Diverse Gene Expression and Dna Methylation Profiles Correlate with Differential Adaptation of Breast Cancer Cells to the Antiestrogens Tamoxifen and Fulvestrant

The development of targeted therapies for antiestrogen-resistant breast cancer requires a detailed understanding of its molecular characteristics. To further elucidate the molecular events underlying acquired resistance to the antiestrogens tamoxifen and fulvestrant, we established drug-resistant sublines from a single colony of hormone-dependent breast cancer MCF7 cells. These model systems allowed us to examine the cellular and molecular changes induced by antiestrogens in the context of a uniform clonal background. Global changes in both basal and estrogen-induced gene expression profiles were determined in hormone-sensitive and hormonal-resistant sublines using Affymetrix Human Genome U133 Plus 2.0 Arrays. Changes in DNA methylation were assessed by differential methylation hybrid-ization, a high-throughput promoter CpG island microarray analysis. By comparative studies, we found distinct gene expression and promoter DNA methylation profiles associated with acquired resistance to fulvestrant versus tamoxifen. Fulvestrant resistance was characterized by pronounced up-regulation of multiple growth-stimulatory pathways, resulting in estrogen receptor A (ERA)–independent, autocrine-regulated proliferation. Conversely, acquired resistance to tamoxifen correlated with maintenance of the ERA-positive phenotype, although receptor-mediated gene regulation was altered. Activation of growth-promoting genes, due to promoter hypomethylation, was more frequently observed in antiestrogen-resistant cells compared with gene inactivation by promoter hypermethylation, revealing an unexpected insight into the molecular changes associated with endocrine resistance. In summary, this study provides an in-depth understanding of the molecular changes specific to acquired resistance to clinically important antiestrogens. Such knowledge of resistance-associated mechanisms could allow for identification of therapy targets and strategies for resensitiza-tion to these well-established antihormonal agents.


Introduction
The steroid hormone estrogen is strongly implicated in the development and progression of breast cancer (1).The primary mediator of estrogen action in breast cancer cells is estrogen receptor a (ERa), a ligand-activated transcription factor (1). Consequently, the leading drugs used for endocrine therapy of breast cancer all block ERa activity, including antiestrogens (i.e., tamoxifen and fulvestrant) and aromatase inhibitors (2).Despite the efficacy and favorable safety profile of these agents, the use of endocrine therapy is limited by the onset of drug resistance, in which most patients who initially respond to endocrine therapy eventually relapse (2).
In breast cancer cells, ERa can mediate ''genomic'' regulation of gene transcription and ''nongenomic'' activation of various protein kinase cascades [e.g., Src homology and collagen/growth factor receptor binding protein 2/SOS/mitogen-activated protein kinase (MAPK), phosphatidylinositol 3-kinase/AKT, and cyclic AMP (cAMP)/protein kinase A (PKA) pathways; ref . 3].As the transcriptional activity and target gene specificity of ERa and subsequent cellular response(s) to ligands are determined by complex combinatorial associations of ERa with coregulators, other transcription factors, and membrane-initiated signaling pathways (3)(4)(5)(6)(7), a myriad of receptor interactions may become altered during the acquisition of antiestrogen resistance.
ERa transcriptional activity is mediated by a constitutively active AF-1 and a ligand-regulated AF-2 (8).17h-Estradiol (E2), the primary ligand for ERa, binds to the ligand-binding domain (LBD) and induces a conformational change in the AF-2 domain, resulting in coregulator recruitment and transcription regulation followed by rapid ERa degradation (8,9).The antiestrogen tamoxifen, which competes with E2 for LBD binding, induces a conformational change distinct from the E2-ERa complex, leading to inactivation of the AF-2 domain and receptor stabilization (10,11).However, tamoxifen-bound ERa is capable of binding to DNA and regulating gene transcription, either directly, through the AF-1 domain, or indirectly, by sequestering coregulators away from other transcription factors (12).In addition, tamoxifen can act as an agonist to elicit nongenomic signaling through membrane ERa (13).These observations suggest that the action of tamoxifen is not limited to diminished estrogen-induced gene regulation.However, the mechanism(s) remains unclear of how the complex, multifactorial actions of this drug on gene expression and nongenomic signaling contribute to the acquisition of breast cancer tamoxifen resistance.
In contrast to tamoxifen, the antiestrogen fulvestrant is recommended for use in postmenopausal women whose disease has progressed after first-line endocrine therapies (such as tamoxifen and aromatase inhibitors).The mechanism of action of this so-called ''pure antagonist'' differs markedly from tamoxifen.Fulvestrant inhibits cytoplasm-to-nucleus ERa translocation, dimerization, and DNA binding of ERa as well as inducing its cytoplasmic aggregation, immobilization to the nuclear matrix, and proteasomal degradation (14).As a consequence of these actions, both ERa-mediated genomic gene regulation and nongenomic signaling are attenuated, leading to complete suppression of ERa signaling pathways (14,15).Despite the potent effects of fulvestrant, tumors eventually develop resistance to this selective ER down-regulator (16), although the underlying mechanism(s) of this phenomenon remains poorly understood.
