Arginine metabolism regulates human erythroid differentiation through hypusination of eIF5A

Hematopoietic stem cells (HSCs) possess 2 fundamental characteristics: the capacity for self-renewal and the sustained production of all blood cell lineages. The differentiation of HSCs to the erythroid lineage is distinctive in that progressive mitoses result in the production of daughter cells that differ, morphologically and functionally, from their parent cell, with terminal erythroid differentiation generating enucleated reticulocytes.

The maturation of erythroid precursors is regulated by transcription factors, cell-cell contacts, oxygen sensors, and cytokines, most notably, erythropoietin (EPO).^1-6 Notably though, erythropoiesis is also a formidable process at a metabolic level. The production of 200 × 10⁷ red cells per day is highly dependent on iron, glucose, fatty acid, and amino acid metabolism. Amino acids contribute to multiple aspects of erythroblast metabolism, ranging from precursors of nucleic acids, mTOR sensing, production of tricarboxylic acid cycle intermediates, and maintenance of intracellular redox, among others.^7-15 Arginine, a semiessential amino acid,¹⁶ affects erythroid progenitors^17,18 but its function in erythroid differentiation is unknown.

Arginine catabolism results in the generation of a diverse range of products, including nitric oxide, urea, creatine, proline, glutamate, agmatine, and polyamines, including spermidine. Of note, the spermidine polyamine is required for hypusination of eukaryotic translation initiation factor 5A (eIF5A). After hypusination, eIF5A promotes translation elongation as well as translation termination.^19-21 Here, we show that the spermidine-dependent hypusination of eIF5A is required for erythroid lineage commitment as well as for terminal erythroid differentiation. The critical nature of protein synthesis in erythropoiesis is highlighted by the finding that mutations in ribosomal protein (RP) genes underlie the hypoplastic anemia associated with Diamond-Blackfan anemia (DBA) and myelodysplastic syndromes (MDSs) with chromosome 5q deletions [del(5q)].^22,23 In light of our results showing the importance of eIF5A-stimulated translation in erythroid differentiation, we evaluated the impact of RP haploinsufficiency on hypusination. Strikingly, our data identify a novel link between RPs and hypusination in the regulation of erythropoiesis: haploinsufficiency of RPs significantly decreased hypusinated eIF5A (eIF5A^H) in erythroid progenitors. Thus, our study reveals a mechanism via which RPs regulate hypusination, thereby governing erythroid differentiation.

All studies involving human samples were conducted in accordance with the declaration of Helsinki. CD34⁺ cells were isolated after informed consent and approval by the respective institutional review boards. Samples from patients with and without MDS were obtained from the HemoDiag prospective cohort of patients with hemopathy (#NCT02134574, Centre Hospitalier Universitaire de Montpellier). CD34⁺ cells were differentiated as detailed in supplemental Methods, which is available on the Blood website.

Lentiviral pLKO.1 plasmids were generated harboring short hairpin RNAs (shRNAs) directed against SLC7A1, deoxyhypusine synthase (DHPS), and RPL11,^24,25 as described in the supplemental Methods.

Expression of surface markers was monitored using the appropriate fluorochrome-conjugated antibodies (supplemental Table 2) and detailed in the supplemental Methods.^1,5 

Mitochondrial biomass and activity were monitored by flow cytometry; mitochondrial and glycolysis stress tests were performed on an XFe-96 Analyzer (Seahorse Bioscience, Agilent); and glutamine and glucose uptake were evaluated using radioactively labeled substrates, as detailed in the supplemental Methods.

Protein synthesis rate was monitored by the Click-iT Plus OPP Protein Synthesis Assay, as described by the manufacturer (Molecular Probes).

After transfer of protein gels to polyvinylidene fluoride, membranes were incubated with 5% nonfat dry milk, incubated with the indicated antibodies, and followed by peroxidase anti-rabbit or anti-mouse immunoglobulin, as described in the supplemental Methods.

EPO-stimulated CD34⁺ cells were treated with N1-guanyl-1,7-diaminoheptane (GC7) (5 μM) for 48 hours. Cells from 3 donors were washed, and dry pellets were conserved at −80 °C until extraction (supplemental Methods).

