The Key to immortality? What is it?

[n9newunsixx5150]

What is the key to immortality? please tell me...

Self-realization.

 

[Mi$ta-Murda187]

Thanks for the reply

 

[Y-S]

Aubrey De Grey

 

[MEXCOM]

Not dying. Any other questions you need answered?

 

[ThaG]

Mammalian sirtuins—emerging roles in physiology, aging, and calorie restriction


Sir2 is an NAD-dependent deacetylase that connects metabolism with longevity in yeast, worms and flies. Mammals contain seven homologs of yeast Sir2, SIRT1–7. Here, we review recent findings demonstrating the role of these mammalian sirtuins as regulators of physiology, calorie restriction, and aging. The current findings sharpen our understanding of sirtuins as potential pharmacological targets to treat the major diseases of aging.



Silent information regulator (Sir) proteins regulate lifespan in multiple model organisms. In yeast, an extra copy of the SIR2 gene extends replicative lifespan by 50%, while deleting Sir2 shortens lifespan (Kaeberlein et al. 1999Go). Sir2 silences chromatin, enables DNA repair, and is involved in chromosome fidelity during meiosis (Blander and Guarente 2004Go). Sir2 promotes longevity by suppressing the formation of toxic extrachromosomal rDNA circles (ERCs) in yeast (Sinclair and Guarente 1997Go). The Caenorhabditis elegans ortholog sir-2.1 also extends worm lifespan (Tissenbaum and Guarente 2001Go), but by a distinct mechanism. Sir-2.1 requires the worm forkhead protein DAF-16 for lifespan extension (Tissenbaum and Guarente 2001Go). While earlier models suggested sir-2.1 might function by down-regulating insulin signaling, more recent findings show that sir-2.1 binds to DAF-16, activating it directly (Berdichevsky et al. 2006Go). Moreover, sir-2.1 does not respond to changes in insulin signaling, but, rather, is activated by stress treatments, such as heat shock and oxidative damage (Berdichevsky et al. 2006Go). Likewise, an increase in the dosage of Drosophila Sir2 extends lifespan (Rogina and Helfand 2004Go).

Although a link between Sir2 and longevity was clear, its enzymatic activity remained elusive for years. An early clue came from the observation that CobB, an Escherichia coli homolog of Sir2, could catalyze the phosphoribosyltransferase reaction in cobalamin biosynthesis (Tsang and Escalante-Semerena 1998Go). Thus, it was predicted and demonstrated that Sir2 possessed NAD-dependent ADP-ribosyltransferase activity (Frye 1999Go; Tanny et al. 1999Go). Subsequent reports revealed that yeast and mammalian sirtuins catalyze a novel and robust reaction, NAD-dependent histone deacetylation, unavoidably linking Sir2 activity with metabolism (Imai et al. 2000Go; Landry et al. 2000Go; Smith et al. 2000Go). Mechanistically, ADP-ribosylation and deacetylation reactions by sirtuins are similar because they cleave NAD in each reaction cycle (Fig. 1) (Grubisha et al. 2005Go). During each cycle, deacetylation generates the novel metabolites, 2' and 3'-O-acetyl-ADP-ribose (Tanner et al. 2000Go; Sauve et al. 2001Go; Tanny and Moazed 2001Go), which may be important regulators of physiology (Grubisha et al. 2006Go). Today, more than a dozen nonhistone deacetylation substrates are known, several of which are described below.

Calorie restriction (CR) is a dietary regimen that extends the lifespan of every organism tested to date. Specifically, CR extends the lifespan of yeast (Lin et al. 2002Go), spiders (Austad 1989Go), flies (Loeb and Northrop 1917Go), fish (Comfort 1963Go), and rodents (McCay et al. 1935Go; Austad 1989Go). Sir2 is required for lifespan extension by CR in yeast, worms, and flies (Lin et al. 2000Go; Rogina and Helfand 2004Go; Wang and Tissenbaum 2006Go). In yeast, CR (0.5% glucose), was previously shown to increase mitochondrial function and to up-regulate SIR2 activity (Lin et al. 2002Go, 2004Go). However, in this case, the mitochondrial activation is SIR2-independent, suggesting that it lies upstream of SIR2. A more severe CR regimen (0.05% glucose) extends yeast replicative lifespan by a different mechanism that is apparently independent of both SIR2 and mitochondrial respiration (Kaeberlein et al. 2004Go, 2005Go). Finally, SIR2 has no effect on yeast survival under starvation conditions, and appears to actually reduce survival of certain exceptionally long-lived mutant strains (Fabrizio et al. 2005Go).

Do sirtuins regulate human longevity? Mammals have seven Sir2 homologs (sirtuins, SIRT1–7). These proteins have a highly conserved NAD-dependent sirtuin core domain, first identified in the founding yeast SIR2 protein, making them good candidates as lifespan regulators (Frye 2000Go). As highlighted in this review, mammalian sirtuins have diverse cellular locations, target multiple substrates, and affect a broad range of cellular functions (Table 1). In this review, we emphasize an emerging theme in the field of aging—the regulation of oxidative stress, DNA damage, and metabolism by mammalian sirtuins.


Three mammalian sirtuins (SIRT1, SIRT6, and SIRT7) are localized to the nucleus. SIRT1 is most extensively studied, has more than a dozen known substrates, and is a guardian against cellular oxidative stress and DNA damage. Moreover, SIRT1 plays a prominent role in metabolic tissues, such as pancreas, fat, and liver. SIRT6 and SIRT7 may also be important regulators of DNA damage and metabolism, respectively.


While a role for mammalian sirtuins in lifespan regulation has not been directly determined, evidence suggests that the Sir2 ortholog, SIRT1, may regulate many physiological processes known to be affected during aging and which are altered by CR (Fig. 2). SIRT1 deacetylates a large number of substrates, including p53, Ku70, NF-{kappa}B, and forkhead proteins to affect stress resistance in cells (Luo et al. 2001Go; Vaziri et al. 2001Go; Brunet et al. 2004Go; Cohen et al. 2004Go; Motta et al. 2004Go; Yeung et al. 2004Go), which may relate to the observed stress resistance conferred by CR. SIRT1 also regulates the activities of the nuclear receptor PPAR{gamma} and PGC-{alpha} (see below) to influence differentiation of muscle cells, adipogenesis, fat storage in white adipose tissue, and metabolism in the liver, suggesting a possible connection between this sirtuin and diets that promote leanness and longevity (Fulco et al. 2003Go; Picard et al. 2004Go; Rodgers et al. 2005Go). The observed induction of SIRT1 protein during CR is consistent with this idea (Cohen et al. 2004Go; Nisoli et al. 2005Go). These activities link SIRT1 to known physiological effects of CR, and suggest that this sirtuin may help mediate CR in mice.



SIRT1 regulation of insulin and glucose homeostasis

A critical component of the physiology of CR is increased insulin sensitivity and corresponding reductions in blood glucose and insulin levels (Barzilai et al. 1998Go; Dhahbi et al. 2001Go). Pancreatic ß-cells help to maintain glucose homeostasis by secreting insulin in response to glucose. Metabolism of glucose in these cells by glycolysis generates pyruvate, which enters mitochondria where it can be converted to CO2 by the TCA cycle. NADH made by this metabolic process drives electron transport and ATP synthesis. The increased ATP/ADP ratio causes closure of KATP channels and depolarizes the plasma membrane leading to an influx of Ca2+, which triggers fusion of secretory vesicles containing insulin to the cell membrane.

Two recent studies in mice have demonstrated that SIRT1 positively regulates glucose-stimulated insulin secretion in pancreatic ß-cells (Moynihan et al. 2005Go; Bordone et al. 2006Go). ß-Cell-specific SIRT1-overexpressing (BESTO) mice demonstrate increased insulin secretion in response to glucose (Moynihan et al. 2005Go). Conversely, SIRT1-/- mice or their isolated islets show blunted insulin secretion (Bordone et al. 2006Go). Both studies find that SIRT1 represses transcription of the mitochondrial uncoupling protein UCP-2 gene, which uncouples mitochondrial respiration from ATP production and reduces the proton gradient across the mitochondrial membrane. Thus, by blocking UCP-2 function, SIRT1 promotes more efficient energy generation. Indeed, BESTO islets demonstrate higher ATP levels (Moynihan et al. 2005Go), while islets from SIRT1 knockout (KO) mice do not elevate ATP production in response to glucose (Bordone et al. 2006Go). Interestingly, SIRT1-mediated repression of UCP-2 is alleviated by acute food deprivation (Bordone et al. 2006Go), which may further dampen ATP synthesis and the insulin responsiveness of ß-cells during starvation. While SIRT1 protein level is not affected by this condition, there is a decrease in the NAD/NADH ratio, which may reduce SIRT1 activity in pancreas (Bordone et al. 2006Go). The presence of UCP-2 in fasted animals may also ease the transition to metabolic activity after the next feeding and prevent hyperpolarization of the mitochondrial membrane and the corresponding production of reactive oxygen species. While these studies demonstrate that SIRT1 activity may be down-regulated in ß-cells during fasting, it is not known whether SIRT1 regulates insulin secretion during CR, or plays any role in pathologies demonstrating impaired insulin secretion.

Another study raises the possibility that SIRT1 promotes the survival of pancreatic ß-cells during oxidative stress (Kitamura et al. 2005Go). In stressed ß-cells, the forkhead protein FOXO1 moves into the nucleus and activates the ß-cell transcription factors, NeuroD and MafA, and provides stress resistance. As described above, SIRT1 binds to and regulates forkhead transcription factors both negatively and positively (Brunet et al. 2004Go; Motta et al. 2004Go). Kitamura et al. (2005)Go show that nuclear FOXO1 associates with SIRT1 in PML (promyelocytic leukemia) bodies in stressed cells. They suggest that SIRT1 deacetylates FOXO1 at that location to activate the protein and provide stress resistance (Accili and Arden 2004Go). This process bears similarity to sir-2.1 in C. elegans discussed above, which functions as a coactivator of the DAF-16 protein after it translocates to the nucleus in oxidatively or heat stressed worms (Berdichevsky et al. 2006Go).

Glucose homeostasis is maintained by the liver, in addition to pancreatic ß-cells, in response to changing nutrient conditions. During fasting, hepatocytes induce gluconeogenesis to supply other tissues with glucose. Several new studies have revealed that this nutrient response is under tight control of SIRT1 activity, providing another link between SIRT1 and metabolism. In cultured hepatocytes, SIRT1 interacts with and deacetylates nuclear FOXO1, promoting FOXO1-dependent transcription of hepatic gluconeogenic genes upon stress (Frescas et al. 2005Go). In the liver, the transcriptional coactivator PGC-1{alpha} also drives expression of the gluconeogenic pathway. SIRT1 deacetylates and activates PGC-1{alpha} to coordinate the increase in expression of gluconeogenic genes with the repression of glycolytic genes during fasting (Rodgers et al. 2005Go). However, in a neuronal cell line, overexpression of SIRT1 decreases the activity of PGC1-{alpha} and mitochondrial function (Nemoto et al. 2005Go), suggesting that the relationship between SIRT1 and PGC-1{alpha} may be complex.


SIRT1 and neuron function

SIRT1 has also been linked to the survival of neurons. It is interesting to note that CR protects against neurodegenerative pathology in mouse models for Alzheimer's (Zhu et al. 1999Go; Patel et al. 2005Go) and Parkinson's (Duan and Mattson 1999Go). SIRT1 can promote survival in cultured neuronal cells as an antiapoptotic factor, perhaps through down-regulating the proapoptotic factors, p53 (Luo et al. 2001Go; Vaziri et al. 2001Go) and FOXO (Brunet et al. 2004Go; Motta et al. 2004Go). Even more interestingly, SIRT1 may be involved in the axonal protection observed in the Wallerian strain of mice (Araki et al. 2004Go), which have a translocation that increases levels of the NAD biosynthetic enzyme nicotinamide mononucleotide adenylyl-transferase 1 and renders peripheral axons more stable after a neuronal insult. Indeed, NAD itself provides protection to axons in cultured dorsal root ganglia. One study shows that the effects of NAD and the Wallerian strain are dependent on SIRT1, leading to the conclusion that this sirtuin is neuroprotective (Araki et al. 2004Go). However, another study did not observe a difference in the response of dorsal root ganglion from SIRT1 KO mice (Wang et al. 2005Go). It is interesting to note that the time course of axonal degradation in the two studies is different, suggesting that there may be two different neuroprotective processes induced by NAD— one dependent on SIRT1 and one not. Future studies should resolve these apparent experimental discrepancies.

SIR2 orthologs can also protect against neuronal dysfunction due to polyglutamine toxicity in C. elegans and mammalian cells (Parker et al. 2005Go). One study shows that neurotoxicity in worms is spared by the age-1 mutation, which reduces insulin-like signaling (Morley et al. 2002Go), or in a transgenic strain overexpressing sir-2.1 (Parker et al. 2005Go). Like the effects of the age-1 or sir-2.1 transgenes in extending lifespan, neuroprotection requires the forkhead protein DAF-16 (Parker et al. 2005Go). As might be expected, polyglutamine toxicity is exacerbated in daf-16 or sir-2.1 mutants. The putative SIR2-activating polyphenol, resveratrol (Howitz et al. 2003Go), also protects against cell death in striatal neurons with the Huntingtons Disease allele htt (109Q) (Parker et al. 2005Go). Finally, ß-amyloid-induced death of microglia is spared by overexpression of SIRT1 or resveratrol treatment (J.Chen et al. 2005Go). In toto, the above studies raise the possibility that activation of SIRT1 may be a novel strategy to protect against neurodegenerative diseases. We note, however, that the protective effect of SIRT1 may not be universal in all cell types, since SIRT1-/- mouse embryonic fibroblasts (MEFs) actually survive better in culture and bypass senescence (Chua et al. 2005Go).


