Ca²⁺–mitochondria axis drives cell division in hematopoietic stem cells

Hematopoietic stem cells (HSCs) play a key role in the lifelong maintenance of hematopoiesis through self-renewal and multilineage differentiation. Adult HSCs reside within a specialized microenvironment of the bone marrow (BM), called “niche,” in which they are maintained in a quiescent state. Because the loss of HSC quiescence leads to the exhaustion or aging of stem cells through excess cell division, the maintenance of quiescence in HSCs is essential for hematopoietic homeostasis (Mendelson and Frenette, 2014). A feature of quiescent HSCs is their low baseline energy production; quiescent HSCs exhibit low mitochondria membrane potentials (ΔΨ_m) and rely on glycolysis (Suda et al., 2011; Ito and Suda, 2014). Likewise, HSCs with a low ΔΨ_m show greater engraftment, compared with cells with high ΔΨ_m (Vannini et al., 2016). These reports exhibit that the maintenance of quiescent HSCs do not rely on mitochondrial metabolism.

Upon stress hematopoiesis, HSCs are forced to exit quiescence and either self-renew or differentiate to mature hematopoietic cells. HSCs exit quiescence and actively cycle upon interferon treatment or 5-fluoruracil (5-FU)–induced BM suppression (Harrison and Lerner, 1991; Essers et al., 2009; Baldridge et al., 2010). The mechanism that determines whether HSCs self-renew or differentiate during stress hematopoiesis remains unclear. The study on the activation of HSCs has not been progressed much compared with quiescent HSCs. Indeed, in addition to the low frequency of active HSCs at steady-state, a definition or prospective marker that distinguishes between quiescent and active HSCs at steady-state has not been well established. Moreover, stress hematopoietic events change the phenotypes of HSCs in BM, thereby making the accurate identification of HSCs in numbers difficult (Pietras et al., 2014), which appears to constitute a bottleneck in the study concerning active HSCs.

The influx of Ca²⁺ into mitochondria is required for the activation of mitochondria (Hajnóczky et al., 1995; Jouaville et al., 1999). Since the up-regulation of intracellular Ca²⁺ level triggers mitochondrial Ca²⁺ level (Hajnóczky et al., 1995), the control of the former appears to play a key role in mitochondrial activity. Intracellular Ca²⁺ level is regulated by ER-mediated release/uptake of Ca²⁺, Ca²⁺ channel–mediated influx, and the efflux by Ca²⁺ pump or Na⁺/Ca²⁺ exchanger. Recently, purine receptors including P2X, P2Y and adenosine receptors were reported to be involved in the regulation of intracellular Ca²⁺ (Ralevic and Burnstock, 1998; Svenningsson et al., 1999; Jiang et al., 2017). Although P2Y14 receptor is known for regulating HSCs under stress (Cho et al., 2014), the role of Ca²⁺ level in HSC maintenance still remains largely unknown.

In this study, we elucidated the mechanism underlying the initiation of cell division in HSC during stress hematopoiesis. We mainly focus on the change of energy metabolism in HSCs after BM suppression following 5-FU administration. While quiescent HSCs show low ΔΨ_m, enhanced ΔΨ_m as a result of increased intracellular Ca²⁺ level is required for HSC division in vivo and in vitro. Moreover, we found that extracellular adenosine negatively regulates ΔΨ_m of HSCs after 5-FU administration. Importantly, when HSC divisions were induced, the appropriate suppression of ΔΨ_m achieved both cell division and the maintenance of HSC functions. Our data indicate that the Ca²⁺–mitochondria pathway plays a key role not only in initiating HSC divisions but also determining self-renewing or differentiation divisions.

To examine the mechanism underlying HSC cell cycle entry, we first focused on the change of a HSC population after BM suppression following 5-FU administration. Although CD150⁺CD48⁻c-Kit⁺Sca-1⁻lineage⁻ (CD150⁺CD48⁻ KSL; SLAM KSL) cells have been regarded as one of most reliable fractions for HSC identification, these cells were drastically reduced at 4 d after 5-FU administration (Fig. 1 A). All mice treated with this dose (250 mg/kg) of 5-FU could survive for >3 mo (unpublished data), and it is likely that 5-FU administration appears to alter the expression pattern of Sca-1 or c-Kit in HSCs rather than the drastic depletion of HSCs. To circumvent the change in HSC surface marker phenotype during the recovery from 5-FU–induced BM suppression, we used Endothelial protein C receptor (EPCR)-based fraction, lineage⁻EPCR⁺CD150⁺CD48⁻ (L⁻ESLAM), for the identification of HSCs, since we and others previously found that EPCR contributes to the accurate identification of HSCs without relying on Sca-1 or c-Kit (Kent et al., 2009; Umemoto et al., 2017). L⁻ESLAM population was certainly decreased by 5-FU administration, but could be stably detected (Fig. 1 B). Indeed, most of the L⁻ESLAM population showed Sca-1⁺c-Kit⁺ phenotype in untreated mice, while L⁻ESLAM cells in 5-FU–administrated mice were different from Sca-1⁺c-Kit⁺ gate (Fig. 1 C). However, L⁻ESLAM cells started to regain Sca-1⁺c-Kit⁺ phenotype at 10 d after 5-FU administration and almost completed after 14 d (Fig. 1 C). Moreover, c-kit^Low L⁻ESLAM cells derived from mice treated with 5-FU after 4 d showed similar engraftment and lineage contributions compared with c-kit^High L⁻ESLAM cells obtained from untreated mice after primary and secondary transplantation (Fig. 1, C–F). These data indicate that L⁻ESLAM fraction not only more accurately determines HSCs in 5-FU–treated mice rather than a fraction based on Sca-1⁺c-Kit⁺ population, but also equivalently identify stem cells both before and after 5-FU administration without any changes of gating.

