Modulation of drug resistance by artificial transcription factors

Acquired resistance to anticancer drugs is a formidable problem that hampers current chemotherapeutic treatments. Tumor cells develop drug resistance by dysregulating expression of one or more genes that affect a variety of mechanisms (1). Commonly dysregulated genes include ABC transporters (e.g., the MDR1 gene product), genes involved in drug metabolism (e.g., cytochrome P450), genes controlling cell cycle progression (e.g., cyclin inhibitors), and antiapoptotic genes (e.g., the bcl family members; refs. 2, 3). The MDR-1 gene confers resistance to several drugs, including paclitaxel, doxorubicin, and vincristine. Inhibitors of MDR-1 have been developed, including small molecules, ribozymes (4, 5), antisense oligonucleotides (6, 7), small interfering RNA (siRNA; refs. 8, 9), and de novo designed artificial transcription factors (ATF; ref. 10). However, the multiplicity of molecular responses that trigger drug resistance in cancer cells, or multifactorial multidrug resistance (1), poses a fundamental limitation for single, directed, gene targeting approaches.

In contrast to inhibitors that target single genes, natural transcription factors typically recognize DNA sequences located in several promoters, thereby orchestrating expression of many genes and subsequent cellular responses. Transcription factors, such as p53, a tumor suppressor lost or mutated in 50% of human tumors (11–13), play a central role in regulating multiple targets in response to many stress responses, including DNA-damaging signals (11–19). Like natural transcription factors, ATFs may recognize DNA sequences located in several promoters and thus can regulate expression of more than one gene. In addition, ATFs can be designed to mediate both up- and down-regulation of targeted genes. When the DNA-binding domain (DBD) of an ATF is linked to an activation domain, the resulting transcriptional activator is able to up-regulate target gene expression, whereas linkage of a repression domain down-regulates target gene expression (20). We reasoned that multiplicity of targeting, together with transcriptional modulation of gene expression mediated by ATFs, could be used to efficiently regulate resistance to anticancer drugs in cancer cell lines. To test the capability of ATFs to down-regulate resistance to anticancer drugs, we have adopted a genome-wide screening strategy. We have described previously the construction of ATF libraries by recombination of sequence-specific zinc finger DBDs (21–23). In this article, we delivered three-finger (3ZF) and six-finger (6ZF) zinc finger activator libraries into HeLa cells to identify and regulate gene pathways involved in drug resistance.

The pMX-3ZF-library-VP64-IRES-GFP and the pMX-6ZF-library-VP64-IRES-GFP were transfected into 293 cells that express the gag-pol genes as described (21, 23). Viral supernatant was used to infect 3 × 10⁵ and 10⁸ HeLa cells, plated in 20 × 150 cm tissue culture dishes, for the 3ZF and 6ZF libraries, respectively. After 72 h, the cells were treated with 200 nmol/L paclitaxel for 3 days. As a control, mock-transduced cells were treated with the same concentration of paclitaxel. After drug treatment, the cells were washed once with PBS and once with medium and allowed to proliferate for a period of 1 week (to avoid the selection of nonproliferating, terminally differentiated cells). After the drug treatment, no cells survived from mock-transduced plates, but we obtained up to five resistant colonies per plate from the library-transduced cells. Genomic DNA was extracted from surviving cells and the DNA encoding the zinc fingers was extracted by PCR and recloned into the pMX-IRES-GFP vector as described (21, 23). Single clones were sequenced and processed as described below for functional analysis.

Individual 3ZF or 6ZF proteins expressed in the retroviral vector pMX-ZF-VP64-IRES-GFP were individually transduced into HeLa cells. An XTT (Roche) survival assay to determine the fraction of surviving cells was done following the manufacturer's instructions. The experiment was repeated at least three times with independent transductions to ensure reproducibility. Statistical significance between groups (ATF-transduced clones and control cells) was determined using a Student's t test (P < 0.05 was considered significant).

The zinc fingers were cloned in pMalC2 vector (NEB), which enables expression of the zinc finger DBD as a COOH-terminal fusion with the bacterial carrier protein maltose-binding protein. The zinc finger DBD-maltose-binding protein were expressed in Escherichia coli XL-1 and purified using an amylose resin as described by the manufacturer's instructions (NEB). Biotinylated oligonucleotides comprising the ATF-binding sites were added to streptavidin-coated 96-well ELISA plates as described (24). Purified protein in binding buffer [ZFA buffer containing 5 mmol/L DTT/1% bovine serum albumin/sonicated salmon sperm 140 ng/μL (Sigma)] was added at different concentrations in the linear range of this assay. Binding was normalized to 1 with the cognate substrate signal for each protein. DNA mobility shifts and K_d determinations were done as described (24).

