IntroductionIn 1979, researchers studying the chicken globin locus noted that methylation of cytosines within CpG dinucleotides of gene promoter DNA correlated with transcriptional silencing1. This association was confirmed in humans, where the global demethylating drug 5-azacytidine was found to reverse silencing of the γ-globin genes HBG1 and HBG2 (hereafter referred to as HBG) in humans and non-human primates, thereby strengthening the correlation between CpG methylation and gene repression2,3,4. Since then, numerous studies have aimed to understand the correlation between CpG methylation and HBG silencing, as reversing repression of HBG to induce fetal hemoglobin (HbF, α2γ2) production is an established therapeutic strategy for β-hemoglobinopathies caused by mutations in the adult-expressed β-globin gene (HBB)5,6,7,8. However, it has been difficult to demonstrate definitively whether the association between global CpG methylation and HBG gene silencing in humans is causal or correlative. Indeed, a major role for CpG methylation in HBG silencing has recently been brought into question9.In a forward genetic screen, we identified Ubiquitin-like with PHD and Ring Finger Domains 1 (UHRF1) as a mediator of HBG gene silencing. The UHRF1 protein facilitates maintenance of CpG methylation during DNA synthesis by recruiting DNA methyltransferase 1 (DNMT1) to hemi-methylated CpG sites at replication forks10,11. Loss of UHRF1 in adult-type HBB-expressing erythroid cells caused global CpG demethylation and HBG gene activation, while local re-methylation at the HBG promoters via epigenomic editing restored silencing. Conversely, HBG was activated by targeted promoter demethylation in the adult-type erythroid cell line HUDEP2 and in primary erythroblasts derived from CD34+ hematopoietic stem and progenitor cells (HSPCs). Taken together, our results demonstrate that CpG methylation at the HBG promoters causes reversible gene silencing.ResultsMethylation maintenance factor UHRF1 is a regulator of HbFTo identify regulatory mechanisms controlling HBG expression, we performed a CRISPR/Cas9 screen in the adult-type erythroid cell line HUDEP2, which primarily expresses HBB12. HUDEP2 cells expressing Cas9 were transduced with a lentiviral vector library encoding 3143 single guide RNAs (sgRNAs) targeting 776 genes encoding components of the ubiquitin proteasome system, fractionated by immune-flow cytometry according to HbF expression, and analyzed for sgRNA content by next-generation sequencing (NGS)13. sgRNAs that were enriched in HbFhigh cells targeted genes encoding known repressors of HBG transcription (ZBTB7A, VHL, BCL11A) and the BCL11A protein binding motif in the HBG promoters, which is mutated in some individuals with Hereditary Persistence of Fetal Hemoglobin (HPFH) (Fig. 1a). In addition to these controls, there was enrichment of sgRNAs targeting UHRF1, a DNMT1 cofactor in the CpG methylation maintenance pathway10,14. Lentiviral transduction of two individual UHRF1 sgRNAs into Cas9-expressing HUDEP2 cells raised HBG mRNA and HbF protein levels, validating the initial screening results (Fig. 1b, c). Disruption of UHRF1 in CD34+ HSPCs followed by in vitro differentiation raised HBG and HbF in erythroid progeny to nearly the same levels as disruption of the BCL11A erythroid enhancer according to a clinically approved strategy15 (Fig. 1d-f and Supplementary Fig. 1a). Consistent with its fundamental role as a DNMT1 cofactor, UHRF1 depletion caused global CpG demethylation, which accompanied the induction of HBG and several other genes (Fig. 1g, h). However, there were no changes in mRNAs encoding known regulators of HBG transcription, including BCL11A, ZBTB7A, or MYB in UHRF1-disrupted cells (Fig. 1h).Fig. 1: A CRISPR/Cas9 screen identifies UHRF1 as a repressor of fetal hemoglobin (HbF) production.a Single guide RNAs (sgRNAs) that were enriched in HUDEP2 cells expressing high levels of HbF (see Methods for details). The y-axis shows the log2 fold-change (FC) in sgRNAs comparing HbFHigh to HbFLow cells. Dots represent the average for four sgRNAs targeting the same gene. Positive controls for repressors of HbF, including BCL11A, ZBTB7A, VHL, and a BCL11A repressor binding element in the HBG (γ-globin) promoter (labeled HFPH) are shown. UHRF1 is indicated as an orange point. b Cas9-expressing HUDEP2 cells were transduced with lentiviral vector encoding one of two different UHRF1 sgRNAs (g1, g2) or non-targeting (NT) sgRNA control, induced to undergo erythroid maturation, and analyzed after 5 days. Graph shows HBG (HBG2/HBG1) mRNA as a fraction of total mRNA (HBG/(HBG + HBB)) determined by quantitative real-time qPCR (Data are presented as mean ± s.d. of three independent experiments). c %HbF protein (HbF/(HbF + HbA)) determined by ion exchange high-performance liquid chromatography (HPLC) analysis in two independent experiments described in panel b. Bars represent mean values. d-f Normal human peripheral blood CD34+ cells from three independent donors were modified by electroporation of ribonucleoprotein (RNP) consisting of Cas9 + sgRNA targeting the erythroid-specific enhancer in BCL11A intron 2 (BCL11A enh) or UHRF1 protein coding regions, cultured in maintenance medium, and analyzed after 12 days. sgRNA targeting Adeno-associated virus integration site 1 (AAVS1) was used as control. d %HBG mRNA determined by qPCR. (Graph shows the mean percentage ± s.d. of three biological replicate experiments). e HPLC analysis of cellular hemoglobin content with %HbF indicated. (Graph shows the mean percentage ± s.d. of three biological replicate experiments) f Representative flow cytometry plots showing HbF immunostaining cells (F-cells). g Density plots showing global methylation status determined by Methylation EPIC 850 K array in UHRF1-disrupted or control HUDEP2 cells. h RNA-seq analysis showing transcripts per million mapped reads (TPM) in UHRF1-disrupted or control HUDEP2 cells. Each dot represents an individual gene (average of three technical replicate experiments). **P