Expression of ERCC1 Antisense RNA Abrogates Gemcitabine-mediated Cytotoxic Synergism with Cisplatin in Human Colon Tumor Cells Defective in Mismatch Repair but Proficient in Nucleotide Excision Repair¹

dFdC³ has recently been recognized as a new anticancer agent with significant activity alone against a broad spectrum of tumors (1, 2). Furthermore, the relatively mild toxicity profile of the drug has generated considerable enthusiasm for including dFdC in combinations with other active anticancer drugs (3, 4, 5, 6, 7, 8, 9). Numerous combinations have been evaluated, and dFdC plus CDDP for the treatment of advanced non-small cell lung cancer has recently been shown to produce a more favorable response rate than that seen with the standard treatment of CDDP and etoposide (10). At the time this trial was initiated, there was little experimental evidence to guide the optimal design of the schedule of the two drugs in combination. Subsequently, investigations in cells in culture demonstrated synergistic cell killing (11, 12, 13). Furthermore, studies in tumor-bearing mice have suggested a schedule dependency such that dFdC and CDDP should be administered either simultaneously or within a short interval of one another to achieve the greatest therapeutic activity (14, 15). Although it has been suggested that the optimal schedules would increase the association of platinum with DNA, no mechanism by which this could occur has been defined.

It is known that the cell death caused by nucleoside analogues such as 1-β-d-arabinofuranoxylcytosine (16) and fludarabine (17) as well as dFdC (18, 19, 20) is associated with their incorporation into DNA during replication. Thus,it has been postulated that resynthesis of DNA patches during the various types of DNA repair that are initiated by radiation and chemotherapy damage to DNA would provide the opportunity for nucleotide analogue incorporation in cells that are not involved in DNA replication (21). Some evidence consistent with this hypothesis has been obtained for fludarabine (22, 23, 24, 25, 26). However, the mechanism(s) by which dFdC produces synergistic cytotoxicity with CDDP are not clear.

It is generally believed that CDDP exerts its cytotoxic effects by disrupting the DNA structure in cells through the formation of intrastrand adducts and interstrand cross-links. The former account for >90% of the total DNA lesions, whereas the latter represent about 3% of the lesions (27). CDDP adducts are repaired by the NER pathway (28), and the interstrand cross-links are repaired by the recombination repair pathway (29). The sensitivity of tumor cells to CDDP has been shown to be inversely correlated with cellular NER capability (30, 31). Contrarily, tumors that are defective in MMR become more resistant to CDDP than their MMR-proficient counterparts (32, 33, 34, 35). It has been postulated that MMR proteins serve as a detector of the CDDP-DNA adducts, and the recognition triggers cAbl-mediated apoptosis rather than inducing the repair of the adducts (36). Alternatively, as proposed by the so-called futile replication/repair cycles (37), MMR proteins recognize CDDP-DNA adducts from an error-prone translesion synthesis, but the repair proteins fail to correct the lesion because the repair is being directed at the nascent strand; repeats of the futile cycle result in DNA strand breaks that lead to cell death. Despite these postulations,the mechanisms of MMR protein-mediated CDDP cytotoxicity still await to be determined. Nevertheless, considering that tumors are often found to be defective in MMR, it is of interest to determine the role of the NER process in CDDP sensitivity and its cytotoxic synergy with dFdC in MMR-deficient tumor cells.

The NER pathway consists of damage recognition, dual incision/excision,repair synthesis, and ligation steps (28). Approximately 30 polypeptides that participate in this repair process have been identified (38). Among them, the ERCC1 protein is an important factor in the incision process—the rate-limiting step of the pathway. ERCC1 is a 15-kb repair gene located on human chromosome 19. Alternative splicing of ERCC1 precursor RNA produces RNA species of 1.1 kb or larger (39). ERCC1 forms a heterodimer with XPF, and the ERCC1/XPF complex is responsible for the incision to cleave the damaged strand at the phosphodiester bonds between 22 and 24 nucleotides 5′ to the lesion. A functional ERCC1 is important in the repair of CDDP-DNA adducts and in CDDP sensitivity in intact cells (40). ERCC1 has also been implicated to involve in the recombination repair of DNA interstrand cross-links (41), although the pathway of recombination repair is much less clear.

