Alpha-1 antitrypsin deficiency (AATD) is characterized by both chronic lung disease due to loss of wild-type AAT (M-AAT) antiprotease function and liver disease due to toxicity from delayed secretion, polymerization, and aggregation of misfolded mutant AAT (Z-AAT). The ideal gene therapy for AATD should therefore comprise both endogenous Z-AAT suppression and M-AAT overexpression. We designed a dual-function rAAV3B (df-rAAV3B) construct, which was effective at transducing hepatocytes, resulting in a considerable decrease of Z-AAT levels and safe M-AAT augmentation in mice. We optimized df-rAAV3B and created two variants, AAV3B-E12 and AAV3B-G3, to simultaneously enhance the concentration of M-AAT in the bloodstream to therapeutic levels and silence endogenous AAT liver expression in cynomolgus monkeys. Our results demonstrate that AAV3b-WT, AAV3B-E12, and AAV3B-G3 were able to transduce the monkey livers and achieve high M-AAT serum levels efficiently and safely. In this nondeficient model, we did not find downregulation of endogenous AAT. However, the dual-function vector did serve as a potentially \"liver-sparing\" alternative for high-dose liver-mediated AAT gene replacement in the context of underlying liver disease.

Capsid engineering of adeno-associated virus (AAV) can surmount current limitations to gene therapy such as broad tissue tropism, low transduction efficiency, or pre-existing neutralizing antibodies (NAb) that restrict patient eligibility. We previously generated an AAV3B combinatorial capsid library by integrating rational design and directed evolution with the aim of improving hepatotropism. A potential isolate, AAV3B-DE5, gained a selective proliferative advantage over five rounds of iterative selection in hepatocyte spheroid cultures. In this study, we reanalyzed our original dataset derived from the AAV3B combinatorial library and isolated variants from earlier (one to three) rounds of selection, with the assumption that variants with faster replication kinetics are not necessarily the most efficient transducers. We identified a potential candidate, AAV3B-V04, which demonstrated significantly enhanced transduction in mouse-passaged primary human hepatocytes as well as in humanized liver chimeric mice, compared to the parental AAV3B or the previously described isolate, AAV3B-DE5. Interestingly, the AAV3B-V04 capsid variant exhibited significantly reduced seroreactivity to pooled or individual human serum samples. Forty-four percent of serum samples with pre-existing NAbs to AAV3B had 5- to 20-fold lower reciprocal NAb titers to AAV3B-V04. AAV3B-V04 has only nine amino acid substitutions, clustered in variable region IV compared to AAV3B, indicating the importance of the loops at the top of the three-fold protrusions in determining both transduction efficiency and immunogenicity. This study highlights the effectiveness of rational design combined with targeted selection for enhanced AAV transduction via molecular evolution approaches. Our findings support the concept of limiting selection rounds to isolate the best transducing AAV3B variant without outgrowth of faster replicating candidates. We conclude that AAV3B-V04 provides advantages such as improved human hepatocyte tropism and immune evasion and propose its utility as a superior candidate for liver gene therapy.

AAV virion biology is still lacking a complete understanding of the role that the various structural subunits (VP1, 2, and 3) play in virus assembly, infectivity, and therapeutic delivery for clinical indications. In this study, we focus on the less studied adeno-associated virus AAV3B and generate a collection of AAV plasmid substrates that assemble virion particles deficient specifically in VP1, VP2, or VP1 and 2 structural subunits. Using a collection of biological and structural assays, we observed that virions devoid of VP1, VP2, or VP1 and 2 efficiently assembled virion particles, indistinguishable by cryoelectron microscopy (cryo-EM) from that of wild type (WT), but unique in virion transduction (WT > VP2 > VP1 > VP1 and 2 mutants). We also observed that the missing structural subunit was mostly compensated by additional VP3 protomers in the formed virion particle. Using cryo-EM analysis, virions fell into three classes, namely full, empty, and partially filled, based on comparison of density values within the capsid. Further, we characterize virions described as \"broken\" or \"disassembled\" particles, and provide structural information that supports the particle dissolution occurring through the two-fold symmetry sites. Finally, we highlight the unique value of employing cryo-EM as an essential tool for release criteria with respect to AAV manufacturing.