Nonclinical evaluation of renizgamglogene autogedtemcel for SCD and TDT

Abstract

Sickle cell disease and transfusion-dependent β-thalassemia can be treated by fetal hemoglobin upregulation. Disruption of the distal BCL11A binding site at the HBG1/2 promoters to induce fetal hemoglobin using either SpCas9 or AsCas12a mimics multiple hereditary persistence of fetal hemoglobin mutations. AsCas12a showed higher editing efficiency, higher specificity, and increased fetal hemoglobin induction potential compared with SpCas9. AsCas12a-edited healthy donor CD34⁺ cells exhibited long-term, multi-lineage, and polyclonal engraftment in immunocompromised mice. High-level fetal hemoglobin induction was observed in erythroid progeny derived in vivo from edited healthy donor CD34⁺ cells and sickle cell disease or transfusion-dependent β-thalassemia donor CD34⁺ cells in vitro. In erythroid cells from patients with sickle cell disease, gene editing reduced sickling and improved rheological behaviors under deoxygenated conditions. In erythroid cells from patients with β-thalassemia, gene editing ameliorated ineffective erythropoiesis and significantly increased hemoglobin content per cell. A comprehensive off-target editing evaluation in edited CD34⁺ cells showed AsCas12a to be highly specific, with no off-target editing detected. In summary, editing CD34⁺ cells at the HBG1/2 promoter distal BCL11A binding site using AsCas12a phenocopied hereditary persistence of fetal hemoglobin mutations, demonstrating its potential as a gene editing approach for the treatment of β-hemoglobinopathies.

Graphical abstract

Upregulation of fetal hemoglobin is a known treatment strategy for β-hemoglobinopathies. This study by Marco and colleagues shows that editing CD34⁺ cells at the HBG1/2 promoter regions with AsCas12a is highly efficient and specific and robustly induces fetal hemoglobin. Gene editing of this genomic region with AsCas12a therefore has potential as an effective gene editing approach.

Introduction

Hemoglobin (Hb) is a metalloprotein tetramer composed of two α-like globin chains and two β-like globin chains. During human development, α- and β-like globin genes undergo an orderly switch in expression. In the fetus, the primary Hb expressed is fetal (HbF, α2γ2), which is replaced by adult Hb (α2β2) postnatally. Insufficient or abnormal globin expression leads to hemoglobinopathies. Sickle cell disease (SCD) and β-thalassemia constitute the two major types of β-hemoglobinopathies. In SCD, substitution of glutamic acid with valine at position 6 of the β-globin (HBB) chain produces sickle Hb (HbS, α2β^s2). When deoxygenated, HbS polymerizes, causing red blood cells (RBCs) to become rigid and obstruct microvascular blood flow, leading to painful vaso-occlusive crises, progressive end organ damage, and reduced life expectancy.¹ In β-thalassemia, mutations result in decreased or absent HBB expression. The imbalance in α-globin (HBA) and HBB synthesis results in an excess of free unpaired HBA chains that form toxic aggregates in erythroid precursors and RBCs, causing ineffective erythropoiesis and hemolytic anemia.² Patients with the most severe form of β-thalassemia, termed transfusion-dependent β-thalassemia (TDT), require lifelong blood transfusions to survive.

Individuals with SCD and TDT are typically born asymptomatic owing to the high levels of HbF expressed early in life; disease manifestations occur as γ-globin (HBG) production declines.³ Reactivation of HbF, through treatments such as hydroxyurea, reduces the frequency of painful episodes in SCD⁴ and decreases transfusion requirements in TDT.⁵ Reactivation of HbF can also be achieved via genetic manipulation. Recently, exagamglogene autotemcel (exa-cel), an autologous hematopoietic stem and progenitor cell (HSPC) product consisting of CD34⁺ cells edited with clustered regularly interspaced short palindromic repeats (CRISPR)-Streptococcus pyogenes Cas9 (SpCas9)⁶ at the erythroid-specific enhancer of the BCL11A gene to upregulate HbF, has been approved for the treatment of SCD and TDT.⁷ BCL11A binds to the distal TGACCA motif at HBG gene (HBG1 and HBG2, denoted as HBG1/2) promoters to suppress HBG expression.⁸ Genome editing of the erythroid-specific enhancer of the BCL11A gene downregulates BCL11A expression in the erythroid lineage, resulting in the upregulation of HBG and HbF.⁹ In clinical studies where patients received exa-cel, 97% of patients with SCD were vaso-occlusive crisis-free for ≥12 consecutive months,¹⁰ and 91% of patients with TDT achieved transfusion independence.¹¹ The regulatory approval of exa-cel supports the potential safety and efficacy of genome editing-mediated HBG reactivation for the treatment of β-hemoglobinopathies.

We previously characterized the biological consequences of editing human CD34⁺ HSPCs at either the −110 distal BCL11A binding site at the HBG1/2 promoters or the GATA1 binding site at the +58 BCL11A erythroid enhancer using SpCas9 ribonucleoprotein (RNP).¹² We found that editing of the distal BCL11A binding site at the HBG1/2 promoters generated robust HbF induction, with no effect on erythropoiesis. Other groups have also showed promising HbF induction by using SpCas9 to edit the −110 distal BCL11A binding site at the HBG1/2 promoters.^13,14 Clinical studies were designed to evaluate an investigational HBG1/2 promoter-edited CD34⁺ cell therapy, renizgamglogene autogedtemcel (reni-cel), in patients with SCD and TDT. Reni-cel is composed of autologous HSPCs edited using an engineered and highly specific CRISPR-Acidaminococcus sp. Cas12a (AsCas12a)¹⁵ at the distal BCL11A binding site of the HBG1/2 gene promoters. Patients with SCD or TDT who received reni-cel demonstrated rapid and sustained increases in HbF and early correction of anemia.^16,17,18 Here, we describe the nonclinical evaluations that supported the first-in-human clinical trial for reni-cel.

Results

Selection of AsCas12a nuclease for development

Because the BCL11A binding site can be targeted by either SpCas9 or AsCas12a (Figure 1A), we performed nonclinical evaluations to determine whether SpCas9 or AsCas12a is better suited for further clinical development in β-hemoglobinopathies. The editing efficiency of a variant of AsCas12a, which was engineered to have high efficiency in addition to intrinsic high specificity,¹⁵ and of wild-type (WT) SpCas9⁶ RNPs was compared in mobilized CD34⁺ cells from healthy donors (Figure 1B; Table S1). Overall, the editing efficiency of AsCas12a trended higher than that of SpCas9 at all concentrations tested. Comparison of insertion or deletion (indel) profiles of AsCas12a- and SpCas9-edited CD34⁺ cells shows that both nucleases generated an array of non-homologous end joining and microhomology-mediated end joining (MMEJ) indels (Figure S1). The most frequent indel induced is an 18-bp MMEJ deletion by AsCas12a and a 13-bp MMEJ deletion by SpCas9 that mimics the natural hereditary persistence of HbF (HPFH) mutation. Aside from the top MMEJ deletions, indels generated by AsCas12a were larger than those generated by SpCas9, especially on the 3′ (right) side (Figures 1C and S2), and indels >3 bp induced more HbF than indels ≤3 bp (Figure S2B). Overall, indels generated by AsCas12a delete more bases of the BCL11A binding site than those generated by SpCas9. Consistent with this observation, AsCas12a was significantly more potent in inducing HbF than SpCas9 (Figure 1D; Table S2), although the results showed considerable variability, possibly due in part to the use of multiple donors. The editing rates in erythroid cells were comparable (Figure 1E). Additional information about which experiments shared donors in Figures 1D and 1E can be found in Tables S1 and S2.

Figure 1.