SIRT1 and CR

The above studies suggest that important physiological processes triggered by CR in mammals are regulated by SIRT1, making it vital to know whether mammalian sirtuins also regulate changes during CR. Recent experiments have directly related SIRT1 function to CR in mice. CR induces the endothelial nitric oxide synthase (eNOS), and results in an increase in mitochondrial biogenesis (Nisoli et al. 2005Go). Moreover, this mitochondrial induction by CR does not occur in eNOS deficient mice. Interestingly, the SIRT1 gene is activated by NO in vivo and in vitro (Nisoli et al. 2005Go), tracing a pathway in which CR induces NO production and activates mitochondrial biogenesis and SIRT1.

A more direct demonstration of the requirement of SIRT1 in CR involved placing SIRT1 KO mice on this diet (D.Chen et al. 2005Go). Although the KO mice show changes in blood glucose, triglycerides, and IGF-1 similar to wild-type controls, there is a large difference in one interesting output of CR. Wild-type mice show a fivefold to 10-fold increase in physical activity, which has been observed previously and may represent a foraging instinct induced by food insufficiency, but KO mice do not display any increase in activity. SIRT1 KO mice move as well or better than wild type when challenged by other means; i.e., rotarod or treadmill. This study indicates the first requirement for SIRT1 for at least one phenotype triggered by mammalian CR.


SIRT6 regulates DNA repair

SIRT6 is a nuclear protein widely expressed in mouse tissues (Liszt et al. 2005Go; Michishita et al. 2005Go). Original reports demonstrated that SIRT6 has a weak to absent in vitro deacetylate activity (North et al. 2003Go; Liszt et al. 2005Go). However, SIRT6 has also been shown to demonstrate a robust auto-ADP-ribosyltransferase activity (Liszt et al. 2005Go).

Recent work has provided insight into the diverse physiological functions of SIRT6 (Mostoslavsky et al. 2006Go). SIRT6 KO mice display premature aging symptoms, including loss of subcutaneous fat and decreased bone density, and die within 4 wk after birth. These phenotypes contrast with those of SIRT1 KO animals, which are postnatal lethal on an inbred strain (Cheng et al. 2003Go; McBurney et al. 2003Go). Outbred SIRT1 KO animals can survive into adulthood, but demonstrate a severe phenotype, including small size, delayed bone mineralization, defective skeletal closure, delayed eyelid opening, and sterility (Cheng et al. 2003Go; McBurney et al. 2003Go; Lemieux et al. 2005Go).

SIRT6 KO mice exhibit a deficiency in one specific form of DNA repair, the base excision repair (BER) (Mostoslavsky et al. 2006Go). MEFs lacking SIRT6 demonstrate impaired proliferation and enhanced sensitivity to DNA-damaging agents. SIRT6 KO MEFs demonstrate genomic instability in the form of chromosomal translocations, fragments, gaps, and detached centromeres. These defects can be rescued by overexpression of the DNA polymerase involved in BER, Polß. Furthermore, SIRT6 KO MEFs exhibit normal cell cycle checkpoints, end-joining, and double-strand break DNA repair. How SIRT6 regulates BER is still unknown. One might hypothesize that this sirtuin ADP-ribosylates a substrate protein involved in BER, which could be a component of the repair machinery or the chromatin at the site of DNA damage.

SIRT6 KO mice also display interesting metabolic phenotypes: low levels of circulating IGF-1 and hypoglycemia that becomes progressively more severe with age (Mostoslavsky et al. 2006Go). It will be interesting to know whether these metabolic changes are due to a direct role for SIRT6 in regulating IGF-1 and glucose homeostasis or are an indirect consequence of DNA damage that accumulates in these mutant mice.


SIRT7 promotes rRNA transcription

SIRT7 localizes to the nucleolus of human cells (Michishita et al. 2005Go; Ford et al. 2006Go). Interestingly, SIRT7 expression correlates with growth (Ford et al. 2006Go)—it is abundant in tissues with high proliferation, such as liver, spleen, and testes. By contrast, SIRT7 expression is absent or low in nonproliferating tissues, like heart, brain, and muscle.

Recent work has shown that SIRT7 may regulate cellular growth and metabolism (Ford et al. 2006Go). In the nucleolus, SIRT7 associates with rDNA and interacts with RNA polymerase I (Pol I). Overexpressing SIRT7 increases rRNA transcription and RNA inhibition of SIRT7 decreases transcription, showing that this sirtuin activates Pol I transcription (Ford et al. 2006Go).

An NAD-dependent deacetylase activity has not been observed for SIRT7 (North et al. 2003Go), but the amino acid residues that bind NAD in the conserved sirtuin core domain are required for SIRT7 activity (Ford et al. 2006Go), suggesting a role for NAD-dependent regulation. SIRT7 thus appears to regulate cell growth and metabolism in response to changing metabolic conditions by driving ribosome biogenesis in dividing cells. It is interesting that both SIRT7, as an activator of rRNA transcription, and SIRT1, as an inhibitor of p53 and FOXO, have features that are progrowth and prosurvival for cells.


Cytoplasmic sirtuins

To date, only SIRT2 is reported to be localized mainly in the cytoplasm (North et al. 2003Go; Michishita et al. 2005Go), while a fraction of SIRT2 is nuclear (North et al. 2003Go). Interestingly, SIRT1 is also reported to be a cytoplasmic protein in pancreatic {alpha}-cells (Imai et al. 2000Go). These findings lead us to speculate that mammalian sirtuins may shuttle between the nucleus and cytoplasm, depending on cell type or environmental stimuli.


SIRT2

Mammalian SIRT2 is a predominantly cytoplasmic protein (Dryden et al. 2003Go; North et al. 2003Go; Michishita et al. 2005Go), colocalizes with tubulin, and can deacetylate a number of substrates in vitro, including a-tubulin (North et al. 2003Go) and histones, although the physiological consequences of a-tubulin deacetylation by SIRT2 are not yet clear. The yeast ortholog of SIRT2, Hst2, can function in parallel to SIR2 in certain strains with respect to lifespan extension and rDNA silencing (Lamming et al. 2005Go). Therefore, it will be interesting to determine the lifespan of SIRT2-overexpressing mice, or to determine whether CR is partly mediated by SIRT2 using SIRT2 KO animals.

Cell culture studies demonstrate SIRT2 may be important in regulating mammalian cell cycle. SIRT2 protein levels increase during mitotic phase of the cell cycle and its overexpression delays mitosis (Dryden et al. 2003Go). Consistent with the idea that SIRT2 may restrain the cell cycle, expression of this sirtuin is down-regulated in human gliomas, compared with normal brain samples (Hiratsuka et al. 2003Go). SIRT2 colocalizes with chromatin during the G2/M transition, a period in which the nuclear membrane has broken down (Vaquero et al. 2006Go). Both SIRT2 and Hst2 show a preference for deacetylating histone H4 at Lys16 in vitro, and SIRT2 KO mouse embryonic fibroblasts (MEFs) display hyperacetylated H4K16 during mitosis. Since SIRT1 also deacetylates H4K16, SIRT2 and SIRT1 may function redundantly, at least during the M phase of the cell cycle. SIRT2 may also regulate other phases of the cell cycle, since G1 is extended and S is shortened in SIRT2-/-MEFs.


Mitochondrial sirtuins—key regulators of metabolism

Mitochondria are dynamic organelles that regulate nutrient utilization to provide the cell with energy even during dramatic changes in diet and development. Mitochondria also play a central role in mediating apoptosis in response to DNA damage or oxidative stress. These organelles are the primary site of reactive oxygen species (ROS) generation within the cell, and increased oxidative damage is proposed to be one cause of mammalian aging (Harmon 1956Go; Wallace 2005Go).

The mitochondrial localization of SIRT3–5 is especially intriguing because mitochondrial dysfunction is associated with mammalian aging and many diseases, including diabetes, neurodegenerative diseases, and cancer (Wallace 2005Go). Do mitochondrial sirtuins regulate metabolism, the oxidative stress response, and ultimately, mammalian aging? It is important to note that although SIRT1 is not itself physically associated with mitochondria, as described above, it also impacts mitochondrial functions. Lifespan analysis of animals with varying SIRT3–5 level has not been performed; however, there is growing evidence linking mitochondrial sirtuins with regulating energy usage and even human lifespan.


SIRT3

SIRT3 was the first sirtuin shown to be localized to the mitochondria of mammalian cells (Onyango et al. 2002Go; Schwer et al. 2002Go; Michishita et al. 2005Go). SIRT3 is localized to the mitochondrial matrix and cleavage of its signal sequence is necessary for enzymatic activity (Schwer et al. 2002Go). SIRT3 deacetylates multiple substrates in vitro including histone peptides (Onyango et al. 2002Go; Schwer et al. 2002Go) and tubulin (North et al. 2003Go).

The biological functions of SIRT3 are beginning to emerge. SIRT3 is expressed in brown adipose tissue and induced by cold exposure (Shi et al. 2005Go). Moreover, the deacetylase activity of SIRT3 is reported to be required for the induction of uncoupling protein 1 (UCP-1). SIRT3 also appears to regulate mitochondrial functions, as its overexpression increases respiration, while decreasing reactive oxygen species production (Shi et al. 2005Go).

Two recent studies demonstrate that SIRT3 may regulate the activity of acetyl-CoA synthetase (AceCS) (Hallows et al. 2006Go; Schwer et al. 2006Go), representing a striking, conserved activity with the bacterial sirtuin, cobB (Tsang and Escalante-Semerena 1998Go; Starai et al. 2002Go). AceCS uses acetate, CoA, and ATP to form acetyl-CoA, which is an intermediate in the TCA cycle, and is also required for cholesterol and fatty acid synthesis. Acetylation of mitochondrial AceCS (AceCS2) inactivates the enzyme, whereas deacetylation by SIRT3 activates it. Interestingly, SIRT1 can deacetylate and activate the cytosolic form of AceCS (AceCS1). These data suggest that SIRT3 may play a role in regulating the entry of carbons from acetate into central metabolism. It will be important to assess the in vivo relevance of these findings using SIRT3-/- mice. In sum, SIRT3 may be especially important under conditions of energy limitation—i.e., during fasting or CR to ensure full incorporation of dietary or ketone-derived acetate into metabolism.

In human population studies, polymorphisms within the SIRT3 gene have been linked to longevity. The G477T transversion, while not affecting the amino acid sequence, associates with survivalship of elderly males (Rose et al. 2003Go) and may signify a haplotype promoting longevity. The same group found that a variable number of tandem repeats (VNTR) enhancer within SIRT3 also associates with lifespans >90 yr (Bellizzi et al. 2005Go). These findings will need to be validated in larger samples, but suggest that the expression of SIRT3 may promote longevity in humans and raise the importance of performing lifespan experiments in mice that overexpress or lack SIRT3.


SIRT4

SIRT4 is another mitochondrial protein (Michishita et al. 2005Go) that regulates energy usage. SIRT4 lacks detectable deacetylase activity (North et al. 2003Go), but demonstrates ADP-ribosyltransferase activity. SIRT4 plays an important role in regulating amino acid-stimulated insulin secretion (AASIS) in pancreatic ß-cells by ADP-ribosylating and inhibiting glutamate dehydrogenase (GDH) (Fig. 3; Haigis et al. 2006Go). GDH converts glutamate into {alpha}-ketoglutarate, a TCA cycle intermediate. GDH-activating mutations cause hyperinsulinism in humans showing that this enzyme regulates insulin secretion by gating the flow of amino acids into central metabolism in ß-cells (Stanley et al. 1998Go). SIRT4 KO mice have no gross abnormalities, but display higher GDH activity and higher levels of circulating insulin. SIRT4 KO mice have elevated AASIS, and strikingly, unlike wild type, they secrete insulin in response to glutamine.



SIRT4 may also connect insulin secretion with CR. Islets isolated from CR mice demonstrate increased AASIS similar to islets from ad libitum fed SIRT4 KO mice (Haigis et al. 2006Go). Down-regulation of SIRT4 during CR appears to mediate this effect because GDH from islets of CR mice is less ADP-ribosylated and more active than GDH from control islets. A similar change in GDH is found in the liver of CR mice. We suggest that SIRT4 coordinates a physiological response in liver and ß-cells during energy limitation (Fig. 3). In the liver, the flow of carbon from amino acids into gluconeogenesis would be increased, and in ß-cells the ability amino acids to trigger insulin secretion would be elevated. Since overall insulin secretion is clearly lower during CR because of reduced blood glucose, it seems likely that the spectrum of insulin secretagogues is thus shifted from carbohydrates to amino acids.