It has been shown that incorporation of dFdC into DNA would cause cell death (18, 19, 20), whereas synthesis of DNA repair patches would provide the opportunity for nucleotide analogue incorporation. Furthermore, recent evidence indicates that dFdC activates protein kinase C in human ovarian cancer cells (42) and that serine/threonine-specific protein phosphorylation plays an important role in modulating NER activity (43). Based on these findings, it is conceivable that dFdC may act as a NER inhibitor for DNA damage induced by CDDP and that NER inhibition may become an important mechanism through which dFdC mediates cytotoxic synergism with CDDP.

In this report, we selected an NER-proficient, MMR-deficient cell line to test this hypothesis. CP2.0 human colon tumor cell line, a NER-competent, CDDP-resistant subline derived from LoVo (44), was found to be deficient in MMR due to the loss of both alleles of the hMSH2 MMR gene (45). Using an in vitro repair assay and employing CP2.0 whole-cell extracts and CDDP-damaged plasmids as the substrate, we provide direct evidence that dFdCTP inhibits the repair of CDDP-induced DNA lesions. Furthermore, using an ERCC1 antisense RNA approach, we demonstrate that down-regulation of CDDP lesion-repair activities abolishes the cytotoxic synergism between dFdC and CDDP. These data strongly suggest a close link between dFdC-mediated inhibition of the damage repair and the cytotoxic synergism of dFdC and CDDP in NER-proficient, MMR-deficient CP2.0 human colon tumor cells.

dFdC was provided by Lilly Research Laboratories (Indianapolis,IN), and dFdCTP was synthesized by Sierra Bioresearch (Tucson, AZ). CDDP was purchased from Sigma Chemical Co. (St. Louis, MO).[α-³²P]dCTP (3000 Ci/mol) and[α-³²P]dATP (3000 Ci/mol) were obtained from Dupont-New England Nuclear (Boston, MA). Enzymes were purchased from Life Technologies, Inc. (Gaithersburg, MD).

The CP2.0 CDDP-resistant clone originally developed from a LoVo human colon carcinoma cell line was grown in Ham’s F-10 medium. The characteristics of this clone have been described previously (44). The clones (designated as AS-1 to AS-8) selected from transfected CP2.0 cells with the ERCC1 antisense expression plasmid were grown in Ham’s F-10 medium.

The cell-free system of Hansson and Wood (46) was used with minor modifications (24). Briefly, CP2.0 whole-cell extracts (100 μg) were incubated with 300 ng each of CDDP-damaged Bluescript KS⁺ plasmid (pBS) and nondamaged pHM plasmid (a gift from Dr. R. D. Wood, Imperial Cancer Research Fund, South Mimms, Herts, England) in the presence or absence of dFdCTP for 4 h at 30°C. The reaction buffer contained 45 mmHEPES-KOH (pH 7.8), 70 mm KCl, 7.4 mmMgCl₂, 0.9 mm DTT, 0.4 mmEDTA, 2 mm ATP, 2.5 μg of creatine phosphokinase, 3.4%glycerol, 18 μg of BSA, 20 μm each dGTP and dTTP, 4μ m each dATP and dCTP, and 2 μCi[α-³²P]dATP or[α-³²P]dCTP. At the end of the incubation,plasmid DNAs were purified, linearized with BamHI, and separated by gel electrophoresis. The incorporation of radioactive nucleotides was quantified after normalization for gel loading. DNA repair activity, expressed as specific incorporation of[α-³²P]dATP or[α-³²P]dCTP, was determined by subtracting the nonspecific incorporation in the nondamaged pHM control plasmid from the total incorporation in the CDDP-damaged pBS substrate.

The cytotoxicity was evaluated by a clonogenic assay. Exponentially growing cells were treated simultaneously with a 1:25 ratio of dFdC(0.1–1.0 μm for both CP2.0 and AS clones) and CDDP(2.5–25.0 μm for CP2.0 cells and 1.0–12.5μ m for AS clones) for 24 h before the cells were plated for determination of colony formation. The cytotoxic interaction of the two agents was analyzed by the median-effect method of Chou and Talalay (47) using the data from the survival curves. Theoretically, the combination index (CI) value determined by this method represents the ratio of the combination dose to the sum of the single-agent doses at an isoeffective level. Therefore, CI values <1 indicate synergism, values >1 show antagonism, and a value of 1 indicates additive effects.

