A poor prognosis gene transcription signature associated with PIK3CA mutation in ER+ breast cancer.

Loi S, Haibe-Kains B, Sotiriou C, Lallemand F, Piccart MJ, Wayne P and McArthur GA

Background: The phosphathidylinositol-3-kinase (PI3K) signaling pathway is commonly deregulated in breast cancer tumor biology. This pathway contains many potential therapeutic targets. However, the literature reporting the association between histopathological correlates and clinical outcome has been inconsistent probably due to the multiple mechanisms of activation and complex downstream interactions.

Methods: 173 estrogen receptor positive breast cancers (ER+BC )were sequenced for PIK3CA activating mutations. Corresponding gene expression profiles were available for 161 patients. Ingenuity Pathway Analysis software was used to generate interaction networks and test for enrichment of cellular functions between PIK3CA mutation positive and normal samples.   

Results: PIK3CA mutations were found in 45 samples (26%). The majority of mutations (80%) were found on exon 20. There was no significant association with mutation and lymph node status, tumor grade or age, and a borderline association with tumor size (p=0.056). No difference in mean gene expression level of AKT, PTEN, ERBB2, STMN between PIK3CA mutation positive samples and wild type was found. In contrast, TP53 and MYC levels were significantly different (p= 0.03, p= 0.01 respectively). There was no association between mutation status and clinical outcome (log rank p value =0.3). Contrasting gene expression profiles of mutation positive samples and wild type revealed 81 significantly differentially expressed genes (p<0.001) highly enriched for genes involved in cellular morphology, protein synthesis and gene expression. A gene-expression and interaction-based outcome predictor consisting of 35 genes was then constructed. This gene set could predict clinical outcome in 1228 independent ER+ BC samples HR: 1.39(95%CI 1.23-1.58) p<0.001.

Conclusions: PIK3CA mutations per se were not significantly associated with a distinct phenotype or clinical outcome in our dataset of 173 ER+ BC samples. However, we have identified important gene interaction networks related to PI3K pathway signaling due to PIK3CA mutations which may be useful for prognostic and predictive markers and therapeutic development in ER+ BC.