Interrupting ERa function by antiestrogens can result in epigenetic modification of chromatin and altered gene expression (17,18).DNA methylation occurs in CpG dinucleotides, which are concentrated to form CpG islands in the promoter region of f70% of human genes (19).Hypermethylation of CpG islands in gene promoters often leads to inactivation of transcription, and an inverse relationship between promoter methylation levels and transcriptional activity has been well documented (20).Our recent studies show that depleting ERa with small interfering RNA (siRNA) in breast cancer cells triggers repressive chromatin modifications and DNA methylation in a set of ERa target promoters, resulting in transcriptional silencing of the corresponding genes (17).Whether interrupting ERa function by tamoxifen or fulvestrant can similarly affect DNA methylation patterns has not been explored.
The purpose of the current study was to identify molecular changes associated with acquired tamoxifen or fulvestrant resistance.To achieve this objective, we compared gene expression and DNA methylation profiles in estrogen-responsive MCF7 human breast cancer cells and tamoxifen-and fulvestrant-resistant MCF7 derivatives.Collectively, our results indicate that significant changes in downstream ERa target gene networks contribute to the acquisition of tamoxifen resistance; in contrast, loss of ERa signaling pathways, activation of compensatory growth-stimulatory cascades, and global remodeling of gene expression patterns underlie acquired resistance to fulvestrant.Finally, we report, for the first time, a prominent role for promoter hypomethylation of oncogenes in the acquisition of breast cancer antiestrogen resistance.
Preparation of cell extracts, immunoblotting, and luciferase assay.Before all experiments, MCF7-T and MCF7-F cells were cultured in hormone-free medium for 1 week to deplete any residual OHT or fulvestrant.Cells were cultured in basal medium for 3 days.Preparation of whole-cell extracts, immunoblotting, and luciferase analyses were done as described previously (21,22).To determine h-catenin activity, cells were transfected with TOPflash or FOPflash (23), along with CMVh-gal as internal control for transcription efficiency.h-Catenin activity was determined by dividing the OT-FLASH value by the OF-FLASH value.
Cell proliferation and clonogenicity assays.Cell proliferation assays were done as described previously (22).To examine clonogenic activity, cells were plated (300 per well) in six-well plates, cultured for 2 weeks, and stained with 0.5% methylene blue in 50% methanol.Colonies that contained z50 cells were scored.
RNA preparation and microarray hybridization.Cells were cultured in basal medium for 3 days and treated with E2 (10 À8 mol/L) for 4 hours.Total RNA was prepared using the Qiagen RNeasy Mini kit.A DNase I digestion step was included to eliminate DNA contamination.cRNA was generated, labeled, and hybridized to the Affymetrix Human Genome U133 Plus 2.0 Arrays by the Center for Medical Genomics at Indiana University School of Medicine (Bloomington, IN). 6 Microarray data analysis and validation.The hybridized Human Genome U133A 2.0 Array was scanned and analyzed using the Affymetrix Microarray Analysis Suite version 5.0.The average density of hybridization signals from four independent samples was used for data analysis, and genes with signal density <300 pixels were omitted from the data analysis.Ps were calculated with two-sided t tests with unequal variance assumptions, and a P value of <0.001 was considered to be significant.The following pair-wise comparisons were conducted: E2 versus untreated for each subline to identify E2-responsive genes and untreated MCF7-T or untreated MCF7-F versus untreated MCF7 to identify genes whose basal expression levels were altered in MCF7-T or MCF7-F.The fold change was described as a positive value when the expression level was increased and a negative value when the expression level was reduced.False discovery rate was set at 0.1 in the data analysis.To confirm the gene expression data from microarray analysis, quantitative PCR was used to examine the mRNA levels of a subset of genes (Supplementary Fig. S2).The quantitative PCR results showed a high degree of correlation to the microarray data.
Differential methylation hybridization.Genomic DNA was prepared using Qiagen DNeasy Tissue kit.Differential methylation hybridization was done as described previously (17) using a customized 60-mer oligonucleotide microarrays, which contain f44,000 CpG-rich fragments from f12,000 promoters of defined genes.The fold change in methylation density was described as a positive or negative value when methylation density was increased or decreased, respectively, compared with MCF7.

