Inhibition of NR2F2 restores hormone therapy response to endocrine refractory breast cancers

Abstract

Endocrine resistance is frequently encountered in estrogen receptor alpha positive (ER+) breast cancer patients, often due to somatic mutations such as Neurofibromin 1 (NF1) loss. The mechanism(s) by which ER-directed proliferation is lost in such cases is unknown, thwarting pharmacologic strategies involving continued endocrine therapy. In this study, we identified NR2F2, an orphan nuclear receptor, to be essential for NF1 loss induced endocrine resistance. An induction of NR2F2, via activation of the MAPK pathway upon NF1 loss or other MAPK pathway genetic alterations, was observed in ER+ cell models and patient samples. Mechanistically, increased NR2F2 orchestrates a repressed ER transcriptional program by repartitioning the ER cistrome, altering the balance of its associated transcriptional coregulators and modifying global chromatin accessibility. Accordingly, genetic knockout or pharmacologic inhibition of NR2F2 restores the antitumor responses to hormone therapy in multiple endocrine refractory models. Together, our findings identify the nuclear receptor NR2F2 to be a modulator of the hormone response pathway in breast cancer whose inhibition represents a promising anticancer strategy.

One Sentence Summary

NR2F2 promotes endocrine resistance by repressing the ERα transcriptional program and represents a promising therapeutic target for refractory breast cancers

Introduction

Estrogen receptor alpha (ER) serves as a master regulator of the transcriptome in both normal and transformed mammary cells, binding to and regulating transcriptional activation of thousands of loci (1). Exogenous exposure to ER agonists that stimulate ER transcriptional activation has been shown to increase breast cancer risk (2-4) while drugs that inhibit ER activity are highly effective anticancer therapies (5-8). Hormone therapies, including aromatase inhibitors, tamoxifen and fulvestrant, are widely used systemic therapies against ER positive breast cancers. Clinical resistance to hormone therapy commonly occurs via different acquired genetic alterations (9-18), with loss-of-function (LOF) mutations of neurofibromin 1 (NF1) being among the most common (12). Previous studies have confirmed the ability of NF1 loss to promote endocrine resistance but do not fully explain how these alterations disrupt the ER-directed growth program or whether this could be exploited therapeutically.

In this work, we investigated the dependencies of hormone-resistant breast cancers using the NF1 mutant context and identified the orphan nuclear receptor NR2F2 as an essential mediator of endocrine resistance and tumor growth. NR2F2 upregulation is found to suppress the ER-dependent growth program such that its depletion or inhibition restores endocrine response, leading to tumor regressions in multiple endocrine refractory tumors, nominating NR2F2 as a broadly relevant target for ER+ breast cancers.

Results

NR2F2 mediates survival of NF1 loss ER+ tumors

To characterize the scope of NF1 alterations in hormone receptor positive (HR+) breast cancer, we surveyed the genetic landscape using MSK-IMPACT from a cohort of over 2,600 clinically annotated HR+ and HER2 non-amplified breast cancers (Fig. 1A). Somatic mutations of NF1 were observed in 5.2% of patient samples and were found at a higher proportion in metastatic than primary tumors (6.9% vs 2.5%, p<0.0001), including 2.5% vs 1.3% mutations without loss of heterozygosity (LOH), 3.6% vs 0.7% mutations with LOH, and 0.8% vs 0.1% homozygous deletions (Fig. 1A, Table S1). Given the prevalence of NF1 altered tumors, we sought to identify unique vulnerabilities for this context by performing CRISPR/Cas9 screens in wildtype (WT) and NF1 knockout (KO) MCF7 cells (Fig. S1A). We identified 16 candidates with at least three sgRNAs specifically dropped out in NF1 KO cells (Fig. S1B and table S2). These candidates included several nuclear proteins, including epigenetic modifiers CREBBP, SMARCD1, KAT2A, KDM5C, HDAC2 and DOT1L, the orphan nuclear receptor NR2F2, protein phosphatase DUSP4, and SUMO-protein ligase CBX4. NR2F2 scored as the top candidate as its three sgRNAs were the most depleted (Fig. 1B). NR2F2 has been previously noted to have roles in development (19-22), and cancer progression in various lineages including prostate cancer (23), breast cancer (24) and squamous cell skin cancer (25). Moreover, NR2F2 has been incidentally found to interact with ER in broad proteomic screens (16, 26-28).

Fig. 1. NR2F2 mediates cell growth and survival of NF1 loss ER+ breast cancer.

(A) The frequencies of NF1 genomic alterations (mutations with or without loss of heterozygosity and homozygous deletion) in primary and metastatic ER+ breast cancer samples from MSK-IMPACT dataset.

(B) Ranked log2 fold change (FC) of individual sgRNA specifically depleted in NF1 loss MCF7 cells from data in Fig. S1B.

(C) Tumor growth curve of xenografted MCF7 cells transduced with the indicated sgRNAs. 5 mice in each group.

(D) Competition cellular growth assay results showing relative cell growth of MCF7 sgNT, sgNF1, or DKO cells treated with 100nM fulvestrant (Ful) (n = 4) or estrogen-depleted media (n = 4) for 7 days.

(E) Tumor growth of MCF7 sgNT, sgNF1 and DKO xenografts treated with 5mg Ful per mouse per week. 5 mice in each group.

(F) Relative tumor volume change compared to the initial tumor volumes of NF1 loss sgNT or sgNR2F2 PDX tumors at the indicated days. Tumors were treated with vehicle or 5mg Ful/mice, twice per week, 5 mice in each group.

(G) Tumor growth of PDOX_64aS1 models (sgNT, sgNF1 and DKO) treated with either vehicle or 3 mg Ful per mouse per week. 5 mice in each group.

