Metabolic profiling and perturbations of human CD34⁺ hematopoietic stem and progenitor cells to regulate HSPC function

Key Points

• •Metabolomic and transcriptomic profiling highlights Gln and AA as distinct developmental metabolic cues in human HSPCs.
• •Gln inhibition promoted HSPC quiescent state and multilineage engraftment, and AA enhanced megakaryocytic differentiation.

Visual Abstract

Abstract

Hematopoietic stem cells (HSCs) possess self-renewal and multilineage differentiation abilities to generate blood cells and sustain hematopoiesis. Recent studies indicate that HSC metabolism is crucial for regulating their function and cell fate determination in mammals. However, a comprehensive understanding of the metabolic landscape of human HSCs across distinct developmental stages remains lacking. In this study, we performed untargeted metabolomics of lin⁻CD34⁺ hematopoietic stem and progenitor cells (HSPCs) from human fetal liver, umbilical cord blood, and adult bone marrow, revealing different developmentally associated metabolic signatures that may shape HSPC function across ontogeny. Metabolomic analysis identified d-glutamine (Gln) and arachidonic acid (AA) as metabolites exhibiting distinct abundance during HSPC development, suggesting their potential roles in modulating HSPC function. Transcriptomic profiling after specific metabolic treatments further revealed distinct gene expression programs associated with lineage commitment, stemness maintenance, and metabolic regulation. Functional assays demonstrated that inhibition of Gln metabolism with DON (6-diazo-5-oxo-l-norleucine) induced HSPCs into quiescent cell states, improving the engraftment of HSPCs and myeloid differentiation. In contrast, exogenous AA supplementation in HSPC culture promoted proliferation and significantly enhanced megakaryocytic differentiation both in vitro and in vivo. Collectively, our study profiled the metabolic landscape of human HSPCs from embryonic to adult period through the newborn stages, suggesting that metabolic modulation could regulate HSPC function. These findings provide novel mechanistic insights and potential strategies for metabolite-based interventions to promote and enhance human HSPC function, with broad implications for basic research and regenerative medicine.

Introduction

Hematopoietic stem cells (HSCs), which sustain lifelong blood production through self-renewal and multilineage differentiation, emerge from hemogenic endothelial cells in the aorta-gonad-mesonephros region, expand in the fetal liver (FL), become predominantly quiescent in adult bone marrow (BM), and functionally decline with aging.¹ Under homeostatic conditions, most HSCs remain quiescent to preserve their self-renewal capability and resist oxidative stress or DNA damage, whereas only a small fraction actively replenishes blood cells.² Quiescent HSCs can rapidly enter the cell cycle, expand, and differentiate into lineage cells in response to physiological stress or injury. Once the restoration of homeostasis is achieved, activated HSCs can return to dormancy, thereby ensuring the long-term maintenance of the HSC pool.3, 4, 5 These dynamic transitions are tightly governed by multiple signaling pathways and microenvironmental factors.⁶

Recent advances in single-cell multiomics have enabled high-resolution dissection of hematopoietic ontogeny at the transcriptomic, metabolomic, and proteomic levels.7, 8, 9 It is now widely acknowledged that metabolic programs do not just mirror the cellular state, but can actively shape stem cell functional states and lineage potential.^10,11 During ontogeny, HSCs exhibit dynamic metabolic shifts that correspond to their functional transitions and the sensing of microenvironmental changes. Embryonic and fetal hematopoietic stem and progenitor cells (HSPCs) display heightened metabolic activity, including increased oxidative phosphorylation, fatty acid oxidation, and glutamine (Gln) use,¹² supporting rapid proliferation and lineage output of HSPCs. In contrast, adult BM (aBM) HSCs prefer to rely on anaerobic glycolysis and fatty acid metabolism to preserve their quiescence and long-term self-renewal capability.¹³ These metabolic adaptations are closely tied to the hematopoietic developmental stages, alterations in the microenvironment, and lineage output demands.

In this study, we profiled the metabolic landscapes of human lin⁻CD34⁺ HSPCs from FL, umbilical cord blood (UCB), and aBM using untargeted metabolomics. By integrating multivariate clustering and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment, we identified developmentally associated metabolic signatures and prioritized Gln and arachidonic acid (AA), which showed development-specific abundance. We further investigated the functional changes of human HSPCs after targeted intervention experiments using Gln inhibition and AA supplementation through in vitro and in vivo functional assays. Our findings underscore the role of metabolism-driven regulation in HSPC function, provide a metabolomic framework for understanding developmental hematopoiesis, and highlight potential avenues for metabolic modulation in regenerative therapies.

Materials and methods

Human samples and HSPC isolation

Human FL (9-10 postconception weeks, n = 7), UCB (n = 15), and aBM (n = 8) samples were obtained with informed consent under the institutional review board approval (the Fifth Medical Center of the Chinese PLA General Hospital, KY-2023-3-22-1). For untargeted metabolomics, 3 independent biological replicates were analyzed for each developmental stage (FL/UCB/aBM). aBM donors were healthy males aged 30 to 40 years, whereas UCB samples were obtained from full-term pregnancies without metabolic/hematologic disorders and FL specimens were confirmed to be free of known developmental abnormalities. HSPCs were processed within 6 hours of collection. Lin⁻CD34⁺ cells were enriched by magnetic-activated cell sorting and purified by fluorescence-activated cell sorting (Sony MA900).