Results
Establishment and characterization of breast cancer cell lines with acquired antiestrogen resistance.The cell line MCF7 is a standard in vitro model for hormone-sensitive breast cancer

Cancer Research
Cancer Res 2006; 66: (24).December 15, 2006 (24).Consequently, we chose this cell line to investigate molecular changes associated with acquired resistance to tamoxifen and fulvestrant.MCF7 cultures are likely heterogeneous in nature (25); thus, to avoid selecting clonal variants with intrinsic drug resistance, we used a single estrogen-responsive MCF7 clone stably transfected with an ERa-responsive luciferase reporter (ERE-pS2-Luc; ref. 21) to derive sublines resistant to tamoxifen (MCF7-T) or fulvestrant (MCF7-F).The stably integrated ERE-pS2-Luc reporter was used to monitor ERa transcriptional activity.Because the three sublines used in the study were derived from a single MCF7 colony, cellular and molecular alterations observed in the drug-resistant sublines are likely due to an adaptive process in response to primary drug action.The overall scheme used to develop our model system is illustrated in Supplementary Fig. S1.
Cell morphology changes associated with acquired antiestrogen resistance are shown in Fig. 1A.MCF7-T cells were similar to MCF7 cells in appearance, growing as tightly packed colonies with limited cell spreading.MCF7-F cells, by contrast, showed reduced cell-cell contacts compared with MCF7 or MCF7-T cells and were loosely attached to the culture surface.
We next examined the expression levels of ERa mRNA and protein in the three sublines by quantitative PCR and immunoblot analyses, respectively (Fig. 1B).To avoid the effects of estrogen and antiestrogens, the cells were cultured in drug-free medium for 1 week followed by basal medium for 3 days before examining ERa content.Compared with MCF7 cells, ERa mRNA levels in MCF7-T and MCF7-F cells were decreased by 50% and 90%, respectively.Immunoblot analysis showed a 2-fold increase in ERa protein level in MCF7-T cells compared with a 90% decrease in receptor protein levels in MCF7-F cells.ERa transcriptional activity in these sublines was examined by monitoring the expression levels of the stably integrated ERE-pS2-Luc (Fig. 1C).Compared with MCF7, basal luciferase activity was higher in MCF7-T (f2-fold versus MCF7), likely due to the elevated protein level of ERa.E2 treatment increased ERE-pS2-Luc activity, which was inhibited by cotreatment with OHT or fulvestrant in both MCF7 and MCF7-T cells.
Figure 2. Expression of E2-responsive genes in MCF7, MCF7-T, and MCF7-F cells.A, Venn diagrams showing the numbers of E2-responsive genes in MCF7, MCF7-T, and MCF7-F cells.E2-responsive genes were defined as genes whose expression levels were up-regulated or down-regulated by E2 (10 À8 mol/L, 4 hours) by >2-fold and assigned to nine groups: up-regulation by E2 in all three sublines, MCF7 and MCF7-T only, MCF7 and MCF7-F only, MCF7 only, and MCF7-T only; down-regulation by E2 in MCF7 and MCF7-T only, MCF7 only, MCF7-T only, and MCF7-F only.B, two-dimensional hierarchical clustering of E2-responsive genes in MCF7, MCF7-T, and MCF7-F, untreated or E2 treated (10 À8 mol/L, 4 hours).Each row represents a single gene.Red, genes with high expression levels; green, genes with low expression levels.The similarities in the expression pattern among sublines are presented as a ''condition tree'' on the top of the matrix.C, Venn diagrams showing the number of E2-responsive genes that exhibited significant changes in basal expression levels in MCF7-T and MCF7-F cells.
Hatched areas with white lines, number of genes commonly altered in both MCF7-T and MCF7-F cells.Cutoff was set as fold change >2 (up-regulated or down-regulated, versus MCF7).
These observations suggest that ERa retains its transcriptional activity and sensitivity to different ligands in MCF7-T cells.By contrast, basal luciferase activity was dramatically elevated in MCF7-F (f20-fold versus MCF7), and no effect of E2, OHT, or fulvestrant on ERE-pS2-Luc expression was observed (Fig. 1C), showing that the integrated ERE-pS2-Luc reporter became constitutively activated through an ERa-independent mechanism in the fulvestrant-resistant subline.Together, these results indicate that the acquisition of tamoxifen resistance is associated with retention of functional ERa, whereas acquired fulvestrant resistance is accompanied by loss of ERa protein and E2-induced gene transactivation.