We note that the apparent down-regulation of SIRT4 during CR in ß-cells and liver goes against the expectation that sirtuin activity should increase during this dietary regimen. However, it is consistent with the observed reduction in the NAD/NADH ratio in liver of CR mice (Hagopian et al. 2003bGo) and the observed increase in gluconeogenesis in this organ (Hagopian et al. 2003aGo). We suggest that a shift from carbohydrates to fat as preferred energy source during CR may help drive down the NAD/ NADH ratio in these tissues and thus moderate these metabolic changes.


SIRT5

SIRT5 remains the least characterized sirtuin to date. SIRT5 is described as a mitochondrial protein (Michishita et al. 2005Go) and has weak deacetylase activity (North et al. 2003Go), but does not appear to possess an ADP-ribosyltransferase activity (Haigis et al. 2006Go). Possible physiological substrates of SIRT5 and its role in mammalian cells are not yet known.


Summary and perspective

Sirtuins have emerged as key antiaging genes in model organisms. The NAD-dependence of these proteins links them unavoidably to the metabolic activity of cells. In several organisms, sirtuins have been shown to be regulated by and to mediate the effects of the dietary regimen CR. Moreover, mammalian sirtuins have been implicated in stress resistance and numerous metabolic pathways, including adipogenesis, gluconeogenesis, and insulin and glucose homeostasis.

While it may be years before we know whether sirtuins regulate mammalian lifespan, current data suggests that these proteins are regulated by diet and in turn, regulate multiple facets of physiology, making them interesting therapeutic targets for metabolic and neurodegenerative diseases (Fig. 4). Several important questions need to be resolved to improve our understanding of sirtuin biology and their therapeutic potential. First, what other functions can be ascribed to the SIRT1–7 proteins? We predict that like SIRT1, SIRT2–7 will have multiple targets and may affect many biologies. Second, how is sirtuin function regulated? Two known mechanisms of sirtuin regulation are (1) its protein induction during CR or fasting (Cohen et al. 2004Go; Nisoli et al. 2005Go; Rodgers et al. 2005Go) and (2) regulation by metabolites NAD, NADH, and nicotinamide (Anderson et al. 2003aGo, bGo; Lin et al. 2004Go). Third, are all seven sirtuins regulated in the same direction by CR in any given tissue? If the NAD/ NADH ratio (or nicotinamide concentration) is a primary determinant of regulation, it is possible that all seven sirtuins will be regulated similarly by CR in a given cell, since the NAD/NADH ratio has the potential to equilibrate throughout cellular compartments by shuttle systems. Another likely possibility is that the concentration of NAD and its metabolites are regulated compartmentally. For example, enzymes involved in NAD biosynthesis are found in the nucleus, peroxisome, Golgi apparatus, mitochondria, and cytoplasm (Yang et al. 2006Go). Fourth, does the NAD/NADH ratio change in different directions depending on the tissue in response to CR? This possibility seems likely, since CR induces distinct metabolic changes in different tissues (e.g., activation of gluconeogenesis in liver and fat loss in white adipose tissue). Moreover, fasting decreases SIRT1 activity in the pancreas (Bordone et al. 2006Go), while increasing its activity in the liver (Rodgers et al. 2005Go). This last question is important, because if sirtuins are activated by CR in some tissues but repressed in others, genetically altered mice (i.e., SIRT knockout or overexpressed) or pharmacological interventions to activate or repress a sirtuin systemically (see below) may mimic CR in only a segmental fashion.



In the next few years the answers to these and other questions will auger how well pharmacological agents that target sirtuins will serve as CR mimetics. This path of drug intervention is especially compelling, because CR mitigates many major diseases in rodent models. We can hope that new classes of drugs are on the horizon to deliver broad benefits for these diseases.

 

[ThaG]

Mitochondria, Oxidants, and Aging

The free radical theory of aging postulates that the production of intracellular reactive oxygen species is the major determinant of life span. Numerous cell culture, invertebrate, and mammalian models exist that lend support to this half-century-old hypothesis. Here we review the evidence that both supports and conflicts with the free radical theory and examine the growing link between mitochondrial metabolism, oxidant formation, and the biology of aging.

Article Outline

Main Text

The Free Radical Theory
Mitochondrial ROS Generation: The Belly of the Beast
Where Are ROS Generated in the Cytochrome Chain?

Basal Rate and Elimination of ROS Generation
Age-Related Changes in Mitochondrial Function
Mitochondria and Aging in Yeast and Worms
Aging in Flies
Mammalian Models of Aging
Conclusions
Acknowledgements
References



Main Text

Many believe that the seeds of aging can be traced back to a chance encounter that occurred sometime between one and two billion years ago. The event of note involved the presumed incorporation of an uninvited eubacteria into an Archea-type host. This was presumably not the first such encounter between a host cell and invading bacteria. Nonetheless, it is generally presumed that previous interactions had led to the death of the invader, the host, or more likely both organisms. However, on this day, rather then a case of mutual destruction, the results of this incursion led to an agreement between host and bacteria that has persisted more or less intact until this day. Although most of the initial details are unknown, over time, the two initial antagonists began to rely on each other in ever more intricate and interconnected ways. The eubacteria is believed to have evolved into the mitochondria, able to safely replicate within limits inside its host. Slowly, as it became more comfortable in its new environment, more and more responsibility for its own replication and maintenance was shifted to the cell. Indeed, today, of the thousand or so proteins that make up the mitochondria, only a handful are still encoded by mitochondrial DNA. The host, too, began to rely more heavily on its onetime invader. It began to see the advantage of specialization and compartmentalization. Although the cell still retained the capacity to produce energy independent of the mitochondria, more and more of the day-to-day responsibility was turned over to this new organelle. Suddenly, the ability of these onetime unwanted invaders to efficiently produce chemical energy seemed to allow the possibility of powerful muscles or prodigiously beating hearts.

Our story might have ended there, a heart-warming tale of two potential enemies that joined forces to work together for the common good. Yet old habits are indeed hard to break. And although it is true that, on that day, the invading eubacteria did not immediately kill its host, it has also become increasingly clear that it may not have entered the agreement with full disclosure. For, although mitochondria are marvels of energy production, they also have another, less beneficial legacy in the cell. Increasingly, this other property, the continuous production of potentially harmful reactive oxygen species (ROS), has become a central focus of aging research. The purpose of this review is therefore to reexamine the terms of a now nearly two billion-year-old agreement and in particular to evaluate to what degree the mitochondrial production of ROS and the cellular response to oxidative stress may be a major determinant of how we age.
The Free Radical Theory

It has been nearly 50 years since Harman proposed the “free radical theory” of aging (Harman, 1956). The initial theory suggested that aging, as well as the associated degenerative diseases, could be attributed to deleterious effects of free radicals on various cell components. When originally proposed, the notion that cells actually produced free radicals remained unproven and hotly debated. Even after Harman’s proposal, this controversy raged for the next decade or so, until it was ultimately settled with the discovery of the enzyme superoxide dismutase (McCord and Fridovich, 1969). The existence of an intracellular enzyme whose sole function was to scavenge superoxide seemed indirect but powerful evidence that cells must continuously produce their own free radicals. Nonetheless, with relatively little direct experimental support, Harman initially speculated that the source of age-inducing free radicals was most likely “the interaction of the respiratory enzymes involved in the direct utilization of molecular oxygen.” Such a hypothesis was also generally consistent with early theories suggesting a correlation between overall metabolic rate and life span. In the 1920s, Pearl proposed the “rate of living” hypothesis that directly linked metabolic rate with the longevity of an organism (Pearl, 1928). Pearl was unclear what the precise mechanism was that linked metabolism with life span and therefore suggested that some vital cellular element was somehow consumed in proportion to overall metabolic rate. His concept was that, when this unknown but vital element was exhausted, death occurred.

Since Harman’s initial formulation, ensuing experimentation has solidified but not proven his underlying theory. Today, although aerobic metabolism and the corresponding generation of ROS remain the most widely accepted cause of aging, substantial gaps and unknowns persist. Indeed, fundamental questions regarding what governs the relationship between overall metabolic rate and the production of ROS remain unclear. Perhaps more important, relatively little is known about what are the relevant intracellular targets for ROS and how oxidative modification of these targets might influence life span. Fifty years after its formulation, we now know that cells make free radicals on a continuous basis, yet we are still uncertain as to whether they cause or merely correlate with aging.
Mitochondrial ROS Generation: The Belly of the Beast

ROS are generated in multiple compartments and by multiple enzymes within the cell. Important contributions include proteins within the plasma membrane, such as the growing family of NADPH oxidases (Lambeth 2004); lipid metabolism within the peroxisomes; as well as the activity of various cytosolic enzymes such as cyclooxygenases. Although all these sources contribute to the overall oxidative burden, the vast majority of cellular ROS (estimated at approximately 90%) can be traced back to the mitochondria. The generation of mitochondrial ROS is a consequence of oxidative phosphorylation, a process that uses the controlled oxidation of NADH or FADH to generate a potential energy for protons (ΔΨ) across the mitochondrial inner membrane. This potential energy is in turn used to phosphorylate ADP via the F1-F0 ATPase. At several sites along the cytochrome chain, electrons derived from NADH or FADH can directly react with oxygen or other electron acceptors and generate free radicals. In the past, the generation of ROS or other free radicals was thought of as “slippage” or an unproductive side reaction. More recently, as will be discussed later in more detail, it has been proposed that mitochondrial ROS may actually be important in various redox-dependent signaling processes (Nemoto et al., 2000, Werner and Werb, 2002 and Dada et al., 2003) as well as in the aging clock.

A schematic diagram of the flow of electrons through the cytochrome chain is presented in Figure 1. NADH, generated by associated Krebs cycle dehydrogenases (DH) (Robinson and Srere 1985), is initially oxidized at site I. As the electrons from NADH are passed to the first mobile electron acceptor, oxidized coenzyme Q, the energy is dissipated by the ejection of protons. Coenzyme Q can also accept electrons from the site II complex donated by FADH, thereby bypassing site I and one proton ejection site. Coenzyme Q next donates electrons to cytochrome b in site III in a near potential energy neutral process. In site III, the electrons are passed to cytochrome c1 with the dissipative ejection of protons. Cytochrome c1 transfers its electrons to the second mobile element in the cytochrome chain, cytochrome c. Cytochrome c in turn reduces cytochrome a,a3 (i.e., cytochrome oxidase [COX] referring to the terminal electron acceptor) in site IV, which ultimately reduces molecular oxygen to form water. This final dissipation of the redox energy in NADH/FADH at site IV is also associated with a final ejection of protons. In this manner, the cytochrome chain transforms the redox energy of the rather stable molecules NADH and FADH into a ΔΨ across the inner mitochondrial. In this complex reaction sequence, several important questions naturally arise: Where are ROS generated? What is the basal rate of mitochondrial ROS generation? What regulates mitochondria ROS generation? How are ROS eliminated?



Where Are ROS Generated in the Cytochrome Chain?

The two major sites for ROS generation are believed to be at sites I and III where large changes in the potential energy of the electrons, relative to the reduction of oxygen, occur (see Figure 1). Experimental manipulations that increase the redox potential of site I (Kushnareva et al., 2002) or site III (Chen et al., 2003) generally increase the rate of ROS generation, supporting the notion that the redox potential of these reactive sites is important in free radical formation. Site I remains the least understood of the cytochrome chain elements (Yagi and Matsuno-Yagi, 2003). This multisubunit complex is believed to be composed of not, vert, similar46 proteins with a combined molecular weight exceeding 1 MDa and is thought to contain at least one bound flavin mononucleotide (FMN) and eight iron sulfur groups. Both the iron sulfur groups (Genova et al., 2001) and FMN sites have been implicated in ROS generation (Liu et al., 2002). At site III, an even more complicated story is found with the Q cycle contributing to the generation of O2− through reduced ubisemiquinone either on the inner or outer membrane surfaces (Turrens et al., 1985 and Aguilaniu et al., 2003 ). Surprisingly, the terminal oxidation step at cytochrome oxidase is not believed to be a significant source of ROS in intact systems, despite the fact that this site is capable of in vitro ROS generation (Fridovich and Handler, 1961).

Since the precise mechanisms of ROS generation are unknown, we can construct a simple relationship that might globally describe the generation of ROS in the mitochondria. First, we will define the term Eox as the net driving force for the reduction of oxygen. Eox can be estimated as the difference between the redox potential for donating a single electron to oxygen (Eo = −160 mV, Wood [1988]) and the redox potential of a particular electron donor at a given reaction site. Also important is the oxygen tension, PO2, and a hypothetical first order reaction constant, Kox, resulting in the following simple equation for total net ROS generation (QROS):

In the equation, “site” represents all of the mitochondrial ROS-generating sites in a given cell. Since the concentration of aqueous O2 is significantly higher than that of O2− or H2O2 (Cadenas and Davies 2000), under most conditions, the reverse reaction rate can be ignored.

