Forced enhancer-promoter rewiring to alter gene expression in animal models

Transcriptional enhancers can be in physical proximity of their target genes via chromatin looping. The enhancer at the β-globin locus (locus control region [LCR]) contacts the fetal-type (HBG) and adult-type (HBB) β-globin genes during corresponding developmental stages. We have demonstrated previously that forcing proximity between the LCR and HBG genes in cultured adult-stage erythroid cells can activate HBG transcription. Activation of HBG expression in erythroid cells is of benefit to patients with sickle cell disease. Here, using the β-globin locus as a model, we provide proof of concept at the organismal level that forced enhancer rewiring might present a strategy to alter gene expression for therapeutic purposes. Hematopoietic stem and progenitor cells (HSPCs) from mice bearing human β-globin genes were transduced with lentiviral vectors expressing a synthetic transcription factor (ZF-Ldb1) that fosters LCR-HBG contacts. When engrafted into host animals, HSPCs gave rise to adult-type erythroid cells with elevated HBG expression. Vectors containing ZF-Ldb1 were optimized for activity in cultured human and rhesus macaque erythroid cells. Upon transplantation into rhesus macaques, erythroid cells from HSPCs expressing ZF-Ldb1 displayed elevated HBG production. These findings in two animal models suggest that forced redirection of gene-regulatory elements may be used to alter gene expression to treat disease.


INTRODUCTION
Transcriptional enhancers can be located at great genomic distances from their target promoters. Chromatin looping that physically juxtaposes enhancers and promoters is one mechanism by which regulatory information may be transmitted, even though there is a vigorous debate about how widespread this mechanism is and how "proximity" translates into actual physical distances. [1][2][3][4] The b-globin gene cluster comprises a distal enhancer, termed locus control region (LCR), that is required for high-level transcription of all b-type globin genes in erythroid cells. 5,6 Based on chromosome conformation capture (3C) experiments, the LCR is thought to contact globin genes 7,8 requiring erythroid-specific transcription factors, 9 including GATA1 and its coregulators FOG1 and Ldb1. 10,11 "Contact" is operationally defined here as an increased 3C signal with no implications for actual Euclidean distance.
LCR-gene contacts occur in a developmental-stage-specific manner. Thus, the LCR is in proximity with the fetal-type (HBG1/2) and adult-type (HBB and HBD) globin genes during fetal and adult stages, respectively. 12,13 Interest in the molecular underpinnings of the switch from HBG to HBB expression, which occurs around the time of birth, has been stimulated by the recognition that elevated fetal hemoglobin (hemoglobin F [HbF]) levels benefit patients with sickle cell disease (SCD), a devastating disorder caused by a point mutation affecting the HBB gene (20A>T, Glu6Val). [14][15][16] HbF consists of the products of the HBG and HBA genes (g 2 a 2 ), while adult hemoglobin consists predominantly of the products of the HBB and HBA genes (b 2 a 2 ).
Our prior work examined the cause-effect relationship of looped enhancer-promoter contacts by forcing LCR-promoter interactions using an artificial "looping factor." Specifically, designer zinc-finger (ZF) proteins engineered to bind to the murine b-globin promoter were fused to the dimerization domain (DD) of the GATA1 cofactor Ldb1 (ZF-Ldb1). Ldb1 was chosen because it had been shown to be essential for enhancer-promoter contacts at the b-globin locus. 11 ZF-Ldb1 was capable of promoting LCR-promoter contacts and augmenting transcription. 17 Similar results were obtained when tethering Ldb1 via dCas9. 18 Moreover, ZF-mediated tethering of Ldb1 to the HBG genes in adult human erythroid cells increased HBG transcription at the expense of HBB and HBD transcription, 19,20 consistent with a model in which HBG and HBB/HBD genes may compete for LCR activity [21][22][23][24] ( Figure 1A). Together, these results suggested that engineering genome folding to rewire enhancer-promoter contacts might be a strategy for reactivation of the HBG genes to treat SCD. 25 The aforementioned studies were carried out with cultured erythroid cells in vitro. Here we extended these proof-of-concept studies to preclinical murine and non-human primate models with the goal to examine whether the forced reshaping of enhancer-promoter contacts might be a viable strategy to alter gene expression for therapeutic purposes.

