Systematic multi-trait AAV capsid engineering for efficient gene delivery

Abstract

Broadening gene therapy applications requires manufacturable vectors that efficiently transduce target cells in humans and preclinical models. Conventional selections of adeno-associated virus (AAV) capsid libraries are inefficient at searching the vast sequence space for the small fraction of vectors possessing multiple traits essential for clinical translation. Here, we present Fit4Function, a generalizable machine learning (ML) approach for systematically engineering multi-trait AAV capsids. By leveraging a capsid library that evenly samples the manufacturable sequence space, reproducible screening data are generated to train accurate sequence-to-function models. Combining six models, we designed a multi-trait (liver-targeted, manufacturable) capsid library and validated 89% of library variants on all six predetermined criteria. Furthermore, the models, trained only on mouse in vivo and human in vitro Fit4Function data, accurately predicted AAV capsid variant biodistribution in macaque. Top candidates exhibited high production yields, efficient murine liver transduction, up to 1000-fold greater human hepatocyte transduction, and increased enrichment, relative to AAV9, in a screen for liver transduction in macaques. The Fit4Function strategy ultimately makes it possible to predict cross-species traits of peptide-modified AAV capsids and is a critical step toward assembling an ML atlas that predicts AAV capsid performance across dozens of traits.

Natural Adeno-Associated Virus Serotypes and Engineered Adeno-Associated Virus Capsid Variants: Tropism Differences and Mechanistic Insights

“…More recently, the employment of the AAV9-CMV-Express platform, which relies on the recovery of library transcripts to ensure the selection of transcriptionally functional variants [ 260 ] ( Figure 5 ), led to the identification of the AAV9-derived variant AAV-BI30 [ 261 ]. This variant showed an overall higher, broader, more selective, and strain-compatible transduction of endothelial cells in vitro and in vivo compared to previously identified variants such as AAV-BR1.…”

“…Recently, a novel pipeline leveraging machine learning, called Fit4Function, has emerged as a significant conceptual and technological advance, proving successful in generating novel AAV capsid variants de novo [ 260 ]. Fit4Function is an in silico prediction model that incorporates multi-function properties to find variants optimized for several properties simultaneously.…”

Targeting AAV vectors to the central nervous system by engineering capsid–receptor interactions that enable crossing of the blood–brain barrier

“…SVAE-generated variants were then evaluated in silico for predicted production fitness to preempt a high proportion of the variants failing to be produced at detectable levels in the virus library [ 38 ]. For the saturation mutagenesis approach, we generated all possible 7-mer sequences containing these motifs without filtering by predicted production fitness.…”

“…That is, if gen_variants is the full set of novel variants generated for either receptor, for a variant v in gen_variants, the joint score joint_score( v ) of v is defined as follows: This split accounts for the SVAE’s lack of production fitness knowledge, i.e., the VAE model may not understand the destabilizing effects of certain AAs in the context of our 7-mer insert, such as cysteine (C) or tryptophan (W). The production fitness predictor is described in a separate article [ 38 ].…”

An AAV capsid reprogrammed to bind human Transferrin Receptor mediates brain-wide gene delivery

“…The AAV9 variant capsids carrying CAG-NLS-mScarlet-P2A-Luciferase-WPRE-SV40pA or CMV/CBA- GBA1 -HA-pA were either produced in suspension followed by iodixanol purification and titered as previously described ( 64 ) or produced using adherent cells as previously described ( 65 ). Production, purification methods, and post-purification yields are provided in figure S8.…”