Base-editing-mediated dissection of a γ-globin cis-regulatory element for the therapeutic reactivation of fetal hemoglobin expression

Abstract

Sickle cell disease and β-thalassemia affect the production of the adult β-hemoglobin chain. The clinical severity is lessened by mutations that cause fetal γ-globin expression in adult life (i.e., the hereditary persistence of fetal hemoglobin). Mutations clustering ~200 nucleotides upstream of the HBG transcriptional start sites either reduce binding of the LRF repressor or recruit the KLF1 activator. Here, we use base editing to generate a variety of mutations in the -200 region of the HBG promoters, including potent combinations of four to eight γ-globin-inducing mutations. Editing of patient hematopoietic stem/progenitor cells is safe, leads to fetal hemoglobin reactivation and rescues the pathological phenotype. Creation of a KLF1 activator binding site is the most potent strategy - even in long-term repopulating hematopoietic stem/progenitor cells. Compared with a Cas9-nuclease approach, base editing avoids the generation of insertions, deletions and large genomic rearrangements and results in higher γ-globin levels. Our results demonstrate that base editing of HBG promoters is a safe, universal strategy for treating β-hemoglobinopathies.

Fig. 1. LRF BS disruption and KLF1 BS creation in the HBG1/2 promoters in K562 cells.

a Schematic representation of the β-globin locus on chromosome 11, depicting the 5’ hypersensitive sites of the locus control region (5’ LCR HSs; gray boxes), HBE1, HBG2, HBG1, HBD, and HBB genes (colored boxes), the HBG2 and HBG1 promoters (white boxes) and the 3’ hypersensitive to DNase I site (3’HS). The sequence of the HBG2 and HBG1 identical promoters, from –212 to –179 nucleotides upstream of the HBG transcription start sites, is shown below. Red and green ovals indicate LRF repressor and KLF1 activator. HPFH mutations identified in the HBG1 and/or HBG2 promoters are highlighted by black arrows, and HPFH mutations that can be reproduced by ABEs or CBEs are highlighted in green and red, respectively. The percentage of HbF expression in heterozygous HPFH carriers and carriers of SCD (*) or β-thalassemia (**) is indicated in brackets. The sequence of LRF BS upon generation of the LRF 4C, LRF 8C, LRF 2T, and KLF1 profiles is presented, and modified bases are highlighted in red and green. b ChIP–qPCR analysis of LRF at HBG1/2 promoters in edited and control (mock-transfected) K562 cells. ChIP was performed using an antibody against LRF. HBG prom pair of primers was used to amplify the HBG1/2 promoters. DEFB122 served as a negative control. Data were normalized to the values observed at the KLF1 locus (positive control). Data are expressed as mean ± SEM (n = 3 biologically independent experiments) (left panel). C-G to T-A or A-T to G-C base-editing efficiency of the input and the LRF immunoprecipitated fractions was calculated by the EditR software in samples subjected to Sanger sequencing. Data are expressed as mean ± SEM (n = 3 biologically independent experiments) (right panel). *P = 0.0140; **P = 0.0040 (two-way ANOVA with Dunnett correction for multiple comparisons). Source data are provided as a Source Data file.

Fig. 2. LRF BS disruption and KLF1 BS creation in the HBG1/2 promoters of SCD HSPC-derived erythroblasts.

