Mechanisms involved in hematopoietic stem cell aging

Abstract

Hematopoietic stem cells (HSCs) undergo progressive functional decline over time due to both internal and external stressors, leading to aging of the hematopoietic system. A comprehensive understanding of the molecular mechanisms underlying HSC aging will be valuable in developing novel therapies for HSC rejuvenation and to prevent the onset of several age-associated diseases and hematological malignancies. This review considers the general causes of HSC aging that range from cell-intrinsic factors to cell-extrinsic factors. In particular, epigenetics and inflammation have been implicated in the linkage of HSC aging, clonality, and oncogenesis. The challenges in clarifying mechanisms of HSC aging have accelerated the development of therapeutic interventions to rejuvenate HSCs, the major goal of aging research; these details are also discussed in this review.

Introduction

A pioneering study by Leonard Hayflick described the discovery of cellular senescence, with the demonstration that cultured normal human cells have a limited capacity for cell division [1]. This phenomenon, termed the “Hayflick limit,” is now known to be due to telomere attrition [2]. Subsequent analyses of genetically manipulated model organisms, such as yeasts and nematodes have revealed bona fide factors determining the length of life, including signaling pathways and transcription factors [3]. Recent advances in sequencing technologies have enabled the understanding of global alterations of transcription and epigenetics in the aging process. Other studies on aging using mouse models mimicking human progeria and the naked mole-rat, a mouse-sized rodent with exceptional longevity, have provided advanced knowledge of aging [4, 5]. Multicellular organisms consist of many kinds of cells that cooperatively organize multiple organ systems. Each organ, including the hematopoietic system, is properly maintained by tissue stem cells to antagonize age-associated tissue damage. However, it is not possible to permanently maintain their youthful states, since stem cells become gradually exhausted with age. This process, termed stem cell aging, is recognized as a major cause of an age-dependent organ dysfunction and is central to the entire aging process.

In this review, we focus on aging of the hematopoietic system in terms of the aging of hematopoietic stem cells (HSCs). We first discuss general causes of stem cell aging and their impacts on HSC functions. The causes range from cell-intrinsic mechanisms (i.e. DNA damage, epigenetics, metabolism, and polarity shift) to cell-extrinsic mechanisms (i.e. microenvironment, inflammatory, and immunity). These mechanisms are influenced by genetic factors and environmental factors, including our lifestyles. We further focus on another aspect of HSC aging, that is, clonality. This is a typical feature of cancer, but now there is growing evidence that aged HSCs are exposed to a clonal competition, even in the healthy condition, which could be an origin of hematological malignancies [6–8] and increase the risk of systemic disorders, such as cardiovascular diseases and inflammatory diseases [9]. Knowledge of these mechanisms will help in developing interventions against age-related chronic diseases. The review concludes with a consideration of the current situation and future prospects of HSC rejuvenation.

Age-related changes in the hematopoietic system

Even in healthy people who do not develop any hematologic diseases, the hematopoietic capability changes with age. This reflects a time-dependent deterioration of HSC functions and the bone marrow microenvironment. Just like other organ systems, functions of the hematopoietic system are also affected by individual lifestyles, such as dietary composition and capability of movement, that especially vary among elderly people. In general, the difference in hematopoietic status among individuals becomes greater as time passes, which leads to the so-called “heterogeneity.” Despite this, elderly people are commonly vulnerable to several specific diseases related to aging of the hematopoietic system in common: anemia, infection, and malignancy.

Anemia and its potential progression to myeloid malignancies

The most common hematologic disease among elderly people is anemia. The prevalence of anemia exceeds 20% at the age of 85 [10, 11]. Approximately one-third of anemia in the elderly is caused by deficiencies of iron, folate, and vitamin B12, indicating the strong influence of lifestyle changes in hematopoiesis among elderly people. Another distinctive finding of hematopoiesis in the elderly is chronic inflammation, which causes 20% of elderly anemia [11]. In one-third of cases, the cause of anemia has not been identified. They include patients with idiopathic cytopenia of undetermined significance (ICUS) [11–13]. Notably, clonal expansion of HSCs with somatic mutations is found in 30% of ICUS patients, which is referred to as clonal ICUS (CCUS). These patients are at higher risk of developing myeloid neoplasms [12, 13], implying a close relationship between unexplained anemia and myeloid transformation among elderly people.

Immunosenescence

Elderly people are susceptible to various infectious diseases. Pneumonia is a notable cause of death in the elderly globally [14]. Age-associated alterations in the immune system, often referred to as “immunosenescence” or “senescent immune remodeling,” have been suggested to be involved in the vulnerability to infections among elderly people [15]. Distinctive age-related changes in the immune system include decreased numbers of T and B lymphocytes and thymic involution, which are thought to be partly provoked by a compromised differentiation of aged HSCs toward the lymphoid lineage [16–20]. It is also reported that repertoires of naïve CD4 and CD8 T cells decline with age accompanied by an inequality of clone size, suggesting clonal selection among naïve T cells [21–23]. These changes lead to dysfunction of adaptive immunity and a subsequent predisposition to cancers, a representative aspect of aging.

Previous studies have reported an age-associated change of WNT signaling in HSCs as a molecular mechanism underlying a functional decline of T cells. Increased expression of Wnt5a in aged HSCs induces a switch from β-catenin-dependent canonical Wnt signaling, which plays a pivotal role in T cell differentiation [24], to non-canonical Wnt signaling, leading to myeloid-biased differentiation of HSCs [25]. The enhanced expression of Wnt5a further causes activation of small Rho GTPase Cdc42 to induce cytoskeletal protein “polarity shift,” one of the characteristic features of aged HSCs [26]. The section on “Polarity” describes this is detail.

Aging of HSCs

Age-dependent organ dysfunction is closely associated with impaired regenerative potential of tissue stem cells [27]. In addition to a functional decline, they can detach from the stem cell pool due to terminal differentiation, apoptosis, and cellular senescence caused by various intrinsic and extrinsic factors over time. Thus, stem cell aging consists of chronological changes in both quality and quantity of the stem cell pool. Basically, HSCs have a lot in common with stem cells in other tissues regarding the principles of aging.

HSCs play an integral role in life-long hematopoiesis to satisfy the huge demand for blood cells. Functional deterioration of the HSC compartment is accountable for aging of the hematopoietic system, leading to age-related changes in hematopoiesis that include anemia, myeloid-biased differentiation, immunosenescence, infection, and hematological malignancy [28, 29]. HSCs have the potentials to self-renew and differentiate into all blood lineages [30]. Their functions are experimentally evaluated by long-term engraftment and multi-lineage differentiation potentials in competitive transplantation assays. Reflecting the functional decline, aged HSCs exhibit impaired repopulating potential and defective differentiation compared to young HSCs. Intriguingly, the number of surface marker-defined phenotypic HSCs is paradoxically increased during aging despite their impaired capacities for engraftment, differentiation, and homing [16, 31, 32]. It is likely that a functional decline of aged HSCs could be compensated by expansion of phenotypic HSCs.

One of the characteristic features of HSCs is reversible cell cycle arrest (quiescence) during steady state hematopoiesis. Quiescence is an important mechanism to preserve stem cell identity. Loss of quiescence results in the excessive proliferation and subsequent exhaustion of themselves [33]. A recent study using label-retaining assays demonstrated that label-retaining HSCs divided symmetrically four times throughout life and minimally contributed to homeostatic hematopoiesis in mice [19]. On the other hand, the pool of non-label-retaining HSCs with a myeloid-restricted repopulating potential steadily expanded over time and eventually dominated the HSC compartment in aged mice. These data indicate that the aged HSC compartment is highly heterogenous and consists of rare HSCs with a limited divisional history and predominant progenitors with limited self-renewal and myeloid-biased differentiation potentials.

Senescence

Another state of the cell cycle arrest is cellular senescence. This state is clearly distinguished from quiescence in that cells in a quiescent state can re-enter the cell cycle in response to external demand, while cells in a senescent state cannot [34]. As mentioned above, this phenomenon was originally described by Hayflick in human fibroblasts [1]. Since then, a number of studies have shown that a wide variety of cells, including tissue stem cells, can undergo cellular senescence over time, implying its contribution to aging [35–42]. A typical features of senescent cells is increased expression of cell cycle inhibitors (p16^INK4A, p21^CIP1, p53, and RB) [43–46] and senescence-associated β-galactosidase (SA-β-gal) [35]. They also secrete inflammatory cytokines as a phenomenon called the “senescence-associated secretory phenotype (SASP)” [47, 48] and can exhibit tetraploidy [49].

In particular, the expression of p16^INK4A is most important. Its expression level is markedly elevated with age in the majority of tissues and is a hallmark of the senescent cells [40, 50]. Janzen et al. found that aged HSCs express p16^INK4A and its depletion can ameliorate the repopulation capacity in mice [39], indicating a potential role of senescence in HSC aging. In addition to HSCs, a series of studies have demonstrated that ablation of p16^INK4A in other tissue stem cells also restored regenerative potential, but increased susceptibility to malignant transformation [37–39, 51]. Given these facts, it is conceivable that senescence is a safeguard system to prevent the growth of dysfunctional stem cells and potentially cancerous cells. Normally, senescent cells are detected and removed by the immune system [52–55]. Yet it is plausible that this process would be retarded owing to immunosenescence, which can lead to accumulation of senescent cells and subsequent aging of the tissue. Pro-inflammatory cytokines are SASP factors that are secreted by senescent cells. These cytokines play a beneficial role in expediting clearance of themselves by immune cells [52–55]. In contrast, they have been shown to be a breeding ground for chronic inflammatory milieu and promote tumor development [56–59].

