Nonclinical evaluation of HBG1/2 and BCL11A as genome-editing targets for the treatment of β-hemoglobinopathies

TO THE EDITOR:

The hemoglobin tetramer is composed of 2 α-like globin chains and 2 β-like globin chains. Sickle cell disease (SCD) and β-thalassemia are β-hemoglobinopathies characterized by abnormal and inadequate production of β-globin, respectively. Patients with SCD and transfusion-dependent β-thalassemia (TDT) are usually born asymptomatic because of high levels of γ-globin, which forms fetal hemoglobin (HbF; α2γ2). Symptoms develop postnatally as γ-globin is replaced by β-globin (Figure 1A).¹ However, coinheritance of genetic variants of hereditary persistence of HbF attenuates the severity of both diseases. Therefore, reactivating γ-globin and increasing HbF levels is a promising treatment strategy for SCD and TDT.^1,2 

Silencing γ-globin is partially mediated by the transcription factor B-cell lymphoma/leukemia 11A (BCL11A), which binds to the promoters of the γ-globin genes, HBG1 and HBG2 (denoted HBG1/2; Figure 1B).³ Editing patient CD34⁺ hematopoietic stem and progenitor cells (HSPCs) at the BCL11A erythroid enhancer +58 GATAA motif or the distal BCL11A binding site at the HBG1/2 promoters’ −110 regions can permanently disrupt this pathway and reactivate γ-globin expression.^4-8 Both approaches have demonstrated clinically meaningful HbF induction and improved clinical outcomes in patients with SCD and TDT.^5-8 In clinical studies in which patients received BCL11A erythroid enhancer–edited HSPCs, 97% of patients with SCD were free of vaso-occlusive crisis for ≥12 consecutive months;⁶ and 91% of patients with TDT achieved transfusion independence.⁸ These studies validate the safety and effectiveness of this approach and supported the approval of exagamglogene autotemcel for treating SCD and TDT, with long-term studies ongoing.⁹ Notwithstanding, there is an opportunity to explore other mechanisms of HbF reactivation for optimal patient outcomes. Clinical evaluation of an investigational HBG1/2 promoter-edited CD34⁺ cell therapy, renizgamglogene autogedtemcel, is ongoing, with rapid and sustained increases in HbF and early correction of anemia having been observed in treated patients with SCD and TDT.^7,10,11 Additionally, renizgamglogene autogedtemcel has demonstrated early engraftment and a favorable safety profile to date. In this nonclinical study, we evaluated whether editing the HBG1/2 promoters or the BCL11A erythroid enhancer led to biological consequences beyond HbF induction based on target site.

Mobilized CD34⁺ cells from healthy donors were edited with CRISPR-associated protein 9 (Cas9) ribonucleoproteins (RNPs) at either the HBG1/2 promoters’ −110 region in which multiple genetic variants of hereditary persistence of HbF reside^3,12 or the BCL11A erythroid enhancer^5,13,14 (Figure 1C; supplemental Table 1; see supplemental Data for detailed methods). BCL11A-targeting RNPs were more efficient than HBG1/2-targeting RNPs in creating insertions and/or deletions (indels) at the target site (Figure 1D). Noteworthy, simultaneous cleavage of the HBG1 and HBG2 promoters may lead to a 4.9-kilobase deletion of the intervening sequence, creating an HBG1-2 hybrid promoter with or without indels. This 4.9-kilobase deletion is estimated to occur in ∼30% to 40% of the loci.^4,12 When differentiated into erythroid cells in vitro, both targeting approaches led to robust HbF expression at ∼35% (Figure 1E), which is consistent with previous publications^5,12,14 and above the 30% HbF threshold for therapeutic relevance.¹⁵ 