We next examined the change of HSC (L⁻ESLAM) number after 5-FU administration. Although BM nucleated cell number was decreased at 4 d after 5-FU treatment (Fig. 1 G), the number of L⁻ESLAM HSCs recovered from day 3 and later (Fig. 1 H). Consistent with these results, uptake of EdC in L⁻ESLAM cells was significantly increased from 3 d after 5-FU administration (Fig. 1 I). These data indicate that HSC division is initiated from between 3 and 4 d after 5-FU administration.

Indeed, ∼80% or >95% of L⁻ESLAM cells show the uptake of EdC for 24 (Fig. 1 I) or 48 h (Fig. S1 A) from 4 d after 5-FU administration, respectively, which suggests that most of L⁻ESLAM cells in mice treated with 5-FU after 4 d prospectively initiate cell division. Therefore, to examine the mechanism of HSC division initiation after 5-FU administration, “quiescent” HSCs (derived from untreated mice) and “cycling” HSCs (derived from mice treated with 5-FU after 4 d) were subjected to RNA sequencing (RNA-seq; Fig. 2 A). Consistent with the cell cycle state of HSCs, gene set enrichment analysis (GSEA) showed that gene sets related to “mitotic spindle” and “G2M check point” were enriched within up-regulated genes in cycling HSCs (Table 1 and Fig. 2 B). In addition, up-regulated genes in cycling HSCs showed significant enrichment of gene sets related to “mTOR1 signaling,” “glycolysis,” “reactive oxygen species,” and “oxidative phosphorylation” (Table 1 and Fig. 2 B). These data suggest that the cell cycle entry of HSCs is associated with a change of energy metabolism. Since “quiescent” L⁻ESLAM HSCs originally exhibited low ΔΨ_m compared with other progenitor or mature cell fractions at steady-state (Fig. S1 B), we examined whether mitochondrial activity was changed between “cycling” and “quiescent” HSCs. To analyze the mitochondrial function of HSCs after 5-FU administration, we used red fluorescence of JC-1 (JC-1 Red), an indicator for ΔΨ_m, and found that ΔΨ_m was significantly enhanced in “cycling” HSCs at 4 d after 5-FU administration compared to “quiescent” HSCs (Fig. 2 C). Simultaneously, “cycling” HSCs showed increased green fluorescence of JC-1 (JC-1 Green), which reflects mitochondrial mass, compared with “quiescent” HSCs (Fig. S1 C). Moreover, mitochondrial superoxide level and intracellular ATP content was also increased in “cycling” HSCs (Fig. 2, D and E). Furthermore, “cycling” HSCs appear to show increased glycolysis, as indicated by up-regulated uptake of 2-deoxy-2-[(7-nitro-2,1,3-benzoxadiazol-4-yl)amino]-d-glucose (2-NBDG), known as a fluorescent glucose analogue (Fig. S1 D). These data indicated that “cycling” HSCs derived from mice treated with 5-FU after 4 d show more activation of energy metabolism through the enhancement of mitochondrial functions along with glycolysis, compared with “quiescent” HSCs. Importantly, the time course of ΔΨ_m up-regulation (Fig. 2 C) and EdC uptake (Fig. 1 F) indicates that mitochondrial function in HSCs is enhanced immediately before the initiation of cell division.

As the influx of Ca²⁺ into mitochondria is essential for the enhancement of mitochondrial activity (Hajnóczky et al., 1995; Jouaville et al., 1999), we next examined changes in intracellular or mitochondrial Ca²⁺ level in HSCs after 5-FU administration by using Flou-4 or Rhod-2 staining, respectively (Hawkins et al., 2010; Fukumori et al., 2013; Bidaux et al., 2015). A gradual increase in intracellular Ca²⁺ level observed in HSCs after 5-FU treatment (Fig. 2 F) with significant increase in mitochondrial Ca²⁺ level at 3 d (Fig. 2 G), which coincided with enhanced ΔΨ_m (Fig. 2 C). These data suggest that a simultaneous increase in both ΔΨ_m and mitochondrial Ca²⁺ level following enhanced intracellular Ca²⁺ level is required for quiescent HSCs to initiate cell division.

