Preparation of single-cell suspensions

The FL samples were collected within two hours after the termination of pregnancy, and the FL tissues were cut into small pieces and digested with collagenase VI for 15–20 min at 37 °C with periodic mixing. Then, the FL single-cell suspension was filtered through a 100 µm cell strainer. The FL, UCB, and aBM mononuclear cells (MNCs) were isolated by density gradient centrifugation according to the manufacturer’s protocol (Ficoll-Paque, GE Healthcare) and filtered through a 70 µm cell strainer. All experiments we performed were on freshly isolated samples within six hours to avoid the perturbations induced by the freezing and thawing of cells.

Flow cytometry and MACS

After staining with CD34-PE (581, 343506, BioLegend) and CD45-FITC (HI30, 560976, BD), the proportion of CD34⁺CD45^mid cells was analyzed with a FACS Calibur. CD34⁺ cells were purified from FL, UCB, and aBM monocular layers via two rounds of magnetic positive cell selection (MACS, Miltenyi Biotec) with CD34 microbeads and MS columns (Miltenyi Biotec). Lin⁻CD34⁺ HSPCs were sorted on a Sony MA900 Multi-Application Cell Sorter. Human Lineage antibody cocktail including anti-human CD2, CD3, CD14, CD16, CD19, CD56, and CD235a (eFluor™ 450, clone: RPA-2.10, OKT3, 61D3, CB16, HIB19, TULY56, HIR2, 22-7775-72, eBiosciences). The FACS-sorted cells were washed with PBS and centrifuged to discard the supernatant, flash-frozen, and stored at −80 °C for subsequent proteomics analysis.

To calculate the proportions of Ficoll-purified human FL, UCB, and aBM CD90⁺ HSCs (Lin⁻CD34⁺CD38⁻CD45RA⁻CD90⁺), we stained the MNCs with anti-human Lineage-eFluor™ 450, anti-human CD45-BV421 (HI30, 563879, BD), anti-human CD34-FITC (581, 343503, Biolegend), anti-human CD38-PE-Cy7 (HB-7, 356608, Biolegend), anti-human CD45RA-BV605 (5H9, 740424, BD), and anti-human CD90-APC (5E10, A15726, Life) antibodies. To compare the expression of surface markers of FL, UCB, and aBM HSPCs for FACS validation, we stained the MNCs with anti-human HLA-DR/DP/DQ-FITC (Tü39, 361705, BioLegend), anti-human CD133-APC (TMP4, 17-1338-42, eBioscience), and anti-human CD144-BV786 (55-7H1, 565672, BD). Flow cytometry was performed on a BD FACSymphony™ A5. Data analysis was performed using Flowjo V10.8.1 software (http://www.flowjo.com).

Cell culture

Freshly isolated UCB-derived Lin⁻CD34⁺ cells were incubated with buthionine sulfoximine (BSO, 125 µM) or N-acetyl-L-cysteine (NAC, 100 µM) or both for 2 days in serum-free conditions with Stem-Span SFEM II (StemCell Technologies, 9655) supplemented with human TPO, SCF, and FLT3-L (50 ng/mL each) (Peprotech, AF-300-18, AF-300-07; AF-300-19) and 1% penicillin–streptomycin (P/S).

Colony-forming assays

For methylcellulose colony-forming assays, 200 FACS-sorted HSPCs (Lin⁻CD34⁺CD38^–) were plated with methylcellulose-based medium containing recombinant cytokines for human cells (Methocult, StemCell Technologies, H4434, contains SCF, IL-3, EPO, and GM-CSF) onto 24-well plates. Colonies were counted and morphologically assessed after 14 days.

Cell cycle

Each group of cells labelled with human HSPC markers was washed twice with cold PBS. Cells were fixed and permeabilized with BD Cytofix/Cytoperm and stained with Ki67-PE (SolA15, 12-5698-82, eBioscience) and Hoechst33342 (10 µg/ml).

Apoptosis analysis

Each group of cells labelled with human HSPC markers was washed twice with cold PBS and incubated with Annexin V (KeyGEN BioTECH) and 7-AAD (Invitrogen) for 30 min at 4 °C according to the manufacturer’s instructions. The proportion of apoptotic cells was detected by flow cytometry within 1 h.

