ATP citrate lyase controls hematopoietic stem cell fate and supports bone marrow regeneration

Abstract

In order to support bone marrow regeneration after myeloablation, hematopoietic stem cells (HSCs) actively divide to provide both stem and progenitor cells. However, the mechanisms regulating HSC function and cell fate choice during hematopoietic recovery remain unclear. We herein provide novel insights into HSC regulation during regeneration by focusing on mitochondrial metabolism and ATP citrate lyase (ACLY). After 5‐fluorouracil‐induced myeloablation, HSCs highly expressing endothelial protein C receptor (EPCR^high) were enriched within the stem cell fraction at the expense of more proliferative EPCR^Low HSCs. These EPCR^High HSCs were initially more primitive than EPCR^Low HSCs and enabled stem cell expansion by enhancing histone acetylation, due to increased activity of ACLY in the early phase of hematopoietic regeneration. In the late phase of recovery, HSCs enhanced differentiation potential by increasing the accessibility of cis‐regulatory elements in progenitor cell‐related genes, such as CD48. In conditions of reduced mitochondrial metabolism and ACLY activity, these HSCs maintained stem cell phenotypes, while ACLY‐dependent histone acetylation promoted differentiation into CD48⁺ progenitor cells. Collectively, these results indicate that the dynamic control of ACLY‐dependent metabolism and epigenetic alterations is essential for HSC regulation during hematopoietic regeneration.

ACLY‐dependent coupling of metabolic state and chromatin accessibility facilitates recovery of the blood system after myeloablation.

Introduction

Hematopoietic stem cells (HSCs) govern the lifelong maintenance of hematopoietic homeostasis via self‐renewal and multi‐lineage differentiation. In the steady state, adult HSCs maintain a quiescent state with low mitochondrial metabolism within a specialized microenvironment of bone marrow (BM), called the “niche” (Trumpp et al, 2010; Ito & Suda, 2014; Nakamura‐Ishizu et al, 2020). However, upon stress hematopoiesis, such as inflammatory conditions or myeloid ablation, HSCs exit quiescence and actively divide to regenerate the hematopoietic system through an appropriate balance between self‐renewal and differentiation (Harrison & Lerner, 1991; Essers et al, 2009; Baldridge et al, 2010). Under these conditions, cytokines, niche factors, and intrinsic regulation in HSCs are reportedly involved in the promotion of BM regeneration (Zhao et al, 2014; Zhou et al, 2017; Chen et al, 2019; Severe et al, 2019; Sezaki et al, 2020; Sudo et al, 2021). However, the mechanisms by which HSC functions are regulated during BM regeneration remain unclear. We previously demonstrated that the up‐regulation of the mitochondrial membrane potential (ΔΨ_m) contributed to the initiation of HSC divisions after the administration of 5‐fluorouracil (5‐FU) (Umemoto et al, 2018); however, the mechanisms by which HSCs select self‐renewal or differentiation currently remain unknown.

Previous studies reported that quiescent HSCs showed low ΔΨ_m in the steady state; therefore, high ΔΨ_m has been recognized as a type of differentiation indicator (Maryanovich et al, 2015; Vannini et al, 2016, 2019). However, HSCs with high ΔΨ_m derived from 5‐FU‐treated mice showed equivalent engraftment activity to quiescent HSCs derived from untreated mice (Umemoto et al, 2018). Moreover, a recent study revealed that HSCs with high ΔΨ_m had enhanced stem cell functions after transplantation (Mansell et al, 2020). Therefore, the activation of mitochondrial metabolism does not always imply a decrease in stem cell function. We herein attempted to elucidate the mechanisms by which HSCs convert the effects of their metabolic status into stem cell regulation.

ATP citrate lyase (Acly) is an enzyme that converts citrate into acetyl‐coenzyme A (Acetyl‐CoA) in the cytoplasm. Although pyruvate dehydrogenase within mitochondria also synthesizes Acetyl‐CoA from pyruvate through the tricarboxylic acid (TCA) cycle, Acetyl‐CoA cannot pass through the mitochondrial membrane. Therefore, Acly‐synthesized Acetyl‐CoA plays an essential role in the acetylation of several proteins outside mitochondria as well as lipid synthesis (Zaidi et al, 2012). The Acly‐dependent synthesis of Acetyl‐CoA is closely associated with mitochondrial metabolism because citrate, a substrate of Acly, translocates to the cytoplasm from mitochondria through corresponding transporters. Acly reportedly contributed to histone acetylation in several cells under high glucose conditions (Wellen et al, 2009; Deb et al, 2017). In addition, the inhibition of the mitochondrial citrate transporter, Slc25a1, also suppressed cytoplasmic Acetyl‐CoA (Tan et al, 2020). Therefore, Acly is recognized as one of the key factors linking metabolic and epigenetic regulation.

In the present study, we provide a novel insight into HSC regulation during BM regeneration by examining the metabolic and epigenetic status of HSCs after the administration of 5‐FU. In the early phase after the administration of 5‐FU, HSCs highly expressing the stem cell marker, endothelial protein C receptor (EPCR), which originally has the potential to maintain stem cell features under a high metabolic status, predominantly remained within BM. These HSCs enabled stem cell expansion by promoting histone acetylation following enhancements in mitochondria‐Acly activity. When HSCs gained enhanced differentiation potential accompanied by the increased accessibility of enhancers in progenitor cell‐related genes in the late phase during BM regeneration, Acly‐mediated histone acetylation allowed these HSCs to provide progenitor cells. Therefore, the regulation of Acly activity in accordance with differentiation potential is essential for HSC regulation during BM regeneration from myeloablative stress.

Results

HSC properties change from the early to late phase during BM regeneration after the administration of 5‐FU

To analyze HSCs during BM regeneration after 5‐FU‐induced myeloablation, we utilized the EPCR‐based definition (Umemoto et al, 2018). Lineage⁻EPCR⁺CD150⁺CD48⁻ (L⁻ESLAM) cells represented almost the same fraction as EPCR⁺CD150⁺CD48⁻c‐Kit⁺Sca1⁺lineage⁻ (ESLAM KSL) cells in the steady state (Fig EV1A). Although the majority of L⁻ESLAM cells highly expressed Sca1 after the administration of 5‐FU (Fig EV1B), the expression of c‐Kit in the L⁻ESLAM fraction was markedly altered (Fig EV1C). However, we previously confirmed that c‐Kit^Low L⁻ESLAM cells derived from 5‐FU‐treated mice showed equal long‐term repopulating (LTR) activity to c‐Kit^High L⁻ESLAM cells derived from untreated mice (Umemoto et al, 2018). In addition, after carboxyfluorescein succinimidyl ester (CFSE)‐labeled L⁻ESLAM cells obtained from untreated mice were transplanted without irradiation into 5‐FU‐treated mice, donor cells showed a similar expression level of c‐Kit to recipient HSCs under every condition (Fig EV1D). These results indicate that c‐Kit expression in HSCs is largely affected by the surrounding environment during BM regeneration. Therefore, the L⁻ESLAM definition enables a precise analysis of HSCs during BM regeneration after the administration of 5‐FU without the influence of altered c‐Kit levels.

Figure EV1. c‐Kit levels in lineage⁻EPCR⁺CD150⁺CD48⁻ (L⁻ESLAM) HSCs markedly changed during BM regeneration after the administration of 5‐FU (Related to Fig 1).

• A–C
  The frequency of Sca‐1⁺c‐Kit⁺ cells (A) or Sca‐1⁺ cells (B) and c‐Kit expression levels (C), in the L‐ESLAM fraction before and after the administration of 5‐FU. Data are shown as means ± SD (n = 4, two independent experiments, *P < 0.01 by the t‐test).
• D
  According to the indicated schedule, 10,000 L⁻ESLAM HSCs (Ly5.2) were transplanted without irradiation into untreated or 5‐FU‐treated mice, and c‐Kit levels in the donor or recipient L⁻ESLAM fraction were subsequently examined. Numbers within each histogram of donor cells represent the number of CFSE⁺ cells detected. The graph depicts the relative MFI of c‐kit in the indicated cells. Data are shown as means ± SD (n = 4, two independent experiments, N.S.; not significant by the t‐test).

Using the L⁻ESLAM definition, we confirmed that the administration of 5‐FU markedly decreased not only proliferative progenitors, including CD48⁺ progenitor cells, but also HSCs until 3 days after the administration of 5‐FU (Fig 1A). HSC numbers began to recover from 3 days after the administration of 5‐FU and increased until 12 days (Fig 1A). In addition, a focus on the recovery phase when HSC numbers increased (from 3 to 12 days after the administration of 5‐FU) revealed that the number of HSCs was higher than that of CD48⁺ progenitor cells in the early phase (3–7 days), whereas CD48⁺ progenitor cells markedly increased and became predominant in the late phase (9–12 days) (Fig 1A). HSCs in both phases similarly showed the significant enrichment of genes related to cell division among up‐regulated genes from that in the steady state (Fig 1B), and exhibited markedly higher cell division activity than quiescent HSCs (Fig 1C). However, the gene expression pattern in HSCs was clearly distinguishable between the early and late phases (Fig 1D). In HSCs, the expression of genes related to energy metabolism was higher in the early phase than in the late phase (Fig 1E). Moreover, consistent with cell division capacity, the expression of cell division‐related genes was up‐regulated in the early phase (Fig 1C and E). The potential to generate CD48⁺ progenitor cells during the ex vivo culture was also enhanced in HSCs from the end of the early phase (7 days after the administration of 5‐FU) (Fig 1F). These results suggest that the metabolic status, cell division capacity, and stem cell potential of HSCs changed from the early to late phase of hematopoietic recovery after the administration of 5‐FU.