The constitutive activation of the oncogenic signaling pathway by mutations, including Ras, Raf, and Pten, can cause oncogene-induced senescence (OIS), which is regarded as a defensive mechanism to prevent aberrant proliferation of oncogenic cells [43, 45, 60–62]. Recently, one study reported a model of OIS in hematopoietic cells using a humanized mouse model, in which human hematopoietic stem and progenitor cells (HSPCs) expressing the mutant form of BRAF^V600E were transplanted into the immunocompromised NSG mice to activate the BRAF/MAPK pathway [63]. These recipient mice presented with myeloid-restricted hematopoiesis and died of bone marrow failure. HSPCs expressing BRAF^V600E exhibit typical senescent phenotypes, such as cell cycle arrest, and increased levels of p16^INK4A, SA-β-gal, and inflammatory cytokines. Importantly, SASP factors from BRAF^V600E-expressing HSPCs can induce senescence in BRAF^V600E-non-expressing by-stander cells. Pharmacological inhibition of tumor necrosis factor-alpha (TNF-α) can ameliorate senescence of by-stander cells, lymphoid differentiation, and disease progression in recipient mice.

These data suggest that senescent HSCs not only impair their functions but also exacerbate the surrounding microenvironment through a paracrine senescence mediated by SASP factors. However, whether HSCs actually undergo senescence is still controversial, and further research is required to determine the significance of senescence in HSC aging [64, 65].

Mechanisms of HSC aging

While age-dependent phenotypic changes of hematopoiesis are well-known, the underlying mechanisms of HSC aging remain less clear. Aging process is so stochastic affected by multiple internal and external factors that individual life spans greatly vary, which makes this issue complicated. Nevertheless, previous studies have clarified the nature of aging through analyses of genetically manipulated model organisms and progeria mouse models [3, 5]. In this section, we discuss the molecular bases of stem cell aging including cell-intrinsic (DNA damage, telomere attrition, epigenetics, metabolism, and polarity) and cell-extrinsic (inflammation and microenvironment) factors.

Cell-intrinsic factors

The most likely explanation for HSC aging is deterioration of cellular components over time, including damage to proteins, lipids, and nucleic acids. DNA damage is permanently inheritable and can lead to functional decline and also malignant transformation of HSCs. Analogous to DNA damage, epigenetic modifications could be inherited in progeny as an “epigenetic memory” to affect the cellular functions. Furthermore, several lines of evidence have demonstrated that metabolically inactivated state is a prerequisite for maintenance of HSCs. These findings indicate that dysregulation of metabolic homeostasis secured by nutrient sensing pathways, antioxidant mechanisms, and proteostasis is responsible for an age-dependent deterioration of HSC functions. Cell-extrinsic factors affecting HSC aging are summarized in the accompanying figure and table (Fig. 1 and Table 1).

Fig. 1.

Overview of cell-intrinsic factors contributing to HSC aging. Aging is an integral process mediated by both internal and external factors. Concerning cell-intrinsic factors, the integrity of mitochondria and epigenetics are particularly critical for maintaining HSC functions. Mitochondrial activity is properly regulated by nutrient sensing pathways, such as AMPK and PI3K/AKT/mTOR signaling. ROS, byproducts of aerobic respiration in mitochondria, can damage proteins, lipids, and nucleic acids to impair mitochondrial function and genome integrity. These toxic effects of ROS are ameliorated by protective mechanisms, involving FOXOs, ATM, Sirtuins, and other molecules. Mitochondria with excessive activity are degraded by autophagy to maintain quiescence of HSCs with a low metabolic state. Perturbation of mitochondrial integrity caused by dysregulation of these mechanisms could lead to HSC aging. Epigenetics factors also influence HSC aging, as suggested by a clonal expansion of HSCs harboring mutations in epigenetic factors, including DNMT3A, TET2, ASXL1, among elderly people. HSCs with these mutations are characterized by increased self-renewal and impaired differentiation potential, leading to elevated susceptibility of the elderly to hematological malignancies

Table 1.

Cell-intrinsic factors affecting HSC aging

Age-related changes	Description	Rationale
Senescence
Telomere attrition (replication)	Irreversible cell cycle arrest induced by activation of p53/p21CIP1 and p16INK4A/RB pathways	Increased expression of p16 INK4A in aged HSCs and its depletion ameliorates their repopulating potentiala [39]
DNA damage
Oxidative stress
Oncogene (e.g. RAS, RAF, PTEN)		HSPCs expressing BRAF V600E exhibit senescent phenotypes (e.g. upregulation of p16 and SA-β-gal, SASP) [63]
Genome instability
DNA damage	Can be caused by…	Increased DNA damage in aged HSCs [68–70]
	DNA repair through error-prone NHEJ [71]
		Mice defective in DDR-related genes show dysfunction of HSCs [69, 74–79]
	Attenuation of DNA damage response with age [68, 72]
	ROS production by OXOPHOS in mitochondria
Telomere attrition	Telomere shortening caused by replication	Loss of telomeric DNA in HSCs over time [84]
		Terc -deficient mice show impaired repopulating potential [85]
Epigenetics
Dysregulated DNA methylation	Aging is associated with…	Loss of Dnmt1 impairs self-renewal and differentiation potential of HSCs [93]
	Increase in methylation levels overall [65]	Loss of Dnmt3a increases self-renewal of HSCs at the expense of differentiation potential [94]
	Hypomethylation in genes related to HSC maintenance [65]	Deletion of Tet2 enhances self-renewal of HSCs and causes myelo-proliferation [95, 96]
	Hypermethylation in differentiation-related genes [65]
	Differential methylation in genes relevant to WNT, cadherin, and cell-adhesion pathways [97]
	Decreased levels of 5hmC, Dnmt, and Tet with age [65]
	CH harboring mutations in DNMT3A and TET2 in the elderly [6–8]
Dysregulated histone modifications	(Sun et al. Murine HSCs Ref [65]) Aging is relevant to…	Fine-tuning of H3K4me3 and H3K27me3 levels is indispensable for proper maintenance and differentiation of HSCs [98–103]
	Increased number and broadened width of H3K4me3 peaks
	Increased intensities of H3K27me3 peaks without change in number
	Increase in the number of bivalent domains
	(Adelman et al. Human HSCs Ref. [97]) Aging is relevant to…
	Decreased H3K27ac, H3K4me1, and H3K4me3 levels	ASXL1-MT induces expansion of HSCs with impaired repopulation and differentiation potentials over time [240]
	Decreased H3K27ac levels at active enhancers
	Decreased H3K4me3 levels at the bivalent promoters
	Downregulation of epigenetic regulators and hematopoietic transcription factors accompanied by changes in histone modification
	CH harboring mutations in ASXL1 in the elderly [6–8]
Epigenetic drift	Stochastic changes in epigenetic modifications can be inherited by their offspring as an "epigenetic memory"	Clonal behavior of HSCs is determined by epigenetic memory (DNA methylation, chromatin accessibility) during aging [106]
		Epigenetic modifications become increasingly heterogenous among individual cells with the passage of time [107]
		iPS cells derived from aged HSCs are capable of redifferentiating into HSCs whose functions are comparable with those of young HSCs [257, 258]
Metabolism
Reactive oxygen species (ROS)	ROS-mediated peroxidation of mitochondrial membrane lipids causes proton leak to elevate ROS levels over time [122]	ROS-mediated peroxidation of mitochondrial membrane lipids causes proton leak to elevate ROS levels over time [122]
	HSCs are equipped with protective mechanisms against oxidative insults (e.g. FOXO [113, 114], ATM [74], SIRT [147, 148])	HSCs are equipped with protective mechanisms against oxidative insults (e.g. FOXO [113, 114], ATM [74], SIRT [147, 148])
Mitochondrial dysfunction	Aging is associated with mitochondrial dysfunction and increased mutations in mitochondrial DNA [115, 118–121]	Age-associated decline in mitochondrial activity causes myeloid-biased differentiation and impairs engraftment of HSCs [265]
		Deletion of Uqcrfs1, a component of mitochondrial respiratory chain, impairs functions of HSCs [117]
		Mitochondrial mutator mice exhibit impaired differentiation potential of HSCs [116]
Dysregulation of nutrient sensing pathways	PI3K/AKT/mTOR pathway upregulates mitochondrial biogenesis to promote aerobic respiration [132]	Overactivation of PI3K/AKT/mTOR pathway leads to loss of quiescence and exhaustion of HSCs [133–136, 139]
		Age-associated activation of mTOR causes expansion of HSCs with impaired repopulation and myeloid-biased differentiation potentials [239, 262]
Proteostasis	Low rates of protein synthesis enable HSCs to degrade and accumulate misfolded proteins [143]	An artificial increase in misfolded proteins reduces the HSC pool and impairs HSC functions [142]
	Degradation of aberrant proteins and organelles (e.g. activated mitochondria) via macroautophagy [141]	Aged HSCs with low autophagy levels exhibit mitochondrial activation and impaired repopulating potential [124]
	Selective degfradation of target proteins through chaperon-mediated autophagy (CMA) [126]	CMA activity in HSCs declines with age [126]
		HSCs defective in CMA impair glycolysis and fatty acid metabolism to reduce self-renewal [126]
Sirtuins	Sirt1: decrease ROS levels via Foxo3 activation [147] Sirt2: inhibit age-related activation of NLRP3 inflammasome [150] Sirt3: deacetylate SOD2 to improve its function and to reduce oxidative stress [148] Sirt6: interact with LEF1 to deacetylate H3K56ac and downregulate Wnt target genes [149] Sirt7: inhibit NRF1 to activate mitochondrial unfolded protein response [127]	Sirt1 [147], Sirt2 [150], Sirt3 [148], and Sirt7 [127] are downregulated in aged HSCs and their genetic ablations impair HSC functions
		Overexpression of Sirt1, Sirt2, Sirt3, or Sirt7 restores functions of aged HSCs through Foxo3 activation [147], NLRP3 inflammasome inactivation [150], reduction in ROS levels [148], and activation of the mitochondrial unfolded protein response [127], respectively
Polarity shifts
	CDC42 causes a loss of polarity, including apolar distribution of tubulin and H4K16ac, and disturbed distribution of heterochromatin [25, 26, 151]	Aged HSCs upregulate Wnt5a to activate non-canonical Wnt signaling and Cdc42, leading to changes in polarity and chromatin architecture accompanied by myeloid-skewed differentiation and impaired repopulating potential [25, 26, 151]

