The equilibrative nucleoside transporter ENT1 is critical for nucleotide homeostasis and optimal erythropoiesis

The commitment of hematopoietic stem cells (HSCs) to the erythroid lineage is a complex process controlled by extrinsic factors as well as the physical interaction of HSCs with their microenvironment.^1-5  Nutrient resources and their transport has recently emerged as key regulators of HSC proliferation and lineage commitment.^6-8  Notably, a sufficient intracellular nucleotide pool has been shown to be essential for erythroid lineage commitment of HSCs.⁶  Nucleotide biosynthesis is regulated by the intracellular availability of glutamine and the shunting of glucose from glycolysis pathway through the pentose phosphate pathway.⁶  Nucleotides are indispensable precursors for RNA and DNA synthesis and are extremely important for cell division.^9,10  Balanced pools of intracellular nucleotides are crucial to avoid DNA damage in erythroid and lymphoid lineages.^11-14  Furthermore, nucleotide homeostasis is also crucial for terminal erythroid differentiation and the generation of mature red blood cells (RBCs). Cyclic adenosine monophosphate (cAMP)- and guanosine 3′,5′-cyclic monophosphate (cGMP)-mediated signaling is associated with the regulation of erythropoiesis through the modulation of numerous signaling pathways^15-23  and the activation of important erythroid-specific transcription factors.^24,25 

HSCs lack the capacity to carry out de novo nucleoside synthesis,²⁶  and thus rely on membrane transporters to acquire nucleosides for DNA and RNA synthesis and adenosine triphosphate (ATP) production.²⁶  The major source of nucleoside recycling is the degradation of nucleic acids, arising from autophagic (or phagocytic) pathways²⁶  and the degradation of nucleotides in the extracellular compartment. The equilibrative nucleoside transporter 3 (ENT3) is the major nucleoside transporter responsible for maintaining nucleoside homeostasis in lysosomal and mitochondrial compartments.^27,28  In humans, ENT3 alteration causes numerous genetic disorders with a wide spectrum of clinical manifestations including hematological abnormalities.^29-33  Mechanistically, the altered commitment of ENT3-deficient HSCs is caused by the loss of lysosomal adenosine transport, which impedes autophagy-regulated HSC fates.³³  However, ENT1 is the major plasma membrane transporter in HSCs that ensures nucleoside uptake from the extracellular compartment.^34,35  ENT1 plays a central role in the homeostasis of plasma adenosine which is generated extracellularly by the degradation of ATP via the ectonucleotidases CD39 and CD73.^36,37  Although extracellular uptake of nucleosides is an important step in nucleotide biosynthesis, its specific contribution in driving HSC lineage commitment remains to be defined. This is of critical interest as humans with ENT1 deficiency, because of a c.589+1G>C mutation in the SLC29A1 gene encoding ENT1, have recently been described.³⁸ 

Here, we demonstrate that ENT1 deficiency in humans and mice is associated with erythroid defects and identify a functional link between ENT1 and ABCC4-mediated cyclic nucleotide metabolism in erythroid differentiation. Our genetic and pharmacological findings open new avenues for the development of novel therapeutic strategies for the treatment of anemia.

Detailed protocols for all methods are provided in supplemental Methods, available on the Blood Web site.

Fresh peripheral blood samples were obtained from healthy donors from the Etablissement Français du Sang and ENT1_null patients after obtaining written informed consent.

Human primary CD34 cells, isolated from the peripheral blood of patients and healthy donors or from cord blood, were cultured as described in the supplemental Methods.

ENT1 was downregulated in CD34⁺ progenitors using a short hairpin RNA (shRNA) approach.

Phosphoproteomic analyses were performed using the label-free proteomic methodology. Intracellular pools of nucleotides were quantified by high-performance liquid chromatography mass spectrometry.

All protocols involving animal studies were reviewed and approved by the Institutional Animal Welfare Committee of the University of Texas Health Science Center at Houston.

