Most invasive lobular breast cancers (ILC) are of the luminal A subtype and are strongly hormone receptor–positive. Yet, ILC is relatively resistant to tamoxifen and associated with inferior long-term outcomes compared with invasive ductal cancers (IDC). In this study, we sought to gain mechanistic insights into these clinical findings that are not explained by the genetic landscape of ILC and to identify strategies to improve patient outcomes. A comprehensive analysis of the epigenome of ILC in preclinical models and clinical samples showed that, compared with IDC, ILC harbored a distinct chromatin state linked to gained recruitment of FOXA1, a lineage-defining pioneer transcription factor. This resulted in an ILC-unique FOXA1–estrogen receptor (ER) axis that promoted the transcription of genes associated with tumor progression and poor outcomes. The ILC-unique FOXA1–ER axis led to retained ER chromatin binding after tamoxifen treatment, which facilitated tamoxifen resistance while remaining strongly dependent on ER signaling. Mechanistically, gained FOXA1 binding was associated with the autoinduction of FOXA1 in ILC through an ILC-unique FOXA1 binding site. Targeted silencing of this regulatory site resulted in the disruption of the feed-forward loop and growth inhibition in ILC. In summary, ILC is characterized by a unique chromatin state and FOXA1–ER axis that is associated with tumor progression, offering a novel mechanism of tamoxifen resistance. These results underscore the importance of conducting clinical trials dedicated to patients with ILC in order to optimize treatments in this breast cancer subtype.

Significance:

A unique FOXA1–ER axis in invasive lobular breast cancer promotes disease progression and tamoxifen resistance, highlighting a potential therapeutic avenue for clinical investigations dedicated to this disease.

See related commentary by Blawski and Toska, p. 3668

Invasive lobular carcinoma (ILC) is the second most common histologic subtype of breast cancer, accounting for 10% to 15% of all cases. The classic variant of ILC is characterized by relatively uniform and noncohesive cells that grow as a single file infiltrating the stroma. This growth pattern is attributed to the loss of E-cadherin, the hallmark of ILC (1, 2), and can render physical exam and mammographic diagnosis challenging (3, 4). Although the majority of ILCs are of low or intermediate grade, express high levels of estrogen receptor (ER), and of luminal A subtype, several studies suggest that ILC long-term outcomes are inferior to stage-matched invasive ductal carcinoma (IDC; ref. 5). Additionally, the retrospective analysis of the BIG-1-98 adjuvant endocrine trial demonstrated that the magnitude of the inferior benefit from tamoxifen compared with an aromatase inhibitor (AI) was greater in ILC compared with IDC (6), indicative of relative resistance to tamoxifen treatment in early-stage treatment-naïve ILC.

Several studies have revealed differences in the mutational landscape of ILC compared with IDC. In addition to the loss of E-cadherin, ILC is characterized by a higher frequency of FOXA1 mutations (7, 8), which are found in 7% to 9% of primary ILC versus 2% in primary IDC. In primary, treatment-naïve, ILC tumors, FOXA1 mutations are associated with increased expression of FOXA1 and unique transcriptional profiles (7). More recently, FOXA1 mutations were shown to be associated with endocrine resistance in metastatic ER-positive (ER+) breast cancer (9). FOXA1 is a pioneer transcription factor (TF) that mediates ER transcriptional activity (10–13). Thus, the increase in FOXA1 mutations in ILC and the disparate patterns of response to the different classes of endocrine treatment suggest that ILC may have a unique ER axis compared with IDC. In support of this notion, previous work has shown that ILC cells are characterized by a unique transcriptional response to estrogen (14). However, the mechanism by which the ER transcriptional axis is altered in ILC models lacking FOXA1 mutations remains elusive. In this study, we performed a comprehensive analysis of the epigenome of ILC in preclinical models and clinical samples with the aim to provide mechanistic insights to explain the distinctive responses to estradiol and tamoxifen in ILC versus IDC and to inform us on new potential therapeutic approaches for ILC.

Cell lines

MCF7, T47D, and MDA-MB-134-VI (MDAMB134) cells were purchased from ATCC; SUM44PE (SUM44) cells were a gift by Dr. Stephen Ethier. All the cells were authenticated by short tandem repeat profiling at Bio-Synthesis and regularly tested for Mycoplasma contamination by the Mycoalert Detection Kit (Lonza). MCF7 cells were maintained in DMEM and T47D cells in RPMI supplemented with 10% heat-inactivated FBS and 1% penicillin/streptomycin (P/S). MDAMB134 cells were maintained in L-15 with 20% FBS and 1% P/S. SUM44 were maintained in Ham's F-12 supplemented with insulin (5 μg/mL), hydrocortisone (1 μg/mL), fungizone (2.5 μg/mL), transferrin (5 μg/mL), T3 (6.6 ng/mL), ethanolamin (5 mmol/L), NaSe (8.7 ng/mL; all from Sigma-Aldrich), gentamicin (25 μg/mL), HEPES (10 mmol/L), BSA (0.1%; all from Thermo Fisher Scientific). For hormone-depleted (HD) conditions, cells were kept in phenol-red free medium supplemented with 10% heat-inactivated charcoal-stripped (CS)-FBS (with the exception of SUM44; 20% for MDAMB134) and 1% P/S. MCF7, T47D, and SUM44 were incubated at 37°C in 5% CO2. MDAMB134 cells were incubated at 37°C without CO2.

Human tissue studies

The ILC metastatic ascitic fluid collection was conducted in accordance with the Declaration of Helsinki ethical guidelines and approved by the Dana-Farber/Harvard Cancer Center institutional review board (Protocol 93-085). Participants signed written informed consent before the collection. Red blood cells and dead cells were removed by Ficoll (Sigma-Aldrich) followed by the selection of cancer epithelial cells using EpCam selective beads (Dynabeads). Cancer epithelial cells were subjected to FOXA1, ER, and H3K27ac chromatin immunoprecipitation sequencing (ChIP-seq).

ATAC-seq

For the global chromatin accessibility experiments, ductal and lobular cell lines were cultured for three days in HD conditions and then treated with 10 nmol/L estradiol for 45 minutes. Nuclei of 100,000 cells fixed in 1% formaldehyde were processed as previously reported (15). Briefly, cells were resuspended in 1 mL of cold assay for transposase-accessible chromatin sequencing (ATAC-seq) resuspension buffer (RSB; 10 mmol/L Tris-HCl pH 7.4, 10 mmol/L NaCl, and 3 mmol/L MgCl2 in water) and centrifuged. Cell pellets were then resuspended in ATAC-seq RSB (0.1% NP40, 0.1% Tween-20, and 0.01% digitonin) and incubated on ice. After lysis, ATAC-seq RSB containing 0.1% Tween-20 (without NP40 or digitonin) was added. Nuclei were centrifuged and then were resuspended in 50 μL of transposition mix (25 μL 2× TD buffer, 2.5 μL transposase (100 nmol/L final), 16.5 μL PBS, 0.5 μL 1% digitonin, 0.5 μL 10% Tween-20, and 5 μL water). Transposition reactions were incubated at 37°C for 30 minutes. Reactions were cleaned up with Zymo DNA Clean and Concentrator5 columns.

ChIP sequencing

ChIP experiments were conducted as described previously (16) and were done in duplicates. MCF7, T47D, MDAMB134, and SUM44 cells were cultured for three days in HD conditions and then treated with 10 nmol/L estradiol for 45 minutes. For the ChIP experiments with tamoxifen, we used 10 nmol/L 4-hydroxytamoxifen for 45 minutes. Chromatin from 20 million formaldehyde-fixed cells was sonicated to a size range of 200 to 300 bp. Solubilized chromatin was immunoprecipitated with a mix of the ER antibodies Ab10 (Thermo Fisher Scientific) and SC-543 (Santa Cruz), a mix of the FOXA1 antibodies ab5089 and ab23738 (Abcam), GATA3 antibody D13C9 (Cell Signaling) or H3K27ac antibody (C15410196, Diagenode). The same antibodies were used in each experiment for all the cell lines. The samples were reversed crosslinked, treated with proteinase K, and DNA was extracted. Libraries were sequenced using 75 bp paired-end reads on the Illumina Nextseq500 at the Dana-Farber Cancer Institute.

RNA sequencing

Ductal and lobular cell lines were cultured for three days in HD conditions. After washing the cells with PBS, cells were incubated for 12 hours in HD medium with 10 nmol/L estradiol, or DMSO treatment. Total RNA was extracted using the QIAGEN RNeasy kit with DNase digestion. RNA concentrations were measured by NanoDrop, and the quality of RNA was determined by a Bioanalyzer. For all cell line studies, samples were analyzed in at least duplicates. RNA-seq libraries were made using the TruSeq RNA Sample Preparation Kit (Illumina). Samples were sequenced on an Illumina Nextseq500.

Generation of knockout cells (CRISPR, ShRNA, and siRNA)

CRISPR/Cas9 cells

Construction of lenti-CRISPR/Cas9 vectors targeting E-cadherin (CDH1) in the two ductal cell lines was performed following the protocol associated with the backbone vector (49535, Addgene). The following sgRNA for CDH1 sequences were used:

  • gRNA1 F:5′ CACCGAGCTGGCTGACATGTACGG 3′

  • gRNA1 R: 5′ AAACCCGTACATGTCAGCCAGCTC 3′

DOX-inducible ShFOXA1

FOXA1 shRNA constructs (gift from Novartis) were generated by inserting annealed oligonucleotides into EcoRI/AgeI-digested pLKO-Tet-On plasmid.

The FOXA1 sequence to target was: 5′ GCAGATGTCTTTAAATGAAAT.