Next, we examined the relationship between ΔΨ_m and intracellular Ca²⁺ level, when HSC division is induced by cytokine stimulation using the combination of stem cell factor (SCF) and thrombopoietin (TPO) in vitro. Under culture conditions that most of HSCs show more than one cell division within 48 h (5 or 50 ng/ml; Fig. 3 A), ΔΨ_m of HSCs was significantly enhanced, which was dependent on the length of culture time (Fig. 3 B). Even culture at low concentration of cytokines (0.05 or 0.5 ng/ml) for 6 h also led to enhanced ΔΨ_m in HSCs compared with uncultured state (Fig. 3 B). However, increase in ΔΨ_m ceased at 18 h later (Fig. 3 B) and HSCs neither survive (0.05 ng/ml; Fig. 3 C) nor show further cell division (0.5 ng/ml) beyond this point (Fig. 3 A). Interestingly, the magnitude of ΔΨ_m during each culture for 18 h greatly correlated to intracellular Ca²⁺ level (Fig. 3, B, D, and E) in vitro. Similarly, mitochondrial Ca²⁺ level was also increased along with intracellular Ca²⁺ level (Fig. 3 F). Although only 20% of HSCs divided at 24 h (Fig. 3 A), intracellular and mitochondrial Ca²⁺ level and ΔΨ_m in most of HSCs already increased at 18 h of culture (Fig. 3, B, D, and F). These data support that both intracellular Ca²⁺ level and mitochondrial activity predisposed HSC division.

Next, to examine the role of Ca²⁺–mitochondria pathway for HSC division, we treated cultured HSCs with Nifedipine, a blocker of L-type voltage-gated Ca²⁺ channels (LTCCs). Interestingly, Nifedipine significantly cancelled the increase of intracellular Ca²⁺ (Fig. 3 G) as well as mitochondrial functions including ΔΨ_m (Fig. 3 H and Fig. S2) under conditions that induced HSC division. Similarly, Isradipine, alternative blocker of LTCCs, also suppressed both intracellular Ca²⁺ level and mitochondrial activity (Fig. S3). These data indicate that the activation of Ca²⁺–mitochondria pathway is mainly regulated through extracellular Ca²⁺ intake. Moreover, HSC division was significantly suppressed by Nifedipine treatment (Fig. 3 I). However, CFSE dilution assay revealed that most HSCs show more than one cell division after 4 d culture even in the presence of Nifedipine, but less frequently divide in a low cytokine culture (Fig. 3 J), suggesting that Nifedipine prolonged cell division intervals.

To examine the mechanism of how Nifedipine treatment prolongs the interval of cell divisions, we tested the expression of cell cycle–related genes in HSCs cultured with or without Nifedipine. Importantly, Nifedipine treatment significantly suppressed late G1 phase–related cyclins (Cyclin E; Ccne1 and Ccne2) without negatively affecting the expression of early G1 phase–related cyclins (Cyclin D; Ccnd1, Ccnd2, and Ccnd3; Fig. 4 A). In addition, the expression of Cdkn1a, also known as p21, was enhanced in Nifedipine-treated HSCs (Fig. 4 B). Consistent with these result, Nifedipine treatment decreased the phosphorylation of CDK4, CDK6, and Rb, which are known to induce Cyclin E at the late G1 phase (Fig. 4 C). These data indicate that the suppression of Ca²⁺–mitochondria pathway during HSC division decreases the expression of Cyclin E through suppression of phosphorylation of CDK4/6-Rb axis, possibly prolonging cell division intervals.

To examine whether Nifedipine-mediated prolonged interval of cell division affects HSC function, we analyzed the effect of cell division on HSCs using CFSE-labeling. Although HSCs cultured for 4 d under control conditions showed significant increase in CD48 expression as cell division number was increased, the addition of Nifedipine maintained a low level of CD48 expression after cell divisions (Fig. 5 A). 2-d culture with untreated controls still increased CD48 expression, even though the CFSE pattern was similar to 4-d culture with Nifedipine (Fig. 5 A). Similarly, the frequency of phenotypic HSCs within three-cell-division population was significantly decreased under control conditions (for 2 d), compared with one-cell-division population, whereas Nifedipine-treated HSCs (for 4 d) maintained ESLAM (EPCR⁺CD150⁺CD48⁻) KSL phenotype even after three cell divisions (Fig. 5 B). Although L⁻ESLAM fraction was used to determine HSCs after 5-FU, we used ESLAM KSL fraction to carefully determine HSCs after the culture, since cultured HSCs show c-Kit⁺ Sca-1⁺ phenotype (Zhang and Lodish, 2005). Consistent with these results, three-cell-division population under control conditions (day 2 of culture) showed little repopulation activity upon competitive BM transplantation. In contrast, three-cell-division populations in Nifedipine treated HSCs (day 4 of culture) exhibited the engraftment in most recipients (Fig. 5 C). In addition, Nifedipine hardly affected linage contributions within recipients (Fig. 5 D). On the other hand, although low cytokine conditions similarly exhibited low intracellular Ca²⁺ level and ΔΨ_m (Fig. 3, E and F), the maintenance of HSCs failed even after one cell division (Fig. 5 B). These results suggest that the suppression of Ca²⁺–mitochondria pathway in the presence of cytokine stimulation contributes to the maintenance of HSCs after cell division.