Testing of ZF-Ldb1 constructs in vitro
ZF-Ldb1 constructs containing the DD of Ldb1 fused to the ZF domain were inserted upstream of an IRES-GFP cassette into a vector containing the promoter of the ankyrin (ANK1) gene to drive erythroid-specific expression. We tested four different ZF moieties targeting both HBG genes (HBG1 and HBG2, which are adjacent to each other and are thought to have arisen through gene duplication) -GG23 (pcl20-ANK1-GG23DD-IRES-GFP), GG38 (pcl20-ANK1-GG38DD-IRES-GFP), GG52 (pcl20-ANK1-GG52DD-IRES-GFP), and GG63 (pcl20-ANK1-GG63DD-IRES-GFP), with each ZF targeting different areas of the HBG promoters ( Figure 1B). When tested in fetal-type K562 cells, all ZF-Ldb1 constructs bound to their targets, albeit with varying intensities ( Figure 1C). 17,19,20 To test the ability of ZF-Ldb1 to induce HbF synthesis in primary adult erythroid cells, we lentivirally transduced human CD34 + cells and differentiated them toward the erythroid lineage in a three-phase culture system, as described previously. 20,26 Significant induction of g-globin (the product of the HBG genes) in GFP + cells was achieved with all tested vectors, as measured by cation-exchange high-performance liquid chromatography (HPLC) ( Figure 1D). The GG38-, GG63-, and GG52-based constructs were the strongest HbF inducers, with levels similar to those achieved previously with a ZF construct termed GG1-Ldb1, which targets the same region of the HBG promoter as does GG63. 19 There was no direct correlation of ZF-Ldb1 binding intensity, as measured by chromatin immunoprecipitation (ChIP) in K562 cells and HbF induction in CD34 + cells, which might be due to different binding site distances from the transcriptional start sites or differences between K562 cells and primary human cells.
To determine which ZF-Ldb1 constructs maximized induction of HBG while minimizing off-target effects, RNA sequencing (RNAseq) was performed on day 13 of in vitro erythroid differentiation of human primary CD34 + cells and compared with GFP empty vectortransduced cells. GG38, GG63, GG52, and GG1 increased HBG1/2 mRNA levels with a concomitant decrease in the mRNA levels of the adult type genes HBB in all samples and HBD in GG38-, GG63-, and GG1-treated samples ( Figure 1E). Importantly, GG38 and GG63 showed the fewest differentially expressed genes (DEG = 33 and 17, respectively; absolute fold change >1.5, false discovery rate [FDR] < 0.05) compared with GG52 (DEG = 409) or GG1 (DEG = 677) ( Figure 1F; Tables S1-S4). Thus, we focused the remainder of our studies on the GG38-Ldb1-and GG63-Ldb1-expressing vectors. Our previous work in human cell cultures showed that forced LCR-HBG contacts can stimulate HBG transcription in a manner well tolerated by the cells. 19,20 To test the efficacy of this approach in a whole-animal model, we employed the BERK mouse model. BERK mice lack murine aand b-type globin genes and instead carry a transgene that includes a compressed form of the human LCR and the human HBA, HBG, HBD, and HBB genes, the latter containing the SCD mutation. 27 The impaired health of homozygous transgene-carrying animals results in poor breeding and intolerance to myeloablative conditioning, which is required for autologous bone marrow transplantation. However, heterozygous animals are healthier, enabling transplantation studies. We tested the efficacy of the GG38-and GG63-carrying vectors by performing myeloablative transplantation of mouse bone marrow lentivirally transduced with either GFP-only control or GG38/GG63-Ldb1-GFP constructs into heterozygous (mouse b A human b S ) BERK mice, followed by analysis  19 ). Red + indicates the relative degree of gene expression. (B) Specific target sites within the b-globin locus for each of the ZF-Ldb1 constructs. (C) Anti-HA ChIP-qPCR performed in K562 cells shows the specificity of all zinc-finger (ZF) constructs to the HBG locus. Data are displayed as signal in the immunoprecipitation (IP) fraction compared with total input; n = 1. (D) Expression levels of HbF as a percentage of total globins (HbF + HbA) by cation-exchange HPLC of GFP + cells on day 15 of CD34 + in vitro culture following lentiviral transduction of CD34 + primary human cells with either GFP control or ZF-Ldb1 constructs. n = 2-3 independent donors. Statistical analyses were done using one-way ANOVA. Error bars represent standard deviation. ns, not significant; *p < 0.05; **p < 0.01. (E) RNA-seq analysis of in-vitro-differentiated erythroid cells on day 13 of culture following lentiviral transduction of CD34 + primary human cells with either GFP control or ZF-Ldb1 constructs. n = 2 independent donors. Note that multiple genes may be overlapping in this visualization. Red dot, absolute fold change > 4 and false discovery rate (FDR) < 0.05; blue dot, absolute fold change between 1.5 and 4 and FDR < 0.05; black dot, FDR > 0.05 (not significantly changed). (F) Number of DEGs (absolute fold change greater than 1.5 and FDR < 0.05) among ZF constructs. TSS, transcription start site; HS2, DNAse hypersensitivity site 2 of the LCR; HBE, ε-globin; HBG, g-globin; HBD, d-globin; HBB, b-globin; LCR, locus control region.
of HbF and HBG expression 10 weeks post transplantation (Figure 2A). Reverse-phase HPLC analysis of GFP + bone marrow cells showed a 2.7-fold induction of HbF in GG38-transplanted mice and a 3.6-fold induction in GG63-transplanted mice ( Figure 2B). Accordingly, western blot analysis revealed significantly increased expression of g-globin in the GG38 and GG63 cohorts (Figures 2C and S1). Because the BERK mice used in this study were heterozygous for the human globin genes with minimal symptoms of SCD, it precluded examination of disease parameters, such as erythroid cell indices, spleen size, or urine concentration. Taken together, these data reveal that forced chromatin looping can be employed to activate HBG expression in a whole-animal model.

