Disrupting ZBTB7A or BCL11A binding sites reactivates fetal hemoglobin in erythroblasts from healthy and β⁰-thalassemia/HbE individuals

Abstract

CRISPR/Cas9 genome editing has emerged as a promising treatment for genetic diseases like β-thalassemia. Editing γ-globin promoters to disrupt ZBTB7A/LRF or BCL11A binding sites has shown potential for reactivating fetal hemoglobin and treating sickle cell disease. However, its application to β⁰-thalassemia/HbE disease remains unclear. This study utilized CRISPR/Cas9 to disrupt these sites in mobilized CD34 + hematopoietic stem /progenitor cells from healthy donors and β⁰-thalassemia/HbE patients. The editing efficiency for the BCL11A site (75–92%) was higher than for the ZBTB7A/LRF site (57–60%). Both disruptions similarly increased fetal hemoglobin production in healthy donors (BCL11A 26.2 ± 1.4%, ZBTB7A/LRF 27.9 ± 1.5%) and β⁰-thalassemia/HbE cells (BCL11A 62.7 ± 0.9%, ZBTB7A/LRF 64.0 ± 1.6%). Off-target effects were absent in BCL11A-edited cells but observed at low frequencies in ZBTB7A/LRF-edited cells. Neither disruption significantly affected erythroid differentiation. These findings highlight the comparable contributions of ZBTB7A/LRF and BCL11A binding sites to γ-globin reactivation. CRISPR/Cas9 editing of either site may offer a potential therapeutic strategy for β⁰-thalassemia/HbE disease.

Introduction

β-Thalassemias, a common genetic disorder caused by a β-globin gene mutation that results in the reduction or loss of production of the β-globin chain, are a serious public health concern worldwide. Clinical severity varies widely, ranging from mild anemia to transfusion dependency, with complications affecting growth and overall development^1,2. β⁰-thalassemia/HbE is a severe form of β-thalassemia, predominantly affecting Southeast Asia populations¹. It results from the compound heterozygous inheritance of a β⁰-thalassemia mutation, which leads to the absence of β-globin production and the hemoglobin E (HbE, HBB: c.79G > A) variant. This variant causes structural hemoglobin abnormalities and reduced β-globin synthesis due to aberrant splicing³. The anemia associated with β⁰-thalassemia/HbE disease is driven by impaired hemoglobin production and an imbalance between α- and non-α-globin chain synthesis, leading to ineffective erythropoiesis and increased erythroid cell destruction⁴. At present, one of the curative treatments for transfusion-dependent thalassemia is allogeneic hematopoietic stem cell transplantation. However, the lack of HLA-matched siblings, the risk of post-transplant complications (graft rejection or graft-versus-host disease), and the high cost have restricted its availability for most patients⁵. Inducing fetal hemoglobin (HbF) production can alleviate the severity of β-thalassemia by ensuring an appropriate balance between α- and non-α-globin chain levels⁶. The γ-globin promoter contains transcriptional repressors (ZBTB7A/LRF, BCL11A) binding sites necessary for hemoglobin switching⁷. Hereditary persistence of fetal hemoglobin (HPFH) mutations in γ-globin promoters creates de novo binding sites for erythroid transcriptional activators or disrupts existing transcriptional repressor binding sites, leading to increased HbF production. Point mutations of the γ-globin promoter at − 113A > G, − 175T > C, and − 198T > C are benign changes that lead to HPFH, which generate binding sites for the HbF activators GATA1, TAL1, and KLF1, respectively^8–10. In addition, the introduction of HPFH-associated mutations around the − 200-bp cluster (− 195C > G, − 196C > T, − 197C > T, − 201C > T, and − 202C > T/G) and − 115-bp cluster (− 114C > A, − 117G > A, and a 13-bp deletion) upstream from the transcription start site (TSS) was demonstrated to disrupt the binding sites of the two major γ-globin repressors, ZBTB7A/LRF and BCL11A, respectively¹⁰. The development of CRISPR/Cas9 technology provides the opportunity to reactivate fetal hemoglobin synthesis and develop a promising approach for the genetic repair of β-hemoglobinopathies. Previous reports have described CRISPR/Cas9-mediated β-globin gene correction via the homology-directed repair (HDR) pathway in K562 and CD34 + hematopoietic stem/progenitor cells (HSPCs) from sickle cell disease (SCD) patients^11–13. CRISPR/Cas9 genome editing to create HPFH mutation in CD34 + HSPCs resulted in increased expression of γ-globin genes^14,15. A previous report also showed that genetic variation of individual erythroid-lineage-specific enhancers of human BCL11A, known as DNase I hypersensitive sites (DHSs), + 55, + 58, and + 62, was associated with variation in HbF induction^16–18. Preclinical studies have also demonstrated the disruption of transcriptional repressor binding sites in the γ-promoter or human BCL11A enhancer by CRISPR/Cas9, which promoted HbF reactivation in cell lines and human erythroid cells derived from patients with SCD and β-thalassemia^19–21. Recently, a successful clinical trial showed that CRISPR/Cas9-mediated disruption of γ-globin promoters led to sustained increases in HbF levels and clinical improvement in SCD patients²². Several studies applying CRISPR/Cas9-based genome editing of the ZBTB7A/LRF or BCL11A binding sites in the γ-globin promoter have demonstrated a variety of indel frequencies, and increments of HbF levels in K562, HUDEP-2 and CD34 + HSPCs^19–21,23. However, the preclinical profiles of indel-frequency and off-target analysis associated with CRISPR/Cas9-nuclease mediated DNA double-stranded break in β⁰-thalassemia/HbE, the most common β-thalassemia disease in Southeast Asian countries including Thailand, remain limited. In this present study, we explored the therapeutic potential of using CRISPR/Cas9 to disrupt the binding sites of two well-known transcriptional repressors, ZBTB7A/LRF (-197) and BCL11A (-115) in the γ-globin promoters in erythroblasts derived from CD34 + HSPCs of β⁰-thalassemia/HbE patients (Fig. 1A).