Top oligo: CCGGGCAGATGTCTTTAAATGAAATCTCGAGATTTCATTTAAAGACATCTGCTTTTT

Bottom oligo: AATTAAAAGCAGATGTCTTTAAATGAAATCTCGAGATTTCATTTAAAGACATCTGC

Lentivirus was produced in 293T cells to infect cells in media containing polybrene (8–10 μg/mL). Cells were selected with puromycin treatment after transduction. 1 μg/mL of doxycycline was used for the induction of the shFOXA1.

siRNA

MCF7 and MDA134 cells were seeded in 6- or 24-well plates and transfected with 20 nmol/L siRNA oligos by Lipofectamine RNAiMax reagent (Life Technology). Knockdown efficiency was determined after 72 hours of transfection. Cell counting was performed after 3 and 7 days of transfection. The ON-TARGET siRNA oligos targeting ESR1 (LQ-003401-00), FOXA1 (J-010319-05 and J-010319-06), CASZ1 (LQ-020764-02), or nontargeting control (D-001810-01) were all purchased from Dharmacon.

CRISPRi FOXA1 enhancer

Stable dCas9-KRAB-expressing MCF7 and SUM44 cell lines were generated by lentiviral transduction using Lenti-dCas9-KRAB-blast (Addgene #89567) transfer vector and pMD2.G (Addgene; #12259) and psPAX (Addgene; #12259) lentiviral packaging vectors. Transduced cells were selected and maintained in blasticidin (10 μg/mL) selective culture media. Guide RNAs (gRNA) were designed against peak 1 and nonhuman targeting control (NTC) gRNA was included and cloned as described previously (detailed protocol available at http://www.broadinstitute.org/rnai/public/resources/protocols). Briefly, for each gRNA, complementary single-stranded oligonucleotides were synthesized (Invitrogen), phosphorylated, annealed, and gRNA cassettes were ligated into pXPR_BRD003 gRNA expression vector containing puromycin selection marker. After bacterial transformation (Stbl3 cells, Invitrogen) individual clones were picked and regenerated, and correct gRNA sequences were confirmed by Sanger sequencing.

Lentiviral particles were generated for each sgRNA experiment by transforming HEK-293T cells with gRNA transfer vectors and lentiviral packaging mix (pMD2.G psPAX). Lentiviral particle containing media were collected and filtered using 0.45-μm pore size syringe filters (Corning) after 48 hours posttransfection and used for the treatment of previously plated breast cancer cell lines. Media were changed after 24 hours and replaced by puromycin (2 μg/mL) containing the selective culture medium. Puromycin-resistant cells were collected after 7 days of selection for proliferation studies and RNA for cDNA.

The coordinates of the region of interest (P1): chromosome 14 from 38,067,100 to 38,071,000, Human hg19. Working guides were:

  • gP1: caccgAGGAGCTACTAGACCAGTAA and aaacTTACTGGTCTAGTAGCTCCTC

  • NTC: caccGCGACCCAAATGCACCCTTT and aaacAAAGGGTGCATTTGGGTCGC

Quantification and statistical analysis

Statistical analyses for cell proliferation studies were performed using two-sided Student t tests, and P values less than 0.05 were considered statistically significant. Error bars represent the ± SEM. Cell line experiments testing cell proliferation were all performed in triplicates.

Survival analysis

Calculation of average modified Z-score (AveMZ)

Breast tumor gene-expression profiles with clinicopathologic data were obtained from METABRIC (17, 18) via authorized access in Synapse (synapse.sagebase.org). To calculate the FOXA1-ILC_120 gene set expression in the METABRIC data set, we used the modified Z-score to transform the gene-expression data because it relies on the median and is less influenced by outliers when compared with the standard Z-score based on the mean. The score was calculated by the formula of (X-MED)/(1.486*MAD), where X is the log-transformed gene-expression value. MED is the median level of X across samples, and MAD (≠ 0) is the median absolute deviation calculated by MAD = median (|X − median(X)|). The signature score of the target gene set was presented as AveMZ by calculating the mean of modified Z-scores of the signature genes.

Kaplan–Meier plots were produced using the R survminer v0.4.8 package to display the survival probabilities per group as a function of time. Groups of ER+ breast cancer cohorts were stratified by the median cutoff of the gene signature score. Hazard ratio (HR) of stratum and P value were calculated using the Wald estimates and likelihood ratio test, respectively, in the Cox proportional hazards model using the R survival v3.2-3 package.

Additional methods are in the Supplementary Materials.

Data availability

The whole-genome sequencing, RNA-seq, ChIP-seq, ATAC-seq, and Hi-ChIP-seq data have been deposited in the Gene-Expression Omnibus database under GSE152367.

ILC has a unique chromatin state that is tightly linked to FOXA1 recruitment

To study the epigenetic landscape and identify the genome-wide active regulatory elements accessible to TF binding in ILC versus IDC, we first conducted ATAC-seq in cell line models of ER+ ILC [(MDA-MB-134-VI (MDAMB134) and SUM44PE (SUM44)] and ER+ IDC (MCF7 and T47D). Unsupervised sample-to-sample correlation analysis segregated the ILC from IDC models, revealing a fundamental unique chromatin state of ILC (Fig. 1A). Differential analysis (L2FC>1 and <−1, q < 0.01) showed 11,777 sites significantly gained in ILC and 5,444 sites significantly gained in IDC (Fig. 1B; Supplementary Fig. S1A). The top enriched motifs in the ILC-gained sites were FOXA1 motifs, followed by motifs of AP2γ, a key regulator of ER (19), and the estrogen response element (ERE; Fig. 1C; Supplementary Fig. S1A). In contrast, the CTCF motif was the top motif in the IDC gained and the nondifferentiated accessible sites (Fig. 1D; Supplementary Fig. S1A).

Figure 1.

ILC has a unique chromatin cell state. A, Sample-to-sample correlation of chromatin accessibility based on ATAC-seq by the Euclidean distance between rows/columns and Ward's method of ILC cells [MDAMB134 (MDA134) and SUM44] and IDC cells (MCF7 and T47D) after 10 nmol/L β-estradiol (E2) stimulation (cells were grown in HD conditions for 3 days, followed by a 45-minute treatment with 10 nmol/L E2). Shown in the plot are results of replicates. B, Tornado plots of chromatin-accessible sites gained in ILC cells (11,777 peaks; blue) and gained in the IDC cells (5,444 peaks; red; Log2FC > 1 or <−1, Q < 0.01). Chromatin-accessible sites are shown in a horizontal window of ±2 kb from the peak center. C and D, Ranking of motifs enriched in the ILC- (C) and IDC- (D) gained accessible sites based on P value. E, Sample-to-sample correlation heatmap of open chromatin sites in TCGA ER+ breast cancer tumors applying only the chromatin-accessible sites gained in the ILC cell line models (11,777 peaks). Samples are clustered by the Euclidean distance between rows/columns and Ward's method. Samples cluster to three groups including an ILC-enriched group (Fisher exact test). F, Tornado plots of chromatin-accessible sites lost when FOXA1 is downregulated by a DOX-inducible shRNA after 3 days of DOX in presence of HD and 45 minutes 10 nmol/L E2. G and H, Venn diagrams of chromatin-accessible sites upregulated in MDA134 in comparison with MCF7 (red; G) or upregulated in lobular cells in comparison with ductal cells (red; H) and the chromatin-accessible sites lost by downregulation of FOXA1 by shRNA (green).

Figure 1.

ILC has a unique chromatin cell state. A, Sample-to-sample correlation of chromatin accessibility based on ATAC-seq by the Euclidean distance between rows/columns and Ward's method of ILC cells [MDAMB134 (MDA134) and SUM44] and IDC cells (MCF7 and T47D) after 10 nmol/L β-estradiol (E2) stimulation (cells were grown in HD conditions for 3 days, followed by a 45-minute treatment with 10 nmol/L E2). Shown in the plot are results of replicates. B, Tornado plots of chromatin-accessible sites gained in ILC cells (11,777 peaks; blue) and gained in the IDC cells (5,444 peaks; red; Log2FC > 1 or <−1, Q < 0.01). Chromatin-accessible sites are shown in a horizontal window of ±2 kb from the peak center. C and D, Ranking of motifs enriched in the ILC- (C) and IDC- (D) gained accessible sites based on P value. E, Sample-to-sample correlation heatmap of open chromatin sites in TCGA ER+ breast cancer tumors applying only the chromatin-accessible sites gained in the ILC cell line models (11,777 peaks). Samples are clustered by the Euclidean distance between rows/columns and Ward's method. Samples cluster to three groups including an ILC-enriched group (Fisher exact test). F, Tornado plots of chromatin-accessible sites lost when FOXA1 is downregulated by a DOX-inducible shRNA after 3 days of DOX in presence of HD and 45 minutes 10 nmol/L E2. G and H, Venn diagrams of chromatin-accessible sites upregulated in MDA134 in comparison with MCF7 (red; G) or upregulated in lobular cells in comparison with ductal cells (red; H) and the chromatin-accessible sites lost by downregulation of FOXA1 by shRNA (green).

Close modal

To test the clinical relevance of these findings, we analyzed the ER+ breast cancer samples of The Cancer Genome Atlas Program (TCGA) that were included in the Pan-Cancer ATAC-seq study (20). This analysis consisted of 58 ER+ breast cancers with different histologic subtypes (IDC: N = 38, ILC: N = 14 and other histologic subtypes: N = 6). None of these samples harbored FOXA1 mutations. Unsupervised sample-to-sample correlation of the chromatin-accessible sites gained in the ILC model cell lines segregated the samples into three main clusters: a FOXA1 low cluster (biologically an ER-negative cluster), ILC-enriched cluster (including 11 of the 14 ILC samples, Fisher exact test P <0.001), and IDC-enriched cluster (Fig. 1E). Thus, ILC can be characterized and identified by the unique chromatin-accessible sites.

Because FOXA1 was the top motif enriched in the ILC-gained chromatin-accessible sites, we next tested the impact of FOXA1 silencing on chromatin accessibility by engineering MDAMB134 cells that stably express a doxycycline (DOX) inducible shFOXA1 (Supplementary Fig. S1B). Silencing of FOXA1 resulted in the loss of 6,215 accessible sites (L2FC <−1, q < 0.01) without a gain in accessible sites (Fig. 1F). In control cells not expressing the shFOXA1 construct, DOX treatment had no significant effect on chromatin accessibility (L2FC <−1, q < 0.01; Supplementary Fig. S1C). Importantly, the lost sites significantly overlapped with the chromatin-accessible sites gained in MDAMB134 compared with MCF7 cells or gained in ILC versus IDC cell models (Fig. 1G and H).

