Characterization of AAV integrations and rearrangements from long and short reads with RAAVioli

Recombinant Adeno Associated Viral (rAAV)-based gene therapy (GT) applications have been successfully exploited for the treatment of several disorders. rAAV mainly remains episomal in the nucleus of transduced cells, however, numerous studies demonstrated integration of fragmented or full-length AAV DNA within the transduced cell genome where double-strand DNA breaks (DSBs) or nicks have occurred. Yet, preclinical studies revealed the occurrence of hepatocellular carcinoma and clonal expansion events consequent to rAAV insertions, posing safety concerns for their clinical use. However, bioinformatics tools able to identify AAV integration sites (IS) and characterize vector rearrangements are still missing. Here, we collected data from a humanized liver mouse model, where human primary hepatocytes have been transduced ex-vivo or in-vivo with a tomato expressing AAV. PCR amplicons or DNA fragments containing AAV vector portions were sequenced by both short paired- end and long reads and then analyzed by RAAVioli (Recombinant Adeno-Associated Viral IntegratiOn analysis), to characterize vector rearrangements and IS. Python and R scripts parse the alignments to identify IS and reconstruct rearrangements using CIGAR strings. We retrieved 811 and 370 IS from short paired-end Illumina reads and long PacBio reads respectively, confirming the higher efficiency of PCR-based approach in IS retrieval. The distribution of AAV IS was sparse in the human genome similarly in both datasets, and Albumin gene was the most targeted gene as expected. Furthermore, 32 ISs were in common between the two datasets, demonstrating the reliability of RAAVioli independently from sequencing platform adopted. Both datasets showed a similar percentage (~25%) of fragments with AAV rearrangements, however more than 2 rearrangements per fragment were retrieved only in long PacBio reads. Precision and accuracy of RAAVioli pipeline was assessed through simulated datasets obtaining scores >0.95 in IS identification and rearrangement characterization. These data demonstrated that RAAVioli is a comprehensive and flexible bioinformatic tool that can efficiently map AAV IS using long and short paired ends sequencing reads. These approaches are fundamental to characterize AAV integration and recombination events in gene therapy and gene editing applications, allowing and improving the assessment of safety in AAV studies.