Impact of intratumoral heterogeneity on drug responses: Unmasked by single-cell transcriptome analysis

Abstract

Introduction: Understanding distinct genomic signatures of a patients cancer is required to design and predict accurate therapeutic responses. Current approaches regarding heterogeneous cancer cells en masse as a pooled population, however, hardly reflect complete genomic landscape of tumor diversity, missing potentially important minor subclones implications such as metastasis and drug resistance.

Methods: To dissect intratumoral heterogeneity and discover unique patterns of subclonal behaviors against drug treatment responses in functional modalities and signaling pathways out of heterogeneous population derived from a cancer patient, tumor transcriptome was characterized at single-cell resolution by utilizing single-cell RNA sequencing (scRNA-seq).

Results: First, in a lung adenocarcinoma PDX model*, individual cells showed mosaic expression of SNVs and differential gene expression. We could cluster the PDX cells into three distinct groups according to the presence of KRAS G12D mutation and transcriptome-based risk scores (RS). A single cell group with KRAS G12D/high RS was activated in the RAS-MAPK signaling pathway, and targeted by anti-cancer drugs such as docetaxel and BKM120. In comparison, a KRAS G12D/low RS group showed inactive RAS-MAPK signaling despite the expression of KRAS G12D. The drug-resistant population recapitulated gene expression signatures of the KRAS G12D/ low-RS group. Second, in a metastatic renal cell carcinoma PDX model, enriched subclonality was identified in a metastasis tumor with activated expression signatures of epithelial-mesenchymal transition and poor prognosis. Integrated analysis of transcriptome profiling and drug screening identified the most effective anti-cancer drugs. Furthermore, singularity of single cells with mutually exclusive activation of EGFR and SRC signaling pathways could suggest the potential of combination therapy, and its efficacy was validated in 2D and 3D in vitro and in vivo models.

Conclusions: Patient-derived xenograft cells showed heterogeneous profiles in SNVs and gene expression, which could cluster them into subclones with differential responses to anti-cancer drugs. Taken together, our approach of single-cell RNA sequencing on tumors provides insights into the more accurate strategy to identify minor but potentially important subclones that are relevant to drug resistance.