The multifaceted role of mitochondria in HSC fate decisions: energy and beyond

Abstract

Hematopoietic stem cells (HSCs) have the properties to self-renew and/or differentiate into all-mature blood cell lineages. The fate decisions to generate progeny that retain stemness properties or that commit to differentiation is a fundamental process to maintain tissue homeostasis and must be tightly regulated to prevent HSC overgrowth or exhaustion. HSC fate decisions are inherently coupled to cell division. The transition from quiescence to activation is accompanied by major metabolic and mitochondrial changes that are important for cell cycle entry and for balanced decisions between self-renewal and differentiation. In this review, we discuss the current understanding of the role of mitochondrial metabolism in HSC transition from quiescence to activation and fate decisions.

Hematopoietic stem cells (HSCs) are at the apex of the hematopoietic hierarchy. Multipotent by nature, they have the potential to differentiate into all-mature hematopoietic lineages to replenish the blood system, at a steady state or under stress conditions. HSCs mostly reside in a quiescent state to prevent their premature exhaustion due to replicative stress. They divide from time to time to produce highly proliferative progenitors that commit to differentiation, and daughter HSCs that return to a quiescent state to maintain stemness characteristics, known as self-renewal. The decisions to self-renew or to commit to differentiation are critically important to generate adequate numbers of both HSCs and progeny while preventing HSC overgrowth or exhaustion and thus has been an intense area of research. HSC fate decisions inherently occur during cell division, which means that fate decision mechanisms need to be investigated in the context of HSC transitioning from quiescence into the cell cycle. One critical aspect of HSC activation is metabolic remodeling. We know that quiescent or dormant HSCs possess low-metabolic activity, including low-mitochondrial activity [1,2]. Residing in a hypoxic niche, HSCs have high-HIF-1a expression and can use anaerobic glycolysis for their energy needs [3]. During anaerobic glycolysis, the glucose substrate pyruvate is diverted from being incorporated into mitochondria, thus reducing the citrate cycle and mitochondrial respiration. Maintaining anaerobic glycolysis is critical to maintaining HSC quiescence and protecting quiescent HSCs from the negative effects of oxidative stress [3]. This is in sharp contrast with the progenitor compartment that heavily relies on mitochondrial oxidative phosphorylation (OXPHOS) to produce ATP (adenosine triphosphate) for their proliferation and differentiation [1,2]. Interestingly, exit from quiescence is accompanied by remarkable metabolic changes in HSCs, at the cellular, metabolic, and transcriptomic levels [4–9]. Notably, HSC activation causes a drastic increase in mitochondrial activity, higher levels of OXPHOS, and reactive oxygen species. Interestingly, glycolysis activity also increases with HSC activation, and active HSCs have higher glucose uptake than quiescent HSCs [8,10,11], which is necessary for commitment to differentiation [11]. Our understanding of the role of mitochondria in HSC fate decisions has drastically increased in recent years and is much more complex than previously thought. This review intends to summarize what we learned about mitochondrial metabolism of the HSC pool when HSC are quiescent and for HSC fate decisions during cell division. It should be noted that several surface markers are used to identify murine HSCs for experimental studies. HSCs are commonly identified as Lineage-Sca1+c-Kit+(LSK) CD48 CD150+ (or LSK-SLAM (Signaling lymphocytic activation molecule) [12], which contains both long-term-HSC (LT-HSC) and multipotent stem and progenitor cells. This population can then be further enriched for LT-HSCs by their CD34 and CD135/Flk2/Flt3 negativity [13,14], as well as endothelial protein C receptor (EPCR or CD201) positivity [15,16]. HSCs defined by different combinations of markers may behave differently; this should be taken into consideration when interpreting these studies.

HSCS HAVE HIGH-MITOCHONDRIAL CONTENT

Mitochondria play an important role in the supply of energy and metabolic activity in response to cellular demand [17]. Mitochondria are very dynamic organelles, whose numbers and organization can vary greatly [18–20]. Mitochondrial numbers and organization are controlled by fusion and fission [18], as well as turnover—a balance between biogenesis and mitophagy—the selective process of removing damaged mitochondria [21–23]. Mitochondria are both bioenergetic and biosynthetic organelles. Considered the powerhouse of the cells, they produce energy in the form of ATP by a functional electron transport chain (ETC), which is linked to oxygen consumption (OXPHOS). Fatty acid oxidation (FAO) is a potent catabolic program that generates acetyl-CoA and can fuel ETC to generate ATP. The mitochondrial tricarboxylic acid (TCA) cycle also produces numerous metabolite intermediates, including citrate and oxaloacetate, to fulfill anabolic demands. Acetyl-CoA produced in the mitochondria can be converted to citrate, which can then be exported to the cytosol and converted back to acetyl-CoA to be used for lipid synthesis or protein and histone acetylation. Finally, mitochondria are also the hosts of the one carbon folate cycle that generates metabolites used for nucleotide production and S-adenosyl-methionine.

Because quiescent HSCs have low-mitochondrial activity, it was long thought that HSCs possessed only a few mitochondria. Cumulative evidence has now demonstrated that the mitochondrial content is high in young HSCs [24]. Mitochondrial content was underestimated for a long time until the use of the mitochondrial reporter mito-Dendra2 mouse model that expresses the fluorescent protein Dendra2 fused to the mitochondrial-targeted element of Cox8, a protein expressed in the inner membrane of mitochondria. De Almeida et al. [24] showed that HSCs have the highest mito-Dendra2 signals. The fact that HSCs have high-mitochondrial content was then confirmed using quantification of mitochondrial DNA. Interestingly, mito-Dendra2 fluorescence intensity varies among HSCs, such that HSCs can be separated into mito-Dendra2^Lo and mito-Dendra2^Hi subsets [25]. Mito-Dendra2^Lo and mito-Dendra2^Hi subsets represent HSC subsets with different mitochondrial content. Interestingly, mito-Dendra2^Hi HSCs had lower expression of the cell cycle-associated genes Myc and Cdk6, but higher expression of the negative cell cycle regulator Cdkn1c, compared with mito-Dendra2^Lo HSCs. Recipients transplanted with mito-Dendra2^Hi HSCs exhibited significantly higher peripheral blood chimerism, suggesting that long-term HSCs are enriched in the mito-Dendra2^Hi HSC fraction [25].