Based on this simplistic approach, any perturbation to oxidative phosphorylation that changes these terms, including the number of mitochondria or cytochrome chain equivalents within a cell, would increase the production of ROS. The regulation of QROS can be evaluated for each one of these elements. For instance, putting the mitochondria in a reduced state without ADP or Pi for oxidative phosphorylation (state 4) (Loschen et al., 1971) or the addition of electron transport inhibitors that secondarily increase the Eox of site I or site III all tend to increase QROS (Staniek and Nohl, 2000, Aguilaniu et al., 2003 and Loschen et al., 1971 ). Experimentally, a large increase in ROS formation is often seen in the condition known as reverse electron flow. This is usually achieved when a site II substrate, succinate, is added in the presence of a site III inhibitor, thereby generating a reverse flow of electrons from site II to site I (Loschen et al., 1971 and Aguilaniu et al., 2003 ). Reverse electron flow might also be responsible for the high ROS generation occurring with fatty acid oxidation that also generates electrons for site II via FADH (Boveris et al., 1972 and Boveris and Chance, 1973). Another “endogenous” example in which Eox and ROS are increased is with the release of cytochrome c during apoptosis. In this condition, the absence of cytochrome c results in a block in electron flow and a rise in the Eox of site I (Kushnareva et al., 2002) with a subsequent rise in ROS produced at this site (Kushnareva et al., 2002). This does not appear to happen at site III since in the absence of electron transport there is also an inhibition of Q cycle turnover that is required for generating the active ubisemiquinone (Turrens et al., 1985). Finally, another important physiological regulator of Eox is the family of uncoupling proteins (UCPs) (Echtay et al., 2002 and Casteilla et al., 2001). Since Eox and ΔΨ are coupled through the proton ejection process, these parameters are proportional under most conditions. If ΔΨ is reduced by the action of uncoupling proteins “leaking” protons across the inner membrane (see Figure 1), then predictably Eox and therefore QROS would be decreased. Thus, uncoupling has been proposed as an important mechanism to reduce ROS levels (Casteilla et al., 2001 and Brand et al., 2004). Interestingly, UCPs might also be directly activated by superoxide anions, thereby providing an overall feedback circuit for ROS production (Echtay et al., 2002).

An additional recent proposal for ROS regulation is that the entry of electrons into and through the cytochrome chain, especially at the level of the DH-site I complex, is highly regulated. This electron gate would presumably only permit oxidative phosphorylation to occur when it was required by cellular energetic needs (Bose et al., 2003 and Joubert et al., 2004). This mechanism potentially permits the modulation of oxidant formation without the requirement to dissipate ΔΨ through the energetically nonproductive action of UCPs. There is also growing appreciation that the activity of components of the electron transport chain can also be altered by covalent modification and that these modifications may be important in ROS formation (Ludwig et al., 2001). Finally, oxygen tension represents another variable known to regulate the rate of mitochondrial ROS production (Turrens et al., 1982). The mechanisms underlying the matching of tissue oxygen level to metabolic demand are poorly understood. Yet it is clear that tissue pO2 can be dynamically regulated, and, indeed, the increase in venous oxygen content observed during brain activation is the physiological parameter that forms the basis for fMRI detection (Ogawa et al., 1992).

From this brief review, it is clear that there are indeed many factors that can regulate mitochondria ROS generation. Presumably, this complexity contributes in part to the numerous conflicting reports in the literature regarding the nature, control, and degree of mitochondrial ROS generation. The above discussion does, however, raise an important caveat concerning the relationship between oxidant formation and metabolic rate (i.e., oxygen consumption). In general, the higher the rate of oxidative phosphorylation and oxygen consumption, the lower the overall value of cytochrome chain Eox. Consistent with this notion was the early observation obtained from isolated mitochondria that augmenting oxidative phosphorylation resulted in a reduction, not an augmentation, of ROS generation (Loschen et al., 1971). Although it is commonly assumed that an increase in oxygen consumption produces an increase in ROS production, we would argue that this positive correlation is only true if the increase in oxygen consumption was secondary to a higher tissue pO2 or an increase in the number of “sites”, i.e., functional mitochondria. In contrast, an increase in oxygen consumption in the setting of constant tissue oxygen concentrations and a fixed number of mitochondria would favor a decrease in ROS levels.
Basal Rate and Elimination of ROS Generation

There is a wide variance in the literature regarding what percentage of basal mitochondrial oxygen consumption ultimately leads to ROS generation. This is not surprising, since most of these conclusions are based on isolated mitochondria studies in which the Eox of many of the redox elements were held at very unphysiological states or under conditions in which reverse electron flow was possible. Based on these initial observations (for review, see Chance et al. [1979]), it was suggested that not, vert, similar2% of the total oxygen consumption was funneled to ROS generation. This number has been widely cited despite the fact that, even in these early studies, it was appreciated that the ROS measurements were made under artificial conditions. Subsequent studies, under more physiological conditions, have reduced this basal value to not, vert, similar0.2% (Staniek and Nohl, 2000 and Aguilaniu et al., 2003 ). This lower value does not imply that the baseline production is unimportant but does suggest that the mitochondrial ROS load on most tissues may not be as severe as once thought.

The discovery of superoxide dismutase (SOD), as discussed previously, was a major step in establishing the generation of ROS or H2O2 in the mitochondria. Two intracellular SOD enzymes exist within the cell: SOD2, a manganese-dependent enzyme in the matrix, and SOD1, a copper-containing enzyme primarily in the cytosol. Both of these enzymes convert O2− into H2O2 that is then further deactivated by catalase (Radi et al., 1991) to water and oxygen or by the various glutathione peroxidases to reduced glutathione and water. The discovery of the peroxiredoxins (Chang et al., 2004) represents yet another family of important scavenging enzymes in the mitochondria (Taylor et al., 2003). Superoxide that is not immediately scavenged can directly react with oxidized cytochrome c (Joubert et al., 2004 and Butler et al., 1975) or cytochrome oxidase (Orii, 1982). The ability of the matrix to withstand large transient loads of free radicals with minimal damage using these protection mechanisms has been demonstrated during the extremely rapid photooxidation of not, vert, similar1 mM matrix NADH to NAD+ + 2ROS in intact mitochondria (Joubert et al., 2004). These highly efficient scavenging systems suggest that any measured release of ROS from mitochondria may represent only a small fraction of the total ROS generated.
Age-Related Changes in Mitochondrial Function

A number of studies have demonstrated that mitochondrial integrity declines as a function of age (Shigenaga et al., 1994). Age-dependent increases in the level of damaged DNA have been commonly assessed through biomarkers such as the formation of 8-oxo-2′-deoxyguanosine (oxo8dG). In postmitotic tissue such as brain, the levels of oxo8dG are significantly higher in mitochondrial compared to nuclear DNA (Richter et al., 1988). Reasons for these differences are thought to include the proximity of mitochondrial DNA to the source of oxidants and the lack of any protective histone covering. This postulated and observed increased sensitivity of mitochondrial DNA to oxidative damage has led to the concept of the “vicious cycle” in which an initial ROS-induced impairment of mitochondria leads to increase oxidant production that, in turn, leads to further mitochondrial damage.

Experimental evidence both for and against the vicious cycle exists. For instance, many studies have demonstrated that old mitochondria appear morphologically altered and functionally produce more oxidants and less ATP (Shigenaga et al., 1994). Nonetheless, other investigators have recently criticized the methodology employed in some of these studies (Maklashina and Ackrell, 2004). Therefore, whether or not there is a significant impairment of electron transport activity as mitochondria age remains somewhat of an open question. Recent genomic studies, however, suggest that, transcriptionally, components of the electron transport chain are indeed affected by aging. In one interesting study, the authors compared microarray data between two organisms (C. elegans and Drosophila) as they aged in an effort to obtain a consensus aging transcriptisome. In both species, there was a small but approximate 2-fold decrease in a large set of genes involved in ATP synthesis and mitochondrial respiration (McCarroll et al., 2004). Although these studies would seem to support the vicious cycle concept, two caveats are worth mentioning. First, these authors were studying nuclear-encoded, not mitochondria-encoded, transcripts. Second, the exact timing of the downregulation for these transcripts occurred when the animal was at the early adult stage. This transcriptional change therefore appears to occur before the usually observed decline in mitochondrial function and presumably also before one would expect the cumulative effects of oxidants to begin having their peak effects.

Since the formation of ROS species is a function of ambient oxygen concentration (Turrens, 2003), the cellular and organismal response to high oxygen concentrations may represent an insightful stress to explore the mechanisms of aging. Here again, transcriptional profiling may provide a glimpse of underlying mechanism. For instance, comparison of the gene expression patterns of Drosophila undergoing normal aging and those flies exposed to acute hyperoxia revealed significant concordance (Landis et al., 2004). In another recent study, the biological resistance to hyperoxia was used as a genetic screen to obtain Drosophila mutants that are either overly sensitive or resistant to this stress (Walker and Benzer, 2004). A careful analysis of the morphological changes that mitochondria underwent after high oxygen exposure demonstrated that individual mitochondria develop a previously unknown “swirl” phenomenon. This altered morphology presumably occurs due to a rapid reorganization of mitochondrial cristae in response to oxidative stress. Interestingly, mitochondria from older flies have significantly more swirls then younger flies, and mutants selected for increased swirl formation have a significantly shorter life span.

The cellular response to high oxygen also supports a role for intracellular oxidants as at least one important determinant of the life span of mammalian cells in culture. Cellular senescence is an interesting biological phenomenon whereby nonimmortalized cells, after a discrete number of passages, undergo a permanent withdrawal from the cell cycle. The senescent state is accompanied by consistent morphological and biochemical changes, suggesting it may be programmed in much the same way as differentiation or apoptosis. Significant questions persist as to whether the molecular mechanisms underlying cellular senescence are relevant to overall organismal aging. With that said, it has been recognized for some time that lowering the ambient oxygen concentration can significantly extend the life span of primary cells in culture (Packer and Fuehr, 1977). Similar prolongation of cellular life span can be achieved by augmenting antioxidant levels. For instance, increasing the level of superoxide dismutase extends the life span of primary fibroblasts as well as decreasing the rate of telomere shortening (Serra et al., 2003). Conversely, knockdown of SOD using RNAi was demonstrated to induce senescence (Blander et al., 2003). Interestingly, reducing SOD by RNAi resulted in the induction of p53, and this induction was required for senescence. Cellular induction of p53 can result in either apoptosis or senescence, and there is some evidence that the decision for what cell fate pathway is chosen may depend on the intracellular level of ROS (Macip et al., 2003). In a similar fashion, expression of an activated form of Ras proteins can induce senescence in some primary fibroblasts (Serrano et al., 1997). This Ras-induced senescence is also accompanied by p53 induction as well as a rise in ROS levels. Again, either antioxidant augmentation or lowering the level of ambient oxygen rescued Ras-expressing cells from entering senescence (Lee et al., 1999). Recently, seladin-1, a gene previously implicated in cholesterol metabolism, was implicated as an important redox-sensitive intermediary between Ras and p53 (Wu et al., 2004).
Mitochondria and Aging in Yeast and Worms

The use of simple organisms such as worms and flies would seem the ideal testing ground for a free radical or metabolic theory of aging. Indeed, there are a growing number of studies that support a central role for mitochondrial metabolism in the aging process. While it is impossible for us to highlight all such studies, we will attempt to review a selected number of what we view as some of the most important observations.

Perhaps the most straightforward relationship between metabolic rate and aging is the observation that global manipulations that effect metabolism or metabolic substrates can also lead to alterations in life span. For instance, lowering ambient temperature in worms or flies slows the metabolic rate and also results in an concomitant extension of life span (Miquel et al., 1976). Similarly, in yeast, a simple reduction of available glucose in the media results in life extension. This paradigm has been used as a model of caloric restriction (CR), and experimental evidence suggest that life extension under these conditions requires the NAD-dependent deacetylase Sir2 (Lin et al., 2002). It has also been shown that CR activates Sir2 activity, although the precise details remain controversial. More relevant to our discussion is that S. cerevisiae appear to shift their utilization of glucose from fermentation under normal conditions to predominantly mitochondria-based aerobic respiration when glucose is limiting. This metabolic shift results in an increase in overall oxygen consumption (Lin et al., 2002). In yeast, therefore, CR leads to longer life but a presumably higher metabolic rate. These observations would appear to conflict with the straightforward predictions of the free radical theory of aging that would suggest that higher rates of aerobic metabolism would decrease life span. Nonetheless, as discussed previously, the relationship between oxygen consumption and ROS generation is complex. Further measurements in this model of the level of oxidative stress under normal and CR conditions are therefore needed.

The role of Sir2 and its mammalian ortholog SIRT1 is an active area of research, and there are a number of recent reviews that describe the biology of this family of proteins (Denu, 2003, North and Verdin, 2004 and Blander and Guarente, 2004). Relevant to our discussion, Sir2, although essential for the yeast CR response, does not appear to directly alter antioxidant defenses in yeast (Lin et al., 2002). Interestingly, however, in aging yeast there appears to be an accumulation of oxidatively modified proteins that accumulate with replicative age. Under normal conditions, these modified proteins are asymmetrically distributed between mother and daughter cell, and the daughter cell is usually born without inheriting its equal share of damaged proteins. This sparing of the daughter cell does not occur in Sir2-deficient yeast, suggesting that the deacetylase may be important for protecting the cell against the consequences of oxidative damage (Aguilaniu et al., 2003). Consistent with this, the mammalian deacetylase SIRT1 also appears to protect cells from direct oxidative stress (Luo et al., 2001 and Brunet et al., 2004).

