UC Berkeley

With the recent advent of several generations of targeted DNA nucleases, most recently CRISPR/Cas9, genome editing has become broadly accessible across the biomedical community. The capacity of these nucleases to modify specific genomic loci associated with human disease could render new classes of genetic disease, including autosomal dominant or even idiopathic disease, accessible to gene therapy. In parallel, the emergence of adeno-associated virus (AAV) as a clinically important vector raises the clear possibility of integrating these two technologies towards the development of gene editing therapies. Though clear challenges exist, numerous proof of concept studies in preclinical models offer exciting promise for the future of gene therapy.

Retinal degenerative diseases are a major area of therapeutic interest today. The accessibility and well-characterized nature of the retina have led to gene therapeutic approaches for recessive retinal degenerative diseases that have begun to reach the clinic, but dominant diseases, such as many forms of autosomal dominant retinitis pigmentosa, require ablation of the disease allele rather than simple replacement. I use AAV as a delivery vector for CRISPR/Cas9 to treat the form of autosomal dominant retinitis pigmentosa caused by a P23H mutation in the mouse rhodopsin Rho gene. After optimizing the delivery route and the targeting guide RNA, I knocked out all native mouse alleles in rod photoreceptors and replaced them with a copy of Rho that cannot be targeted by the gene editing nucleases. Analysis of rods confirms successful gene editing of the Rho locus, and long-term live tracking of the outer nuclear layer shows modest rescue of photoreceptors over time, demonstrating the viability of this technique for direct editing and rescue of photoreceptors in retinal genetic disease.

While therapeutic gene editing is rising to prominence today, a major concern for in vivo gene editing is the persistence of virally delivered gene editing molecules. Long-term expression of Cas9 could lead to dangerous, undesired off-target editing inside of edited cells, or conversely, edited cells could be targeted for destruction by the immune system due to the presence of foreign protein expression, negating the benefit of correcting such cells. I therefore designed a transient CRISPR/Cas9 system for AAV delivery. I integrated copies of the desired genomic target site flanking the Cas9 vector itself such that upon functional expression of Cas9 and its guide RNA, the delivered Cas9-containing AAV genome would also be targeted for cleavage and destruction by Cas9, generating an effective pulse of Cas9 activity. I show that such self-inactivating constructs can successfully self-inactivate while retaining desired genomic editing activity, creating a temporal dynamic in which expression of Cas9 rises and then falls, and the transience conferred by this system can significantly reduce even short-term off-target editing rates. Further, I show that in contexts where the vector is targeted for destruction by Cas9 before desired genomic editing can occur, incorporation of mismatches in the integrated target sites can tune the rate at which self-inactivation occurs and increase on-target editing rates. I thus demonstrate the functionality of transient gene editing constructs delivered in AAV, improving the potential safety and viability of in vivo therapeutic gene editing.

The use of AAV for therapeutic purposes requires AAV variants capable of infecting the target cell type in its \textit{in vivo} context. As natural AAV serotypes may not be optimized for infection in a given context, library-based approaches of library selection and directed evolution have succeeded at generating novel characteristics for AAV vectors. Developing screens and selections for multiple traits, however, can be difficult, and combining the results of independent selections may not be feasible. I show that two libraries with local compatibility for selected mutations can be combined to gain the traits of both. I take peptide inserts and variant loops selected for photoreceptor transduction from the vitreous and combine them with a variant selected from an AAV DNA shuffle library created via the SCHEMA algorithm for greater infectivity of neural stem cells and antibody evasion. The shuffle block in the SCHEMA corresponding to the location of peptide insert variant is notably from AAV2, the same serotype the photoreceptor libraries were generated in, increasing the likelihood of compatibility due to an identical local amino acid microenvironment. I show that combining these mutations more reliably confers both greater retinal infectivity as well as antibody evasion properties. I thus demonstrate a more reliable and robust approach to successfully combining results of powerful library selection methods to generate AAV vectors with multiple desired traits. Together, these results demonstrate the viability of retinal gene therapy with AAV-based gene editing approaches and provide additional tools for improved safety and efficacy of AAV-based gene editing therapy.