Anticancer drugs affect the alternative splicing of Bcl-x and other human apoptotic genes

Inducing an apoptotic response is the goal of most current chemotherapeutic interventions against cancer. However, little is known about the effect of chemotherapeutic agents on the alternative splicing of apoptotic genes. Here, we have tested 20 of the mainstream anticancer drugs for their ability to influence the production of Bcl-x splice isoforms. We find that many drugs shift splicing toward the proapoptotic Bcl-x_S splice variant in 293 cells. The drugs modulate splicing decisions most likely through signaling events because the splicing switch is not compromised by inhibiting de novo protein synthesis or the activity of caspases. Several drugs also shift Bcl-x splicing in cancer cell lines (MCF-7, HeLa, PC-3, PA-1, and SKOV-3), but the set of active drugs varies between cell lines. We also examined the effect of anticancer agents on the alternative splicing of 95 other human apoptotic genes in different cell lines. Almost every drug can alter a subset of alternative splicing events in each cell line. Although drugs of the same class often influence the alternative splicing of the same units in individual cell lines, these units differ considerably between cell lines, indicating cell line–specific differences in the pathways that control splicing. [Mol Cancer Ther 2008;7(6):1398–409]

Apoptosis or programmed cell death is a major pathway in the complex homeostatic balance between cellular proliferation and cell death. Conserved throughout evolution, apoptosis represents the endpoint of a cascade of events triggered by a variety of intracellular or extracellular stimuli including DNA damage, telomere dysfunction, oncogenic mutations, growth factor depletion, and hypoxia (1–4). Apoptosis is realized through the extrinsic (death receptor–dependent) or intrinsic (mitochondria–mediated) pathway, both routes activating caspases to execute cell death.

Mainstream anticancer agents act by triggering the apoptotic pathway (5–9). However, intrinsic alterations in the apoptotic pathway are a hallmark of cancer cells (2, 10, 11) and are considered to be a major cause of drug inefficacy (12, 13). Alternatively or in addition, the promotion of genomic instability by some cytotoxic agents may provide an opportunity for the emergence of additional lesions in the apoptotic pathway, in turn giving rise to even more aggressive cancers that become resistant to subsequent therapies (14).

The majority of the genes involved in apoptosis are alternatively spliced, in many cases to produce isoforms with opposing functions. Examples of this type are found in every category of apoptotic regulators including transmembrane receptors (e.g., Fas, Fas ligand, and LARD), adaptor molecules (e.g., Bcl-x, Bak, Apaf1, survivin, Mcl-1, and TRAF2), caspases (e.g., caspase-1, caspase-2, caspase-6, caspase-7, caspase-8, and caspase-9), and executors (e.g., FLIP and ICAD; ref. 15). The alternative splicing of apoptotic genes is often altered in cancers (16–18). Although a direct contribution of these changes to carcinogenesis and tumor development is in general poorly documented, the overexpression of specific antiapoptotic splice isoforms can protect cancer cells and contribute to their resistance to chemotherapeutic drugs. For example, the antiapoptotic Bcl-x_L isoform is overexpressed in many types of cancers (for examples, see refs. 19–21), and increasing the expression of the proapoptotic Bcl-x_S variant can kill cancer cells or improve their sensitivity to chemotherapeutic agents (17, 22, 23). Likewise, overexpressing the proapoptotic variant of caspase-2 in lung cancer cells sensitizes them to apoptosis on treatment with cisplatin (24).

Although alternative splicing is an important mechanism regulating apoptosis, the effect of chemotherapeutic agents on the alternative splicing of apoptotic genes has rarely been examined. Camptothecin, homocamptothecin derivatives, etoposide, amsacrine, doxorubicin, and mitoxantrone favor the formation of the proapoptotic caspase-2 isoform in U937 and HeLa cells (25, 26). A shift toward the proapoptotic Bcl-x_S isoform occurs when P388 cells are treated with a topoisomerase I inhibitor (27) and when cardiomyocytes are treated with daunorubicin (28). The gemcitabine-mediated increase in ceramide levels also stimulates the production of the proapoptotic splice isoforms of Bcl-x and caspase-9 in A549 cells (29). Finally, we have recently reported that the apoptotic inducer and protein kinase C inhibitor staurosporine can promote a splicing switch in 293 cells to favor the production of the proapoptotic Bcl-x_S isoform (30).

