Normal Cells as Well as a Newly-defined Subset of Cancer Cells Are Resistant to TRAIL-mediated Apoptosis.
We evaluated the cell killing effect of TRAIL on various normal and cancer cell lines. As previously reported by others (1, 3), normal cells (fibroblasts) were resistant to TRAIL treatment(Fig. 1, A and B). In contrast, cancer cells showed a variable response to TRAIL (Fig. 1). HCT116, H460, PA1, SKBr3, and SW480 were sensitive to TRAIL. TRAIL sensitivity was defined as a <75% cell viability at 16 h after TRAIL treatment is measured by the TRAIL MTT assay. A875, FADU, J82,and SKOV3 cells were found to be resistant to TRAIL. Human Papillomavirus E6-expressing HCT116, H460, and PA1 cells were relatively more resistant to TRAIL than the neocounterparts (Fig. 1,A). Long-term (7 days) TRAIL treatment of cell lines (Fig. 1 B) showed nearly the same result as the short-term (16 h)MTT assay results. Based on the observations from the long-term TRAIL treatment assay, certain fractions of cells showed resistance to TRAIL,although the majority of the cells were killed by TRAIL treatment.
Taken together, those results suggest that there is a subgroup of TRAIL-resistant cancer cells and that to a degree, wild-type p53 may modulate TRAIL responsiveness. We further explored the molecular basis of TRAIL resistance in cancer cells.
Correlation between TRAIL Receptor Expression and TRAIL Sensitivity.
To determine whether there is any correlation between TRAIL sensitivity and the expression level of TRAIL receptors, a semiquantitative RT-PCR assay was performed (Fig. 2). The number of PCR cycles required for linear amplification and detection was initially determined for each death receptor (Fig. 2,A). KILLER/DR5 was expressed in all cell lines tested (Fig. 2, B and C), and its mRNA expression level did not correlate with TRAIL sensitivity (Fig. 3,B). In contrast, the expression level of DR4 varied among different cell lines (Fig. 2,B). For example, in normal fibroblast cells,DR4 expression was very low or not detectable (Fig. 2,B, Lanes 1 and 2). Cancer cell lines except J82 and SKOV3 that expressed DR4 were sensitive to TRAIL regardless of p53 status (Fig. 1, Fig. 2,B, and Fig. 3,A; see below). PA1, A875, and FADU cells did not express detectable DR4 protein (Fig. 2,B, Lanes 5, 6, and 9). DR4 protein was highly expressed in HCT116, H460, and SW480 cells (DR4 in Fig. 4, Lanes 3, 4, and 7), and they were the most sensitive cell lines to TRAIL(Fig. 1, A and B). The antiapoptotic TRAIL receptors, TRID and TRUNDD, were also expressed in cancer cells. TRID was expressed in all of the cell lines except PA1 cells, whereas TRUNDD was not expressed in H460, A875, SKBr3, and FADU cell lines (Fig. 2,B, Lanes 3, 6, 7, and 9). The high expression of TRID or TRUNDD in the normal cell lines HS27 or WI38 is consistent with previous results implicating high decoy receptor expression as a mechanism of TRAIL resistance. However, neither TRID nor TRUNDD levels adequately explain the observed patterns of TRAIL sensitivity in the panel of cancer cells (Fig. 3, D-F). The presence of DR4 alone (r = 0.769; P =0.006) or DR4 and KILLER/DR5 (r = 0.786, P = 0.004) appeared to correlate better with TRAIL sensitivity of cancer cells than the expression of decoy receptors(Fig. 3, A and C).
FLIP Expression Correlates Well with TRAIL Resistance.
Cellular FLIP is an inhibitor of caspase activation and may be overexpressed in human cancer cells (33). We determined whether the expression level of FLIP might correlate with TRAIL sensitivity. We detected FLIP expression in five of six TRAIL-resistant cell lines including normal cells A875, J82, and SKOV3 (FLIP in Fig. 4, Lanes 1, 2, 8, 10, and 11) but only in one (PA1)of five TRAIL-sensitive cell lines (FLIP in Fig. 4, Lane 5). These results suggest that high expression of FLIP may be the another important determinant of TRAIL resistance (χ2; P < 0.05).
K441R Polymorphism Found in the Death Domain of DR4.