Optimization of ZF-Ldb1 constructs for use in rhesus macaques
Recent advances in lentivirus-based vector development for red cell disorders suggest that addition of the BCL11A +58 enhancer enhances erythroid-specific expression of transgenes in vivo. 28 We therefore paired the BCL11A +58 enhancer with either the ANK1 or glycophorin A (GPA) promoter to drive expression of the GG38/GG63-Ldb1 constructs ( Figure 3A). In addition, the vectors contained a P2A sequence followed by a Venus or mEmerald cassette, allowing monitoring of gene expression by flow cytometry. Primary human CD34 + cells from two independent healthy donors were transduced with the vectors at MOIs of 20 or 100, matured ex vivo using a two-phase erythroid differentiation culture system, and analyzed for transgene expression by flow cytometry and for HbF induction by HPLC ( Figure 3B). Transduced cells displayed no alterations in viability, as determined by flow cytometry (Figure S2). Furthermore, the ANK1 promoter led to more sustained transgene expression at later stages of erythroid differentiation, regardless of the construct expressed ( Figure S2). HPLC analysis on days 9 and 12 showed that the ANK1 promoter-driven constructs produced the highest levels of HbF for both ZF constructs, with HbF levels of nearly 40% seen with the GG38 construct and 40%-50% with the GG63 vector ( Figures 3C and S3). Notably, the +58/ANK1-GG63-P2A-mEmerald vector and the +58/ANK1-GG38-P2A-Venus vector triggered significant HbF induction (33%-66% HbF, 2.5-6-fold HbF induction) with high sustained transgene expression (Venus + 33%-70%) and vector copy numbers (VCNs) (1.5-4.3) (Figures 3C, S2, and S3), suggesting that these two vectors were the best for further use.

ZF-Ldb1 constructs robustly induce HbF in adult rhesus erythroid cells ex vivo
To test whether HbF levels could be raised via forced enhancerpromoter looping in a non-human primate model, we tested our vectors in cultured rhesus macaque cells. Rhesus macaque CD34 + progenitor cells were isolated from granulocyte colony-stimulating factor (G-CSF)-mobilized peripheral blood (PB) as reported previously, 29  we transduced rhesus macaque erythroblasts derived from peripheral blood mononuclear cells (PBMCs) with these vectors, followed by in vitro erythroid differentiation. Resulting HbF levels ranged from 48.4%-82.8% compared with GFP only (38.3%) and NTC (32.6%) ( Figure S4A). In CD34 + hematopoietic stem and progenitor cells (HSPCs) and PBMCderived erythroid cells, ANK1 promoter-and GG38-containing constructs induced higher HbF levels than GPA promoter-and GG63containing constructs ( Figures 4A and S4A). Similarly, the mEmerald or Venus signal was lower in GPA groups compared with ANK1 groups ( Figures 4B and S4B). These results identified the ANK1-GG38 construct as the most robust in terms of expression and HbF induction.