A homozygous splicing mutation in the SLC29A1 gene was previously identified in 3 Augustine-null phenotype siblings (P1, P2, P3), resulting in total deficiency of ENT1³⁸  (Figure 1A; supplemental Figure 1). RBC indices of the 3 ENT1_null probands revealed macrocytosis (mean corpuscular volume, 104-108 fL) and increased cell hemoglobin (Hb) content (mean hemoglobin concentration, 34-37 pg) but normal cell Hb concentration (mean corpuscular volume concentration) (supplemental Table 1). The normal reticulocyte counts in all probands are consistent with the absence of anemia and hemolysis (normal lactate dehydrogenase and haptoglobin values). However, morphological examination revealed >40% of RBCs with altered cell shapes and sizes, a characteristic of anisopoikilocytosis, in all 3 patients (Figure 1B). Although RBCs from P1 and P2 showed similar morphological alterations, the blood smear of P3 revealed a more severe phenotype with the presence of elliptocytes and a higher percentage of schizocytes (Figure 1B-C). Furthermore, in conjunction with more severe morphological abnormalities, RBC membrane deformability was impaired in the P3 proband as reflected by the extent of increase in Elongation Index as a function of increasing applied shear stress³⁹  (Figure 1D). ENT1_null RBCs displayed heterogenous osmotic gradient ektacytometry profiles, confirming distinct RBC changes in these patients with the profile in P3 consistent with elliptocytosis (Figure 1E). This macrocytosis was not due to deficiency of either vitamin B₁₂ or folate in ENT1_null probands because their levels were within the normal range (539 ± 134 pmol/L and 9.2 ± 0.2 nmol/L, respectively). Except for moderate ectopic mineralization also observed in slc29a1^−/− mice,⁴⁰  ENT1_null probands did not exhibit significant nonhematological clinical manifestations.³⁸ 

To further evaluate the RBC membrane defect in ENT1_null RBCs, we performed a comparative global phosphoproteomic analysis of RBC membrane proteins. This analysis was performed because changes in cyclic nucleotide levels have long been known to affect protein kinases and thereby the phosphorylation state of the cell.⁴¹  Phosphoproteome analysis identified 501 serine, threonine, and tyrosine phosphorylation sites in 182 cytoskeletal and transmembrane proteins across all samples (supplemental Excel file). Ten serine sites were differentially hypophosphorylated in 8 proteins from ENT1_null RBCs (Figure 1F, P < .01; supplemental Figure 2); these proteins included membrane proteins such as spectrin, ankyrin, adducin, and protein 4.1 as well as non-membrane proteins such as ALDOA. Notably, these membrane proteins form a complex that is critical for membrane stability and RBC morphology.^42,43  Interestingly though, Triton X-100 extractability of these proteins from ENT1_null RBCs was unchanged compared with controls (supplemental Figure 3), suggesting that other molecular mechanism(s) besides spectrin-based skeletal complexes underlie the abnormal RBC morphology of ENT1_null individuals. On the other hand, 2 proteins were hyperphosphorylated in ENT1_null RBCs compared with controls; the chloride nucleotide-sensitive channel CLNSA1 (also called pICln) and PNPLA6 (Figure 1F; P < .01). Interestingly, CLNSA1 has been shown to regulate RBC volume via its interaction with the 4.1R protein.⁴⁴  Moreover, some differences in protein phosphorylation can be appreciated among ENT1_null ghosts (supplemental Excel file).

Because ENT1 is the major nucleoside transporter in RBCs,⁴⁵  we evaluated the nucleotide-derived metabolites in ENT1_null RBCs by high-performance liquid chromatography mass spectrometry. Importantly, we noted that the intracellular levels of both purine and pyrimidine nucleotides (dCTP, dCDP, common myeloid progenitor [CMP], GTP, and dATP) were significantly increased in ENT1_null RBCs compared with control RBCs (Figure 1G; P < .05, P < .01, and P < .001). This accumulation of deoxynucleotides (dNTPs) in ENT1_null RBCs could either be due to the altered dNTP uptake from the plasma or to an inappropriate uptake of dNTPs by the mitochondria of ENT1_null erythroblasts because ENT1 has been implicated in dNTP transport across the mitochondrial membrane.^46-48  Concerning the cyclic nucleotides, there were no statistically significant differences in cAMP levels between patient and control RBCs (Figure 1G). Interestingly though, cGMP levels were extremely variable, with high levels in RBCs from P1 and P2 and very low levels in P3.