Data were presented as mean ± SEM. P values were calculated using Fisher’s exact test followed by Benjamini-Hochberg multiple testing correction (A), two-tailed multiple Mann Whitney test [(C), (E), and (G)] or one-way ANOVA followed by Tukey’s test [(D) and (F)]. n.s., not significant, *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001 (p value or adjusted p value definition applies to all Figures).

For initial validation, we analyzed NF1 and NR2F2 expression in a panel of ER+ cell lines (Fig. S1C) demonstrating their co-expression. To confirm NR2F2 dependency in NF1 loss ER+ cells, we generated control (sgNT) and NR2F2 KO cells (Fig. S1D) and performed competitive cell growth assay (Fig. S1E). The NF1 KO-induced growth advantage was eliminated upon NR2F2 KO in the dual NF1/NR2F2 KO (DKO) cells (Fig. S1F). The reduced fitness of DKO cells was confirmed in two additional ER+ cell lines, T47D and MDA-MB-361 (Fig. S1D, G). To investigate the relevance of NR2F2 to tumor growth in vivo, MCF7 isogenic cells were grown as xenografts and tumor growth was monitored. We found MCF7 NF1 KO xenografts grew faster compared to WT control tumors, while loss of NR2F2 suppressed the growth of NF1 KO tumors to nearly that of the control tumors (Fig. 1C).

We next investigated the contribution of NR2F2 to NF1 loss mediated endocrine resistance. Compared to control cells, NF1 KO cells showed substantial outgrowth on fulvestrant or estrogen depletion treatment, whereas there was no significant difference observed for DKO cells (Fig. 1D, Fig. S1H-J). Moreover, NR2F2 knockdown by inducible shRNAs restored the endocrine sensitivity of NF1 loss cells (Fig. S1K-M).

Lastly, the potential of NR2F2 depletion to restore endocrine sensitivity in vivo was assessed using isogenic MCF7 xenografts, patient-derived xenografts (PDX) and patient-derived organoid-based xenograft (PDOX) models carrying endogenous or CRISPR-Cas9 edited NF1 and/or NR2F2 loss (Table S3). As expected, NF1 KO MCF7 tumors proved resistant to fulvestrant while DKO tumors displayed equivalent tumor regression as control tumors (Fig. 1E). We knocked out NR2F2 in the PDX_7aS1 model which carries an NF1 truncating mutation (Q1841Lfs*20) and selected sgNT control and sgNR2F2 tumors with comparable steady-state ER expression for further study (Fig. S1N). We also established treatment-naïve ER+ primary tumor derived organoids (PDO) and PDO-based xenografts (PDOX) with NF1 and/or NR2F2 KO (Fig. S1O). The PDX_7aS1 and PDOX_64aS1-sgNF1 tumors did not respond to fulvestrant treatment, while NR2F2 depletion (PDX_7aS1-sgNR2F2 and PDOX_64aS1-DKO) showed near complete inhibition of tumor growth (Fig. 1F-G). Altogether, the data reveal that NR2F2 is essential for NF1 loss-mediated tumor growth and endocrine resistance in ER+ breast cancers.

NF1 loss upregulates NR2F2 via the MAPK pathway

NF1 loss has been linked to endocrine resistance through its GAP activity for RAS GTPases (12, 17). We observed an induction of NR2F2 expression across NF1 KO models (Fig. S1C, S1N). We interrogated the RAS effector MAPK pathway using selective MEK and ERK kinase inhibitors and found that NR2F2 expression in NF1 KO cells was suppressed in a dose dependent manner (Fig. 2A and Fig. S2A-C). Conversely, inhibition of another RAS effector pathway, PI3K-AKT, had no effect on NR2F2 (Fig. 2A and Fig. S2A, C). Moreover, overexpression of a constitutively active, RAF-independent MEK1 mutant (29), MEK1 ΔE102-I103, increased NR2F2 expression in MCF7 and T47D cells while the ERK inhibition blocked this increase (Fig. S2D). By contrast, overexpression of a constitutively active AKT1 allele (30), myrAKT1, that activated AKT signaling but reduced MAPK activity, suppressed the expression of NR2F2 in both cell lines (Fig. S2E). The induction of NR2F2 was similarly observed in additional contexts involving MAPK activation such as overexpression of KRAS (Fig. S2F). To further examine the impact of NF1 loss on NR2F2 expression, we generated isogenic cell models of triple negative breast cancer (TNBC), including MDA-MB-453, HCC1395, BT549 and CAL51. Unlike our observations in ER+ cell lines, NF1 KO in these ER-negative models did not promote NR2F2 expression (Fig. S2G). These findings suggest that the regulatory relationship between NF1 and NR2F2 may have subtype-specific differences in breast cancer.

Fig. 2. NF1 loss induces endocrine resistance via NR2F2.

(A) Immunoblotting results with the indicated antibodies of MCF7 sgNT and sgNF1 cells treated with DMSO, 2 μM SCH772984, 2 μM trametinib, 1 μM BYL719 or 2 μM GDC0068 for 24hrs.

(B) NR2F2 mRNA abundances in ER+ patients from METABRIC cohort. NF1 WT/MEK signature^low subset: patients without NF1 genomic alterations and low MEK_Up.V1_Up signature (n = 413); NF1 WT/MEK signature^high subset: patients without NF1 genomic alterations and high MEK_Up.V1_Up signature (n = 374); NF1 loss subset: patients with NF1 genomic alterations including homozygous deletion, heterozygous deletion and driver mutations (n = 227).

(C) Relative cell growth of MCF7 and T47D cells overexpressing NR2F2 compared to empty vector (EV) expressing control cells with the treatment of DMSO or 100nM Ful measured using competition cellular growth assay (n = 3).

(D) Representative clonogenic proliferation assay results of MCF7 and T47D cells overexpressing EV or NR2F2 treated with DMSO or 100nM Ful for 7 days (n ≥ 2).

(E) Tumor growth of MCF7 xenografts transduced with EV or NR2F2 treated with vehicle or 3mg Ful per mouse per week. 5 mice in each group. The day started the treatment was shown as day 0. Tumor volumes were measured every three or four days.

