Abstract

Transcription factors are direct effectors of altered signaling pathways in cancer and frequently determine clinical outcomes in cancer patients. To uncover new transcription factors that would determine clinical outcomes in breast cancer, we systematically analyzed gene expression data from breast cancer patients. Our results revealed that Forkhead box protein M1 (FOXM1) is the top-ranked survival-associated transcription factor in patients with triple-negative breast cancer. Surprisingly, silencing FOXM1 expression led breast cancer cells to become more sensitive to doxorubicin (Dox). We found that FOXM1-dependent resistance to Dox is mediated by regulating DNA repair genes. We further demonstrated that NFκB1 interacts with FOXM1 in the presence of Dox to protect breast cancer cells from DNA damage. Finally, silencing FOXM1 expression in breast cancer cells in a mouse xenograft model significantly sensitized the cells to Dox. Our systematic approaches identified an unexpected role of FOXM1 in Dox resistance by regulating DNA repair genes, and our findings provide mechanistic insights into how FOXM1 mediates resistance to Dox and evidence that FOXM1 may be a promising therapeutic target for sensitizing breast cancer cells to Dox.

Introduction

Transcription factors (TFs) are regulatory endpoints of signaling pathways, and their altered activities result in the development of many human diseases, including cancer ( 1 , 2 ). TFs function as oncogenes or tumor suppressors in diverse human cancers ( 2 , 3 ). Many TFs are frequently overexpressed in human cancers and determine clinical outcomes in cancer patients ( 1 , 2 ). For example, expression levels of MYC , JUN and STAT3 are significantly associated with unfavorable clinical outcomes in patients after treatment ( 4–8 ). In breast cancer, the function of TFs dictating patient prognosis has been well known ( 9 , 10 ). Expression level of estrogen receptor (ESR1), GATA3 and FOXA1 is directly correlated with patient overall survival because those genes influence endocrine therapy and chemoresponse effect ( 9 , 10 ). Furthermore, TF ESR1 and progesterone receptor ( PGR ) are a good classifier discriminating molecular subtype of breast cancer ( 9 , 10 ). The Human Genome Project has estimated that more than 3000 TFs exist ( 1 ). As many of them are expected to be involved in the development of cancer and thus are potential therapeutic targets, it is important to develop a strategy that can easily identify oncogenic TFs as potential therapeutic targets.

In the current study, using systematic analysis, we identify that Forkhead box protein M1 (FOXM1) is a poor prognosis TF conferring Dox resistance in breast cancer.

Materials and methods

Selection of genes significantly associated with overall survival of breast cancer patients

To avoid including too many potential false positives, we used two independent data sets of breast cancer patient [the Netherlands Cancer Institute (NKI) ( 9 ) and University of North Carolina (UNC)] ( 10 ) cohorts containing both genome-wide expresssion data and patient survival data, and we analyzed each data set independently. We first dichotomized patients according to expression levels of individual genes (high/low) in microarray data by using the median as the cutoff. We estimated the prognostic significance of individual genes by applying Kaplan–Meier analysis and the log-rank test to dichotomized patients. The association of a gene with overall survival was considered significant if P < 0.01 by the log-rank test. When this criterion was applied to the two independent data sets, only 104 genes remained significant in both patient cohorts and were thus considered to be survival-associated genes ( Figure 1A and Supplementary Figure 1 is available at Carcinogenesis Online).

Fig. 1.

FOXM1is an oncogenic transcription factor in breast cancer. (A) Venn diagram representing the number of genes significantly associated with overall survival in two independent cohorts ( P < 0.01 by log-rank test); 104 genes were shared by the two cohorts and were thus considered to be SAGs. (B) Number of genes whose expression is significantly correlated ( P < 0.001) with that of 11 TFs in the NKI cohort. Pearson’s correlation coefficient was used to estimate correlation. (C and D) Patients in the NKI and UNC cohorts were stratified into the five subtypes according to intrinsic gene expression signature in previous studies ( 23 ) and box plots used to visualize FOXM1. Statistical significance was estimated by ANOVA. HER2, human epidermal growth factor receptor 2; ER, estrogen receptor. (E and F) Receiver operating characteristic curve analysis of FOXM1 expression probability over 3-year recurrence of breast cancer in the NKI and UNC cohorts. Receiver operating characteristic curve analysis was performed to evaluate the association of FOXM1 gene expression levels with recurrence by determining the area under curve (AUC) estimated through the concordance index. The corresponding P valuesss are from the one-sided Wilcoxon’s rank test (CI, confidence interval).

Fig. 1.

FOXM1is an oncogenic transcription factor in breast cancer. (A) Venn diagram representing the number of genes significantly associated with overall survival in two independent cohorts ( P < 0.01 by log-rank test); 104 genes were shared by the two cohorts and were thus considered to be SAGs. (B) Number of genes whose expression is significantly correlated ( P < 0.001) with that of 11 TFs in the NKI cohort. Pearson’s correlation coefficient was used to estimate correlation. (C and D) Patients in the NKI and UNC cohorts were stratified into the five subtypes according to intrinsic gene expression signature in previous studies ( 23 ) and box plots used to visualize FOXM1. Statistical significance was estimated by ANOVA. HER2, human epidermal growth factor receptor 2; ER, estrogen receptor. (E and F) Receiver operating characteristic curve analysis of FOXM1 expression probability over 3-year recurrence of breast cancer in the NKI and UNC cohorts. Receiver operating characteristic curve analysis was performed to evaluate the association of FOXM1 gene expression levels with recurrence by determining the area under curve (AUC) estimated through the concordance index. The corresponding P valuesss are from the one-sided Wilcoxon’s rank test (CI, confidence interval).

Quantitative reverse transcription–polymerase chain reaction

Total RNA was extracted from the indicated cell lines by using a mirVana RNA isolation labeling kit (Ambion) according to the manufacturer’s instructions, and RT–PCR was assayed by using real-time quantitative reverse transcription–polymerase chain reaction (qRT–PCR) with TaqMan primers specific to each gene (Applied Biosystems). Real-time PCR was performed using the 7700HT real-time PCR system with a 96-well block module (Applied Biosystems). Cycling conditions were 45°C for 30min and 95°C for 10min, followed by 40 cycles of 95°C for 15 s and 60°C for 60 s. Relative amounts of mRNA were calculated from the threshold cycle number using expression of cyclophilin A as an endogenous control. All experiments were performed in triplicate and the values averaged.