The dauer phase in C. elegans represents another global metabolic shift that is relevant to life span determination. Under optimal conditions, a developing worm passes quickly from a second stage larva (L2) to a third stage larva (L3). When conditions are less then optimal, perhaps because of a lack of food or overcrowding, rather than proceeding to the L3 stage, the worm enters an alternative stage known as dauer. In this somewhat suspended state, the worm can exist for up to 3–6 months. This represents a significant increase in life span, since, under normal conditions, the worm survives for less than a month. Once in the dauer stage, the larvae undergo a number of morphological and metabolic changes. Among these changes is the development of a hard and impermeable cuticle. The dauer larvae no longer actually feeds but rather depends on internal (primarily fat) stores to maintain their energetic needs. Of particular interest is that there is also a significant shift away from the use of the tricarboxylic acid (TCA) cycle and the use of electron transport and a heavier reliance on alternative energetic pathways (Wadsworth and Riddle, 1989). The entry into dauer also results in a nearly 4-fold decrease in oxygen consumption when compared to third stage larva (Vanfleteren and De Vreese, 1996).

Entry into dauer is stimulated by harsh external environmental conditions and is under the control of several distinct signaling pathways. One such well-studied pathway is homologous to the insulin/IGF pathway in mammalian tissues. As discussed in more detail in the accompanying reviews, this pathway ultimately controls the activity of DAF-16, a transcription factor that belongs to the Forkhead family of proteins. Mutations along this pathway that moderately increase the activity of DAF-16 appear to result in an extension of life span, while stronger mutations result in a constitutive dauer phenotype. Typically, the long-lived mutants, like their dauer counterparts, are more resistant to harsh environmental conditions such as increased heat or exposure to oxidative stress (Johnson et al., 2001). The basis for this heartiness appears in part to lie in the recently revealed transcriptional targets of DAF-16. Three separate groups have analyzed the DAF-16 transcriptosome. Each used slightly different methodologies, but all groups agree that DAF-16 regulates a large set of genes that modulate oxidative stress, as well as genes involved in overall metabolic regulation (Murphy et al., 2003, Lee et al., 2003a and McElwee et al., 2003). Interestingly, the property of modulating oxidative stress appears well conserved, as the mammalian ortholog of DAF-16, the transcription factor Foxo3a, also regulates many of the same antioxidant proteins (Nemoto and Finkel, 2002 and Kops et al., 2002).

Included among the metabolic genes induced by DAF-16 are a number of enzymes involved in mitochondrial transport as well as in fat metabolism and utilization. For instance, DAF-16 appears to regulate genes involved in the glyoxylate cycle, a process of fat utilization that usually involves the peroxisomes. Indeed, a recent intriguing hypothesis is that a number of long-lived C. elegans mutants can best be understood by tracing the various energy-generating pathways in worms (Rea and Johnson, 2003). In particular, these authors suggest that diversion away from the TCA cycle and the classic electron transport chain and toward alternative energy pathways is a common element of both the dauer as well as many long-lived phenotypes. This hypothesis therefore differs slightly from the previously discussed metabolic adaptation that occurs during CR in yeast (Lin et al., 2002). Nonetheless, in worms, these alternative energetic pathways, by reducing flux through the electron transport chain, were postulated to result in reduced ROS generation (Figure 2). It is important to note that the increased reliance on these alternative pathways can often result in energetically crippled, albeit long-lived animals. Mitochondrial-based metabolism and the TCA cycle presumably evolved in large part because of the ability to produce the most ATP molecules per unit of nutrient consumed. Reducing an organism’s reliance on such pathways may allow a worm to survive for an extended period of time in the controlled laboratory environment but would probably place this animal at a significant disadvantage in the real world, where only the fastest and reproductively fittest survive. Finally, in the context of the whole organism, it will be interesting to understand how circulating hormones such as insulin or IGF-1 may affect these same global parameters and in particular whether these ligands may shorten life by increasing the dependence on TCA-mediated metabolism.




The link between mitochondrial metabolism and longevity is also supported by several studies demonstrating that direct disruption of the electron transport chain can have a significant effect on life span. One of the first such long-lived mutants characterized is isp-1. This mutant consists of a missense mutation in a component of complex III of the respiratory chain (Feng et al., 2001). Isp-1 mutants demonstrate a generalized slowing in many developmental milestones and a corresponding reduction in oxygen consumption, consistent with a block in electron transport. The results from these initial isp-1 studies have been significantly extended by two RNAi-based screens to uncover additional genes that regulate life span. In one study, it was demonstrated that reducing the level of various components of respiratory chain complexes I, III, IV, or V resulted in smaller but longer-lived worms (Dillin et al., 2002). Interestingly this phenotype was only evident if RNAi treatment began at an early age, since similar treatment after the worm had reached the young adult stage did not prolong life. Similarly, a systematic RNAi screen that sought to inactivate over 5600 random C. elegans genes also implicated the mitochondria (Lee et al., 2003b). Although a large number of life span-determining genes were identified in this screen, the largest functional class appeared to be genes that, in some fashion, regulate mitochondrial activity. This class, comprising nearly 15% of the total pool of life-extending genes, included a host of electron transport chain elements, mitochondrial carriers, and mitochondrial ribosomal subunits. In general, these mitochondrial mutants, although long lived, had grossly altered mitochondrial morphology as well as significantly reduced oxygen consumption and steady-state ATP levels. In contrast to these consistent findings regarding energetics, there was a much more heterogeneous response of these mutants to exogenous superoxide or hydrogen peroxide challenge. While this latter result was interpreted as a lack of consistency for these mutants with regard to endogenous ROS levels, it is important to note that there have been few if any reports that directly measure levels of reactive oxygen species in living worms. Although oxidant levels are commonly assessed in mammalian tissue culture cells through the use of cell-permeant dyes or spin trapping reagents, such methods do not appear to work in adult C. elegans due to their hard exterior cuticle. As such, it is unclear what the resting levels of ROS were in these various mitochondrial crippled but long-lived mutants. Until such direct measurements are made in these mutants, the precise role of ROS levels in the observed increase in life span cannot be fully assessed.

Several other mutants that regulate life span are also worth mentioning here. Although life-shortening mutants are in general less instructive then life-extending mutants, studies with the mev-1 strain stand out as particularly notable exception. This short-lived mutant was initially isolated because of increase sensitivity to the ROS generator methyl violet. Subsequent studies have demonstrated that the mutation maps to a subunit of complex II (Ishii et al., 1998). Mev-1 animals have significant mitochondrial structural abnormalities and indirect evidence of increased ROS generation (Senoo-Matsuda et al., 2001). These worms also were shown recently to exhibit increased levels of nuclear DNA damage (Hartman et al., 2004), arguing that mitochondrial oxidants might be an important source of overall genomic instability. This conjecture is also supported by observations in mice that are heterozygous for mitochondrial superoxide dismutase (Sod2+/−). These mice also demonstrate an increased incidence of nuclear DNA damage as well as a significant increase in tumor formation (Van Remmen et al., 2003). Together, these observations raise important questions as to what are the most important intracellular targets for mitochondrial-derived oxidants. For instance, are the damaging effects of oxidants generated within the mitochondria confined to the mitochondria, or do they extend to the nucleus? Although the mev-1 and Sod2 experiments suggest that genomic DNA may indeed be an important target, it remains unclear whether, under normal circumstances, the production of mitochondria-derived ROS are high enough and the half-life of oxidants are long enough to cause significant nuclear damage.

One additional mutant in C. elegans is worth mentioning since it provides another important potential link between metabolism, oxidants, and aging. A long-lived mutant known as clk-1 lacks an enzyme required in the biosynthesis of ubiquinone (Ewbank et al., 1997). As previously discussed, ubiquinone, also known as coenzyme Q, is an important and well-conserved electron acceptor for both complex I- and II-dependent respiration. The synthesis of ubiquinone involves the sequential addition of isoprenyl subunits to the tail of the molecule, and long-lived clk-1 mutants accumulate a precursor of coenzyme Q containing eight rather the usual nine isoprenyl subunit tail. While all are agreed that clk-1 mutants are developmentally slowed and long lived, controversy remains as to whether metabolic rate as assessed by total oxygen consumption is reduced or unchanged in the clk-1 mutants (Branicky et al., 2000 and Braeckman et al., 2002). Interestingly, although coenzyme Q is often sold in health stores as a beneficial, life-extending antioxidant, withdrawal of the compound from the diet of wild-type worms can actually increase life span by 60% (Larsen and Clarke, 2002). Two other recent observations of the clk-1 mutant provide potentially important insights. The first is that a very recent study using intact functional mitochondria has found a specific deficit in complex I-dependent respiration in clk-1 mutants, while complex II-dependent respiration was normal (Kayser et al., 2004). This contrasts with previous measurements made with isolated cytochrome complexes that demonstrated a modest reduction in complex II activity (Felkai et al., 1999). It is presently unknown how this specific deficit in oxidative phosphorylation is achieved since, as mentioned, coenzyme Q was previously thought to be involved in both complex I- and II-dependent respiration. Nonetheless, these results suggest a potential caveat in measuring total oxygen consumption or isolated individual mitochondrial complex activity and assuming that electron transport is unchanged. Although difficult, more precise biochemical studies using intact functional mitochondrial preparations derived from either wild-type or long-lived C. elegans animals would seem to be an important avenue of investigation. Another observation that might be relevant to understanding the link between the clk-1 mutant and oxidants involved an attempt to understand why these animals undergo a profound slowing of multiple developmental processes. A recent study implicated a decrease in cytoplasmic ROS levels as the culprit for these developmental delays in the clk-1 animals (Shibata et al., 2003). Interestingly, these altered levels of cytoplasmic ROS appear to effect development not as random, destructive elements but rather as specific signaling molecules. In the case of clk-1 mutants, ROS appear to act in part as downstream effectors of the small GTPase Ras to modulate developmental signals. This result is consistent with a growing appreciation that a number of signaling pathways are regulated in some fashion by redox modulation (Finkel, 2003). In these varied cases, ROS are acting as specific signaling molecules. If, in fact, alterations in ROS regulate the developmental delay in clk-1 by acting as specific messengers, it raises the possibility that ROS may potentially regulate aging in a similar, specific fashion.
Aging in Flies

Many of the pathways operational in C. elegans also appear to regulate life span in Drosophila. In particular, the insulin/IGF pathway appears as an important regulator of longevity. For instance, a null mutation in the insulin substrate chico results in an approximate 40% increase in maximal life span of female flies (Clancy et al., 2001). Similarly, increased activity of the Forkhead transcription factor dFoxo appears to slow aging in Drosophila as it does in C. elegans (Giannakou et al., 2004). Drosophila have also been extensively used to assess the role of antioxidant enzyme overexpression in determining life span. Several studies have suggested that expression of superoxide dismutase alone or in conjunction with the peroxide scavenging enzyme catalase could slow aging (Orr and Sohal, 1994, Parkes et al., 1998 and Sun and Tower, 1999). More recent evidence suggests that these initial results, which would strongly support the free radical theory of aging, may have been inadvertently biased by using relatively short-lived control strains. Repeat experiments in long-lived strains suggest that the expression of various antioxidant proteins is without significant effect (Orr et al., 2003).

Another instructive Drosophila life span mutant is the colorfully named Indy (I’m not dead yet, a tribute Monty Python fans will recognize). This mutant demonstrates a 50% increase in maximal life span (Rogina et al., 2000). Analysis of the Indy gene product demonstrated that it belongs to a family of proteins that, in mammalian cells, handle the uptake of Krebs cycle intermediates such as citrate and succinate. While this mutant would seem to fit the general paradigm of altered mitochondrial metabolism leading to altered life span, it should be noted that direct measurement of metabolic rates in Indy mutants revealed they were indistinguishable from wild-type flies (Marden et al., 2003). Similarly, dietary-restricted flies and flies with mutant chico genes also failed to have detectable alterations in their overall metabolic rate and/or oxygen consumption (Hulbert et al., 2004). Therefore, as attractive as the notion is that mutants such as chico and Indy have altered metabolism and that these metabolic alterations are the basis for their longevity, direct proof that this mechanism is operational remains elusive.

One additional interesting Drosophila mutant originally isolated from a genetic screen is the long-lived strain methuselah (mth). This mutant exhibited a 35% increase in average life span and was resistant to numerous stressors including heat, starvation, and oxidants (Lin et al., 1998). Analysis of the mth gene product demonstrated that it belonged to the family of GTP binding protein-coupled receptors (GPCR), the family of seven transmembrane spanning receptors that modulate a host of signaling pathways. These results suggested that, at least in Drosophila, GPCR-regulated pathways might modulate stress resistance. It also raised the possibility that small molecule antagonists for the endogenous ligand of mth may extend life span. Until recently, however, the endogenous ligand of mth has remained unknown. However, a fascinating new study suggested that the cognate ligand is encoded by the Drosophila gene product stunted (Cvejic et al., 2004). There are two splice variants of the stunted gene, one encoding a 60 amino acid peptide and the other a slightly smaller 56 amino acid peptide. Both peptides encode for a highly conserved protein that corresponds to the epsilon (Porson) subunit of the F1F0-ATP synthase of the electron transport chain. This provocative result suggests a potentially important connection between metabolism, stress resistance, and GPCR signaling. Nonetheless, it also raises a number of important unanswered questions including how a presumed mitochondrial protein can function as a ligand for a cell surface receptor. Finally, the intersection between GPCR signaling and metabolism was recently strengthen by another fascinating observation that the citric acid intermediates succinate and α-ketoglutarate are actually the endogenous ligands for two separate mammalian orphan GPCR (He et al., 2004). Together, these results hint that classical cell surface signal transduction pathways and intracellular metabolism may be significantly more interconnected than previously suspected.

