Over the past decade, directed-evolution method was extensively used to generate novel AAV variants for targeted gene therapy in the central nervous system (CNS). These studies involved screening different peptide insertion libraries designs, considering serotype, insertion positioning, and peptide size. In this study, we employed NGS-guided in vivo screening to compare the performance of three AAV9 peptide insertion library designs: 7-mer peptide, 12-mer peptide, and liver-de-targeted 7-mer peptide. After each round of screening, a comprehensive bioinformatic analysis was conducted to monitor the process and compare CNS-derived peptides with off-target tissues.

These screenings revealed that candidates derived from the 12-mer insertion and liver-de-targeted 7-mer libraries exhibited improved performance compared to the unmodified 7-mer library. Subsequently, the CNS targeting properties of 28 selected mutant 7-mer and the 12-mer candidates were tested in mice using both a barcoded approach and as single variants. Among the 7-mer variants <50% showed higher CNS-expression compared to parental control vectors, while 100% of the 12-mer candidates outperformed AAV9 in terms of CNS expression and liver de-targeting. The two 12-mer candidates, AAV9-CNS_005 and AAV9-CNS_006, demonstrated enhanced CNS transduction, outperforming AAV9 by ~26 and 40-folds respectively, and AAV9-PHP.eB by ~4-5-folds. Immunohistochemistry analysis confirmed that both AAV9-CNS_005 and AAV9-CNS_006, like AAV9-PHP.eB, primarily targeted astrocytes, endothelial cells, and neurons in the CNS, while minimal expression was observed with AAV9.

Overall, this study emphasizes the importance of optimizing peptide library design, refining the screening process, and using stringent bioinformatic analysis to identify efficient and specific AAV variants.