Cells were collected and the RNA was extracted using the RNeasy extraction kit (Qiagen) and reverse transcribed using the high-capacity cDNA Archive kit (Applied Biosystems). Real-time PCR conditions and DNA microarray experiments are described in Supplementary Material.⁵

The WAF1/CIP1 promoter reporter plasmid was kindly provided by Carol Prives (Columbia University). For the construction of a human DMP1 promoter reporter, we amplified a 2.3-kb genomic fragment 5′ of exon 1 using a PCR approach and oligonucleotide primers (forward 5′-GAGCTCTTCATTCCTCCATTAGCACAGCAATCTCCATCAGC-3′ and reverse 5′-CTCGAGTCCGGGCACTTTGGAAGAACCAGGATGGAAGCTC-3′); the primer used incorporated SacI and XhoI restriction sites into the fragment. The resulting promoter fragment was cloned into the SacI and XhoI sites of the pGL2-Basic vector (Promega). For transfection experiments, 293T cells were seeded into a 24-well plate at a density of 3 × 10³ cells per well. After 24 h, cells were transfected using a modified calcium phosphate method (25). For the reporter assay, 100 ng reporter plasmid was cotransfected with 50 or 200 ng zinc finger expression plasmid or pcDNA3.1 vector. pRL-TK Renilla luciferase vector (4 ng) was cotransfected in each experiment as an internal control for transfection efficiency. After 48 h, cell lysates were prepared and luciferase activity was measured according to the manufacturer's Dual-Luciferase reporter assay protocol (Promega). Each transfection was done in triplicate and data are presented as the mean ± SE.

Individual clones of ATF-transduced cells were plated in triplicate in a 96-well format tissue culture plates (2,000 cells per well). Cells were allowed to proliferate for 4 days. Growth was assessed by XTT survival analysis (Roche), done every 24 h, according to the manufacturer's instructions.

Individual clones of ATF-transduced cells were harvested (∼10⁶ cells per clone) and treated with 0.5 mL PBS containing RNase A (0.02 mg/mL; Qiagen) for 30 min in the dark and then chilled on ice for 10 min. A 25 μL aliquot of 1 mg/mL propidium iodide (Sigma) was added. Cells were incubated in the propidium iodide solution for 1 h at room temperature in the dark and then analyzed by flow cytometry using a FACSCalibur (BD). The percentages of cells in the G₀-G₁, S, and G₂-M phases of the cell cycle were determined using CellQuest (BD) and ModFit software (BD). The experiment was done three times with three independent transductions to ensure reproducibility.

Three- and six-finger libraries linked to the VP64 activation domain were delivered into HeLa cells using the pMX-IRES-GFP retroviral vector (26). We chose HeLa cells because this cell line is sensitive to the drug paclitaxel and expresses wild-type p53. The conditions described below were the same as described for peptide library selections using retroviral libraries in HeLa cells (27). As described elsewhere (21), a 3ZF library with a complexity of ∼10⁴ different clones can recognize most 9-bp sequences of type 5′-(G/A/TNN)₃-3′. A 6ZF library of complexity 7.4 × 10⁷ allows recognition of most 18 bp of sequences of the type 5′-(G/A/TNN)₆-3′. We transduced 3 × 10⁵ HeLa cells with a 3ZF library and 10⁸ HeLa cells with a 6ZF library. After 72 h, the transduced cells were treated with 200 nmol/L paclitaxel. After an additional 72 h, virtually all untransduced cells became apoptotic and detached from the plate. However, several 3ZF- and 6ZF-transduced colonies were able to survive to paclitaxel treatment. Retroviral clones that survived in this assay were replated and propagated for a period of 1 week in absence of drug as described (28). The genes encoding the 3ZF and 6ZF DBDs were extracted by PCR using genomic DNA from the transduced cells and recloned into the pMX-IRES-GFP vector as described (21, 23).

