Irrespective of their source, MSCs enter a state of replicative senescence (i.e., in vitro aging, also known as the Hayflick limit) after repeated serial passage in culture when the cells stop dividing after a certain number of PDs (Hayflick and Moorhead, 1961). The maximum number of PDs that can be achieved by MSCs is estimated to be 30 to 40 (Banfi et al., 2002; Baxter et al., 2004). No clear information on passage number has been provided for the 15 MSC products approved to date for clinical use (Table 1). However, given that the differentiation potential of MSCs decreases after extended passages, low-passage cultures are recommended for clinical-scale expansion of cultures (Lechanteur et al., 2016). In the following sections, we discuss heterogeneity and biological and functional changes in MSC senescence (Figure 2).

Despite their global features as defined by the ICST, MSCs are complex cell populations that exhibit heterogeneity depending on the donor, tissue source, and whether they are clonal populations or single cells (Phinney, 2012). MSC heterogeneity comprises proliferation rate, morphology, immunophenotype, multilineage differentiation potential, and senescence (Schellenberg et al., 2012). In symmetric cell division, a self-renewing parent cell divides into two daughter cells with comparable shape and differentiation potential. In contrast, asymmetric cell division yields a self-renewing cell and a non-dividing cell that becomes senescent in culture. These dynamics result in an initially dominant cell population being overtaken by other clonal populations after multiple passages (Figure 3). Heterogeneity in the proliferation potential of cultured MSCs manifests morphologically as subpopulations of small, round, rapidly proliferating cells and slowly dividing, large flattened cells (Mets and Verdonk, 1981; Colter et al., 2001). Using the limiting dilution assay at later passages, it was determined that not every cell is capable of clonal expansion and colony formation at the time of culture establishment (Schellenberg et al., 2012). More importantly, the number of colony-forming unit (CFU) fibroblasts decreased continuously during culture expansion, and were scarcely detected after >20 passages (Schellenberg et al., 2012). Likewise, clonal analysis of single-cell–derived colonies has suggested that not every cell has trilineage (i.e., osteogenic, adipogenic, and chondrogenic) potential (Wagner et al., 2008), and subsets with high differentiation potential rapidly decline in number after a few passages (Schellenberg et al., 2012). The age-related heterogeneity of MSCs is thought to be associated with epigenetic status; subpopulations with variable expression of stem cell antigen (Sca)-1 regained the Sca-1 profile of the parent cell after 4–8 days of culture, which was accompanied by epigenetic changes at the lymphocyte antigen 6 complex promoter (Hamidouche et al., 2017). Given the relevance to clinical efficacy, molecular markers of aging in cultured MSCs are needed to reflect senescence-associated alterations. The current understanding of aging-related cellular changes is based primarily on homogeneous bulk cell-derived data; emerging tools for single-cell analysis can help to define the heterogeneity of MSCs.

As MSC populations with a large proportion of senescent cells are less effective when transplanted, it is critical to detect senescent MSCs during expansion. Surface marker profiling is one approach for identifying and purifying senescent cells from a culture. For example, the expression of CD146—also known as melanoma cell adhesion molecule (MCAM)–was downregulated in MSCs derived from aged donors compared to those from young donors, as well as in MSCs after prolonged in vitro expansion (Gnani et al., 2019). Furthermore, low but not high CD146 expression was associated with a senescent phenotype in MSCs (Jin et al., 2016), and CD146⁺ MSCs showed increased migratory potential toward degenerating tissues (Wangler et al., 2019). CD264 is another surface marker of in vitro aging in MSCs that is unrelated to the chronologic age of the donor (Madsen et al., 2017); cells expressing this protein exhibit increased senescence-associated β-galactosidase (SA-β-gal) activity and reduced differentiation potential and colony-forming efficiency compared to CD264^– MSCs (Madsen et al., 2020). Other surface markers that show altered expression with in vitro/in vivo aging are summarized in Table 2. It should be noted that although these molecules are expressed to varying degrees during the aging process and may be associated with functional changes, there is presently no consensus on whether they can serve as the gold standard for selective purification of young vs. old MSC populations.