Seahorse assays

The oxygen consumption rate (OCR) and the proton efflux rate (PER) were measured using a Seahorse XFe96 Extracellular Flux Analyser (Agilent). OCR and PER were performed with a Seahorse XF Cell Mito Stress Test Kit (Agilent, 103015–100) and a Seahorse XF Glycolytic Rate Assay Kit (Agilent, 103344–100), respectively. A mixture of mitochondrial pressure medium was prepared by adding 200 µL of glucose (1 M), pyruvate (100 mM), and glutamine (200 mM) to 19.4 mL of RPMI medium (Agilent, 103576–100) and storing it according to the manufacturer's instructions. The glycolytic rate assay medium was similar to the medium above for measuring mitochondrial pressure. The OCR and PER were measured according to the manufacturer's protocol. Briefly, we pipetted 180 µL of a cell suspension of 1×10⁵ purified FL and UCB Lin⁻CD34⁺ HSPCs into each sample well of fibronectin (Gibco, 33016015)-coated 96-well XFe plates. After three baseline OCR measurements, the cells were sequentially exposed to oligomycin (1.5 µM), carbon-cyanide-4 (trifluoromethyl) phenazine (FCCP; 2 µM), and rotenone/antimycin A (Rot/AA; 1 µM). Three measurements were recorded after each injection. After three baseline PER measurements, the cells were sequentially exposed to a Rot/AA mixture (0.5 μM) and 2-deoxyglucose (2-DG; 50 mM). After normalization by cell number/well, the results were analyzed using the Wave program 2.6.3.5 (Agilent) and GraphPad Prism 9.

Metabolic state analysis of HSPCs

Sorted FL, UCB, and aBM HSPCs (1×10⁵ per sample, n=4) were lysed, and the intracellular GST activity was measured using a Glutathione S-transferase (GST) Activity Assay Kit (BC0355, Solarbio) according to the manufacturer's instructions. GSH/GSSG levels were assessed using a GSH and GSSG assay kit (S0053, Beyotime). For pyruvate content analysis, sorted FL and UCB HSPCs (1×10⁵ per sample, n=3) were measured using a Pyruvate Assay Kit (BC2205, Solarbio) according to the manufacturer's instructions. For 2-NBDG (2-[N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)amino]-2-deoxy-D-glucose) glucose uptake, 5×10⁶ MNCs were first stained with human HSPC phenotypic markers (Lineage cocktail, CD34, CD38 antibodies) were stained with 50 μg/mL 2-NBDG (ab235976, Abcam) in 1 mL of PBS containing 0.5% BSA at 37 °C in incubation bins for 1 h, and flow cytometry was performed on a BD FACSymphony™ A5 (excitation/emission=485/535 nm). For reactive oxygen species (ROS) analysis, 5×10⁶ MNCs were first stained with HSPC phenotypic markers and suspended with the CellROX^® Reagent (C10422, Invitrogen) at a final concentration of 5 μM and incubated for 30 min at 37 °C (excitation/emission=640/665 nm).

Measurement of the protein synthesis rate

A total of 5×10⁶ MNCs were first stained with HSPC phenotypic markers and mixed with 50 mM OP-Puro (C10456, Invitrogen) for 1 h, fixed for 20 min, and then detected with a Click-iT^Ⓡ OPP Reaction Buffer Kit (C10456, Invitrogen) according to the manufacturer’s instructions (excitation/emission=495/519 nm).

Mitochondrial analysis

The mitochondrial mass and active mitochondrial content were assessed by MitoTracker Green (MTG) (C1048, Beyotime) and MitoTracker Red CMXRos (C1035, Beyotime) staining, respectively. Briefly, 5×10⁶ MNCs were first stained with human HSPC phenotypic markers and then stained with MTG (100 nM) (excitation/emission=490/516 nm) or MitoTracker Red CMXRos (200 nM) (excitation/emission=579/599 nm) in 2 mL of HBSS (C1029, Beyotime) at 37 °C in incubation bins for 30 min.

