AKT Alters Genome-Wide Estrogen Receptor α Binding and Impacts Estrogen Signaling in Breast Cancer

Abstract

Estrogen regulates several biological processes through estrogen receptor α (ERα) and ERβ. ERα-estrogen signaling is additionally controlled by extracellular signal activated kinases such as AKT. In this study, we analyzed the effect of AKT on genome-wide ERα binding in MCF-7 breast cancer cells. Parental and AKT-overexpressing cells displayed 4,349 and 4,359 ERα binding sites, respectively, with ∼60% overlap. In both cell types, ∼40% of estrogen-regulated genes associate with ERα binding sites; a similar percentage of estrogen-regulated genes are differentially expressed in two cell types. Based on pathway analysis, these differentially estrogen-regulated genes are linked to transforming growth factor β (TGF-β), NF-κB, and E2F pathways. Consistent with this, the two cell types responded differently to TGF-β treatment: parental cells, but not AKT-overexpressing cells, required estrogen to overcome growth inhibition. Combining the ERα DNA-binding pattern with gene expression data from primary tumors revealed specific effects of AKT on ERα binding and estrogen-regulated expression of genes that define prognostic subgroups and tamoxifen sensitivity of ERα-positive breast cancer. These results suggest a unique role of AKT in modulating estrogen signaling in ERα-positive breast cancers and highlights how extracellular signal activated kinases can change the landscape of transcription factor binding to the genome.

Estrogen-mediated transcriptional regulation has been studied quite extensively because of its relevance to breast cancer (1, 26). It also serves as a model system to study integration of several signaling events to transcription (26). The transcriptional effects of estrogen are largely mediated by two receptors belonging to nuclear receptor superfamily: estrogen receptor α (ERα) and ERβ (26). ERs bind DNA either directly via estrogen response elements (ERE) or tethered through other transcription factors (8, 30, 47). Genome-wide analyses of ERα binding to DNA by chromatin immunoprecipitation combined with microarrays (ChIP-on-chip) have revealed the presence of complete (GGTCAnnnTGACC) or half-ERE (GGTCA)-like sequences in ∼95% of ERα binding regions (13, 36). These studies also revealed enrichment of binding sites for C/EBP, Oct, and Forkhead transcription factors adjoining ERE sequences in the ERα binding regions (13). Importantly, only ∼5% of ERα binding sites are within 5-kb promoter-proximal region, with a great majority being found in the intronic or distal locations of the transcription start site of a gene (13, 36).

The transcriptional activity of ERα is regulated by distinct posttranslational modifications, including phosphorylation. Several extracellular signal-activated kinases phosphorylate ERα and regulate DNA binding, transactivation, coregulator interaction, stability, and/or subcellular distribution. Kinases known to phosphorylate ERα include mitogen-activated protein (MAPK), IKKα, RSK, AKT/PKB, p38 kinase, PKA, Src, cyclin A/cdk2, and cdk7 (1, 11, 27, 31, 45, 48). It has been suggested that changes in the phosphorylation status of the receptor contributes to ERα dysfunction in various pathological conditions, including breast cancer (1).

The effects of phosphorylation on ERα activity have been studied by many methods, including in vitro DNA-binding assays, ChIP analysis of few endogenous target genes, and transient-transfection experiments with synthetic ERE-containing reporters or with the promoter regions of endogenous target genes (2, 11, 27, 45, 48). Kinase-induced changes in ERα DNA-binding activity in a genome-wide scale have not been analyzed. Here, we focus on the serine/threonine kinase AKT/PKB because it phosphorylates ERα, confers anti-estrogen resistance, and is aberrantly activated in ∼50% of human malignancies (17, 28). In addition, transgenic mice overexpressing AKT1 in the mammary gland predominantly develop ERα-positive mammary tumors when exposed to chemical carcinogen (6). Several other studies also intimately link AKT to ERα-positive breast cancers (5, 40).

We examined the effect of AKT on genome-wide ERα binding by ChIP-on-chip assay using MCF-7 breast cancer cells as a model system. AKT-mediated changes in ERα binding were then compared to estrogen-regulated gene expression in these cells. In addition, the expression pattern of genes that displayed AKT-dependent changes in expression and/or ERα binding were evaluated in publicly available primary tumor gene expression data. These studies reveal that AKT significantly alters ERα binding to genes in transforming growth factor β (TGF-β), nuclear factor-κB (NF-κB), and E2F activated signaling pathways. In addition, AKT may change secondary estrogen target gene expression by modulating basal and/or estrogen-regulated expression of E2F family transcription factors (8, 52).

MATERIALS AND METHODS

Cell culture, cell growth assays, and reagents.

We used bicistronic retrovirus (pqCXIP vector; BD Biosciences, Palo Alto, CA)-mediated gene transfer to generate parental MCF-7 cells with retrovirus vector alone (MCF-7p) or constitutively active AKT (MCF-7AKT). Cells infected with retroviruses were selected by using 1 μg of puromycin/ml. Constitutively active AKT construct has been described previously (11). Cells were maintained in minimal essential medium plus 10% fetal calf serum and transferred to phenol red-free minimal essential medium plus 5% charcoal-dextran treated serum for 4 days prior to experiments. To measure the effect of estrogen and TGF-β on proliferation, 1,000 cells/well were plated in 96-well plates (eight wells for each treatment conditions), and proliferating cells were quantitated by using bromodeoxyuridine-enzyme-linked immunosorbent assay (EMD Biosciences, San Diego, CA). Experiments were repeated five times, and the mean and standard deviation values are presented in the text.

ChIP-on-chip analysis.

Chromatin preparation, ChIP, and hybridization were performed as described previously, except that samples were hybridized to a seven-chip set (Affymetrix GeneChIP human titling 2.0R array set; catalog no. 900772) and analyzed by using the human genome version 18 (Hg.18) (13). A model-based analysis of tiling-array analysis was performed according to methods defined previously on ChIP-on-chip data sets in MCF7 cells (25, 34). The analysis of estrogen-treated MCF-7p cells utilized two complete whole-genome data sets and four control input data sets from the same cells. The estrogen-treated MCF-7AKT analysis utilized two complete whole-genome data sets and three control input data sets with background DNA from the same cells. The data sets were evaluated at different false-discovery rates (FDRs) between 1 and 20%, with a bandwidth of 300 bases to extend from the position being analyzed and with a maximum gap of 300 bp between positive probes. The minimum number of probes for a MATcH score analysis was ten (25).

Overlap analysis was performed within an in-house script to identify binding regions between multiple ChIP-on-chip data sets that overlap in genomic space. Sites were considered shared if they touched each other in two cell types. Overlap analysis was completed at each FDR cutoff between 1 and 20%. Final numbers from the 5 and 20% FDRs were utilized for calculations. Each of the possible combinations of site overlap analysis was considered for MCF-7p(FDR20%), MCF-7AKT(FDR20%), MCF-7p(FDR5%), and MCF-7AKT(FDR5%) to generate the relevant Venn-diagram (25). Some of the ERα binding sites identified in this analysis were further confirmed by ChIP, followed by quantitative PCR (Q-PCR) as described previously (13). The data from three to five independent ChIP-Q-PCR experiments are presented.

Microarray analysis.

