UC Berkeley

Inherited retinal degenerations, effecting about 1:3000 people, have historically been challenging to treat. A large number of mutations have been identified in these patients, which share a similar phenotype of a gradual loss of photoreceptor cells leading to blindness. In recent decades, a promising new form of treatment called gene therapy has been developed to address patient needs. Encouragingly, there have been several successful clinical trials using adeno-associated virus (AAV) to deliver genetic cargo to the necessary retinal cells. Viral vectors based on AAV have seen success because of the ability of the virus to achieve high transfection efficiency and a strong safety profile. The seminal gene therapy clinical trial utilizing AAV was the delivery of RPE65 to the retinas of patients with Leber congenital amaurosis (LCA), which proved to have long-term safety and showed significant efficacy. But, challenges still remain for broad application of AAV. It is therefore necessary to develop AAV variants with novel properties in order to address the wide range of retinal degenerations.

There are a number of natural AAV serotypes that exhibit different tropisms for different retinal cells allowing for the potential of treatments targeted to different cell types. Importantly, the route of delivery strongly affects the cells targeted and depends on whether the virus is administered subretinally or intravitreally. Natural AAV serotypes perform poorly when delivered intravitreally due to structural barriers, such as the inner limiting membrane. Therefore, initial clinical trials chose to deliver AAV subretinally, which has the advantage of local delivery to the photoreceptors or retinal pigment epithelium and limited immune response. But, expression is restricted to only a portion of the retina and there is a risk of damage from retinal detachment in a tissue that is already degenerating. Intravitreal injection would be preferred and would allow for pan-retinal expression and would not cause retinal detachment. Importantly, recent evidence has shown species specificity for AAV variants necessitating future AAV engineering approaches to take into account the specific challenges of the human retina. One method of AAV engineering, directed evolution, has yielded many novel AAV variants with advantageous properties. Directed evolution is a biomimetic process that emulates how viruses naturally evolve. Large libraries of highly diverse AAV variants are iteratively selected utilizing pressures for a desired attribute.

This dissertation focuses on the directed evolution of novel variants that are able to intravitreally deliver cargo in large animal models. Canines serve as a strong preclinical model for retinal degenerations and contain a visual streak for high acuity vision, which is lacking in rodents. Directed evolution was performed in a canine model and variants were selected utilizing deep sequencing of each round of evolution resulting in AAVs that had high outer nuclear layer or bipolar cell tropism when delivered intravitreally. Directed evolution was also performed in non-human primates (NHPs) because the retina shares many similarities to the human retina, including a cone-rich macula and fovea, and has the highest chance of yielding the most clinically relevant AAVs.

To generate a NHP variant, several new approaches and tools were developed. First, novel AAV libraries were generated including AAV4-, AAV5-, and Ancestral-based peptide insertion libraries. Multiple plasmid backbones with unique restriction and primer binding sites were created to allow for tuning of different pools of libraries throughout the directed evolution process. A packaging strategy that mitigated the possibility of replication and immune response during NHP directed evolution was developed leading to a 10-fold increase in safety. Serum screening for animals without pre-existing antibodies to AAV also increased safety. Implementing these tools and strategies for directed evolution in a NHP model led to an AAV2-based peptide insertion variant that showed increase cone targeting and decreased ganglion cell targeting when intravitreally delivered. For both the canine and NHP directed evolution studies, deep sequencing of each round of selection revealed “hidden” variants and elucidated the biases in the selection process due to initial overrepresentation of particular variants. An additional study to increase the packaging capacity of AAV was performed and demonstrated the size limits of AAV cargo. The research performed in this dissertation led to AAV variants with increased therapeutic properties and insights into the mutability of AAV and performance of individual variants throughout the directed evolution process.