Gene therapy, the introduction of genetic material into a patient to address the underlying causes of disease, has shown strong clinical potential to treat a variety of genetic disorders. There is no scarcity of disease applications for gene therapy and gene editing; rather, the grand challenge in this field is the development of technology platforms for gene delivery that can safely and efficiently mediate stable gene expression. Viral vectors based on adeno-associated virus (AAV) have emerged as successful gene delivery vehicles due to their natural efficacy at circumventing biological barriers and achieving high transduction efficiency. AAV is thus a promising gene therapy vehicle; however, human therapeutic needs demand delivery properties that at best may have conferred AAVs with no selective advantages during natural evolution and at worst may be at odds with natural selection. As a result, there have been considerable efforts to engineer AAV vectors to meet biomedical needs. One successful approach, directed evolution, emulates how viruses naturally evolve – iterative rounds of genetic diversification and selection for improved function – but with selective pressures that can be designed to result in therapeutically useful viruses. Directed evolution has been applied to generate new viral variants with altered gene delivery specificities, but existing AAV libraries and selection strategies must be improved to overcome remaining clinical challenges. These include biological transport barriers that limit viral access to target tissues, poor infectivity of clinically important cell types, and inability to control which tissues are transduced. Such challenges have motivated my dissertation research to engineer new AAV libraries, design improved selection strategies, and evolve AAV vectors that address unmet therapeutic needs.

Significant improvements to naturally occurring AAV variants are needed to address the clinical challenges highlighted above, yet these naturally occurring variants often lack the evolutionary plasticity required to tolerate new mutations. In contrast, ancestral sequences are by definition highly evolvable, having given rise to modern AAV variants. I computationally designed and experimentally constructed an ancestral AAV library to access novel viral sequences with enhanced infectious properties, and to gain insights into AAV’s evolutionary history. I found that ancestral AAV variants were broadly infectious, a property that may have enhanced vector spread during the natural evolution of the virus. Moreover, variants selected for muscle tropism mediated up to 31-fold higher gene expression in muscle compared to AAV1, a serotype clinically utilized for muscle delivery, highlighting their potential for gene therapy.

Another method to engineer AAV variants with new biological properties is to shuffle the DNA sequences of naturally occurring AAVs. I applied the computational algorithm SCHEMA to predict the optimal crossover locations for DNA recombination of the AAV capsid gene. SCHEMA calculates the number of amino acid interactions in the capsid crystal structure that are broken upon creation of a chimeric AAV. I designed and constructed a SCHEMA library of over 1.6 million chimeric AAV variants, each representing a possible solution to a therapeutic challenge. I next developed an in vivo Cre-dependent selection strategy to drive convergence from millions of variants to a select few that target adult neural stem cells within the central nervous system. Adult neural stem cells confer plasticity to the central nervous system and have neuroregenerative capabilities that may be harnessed to treat disease. Significant progress has been made in elucidating the molecular mechanisms that govern neural stem cell maintenance and neurogenesis in the subventricular zone (SVZ), the largest germinal niche in the adult mammalian brain. Our understanding of stem cell biology has improved, yet the ability to genetically manipulate endogenous stem cell populations in situ remains challenging due to inefficient vehicles for gene delivery. After three rounds of in vivo selection I identified a novel chimeric AAV, SCH9, which mediates 24-fold higher GFP expression and 12-fold greater transduction volume than AAV9 in the SVZ, and efficiently infects adult neural stem cells. Interestingly, I found that SCH9 combines properties from multiple AAV parent serotypes in its ability to utilize both galactose and heparan sulfate proteoglycans as receptors for cell transduction. Moreover, SCH9 is less susceptible to neutralizing antibodies than the AAV serotypes from which it is derived. In summary, my dissertation research has resulted in new technologies for AAV directed evolution and novel vectors with therapeutic promise.