The availability of intracellular arginine is regulated in large part by the cell-surface expression of the SLC7 cationic amino acid transporter (CAT) family.^26,27 Erythroid differentiation of human CD34⁺ hematopoietic stem and progenitor cells (HSPCs), induced by recombinant EPO (supplemental Figure 1A-C), results in a significant upregulation of SLC7A1 (P < .05) (Figure 1A), as well as arginine uptake (P < .05) (Figure 1B; supplemental Figure 1D). Interestingly, this increase in arginine transport does not represent a global augmentation in amino acid uptake; rEPO stimulation did not augment transport of glutamine (supplemental Figure 1D), an amino acid critical for erythroid development.^10,11 Cell-surface SLC7A1 levels then decreased during terminal erythroid differentiation (P < .001) (Figure 1C), associated with a fivefold reduction in arginine uptake from days 4 to 10 of differentiation (P < .01) (Figure 1D). During differentiation, arginine was transported into these cells by SLC7A1/CAT1 and not y+LAT/4F2hc;²⁸ N-ethylmaleimide, a pharmacological agent that blocks CAT1 but not y+LAT,²⁹ inhibited arginine uptake to levels detected in red blood cells (P < .0001) (supplemental Figure 1E).

To assess whether SLC7A1-mediated arginine metabolism plays a key role in erythropoiesis, we downregulated SLC7A1 expression in cytokine-stimulated progenitors with a pre-erythroid progenitor phenotype,³ characterized as CD34⁺IL3Rα^LoCD36⁻ (supplemental Figure 1F). Progenitors transduced with an SLC7A1 shRNA lentiviral vector, also harboring the green fluorescent protein (GFP) reporter gene, exhibited significantly reduced SLC7A1 transcripts (P < .05) (supplemental Figure 1G) and cell-surface SLC7A1 levels (P < .001) (Figure 1E), compared with HSPCs transduced with a control (shCTRL) vector. SLC7A1 downregulation decreased expansion by twofold (P < .0001) (supplemental Figure 1H) but more importantly, these cells were negatively selected; shSLC7A1-transduced progenitors (GFP⁺) decreased from 54% to 9% in a representative experiment (P < .001) (Figure 1G). Importantly, these progenitors were not negatively selected after expansion in nonerythroid conditions (supplemental Figure 1I). Consistent with an important function for SLC7A1 in erythroid differentiation, its downregulation resulted in the reduction of the IL3R⁻GlyA⁻CD34⁻CD36⁺ colony-forming unit (CFU)–enriched subset (P < .01) (Figure 1H) and decreased the expression of GlyA, CD36, and CD71 erythroid markers (P < .0001) (Figure 1I; supplemental Figure 1J). Notably though, these EPO-signaled HSPCs maintained their myeloid lineage potential, with more than twofold increase in the differentiation of CD11b⁺ myeloid cells (P < .01) (Figure 1I).

We next evaluated the impact of SLC7A1 downregulation on the erythroid specification of primitive CD34⁺CD38⁻ progenitors. After transduction with an shCTRL or shSLC7A1/CAT1 vector, fluorescence-activated cell sorter–sorted GFP⁺CD34⁺CD38⁻ bone marrow (BM) progenitors (supplemental Figure 2A) were evaluated in a single-cell StemMACS HSC-CFU Assay Kit (Miltenyi). Assessment of burst-forming unit erythroid progenitors revealed a more than threefold decrease in the number of colonies generated after SLC7A1/CAT1 downregulation (P < .05) (Figure 1J; supplemental Figure 2B). However, the generation of CFU–granulocyte-macrophage, albeit lower, was not significantly attenuated as compared with control conditions (Figure 1J). Furthermore, GFP⁺CD34⁺CD38⁻ BM progenitors evaluated in EPO-stimulated liquid culture conditions exhibited attenuated induction of GlyA and only minimal expansion of shCAT1-transduced cells after 7 days (compared with a ninefold expansion of shCTRL-transduced cells, supplemental Figure 2C). These data correlated with a reduced growth of single-sorted GFP⁺CD34⁺CD38⁻ cord blood progenitors after shSLC7A1/CAT1 transduction, with growth in 24 of 177 wells as compared with 81 of 177 wells, respectively. Micrographs of wells seeded with 100 sorted GFP⁺CD34⁺CD38⁻ shCTRL-transduced cord blood progenitors showed a larger area than after shSLC7A1/CAT1 transduction and a loss of transgene (shSLC7A1 with GFP) in GlyA⁺ erythroblasts generated from 2 of the 3 wells, attesting to the importance of this transporter (supplemental Figure 2D). Together, these data strongly suggest that SLC7A1 plays a nonredundant role in regulating the erythroid differentiation of human HSPCs.