Individual ATF clones were reintroduced into HeLa cells and individually challenged with paclitaxel. From a total of 15 3ZF clones and 30 6ZF clones tested individually, we isolated several ATFs able to promote paclitaxel resistance in a survival assay (Table 1; Fig. 1A). One of the 6ZF sequences was identical from two independent clones, indicating selective pressure for particular ATF members. As shown in Fig. 1A and Supplementary Fig. S1,⁵ selected ATF induced different degrees of paclitaxel resistance. The strongest effect was achieved with a 3ZF protein, 3ZF-1-VP. Each of the ATF that promoted paclitaxel resistance also protected HeLa cells from the drug etoposide, a topoisomerase II inhibitor (Fig. 1B; Supplementary Fig. S2)⁵ at concentrations higher than 5 μmol/L. The ATF did not have a significant effect on toxicity of the drug doxorubicin (data not shown). In contrast to results from a recent report (29), we were not able to select ATFs that imparted significant resistance to paclitaxel or other anticancer drugs by using transient expression vectors, even after five rounds of selection.

To investigate whether the induction of drug resistance phenotype by ATFs depended on the nature of the effector domain, we expressed the DBDs as fusions with a strong transcriptional repression domain (Kruppel domain). All selected proteins, except ATF 3ZF-6-KD, had no effect on the drug resistance phenotype relative to the cell line (Supplementary Fig. S3).⁵ ATF 3ZF-6 promoted drug resistance independently of the effector domain; this zinc finger protein might act either through simple DNA binding and competition with endogenous proteins or through post-transcriptional regulation. The other ATF selected in this screen required the activation domain, and therefore transcriptional gene activation, for phenotype induction.

Next, we determined whether the selected 3ZF and 6ZF DBDs were able to bind their predicted DNA substrates in vitro. These substrates were defined according to the nature of the DNA-binding α-helical sequences indicated in Table 1. DBDs were expressed as COOH-terminal fusions with bacterial maltose-binding protein. Electrophoretic mobility shift assays and DNA-binding specificity ELISA were used to determine binding affinities and specificities. As shown in Fig. 2, all zinc finger-maltose-binding protein fusions bound predicted target sites with high sequence selectivity. These results show that ATFs isolated from functional phenotypic screens retain their predicted substrate specificity. The affinities of the zinc finger proteins were in the 1 to 5 nmol/L range (Table 1). As expected, 6ZF proteins had higher affinities than 3ZF proteins due to the presence of additional zinc finger-DNA interactions.

Next, we investigated whether the ATF enhanced survival of HeLa cells treated with anticancer drugs with unrelated mechanisms of action. We focused on ATF 3ZF-1-VP because this protein had the strongest activity and protected cells from higher concentrations of paclitaxel than the other ATFs evaluated (Fig. 1; Supplementary Fig. S1).⁵ We challenged HeLa cells retrovirally transduced with 3ZF-1-VP with several commonly used anticancer drugs. As shown in Fig. 3, 3ZF-1-VP enhanced drug resistance to cisplatinum, a DNA-damaging drug, and vincristine, a microtubule-destabilizing drug. ATF 3ZF-1-VP did not alter susceptibility to 5-fluorouracil. The results of these experiments showed that the ATF 3ZF-1-VP protected HeLa cells not only from paclitaxel (the drug used in the selection) but also from anticancer drugs with unrelated mechanisms of action.

Our results suggested that ATF 3ZF-1-VP was able to promote multidrug resistance in a manner dependent on the linked transactivation domain. We investigated whether this same DBD linked to a transcriptional repressor, the Kruppel domain, could down-regulate the drug resistance phenotype. As shown in Fig. 3, HeLa cells transduced with ATF 3ZF-1-KD exhibited enhanced sensitivity to several drugs, particularly vincristine, cisplatinum, and 5-fluorouracil, compared with cells transduced with a control retrovirus. ATF 3ZF-1-KD did not significantly enhance sensitivity to carboplatin. This may indicate that specific mechanisms of drug resistance are associated with certain drugs.

We next attempted to determine the molecular mechanisms by which 3ZF 1-VP induces drug resistance. We first studied whether intact p53 was necessary for 3ZF-1-VP-mediated drug resistance. The p53 protein is mutated in 50% of human tumors and has been shown to play a central role in inducing growth arrest and apoptosis in response to DNA-damaging signals (11–13). Overexpression of known targets of p53, such as p21^WAF1/CIP1, is correlated with antiproliferative responses, G₀-G₁ cell cycle arrest (12, 13, 30, 31) and responsiveness to chemotherapy (14, 17, 32). The p53 pathway is inactivated in HeLa cells as the papillomavirus E6 protein binds to the p53 protein and targets it for ubiquitin-mediated degradation (33). It has been reported, however, that p53 can be reactivated in HeLa cells by expression of the papillomavirus E2 protein or by treatment with leptomycin B or actinomycin D (34).