To construct the antisense ERCC1 expression recombinant(pERCC1/AS), a BamHI/PvuII fragment derived from a pE12–12 plasmid (kindly provided by Dr. J. H. J. Hoeijmakers, Erasmus University, Rotterdam, the Netherlands) containing full-length ERCC1 cDNA was inserted in the inverted orientation at the downstream end of the cytomegalovirus (CMV) promoter in the pCEP4 vector (Invitrogen, Carlsbad, CA). To generate ³²P-labeled RNA probes, pBK-ERCC was constructed by inserting the ERCC1 full-length cDNA fragment into the BamHI/HindIII multiple cloning sites of the pBK-CMV vector (Stratagene, La Jolla, CA) so that in vitrotranscription of linearized pBK-ERCC would have the sense RNA transcribed by the T3 polymerase and the antisense RNA by the T7 polymerase.

CP2.0 cells growing in the logarithmic phase were transfected with purified pERCC1/AS using lipofectin according to the manufacturer’s instructions (Life Technologies, Inc.). The pERCC1/AS-transfected clones were selected using limiting dilution when the hygromycin (400μg/ml)-resistant colonies appeared; the selection was performed by determining the levels of ERCC1 mRNA and antisense RNA expression in cell colonies by Northern blotting.

RNA was prepared using the acid guanidinium thiocyanate phenol/chloroform method (22), separated on 1%formaldehyde-agarose gels, transferred to a nylon membrane, and hybridized with ³²P-labeled strand-specific RNA probes. Eight pERCC1/AS-transfected clones (AS-1 to AS-8) that expressed various levels of antisense ERCC1 transcripts were selected for the determination of their repair activities using in vitro DNA repair assay, and the cytotoxic interaction of dFdC and CDDP was further evaluated in these selected clones. Cells transfected with an empty vector were included in the experiment to serve as a mock transfection control (CP/vector).

The Wilcoxon signed-rank test was used for the comparison between the use of [α-³²P]dATP and[α-³²P]dCTP in the in vitro repair assay for measuring the effect of dFdCTP on the DNA repair activity. Univariate linear regression was applied to assess the relationships between the level of ERCC1 antisense RNA and ERCC1 mRNA; the ERCC1 antisense RNA and DNA repair activity; and the repair activity and the cytotoxic synergism. Spearman’s rank-order correlation coefficient, r, was used to assess the relationship between two variables that might not be normally distributed. A P < 0.05 was regarded as statistically significant.

To determine the effect of dFdCTP on repair, the in vitroDNA repair synthesis assay was adapted for these experiments. The pBS plasmid DNA containing CDDP-induced adducts was used as the substrate for the repair enzymes in the whole-cell extracts. The repair activity was measured by the specific incorporation of[α-³²P]dATP into damaged plasmid during repair synthesis. dFdCTP inhibited[α-³²P]dATP incorporation in a dFdCTP dose-dependent manner, with an IC₅₀ value of 69.6 ± 5.1 μm. When[α-³²P]dCTP was used in the repair reaction,the inhibitory effect of dFdCTP was greater (IC₅₀= 51.2 ± 4.7 μm; Fig. 1). The marginally greater inhibitory effects (P = 0.06, Wil-coxon signed-rank test) may be attributed to competition of dFdCTP with dCTP for DNA incorporation;this was suggested by the findings that addition of dCTP to the reaction mixture resulted in reduction of the inhibitory effect of dFdCTP on [α-³²P]dATP incorporation (Fig. 2). The inhibition by dFdCTP was almost completely reversed when the ratio of exogenous dCTP:dFdCTP reached 1:6.7 (7.5 μm:50 μm;Fig. 2). In contrast, the addition of dATP did not reverse the inhibitory effect of dFdCTP (data not shown).

To determine whether the inhibition of DNA lesion repair plays a contributory role in the synergism, we analyzed the cytotoxic interaction of dFdC with CDDP in NER-proficient, MMR-deficient CP2.0 cells. Fig. 3,A shows the survival of CP2.0 cells treated with various concentrations of dFdC and CDDP alone and in simultaneous combination. The data reveal a cooperative cytotoxic effect of the two agents. The precise mode of the cooperative interaction was further analyzed by the median-effect method. Fig. 3,B shows that the slopes of the median-effect plots for the single agents were not parallel, indicating that the two agents had different modes of action. Moreover, the slope of the combination did not parallel either slope of the single agents,suggesting that the cytotoxicity resulting from the interaction of these two agents may be produced through an effect that is generally not produced by either drug as a single agent. The CI values, analyzed using the conservative isobologram and shown in the fraction affected-CI plot constructed by computer analysis (Fig. 3 C),were between 0.51 and 0.27 at 0.6–1.0 inhibition levels, indicating that combined dFdC and CDDP produced significantly synergistic cytotoxic effects in CP2.0 cells, consistent with the findings by others in different tumor cell lines (11, 48).

