The GATA1 N terminus coordinates metabolic reprogramming in erythropoiesis

Key Points

• •GATA1 contributes to metabolic reprogramming through its regulation of the PKM gene.
• •Enhanced glycolysis is a feature of RPS19-mutant DBA, which is associated with dysregulation of GATA1.

Visual Abstract

Abstract

Mutations in GATA1 that cause skipping of exon 2, which encodes the N terminus, are associated with the myeloid leukemia of Down syndrome and Diamond-Blackfan anemia (DBA). To elucidate the molecular function of this N-terminal region, we used single-cell RNA sequencing (scRNA-seq) on fetal liver cells from Gata1-mutant embryos that express only the short isoform of GATA1 (GATA1s) lacking the N terminus of full-length GATA1 (GATA1FL). scRNA-seq revealed defects in erythropoiesis and aberrant upregulation of glycolytic genes, including PKM, which encodes pyruvate kinase to catalyze the final and irreversible step of glycolysis. Using precision nuclear run-on sequencing and cleavage under targets and release using nuclease (CUT&RUN) after acute GATA1 deletion in erythroid cells, we identified PKM as a direct target of GATA1. Substitution of GATA1FL with GATA1s induced histone lactylation at the PKM promoter, increased pyruvate kinase M (PKM) expression and activity, and enhanced glycolytic flux in erythroid progenitors, without affecting mitochondrial respiration. Importantly, PKM expression is also significantly elevated in patients with DBA with RPS19 mutations, which is associated with reduced levels of GATA1, further supporting a link between GATA1s-driven defective erythropoiesis and dysregulated glycolysis. Together, these findings reveal that GATA1 controls not only heme metabolism but also glycolytic reprogramming.

The zinc finger transcription factor GATA1 is a master regulator of erythropoiesis, coordinating the differentiation program. In some conditions such as Diamond-Blackfan anemia and associated with some mutations in Down syndrome, a short form of GATA1 is dominant. Ling and colleagues define the function of this isoform, demonstrating in primary erythroid cells how it can reprogram metabolism through its regulation of PKM, the gene encoding pyruvate kinase M1/2.

Introduction

The zinc finger transcription factor GATA1 is expressed in several hematopoietic lineages and is indispensable for their development, particularly erythropoiesis and megakaryopoiesis. Studies have demonstrated the embryonic lethality of complete loss of GATA1, with Gata1-null mice dying from severe anemia between embryonic days 10.5 and 11.5 (E10.5-E11.5).¹ GATA1 is essential for both primitive (embryonic) and definitive (adult) erythropoiesis. Erythroid progenitors that lack GATA1 undergo apoptosis and fail to generate mature erythrocytes.² Of note, inherited mutations in GATA1 are associated with several related red blood cell disorders. These disorders include dyserythropoietic anemia and thrombocytopenia, which are caused by mutations in the first zinc finger domain of GATA1.^3,4 Diamond-Blackfan anemia (DBA), a specific type of dyserythropoietic anemia, is caused by mutations that lead to the expression of only the short isoform of GATA1 (GATA1s), which lacks the N-terminal 83 amino acids.⁵ Additionally, hemolytic anemia is caused by missense mutations in an intrinsically disordered region of GATA1.⁶ Although the mechanism by which GATA1 zinc finger mutations cause anemia is well understood, the requirement of the GATA1 N terminus, which is absent in the shorter GATA1s isoform, remains elusive.

Although the role of GATA1 in heme metabolism has been well established, its contributions to other metabolic processes required for erythroid cell function, including energy metabolism, are less well understood. During erythropoiesis, mitochondria biogenesis and metabolism are precisely regulated, particularly the dynamic shift from glycolysis to oxidative phosphorylation (OXPHOS), followed by a return to glycolysis during late stages of maturation.^{7, 8, 9} As hematopoietic progenitor cells commit to the megakaryocyte-erythroid progenitor stage, they show increased OXPHOS and mitochondrial activity. Early erythroid progenitors (ProE), which require substantial energy for rapid expansion, show a mix of glycolysis and mitochondrial metabolism. As differentiation proceeds, cells gradually shut down mitochondrial metabolism and rely on glycolysis, and the pentose phosphate pathway becomes crucial for maintaining redox balance.¹⁰ Erythroid disorders associated with dysregulation of these 2 metabolic shifts, such as hemolytic anemia, are caused by mutations in pyruvate kinase L/R (PKLR), which disrupt ATP generation (PK deficiency), or glucose-6-phosphate dehydrogenase mutations affect red blood cell metabolism, making cells more susceptible to oxidative stress (glucose-6-phosphate dehydrogenase deficiency).⁷ However, the contribution of GATA1 to regulation of glycolysis and its potential contribution to pathogenesis of anemia in the presence of GATA1 mutations remain unexplored.

Methods

Animals

Mice expressing only GATA1s (refer to Gata1s) were provided by Stuart Orkin (Children’s Hospital, Boston, MA). Genotyping was performed as previously described.¹¹

Cell culture and differentiation

HUDEP1 (human umbilical cord blood-derived erythroid progenitor 1) and HUDEP2 cells were kindly provided by Yukio Nakamura (RIKEN BioResource Research Center, Japan). Cells were maintained in expansion medium consisting of SFEM II supplemented with 3 U/mL erythropoietin (EPO), 50 ng/mL human stem cell factor (hSCF), 1μM dexamethasone, and 1 μg/mL doxycycline, at a density of <0.5 × 10⁶ cells per mL. For differentiation, cells were cultured in Iscove modified Dulbecco medium supplemented with 2% fetal bovine serum, 3% human AB serum, 3 U/mL EPO, 10 μg/mL human insulin, 3 U/mL heparin, 1 mg/mL human holo-transferrin, 50 ng/mL hSCF, and 1 μg/mL doxycycline. Cell identity was authenticated using the PowerPlex Fusion System (Promega) performed at the Hartwell Center (St. Jude Children's Research Hospital) and cultures were negative for mycoplasma as assessed by the MycoStrip mycoplasma detection kit (InvivoGen; catalog no. rep-mys-50).