ZF-Ldb1 constructs induce HbF in transplant-derived erythroid cells in vivo
Because the ANK1-GG38 combination was found to be superior in terms of HBG induction, we used this construct in autologous myeloablative rhesus macaque transplants. After transduction of G-CSF-and plerixafor-mobilized CD34 + HSPCs, a fraction of cells was frozen, and another fraction was differentiated toward the erythroid lineage in vitro and examined for VCN and HBG expression ( Figure S5). While the percentage of Venus + cells was relatively low (9.96%-14.70%) in differentiated cells, the VCN ranged between 4.3-7.8, and g-globin levels (45.6%-46.5%) were elevated compared with NTC groups (15.9%-20.9%) ( Figures 4C  and 4D). Two animals, ZM31 and ZL52, were infused with 3.22-4.72 Â 10 6 CD34 + HSPCs/kg and engrafted with typical reconstitution kinetics (Table 1). Blood parameters in transplanted animals were within normal ranges after engraftment throughout the study ( Figure S6).

Long-term increases in HbF levels in ZF-Ldb1-expressing erythroid cells
After cell infusions, PB samples were collected regularly to evaluate VCNs in cell fractions, HbF + cell (F-cell) percentage, and g-globin expression in erythroid cells from transplanted animals. Stable VCNs in myeloid ($0.3 for ZM31, $0.6 for ZL52) and lymphoid ($0.3 for ZM31, $0.8 for ZL52) cells were detected at 135 weeks after transplantation ( Figure 5A). In bulk analyses, F-cells ( Figure 5C) and g-globin expression ( Figure 5D) were high only early post transplantation but receded to control animals' levels 30-45 weeks post transplantation, consistent with decreased Venus reporter expression (Figure 5B). Although the VCNs were considerable, Venus + erythroid cells were limited (1%-3%) and not substantially different than in other cell fractions, including myeloid and lymphoid cells for both an-imals ( Figure 5B), suggesting that decreased transgene expression was not simply a consequence of loss of Venus expression in enucleated circulating erythroid cells. To investigate whether ZF-Ldb1 expression was toxic to erythroid cells, we evaluated the VCN in CD20 + B-cells, CD45 À CD71 + erythroid progenitors, and all other CD45 + cells (CD45 + CD20 À ) from bone marrow mononuclear cells of the transplanted animals. The results show that erythroid progenitors had VCNs comparable with those in other cell lineages tested, suggesting that ZF-Ldb1 expression was not selectively toxic to erythroid progenitors ( Figure S7).
Although ZF-Ldb1 was not stably expressed over time in bulk populations, we hypothesized that the subset of cells expressing ZF-Ldb1 (Venus + cells) might still respond to its activity. We sorted Venus +  Figure 6A). We investigated whether there was any difference in g-globin expression relative to the degree of Venus expression. We sorted Venus-mid and Venus-high populations approximately 2 years post transplantation and evaluated g-globin expression levels by RP-HPLC. No differences were observed between mid-and high-Venus-expressing populations (Figure 6B). Nevertheless, these data indicate the long-term functionality of the ZF-Ldb1 construct to induce HbF expression and are consistent with sustained HbF induction in GFP + cells in our BERK murine transplant studies utilizing the GG38-Ldb1 and GG63-Ldb1 constructs ( Figure 2). Taken together, our data suggest that forced enhancer-promoter rewiring in murine and macaque models allows significant increases in HBG gene expression.