Early-passage MSCs are small and have a fibroblast-like spindle shape but acquire a hypertrophic and flat morphology with more podia and actin stress fibers upon extended culture (Stenderup et al., 2003; Mauney et al., 2004; Stolzing and Scutt, 2006). MSCs from passages 1 to 3 have a uniform size but begin to enlarge at passage 5, such that cells at passages 6–9 are 4.8-fold larger than passage 1 cells (Oja et al., 2018). Morphologic features can predict how well cells can adapt to a given condition. Cell and nuclear morphology in MSCs in the first 3 days of osteogenic induction was found to be closely correlated with their long-term (35-day) mineralization capacity (Marklein et al., 2016). A subset of aged MSCs with a small cell size had ATP levels equivalent to those in young MSCs, whereas levels in large-sized cells were comparable to those in the aged parent MSC population (Block et al., 2017). Increased cell size and granularity were positively correlated with MSC autofluorescence; the latter has therefore been proposed as a non-invasive, real-time quantifiable marker for cellular senescence (Bertolo et al., 2019).

Animal cells sense and correct deviations in size by adjusting cell cycle length as well as growth rate (Miettinen et al., 2014), which is increased in small cells and reduced in large cells (Ginzberg et al., 2018). Loss of cell size uniformity can indicate abnormal biosynthesis or cell cycle progression. While the regulation of size homeostasis in relation to senescence is not fully understood, cell enlargement can distinguish senescent subpopulations in culture. Among aged MSCs, SA-β-gal activity is increased in large as compared to small-sized cells (Block et al., 2017). Notably, MSCs immortalized by SV40 (Negishi et al., 2000) or telomerase transfection (Kobune et al., 2003) are significantly smaller than their parent cells.

The enlargement of aging cells and their transformation to a hypertrophic morphology is accompanied by biological changes. A decline in proliferative capacity was reported in MSCs derived from old patients as compared to their healthy young counterparts (Banfi et al., 2002) and in long-term MSC cultures regardless of the cell source. MSCs from young donors had greater mitotic activity (41 ± 10 vs. 24 ± 11 PDs), slower progression to senescence, and an increased rate of proliferation (0.09 ± 0.02 vs. 0.05 ± 0.02 PDs/day) than those from old donors (Stenderup et al., 2003). CFU is a retrospective parameter describing the clonogenic potential of a single cell; decreases in CFU and average colony size are correlated with MSC aging in vitro (Liu et al., 2004).

In addition to impaired proliferation, aging negatively affects MSC migration and homing ability (Liu et al., 2017). Directed migration toward stimuli by MSCs is critical for better functional outcomes in cell-based therapy. Impaired migratory capacity in senescent MSCs in response to pro-inflammatory signals was found to be closely associated with activator protein (AP)-1 pathway inhibition (Sepulveda et al., 2014). Cell migration involves the reorganization of the actin cytoskeleton (Le Clainche and Carlier, 2008). MSCs derived from old donors exhibit reduced response to biological and mechanical signals because their actin cytoskeleton is less dynamic (Kasper et al., 2009). Gene expression profiling has identified several cytokines and chemokines and their receptors important for cell migration–including stromal cell-derived factor 1 (SDF-1) and its receptor chemokine receptor type 4 (CXCR4), tumor necrosis factor receptor (TNFR), IFN-γ receptor (IFNGR), and C-C motif chemokine receptor 7 (CCR7)—that are downregulated in aged MSCs as compared to younger cells (Geissler et al., 2012; Bustos et al., 2014).

