Somatic Mutations and Aging

There are about 10 trillion cells in our body, and each suffers approximately 70,000 DNA lesions every day (Lindahl & Barnes, 2000). Our cells are under constant exposure to external insults, including ionizing radiation (e.g., UV, cosmic rays) and environmental toxins, and internal insults, including free radicals and replication errors. A sophisticated set of repair mechanisms has evolved to repair different types of DNA lesions (Kunkel, 2015). However, even in the presence of an intact repair system, DNA mutations still accumulate with age. In human cortex, 8‑oxoguanine, a damage caused by reactive oxygen species (ROS), increases with age (Lu et al., 2004). H2AX, which is a histone subtype that gets phosphorylated and recruited to the site of double strand breaks (DSB) (Paull et al., 2000), increases with age in humans (Scaffidi & Misteli, 2006), mice (Sedelnikova et al., 2004), and monkeys (Herbig, Ferreira, Condel, Carey, & Sedivy, 2006). The phosphorylated H2Av (a homologue of H2AX in flies) foci persist in germ cells of old compared to young flies after the induction of DSB (Delabaere et al., 2017), which implies that DNA damage repair slows down with age in flies. Also, p53 binding protein (53BP1), another marker of DSB, accumulates with age in the skin of living baboons (Herbig et al., 2006). The persistent activation of repairing proteins may imply that the increased level of DNA damage exceeds the capacity of the DNA repair system or that DNA repair efficiency declines with age. In either case, evidence holds that genomic instability increases with age, but its role as a driver of aging, as likely as it seems, is yet to be experimentally scrutinized.

Progeria, a syndrome described by accelerated aging or premature aging, is driven by genomic instability; the Hutchinson-Gilford Progeria syndrome (HGPS) and Werner syndrome are the best studied. HGPS is known to be caused by de novo mutations in the LMNA gene, which encode for A-type lamins (lamin A/C) (Eriksson et al., 2003). Such mutations alter nuclear morphology and function, and cause genomic instability (Gonzalo & Kreienkamp, 2015). HGPS patients often die in their teens (Hennekam, 2006), but their cells exhibit age-related dysfunction, including genomic instability (Liu et al., 2005), senescence (Liu et al., 2005), altered gene expression (Prokocimer, Barkan, & Grunebaum, 2013), and nuclear envelope irregularities (Goldman et al., 2004). Notably, the de novo mutation found in HGPS syndrome patients also accumulates with age in normal aging human skin from different sites of the body (McClintock et al., 2007). Werner syndrome is caused by mutations in the WRN gene, which encodes for a protein known to act as a helicase and exonuclease, which are key processes in DNA replication and repair (Gray et al., 1997). Further, recent evidence suggests that the WRN protein is important in maintaining telomeres (described below in the “Telomeres” subsection) (Crabbe, Verdun, Haggblom, & Karlseder, 2004; Edwards, Machwe, Chen, Bohr, & Orren, 2015; Sun et al., 2017). Although it is debatable whether progeroid syndromes are an accelerated version of normal aging, there is much to be learned about the molecular mechanisms of aging from these unusual conditions (Burtner & Kennedy, 2010).

Epigenetics

Epigenetics, in a broad sense, refers to chromosomal modifications that regulate genes without altering the underlying DNA. Waddington coined the term “epigenetics” to describe changes in phenotype without changes in genotype (Waddington, 1942a, 1942b). Epigenetic regulation can largely happen at the DNA, protein (histone), and/or chromatin level, which are subject to changes during aging (Benayoun, Pollina, & Brunet, 2015). For instance, aging can lead to chromatin remodeling, histone loss, histone variant incorporation, histone modification alterations, and DNA methylation pattern shifts (Pal & Tyler, 2016; Sen, Shah, Nativio, & Berger, 2016). Here we discuss the role of epigenetics in genomic instability.