Cell culture and chemotherapeutic agents

Breast cancer cell lines and U2OS cells were purchased from American Type Culture Collection. MDA-MB-231 cells were grown in RPMI 1640 medium supplemented with 10% fetal bovine serum (FBS). MCF-7 cells were maintained in Dulbecco’s modified Eagle medium/nutrient mixture F-12 supplemented with 10% FBS. U2OS cells were maintained in McCoy’s 5A medium supplemented with 10% FBS with glutamine, penicillin and streptomycin. Cells were incubated at 37°C in a humidified incubator with 5% CO 2 . Paclitaxel, Dox, cisplatin, etoposide and mitoxantrone were dissolved in dimethyl sulfoxide (DMSO). 5-FU was dissolved in sterile water. All the chemotherapeutic agents were obtained from Sigma.

Plasmids and luciferase assay

pCDNA3NFκB1 plasmid was purchased from Addgene. pGL3- PLK4 promoter was described previously ( 11 ). To generate pGL3- EXO1 and pGL3- RFC4 , the promoter region from −2000 to +1 was cloned by PCR from human genomic DNA. Sequences were verified by automatic sequencer. For luciferase-based reporter assays, cells were transfected with indicated reporter genes and plasmids using Fugene6 (Roche) according to the manufacturer’s instructions. After 48h, cells were harvested to measure luciferase activity, which was normalized with β-galactosidase.

Western blotting

Cells were maintained and western blotting was performed as described previously ( 18 ). Antibodies used were anti-FOXM1 (sc-500; Santa Cruz Biotechnology), NFκB1 (#3035; Cell Signaling Technology), β-actin (H-196; Santa Cruz) and α-tubulin (#3873; Cell Signaling).

Small interfering RNA

siLuc, siFOXM1, siEXO1, siCDC45L, siRFC4 and siPLK4 were described previously ( 12–16 ). SiPOLE2 and siNFκB1 were purchased from Ambion. Cells were transfected with indicated small interfering RNA (siRNA) using Oligofectamine (Invitrogen) for 48 or 72h in the presence of indicated chemotherapy drugs.

3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay

The 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay was used to assess cell proliferation after drug treatment with or without specific siRNA. Cells were plated into 96-well plates, transfected with specific siRNA and incubated for 36 h. Cells were then exposed to indicated chemotherapy drugs at indicated concentrations for 36 h. After treatment, 20 µl of MTT was added to each well, the wells were incubated for 3 h and 100 µl of DMSO was then added. The absorbance of individual wells was read at 570nm.

Chromatin immunoprecipitation assay

Chromatin immunoprecipitation (ChIP) experiments were performed as described previously ( 18 ). Antibodies used were anti-FOXM1 (K-19) and rabbit normal IgG (SC-2027), both from Santa Cruz Biotechnology. Primer sequences used are listed in Supplementary Table 4 is available at Carcinogenesis Online.

Microarray

Total RNA was extracted from MDA-MB-231 cells by using a mirVana RNA isolation labeling kit (Ambion). We used 500ng of total RNA for labeling and hybridization, according to the manufacturer’s protocols (Illumina). After the bead chips were scanned with a BeadArray Reader (Illumina), the microarray data were normalized using the quantile normalization method in the Linear Models for Microarray Data (LIMMA) package in the R language environment ( 20 ). The expression level of each gene was log2-transformed before further analysis. Primary microarray data are available in the National Center for Biotechnology Information Gene Expression Omnibus public database (microarray platform, GPL6947; microarray data, GSE25741).

Homologous recombination assay

The homologous recombination (HR) assay was described previously ( 17 , 18 ). Briefly, U2OS cells stably expressing DR (tandem repeated) GFP reporter were transfected with siFOXM1 or siLuc. After I- Sce I endonuclease was delivered to the cells by transient transfection, GFP-positive cells were sorted and analyzed by flow cytometry to indicate the efficiency of HR repair.

Stable isotope labeling by amino acids in cell culture method

Stable isotope labeling by amino acids in cell culture (SILAC) kits were obtained from Invitrogen. As described previously ( 19 ), MDA-MB-231 cells were grown in RPMI 1640 medium lacking arginine and lysine supplemented with dialyzed FBS. The medium was supplemented with heavy 13 C 6 - l -arginine and 13 C 6 - l -lysine or with 12 C 6 - l -arginine and 12 C 6 - l -lysine. Cells were passaged five times in ‘heavy’ or ‘light’ SILAC medium to ensure complete labeling. Then, MDA-MB-231 cells labeled with light medium were exposed to 5 µM Dox for 36h. Lysates from MDA-MB-231 cells labeled with the heavy and light medium were mixed together, purified and immunoprecipitated with FOXM1 antibody. The immunoprecipitates were washed thrice with NETN buffer (20mM Tris–HCl [pH 8.0], 100mM NaCl, 1mM EDTA, 0.5% NP40). The washed beads were boiled with Laemmli buffer and subjected to SDS-PAGE (4–20% Tris/glycine Novex gel; Invitrogen). The Coomassie brilliant blue-stained protein bands were excised and destained with 50mM ammonium bicarbonate solution in 50% methanol. Gel pieces were then washed in HPLC water overnight. After washing, gel pieces were digested with 100ng of trypsin in 50mM NH 4 HCO 3 (pH 8.5) for 4h in 15 µl. After digestion, peptides were extracted by adding 200 µl of acetonitrile. The supernatants were dried in a SpeedVac dryer.

Protein identification (LC-MS/MS analysis)

Each dried sample was dissolved in 20 µl of 5% methanol/95% water/0.01% formic acid solution and injected into the Surveyor HPLC system (Thermo Finnigan) using an autosampler. A 100mm × 75 µm C18 column (5 µm, 300 Å pore diameter, PicoFrit; New Objective) with mobile phases of A (0.1% formic acid in water) and B (0.1% formic acid in methanol) was used, with a gradient of 5–95% of phase B over 15min followed by 95% of phase B for 5min at 200 nl/min. Peptides were directly electrosprayed into the mass spectrometer (Finnigan LTQ, Thermo Finnigan) using a nanospray source. The LTQ was operated in the data-dependent mode to acquire fragmentation spectra of the 20 strongest ions.

