Role of BAX in the Apoptotic Response to Anticancer Agents

See allHide authors and affiliations

Science  03 Nov 2000:
Vol. 290, Issue 5493, pp. 989-992
DOI: 10.1126/science.290.5493.989


To assess the role of BAX in drug-induced apoptosis in human colorectal cancer cells, we generated cells that lack functional BAX genes. Such cells were partially resistant to the apoptotic effects of the chemotherapeutic agent 5-fluorouracil, but apoptosis was not abolished. In contrast, the absence of BAX completely abolished the apoptotic response to the chemopreventive agent sulindac and other nonsteroidal anti-inflammatory drugs (NSAIDs). NSAIDs inhibited the expression of the antiapoptotic protein Bcl-XL, resulting in an altered ratio of BAX to Bcl-XL and subsequent mitochondria-mediated cell death. These results establish an unambiguous role forBAX in apoptotic processes in human epithelial cancers and may have implications for cancer chemoprevention strategies.

The induction of apoptosis, or programmed cell death, in cancer cells is thought to be fundamental to the success of treatments for cancer. The Bcl-2 family members are intimately involved in the apoptosis (1,2), but the role of these proteins in drug-induced death has been confusing. BAX, the prototypic death-promoting member of theBcl-2 family, provides a good example of the complications that have arisen. Many studies have relied on overexpression of BAX protein, conditions that may not faithfully reproduce its normal activity, and have yielded conflicting results (3). Drugs induce endogenous BAX expression through p53-dependent transcription in some cancer cell lines, but not others (4). In mice, BAX plays no role in the most well-studied examples of drug- or radiation-induced and p53-dependent apoptosis, involving thymocytes and intestinal epithelium (5–7). Unlike the human gene, the murineBAX gene has no p53-binding site in its promoter (8). Nevertheless, BAX deficiency promotes drug resistance in murine fibroblasts by partially attenuating p53-dependent apoptosis, but only when such cells are transformed with the adenoviralE1A oncogene (3, 9). This picture is further confounded by the finding that BAX deficiency can promote rather than inhibit apoptosis in some murine cell types (10).

The most important targets of chemotherapeutic agents are human epithelial cells, which give rise to the vast majority of naturally occurring cancers. To clarify the role of BAX in drug-induced apoptosis in such cells, we created and studied isogenic derivatives that differ only in the presence or absence of theBAX gene. HCT116 colorectal cancer cells were chosen as the parental cells because they contain normal p53 andBAX genes and undergo apoptosis in response to 5-fluorouracil (5-FU) and sulindac. 5-FU is the mainstay of treatment for colorectal cancer and is an antimetabolite that induces cell death in a p53-dependent manner (11). Sulindac is the prototypic chemopreventive agent for patients with colorectal cancer predisposition and is a nonsteroidal anti-inflammatory drug (NSAID) that binds to and inhibits cyclooxygenases and other cellular proteins (12–14).

To obtain isogenic cells differing in BAX status, we first exploited the innate propensity of mismatch repair (MMR)–deficient cells to mutate mononucleotide tracts (15). BAXcontains an unstable G8 tract at nucleotides 114 to 121 (codons 38 to 41) that is often mutated in MMR-deficient tumors (16). Through analysis of unselected subclones (17), we found that 2% of HCT116 cells had two intact BAX alleles, 94% had one intact allele (+/−, with one allele containing a deletion or insertion of a G within the G8 tract), and 4% had two mutant alleles (−/− cells). Western blots confirmed the absence of BAX protein in theBAX−/− cells (Fig. 1A). In BAX +/+and BAX +/− cells, there was a slight induction of BAX protein by agents that activate p53 (Fig. 1A). Induction of p53 by 5-FU caused equivalent amounts of apoptosis inBAX +/+ and heterozygousBAX +/− cells. The extent of apoptosis was somewhat reduced in the BAX−/− cells, butBAX deficiency did not recapitulate the effects ofp53 deficiency, which provided nearly complete protection against 5-FU–induced apoptosis (Fig. 1B) (11). In marked contrast, BAX deficiency completely eliminated the apoptosis induced by sulindac (Fig. 1B). In this case, p53 deficiency had no effect, indicating that the apoptosis induced by sulindac wasp53-independent. A similar effect of BAX deficiency was observed on the apoptosis induced by indomethacin, another NSAID, but little protection was observed against ceramide, a non-NSAID inducer of apoptosis [Fig. 1B and Web fig. 1 (18)].

