Ex vivo editing of human hematopoietic stem cells for erythroid expression of therapeutic proteins

Abstract

Targeted genome editing has a great therapeutic potential to treat disorders that require protein replacement therapy. To develop a platform independent of specific patient mutations, therapeutic transgenes can be inserted in a safe and highly transcribed locus to maximize protein expression. Here, we describe an ex vivo editing approach to achieve efficient gene targeting in human hematopoietic stem/progenitor cells (HSPCs) and robust expression of clinically relevant proteins by the erythroid lineage. Using CRISPR-Cas9, we integrate different transgenes under the transcriptional control of the endogenous α-globin promoter, recapitulating its high and erythroid-specific expression. Erythroblasts derived from targeted HSPCs secrete different therapeutic proteins, which retain enzymatic activity and cross-correct patients' cells. Moreover, modified HSPCs maintain long-term repopulation and multilineage differentiation potential in transplanted mice. Overall, we establish a safe and versatile CRISPR-Cas9-based HSPC platform for different therapeutic applications, including hemophilia and inherited metabolic disorders.

Fig. 1. Editing of selected sites in the α-globin locus minimally affects globin production.

a Locations of gRNA on HBA2 gene. All gRNA except 74 (in purple) target both HBA1 and HBA2. Selected gRNA are highlighted. b K562-Cas9 screening of gRNA targeting different region of the α-globin locus (5′ untranslated region/start codon (5′UTR/ATG), intron 1 (IVS1) or intron 2 (IVS2)). Each bar is a different gRNA, each dot a different experiment. Editing efficiency is expressed as percentage of modified HBA alleles (bars represent mean; data from 1, 2 or 3 biological replicates). The gRNA selected for each region are highlighted. c Editing efficiencies in HSPCs in erythroid liquid culture (circles) or in BFU-E (burst-forming unit-erythroid, squares). Lines show mean (2-4 donors; n = 5 KO, n = 6 AAVS1, n = 9 5′UTR and IVS2). d Colony-forming cell (CFC) frequency in edited HSPCs (mean ± SD; n = 2, n = 4 for IVS2). CFU-GEMM, granulocyte, erythroid, macrophage, megakaryocyte; BFU-E, burst-forming unit-erythroid; CFU-G/GM, granulocyte-macrophage. e Flow-cytometry analysis of erythroid markers upon differentiation of edited HSPCs, day12 (green 5′UTR, purple IVS2 and gray AAVS1). Results are shown as mean ± SD (n = 3 biological replicates, 3 different donors; p = 0.94, one-way ANOVA Tukey’s test). f HPLC analysis of hemoglobin monomers of erythroblasts derived from edited HSPCs (n = 6 5′UTR, n = 5 IVS2, n = 3 KO and AAVS1, n = 4 UT; 4 donors). InDels percentage mean is indicated. The ratio α/β-like globins in normal cells is close to 1 (red dashed line). Black lines show mean; BFU-E (squares), erythroid liquid culture (circles) (***p < 0.001 vs UT, 5′UTR, IVS2, AAVS1; one-way ANOVA, Tukey’s test). g HPLC analysis of hemoglobin tetramers of erythroblasts derived from edited HSPCs (same samples as in f). Every tetramer is reported as % of total hemoglobins (*p < 0.05; **p < 0.01, ***p < 0.001; two-way ANOVA, Dunnet’s test). Black lines show mean; BFU-E (squares), erythroid liquid culture (circles). Hb Bart, γ₄ (p = 0.017 5′UTR, p = 0.013 IVS2, p = 0.006 AAVS1 vs KO); HbH, β₄ (p = 0.038 5′UTR, p = 0.035 IVS2, p = 0.032 AAVS1 vs KO); HbF, fetal hemoglobin, α₂γ₂ (p < 0.001 5′UTR and AAVS1, p = 0.002 IVS2 vs KO); HbA, adult hemoglobin, α₂β₂ (p < 0.001 vs KO). Source data are in the Source Data file.

Fig. 2. Transgene integration into the α-globin locus results in robust erythroid-specific expression.

a AAV6 donors used for KI experiments in 5′UTR (top) and IVS2 (bottom) of the α-globin genes. Both vectors contain a promoterless GFP with bovine growth hormone polyA (pA), followed by a phosphoglycerate Kinase (PGK) promoter with a puromycin selection marker (puro) and simian virus polyA (pA). This cassette is flanked by 250 bp homology arms (homology) to gRNA target. IVS2 trap also contains a synthetic intron (IVS), a splice acceptor (SA) and a self-cleaving peptide (2A). ITR, Inverted terminal repeats. b Representative histograms of GFP expression of HUDEP-2 KI cells at day 0 (light pink), day 7 (red) and day 9 of erythroid differentiation (dark red). Untreated HUDEP-2 are shown in gray (n = 1). c Barplot of GFP median fluorescent intensity (MFI) as in b. d Schematic representation of HSPC targeting experiments. e Representative histograms of GFP expression of 5′UTR KI (red fill) and AAV6 only HSPCs (gray line) during erythroid differentiation. Percentage of GFP positive cells is indicated (n = 3 different donors). f GFP median fluorescent intensity (MFI) during differentiation of 5′UTR KI HSPCs (lines indicate mean, n = 3 different donors indicated by open, gray and black circles). g Representative dot plots showing GFP expression in erythroid (GYPA+) and leukocytes (hCD45+) CFC. h Representative overlay images (bright field and GFP channel) of different erythroid progenitor-derived colonies (n = 24). Scale bars in red indicate 200 μm. CFU-GEMM, granulocyte, erythroid, macrophage, megakaryocyte; BFU-E, burst-forming unit-erythroid; CFU-E, erythroid. Source data are in the Source Data file.

