Ligand-binding Domain–activating Mutations of ESR1 Rewire Cellular Metabolism of Breast Cancer Cells

About 75% of all breast cancers express estrogen receptor-α (ER), and inhibition of ER activity by endocrine treatments is an effective and safe treatment strategy for patients harboring ER⁺ breast cancer. However, some patients with metastatic breast cancer do not respond to any form of endocrine treatment (de novo resistance), and virtually all patients who initially respond eventually develop endocrine resistance (acquired resistance). We and others have identified mutations in the ER ligand-binding domain (LBD) that lead to a conformational change mimicking the activated ligand bound receptor and serve as a novel mechanism of resistance to endocrine therapy (1–5). The most common mutations of the ER are substitution of Asp-538 to glycine (D538G) and substitution of Tyr-537 to serine (Y537S; refs. 6–8). These and other activating mutations where identified in up to 39% of heavily pretreated patients with metastatic breast cancer (6, 7, 9).

Accumulating clinical and laboratory data suggest a unique phenotype of breast cancers expressing these mutations. Thus, LBD mutations were shown to enhance colony formation by breast cancer cells (10), we and others noted an association between the presence of D538G mutation and increased frequency of liver metastases (2), and analysis of the SoFea and PALOMA clinical trials, indicated shortened progression free-survival of patients with tumors harboring LBD mutations (6, 11). These data suggest that breast cancers harboring these LBD-activating mutations may constitute a new subgroup of breast cancers, characterized by increased invasiveness and tissue tropism, translating into a more aggressive clinical behavior.

A possible candidate mediating aggressive behavior of ER⁺ breast cancer cells is the PI3K-AKT-mTOR pathway (12). Activating mutations in this pathway occur in over 40% of ER⁺ breast cancers (6, 7, 9) and activation of this pathway has been linked to endocrine-resistant breast cancer (13, 14). Accordingly, inhibition of mTOR by everolimus is an effective strategy to overcome endocrine resistance in patients with metastatic breast cancer (14). Activation of the PI3K-AKT-mTOR pathway may be associated not only with endocrine resistance but also with a more aggressive disease behavior (13, 14). However, the interaction between the presence of LBD mutations and activation of the PI3K-AKT-mTOR has not been characterized yet.

One of the hallmarks of cancer is reprogramming of energy metabolism, which provides cancer cells appropriate conditions required for enhanced proliferation and survival. The ER pathway may play a role in rewiring breast cancer cells' metabolism. Thus, ER⁺ and ER⁻ breast cancer have different metabolic signatures, with ER⁺ tumors showing activation of fatty acid metabolism, transport, (15), increased glucose uptake, and glycolysis (16), as well as an increase in amino acid–associated metabolites (17). None the less, to our knowledge, the metabolic effects of activating LBD mutations have not been characterized yet.

In this study, we sought to characterize the oncogenic properties of breast cancer cells harboring LBD-activating mutations. Our results indicate aggressive behavior of these cells, as well as activation of PI3K-AKT-mTOR. Transcriptomic and metabolomic analyses revealed unique metabolic profile, distinct from that of activated ER cells, characterized by increased tricarboxylic acid (TCA) cycle and reliance on glutamine. Thus, the data indicate that LBD-activating mutations rewire breast cancer cell metabolism.

The ERE-luciferase reporter construct, kindly provided by D. Harris, (University of California, CA), consists of two repeats of the upstream region of the vitellogenin ERE promoter. Generation of pcDNA3 ER-WT and 538G-ER constructs was described previously (2). 537S-ER-pcw107-V5 and HcRed-pcw107-V5 (empty vector) were a gift from David Sabatini & Kris Wood (Addgene plasmid catalog nos. 64634 and 64647; ref. 18). WT-ER-pcw107-V5 was generated using the In-Fusion HD cloning kit according to the manufacturer's instructions, using 537S-ER-pcw107-V5 plasmid as a template for PCR. Primers used were as follows: 5′-primer: CTCTATGACCTGCTGCTGGAGATGCT; 3′-primer: CAGCAGGTCATAGAGGGGCACCACGTT. 538G-ER-pcw107-V5 was generated using the In-Fusion HD cloning kit as well, using WT-ER-pcw107-V5 plasmid as a template for PCR. Primers used were as follows: 5′-primer: CCTCTATGGCCTGCTGCTGGAGATGCTG; 3′-primer: AGCAGGCCATAGAGGGGCACCACGTTC. All subcloned constructs were sequenced. pCMV-VSV-G was a gift from Bob Weinberg (Addgene plasmid catalog no. 8454; ref. 19), psPAX2 was a gift from Didier Trono (Addgene plasmid catalog no. 12260).