(F) Relative cell growth of MCF7 (left panel, n = 6) and T47D cells (right panel, n = 5) transduced with two individual sgRNAs targeting NR2F2 with the treatment of 100nM Ful, 500nM 4-OHT or estrogen-depleted (ED) media for 7 days as measured by competition cellular growth assay.

Data were shown as median with interquartile range (B) or mean ± SEM [(C), (E), and (F)]. P values were calculated using one-way ANOVA followed by Tukey’s test (B), two-tailed unpaired Student’s t tests with Welch’s correction (C), two-tailed Mann Whitney test (E) or two-way ANOVA followed by Tukey’s test (F).

To investigate whether the correlation between NR2F2 expression and NF1 mutation status or MAPK pathway activity could be observed in breast cancer patients, we analyzed NR2F2 mRNA expressions in ER+ samples from the METABRIC dataset (31, 32). NR2F2 expressions were more abundant in tumors with either deleterious NF1 mutations or high MAPK pathway activity (Fig. 2B, Fig. S2H). These data imply that NF1 loss mediates its tumor promoting and endocrine resistant phenotypes in part via MAPK pathway-mediated induction of NR2F2 expression.

NR2F2 mediates endocrine resistance in ER+ breast cancer models

To determine whether induction of NR2F2 expression is sufficient to recapitulate NF1 loss-induced phenotypes, we stably overexpressed NR2F2 in MCF7 and T47D cells (Fig. S2I) and evaluated their responsiveness to fulvestrant. While the growth rate of NR2F2 overexpressing (OE) cells was slightly reduced (Fig. 2C, D), these cells were highly resistant to fulvestrant (Fig. 2C, D, and Fig. S2J). The effect of NR2F2 OE on fulvestrant response was confirmed across multiple additional ER+ cell lines, including CAMA1, EFM19, ZR751 and HCC712 (Fig. S2K). NR2F2 OE similarly mediated resistance to additional endocrine therapies including 4-hydroxytamxifen (4-OHT) and estrogen deprivation (Fig. S2L). To determine the role of NR2F2 in vivo, we generated orthotopic MCF7 xenografts expressing empty vector (EV) or NR2F2 and monitored their responses to fulvestrant treatment. Whereas NR2F2 OE tumors initially showed comparable fulvestrant sensitivity as the control tumors, they eventually manifested tumor progression, unlike control tumors (Fig. 2E). Additionally, we found that NR2F2 KO enhanced the response of NF1 WT cells to endocrine therapies by 20-30% (Fig. 2F and Fig. S1K-M, S2M-N). Together, our data point to a critical role of NR2F2 in limiting the response of breast cancer cells to endocrine therapy.

NR2F2 interferes with ER signaling

To elucidate the basis for NR2F2 regulated endocrine response, we performed RNA sequencing (RNA-seq) of MCF7 control, NF1 KO, NR2F2 OE, NR2F2 KO or DKO cells (Fig. S3A). We first compared gene expression alterations induced by NF1 KO and NR2F2 OE and identified approximately 700 overlapping transcripts altered for both endocrine resistant models (Fig. 3A and Fig. S3B). Pathway Analysis revealed that these overlapping genes were highly enriched in estrogen response gene sets and endocrine resistance gene sets (Fig. 3A and table S4). Consistent with a prior report on the role for NR2F2 in promoting epithelial-mesenchymal transition (EMT) (25), we also observed enrichment of the EMT signature. By contrast, we observed the opposite differential gene expression patterns induced by NF1 KO and NR2F2 KO, many of which were abolished in DKO cells, including the genes congruously altered in NF1 KO and NR2F2 OE cells (Fig. S3A and Fig. 3B). Consistently, gene set enrichment analysis (GSEA) using hallmark gene sets revealed that, among the top enriched gene signatures, estrogen response gene sets were negatively enriched in NF1 loss cells and positively enriched in NR2F2 loss cells, whereas these signatures were either slightly or not enriched in DKO cells (Fig. 3C and table S5). The expression changes of multiple canonical ER target genes were validated by RT-qPCR (Fig. 3D). As ER signaling is important for controlling various cell cycle-related genes, we analyzed NF1 and NR2F2 regulated cell cycle-related ER target genes using GSEA and EstrogenDB. We found NR2F2 may impact expression of a subset of proliferative ER targets (GINS2, SFN, JAK2, RBBP8, MCM7, PKMYT1, SKP2 and RBL2) (Fig. S3C). These data suggest that NF1 loss might alter ER-dependent transcription via its induction of NR2F2.

Fig. 3. NR2F2 regulates the ER transcriptional program.

(A) Venn diagram and enriched signatures of the overlapped differentially expressed genes with the same change directions between MCF7 NF1 loss and NR2F2 OE cells compared to control cells.

(B) Scatter plots showing the expression of up- and down-regulated genes from Fig. 3A in NR2F2 KO and DKO cells.

(C) Gene set enrichment analysis (GSEA) plot of Hallmark estrogen response early and late gene sets in MCF7 sgNF1, sgNR2F2 and DKO cells. Normalized enrichment scores (NES) and adjusted p values obtained from the GSEA are shown.

(D) RT-qPCR validation of the expression of multiple ER canonical target genes in sgNT, sgNF1 and DKO cells. At least two independent biological experiments were conducted.

(E) Gene Set Variation Analysis (GSVA) results of estrogen response gene sets in ER+ patient samples from METABRIC dataset with high or low NR2F2 mRNA expression (top panel), or with high or low NF1 mRNA expression (bottom panel). High or low expression was defined based on the median values of expression.

Data were shown as median with interquartile range [(B) and (E)] or mean ± SEM (D). P values were calculated using one-way ANOVA followed by Games-Howell's test (B) or Tukey’s test (D), or two-tailed Student’s t test (E).