Fig. 1.

CRISPR/Cas9-mediated double-strand break on a transcriptional repressor binding site in the γ-globin gene promoter in CD34 + HSPCs. (A) The schematic representation of the γ-globin promoter on chromosome 11 illustrates sgRNA targeted on LRF (− 197) and BCL11A (− 115) binding regions. Genome editing at these sites induces indel mutations through non-homologous end joining (NHEJ) and microhomology-mediated end joining (MMEJ) DNA repair pathways. (B) Editing efficiency, represented as the percentage of indels generated by CRISPR/Cas9-mediated genome editing using sg-LRF and sg-BCL11A in erythroblasts, as determined by deep sequencing (n = 3 per group). (C,D) Deep sequencing analysis showing genome editing profiles at the LRF and BCL11A binding sites in the γ-globin promoter in edited erythroblast-derived CD34 + HSPCs. The reference sequence is displayed at the top, with the PAM sequence highlighted in a red box and the predicted cleavage position indicated by a blue dashed line. Each row represents an observed sequence aligned to the reference sequence. Insertions are outlined in red boxes, deletions are shown as dashes, and substitutions are shown in bold text. The frequency of each sequence is displayed as a percentage of the total reads, with the corresponding read count in parentheses. Unedited alleles that perfectly match the reference sequence are marked with a red asterisk. (E,F) Measurements of on-target, indel-mutation and MMEJ frequencies after genome editing in samples from healthy donors and β⁰-thalassemia/HbE patients (n = 3 per group).

Results

Targeting CRISPR/Cas9 to disrupt ZBTB7A/LRF or BCL11A binding site in γ-globin promoter

To determine the specificity and efficiency of CRISPR/Cas9-mediated disruption of the ZBTB7A/LRF or BCL11A binding site in the γ-globin promoter, mobilized CD34 + HSPCs from healthy donors (n = 3) and β⁰-thalassemia/HbE (n = 3) were initially electroporated with specific ribonucleoprotein (RNP) containing Cas9 complexed with sgRNA targeting the ZBTB7A/LRF binding site at HBG-197 (sg-LRF) and the BCL11A binding site at HBG-115 (sg-BCL11A). These sgRNA sequences were obtained from previously published and extensively characterized studies in healthy donors and SCD HSPCs^19–21. Deep sequencing analysis revealed the distribution of indel frequencies at the targeted genomic regions. The editing efficiencies of sg-LRF and sg-BCL11A were approximately 69.4 ± 7.4% and 84.9 ± 17.1% in healthy donor cells and 68.2 ± 12.2% and 88.5 ± 3.1% in β⁰ thalassemia/HbE cells, respectively (Fig. 1B). The results showed the indel mutation profiles, ranked by the number of reads in each allele demonstrated in genome editing profiles at the ZBTB7A/LRF binding site (Fig. 1C) and BCL11A binding site (Fig. 1D) in the γ-globin promoter. We summarized the ten most frequently identified reads, demonstrating various indel frequencies at the ZBTB7A/LRF binding site (Fig. 1E) and the BCL11A binding site (Fig. 1F). The 6-bp deletion was most frequently observed at the ZBTB7A/LRF edited binding site in both healthy donors and β⁰-thalassemia/HbE cells (10.8 ± 1.5% vs. 11.1 ± 0.3%, respectively) (Fig. 1E). Additionally, the 13-bp deletion was the most common indel mutation detected at the BCL11A edited binding site in healthy donors and β⁰-thalassemia/HbE cells (21.1 ± 0.6% vs. 21.6 ± 2.1%, respectively) (Fig. 1F).