Cell lines were originally obtained from the ATCC and authenticated with the DNA markers used by ATCC. MCF-7 and T47D cells were grown in DMEM containing 10% FBS. For particular experiments, cells were grown with DMEM without glutamine or glucose, and in these experiments glucose and/or glutamine (both from Life Technologies) were added separately. For all E2 (obtained from Sigma) and rapamycin (obtained from Cayman Chemical) studies, cells were cultured in phenol-free media using 10% charcoal-treated serum for 2 days before treatment. All transfections, except for lentiviral system, used jetPEI (Polyplus-transfection SA).

537S-ER-pcw107-V5, 538G-ER-pcw107-V5, WT-ER-pcw107-V5, or empty pcw107-V5 vectors were transfected along with pCMV-VSV-G and psPax2 (2μg: 2μg: 2μg) into HEK-293T cells using the calcium phosphate coprecipitation method. In brief, HEK-293T cells were seeded in 6-well plate and grown to 80% confluency. Plasmid DNA (6 μg) was mixed with 250 μL of 0.25 mol/L CaCl₂, incubated for 20 minutes at room temperature, and then mixed with 250 μL of sterile HBSx2 (NaCl 0.28 mol/L, HEPES 0.05 mol/L, and Na₂HPO₄ 1.5 mmol/L; pH = 7.05). The mixed solution was added to cells for 24 hours. Conditioned medium was harvested twice, 48 hours and 72 hours after transfection, and conditioned media containing viral particles was filtered through 0.45-μm filters. For virus infection, MCF7 or T47D cells were incubated with conditioned media containing virus particles supplemented with polybrene (8 μg/mL) for 8 hours. Stably infected cells were selected by puromycin (0.5 μg/mL). Two weeks later, single colonies were generated by seeding half-cell per well into 96-well plate, and genomic DNA of the clones was sequenced to confirm insertion of the mutation. Two clones of each wild-type (WT), 538G-ER, or 537S-ER were cultured routinely separately and the clones of each genotype were mixed prior to each experiment.

Viability and proliferation were assessed using the 3-(4, 5-dimethylthiazol-2-yl)-2, 5-diphenyltetrazolium bromide (MTT) assay as described previously (20). For the assay, cells (WT, 538G-ER, or 537S-ER–expressing cells) were plated in 96-well plate (3,000 cells/well). After 24 hours, 10 wells of each cell type (WT and mutant) served as day 0. Experiment plate was treated with E2 or control vehicle for 72 hours, MTT was added for 1.5 hours, medium was aspirated, and MTT was dissolved with 50 μL DMSO. Absorbance was determined at 570 nm using a multichannel plate spectrophotometer. Data were subsequently analyzed by comparing live cell counts on day 0 with live cell counts on day 3. Each condition was assessed from 15 replicates and three independent experiments were conducted.

Cells were plated in 96-well plates at a density of 3,000 cells per well, 15 wells per treatment. A day later, cells were treated with E2 or control vehicle and incubated at 37°C, 5% CO₂ for 24 hours. Glutaraldehyde (2.5%) was diluted 1:5 into cells for 10 minutes and then cells were washed three times with ddH₂O. Cells were incubated with 100 μL of methylene blue stain [1% methylene blue in borate buffer (pH 8.5)] for 1 hour at room temperature. After removing the methylene blue stain, cells were washed four times with dH₂O and 100μL of 0.1 mol/L HCl was added into each well. The absorbance was read with a microplate reader at 650 nm.

Migration was assessed using the wound-healing (“scratch”) assay as described previously (2). For the assay, MCF-7 cells were transfected with the indicated constructs and grown to confluency in a 6-well plate in phenol-free media with 10% charcoal-treated serum. In some experiments MCF-7 cells stably expressing WT or 537S-ER were grown to confluency. The cell monolayer was then scraped in a straight line with a 200-μL tip and photographed at indicated time.

The assay was conducted as described previously (21). Briefly, 2 × 10⁴ cells/well were seeded in noble agar (Sigma) at a final concentration of 0.3%, on top of 0.5% base agar in complete media with phenol red. Cells were refed twice a week with phenol-free media with 10% charcoal-treated serum. Triplicates were performed for each condition and cells were incubated for 2–4 weeks at 37°C in 5% CO₂ atmosphere. Colonies were then counted and photographed.

Cells were harvested, lysed, and the total protein was extracted with RIPA buffer (50 mmol/L Tris–HCl pH 7.4, 150 mmol/L NaCl, 1% NP-40, 0.25% Na-deoxycholate, 1 mmol/L EDTA, and 1 mmol/L NaF), together with a protease and phosphatase inhibitor cocktails (Sigma). Lysates were resolved on 10% SDS-PAGE and immunoblotted with the indicated antibodies. ERα (Santa Cruz Biotechnology, sc-8002), β-actin (Sigma, A5441), and anti-mTOR, anti-p-AKT, and anti-p70s6k (Cell Signaling Technology, catalog nos. 9234, 9271, and 9234).

Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) measurements were performed with a Seahorse XF 96 Analyzer from Seahorse Bioscience (Agilent) according to the manufacturer's instructions. Cells expressing WT, Y537S, or D538G mutation were plated at a density of 1 × 10⁴ cells per well, treated with ethanol (0.003%) as control or E2 (10 nmol/L), 10 replicates for each treatment. On the day of analysis, media was changed to Seahorse XF Base Medium supplemented with 1 mmol/L glutamine for the glycolysis stress test, or 1 mmol/L pyruvate, 2 mmol/L glutamine, and 10 mmol/L glucose for the mito stress test followed by incubation at 37°C in a non-CO₂ incubator for 1 hour.

Mitochondrial respiration (OCR) was measured by the XF-96 Extracellular Flux Analyzer using Cell Mito Stress Test Kit from Seahorse Biosciences under basal conditions followed by the sequential addition of oligomycin (2 μmol/L), FCCP (0.5 μmol/L), rotenone (0.5 μmol/L), and antimycin A (0.5 μmol/L). Glycolysis activity (ECAR) was measured using glycolysis test kit following sequential addition of glucose (10 μmol/L), oligomycin (1 μmol/L), and 2-deoxyglucose (2-DG; 50 μmol/L). After each injection, four time points were recorded with approximately 35 minutes between each injection. The OCR and ECAR were automatically recorded and calculated by the Seahorse XF-96 Software. OCR and ECAR were normalized to cell number. Each experiment was repeated at least three times, and a representative experiment is shown. The basal respiration, proton leak, and ATP production are calculated by the Seahorse software as described in the Seahorse Report Generator manual (https://www.agilent.com/cs/library/usermanuals/public/Report_Generator_User_Guide_Seahorse_XF_Cell_Mito_Stress_Test_Single_File.pdf).

MCF-7 cells were seeded in 6-well plates (2 × 10⁵ cells/well). Cells were transfected with either WT-ER or 538G-ER for 48 hours. The experiment was performed in full medium containing 10% FBS. Forty-eight hours after transfection, RNA was extracted using TRizol reagent according to the manufacturer's instructions. Gene expression analysis was conducted using Affymetrix GeneChip Human Gene 1.0 ST arrays. Using Partek Genomics Suite, a list of genes differentially expressed by 538G-ER compared with WT-ER was generated (>1.25 or <-1.25-fold change, P <0.05). Web-based applications and public databases (DAVID, WebGestalt, GeneAnalytics, and String; refs. 22–25) were used to functionally categorize the genes and their regulation by the ER pathway.

MCF-7 cells stably expressing WT-ER or 537S-ER were grown in triplicates with 7 mmol/L glucose and/or 8 mmol/L glutamine treated with E2 (10 nmol/L) for 24 hours. Total RNA was extracted using the High Pure RNA Isolation Kit (Roche). RNAseq and bioinformatics were conducted at the INCPM (The Mantoux Bioinformatics Institute of the Nancy and Stephen Grand Israel National Center for Personalized Medicine and Weizmann Institute of Science, Rehovot, Israel). For sequencing, briefly, 500 ng of total RNA was fragmented followed by reverse transcription and second strand cDNA synthesis. The double-strand cDNA was subjected to end repair, A-base addition, adapter ligation, and PCR amplification to create barcoded libraries. Libraries were evaluated by Qubit and TapeStation. Sequencing was conducted with NextSeq SR75 v2 (Illumina) at 75 cycles, single read kit. The output was approximately 21 million reads per sample.

Bioinformatics: Poly-A/T stretches and Illumina adapters were trimmed from the reads using cutadapt (26); resulting reads shorter than 30 bp were discarded. Reads were mapped to the Homo Sapiens GRCm38 reference genome using STAR (ref. 27; with EndToEnd option and outFilterMismatchNoverLmax = 0.04), supplied with gene annotations downloaded from Ensembl release 88. Expression levels for each gene were quantified using HTSeq-count (28), using the Ensembl annotations. Differentially expressed genes were identified using limma-voom (29). Pipeline was run using snakemake (30). Pathway and function enrichment and heatmaps of genes associated with specific pathways were generated.

Chromatin immunoprecipitation (ChIP) assay was carried out with the ChIP Assay Kit (Millipore Temecula) as described elsewhere (31) using anti-ERα antibody (Millipore Temecula). ER-bound DNA was amplified by qRT-PCR with primers spanning AP1 motif on the promoters of SNAI, SNAI2, and MMP1, or spanning ERE site at GREB1 promoter (32):

Two days after transfection with the various constructs, total RNA was prepared using the High Pure RNA Isolation Kit (Roche). Total RNA (1 μg) was reverse transcribed using qScript cDNA Synthesis Kit (Quanta Biosciences). qRT-PCR was used to determine mRNA level. Primers were designed using Primer Express (Applied Biosystems) and synthesized by Integrated DNA Technologies and are listed below. Amplification reactions were performed with Platinum qPCR SuperMix in triplicate using StepOne Plus (Applied Biosystems). PCR conditions: 50°C for 2 minutes, 95°C for 2 minutes, followed by 40 cycles of 95°C for 15 seconds, 60°C for 45 seconds.