DISCUSSION
Our prior work showed that ZF-mediated tethering of Ldb1 to the HBG promoter can reactivate its expression in normal and SCD patient-derived cultured adult erythroblasts. 19,20 More recently, a study using transcriptional activator-like effectors (TALEs) to tether Ldb1 to the HBG promoters has shown similar results. 30 Here, we find that the optimization of our ZF-Ldb1 vector achieved HbF induction comparable with our prior studies 19,20 ( Figure 1D) while reducing offtarget effects on the transcriptome (Figures 1E and 1F). Addition of the BCL11A +58 enhancer in combination with the ANK1 promoter and the GG38 and GG63 ZF moieties achieved considerable HbF expression in rhesus macaques and human primary cells. Furthermore, we show for the first time in two independent animal models that forced proximity of the LCR with the HBG promoters augments HbF expression in vivo. A naturally occurring gain-of-function mutation at the HBG genes that triggers recruitment of the Ldb1 complex causes elevated HBG expression throughout adult life, 31 providing genetic evidence that Ldb1 tethering to the HBG genes may be a potent means to reactivate them.
Although the number of F-cells and HbF levels were significantly upregulated in the first 3 months post transplantation in rhesus macaques, levels gradually decreased and stabilized to around pre-transplantation levels 30-40 weeks after transplantation. VCNs in the transplanted animals remained stable throughout the 135 weeks of follow up. Thus, the drop in F-cells and HbF levels was not due to loss of vector-containing cells over time. Although it is established that transplantation itself induces a transient rise in HbF in early transplantation periods, 32 we have demonstrated previously that transplantation of rhesus macaques with CD34 + HSPCs transduced with a GFP vector induced lower levels (2.53%) of HbF only during the early phase of transplantation, 33 suggesting that ZF-Ldb1 expression contributes to the observed HbF induction. This notion is supported by Venus + RBCs having 8%-24% HBG levels, while Venus À RBCs displayed HBG expression levels similar to controls.
One major limitation of the study is that Venus expression remained low despite high VCNs ( Figure 5). This is potentially due to low gene expression in rhesus cells from use of a polycistronic vector with Venus in the third position 28 or gene silencing, possibly via DNA methylation of the integrated vector. Decreased expression in vivo, perhaps because of positional effects, 34 could explain why some cells continue to have Venus expression long term, while others do not express detectable Venus reporter despite VCN data suggesting that most of the cells carry the vector. We have shown recently that while shRNA-mediated BCL11A knockdown in rhesus macaques induced HbF only transiently, sustained shRNA expression and increased HbF production was achieved by incorporating a truncated form of the human erythropoietin receptor (thEpoR) into the vector. 28 thEpoR expression provides a selective advantage to transduced erythroid precursor cells. We envision that this approach might be similarly beneficial in our system to achieve sustained ZF-Ldb1 expression in vivo.
The use of clinically relevant vectors, 35 the identification of key regulatory regions, including the BCL11A +58 erythroid-specific enhancer, 36,37 and the application of the ANK1 insulator-promoter in vector development 38 have driven much of the initial optimization shown in this study. Further improvement of the forced chromatin looping strategy to achieve robust, sustained, and pancellular expression might provide an avenue for future therapeutic applications. Tethered Ldb1 is capable in non-erythroid cells, such as neuronal progenitor cells, of efficiently generating long-range contacts between genomic segments much farther apart than at the b-globin locus ($500 kb), 39 suggesting that forcing long-range chromatin contacts via Ldb1 tethering might be more broadly applicable.
Taken together, our findings serve as a proof of concept demonstrating that forced chromatin looping may be developed to become a useful tool for treatment of genetic disorders. ZF domains may be less antigenic compared with TALEs or Cas9, can be engineered to a high level of target specificity, and can be used to tether protein moieties that, in turn, reshape chromosome conformation. In this manner, via engagement of a powerful distal enhancer, such as the LCR, gene expression changes may be increased to levels not achievable by simple fusion to ZF moieties of conventional transactivation domains. We envision that altering nuclear architecture in a manner that re-connects enhancers with their natural target genes might be applicable to diseases in which enhancer-promoter   for the treatment of b-thalassemia major 42,43 and SCD. 44 Recent studies have identified potential therapeutic targets for rewiring of enhancer-promoter interactions, including EPHA4-related limb malformations 40 and SOX9-mediated craniofacial disorders, 45 among others. 46 One advantage of forced enhancer-promoter rewiring might be that it not only reduces expression of the mistargeted gene but also restores expression of the natural enhancer target gene. Thus, advances such as those described here may help to allow targeted treatment of enhanceropathies across the age spectrum for a wide variety of inherited and acquired disorders that involve mis-wired or dysfunctional enhancers.
It is important to note that advances in gene editing/manipulation technologies have yielded attractive new approaches for the treatment of SCD, including altering the expression of BCL11A 47,48 or the HBG genes directly, 24,49,50 or to convert the SCD mutation into a harmless variant. 51 In light of these developments, we are not proposing to use forced chromatin looping as a therapeutic modality in this context. Instead, we provided proof of concept that rewiring of enhancer-promoter contacts might be a means to alter gene expression in settings where regulatory elements or genome architectural features are disrupted. That said, there are additional considerations that might make forced chromatin looping an advantageous strategy. Redirecting an enhancer away from its original target is expected to dampen transcription of the target gene. In the case of the b-globin cluster, our prior work has demonstrated that the gain in HBG transcription was accompanied by a reciprocal reduction in HBB expression. 19 In the case of SCD, in which a mutant form of HBB is the root cause of the disease, this approach combines the benefits of increased HbF levels with a reduction of the toxic form of HBB. These combinatorial effects might be beneficial when considering this strategy for the treatment of inherited diseases and enhanceropathies in cases where gene editing is not applicable.

