The Cyclin-Dependent Kinase Inhibitor Flavopiridol Potentiates the Effects of Topoisomerase I Poisons by Suppressing Rad51 Expression in a p53-Dependent Manner

The results of a phase I clinical trial of the topoisomerase I (Topo I) poison CPT-11 followed by the cyclin-dependent kinase inhibitor flavopiridol in patients with advanced solid tumors indicate that patients whose tumors were wild-type, but not mutant, for p53 obtained the most clinical benefit from this combination therapy. We elected to elucidate the mechanistic basis for this effect in isogenic-paired HCT116 colon cancer cells that were either wild-type (+/+) or null (−/−) for p53. With the combination therapy of SN-38 (the active metabolite of CPT-11) followed by flavopiridol, the induction of apoptosis was 5-fold greater in the p53+/+ cells compared with the p53−/− cells. This sequential treatment induced phosphorylation of p53 at Ser¹⁵, which interacted with Rad51, a DNA repair protein involved in homologous recombination. Rad51 bound to p53-Ser¹⁵ within the first 5 hours of combination therapy, and then was transcriptionally suppressed at 24 hours by flavopiridol only in p53+/+ cells. Microarray analysis also revealed suppression of Rad51 in a p53-dependent manner. Depletion of Rad51 by small interfering RNA (siRNA) sensitized both p53+/+ and p53−/− cells to SN-38–induced apoptosis with increase of γH2AX, a marker of DNA damage. Conversely, overexpression of Rad51 rescued p53+/+ cells from SN→F–induced apoptosis. Because flavopiridol inhibits Cdk9, we found that inhibition of Cdk9 by DRB or by siRNA could recapitulate the flavopiridol effects, with suppression of Rad51 and induction of apoptosis only in p53+/+ cells. In conclusion, after DNA damage by Topo I poisons, flavopiridol targets homologous recombination through a p53-dependent down-regulation of Rad51, resulting in enhancement of apoptosis. [Cancer Res 2008;68(7):2312–20]

SN-38 (the active metabolite of CPT-11) stabilizes the topoisomerase I (Topo I)–DNA complex, and its collision with the DNA replication fork leads to the generation of permanent strand breaks and G₂ arrest. Flavopiridol, a pan–cyclin-dependent kinase inhibitor, has been reported to induce DNA damage (1) and inhibit tumor cell growth through cell cycle arrest at G₁ or G₂ (2). We have previously reported that flavopiridol at noncytotoxic, nanomolar concentrations significantly enhances the induction of apoptosis by the chemotherapeutic agent mitomycin C and paclitaxel in gastric and breast cancer cell lines (3, 4). Also, the addition of flavopiridol to HCT116 cells enhanced the induction of apoptosis by SN-38 and increased the efficacy of CPT-11 in HCT116-treated xenografts (5). These events were highly sequence dependent such that SN-38 or CPT-11 needed to be administered before flavopiridol both in vitro and in vivo to induce this effect.

Based on this preclinical data, a phase I clinical trial of CPT-11 followed by flavopiridol was initiated. This study indicated that patients whose tumors were wild-type, but not mutant, for p53 obtained the most clinical benefit from this combination therapy (6). In view of this, we elected to determine the role of p53 in the enhancement of SN-38–induced apoptosis by flavopiridol. After DNA damage, the level of p53 increases and induces a range of responses, including cell cycle arrest, apoptosis, senescence, and DNA repair. This depends on the type and dose of the genotoxic stress. Recent studies have shown that p53 represents a central player in DNA repair by repressing homologous recombination, counteracting error-prone double-strand breaks, and promoting genomic stability (7). Activation of p53 is also mediated by posttranslational modifications by specific kinases, such as the ataxia-telangiectasia mutated (ATM) and the ATM and Rad3 related (8, 9). When DNA replication is blocked, p53 becomes phosphorylated at Ser¹⁵ and associates with key enzymes of homologous recombination, such as the MRN complex, Rad51 and Rad54 (10, 11). Rad51 is central to homologous recombination–mediated double-strand break repair, by forming nucleoprotein filaments and mediating strand exchange between DNA duplexes (12, 13). Aberrant amounts of Rad51 have been observed in a number of transformed cell types (14) and may induce malignant transformation. Biochemical investigations showed that p53 inhibits Rad51-mediated strand exchange in vitro (11, 15, 16). Furthermore, p53 modulates homologous recombination by the transcriptional repression of the RAD51 gene (17). Therefore, we elected to determine whether the enhancement of SN-38–induced apoptosis was p53 dependent and the role of Rad51 in this process.

