Gene expression and survival analysis of breast cancer sub types defined by immunohistochemistry

Abstract

Introduction: Breast cancer in the clinical setting may be divided into four subtypes (luminal A, luminal B, HER2 and basal) by immunohistochemical (IHC) analysis. This project aims to 1) compare subtype classification by IHC vs. the PAM50 gene classifier, 2) investigate the gene expression profiles of the IHC subtypes and 3) investigate the prognostic implications of the IHC subtypes. Method: The concordance between the IHC and PAM50 classifiers was assessed in a group of 405 breast cancers used in a previous study by Breffer et al. [1]. For RNA-seq analysis, data for a group of 105 cancers subtyped by IHC were obtained from The Cancer Genome Atlas Program (TCGA), National Cancer Institute (NCI) [2]. The data were analysed with limma and edgeR to obtain the gene expression profiles for the IHC subtypes. An exploratory unsupervised hierarchical cluster analysis was performed to help understand the usefulness of these gene sets in providing information on IHC cancer subtypes. Lastly, breast cancer specific overall survival among the IHC subtypes was assessed in a cohort of 292 cancers from the Garvan Insititute. Results: The concordance between subtyping by PAM50 and IHC among the 405 tumours was 59%. For cancers classified as HER2 by IHC (HER2 IHC cancers), 94% (16 out of 17) were also classified as HER2 by PAM50. For basal IHC cancers the concordance with PAM50 was 86% (54 out of 63). For luminal A and luminal B IHC cancers, the concordance with PAM50 was 59% (148 out of 251) and 28% (21 out of 74) respectively. Linear modeling of RNA-seq data showed basal IHC cancers were enriched for genes associated with tumour hypoxia, activation of the beta-catenin pathway, progression through the cell cycle and stem cell proliferation. Whereas as luminal A cancers showed down-regulation of genes associated with progression through the cell cycle and cell migration. An exploratory unsupervised hierarchical cluster analysis by the PAM50 gene set supported the usefulness of these genes in understanding the differences between the subtypes. Survival analysis using the Cox regression model showed, compared to luminal A IHC cancers, basal (HR = 4.08, 95% CI: 2.02 – 5.77), and HER2 (HR = 4.08, 95% CI: 2.14 – 7.78) cancers were associated with shorter breast cancer specific overall survival (p = 0.018). A tendency for poorer survival was seen for luminal B IHC cancers (HR = 2.04, 95% CI: 0.98 – 4.25) was observed. Conclusion: Breast cancer IHC subtypes have distinctive gene expression profiles that relate to their biological behaviour. Subtyping by IHC yields important prognostic information that may aid the clinical management of breast cancers.