AsCas12a is superior to SpCas9 with regard to editing HBG1/2 promoters

(A) Target sequences for the AsCas12a and SpCas9 RNPs targeting the distal TGACCA BCL11A binding motif at the HBG1/2 promoters. (B) Healthy donor CD34⁺ cells were electroporated with AsCas12a RNP complexed at 2:1 gRNA to protein ratio or SpCas9 RNP complexed at 4:1 gRNA to protein ratio at the concentration indicated. Indel rates at 1 day after electroporation were determined. N values were n = 8–10 per concentration for AsCas12a, and n = 4–7 per concentration for SpCas9. Four donors were used for AsCas12a and six were used for SpCas9 (see Table S1). (C) Indel profiles of AsCas12a- or SpCas9-edited CD34⁺ cells. Shown are the normalized profiles (each with a maximum value of 1) for the detected deletions at each base on the target region. n = 4 for AsCas12a and for SpCas9. The red bar marks the distal BCL11A binding motif (TGACCA). (D) Healthy donor CD34⁺ cells electroporated with 4–8 μM AsCas12a or SpCas9 RNP and placed in erythroid differentiation conditions for 18 days. HbF was measured using reverse-phase ultra-high-performance liquid chromatography, where individual globin chains were eluted. HbF percentage is calculated as γ/(γ+β). N values were n = 22 for control, n = 16 for AsCas12a, and n = 13 for SpCas9. Ten donors were used for the control, 7 for AsCas12a, and 4 for SpCas9 (Table S2). (E) Healthy donor CD34⁺ cells electroporated with 4–8 μM AsCas12a or SpCas9 RNP and placed in erythroid differentiation conditions. Indel rates were determined at day 14 erythroid culture. N values were n = 8 for control, n = 12 for AsCas12a, and n = 11 for SpCas9. (F) Number of on- and off-target cut sites identified using Digenome-seq for 25 randomly selected matched sites plus the HBG1/2 target site for AsCas12a and SpCas9. The generalized target sequence and the PAM sequences for AsCas12a and SpCas9 are depicted. Each dot represents one sample. gRNA, guide RNA; HbF, fetal hemoglobin; Indel, insertions or deletions; RNP, ribonucleoprotein. Horizontal bar represents the mean value of the respective treatment group for (B)–(D). Two-way ANOVA was used for (B) and (F), and one-way ANOVA with Tukey’s multiple comparison tests was used for (C) and (D) to determine whether the mean values of the different treatment groups were statistically significantly different. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001.

The specificity of AsCas12a and SpCas9 was compared using in vitro digested whole-genome sequencing (Digenome-seq).¹⁹ Digenome-seq off-target cut sites were identified for AsCas12a and SpCas9 RNPs targeting the distal BCL11A binding site at the HBG1/2 promoters and a survey of 25 other “matched sites” randomly selected across the genome (Table S3). These random sites were chosen because they contained target sequences with the required protospacer adjacent motif (PAM) sequences for both AsCas12a and SpCas9 (a 5′ TTTV PAM for AsCas12a and a 3′ NGG PAM for SpCas9), enabling direct comparison of specificity for the same DNA sequence. Off-target cut sites were defined as those with cleavage activity within a mismatch or gap threshold of up to six bases. The number of Digenome-seq off-target cut sites varied depending on the target sequence, but AsCas12a consistently exhibited higher specificity than SpCas9 at all sites evaluated (Figure 1F; Table S4). Further experiments will have to be performed to verify whether the superior specificity of AsCas12a against SpCas9 is also observed in cells. At the HBG1/2 target site, Digenome-seq identified 179 off-target candidates for SpCas9 compared with only 16 for AsCas12a.

Based on the higher editing efficiency, HbF induction, and specificity, AsCas12a was selected to target the BCL11A binding site at the HBG1/2 promoters for further development.

In vitro characterization of AsCas12a-edited healthy donor CD34⁺ cells

CD34⁺ cells are heterogeneous populations comprising long-term hematopoietic stem cells (LT-HSCs), multipotent progenitors (MPPs), and lineage-restricted progenitors.²⁰ LT-HSCs exhibit long-term reconstitution capacity and are able to differentiate into all lineages to maintain the adult hematopoietic system.²⁰ To understand whether AsCas12a can efficiently edit different subpopulations of CD34⁺ cells, editing rates were determined in phenotypic common myeloid progenitors (CMPs), MPPs, and HSCs at day 2 after electroporation (Figure 2A). High levels of indels were achieved by AsCas12a in all three subpopulations. Although higher levels of editing were detected in CMPs and MPPs, indel levels above 80% were attained in phenotypic HSCs. These data suggested that electroporating CD34⁺ cells with AsCas12a RNP targeting the HBG1/2 promoters can result in efficient editing of HSCs, the cell population that is critical for the durability of the autologous gene-edited cell therapy for the treatment of β-hemoglobinopathies.

Figure 2.

AsCas12a edits HSPCs efficiently and leads to edited allele-dependent HbF upregulation

(A) Healthy donor CD34⁺ cells were electroporated with AsCas12a RNP at 3–8 μM. At day 2 after electroporation, common myeloid progenitor (CMP, CD34⁺CD38⁺CD123⁺CD45RA⁻), multi-potent progenitor (MPP, CD34⁺CD38^lowCD90⁻ CD45RA⁻), and hematopoietic stem cells (HSCs, CD34⁺CD38^lowCD90⁺CD45RA⁻) were flow-sorted, and indel levels were determined. Indel levels were determined at day 2 after electroporation; n = 3. (B) The proportion of cells with deletion of the 4.9-kb intervening region between HBG1 and HBG2 was evaluated in the total CD34⁺ cells or subpopulations at day 2 after electroporation using a droplet digital PCR assay; n = 3. (C) Healthy donor CD34⁺ cells were electroporated without RNP or with 2 μM AsCas12a RNP, and individual cells were cultured in erythroid differentiation conditions. Individual erythroid clones were genotyped and HbF levels were measured by reverse-phase ultra-high-performance liquid chromatography. HbF percentage is calculated as γ/(γ+β). HbF levels are collated based on the number of HBG alleles with indels that each clone had. The dotted horizontal line marks 30% HbF expression. Each dot represents one sample; solid horizontal lines represent the mean value of each group. N values were n = 233 for control and n = 180 for edited. PCR, polymerase chain reaction. Repeat-measures one-way ANOVA with Dunnett’s multiple comparison test was used for (A) and (B), and one-way ANOVA with Dunnett’s multiple comparison test was used for (C) to determine whether the mean values of the different groups were statistically significantly different. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001.

As the HBG1 and HBG2 promoters both contain the target sequence, simultaneous cleavage of both target sites can result in deletion of the 4.9-kb intervening fragment and loss of the HBG2 gene. A droplet digital polymerase chain reaction (ddPCR) assay was developed to measure the frequency of the 4.9-kb deletion. Evaluation of phenotypic CMPs, MPPs, and HSCs 2 days post-electroporation revealed that CMPs had the highest rate of 4.9-kb deletion (Figure 2B). Interestingly, although HSCs had similar levels of indels compared with total CD34⁺ cells, the rate of 4.9-kb deletion was significantly lower in HSCs than in total CD34⁺ cells. Editing with SpCas9 showed rates of the 4.9-kb deletion slightly higher than the ones obtained for AsCas12a (Figure S3).

As there are a total of four alleles of HBG genes per cell, we interrogated how the levels of HbF expression were impacted by the number of HBG alleles edited. A total of 413 erythroid colonies, each derived from a single CD34⁺ cell, were expanded in erythroid culture conditions. Colonies were then sequenced to determine allelic editing and globin chains analyzed to determine the levels of HbF expression (Figures 2C and S4, with additional information about 4.9-kb deletions and inversions). The number of HBG alleles with indels was proportional to the level of HbF induction. Although a single allele indel was sufficient to significantly increase the average HbF levels compared with controls, having two or more alleles with indels increased average HbF levels to >30% of total Hb, a threshold level accepted as sufficient to prevent SCD RBC sickling under hypoxic conditions.²¹

AsCas12a editing of SCD CD34⁺ cells reduces sickling and improves rheological behavior in their erythroid progeny

Mobilized peripheral blood CD34⁺ cells were obtained from patients with SCD and electroporated with HBG1/2-targeted AsCas12a RNP. AsCas12a RNP edited SCD CD34⁺ cells efficiently, achieving approximately 90% indels (Figure 3A). Genome editing of SCD CD34⁺ cells generated HbF levels that represented 50% of total Hb after erythroid differentiation (Figure 3B). To test whether this level of HbF induction is sufficient to inhibit RBC sickling, RBC progeny of unedited CD34⁺ and edited CD34⁺ cells were subjected to sodium metabisulfite-induced hypoxia. Average sickling frequency reduced from 38.3% in untreated SCD RBCs to 10.6% in edited SCD RBCs, which is a 72% reduction in sickling after editing SCD CD34⁺ cells (Figure 3C). To evaluate whether HbF induction decreases the rigidity and improves the deformability of SCD RBCs when deoxygenated, cultured RBCs were analyzed on the LoRRca ektacytometer²² where shear stress was applied under decreasing oxygen tension. In edited CD34⁺ cells from donors with SCD, RBC progeny were more resistant to hypoxia, and a lower oxygen tension under hypoxic conditions was required to induce sickling after editing compared with unedited CD34⁺ cells (Figure 3D). In addition, when deoxygenated, edited SCD RBCs were more flexible than untreated controls, as demonstrated by the higher minimum elongation index (EI) in the edited RBCs (Figure 3E). To assess whether reduced sickling and increased flexibility of SCD RBCs derived from edited CD34⁺ cells would lead to improved rheological behavior, cultured RBCs were flowed through a microfluidic platform mimicking the microvasculature. The edited SCD RBCs were able to maintain flow velocity significantly better than untreated SCD RBCs under a range of reduced oxygen tension (Figure 3F). These data suggest that edited SCD RBCs have improved deformability that would allow them to better traverse through microvasculature.