THE PLASTICITY OF THE MITOCHONDRIAL NETWORK DETERMINES THE TRANSITION BETWEEN HSC QUIESCENT AND ACTIVATED STATES

Mitochondrial Membrane Potential

Despite possessing high-mitochondrial content, quiescent HSCs have immature mitochondria in their structure/morphology [11,25,26]. HSC mitochondria exhibit a punctate morphology with underdeveloped cristae and low-membrane potential, which is an indication of low-proton pump activity (Figure 1). The mitochondrial membrane potential (MMP) is detected by the cell-permeable mitochondrial dye tetramethylrhodamine methyl ester (TMRE). Several groups have shown that the level of TMRE intensity is lower in the most immature HSC population than in other hematopoietic populations [4,10,11,27–29]. Interestingly, long-term repopulation activity is found in the TMRE low (TMRE^Lo) fraction of phenotypically defined HSCs as LSK-SLAM cells or within LSK cells, suggesting that low-mitochondrial activity defines HSC functional activity [11,27–29]. TMRE^Lo LSK-SLAM are considered dormant, found in the label-retention cells, and possess low levels of messenger ribonucleic acid (mRNA) and protein syntheses [11]. Even after in vitro activation and exposure to cytokines, low-mitochondrial activity marks cells that retain long-term repopulation activity. Maintaining low-mitochondrial activity in vitro by blocking the establishment of a membrane potential using a mitochondrial uncoupler seems to be sufficient to maintain stemness after several days in culture [29]. Maintaining low-mitochondrial activity depends on low Myc expression and low-mTOR (mammalian target of rapamycin) activity [3,4,27,30–32]. Low-mitochondrial activity also depends on the mitochondrial carrier homolog 2 (MTCH2) [33]. MTCH2 is a member of the SLC25 family of nuclear-encoded transporters that are localized in the inner mitochondrial membrane. It is a truncated BH3-interacting domain death agonist (tBID)-like interacting protein that recruits tBID to mitochondria, and activates Bax/Bak to control apoptosis. In HSC-enriched LSK cells, MTCH2 limits the expression of the mitochondrial respiratory chain complex to maintain low-OXPHOS levels and low MMP [33]. In addition, MTCH2 maintains balanced NAD/NADH levels for optimum redox control to limit mitochondrial ROS (reactive oxygen species, mitochondrial volume and thus prevents abnormal LSK proliferation. MTCH2 is regulated downstream of the “ataxia telangiectasia mutated” (Atm)-BID axis. ATM controls cellular apoptosis in response to irradiation-induced DNA damage in the bone marrow. ATM is important to maintain HSC repopulation activity by regulating oxidative stress [34]. Interestingly, MTCH2 loss protected LSK cells from irradiation-induced death in vivo by conferring LSK and LSK-SLAM resistance to irradiation-induced caspase-3 activation/-cleavage in vivo and apoptosis [33]. These findings underscore the central role of mitochondria in DNA-damage response, an area of research that is largely underinvestigated.

Figure 1.

HSC metabolic activity in a quiescent and activated state. Quiescent HSCs exhibit low-metabolic and mitochondrial activity. HSCs have high hypoxia-factor 1a expression and rely on anaerobic glycolysis, FAO, and high-lysosomal functions. HSCs have high-mitochondrial content but low MMP that are immature in shape and structure. However, actively cycling HSCs increase their metabolism, including increased nutrient uptake, increased mTOR activity, remodeling of the mitochondrial network into a connected network that is highly dynamic between fission (Drp1 activity) and fusion, and increased glycolysis. Cell cycle entry is initiated by increased intracellular and mitochondrial calcium, and increased MMP and activity. HSC activity is characterized by anabolic pro-growth metabolic activity, including increased protein, lipid, and nucleotide synthesis. Mitochondria can produce acetyl-CoA that can be used for epigenetic modifications. 1C=One carbon cycle; FA=fatty acid; FAO=fatty acid oxidation; HSC=hematopoietic stem cell; mTOR=mammalian target of rapamycine; PDK=pyruvate dehydrogenase kinase; PPP=pentose phosphate cycle; TCA=tricyclic acid cycle.

The MMP of quiescent HSC is lower than in lineage-committed cells, but it is not negative. The relative levels of MMP in HSCs depend on the measurement conditions [35,36]. Different intensity levels of TMRE can be detected by flow cytometry when the dye has not reached equilibrium [35,36]. TMRE dye can be sensitive to multidrug resistance (MDR) activity and thus can be modulated by using drugs that inhibit MDR activity, such as verapamil or the Ca²⁺-independent MDR inhibitor, Cyclosporin H [26,35]. Other studies have reported that modulation of MDR activity did not affect the staining intensity of TMRE in LSK-SLAM [11]. Nevertheless, using agents to block MDR activity, HSCs (LSK CD135−CD150⁺CD48⁻) were shown to have a positive MMP that is enabled by the minimum proton flow through ATP synthesis (or complex V), but the rates of respiration and phosphorylation are low. This is due to an imbalanced expression of the diverse complexes of the respiratory chain. HSCs have high expression of ETC complex II, which sustains complex III in proton pumping, but the expression levels of complex I or V are relatively low [26].

