Unraveling Vulnerabilities in Endocrine Therapy-Resistant HER2+/ER+ Breast Cancer

Abstract

Breast tumors overexpressing human epidermal growth factor receptor (HER2) confer intrinsic resistance to endocrine therapy (ET), and patients with HER2/estrogen receptor–positive (HER2+/ER+) breast cancer (BCa) are less responsive to ET than HER2–/ER+. However, real-world evidence reveals that a large subset of patients with HER2+/ER+ receive ET as monotherapy, positioning this treatment pattern as a clinical challenge. In the present study, we developed and characterized 2 in vitro models of ET-resistant (ETR) HER2+/ER+ BCa to identify possible therapeutic vulnerabilities. To mimic ETR to aromatase inhibitors (AIs), we developed 2 long-term estrogen deprivation (LTED) cell lines from BT-474 (BT474) and MDA-MB-361 (MM361). Growth assays, PAM50 subtyping, and genomic and transcriptomic analyses, followed by validation and functional studies, were used to identify targetable differences between ET-responsive parental and ETR-LTED HER2+/ER+ cells. Compared to their parental cells, MM361 LTEDs grew faster, lost ER, and increased HER2 expression, whereas BT474 LTEDs grew slower and maintained ER and HER2 expression. Both LTED variants had reduced responsiveness to fulvestrant. Whole-genome sequencing of aggressive MM361 LTEDs identified mutations in genes encoding transcription factors and chromatin modifiers. Single-cell RNA sequencing demonstrated a shift towards non-luminal phenotypes, and revealed metabolic remodeling of MM361 LTEDs, with upregulated lipid metabolism and ferroptosis-associated antioxidant genes, including GPX4. Combining a GPX4 inhibitor with anti-HER2 agents induced significant cell death in both MM361 and BT474 LTEDs. The BT474 and MM361 AI-resistant models capture distinct phenotypes of HER2+/ER+ BCa and identify altered lipid metabolism and ferroptosis remodeling as vulnerabilities of this type of ETR BCa.

Globally, breast cancer (BCa) remains the leading cause of cancer-related mortality in women (1). Human epidermal growth factor receptor positive/hormone receptor positive (HER2+/HR+) tumors account for approximately 10% of all BCa cases and 60% of HER2+ tumors (2). HER2+/estrogen receptor positive (ER+) and HER2+/ER− BCa have distinct biology and are proposed as 2 distinct disease subtypes (3, 4). Prior and recent studies reported favorable prognosis and overall survival (OS) as well as lower-grade tumors in HER2+/ER+ BCa than in patients with HER2+/ER− BCa (5-8). However, HER2+/ER+ BCa has low response rates to anti-HER2 therapy compared with HER2+/ER− tumors (9, 10), has a limited antiproliferative response to endocrine therapy (ET), and is at a higher risk of recurrence (2, 10, 11).

Concurrent blockade of both HER2 and ER pathways in patients with advanced/metastatic HER2+/ER+ BCa is becoming increasingly evident as an effective treatment strategy. In clinical trials, postmenopausal women receiving combinations of ET plus anti-HER2 as first-line therapy showed improvements in progression-free survival and/or other clinical benefits (12-14). Likewise, a retrospective study of the national cancer database showed the highest 5-year OS among those receiving this combination therapy vs ET alone, chemotherapy alone or anti-HER2+ chemotherapy (15). Unfortunately, outside of clinical trials a large subset (36.7–60%) of patients with HER2+/ER+ receive hormonal treatment as a monotherapy (15, 16). Although current guidelines recommend anti-HER2 agents (trastuzumab + pertuzumab; TP) plus chemotherapy as first-line standard of care for advanced HER2+ disease irrespective of HR status, endocrine monotherapy remains an option for selected patients with HER2+/ER+ who have low disease burden or intolerance to chemotherapy (17, 18). Moreover, antibody-based anti-HER2 therapy is mainly given to patients with primary HER2+/ER+ BCa for 1 year, while ET is usually given longer (at least 5 years) (2). Thus, residual HER2+/ER+ disease after surgery, not subjected to anti-HER2 therapy, remains at high risk of incomplete response or resistance to prolonged ET.

A shorter time to recurrence in patients with HER2+ disease, compared with HER2–, treated with single-agent ET as tamoxifen or anastrozole was reported (19). Likewise, compelling evidence from preclinical and clinical data strongly suggests that HER2 overexpression confers intrinsic resistance to ET and that HER2+/ER+ BCa is less responsive to ET than HER2-/ER+ tumors (2, 10, 20). Multiple studies have further established that bidirectional crosstalk between ER and HER2 signaling pathways mediates resistance to ET (21-23). While acquired ET resistance is extensively studied in HER2–/ER+ BCa, preclinical models of HER2+/ER+ BCa that have acquired resistance to ET are lacking. Existing models of resistance have been generated by genetic manipulations of ER and/or HER2, for instance (24, 25).

Collectively, there is an unmet need to develop, characterize, and study therapeutic responses of ET-resistant (ETR) HER2+/ER+ models to HER2- and other ER-directed therapies. In this study, we establish ETR variants through long-term estrogen deprivation (LTED) to mimic resistance to aromatase inhibitors (AIs) from 2 HER2+/ER+ BCa cell lines (BT-474 and MDA-MB-361) (26, 27). To identify therapeutic vulnerabilities between AI-responsive parentals and ETR-LTEDs of HER2+/ER+ BCa cell lines, we performed growth assays, PAM50 molecular subtyping, and genomic and transcriptomic analyses. Our data show that the BT-474 and MDA-MB-361 ETR models capture distinct phenotypes of HER2+/ER+ BCa and highlight altered lipid metabolism and ferroptosis remodeling as features of this ETR BCa.

Materials and Methods

Treatments, Cell Lines, and Cell Culture

Trastuzumab and pertuzumab were obtained from Genentech, SYTOX green (S7020) and Hoechst 33342 (H1399) from ThermoFisher, fulvestrant (ICI) (S1191) from Selleck Chemicals, 17β-estradiol (E8875) from Sigma Aldrich, and RSL3 (HY-100218A) from MedChemExpress.