To determine whether drug resistance mediated by this 3ZF-1-VP required p53, we first studied the effect of ATF on drug resistance in cell lines other than the HeLa. As shown in Fig. 4A, ATF 3ZF-1-VP was able to enhance paclitaxel resistance in the MDA-MB-435 cell line and in a Kaposi's sarcoma line. A mutated form of p53 is expressed in MDA-MB-435 cells (35) and the p53 pathway is inactivated by herpesvirus proteins in the Kaposi's line (36). The fact that ATF 3ZF-1-VP had some activity in cells with mutated or inactivated p53 suggests that the mechanism of ATF 3ZF-1-VP does not completely depend on p53.

This was further verified by inhibition of p53 expression using a siRNA. HeLa cells were transduced with a lentiviral vector expressing a siRNA complementary to a region of p53 mRNA (37). HeLa cells stably expressing this construct showed 90% to 95% reduction in p53 mRNA expression, compared with the parental HeLa cell line, as assessed by real-time reverse transcription-PCR quantification (Fig. 4C). HeLa cells transduced with a lentiviral vector encoding a p53 siRNA or with a control lentiviral vector that expressed a non-p53 siRNA were first treated with hygromycin to select for cells that had stably integrated vectors. These cells were next transduced with vector that allowed expression of ATF 3ZF-1-VP and were then challenged with paclitaxel. The percentage of cells surviving the drug was assessed by XTT assay. ATF expression in HeLa cells was verified by flow cytometry using GFP as a control.

As shown in Fig. 4B (top), the p53 siRNA had little effect on survival of cells as concentrations of paclitaxel were increased. This was expected because the p53 tumor suppressor is inactivated at the protein level in HeLa cells. However, when p53 siRNA was coexpressed with ATF 3ZF-1-VP, there was a significant reduction (20-40%) in drug resistance relative to cells that expressed only 3ZF-1-VP (Fig. 4B, bottom). This indicates that ATF 3ZF-1-VP activates both p53-dependent and p53-independent pathways to induce the drug-resistant phenotype. Thus, ATF 3ZF-1-VP could activate some targets involved in control of cell growth that are independent of the p53 pathway or targets downstream of p53 itself. The p53 dependence was also observed in other ATF-transduced cells, such as ATF 3ZF-1-VP, where the p53-siRNA significantly affected the ability of transduced cells to survive paclitaxel treatment. Consistent with this p53 dependence, we found that ATF-transduced cells exhibited lower proliferation rates relative to control cells (Supplementary Fig. S4).⁵

The experiments described above suggested that the selected ATF could activate p53-dependent pathways to promote survival. We then investigated the expression of genes participating in p53-dependent pathways in ATF-transduced cells by real-time expression analysis (Fig. 5A). We analyzed the expression of two targets of p53: the cyclin-dependent kinase inhibitor p21^WAF1/CIP1 and the zinc finger protein wild-type p53-induced gene 1 (wig-1). p21^WAF1/CIP1 induces cell cycle arrest and antiproliferative responses and is up-regulated in cell lines resistant to paclitaxel (14–16, 32) and in breast tumors exposed to chemotherapy (17). wig-1 is a zinc finger, RNA-binding protein that mediates antiproliferative responses dependent on p53 (38–40). We also evaluated p53 mRNA levels and mRNA levels of another important tumor suppressor p14^ARF. p14^ARF stabilizes and activates p53 by binding and neutralizing the functions of a negative regulator of p53, the protein MDM2, resulting in growth arrest (11). The transcription factor hDMP1 is a direct activator of p14^ARF (41, 42).

To quantitatively analyze RNA expression, we first transduced HeLa cells with ATF-encoding retroviruses and control viruses containing no zinc fingers. RNA was extracted from transduced cells and processed for real-time expression analysis. Data were normalized to the sample that did not express a zinc finger. As shown in Fig. 5B, some of the ATFs that induced drug resistance up-regulated genes in the p14^ARF-MDM2-p53 pathway. We did not see significant variations in MDM2 mRNA levels in ATF-transduced cells (data not shown).