In eukaryotes, DNA methylation is a major epigenetic modification that occurs on cytosines, especially in clusters known as “CpG islands,” which are GC-rich regions with high densities of CpGs that typically occur at or near the transcription start site of genes. Imbalanced DNA methylation may dysregulate genes involved in DNA repair and chromosomal stability, and hence contribute to genomic instability (Dodge et al., 2005; Karpf & Matsui, 2005; Meng et al., 2015). Recently, studies on age-dependent DNA methylation changes in multiple tissues provide insight into a more accurate prediction of “biological age” as compared to “chronological age” (Horvath & Raj, 2018), which may be a more accurate predictor of mortality (Levine, 2012). These studies stem from a collective interest in identifying reliable biomarkers of aging (Warner, 2004) and the fact that DNA methylation patterns change with age (Bell et al., 2012; Christensen et al., 2009; Hernandez et al., 2011; Maegawa et al., 2010; Rakyan et al., 2010). Currently, there are three major “clocks” that provide comparable and complementary estimates of biological age (Horvath & Raj, 2018) that are based on DNA methylation data derived from (a) blood-derived DNA (71 CpGs) (Hannum et al., 2013); (b) multiple tissues types (353 CpGs), including 51 healthy tissues and cell types (Horvath, 2013); and (c) clinical data regressed on blood-derived DNA (513 CpGs) (Levine et al., 2018). The connection between the epigenetic clock and genomic instability has been speculated, but further research is needed to provide experimental evidence (Horvath & Raj, 2018).

The DNA in the nucleus wraps around the core histone proteins, forming a nucleosome that in turn compacts into a higher-level structure known as chromatin (Luger, Mäder, Richmond, Sargent, & Richmond, 1997). Such complex nucleoprotein structures confer not only high-level organization to contain a large amount of data that can be dynamically and rapidly accessed as needed, but also protection of our DNA code from damaging insults. Histone proteins (H2A, H2B, H3, and H4) possess a lysine-rich tail that projects from the nucleosome, and are subject to several post-translational modifications (PTMs) that regulate genetic processes, including chromatin compaction/dynamics and transcription. Well-studied histone PTMs include acetylation, methylation, phosphorylation, and ubiquitylation, which work together under a so-called “histone code” (Jenuwein & Allis, 2001). Although it is not clear whether histone methylation directly regulates aging, much correlative evidence of its involvement in aging exists. Methylation of lysine residues of histone 3 at positions 4 and 36 (H3K4 and H3K36) are “active” marks that are enriched in euchromatin, whereas H3K9 and H3K27 methylation are “repressive” marks that are enriched in heterochromatin. A wide redistribution of these methylated histone markers has been reported in senescent human cells and fibroblasts from HGPS patients (Shah et al., 2013). Histone acetylation can directly regulate DNA-histone interaction by neutralizing the positively charged lysine residues of the histone N termini, decreasing their affinity for DNA (Grunstein, 1997). Histone acetyltransferases (HATs) and deacetylases (HDACs) dynamically regulate the acetyl status of histones and, in turn, chromatin architecture. Notably, sirtuins are HDACs that are also key regulators of aging and age-related diseases (Longo & Kennedy, 2006). In yeast, the histone deacetylase silent information regulator 2 (Sir2), which is the founding member of the sirtuin family, silences transcription and stabilizes repetitive DNA (rDNA), but relocalizes to sites of genomic instability during aging or DNA damage; Sir2 overexpression extended yeast replicative life span, whereas its loss shortened it (Sinclair & Guarente, 1997). SIRT1, the mammalian homolog of Sir2 (Oberdoerffer et al., 2008; Wang et al., 2008), and SIRT6 (Mao et al., 2011; Mostoslavsky et al., 2006) are required for genomic stability.

