Attachment

differentiation. This resulted in the reduction of ROS to background level and improved erythroid differentiation/expansion of thalassemia and SCD CD34+ cells and suggests that the in vivo base editing could be curative in -thalassemia and SCD patients. Several factors could have contributed to the high efficacy of in vivo editing: i) the greatly enhanced catalytic activity of ABE8e (30). ii) the high level of ABE8e expression in HSC mediated by the EF1 promoter. This is supported by comparative studies with HDAd-PGK.ABE8e where ABE8e expression was driven by the relatively weak PGK promoter and rates of editing (both ex vivo and in vivo) and -globin reactivation were significantly lower (Fig.S11, specifically S11I). iii) the prolonged expression of the base editing machinery from episomal HDAd vector in some transduced HSCs. The VCN declined only by one order of magnitude within 6 weeks, which would imply that the editing machinery is active during that time. This is in part supported by the ex vivo study (performed without O6BG/BCNU selection), where an increase of editing rates from the transplant (~38%) to >95% in week 16 mice was observed. We speculate that the extended presence of the HBG1/2 sgRNA and ABE8e in HSCs could increase the probability for editing of poorly accessible target sites (i.e. sites blocked by heterochromatin at a certain stage of HSC differentiation). In this context, stimulation of cell division by O6BG/BCNU selection could have further increased target site accessibility. As the data in secondary recipients indicate, editing must have occurred in primitive long-term repopulating HSCs (before cell division and differentiation). iv) Bystander editing of adenines near the -113 target site, specifically -116 A>G, which would destroy the BCL11A binding motif and, most likely, further inhibit the binding of the repressor. The -116 A>G bystander conversion was found in >99% of -113 A>G edited alleles (Fig.S12). Also, in the absence of the -110 A>G substitution, a GATA binding motif would be created similar to the -113A>G HPFH (31) . This would enable binding of the transcriptional activator GATA1 to the destroyed BCL11A motif and could further increase -globin re-activation. v) Expansion of edited HSPCs with episomal HDAdEF1.ABE8e genomes by early treatment with O6BG/BCNU capitalizing on the prolonged presence of episomal vector genomes. While we did not perform in vivo studies without O6BG/BCNU treatment, the in vitro data with CD34+ cells from healthy donors (Fig.1F) and -thalassemia patients (Fig.2), illustrate the effect of selection. Notably, in -thalassemia and SCD patients, -globin-expressing erythroid cells would have a survival advantage created by the disease background. In this context, we expect that rates of -globin reactivation achieved with our approach could be even higher in patients with hemoglobinopathies and the in vivo selection with O6BG/BCNU on demand.