Contrary to our expectation that DR4-expressing cells should be sensitive to TRAIL, J82 and SKOV3 were resistant to TRAIL treatment. Previously, there was a report indicating that Fas carrying a mutation in the death domain region could act as a dominant-negative inhibitor of Fas-induced cell killing (25). To investigate whether there is a DNA sequence change in the death domain of DR4 in J82 and SKOV3 cells, RT-PCR and DNA sequencing was performed. Sequencing results showed that there is an A-to-G alteration in nucleotide 1322 of DR4 both in SKOV3 and J82 cells (Fig. 5,A and data not shown). This A-to-G transition resulted in the conversion of the amino acid lysine(codon 441) to arginine. To determine whether this alteration is present in normal populations, genomic DNA was isolated from total blood drawn from 10 normal healthy volunteers, and PCR cycle sequencing was performed. The results revealed that 2 (donor 1 and 10) of 10(20%) normal individuals have the base change (Fig. 5,B),and thus, we refer to the alteration as a polymorphism. The polymorphism was found in donors 1 and 10, and SKOV3 was heterozygous in all cases Fig. 5..
Effect of the K441R Polymorphism in the Death Domain of DR4 on DR4-mediated Cell Killing.
To determine whether the K441R polymorphism has any effect on DR4-mediated cell killing, we generated DR4 mammalian expression constructs containing the polymorphism by using site-directed mutagenesis (Fig. 6, A and B). Upon transfection, we found that polymorphic DR4 was less effective in cell killing than its wild-type counterpart (Fig. 6, C and D). In addition, polymorphic DR4 showed an inhibitory effect toward cell killing by wild-type DR4. A potent dominant-negative effect of the K441R polymorphism was observed when the cytoplasmic DR4 (CD/DR4) was expressed. The CD/DR4 (G) rather than f/DR4 (G) showed a nearly complete inhibition of DR4-mediated cell killing (Fig. 6 D).
These results suggest, at least in terms of TRAIL sensitivity, that the K441R polymorphism in the death domain of DR4 makes cells relatively resistant to TRAIL treatment, although they express DR4 on their cell surface. Thus, this polymorphism found in J82 and SKOV3 could contribute to TRAIL resistance.
Cell Killing by Combination of Doxorubicin and TRAIL in TRAIL-resistant Cell Lines.
Normal cells such as HS27 and WI38 are resistant to TRAIL in part due to a low or undetectable expression of DR4, a high expression level of decoy receptors, and a high expression level of FLIP (Fig. 2 and Fig. 4). However, when these cells were treated with the combination of doxorubicin and TRAIL, viability was dramatically reduced (Fig. 7,A) and PARP cleavage became evident (Fig. 7,B). Western immunostaining (Fig. 7,C) showed that there was a significant induction of KILLER/DR5 protein expression. This induction of KILLER/DR5 by doxorubicin may sensitize normal cells to TRAIL-mediated cell killing. These results suggest that an increase in the ratio of expression between proapoptotic and antiapoptotic molecules may reset the responsiveness of the cells from resistant to sensitive. There was no change in the level of DR4 or FLIP expression after doxorubicin treatment (Fig. 7 C).
p53 function was compromised in all of the TRAIL-resistant cancer cell lines tested in this study either by mutation (J82, FADU, and SKOV3) or by the overexpression of MDM2 (A875; Ref. 32). Thus, an exposure to a DNA damaging agent such as doxorubicin might not be expected to result in the p53-dependent KILLER/DR5 induction observed in the normal cells. Nevertheless, when those cells were treated with both doxorubicin and TRAIL, PARP cleavage became evident (PARP in Fig. 8 C,Lanes 4, 8, 12, and 16).
Because there were no changes in the expression level of DR4, DR5, or FLIP after doxorubicin treatment in TRAIL-resistant cancer cell lines(data not shown), we investigated the effect of TRAIL or doxorubicin on the activation of caspases. In terms of doxorubicin sensitivity,TRAIL-resistant cancer cell lines can be divided into doxorubicin-sensitive (FADU) and doxorubicin-resistant (A875, J82, and SKOV3) cells (Fig. 8 C and morphological data not shown).
In doxorubicin-sensitive FADU cells, caspase 8 was activated by doxorubicin treatment alone (caspase 8 in Fig. 8,C, Lane 7). Caspase 9 was also activated by doxorubicin treatment alone in FADU cells (caspase 9 in , Lane 7). Unexpectedly, however,although there was activation of caspases 8 and 9 (“initiator”caspases) in doxorubicin-treated FADU cells, we did not observe complete procaspase 3 (“executioner” caspase) depletion (caspase 3 in , Lane 7). In the doxorubicin-resistant cell lines(A875, J832, and SKOV3), caspase activation was not observed after exposure to either doxorubicin alone or TRAIL alone (Fig. 8,C). Interestingly, caspases 8, 9, and 3 became activated after exposure to the combination of doxorubicin and TRAIL (caspases 8,9, and 3 in Fig. 8,C, Lanes 4, 12, and 16). In contrast to TRAIL-resistant cancer cells, cleavage of caspases 8, 9, and 3 was observed after TRAIL treatment of the TRAIL-sensitive HCT116 colon cancer cell line (Fig. 8,A). When HCT116 was treated with TRAIL, PARP cleavage was evident by 4 h after TRAIL addition, and caspases 8, 9, 3, and 7 became activated at approximately the same time point (4 h after the TRAIL addition; Fig. 8 B).