Statistical analysis of microarray data and survival analysis

The random-variance t -test was used to identify genes differentially expressed between the two classes that were compared using BRB-ArrayTools ( 20 ). The random-variance t -test is an improvement over the standard separate t -test as it allows information to be shared among genes about within-class variation without assuming that all genes have the same variance. Gene expression differences were considered significant if P < 0.001. Kaplan–Meier analysis and the log-rank test were used to estimate patient prognosis.

Xenograft experiments

Female athymic nude mice (NCr-nu) were purchased from the NCI-Frederick Cancer Research and Development Center (Frederick, MD) and maintained as described previously ( 21 ). All mouse studies were approved by the Institutional Animal Care and Use Committee. Mice were cared for in accordance with guidelines set forth by the American Association for Accreditation of Laboratory Animal Care and the US Public Health Service Policy on Human Care and Use of Laboratory. Tumor implantation, siRNA incorporation into dioleoylphosphatidylcholine (DOPC) nanoliposomes and delivery in vivo were carried out as described previously ( 21 ). For FOXM1 silencing experiments, mice were randomized into one of the following treatment groups ( n = 10 per group): (1) control siRNA-DOPC (150 µg/kg i.v. twice weekly), (2) control siRNA-DOPC + Dox (4mg/kg/week), (3) FOXM1 siRNA-DOPC (150 µg/kg i.v. twice weekly) and (4) FOXM1 siRNA-DOPC + Dox. Control siRNA sequences are UUCUCCGAACGUGUCACGU [dT][dT], and FOXM1 siRNA sequences are the same as those used for cell line experiments. Tumor cells from breast cancer mouse models (MDA-MD-231; 1.0 × 10 6 ) were injected into the mammary fat pad into mice on day 0, and treatment was started on day 7. Five weeks later, mice were subjected to necropsy, and tumors were harvested.

Immunohistochemistry

Fresh-frozen (OCT) sections from xenograft tumors were stained for FOXM1 and Ki-67, and 5-µm-thick formalin-fixed, paraffin-embedded specimens were stained for cleaved caspase-3. To quantify microvessel density, cell proliferation index (percentage of Ki-67-positive cells) and percentage of cleaved caspase-3-positive cells, we studied five randomly selected 0.159-mm 2 fields for each tumor at 100× magnification.

Results

Systematic analysis reveals that FOXM1 is an oncogenic TF in breast cancer

To identify potential key mediators (drivers) that might determine survival of breast cancer patients, we systematically analyzed ( Supplementary Figure 1 is available at Carcinogenesis Online) gene expression data from two large cohorts of breast cancer patients by applying Kaplan–Meier analysis and the log-rank test. Of 104 genes identified as survival-associated genes ( Figure 1A and Gene list on Supplementary Table 1 is available at Carcinogenesis Online), 11 were TFs. As expected, our list contained some previously identified prognostic TFs. For example, ESR1 and GATA3 have been identified as good prognostic genes ( 10 , 22 ). As targets of TFs are regulated at the transcriptional level, we hypothesized that the biological activity of TFs would best be reflected in the number of mRNAs correlated with expression patterns of TFs. Thus, we prioritized TFs according to the degree of association between their expression patterns and those of the rest of the genes. The degree of association was estimated by the number of genes whose expression pattern was significantly correlated ( P < 0.001) with that of 11 TFs in the NKI cohort ( Figure 1B and Supplementary Figure 2 is available at Carcinogenesis Online). From this priority rank, we selected FOXM1 for further characterization because it had the highest number of correlated genes among poor prognostic genes. In addition, FOXM1 expression is highest in the basal-like subtype of breast cancer, genomic subtype equivalent to clinically defined poor prognosis triple-negative breast cancer ( Figure 1C and D ) ( 23 ). As FOXM1 expression was significantly associated with higher recurrence rates after treatment in receiver operating characteristic analysis ( Figure 1E and F )

Fig. 2.

FOXM1 mediates resistance to Dox in breast cancer. (A–D) Indicated cell lines were transiently transfected with siFOXM1 or siLuc. After 36h, cells were treated with increasing concentrations of indicated drugs for 36h, and proliferation rates were measured by MTT assay. Values shown were normalized to control cells and represent mean ± SD. Student’s t -test was used for comparisons (* P < 0.05, ** P < 0.01, *** P < 0.005).

Fig. 2.

FOXM1 mediates resistance to Dox in breast cancer. (A–D) Indicated cell lines were transiently transfected with siFOXM1 or siLuc. After 36h, cells were treated with increasing concentrations of indicated drugs for 36h, and proliferation rates were measured by MTT assay. Values shown were normalized to control cells and represent mean ± SD. Student’s t -test was used for comparisons (* P < 0.05, ** P < 0.01, *** P < 0.005).

FOXM1 is involved in Dox resistance

We examined whether FOXM1 plays a role in chemoresistance in breast cancer cells. After silencing FOXM1 expression with specific siRNA ( Supplementary Figure 3 is available at Carcinogenesis Online), we carried out cell viability assays in basal-like estrogen receptor (ER)–negative (MDA-MB-231) and ER-positive (MCF-7) breast cancer cells with a panel of chemotherapy agents used to treat breast cancer patients: the DNA topoisomerase IIα (TOP2A) inhibitor Dox, a microtubule stabilizer (paclitaxel), a DNA alkylating agent ( cis -diammineplatinum(II); cisplatin) and an antimetabolite (5-fluorouracil; 5-FU). Interestingly, when FOXM1 expression was silenced, MDA-MB-231 cells became most sensitive to Dox although sensitivity to other cytotoxic agents was increased in general ( Figure 2A ), suggesting that FOXM1 might account for the resistance of breast cancer cells to Dox. Although the viability of MDA-MB-231 cells decreased upon treatment with Dox in a dose-dependent manner, the increase in sensitivity to Dox on silencing FOXM1 expression was independent of dose ( Figure 2B ), suggesting the presence of a specific mechanism of resistance to Dox by FOXM1. Furthermore, results from clonogenicity assay also clearly demonstrated increased sensitivity to Dox upon silencing FOXM1 expression ( Supplementary Figure 4 is available at Carcinogenesis Online). Silencing FOXM1 expression in MCF-7 cells increased their sensitivity to Dox, and the baseline sensitivity of MCF-7 cells was greater and the magnitude of the change in sensitivity was less than in MDA-MB-231cells, suggesting that FOMX1 might play a greater role in Dox resistance in basal-like ER-negative breast cancer cells (MDA-MB-231) than in ER-positive breast cancer cells ( Figure 2C ). We next investigated whether the increase in sensitivity on silencing FOXM1 expression is universal to TOP2 inhibitors, which can be subdivided into intercalating (Dox, mitoxantrone and mAMSA) and nonintercalating (etoposide, teniposide and fluoroquinolones) poisons ( 24 ). Interestingly, although silencing FOXM1 expression significantly increased the sensitivity of MDA-MB-231 cells to mitoxantrone, it did not change their sensitivity to etoposide, indicating the presence of a specific mechanism of resistance to intercalating TOP2 inhibitors by FOXM1 in breast cancer ( Figure 2D ).