^aWhether HSCs undergo cellular senescence is still controversial since some studies showed aged HSCs did not upregulate p16[64, 65]

Genome integrity

DNA damage

Aging is accompanied by accumulation of DNA damage throughout life. Long-lived stem cells, such as HSCs, are continuously exposed to both endogenous (e.g. ROS, replication stress, telomere attrition) and exogenous (e.g. ultraviolet, ionizing radiation, chemical compound, metabolite) genotoxins to increase the number of phosphorylated histone H2AX (γ-H2AX) foci in the nucleus and comet tail moment, both of which are markers for DNA strand breaks [66, 67], indicating the accumulation of DNA damage with age [68–70].

HSCs exclusively employ non-homologous end joining (NHEJ), rather than homologous recombination (HR), for double-strand break repair, the most harmful form of DNA damage, as HSCs are principally in the G0 phase. However, NHEJ is potentially mutagenic and can be the cause of malignant transformation since it is an error-prone manner of DNA repair compared to HR [71]. HSCs in a quiescent state have an attenuated DNA damage response (DDR) and accumulate DNA damage with age, which is repaired when HSCs enter the cell cycle [68, 72]. On the other hand, Bettina et al. argued that the efficiency of DNA repair is comparable between young and aged HSCs, making age-related functional changes in DDR more controversial [70]. Previous findings from patients with premature aging diseases, commonly known as progeroid syndromes, have further implied a causative role of DNA damage in the aging process. Individuals suffering from progeria syndromes frequently harbor mutations in genes related to DDR and exhibit an age-dependent increase in the burden of DNA damage compared to healthy people [73]. Genetically engineered mice defective in DDR-related genes showed progeroid-like phenotypes, including a functional decline of HSCs [69, 74–79].

DNA damage is also a potent inducer of cellular senescence to preclude tumorigenesis caused by DNA damage-induced somatic mutations, and can induce differentiation and apoptosis through activation of p53 [80, 81]. It appears that cellular senescence, differentiation, and apoptosis are all reasonable strategies to secure the integrity of HSCs by eliminating irreversibly damaged cells. However, they can deplete the HSC pool, as suggested by exhaustion of HSCs deficient in Mdm2, an E3 ubiquitin ligase targeting p53 for degradation [82].

Telomere attrition

To prevent genomic instability caused by the end-protection problem, repetitive sequences called telomeric repeats are located in the chromosome ends [83]. In response to telomere attrition, telomerase can catalyze the addition of telomeric DNA to the end of telomeres. However, as the majority of somatic cells in humans have little telomerase activities, telomere shortening does occur with age, including HSCs [2, 84]. This can cause cellular senescence as represented by the aforementioned Hayflick limit [1]. HSCs from telomerase RNA component (Terc)-deficient mice exhibited impaired repopulating potential [85]. Nevertheless, overexpression of the telomerase reverse transcriptase (Tert) could not improve HSC functions [86], suggesting that telomere-independent mechanisms may constrain the repopulating potential of HSCs. Notably, some patients with aplastic anemia and dyskeratosis congenita harbor mutations in genes involved in maintenance of telomeric DNA, including TERC and TERT [87–91]. Individuals with these diseases exhibit accelerated telomere shortening, predisposition to malignancy, bone marrow failure, and premature death. These consequences demonstrate that the telomerase activity is indispensable for sustaining life-long homeostatic hematopoiesis in humans.

Epigenetics

Much effort has been made to identify the characteristic features of gene expressions in the physiological aging of HSCs. Aging is associated with a fluctuation of gene expressions related to stemness and differentiation of HSCs [17], which could be due to age-dependent changes in regulatory mechanisms such as epigenetic modifications [65]. Like genome stability, epigenomic integrity is crucial to retain the identity of HSCs since epigenetics can be inherited from parental cells to daughter cells during the self-renewal process. Accumulating evidence indicates that age-related changes in epigenetic affect stem cell aging.

DNA methylation

Cytosine residues of CpG dinucleotides are methylated by DNA methyltransferase (DNMT) to produce 5-methylcytosine (5mC). DNMT3A and DNMT3B are responsible for de novo methylation and DNMT1 maintains methylation [92]. 5mC is further oxidized to 5-hydroxymethylcytosine (5hmC) by the ten–eleven translocation (TET) family enzyme, leading to loss of methylation [92]. DNA methylation is indispensable for maintaining HSC functions. Loss of DNMT1 impairs self-renewal and differentiation potentials of HSCs [93], whereas DNMT3A-null HSCs increase self-renewal at the expense of differentiation potential [94]. Deletion of Tet2 in mice also causes enhanced self-renewal and expansion of HSCs [95, 96], suggesting that fine-tuning of the DNA methylation level is critical for proper function of HSCs. Genome-wide bisulfite sequencing has revealed elevated global methylation levels as a whole in murine aged HSCs, while those of aged stem cells in other tissues were decreased [65]. Of note, hypermethylation in differentiation-related genes and hypomethylation in genes related to HSC maintenance have been demonstrated in aged HSCs, indicating chronological deterioration of HSC functions. In addition, the level of 5hmC, Dnmt, and Tet is reduced in aged HSCs [65], implying that mechanisms for maintenance of DNA methylation are perturbed with age. Another study using human HSCs from healthy donors identified differentially methylated regions (DMR) that were enriched for pathways related to HSC biology, including WNT and cadherin pathways [97]. Notably, a subset of age-associated DMRs were commonly disrupted in samples from patients with acute myeloid leukemia (AML), which suggests the potential role of age-related changes in DNA methylation patterns in the development of AML.

Histone modifications

Histone modifications are other typical epigenetic mechanisms to regulate gene expressions. An active mark H3K4me3 and a repressive mark H3K27me3 established by the Trithorax (TRX) group proteins and Polycomb-repressive complex 2 (PRC2), respectively, are the principal modifications in a wide range of stem cells [98]. Genetic perturbation of these histone modulators in mice reportedly provoked disturbance of self-renewal and differentiation potentials of HSCs, indicating the essential role of these histone modifications in maintenance and fate determination of HSCs [99–103].

Chromatin immunoprecipitation sequencing (ChIP-seq) analysis has revealed age-related changes in epigenetic modifications in HSCs. Sun et al. demonstrated that H3K4me3 peaks were increased and their width were broadened in aged murine HSCs [65]. In particular, genes involved in stem cell self-renewal showed an increase in H3K4me3 breadth in aged HSCs. Signal intensities of H3K27me3 were elevated without changing the number of them, although its significance in HSC aging is undetermined. Loci modified by both activating H3K4me3 and repressive H3K27me3 marks are termed “bivalent domains,” which represent a poised state for gene expression. A poised state in stem cells is the threshold state that allows the differentiation-related genes to switch their expressions to either activated or suppressed state [98]. Aged HSCs displayed an approximately four-fold increase in the number of bivalent domains compared to those of young HSCs [65], although the impact of an increase in bivalent domains on HSC functions remains unexplained.

Another study by Adelman et al. characterized the age-related changes in histone modification using human HSCs [97]. They showed that the levels of H3K27ac, H3K4me1, and H3K4me3 were decreased across the genome during aging. The observed changes do not appear to be stochastic as they were reproducibly detected at the specific loci. These epigenetic changes were evident at active enhancers annotated with genes involved in the immune signaling, where H3K27ac levels were decreased. Notably, aging was associated with a significant reduction of H3K4me3 levels at the bivalent promoters with minimal changes in H3K27me3 levels, leading to a switch from the bivalent state to the repressive state. These age-related changes in histone modification were accompanied by downregulation of several epigenetic regulators and hematopoietic transcription factors. Moreover, a subset of the epigenetic changes in aged HSCs were commonly observed in AML samples from both young and elderly patients, suggesting the contribution of epigenetic reprogramming during aging to leukemogenesis.