Transcriptome analysis of day-11 in vitro–differentiated erythroblasts from unedited (control) or edited CD34⁺ cells (mean ± standard deviation editing rate: HBG1/2, 98.68% ± 0.09%; BCL11A, 99.22% ± 0.08%) revealed 5 differentially expressed genes (DEGs) in HBG1/2-edited samples and 110 DEGs in BCL11A-edited samples (Figure 1F). In the HBG1/2-edited samples, 4 DEGs were globin related: δ-globin, β-globin locus transcript 3, HBG1, and HBG2. In contrast, most DEGs in BCL11A-edited samples were not globin related. Consistent with BCL11A’s transcriptional repressor role,¹⁶ most of the 110 DEGs in BCL11A-edited samples were upregulated, including BCL11A primary targets¹⁷ (Figure 1F; supplemental Figure 1; supplemental Data). Multiple DEGs were related to cell processes such as apoptosis, differentiation, and proliferation; and 2 DEGs have been implicated in dysregulated erythropoiesis (Figure 1G). These data contradict previous studies showing minimal transcriptomic perturbation after BCL11A erythroid enhancer editing.^4,17 Differences in methodologies may explain this discrepancy. We evaluated day-11 immature erythroblasts, whereas Lamsfus-Calle et al evaluated day-21 erythroid cells, likely reticulocytes.⁴ Rajendiran et al used base editing, which led to a low BCL11A erythroid enhancer disruption and modest HbF induction.¹⁷ Interestingly, when greater BCL11A downregulation was achieved via base editing of the BCL11A initiation codon, they observed a similar pattern of perturbation to BCL11A primary targets¹⁷ as we did (supplemental Figure 1).

These data suggest that editing the BCL11A erythroid enhancer, which leads to BCL11A downregulation in erythroid cells,¹⁴ causes broader transcriptome perturbation in in vitro–derived erythroblasts vs editing the HBG1/2 promoters. Although no overt differences in erythroid maturation in culture were observed (supplemental Figure 2), as previously found with BCL11A exon 2 editing,^17,18 we cannot dismiss that the breadth of transcriptome perturbation may have biological consequences in vitro or in vivo.

To evaluate the engraftment potential of CD34⁺ cells edited at the 2 different target sites, immunodeficient NBSGW mice were infused with edited and control CD34⁺ cells.^19,20 At 16 to 18 weeks after infusion, successful and robust engraftment of human cells was observed in all groups (Figure 2A; supplemental Figures 3A and 4A). Comparable levels of human chimerism and HSPC, B-cell, and neutrophil engraftment in the bone marrow (BM) were observed regardless of the editing approach. However, although long-term engrafted HBG1/2-edited CD34⁺ cells produced similar levels of erythroid cells as unedited CD34⁺ cells, reduced human erythroid output was observed with BCL11A-edited CD34⁺ cells (Figure 2A; supplemental Figure 3A). Normalization of human erythroid output to HSPC frequency in the BM of corresponding animals corroborates that BCL11A-edited HSPCs produced fewer erythroid cells than unedited HSPCs in vivo (Figure 2B). In a study by Wu et al using the same animal model, deficient erythroid production was not observed after BCL11A erythroid enhancer editing.¹⁴ However, the administration route, number of CD34⁺ cells infused, and gating strategies were different, and relatively few animals were evaluated. Furthermore, although no erythroid reduction occurred in 1 rhesus monkey transplanted with BCL11A erythroid enhancer–edited rhesus CD34⁺ cells,²¹ Bcl11a conditional knockout mice, in which BCL11A is reduced by 70% in erythroid cells, do develop mild anemia.²² 

Erythroid progeny of CD34⁺ cells edited at either target site expressed more HbF in vivo than unedited controls (Figure 2C). Interestingly, HBG1/2-edited erythroid cells showed stronger HbF induction than BCL11A-edited erythroid cells in vivo. Lower HbF induction in BCL11A-edited erythroid cells is potentially because of residual BCL11A¹⁴ suppressing γ-globin and/or the lack of sufficient erythropoietic stress²¹ driving optimal HbF expression in vivo.

Indel evaluation showed that CD34⁺ cells edited at either target site successfully engrafted in the mouse BM. The editing levels in BM-engrafted human cells were equivalent to, or higher than, those in infused CD34⁺ cells (Figure 2D; supplemental Figure 3B). Mice that received HBG1/2-edited CD34⁺ cells showed similar editing levels across BM-isolated human HSPCs, B cells, neutrophils, and erythroid cells (Figure 2E). In contrast, in mice receiving BCL11A-edited CD34⁺ cells, human erythroid cells had lower editing levels than HSPCs and other blood lineages. Similar trends were observed in 2 independent experiments (supplemental Figures 3C and 4B).