To specifically evaluate the role of arginine in the commitment of HSPCs to an erythroid lineage fate, progenitors were stimulated with rEPO in media depleted of arginine. In arginine-depleted media, a significantly lower percentage of progenitors adopted a CFU-E fate (P < .0001) (Figure 2A), and these cells exhibited a significantly decreased proliferation capacity (supplemental Figure 3A). Moreover, under conditions of arginine depletion, burst-forming unit erythroid progenitors–enriched CD34⁺CD36⁻ cells (supplemental Figure 3B) exhibited a reduced differentiation toward a CD34⁻CD36⁺ CFU-E stage (supplemental Figure 3B). In agreement with the SLC7A1 knockdown experiments (Figure 1), arginine depletion inhibited erythroid differentiation at both days 3 and 7 of rEPO stimulation (P < .05) (Figure 2B; supplemental Figure 3C). These data highlight the critical role of arginine in promoting the commitment of progenitors to an erythroid fate.

Because SLC7A1 levels decreased during terminal erythropoiesis (Figure 1C), we assessed whether arginine was required during this stage of differentiation. When arginine was depleted from erythroblasts at day 3 of differentiation, all cells upregulated GlyA by day 7 (P < .05) (Figure 2C), albeit with a decrease in cell expansion (P < .001) (supplemental Figure 3D). Notably though, terminal differentiation as well as enucleation was markedly inhibited (P < .01) (Figure 2D). These data were somewhat surprising given the loss in cell-surface SLC7A1 levels upon terminal differentiation (Figure 1C). We therefore evaluated the possibility that intracellular arginine was regulated by arginase expression. Importantly, Arg2 was upregulated by 12-fold during erythroid differentiation and remained elevated in late-stage erythroblasts (P < .01) (Figure 2E), whereas Arg1 was upregulated at very late-stage terminal differentiation (supplemental Figure 3E). Thus, the strong induction of these arginases may be associated with the continued requirement for extracellular arginine throughout the erythroid differentiation process.

Arginine is a precursor of numerous metabolic intermediates that are generated through distinct catabolic pathways (supplemental Figure 3F). Because neither arginine-derived synthesis of creatine nor nitric oxide were required for erythropoiesis (supplemental Figure 3G), we evaluated the role of arginine as a precursor of polyamines (Figure 3A). Notably, polyamine biosynthesis was strictly required for erythroid differentiation: inhibition of polyamine biosynthesis, via the targeting of ornithine carboxylase or spermidine synthase with α-difluoromethylornithine (DFMO) and trans-4-methyl cyclohexylamine (MCHA), respectively, inhibited expansion and abrogated erythroid differentiation (P < .001 to P < .0001) (Figure 3B-C; supplemental Figure 4A-C). In accord with the arginine deprivation experiments, progenitors treated with either DFMO or MCHA maintained their potential to differentiate to a myeloid cell fate, with up to 80% CD11b⁺ cells (Figure 3B-C). Similarly, activation of SAT1-mediated polyamine catabolism by N1, N11-diethylnorspermine decreased GlyA (P < .0001) while supporting CD11b induction (P < .05) (Figure 3D). Spermidine, and not spermine, was specifically required for erythropoiesis, as inhibiting spermine synthase with N-(3-amino-propyl)cyclohexylamine did not negatively affect erythroid differentiation (Figure 3E). Furthermore, in contrast with the requirement of arginine for both erythroid commitment and terminal erythroid differentiation, polyamine biosynthesis was a prerequisite for erythroid commitment but not for terminal erythroid differentiation. DFMO inhibited the generation of erythroblasts but did not modulate either erythroblast differentiation or enucleation (supplemental Figure 4D-E). Furthermore, enucleation was actually inhibited by spermidine (supplemental Figure 4F), likely because of spermidine’s potential to support mitochondrial respiration, a metabolic state that negatively affects enucleation.^14,30 

To assure that the impact of pharmacological inhibitors was directly coupled to polyamine biosynthesis, we performed rescue experiments and found that ectopic putrescine and spermidine restored the DFMO- and MCHA-mediated attenuation of erythroid differentiation (P < .0001) (Figure 3F; supplemental Figure 4G). Interestingly, under control conditions, erythroid differentiation was not dependent on polyamine transport from extracellular stores (Figure 3G). However, after DFMO-mediated abrogation of polyamine biosynthesis, treatment with the AMXT-1501 polyamine transporter inhibitor³¹ abrogated the potential of spermidine to rescue erythropoiesis (Figure 3G; supplemental Figure 4H). These data point to an evolutionary redundancy; intracellular polyamine synthesis is sufficient for erythroid differentiation under physiological conditions, but under stress conditions where intracellular polyamine levels are limiting, the transport of extracellular polyamines promotes erythropoiesis.