To further assess the relationship between chemotherapeutic agents and the alternative splicing of apoptotic genes, we have examined the effect of 20 mainstream anticancer drugs. We find that many anticancer drugs can shift Bcl-x pre-mRNA splicing to the proapoptotic Bcl-x_S splice variant in different cell lines. Interrogating other alternative splicing units reveals global effects of anticancer drugs on alternative splicing.

The human 293 cell line was from Invitrogen (EcR-293) and the human carcinomas of breast (MCF-7), prostate (PC-3), ovary (PA-1), and cervix (HeLa S3) were from the American Type Culture Collection. The ovarian cancer cell line SKOV-3ip1 (herein called SKOV-3) was kindly provided by Janet Price (The University of Texas M. D. Anderson Cancer Center; ref. 31). EcR-293 and HeLa S3 cells were grown in DMEM. MCF-7 cells were grown in Eagle's MEM supplemented with 10 μg/μL insulin. PC-3 cells were grown in Ham's medium/nutrient mixture F-12. PA-1 and SKOV-3 cells were grown in DMEM/F-12. All media were supplemented with 10% fetal bovine serum. To achieve ∼60% confluence before drug treatment, cells were seeded in six-well plates (35 mm) at the following densities: EcR-293 cells at 2 × 10⁵ per well, HeLa and MCF-7 cells at 1 × 10⁵ per well, PC-3 cells at 5 × 10⁴ per well, PA-1 cells at 7.5 × 10⁴ per well, and SKOV-3 cells at 6 × 10⁴ per well.

Drugs used and their modes of action are listed in Table 1. Drugs were obtained from the Service Pharmaceutique du Centre de Chimiothérapie at the Centre Hospitalier de l'Université de Sherbrooke. Drugs were diluted according to instructions from the manufacturers immediately before addition to cell culture medium. Cells were treated with individual drugs for 24 h at 37°C, 5% CO₂. The final concentrations of drugs used were 1 μmol/L gemcitabine, 5 μmol/L cytaralbine, 5 μmol/L capecitabine, 5 μmol/L methotrexate, 5 μmol/L 5-fluoruracil, 10 μmol/L vinorelbine, 10 μmol/L vincristine, 10 μmol/L paclitaxel, 10 μmol/L docetaxel, 0.5 μmol/L topotecan, 10 μmol/L etoposide, 25 μmol/L daunorubicin, 5 μmol/L epirubicin, 5 μmol/L dactinomycin, 10 μmol/L cisplatin, 10 μmol/L oxaliplatin, 50 μmol/L cyclophosphamide, 10 μmol/L chlorambucil, 10 μmol/L dacarbazine, and 10 μmol/L tamoxifen. Staurosporine (Roche Diagnostics) was used at a final concentration of 50 nmol/L.

The inhibitor of caspases z-VAD-fmk (Calbiochem) was used at a final concentration of 50 μmol/L and cells were treated for 2 h before anticancer drugs were added. After treatments, two thirds of the cells from each well were collected for protein extraction and the rest was used for RNA extraction. Whole-cell extracts were prepared by sonication in Laemmli sample buffer and protein concentration was determined using the Lowry method. Protein samples (20 μg/well) were fractionated on a 9% SDS-PAGE (29:1 acrylamide/bisacrylamide) and transferred onto a Hybond-C nitrocellulose membrane. Immunodetection was done according to standard protocols using a dilution of 1:500 of the rabbit anti-poly(ADP-ribose) polymerase (PARP) antibody (Biosource) and 1:5,000 of the horseradish peroxidase–conjugated anti-rabbit secondary antibody (Amersham Biosciences).

The transcription elongation inhibitor DRB and the protein translation inhibitor cycloheximide were used at the concentrations indicated.

For the conventional reverse transcription-PCR (RT-PCR) analysis of Bcl-x products, total RNA was extracted from cells using TRIzol (Invitrogen) and treated with DNase I for 10 min at 37°C followed by a treatment of 5 min at 70°C. Reverse transcription was done with a random hexamer using the OmniScript RT kit (Qiagen) at 37°C for 1 h. One tenth of the cDNA material was used for PCR. The primers used to amplify the Bcl-x alternative splicing events were BX3 (ATGGCAGCAGTAAAGCAAGCG) and BX2 (TCATTTCCGACTGAAGAGTGA). [α-³²P]dCTP (Amersham Biosciences) was added to PCR mixtures. PCR products were fractionated onto a 4% polyacrylamide gel. Dried gels were exposed on screens that were scanned on a STORM 860 PhosphorImager (Amersham Biosciences) and bands were quantitated using the ImageQuant software.