The strongest activation of the p53 and p14^ARF pathways was mediated by ATF 3ZF-6-VP, 3ZF-16-VP, and 6ZF-30-VP. Because hDMP1α, the full-length isoform expressed in humans (42), initializes an upstream genetic cascade by transactivating p14^ARF (41, 42), we looked for a direct interaction of the ATF with the proximal promoter of hDMP1α. As shown in Fig. 5C, ATF 6ZF-30-VP and, to a lesser extent, 3ZF-1-VP were able to activate the proximal hDMP1α promoter in a reporter transactivation assay. This activation was specific because these ATFs were not able to transactivate the p21^WAF1/CIP1 promoter. A search for 6ZF-30-VP-binding sites in the hDMP1α promoter revealed an 18-bp sequence located 10 bp upstream of the transcription start site that matched 13 of the 18 bp in the predicted 6ZF-30-VP recognition site. Very close to that site (at position -20) was another potential 6ZF-30-VP-binding site that matched 11 of 18 bp in the predicted 6ZF-30-VP site (Fig. 5C). A deletion of the promoter region comprising these two potential binding sites totally abolished the transactivation of the promoter by 6ZF-30-VP in a luciferase assay (data not shown). These results further suggest that the hDMP1α promoter is a direct target of 6ZF-30-VP.

ATF 3ZF-1-VP weakly transactivated the hDMP1α reporter. A search for 3ZF-1-VP-binding sites revealed a potential 3ZF-1-VP site located at position -1,042. This transactivation was apparently too weak to efficiently activate the endogenous gene because no up-regulation was detected by real-time analysis. wig-1 was slightly, but significantly, up-regulated by ATF 3ZF-1-VP. This up-regulation was confirmed by DNA microarray experiments (Table 2; Supplementary Material) that showed a 3-fold increase in expression of this target in ATF 3ZF-1-VP-transduced HeLa cells compared with control cells. As shown in Table 2 (Supplementary Material), ATF 3ZF-1-VP also up-regulated p53-independent targets. This could explain the activity of this ATF in different cell lines with different drugs. Some of these targets are directly associated with inhibition of cell growth, such as the tumor suppressor retinoic acid responder-1. The up-regulation of β-tubulin may explain the enhanced resistance to the microtubule-destabilizing drugs, paclitaxel and vincristine. These drugs act by binding to β-tubulin and inducing a blockade in the cytokinesis step of mitosis, resulting in apoptosis (43).

Resistance to anticancer drugs represents a major problem for the long-term effectiveness of chemotherapy. In cancer cells, dysregulation of several different genetic cascades, including both p53- and p53-independent responses (1–3), can lead to drug resistance. Thus, novel therapeutic interventions that inhibit the mechanism of drug resistance are desired. ATF that target many pathways simultaneously would enable a multifunctional and effective regulation of the drug resistance phenotype. To obtain ATF regulators of drug resistance, we have used a genome-wide genetic approach. We delivered libraries of 3ZF and 6ZF ATF domains into a paclitaxel-sensitive cell line and then challenged the transduced population with paclitaxel. Surviving cells provided information on mechanisms of chemoresistance and the selected ATF may ultimately serve as therapeutic agents.

We reasoned that a 3ZF library with 9-bp binding sites would be able to regulate a larger number of genomic targets than the complex 6ZF library that interacts with 18-bp genomic sites. Any given 3ZF has the potential to bind ∼10,000 sites in the genome. Indeed, we found that the most effective ATF regulator isolated from our screen was a 3ZF protein, ATF 3ZF-1-VP. ATF 3ZF-1-VP was able to induce paclitaxel resistance in both p53-deficient and p53-mutant backgrounds. The additional result that siRNA-mediated knockdown of p53 in ATF 3ZF-1-VP-transduced cells diminished the drug resistance activity of this protein by 20% to 40% indicated that ATF 3ZF-1-VP used both p53-dependent and p53-indendent pathways to promote survival. The activity in absence of p53 mRNA was consistent with the activity of this protein in p53-deficient backgrounds.

We used real-time expression analysis to determine whether targets or regulators of the p53 tumor suppressor were up-regulated in ATF-expressing cells. Notably, p21^WAF1/CIP1 and wig-1 were up-regulated in several ATF-transduced cells compared with controls. p21^WAF1/CIP1 is a cyclin-dependent kinase inhibitor and its overexpression has been associated with resistance to paclitaxel (14, 32) and other anticancer drugs (15, 16). In particular, ErbB-2 overexpression in human breast carcinomas has been correlated with up-regulation of p21^WAF1/CIP1 and poor response to chemotherapy (16). wig-1 encodes a putative RNA-binding zinc finger protein activated by p53 that induces growth arrest in colony formation assays (38, 39). Other proteins up-regulated by the ATF included the tumor suppressors p14^ARF and hDMP1α, a transcription factor that directly activates p14^ARF transcription (41, 42). It has been shown that hDMP1α induces cell cycle arrest in primary mouse fibroblast in an ARF-dependent manner. p14^ARF acts to stabilize p53 at protein level by interference with MDM2 function (44). p14^ARF expression induces both p53-dependent and p53-independent antiproliferative genes. Both wig-1 and p21^WAF1/CIP1 are up-regulated by ectopic expression of p14^ARF (43). We observed specific up-regulation of hDMP1α, p14^ARF, p21^WAF1/CIP1, and wig-1 in cells transduced with ATF 3ZF-6-VP and 6ZF-30-VP.