Fig. 3.

65 genes confer Dox resistant by FOXM1 in breast cancer. (A) Gene expression profiles of MDA-MB-231 cells were collected after treating cells with siLuc alone, siFOXM1 alone, siLuc + Dox or siFOXM1 + Dox. (B) Expression patterns of 65 genes identified as Dox-sensitizing genes from analysis of gene expression profile data from MDA-MB-231 cells. Data are presented in matrix format; each row represents an individual gene, and each column represents a tissue. Each cell in the matrix represents the expression level of a gene feature in treated cells. In the cells, red and green reflect relatively high and low expression levels of genes, respectively, as indicated in the scale bar (a log2-transformed scale). (C) MDA-MB-231 breast cancer cells were transiently transfected with siFOXM1 or control siLuciferase (siLuc.) After 36h, cells were treated with Dox. We analyzed 30ng of total RNA from transfected cells by qRT–PCR using gene-specific primers as indicated. Student’s t -test (two-tailed) was used to estimate the significance of gene expression changes. All expression levels changed significantly compared with siLuc ( P < 0.01). Results are shown as mean ± SD from three independent replicates. (D) MDA-MB-231 cells were transiently transfected with indicated gene-specific siRNA or siLuc as control. After 36h, cells were treated with Dox (1 or 10 µM) for 36h, and proliferation rates were measured by MTT assay. Values shown were normalized to control cells and represent mean ± SD. Student’s t -test was used for comparisons.

Fig. 3.

65 genes confer Dox resistant by FOXM1 in breast cancer. (A) Gene expression profiles of MDA-MB-231 cells were collected after treating cells with siLuc alone, siFOXM1 alone, siLuc + Dox or siFOXM1 + Dox. (B) Expression patterns of 65 genes identified as Dox-sensitizing genes from analysis of gene expression profile data from MDA-MB-231 cells. Data are presented in matrix format; each row represents an individual gene, and each column represents a tissue. Each cell in the matrix represents the expression level of a gene feature in treated cells. In the cells, red and green reflect relatively high and low expression levels of genes, respectively, as indicated in the scale bar (a log2-transformed scale). (C) MDA-MB-231 breast cancer cells were transiently transfected with siFOXM1 or control siLuciferase (siLuc.) After 36h, cells were treated with Dox. We analyzed 30ng of total RNA from transfected cells by qRT–PCR using gene-specific primers as indicated. Student’s t -test (two-tailed) was used to estimate the significance of gene expression changes. All expression levels changed significantly compared with siLuc ( P < 0.01). Results are shown as mean ± SD from three independent replicates. (D) MDA-MB-231 cells were transiently transfected with indicated gene-specific siRNA or siLuc as control. After 36h, cells were treated with Dox (1 or 10 µM) for 36h, and proliferation rates were measured by MTT assay. Values shown were normalized to control cells and represent mean ± SD. Student’s t -test was used for comparisons.

Fig. 4.

FOXM1 regulates DNA repair genes. (A and B) Cells were transfected with siRNA or siLuc, and the HR repair rate of I-SceI-induced DSBs was measured in U2OS cells. Each value is relative to the percentage of GFP-positive cells in I-SceI-transfected cells without siRNA transfection. Results are shown as mean ± SD from three independent experiments; Stsudent’s t -test was used for comparisons. Western blotting demonstrated effective FOXM1 knockdown. –I-SceI: absence of I-SceI endonuclease; +I-SceI: presence of I- Sce I endonuclease. (C and D) MDA-MB-231 cells were transiently transfected with indicated reporters, cDNA (FOXM1) and siRNAs. Cells were used to measure luciferase activity. (E) Schematic representation of promoter region of DNA repair genes ( EXO1 , PLK4 and RFC4 ). (F) ChIP assay was performed in MDA-MB-231 cells with FOXM1 antibody. Recruitment of FOXM1 to the indicated gene promoters was analyzed using primers specific to the indicated promoter region. IgG was used as an internal control. All results are shown as mean ± SD from three independent replicates (* P < 0.05, ** P < 0.01, *** P < 0.005).

Fig. 4.

FOXM1 regulates DNA repair genes. (A and B) Cells were transfected with siRNA or siLuc, and the HR repair rate of I-SceI-induced DSBs was measured in U2OS cells. Each value is relative to the percentage of GFP-positive cells in I-SceI-transfected cells without siRNA transfection. Results are shown as mean ± SD from three independent experiments; Stsudent’s t -test was used for comparisons. Western blotting demonstrated effective FOXM1 knockdown. –I-SceI: absence of I-SceI endonuclease; +I-SceI: presence of I- Sce I endonuclease. (C and D) MDA-MB-231 cells were transiently transfected with indicated reporters, cDNA (FOXM1) and siRNAs. Cells were used to measure luciferase activity. (E) Schematic representation of promoter region of DNA repair genes ( EXO1 , PLK4 and RFC4 ). (F) ChIP assay was performed in MDA-MB-231 cells with FOXM1 antibody. Recruitment of FOXM1 to the indicated gene promoters was analyzed using primers specific to the indicated promoter region. IgG was used as an internal control. All results are shown as mean ± SD from three independent replicates (* P < 0.05, ** P < 0.01, *** P < 0.005).