Figure 3.

Efficient editing of SCD CD34⁺ cells by AsCas12a results in amelioration of disease phenotypes in their erythroid progeny

CD34⁺ cells from four donors with SCD were electroporated with 6 μM of the HBG1/2-targeted AsCas12a RNP. One SCD sample had a sufficient number of cells for two independent electroporations. (A) Indel levels were measured at day 3 after electroporation; n = 5. (B) CD34⁺ cells were placed in erythroid differentiation conditions for 18 days, and HbF was measured by reverse-phase ultra-high-performance liquid chromatography. HbF percentage is calculated as γ/(γ+β); n = 5. (C) Cultured SCD erythrocytes were incubated with sodium metabisulfite, and the percentage of cells that sickled was determined; n = 4. (D) Cultured SCD erythrocytes were analyzed on a LoRRca ektacytometer to measure deformability under shear stress, expressed as elongation index, when subjected to decreasing levels of oxygen. The point of sickling, representing the relative oxygen pressure when the SCD RBCs started to sickle during deoxygenation, is plotted for each SCD RBC sample; n = 5. (E) The minimum elongation index of each SCD erythrocyte sample, approximating the flexibility of erythrocytes when deoxygenated, is shown; n = 5. (F) Rheological behavior of cultured RBCs under varying concentrations of oxygen evaluated using a microfluidic platform. The percentage velocity drop was calculated based on the differences between velocity at a specified oxygen concentration and at atmospheric levels of oxygen (21%). N values were n = 6 untreated SCD samples, n = 6 RNP-treated SCD samples, and n = 4 untreated normal samples. RBC, red blood cell; SCD, sickle cell disease. For (A)–(E), each dot represents one sample, and the horizontal line represents the mean value for the treatment group. For (C)–(E), each line connects the untreated and RNP-treated cells from the same donor in the same experiment. For (F), the mean ± standard deviation is shown for each treatment group at the specified oxygen tension. Unpaired Student’s t test was performed for (B), paired Student’s t test was performed for (C)–(E), and two-way ANOVA was performed for (F) to determine whether the differences between edited samples and unedited samples are statistically significant. ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001.

AsCas12a editing of TDT CD34⁺ cells improves erythropoiesis and increases Hb content

Mobilized peripheral blood CD34⁺ cells were obtained from patients with TDT and electroporated with HBG1/2-targeted AsCas12a RNP, where approximately 90% indels were achieved (Figure 4A). As a hallmark of TDT is ineffective erythropoiesis,² TDT CD34⁺ cells were placed in erythroid culture conditions to evaluate cellular health and terminal maturation over the course of the culture. Unedited and edited TDT CD34⁺ cells were able to commit to erythroid lineage comparably, with >98% cells expressing erythroid marker CD235a by day 18 of culture (Figure 4B). The percentage of cells reaching terminal maturation and expelling their nuclei steadily increased over time for both unedited and edited samples (Figure 4C). However, edited TDT cells had improved cell viability (Figure 4D), so a larger fraction of erythroid cells survived to terminal maturation compared with unedited cells, nearly doubling the frequency of enucleated cells by day 18 (Figure 4C). Similar data were obtained from additional lots of CD34⁺ cells from two independent TDT donors (Figure S5) and three healthy donors (Figure S6).

Figure 4.

Efficient editing of TDT CD34⁺ cells by AsCas12a results in improved erythropoiesis and increased hemoglobin content

CD34⁺ cells from three donors with TDT were electroporated with 6 μM of the HBG1/2-targeted AsCas12a RNP. Their genotypes were donor 1 (compound heterozygous β⁰/β⁺), donor 2 (homozygous β⁰/β⁰), and donor 3 (compound heterozygous β⁰/β⁺). See more details in Table S9. (A) Indel levels were measured at day 3 after electroporation; n = 4, as each TDT sample had sufficient cells for two independent electroporations. (B) TDT CD34⁺ cells were cultured in erythroid differentiation conditions, and the expression of erythroid marker CD235a was measured by flow cytometry throughout the culture. (C) Terminal maturation of TDT erythroid cells was determined based on the frequency of erythroid cells that have expelled their nucleus by flow cytometry analysis throughout the culture. (D) Frequency of TDT erythroid cells that were non-viable throughout the course of the 18-day culture was determined by flow cytometry. (E) Globin mRNA levels were quantified and normalized to GAPDH levels by NanoString nCounter in day 14 erythroid cells. (F) Globin chains in day 18 erythroid cells were determined by reverse phase ultra-high-performance liquid chromatography. The amount of globin chain and total hemoglobin content per cell was calculated by plotting the area under the curve against a standard curve generated using a titration series of lysed RBCs with known hemoglobin concentration. TDT, transfusion-dependent β-thalassemia. Each dot (A) represents one independent electroporation. (B)–(D) Mean ± standard deviation of technical triplicates at each time point. For (B)–(D), three independent TDT donors were tested, and data from one representative donor are shown. Data from two additional TDT donors can be found in Figure S5, and from three additional healthy donors in Figure S6. Each dot in (E) and (F) represents the value from each technical replicate. The horizontal lines in (A), (E), and (F) represent the mean values of the treatment group. Two-way ANOVA was performed for (B)–(D), and Student’s t test was performed for (E) and (F) on each TDT donor to determine whether the differences between edited samples and unedited samples are statistically significant. ∗∗∗∗p < 0.0001.

β-Thalassemia is caused by a quantitative deficiency of HBB synthesis, resulting in erythrocytes that are hypochromic and, more commonly, prone to ineffective erythropoiesis as a consequence of a globin chain imbalance and low Hb tetramer content. Unedited and edited TDT CD34⁺-derived erythroid cells were also assessed for globin synthesis. Day 14 cultured erythroid cells were harvested for globin mRNA quantification, and day 18 cultured erythroid cells were harvested for globin protein analysis. Editing of TDT CD34⁺ cells resulted in significantly increased HBB, HBG, and HBA mRNA levels in erythroid cells as compared with their respective unedited controls (Figure 4E). Significantly reduced δ-globin mRNA levels were detected in edited samples. Globin protein subunit analysis detected no β-globin present in the cell lysates (Figure 4F). Editing of TDT CD34⁺ cells resulted in significantly increased HBG and HBA per cell compared with their respective unedited controls. The overall increased HBA level is indicative of its stabilization by HBG, which not only improved the survival of erythroid cells (Figure 4D) but also increased the total Hb content per cell (Figure 4F). On average, an approximately 2-fold increase in total Hb content per cell was observed in edited TDT erythroid cells compared with unedited TDT erythroid cells (Figure 4F). These data suggest that editing at HBG1/2 promoters by AsCas12a has the potential to significantly ameliorate globin chain imbalance and anemia associated with TDT by improving erythropoiesis and increasing total Hb content per cell.