Response of MCF7-T and MCF7-F cells to estrogen, antiestrogens, and growth factors.We next examined growth rates of the three sublines and cell growth in response to E2, OHT, fulvestrant, EGF, and IGF-I (Fig. 1D).In basal growth medium, doubling times for MCF7 and MCF7-F were 6 and 5 days, respectively.By contrast, MCF7-T cells underwent growth arrest in this medium, showing only a 1.5-fold increase in cell number during a 9-day culture in basal medium, indicating that MCF7-T cells are dependent on a higher concentration of serum or the presence of OHT for proliferation.To compare the sensitivities to E2, OHT, and fulvestrant among MCF7, MCF7-T, and MCF7-F cells, dose responses were examined.E2 treatment increased the growth rate of MCF7 cells but showed no effect on MCF7-T or MCF7-F cells.OHT treatment inhibited the growth of MCF7 but not that of MCF7-T or MCF7-F.Fulvestrant inhibited the growth of MCF7 and, to a lesser extent, MCF7-T but exhibited no effect on MCF7-F cells.These observations are consistent with previous reports showing that OHT-resistant cells remained responsive to fulvestrant with a reduced sensitivity, whereas fulvestrant-resistant cells were crossresistant to OHT (26,27).Treatment with EGF or IGF-I increased the growth of MCF7 and MCF7-T cells.However, compared with MCF7 cells, MCF7-T cells were more sensitive to EGF but less sensitive to IGF-I.No effects of either EGF or IGF-I on MCF7-F cells were seen.Collectively, these results indicate that acquisition of resistance to tamoxifen and fulvestrant involves differential sensitivity to estrogen, antiestrogens, and growth factors.
Expression profiling of E2-responsive genes.To investigate whether aberrant changes in basal expression patterns and estrogen responsiveness of ERa target genes contribute to antiestrogen resistance, we analyzed gene expression patterns among MCF7, MCF7-T, and MCF7-F, untreated or treated with E2 (10 À8 mol/L) for 4 hours, as most direct ERa target genes are either induced or suppressed in that period (28,29).The Affymetrix Human Genome U133 Plus 2.0 Array, containing 47,000 probe sets for human transcripts, was used.A total of 360, 175, and 7 genes were found to be E2 responsive ( fold change z2, decreased or increased by E2) in MCF7, MCF7-T, and MCF7-F cells, respectively (Fig. 2A; Supplementary Table S1).Among the 360 E2-responsive genes identified in MCF7 cells, 89 (25%) were also similarly regulated by E2 in MCF7-T cells, whereas 271 (75%) were no longer inducible by E2 in the MCF7-T cells.Based on these results, we suggest that the acquisition of tamoxifen resistance is associated with altered regulation of a cohort of E2-inducible genes.The development of fulvestrant resistance, by contrast, is associated with almost complete loss of E2-induced gene regulation.
A two-dimensional (gene tree and condition tree) hierarchical clustering program 7 was used to analyze the expression patterns of the E2-responsive genes in MCF7, MCF7-T, and MCF7-F, untreated A, Venn diagrams showing the number of genes whose basal expression levels were altered (versus MCF7) in MCF7-T and MCF7-F cells.Hatched areas with white lines, number of genes commonly altered in both MCF7-T and MCF7-F cells.Cutoff was set as fold change >3 (up-regulated or down-regulated, versus MCF7).B, Venn diagrams showing the number of genes with altered promoter methylation intensity (versus MCF7) in MCF7-T and MCF7-F cells.Hatched areas with white lines, number of genes commonly hypermethylated or hypomethylated in both MCF7-T and MCF7-F cells.The cutoff was set as fold change in methylation intensity >2 (decreased or increased, versus MCF7).C, correlation of changes in promoter methylation to gene expression.Genes showing >2-fold change in relative methylation density (X axis ) and mRNA level (Y axis ) in comparison with MCF7 are depicted in the figure and divided into four categories: genes with promoter hypomethylation and increased expression (a); genes with promoter hypermethylation and increased expression (b); genes with promoter hypomethylation and decreased expression (c ); and genes with promoter hypermethylation and decreased expression (d).
or treated with E2 (10 À8 mol/L, 4 hours; Fig. 2B).Based on similarities in the expression profiles of the E2-responsive genes among sublines (presented as a ''condition tree'' on the top of the matrix in Fig. 2B), MCF7-T and MCF7 cells clustered together and MCF7-F cells clustered on a separate branch, suggesting that MCF7-T cells were more similar to the parental MCF7 than MCF7-F cells.
We then examined MCF7-T and MCF7-F cells for changes in basal expression levels of the 360 E2-responsive genes identified in MCF7 cells (Supplementary Table S1).The numbers of E2 up-regulated and down-regulated genes that showed significant changes in basal expression levels in MCF7-T or MCF7-F cells were presented in Fig. 2C.Among the total of 231 genes displaying altered basal expression in either MCF7-T or MCF7-F cells, only 39 (17%; Fig. 2C, hatched areas with white lines) were coordinately altered in both sublines.These results indicate an association between acquired resistance to tamoxifen and fulvestrant and altered basal expression levels of distinct subsets of E2-responsive genes.