FOXM1 govern DNA repair involved in Dox resistance

To gain mechanistic insights into how FOXM1-depleted cells become sensitized to Dox, we carried out microarray experiments to identify genes whose expression is directly altered by silencing FOXM1 expression and Dox treatment. We first identified genes whose expression is significantly altered when cells were treated with siFOXM1 + Dox by comparing the siFOXM1 + Dox group with the other treatment groups (siLuc, siFOXM1 and siLuc + Dox) as controls (gene list A) (Figure 3A). We next identified genes whose expression was altered on Dox treatment by comparing the siFOXM1 group with the siFOXM1 + Dox group (gene list B) and the siLuc group with the siLuc + Dox group (gene list C). To remove genes whose altered expression is due to the general toxicity of Dox treatment, we excluded genes in gene list B and C from gene list A. A series of analysis yielded 65 genes whose expression is only specific to Dox treatment in FOXM1-depleted breast cancer cells. Genes were selected according to statistical significance ( P < 0.001, by Student’s t -test) and fold difference (2-fold). We identified 65 genes as significantly associated with Dox sensitization in FOXM1-depleted cells (Figure 3B and Supplementary Table 2 is available at Carcinogenesis Online (65 gene list). Interestingly, many of these 65 genes are known to be involved in DNA repair (e.g. EXO1 , CDC45L , RFC4 , POLE2 and PLK4 ). The altered expression of these five genes in treated cells was also validated by qRT–PCR (Figure 3C).

We next investigated whether the DNA repair genes regulated by FOXM1 participate directly in Dox sensitization. When expression of these DNA repair genes was silenced by specific siRNAs ( Supplementary Figure 5 is available at Carcinogenesis Online), MDA-MB-231 cells became more sensitive to Dox treatment (Figure 3D), suggesting that downregulation of these genes might account for sensitization of MDA-MB-231 cells to Dox on silencing of FOXM1 expression. Not surprisingly, expression of these DNA repair genes is significantly associated with prognosis in breast cancer patients ( Supplementary Figure 6 is available at Carcinogenesis Online).

Fig. 5.

Proteomic screen identifies NFκB1 as a FOXM1 binding protein. (A) Schematic diagram of the proteomic screen for FOXM1 binding proteins responding to Dox using SILAC method. MDA-MB-231 cells were labeled with 13 C-arginine, 13 C-lysine (heavy medium) or 12 C-arginine, 12 C-lysine (light medium), respectively. Cells are treated with DMSO or Dox (5 µM) for 36h, and cell lysates were used for mass spectrometry analysis. (B) Immunoprecipitation of cultured cells with anti-FOXM1antibody. FOXM1 binding proteins. (C) Screen results using commercially available software showing the SILAC ratios (heavy/light) for the proteins identified by LC-MS/MS. Differentially bound proteins identified by LC-MS/MS are indicated on a log 2 scale. Circles at the top represent proteins with increased interaction with FOXM1 in the presence of Dox, and those at the bottom represent proteins with decreased interaction with FOXM1 in the presence of Dox. is highlighted in red. (D) Coimmunoprecipitation (IP) of NFκB1 with FOXM1 analyzed by western blotting of MDA-MB-231 cells treated with DMSO or Dox. β-actin was used as a negative control. (E) MDA-MB-231 cells were transfected with indicated siRNAs in the presence of Dox or DMSO, and cell lysates were used for western blotting with indicated antibodies.

Fig. 5.

Proteomic screen identifies NFκB1 as a FOXM1 binding protein. (A) Schematic diagram of the proteomic screen for FOXM1 binding proteins responding to Dox using SILAC method. MDA-MB-231 cells were labeled with 13 C-arginine, 13 C-lysine (heavy medium) or 12 C-arginine, 12 C-lysine (light medium), respectively. Cells are treated with DMSO or Dox (5 µM) for 36h, and cell lysates were used for mass spectrometry analysis. (B) Immunoprecipitation of cultured cells with anti-FOXM1antibody. FOXM1 binding proteins. (C) Screen results using commercially available software showing the SILAC ratios (heavy/light) for the proteins identified by LC-MS/MS. Differentially bound proteins identified by LC-MS/MS are indicated on a log 2 scale. Circles at the top represent proteins with increased interaction with FOXM1 in the presence of Dox, and those at the bottom represent proteins with decreased interaction with FOXM1 in the presence of Dox. is highlighted in red. (D) Coimmunoprecipitation (IP) of NFκB1 with FOXM1 analyzed by western blotting of MDA-MB-231 cells treated with DMSO or Dox. β-actin was used as a negative control. (E) MDA-MB-231 cells were transfected with indicated siRNAs in the presence of Dox or DMSO, and cell lysates were used for western blotting with indicated antibodies.

Fig. 6.

NFκB1 regulates DNA repair genes via FOXM1 (A) MDA-MB-231 cells were transiently transfected with siNFκB1 or siLuc. We analyzed 30ng of total RNA from transfected cells by qRT–PCR using gene-specific primers as indicated. (B) MDA-MB-231 cells were transiently transfected with siNFκB1 or siLuc. After 36h, cells were treated with increasing concentrations of Dox for 36h, and proliferation rates were measured by MTT assay. (C) HR assay was performed with NFκB1-depleted cells. (D and E) MDA-MB-231 cells were transfected with 6xFOXM1 reporter plasmid and indicated expression plasmids or siNFκB1. Cell lysates were used to measure luciferase activity. All results are shown as mean ± SD from three independent replicates (* P < 0.05, ** P < 0.01, *** P < 0.005; Student’s t -test).

Fig. 6.

NFκB1 regulates DNA repair genes via FOXM1 (A) MDA-MB-231 cells were transiently transfected with siNFκB1 or siLuc. We analyzed 30ng of total RNA from transfected cells by qRT–PCR using gene-specific primers as indicated. (B) MDA-MB-231 cells were transiently transfected with siNFκB1 or siLuc. After 36h, cells were treated with increasing concentrations of Dox for 36h, and proliferation rates were measured by MTT assay. (C) HR assay was performed with NFκB1-depleted cells. (D and E) MDA-MB-231 cells were transfected with 6xFOXM1 reporter plasmid and indicated expression plasmids or siNFκB1. Cell lysates were used to measure luciferase activity. All results are shown as mean ± SD from three independent replicates (* P < 0.05, ** P < 0.01, *** P < 0.005; Student’s t -test).

