Engineering of the endogenous HBD promoter increases HbA2

The β-hemoglobinopathies, such as sickle cell disease and β-thalassemia, are one of the most common genetic diseases worldwide and are caused by mutations affecting the structure or production of β-globin subunits in adult hemoglobin. Many gene editing efforts to treat the β-hemoglobinopathies attempt to correct β-globin mutations or increase γ-globin for fetal hemoglobin production. δ-globin, the subunit of adult hemoglobin A2, has high homology to β-globin and is already pan-cellularly expressed at low levels in adult red blood cells. However, upregulation of δ-globin is a relatively unexplored avenue to increase the amount of functional hemoglobin. Here, we use CRISPR-Cas9 to repair non-functional transcriptional elements in the endogenous promoter region of δ-globin to increase overall expression of adult hemoglobin 2 (HbA2). We find that insertion of a KLF1 site alone is insufficient to upregulate δ-globin. Instead, multiple transcription factor elements are necessary for robust upregulation of δ-globin from the endogenous locus. Promoter edited HUDEP-2 immortalized erythroid progenitor cells exhibit striking increases of HBD transcript, from less than 5% to over 20% of total β-like globins in clonal populations. Edited CD34 +hematopoietic stem and progenitors (HSPCs) differentiated to primary human erythroblasts express up to 46% HBD in clonal populations. These findings add mechanistic insight to globin gene regulation and offer a new therapeutic avenue to treat β-hemoglobinopathies.


Introduction
Red blood cells, also known as erythrocytes, are packed with hemoglobin and circulate the body to supply all organs with oxygen. Hemoglobin is a hetero-tetrameric protein made up of two α-like and two -like subunits. Hemoglobin A1 (HbA1) accounts for approximately 97% of the hemoglobin expressed in adults and is composed of two -globin subunits (HBA) and two −globin (HBB) subunits. Hemoglobin A2 (HbA2) accounts for the remaining 2-3% of hemoglobin expressed in adults and is composed of two -globin subunits and two -globin (HBD) subunits 1 .
The -hemoglobinopathies, such as sickle cell disease (SCD) and -thalassemia, are caused by mutations in HBB that effect the structure or expression of -globin. The clinical hallmarks include hemolytic anemia and vaso-occlusion, which can lead to acute and chronic pain and organ damage. Clinical management is limited to frequent blood transfusions and life-long treatment of anemia and pain crises. The only curative approach is allogeneic stem cell transplantation, which is dependent upon HLA-identical donor availability 2 . Fetal hemoglobin (HbF), which is the predominant hemoglobin expressed before birth, has anti-sickling properties and its re-expression is frequently pursued as a treatment for -hemoglobinopathies 3 . While increasing HbF has shown to be clinically effective to combat SCD, studies have also validated the in vitro and in vivo anti-sickling abilities of -globin using a humanized mouse model of SCD [4][5][6][7] .
Increased HbA2 expression has some potential advantages over HbF that suggests it could provide an alternative avenue for to ameliorate -hemoglobinopathies. While HbF expression is completely silenced after birth, HbA2 is weakly transcriptionally active and expressed pancellularly in all adult red blood cells 8,9 . Additionally, -globin shares 93% homology to -globin, suggesting that HbA2 may be a better replacement relative to HbF, whose -globin subunit shares 73% homology to -globin. Finally, HbF is known to bind oxygen more tightly than HbA1, an evolutionary advantage for the fetus to draw oxygen from the maternal blood source, while HbA2 has a similar oxygen-binding capacity as HbA1 10,11 . On-going clinical trials re-express HbF to extremely high levels not commonly seen even in individuals with Hereditary Persistence of Fetal Hemoglobin 12 . The effect of extreme maternal HbF re-expression during pregnancy is currently unknown.
The genes for −globin (HBB) and -globin (HBD) are located in the -like globin cluster and regulated by the same control region. The -like globin cluster is located on chromosome 11, and harbors the five -like genes: HBB (-globin gene), HBD (-globin gene), HBG1 and HBG2 (-globin genes), and HBE (-globin gene). -globin and -globin both comprise 147 amino acids and differ at only 10 positions 13 . The extreme difference in expression levels between these two globins is not due to protein instability or differences in translation, but instead results from a lower transcription rate 1 . The globin genes are arranged in order of their expression during development and regulated by contact to a distal enhancer called the Locus Control Region that contains five active DNase Hypersensitivity Sites 14 . A comparative genomics study has shown that, compared to the HBB promoter, the HBD promoter has mutations in multiple transcriptional elements including a KLF1, CP1, -DRF and TFIIB binding site (Fig. 1A) 15 .
Previous studies using transgenic approaches have shown that inclusion of a KLF1 motif in the HBD promoter can drive exogenous expression of -globin 6, 16,17 . However, these studies do not reflect the complex chromosomal context and extensive epigenetic regulation at the -like globin cluster. Due to the large size of the -like globin locus, transgenic studies have mostly included only a subset of the genes of the -like globin locus and a minimal region of the LCR.
They therefore do not necessarily predict the biological outcomes of perturbations at the native -globin locus 18 .
The expression of globin genes is a tightly regulated development process. There are currently no drugs or therapeutic approaches to increase HbA2 for the -hemoglobinopathies.
Using CRISPR-Cas9 genome editing, we used homology directed repair at the endogenous HBD promoter to engineer the transcriptional elements present in HBB. We find that insertion of single transcriptional elements to the endogenous promoter is insufficient for -globin upregulation.
However, insertion of KLF1, -DRF, and TFIIB motifs drive high expression of -globin from the endogenous locus in HUDEP-2 cells and primary erythroblasts. This leads to reconstitution of high levels of HbA2, over 10-fold increase compared to WT unedited cells. Our work adds mechanistic insight to the globin gene regulation at the -like globin cluster and suggests a potential therapeutic avenue to upregulate HbA2 for the -hemoglobinopathies.