MDA-MB-361 (MM361; RRID:CVCL_0620) and BT-474 (BT474; RRID:CVCL_0179) human breast carcinoma cell lines were obtained from the Tissue Culture and Biobanking Shared Resource at Lombardi Comprehensive Cancer Center and were routinely checked for Mycoplasma contamination. Both cell lines were cultured in improved minimum essential medium (IMEM; Gibco) supplemented with 10% fetal bovine serum (FBS) at 37 °C in a humidified atmosphere containing 5% CO₂. We generated 2 LTED variants A and B cells from their corresponding parental cell lines by chronically passaging parental cell lines in phenol red-free IMEM (Gibco) supplemented with 10% charcoal-stripped bovine serum (CSS; Vita Scientific) for over 6 months. The LTED A and B variants for MM361 and BT474 were generated from 2 different flasks of the same passage of their corresponding parental cell line. After 6 months in these conditions, the derived cells were used and continuously cultivated in 10% CSS phenol red-free IMEM.

Growth Assays

Differences in growth kinetics between parental and corresponding LTEDs were evaluated by the trypan blue exclusion assay. Cells were seeded at a density of 100 000 cells/well in 24-well plates. After trypsinization and staining with trypan blue, live cell counts using the Countess II Automated Cell Counter (ThermoFisher Scientific) were recorded on days 0 (24 hours), 2, 4, 6, 8, and 10 from cell seed.

For crystal violet assays, parental and LTED cells were seeded in 96-well plates at 5000 cells/well. Forty-eight hours later, cells were subjected to indicated treatments for 6 additional days. Treatments were replenished 3 days after treatment onset. At the end of the experiments, cells were stained with 0.5% crystal violet in 25% methanol. Once plates were dried, the stain was resolubilized with citrate buffer, and absorbance measurements were obtained from the ELx808 plate reader (BioTek).

Cell Death Quantification

Cells were seeded in 24-well plates at a cell density of 150 000 cells/well, incubated overnight, then subjected to treatments as indicated. After 72 hours of treatments, cells were stained with both Hoechst 33342 (0.1 μg/mL) to monitor total cell number, and Sytox green (5 μM) to monitor dead cells. Images were acquired using the Olympus IX71 microscope and subsequently analyzed by Image J. Percentage cell death was calculated as Sytox green cell number over total cell number.

For drug interaction analyses in Sytox assays, the Relative Index (RI) method (28) was used. Percentage viability values (100 - % cell death) relative to vehicle for each treatment group were obtained and used for RI calculations using the equation [TP alone * RSL3 alone]/[TP + RSL3].

Western Blot Analysis

Whole-cell protein extracts were denatured, resolved on NuPAGE 4% to 12% Bis-Tris gels (ThermoFisher), and either transferred onto nitrocellulose membranes using iBlot 2 dry transfer apparatus (ThermoFisher) or PVDF membranes using the BioRad wet transfer apparatus. After blocking, blots were probed overnight, and the following primary antibodies were used: ER (Cell Signaling Technology Cat# 8644, RRID:AB_2617128), epidermal growth factor receptor (EGFR; Cell Signaling Technology Cat# 2232, RRID:AB_331707), HER2 (Cell Signaling Technology Cat# 2242, RRID:AB_331015), pHER2 (Cell Signaling Technology Cat# 2241, RRID:AB_2099407), HER3 (Cell Signaling Technology Cat# 12708, RRID:AB_2721919), HER4 (Cell Signaling Technology Cat# 4795, RRID:AB_2099883), AKT (Cell Signaling Technology Cat# 4691, RRID:AB_915783), pAKT (Cell Signaling Technology Cat# 9271, RRID:AB_329825), β-actin (Cell Signaling Technology Cat# 3700, RRID:AB_2242334), vinculin (Cell Signaling Technology Cat# 13901, RRID:AB_2728768), 4-hydroxynonenal (4-HNE) (Abcam Cat # ab46545, RRID:AB_722490), GPX4 (Abcam Cat# ab125066, RRID:AB_10973901), malondialdehyde (MDA) (Thermo Fisher Scientific Cat# MA5-27559, RRID:AB_2735264), and GAPDH (Proteintech Cat# 60004-1-Ig, RRID:AB_2107436). After 1 hour of incubation with secondary antibody (either anti-mouse [Cell Signaling Technology Cat# 7076, RRID:AB_330924] or anti-rabbit [Cell Signaling Technology Cat# 7074, RRID:AB_2099233]) and washes, antigen–antibody complexes were detected by the chemiluminescence WesternBright ECL Detection Reagent (Advansta) and imaged using the Amersham Imager 600 (GE Healthcare Life Sciences).

Real-time Polymerase Chain Reaction

Total RNA of biological triplicates was extracted from cells using the PureLink RNA Mini Kit (ThermoFisher) and converted to cDNA with iScript cDNA Synthesis Kit (BioRad) according to the manufacturer's instructions. Expression of target genes with specific primers (sequences listed in Table S1 (29); synthesized by Integrated DNA Technologies) was measured by real-time polymerase chain reaction (qRT-PCR) using the iTaq Universal SYBR Green Supermix (BioRad) and QuantStudio 12 K Flex Real-Time PCR System (ThermoFisher). Data were normalized to the ß-actin housekeeping gene and analyzed by the ΔΔCT method.

Single-Cell RNA Sequencing

Single-cell RNA sequencing (scRNAseq) was performed using the Drop-seq approach described by Macosko et al (30) with updates and detailed workflow from the McCarroll Laboratory at Harvard Medical School (https://mccarrolllab.org/dropseq/). To generate the droplet emulsion and cell encapsulation we used a Dolomite-Bio μEncapsulation system (Dolomite-Bio) according to the manufacturer's instructions. Barcoded beads were purchased from Chemgenes, as specified by Macosko et al (30). Briefly, trypsinized MM361 parental and LTED cells were washed and resuspended in phosphate-buffered saline with 0.01% bovine serum albumin. The single-cell suspension, bead suspension, and the droplet generation oil were loaded into their respective containers and connected via high precision pumps to the scRNA chip as part of the microencapsulation system. The flow rates were adjusted so that each droplet generated encapsulated 1 bead and 1 cell and monitored under the system's high speed digital microscope. Droplets were collected, and a small aliquot was examined microscopically to ensure uniformity of bead size and occupancy. The beads were washed and used for reverse transcription/cDNA synthesis and PCR according to the drop-seq workflow. The PCR products were purified, pooled, and quantified in a BioAnalyzer High Sensitivity Chip (Agilent). Sequencing libraries were prepared using the Nextera XT DNA sample prep kit (Illumina Inc) following the manufacturer's protocol. The libraries were purified, quantified, and sequenced by GenWiz on an Illumina High-Seq instrument using 2× 150 nt reads.

