ATF3-Induced Mammary Tumors Exhibit Molecular Features of Human Basal-Like Breast Cancer

Abstract

Basal-like breast cancer (BLBC) is an aggressive and deadly subtype of human breast cancer that is highly metastatic, displays stem-cell like features, and has limited treatment options. Therefore, developing and characterizing preclinical mouse models with tumors that resemble BLBC is important for human therapeutic development. ATF3 is a potent oncogene that is aberrantly expressed in most human breast cancers. In the BK5.ATF3 mouse model, overexpression of ATF3 in the basal epithelial cells of the mammary gland produces tumors that are characterized by activation of the Wnt/β-catenin signaling pathway. Here, we used RNA-Seq and microRNA (miRNA) microarrays to better define the molecular features of BK5.ATF3-derived mammary tumors. These analyses showed that these tumors share many characteristics of human BLBC including reduced expression of Rb1, Esr1, and Pgr and increased expression of Erbb2, Egfr, and the genes encoding keratins 5, 6, and 17. An analysis of miRNA expression revealed reduced levels of Mir145 and Mir143, leading to the upregulation of their target genes including both the pluripotency factors Klf4 and Sox2 as well as the cancer stem-cell-related gene Kras. Finally, we show through knock-down experiments that ATF3 may directly modulate MIR145/143 expression. Taken together, our results indicate that the ATF3 mouse mammary tumor model could provide a powerful model to define the molecular mechanisms leading to BLBC, identify the factors that contribute to its aggressiveness, and, ultimately, discover specific genes and gene networks for therapeutic targeting.

Keywords: ATF3; Mir143; Mir145; RNA-Seq; basal-like; breast cancer; miRNA; mouse model.

Figure 1

Gene expression patterns of ATF3-induced tumors are consistent with those of human basal-like breast cancer (BLBC). (A) Unsupervised hierarchical clustering and heatmap presentation of all DEGs were generated from three pairs of matched, adjacent, normal mammary glands (M) and mammary tumors (T) from BK5.ATF3 transgenic mice using the hclust and heatmap2 functions in R with gene expression values calculated using edgeR. Red represents upregulated genes and green represents downregulated genes. The expression values are log2 fold changes between tumor and adjacent normal tissues (fold change ≥ 2.0, FDR ≤ 0.05). (B) Dot plots showing the difference in expression of selected genes in BK5.ATF3 mice for both normal tissue (M) and tumor tissue (T). *** indicates a significance of FDR < 0.005. (C) Unsupervised hierarchical clustering and heatmap presentation comparing expression of 3979 differentially expressed genes (DEGs) identified in the BK5.ATF3 mammary tumors and their human homologs. The expression profiles for the 3979 genes in 547 human breast cancer samples representing Basal-like, LuminalA, Luminal B, Her2, and Normal-like subtypes were compared with the mouse DEGs. The colored bar across the top indicates the subtypes of the patient samples depicted in the figure key: Basal-like (red), LumA (blue), LumB (green), Her2 (purple), and Normal (yellow). The horizontal dendogram on the top represents the 547 patient samples, and the vertical dendogram on the left represents 3979 DEGs.

Figure 2

Signaling pathway signatures of ATF3-induced tumors are similar to those of human BLBC. (A) GSEA enriched gene sets cell cycle (a), p53 (b), EMT (c), TGF-beta (d), and Hedgehog signaling pathway (e). (B) GSEA enriched gene sets Wnt pathways (a), canonical Wnt pathway (b), and non-canonical Wnt pathway (c). (C) Heatmaps showing the expression of all Wnt pathway genes in BK5.ATF3 mice (a), expression of canonical Wnt genes in mouse tumors (T11035, T11640, T9232) and matched non-tumor, transgenic mammary gland tissue (M11035, M11640, M9232) (b), and expression of non-canonical Wnt genes, as in b (c). (D) Heatmap showing the expression of genes in common between the BK5.ATF3 mouse model and the GSEA-identified enriched genes from human basal-like breast cancer (BLBC) sample data contained in TCGA using the human biological process gene sets in Gene Ontology. For (A) and (B), the GSEA analysis was performed using GSEAPreranked (Subramanian, Tamayo et al. 2005) against gene sets from V5.1 of the Molecular Signatures Database (MSigDB). Gene pre-ranking was based on the p-value obtained from the edgeR analysis. For (A), gene sets are from V5.1 of the Molecular Signatures Database (MSigDB). For (B), the Wnt pathway gene set (a) was compiled based on GO Ontology, DAVID, the Wnt Homepage, and the scientific literature. The Wnt pathway gene set (a) was sub-categorized into canonical Wnt (b) and non-canonical Wnt (c) gene sets.

Figure 3

Comparison of BK5-ATF3 tumor data based on the expression of five intrinsic gene sets used to characterize murine models of breast cancer [22]. (A) Boxplot of expression log2ratio values for each mouse gene subtype signature (proliferation, basal-like, lactating, luminal, and claudin-low). The log2ratio values were from RNA-Seq and scaled so that the sum of squares in each subtype is 1.0. Only the proliferation and basal gene sets had average log2 ratios significantly different than zero (p < 0.05). (B–E) Heatmaps depicting four of the five gene sets. (B) Proliferation. (C) Basal-like. (D) Lactating. (E) Luminal-like. For the boxplots, all the genes of each subtype were used. For the heatmap, the significantly differential genes of each subtype in RNA-Seq were used.

Figure 4

Identification and validation of differentially expressed miRNA genes. (A) Heatmap of 34 significantly differentially expressed miRNA genes between four pairs of matched tumors (T) and their adjacent mammary tissues (M) from ATF3-expressing mammary glands. (B) Correlation of gene expression between miRNA genes identified by a microarray and selected for validation with RT-qPCR assays using RNA from a total of six tumors and six paired, normal, mammary gland controls, including the four pairs used for the microarray experiments. (C) IPA analysis of the integrated mRNA and miRNA datasets.

Figure 5

Expression of miR-145 and miR-143 and their downstream, stem cell-related gene targets in normal, transgenic mammary tissues (white) and ATF3-induced tumors (black). (A) Relative expression of miR-145 and the pluripotency factors Klf4 and Sox2. (B) Relative expression of miR-143 and its target gene Kras. ** p < 0.05. *** p < 0.005.

Figure 6

Effects of ATF3 knockdown on the expression of selected genes in the TNBC cell line MDA-MB-157 determined by RT-qPCR. (A) Plot showing the relative expression of ATF3 in the presence of either a scrambled shRNA (white), or an ATF3 shRNA (black). (B) As in A, but for miR-145. (C) As in A, but for miR-143. (D) As in A but for ATF3, JUN, TCF7, WNT3A, KLF4, SOX2, OCT4, MYC, and KRAS. ** p < 0.05, *** p < 0.005.