Efficiently searching for enhancers and their target genes in the human genome

Abstract

A single 3 billion letter genome contains the instructions for the 37 trillion diverse cells that make up one human. To accomplish this, the ~21,000 human genes are expressed and perform function in highly specific combinations per cell. Yet, only 2% of the genome codes for genes. The remaining 98% is made up of highly complicated, loosely patterned DNA referred to as “noncoding sequence”. Functional noncoding DNA elements (first termed “enhancers” in 1981) regulate cell-type specific gene expression. Like genes, enhancers disruption is known to cause genetic disease. How can we efficiently search for enhancers within the expansive noncoding genome? The new genome engineering technology CRISPR/Cas9 enables parallelized pooled perturbations to efficiently screen enhancers and the genes they target. In this dissertation, I will cover my development of new pooled methods to screen the noncoding genome. In the first chapter, I introduce the motivation for these methods, the history of enhancers, their current definitions, and emerging technologies for enhancers’ at-scale characterization. In Chapter 2, I describe a method we devised to scan thousands of CRISPR-induced kilobase-sized deletions ("ScanDel") across a desired noncoding region, programming one unique deletion per cell in a pool and phenotyping them in multiplex by pooled functional selection. However, ScanDel and its contemporaries are limited to evaluating enhancers for their effect upon a single gene. In Chapter 3, I describe a second method designed to overcome this limitation, in which large numbers of CRISPR perturbations are introduced to each cell, followed by single-cell transcriptome sequencing to read out their effect upon any transcript. With this method, we effectively evaluated >70,000 potential enhancer-target gene relationships in one experiment. In Chapter 4, I describe a potential path forward to cataloguing all enhancers in the human genome, and how we might do the same for noncoding variants in human disease.