Raw data were imported into the Seurat R package (31). All cells with unique feature counts between 200 and 2500 and a percentage of mitochondrial reads less than 5% were selected for further analysis. After scaling, dimensional reduction, and cell type–specific marker identification with default parameters, we discovered 7 different clusters among LTED and parental cell lines. Differentially expressed genes (P < .05) were calculated between all 3 experimental groups. A combined LTED and parental cell lines uniform manifold approximation and projection (UMAP) was generated with 10 PC and 0.5 resolution. We also performed PAM50 molecular classification using genefu R package (32) at single-cell resolution; a treatment split UMAP was generated with the same settings.

Gene Set Enrichment Assay

Differentially expressed genes were matched against the REACTOME signature from the Human Molecular Signature Database using the gene set enrichment assay (GSEA) portal (http://software.broadinstitute.org/gsea/index.jsp) (33), with false discovery rate (FDR) q-values <0.05. The top 20 enriched pathways and their corresponding log (P value) were graphed.

Whole Genome Sequencing

DNA was extracted from cells using the DNeasy Blood and Tissue Kit (Qiagen) according to manufacturer's instructions. The Genomics and Epigenomics Shared Resource at Georgetown University Medical Center performed the whole genome sequencing (WGS). Paired-end, indexed libraries for human WGS were constructed from 1.0 μg of gDNA using the TruSeq DNA PCR-Free Library Prep Kit (Illumina) according to the manufacturer's instructions. Briefly, DNA was fragmented using a Covaris M220 Focused-ultrasonicator (Covaris) using settings for a 350-bp insert size. Library quality was assessed with a BioAnalyzer 2100 using the High Sensitivity DNA kit (Agilent Technologies). The libraries were quantified using the Kapa Library Quantification Kit Illumina Platforms (Kapa Biosystems). The denatured and diluted libraries were sequenced on a NextSeq 550 System (Illumina) using v2.5 High Output 300 cycle kit with 1% PhiX to an average sequencing depth of 50× coverage.

The quality of raw sequence data (fastq or fasq.gz files) was checked by FastQC v0.11.9 (34), and Cutadapt v3.5 was used for adapter trimming of raw data (35). After trimming, reads with low quality (quality score <33, error rate >10%) and lengths less than 25 bp were eliminated. Processed reads were then aligned to GRCh38 reference sequence using the bwa v0.7.17 paired-end mode (36). Mutation detection was conducted in Genome Analysis Toolkit v4.1.9.0 (37) following the best practice for variant calling workflow. SigProfiler (38) on COSMIC Mutational Signatures version 3.2 was used to classify single base substitution (SBS) in WGS data.

Statistical Analysis

All results are expressed as mean ± SEM. Prism 9 (GraphPad Software) was used for data analysis. Unless otherwise noted, one-way analysis of variance (ANOVA) followed by Tukey's or Dunnett's multiple comparison tests were employed to evaluate statistical significance between groups and P < .05 was considered to be statistically significant.

Results

Patients of the HER2+ BCa Subtype Expressing High Levels of ESR1 Exhibit Lower Risk for Metastasis

We analyzed a publicly available database to evaluate patients’ survival rates of HER2+ BCa stratified by their HR status. Regardless of HR status, 5-year survival rates from the SEER database for patients with HER2+ BCa decrease with disease progression and are the lowest for late (distant)-stage HER2+ BCa (Fig. 1A). Additionally, there is a significant decrease in the survival rates of patients with regional and distant stages of HR–/HER2+ when compared with the same stages of HR+/HER2+ BCa. Next, we specifically focused on ER and compared survival rates of patients with HER2+ BCa as stratified by high and low estrogen receptor alpha gene (ESR1) expression. Ten-year survival data from the Kaplan–Meier plotter database (39) show that higher ESR1 expression trends with better patient survival outcomes, as observed for overall survival (OS), relapse-free survival, and distant metastasis-free survival (DMFS) (Fig. 1B-1D). However, only patients expressing high levels of ESR1 show a statistically significant higher probability for DMFS (logrank P = .035, Fig. 1D), and thus lower risk for metastasis. This lower susceptibility for metastasis suggests correlation of high ER expression with better prognosis within the advanced setting of HER2+ BCa disease.

Figure 1.

Survival analysis for HER2+ BCa regarding their ER status. (A) Compared with early-stage disease, 5-year survival rates from the SEER database (https://seer.cancer.gov/statistics-network/explorer, accessed on February 6, 2023) are lowest for patients with late (distant)-stage HER2+ BCa subtype, regardless of their HR status. Error bars represent SEM. One-way ANOVA followed by Tukey's multiple comparison was performed to compare between groups; ***P < .001 denotes statistical significance. (B-D) Patients with the HER2+ BCa subtype expressing high levels of ESR1 have a significantly lower probability for metastasis. Ten-year survival curves for patients with HER2+ BCa stratified by high and low gene expression of ESR1 (205225_at). Plots were generated from Kaplan–Meier plotter database for BCa (www.kmplot.com, accessed on February 2, 2023) and show hazard ratio (HR) at 95% confidence, log-rank P values, and number of patients (n). OS, overall survival; RFS, relapse-free survival; DMFS, distant metastasis-free survival.

