Attachment

of magnitude lower, respectively. NGS of the target area confirmed the A>G conversion rates detected by Sanger sequencing and determined that indels (mostly deletions) were <1.5% of all reads (Fig.5F, S5). HSCs contain two HBG promoters (HBG1 and HBG2) on each allele. To assess editing of the four target sites on a single cell level, we isolated individual progenitor colonies and sequenced their DNA (n=20 colonies). The data showed 100% bi-allelic editing of both promoters (Fig.5G). -globin protein was detected by flow cytometry in 80-90% of peripheral RBCs starting after week 8 (Fig.6A), which is in accordance with the HSC engraftment/expansion kinetics (Fig.5B) and previous studies with CRISPR/Cas9 (36). Relative -globin expression levels were measured on the mRNA and protein levels. HBG mRNA levels were ~35% of human HBB mRNA levels and ~22% of mouse HBA and HBB mRNA (Fig.6B). In mice transplanted with ex vivo edited Lin- cells, human -globin levels were about 30% of human - globin chains and ~20% of mouse - or -globin chains (Fig.6C). Editing rates and -globin reactivation were stable in secondary recipients demonstrating that editing occurred in long-term repopulating HSCs (Fig.S6). In summary, HDAd-EF1.ABE8e confers efficient ex vivo HSC editing without toxicity associated with the transduction and editing processes. After editing, -globin levels were 35% of human -globin levels without any O6BG/BCNU in vivo selection. In vivo HSC genome editing. One of our hypotheses was that after in vivo HSC transduction, HDAd genomes and, consequently, ABE8e and mgmtP140K expression would be gradually lost due to cell division and vector genome degradation (Fig.S7). We therefore designed a new in vivo HSC transduction/selection regimen shown in Fig.7A. After mobilization of CD46/-YAC transgenic mice and intravenous injection of a single HDAd-EF1.ABE8e vector, O6BG/BCNU treatment was started at day 2 and repeated at days 12 and 26 to capitalize on mgmtP140K expression from episomal HDAd-EF1.ABE8e genomes and to dilute out the vector by stimulating cell division. The latter was demonstrated by qPCR that measured vector copy numbers (VCN) in bone marrow MNCs. In in vivo transduced "primary" mice, during the 16-week experiment, the VCN declined by 2-4 orders of magnitude (Fig.7B), resulting in undetectable ABE8e and mgmtP140K mRNA levels at week 16 after transduction (data not shown). Base editing was measured by Sanger sequencing in PBMCs (Fig.7C). A>G conversion rates increased after selection to 50-90% (average 67%) for the -113 target site with a similar pattern of bystander adenine editing seen in the ex vivo study. Furthermore, conversion rates were similar in PBMCs, splenocytes, and bone marrow MNCs (Fig.7D). Editing rates were comparable for CD3+ (T-cell), CD19+ (B-cell), Gr-1+ (myeloid), and Ter-119+ (erythroid) lineages in the bone marrow, indicating that editing occurred in multi-potential HSCs and that HBG1/2