Global gene expression profiles associated with acquired antiestrogen resistance.We did gene expression profiling of MCF7, MCF7-T, and MCF7-F cells after 3 days of culture in basal medium (Supplementary Table S2).We considered only those genes that displayed a 3-fold or greater change in basal expression level in the antiestrogen-resistant sublines.Altered expression of 371 genes was observed in MCF7-T cells, with nearly an equal number of upregulated and down-regulated genes (184 and 187, respectively; Fig. 3A).In MCF7-F cells, altered expression of 2,518 genes was observed, with more genes up-regulated (1,753 up-regulated versus 765 down-regulated; Fig. 3A).Only 138 genes were coordinately altered in both MCF7-T and MCF7-F cells (81 genes up-regulated and 57 genes down-regulated; Fig. 3A, shadowed areas with white lines), and 233 and 2,380 genes were uniquely altered in MCF7-T and MCF7-F, respectively (Fig. 3A).This result revealed that distinct molecular changes are associated with tamoxifen and fulvestrant resistance.Furthermore, the acquisition of fulvestrant resistance was associated with a dramatic remodeling of global gene expression, with gene up-regulation more prevalent that gene down-regulation.
Although the functions of the genes with altered expression in either MCF7-T or MCF7-F cells were diverse, they could be organized into different functional categories using the Kyoto Encyclopedia of Genes and Genomes database 8 and Gene Ontology algorithms. 9Signaling pathways coordinately altered at multiple levels and known to be involved in growth regulation are listed in Table 1.In MCF7-T cells, five families of genes were prominently altered: (a) PKA pathway, (b) caveolins, (c) Annexins and S100 calcium-binding proteins, (d) MAPK phosphatases, and (e) inhibitor of differentiation proteins.In addition, of the 371 altered genes in MCF7-T cells, 40% were E2-responsive genes (Supplementary Table S1), suggesting that remodeling of the ERa target gene network is a mechanism underlying acquisition of tamoxifen resistance.In MCF7-F cells, prominently altered pathways included EGFR and ErbB2 and related proteins, cytokines/ cytokine receptors, Wnt/h-catenin pathway, Notch pathway, and IFN signaling pathway/IFN-inducible genes (Table 1), showing an overall up-regulation of growth-stimulatory pathways in fulvestrant-resistant cells.
Correlation between genes altered in antiestrogen-resistant cells and known prognostic markers of breast tumors.To assess the potential clinical relevance of our findings, we examined the expression levels of multiple breast cancer prognostic markers (30,31) in MCF7-T and MCF-F.We observed up-regulation ( fold change z2; P < 0.001) of 16 and 47 poor prognostic markers in MCF7-T and MCF7-F, respectively (Supplementary Table S4).Conversely, we observed down-regulation of 4 and 9 good prognostic markers in MCF7-T and MCF7-F, respectively (Supplementary Table S4).Next, we examined the expression levels of genes previously associated with clinical outcome of breast tumors treated with tamoxifen (32)(33)(34).In MCF7-T, seven tamoxifenresistant markers were up-regulated, whereas three tamoxifenresponsive markers were down-regulated (Supplementary Table S5).Finally, we examined the expression levels of known ERa signature genes (30,35).In MCF7-T and MCF7-F, we observed down-regulation of 41 and 138 signature genes of ERa-positive tumors, respectively (Supplementary Table S6).Conversely, we observed up-regulation of 60 and 206 signature genes of ERanegative tumors in MCF7-T and MCF7-F, respectively (Supplementary Table S6).Collectively, our observations that subsets of potentially clinically relevant genes are altered in MCF7-T and MCF7-F support the notion that these cell lines may be valuable models for investigating the molecular events underlying the development of antiestrogen resistance in human breast cancer.
DNA methylation profiles associated with acquired resistance.Our recent studies showed that ERa depletion by siRNA in breast cancer cells led to progressive DNA methylation of genes normally regulated by ERa (17).This finding prompted us to investigate whether long-term treatment with tamoxifen or fulvestrant could cause changes in DNA methylation.Thus, we examined promoter methylation in using differential methylation hybridization and a customized 60-mer microarray containing 44,000 CpG-rich fragments from 12,000 promoters of defined genes.Genes showing altered promoter methylation intensities ( fold change z2, versus MCF7) are listed in Supplementary Table S3.In MCF7-F cells, 281 genes showed altered promoter methylation, with 240 (86%) hypomethylated genes (Fig. 3B).In MCF7-T, 160 genes showed altered promoter methylation, with 124 (77.5%) hypomethylated genes (Fig. 3B).Comparing the promoter methylation profiles, we found that only 16 promoters were commonly hypermethylated or hypomethylated in both resistant sublines (Fig. 3B, shadowed areas with white lines), suggesting that distinct sets of promoters are targeted for epigenetic modification by tamoxifen and fulvestrant.