The MMP increases when HSCs enter the cell cycle, both in vitro and in vivo, at a steady state and under regenerative conditions following 5FU-induced myeloablation [4,10,11,27–29,37]. This is associated with changes in mitochondrial shape in which mitochondria become more elongated and are organized into an interconnected network [10,11] and also increased expression of ETC proteins to form a fully functioning mitochondrial respiratory system [26]. This mitochondrial maturation is thought to be coupled with glycolysis, thus enabling aerobic glycolysis that is necessary both for HSC proliferation and differentiation [11,26] (Figure 1). For instance, the deletion of the mitochondrial complex III subunit Rieske iron-sulfur protein (RISP) in adult hematopoietic cells impaired mitochondrial respiration in lineage-negative cells and decreased quiescence of LSK-SLAM cells resulting in severe pancytopenia and lethality [38]. During development, RIPS deficiency impaired hematopoietic cell differentiation [38]. Deletion of PTEN-like mitochondrial phosphatase (PTPMT1) also resulted in defective mitochondrial respiration and activation. These mice die of bone marrow failure due to impaired LSK-SLAM differentiation [39]. Therefore, mitochondrial respiration activity is important for HSC functions and their commitment to differentiation.

Mitochondrial Calcium Homeostasis

A change in MMP with HSC activation is linked to changes in intracellular Ca²⁺ levels. Mitochondria uptake, release, and control Ca²⁺ chelation in the matrix, which allows storage of vast amounts of Ca²⁺, and is necessary for the regulation of calcium homeostasis. The influx of Ca²⁺ into mitochondria is necessary for mitochondrial activation. Calcium buffering by mitochondria has proved to be critical for quiescent HSCs to initiate cell division. Umemoto et al. [37] discovered that in 5FU-treated mice HSCs, identified as Lin-EPRC+Sca +CD48 CD150+ (L-ESLAM), which more accurately identified cycling HSC than c-Kit [37], MMP and Ca2+ levels increase at day 1 following 5FU exposure, before HSC division, which is approximately 3 days following 5FU treatment. At the same time, mitochondrial mass and glycolysis increase, indicating that mitochondrial function in HSCs is enhanced immediately before the initiation of cell division [37]. The same is true for HSC division ex vivo, in which the increase in MMP correlated with intracellular and mitochondrial Ca²⁺ levels and occurred before HSC division (Figure 1). Ca²⁺ levels are important for the activity of 3 key enzymes of mitochondrial metabolism: ketoglutarate dehydrogenase (KGDH), isocitrate dehydrogenase (IDH), and pyruvate dehydrogenase (PDH); and thus for mitochondrial respiration and boosting NADH/NAD+ (Nicotinamide adenine dinucleotide phosphate/Nicotinamide adenine dinucleotide+) ratio, as a result, increased mitochondrial Ca²⁺ levels are necessary for cell cycle activation. Although the increase of intracellular Ca²⁺ and MMP is important for cell cycle initiation, limiting the levels of Ca2+ and MMP during HSC activation and slowing down the cell cycle better maintained HSC repopulation activity in competitive transplant studies even after 3 cell divisions. An interesting mechanism was also proposed in these studies, in which HSC MMP is indirectly regulated by extracellular purine metabolites, adenosine, that are secreted by myeloid progenitors. In the presence of myeloid progenitors, adenosine can suppress the Ca2+−mitochondria pathway through adenosine A2 receptors, hence limiting HSC MMP [37]. These findings point to the importance of communication between hematopoietic cells for HSC homeostasis. These findings also imply that mitochondrial activity and cell cycle length are coupled to maintain HSC functions. This is very interesting as cell cycle length is known to stratify HSC activity in which HSC repopulation potential is inversely correlated with the time-to-cell cycle entry [40–47]. It will be important to examine how mitochondria control cell cycle length and how this ultimately determines HSC potency, as well as whether it involves mitochondrial energy production or nutrient metabolism.

Interestingly, Igf2bp2, an RNA-binding protein downstream of Lin28b/Hmga2, which regulates messenger RNA stability and translation, is emerging as an important regulator of mitochondrial metabolism in HSCs, along with stemness markers and protein synthesis. Igf2bp2 controls the expression of the mitochondrial respiratory chain complex assembly, mitochondrial organization, and metabolic processes. Igf2bp2 deficiency in mice leads to HSC defects that resemble an aging phenotype with impairment in HSC repopulation potential, slow cell cycle, and myeloid-bias [48].

Mitochondrial Dynamism Quality Control

The mitochondrial network is very dynamic, constantly fusing into a connected network or dividing into smaller units. Mitochondrial dynamism is tightly regulated by well-conserved pathways. Mitochondrial fusion is controlled by mitofusins 1 (Mfn1) and 2 (Mfn2) and optic atrophy 1 (Opa1) and 3 (Opa3). Opa1 is also associated with stabilization of the mitochondrial respiratory chain and controls mitochondrial cytochrome C release-induced apoptosis. Mitochondrial fission is regulated by dynamin-related protein 1 (Drp1) and fission protein 1 (Fis1) [18–20]. Mitochondrial biogenesis is mainly regulated by PGC1a (Peroxisome proliferator-activated receptor-gamma coactivator), TFAM (transcription factor A mitochondrial), and NRF1 (nuclear respiratory factor 1) [49]. Mitochondrial dynamism is important to adapt cells to energy demand. When energy demand is high, mitochondria are fused, and mitochondrial OXPHOS is favored. Mitochondrial fusion also enables the “mixing” of mitochondrial membrane proteins for repair mechanisms.

