Distinct mechanisms of resistance to fulvestrant treatment dictate level of ER independence and selective response to CDK inhibitors in metastatic breast cancer

Abstract

Background: Resistance to endocrine treatment in metastatic breast cancer is a major clinical challenge. Clinical tools to predict both drug resistance and possible treatment combination approaches to overcome it are lacking. This unmet need is mainly due to the heterogeneity underlying both the mechanisms involved in resistance development and breast cancer itself.

Methods: To study the complexity of the mechanisms involved in the resistance to the selective estrogen receptor degrader (SERD) fulvestrant, we performed comprehensive biomarker analyses using several in vitro models that recapitulate the heterogeneity of developed resistance. We further corroborated our findings in tissue samples from patients treated with fulvestrant.

Results: We found that different in vitro models of fulvestrant resistance show variable stability in their phenotypes, which corresponded with distinct genomic alterations. Notably, the studied models presented adaptation at different cell cycle nodes to facilitate progression through the cell cycle and responded differently to CDK inhibitors. Cyclin E2 overexpression was identified as a biomarker of a persistent fulvestrant-resistant phenotype. Comparison of pre- and post-treatment paired tumor biopsies from patients treated with fulvestrant revealed an upregulation of cyclin E2 upon development of resistance. Moreover, overexpression of this cyclin was found to be a prognostic factor determining resistance to fulvestrant and shorter progression-free survival.

Conclusions: These data highlight the complexity of estrogen receptor positive breast cancer and suggest that the development of diverse resistance mechanisms dictate levels of ER independence and potentially cross-resistance to CDK inhibitors.

Keywords: Cell cycle inhibitors; Cyclin E2; Endocrine treatment resistance; Fulvestrant; Metastatic breast cancer.

Fig. 1

In vitro generated fulvestrant-resistant breast cancer cells downregulate ER signaling and proliferate independently of estrogen. a Dose-response curves for parental (-P, black lines) and fulvestrant-resistant (-FR, red lines) cells from six different breast cancer cell lines in response to 6-day treatment with fulvestrant (1 pM to 100 μM). Graphs represent combined data (average ± SEM) from two biological experiments with three technical replicates each. IC50 values are reported in Additional file 3C. b ERE reporter activity in parental and fulvestrant-resistant CAMA-1, MCF7, and HCC1428 cells after treatment for 24 h with (+F) or without (-F) supplementation of 100 nM fulvestrant. Graphs represent combined data (average ± SEM) from two biological experiments with two technical replicates each. Statistical differences were determined with one-way ANOVA and Tukey’s post-hoc test, *** represents p value ≤ 0.0001 between untreated parental and fulvestrant-resistant cells. Data for the ZR-75-1, T47D, and EFM-19 models are shown in Additional file 3E. c Western blotting for ERα expression in parental and fulvestrant-resistant CAMA-1, MCF7, and HCC1428 cells after treatment for 24 h with 100 nM fulvestrant (+F) or DMSO (-F). Representative data from three biological replicates. β-actin or β-tubulin was used as loading control. Quantification of band intensities is presented in Additional file 4A. Data for the ZR-75-1, T47D, and EFM-19 models are shown in Additional file 3F. d Proliferation curves for parental and fulvestrant-resistant CAMA-1, MCF7, and HCC1428 cells in estrogen-depleted media with (+F) or without (ctrl) supplementation of 100 nM fulvestrant. Graphs represent combined data (average ± SEM) from two biological experiments with three technical replicates each. Statistical differences were determined with two-way ANOVA and Tukey’s post-hoc test, *** represents p value ≤ 0.0001 and * ≤0.05 between untreated parental (black solid lines) and fulvestrant-resistant (red solid lines) cells at given time-points

Fig. 2

Fulvestrant-resistant models show variable stability upon fulvestrant withdrawal. a–c Fulvestrant dose-response curves (5 nM to 10 μM) and western blotting for ERα expression in fulvestrant-resistant (-FR) CAMA-1 (a), MCF7 (b), and HCC1428 (c) cells cultured either continuously with fulvestrant (+F, red solid lines) or after removal of fulvestrant (-F, red dotted lines) from the growth media for the indicated times (week 1-week 9). Parental (-P) cells cultured without fulvestrant were used as control (black solid lines). Graphs represent combined data (average ± SEM) from three biological replicates with at least three technical replicates each. Samples for western blotting were collected at each time-point and ERα protein expression was assessed. β-actin or β-tubulin was used as loading control. Representative data from three biological replicates is presented under each graph. Quantification of band intensities is presented in Additional file 4B-D. Stars indicate differences between fulvestrant-resistant cells grown with (red, solid lines) or without (red, dotted lines) fulvestrant. Stars are indicated at every other data point due to restricted space. Data for the ZR-75-1, T47D, and EFM-19 models are shown in Additional file 5