Growth Patterns and Responses to HER2- and ER-targeted Therapies of HER2+/ER+ LTEDs Differ From Their Parental Counterparts

To mimic acquired ETR to AI, we generated 2 LTED variants (A and B) from each of the 2 HER2-amplified, HER2+/ER+ BCa cell lines (Fig. 2A and 2B): BT474 and MM361 (26, 27). We first compared proliferation behaviors of LTEDs vs their corresponding parental cell lines grown in media supplemented with either FBS or CSS (CSS represents short-term hormone starvation). BT474 LTEDs had significantly lower growth rates than their respective parentals grown in FBS media but were not different from parentals in CSS media (Fig. 2C, left panel). CSS media are deprived of estrogen and other steroid hormones as well as small molecular weight growth factors. Therefore, the observed lower growth rates of BT474 LTEDs and parentals grown in CSS media could be explained by estrogen and/or growth factor deprivation. On the other hand, MM361 LTEDs grew faster than parentals in CSS and FBS media (Fig. 2C, right panel). Increased growth rates of MM361 LTEDs in CSS media are suggestive of hormone-independent growth and a more aggressive nature (LTEDs being more aggressive than parentals), consistent with the isolation of MM361 from brain metastatic BCa (40).

Figure 2.

Growth pattern and response to ER- and HER2-targeted therapies of HER2+/ER+ LTED variants differ from their parental counterparts. (A) Western blot validation that BT474 and MM361 express HER2 and ER. (B) Schematic illustrates our preclinical modeling of ER-targeted therapy resistance by establishing 2 LTED variants (A and B) for each of the BT474 and MM361 cell lines. Created with BioRender.com. (C) Growth curves representing viable cell numbers for parental (BT474 and MM361) and their LTED-derived (A and B) variants. Parental cells were cultured in either normal growth media (IMEM supplemented with 10% FBS) or hormone-deprived media (phenol red–free IMEM supplemented with 10% CSS). Parental cells of the CSS group were hormone deprived for 72 hours before seeding. Plots represent mean ± SEM from 2 independent experiments, each performed in triplicates and asterisks denote significant changes using one-way ANOVA followed by Dunnett's multiple comparison test. (D) Parental MM361 is intrinsically resistant to TP treatments whereas both MM361 and BT474 LTED models are less responsive to ICI. Crystal violet growth assay after 6 days of treatments with either vehicle, 10 nM E2 + 1 μM ICI or 1 μg/mL T + 1 μg/mL P quantifying viable cell numbers. Representative graph with data presented as mean ± SEM from 3 independent experiments with 6 readings and analyzed by ANOVA using Dunnett's multiple comparison test. *P < .05, **P < .01, and ***P < .001 were considered statistically to be significant.

Next, we evaluated our HER2+/ER+ parental/LTED pair growth response to HER2- and ER-targeted therapies: dual HER2 blockade by TP, and ICI as a selective estrogen receptor degrader (SERD). For BT474, LTED growth was significantly inhibited by TP and minimally inhibited by ICI, whereas either TP or ICI caused significant growth inhibition of parental cells (Fig. 2D, left panel). Compared with parental MM361, both MM361 LTED variants were also less responsive to ICI but more responsive to TP. Additionally, results from Western blot analysis showed no difference in ER protein levels of BT474 and MM361 LTEDs upon treatment with ICI vs vehicle, supporting the loss of ICI-mediated growth inhibition in HER2+/HR+ ETR LTEDs (Fig. S1A and S1B (29)). Conversely, ICI reduced ER protein levels of both BT474 and MM361 parentals. Altogether, the data show that our 2 HER2+/ER+ LTED models differ from their parental counterparts in baseline growth and responses to HER2- and ER-targeted therapies.

MM361, but not BT474, LTEDs Lose ER Expression and 17β-Estradiol-Induced ER Target Gene Expression

We further characterized our HER2+/ER+ LTED models to determine whether alterations in the basal expression of ER and HER family members may explain observed differential growth inhibition of parental/LTED pair models to anti-HER2 and ER treatments. When compared with their corresponding parentals, BT474 LTEDs showed no substantial changes in ER protein or ESR1 transcript levels, whereas a significant reduction of both was observed in MM361 LTEDs (Fig. 3A-3C). Additionally, ER transcriptional activity was confirmed by increased expression of ER target genes (PGR and TFF1) upon 17β-estradiol (E2) stimulation only in LTEDs of BT474, but not MM361 (Fig. 3D). Protein analysis of the HER family revealed a significant upregulation of EGFR and a modest increase of HER3 in BT474 LTEDs (Fig. 3A and 3B). A significant increase in HER2 and/or modest increase in EGFR (which is able to form heterodimers with HER2 (41)) protein levels of MM361 LTEDs (Fig. 3A and 3B) may explain why these cells responded better to growth inhibition by TP than parentals (Fig. 2D). LTED variants of both BT474 and MM361 cell lines showed activation of the prosurvival AKT signaling (pAKT) downstream of HER2 (Fig. 3A and 3B).

Figure 3.

Expression and genomic activity of ER is either preserved or lost in our HER2+/ER+ LTED models. (A) Western blot analysis show that ER protein is drastically reduced in the MM361 LTED models but preserved in the BT474 LTEDs. (B) Blots from (A) were analyzed by densitometry using ImageJ. Protein expressions were normalized to GAPDH and graphed as mean ± SEM from at least 2 independent experiments. ANOVA followed by Dunnett's multiple comparison test were performed to compare groups and statistical significance is shown on the graph. (C, D) Expression of ESR1 and ER target genes (PGR and TFF1) are maintained in BT474 LTEDs but drastically reduced in MM361 LTEDs. mRNA levels of ER and selected ER target genes under basal conditions or after estrogen stimulation (with either 1 nM or 10 nM E2 for 72 hours) were analyzed by qRT-PCR and data represented as mean ± SEM of fold changes in transcript expression from 4 independent experiments. Asterisks denote significant changes using ANOVA followed by Dunnett's multiple comparison test. *P < .05, **P < .01, and ***P < .001 were considered statistically significant.