Gene expression analysis was performed in quadruplicate using RNA from ethanol or 4-h-estrogen-treated MCF-7p and MCF-7AKT cells. Integrity and quality of samples were measured with a 2100 Bioanalyzer. Samples that passed the checks were treated with the rigorous method of Turbo DNA-free (catalog no. AM1907; Ambion). After the digestion, samples were requantified by using a spectrophotometer. Then, 200-ng portions of the samples were used for whole-genome amplification using an Illumina TotalPrep RNA amplification kit (catalog no. AMIL1791; Ambion). Samples were quantified again, and 1.5-μg portions of each sample were hybridized on Sentrix BeadChip arrays for gene expression Human-6 V2 (catalog no. BD-25-112) at 58°C for 19 h with agitation. The hybridized arrays were washed according to the updated wash procedure, labeled with Cy3-streptavidin (catalog no. PA43001; GE Healthcare), and scanned with an Illumina BeadArray reader. Logarithmic ratios of signal intensities were normalized by using quantile normalization within four arrays in each condition. The normalized signals were further adjusted by aligning median values across four conditions. After genes whose expression levels were not reliably detected in at least half of samples in all of the four conditions (detection P value of >5%) were filtered, differentially expressed genes were analyzed by using the Student t test and two-way analysis of variance.

RNA preparation, RT-PCR, and Q-PCR.

RNA was isolated by using an RNeasy kit from Qiagen (Valencia, CA). Reverse transcription-PCR (RT-PCR) was performed using the single-step RT-PCR kit from Invitrogen (Carlsbad, CA). For Q-PCR, RNA was reverse transcribed using a single-stranded cDNA synthesis kit (Invitrogen) and subjected to Q-PCR using Sybr green (Applied Biosciences, Foster City, CA).

Sample preparation and Western blot analysis.

Cytoplasmic, nuclear, membrane, soluble, and whole-cell lysates were prepared by using previously described methods and subjected to Western blotting (44, 56). ERα antibody for Western was from Chemicon (Temecula, CA), whereas AKT antibody was purchased from Cell Signaling (Danvers, MA). ChIP quality ERα antibodies were from NeoMarkers (Fremont, CA) and Santa Cruz Biotechnology (Santa Cruz, CA); AIB1 antibody was purchased from Santa Cruz Biotechnology.

Sequence conservation analysis.

The 4349 and 4359 ERα ChIP-enriched regions in MCF-7p and MCF-7AKT cells were aligned at their centers and expanded to 3,000 bp in each direction. Within this region, the phastCons scores for each position were retrieved from UCSC genome browser (http://genome.ucsc.edu), which was calculated based on a phylogenetic hidden Markov model that measures the evolutionary conservation in 17 vertebrates, including mammalian, amphibian, bird, and fish species.

Motifs enriched in ER binding regions.

Position-specific score matrices (PSSM) were used to calculate the possibility of a specific transcription factor binding at one genomic locus, as described previously (37).

The ERα ChIP-enriched regions were scanned for transcription factor-binding motifs using 741 PSSM documented in the TRANSFAC database. The highest matching score during this scanning was used to evaluate the binding potential of the candidate transcription factor in the specific ERα ChIP-enriched region. The background nucleotide sequences were selected from all of the 1,000-bp promoter sequences upstream of the transcription starting sites of known genes that are not overlapped with ERα ChIP-enriched region. We did not use a randomly selected intergenic region as background sequences to avoid the sequence bias caused by the unique features in the promoter regions, such as higher levels of GC content. For each PSSM, a unique cutoff score was determined by the lowest matching score where the density of positively identified binding sites in the ChIP-enriched regions is five times more than the one in the background oligonucleotide sequences (or FDR of ≤20%). This analysis enables selecting transcription factors whose binding sites are enriched in ChIP-selected regions compared to background promoters.

RESULTS

AKT changes ERα DNA binding to the genome.

We, among others, have previously shown estrogen-independent activation of ERα and anti-estrogen and chemotherapy resistance of MCF-7 breast cancer cells overexpressing constitutively active AKT (11, 41). The constitutively active AKT overexpressed in these cells is distributed in the plasma membrane, cytoplasm, and nucleus, suggesting that it functions in multiple cellular compartments (Fig. 1A). The endogenous AKT is minimally active in MCF-7 cells under the growth condition used in our studies, and estrogen causes a modest increase in AKT activity after 2 to 4 h of treatment (57). Overexpressed CA-AKT is functional based on the elevated S6 kinase activity in CA-AKT-overexpressing cells (data not shown). We performed ChIP with ERα antibody with or without estrogen treatment for 1 h of parental (MCF-7p) and CA-AKT-overexpressing (MCF-7AKT) cells and hybridized ChIP DNA to the Affymetrix human tiling microarrays representing the entire nonrepetitive human genome sequence, as described previously, with the exception that the array used was a seven-chip set without mismatch probes and with the human genome version 18 (Hg.18) (13). Both data sets with the MCF-7p ChIP-on-chip and MCF-7AKT ChIP-on-chip were evaluated across all FDR values from 1 to 20% and examined for overlaps in genomic space between the data sets. The percentage of overlaps was at the highest at the low FDR values and decreased as the stringency for the cutoff was reduced to an FDR of 20%. The maximum percentage of sites shared by the two datasets was 60% at the cutoff of an FDR of 5%. That number steadily decreased to 40% at the cutoff of an FDR of 20%. The sites that were shared in the genomic space almost always were direct overlaps with the same binding center, suggesting that ERα binding was utilizing the same binding site in both cell types (13). The number of shared and unique ERα binding sites in two cell types is shown in Fig. 1B. Figure S1 in the supplemental material shows a Venn diagram for the number of unique and overlapping sites in two cell types at FDRs of 5 and 20%. Totals of 4,349 and 4,359 ERα binding sites were detected in MCF-7p and MCF-7AKT cells, respectively, with an FDR of 5%. Approximately 40% of ERα binding sites are unique to MCF-7AKT cells, suggesting significant effects of AKT on ERα binding in vivo.

FIG. 1.

AKT affects ERα binding to the genome. (A) Subcellular distribution pattern of CA-AKT and its effect on ERα levels. Nuclear, cytoplasmic, soluble, and membrane fractions of MCF-7p (control) and MCF-7AKT cells were subjected to Western analysis for AKT and ERα. Blots were reprobed for PARP to ensure the enrichment of nuclear proteins in the nuclear fraction. (B) Summary of ERα DNA-binding pattern to the genome of MCF-7p and MCF-7AKT cells after treatment with estrogen (10⁻⁸ M for 1 h). (C) AKT-mediated changes in ERα DNA-binding patterns to representative genes are shown. Open arrows indicate the direction of the gene in relation to chromosomal location, and black bars represent ERα binding sites in two cell types. (D) ERα binding pattern to chromosome 1 in two cell types with or without estrogen treatment.

The ERα binding pattern in both cell types upon estrogen treatment to a few representative genes and to chromosome 1 is shown in Fig. 1C and D. AKT displayed gene-specific effects, as demonstrated by either an increase in the number (for example, GREB-1, FGFR2, and SLC7A5) or a decrease in the number (for example, EHF and SMAD3) of ERα binding sites around these genes. We used ChIP-PCR and ChIP-Q-PCR to verify the effects of AKT on ERα binding to few estrogen responsive genes (Fig. 2). Estrogen-dependent ERα binding to the enhancer-3 region of GREB-1 was observed in both MCF-7p and MCF-7AKT; it was more sustained in MCF-7AKT cells (Fig. 2A). No amplified product was detected when ChIP was done with control immunoglobulin G, suggesting the specificity of the ERα antibody.

FIG. 2.