Transposable Elements

Transposable elements (TEs) are genetic units that can jump around the genome, rearrange the chromosome by homologous recombination, and control gene expression (Rebollo, Romanish, & Mager, 2012; Slotkin & Martienssen, 2007). There are two major types of TEs: (a) retrotransposons that transcribe from DNA and then reverse transcribe into a new location, using their inherently encoded reverse transcriptase (copy-and-paste mechanism), and (b) DNA transposons that jump out of the genome and self-propagate using their inherently encoded transposase (cut-and-paste mechanism) (Huang, Burns, & Boeke, 2012). The large majority of TEs in humans consist of retrotransposons (Huang et al., 2012). TEs make up nearly 50% of our genome (de Koning, Wanjun, Todd, Batzer, & Pollock, 2011; Mills, Bennett, Iskow, & Devine, 2007), and although previously considered by some to be “junk” or “selfish” DNA fragments, they can contribute to the evolution of our genome (Biémont & Vieira, 2006). However, their “jumpy” nature can also have an obvious negative impact on our genome, including gene inactivation and genomic instability. Therefore, our cells have developed a highly regulated system to keep TE mobilization well under control. TEs are mostly found in heterochromatin, the suppressed areas of the genome, and are thus kept inert under healthy conditions (Riddle et al., 2011). In fact, although TE sequences can take up as much as 50% of our genome, only a fraction (<0.05%) remain active (Mills et al., 2007).

Notably, recent studies have shown that TE activity can increase with age, and contribute to genomic instability, in various model organisms. In yeasts, retrotransposon mobility is increased and associated with chromosome rearrangement in old cells (Maxwell, Burhans, & Curcio, 2011). In worms, capsids, products of a retrotransposon, accumulate in the gonads with age at 15°C (Dennis, Sheth, Feldman, English, & Priess, 2012). In flies, many TEs become activated with age, which can be blocked by stabilizing heterochromatin and prevented by dietary restriction (DR) (Orr, 2016; Wood et al., 2016). Also, in fly brains, several TEs showed increased activity with age, and the inactivation of a gene that is involved in silencing TEs further exacerbated brain TE activity and caused progressive age-dependent memory decline and, ultimately, shortened lifespan (Li et al., 2013).

Telomeres

Human telomeres are short DNA repeats that “cap” the ends of linear chromosomes to provide genomic stability (Blackburn & Gall, 1978) as a solution to their inherent end-protection and replication problems (de Lange, 2009; Olovnikov, 1996). Telomeres were first reported by Barbara McClintock (McClintock, 1941) and by Hermann Muller (Muller, 1938) as unique chromosomal ends that differ from those of broken lesions. Telomeric ends consist of repeated DNA sequences (TTAGGG) with an overhang at the lagging strand. This resembles a double-stranded DNA break lesion, and to avoid DNA repair mechanisms acting on them, telomeric ends assume a protective loop formation, mediated by telomeric repeat-binding factor 2 (TRF2) (Griffith et al., 1999). General DNA polymerases are incapable of completely replicating the telomeric ends and require special DNA polymerases known as telomerases. Notably, telomerases are generally not expressed in somatic cells. This leads to a progressive and cumulative loss of telomere length with each cell division. In fact, Alexey Olovnikov proposed in 1971 that the limited division potential of a cell (known as the Hayflick limit; see also the “Cellular Senescence and Aging” section) may be driven by the progressive shortening of the telomeric length with every replication (Olovnikov, 1996).

However, telomere attrition with the uncapping of linear chromosomal ends per se does not appear to be the driver of aging. For instance, a comparative analysis of over 60 mammalian species shows that there is an inverse correlation between telomere length and lifespan (Gomes, et al., 2011). Also, laboratory mice have longer telomeres compared to human telomeres (> 30 Kb vs. 10–15 Kb, respectively) (Campisi, Kim, Lim, & Rubio, 2001), yet live substantially less time than humans. In addition, telomerase activity has been suggested to have co-evolved with body mass rather than lifespan in 15 different rodent species (Seluanov et al., 2007).

