A tumour suppressive relationship between mineralocorticoid and retinoic acid receptors activates a transcriptional program consistent with a reverse Warburg effect in breast cancer

Abstract

Background: The role of nuclear receptors in both the aetiology and treatment of breast cancer is exemplified by the use of the oestrogen receptor (ER) as a prognostic marker and treatment target. Treatments targeting the oestrogen signalling pathway are initially highly effective for most patients. However, for the breast cancers that fail to respond, or become resistant, to current endocrine treatments, the long-term outlook is poor. ER is a member of the nuclear receptor superfamily, comprising 48 members in the human, many of which are expressed in the breast and could be used as alternative targets in cases where current treatments are ineffective.

Methods: We used sparse canonical correlation analysis to interrogate potential novel nuclear receptor expression relationships in normal breast and breast cancer. These were further explored using whole transcriptome profiling in breast cancer cells after combinations of ligand treatments.

Results: Using this approach, we discovered a tumour suppressive relationship between the mineralocorticoid receptor (MR) and retinoic acid receptors (RAR), in particular RARβ. Expression profiling of MR expressing breast cancer cells revealed that mineralocorticoid and retinoid co-treatment activated an expression program consistent with a reverse Warburg effect and growth inhibition, which was not observed with either ligand alone. Moreover, high expression of both MR and RARB was associated with improved breast cancer-specific survival.

Conclusion: Our study reveals a previously unknown relationship between MR and RAR in the breast, which is dependent on menopausal state and altered in malignancy. This finding identifies potential new targets for the treatment of breast cancers that are refractory to existing therapeutic options.

Keywords: Breast cancer; Gene expression profiling; Mineralocorticoid receptor; Retinoic acid receptor; Warburg effect.

Fig. 1

SCCA analysis of NR and coregulator expression in normal and malignant breast. Sparse Canonical Correlation Analysis identified strong inter-correlation between a set of nuclear receptors that include MR and 3 retinoic acid receptors and a set of coregulators in pre-menopausal normal. This strong inter-correlation is decreased in post-menopausal normal, pre-menopausal cancer and post-menopausal cancer. Heatmap colours represent pairwise Spearman rank correlations between each gene pair in the different sample cohorts of the TLDA dataset

Fig. 2

Co-expression-based identification of potential functional crosstalk between MR and RARβ. a Percentage of MR co-expressed genes shared with other nuclear receptors based on RNA-Seq expression profiles of normal and breast cancer tissues from the TCGA BRCA dataset. b Expression of MR, RARB, RARG and RARA in normal breast and breast cancer samples of the TCGA RNA-Seq dataset of 106 normal and 988 breast cancer samples. c Venn diagrams showing the overlap between co-expressed genes identified in normal and cancer samples for MR and RARB

Fig. 3

Differentially expressed genes in ALDO-, RA- and ALDO+RA-treated MCF-7 cells. Gene expression profiling of MR-inducible MCF-7 breast cancer cells treated for 6 h with aldosterone (10 nM), all-trans retinoic acid (1 μM) and 17ß-estradiol (10 nM) or their combination and profiled on Illumina’s HT-12 gene expression bead arrays. a MR mRNA expression, measured by RT-qPCR, uninduced and induced with doxycycline. b MR protein expression uninduced and induced with doxycycline. c Numbers of genes up- or downregulated in each treatment condition. d Venn diagram of the gene overlaps between treatment conditions. e Heatmap of the log2 fold change of genes differentially expressed in each treatment condition. Genes are grouped into different gene clusters and annotated with whether the genes are annotated to be involved in the glycolysis or oxidative phosphorylation hallmarks according to MSigDB. f Heatmap presentation of Biological Processes GO terms significantly enriched (enrichment P value ≤ 0.005 and FDR < 10%) in gene clusters annotated in e, only gene clusters with significantly overrepresented GO terms are included in this heatmap. Cell colours represent the enrichment P values. Grey indicates no significant GO term was identified

Fig. 4

Effect of ALDO+RA treatment on cellular metabolism. a The fold change in expression on ALDO+RA treatment in MR-inducible MCF-7 cells for differentially expressed metabolic genes. b Schematic diagram of the glycolysis pathways (adapted from [31]), highlighting in red the key genes that show differential expression in the ALDO+RA treatment. c Heatmap of the expression level of key genes involved in glucose metabolism in normal samples, cancer samples, cancer samples expressing high level of both MR and RARB and cancer samples expressing low level of both MR and RARB from the TCGA breast cancer RNA-Seq dataset

Fig. 5

Effect of MR and RAR signalling on breast cancer cell proliferation. The meta-PCNA genes which are the top 1% of genes most highly correlated with the proliferating cell nuclear antigen (PCNA), a well-known proliferation marker is used to calculate the meta-PCNA index which is defined as the median expression of meta-PCNA genes. a PCNA index as a proliferation measure for MR-inducible cells under different treatment conditions. Here, Dox is used to induce MR; therefore, Dox alone is equivalent to vehicle treatment of MR containing cells. P values are based on paired Wilcoxon signed-rank test of the expression of meta-PCNA genes for each treatment comparison of interest. b Expression of meta-PCNA genes under different treatment conditions of MR-inducible MCF-7 breast cancer cells

Fig. 6

Survival analysis of breast cancer tissues expressing high MR and RARB. a Scatter plot of expression of MR against RARB in breast cancer samples from the METABRIC dataset. A receptor is classified as highly or poorly expressed in a sample if its expression in that sample is in the top or bottom 20% of its expression range respectively. b Kaplan-Meier analysis showing samples expressing a high level of both MR and RARB have better survival outcome. c All breast cancer samples from the METABRIC breast cancer microarray dataset were classified into four categories: double-positive cases express both MR and RARB based on probe expression cut-off defined in Fig. S3 (blue bars); MR_only cases express MR but not RARB (yellow bars); RARB_only cases express RARB but not MR (red bars); and not_expressed cases do not express MR or RARB (grey bars). The stacked column plot shows the percentage of each of the four categories within each PAM50 subtype from the METABRIC dataset. Here the classification of each breast cancer sample into PAM50 subtype is as defined by the METABRIC publication. d Kaplan-Meier plot for each of the four categories of MR and RARB expression, for all breast cancer samples with survival data in the METABRIC dataset, sub-divided by PAM50 subtype. Due to having only one sample categorized as expressing only RARB, there is no survival curve for this category in the normal-like subtype

Fig. 7

Features of the reverse Warburg effect in oxidative cancer cells and cancer-associated fibroblasts. The features of the reverse Warburg effect are shown: epithelial cancer cells induce the Warburg effect (aerobic glycolysis) in neighbouring tumour-associated fibroblasts while proliferating cancer cells themselves show a preference towards oxidative phosphorylation