Fig. 3

Fulvestrant-resistant models show variable response to estrogen stimulation and tamoxifen treatment. a, b Relative proliferation of parental and fulvestrant-resistant CAMA-1, MCF7, and HCC1428 cells in estrogen depleted (a) or normal (b) growth media with or without supplementation with 1 nM estradiol (E2) (a) or 100 nM 4-hydroxytamoxifen (4-OHT) (b) for 6 days. Each graph represents combined data (average ± SEM) from two biological experiments with three technical replicates each. Statistical differences were determined using one-way ANOVA with Tukey’s post-hoc test. *** represents p value ≤ 0.0001, ** ≤ 0.001, ns represents no statistical differences. Stars and “ns” in a indicate statistical differences compared to -E2 for each cell model unless indicated otherwise. Stars and “ns” in b indicate statistical differences compared to -4-OHT for each cell model unless indicated otherwise. Data for the ZR-75-1, T47D, and EFM-19 models are shown in Additional file 6

Fig. 4

Fulvestrant-resistant models show differences in genomic regulation but similarities in transcriptional output. a Histograms of the difference in copy number estimates between fulvestrant-resistant and corresponding parental cell lines for all six models analyzed by Illumina GSA arrays. If no difference exists between cell lines, the histogram would show one narrow spike at 0. Observation of smaller subpopulations of SNP probes with difference in copy number between resistant and parental cells indicates that copy number alterations have occurred. BAF and LogR files are shown in Additional file 7. b Graphical illustration of pattern correlation analysis identifying genes that are downregulated in parental cells upon treatment with 100 nM fulvestrant for 24 h, and then upregulated upon development of fulvestrant resistance. c Venn diagram of genes following this pattern. One hundred sixty genes were found significant across CAMA-1, MCF7, and T47D models. d Gene expression heatmap of 160 genes (rows) found to be significant based on correlation and expression variance cutoffs in the pattern correlation analysis in all three cell lines. Samples (columns) are ordered by (i) cell line, (ii) treatment type (parental control, parental with 24 h fulvestrant treatment (F 24 h), and resistant (FR, F 24 h)), and (iii) biological replicate. Green corresponds to decreased expression (lower limit − 1.5 in expression) and red to increased expression (upper limit 1.5 in expression). Eighteen genes present in the KEGG pathway “Cell cycle” are indicated to the right. All 160 genes are listed in Additional file 1. e To determine commonly enriched processes, the list of 160 identified genes was subjected to gene ontology annotation using the DAVID bioinformatics tool (see also Additional file 8). Data were visualized using the REVIGO web server. Data are displayed on the x- and y-axis based on semantic similarity; hence, more similar GO terms cluster more closely together. The size of each node represents the log10 p value of enrichment, the larger the node, the smaller the p value, and the more significant the enrichment. Most significantly enriched processes are presented

Fig. 5

Fulvestrant-resistant models present differential regulation of cell cycle proteins and different sensitivity to CDK inhibition. a Western blotting for expression of cell cycle regulating proteins in parental and fulvestrant-resistant CAMA-1, MCF7, HCC1428, and ZR-75-1 cells after treatment for 24 h with 100 nM fulvestrant (+F) or DMSO control (-F). Representative data from three biological replicates. β-actin (CAMA-1, MCF7, and ZR-75-1) or β-tubulin (HCC1428) were used as loading control. Quantification of band intensities is presented in Additional file 4H-K. Data for the T47D and EFM-19 models are shown in Additional file 9B. b, c Dose-response curves for parental and fulvestrant-resistant CAMA-1, MCF7, HCC1428, and ZR-75-1 cells in response to 6-day treatment with increasing concentrations of the CDK4/6 inhibitor palbociclib (1 nM to 5 μM) (b) or a CDK1/2 inhibitor (CDKi3, 100 pM to 50 μM) (c). Graphs represent combined data (average ± SEM) from three biological experiments with three technical replicates each. IC50 values are shown in Additional file 9C and E. Data for the T47D and EFM-19 models are shown in Additional file 9D and F. d Dose-response curves for parental (-P, black solid lines), fulvestrant-resistant (-FR, red solid lines), palbociclib-resistant (-PalbRes, black dotted lines), and double fulvestrant- and palbociclib-resistant (-FR-PalbRes, red dotted lines) CAMA-1 and MCF7 cells in response to 6-day treatment with fulvestrant (10 pM to 50 μM). Graphs represent combined data (average ± SEM) from minimum two biological experiments with three technical replicates each. IC50 values are shown in Additional file 9J. Dose-response curves in response to palbociclib are shown in Additional file 9H

Fig. 6

Cyclin E2 levels increase after progression on fulvestrant and correlate with clinical outcome. a Representation of cyclin E2 expression in samples from three patients with ER+ breast cancer before initiation of fulvestrant treatment (pre) and upon progress during treatment (post) as determined by immunohistochemical stainings. Quantifications of cyclin E2 scoring for each pre- and post- sample are shown to the left and representative stainings are shown to the right. b Evaluation of progression-free survival (PFS) in tumors with cyclin E2 low versus high expression in ER+ breast cancer patients treated with fulvestrant in the advanced setting. Kaplan-Meier plots for PFS are shown to the left and representative stainings for low and high cyclin E2 expression are shown to the right. p value represents log-rank test for PFS between patients with high and low levels of cyclin E2 expression. Scale-bar represents 25 μm in (a, b). c Cox multivariate regression analysis of PFS according to cyclin E2 expression levels, age at diagnosis of metastasis, and site of recurrence