Identification of candidate mediators of chemoresponse in breast cancer through therapy-driven selection of somatic variants

Abstract

Purpose: More than a third of primary breast cancer patients are treated with cytotoxic chemotherapy, typically without guidance from predictive markers. Increased use of neoadjuvant chemotherapy provides opportunities for identification of molecules associated with treatment response, by comparing matched tumour samples before and after therapy. Our hypothesis was that somatic variants of increased prevalence after therapy promote resistance, while variants with reduced prevalence cause sensitivity.

Methods: We performed systematic analyses of matched pairs of cancer exomes from primary oestrogen receptor-positive/HER2-negative breast cancers (n = 6) treated with neoadjuvant epirubicin/cyclophosphamide. We identified candidate genes as mediators of chemotherapy response by consistent subclonal changes in somatic variant prevalence through therapy, predicted variant impact on gene function, and enrichment of specific functional pathways. Influence of candidate genes on breast cancer outcome was tested using publicly available breast cancer expression data (n = 1903).

Results: We identified 14 genes as the strongest candidate mediators of chemoresponse: TCHH, MUC17, ARAP2, FLG2, ABL1, CENPF, COL6A3, DMBT1, ITGA7, PLXNA1, S100PBP, SYNE1, ZFHX4, and CACNA1C. Genes contained somatic variants showing prevalence changes in up to 4 patients, with up to 3 being predicted as damaging. Genes coding for extra-cellular matrix components or related signalling pathways were significantly over-represented among variants showing prevalence changes. Expression of 5 genes (TCHH, ABL1, CENPF, S100PBP, and ZFHX4) was significantly associated with patient survival.

Conclusions: Genomic analysis of paired pre- and post-therapy samples resulting from neoadjuvant therapy provides a powerful method for identification of mediators of response. Genes we identified should be assessed as predictive markers or targets in chemo-sensitization.

Keywords: Chemoresistance; Exome sequencing; Sensitization; Somatic variants.

Fig. 1

Expression of candidate genes correlated with breast cancer outcomes. Expression levels of candidate genes in Table 2 were analysed for correlations with survival from breast cancer using the METABRIC dataset [19], by comparing the distribution of levels between patients who died of their cancer to those that did not using ‘violin’ plots (left of each pair), and by Kaplan–Meier analyses after expression was dichotomized using receiver operator curve analyses into low and high groups (right of each pair). For violin plots, median and quartiles are shown (horizontal lines) and significance was tested using 2-tailed Mann–Whitney U tests. For Kaplan–Meier analyses, significance was tested using log rank tests. Significant correlations only are shown; PLXNA1 and SYNE1 were significant in only the first analysis