In addition, gene expression analysis using RNA-Seq revealed that Nifedipine-treated ESLAM KSL HSCs within three-cell-division populations showed significantly enhanced expression of HSC markers/regulators, compared with control ESLAM KSL cells within three-division populations (Fig. 5 E). Hoxb4, Hoxb5, Pdzk1ip1, and Fdg5 were well-known as markers of HSCs, and Mecom, Spi1, Egr1, Jun, Gfi1, Foxo3, Ndn, and cdk1nc contribute to the regulation of self-renewal or quiescence (Cheng et al., 2000; Zeng et al., 2004; Iwasaki et al., 2005; Miyamoto et al., 2007; Min et al., 2008; Kubota et al., 2009; Hills et al., 2011; Kataoka et al., 2011; Akada et al., 2014; Gazit et al., 2014; Chen et al., 2016; Sawai et al., 2016). Consistent with the up-regulation of Gfi1, Foxo3, Ndn, and Cdkn1a, which are downstream genes of p53 (Liu et al., 2009; Renault et al., 2011), the gene set “HALLMARK_P53_PATHWAY” was significantly enriched within up-regulated genes in Nifedipine-treated HSCs (Table 2). Moreover, transcriptome data revealed that up-regulated genes in Nifedipine-treated HSCs exhibited the enrichment of genes with the binding motif of Foxo3 (Fig. 5 F) or FOXO3-targeted genes (Fig. 5 G). Furthermore, down-regulated genes in Gfi1-deficient LSK cells (compared with wild-type cells) was also enriched within up-regulated genes in Nifedipine-treated HSCs (Fig. 5 H). Collectively, these data suggest that the suppression of Ca²⁺–mitochondria pathway contributes to the maintenance of HSCs during cell division, through the up-regulation of p53-related genes.

So far, our data indicate that the appropriate regulation of Ca²⁺–mitochondria pathway achieves both HSC division and the maintenance of stem cell functions in vitro. However, the mechanism of how Ca²⁺–mitochondria pathway is regulated in cycling HSCs after 5-FU administration is still unknown. Although increased concentration of SCF and TPO enhances ΔΨ_m of HSCs in vitro (Fig. 3 B), the in vivo levels of BM SCF and TPO are not significantly altered between untreated and 5-FU–treated mice (Fig. 6 A). These results indicate that enhanced HSC ΔΨ_m after 5-FU administration does not depend on cytokine concentration. Therefore, we focused on a mechanism that maintains low ΔΨ_m of steady-state HSCs without affecting their viability, because 48 h culture at 0.05 ng/ml SCF and TPO similarly led to low ΔΨ_m of HSCs (Fig. 3 B) but was insufficient for most of HSCs to survive (Fig. 3 C).

As 5-FU treatment led to cell cycle entry of HSCs following the depletion of cycling cells, we hypothesized that the presence or absence of cycling hematopoietic cells with high ΔΨ_m may affect HSC ΔΨ_m in vivo. To validate this hypothesis, we examined lineage⁻c-kit⁺ fraction as a regulator for HSC ΔΨ_m, because these cells were drastically decreased after 5-FU administration (Fig. 6 B), even if decreased expression of c-kit in HSCs is considered (Fig. 1 C). Since myeloid progenitors (MPs; linage⁻c-Kir⁺Sca-1⁻) showed higher ΔΨ_m within lineage⁻c-kit⁺ fraction (Fig. 6 C), we examined the ΔΨ_m of HSCs in co-culture with MPs (Fig. 6 D). Interestingly, the co-culture with MPs significantly decreased ΔΨ_m of HSCs without affecting HSC viability, when compared with HSCs cultured without MPs (Fig. 6, D and E). Next, we predicted extracellularly secreted purine metabolites, adenosine as a mediator that is secreted by MPs to regulate HSC ΔΨ_m. The combination of antagonists for an adenosine receptor (SCH442416 for Adenosine A2a Receptor [ADORA2A]; PSB1115 for Adenosine A2b Receptor [ADORA2B]) completely cancelled the effect of MEPs on HSC ΔΨ_m (Fig. 6 D). Consistent with this result, HSCs showed expression of Adora2a and Adora2b, but not Adora1 and Adora3 (Fig. 6 F). Moreover, the addition of adenosine suppressed the ΔΨ_m of HSCs during the culture (Fig. 6 G) and decreased intracellular Ca²⁺ level in HSCs (Fig. 6 H). These data indicate that extracellular adenosine has a potential to suppress Ca²⁺–mitochondria pathway through adenosine A2 receptors. Importantly, the co-culture with MPs greatly suppressed ΔΨ_m of HSCs through adenosine A2 receptors, compared with the effect of adenosine alone (Fig. 6 F). In addition, this suppressive effect is relatively specific for MPs, as indicated by the results that the co-culture with BM CD45⁺ cells failed the suppression of HSC ΔΨ_m (Fig. S4). Therefore, these data suggested that 5-FU-induced ablation of cycling cells, especially MPs, which provide adenosine may enhance ΔΨ_m of HSCs.