Figure 1. The HSC potential is changed during BM regeneration after the administration of 5‐FU.

1. Kinetics of cell numbers in Lineage⁻EPCR⁺CD150⁺CD48⁻ (L‐ESLAM) HSCs, Lineage⁻EPCR⁺CD150⁺CD48⁺ (CD150⁺CD48⁺) progenitor cells, and Lineage⁻EPCR⁺CD150⁻CD48⁺ (CD150⁻CD48⁺) progenitor cells after the administration of 5‐FU (250 mg/kg, i.v.). The small graph is an enlargement of 0 to 8 days after the administration of 5‐FU. (* or **) in the small graph represents the statistical analysis of HSC between 3 and 4 or 7 days, respectively. Data are shown as means ± SD (n ≥ 4, two or more independent experiments, *P < 0.01, **P < 0.05 by the t‐test).
2. The enrichment plots of genes related to cell cycle in the comparison of the steady state with the early (upper) or late phase (lower). NES and q values represent normalized enrichment scores and FDR, respectively.
3. The frequency of EdC⁺ cells within L⁻ESLAM HSCs 24 h after the administration of EdC (150 mg/kg, i.p.). Data are shown as means ± SD (n ≥ 4, two independent experiments, *P < 0.01 by the t‐test).
4. A principal component analysis (PCA) using the transcriptome data of L⁻ESLAM HSCs derived from untreated (steady state) or 5‐FU‐teated mice at 6 (early phase) or 10 days (late phase).
5. Comparison of gene expression patterns in HSCs between the early (6 days) and late phases (10 days) on GSEA. The graph shows normalized enrichment scores (NES). Bold and italic mean gene sets show (P < 0.05, FDR < 0.05) or (P < 0.05), respectively.
6. The frequencies of phenotypic HSCs (Orange gate) and CD48⁺ progenitor cells (blue gate) after a 4‐day culture of HSCs derived from 5‐FU‐treated mice (at 6–9 days). Data are shown as means ± SD (n = 4, two independent experiments, *P < 0.01 by the t‐test).

To confirm whether the output of HSC divisions differs between the early and late phases after the administration of 5‐FU, we transplanted dye‐labeled HSCs without irradiation into mice treated with 5‐FU after 4 (early phase) or 9 days (late phase), and subsequently analyzed donor‐derived cells after 2 days (Fig 2A). Transplanted HSCs divided within the BM of 5‐FU‐treated recipients; however, cell division numbers were higher in the early phase than in the late phase (Fig 2B and C: histogram). Donor cells within untreated recipients maintained an undivided status (Fig 2B and C: histogram). These cell division statuses of donor cells were consistent with the expression level of genes related to cell division in host HSCs before and after the administration of 5‐FU (Figs 1E and 2D). Importantly, the frequency of phenotypic HSCs within divided cells in the late phase was significantly lower than that in the early phase due to the generation of more CD48⁺ progenitor cells (Fig 2C). Moreover, the maintained L⁻ESLAM fraction exhibited weaker LTR activity in late phase than in the early phase (Fig 2E). While HSCs in the late phase showed a slightly decreased frequency of myeloid lineage cells in the peripheral blood of primary recipients, a marked difference was not observed between the early and late phases (Fig 2F). These results indicate that HSCs generated in the early phase exhibited a greater stem cell potential than those in the late phase.

Figure 2. Cell fate and stem cell activity differed after HSC divisions between the early phase and late phase following the administration of 5‐FU.

• A
  An experimental model to examine divided HSCs within 5‐FU‐treated mice using the transplantation of dye (CFSE or CytoTell^TMRed 650)‐labeled HSCs.
• B
  The frequency of L⁻ESLAM HSCs within the divided fraction in 5‐FU recipients or undivided fraction in untreated recipients. Numbers within each histogram represent the CFSE⁺ cell number detected. Data are shown as means ± SD (n ≥ 4, two independent experiments, *P < 0.01, **P < 0.05 by the t‐test).
• C
  Cell division number in donor‐derived cells within untreated or 5‐FU recipients. Data are shown as means ± SD (n ≥ 4, two independent experiments, *P < 0.01, **P < 0.05, N.S.; not significant by the t‐test).
• D
  The enrichment plots of cell cycle‐related gene sets in the comparison of the early phase with the late phase.
• E, F
  Fifty L⁻ESLAM HSCs within the divided fraction of donor‐derived cells in 5‐FU recipients or undivided fraction in untreated recipients were transplanted along with 2 × 10⁵ competitor cells. After 20 weeks, the chimerism of peripheral blood was examined. White or gray circles within plots represent donor cell chimerism in multilineage‐reconstructed recipient mice (> 0.5% for myeloid and B‐ and T‐lymphoid lineages) or unreconstructed mice after 20 weeks, respectively (E). Black bars show the average (n ≥ 9, two independent experiments, *P < 0.01, **P < 0.05 by the t‐test). Graphs depict the frequency of each lineage within donor‐derived cells (% Ly5.2⁺ cells) in the peripheral blood of recipient mice. Data are shown as means ± SD (n > 9, **P < 0.05 by the t‐test, two independent experiments) (F).

Chromatin accessibility pattern and histone acetylation in HSCs markedly change during BM regeneration

To further confirm the differential potential of HSCs between the early and late phases, we performed ATAC‐seq. The chromatin accessibility pattern in HSCs markedly changed after the administration of 5‐FU, with further alterations being observed from the early to late phase (Fig 3A–C). We also performed a Cleavage Under Targets and Tagmentation (CUT & Tag) assay for H3K27 acetylation (H3K27ac), which is the active marker of promoters and enhancers, using HSCs in each phase and CD150⁺CD48⁺ progenitor cells in the late phase to extract potential cis‐regulatory regions from each of the ATAC peaks detected (Fig 3D). Importantly, genes associated with accessibility‐increased or ‐decreased cis‐regulatory regions in the late phase were significantly enriched among genes up‐regulated in CD150⁺CD34⁺ progenitor cells or CD150⁺CD34⁻ HSCs, respectively (Fig 3E and F). These results suggest that the differentiation potential of HSCs is enhanced from the early and late phases through the increased accessibility of cis‐regulatory regions in progenitor cell‐related genes, such as Cd48, which is consistent with their greater potential to provide CD48⁺ progenitor cells (Figs 1D and 2C). The binding motifs of well‐known hematopoietic regulators, such as PU.1, RUNX, and Scl (Curtis et al, 2004; Lecuyer & Hoang, 2004; Iwasaki et al, 2005; Wang et al, 2014; de Bruijn & Dzierzak, 2017) were significantly enriched within accessibility‐increased regions from the early to late phase (Fig EV2A and B). Moreover, each Runx family members had accessibility‐increased enhancer regions within or close to each gene locus (Fig EV2C). Importantly, these regions appeared to be activated in multi‐potent progenitors (MPPs) rather than HSCs, as evidenced by enhanced H3K27 levels in MPPs (Fig EV2C). Therefore, these transcription factors may be involved in the change observed in chromatin accessibility in HSCs from the early to late phase.

Figure 3. Chromatin accessibility in HSCs changed during BM regeneration after the administration of 5‐FU.

• A
  A principal component analysis (PCA) using the ATAC‐seq data of HSCs in the steady state (untreated), early phase (5‐FU 6 days), or late phase (5‐FU 10 days).
• B, C
  A scatter plot comparing ATAC‐seq peaks between untreated HSCs and those in the early phase (B) or between those in the early and late phases (C).
• D
  ATAC peaks overlapping with H3K27ac peaks detected in HSCs in each phase or CD48⁺ progenitor cells in the late phase after the administration of 5‐FU were selected as cis‐regulatory regions.
• E
  Scatter plots comparing peaks within cis‐regulatory regions in L⁻ESLAM HSCs between the early and late phases after the administration of 5‐FU.
• F
  The enrichment of genes related to “accessibility‐increased regions at the late phase (Brown dot in Fig 3E)” (left) or “accessibility‐decreased regions (Green dot in Fig 3E)” (right) within genes up‐regulated in CD150⁺CD34⁻ HSCs or CD150⁺CD34⁺ progenitor cells.
• G
  A scatter plot comparing the peaks of cis‐regulatory regions between the steady state and early phase after the administration of 5‐FU.
• H
  The enrichment of genes related to “accessibility‐increased regions from the steady state to early phase (Purple dot in Fig 3G)” in the comparison between the steady state and early phase after the administration of 5‐FU.
• I
  The enrichment of genes related to “accessibility‐increased regions from the early to late phase (Brown dot in Fig 3E)” in the comparison between the early and late phases after the administration of 5‐FU.
• J
  ATAC and H3K27ac peaks in the promoter and enhancer regions within the CD48 locus. ChIP‐seq data for transcription factors obtained from the public database (GSE22178) were merged.
• K
  CD48 mRNA levels in HSCs after the administration of 5‐FU. Data are shown as means ± SD (n = 4, two independent experiments, N.S.; not significant by the t‐test).
• L
  Average plot (top) and heatmap (bottom) of H3K27ac reads in HSCs in the early and late phases within accessibility‐decreased (left: Black dot in Fig 3H), ‐unchanged (center: Blue dot in Fig 3H), and ‐increased cis‐regulatory regions (right; Purple dot in Fig 3H) in the early phase.
• M
  Average plot (top) and heatmap (bottom) of H3K27ac reads in HSCs in the early and late phases within accessibility‐decreased (left: Green dot in Fig 3E), ‐unchanged (center: Blue dot in Fig 3E), or ‐increased regions (right: Brown dot in Fig 3E) in the late phase.