A basic strategy for MSC-based regeneration is to replace cells that are lost or impaired by disease with functional cells. It was previously thought that MSCs exert their therapeutic effect through trans-differentiation. During in vitro culture, MSCs progressively lose their capacity to differentiate into adipogenic and osteogenic lineages although the preferred fate is debated, with some studies suggesting that aging shifts the balance in favor of adipocytes at the expense of osteoblastogenesis (Stolzing et al., 2008), and others reporting that osteogenic activity is preserved or even increased in late passages (Wagner et al., 2008) or that both osteogenic and adipogenic potential is lost (Geissler et al., 2012). Age-associated changes in differentiation potential may be related to altered susceptibility to reactive oxidative species (ROS) and apoptosis (Bruedigam et al., 2010); MSCs undergoing osteoblast differentiation showed dose-dependent increases in apoptosis and ROS accumulation upon treatment with rosiglitazone, whereas adipogenesis was unaffected (Jiang et al., 2008). Lineage bias in differentiation is regulated by key signaling pathways, intracellular oxidative stress, and transcriptional and post-transcriptional mechanisms. Gene expression analysis has revealed an age-related downregulation of osteoblast transcription factors such as core binding factor α1 (CBFA1), runt-related transcription factor 2 (Runx2), and distal-less homeobox 5 (DIx5) as well as collagen and osteocalcin, and upregulation of adipogenic factors such as peroxisome proliferator-activated receptor-γ (PPAR-γ) and adipocyte fatty acid-binding protein (aP2) (Jiang et al., 2008). The target genes activated by PPAR-γ are related to lipid metabolism and adipocyte differentiation. Thus, the age-related increase in PPAR-γ expression shifts the fate of MSCs toward adipogenesis, and Wnt/β-catenin signaling regulates MSC differentiation by suppressing PPAR-γ and biasing differentiation toward osteoblastogenesis (Xu et al., 2016).

Aging cells acquire a senescence-associated secretory phenotype (SASP) involving the secretion of proteins that can affect the behavior of neighboring cells via autocrine/paracrine mechanisms (Borodkina et al., 2018; Campisi et al., 2019). MSCs have potent anti-inflammatory and immunosuppressive functions and thus have therapeutic potential for inflammation-related diseases. Aged MSCs have a diminished capacity for inhibiting the proliferation of allogeneic peripheral blood mononuclear cells compared to younger cells (Gnani et al., 2019). The activation of SASP factors such as interleukin 6 (IL-6), IL-8, and monocyte chemotactic protein 1 (MCP1) in the conditioned medium of aged MSCs was shown to be increased compared to young MSC cultures at early passages, an effect that was exacerbated at late passages (Gnani et al., 2019). Secretion of SASP-related chemokines/cytokines not only drives responses that reinforce senescence in a cell-autonomous manner but also acts on neighboring cells via a paracrine mechanism to accelerate senescence. For example, factors secreted by aged MSCs were shown to activate pro-inflammatory gene expression in young hematopoietic stem cells and decreased their clonogenic potential (Gnani et al., 2019).

The beneficial effects of MSC-based therapy are attributable to the action of pro-angiogenic paracrine factors. In aged MSCs, the secretion of these factors—including vascular endothelial growth factor (VEGF), placental growth factor (PGF), and hepatic growth factor (HGF)—is reduced, whereas that of anti-angiogenic factors such as thrombospondin-1 (TBS1) and plasminogen activator inhibitor-1 (PAI-1) is increased. Thus, age negatively affects angiogenesis and directly undermines the therapeutic efficacy of MSCs (Efimenko et al., 2011; Khan et al., 2011).

β-D-Galactosidase (β-Gal) is a eukaryotic hydrolase localized in the lysosome that is active at the optimal pH (6.0) in senescent cells but is absent in proliferating cells (Dimri et al., 1995). SA-β-gal activity is suggested as the gold standard for evaluating senescence in cells and can be detected by cytochemistry/histochemistry and fluorescence-based methods. However, when used in combination with other markers, it can yield false-positive/negative results in quiescent cells or upon stress (de Magalhaes et al., 2004; Yang and Hu, 2005). Senescence-associated lysosomal α-L-fucosidase (SA-α-Fuc) has recently been identified as a more robust biomarker in all types of cellular senescence (Hildebrand et al., 2013; Singh and Piekorz, 2013), but there is still limited evidence for its sensitivity and specificity in distinguishing senescent MSCs.