In addition to the ablation of MPs, extracellular adenosine level was decreased within BM after 5-FU administration (Fig. 7 A), which may also contribute to enhanced ΔΨ_m of HSCs. Indeed, the administration of CV1808, an agonist of Adenosine A2 receptors, after 5-FU treatment suppressed HSC both ΔΨ_m (Fig. 7 B) and HSC division (Fig. 7 C), suggesting that decreased extracellular adenosine level triggers HSC division in 5-FU–treated mice. However, 5-FU administration only reduced 20% of baseline adenosine level. (Fig. 7 A). Furthermore, 5-FU administration did not alter the expression of Adora2a or Adora2b (Fig. 6 E). Moreover, adenosine alone could also negatively regulate their ΔΨ_m and intracellular Ca²⁺ level (Fig. 6, F and G). Furthermore, under high cytokine conditions that induce HSC division, the treatment with CV1808 suppressed Ca²⁺–mitochondria pathway (Fig. S5, A–E), and achieved both prolonged interval of cell division and the maintenance of low CD48 expression after cell divisions (Fig. S5 F). These data suggest that remaining adenosine alone may still affect the regulation of HSC division. To address this, we examined the effect of extracellular adenosine on HSCs in 5-FU–treated mice using antagonists for both ADORA2A and ADORA2B. Interestingly, these antagonists led to further enhance ΔΨ_m of HSCs after 5-FU treatment (Fig. 7 D). Although ADORA2A and ADORA2B antagonists treatment did not affect BM nucleated cell number (Fig. 7 E), the frequency or number of L⁻ESLAM HSC fractions were decreased (Fig. 7, F and G). Collectively, our finding exhibit that the intensity of extracellular adenosine effect not only determines HSC ΔΨ_m, but also contributes to the maintenance of HSCs after 5-FU administration (Fig. 7 H).

The balance between quiescence and divisions in HSCs is essential for hematopoietic homeostasis. However, the mechanism of how HSCs switch from quiescence to cycling during stress hematopoiesis is unclear. Here we show that Ca²⁺–mitochondria pathway plays a key role in the regulation of HSC division. Indeed, Ca²⁺–mitochondria pathway was activated before HSCs started to divide in vitro and in vivo (Figs. 1–3). LTCC blockers, Nifedipine (Fig. 3, G and H; and Fig. S2) or Isradipine (Fig. S3), suppress Ca²⁺–mitochondria pathway of HSCs and also prolong the interval of cell division through the suppression of genes that regulate late G1 phase (e.g., Cyclin E; Figs. 3 and 4). Thus, our findings indicate that the regulation of HSC cell cycle depends on the Ca²⁺–mitochondria pathway. Moreover, the Ca²⁺–mitochondria pathway appears to be accompanied with up-regulated potential for the uptake of glycose (Fig. S1 D). Since glycolysis is linked with mitochondrial energy metabolism, up-regulated glycolysis in HSCs may also contribute to the activation of mitochondrial functions for the initiation of cell division.

In addition to exhibiting prolonged interval of cell divisions through suppressing Ca²⁺–mitochondria pathway, Nifedipine-treated HSCs maintained stem cell features, indicative of undifferentiated cell divisions (Fig. 5). Similarly, an agonism of adenosine A2 receptors led to both the suppression of Ca²⁺–mitochondria pathway and slow cell divisions maintaining stem cell phenotype after HSC division (Fig. S5). In addition, ΔΨ_m of cycling HSCs in 5-FU–treated mice was also suppressed by this agonism, which maintained HSCs (Fig. 7). In fact, GSEA showed that common gene sets such as p53, IL-2/STAT5, and Notch pathways were enriched significantly within up-regulated genes in both cycling HSCs derived from 5-FU–treated mice (Table 1) and Nifedipine-treated HSCs (Table 2). These data indicate that the effect of an agonist of adenosine A2 receptors on HSCs is recapitulated with that of LTCC blockers. Indeed, adenosine A2 receptors reportedly inhibit calcium influx via LTCCs in rod photoreceptors (Stella et al., 2002). Therefore, these data and this previous study suggest that the agonism of adenosine A2 receptors is involved in the inhibition of Ca²⁺ influx, thereby contributing to the maintenance of HSCs during continuous divisions through the suppression of Ca²⁺–mitochondria pathway. However, it has been generally known that adenosine A2 receptors up-regulate intracellular Ca²⁺ level through the activation of Ca²⁺ channels via GαS or Gαq (Ham and Evans, 2012). However, our findings have been also supported by the recent report showing that adenosine A2a receptor contributes to the suppression of reactive oxygen species generation, a function of mitochondria (Hirata et al., 2018). Therefore, adenosine A2 receptors may negatively affect Ca²⁺–mitochondria pathway in HSCs, probably through an unknown mechanism for the inhibition of Ca²⁺ influx.

Similar to HSCs treated with Nifedipine in high cytokine culture, HSCs cultured in a low concentration of cytokines (undivided conditions) showed suppressed Ca²⁺–mitochondria pathway and cell division (Fig. 3). However, a low level of cytokine was insufficient for the maintenance of HSC phenotype as indicated by significant decrease in phenotypic HSCs after cell division (Fig. 5 A). Importantly, Nifedipine treatment did not negatively affect the up-regulation of early G1 phase–related genes (e.g., Cyclin D) that was induced by high cytokine concentration (Fig. 4 A). In addition, Nifedipine-treated HSCs showed enhanced expression of early response genes (e.g., Egr1 and Jun), compared with HSCs cultured under control conditions (Fig. 5 E). These data suggest that the suppression of Ca²⁺–mitochondria pathway does not disturb the reception of cytokine stimulations. Thus, the combination of Ca²⁺–mitochondria pathway and cytokine stimulations are crucial for HSC self-renewal division.