For the global analysis of splice isoforms, 2 μg total RNA was reverse transcribed using random hexamers and the Omniscript reverse transcriptase (Qiagen) in a final volume of 20 μL. cDNA (40 ng) was amplified with 0.375 units/15 μL HotStarTaq DNA polymerase (Qiagen) in the buffer provided by the manufacturer, without MgCl₂ and in the presence of the specific primers for each splicing unit (at concentrations ranging from 0.3 to 0.6 μmol/L) and deoxynucleotide triphosphates. The list of genes and the sequence of the oligos used are listed in Supplementary Table S2.³

Reactions were carried out in the GeneAmp PCR system 9700 (Applied Biosystems). A first cycle of 15 s at 95°C was followed by 35 cycles of 30 s at 94°C, 30 s at 55°C, and 1 min at 72°C. The reaction was ended with the extension step of 10 min at 72°C. Visualization and analysis of amplified products was done using the LabChip HT DNA assay on an automated microfluidic station (Caliper). All oligonucleotides used to amplify the set of 96 alternative splicing units are listed in Supplementary Table S2³ and the alternative splicing units being monitored are described in Revil et al. (30).

A set of 96 alternative splicing units from human apoptotic genes was selected from the AceView database (32). Sets of primers mapping in the exons flanking the alternative splicing events were designed using Primer3 with default variables (33). Analysis of RT-PCR results using the LabChip HT DNA assay yielded products with raw molarity and size values. We only considered products whose size ranged within 10% of the size of expected products. In the case of products associated with multiple isoforms, the closest one, using absolute size difference with the expected product, was chosen. The clustering of the data of splicing units, where expression of a minimum of two and maximum of three splicing isoforms were expected, was done with R (34) and RPy (R for Python) using the function heat map. We used unscaled data and a hierarchical clustering algorithm that used binary distance and ward hierarchical clustering.

We tested 20 drugs that belong to various groups based on their structure or putative mechanisms of action (see Table 1). The following cell lines were used: the human embryonic kidney 293 cell line, the cervical cancer cell line HeLa S3, the breast cancer cell line MCF-7, the prostate cancer cell line PC-3, and the ovarian cancer cell lines PA-1 and SKOV-3. Our reporter for alternative splicing was the endogenous Bcl-x pre-mRNA, which is spliced to produce primarily the antiapoptotic Bcl-x_L isoform in all untreated cell lines (Fig. 1). For each drug, we used concentrations that were inferior to those promoting extensive cell death based on visual inspection after 24 h of treatment. Following RNA extraction, changes in the relative abundance of the Bcl-x_L and Bcl-x_S splice variants were determined by RT-PCR (Fig. 1A; data not shown). Figure 1B plots the average changes in the percentage of Bcl-x_S product in triplicate for all drugs in all cell lines, except for SKOV-3 cells, which were done in duplicate. We considered a drug-induced change significant when it was >2 SD from the average values of the corresponding untreated samples (P < 0.05).

Twelve of the 20 anticancer drugs stimulated the production of Bcl-x_S in 293 cells (Fig. 1B). Previously, we observed that staurosporine shifted Bcl-x splicing in 293 cells but not in cancer cell lines (30). Here, several compounds affected the levels of Bcl-x splice products in the cancer cell lines MCF-7, HeLa, PC-3, PA-1, and SKOV-3. Oxaliplatin and cisplatin were active in all cell lines, whereas epirubicine was active in all cell lines, except PA-1. Gemcitabine, methotrexate, etoposide, daunorubicin, and dactinomycin improved Bcl-x_S production in at least four cell lines. Many drugs shifted the abundance of Bcl-x products in only one or two cell lines. For example, fluorouracil and cyclophosphamide had a significant effect only in 293 and SKOV-3 cells, respectively. Except for docetaxel in HeLa cells, drugs that interfere with microtubule formation did not affect Bcl-x splicing. This is unlikely to be due to suboptimal concentrations or poor drug intake because these drugs promoted PARP cleavage in 293 cells (Fig. 2C). Moreover, concentrations of drugs that increase cell death did not promote a switch toward Bcl-x_S for drugs that failed to do so at lower concentrations (data not shown). Thus, although drugs changed the relative proportion of Bcl-x splice isoforms, the set of active drugs vary between cell lines.