These data indicated that the absence of ENT1 results in a greater phenotypic change in red cells from patient P3 as compared with patients P1 and P2. We therefore performed whole exome sequencing to assess whether other genetic variant(s) might account for these differences (Figure 1H). Analyzing whole exome sequencing data by filtering common pathogenic variants shared by P1 and P2, and not P3, led to the identification of a heterozygous loss-of-function variant in the ABCC4 gene (c.559 G>T; p.Gly187Trp). ABCC4 encodes a cyclic nucleotide exporter highly expressed in erythroid precursors (Figure 1I).⁴⁹  Interestingly, this finding is consistent with the 6.3- and 5.2-fold higher level of cGMP in P1 and P2 RBCs, respectively, compared with controls (Figure 1G). Moreover, cGMP levels in P3 RBCs were actually 2.8-fold lower than in control RBCs. Thus, the ABCC4 mutation in P1 and P2 is associated with a significant increase in RBC cGMP levels compared with P3 and healthy controls, and may function to compensate for the loss of ENT1.

Having documented alterations in mature erythrocytes from ENT1_null individuals, we investigated the role of ENT1 during erythropoiesis. Ex vivo erythropoiesis of CD34⁺ cells from the peripherical blood of all 3 ENT1_null patients was investigated. CD34⁺ progenitors from P1 and P2 underwent erythroid differentiation in the presence of rEPO with increased expression of the glycophorin A (GPA) and Band3 erythroid markers, to similar levels as control progenitors (Figure 2A). Furthermore, in vitro reticulocyte levels for both patients at day 12 were similar to controls (40 ± 2% vs 38 ± 6% in 2 independent experiments, respectively), indicating that the absence of ENT1 in erythroid precursors from P1 and P2 did not inhibit erythroblast enucleation (Figure 2B). Interestingly, macrocytic reticulocytes were observed from the ex vivo erythroid culture of P2 (supplemental Figure 4).

In contrast to P1 and P2, CD34⁺ progenitors from P3 patient failed to undergo ex vivo erythroid differentiation (n = 4; Figure 2A). Only 16% ± 5% of P3 cells acquired GPA expression at day 5 compared with 88% ± 9% for controls (Figure 2A,C; P < .0001). Furthermore, within the GPA⁺ subset, only 3% ± 2% erythroblasts exhibited an upregulation of Band3 vs 24% ± 6% in control erythroblasts (Figure 2C; P < .001). This markedly delayed erythroid differentiation was associated with increased cell death (supplemental Figure 5) and markedly decreased cell proliferation (Figure 2D; P < .01). The ineffective in vitro erythropoiesis of P3 progenitors was confirmed by May-Grunwald-Giemsa (MGG) staining, revealing immature erythroblasts in controls conditions, whereas myeloid cells were noted in cultures of P3 patient (Figure 2E). Furthermore, only 4% of P3 progenitors progressed to the erythroid colony-forming unit (CFU-E) stage, defined as IL3R⁻CD34⁻CD36⁺ compared with 33% of control progenitors (Figure 2F). Consistent with these data, P3 CD34⁺ cells failed to give rise to CFU-E colonies in methylcellulose assays, whereas control cells were able to form large burst-forming unit erythroid and CFU-E colonies (Figure 2G). Interestingly, this defect may be due to alterations in the characteristics of CD34⁺ progenitors migrating to the peripheral circulation, whereas CD34⁺ cells generally account for ∼0.5% of PBMC,⁵⁰  and we detected 0.63% CD34⁺ cells in a representative healthy donor, the percentage of CD34⁺ progenitors in P3 was much lower, at 0.17% (supplemental Figure 6). Altogether, our findings highlight a difference in the ability of circulating CD34⁺ progenitors from P1/P2, compared with P3, to undergo ex vivo erythroid differentiation.