MM361 LTEDs Gain Basal-Like and HER2-Enriched Intrinsic Subtypes and Exonic Mutations in Genes Encoding Transcription Factors and Chromatin Modifiers

Resistance to ET may arise due to changes in gene transcription and/or mutational alterations. The loss/downregulation of ER and increase in HER2 expression in the MM361 LTED model (Fig. 3), coupled with their more rapid adaptation to LTED conditions and increased growth rate vs parentals (Fig. 2C), raised the question of whether this model has shifted its intrinsic subtypes. Data from the SONABRE registry study supports this notion, reporting that HER2+/ER+ BCa has the highest rate of subtype discordance at metastatic presentation (42). We therefore prioritized the MM361 LTED model for further transcriptomic and genomic analyses to identify putative molecular mechanisms driving resistance. We first performed scRNAseq and analyzed the molecular subtype of individual cells using PAM50. MM361 LTEDs intrinsic subtyping is discordant from its parentals. As might be predicted from their loss/downregulation of ER expression at the mRNA and protein levels, LTEDs demonstrated a shift towards non-luminal HER2-enriched (HER2-E) and basal-like phenotypes (Fig. 4A and 4B). Additionally, we performed whole-genome sequencing to identify mutations gained in MM361 LTED. Analysis of SBS mutational signatures from our WGS data showed an enrichment in C to T base substitutions (SBS1) as well as T to C base substitutions (SBS5) across the genome of MM361 LTEDs compared with the parental cell line (Fig. 4C). These aging-related, clock-like mutational signatures are prevalent in many cancer types, though SBS1 mutations were recently reported to be enriched in BCa metastases in an age-independent manner (43). Next, we focused on mutated genes at their exonic regions. MM361 LTEDA and B shared a total of 70 genes bearing mutations in their exonic regions when compared with their parental cells (Fig. 4D). After excluding the 22 genes with silent mutations and characterizing the mutational signatures of the remaining 48, C to T and C to A were the most predominant mutation types (Fig. 4E; Table S2 (29)). Most of these mutated genes were modestly deleterious (Fig. 4F) as predicted by Varmap (44). These genes mainly encode for transcription factors and chromatin modifiers, namely HEY1, CHD4, MAFF, PRDM14, SATB2, SUPT6H, ZNF135. Importantly, mutation of 1 or more of these genes is significantly enriched in ETR advanced BCa tumors that are HER2+ and/or progesterone receptor negative (Fig. S2A and S2B (29)). Together, these data show that a shift towards more aggressive intrinsic molecular subtypes (basal and HER2 enriched) and acquiring mutated transcription factors and chromatin modifiers accompanies the development of ETR in the MM361 LTED model.

Figure 4.

MM361 LTEDs gain basal-like, HER2-E intrinsic subtypes and harbor mutations in transcriptional and chromatin regulatory factors. (A) UMAP plot of scRNAseq data grouped by predefined PAM50 intrinsic molecular subtypes. (B) Bar graph quantifies the distribution of intrinsic subtypes. Chi-square analysis was performed on raw cell numbers to test for differences in the proportions of intrinsic subtypes. (C-F) Characterizing MM361 LTED mutations using WGS. (C) Single base substitution (SBS) mutational signatures for the entire genome of MM361 LTEDs. (D) Venn diagram intersection showing shared mutated genes at their exonic regions of MM361 LTEDA and B vs parental cells. (E) Quantifying SBS for non-silent mutated genes at exonic regions. (F) Mutational effect predictions using Varmap (https://www.ebi.ac.uk/thornton-srv/databases/VarMap, accessed on August 11, 2022).

Lipid Metabolism Pathways are Upregulated in the MM361 LTEDs

Using our scRNAseq data, we performed unsupervised clustering using Seurat to identify overrepresented and underrepresented cell populations in MM361 parental and LTED A and B cells (Fig. S3 (29)). We identified 7 clusters, with clusters 2 and 5 enriched in parental cells and clusters 0 and 6 enriched in LTED cells (Fig. S3A to S3C (29)). GSEA analysis showed that parental-dominant clusters 2 and 5 were enriched with genes regulating estrogen response, glycolysis, tricarboxylic acid cycle, and oxidative phosphorylation (Fig. S3D (29)). By contrast, LTED-dominant clusters showed a significant overrepresentation of fatty acid metabolism (in the more abundant cluster 0), and RNA processing pathways (eg, nonsense-mediated decay) (in the less abundant cluster 6). We then constructed “pseudo-bulk” profiles from scRNAseq and performed differential expression analysis of transcriptomic changes that characterize MM361 LTEDA and B vs parental cells. Transcriptional profiling at a FDR <0.05 and a log2 fold change ≤−0.5 and ≥0.5 identified a total of 356 differentially expressed genes (DEGs) between parental and LTEDA, and 674 DEGs between parental and LTEDB (Fig. 5A). The intersection of DEGs from LTED vs parental comparisons resulted in 202 DEGs between both LTED lines and parental cells, with 156 and 46 genes being upregulated and downregulated, respectively (Fig. 5B and Table S3 (29)). We validated the transcriptomic analysis of MM361 LTEDs and confirmed the increased expression of eleven upregulated DEGs using qRT-PCR (Fig. 5C). Importantly, ESR1 and TFF1 were among the top downregulated genes and ERBB2 expression was upregulated (Fig. S4 (29)), consistent with data shown in Fig. 3. Using LISA (45) to predict the transcriptional regulators (TRs) of these DEGs (Table S3 (29)), we found that the steroid hormone receptors PGR, ESR1, and AR, as well as the pioneer factors GRHL2 and FOXA1 were among the most enriched TRs in common for both upregulated and downregulated DEGs. Predicted TRs associated exclusively with downregulated DEGs included E2F1, RARA, and MAFB (a homolog of MAFF, which we identified as mutated in LTED cells, Table S2 (29)). Predicted TRs for upregulated DEGs included GATA2, PPARG, and CEBPA.

Figure 5.

Lipid metabolism pathways are upregulated in LTEDs of the MM361 cell line. (A) Transcriptional profiling from scRNAseq at FDR < 0.05 and log2 fold change ≤−0.5 and ≥0.5 identified 356 DEG between parental and LTEDA and 674 DEG between parental and LTEDB. (B) A total of 202 DEGs are shared in both LTED lines vs parental cells. Venn diagram intersections show 156 shared upregulated and 46 shared downregulated DEGs in the MM361 LTEDs vs parental. (C) qRT-PCR was used to confirm increased expression of selected upregulated genes identified from the scRNAseq transcriptomic profiling. Data presented as mean ± SEM of fold changes in expression of each gene (LTEDs relative to parental cells) from 3 independent experiments. Asterisks denote significant changes relative to parental MM361 using ANOVA followed by Dunnett's multiple comparison test. (D) Top 20 enriched biological processes in MM361 LTEDs. Overlapped upregulated and downregulated DEGs (from B) in MM361 LTEDs were analyzed by GSEA using the REACTOME gene set, with FDR q-values <0.05. Red- and blue-filled boxes highlight metabolic pathways. (E) UMAP plots for ALOX15B and GPX4, among the upregulated genes of the lipid metabolism pathway being enriched in the MM361 LTEDs.