The following primers were used:

WT-ER- and 537S-ER–expressing cells (in quadruplicates) were grown in phenol-free media with 10% charcoal-treated serum and treated with E2 (10 nmol/L) for 24 hours. Frozen cells (100 mg) were submitted to Metabolon, Inc for sample extraction and analysis. In brief, samples were prepared using the automated MicroLab STAR System from Hamilton Company. Several recovery standards were added prior to the first step in the extraction process for QC purposes. To remove protein and to recover chemically diverse metabolites, proteins were precipitated with methanol under vigorous shaking for 2 minutes followed by centrifugation. The resulting extract was divided into five fractions: two for analysis by two separate reverse-phase (RP)/UPLC-MS/MS methods with positive ion mode electrospray ionization (ESI), one for analysis by RP/UPLC-MS/MS with negative ion mode ESI, one for analysis by HILIC/UPLC-MS/MS with negative ion mode ESI, and one sample was reserved for backup. Samples were placed briefly on a TurboVap (Zymark) to remove the organic solvent. The sample extracts were stored overnight under nitrogen before preparation for analysis. The sample extract was dried, and then reconstituted in solvents compatible to each of the four methods. Each reconstitution solvent contained a series of standards at fixed concentrations to ensure injection and chromatographic consistency. One aliquot was analyzed using acidic positive ion conditions, chromatographically optimized for more hydrophilic compounds while another aliquot was optimized for more hydrophobic compounds. Another aliquot was analyzed using basic negative ion optimized conditions using a separate dedicated C18 column. The basic extracts were gradient eluted from the column using methanol and water. The fourth aliquot was analyzed via negative ionization following elution from a HILIC column (Waters UPLC BEH Amide 2.1 × 150 mm, 1.7 μm) using a gradient consisting of water and acetonitrile with 10 mmol/L ammonium formate. Mass spectrometry (MS) analysis alternated between MS and data-dependent MSn scans using dynamic exclusion. The scan range varied slighted between methods but covered 70–1000 m/z. Statistical analysis of log-transformed data was conducted using “R” (http://cran.r-project.org/) or JMP.

Mice maintenance and experiments were carried out under institutional guidelines of the Sourasky Medical Center (Tel Aviv, Israel) in accordance with current regulations and standards of the institution Animal Care and Use Committee. We used an orthotopic model to test the tumorigenic properties of MCF-7 cells stably expressing WT-ER, 537S-ER, or 538G-ER. Female athymic nude mice (Balb/c background), 4–6 weeks of age were purchased from Envigo RMS. The mice were housed and maintained in laminar flow cabinets under specific pathogen-free conditions. Tumors were induced by injecting 0.5 × 10⁶ cells/100 μL DMEM with 5% FCS into mammary fat pad, 7 mice per cell line. Local tumors were measured twice a week using a caliper. Tumor volume was evaluated by the ellipsoid volume calculation formula 0.5 × (length × width²).

Results are presented as mean ± SD. Tumor volume graph is presented as mean ± SEM. Continuous variables were compared using t test. All significance tests were two-tailed and a P < 0.05 was considered as statistically significant. In the Metabolom study Welch two-sample t test was used. The random forests were created using a cross-validation method where each tree is created leaving out a subset of samples. To validate the tree, each sample receives votes for group placement for each tree that was excluded from the creation of. On the basis of those votes, each sample is “predicted” to be in one of the groups. The predictive accuracy is based on the comparison of these predictions from the validation with the actual groups that the samples are in.

We have previously shown, using overexpression studies, that 538G-ER enhances proliferation and migration of breast cancer cells (2). Here we aimed to test the effects of mutated ER expressed at physiologic levels on breast cancer cells. For this purpose, we generated MCF-7 cells stably expressing either WT-ER, 537S-ER, or 538G-ER, under PGK promoter (18), allowing expression of ER at near physiologic levels at protein level and mRNA level (Supplementary Fig. S1A and S1B). Several single-cell clones were generated for each ER and two with typical activity (Supplementary Fig. S1C and S1D), were verified for exogenous ER expression, were selected, grown separately, and mixed prior to each experiment.

The viability of 537S-ER, 538G-ER, and of WT-ER MCF-7 was assessed using MTT (Fig. 1A) and methylene blue assays (Fig. 1B). For MTT, cells were grown in full serum and medium containing phenol red for 48 or 72 hours. Both 537S-ER- and 538G-ER–expressing cells exhibited significantly enhanced proliferation compared with WT-ER–expressing cells (P < 0.05 for all comparisons; Fig. 1A). For methylene blue assay, cells were grown in estrogen-depleted medium and treated with either a vehicle or E2 for 72 hours. Both 537S-ER- and 538G-ER–expressing cells exhibited significantly enhanced proliferation, either with or without E2 treatment (P < 0.05 for all comparisons; Fig. 1B).