In HSCs, the mitochondrial network organization is plastic and varies with the type of HSCs and their activated state. HSCs are heterogenous in their functional behavior. It is well documented that HSC subsets can generate myeloid-bias, balanced or lymphoid-bias progeny. Myeloid-bias HSCs are enriched in the CD150^Hi fraction of LSK-SLAM, whereas HSCs with extensive lymphoid potential are enriched in the CD150^Lo fraction [50]. Interestingly, CD150^Lo HSCs expressed more Mfn2 mRNA and protein than did CD150^Hi HSCs and had longer mitochondria [51]. The functional importance of these findings was shown in transplant studies, including at the single-cell level, in which a defect in long-term lymphoid repopulation and quiescence of lymphoid-bias HSCs was observed in recipients of Mfn2-deficient bone marrow (BM) cells [51]. Mfn2, but not Mfn1, is known to tether mitochondria to the endoplasmic reticulum (ER) to inhibit NFAT (neclear factor of activated T cells), thereby enhancing intracellular calcium (Ca_i²⁺) buffering [52]. Indeed, Ca_i²⁺ was increased in Mfn2-deficient HSCs as well as in CD150^Hi compared with CD150^Lo HSCs, which was associated with ER stress [51]. How Mfn2 is regulated during HSC division was not examined. It would be important to examine this pathway in the context of HSC activation during bone marrow regeneration, given the importance of Ca2⁺ in the coupling of HSC MMP, cell cycle activation, and fate [37].

Our group has described the importance of mitochondrial morphology plasticity in HSCs during bone marrow regenerative stress [10]. Consistent with prior reports [4], we showed that mitochondrial activity increased with acute LSK-SLAM activation, albeit without changing the overall mitochondrial content. In quiescent HSCs, mitochondria are round and dispersed in the cells. During cell division, mitochondria become more connected into filaments that extend throughout the cell body [10]. At the same time, the mitochondrial network is highly dynamic with numerous fission and fusion events. Mitochondria are also highly mobile by gliding along the microtubule network [53]. Using time-lapse imaging, we showed that in activated LSK-SLAM, the organization of the mitochondria network changes every few minutes. Before division, the mitochondrial network dramatically fragments into smaller units and homogenously segregates to daughter cells, at least in young HSCs. Mitochondrial dynamism is regulated by Drp1 in HSCs. We showed that during LSK-SLAM activation, Drp1 accumulates and oligomerizes around mitochondria to trigger fission events [10]. Genetic loss of Drp1 caused severe mitochondrial aggregation in LSK-SLAM that impairs mitochondrial motility, mitochondrial fission, and proper segregation to daughter cells during division. It impaired HSC repopulation activity in competitive transplant studies. Interestingly, the frequency of HSCs is higher in the BM of recipients of Drp1^—/— cells than wild-type (WT) cells 4 months posttransplant, indicating that Drp1 loss blocks HSC differentiation and amplifies low-functioning HSCs [10]. Hence, mitochondrial dynamism is critically important for HSC self-renewal. We have not observed lineage imbalance in Drp-1-deficient BM, suggesting that mitochondrial dynamism is important across all HSC subsets.

Removal of Damaged Mitochondria: Mitophagy or not Mitophagy?

Removing damaged mitochondria is a necessary process to maintain mitochondrial homeostasis. Damaged mitochondria can be removed using a nonspecific process in which mitochondria are engulfed into autophagosomes and degraded, known as macroautophagy, or using a selective mechanism in which mitochondria are tagged for degradation, known as mitophagy. The regulation of mitophagy is well conserved [21–23]. The canonical mitophagy pathway depends on PTEN-induced putative kinase 1 (Pink1) and the E3 ligase Parkin (encoded by park2). In this pathway, damaged and depolarized mitochondria accumulate Pink1 on the outer membrane, which then recruits Parkin or Arih1 to induce mitochondrial protein ubiquitination. Ubiquitinated proteins are then recognized by adapters (e.g., p62), which recruit LC3 to target mitochondria to the autophagosome. Alternatively, mitochondrial damages induce the activation of outer mitochondrial membrane receptors Bnip3, Bnip3l, Fundc1, Bcl2l13, or Fkbp8 that can directly bind LC3, thus targeting mitochondria for degradation independent of Pink1. Fission of damaged mitochondria from the rest of the network is usually needed before mitophagy and can occur in a Drp1-dependent manner [23,54] (Figure 2).

Figure 2.

Mechanism of removal of mitochondria by mitophagy. Mitochondria can be removed by Pink1-dependent mitophagy in which Pink1 is stabilized at the mitochondrial membrane when depolarization occurs. Depolarization can be a result of high reactive oxygen species. Pink1 recruits the ubiquitin ligase Parkin to ubiquitinate substrate on the mitochondrial membrane and recruit LC3 receptors. LC3 and the autophagy machinery are then activated. In the Pink1-independent pathway, defective mitochondria express outer membrane receptors, including Bnip3l, Fundc1, and Fkpb8, that recruit and bind LC3 directly, to activate the autophagy machinery. Bnip3l and Fundc1 can be activated in a manner dependent on ULK1. Atad3=ATPase family AAA domain-containing protein 3; Bnip3l=BCL2/adenovirus E1B 19 kDa protein-interacting protein 3 like; Fkbp8=FK506-binding protein 8; Fundc1=FUN14 domain-containing protein 1; Ulk1=UNC51-like kinase-1.

Removal of damaged mitochondria is unambiguously necessary for HSC functions. This has been shown in studies in which regulators of the autophagy machinery are deleted [4,55]. For instance, Atg12 loss in adult mice increased mitochondrial content and MMP in HSC (LSK-SLAMFlk2-). This increased mitochondrial activity caused decreased HSC quiescence, and impaired HSC regenerative capacity [4]. Likewise, vavCre-Atg7 mice develop progressive anemia, splenomegaly, and lymphadenopathy and survive only for 12 weeks [56]. In competitive repopulation assays, VavCre-Atg7 bone marrow cells failed to contribute to short- and long-term reconstitution in lethally irradiated hosts. Interestingly, Atg7 deficiency at the adult stage hardly affected the chimerism of donor-derived cells by 3 months after transplantation, unlike using the Vav-Cre system. Hence, autophagy seems to be essential to maintain low-mitochondrial metabolic activity and for HSC functions but may be dispensable for preserving stemness during rapid HSC divisions in vivo [56]. These studies imply that autophagy is necessary to maintain low-mitochondrial activity for HSC quiescence but may be less important when HSCs are actively cycling.

