A unified model of human hemoglobin switching through single-cell genome editing

Key mechanisms of fetal hemoglobin (HbF) regulation and switching have been elucidated through studies of human genetic variation, including mutations in the HBG1/2 promoters, deletions in

the β-globin locus, and variation impacting BCL11A. While this has led to substantial insights, there has not been a unified understanding of how these distinct genetically-nominated

elements, as well as other key transcription factors such as ZBTB7A, collectively interact to regulate HbF. A key limitation has been the inability to model specific genetic changes in

primary isogenic human hematopoietic cells to uncover how each of these act individually and in aggregate. Here, we describe a single-cell genome editing functional assay that enables

specific mutations to be recapitulated individually and in combination, providing insights into how multiple mutation-harboring functional elements collectively contribute to HbF expression.

In conjunction with quantitative modeling and chromatin capture analyses, we illustrate how these genetic findings enable a comprehensive understanding of how distinct regulatory mechanisms

can synergistically modulate HbF expression.

The regulation of fetal hemoglobin (HbF) has been of substantial interest, both for its value to enable improved therapies to elevate HbF as a treatment in sickle cell disease and

β-thalassemia, as well as for its broader implications as a paradigm for understanding the developmental control of gene expression1,2,3. A number of studies have provided insights into HbF

regulation through the identification and analysis of naturally occurring mutations impacting this process. Such variants have been extensively characterized at two distinct loci: (1) in the

gene encoding the BCL11A transcription factor and (2) within the β-globin gene locus that harbors the HbF genes, HBG1 and HBG2. Both common and rare variants in the BCL11A gene alter HbF

expression in erythroid cells, with rare loss-of-function variants resulting in substantially increased HbF4,5,6,7,8,9,10. Other studies focused on the β-globin locus have identified a

number of single-nucleotide variants (SNVs) and small deletions in the HBG1 and HBG2 proximal promoters that allow upregulation of HbF levels to varying extents (Fig. 1a and Supplementary

Data 1)11,12. Recent studies have begun to elucidate how specific variants in these proximal promoters act by either preventing or facilitating the interactions of trans-acting regulatory

factors, most notably the key HbF silencing factors BCL11A and ZBTB7A, with specific sequences13,14,15,16. In addition to variants affecting the HBG1/2 proximal promoters, large deletions

that span the entirety of the adult β-globin HBB and HBD genes also increase HbF expression to varying extents. Such deletions can be broadly classified into two categories: those that have

higher HBG1/2 mRNA and therefore HbF production, termed hereditary persistence of fetal hemoglobin (HPFH) deletions, and those that are characterized by lower HbF production with resultant

globin chain imbalance, termed δβ-thalassemia (Fig. 1a and Supplementary Fig. 1a). We and others have suggested that a 3.5 kb region upstream of the HBD gene may underlie the difference

between these two groups of deletions, although this remains to be functionally tested17,18,19.

a Schematic of HPFH SNVs and HBG-Δ13bp deletion at the proximal γ-globin promoter and large deletions within the β-globin locus that have been reported in individuals with δβ-thalassemias

(blue) and HPFH (purple). SNV single-nucleotide variant, HPFH hereditary persistence of fetal hemoglobin, HBG-Δ13bp 13 bp deletion at the proximal HBG1/2 promoter −101 to −114. b Gene

expression analysis for γ-globin (HBG1/2) and β-globin (HBB) mRNA in erythroid burst-forming units (BFU-E) derived from HSPC-derived erythroblasts upon genome editing of HBG-Δ13bp region in

the HBG1, HBG2, or both promoters, n = 3 biologically independent experiments. Results are shown as mean ± SEM (P values are labeled on the top of each comparison. *P