a Experimental protocol used for base-editing experiments in non-mobilized SCD HSPCs. A BE-, a sgRNA- and a GFP- (optional for enzyme plasmids that do not contain a GFP cassette) expressing plasmid were co-transfected in SCD HSPCs and 18 h post-transfection GFP⁺ cells were FACS-sorted based on GFP medium (med) and high (high) expression. b C-G to T-A or A-T to G-C base-editing efficiency, calculated by the EditR software in samples subjected to Sanger sequencing. The LRF 4C editing profile was obtained by pooling data from CBE-NRCH-, CBE-SpG- and CBE-SpRY- treated samples. Data are expressed as mean ± SEM (n = 12 (LRF 4C), n = 4 (LRF 8C), n = 4 (LRF 2T med), n = 4 (LRF 2T high), n = 3 (KLF1 med), n = 4 (KLF1 high) biologically independent experiments, 4 donors). c Frequency of InDels, measured by TIDE analysis for control, base- and Cas9-edited samples subjected to Sanger sequencing. The insertion or deletion of a C (±1 nt) in the homopoly-C stretch of the LRF 2T profile was separated from the overall frequency of InDels, as it was considered a sequencing error (Supplementary Note 2). Data are expressed as mean ± SEM (n = 12 (control), n = 12 (LRF 4C), n = 4 (LRF 8C), n = 4 (LRF 2T med), n = 4 (LRF 2T high), n = 3 (KLF1 med), n = 3 (KLF1 high, n = 3 (Cas9 med), n = 3 (Cas9 high) biologically independent experiments). ****P ≤ 0.0001 (ordinary one-way ANOVA with Dunnett correction for multiple comparisons). d Frequency of the 4.9-kb deletion, measured by ddPCR, for base- and Cas9-edited samples. Data are expressed as mean ± SEM (n = 8 (control), n = 12 (LRF 4C), n = 5 (LRF 8C), n = 3 (LRF 2T med), n = 3 (LRF 2T high), n = 3 (KLF1 med), n = 3 (KLF1 high), n = 3 (Cas9 med), n = 4 (Cas9 high) biologically independent experiments). *P = 0.0125 (ordinary one-way ANOVA with Dunnett correction for multiple comparisons). e Analysis of HbF and HbS by cation-exchange HPLC in SCD patient RBCs. We calculated the percentage of each Hb type over the total Hb tetramers. The base-editing efficiency is indicated for each sample in the lower part of the panel. Data are expressed as single values or as mean ± SEM (n = 4 (control), n = 6 (LRF 4C), n = 2 (LRF 8C), n = 2 (LRF 2T med), n = 1 (LRF 2T high), n = 1 (KLF1 med), n = 1 (KLF1 high), n = 1 (Cas9 med), n = 2 (Cas9 high) biologically independent experiments, 2 donors). *P = 0.0141 for LRF 4C, or P = 0.0380 for LRF 8C; ****P ≤ 0.0001 (two-way ANOVA with Dunnett correction for multiple comparisons). f Flow cytometry histograms showing the percentage of HbF- and HbS- expressing cells in GYPA^+high population for unstained (GYPA stained only), control (transfected with TE buffer for donor 1 and transfected with TE buffer or with CBE-SpRY plasmid and a sgRNA targeting the unrelated AAVS1 locus for donor 2) and edited samples. g Time-course measurement of the frequency of non-sickle cells upon O₂ deprivation in control (transfected with TE buffer for donor 1 and transfected with TE buffer or with CBE-SpRY plasmid and a sgRNA targeting the unrelated AAVS1 locus for donor 2) and edited samples. Data are expressed as single values or as mean ± SEM (n = 2 (control), n = 3 (LRF 4C), n = 1 (LRF 8C), n = 1 (LRF 2T med), n = 1 (LRF 2T high), n = 1 (KLF1 med), n = 1 (KLF1 high), n = 1 (Cas9 high) biologically independent experiments, 2 donors). h Correlation between HBG mRNA relative expression and base-editing efficiency in single BFU-E colonies (1 donor). HBG mRNA expression was normalized to HBA1/2 mRNA and expressed as a percentage of the total HBB + HBG mRNA. Base-editing efficiency was calculated by the CRISPRESSO 2 software in samples subjected to NGS. Colonies highlighted by a black outline carried InDels. *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001 (KLF1: R² = 0.7299, Y = 0.8896*X + 31.37, P < 0.0001 non-zero slope significance; LRF 2T: R² = 0.8574, Y = 0.5977*X + 25.07, P < 0.0001 non-zero slope significance; LRF 8C: R² = 0.6102, Y = 0.4824*X + 23.77, P < 0.0001 non-zero slope significance; LRF 4C: R² = 0.1678, Y = 0.2529*X + 30.52, P = 0.0522 non-zero slope significance; Multiple t test). BFU-Es edited at the AAVS1 locus were used as negative controls. i Frequency of InDels, measured by the CRISPRESSO 2 software, for edited or control (AAVS1) single BFU-E colonies (KLF1 n = 17; LRF 2T n = 17; LRF 8C n = 11; LRF 4C n = 15; AAVS1 n = 8; 1 donor). j Frequency of the 4.9-kb deletion, measured by ddPCR, for edited or control (AAVS1) single BFU-E colonies (KLF1 n = 11; LRF 2T n = 11; LRF 8C n = 9; LRF 4C n = 15; AAVS n = 8; 1 donor). k Frequency of chromosome 11 loss, as indicated by the ratio of CARS (p arm) and PODL1 (q arm), measured by ddPCR, for edited or control (AAVS1) single BFU-E colonies (KLF1 n = 16; LRF 2T n = 15; LRF 8C n = 10; LRF 4C n = 14; AAVS1 n = 9; 1 donor). Source data are provided as a Source Data file.