Figure EV2. An analysis of accessibility‐increased regions in HSCs from the early to late phase after the administration of 5‐FU (Related to Fig 3).

1. The top 5 most enriched motifs within accessibility‐increased ATAC peaks (Late/Early > 2 in Fig 3C) in the late phase after the administration of 5‐FU.
2. Top 5 most enriched motifs within the accessibility‐increased cis‐regulatory regions in the late phase (Late/Early > 2 (Brown dot) in Fig 3E).
3. Genome maps showing three enhancer regions related to the Runx family. ChIP‐seq data obtained from the public database (H3K27ac in long‐term (LT)‐HSCs, short‐term (ST)‐HSC, and multi‐potent progenitors (MPPs); indicated transcription factors in HPC7) were merged.

However, HSCs in the late phase maintained stem cell phenotypes, even under the increased accessibility of enhancers in progenitor cell‐related genes. Therefore, we examined whether accessibility‐increased regions contribute to actual gene expression in HSCs during BM regeneration after the administration of 5‐FU. Genes associated with accessibility‐increased cis‐regulatory regions in the early phase relative to the steady state were significantly enriched among genes up‐regulated in the early phase (Fig 3G and H). However, a correlation was not observed between genes associated with accessibility‐increased regions from the early to late phase and genes up‐regulated in the late phase (Fig 3E and I). The promoter and enhancer regions within the CD48 locus in the late phase showed increased accessibility from the early to late phase (Fig 3J), whereas the mRNA expression of CD48 did not markedly differ between both phases (Fig 3K). Therefore, the increased accessibility of cis‐elements does not significantly reflect actual gene expression in the late phase relative to that in the early phase.

After increasing the accessibility of enhancer regions, H3K27ac is generally required to activate gene expression (Choukrallah & Matthias, 2014). H3K27ac levels within accessibility‐increased or ‐unchanged cis‐regulatory regions in the early phase were enhanced from the steady state to the early phase (Fig 3L). In contrast, accessibility‐increased cis‐regulatory regions in the late phase showed no changes in H3K27ac levels from the early to late phase (Fig 3M). The accessibility of promoter and enhancer regions within the CD48 locus was enhanced from the early to late phase, whereas H3K27ac levels remained unchanged (Fig 3J). Since genes associated with accessibility‐increased cis‐regulatory regions from the early to late phase included many progenitor cell‐related genes (Fig 3F), the present results suggest that unchanged H3K27ac levels within these accessibility‐increased regions contributed to the maintenance of stem cell phenotypes in HSCs in the late phase.

Acly contributes to HSC expansion in the early phase after the administration of 5‐FU

Since the regulation of H3K27ac on not only accessibility‐increased but also accessibility‐unchanged regions in HSCs markedly differed between the early and late phases, we examined the kinetics of global H3K27ac levels in HSCs during BM regeneration after the administration of 5‐FU. Global H3K27ac levels in HSCs significantly increased in the early phase, but decreased in the late phase close to the level of the steady state (Fig 4A). Acly plays a key role in the regulation of histone acetylation through the generation of cytoplasmic Acetyl‐CoA, a substrate for histone acetyltransferases (HATs) (Wellen et al, 2009; Deb et al, 2017) (Fig 4B). Similar to H3K27ac, the mRNA and protein levels of Acly in HSCs increased in the early phase after the administration of 5‐FU and decreased in the late phase (Fig 4C and D). Moreover, increases in Acly phosphorylation at Ser455, which has been shown to enhance enzymatic activity (Potapova et al, 2000), were observed in the early phase (Fig 4E). HSCs in the early phase also exhibited greater glucose uptake and higher intracellular levels of glutamate (Glu), a metabolite of glutamine, citrate, and total acetylated lysine, which reflect intracellular Acetyl‐CoA levels (Houston et al, 2020) (Fig 4F–I). Although HSCs in the late phase similarly showed greater glucose uptake and increased intracellular levels of Glu (Fig 4F and G), citrate and total acetylated lysine levels were both markedly lower than those in the early phase (Fig 4H and I). Mitochondrial metabolism contributes to the Acly‐dependent synthesis of Acetyl‐CoA through the supply of citrate, which is a substrate of Acly (Zaidi et al, 2012; Lin et al, 2013; Tan et al, 2020) (Fig 4B). Importantly, ΔΨ_m was significantly enhanced in the early phase and suppressed in the late phase (Fig 4J). Consistent with the change in ΔΨ_m, the expression of genes related to oxidative phosphorylation was higher in the early phase than in the steady state or late phase (Figs 1E and 4K). Moreover, the treatment with CTPI‐2 (an inhibitor of mitochondrial citrate transporters) suppressed H3K27 acetylation in HSCs in the early phase, similar to BMS303141 (an inhibitor of Acly) (Fig 5A–C). Therefore, these results indicate that ΔΨ_m and Acly‐mediated metabolism, leading to histone acetylation in HSCs, were markedly enhanced in the early phase, but suppressed in the late phase during BM regeneration.

Figure 4. HSCs showed enhanced mitochondria‐Acly activity leading to histone acetylation in the early phase after the administration of 5‐FU.

• A
  Global H3K27ac in HSCs during BM regeneration after the administration of 5‐FU. Data are shown as means ± SD (n = 4, two independent experiments, *P < 0.01, **P < 0.05 by the t‐test).
• B
  The relationship between the TCA (tricarboxylic acid cycle) cycle within mitochondrial metabolism, Acly, and histone acetylation.
• C–E
  mRNA (C), protein (D), and phosphorylation levels of Ser455 (E) in Acly in HSCs after the administration of 5‐FU. Data are shown as means ± SD (n ≥ 4, two independent experiments, *P < 0.01, N.S.; not significant by the t‐test).
• F
  The potential for glucose uptake by HSCs obtained from untreated or 5‐FU‐treated mice was estimated using 2‐NBDG. Data are shown as means ± SD (n ≥ 3, two independent experiments, *P < 0.01, N.S., not significant by the t‐test).
• G–I
  The levels of intracellular Glu (G), citrate (H), and total acetylated lysine (I) in HSCs obtained from untreated or 5‐FU‐treated mice. Data are shown as means ± SD (n ≥ 3, two independent experiments, *P < 0.01, **P < 0.05, N.S., not significant by the t‐test).
• J
  ΔΨ_m in L⁻ESLAM HSCs after the administration of 5‐FU. Histograms represent the fluorescence intensity of JC‐1 Red, which is polymer‐derived fluorescence generated in response to ΔΨ_m. The graph depicts normalized ΔΨ_m by the ratio of the fluorescence intensity of JC‐1 Red to JC‐1 Green, which indicates monomer‐derived fluorescence. Data are shown as means ± SD (n ≥ 4, more than two independent experiments, *P < 0.01, **P < 0.05 by the t‐test).
• K
  The enrichment of the gene set “HALLMARK_OXIDATIVE_PHOSPHORYLATION” within differentially expressed genes of the steady state versus the early phase. NES and q values represent normalized enrichment scores and FDR, respectively.

Figure 5. Acly‐mediated histone acetylation was essential for HSC expansion in the early phase after the administration of 5‐FU.

• A
  Schema of experimental designs and the targets of inhibitors. Serial treatments with CTPI‐2 (50 mg/kg/shot/day, i.p.), BMS303141 (100 mg/kg/shot/day, p.o.), A‐485 (20 mg/kg/shot/day, i.p.), or TV‐3664 (10 mg/kg/shot/day, p.o.) were performed 3, 4, and 5 days after the administration of 5‐FU.
• B, C
  The effects of CTPI‐2 (B) or BMS303141 (BMS) (C) on H3K27ac in HSCs in the early phase after the administration of 5‐FU. Data are shown as means ± SD (n ≥ 3, two independent experiments, *P < 0.01, **P < 0.05 by the t‐test).
• D, E
  The effects of CTPI‐2 (D), BMS303141 (BMS) (E), A‐485 (D), and TV‐3664 (E) on HSC expansion in the early phase after the administration of 5‐FU. Numbers in the dot plots represent the frequencies of the indicated fractions within total BM cells. Data are shown as means ± SD (n ≥ 4, two independent experiments, *P < 0.01, **P < 0.05 by the t‐test).
• F, G
  Average plot (top) and heatmap (bottom) of H3K27ac reads in the cis‐regulatory regions of genes belonging to the gene set “HALLMARK_MITOTIC_SPINDLE” (F) or “HALLMARK_G2M_CHECKPOINT” (G) in HSCs in the steady state or early phase.
• H
  The enrichment of the gene set “GOBP_FATTY_ACID_BIOSYNTHETIC_PROCESS” within differentially expressed genes of the steady state versus the early phase.
• I
  Average plot (top) and heatmap (bottom) of H3K27ac reads in the cis‐regulatory regions of genes having the GO term “fatty acid biosynthetic process” in HSCs in the steady state or early phase.