Because we showed that FOXA1 has a role in facilitating the ILC-unique chromatin state, we performed FOXA1 ChIP-seq in all four models and primary ILC cells obtained from a malignant peritoneal effusion from a patient with metastatic ER+ ILC (primary ILC Met). The majority of FOXA1 binding sites were in enhancer regions, including intronic and intergenic regions (Supplementary Fig. S2A). FOXA1 binding clustered the ILC models, including the primary ILC Met cells, and separated them from the IDC models (Fig. 2A). This clustering was driven by 12,247 ILC-gained peaks (L2FC > 1, q < 0.01; Fig. 2B). In contrast, there was a marginal number of peaks gained in IDC. As expected, FOXA1 motifs were the most significantly enriched motifs in the gained and nondifferentiated peaks (Fig. 2B). In addition, the ERE motif was among the enriched motifs (Fig. 2B). In support of the role of FOXA1 in the ILC-unique chromatin state, the sites with gained chromatin accessibility in ILC had increased FOXA1 binding in ILC cells (Fig. 2C). In addition, the chromatin-accessible sites that were lost in the MDAMB134 cells after FOXA1 silencing had increased FOXA1 binding in MDAMB134 cells compared with MCF7 cells (Fig. 2D). Strikingly, all the chromatin-accessible sites gained in ILC overlapped with the ILC-gained FOXA1 chromatin recruitment (Fig. 2E). Moreover, there was a significant correlation between the intensity of the ILC-gained FOXA1 binding and the intensity of gained chromatin accessibility in MDAMB134 and SUM44 cells (Spearman correlation coefficient = 0.327, P < 8.33e−98 for MDAMB134 and Spearman correlation coefficient = 0.24, P < 1.72e−50 for SUM44; Fig. 2F; Supplementary Fig. S2B). In aggregate, these results indicate that ILC has a unique chromatin state that is interconnected with the reprogramming of FOXA1 recruitment.

Figure 2.

FOXA1 reprogramming in ILC is linked to the ILC-unique chromatin state. A, Sample-to-sample correlation (Euclidean distance between rows/columns and Ward's method) of FOXA1 binding sites correlation plots between all four cell lines in replicates [MCF7, T47D, MDA134 (MDAMB134), and SUM44] and the primary ILC cells isolated from a malignant peritoneal effusion from a patient with ER+ metastatic ILC (ILC Met). B, Tornado plots of FOXA1 binding sites (12,427 sites) gained in ILC (MDA134 and SUM44) compared with IDC cells (MCF7 and T47D) and the union of the nondifferentiated sites (log2FC > 1 or <−1, Q < 0.01). Table of the top motifs enriched in the FOXA1 binding sites gained in ILC versus IDC cells (top) and FOXA1 binding sites nondifferentiated between the ILC and IDC sites (bottom). C, Quantitative normalized signal of FOXA1 binding based on FOXA1 ChIP-seq in the ILC-gained chromatin-accessible sites based on the ATAC-seq analysis. D, Quantitative normalized signal of FOXA1 binding based on FOXA1 ChIP-seq in the chromatin-accessible sites lost in MDA134 after FOXA1 silencing based on the ATAC-seq analysis. E, Comparison of log2FC between ILC (MDA134 and SUM44) and IDC (MCF7 and T47D), FOXA1 binding sites (FOXA1 ChIP-seq; x-axis) versus the log2FC comparing chromatin accessibility (ATAC-seq; y-axis) between ILC (MDA134 and SUM44) and IDC (MCF7 and T47D). F, Intensity of binding in the intersecting sites of ILC-gained FOXA1 binding and ATAC-seq in MDA134 versus MCF7 cells. Spearman correlation and P value are reported.

Figure 2.

FOXA1 reprogramming in ILC is linked to the ILC-unique chromatin state. A, Sample-to-sample correlation (Euclidean distance between rows/columns and Ward's method) of FOXA1 binding sites correlation plots between all four cell lines in replicates [MCF7, T47D, MDA134 (MDAMB134), and SUM44] and the primary ILC cells isolated from a malignant peritoneal effusion from a patient with ER+ metastatic ILC (ILC Met). B, Tornado plots of FOXA1 binding sites (12,427 sites) gained in ILC (MDA134 and SUM44) compared with IDC cells (MCF7 and T47D) and the union of the nondifferentiated sites (log2FC > 1 or <−1, Q < 0.01). Table of the top motifs enriched in the FOXA1 binding sites gained in ILC versus IDC cells (top) and FOXA1 binding sites nondifferentiated between the ILC and IDC sites (bottom). C, Quantitative normalized signal of FOXA1 binding based on FOXA1 ChIP-seq in the ILC-gained chromatin-accessible sites based on the ATAC-seq analysis. D, Quantitative normalized signal of FOXA1 binding based on FOXA1 ChIP-seq in the chromatin-accessible sites lost in MDA134 after FOXA1 silencing based on the ATAC-seq analysis. E, Comparison of log2FC between ILC (MDA134 and SUM44) and IDC (MCF7 and T47D), FOXA1 binding sites (FOXA1 ChIP-seq; x-axis) versus the log2FC comparing chromatin accessibility (ATAC-seq; y-axis) between ILC (MDA134 and SUM44) and IDC (MCF7 and T47D). F, Intensity of binding in the intersecting sites of ILC-gained FOXA1 binding and ATAC-seq in MDA134 versus MCF7 cells. Spearman correlation and P value are reported.

Close modal

The ILC-unique FOXA1 cistrome reprograms the ER transcriptional network

Given the known role of FOXA1 in facilitating ER binding (10–13), we next sought to determine how the ER axis is affected by the reprogramming of the FOXA1 cistrome in ILC. To this end, we performed ER ChIP-seq in the ILC and IDC models. As shown previously (14), ligand-dependent ER ChIP-seq of all four cell models showed that majority of ER binding sites were in enhancer regions (Supplementary Fig. S3A). Unsupervised clustering of the ER binding sites segregated the ILC models, including the primary ILC Met, from the IDC models (Fig. 3A). This clustering was driven by 6,885 peaks that were shared between MDAMB134 and SUM44 cells and gained compared with the IDC models (log2FC >1 or <−1, q < 0.01; Fig. 3B). As expected, ILC ER-gained binding sites were enriched in the ERE motif, followed by FOXA1 motifs (Fig. 3A; ref. 21). Importantly, comparison of ER with FOXA1 ChIP-seq showed that the ER binding sites gained in ILC had increased FOXA1 binding in ILC versus IDC, and there was a significant overlap between the ER sites gained in ILC and the FOXA1 ILC-gained sites (44% of the gained ER binding sites overlapped with FOXA1-gained binding sites, P < 0.0001; Fig. 3C and D). There were 9,367 ILC-unique FOXA1 peaks that did not overlap with ER binding, which implies that FOXA1 has functions that are ER independent. Thus, the ILC-gained FOXA1 binding sites are tightly linked to the ER-gained binding but likely have functions that are independent of ER binding.

Figure 3.

The ER cistrome in ILC cell line models. A, ER ChIP-seq sample-to-sample correlation plot. IDC (MCF7 and T47D), ILC (MDA134 and SUM44), and primary ILC metastatic cells isolated from malignant peritoneal effusion from a patient with ER+ metastatic ILC (ILC MET) samples are clustered by the Euclidean distance between rows/columns and Ward's method. B, Tornado plots of ER binding sites gained in ILC cells compared with IDC cells (6,885 peaks) and the union of the nondifferentiated sites (59,286; log2FC > 1 or <−1, Q < 0.01). ER binding sites are shown in a horizontal window of ±2 kb from the peak center. Table of the top motifs enriched in the ER binding sites gained in ILC versus IDC cells (top) and ER binding sites nondifferentiated (bottom) comparing the ILC and IDC sites. C, Quantitative normalized signal of FOXA1 binding sites (FOXA1 ChIP-seq) on sites of ILC-gained ER binding based on ER ChIP-seq in ILC cells (SUM44 and MDA134) and IDC cells (MCF7 and T47D). D, Venn diagram of FOXA1 ILC-gained peaks (12,427; blue) and the ER ILC-gained peaks (6,885; orange) showing 3,060 peaks overlapping (sharing at least 1 bp) between the two factors. E, Ranking of the motifs enriched in H3K27ac binding sites gained in ILC cells (SUM44 and MDA134) versus IDC cells (MCF7 and T47D) based on P values of enrichment analysis. F, Overlap between FOXA1 ILC-gained peaks (12,427) compared with IDC (blue) and the H3K27ac ILC-gained peaks (4,569) compared with IDC (purple). G, RNA-seq sample-to-sample correlation based on Euclidean distance between rows/columns and Ward's method of IDC and ILC cells after HD and 12 hours of β-estradiol treatment. H, Comparison of differentially expressed genes between HD and 12 hours of β-estradiol (E2) treatment in ILC (SUM44 and MDA134) and IDC (MCF7 and T47D) cells. Genes with differential expression of Q < 0.01 are assigned to each category by the color scheme.

Figure 3.

