Cross-regulation between FOXA1 and ErbB2 Signaling in Estrogen Receptor-Negative Breast Cancer^1,2

Tissue Microarray Cohort and Immunohistochemistry

Three sets of breast cancer tissue microarray (TMA) slides were obtained from Pantomics (http://www.pantomics.com/tissue-arrays/Systems.htm#Breast_disease_tissue_arrays; BRC1501-3, Richmond, CA) as they were published before [23]. The TMA slides constituted of a total of 79 ER- breast tumors with known ErbB2 status. Immunohistochemistry (IHC) staining was performed using EnVision+ System-HRP (DakoCytomation, Sydney, Australia). Antigen retrieval was carried out using Target Retrieval Solution (DakoCytomation). Primary antibody incubation was performed with FOXA1 rabbit polyclonal antibody (ab23738) from Abcam (Cambridge, United Kingdom) at 1:200 dilution [24,25]. TMA cores were examined at 60x magnification using a light microscope (Nikon Instruments, Inc, Tokyo, Japan). A total of 1000 nuclei per each TMA slide were assessed, and the average score for two replicate slides was used for the analysis.

Scoring for FOXA1 was carried out based on a previously described semiquantitative scoring system [26]. On the basis of this system, percentage (P: score of 1 for each 10% positive tumor nuclei such as 10 = 91%–100%) and intensity (I: 1+ to 3+) of nuclear expression were multiplied to generate a numerical score (S = P x I), which could vary between 0 and 30 [26]. In our study, FOXA1 overexpression (FOXA1+) was defined as S > 15 (e.g., intense nuclear staining [3+] in >50% of cells). Using the median score as a cutoff has been previously used to assess protein overexpression in immunohistochemistry studies [27,28].

Cell Culture

Breast cancer cell lines MDA-MB-453, HCC-1954, and MCF-7 were obtained from the American Type Culture Collection (Manassas, VA). All tissue culture media were obtained from Invitrogen (Melbourne, Australia). MDA-MB-453 and HCC-1954 cell lines were cultured in L15 medium, 10% fetal bovine serum, and RPMI 1640 medium, respectively. The MCF-7 cell line was cultured in Dulbecco modified Eagle medium/F12 medium and 10% fetal bovine serum. Treatment with MEK inhibitor, CI-1040 (PD184352; Selleck Chemicals, Houston, TX) at 10 µM concentration, was carried out for 24 hours in full media.

Western Blot Analysis

FOXA1 rabbit polyclonal antibody was obtained from Abcam. Rabbit monoclonal ERK1/2 antibody and rabbit monoclonal phospho-ERK1/2 (Thr202/Tyr204) were obtained from Cell Signaling (Danvers, MA). Western blots were carried out at 1:1000 dilution of each primary antibody using 10 and 20 µg of cell lysates for the total and phospho-proteins, respectively. Protein concentrations from the cell isolates were measured using the BCA Protein Assay Kit (Thermo Scientific, Melbourne, Australia). Rabbit polyclonal α-tubulin antibody (Abcam) was used as the loading control. Analysis of band densities was performed using Bio-Profil Densitometer Software (Vilber Lourmat, Eberhardzel, Germany). All fold changes in band densities were measured relative to the control groups. Western blots were done in two biologic replicates, and the average fold change was shown for each set of experiments.

Real-time Polymerase Chain Reaction Analysis

Total RNA extraction was performed as described before [29]. Real-time polymerase chain reaction (RT-PCR) to assess the expression levels of FOXA1 (assay ID: Hs00270129_m1), RELB (assay ID: Hs00232399_m1), and TLE3 (assay ID: Hs00183222_m1) was carried out using TaqMan Gene Expression Assays (Applied Biosystems, Melbourne, Australia) as instructed by the manufacturer. Housekeeping gene RPLP0 (Applied Biosystems) was used as a control. Relative gene expression = gene expression in the knockdown group / average gene expression in the control group. All experiments were performed in three biologic replicates.

