82. Mapping the Humoral Immune Response to AAV by Molecular Docking and Cryo-Electron Microscopy for the Design of Next-Generation AAV Vectors

Abstract

Despite constant improvements in gene therapy with AAV vectors, including use of new serotypes, routes of administration, and transgene cassette engineering, immune responses continue to be a major obstacle to therapeutic translation. Neutralizing antibodies (NAb) against the capsid generated by prior viral infection or AAV vector administration significantly reduce not only the effective patient population but also the overall efficacy of an AAV-based gene therapy. Here, by cloning out and evaluating anti-AAV antibodies from singly-sorted memory B cells from seropositive individuals, we have designed an approach that allows us to look at the humoral immune response globally, to hone in on the immunogenic regions of the capsid itself, and to compare responses between individuals in an unbiased and therapeutically-relevant setting. In this study, we screened a panel of 30 normal human donors, selected one with high pre-existing NAb titers (1:320 for AAV2, 1:40 for AAV3B), then sorted out switched memory B cells by negative selection, seeding on irradiated ms3T3-CD40L feeder cells and culturing for 2 weeks in the presence of IL-2 and IL-21. Following supernatant screening for anti-AAV antibody production, over 100 AAV-reactive clones were identified. After isolation and cloning using nested PCR, antibodies were evaluated for AAV capsid binding as well as neutralizing capacity. To date, all antibodies demonstrated binding to AAV2 and AAV3B as well as a panel of additional AAV serotypes (8, 9, rh10, rh32.33), suggesting that AAV-binding, yet non-neutralizing antibodies may possess broad serotype specificity.To identify the epitopes for these anti-AAV antibodies, we first took a high-throughout, predictive approach. Antibody variable region sequences were placed into a generic antibody scaffold and their three-dimensional structure modeled using Kotai Antibody Builder and RosettaAntibody followed by validation using COOT. The resulting structures were then iteratively docked onto the published structure of AAV3B using PIPER to identify the most energetically-favorable binding conformation. Thus far, the vast majority of footprint residues lie in variable regions of the capsid comprising and surrounding the 3-fold spikes. For a number of antibodies, initial prediction-directed capsid alanine scanning experiments have shown decreased antibody binding at the predicted residues, supporting the use of this approach. More comprehensive mutagenesis experiments are underway to further validate the approach and more completely map immunogenic epitopes of the AAV capsid proteins. In addition, cryo-EM analysis is currently underway for a number of these Fab-AAV complexes for additional, direct observation of the repertoire of binding footprints. These studies will provide information critical to understanding the antibody response to AAV and guide future attempts to rationally design next-generation capsids that are able to evade the anti-AAV antibody response.