ChIP analysis of representative genes showing time-dependent changes in ERα and AIB1 recruitment in two cell types. (A) ChIP analysis of enhancer-3 region of GREB-1 for ERα binding. A total of 20 pg of ChIP DNA obtained using ERα antibody or control IgG was subjected to 30 cycles of PCR using GREB-1 enhancer-3-specific primers. (B) ChIP-Q-PCR analysis of ERα binding to select genes. GREB-1 enhancer-3 was analyzed again using ChIP DNA from independent experiments. Averages and standard deviations from three to five experiments are shown. ERα bound to SMRT-element-2 in MCF-7AKT cells in an estrogen-independent manner. Also, there was a difference in the duration of ERα binding to GREB-1 Enh-3 in the two cell types. *, P < 0.05 MCF-7p versus MCF-7AKT for the indicated time point. (C) Effect of AKT on recruitment of AIB1 to ERα binding regions of GREB-1 enhancer-3 and SMRT gene region 1. *, P = 0.02 (MCF-7p versus MCF-7AKT).

The SMRT gene contains multiple ERα binding regions, and one of these regions was unique to MCF-7AKT cells in a ChIP-on-chip assay (Fig. 1C). In ChIP-Q-PCR, the binding of unliganded ERα to this region (SMRT-element-2) was observed only in MCF-7AKT cells; estrogen modestly enhanced ERα binding to this region (Fig. 2B). Although ChIP-on-chip assay failed to detect ERα binding to this region in estrogen-treated MCF-7p cells, ChIP-Q-PCR revealed estrogen-inducible ERα binding to this site, albeit at a lower level than in MCF-7AKT cells. The pattern of estrogen-inducible ERα binding to the other region of SMRT (SMRT-1) was similar in both cell types, although the level of ERα binding under estrogen-treated condition was higher in MCF-7AKT cells than in MCF-7p cells. We have confirmed ERα binding to regions detected in DSCAM, NRIP, SLC7A5, and cyclin D1 genes by ChIP-on-chip assay (data not shown). Overall, we consistently observed prolonged ERα binding to select target sequences in MCF-7AKT cells compared to MCF-7p cells. Non-ERα-binding regions were not amplified in an estrogen-dependent manner in PCRs containing ERα-ChIP DNA (data not shown). The primer sequences used for Q-PCR are provided in Table S1 in the supplemental material.

We used ChIP-Q-PCR to examine the effect of AKT on coactivator recruitment to select ERα binding sites. We selected AIB1 as the coactivator for the present study because it is overexpressed in MCF-7 cells and in breast cancer (3, 43, 50). Also, AKT stabilizes AIB1 protein (21). The kinetics of AIB1 recruitment to ERα binding sites of GREB-1 was similar in both cell types, although overall AIB recruitment was higher in MCF-7AKT cells (Fig. 2C). Recruitment of AIB1 to SMRT-1 site was much lower than to GREB-1 site, and it was significantly lower in MCF-7AKT cells than in MCF-7p cells (P = 0.02 at 4 h of estrogen treatment). Although the trend of the AIB1 binding pattern was similar in all experiments, there was marked experimental variation in the n-fold change in binding, possibly reflecting a profound influence of modest changes in growth conditions on binding of this coactivator to the genome. Both MCF-7p and MCF-7AKT cells under both basal and estrogen-treated conditions contained similar levels of acetylated histone H3 (data not shown). Taken together, AKT mainly alters the pattern of ERα recruitment to the genome and may affect site-specific coactivator recruitment.

Sequence conservation and ERα binding site distribution pattern in two cell types in relation to transcription start sites.

We mapped the location of ERα binding sites relative to the transcription start sites of known genes from RefSeq in two cell types. As reported previously for MCF-7 cells (13, 36), the majority of ERα binding sites were detected away from the transcription start site, with only 9.6 and 10.9% of the sites in MCF-7p and MCF-7AKT cells, respectively, within the 1-kb promoter region (Fig. 3A). ERα binding sites in both cell types showed a high degree of conservation across multiple vertebrate species (Fig. 3B), and regions that are unique in MCF-7p (not overlapping with any AKT-region) have a higher conservation score than ones that are unique in MCF-7AKT cells (P < 10⁻¹⁶, Mann-Whitney U test, Fig. 3C). Whether these differences in conservation score contribute to species-specific differences in the effects of AKT on ERα binding is not known. Nonetheless, the high degree of sequence conservation of the ERα binding region unique to MCF-7AKT cells suggests that these sites are bona fide ERα binding sites and that their accessibility to ERα is controlled by extracellular signals and/or tissue-specific chromatin modifications.

FIG. 3.

Sequence conservation and ER binding site distribution pattern in two cell types in relation to transcription start sites. (A) ERα binding site density in relation to the transcription start site. (B) Sequence conservation of all ERα binding sites in both cell types. Sequences flanking ERα binding regions are not conserved. (C) Sequence conservation of ERα binding sites that are unique to both cell types. Note the higher conservation of ERα binding sites in MCF-7p cells compared to MCF-7AKT cells (P < 10⁻¹⁶).

Enrichment of other transcription factor binding sites in the ERα binding region in two cell types.

AKT may influence ERα binding by at least three mechanisms: (i) AKT-mediated ERα phosphorylation or coregulator interactions convert low-affinity ERα binding sites to high-affinity binding sites and vice versa; (ii) AKT expands the repertoire of transcription factors to which ERα interacts, which permits ERα association with DNA through protein-protein interaction; and (iii) AKT increases the levels and/or DNA-binding activity of cooperating transcription factors such as FOXA1 and AP-1 that bind to sites adjacent to the ERα binding region (13). To explore these possibilities, we scanned the ERα binding region for enrichment of specific transcription factor binding motifs that are common and distinct to each cell type. This was achieved by using the position weight matrices documented in the TRANSFAC (29), which permitted unbiased characterization of 741 transcription factor binding sites. The majority of the ERα binding region contained half or a complete ERE motif, with some genes having both half and full EREs (Fig. 4A).

FIG. 4.

Enriched motifs within ERα ChIP-enriched regions in two cell types. (A) Enriched motifs in ERα binding region in two cell types. The y axes in the left panel demonstrate the ratio of the numbers of specific binding sites in the ChIP-enriched regions to the ones in randomly selected promoter sequences (1,000-bp upstream transcription starting site of the genes that are not ERα targets) or the FDR at different matching score cutoffs. The numbers in the Venn diagrams are calculated based on the matching score cutoffs that results in a 20% FDR. Motif enrichment was similar in both cell types, even within binding sites that are unique to a cell type. (B) Alignment of enriched motifs with the ERE. Nucleotide sequences indicated in red are shared with ERE.

We next compared the distribution pattern of ERα binding sites in two cell types in relation to closely located genes. An ERα binding region was assigned to a gene if it was located within 20 kb 5′ of the transcription start site or within 20 kb 3′ of the transcription termination site. We emphasize that such categorization is not meant to implicate ERα binding to estrogen-regulated expression of the target gene. Based on these criteria, 1,294 genes in both cell types were associated with ERα binding sites with half and/or full ERE elements. However, 373 and 614 unique genes in MCF-7p and MCF-7AKT cells, respectively, were associated with the ERα binding region with full or half ERE elements. Complete ERE sequences were associated with 1,140 genes that are common in both cell types, 286 genes in MCF-7p cells, and 544 genes in MCF-7AKT cells. The list of genes associated with half or full EREs is provided in Table S2 in the supplemental material. The percentage of genes with full-length ERE and ERE half sites in our studies is modestly different than that in a recently published study using the Hg.17 version (13, 36). In the present study, 86% of the ERE sites are full length (compared to 71% in the previous study) (36). Nonetheless, as expected, ERE elements are the predominant elements enriched in ERα binding regions.