To specifically assess the impact of ENT1 on erythropoiesis, we pursued an shRNA-mediated ENT1-knockdown strategy as illustrated in Figure 3A. Monitoring the surface expression profile of ENT1 during erythropoiesis by flow cytometry showed a high-level expression of ENT1 on progenitor cells before rEPO stimulation (day 0) followed by a gradual decrease during terminal erythroid maturation (Figure 3B). Transduction of cord blood CD34⁺ progenitors with an shENT1 lentiviral vector resulted in a pronounced reduction in SLC29A1 messenger RNA (mRNA) levels (90% ± 4%; Figure 3C) and a concomitant decrease in cell surface expression of ENT1 (40% ± 8% at day 0; supplemental Figure 7A-B). Although ENT1 knockdown did not significantly alter expression of CD34, CD36, or IL3R before rEPO-induced erythroid differentiation (day 7; supplemental Figure 8), shENT1-transduced progenitors were counterselected following addition of EPO. The percentages of GFP⁺ cells decreased by more than 50% between day 0 and day 5 of erythroid differentiation and by more than 80% at day 7 (Figure 3D). The negative impact of ENT1 knockdown was specific for erythroid differentiation as the percentage of shENT1-GFP⁺ progenitors remained stable upon expansion of CD34⁺ progenitors in the absence of rEPO (Figure 3D). In addition, the EPO-induced expansion of GFP⁺ cells was significantly lower for shENT1-transduced compared with shControl-transduced progenitors (Figure 3E). These findings strongly suggest that ENT1 expression is crucial for erythroid, but not nonerythroid, differentiation.

To address the effect of ENT1 downregulation on terminal erythroid differentiation, we monitored the surface expression of the GPA and CD71 erythroid markers. EPO-induced erythroid differentiation of hematopoietic progenitors was significantly attenuated following ENT1 downregulation and, notably, this effect was only detected in the shENT1-transduced subset (Figures 3F-G). At day 3 of differentiation, only 42% ± 5% and 41% ± 8% of shENT1-transduced cells expressed GPA and CD71, respectively, compared with 55% ± 9% and 66% ± 5% of shControl-transduced cells (P < .01 and P < .05; Figure 3G). Consistent with this finding, erythroblast enucleation was significantly reduced in shENT1-GFP-transduced cells (5% ± 2% vs 20% ± 6%, P < .05; Figure 3H). In addition, flow cytometry analysis revealed that >25% of shENT1-transduced progenitors expressed the CD11a and CD33 myeloid markers compared with only 4% of control cells (P < .05; Figure 3I). Overall, these findings reveal an important role for ENT1 in erythroid differentiation.

To assess whether the absence of ENT1 in humans results in similar alterations in ENT1_null mice, we analyzed the complete blood count in 8- to 12-week-old Ent1^−/− mice compared with their wild-type (WT) counterparts. Both male and female Ent1^−/− mice exhibited a macrocytosis and increased mean hemoglobin concentration that was similar to that detected in ENT1-deficient patients (supplemental Table 2); blood smears revealed an increased size of RBCs in Ent1^−/− mice (Figure 4A), as well as a significantly increased mean corpuscular volume (P < .01). Additionally, RBC numbers and hematocrit were reduced in both male and female Ent1^−/− mice compared with gender-matched control mice (supplemental Table 2). These findings point to similarities as well as differences in the erythroid phenotype resulting from ENT1 deficiency in humans and mice.