The concurrence of the dFdCTP-induced repair inhibition and dFdC-mediated cytotoxic synergism supports an association between them. To confirm that the inhibition of repair activity contributes to cytotoxic synergism, we examined the effect of the constitutive expression of ERCC1 antisense RNA on the interaction of dFdC and CDDP. CP2.0 cells were transfected with pERCC1/AS, an expression vector with a constitutively active promoter. Eight clones (AS-1 to AS-8) expressing various levels of ERCC1 antisense RNA (Fig. 4,A) were selected from among 50 hygromycin-resistant clones for characterization of ERCC1mRNA expression (Fig. 4,B), repair activity (Fig. 4,C), and cytotoxic synergism (Table 1). The relationship between the levels of ERCC1 antisense RNA and sense RNA and the relationship between the ERCC1 antisense RNA level and the cells’ repair activity were evaluated using Spearman’s rank correlation test. The statistical analysis revealed that the expression of ERCC1antisense RNA correlated inversely with the level of ERCC1sense expression (r = −0.964; P =0.004; Fig. 4,D) and with the level of the repair activity(r = −0.952; P = 0.004; Fig. 4,E). The same test was used to assess the association between the repair activity and cytotoxic synergism. As shown in Fig. 5, there was a statistically significant correlation between the repair activity and cytotoxic synergism(r = 0.824; P = 0.012) in these transgenic clones. In other words, the level of down-regulated repair activity significantly correlated with the reduction in synergism. These results strongly suggest a close relationship between DNA repair activity and the cytotoxic synergism of combined dFdC and CDDP in MMR-deficient CP2.0 cells.

Using a cell-free system in which repair-patch synthesis activity was quantified by the specific incorporation of a radioactive nucleotide into a CDDP-damaged plasmid, we demonstrated that dFdCTP inhibits the repair of CDDP-induced DNA lesions in vitro(Fig. 1). The mechanism responsible for the inhibition is presently unclear. However, our finding that dFdCTP inhibited the incorporation of [α-³²P]dATP as well as[α-³²P]dCTP suggests that dFdCTP may work through a dual mechanism: competition with dCTP for the binding site of the polymerase and prevention of chain elongation after incorporation into the DNA. The recent finding that fludarabine triphosphate induces the formation of a truncated repair patch in repairing a site-specific CDDP-DNA adduct in vitro (49) supports the possibility that dFdCTP inhibits NER through premature termination of repair patch elongation.

The inhibition by dFdCTP of repair of CDDP-induced DNA lesions was concurrent with the cytotoxic synergism of the two agents (Fig. 3),supporting an association between the inhibition of DNA repair and the cytotoxic synergism in MMR-deficient but NER-proficient CP2.0 cells. A cause-effect relationship between the DNA repair inhibition and cytotoxic synergism is further supported by the results from the antisense experiment. In this experiment, the increased expression of the antisense RNA clearly resulted in reduced levels of ERCC1 mRNA and repair activity in the cells (Fig. 4). Furthermore, the expression of ERCC1 antisense RNA resulted in a simultaneous abrogation of DNA repair activity and cytotoxic synergism, in which the level of down-regulated DNA repair activity was found to correlate with the degree of the reduction in the synergism(Fig. 5).

It should be noted that although the clones that stably expressed ERCC1 antisense RNA displayed reduced cytotoxic synergism,they exhibited greater sensitivity to CDDP than did their parental CP2.0 cells (Table 1). These findings are consistent with those by Perfetti et al. (50) that the rates of cell kill after UV irradiation correlate inversely with ERCC1mRNA levels in various cell lines. Because no other biological characteristics, such as cell morphology, growth rate, or cell-cycle distribution were significantly altered in these transgenic clones(data not shown), the increase in CDDP sensitivity may be attributed to the reduced DNA repair capability as a result of a reduced level of ERCC1 expression.