FOXA1 Knockdown in Cell Lines

FOXA1 knockdown (KD) was carried out as described before [30] in MDA-MB-453 and HCC-1954 cells using the following small interfering RNA (siRNA) oligos (duplex; Sigma-Aldrich, Sydney, Australia): D1, 5′CACACAAACCAAACCGUCA; D2, 5′UGACGGUUUGGUUUGUGUG. TLE3 KD was carried out using the following two sets of siRNA oligos (duplex; Sigma-Aldrich): set 1: D1, 5′GUCAUCUACUAAACAAGAA; D2, 5′UUCUUGUUUAGUAGAUGAC; and set 2: D1, 5′GUCAUCUACUAAACAAGAA; D2, 5′UUCUUGUUUAGUAGAUGAC. Transfection of siRNA oligos using Lipofectamine RNAiMAX (Invitrogen) was carried out with a reverse transfection method as instructed by the manufacturer. The final siRNA duplex concentration was 20 nM for the KD experiments. Cells transfected with siRNA Universal Negative Control No. 1 (Sigma-Aldrich) were used as controls. In all experiments, the effects of KDs were assessed 72 hours after the siRNA transfections.

Epidermal Growth Factor Pathway Arrays

TaqMan Human Epidermal Growth Factor (EGF) Pathway Arrays constituting of 96 genes were obtained from Applied Biosystems (Part No: 4418774, the layout of this array is available under the following link: http://www3.appliedbiosystems.com/cms/groups/portal/documents/web_content/dev_043045.gif). Transfections with FOXA1-siRNA oligos and control-siRNA were carried out in MDAMB-453 and HCC-1954 cell lines as described previously. Seventy-two hours after transfections, total RNA was extracted for RT-PCR analysis. The efficiency of FOXA1-KD was validated using RT-PCR as described previously. A total of eight arrays were applied for two biologic replicate experiments with each of the FOXA1-siRNA and controlsiRNA oligos in MDA-MB-453 and HCC-1954 cell lines (four arrays per cell line).

After RT-PCR experiments using human EGF pathways arrays, cycle threshold (C_T) values in each array plate were normalized to the mean C_T values of four housekeeping genes on the plates (18S, GAPDH, HPRT1, and GUSB) to obtain ΔC_T values for each gene. Next, the average ΔC_T values for the control-siRNA plates in each cell line were deducted from the ΔC_T values of the corresponding genes on FOXA1-siRNA plates to obtain ΔΔC_T values. Finally, relative gene expressions were calculated using 2^-ΔΔCT formula as described before [31].

Luciferase Reporter Assays

Full-length complementary DNA (cDNA) clones for CREB1, c-Jun, c-Fos, and AP2α were obtained from Open Biosystems (Thermo Scientific). Full-length cDNA clone for Elk1 was obtained from GeneCopoeia (Rockville, MD). The clones were validated by restriction digestion/sequencing and then subcloned into a pcDNA3.1 vector (Invitrogen) to generate expression constructs. The sequence of 1.5-kb promoter region of FOXA1 was obtained using Ensembl Genome Browser (http://www.ensembl.org/index.html) and PCR generated using the following primer set: forward, GCGCGGTACCCGCAGTGCAGATGCGTTCCC; reverse, GCGCGAGCTCACCTCCTGCGTGTCTGCGTAGT. Subsequently, FOXA1 promoter was cloned into a pGL3 luciferase reporter vector (Promega, Sydney, Australia) and validated by restriction digestion/sequencing.

To carry out the reporter assays, MCF-7 cells were cotransfected with the FOXA1 reporter vector and each of the transcription factors using ExGen 500 reagent (Fermentas Life Sciences, St Leon-Rot, Germany). The Renilla pRL-TK vector (Promega) was used as an internal control reporter. Cotransfection with FOXA1 reporter vector and an empty pcDNA vector was used as a control. Fortyeight hours after the transfections, reporter activities were measured using Dual-Glo Luciferase Assay System (Promega) in an Orion II Microplate Luminometer (Berthold Detection Systems, Pforzheim, Germany). The response ratios for transcription factors and control were measured relative to the internal control reporter (relative response ratio). All reporter assays were carried out in four biologic replicates.

