Advancing and Applying Precision Genome Editing Technologies

Abstract

Since the discovery of the chemical structure of DNA, researchers have endeavored to understand how the chemical properties of this complex molecule provide a blueprint for biological function. The ability to modulate cellular genomic DNA has afforded an ever-growing understanding of how the chemical integrity of DNA is maintained to preserve the biological source code. In Chapter 1, I discuss how the efforts of the community have continued to unlock the secrets of the genome and how these discoveries have initiated an era of genomic medicine, where, for the first time, researchers can directly correct genetic drivers of disease. The progression from meganucleases to zinc finger nucleases (ZFNs), transcription activator-like effector nucleases (TALENs), and clustered regularly interspaced short palindromic repeats (CRISPR) in combination with CRISPR associated (Cas) effectors has continually improved the efficiency and flexibility with which we can modify genomic DNA. Precision genome editing tools like base editors and prime editors¬ enable efficient and precise correction of various genome edits. Base editors are precision genome editing reagents capable of installing C•G-to-T•A mutations (CBEs) and A•T-to-G•C mutations (ABEs). In Chapter 2, I discuss my discovery that base editor expression bottlenecks editing efficiency and the subsequent improvements we made to dramatically increase the activity of both CBE and ABE base editors. By improving the nuclear localization signal (NLS) and optimizing the codon usage of base editor constructs, we greatly improved the efficiency of BE4- and ABE-mediated genome editing, in most tested cases we acheived ~80% target base conversion in HEK293T cells. We also developed novel deaminase domains that afford higher editor expression and consequently improved editing efficiencies by performing Ancestral Sequence Reconstruction (ASR) of the component cytidine deaminase domain to create a novel CBE, “AncBE4max”. We show that optimized “BE4max”, “AncBE4max”, and “ABEmax” base editors are especially enabling under unfavorable conditions such as when delivery of constructs is limiting. These optimized editors were used to correct several pathogenic SNPs in a variety of mammalian cell types with substantially higher efficiencies than BE4 and ABE. AncBE4max, BE4max, and ABEmax substantially expand the capabilities of both cytidine and adenine base editing. Collectively, these variants improved the performance of CBEs and ABEs, established deaminase ancestral reconstruction as a core capability of the lab, and still represent the foundational constructs for CBE and ABE development for both research and therapeutic applications of base editors. While the “BEmax” optimizations improved the general editing performance of both CBEs and ABEs across a number of tested conditions, the targeting scope and precision are limited, constraining the usefulness of these tools. In Chapter 3, I briefly discuss efforts in the field that focused on expanding the targetable genome space (i.e. by developing Cas effectors with different PAM preferences) and other efforts directed at improving the precision of base editing outcomes. I then describe efforts I assisted in to extend the substrate scope of the component deaminase domains of CBEs and ABEs. This work was led by Ben Thuronyi who created a system for phage-assisted continuous evolution of base editors (BE-PACE). This circuit was then used to evolve cytosine base editor (CBE) variants that overcome target sequence context constraints of canonical APOBEC1 deaminase-derived CBEs. Three variants were generated and tested. First, an APOBEC1 deaminase variant, evoAPOBEC1, was evolved that exhibits up to 26-fold improvement in editing GC, a disfavored context for wild-type APOBEC1, while maintaining high levels of editing in all other sequence contexts tested. Second, evolution of an ancestrally reconstructed cytidine deaminase resulted in evoFERNY, a cytidine deaminase that is 29% smaller than APOBEC1, and supports efficient cytosine base editing in all tested sequence contexts. Third, we evolved a highly active CDA1 variant, evoCDA1, which achieved dramatically improved editing at difficult target sites in the CBE context. These evolved CBEs enabled greatly improved base editing of SNPs associated with genetic deafness, Alzheimer’s disease, and Wolfram syndrome in primary cells and cell lines. This same circuit was then used by Michelle Richter and Kevin Zhao to evolve novel ABE deaminase domains (ABE8s) with improved activity across all tested substrates. These projects helped generate novel base editing reagents and extended the targeting