Fig. 3. RNA-mediated base editing in SCD HPSCs.

a Experimental protocol used for base-editing experiments using BE mRNAs in SCD HSPCs (2 plerixafor-mobilized donors and 1 non-mobilized donor). A BE mRNA and a chemically modified sgRNA were co-transfected in SCD HSPCs. Cells were differentiated into mature RBCs or underwent a CFC assay. b–e C-G to T-A (b, c) or A-T to G-C (d, e) base-editing efficiency, calculated by the EditR software in samples subjected to Sanger sequencing at early (Day 6) or late (Day 13) time points during the in vitro erythroid differentiation protocol. Data are expressed as mean ± SEM (n = 3 biologically independent experiments, 3 donors). f Frequency of InDels, measured by TIDE analysis, for control, base- and Cas9- edited samples. Data are expressed as mean ± SEM (n = 3 biologically independent experiments). ****P ≤ 0.0001 (ordinary one-way ANOVA with Dunnett correction for multiple comparisons). g Frequency of the 4.9-kb deletion, measured by ddPCR, for control, base- and Cas9- edited samples. Data are expressed as mean ± SEM (n = 3 biologically independent experiments). ****P ≤ 0.0001 (ordinary one-way ANOVA with Dunnett correction for multiple comparisons). h CFC frequency for control and edited samples. Data are expressed as mean ± SEM (n = 3 biologically independent experiments, 3 donors). No statistical differences were observed between control and edited samples (two-way ANOVA). i RT-qPCR analysis of β^S- and γ-globin mRNA levels in SCD patient erythroblasts at day 13 of erythroid differentiation. β^S- and γ-globin mRNA expression was normalized to α-globin mRNA and expressed as a percentage of the β^S-+γ- globins mRNA. Data are expressed as mean ± SEM (n = 3 biologically independent experiments, 3 donors). *P = 0.0058; ***P = 0.0003; ****P ≤ 0.0001 (two-way ANOVA with Dunnett correction for multiple comparisons). j Expression of ^Gγ-, ^Aγ-, γ(^Gγ + ^Aγ)-globin chains measured by RP-HPLC in SCD patient RBCs. γ-globin expression was normalized to α-globin. Data are expressed as mean ± SEM (n = 3 biologically independent experiments, 3 donors). *P = 0.0133 for LRF 4C, or P = 0.0257 for LRF 8C; ***P = 0.0002; ****P ≤ 0.0001 (two-way ANOVA with Dunnett correction for multiple comparisons). k Analysis of HbF and HbS by cation-exchange HPLC in SCD patient RBCs. We calculated the percentage of each Hb type over the total Hb tetramers. Data are expressed as mean ± SEM (n = 3 biologically independent experiments, 3 donors). **P = 0.0013; ***P = 0.0003; ****P ≤ 0.0001 (two-way ANOVA with Dunnett correction for multiple comparisons). l Flow cytometry histograms showing the percentage of HbF- and HbS-expressing cells in GYPA⁺ population for unstained (GYPA stained only), control (untreated, or transfected with TE buffer, or transfected with a BE mRNA only, or transfected with a BE mRNA and a sgRNA targeting the unrelated AAVS1 locus) and edited samples. m Frequency of HbF- and HbS- expressing cells in GYPA⁺ population for unstained, control and edited samples. Data are expressed as mean ± SEM (n = 3 biologically independent experiments, 3 donors). *P = 0.0105; **P = 0.0046 for LRF 8C and KLF1, or P = 0.0011 for LRF 2T; ***P = 0.0006 (two-way ANOVA with Dunnett correction for multiple comparisons). n Frequency of sickling cells upon O₂ deprivation in control and edited samples. Data are expressed as single values or as mean ± SEM (n = 3 biologically independent experiments, 3 donors). o Representative photomicrographs of SCD patient RBCs under hypoxia conditions. Red arrows indicate sickling RBCs, and green arrows indicate normal RBCs. Source data are provided as a Source Data file.