In addition to histone acetylation, the CTPI‐2 or BMS303141 treatment simultaneously suppressed HSC expansion in the early phase after the administration of 5‐FU (Fig 5D and E). Moreover, the suppressive effects of these inhibitors on HSCs in the early phase were observed in the ex vivo culture (Fig EV3). A‐485 (an inhibitor of histone acetyltransferase) also suppressed HSC expansion in the early phase (Fig 5D). In addition, the locus of cell cycle‐related genes, which was up‐regulated in the early phase, showed higher H3K27ac levels in the early phase than in the steady state (Figs 1B and 5F and G). These results indicate that Acly‐mediated histone acetylation contributes to HSC expansion in the early phase, possibly by activating the expression of cell cycle‐related genes.

Figure EV3. The suppression of Acly‐related metabolism in vitro in HSCs in the early phase (Related to Fig 5).

• A, B
  The number or L⁻ESLAM phenotypic HSCs after a 4‐day culture of HSCs in the early phase after the administration of 5‐FU in the presence or absence of CTPI‐2 (50 μM), an inhibitor of mitochondrial citrate transporters, or BMS303141 (BMS; 100 μM), an inhibitor of Acly. Data are shown as means ± SD (n ≥ 3, two independent experiments, *P < 0.01 by the t‐test).
• C
  H3K27ac levels after a 1‐day culture of HSCs in the early phase after the administration of 5‐FU with or without CTPI‐2 or BMS303141. Data are shown as means ± SD (n ≥ 4, two independent experiments, *P < 0.01 by the t‐test).

Similar to histone acetylation, fatty acid synthesis occurs downstream of Acly‐related metabolism. TVB‐3664 (an inhibitor of fatty acid synthase; Fasn) suppressed HSC expansion in the early phase (Fig 5E), suggesting that Acetyl‐CoA‐dependent fatty acid synthesis plays a key role during HSC expansion in the early phase. In addition, HSCs in the early phase showed the up‐regulated expression of genes related to fatty acid synthesis (Fig 5H), accompanied by increased H3K27ac levels in the locus of these genes (Fig 5I). These results suggest that Acly‐related metabolism played an essential role in enhancing not only histone acetylation but also fatty acid synthesis during HSC expansion in the early phase.

EPCR^High HSCs originally have the potential for stem cell expansion under a high metabolic status

Although low ΔΨ_m is generally required for the maintenance of HSCs in the steady state (Vannini et al, 2016, 2019), HSC expansion in the early phase after the administration of 5‐FU was induced under high ΔΨ_m (Fig 4J and K) (Umemoto et al, 2018). To address this discrepancy, we focused on the enhanced frequency of cells with high EPCR levels within the L⁻ESLAM fraction in the early phase (Fig 6A). The administration of 5‐FU in mice transplanted with EPCR^High HSCs (EPCR expression level: top 50% within the L‐ESLAM fraction) or EPCR^Low HSCs (EPCR expression level: lower 50% within the L‐ESLAM fraction) without irradiation significantly decreased the chimerism of donor cells only in mice with EPCR^Low HSC‐derived cells (Fig 6B). HSC numbers decreased immediately after the administration of 5‐FU (Fig 1A), suggesting that EPCR^High HSCs were enriched within BM in the early phase after the administration of 5‐FU due to reductions in EPCR^Low HSCs. EPCR⁺ HSCs (EPCR expression level: lower 30% within the L‐ESLAM fraction) showed the higher expression of genes related to the cell cycle than EPCR⁺⁺⁺ HSCs (EPCR expression level: top 30% within the L‐ESLAM fraction) (Figs 6A and EV4A). Moreover, EPCR⁺⁺⁺ HSCs exhibited a weaker proliferative capacity than EPCR⁺ HSCs in vitro (Fig EV4B) due to the inclusion of more clones with low proliferative activity (Fig EV4C). Since 5‐FU mainly targets proliferative cells, these results support 5‐FU preferentially eliminating low EPCR‐expressing HSCs rather than high EPCR‐expressing HSCs.

Figure 6. EPCR^High cells remaining after the administration of 5‐FU originally have the potential to maintain stem cell features under a high metabolic status.

1. Expression levels of EPCR in the L⁻ESLAM fraction in the steady state or early phase after the administration of 5‐FU. Indicated gates within histograms were selected based on the top, middle, or lower 30% EPCR expression level in HSCs in the steady state. (n = 4, two independent experiments, *P < 0.01 by the t‐test).
2. According to the indicated schedule, EPCR^High or EPCR^Low L⁻ESLAM HSCs (Ly5.2) were transplanted into non‐irradiated Ly5.1 mice, and the chimerism of donor‐derived cells in the L⁻ESLAM fraction was subsequently examined after the administration of PBS or 5‐FU. Indicated gates within the dot plot for EPCR versus c‐Kit were selected based on top or lower 50% EPCR expression levels in HSCs in the steady state. The graph depicts the frequency of donor (CD45.2⁺)‐derived cells within the L⁻ESLAM fraction. Data are shown as means ± SD (n = 4, two independent experiments, *P < 0.01, N.S., not significant by the t‐test).
3. The UMAP result of single‐cell ATAC‐seq using EPCR^High or EPCR^Low HSCs in the steady state, as defined in Fig 5B. Green and red plots represent the EPCR^High and EPCR^Low fractions, respectively. Numbers within brackets show the number used in this plot.
4. ATAC peaks overlapping with H3K27ac peaks detected in HSCs in each phase or CD48⁺ progenitor cells in the late phase were selected as cis‐regulatory regions.
5. A scatter plot comparing the peaks of cis‐regulatory regions between EPCR⁺⁺⁺ and EPCR⁺ HSCs in the steady state.
6. The enrichment of genes related to “accessibility‐increased regions in the EPCR⁺ fraction relative to the EPCR⁺⁺⁺ fraction (Purple dot in Fig 6E)” within genes up‐regulated in CD150⁺CD34⁻ HSCs or CD150⁺CD34⁺ progenitor cells.
7. Fifty indicated cells were transplanted with 2 × 10⁵ competitor cells into irradiated recipients. After 12 weeks, the chimerism of peripheral blood was analyzed. White or gray circles represent donor cell chimerism in multilineage‐reconstructed recipient mice (> 0.5% for myeloid and B‐ and T‐lymphoid lineages) or unreconstructed mice after 20 weeks, respectively. Black bars show the average (n ≥ 7, two independent experiments, *P < 0.01 by the t‐test).
8. Fold changes in the phenotypic HSC number after a 4‐day culture of the indicated HSC fractions. Data are shown as means ± SD (n = 4, two independent experiments, *P < 0.01 by the t‐test).
9. Venn diagrams represent the estimated frequencies of EPCR⁺⁺⁺, EPCR⁺⁺, and EPCR⁺ HSC‐derived cells within CD48⁻HSC or CD48⁺ progenitor cells after a 4‐day culture of HSCs derived from untreated mice. These frequencies were calculated based on the total cell number and frequency of each fraction after the culture of EPCR⁺⁺⁺, EPCR⁺⁺, and EPCR⁺ HSCs.
10. ΔΨ_m in CD48⁻ HSCs and CD48⁺ progenitor cells after a 4‐day culture of HSCs in the steady state. Data are shown as means ± SD (n ≥ 4, two independent experiments, *P < 0.01, N.S.; not significant by the t‐test).
11. One hundred indicated cells were transplanted with 2 × 10⁵ competitor cells into irradiated recipients. After 20 weeks, the chimerism of peripheral blood was analyzed. White or gray circles represent donor cell chimerism in multilineage‐reconstructed recipient mice (> 0.5% for myeloid and B‐ and T‐lymphoid lineages) or unreconstructed mice after 20 weeks, respectively. Black bars show the average (n ≥ 11, two independent experiments, *P < 0.01 by the t‐test).

Figure EV4. Low EPCR‐expressing HSCs showed higher proliferative activity and potential for the generation of CD48⁺ progenitor cells (Related to Fig 6).

• A
  The enrichment of the gene set “HALLMARK_G2M_CHECKPOINT” within differentially expressed genes of EPCR⁺⁺⁺ versus EPCR⁺ HSCs.
• B
  Total cell number after a 4‐day culture of EPCR⁺⁺⁺, EPCR⁺⁺, and EPCR⁺ HSCs. Data are shown as means ± SD (n = 4, two independent experiments, *P < 0.01 by the t‐test).
• C
  Cell number after a single‐cell culture for 4 days using EPCR⁺⁺⁺, EPCR⁺⁺, and EPCR⁺ HSCs. Data are shown as means ± SD (n = 3, three independent experiments, *P < 0.01, **P < 0.05 by the t‐test).
• D
  ΔΨ_m in HSCs obtained from untreated mice before or after 1 day of culture. Data are shown as means ± SD (n = 4, two independent experiments, *P < 0.01 by the t‐test).
• E
  The enrichment plots of genes related to oxidative phosphorylation in the comparison of the 1‐day culture with the steady state. NES and q values represent normalized enrichment scores and FDR, respectively.
• F, G
  Levels of Acly (F) and H3K27ac (G) in HSCs obtained from untreated mice before or after a 1‐day culture. Data are shown as means ± SD (n ≥ 3, two independent experiments, *P < 0.01 by the t‐test).
• H
  The frequencies of the indicated fractions after a 4‐day culture of each HSC subtraction divided based on EPCR expression levels. Data are shown as means ± SD (n = 4, two independent experiments, *P < 0.01 by the t‐test).
• I
  The frequencies of the CD150⁺CD48⁻ and CD48⁺ fractions within the lineage⁻EPCR⁺ fraction after a 4‐day culture of HSCs obtained from untreated mice with or without BMS303141 (BMS). Data are shown as means ± SD (n ≥ 6, two independent experiments, *P < 0.01, by the t‐test).
• J
  Global H3K27ac after 2 days of the culture of L‐ESLAM HSCs derived from untreated mice with or without 50 μM BMS303141 (BMS). Data are shown as means ± SD (n ≥ 4, two independent experiments, *P < 0.01, **P < 0.05 by the t‐test).

