Engineering CRISPR-Associated Transposases for Programmable DNA Integration in Human Cells

Abstract

Programmable large DNA integration in human cells has the potential to enable mutation-agnostic gene therapies for loss-of-function diseases and accelerate the development of next-generation cell therapies. However, existing large DNA integration technologies remain limited by deleterious double-strand breaks, restricted cargo capacity, off-target integration, or extensive target-specific optimization that limits their programmability. CRISPR-associated transposases (CASTs), naturally occurring bacterial RNA-guided DNA integrases, overcome many of these limitations by enabling programmable, site-specific DNA integration with minimal byproducts while supporting integration of multi-kilobase DNA cargoes. Despite high activity in bacterial cells, wild-type CAST systems exhibit minimal activity in human cells. In this thesis, I describe the development of a laboratory-evolved and engineered CAST system for efficient programmable DNA integration in human cells through the complementary application of directed evolution, protein engineering, and high-throughput screening. In the first part of this thesis, I describe the development of a human-active CAST system. We initially evolved each protein component using phage-assisted continuous evolution (PACE). While repeated evolution of the QCascade DNA-targeting complex failed to improve mammalian DNA integration because bacterial selection pressures did not adequately enrich for variants with improved mammalian activity, evolution of the transposition machinery improved integration efficiencies by several orders of magnitude. Combined with structure-guided engineering of the QCascade complex, we generated evoCAST, a laboratory-evolved CAST system that enhances programmable DNA integration in human cells by more than 200-fold. EvoCAST mediates efficient integration of multi-kilobase DNA cargoes across diverse genomic target sites while retaining the favorable characteristics of wild-type CAST systems, including high programmability, minimal byproducts, and low levels of off-target integration. While evoCAST enables efficient, site-specific integration of kilobase-sized DNA cargoes in human cells, integration efficiencies differ considerably across spacer sequences, and activity remains variable across clinically relevant cellular contexts. These bottlenecks can in part be attributed to an incomplete understanding of the molecular determinants governing crRNA function, target site selection, and transposon end architecture. In the second part of my thesis, I describe the development of pooled, large-scale screening platforms to systematically dissect these nucleic acid components and establish general design principles governing programmable DNA integration by evoCAST. Large-scale spacer screening identified predictive rules for spacer design and enabled the development of a machine learning model that accurately predicts and ranks evoCAST spacer activity. Systematic analysis of CRISPR repeat sequences revealed extensive mutational flexibility and defined a truncated guide RNA that facilitates evoCAST applications while preserving full activity. Library screening of transposon end architectures identified compact, highly active transposon ends incorporating optimized TnsB binding site substitutions, tolerated deletions, and preferred 5-bp transposon flanking sequences. These minimized transposon ends are 62 bp shorter than the canonical architecture, reducing the size of the post-transposition scar and the potentially cytotoxic double-stranded DNA component, streamlining evoCAST applications. Finally, combining optimized crRNA and transposon end architectures yielded an engineered donor that consistently improved integration efficiencies and consistency across multiple genomic loci and human cell types, including HEK293T cells, HuH7 cells, and primary fibroblasts from three patients. Together, these studies establish generalizable design principles governing both the protein and nucleic acid determinants of programmable DNA integration. By integrating laboratory evolution, rational engineering, and large-scale functional screening, this work establishes a versatile and scalable framework for engineering increasingly efficient CRISPR-associated transposases for programmable genome integration in human cells.