Fig. 4. RNA-mediated base editing in β-thalassemia HSPCs.

a, b C-G to T-A (a) or A-T to G-C (b) base-editing efficiency, calculated by the EditR software in β-thalassemic erythroblasts subjected to Sanger sequencing. Data are expressed as mean ± SEM (n = 2 biologically independent experiments, peripheral blood-derived non-mobilized HSPCs from 2 donors harboring the CD39/IVS1-110 G > A and CD 8/9 + G/IVS1-110 G > A mutations, respectively). c Frequency of InDels, measured by TIDE analysis, for base-edited samples subjected to Sanger sequencing. Data are expressed as mean ± SEM (n = 2 biologically independent experiments). d CFC frequency for control (transfected with TE buffer) and base-edited samples. Data are expressed as mean ± SEM (n = 2 biologically independent experiments, 2 donors). No statistical differences were observed between control and edited samples (two-way ANOVA). e RT-qPCR analysis of β-like globin mRNA levels in β-thalassemia patient erythroblasts at day 13 of erythroid differentiation. β-like globin mRNA expression was normalized to α-globin mRNA. Data are expressed as mean ± SEM (n = 2 biologically independent experiments, 2 donors). *P = 0.0270; ****P ≤ 0.0001 (two-way ANOVA with Dunnett correction for multiple comparisons). f Analysis of HbF, HbA and HbA₂ by cation-exchange HPLC in β-thalassemia patient RBCs. We calculated the percentage of each Hb type over the total Hb tetramers. Data are expressed as mean ± SEM (n = 2 biologically independent experiments, 2 donors). ****P ≤ 0.0001 (two-way ANOVA with Dunnett correction for multiple comparisons). g Frequency of HbF-expressing cells in the GYPA⁺ population for control and edited samples, as measured by flow cytometry. Data are expressed as mean ± SEM (n = 4 biologically independent experiments, 2 donors). **P = 0.0098 for LRF 8C, or P = 0.0033 for KLF1 (ordinary one-way ANOVA with Dunnett correction for multiple comparisons). Representative flow cytometry histograms showing HbF⁺ cells in GYPA⁺ populations for control and base-edited samples are presented below the graph. h Expression of β-, δ-, ^Gγ-, ^Aγ- and γ- (^Gγ- + ^Aγ-) globin chains measured by RP-HPLC in β-thalassemia patient RBCs. β-like globin expression was normalized to α-globin. The ratio α/non-α globins is reported on top of the graph. Data are expressed as mean ± SEM (n = 2 biologically independent experiments, 2 donors). *P = 0.0148 (two-way ANOVA with Dunnett correction for multiple comparisons). i Analysis of α-globin precipitates by cation-exchange HPLC in β-thalassemia patient RBCs. We calculated the proportion of α-globin precipitates over the total Hb tetramers. Data are expressed as mean ± SEM (n = 2 biologically independent experiments, 2 donors). ***P = 0.0009 for LRF 8C, or P = 0.0004 for KLF1 (one-way ANOVA with Dunnett correction for multiple comparisons). j Frequency of enucleated cells at day 6, 9, 13, 16, and 20 of erythroid differentiation, as measured by flow cytometry analysis of DRAQ5 nuclear staining in control and edited samples. Data are expressed as mean ± SEM (n = 2 biologically independent experiments, 2 donors). Representative flow cytometry histograms