Studying Tumor Evolution with Emerging Sequencing Technologies

Abstract

The path along which heterogeneous subpopulations within a tumor arise can be modeled as an evolutionary process, and a better understanding of such process may improve cancer diagnostics, prognostics, and treatment outcome. How to construct more accurate tumor phylogenies and how to leverage the inferred phylogenetic relations among tumor subclones in understanding the disease are both key in the study of tumor evolution. The advances in novel sequencing technologies present opportunities to examine crucial aspects of the evolution of a tumor, beyond the types of information short-read genomic and transcriptomic sequencing technologies offer. As such, this work leverages three such technologies - single-cell bisulfite sequencing, long-read sequencing, and spatial transcriptomics - in answering key questions in tumor evolution.

In the first part of the thesis we present the first distance-based method to jointly construct a tumor phylogeny and identify lineage-informative CpG sites from single-cell methylation sequencing data. The tool provides a fast and accurate way to examine tumor evolution through the methylome at a single-cell resolution and allows the incorporation of matched RNA sequencing information whenever applicable.

Next, we present a novel framework to study both the genomic and epigenomic changes during tumor evolution using Nanopore long-read sequencing data of single-cell derived tumor sublines. Single nucleotide variants, structural variants, copy number alterations, as well as CpG methylation were profiled. The clonal resolution this data set uniquely affords allowed the inference of active mutational processes, confident identification of parallel CNAs, and detection of differentially methylated regions, all at a subclonal level.

Finally, we present a framework to examine tumor subclones at spatial resolution. Melanoma tumors from mice in both control and anti-CTLA-4 treatment group were subjected to both Smart-seq2 single-cell full length transcriptomics sequencing and 10x Visium spatial transcriptomics sequencing. Six tumor subclones were identified using mutations profiled from the Smart-seq2 data, and their proportion were estimated from the spatial transcriptomics slides via cell type deconvolution. The associations between tumor subclones with various stromal and immune cell types were accessed with established spatial statistics. By contrasting subclone-immune activities between responding and non-responding tumors, we show that there exist subclonal differences in their response to the immune microenvironment.