Another interesting paradigm suggests that replicative senescence is driven by persistent DNA damage responses (DDR) at critically short telomeres that exhibit irreparable DNA damage (d’Adda di Fagagna et al., 2003; Fumagalli et al., 2012; Herbig, Jobling, Chen, Chen, & Sedivy, 2004). Indeed, the uncapped telomere recruits activated proteins involved in DSB repair at the telomeric site (Takai et al., 2003) and causes homologous recombination (HR)-dependent deletions (Wang et al., 2004). Notably, neither the DNA damage checkpoint nor DSB repair occur during mitosis, which appears to be an inherent mechanism to prevent sister telomere fusions that produce dicentric chromosomes and aneuploidy (Orthwein et al., 2014). This accords with Barbara McClintock’s earlier report that the chromatin ends are different from DSB in that HR was never observed (McClintock, 1941). Telomeric uncapping–induced recruitment of DSB repair proteins is also seen in human fibroblasts (d’Adda di Fagagna et al., 2003).

The African turquoise killifish, an emerging vertebrate model to study aging and age-related diseases (Kim, Nam, & Velenzano, 2016; Valdesalici & Cellerino, 2003), also show age-dependent shortening of telomeres, which are of comparable length to humans (5–7 Kb), in muscles and liver (Harel et al., 2015; Hartmann, Reichwald, & Lechel, 2009). Further, forced telomere attrition by mutating a component of telomerase (i.e., TERT) recapitulated key aspects of the human dyskeratosis congenita syndrome. TERT mutant killifish exhibited dysfunction in proliferative tissues, including the blood, intestine, and gonads, and showed precancerous changes in the gut (Harel et al., 2015). With the killifish genome now assembled and annotated (Valenzano et al., 2015), and powerful genetic tools made available (Harel, Valenzano, & Brunet, 2016), this shortest-lived laboratory model of vertebrate aging may provide deeper insight on genomic instability and aging. Telomeres have also been implicated in senescence, as discussed in the “Cellular Senescence and Aging” section.

mtDNA Instability

mtDNA is present in hundreds to thousands of copies per cell and replicated independent of the cell cycle (Bogenhagen & Clayton, 1977). Numerous studies have demonstrated that mitochondrial mutations increase with age in both animal models and in human tissues, including brain, heart, colon, and skeletal muscle (Bua et al., 2006; Cortopassi & Arnheim, 1990; Gardner, Payne, Hovarth, & Chinnery, 2015; Greaves et al., 2014; Kennedy, Salk, Schmitt, & Loeb, 2013; Larsson, 2010; Pikó, Hougham, & Bulpitt, 1988), although the levels and kind of mutations appear to differ between tissues and even within tissues (Soong, Hinton, Cortopassi, & Anheim, 1992). It has been documented that both point mutations and deletions of mtDNA accumulate with age (Ameur et al., 2011; Bua et al., 2006; Cortopassi & Arnheim, 1990; Gardner et al., 2015; Greaves et al., 2014; Kennedy et al., 2013; Larsson, 2010; Pikó et al., 1988; Vermulst et al., 2007). The major source of mtDNA point mutations with age has been shown to be replicative infidelity in rodents and humans, rather than oxidative stress as once considered by many (discussed in detail in the “Mitochondria, Free Radicals, and Aging” section) (Ameur et al., 2011; Vermulst et al., 2007). Further, some mutations can persist and cause heteroplasmy, whereby a mix of wild-type and mutant mtDNA co-exist in a cell (Kennedy et al., 2013; Stewart & Chinnery, 2015) or in a single mitochondrion (Morris et al., 2017). Even low-level heteroplasmy, which is common in humans (Payne et al., 2012), in the context of somatic mutations can contribute to aging and neurological defects (Ross et al., 2013).