The pleiotropic nature of polyamines has made it difficult to define their specific role(s).³² Notably though, spermidine serves as a precursor for the generation of the natural amino acid hypusine [N^ε-4-amino-2-hydroxybutyl(lysine)], an essential posttranslational modification of eIF5A. Hypusine is formed by the conjugation of the aminobutyl moiety of spermidine to lysine-50 of human eIF5A.³³ Remarkably, eIF5A is the only protein in which this modification is known to occur, and its presence is required for its function, promoting translation elongation as well as translation termination.^19-21 The synthesis of hypusine is catalyzed through sequential enzymatic steps involving DHPS and deoxyhypusine hydroxylase (DOHH) (Figure 4A).

The addition of the spermidine analog GC7, one of the most potent inhibitors of DHPS³⁴ (Figure 4A), blocked the EPO-induced generation of CFU-E (P < .0001) (Figure 4B). Furthermore, myeloid differentiation occurred despite a decrease in progenitor expansion in this condition (P < .0001) (supplemental Figure 5A). CD11b⁺ cells were increased by 30-fold by day 7 of EPO-induced differentiation (P < .05) (Figure 4C). This increase was accompanied by a massive loss of erythroid progenitors (P < .0001) (Figure 4C-D; supplemental Figure 5B-C). Furthermore, inhibition of DHPS activity at day 3 attenuated the differentiation of late-stage CD49d⁻ orthochromatic erythroblasts and enucleation (P < .01) (supplemental Figure 5D). A pharmacological inhibitor of DOHH, the iron chelator ciclopirox olamine (supplemental Figure 5E),³⁵ also significantly inhibited the induction of GlyA to levels <10% of that detected in control conditions (P < .001) (supplemental Figure 5F).

To directly evaluate hypusination, we targeted DHPS using an shRNA-mediated approach. shDHPS-transduced progenitors exhibited a 61% ± 3% decrease in DHPS messenger RNA (P < .01) (supplemental Figure 5G) and a decreased expansion (P < .0001) (supplemental Figure 5H). Moreover, progenitors with downregulated DHPS were massively counterselected during erythroid differentiation (P < .001) (Figure 4E). Similarly to the impact of GC7, shDHPS-transduced progenitors exhibited a decreased generation of CFU-E (P < .01) (Figure 4F) and a significantly attenuated upregulation of erythroid markers (P < .0001) (Figure 4G). The negative impact of DHPS downregulation was also restricted to erythroid-committed progenitors; in the absence of rEPO, shDHPS-transduced progenitors were not negatively selected (supplemental Figure 5I), and there was a threefold expansion of myeloid cells (P < .001) (Figure 4G). Thus, in a manner analogous to conditions of decreased arginine uptake, inhibition of DHPS-attenuated erythroid but not myeloid differentiation.

The data presented above strongly suggest a critical role for hypusination in erythroid differentiation. However, it is notable that eIF5A^H has not previously been evaluated as a function of erythroid differentiation. We found that eIF5A levels increased within 4 days of EPO stimulation (P < .01) (Figure 5A) and then dropped during terminal differentiation between polychromatic and orthochromatic erythroblasts (P < .05) (Figure 5B; supplemental Figure 6A).

The potential impact of hypusination on protein synthesis was evaluated as a function of the incorporation of the alkyne analog of puromycin, O-propargyl-puromycin, forming covalent conjugates with nascent polypeptide chains.^36,37 Within 1 day of EPO stimulation, O-propargyl-puromycin incorporation was significantly increased, indicating an augmented level of protein synthesis (P < .05) (supplemental Figure 6B), and then decreased during terminal differentiation (Figure 5C-D). Interestingly, the dramatic drop in protein synthesis occurred earlier than the drop in hypusination, between the basophilic and polychromatic erythroblast stages (P < .001) (Figure 5D). These data correlate with previous findings showing that total protein levels also drop between the basophilic and polychromatic stages of differentiation.³⁸ 

To assess whether hypusination in hematopoietic progenitors is directly regulated by arginine catabolism, we evaluated hypusination after arginine deprivation and polyamine biosynthesis inhibition. Both these conditions dramatically reduced hypusination without having a significant impact on eIF5A levels themselves (supplemental Figure 6C-D). Furthermore, polyamine biosynthesis was directly required for hypusination, and putrescine and spermidine rescued the DFMO-mediated inhibition of hypusination (P < .001) (supplemental Figure 6D). Hypusination also directly affected the activation state of the mTOR pathway, integrating environmental clues into cellular responses. Phosphorylation of RPS6, a downstream reporter of mTOR activity, was inhibited by GC7 and DFMO to levels comparable with those detected after arginine deprivation (supplemental Figure 6E). Thus, hypusination in hematopoietic progenitors is strictly dependent on arginine-derived polyamine biosynthesis and regulates mTOR signaling cascades.