However, it has remained somewhat unclear in which context mitochondria are removed via mitophagy and what is the role of mitophagy in HSC quiescence versus activation. For instance, genetic loss of Park2 does not alter HSC functions [4], suggesting that either mitophagy is not active in HSCs and that mitochondria are solely removed via macroautophagy or that compensatory mechanisms exist. Ito et al. [5] reported that the self-renewal of purified Tie2+ HSCs relies on mitophagy in vitro, especially in the context of Pparδ agonist and FAO activation with GW501516. They show that Parkin and Pink1 expression as well as the mitophagy flux increase upon FAO activation in Tie2+ HSC in vitro [5]. Consistently, silencing park2 by RNA interference-mediated knockdown decreased mitophagic flux in Pparδ agonist-treated cells and greatly decreased their regenerative potential in subsequent transplant studies [5]. These findings clearly show the importance of the Park2-mediated mitophagy in response to Pparδ agonist in vitro and in supporting HSC regenerative potential. These data also suggest that mitophagy is important when HSCs are actively cycling. Another study reported the importance of regulated Pink1-/Park2-mediated mitophagy in HSPC functions in vivo [57]. In this study, the expression of mitophagy regulators increased in activated LSK cells during the recovery phase following 5FU treatment, which was necessary for bone marrow reconstitution. Together, these studies indicate that Pink1-dependent mitophagy is important for HSC functions but it may be context-dependent, implying that Pink-1-independent mitophagy may play a role as well. Hence, we need to examine the role of mitophagy in HSCs in a context-specific manner and investigate the role of noncanonical mitophagy pathways.

Many questions on the role of mitophagy in HSC functions remain. Studies discussed here do not address directly the questions of when and how mitophagy is used by HSCs in vivo. It is still unclear what the role of mitophagy is in the transition of HSCs between quiescence and activation, whether it drives HSCs to return to quiescence following activation and/or whether it is important for self-renewal activity during cell division. Data on actual levels of mitophagy flux in HSCs in vivo are still missing, under both homeostatic and regenerative conditions. Studying mitophagy using genetic loss of regulators of mitophagy or autophagy in which HSCs and their progeny are targeted has great limitations as an effect on HSC quiescence could be secondary to a loss in progenitor functions. It is also unclear if mitophagy is homogenous across various HSC subsets. The use of a mitophagy reporter mouse system will be instrumental in answering these important questions. Furthermore, it will be important to determine whether mitophagy serves as a quality control mechanism to remove damaged organelles or whether mitochondrial content is recycled and used as a source of nutrients. Important work on lysosome functions has offered insights into the role of mitochondria removal in HSC functions. Lysosomes are major components of cellular degradative machinery, but they also serve as signaling hubs for anabolic processes. Lysosomes also reuse and store metabolites and thus have both pro- and antigrowth activities. HSCs possess large lysosomes. Lysosomes are important to maintain HSC functions, both in the murine and the human system [11,58]. But how lysosomes work in HSCs is not well understood. In the human system, lysosomal activity is high in quiescence CD34+CD38 HSPC and limits HSC metabolic and mitogenic activation by degradation of the transferrin receptor, reducing lysosome activation by knocking down the transcription factor Transcription Factor EB (TFEB) and reducing HSC functions. However, overexpressing TFEB promotes quiescence and full self-renewal potential of the HSC pool [58]. During HSC activation, MYC engages biosynthetic processes while repressing lysosomal catabolism, thus driving HSC activation. These findings suggest that HSCs transition from a catabolic to anabolic state with cell activation. In murine HSCs, although LSK-SLAM cells possess high numbers of acid lysosomes with high-lysosomal flux, the actual lysosomal degradation activity might be low. In this context, the study showed that blocking lysosomal activation and amino acid release, using concanamycin A (ConA), a specific inhibitor of the vacuolar H⁺ -adenosine triphosphatase ATPase (v-ATPase), enhanced quiescence and potency of LSK-SLAM repopulation activity. Lysosomes were particularly enriched with mitochondria [11]. Liang et al. [11] suggested that mitochondria are engulfed into lysosomes but are not fully degraded and that cargo are not released into the cytosol when HSCs are quiescent. The initiation of cargo degradation may be important for quiescence exit to provide nutrients for cellular activation [11]. In this case, mitophagy is important to maintain HSC quiescence as well as for cell cycle initiation. Hence, these studies, seemingly in contradiction, likely reveal that lysosomes may have both a pro- and antigrowth function in HSCs, which requires more investigation.