Generation of a degradable system for FKBP12F36V-GATA1 expression

HUDEP2 GATA1 FKBP-V tagged cells were generated using CRISPR-CRISPR–associated protein 9, as previously described.¹² Clones were screened for the desired modification via polymerase chain reaction–based assays and confirmed with sequencing. Editing construct sequences and relevant primers are listed in supplemental Table 1, available on the Blood website.

Seahorse glycolysis stress test

The glycolysis stress test was performed to assess the extracellular acidification rate using the XFe96 extracellular flux analyzer (Agilent Technologies), following the manufacturer’s protocol provided with the Seahorse XF glycolysis stress test kit (Agilent; catalog no. 103020-100). Briefly, HUDEP1 cell lines expressing either GATA1 full-length (GATA1FL) or GATA1s (G1s) were cultured for 24 hours in differentiation medium. After 24 hours, cells were harvested and washed twice with prewarmed Seahorse XF RPMI 1640 medium, no phenol red (pH 7.4; Agilent, catalog no. 103576-100) supplemented with 2mM l-glutamine (Agilent, catalog no. 103579-100). Cells were then seeded onto Cell-Tak (Corning, catalog no. 354240) coated Seahorse XFe96 cell culture microplates at a density of 1.0 × 10⁵ cells per well, centrifuged at 100g for 1 minute, and incubated in a non-CO₂, 37°C humidified incubator for 1 hour. After incubation, cells were subjected to sequential injections of glucose (10mM), oligomycin (1μM), and 2-deoxyglucose (50mM), and extracellular acidification rate was monitored for 2 hours. Sorted CD71^hiCD235a⁻ erythroid progenitors derived from human CD34⁺ cells were cultured overnight before the glycolysis stress test to eliminate the influence of sorting stress. After the assay, cell counts, and normalization were performed using Gen5 software (Agilent, BioTek Cytation5) with the high-contrast brightfield module. Key glycolytic parameters were calculated using the Seahorse XF real-time glycolysis stress test report generator and analyzed in GraphPad Prism (GraphPad, San Diego, CA).

CRISPR/CRISPR–associated protein 9 gene editing and erythroid differentiation of human CD34⁺ cells

Healthy human CD34⁺ cells were isolated from fresh granulocyte colony-stimulating factor human mobilized Leukopak, purchased from CGT Global. These cells were cultured in recovery media (x-VIVO 10, 100 ng/mL hSCF, 100 ng/mL human fetal liver kinase 3 ligand), and 100 ng/mL human thrombopoietin overnight before ribonucleoprotein electroporation delivery,¹³ as previously outlined. Three days after gene editing, the cells underwent 2-phase differentiation as previously described.¹³ Cells from 5-day differentiation were collected for flow cytometry sorting. The assays were conducted after overnight recovery from sorting in the same medium. Media, guide RNA sequences, and cytokines used are detailed in supplemental Table 1.

Isotope tracing and gas chromatography–mass spectrometry (GC-MS) analysis

HUDEP1 cells expressing GATA1FL (clone GATA1FL1) and GATA1s (clone G1s_1D5) were cultured in differentiation medium for 24 hours before the tracing experiment. For the [U-¹³C]glucose tracing, 1 million cells were collected and washed once with prewarmed phosphate-buffered saline, then resuspended in tracing medium (Dulbecco modified Eagle medium supplemented with 5% dialyzed fetal bovine serum, 3 U/mL EPO, 10 μg/mL human insulin, 3 U/mL heparin, 1 mg/mL human holo-transferrin, 50 ng/mL hSCF, 1 μg/mL doxycycline, 2mM glutamine, and 10mM [U-¹³C]glucose). Cells were then incubated in a tissue culture incubator for 3 hours. After incubation, cells were collected and washed twice with ice-cold saline, without disturbing the pellets. The washed cell pellets were resuspended in prechilled (−20°C) 80% methanol and subjected to 3 freeze-thaw cycles using liquid nitrogen and a 37°C water bath. Metabolites were extracted after the freeze-thaw cycles and centrifugation at 10 000g for 10 minutes. The supernatants were evaporated to dryness and derivatized by methoximation and tert-butyldimethylsilyl derivatization. The derivatized samples were then analyzed by GC-MS to determine isotopolog distributions. Metabolite data acquisition was performed on an Agilent 8890 gas chromatograph coupled to an Agilent 5977C mass selective detector, with data collected using Agilent MassHunter software. Samples were separated on an Agilent J&W HP-5MS UI capillary column (30.0 m × 0.25 mm, 0.25 μm film thickness) with helium as the carrier gas. Chromatographic peaks were integrated using El-MAVEN (Elucidata), and mass isotopolog distributions were corrected for the natural abundance of ¹³C with IsoCor (version 2.2.2).

Additional methods are provided in the supplemental Methods.

Animal research was approved by the St. Jude Children’s Research Hospital institutional animal care and use committee.