Telomeres are specialized nucleoprotein caps containing repetitive nucleotide sequences that protect chromosomes from end-to-end fusion and prevent the loss of genetic information during DNA replication (Sahin and Depinho, 2010). Telomeres shorten with every cell division and senescence is triggered when they reach a critical length (Baird et al., 2003). As such, telomere length has been used to estimate replicative history and predict senescence in MSCs (Montpetit et al., 2014). There is increasing evidence of an association between diminished proliferative capacity and telomere shortening in MSCs. However, the exact telomere length in senescent MSCs and whether it differs according to cell source, culture conditions, and measurement method is unclear. For example, a telomere length of 10 kb was proposed as a threshold for senescence (Baxter et al., 2004), although another study reported a length of 6.8 ± 0.6 kb in senescent cells (Oja et al., 2018). In addition, a recent study described a mechanism of senescence that is independent of cell division and telomere length, involving activation of classical senescence-associated pathways and yielding a non-canonical SASP (Anderson et al., 2019). Although this phenomenon was first reported in post-mitotic cardiomyocytes, it may also occur in MSCs. Thus, telomere shortening has limitations for the measurement of senescence, and other markers may be more informative under certain conditions.

The senescent state is characterized by cell cycle arrest. Senescence-associated growth arrest is maintained by the activation of several pathways including phosphorylated inhibitor of cyclin-dependent kinase 4A (p16^INK4A)/phosphorylated retinoblastoma (pRb) and p53/p21^WAF1 signaling (Campisi, 2005). p16^INK4A is an inhibitor of cyclin-dependent kinase (CDK) and induces premature cell senescence via telomere-dependent and -independent mechanisms (Serrano et al., 1993). p16^INK4A level was shown to increase with chronological age or PDs of MSCs in culture, and a large proportion of the p16^INK4A-positive cells were negative for the proliferation marker Ki67 and positive for SA-β-gal. Inhibiting p16^INK4A reduced the number of senescent MSCs and conferred cells with the ability to proliferate (Shibata et al., 2007). Similarly, overexpressing p21^WAF1—a CDK inhibitor that acts by dephosphorylating pRb–increases cellular senescence, as evidenced by elevated SA-β-gal activity and telomere shortening (Huang et al., 2004), while inhibiting p21^WAF1 in senescent cells restored their replicative capacity. However, p21^WAF1 depletion was less efficient at preventing senescence than p53 depletion, suggesting that the latter acts through p21^WAF1-independent mechanisms to exert this effect (Gire and Dulic, 2015).

Mesenchymal stem cells-derived microvesicles (MSC-MVs) that mimic the senescent state of the parent MSC have recently emerged as a potential cell-free biomarker for cellular senescence. Senescent late-passage MSCs secrete larger amounts of MSC-MVs of smaller size than those in early passages, and CD105 expression in MSC-MVs decreased with senescence in parent MSCs. RNA sequencing results suggest that most genes that are highly expressed in senescent MSC-MVs are involved in aging-related diseases (Lei et al., 2017). Functionally, senescent MSC-MVs have a lower capacity to promote osteogenesis (Lei et al., 2017) and recruit macrophages and fail to alter macrophage phenotypes (Huang et al., 2019).

Besides the abovementioned markers, new tools and approaches have been proposed to monitor MSC aging such as SiR-actin, a fluorogenic F-actin specific probe that can be used to evaluate actin turnover (Mishra et al., 2019). CyBC9 (another fluorescent probe) combined with high-throughput screening revealed accumulation of mitochondria in senescent MSCs that presumably resulted from the loss of membrane potential (Ang et al., 2019). Thus, senescence can be characterized not by a universal biomarker, but by a set of non-exclusive markers in conjunction with specific biological features.

Senescence is a multistep process involving various mechanisms that have not been fully elucidated. One of these is irreversible cell cycle arrest—typically in response to DNA damage—in the presence of growth-promoting stimuli. DNA damage accumulates throughout the lifetime of an organism as a result of DNA replication errors and exposure to endogenous and exogenous mutagens. The DNA damage response (DDR) network can sense and initiate repair of mutations and thereby slow their accumulation. Activation of the DDR network can transiently halt cell cycle progression through stabilization of p53 and transcriptional activation of the CDK inhibitor p21. However, if DNA damage persists, p16^INK4A is activated via p38-mitogen-activated protein kinase-mediated mitochondrial dysfunction and ROS production. This results in CDK inhibition and activation of the tumor suppressor Rb1, which induces the onset of senescence (Figure 4; Zhang et al., 2007; Ciccia and Elledge, 2010).