Data from the 3 independent experiments were combined for correlation analyses between editing levels in isolated HSPCs, B cells, neutrophils, or erythroid cells (Figure 2E). Because HSPCs produce downstream lineages, editing levels in the differentiated progeny should closely resemble the levels in the HSPCs if unedited and edited CD34⁺ cells have similar downstream progeny output capability. Indeed, editing levels in B cells and neutrophils were highly similar to those in corresponding HSPCs, regardless of editing site. Editing levels in erythroid cells derived from HBG1/2-edited CD34⁺ cells also closely resembled the corresponding HSPCs. In contrast, erythroid cells derived from BCL11A-edited CD34⁺ cells had lower editing levels than corresponding HSPCs, except when editing levels were very high (∼100%) with almost no unedited CD34⁺ cells present. This suggests that BCL11A-edited CD34⁺ cells generate fewer erythroid cells than unedited CD34⁺ cells, causing an underrepresentation of edited erythroid cells in the mouse BM. The erythroid-specific nature of this observation reflects the strict lineage specificity of this enhancer element.^13,23 

An additional correlation analysis on the compiled data set evaluated the impact of HSPC editing levels on erythroid production. Although the editing levels in HSPCs derived from HBG1/2-edited CD34⁺ cells did not affect overall erythroid output, a negative correlation was observed in HSPCs derived from BCL11A-edited CD34⁺ cells (Figure 2F; supplemental Figure 4C). This corroborates that BCL11A erythroid enhancer editing adversely affects the ability of CD34⁺ cells to generate erythroid cells in a xenotransplantation setting. Whether the DEGs observed in in vitro–derived erythroblasts could account for such observations in vivo requires additional investigation. Because both editing approaches use Cas9, the different in vivo erythroid output potential of HBG1/2-edited vs BCL11A-edited HSPCs is likely a biological consequence of target site selection; this should be broadly applicable to other editing modalities.^24,25 

These data suggest that BCL11A is involved in erythroid development and that editing the BCL11A erythroid enhancer in CD34⁺ cells negatively affects erythroid production in a xenotransplantation setting. In contrast, editing BCL11A binding sites at the HBG1/2 promoters enabled robust HbF induction without adversely affecting erythropoiesis. The ability of edited CD34⁺ cells to generate erythroid cells efficiently may have implications for sustained correction of anemia in a clinical setting, but this remains to be substantiated. Collectively, these preclinical data suggest that editing the HBG1/2 promoters is a promising alternative to editing BCL11A for the durable treatment of β-hemoglobinopathies.

Acknowledgments: The authors thank Georgia Giannoukos, James Bochicchio, Jack Heath, Patricia Sousa, Gregory Gotta, and Edouard de Dreuzy for experimental and sequencing support. The authors acknowledge useful discussions with Stephen Sherman, Haiyan Jiang, Kate Zhang, Charles Albright, Mark Shearman, Tanya M. Teslovich, and Linda Burkly. Medical writing support was provided by Bertha Jane Vandegrift of Porterhouse Medical US and was funded by Editas Medicine Inc, Cambridge, MA, in accordance with good publication practice guidelines.

This study was funded by Editas Medicine, Inc.

Contribution: K.-H.C. conceived the study; T.J., M.J., S.R.B., L.L., P.Z., Y.S., and K.-H.C. performed experiments; T.J., M.J., S.R.B., L.L., P.Z., E.M., M.W., S.G., and K.-H.C. analyzed results; K.-H.C. and E.M. interpreted the results; all authors wrote and revised the manuscript.

Conflict-of-interest disclosure: T.J., M.J., S.R.B., L.L., P.Z., Y.S., K.-H.C, E.M., M.W., and S.G. are current employees and shareholders of Editas Medicine, Inc.

Correspondence: Kai-Hsin Chang, Editas Medicine Inc, 11 Hurley St, Cambridge, MA 02141; email: kaihsin.chang@editasmed.com.