Results

Single-cell RNA sequencing (scRNA-seq) highlights defects in erythropoiesis in Gata1s mutant embryos

We, and others, have shown that the N terminus of GATA1 is essential for normal development of the erythro-megakaryocytic lineages.^14,15 Mechanistic studies have revealed that the absence of the N terminus does not abolish the ability of GATA1s to bind chromatin but rather leads to the inability to recruit cofactors that mediate chromatin modifications.¹⁴ To investigate the function of the N terminus of GATA1 in hematopoiesis, we conducted 10X genomics single-cell transcriptome analysis on mouse fetal liver cells that exclusively expressed GATA1s¹⁵ from mice at E13.5 (supplemental Figure 1A). We obtained 84 120 high-quality cells from 3 wild-type embryos (32 858 cells) and 4 Gata1s embryos (51 262 cells) for subsequent analysis. From total fetal liver cells, we isolated erythrocytes at all stages of maturation, including the megakaryocyte-erythroid progenitor, erythroid progenitor (EP), proerythroblast (ProE), basophilic erythroblast (BasoE), polychromatic erythroblast (PolyE), and orthochromatic erythroblast (OrthoE). Other hematopoietic stem and progenitor cells (HSPC) and megakaryocytes (MK) were also detected (Figure 1A-B; supplemental Figure 1B). RNA velocity analysis revealed 3 distinct trajectories in Gata1s and wild-type control embryos, corresponding to the erythroid, megakaryocyte, and granulocyte/monocyte lineages (Figure 1C). Upon comparing the distribution of cells between Gata1s and controls, 2 distinct differences are obvious: a shift of erythrocytes toward an immature state, and an expansion of the megakaryocyte population (Figure 1D). Orthochromatic erythroblasts were reduced by 72% in Gata1s embryos, representing just 15.5% of the cells in Gata1s embryos compared with 55% in wild-type controls (Figure 1E-F). This reduction was accompanied by the accumulation of all upstream erythroblast populations. By contrast, the megakaryocyte population showed a 14.6-fold expansion in Gata1s embryos (Figure 1E). We also observed downregulation of GATA1 target genes in all populations of GATA1s erythroid cells (supplemental Figure 1C).

Figure 1.

Erythropoiesis is disrupted in Gata1s mutant embryos. (A) UMAP plot showing cell-type distributions from male E13.5 fetal liver based on scRNA-seq data. Colors denote distinct cell types. (B) UMAP plots displaying expression levels of representative marker genes for various lineages: HSPC (Kit, Runx1, Gata2), megakaryocyte (Itga2b), erythroid (Tfrc, Gypa, Bpgm, Hba-x), and myeloid (Itgam). (C) RNA velocity analysis (Velocyto) overlaid on UMAP of fetal liver cells from WT and Gata1s (G1s) embryos, showing inferred differentiation trajectories across hematopoietic and erythroid stages (HSPC > MEP > EP > ProE > BasoE > PolyE > OrthoE). (D) Density UMAP plots visualizing the distribution of cells in WT and G1s fetal livers, highlighting differences in local cell density. (E) Pie charts quantifying the relative proportion of each cell type in WT and G1s fetal livers. (F) Bar graph comparing the proportions of erythroid populations between Gata1s (G1s) and wild-type (WT) cells. (G) Heat map showing NES from GSEA of selected pathways enriched across erythroid differentiation stages (HSPC > MEP > EP > ProE > BasoE > PolyE > OrthoE). Erythroid lineage development (blue arrows), Myc-aging-related (orange arrows), and glycolysis-related (red arrows) pathways are highlighted. Asterisks (∗) denote FDR < 0.05. GMP, granulocyte-monocyte progenitor; Gran, granulocyte; HSPC, hematopoietic stem and progenitor cells; LEC, liver endothelial cell; MEP, megakaryocyte-erythroid progenitor; MK, megakaryocyte; Mono, monocyte; OrthoE, orthochromatic erythroblast; PolyE, polychromatic erythroblast; UMAP, uniform manifold approximation and projection; WT, wild-type; NES, normalized enrichment scores.

Glycolytic gene expression is dysregulated in Gata1s erythroblasts

We performed gene set enrichment analysis on the scRNA-seq data to characterize pathways that are affected by the loss of the GATA1 N terminus in various erythroid populations. We observed that erythroid lineage development-related and heme metabolism-related pathways were negatively enriched in Gata1s erythroblasts, as anticipated (Figure 1G). Additionally, we detected the enrichment of erythroblast/EP aging-related pathways, which was accompanied by the dysregulation of MYC targets that have been previously associated with erythropoiesis.^{16, 17, 18} Notably, glycolysis-related pathways, including pyruvate metabolism, glucose metabolism, glucose-6-phosphate, and fructose-6-phosphate pathways, exhibited significant alterations in Gata1s erythroblasts (Figure 1G; supplemental Figure 1D). Furthermore, the Hallmark_Glycolysis was significantly enriched in ProE and BasoE erythroblasts (supplemental Figure 1E-F).

Overexpression of GATA1s rescues survival of EPs but is unable to support terminal differentiation

Next, we leveraged the human EP cell line, HUDEP1 cells,¹⁹ to probe the relationship between GATA1 and energy metabolism. CRISPR knockout of either the FL isoform (GATA1FL), mediated by targeting the first translation start codon (Met 1), or of both FL and GATA1s by targeting the second translational start codon, Met 84, revealed that HUDEP1 cells cannot survive without GATA1FL under expansion conditions (Figure 2A). Given a previous study that showed that overexpression of GATA1s could rescue anemia in Gata1-deficient mice,²⁰ we tested whether increased levels of GATA1s would allow GATA1-targeted HUDEP1 cells to expand. Indeed, we found that overexpression of GATA1s could rescue the survival of GATA1FL-null HUDEP1 cells under expansion conditions. These cells were generated by CRISPR-mediated deletion of the entire exon 2 or introduction of a splicing mutation in exon 2 (V74L)⁵ in cells that overexpress GATA1s (Figure 2B). Of note, these cells exhibited slower growth than parental HUDEP1 cells (Figure 2C). Furthermore, HUDEP1 cells expressing only GATA1s died during differentiation and failed to undergo hemoglobinization (Figure 2D). By contrast, HUDEP1 cells engineered to express only GATA1FL, achieved by CRISPR substitution of the second start codon (Met 84I), exhibited normal growth and differentiation (Figure 2C-D), indicating that expression of GATA1s is not essential.