Cell culture. HCT116 human colon carcinoma cells and its p53-null variant were a gift from Dr. Vogelstein (Johns Hopkins University, Baltimore, MA). Cells were grown in McCoy's 5A medium supplemented with 10% fetal bovine serum. SN-38 (5 mmol/L stock, Pharmacia) was prepared in DMSO. Flavopiridol (4.5 mmol/L) was prepared in water. z-VAD was purchased from MBL International. Cycloheximide and Lnll were from Sigma Aldrich.

Apoptosis assays. The measurement of apoptosis by quantitative fluorescence microscopy (QFM) with 4′,6′-diamidino-2-phenylindole (Sigma Chemical Co.) for nuclear chromatin was performed as previously described (3). Apoptosis was also measured using Annexin V–FITC Apoptosis detection Kit II (BD PharMingen) following the manufacturer's instructions. The statistical significance of the experimental results was determined by the two-sided t test.

Immunoblotting. Cells were lysed in radioimmunoprecipitation assay buffer supplemented with protease inhibitor (Complete Mini, Roche Diagnostics) and 1 mmol/L NaVO₃. Equal amounts of protein were loaded on 4% to 12% PAGE gels (Invitrogen). The membranes were probed with antibodies for p53 (Santa Cruz Biotechnology, Inc.), Rad51 (Calbiochem), p53-Ser¹⁵ (Cell Signaling), and α-tubulin (Upstate Biotechnology). Equivalent amounts of proteins from the cleared lysates were incubated with Rad51 antibody for 3 h at 4°C. Then, 30 μL of agarose beads (Upstate Biotechnology) were added for an additional hour. Immunocomplexes were washed in lysis buffer and suspended in 4× SDS sample buffer for Western blot analysis.

Quantitative real-time PCR. RNA was extracted with TriPure Isolation Reagent (Roche Diagnostic). Reverse transcription of 1 μg of RNA was done using the SuperScript III First-Strand Synthesis System (Invitrogen) following the manufacturer's instructions. All quantitative real-time reverse transcription PCR assays were done on the 7300 Real Time PCR System (Applied Biosystems). TaqMan gene expression assays (Applied Biosystems), which include gene-specific probe and primer sets, were used to detect both Rad51 and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA. The relative expression of Rad51 was calculated by normalizing to GAPDH mRNA. Graphs are the means of three separate experiments.

Microarray data analysis. cDNA was synthesized in the presence of oligo(dT)24-T7 from Genset Corp. cRNA was prepared using biotinylated UTP and CTP and was hybridized to HG U113A oligonucleotide arrays (Affymetrix, Inc). Data preprocessing and statistical analysis were carried out in R³

and Bioconductor.⁴ The expression intensities were normalized using the Robust Multiarray Average method (18), which includes (a) background adjustment, (b) quantile normalization across arrays, and (c) probe-level expression measure summarization using median polish on the log 2 scale for each probe set. Differential expression analysis was performed to identify genes that were differentially expressed between SN alone–treated versus SN→F–treated samples in the parental cell line and in the p53-deficient cell line, respectively. An empirical Bayes t test was applied to each gene (19). This test is a modified version of the conventional t test and is more suitable for data with small sample size. A P value cutoff of 0.001 was used to select differentially expressed genes (P ≤ 0.001), whereas a P value cutoff of 0.50 was used to select nondifferentially expressed genes (P ≥ 0.50). Unsupervised hierarchical clustering was done to look for samples and genes with similar gene expression patterns. The samples were clustered using the Euclidean distance as the dissimilarity measure and the Ward function as the joining method; the genes were clustered using the Euclidean distance and the Complete function. The results of the cluster analysis were displayed a heat map.