By analyzing the methylation status of the 360 E2-responsive genes identified in MCF7 cells (Supplementary Table S1), we found a total of eight genes with altered methylation in MCF7-T (ID4 and FABP5) and MCF7-F (FHL2, FUT4, MICAL2, P2RY2, PIK3R3, and USP31 ).This observation suggests that the acquisition of antiestrogen resistance is not associated with changes in promoter methylation status of early E2-responsive genes.
To correlate changes in promoter methylation and basal gene expression levels, a linear regression analysis was done.An inverse correlation was observed between promoter methylation intensities and mRNA expression levels (P < 0.05; Fig. 3C), such as decreased basal expression levels were associated with increased promoter methylation.For a subset of genes listed in Table 2, increased mRNA expression levels were correlated with hypomethylation or vice versa (decreased mRNA expression levels were correlated with hypermethylation).Taken together, these  results show that MCF7-F and MCF7-T display highly divergent DNA methylation patterns; furthermore, promoter hypomethylation was more prevalent in antiestrogen-resistant sublines than in MCF7 cells.EGFR/ErbB2 and Wnt/B-catenin signaling pathways and antiestrogen resistance.As several signaling pathways showed coordinate alteration of multiple components in MCF7-T and MCF7-F (Table 1), we first examined the role of EGFR/ErbB2 in supporting estrogen-independent cell proliferation.Immunoblotting analysis revealed that EGFR is up-regulated and activated (phosphorylated) in MCF7-F and ErbB2 is up-regulated and activated (phosphorylated) in both MCF7-T and MCF7-F cells (Fig. 4A, top).Cell proliferation assays showed that 4557W (an inhibitor of both EGFR and ErbB2) and AG879 (an ErbB2-specific inhibitor) both inhibited cell proliferation of MCF7-T and MCF7-F but not MCF7 (Fig. 4A, bottom).We also examined the effect of PD15303, an EGFR-specific inhibitor.At 10 Amol/L PD15303, only the growth of MCF7-F was inhibited; however, 30 Amol/L PD15303 completely blocked MCF7-T and MCF7-F cell growth but only partially inhibited MCF7 growth (Fig. 4A, bottom).
Next, we examined the role of the Wnt/h-catenin pathway in supporting estrogen-independent cell growth.Immunoblotting analysis revealed that h-catenin is up-regulated in both MCF7-T and MCF7-F but only activated in MCF7-F, indicated by the presence of h-catenin in the nuclear fraction (Fig. 4B).To inhibit hcatenin activity, we used EGCG (36).Cell proliferation and clonogenicity assays were used to show that inhibiting h-catenin activity blocked MCF7-F cell growth but not MCF7-T or MCF7 (Fig. 4B, bottom).Reporter analysis using a TOPflash construct (23) confirmed that h-catenin-mediated gene transcription was increased in MCF7-F, which was eliminated by EGCG treatment (Fig. 4B, bottom).Taken together, these results show that the EGFR/ ErbB2 pathway plays an important role in supporting MCF7-T and MCF7-F cell growth as well as the involvement of h-catenin activation in fulvestrant resistance.

Discussion
Based on the unique molecular actions of tamoxifen and fulvestrant, we hypothesized that the two antiestrogens induce distinct adaptive responses in breast cancer cells and subsequently promote the emergence of drug-resistant cells with specific molecular characteristics.To test this possibility, we generated breast cancer cells with acquired resistance to either tamoxifen or fulvestrant and did global gene expression and DNA methylation analyses on the resistant cells.In our model system, to avoid clonal selection of variants with intrinsic drug resistance from a heterogeneous population, we isolated a single estrogen-responsive MCF7 clone to subsequently derive the tamoxifen-and fulvestrantresistant sublines.Hormone-free medium was used during the selection process to exclude interference from estrogens.Consequently, the cellular and molecular changes identified in our model systems can be considered ''acquired traits'' in response to tamoxifen and fulvestrant treatment.Although originating from the same MCF7 clone, the sublines developed strikingly divergent phenotypes (Fig. 1) and molecular characteristics (i.e., gene expression and promoter methylation patterns).
In MCF7-T cells, expression of a functional ERa was maintained, and the cells responded to E2 treatment with altered gene expression (Fig. 1B; Supplementary Table S1).Although E2stimulated cell growth was no longer observed in the MCF7-T subline, the cells retained a transcriptional response to E2 and remained sensitive to growth inhibition by fulvestrant (Fig. 1D), suggesting that ERa signaling continues to contribute to growth regulation after the acquisition of tamoxifen resistance.Comparative analysis revealed that E2-responsive gene profiles were markedly different between MCF7 and MCF7-T (Fig. 2), suggesting that different groups of genes were targeted by ERa in the parental MCF7 cells compared with the tamoxifen-resistant subline.