The cytokine TRAIL is a promising agent for cancer therapy and is presently under investigation (6, 7). The importance of TRAIL as a potential anticancer agent is that it appears to be a potent cancer-specific cytotoxic drug and is not as toxic as other cytokines. TNF-α or Fas have not been successful in clinical trials when administered systemically because of toxicity (3, 4).
Our results provide novel basic information relevant to TRAIL therapy of cancer in the following respects. First, we report that TRAIL resistance is mainly determined by the expression of its proapoptotic death receptors, especially DR4 (r = 0.769, P = 0.006). In fact, cell lines that were resistant to TRAIL were found to have a relatively low or undetectable expression level of DR4. Normal cell lines, such as HS27 and WI38, which are resistant to TRAIL, have extremely low expression of DR4 mRNA or protein (Fig. 2,B, Fig. 3,A, and Fig. 4), and a subgroup of TRAIL-resistant cells also have low or undetectable DR4 expression (Fig. 2,B and Fig. 4). For DR4 expression alone, aχ 2 analysis revealed that this parameter is a highly significant predictor of TRAIL sensitivity when expression is high versus low or undetectable (P < 0.01). For the χ2 analysis, high expression was defined as DR4/GAPDH > 50 as shown in Fig. 2,C. It is important to note that mRNA levels do not always correlate with protein levels and that the strength of the correlation between DR4 expression and TRAIL sensitivity (Fig. 2 and Fig. 3) might be stronger or weaker if the measured DR4 protein levels (Fig. 4) were actually quantitated. The expression of KILLER/DR5, however, does not correlate well with TRAIL sensitivity (Fig. 2 and Fig. 3,B). Our observation is supported by a recent report that TRAIL sensitivity in melanoma cells correlates well with the expression level of DR4 (24). Contrary to our observation, J82 and SKOV3 expressed DR4 (Fig. 2,B and Fig. 4) but were resistant to TRAIL treatment. A previous report that mutation in the death domain region of Fas can act as in a dominant-negative fashion in cell killing (25)prompted us to examine the death domain region of DR4 in J82 and SKOV3 cells. Indeed, J82 and SKOV3 have an A-to-G alteration at codon 441 in the death domain region of DR4 (Fig. 5,A). However, that change is also found in 20% (2 of 10) of a normal population and thus,we refer to the DR4 K441R alteration as a polymorphism. Polymorphic DR4 acted in a dominant-negative manner in DR4-mediated cell killing (Fig. 6, C and D). We make no claim about any disease susceptibility associated with the K441R polymorphism in the DR4 gene. However, the presence of the K441R DR4 polymorphism in cancers may reduce their sensitivity to TRAIL, at least in vitro.
It is important to note the differences observed when full-length versus cytoplasmic domain expression constructs were used to express DR4. In particular, Fig. 6, C and Ddemonstrates that the cytoplasmic domain of DR4 does not itself induce cell death when it contains 441R. In addition, this variant of the cytoplasmic is capable of completely inhibiting death induced by the 441K allele. However, full-length DR4 containing the K441R mutation does not share these properties. Instead, full-length DR4 containing the 441R allele induces apoptosis in ∼50% of transfected cells and poorly inhibits killing by the full-length 441K allele (Fig. 6,D, right). These results suggest that the polymorphic 441R allele may contribute but cannot alone explain the observed resistance to TRAIL in certain cancer cell lines (J82 and SKOV3). These cell lines express somewhat increased levels of FLIP (Fig. 4), which may also contribute to their resistance to TRAIL (see below).
Second, the inhibitor of caspase activation FLIP may confer resistance to TRAIL at a point downstream of the death receptors. We found that 83% (five of six cell lines) of TRAIL-resistant cell lines showed a detectable expression of FLIP, whereas only one of five (20%)TRAIL-sensitive lines expressed FLIP (Fig. 4;χ 2; P < 0.05). However, the fact that FLIP-expressing PA1 cells are sensitive to TRAIL suggests that even in the presence of FLIP, cells can be killed if there is enough of an input signal for inducing apoptosis.
