UC Berkeley

Gene delivery vehicles, or vectors, based on adeno-associated viruses (AAV) have demonstrated success in both preclinical disease models and recently in human clinical trials for several disease targets, including muscular dystrophy, hemophilia, Parkinson's disease, Leber's congenital amaurosis, and macular degeneration. AAV gene therapy vectors have become increasingly utilized because the parent virus is nonpathogenic in humans, they can transduce both dividing and non-dividing cells, and they are efficient for some important cell and tissue types. AAV's simple genome contains two open reading frames, which encode the nonstructural proteins needed for viral replication and virus assembly (rep) and the three structural proteins that assemble to form a 60-mer viral capsid (cap). To create a gene therapy vector, a therapeutic gene of interest is inserted in place of the viral open reading frames to be packaged during vector production. Despite its considerable promise and emerging clinical success, several challenges impede the broader implementation of AAV gene therapy, including the prevalence of anti-AAV neutralizing antibodies in the human population due to natural exposure to the parent virus, low transduction of a number of therapeutically relevant cell types, and an inability to overcome physical transport barriers in the body. These challenges arise since the demands we place on AAV vectors are often different from or even at odds with the properties nature bestowed on their parent viruses. Viral directed evolution - the iterative generation of large, diverse libraries of viral mutants and selection for variants with specific properties of interest - offers a promising means to address these problems.

Directed evolution is a high-throughput, molecular engineering approach that our group has adapted and implemented to create AAV variants with novel properties, such as altered receptor binding, altered cell transduction, and altered tissue transduction in the body. In general, the method emulates the process of natural evolution, in which repeated genetic diversification and selection enable the accumulation of key mutations or genetic modifications that progressively improve a molecule's function, even without knowledge of the underlying mechanistic basis for the problem. For AAV, this process has involved mutating wild-type AAV cap genes to create large genetic libraries, which can be packaged to generate libraries of viral particles, each of which is composed of a variant capsid surrounding a viral genome encoding that capsid. A selective pressure - such as high-affinity antibodies against the AAV capsid, the ability to infect adult neural stem cells, or the ability to infect human pluripotent stem cells - is then applied to promote the emergence of variants able to surmount these barriers. After each such selection step, the successful variants can be recovered and used as the starting material for the next selection step to further enrich for improved variants. After several such selection steps, the resulting cap gene pool is subjected to additional mutagenesis and selection. After several rounds of mutagenesis and selection, the resulting variants can be analyzed individually for the desired property.

Using directed evolution, I have engineered several novel AAV variants capable of enhanced gene delivery in several applications. First, variants selected in the presence of pooled human antibodies were 2- to 35-fold less susceptible to neutralization by anti-AAV antibodies compared to parental AAV in vitro. The antibody neutralization properties also translated to enhanced transduction in an in vivo mouse model in the presence of neutralizing antibodies. The isolation of such novel variants resistant to anti-AAV antibodies may enable the future treatment of patients with pre-existing immunity that are currently ineligible for AAV gene therapy. Second, a novel variant selected for the ability to infect adult neural stem cells was capable of more efficient gene delivery to rat, mouse, and human neural stem cells in vitro. Furthermore, the variant demonstrated more efficient and specific transduction of rat and mouse neural stem cells in vivo compared to natural AAV serotypes. Gene delivery to neural stem cells using this variant could be used as a gene therapy option to better harness these cells for tissue regeneration. Finally, a novel variant selected for the ability to infect human pluripotent stem cells was able to transduce several human embryonic stem cell and induced pluripotent stem cell lines 3-fold more efficiently than natural AAV serotypes, which enabled a 10-fold increase in the efficiency of a genome-editing technique termed gene targeting. These results demonstrate that engineered and evolved AAV vectors are highly promising for a range of applications from the lab to the clinic.