Fetal hemoglobin reactivation by targeted disruption of ZBTB7A/LRF or BCL11A binding sites in γ-globin promoter and erythroid differentiation

Disruption of the ZBTB7A/LRF or BCL11A binding sites of the γ-globin promoter resulted in a significant increase in γ-globin transcripts. In healthy donor cells, the fold changes of γ-globin were 7.5–11.4 for sg-LRF and 6.1–11.2 for sg-BCL11A. The changes in β⁰-thalassemia/HbE cells were 4.0–5.3 folds for sg-LRF and 2.7–3.2 folds for sg-BCL11A. The α- and β-globin transcripts were present at similar levels upon comparing edited cells and controls (Fig. 2A, B). The gene expression of LRF and BCL11A in both healthy donors and β⁰-thalassemia/HbE cells (Fig. 2C, D) did not significantly change after CRISPR/Cas9-mediated disruption of the ZBTB7A/LRF and BCL11A binding sites in the γ-globin promoter.

Fig. 2.

Gene expression of globins and transcription factors after disruption of ZBTB7A/LRF or BCL11A binding sites in erythroid cells derived from CD34 + HSPCs. (A,B) Relative quantification of globin gene expression (α-, β- and γ- globin) analyzed by RT-qPCR and normalized to RPS18 (n = 3 per group). (C,D) Expression of transcriptional repressors  BCL11A and LRF, normalized to GAPDH (n = 3 per group). Values are represented as means ± SD. ***p < 0.0001, **p < 0.001, *p < 0.05 compared to control.

Hemoglobin analyses by cation exchange-high performance liquid chromatography (HPLC) demonstrated that editing the γ-globin promoters in healthy donor cells at ZBTB7A/LRF and BCL11A binding sites significantly increased fetal hemoglobin levels. Specifically, the HbF induction levels were 27.9 ± 1.5% for sg-LRF and 26.17 ± 1.4% for sg-BCL11A, compared to the basal level of 3.4 ± 0.9%; all p < 0.0001). In β⁰-thalassemia/HbE, editing the γ-globin promoters using sg-LRF and sg-BCL11A yielded significant increases in HbF levels, reaching up to 64.0 ± 1.6% (p < 0.001) and 62.7 ± 0.9% (p < 0.0001), respectively, compared to the basal level of 27.87 ± 0.9% in controls. These increases in HbF levels led to proportional decreases in HbA levels in healthy donors or HbE levels in β⁰-thalassemia/HbE erythroid cells (Fig. 3A). The reverse-phase HPLC analysis revealed reactivation of the γ-globin chain. The ratio of γ-globin to α-globin was significantly increased in edited cells compared to control cells in both cell types, as shown in Fig. 3B. The percentage of γ^G-globin chain and γ^A-globin chain was increased considerably after genome editing by sg-LRF and sg-BCL11A compared to controls in healthy donors and β⁰-thalassemia/HbE (Fig. 3C). Remarkably, genome editing in the γ-globin gene promoters predominantly resulted in the induction of the γ^A-globin chain rather than the γ^G-globin chain in both cell types, as shown in Fig. 3C.

Fig. 3.

Reactivation of fetal hemoglobin and globin chain analysis after disruption of ZBTB7A/LRF or BCL11A binding site in erythroid cells derived from CD34 + HSPCs. (A) Cation-exchange HPLC shows the reactivation of HbF synthesis on day 14. (B) Reverse phase HPLC shows globin chain analysis and the relative ratio of γ/α-globin chain induction in edited cells. (C) The expression of globin chain (α, β, β^E, γ^G and γ^A) after genome editing (n = 3 per group). Values are represented as means ± SD. ***p < 0.0001, **p < 0.001, *p < 0.05 compared to control.

We found that disruption of the ZBTB7A/LRF and BCL11A binding sites in the γ-globin promoter did not significantly affect erythroid differentiation in the healthy donors and β⁰-thalassemia/HbE cells (Fig. 4A, B). This was demonstrated by the comparable erythroid maturation profile observed in the edited cells and control cells. Flow cytometry dot plot and gating strategy are provided in the supplemental Figure S1.

Fig. 4.

Erythroid cell differentiation after disruption of ZBTB7A/LRF or BCL11A binding site in erythroid cells derived from CD34 + HSPCs. (A) Cell morphology of control cells and edited erythroid cells on day 10 and day 14, stained with Giemsa reagent, visualized by light microscope (scale bar = 10 μm). A total of 250–300 cells were counted per donor (n = 3 per group). The stage-specific normoblast distributions are represented in bar graphs. The error bars represent donor-to-donor variation, indicating the variability observed between individual samples. Values are represented as means ± SEM. (B) Erythroid maturation analysis by flow cytometry showing the expression of transferrin receptor (CD71) and glycophorin A (GPA) in gate R1( CD71^High GPA⁺ ) early stage population, R2( CD71^Med GPA⁺ ) transition stage population, R3( CD71^Low GPA⁺ ) late stage population on day 10 and day 14 during erythroid differentiation in healthy donors and β⁰-thalassemia/HbE patients (n = 3 per group). Values are represented as means ± SD.