Untargeted metabolomics

For each sample, 1 × 10⁵ cells were extracted and profiled by ultraperformance liquid chromatography–tandem mass spectrometry (UPLC–MS/MS) in positive/negative electrospray ionization (ESI⁺/ESI⁻) modes with pooled quality control samples for run stability. Data (.mzXML) were processed using XCMS, a software package for processing mass spectrometry-based metabolomics data, and features with >50% missing values or relative standard deviation >50% were removed. Principal component analysis/orthogonal partial least squares discriminant analysis were used to summarize variance. Differentially abundant metabolites were defined by variable importance in projection >1 and P < .05, followed by KEGG enrichment and fuzzy c-means clustering analyses.

HSPC culture and metabolite treatment

Lin⁻CD34⁺ HSPCs derived from UCB, FL, and aBM were cultured in serum-free medium supplemented with stem cell factor, FMS-like tyrosine kinase 3 ligand, and thrombopoietin and treated for 48 hours with vehicle control, DON (6-diazo-5-oxo-l-norleucine; 2.5μM), or AA (50μM). Untreated cultures served as controls.

Xenotransplantation assay in NOG-W41 mice

UCB-derived CD34⁺ HSPCs were pretreated ex vivo with DON or AA for 48 hours and then transplanted into female immunodeficient NOD.Cg-Prkdc^scid Il2rg^tm1SugKit^{em1(V831M)Jic}/JicCrl (NOG-W41) recipient mice, aged 8 to 12 weeks, without irradiation. Human engraftment and lineage output were subsequently assessed in peripheral blood (PB) and hematopoietic tissues at the indicated time points (see the supplemental Methods). All animal studies were conducted in accordance with the guidelines of the Fifth Medical Centre, Chinese People’s Liberation Army General Hospital (IACUC-2023-0022).

RNA sequencing (RNA-seq) and analysis

TruSeq stranded mRNA libraries (Illumina) were sequenced using NovaSeq 6000 (PE150), aligned to GRCh38 using Hierarchical Indexing for Spliced Alignment of Transcripts 2, quantified with StringTie, and analyzed using DESeq2, a software package for differential expression analysis of RNA-seq count data (adjusted P < .05, |log₂FC| > .5). Gene Ontology/KEGG and Gene set enrichment analysis were performed using clusterProfiler (MSigDB v7.4).

Statistical analysis

All experiments were repeated at least 3 times, independently. For all experiments, analysis between the 2 groups were performed using a standard 2-tailed Student t test. Comparisons among >3 groups were analyzed using 1-way analysis of variance followed by Tukey’s multiple comparison test. A value of P < .05 was considered statistically significant. Data with statistical significance (∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001) are shown in figures. Detailed materials and methods are provided in the supplemental Methods.

Ethics approval and consent to participate

Collection of all human samples was approved by the Medical Ethics Committee of the Fifth Medical Center of Chinese PLA General Hospital (approval ID: KY-2023-3-22-1; Study on the metabolic regulation of HSC development and regeneration, approved on 30 March 2023). Written informed consent was obtained from all healthy donors before sample collection. Donor privacy was strictly protected, and all collected biological materials were used solely for the purposes of this study.

Results

Metabolomic profiling reveals distinct developmental signatures in human FL, UCB, and aBM HSPCs

To delineate metabolic heterogeneity of human HSPCs across ontogenic stages, we performed untargeted metabolomic profiling of lin⁻CD34⁺ HSPCs isolated from FL, UCB, and aBM. Each sample consisted of 1 × 10⁵ cells, processed by UPLC-MS/MS in both positive and negative ion modes. In total, 779 unique metabolites were detected across all samples, including 470 features in positive ion mode and 390 features in negative ion mode, with partial overlap between modes. A schematic overview of the experimental design encompassing sample preparation, multiomics analyses, and downstream functional assays, is shown in Figure 1A. These metabolites spanned diverse chemical classes, with lipids (34%) and organic acids (14.6%) being the most abundant (supplemental Figure 1A). Violin plots demonstrated consistent metabolite numbers and abundance distributions across biological replicates (supplemental Figure 1B), confirming the robustness of the metabolomic dataset.

Figure 1.

Metabolic profile of human HSPCs from embryonic to adult development. (A) The schematic diagram of the experimental design. CD34⁺ HSPCs were derived from human FL, UCB, and aBM stages. Mononuclear cells (MNCs) were obtained by density gradient centrifugation, followed by magnetic-activated cell sorting enrichment and fluorescence-activated cell sorting (FACS) sorting of HSPCs. Then, the metabolomics of human CD34⁺ HSPCs were analyzed. Based on the results of metabolomics analysis, bulk transcriptome sequencing was further performed after 48 hours of UCB CD34⁺ intervention with additives. Next, the relevant functional verification was conducted, including cell-cycle, cell-division, energy-related indicators, FACS analysis, differentiation, and CFU analysis. Finally, transplantation strategy with ex vivo expanded UCB HSPCs with control, DON, or AA treatment. Human hematopoietic reconstitution was assessed at 12 weeks. (B) Principal component analysis (PCA) of HSPCs from distinct developmental origins in positive (left) and negative (right) ion modes. Samples from FL (yellow), UCB (orange), and aBM (red) stages were shown with 3 biological replicates per group. PCA1 accounts for 48% and 73% of the variance in positive and negative ion modes, respectively. Ellipses represent 95% confidence intervals. (C) Venn diagram showing the overlap of differentially abundant metabolites between each pair of HSPCs at 3 developmental stages in positive (left) and negative (right) ion modes. The number of metabolites in each overlap was indicated in the corresponding circles. (D) Heat maps showing the representative differentially abundant metabolites among human FL, UCB, and aBM HSPCs under the positive (left) and negative (right) ion modes. Differentially abundant metabolites were analyzed at a threshold of variable importance in projection >1 and P < .05. Flt3L, FMS-like tyrosine kinase 3 ligand; SCF, stem cell factor; SFEM, serum-free expansion medium; TPO, thrombopoietin.