Figure 2.

Overexpression of GATA1s partially compensates for loss of full-length GATA1 in erythroid cell expansion and enhances glycolysis. (A) Cell fitness/dependency assays were conducted following electroporation of ribonucleoprotein complexes targeting either the first start codon (Met1) or the alternative start codon (Met84) under expansion conditions. Disruption of Met1 allows translation from Met84, resulting in exclusive GATA1s expression, whereas disruption of Met84 prevents production of any GATA1 protein (top). Editing efficiency was assessed by targeted deep sequencing up to 21 days after electroporation (bottom). (B) Western blot analysis confirming GATA1 isoform expression in parental HUDEP1 cells, and in single-cell-derived clones expressing either only full-length GATA1 (GATA1FL) or only the short isoform (GATA1s, G1s). Expression of only full-length GATA1 was achieved by introducing an M84I substitution using CRISPR-based genome editing (clones GATA1FL#1 and GATA1FL#2). Two CRISPR-based strategies were used to eliminate GATA1FL expression: deletion of exon 2 (G1s_1D clones and G1s_2D clones) and knockin of the splicing mutation V74L (1KI9 clone). (C) Proliferation of HUDEP1 parental cells and isoform-specific clones was monitored under expansion conditions using IncuCyte live-cell imaging. (D) Two representative HUDEP1 clones expressing only GATA1s (1D5 and 1KI9), along with parental HUDEP1 and GATA1FL1 clone, were subjected to erythroid differentiation. Representative cytospin images (250 rpm, 5 minutes, room temperature) followed by May-Grünwald Giemsa staining after 4 days of differentiation. Scale bar, 25 μm. The bottom image depicts cell pellets from the same differentiation experiment. (E) Seahorse XF Pro glycolysis stress test performed on 2 representative HUDEP1 clones expressing only GATA1s (1D5 and 1KI9), alongside parental HUDEP1 and GATA1FL1 clone, following 24 hours of erythroid differentiation. Traces show extracellular acidification rate (ECAR) in response to sequential injection of 20mM glucose, 1μM oligomycin, and 100mM 2-deoxy-d-glucose (2-DG). (F) Quantification of glycolytic parameters from panel E, including glycolysis, glycolytic capacity, and glycolytic reserve. Data are presented as bar graphs. (G) Seahorse XF Pro mitochondrial stress test performed on the same 24 hour differentiated HUDEP1 clones to assess glucose oxidation. Oxygen consumption rate (OCR) was measured in response to sequential injection of 20mM glucose, 1μM oligomycin, 1μM FCCP [carbonyl cyanide-4-(trifluoromethoxy)phenylhydrazone], 1μM rotenone, and 2μM antimycin A. ∗∗P < .001; ∗∗∗∗P < .0001. 2-DG, 2-deoxy-d-glucose; ns, no statistical significance; OCR, oxygen consumption rate.

Erythroid cells expressing GATA1s exhibit hyperactive glycolysis under differentiation conditions

Next, we assessed the effect of loss of GATA1FL on energy metabolism in HUDEP1 cells under expansion and differentiation conditions. Seahorse glycolysis stress tests revealed a marked increase in glycolytic activity in HUDEP1 clones expressing GATA1s under differentiation conditions, whereas that of HUDEP1 cells expressing GATA1FL were similar to that of wild-type HUDEP1 cells (Figure 2E). Specifically, HUDEP1 clones expressing only GATA1s (1KI9 and 1D5) exhibited elevated glycolysis, glycolytic capacity, and glycolytic reserve under differentiation conditions (Figure 2F), whereas these changes were not observed during expansion (supplemental Figure 2A-B). Furthermore, mitochondrial respiration and ATP production showed no significant differences under either condition (Figure 2G; supplemental Figure 2C).

PKM2 is upregulated in erythroblasts expressing only GATA1s

Next, we performed bulk mRNA sequencing on HUDEP1 clones expressing only GATA1FL or only GATA1s. Results from gene set enrichment analysis largely recapitulated the results from the scRNA-seq analysis of erythroblasts in that Hallmark_Glycolysis pathway was positively enriched in differentiated HUDEP1 cells expressing GATA1s, but not the undifferentiated cells (Figure 3A; supplemental Figure 3A). By analyzing the genes involved in glycolysis, we observed that many of them exhibited significant changes in GATA1s HUDEP1 cells (Figure 3B). This finding aligns with the observations made from mouse scRNA-seq data (supplemental Figure 3B-E). The differentiation conditions had a more pronounced effect on the HUDEP1 cells expressing GATA1s. Notably, the expression of the other PK-encoding gene, PKLR, was decreased in HUDEP1 cells expressing only GATA1s, which is consistent with the finding that GATA1 binds and regulates PKLR expression in human erythroblasts (Figure 3B; supplemental Figure 3F).²¹ Conversely, it was mildly increased in Gata1s mice during the EP stage (8.9% increase), ProE stage (3.4% increase), and BasoE stage (18.3% increase; supplemental Figure 3B-E). During these stages, the glycolytic pathway was significantly positively enriched (Figure 1G).

Figure 3.