showing the DRAQ5^- cell population for unstained, control, and edited samples are presented below the graph. ***P = 0.0003; ****P ≤ 0.0001 (two-way ANOVA with Dunnett correction for multiple comparisons). k Cell size of enucleated cells (DRAQ5^-) at day 13, 16, and 20 of erythroid differentiation, as measured by flow cytometry using the median of forward scatter (FSC) intensity, and normalized to HD RBCs. Data are expressed as mean ± SEM (n = 2 biologically independent experiments, 2 donors). Representative flow cytometry contour plots showing the FSC of DRAQ5^- cell population for control and edited samples are reported below the graph. **P = 0.0033 for D13, or P = 0.0021 for D16, or P = 0.0034 for D20; ****P ≤ 0.0001 (two-way ANOVA with Dunnett correction for multiple comparisons). l–n Frequency of CD36⁺ (l), CD71⁺ (m) and GYPA⁺ (n) cells at day 6, 9, 13, 16, and 20 of erythroid differentiation, as measured by flow cytometry analysis of CD36, CD71, and GYPA erythroid markers. Data are expressed as mean ± SEM (n = 2 biologically independent experiments, 2 donors). Representative flow cytometry histograms showing the CD36⁺ (l), CD71⁺ (m) and GYPA⁺ (n) cell populations for unstained, control, and edited samples are presented below the graph. *P = 0.0241 for CD36/D16, or P = 0.0190 for CD36/D20, or P = 0.0307 for CD71; **P = 0.0072 for CD36, or P = 0.0020 for CD71/LRF 8C, or P = 0.0028 for CD71/KLF1; ***P = 0.0002; ****P ≤ 0.0001 (two-way ANOVA with Dunnett correction for multiple comparisons). o Frequency of α4-Integrin⁺, BAND3⁺ and α4-Integrin⁺/BAND3⁺ in 7AAD^-/GYPA⁺ cells at day 6, 9, 13, 16, and 20 of erythroid differentiation, as measured by flow cytometry analysis of α4-Integrin and BAND3 erythroid markers. Data are expressed as mean ± SEM (n = 2 biologically independent experiments, 2 donors). Representative flow cytometry contour plots showing the α4-Integrin⁺, BAND3⁺ and α4-Integrin⁺/BAND3⁺ cell populations for unstained, control, and edited samples are reported below the graph. p Frequency of apoptotic cells (Annexin V⁺-cells) in control and edited samples at day 13 of erythroid differentiation, as measured by flow cytometry. Data are expressed as mean ± SEM (n = 2 biologically independent experiments, 2 donors). Representative flow cytometry contour plots showing the Annexin V⁺ cell populations for unstained, control, and edited samples are reported on the right side of the graph. **P = 0.0094 for LRF 8C, or P = 0.0014 for KLF1 (one-way ANOVA with Dunnett correction for multiple comparisons). q Frequency of ROS-containing cells (DCFDA⁺ cells) in control and edited samples at day 20 of erythroid differentiation, as measured by flow cytometry analysis in DRAQ⁺ and DRAQ^- cells. Data are expressed as mean ± SEM (n = 2 biologically independent experiments, 2 donors). Representative flow cytometry contour plots showing the DCFDA⁺ cell populations for unstained, control, and edited enucleated samples are reported on the right side of the graph. Source data are provided as a Source Data file.

Fig. 5. DNA damage and immune responses in HSPCs upon RNA-mediated base editing and RNP-mediated Cas9 treatment.