Epigenetic drift

The driving force of epigenetic changes in the aging process remains elusive. One possible explanation is an age-associated downregulation of epigenetic factors, including DNMT3A, TET2, and EZH2, a component of PRC2, which could perturb epigenetic modifications over time [65]. In contrast, recent studies have provided another view regarding this issue. Epigenetic modifications of HSCs can be subject to stochastic changes caused by a fluctuation of epigenetic regulations to be inherited by their offspring as an “epigenetic memory.” This could be a principal cause of heterogeneity in aged HSCs [104, 105]. Yu et al. used endogenous fluorescent tagging in mice to reveal the heterogeneity of hematopoietic cells in vivo. The authors demonstrated that when HSCs were transplanted into recipient mice, each clone contributed to hematopoiesis in the same way as it had done in donor mice. In addition, each clone behaved in the same manner in response to inflammation and genotoxic insult in all recipient mice. These traits of HSCs were associated with distinctive DNA methylation and chromatin accessibility patterns of each clone that were preserved under the stress conditions, indicating that an epigenetic memory is one of the determinant factors for the fate decision of HSCs [106].

Cumulative changes of epigenetic modifications may cause fluctuation of gene expression to affect growth potential over time. This phenomenon is termed “epigenetic drift,” which could contribute to loss of identity as a stem cell and may be the basic principle for acquiring a clonal advantage of HSCs [28, 107]. Further details are provided in the section dealing with “Clonal hematopoiesis”.

Metabolism

Reactive oxygen species (ROS)

Mitochondria are the principal powerhouse of cells. In the mitochondria, oxygen incorporated into blood is utilized in oxidative phosphorylation (OXPHOS) to produce adenosine triphosphate (ATP), an “energy currency” of cells, in the course of aerobic respiration. ROS are byproducts of OXPHOS, which mediate tissue damage caused by oxidation of proteins, lipids, and nucleic acids [108]. Since Denham Harman postulated “the free radical theory of aging” in the 1950s [109], a growing body of evidence has suggested the validity of this theory, including aging of the hematopoietic system.

Several studies have indicated a causative role of ROS in a functional decline of HSCs. In one study, HSCs expressing high levels of ROS transplanted into recipient mice impaired the repopulating potential and myeloid-skewed differentiation, whereas transplantation of HSCs with low levels of ROS retained higher repopulating potential [110]. Moreover, an age-dependent deterioration of repopulating potential associated with increased levels of ROS and DNA damage was observed in HSCs from elderly healthy people. In another study, treatment with the antioxidant N-acetylcysteine (NAC) restored the repopulating potential of HSCs accompanied with reduction of ROS and DNA damage, indicating the adverse effects of ROS on HSCs [111].

HSCs are equipped with several protective mechanisms against oxidative insults. The forkhead box protein O (FOXO) family of transcription factors are longevity factors that are also critical in anti-oxidative defense of HSCs [112]. FOXO proteins regulate expression of reducing enzymes, including catalase and manganese superoxide dismutase (MnSOD). Deletion of the Foxo gene decreased the expression of anti-oxidative genes, leading to increased ROS levels, loss of quiescence, and impaired repopulating potential of HSCs [113, 114]. In addition, the ataxia telangiectasia mutated (ATM) protein can protect HSCs from oxidative stress, although the underlying mechanisms remain unclear [74]. Mice deficient in either Foxo or Atm exhibited elevated ROS levels and age-related functional deterioration of HSCs compared to those of wild-type mice, implying the importance of anti-oxidative defense mechanisms under the physiological aging of HSCs.

Mitochondrial dysfunction

Aging is accompanied by mitochondrial dysfunction [115], which can contribute to aging of HSCs [116]. The importance of proper mitochondrial functions in HSCs is exemplified by a study in which the deletion of Uqcrfs1, a component of mitochondrial respiratory chain, impaired the differentiation potential and quiescence of HSCs [117]. One of the reasons for an age-dependent deterioration of mitochondrial function is increased mutations in mitochondrial DNA (mtDNA) [118–121]. As ROS are exclusively generated in a respiratory chain at the mitochondrial inner membrane, mtDNA could directly suffer from ROS-mediated DNA damage [122]. One study using mutator mice harboring a proofreading-defective mitochondrial DNA polymeraseγ demonstrated that mutations in mtDNA caused a decline in mitochondrial membrane potential (MMP) of HSCs and premature aging phenotypes, indicating the requirement of mtDNA integrity for maintenance of HSC functions [116]. Furthermore, the proteostasis mechanisms maintaining mitochondrial quality, such as mitochondrial unfolded protein response and mitophagy, have been shown to contribute to keep HSC functional [123–128]. The details are provided in the section dealing with “Proteostasis” described below.

Nutrient sensing pathways

Dormant HSCs utilize anaerobic glycolysis for energy production in the hypoxic environment of the bone marrow niche to prevent damage from ROS generated by aerobic respiration in mitochondria [129, 130]. To achieve metabolic homeostasis in HSCs, mitochondrial activity is strictly regulated by multiple signaling pathways. All the pathways have been reported to be involved in aging and longevity [131].

In response to external proliferating stimuli, such as Insulin/Insulin-like growth factor (IGF) signaling, the phosphoinositide 3-kinase (PI3K)/AKT pathway activates mammalian target of rapamycin (mTOR) to upregulate mitochondrial biogenesis, increase ATP production, and promote anabolic processes [132]. Overactivation of PI3K/AKT pathway induces excessive mitochondrial biogenesis and ROS production, leading to loss of quiescence and exhaustion of HSCs [133–136], whereas its inactivation severely impairs repopulating capacity of HSCs [137]. In contrast, AMP-activated protein kinase (AMPK), which is activated by liver kinase B1 (LKB1)-mediated phosphorylation, promotes glycolysis and inhibits mTOR activity to suppress an energy-consuming anabolic process by sensing an increase in AMP/ATP ratios [132]. One study described that ablation of Lkb1 led to loss of quiescence, impaired repopulating potential, and exhaustion of HSCs associated with reduced mitochondrial activity in AMPK-dependent and AMPK-independent manners [138].

Both pathways eventually converge on mTOR, implying the importance of its proper regulation for HSC maintenance. Indeed, loss of tuberous sclerosis complex (Tsc1), a negative regulator of mTORC1, can induce the entry of HSCs into the cell cycle and impairment of its self-renewal potential [139]. On the contrary, inactivation of mTORC1 via depletion of Raptor, a component of mTORC1, can perturb HSC functions to cause pancytopenia [140]. Taken together, these findings suggest that metabolic pathways require fine-tuning to maintain HSC functions and could be possible therapeutic targets for rejuvenation of aged HSCs.

Proteostasis

Similar to genome stability, the homeostasis of proteins (proteostasis) is critical for proper cellular functions that include synthesis, folding, trafficking, and degradation to ensure the integrity of proteins. Intrinsic and extrinsic stresses cause unfolding and misfolding of proteins, which is resolved by proteostasis mechanisms. However, dysregulation of proteostasis causes the accumulation of aberrant proteins and toxic aggregations responsible for cellular damage, tissue malfunction, and aging [141]. HSCs can accumulate damaged and misfolded proteins over time due to their longevity. Nevertheless, the amount of these deranged proteins is maintained at a low level in HSCs compared to that of myeloid progenitors [142]. It has been demonstrated that rates of protein synthesis in HSCs are low compared to any other populations of hematopoietic cells, which could allow HSCs to prevent the synthesis of aberrant proteins and also to degrade misfolded proteins [143]. The observation that an artificial increase in misfolded proteins reduced the HSC pool and impaired HSC functions indicates the fundamental role of proteostasis in the homeostatic hematopoiesis [142].

The more active proteostasis mechanism against misfolded proteins, termed “autophagy-lysosome system,” has been shown to be a prerequisite for maintaining integrity of HSCs. The most characterized form of autophagy is macroautophagy, where autophagosomes encompassing target proteins and organelles are transported to lysosomes for degradation [141]. HSCs deficient in proteins involved in autophagosome formation, including autophagy-related 5 (Atg5), Atg7, and Atg12, exhibit mitochondrial activation and increased ROS levels, which in turn leads to impaired self-renewal and differentiation potentials, establishing the critical role of autophagy in quality control of mitochondria as well as HSCs [123–125]. An age-dependent decline of macroautophagy appears to be accountable for dysfunction of diverse tissues [141]; however, the capacity for macroautophagy reportedly remains intact in aged HSCs [144]. Remarkably, a study by Ho et al. demonstrated that aged HSCs with low autophagy levels displayed mitochondrial activation, ROS elevation, and increased DNA damage associated with impaired regeneration potential, which are all typical features of aged HSCs. By contrast, aged HSCs with high autophagy levels showed the opposite results to those with low autophagy levels, which was rather akin to young HSCs. The same experiments using young HSCs showed no difference in their repopulating potential according to autophagy levels, implying the specific dependence on autophagy in aged HSCs [124]. Together, these results suggest that macroautophagy maintains quiescence of HSCs with a low metabolic state to prevent an age-dependent functional decline of HSCs by eliminating activated mitochondria.