The switch in the relative abundance of Bcl-x products could reflect a change in alternative splicing or in the differential stability of the Bcl-x mRNA isoforms. For staurosporine, we previously ruled out an effect on Bcl-x mRNA stability because active transcription was required for the drug-mediated increase in Bcl-x_S (30). To determine if this conclusion also applied to the set of anticancer drugs that increase the relative abundance of Bcl-x_S in 293 cells, cells were pretreated with the transcriptional elongation inhibitor DRB (Fig. 2A) before treatment with the anticancer drugs. The results indicate that all drug-induced increases in Bcl-x_S require active transcription, supporting the view that the shifts are caused primarily by alterations in splice site selection.

To assess if de novo protein synthesis was required for the effect of the drugs on Bcl-x splicing in 293 cells, we tested the effect of the active drugs in the presence of the protein synthesis inhibitor cycloheximide. As reported previously (30), cycloheximide by itself stimulates the production of Bcl-x_S mRNA product in a concentration-dependent manner (Fig. 2B, control lanes). However, when the stimulation obtained with an anticancer drug was superior to that of cycloheximide alone, cycloheximide never reduced the level of stimulation (Fig. 2B), suggesting that the drug-induced splicing switches did not require de novo protein synthesis.

Chemotherapeutic agents may also exert their effect on Bcl-x splicing by inducing apoptosis, which can promote the caspase-dependent degradation of components of the splicing machinery. Indeed, spliceosomal components and RNA-binding proteins can be targets for caspases (35–38). To ask whether caspase activation is required for the effect on Bcl-x splicing in 293 cells, drug treatments were done with or without the pancaspase inhibitor z-VAD-fmk. Caspase activation was monitored through the cleavage of PARP, a preferential substrate for caspase-3 and caspase-7. PARP cleavage was induced by all drugs, except dactinomycin, dacarbazine, and fluorouracil (Fig. 2C). The level of PARP cleavage did not correlate with the shift in Bcl-x splicing, an indication that this shift might not be controlled by caspases. Although z-VAD-fmk itself had no significant effect on the splicing profile of Bcl-x in 293 cells, it prevented drug-induced PARP cleavage. In addition, z-VAD-fmk did not prevent the Bcl-x splicing shift induced by gemcitabine, methotrexate, topotecan, etoposide, daunorubicin, epirubicine, cisplatin, and oxaliplatin (Fig. 2D). Thus, the effect of anticancer agents on Bcl-x splicing in 293 cells does not require the activation of caspases.

Given that de novo protein synthesis and protein degradation are not required to obtain the drug-induced switch in Bcl-x splicing, the drugs are most likely activating endogenous signaling pathways. Notably, a switch toward Bcl-x_S was obtained by inhibiting protein kinase C or by activating Fyn kinase or protein phosphatase 1 (29, 30, 39). Whereas staurosporine and tamoxifen are known protein kinase C inhibitors (40, 41), the signaling events mediating the effect of the other drugs on Bcl-x alternative splicing remain to be identified.

The effect of anticancer drugs on Bcl-x splicing led us to ask if other alternative splicing units were sensitive to anticancer agents. A collection of 96 splicing units from apoptotic genes was interrogated in the 293, MCF-7, PC-3, PA-1, and SKOV-3 cell lines. The relative abundance of the RT-PCR products was determined by fractionation on microfluidic capillary workstations (42). Based on size estimation, all products corresponded to known splicing events and no novel splicing products were uncovered in this analysis. RT-PCR yielding products that were below an arbitrary threshold of sensitivity (2 nmol/L) were not considered. Percentage values for splicing products were compared with the values obtained with corresponding mock-treated samples. Given that 95% of the variations seen for all controls were within 15%, a change was declared significant when the difference was equal or superior to 15% (P < 0.05). The direction of the shift (e.g., exon inclusion or skipping) was not considered in this analysis. Splicing units that did not shift once in any cell lines are not included in the figures.

