Aims: HbE/β-thalassemia is the commonest form of severe β-thalassemia, and comprises approximately 50% of all cases worldwide1. HbE/β-thalassemia is caused by the HbE codon 26 G>A mutation on one allele and any β0-thalassemia mutation on the other. There is a reduction in β-globin production, resulting in a relative excess in α-globin chains that leads to ineffective erythropoiesis. Importantly, individuals with a mutation on one, but not two, alleles have β-thalassemia trait, a carrier state with a normal phenotype shared by 1.5% of the world’s population2. Recent gene therapy and gene editing approaches have been developed to treat β-thalassemia but do not directly repair the causative mutation in-situ. Gene replacement approaches rely on lentiviral vector-based sequence insertion or homology directed repair (HDR). HbF induction strategies rely on non-homologous end joining targeting of enhancers in-trans. These approaches, whilst variably successful, are associated with potential safety concerns. Methods: Adenine base editors (ABEs) circumvent these problems by directly repairing pathogenic variants in-situ through deamination. ABEs catalyse A-T to G-C transitions. Conversion of the HbE codon to WT through base editing is an attractive strategy to recapitulate the phenotypically normal β-thalassemia trait state without potentially harmful double-strand breaks or random vector insertions. ABEs are able to convert the HbE codon (AAG) to wild-type (GAG), but also to GGG or AGG (Fig A). GGG at codon 26 is found in a naturally occurring haemoglobin, Hb Aubenas3. Heterozygotes have normal red cell indices and are phenotypically normal. We electroporated the latest generation of ABE8 editors4 as mRNA into 3 different severe HbE/β-thalassemia donor HSPCs with sgRNAs targeting the HbE codon. Results: The mean conversion from the HbE codon to a normal or normal variant in unselected cells was 86.2 (SD±8.1%, Fig B). The indel rate from inadvertent on-target Cas9 cleavage was below 0.5%. Edited cells did not show any perturbations in erythroid differentiation as assessed by Immunophenotyping and cellular morphology. In differentiated erythroid cells, RT-qPCR showed a mean fall in the α/β mRNA ratio to 0.65±0.08 (unedited patient cells normalised to 1, n=5, Figure C), indicating a reduction in the excess α-globin gene expression. Protein analysis by CE-HPLC showed a 3.6-fold reduction in HbE levels (SD±1.3) and a 13.5-fold increase in HbA/Hb Aubenas (SD±2.4, Fig D & E). In serial NSG-murine xenotransplantation experiments, base edited cells were found to persist in secondary transplants, showing editing is possible in long-term HSCs (mean editing efficiency 34.5%, Fig F). Potential off-target effects were assessed in-vitro by CIRCLE-seq5; most candidate sites were in intergenic and intronic regions (Fig G & H). The top 250 sites were sequenced using deep targeted NGS. Only 5 sites showed OT deaminations at low levels in patient cells (mean 1.5%). We developed a machine learning model to assess potential OT-effects on chromatin accessibility, at all candidate sites in 49 different blood cell types6. Only 17 potential edits were predicted to result in a significant change in chromatin state (Fig I). 3 of these were in the top 250 sites sequenced previously, and none showed deamination in-vivo. Conclusion: Together these data provide robust evidence for base editing being used as an effective and safe therapeutic strategy for HbE/β-thalassemia.
Read full abstract