PUM1 mediates the posttranscriptional regulation of human fetal hemoglobin

β-globin switching from fetal γ-globin (HBG1 and HBG2) to adult β-globin (HBB) is a developmental process that occurs at around the time of birth in erythrocytes, and the mechanisms that allow γ-globin re-expression in adult erythrocytes are exploited as an effective therapeutic strategy to ameliorate sickle cell anemia and β-thalassemia.¹ Unlike transcriptional and epigenetic regulation, posttranscriptional regulation of β-globin switching is poorly understood, with few reports on its physiological and clinical relevance.^2-9 Here, we report that Pumilo 1 (PUM1), an RNA-binding protein with no previously reported functions in erythropoiesis, is a novel target of the erythroid master transcription factor erythroid Krüppel-like factor (EKLF/KLF1) and has a role in the posttranscriptional regulation of fetal hemoglobin (HbF).

Human blood samples were collected after approval by the Institutional Review Board at Cleveland State University. Eklf^+/+ and Eklf^−/− murine erythroid cells, human umbilical cord blood–derived erythroid progenitor 2 (HUDEP2) cells, and the K562 cell line were grown and differentiated as described previously.^10,11 Primary adult human CD34⁺ cells (hematopoietic stem and progenitor cells [HSPCs]) from mobilized peripheral blood collected via apheresis from healthy adult donors were purchased from Yale Cooperative Center of Excellence in Hematology and were expanded and differentiated as described before.¹² 

Click-iT Nascent RNA Capture Kit (Thermo Fisher Scientific) was used according to the manufacturer’s protocol to assess transcript stability. Polyribosome fractionation in HUDEP2 cells was performed as described previously.¹³ RNA immunoprecipitation (RIP) was performed using RIPAb+ PUM1 polyclonal antibody and primer set (Millipore Sigma) and Magna-RIP RNA-Binding Protein Immunoprecipitation Kit (Millipore Sigma) according to the manufacturer’s instructions. Detailed protocols for the above techniques and for lentiviral transductions to knock down and overexpress PUM1, quantitative reverse transcriptase polymerase chain reaction, western blotting, flow sorting, F-cell staining, Hb high-pressure liquid chromatography (HPLC), and statistics, along with the list of antibodies and primers are available in the supplemental Materials and Methods.

We previously created an ex vivo primary cell culture system to expand our understanding of how EKLF mediates the precise changes leading to terminal erythropoiesis and enucleation.¹¹ RNA sequencing analysis on the Eklf^+/+ and Eklf^−/− murine erythroid cells and chromatin immunoprecipitation sequencing analysis in human erythroid cells¹⁴ identified a novel EKLF target, PUM1, which is upregulated specifically during erythroid terminal differentiation (Figure 1A-B). PUM1, a member of the Pumilio RNA-binding protein (PUF) family of sequence-specific RNA-binding proteins, acts as a posttranscriptional repressor by binding to the 3′ untranslated region (3′-UTR) of messenger RNA (mRNA) targets and impairing their stability and/or translational efficiency.¹⁵ Consistent with these functions, we observed PUM1 in the cytoplasm before and after erythroid terminal differentiation (supplemental Figure 1).

EKLF is a known regulator of Hb switching,¹⁶ so to identify the functions of PUM1 in erythroid terminal differentiation, we first tested its ability to regulate globin levels. Fetal γ-globin is expressed from 2 genes, HBG1 (^Aγ) and HBG2 (^Gγ)¹⁷; HBG1 has 2 core PUM1 consensus binding sites in its 3′-UTR, but HBG2 and other fetal and adult globins do not (supplemental Figure 2). Because HbF comprises <1% of total Hb in adult erythroid cells,¹⁸ we asked whether PUM1 suppressed γ-globin expression specifically in these cells. We tested this by knocking down PUM1 in HUDEP2 immortalized human erythroid progenitor cells that exhibit an adult Hb profile. We observed a modest increase (∼2.5-fold) in levels of the γ-globin transcript (supplemental Figure 3), but the increase in the γ-globin protein levels after PUM1 knockdown was more than 12-fold by western blot analysis, and HPLC showed 22.6% HbF levels in PUM1 knockdown cells compared with no detectable HbF in control cells (Figure 1C-D). Similar results were also observed in primary erythroid cells derived from human adult HSPC CD34⁺ cells (Figure 1E-F; supplemental Figure 4A-B). Conversely, overexpression of PUM1 in K562 erythroleukemia cells that express high endogenous HbF¹⁹ led to a reduction of γ-globin (Figure 1G).

