Gates Package — page 823

editing (and -globin reactivation) did not provide a proliferative advantage or disadvantage for any specific lineage in -YAC mice (Fig.7E). NGS demonstrated again a high specificity for A>G conversions with less than 1.5% of indels (Fig.7F-H, S8). Editing of the four HBG target sites on a single cell level was less efficient than in the ex vivo setting (Fig.7I). 66.7% and 71.4% of colonies had biallelic conversion in the HBG1 and HBG2 promoters, respectively. 21% (HBG1) and 23.6% (HBG2) of colonies had mono-allelic -113A>G conversions. The percentages without HBG1 and HBG2 editing were 21 and 23.8%, respectively. Flow cytometry measuring -globin in peripheral RBCs showed a gradual increase from 30% (average at week 4) to 70% -globin+ RBCs (average at week 16-P) (Fig.8A). A similar kinetic has been observed in our previous studies (12, 37). Levels of human -globin (hHBG) mRNA relative to human HBB mRNA were 30% on average at week 16 after in vivo transduction (Fig.8B). Relative to mouse HBA and HBB mRNA, the level of hHBG mRNA was 12% and 15%, respectively. Human -globin protein levels, measured by HPLC of erythrocyte lysates, were about 25% of human -globin chains and ~15%/20% of mouse - or -globin chains, respectively (Fig.8C). The presence of HbF in peripheral RBCs of in vivo treated CD46/-YAC mice was also shown by staining of cytospins with a human HbF specific antibody (Figs.8D and E). The signals show a spectrum of intensity, most likely related to how many of the four target sites were edited. HbF staining on human peripheral blood and cord blood RBCs were used as negative and positive controls, respectively. Editing rates and percentages of -globin+ RBCs were maintained in secondary recipients (Fig.S9). In summary, in vivo HSC transduction by a single intravenous HDAd injection into mobilized β-YAC/CD46 mice followed by early in vivo selection resulted in >60% editing at the -113A site. 25% -globin of human β-globin is expressed in >60% of erythrocytes in healthy mice. Off-target editing. Using non-integrating vectors eliminates the risk of insertional mutagenesis. However, the application of genome editing enzymes bears the possibility of genotoxicity due to off-target editing. To assess this, we screened potential off-target editing sites using "Circularization for In vitro Reporting of Cleavage Effects by sequencing" (CIRCLE-seq), a highly sensitive in vitro method capable of interrogating genome-wide off-target activity (38) (Fig.9). CD46/β-YAC genomic DNA was cleaved with SpCas9 coupled with sgHBG#2 guide RNA for CIRCLE-Seq. The sgHBG#2 RNA was identical to that used in the HDAd-EF1.ABE8e vector. CIRCLE-Seq identified a total of 272 candidate off-target sites (Supplementary Excel file S1). To investigate off-target editing in in vivo transduced mice, we performed targeted deep sequencing for the top 20 out of the 272 sites based on the CIRCLE-seq read count. Genomic DNA from the mouse with the highest on-target editing (65%) (week 16-S, secondary) and a naïve mouse