Deep mutational scanning of mammalian loci using CRISPR-Cas9 and multiplex HDR

Abstract

Functional consequences of genetic variants are best studied in their endogenous chromosomal context. Gene editing by homology-directed repair can introduce such predetermined genetic changes into chromosomal DNA. In this thesis, I develop methods to generate tens to hundreds of genetic variants, expressed from a native chromosomal context, and simultaneously evaluate their phenotypic impact. This approach involves repair of Cas9-derived double strand breaks (DSBs) from oligonucleotide repair template libraries containing controlled levels of nucleotide heterogeneity. Cell populations are then purified based on a phenotypic assay and subjected to deep amplicon sequencing at the target site to link genotype with phenotype. In the first chapter, I developed a bioinformatics pipeline for the processing of Illumina sequencing reads containing nucleotide variants, and validate this pipeline in silico. As a proof-of-principle, in the second chapter I then introduced nucleotide variants across 8 codons of a chromosomal GFP transgene in mouse embryonic stem cells. The functional impact of these variants was quantified, with the results benchmarked against an existing episomal dataset, and by in silico modelling of mutant protein structure. In the final chapter, I applied this pipeline to analyse a CRISPR deep mutational scanning dataset incorporating all possible amino acid substitutions within a region of β-catenin, a component of the Wnt signalling pathway, that is a mutational hotspot in many types of cancer. The functional impact of these clinically relevant variants was assessed using a fluorescent reporter of Wnt signalling. By combining the resulting functional scores with mutational signature data from genome sequencing of different tumour types, I finally dissect the relative contribution of mutational bias and natural selection to the different patterns of amino acid substitutions found in different tumour types.