Our data underline the importance of the fine-tuned homeostasis required to maintain PUM1 protein levels, because we observed that even slight perturbations in PUM1 levels result in gross γ-globin changes, as was previously reported in patient mutations that reduced PUM1 levels by 25% in other tissues.²⁰ Importantly, PUM1 knockdown did not affect the progression of erythropoiesis either in primary adult human HSPCs or in HUDEP2 cells (Figure 1H; supplemental Figure 5A-B). Furthermore, as shown in supplemental Figure 6, knockdown of PUM1 did not lead to changes in the levels of known γ-globin regulators such as EKLF, BCL11A, and ZBTB7A.^21-23 

PUM1 regulates gene expression in mammals by degrading target mRNA and/or inhibiting translation of target mRNAs.¹⁵ Native RIP of PUM1 pulled down γ-globin mRNA compared with β-globin and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA (Figure 2A; supplemental Figure 7), thus confirming a unique and direct role for PUM1 in regulating γ-globin. We then performed a nascent mRNA degradation assay in which we pulsed HUDEP2 cells with ethylene uridine (EU) ribonucleotide homologs and then washed and analyzed the newly synthesized EU-incorporated mRNA at different time points. The EU-incorporated HBG1, HBG2, and HBB β-type globin mRNA levels were reduced over time in the control cells, but only HBG1 mRNA levels were relatively stabilized after PUM1 was knocked down. We observed this after either a short 24-hour chase or a longer 3-day chase, suggesting that PUM1 specifically mediates HBG1 mRNA degradation (Figure 2B; supplemental Figure 8A). The delay in the degradation of HBG1 upon PUM1 knockdown is further captured in a time course of EU incorporation and chase (supplemental Figure 8B). Next, polysome profiling of the control and PUM1 knocked-down HUDEP2 cells showed a specific increase in the HBG1 mRNA levels in polyribosomal fractions compared with the monosomal fractions, suggesting that HBG1 mRNA, unlike HBG2 and HBB mRNA, is more actively translated under reduced PUM1 levels (Figure 2C-D; supplemental Figure 9A-D). These results demonstrate that PUM1 regulates HBG1 at the level of mRNA stability and also at translation and thus serves as a posttranscriptional regulator of γ-globin in adult human erythroid cells.

The diverse roles of PUM1 in human pathology imply that it has distinct cell type–dependent roles during development.²⁰ Therefore, we investigated whether patient mutations in PUM1 could result in high HbF levels. We identified a 5-year-old child with PUM1-associated developmental disability, ataxia, and seizure (PADDAS) who harbored a novel heterozygous PUM1 mutation (p.(His1090Profs∗16); c.3267_3270delTCAC). The mutation, a frameshift in the RNA-binding domain, introduces 16 new amino acids and a premature stop codon (Figure 2E). We observed elevated HbF levels in the patient (above the accepted reference range), with a more than 10-fold increase over the levels in healthy parents, as analyzed by HPLC and by modified Kleihauer-Betke staining for F cells (Figure 2F-G). The complete blood count in the patient suggested that elevated HbF was not due to anemia (supplemental Figure 10).