Anchorage-independent growth is a hallmark of the neoplastic phenotype. To assess the effects of mutated ER expression on anchorage-independent growth, growth of these cells in soft agar was examined in the presence or absence of E2. In the absence of E2, WT-ER cells did not form colonies whereas 537S-ER cells successfully grow under these conditions (P < 0.05; Fig. 1C). Upon addition of E2, expression of 537S-ER was associated with a 300% increase of colony formation and growth compared with WT-ER (P < 0.05; Fig. 1C). Similar results were observed with 538G-ER cells (**, P < 0.01; ***, P < 0.005; Fig. 1D).

Next we aimed to determine the tumorigenicity profile of mutated ER–expressing cells compared with WT-ER cells, to form tumors in nude mice. Nude mice were inoculated into the mammary fat pad with WT-ER, 537S-ER, or 538G-ER MCF-7 cells, 7 mice per group. Tumor volume was measured twice a week (Fig. 1E). Twenty days postinoculation, the tumors formed by the 538G mutant were 3-fold larger than those formed by the WT (P < 0.05; Fig. 1E), while the 537S were 5-fold larger than those formed by the WT (P < 0.05; Fig. 1E).

As depicted above (Fig. 1) the mutated ERs confer an aggressive phenotype to breast cancer cells beyond that of activated WT-ER. We therefore hypothesized that the LBD-mutated ER may show differential regulation of gene expression compared with WT-expressing cells, even in the presence of estrogen. To test this, we performed an mRNA expression screen of MCF-7 cells transfected with either 538G-ER or WT-ER and grown in conditions that WT-ER is also transcriptionally active, that is, full serum (without charcoal stripping) and medium containing phenol red. Hierarchical clustering showed 72 differentially expressed genes in 538G-ER cells compared with WT-ER genes (1.25-fold, P < 0.05), of them 59 were upregulated in the mutant compared with WT (Fig. 2A) and a list of most upregulated genes associated with aggressiveness is detailed (Table 2B). Classic ERE-regulated transcripts (e.g., GREB1 and PR) were not upregulated in this screen, indicating that the differentially regulated genes are not associated with classic ER-affected genes but with a unique effect of 538G-ER. Function analysis of these genes using GeneAnalytics revealed a significant enrichment of ECM remodeling, ECM degradation, and matrix metalloproteinases pathways (Fig. 2C; Supplementary Fig. S2A). Importantly, we observed also signaling pathways associated with tumorigenesis like G-protein Ras family GTPases and TGFβ. Most of the array results were validated using qRT-PCR (Fig. 2F and G) and validated genes included genes known to be associated with invasion and metastases (e.g., SERPIN9, SNAI1, SNAI2, MMP1, and MMP13), as well as tumor metabolism (e.g., PDK4 and ACSS1). The expression of these genes was also studied in T47D, expressing either 538G-ER or 537S-ER or WT-ER. While similar pattern was observed, a differential expression of certain genes was noted. For example, 537S-ER induced a nearly 30-fold increased expression of MGP in MCF-7 cells, whereas in T47D its effect was much milder.

We next conducted transcriptomic analysis and compared MCF-7 cells expressing 537S-ER without E2 treatment with WT-ER treated with E2. As noted for 538G-ER–expressing cells, pathway enrichment analysis of untreated 537S-ER cells revealed an increase in genes related to extracellular matrix degradation and to ERK and AKT pathways, even when compared with WT-ER–treated cells (score 31.86, P < 0.0001; Fig. 2D; Supplementary Fig. S2B). This analysis demonstrated a clear separation between E2-regulated genes and genes that are uniquely regulated by 537S-ER activity (Fig. 2E).

To decipher whether the mutated ERs bind directly to the promotors of the genes that were upregulated in the mRNA expression screen, we performed a ChIP assay and assessed binding of the ER to the promoters of SNAI1, SNAI2, MMP1, or GREB1 as a control for ER binding. The ER can bind apart of ERE other motifs like FOXA1 and AP1 (33). As no ERE- or FOXA1-binding sites are present on these promoters (except for GREB1) we assessed binding to AP1 sites. ChIP assay indicated that while all ERs bind to the GREB1 promoter, neither the WT-ER nor the mutated ERs bind directly to SNAI1, SNAI2, or MMP1 promoters (Fig. 2H).

One of the hallmarks of cancer is the unique metabolism of cancer cells (34). As cells expressing the LBD-activating mutations show a more aggressive phenotype, we hypothesized that they will also show different metabolic activity. To test this we first analyzed activity of the PI3K-AKT-mTOR pathway, a major regulator of breast cancer cell metabolism (13). In accordance with a recent observation (10), we noted activation of the PI3K-AKT-mTOR pathway in MCF-7 cells stably expressing 537S-ER or transiently expressing 538G-ER and noted enhanced phosphorylation of AKT, mTOR and its down-stream target S6K (Supplementary Fig. S4A). We aimed to elucidate whether this pathway underlies the increased activity of the mutants. Gene expression analysis of CST1 (gene activated only by mutated ER, Fig. 2) and migration assay showed that mTORC1 inhibition using rapamycin only partially inhibited these activities (Supplementary Fig. S4B and S4C).