Figure 1

Drug effects on cells of varying BAXgenotypes. (A) BAX protein expression in cells treated with 5-FU. HCT116 cells with the indicated BAX genotypes were treated for 48 hours with 5-FU (375 μM, Roche). Equal amounts of total cellular proteins from these cells were separated by SDS–polyacrylamide gel electrophoresis and subjected to immunoblotting with a polyclonal antibody to BAX (N-20, Santa Cruz). The molecular size of the BAX protein is 21 kD. (B) Fraction of cells undergoing apoptosis after 48 hours of treatment with the indicated drugs. Cells of the indicated BAX genotype were tested, along with cells in which p53 orp21 genes were disrupted through gene targeting (p53 KO and p21 KO, respectively) (24). 5-FU (Roche), sulindac sulfide (Merck), and indomethacin (Sigma) were used at final concentrations of 375, 120, and 500 μM, respectively. Cells (attached plus those floating in the medium) were harvested at varying time points after drug treatment and fixed in a solution containing final concentrations of 3.7% formaldehyde, 0.5% Nonidet P-40, and Hoescht 33258 (10 μg/ml) in phosphate-buffered saline. Apoptosis was assessed through microscopic visualization of condensed chromatin and micronucleation after staining of cells with DAPI (25). At least two independent experiments were carried out for each condition, and a minimum of 300 cells were counted for each measurement. The sulindac and indomethacin responses of the BAX-deficient cells were significantly different from the other cells tested (P< 10−8, Student's t test).

To demonstrate that BAX deficiency was responsible for the observed effects, we prospectively inactivated the wild-type (WT) BAX allele in a BAX heterozygote. A targeting vector was constructed (Fig. 2A) and transfected into HCT116 cells for this purpose (19). We recovered two clones in which the remaining WT BAX allele had been targeted (KO1 and KO2) (19), as verified by Southern and Western blotting (Fig. 2B) and DNA sequencing [Web fig. 2 (18)]. Treatment of these cells with 5-FU and NSAIDs (Fig. 2C) revealed that they behaved identically to the BAX−/− clones shown in Fig. 1; deletion of BAX provided slight protection against ceramide-induced apoptosis [Web fig. 3 (18)], partial protection against 5-FU–induced apoptosis (Fig. 2C), and profound protection against the apoptosis induced by the NSAIDs sulindac and indomethacin (Fig. 2C). Controls for these experiments were provided by heterozygous sister clones that arose after transfection with the targeting vectors; the behavior of these BAX+/− clones was identical to that of the HCT116 parental cells, as expected (Fig. 2C).

Figure 2

Targeted deletion of BAX.(A) BAX genomic locus and the targeting construct. The targeting construct consists of two homologous arms and a hygromycin-resistant gene (Hygro). The Eco RI sites within theBAX gene and the targeting construct and position of the probe used for Southern blotting are shown. Boxes I to VI represent BAX exons 1 through 6. (B) Southern blot after Eco RI digestion of genomic DNA of selected clones (top). Fragments corresponding to the wild-type allele and targeted allele are 10.0 and 8.4 kb, respectively. Western blots of proteins from these clones are shown at the bottom; an antibody to BAX was used to detect the 21-kD BAX polypeptide. (C) Fraction of cells undergoing apoptosis after drug treatment. Cells were treated with the indicated drugs as described in Fig. 1B for the times shown on the xaxis. “BAX +/−” refers to a sister clone isolated from the transfection experiment used to generate the knockout clones. The fraction of apoptotic cells was assessed by fluorescence microscopy of DAPI-stained cells.

If BAX deficiency so profoundly affects the sensitivity to NSAIDs, one might expect that parental cell populations treated with NSAIDs would be enriched in cells with mutations of BAX.This hypothesis was tested by recovery of the clones growing after indomethacin treatment (Fig. 3A). Of 60 clones analyzed, 70% (42 out of 60) had insertions or deletions in the G8 tracts of both BAX alleles, as compared with only 4% (4 out of 96) in the parental population (P < 10−12, χ2 test). The clones withBAX mutations were resistant to both indomethacin and sulindac (Fig. 3B).

Figure 3

Cells selected for resistance to indomethacin contain BAX mutations. (A) HCT116 cells were treated with indomethacin (500 μM) for 72 hours. Cells were harvested and plated into 96-well plates without drug. Single clones were recovered and expanded. PCR analysis of theBAX gene shows that 17 out of 24 clones that arose after indomethacin treatment contained mutations of both BAXalleles (lower panel), whereas such mutations were rare in clones that arose in the absence of indomethacin treatment (upper panel). Most mutants had a deletion of a G within the normal G8tract, whereas clones 13 and 21 had an insertion (creating G9) and clone 24 had a 2–base pair deletion (creating G6). (B) Clones 2 and 3 (labeled Indo Clone #2 and #3) were expanded and tested for apoptosis after indomethacin treatment, along with control clones of the indicated genotypes.

What are the mechanisms by which NSAIDs cause BAX-dependent apoptosis? Previous studies have shown that alterations in the ratio between proapoptotic and antiapoptotic members of the Bcl-2family, rather than the absolute expression level of any singleBcl-2 family member, can determine apoptotic sensitivity (20). Bcl-2 was not detectably expressed before or after treatment with NSAIDs, but the expression of the antiapoptotic Bcl-XL protein was substantially reduced by NSAIDs, both in parental HCT116 cells and in their p53−/− derivatives (Fig. 4A). The inhibition ofBcl-XL expression was also observed at the RNA level [Web fig. 4 (18)], suggesting a transcriptional basis for this decrease.