Animals
Mice were maintained at the Children'  were obtained from the University of Pennsylvania Human Immunology Core (Philadelphia, PA, USA). CD34 + cells were purified by using a MACS MicroBead Kit (Miltenyi) and cultured in a 3-phase in vitro culture, as described previously. 52 Lentiviral stocks were generated by co-transfection of the previously described gene transfer plasmids (pcl20-ANK1-GFP, pcl20-ANK1-GG23DD-IRES-GFP, pcl20-ANK1-GG38DD-IRES-GFP, pcl20-ANK1-GG63DD-IRES-GFP, pcl20-ANK1-GG52DD-IRES-GFP, or pcl20-ANK1-GG1DD-IRES-GFP 19 ) together with the envelope plasmid (vesicular stomatitis virus [VSV]-G), the packaging plasmid (pMDLg/p RRE), and the pRSV-REV plasmid into HEK293T cells, and the viral supernatant harvested and concentrated by ultracentrifugation using a protocol described previously. 20 The ZF proteins tested in the present study bind to sequences located at À443 bp (GG23), À182 bp (GG38), À114 bp (GG63 and GG1), and À45 bp (GG52) with respect to the HBG1/2 transcription start sites, following previously established design principles, 19,53 including target sites located in accessible chromatin as well as at sites where there are no known transcription factor binding sites. Specifically, the tested ZF constructs target the following sequences located in the g-globin promoter region: GG23, 5 0 -TTAGGCATAG GTCCAGGATT-3 0 ; GG38, 5 0 -TTGAGATAGTGTGGGGAAGG-3 0 ; GG63 and GG1, 5 0 -GGTCAAGGCAAGGCTGGCCA-3 0 ; GG52, 5 0 -A GCCGCCGGCCCCTGGCCTC-3 0 . Viral titration was performed using K562 cells and titered to the virus transduction unit of 1E8-1E9 particles/mL. K562 or CD34 cells were spin-infected at 2,250 rpm at room temperature for 1.5 h as described previously, 19 and GFP + cells were sorted by fluorescence-activated cell sorting (FACS) on day 2 post infection. Cation-exchange HPLC for quantification of HbF was performed as described previously. 52,54 ChIP Anti-hemagglutinin (HA) ChIP-qPCR was performed in K562 cells using human-specific primers for HS2, HBE, HBG, HBD, HBB, and CD4 as described previously. 19 RNA extraction, qRT-PCR, and RNA-seq RNA samples were harvested in TRIzol (Thermo Fisher Scientific) and purified with the RNeasy Mini Kit (QIAGEN). For RNA-seq, 200 ng RNA, isolated as just described, was depleted for ribosomal RNA by using the Ribo-Zero removal reagents and protocol from the ScriptSeq Complete Kit (Illumina), followed by purification with the RNeasy MinElute Clean Up Kit (QIAGEN). Sequencing libraries were constructed and analyzed, libraries were sequenced in pairedend mode, and reads were processed as described previously. 26,55 Murine bone marrow isolation and lineage -Sca1 + ckit + (LSK) sorting All BERK mice used for experiments were heterozygous for the sickle transgene (mouse b A human b S ), confirmed by RP-HPLC with appropriate mouse and human globin controls. Male BERK mice were euthanized at 2-3 months of age by CO 2 narcosis. Cells from bone marrow were harvested by flushing the tibiae, femora, and hip bones with PBS containing 0.5% BSA and 2 mM EDTA. Lineage À cells from total bone marrow were isolated using the Lineage Cell Cultured LSK cells were transduced with lentiviruses on day 2 using a RetroNectin (T100B, Takara)-based method. 56 Briefly, lentivirus carrying pcl20-ANK1-GFP, pcl20-ANK1-GG38DD-IRES-GFP, or pcl20-ANK1-GG63DD-IRES-GFP were loaded into RetroNectin coated 12-well plates to aim for a transduction efficiency of 25%-30%. Cultured LSK cells were transferred to the lentivirus-loaded plates and incubated for 1 more day. On day 3, 285K LSK cells were mixed with 500,000 Sca1-depleted competitor BM cells and injected retro-orbitally into lethally irradiated (12 Gy, split dose, Orthovoltage Precision X-ray) recipient heterozygous BERK mice. GFP + BM cells (constituting 5%-10% of transplanted mouse bone marrow) were isolated via FACSAria-based cell sorting.

RP-HPLC analysis of transplanted BERK mice
For BERK murine transplant analysis, RP-HPLC analyses were conducted as reported previously, with slight modifications. 20,57 Briefly, murine whole bone marrow or GFP + bone marrow samples were lysed in 100 mL MilliQ water and centrifuged at 10,000 Â g for 10 min. A 50-mL aliquot of the supernatant was injected on a Hitachi D-7000 HSM series apparatus (Hitachi Instruments, San Jose, CA, USA) using an Aeris 3.6-mm, Widepore C4 200-Å, liquid chromatography (LC) 100 Â 4.6-mm column (Phenomenex, Torrence, CA, USA) and a gradient from 20%-60% acetonitrile in 0.1% trifluoroacetic acid over 25 min with UV detection at 215 nm. Types and relative quantities of hemoglobins in samples were assessed by comparison with standard hemoglobin controls.