To identify biological processes that are enriched within the LTED variants, we performed GSEA on the upregulated and downregulated shared gene lists of both LTED lines vs parental cells. Genes upregulated in LTEDs were enriched in pathways involved in defective apoptosis and immune-related pathways. Conversely, pathways involved in estrogen and other nuclear receptor signaling were enriched within the downregulated gene set (Fig. 5D). Notably, metabolic pathways were among the enriched pathways in the MM361 LTEDs. While lipid metabolism was upregulated, pathways of glucose metabolism, glycolysis, and gluconeogenesis were all downregulated (Fig. 5D), consistent with our unsupervised clustering analysis results (Fig. S3 (29)). The upregulation of lipid metabolism pathways is also supported by the observed enrichment of putative TRs PPARG and CEBPA from our upregulated DEGs (Table S3 (29)), which are key activators of lipid metabolism (46). Our data suggest notable metabolic remodeling of MM361 LTEDs, specifically upregulation of lipid metabolism.

Dual Targeting of HER2 and GPX4 Increase Cell Death of HER2+/ER+ LTEDs

ALOX15B and GPX4 are 2 of the upregulated genes of lipid metabolism pathways enriched in the MM361 LTEDs (Fig. 5D and 5E). In addition, ALOX15B and GPX4 were among the 13 MM361 LTED DEGs that overlapped with ferroptosis-related genes (Fig. S5 (29)). Ferroptosis is an iron-dependent cell death that is triggered by lipid peroxidation (47). Interestingly, both upregulated proteins are opposing key players in regulating ferroptosis (48); ALOX15B induces ferroptosis, whereas GPX4 inhibits it. We validated the upregulation of ALOX15B mRNA in MM361 LTEDs (Fig. 5C), and GPX4 protein levels in both MM361 and BT474 LTEDs (Fig. 6A and S6A (29)).

Figure 6.

Dual targeting of HER2 and GPX4 increase cell death of MM361 LTEDs. (A) Representative immunoblot of 3 independent experiments showing enhanced expression of GPX4 protein in MM361 LTEDs. (B-D) Cells treated with either vehicle, 1 μg/mL T + 1 μg/mL P, 1 μM RSL3, or a combination of both for 72 hours then stained with Hoechst to monitor total cell number (gray nuclei), and Sytox green to monitor dead cells (green nuclei) as in panels B and C, or analyzed by Western blot analysis as in D. (B, C) Sytox staining experiments show that cotreatments of TP + RSL3 induce the highest cell death in MM361 LTEDs. (B) Cell death representative images and (C) quantifications from 2 independent experiments with at least 10 fields analyzed for each and data presented as mean ± SEM of cell death percentages. Asterisks denote significant changes using one-way ANOVA followed by Tukey's multiple comparison test; *P < .05 and ***P < .001 were considered to be statistically significant. RI value (28) for LTEDA = 1.03 (additive), and LTEDB = 1.02 (additive). (D) Treatment effects on GPX4 and pAKT protein levels were assessed with Western blot analysis. Immunoblots are a representation of 2 independent experiments. Blots were analyzed by densitometry using ImageJ. Protein expressions were normalized to vinculin and graphed as mean ± SEM from 2 independent experiments. ANOVA followed by Dunnett's multiple comparison test were performed to compare groups and statistical significance is shown on the graph.

Inhibiting the PI3K-AKT-mTOR signaling pathway— shown in Fig. 3 to be strongly activated in both MM361 and BT474 LTED cells—can sensitize BCa cells to ferroptosis induction, in part by reducing SREBP-mediated lipogenesis (49). Therefore, we hypothesized that LTED cells of HER2+/ER+ BCa are resistant to ferroptosis, due to GPX4 upregulation, and that dual targeting of HER2 (upstream of PI3K-AKT-mTOR) and GPX4 may sensitize these LTEDs to ferroptosis-induced cell death. In MM361 LTEDs, significant cell death was observed following HER2 inhibition by TP, but not GPX4 inhibition by RSL3 (Fig. 6B and 6C). In BT474 LTEDs, TP and RSL3 alone each significantly increased cell death (Fig. S6B and S6C (29)). Importantly, in both LTED models, cell death was significantly enhanced by the combination of TP + RSL3. Consistent with this, combinatorial treatments of TP + RSL3 drastically inhibited cell growth of both LTED models (Fig. S6D (29)). As expected, TP alone reduced the activation of AKT (pAKT) and RSL3 alone reduced protein levels of GPX4 (Fig. 6D), but no further reduction was observed with TP + RSL3 vs that of the single treatments (Fig. 6D). To test whether the increased cell death induced by the TP + RSL3 combination was due to ferroptosis, we measured the levels of 2 ferroptosis lipid peroxidation markers, 4-HNE and MDA (Fig. S7 (29)). RSL3 alone slightly increased levels of both 4-HNE and MDA, but TP + RSL3 did not further enhance the levels of 4-HNE and MDA protein conjugates. Together, these data suggest that increased lipid metabolism in HER2+/ER+ BCa LTED models is accompanied by reprogramming of ferroptosis pathways, and that the combination of anti-HER2 agents and a ferroptosis inducer is an effective approach to inducing cell death.