Edited healthy donor CD34⁺ cells engraft immunodeficient NBSGW efficiently and produce high levels of HbF in their erythroid progeny

Mobilized CD34⁺ cells from three healthy donors were electroporated with 0.5–8 μM of HBG1/2-targeted AsCas12a RNP and infused into immunodeficient NOD.Cg-Kit^W−41J Tyr⁺ Prkdc^scid Il2rg^tm1Wjl (NBSGW) mice^23,24 to evaluate the engraftment potential of edited CD34⁺ cells. At the time of infusion, indel levels of CD34⁺ cells ranged from 46.52% to 89.82% for doses ranging from 0.5 to 8 μM (Figure 5A). At 16 weeks after infusion, high long-term indel levels were achieved in the bone marrow (BM) of all groups of mice that received cells generated with 4–8 μM RNP. When comparing the indel levels at week 16 post-infusion to those in the CD34⁺ cells prior to infusion, no reduction in indel levels was observed except at a low RNP concentration (0.5 μM), demonstrating that AsCas12a can edit LT-HSCs efficiently ex vivo. A reduction in the fraction of MMEJ indels was observed at 16 weeks (Figure S7). In addition to unfractionated BM, human CD19⁺ B cells, CD15⁺ neutrophils, CD235a⁺ erythroid cells, and lineage (Lin)⁻CD34⁺ HSPCs were also sorted from mouse BM for indel assessment. Indel levels were comparable across all lineages (Figure 5B). Correlation analyses show that the indel levels in committed lineages highly resembled those in their corresponding HSPCs (Figure 5C), supporting the notion that the edited HSPCs are comparable to unedited HSPCs in producing downstream progeny with no growth advantages or disadvantages.

Figure 5.

AsCas12a-edited healthy donor CD34⁺ cells retain HSC functionalities and generate erythroid cells with high levels of HbF expression in vivo

Healthy donor CD34⁺ cells were electroporated with AsCas12a RNP at the concentration indicated, and each NBSGW mouse received 1 × 10⁶ cells via intravenous tail vein injection. Animals were euthanized at 16 weeks after engraftment. (A) Indel levels were determined in the CD34⁺ cells at 1 day after electroporation prior to infusion (pre) and in mouse BM. (B) Indel levels were determined in human CD19⁺ B cells, CD15⁺ neutrophils, CD235a⁺ erythroid cells, and Lin⁻CD34⁺ HSPCs sorted from mouse BM. (C) Data from three independent in vivo experiments were compiled and correlation analyses were conducted for B cells, erythroid cells, or neutrophils against HSPCs. The black angled line represents the identity line (y = x). (D) The 4.9-kb deletion levels were determined in the CD34⁺ cells at 1 day after electroporation prior to infusion (pre) and in mouse BM. (E) Human chimerism percentage calculated as hCD45⁺/(h+m)CD45⁺ in BM. (F) Human lineage distribution in BM. B cell percentage is calculated as hCD19⁺/hCD45⁺, neutrophil percentage is calculated as hCD15⁺/hCD45⁺, erythroid percentage is calculated as hCD235a⁺/total cell, and HSPC percentage is calculated as hCD34⁺/hCD45⁺. (G) Mouse BM was plated in semi-solid methylcellulose media to evaluate the frequency of colony-forming unit granulocyte and monocyte (CFU-GM), burst-forming unit erythroid (BFU-E), and CFU-granulocyte, erythrocyte, monocyte, megakaryocyte (GEMM). (H) HbF expression by human erythroid cells flow sorted from mouse BM. HbF was measured using reverse-phase ultra-high-performance liquid chromatography, where individual globin chains were eluted. HbF percentage is calculated as γ/(γ+β). (I) Frequencies of HbF⁺ human erythrocytes in mouse BM as determined by immunostaining. For (A)–(I), each bar represents one pre-infusion sample; each dot represents one mouse sample. Horizontal bar represents the mean value of the respective treatment group. For (E)–(I), one-way ANOVA with Dunnett’s multiple comparison test was performed for experiments 1 and 2, and Student’s t test was performed for experiment 3 to determine whether the mean values of the treatment groups were statistically significantly different from the control groups. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗∗p < 0.0001.

The frequencies of the 4.9-kb deletion were evaluated in pre-infusion CD34⁺ cells and sorted long-term engrafted HSPCs. Deletion of the 4.9-kb fragment occurred less frequently in long-term engrafted HSPCs than in pre-infusion samples, reduced by approximately 50% (Figure 5D). This contrasts with the 16-week indel levels, which were similar to or higher than pre-infusion indel levels (Figure 5A). It should be noted that the indels and 4.9-kb deletion data obtained in vivo at week 16 post-infusion (Figure 5D) are consistent with the data from the ex vivo sorted phenotypic HSCs at day 2 post-electroporation (Figures 2A and 2B). The lower frequency of 4.9-kb deletion observed in HSCs is probably due to the less efficient utilization of the MMEJ DNA repair pathway by LT-HSCs.²⁵

In addition to the editing levels, the engraftment of various human lineages was evaluated by flow cytometry. Human CD45⁺ cells constituted most of the mouse BM (Figure 5E). Most of the human cells were B cells, although neutrophils, erythroid cells, and immature HSPCs were also detected (Figure 5F). There were no consistent differences in the human lineage engraftment between mice that received control unedited cells and mice that received edited cells. There were some differences in long-term erythroid frequencies between mice that received unedited cells and edited cells in one experiment. However, such differences were not detected in the other two experiments where edited cells were generated with higher RNP concentrations and hence were likely incidental. Furthermore, clonogenic potential of human cells in mouse BM was also evaluated. At 16 weeks post-infusion, unedited and edited CD34⁺ cells generated comparable frequencies of myeloid progenitors, including colony-forming unit granulocyte and monocyte (CFU-GM), burst-forming unit erythroid (BFU-E), and CFU-granulocyte, erythrocyte, monocyte, megakaryocyte (GEMM) in mouse BM (Figure 5G), further demonstrating that the editing does not impact HSPC functionalities.

HbF induction in long-term generated erythroid cells was evaluated in CD235a⁺ erythroid cells sorted from BM. Consistent with sustained indel levels detected in long-term engrafted HSPCs and derived erythroid cells (Figure 5B), robust HbF induction was achieved in erythroid cells derived from edited CD34⁺ cells after 16 weeks of engraftment (Figure 5H). Long-term HbF expression was pan-cellular (Figure 5I), which is critical for its clinical benefit.²¹

Edited healthy donor CD34⁺ cells lead to polyclonal engraftment in immunodeficient NSG mice

Potential tumorigenicity of edited CD34⁺ cells was evaluated in immunodeficient NOD.Cg-Prkdc^scid Il2rg^tm1Wjl (NSG) mice. Three independent lots of healthy donor CD34⁺ cells were electroporated with 8 μM HBG1/2-targeted AsCas12a RNP, which resulted in indel levels of approximately 90% (Figure S8A), infused into irradiated NSG mice, and followed for 20 weeks. The RNP concentration was selected to ensure that the maximum level of editing was achieved. Mock-electroporated CD34⁺ cells from the same donors were used as controls. Assessment of toxicity and tumorigenicity was based on mortality, clinical observations, body weight, engraftment, clonality, hematology, clinical chemistry, and anatomic pathology. All mice showed human engraftment and multilineage differentiation without differences between groups that received cells electroporated with AsCas12a RNP and groups that received control cells (data not shown). There were no toxicities or evidence of tumorigenicity related to the infusion of edited CD34⁺ cells in any of the analyses conducted (data not shown). Edited cells from three donors were analyzed using UDiTaS,²⁶ and only low levels of resections (<1.58%) and translocations (<0.23%) were detected (Table S5).

Polyclonality was assessed by evaluating the diversity of unique indels in pre-infusion CD34⁺ cells, in whole blood (WB) throughout the course of the experiment and in BM at necropsy. Editing with AsCas12a RNP resulted in a diverse indel profile. Using a frequency threshold of ≥0.1%, around 100–150 unique indels were identified in the pre-infusion samples (Figure 6A). Albeit with some fluctuations, this repertoire remained highly polyclonal throughout the course of the study in WB and in terminal BM. The most frequent indel prior to infusion, an 18-bp MMEJ deletion (Figure S8A), persisted in vivo but at lower frequencies, which ranged from 7% to 19% in WB and 5% to 15% in BM at experiment week 21 (Figure S8B), although some animal-to-animal variabilities were observed (Figure S8C). This was associated with an increased diversity of frequent indels (defined by a ≥1% frequency cutoff), with approximately 10 unique indels in the pre-infusion samples, compared to approximately 20 unique indels in vivo (Figure 6B). Clonal dominance was assessed by following the predominant indels frequency (using the top two indels combined, which may be carried by a single clone on both alleles). The top two indels contributed to approximately 25% of the reads prior to infusion and stabilized at approximately 20% by 20 weeks post-infusion (experiment week 21), demonstrating no overt clonal outgrowth (Figure 6C).