To further characterize the effect of 537S-ER on the metabolic activity of breast cancer cells, we conducted a metabolomic profiling of MCF-7 cells expressing either WT-ER or 537S-ER (described above, Fig. 1). For the analysis, cells were treated with either a vehicle control or E2, harvested 24 hours later, and a metabolomic screen was conducted as described under “Material and Methods”. Treatment of WT-ER cells with E2 yielded 209 metabolite showing significant change (P < 0.05, either up or downregulation), compared with untreated cells (E2-dependent metabolites). However, comparisons of differentially regulated metabolites in 537S-ER–expressing cells compared with untreated or E2-treated WT-ER–expressing cells revealed 370 and 225 metabolites, respectively. This indicates additional activity gained by cells expressing the mutated ER. Treatment of 537S-ER cells with E2 did not lead to major metabolic changes (Fig. 3A).

To further reveal a metabolic profile of 537S-ER cells, which is distinct from that of activated WT-ER, the identity of the different metabolites shared by each of the above groups was analyzed (Fig. 3B). Treatment of WT-ER with E2 resulted in a decrease of 183 metabolite, most of them (150/183, 82%) also observed in WT-ER versus 537S-ER. However, these metabolites comprise only a minority (150/370, 47%) in WT-ER versus 537S-ER comparison. Likewise, the comparison between 537S-ER and WT+E2 yielded a decrease in 164 metabolite, most of them (136/164, 83%) observed in WT-ER versus 537S-ER. But, again, these metabolites comprise only a minority (178/370, 42%) in WT-ER versus 537S-ER comparison. Importantly, only 65 metabolites were downregulated in both groups (WT vs. WT+E2 and 537 vs. WT+E2, 35% and 39%, respectively).

Random Forest classification using named metabolites of untreated WT-ER–expressing cells versus E2 treated gave a predictive accuracy of 100% (Fig. 3C). The top-ranking metabolites and pathways identified were energy, plasma membrane lipids, and amino acid metabolites. These are all considered to be E2-dependent metabolites. Classification of untreated WT-ER versus 537S-ER–expressing cells gave a predictive accuracy of 100% (Fig. 3D). The top-ranking metabolites and pathways identified were amino acid metabolites and plasma membrane associated lipids. This list is expected to include both E2-dependent and -independent metabolites. Classification of E2-treated WT-ER–expressing cells compared with 537S-ER–expressing cells gave a predictive accuracy of 87.5% (Fig. 3E). The top-ranking metabolites and pathways in this comparison include plasma membrane lipids, amino acids metabolism, and carbohydrate metabolism and these are expected to be E2 independent. Data generated from the metabolomics screen was used for a metabolic pathway enrichment analysis. The major pathways differentially affected in 537S-ER–expressing cells were glycolysis and TCA cycle (Fig. 4). Main glycolytic intermediates were found to be reduced by more than 50% in 537S-ER versus WT-ER–expressing cells (Fig. 4A, both E2-treated, P < 0.05). On the other hand, expression of 537S-ER led to an increase in the first half of TCA cycle metabolites (i.e., citrate, isocitrate, aconitate, and α-ketoglutarate) and a decrease in the second half (i.e. succinate, fumarate, and malate; Fig. 4B).

Direct analysis of ECAR, indicating glycolytic rate, and OCR, indicating mitochondrial activity, of MCF-7 cells expressing either WT-ER, 537S-ER, or 538G-ER was conducted using Seahorse technology and revealed similar glycolytic rate for both cells (Fig. 5A), but enhanced mitochondrial respiration of 537S-ER–expressing cells compared with WT-ER cells (P < 0.01; Fig. 5B). Similar results were obtained in 538G-ER–expressing cells (Fig. 5C and D). These data indicate E2-independent increased metabolic activity of MCF-7 cells expressing LBD-activating mutations. Similar results were obtained in 537S-ER T47D cells (Supplementary Fig. S5). Interestingly, the mitochondrial activity of 538G-ER T47D cells was not elevated (Supplementary Fig. S5). Similarly, it was recently published that the IGF-1R pathway is more responsive in 537S-ER T47D cells, whereas T47D-expressing 538G-ER cells did not exhibit a similar activity (35).

We assessed the effect of rapamycin on cellular metabolism. Interestingly, the results indicated that addition of rapamycin increased the mitochondrial activity of all the cells, whereas the glycolytic rate was slightly lower upon rapamycin addition, although did not reach statistical significance (Supplementary Fig. S4D and S4E).