Erythropoiesis is initiated from blood marrow (BM)-derived HSCs that differentiate to common myeloid progenitors (CMPs), followed by commitment to megakaryocyte erythroid progenitors (MEPs) vs granulocyte myeloid progenitors (GMPs). MEPs generate erythroid progenitors that undergo terminal differentiation to generate mature RBCs^51,52  (Figure 4B). Given that Ent1^−/− mice displayed mild anemia, we evaluated the phenotype of hematopoietic progenitors in their BM compared with WT mice. There were no differences in the number of Lin⁻cKit⁺ cells between both mice groups (Figure 4C). However, Ent1^−/− mice displayed a significant decrease in CMP (CD34⁺CD16/32⁻) and MEP (CD34⁻CD16/32⁻) cells but an increased GMP (CD34⁺CD16/32⁻) cells compared with WT mice (P < .05; Figure 4C). These data point to decreased erythroid differentiation and indeed, the percentage of Ter119⁺ erythroid cells in the BM of Ent1^−/− mice was lower than WT mice, whereas the level of Gr-1⁺ granulocytes was increased (P < .05, Figure 4D).

In mice, the spleen is the major site for stress erythropoiesis.⁵³  Surprisingly, we found that Ent1^−/− mice displayed splenomegaly (P < .05; Figure 4E) compared with their WT counterparts at 8 to 12 weeks of age, raising an intriguing possibility of an increased erythropoiesis to counteract the anemic state of these mice. Interestingly, flow cytometry analysis of erythroid precursors in the spleen showed an increase in early progenitors (populations I and II) with equivalent levels of later progenitors (Figure 4F). Thus, there is an increase in the total splenic Ter119⁺ population in Ent1^−/− mice (Figure 4G). To rule out the possibility that the splenomegaly in Ent1^−/− mice is due to increased clearance of RBCs by the spleen, we measured red cell life span by labeling all circulating erythrocytes in vivo by injection of N-succinimidyl-6-[biotinamido] hexanoate. The percentage of biotinylated erythrocytes in the circulation in Ent1^−/− mice was not decreased compared with WT mice (Figure 4H), strongly suggesting that increased RBC destruction was not responsible for the splenomegaly detected in Ent1^−/− mice.

The abnormal protein phosphorylation and nucleotide metabolome in human mature ENT1_null RBCs, in conjunction with the increased adenosine plasma levels in ENT1_null probands (P < .0001; supplemental Figure 9), as previously reported for Ent1^−/− mice,⁵⁴  confirm the importance of adenosine as a substrate of ENT1. To address the question of whether ENT1 regulates erythropoiesis via its capacity to transport adenosine, we performed ex vivo differentiation of CD34⁺ progenitors from healthy donors following addition of adenosine. Enhanced terminal erythroid differentiation was noted in the presence of adenosine (15 µM), a concentration that did not alter proliferation or survival (not shown). Differentiation was monitored as a function of α4-integrin/Band3 staining on GPA⁺ cells, with an α4-integrin^low/Band3^low phenotype representing the most differentiated erythroblasts. Notably, adenosine significantly accelerated erythroid differentiation in control progenitors (Figures 5A-B); this finding was also confirmed by MGG staining (Figure 5C). However, adenosine did not enhance erythroid differentiation following shRNA-mediated knockdown of ENT1, as monitored by GPA and Band3 expression (Figures 5D-E), and did not rescue shENT1-transduced progenitors from being negatively selected (supplemental Figure 10). These data are in accord with ENT1’s role as a major adenosine transporter expressed in erythroid precursors.^34,35  Furthermore, shENT1-transduced progenitors upregulated myeloid lineage markers despite the presence of rEPO and adenosine (Figure 5D). These findings demonstrate the inability of adenosine to enhance erythroid differentiation in the absence of ENT1, strongly suggesting that ENT1 regulates erythroid differentiation via its role as adenosine transporter.

Because adenosine is an important precursor in the synthesis of other adenine nucleotides, we analyzed the nucleotide metabolome of control erythroblasts cultured in the presence of adenosine. Interestingly, the cyclic nucleotides cAMP and cGMP were significantly increased in erythroblasts treated with adenosine (Figure 5F). Phosphorylation of the cAMP-response element-binding (CREB) protein, a transcription factor activated by cyclic nucleotides, decreased as a function of erythroid differentiation (supplemental Figure 11). However, under conditions where extracellular adenosine was increased, CREB phosphorylation was increased (Figure 5G; supplemental Figure 12). Accordingly, protein kinase A (PKA) phosphorylation was also significantly increased in the presence of adenosine, confirming an adenosine-stimulated cyclic nucleotide biosynthesis (Figure 5H). Together, these data reveal the importance of ENT1 in mediating adenosine uptake to regulate cyclic nucleotide biosynthesis and erythropoiesis.