Detection of a large deletion in the γ-globin promoter and off-target analysis

As γ^G- and γ^A-globin promoters share identical base sequences, a large 4.9-kb deletion spanning the γ^G-globin gene and the flanking sequences was generated as a result of the γ-globin promoter editing. Gap-PCR demonstrated this deletion in sg-LRF- and sg-BCL11A-edited healthy donors and β⁰-thalassemia/HbE cells. Semi-quantitative Gap-PCR was used to confirm the 4.9-kb deletion of the γ^G-globin gene in sg-LRF and sg-BCL11A edited healthy donors and β⁰-thalassemia/HbE cells (Fig. 5A). The results showed that editing the LRF and BCL11A binding sites yielded two PCR amplicons (full amplicon size = 6,868-bp, amplicon size with the 4.9-kb deletion = 1,865-bp) in both cell types.

Fig. 5.

Detection of 4.9-kb deletion and off-target analysis after disruption of ZBTB7A/LRF or BCL11A binding site in erythroid cells derived from CD34 + HSPCs. (A) Agarose gel electrophoresis showing a large deletion after genome editing. (4.9-kb deletion = 1,865-bp., full amplicon = 6,868-bp). (B) Deep sequencing by NGS to determine the percentage of off-target cleavage sites (indel mutations) in editing cells compared to control cells (n = 3 per group). Values are represented as means ± SEM.

We analyzed the top five previously reported potential off-target (OT1–OT5) cleavage sites for each sgRNA based on prior GUIDE-seq and CIRCLE-seq studies^19,20. The numbering of OT1–OT5 in Fig. 5B follows the ranking of off-target activity reported in these studies, with OT1 representing the site with the highest predicted off-target potential and OT5 the lowest. Amplicon deep sequencing revealed that off-targeting was not detected with OT3–OT5 of sg-LRF or with OT1–OT5 of sg-BCL11A at the detection threshold for amplicon deep sequencing of approximately 0.01% to 0.1%. However, indel mutations were detected at low frequencies with OT1-sg-LRF (2.5 ± 0.7% vs. 2.6 ± 1.3%) and OT2-sg-LRF (1.6 ± 0.3% vs. 2.0 ± 1.5%) in the healthy donor and β⁰-thalassemia/HbE cells, respectively (Fig. 5B).

Discussion

The CRISPR/Cas9 nuclease system constitutes a major technological genome editing advancement. Non-homologous end joining repair, facilitated by single-guide RNA incorporated into the Cas9 nuclease for site-specific cleavage to DNA double-strand break-induced indel mutations, has substantial promise for clinical treatments^24–29. One potential application is in cases of compound heterozygosity for the β⁰-thalassemia/HbE genotype, a major public health problem in Southeast Asia, including Thailand¹. Therapeutic approaches aimed at reactivating HbF production in β-thalassemia by techniques leading to the downregulation of transcription factors involved in γ-globin silencing, including BCL11A, ZBTB7A/LRF, LSD1, and KLF1, have been reported^30–33. However, reducing the levels of some transcriptional repressors, such as KLF1, might also affect the maturation of erythroid cells^34,35.

HPFH mutations involving the transcriptional repressor binding sites in the γ-globin promoter suggested the potential benefit of mimicking HPFH mutation to reactivate HbF expression. Leslie Weber and colleagues demonstrated that disrupting two clusters in the γ-globin promoters, the binding sites of ZBTB7A/LRF (− 195, − 196, − 197) and BCL11A (− 115), significantly increased HbF synthesis in erythroblasts derived from patients with SCD. The ZBTB7A/LRF binding site is in the − 200 region, where the − 197 sgRNA-mediated disruption exhibits the highest genome editing efficiency and HbF production²⁰.

Here, we performed single electroporation with specific sg-LRF or sg-BCL11A in each donor and compared the editing results of individual binding sites performed using the same sample. Specifically, we selected − 197 and − 115 sgRNA for binding site genome editing, with our results showing that the editing efficiency of BCL11A (− 115, approximately 90%) binding site was higher than that for ZBTB7A/LRF (− 197, approximately 70%). These editing efficiencies were the primary contributors to HbF and were consistent with previous reports^19,20. The discrepancy in editing efficiency may be attributed to differences in chromatin accessibility and the structural characteristics of the two target loci. The BCL11A binding site is generally more accessible whereas the LRF binding site may be surrounded by more complex regulatory regions or chromatin architecture, which could reduce editing efficiency at this site³⁶. Additionally, the nucleotides in both the PAM-distal and PAM-proximal regions of the sgRNA significantly correlate with on-target editing efficiency. The genomic context, GC content, and secondary structure of the sgRNA are also key factors contributing to cleavage efficiency. Many studies suggest higher editing efficiencies are achieved when the GC content of the sgRNA-targeted region is between 40 and 60%³⁷. The GC-rich of LRF binding site (5-GGTGGTGG-3) may contribute to lower cleavage efficiency compared to the BCL11A binding site (5-TGACCA-3). The predominant indel mutations detected were 13-bp deletion at the  − 115 sgRNA target site and 6-bp deletions at the  − 197 sgRNA target site, which were mediated by Microhomology-Mediated End Joining (MMEJ) in both healthy donor and β⁰-thalassemia/HbE cells. The editing efficiency at the LRF binding site was significantly lower than at the BCL11A binding site; however, HbF induction remains comparable. This may be due to some indel mutations at the BCL11A binding site that do not completely abolish BCL11A-mediated repression, allowing residual suppression of HbF expression. Our results are consistent with previous findings³³, confirming that ZBTB7A/LRF and BCL11A independently repress γ-globin gene expression.