As expected from these results, inhibition of DHPS, by either GC7 or an shDHPS, resulted in a 75% loss of hypusine as compared with that in control progenitors (P < .0001 and P < .01) (Figure 5E-F). Moreover, these modulations resulted in functional changes in protein synthesis. Nascent protein synthesis in GC7-treated progenitors decreased to levels detected in cells treated with a global inhibitor of protein synthesis (P < .0001) (Figure 5G-H).

To specifically evaluate the proteins that were regulated by hypusinated eIF5A in erythroid progenitors, we analyzed progenitors differentiated for 48 hours in the absence or presence of GC7 by quantitative liquid chromatography tandem mass spectrometry–based proteomics using tandem mass tags.³⁹ Notably, and in agreement with the impact of GC7 on erythroid differentiation, some of the most highly downregulated proteins were hemoglobin subunits and erythroid spectrin chains (Figure 5I). Furthermore, DHPS was significantly upregulated after GC7 treatment, pointing to a potential compensatory feedback loop.

Gene set enrichment analysis and protein-protein interaction network clustering analysis revealed a massive loss of proteins involved in the translation of mitochondrial proteins (Figure 5J; supplemental Figure 7A). Our proteomics analyses identified 824 of 1158 mitochondrial proteins; after attenuation of eIF5A^H, 620 proteins were downregulated and 204 were upregulated (supplemental Figure 7B), with a dramatic decrease in proteins involved in oxidative phosphorylation (supplemental Figure 7C). Importantly, eIF5A functions in cytoplasmic, not mitochondrial, protein synthesis, and thus, the impact on mitochondrial proteins may be either direct or indirect.⁴⁰ Although the mechanisms regulating the massive impact of hypusination on mitochondrial proteins is not known, this effect has been detected in multiple other cell systems, including renal cells,⁴¹ brain,⁴² hepatic cells,⁴³ cardiomyocytes,⁴⁴ macrophages,⁴⁵ and T cells.⁴⁶ In the context of red cells, these data are of much interest given previous work showing the importance of mitochondrial reactive oxygen species in erythroid differentiation.^14,30,47 Indeed, we found that the specific inhibition of mitochondrial protein synthesis with chloramphenicol significantly decreased erythroid progenitor differentiation (P < .001), whereas CD11b⁺ myeloid cells continued to be generated (P < .05) (supplemental Figure 7D). Thus, although there is a large literature reporting on the importance of protein synthesis during erythropoiesis,^22,48,49 our results highlight the critical role of cytoplasmic hypusine-dependent protein synthesis in the mitochondrial translation apparatus required for erythroid differentiation.

To decipher the impact of hypusination on the synthesis of proteins involved in mitochondrial metabolism, we first evaluated the expression of mitochondrial complex enzymes. Interestingly, complexes II, III, and V were all upregulated by EPO-induced differentiation (P < .05) (supplemental Figure 8A). Notably though, both GC7 and DHPS downregulation resulted in decreased expression of all 5 mitochondrial complex enzymes, whereas actin levels remained stable (Figure 6A). Thus, the expression levels of this subset of mitochondrial proteins, synthesized in the cytoplasm, are distinctively sensitive to eIF5A^H levels.

Based on the importance of eIF5A^H for the synthesis of mitochondrial proteins, we studied mitochondrial metabolism in these cells. Although GC7 treatment and DHPS downregulation both diminished mitochondrial biomass and polarization potential (P < .01) (supplemental Figure 8B-C), decreases were <1.2-fold. However, the impact on metabolism was much more striking; GC7 almost completely abrogated oxidative phosphorylation (OXPHOS) in these progenitors, and their glycolytic reserve was decreased (Figure 6B; supplemental Figure 8D). Similarly, the inhibition of polyamine biosynthesis by either DFMO or N1, N11-diethylnorspermine decreased basal respiration and spare respiratory capacity by 2.5- to 4.5-fold (supplemental Figure 8E). This loss of OXPHOS, albeit with maintained glycolysis, was also detected in progenitors with downregulated DHPS (Figure 6C; supplemental Figure 8F), again highlighting the critical role of hypusination for mitochondrial metabolism. Ciclopirox, an inhibitor of hypusination and mitochondrial respiration,⁵⁰ also significantly decreased basal and maximal oxygen consumption (P < .05 and P < .001) (supplemental Figure 8G).