The TCA cycle can be fed by either pyruvate derived from glucose and glycolysis or alternative sources, among them glutamine (36). Thus, glutamine can undergo glutaminolysis by glutamine synthetase to yield glutamate, which can be further deaminated into α-ketoglutarate and enter the TCA cycle. Glutamine and glutamate metabolism correlates with the expression of EMT-associated transcription factors (37) and glutaminolysis may serve as a driver of invasiveness of breast cancer cells (38). To analyze the contribution of glutamine to the aggressive behavior of 537S-ER–expressing cells, MCF-7 cells expressing either 537S-ER or WT-ER, were grown in the presence of E2 under glucose deprivation conditions (described under “Material and Methods”). Transcriptomic analysis by RNAseq revealed enrichment in pathways related to extracellular matrix degradation and ER stress in 537S-ER compared with E2-treated WT-ER cells (score 21, P < 0.05; Fig. 6A; Supplementary Fig. S6). A heatmap of the relevant ECM degradation–related genes was generated, and revealed a striking difference between WT and 537S-ER cells (two distinct sample clusters). Importantly, it revealed two gene clusters (MMP1, LOX) and (ITGA5, LRP4) in which their expression is upregulated specifically when 537S-ER cells were grown with glutamine as a sole carbon source (Fig. 6B). Interestingly, classic ERE-regulated genes were upregulated by both E2-treated WT and 537S-ER cells mostly when incubated with glucose and not glutamine (Fig. 6B). Of interest, 537S-ER cells grown with glucose form a separate cluster with highest increase in these genes. This was validated by direct measurement of ECM degradation–associated gene MMP1 and of classic ERE-regulated gene progesterone receptor (PR) under glucose and/or glutamine deprivation (Fig. 6C). Expression of PR was increased approximately 20-fold in WT cells treated with E2 and approximately 45-fold in 537S-ER, under exposure to glucose. Yet, expression of MMP1, increased only in 537S-ER and in the presence of glutamine. Similar results were obtained in 538G-ER-expressing cells (Fig. 6D). We aimed to reveal whether the aggressive-related gene expression is reflected also in the aggressive behavior of the cells. Cell migration experiments showed that while WT-ER–expressing cells required glucose for migration, 537S-ER–expressing cells were able to migrate in the presence of either glucose or glutamine (Fig. 6E). Thus, these results suggest that 537S-ER cells, in opposite to WT-ER cells, can use glutamine as a carbon source to promote their aggressive phenotype (summarized in Fig. 6F).

The detection of LBD-activating mutations of ESR1 was shortly followed by clinical observations suggesting an association between these mutations and adverse outcome (6). These observations raised the question whether the mutated ER is simply an activated ER or does it harbor a gain-of-function phenotype. While previous studies focused mostly on ligand-independent activity of the mutated receptor and examined its activity under estrogen-deprivation conditions (10, 39), we studied its activity in the presence of estrogen, under conditions better mimicking its activity in patients with breast cancer. We show here that harboring LBD-activating mutations leads to metabolic activity associated with aggressiveness and is associated with higher tumorigenic characteristics of breast cancers cells in vitro and in vivo.

We initially noted activation of the mTORC1 in cells expressing the mutated ER. As the mTOR pathway is a major regulator of cell metabolism, we hypothesized that the metabolic activity of these cells will be different from cells expressing WT-ER. The results we present suggest that mTORC1 pathway activation contributes to the enhanced activity of mutated ERs, although other factors play an important role in their enhanced aggressive phenotype. Several studies described metabolic alterations in breast cancer cells. Yet, the metabolic landscape of ER⁺ breast cancer cells and the metabolic effects of estrogen have not been elucidated. While a single study noted increased glucose uptake and glycolysis in MCF-7 cells following E2 treatment (18) a metabolic screen of relatively small number of metabolites could not verify changes in energy metabolites (19). A metabolomic profiling of tumors noted differences between ER⁺ and ER⁻ tumors but did not evaluate the activity of estrogen (40). For this study we used a metabolomic profiling platform that has been thoroughly validated (41). The metabolic profiling enabled a clear classification and separation between untreated WT-ER cells, E2-treated WT-ER cells, and 537S-ER cells.

This comprehensive analysis shed new light on the metabolic effect E2 imposes on WT-ER–expressing cells and suggest a role for E2 in regulating energy, plasma membrane lipids, and amino acid metabolites. Possibly, analysis of specific metabolites (detailed in Fig. 3C) can serve as a novel tool for the evaluation of activity of the ER pathway and aid in the classification of breast cancers.

The metabolic profiling enabled also a clear classification and separation between E2-activated WT cells and 537S-ER–expressing cells. Our data show significant changes in the levels of more metabolites in 537S-ER cells compared with WT (370 vs. 209), and most of the differentially regulated metabolites were E2-independent. Thus, the unique additional activity of LBD mutations directly affects metabolic activity. Analysis of the top ranking metabolites showing differential levels indicated plasma membrane lipids, amino acid metabolites, and carbohydrates as the most common metabolic pathways affected by 537S-ER (Fig. 3).These findings further support the transcriptomic data regarding the additive effect of the LBD-activating mutations, even compared with E2 treatment.