FOXM1 directly regulates DNA repair genes

Because HR is one of the mechanisms for repairing DNA double-strand breaks (DSBs) generated by genotoxic reagents, including Dox, and because inhibition of HR leads to increased sensitivity to chemotherapy drugs in human cancers ( 25 , 26 ), we next examined the contribution of FOXM1 to HR by measuring the repair rate of defined genomic DSBs induced at an I- Sce I restriction site in a defective GFP reporter gene ( 17 , 18 ). HR repair was significantly impaired in FOXM1 -depleted cells, as evidenced by a significant decrease (50%) in HR-repaired GFP-positive cells (Figure 4A). Furthermore, silencing RFC4 , POLE2 and PLK4 expression significantly decreased HR (Figure 4B). Expression of these genes is directly regulated by FOXM1, as evidenced by reporter assays and ChIP assays demonstrating that FOXM1 directly bound to EXO1 , RFC4 and PLK4 promoter regions (Figure 4C–F). Taken together, these results strongly suggest that increased sensitivity to Dox on silencing FOXM1 expression is due largely to downregulation of DNA repair genes.

NFκB1 is a binding partner of FOXM1 conferring Dox resistance

We next sought to identify FOXM1 binding partners in Dox-treated cells by using a proteomic approach with the SILAC strategy. Whole-cell lysates from MDA-MB-231 cells grown in indicated culture conditions (Figure 5A) were immunoprecipitated with anti-FOXM1 antibody (Figure 5B). We identified 415 proteins as FOXM1 binding proteins (Figure 5C and Supplementary Table 3 is available at Carcinogenesis Online). Interestingly, we identified the TF NFκB1 (p50/p105) as one of the proteins whose interaction with FOXM1 is significantly increased in Dox-treated cells. We later confirmed this interaction by reciprocal immunoprecipitation assay with anti-NFκB1 in MDA-MB-231 cells treated with Dox (Figure 5D). In addition, NFκB1 expression increased significantly on Dox treatment, whereas FOXM1 expression did not change (Figure 5E). Previous reports demonstrated that NFκB1 expression is associated with drug resistance in human cancer ( 27 ), suggesting that interaction of FOXM1 with NFκB1may account for chemoresistance in cancer cells. Interestingly, expression of the DNA repair genes EXO1 , RFC4 , POLE2 and PLK4 decreased significantly when NFκB1 expression was silenced by specific siRNA (Figure 6A). Moreover, NFκB1-depleted cells became more sensitive to Dox treatment (Figure 6B) ( 27 ). Taken together, these results clearly indicate that NFκB1 in Dox-treated cells co-operatively interacts with FOXM1 to regulate expression of DNA repair genes and protect cancer cells from DSBs induced by Dox. This idea is further supported by our finding that NFκB1-depleted cells were also defective in HR (Figure 6C). Reporter assays with 6xFoxM1 responsive promoter ( 28 ) clearly demonstrated that NFκB1 is a bona fide coactivator of FOXM1 (Figure 6D). FOXM1 transcriptional activity was significantly decreased by silencing endogenous NFκB1 expression, but overexpression of exogenous NFκB1 rescued this decrease (Figure 6E).

Therapeutic efficacy of FOXM1 silencing with Dox treatment

Next, we further determined the effect of silencing FOXM1 expression on xenograft tumor growth with Dox treatment. As expected, treatment with control siRNA + Dox reduced tumor weight significantly compared with treatment with control siRNA alone ( Figure 7A). Treatment with FOXM1 siRNA + Dox resulted in an even greater reduction in tumor weight compared with treatment with control siRNA + Dox or FOXM1 siRNA alone, recapitulating the effects of combination treatment in vitro . Efficient silencing of FOXM1 by siRNA was confirmed by immunostaining of FOXM1 combination treatment significantly activated caspase-3 and decreased cell proliferation, as evidenced by immunostaining of cleaved caspase-3 and Ki-67 (Figure 7B). These xenograft experiments clearly demonstrated that silencing FOXM1 can greatly improve the therapeutic efficacy of Dox in breast cancer.

Discussion

FOXM1 is the proliferation-specific oncogenic TF and overexpressed in a wide range of tumor types examined, including those of mammary, neural, gastrointestinal and reproductive origin ( 29 ). This expression pattern is attributed to the ability of FOXM1 to transactivate genes ( SKP2 , CDC25B , Aurora B Kinase and Survivin ) required for cell cycle progression, suggesting that FOXM1 is a good therapeutic target for human cancer ( 29 , 30 ). Our systematic analysis in breast cancer suggests an oncogenic function for FOXM1 , which is significantly higher in the basal-like subtype that comprise 75% of triple-negative tumors and shows poor prognosis (Figure 1) ( 31 ). Although many patients with ER-positive and human epidermal growth factor receptor 2 (HER2)–positive breast cancer have received great benefit from treatments targeting ER and HER2, chemotherapy is the only available systemic treatment for patients with triple-negative breast cancer, and for these patients, the prognosis remains very poor because of high recurrence rates after treatment and development of resistance during the course of treatment ( 32 , 33 ). Because chemotherapy is currently the sole option for basal-like patients, we explored whether FOXM1 expression had any effect on chemosensitivity in vitro. We observed a statistically significant increase in sensitivity to two TOP2A inhibitors, Dox and mitoxantrone but not etoposide, in FOXM1 -depleted MDA-MB-231 (Figure 2). Dox treatment to cancer cell created double-stranded DNA breaks, and DNA repair genes were induced to rescue the cell from the DNA damage. Our gene expression data clearly demonstrated that DNA repair genes were directly regulated by FOXM1 (Figure 3). Thus, silencing FOXM1 expression significantly sensitizes basal-like breast cancer cells to Dox. Other investigators have also shown that FOXM1 might be involved in resistance to other chemotherapy drugs (cisplatin, paclitaxel and epirubicin) ( 34 , 35 ). As shown in Figure 2A , when compared with cisplatin, paclitaxel and Dox, depletion of FOXM1 expression greatly sensitized to Dox-resistant MDA-MB-231 cells to Dox. Furthermore, previous study also demonstrated that FOXM1 regulates DNA repair genes (XRCC1 and BRCA2) and contributes to DNA damage response although they suggested different mechanism of action regarding FOXM1 function on DNA repair systems, suggesting FOXM1 clearly involved in DNA repair pathway via various repair mechanism ( 36 ).We further demonstrated that FOXM1-dependent resistance to Dox is mediated by FOXM1 interaction with NFκB1 to regulate multiple DNA repair genes and enhance HR ( Figures 5 and 6 ). NFκB has been shown to be one of the principal mediators of resistance to chemotherapy ( 27 ). As a result, a number of pharmacological strategies to inhibit NFκB activation have been developed. Previous study suggested that suppression of NFκB1 overcomes Dox-resistance in vitro and in vivo ( 27 ).