Due to the extensive homology between the human HBG1 (γ^A-globin) and HBG2 (γ^G-globin) promoter, including the recognition site for sg-LRF and sg-BCL11A. CRISPR/Cas9-mediated genome editing at ZBTB7A/LRF or BCL11A binding site on the HBG promoter always carries potential risks to generate large deletions and gene inversions. Previous studies have reported that simultaneous genome editing at the ZBTB7A/LRF and BCL11A binding induces on-target DNA double-strand breaks, leading to approximately 10–30% occurrence of the 4.9-kb deletion containing HBG2 genes (γ^G-globin)^19,20. Consistent with these findings, our study detected a 4.9-kb deletion by Gap-PCR in approximately 40–50% following genome editing of either −115 sgRNA or −197 sgRNA in healthy donors and β⁰-thalassemia/HbE cells (Fig. 5A). Notably, the 4.9-kb loss resulted in the complete loss of the HBG2 gene; however, HbF expression was still elevated, primarily due to increased expression of the HBG1 gene. This finding is supported by reverse-phase HPLC analysis, which revealed a tremendous increase in γ^A-globin levels compared to γ^G-globin in editing cells (Fig. 3C), consistent with observations from previous studies^13,19,20. In both healthy donors and β⁰-thalassemia/HbE patients, sg-BCL11A-edited cells exhibited lower levels of γ^G-globin and higher levels of γ^A-globin compared to sg-LRF-edited cells. Despite these differences, the total γ-globin content (combined γ^G- and γ^A-globin chains) remained comparable across all edited cells. This suggests that the loss of the HBG1 gene resulted in a compensatory balance between γ^G- and γ^A-globin chains, maintaining an equivalent level of HbF induction. Notably, our study showed that an increase in the proportion of HbF relative to total hemoglobin as a result of BCL11A binding site disruption (− 115 sgRNA) was comparable to that observed previously upon disruption of the core erythroid-specific enhancer of the BCL11A (+ 58) gene itself in β⁰-thalassemia/HbE¹⁷. This evidence supports the concept that BCL11A is a major transcriptional repressor of the γ-globin gene in adult erythroid ontogeny^32,33. Double CRISPR/Cas9-mediated genome editing, which combines BCL11A enhancer editing with editing of the γ-globin promoter at either − 115 sgRNA or − 197 sgRNA concurrently, has been shown to enhance reactivation of HbF more effectively than single genome editing in both β-thalassemia cells and healthy donors²¹. However, the extent of off-target effects and genetic rearrangement of double genome editing sites remains unclear compared with that of single genome editing.

Previous studies have demonstrated that genome editing at the + 58 BCL11A erythroid enhancer^18,21 and HBG-115²¹ in β-thalassemia patient-derived cells effectively ameliorates disease phenotypes by restoring α/β-globin chain balance. Their results had been shown to increase the frequency of HbF⁺ cells, induce HbF expression, enhance erythroid maturation (as indicated by higher enucleation rates), and significantly reduce reactive oxygen species levels compared to untreated cells. However, the erythroid maturation in sg-LRF and sg-BCL11A edited cells in this study was slightly improved (Fig. 3B), as represented by the R3 gating; however, it was not significantly different from that in unedited β⁰-thalassemia/HbE cells that might be caused by various of intrinsic stress in β⁰-thalassemia/HbE cells.

Recently, CRISPR/Cas9 genome editing has effectively disrupted γ^G- and γ^A-globin promoters in SCD, as demonstrated in preclinical experiments and a clinical study (NCT04443907)²². The clinical trials RUBY (NCT04853576)³⁸ and EdiThal (NCT05444894)³⁹ involved patients with SCD and transfusion-dependent β-thalassemia major, respectively. These clinical trials suggested that the 4.9-kb intergenic deletion commonly emerges upon γ-globin promoter genome editing. In the present study, we employed Gap-PCR as a qualitative method to confirm the presence of this 4.9-kb deletion. We acknowledge that accurate quantification requires methods such as ddPCR or qPCR, which will be considered in future investigations. While these studies observed the presence of a 4.9-kb deletion in edited cells, they reported favorable therapeutic outcomes with no adverse effects attributable to the gene-editing products. These findings suggest that CRISPR/Cas-mediated genome editing targeting the γ-globin promoters with the generation of the 4.9-kb deletion is safe and effective, making it a promising approach for therapeutic applications. Moreover, the adenine or cytosine base editing strategy to create HPFH mutations in the − 200 clusters of the γ-globin promoter also produced the 4.9-kb deletion and provided a range of off-target activity of base editing in SCD-derived erythroblasts⁴⁰.