To further confirm the role of NR2F2 in ER-dependent transcription, we examined transcriptional profiles of these cell models upon estrogen stimulation or fulvestrant treatment. Indeed, we observed impaired expression of estrogen response genes in NF1 KO and NR2F2 OE cells, which were enhanced in NR2F2 KO cells and restored in DKO cells (Fig. S3D, E). Both NF1 loss and NR2F2 OE suppressed the gene expression response to fulvestrant, while the transcriptional output was more potently repressed by fulvestrant in NR2F2 KO cells (Fig. S3F). These results revealed that NR2F2 plays an integral role in linking NF1 loss to the ER-dependent transcriptional program. We next investigated whether NR2F2 expression might be linked with ER activity in clinical samples from the METABRIC cohort. We stratified ER+ tumors into two groups, NR2F2-low and NR2F2-high, according to their mRNA expression. We found that NR2F2-low ER+ patients had higher ER transcriptional activity compared to NR2F2-high patients (Fig. 3E). Consistent with our in vitro results, inverse ER transcriptional activities were observed in NF1 mRNA expression low and high samples (Fig. 3E).

NR2F2 modulates ER chromatin binding

Given the role of NR2F2 as a nuclear receptor in modulating ER-dependent transcription, we hypothesized that NR2F2 may regulate ER chromatin binding. To this end, we performed chromatin immunoprecipitation followed by sequencing (ChIP-seq) of NR2F2 and ER ChIP-seq to map their genome-wide binding sites. We found that about 80% of NR2F2 binding sites distributed in the gene body (~50%) and intergenic (~30%) regions, and ~18% of NR2F2 peaks were localized at promoter regions (Fig. S4A). Compared to control cells, NF1 KO cells displayed a remarkable increase of NR2F2 chromatin binding events (~40% of total peaks), and a decrease at a small set of sites (~4% of total peaks) (Fig. S4A). Motif analysis identified enrichment of GRHL2, JunB, FOXA1, GATA3, TRPS1 and ER motifs in NR2F2 peaks (Fig. S4B and table S6), suggesting the potential cooperation of NR2F2 with these transcription factors. Consistently, we observed intensive overlaps of ER and NR2F2 chromatin binding regions in both control and NF1 KO cells (Fig. 4A). We then analyzed their physical distances and found that the center of more than 40% of ER binding peaks in both cell models are within 200bp of NR2F2 peak centers (Fig. 4B, C). The proximity of ER and NR2F2 peaks, and the sequence consensus of canonical NR2F2 motif and estrogen response element (ERE) half-site, suggests ER and NR2F2 may share identical binding sites within these regions (Fig. 4B, C). To corroborate ER and NR2F2 binding to the same or proximal chromatin sites, we performed NR2F2-ER proximity ligation assays (PLA) and NR2F2_ChIP-ER_ReChIP. We found that NR2F2 interacts with ER, with the interaction being increased in NF1 KO cells (Fig. 4D, S4C). The NR2F2_ChIP-ER_ReChIP results revealed that ER co-occupied at approximately 40% of NR2F2 chromatin binding sites. Additionally, the ERE motif was enriched within NR2F2-specific sites lacking ER co-occupancy. These findings suggest that NR2F2 and ER may either bind together or compete at the same chromatin sites.

Fig. 4. NR2F2 modulates ER chromatin binding.

(A) Venn diagram showing shared regions bound by ER and NR2F2 in MCF7 sgNT and sgNF1 cells under full media condition. Canonical ER and NR2F2 binding motifs were shown.

(B) Distances of ER and NR2F2 ChIP-seq peak centers in MCF7 sgNT (top panel) and sgNF1 (bottom panel) cells. Percentages of peaks within 200 bp were shown.

(C) Representative ChIP-seq tracks of ER and NR2F2 in MCF7 sgNT, sgNF1, sgNR2F2 or DKO cells at loci of chromosome 21 (top) and chromosome 14 (bottom).

(D) The colocalization of NR2F2 and ER were determined using the Proximity ligation assay and confocal imaging. Anti-ER antibody (#8644) from Cell Signaling was used. The number and mean intensity of NR2F2 and ER interaction foci were quantified using ImageJ and shown as mean ± SEM. Eight images with at least 30 cells each were analyzed for each condition. P values were calculated using two-tailed unpaired Student’s t tests with Welch’s correction.

(E) Average density plots (top panels) and heatmap (bottom panels) showing dynamic ER ChIP-seq peak changes in MCF7 sgNT, sgNF1, sgNR2F2 and DKO cells. Peak numbers of each cluster were shown.

(F) Representative homer known motifs enriched in each ER ChIP-seq cluster.

(G) Average density plots (top panels) and heatmap (bottom panels) of NR2F2 chromatin binding peaks within ER ChIP-seq clusters.

Profiling of ER binding revealed that NF1 loss, NR2F2 depletion or DKO altered ER genome-wide bindings compared to control cells (Fig. S4E). We next compared the binding patterns of these differential ER binding peaks cross cell models by subdividing them into 5 clusters using k-means clustering analysis (Fig. 4E). NR2F2 KO alone increased ER bindings at sites of clusters 3-4 and a decrease in clusters 1, 2 and 5. NF1 loss specifically increased ER binding events at sites of cluster 1 and strongly reduced ER binding at most sites including clusters 2-5. NF1 loss enhanced or impaired ER bindings in clusters 1-3 were notably restored to comparable state as control cells upon NR2F2 KO in DKO cells. We validated ER binding changes at multiple sites using ChIP-qPCR (Fig. S4F). Homer known motif enrichment analysis revealed the most remarkable enrichments of multiple transcription factor binding motifs within clusters 2 and 3, including nuclear receptors such as ER and NR2F2 motifs, forkhead box protein such as FOXA1 motif, GATA3/TRPS1 motif, AP2 motif, GRHL2 motif, AP-1 TFs such as JunB motif and TEAD motif (Fig. 4F and table S7). Consistently, our NR2F2 ChIP-seq analysis confirmed the occupancy of NR2F2 at dynamic ER sites (Fig. 4G), with the most intensive binding in clusters 2 and 3. NR2F2 binding in ER clusters 1-3 were either increased or decreased in NF1 KO cells, suggesting that NR2F2 could assist or interfere with ER chromatin binding at these sites.

