Abstract
Systemic administration of adeno-associated virus (AAV) vectors for spinal cord gene therapy has challenges including toxicity at high doses and pre-existing immunity that reduces efficacy. Intrathecal delivery of AAV vectors into the cerebral spinal fluid (CSF) can avoid many of the issues of systemic delivery, although achieving broad distribution of the vector and transgene expression throughout the spinal cord is challenging and vector entry to the periphery occurs, sometimes initiating hepatotoxicity. Here we performed two rounds of in vivo biopanning in non-human primates (NHPs) with an AAV9 peptide display library injected intrathecally and performed insert sequencing on DNA isolated from either whole tissue (conventional selection), isolated nuclei, or nuclei from transgene-expressing cells. A subsequent barcoded pool of candidates and AAV9 was compared at the DNA (biodistribution) and RNA (expression) level in spinal cord and liver of intrathecally injected NHPs. Most of the candidates displayed enhanced biodistribution compared to AAV9 at all levels of spinal cord ranging from 2 to 265-fold. Nuclear isolation or expression-based selection yielded 4 of 7 candidate capsids with enhanced transgene expression in spinal cord (up to 2.4-fold), while no capsid obtained by conventional selection achieved that level. Furthermore, several capsids displayed lower biodistribution to the liver of up to 1,250-fold, compared to AAV9, providing a remarkable on target/off target biodistribution ratio. These capsids may have potential for gene therapy programs directed at the spinal cord and the selection method described here should be useful in clinically relevant large animal models.