Analysis of basal gene expression levels revealed a subset of 371 genes with altered expression in MCF7-T cells; a significant number of these (f40%) were E2 responsive, suggesting that genes normally regulated by ERa are targeted for molecular alteration during acquisition of tamoxifen resistance.Based on these findings, we suggest that breast cancer cells with acquired tamoxifen resistance continue to use ERa to support cell growth/survival but through an altered ER target gene network.Functional analysis of altered genes in MCF7-T cells revealed that several signaling pathways were coordinately up-regulated at multiple levels, including PKA pathway, caveolins, Annexins and S100 calciumbinding proteins, MAPK phosphatases, and inhibitor of differentiation proteins (Table 1).Deregulation of these pathways has previously been implicated in breast cancer pathogenesis (37)(38)(39)(40)(41), but their precise roles in tamoxifen action and acquired resistance remain to be established.
In contrast to MCF7-T cells, the MCF7-F cells showed dramatically reduced expression of ERa and were refractory to E2-induced gene regulation and growth stimulation (Fig. 1; Supplementary Table S1).A large number of signature genes of ERa-positive tumors were significantly down-regulated in MCF7-F, suggesting

Cancer Research
Cancer Res 2006; 66: (24).December 15, 2006 that acquired fulvestrant resistance is coupled with the generation of ERa-negative phenotype.One striking observation from the global gene expression analysis is the up-regulation of multiple growth-regulatory pathways in MCF7-F cells, including EGFR/ ErbB2 and related proteins, cytokines/cytokine receptors, Wnt/hcatenin pathway, and Notch pathway (Table 1).We showed that both EGFR/ErbB2 and Wnt/h-catenin pathways play a role in supporting estrogen-independent cell growth of MCF7-F, and the contribution of the other signaling pathways to the development of the resistant phenotype remains unclear.
Although aberrant promoter methylation is an early event in tumorigenesis and frequently observed in breast tumors (42), to our knowledge, a role for DNA methylation in remodeling gene expression patterns associated with acquired antiestrogen resistance has not been reported.Our genome-wide promoter methylation analysis of MCF7-T and MCF7-F cells showed that tamoxifen and fulvestrant can cause hypermethylation or hypomethylation of particular CpG-rich loci, resulting in distinct promoter methylation patterns.We predicted, based on our previous study (17), that inhibition of ERa signaling would result primarily in gain of methylation on promoter regions of ERa direct target genes (i.e., hypermethylation of CpG-rich loci).In contrast to our hypothesis, the promoter methylation status of only eight E2-responsive genes (FABP5, FHL2, FUT4, ID4, MICAL2, P2RY2, PIK3R3, and USP31) was found to be altered in the antiestrogenresistant cells.One possible explanation is that, although most early E2-responsive genes in MCF7 cells are involved in cell growth control, their inactivation by promoter methylation could result in cell growth arrest or death.Thus, only cells without hypermethylation of ERa target genes, perhaps due to defective DNA methylation, may be able to escape the detrimental effects of antiestrogens.In support of this possibility, promoter hypomethylation was more prevalent than promoter hypermethylation in the antiestrogen-resistant sublines.Intriguingly, our observation agrees with a previous study reporting that the DNA methylation inhibitor 5-azacytidine promoted the generation of antiestrogen resistance colonies from hormone-sensitive breast cancer ZR-75-1 cells (43).
Most current studies on cancer-related DNA methylation have been focused on suppression-linked promoter hypermethylation of tumor suppressors (44).However, a correlation between hypomethylation of promoter regions and transcriptional activation of tumor-promoting genes in tumors has been described (45,46).Several genes that showed increased basal expression levels and promoter hypomethylation in MCF7-T or MCF7-F cells were found to be up-regulated in cancer cells and possess oncogenic activity, such as CDH2, ID4, ANXA4, BRAF, CTNNB1, and Wnt11 (47)(48)(49)(50).Taken together, our results suggest that promoter hypomethylation plays a role in the development of antiestrogen resistance.Further studies are required to elucidate how other epigenetic events, such as histone modification and chromatin remodeling, contribute to altered promoter methylation and acquired antiestrogen resistance in breast cancer cells.
In this first study to provide a detailed analysis of the ERa target gene network, global gene expression, and DNA methylation profiles in tamoxifen-and fulvestrant-resistant cells, we show that the acquisition of resistance to tamoxifen and fulvestrant involves distinctly different pathways (summarized in Supplementary Fig. S3).Tamoxifen resistance is associated with the maintenance of the ERapositive phenotype and use of an altered ERa signaling network to promote cell proliferation/survival.Acquired resistance to fulvestrant is an ERa-independent phenomenon, using multiple growthstimulatory pathways to establish autocrine-regulated proliferation.