Sample preparation for proteomics analysis

Owing to the limited number of samples, the donor samples were analyzed in single replicates, and different individuals were considered biological replicates. The number of samples/donors was limited by the material available from the hospital for this study. We generated proteomic data for Lin⁻CD34⁺ HSPCs from FL (n=4), UCB (n=5), and aBM (n=4) (10⁵ cells per sample) samples. All of the samples were processed via an in-solution digestion method with an EasyPept Ex very microprotein extraction kit (OSFP0004-8X, Omicsolution). In brief, a detergent solution (50 mM ammonium bicarbonate (ABC) solution containing 0.1% RapiGest and 0.4 μL of 5×cocktail with a final concentration of 20 mM) was introduced to each sample and incubated for 2 h at 37 °C. Following cooling to room temperature, iodoacetamide (IAM) (40 mM) was added, and the mixture was mixed at 600 rpm for 1 min. This mixture was kept in darkness for 30 min. Subsequently, TEPC (Tris (2-carboxyethyl) phosphine) (3 µL) and CAA (chloroacetamide) (5 µL) were sequentially introduced into the mixture. The reaction was then terminated by heating to 95 °C, and trypsin (enzyme:protein=1:2 (w/w)) digestion was carried out at 37 °C for 2 h. Finally, the peptides from each sample were purified, concentrated via vacuum centrifugation, and reconstituted in a solution containing 0.1% (v/v) formic acid. The peptide concentration was assessed by measuring the UV spectral density at 280 nm. For the data-independent acquisition (DIA) experiment, an index retention time (iRT) calibration peptide was included at a concentration of 400 ng per sample.

MS assay for data-independent acquisition (DIA)

LC–MS/MS analysis was performed on a timsTOF Pro mass spectrometer (Bruker Daltonics, Bremen, Germany) coupled with a nanoElute liquid chromatograph (Bruker Daltonics, Bremen, Germany) with a 75 μm×25 cm long column (Thermo Scientific) (1.9 μm id) for a 90 min gradient. Peptide separation was performed at a flow rate of 300 nL/min with mobile phases A (0.1% (v/v) formic acid in water) and B (0.1% (v/v) formic acid in acetonitrile) along with a 90 min gradient as follows: 0–70 min, 2–22% B; 70–78 min, 22–37% B; 78–83 min, 37–95% B; and 83–90 min, 95% B. The mass spectrometer was operated in positive ion mode and defined ion mobility MS from TIMS scans in a single 100 ms 100–1700.32 window mass range from the m/z-ion mobility plane. The collision energy increased linearly with mobility during parallel accumulation–serial fragmentation (PASEF) MS/MS scanning, ranging from 20 eV at 1/K0=0.60 Vs/cm² to 59 eV at 1/K0=1.60 Vs/cm².

MS data analysis

The DIA data were analyzed using a database search of Spectronaut™ 14.4.200727.47784. The main software parameters were set as follows: the retention time prediction type was dynamic iRT, MS2 level correction interference was enabled, and cross-run normalization was enabled. All of the results were filtered according to the q value cut-off of 0.01 (equivalent to an FDR<1%). The database information used in our search database was “Swissprot-Home-sapiens-20405-20230103".

PCA

After normalization to the total peak intensity, the processed data were uploaded and imported into SIMCA-P (version 14.1, Umetrics, Umea, Sweden), where multivariate data analysis, including Pareto-scaled PCA, was performed.

Pearson correlation coefficient

Pearson correlation coefficient analysis is a statistical method that measures the linear relationship between two variables by calculating the product-moment correlation coefficient between them. This correlation coefficient was between −1 and+1, where+1 represented a complete positive correlation (one variable increased and the other increased).

KEGG annotation

In accordance with the annotation procedure, the online KEGG database (http://geneontology.org/) was used to identify the KEGG homology of the studied proteins and localize them in the KEGG pathways.

Enrichment analysis

Enrichment analysis was performed using Fisher’s exact test, which uses whole quantified proteins as the background dataset. Afterwards, Benjamini‒Hochberg correction was applied for multiple tests to adjust the derived p values. Only functional categories and pathways with a p value below the 0.05 threshold were considered significant.

Fuzzy C-means clustering

The protein expression trends were illustrated to analyze the overall expression patterns of all DEPs at two consecutive stages among FL, UCB, and aBM HSPCs. Fuzzy processing was performed via Mfuzz software. The fuzzy c-means (FCM) algorithm was used for analysis, and all proteins were divided into nine different expression modules on the basis of their expression trends.

Real-time quantitative (RT‒PCR)

We extracted total cellular RNA from Lin⁻CD34⁺ FL, UCB, and aBM cells using TRIzol reagent (Invitrogen, USA). The RNA was reverse transcribed to cDNA using the PrimeScript RT Reagent Kit (RR047A, Takara), subjected to PCR, and placed in a fluorescence quantitative PCR instrument for RT‒PCR data analysis.