Although all ENT1_null patients exhibited macrocytosis, the impact of ENT1 on RBC deformability was more pronounced in P3 compared with P1/P2. Furthermore, in vitro erythroid differentiation was altered in P3 progenitors but was not significantly attenuated in P1 and P2 progenitors. The attenuated erythroid phenotype of P1 and P2 compared with P3 was associated with increased RBC cyclic nucleotide levels (Figure 1G). In addition, P2 erythroblasts exhibited an increased accumulation of cAMP and cGMP compared with control erythroblasts, whereas shENT1-transduced erythroblasts exhibited a reduced level of cAMP (supplemental Figures 13 and 14). Thus, ENT1 may regulate erythropoiesis via cyclic nucleotide metabolism, and our data further suggest that a mutation in ABCC4 compensates for the absence of ENT1, attenuating the erythroid phenotype in patients P1 and P2.

To test this hypothesis, we evaluated the effect of an ABCC4 inhibitor (MK-571) on ex vivo erythropoiesis of healthy donors (supplemental Figure 15). As shown in Figure 6, MK-571 significantly enhanced erythropoiesis as indicated by the increased GPA expression (Figure 6A). Moreover, MK-571 increased terminal erythroblast maturation and dramatically increased enucleation by approximately twofold (P < .05; Figure 6B-C). MK-571 also enlarged the pool of erythroid cells as shown by a significant increase in expansion without changes in apoptosis or cell death (supplemental Figures 115B-C).

To specifically evaluate the impact of ABCC4 function on erythroid differentiation in the murine system, we inhibited ABCC4 function in Ent1^−/− mice with MK-571.⁵⁵  Retro-orbital administration of MK-571 (10 mg/kg for 6 days) increased total RBCs, Hb, and hematocrit in Ent1^−/− mice to levels similar to those detected in WT mice (Figure 6D). Interestingly though, MK-571 treatment did not alter red cell macrocytosis in Ent1^−/− mice, strongly suggesting that erythropoiesis, but not macrocytosis, is regulated by ABCC4 in the absence of ENT1. Moreover, our data suggest that ABCC4 negatively regulates erythroid differentiation even in the presence of ENT1. MK-571 treatment significantly increased total RBCs, Hb production, and hematocrit in WT mice. MK-571 treatment increased the percentage of circulating reticulocytes in both Ent1^−/− and WT mice (Figure 6D), supporting this newly described association between ABCC4 and ENT1 in erythropoiesis. We then evaluated the potential impact of this drug on early stages of erythroid commitment. MK-571 treatment led to an increased percentage of BM MEPs in both WT and Ent1^−/− mice (P < .05 and P < .001, respectively; Figure 6E; supplemental Figure 16), but did not alter the percentage of Lin⁻/cKit⁺ progenitors in either group (Figure 6F). This increase was associated with a significantly decreased percentage of BM GMPs in WT mice as well as BM CMPs in Ent1^−/− mice (P < .05; Figure 6E). This MEP/CMP shift was coupled to a significant increase in BM Ter119⁺ erythroid cells in both WT and Ent1^−/− mice and a decreased percentage of Gr-1⁺ cells in BM of WT mice (Figure 6G). These experiments, performed on patient cells and following pharmacological manipulation of both human and murine ENT1^−/− progenitors, highlight a novel crosstalk between ABCC4 and ENT1 in regulating erythroid differentiation.