The Mitochondria Free Radical Theory of Aging

Emerging Mitochondrial Roles in Aging

Increasing evidence that mitochondria act as signaling organelles favors the concept that mitochondrial communication may be involved in important cellular processes (Chandel, 2015; Quiros, Mottis, & Auwerx, 2016), and ultimately aging (Bratic & Larsson, 2013; Durieux, Wolff, & Dillin, 2011; Hill, Sataranatarajan, & Remmen, 2018; Karpac & Jasper, 2013; López-Otin, Galluzzi, Freije, Madeo, & Kroemer, 2016; Melber & Haynes, 2018; Shadel & Horvath, 2015; Sun, Youle, & Finkel, 2016). Eukaryotic cells are functionally compartmentalized into organelles that have specific roles. Mitochondria and the nucleus are unique in the sense that they contain genomes. To dynamically coordinate the complex cellular components, a highly sophisticated communication system must exist. Indeed, mitochondria and the nucleus are known to orchestrate a wide range of cellular functions, and the privileged communication between these two is widely known as “mitonuclear communication” (Quiros et al., 2016). Although the messages transmitted through mitonuclear communication is predicted to be broad, cellular homeostasis is a common theme in recent literature (Hill et al., 2018). For example, ROS signaling (Shadel & Horvath, 2015), the mitochondrial unfolded protein response (UPRmt) (Nargund et al., 2012; Qureshi, Haynes, & Pellegrino, 2017), and mitochondrial damage-associated molecular patterns (mtDAMPs) that signal mitochondrial damage (Galluzzi, Kepp, & Kroemer, 2012; Wilkins, Weidling, Ji, & Swerdlow, 2017) all strive to bring balance back to the cell. Notably, mitochondrial signals can also regulate distal tissues in worms (Durieux et al., 2011; Woo & Shadel, 2011).

Until the early 2000s, mitochondrial retrograde signaling (i.e., signals transmitted from mitochondria) was entirely considered to be mediated by nuclear-encoded factors, secondary metabolites, or transient molecules (Lee, Yen, & Cohen, 2013; Quiros et al., 2016). However, identification of small genes in the mtDNA, collectively referred to as mitochondrial-derived peptides (MDPs), has provided another layer of mitochondrial communication (Kim, Son, Benayoun, & Lee, 2017; Lee et al., 2013). In fact, the past several years have witnessed increasing discoveries of very small peptides in both the nuclear and mitochondrial genomes (Kim et al., 2017; Makarewich & Olson, 2017). In the mtDNA, there are now eight peptides, including humanin (Guo et al., 2003; Hashimoto et al., 2001; Ikonen et al., 2003), MOTS-c (Lee et al., 2015), and SHLP1-6 (Cobb et al., 2016). Notably, metabolic stress can trigger MOTS-c to directly regulate the nuclear genome to enhance cellular resilience, indicating that the mitochondrial and nuclear genomes are fundamentally integrated at the genetic level (Kim, Son, Benayoun, & Lee, 2018). In fact, the compatibility between the mitochondrial and nuclear genomes has been shown to be important. Mice engineered to host mitochondria from a different genetic background (i.e., forced mismatch of mitochondrial and nuclear genomes) have altered cellular bioenergetics, mitochondrial ROS generation, and resistance to cardiac stress and atherogenic diet (Betancourt et al., 2014; Dunham-Snary & Ballinger, 2015; Fetterman et al., 2013). Although still on hold in the United States, the United Kingdom has approved mitochondrial replacement therapy, which would allow the genetic adoption of mtDNA from another woman (also referred to as “three-parent babies”), without understating the consequences of incompatible mtDNA-nDNA combinations and dysregulated mitonuclear communication (Hamilton et al., 2015). On that line, aging may be underwritten by the decline in communication robustness and efficiency, which reduces the cells’ ability to dynamically respond and adapt to the environment, and ultimately cellular fitness.

Lastly, our understanding of the evolutionary roots of mitochondria in aging is underdeveloped. This is especially interesting in the context of genomic co-evolution of the mitochondrial and nuclear genomes that have establish a tightly coordinated bi-genomic system—how would the force of natural selection apply under these conditions?