Figure 1 .
Figure 1.Characterization of tamoxifenand fulvestrant-resistant sublines.A, phase-contrast photomicrographs of MCF7, MCF7-T, and MCF7-F cells.All cells were in log growth phase.Magnification, Â10.B, ERa mRNA levels were determined by quantitative PCR and normalized to ERa mRNA level in MCF7 cells.Columns, mean (n = 3); bars, SE.ERa protein levels were determined by immunoblotting with a specific ERa antibody.GAPDH was used as loading control.To determine the relative level of ERa in MCF7-F cells, the ERa level in 50 Ag MCF7-F protein extract was compared with that in various amounts of MCF7 protein extracts.C, ERa transcriptional activity was determined by measuring luciferase expression driven by the stably integrated reporter ERE-pS2-Luc.Cells were treated with indicated doses of E2 alone (left), 10 À8 mol/L E2 in combination with indicated doses of OHT (middle ), or 10 À8 mol/L E2 in combination with indicated doses of fulvestrant (right ).Luciferase activity (unit/Ag protein) was normalized with protein concentration.Points, mean (n = 4); bars, SE.D, comparison of cell growth rates among MCF7, MCF7-T, and MCF7-F.To determine growth rates in basal medium, cells were plated in 96-well dishes (2,000 per well) in basal medium for the indicated time and cell numbers were determined by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay.Relative cell growth rates (drug versus vehicle) in the presence of indicated doses of E2, OHT, fulvestrant, EGF, and IGF-I were also examined, and cell numbers were determined by MTT assay after 7-day treatment.Points, mean (n = 6); bars, SE.

Figure 3 .
Figure 3. Alterations in global gene expression and promoter methylation.A, Venn diagrams showing the number of genes whose basal expression levels were altered (versus MCF7) in MCF7-T and MCF7-F cells.Hatched areas with white lines, number of genes commonly altered in both MCF7-T and MCF7-F cells.Cutoff was set as fold change >3 (up-regulated or down-regulated, versus MCF7).B, Venn diagrams showing the number of genes with altered promoter methylation intensity (versus MCF7) in MCF7-T and MCF7-F cells.Hatched areas with white lines, number of genes commonly hypermethylated or hypomethylated in both MCF7-T and MCF7-F cells.The cutoff was set as fold change in methylation intensity >2 (decreased or increased, versus MCF7).C, correlation of changes in promoter methylation to gene expression.Genes showing >2-fold change in relative methylation density (X axis ) and mRNA level (Y axis ) in comparison with MCF7 are depicted in the figure and divided into four categories: genes with promoter hypomethylation and increased expression (a); genes with promoter hypermethylation and increased expression (b); genes with promoter hypomethylation and decreased expression (c ); and genes with promoter hypermethylation and decreased expression (d).

Figure 4 .
Figure 4. Roles of EGFR/ErbB2 and Wnt/h-catenin pathways in estrogenindependent proliferation of MCF7-T and MCF7-F.A, inhibition of EGFR/ErbB2 activity prevents cell growth of MCF7-T and MCF7-F.EGFR and ErbB2 protein and phosphorylation levels in whole-cell lysates were examined by immunoblots.To examine cell growth rates, MCF7 cells (in growth medium), MCF7-T cells (in hormone-free medium with 100 nmol/L OHT), and MCF-F cells (in hormone-free medium with 100 nmol/L fulvestrant) were treated with EGFR/ErbB2 inhibitors as indicated.Cell numbers were determined by MTT assay after a 7-day treatment period.Points, mean (n = 6) relative cell growth rate (drug versus vehicle); bars, SE.B, inhibition of h-catenin activity prevents cell growth of MCF7-F.Level of h-catenin protein in whole-cell lysate and nuclear fraction was examined by immunoblots.Cell growth rates in the presence of EGCG were determined as in (A).To examine clonogenic activity, cells were treated with EGCG for 2 weeks.Colonies that contain >50 cells were scored.Points, mean (n = 6) relative clonogenic activity (drug versus vehicle); bars, SE.To examine h-catenin transcription activity, cells were transfected with TOPflash or FOPflash construct and treated with EGCG for 16 hours as indicated.The transcription activity of h-catenin was presented as the ratio of TOPflash against FOPflash.

Table 2 .
Genes with altered expression and promoter methylation in MCF7-T and MCF7-F cells

Table 2 .
Genes with altered expression and promoter methylation in MCF7-T and MCF7-F cells (Cont'd)