Figure 6.

AsCas12a-edited healthy donor CD34⁺ cells lead to polyclonal engraftment in irradiated NSG mice

Three batches of healthy donor CD34⁺ cells were electroporated with 8 μM of the HBG1/2-targeted AsCas12a RNP, intravenously infused into irradiated NSG mice, and followed for 20 weeks. Unique indels were identified and tracked in WB throughout the course of the experiment and in BM at the necropsy. (A) Number of unique indels that were detected at a frequency of ≥0.1% of the reads. (B) Number of unique indels that were detected at a frequency of ≥1% of the reads. (C) Combined frequency of the top two indels. Data are shown as mean ± standard deviation. Groups 1, 3, and 5 received unedited control cells; therefore, data are not shown. BM, bone marrow; F, female (shown as open circle); M, male (shown as closed circle); Pre, pre-infusion sample (shown as closed square); WB, whole blood. Numbers that precede gender denote experiment week, with the week of infusion defined as experiment week 1. n = 12–16 per gender.

Specificity evaluation demonstrates HBG1/2-targeted AsCas12a RNP to be specific

To evaluate potential off-target editing, a two-phased approach was taken, comprising a discovery phase and a verification phase. In the discovery phase, three orthogonal methods were employed to identify potential off-target candidates: an in silico method (CRISPR-Cas-aware aligner for in silico off-target search),²⁷ a biochemical method (Digenome-seq),¹⁹ and a cellular assay (genome-wide, unbiased identification of DNA double-stranded breaks enabled by sequencing).²⁸ A total of 319 potential off-target candidate sites were identified, 90% of which were derived from in silico predictions (Figure 7). In the verification phase, a multiplex rhAmpSeq panel was established to evaluate whether the off-target editing did occur in edited CD34⁺ cells from three healthy donor lots, six SCD patient lots, and three TDT patient lots. Of the 319 identified sites, 310 were successfully evaluated, and no off-target editing was observed, supporting the high-specificity and low-genotoxicity risk of the HBG1/2-targeted AsCas12a RNP (Table S6). A risk assessment was conducted for the remaining nine sites that could not be evaluated. These sites were located in intergenic or intronic regions, and their target genes did not appear in the COSMIC Cancer Gene Census database. No overt risk to patients was identified if off-target editing had occurred at these sites (Figure 7).

Figure 7.

Off-target evaluation shows that the HBG1/2-targeted AsCas12a is highly specific

Off-target evaluation of the HBG1/2-targeted AsCas12a RNP is divided into a discovery phase and a verification phase. Three orthogonal methods were used during the discovery phase. Off-target candidates identified during the discovery phase were evaluated in lots from healthy donors, lots from patients with SCD, and lots from patients with TDT during the verification phase, and no off-target editing was detected. Risk assessment was performed on unevaluated sites within the potential off-target candidates.

Discussion

SCD and TDT are two inherited blood disorders caused by genomic variations affecting the function or the production of adult β-globin. Reactivation of developmentally silenced fetal HBG can ameliorate both diseases. Silencing of HBG is mediated partially through binding of the transcription factor BCL11A to the distal TGACCA motif at the HBG1 and HBG2 promoters. Recently, exa-cel, a BCL11A erythroid enhancer-edited autologous HSPC product, was approved for the treatment of SCD and TDT. Although exa-cel has been shown to be effective, particularly in reducing acute manifestations, analyses of the long-term safety and durability of clinical benefit are ongoing, and as such, there remains a need for novel cell medicines that offer the potential for expanded therapeutic benefit to patients. To this end, nonclinical evaluation comparing HBG1/2-edited and BCL11A-edited CD34⁺ cells demonstrated that targeting HBG1/2 promoters can be an alternative to editing the BCL11A erythroid enhancer for HbF induction, reproducing the naturally occurring HPFH phenotype and without the risk of transcriptome perturbation or reduced erythroid production.¹²

In addition to the target site selection, the choice of gene editing enzyme also plays a critical role in ensuring an optimized approach for therapeutic applications. Currently, four classes of nucleases (meganucleases, zinc finger nucleases, transcription activator-like effector nuclease, and CRISPR-Cas) have been tested in the clinic (ClinicalTrials.gov). CRISPR-Cas, with its easily customizable features and a multitude of variants to target different PAM sequences, has been under extensive nonclinical evaluations for potential clinical applications. SpCas9 and AsCas12a are two classes of Cas enzymes that have been evaluated in the clinic. There are major differences between the two enzymes that may impact the consideration of selecting one over the other for a specific application, such as the PAM sequence (TTTV for AsCas12a and NGG for SpCas9), staggered vs. blunt end cleavage, and the size of the guide RNA (gRNA).^6,29,30 Importantly, AsCas12a is less tolerant of mismatches in the RNA/DNA hybrid within the enzyme active site (R-loop) compared with SpCas9, conferring upon it better specificity.^31,32 WT AsCas12a suffers from having lower potency compared with WT SpCas9, but protein engineering has overcome the potency limitation while retaining the high specificity of the enzyme.¹⁵ Using a variant of the engineered AsCas12a, we demonstrated that the HBG1/2-targeted AsCas12a RNP is more potent in editing CD34⁺ cells (Figure 1B), more efficient in disrupting the BCL11A binding motif (likely owing to larger indels induced by staggered cut) (Figure 1C), and leads to higher HbF induction than SpCas9 (Figure 1D). A limitation of this study is that different donors were used for the AsCas12a and SpCas9 experiments.

The Digenome-seq analysis of 25 matched sites plus the HBG1/2 target sites further substantiated the superior specificity of engineered AsCas12a compared with WT SpCas9 (Figure 1F). The specificity of the HBG1/2 targeted AsCas12a RNP was further evaluated thoroughly using a two-phased approach where in silico, biochemical, and cellular assays were used to broadly identify potential off-target candidates, which were subsequently interrogated by sequencing in edited CD34⁺ cells from healthy donors and patients with SCD and patients with TDT. No evidence of off-target editing was detected. We note that other authors have tested the specificity of the SpCas9 guide targeting the BCL11A binding motif using a combination of bioinformatic, in vitro, and in vivo approaches, and no off-target sites were verified.¹⁴

The objective of editing the HBG1/2 promoters is to induce HBG and subsequently HbF expression to counter sickling mediated by HbS polymerization in SCD and to reduce ineffective erythropoiesis caused by excess unpaired HBA and increase Hb content in TDT. Although the simultaneous cleavage of the HBG1 and HBG2 promoters can lead to deletion of the intervening sequence and result in the removal of HBG2, the frequency of the 4.9-kb deletion induced by AsCas12a RNP did not prevent induction of therapeutically relevant levels of HbF (Figures 1D and 3B), given that having two edited alleles was sufficient to lead to the expression of ≥30% HbF (Figure 2C). Indeed, the findings of significant reduction in sickling, improved resistance to hypoxic conditions, and improved flexibility and flow through microfluidic channels under hypoxia demonstrated phenotypic improvement in erythroid cells derived from SCD CD34⁺ cells as a result of gene editing (Figure 3). These data suggest that receiving the AsCas12a-edited CD34⁺ cells as a treatment may result in reduction in vaso-occlusive crises, hemolysis, and other associated complications caused by sickling for patients with SCD. This hypothesis was substantiated in the RUBY clinical trial (NCT04853576), where normalization of Hb, reduction or normalization of hemolytic markers, and reduction in painful vaso-occlusive crises were reported in patients with SCD who received reni-cel.^16,18 Furthermore, nonclinical evaluation of edited TDT CD34⁺ cells found significantly improved erythropoiesis, including reduced cell death and better terminal maturation, as well as significantly increased Hb content per cell (Figure 4). These data suggest that this approach has the potential to address ineffective erythropoiesis and anemia, both of which are hallmarks of TDT. Consistent with the nonclinical observations, patients with TDT who received reni-cel infusion in the EdiThal clinical trial (NCT05444894) remained transfusion independent for a range of 1.2–9.9 months.¹⁷