The loss of mitochondrial metabolism under conditions where hypusination was negatively affected led us to test the hypothesis that drivers of the tricarboxylic acid cycle would augment erythroid differentiation. Succinate, by bypassing complex I, can potentially provide a fuel source for compromised mitochondria.⁵¹ Furthermore, our protein analyses revealed a significant loss of the complex II SDHB enzyme (Figure 6A), thereby attenuating the generation of succinate. Notably, the addition of succinate to GC7-treated EPO-induced progenitors significantly upregulated OXPHOS (P < .01) (Figure 6D), a response not induced by acetate or α-ketoglutarate (data not shown). Even more importantly, ectopic succinate promoted the erythroid lineage commitment of GC7-treated hematopoietic progenitors (P < .01) (Figure 6E). Together, these data reveal an interplay between hypusination and mitochondrial metabolism in regulating erythroid lineage commitment and differentiation.

Consistent with its role in translation and binding in the E site of the ribosome, eIF5A interacts with proteins associated with ribosome function, including RPL and RPS proteins.⁵² Of note, mutations resulting in impaired ribosome biogenesis, encompassing >20 RPs have been found to result in DBA. This rare congenital disease predominantly affects erythroid lineage cells and is characterized by macrocytic anemia and BM failure.⁵³ Furthermore, haploinsufficiency of the RPS14 gene results in macrocytic anemia in del(5q) in patients with MDS.⁵⁴ However, the potential association between RP function and eIF5A^H has not been previously evaluated. Notably, we found that eIF5A^H was significantly lower in immortalized haploinsufficient Rps14^+/− progenitors⁵⁵ than in control Mx1Cre⁺ progenitors (P < .01) (supplemental Figure 9A). Moreover, basal oxygen consumption as well as spare respiratory capacity were significantly lower in Rps14^+/−Mx1Cre⁺ than in Mx1Cre⁺ progenitors (P < .01) (supplemental Figure 9B), thereby linking RPS14 expression to hypusination and mitochondrial metabolism.

We further assessed hypusination in CD34⁺ progenitors from del(5q) in patients with MDS. Similar to Rps14-haploinsufficient murine progenitors, progenitors from 3 patients with del(5q)-MDS (supplemental Table 1) exhibited high expression of eIF5A, but hypusination levels were significantly reduced (P < .05) (Figure 7A). Moreover, after rEPO-induced differentiation, eIF5A^H decreased in del(5q)-MDS progenitors, accompanied by ineffective erythroid differentiation (Figure 7A; supplemental Figure 9C). This defective hypusination was specific to EPO-stimulated CD34⁺ progenitors in 2 patients with MDS, where we were able to evaluate hypusination in another cell lineage, activated T cells, which exhibited a more than twofold increase in hypusination (supplemental Figure 9D). Thus, in both murine and human progenitors, the ineffective erythropoiesis linked to RPS14 haploinsufficiency is associated with a loss of hypusinated eIF5A.

To further evaluate the link between RPs and hypusination, we studied eIF5A^H in CD34⁺ progenitors from 2 patients with DBA with RPL11 haploinsufficiency (supplemental Table 1). For 1 patient with DBA, where sufficient numbers of freshly isolated BM CD34⁺ progenitors were available for evaluation, neither eIF5A nor hypusination were attenuated (Figure 7B). Notably though, after EPO-induced stimulation, hypusination decreased by threefold relative to control CD34⁺ progenitors, and in a second RPL11-haploinsufficient patient, hypusination as well as eIF5A levels in EPO-stimulated peripheral CD34⁺ progenitors were more than sixfold lower than those in control (Figure 7B). To assess whether this phenotype would be recapitulated by directly downregulating RPL11, an shRNA against this target was introduced into healthy CD34⁺ progenitors. As expected, RPL11 downregulation (supplemental Figure 9E) decreased erythroid differentiation, whereas myeloid cells continued to be generated (Figure 7C; supplemental Figure 9F).^24,25 Strikingly, the shRNA-mediated downregulation of RPL11 significantly decreased hypusination, despite the maintenance of eIF5A levels (P < .01) (Figure 7D). Moreover, the attenuated hypusination in these progenitors directly correlated with a loss in mitochondrial metabolism (Figure 7E; supplemental Figure 9G). Thus, these data reveal a novel role for RPs in controlling eIF5A^H in erythroid progenitors, resulting in a regulated coordination of mitochondrial translation and metabolism with subsequent erythroid differentiation.

Translation is a critical component in the regulation of multiple developmental pathways, cell activation, and cancer progression.^56-59 Within the broad realm of protein synthesis, hypusinated eIF5A drives translational programs affecting cancer, diabetes, and infectious diseases.^20,60-62 Notably though, the impact of eIF5A in erythropoiesis, a process that is particularly sensitive to translational regulation, has not been evaluated. The data that we present here show that optimal protein synthesis in erythroid progenitors is dependent on SLC7A1-mediated arginine uptake, resulting in the hypusination of eIF5A by the arginine-catabolized spermidine moiety. Under conditions where hypusination is suboptimal, hematopoietic progenitors are not capable of undergoing erythroid differentiation but maintain their myeloid lineage fate. Thus, our study reveals a direct link between arginine metabolism and eIF5A-induced protein synthesis in erythroid-myeloid lineage commitment. Furthermore, we uncover a critical role for eIF5A activity in the defective erythropoiesis associated with ribosomopathies.