Chromatin Immunoprecipitation Assays

Chromatin immunoprecipitation (ChIP) assays were performed in MDA-MB-453 cell line using ChIP Assay Kit (USB Corporation, Cleveland, OH) as instructed by the manufacturer [32]. ChIP-grade rabbit monoclonal CREB1 (ChIPAb+ CREB Kit; Millipore, Melbourne, Australia), rabbit polyclonal AP2α (Abcam), rabbit polyclonal c-Fos (Abcam), rabbit polyclonal FOXA1 (Abcam), and rabbit polyclonal TLE3 (Proteintech, Chicago, IL) antibodies were applied at 4 µg per assay.

Two primer sets for each of the FOXA1, RELB, and PAK1 promoters were used for the end point RT-PCR amplification using SYBR green method (Applied Biosystems). FOXA1 promoter primers included the following: forward primer set 1, TCACTTGGCTCGGCTGACT (start: -960); reverse primer set 1, AGCCCCACTTTTGCTTCGT (start: -902); forward primer set 2, CTCCCCATTTCCCTCTTTCC (start site: -565); and reverse primer set 2, TGTTTGCAAAGCAGTGTAATTGG (start site: -505). RELB promoter primers included the following: forward primer set 1, CGCCCGGCCTCCAA (start site: -714); reverse primer set 1, GACCAGCGCCTGGAACAC (start site: -664); forward primer set 2, GGTAAGGCGTGATGGCTCTAAG (start site: -381); and reverse primer set 2, CGTCGCTGCCCATTGC (start site: -319). PAK1 promoter primers included the following: forward primer set 1, CTCTTGAGCCCTGGAGTTTGA (start site: -877); reverse primer set 1, CAGGCTGGAGTGCAGTGGTA (start site: -815); forward primer set 2, TGAATGGGTGTGGCTGTGTT (start site: -567); and reverse primer set 2, TGGACTGTATCAAGGGTCAGGAA (start site: -494). Amplification of input chromatin before immunoprecipitation at a dilution of 1:50 was used as a positive control. ChIP assays using nonspecific antibody (rabbit IgG) served as a negative control. The assays were carried out in three replicates and copy number changes were calculated as -Log2 value for each experimental set.

Cell Viability Assay

FOXA1-KD in MDA-MB-453 and HCC-1954 cells was carried out using reverse transfection as described previously. A total of 10,000 cells transfected with either FOXA1-siRNA or control-siRNA were seeded per each well of a 96-well plate. Seventy-two hours after transfections, cell viability was assessed using Vybrant MTT Proliferation Assay Kit (Invitrogen) as instructed by the manufacturer. MTT assays were performed in eight biologic replicates, and absorbance at 570 nm was measured using a plate reader.

Bioinformatics and Statistical Analysis

Analysis of gene expression correlations. Expressions of FOXA1, AP2α, and ErbB2 in a total of 30 ER- breast cancer cell lines were extracted from the microarray data published by Neve et al. [33]. Expression data were normalized to the median expression of each gene across the cohort of ER- cell lines. Correlation coefficients between the normalized expression values of FOXA1 with those of ErbB2 and AP2α were obtained using the Pearson method.

Promoter analysis. Sequences of the 1-kb promoter region of FOXA1, RELB, and PAK1 genes were obtained using Ensembl Genome Browser (http://www.ensembl.org/index.html). Identification of putative binding sites in the promoter regions was carried out using PATCH public 1.0 software (http://www.gene-regulation.com/cgibin/pub/programs/patch/bin/patch.cgi).

Bioinformatics and statistical analysis. Heat map was generated using Spotfire DecisionSite for Functional Genomics (TIBCO, Somerville, MA). Biostatistical analysis was carried out using IBM SPSS Statistics 20 (Armonk, NY). The Mann-Whitney U test was applied for the comparison of nonparametric data. All error bars depict ±2 SEM.

FOXA1 Is Frequently Overexpressed in ErbB2+/ER- Breast Tumors

To assess the association between FOXA1 and ErbB2 overexpression in ER- breast cancer, we carried IHC staining for FOXA1 protein expression in a TMA cohort of 79 ER- tumors with known ErbB2 status as described in the Materials and Methods. This cohort constituted of a total of 36 ErbB2+ tumors defined as 3+ IHC staining for ErbB2 and 43 ErbB2- tumors. After IHC scoring of FOXA1 staining, we compared the percentage of FOXA1+ tumors, as defined by S > 15, between ErbB2+ and ErbB2- groups. Notably, we observed that ErbB2+ tumors had approximately 2.5-fold higher frequency of FOXA1 overexpression at 72% ± 7% compared with that of ErbB2- tumors at 30% ± 7% (P < .01; Figure 1, A–C). In addition, we analyzed the data using a cutoff of S > 24 as applied by Mehta et al. [34]. This approach also demonstrated a higher level of FOXA1 expression in ErbB2+ compared with ErbB2- tumors (Figure W1).