In terms of chromatin accessibility patterns, EPCR^High HSCs were clearly distinguished from EPCR^Low HSCs in the steady state, as evidenced by the results of single‐cell ATAC‐seq showing the differential distribution of EPCR^High and EPCR^Low HSCs in a plot of UMAP (Fig 6C). Furthermore, genes related to accessibility‐increased cis‐regulatory regions in EPCR⁺ HSCs relative to EPCR⁺⁺⁺ HSCs were significantly enriched among genes up‐regulated in progenitor cells (Fig 6D–F). EPCR⁺⁺⁺ HSCs also showed markedly higher LTR activity than EPCR⁺ HSCs (Fig 6G). Therefore, high EPCR‐expressing HSCs exhibited a greater stem cell potential than low EPCR‐expressing HSCs.

Importantly, when not only ΔΨ_m but also Acly and H3K27ac levels in HSCs obtained from untreated mice were enhanced during the ex vivo culture (Fig EV4D–G), EPCR⁺ HSCs provided more CD48⁺ progenitor cells. In contrast, EPCR⁺⁺⁺ HSCs showed an enhanced frequency of phenotypic HSCs (Fig EV4H), similar to HSCs in the early phase (Fig 1F; 6 days after the administration of 5‐FU). After the culture, EPCR⁺⁺⁺ HSCs showed a significantly increased phenotypic stem cell number, whereas the EPCR⁺⁺ and EPCR⁺ fractions did not (Fig 6H). Therefore, HSCs maintained under an ex vivo culture were mainly derived from high EPCR‐expressing HSCs (Fig 6I). Moreover, although these HSCs still showed similarly enhanced ΔΨ_m to generated CD48⁺ progenitor cells (Fig 6J), they engrafted in all irradiated recipients after transplantation (Fig 6K). These results suggest that high EPCR‐expressing HSCs originally exhibited not only higher LTR activity but also the potential to maintain stem cell features under high mitochondria‐Acly activity, thereby enabling HSC expansion under a high metabolism status in the early phase after the administration of 5‐FU.

The suppression of Acly‐mediated metabolism contributes to the maintenance of stem cell phenotypes in HSCs in the late phase

CD48⁺ progenitor cells generated during the ex vivo culture were mainly derived from low EPCR‐expressing HSCs (Figs 6I and EV4H). Importantly, CD48⁺ progenitor cell generation was suppressed by the Acly inhibitor along with H3K27ac levels (Fig EV4I and J). These results suggest that the suppression of Acly‐related metabolism contributed to the maintenance of stem cell phenotypes, particularly in low EPCR‐expressing HSCs rather than in high EPCR‐expressing HSCs. Similar to low EPCR‐expressing HSCs, HSCs in the late phase showed the increased accessibility of cis‐regulatory regions in progenitor cell‐related genes (Figs 3F and 6F), while HSC phenotypes were still maintained under suppressed mitochondria‐Acly activity (Fig 4). However, the CD48⁺ progenitor cells that appeared in the late phase showed higher levels of global H3K27ac, Acly, total acetylated lysine, and ΔΨ_m than maintained HSCs (Fig 7A–E). Importantly, CD48⁺ progenitor cells exhibited increased H3K27ac levels within the locus of up‐regulated genes, but not down‐regulated genes (Fig 7F). CD48⁺ progenitors showed higher H3K27ac levels in both the promoter and enhancer within the CD48 locus than CD48⁻ HSCs (Fig 7G) as well as increased mRNA expression levels (Fig 7H). Since the accessibility of both the promoter and enhancer in the CD48 locus was increased in HSCs from the early to late phase (Figs 3G and 7G), we compared H3K27ac levels between maintained CD48⁻ HSCs and CD48⁺ progenitor cells in the late phase by focusing on the chromatin accessibility pattern in HSCs. CD48⁺ progenitors showed higher H3K27ac levels within accessibility‐increased regions in HSCs in the late phase, but not within other regions, than CD48⁻ HSCs (Fig 7I). Moreover, the 192 regions extracted from these regions showed more than two‐fold higher H3K27ac levels in CD48⁺ progenitor cells than in maintained HSCs (Fig 7J: Venn diagram), and genes related to these H3K27ac‐increased regions were significantly enriched among genes up‐regulated in CD48⁺ progenitor cells (Fig 7J: GSEA plot). These results suggest that increased H3K27ac levels allowed HSCs in the late phase to differentiate into CD48⁺ progenitor cells.

Figure 7. CD48⁺ progenitor cells in the late phase show increased H3K27ac levels within accessibility‐increased regions in HSCs from the early to late phase.

• A–D
  Global H3K27ac (A), Acly levels (B), acetylated lysine (C), and ΔΨ_m (D) in CD48⁻ HSCs and CD48⁺ progenitor cells (HPCs) in the late phase after the administration of 5‐FU. Data are shown as means ± SD (n = 4, two independent experiments, *P < 0.01 by the t‐test).
• E
  The enrichment of the gene set “HALLMARK_OXIDATIVE_PHOSPHORYLATION” within differentially expressed genes between CD48⁻ HSCs and CD48⁺ progenitor cells. NES and q values represent normalized enrichment scores and FDR, respectively.
• F
  Average plot (top) and heatmap (bottom) of H3K27ac reads in CD48⁻ HSCs and CD48⁺ HPCs in the late phase within cis‐regulatory regions related to genes with lower (left) or higher (right) expression levels in CD48⁺ HPCs than in CD48⁻ HSCs.
• G
  ATAC peaks of HSCs in the early or late phase, and H3K27ac peaks in CD48⁻ HSCs or CD48⁺ HPCs in the late phase within the promoter and enhancer regions in the CD48 locus.
• H
  CD48 mRNA levels in CD48⁻ HSCs and CD48⁺ HSCs in the late phase after the administration of 5‐FU. Data are shown as means ± SD (n = 4, two independent experiments, *P < 0.01 by the t‐test).
• I
  Average plot (top) and heatmap (bottom) of H3K27ac reads in CD48⁻ HSCs and CD48⁺ HPCs in the late phase within accessibility‐decreased (left: Green dot in Fig 3E), ‐unchanged (center: Blue dot in Fig 3E), or ‐increased cis‐regulatory regions (right: brown dot in Fig 3E) in HSCs in the late phase.
• J
  Within accessibility‐increased cis‐regulatory regions in the late phase (brown dot in Fig 3E), 192 regions showed more than two‐fold increases in H3K27ac levels in CD48⁺ progenitor cells than in maintained HSCs (Venn diagrams). The GSEA plot represents the enrichment of the gene set related to these H3K27ac‐increased regions within differentially expressed genes between CD48⁻ HSCs and CD48⁺ progenitor cells. NES and q values represent normalized enrichment scores and FDR, respectively.

When HSCs in the late phase provided CD48⁺ progenitor cells during the ex vivo culture (Fig 1F), their ΔΨ_m and Acly levels were markedly enhanced (Fig 8A and B). Importantly, the treatment with BMS303141, SB204990 (an alternative inhibitor of Acly), and CTPI‐2 significantly suppressed the generation of CD48⁺ progenitor cells after the culture of HSCs in the late phase (Fig 8C). Although these inhibitors negatively affected cell proliferation (Fig 8D), the generation of CD48⁺ progenitor cells was further suppressed (Fig 8E). Therefore, Acly‐mediated metabolism contributes to the generation of CD48⁺ progenitor cells from HSCs in the late phase. These inhibitors simultaneously suppressed global H3K27ac during the culture (Fig 8F). Furthermore, A‐485, an inhibitor of HATs, suppressed the appearance of CD48⁺ cells, resulting in the maintenance of the CD48⁻ phenotype (Fig 8G). Similarly, A‐485 exerted negative effects on cell proliferation (Fig 8H); however, the suppressive effect on CD48⁺ cell numbers was greater (Fig 8I). Therefore, the suppression of Acly‐dependent histone acetylation prevents the generation of CD48⁺ progenitor cells from HSCs in the late phase.

Figure 8. The mitochondria‐Acly activity regulates the production of CD48⁺ progenitor cells during BM regeneration after the administration of 5‐FU.