Here, we show that ENT1_null RBCs exhibit a defective phosphorylation of membrane proteins that is associated with distinct misshaped cells, confirming the crucial role of ENT1-mediated nucleoside transport in the regulation of metabolic pathways and kinase activities. Although the finely tuned regulation of membrane protein phosphorylation has previously been shown to be critical in the control of cell volume and membrane organization via ion channel activities and protein–protein interactions, respectively,^42,56-59  our finding of dNTPs accumulation in ENT1_null RBCs, reveals an important role of ENT1 in deoxynucleotide recycling and nucleotide homeostasis during erythroid diffrentiation.⁴⁸  Balanced nucleotide pools, controlling replication stress, DNA damage, protein kinases, as well as transcription factors,^19,23,60-62  have been shown to regulate hematopoietic lineage fate.^13,14  Indeed, we provide unequivocal demonstration that ENT1-mediated adenosine transport regulates the intracellular level of cyclic nucleotides in hematopoietic progenitors, altering erythroid differentiation under conditions where the phosphoproteome of ENT1_null RBCs is abnormal. Finally, our data show that knockdown of ENT1 in human CD34⁺ progenitors impede in vitro erythropoiesis while increasing ENT1-mediated adenosine transport enhances erythroid differentiation.

The absence of anemia in ENT1_null patients is particularly intriguing considering the numerous erythroid alterations, including macrocytosis, anisopoikilocytosis, and abnormal deformability. These data are even more surprising given that Ent1^−/− mice exhibit a reduced RBC count, Hb, and hematocrit. Interestingly though, red cells in Ent1^−/− mice are not destroyed and ENT1_null patients exhibit normal hemolysis markers. These data suggest that the lifespan of murine and human ENT1_null RBCs is not shortened (supplemental Table 1). In this regard, it is notable that neither S-adenosyl-methionine nor nucleotides are significantly altered in mature murine Ent1^−/− red cells.⁶³  An attractive hypothesis to explain differences in the role of Ent1 in murine vs human red cell differentiation may be related to the hypoxia-mediated up regulation of key metabolites, including nucleotides, within the low-oxygen BM microenvironment.^54,64-66 

The hypoxic microenvironment of the BM niche influences the metabolic pathways used by HSCs and consequently their erythroid differentiation. In line with these findings, we have recently shown that ENT1 and adenosine signaling are involved in the adaptation to high-altitude hypoxia.⁴⁵  Thus, molecules and signaling networks that are yet to be identified may complement ENT1 function in the murine spleen where Ter119⁺ cells were increased as compared with bone marrow where erythroid differentiation was negatively affected. Furthermore, although the specific ablation of Ent1 in erythroblasts, using an EpoR-Cre, also resulted in macrocytosis,⁶⁷  RBC numbers as well as other RBC parameters were less affected than those reported here using a complete Ent1 knockout. These data therefore raise the important possibility that ENT1 plays a role in erythroid differentiation at a progenitor stage upstream of EpoR⁺ erythroblasts.

Our findings reveal a critical role of ABCC4 in regulating erythroid differentiation in the absence of ENT1; moreover, we find that inhibition of ABCC4 enhances human and murine erythroid differentiation under physiological conditions. Notably, we identified ABCC4 as a crucial factor by documenting its role in explaining the differences in the ex vivo erythroid differentiation potential of P1 and P2 progenitors compared with P3 progenitors. Mechanistically, we found that P1/P2 RBCs exhibited high levels of intracellular cyclic nucleotides, whereas P3 RBCs are characterized by low cyclic nucleotide levels. This significant difference was associated with a heterozygous loss-of-function variant in the ABCC4 cyclic nucleotide exporter in P1/P2, but not in P3. The finding that ABCC4 inhibitor significantly rescued the anemic phenotype and increased erythroid differentiation in Ent1^−/− mice strongly suggests a crosstalk between ABCC4 and ENT1 in regulating erythropoiesis. Notably, though, ABCC4 inhibition did not correct the macrocytic phenotype of these Ent1^−/− mice, indicating that in our murine model, the macrocytic red cell phenotype is ABCC4-independent. Collectively, these data indicate that the intracellular level of cyclic nucleotides is crucial for in vivo erythropoiesis and highlight the interdependence of the ABCC4 and ENT1 transporters in erythropoiesis.