CD34⁺ cells are a heterogeneous cell population that contains long-term repopulating cells and short-term progenitors. For the treatment to be durable, the LT-HSCs must be efficiently edited. Editing of LT-HSCs by the HBG1/2-targeted AsCas12a RNP was evaluated both in vitro by flow-activated sorting of phenotypic LT-HSCs based on cell surface marker expression (Figure 2) and in vivo in unconditioned NBSGW mice (Figure 5) and in irradiated NSG mice (Figure 6). Both in vitro and in vivo data demonstrated that the LT-HSCs were efficiently edited, with indels >80%. Interestingly, the editing levels in bulk CD34⁺ cells at 1 day after electroporation were indicative of, or even underestimated, the editing levels observed in the long term (Figure 5A). It should be noted that the 4.9-kb deletion frequency was lower in LT-HSCs than in bulk CD34⁺ cells (Figures 2B and 5D), potentially due to the less-efficient utilization of the MMEJ DNA repair pathway by LT-HSCs.²⁵ AsCas12a-edited HSPCs retained their functionality and were able to lead to multi-lineage engraftment with no lineage skewing observed, which is consistent with the nonclinical observation previously obtained with SpCas9-edited HSPCs¹² and the clinical observations that patients who received reni-cel achieved rapid tri-lineage engraftment and maintained stable editing in their peripheral blood and BM for a mean of 6.2 months post-infusion.^16,17,18 Importantly, long-term, robust HbF induction in a pan-cellular fashion was observed nonclinically (Figures 5H and 5I), which would predict a potentially durable clinical benefit for patients with β-hemoglobinopathies.

One of the major concerns of developing genome-edited medicine is inadvertently generating leukemia-initiating cells. This risk was evaluated using the well-characterized irradiated NSG mouse model, known to support long-term engraftment of CD34⁺ cells^33,34; propagation of human tumor- and leukemia-initiating cells^35,36,37,38; and to exhibit a prolonged lifespan with a low incidence of spontaneous murine tumors.³⁴ Throughout the 20-week experiment, no evidence of human neoplasm was observed following the infusion of HBG1/2-edited human CD34⁺ cells, and a highly polyclonal engraftment was observed in both WB and BM (Figures 6A and 6B). No overt clonal outgrowth was detected based on predominant indel analysis (Figure 6C). These data support that this gene editing approach carries a low risk of tumorigenicity, although this remains to be substantiated in a clinical setting.

In summary, these studies demonstrate that AsCas12a RNP efficiently edits the HBG1 and HBG2 promoter distal BCL11A binding sites in normal, SCD, and TDT CD34⁺ cells. The editing strategy is compatible with maintaining the function of the LT-HSCs. Efficient editing in LT-HSCs and robust pan-cellular HbF expression predict a long-lasting therapeutic benefit. The levels of HbF induced are likely to be therapeutically relevant, as they are sufficient to ameliorate the sickling and improve rheological behaviors of SCD RBCs, improve erythropoiesis, and increase Hb content in TDT erythroid cells. Off-target evaluation and tumorigenicity studies did not identify an overt risk of genotoxicity or generation of cells with uncontrolled proliferative capacity. These data support the development of a therapy for potential treatment of β-hemoglobinopathies where patients’ LT-HSCs cells are edited at the HBG1 and HBG2 promoter distal BCL11A binding site using AsCas12a.

Materials and methods

Editing of CD34⁺ cells

CD34⁺ cells were cultured for 2 days at 37°C, 5% carbon dioxide (CO₂) in X-VIVO 10 media (Lonza Biologics, Walkersville, MD), supplemented with 1× GlutaMAX (Thermo Fisher Scientific, Waltham, MA), 100 ng/mL stem cell factor (SCF), 100 ng/mL thrombopoietin, and 100 ng/mL FMS-like tyrosine kinase 3 ligand (Flt3L) (all from PeproTech, Rocky Hill, NJ) at a density of 1 × 10⁶ cells/mL. After culture, CD34⁺ cells were washed and resuspended in MaxCyte electroporation buffer (MaxCyte, Gaithersburg, MD) containing pre-complexed SpCas9 (2:1 or 4:1 gRNA to protein ratio) or a variant of AsCas12a Ultra¹⁵ (Integrated DNA Technologies, Coralville, IA) (complexed at 2:1 gRNA to protein ratio) at 0.5–8 μM/mL as specified in the figure legends to a final cell density of 10 × 10⁶–62.5 × 10⁶/mL except if the cell number was limiting when patient materials were utilized. The target site sequences used are shown in Figure 1A. Electroporation was carried out using a MaxCyte GT electroporation device, per the manufacturer’s instructions. Cells were cultured in supplemented X-VIVO 10 media for up to 3 days post-electroporation.

HSPC sort

CD34⁺ cells at day 2 post-electroporation were stained with anti-human CD34-BV421, anti-human CD38-BV605, anti-human CD123-BV711, anti-human CD45RA-fluorescein isothiocyanate (FITC), and anti-human CD90-R-phycoerythrin-cyanine7 (PE-Cy7) (all from BioLegend, San Diego, CA) for 30 min at 4°C. Samples were washed with phosphate-buffered saline (PBS)-supplemented 0.5% bovine serum albumin (BSA) (Miltenyi Biotec, Auburn, CA) and resuspended in PBS with 7-amino-actinomycin D (eBioscience, San Diego, CA) for dead cell exclusion. HSCs (CD34⁺ CD38^lowCD90⁺CD45RA⁻), MPPs (CD34⁺CD38^lowCD90⁻CD45RA⁻), and CMPs (CD34⁺CD38^highCD123⁺CD45RA⁻) were sorted on a BD FACSAria Fusion cell sorter (BD Biosciences, Franklin Lakes, NJ).

Determination of indels

Genomic DNA was extracted from samples using QuickExtract (Lucigen, Middleton, WI), per the manufacturer’s instruction. The indel percentage was quantified by next-generation sequencing on the Illumina MiSeq platform (Illumina, San Diego, CA), as described in Giannoukos et al.²⁶ Briefly, two rounds of amplification were performed: round one targeted the HBG1 and HBG2 promoter region (see Table S7; highlighted in bold are the specific target sequences), and round two added the full-length Illumina adapter sequence. The PCR product for both rounds of amplification was purified using (0.9×) Agencourt AMPure XP beads (Beckman Coulter, Indianapolis, IN), per the manufacturer’s protocol. The second round of PCR product purification was followed by 300- to 1,200-bp size selection on the BluePippin (Sage Science, Beverly, MA) and loaded on the Illumina MiSeq with 10% PhiX (Illumina). Analysis of indel rates was done as described in Bothmer et al.³⁹ To generate the deletion profiles in Figure 1C, the cigar strings for the reads in the alignment BAM were examined, and the percentage deletion for each base within the target region was calculated. All profiles were normalized to a maximum of one. To quantify the frequencies of the observed indels for polyclonality assessment (Figures 6 and S8), the reads were first parsed and assigned to HBG1 or HBG2 based on the canonical differences in the reference genome hg38. The reads assigned to HBG1 were discarded so that only reads from two alleles (both HBG2) were quantified. Deletions were classified as MMEJ if stretches of ≥2 bp were detected at the deletion boundary and repeated in the region immediately flanking the other end of the deletion.

4.9-kb deletion frequency determination

To assess the frequency of the 4.9-kb fragment deletion, two ddPCR reactions, one for reference amplicon and one for on-target amplicon, were designed. The sequences of the primers and probes are listed in Table S8. Extracted genomic DNA, primers, and probes were added to the ddPCR Supermix (Bio-Rad, Hercules, CA) to a well of a 96-well PCR plate. Droplets were generated using a droplet generator (Bio-Rad). The plate was then moved to the thermocycler, and the following protocol was run: 1 cycle of 10 min at 98°C, 40 cycles of 30 s at 94°C, and 2 min at 94°C, followed by 1 cycle of 10 min at 98°C and hold at 4°C. Afterward, the plate was moved to the droplet reader (Bio-Rad) for quantification and counting of droplets that are (1) negative, (2) positive for both the on-target amplicon and the reference amplicon, (3) positive for the on-target amplicon only, and (4) positive for the reference amplicon only. The equation used to calculate the frequency of 4.9-kb deletions (Equation S1) can be found in the supplemental information.