The ER cistrome in ILC cell line models. A, ER ChIP-seq sample-to-sample correlation plot. IDC (MCF7 and T47D), ILC (MDA134 and SUM44), and primary ILC metastatic cells isolated from malignant peritoneal effusion from a patient with ER+ metastatic ILC (ILC MET) samples are clustered by the Euclidean distance between rows/columns and Ward's method. B, Tornado plots of ER binding sites gained in ILC cells compared with IDC cells (6,885 peaks) and the union of the nondifferentiated sites (59,286; log2FC > 1 or <−1, Q < 0.01). ER binding sites are shown in a horizontal window of ±2 kb from the peak center. Table of the top motifs enriched in the ER binding sites gained in ILC versus IDC cells (top) and ER binding sites nondifferentiated (bottom) comparing the ILC and IDC sites. C, Quantitative normalized signal of FOXA1 binding sites (FOXA1 ChIP-seq) on sites of ILC-gained ER binding based on ER ChIP-seq in ILC cells (SUM44 and MDA134) and IDC cells (MCF7 and T47D). D, Venn diagram of FOXA1 ILC-gained peaks (12,427; blue) and the ER ILC-gained peaks (6,885; orange) showing 3,060 peaks overlapping (sharing at least 1 bp) between the two factors. E, Ranking of the motifs enriched in H3K27ac binding sites gained in ILC cells (SUM44 and MDA134) versus IDC cells (MCF7 and T47D) based on P values of enrichment analysis. F, Overlap between FOXA1 ILC-gained peaks (12,427) compared with IDC (blue) and the H3K27ac ILC-gained peaks (4,569) compared with IDC (purple). G, RNA-seq sample-to-sample correlation based on Euclidean distance between rows/columns and Ward's method of IDC and ILC cells after HD and 12 hours of β-estradiol treatment. H, Comparison of differentially expressed genes between HD and 12 hours of β-estradiol (E2) treatment in ILC (SUM44 and MDA134) and IDC (MCF7 and T47D) cells. Genes with differential expression of Q < 0.01 are assigned to each category by the color scheme.

Close modal

To start to elucidate the transcriptional significance of the ILC-unique chromatin state and the reprogramming of the FOXA1–ER axis, we first performed H3K27ac ChIP-seq, a histone mark of active enhancers. Sample-to-sample correlation of H3K27ac ChIP-seq segregated ILC from IDC. This segregation was driven by 4,569 ILC-gained sites (Supplementary Fig. S3B and S3C). The number of sites gained in IDC was limited. The top motif in the ILC-gained H3K27ac sites was FOXA1 followed by ERE (Fig. 3E; Supplementary Fig. S3C). Close to 50% of the ILC H3K27ac-gained sites overlapped with the FOXA1 ILC-gained sites (Fig. 3F). Consistent with these findings, RNA-seq analysis of the four cell lines in estradiol-stimulated conditions segregated the ILC from the IDC models (Fig. 3G), and most of the estradiol regulated genes were different in ILC compared with IDC (Fig. 3H). Taken together, these results support a previous study that showed disparate ligand-stimulated transcription in ILC versus IDC (14) and provide initial evidence for the role of FOXA1 in the ILC-unique transcription.

The ILC–FOXA1 gene set is associated with increased risk of recurrence in ILC tumors of luminal A molecular subtype

To test if the ILC-unique FOXA1 chromatin recruitment is driving the ligand-stimulated transcriptional differences between ILC and IDC, we integrated the RNA-seq and ChIP-seq data by applying Binding and Expression Target Analysis (BETA; ref. 22). BETA basic showed that the FOXA1 binding sites gained in ILC are significantly associated with the upregulation of genes with increased expression in ILC versus IDC in estradiol-stimulated conditions (Supplementary Fig. S4A). This association is highlighted at the level of single genes in the volcano plot in Fig. 4A, which shows the overlap between genes predicted to be gained by FOXA1 binding based on BETA minus and the genes upregulated in ILC versus IDC based on the RNA-seq differential expression analysis (Fig. 4A; Supplementary Fig. S4B). To further substantiate the link between the FOXA1-gained sites and FOXA1-dependent transcription in ILC, we tested the association between the ILC-gained binding sites and the genes regulated by the silencing of FOXA1 using the transcriptomic analysis of MDAMB134 cells with and without DOX-induced silencing of FOXA1. BETA basic analysis showed that the ILC FOXA1-gained sites were significantly associated with the genes that were downregulated after FOXA1 silencing (Supplementary Fig. S4C and S4D).

Figure 4.

FOXA1 drives the ILC-unique transcriptome. A, Volcano plot depicting differential (DESeq2) expression comparing RNA-seq of ILC cells (MDA134 and SUM44) versus IDC cells (MCF7 and T47D) after β-estradiol (E2) stimulation. Shown in yellow are the genes with significant differential expression ([log2FC] > 1, Q < 0.01. DESeq2). Red dots, genes with significant differential expression ([log2FC] > 1, Q < 0.01; DESeq2) and are regulated by FOXA1 ILC-gained binding sites based on BETA minus analysis. B, Hallmark pathways enriched in the genes regulated by the ILC-gained FOXA1 sites based on BETA basic. The normalized enrichment score (NES) is represented on the x-axis; the number of genes in the data set is the count number represented by the circle size, q < 0.25. C, Enrichment plot from GSEA showing enrichment of the FOXA1_ILC_120 gene set derived from the ILC-gained binding sites in ILC versus IDC ER+ breast cancers in the METABRIC cohort. D, Distant-free survival in patients with luminal A molecular subtype ILC from the METABRIC cohort comparing patients with high versus low expression of the FOXA1_ILC_120 gene set. E, Distant-free survival in patients with luminal B molecular subtype type ILC from the METABRIC cohort comparing patients with high versus low expression of the FOXA1_ILC_120 gene set. F, Distant-free survival in patients with luminal A molecular subtype IDC from the METABRIC cohort comparing patients with high versus low expression of the FOXA1_ILC_120 gene set. G, Distant-free survival in patients with luminal B molecular subtype IDC from the METABRIC cohort comparing patients with high versus low expression of the FOXA1_ILC_120 gene set.

Figure 4.

FOXA1 drives the ILC-unique transcriptome. A, Volcano plot depicting differential (DESeq2) expression comparing RNA-seq of ILC cells (MDA134 and SUM44) versus IDC cells (MCF7 and T47D) after β-estradiol (E2) stimulation. Shown in yellow are the genes with significant differential expression ([log2FC] > 1, Q < 0.01. DESeq2). Red dots, genes with significant differential expression ([log2FC] > 1, Q < 0.01; DESeq2) and are regulated by FOXA1 ILC-gained binding sites based on BETA minus analysis. B, Hallmark pathways enriched in the genes regulated by the ILC-gained FOXA1 sites based on BETA basic. The normalized enrichment score (NES) is represented on the x-axis; the number of genes in the data set is the count number represented by the circle size, q < 0.25. C, Enrichment plot from GSEA showing enrichment of the FOXA1_ILC_120 gene set derived from the ILC-gained binding sites in ILC versus IDC ER+ breast cancers in the METABRIC cohort. D, Distant-free survival in patients with luminal A molecular subtype ILC from the METABRIC cohort comparing patients with high versus low expression of the FOXA1_ILC_120 gene set. E, Distant-free survival in patients with luminal B molecular subtype type ILC from the METABRIC cohort comparing patients with high versus low expression of the FOXA1_ILC_120 gene set. F, Distant-free survival in patients with luminal A molecular subtype IDC from the METABRIC cohort comparing patients with high versus low expression of the FOXA1_ILC_120 gene set. G, Distant-free survival in patients with luminal B molecular subtype IDC from the METABRIC cohort comparing patients with high versus low expression of the FOXA1_ILC_120 gene set.

Close modal

We next used BETA basic to identify the genes that are direct targets of FOXA1 unique to ILC in estradiol-stimulated conditions and generated a ranked product gene set termed the FOXA1–ILC gene set (Supplementary Table S1A). Ranked Gene Set Enrichment Analysis (GSEA) revealed that the FOXA1–ILC gene set is enriched in genes involved in key cancer pathways (ER response pathways signaling; refs. 23, 24), TGFβ signaling (25, 26), hypoxia (25), and NFKB signaling (Q < 0.25; Fig. 4B; refs. 27, 28). We next generated a more stringent gene set of 120 FOXA1–ILC genes by selecting the genes with a ranked product P <0.001 (FOXA1_ILC 120). To study the clinical significance of this gene set, we turned to the METABRIC cohort. We first determined that the FOXA1_ILC 120 is enriched in primary ER+ ILC tumors compared with primary ER+ IDC tumors (Fig. 4C). Importantly, high versus low expression of the FOXA1_ILC 120 gene set was associated with a significantly lower distant relapse-free survival (HR = 3, P < 0.039) in patients with ILC and not IDC luminal A tumors. This association was not seen in patients with luminal B breast cancers. Thus, the FOXA1_ILC 120 is a potential signature of high-risk ILC tumors that are driven by ER but not of tumors that are less dependent on ER signaling, which are enriched in genetic alterations or other pathways associated with treatment resistance, such as P53 mutations (Fig. 4DG; ref. 29). Hence, this signature could potentially identify the ILC patients who might benefit from improved endocrine treatment strategies.

We next looked at the overlapping FOXA1/ER/H3K27ac sites gained in ILC and integrated these gained sites with the RNA-seq differential expression comparing ILC and IDC in estradiol-stimulated conditions by applying BETA basic to identify single genes that are direct targets of FOXA1 and ER and are upregulated in ILC. CASZ1 was the top-ranked gene, and RET, and SNAI1 were among the top-ranked genes that met these criteria (Supplementary Table S1B; Supplementary Fig. S4E). The protein levels of these three genes were increased in MDAMB134 versus MCF7 cells and downregulated by the silencing of ER or FOXA1 (Supplementary Fig. S4F). Previous studies have shown that RET and SNAIL are ER targets with roles in resistance to endocrine treatment (30–32). In contrast, CASZ1, a zinc finger TF that regulates transcription by binding to the nucleosome remodeling and histone deacetylase complex (33), has not been implicated as an important gene in breast cancer. We first confirmed that CASZ1 expression is significantly higher in ER+ ILC versus ER+ IDC in the METABRIC cohort (Supplementary Fig. S4G). In the MDAMB134 model, silencing of CASZ1 resulted in cell growth inhibition (Supplementary Fig. S4H) and downregulation of genes with key roles in tumor growth, including CDK2 and genes related to receptor tyrosine kinase signaling such as FGFRL1, FGFR4, MAPK9, and PIK3R2 (Supplementary Fig. S4I and S4J). Collectively, these results show that the ILC-unique FOXA1 cistrome mediates the unique response to estradiol and the transcription of genes involved in tumor progression and poor outcomes.