Because hDMP1α is an upstream regulator of p14^ARF, we investigated whether these ATFs directly transactivated the hDMP1α gene promoter. We showed strong transactivation of the hDMP1α promoter by ATF 6ZF-30-VP. This transactivation was specific because this ATF was not able to activate the proximal p21^WAF1/CIP1 promoter. An analysis of the DNA sequence of the hDMP1 promoter revealed a putative 18-bp site at position -10 that matched 13 of 18 bp in the predicted binding site for ATF 6ZF-30-VP and a closely related sequence located 20 bp upstream of the first site that matched 11 of 18 bp in the predicted binding site. It is likely that activation requires occupancy of both sites to compensate for reduced binding at these imperfect sites. Consistent with activation of p53 genes, several of the selected ATF decreased the proliferation rate of HeLa cells and induced G₀-G₁ cell cycle arrest.

Treatment with microtubule-stabilizing drugs, such as paclitaxel or vincristine, the topoisomerase II drug, etoposide, or DNA-damaging drugs results in a G₂-M arrest and induction of apoptosis (14, 44). Paclitaxel acts by binding β-tubulin and preventing microtubule depolymerization and results in incomplete mitosis and induction of apoptosis (43). Cyclin-dependent kinases play an essential role in controlling cell cycle transitions (45). One possible mechanism used to escape the mitotic arrest and induction of apoptosis mediated by anticancer drugs is to dysregulate the cell cycle checkpoints to prevent or delay the entry to mitosis. In breast cancer, ErbB-2 up-regulation is associated with enhanced p21^WAF1/CIP1 expression (16), inhibition of the cyclin p34^cdc2 (16), and paclitaxel resistance (45). Overexpression of ErbB2 oncoprotein also results in arrest in the G₀-G₁ phases of the cell cycle (16). In light of this, it may be that the selected ATF cause dysregulation of the cell cycle leading to G₀-G₁ arrest.

DNA microarray analysis of ATF 3ZF-1-VP, our most effective regulator, showed regulation of p53-dependent genes, such as wig-1, and p53-independent genes with different mechanisms of action, such as β-tubulin and tumor suppressor genes. Our array analysis also revealed specific regulation of important signaling molecules involved in tumor progression, such as ephrin-A4 (46). The multigene nature of the ATF 3ZF-1-VP-mediated transcriptional response probably reflects the effectiveness of this ATF in regulating resistance to many drugs, including paclitaxel, etoposide, vincristine, cisplatinum, and carboplatin.

Importantly, we found that linkage of the DBD of ATF 3ZF-1-VP with a transcriptional repression domain (Kruppel domain) resulted in down-regulation of the drug resistance phenotype, particularly for the microtubule-destabilizing drug vincristine and the DNA-damaging drugs cisplatinum and 5-fluorouracil. Increased sensitivity to 5-fluorouracil, Tomudex, and Taxol has been described by Dolnick et al. by treating colon cancer cells with 3-oxododecanoyl homoserine lactone [3-oxo-C12-(L)-HSL; ref. 47]. The effects of this compound were associated with changes in microtubule and cytoskeleton proteins (α-tubulin and β-actin expression). More experiments are required to determine which genes are targeted by the repressor ATF.

Our observations raise the possibility that ATF might be used not only to block activation of multiple pathways associated with drug resistance but also to sensitize cells to chemotherapy treatments. In conclusion, our data show that ATF can be used to regulate drug resistance by activation of p53-dependent and p53-independent pathways. Furthermore, our analysis proves that ATF libraries, in combination with siRNA technology, can be used to precisely map promoters, genes, and pathways involved in drug resistance.

We thank Dave Valente for technical support and Drs. K.G. Wiman and C. Prives for kindly providing wig-1 expression and WAF1/CIP1 promoter plasmids, respectively.