Reactive oxidative species are a group of oxygen-containing small molecules. Approximately 90% of ROS are generated endogenously by the mitochondrial electron transport machinery (Poyton et al., 2009) but extrinsic factors such as radiation, ultraviolet light, hypoxia, and low temperature can also increase their production. As a metabolic by-product, ROS induce oxidation and various cellular responses through the generation of reactive secondary metabolites. While physiologic levels of ROS are necessary for proliferation and differentiation, an excess can trigger cellular senescence, including in MSCs (Sart et al., 2015). ROS induce DNA damage and accelerates telomere erosion, both of which activate the DDR. Senescent MSCs have elevated levels of ROS compared to normal cells that cause persistent DDR activation, thereby forming a positive feedback loop in the progression of senescence (Figure 4; Yang et al., 2015).

Reactive oxidative species accumulate with advancing age, leading to decreased mitochondrial metabolism. Mitochondrial DNA (mtDNA) encodes 13 polypeptides of the mitochondrial oxidative phosphorylation (OXPHOS) enzyme complexes, 22 tRNAs for mitochondrial biosynthesis, and 16S and 18S rRNA for mitochondria. Because of the absence of efficient repair mechanisms, the mutation rate of mtDNA is higher than that of nuclear DNA. A point mutation in mitochondria-encoded subunit 3 of cytochrome c oxidase was found to be associated with enhanced tissue degeneration and risk of premature aging (Niemann et al., 2017). The mitochondrial free radical theory of aging posits that age-dependent accumulation of mtDNA abnormalities, impaired OXPHOS, and altered expression of antioxidant enzymes lead to increased ROS production, which, in turn, results in progressive mitochondrial dysfunction and global cellular damage in a positive feedback loop (Figure 4; Wang C. H. et al., 2013). However, other studies suggest that mtDNA mutations and mitochondrial dysfunction affect aging independently of ROS production, based on the unexpected observation that genetic manipulations that increased mitochondrial ROS levels and oxidative damage did not accelerate aging in mice (Zhang et al., 2009), whereas those that impaired mitochondrial function without increasing ROS enhanced aging (Kujoth et al., 2005).

Autophagy is a critical process for maintaining cellular homeostasis under physiologic and pathologic conditions. By removing damaged cellular components including proteins and mitochondria, autophagy prevents age-related cellular injury (Wirawan et al., 2012) and allows stem cells to avoid the transformation from a reversible quiescent (G0) state to irreversible senescence (Garcia-Prat et al., 2016). Although ROS are known to stimulate autophagy, the age-related increase in ROS levels reduces autophagic capacity (Yamamoto et al., 2016). In fact, aged MSCs exhibit reduced autophagy, which is correlated with diminished self-renewal capacity and regenerative potential and replicative exhaustion. From a mechanistic standpoint, decreased autophagy results in the loss of proteostasis and increases mitochondrial activity, oxidative stress, and metabolism state in MSCs (Figure 4; Revuelta and Matheu, 2017).

Key aspects of the mechanisms of MSC aging remain unknown, but epigenetic changes (i.e., those occurring in the absence of DNA sequence alterations) such as DNA methylation, histone modification, and chromatin remodeling may play a role (Ozkul and Galderisi, 2016; Cakouros and Gronthos, 2019). For example, dysregulated expression of Brahma-related gene 1 (BRG1), a component of ATP-dependent chromatin remodeling complexes, was associated with senescence in MSCs via regulation of NANOG methylation status (Squillaro et al., 2015). Additionally, microenvironmental and hormonal conditions are important factors contributing to MSC aging in vivo. As MSCs exist in a semi-static state, replicative exhaustion is unlikely to occur (Ganguly et al., 2017). There is increasing evidence that the in vivo cellular aging process is caused by chronologic aging of the host and is accelerated by conditions such as obesity and systemic inflammation (Frasca et al., 2017; Franceschi et al., 2018).

Strategies allowing the generation of large numbers of MSCs that have retained their stemness are needed for clinical applications. Here, we summarize current research efforts to prevent MSC senescence.