scope of base editors, highlighting the possibility for in vivo application of these tools. In Chapter 4, I describe project I contributed to that was led by Jon Levy, in which we developed an in vivo delivery system for CBEs and ABEs. This study describes base editor delivery via adeno-associated viruses (AAVs), a clinically validated viral delivery platform. Base editors, in addition to its promoter, regulatory elements, and a cognate sgRNA and its U6 promoter exceed the packaging limit of AAVs. We overcame this packaging limitation by using trans-splicing inteins that reconstitute full-length base editors in cells that have been transduced with viruses expressing both the N- and C-terminal editor halves. I then describe my work that uses this editing system for the correction of Hutchinson-Gilford Progeria Syndrome (HGPS) in an in vivo mouse model of this devastating disease. HGPS is a fatal disease typically caused by a single dominant-negative C•G-to-T•A mutation (c.1824 C>T, G608G) in LMNA, the nuclear lamin A gene. This mutation activates a cryptic splice site that results in production of progerin, a toxic protein that causes rapid aging and a greatly shortened average lifespan of approximately 14 years1-7. Adenine base editors (ABEs) convert targeted A•T base pairs to G•C base pairs with minimal byproducts and without requiring double-strand DNA breaks or donor DNA templates8-10. The remainder of Chapter 4 describes efforts I led to use an ABE to directly correct the pathogenic LMNA c.1824 C>T mutation in cultured patient-derived fibroblasts and in a mouse model of HGPS. Lentiviral delivery of ABE to cultured fibroblasts results in very efficient (~90%) correction of the pathogenic allele to wild-type LMNA, mitigation of RNA mis-splicing, reduction of progerin protein levels, and correction of nuclear abnormalities. Unbiased DNA off-target editing analysis and transcriptome-wide RNA off-target editing analysis did not detect off-target activity in treated patient-derived fibroblasts. In transgenic mice expressing two copies of the human LMNA c.1824 C>T variant, adeno-associated virus 9 (AAV9)-mediated delivery of ABE via retro-orbital injection resulted in substantial correction of the pathogenic mutation (~20-60% in unsorted cells from a variety of organs 6 months after injection), restoration of normal RNA splicing, and reduction of progerin protein. In vivo base editing also dramatically rescued vascular pathology including recovery of vascular smooth muscle cell counts and prevention of adventitial fibrosis, with minimal progerin-positive smooth muscle nuclei. A single injection of ABE AAV9 at P14 improved animal vitality and greatly extended median lifespan from 215 days to 510 days, approaching old age in C57BL/6 mice. These findings support the potential of in vivo base editing to treat HGPS, and other genetic diseases, by directly correcting the root cause of the disease. The successful application of base editors to the correction of a pathogenic allele is an exciting possibility that highlights the value of having a broad genome editing toolbox for installing all manner of genome edits with high precision and efficiency. In Chapter 5, I discuss my efforts to expand the precision genome editing toolkit through the development of C•G-to-G•C base editors (CGBEs). This project describes the development of a suite of novel engineered CGBEs paired with machine learning models to enable efficient and high-purity C•G-to-G•C base editing at a much larger fraction of target sites and with greater predictability compared to using any single CGBE alone. I performed a CRISPRi screen targeting 476 genes associated with DNA repair or DNA metabolism to identify factors that affect C•G-to-G•C editing outcomes and used the resulting information together with results from cytosine base editor modification experiments to engineer a panel of CGBEs with diverse editing profiles. With collaborators, we characterized ten promising CGBEs on a library of 10,638 genomically integrated target sites in mammalian cells and used the resulting data to train machine learning models that accurately predict the purity and yield of edited outcomes (R = 0.90). These CGBEs enable correction to wild-type amino acid coding sequence of 546 disease-related transversion SNVs with >90% precision and up to 70% efficiency. We demonstrate that machine learning prediction of optimal CGBE and sgRNA choice enables high-purity transversion base editing at >4-fold more target sites than can be achieved using any single CGBE. Together, these tools expand the capabilities and utility of base editing. I also briefly discuss my involvement in the development of a novel genome editing technique termed Prime Editing.