ERα binding regions displayed enrichment of several other non-ERE motifs in a significant manner. These include the peroxisome proliferator activator receptor response element, the AP-1 element, the BACH1 element, the androgen receptor response element, the NF-E2 element, and the glucocorticoid receptor binding site. The genes associated with ERα binding region and one or more of the above elements are listed in Table S2 in the supplemental material. Among the above elements, the peroxisome proliferator activator receptor response element shows maximum homology with ERE (Fig. 4B). Alignment of other elements with ERE is less than 50%. Thus, BACH1, AP-1, NF-E2, the androgen receptor, and the glucocorticoid receptor may correspond to factors that cooperate with ERα. Significantly, we did not observe enrichment of a unique transcription factor binding site within ERα binding regions that is specific for either MCF-7p or MCF-7AKT cells, suggesting that AKT did not expand the repertoire of transcription factors to which ERα interacts.

Correlating AKT-mediated changes in ERα DNA binding with E2-regulated gene expression.

To determine whether the observed differences in ERα binding pattern between two cell types correlate with changes in estrogen-regulated gene expression, we performed gene expression analysis of MCF-7p and MCF-7AKT cells with or without estrogen treatment for 4 h. The number of estrogen-regulated genes jumped from 833 in MCF-7p cells to 1,063 in MCF-7AKT cells (P < 0.05; >1.5-fold change) with 357 genes differentially expressed in two cell types (only P < 0.05, but no n-fold change was considered) (Table 1). Totals of 299 and 363 of the estrogen-regulated genes in MCF-7p and MCF-7AKT cells, respectively, were associated with ERα binding sites. Lists of estrogen-regulated genes in the two cell types, along with the number of ERα binding sites associated with each gene is provided in Tables S3 to S6 in the supplemental material. Overall, ∼40% of estrogen-regulated genes were associated with ERα binding sites; whether these sites are functionally involved in estrogen-regulated expression of the corresponding gene needs to be investigated.

TABLE 1.

Summary of estrogen-regulated gene expression patterns and ERα binding sites in MCF-7p and MCF-7AKT cells

Gene category	No. of genesa
Estrogen-induced genes in MCF-7p cells (P < 0.05; >1.5-fold change)	480
Estrogen-induced genes in MCF-7AKT cells (P < 0.05; >1.5-fold change)	701
Estrogen-repressed genes in MCF-7p cells (P < 0.05; values, <0.67)	353
Estrogen-repressed genes in MCF-7AKT cells (P < 0.05; values, <0.67)	362
Genes associated with ERα binding sites in estrogen-treated MCF-7p cells	211/88
Genes associated with ERα binding sites in estrogen-treated MCF-7AKT cells	281/82
Genes differentially expressed upon E2 treatmentb	357
Genes differentially expressed in MCF-7p cells associated with ERα binding sites	94
Genes differentially expressed in MCF-7AKT cells associated with ERα binding sites	93
Differentially expressed genes associated with ERα binding sites in only MCF-7p cells	21
Differentially expressed genes associated with ERα binding sites in only MCF-7AKT cells	22
Differentially expressed genes associated with different no. of ERα binding sites	27
Genes that are differentially expressed without an associated change in ERα binding	45

^aValues are expressed as number induced/number repressed where applicable.

^bIncluding those that are induced or repressed in both cell types but at different levels.

We next examined the ERα binding pattern among genes that are differentially expressed in two cell types. Only 94 and 93 of these genes, respectively, associate with ERα binding sites in MCF-7p and MCF-7AKT cells (Table 1). Differentially estrogen-regulated genes in the two cell types with ERα binding sites could be further classified into four groups: (i) those with ERα binding sites in only MCF-7p cells (Table 2); (ii) those with ERα binding sites in only MCF-7AKT cells (Table 3); (iii) those with different number of ERα binding sites in two cell types (Table 4); and (iv) those with the same number of binding sites but differential estrogen effects (Table 5). Overall, AKT-induced changes in ERα binding correlated with elevated induction or repression of genes and in some cases loss of repression or induction. For example, CYP26A1, a gene involved in retinoic acid metabolism (22) was induced at a much higher level in MCF-7AKT cells despite a reduced number of ERα binding sites in these cells (Table 4). LRFN5 gained an ERα binding site in MCF-7CA-AKT cells and was repressed by estrogen in only AKT cells.

TABLE 2.

Genes associated with ERα binding sites in only MCF-7p cells and are differentially expressed in two cell types^a

GenBank no.	Gene symbol	Strand	No. of binding sites in MCF-7p+E2	Fold change upon E2		Differential expression in MCF-7p and MCF-7AKT	P
				MCF-7p	MCF-7AKT
NM_018948	ERRFI1	−	1	1.14*	1.39*	1.21	4.8E-02
NM_001006605	FAM69A	−	1	1.02	0.92*	0.90	2.0E-02
NM_032044	REG4	−	1	1.06	0.87*	0.83	2.0E-02
NM_004419	DUSP5	+	1	1.20*	1.74*	1.45	2.5E-03
NM_001005463	EBF3	−	1	1.04	0.92*	0.88	1.1E-02
NM_007173	PRSS23	+	1	1.98*	2.47*	1.25	4.2E-02
NM_005892	FMNL1	+	1	1.06	0.94	0.88	2.4E-02
NM_001544	ICAM4	+	1	1.07	0.90*	0.84	6.9E-03
NM_024656	GLT25D1	+	1	1.60*	1.93*	1.21	4.5E-02
NM_006343	MERTK	+	4	1.47*	1.89*	1.28	4.0E-03
NM_003592	CUL1	+	1	1.08	1.20*	1.12	3.0E-02
NM_018437	HEMGN	−	1	1.02	1.11*	1.09	4.5E-02
NM_001701	BAAT	−	1	1.07	0.92*	0.86	3.4E-02
NM_015597	GPSM1	+	1	1.21*	1.49*	1.23	3.0E-03
NM_003115	UAP1	+	1	0.80*	1.09	1.36	4.7E-02
NM_016545	IER5	+	1	0.86*	1.09	1.26	2.9E-02
NM_012342	BAMBI	+	2	0.83*	1.10	1.32	1.2E-03
NM_144973	MGC24039	−	1	0.96	1.13*	1.18	4.9E-02
NM_024089	KDELC1	−	1	0.97	1.11*	1.14	3.2E-02
NM_033643	RPL36	+	1	0.83*	0.96*	1.16	2.0E-03
NM_001125	ADPRH	+	1	0.92*	1.04	1.13	3.6E-02
NM_012135	FAM50B	+	1	0.79*	0.88*	1.11	4.9E-02

^aDifferential expression in MCF-7p cells and MCF-7AKT cells of >1: estrogen-induced gene upregulation is greater in MCF-7AKT cells than in MCF-7p cells or estrogen-mediated downregulation is greater in MCF-7p cells than in MCF-7AKT cells. Differential expression in MCF-7p cells and MCF-7AKT cells of <1: estrogen-induced gene upregulation is greater in MCF-7p cells than in MCF-7AKT cells, or estrogen-mediated downregulation is greater in MCF-7AKT cells than in MCF-7p cells. *, genes whose expression was significantly different in untreated and estrogen-treated cells.

TABLE 3.