Plasmid and small interfering RNA transfections. Rad51 full-length cDNA was cloned from HCT116 cells using the Titan One Step reverse transcription-PCR Kit (Roche Applied Science) and inserted in the mammalian expression vector pcDNA3.1 (Invitrogen). The Rad51/P186 mutant (11) was generated using the QuikChange II Site-Directed Mutagenesis Kit (Stratagene). Cdk9 (SMART-Pool) and Rad51 small interfering RNA (siRNA) were obtained from Dharmacon, Inc. The sequence for Rad51 siRNA was GAGCUUGACAAACUACUUC. siRNAs were transfected using OligofectAMINE reagent (Invitrogen).

Flavopiridol enhances SN-38–induced apoptosis in a p53-dependent manner. Human colon carcinoma HCT116 cells with wild-type p53 (p53+/+) and its p53 null isogenic sub–cell line (p53−/−) were treated with 20 nmol/L SN-38 and 300 nmol/L flavopiridol individually and sequentially, as SN-38 for 24 hours (SN₂₄) followed by either flavopiridol (SN₂₄→F₂₄) or drug-free medium for 24 hours (SN₂₄→ND₂₄). It has been shown that under these treatment conditions, SN-38 induces DNA damage and a potent G₂ cell cycle arrest in both cell lines at 24 hours, with p53 and p21 up-regulation in the p53+/+ cells (5) and activation of chk1 kinase in the p53−/− cells (20). As determined by QFM (Fig. 1A), SN-38 or flavopiridol alone did not induce significant apoptosis. The sequential treatment of p53+/+ cells with SN₂₄→F₂₄ induced apoptosis in 23.5 ± 2.1% of the cells, which was significantly greater than that observed for SN₂₄→ND₂₄ (3.8 ± 1.2%). However, the SN₂₄→F₂₄ treatment induced 4.7 ± 1.1% apoptosis in p53−/− cells. This was further evaluated with Annexin V+ staining (Fig. 1B). With this method, 18% to 20% of both p53+/+ and p53−/− cells became positive for Annexin V after SN-38 treatment. With the addition of flavopiridol, Annexin V+ increased to 37.6% in the p53+/+ cells (P < 0.001 versus SN-38 alone–treated cells), whereas there was no significant change in the p53−/− cells. Thus, the enhancement of SN-38–induced apoptosis in G₂-arrested cells by flavopiridol seems to be p53 dependent.

In view of this, we investigated the effects of SN-38 and flavopiridol on p53 induction in the p53+/+ cells. As shown in Fig. 2A, both conditions up-regulated p53 protein levels over time. In contrast, with SN-38 for 24 hours followed by drug-free medium for up to 24 hours (SN₂₄→ND₂₄), there was a time-dependent decrease in p53 expression. However, the sequential treatment of SN-38 for 24 hours followed by flavopiridol for 3 to 24 hours (SN₂₄→F_x) induced a time-dependent increase in total p53, as well as the induction of phosphorylated p53 at Ser¹⁵. Under these conditions, Rad51 was induced by SN-38 at 24 hours, which was sustained when the cells were then changed to drug-free medium. In contrast, Rad51 was down-regulated by flavopiridol, both as a single agent and in sequential treatment with SN-38 (SN₂₄→F_x).

Phosphorylated p53 at Ser¹⁵ has been found to interact with Rad51 and inhibit its activity during DNA repair (10). Interestingly, coimmunoprecipitation experiments (Fig. 2B,, right) showed that p53 bound to Rad51 after 5 hours of the sequential SN→F₅ treatment, but not with SN-38 alone for 24 hours (SN→F₀). This interaction was sustained for the next 16 hours, until Rad51 was down-regulated at 24 hours by flavopiridol (Fig. 2B).