The Somatotropic Axis and Aging

Growth hormone is a key regulator of somatic growth and metabolism either directly or indirectly via effectors such as IGF-1 (Vijayakumar, Yakar, & LeRoith, 2011). GH blocks insulin action, stimulates lipolysis, and inhibits lipogenesis, whereas IGF-1 has opposite effects on glucose and lipid metabolism (Vijayakumar et al., 2011). GH is produced and secreted by somatotropic cells in the anterior pituitary under the regulation of hypothalamic neuropeptides (i.e., growth hormone–releasing hormone and somatostatin). GH deficiency at the level of production or recognition by GH receptors (GHR) of its target tissues causes dwarfism but confers remarkable lifespan extension in mice (Bartke, Sun, & Longo, 2013). Ames, Snell, and Little mice are GH-deficient mutants that exhibit dwarfism and lifespan extension (> 50%, > 40%, and > 20%, respectively) (Brown-Borg, Borg, Meliska, & Bartke, 1996; Flurkey,Papaconstantinou, Miller, & Harrison, 2001). Also, mice that are GH insensitive because of mutations in GHR (GHRKO) are dwarfs and live approximately 50% longer than their wild-type counterparts (Coschigano, Clemmons, Belush, & Kopchick, 2000; Zhou et al., 1997). On the contrary, transgenic mice that overexpress GH show a 50% reduction in lifespan (Bartke, 2003) and increased cancer incidence, and exhibit various age-related dysfunctions in the kidney and liver (Wolf et al., 1993). The GHRKO mice model the Laron syndrome in humans, which is also described by GH insensitivity due to GHR mutations, leading to dwarfism (Guevara-Aguirre et al., 2011; Guevara-Aguirre et al., 2015). Notably, an Ecuadorian cohort with Laron syndrome was found to have only one nonlethal malignancy and no cases of diabetes, in contrast to a prevalence of 17% for cancer and 5% for diabetes in control subjects (Guevara-Aguirre et al., 2011). Further, those with Laron syndrome had greater regional cortical surface area, enhanced cognitive performance, and greater task-related activation in frontal, parietal, and hippocampal regions compared to their unaffected counterparts (Nashiro, Guevara-Aguirre, Braskie, Hafzalla, &Velasco, 2017). Unlike GHR disruption, homozygous inactivation of IGF-1 receptor (IGF-1R) is neonatal lethal in mice; homozygous IGF-1 knockout had a variable effect on lifespan (J.-P. Liu, Baker, Perkins, Robertson, & Efstratiadis, 1993). However, heterozygous knockout of IGF-1R (Igf1r^+/–) from birth showed a 26% increase in mean lifespan in mice; female Igf1r^+/– mice lived 33% longer, whereas male Igf1r^+/– mice only had a 16% increase (not statistically significant) (Holzenberger et al., 2003). Interestingly, Igf1r^+/– mice were normal in their size, metabolism, physical capacity, and fertility, but were more resilient to oxidative stress, a hallmark of longevity (Holzenberger et al., 2003). Recently, late-life inactivation of IGF-1R has also been shown effective in extending health/lifespan of female mice (K. Mao et al., 2018). In humans, overrepresented heterozygous mutations in the IGF1R gene in centenarians revealed a functional variant that reduces IGF-1R activity, providing a correlation between IGF-1 signaling and longevity (Suh et al., 2008).

Interestingly, the GH/IGF-1 axis is also linked to mitochondria via MDPs. Circulating humanin levels were higher in those with Laron syndrome, mouse models with deficiencies in GH/IGF-1, and after GH injection in children; humanin levels seems to be regulated at the level of IGF-1 (C. Lee et al., 2014).

Environmental Factors

We are continually exposed to toxic agents in the air we breathe. In a recent review from the World Health Organization (WHO), more than 80% of the population in the WHO European Region lives in cities with particulate matter (PM) levels that are higher than the WHO Air Quality Guidelines (2013). Exposure to airborne combustion engine–derived PM has been associated with cardiovascular and respiratory dysfunctions (Künzli et al., 2010). Exposure to PM was also positively correlated with increased inflammation (e.g., C-reactive protein [CRP]) and oxidative stress in humans (Chuang, Chan, Su, Lee, & Tang, 2007; Hajat et al., 2015; Ostro et al., 2014). PM smaller than 2.5 μ‎m in aerodynamic diameter (PM2.5) are of special concern and have been linked to many diseases, including cardiovascular and respiratory diseases, diabetes, and neurodevelopment and cognitive function, as well as increased mortality (Jerrett et al., 2013). Interestingly, mice exposed to nanoscale (< 200 nm) PM (nPM) derived from urban freeway air (Los Angeles, CA) showed elevated cytokines levels (IL-1α‎ and TNFα‎) in the cerebral cortex (Morgan et al., 2011).