ENO1-F: 5’-AGCCAGTGCAGGAATCCAGGTAG-3’

ENO1-R: 5’-ACTCGGTCACGGAGCCAATCT-3’

PROM1-F: 5’-CACTACCAAGGACAAGGCGTTCA-3’

PROM1-R: 5’-CGCTGGTCAGACTGCTGCTAAG-3’

ESAM-F: 5’-GGTGCTGTGCCTGTGATGGT-3’

ESAM-R: 5’-CCACTTCCTCTTCTCCTTCTGTCT-3’

SDHB-F: 5’-TCAGGAAGGCAAGCAGCAGTATCT-3’

SDHB-R: 5’-GATGGTGTGGCAGCGGTATAGAGA-3’

GSTP1-F: 5’-CCAGAACCAGGGAGGCAAGA-3’

GSTP1-R: 5’-GAGGCGCCCCACATATGCT-3’

LANCL1-F: 5’-TGAGTTCTCACAACGCTTGAC-3’

LANCL1-R: 5’-CGAGGGTCTGCTGATTTCAGG-3’

GSTM2-F: 5’-AAATGCTGAAGCTCTACTCACAGTTTC-3’

GSTM2-R: 5’-GGCTCAAATACTTGGTTTCTCTCAAG-3’

GSS-F: 5’-GGAACATCCATGTGATCCGAC-3’

GSS-R: 5’-GCCATCCCGGAAGTAAACCA-3’

β-actin-F: 5’-TCCATCATGAAGTGTGACGT-3’

β-actin-R: 5’-GAGCAATGATCTTGATCTTCAT-3’.

Subsets of human HSPC proteins undergo fetal-to-adult transition with different patterns

Next, we calculated the DEPs between HSPCs at different developmental stages to identify their protein characteristics (fold change>1.5 or<0.667 and p<0.05) (Fig. 2A, Additional file 1: Fig. S2B, Additional file 3). The subcellular distribution of DEPs was shown in Additional file 1: Fig. S2C. These DEPs were used for downstream analyses. Statistical analysis revealed 1,358 DEPs between fetal and newborn HSPCs. The most significant fraction (1,088 proteins) of these proteins was highly expressed in FL HSPCs. Surprisingly, compared with FL or aBM HSPCs, UCB HSPCs presented very few DEPs, which suggested that UCB HSPCs presented more similar protein expression features to FL and BM HSPCs but fewer unique expression features. The proteins differentially expressed between different pairs of cell populations were analyzed via a Venn diagram (Fig. 2B, Additional file 1: Fig. S4A–B, Additional file 5). To identify the dynamic changes in HSPC development between neighbouring cell populations, fuzzy c-means (FCM) [33] was applied to cluster global protein expression profiles. We observed nine distinct clusters of protein expression patterns representing proteins that were differentially regulated (Fig. 2C, Additional file 1: Fig. S3, Additional file 4). Among these, clusters 4 and 6 represented proteins upregulated in fetal HSPCs and then downregulated after birth. In contrast, clusters 5 and 7 represented proteins that were downregulated in fetal HSPCs and upregulated after birth.

Open in a separate window

Fig. 2

Differential protein levels throughout human HSPC development. A Changes in the proteomics of upregulated and downregulated proteins between each pair in FL, UCB, and aBM HSPCs. DEPs were analyzed at thresholds of FC>1.5 or FC<0.67 and p<0.05. B Overlapping DEPs between each pair of stages: FL vs. UCB, UCB vs. aBM, and FL vs. aBM. C Mfuzz analysis is divided into nine different expression trends on the basis of the expression trends of all DEPs of each stage for human HSPCs. The X-axis represents the other groups, and the Y-axis represents the expression change after homogenization. The lines of each cluster refer to a class of proteins expressed in proteins. Each cluster identifies its representative proteins. D Heatmap showing the -log₁₀ transformed significant value (p value) of the KEGG pathway terms describing each of the 9 clusters. E KEGG pathway enrichment analysis of terms significantly upregulated and downregulated in fetal HSPCs versus UCB and aBM HSPCs

Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis was subsequently applied to identify pathways enriched in each cluster. Notably, fatty acid degradation, oxidative phosphorylation (OXPHOS), the tricarboxylic acid (TCA) cycle, the peroxisome proliferator-activated receptor (PPAR) signaling pathway, and ribosome were enriched in clusters 4 or 6, with higher expression in fetal HSPCs (Fig. 2D–E). In contrast, chemokines, immune-related pathways (such as T and B-cell receptors), signalling pathways (Wnt, MAKP, TGF-β, and FoxO), antigen processing and presentation, proteasome, spliceosome (involved in mRNA maturation), glycolysis/gluconeogenesis, and the pentose phosphate pathway (PPP) were among the top differentially regulated pathways enriched in clusters 5 or 7, showing an upregulated pattern during human HSPC development (Fig. 2D–E). The FoxO signalling pathway, which plays essential roles in response to physiological oxidative stress and the detoxication of ROS [34], was active in UCB and aBM HSPCs but not in FL HSPCs (Fig. 2D).