Figure 4

Mechanisms underlyingBAX-dependent, NSAID-induced apoptosis. (A) Expression of Bcl-2 family proteins after sulindac treatment. Parental HCT116 cells and p53 knockout cells were treated with sulindac sulfide (120 μM). Equal amounts of total cellular proteins collected at the indicated time points were analyzed by immunoblotting with antibodies against BAX, Bcl-2 (N-19, Santa Cruz), Bcl-XL (Transduction Laboratories), and α-tubulin (TU-02, Santa Cruz ) as loading control. The lane labeled “+” represents protein from HL-60 cells, which normally express Bcl-2. (B) Bax and Bcl-XL proteins from the indicated colorectal cancer cell lines were analyzed by immunoblotting after treatment with either sulindac or indomethacin. (C) Caspase 9 Western blot of parental andBAX knockout cells after treatment with sulindac for the indicated times (in hours). Equal amounts of total cellular proteins were separated and immunoblotted with a caspase 9–specific antibody (H-83, Santa Cruz), which recognizes the COOH-terminus of the protein. The molecular size of the intact caspase 9 polypeptide is 46 kD. A degraded fragment of caspase 9 (37 kD) is detected in parental cells, but not in BAX knockout cells. (D) Colony-formation assays. (Top) About 5000 cells were treated with indomethacin as described in Fig. 1B. Cells (attached plus those floating in the medium) were harvested after 72 hours of treatment and plated into T25 flasks without drug. (Bottom) Cells were transfected with a Bcl-XLexpression vector or a control (empty) vector and treated with zeocin (to select for transfectants) or zeocin plus indomethacin for 48 hours (26). In all cases, clones were visualized by crystal violet staining 7 to 10 days later, and the results shown are representative of at least three independent experiments. BAX gene disruption resulted in a greater than 100-fold increase in colony number compared with parental cells, whereas exogenous Bcl-XL expression resulted in a greater than 10-fold increase in colony number compared with cells transfected with the empty control vector.

The hypothesis that the ratio between BAX and Bcl-XLproteins plays a major role in sensitivity to NSAIDs led to several testable predictions. First, this ratio should change in other colorectal cancer cell lines treated with sulindac. This prediction was tested in a total of eight cell lines. Despite the heterogeneity in response to sulindac previously noted in such lines (21), we found that five of eight lines responded similarly, with no increase in BAX but substantial increases in the ratio of BAX:Bcl-XLdue to decreases in Bcl-XL expression (Fig. 4B). Second, NSAIDs with little structural similarity to sulindac should induce similar changes in the ratio between BAX and Bcl-XL; this prediction was confirmed by analysis of cells treated with indomethacin (Fig. 4B). Third, the change in BAX/Bcl-XL ratio should cause apoptosis through a mitochondrial pathway (1,2). This pathway was indeed affected in sulindac-treated cells, as the mitochondrial membrane potential was disrupted [Web fig. 5 (18)], caspase 9 was activated (Fig. 4C and Web fig. 6), and DNA was subsequently degraded in cells with intact BAXgenes but not in BAX-deficient cells [Web figs. 7 and 8 (18)]. Fourth, a biologically significant difference in NSAID sensitivity should be reflected in standard colony-formation assays, in which not just apoptosis but all parameters related to cellular growth and death can be simultaneously assessed. We found thatBAX disruption led to a profound difference in sensitivity to NSAIDs, whereas disruptions of p53 or p21 had no effect in standard colony-formation assays (Fig. 4D and Web figs. 9 and 10 (18)]. Finally, the forced expression of Bcl-XL should rescue cells with intact BAX genes from NSAID-mediated apoptosis. This prediction was confirmed by introducing a Bcl-XL expression vector into HCT116 cells and assessing colony formation after treatment with NSAIDs (Fig. 4D).

It was not expected that deletion of a single gene could so profoundly affect cell death in a human cancer cell. Previous experiments have shown that NSAIDs induce heterogeneous changes in human tumor cells, including growth arrest, apoptosis, and necrosis (21). Despite this heterogeneity, we found that most colorectal cancer cell lines exhibited markedly similar changes in BAX:Bcl-XLratios, whereas the others died through mechanisms that were independent of BAX:Bcl-XL. Furthermore, the importance ofBAX was rigorously demonstrated in HCT116 cells through three different approaches for generating cells with disruptedBAX genes. In addition to their implications for understanding basic determinants of drug responsiveness in human cancer cells, these results may have important clinical implications. It is currently believed that chemoprevention offers the best hope for nonsurgical management of patients with hereditary predispositions to colorectal cancer. The most common form of such predisposition is hereditary nonpolyposis colorectal cancer, which is caused by defects in mismatch repair (22). Our results suggest that such tumors may easily develop resistance to NSAIDs through an inherent instability in the mononucleotide tract in BAX. By analogy with the successful strategy used to combat the highly mutable retroviruses that cause acquired immunodeficiency syndrome (23), it may therefore be important to consider combinations of chemopreventive drugs, rather than single agents, in such patients.

Note added in proof: It has recently been shown that tumor cells without BAX undergo less apoptosis when xenografted in mice (27).

  • * To whom correspondence should be addressed. E-mail: vogelbe{at}


View Abstract

Navigate This Article