Aligned with ER chromatin binding changes in NF1 KO and DKO cells under full media condition, a high proportion of NF1 KO decreased or increased estrogen-stimulated ER chromatin binding events (71.4% of total dynamic ER peaks) were restored to the control state in DKO cells (Fig. S4G). Together, these data suggest that NF1 loss promotes a redistribution of the ER cistrome globally via its induction of NR2F2.

NR2F2 interacts with multiple chromatin remodelers and transcriptional coregulators

Nuclear receptors including ER regulate transcription through the recruitment of numerous regulatory co-factors, such as other TFs, chromatin remodeling complexes, coactivators, and corepressors (33, 34). The regulation of NR2F2 on ER-dependent transcriptome prompted us to further ascertain whether and how NR2F2 induction influences coregulators dynamics. We performed quantitative multiplexed rapid immunoprecipitation mass spectrometry of endogenous proteins (qPLEX-RIME) (28) to map NR2F2 interactomes on chromatin in the sgNT control, NF1 KO and DKO MCF7 cells with four independent biological replicates and using IgG as control (Fig. S5A). We identified 2 and 237 proteins specifically enriched in sgNT control and NF1 KO cells respectively, and 190 proteins in both cells including the bait protein NR2F2 (Fig. 5A, Fig. S5B-C and table S8). The interaction of NR2F2 with 114 shared proteins were significantly (p<0.05) reinforced in sgNF1 cells (table S8), including ER which we confirmed by PLA (Fig. 4D and S4E). Several proteins have been previously detected to interact with NR2F2, including ER (16, 27, 28), CBP (28), NCOA3 (28), POLR2A (28), , SMAD4 (19), BCL11B (35), ARID1A and BRG1 (16). Grouping interactors based on their known function(s), we found that about 60% of NR2F2 interacting proteins are nuclear receptors (NRs) (Fig. 5B), components of chromatin remodeling complexes such as the SWI/SNF complex (Fig. 5C), transcriptional corepressor complexes like HDAC1/2 containing NuRD and CoREST complexes (Fig. 5D), transcriptional coactivators like CBP/p300 and COMPASS complex (Fig. 5E), transcription factors (TFs) such as GRHL1/2, JunB, FOXA1, SMAD3/4, GATA3 and STAT3 (Fig. S5D), and chromatin organization and stability related proteins (Fig. S5B-C).

Fig. 5. NR2F2 interacts with chromatin remodeling complexes and transcriptional coregulators.

(A) Venn diagram showing the numbers of sgNT cells specific, sgNF1 cells specific and shared NR2F2 interacting proteins.

(B-E) Fold changes of the indicated NR2F2 interactors in the anti-NR2F2 samples compared to IgG control in MCF7 sgNT and sgNF1 cells. (B) Nuclear receptors, (C) components of ATP-dependent chromatin remodeling complexes, (D) components of transcription corepressor complexes and (E) components of transcription coactivator complexes interact with NR2F2 either in MCF7 sgNT or sgNF1 cells. NR2F2 interactomes were characterized using qPLEX-RIME with four biological replicates of each cell line.

(F) Comparison of fold changes of transcriptional corepressors and coactivators interacting with NR2F2 in MCF7 sgNT and sgNF1 cells. Data were shown as median with interquartile range and analyzed using two-way ANOVA followed by Tukey's test.

To further establish the association of these cooperating factors with NR2F2 on chromatin, we assessed the recruitment of nuclear receptors ER and ESRRA, chromatin remodelers DPF2, SMARCE1 and SMARCA5, transcriptional corepressors HDAC2, MTA1/3, GATAD2B, CTBP1, RCOR1, ZNF217 and SIN3A, transcriptional coactivators NCOA3 and EP300, and TFs such as FOXA1 and GATA3 at NR2F2 binding sites using reference ChIP-seq data in MCF7 cells from the Encyclopedia of DNA elements (ENCODE). We found that most NR2F2 chromatin binding sites showed substantial co-occupancies with the majority of the forementioned factors (Fig. S5E-J). These results are consistent with the motif analyses for NR2F2 binding regions (Fig. S4B).

The finding of NR2F2 interactions with coactivator and corepressor complexes raised the question as to the functional outcome of these interactions upon NR2F2 expression changes. To address this question, we compared the fold changes of NR2F2 interacting coactivators and corepressors in control and NF1 KO cells (Fig. 5F). NR2F2 displayed higher affinity to several corepressors, such as TLE5, NCOR1 and DNTTIP1, in both control and NF1 KO cells. NR2F2 showed similar balance of interactions with coactivators and corepressors in control cells, while relative interactions with corepressors were dramatically induced compared to coactivators in NF1 KO cells, suggesting a reordering of NR2F2-mediated regulatory cofactors dynamics. In sum, this data point to an enrichment of NR2F2 interaction with corepressors, which may be a key mechanism for suppression of ER function and dependence in NF1 KO cells.

NR2F2 modulates chromatin accessibility in NF1 loss cells

Given the interaction of NR2F2 with chromatin structure related proteins, such as epigenetic modifiers, chromatin remodelers, and components of nucleosomes, cohesin and heterochromatin (Fig. S5B-C, table S8), we speculated that chromatin accessibility, including ER-regulated regions, might also be altered in NF1 loss cells. To test this hypothesis, we performed the assay for transposase-accessible chromatin sequencing (ATAC-seq). Compared to control cells, NF1 KO, NR2F2 KO and DKO altered the chromatin accessibility at 25148 (35.5%), 5985 (10.4%) and 10135 (15.5%) sites respectively (Fig. S6A). A high percentage of decreased or lost dynamic peaks in NF1 loss cells were located at promoter regions while gained or increased peaks were mainly in distal intergenic and gene body regions. The lost or gained dynamic peaks in NR2F2 loss or DKO cells displayed slightly increased distributions in distal intergenic and gene body regions, indicative of enhancers.