Protein translational machinery-related pathways, such as ribosome and ribosome biogenesis, were more enriched in FL HSPCs than in UCB and aBM HSPCs (Fig. 2D–E, Additional file 1: Fig. S3A). Both ribosomal assembly and protein synthesis have been implicated in HSC function and regeneration [26, 35]. A low level of protein synthesis is essential for maintaining HSC metabolic homeostasis, whereas an alteration in protein synthesis impairs HSC function [26, 36]. Homeostatic or quiescent HSCs restrict protein synthesis for long-term maintenance. In contrast, proliferating HSCs undergo excessive protein synthesis to support their drastic expansion, which may increase the number of misfolded proteins and impair HSC function. Human FL hematopoiesis requires a highly regulated protein synthesis rate; it acts as an essential niche for hematopoietic differentiation and HSC expansion until birth [1, 32]. In addition, peroxisome-related proteins were also upregulated in FL HSPCs, whereas the spliceosome and proteasome were highly expressed in UCB and aBM HSPCs (Fig. 2D–E, Additional file 1: Fig. S3A).

Taken together, our proteome data revealed dynamic and complex protein expression patterns during the fetal-to-adult transition process in human HSPC development, which occurred progressively along with a continuum of HSPC maturation.

Metabolic switch of human HSPCs from fetal to adulthood through the perinatal period

Compared with HSPCs in the UCB or aBM stages, those in the FL stage proliferate extensively, resulting in different metabolic demands [24]. However, the metabolic pathways responsible for the production of energy and protein synthesis at the human fetal, newborn, and adult stages have not been described comprehensively. As shown in Fig. 2E and Additional file 1: Fig. S3B, the most notable proteome differences between HSPCs at the three different developmental stages were related to central carbon metabolism (summarized in Fig. 3). Proteome analysis revealed that the rate-limiting enzymes involved in facilitating the entry of glucose into glycolysis/gluconeogenesis, such as hexokinase 1 (HK1), glycogen phosphorylases liver form (PYGL), phosphoglucose mutase 2 (PGM2), glycolytic enzymes aldolase C (ALDOC), triosephosphate isomerase (TPI1), and enolase-1/2 (ENO1/2), were overexpressed in UCB or aBM HSPCs (Fig. 3A). These findings indicated that adult HSPCs rely primarily on glycolysis to supply energy. HK1, the enzyme that catalyses the first rate-limiting step of glycolysis, exhibited an age-associated increase among human BM CD34⁺ HSPCs [20]. In addition, we detected a significantly increased level of ENO1 in UCB HSPCs compared with their FL counterparts via RT‒PCR (Fig. 3B). The quiescent HSPCs have low energy requirements and are believed to depend on glycolysis rather than mitochondria. They are sustained by a hypoxic ecological niche and display great self-renewal capacity [37–39].

Open in a separate window

Fig. 3

Prominent changes in central carbon metabolism occur in HSPCs at three developmental stages. A–B Glucose metabolism (glycogen metabolism and PPP) A and the TCA cycle B in FL (brown), UCB (dark grey), and aBM (peacock blue) HSPCs are depicted as unidirectional arrows representing unidirectional reactions and bidirectional arrows indicating bidirectional reactions. The gene names of the respective enzymes are written in capital letters. Green indicates the enzymes upregulated in fetal HSPCs, red indicates the enzymes downregulated in fetal HSPCs, and black indicates which enzyme was not changed (see also Fig. 4). Fold change (FC) proteomics expression data of enzymes are shown as histograms in the indicated comparisons. Four to six independent experiments were performed per enzyme. The significance of the proteomic data is indicated by one-way ANOVA with Tukey’s multiple comparison test. *p<0.05, **p<0.01, ***p<0.001, and ****p<0.0001. C–D RT‒PCR analysis (n=3) of ENO1 and SDHB in FL and UCB HSPCs. The significance is indicated by an unpaired Student’s t test. *p<0.05 and **p<0.01