Fig. 3. F9 KI into the α-globin locus results in expression and secretion of functional enzyme.

a AAV6 donor used for KI experiments of FIX Padua. b FIX KI efficiency in HUDEP-2 cells was measured by flow cytometry (light green) or ddPCR specific for on-target integration (dark green) before and after sorting (n = 1). c Quantification of FIX mRNA in KI HUDEP-2 upon differentiation (mean ± SD, n = 2 undifferentiated, n = 3 differentiated). d Quantification of FIX secretion in medium of HUDEP-2 clones (n = 28) with monoallelic or biallelic KI (ELISA), as detected by on target ddPCR analysis (AAV-genome junction amplification). Lines represent median. e KI efficiency in HSPCs at day 9 of erythroid differentiation. Lines represent mean (n = 4). f, g FIX expression during HSPC differentiation at RNA (f, qPCR; n = 2 day 9; n = 4 day 12) and protein level (g, ELISA on supernatants, n = 3 day 7; n = 4 day 9 and 12; 3 donors). Bars represent mean ± SD. h Comparison of FIX antigen (ELISA) and activity (aPTT) in supernatants of KI HSPCs (mean; n = 2). i, j Comparison of FIX RNA at day 9 and 12 of erythroid differentiation (i) and protein (j) in KI HSPCs (AAV + RNP) vs HSPCs transduced with an erythroid-specific lentiviral vector (LvEry FIX). Bars represent mean (**p = 0.003 t-test Holm-Sidak correction for RNA at day 12; p = 0.08 for protein, n = 2). k Integration pattern in single BFU-E (2 donors): no integration (0), monoallelic (1) and biallelic KI (2). Source data are in the Source Data file.

Fig. 4. Expression and therapeutic potential of different lysosomal enzymes.

a AAV6 donor used for KI experiments. All enzymes were tagged with hemagglutinin tag (3xHA). b Transcript upregulation of different enzymes in targeted HUDEP-2 upon differentiation (qPCR, n = 2, mean). Fold increase is indicated. c Representative western blot detecting different enzymes (HA-tag and anti-β tubulin) of targeted HUDEP-2 upon differentiation (n = 2). d KI efficiency of LAL-AAV6 in HSPCs at day 9 (lines indicate mean; AAV n = 4, AAV + RNP n = 5). e Representative western blot of LAL in HSPC lysates, supernatants and BFU-E in untreated (UT), transduced (AAV) and KI-HSPCs (AAV + RNP). Anti-HA tag and anti-β tubulin antibodies were used. f Quantification of secreted LAL during erythroid differentiation. Anti-LAL antibody was used. Data are shown as fold increase over untreated cells (UT, donor=2). g LAL activity in HSPC supernatants during erythroid differentiation, data are shown as fold increase over untreated cells (UT, n = 2). h Integration pattern in single BFU-E: no integration (0), monoallelic (1) and biallelic (2). Mean ± SD, donor = 2). i Uptake of erythroid-expressed LAL by WD fibroblasts, measured by western blot or activity assay (mean; n = 2). j Cholesterol levels in WD fibroblasts after incubation with conditioned medium from untreated (UT) or LAL KI-erythroblasts. WT fibroblasts are shown as control (n = 4; p = 0.003 WD-UT vs WD-LAL; p < 0.001 WD-LAL vs WT; one-way ANOVA, Tukey’s test). k Nile Red staining in WD fibroblasts after incubation with conditioned medium from untreated (UT) or LAL KI-erythroblasts. WT fibroblasts are shown as control. Black lines indicate mean ± SD; number of fibroblasts analyzed is indicated. (***p < 0.001 one-way ANOVA, Tukey’s test). Source data are in the Source Data file.

Fig. 5. KI HSPCs engraft NSG mice and express LAL upon erythroid differentiation.

a Schematic representation of engraftment experiments. b Percentage of human CD45+/HLA-ABC+ cells in hematopoietic organs of mice. BM = bone marrow. (UT and AAV n = 3; AAV + RNP n = 8). c GFP positive cells in peripheral blood of transplanted mice over time. Line indicates mean (n = 4 week 4; n = 8 week 8 and 16). d Edited cells in bone marrow of transplanted mice. GFP is expressed as percentage of CD45+ cells, mean is shown (*p = 0.012 AAV vs AAV + RNP, two-tailed Mann–Whitney test). e GFP positive cells in HSPCs (CD34), myeloid (CD33), B (CD19) and T (CD3) cells in bone marrow of transplanted mice. Each line represents one animal (n = 8). f GFP positive cells in HSPCs (CD34) and in a more primitive HSPC subset (CD34⁺CD38⁻) in bone marrow of transplanted mice (n = 4; ns: p = 0.8, two-tailed Mann Whitney test). g Western blot on CD34-derived BFU-E from mice engrafted with untreated (UT), transduced (AAV) or KI HSPCs (AAV + RNP). Anti-HA tag and anti-β tubulin antibodies were used (n = 2, BFU-E pooled from 2 mice). Source data are in the Source Data file.