MicroRNA expression and gene regulation drive breast cancer progression and metastasis in PyMT mice

Abstract

Background: MicroRNAs (miRNAs) are small non-coding RNA molecules of about 22 nucleotides which function to silence the expression of their target genes. Numerous studies have shown that miRNAs are not only key regulators in important cellular processes but are also drivers in the development of many diseases, especially cancer. Estrogen receptor positive luminal B is the second most common but the least studied subtype of breast cancer. Only a few studies have examined the expression profiles of miRNAs in luminal B breast cancer, and their regulatory roles in cancer progression have yet to be investigated.

Methods: In this study, using polyoma middle T antigen (PyMT) mice, a widely used luminal B breast cancer model, we profiled microRNA (miRNA) expression at four time points that represent different key developmental stages of cancer progression. We considered the expression of both miRNAs and messenger RNAs (mRNAs) at these time points to improve the identification of regulatory targets of miRNAs. By combining gene functional and pathway annotation with miRNA-mRNA interactions, we created a PyMT-specific tripartite miRNA-mRNA-pathway network and identified novel functional regulatory programs (FRPs).

Results: We identified 151 differentially expressed miRNAs with a strict dual nature of either upregulation or downregulation during the whole course of disease progression. Among 82 newly discovered breast-cancer-related miRNAs, 35 can potentially regulate 271 protein-coding genes based on their sequence complementarity and expression profiles. We also identified miRNA-mRNA regulatory modules driving specific cancer-related biological processes.

Conclusions: In this study we profiled the expression of miRNAs during breast cancer progression in the PyMT mouse model. By integrating miRNA and mRNA expression profiles, we identified differentially expressed miRNAs and their target genes involved in several hallmarks of cancer. We applied a novel clustering method to an annotated miRNA-mRNA regulatory network and identified network modules involved in specific cancer-related biological processes.

Keywords: Breast cancer; Cancer progression; Metastasis; MicroRNA; PyMT mouse model; Regulatory modules.

Fig. 1

MicroRNA (miRNA) sequencing data analysis. a Identification of stage-specific differentially expressed miRNA-messenger RNA (mRNA) regulatory network. At each time point, for a differentially expressed miRNA and one of its mRNA targets predicted based on their sequence complementarity from the m3RNA database, we calculated the correlation between their expression levels (reads per kilobase million (RPKM)) from all 24 samples. Any miRNA-mRNA interaction with positive correlation was filtered out. Thus, only miRNA-mRNA pairs with opposite differential expression were included in the regulatory network for each cancer developmental stage. b Identification of overall transition pattern-specific miRNA-mRNA regulatory network. We classified miRNAs into 27 groups based on the overall expression patterns of the three consecutive stage transitions. In each group, only mRNA-miRNA pairs with opposite transition patterns were considered. c Identification of miRNA regulatory modules. We first annotated genes in a miRNA-mRNA regulatory network with functional and structural terms from Gene Ontology Biological Process, Kyoto Encyclopedia of genes and genomes (KEGG) pathways, Panther pathways, and Interpro domains. Using a maximal biclique analysis followed by bi-clustering, we then identified sets of coherently related genes and annotation terms. The miRNA regulatory modules were formed by adding miRNAs to corresponding sets that they potentially regulate

Fig. 2

Transition patterns of microRNA (miRNA) expression during cancer progression. Each gray line, re-based to 0 at week 6, shows the pseudo fold-changes (ϕFCs) of a particular miRNA at the four assayed time points. Expression change during each stage transition was discretized into increase (+), no change (0), or decrease (–). All miRNAs with the same transition pattern, e.g., p–00, are plotted together in one plot, in which the transition pattern and the number of miRNAs are given at the top. The red line in each plot indicates the median ϕFCs. Only positive and negative patterns without opposite changes are shown here. See Additional file 1: Figure S2 for the whole set of 27 transition patterns

Fig. 3

MicroRNA (MiRNA) expression analysis and miRNA-messenger RNA (mRNA) interaction validation. a MiRNA expression profiles during breast cancer (BRCA) progression in PyMT mice. The hierarchical clustering of miRNAs was carried out with the Ward method and the Euclidean distances of their expression profiles. The red dots indicate the miRNAs already present in another breast cancer study. b Comparison of PyMT miRNA-mRNA interactions with their human cancer orthologs. From the CancerMiner database for each human orthologous miRNA-mRNA pair, we calculated the average of recurrent scores for all cancer types and obtained the specific score for breast cancer and the minimum score for any cancer type. The more negative the score, the more reliable the validation of the interaction. Therefore, 0 was used as the threshold to differentiate false and true positives. c Score distributions of human orthologs of PyMT miRNA-mRNA interactions in CancerMiner. The negative values indicate the interactions are validated by all three different datasets

Fig. 4

Functional enrichment analysis. The statistically enriched Gene Ontology (GO) terms about biological processes were identified among genes potentially regulated by differentially expressed microRNAs (miRNAs) at each assayed time point. The size of the dot indicates the number of genes with the annotation of the term. Its color indicates the statistical significance of the enrichment. Its absence indicates the annotation is not enriched at the given time point. Biological terms appear in an alphabetical order in the figure. The full enrichment analysis results are included in the supplementary material (Additional file 2)

Fig. 5

MicroRNA (MiRNA) gene targets and hallmarks of cancer. Enriched functional and pathway terms were combined according to the hallmarks of cancer [38]. The heatmap indicates the number of miRNA-regulated genes annotated to each hallmark at each time point. An additional category, 'Neural genetics', was added to combine biological terms related to neural processes and disorders. Red indicates high transcriptional activity for the corresponding hallmark and time point. The heatmap suggests that activation invasion and metastasis, sustaining proliferative signaling, and evading growth suppressors are more active, especially in the transition to the carcinoma stages

Fig. 6

Regulation of fatty acid metabolism. a Expression profiles along all breast cancer subtypes in The Cancer Genome Atlas dataset of OLR1, FAAH, SNCA, ME1, PPARA and PPARG genes. FC fold change (expression in tumor samples divided by that in normal samples). b Recurrent regulatory network of fatty acid metabolism genes (c) Expression profiles of fatty acid metabolism genes and miRNA regulators in the PyMT mouse model. Her2 human epidermal growth factor receptor 2, LumA and LumB luminal A and luminal B