Targeting the endogenous HBD promoter to introduce functional HBB promoter elements
We aligned the promoter sequences of HBB and HBD to identify transcriptional elements missing in the HBD promoter (Fig 1a). This highlighted multiple mutations and deletions in the KLF1, CP1, -DRF, and TFIIB binding motifs. To re-engineer the endogenous HBD promoter, we employed CRISPR-Cas9 induced homology directed repair (HDR) gene editing 19,20 .
We designed and tested three sgRNAs targeting the HBD promoter ( Fig S1A) and three HDR templates that would incorporate base pairs needed to complete the transcriptional element motifs (Fig 1b). The HDR templates were designed as single stranded oligo donor nucleotides (ssODN) to either insert a KLF1 (K) sequence, a -DRF and TFIIB (DT) sequence, or all three elements (KDT). The -DRF and TFIIB motifs were designed into a single template because they are separated by only 2 base pairs. A previous transgenic study showed that the CP-1 site plays a minor role in HBB transcription, and was therefore omitted from our designs 21 . We maintained the ordering between motifs to mimic the HBB promoter and paid particular attention to limiting the number of mutated base pairs to increase the likelihood of successful HDR editing.
We performed editing in HUDEP-2 cells, an immortalized cell line capable of differentiating into hemoglobin-producing erythroid cells 22 , using Cas9 ribonucleoprotein (RNP) and an ssODN.
After optimizing conditions and testing multiple sgRNAs (Table S1), we found NHEJ rates ranging from 56 -74% and HDR efficiencies from 13 -27% as measured by next generation sequencing (NGS) (Fig 1c). The edited pools were differentiated to erythroblasts and qRT-PCR measurements were taken to assess the effects on HBD expression (Fig 1d). Despite roughly equal HDR rates between all promoter designs, we observed a statistically significant increase in HBD only for the KDT design with three elements.

