Adeno-associated virus (AAV) vectors are quickly becoming the vectors of choice for therapeutic gene delivery. To date, hundreds of natural isolates and bioengineered variants have been reported. While factors such as high production titer and low immunoreactivity are important to consider, the ability to deliver the genetic payload (physical transduction) and to drive high transgene expression (functional transduction) remains the most important feature when selecting AAV variants for clinical applications. Reporter expression assays are the most commonly used methods for determining vector fitness. However, such approaches are time consuming and become impractical when evaluating a large number of variants. Limited access to primary human tissues or challenging model systems further complicates vector testing. To address this problem, convenient highthroughput methods based on next-generation sequencing (NGS) are being developed. To this end, we built an AAV Testing Kit that allows inherent flexibility in regard to number and type of AAV variants included, and is compatible with in vitro, ex vivo, and in vivo applications. The Testing Kit presented here consists of a mix of 30 known AAVs where each variant encodes a CMV-eGFP cassette and a unique barcode in the 3¢-untranslated region of the eGFP gene, allowing NGS-barcode analysis at both the DNA and RNA/cDNA levels. To validate the AAV Testing Kit, individually packaged barcoded variants were mixed at an equal ratio and used to transduce cells/tissues of interest. DNA and RNA/cDNA were extracted and subsequently analyzed by NGS to determine the physical/functional transduction efficiencies. We were able to assess the transduction efficiencies of immortalized cells, primary cells, and induced pluripotent stem cells in vitro, as well as in vivo transduction in naïve mice and a xenograft liver model. Importantly, while our data validated previously reported transduction characteristics of individual capsids, we also identified novel previously unknown tropisms for some AAV variants.

Use of the prototypical adeno-associated virus type 2 (AAV2) capsid delivered unexpectedly modest efficacy in an early liver-targeted gene therapy trial for hemophilia B. This result is consistent with subsequent data generated in chimeric mouse-human livers showing that the AAV2 capsid transduces primary human hepatocytes in vivo with low efficiency. In contrast, novel variants generated by directed evolution in the same model, such as AAV-NP59, transduce primary human hepatocytes with high efficiency. While these empirical data have immense translational implications, the mechanisms underpinning this enhanced AAV capsid transduction performance in primary human hepatocytes are yet to be fully elucidated. Remarkably, AAV-NP59 differs from the prototypical AAV2 capsid by only 11 aa and can serve as a tool to study the correlation between capsid sequence/structure and vector function. Using two orthogonal vectorological approaches, we have determined that just 2 of the 11 changes present in AAV-NP59 (T503A and N596D) account for the enhanced transduction performance of this capsid variant in primary human hepatocytes in vivo, an effect that we have associated with attenuation of heparan sulfate proteoglycan (HSPG) binding affinity. In support of this hypothesis, we have identified, using directed evolution, two additional single amino acid substitution AAV2 variants, N496D and N582S, which are highly functional in vivo. Both substitution mutations reduce AAV2's affinity for HSPG. Finally, we have modulated the ability of AAV8, a highly murine-hepatotropic serotype, to interact with HSPG. The results support our hypothesis that enhanced HSPG binding can negatively affect the in vivo function of otherwise strongly hepatotropic variants and that modulation of the interaction with HSPG is critical to ensure maximum efficiency in vivo. The insights gained through this study can have powerful implications for studies into AAV biology and capsid development for preclinical and clinical applications targeting liver and other organs.

Recent clinical successes in gene therapy applications have intensified interest in using adeno-associated viruses (AAVs) as vectors for therapeutic gene delivery. Although prototypical AAV2 shows robust in vitro transduction of human hepatocyte–derived cell lines, it has not translated into an effective vector for liver-directed gene therapy in vivo. This is consistent with observations made in Fah−/−/Rag2−/−/Il2rg−/− (FRG) mice with humanized livers, showing that AAV2 functions poorly in this xenograft model. Here, we derived naturally hepatotropic AAV capsid sequences from primary human liver samples. We demonstrated that capsid mutations, likely acquired as an unintentional consequence of tissue culture propagation, attenuated the intrinsic human hepatic tropism of natural AAV2 and related human liver AAV isolates. These mutations resulted in amino acid changes that increased binding to heparan sulfate proteoglycan (HSPG), which has been regarded as the primary cellular receptor mediating AAV2 infection of human hepatocytes. Propagation of natural AAV variants in vitro showed tissue culture adaptation with resulting loss of tropism for human hepatocytes. In vivo readaptation of the prototypical AAV2 in FRG mice with a humanized liver resulted in restoration of the intrinsic hepatic tropism of AAV2 through decreased binding to HSPG. Our results challenge the notion that high affinity for HSPG is essential for AAV2 entry into human hepatocytes and suggest that natural AAV capsids of human liver origin are likely to be more effective for liver-targeted gene therapy applications than culture-adapted AAV2.