Principal component analysis revealed clear separation of metabolic profiles among different developmental stages of human HSPCs (Figure 1B), indicating substantial metabolic heterogeneity across developmental stages. This separation was further validated by orthogonal partial least squares discriminant analysis (supplemental Figure 1C-D), which demonstrated robust discrimination among the 3 groups based on both global and differential metabolite features, with R²Y and Q² values >0.5 in all comparisons, supporting the stability and predictive power of the model.¹⁴

To characterize metabolic alterations across developmental stages, we applied multivariate analysis and identified 130 differentially abundant metabolites using a significance threshold of variable importance in projection >1 and P value < .05. These were predominantly lipid species (>80%), mainly glycerophospholipids (46.2%), fatty acids and derivatives (30.0%), and bile acids (9.2%). Venn diagram analysis revealed developmentally associated metabolites uniquely enriched in each pairwise comparison (Figure 1C). Although phosphatidylcholine was shared across all 3 groups, the limited number of overlaps in metabolites further underscored the divergence of metabolic states across ontogeny. Heat map analysis revealed distinct metabolite enrichment patterns among stages, with each developmental group exhibiting unique metabolic signatures (Figure 1D). Compared with UCB HSPCs, FL HSPCs showed 35 upregulated metabolites in positive ion mode, particularly lipid-related species such as 1-stearoyl-2-arachidonoyl-sn-glycerol, uridine, and 3-amino-2,3-dihydrobenzoic acid, as well as 30 upregulated metabolites in negative ion mode. Compared with UCB HSPCs, 31 metabolites in positive ion mode were more abundant in aBM HSPCs, including organic acid derivatives such as linoleoylglycine and cholic acid (Figure 1D; supplemental Figure 1E). Together, these results suggested that human HSPCs exhibit metabolic heterogeneity across developmental stages.

Differential metabolite levels throughout human HSPC development

To identify dynamic metabolite changes during human HSPC development, we performed KEGG pathway enrichment analysis by combining all differentially abundant metabolites from both positive and negative ion modes of each developmental stage. KEGG analysis revealed developmental stage-specific upregulated pathway modules (Figure 2A). All developmentally associated metabolites were grouped into 9 trajectory clusters according to their abundance changes (Figure 2B; supplemental Figure 2A). Interestingly, D-Gln and AA exhibited unique patterns among the developmentally associated metabolites. D-Gln showed higher abundance in UCB and aBM HSPCs, which participated in the glutamatergic synapse (cluster 2), consistent with its roles in redox control, nitrogen handling, and biosynthetic support in postnatal hematopoiesis.¹⁵ Conversely, AA level was more abundant in FL and aBM HSPCs, which participated in necroptosis and platelet activation pathways (cluster 9), suggesting its essential role in cell survival, inflammation, and thrombopoiesis.¹⁶ Together, these metabolite patterns indicated that AA might dominate early-stage inflammatory and proliferative responses, and Gln might support biosynthetic and antioxidative functions in postnatal stages. Prenatal HSPCs enriched pathways associated with immune signaling and early cellular metabolism, which were consistent with prominent lipid-inflammatory programs associated with AA. In contrast, pathways involved in redox and biosynthesis were enriched in postnatal HSPCs, which was consistent with Gln supporting anabolic and redox programs during transitional and adult phases (supplemental Figure 2B).

Figure 2.

Differential metabolic pathways and key metabolites throughout human HSPC development. (A) KEGG pathway enrichment analysis based on differentially abundant metabolites identified is enriched within FL, UCB, and aBM HSPCs. The Sankey diagram illustrated the enriched metabolites and pathways. In the bubble plot, the color of the points represent the enrichment −log₁₀(P value). The size of the bubbles reflect the number of enriched metabolites in each pathway. (B) Heat map of metabolites in significantly enriched KEGG pathways. The color scale represents relative metabolite abundance changes (with red indicating higher abundance and blue indicating lower abundance). Below the heat map, the trends in metabolite abundance within Mfuzz clusters are shown, with the x-axis representing different groups and the y-axis representing the normalized metabolite abundance changes. Each cluster is labeled with its representative metabolites. (C) Metabolic map of AA and Gln during human HSPC development. FL (yellow), UCB (orange), and aBM (red). Arrows indicate metabolic reactions. Proteomic data were referenced from our previous study (Xiong et al¹⁷). Metabolites involved in each metabolic step are shown within rectangles, and the corresponding enzyme names are indicated in uppercase letters. The size of rectangles represents metabolite abundance level, and the size of circles represents protein expression level. Protein expression levels are displayed as a violin diagram for each comparison, with 4 to 6 independent biological replicates per protein. Statistical significance of proteomic data was determined by 1-way analysis of variance followed by Tukey’s multiple comparison test. ∗P < .05; ∗∗P < .01. ALOX5, arachidonate 5-lipoxygenase; ALOX12, arachidonate 12-lipoxygenase; CYP, cytochrome P450; EGFR, epidermal growth factor receptor; Epox, epoxides; GLS, glutaminase; GLUD1, glutamate dehydrogenase 1; GnRH, gonadotropin-releasing hormone; LTs, leukotrienes; ns, nonsignificant; PG, prostaglandins; PL, phospholipids; PTGS1, prostaglandin-endoperoxide synthase 1; TCA, tricarboxylic acid cycle.