Western blot analysis
Western blot analysis was performed as described previously. 26 Briefly, fluorescent western blotting was performed on Immobilon-FL polyvinylidene difluoride membranes (Millipore), blocking was performed in Odyssey blocking buffer (LI-COR Biosciences), and antibody staining was performed in blocking buffer diluted 1:1 in Tris-buffered saline-0.1% Tween 20. Primary staining was performed overnight with gentle shaking at 4 C, and secondary staining was performed for 1 h at room temperature. Primary antibodies included g-globin

Vector packaging and lentiviral production
Previously published protocols were modified for use in erythroid ZF vector optimization. 58,59 Briefly, vectors were packaged with chimeric simian human immunodeficiency virus (SHIV), produced in HEK293T suspension cultures, Mustang Q purified (XT5 ion-exchange capsule, Pall Life Sciences), and concentrated by diafiltration in X-VIVO 10 medium (Lonza, Walkersville, MD, USA) to achieve a final concentration of about 50-fold from the starting material. The vectors were then aliquoted and stored at À80 C before titration using HOS cells as described previously. 59 Vector optimization, transduction, and CD34 + erythroid differentiation testing CD34 + erythroid differentiation of transduced cells was performed using a two-phase system as described previously. 59 Flow cytometry of rhesus macaque mEmerald (with a 515/20 filter, Chroma Technology) and Venus (with a 525/LP filter, BD Biosciences) was performed using a CytoFlex (Beckman Coulter) instrument on days 5, 7, 9, and 12 of differentiation. FlowJo v.10 (FlowJo, Ashland, OR, USA) was used for data analysis and graphic representation. Samples from days 9 and 12 of erythroid differentiation were prepared for cation-exchange HPLC analysis by isolating 100,000 erythroid cells, washing with PBS, lysing in hemolysate reagent (Helena Laboratories, 5125), and centrifuging at 14,000 Â g for 10 min. Quantification of hemoglobin tetramers and individual globin chains was performed using ion-exchange columns on a Prominence HPLC system (Shimadzu). Proteins eluted from the column were identified at 220 and 415 nm with a diode array detector. The relative amounts of hemoglobins or individual globin chains were calculated from the area under the 415-nm peak, with percentage of HbF = [HbF/(HbA + HbF)] Â 100. Types and relative quantities of hemoglobins in samples were assessed by comparison with standard hemoglobin controls.

Rhesus macaque PBMC and CD34 + isolation and transplantation
Rhesus macaque PBMCs were isolated from PB using Ficoll-Paque PLUS density gradient medium (GE Healthcare) following manufac-turer recommendations. G-CSF-mobilized (Amgen) and plerixaformobilized (Amgen) rhesus macaque CD34 + HSPCs were collected as described previously. 29,60 In short, the animals were treated with a 5-day course of 15 mg/kg G-CSF (Amgen) subcutaneously and a single subcutaneous dose of 1 mg/kg AMD3100 (Sigma-Aldrich) on the morning of the fifth day, 3-4 h before leukapheresis. A small-volume leukapheresis procedure was performed using a CS3000 cell separator (Baxter Fenwal), and CD34 + HSPCs were immunoselected using a rhesus macaque CD34 + antibody (clone 12.8, Fred Hutchinson Cancer Research Center) and an anti-mouse IgM bead (Miltenyi Biotech). A CD34 + antibody (550761, BD Biosciences) was used to determine the purity of the immunoselected population.
Rhesus macaque CD34 + lentiviral transduction, ex vivo erythroid differentiation, and transplantation Rhesus macaque PBMCs and mobilized CD34 + HSPCs were transduced with lentiviral particles expressing ZF constructs using a high-density transduction protocol as described previously. 35 Briefly, progenitor cells were pre-stimulated in serum-free X-VIVO10 medium (Lonza) containing SCF, Flt-3, and TPO (100 ng/mL each, R&D Systems) at a density of 2 Â 10 6 cells/mL for 24 h. Then, the medium was replaced with fresh pre-stimulation medium with growth factors and ZF constructs at MOI 10 and 50 for PBMCs and CD34 + HSPCs, respectively. After 24 h of transduction, the majority of cells were cryopreserved for later transplantation; a small portion of the transduced CD34 + HSPCs were used for RBC differentiation to confirm VCN, reporter gene (Venus), and HbF induction using a previously reported protocol. 61 Briefly, 1 Â 10 5 cells/mL were inoculated into tissue culture plates (Corning) coated with irradiated OP-9 cells (ATCC, Crl-2749) in erythroid proliferation medium consisting of Iscove's modified Dulbecco's medium (Mediatech) with 10 ng/mL SCF, 1 ng/mL IL-3 (R&D Systems), 2 U/mL erythropoietin (EPO, Amgen), 10 À6 M dexamethasone (VETone), 10 À6 M estradiol (Pfizer), and 20% FBS (Gibco) for 6 days. Then, cells were incubated for an additional 8 days in maturation medium containing 10 ng/mL insulin (Lilly), 500 mg/mL holo-transferrin (Sigma-Aldrich), 2% (w/v) bovine serum albumin (Roche), 2 U/mL EPO, and 20% FBS. DNA and protein samples were isolated from differentiated cells to determine VCN and g-globin protein expression, respectively. Lentivirally transduced CD34 + HSPCs (specific cell numbers are listed in Table 1) were then reinfused into rhesus macaques following 5 Gy Â 2 total body irradiation at a dose of 0.6 Gy/min using a Co-60 irradiator located at the Armed Forces Radiobiology Research Institute (Bethesda, MD, USA).