As reported previously (40), we noted, by both, Seahorse technology and metabolic profiling, increased glycolysis following E2 treatment. Yet, we did not observe increased glycolysis in 537S-ER cells. On the other hand, we observed increased mitochondrial activity in 537S-ER cells. Thus, analysis of these cells indicated marked increase in levels of metabolites associated with the first half of the TCA cycle (citrate to α-ketoglutarate), along with an increase in mitochondrial activities as evidenced by Seahorse analysis and increased ATP levels. Interestingly, 537S-ER led to a decrease in the levels of metabolites associated with second half of TCA cycle (succinate–malate). Several reasons may lead to this phenomenon. It is possible that these TCA cycle intermediates are shunted to other pathways, including amino acid or nucleotide biosynthetic pathways, thus leading to a decrease in their levels. Alternatively, pleurosis takes place in cancer cells when there is a shortage in certain TCA cycle intermediates and glutamine is an important participant in these reactions (36). It is possible that cells expressing 537S-ER engage glutamine in other pathways, thus not allowing it to participate in the pleurosis reactions. The mechanism underlying this needs to be further studied, using labeled glucose and glutamine.

537S-ER and WT-ER cells showed differential dependency on glucose and glutamine. Thus, while WT-ER cells rely on glucose and cannot utilize glutamine as an alternative carbon source, 537S-ER cells may use either glucose or glutamine interchangeably (Fig. 6D). This evidently yields an advantage to cells exposed to different environmental conditions. Our study further implies that the exposure to glutamine in the absence of glucose can, by itself, mediate aggressive phenotype acquired by breast cancer cells expressing 537S-ER. Thus, the transcription of MMP1, a gene strictly regulated by 537S-ER expression and not WT-ER, was observed under glutamine exposure and glucose starvation, while expression of PR, a gene classically regulated by E2-stimulated ER, depends mainly on glucose. The aggressive behavior which is supported by glutamine was exemplified also by migration assay. Thus, while WT-ER–expressing cells require glucose to migrate, 537S-ER–expressing cells migrate under either glucose or glutamine deprivation, it is not clear what role glucose or glutamine play in cell proliferation. This is the subject of our next studies. The ability of cells to thrive under glucose deprivation bestows a selective advantage and also ability to grow under specific microenvironment condition where glucose is scarce. Glutamine dependency was demonstrated in colorectal cancer cells were PIK3CA mutations were shown to reprogram glutamine metabolism. Thus, expression of mutant PIK3CA led to substantially more conversion of glutamine to a-ketoglutarate in order to replenish the TCA cycle to generate more ATP. This resulted in a more aggressive behavior as evidenced in higher proliferation rate and increased tumor growth (42). A correlation between glutamine dependency and increased aggressiveness was also seen in ovarian cancer, where it was shown that glutamine is required to promote cell proliferation (43). In addition, it was demonstrated that certain ovarian cancer cell lines are glutamine-dependent while others are glucose dependent, and importantly, glutamine dependency correlated with migration and invasiveness (44).

Interestingly, most of the genes uniquely regulated by the mutated ERs do not contain an ERE motif in their promoter region. Indeed, the ChIP assay results also confirm that the mutated ERs do not directly bind these genes' promoters. These results are in agreement with a recently published study that revealed a unique transcriptional activity of mutated ER, and this activity is mediated by ERE regions (33).

Taken together, this study shows for the first time that LBD mutations possess a more aggressive phenotype, beyond that of activated WT-ER, and this phenotype is associated with a unique gene signature and metabolic pattern. Revealing these unique properties may lead to the discovery of novel therapeutic targets that can be exploited to develop specific inhibitors for tumors expressing LBD mutations.

I. Wolf reports receiving speakers bureau honoraria and is a consultant/advisory board member for Roche. No potential conflicts of interest were disclosed by the other authors.

Conception and design: L. Zinger, M. Pasmanik-Chor, T. Rubinek, I. Wolf

Development of methodology: L. Zinger, T. Rubinek, I. Wolf

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Zinger, K. Merenbakh-Lamin, A. Klein, S. Journo, T. Boldes, T. Rubinek, I. Wolf

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Zinger, K. Merenbakh-Lamin, A. Klein, A. Elazar, T. Boldes, M. Pasmanik-Chor, A. Spitzer, T. Rubinek, I. Wolf

Writing, review, and/or revision of the manuscript: L. Zinger, M. Pasmanik-Chor, T. Rubinek, I. Wolf

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Zinger, T. Rubinek, I. Wolf

Study supervision: T. Rubinek, I. Wolf

This work was financially supported by the Israel Science Foundation to I. Wolf (grant nos. 1320/14 and 2385/15); the Israel Cancer Association to I. Wolf (grant no. 20160053); The Parasol Center for Women's Cancer Research, The Parasol Foundation; Djerassi-Elias Oncology Institute, CBRC, Tel Aviv University, Tel Aviv, Israel to I. Wolf; The Margaret Stultz foundation, the Sackler Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel to I. Wolf.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.