Hypusinated eIF5A has been identified as required for the translation elongation process rather than the initiation process,^63,64 especially important for facilitating the translation of proteins containing polyproline residues.^65-67 Notably, eIF5A is the only protein in eukaryotes and archaea to contain hypusine.^19,64,68 The downstream effectors via which hypusinated eIF5A regulates cell differentiation are still being elucidated, but recent studies show the importance of hypusinated eIF5A in mitochondrial translation.^45,46 The mechanisms via which hypusinated eIF5A stimulates translation of messenger RNAs encoding mitochondrial RPs have not yet been elucidated, but it is intriguing to speculate that in addition to a direct effect, it may modulate expression of chaperones involved in the assembly of mitochondrial proteins. Moreover, eIF5A may interact directly with mitofusin, a protein that is critical in mitochondrial clustering and function.⁴⁰ Irrespective of the mechanism(s), eIF5A-induced mitochondrial programming reverses B-cell senescence,⁶⁹ promotes vaccine immunogenicity,⁷⁰ and protects the brain from premature aging,^42,71 among others.

The findings that we present here demonstrate the critical nature of eIF5A-dependent mitochondrial function in regulating the erythroid commitment of HSPCs. Erythroid differentiation requires mitochondrial function^{11,14,47,72,73} and we now find that in the absence of eIF5A activity, mitochondrial complex enzymes are massively decreased, coupled with an almost complete loss of oxygen consumption. Thus, eIF5A-dependent OXPHOS is a sine qua non for erythroid differentiation. Furthermore, the abrogated erythroid commitment of DHPS-attenuated progenitors can be partially rescued by succinate. Interestingly, although succinate can increase oxygen consumption, its accumulation can modulate cell physiology through succinylation of substrates,⁷⁴ as well as by generating a condition of “pseudohypoxia” secondary because of stabilization of HIF-α prolyl hydroxylase.⁷⁵ 

The rapid initiation of translation allows erythroid progenitors to immediately respond to environmental cues, stimulating erythroid differentiation programs.^49,76 It is therefore not surprising that multiple translation initiation factors, such as eIF4G, eIF2α, and EIF4EBP1, have been found to regulate erythropoiesis.^77-81 Moreover, haploinsufficiency of RPs affects the translation of specific transcripts,^22,23,82-84 resulting in anemia in patients with DBA and del(5q) MDS.^85,86 The mechanisms via which RP deficiency regulates hypusine levels are not known, but one hypothesis is that RP-mediated changes in DHPS, DOHH, spermidine, and/or the substrates and cofactors mediating their activities, including oxygen concentration,⁸⁷ can lead to reduced hypusine levels. Furthermore, eIF2 plays an important role in erythroid differentiation,⁷⁹ and as colliding ribosomes have recently been shown to trigger eIF2 phosphorylation,⁸⁸ impaired eIF5A function may cause translating ribosomes to collide on poly-Pro motifs, contributing to the pathophysiology of DBA. More generally, variations in hypusination as a function of cell type/activation highlight the intricate regulation of metabolic and translational pathways. Considering our data, it is opportune to assess the impact of ribosomal availability on eIF5A-dependent translation, bringing a novel perspective to our understanding of the defective protein synthesis characterizing ribosomopathies.

The link that we uncover between RPs and eIF5A activity opens new therapeutic perspectives for patients with ribosomopathies. Notably, L-leucine has recently been found to increase protein synthesis in both DBA and del(5q) MDS erythroblasts, resulting in improved erythroid differentiation^23,89 as well as an erythroid response in ∼15% of patients with DBA.⁹⁰ We hypothesize that the erythroid protein subsets regulated by the leucine-mediated induction of the eIF4F complex⁹¹ may differ from those that are dependent on hypusinated eIF5A. It will be important to assess whether the defective hypusination that we detected in RPS14- and RPL11-haploinsufficient progenitors is also a characteristic of patients with RPS19^+/− DBA. eIF5A^H can potentially be increased by arginine supplementation or more directly by spermidine itself. Spermidine may have a greater impact than arginine when eIF5A activity is limiting; spermidine exhibited therapeutic efficacy in cognition and^42,92,93 reversed B-cell senescence⁶⁹; and improved vaccinal responses in the elderly.⁷⁰ In erythroid differentiation, results have been disparate,^94,95 potentially because of the presence of amine oxidases in fetal calf serum, which catalyze the generation of hyperoxide from spermidine.⁹⁶ Although a therapeutic strategy designed to deliver high doses of spermidine has not yet been established, a prospective cohort study found a significant correlation between high dietary spermidine intake and cognitive function.⁹³ The data that we present here support the clinical evaluation of combined nutrient supplementation in patients with ribosomopathies, with the goal of promoting eIF4F- as well as eIF5A-dependent protein synthesis.