Aberrant PKM2 expression in erythroblasts expressing only GATA1s. (A) GSEA of bulk RNA-seq data comparing 4-day differentiated HUDEP1 GATA1s clones (1D5 and 1KI9) to the GATA1FL1 clone. The top 10 enriched pathways are shown. (B) Heat map showing differential expression of glycolytic enzymes in HUDEP1 cells expressing only GATA1s vs GATA1FL under expansion and 4-day differentiation conditions. Asterisks (∗) indicate transcripts with adjusted P value (padj) <.05 and >30% change in expression. (C) Western blot analysis confirming PKM2 expression using a PKM2-specific antibody in HUDEP1 clones expressing either GATA1FL or GATA1s after 24 hours of differentiation. Phosphorylation of PKM2 at Y105 and S37 was also assessed. As expected, PKM2pY105 was absent from the nuclear fraction, whereas PKM2pS37 was detected in both cytosolic and nuclear fractions. (D) IGV snapshots showing CUT&RUN profiles of GATA1, H3K27ac, and H3K4me3 occupancy at the PKM gene locus in 4-day differentiated HUDEP1 clones expressing either GATA1FL or GATA1s (1D5 and 1KI9). Dashed lines indicate GATA1 binding regions at the PKM locus. Cyto, cytosolic; diff, differentiation; FC, fold change; FDR, false discovery rate; GRAMD2A, GRAM domain-containing protein 2A; IGV, Integrative Genomics Viewer; PARP6, polyADP-ribose polymerase family member 6.

We found that the PKM gene, which regulates the final and irreversible step of glycolysis, was consistently upregulated in both mouse and human models. PKM generates 2 protein products, PKM1 and PKM2, depending on the usage of exon 9 or exon 10, respectively. PKM1 is a constitutively active enzyme, predominantly expressed in tissues that require high energy, such as muscle and the brain, whereas PKM2 enables allosteric regulation and dimer-to-tetramer switching, which is crucial for metabolic flexibility. Through analysis of mRNA sequencing data obtained from mouse erythroblasts (GSE130156)¹⁴ and human HUDEP1 cells, we confirmed that PKM2 is the predominant form in erythroid cells in both mice and humans (supplemental Figure 4A-B). Moreover, we observed higher PKM mRNA levels in clones expressing GATA1s, suggesting that either GATA1s activates, or GATA1FL suppresses, PKM2 expression (Figure 3B; supplemental Figure 3F). We confirmed the elevated expression of PKM2 in GATA1s HUDEP1 cells at the protein level (Figure 3C). Furthermore, the accumulation of phosphorylated PKM2 on serine 37, which has been reported previously to promote the translocation of PKM2 into the nucleus and regulate gene transcription, was also detected.^22,23 Additionally, phosphorylation of PKM2 on tyrosine 105, which has been previously reported to negatively regulate PKM2 activity by inhibiting the formation of active, tetrameric PKM2, was also increased.²⁴ Finally, through the cleavage under targets and release using nuclease (CUT&RUN) assay, we determined that GATA1 binds to the PKM gene (Figure 3D; supplemental Figure 3G). Furthermore, we detected an increase in H3K4me3 (histone H3 lysine 4 trimethylation) on the gene promoter region and accumulation of H3K27ac (histone H3 lysine 27 acetylation) at GATA1-bound regions in the absence of GATA1FL under differentiation conditions, which is consistent with the elevated gene transcription (Figure 3D).

GATA1 represses PKM expression during erythropoiesis

To further investigate the regulation of PKM, we constructed an acute GATA1 degradation model by introducing a FKBP12^F36V tag at the N terminus of GATA1 in HUDEP2 clone that expresses only GATA1FL (Figure 4A). Western blot analysis revealed that GATA1 was substantially reduced 2 to 3 hours after the addition of 100nM dTAG-13 whereas protein level was restored ∼24 hours after the washout of dTAG-13 (Figure 4B). We performed precision run-on sequencing, which provides single-nucleotide resolution for the detection of nascent transcripts, on the cells after induction of GATA1 degradation by dTAG-13 (0.5 hours, 3 hours, and 6 hours) under expansion conditions and differentiation conditions. Among the genes encoding enzymes involved in the glycolysis pathway, only PKM exhibited upregulation 3 hours after GATA1 degradation, whereas PFKFB4 exhibited downregulation 3 hours after GATA1 degradation (supplemental Figure 5A-B). Notably, PKM transcripts doubled after 3 hours of GATA1 degradation (Figures 4C-E; supplemental Figure 5C-D). This increase was further amplified at the 6-hour time point. Again, the elevation of nascent PKM transcripts is pronounced in differentiation conditions. Furthermore, we observed decreased GATA1 occupancy around the PKM gene promoter region after the addition of dTAG-13 (Figure 4F). Together, these results suggest that PKM is a direct target of GATA1, and its expression is suppressed by GATA1 during erythrocyte maturation. However, GATA1s is unable to repress PKM gene expression.

Figure 4.

GATA1 binds and directly regulates the PKM gene. (A) Schematic illustrating the strategy used to generate an FKBP12^F36V-tagged GATA1 acute degradation system in HUDEP2 cells. (B) Western blot confirming degradation of GATA1 protein following treatment with 100nM dTAG-13. Restoration of GATA1 expression was assessed after dTAG-13 washout. (C-D) Precision nuclear run-on sequencing (PRO-seq) was performed at 0.5 hours, 3 hours, and 6 hours after GATA1 degradation. IGV snapshots show nascent transcription at the PKM gene locus under both expansion (expan) and 24-hour differentiation (diff) conditions. (E) Bar graph quantifying normalized nascent transcription levels of the PKM gene from the data shown in panels C-D. Asterisks (∗) indicate statistically significant changes (padj <.05). (F) IGV snapshots of CUT&RUN data showing dynamic GATA1 occupancy at the PKM locus at 0.5 hours, 3 hours, and 6 hours following GATA1 degradation, corresponding to the PRO-seq time course in panels C-E. Diff, differentiation; DMSO, dimethyl sulfoxide; Norm., normalized.