We grouped these dynamic peaks into 5 clusters using k-means clustering analysis to compare their changes cross cell models (Fig. 6A). NR2F2 KO cells had increased or decreased chromatin accessibility at sites of clusters 2 and 5, respectively. In accord with the increased interactions of NR2F2 with chromatin modifiers containing coactivator and corepressor complexes (Fig. 5C, D), NF1 loss was associated with gained accessibility at 6.92% of sites (cluster 1) and a widespread decrease in chromatin accessibility at 93.08% of these dynamic sites (clusters 2-5) compared to control cells. Importantly, all NF1 KO induced chromatin accessibility changes were completely or partially rescued by NR2F2 depletion in DKO cells, suggesting that NR2F2 is a key mediator of these events. Integration of ATAC-seq and NR2F2 ChIP-seq data revealed remarkable NR2F2 binds at the differentially accessible sites, with most sites displayed enhanced or gained NR2F2 binding in NF1 loss cells (Fig. 6B). Chromatin accessibility and NR2F2 binding changes suggest that NR2F2 could both positively and negatively regulate chromatin accessibility, which most likely depends on its associated chromatin regulatory proteins, consistent with the protein-protein interaction analyses.

Fig. 6. NR2F2 modulates chromatin accessibility.

(A) Average density plots (top panels) and heatmap (bottom panels) showing dynamic ATAC-seq peak changes in MCF7 sgNT, sgNF1, sgNR2F2 and DKO cells. Peak numbers of each cluster were shown.

(B) Average density plots (top panels) and heatmap (bottom panels) of NR2F2 chromatin binding peaks within ATAC-seq clusters.

(C) Association of NF1 KO altered and DKO restored genes (Fig. S3A) with dynamic ATAC-seq peaks in each cluster (Fig. 5F).

(D) Chromatin accessibility of altered ER peaks from Fig. 4D.

We next did integrative analysis to elucidate the correlation between chromatin accessibility changes and gene expression changes. We idefined a set of genes that were up- or down-regulated in NF1 KO cells but completely or partially restored in DKO cells. We found that NF1 KO cells gained accessible chromatin regions in cluster 1, which were linked to upregulated genes, and chromatin regions in clusters 2-5 were associated with both upregulated and downregulated genes (Fig. 6C). Our data suggest that NR2F2-modulated chromatin accessibility changes contribute to gene expression changes, which most presumably also involve the regulation of transcription cofactors.

Homer Known motif analysis revealed enriched forkhead box protein binding motifs such as FOXA1, AP-1 binding motifs such as JunB, GRHL2 motif, as well as ER and NR2F2 motifs, in many ATAC-seq clusters (Fig. S6B and table S9). The enrichment of ER motif suggests that chromatin accessibility changes might be correlated with ER binding alterations. We investigated the chromatin accessibility of ER dynamic binding sites (Fig. 4D) in these models. NR2F2 KO cells showed increased chromatin accessibility at ER ChIP-seq cluster 3 sites. While NF1 KO cells has decreased accessibility of all ER dynamic binding sites, this decrease was rescued in DKO cells (Fig. 6D). Therefore, decreased chromatin accessibility may also contribute to decreased ER binding in NF1 loss cells.

Taken together, NR2F2 mediates endocrine resistance in NF1 loss cells via impairing ER signaling, which was achieved through modulating ER cistrome, transcriptional coregulators dynamics and chromatin accessibility.

NR2F2 is a druggable target in multiple endocrine resistant models

As ER dependence is lost in many different breast cancers that develop resistance to antiestrogen therapy, we examined the significance of NR2F2 in several models of clinical endocrine resistance. We first investigated KRAS overexpression as an alternative MAPK pathway activating mechanism and observed increased NR2F2 expression upon KRAS OE (Fig. S2F). NR2F2 KO significantly (p<0.0001) mitigated the fulvestrant resistance in KRAS OE cells (Fig. 7A). Genomic deletion of tumors suppressors like PTEN or ARID1A have recently been found to be recurrent mechanisms of endocrine resistance that alter ER signaling and/or chromatin accessibility (12, 16, 18, 36). To test whether NR2F2 could be targeted in these contexts, we knocked out NR2F2 in ARID1A KO or PTEN KO MCF7 cells (Fig. S7A). We found that NR2F2 depletion significantly (p<0.0001) sensitized both ARID1A KO and PTEN KO cells to fulvestrant and/or estrogen depletion treatment (Fig. 7B, C).

Fig. 7. NR2F2 is an actionable target in multiple endocrine refractory breast cancer models.

(A) Cell viabilities of MCF7 EV, KRAS OE and KRAS OE&NR2F2 KO cells treated with 50nM Ful or combined with the serial diluted NR2F2 inhibitors measured at day 7 using CellTiter-Glo assay (n = 5).

(B) Competition cellular growth assay results showing relative cell growth of MCF7 sgNT, sgPTEN, or DKO cells treated with 100nM Ful or estrogen-depleted media for 7 days (n = 3).

(C) Competition cellular growth assay results showing relative cell growth of MCF7 sgNT, sgARID1A, or DKO cells treated with 100nM Ful or estrogen-depleted media for 7 days (n = 3).

(D) Representative clonogenic proliferation assay results of MCF7 sgNT, sgNR2F2, sgNF1 and DKO cells treated with 50nM Ful in combination with DMSO, 2 μM or 5 μM NR2F2 inhibitor Z021 for 7 days (n ≥ 2).