The requirement of ERCC1 protein for NER of intrastrand adducts in mammalian cells has been well documented (40, 51, 52), but whether the ERCC1 protein is also involved in the recombination repair of interstrand cross-links is less clear. Based on the findings that ERCC1-deficient mouse fibroblasts were hypersensitive to UV irradiation but were only moderately sensitive to mitomycin C, it has been suggested that the ERCC1 protein is indispensable for NER but may not be essential for the recombination-mediated repair of interstrand cross-links (53). On the other hand, it has also been reported that ERCC1-mutant cell lines are hypersensitive to alkylating agents (41). Recently, Li et al.(54) demonstrated that, in an in vitrorecombination repair assay, the deficiency of DNA repair activity in the extracts from ERCC1- and XPF-deficient cells could be restored by adding the ERCC1/XPF heterodimer to the reaction mixture, which suggests that ERCC1 is required for efficient recombination repair. Thus, the abrogation of the cytotoxic synergism caused by the expression of ERCC1 antisense could result from the antisense ERCC1-mediated inhibition of both NER and recombination repair pathways.

In addition to the MMR proteins, other non-NER-related proteins, such as HMG domain-containing proteins (55, 56) and histone H1 proteins (57), have been reported to bind CDDP-DNA adducts. The involvement of these proteins may affect the efficiency of the NER process. In this regard, it is logical to question to what degree the observed dFdCTP-mediated NER inhibition represents a secondary effect of dFdC interaction with these proteins. Further experimentation would be required to explore these links. However,although the possibility exists that dFdC may produce its repair-inhibitory effect through inhibiting the synthesis of these binding proteins (e.g., by self-incorporation into the repair protein-encoding genes), the hMutSα MMR protein (heterodimer of hMSH2 and hMSH6) and HMG1-box protein have been shown to bind preferentially to 1,2-d(GpG) and d(ApG) adducts. The same proteins,however, poorly recognize 1,3-d(GpNpG) CDDP-induced DNA adducts (32, 56). Considering that d(GpG) and d(ApG) adducts together account for about 85% of the total CDDP-induced DNA lesions,fractions of these adducts could conceivably escape from binding to the proteins in the cells. As to MMR proteins, although presently prevailing models suggest that cAbl-mediated apoptosis (36) and futile replication/repair cycle-induced cell death (37) may be the mechanisms responsible for CDDP cytotoxicity, it remains to be determined what roles the MMR system may play in dFdC-mediated cytotoxic synergism with CDDP.

van Moorsel et al. (15) recently showed that incorporation of dFdCTP into DNA could facilitate the binding of CDDP to DNA in cell line models as assessed by the Pt-DNA formation measured by atomic absorption spectrometry. In the same report, however, they found no increment of Pt-DNA retention induced by dFdC in the cell lines tested. Therefore, their results would argue against our theory that dFdC-mediated repair inhibition is primarily responsible for the synergistic interaction. Nevertheless, our results clearly show that dFdC inhibits the resynthesis step of the repair process. Such a mode of DNA repair inhibition by dFdC would unlikely result in increased DNA platination and adduct retention. The present data support the dFdC-mediated DNA repair inhibition as a mechanism responsible for the cytotoxic synergy of dFdC and CDDP; however, other mechanisms, such as changes in the intracellular accumulation and cross-linking efficiency of CDDP, cannot be totally excluded.

The IC₅₀ value (69.6 ± 5.1μ m) of dFdCTP for the inhibition of repair of CDDP-induced DNA lesions is in the range of concentration readily achievable in the clinic (58). The self-potentiating action of dFdC would promote its phosphorylation efficiency (59) and slow its elimination (60) by cells. These biological characteristics of dFdC aid in enhancing the intracellular level of dFdCTP relative to dCTP and consequently favor the incorporation of the former into DNA. In addition, once dFdCTP is incorporated into DNA, this nucleotide analogue is difficult to excise by proofreading exonucleases (20, 61). This characteristic would further enhance the agent’s inhibitory effect during the DNA repair synthesis process in cells.

In conclusion, the results of this study provide direct evidence that dFdCTP inhibits the repair of CDDP-induced DNA lesions and suggest that the incorporation of dFdCTP into the repair patch may trigger signaling pathways that lead to cell death. The evidence also suggests that dFdC-mediated inhibition of repair of CDDP-induced lesions plays a critical role in the cytotoxic synergism of dFdC and CDDP in NER-competent, MMR-deficient CP2.0 human colon tumor cells.