We also demonstrated that a purine metabolite, adenosine, acts as a regulator of HSCs by suppressing the Ca²⁺–mitochondria pathway in HSCs. (Figs. 6 and 7). Adenosine decreases ΔΨ_m of HSCs in co-culture of HSCs with MPs without affecting their viability in vitro (Fig. 6). This suggests that MPs which are depleted with 5-FU plays a key role in HSC quiescence through providing extracellular adenosine. In addition, regulatory T cells reportedly regulates HSC quiescence through supplying extracellular adenosine (Hirata et al., 2018). Therefore, our results suggest that MPs act as a novel provider of adenosine to contribute to HSC quiescent through suppressing Ca²⁺–mitochondria pathway. In addition to the changes in adenosine level, the ablation of surrounding hematopoietic cells after 5-FU administration may induce the dissolution of hypoxia. As non-HSC hematopoietic cells show a high ΔΨ_m, the ablation of these cells after 5-FU administration may cause a decrease in the consumption of O₂ within the BM. With that, we confirmed an increase in the number of erythrocytes within the BM after 5-FU treatment (unpublished data). In addition, hypoxia is well known as one of factors for the regulation of HSC quiescence (Takubo et al., 2010), and the dissolution of hypoxia may attenuate HSC features through alteration of ΔΨ_m (Mantel et al., 2015). Thus, these data and this previous study support the possibility that the dissolution of hypoxia within BM after 5-FU treatment also contributes to enhanced ΔΨ_m of HSCs.

Although both ΔΨ_m and intracellular Ca²⁺ were simultaneously enhanced after cytokine stimulation in vitro (Fig. 3), the enhancement of ΔΨ_m lagged behind increased intracellular Ca²⁺ level in HSCs after 5-FU administration (Fig. 2). Since the mitochondrial Ca²⁺ level was enhanced simultaneously at the same timing as ΔΨ_m after 5-FU administration (Fig. 2), quiescent HSCs in vivo may be resistant to an influx of Ca²⁺ into mitochondria. Mitochondrial calcium uniporter (MCU) plays a key role in the influx of Ca²⁺ from cytosol to mitochondria (Kirichok et al., 2004) as well as the regulation of mitochondrial energy metabolism (Tarasov et al., 2012). Therefore, the regulation of MCU may be involved in the resistance for an influx of Ca²⁺ into mitochondria in quiescent HSCs within BM niche.

In conclusion, this study demonstrates that the activation of Ca²⁺–mitochondria pathway precedes HSC division. The appropriate suppression of Ca²⁺–mitochondria pathway is necessary for HSC self-renewal division. Therefore, the regulation of Ca²⁺–mitochondria pathway is crucial for the determination of HSC cell fate (self-renewing or differentiation division). We indicate that adenosine acts as a suppressor of Ca²⁺–mitochondria pathway to regulate HSC maintenance in vivo. Thus, our findings shed new light into understanding the mechanism for the HSC self-renewal and may also contribute to the development of ex vivo expansion of HSCs for therapeutic applications.

C57BL/6-Ly5.2 and C57BL/6-Ly5.1 mice were obtained from Sankyo Labo Service Corporation or Japan SLC Inc. Each strain used was between 8 and 12 wk of age. All animal experiments were performed according to the Guidelines of Kumamoto University on Animal Use (Approval No. 29-091).

The following monoclonal antibodies were used for cell sorting and flow cytometric analysis of surface markers: anti-c-Kit (2B8), anti-CD150 (TC15-12F12.2), anti-CD48 (HM48-1), anti-EPCR (RAM34; eBioscience), anti–Sca-1 (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), and anti-CD8 (53-6.72) antibodies. All antibodies were obtained from BioLegend unless otherwise noted.

Suspensions of BM cells were prepared from mice as described previously (Umemoto et al., 2006, 2012).

To induce BM suppression, mice were intravenously injected with 250 mg/kg 5-FU (Kyowa Hakko Kirin). At indicated time after the last administration, BM cells were removed for several analyses. Total nucleated cell numbers were quantified using TC20 Automated Cell Counter (Bio-Rad Laboratories) after Turk staining.

We used either FACS Aria III (BD Biosciences) or FACS CANTO II (BD Biosciences) for cell sorting and flow cytometric analyses, as described previously (Umemoto et al., 2006, 2012).

Long-term competitive repopulation assays were performed by transplanting the indicated cells derived from C57BL/6-Ly5.1 congenic mice into lethally irradiated (10 Gy) C57BL/6-Ly5.2 mice through i.v., as described previously (Umemoto et al., 2012). 20 wk after transplantation, recipient mice with donor cell chimerism (>0.1% for myeloid and B- and T-lymphoid lineages) were considered to be multilineage-reconstituted mice (positive mice). For serial transplantation, 10⁷ whole BM cells were obtained from primary transplanted mice and transplanted into secondary irradiated recipient mice.