In contrast, the proteins involved in OXPHOS and the TCA cycle presented increased expression in FL HSPCs (Fig. 3C). Pyruvate dehydrogenases (PDHB and PDHX) were highly expressed in FL HSPCs, promoting entry into the TCA cycle [40]. Furthermore, FL HSPCs presented increased pyruvate levels, although the difference was insignificant (Fig. 4G), consistent with increased mitochondrial OXPHOS. The enzymes involved in the TCA cycle (such as ACO2, IDH1, SDHA, SDHB, and FH) were abundant at the fetal stage (Fig. 3C-D). FH has been demonstrated to be a critical metabolic regulator of HSC self-renewal and differentiation [41]. These significantly differentially expressed enzymes could be targeted to intervene in the energy metabolism of human HSPCs and further influence their fate.

Open in a separate window

Fig. 4

The metabolic landscape of human HSPCs among FL, UCB, and aBM. A Changes in purine, AA, fatty acid, ketogenesis, glutamine, and glutathione metabolism between FL (brown), UCB (dark grey), and aBM (peacock blue) HSPCs. Green indicates the enzymes upregulated in fetal HSPCs (see Fig. 3). Fold changes in the proteomeo data of enzymes are shown as histograms in the indicated comparison. Four to six independent experiments were performed per enzyme. The significance of the proteomic data is indicated by one-way ANOVA with Tukey’s multiple comparison test. *p<0.05, **p<0.01, and ***p<0.001. B Scheme of glutamine-related metabolism and glutathione metabolism replenishment. C Proteomic expression data of glutathione metabolism-related enzymes are shown as histograms. The significance of the proteomic data is indicated by one-way ANOVA with Tukey’s multiple comparison test. *p<0.05, **p<0.01, ***p<0.001, and ****p<0.0001. D RT‒PCR analysis (n=3) of GSTP1, LANCL1, GSTM2, and GSS in FL and UCB HSPCs. *p<0.05, **p<0.01, and ***p<0.001. E GST levels of FL, UCB, and aBM HSPCs. Significance is indicated by one-way ANOVA with Tukey’s multiple comparison test. *p<0.05, **p<0.01, ***p<0.001, and ****p<0.0001. F GSH and GSSG levels in FL and UCB HSPCs. Significance is indicated by the Student’s t test. *p<0.05, **p<0.01, and ***p<0.001. G Pyruvate levels in FL and UCB HSPCs as indicated (n=3)

In addition to these mitochondria-related pathways, amino acid (AA) metabolism (especially valine, arginine, and proline), fatty acid degradation, lipoic acid metabolism, and steroid biosynthesis were among the pathways most differentially regulated in FL HSPCs compared with those in UCB or aBM HSPCs (Fig. 2D/4A). Enzymes involved in glutamine metabolism, such as glutaminase (GLS) and glutamate dehydrogenase 1 (GLUD1), were highly enriched in FL HSPCs (Fig. 4B). However, the enzymes that participate in glutathione (GSH) metabolism, such as S-transferases (GSTs), GSTP1, LANCL1, and GSTM2, as well as the GSH synthesis enzymes GSR and GPX4, presented increased expression patterns in UCB or aBM HSPCs (Fig. 4B-C). We also detected significantly increased levels of GSTP1, LANCL1, GSTM2, and GSS in UCB or aBM HSPCs compared with their FL counterparts via RT‒PCR (Fig. 4D). Differences in mRNA expression were detected mainly before and after birth, not between newborn and adult HSPCs. In addition, UCB and aBM HSPCs displayed high GST activity (Fig. 4E). Interestingly, compared with UCB HSPCs, FL HSPCs presented significantly increased levels of free glutathione pools, both reduced (GSH) and oxidized (GSSG) (Fig. 4F). We hypothesize that UCB HSPCs may have a higher glutathione transfer metabolism rate to consume GSH and GSSG than FL HSPCs, which needs further investigation.

Switching from the oxidative pathway to the glycolytic pathway coincides with the development of human HSPCs

Seahorse analysis was used to measure the oxygen consumption rate (OCR), which was indicative of OXPHOS. In contrast, the proton efflux rate (PER) in the glycolytic rate assay provides accurate measurements of glycolytic rates under basal conditions and compensatory glycolysis following mitochondrial inhibition, which indicates glycolysis. As shown in Fig. 5A-E, compared with FL HSPCs, UCB HSPCs demonstrated a lower OCR rate, basal respiration, maximal respiration, spare respiratory capacity, ATP production, and nonmitochondrial oxygen consumption. In contrast, UCB HSPCs displayed relatively greater basal glycolysis and compensatory glycolysis via PER measurements. Taken together, these findings are consistent with our findings that FL HSPCs primarily use OXPHOS to meet their energy demands and have a more active metabolic state than UCB HSPCs [11].