Another type of autophagy is referred to as chaperone-mediated autophagy (CMA). In CMA, only proteins bearing a specific targeting motif, KFERQ-like pentapeptide sequence, are degraded. Target proteins with this motif are recognized by the cytosolic chaperone HSC70 and are subsequently transported to lysosomes, where they are internalized via the lysosome-associated membrane protein type 2A (LAMP2A) receptor and then degraded [145]. Dong et al. described reduced CMA activity in aged HSCs compared to young HSCs, whereas no change in CMA activity was observed in granulocyte-myeloid progenitors (GMPs) during aging. Mice lacking Lamp2a showed impaired glycolysis and fatty acid desaturase 2 (FADS2)-dependent fatty acid metabolism to reduce the self-renewal of HSCs, probably due to the failure to CMA-mediated degradation of enzymes involved in these metabolic processes, including FADS2. Notably, genetic or pharmacological activation of CMA was reported to restore glycolysis and FADS2-dependent fatty acid metabolism, leading to improvement of aged HSC functions [126].

Taken together, these data indicate the specific requirement of protein integrity for maintenance and function of HSCs, and that functional declines in proteostasis mechanisms contribute to the aging of HSCs.

Sirtuins

The connection between the cellular metabolic state and HSC function has also been described in studies focusing on the sirtuin family of NAD⁺-dependent protein deacetylases that regulate mitochondrial biogenesis, anti-oxidative defense, and protein folding. Mammalian sirtuin proteins are composed of seven homologs of the yeast Sir2 protein and are localized in the nucleus (Sirt1, Sirt6, Sirt7), cytoplasm (Sirt2), or mitochondria (Sirt3, Sirt4, Sirt5), implying their pleiotropic roles in metabolic homeostasis [146].

Sirt1 decreases ROS levels to maintain HSC functions by promoting nuclear localization and subsequent activation of Foxo3 [147]. Sirt3 and Sirt6 deacetylate target proteins to prevent HSCs from functional decline with age. Sirt3 deacetylates SOD2 to improve its function and reduce oxidative stress, which maintains mitochondrial functions and preserves the self-renewal capacity of HSCs [148]. Sirt6 interacts with lymphoid enhancer binding factor 1 (LEF1), a Wnt signaling transcription factor, to deacetylate histone H3 at lysine 56 and downregulate Wnt target genes, which in turn contributes to maintaining HSC functions [149]. Recent studies have identified mitochondrial proteostasis as an integral mechanism for homeostatic hematopoiesis. In one study, Sirt7 inhibited nuclear respiratory factor 1 (NRF1) to suppress mitochondrial biogenesis, leading to the activation of a mitochondrial unfolded protein response to ensure the integrity of HSCs [127]. In another study, Sirt2 inhibited age-related activation of NACHT, LRR, and PYD domains-containing protein 3 (NLRP3) inflammasome caused by mitochondrial oxidative damage and protein folding stress to retain HSC functions during aging [150]. (See also the “Proteostasis” section in detail).

The significance of these findings in aging research comes from the observations that expression levels of Sirt1, Sirt2, Sirt3, and Sirt7 are decreased during aging and that their exogenous expressions can ameliorate HSC functions [127, 147, 148, 150]. These data support the potential of sirtuins as therapeutic targets for rejuvenation of HSCs.

Polarity

One of the characteristic features of aged HSCs is a “polarity shift.” Florian et al. demonstrated that young HSCs are in a polar state in which cytoskeletal proteins are asymmetrically distributed, while aged HSCs are in an apolar state [26]. Mechanistically, aged HSCs upregulate Wnt5a to induce transition from canonical to non-canonical Wnt signaling, resulting in activation of a small Rho GTPase Cdc42. The activated Cdc42 causes a polarity shift through regulation of cytoskeletal structures mediated by actin and tubulin organizations. Moreover, activation of Cdc42 induces downregulation of nuclear envelope protein Lamin A/C, leading to apolar distribution of H4K16ac, altered compartmentalization of heterochromatin, and increased nuclear volume in aged HSCs [151]. Treatment with Wnt5a reportedly induced Cdc42 upregulation, apolarity, myeloid-skewed differentiation, and impaired repopulating potential of young HSCs. All these features were similar to aged HSCs, demonstrating a causal link between a polarity shift and HSC aging [25].

Cell-extrinsic factors

In response to external demand for hematopoiesis, HSCs proliferate and differentiate to produce blood cells to cope with emergencies such as hemorrhage and infection. Age-related changes of the extracellular milieu can also affect HSC functions that account for characteristic features of hematopoiesis observed in the elderly. Specifically, chronic inflammation frequently observed in the elderly can promote the initiation of age-related diseases in multiple tissues. The hematopoietic system has recently been shown to be no exception. Cell-extrinsic factors affecting HSC aging are summarized in Table 2.

Table 2.

Cell-extrinsic factors affecting HSC aging

Age-related changes	Description	Rationale
Niche aging	Age-related changes of the bone marrow microenvironment can impact on HSC functions	Aged bone marrow cells transplanted into irradiated young recipient mice exhibit a high engraftment and a balanced differentiation [17, 206]
Changes in vasculature	Arterioles are shortened and decreased with age while sinusoids are preserved [184, 187–189]	Aged HSCs detach from arteriolar regions, whereas their proximity to sinusoidal niche are preserved [184, 188, 189]
		Label-retaining HSCs with highest regenerative potential predominantly reside in the sinusoidal niche in aged mice [184]
Dysfunction of ECs	Aged ECs exhibit increased vascular leakiness to elevate ROS levels associated with downregulation of Scf and Cxcl12 [187, 190]	Infusion of aged ECs after myeloablative irradiation abrogates repopulating potential of young HSCs in transplant experiments [190]
Dysfunction of MSCs	Aged MSCs exhibit skewed differentiation into adipocytes [191–193]	Adipocytes impair repopulating potential of HSCs [196, 197]
	Aged MSCs present cellular senescence, reduced colony forming ability, and downregulation of Cxcl12, Scf, and Angpt1 [188, 198, 199]	Deletion of Cxcl12 or Scf from LepR + MSCs depletes HSCs [160, 182]
Inflammaging	Aging is associated with a chronic, low-grade, systemic inflammation [15, 201]	The levels of inflammatory cytokines are elevated in the bone cavity [189, 202–204]
	Senescent stromal cells (e.g. MSCs) can secrete SASP factors to cause inflammatory milieu, leading to a functional decline of HSCs [189, 198, 204, 206–208]	Aberrant inflammation can provoke exhaustion of HSCs, bone marrow failure, and thrombotic disorders [9, 209, 210, 214, 215, 218–221]
		Aged MSCs secrete SASP factors to induce expression of pro-inflammatory genes and reduce clonogenic potential [198]
		Oral administration of the senolytic drug ABT263 mitigates engraftment and myeloid-biased differentiation of aged HSCs [269]
(+ Clonal hematopoiesis)	HSCs harboring CHIP-related mutations acquire growth advantage over wild-type HSCs under the inflammatory environment [245, 247, 248]	Loss of Dnmt3a in HSCs disrupts DNA methylations to outgrow wild-type HSCs under the chronic inflammation [245]
		Tet2 -deficient HSCs display growth advantage and myeloid-biased differentiation by inflammatory stimulations [247, 248]
	CHIP-related genes function as a suppressor of the inflammatory response to prevent inflammaging and thrombotic diseases [221, 250]	Tet2 -deficient HSCs produce a higher level of IL-6 compared to wild-type HSCs in response to LPS injection [250]
		Tet2 loss promotes NLRP3 inflammasome-mediated secretion of IL-1β by macrophages to deteriorate atherosclerosis [221]

Microenvironment

Overview of the niche

HSCs require an extremely specialized microenvironment termed a niche to retain their identity as stem cells [152, 153]. Previous studies have identified the roles of niche cells in hematopoiesis using genetically manipulated mice that enable the tracing or depletion of a specific cell lineage and the deletion of genes encoding soluble factors combined with high-resolution microscopy imaging of bone marrow. According to the current understanding, the basic factors in the niche that affect the HSC fate are the location in the bone cavity, stromal cells in the specific location, and soluble factors secreted by stromal cells.

Initial research identified the endosteal niche located in the vicinity of a bone surface, where osteoblasts are expected to maintain HSC functions [154–159]. However, subsequent studies have indicated that osteoblasts do not have an essential role in HSC maintenance, although they may regulate lymphogenesis, and indicate the dependence of HSCs on vascular niche [160]. The vascular niche comprises the arteriolar niche, which is innervated by β2 and β3 adrenergic nerves [161–163], and the sinusoidal niche [160, 164]. Mesenchymal stem cells (MSCs) and endothelial cells (ECs) have HSC-supportive roles in the vascular niche. MSCs retain the ability to self-renew and differentiate into multi-lineage cells, such as osteoblasts, adipocytes, and chondroblasts [165]. ECs organize the inner wall of blood vessels. MSCs are typically marked as Nestin (Nes)-green fluorescent protein (GFP)-positive cells where GFP expression is controlled by regulatory elements of the nestin gene [166]. These cells are highly enriched for soluble factors indispensable for HSC maintenance, including stem cell factor (SCF) and CXC-chemokine ligand 12 (CXCL12), although the levels of these factors in ECs are much lower than the levels in MSCs [167]. Nes-GFP bright peri-arteriolar cells were shown to be positive for the classical pericyte marker NG2. By contrast, Nes-GFP dim peri-sinusoidal cells overlapped with leptin receptor (LepR)-expressing cells that were enriched for Cxcl12 [167–170]. In addition, differentiated blood cells, such as megakaryocytes [171–174], macrophages [175–179], and regulatory T cells [180, 181], have been shown to be involved in HSC maintenance, indicating the role of their progeny as a niche factor.

