Characterization of AAV Integrations in preclinical models of gene therapy using RAAVioli pipeline with long and short sequencing reads

Adeno-Associated Viral vectors (AAVs) have been widely used in gene therapy to treat various genetic disorders. Although they are typically considered episomal vectors, sev- eral studies have shown that both fragmented and full-length AAV DNA can integrate into the genomes of host cells leading to hepatocellular carcinoma and clonal expansion in some pre- clinical models. However, methods and bioinformatic tools that provide a reliable and efficient assessment of AAV integration sites (IS) are required. Here, we propose a sonication-based PCR approach combined with short-read sequencing and a bioinformatics pipeline called RAAVIoli (Recombinant Adeno-Associated Viral Integration analysis) to characterize AAV integration sites (IS) and vector rearrangements. RAAVioli utilizes Python and R scripts to parse alignments, identify IS, and reconstruct vector rearrangements using CIGAR strings. The robustness of our approach was dem- onstrated by characterizing AAV IS in a humanized liver mouse model, where human primary hepatocytes were transduced with a tomato-expressing AAV. In this model, vector insertions were previously characterized using an AAV-specific probe base selection method and long-read PacBio sequencing. Sequencing reads were analyzed in both cases using the same RAAVioli pipeline. A greater number of AAV IS were identified by PCR/ short-read sequencing (N 1⁄4 730) as compared to the long-read sequence approach (N 1⁄4 370). AAV IS distributed similarly within the human genome showing the typical preference of tar- geting CpG islands and transcriptional start sites. Moreover, 32 IS were shared among the two datasets, demonstrating the consis- tency of the results obtained independently from the sequencing platform adopted. Although the short-read sequencing method identified a lower number of IS (*10%) with rearranged AAV genomes compared to the long-read sequencing approach (*25%), it demonstrates significantly greater efficiency in retrieving AAV insertion sites. Thanks to this platform, we recently characterized AAV IS in a liver-directed gene editing approach used to correct Wilson Disease (WD). WD is an autosomal recessive disorder caused by mutations in the ATP7B gene which is primarily expressed in hepatocytes and is involved in copper metabolism. Here, genome editing was accomplished by integrating a promoter- less human mini-ATP7B cDNA into the albumin locus (Alb- ATP7B) using a nuclease-free homology-directed repair (HDR) mechanism. The treatment corrected the disease pheno- type in Atp7b-/- mice by promoting liver repopulation of edited hepatocytes. Analysis of AAV IS revealed that vector insertion did not only cluster near or within the albumin-edited site but also distributed throughout the entire gene, exhibiting a strand-orientation bias. Indeed, 80% of the Albumin IS have the integrated vector genome oriented in the same direction as the targeted gene. Interestingly, this phenomenon was not observed in Atp7b-/- mice injected with a control vector expressing GFP. These data suggest the occurrence of a selec- tion mechanism favoring the survival of hepatocytes express- ing the therapeutic transgene consequently to integration events that are not driven by HDR-mediated mechanism. In summary, our work indicates that the development of reli- able methods for the characterization of AAV integrations is fun- damental to providing insights into the safety and efficacy of these vectors in several gene therapy applications.