Following 5-FU treatment, 150 mg/kg EdC (Tokyo Chemical Industry Co.) was i.p. administrated at an indicated point. After 24 or 48 h, cells that uptake EdC were determined by Click-iT Plus EdU Alexa Fluor 488 Flow Cytometry Assay kit (Thermo Fisher Scientific), according to the manufacturer’s instruction.

We performed RNA-seq as previously reported (Hayashi et al., 2018) with minor modification. In brief, using 100 sorted cells, the first strand of cDNA was synthesized by using PrimeScript RT reagent kit (TAKARA Bio Inc.) and not-so random primers. Following the synthesis of the first strand, the second strand was synthesized by using Klenow Fragment (3′-5′ exo-; New England Biolabs Inc.) and complement chains of not-so random primers. Using purified double-strand cDNA, the library for RNA-Seq was prepared and amplified using Nextera XT DNA sample Prep kit (Illumina Inc.). This prepared library was sequenced on Next-Seq system (Illumina Inc.), according to the manufacturer’s instruction. In addition, each obtained read was mapped to the reference sequence “GRCm38/mm10” using CLC genomic workbench v11.0.0 (Qiagen), and expression levels were normalized and subjected to the statistical analyses based on EdgeR. All RNA-seq data were deposited in the Gene Expression Omnibus (GEO) under accession no. GSE111118. Transcriptome data were subjected to GSEA using GSEA v3.0.0 software, available from the Broad Institute (Subramanian et al., 2005). All Gene sets were obtained from the database of Broad Institute unless otherwise stated. Gfi1-dependent gene set “DOWN_IN_GFI1_KO_KSL” was extracted by a threshold setting at more than twofold changed and P < 0.05, after gene expression pattern between Wt and Gfi1 KO LSK cells in GEO under accession no. GSE20282 were compared.

The ΔΨ_m, intracellular Ca²⁺, mitochondrial Ca²⁺, and or mitochondrial superoxide level of indicated cells were determined according to the manufacturer’s instruction using MitoProbe JC-1 Assay kit (Thermo Fisher Scientific), Fluo-4, AM (Thermo Fisher Scientific), Rhod-2, AM (Thermo Fisher Scientific), or MitoSOX Red Mitochondrial Superoxide Indicator (Thermo Fisher Scientific), respectively. In brief, cells were stained with 2 µM JC-1, 1 µM Fluo-4, AM, 1 µM Rhod-2, AM, or 5 µM MitoSOX Red for 30 min. After staining, each fluorescent intensity was determined using a flow cytometer.

Four hundred sorted HSCs were suspended using 25 µl S-clone SF-03 medium supplemented with 0.5% bovine serum albumin, and subsequently equivalent volume of “Cellno” ATP ASSAY reagent (TOYO B-Net Co.) was added. After 10 min, the intensity of luminescence in each sample was measured by Multi-mode Plate reader Synergy H1 (BioTek Instruments).

After total BM cells were stained by APC-conjugated anti-EPCR antibody, magnetic beads–conjugated anti-APC antibody (Miltenyi Biotec) was used as secondary antibody for cell sorting by AutoMACS pro (Miltenyi Biotec). Obtained EPCR⁺ cells were cultured with 2-NBDG using 2-NBDG Glucose Uptake Assay kit according to the manufacturer’s instruction (BioVision), before staining for the identification of L⁻ESLAM cells. The potential for glucose uptake in L⁻ESLAM cells was determined based on the fluorescence intensity of 2-NBDG by a flow cytometer.

As described previously (Umemoto et al., 2012, 2017), CD150⁺CD48⁻KSL HSCs were sorted and cultured for 5 d in S-Clone SF-03 medium (Sanko-Junyaku Co.) supplemented with 0.5% bovine serum albumin (Sigma), 0.05∼50 ng/ml mouse SCF and 0.05∼50 ng/ml mouse TPO (all from R&D Systems). For single cell culture, CD150⁺CD48⁻KSL HSCs were clonally cultured into 96-well U-bottom plate, and subsequently checked for cell division in each well under a light microscopy after 24 and 48 h. Moreover, to examine an origin of Ca²⁺ influx, 65 µM Nifedipine (Sigma), a Ca²⁺ channel blocker (Shenandoah Biotechnology), or 100 µM 2-Aminoethoxydiphenylborane (2-APB), an inhibitor of IP3 receptor (Sigma), was supplied into the medium.

mRNA expression was assessed using quantitative RT-PCR as described previously (Umemoto et al., 2017).

Intracellular staining was performed using a PerFix EXPOSE kit (Beckman Coulter) according to the manufacturer’s instructions. In brief, sorted CD150⁺CD48⁻KSL cells were fixed, permeabilized, and staining first with antibodies against CDK4^pT172 (9H2L7; Thermo Fisher Scientific), CDK6^pT177 (16HCLC; Thermo Fisher Scientific), and Rb^pT780 (D20B12; Cell Signaling Technology), and then with a PE-conjugated secondary antibody against rabbit IgG (BioLegend). After staining, the cells were analyzed using flow cytometry.

