Selective pressure of endocrine therapy activates the integrated stress response through NFκB signaling in a subpopulation of ER positive breast cancer cells

Abstract

Background: While estrogen receptor (ER) positive breast tumors generally respond well to endocrine therapy (ET), up to 40% of patients will experience relapse, either while on endocrine therapy or after ET is completed. We previously demonstrated that the selective pressure of tamoxifen activates the NFκB pathway in ER + patient tumors, breast cancer cell lines, and breast cancer xenograft tumors, and that this activation allows for survival of a subpopulation of NFκB + cells that contribute to cell regrowth and tumor relapse after ET withdrawal. However, the mechanisms contributing to the expansion of an NFκB + cell population on ET are unknown.

Methods: Here, we utilized single-cell RNA sequencing and bioinformatics approaches to characterize the NFκB + cell population and its clinical relevance. Follow-up studies were conducted to validate our findings and assess the function of the integrated stress response pathway in breast cancer cell lines and patient-derived models.

Results: We found that the NFκB + population that arises in response to ET is a preexisting population is enriched under the selective pressure of ET. Based on the preexisting NFκB + cell population, we developed a gene signature and found that it is predictive of tumor relapse when expressed in primary ER + tumors and is retained in metastatic cell populations. Moreover, we identified that the integrated stress response (ISR), as indicated by increased phosphorylation of eIF2α, occurs in response to ET and contributes to clonogenic growth under the selective pressure of ET.

Conclusions: Taken together, our findings suggest that a cell population with active NFκB and ISR signaling can survive and expand under the selective pressure of ET and that targeting this population may be a viable therapeutic strategy to improve patient outcome by eliminating cells that survive ET. Understanding the mechanisms by which breast cancer cells survive the selective pressure of ET may improve relapse rates and overall outcome for patients with ER + breast tumors.

Keywords: Breast cancer; Endocrine therapy; Estrogen receptor; Integrated stress response; NFκB.

Fig. 1

Activation of NFκB by different ETs in preclinical ER + breast cancer models. a–c Clonogenic assays were conducted in MCF-7-NFκB-RE-GFP and T47D-NFκB-RE-GFP cell lines cultured in growth media (GM) in the presence or absence of 1 μM 4OHT or ICI, or in ED conditions for 2 weeks. a Colony confluence (area covered by colonies) was quantified using a Celigo imaging cytometer. b The percentage of GFP + confluence per condition was determined using Celigo imaging. c Representative images of colonies for the MCF-7-NFκB-RE-GFP cell line grown in GM ± 4OHT for 2 weeks. Scale bar: 200 µm. d HCI-003 and HCI-017 PDxOs were grown under the selective pressure of ETs for 2 weeks. Data represent total organoid area per µm² on day 14 as determined by the Incucyte S3 organoid module. e, f Expression of ER target genes e and NFκB target genes f was determined by QPCR in HCI-003 and HCI-017 PDxOs treated with ET for 2 weeks. The ER target genes were used as controls for ET. The heatmap represents Fold Change for each gene relative to GM control

Fig. 2

Single-cell RNA sequencing identifies an NFκB + cell population. a scRNA sequencing was conducted on MCF-7 cells treated with 4OHT for 2 weeks in clonogenic conditions. Bi-dimensional representation of 648 single-cell transcriptomes is shown (UMAP). b Box plots represent the Hallmark NFκB signature pathway score (median, interquartile values, range, and outliers) for each cell cluster. FEA for each cluster was performed but threshold for enrichment was not reached for either (Cluster 0: AUC 0.56, P = 0.02; Cluster 1: AUC 0.44, P = 0.02). c Bar plot showing the distribution of z-scores for the Hallmark NFκB signature on a per cell basis. d NFκB + cells (green) and NFκB- cells (gray), based on individual cell z-scores, are distributed across cell clusters defined in a. e FEA was performed for additional known NFκB genes signatures from MsigDB. AUC values are shown in a heatmap, and P-values are presented in Additional file 2: Table S1

