Identification of small molecule agonists of fetal hemoglobin expression for the treatment of sickle cell disease

Abstract

Induction of fetal hemoglobin (HbF) has been shown to be a viable therapeutic approach to treating sickle cell disease and potentially other β-hemoglobinopathies. To identify targets and target-modulating small molecules that enhance HbF expression, we engineered a human umbilical-derived erythroid progenitor reporter cell line (HUDEP2_HBG1_HiBiT) by genetically tagging a HiBiT peptide to the carboxyl (C)-terminus of the endogenous HBG1 gene locus, which codes for γ-globin protein, a component of HbF. Employing this reporter cell line, we performed a chemogenomic screen of approximately 5000 compounds annotated with known targets or mechanisms that have achieved clinical stage or approval by the US Food and Drug Administration (FDA). Among them, 10 compounds were confirmed for their ability to induce HbF in the HUDEP2 cell line. These include several known HbF inducers, such as pomalidomide, lenalidomide, decitabine, idoxuridine, and azacytidine, which validate the translational nature of this screening platform. We identified avadomide, autophinib, triciribine, and R574 as novel HbF inducers from these screens. We orthogonally confirmed HbF induction activities of the top hits in both parental HUDEP2 cells as well as in human primary CD34+ hematopoietic stem and progenitor cells (HSPCs). Further, we demonstrated that pomalidomide and avadomide, but not idoxuridine, induced HbF expression through downregulation of several transcriptional repressors such as BCL11A, ZBTB7A, and IKZF1. These studies demonstrate a robust phenotypic screening workflow that can be applied to large-scale small molecule profiling campaigns for the discovery of targets and pathways, as well as novel therapeutics for sickle cell disease and other β-hemoglobinopathies.

Fig 1. Engineering of HUDEP2_HBG1_HiBiT reporter cell line.

(A) Schematic representation of HiBiT protein complementation assay. Cells expressing a fusion protein of target and small HiBiT reconstituted with large subunit LgBiT to form a NanoLuc complex, which generates a luminescent signal in the presence of added furimazine substrate. (B) Schematic representation of insertion of HiBiT tag to the C-terminus of the HBG1 gene within the γ-globin locus. The stop codon of HBG1 gene is in red. The gRNA sequence and corresponding PAM sequence are indicated. (C) The sequence of gRNA that is designed to selectively target the HBG1 gene. Alignment of C-terminal portions of HBG1 and HBG2 genes. The sequence fragment with a single nucleotide difference between HBG1 and HBG2 genes is indicted by the asterisk, which overlap with the gRNA target sequence.

Fig 2. Validation of HUDEP2_HBG1_HiBiT reporter cell line.

(A) Representative FACS results of HUDEP2_HBG1_HiBiT reporter cells treated with increasing concentrations of pomalidomide in differentiation media for 6 days. Gated populations of HbF positive cells and dead cells are indicated. (B) FACS data from (A) plotted as fold changes of percentage of HbF positive cells compared to DMSO as the negative control (n = 2). (C) Fold changes in HiBiT luminescence signals of HUDEP2_HBG1_HiBiT reporter cells treated with increasing concentrations of pomalidomide compared to DMSO as the negative control (n = 3). (D) Correlation of fold changes of HiBiT luminescence signals detected by Nano-Glo HiBiT lytic detection reagent and percentages of HbF positive cells detected by flow cytometric analysis. (E) Quantification of α, β, γ-globin proteins and HiBiT-tagged protein expression levels by western blot from S1 Fig. Fold changes of protein expression levels were calculated following normalization to GAPDH and then relative to DMSO treated samples. (F) Correlation of fold changes of HiBiT luminescence signals detected by Nano-Glo HiBiT Lytic Detection reagent and γ-globin protein expression levels detected by western blot analysis.

Fig 3. Assay optimization for high-throughput screening.

Pomalidomide-treated HUDEP2_HBG1_HiBiT cells were harvested for HiBiT luminescence signal detection after 4 (A), 5 (B, D) or 6 (C) days of incubation. Fold changes of HiBiT luminescence signals compared to DMSO treatment samples are graphed. Average values of thirty-two samples for each treatment are presented. Z’ factors were calculated using GraphPad Prism 8 software.

Fig 4. High throughput chemogenomic screen for HbF inducers.

(A) Schematic diagram of HUDEP2_HBG1_HiBiT chemogenomic screen for identification of compounds and targets that up-regulate fetal hemoglobin gene expression. (B) Compound activity distribution and hit calling strategy. Primary hits were selected based on HiBiT luminescence signal +3xSTDEV above the average HiBiT luminescence signals of DMSO treated samples (dotted line). (C) HUDEP2_HBG1_HiBiT cells were treated with the indicated compounds at final concentrations of 0, 0.37, 1.1, 3.3 μM, and 10 μM for 5 days. HiBiT luminescence signals and relative cell viability were detected by Nano-Glo HiBiT Lytic Detection reagent and CellTiter Glo reagent, respectively. The data were normalized to DMSO controls for fold changes of HiBiT luminescence signals, and as 100% for the relative cell viability.

Fig 5. Validation of the HbF inducing drugs in human primary CD34+ cells.

Human peripheral blood CD34+ cells treated with lead hit compounds and profiled by FACS for HbF expression. (A) Representative images of FACS analyses are shown. The percentages of fetal globin positive cells are indicated. (B) Percentages of HbF positive cells for each treatment are graphed.

Fig 6. Investigation of molecular targets of lead hit compounds.

Parental HUDEP2 cells were treated with the indicated compounds for 6 days. (A) Induction of HbF positive cells detected by FACS analysis. The average values and standard deviation of percentages of HbF positive cells from three replicates are graphed. (B) Western blot images of BCL11A, IKZF1, LRF/ZBTB7A, CRBN, and GAPDH proteins expression in DMSO- and compound-treated HUDEP2 cells. GAPDH was used as the loading control. (C) Quantification of BCL11A, IKZF1, LRF/ZBTB7A, and CRBN proteins levels from (B). Fold changes of protein expression levels were normalized with GAPDH controls and then fold change values determined relative to DMSO treated samples.