Abstract P2-11-10: Intra-tumor heterogeneity affects multi-gene test prognostic risk stratification

Abstract

Abstract INTRODUCTION: Multigene tests (MGTs) have entered clinical practice and patient management. All MGTs are currently performed in reference laboratories using RNA isolated from a tumor section. The effect of intra-tumor heterogeneity on detection of genes within these MGTs, and the subsequent ability of to predict prognostic risk, has received little attention. To examine this directly, we used nCounter analysis to measure the expression of genes from five MGTs (OncotypeDx, Mammaprint, EndoPredict, PAM50, and Breast Cancer Index) in formalin fixed paraffin embedded tumor sections compared to cores taken from the tumor block. METHODS: We selected 71 estrogen receptor (ER) positive node-negative tumors, all of which had clinical OncotypeDx scoring performed at Genomic Health Inc, and for which ER, PR, HER2 and Ki67 clinical pathological measurements were available. For each patient we cut a 5uM section, and cut a 0.6mm core from a representative part of the tumor. If the tumor had an area of high focal Ki67 (n = 26), low PR (n = 13), or both (n = 5) cores were cut from these areas. In total we processed 181 samples. RNA was isolated using Qiagen RNAeasy kit, and 100ng used for nCounter analysis using a custom codeset incorporating five MGTs (n = 142 genes of which 12 are reference genes). MGTs were normalized to their respective reference genes, and nCounter OncotypeDx scores were scaled to the clinical OncotypeDx score for estimation of risk recurrence. RESULTS: Hierarchical clustering using all of the MGT genes combined (normalized to all reference genes) showed that the majority (61) of tumor samples clustered by patient, indicating greater inter than intra-tumor heterogeneity. However, when individual MGTs were examined alone, the intra-tumor heterogeneity increased. We found high correlation between Oncotype genes expression in the whole section versus representative tumor cores (r = 0.94). However, areas of low Ki67 and high PR staining showed low Ki67 and high PR gene expression, and consequently a slightly higher but not statistically significant recurrence score. For 17/75 (22.6%) patients, Oncotype Dx recurrence scores crossed the boundaries for low, intermediate and high risk. Conclusion: Inter-tumor heterogeneity is greater than intra-tumor heterogeneity for MGTs, and for the majority of patients a core of tumor gives gene expression measurements that are highly similar to a whole tumor section. This suggests that the tumor microenvironment (in our sample set mainly adipocytes) contribute little to the mRNA levels of MGTs. However, regions of high Ki67 or low PR within tumors have slightly higher OncotypeDx recurrence scores. The finding that recurrence scores differ by the tumor region that is sampled is consistent with other recent studies of intra-tumor genetic and transcriptomic heterogeneity, and indicates the potential clinical importance of these observations. Citation Information: Cancer Res 2013;73(24 Suppl): Abstract nr P2-11-10.