(E) Representative clonogenic proliferation assay results of MCF7 sgNT, sgARID1A, and sgPTEN cells treated with DMSO, 50nM Ful, 50nM Ful + 5 μM NR2F2 inhibitor Z021 for 7 days (n ≥ 2).

(F) Tumor growth of MCF7 sgNF1 xenografts treated with vehicle, 5mg Ful, 5mg Ful plus 3 or 6 mg/kg NR2F2 inhibitor Z021. Five mice in each group.

(G) Tumor volume changes of PDX models treated with vehicle, the indicated concentration of Ful and/or NR2F2 inhibitor Z021 at the shown dates compared to the initial volumes. Each dot represents a single mouse.

(H) Tumor growth of PDOX_64aS1-sgNT tumors treated with vehicle or 3 mg/mouse of Ful, and PDOX_64aS1-sgNF1 tumors treated with 3mg/mouse of Ful plus 6 mg/kg of NR2F2 inhibitor Z021. Five mice in each group.

All data were presented as mean ± SEM. P values were calculated using two-way ANOVA (A) or one-way ANOVA [ (B), (C), (G)] followed by Tukey’s test, or two-tailed multiple Mann Whitney test [(F) and (H).

An endogenous ligand of NR2F2 has not yet been identified, limiting strategies to block ligand production. However, prior reports have described features of the ligand binding pocket leading to the development of selective inhibitors that could downmodulate NR2F2 activity (37-39). We selected five compounds with high avidity for NR2F2 to assess for their antitumor effects and potential to restore endocrine responsiveness (39). The selected inhibitors, combined with fulvestrant, all exerted more potent antiproliferative effects in NR2F2 expressing cells than in NR2F2 KO or DKO cells, with the most potent inhibition in NF1 KO cells (Fig. S7B), implying their specificity. To confirm the binding of NR2F2 inhibitors to NR2F2, we performed thermal shift assays (TSA) using purified NR2F2 ligand binding domain (LBD) and the inhibitor Z021. We measured a melting temperature (Tm) of 59.88 ± 1.6 °C in the presence of saturating Z021, and a Tm of 57.70 ± 0.69 °C in the absence of Z021 (Fig. S7C). This approximately 2°C positive shift in the Tm further implies direct binding of Z021 to the NR2F2 LBD. Like the results of NR2F2 KO, drug combination studies demonstrated that NR2F2 inhibition by these inhibitors could effectively reverse NF1 KO induced-fulvestrant resistance (Fig. 7D and Fig. S7B, D). To assess the generalizability, we tested the NR2F2 inhibitors in other endocrine resistant contexts, including KRAS OE, ARID1A KO and PTEN KO. In each model of these genotypes, we found that NR2F2 inhibitors could partially or completely reverse the fulvestrant resistant phenotype (Fig. 7E, Fig. S7E-G).

To ascertain the in vivo relevance of these inhibitors, we first sought to establish whether continuous inhibition of NR2F2 might have any untoward effects on the health of murine models. After two weeks of daily Z021 (6 mg/kg) treatment, no overt pathologic changes in various organs (e.g., liver, kidney, heart, and spleen) were observed (Fig. S7H). We next assessed Z021’s antitumor effects with different doses in combination with fulvestrant in the NF1 loss models. NF1 KO MCF7 xenografts showed marginal tumor growth suppression with fulvestrant, but the addition of Z021 showed a dose dependent effect with 3mg/kg further inhibiting tumor growth and 6mg/kg caused complete growth suppression (Fig. 7F). Importantly, combinatorial treatment significantly suppressed the tumor growth in multiple endocrine resistant PDX models, including PDX_7aS1 (p<0.0001), PDX_50aS1 (p<0.01), PDX_WHIM18 (p<0.05) (Fig. 7G). Consistent with NR2F2 KO (Fig. 1G), NR2F2 inhibition restored fulvestrant sensitivity in PDOX_64aS1-sgNF1 tumors to similar rates as PDOX_64aS1-sgNT tumors (Fig. 7H). Together, these data reveal the ability of NR2F2 inhibition to restore the intrinsic hormone response in multiple ER+ breast cancer models.

Discussion

In this study, we identify the orphan nuclear receptor, NR2F2, to be a highly specific therapeutic target in HR+ breast cancer as its antagonism can restore the intrinsic estrogen sensitivity. The unique and essential roles for this receptor have implications for our understanding on the regulation of nuclear receptor function both in the mammary gland and in the treatment of patients with ER+ breast cancer.

The robust tumor promoting functions of ER have been broadly confirmed by the activity of antiestrogens in breast cancer as well as the widely prevalent finding of acquired, constitutively active mutations in ESR1 among patients (10, 11) treated with endocrine therapies. The additional finding of acquired mutations in genes involved in MAPK signaling among ESR1 WT but hormone-resistant cancers (12) was somewhat surprising in this context. While MAPK activating mutations are widely oncogenic, the mechanism(s) by which this pathway could suppress the intrinsic response to antiestrogens was not immediately apparent. To uncover the specific targets and programs essential for MAPK mediated resistance, we conducted CRISPR screens in WT and NF1 KO cells which yielded a nuclear receptor, NR2F2, as the top hit. We found that NR2F2 loss indeed impaired the growth and restored the endocrine response of NF1 KO cells, while its overexpression could induce endocrine resistance in multiple model systems. These findings pointed to a potential direct role for NR2F2 in the endocrine response pathway in breast cancer.

