Molecular profiling and quantitative image analysis reveal spatial intratumor heterogeneity in TNBC.

Abstract

e12536 Background: Triple negative breast cancer (TNBC) is a molecularly complex and heterogeneous subtype with distinct biological features and clinical behavior. Extensive intra-tumor heterogeneity is suspected to be a major cause of therapeutic failure. Spatially-distinct parts of the tumor harbor diverse and divergent clonal populations of cancer cells. Thus it is likely increasing the likelihood that some of those clones resist treatment, expand in numbers and eventually, repopulate the tumor leading to recurrence and spread. Therefore, a deeper understanding of this complexity is fundamental to gaining insights into this clinically important issue. In this study, we used RNA sequencing and machine learning approaches to examine the molecular and phenotypic/cellular profiles of various tumor samples from the same patient tumor to reveal spatial intra-tumor heterogeneity in TNBCs. Methods: We used 34 samples (2-4 samples from each patient tumor) from a total of 11 unique TNBC patients. RNA-sequencing was performed to quantify differential gene expression and machine learning (ML) approaches (deep learning regression modeling) were used to quantify the percentage of tumor and tumor-infiltrating lymphocytes (TILs) in the H&E stained tissue sections. The extent of concordance/discordance between the multiple tumor samples that originated from the same patient was analyzed by analyzing the intra- and inter-patient variance of normalized tumor cell, TIL % and gene expression. We also performed pathway analysis to identify signaling pathways dysregulated within (intra-tumoral heterogeneity) and between tumors (inter-tumoral heterogeneity) and performed molecular subtype analysis. Results: We observed that gene expression variance as higher within-patient (intra-tumor) compared to between-patient (inter-tumor). Tumor samples from 70% of patients showed different molecular subtypes representing extensive intra-tumor heterogeneity. Our ML-based image analysis showed that intra-patient tumor cell and TILs density/percentage variance was greater than inter-patient variance. In addition, patients with high within-patient gene expression variability had a high tumor and TIL variance. Among the within-patient expression variability, the genes associated with the PLK1 and Notch signaling were enriched. Conclusions: Our results suggest that TNBCs exhibit higher intra-tumor gene expression and cellular variance compared to inter-tumor gene expression and cellular variance suggesting higher intra-tumor heterogeneity in TNBCs.