GATA1s enhances PKM activity and promotes glycolysis during differentiation

To assess glycolytic activity, we performed stable isotope tracing using [U-¹³C]glucose in HUDEP1 cells after 24-hour differentiation. Metabolites from glycolysis and the trichloroacetic acid cycle were analyzed by GC-MS. We observed that GATA1s-expressing cells showed increased labeling of glycolytic intermediates, including 3-phosphoglycerate, phosphoenolpyruvate, lactate, and pyruvate, compared with FL control cells (Figure 5A-B). The increased ratio of ¹³C-labeled pyruvate to phosphoenolpyruvate further supports enhanced PKM2 activity in GATA1s cells (Figure 5B). Consistently, we observed higher levels of PK activity and intracellular lactate in differentiated GATA1s HUDEP1 cells (Figure 5C-D).

Figure 5.

Glycolysis is hyperactivated in erythroblasts expressing only GATA1s. (A) Schematic representation of the metabolic fate of uniformly labeled ¹³C₆-glucose through glycolysis and the TCA cycle. (B) ¹³C₆-glucose tracing of glycolytic and TCA cycle metabolites (as shown in panel A) in 24-hour differentiated HUDEP1 cells expressing either GATA1FL or GATA1s (1D5). The ratio of labeled PEP isotopologs is shown on the right. (C) Intracellular PK activity measured in 24-hour differentiated HUDEP1 cells expressing GATA1FL or GATA1s (1D5 and 1KI9) using the Abcam PK Assay Kit (ab83432). (D) Extracellular lactate levels measured from culture media of 24-hour differentiated HUDEP1 cells expressing GATA1FL or GATA1s (1D5 and 1KI9), using the Abcam L-Lactate Assay Kit (ab65330). In panels B-D, asterisks indicate statistically significant differences. (E) CUT&RUN heat maps comparing H3K27la and H3K18la histone modifications in HUDEP1 cells expressing GATA1s (1D5) vs GATA1FL after 24 hours of differentiation. (F) IGV snapshots of CUT&RUN tracks for H3K27ac, H3K27la, and H3K18la at the PKM locus in HUDEP1s (1D5) vs GATA1FL cells. (G) Culture media lactate concentration in FL1 and 1D5 HUDEP1 clones treated with vehicle (DMSO) or GSK2837808A. (H) Western blot of lactylated lysine in extracts from GATA1FL1 and G1s_1D5 clones treated with DMSO or GSK2837808A. (I) Flow cytometry plots of erythroid differentiation in GATA1s_1D5 clone treated with varying doses of GSK2837808A at days 3 and 5. ∗P < .05; ∗∗P < .01. Acetyl-CoA, acetyl coenzyme A; diff, differentiation; GSK, GSK2837808A; ns, no statistical significance; Pyr, pyruvate; Suc, succinate; TCA, trichloroacetic acid; TES, transcription end site; TSS, transcription start site.

Next, we examined protein lactylation, a posttranslational modification induced by intracellular lactate accumulation.²⁵ In particular, histone lysine lactylation has been shown to have a profound impact on transcription, similar to other histone epigenetic modifications.^26,27 To assess histone lactylation, we performed CUT&RUN with antibodies against H3K27la and H3K18la in HUDEP1 cells after 24 hours of differentiation. Compared with cells expressing GATA1FL, GATA1s cells exhibited a global increase in H3K27la and H3K18la (Figure 5E). Furthermore, both H3K27la and H3K18la were highly enriched at the PKM locus in GATA1s HUDEP1 cells, colocalizing with the active histone mark H3K27ac (Figure 5F). In contrast, no histone lactylation was enriched at the PKLR locus, suggesting that only a specific subset of genes is selectively activated by GATA1s-driven glycolytic reprogramming (supplemental Figure 6A). This increase in histone lactylation likely reinforces the persistent expression of PKM in GATA1s mutant cells.

To confirm the results obtained with HUDEP1 clones in a primary cell setting, we performed CUT&RUN for lactylated histones in murine CD71^hiTer119^hi cells collected from Gata1s and wild-type littermate embryos at E13.5. As observed in GATA1s HUDEP1 cells, H3K27ac, H3K27la, and H3K18la were increased along the Pkm gene in Gata1s erythroblasts compared with wild-type cells (supplemental Figure 6B). A global increase in lactylation on H3K27 and H3K18 was also consistent with the findings in human HUDEP1 cells (supplemental Figure 6C). To further investigate the effects of histone lactylation on gene expression, we analyzed our previous mRNA sequencing data obtained from wild-type mouse CD71^hiTer119^hi erythroblasts (GSE130156) to examine the histone lactylation occupancy on differentially expressed genes, which are categorized into low expression (0%-33.3%), medium expression (33.3%-66.6%), and high expression (66.6%-100%). The results confirmed a positive correlation between gene expression level and histone lactylation (supplemental Figure 6D). Furthermore, by analyzing the dysregulated genes in Gata1s erythroblasts, upregulated genes exhibited elevated occupancy of H3K27ac, H3K27la, and H3K18la; conversely, downregulated genes showed reduced occupancy of these histone modifications (supplemental Figure 6E-F). Collectively, the findings in HUDEP1 cells are well validated in primary mouse erythroblasts.

Next, to determine whether increased lactate levels contribute to the impaired differentiation of GATA1s-mutant HUDEP1 cells, we treated cells with the lactate dehydrogenase A inhibitor, GSK2837808A. We found that the level of lactate in the GATA1FL1 and G1s_D5 clones was significantly reduced with drug treatment and that this was associated with decreased lactyllysine (Figure 5G-H). Of note, the reduction in lactate afforded by GSK2837808A improved the differentiation trajectory of the G1s_D5 clone as evidenced by the increased proportions of CD49d⁺Band3⁻ and CD49d⁺Band3⁺ cells (Figure 5I). Lactate dehydrogenase A inhibition did not, however, rescue the growth defect of GATA1s mutant HUDEP1 cells (supplemental Figure 6G).