Microbiome

The human body is colonized by as many bacterial cells as its own (near 1:1 ratio) (Sender, Fuchs, & Milo, 2016). Most commensal bacteria reside in the colon, and thus much interest has been placed on the gut microbiome. The gut microbiota composition of the elderly differs from that of younger individuals (Claesson et al., 2011), and loss of core microbiota diversity is associated with increased frailty (O’Toole & Jeffery, 2015). Our immune system is deeply connected to the microbiome; the microbiome-host interactions can regulate inflammatory cytokine production capacity (~10% in cytokine variability) (Schirmer et al., 2016). Further, intestinal permeability may increase with age, leading to a “leaky gut.” In fact, Elie Metchnikoff, who coined the term “gerontology,” suggested that “leaky gut” results in displaced bacteria, and that in turn causes chronic systemic inflammation (Metchnikoff, 2004). In a recent study, ileal biopsies of young and old individuals showed increased inflammatory cytokine expression (IL‑6), reduced barrier integrity, and reduced inflammatory response to bacterial challenges (IL‑6 and -8) (Man et al., 2015). Age-dependent loss of intestinal integrity is also observed in flies and is associated with chronic inflammation and death (Rera, Clark, & Walker, 2012). In mice, age-dependent microbial dysbiosis promoted loss of intestinal barrier integrity and systemic inflammation; mice under germ-free conditions were protected from age-dependent inflammation (Thevaranjan et al., 2017).

Drugs could also affect the gut microbiome. For instance, in addition to conventional antibiotics, other pro-longevity drugs, such as rapamycin (Bitto et al., 2016) and metformin (F.Cabreiro et al., 2013; Wu et al., 2017), also have antibacterial actions that may have mediated their effects on lifespan. Another study has shown that antibiotic-induced bacterial change in the guts of mice under a high-fat diet or control lowered the inflammation markers (Cani et al., 2008). Microbes have been shown to modulate the rate of aging in invertebrate model organisms, including worms and flies (Heintz & Mair, 2014). Further studies to understand the role of the gut microbiota in mediating pharmacological interventions, especially those administered orally, that extend lifespan and/or health span in model organisms may provide much insight.

Cellular Senescence

Cellular senescence has been traditionally described as a permanent cell cycle arrest. Cells were once thought to be immortal and capable of dividing indefinitely (Witkowski, 1985), indicating that organismal aging was a problem of a multicellular body as a whole. Hayflick and Moorehead challenged this notion with their publication that described the first demonstration of finite cellular replicative capacity (Hayflick & Moorhead, 1961). They showed that human fibroblast cultures could divide 40 to 50 times before entering a nondividing state, and speculated that they had a role in aging. Interestingly, cells from older individuals had reduced dividing capacity compared to cells from younger individuals, suggesting that the replicative limitations of a given cell were predetermined from birth (Hayflick, 1965).

Hayflick’s proposal has been referred to as replicative senescence, which is associated with critical shortening of telomeres, segments at the end of nuclear chromosomal DNA. Telomere attrition, which induces persistent DDR, is further discussed in the “Genomic Instability and Aging” section. Telomeric shortening can induce cellular senescence (Herbig et al., 2004), and increasing telomerase activity can immortalize human cells (Counter et al., 1992; Kiyono et al., 1998). Therefore, long telomeres may seem to be a preventative factor against aging by reducing replicative cellular senescence. However, murine cells senesce regardless of their long telomeres (Itahana & Dimri, 2004), and an inverse relationship is observed in the wild between telomeric length and lifespan (Gomes et al., 2011). Therefore, multiple mechanisms may be involved in determining the relationship between telomere length and senescence.