Flavopiridol down-regulates Rad51 in a p53-dependent manner. Thus, it would seem that with this sequential therapy at early time points, Rad51 binds to p53-Ser¹⁵; however Rad51 expression is suppressed over time. Rad51 is an important regulator of homologous recombination and is reportedly inhibited by p53, not only through physical interaction but also by transcriptional repression (17). Thus, we tested whether flavopiridol could differentially regulate Rad51 depending on the p53 status of the cells. HCTp53+/+ and p53−/− cells were treated with SN-38 for 24 hours followed by flavopiridol for the indicated times. Rad51 declined in p53+/+ cells, with almost complete down-regulation after 20 hours of flavopiridol treatment (Fig. 3A). In contrast, Rad51 continued to be expressed over time in p53−/− cells. In addition, γ-H2AX showed a time-dependent increase only in the p53+/+ cells, consistent with an increase in double-strand DNA breaks.

To extend these findings for the p53 dependency of Rad51 suppression, we tested the effect of SN-38 and flavopiridol on a panel of cell lines that differ in their p53 status. As shown in Fig. 3B, the suppression of Rad51 with the sequential therapy of SN→F was observed only in cell lines that are wild-type for p53, including gastric cancer (MKN-74) and melanomas (M173 and M19). Still, no suppression of Rad51 was observed in three colon cancer cell lines (H747, HT29, and SW620) that are mutant for p53. This is in contrast to the HCTp53−/− cells, which exhibited a decrease in Rad51 expression at later time points, suggesting slight differences in cell-specific responses to flavopiridol among p53 mutant cell lines. The suppression of Rad51 expression in the p53 wild-type cell lines again correlated with the enhancement of SN-38–induced apoptosis by QFM (Fig. 3C). In contrast, there was no increase in apoptosis with SN→F in the p53 mutant cell lines. In control experiments, to exclude proteasome- and caspase-mediated effects on Rad51 down-regulation, we used the pan-caspase inhibitor z-VAD or the proteasome inhibitor Lnll together with the SN→F treatments. Under these conditions, the time-dependent suppression of Rad51 by SN→F in the p53+/+ cells could not be restored by the addition of either compounds (Fig. 3D). Furthermore, Rad51 protein stability was assessed in both cell lines by adding cycloheximide to the SN→F treatments, and we found no difference in protein decay by 6 hours (Fig. 3E).

Next, we compared Rad51 mRNA levels by real-time PCR in both cell lines. The addition of flavopiridol to the SN-38–treated cells induced significant reductions in Rad51 mRNA in the p53+/+ cells over time (Fig. 4A). In contrast, Rad51 mRNA levels did not decrease significantly with flavopiridol in the p53−/− cells. To further investigate the effect of flavopiridol on the expression of p53-dependent genes, we performed microarray analysis, using Affymetrix U133A GeneChip arrays, to identify genes that were differentially expressed between SN-38 alone versus SN→F–treated samples in the p53+/+ cells, but not in the p53−/− cell line. There are 116 genes that are down-regulated by flavopiridol in the HCTp53+/+ cell line but unaffected in the HCTp53−/− cells, which we will call “differentially suppressed genes” (Supplementary Data). Among the 116 genes, an unsupervised hierarchical clustering was performed to look for samples and genes with similar expression patterns (Fig. 4B). Some of these differentially expressed genes encode proteins critical for cell cycle (Cdk2, cyclin D3, p19), regulation of transcription (GTF2A1, NR3C1, POLR2L, MED12, ZNF337, DNAJC1), and regulation of apoptosis (FAIM3). We also found significant suppression of two genes involved in DNA repair, one of them being Rad51. The other is HMGB1, a member of the large family of HMG-box proteins, which has been associated with mammalian excision base repair in the setting of oxidative stress (21). Taken together, these results suggest that Rad51 is down-regulated by flavopiridol in a p53-dependent manner. Furthermore, a survey of genes indicates that Rad51 is the only gene suppressed in a p53-dependent manner that has been directly linked to the process of homologous recombination.