MITOCHONDRIA AS DRIVERS OF HSC FATE DECISIONS

The decision between HSC self-renewal and commitment to differentiation inherently occurs during cell division. Mitochondria and their metabolic products have emerged as critical regulators of fate decisions of stem cells beyond their roles in energy transition for cell cycle initiation. HSCs can divide asymmetrically in which a single HSC can produce 2 daughter cells that adopt distinct fate (asymmetric cell division [ACD]). HSCs can also divide symmetrically to generate daughter cells of similar fate that can be stem cells (symmetric self-renewing division) or differentiated cells (symmetric differentiating cell division). The partitioning of fate-determining factors during cell division, equal or unequal, determines the outcome of cell division, symmetric or asymmetric, respectively. Demonstrating ACD requires linking factors that alter cell fate with their asymmetric distribution. This has been a challenging task in HSCs because of limited knowledge of factors that determine HSC identity, and because HSCs are retrospectively defined by their ability to generate mature cells, which means that examining HSC fate depends on the behavior of the progeny. In yeast, organelles are asymmetrically segregated during division in which the mother cell keeps old mitochondria to generate a “younger” daughter cell [59]. In contrast, human mammary epithelial-like stem cells keep the newly generated mitochondria to maintain stemness properties [60]. In HSCs, a functional link between the asymmetric segregation of mitochondria during cell division and distinct daughter cell fate is being made. The first evidence came in studies from Ito et al. [5] showing the functional importance of mitochondrial FAO in HSC asymmetric division [7]. Fatty acids (FAs) are transferred into mitochondria via the mitochondrial carnitine palmitoyltransferase I (CPT1) where they are broken down into acetyl-CoA, known as β-oxidation. Inhibition of FAO using etomoxir treatment during the active phase of bone marrow regeneration following bone marrow transplantation lowered ATP levels in LSK cells and reduced HSC long-term repopulation activity [7]. In vitro, inhibiting FAO in CD34–LSK cells reduced their repopulating activity, suggesting that FAO is an important metabolic pathway to maintain HSC functions during activation. Interestingly, they further demonstrated that FAO is essential for CD34–LSK asymmetric self-renewing divisions. For this, they used single-cell transplantation to robustly assess stemness characteristics of the daughter cells of a single CD34− LSK division. When FAO is inhibited during cell division, CD34− LSK cells produce daughter cells that no longer possess repopulating activity, indicating that FAO controls the cell-fate decision of dividing HSCs [7] (Figure 3). This study implies that FAO activity can determine HSC fate decision to self-renew or to commit to differentiation and that the daughter cell that maintains high FAO activity has a greater chance to maintain stemness characteristics. It also means that modulating HSC metabolic activity is sufficient to alter their fate. The use of the pharmacologic approach was critically important to demonstrate causality because it enabled manipulating FAO activity during HSC division leaving the progeny intact, thus ensuring that the outcome is not due to changes in the function of the progeny. This cannot be assessed using traditional genetic approaches in which parent cells and their progeny are targeted. Ito et al. [5] then demonstrated that FAO maintains HSC functions by promoting mitophagy. They show that mitochondria are cleared more quickly due to enhanced mitophagy activation in Tie2-GFP+ HSCs treated with a PPAR agonist that enhances FAO in a manner dependent on Parkin and Pink1. However, whether it involves asymmetric segregation of mitochondrial subsets with different FAO activity was not assessed but would be interesting to investigate.

Figure 3.

Asymmetric distribution of organelles in HSC. Quiescent HSCs possess numerous lysosomes and mitochondria with low-membrane potential. Cell cycle entry is accompanied by increased mitochondria activity and membrane potential, increased calcium, and increased lysosomal degradation activity. When fully activated, HSCs switch to anabolic metabolism to increase protein, nucleotide, and lipid synthesis. An ill-defined process of sorting out the various organelles occurs during mitosis such that lysosomes and active mitochondria are asymmetrically segregated to daughter cells. The daughter cell receiving high lysosome content but mitochondria with low-membrane potential tends to retain stemness characteristics whereas the daughter cell receiving active mitochondria that exhibit high membrane potential tends to commit to differentiation. HSC=Hematopoeitic stem cell.

The Schroeder group showed that both murine LSK−SLAM−CD34− cells and human CD34+CD38−CD90 +CD49f+ can asymmetrically segregate MMP^Hi mitochondria to daughter cells [61,62]. Daughter cells receiving more active mitochondria also received more Myc and upregulated the transferrin receptor CD71, which was shown to predict a commitment to differentiation. Interestingly, there was an inverse correlation with lysosome inheritance. In this case, daughter cells receiving fewer active mitochondria inherited more lysosomes, which was associated with slow cell cycle duration and HSC activity in vitro. However, low levels of lysosomes were also shown to correlate with myeloid differentiation and speed of cell-fate acquisition [61,62]. This suggests that fate bifurcations of daughter cells are controlled by asymmetric inheritance of mitochondria and lysosomes. These findings are in line with the fact that mouse and human HSC activity is associated with low-mitochondrial activity but high-lysosomal activity [11,58]. Asymmetric division of active mitochondria was also described in an independent study [63] (Figure 3). This is interesting as the overall mitochondrial content remains equally distributed to daughter cells during HSC division [10]. These findings imply that the mitochondrial network of HSCs is heterogenous and comprised of several mitochondrial populations that produce different metabolites and that are not randomly distributed during HSC division to instruct daughter cell fate.

The mechanism that controls the asymmetric partitioning of mitochondria and lysosomes during HSC division is not known but probably depends on cytoskeleton regulation, as in other systems. ACD requires the establishment of cell polarity that is dynamically coordinated by cytoskeleton reorganization, enabling intracellular compartmentalization. Because organelles are bound to the cytoskeleton, the distribution of organelles depends on the cytoskeleton. However, how mitochondrial metabolism controls the fate of the daughter cells is completely unclear and how FAO metabolites are used in HSC fate decisions is not known. Acetyl-CoA generated from FAO can enter the TCA cycle to fuel OXPHOS and energy production or be transferred back to the cytoplasm via reversible citrate conversion to participate in anabolic processes or used for posttranslational acetylation modifications. Understanding how acetyl-CoA is used will inform how metabolic pathways control HSC fate decisions. Furthermore, mitochondrial functions go well beyond energy production. Mitochondria play important roles in iron metabolism and produce the building blocks for the synthesis of macromolecules, including lipids and nucleotides. How these pathways are used by HSCs and contribute to the control of fate decisions needs to be investigated in detail.