To further place these metabolic data into the overall context of human HSPC development, we integrated the untargeted metabolomic data from this study with our previously published proteomic data of HSPCs across different developmental stages.¹⁷ We constructed a schematic overview of the related metabolic pathways, including Gln metabolism, the tricarboxylic acid cycle, fatty acid synthesis, and downstream AA metabolism (Figure 2C). Metabolite abundances showed that key intermediates, including Gln, citrate, and AA, varied across developmental stages, highlighting dynamic metabolic remodeling during HSPC ontogeny.

Further analysis of our published proteomic data indicated that the expression changes of several key metabolic enzymes were consistent with the abundance trends of the corresponding metabolites. Glutaminase, which catalyzes the conversion of Gln to glutamate, showed reduced abundance during development, suggesting a developmental remodeling of glutaminolysis. This was consistent with the decreased levels of glutamate, indicating reduced Gln catabolism. In addition, prostaglandin-endoperoxide synthase 1, arachidonate 5-lipoxygenase, and arachidonate 12-lipoxygenase, which were involved in AA metabolism, also exhibited elevated expression levels consistent with the increased AA metabolite levels (Figure 2C), further supporting the elevated metabolic activity of AA metabolism upon HSPC development.

Together, these results revealed a coordinated metabolic reprogramming of human HSPCs during ontogeny, reflecting their adaptation to varying proliferative demands and environmental conditions across developmental stages.

Multiomics analysis identifies candidate metabolites that regulate HSC function

To further substantiate our rationale for metabolic intervention, we examined whether our findings aligned with published datasets characterizing hematopoietic metabolism across developmental hierarchies (Figure 3A). In a population RNA-seq study reported by Lalioti et al, Gln-related metabolic genes exhibited lower expression in HSPCs but became progressively upregulated in progenitors.¹⁸ Together, these observations suggest that Gln metabolism may become more active during differentiation and that attenuating its flux might help HSPCs maintain a more primitive cellular state. In contrast, AA metabolism showed an opposite pattern. In the same study, low-input untargeted lipidomic analysis further showed that AA-containing phospholipids, including PE(38:4), were more abundant in downstream progenitor subsets than in HSPCs, supporting the notion that AA-associated lipid remodeling is linked to differentiation rather than stem-state maintenance. In addition, an independent single-cell metabolomic analysis found that AA abundance increased along the dormant-to-active HSC continuum¹⁹ (Figure 3B). Supporting this, Zeng et al²⁰ further showed that downstream AA-derived lipid mediators, including prostaglandin E2 and thromboxane B2, were present at low levels in HSCs but progressively increased across committed progenitors. These metabolites are known to modulate inflammation, proliferation, and platelet production, suggesting that elevated AA availability may accompany or facilitate lineage priming and activation programs.

Figure 3.

Multiomics analysis reveals the regulation of Gln- and AA-associated pathways in human HSPCs. (A) Schematic overview of Gln metabolism and downstream lipid remodeling, showing representative metabolites and key enzymes. Red and blue labels indicate progressive increases or decreases, respectively, from HSCs to progenitor populations, as derived from external low-input targeted metabolomic and population RNA-seq transcriptomic data sets (Lalioti et al¹⁸). (B) Left: phosphatidylethanolamine PE (38:4) levels in human HSCs and progenitors from lipidomic profiling (Lalioti et al¹⁸). Middle: AA abundance across dormant and activated murine HSC subsets from single-cell metabolomic analysis (Cao et al¹⁹). Right: violin plots showing levels of prostaglandin E2 and thromboxane B2 across hematopoietic progenitor populations, based on human single-cell metabolomics (Zeng et al²⁰). (C) Schematic diagram of experimental design. (D-E) Gene Ontology (GO) enrichment of upregulated and downregulated genes after DON treatment (D) and after AA treatment (E). Red bars indicate upregulated pathways, and blue bars indicate downregulated pathways. Green shading denotes GO terms associated with Metabolic pathways and lipid transport, light blue shading denotes Cell proliferation and differentiation, and light pink shading denotes Immunity and trafficking pathways. (F) Gene expression changes related to stemness, immune regulation, and lineage differentiation in HSPCs after AA or DON treatment. Control is shown in gray, the AA-treated group in dark red, and the DON-treated group in blue. α-KG, α-ketoglutarate; CLP, common lymphoid progenitor; CMP, common myeloid progenitor; GMP, granulocyte-macrophage progenitor; MEP, megakaryocyte-erythroid progenitor.

To further elucidate the potential impact of candidate metabolites, we treated human UCB-derived CD34⁺ HSPCs with DON (glutaminase inhibitor) or AA, and then conducted bulk transcriptomic sequencing (RNA-seq) analysis (Figure 3C). Gene Ontology enrichment analysis showed that HSPCs after DON treatment significantly enriched the terms related to immune homeostasis and hematopoiesis-related signaling pathways, including Wingless-related integration site and bone morphogenetic protein signaling pathways (Figure 3D), suggesting that inhibition of Gln metabolism may promote maintenance of the stemness and immune function of HSPCs. In contrast, AA treatment on human HSPCs significantly enriched the terms that related to megakaryocyte differentiation and platelet production (Figure 3E), implying that AA supplementation may promote megakaryocytic differentiation of human HSPCs.