• A, B
  ΔΨ_m (A) and Acly levels (B) in HSCs derived from 5‐FU treated mice in the late phase before and after a 1‐day culture. Data are shown as means ± SD (n ≥ 3, two independent experiments, *P < 0.01 by the t‐test).
• C
  The frequencies of the CD150⁺CD48⁻ and CD48⁺ fractions within the lineage⁻EPCR⁺ fraction after a 4‐day culture of HSCs in the late phase with or without BMS303141 (BMS; 100 μM), SB204990 (SB; 100 μM), an alternative Acly inhibitor, or CTPI‐2 (50 μM). Data are shown as means ± SD (n = 4, two independent experiments, *P < 0.01, by the t‐test).
• D, E
  Total cell number (D) or CD48⁺ cell number (E) after the 4‐day culture of HSCs in the late phase with or without BMS303141 (BMS), SB204990 (SB), or CTPI‐2. Numbers within the graph represent the inhibitory effects of each treatment (fold decrease by an inhibitor treatment). Data are shown as means ± SD (n = 4, two independent experiments, *P < 0.01, by the t‐test).
• F
  Global H3K27ac in HSCs in the late phase before or after a 1‐day culture with or without BMS303141 (BMS), SB204990 (SB), or CTPI‐2. Data are shown as means ± SD (n ≥ 4, two independent experiments, *P < 0.01, by the t‐test).
• G
  The frequencies of CD150⁺CD48⁻ and CD48⁺ fractions within lineage⁻EPCR⁺ after HSCs in the late phase were cultured for 4 days with or without A‐485 (1 μM), an inhibitor of histone acetyltransferase. Data are shown as means ± SD (n = 4, two independent experiments, *P < 0.01 by the t‐test).
• H, I
  Total cell number (H) or CD48⁺ cell number (I) after a 4‐day culture of HSCs in the late phase with or without A‐485. Numbers within the graph represent the inhibitory effects of each treatment (fold decrease by an inhibitor treatment). Data are shown as means ± SD (n = 4, two independent experiments, *P < 0.01, by the t‐test).
• J
  The frequencies of CD150⁺CD48⁻ and CD150⁺CD48⁺ fractions within lineage⁻EPCR⁺ after a serial CTPI‐2 treatment (50 mg/kg/shot/day) following the administration of 5‐FU in mice transplanted with 15,000 HSCs without irradiation. Data are shown as means ± SD (n = 4, two independent experiments, **P < 0.05 by the t‐test).

However, since HSCs in the late phase already showed lower Acly‐related metabolism, the mechanisms by which CD48⁺ progenitor cells in the late phase enhanced Acly‐mediated metabolism for their differentiation remain unclear. CD48⁺ progenitor cells started to significantly increase in number from 7 days after the administration of 5‐FU (Fig 1A). This timing was consistent with HSCs exhibiting a greater potential to generate CD48⁺ cells in vitro (Fig 1F). HSCs maintained higher levels of H3K27ac, Acly, total acetylated lysine, and ΔΨ_m in the intermediate phase, namely, 7–8 days after the administration of 5‐FU, than in the steady state or late phase (Fig 4A, D, I and J). These results suggest that the regulation of Acly‐mediated metabolism in HSCs in the intermediate phase contributed to the appearance of CD48⁺ progenitor cells in vivo. The treatment with BMS303141 or CTPI‐2 significantly decreased the number of CD48⁺ progenitor cells that appeared in the late phase (Fig EV5A and B). However, since HSC numbers similarly decreased (Fig EV5A and B), it remained unclear whether this metabolic suppression affects the generation of progenitor cells from HSCs in this context. To more clearly confirm this, we transplanted HSCs (Ly5.2) into Ly5.1 mice without irradiation and analyzed donor cells after the treatment with CTPI‐2 following the administration of 5‐FU (Fig 8J). This inhibitor treatment decreased the frequency of CD48⁺ progenitor cells and increased the phenotypic HSC frequency within donor‐derived cells (Fig 8J). These results suggest that Acly‐mediated metabolism contributes to the production of CD48⁺ progenitor cells from HSCs with an enhanced differentiation potential in the intermediate phase during BM regeneration.

Figure EV5. Effects of inhibitor treatments on HSCs in the late phase and the role of the mitochondria‐Acly axis in HSCs during BM regeneration (Related to Fig 8).

• A, B
  The effects of CTPI‐2 (A) or BMS303141 (BMS) (B) on the number of HSCs and CD48⁺ progenitor cells in the late phase after the administration of 5‐FU. CTPI‐2 (50 mg/kg/shot/day) or BMS (100 mg/kg/shot/day) was serially treated 6, 7, and 8 days after the administration of 5‐FU. Numbers in the dot plots represent the frequency of the indicated fractions within all BM cells. Data are shown as means ± SD (n = 4, two independent experiments, **P < 0.05 by the t‐test).
• C
  Proposed model (The steady state). HSCs show a quiescent state in vivo under low ΔΨ_m and Acly levels. EPCR^low HSCs have a chromatin accessibility pattern that enhances the differentiation potential, and EPCR^High HSCs have a higher self‐renewal capacity. (The early phase after the administration of 5‐FU) After the administration of 5‐FU, EPCR^High HSCs were enriched, resulting in self‐renewing divisions for HSC expansion by enhancing histone acetylation under high mitochondria‐Acly activity. (The late phase after the administration of 5‐FU) When HSCs gained an enhanced differentiation potential through the increased accessibility of progenitor cell‐related cis‐regulatory regions, the suppression of histone acetylation mediated by decreased mitochondria‐Acly activity contributed to the maintenance of stem cell phenotypes by preventing the expression of genes showing increased accessibility.

Discussion

We herein revealed two distinctive phases (the early phase and late phase) based on changes in the metabolic status of HSCs in BM recovery from myeloablation. HSC expansion in the early phase was induced under high ΔΨ_m, while HSCs with an enhanced differentiation potential showed low ΔΨ_m in the late phase (Fig EV5C). Previous studies, including ours, demonstrated that quiescent or self‐renewing HSCs showed low ΔΨ_m in steady‐state hematopoiesis (Vannini et al, 2016, 2019; Nakamura‐Ishizu et al, 2018). However, we showed that the level of ΔΨ_m was not a hallmark of HSCs, it was more dependent on the mitochondrial status based on this study on BM recovery. Importantly, we found that HSCs may be subclassified by the expression levels of EPCR. EPCR^High HSCs are more primitive than EPCR^Low HSCs and originally have the potential to maintain stem cell features even under enhanced ΔΨ_m (Figs 6 and EV4D–H). Moreover, EPCR^Low HSCs were more proliferative (Fig EV4C–E) (Rao et al, 2021). Since 5‐FU generally targets actively dividing cells, its administration enriched EPCR^High HSCs within BM in the early phase of BM recovery by decreasing EPCR^Low HSCs (Fig 6A and B). Therefore, the present results indicate that HSC expansion under enhanced ΔΨ_m in the early phase is enabled by EPCR^High HSCs without largely decreasing stem cell activity.

PAR1 signaling reportedly regulates the retention of EPCR‐expressing HSCs within BM (Gur‐Cohen et al, 2015). After myeloablation, cytokines within serum or BM or the behaviors and composition of niche cells markedly change (Zhao et al, 2014; Zhou et al, 2017; Chen et al, 2019; Severe et al, 2019; Sudo et al, 2021). These environmental changes may affect HSC retention within BM, resulting in the selection of some specific HSC subfractions.

Acly is a transferase that catalyzes the conversion of citrate to Acetyl‐CoA. It functions as a common substrate for not only histone acetylation but also de novo lipid synthesis. We herein identified Acly as an essential regulator of stem cell functions during BM regeneration through the modification of histone acetylation. In the early phase, enhanced mitochondria‐Acly activity contributes to the expansion of HSCs through not only histone acetylation but also fatty acid synthesis (Figs 4 and 5). Previous studies reported that a histone deacetylase inhibitor contributed to human HSC expansion (De Felice et al, 2005; Chagraoui et al, 2021), which supports Acly‐mediated H3K27ac in HSCs being required for stem cell expansion in the early phase. Moreover, HSCs in the early phase showed the up‐regulated expression of genes related to fatty acid synthesis (Fig 5H), accompanied by enhanced H3K27ac levels within the locus of these genes (Fig 5I). Therefore, Acly‐synthesized acetyl‐CoA in these HSCs may contribute to fatty acid synthesis not only as a source of fatty acids but also through histone acetylation to activate the expression of its related genes.

In contrast, in the late phase of recovery, HSCs showed an enhanced differentiation potential, as evidenced by chromatin accessibility pattern (Fig 3F) as well as the behavior in vitro (Fig 1F). However, these HSCs enable the maintenance of stem cell phenotypes under low mitochondria‐Acly metabolic pathway, since poised cis‐regulatory regions in progenitor cell‐related genes remain inactive due to insufficient histone acetylation activity (Figs 3, 4 and 7). On the other hand, HSCs in the intermediate phase also showed an enhanced differentiation potential (Fig 1F), but still increased mitochondrial‐Acly activity (Fig 5A, D and K). CD48⁺ progenitor cell numbers within BM started to significantly increase, particularly from this phase (Fig 1A), and the inhibition of Acly‐mediated metabolism suppressed the generation of CD150⁺CD48⁺ progenitor cells from HSCs in this phase (Fig 8J). Therefore, the present results suggest that the regulation of Acly‐mediated metabolism is essential for fate decisions in HSCs enhancing the differentiation potential during BM regeneration. Previous studies reported that extracellular adenosine contributes to the suppression of ΔΨm in HSCs (Hirata et al, 2018; Umemoto et al, 2018). In the present study, we found that lineage⁻EPCR⁻c‐Kit⁺ cell numbers markedly decreased in the early phase after the administration of 5‐FU, but recovered in the late phase (Fig 1A). These cells provide adenosine to suppress ΔΨm in HSCs (Umemoto et al, 2018). Therefore, an adenosine‐related system may be involved in the suppression of ΔΨm in HSCs in the late phase, similar to the steady state.