Dox, which is a frequently used chemotherapeutic agent against breast cancer, has been reported to induce acute and chronic toxicities and immunosuppression ( 37 , 38 ). Thus, novel treatment options are needed to reduce dosages of Dox and consequently unwanted toxicities while enhancing its therapeutic efficacy in treating breast cancer ( 37 ). Recent study identified the thiazole antibiotics siomycin A and thiostrepton as potential inhibitors of FOXM1 ( 39 ). Thus, it will be interesting to test that combination of thiazole antibiotics with Dox would overcome resistance of breast cancer cells to Dox or reduce dosage of Dox in routine chemotherapies.

In this study, we have evaluated the potential therapeutic benefit of FOXM1 silencing with Dox in vitro cell culture systems and tested its potential clinical relevance in animal models of breast cancer. Thus, our study opens up new opportunities to test these new drugs as chemotherapy drugs in combination with Dox in future preclinical studies. Together, our results clearly demonstrate that FOXM1 is a promising therapeutic target to improve current systemic chemotherapy to treat breast cancer patients, especially those with triple-negative tumors.

Supplementary material

Supplementary Tables 1–4 and Figures 1–6 can be found at http://carcin.oxfordjournals.org/

Fig. 7.

Effect of silencing FOXM1 expression with or without Dox treatment on growth and apoptosis in an orthotopic breast cancer model. (A) Growth of MDA-MB-231 tumors in mice treated with FOXM1 siRNA-DOPC or control siRNA-DOPC + Dox. Tumor volume was measured at day 37. (B) Ki-67 and cleaved caspase-3 staining of MDA-MB-231 tumor tissues treated with FOXM1 siRNA, control siRNA or control siRNA + Dox. The positive cells per high-power field (HPF) were counted and are shown in the graphs. All results are shown as mean + SD. DOPC, dioleoyl phosphatidylcholine nanoliposomes.

Fig. 7.

Effect of silencing FOXM1 expression with or without Dox treatment on growth and apoptosis in an orthotopic breast cancer model. (A) Growth of MDA-MB-231 tumors in mice treated with FOXM1 siRNA-DOPC or control siRNA-DOPC + Dox. Tumor volume was measured at day 37. (B) Ki-67 and cleaved caspase-3 staining of MDA-MB-231 tumor tissues treated with FOXM1 siRNA, control siRNA or control siRNA + Dox. The positive cells per high-power field (HPF) were counted and are shown in the graphs. All results are shown as mean + SD. DOPC, dioleoyl phosphatidylcholine nanoliposomes.

Funding

The study was funded in part by the intramural faculty fund of The University of Texas MD Anderson Cancer Center, the Basic Science Research Program through the National Research Foundation funded by MEST of Korea (2011-0001047); the Shared Research Equipment Assistance, the National Institutes of Health (RC2GM092599, U54 CA151668, CA128797); the Department of Defense (BC085265); MD Anderson Cancer Center is partially funded by the National Institutes of Health through Cancer Center Support Grant CA016672.

Abbreviations

    Abbreviations
  • Dox

    Doxorubicin

  • ER

    estrogen receptor

  • FOXM1

    Forkhead box protein M1

  • HER2

    human epidermal growth factor receptor 2

  • NFκB

    nuclear factor kappa-light-chain-enhancer of activated B cells

  • PR

    progesterone receptor

  • qRT–PCR

    quantitative reverse transcription–polymerase chain reaction

  • TF

    transcription factors

Acknowledgements

We thank Y. Sun for pGL3- PLK4 promoter construct and P. Raychaudhuri for pEGFP-FOXM1 and 6xFOXM1 responsive promoter (CDX2-Luc).

Y.Y.P. and J.S.L. conceived and designed the study. Y.Y.P. performed the experiments. G.P. performed the HR assay. N.B.J., C.R.A., G.L.B. and A.K.S. carried out animal experiments and collected samples. Y.Y.P., S.B.K. and J.S.L. analyzed the microarray data sets. S.Y.L. and K.K. contributed to the interpretation of data. S.Y.J. performed LC-MS/MS and analyzed the proteomic data. S.R.L. and S.H.L. cloned the reporter constructs. Y.Y.P. and J.S.L. wrote the manuscript. All authors approved the final version of the manuscript.

Conflict of Interest Statement : None declared.