Discussion

To date, models of acquired ETR have been almost exclusively developed for HER2–/ER+ BCa, and not HER2+/ER+. In this study, we generated and characterized models of the latter without using genetic manipulation approaches. We report that ETR BT474 and MM361 AI-resistant (LTED) models capture distinct phenotypes of HER2+/ER+ BCa as supported by the following results. Compared with parentals, MM361 LTEDs grew faster, lost ER, and increased HER2 expression, whereas BT474 LTEDs grew slower and showed no substantial changes in ER or HER2 expression. Additionally, growth inhibition by anti-HER2 (TP) treatment was much more pronounced in the LTEDs vs parental MM361, unlike the BT474 model, where both LTEDs and parentals benefited from TP. Differences in ER and HER2 expression in the LTEDs of MM361 and BT474 models may suggest distinct resistance mechanisms to ET partly due to the differential distribution of intrinsic subtypes within these 2 cell lines. It could imply that BT474 ETR behaves more like luminal whereas MM361 LTEDs behave more like non-luminal HER2+ BCa and hence the more aggressive nature of the latter. It is worth noting that CSS removes estrogen and other steroid hormones as well as small molecular weight growth factors. Therefore, some of our reported observations within this study could be attributed to hormone and/or growth factor deprivation.

HR and HER2 receptor expression, or lack thereof, are critical in guiding BCa treatments. However, pathology-based immunohistochemistry does not fully recapitulate intrinsic biology. Instead, substantial discrepancies exist between immunohistochemistry-based and gene expression-based PAM50 subtyping (50). Our results show that MM361 LTEDs intrinsic subtyping is discordant from its parentals. We observed a loss in ER/ESR1 accompanied by a gain in HER2/ERBB2 expression and, thus, molecular switching to non-luminal HER2-E phenotype and sustained activation of AKT signaling. These alterations altogether explain why our ETR MM361 LTEDs were less responsive to ICI but more responsive to anti-HER2 targeted therapies vs their corresponding TP-resistant parentals. HER2-E subtype represents ∼75% HER2+/HR– and 30% of HER2+/HR+ tumors (50, 51), is associated with anti-HER2 sensitivity (50), and is a predictive biomarker of poor response and resistance to AI (10). Our findings are supported by several studies. First, data from the SONABRE registry (NCT03577197) show that among all BCa subtypes, patients with HR+/HER2+ advanced BCa exhibit the highest receptor subtype discordance rate between primary tumor and metastatic lesions as they convert mainly to HR+/HER2– and HR-/HER2+ (42). In other words, ER and HER2 expression varied during tumor progression, and crosstalk between ER and HER2 was proposed as the leading cause of this discordance (42). Second, HER2 and ER expression are negatively correlated in HER2+ BCa tumors (52). Third, cell lineage tracing experiments in mice showed that HER2+/ER+ cells can lose ER expression and be converted into highly aggressive HER2+/ER– BCa cells (53). Lastly, HER2 overexpression (with or without EGFR) activates downstream AKT and MAPK pathways to circumvent ER inhibition and promotes growth of ETR BCa (2, 25, 54). ER loss in our MM361 ETR model could be explained by ESR1 promoter methylation and/or decreased FOXA1 and GATA3, which are required for normal ER expression (53). Additionally, losing the ER-regulated cluster of let-7c, miR99a, and miR125b (55) could explain our upregulated HER2 expression.

Beyond the HER2-E intrinsic subtype, acquiring mutations in genes encoding for transcription factors and chromatin modifiers could underlie mechanisms of AI resistance and/or the aggressive nature of the MM361 LTED model. For instance, integrated genomic analysis of responding vs resistant ER+ breast tumors from a patient who developed resistance to letrozole treatment revealed an acquired mutation in the CHD4 gene (56). SUPT6H protein levels decrease in poorly differentiated breast tumors, and SUPT6H is a requirement for ERα transcriptional activity and maintenance of chromatin structure in BCa cells (57). Also, SATB2 induces transformation of mammary epithelial cells into progenitor-like cells and its knockdown attenuates proliferation and epithelial–mesenchymal transition of BCa cells (58). What role the mutations in these genes that we observe in MM361 LTEDs play in the ETR phenotype is a subject of future study.

BCa cells carrying PIK3CA activating mutation (like the parental BT474 and MM361 cells (59)), and thus an activated PI3K-AKT-mTOR signaling pathway, are resistant to ferroptosis induction by the GPX4 inhibitor RSL3 (49). Consistent with this, inhibiting an activated PI3K-AKT-mTOR signaling pathway sensitizes BCa cells to ferroptosis induction by RSL3 (49). In the current study, we show that dual targeting of HER2 and GPX4 increases cell death of HER2+/ER+ LTEDs but not levels of ferroptosis lipid peroxidation markers beyond that of single-agent treatments. One possible explanation could be that GPX4 inhibition may not be sufficient to overcome ferroptosis resistance in our LTEDs. In other words, elevated GPX4 may not be the sole cause of ferroptosis resistance in MM361 LTEDs; instead, several other ferroptosis-related DEGs may mediate this resistance. Dysregulated oxidative and iron homeostasis exemplified by upregulation of ferroptosis suppressor genes (TMBIM4 and GPX4), antioxidants (PRDX2), iron chelators (LCN2), and/or downregulation of ferroptosis inducing genes (RPL8 and HILPDA) could be responsible for ferroptosis resistance in our LTEDs (Fig. S3 (29)). Moreover, multiple ferroptosis regulatory pathways that are independent of GPX4 have recently been discovered (60). For instance, cisplatin-resistant–derived exosomes secrete microRNAs that increase the expression of ferroptosis suppressor protein 1 and thus enhance resistance of cancer cells to ferroptosis (61). Along the same lines, LTED cells treated with TP + RSL3 may protect themselves from ferroptosis using a similar mechanism. In our hands, TP did not reduce GPX4 nor enhance expression of ferroptosis lipid peroxidation markers. However, a recent study reported that trastuzumab triggers ferroptosis by reducing GPX4 whilst increasing ROS levels in embryonic rat myoblast (H9c2) cells (62). This inconsistency could be attributed to variances in cell context, doses and/or duration of treatments.