We next tested whether FOXA1 overexpression could be predicted based on ErbB2 status using receiver operating characteristic analysis. This method demonstrated an area under the curve (AUC) of 0.71 for predicting FOXA1+ tumors based on ErbB2 status, indicating that ErbB2 status can classify FOXA1 expression pattern in ER. breast tumors (P < .01; Figure 1D). Altogether, these findings suggest that FOXA1 overexpression is associated with ErbB2+ status in ER- breast tumors.

FOXA1 Expression Correlates with ErbB2 and AP2α Transcript Levels in ER- Cell Lines

We further assessed the association between FOXA1 and ErbB2 expression patterns using the expression microarray data available from a cohort of thirty ER- breast cancer cell lines [33]. In addition, we examined a possible association between the expression of FOXA1 and AP2α using this data set. AP2α is a transcription factor with an established role in the regulation of ErbB2 expression and is also a top-ranking gene in the steroid-response gene signature [3,35–40]. Therefore, a possible association between FOXA1 and AP2α expression would have potential biologic significance in ErbB2+/ER- breast cancer.

The expression data for FOXA1, ErbB2 and AP2α were extracted and normalized as described in the Materials and Methods (Table W1). We next measured Pearson correlation coefficients (CCs) between the expression of FOXA1 and that of ErbB2 and AP2α. Notably, FOXA1 expression had significant correlations with the expression of both ErbB2 (CC = 0.56, P < .01; Figure 1E) and AP2α (CC = 0.6, P < .01; Figure 1F). These data further confirm a robust association between FOXA1 and ErbB2 expression patterns in ER- breast cancer. In addition, AP2α and ErbB2 expressions have a similar level of correlation with that of FOXA1 in ER- cell lines.

ERK Regulates FOXA1 Expression

We next investigated the underlying mechanism for the observed overexpression of FOXA1 in ErbB2+/ER- breast tumors. A possible mechanism for this association is the regulation of FOXA1 transcription by the ErbB2 signaling pathway. To assess this possibility, we first examined the effect of ERK, an important downstream mediator of ErbB2 signaling, on FOXA1 expression using MEK inhibitor CI-1040 [9]. MDA-MB-453 and HCC-1954 cell lines were treated with CI-1040 at 10 µM concentration for 24 hours, and vehicle (DMSO)-treated cells were used as controls. FOXA1 expression was then assessed using RT-PCR and Western blot analysis. We observed that MEK inhibition reduced the level of FOXA1 transcript and protein levels by approximately 35% (Figure 2, A and B). These findings suggest that ERK signaling regulates FOXA1 expression in molecular apocrine cells.

FOXA1 Is a Target of CREB1, AP2α, and c-Fos Transcription Factors

To further identify a mechanism for our observations, we studied whether FOXA1 is regulated by some of the key transcription factors in the ErbB2-ERK signaling pathway. In this respect, we examined CREB1, Elk1, c-Jun, and c-Fos, which are well-characterized downstream mediators of ErbB2-ERK signaling [41–46]. In addition, we investigated a potential role for AP2α in the regulation of FOXA1 because we observed a significant correlation between AP2α and FOXA1 expression (Figure 1F) as well as the fact that AP2α is a critical regulator of ErbB2 transcription [35–40].

Examination of 1-kb FOXA1 promoter region identified several putative binding sites for the previously mentioned transcription factors (Figure 2C). We subsequently used luciferase reporter assays to examine the effects of these predicted transcription factors on the regulation of FOXA1 promoter. Owing to a high degree of transfectability, MCF-7 cells were used for the reporter assay experiments [9,32]. MCF-7 cells were cotransfected with the FOXA1 reporter vector and each of the CREB1, c-Jun, c-Fos, AP2α, and Elk1 expression constructs. Cotransfection with the FOXA1 reporter vector and an empty pcDNA vector was used as a control. Forty-eight hours after the transfections, reporter activities were measured, and relative response ratios were calculated as described in Materials and Methods. We observed a marked increase in FOXA1 reporter activity with CREB1 by approximately 12-fold (P < .01; Figure 2D). In addition, there was an approximately three-fold increase in FOXA1 reporter activity with AP2α and c-Fos constructs (P < .01; Figure 2D). However, c-Jun and Elk-1 did not activate the FOXA1 promoter (Figure 2D).