Downregulation of E-cadherin in IDC or upregulation of E-cadherin in ILC is not sufficient to induce a unique ER transcriptional program

Although the loss of E-cadherin in ILC leads to the degradation of beta-catenin, p120 translocates to the cytoplasm or nucleus and in these compartments, it has a role in signaling and transcription regulation (34). We therefore asked if the loss of E-cadherin and the release of membrane-bound p120 contribute to the ILC-unique ER cistrome and transcriptional response to estradiol. To test this, we silenced CDH1 in MCF7 and T47D cells with a CRISPR cas9-gRNA (Supplementary Fig. S5A and S5B). The decrease in E-cadherin expression led to increased migratory capacity but did not affect cell growth (Supplementary Fig. S5C and S5D). In addition, decreased E-cadherin expression resulted in limited transcriptional changes in MCF7 and T47D cells and the downregulated genes were involved in pathways of cell adhesion molecules among other pathways (Supplementary Fig. S5E–S5H). In addition, silencing of E-cadherin did not lead to significant changes in ER binding or H3K27 acetylation in estrogen-stimulated conditions (Supplementary Fig. S5I and S5J). Consistent with these results, there were limited transcriptional changes after estrogen stimulation when comparing the cells with and without E-cadherin downregulation (Supplementary Fig. S5K and S5L). Specifically, FOXA1 and ER expression was not altered after E-cadherin silencing. Additionally, stable overexpression of E-cadherin in MDAMB134 cells with a DOX-inducible construct resulted in a limited number of transcriptional changes at 3 days and 3 weeks (Supplementary Fig. S5M–S5O). Taken together, silencing of E-cadherin by itself was not sufficient in these models for the rewiring of the FOXA1–ER axis and the altered response to estrogen seen in ILC compared with IDC.

A circuitry of FOXA1 upregulation mediated by a unique superenhancer

Because we showed that aberrant FOXA1 activity is central to the epigenetic landscape in ILC, we sought to understand the mechanism of the altered FOXA1 recruitment in ILC. Mutations in TFs can contribute to altered binding and tumor progression (9, 16); however, whole-genome sequencing of the ILC models did not identify FOXA1 mutations (Supplementary Fig. S6A). The overexpression of TFs can also alter the regulatory activity of cancer cells (35) and, in fact, we observed mRNA and protein overexpression of FOXA1 in ILC compared with IDC (Fig. 5A; Supplementary Fig. S4F). Because we did not identify FOXA1 copy-number variations or mutations in the promoter region to explain the upregulation of FOXA1 (Supplementary Fig. S6A; refs. 24, 36), we interrogated the TFs that regulate FOXA1 expression (cistromeDB toolkit; ref. 37). Similar to other key lineage-defining TFs that have been found to be autoinduced (38), the TF with the highest FOXA1 regulatory potential was FOXA1 itself. ER and GATA3, the other two key determinants of ER+ breast cancer, are also potent FOXA1 regulators (Fig. 5B). Interestingly, examination of FOXA1 binding upstream to the FOXA1 transcription start site (TSS) revealed a FOXA1 peak in the enhancer region (>10 kb from the TSS) that was unique to the ILC models (P1; Fig. 5C). H3K27 acetylation in this region was also different in the ILC models compared with IDC and extended at least 3 kb upstream in ILC overlapping with the P1 peak (Fig. 5D). Based on the H3K27ac ChIP-seq analysis this region upstream to the FOXA1 promoter is a superenhancer region in all four models (Supplementary Table S1C). However, the boundaries of the superenhancer region in the ILC models extended beyond those of the IDC superenhancer and only in the ILC models, the superenhancer region overlapped with P1. We also detected the ILC-unique FOXA1 P1 peak and superenhancer region in the primary ILC Met, demonstrating clinical relevance of the P1 peak and unique superenhancer region (Fig. 5C and D). Because superenhancers are characterized by the binding of multiple tissue-specific TFs (39), we looked at the binding of ER and GATA3, the other top-ranked FOXA1 regulators, and saw increased ER and GATA3 binding in the ILC-unique superenhancer region (Fig. 5E). Furthermore, H3K27ac Hi-ChIP detected ILC-unique loops in this region (Supplementary Fig. S6B). Next, we tested the regulatory function of P1 by selective targeting of P1 with dCAS9-KRAB and a gRNA targeting the summit of this peak (details of the coordinates are in Materials and Methods). In the SUM44 ILC cells, inactivation of P1 decreased the expression of FOXA1 at the mRNA and protein level, led to the regression of cell growth and decreased the expression of canonical ER target genes including PgR, MYC, and CCND1 (Fig. 5FI; Supplementary Fig. S6C). Among these genes, MYC and CCDN1 have key roles in cell proliferation. The expression of ER was also downregulated. This is consistent with the downregulation of ER expression when FOXA1 was silenced with siFOXA1 in this study (Supplementary Fig. S4F) and a previous study that demonstrated the downregulation of ER in response to FOXA1 silencing (40). In contrast, inactivation of P1 in MCF7 cells did not affect FOXA1 expression or cell growth despite the sensitivity of MCF7 to a siFOXA1 (Fig. 5J and K; Supplementary Fig. S6D). The activity of dCAS9-KRAB in MCF7 cells was confirmed by infecting the MCF7 cells engineered to stably express dCAS9-KRAB with a BFP-GFP reporter construct that includes a gRNA that targets the GFP cassette. In these cells, expression of dCAS9-KRAB along with the gGFP resulted in the loss of expression of BFP and GFP (Supplementary Fig. S6E). Collectively, our results support FOXA1 overexpression in the ILC models at least partly through autoinduction by a unique FOXA1 binding site within a superenhancer.

Figure 5.

FOXA1 autoinduction in ILC through a unique superenhancer and FOXA1 binding site. A, RNA expression (FPKM values) of FOXA1 in IDC (MCF7 and T47D) and ILC (MDA134 and SUM44) cells. t test, ****, P < 0.00001. B, Regulatory potential of the transcription factors that regulate FOXA1 based on the CistromeDB toolkit (dbtoolkit.cistrome.org). C and D, ChIP-seq tracks showing FOXA1 (C) and H3K27acetylation (D) in the IDC (MCF7 and T47D (red and orange tracks) and ILC (MDA134 and SUM44; blue and purple tracks) cell lines and primary cells from ILC metastasic peritoneal effusion (ILC met, green tracks). E, ER and GATA3 ChIP-seq tracks for the IDC (MCF7 and T47D; red and orange tracks) and ILC (MDA134 and SUM44; blue and purple tracks) cell lines. F, Cartoon of CRISPRi action at the P1 site. Enlargement showing the gRNA used to target the FOXA1 binding region. G, mRNA levels of FOXA1 in SUM44 lobular cells by RT-qPCR in presence of the gCTR and gP1 at day 7 after transduction. The mRNA expression was normalized to the GAPDH housekeeping gene, and expression levels are presented as 2–ΔΔCT compared with control. H, Cell proliferation assay after 14 days in SUM44 cells of control cells and after suppression of the FOXA1 P1 (gP1) using CRISPRi. Error bars, ± SEM, n = 3. I, Western blot of whole-cell lysates for FOXA1 in SUM44 cells in the presence of the control guide (gCTR) and the guide RNA targeting P1 (gP1) at day 9 after transduction. J, FOXA1 mRNA levels of MCF7 in the presence of the gCTR and the gP1. K, Cell proliferation assay in MCF7 cells in the presence of the gCTR and the gP1. *, P < 0.05; ****, P < 0.0001; ns, not significant.

Figure 5.

FOXA1 autoinduction in ILC through a unique superenhancer and FOXA1 binding site. A, RNA expression (FPKM values) of FOXA1 in IDC (MCF7 and T47D) and ILC (MDA134 and SUM44) cells. t test, ****, P < 0.00001. B, Regulatory potential of the transcription factors that regulate FOXA1 based on the CistromeDB toolkit (dbtoolkit.cistrome.org). C and D, ChIP-seq tracks showing FOXA1 (C) and H3K27acetylation (D) in the IDC (MCF7 and T47D (red and orange tracks) and ILC (MDA134 and SUM44; blue and purple tracks) cell lines and primary cells from ILC metastasic peritoneal effusion (ILC met, green tracks). E, ER and GATA3 ChIP-seq tracks for the IDC (MCF7 and T47D; red and orange tracks) and ILC (MDA134 and SUM44; blue and purple tracks) cell lines. F, Cartoon of CRISPRi action at the P1 site. Enlargement showing the gRNA used to target the FOXA1 binding region. G, mRNA levels of FOXA1 in SUM44 lobular cells by RT-qPCR in presence of the gCTR and gP1 at day 7 after transduction. The mRNA expression was normalized to the GAPDH housekeeping gene, and expression levels are presented as 2–ΔΔCT compared with control. H, Cell proliferation assay after 14 days in SUM44 cells of control cells and after suppression of the FOXA1 P1 (gP1) using CRISPRi. Error bars, ± SEM, n = 3. I, Western blot of whole-cell lysates for FOXA1 in SUM44 cells in the presence of the control guide (gCTR) and the guide RNA targeting P1 (gP1) at day 9 after transduction. J, FOXA1 mRNA levels of MCF7 in the presence of the gCTR and the gP1. K, Cell proliferation assay in MCF7 cells in the presence of the gCTR and the gP1. *, P < 0.05; ****, P < 0.0001; ns, not significant.