First, we compared the Bcl-x (BCL2L1) portion of this experiment with our previous manual assessment. Twenty-seven of the 29 cases of drugs affecting Bcl-x splicing in different cell lines had been detected in our first analysis (Fig. 1). Because more hits had been identified manually, the automated procedure was therefore highly specific (0.91) but less sensitive (0.55). Of the 15 cases of splicing modulation that did not match between the two analyses, this number was reduced to 8 when we included differences of ≥12% rather than ≥15%. Thus, based on the analysis of Bcl-x, there is a very good fit (84% overall concordance) between the two independent sets of experiments.

The number of significant shifts for each drug in each cell line is shown in Table 1. Overall, treatment of all cell lines with 20 anticancer drugs shifted alternative splicing in 12% of the cases. Except for two cases (vinorelbine in PA-1 and docetaxel in SKOV-3), all the drugs shifted at least one alternative splicing event in all cell lines. Splicing decisions in 293 cells were more responsive to drugs than in the cancer-derived cell lines.

To analyze the effect of different drugs in individual cell lines, we did a hierarchical two-dimensional clustering analysis (Fig. 3). The x axis of the clustered maps plots drug response according to alternative splicing units in the y axis. This clustering analysis reveals a tendency for drugs of the same group to be neighbors. For example, in 293 cells, antimetabolites and mitotic inhibitors formed relatively distinct clusters, whereas alkylating agents (platinum agents and DNA cross-linkers) clustered together. In PA-1 cells, platinum agents and several antimetabolites formed distinct clusters. Various topoisomerase inhibitors and intercalating agents were also clustering closely in MCF-7, PC-3, and SKOV-3 cells. Thus, drugs with common modes of action had a tendency to affect the same alternative splicing events in individual cell lines. This characteristic was also detected globally when target units for all cell lines are plotted relative to drugs (Fig. 4). Two distinct clusters were observed: one made of the antimetabolites and mitotic inhibitors and the other one containing the alkylating agents and the topoisomerase inhibitors. Thus, drug identity is making a significant contribution toward determining alternative splicing hits.

Although drugs of the same class were often neighbors, the hierarchical relationship between distinct clusters appeared to vary considerably between cell lines (Fig. 3). To get a better appreciation of this cell line–specific signature, we did a hierarchical clustering that helps see how individual splicing units reacted to individual drugs in all cell lines (Fig. 5). In this case, we only considered splicing events that could be monitored in all five cell lines (38 alternative splicing units) to ignore possible cell line–specific expression biases that might drive the clustering. The dendrogram clearly reveals that cell line identity was driving the clustering. Distinct clusters included a majority of the hits in 293, SKOV-3, PC-3, and MCF-7. For PA-1, hits were dispersed in three major subclusters, with one cluster positioned near the SKOV-3 hits, possibly reflecting their common ovarian origin. Overall, the coherence by cell line was good as indicated by a κ statistic of 0.65 (two-tailed 95% confidence interval, 0.53-0.76). Thus, despite a clear relationship between drug classes and splicing targets, cell line identity is also making an important contribution to determine which splicing units react to drugs.

Our global analysis can also be examined from the perspective of alternative splicing events because distinct units may share a common mode of regulation if their splicing is affected by the same drugs. For example, the TIA1 and TIAF1 splicing units belong to two related genes and both reacted to cisplatin and topotecan in 293 cells (Fig. 5). Likewise, TRAF2 and TRAF4 reacted to the same three drugs in SKOV-3 cells (Fig. 5). The case for common regulatory pathways is also supported when distinct units respond to the same drugs in different cell lines. For example, the response of alternative splicing units in genes of the Bcl-2 family (BCL2L11 and BCL2L12; BCLAF1 and BAX) and tumor necrosis factor receptors (TNFRSF6 and TNFRSF10B) was sufficiently similar in different cell lines to form distinct sets of adjacent or close neighbors (Fig. 5), suggesting that members in each set share common regulatory pathways. The other end of the spectrum is exemplified by BRCA1 and CASP9, which were affected by 18 and 19 drugs in SKOV-3 and 293 cells, respectively (Fig. 5; Supplementary Table S1),³ suggesting that numerous pathways may converge to regulate the alternative splicing of these units in SKOV-3 and 293 cells.