It remains elusive which bone marrow niche is predominant in steady state hematopoiesis. Depletion of Cxcl12-abundant reticular (CAR) cells in MSCs or ECs has been shown to decrease HSCs, whereas that in osteoblasts had no effect on HSCs, emphasizing the fundamental role of the vascular niche in HSC maintenance [160, 164, 167, 169, 182–185]. However, further research seems to be required to reach consensus on the specific role of arteriole and sinusoidal niches [167, 169, 183–186]. The structure of the HSC niche and roles of niche factors are summarized in detail in Fig. 2.

Fig. 2.

Aging of the bone marrow niche. HSC aging can be caused by age-related changes in extracellular factors, including tissue structures, stromal cells, and soluble molecules. The HSC niche is composed of an endosteal niche and a vascular niche that is further classified into arterial niche, which is innervated by the sympathetic nerve, and the sinusoidal niche. Recent studies suggest that aged HSCs predominantly reside in the sinusoidal niche accompanied by arterial shortening and diminution with age. Regarding innervation of the bone marrow, sympathetic denervation and a functional switch of neurotransmission from β3-AR to β2-AR during aging are responsible for HSC aging. Further, an age-dependent functional decline of MSCs and ECs that secrete prerequisite niche factors for HSC maintenance such as CXCL12 and SCF has a causative role in HSC aging. In addition, aged MSCs exhibit skewed differentiation into adipocytes to inhibit HSC functions. Progenies of HSCs like megakaryocytes and regulatory T cells regulate HSC functions. Megakaryocytes increase with age and secrete TPO, CXCL4, and TGF-β to maintain the quiescent state of HSCs in the sinusoidal niche. The distance between HSCs and megakaryocytes increases with age, implying the contribution to age-associated decline in HSC functions. Regulatory T cells confer immune-privilege on HSCs through secretion of IL-10 and adenosine, which may be crucial for engraftment of transplanted HSCs

Niche aging

As with other tissues, the microenvironment of the bone marrow is subject to both internal and external stressors, which can lead to a decline in HSC-supportive functions. Indeed, aging is associated with remodeling of the bone cavity. The bone marrow vasculature changes in aged mice, where arterioles are shortened and decreased while sinusoids are preserved [184, 187–189]. Consistent with these observations, aged HSCs detach from endosteal and arteriolar regions, whereas their proximity to the sinusoidal niche are preserved [184, 188, 189]. Saçma et al. demonstrated that label-retaining HSCs, which retained the highest regenerative potential in aged mice, predominantly resided in the sinusoidal niche, whose morphology and number were uniquely preserved during aging [184]. These data imply a transition of HSCs from endosteal or arteriole niche to the sinusoidal niche during aging.

Moreover, aged mice exhibit an increase in vascular leakiness associated with elevated ROS levels and decreased expressions of HSC-supportive factors, such as Scf and Cxcl12 in ECs [187, 190]. Interestingly, infusion of aged ECs after myeloablative irradiation reportedly abrogated the repopulating potential of young HSCs compared to the infusion of young ECs, suggesting that the aging of ECs contributes to the functional deterioration of HSCs [190]. MSCs exhibit skewed differentiation into adipocytes with age [191–193], which could account for bone fragility and fatty marrow observed in the elderly [194, 195]. Adipocytes impair the repopulating potential of HSCs and delay the healing of bone fracture, implying the contribution of MSC dysfunction to an age-related involution of skeletal and hematopoietic systems [196, 197]. Furthermore, aged MSCs exhibit cellular senescence, reduced colony forming ability, and downregulation of HSC maintenance factors, including Cxcl12, Scf, and Angpt1, indicating a functional impairment of MSCs over time [188, 198, 199].

Maryanovich et al. demonstrated that age-associated attrition of arterioles is accompanied with loss of innervation by sympathetic nerves [188]. The authors described that surgical denervation of young bone marrow caused a decline in repopulating potential of HSCs and a decrease in the number of arterioles accompanied by dysfunction of MSCs, reminiscent of physiological aging. However, the impact of age-dependent alterations of the sympathetic nervous system is somewhat controversial, given that the findings of Ho et al. were opposite in nature. The authors described that sympathetic innervation was increased in aged bone marrow [200]. They argued that a functional switch of neurotransmission from β3-adrenergic receptor (AR) to β2-AR was the cause of HSC aging. Despite their conflicting findings, both studies emphasized the importance of β3-AR signaling in maintaining HSC functions, as treatment with β3-adrenergic agonist improved functions of HSCs in aged mice or the progeria mouse model.

Inflammation

One of the most important topics to be elucidated in hematopoietic aging is a chronic, low-grade, systemic inflammation termed “inflammaging.” Inflammaging manifests as anemia, myeloid-biased hematopoiesis, thrombocytosis, and immunosenescence associated with elevated serum levels of inflammatory cytokines that include interleukin (IL)-6 and C-reactive protein (CRP) [15, 201]. It is inevitable that HSCs are subjected to inflammaging since the levels of inflammatory cytokines in the bone cavity are elevated with age [189, 202–204]. Previous studies have suggested that infection, immunosenescence, and SASP may underlie inflammaging, although the accurate source of inflammatory cytokines remains elusive. Aged macrophages [202], plasma cells [205], and stromal cells [189, 204, 206–208] in the bone marrow display increased levels of inflammatory cytokines, which in turn led to myeloid- and platelet-biased hematopoiesis. These data suggest that the inflammatory environment in the bone cavity due to aging of stromal cells plays a causative role in inflammaging of HSCs.

The response of HSCs to inflammatory stimuli is quite similar to their age-dependent functional changes. Inflammatory signals stimulated by IL-1 [209], TNF-α [207], interferon (IFN)-α [210], IFN-γ[211], lipopolysaccharide (LPS) [212, 213], and polyinosine-polycytidylic acid (pIpC) [210, 214, 215] produce transient proliferation and myeloid-skewed differentiation of HSCs at the expense of lymphogenesis. These effects are associated with changes in the subpopulation of HSPCs, where myeloid-biased HSCs and GMPs are increased while common lymphoid progenitors (CLPs) are decreased [207, 209, 213]. In addition, IL-1, TNF-α, LPS, and pIpC induce expansion of platelet-biased phenotypic HSCs [202, 216] and IL-6 upregulates thrombopoietin (TPO) levels to enhance thrombopoiesis [217].

These responses to inflammatory stimuli are beneficial defense mechanisms that expedite tissue repair, where HSCs exhibit myeloid-biased differentiation to meet the demand for myeloid cells and platelet-biased differentiation to replenish platelets for hemostasis in exchange for functional integrity of HSCs. However, it should be kept in mind that an aberrant inflammation can provoke exhaustion of HSCs [209, 210, 214, 215, 218], bone marrow failure [219, 220], and thrombotic disorders [9, 221]. Moreover, inflammation-driven proliferation of HSCs stimulates increased levels of mitochondrial ROS and DNA damage [215], suggesting that inflammation can be a cause of mutations, as evident by some inflammatory diseases (e.g. hepatic cirrhosis, ulcerative colitis, and Barrett's esophagus) [222–226].

Clonality

Aging is accompanied by clonal competition

As described earlier, an age-dependent functional decline of HSCs is often associated with paradoxical expansion of phenotypic HSCs (pHSCs) [16, 31]. This observation implies that HSCs are affected by a driving force for proliferation over time, which may lead to outgrowth of a malignant clone (i.e. cancer), one of the most problematic aspects in the aging of organisms. This prediction is illustrated by a phenomenon termed “clonal hematopoiesis of indeterminate potential (CHIP)”, where blood cells harboring somatic mutations are clonally propagated in spite of the normal hematologic state [6–8]. The current criterion for CHIP is the presence of somatic mutations with a variant allele frequency (VAF) exceeding 2% in blood cells (corresponding to 4% of blood cells with heterozygous mutations) [227]. This value is determined by the detection limit of standard sequencing technologies. Sequence analyses combined with epidemiologic studies have elucidated that CHIP is unexpectedly prevalent among healthy people and commonly increases with age. The occurrence of CHIP is approximately 10% for healthy people over the age of 65. In subsequent studies, CHIP was observed in virtually all of healthy people when the detection limit was lowered [228, 229], indicating that CHIP is an inevitable condition during aging of the hematopoietic system.

Recent studies have clarified the clonal history of HSCs during chronological aging [106, 230–236]. Throughout life, each HSC can acquire somatic mutations stochastically to create a heterogenous HSC pool, which in turn leads to interclonal competition that is subject to cell-intrinsic and -extrinsic factors. This results in either exhaustion, equilibrium, or expansion of the HSC clone over time [6, 237, 238]. Only a few types of mutations that confer moderate self-renewal capacity on HSCs appear to contribute to the formation of discerning clones, since most of mutations are expected to evoke either adverse effects to reduce fitness or neutral effects to confer virtually no effect on proliferation. The most frequently mutated genes in CHIP individuals are the epigenetic regulators DNMT3A, TET2, and ASXL1, which account for approximately 70% of these genes, followed by splicing factors (SF3B1, SRSF2, U2AF1), signaling molecules (JAK2, CBL), and DDR genes (TP53, PPM1D) [6–8]. Notably, mutations observed in CHIP are similar to those in myeloid neoplasms and individuals with CHIP are at increased risk for hematological malignancies and cardiovascular diseases [7, 9]. These findings have established CHIP as a predisposing factor for age-related disorders including both malignant and non-malignant conditions.