MITOCHONDRIA IN REPLICATIVE STRESS AND AGING

HSCs have high regenerative potential and can reconstitute the entire blood system on transplantation. For this, HSC activation is a necessary step to generate billions of mature blood cells. Yet, the high replication rate that is needed to replenish the blood system causes permanent injury to HSCs. HSCs maintain high-oxidative stress and activation of stress response pathways, including p38^MAPK and TGF-β signaling, which are causal factors of HSC functional decline and premature aging, termed replicative aging [31,64]. Similarly, we have shown that following the stress of bone marrow transplantation, the mitochondrial network is no longer plastic and fails to properly organize in LSK-SLAM HSCs due to loss of Drp1 activity [10]. LSK-SLAM HSC mitochondria are structurally abnormal and swollen, exhibit lower membrane potential, are less dynamic, and as a result organize in large aggregates in the cells. These findings indicate that HSCs fail to repair their mitochondria after replicative stress. One consequence of this is that mitochondria no longer are symmetrically inherited by daughter cells during single LSK-SLAM division [10]. The consequences of maintaining a less-dynamic mitochondrial network are that HSCs fail to retransition into a transcriptional activation state during the cell cycle [10]. These findings are in line with the concept that mitochondria are important for HSC activation [37]. The accumulation of abnormal mitochondria in LSK-SLAM HSC following transplantation or when Drp1 is absent strongly suggests that the daughter cells that inherit abnormal mitochondria remain HSCs, albeit less functional, whereas the daughter cells that receive normal mitochondria likely go on to differentiate; however, it needs to be demonstrated using the in vivo paired-daughter cell assay and by transplantation of the daughter cells. The fact that HSCs can keep abnormal mitochondria is interesting in the context of HSC division memory. As suggested by Bernitz et al., [65] HSCs seem to count and remember their division history in a model termed “generation-aging” in which the generating potential of a stem cell is progressively lost with prior division history. Keeping abnormal mitochondria could be used by HSCs as a mitotic clock to remember prior division history and progressively limit HSC functions. Analyzing the chromatin landscape of HSCs with different division histories and understanding the role of mitochondria in epigenetic memory will be needed to address this hypothesis.

The HSC compartment also changes with physiologic or chronologic aging. The HSC pool expands phenotypically, but HSC self-renewal and regenerative functions decline. Old HSCs exhibit myeloid skewing and an expansion of myeloid-restricted repopulating cells [66]. We and others have shown that aged HSCs (LSK-SLAMs) accumulate abnormal mitochondria that have lower MMP and are different in shape [4,67] as in HSCs after replicative stress [10]. Interestingly, low-mitochondrial activity is associated with lower transcription rate and cell cycle progression [68]. Label-retention experiments in older mice suggest that aged HSC progressively lengthen periods between divisions [65]. Single-cell RNA sequencing comparing young and aged LSK-SLAM HSCs or LSK-SLAM Flt3^— HSC revealed a lower frequency of cells in the G₁ phase among old compared with young HSCs [69] and exhibited a delay in differentiation due to cell cycle arrest and imbalance of cell cycle regulators [70]. In humans, aging is also accompanied by an accumulation of phenotypic HSCs that have decreased repopulation activity [71] and prolonged G₁ phase due to desensitization to mitogenic stimulation by extrinsic growth factors [72]. Interestingly, HSCs exhibit signs of mitochondrial aging at a relatively “young” age. We have shown that MMP starts to decline in LSK-SLAM as early as 12 months [67]. The mitochondrial network is also changing with HSC exhibiting more swollen mitochondria [67]. Expression of ETC genes is lower. Interestingly, MMP values remain heterogenous in old HSCs with MMP^Lo and MMP^Hi subsets. Old MMP^Hi LSK-SLAM HSCs, similar to young MMP^Hi LSK-SLAM HSCs, transcribe RNA faster than MMP^Lo LSK-SLAM HSCs [68]. The MMP level drives the transcriptional rate as in vivo treatment with the mitochondrial uncoupler carbonyl cyanide m-chlorophenyl hydrazine (CCCP) resulted in a reduction of MMP concomitant with a significant decrease in the transcription rate of HSCs. One mechanism is likely the loss of mitochondrial quality control mechanism, such as the mitochondrial unfolded protein response (UPRmt) that acts as a metabolic checkpoint that regulates HSC maintenance [73]. Interestingly, Igf2bp2 expression and its target genes are almost completely lost during aging, perhaps explaining some of the mitochondrial defects associated with aged HSC, including loss of MMP, senescent phenotype, and slow cycling [74]. Finally, microenvironmental factors such as insulin growth factor 1 (Igf-1) also play a role [67]. A decline in Igf-1 in the bone marrow microenvironment appears to initiate LSK-SLAM HSC aging in mid-age mice. But, interestingly, Igf1 supplementation can restore HSC MMP [67].

MITOCHONDRIA AS PROMISING TARGETS TO IMPROVE HSC FUNCTIONS

The findings discussed in this review strongly indicate that mitochondria are central regulators of HSC functions, become defective with age and replicative stress, and cause HSC functional decline. Because mitochondrial functions can be manipulated, mitochondria have become an attractive target for improving HSC functions. A recent study showed that boosting mitochondrial clearance with nicotinamide riboside (NR) can preserve HSC functions in vivo and in vitro [63]. LSK-SLAMCD34− HSCs exhibit a relatively low NAD⁺/NADH (Nicotinamide adenine dinucleotide+/Nicotinamide adenine dinucleotide phosphate) content. NAD⁺ is an essential coenzyme for energy metabolic pathways, including aerobic glycolysis, the TCA cycle, OXPHOS, FA metabolism, and antioxidant defense. NAD+ can be produced via different routes, including via de novo production from quinolinic acid or via salvage pathways (classical and alternative) from nicotinic acid, nicotinamide (NAM), or NR. The active pathway in LSK-SLAMCD34− seems to be the classical salvage pathway—involving Nmnat3 and Nrk1, which can use NR but not NAM to generate NAD+. Treating mice with NR reduced MMP in LSK-SLAMCD34− both after only 1 week of treatment in vivo. It also reduced MMP in dividing HSCs in vitro for 2 days in vitro [63]. Interestingly, reducing MMP was associated with reduced energy metabolism that slowed entry into the first division without completely blocking cell division. Hence, NR curtails metabolic activation enabling slower cell cycle to preserve HSC functions. Mechanistically, NR decreases MMP by increasing mitochondrial clearance. NR is known to induce mitochondrial clearance by activating the autophagy machinery via the NAD+-dependent deacetylase Sirtuin family proteins and by boosting mitochondrial dynamism. NR also activates the mitochondrial stress response, including mitochondrial stress-associated chaperones Hsp60 and Hspa9, and the mitochondrial protease LonP1 to increase mitochondrial clearance. Interestingly, NR increased proliferative asymmetry in LSK-SLAMCD34− HSCs [63]. In this case, lowering MMP is beneficial to maintain HSC functions, which is consistent with studies on the mitochondrial−Ca2+ axis discussed here [37].