Close modal

The reprogrammed FOXA1 cistrome drives tamoxifen resistance in ILC

Similar to the clinical observation of relative resistance to tamoxifen in ILC (6), MDAMB134 and SUM44 cells were resistant to tamoxifen (IC50 of close to 1 μmol/L for both ILC models compared with an IC50 of 1 nmol/L and 2.2 nmol/L for T47D and MCF7, respectively; Fig. 6A). Because the ILC models have a longer doubling time compared with the IDC models, we confirmed the resistance to tamoxifen by growth rate metrics, which are independent of the division rate of the assayed cells (41). The GR50 for tamoxifen was 8- to 30-fold higher in the ILC models compared with IDC (Fig. 6B). Tamoxifen resistance was not due to ligand- or ER-independent growth, as evidenced by the high sensitivity of MDAMB134 cells to estrogen deprivation and ER silencing, as in MCF7 cells (Fig. 6C and D; Supplementary Fig. S7A and S7B).

Figure 6.

The mechanism of tamoxifen resistance in ILC. A and B, Dose-response curves of 4-hydroxytamoxifen (tamoxifen) treatment in ductal and lobular cell lines. A, Curves normalized to vehicle control. IC50 values are in the range of 1 μmol/L for both ILC models (MDA134 and SUM44), compared with an IC50 of 1 nmol/L and 2.2 nmol/L for T47D and MCF7, respectively. B, GR50s values are 240 nmol/L for MCF7, 60 nmol/L for T47D, 1,920 nmol/L for MDA134, and 1,680 μmol/L for SUM44. C, Cell proliferation curves of MDA134 cells followed for seven days without (ED) or with estradiol (E2; 10 nmol/L). Error bars, ± SEM, n = 3. **, P <0.01. D, Cell proliferation studies of MDA134 cells in full medium conditions including cells transfected with an siControl (siCTR) and cells with silencing of ER (siER_1 and siER_2). Error bars, ± SEM, n = 3. **, P < 0.01. E, Tornado plots of ER binding sites lost in MCF7 treated with 4-hydroxytamoxifen (TAM) compared with β-estradiol (E2) treatment. F, Tornado plots of ER binding sites lost in MCF7 treated with TAM compared with E2 treatment and unchanged in the MDA134 cells in E2- and TAM-treated conditions (log2FC > 1 or <−1, q < 0.01). G, Quantitative normalized signal of FOXA1 ChIP-seq binding at ER binding sites lost in MCF7 cells with TAM treatment but unchanged in MDA134 cells. H, Quantitative normalized signal of ER ChIP-seq binding on ER sites lost in MCF7 cells but retained in MDA134 cells, in MDA134 cells treated with TAM (with and without DOX induction of FOXA1 silencing (shFOXA1). I, Comparison of differentially expressed genes between β-estradiol and 4-hydroxytamoxifen in MCF7 and MDA134. Genes with FDR < 0.05 are assigned to each category by the color scheme. There were 711 shared differentially expressed genes, 1,592 genes differentially expressed in MCF7 cells only, and 649 genes differentially expressed exclusively in MDA134. J, BETA basic plot of the activating and repressive function of the ER binding sites lost in MCF7 but conserved in MDA134 after tamoxifen treatment. Red line, genes upregulated; purple line, genes downregulated with 4-hydroxytamoxifen versus β-estradiol treatment in MCF7 cells. Black dashed line, nondifferentially expressed genes as background. The P value is based on the Kolmogorov–Smirnov test. K, Hallmark pathways enriched in the genes determined to be downregulated by 4-hydroxytamoxifen compared with E2 treatment and regulated by the ER binding sites lost in MCF7 cells after tamoxifen treatment but unchanged in MDA134. The normalized enrichment score (NES) is represented on the x-axis, and the number of genes in the data set is the count number represented by the circle size, q < 0.25.

Figure 6.

The mechanism of tamoxifen resistance in ILC. A and B, Dose-response curves of 4-hydroxytamoxifen (tamoxifen) treatment in ductal and lobular cell lines. A, Curves normalized to vehicle control. IC50 values are in the range of 1 μmol/L for both ILC models (MDA134 and SUM44), compared with an IC50 of 1 nmol/L and 2.2 nmol/L for T47D and MCF7, respectively. B, GR50s values are 240 nmol/L for MCF7, 60 nmol/L for T47D, 1,920 nmol/L for MDA134, and 1,680 μmol/L for SUM44. C, Cell proliferation curves of MDA134 cells followed for seven days without (ED) or with estradiol (E2; 10 nmol/L). Error bars, ± SEM, n = 3. **, P <0.01. D, Cell proliferation studies of MDA134 cells in full medium conditions including cells transfected with an siControl (siCTR) and cells with silencing of ER (siER_1 and siER_2). Error bars, ± SEM, n = 3. **, P < 0.01. E, Tornado plots of ER binding sites lost in MCF7 treated with 4-hydroxytamoxifen (TAM) compared with β-estradiol (E2) treatment. F, Tornado plots of ER binding sites lost in MCF7 treated with TAM compared with E2 treatment and unchanged in the MDA134 cells in E2- and TAM-treated conditions (log2FC > 1 or <−1, q < 0.01). G, Quantitative normalized signal of FOXA1 ChIP-seq binding at ER binding sites lost in MCF7 cells with TAM treatment but unchanged in MDA134 cells. H, Quantitative normalized signal of ER ChIP-seq binding on ER sites lost in MCF7 cells but retained in MDA134 cells, in MDA134 cells treated with TAM (with and without DOX induction of FOXA1 silencing (shFOXA1). I, Comparison of differentially expressed genes between β-estradiol and 4-hydroxytamoxifen in MCF7 and MDA134. Genes with FDR < 0.05 are assigned to each category by the color scheme. There were 711 shared differentially expressed genes, 1,592 genes differentially expressed in MCF7 cells only, and 649 genes differentially expressed exclusively in MDA134. J, BETA basic plot of the activating and repressive function of the ER binding sites lost in MCF7 but conserved in MDA134 after tamoxifen treatment. Red line, genes upregulated; purple line, genes downregulated with 4-hydroxytamoxifen versus β-estradiol treatment in MCF7 cells. Black dashed line, nondifferentially expressed genes as background. The P value is based on the Kolmogorov–Smirnov test. K, Hallmark pathways enriched in the genes determined to be downregulated by 4-hydroxytamoxifen compared with E2 treatment and regulated by the ER binding sites lost in MCF7 cells after tamoxifen treatment but unchanged in MDA134. The normalized enrichment score (NES) is represented on the x-axis, and the number of genes in the data set is the count number represented by the circle size, q < 0.25.

Close modal

To start to understand the mechanism of relative resistance to tamoxifen in ILC, we performed ER ChIP-seq after tamoxifen or estradiol treatment in MCF7 and MDAMB134 cells. In MCF7 cells, tamoxifen compared with estradiol led to the loss of 16,022 ER binding sites, whereas in MDAMB134 cells, there were 37 lost and 236 gained ER binding sites (log2FC > 1 or <−1 and q < 0.01; Fig. 6E; Supplementary Fig. S7C). When comparing the two cell lines and two treatment conditions, we identified 5,561 peaks that were lost in MCF7 cells but retained in MDAMB134 cells after tamoxifen treatment (log2FC > 1 or <−1 and q < 0.01; Fig. 6F). These ER peaks had increased FOXA1 binding and chromatin accessibility in ILC cells compared to IDC cells (Fig. 6G and Supplementary Fig. S7D). Furthermore, downregulation of FOXA1 expression with DOX-induced shFOXA1 in MDAMB134 cells (Supplementary Fig. S1F) resulted in growth inhibition and a loss of ER binding in the sites maintained after tamoxifen treatment (Supplementary Fig. S7E; Fig. 6H). Of note, we showed that FOXA1 silencing leads to decreased ER expression (Supplementary Fig. S4F); hence, these functional consequences of FOXA1 silencing can also be attributed to its role in facilitating ER binding and regulating the expression of ER.

To study the transcriptional consequences of the disparate effects of tamoxifen treatment on ER binding in ILC versus IDC, we performed RNA-seq with and without tamoxifen treatment in MCF7 and MDAMB134 cells. In line with the differences in the changes in ER binding, there were differences in the transcriptional changes in response to tamoxifen treatment. A comparison of the transcriptional changes induced by tamoxifen in MCF7 and MDAMB134 revealed 1,592 genes differentially expressed in MCF7 cells only, 649 genes differentially expressed in MDAMB134 only, and 711 genes differentially expressed in both cell models (Fig. 6I). BETA basic analysis showed that the ER binding sites lost in MCF7 cells but retained in MDAMB134 after tamoxifen treatment had a significant association with the genes downregulated with tamoxifen treatment in MCF7 cells (Fig. 6J). GSEA of the ranked genes regulated by the MDAMB134 tamoxifen-maintained ER binding sites in MCF7 cells showed that these genes are involved in pathways that are key therapeutic targets of tamoxifen, such as response to estrogen, MYC targets, and epithelial–mesenchymal transition (Fig. 6K, Supplementary Table S1D; ref. 42). In aggregate, in MCF7 cells tamoxifen treatment results in the loss of a set of ER binding sites, and this is associated with the downregulation of genes important for the therapeutic response to tamoxifen. In contrast, in MDAMB134 cells, ER binding at these sites is retained in a FOXA1-dependent manner.

In this study, we found that ILC has a unique chromatin state associated with the altered FOXA1 recruitment. The ILC-unique epigenetic state provides one mechanism to explain the disparate transcriptional responses to estrogen and resistance to tamoxifen detected in the ILC cell models despite the high ER expression and ER dependency. These results provide new fundamental insights to the unique biology of ILC and may have important therapeutic implications.

Several mechanisms of aberrant FOXA1 signaling have been described previously, including FOXA1 mutations and amplifications (7–9, 24). Though the increased frequency of FOXA1 mutations in ILC suggests that altered FOXA1 signaling is important in the biology of ILC, the majority of ILCs do not harbor a FOXA1 mutation. This implies that other mechanisms of aberrant FOXA1 signaling may be enriched in ILC. Indeed, we identified another mechanism of altered FOXA1 signaling in cell lines and metastatic ILC cells isolated from a patient with a metastatic peritoneal effusion. Herein, we provided evidence for a feed-forward circuit of FOXA1 autoinduction involving a unique superenhancer and FOXA1 binding site with the corecruitment of other TFs including ER and GATA3. The corecruitment of these TFs may suggest that these TFs regulate the transcriptional activity of each other. This is supported by previous studies that showed an interplay between the FOXA1, ER, and GR TFs in which both ER and GR can alter the distribution of FOXA1 binding (43, 44). Our study demonstrates how the overexpression of a single key TF can lead to sustained effects on an entire transcriptional program and can facilitate tumor progression and treatment resistance.