Open in a separate window

Fig. 5

Higher mitochondrial function and increased glucose uptake support the oxidative consumption of FL HSPCs. A OCRs were examined in FL and UCB HSPCs via the Seahorse assay. The error bars represent the mean±SEM. B Levels of basal respiration, maximal respiration, spare respiratory capacity, ATP production, and nonmitochondrial oxygen consumption were examined. The error bars represent the mean±SEM; the significance is indicated by unpaired Student’s t tests. *p<0.05. C PERs of FL and UCB HSPCs were measured. The error bars represent the mean±SEM. D–E The levels of basal glycolysis (D) and compensatory glycolysis (E) were examined. The error bars represent the mean±SEM. F–G Flow cytometric analysis of 2-NBDG glucose uptake in Lin⁻CD34⁺CD38⁻ (F) and Lin⁻CD34⁺CD38⁺ (G) cells at three developmental stages. H–I Flow cytometric analysis of total mitochondrial mass (MTG) in Lin⁻CD34⁺CD38⁻ (H) and Lin⁻CD34⁺CD38⁺ (I) cells at three developmental stages. J–K Flow cytometric analysis of the active mitochondrial content (MitoTracker Red CMXRos) in Lin⁻CD34⁺CD38⁺ (J) and Lin⁻CD34⁺CD38⁻ (K) cells at three different developmental stages. L–M Flow cytometric analysis of cellular ROS levels in Lin⁻CD34⁺CD38⁻ (L) and Lin⁻CD34⁺CD38.⁺ (M) cells at three developmental stages. Significance is indicated by one-way ANOVA with Tukey’s multiple comparison test. *p<0.05, **p<0.01, ***p<0.001, and ****p<0.0001

As the proteomic profile and Seahorse analysis revealed an increased mitochondrial respiratory rate in fetal HSPCs, we investigated whether human HSPCs at different developmental stages differed in terms of mitochondrial content and activity, glucose uptake, ROS production, and ribosomal translation. To obtain a sufficient number of cells for assessing mitochondrial function, we focused on the Lin⁻CD34⁺CD38⁻ population. MTG and MitoTracker Red CMXRos were used to quantify the total and active mitochondria, respectively (Fig. 5F/5H, Additional file 1: Fig. S5A-B). We also measured glucose uptake using 2-NBDG, a fluorescently tagged glucose analogue (Fig. 5J). Protein synthesis was analyzed via an O-propargyl-puromycin (OP-puro) incorporation assay (Additional file 1: Fig. S5G-H). FL Lin⁻CD34⁺CD38⁻ showed higher total and active mitochondria content than that in its UCB and aBM counterparts. An apparent increase in ROS production was noted in the FL Lin⁻CD34⁺CD38⁻ population compared with that in the UCB and aBM populations (Fig. 5L). The level of glucose uptake also increased in the FL Lin⁻CD34⁺CD38⁻ population. The same pattern was also shown in the Lin⁻CD34⁺CD38⁺ HPC population (Fig. 5G/5I/5K/5M), which was consistent with our proteomic analysis of the Lin⁻CD34⁺ HSPC population (Additional file 1: Fig. S5C-F).

Distinct response of human Lin⁻CD34⁺CD38⁻ HSPCs and Lin⁻CD34⁺CD38⁺HPCs to perturb glutathione metabolism

The enzymes (GSTP1, GSTM2, and LANCL1) involved in glutathione metabolism showed increased expression patterns in human HSPCs after birth (Fig. 4C-D). We investigated whether perturbing glutathione metabolism affects ROS production, the metabolic state, and the expansion of human HSPCs (Fig. 6A). BSO, an inhibitor of glutathione synthetase, can lead to an increase in intracellular ROS levels [42]. Treatment with BSO (125 µM) in vitro for two days significantly decreased HSC expansion (Lin⁻CD34⁺CD38⁻), leading to differentiation and significantly increased levels of ROS (Fig. 6B/6C/6I). Next, we asked whether the pharmacological inhibition of ROS elevation could protect human HSPCs from functional degradation. The antioxidant agent N-acetyl-L-cysteine (NAC, 100 µM) protected human HSPCs from BSO-induced HSPC differentiation and elevated ROS levels (Fig. 6B/6C/6I).