The theme of stress signaling and life span has also been bolstered by another recent report in which the authors demonstrated that JNK-dependent pathways were stimulated by cellular stress and were responsible for inhibiting the toxic effects of intracellular ROS (Wang et al., 2003). Augmenting JNK signaling in Drosophila resulted in less evidence of oxidative damage and a significant increase in life span. Given that JNK activity is also regulated by intracellular oxidants in mammalian cells, this suggests an important and perhaps evolutionary conserved target for intervention. Nonetheless, these results also suggest that there are many potential intracellular targets for ROS, and it is not yet clear which of these targets is the most relevant for regulating life span (Figure 3).


Mammalian Models of Aging

The development of mammalian models of aging in many ways appears to support and converge with the previously described invertebrate studies. For instance, there are a number of long-lived mutant mice that appear to modulate the insulin/IGF-1 pathway in a manner that is analogous to the daf-2/age-1/daf-16 mutants in C. elegans. Among these previously described mice are the Ames, Snell, and Laron strains. All of these mice are smaller then their wild-type counterparts, with their dwarf status a result of deficiencies in several secreted factors including growth hormone, thyroid hormone, and often IGF-1 (Quarrie and Riabowol, 2004). Besides their smaller size, these long-lived strains exhibit other potentially important characteristics such as a decrease in basal body temperature and a modest increase in antioxidant scavenging capacity. Recently, the role of the insulin/IGF-1 pathway in mammalian aging was examined in a more directed fashion. Two studies have suggested that these pathways are indeed important determinants of life span. Although complete knockout of the IGF-1 receptor (IGF-1R) is lethal, heterozygotes are viable. Analysis of the life span of such animals demonstrated that female heterozygotes lived approximately 30% longer than control wild-type mice, while male heterozygotes had roughly a 15% life span extension (Holzenberger et al., 2003). Similar to the daf-2 mutants, IGF-1R heterozygotes also exhibit an increase in oxidative stress resistance. A related study in which an adipose-specific knockout of the insulin receptor was analyzed also demonstrated a significant (approximately 18%) increase in life span in these animals (Bluher et al., 2003). Again, this life span extension was a result of inhibiting insulin signaling in a tissue-specific fashion, as the complete abrogation of insulin signaling is incompatible with life. These results are reminiscent of observations made in C. elegans, in which strong mutations in the daf-2 pathway result in a constitutive dauer phenotype with complete growth arrest, while weaker mutations appear to have the potential to increase life span.

Another interesting example of a long-lived mammalian species is the targeted disruption of the p66shc gene. These mice live 30% longer than control mice and also exhibit an increased resistance to oxidative stress (Migliaccio et al., 1999). The p66shc protein belongs to a family of adaptor molecules that regulate protein-protein interaction for a number of cell surface receptors, including the insulin receptor. Interestingly, cells derived from p66shc-disrupted animals also exhibit a decrease in basal and stress-induced levels of ROS (Nemoto and Finkel, 2002 and Trinei et al., 2002). Since the measured level of ROS represents a balance between oxidant generation and scavenging, alterations in ROS levels can come from either decrease generation or increased destruction. There is some evidence that p66shc may regulate mammalian Forkhead activity and that this in turn may lead to an increase in antioxidants such as catalase and superoxide dismutase (Nemoto and Finkel, 2002 and Kops et al., 2002). In addition, recent evidence that a fraction of p66shc localizes to the mitochondria suggests that this protein may also be involved in regulating mitochondrial oxidant generation, although no direct proof of this currently exists (Orsini et al., 2004).

One other recent study in genetically altered mice provided some of the best support for the previously described vicious cycle of mitochondrial damage leading to aging. These investigators developed a knockin mice that expressed a proofreading-deficient form of a nuclear-encoded mitochondrial DNA polymerase (Trifunovic et al., 2004). Mice expressing this altered polymerase exhibited a significantly higher number of mitochondrial point mutations as well as deletions. Interestingly, these mutations did not significantly increase with the age of the animal, suggesting that they tended to occur at some early point in development. The mice containing the altered polymerase exhibited a significantly shortened life span as well as the appearance of a number of age-related phenotypes, including hair loss, kyphosis, and reduced fertility. There was also a reduction in respiratory chain activity and ATP generation in postmitotic tissue such as heart. All of these biochemical and physiological changes appear, therefore, to be secondary to the initial defect exhibited in the mitochondrial DNA and suggest that damaged mitochondria can accelerate the aging process. The reciprocal experiment involving mice that are genetically engineered to have a reduced rate of mitochondrial DNA mutations has not yet been performed. If these mice were, in fact, capable of living longer, the central role of mitochondrial damage in the aging process would be significantly strengthened.

Finally, in addition to genetically altered mice, examination of normal variations within outbred strains has yielded some important insights concerning the relationship between metabolism and life span. The classic rate of living theory would predict that increased metabolic rate would lead to increased ROS generation and reduced life span. As mentioned earlier, mitochondrial uncoupling represents an important exception to the previously believed positive correlation between metabolic rate and oxidant formation. This notion that inefficiency in mitochondrial ATP generation may be necessary to reduce ROS generation has led to an “uncoupling to survive” hypothesis (Brand, 2000). Support of this comes from a recent study demonstrating that, in an outbred strain of mice, those animals with higher metabolic intensities and oxygen consumption lived longer then animals with lower metabolic intensities (Speakman et al., 2004). These long-lived, high metabolic rate mice also had significant increases in their degree of metabolic uncoupling, suggesting that these animals may decrease ROS generation even in the setting of increased oxygen consumption by reducing the mitochondrial membrane potential.
Conclusions

Some 50 years after its initial proposal, the free radical theory remains perhaps the most vigorous contender to explain the basis of aging in a wide range of species. Certainly, as described here, there are a host of either short- or long-lived organisms that appear to have changes in mitochondrial metabolism, ROS generation, or oxidative stress resistance as their primary alteration. This concurrence in numerous species does not prove causality but does at least hint at a strong underlying relationship. The additional observations that direct oxidant challenge can mimic many of the cellular and transcriptional changes seen with aging also strengthens the link between the level of ROS and the rate of aging. In Harman’s original hypothesis, he suggested that both aging and age-related diseases were regulated by intracellular free radicals. The recent discovery that certain metabolic genes involved in the TCA cycle can act as tumor suppressors (Pollard et al., 2003) and that genes that slow overall aging also slow the development of chronic disease such as atherosclerosis (Napoli et al., 2003) suggest that Harman’s initial intuition may ultimately prove correct. In the end, the seeds of both our immediate vitality and our ultimate mortality would seem to be intertwined in the combustible combination of nutrients and oxygen that continuously occurs within our mitochondria. Economists often warn us that there is no such thing as a free lunch. Biologists are beginning to understand that this maxim appears to apply not only to economic models but equally well to an agreement negotiated billions of years ago between an unsuspecting host and an uninvited intruder.

 

[ThaG]

DNA Repair, Genome Stability, and Aging

Aging can be defined as progressive functional decline and increasing mortality over time. Here, we review evidence linking aging to nuclear DNA lesions: DNA damage accumulates with age, and DNA repair defects can cause phenotypes resembling premature aging. We discuss how cellular DNA damage responses may contribute to manifestations of aging. We review Sir2, a factor linking genomic stability, metabolism, and aging. We conclude with a general discussion of the role of mutant mice in aging research and avenues for future investigation.

Article Outline

Introduction
DNA Damage, Reactive Oxygen Species, and Aging

Reactive Oxygen Species: An Important Source of Age-Related DNA Damage

Cellular Senescence: A Link between Cellular Damage and Aging?

Replicative Senescence Differs between Human and Mouse Cells
Cellular Senescence and Aging

DNA Repair and Aging
The ATM-p53 Axis: A Link between the DNA Damage Response and Aging?

ATM and Aging
p53 and Aging

A Putative Role for NHEJ in Suppressing Aging

NHEJ in DSB Repair and Telomere Maintenance
A Potential Relationship between Ku80 Deficiency, DNA-PKcs Deficiency, and Aging

Potential Roles for Rad50 and BRCA1 in HR and Aging

HR in DSB Repair
BRCA1 in HR and Other Cellular Processes
BRCA1 and Rad50 Hypomorphs Show Aging Phenotypes

Single-Stranded DNA Lesions and Aging

Nucleotide Excision Repair
Some NER Defects Lead to Premature Aging Phenotypes

WRN and Aging
Sir2: A Link between Metabolism, Genome Stability, and Life Span
Conclusions
Acknowledgements
References



Introduction

Although aging is nearly universally conserved among eukaryotic organisms, the molecular mechanisms underlying aging are only beginning to be elucidated. A useful conceptual framework for considering the problem of aging is the Disposable Soma model (Kirkwood and Holliday, 1979). This model proposes that organisms only invest enough energy into maintenance of the soma to survive long enough to reproduce. Aging occurs at least in part as a consequence of this imperfect maintenance, rather than as a genetically programmed process. Although aging may involve damage to various cellular constituents, the imperfect maintenance of nuclear DNA likely represents a critical contributor to aging. Unless precisely repaired, nuclear DNA damage can lead to mutation and/or other deleterious cellular and organismal consequences. Damage to both nuclear DNA, which encodes the vast majority of cellular RNA and proteins, and mitochondrial DNA have been proposed to contribute to aging (Karanjawala and Lieber, 2004). The reader is referred to the review by Balaban et al. in this issue of Cell concerning the potential role of mitochondrial DNA damage in aging (Balaban et al., 2005). Nuclear DNA is an attractive target for aging-related changes since it must last the lifetime of the cell, unlike other cellular constituents, which can be replaced. In addition, the nuclear genome is present at only two to four copies per cell, rendering it potentially very vulnerable to lesions; by contrast, the mitochondrial genome is present at several thousand copies per cell. Many mutants with phenotypes that resemble premature aging possess defects in nuclear DNA repair. Therefore, we focus on the role of nuclear DNA damage in mammalian aging, emphasizing recent work in mouse models.
DNA Damage, Reactive Oxygen Species, and Aging

A large body of evidence argues that DNA damage and mutations accumulate with age in mammals (Vijg, 2000). Cells harboring mutations at defined loci have been shown to increase with age in humans and mice. Cytogenetically visible lesions such as translocations, insertions, dicentrics, and acentric fragments also accumulate in aging mammalian cells. Mice with integrated reporter arrays have allowed estimates of the age-related occurrence of both point mutations and larger genomic rearrangements. Such analyses have revealed considerable variation in mutation spectra between different tissues. These differences likely reflect functional characteristics of those tissues, such as mitotic rate, transcriptional activity, metabolism, and the action of specific DNA repair systems.
Reactive Oxygen Species: An Important Source of Age-Related DNA Damage

There are many sources of DNA damage. In addition to external sources, such as ionizing radiation and genotoxic drugs, there are also cell-intrinsic sources, such as replication errors, spontaneous chemical changes to the DNA, programmed double-strand breaks (DSBs) (in lymphocyte development), and DNA damaging agents that are normally present in cells. The latter category includes reactive oxygen species (ROS), such as superoxide anion, hydroxyl radical, hydrogen peroxide, nitric oxide, and others. Major sources of cellular ROS production are the mitochondria, peroxisomes, cytochrome p450 enzymes, and the antimicrobial oxidative burst of phagocytic cells. ROS can cause lipid peroxidation, protein damage, and several types of DNA lesions: single- and double-strand breaks, adducts, and crosslinks. The situation in which ROS exceed cellular antioxidant defenses is termed oxidative stress. As normal byproducts of metabolism, ROS are a potential source of chronic, persistent DNA damage in all cells and may contribute to aging (Sohal and Weindruch, 1996). The ROS theory of aging is discussed in depth in this issue of Cell by Balaban et al. (2005). In brief, longer-lived species generally show higher cellular oxidative stress resistance and lower levels of mitochondrial ROS production than shorter-lived species. Caloric restriction, an intervention that extends life span in many organisms, likely decreases ROS production (Barja, 2004). In model organisms, many mutations that promote longevity concomitantly increase oxidative stress resistance (Finkel and Holbrook, 2000). In addition, levels of 8-oxoguanine (oxo8dG), a major product of oxidative damage to DNA, accumulate with age (Hamilton et al., 2001).

The potential importance of oxidative damage to DNA in age-related dysfunction is highlighted by a recent study of postmortem human brain tissue (Lu et al., 2004), which found that many nuclear genes involved in critical neural functions show reduced expression after age 40, concomitant with elevated levels of oxo8dG and DNA damage in their promoters. However, the high levels of oxidative DNA damage found by these investigators is at odds with other much lower estimates of the amount of oxidatively damaged DNA in cells (Hamilton et al., 2001). This highlights the difficulty of accurately measuring ROS and oxidative damage experimentally. Additional unresolved issues concerning the ROS theory of aging exist that will require clarification. If ROS are indeed an important source of aging-associated damage, it is currently unclear whether nuclear or mitochondrial DNA is the most relevant functional target (Barja and Herrero, 2000 and Hamilton et al., 2001). Among long-lived mutants, the correlation with oxidative stress resistance is a frequent but not universal one: in some mutants, longevity occurs despite unchanged resistance to ROS or other forms of stress. Additionally, mice heterozygous for a mutation in the mitochondrial enzyme that processes superoxide, Sod2, live out a normal lifespan, despite accumulating higher levels of nuclear and mitochondrial 8oxodG with age (Van Remmen et al., 2003). Thus, while several lines of evidence argue for important role for ROS in aging, many questions remain regarding this hypothesis. In this regard, it is possible that additional mutations may be required to fully unveil potential effects of ROS (see below).
Cellular Senescence: A Link between Cellular Damage and Aging?