In vitro erythroid differentiation

Erythroid cells were derived from CD34⁺ cells using three-step differentiation conditions as described in Giarratana et al.⁴⁰ with slight modifications. Briefly, CD34⁺ cells were placed in step 1 media consisting of Iscove’s modified Dulbecco’s medium supplemented with 1× GlutaMAX, 100 U/mL penicillin, 100 μg/mL streptomycin (all from Invitrogen, Waltham, MA), 5% human AB⁺ plasma (Octapharma, Paramus, NJ), 330 μg/mL human holo-transferrin, 10 μg/mL human insulin, 2 U/mL heparin (all from Sigma-Aldrich, St. Louis, MO), 3 U/mL recombinant human erythropoietin, 100 ng/mL SCF, and 5 ng/mL interleukin (IL)-3 (all from PeproTech) with (for data presented in Figure S1B) or without (for data presented in Figure 2) 1 μM/mL hydrocortisone (Sigma-Aldrich) at 2 × 10⁴ cells/mL. On day 7 of differentiation, cells were transferred at 1 × 10⁵ cells/mL to step 2 media, where IL-3 and hydrocortisone were removed. On day 11 of differentiation, cells were transferred to step 3 media at 1 × 10⁶ cells/mL, where SCF was removed. On day 14 of differentiation, cells were sub-cultured at 5 × 10⁶ cells/mL in step 3 media. For data presented in Figure 3, human AB⁺ plasma was replaced with KnockOut Serum Replacement (Invitrogen) in step 3 media. Cells were harvested on day 18 for downstream analyses. For in vitro erythropoiesis assessment of TDT CD34⁺ cells, the cell suspension was stained with anti-CD235a-FITC, NucRed (Invitrogen), and 4′,6-diamidino-2-phenylindole (DAPI) (BD Biosciences) throughout the culture to evaluate the frequency of cells that have committed to erythroid lineage (CD235a⁺/DAPI⁻), reached terminal maturation, and expelled their nuclei (NucRed⁻/CD235a⁺DAPI⁻), and the frequency of erythroblasts that were not viable (DAPI⁺/hCD235a⁺NucRed⁺). Cells were acquired using a Guava easyCyte flow cytometer (MilliporeSigma, Burlington, MA) and analyzed using GuavaSoft InCyte 3.3 (MilliporeSigma).

Single-cell erythroid clonal expansion

CD34⁺ cells, plated as a single cell per well in 384-well plates, were cultured in step 1 media for 7 days. On day 7, 30 μL step 1 media was replaced with 30 μL step 2 media. On day 11, individual erythroid clones were transferred from 384-well plates to 96-well plates preloaded with 100 μL of step 3 media per well. On day 14, cells were pelleted and resuspended in 150 μL of fresh step 3 media. For each colony, 10 μL was harvested for indel analysis. On day 18, cultures were harvested for Hb assessment. The HBG1 and the two HBG2 haplotypes could be distinguished by the presence of variants along the amplicon (indicated by their position and sequence):

• HBG1: 277:4_323G_359:CCT_369:T_376:G_408:T_450:C

• HBG2-Q: 277:4_323A_359:TCT_369:C_376:A_408:C_450:T

• HBG2-P: 277:4_323A_359:TCC_369:C_376:A_408:C_450:T

When the four HBG1 and HBG2 alleles are present, their contributions are HBG1 (50%), HBG2-Q (25%), and HBG2-P (25%). A single 4.9-kb deletion results in the disappearance of one of the HBG1 alleles, resulting in the following contributions: HBG1 (33%), HBG2-Q (33%), and HBG2-P (33%). A single 4.9-kb inversion renders one HBG1 and one HBG2 undetectable, resulting in HBG1 (50%) and HBG2-Q or HBG2-P (50%). Two 4.9-kb deletions result in HBG2-Q (50%) and HBG2-P (50%). One 4.9-kb deletion and one 4.9-kb inversion result in HBG2-Q or HBG2-P (100%).

The cells used to generate Figure S2 data had one HBG1 allele encoding AγI and the other encoding AγT, which differ at amino acid 75 (isoleucine vs. threonine) and can be separated by liquid chromatography. This enabled the direct association of indels at the specific HBG1 promoter with changes in expression of the corresponding Aγ variant.

Hb expression assessment by RP-UPLC

Expression of various Hb subunits was analyzed using reverse-phase ultra-performance liquid chromatography (RP-UPLC) modified from a method described by Masala and Manca.⁴¹ Erythroid cells were lysed in water and underwent four freeze-thaw cycles. Lysate was injected onto an Agilent 1290 UPLC system (Agilent, Santa Clara, CA) equipped with Waters ACQUITY UPLC Protein BEH C4 Columns (Waters, Milford, MA). Elution was obtained using a 20-min linear gradient of 38.75%–43.75% acetonitrile in water (MilliporeSigma) and trifluoroacetic acid (MilliporeSigma) at a constant concentration of 0.1%. Elution was followed at 220 nm with no reference wavelength. The globin chains were eluted in the following order: β, α, Gγ, and Aγ (Figure S9A). Area under the curve (AUC) of each peak approximated the relative abundance of each globin chain. Level of HbF expression was calculated as (Aγ + Gγ)/(Aγ + Gγ + β) (%). The Hb content per cell was determined based on a standard curve generated using whole peripheral blood with known Hb concentration from a healthy donor. The certificate of analysis provided by the vendor indicated 11 g/dL Hb and 4.08 × 10⁹ cells/mL RBC. We washed 4.9 mL blood twice with 40 mL PBS with a centrifugation speed of 500 × g for 5 min. The pellet was resuspended in 10.78 mL UPLC-grade water to bring the Hb concentration to 0.05 g/mL and lysed via four freeze-thaw cycles. A portion of the hemolysate was filtered through a 0.2-μm filter and further diluted to make a stock solution of 3 μg/μL Hb. From this stock, serial dilutions were prepared to make hemolysates, with Hb concentrations ranging from 0.1 to 0.7 μg/μL. The total AUC of β, α, Gγ, and Aγ chains combined was determined for each dilution, and a standard curve was generated (Figure S9B). A simple linear regression analysis was performed to generate an equation used to determine the Hb content per cell in experimental samples. In Figure S2B, the Aγ I was separated from the normal Aγ since they elute at different times.

Sickling analysis

Cultured RBCs derived from SCD CD34⁺ cells were mixed with sodium metabisulfite to a final concentration of 1%. One drop (approximately 20 μL) of this cell suspension was then placed on a microscopic slide, covered with a coverslip and the edges were sealed with nail varnish. Slides were stored at room temperature for 1–4 h before being imaged under a microscope and analyzed for morphological changes.

Deformability analysis

Fifteen million cultured erythrocytes after filtration were collected and washed with PBS-0.5% BSA. Cells were centrifuged at 500 × g for 5 min, resuspended in 1.5 mL OXY ISO solution (RR Mechatronics, Zwaag, the Netherlands), and mixed gently by inversion. Cell suspension was then loaded onto the LoRRca ektacytometer, where Oxygenscan was performed per the manufacturer’s instruction to measure the deformability of erythrocytes under shear stress when deoxygenated. Erythrocyte deformability was expressed as EI, and “point of sickling” represented the relative oxygen pressure when the SCD erythrocytes started to sickle and lost >5% of the EI during deoxygenation.