Previous studies identified several mechanisms of tamoxifen resistance unique to ILC models (45, 46). In this study, we showed another mechanism of tamoxifen resistance in ILC. Tamoxifen is a selective estrogen receptor modulator that inhibits the transcriptional activity of ER by shifting the balance between coactivator and corepressor binding (47). Resistance specific to tamoxifen has been attributed mainly to a shift toward increased coactivator versus corepressor binding, which can lead to ER agonistic activity and explain tumor regression after tamoxifen withdrawal (48, 49). Here we show that in MCF7 cells, tamoxifen treatment leads to loss of ER binding in selective sites and reduces ER-mediated transcription. Our results suggest that gained FOXA1 recruitment in ILC can counteract this activity of tamoxifen and preserve ER binding. In addition, FOXA1 silencing was sufficient to inhibit this ER binding either through the downregulation of ER expression or by the perturbation of ER binding. Together, these results highlight the importance of developing approaches for targeting traditionally undruggable targets such as FOXA1.

We provide proof of principle for tumor growth inhibition by selective targeting of a unique FOXA1 binding site associated with a superenhancer that regulates the expression of FOXA1 itself. The advantage of this type of therapeutic approach over current epigenetic strategies is 2-fold: (i) Substantial efficacy because of the disruption of an autoregulatory loop that is self-sustained and regulates multiple genes involved in tumor progression. (ii) High precision given the disruption of a selective superenhancer in selective cells as opposed to current therapeutic approaches that target epigenetic regulators and modulate the transcription of multiple genes in a nonselective manner (50). Importantly, selective inhibition of a regulatory region with a small molecule has been shown to be feasible (51). However, of more immediate clinical relevance, we show that ILC is highly dependent on a unique ER transcriptional axis and cell growth remains ligand dependent. These results support preclinical studies of ILC models and clinical trials dedicated to patients with ILC to investigate the oral selective ER degraders and other novel endocrine treatments currently in clinical development.

M.L. Freedman reports personal fees and other support from Precede outside the submitted work. R. Schiff reports grants from Breast Cancer Research Foundation during the conduct of the study; grants from Gilead Sciences, Puma, Biotechnology Inc, personal fees from Macrogenics, and Wolters Kluwer/UpToDate outside the submitted work; and reports a pending patent (via institution), which is truly unrelated to the currently submitted work, but is still disclosed below: NRF Ref. BAYM.P0312US.P1-1001123973: “A multiparameter classifier to predict response to HER2-targeted therapy without chemotherapy in HER2-positive breast cancer.” No revenue was received. R. Jeselsohn reports other support from Pfizer and grants from Lilly during the conduct of the study; personal fees from Luminex and other support from Pfizer outside the submitted work. No disclosures were reported by the other authors.

A. Nardone: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. X. Qiu: Data curation, formal analysis, supervision, writing–review and editing. S. Spisak: Data curation, investigation, writing–review and editing. Z. Nagy: Validation, investigation, writing–review and editing. A. Feiglin: Data curation, formal analysis, methodology, writing–review and editing. A. Feit: Data curation, formal analysis, methodology, writing–review and editing. G. Cohen Feit: Investigation, writing–review and editing. Y. Xie: Formal analysis, writing–review and editing. A. Font-Tello: Investigation, writing–review and editing. C. Guarducci: Investigation, writing–review and editing. F. Hermida-Prado: Investigation, writing–review and editing. S. Syamala: Investigation, writing–review and editing. K. Lim: Investigation, writing–review and editing. M. Munoz Gomez: Investigation, writing–review and editing. M. Pun: Investigation, writing–review and editing. M. Cornwell: Formal analysis, investigation, writing–review and editing. W. Liu: Investigation, writing–review and editing. A. Ors: Investigation, writing–review and editing. H. Mohammed: Investigation, writing–review and editing. P. Cejas: Formal analysis, writing–review and editing. J.B. Brock: Formal analysis, writing–review and editing. M.L. Freedman: Supervision, writing–review and editing. E.P. Winer: Conceptualization, funding acquisition, writing–review and editing. X. Fu: Formal analysis, writing–review and editing. R. Schiff: Conceptualization, writing–review and editing. H.W. Long: Conceptualization, supervision, methodology, writing–review and editing. O. Metzger Filho: Conceptualization, supervision, writing–review and editing. R. Jeselsohn: Conceptualization, supervision, funding acquisition, validation, methodology, writing–original draft, writing–review and editing.

This work was conducted with support from the Maor Foundation (to E. Winer, O. Metzger Filho, and R. Jeselsohn) and NIH (5RO1CA237414-02 and 1K08CA191058-01A1 to R. Jeselsohn; 5RO1CA193910-03 to M.L. Freedman). The authors thank Ms. Cheri Fox for her support, important insights, and helpful discussion.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.

Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).