Epigenetic drift as a driving force of clonal advantage

The molecular mechanisms by which mutations detected in CHIP confer a clonal advantage on HSCs have been intensively investigated using genetically engineered mouse models. Among them, the impact of mutations in epigenetic regulators on hematopoiesis has been studied in detail. Deletion of Dnmt3a or Tet2 in mice was reported to disrupt DNA methylations and gene expressions to enhance self-renewal of HSCs, leading to the expansion of the HSC compartment [94–96, 239]. Moreover, taking advantage of a genetic mosaic mouse model, HSCs expressing C-terminally truncated form of Asxl1 (Asxl1-MT) have been shown to be increased over time and eventually occupy the HSC compartment in vivo [240]. These data suggest that HSCs with mutations in epigenetic regulators acquire a clonal advantage to outcompete their normal counterparts, recapitulating CHIP in humans.

The reason why epigenetic regulators are most frequently mutated in CHIP remains unclear. Single-cell analyses revealed that chromatin modifications and DNA methylations became increasingly heterogenous among individual cells with the passage of time [107]. Such time-dependent alterations of the epigenome are presumed to be transmitted to offspring as an epigenetic memory to induce epigenetic drift (details are provided in the section dealing with “Epigenetic drift”). It is possible that mutations in epigenetic regulators augment the fluctuation of the epigenome to enhance self-renewal of HSCs, which can underlie the clonal advantage of HSCs and subsequent CHIP in the elderly.

Extrinsic factors causing clonal expansion

External stressors can endow HSCs with a growth advantage in the context of specific mutations. Coombs et al. observed CHIP in 25% of cancer patients, which was related to prior radiation therapy and smoking history. Compared to general population, DDR-related genes TP53, PPM1D, and ATM were more frequently mutated in these patients [241]. In line with this finding, hematopoietic cells with Tp53 or Ppm1d mutations in mice acquire a survival advantage under the condition of radiation and anticancer drug treatments [242–244]. These results suggest that HSCs harboring mutations in DDR-related genes reduce the susceptibility to DNA damage, which is advantageous for cellular survival compared to wild-type counterparts.

Another important cell-extrinsic factor that affects the clonality of HSCs is inflammation. As described earlier, inflammatory stimuli act on HSCs to proliferate to cope with emergencies. Meanwhile, the HSC pool remains in equilibrium over the long-term regardless of inflammation since HSCs undergo terminal differentiation, apoptosis, and cell cycle arrest to suppress excessive proliferation. Nevertheless, recent evidence has suggested that HSCs with CHIP-related mutations override these regulatory mechanisms to acquire a growth advantage over wild-type HSCs under the inflammatory milieu. In response to activated IFN-γ signaling caused by chronic mycobacterial infection, HSCs from Dnmt3a^−/− mice were reported to preferentially expand over those from wild-type mice in competitive transplantations accompanied by reduced differentiation through hypermethylation of the Fos and Jun family of pro-differentiation factors [245]. This study suggests that the loss of Dnmt3a in HSCs disrupts DNA methylations of pro-differentiation genes to reduce differentiation potential, while retaining self-renewal capacity to outgrow wild-type HSCs under the chronic inflammation. Consistent with this study, deletion of Cebpa triggers an aberrant expansion of HSCs accompanied by a deficit in Cebpa-dependent myeloid differentiation in the context of chronic IL-1 treatments, but not in the absence of inflammatory stimuli [246]. These findings further indicate that differentiation defects underlie a chronic inflammation-mediated expansion of the HSC pool.

In addition, several studies have shown that Tet2-deficient HSCs show increased susceptibility to inflammatory stimuli. Meisel et al. demonstrated that an increase in IL-6 level caused by systemic bacterial dissemination in Tet2^−/− mice enforced propagation and myeloid-biased differentiation of HSPCs [247]. Another study revealed that HSPCs from Tet2^−/− mice showed activation of non-canonical nuclear factor-kappa B (NF-κB) signaling via the upregulation of A20 (Tnfip3) in the presence of low-dose chronic inflammation [248], which may lead to a growth advantage of Tet2-deficient HSCs. Furthermore, it has been reported that Tet2 itself functions as a direct suppressor of the inflammatory response. Zhang et al. demonstrated that Tet2 recruits Hdac2 to deacetylate and downregulate IL-6 independent of the DNA hydroxymethylation mechanism [249]. In line with this finding, Tet2-deficient HSCs produce a higher level of IL-6 compared to wild-type HSCs in response to LPS injection [250]. Moreover, Tet2 loss promotes NLRP3 inflammasome-mediated secretion of IL-1β by macrophages to elevate inflammatory response and deteriorate atherosclerosis [221].

Taken together, these studies indicate that mutations in CHIP-related genes render HSCs susceptible to inflammatory stimuli, thereby conferring a growth advantage on HSCs at the expense of differentiation potential. It is plausible that inflammatory environment in aged bone marrow niche contributes to a relative selective advantage of mutated HSCs, which is further potentiated by inflammatory cytokines secreted by themselves as well as macrophages harboring mutations.

Rejuvenation

The general idea for HSC rejuvenation is that pharmacological or genetic manipulations that counteract the cause of aging described above can be a preventive approach. Although currently there are no criteria for HSC rejuvenation, it is widely accepted that mitigation of the phenotypes commonly observed in aged HSCs, such as impaired repopulating potential and myeloid-biased differentiation, is considered to achieve rejuvenation.

One issue regarding HSC rejuvenation is that interventions directed at the external milieu may not be sufficient to restore functions of aged HSCs. Heterochronic parabiosis, a technique where young and aged animals are surgically combined to share circulation, has demonstrated that soluble factors derived from young animals have anti-aging effects on aged stem cells, including those resident in the liver, muscles, and neurons [251–254]. However, Ho et al. recently reported that aged HSCs were refractory to rejuvenation by heterochronic parabiosis [255]. Furthermore, Kuribayashi et al. demonstrated that aged HSCs transplanted in young mice without preconditioning showed no significant functional rejuvenation [256]. Nevertheless, there is enough evidence, showing that interventions to extracellular milieu can rejuvenate HSC functions successfully. Since diverse factors contribute to HSC aging, it may be critical to use multiple drugs with different mechanisms to achieve rejuvenation of HSCs.

The study from Kuribayashi et al. described above also demonstrated the important findings that the young niche partially rejuvenated transcriptional profiles of aged HSCs, while DNA methylation profiles of those were not changed. This result underscores the importance of epigenetic reversion to a younger state for rejuvenating HSCs.

Epigenetics

The experimental rationale for the causative role of epigenetic memory in HSC aging comes from one study where induced pluripotent stem (iPS) cells derived from aged HSCs were capable of re-differentiating into HSCs whose functions were comparable with those of young HSCs [257, 258]. Nitta et al. revealed that overexpression of Bmi1, a component of PRC1, maintained repression of its target genes that were derepressed with age and enforced expression of HSC signature genes, which ameliorated the age-related decline in HSC functions [259]. In addition, two groups reported that treatment with ascorbate (vitamin C), a cofactor for Fe- and 2-oxoglutarate-dependent dioxygenases, restored 5hmC levels of Tet2^−/− HSCs to suppress self-renewal and myeloid-biased differentiation [260, 261]. Since TET2 is one of the most mutated genes in individuals with CHIP, treatment with ascorbate could be a promising strategy to intervene in this condition. Together, these findings indicate that epigenetic modifications may be valuable to rejuvenate aged HSCs and to suppress an age-dependent clonal expansion of HSCs.

Nutrient sensing pathways

Anabolic signaling, including Insulin/IGF-1 and mTOR that sense energy abundance, accelerates tissue aging and dysfunction of stem cells. On the other hand, catabolic signaling, including AMPK and sirtuins that sense energy scarcity, can reportedly extend longevity and rejuvenate stem cells [132]. Prolonged fasting, a typical catabolic situation, decreased myeloid-biased HSCs and restored balanced differentiation in aged mice through suppression of IGF-1/PKA signaling [262]. Zheng et al. showed that mTOR was activated in aged HSCs and its inhibition by treatment with the mTOR inhibitor rapamycin restored impaired repopulating potential and myeloid-biased differentiation of aged HSCs. Intriguingly, rapamycin treatments also suppressed an age-related expansion of the HSC pool [263]. Likewise, we recently demonstrated that rapamycin treatment ameliorated aberrant expansion and differentiation defect of aged HSCs expressing Asxl1-MT [240].

These data raise the possibility that inhibiting the mTOR pathway improves the functions of aged HSCs and prevents a clonal expansion of HSCs observed in the elderly with CHIP. Paradoxically, an age-dependent decline in the level of IGF1 was observed in the bone marrow microenvironment. This attenuated PTEN/AKT/mTOR signaling, which is responsible for some hallmarks of HSC aging, including increased DNA damage, polarity shift, impaired mitochondrial activity, and myeloid-biased differentiation of HSCs [264]. The collective data indicate that modulation of IGF-1/AKT/mTOR signaling pathways may confer beneficial effects on aged HSCs to rejuvenate their functions.