a C-G to T-A or A-T to G-C base-editing efficiency, calculated by the EditR software in mobilized HD HSPC samples subjected to Sanger sequencing 6 days post-transfection. Data are expressed as mean ± SEM (n = 3 biologically independent experiments, 3 donors). b C-G to T-A or A-T to G-C base-editing efficiency, calculated by the EditR software in mobilized HD HSPC samples subjected to Sanger sequencing 12h-, 24h-, 48h-, and 6 days post-transfection. Data are expressed as mean ± SEM (n = 2–3 biologically independent experiments, 2–3 donors). c Frequency of InDels, measured by TIDE analysis, for control, base- and Cas9-edited samples. Data are expressed as mean ± SEM (n = 3 biologically independent experiments, 3 donors). ***P = 0.0005 (ordinary one-way ANOVA with Dunnett correction for multiple comparisons). d Frequency of the 4.9-kb deletion, measured by ddPCR, for control, base- and Cas9- edited samples. Data are expressed as mean ± SEM (n = 3 biologically independent experiments, 3 donors). ****P ≤ 0.0001 (ordinary one-way ANOVA with Dunnett correction for multiple comparisons). e RT-qPCR analysis of genes activated by RNA stimuli, 24 h post-transfection in HD HSPCs. TNF-α, IL-6, IL-12, IFN-α, IFN-β, TLR7, TLR8, RIG-I mRNA expression was normalized to GAPDH mRNA. LPS-activated macrophages were used as a positive control. Data are expressed as mean ± SEM (n = 3 biologically independent experiments, 3 donors). ****P ≤ 0.0001 (ordinary one-way ANOVA with Dunnett correction for multiple comparisons). f RT-qPCR analysis of CDKN1 (p21), 24, 48 and 72 h post-transfection in HD HSPCs. CDKN1 mRNA expression was normalized to GAPDH mRNA. Data are expressed as mean ± SEM (n = 3 biologically independent experiments, 3 donors). *P = 0.0292 for 24 h, or P = 0.0259 for 48 h, or P = 0.0131 for 72 h; **P = 0.0019; ****P ≤ 0.0001 (ordinary one-way ANOVA with Dunnett correction for multiple comparisons). Source data are provided as a Source Data file.

Fig. 6. Transcriptomic analysis of HD HSPCs after RNA-mediated base editing and RNP-mediated Cas9 treatment.

a RNA-seq analysis was performed 48 h after transfection. Mean-difference plots show differentially expressed genes of edited samples over control samples. Genes that are not statistically significant (false discovery rate [FDR] > 0.05) differentially expressed are depicted by black dots. Genes that are statistically significant (FDR < 0.05) upregulated or downregulated are depicted by red and blue dots, respectively. The enzyme used and the profile generated are indicated on top of each plot. b Expression of statistically significant (FDR < 0.05) upregulated or downregulated genes, as measured by the log2-fold change of FPKM of edited samples over control samples. Data are expressed as mean ± SEM (n = 3 biologically independent experiments, 3 donors). c Strip plots showing the variant allele frequency of A > G mutations or C > T mutations in RNA observed in HSPCs obtained from three different HD. The total number of mutations are indicated above each sample. Source data are provided as a Source Data file.

Fig. 7. sgRNA-dependent DNA off-target activity of the base-editing system.

a–c sgRNA-dependent off-target DNA sites, as evaluated by GUIDE-seq analysis, of LRF_bs_3 (LRF 4C) (a), LRF_bs_2 (LRF 8C) (b), and KLF1_bs_1 (LRF 2T and KLF1) (c) sgRNAs in HEK293T cells. sgRNAs were coupled with a Cas9-nuclease corresponding to the Cas9 nickase included in the base editor (Cas9-SpRY for LRF_bs_3 and LRF_bs_2 sgRNAs, and Cas9 for KLF1_bs_1 sgRNA). The protospacer targeted by each sgRNA and the PAM are reported on top of each panel, followed by the off-target sites and their mismatches with the on-target (highlighted in color). The number of sequencing reads, the chromosomal coordinates (Human GRCh37/hg19), and the site of each off-target are reported. d–f Frequency of C-G to T-A (d and e) or A-T to G-C (f) base-editing conversion at on-target and off-target (OT) sites, for control and LRF 4C (d), LRF 8C (e), LRF 2T (f), and KLF1 (f) samples, as measured by targeted NGS sequencing. Data are expressed as individual values and median (n = 3 biologically independent experiments, 3 donors). *P = 0.0107 for d, or P = 0.0299 for f; **P = 0.0022; ****P ≤ 0.0001 (two-way ANOVA with Sidak (d and f) or Tukey (f) correction for multiple comparisons). g Venn diagrams showing the overlapping of C > T or A > G single-nucleotide variants in exons, in control (Ctr), CBE-SpRY-OPT2-, or ABEmax- treated HSPCs obtained from three different HD. Source data are provided as a Source Data file.