ROS and many other DNA-damaging agents can cause cells to enter a state of irreversible cell cycle arrest accompanied by characteristic morphologic and functional alterations, termed senescence (Ben-Porath and Weinberg, 2004). The induction of senescence depends on pathways involving the p53 and Rb proteins (Figure 1). Cellular senescence has been best characterized in cultures of human fibroblasts and mouse embryo fibroblasts (MEFs), which cease expanding after repeated passage in culture, a process termed replicative senescence. Replicative senescence has been employed as a cellular model for aging; many mutations in DNA repair genes that cause premature aging phenotypes also confer premature replicative senescence (Table 1).




Activated p53 induces senescence via a complex gene expression program that includes induction of p21. The role of Rb in senescence involves repression of E2F target genes as well as alterations in chromatin structure. Senescent cells may contribute to aging via depletion of stem cell pools and/or elaboration of factors that interfere with tissue function. Factors elaborated by senescent cells may also stimulate the growth of epithelial tumors (Krtolica and Campisi, 2002).

Table 1.

Features of Selected Models of Premature Aging in the Mouse
Mutant Cellular Process Affected Tissues Affected Increased Rate of Cancer? Accelerated Fibroblast Senescence? Citations
Atm DSB signaling/repair Cerebellum, Gonad, Hematopoietic organs, Thymus Yes Yes (Ito et al., 2004; Shiloh and Kastan, 2001)
Bub1bH/H Spindle assembly checkpoint Bone, Lens, Skin, Gonad No Yes (Baker et al., 2004)
BRCA1Δ11/Δ11/p53+/− DSB repair, other Bone, Eye, Heart, Intestine, Liver, Lymphocytic hyperplasia, Testes +others Yes Yes (Cao et al., 2003b)
DNA-PKcs NHEJ, other Bone, Intestine Yes No (Espejel et al., 2004b)
Ercc1, XPF Nucleotide excision repair, crosslink repair, other Liver (Ercc1, XPF) + Brain, Kidney, Skin, Spleen—Ercc1 No Yes (Ercc1) (McWhir et al., 1993; Tian et al., 2004; Weeda et al., 1997)
Ku80 NHEJ, other Bone, Liver, Skin No Yes (Vogel et al., 1999)
Dysfunctional p53 (p53m/+) DNA damage response Bone, Hair, Lymphoid tissue, Skin No ND (Tyner et al., 2002)
Dysfunctional p53 (p44 Tg) DNA damage response Bone, Testes No Yes (Maier et al., 2004)
PASG/Lsh DNA methylation Bone, Hair, Kidney, Skin, Thymus No Yes (Sun et al., 2004)
PolgA Mitochondrial DNA polymerase Bone, Hair, Heart, Hematopoietic organs, Skin, Testes No ND (Trifunovic et al., 2004)
Rad50S/S DSB repair Hematopoietic organs, Testes Yes No (Bender et al., 2002)
Terc Telomere maintenance Hematopoietic organs, Hair, Heart, Intestine, Myometrium, Skin, Testes Conflicting results Yes (Espejel et al., 2004a; Lee et al., 1998)
Terc/Atm Telomere maintenance/DSB signaling/repair Bone, Brain, Hair, Hematopoietic organs, Intestine No Yes (Wong et al., 2003)
Terc/DNA-PKcs Telomere maintenance/NHEJ, other Intestine, Various No ND (Espejel et al., 2004a)
Terc/Parp-1 Telomere maintenance/ DNA repair Various No ND (Espejel et al., 2004a)
Terc/Ku80 Telomere maintenance/NHEJ, other Intestine, Various No ND (Espejel et al., 2004a)
Terc/Wrn Telomere maintenance/DNA repair Bone, Endocrine, Gonad, Hair, Intestine, Lens, Skin, Spleen Yes Yes (Chang et al., 2004)
Terc/Wrn/Blm Telomere maintenance/DNA repair Bone, Endocrine, Gonad, Hair, Intestine Yes Yes (Du et al., 2004)
TopIIIbeta Topoisomerase Kidney, Lymphocytic infiltrates, Pancreatic islets, Skin, Testes No ND (Kwan et al., 2003)
Wrn DNA repair None No Yes (Lebel and Leder, 1998; Lombard et al., 2000)
XPA/CSB NER, transcription Cerebellum No ND (Murai et al., 2001)
XpdTTD NER, transcription Bone, Hair, Ovary, Skin Yes ND (de Boer et al., 2002)
XpdTTD/XPA NER, transcription Bone, Hair, Skin No ND (de Boer et al., 2002)

Replicative Senescence Differs between Human and Mouse Cells

In many primary human cell lines, replicative senescence occurs secondary to attrition of the ends of the chromosomes, the telomeres, in a process termed “intrinsic senescence” (Itahana et al., 2004). In mammalian cells, the ends of the chromosomes consist of a terminal 3′ single-stranded tail, the G strand overhang, which is buried into adjacent double-stranded repetitive telomeric DNA, forming a protective “t loop” higher-order structure (de Lange, 2002). This t loop is stabilized by a “D loop,” or displacement loop: the region formed between the invading end of the telomere into adjacent double-stranded DNA. The G strand overhang is the substrate for the enzyme telomerase, which employs an RNA template, Terc, to extend telomeres during S phase, thereby countering the natural shortening of telomeres that would otherwise occur with each cell division. Telomerase access to its substrate is, in turn, regulated by telomeric proteins, which also modulate the conformational changes required for telomere replication and subsequent reestablishment of a protective end structure. In many types of human cells, including fibroblasts, telomeres shorten with each successive generation. Critically short telomeres trigger the onset of senescence through a process that may involve loss of the t loop structure and/or loss of protective proteins (referred to as “uncapping”). Such uncapped telomeres are then recognized by the cell cycle checkpoint machinery as DNA damage, leading to cell cycle arrest (Ben-Porath and Weinberg, 2004).

In contrast to primary human fibroblasts, which express very low levels of telomerase activity, MEFs derived from M. musculus possess long telomeres and easily detectable telomerase activity (Itahana et al., 2004). Thus, MEFs ordinarily do not senesce due to telomeric attrition. Instead, senescence of wild-type MEFs occurs primarily in response to oxidative DNA damage incurred during cell culture (Parrinello et al., 2003) in a process termed “extrinsic senescence” (Itahana et al., 2004). These differences between human and mouse cells with respect to senescence are not as great as they may seem. Terc deficiency in mice leads to progressive telomere attrition in successive mouse generations and rapid senescence in embryonic fibroblasts derived from late-generation animals (Espejel and Blasco, 2002). Replicative life span of human cells is altered by culture under different oxygen tensions, and telomere shortening in these cells is itself accelerated by elevated levels of oxidative stress. It is now clear that senescence can be induced by a variety of different types of cellular injury (Figure 1).
Cellular Senescence and Aging

What is the relationship between cellular senescence and aging? The evidence that cellular senescence actually plays a role in aging is correlative: senescent cells accumulate in vivo in mammals with increasing age and at sites of pathology (Itahana et al., 2004), and many mouse and human models of premature aging are accompanied by premature cellular senescence in vitro (Table 1). Of note, late-generation Terc-deficient mice show some signs of accelerated aging (Lee et al., 1998 and Rudolph et al., 1999). Two general models have been proposed to explain how cellular senescence may contribute to aging (Krtolica and Campisi, 2002 and Pelicci, 2004). First, senescence of progenitor or stem cells themselves could impair tissue renewal. In this regard, the Polycomb group repressor Bmi1 appears to control levels of hematopoietic stem cells via negatively regulating the induction of senescence specifically in these stem cells. Second, senescent cells secrete proteases and other factors that may disrupt tissue function. In this regard, senescence has a complex relationship with neoplasia. Senescence has been postulated to occur as a tumor suppressor mechanism, whereby cells that have undergone a genotoxic insult and therefore possess the potential for neoplastic transformation enter a state in which they are incapable of dividing (Krtolica and Campisi, 2002). However, senescent stromal cells can actually promote the growth of epithelial cancers, malignancies that occur with increased incidence in the elderly. Senescence has been offered as an example of “antagonistic pleiotropy,” a process that is beneficial in young organisms but deleterious later in life: senescence suppresses cancer by preventing potentially tumorigenic cells from dividing but may potentially contribute to organ dysfunction in the aged through a variety of mechanisms, perhaps even contributing to neoplasia in this setting (Krtolica and Campisi, 2002). However, a causal relationship between cellular senescence and organismal aging has yet to be proved.
DNA Repair and Aging

The accurate maintenance of nuclear DNA is critical to cellular and organismal function, and therefore, numerous DNA repair systems have evolved. There is some evidence that the intrinsic fidelity and activity of such systems in different species may influence the rate of age-associated functional decline (Hart and Setlow, 1974), although such studies need to be performed with modern methodology. As outlined below, the efficiency of cellular DNA repair machinery itself may decline with age. Many different types of DNA lesions exist. DNA DSBs are repaired via the nonhomologous end-joining (NHEJ) and homologous recombination (HR) pathways, whereas lesions on a single strand of DNA are repaired via base excision repair (BER) and nucleotide excision repair (NER) (and its subpathways). In mice and humans, mutations in certain DNA repair genes lead to phenotypes that, in some respects, caricature aging. For reference, the features of aging as it occurs in wild-type mice are enumerated in Table 2, and the features of mouse DNA repair/metabolism mutants showing some features of aging are listed in Table 1. We now turn to discussing specific gene defects and their relevance to aging.

Table 2.

Common Pathologic Features of Aging in Mice (after Bronson and Lipman [1991]; Cao et al. [2003a])
Hyperplasia/Neoplasia
Adrenal hyperplasia
Angiosarcoma
Harderian gland adenoma
Endometrial hyperplasia
Lung adenoma
Lymphoma
Mammary gland adenocarcinoma
Mast cell tumor
Ovarian cystadenoma
Paraovarian cyst
Pituitary adenoma
Sarcoma
Thyroid follicular cell hyperplasia
Uterine leiomyoma/leiomyosarcoma
Leukocytic Infiltrates
Kidney
Liver
Lung
Mesentery/omentum
Perineurium
Salivary gland
Genitourinary system
Hydronephrosis
Ovarian/testicular atrophy
Seminal vesicle dilation
Renal tubular dilation
Bone
Decreased cancellous bone
Degenerative joint disease
Molar teeth periodontitis
Proliferations in the head/spine
Neurological
Hydrocephalus
Neuronal lipofuscinosis
Radiculopathy
White matter gliosis
Other
Amyloidosis
Fatty change of the liver
Focal myocardial degeneration
Hepatocyte polyploidization
Thymic involution

The ATM-p53 Axis: A Link between the DNA Damage Response and Aging?

Unrepaired or improperly repaired DSBs have serious potential consequences for the cell: cell death, senescence, dysregulation of cellular functions, genomic instability, and, in higher organisms, oncogenic transformation. The initial step in DSB repair is detection of the lesion, and this step involves the ataxia-telangiectasia mutated (ATM) kinase, other related phosphatidylinositol 3-kinase-like kinases (PIKKs), and other proteins (Figure 2; Bassing and Alt, 2004). Activated ATM phosphorylates numerous proteins involved in the G1/S, intra-S, and G2/M checkpoint responses and additionally phosphorylates factors involved in DNA repair (Bassing and Alt, 2004).




An important target of ATM is the p53 protein, which plays a role in the cellular response to numerous genotoxic insults including DSBs. Phosphorylation of p53 by ATM and kinases downstream of ATM is a major mechanism leading to upregulation of p53 levels and activity, although p53 also can be activated via ATM-independent mechanisms. Activated p53 stimulates or represses transcription of many target genes and coordinates checkpoint, senescence, and apoptosis pathways in response to DSBs and other signals. In this regard, p53, through modulation of various downstream targets, triggers arrest/senescence or apoptosis (Meek, 2004). The exact factors that determine the differential outcomes of this complex program are not yet completely elucidated but vary with the cell type, as well as the kind, intensity, and duration of the damage.
ATM and Aging

Perturbations in ATM function can lead to symptoms of accelerated aging. Patients with mutations in the ATM gene suffer from Ataxia-Telangiectasia (AT), a condition characterized by a prematurely aged (progeroid) appearance, immunodeficiency, cerebellar degeneration, and cancer (Shiloh and Kastan, 2001). ATM deficiency in mice recapitulates many of these phenotypes, although the progeroid features of the mouse models are less prominent than in the human disease. Given the many targets of ATM, it is difficult to trace the progeroid appearance of AT patients to a specific function of this protein although several possible mechanisms are conceivable (see below). In this context, the phenotype of ATM-deficient cells may be relevant. In particular, such cells manifest sensitivity to DSB-inducing agents and have marked genomic instability; they also grow poorly and senesce prematurely in culture (Barzilai et al., 2002).