The suppression of Rad51 is, in part, mediated by inhibition of Cdk9. Flavopiridol is a potent inhibitor of Cdk9, which phosphorylates the COOH-terminal domain of RNA polymerase II and promotes initiation and elongation of mRNA transcripts (22). To address the possibility that Cdk9 inhibition by flavopiridol may directly suppress the expression of Rad51, we used a specific Cdk9 inhibitor, DRB, and a specific Cdk9 siRNA in HCTp53+/+ and HCTp53−/− cells. Both DRB (Fig. 5A) and Cdk9 depletion by siRNA (Fig. 5C) mimicked flavopiridol effects in SN-38–treated p53+/+ cells, with down-regulation of Rad51, induction of p53, and increase in apoptosis (Fig. 5A–D). In p53−/− cells, however, Rad51 did not change after Cdk9 down-regulation by siRNA (Fig. 5C), slightly decreased only at later time points with SN→DRB treatment (Fig. 5A), and demonstrated no significant increase in apoptosis (Fig. 5B and D). These results suggest that down-regulation of Rad51 by flavopiridol is, at least in part, due to the inhibition of Cdk9, and that the induction of apoptosis under these treatment conditions still requires the presence of p53.

Depletion of Rad51 sensitizes HCT116 cells to single-agent SN-38. Next, we tested whether the depletion of Rad51 would recapitulate the effect of flavopiridol, in the presence or absence of p53. Both cell lines were transfected with a specific Rad51 or control siRNA and then treated with SN-38. As shown in Fig. 6A, Rad51 was completely down-regulated in both p53+/+ and p53−/− cell lines after siRNA transfection. There was evidence of induction of γH2AX (Fig. 6A) and low levels of spontaneous apoptosis in both cell lines (Fig. 6B) when compared with siRNA controls. Treatment with SN-38 for 24 hours (SN₂₄) induced 23.7 ± 1.7% apoptosis in p53+/+ cells and 19.2 ± 0.3% in p53−/− cells (Fig. 6B). These results suggest that Rad51 plays a role in the protection from apoptosis to single-agent SN-38 treatment in both cell lines.

Overexpression of Rad51 results in resistance to SN→F according to p53 status. The role of Rad51 in SN→F–induced apoptosis was further investigated by generating Rad51-overexpressing cells. In these experiments, we also used the mutant Rad51/186P (11), which lacks the ability to bind p53. The constructs were transfected in p53+/+ cells 24 hours before SN→F treatment. Rad51 protein levels are shown in Fig. 6C. Overexpression of Rad51 and Rad51/P186 decreased the percentage induction of apoptosis by SN→F from 21 ± 1% of vector transfectants to 11 ± 1% and 8 ± 1%, respectively (Fig. 6D). This also corresponded to a decrease in γH2AX expression, demonstrating that with SN→F, Rad51 overexpression decreased DNA damage, especially when it could not bind to p53. Taken together, these results indicate that Rad51 plays a central role in the resistance to SN-38 and that flavopiridol induces a p53-dependent suppression of Rad51 that results in induction of apoptosis.

Homologous recombination mediates the error-free repair of DNA, mostly during the late S-G₂ phases of the cell cycle (23, 24). In the presence of high doses of DNA-damaging agents or defects in the DNA repair machinery, the repair is unsuccessful and this may lead to permanent cell cycle arrest or apoptosis. Our studies indicate that flavopiridol converted Topo I–induced cell cycle arrest into apoptosis through a p53-dependent transcriptional suppression of Rad51. This seems to involve two steps. First, with sequential SN→F treatment, there is early binding of p53-Ser¹⁵ to Rad51, which seems to be sufficient to inhibit the process of Rad51-mediated DNA repair that occurs after SN-38–induced DNA damage. In fact, mutation of Rad51 at its binding domain to p53 significantly prevented the induction of apoptosis with the sequential therapy. This is then followed by a second event of down-regulation of Rad51 at the mRNA level.

Our studies indicate that this effect is not due to caspase- or proteasome-mediated degradation of Rad51 or due to differences in protein stability. This transcriptional suppression of Rad51 with sequential SN→F and the resulting induction of apoptosis are absolutely dependent on the presence of wild-type p53. This p53-dependent effect on Rad51 suppression was further confirmed using an unsupervised hierarchical clustering of gene expression profiles. Among 116 genes suppressed by flavopiridol on SN-38–treated HCTp53+/+ cells, only two genes involved in DNA repair were identified, and one of them was Rad51. Although these results do not exclude the effect of the transcriptional suppression of these other genes in this combination therapy, it indicates that when flavopiridol is combined with a topoisomerase poison the suppression of Rad51 does play a significant role in the enhancement chemotherapy-induced apoptosis.