The authors thank all members of their laboratories for discussions and scientific critique. The authors are grateful to Jean-Luc Veyrune and the Unité of Thérapie Cellulaire, as well as the Clémentville Clinic (Montpellier, France), for their generous efforts in providing access to cord blood samples, Sandra Manceau and Pierre Buffet of the GR-Ex for their precious assistance with the GR-Ex clinical protocol and patient collection, Isabelle Marie for her valued help with the Diamond-Blackfan anemia (DBA) registry, Thierry Leblanc and Flore Sicre De Fontbrune for their wonderful help with one of the patients with DBA, and Ludivine David Nguyen for her expertise in the molecular DBA diagnosis. The authors are thankful to Steven A. Carr and Karl Clauser for their insightful input on proteomics analyses.

This work was supported by generous funding from the National Institutes of Health (NIH), National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) (DK32094) (N.M., S.K., and N.T.), Laboratory of Excellence for Red Cells [(LABEX GR-Ex)-ANR Avenir-11-LABX-0005-02] (P.G.-M., M.S., L.D.C., S.K., and N.T.), NIH, NIDDK R01 DK060581 and R01 DK124906 (R.G.M.), EJPRD/ANR-19-RAR4-0016 and EJP RD COFUND-EJP 825575 (L.D.C.), NIH, National Heart, Lung, and Blood Institute (NHLBI) (HL144436 and HL152099) (L.B.), the FRM, ARC, the French national (ANR) research grants (NutriDiff), the French laboratory consortiums (LABEX) EpiGenMed and GR-Ex, and by intramural NIH funding from the NHLBI (C.E.D.), the Eunice Kennedy Shriver National Institute of Child Health and Human Development grant ZIA HD001010 (T.E.D.), and the National Cancer Institute's Center for Cancer Research (ZIA BC 011924) (N.T.). P.G.-M. was supported by a fellowship from the Clarin-COFUND EU Program (Principado de Asturias, Spain), I.P. was supported by a fellowship from EpiGenMed, and A.J. is supported by a fellowship from the French Ministry of Health.

Contribution: P.G.-M., S.K., and N.T. conceived the study; P.G.-M., I.P., M.E.O., A.J., J. Papoin, H.Y., J.G., J. Platon, S.W.S.K., K.L.M., R.G.M., M.P., T.K., M.B.-C., N.P.-S., M.S., T.E.D., N.M., L.D.C., N.D.U., L.B., S.K., and N.T. were involved in study design; K.L.M., S.L., D.J.Y., F.P., A.N., G.C., C.E.D., and L.D.C. participated in the analyses of patients, provided samples, and provided clinical information; P.G.-M., I.P., M.E.O., A.J., J. Papoin, H.Y., J.G., J. Platon, S.W.S.K., K.L.M., M.D., M.P., T.K., S.B., and M.B.-C. performed experiments; all authors participated in data analysis; P.G.-M., S.K., and N.T. wrote the initial manuscript with extensive edits from I.P., M.E.O., M.S., T.E.D., N.M., N.D.U., and L.B.; and K.L.M., V.D., and V.S.Z. provided important critical input.

Conflict-of-interest disclosure: M.S., S.K., and N.T. are inventors on patents describing the use of RBD ligands but N.T. no longer has any patent rights. M.S. is the cofounder of METAFORA-biosystems, a start-up company that focuses on metabolite transporters under physiological and pathological conditions. The remaining authors declare no competing financial interests.

Correspondence: Pedro Gonzalez-Menendez, Departamento de Morfología y Biología Celular, School of Medicine, Universidad de Oviedo, Julián Claveria 6, 33006 Oviedo, Spain; e-mail: gonzalezmpedro@uniovi.es; Sandrina Kinet, Institut de Génétique Moléculaire de Montpellier, CNRS, 1919 Route de Mende, 34090 Montpellier, France; e-mail: kinet@igmm.cnrs.fr; and Naomi Taylor, Pediatric Oncology Branch, National Cancer Institute, National Institutes of Health, Building 10, Room 1W-3940, 9000 Rockville Pike, Bethesda, MD 20892; e-mail: taylorn4@mail.nih.gov.