References

1.
Levine
M.
et al
2003
Transcription regulation and animal diversity
.
Nature
  ,
424
,
147
––
151
.
2.
Darnell
J.E.
Jr
2002
Transcription factors as targets for cancer therapy
.
Nat. Rev. Cancer
  ,
2
,
740
––
749
.
3.
Vogt
P.K.
2001
Jun, the oncoprotein
.
Oncogene
  ,
20
,
2365
––
2377
.
4.
Quinn
D.I.
et al
2005
Molecular markers of prostate cancer outcome
.
Eur. J. Cancer
  ,
41
,
858
––
887
.
5.
Brooks
T.A.
et al
2009
The role of supercoiling in transcriptional control of MYC and its importance in molecular therapeutics
.
Nat. Rev. Cancer
  ,
9
,
849
––
861
.
6.
Yu
H.
et al
2009
STATs in cancer inflammation and immunity: a leading role for STAT3
.
Nat. Rev. Cancer
  ,
9
,
798
––
809
.
7.
Bild
A.H.
et al
2006
Oncogenic pathway signatures in human cancers as a guide to targeted therapies
.
Nature
  ,
439
,
353
––
357
.
8.
Nakanishi
C.
et al
2005
Nuclear factor-kappaB inhibitors as sensitizers to anticancer drugs
.
Nat. Rev. Cancer
  ,
5
,
297
––
309
.
9.
van de Vijver
M.J.
et al
2002
A gene-expression signature as a predictor of survival in breast cancer
.
N. Engl. J. Med.
  ,
347
,
1999
––
2009
.
10.
Oh
D.S.
et al
2006
Estrogen-regulated genes predict survival in hormone receptor-positive breast cancers
.
J. Clin. Oncol.
  ,
24
,
1656
––
1664
.
11.
Li
J.
et al
2005
SAK, a new polo-like kinase, is transcriptionally repressed by p53 and induces apoptosis upon RNAi silencing
.
Neoplasia
  ,
7
,
312
––
323
.
12.
Wang
I.C.
et al
2005
Forkhead box M1 regulates the transcriptional network of genes essential for mitotic progression and genes encoding the SCF (Skp2-Cks1) ubiquitin ligase
.
Mol. Cell Biol.
  ,
25
,
10875
––
10894
.
13.
Gravel
S.
et al
2008
DNA helicases Sgs1 and BLM promote DNA double-strand break resection
.
Genes Dev.
  ,
22
,
2767
––
2772
.
14.
Rodriguez
R.
et al
2008
Apoptosis induced by replication inhibitors in Chk1-depleted cells is dependent upon the helicase cofactor Cdc45
.
Cell Death Differ.
  ,
15
,
889
––
898
.
15.
Arai
M.
et al
2009
The knockdown of endogenous replication factor C4 decreases the growth and enhances the chemosensitivity of hepatocellular carcinoma cells
.
Liver Int.
  ,
29
,
55
––
62
.
16.
Cizmecioglu
O.
et al
2008
Plk2 regulated centriole duplication is dependent on its localization to the centrioles and a functional polo-box domain
.
Cell Cycle
  ,
7
,
3548
––
3555
.
17.
Pierce
A.J.
et al
1999
XRCC3 promotes homology-directed repair of DNA damage in mammalian cells
.
Genes Dev.
  ,
13
,
2633
––
2638
.
18.
Peng
G.
et al
2009
BRIT1/MCPH1 links chromatin remodelling to DNA damage response
.
Nat. Cell Biol.
  ,
11
,
865
––
872
.
19.
Ong
S.E.
et al
2006
A practical recipe for stable isotope labeling by amino acids in cell culture (SILAC)
.
Nat. Protoc.
  ,
1
,
2650
––
2660
.
20.
Simon
R.
et al
2007
Analysis of gene expression data using BRB-ArrayTools
.
Cancer Inform.
  ,
3
,
11
––
17
.
21.
Thaker
P.H.
et al
2006
Chronic stress promotes tumor growth and angiogenesis in a mouse model of ovarian carcinoma
.
Nat. Med.
  ,
12
,
939
––
944
.
22.
Badve
S.
et al
2009
Oestrogen-receptor-positive breast cancer: towards bridging histopathological and molecular classifications
.
J. Clin. Pathol.
  ,
62
,
6
––
12
.
23.
Fan
C.
et al
2006
Concordance among gene-expression-based predictors for breast cancer
.
N. Engl. J. Med.
  ,
355
,
560
––
569
.
24.
Nitiss
J.L.
2009
Targeting DNA topoisomerase II in cancer chemotherapy
.
Nat. Rev. Cancer
  ,
9
,
338
––
350
.
25.
Gullotta
F.
et al
(2010)
Targeting the DNA double strand breaks repair for cancer therapy
.
Curr. Med. Chem.
 
26.
Helleday
T.
(2010)
Homologous recombination in cancer development, treatment and development of drug resistance
.
Carcinogenesis
  ,
31
,
955
––
960
.
27.
Mi
J.
et al
2008
RNA aptamer-targeted inhibition of NF-kappa B suppresses non-small cell lung cancer resistance to doxorubicin
.
Mol. Ther.
  ,
16
,
66
––
73
.
28.
Kalinichenko
V.V.
et al
2004
Foxm1b transcription factor is essential for development of hepatocellular carcinomas and is negatively regulated by the p19ARF tumor suppressor
.
Genes Dev.
  ,
18
,
830
––
850
.
29.
Myatt
S.S.
et al
2008
Targeting FOXM1
.
Nat. Rev. Cancer
  ,
8
,
242
.
30.
Costa
R.H.
(
2005
) FoxM1 dances with mitosis
.
Nat. Cell Biol.
  ,
7
,
108
––
110
.
31.
Rakha
E.A.
et al
(
2009
) Triple-negative/basal-like breast cancer: review
.
Pathology
  ,
41
,
40
––
47
.
32.
Dawson
S.J.
et al
2009
Triple negative breast cancers: clinical and prognostic implications
.
Eur. J. Cancer
  ,
45
(suppl. 1),
27
––
40
.
33.
Carey
L.
et al
Triple-negative breast cancer: disease entity or title of convenience?
Nat. Rev. Clin. Oncol.
  ,
7
,
683
––
692
.
34.
Carr
J.R.
et al
FoxM1 mediates resistance to herceptin and paclitaxel
.
Cancer Res.
  ,
70
,
5054
––
5063
.
35.
Kwok
J.M.
et al
FOXM1 confers acquired cisplatin resistance in breast cancer cells
.
Mol. Cancer Res.
  ,
8
,
24
––
34
.
36.
Tan
Y.
et al
2007
Chk2 mediates stabilization of the FoxM1 transcription factor to stimulate expression of DNA repair genes
.
Mol. Cell Biol.
  ,
27
,
1007
––
1016
.
37.
Smith
L.A.
et al
Cardiotoxicity of anthracycline agents for the treatment of cancer: systematic review and meta-analysis of randomised controlled trials
.
BMC Cancer
  ,
10
,
337
.
38.
Bandyopadhyay
A.
et al
Doxorubicin in combination with a small TGFbeta inhibitor: a potential novel therapy for metastatic breast cancer in mouse models
.
PLoS One
  ,
5
,
e10365
.
39.
Gartel
A.L.
A new target for proteasome inhibitors: FoxM1
.
Expert Opin. Investig. Drugs
  ,
19
,
235
––
242
.