Advancing the capabilities of base editing and protein delivery

Abstract

Genome editing technologies can in principle enable powerful new therapeutic strategies, including possible cures for deadly genetic disorders. Realizing the therapeutic potential of genome editing requires the ability to precisely correct disease-causing mutations in the relevant cell types without inducing unwanted off-target mutations or other negative consequences. In this thesis, I describe advances that increase the safety and therapeutic applicability of genome editing, including the development of high-fidelity base editors and improved methods for delivering genome editing agents into target cells within the body. In Chapter 1, I provide an overview of current genome editing technologies, including nucleases, base editors, and prime editors. I also introduce viral and non-viral methods for delivering genome editing agents into cells and compare the benefits and drawbacks of existing delivery modalities. In Chapter 2, I describe my work to characterize and minimize genome-wide off-target base editing. Base editors are precision genome editing agents that can precisely correct pathogenic single-nucleotide mutations in genomic DNA. However, the original base editors were known to randomly generate undesired off-target mutations throughout the genome, which could be problematic for therapeutic applications. Furthermore, due to their low frequency, these off-target mutations were challenging to detect using standard methods. We first developed new methods for detecting these rare off-target mutations that did not involve whole-genome sequencing and were therefore rapid, high-throughput, and cost-effective. Using these methods, we engineered new high-fidelity base editors that exhibited substantially minimized off-target editing compared to the original base editors. We also observed that off-target editing levels were reduced when delivering base editor ribonucleoprotein complexes into cells instead of base editor-encoding DNA, highlighting the benefits of transiently delivering genome editing agents. Collectively, our findings provide new methods for rapidly quantifying off-target base editing and illuminate several strategies for reducing off-target base editing. In Chapter 3, I describe my work to develop a new method for safely and efficiently delivering therapeutic proteins into cells using engineered virus-like particles (eVLPs). While protein delivery of genome editing agents was known to offer safety advantages compared to nucleic acid delivery by viral vectors, no general strategy for delivering genome editing proteins into cells within living animals (in vivo) had been previously reported. To address this unmet need, we developed eVLPs that efficiently package and deliver genome editing proteins instead of viral genomes. In eVLPs, genome editor cargo proteins are fused to retroviral gag capsid proteins, thereby directing the packaging of the desired cargos into virions that lack any viral genetic material. To unlock the potential of eVLPs to mediate efficient in vivo delivery, we systematically engineered eVLP architectures to improve cargo loading into the particles as well as cargo release into the target cells. The resulting eVLPs were now potent enough to enable efficient in vivo delivery of genome editing proteins into the mouse brain, eye, and liver. We demonstrated that our new protein delivery platform could be used to mediate therapeutic base editing in mice, including one-time treatments for genetic blindness and hypercholesterolemia. Compared to viral and non-viral delivery methods currently used in the clinic, our eVLP platform offers a more favorable safety profile without sacrificing delivery efficiencies. In Chapter 4, I describe my work to develop eVLPs with improved properties using directed evolution. The ability to perform screens or selections with large libraries of eVLPs would enable the systematic identification of eVLP variants with desired properties. However, because eVLPs lack viral genetic material, they cannot be evolved using existing methods for evolving viruses in the laboratory. I report a new, general scheme for evolving eVLPs with desired properties. In this scheme, the identity of a particular eVLP library member is uniquely encoded by a barcode sequence present on the guide RNA molecules that are loaded within that particular eVLP library member. Therefore, the identity of the eVLP variants that survive after selection can be determined by sequencing the barcodes of the guide RNAs present after selection. I applied this scheme to evolve eVLP capsid variants with improved properties. Initial screens for improved eVLP production and/or transduction revealed the effects of different capsid mutants on eVLP properties and identified capsid mutants that exhibit improved potency relative to previous-best eVLPs. This work establishes the barcoded eVLP scheme as a promising and generalizable platform for evolving eVLP variants with desired properties. I conclude by discussing perspectives on the future of genome editing and protein delivery modalities, including opportunities for future improvements and prospects for clinical translation.