To gain further knowledge on key gene expression changes after AA and DON treatment on human HSPCs, we examined several functionally relevant genes from 3 major categories, including stemness, immune function, and lineage differentiation (Figure 3F). Among them, genes closely related to hematopoiesis and stem cell maintenance, such as HES1 and NOTCH1, as well as immune function–related genes, like LRRK2 and HLA-DMB, showed enhanced expression levels after DON treatment. In contrast, genes such as MERTK, ILK, and PLA2G4, which were associated with megakaryocyte differentiation, activation, and proliferation, were significantly upregulated in the AA treatment group. Together, these transcriptomic findings provide mechanistic clues underlying the divergent roles of Gln and AA metabolism in regulating HSPC function.

Gln metabolism inhibition by DON promotes multilineage engraftment and myeloid differentiation of human HSPCs

To evaluate the role of Gln in human hematopoiesis, we treated human CD34⁺ HSPCs with DON. Functional assays using UCB HSPCs were presented as the primary data set, and parallel experiments were also performed using FL and aBM HSPCs. Concentration gradient experiments revealed that 2.5μM was the optimal concentration for DON (supplemental Figure 3A-B). DON treatment altered the distribution of CD34/CD38 phenotypic subsets across FL, UCB, and aBM HSPCs, resulting in an increased CD34⁺CD38⁻ fraction (Figure 4A-B; supplemental Figure 3C). In UCB HSPCs, DON treatment significantly reduced overall cell expansion, largely driven by a marked loss of CD34⁺CD38⁺ cells, whereas CD34⁺CD38⁻ cells were slightly increased. A similar trend was observed in FL HSPCs, whereas BM HSPCs showed only a modest reduction in proliferation with minimal changes in these compartments (supplemental Figure 3D). Notably, cell-cycle profiling demonstrated that DON treatment induced a shift toward quiescence, characterized by an increase in G₀-phase fractions and a drastic reduction in S/G₂/M-phase fractions in both CD34⁺CD38⁺ hematopoietic progenitor cells (HPCs) and CD34⁺CD38⁻ HSPCs across FL/UCB/aBM HSPCs samples (Figure 4C; supplemental Figures 3E-G and 4A-B). These results suggested a shift toward a more quiescence-associated state. Meanwhile, the Cell Trace Violet (CTV) division assay confirmed that DON-treated UCB HSPCs underwent fewer cell divisions compared with controls over the culture period (Figure 4D).

Figure 4.

Inhibition of Gln use preserves HSPC quiescence and remodels their differentiation potential. (A) Representative flow cytometry plots showing the frequencies of CD34⁺CD38⁻, CD34⁺CD38⁺, CD34⁻CD38⁺, and CD34⁻CD38⁻ cells in control and DON-treated UCB HSPCs after 48 hours. (B) Bar graphs showing the frequencies of CD34⁺CD38⁻, CD34⁺CD38⁺, CD34⁻CD38⁺, and CD34⁻CD38⁻ cells in control and DON-treated FL, UCB, and aBM HSPCs after 48 hours. (C) Bar graphs showing the percentage of cells in different cell-cycle phases (G₀, G₁, S/G₂/M) in CD34⁺CD38⁻ HSPCs derived from FL, UCB, and aBM after DON treatment. (D) Representative CTV histograms showing cell division profiles of UCB HSPCs in control (gray) and DON-treated (blue) groups at day 1, day 2, day 3, and day 4 of culture. The vertical dashed lines indicate successive rounds of cell division (division 0-7), with progressive dilution of CTV fluorescence corresponding to proliferative activity. (E) Representative CFU assay images (original magnification ×40) (left), quantification of UCB and aBM HSPCs cultured (right) in methylcellulose medium with or without DON treatment. Colonies were classified as CFU-granulocyte–macrophage, BFU-E, and CFU-granulocyte–erythroid–macrophage–megakaryocyte. Bar graphs show the number of colonies formed per 200 seeded cells. (F-G) Representative flow cytometry plots (F) showing CD45⁺CD33⁺ myeloid differentiation in control- and DON-treated UCB HSPCs after MS5 coculture, and quantification (G) of the percentages of CD45⁺CD33⁺ myeloid cells in control and DON-treated FL, UCB, and aBM HSPCs after differentiation culture. (H-I) Representative flow cytometry plots (H) showing CD235a⁺ erythroid differentiation in control and DON-treated UCB HSPCs after MS5 coculture, and quantification (I) of the percentages of CD235a⁺ erythroid cells in control and DON-treated FL, UCB, and aBM HSPCs after differentiation culture. Statistical significance is determined using an unpaired Student t test. Error bars denote mean ± standard error of the mean (SEM); ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001; n ≥ 3. BFU-E, erythroid burst-forming unit; GEMM, granulocyte-erythroid-macrophage-megakaryocyte; GM, granulocyte-macrophage; ns, nonsignificant.

Importantly, colony-forming unit (CFU) assays demonstrated that Gln inhibition by DON enhanced the clonogenic capacity of UCB and aBM HSPCs (CFU-granulocyte–macrophage and CFU-granulocyte–erythroid–macrophage–megakaryocyte) (Figure 4E; supplemental Figure 4C). Lineage differentiation assays across all HSPCs from 3 developmental stages revealed that myeloid differentiation potential was significantly enhanced after DON treatment, as evidenced by increased CD33⁺ cell outputs and a corresponding rise in CFU-GM colony formation, accompanied by reduced CD235a⁺ cells (Figure 4F-I; supplemental Figure 4D-E). In contrast, lymphoid outputs showed only modest, stage-dependent changes (supplemental Figure 4F-H). These data suggested that Gln inhibition in HSPCs maintained their quiescent cell state, enhanced their myeloid differentiation potential, and reduced erythroid outputs.