Quantitative PCR
Genomic DNA was extracted from ex-vivo-differentiated cells, PB myeloid cells, or lymphoid cells isolated using the DNeasy Blood and Tissue kit (QIAGEN) following the manufacturer's protocol. VCNs were measured by a QuantStudio 6 Flex real-time PCR system (Thermo Fisher Scientific) using a self-inactivating long terminal repeat (SIN-LTR) probe and primers and ribosomal RNA probe and primers (TaqMan ribosomal RNA control reagents, Applied Biosystems). 62

Flow cytometry
For rhesus macaque analysis, rhesus macaque mEmerald and Venus expression in ex-vivo-transduced cells or PB cell fractions were analyzed using flow cytometry (FACSCanto, BD Biosciences). The F-cell percentage in PB RBCs of the transplanted animals was determined by a previously published protocol using a primary HbF-specific antibody (551796, BD Biosciences) and an allophycocyanin (APC)-conjugated secondary antibody (550874, BD Biosciences). 33

Hemoglobin electrophoresis
Hemoglobin content in differentiated PBMCs and CD34 + HSPCs was determined using cellulose acetate membranes according to the manufacturer's instructions (Helena Laboratories, TX, USA).

RP-HPLC analysis of transplanted rhesus macaques
Rhesus PB RBCs or ex-vivo-differentiated progenitor cells were washed with phosphate-buffered saline (Corning Cellgro) three times. After lysing RBCs in 25-100 mL HPLC-grade water, cells were centrifuged at 16,000 Â g for 10 min. The supernatant was added to 2.5-10 mL of 100 mmol/L TCEP (tris(2-carboxyethyl)phosphine; Thermo Fisher Scientific) and incubated for 5 min at room temperature. Then, the mixture was added to 22.5-85 mL of 0.1% trifluoroacetic acid/32% acetonitrile, and the solution was briefly pulse vortexed three times. The reduced solution (10-40 mL) was analyzed at a 0.7 mL/min flow rate for 58 min using the Agilent 1100 HPLC (Agilent Technologies) equipped with an RP column (Aeris 3.6-mm Widepore C4 200-Å, 250 Â 4.6 mm, Phenomenex), and guard cartridge (AJ0-4330, Phenomenex). Solvent A (0.12% trifluoroacetic acid (TFA) in water) and solvent B (0.08% TFA in acetonitrile) were used with a gradient for separation of globin protein of 35% solvent B, followed by changes in the percentage of solvent B as follows: 3 min at up to 41.2%, 3 min at up to 41.6%, 5 min at up to 42%, 4 min at up to 42.4%, 6 min at up to 42.8%, 6 min at up to 44.4%, 6 min at up to 47%, 14 min at up to 100%, and re-equilibration for 11 min at 35%. The globin-chain peaks were detected at 215 nm and confirmed by an Agilent HPLC-6224 mass spectrometer equipped with an electrospray ionization (ESI) interface and a time-of-flight mass detector (Agilent Technologies) as described previously. 38

DATA AVAILABILITY
RNA-seq data were deposited in the Gene Expression Omnibus with accession number GSE221182. For additional original data requests, please contact the corresponding author.