We measured the expression level of five genes (DR4, KILLER/DR5, TRID,TRUNDD, and FLIP) and tested for correlations with TRAIL sensitivity. The expression of two of the parameters (DR4 and FLIP) appeared to independently correlate with TRAIL sensitivity. From the regression analysis shown in Fig. 3, the P value for the DR4 correlation with TRAIL sensitivity is 0.006 (see legend of Fig. 3). Thus, we would have had to test 167 variables to reach the 0.006 level of significance at random for DR4 due to the effect of multiple testing. Moreover, the design of our study was hypothesis driven, with a biological basis giving a reasonable pretest probability of certain correlations. For example, we tested biologically plausible determinants of TRAIL sensitivity. One of the concerns with multiple correlations arises when one tests a very large number of variables (without a hypothesis), such as in a questionnaire with several hundred questions or perhaps a query of an expression of several thousand genes on a DNA microarray chip, and then develops the hypothesis based on any observed correlations at the P < 0.05 level. Of course, if one tests enough variables, there is a random chance that a few will appear to be significant but will actually be meaningless. Thus, because we believed that correcting for multiple testing artifacts would not significantly alter our Ps or conclusions, we have not corrected our calculations for the effects of multiple comparisons. Thus, there is a small chance that our analysis may be limited by the effects of multiple comparisons, and it remains to be seen if others will find a similar significance of DR4 and FLIP expression levels using larger sample sizes and testing fewer variables.
Third, the targeted destruction of p53 to generate otherwise isogenic cancer cell lines revealed that TRAIL sensitivity could be modulated somewhat by p53 (Fig. 1). This is a preliminary observation that requires further investigation. It is clear from our data that wild-type p53 is not required for the apoptotic response to TRAIL.
Fourth, the combination of doxorubicin and TRAIL can kill TRAIL-resistant cancer cells, although each treatment alone cannot effectively kill the cells. The mechanism(s) of this additive killing is not clear yet. We have ruled out changes in the expression level of death receptors or FLIP as a basis for enhanced cell killing by doxorubicin plus TRAIL (data not shown). The fact that FADU cells show caspase 8 and 9 activation upon doxorubicin treatment suggests that the caspase activation axis from caspase 8 through Bcl2 inhibitory protein(Bid) to caspase 9 might be intact in FADU cells but not in other TRAIL-resistant cell lines (Fig. 8,C). As recently reported (22) and observed in our experiments, doxorubicin and TRAIL could activate caspases in augmenting the killing effect. However, although TRAIL resistance can be overcome by combined treatment with doxorubicin, careful consideration should be given to the dose of doxorubicin given the observed sensitization of normal cells to TRAIL-mediated apoptosis (Fig. 7).
Fifth, among TRAIL-sensitive cancer cells, a certain fraction appears to be resistant to TRAIL-mediated killing (Fig. 1 B). A recent report also showed that subclones of TRAIL-sensitive cancer cells display a variable response to TRAIL, although the expression level of TRAIL death receptors or FLIP was not changed (24). We do not know the underlying mechanism of this TRAIL resistance yet.
Our findings suggest that although TRAIL may be useful as a therapeutic agent in cancer, particular attention to molecular determinants of sensitivity needs to be considered to optimize such therapy. TRAIL does not appear to have harmful effects toward normal cells and can kill cancer cells irrespective of p53 status if wild-type DR4 is expressed on their cell surface. Our results also indicate that doxorubicin can sensitize cells to TRAIL-mediated cell killing in vitro,thereby raising hopes that such a strategy may be useful in cancer therapy.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Supported in part by NIH Grants CA75138-01 and CA75454-01.
The abbreviations used are: TRAIL, tumor necrosis factor-related apoptosis-inducing ligand; Ab, antibody; TNF,tumor necrosis factor; MTT,3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide; GAPDH,glyceraldehyde-3-phosphate dehydrogenase; RT-PCR, reverse transcription-PCR; TRUNDD, TRAIL decoy receptor containing a truncated death domain; TRID, TRAIL decoy receptor lacking an intracellular domain; KILLER/DR5, p53-regulated proapoptotic KILLER/death receptor 5;FLIP, FLICE inhibitory protein; PARP, poly ADP-ribose polymerase; FADD,FAS-associated death domain protein; CMV-β-gal, cytomegalovirusβ-galactosidase; mAb, monoclonal Ab.