Senescent cells in vitro can be characterized by enlarged and flat morphology (Lee et al., 2006), growth arrest, resistance to apoptosis, and altered gene expression (Campisi & d’Adda di Fagagna, 2007). There are biomarkers to detect cellular senescence, including senescence-associated beta-galactosidase (B. Y. Lee et al., 2006), loss of lamin B1 (Freund, Laberge, Demaria, & Campisi, 2012), and the complex SASP secretome (Coppé et al., 2008), which are best used in combination. However, senescence is much more difficult to identify in vivo, largely because of technical limitations, and thus poorly understood in living organisms (Childs, Durik, Baker, & Deursen, 2015). In addition, tissue specificity may further complicate senescence assessment in vivo, based on the fact that SASP gene expression at the transcriptional level varies by tissue type in both young and old mice (Hudgins et al., 2018).

Depending on the driver and the mechanism of activation, senescent cells can be classified into either acute or chronic senescent cells (Sturmlechner, Durik, Sieben, Baker, & van Deursen, 2017). Acute senescent cells arise from a defined set of regulated processes and are then cleared upon completing their physiological duties during a transient existence (e.g., embryonic development [Muñoz-Espín et al., 2013] and wound healing [Demaria et al., 2014]) (Muñoz-Espín & Serrano, 2014). However, chronic senescent cells are driven by diverse stressors over time and are poorly cleared, leading to their accumulation and continuous SASP production, ultimately accompanying pathological consequences. DNA damage, oxidative stress, suboptimal culture conditions, therapies, and other possible stressors, other than telomere erosion, can induce cellular senescence (Acosta et al., 2008; Acosta et al., 2013; D. J. Baker, Alimarah, van Deursen, Campisi, & Hildesheim, 2017; Parrinello et al., 2003). Such alternative mechanisms could partially explain how murine cells that possess long telomeres also undergo senescence (Parrinello et al., 2003). Further, senescence is a dynamic process that can be divided into three stages: early, full, and deep senescence, depending on the state of chromatin remodeling, SASP expression profiles, and phenotypic diversification, suggesting the possible reversion of early senescent cells back to their previous state (Deursen, 2014).

Senolytics

It is premature to conclude a causal connection between cellular senescence and aging (Jeyapalan & Sedivy, 2008). However, from an antagonistic pleiotropy perspective, cellular senescence can increase the fitness of young individuals by suppressing tumors and regulating embryonic development (Muñoz-Espín et al., 2013) and wound healing (Demaria et al., 2014). However, two major aspects of cellular senescence can directly reduce the fitness in late life: (a) a loss of tissue regeneration due to cell cycle arrest in progenitor cells (cell autonomous) and (b) the production of pro-inflammatory SASP (non-cell autonomous) that can promote various pathologies, including atherogenesis (Childs et al., 2016); kidney disease (Sturmlechner et al., 2017); and, paradoxically, cancer (Ana et al., 2001; Demaria et al., 2017; Jean-Philippe. Kauser, Campsi, & Beauséjour, 2006).

Therefore, if cellular senescence can promote tissue dysfunction during aging and may even directly drive age-related pathology, it may be therapeutically beneficial to eliminate it late in life. Clearance of senescent cells in a progeroid mouse model using a transgenic approach delayed several age-associated pathologies (Baker et al., 2011). Senolytics are drugs that can selectively eliminate senescent cells. The small-molecule senolytic ABT263 can selectively eliminate senescent bone marrow hematopoietic stem cells and senescent muscle stem cells in mice and are able to ameliorate stem cell function during aging (Chang et al., 2015). A recent study supports the concept that senescent cells can be contagious and cause physical impairment (M. Xu et al., 2018). In the study, senescent cells (pre-adipocytes) were transplanted into young mice, which then led to persistent physical dysfunction and triggered cellular senescence to other tissues; old mice that received senescent cell transplant exhibited an even stronger decline of physical capacity. Notably, intermittent administration of a senolytic cocktail, dasatinib plus quercetin, alleviated physical dysfunction and increased post-treatment survival by 36% while reducing mortality hazard to 65%.