In summary, we show that ENT1 plays a key role in nucleotide homeostasis and optimal erythropoiesis and provide both genetic and pharmacological evidence that ABCC4 is a key negative regulator of erythropoiesis in the absence of ENT1, exacerbating a perturbation in cyclic nucleotide homeostasis and downstream erythroid defects. Moreover, we find that inhibition of ABCC4 by MK-571 results in a rescue of the anemic phenotype in Ent1^−/− mice, and an enhanced RBC production in WT mice without a significant reduction in circulating white blood cells. Notably, MK-571 has been tested in clinical trials as a therapeutic molecule to treat asthma,^68-70  pulmonary arterial hypertension,⁷¹  and platelet aggregation^55,72,73  via increasing the intracellular levels of cAMP and cGMP. The impact of MK-571 in pulmonary arterial hypertension was specific to its effects on ABCC4 because deletion of this gene resulted in a notable reduction in hypoxia-induced inflammatory response.⁷¹  MK-571 could therefore represent a novel drug for the therapeutic treatment of anemia.

The authors are indebted to Bérengère Koehl for helpful comments and for reading the manuscript, and to Cédric Vrignaud and Romain Duval for their assistance at CNRGS in this research project. The authors thank Sara El Hoss, Sandrine Laurence, and Niloofar Reihani for their help with methylcellulose cultures, and Martina Moras and Marie Daumur for their assistance.

This work was supported by INSERM, the Institut National de la Transfusion Sanguine, and the Laboratory of Excellence GR-Ex (reference ANR-11-LABX-0051). The Labex GR-Ex is funded by the IdEx program “Investissements d’Avenir” of the French National Research Agency (reference ANR-18-IDEX-0001). This work is also supported by grants from the Fondation Recherche Médicale (N.T.), National Institutes of Health (NIH) National Institute of Diabetes and Digestive and Kidney Diseases (PO1 DK032094) (M.N., S.K., and N.T.), NIH National Heart, Lung, and Blood Institute (HL137990 and HL136969) (Y.X.), and the McGovern Fund at the University of Texas Health Science Center-McGovern Medical School (Y.X.). M.M. was funded by the Ministère de l’Enseignement Supérieur et de la Recherche (Ecole Doctorale BioSPC) and the Institut National de la Transfusion Sanguine. P.G.-M. was funded by the CLARIN-COFUND program from the Principado de Asturias and the European Union.

Contribution: C.L.V.K., Y.C., O.H., M. Sitbon, N.T., T.P., and S.A. conceived the project and obtained funding; M.M., S.K., N.T., and S.A. designed the research study; Y.X. oversaw the design of the in vivo mouse studies; T.P. collected human samples and clinical data of ENT1_null patients; M.M., P.G.-M., and S.K. performed most of the experiments; X.C. conducted the in vivo mouse studies; A.-C.B. and S.S. realized the metabolomic analyses; C.C. and I.C.G. performed the phosphoproteomic studies; Y.Z. performed mouse blood smear experiments; M.M., P.G.-M., X.C., M. Serra, A.K.D., S.K., N.M., T.P., Y.X., N.T., and S.A. analyzed the data; and M.M., N.M., N.T., and S.A. wrote the manuscript with editing from P.G.-M., S.K., M. Sitbon, X.C., C.L.V.K., T.P., and Y.X.

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

Correspondence: Slim Azouzi, INSERM, UMR S1134, INTS, 6 Rue Alexandre Cabanel, 75015 Paris, France; e-mail: slim.azouzi@inserm.fr; Naomi Taylor, Pediatric Oncology Branch, NCI, CCR, NIH, Bethesda, MD 20814; e-mail: taylorn4@mail.nih.gov; Thierry Peyrard, INSERM, UMR S1134, INTS, 6 Rue Alexandre Cabanel, 75015 Paris, France; e-mail: tpeyrard@ints.fr; and Yang Xia, Department of Biochemistry and Molecular Biology, University of Texas McGovern Medical School at Houston, Houston, TX 77030; e-mail: yang.xia@uth.tmc.edu.