DMBA induced mouse mammary tumors display high incidence of activating Pik3caH1047 and loss of function Pten mutations

Abstract

Controversy always existed on the utility of chemically induced mouse mammary carcinogenesis models as valid equivalents for the study of human breast cancer. Here, we performed whole exome and RNA sequencing on long latency mammary tumors (218 ± 27 days) induced by the carcinogen 7,12-Dimethylbenzathracene (DMBA) and short latency tumors (65 ± 11 days) induced by the progestin Medroxyprogesterone Acetate (MPA) plus DMBA in CD2F1 mice. Long latency tumors displayed a high frequency of Pi3kca and/or Pten mutations detected in 11 of 13 (85%) long latency cases (14/22, 64% overall). Eighty-two percent (9/11) of tumors carried the Pik3ca H1047L/R hot-spot mutation, as frequently found in human breast cancer. These tumors were luminal-like and mostly ER/PR+, as in humans. Transcriptome profiling indicated a significant activation of the PI3K-Akt pathway (p=3.82e-6). On the other hand MPA+DMBA induced short latency tumors displayed mutations in cancer drivers not commonly found mutated in human breast cancer (e.g. Hras and Apc). These tumors were mostly basal-like and MPA exposure led to Rankl overexpression (60 fold induction) and immunosuppressive gene expression signatures. In summary, long latency DMBA induced mouse mammary tumors reproduce the molecular profile of human luminal breast carcinomas representing an excellent preclinical model for the testing of PIK3CA/Akt/mTOR pathway inhibitory therapies and a good platform for the developing of additional preclinical tools such as syngeneic transplants in immunocompetent hosts.

Keywords: DMBA; MPA; Pik3ca; Pten; mammary tumors.

Figure 1. Mutational profiles of chemically induced mouse mammary tumors

A. Mutational signature of DMBA alone and MPA+DMBA combined treatment. The mutational signature is displayed using a 96 substitution classification defined by the substitution class and the sequence context immediately 5′ and 3′ to the mutated base. The probability bars for each of the six types of substitutions as well as the mutated bases are displayed in different colors. Y-axis represent the percentage of mutations attributed to a specific mutation type corrected by the frequency of its corresponding trinucleotide across the mouse genome. B. Comparison of mutation rates per Mgbp among treatments. Note the much higher mutation rate found in MPA+Dx4 generated tumors. C. Correlation analysis of mutations per Mgbp and tumor latency in days among tumors. All but one of the short latency tumors were generated by the MPA+Dx4 protocol (red dots).

Figure 2. Cancer driver gene mutations in chemically induced mouse mammary tumors

A. Driver and co-driver mutations in tumor samples. Each column represents one tumor case and each row represents the number of mutations in each gene per tumor. Blue squares, + mutation. B. Distribution of the most prevalent driver/co-driver mutations among short (black bars) and long latency (gray bars) chemically induced tumors. C. The CoMEt test was employed to identify mutually exclusive events between Pik3ca/Pten and Hras mutations across the 22 tumors analyzed. As can be observed Pi3ca/Pten mutations frequently co-occurred in the same tumors and were mostly mutually exclusive with Hras mutations.

Figure 3. Hierarchical clustering of normal and tumor samples based on RNA-Seq profiles

A. The first row represent tumors derived from DMBA (gray boxes) or MPA+DMBA combine protocols (black boxes). The second row represents long (gray boxes) and short latency (black boxes) tumors. The third and fourth rows represent tumors that carry Hras and Pik3ca mutations (black boxes). The fifth row indicates the ER and PR positive (black boxes) and negative (gray boxes) cases according to immunohistochemistry results. Tumor intrinsic subtypes predicted by the PAM50 gene model are highlighted in color codes in the last row. B. Intrinsic tumor subtypes according to latency.

Figure 4. Representative validation of estrogen receptor alpha (ER), progesterone receptor (PR) and luminal and basal cytokeratins (CKs) in mouse mammary tumors

Panels A-D. representative immunohistochemistry results for long latency (254 ds) tumor T12800. Carcinoma with acinar/glandular differentiation displaying strong ER and PR positivity (brown nuclei in A and B respectively), Strongly positive for luminal CK8 (C) and residual expression in myoepithelial like cells for CK5 (D). This tumor displayed mutations in Pik3ca and Pten driver genes as per Exome-Seq analyses. Panels E-F. short latency (60 ds) tumor T12513. This luminal carcinoma displayed mostly glandular differentiation, strong reactivity for ER, PR and CK8 (E-G. respectively) and negative for CK5 H. Exome-Seq for this case revealed mutations on Hras and Atm among other potential cancer driver genes. Panels I-L. Short latency (55 ds) adenosquamous carcinoma T12521, negative for ER (I) and PR (not shown) expression and with mixed expression of luminal (CK8) and basal cytokeratins CK6/CK5 that strongly stained the squamous differentiated regions.

Figure 5. Transcriptomic changes induced by medroxyprogesterone in normal mammary gland and mammary tumors

A. Venn diagram illustrating the identification of a MPA 26-gene expression signature commonly deregulated in normal and tumors samples. B. Functional Gene Ontology enrichment analysis of the MPA-26 gene expression signature. C. Box plot of Rankl mRNA expression levels comparing normal and chemically induced mouse mammary tumors by the various treatment protocols as indicated at the bottom of D. Note the dramatic increase in Rankl expression in normal and tumor samples exposed to MPA treatment. D. Comparative immune scores as determined by the ESTIMATE algorithm of normal and tumor samples of various carcinogenesis protocols as indicated. As can be observed carcinogenesis protocols employing MPA treatment (M+Dx4 and M+Dx2) displayed significantly lower immune scores compared with DMBA alone and normal samples (p<0.001).

Figure 6. Comparative analysis of the frequency of Pik3Ca/PIK3CA hot-spot mutations among chemically induced mouse mammary tumors and human breast carcinomas

PIK3CA mutational profile in primary breast carcinomas was obtained from the TCGA-BRCA project [24].

Figure 7. Analysis of differentially expressed transcript in DMBA-induced mouse mammary tumors

A. Functional enrichment analysis of transcripts differentially expressed among DMBA-induced tumors and normal samples (p-adj < 0.001; Log2FC > 1). B. Heat map of 98 PI3K/AKT related genes differentially expressed between DMBA-induced tumors and normal samples. C. PI3K/AKT pathway diagram highlighting some of the genes mutated (stars) and up-modulated (colored boxes) in DMBA-induced mouse mammary tumors.