Off-target Cas9 nuclease activity continues to be a significant concern in therapeutic applications^41,42. Because of this, we used amplicon deep sequencing to determine the indel mutation frequency at potential off-target sites. The candidates for these potential off-target regions were selected based on previous studies that used sgRNA with sequences similar to ours through GUIDE-Seq and CIRCLE-Seq methods^19,20. Our study detected a low frequency of indel mutations at off-target sites OT-1 and OT-2 of sg-LRF in both edited healthy donor and β⁰-thalassemia/HbE cells. These off-target events were located in the intergenic region of IFNG-IL26 on chromosome 12 (OT-1) and the intergenic region of RUN6-1047 on chromosome 3 (OT-2). These findings suggest that the observed off-target effects are likely non-clinically significant; however, further investigation is warranted to confirm their safety profile. Consistent with our results, a previous study reported that the sg-LRF targeting at positions −196 and −197 induced a low frequency of indel mutations at off-target regions containing 1–3 mismatches. Recently, chromosomal aberration analysis using CAST-seq in primary HSPCs and GUIDE-seq in 293 T cells revealed different off-target effects of genome editing targeting the LRF binding site using − 196 and − 197 sgRNAs⁴³. The combination of these techniques offers a valuable approach for monitoring the safety of genome editing in future studies. Notably, no off-target effects were detected with sg-BCL11A at position −115 in SCD²⁰. Furthermore, recent studies have demonstrated that genetic variations can create protospacer adjacent motif (PAM) sequences, potentially altering the off-target effects of CRISPR-Cas9 gene editing⁴⁴. To ensure the safety of gene editing therapies, future studies should include comprehensive off-target analyses, such as GUIDE-seq and CAST-seq in β⁰-thalassemia/HbE cells to better assess the impact of genetic diversity on editing outcomes.

In conclusion, we demonstrated that CRISPR/Cas9-mediated genome editing in the ZBTB7A/LRF and BCL11A binding sites in the γ-globin promoter significantly increased HbF production in β⁰-thalassemia/HbE disease cells. This therapeutic strategy was achieved without affecting cell maturation. Our findings demonstrated that editing erythroblasts derived from CD34 + HSPCs resulted in a particular frequency of off-target by sgRNA targeting the LRF binding site. The obtained preclinical data suggest that this approach offers a promising strategy for the clinical treatment of β⁰-thalassemia/HbE patients. For future studies, utilizing cell sorting techniques could provide deeper insights into the relationship between specific indel patterns and HbF induction.

Methods

Subjects and sample collection

This study was approved by the Ethics Committee on Human Rights Related to Research Involving Human Subjects, Faculty of Medicine Ramathibodi Hospital, Mahidol University, Thailand (COA. MURA2019/179). All samples were collected after written informed consent was obtained from the patients and/or their legal guardians. All methods were performed in accordance with the relevant guidelines and regulations. Granulocyte colony-stimulating factor-mobilized peripheral blood progenitor cells were collected by leukapheresis from healthy donors (n = 3) and β⁰-thalassemia/HbE patients (n = 3) at the stem cell collection unit, Department of Pediatrics, Faculty of Medicine, Ramathibodi Hospital, Mahidol University, Thailand.

CD34 + HSPCs isolation and gamma genome editing

CD34 + HSPCs were separated from mononuclear cells using a CD34 + positive purification kit with magnetic microbead separation and LS-MACS columns (Miltenyi Biotech, Bergisch Gladbach, Germany), as described in the manufacturer’s protocol. CD34 + HSPCs were pre-cultured in StemSpan SFEM II (STEMCELL Technologies, Vancouver, BC, Canada) in the presence of 100 ng/ml human stem cell factor (hSCF, 300–07; Peprotech), 100 ng/ml human interleukin-6 (hIL-6, 200–06; Peprotech), 100 ng/ml human thrombopoietin (hTPO, 300–18; Peprotech), 100 ng/ml human Flt3-Ligand (hFlt-3L, 300–19; Peprotech), and 10 ng/ml human interleukin-3 (hIL-3, 200–03; Peprotech), for 48 h before being subjected to RNP transfection. The in-house wild-type 3xNLS spCas-9 recombinant protein, as previously validated and reported, was used in this process^17,45. sgRNAs with 2′-O-methyl-3′-phosphorothioate modification at the 5′ and 3′ ends were obtained from Integrated DNA Technologies (Singapore). The sgRNA nucleotide sequences of sg-LRF and sg-BCL11A are shown in Table S1. RNP complexes were assembled at room temperature for 15 min using a 1:3 ratio of Cas9:sgRNA (100 pmol Cas9 and 300 pmol sgRNA). The cultured CD34 + HSPCs (2 × 10⁵ cells) were mixed with RNPs using the P3 Primary Cell 4D-Nucleofector X Kit (Lonza, USA) and electroporated using program DZ100 in 16-well Nucleocuvette Strips. Each sample was individually subjected to single electroporation with specific sg-LRF or sg-BCL11A. Ninety-six hours later, CD34 + electroporated cells were subjected to two-phase in vitro erythroid differentiation.