To assess the binding of CREB1, AP2α, and c-Fos to FOXA1 promoter, we next carried out ChIP assays in MDA-MB-453 cell line. Two sets of primers for FOXA1 promoter in proximity to the predicted binding sites were used for the end point RT-PCR amplification (Figure 2C). Amplification of input chromatin at a dilution of 1:50 and ChIP using nonspecific antibody (rabbit IgG) were applied as positive and negative controls, respectively. Copy number changes were calculated as -Log2 value for each experimental set (Figure 2, E and F). Importantly, we observed a significant enrichment for FOXA1 promoter region with each of the CREB1, AP2α, and c-Fos antibodies using both primer sets (P < .01; Figure 2, E and F). Altogether, these data demonstrate that FOXA1 is a target gene of CREB1, AP2α, and c-Fos transcription factors.

FOXA1 Regulates a Gene Signature in the ErbB2 Signaling Pathway

We next explored whether FOXA1 has a role in transcriptional regulation of the EGFR-ErbB2 signaling pathway. To investigate this possibility, we assessed the effect of FOXA1-KD on the expression of EGFR-ErbB2 signaling genes in MDA-MB-453 and HCC-1954 cell lines using Human EGF Pathway Arrays as described in Materials and Methods. FOXA1-KD was carried out using a siRNA duplex, and the efficiency of KD was evaluated by Western blot analysis and RT-PCR. We observed at least 80% reduction of FOXA1 transcript and protein levels after FOXA1-KD (Figure 3, A and B). It is notable that RT-PCR validations of FOXA1-KD were carried out in all replicate experiments applied for Human EGF Pathway Arrays (Figure 3B).

We subsequently carried out RT-PCR using Human EGF Pathway Arrays and analyzed the expression data as described in Materials and Methods. We identified genes with more than two-fold change in expression between FOXA1-KD and control experiments. Importantly, there was a core gene signature consisting of ten genes that showed more than two-fold change in expression after FOXA1-KD in both MDA-MB-453 and HCC-1954 cell lines (Figure 3C). This FOXA1-regulated gene signature included six upregulated and four downregulated genes (Figure 3C). The upregulated genes were RELB (8.5-fold), c-Fos (3.1-fold), MAP3K1 (3-fold), PRKCE (2.9-fold), R-Ras (2.8-fold), and CDH1 (2.4-fold), with mean fold increase in their expression levels across two cells lines ranging from 2.4- to 8.5-fold (Figure 3C). The downregulated genes included PAK1 (0.29), PRKCB (0.33-fold), VAV3 (0.33-fold), and CSK (0.42-fold), with mean fold reduction in their expression levels across two cells ranging from 0.29- to 0.42-fold (Figure 3C).

Moreover, we identified a separate group of genes, with more than a two-fold change in expression after FOXA1-KD, which were different between MDA-MB-453 and HCC-1954 cell lines (Figure 4, A and B). This differentially regulated gene list in MDA-MB-453 cell line included BRAF, GRB2, IKBKB, MAPK9, and MAP2K7 that were downregulated by 0.27- to 0.49-fold (Figure 4A). In HCC-1954 cell line, CAV2, CBL, and MYC were upregulated by 2.6- to 3.7-fold and DIRAS3, EGF, and RHOB were downregulated by 0.12- to 0.33-fold (Figure 4B).

These findings suggest that FOXA1 regulates the transcription of a gene signature in the ErbB2 signaling pathway. In addition, FOXA1 has both positive and negative regulatory effects on ErbB2 signaling genes as evidenced by down-regulation and up-regulation of these genes after FOXA1-KD, respectively.