Interesting studies in old HSCs show that in this context, enhancing MMP has a beneficial effect on HSC functions. Mitoquinol (Mito-Q), a mitochondrial-targeted coenzyme-Q10 successfully increased MMP of old LSK-SLAM HSCs in vitro [68]. Importantly, successive Mito-Q injection significantly enhanced the MMP of old LSK-SLAM HSCs in vivo. Mito-Q increased old HSC global transcription rate and decreased their levels of intracellular ROS. Mito-Q treatment of old mice significantly reduced the myeloid-bias CD150^Hi HSC subsets, after only a 5-day treatment, both phenotypically and at the transcriptional level. BM reconstituted by old Mito-Q-treated donor HSCs has significantly higher MMP than BM derived from old untreated donor HSCs and is similar in MMP to BM reconstituted by young HSCs. Moreover, both old Mito-Q-treated and young HSCs reconstitute a stem cell pool that is higher in MMP compared with the HSC pool derived from old untreated donors and showed significantly superior peripheral blood engraftment, including improved B-cell output, compared with old untreated HSCs [68]. Thus, short-term in vivo Mito-Q treatment on HSCs is cell intrinsic and can be maintained upon stem cell transplantation with long-term consequences for function. Mito-Q treatment also prevented the onset of anemia that occurred with natural aging. These results raise the possibility that early intervention might prevent or attenuate the onset of an aging phenotype, highlighting the translational potential of targeting mitochondria.

A model is emerging from the studies discussed here, in which mitochondrial functions are coupled to cell cycle activity for HSC homeostasis. Mitochondrial activation is necessary for HSC cell cycle initiation but if unchecked, affects HSC functions, perhaps because HSCs are unable to restore mitochondrial homeostasis after activation. For this, lowering MMP seems to be a promising target to prevent HSC functional decline with activation. NR emerges as a potential booster of hematopoiesis, which could prove useful in vitro for HSC expansion protocols. However, failure to repair mitochondria during replicative stress, at a steady state or under regenerative conditions, ultimately leads to an unhealthy loss in MMP. Defective mitochondria seem also to be associated with hematologic diseases, such as myelodysplastic syndromes that are associated with mutations in splicing factors [75]. In these cases, restoring MMP may be necessary to allow for HSC cell cycle entry, appropriate fate decisions, and blood cell production. Hence, how and when to target mitochondria to improve HSC functions will need to be tailored to the context, including age.

CONCLUSIONS AND PERSPECTIVES

Our knowledge of mitochondria in HSC functions has drastically expanded in the past few years. Mitochondria are emerging as critically important for HSC quiescence, cell cycle initiation, and for HSC fate decisions during cell division. Yet, we still do not know exactly which metabolites are produced by mitochondria, how they are used, and in which context. Mitochondria produce not only energy but also the building blocks for the synthesis of macromolecules, including lipids and nucleotides. Mitochondria produce metabolite intermediates that are used for epigenetic remodeling or for regulating epigenetic modification enzymes. Mitochondria also communicate with the rest of the cells through retrograde signaling pathways to modulate gene expression. Therefore, we need to understand which mitochondrial metabolic pathways are active and when, how mitochondrial metabolites are used, and how it is coordinated with the cell cycle. Only then will we understand the role of mitochondria in HSC fate decisions. With the emergence of HSC heterogeneity, it will also be important to determine whether different HSC subsets rely on distinct metabolic needs and whether specific metabolic pathways drive lineage specification. One aspect of mitochondrial function that has not been studied so far is its relationship with other organelles. Mitochondria form unique contact sites with lysosomes, ER, and lipid droplets that are important for metabolite transfer between organelles and to regulate their respective functions. Given the complex role of lysosomes in HSC functions, mitochondria-lysosome interactions may be important beyond mitophagy and will be interesting to investigate. For this, we will need to develop better imaging techniques and image analytical platforms. Super-resolution imaging with a 3-dimensional view will be necessary to determine how organelles interact and to sort out mitochondria-lysosome contacts from mitophagy events. This is important because mitochondria become defective with replicative stress and physiologic aging. It is also increasingly recognized that mitochondrial defects are associated with hematologic disorders, notably myelodysplastic syndromes. How mitochondrial defects drive aging-related phenotype, and how HSCs adapt to abnormal mitochondria for their survival also need further investigation. This knowledge will guide the development of new therapeutics to fully exploit HSC potential for regenerative purposes.

HIGHLIGHTS.

• Mitochondria are critical regulators of hematopoietic stem cell (HSC) functions

• Mitochondria control HSC cell cycle entry

• Mitochondrial metabolism controls asymmetric HSC fate decisions

• Mitochondria become defective with age and replicative stress

• Mitochondria can be targeted to improve HSC functions

Data Availability

The data sets generated and/or analyzed during the current study are available from the corresponding authors on reasonable request.