Fig. 8. RNA-mediated base editing of the −200 region of HBG promoters in repopulating HSCs.

a Experimental protocol of HSPC xenotransplantation in NBSGW mice. G-CSF-mobilized HD HSPCs or non-mobilized SCD HSPCs were subjected to RNA-mediated base editing. A BE mRNA and a chemically modified sgRNA were co-transfected in HSPCs and cells were xenotransplanted into NBSGW immunodeficient mice. b Engraftment of human cells in NBSGW mice transplanted with control (mock-transfected, transfected with CBE or ABE mRNA alone) and edited (LRF 8C or KLF1) mobilized HD or SCD HSPCs [HD: n = 4 (mock, CBE mRNA, LRF 8C, ABE mRNA), n = 3 (KLF1) mice per group; SCD: n = 3 (mock), n = 2 (LRF 8C), n = 5 (KLF1) mice per group] 16 to 20 weeks post-transplantation. Engraftment is represented as a percentage of human CD45⁺ cells in the total murine and human CD45⁺ cell population, in bone marrow (BM), spleen, thymus, and peripheral blood. Each data point represents an individual mouse. Data are expressed as mean ± SEM. c Human hematopoietic progenitor content in BM human CD45⁺ cells derived from mice transplanted with control and edited HSPCs [HD: n = 4 (mock), n = 3 (CBE), n = 4 (LRF 8C), n = 3 (ABE), n = 2 (KLF1) mice per group; SCD: n = 5 (mock), n = 1 (LRF 8C), n = 2 (KLF1) mice per group]. We plotted the percentage of human CD45⁺ cells giving rise to BFU-E and CFU-GM. Data are expressed as mean ± SEM. d C-G to T-A or A-T to G-C base-editing efficiency, calculated by the EditR software, in input, bone marrow-, spleen-, BFU-E-, CFU-GM-, and peripheral blood-derived HD and SCD human samples subjected to Sanger sequencing. Data are expressed as mean ± SEM [HD-LRF 8 C: n = 3 (Input) biologically independent experiments, n = 4 (bone marrow and blood), n = 3 (spleen, BFU-E, CFU-GM) mice per group; HD-KLF1: n = 3 (Input) biologically independent experiments, n = 3 (bone marrow and spleen), n = 2 (BFU-E, CFU-GM and Blood) mice per group; SCD-LRF 8C: n = 2 (Input) biologically independent experiments, n = 2 (bone marrow, spleen and blood), n = 1 (BFU-E and CFU-GM) mice per group; SCD-KLF1: n = 2 (Input) biologically independent experiments, n = 6 (bone marrow and spleen), n = 2 (BFU-E and CFU-GM), n = 5 (Blood) mice per group]. The frequency of base editing in input cells was calculated in cells cultured in the HSPC medium (pointing-up triangle), in liquid erythroid cultures (rhombus), BFU-E (square) and CFU-GM (pointing-down triangle) colonies. Each data point (circle) represents an individual mouse. HD/LRF8C: **P = 0.021 for bone marrow, or P = 0.0094 for spleen, or P = 0.0037 for BFU-E, or P = 0.0059 for CFU-GM, or P = 0.0042 for Blood. HD/KLF1: *P = 0.0136; **P = 0.0065; ***P = 0.0009. SCD/LRF8C: **P = 0.0056 for Bone marrow, or P = 0.012 for BFU-E; ***P = 0.0004 for Spleen, or P = 0.0010 for CFU-GM, or P = 0.0001 for Blood. SCD/KLF1: **P = 0.0040; ***P = 0.0009; ****P ≤ 0.0001 (two-way ANOVA with Dunnett correction for multiple comparisons). e Base-editing profile for LRF 8C and KLF1 samples, calculated using EditR software, in input, bone marrow-, spleen- and peripheral blood-derived human samples subjected to Sanger sequencing. Data are expressed as mean ± SEM [LRF 8C: n = 5 (Input) biologically independent experiments, n = 6 (bone marrow and blood), n = 5 (spleen and CFU-GM), n = 4 (BFU-E) mice per