UC Berkeley

Gene therapy, the addition of nucleic acids to a patient’s cells to treat or prevent disease, is a technique that has shown robust clinical performance across an array of genetic disorders. Central to the success of any gene therapy is the use of a delivery system that can safely, specifically, and efficiently mediate gene expression in target cells. Some of the most promising in vivo gene delivery approaches developed to date are vectors based on non-pathogenic DNA viruses, which evolved the capacity to deliver biologically active DNA to cells through millions of years of natural evolution. However, natural evolutionary constraints resulted in viruses whose properties are not optimized for manufacturability or therapeutic gene delivery. Thus, strategies to engineer improvements are necessary. Directed evolution, the iterative process of creating genetic diversification and selecting for genotypes that confer improved function, is a powerful engineering approach to engineer optimized phenotypes that did not evolve in nature. Initially used to improve function of simple proteins, directed evolution approaches have since been extended to engineer increasingly complex, multimeric proteins such as the capsids of small viruses. Nonetheless, directed evolution approaches remain limited for highly complex applications for which the protein of interest belongs to a large virus or organism whose genome is not amenable to in vitro library generation, the desired phenotype is conferred by a multigenic pathway or poorly defined sequence space, or a selective pressure linking genotype to function does not exist. This dissertation unveils two novel directed evolution approaches and demonstrates their potential to overcome previous barriers to engineering parvoviruses or poxviruses as scalable platforms for therapeutic gene delivery.

Our first directed evolution approach targeted a recently discovered non-structural protein to improve cellular production of adeno-associated virus (AAV). Gene therapies delivered by AAV have shown significant clinical success in recent years. However, manufacturing sufficient AAV to meet current and projected clinical needs is a significant hurdle to the growing gene therapy industry. The recently discovered membrane-associated accessory protein (MAAP) is encoded by an alternative open reading frame in the AAV cap gene that is found in all presently reported natural serotypes. Recent evidence has emerged supporting a functional role of MAAP in AAV egress, though the underlying mechanisms of MAAP function remain unknown. Here, we show that inactivation of MAAP from AAV2 by a single point mutation that is silent in the VP1 ORF (AAV2-ΔMAAP) decreased exosome-associated and secreted vector genome production. We hypothesized that novel MAAP variants could be evolved to increase AAV production and thus subjected a library of over 1E6 MAAP variants to five rounds of packaging selection into the AAV2-ΔMAAP capsid. Between each successive packaging round, we observed a progressive increase in both overall titer and ratio of secreted vector genomes conferred by the bulk selected MAAP library population. Next-generation sequencing uncovered enriched mutational features, and a resulting selected MAAP variant containing a novel C-terminal domain increased overall GFP transgene packaging in AAV2, AAV6, and AAV9 capsids as well as specific infectivity of secreted recombinant AAV2 and AAV6. This work may be applicable to increasing per-cell AAV output in industrial settings, thus leading to decreased costs and increased access to life-saving gene therapies.

Our second directed evolution approach was developed to overcome previous barriers to engineering large, cytoplasmically replicating DNA viruses as oncolytic virotherapies. Oncolytic viruses, whose tropism has been redirected to cancer cells while leaving healthy cells unharmed, have been widely tested in clinical trials to promote antitumor immunity. Among the clinically explored oncolytic virotherapy platforms, vaccinia virus (VV) is particularly attractive due to its extensive clinical safety profile; viral gene deletions that confer specificity for cancer cells; FDA-approved vaccines, therapeutics, and “kill switch” mechanisms to prevent unintended spread; quick replication cycle; capacity to encode up to 40 kbp of heterologous DNA for therapeutic gene expression; and adjuvant properties promote anti-tumor immune responses. Despite these advantages, VV has yet to reach efficacy endpoints needed to clear Phase III clinical trials. Directed evolution holds promise to overcome existing challenges, such as improving delivery to the solid tumor cores or metastasized tumor sites. However, the development of directed evolution approaches for VV have been limited because its ~190 kbp, repetitive sequence-containing, cytoplasmically replicating genome is poorly amenable to established genetic diversification techniques. To enable directed evolution of VV, we engineered a CRISPR-fusion complex called CytoEvolvR and developed novel methods to perform continuous diversification and selection of improved vectors in cancer cells. Next-generation sequencing and florescence reversion assays showed that optimized versions of CytoEvolvR generate tunable, multiplexable, site-specific, non-PAM-constrained diversification of endogenous loci in VV genomes containing safety deletions. In vitro selections of a double-deleted VV vector for which a VV gene of interest was diversified by CytoEvolvR uncovered enriched SNP and truncation variants. The results of this work may be translatable to improving delivery of in vivo cancer immuno- and gene therapies.