Inevitably, our study has several limitations that must be considered. Future xenograft studies will be necessary to verify and follow up on some of our current findings, such as the response of ETR HER2+/ER+ to ET and anti-HER2 treatments and whether ferroptosis induction combined with anti-HER2 treatments suppresses the growth of these cells in vivo. Despite harboring a sufficient gain in ERBB2 copy number to be classified as HER2+ (amplified (63, 64)), a previous study demonstrated that the parental MM361 cell line exhibits heterogenous expression of ERBB2 and ESR1, and this heterogeneity is not due to heritable genetic differences (64), suggesting a high degree of plasticity and variability. This informed our rationale for performing scRNAseq to capture transcriptional heterogeneity of the parental cells and ETR LTED models. Importantly, our conventional and “pseudo-bulk” analyses of the scRNAseq data both identified alterations in lipid metabolism. However, we sequenced a relatively small number of cells by scRNAseq (<4000 cells per sample), and combined with the fact that scRNAseq is best suited to detecting higher abundance transcripts, we cannot exclude the possibility that we are missing minor populations or subclones that contribute significantly to the ETR phenotype (eg (65, 66)). The gene mutations we identified in LTEDs are constrained by the WGS coverage that we achieved (LTEDA: 53×, LTEDB: 50×, Parental: 52×). Furthermore, deeper sequencing depths may capture and result in a larger number of mutated genes. However, even at the sequencing depths and coverage levels we used, our results may suggest that multiple potential resistance mechanisms exist. Deep sequencing of a different epithelial tumor, colorectal cancer, coupled with an evolutionary analysis, showed that the ∼10⁹ cells in a 1 cm³ lesion will contain every possible resistance mutation in a minor subclone within the lesion (67). This large diversity of resistance mechanisms implies a need for treatment regimens that include therapies too numerous to be given simultaneously in combination (68).

Increasing efforts are being directed toward chemotherapy-sparing regimens for HER2+/ER+ BCa (69-72). However, these treatment strategies should also be tested on HER2+/ER+ BCa that is refractory to or have progressed on ET. Therefore, future studies are needed to address open questions regarding (1) responsiveness of ETR HER2+/ER+ models to dual and triple combinations of anti-HER2, ET/SERDs (including recently approved oral SERD amcenestrant), and CDK4/6 inhibitors, as well as innovative sequencing of these combinations (64, 65, 68, 73), (2) the impact of our identified mutations on development of ETR in HER2+/ER+ BCa, and (3) differences in ER activity between ETR models of HER2+/ER+ and other HER2-ER+.

Conclusions

Characterizing models of HER2+/ER+ that mimic a real-world treatment pattern is necessary to understand mechanisms of resistance to ET and may provide useful information for refining current treatment approaches and improving patients’ outcomes. Here, we report that anti-HER2 targeted therapies effectively inhibit growth of ETR HER2+/ER+ BCa cells that exhibit concurrent loss of ER expression and gain in HER2 and HER2-E phenotype. Our BT474 and MM361 LTED models capture distinct phenotypes of HER2+/ER+ BCa and pinpoint altered lipid metabolism and ferroptosis remodeling as vulnerabilities of this type of ETR BCa.

Abbreviations

4-HNE

4-hydroxynonenal

AI

aromatase inhibitor

ANOVA

analysis of variance

BCa

breast cancer

BT474

BT-474

CSS

charcoal-stripped bovine serum

DEG

differential expressed gene

DMFS

distant metastasis-free survival

E2

17β-estradiol

EGFR

epidermal growth factor receptor

ER

estrogen receptor

ET

endocrine therapy

ETR

endocrine therapy resistant

FBS

fetal bovine serum

FDR

false discovery rate

GSEA

gene set enrichment assay

HER2

human epidermal growth factor receptor 2

HR

hormone receptor

ICI

fulvestrant

IMEM

improved minimum essential medium

LTED

long-term estrogen deprivation

MDA

malondialdehyde

MM361

MDA-MB-361

qRT-PCR

real-time polymerase chain reaction

RI

Relative Index

SBS

single base substitution

scRNAseq

single-cell RNA sequencing

SERD

selective estrogen receptor degrader

TP

trastuzumab + pertuzumab

TR

transcriptional regulator

WGS

whole genome sequencing

Funding

This work was funded by the Department of Defense (DoD) Breast Cancer Research Program awards W81XWH-20-1-0759 and W81XWH-20-1-0760 (to R.B.R. and R.A.B., respectively), the Oak Ridge National Laboratory Director's R&D fund (to M.P. and R.A.B.), and philanthropy support from Lombardi Women at Georgetown Lombardi’s Nina Hyde Center for Breast Cancer Research (to R.B.R.). Fellowship support for H.S. and S.T. was provided by the Tumor Biology Training Grant T32 CA009686 (principal investigator: Dr. Anna T. Riegel). D.M. received support from the Georgetown Regents Scholars Program. Technical services were provided by the Genomics and Epigenomics, Microscopy and Imaging, and Tissue Culture Shared Resources at GUMC, which are supported, in part, by NIH/NCI Cancer Center Support Grant P30 CA051008 (principal investigator: Dr. Louis M. Weiner). The content of this article is the sole responsibility of the authors and does not represent the official views of the DoD or NIH.

Authors Contributions

S.B., H.S., and R.B.R. designed the experiments. S.B. performed most of the experiments, analyzed data, prepared figures, and drafted the manuscript. H.S. performed, analyzed, and graphed some of the RT-PCR and growth assays. M.B. and M.P. performed the scRNAseq experiments and L.J. analyzed the scRNAseq and WGS data. S.T. and D.M. performed most of the western blot analyses. M.D.M. and R.A.B. contributed to the study discussions. R.B.R. supervised the study, wrote parts of the manuscript, reviewed, and revised it. All authors read, revised, and approved the final manuscript.

Disclosures

S.B., H.S., L.J., S.T., D.M., M.B., M.P., and M.D.M. have nothing to declare. R.A.B. consults for AstraZeneca LP and Boehringer Ingelheim, and is the Chief Scientific Officer and Managing Member of Onco-Mind, LLC, which owns patents related to cancer precision medicine. R.B.R. is an Associate Editor for the Journal of the Endocrine Society.

Data Availability

Original data generated and analyzed during this study are included in this published article or in the data repository listed in References ((29); https://doi.org/10.5281/zenodo.8265999). Raw sequencing data have been submitted to Sequence Read Archive (SRA, SUB13842364 for whole-genome sequencing, SUB13842085 for scRNAseq) in association with BioSample deposit SUB13858008.

Ethics Approval and Consent to Participate

Not applicable.

Consent for Publication

All authors agreed to publish this study.