In vitro erythroid differentiation of CD34 + HSPCs

The basal medium for in vitro erythroid differentiation was composed of Iscove’s Modified Dulbecco’s Medium (Gibco, Grand Island, NY, USA) supplemented with 20% fetal bovine serum (Sigma-Aldrich, St. Louis, MO, USA) and 300 μg/mL holo-transferrin (holo-TF; PromoCell, Heidelberg, Germany). Cells were cultured for 4 days in phase I, in which the basal medium was composed of 10 ng/mL hIL-3, 50 ng/mL hSCF, and 2 units/mL of human recombinant erythropoietin (EPO; CILAG GmbH, Zug, Switzerland). After 4 days, suspended cells were collected and re-seeded in phase II medium consisting of basal medium with 5 units/mL EPO. The culture was maintained under an atmosphere of 5% CO₂ at 37 °C for 10 days in a phase II medium.

Indel mutation analysis

Genomic DNA from erythroblasts derived from CD34 + HSPCs was isolated using a Genomic DNA Mini Kit (Geneaid, Taipei, Taiwan). The targeted editing site of the γ-globin promoter was amplified using DreamTaq Green PCR master (Thermo Fisher Scientific, MA, USA) mix using the following cycling conditions: 95 °C for 3 min; 35 cycles of 95 °C for 30 s, 63 °C for 30 s, and 72 °C for 60 s; and 72 °C for 10 min with a specific primer located on the γ-globin promoter. The relevant primer sequences are shown in TableS2. The purified PCR products were subjected to Sanger sequencing. The sequencing data were imported into the web tool ICE to analyze the frequency of indel mutations.

Analysis of 4.9-kb deletion of the γ-globin promoter

The detection of the region encompassing the 4.9-kb deletion at the γ-globin gene promoter was analyzed using gap-PCR. Genomic DNA was amplified using Hercules II Fusion DNA Polymerase (Agilent Technologies, CA, USA) using the following cycling conditions: 95 °C for 2 min; 25 cycles of 95 °C for 10 s, 60 °C for 20 s, and 72 °C for 3.5 min; and finally, 72 °C for 5 min, with the specific primers shown in Table S3. The full-amplicon and large-deletion PCR products were 6.7-kb and 1.8-kb in size, as determined by agarose gel electrophoresis.

RNA isolation and reverse-transcription quantitative PCR (RT-qPCR)

In accordance with the manufacturer’s instructions, total RNA was extracted from erythroid cultures (approximately 2 × 10⁶ cells) using TRIzol Reagent (Thermo Fisher Scientific, MA, USA). cDNA was synthesized by a reverse-transcription reaction using the RevertAid First Strand cDNA synthesis kit (Thermo Fisher Scientific), following the manufacturer’s protocol. RT-qPCR was performed in duplicate with the specific primers (Table S4) using FastStart Essential DNA Green Master Mix (Roche Diagnostics, CA, USA) and analyzed using a LightCycler 96 System (Roche Molecular Systems). All globin gene expression levels were normalized to ribosomal protein S18 (RPS18), whereas the expression levels of transcription factors were normalized to GAPDH. The 2^−∆∆Ct method⁴⁶ was used to analyze the relative fold change.

Erythroid differentiation analysis

Erythroid cell differentiation was monitored via the analysis of cell surface markers using flow cytometry on a FACSVerse flow cytometer (BD Biosciences, San Jose, CA, USA), in which cells were immunostained with allophycocyanin-conjugated anti-transferrin receptor (CD71-APC) (BD Biosciences) and fluorescein isothiocyanate-conjugated anti-glycophorin A (GPA-FITC) (BioLegend, CA, USA) antibodies. In addition, cell maturation was monitored using Giemsa-stained cytospin preparations. Cell morphology was observed under a light microscope.

Quantification of fetal hemoglobin

Hemolysates were prepared from at least 1 × 10⁶ cultured cells on day 14 in VAR-β-THAL Elution buffer 1 (Bio-Rad, CA, USA) and used for high-performance liquid chromatography (HPLC) for hemoglobin type analysis using a Bio-Rad VARIANT II Hemoglobin Testing System with the β-Thalassemia Short Program (Bio-Rad), following the manufacturer’s recommendations.