Induced pluripotent stem cell (iPSC)-derived MSCs (iMSCs) can be passaged more than 40 times without exhibiting features of senescence (Sabapathy and Kumar, 2016). iMSCs retain a donor-specific DNA methylation profile while tissue-specific, senescence-associated, and age-related patterns are erased during reprogramming (Frobel et al., 2014). Recent studies have demonstrated that iMSCs have superior regenerative capacity compared to tissue-derived MSCs in preclinical degenerative disease models (Lian et al., 2010; Chen et al., 2019; Wang et al., 2019). However, the generation of iMSCs from iPSCs requires a significant degree of molecular manipulation, and there are safety concerns regarding the self-renewal and pluripotency of iPSC-derived cells after in vivo transplantation, which have the risk of tumorigenicity and genomic instability (Hynes et al., 2013). In addition, each independent iPSC line has a unique genetic and epigenetic profile that must be characterized. The concept of iMSCs is at its infancy and requires validation from preclinical and clinical studies before it can be clinically useful.

Aging is not a passive or random process but can be modulated through several key signaling molecules/pathways (Kenyon, 2010). Identification of age-related coordinating centers can provide novel targets for therapeutic interventions. Sirtuins (SIRTs) are a class of highly conserved nicotinamide adenine dinucleotide-dependent protein deacylases of which there are 7 (SIRT1–7) in mammals (Vassilopoulos et al., 2011). The role of SIRTs in aging is related to their regulation of energy metabolism, cell death, and circadian rhythm and maintenance of cellular and mitochondrial protein homeostasis (O’Callaghan and Vassilopoulos, 2017). Mitochondrial SIRTs (SIRT3–5) act as stress sensors and regulate protein networks to coordinate the stress response (van de Ven et al., 2017). Overexpression of SIRTs has been investigated as a potential strategy for preventing MSC aging. For instance, SIRT3 expression in MSCs decreased with prolonged culture and its overexpression in later-passage cells restored differentiation capacity and reduced aging-related senescence (Denu, 2017). SIRT1 is required for long-term growth of MSCs and SIRT1 overexpression was shown to delay senescence without loss of adipogenic or osteogenic potential (Yuan et al., 2012). Additionally, SIRT1 expression was shown to be spontaneously upregulated upon osteogenic differentiation and protected MSCs from extracellular oxidative stress (Li et al., 2018).

Genetic engineering has been used to slow MSC aging. Besides SIRTs, several molecules have been identified as potential targets for interventions to prevent senescence (Table 3). Ectopic expression of telomerase reverse transcriptase in MSCs extended their replicative lifespan, which preserved a normal karyotype, promoted telomere elongation, and abolished senescence without loss of differentiation potential (Simonsen et al., 2002). Introduction of Erb-B2 receptor tyrosine kinase 4 (ERBB4) in aged MSCs conferred resistance to oxidative stress-induced cell death and rescued the senescence phenotype (Liang et al., 2019). Knocking down macrophage migration inhibitory factor (MIF) in young MSCs induced senescence; conversely, its overexpression in aged MSCs rejuvenated the cells by activating autophagy (Zhang et al., 2019). However, the risk of malignant transformation remains a major barrier for the use of genetics-based approaches in clinical practice.

Pharmacologic approaches and cytokine supplementation have also shown promise for delaying senescence (Table 3). Inhibiting mechanistic target of rapamycin by rapamycin treatment enhanced autophagy and myogenic differentiation in aged stem cells (Takayama et al., 2017), while pretreatment with MIF rejuvenated MSCs in a state of age-induced senescence by interacting with CD74 and thereby activating 5′AMP-activated protein kinase-Forkhead box O3a signaling (Xia et al., 2015). Pharmacologic antagonism of lysophosphatidic acid, a ubiquitous metabolite in membrane phospholipid synthesis, extended the lifespan of MSCs in culture and increased their clonogenic potential while preserving their capacity for both osteogenic and adipogenic differentiation (Kanehira et al., 2012). However, as the effects of these approaches can vary–for instance, fibroblast growth factor supplementation resulted in the loss of osteogenic/adipocytic differentiation potential in long-term cultures (Gharibi and Hughes, 2012)–further research is needed to evaluate their long-term safety and efficacy.