At the transcriptomic level, we also identified the molecular features underlying DON-induced metabolic quiescence in HSPCs. Gene set enrichment analysis revealed negative enrichment of aerobic respiration pathway in DON-treated cells (Figure 5A). In addition, heat map analysis showed lower expression levels of key genes associated with mitochondrial respiratory chain, including NDUFB3, ATP5MD, NDUFA1, MT-ND3, and ROMO1 (Figure 5B). Downregulation of ROMO1,²¹ a well-known regulator of mitochondrial reactive oxygen species (ROS) production, corresponded with the observed reduction in ROS levels, suggesting a coherent mechanism of redox attenuation.

Figure 5.

DON enhances human multilineage reconstitution of UCB HSPCs in NOG-W41 mice. (A) Gene set enrichment analysis (GSEA) showing negative enrichment of the aerobic respiration pathway in DON-treated UCB-HSPCs. Representative downregulated genes (MT-ND3, NDUFB3, NDUFA1, ATP5MD) are indicated. NES and FDR values are shown. (B) Heat map showing the expression of genes involved in the mitochondrial respiratory chain in control and DON-treated UCB-HSPCs. The color scale represents normalized expression values, with red indicating upregulation and blue indicating downregulation. (C-D) Metabolism analysis in UCB HSPCs after DON treatment. ATP levels (C) and NAD⁺/NADH levels (D) were measured. (E) Representative flow cytometry histograms (left) and quantification (right) showing intracellular ROS levels in CD34⁺CD38⁻ HSPCs from control (gray) and DON-treated (blue) groups. The level of ROS intensity was measured by mean fluorescence intensity. (F) Schematic diagram of transplantation strategy with ex vivo expanded UCB HSPCs with control or DON treatment, and human hematopoietic reconstitution was assessed at 12 weeks. (G) Mean levels of human hCD45⁺ engraftment in PB of NOG-W41 recipient mice at different time points after transplantation. (H) Representative flow cytometry plots of human hematopoietic cell engraftment in recipients’ BM at 12 weeks after transplantation. (I) Percentage of human hCD45⁺ engraftment in the BM and spleen of recipients. (J-K) Frequency of human CD45⁺CD33⁺ (J) and CD235a⁺ (K) cells in the BM of recipients. (L) Percentages of human CD34⁺CD38⁻ HSPC and CD34⁺CD38⁺ HPCs in the BM of recipients. Statistical significance is determined using an unpaired Student t test. Error bars denote mean ± SEM; ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001; n ≥ 3. ATP, adenosine triphosphate; FDR, false discovery rate; NES, normalized enrichment score; ns, nonsignificant.

To resolve the underlying metabolic changes, we measured the intracellular ATP and redox cofactors. DON-treated cells exhibited reduced ATP levels and diminished reduced nicotinamide adenine dinucleotide (NADH) without changes in oxidized nicotinamide adenine dinucleotide (NAD⁺), leading to an elevated NAD⁺ to NADH ratio indicating a shift toward an oxidized redox balance (Figure 5C-D). Interestingly, ROS levels were significantly lower in both CD34⁺CD38⁺ and CD34⁺CD38⁻ compartments compared to control (Figure 5E; supplemental Figure 4I), suggesting that Gln blockade suppressed mitochondrial metabolism and ROS-associated signaling, potentially contributing to the observed quiescent state of HSPCs.²²

To assess the phenotype induced by DON in vivo, UCB HSPCs were pretreated with DON for 48 hours and transplanted into nonirradiated NOG-W41 mice (Figure 5F). PB chimerism showed that DON pretreatment did not affect hCD45⁺ engraftment at week 4, but significantly increased human chimerism at weeks 8 and 12 (Figure 5G). Recipient BM and spleen analysis further illustrated increased engraftment and a redistribution of graft composition, including enhanced myeloid outputs, reduced erythroid contribution, and increased primitive HSPCs (Figure 5H-L).

In summary, Gln metabolism inhibition by DON promoted a metabolically quiescent and functionally preserved state in human HSPCs, characterized by cell-cycle arrest, reduced metabolic activity, and lower redox signaling. Consistently, DON increased human multilineage chimerism in NOG-W41 mice, expanded the primitive HSPCs, and skewed lineage output toward myeloid differentiation with reduced erythroid contribution. These results identified Gln metabolism as a key metabolic pathway that maintains HSPC function in vivo.

AA supplementation enhances megakaryocytic differentiation of HSPCs and is accompanied by redox-linked metabolic reprogramming

To examine whether AA treatment could affect the function or metabolic status of human HSPCs, FL, UCB, and aBM HSPCs were cultured with AA supplementation. Concentration gradient experiments showed that 50μM was the optimal concentration for AA (supplemental Figure 5A-B). AA significantly promoted cell expansion of UCB and aBM HSPCs compared with the controls, with CD34⁻ fractions showing a significant increase, and no significant changes were detected in FL HSPCs (Figure 6A; supplemental Figure 5C). Specifically, this was manifested as a leftward shift in CD34 expression intensity and a decreased CD34⁺CD38⁻ proportion, suggesting that AA promoted the differentiation (Figure 6B-C). Consistent trends were also observed in FL and aBM HSPCs (supplemental Figure 5D). Cell cycle profiling in FL/UCB/aBM HSPCs all revealed that AA treatment promoted the G₁ to S/G₂/M transition in total HSPCs and activated cell proliferation (Figure 6D; supplemental Figures 5E-G and 6A-B). Particularly in the CD34⁺CD38⁻ population, G₀-phase cells were nearly absent, indicating that AA supplementation facilitated the entry into cell cycling. The CTV division assay also revealed enhanced proliferative activity in AA-treated cells, as evidenced by a marked reduction in CTV fluorescence intensity relative to the controls (Figure 6E).