1.
Berx
G
,
Cleton-Jansen
AM
,
Nollet
F
,
de Leeuw
WJ
,
van de Vijver
M
,
Cornelisse
C
, et al
.
E-cadherin is a tumour/invasion suppressor gene mutated in human lobular breast cancers
.
EMBO J
1995
;
14
:
6107
15
.
2.
Derksen
PW
,
Liu
X
,
Saridin
F
,
van der Gulden
H
,
Zevenhoven
J
,
Evers
B
, et al
.
Somatic inactivation of E-cadherin and p53 in mice leads to metastatic lobular mammary carcinoma through induction of anoikis resistance and angiogenesis
.
Cancer Cell
2006
;
10
:
437
49
.
3.
McCart Reed
AE
,
Kutasovic
JR
,
Lakhani
SR
,
Simpson
PT
.
Invasive lobular carcinoma of the breast: morphology, biomarkers and 'omics
.
Breast Cancer Res
2015
;
17
:
12
.
4.
Arpino
G
,
Bardou
VJ
,
Clark
GM
,
Elledge
RM
.
Infiltrating lobular carcinoma of the breast: tumor characteristics and clinical outcome
.
Breast Cancer Res
2004
;
6
:
R149
56
.
5.
Pestalozzi
BC
,
Zahrieh
D
,
Mallon
E
,
Gusterson
BA
,
Price
KN
,
Gelber
RD
, et al
.
Distinct clinical and prognostic features of infiltrating lobular carcinoma of the breast: combined results of 15 International Breast Cancer Study Group clinical trials
.
J Clin Oncol
2008
;
26
:
3006
14
.
6.
Metzger Filho
O
,
Giobbie-Hurder
A
,
Mallon
E
,
Gusterson
B
,
Viale
G
,
Winer
EP
, et al
.
Relative effectiveness of letrozole compared with tamoxifen for patients with lobular carcinoma in the BIG 1-98 trial
.
J Clin Oncol
2015
;
33
:
2772
9
.
7.
Ciriello
G
,
Gatza
ML
,
Beck
AH
,
Wilkerson
MD
,
Rhie
SK
,
Pastore
A
, et al
.
Comprehensive molecular portraits of invasive lobular breast cancer
.
Cell
2015
;
163
:
506
19
.
8.
Desmedt
C
,
Zoppoli
G
,
Gundem
G
,
Pruneri
G
,
Larsimont
D
,
Fornili
M
, et al
.
Genomic characterization of primary invasive lobular breast cancer
.
J Clin Oncol
2016
;
34
:
1872
81
.
9.
Arruabarrena-Aristorena
A
,
Maag
JLV
,
Kittane
S
,
Cai
Y
,
Karthaus
WR
,
Ladewig
E
, et al
.
FOXA1 mutations reveal distinct chromatin profiles and influence therapeutic response in breast cancer
.
Cancer Cell
2020
;
38
:
534
50
.
10.
Eeckhoute
J
,
Keeton
EK
,
Lupien
M
,
Krum
SA
,
Carroll
JS
,
Brown
M
.
Positive cross-regulatory loop ties GATA-3 to estrogen receptor alpha expression in breast cancer
.
Cancer Res
2007
;
67
:
6477
83
.
11.
Carroll
JS
,
Liu
XS
,
Brodsky
AS
,
Li
W
,
Meyer
CA
,
Szary
AJ
, et al
.
Chromosome-wide mapping of estrogen receptor binding reveals long-range regulation requiring the forkhead protein FoxA1
.
Cell
2005
;
122
:
33
43
.
12.
Eeckhoute
J
,
Lupien
M
,
Meyer
CA
,
Verzi
MP
,
Shivdasani
RA
,
Liu
XS
, et al
.
Cell-type selective chromatin remodeling defines the active subset of FOXA1-bound enhancers
.
Genome Res
2009
;
19
:
372
80
.
13.
Theodorou
V
,
Stark
R
,
Menon
S
,
Carroll
JS
.
GATA3 acts upstream of FOXA1 in mediating ESR1 binding by shaping enhancer accessibility
.
Genome Res
2013
;
23
:
12
22
.
14.
Sikora
MJ
,
Cooper
KL
,
Bahreini
A
,
Luthra
S
,
Wang
G
,
Chandran
UR
, et al
.
Invasive lobular carcinoma cell lines are characterized by unique estrogen-mediated gene expression patterns and altered tamoxifen response
.
Cancer Res
2014
;
74
:
1463
74
.
15.
Watt
AC
,
Cejas
P
,
DeCristo
MJ
,
Metzger-Filho
O
,
Lam
EYN
,
Qiu
X
, et al
.
CDK4/6 inhibition reprograms the breast cancer enhancer landscape by stimulating AP-1 transcriptional activity
.
Nat Cancer
2021
;
2
:
34
48
.
16.
Jeselsohn
R
,
Bergholz
JS
,
Pun
M
,
Cornwell
M
,
Liu
W
,
Nardone
A
, et al
.
Allele-specific chromatin recruitment and therapeutic vulnerabilities of ESR1 activating mutations
.
Cancer Cell
2018
;
33
:
173
86
.
17.
Curtis
C
,
Shah
SP
,
Chin
SF
,
Turashvili
G
,
Rueda
OM
,
Dunning
MJ
, et al
.
The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups
.
Nature
2012
;
486
:
346
52
.
18.
Rueda
OM
,
Sammut
SJ
,
Seoane
JA
,
Chin
SF
,
Caswell-Jin
JL
,
Callari
M
, et al
.
Dynamics of breast-cancer relapse reveal late-recurring ER-positive genomic subgroups
.
Nature
2019
;
567
:
399
404
.
19.
Tan
SK
,
Lin
ZH
,
Chang
CW
,
Varang
V
,
Chng
KR
,
Pan
YF
, et al
.
AP-2gamma regulates oestrogen receptor-mediated long-range chromatin interaction and gene transcription
.
EMBO J
2011
;
30
:
2569
81
.
20.
Corces
MR
,
Granja
JM
,
Shams
S
,
Louie
BH
,
Seoane
JA
,
Zhou
W
, et al
.
The chromatin accessibility landscape of primary human cancers
.
Science
2018
;
362
:
eaav1898
.
21.
Cheung
E
,
Acevedo
ML
,
Cole
PA
,
Kraus
WL
.
Altered pharmacology and distinct coactivator usage for estrogen receptor-dependent transcription through activating protein-1
.
Proc Natl Acad Sci U S A
2005
;
102
:
559
64
.
22.
Wang
S
,
Sun
H
,
Ma
J
,
Zang
C
,
Wang
C
,
Wang
J
, et al
.
Target analysis by integration of transcriptome and ChIP-seq data with BETA
.
Nat Protoc
2013
;
8
:
2502
15
.
23.
Qiu
M
,
Bao
W
,
Wang
J
,
Yang
T
,
He
X
,
Liao
Y
, et al
.
FOXA1 promotes tumor cell proliferation through AR involving the Notch pathway in endometrial cancer
.
BMC Cancer
2014
;
14
:
78
.
24.
Fu
X
,
Jeselsohn
R
,
Pereira
R
,
Hollingsworth
EF
,
Creighton
CJ
,
Li
F
, et al
.
FOXA1 overexpression mediates endocrine resistance by altering the ER transcriptome and IL-8 expression in ER-positive breast cancer
.
Proc Natl Acad Sci U S A
2016
;
113
:
E6600
E9
.
25.
Fu
X
,
Pereira
R
,
De Angelis
C
,
Veeraraghavan
J
,
Nanda
S
,
Qin
L
, et al
.
FOXA1 upregulation promotes enhancer and transcriptional reprogramming in endocrine-resistant breast cancer
.
Proc Natl Acad Sci U S A
2019
;
116
:
26823
34
.
26.
Song
B
,
Park
SH
,
Zhao
JC
,
Fong
KW
,
Li
S
,
Lee
Y
, et al
.
Targeting FOXA1-mediated repression of TGF-beta signaling suppresses castration-resistant prostate cancer progression
.
J Clin Invest
2019
;
129
:
569
82
.
27.
Franco
HL
,
Nagari
A
,
Kraus
WL
.
TNFalpha signaling exposes latent estrogen receptor binding sites to alter the breast cancer cell transcriptome
.
Mol Cell
2015
;
58
:
21
34
.
28.
Sikora
MJ
,
Jacobsen
BM
,
Levine
K
,
Chen
J
,
Davidson
NE
,
Lee
AV
, et al
.
WNT4 mediates estrogen receptor signaling and endocrine resistance in invasive lobular carcinoma cell lines
.
Breast Cancer Res
2016
;
18
:
92
.
29.
Cancer Genome Atlas N
.
Comprehensive molecular portraits of human breast tumours
.
Nature
2012
;
490
:
61
70
.
30.
Horibata
S
,
Rice
EJ
,
Mukai
C
,
Marks
BA
,
Sams
K
,
Zheng
H
, et al
.
ER-positive breast cancer cells are poised for RET-mediated endocrine resistance
.
PLoS One
2018
;
13
:
e0194023
.
31.
Scherbakov
AM
,
Andreeva
OE
,
Shatskaya
VA
,
Krasil'nikov
MA
.
The relationships between snail1 and estrogen receptor signaling in breast cancer cells
.
J Cell Biochem
2012
;
113
:
2147
55
.
32.
Gattelli
A
,
Garcia Sola
ME
,
Roloff
TC
,
Cardiff
RD
,
Kordon
EC
,
Chodosh
LA
, et al
.
Chronic expression of wild-type Ret receptor in the mammary gland induces luminal tumors that are sensitive to Ret inhibition
.
Oncogene
2018
;
37
:
4046
54
.
33.
Liu
Z
,
Lam
N
,
Thiele
CJ
.
Zinc finger transcription factor CASZ1 interacts with histones, DNA repair proteins and recruits NuRD complex to regulate gene transcription
.
Oncotarget
2015
;
6
:
27628
40
.
34.
van de Ven
RA
,
Tenhagen
M
,
Meuleman
W
,
van Riel
JJ
,
Schackmann
RC
,
Derksen
PW
.
Nuclear p120-catenin regulates the anoikis resistance of mouse lobular breast cancer cells through Kaiso-dependent Wnt11 expression
.
Dis Model Mech
2015
;
8
:
373
84
.
35.
Sanda
T
,
Lawton
LN
,
Barrasa
MI
,
Fan
ZP
,
Kohlhammer
H
,
Gutierrez
A
, et al
.
Core transcriptional regulatory circuit controlled by the TAL1 complex in human T cell acute lymphoblastic leukemia
.
Cancer Cell
2012
;
22
:
209
21
.
36.
Rheinbay
E
,
Parasuraman
P
,
Grimsby
J
,
Tiao
G
,
Engreitz
JM
,
Kim
J
, et al
.
Recurrent and functional regulatory mutations in breast cancer
.
Nature
2017
;
547
:
55
60
.
37.
Zheng
R
,
Dong
X
,
Wan
C
,
Shi
X
,
Zhang
X
,
Meyer
CA
.
Cistrome Data Browser and Toolkit: analyzing human and mouse genomic data using compendia of ChIP-seq and chromatin accessibility data
.
Quant Biol
2020
;
8
:
267
76
.
38.
Young
RA
.
Control of the embryonic stem cell state
.
Cell
2011
;
144
:
940
54
.
39.
Whyte
WA
,
Orlando
DA
,
Hnisz
D
,
Abraham
BJ
,
Lin
CY
,
Kagey
MH
, et al
.
Master transcription factors and mediator establish super-enhancers at key cell identity genes
.
Cell
2013
;
153
:
307
19
.
40.
Fei
T
,
Li
W
,
Peng
J
,
Xiao
T
,
Chen
CH
,
Wu
A
, et al
.
Deciphering essential cistromes using genome-wide CRISPR screens
.
Proc Natl Acad Sci U S A
2019
;
116
:
25186
95
.
41.
Hafner
M
,
Niepel
M
,
Chung
M
,
Sorger
PK
.
Growth rate inhibition metrics correct for confounders in measuring sensitivity to cancer drugs
.
Nat Methods
2016
;
13
:
521
7
.
42.
Wang
Q
,
Cheng
Y
,
Wang
Y
,
Fan
Y
,
Li
C
,
Zhang
Y
, et al
.
Tamoxifen reverses epithelial-mesenchymal transition by demethylating miR-200c in triple-negative breast cancer cells
.
BMC Cancer
2017
;
17
:
492
.
43.
Paakinaho
V
,
Swinstead
EE
,
Presman
DM
,
Grontved
L
,
Hager
GL
.
Meta-analysis of chromatin programming by steroid receptors
.
Cell Rep
2019
;
28
:
3523
34
.
44.
Swinstead
EE
,
Miranda
TB
,
Paakinaho
V
,
Baek
S
,
Goldstein
I
,
Hawkins
M
, et al
.
Steroid receptors reprogram FoxA1 occupancy through dynamic chromatin transitions
.
Cell
2016
;
165
:
593
605
.
45.
Riggins
RB
,
Lan
JP
,
Zhu
Y
,
Klimach
U
,
Zwart
A
,
Cavalli
LR
, et al
.
ERRgamma mediates tamoxifen resistance in novel models of invasive lobular breast cancer
.
Cancer Res
2008
;
68
:
8908
17
.
46.
Sottnik
JL
,
Bordeaux
EK
,
Mehrotra
S
,
Ferrara
SE
,
Goodspeed
AE
,
Costello
JC
, et al
.
Mediator of DNA damage checkpoint 1 (MDC1) is a novel estrogen receptor coregulator in invasive lobular carcinoma of the breast
.
Mol Cancer Res
2021
;
19
:
1270
82
.
47.
Shang
Y
,
Hu
X
,
DiRenzo
J
,
Lazar
MA
,
Brown
M
.
Cofactor dynamics and sufficiency in estrogen receptor-regulated transcription
.
Cell
2000
;
103
:
843
52
.
48.
Shang
Y
,
Brown
M
.
Molecular determinants for the tissue specificity of SERMs
.
Science
2002
;
295
:
2465
8
.
49.
Howell
A
,
Dodwell
DJ
,
Anderson
H
,
Redford
J
.
Response after withdrawal of tamoxifen and progestogens in advanced breast cancer
.
Ann Oncol
1992
;
3
:
611
7
.
50.
Ediriweera
MK
,
Tennekoon
KH
,
Samarakoon
SR
.
Emerging role of histone deacetylase inhibitors as anti-breast-cancer agents
.
Drug Discov Today
2019
;
24
:
685
702
.
51.
Kurmis
AA
,
Dervan
PB
.
Sequence-specific suppression of androgen receptor-DNA binding in vivo by a Py-Im polyamide
.
Nucleic Acids Res
2019
;
47
:
3828
35
.

Supplementary data