To further investigate how NR2F2 might mediate endocrine resistance, we examined the effects of NF1 and NR2F2 on the transcriptome. We found that NF1 KO and NR2F2 overexpression led to suppression of ER-regulated gene sets while knockout of both NF1 and NR2F2 restored them. In concert, we found ER chromatin binding and chromatin accessibility had the same pattern as attenuated by NF1 loss but restored by concomitant knockout of NR2F2. Taken together, the results point to NR2F2 potentially insulating key portions of the estrogen response pathway akin to the recent findings for other nuclear receptors PGR and AR that modulate ER genomic binding and transcription (40, 41). Moreover, our protein-protein interaction studies revealed the importance of NR2F2-associated coregulator complexes in ER action and chromatin accessibility. Several of the NR2F2 associated transcription coregulators, including GATA3 (42, 43), FOXA1 (15, 44-46), AP1 (47, 48), PBX1 (49, 50) and ARID1A (16, 18), have been previously implicated as direct modulators of ER and the response of mammary derived normal and cancer cells to hormonal manipulation. In line with these findings, we observed that NR2F2 depletion or inhibition could reverse endocrine resistance induced by ARID1A loss. Coupled with the findings that NR2F2 inhibition could restore endocrine response to NF1 KO cells, the data imply that a recurrent mechanism by which ER+ cancers develop resistance to hormone therapy is via amplification of the modulatory effects of NR2F2 upon ER and/or chromatin accessibility. Further studies on the interactions and functions of NR2F2 in mammary gland development may shed further light on the discrete mechanisms by which NR2F2 modifies physiologic ER function and chromatin accessibility.

From a therapeutic standpoint, our data point to NR2F2 inhibition as a strategy for restoring the endocrine sensitivity to ER+ breast cancers. This could offer a more selective approach to targeting MAPK-altered ER+ breast cancers than MEK/ERK inhibitors that have major toxicities due to the ubiquitous roles for this pathway in other cell types. Moreover, the data reveal that there are potential roles for NR2F2 inhibition in several other endocrine resistant contexts including loss of PTEN and ARID1A. Moreover, inhibition of NR2F2 could enhance the activity in tumors that are “sensitive” to antiestrogens, perhaps due to the expression of RTKs or RTK ligands, pointing to an underlying expression of the NR2F2 program and the broad applicability to this therapeutic strategy.

While NR2F2 is a nuclear receptor, its endogenous ligand has not been identified. Nonetheless, the analogy to other nuclear receptors has led to initial discovery efforts revealing the potential to selectively target NR2F2 (37-39). However, the true therapeutic potential for this approach appears to center on the mechanism by which NR2F2 is modulating the ER program rather than as a canonical oncogenic driver. In specifying and modulating the intrinsic endocrine response pathway, NR2F2 reveals itself to be a highly promising drug target for restoring the intrinsic endocrine sensitivity. Further efforts aimed at optimizing the current series of available compounds and evaluating their potencies in more in vivo models, such as a MIND model, may facilitate the translation of these results to the clinic.

This study identified a subset of genes and loci co-regulated by ER and NR2F2, however, NR2F2 clearly has additional regulatory sites and functions beyond its role in endocrine response in breast-derived cells. They physiological functions of NR2F2 are not fully understood, partly due to the absence of a defined ligand-stimulated program. Further studies are needed to evaluate the potential toxicities, therapeutic index, and broader relevance of NR2F2-regulated genes to breast cancer progression.

Materials and Methods

Study design

The objective of the study was to investigate the molecular mechanisms underlying NF1 loss induced endocrine resistance and identify druggable targets to overcome such resistance. To address this, we performed CRISPR-Cas9 knockout screens in ER+ cells, which identified NR2F2 as a key driver of NF1 loss growth advantage and endocrine resistance. Using the PI3K/AKT and MAPK pathway inhibitors, we revealed that NR2F2 was upregulated by the MAPK pathway activation. We validated the role of NR2F2 in mediating endocrine responses using both knockout and overexpression approaches across multiple ER+ cell line models (MCF7, T47D, MDA-MB316, CAMA1, ZR751, EFM19 and HCC712), MCF7-derived xenografts, and patient tumor derived models (PDX_7aS1, PDX_50aS1, PDX_WMIH18 and PDOX_64aS1). Functional assays, including competition cell growth assays, clonogenic assays, and CellTiter-Glo assays, were used to assess the impact of NR2F2 alterations. To elucidate the molecular mechanisms by which NR2F2 regulates endocrine responses, we employed multi-omics approaches, including RNA-seq, ChIP-seq, ChIP-ReChIP, ATAC-seq and qPLEX-RIME. We assessed the correlations of 1) the MAPK pathway activity/NF1 loss and NR2F2 expression and 2) the ER transcriptional activity and NR2F2 expression in patient samples from METABRIC data cohort. Publicly available ChIPseq data were used to confirm their potential chromatin co-binding with NR2F2. We expanded the effect of NR2F2 knockout in the context of other genomic alterations, such as KRAS overexpression, PTEN loss, and ARID1A loss, which are known drivers of endocrine resistance. We also explored whether pharmacological inhibition of NR2F2 has similar effect to restore endocrine sensitivity in multiple resistant models. Sample sizes were established based on previously published studies in the field, and all instances of replication described in this study refer to biological replicates. No data were excluded from the analyses. Detailed experiment designs and exact biological replicate numbers were described in the figure legends or the “Materials and Methods” section of the Supplementary Materials. Patient tumor tissues used for generating research models were collected with informed consent from patients under protocols approved by the MSKCC review board. All in vivo studies were performed in accordance with guidelines approved by the Memorial Sloan Kettering Cancer Center (MSKCC) Institutional Animal Care and Use Committee and Research Animal Resource Center. Mice used in in vivo experiments were randomized into control or treatment groups by different investigators.

Statistical Analysis

Individual data for all experiments with samples sizes less than 20 are provided in data files S1-S10. Statistical analyses were performed using GraphPad Prism 10 and RStudio. All tests were two-tailed, with a 95% confidence interval, and p-values < 0.05 were considered as statistically significant. Details regarding the numbers of biologically independent experiments, individual mice or patient samples, as well as the specific statistical tests used, are included in the respective figure legends or the Materials and Methods section of the Supplementary Materials.

Supplementary Material

Materials and Methods

Figs. S1 to S7

Tables S1 to S11

Data Files S1-S10

References (51-72)