Genes associated with ERα binding sites in only MCF-7AKT cells and are differentially expressed in two cell types^a

GenBank no.	Gene symbol	Strand	No. of binding sites in MCF-7AKT+E2	Fold change upon E2		Differential expression in MCF-7p and MCF-7AKT	P
				MCF-7p	MCF-7AKT
NM_005185	CALML3	+	1	1.10	0.93*	0.84	2.2E-02
NM_005197	FOXN3	−	2	1.03	0.90*	0.88	4.3E-02
NM_000196	HSD11B2	+	2	1.36*	1.63*	1.19	1.1E-02
NM_016095	GINS2	−	1	1.46*	1.70*	1.17	4.3E-02
NM_002274	KRT13	−	2	1.48*	1.91*	1.28	3.9E-02
NM_016077	PTRH2	−	1	1.68*	1.87*	1.12	2.4E-02
NM_006640	SINT1	+	1	1.45*	1.71*	1.18	4.7E-02
NM_004036	ADCY3	−	1	1.49*	1.71*	1.15	4.7E-02
NM_015341	NCAPH	+	1	1.08*	0.94	0.87	2.6E-02
NM_018288	PHF10	−	1	1.01	1.18*	1.17	3.1E-03
NM_019848	SLC10A3	−	1	1.34*	1.67*	1.24	1.3E-03
NM_001066	TNFRSF1B	+	1	0.94*	1.09	1.16	9.6E-03
NM_030952	NUAK2	−	1	0.79*	0.92*	1.16	1.2E-02
NM_173576	MKX	−	1	0.70*	0.55*	0.78	1.3E-02
NM_080760	DACH1	−	2	0.90*	1.05	1.17	8.6E-03
NM_005358	LMO7	+	1	0.74*	0.86*	1.17	4.5E-02
NM_032714	C14orf151	+	1	0.83*	0.98	1.18	4.8E-02
NM_016084	RASD1	−	1	0.45*	0.72*	1.60	8.5E-03
NM_178233	OTOP3	+	1	0.92*	1.05	1.14	3.1E-02
NM_019850	NGEF	−	1	0.93*	1.04*	1.12	1.7E-02
NM_005668	ST8SIA4	−	1	0.63*	0.47*	0.74	3.9E-03

^aDifferential expression in MCF-7p and MCF-7AKT cells of >1: estrogen-induced gene upregulation is greater in MCF-7AKT cells than in MCF-7p cells or estrogen-mediated downregulation is greater in MCF-7p cells than in MCF-7AKT cells. Differential expression in MCF-7p and MCF-7AKT cells of <1: estrogen-induced gene upregulation is greater in MCF-7p cells than in MCF-7AKT cells or estrogen-mediated downregulation is greater in MCF-7AKT cells than in MCF-7p cells. *, genes whose expression was significantly different in untreated and estrogen-treated cells.

TABLE 4.

Genes that showed differential expression in two cell types, which correlated with a change in the number of ERα binding sites^a

GenBank no.	Gene symbol	Strand	No. of binding sites		Fold change		Differential expression in MCF-7p and MCF-7AKT	P
			MCF-7p+E2	MCF-7AKT+E2	MCF-7p	MCF-7AKT
NM_015215	CAMTA1	+	4	6	1.52*	1.88*	1.24	4.0E-02
NM_001408	CELSR2	+	3	4	2.32*	2.98*	1.28	5.3E-03
NM_001014797	KCNMA1	−	7	10	1.08	0.91	0.84	3.0E-02
NM_057157	CYP26A1	+	2	1	1.25*	1.60*	1.28	8.1E-04
NM_015062	PPRC1	+	3	2	2.06*	2.57*	1.25	1.0E-02
NM_007069	HRASLS3	−	2	1	1.20*	1.44*	1.20	4.5E-03
NM_002382	MAX	−	2	1	1.17*	1.39*	1.18	1.7E-02
NM_018119	POLR3E	+	1	2	1.29*	1.44*	1.12	2.5E-02
NM_018667	SMPD3	−	3	2	1.38*	1.17*	0.85	1.6E-02
NM_013370	OSGIN1	+	3	4	1.85*	2.78*	1.50	2.6E-02
NM_002265	KPNB1	+	1	2	1.03	1.23*	1.19	6.0E-03
NM_005884	PAK4	+	3	5	1.37*	1.52*	1.11	4.9E-02
NM_001009565	CDKL4	−	1	2	1.03	0.93*	0.90	4.4E-02
NM_001450	FHL2	−	3	4	2.43*	3.44*	1.41	1.1E-02
NM_003043	SLC6A6	+	2	4	1.32*	1.76*	1.33	2.5E-02
NM_015238	WWC1	+	1	2	2.68*	3.15*	1.18	4.4E-02
NM_003897	IER3	−	3	1	1.48*	1.76*	1.19	2.1E-02
NM_173492	PIP5KL1	−	1	2	1.05	1.34*	1.27	4.4E-02
NM_004878	PTGES	−	1	2	1.98*	3.39*	1.71	5.4E-03
NM_004433	ELF3	+	2	3	0.73*	0.87*	1.19	4.3E-02
NM_012308	FBXL11	+	1	3	0.93*	0.83	0.89	3.5E-03
NM_152447	LRFN5	+	1	2	0.95	0.67*	0.71	1.7E-03
NM_001077182	FSCN2	+	2	1	0.96	1.08*	1.13	1.9E-02
NM_004228	PSCD2	+	2	1	0.86*	0.98	1.13	1.7E-02
NM_016315	GULP1	+	3	1	0.90*	0.72*	0.80	2.6E-02
NM_015234	GPR116	−	2	1	0.84*	0.97	1.16	2.2E-02
NM_015689	DENND2A	−	1	2	0.95	1.10*	1.16	2.9E-02

^aDifferential expression in MCF-7p and MCF-7AKT cells of >1: estrogen-induced gene upregulation is greater in MCF-7AKT cells than in MCF-7p cells or estrogen-mediated downregulation is greater in MCF-7p cells than in MCF-7AKT cells. Differential expression in MCF-7p and MCF-7AKT cells of <1: estrogen-induced gene upregulation is greater in MCF-7p cells than in MCF-7AKT cells or estrogen-mediated downregulation is greater in MCF-7AKT cells than in MCF-7p cells. *, genes whose expression was significantly different in untreated and estrogen-treated cells. Genes with a change of >1 are inducible genes, and genes with a change of <1 are repressed genes.

TABLE 5.

Genes associated with same number of ERα binding sites but differentially regulated by estrogen^a