Flavopiridol has been implicated in the transcriptional repression of numerous genes (25), including the antiapoptotic genes Mcl-1 (26), the X-linked inhibitor of apoptosis protein (XIAP), and c-Kit (27). This effect is mediated by inhibition of the Cdk9/pTEFb complex, which regulates RNA polymerase II (21). These effects have not being reported as p53 dependent. In fact, our microarray analysis of genes suppressed in a p53-dependent manner with SN→F did not indicate differential suppression of Mcl-1 or any of the Bcl-2 family members. Using siRNA against Cdk9 and DRB to target Cdk9 with SN-38, we show suppression of Rad51 in the p53+/+ cells. This would suggest that suppression of Rad51 by Cdk9 inhibition is a p53-dependent process and that flavopiridol represents a novel therapeutic approach in the targeting of this gene.

In the p53+/+ cells, both SN-38 and flavopiridol are shown to up-regulate p53. However, it is only after SN→F treatment that we observe sustained p53 phosphorylation. It has been shown that p53 can reduce the activity of a variety of promoters (28), including those of c-fos (29) and Rb (30). Posttranslational modifications of p53, such as phosphorylation and acetylation, are involved in the modulation of p53 transcriptional activities (31, 32). This increases the possibility that p53 in its phosphorylated state may be playing a role in the transcriptional suppression of Rad51 under these treatment conditions.

Recently, inhibition of cyclin-dependent kinases has been implicated in the reduction of DNA repair mechanisms. For example, the pan–cyclin-dependent kinase inhibitor roscovitine sensitized tumor cell lines to doxorubicin by increasing the amount of double-strand DNA breaks and reducing the efficiency of homologous recombination repair (33). This effect seemed to be predominant in the G₂-M phase of the cell cycle, where homologous recombination was preferentially active. Although DNA-PK activity was suppressed by roscovitine, the effect of roscovitine on Rad51 was not explored.

Rad51 activity is essential during homologous recombination–mediated double-strand break repair, either during replication (34) or following exposure to DNA-damaging agents. Its central role in homologous recombination is supported by studies with Rad51 knockout mice, which showed early embryonic lethality (35). Furthermore, it has been reported that down-regulation of Rad51 enhanced the anticancer activities of cisplatin (36) and increased the radiosensitivity of tumor cells (37). Conversely, a number of studies showed that high levels of Rad51 caused resistance to a variety of chemotherapeutic agents like doxorubicin (38) and 5-fluorouracil (39). We also found that overexpression of Rad51 significantly protected p53+/+ cells from SN→F–induced apoptosis.

For flavopiridol, the targeting of Rad51 seems to be critical for the increase in DNA damage and the enhancement of SN-38–induced apoptosis. This observation and a mechanistic understanding of this process should have significant implications for the development of this agent in solid tumors. Currently, flavopiridol has been shown to have single activity in chronic lymphocytic leukemia, which is believed to be mediated by the flavopiridol targeting of Cdk9 and the suppression of Mcl-1 and bcl-2 family members (40). However, the experience with single-agent flavopiridol in the treatment of solid tumors has been generally disappointing (41, 42). In contrast, clinical activity of flavopiridol in combination with chemotherapy has been observed (6, 43, 44). The ultimate success of this agent will depend on the identification of subpopulation of patients who will most benefit from the combination with chemotherapy. A clinical trial to test this hypothesis with sequential CPT-11 and flavopiridol in patients with gastroesophageal cancer who are wild-type p53 is now being planned, and suppression of Rad51 in serial tumor biopsies will be examined. Therefore, based on these laboratory studies, it is our hope that the targeting of Rad51 with Topo I inhibitors followed by flavopiridol will represent a new approach for the treatment of solid tumors that are wild-type p53.

Grant support: R01 CA67819.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.