Alterations in apoptosis are associated with malignant growth and are a characteristic of nearly all types of cancers (2, 43). Likewise, aberrant profiles of alternative splicing occur frequently in cancer, but the precise contributions of these alterations to malignancy and metastasis remain poorly understood (reviewed in refs. 44, 45). Although the expression of specific splice isoforms likely influences cancer progression and the success of anticancer regimens, little is known about the effect of anticancer drugs on alternative splicing. To assess the general propensity of anticancer drugs to modulate alternative splicing, we have studied the effect of 20 anticancer agents on the alternative splicing of the important apoptotic regulator Bcl-x in 293 cells and five cancer cell lines. The analysis was further expanded to cover a 95 alternative splicing events occurring in apoptotic genes.

Our first conclusion is that a variety of anticancer agents increased the relative abundance of Bcl-x_S transcripts in 293 cells. Our experiments indicate that the drugs affected splicing and not the differential stability of Bcl-x mRNA variants because blocking transcription prevented the drug-induced shifts. Although we have used concentrations of drugs that did not promote massive apoptosis, we nevertheless tested the possibility that this splicing shift might be due to a caspase-dependent degradation of splicing factors. Clearly, the drug-induced Bcl-x splicing shift continued to be observed in the presence of a pan-inhibitor of caspases. Likewise, compromising de novo protein synthesis with cycloheximide did not block the drug-induced shift in Bcl-x splicing. Our results are consistent with the notion that the drugs might alter signaling pathways that impinge on Bcl-x splicing control. Recently, we have observed that protein kinase C is part of a network that represses the use of the 5′ splice site of Bcl-x_S in 293 cells (30). Although protein kinase C inhibitors like staurosporine, Gö6976, and tamoxifen improve the production of Bcl-x_S in 293 cells, these drugs did not affect Bcl-x splicing in the cancer cell lines HeLa, SKOV-3, U-87, MCF-7, and HCT 116 (30). It was therefore of interest to test a wider collection of anticancer drugs to see if Bcl-x splicing remained resistant in cancer cell lines. Notably, for each cancer cell line tested (MCF-7, HeLa, PC-3, PA-1, and SKOV-3), we identified drugs that increase the production of Bcl-x_S. However, the number of drugs active in cancer cell lines was smaller than those active in 293 cells and the composition of the active set varied considerably between cell lines. This cell line specificity was unlikely to be due to deficiencies in drug uptake because almost every drug affected the alternative splicing of other units in each cell line. Hence, our results indicate that the regulatory features of Bcl-x splicing are largely cell line specific. This may reflect cell line–specific differences in the nuclear abundance of splicing regulatory proteins. Alternatively or in addition, cell line–specific differences in the signaling pathways may alter the localization and activity of proteins that control Bcl-x splicing.

Analyzing the effect of drugs on other apoptotic splicing units revealed that different drugs affected a common subset of splicing events in each cell line. However, the composition of these targets varied considerably between cell lines. Thus, the networks of splicing regulation affected by the drugs appear largely cell line specific. Overall, this cell type specificity means that if a drug can shift splicing to proapoptotic products in some cancer cell types, it may not necessarily do so in others.

In contrast to the effect of drugs on Bcl-x splicing, which always shifted splicing toward the proapoptotic isoform, the effects of drugs on other alternative splicing units in apoptotic genes was not as clear cut. In 293 cells, most drugs activated the proapoptotic splicing event in Casp9 and survivin, but the reverse effect was seen on Fas and Siva. This profile did not consistently hold up in other cell lines. The ability of chemotherapeutic drugs to activate signals that simultaneously induce and antagonize apoptosis has been noted previously (25, 46–48), and our analysis expand these observations to splice isoforms with opposing activities.

The effect of anticancer drugs on the alternative splicing of Bcl-x and other apoptotic genes provides a novel rationale for using these compounds. The general ability of most drugs to favor the production of the proapoptotic Bcl-x_S isoform may help promote apoptosis in cancer cells or may sensitize cancer cells to other cytotoxic effects of the drugs (22, 49). Yet, a drug that favors the production of Bcl-x_S may not always encourage the production of proapoptotic variants in other genes. This may partly explain the variable success of these mainstream anticancer agents. Based on the information gathered in this study, it may be worth testing the cell-killing effect of drug combinations that favor the production of proapoptotic isoforms from several genes. It will also be important to continue assessing the function of the many apoptotic isoforms produced by alternative splicing and to determine how chemotherapeutic agents affect their production in different types of cancers. When taken in consideration with other factors, such as p53 status, this information may help improve the efficacies of future anticancer therapies.

No potential conflicts of interest were disclosed.

We thank Claudine Rancourt and Janet Price for the cell lines.