GenBank no.	Gene symbol	Strand	No. of E2 inducible binding sites	Fold change		Differential expression between MCF-7p and MCF-7AKT cells	P
				MCF-7p	MCF-7AKT
NM_013319	UBIAD1	+	1	1.88*	2.18*	1.16	2.4E-02
NM_025077	TOE1	+	2	1.32*	1.59*	1.20	1.8E-02
NM_004428	EFNA1	+	2	0.35*	0.32*	0.91	2.2E-02
NM_002928	RGS16	−	1	1.42*	1.70*	1.19	3.2E-02
NM_001030287	ATF3	+	1	1.27*	1.68*	1.32	4.2E-03
NM_000399	EGR2	−	1	0.97	1.16*	1.19	2.9E-02
NM_018649	H2AFY2	+	1	1.07	1.34*	1.26	5.9E-04
NM_019058	DDIT4	+	1	0.66*	0.88	1.32	7.4E-03
NM_138706	B3GNT6	+	2	1.12*	0.94*	0.84	2.0E-03
NM_000926	PGR	−	1	2.27*	3.14*	1.38	8.0E-05
NM_003979	GPRC5A	+	1	0.72*	0.95	1.32	2.6E-02
NM_005967	NAB2	+	2	1.07	1.24*	1.16	3.6E-02
NM_079836	TUBA3C	−	1	1.35*	1.68*	1.24	6.7E-03
NM_005766	FARP1	+	3	1.52*	1.76*	1.16	2.7E-02
NM_014045	LRP10	+	1	0.86*	0.99	1.16	2.2E-02
NM_014584	ERO1L	−	1	1.20*	1.38*	1.15	3.8E-02
NM_024071	ZFYVE21	+	2	1.29*	1.49*	1.16	2.5E-02
NM_207419	C1QTNF8	−	2	1.13*	1.02	0.90	4.4E-02
NM_138440	VASN	+	3	0.90	1.17	1.30	4.1E-02
NM_001025604	ARRDC2	+	2	0.92	1.04	1.13	4.0E-02
NM_033103	RHPN2	−	1	1.20*	1.44*	1.20	9.0E-03
NM_014681	DHX34	+	2	1.09*	1.23*	1.13	3.6E-02
NM_000979	RPL18	−	1	0.90*	0.97	1.08	2.9E-02
NM_005253	FOSL2	+	1	0.95	1.15*	1.22	2.6E-02
NM_199227	MAP1D	+	1	1.02	0.87*	0.85	1.0E-02
NM_024893	C20orf39	+	1	3.59*	3.26*	0.91	3.4E-02
NM_021809	TGIF2	+	1	2.27*	2.63*	1.16	3.2E-02
NM_005423	TFF2	−	1	1.52*	2.42*	1.59	8.0E-05
NM_001006114	C21orf57	+	1	0.92	1.04	1.13	3.9E-02
NM_007061	CDC42EP1	+	1	1.46*	1.76*	1.21	4.7E-02
NM_003670	BHLHB2	+	1	2.02*	3.00*	1.48	2.9E-02
NM_024513	FYCO1	−	1	0.86*	0.75*	0.88	3.7E-02
NM_002841	PTPRG	+	1	0.88	1.09	1.25	2.4E-02
NM_182920	ADAMTS9	−	1	1.11	1.45*	1.30	3.0E-03
NM_152773	MGC33212	−	1	0.99	0.86*	0.87	4.6E-02
NM_013409	FST	+	2	1.05	1.27*	1.21	1.5E-02
NM_000358	TGFBI	+	1	0.78*	0.93	1.18	2.2E-03
NM_000679	ADRA1B	+	1	0.78*	0.97	1.25	4.6E-02
NM_178012	TUBB2B	−	1	8.86*	13.92*	1.57	2.1E-04
NM_006638	RPP40	−	1	1.62*	1.83*	1.13	1.4E-02
NM_021922	FANCE	+	1	1.20*	1.38*	1.15	1.2E-02
NM_022917	NOL6	−	1	1.91*	2.30*	1.20	3.1E-02
NM_017881	C9orf95	−	1	2.07*	2.38*	1.15	1.8E-02
NM_017443	POLE3	−	1	1.35*	1.50*	1.11	2.4E-02
NM_001633	AMBP	−	2	1.11*	1.30*	1.17	4.1E-02

^aDifferential expression in MCF-7p and MCF-7AKT cells of >1: estrogen-induced gene upregulation is greater in MCF-7AKT cells than in MCF-7p cells or estrogen-mediated downregulation is greater in MCF-7p cells than in MCF-7AKT cells. Differential expression in MCF-7p and MCF-7AKT cells of <1: estrogen-induced gene upregulation is greater in MCF-7p cells than in MCF-7AKT cells or estrogen-mediated downregulation is greater in MCF-7AKT cells than in MCF-7p cells. *, genes whose expression was significantly different in untreated and estrogen-treated cells. Genes with a change of >1 are inducible genes, and genes with a change of <1 are repressed genes.

To further verify whether AKT-induced changes in ERα binding correlates with changes in the estrogen-regulated expression of target gene, we performed semiquantitative RT-PCR (Fig. 5A) or real-time RT-PCR (Fig. 5B). The expression of select genes was measured in untreated and estrogen-treated (for 1 to 24 h) MCF7p and MCF-7AKT cells. The majority of genes displayed significant differences in basal and/or estrogen-regulated expression in the two cell types. For example, the basal expression of HSPB8 and FKBP4 was higher in MCF-7AKT cells, which was further enhanced by estrogen, correlating with gene expression array results (see Table S3 in the supplemental material). Unliganded ERα bound to FKBP4 gene in only MCF-7AKT cells, and estrogen increased the number of binding sites to three. In contrast, in MCF-7p cells we observed only estrogen-dependent ERα binding to two regions (see Table S3 in the supplemental material). HSPB8 displayed an ERα binding site only in estrogen-treated MCF-7AKT cells, possibly reflecting AKT-mediated increase in the affinity of ERα to this gene (see Table S3 in the supplemental material).

FIG. 5.

Estrogen-regulated gene expression pattern in MCF-7p and MCF-7AKT cells. (A) Semiquantitative RT-PCR analyses of representative genes, which are associated with differential ERα binding after estrogen treatment in two cell types. Representative data from two or more experiments of a set of estrogen-inducible and estrogen-repressible genes are shown. (B) Quantitative RT-PCR analysis of representative genes in two cell types treated with estrogen for the indicated times. The data from three experiments, each performed in duplicate, are shown. PCRs contained water as a negative control, and no amplification was observed in the absence of cDNA in the reactions.

Estrogen-inducible expression of ADORA1 was higher in MCF-7AKT clls than in MCF-7p cells, which correlated with an AKT-mediated increase in the number of ERα binding sites from one to two. The estrogen-repressible gene BTG2 displayed one ERα binding site in estrogen-treated MCF-7p cells but none in MCF-7AKT cells (see Table S6 in the supplemental material). Accordingly, estrogen-mediated repression of this gene was more prolonged in MCF-7p cells than in MCF-7AKT cells (Fig. 5A). AKT-mediated changes in ERα binding and coactivator recruitment did not always correlate with changes in expression pattern of the target gene in all assays. For example, estrogen induced ERα binding to four and seven regions, respectively, of GREB-1 in MCF-7p and in MCF-7AKT cells and two of seven sites in MCF7AKT cells bound to ERα in the absence of estrogen (see Table S3 in the supplemental material). However, estrogen-inducible expression of this gene was not significantly different in the two cell types. In fact, in two of three experimental strategies (in gene expression array and RT-PCR but not quantitative RT-PCR), GREB-1 expression was lower in MCF-7AKT cells than in MCF-7p cells (Fig. 5 and see Table S3 in the supplemental material). Thus, AKT-mediated changes in ERα binding may have gene-specific effects, with only certain genes responding to altered ERα binding with increase or decrease in transcription.

AKT may influence secondary estrogen target gene expression through differential basal and/or estrogen-inducible expression of E2F2 and E2F6.

Of 357 genes that displayed differential estrogen-regulated expression in two cell types, only 93 genes are associated with ERα binding sites (Table 1). The rest of the differentially estrogen-regulated genes may correspond to secondary estrogen response genes, and their expression is some how influenced by AKT. Recent studies have shown that secondary estrogen target gene expression is mediated through E2F family transcription factors since majority of genes in this category have E2F binding sites (8, 52). Consistent with this possibility, gene expression array results (see Table S4 in the supplemental material) revealed higher levels of estrogen-inducible expression of E2F2 and E2F6 in MCF-7AKT cells than in MCF-7p cells. To evaluate this further, we examined ERα binding patterns to E2F1, E2F2, E2F6, E2F7, and E2F8. Only E2F2 and E2F6 genes contain ERα binding sites (Fig. 6A). E2F2 contains one ERα binding site close to the transcription start site that is common to both cell types. E2F6 gene contains an ERα binding site ∼15 kb 5′ of the transcription start site in only MCF-7p cells. Thus, AKT may target E2F2 and E2F6 to alter the expression of estrogen secondary response genes.

FIG. 6.