Fig. 3

Integration of two single-cell RNA sequencing datasets from untreated and 4OHT-treated MCF-7 cells. a Single-cell transcriptomes from 2618 parental MCF-7 cells and 648 4OHT-treated MCF-7 cells were integrated and represented bi-dimensionally using Seurat package v.3.1. b The percent change in abundance of cell populations with 4OHT treatment relative to the total population for each group are shown. *P < 0.05, ***P < 0.001, ns = not significant. c FEA was performed for NFκB gene signatures with one representative example shown in box plots. AUC and P-values for other NFκB signatures are presented in Additional file 3: Table S2. d A custom gene signature was derived from differentially expressed genes in tumors of patients receiving neoadjuvant tamoxifen. The signature was used for FEA and the signature score per cluster is presented in boxplots

Fig. 4

Clinical relevance of an NFκB + population signature. a, b The NFκB + Population Signature was interrogated in 1817 ER + breast tumors from the Metabric cohort available in cBioPortal for Cancer Genomics. Histologic grade (a) and patient relapse free survival (b) between tumors + vs. – for expression of the NFκB + Population Signature are displayed. Statistical significance was determined using chi-squared test (a) or log-rank test (b). ***P < 0.001. c Single-cell transcriptomes of from primary and metastatic tumors of PDX models UCD65 (top) and UCD4 (bottom) are represented in UMAP plots. d The proportion of cells in each cluster is indicated by their origin (i.e., primary tumor or metastatic location). e FEA was performed for the NFκB + Population Signature on both datasets with box plots showing signature scores per cluster

Fig. 5

Expression of DEGs from the NFκB + cell population in multiple ER + breast cancer models treated with ETs. a Top DEGs in the NFκB + cell population (Cluster4) are represented in dot plots, with color representing expression level and size representing the percentage of cells in the cluster expressing the gene. b Expression of DEGs was determined by QPCR in MCF-7 and T47D cell lines and HCI-003 and HCI-017 PDxOs treated with different ETs for two weeks. The heatmap represents Fold Change for each gene relative to GM control. c Expression of top DEGs in the original 4OHT-treated cell populations from Fig. 2a are displayed on UMAP plots, representing gene expression level. d Expression of top DEGs is represented in dot plots for NFκB + vs NFκB- cell populations based on NFκB pathway signature score, as in Fig. 2. e Expression of DEGs was determined by QPCR for sorted GFP + cells relative to GFP- cells from the MCF-7-NFκB-RE-GFP cell line treated with 4OHT for 2 weeks

Fig. 6

ISR pathway activation in the NFκB + cell population under selective pressure of ETs. a IPA network analysis for DEGs of the NFκB + cell population. Two central nodes were identified, p65/RelA (activation z-score = 2.017, P = 1.42E-05) and ATF4 (activation z score = 3.549, P = 1.04E-15). b, c FEA was performed to examine enrichment of ATF4 and ISR gene signatures across clusters. d Correlation between a hallmark NFκB signature and the ISR gene was examined for each cluster. Correlation was calculated using Pearson’s Correlation function in RStudio software (r = 0.56; P = 0.001). e Co-immunofluorescence for peIF2α and GFP was performed on MCF-7-NFκB-RE-GFP cells treated with different ETs for 2 weeks in clonogenic conditions. Scatter plot shows analysis of fluorescence intensity for peIF2α and GFP per cell in each treatment group. Intensity was measured by ImageJ software and Pearson’s correlation coefficients and P-values are shown on the graph. Scale bar: 50 µm. f Co-immunofluorescence for peIF2α and GFP was performed as described above. IKK7 (1 µM), an inhibitor of IKKα/β, was added for the last 72 h of culture. Histograms show fluorescence intensity of peIF2α vs. number of cells analyzed in each treatment groups. Scale bar: 50 µm. g The effect of an ISR inhibitor (ISRIB, 10 µM) was examined on growth of GM and 4OHT-treated cells after 2 weeks of treatment. *P < 0.05, ns = not significant