The ex vivo culture showed that Acly suppression decreased CD48⁺ progenitor cell production from HSCs not only in the late phase of BM regeneration (Fig 8C) but also in the steady state (Fig EV4I). In the case of HSCs obtained from untreated mice, CD48⁺ progenitor cells were mainly derived from low EPCR‐expressing HSCs (Fig 6I). Moreover, low EPCR‐expressing HSCs showed the increased accessibility of the cis‐regulatory regions of progenitor‐associated genes (Fig 6E and F), similar to HSCs in the late phase of BM recovery (Fig 3E and F). These results support the key role of Acly suppression in maintaining stem cell phenotypes, particularly in HSCs with an enhanced differentiation potential based on the increased accessibility of cis‐regulatory regions in progenitor cell‐associated genes. The HSC fraction is reportedly not only functionally but also epigenetically heterogenous (Figs 6C–K and EV4H) (Wilson et al, 2015; Morcos et al, 2017; Giladi et al, 2018). Therefore, the low metabolic status in quiescent HSCs may maintain a diversity within the stem cell compartment by preventing the differentiation of stem cells especially in HSCs with enhanced differentiation potential. Similar to HSCs in the steady state, HSCs in the late phase showed suppressed Acly levels (Fig 4C–E), and simultaneously contained more stem cells having the potential to provide progenitor cells than those in the early phase (Fig 1F). Therefore, the HSC compartment may start to reconstruct functional heterogeneity from the late phase during BM regeneration.

Although the kinetics of mRNA levels in HSCs correlated with protein levels (Fig 5B and C), the increase observed in protein levels was slightly greater than that in mRNA levels in the early phase. The acetylation of Acly reportedly results in protein stabilization through the prevention of ubiquitination (Lin et al, 2013). Therefore, when mitochondrial metabolism supports the Acly‐mediated synthesis of Acetyl‐CoA through the supply of citrate (Fig 4H and J), this protein may be simultaneously stabilized via acetylation, resulting in enhanced protein levels. Therefore, mitochondrial activity may play a key role in the positive regulation of Acly‐mediated reactions in HSCs through not only the supply of citrate (a substrate) but also increases in Acly protein levels.

Between the early and late phases during BM regeneration, HSCs showed increased accessibility on the enhancer regions of Runx1, Runx2, and Runx3, which are activated in MPPs rather than in HSCs (Fig EV2C). Since the late phase showed the significant enrichment of the Runx‐binding motif within accessibility‐increased regions from the early phase (Figs 3E and EV2B), the modified expression of the Runx family in response to epigenetic changes in the late phase may be a predisposition for HSC differentiation. Runx1 and 3 expression in HSCs has been extensively examined, and each Runx member is one of the key regulators in HSCs (Kuo et al, 2009; Cai et al, 2011, 2015; Wang et al, 2014). In addition, Runx1 binds to the Scl complex (Wilson et al, 2010). Moreover, PU.1 is a well‐known downstream target of Runx1 (Huang et al, 2008). Furthermore, accessibility‐increased regions from the early to late phase showed the significant enrichment of the Scl as well as PU.1‐binding motif (Figs 3E and EV2B). Therefore, the present results suggest that the transcription network of the Runx family, Scl and PU.1, functions as a switch board for HSC differentiation division.

In conclusion, we herein provide an important insight into HSC regulation during BM regeneration from the aspect of metabolic and epigenetic regulation. Acly is required for stem cell expansion in the early phase of BM regeneration by HSCs with higher LTR activity. Acly allowed HSCs with increased accessibility to the cis‐regulatory regions of progenitor cell‐related genes to differentiate into progenitor cells through histone acetylation. Importantly, a low metabolic status appears to be required in order to maintain stem cell phenotypes, particularly in HSCs showing chromatin accessibility patterns enhancing the differentiation potential, so‐called “primed HSCs.” Therefore, we demonstrated that metabolic regulation in accordance with the differentiation potential of HSCs is critical for stem cell regulation. The manipulation of HSCs based on the present results may contribute to the development of not only the ex vivo expansion of HSCs for therapeutic applications but also therapy promoting BM regeneration after BM transplantation or treatments with anti‐cancer drugs.

Materials and Methods

Animal experiments

C57BL/6‐Ly5.2 and C57BL/6‐Ly5.1 mice were obtained from Sankyo Lab Service Corporation (Tokyo, Japan) or Japan SLC Inc. (Shizuoka, Japan), respectively. Each strain used was between 8 and 12 weeks of age. As previously described (Umemoto et al, 2018), mice were intravenously administered 250 mg/kg 5‐FU (Kyowa Hakko Kirin, Tokyo, Japan). To inhibit mitochondrial citrate transporters or HATs, CTPI‐2 (MedChemExpress, Monmouth Junction, NJ) or A‐485 (MedChemExpress) was administered at 50 or 20 mg/kg/day i.p., respectively, for 3 days. To inhibit Acly or Fasn in vivo, BMB303141 (Selleck, Houston, TX) or TVB‐3664 (Selleck) was administered at 100 or 10 mg/kg/day p.o. for 3 days.

Antibodies for flow cytometry

The following monoclonal antibodies were used in cell sorting and flow cytometric analyses: anti‐c‐Kit (2B8), anti‐CD150 (TC15‐12F12.2), anti‐CD48 (HM48‐1), anti‐EPCR (RAM34, eBioscience, San Diego, CA), anti‐Sca1 (E13‐161.7), anti‐CD45.2 (104), anti‐CD45.1 (A20), anti‐B220/CD45R (RA3‐6B2), anti‐Mac‐1 (M1/70), anti‐Gr‐1 (RB6‐8C5), anti‐CD4 (RM4‐5), anti‐CD8 (53‐6.72), anti‐H3K27ac (D5E4, Cell Signaling Technology, Danvers, MA), anti‐Acly (EP704Y: Abcam, Cambridge, UK), anti‐Acly (phospho‐Ser455) (Biobyt, Cambridge, UK), and anti‐acetylated lysine (15G10) antibodies. All antibodies were obtained from BioLegend (San Diego, CA) unless otherwise noted.

Cell sorting and flow cytometric analyses

We used Moflo XDP (Beckman Coulter, Brea, CA) for cell sorting, as described previously (Umemoto et al, 2018). In addition, FACS CANTO II (BD Biosciences, Franklin Lakes, NJ) or Moflo XDP was used for flow cytometric analyses of freshly uncultured or cultured cells, respectively. In the case of H3K27ac, prior to staining with the antibody, cells were fixed by 4% paraformaldehyde and subsequently permeabilized by 100% methanol. In the case of Acly, phosphorylated Acly, and acetylated lysine, we used a PerFix EXPOSE kit (Beckman Coulter) for cell fixation and permeabilization according to the manufacturer’s instructions.

RNA sequencing

As previously described (Umemoto et al, 2017, 2018; Hayashi et al, 2018), we prepared libraries for RNA‐Seq and conducted deep sequencing using the Next‐Seq 550 or Hi‐seq 2000 system (Illumina Inc., San Diego, CA). Using these transcriptome data, we performed a principal component analysis or gene set enrichment analysis (GSEA) using CLC genomic workbench v11.0.0 (Qiagen, Aarhus, Denmark) and GSEA v3.0.0 software (http://www.broad.mit.edu/gsea) (Subramanian et al, 2005), respectively. All gene sets used in GSEA were obtained from the database of the Broad Institute unless otherwise stated. All gene expression profiles used in GSEA in the present study were derived from RNA‐Seq analyses deposited under the above accession number.

EdC uptake assay

As previously reported (Umemoto et al, 2018), untreated or 5‐FU‐treated mice were administered EdC (150 mg/kg i.p.) (Tokyo Chemical Industry Co., Tokyo, Japan) 24 h prior to the assessment of EdC uptake. Cells showing EdC uptake were identified by the Click‐iT^® Plus EdU Alexa Fluor^® 488 Flow Cytometry Assay Kit (Thermo Fisher Scientific, Waltham, MA), according to the manufacturer’s instructions.

HSC cultures

L⁻ESLAM HSCs were sorted and cultured in S‐Clone SF‐03 medium (Sanko‐Junyaku Co., Tokyo, Japan) supplemented with 10 ng/ml mouse stem cell factor (R&D Systems, Minneapolis, MN), 100 ng/ml mouse thrombopoietin (R&D Systems), and 0.5% fatty acid‐free bovine serum albumin (BSA) (Proliant Biologicals, Ankeny, IA). To suppress Acly, 100 μM BMS303141 (Cayman Chemical, Ann Arbor, MI) or 100 μM SB204990 (R&D Systems) was used. Moreover, to inhibit mitochondrial citrate transporters or HATs, 50 μM CTPI‐2 (MedChemExpress) or 1 μM A‐485 (Selleck) was added to the medium. In the single‐cell culture, the indicated fraction was clonally cultured into 96‐well U‐bottomed plates, and cell numbers were counted under a light microscope 4 days later.