HUDEP-2 cells
Heterogeneously edited pools of cells, containing a mixture of alleles, can mask large effects at a clonal level. We isolated heterozygous and homozygous HDR clones to more accurately assess the effect of each motif edit on HBD expression. We obtained at least one heterozygote and homozygote for each edit, as verified by amplicon NGS sequencing of the HBD promoter (Fig S1). We used ChIP-qPCR to test whether the inserted motifs were functional in recruiting KLF1 and RNA Pol II to the HBD promoter, comparing to the HBB promoter and VEGFA as a positive and negative control, respectively. Due to the unavailability of a TFIIB antibody suitable for ChIP, we performed ChIP for RNA Pol II, which is a binding partner of TFIIB in the pre-initiation complex (Fig 2a). The -DRF motif has been shown to be important for high transcription of HBB, but its bona fide binding factor has not yet been identified 23 . In the homozygous clones K and KDT, which harbor the edited KLF1 sequence, we observed increased binding of KLF1 to the HBD promoter relative to unedited WT HUDEP-2 cells. However, we found RNA Pol II binding at the HBD promoter binding only in the homozygous KDT clone. The homozygous DT clone, which harbors a TFIIB site, did not show significant binding of RNA Pol II.
Taken together, these data indicate that an intact KLF1 site is sufficient to recruit KLF1, but all three transcription factor binding sites are necessary to recruit the transcriptional machinery to the HBD promoter 24 .
We next performed qRT-PCR on all differentiated heterozygous and homozygous clones to look at percentage of HBD produced compared to total -like globins (Fig 2b). HBB expression was unaffected by the K, DT, and KDT knock-ins (Fig S2). Like the pooled editing results, we found that clones with single knock-ins of KLF1 or -DRF and TFIIB motifs alone did not show increased HBD. However, we observed a striking increase in HBD for KDT clones harboring all three motifs. The heterozygous clone showed HBD increases to 10% of total -like globins. The homozygous clone showed increases in HBD from less than 5% to over 20% of -like globins.
Next, hemoglobin protein levels were measured using high pressure liquid chromatography (HPLC) for the differentiated homozygous clones (Fig 2c). Peaks were assigned to hemoglobin complexes based on previous work that performed mass-spectrometry on each globin peak fraction 25 . We found that the transcript-level results are further supported at the protein level. The KLF1 and the -DRF and TFIIB clones did not show any measurable increase in HbA2, while the KDT homozygous showed a large increase in HbA2.

Endogenous editing of the HBD promoter increases HBD in CD34+ derived erythroblasts
To test the effects of HBD promoter engineering in a more clinically relevant cell type for a potential ex vivo therapy to treat -hemoglobinopathies, we edited human CD34+ derived erythroblasts. We used CRISPR-Cas9 to perform pooled knock-in at the HBD promoter in mobilized peripheral blood CD34+ human hematopoietic stem and progenitor cells (HSPCs) with the K, DT, and KDT RNP and ssODNs. We found HDR rates ranging from 18 -25% for all promoter knock-ins (Fig 3a). We differentiated the edited HSPC pools and performed qRT-PCR to measure HBD expression levels. As in HUDEP-2 cells, there was no statistically significant increase in HBD levels for K and DT, and a slight increase in two different donors in the KDT knock-in condition (Fig 3b).
To test the effect of KDT editing on a single-cell level, we isolated and grew colonies from HSPCs under erythroid expansion conditions. Each colony was genotyped and their alleles classified as either unmodified, NHEJ, or HDR (Fig S3A). These CD34+ derived clonal erythroblasts were differentiated and qRT-PCR was performed to determine HBD expression relative to total -like globins (Fig 3c). HBB transcript levels remain similar between unedited or NHEJ clones and KDT knock-in clones (Fig S3B). We observed that colonies with unedited or all NHEJ alleles had an average expression of lower than 3% HBD of total -like globins.
Heterozygous KDT knock-ins from multiple donors increased HBD expression dramatically, between 15 -35% of total -like globins in heterozygous knock-ins and between 17 -37% of total -like globins in homozygous knock-ins.

