Gates Package — page 827

Safety concerns with our approach are related to the intravenous injection of HDAd5/35++ vectors and genotoxicity due to off-target base editing. HDAd vectors are sensed by the RES and can trigger the release/production of pro-inflammatory cytokines. We have developed pharmacological approaches to avoid these responses in mice (2) and NHPs (8). Furthermore, intravenously injected HDAd5/35++ vectors preferentially transduce mobilized HSCs (8, 47), minimizing base editing in non-hematopoietic tissues. Recently, virus-like particles (VLPs) have been used for in vivo editing of murine liver or retina(48). Furthermore, non-viral vehicles based on lipid nanoparticles (LNPs) for mRNA delivery (49, 50) hold great promise for in vivo HSC genome editing. However, these new approaches also face the hurdles accounted with intravascular administration of viral vectors, namely the unproductive sequestration, which require high vector doses, and associated acute toxicity. Furthermore, while viruses have evolved highly efficient mechanisms for cellular uptake and intracellular trafficking, it is not straightforward to incorporate these functions into nonviral delivery vehicles. Clearly, prolonged expression of the editing machinery could increase the frequency of off-target editing and the genotoxicity risk. Our data indicate that off-target DNA or RNA editing is minimal while on-target activity is high. i) No hematological or histological abnormalities were found in primary and secondary mice over a total period of 32 weeks. ii) More than 98% of edits at the target site were the desired A>G conversions; less than 1.5% of reads had small deletions or insertions in the observation window. Interestingly, the indel frequency in secondary mice (at week 16) was significantly lower than in primary mice, indicating a selection against cells with indels over time. iii) NGS of off-target sites predicted by CIRCLE-Seq or CasOFFinder did not show editing in the quantification window. iv) Off-target base editing can manifest as guide-independent, spurious editing of RNA (51, 52) or genomic DNA (30). This, as well as clonal expansion of malignant cells could result in alterations in the transcriptome. While RNA-seq showed the expected activation of human -globin genes (10- to 100-fold log10)) and residual human mgmtP140K mRNA, changes in the mouse transcriptome were minimal (2.0- to 2.15-fold log10) and it is debatable whether these small alterations have biological consequences. Extensive off-target analyses are crucial for a potential clinical translation of the approach. Therefore, in upcoming NHP studies with an improved HDAd-EF1.ABE8e vector, great attention will be given to genomic analyses and potential long-term effects of in vivo base editing.