Transplantation of HSCs into non‐irradiated mice

After sorting, 1 × 10⁴ lineage⁻ EPCR⁺CD150⁺CD48⁻ (L⁻ESLAM) HSCs were stained with 2 μM CFSE (Thermo Fisher Scientific, Waltham, MA) or 1/2,000‐diluted CytoTell^TMGreen (AAT Bioquest, Sunnyvale, CA) for 10 min. Labeled cells were transplanted without irradiation into untreated or 5‐FU‐treated mice, and subsequently analyzed after 2 days. When cell division number was estimated, the highest fluorescent peak served as undivided cells (0 divisions) after some labeled HSCs in each experiment were cultured under undivided conditions as previously described (Umemoto et al, 2018).

Long‐term competitive repopulation assays

As described previously (Umemoto et al, 2018), sorted cells derived from C57BL/6‐Ly5.2 were i.v. transplanted along with 2 × 10⁵ competitor cells derived from C57BL/6‐Ly5.1 congenic mice into lethally irradiated (total of 10 Gy) C57BL/6‐Ly5.2 mice. Twenty weeks after transplantation, recipient mice with donor cell chimerism (> 0.5% for myeloid and B‐ and T‐lymphoid lineages) were considered to be multilineage‐reconstituted mice.

Omni‐ATAC‐seq

Omni‐ATAC‐seq was performed as previously described with minor modifications (Corces et al, 2017). In brief, sorted HSCs were lysed with ATAC‐resuspension buffer containing detergents, which are listed in a previous study. The transposition reaction was performed with TDE1 (Illumina Inc.). After optical amplification by PCR using ATAC‐primers, which are listed in a previous study (Buenrostro et al, 2015), the purification of ATAC‐seq, deep sequencing, trimming, alignment (to mouse genome mm10), peak calling, and quantification were performed by Illumina NextSeq 500 Platform with 38 bps paired‐end reads, Trimmomatic, Bowtie2, MACS2, and HOMER, respectively. HOMER was also used for de novo motif discovery in accessibility‐increased or ‐decreased regions. The ChIP‐seq data of H3K27ac and transcription factors were downloaded from GSE59636 and GSE22178, respectively, and reanalyzed (Lara‐Astiaso et al, 2014; Wu et al, 2014).

CUT & Tag assay

The CUT & Tag assay was performed using CUTANA pAG‐Tn5 (EpiCypher, Durham, NC) according to the manufacturer’s protocol. Briefly, 12,500 sorted cells were lysed with NE buffer (20 mM HEPES‐KOH pH 7.9, 10 mM KCl, 0.1% Triton X‐100, 20% glycerol, 0.5 mM Spermidine, and 1× Roche Complete Protease Inhibitor) to isolate nuclei. The nuclei obtained were bound to BioMag Plus Concanavalin A (ConA) beads (Bang Laboratories, Fishers, IN), followed by an incubation with an anti‐H3K27ac antibody (20 μg/ml; MABI0309, Active Motif, Carlsbad, CA) at 4°C overnight. After the immunoprecipitation reaction, these samples were further incubated with an anti‐mouse IgG antibody (10 μg/ml; Antibodies‐online.com) at room temperature for 30 min, prior to the treatment with CUTANA pAG‐Tn5 at room temperature for 1 h. Samples were then resuspended in Tagmentation buffer (20 mM HEPES‐NaOH pH 7.5, 300 mM NaCl, 10 mM MgCl₂, 0.5 mM Spermidine, and 1× Roche Complete Protease Inhibitor) and incubated at 37°C for 1 h. Samples were then resuspended in 5 µl of SDS release buffer (10 mM TAPS, 0.1% SDS) and incubated at 58°C for 1 h to extract tagmented chromatin. SDS quench buffer (0.67% Triton X‐100) was added to samples, followed by PCR amplification with indexed primers. CUT & Tag libraries were cleaned with SPRIselect beads (Beckman Coulter) and subjected to sequencing using the Illumina NextSeq 500 Platform with 75‐bp single‐end reads. Trimming, alignment (to mouse genome mm10), peak calling, and quantification were performed by Trimmomatic, Bowtie2, and HOMER, respectively. ATAC peaks overlapping with H3K27ac peaks observed in HSCs in the steady state, early phase, or late phase were selected by bedtools and used as cis‐regulatory regions (Figs 3D and 6G) in subsequent analyses.

Measurement of glucose uptake

As previously described (Umemoto et al, 2018), EPCR⁺ cells enriched by AutoMACS pro (Miltenyi Biotec, North Rhine‐Westphalia, Germany) or whole cultured cells were cultured with 2‐NBDG (2‐deoxy‐2‐[(7‐nitro‐2,1,3‐benzoxadiazol‐4‐yl)amino]‐D‐glucose) (BioVision Milpitas, CA) at 37°C for 30 min. After staining for the identification of L⁻ESLAM cells, the potential for glucose uptake in L⁻ESLAM cells was assessed based on the fluorescence intensity of 2‐NBDG using a flow cytometer.

Measurement of intracellular Glu

After sorted cells were denatured by 0.1 N HCl, o‐phthaldialdehyde, 3‐mercaptopropionic acid, and boric acid were added to the cell lysate to fluorescently label free amino acids. Glu levels were detected using the NexeraX2 high‐performance liquid chromatography system (Shimadzu, Kyoto, Japan).

Measurement of citrate

After sorting 50,000 HSCs, citrate levels were assessed using Citrate Assay Kit according to the manufacturer’s instructions (Abcam).

Measurement of ΔΨ_m

As previously reported (Umemoto et al, 2018), to assess ΔΨ_m, cells were stained with 2 μM JC‐1 dye (Thermo Fisher Scientific, Waltham, MA) at 37°C for 30 min, followed by flow cytometric analyses.

Single‐cell ATAC‐Seq

According to the manufacturer’s instructions, sorted HSCs were resuspended in lysis buffer and incubated on ice for 5 min. Lysed cells were washed, strained, and nuclei were resuspended. Isolated nuclei were then used as input following the 10× Genomics Chromium Next GEM single‐cell ATAC Reagent Kits v1.1 manual (10× Genomics, Pleasanton, CA). With a target of 5,000 nuclei recovery, samples were added to the tagmentation reaction, loaded into the Chromium Controller for nuclei barcoding, and prepared for library construction following the manufacturer’s protocol. The resulting libraries were quantified using the KAPA Library Quantification Kit for Illumina platforms (KAPA Biosystems, Wilmington, MA), and sequenced with PE34 sequencing on the NextSeq 500/550 sequencer (Illumina). Sequenced data were processed using Cell Ranger ATAC software, with alignment to the mouse (mm10) genome. Duplicate reads were removed, and only reads mapping as matched pairs and only uniquely mapped reads (mapping quality ≥ 1) were used for further analyses. Alignments were extended in silico at their 3′‐ends to a length of 200 bp and assigned to 32‐nt bins along the genome. Peaks were identified using the MACS 2.1.0 algorithm. Peaks that were on the ENCODE blacklist of known false ChIP‐Seq peaks were removed. Signal maps and peak locations were used as input data to the Active Motif’s proprietary analysis program.

Statistical analysis

Statistical analysis was performed by using an unpaired two‐sided t‐test (GraphPad Prism or Microsoft Excel). Error bars represent standard deviation (SD) from more than three independent biological replicates obtained from more than two independent experiments. Significance was assumed with *P < 0.01, **P < 0.05, N.S., not significant.

Author contributions

Terumasa Umemoto: Conceptualization; Data curation; Formal analysis; Funding acquisition; Validation; Investigation; Visualization; Methodology; Writing—original draft; Project administration; Writing—review & editing. Alban Johansson: Investigation. Shah Adil Ishtiyaq Ahmad: Investigation. Michihiro Hashimoto: Investigation. Sho Kubota: Investigation; Methodology. Kenta Kikuchi: Investigation; Methodology. Haruki Odaka: Investigation; Methodology. Takumi Era: Investigation; Methodology. Daisuke Kurotaki: Investigation; Methodology. Goro Sashida: Supervision; Investigation; Methodology. Toshio Suda: Supervision; Writing—review & editing.

In addition to the CRediT author contributions listed above, the contributions in detail are:

TU and TS designed the present study and wrote the manuscript. TU performed most of the experiments. AJ, SAIA, and MH assisted with several experiments. SK and GS performed ATAC‐Seq. KK and DK performed the CUT & Tag assay. HO and TE measured intracellular Glu levels.

Disclosure and competing interests statement

The authors declare that they have no conflict of interest.

Supporting information

Expanded View Figures PDF

The EMBO Journal (2022) 41: e109463.

Data availability

The datasets and computer code produced in this study are available in the following databases.

• RNA‐seq data: Gene Expression Omnibus GSE164283 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE164283), GSE187376 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE187376), and GSE186956 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE186956).

• ATAC‐seq data: Gene Expression Omnibus GSE164281 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE164281) and GSE187014 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE187014).

• ChIP‐seq data: Gene Expression Omnibus GSE190148 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE190148).

• Single‐cell ATAC‐seq data: Gene Expression Omnibus GSE187444 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE187444).