Enhanced glycolysis is also a feature of human erythroid cells with a GATA1 deficiency

Next, we introduced the GATA1s mutation, by exon-2 deletion,²⁸ into healthy human CD34⁺ cells by CRISPR, cultured the cells for 5 days under erythroid differentiation conditions, and then sorted the CD71⁺CD235a⁻ population. Compared with the adeno-associated virus integration site 1 (AAVS1) single-guide RNA, the single-guide RNA targeting GATA1 led to a significant reduction in GATA1FL with a concomitant increase in GATA1s protein and increased levels of PKM2 and phosphorylated PKM2 (Figure 6A-B). The editing efficiency at this time point approached 44% (Figure 6C). Seahorse glycolysis stress tests revealed elevated glycolysis, glycolytic capacity, and glycolytic reserve under differentiation conditions in the GATA1s-edited erythroid cells derived from CD34⁺ cells of 3 independent donors (Figure 6D-E). Finally, editing of GATA1 also led to increased PK activity as expected (Figure 6F).

Figure 6.

Loss of full-length GATA1 enhances glycolysis of erythroid cells derived from human CD34⁺ cells. (A) Western blots assessing the levels of GATA1, GATA1s, PKM2, and p-PKM2 in CD71⁺CD235a⁻ cells derived from 3 different CD34⁺ cell donors following knockout of full-length GATA1. HSC70 is shown as a loading control. (B) Quantitation of western blot data. (C) Editing efficiency of the CRISPR experiment as measured in sorted CD71⁺CD235a⁻ cells. (D) Seahorse XF Pro glycolysis stress test performed on erythroid cells isolated from cultures of 3 different CD34⁺ cell donors. Traces show ECAR in response to sequential injection of 20mM glucose, 1μM oligomycin, and 100mM 2-deoxy-d-glucose (2-DG). (E) Quantification of glycolytic parameters from panel D, including glycolysis, glycolytic capacity, and glycolytic reserve. (F) PK activity in purified CD71⁺CD235a⁻ cells with and without GATA1 editing. 2-DG, 2-deoxy-d-glucose; p-PKM2, phosphorylated PKM2.

Enhanced glycolysis is a feature of RPS19-mutant erythroid cells

DBA is a congenital bone marrow failure characterized by erythroid hypoplasia that is predominantly caused by mutations in ribosomal protein (RP) genes. Among patients with DBA, 70%-80% harbor heterozygous loss-of-function variants in RP subunits, most commonly RPS19.²⁹ These RP gene mutations have been demonstrated to impair erythroid cell development by a selective reduction in translation of GATA1.³⁰ Moreover, a small proportion of patients with DBA have a GATA1 mutation that exclusively produces GATA1s.⁵ Recent studies have indicated that the restoration of the GATA1FL gene may hold promise as a potential universal gene therapy approach for patients with DBA,³¹ suggesting that a GATA1 deficiency is key to the pathogenesis of DBA. By reanalyzing published single-cell transcriptome data from patients with DBA,³² we found that glycolytic pathway genes such as PKM, ENO1, GAPDH, HK1, LDHA, PGAM1, and TPI1, exhibit elevated expression in EP and megakaryocytic progenitor cells of patients with DBA (Figure 7A). Notably, most alterations in gene expression were more pronounced in the EP population. Furthermore, in the EP population, PKM expression was 2.3-fold higher in all patients and 8.2-fold higher in patients with the RPS19 mutation compared with healthy donors (Figure 7B). By analyzing the gene expression profiles of each subgroup (ie, RPL-DBA and RPS19-DBA), we observed that PKM was selectively elevated in RPS19-DBA samples compared with healthy donors (Figure 7A-B; supplemental Figure 7B). Conversely, the expression of PKLR was not increased; instead, it exhibited a 30% decrease in RPS19-DBA samples compared with healthy donors (supplemental Figure 7A). Although elevated PKM expression is also noted in other lineages, it is substantially less pronounced than in EPs (supplemental Figure 7B).

Figure 7.

Erythroid cells from patients with DBA show evidence of enhanced glycolytic gene expression and loss of RPS19 enhances glycolysis. (A) Heat maps showing differential expression of glycolytic enzymes in all patients with DBA vs healthy donors (left) and patients with DBA with RPS19 mutations vs healthy donors (right). ∗Differentially expressed genes are defined as having an adjusted P value (padj <.05) and a fold change ≥30%. Data represent a reanalysis of scRNA-seq of patients with DBA with any ribosomal gene mutation (left) or only RPS19 mutant cases (right).³² (B) FCs in PKM gene expression corresponding to comparisons shown in panel A. Asterisks indicate statistical significance (∗Padj <.05). (C) Western blot of GATA1, GATA1s, and RPS19 protein levels in CD34⁺ cells 3 days after the RNP electroporation delivery to delete RPS19. H3 is shown as a loading control. (D) Editing efficiency and western blot quantitation of RPS19 and GATA1 protein levels. (E) Seahorse XF Pro glycolysis stress test performed on erythroid cells isolated from cultures of 3 different CD34⁺ cell donors. Traces show extracellular acidification rate (ECAR) in response to sequential injection of 20mM glucose, 1μM oligomycin, and 100mM 2-deoxy-d-glucose (2-DG). (F) Quantification of glycolytic parameters from panel E, including glycolysis, glycolytic capacity, and glycolytic reserve. (G) PK activity in purified CD71⁺CD235a⁻ cells with and without RPS19 editing. 2-DG, 2-deoxy-d-glucose; EoMBP, eosinophil mast cell/basophil; fl/s, GATA1FL versus GATA1s; HSC, hematopoietic stem cell; MPP, multipotent progenitor; MyP, myeloid/monocyte/macrophage/neutrophil progenitor; RNP, ribonucleoprotein.