Here we identify PUM1 (an RNA-binding protein with no previously reported roles in erythropoiesis) as a direct posttranscriptional regulator of Hb switching. The change in the ratio of ^Gγ(HBG2):^Aγ(HBG1) from 3:1 at birth to 2:3 in the small amount of HbF present in the blood of adults may be explained by the existence of exclusive regulatory mechanisms for these 2 genes (such as the PUM1-mediated targeting of HBG1 alone), although how PUM1 itself is regulated during this developmental window is unknown.²⁴ An investigation of available databases identifies an increase in PUM1 transcript levels as the site of erythropoiesis shifts from the fetal liver to bone marrow, similar to previously identified regulators of Hb switching such as ZBTB7A/LRF (supplemental Figure 11).²⁵ Investigation will be required to decipher whether, in addition to direct posttranscriptional regulation, indirect mechanisms could also contribute to the robust induction of γ-globin observed upon PUM1 knockdown in erythroid cells. Future studies will probe the potential for erythroid-specific enhancers of PUM1 expression as an EKLF target.

The HbF induction (∼22% HbF) observed after PUM1 knockdown exceeds the levels considered therapeutic in patients with sickle cell anemia,^26-29 although it is lower than the level of induction observed after modulation of targets such as BCL11A and ZBTB7A/LRF (ranging from ∼23% to 73% HbF in cultured erythroid cells^22,30). Because our studies indicate that PUM1 knockdown does not affect progression through erythroid terminal differentiation, that PUM1 knockout mice are viable with no reported erythroid defects,³¹ and that PUM1 functions as a cytoplasmic regulator, we propose that it could potentially serve as a safe and effective non–gene-altering target toward ameliorating β-thalassemia and sickle cell disease.

The authors thank James J. Bieker, PhD, for sharing reagents, participating in discussions on the project, and providing constructive comments on the manuscript; Jane A. Little, MD, for participating in discussions and helping interpret clinical data; Anchal Agarwal for helping with the polysome fractionation analysis; Sheetal Uppal, PhD, for helping to map patients’ PUM1 mutations; Karamoja Monchamp and Allison Bode for helping with the HPLC study; and Girish Shukla, PhD, for discussions related to the RNA biology and design of experiments.

This study was supported by grants from the National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases (K01DK115686) (M.N.G.), (1U2CDK129440, 1TL1DK132770) (S.E.A), and (U54DK106857), which supported the purchase of CD34⁺ cells from Yale Cooperative Center of Excellence in Hematology; from the National Heart, Lung, and Blood Institute (HL151392) (A.A.K.); and Research Fellowships from Cooley’s Anemia Foundation (M.N.G. and M.R).

Contribution: M.N.G. and M.R. conceptualized the study and wrote the manuscript; R.E., A.R.D., R.M.G., and M.R. performed quantitative reverse transcriptase polymerase chain reaction assays and western blots; R.E., A.R.D., and R.M.G. performed the differentiation assays and cell culturing; M.N.G. performed the Click-iT pulse-chase experiments and RNA immunoprecipitation; S.E.A. performed the immunofluorescence experiment; R.A.W. genotyped the mutations of the patients; S.G. and O.Y.M. ran the fluorescence-activated cell sorted (FACS) samples and helped analyze the FACS data; A.O.-A. helped interpret the deidentified patient data; A.G. performed the polysome fractionation; A.G., A.A.K., M.R., and M.N.G. analyzed the data; Y.M. and U.A.G. helped with the HPLC analysis; and M.R. and M.N.G. processed raw data and created the figures.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Merlin N. Gnanapragasam, Department of Biological, Geological and Environmental Sciences, Center for Gene Regulation in Health and Disease, Cleveland State University, 2399 Euclid Ave, SR276, Cleveland, OH 44115-2406; e-mail: m.gnanapragasam@csuohio.edu; and Mahesh Ramamoorthy, Center for Gene Regulation in Health and Disease, Cleveland State University, 2399 Euclid Ave, SR276, Cleveland, OH 44115-2406; e-mail: m.ramamoorthy@csuohio.edu.