AKT overexpression leads to changes in estrogen-regulated expression of E2F2 and E2F6, which mediates secondary estrogen target gene expression. (A) ERα binding to E2F2 and E2F6 genes in two cell types. Two versions of E2F6 gene are shown: an expanded version depicting both E2F6 and the adjoining GREB-1 genes and a concise version with only E2F6 gene displaying the ERα binding site in MCF-7p cells. GREB-1 is an estrogen-regulated gene with multiple ERα binding sites. The direction of the gene (large arrow) and the ERα binding regions (black bars) are shown. (B) The basal and estrogen-inducible expression of E2F2 and E2F6 in MCF-7p and MCF-7AKT cells was measured by quantitative RT-PCR. The data from three experiments done in duplicate are shown (the asterisk indicates a comparison between MCF7p and MCF7AKT cells under similar treatment conditions [P < 0.05]).

We verified the effect of estrogen on E2F2 and E2F6 expression by quantitative RT-PCR. AKT increased the basal expression of E2F2 with further increase in overall expression in cells treated with estrogen for 24 h (Fig. 6B). We have confirmed elevated basal and estrogen-induced increases in E2F2 protein in MCF-7AKT cells by Western analysis (data not shown). Estrogen-inducible E2F6 expression was significantly higher in MCF-7AKT cells than in MCF-7p cells after 3 h of estrogen treatment, confirming the gene expression array results (Fig. 6B). Depending on the cell type, E2F2 functions as an oncogene or tumor suppressor and is required for cMyc-induced cell proliferation but not apoptosis (15, 32, 54). E2F6 is a transcription repressor, is a component of Bmi-1 polycomb repressor complex, and is suggested to suppress cell proliferation (15, 55). Thus, AKT may change the expression levels of both estrogen-induced and estrogen-repressed secondary target genes either at the basal level (through E2F2) or under estrogen-treated conditions (through E2F6).

To determine how many of the previously described secondary estrogen responses are differentially expressed in the two cell types, we compared our gene expression analysis with that published by Bourdeau et al. (8). A gene was designated as a secondary estrogen response gene if estrogen-inducible expression was observed in untreated but not in cycloheximide-treated MCF-7 cells in the previous study (8) but lacking ERα binding site in our study. This analysis revealed elevated expression of CDT1 and RAB31 in MCF-7AKT cells compared to MCF-7p cells (P < 0.05) (see Table S7 in the supplemental material). Nine other secondary response genes showed differential expression between two cell types, but the differences did not reach statistical significance.

Expression pattern of differentially estrogen-regulated genes in two cell types in primary breast cancers.

We next examined the expression pattern of genes listed in Tables 2 to 5 (both estrogen-inducible and estrogen-repressible genes) in relation to ERα status in primary breast cancer by using a publicly available database (www.oncomine.org). Of 41 genes whose expression data are available, 21 genes are overexpressed in ERα-positive breast cancers (see Table S8 in the supplemental material), and 17 of these are estrogen-inducible genes. AKT increased the expression of all 17 of these genes and reduced the estrogen-mediated repression of 3 of the 4 repressed genes. At least eight genes in this list have previously been linked to AKT. For example, DUSP5, which is overexpressed in ERα-positive breast cancer and is estrogen inducible to a greater extent in MCF-7AKT cells (see Table S8 in the supplemental material), dephosphorylates and inactivates ERK1/2 (10). AKT-mediated inactivation of ERK1/2 has previously been shown to be essential for shifting breast cancer cells from cell cycle arrest to proliferation (58). Therefore, AKT-induced changes in ERα binding and consequent changes in gene expression may contribute to the overexpression of specific genes in ERα-positive breast cancers.

Recently, genomic grade index (GGI), also called MaPQuant, has been used to subclassify ERα-positive breast cancers into two clinically distinct groups (38, 51). This classification relies on the average expression levels of 97 genes that are associated with cell proliferation and histologic grade. ERα-positive patients with a higher GGI respond poorly to tamoxifen therapy and display poor survival. We examined how many of genes in this list are regulated by estrogen and whether AKT influenced their expression. A total of 66 genes (P < 0.05, >1.5-fold change) were estrogen regulated in MCF-7p cells, with 42 being induced by estrogen. Surprisingly, only five of these genes were associated with ERα binding sites, suggesting that secondary estrogen response constitutes the major component of GGI. A total of 75 genes in the list were regulated by estrogen in MCF-7AKT cells; 50 of which were estrogen inducible, and 9 were associated with ERα binding sites. Also, the expression levels of these genes were significantly different from MCF-7p cells, and the total expression level of these genes was higher in MCF-7AKT cells than in MCF-7p cells. Overall, our results suggest that AKT alters GGI by increasing the expression levels of estrogen-regulated genes, which may have prognostic implications.

Signaling pathways affected by AKT-mediated changes in ERα DNA binding.

Four data sets were used to perform Ingenuity Pathway (a software program) analysis to identify direct interaction network among the genes associated with differential ERα binding pattern and/or expression between two cell types. The first data set was 357 genes that showed differential estrogen-regulated expression in two cell types. The second and third data sets were from Tables 2 and 3, representing genes associated with ERα binding sites in MCF-7p and MCF-7AKT cells, respectively. The fourth data set was from Oncomine analysis of genes that are overexpressed in ERα-positive breast cancers (from Table S8 in the supplemental material). All four analyses identified specific effects of AKT on TGF-β response pathways. The TGF-β pathway identified using 357 differentially expressed genes and other data sets are shown in Fig. 7A and Fig. S2 in the supplemental material, respectively. Deregulation of the TGF-β1-mediated growth inhibitory pathway is required for normal ERα-positive cells of the mammary gland to acquire proliferative capacity (20). Also, TGF-β growth-inhibitory effects are gradually lost during cancer progression, and this loss of growth-inhibitory effects in breast cancer is dependent on ERα status (9). Therefore, we examined the effect of TGF-β1 on the proliferation of MCF-7p and MCF-7AKT cells. Consistent with a previous report (39), MCF-7p cells were growth inhibited by TGF-β, and estrogen reversed the growth-inhibitory effect (Fig. 7B). In contrast, MCF-7AKT cells were less sensitive to TGF-β-mediated growth inhibition than were MCF-7p cells (P = 0.01). Reduced sensitivity of MCF-7AKT cells to TGF-β could be due to AKT-mediated ligand-independent activity of ERα or due to ERα-independent function of AKT. Additional pathways identified in this analysis include NF-κB, E2F, retinoic acid, and tumor necrosis factor network (Fig. S2 in the supplemental material). Taken together, our results suggest a significant effect of AKT on pathways that control apoptosis (tumor necrosis factor [TNF] and NF-κB), differentiation/estrogen response (retinoic acid and TGF-β), and proliferation/invasion (TGF-β) in ERα-positive breast cancers.

FIG. 7.

AKT reduces the growth-inhibitory function of TGF-β1. (A) Ingenuity pathway analysis of genes that are differentially expressed in estrogen-treated MCF-7p and MCF-7AKT cells. Genes indicated in green are repressed by estrogen, and those indicated in red are induced by estrogen in either or both cell types. (B) The effect of TGF-β1 on proliferation. Cells were treated with TGF-β (25 ng/ml) with or without estrogen (10⁻¹⁰ M) for 3 days, and the cell proliferation was measured by bromodeoxyuridine-enzyme-linked immunosorbent assay. Averages and standard deviations from four independent experiments are shown (TGF-β-treated MCF-7p cells versus TGF-β-treated MCF-7AKT cells [*, P = 0.01]).