Open in a separate window

Fig. 6

Distinct response of human HSCs and HPCs to disturb glutathione metabolism. A Experimental diagram of BSO/NAC/BSO+NAC treatment of human CD34⁺ HSPCs in vitro. B Number of total CD34⁺ HSPCs after culture in vitro for 48 h with BSO/NAC/BSO+NAC. C Frequency of CD34⁺CD38⁻ and CD34⁺CD38⁺ cells among CD34⁺ HSPCs after 48 h of culture with BSO/NAC/BSO+NAC. D–E Apoptosis analysis of HSPCs cultured for 48 h with BSO/NAC/BSO+NAC. Early (Annexin V⁺7-AAD⁻) and late (Annexin V⁺7-AAD⁺) apoptosis were quantified by flow cytometry in Lin⁻CD34⁺CD38⁻ HSPCs (D) and Lin⁻CD34⁺CD38⁺ HPCs (E). F Cell cycle analysis of CD34⁺CD38⁻ (up) and CD34⁺CD38⁺ (down) cells among CD34⁺ HSPCs after 48 h of culture with BSO/NAC/BSO+NAC. G–J Flow cytometric analysis of total mitochondria (MTG) (G), active mitochondrial content (MitoTracker Red CMXRos) (H), ROS levels (I), and protein synthesis rate (OP-Puro) (J) in Lin⁻CD34⁺CD38⁻ HSCs. (Error bars represent the means±SEM, n>3). Significance was indicated by one-way ANOVA with Tukey’s multiple comparison test (n>3, *p<0.05, **p<0.01, ***p<0.001, and ****p<0.0001)

Unexpectedly, the apoptosis of HSPCs (Lin⁻CD34⁺CD38⁻) and progenitors (Lin⁻CD34⁺CD38⁺) in response to BSO and NAC was significantly different (Fig. 6D-E). The increase in BSO-induced elevated ROS levels significantly increased the number of apoptotic Lin⁻CD34⁺CD38⁻ HSPCs (FC=20.85, p<0.0001, relative to CON), and NAC rescued the cells from apoptosis (FC=5.36, p<0.0001, relative to CON) (Additional file 1: Fig. S6A). This effect appears to be specific to primitive Lin⁻CD34⁺CD38⁻ cells, as no drastic increase in apoptosis was observed in Lin⁻CD34⁺CD38⁺ HPCs (FC=3.60, p<0.0001, relative to CON). In contrast, NAC aggravated the apoptosis of HPCs (FC=6.05, p<0.0001, relative to CON) (Additional file 1: Fig. S6A). Interestingly, BSO treatment decreased the total mitochondrial content while increasing the active mitochondrial content in both the HSPC and HPC populations (Fig. 6G-H, Additional file 1: Fig. S6B-C). These results were also validated by quantifying the cellular ROS levels (Fig. 6I, Additional file 1: Fig. S6D-E). BSO treatment halted the cell cycle in Lin⁻CD34⁺CD38⁻ cells but not in Lin⁻CD34⁺CD38⁺ cells (Fig. 6F). BSO treatment of human CD34⁺ cells also significantly decreased de novo translation (protein synthesis rate), which was confirmed using OP-Puro (Fig. 6J, Additional file 1: Fig. S6F-G).

Taken together, these results demonstrated that the oxidative stress induced by BOS treatment led to ROS accumulation, a decreased protein synthesis rate, and the collective induction of apoptosis in Lin⁻CD34⁺CD38⁻ cells. In contrast, Lin⁻CD34⁺CD38⁺ cells could resist apoptosis to some extent. The antioxidant NAC also induced different responses to rescue BSO-induced elevated apoptosis in HSPCs and HPCs. Glutathione metabolism and ROS play critical roles in human HSC function (especially in mitochondria), and the regulation of glutathione-related metabolism and ROS has the potential to maintain HSC function and longevity and improve HSC regeneration or HSCT in patients.

The authors thank all donors who were permitted to include samples in this study. We thank Dr. B. Liu for the critical discussion and Mr. H. Sun for his help in the flow cytometry experiment (Institute of Infectious Diseases, Fifth Medical Center of Chinese PLA General Hospital). We thank Mr. X. Zhang and Mrs. H. Jiang for the original schematic diagram drawing. The authors declare that they did not use the AI in this study.