Mitochondrial integrity

Consistent with an age-dependent decline in mitochondrial function, Mansell et al. demonstrated a marked reduction of MMP, one of the indicators of mitochondrial activity, in aged HSCs [265]. The authors showed that pharmacological enhancement of MMP in aged HSCs ameliorated myeloid-biased differentiation and long-term engraftment in vivo. On the other hand, Ho et al. described that aged HSCs with high autophagy activity maintained a low metabolic state and self-renewal potential at similar levels as young HSCs, whereas HSCs with low autophagy activity exhibited an overactive mitochondrial respiration and impaired repopulating potential [124]. It remains unclear why autophagy activities differ among aged HSCs. Clarifying its mechanisms will inform novel therapeutic strategies for rejuvenating HSCs.

Proteostasis

Proteostasis mechanisms contribute to retaining HSC functions over time. This is exemplified by a prerequisite for mitochondrial homeostasis mediated by macroautophagy and mitochondrial unfolded protein response to maintain HSC integrity [123–125, 127, 128]. Furthermore, Dong et al. recently revealed that CMA activity in HSCs declined with age to impair glycolysis and fatty acid metabolism, whose pharmacological or genetic activation rejuvenated metabolic states and functions of aged HSCs [126]. The findings implicate CMA as a potential therapeutic target for HSC rejuvenation.

Sirtuins

Sirtuins are influential in longevity and reportedly rejuvenate aged HSCs. As stated earlier, overexpression of Sirt1, Sirt2, Sirt3, or Sirt7, whose expression is downregulated with age, restores the functions of aged HSCs through divergent mechanisms that include Foxo3 activation, NLRP3 inflammasome inactivation, reduction in ROS levels, and activation of the mitochondrial unfolded protein response, respectively [127, 147, 148, 150]. All of them contribute to mitochondrial homeostasis, which further emphasizes the importance of metabolic homeostasis secured by mitochondrial quality control mechanisms.

Polarity

Polarity shift is a distinct feature of aged HSCs, which contributes to a functional decline during aging [25, 26, 151]. An increase in Cdc42 level induced by age-related shift from canonical to non-canonical Wnt signaling is responsible for the polarity shift. Treatment of aged HSCs with the selective Cdc42 inhibitor CASIN restores polarity shift, differentiation defect, and impaired repopulating potential [25, 26]. Repression of Cdc42 activity also reverts distribution of H4K16ac and nuclear localization of heterochromatin to a young state in aged HSCs [151]. These findings imply that changes in polarity and chromatin architecture are potential therapeutic targets for HSC rejuvenation.

Microenvironment

Cell-extrinsic factors that accelerate HSC aging are promising targets for rejuvenation since they can be modulated externally. Specifically, rejuvenation of the aged niche may be a feasible strategy, as aged bone marrow cells transplanted into irradiated young recipient mice exhibited a high engraftment and a balanced differentiation compared with those transplanted into aged mice. Conversely, when young bone marrow cells were transplanted into irradiated aged recipient mice, they showed impaired repopulating potential and myeloid-skewed differentiation compared with those transplanted into young mice, suggesting a dependence of HSC functions on the bone marrow niche [17, 206]. It has been reported that aged ECs exhibit increased vascular leakiness and decreased expression of HSC-supportive factors. Infusion of young ECs into irradiated recipient mice improves the repopulating potential of aged HSCs when compared to that of aged ECs [190]. In addition, sympathetic denervation contributes to a functional decline of both HSCs and MSCs during aging, which is significantly mitigated by treatment with a β3-adrenoreceptor (AR) agonist [188]. Altogether, these data suggest that modulations of the niche functions may provide therapeutic opportunities for the physiological aging of HSCs.

Senolysis

Cellular senescence is a beneficial process, wherein an irreversible cell cycle arrest contributes to suppress uncontrolled cellular proliferation and development of cancers. Although cellular senescence itself is advantageous for tissue homeostasis, accumulation of senescent cells appears to be harmful. It has been shown that an artificial clearance of senescent cells (“Senolysis”) restores an age-related functional deterioration of multiple organs, including heart, kidney, and bone marrow [266–268]. Remarkable effects of senolysis include extended lifespan and delayed tumor onset, indicating potential therapeutic value for longevity. Oral administration of the senolytic drug ABT263, which is a specific inhibitor of the anti-apoptotic proteins BCL-2 and BCL-xL, was shown to effectively removed senescent cells of aged mice and mitigate impaired engraftment and myeloid-biased differentiation of aged HSCs [269]. Whether removal of senescent HSCs contributes to the rejuvenation of the remaining HSCs is still unclear, since whether HSCs actually undergo senescence is still debatable. Alternatively, it is possible that the elimination of senescent stromal cells, such as MSCs, which adopt a senescent state over time [198], reduces secretion of SASP factors to ameliorate inflammatory milieu, contributing to the functional mitigation of aged HSCs.

Perspective

Progress in sequencing technologies brings new opportunities for aging research

Aging is associated with changes in the number of hematopoietic cells in each population as described earlier. Since classical bulk sequencing represents average signals across all cellular populations, it is impossible to determine whether changes in sequencing data are due to population changes or changes that have occurred in each cell. Recent single-cell technologies enable us to interrogate differences among individual cells, which is more suitable to dissect heterogenous populations, such as hematopoietic cells and bone marrow stromal cells [270, 271]. Single-cell RNA-seq (scRNA-seq) analysis revealed that differentiation of aged HSCs is skewed toward myeloid/platelet lineage accompanied by elevated expression of myeloid/platelet-specific genes at the single-cell level [272]. scRNA-seq also demonstrated that mesenchymal progenitors decreased in the number and upregulated adipocyte markers to skew differentiation toward adipocyte lineages [273]. A single-cell transcriptomic atlas of aging tissues—Tabula Muris Senis—is now publicly available, which provides new insights into aging of the hematopoietic system [274]. Taking advantage of this atlas, the authors showed that clonality for both B cell and T cell repertoires are increased in aged mice. This result suggests one of the mechanisms of immunosenescence, where aged individuals manifest a higher vulnerability to new pathogens compared to younger ones.

It is possible that single-cell technologies combined with known aging factors, such as genome stability, epigenetics, metabolism, immunity, polarity, and niche factors uncover the unexpected common mechanisms underlying these age-related changes to provide us an integrated view of HSC aging.

Overcoming clonal heterogeneity

One of the most intriguing recent findings in hematopoiesis is that clonal expansion of hematopoietic cells harboring somatic mutations is observed virtually all people [6–8, 228, 229]. This suggests that our hematopoietic system is composed of multiple clones harboring somatic mutations, presumably causing transcriptional heterogeneity among hematopoietic cells. Single-cell technologies make it possible to investigate genetic mutations and transcriptional status simultaneously. Miles et al. mapped clonal trajectories from CH to AML using single-cell DNA sequencing (scDNA-seq) and found that AML clones exhibited complex evolutional trajectory relative to CH and clonal dominance is affected by mutational combinations in each clone [275]. Moreover, simultaneous analysis of scDNA-seq and immunophenotyping revealed that expression of cell-surface proteins differs among clones depending on the mutational background. Similarly, single-cell multi-omics approach to integrate mutations, transcriptome, and methylome using HSPCs from individuals with DNMT3A mutated CH has also been attempted [276]. These studies will enable a precise evaluation of genotype–phenotype correlation at the single-cell level to overcome heterogeneity of the aged hematopoietic system.

Social demands for aging research

In developed countries, and definitely in the developing countries in the near future, a rapidly aging population is one of the biggest problems in society. Aging is an obvious risk factor for the development of chronic diseases, such as heart disease, cerebrovascular disease, dementia, diabetes, and cancer [277]. These conditions impair the quality of life and increase the burden on medical resources. Therefore, both basic and applied research on aging is crucial to address the accelerating social demands. Advances in understanding the mechanisms of HSC aging could provide opportunities to resist aging, which is expected to resolve some aspects of social problems arising from the rapid aging of the population. Although some research achieved successful rejuvenation of HSCs in vivo, we do not know whether they can be true of human, since most of these findings are obtained from model organisms. Nevertheless, the past studies have uncovered the similarities in the basic concepts of hematopoiesis between human and mouse, which could be extended to aging of the hematopoietic system. This notion is exemplified by an expansion of phenotypic HSCs with myeloid-biased differentiation potential in aged human [278].

People do not live in a uniform manner; thus, aging will be different among different people. Such a heterogeneity can be a barrier to establish successful interventions in HSC aging. Single-cell omics approach will be a valuable strategy to evaluate age-related changes in hematopoiesis annotated with mutational trajectory among individuals. Gathering findings from the past studies and new technologies, the day would come when personalized treatments, combined with a genetic approach, are available to address aging.

Author contributions

The manuscript was written by TF, SA, SG, and TK took part in writing the manuscript. All authors read and approved the final manuscript.

Funding

This work was supported by Grant-in-Aid for Scientific Research (A) (No. 20H00537) (to T.K.), Grant-in-Aid for Scientific Research on Innovative Areas “Stem Cell Aging and Disease” (No. 17H05634), The Tokyo Biochemical Research Foundation (to T.K.), Japanese Society of Hematology (to T.K.).

Data availability

Not applicable.

Declarations

Conflict of interests

The authors have no relevant financial or non-financial interests to disclose.

Ethics approval

Not applicable.

Consent to participate

Not applicable.

Consent to publish

Not applicable.