Differences in FOXA1-Regulated Genes across Cell Lines Have Functional Implications

Because there are a group genes differentially regulated by FOXA1 between MDA-MB-453 and HCC-1954 cell lines (Figure 4, A and B), we next investigated whether these differences in FOXA1-regulated genes have any functional implications in molecular apocrine cells. To assess this, we examined the effect of FOXA1-KD on ERK phosphorylation and cell viability using Western blot analysis and MTT assay, respectively.

In MDA-MB-453 cell line, FOXA1-KD reduced the level of ERK phosphorylation by 45% compared with the control without any change in total ERK level (Figure 5A). Moreover, this effect was associated with a significant decrease in cell viability by approximately 25% (P < .03; Figure 5B). It is notable that ErbB2 level did not change after FOXA1-KD, indicating that ErbB2 expression is not regulated by FOXA1 in this cell line (Figure 5A). In contrast, FOXA1-KD did not affect the level of ERK phosphorylation and cell viability in HCC-1954 cell line (Figure 5, C and D). These findings suggest that FOXA1 expression is necessary for maintaining ERK phosphorylation and cell viability in MDA-MB-453 line as opposed to HCC-1954 line. Therefore, the functional effects of FOXA1 can vary across molecular apocrine cells.

RELB and PAK1 Are FOXA1 Target Genes in ErbB2 Signaling

We subsequently assessed whether selective signature genes are direct targets of FOXA1. For this purpose, we examined FOXA1 binding to the promoters of RELB and PAK1, which were the most upregulated and downregulated genes in the gene signature, respectively (Figure 3C). It is notable that examination of 1-kb promoter regions of RELB and PAK1 identified several putative binding sites for FOXA1 (Figure 6A).

To assess FOXA1 binding to the promoters of RELB and PAK1, we carried out ChIP assays using FOXA1 antibody in MDA-MB-453 cell line. Two sets of primers in proximity to the predicted binding sites were used for each of the RELB and PAK1 promoters to assess the end point RT-PCR amplification (Figure 6A). Amplification of input chromatin at a dilution of 1:50 and ChIP using nonspecific antibody (rabbit IgG) were applied as positive and negative controls, respectively. Copy number changes were calculated as -Log2 value for each experimental set (Figure 6, B and C). ChIP assays demonstrated significant binding of FOXA1 to the promoters of RELB and PAK1 (P < .01; Figure 6, B and C). Moreover, the results were reproducible using both primer sets for each of the promoters (Figure 6, B and C). These findings indicate that RELB and PAK1 are FOXA1 target genes in the ErbB2 signaling pathway.

FOXA1 Binding Partner Corepressor TLE3 Regulates RELB

Our results indicate that FOXA1 negatively regulates a group of genes in the ErbB2 signaling pathway such as RELB. It has been shown that FOXA1 can act as a pioneer factor to recruit transcriptional corepressor TLE3 resulting in gene repression distal to the recruitment site [47,48]. To investigate whether this recruitment of TLE3 would provide a possible mechanism for the negative regulatory effects of FOXA1 observed in our study, we examined the role of TLE3 in the regulation of RELB as an example of a target gene repressed by FOXA1.

We first investigated the effect of TLE3-KD on RELB expression in MDA-MB-453 cell line. TLE3-KD was carried out using two sets of siRNA duplexes and the efficiency of KD experiments were assessed using RT-PCR. We observed approximately 80% reduction in the level of TLE3 expression with both siRNA duplexes (Figure 6D). We next measured RELB expression levels after TLE3-KD using RT-PCR. Notably, there was a significant increase in RELB expression by 8- to 11-fold after TLE3-KD that was reproducible with both sets of TLE3-siRNA duplexes (P < .01; Figure 6E).

We subsequently tested whether TLE3 can be recruited to the RELB promoter regions that bind FOXA1 using ChIP assays with the same primer sets applied for FOXA1-RELB ChIP experiments. Amplification of input chromatin at a dilution of 1:50 and ChIP using nonspecific antibody (rabbit IgG) were applied as positive and negative controls, respectively. Copy number changes were calculated as -Log2 value for each experimental set (Figure 6F). Importantly, we observed significant enrichment for RELB promoter region using TLE3 antibody with both primer sets (P < .01; Figure 6F). These results indicate that, in a similar fashion to FOXA1, TLE3 negatively regulates RELB and is recruited to the same promoter regions of this gene.