Microfluidic rheological assessment

Microfluidic rheological analysis of cultured RBCs derived from SCD CD34⁺ cells was performed at the University of Minnesota in Dr. David Wood’s laboratory, where microfluidic devices were fabricated. Photolithography techniques were used to create master molds on silicon wafers for soft replication with gas-permeable polydimethylsiloxane as described in Lu et al.⁴² The device consisted of a 15-μm blood layer, a 100-μm hydration layer, and a 150-μm gas layer overlain and bonded to a glass slide. Devices were mounted for image acquisition on a Zeiss Axio Vert.A1 (Carl Zeiss, Oberkochen, Germany) in a temperature-controlled environment at 37°C. A pressure regulator was used to create a constant driving force to flow blood samples through the channel at an average velocity of 700 mm/s (PCD-15PSIG, Alicat Scientific, Tucson, AZ). Syringe pumps flowed PBS through the hydration layer at a rate of 500 μL/h to prevent sample dehydration during the experiment. Oxygen conditions on the device were controlled via a gas mixing system that combines air (21% O₂, 5% CO₂, 74% N₂) and nitrogen (95% N₂, 5% CO₂) to achieve the desired oxygen concentration. Cultured RBCs, washed and resuspended in PBS to achieve a 20% target hematocrit, were drawn into the channel where velocity was recorded using a high-speed camera. The blood velocity through the channel was determined using Kanade-Lucas-Tomasi feature tracking in MATLAB (MathWorks, Natick, MA). Velocity data within the channel were collected during repeated cycles of high and low oxygen concentrations. Under a constant driving pressure, the velocity was allowed to reach steady state at atmospheric levels of oxygen (21% O₂) before the supplied oxygen concentration was suddenly dropped to a lower concentration. The average velocity was allowed to stabilize to a steady-state value and recorded during each cycle. The velocity distributions used for the analysis were determined by calculating the fractional drop in velocity between the low-oxygen velocity measurements and the median of the preceding 21% O₂ cycle.

Engraftment experiment using NBSGW mice

Animal studies were performed at the Charles River Accelerator and Development Lab (CRADL, Cambridge, MA). All animal study protocols were reviewed and approved by the CRADL Institutional Animal Care and Use Committee (IACUC; protocol no. 2022-1554). Female NBSGW mice no less than 5 weeks old were purchased from The Jackson Laboratory (Bar Harbor, ME). We administered 1 × 10⁶ CD34⁺ cells to each animal via intravenous tail vein injection. Mice were euthanized by CO₂ asphyxia at 16 weeks after infusion. BM cells were collected and stained with anti-mouse CD45-PE-CF594 (BD Biosciences), anti-human CD34-allophycocyanin (APC), anti-human CD19-PE, anti-human CD15-PE-Cy7, anti-human CD45-APC-Cy7 (all from BioLegend), and anti-human CD235a-FITC (Agilent). DAPI (BD Biosciences) was used for dead cell exclusion. Cells were acquired using a Guava easyCyte flow cytometer and analyzed using GuavaSoft InCyte 3.3. Human cell engraftment and lineage distribution, including chimerism (hCD45⁺/(hCD45⁺ + mCD45⁺)), B cell (hCD19⁺/hCD45⁺), neutrophil (hCD15⁺/hCD45⁺), erythroid cell (hCD235a⁺/total live cells), and HSPC (hCD34⁺/hCD45⁺) frequencies were determined. In addition to immunophenotyping, a separate aliquot of BM cell suspension was stained with anti-human CD45-BV785, anti-human CD15-BV650, anti-human CD34-PE-Cy7, anti-human CD19-PE, anti- Lin cocktail-APC (CD3, CD14, CD16, CD19, CD20, and CD56) (all from BioLegend), and anti-human CD235a-FITC. Human B cell, neutrophils, erythroid cells, and HSPCs (Lin⁻hCD15⁻hCD235a⁻hCD45⁺ hCD34⁺) were also sorted using FACSAria Fusion sorter (BD Biosciences) for indel analysis. An aliquot of sorted erythroid cells was also lysed for Hb assessment.

F cell determination

The frequency of HbF-expressing erythroid cells (F cell) in mouse BM was determined using a method described by Thorpe et al.⁴³ In brief, cells were fixed with 4% paraformaldehyde (Electron Microscopy Sciences, Hatfield, PA), permeabilized with ice-cold acetone (Sigma-Aldrich), washed with PBS-0.5% BSA, and stained with anti-human CD235a-PE (Agilent), anti-Hb-γ antibody-FITC (Santa Cruz Biotechnology, Dallas, TX), and NucRed. F cell percentage was calculated as HbF⁺/CD235a⁺NucRed⁻ × 100% (Figure S10).

Colony-formation assay

We added 240,000 BM cells to 3 mL MethoCult media H4435 (STEMCELL Technologies, Vancouver, BC, Canada). Triplicates of 1-mL aliquot were plated in SmartDish (STEMCELL Technologies) and cultured for approximately 2 weeks. At the end of the culture, wells were imaged with STEMvision (STEMCELL Technologies) and colony-forming cell frequency was determined per the manufacturer’s instructions.

Engraftment experiment using NSG mice

The engraftment experiment using NSG mice was conducted at The Jackson Laboratory (protocol no. 123119 IVSS). All animal study protocols were reviewed and approved by the IACUC. Three lots of mobilized CD34⁺ cells from three independent healthy donors were used for the experiment. Control articles were electroporated without RNP. Test articles were grown for 48 h, electroporated with 8 μM HBG1/2-targeted AsCas12a RNP using the MaxCyte CL-1.1, and grown for an additional 24 h. Control and test articles were cryopreserved and transferred to The Jackson Laboratory. Male and female NSG mice, approximately 6–7 weeks of age, were irradiated with 175 cGy from an X-ray irradiator source. At the time of infusion, each animal received 4 × 10⁶ cells via intravenous tail vein injection. The week of the injection was defined as experiment week 1. Body weight and clinical observations were performed weekly. WB was collected at experiment weeks 5, 9, 13, 17, and 21 for blood smear, flow cytometry, complete blood count (CBC), serum chemistry, and indel analysis. Animals were euthanized at experiment week 21. BM was collected for flow cytometry and indel analysis. CBC, blood smear, and serum chemistry analyses were performed by IDEXX BioAnalytics (Westbrook, ME). Flow cytometry was performed by Flow Contract Site Laboratory (Bothell, WA). Indel analysis was conducted at Editas Medicine. Histopathology evaluations were conducted by Vet Path Services (Mason, OH).

Patient materials

De-identified SCD and TDT CD34⁺ cells used in Figures 3 and 4 were generously provided by J.F.T. (National Heart, Lung, and Blood Institute [NHLBI/NIH]) and M.C.W. (University of California, San Francisco [UCSF]). All patient samples were collected under a research protocol approved by the institutional review board (IRB) at the respective institution (NHLBI protocol no. 17-H-0124, UCSF Benioff Children’s Hospital protocol IRB no. 2016-106, UCSF protocol IRB no. 21-33506). Patients with SCD were homozygous for sickle mutation. TDT donor genotype information can be found in Table S9. SCD and TDT materials used for specificity evaluation in Figure 7 were the manufactured products for the RUBY and EdiThal studies.

Digenome-seq, NanoString, rhAmpSeq, and UDiTaS methods

See Tables S10, S11, S12, and S13, and the supplemental information for details.

Statistical analyses

Data were analyzed using GraphPad Prism 10. Statistical analyses are noted in the figure legends.

Data availability

For original data, please contact info@editasmed.com. High-throughput sequencing data have been deposited in the National Center for Biotechnology Information’s Sequence Read Archive Database under accession code PRJNA1226931 and can be accessed at http://www.ncbi.nlm.nih.gov/bioproject/1226931.

Acknowledgments

This work was funded by Editas Medicine. The authors wish to acknowledge Charles Albright, Kate Zhang, Sandra Teixeira, Mark Shearman, Sean Scott, Tongyao Wang, Terence Ta, Stephen Sherman, and Li Li for their intellectual and technical contributions.

Author contributions

P.S., T.J., E.D., J.M.H., and K.-H.C. conceived the study; P.S., T.J., E.D., J.M.H., K.-H.C., R.V., G.M.G., and S.H. performed the experiments; P.S., T.J., E.D., J.M.H., E.M., G.M.G., G.G., K.-H.C., S.H., and C.J.W. analyzed the results; M.C.W. and J.F.T. provided critical reagents and participated in data discussion; J.A.Z., C.J.W., D.K.W., and K.-H.C. interpreted the results; all authors wrote the manuscript.

Declaration of interests

E.M., R.V., and G.M.G. are current employees and shareholders of Editas Medicine. P.S., T.J., E.D., J.M.H., J.A.Z., G.G., C.J.W., and K.-H.C. are previous employees of Editas Medicine. E.D., J.M.H., J.A.Z., and K.-H.C. are listed as inventors on patent applications related to this work, filed by Editas Medicine.