Figure 6.

AA supplementation promotes the proliferation of HSPCs and megakaryocytic differentiation. (A) Number of total cells, CD34⁺ HSPCs, and CD34⁻ cells in UCB HSPCs after culture in vitro for 48 hours with AA treatment. Data are presented as paired measurements from independent donors. Error bars denote mean ± SEM; ∗P < .05; ∗∗∗P < .001, by paired Student t test. (B) Merge of representative flow cytometry plots (left) and histogram (right) showing the frequencies of CD34⁺CD38⁻, CD34⁺CD38⁺, CD34^-CD38⁺, and CD34⁻CD38⁻ cells in control and AA-treated UCB HSPCs after 48 hours. (C) Bar graphs showing the frequencies of CD34⁺CD38⁻, CD34⁺CD38⁺, CD34⁻CD38⁺, and CD34⁻CD38⁻ cells in control and AA-treated FL, UCB, and aBM HSPCs after 48 hours. (D) The percentages of cells in different cell cycle phases (G₀, G_1, S/G₂/M) in the CD34⁺CD38⁻ HSPCs derived from FL, UCB, and aBM stages after AA treatment are shown. (E) Representative CTV histograms showing cell division profiles of UCB HSPCs cultured with (red) or without (gray) AA treatment for 1 to 4 days. Vertical dashed lines indicate successive rounds of cell division (division 0-7), reflecting progressive dilution of CTV fluorescence during proliferation. (F) Representative flow cytometry plots (left) and bar graphs (right) showing immature CD41a⁺CD42b⁻ megakaryocytic cells and mature CD41a⁺CD42b⁺ megakaryocytes in control and AA-treated groups in FL HSPC (up), UCB HSPC (middle), and aBM HSPC (down) after 14 days of serum-free megakaryocytic differentiation culture. (G-H) The percentage of immature CD41a⁺CD42b⁻ megakaryocytic cells (G) and mature CD41a⁺CD42b⁺ megakaryocytes (H) were summarized. (I-J) The percentage of CD235a⁺ erythroid cells (I) and CD45⁺CD33⁺ myeloid cells (J) of control and AA-treated UCB HSPCs differentiation induced by MS5 coculture was measured by flow cytometry. (K) Percentage of CD19⁺ B cells generated from control and AA-treated FL, UCB, and aBM HSPCs after differentiation culture. Statistical significance is determined using an unpaired Student t test. Error bars denote mean ± SEM; ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001; n ≥ 3. ns, nonsignificant.

Functionally, AA treatment robustly enhanced megakaryocytic differentiation in HSPCs derived from all 3 developmental stages, as evidenced by increased frequencies of both CD41a⁺CD42b⁻ early and CD41a⁺CD42b⁺ mature megakaryocytes (Figure 6F-H). Notably, FL HSPCs exhibited the strongest ability in both early and mature megakaryocytes. Meanwhile, erythroid, myeloid, and T lymphoid outputs showed only modest, stage-dependent changes (Figure 6I-J; supplemental Figure 6C-E). Notably, CD19⁺ cells showed a consistent increasing trend across all 3 developmental stages of HSPCs (Figure 6K). These findings indicated that AA treatment of human HSPCs mainly promotes megakaryopoiesis.

To determine whether AA-driven functional changes were accompanied by their metabolic changes, we quantified intracellular ATP, NAD⁺/NADH, and ROS levels in HSPCs. AA significantly increased ATP production and reduced NAD⁺ to NADH ratio (Figure 7A-B), suggesting enhanced mitochondrial activity and redox imbalance. These shifts likely reflect the elevated level of oxidative phosphorylation, consistent with a transition toward a more bioenergetically active state. AA treatment also markedly elevated intracellular ROS levels in both CD34⁺CD38⁺ HPCs and CD34⁺CD38⁻ HSPCs (Figure 7C; supplemental Figure 6F). This increase in oxidative stress suggested that AA may promote metabolic activation and redox signaling in more differentiated subsets. Given the context-dependent role of ROS in regulating HSPC fate, including its known involvement in promoting lineage commitment, the elevated ROS levels may contribute to the enhanced proliferation and megakaryocytic differentiation induced by AA.²³

Figure 7.

AA enhances metabolic activation and promotes megakaryocytic output in vitro and in vivo. (A-B) Metabolism-related analysis in UCB-HSPCs after AA treatment. ATP levels (A) and NAD⁺/NADH levels (B) were measured. (C) Representative flow cytometry histograms (left) in CD34⁺CD38⁺ HPCs and quantification (right) showing intracellular ROS levels in HPCs and CD34⁺CD38⁻ HSPCs after AA treatment. ROS intensity was measured by mean fluorescence intensity. (D) Schematic diagram of transplantation strategy with ex vivo expanded UCB HSPCs with control or AA treatment, and human hematopoietic reconstitution was assessed at 12 weeks. (E) Percentage of human hCD45⁺ engraftment in recipients’ BM. (F) Lineage differentiation in recipients’ BM. (G) Representative flow cytometry plots (left) and bar graphs (right) of human Mk (human CD45⁺CD41a⁺) in recipients’ BM. (H) Frequency of human CD45⁺CD41a⁺ Mk in recipients’ spleen. Statistical significance was assessed by an unpaired Student t test. P values are indicated as ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001. Error bars denote mean ± SEM; n ≥ 3. ATP, adenosine triphosphate; Mk, megakaryocyte; ns, nonsignificant; SP, spleen.