DISCUSSION

In this study, we describe the effect of AKT overexpression on ERα DNA binding pattern in MCF-7 breast cancer cells. Activation of AKT is observed in several cancers, including breast cancer (17). AKT activation in cell lines and primary breast cancers is associated with anti-estrogen-resistance phenotype (11, 28, 46). In addition, MMTV-AKT1 transgenic mice develop ERα-positive breast cancers upon exposure to chemical carcinogen (6). Thus, AKT may have causative role in ERα-positive breast cancers.

A number of extracellular signaling pathways modulate subcellular localization, dimerization, DNA binding, and transactivation function of ERα (1, 31, 35). However, the effects of these signaling molecules on ERα DNA binding on a global scale have not been reported. Although in vitro DNA binding assays such as electrophoretic mobility shift assays did not detect an effect of AKT on ERα DNA binding, genome-wide analysis on chromatinized DNA revealed significant effects of AKT on ERα binding. A number of genes gained or lost ERα binding upon AKT overexpression, which often correlated with altered estrogen-regulated expression of corresponding genes. Whether AKT-mediated changes in ERα binding to the genome require AKT-mediated phosphorylation of ERα is technically difficult to evaluate. Most of the ERα target genes are epigenetically silent in ERα-negative cells or get silenced in ERα-positive cells upon experimental depletion of ERα (33). This epigenetic silencing precludes any overexpression studies with wild type or phosphorylation-defective mutants of ERα. In addition, the use of natural inducers of AKT such as heregulin, instead of AKT overexpression, is not a feasible approach since most of these inducers simultaneously induce other signaling molecules such as MAPK in MCF-7 cells. MAPK phosphorylates and regulates ERα activity (27, 57).

How AKT-mediated changes in ERα binding pattern alter the overall phenotype of breast cancer cells remains to be determined. It appears that AKT alters basal and/or estrogen-regulated expression of genes in TGF-β, retinoic acid, E2F, NF-κB, and TNF signaling pathways. MCF-7AKT cells proliferate faster than MCF-7p cells when grown under estrogen-free but not under regular growth media. This effect of AKT on proliferation requires ERα since short hairpin RNA against ERα reduced the effect of AKT (data not shown). Thus, AKT-mediated changes in the activity of unliganded ERα may contribute to cell proliferation. The same activity of unliganded ERα may also be responsible for resistance of MCF-7AKT cells to TGF-β. In this respect, AKT increased the expression of BHLHB2, also called Stra13 or DEC1 (Table 5), which mediates the cell survival function of TGF-β1 (19). BHLHB2 is one of the frequently overexpressed genes in ERα-positive breast cancers (15 studies P = 3.3E-13, see Table S4 in the supplemental material). TGIF2, a TGF-β1-inducible homeobox gene that represses TGF-β target genes by interacting with SMAD proteins (42), is induced by estrogen at a much higher level in MCF-7AKT cells than in MCF-7p cells (Table 5). Animal models have shown that proliferation of nondividing ERα-positive normal mammary epithelial cells is restrained by TGF-β and loss of this TGF-β response is required for ERα-positive cells to proliferate in response to estrogen (20). Because AKT activation is observed in the preneoplastic state in breast cancer and AKT overexpression in mammary gland leads to ERα-positive breast cancers upon carcinogen exposure (6, 7), one of the primary functions of activated AKT may be to impair TGF-β-mediated growth inhibition of ERα-positive cells.

AKT caused a significant change in the ERα DNA-binding pattern through unknown mechanisms. The majority of the ERα binding regions that are unique to MCF-7AKT cells contain ERE elements. Even among ERα binding regions that are common to both cell types, we consistently observed prolonged ERα binding to selected regions in MCF-7AKT cells compared to MCF-7p cells, suggesting an influence of AKT on the clearance of ERα from DNA (Fig. 2B). We propose that AKT changes the overall chromatin structure and exposes certain ERE elements to ERα. In this context, ERα binding regions in MCF-7AKT cells is associated with higher levels of AIB1 under estrogen-free conditions; however, the overall histone H3 acetylation levels were the same in both cell types (data not shown). Additional studies with other coregulators involved in ERα activity are required to delineate the effects of AKT on coregulator recruitment to the ERα binding site. In summation, differential estrogen-regulated expression of the genes in MCF-7p and MCF-7AKT cells may involve (i) gene-specific effects of AKT on ERα binding and coactivator recruitment and (ii) the effects AKT on the expression of E2F family transcription factors that control secondary estrogen response (8).

Transcription factor binding site analysis did not reveal enrichment of specific transcription factor binding sites in the ERα binding region of either cell types. However, there was a trend for an increased number of AP-1 binding sites in the ERα binding region in MCF-7AKT cells compared to MCF-7p cells (Fig. 4A). In this context, AKT did increase the levels of AP-1 transcription factor (data not shown), which is considered one of the major ERα tethering factors, particularly for genes that lack EREs and are repressed by estrogen (13). It is possible that an AKT-mediated increase in the levels or posttranslational modification alters the DNA-binding affinity of these transcription factors to specific elements depending on variation in the sequence motifs; this in turn changes the affinity of ERα to the transcription factors themselves or to the adjoining ERE elements.

Among other transcription factor-enriched motifs, BACH1 and NF-E2 (or NF-E2-related factor NRF-2) are of interest since these are redox-regulated transcription factors. Their activity is further regulated at the level of nuclear localization, and at least the activity of NRF-2 is regulated by AKT (14, 24, 49, 53). Thus, BACH1, NF-E2/NRF-2, and AKT may determine estrogen response by directing the ERα DNA-binding pattern, which may be additionally controlled by the redox status of cells. Note that although previous studies have shown AKT-mediated reduction in ERα expression (23), we did not observe differences in ERα expression and localization between MCF-7p and MCF-7AKT cells (Fig. 1); thus, differences in ERα DNA binding between two cell types are independent of the expression and subcellular localization status of ERα.

Although much progress has been made in understanding estrogen signaling, it is still not clear how ERα binding to certain regions leads to activation, whereas binding to other regions leads to transcription repression. Even within a gene, we did not find a correlation between the number of ERα binding sites and the level of target gene induction/repression by estrogen. For example, although AKT increased the number of binding sites around the GREB-1 gene, estrogen-inducible expression was similar or lower in MCF-7AKT cells depending on the assay (Table 2, Table S3 in the supplemental material, and Fig. 5). It is possible that within a gene, ERα binding to certain regions has a negative effect on transcription, while binding to other regions has a positive effect. Chromatin conformation capture assays have revealed a physical interaction between distant enhancer-bound ERα with promoter-bound ERα or the transcription start site through chromatin looping and, depending on the binding site, these interactions may have positive or negative effects on transcription (12, 16). Alternatively, some of the ERα binding may be essential for other functions of estrogen, such as alternative splicing (4). How ERα binding to multiple sites on a gene leads to changes in transcription initiation (the majority of genes), relieving transcription attenuation (as in case of Myb) (18), alternative splicing (as in the case of SMRT, caspase 7, and AXIN [unpublished data]), and polyadenylation (yet to be explored) is an important question that needs to be addressed in the future in order to completely elucidate the effect of estrogen on the cellular proteome.

In summation, our studies reveal AKT-mediated enhancement of estrogen-regulated gene expression; these changes may consequently alter the way estrogen-regulated signaling integrates with TGF-β, NF-κB/TNF, retinoic acid, and E2F pathways. Some of these AKT-mediated changes may involve altered ERα and coactivator binding to the genome and consequent effects on primary and/or secondary estrogen response. Since growth factors activate multiple kinases, including AKT, a comprehensive understanding of growth factor and ERα-estrogen cross talk will require additional genome-wide studies that examine the effects of other kinases on ERα binding and estrogen-induced gene expression.