Globin chain analysis by reverse-phase high-performance liquid chromatography

Erythroid cells were lysed in deionized water and subjected to two freeze–thaw cycles. A clear cell lysate was separated by centrifugation at 14,000×g for 10 min at 4 °C, and the supernatant was transferred into an HPLC micro vial. Analyses were performed on a Waters HPLC Alliance e2695 (Waters Corporation, MA, USA) separation module and detector. The stationary phase was collected on an Aeris 3.6-µm WIDEPORE-C4 200 Å column behind a SecurityGuard UHPLC Wide-pore C18; 4.6 mm guard column (Phenomenex, CA, USA). The globin chain separation was performed as previously described⁴⁷. Empower 3 chromatography software was used for data acquisition and analysis.

Analysis of on- and off-target of sgRNA by amplicon deep sequencing

The off-target sg-LRF and sg-BCL11A cleavage sites were predicted by GUIDE-Seq and Circle-Seq, as previously described^19,20. The top 5 off-target sequences were selected for examination in genome-edited erythroblasts from healthy donors and β⁰-thalassemia/HbE patients. Briefly, genomic DNA of genome-edited erythroblasts derived from CD34 + HSPCs was then subjected to amplification of the selected potential on- and off-target sgRNA cleavage sites using locus-specific primers (Table S5). Two-step PCR amplification with the proofreading enzyme Herculase II Fusion DNA Polymerase (Agilent Technologies) was performed for library preparation. First-PCR reactions used locus-specific primers overhanging the Illumina Nextera handle sequence. PCR thermal cycling was performed as follows: 95 °C for 2 min; 15 cycles of 95 °C for 10 s, 60–63 °C for 20 s, and 72 °C for 20 s; and final extension 72 °C for 5 min. Purified first-PCR amplicons were used as a template, and a unique Illumina Nextera index (Index i5 vs. i7 barcodes) was used for the second PCR reaction. PCR thermal cycling was performed as follows: 95 °C for 2 min; 15 cycles of 95 °C for 10 s, 60 °C for 20 s, and 72 °C for 20 s; and final extension 72 °C for 5 min. The PCR products were purified with AMPure XP beads (Beckman Coulter) and quantified using Qubit™ dsDNA Quantification Assay Kit (Thermo Fisher Scientific). Library quantification, normalization, and pooling were performed as described in the 16S metagenomic sequencing library guide (https://support.illumina.com/downloads/16s_metagenomic_sequencing_library_preparation.html). The Illumina MiSeq platform performed 2 × 250 paired-end sequencing, generating approximately 100,000 paired-end reads per amplicon. Sequencing data were analyzed using CRISPResso2 software version 2.2.14 with the following specific parameters⁴⁸. A minimum alignment identity of 75% was applied to include high-confidence alignments. The quantification window size was set to ± 2-bp around the predicted cleavage site to identify insertion and deletion events. Reads with a mean PHRED quality score below 30 were excluded from the analysis to maintain data quality.

Statistical analysis

All statistical analyses were performed using unpaired Student’s t-test by Prism 8 version 8.4.3 (GraphPad Software, San Diego, CA, USA). Results are presented as means ± SD, and differences were considered significant when the p-value was less than 0.05.

Author contributions

C.W.: Investigation, Validation, Formal analysis, Writing – original draft, Writing – review & editing. P.I.: Investigation. C.S.: Resources. N.J., A. Tub., W.C.: Supervision, Writing – review & editing. P.K., J.P.: Data curation, Formal analysis. D.S., U.A., S.H.: Resources, Supervision. D.S.: Conceptualization, Methodology, Writing – review & editing. A. Tang: Conceptualization, Methodology, Investigation, Validation, Writing – original draft, Writing – review & editing, Supervision. All authors have read and approved the final version of the manuscript.

Funding

This research is supported by Faculty of Medicine Ramathibodi Hospital, Mahidol University, Thailand (Grant ID: RF_64041) and Mahidol University (Grant ID: MRC-IM 02/2565; Basic Research Fund: the fiscal year 2022 Grant ID: BRF1-016/2565).

Data availability

The amplicon deep sequencing datasets generated in this study for both on-target and off-target sgRNA sequences have been deposited in the NCBI Sequence Read Archive (SRA) under accession number PRJNA1211571 (https://www.ncbi.nlm.nih.gov/sra/PRJNA1211571). Additional data or materials supporting this study’s findings are available from the corresponding author upon request.

Declarations

Competing interests

The authors declare no competing interests.

Ethics declarations

This study was performed after obtaining institutional ethical approval (COA.MURA2019/179) from the Ethics Committee on Human Rights Related to Research Involving Human Subjects, Faculty of Medicine Ramathibodi Hospital, Mahidol University, Thailand.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-10791-8.