Following sorted CD150⁺CD48⁻KSL HSCs were stained with 2 µM CFSE (Thermo Fisher Scientific) for 10 min, labeled cells were cultured under indicated conditions. Subsequently, cultured cells with CFSE label were analyzed by a flow cytometer. Highest fluorescent peak in HSCs cultured under undivided conditions (0.5 ng/ml SCF and 0.5 ng/ml TPO) served as undivided cells (zero division).

After changed genes were extracted from RNA-seq data of both three-divided CD150⁺CD48⁻KSL HSCs after the culture in the absence and presence of Nifedipine (more than twofold increase, P < 0.05), de novo motif discovery was undertaken on the 500-bp upstream and 100-bp downstream sequences of transcription start sites within Nifedipine–up-regulated genes using DREME (Bailey, 2011), and TOMTOM (Gupta et al., 2007) was used for motif matching.

After tibias and femurs were flashed out using 500 µl of PBS, cell components were removed from supernatants by the filtration following centrifuge at 300 g for 5 min. The amount of SCF or TPO within BM in untreated or 5-FU–treated mice was assessed by using ELISA kit (CUSABIO).

1,000 CD150⁺CD48⁻KSL HSCs derived from C57BL/6-Ly5.2 mice were cultured with 30,000 MPs (lineage⁻c-Kit⁺Sca-1⁻) or CD45⁺ cells obtained from BM of C57BL/6-Ly5.1 mice in S-Clone SF-03 medium supplemented with 0.5% bovine serum albumin, 0.5 ng/ml mouse SCF, and 0.5 ng/ml mouse TPO. After 48 h, the ΔΨ_m of HSCs was examined by staining with an antibody for CD45.2 following JC-1 staining. Before the analysis using a flow cytometer, 1 µg/ml propidium iodide (Sigma) was added into cell suspension to assess their viability. To confirm the contribution of adenosine to the suppression of ΔΨ_m in HSCs, 10 µM SCH442416 (Sigma), and 10 µM PSB1115 (R&D Systems) were supplied as antagonists for Adora2a and Adora2b, respectively.

After tibias and femurs were flashed out using 500 µl of PBS, cell components were removed from supernatants by the filtration following the centrifuge at 300 g for 5 min. Adenosine level was determined within obtained supernatants using Adenosine Assay kit according to the manufacturer’s instruction (BioVision).

To examine the effect of adenosine on ΔΨ_m in HSCs in vitro, HSCs were cultured under undivided conditions (0.5 ng/ml SCF and 0.5 ng/ml TPO) with 0.1∼10 µM adenosine (Sigma) for 48 h and subjected to JC-1 staining. To investigate the effect on intracellular Ca²⁺ level, immediately after Fluo-4–labeled HSCs were stimulated with 1 µM adenosine, the fluorescent intensity of Fluo-4 was examined by flow cytometry. To examine the role of adenosine A2 receptors in ΔΨ_m or intracellular Ca²⁺ level under divided conditions (50 ng/ml SCF and 50 ng/ml TPO), HSCs were treated with 50 µM CV1808 (R&D Systems) for 48 h and subjected to JC-1 or Fluo-4 staining. For in vivo study, 3 mg/kg CV1808 were administrated via i.v. two or three times every 24 h from 1 d after 5-FU treatment, and subsequently ΔΨ_m or EdC uptake in HSCs were examined at 3 or 4 d after 5-FU administration, respectively. Moreover, both 6 mg/kg SCH442416 and 6 mg/kg PSB1115 were administrated via i.p. four times every 24 h from 3 d after 5-FU treatment. These treated mice were analyzed at 7 d after 5-FU administration.

Fig. S1 shows cell division, mitochondrial statuses and the potential for glucose uptake of HSCs before and after 5-FU administration. Fig. S2 shows the effect of Nifedipine on Ca²⁺ and superoxide level within mitochondria in HSCs. Fig. S3 shows the effect of Isradipine on Ca²⁺–mitochondria pathway in HSCs. Fig. S4 shows little potential for the suppression of HSC ΔΨ_m in mature hematopoietic cells. Fig. S5 shows the effect of CV1808 on Ca²⁺–mitochondria pathway as well as cell division in HSCs.

The authors thank Ms. Miho Kataoka and Ms. Yu Matsuszaki for technical assistances in several experiments.

This study was supported by the National Medical Research Council grant of Singapore Translational Research Investigator Award (NMRC/STaR/0019/2014 to T. Suda), from the Japan Society for the Promotion of Science (JSPS) Grant-in-Aid for Young Scientists (17K16190; to T. Umemoto) and Grant-in-Aid for Scientific Research (26221309 to T. Suda), the Ichiro Kanehara Foundation (to T. Umemoto), Friends of Leukemia Research Fund (to T. Umemoto), and the Shinnihon Foundation of Advanced Medical Treatment Research (to T. Umemoto).

The authors declare no competing financial interests.

Author contributions: T. Umemoto and T. Suda designed the study and wrote the manuscript. T. Umemoto performed most of the experiments. M. Hashimoto and A. Nakamura-Ishizu helped with several experiments. A. Nakamura-Ishizu edited the manuscript. T. Mastumura performed promoter motif analysis.