group; KLF1: n = 5 (Input) biologically independent experiments, n = 9 (Bone Marrow and Spleen), n = 8 (Blood) n = 4 (BFU-E and CFU-GM) mice per group]. f Frequency of InDels, measured by TIDE analysis, in input, bone marrow-, spleen- and peripheral blood-derived human samples subjected to Sanger sequencing. Data are expressed as mean ± SEM [Input: n = 7 (Control), n = 5 (LRF 8C and KLF1) biologically independent experiments; Bone Marrow: n = 11 (Control), n = 6 (LRF 8C), n = 9 (KLF1) mice per group; Spleen: n = 11 (Control), n = 5 (LRF 8C), n = 9 (KLF1) mice per group; Blood: n = 11 (Control), n = 4 (LRF 8C), n = 8 (KLF1) mice per group; BFU-E: n = 9 (Control), n = 4 (LRF 8C), n = 4 (KLF1) mice per group; CFU-GM: n = 10 (Control), n = 5 (LRF 8C), n = 4 (KLF1) mice per group]. g Frequency of the 4.9-kb deletion, measured by ddPCR, in input samples (left panel). Frequency of mice that bear the 4.9-kb deletion in bone marrow-derived human CD45⁺ cells (right panel). Data are expressed as mean ± SEM (n = 11 (control), n = 5 (LRF 8C and KLF1) biologically independent experiments for left panel and n = 6–9 mice per group for right panel). ***P = 0.0002 (ordinary one-way ANOVA with Dunnett correction for multiple comparisons). h A-T to G-C base-editing efficiency at on- and off-target sites, calculated by the EditR software, in input, bone marrow- and spleen-derived HD and SCD human samples subjected to Sanger sequencing. Data are expressed as mean ± SEM [ctr: n = 3 biologically independent experiments; Input: n = 3 (HD), n = 2 (SCD) biologically independent experiments; BM: n = 3 (HD), n = 6 (ON), n = 5 (OT4) mice per group]. ns P = 0.5920 for HD, or P > 0.9999 for SCD; *P = 0.0104; **P = 0.0033 (two-way ANOVA with Dunnett correction for multiple comparisons). i Base editing in single BFU-E colonies derived from engrafting HD HSPCs, calculated by the EditR software. We plotted the frequency of BFU-E with 0, 1, 2, 3, or 4 edited HBG promoters, the frequency of BFU-E with 0, 1 or 2 edited OT4 alleles and the frequency of BFU-E edited only at HBG promoters or OT4 or at both HBG promoters and OT4 (n = 42 BFU-E obtained from 2 different mice). j Frequency of chromosome 11 loss, as indicated by the ratio of CARS (p arm) and PODL1 (q arm), measured by ddPCR, for edited or control (AAVS1) single BFU-E colonies (KLF1 n = 29 biologically independent colonies; AAVS1 n = 6 biologically independent colonies; 1 donor). k RT-qPCR analysis of β-like globin mRNA levels in bone marrow-derived BFU-E. β-like globins mRNA expression was normalized to α-globin mRNA. Data are expressed as mean ± SEM [n = 7 (Mock), n = 3 (ABE mRNA), n = 4 (KLF1) biologically independent experiments; HD (black circles) and SCD (empty circles) samples]. ****P ≤ 0.0001 (two-way ANOVA with Sidak correction for multiple comparisons). l Correlation between γ-globin mRNA relative expression and base-editing efficiency in bone marrow-derived single BFU-E (n = 69). γ-globin mRNA expression was normalized to α-globin mRNA and expressed as a percentage of the total β- and γ- globin mRNA. Base-editing efficiency was calculated by the EditR software in samples subjected to Sanger sequencing (R² = 0.4263, Y = 0.5328*X + 51.81, P < 0.0001 non-zero slope significance; simple linear regression). Source data are provided as a Source Data file.