Finally, to directly assess whether decreased RPS19 is associated with altered metabolism, we knocked out RPS19 by CRISPR in human CD34⁺ cells. We observed decreased levels of RPS19 and both GATA1 and GATA1s in pools of cells 3 days after electroporation (Figure 7C-D). We then cultured the cells for 5 days in erythroid differentiation conditions and sorted CD71⁺CD235a⁻ cells. Of note, we observed decreased levels of RPS19 with concomitant increases in PKM2 and phosphorylated PKM2 protein (supplemental Figure 8A-B). Seahorse glycolysis stress tests revealed that the decreased RPS19 led to increased glycolysis, glycolytic capacity, and glycolytic reserve (Figure 7E-F). RPS19-targeted cells also had elevated PK activity (Figure 7G).

Discussion

Several lines of evidence support the pivotal role of the N terminus of GATA1 in normal erythroid differentiation. These include the phenotype of patients with N-terminal truncations, the inability of GATA1 N-terminal–deleted induced pluripotent stem cells (iPSCs) to generate erythroid cells, and red blood cell defects in mice expressing only the GATA1s. Nevertheless, the molecular mechanisms underlying the erythropoietic defects associated with GATA1 N-terminal mutations remain enigmatic.

GATA1 is well-known to regulate heme metabolism, by positively controlling the expression of key genes such as ALAS2, FECH, and BACH1.^{33, 34, 35} However, a contribution of GATA1 to other metabolic processes including energy metabolism has not been described. Here, we show that GATA1 normally suppresses PKM gene expression during erythropoiesis, whereas GATA1s allows for persistent expression, likely, in part, indirectly through increased histone lactylation, which is a consequence of the increased lactate production. Of note, PK activators show promise as novel treatments for thalassemia, and perhaps sickle cell disease by virtue of their ability to stimulate maturation/survival of EP cells.^{36, 37, 38, 39, 40} Our study suggests that tight control of PK expression is necessary for erythropoiesis and that its dysregulation is associated with GATA1s mutations. In addition, our work suggests that there is a feed-forward mechanism by which loss of GATA1 leads to increased production of lactate, which then increases histone lactylation that further enhances expression of glycolytic genes. Furthermore, phosphorylation of PKM2 at S37 promotes its translocation and induces a positive feedback loop on its own expression and that of many other glycolysis pathway genes.⁴¹ Of note, energy metabolism is tightly regulated during erythropoiesis, with a late switch from OXPHOS to glycolysis; the decreased expression of GATA1 at this stage may facilitate the derepression of PKM, enabling this switch. By contrast, GATA1s leads to activation of PKM in early erythroid precursors, leading to the aberrant premature activation of glycolysis that likely contributes to impaired erythropoiesis.

We also discovered that PKM expression is higher in RPS19 mutant cases of DBA. This is consistent with single-cell data that showed that RPL mutant cases, in which GATA1 expression and activity are preserved, represent distinct entities. The data from the human EP cell line HUDEP1 and patients with RPS19-mutated DBA also revealed aberrant PKLR expression in the absence of GATA1FL. Early in erythropoiesis, PKM is predominant over PKLR but as development proceeds, PKM2 expression is reduced while the erythroid-specific isoform PKR takes over PKM2’s role in glycolysis. However, in the presence of only GATA1s (and in patients with RPS19-mutated DBA), the PKM2/PKR switch fails, resulting in PKM2 hyperactivation and an inability to upregulate PKR. The failure of PKM and PKLR switch may also contribute to the GATA1-mediated erythroid defects in DBA.

Mitapivat, an activator of PKLR and PKM,⁴² is an approved therapy for hemolytic anemia with PK deficiency⁴³ and is being tested in phase 3 clinical trials for patients with α- or β-thalassemia (ClinicalTrials.gov identifiers: NCT04770753 and NCT04770779)⁴⁴ and sickle cell disease (NCT05031780).⁴⁵ Of note, it has been reported that mitapivat reduced anemia and transfusion burden in patients with thalassemia.⁴⁶ Notably PKM2/PKR switch failure has also been observed in β-thalassemia.³⁶ In this study, mitapivat treatment resulted in elevated PKR expression and decreased PKM2 expression at the protein level, which was accompanied by a decrease in PKM2 mRNA, suggesting restoration of the PKM2/PKR switch. In addition, the drug led to reduced expression of PKM2-HIF (hypoxia inducible factor) target glycolytic genes, including lactate dehydrogenase A (Ldha) and pyruvate dehydrogenase kinase 1 (Pdk1) as well as enhanced erythroblast maturation. Additionally, the decreased expression of Ldha was hypothesized to downregulate lactate production in β-thalassemia. Normalizing lactate levels may lead to the restoration of histone lactylation, ultimately resulting in the normalization of gene expression, including PKM2.

Although we were unable to restore erythroid maturation of GATA1s-mutant cells with PKM2 inhibitors, we predict that PKM2/PKLR rescue is more critical for overcoming the defect than only inhibiting PKM2 itself and that inhibition of PKM2 activity cannot rescue nuclear PKM2 activity and histone lactylation. Given the numerous pathways that are controlled by GATA1, rescue of erythropoiesis of GATA1s cells likely requires not only restoration of glycolytic control, but also of heme metabolism and β-globin gene expression.

Conflict-of-interest disclosure: J.D.C. is a scientific advisor for Alethiomics and receives research support from Syndax Pharmaceuticals. The remaining authors declare no competing financial interests.