Discussion
Current gene editing efforts to treat the -hemoglobinopathies include correcting individual HBB mutations 26 , a method which would be limited to specific types of patient mutations, or increasing fetal hemoglobin 3 , a gene normally silenced after birth and with different oxygenbinding capacities than HbA1. In this study, we describe a path to upregulating HbA2, which shares high similarity to HbA1 and is applicable to all -hemoglobinopathy disease mutations.
Previous transgenic studies have shown that insertion of KLF1 alone to the HBD promoter sequence is sufficient to drive expression of HBD 6,16 . However, these studies do not reflect the complex chromatin context and regulation at the -like globin locus. We find that endogenous insertion of a KLF1 motif alone is insufficient to drive HBD expression. Instead, insertion of three motifs, KLF1, -DRF, and TFIIB, is necessary to recruit RNA Pol II and induce high expression of HBD. In CD34+ derived erythroblasts, homozygous knock-in of KLF1, -DRF, and TFIIB leads to increases in HBD up to 35% of total -like globins. To our knowledge, this is the first genomic editing of the HBD promoter that results in increased HbA2. Bulk editing of HBD through insertion of multiple transcriptional elements will rely on high levels of HDR. Increasing HDR outcomes might be achieved in CD34+ cells by a variety of methods such as controlled cell-cycling or modulation of DNA repair factors 27,28 , 29,30 .
Our promoter engineering approach dramatically increases HBD levels, but it is not yet clear whether HBD can be increased to the same levels as HBB. Using ChIP, we observed somewhat lower RNA Pol II binding at the HBD promoter than at the HBB promoter in the homozygous KDT clone. This suggests that further engineering of the HBD promoter could possibly increase HBD transcription. The -globin locus is tightly regulated by developmental epigenetics and erythroid-specific transcription factors. Further work might explore the insertion of more erythroid-specific or synthetic transcription factor binding sites in attempts to further increase in HBD.
It is estimated that roughly 30% of beneficial hemoglobin (10 pg per cell) is sufficient to ameliorate -hemoglobinopathy symptoms 9 . In our experiments, heterozygous and homozygous knock-in of KDT in CD34+ erythroblasts led to 15 -37% HBD relative to total -like globins.
Interestingly, in edited CD34+ erythroblasts, we observed that heterozygous and homozygous KDT populations expressed similar increases in HBD. Further studies should investigate whether heterozygous knock-in of KDT in -hemoglobinopathy cells is sufficient to ameliorate disease phenotypes. For example, one could knock-in KDT to the HBD promoter of SCD patient HSPCs and perform HPLC or microscopy assays to determine the anti-sickling effects of a heterozygous or homozygous knock-in. If heterozygous knock-in of KDT is sufficient, one would see a decrease in sickle hemoglobin produced, or a decreased percentage of sickled red blood cells.
In summary, our presented study provides a novel strategy to increase levels of HbA2 from the endogenous HBD locus that could potentially be applicable as an ex-vivo gene therapy.
In pre-clinical studies, it will be important to quantitatively investigate the amount of HBD that is optimal for improving the function and health of patient red blood cells and explore the overall potential and safety of increasing HbA2 as a therapeutic option for the -hemoglobinopathies.

Figure Legends
The data is presented as mean ± SD of three biological replicates. P value indicates paired, twotailed student t test (ns, non-significant; *, P ≤ 0.05; **, P ≤ 0.01). The three biological replicates from the editing experiment in C) were each differentiated and the data is presented as mean ± SD of three biological replicates. P value indicates paired, two-tailed student t test (ns, non-significant; *, P ≤ 0.05; **, P ≤ 0.01).
Briefly, guide sequences were ordered as oligonucleotides and formed into duplexes using a PCR thermocycler. The DNA template was transcribed to RNA using HiScribe™ T7 High Yield RNA Synthesis Kit (E2040S) following manufacturer protocol. The resulting RNA was purified using RNeasy Mini kit (74104) and Rnase-Free DnaseI Kit (79254).

mPB-HSPCs cell culture and differentiation
For editing for human CD34+ cells, CD34+ mobilized peripheral blood HSPCs were thawed and cultured in SFEM containing CC110 supplement (Stemcell Technologies 02697) for 2 days.  Table S1.

ChIP-qPCR
ChIP was performed as done previously 25

Next-Generation Sequencing (NGS) amplicon preparation and analysis
To confirm homozygosity/heterozygosity of the HBD clones, samples' gDNA extracted with QuickExtract DNA Kit was first amplified by PCR. The primers were designed specifically for NGS, spanning a region of < 200 bp (including the primers sequences) in which the cutsite is asymmetrically placed (e.g., 30-80 bp from the forward or the reverse primer) to capture the edited region. Subsequently, two stubber sequences are added, one for the forward primer

Reference G G G A G G A C A G G A C C A G C A T A A A A G G C A G G G C A G A G T C G A C
sgRNA 93.84% (382159 reads)

K heterozygous
Reference G G G A G G A C A G G A C C A G C A T A A A A G G C A G G G C A G A G T C G A C sgRNA 0.39% (957 reads) 0.42% (1030 reads) 93.01% (230215 reads)