High-Resolution CRISPR Functional Genomics Reveals Critical Vulnerabilities of the NuRD Complex for Fetal Hemoglobin Control

Hemoglobinopathies resulting from mutations of the adult beta-globin gene affect millions worldwide. Persistently elevated levels of fetal hemoglobin (HbF) are without consequence in healthy individuals but convey a major clinical benefit in patients with sickle cell disease and beta-thalassemia. Fetal globin silencing relies on interactions between DNA binding factors and chromatin readers, writers, and erasers acting within multiprotein nuclear complexes. The nucleosome remodeling and deacetylase (NuRD) complex is an epigenetic chromatin modifier, including ATP-dependent nucleosome remodeling and histone deacetylase activities, that has been implicated in HbF control. BCL11A and ZBTB7A, two key transcription factors required for HbF repression in adult erythroid cells, physically interact with NuRD complex members. In order to identify potential target sites within the NuRD complex with maximal HbF de-repression and minimal cellular toxicity potential we set out to dissect the function of the NuRD complex in adult erythroid precursors by dense mutagenesis. We performed CRISPR-Cas9 mediated saturating mutagenesis of all coding regions of human NuRD complex members in human umbilical cord blood-derived erythroid progenitor (HUDEP-2) adult erythroid cells with HbF expression as the primary readout and cell fitness as a counter-screen. This approach distinguished four classes of NuRD complex members: those required for HbF repression but dispensable for cellular fitness (such as MTA2, MBD2, HDAC2, and GATAD2A), those required for both HbF repression and cell fitness (such as CHD4), those only required for cell fitness (such as RBBP4), and many not essential for either HbF repression or cell fitness. The NuRD paralogs essential for HbF repression correlated with the abundance of peptides retrieved by MTA2 immunoprecipitation followed by mass spectrometry, suggesting a NuRD sub-complex that silences fetal globin. We demonstrated that this tiling CRISPR screen approach identifies residues at which short in-frame deletions result in loss of function, including at catalytic residues of CHD4 and HDAC2, additional NuRD domains and conserved regions of unknown significance. By mapping functional data onto existing structures we detected sequences essential for HbF repression within the ELM2 and BAH domains of MTA2 that mediate key protein-protein interactions with HDAC1/2 and with chromatin. At CHD4, we found numerous critical positions where in-frame alleles resulted in robust HbF elevation to a greater magnitude than disruption of any other NuRD member. However nearly all of these HbF-inducing perturbations at CHD4 were inversely correlated with cell fitness. In contrast, we observed that in-frame deletions within the CHDCT2 domain of CHD4 resulted in profound elevation of HbF level but spared the negative impact on cell growth. Overall, these results demonstrate a sub-complex of NuRD essential for HbF silencing and identify key interfaces including at the potent HbF repressor CHD4 which could serve as selective small molecule development targets for therapeutic HbF re-induction.