To assess the effect of AA in vivo, UCB HSPCs were pretreated with AA and transplanted (Figure 7D). The results showed that AA treatment did not significantly change overall human chimerism in recipients’ BM (Figure 7E). Consistent with the in vitro differentiation assays, AA-treated grafts showed increased CD19⁺ B outputs, minimal change in CD33⁺ myeloid cells, and a slight reduction in CD3⁺ T-cell contribution (Figure 7F). More importantly, AA pretreatment markedly enhanced hCD41a⁺ megakaryocytic output in both the BM and spleen of recipient mice after transplantation (Figure 7G-H). These in vivo findings were consistent with the transcriptomic enrichment of megakaryocyte-related programs (Figure 3E).

Taken together, AA promoted cell-cycle entry, metabolic activation, and megakaryocytic differentiation of human HSPCs in vitro. Although AA did not significantly alter overall BM chimerism, it markedly increased megakaryocytic outputs after xenotransplantation. These results identified AA as a developmentally regulated metabolic cue that preferentially drives megakaryocyte differentiation.

Discussion

In this study, we dissected the metabolic features of human HSPCs across 3 developmental stages, such as FL, UCB, and aBM, using untargeted metabolomics. Among the 130 differentially abundant metabolites identified, most were lipids involved in membrane remodeling, fatty acid oxidation, and redox homeostasis. Pathway enrichment analysis indicated that FL and aBM HSPCs were preferentially associated with lipid-related pathways, whereas UCB HSPCs showed stronger associations with amino acid-related modules, including Gln metabolism and nucleotide biosynthesis. Although both FL and UCB are prenatal sources, they diverged in metabolic preferences. FL HSPCs were enriched for fatty-acid derivatives, including AA, whereas UCB HSPCs exhibited higher levels of Gln-related metabolites and intermediates. This divergence suggests that AA- and Gln-related metabolic programs may differentially shape HSPC functional states during hematopoietic maturation.

Based on their distinct developmental abundance patterns, we prioritized AA- and Gln-related axes for in vitro and in vivo perturbation in human HSPCs. First, RNA-seq after Gln-pathway perturbation showed suppression of gene programs related to aerobic respiration and the mitochondrial electron transport chain. Consistently, DON treatment reduced HSPC proliferation and cell division, enhanced myeloid differentiation and reduced erythroid potential, and decreased ROS levels in vitro. Similarly, in vivo transplantation also showed that DON treatment could improve multilineage reconstitution in NOG-W41 mice. Simsek et al reported the low mitochondrial activity in dormant HSCs,²⁴ and Rodriguez-Fraticelli et al demonstrated that long-term hematopoietic stem cells maintain a low metabolic and oxidative state.²⁵

Notably, DON treatment also resulted in lineage bias, which prominently increased myeloid output while suppressing erythroid differentiation. This myeloid-favoring, erythroid-suppressing effect mirrors findings from Oburoglu et al, who reported that Gln supports de novo pyrimidine synthesis, which is crucial for erythropoiesis.²⁶ Although both studies observed reduced erythroid output upon Gln inhibition, our data further place this effect within the context of reduced metabolic activity, mitochondrial suppression, and altered ROS status. Therefore, rather than suggesting definitive dormancy, our integrated results suggest that repression of glutaminolysis is associated with phenotypic and functional features consistent with a more quiescence-associated HSPC phenotype.

Furthermore, transcriptomic profiling of AA-treated cells revealed increased expression of genes such as MERTK, ILK, and PLA2G4A, which were involved in signal transduction, adhesion, and cytoskeletal remodeling.²⁷ For example, MERTK has been shown to modulate HSC homing and niche interaction,²⁸ whereas ILK has been associated with megakaryocytic differentiation.²⁹ Although previous work have implicated AA and its lipid mediators in platelet-related biology and inflammatory signaling,^30,31 our study extends these observations by showing that AA treatment is associated with transcriptional programs related to proliferation and megakaryocytic differentiation, along with directionally consistent functional effects in UCB HSPCs and supportive findings from parallel assays in FL and aBM cultures.

Several limitations of our study should also be mentioned. Because of the limited accessibility of human samples, especially the embryonic samples, the number of biological replicates in our study remained constrained. The present metabolic analysis was conducted on the lin⁻CD34⁺ population, which is insufficient to resolve the heterogeneity within the bulk cell population containing HSCs and different HPC subtypes. Interestingly, we observed a nonlinear developmental pattern in the abundance of certain metabolites, most notably AA, which showed elevated levels in FL, decreased abundance in UCB, and re-elevated expression in aBM. This biphasic trajectory suggests a developmental stage–dependent regulatory role that may also be influenced by the niche of AA that may not align with a simple linear model of hematopoietic maturation.

This study integrated untargeted metabolomic profiling, targeted metabolic perturbations, transcriptomic analysis, and in/ex vivo functional validation to identify developmentally associated metabolic regulators of human HSPCs. We highlighted Gln-related metabolism and AA-linked lipid programs as 2 prominent axes that were differentially engaged across developmental stages and associated with distinct HSPC functional states and lineage outputs (see the Graphical Abstract). Pharmacological inhibition of Gln use with DON enhanced multilineage engraftment of human HSPCs, particularly myeloid differentiation, whereas AA supplementation primarily promoted megakaryocytic differentiation. Collectively, these findings underscore the importance of metabolic remodeling in coordinating human HSPC function and provide a framework for understanding how targeted metabolic modulation may be leveraged to preserve HSPC maintenance or direct lineage commitment.

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