Genomic instability or DNA repair defects of ATM deficiency could contribute to the aging-like features of this disorder (see below). In this context, the premature cellular senescence phenotype of ATM deficiency is rescued by p53 deficiency (Xu et al., 1998), suggesting a role for ATM-independent, p53-dependent checkpoint responses. By extension, such responses also might contribute to premature-aging phenotypes associated with ATM deficiency. Another potential function for ATM in suppressing progeroid phenotypes may be related to reported ATM functions in regulating intracellular ROS levels and sensitivity to these molecules (Barzilai et al., 2002 and Ito et al., 2004). Persistently elevated ROS levels in AT cells might cause chronic damage to DNA and other cellular macromolecules. However, given the lack of an AT-like phenotype of the Sod2+/− mouse, progeroid features of ATM deficiency likely reflect more than increased ROS levels. One possibility would be elevated ROS levels in conjunction with loss of another ATM function, such as ability to respond to ROS-induced DNA damage. Finally, ATM plays an important role in telomere maintenance; AT cells show shortened telomeres and an increased incidence of telomeric fusions (Pandita, 2002). Moreover, mice lacking both ATM and Terc have higher levels of telomeric dysfunction than generation-matched Terc-deficient mutants, as well as proliferative defects in multiple tissues, decreased survival, and clinical evidence of premature aging (Wong et al., 2003). Thus, human AT patients may show aging-like effects, at least in part, as a consequence of telomeric dysfunction, an effect that ordinarily may be masked in the mouse due to the long telomeres of this organism.

The phenotypes of ATM deficiency are influenced by environmental conditions. In the ATM-deficient mouse, the onset of T cell lymphoma can be delayed by treatment with an antioxidant, suggesting that dietary factors impact on the expression of this condition (Schubert et al., 2004). Neurological dysfunction in this model can similarly be ameliorated via antioxidant treatment (Browne et al., 2004). Additionally, the frequency of thymic lymphoma varies among different strains of ATM-deficient mice, a finding that might reflect different background mutations and/or housing conditions and the types of pathogens present (Petiniot et al., 2002). These observations raise the possibility that other aspects of the ATM-deficient phenotype, including the premature aging observed in human patients, may be influenced by environment. This is a theme that we will return to in our discussion of other premature aging models.
p53 and Aging

Various lines of evidence suggest that p53 plays opposing roles in the aging process. While p53 suppresses the onset of malignancy and thereby extends life span, at the same time it promotes cellular senescence and apoptosis in response to DNA damage, potentially contributing to the clinical changes of aging. Thus, p53 function may display antagonistic pleiotropy (Campisi, 2002). The role of p53 in aging cannot be directly tested using p53-deficient mice, as such animals invariably die of malignancy before age-related changes become manifest. However, two mouse strains that express C-terminal p53 fragments along with full-length p53 have been reported to show accelerated aging phenotypes and a lower incidence of malignancy (Maier et al., 2004 and Tyner et al., 2002). It has been proposed that these truncated p53 proteins exert their effects by modifying the activity of endogenous wild-type p53 protein. Only nonphysiologic activation of p53 leads to progeria, as mice expressing extra copies of wild-type p53 under the control of its own promoter do not show signs of premature aging (Garcia-Cao et al., 2002).

Further arguing for a role for p53 in promoting aging, a null mutation in a gene functioning downstream of p53 in the induction of apoptosis, p66Shc, confers oxidative stress resistance and extends mouse life span (Migliaccio et al., 1999). The p66Shc protein shortens murine life span via at least two mechanisms: it increases constitutive intracellular ROS levels, and it promotes cell death in response to oxidative stress. It is unclear how p66shc evolved to play such a role in the cell, since the shorter isoforms of this protein, p52 and p46, fulfill an entirely dissimilar cellular function, transducing signals from tyrosine kinases to ras.
A Putative Role for NHEJ in Suppressing Aging
NHEJ in DSB Repair and Telomere Maintenance

We now turn from a discussion of factors upstream of DNA repair to a consideration of repair pathways themselves and their involvement in aging. NHEJ is one of the two major DSB repair pathways in mammalian cells (Bassing and Alt, 2004). NHEJ is mediated by at least six core factors (Figure 3). Four of these proteins (Ku80, Ku70, Ligase IV, and XRCC4) are conserved from yeast to mammals; they are indispensable for all NHEJ reactions. In contrast, the other two NHEJ factors, DNA-PKcs (DNA-dependent protein kinase catalytic subunit) and Artemis, have evolved more recently and are thought to be required for joining the subset of DNA ends that require processing prior to ligation. Ku70 and Ku80 bind as a heterodimer to DSBs, where they are thought to serve a protective function and to enlist other factors. In this regard, Ku70 and Ku80 recruit DNA-PKcs, and together the three proteins form the DNA-PK holoenzyme. DNA-PK activates Artemis, which functions as an endonuclease to process ends that cannot be directly rejoined. Finally, XRCC4 and Ligase IV, which are likely recruited by Ku, function together to catalyze end ligation itself. NHEJ often occurs concomitant with loss of a few nucleotides at the site of joining.




NHEJ plays a critical role in general DNA DSB repair and, correspondingly, in the maintenance of genomic stability. In addition, NHEJ plays a role in repairing genetically programmed DSBs in the context of antigen receptor variable region gene assembly (V(D)J recombination) in developing lymphocytes. Mice with targeted inactivating mutations in NHEJ genes display phenotypes that reflect loss of these functions (Ferguson and Alt, 2001). All NHEJ-deficient mice suffer from severe combined immunodeficiency as a consequence of an inability to productively rejoin broken V(D)J gene segments in developing B and T cells. NHEJ deficiency is also associated with ionizing radiation-sensitivity and an elevated incidence of spontaneous genomic instability. MEFs deficient for XRCC4, Ligase IV, Ku70, or Ku80 (but not DNA-PKcs or Artemis) senesce prematurely in culture, and mice deficient for these four factors are very small and show widespread neuronal apoptosis during embryogenesis. Premature senescence and neuronal apoptosis, but not small size, are relieved by p53 deficiency. Thus, the former phenotypes occur as a response to, rather than as a direct consequence of, unrepaired DSBs (Ferguson and Alt, 2001). Mice deficient in NHEJ and p53 invariably succumb to pro-B cell lymphomas as a consequence of aberrant repair of V(D)J recombination-associated DSBs in developing B cells, leading to oncogenic translocations (Bassing and Alt, 2004).

In addition to their roles in DSB repair, at least three core NHEJ factors, namely Ku70, Ku80, and DNA-PKcs, localize to telomeres (d'Adda di Fagagna et al., 2001). Deficiency for any of these, as well as for Artemis (Rooney et al., 2003), is associated with an increased frequency of end-to-end chromosomal fusions in MEFs, suggesting a role for these proteins in chromosomal end capping. There is conflicting data on whether NHEJ factors protect against telomere shortening. NHEJ also promotes telomeric fusions in some circumstances. Increased telomere end-to-end fusions are observed in the setting of the telomeric attrition associated with Terc deficiency or inhibition of TRF2, a protein thought to function in end capping, and these end-to-end fusions are eliminated in NHEJ-deficient backgrounds (Espejel et al., 2002a, Espejel et al., 2002b and Smogorzewska et al., 2002). Thus, by ligating uncapped telomeres, NHEJ may actually promote genomic instability. Aside from roles in the NHEJ reaction and in telomere maintenance, Ku70, Ku80, DNA-PKcs, and Artemis have other known or suspected cellular functions, including DSB or checkpoint signaling, whereas XRCC4 and Ligase IV appear to function only in end-ligation during NHEJ. In summary, NHEJ plays crucial roles in general and site-specific DSB repair, in telomere maintenance, and in maintenance of genomic stability.
A Potential Relationship between Ku80 Deficiency, DNA-PKcs Deficiency, and Aging

NHEJ has been proposed to play a causative role in the aging process. DSBs are frequent events in mammalian somatic cells, where they are also very commonly repaired by NHEJ. In addition, genetic studies support the notion that NHEJ plays an important role in the repair of ROS-induced DNA lesions (Karanjawala and Lieber, 2004). Since NHEJ can delete a few nucleotides at sites of DSB repair, this process theoretically could lead to accumulation of mutations and contribute to cellular decline and aging (Karanjawala and Lieber, 2004). Moreover, in the absence of NHEJ, DSBs often are repaired concomitant with large deletions and/or translocations; thus, absent or even decreased levels of NHEJ also might contribute to accelerated aging.

The observation that phenotypes resembling accelerated aging have been described in one strain each of Ku80- and DNA-PKcs-deficient mice potentially supports this model. Thus, a line of Ku80-deficient mice prematurely exhibits age-specific changes including osteopenia, atrophic skin, liver lesions, and shortened life span (Vogel et al., 1999). Likewise, a strain of DNA-PKcs-deficient mice recently has been noted to exhibit age-related pathologies, with osteopenia, intestinal atrophy, thymic lymphoma, and reduced longevity (Espejel et al., 2004b). There are several potential explanations to account for how NHEJ deficiency could cause aging-like phenotypes. As described above, spontaneous DSBs in Ku80- or DNA-PKcs-deficient animals could directly impair cellular function by promoting DNA deletions. In this context, Ku80-deficient mice are found to have a low rate of mutations at a marker locus (Rockwood et al., 2003); such an assay might not detect large deletions or rearrangements, however. Alternatively, unrepaired DSBs could trigger elevated levels of cellular senescence and apoptosis. In this regard, p53-dependent responses are critical in mediating the premature senescence and neuronal apoptosis phenotypes of certain NHEJ mutants (Ferguson and Alt, 2001).

Despite the phenotypes of the Ku80- and DNA-PKcs-deficient mice, several observations suggest that any potential roles for NHEJ in suppressing aging might be more complicated. Ku70- and Artemis-deficient mice thus far have not been reported to show premature aging phenotypes. If Ku70-deficient mice actually lack an aging phenotype, the apparent Ku80-deficient progeroid phenotype must be rationalized in the context of the finding that targeted ablation of Ku70 results in dramatically reduced Ku80 levels (Gu et al., 1997). In addition, aging-related phenotypes thus far have not been reported in several other independently generated DNA-PKcs mutant mouse strains. Ligase IV- and XRCC4-deficient mice show embryonic lethality due to widespread neuronal apoptosis, precluding aging analyses. However, the lack of a consistent aging-like phenotype in other lines of long-lived Ku-, DNA-PKcs-, or Artemis-deficient mice is difficult to explain if NHEJ plays a general role in delaying manifestations of aging (Karanjawala and Lieber, 2004). In this regard, it is conceivable that some aging-related phenotypes might have been missed due to lack of thorough examination of other strains of NHEJ-deficient mice.

Background mutations in various strains of NHEJ-deficient mice, either exacerbating or suppressing the progeroid manifestations of NHEJ deficiencies, might contribute to the apparently discrepant phenotypes of different lines. This possibility, which is relevant to any gene deficiency/aging model, warrants further examination and is discussed in other contexts below. Also, given that both the Ku80- and DNA-PKcs-deficient mice show evidence of ongoing inflammation, potentially indicative of infection, it is conceivable that some aging-like phenotypes might occur directly due to the effects of chronic infection in the setting of immunodeficiency, rather than due to impaired DNA repair. Osteopenia, for example, could result from elevated glucocorticoid levels induced by chronic physiologic stress from infection or malnutrition associated with intestinal atrophy. The degenerative changes in afflicted strains of DNA-PKcs- and Ku80-deficient mice affect only a limited subset of organs. Therefore, either these NHEJ factors are only involved in suppressing aging-related changes in certain tissues or these degenerative changes occur for reasons distinct from those that contribute to aging in wild-type animals. Of note, degenerative changes in these models occur in both highly proliferative (intestine and skin) and relatively less proliferative (bone and liver) tissues, apparently sparing many other tissues of both types. These observations argue against a simple relationship between mitotic status and dependence on NHEJ in the suppression of aging.

In addition to the above considerations, other models for NHEJ factor functions in suppressing aging may be postulated. It is notable that efficiency of DSB repair may decline with age in budding yeast (McMurray and Gottschling, 2003) and in mammalian tissues (Sedelnikova et al., 2004 and Singh et al., 2001), as well as in senescent mammalian cells (Seluanov et al., 2004). Theoretically, such an age-related decline in DSB repair might contribute to increased mutations and genomic rearrangements preferentially near the end of life, although such a decline has not been rigorously proven. Also, loss of other functions of DNA-PKcs or Ku80, such as telomere end capping and/or DNA damage signaling, could contribute to aging phenotypes. In this context, targeted inactivation of either Ku80 or DNA-PKcs, but not in Ligase IV or XRCC4, demonstrates synthetic lethality with ATM deficiency. These observations argue for overlapping functions between Ku/DNA-PKcs and ATM that do not involve classical NHEJ, perhaps related to telomere maintenance (Sekiguchi et al., 2001). Moreover, Ku80/Terc and DNA-PKcs/Terc double deficient mice show exacerbation of intestinal atrophy and other aging-like phenotypes of Terc deficiency (Espejel et al., 2004a). Late generation Terc-deficient mice have an uncharacterized DSB repair defect (Wong et al., 2000); loss of NHEJ in the context of dysfunctional telomeres would be predicted to further compromise DSB repair. In this regard, cells deficient in Ligase IV and ATM or Terc and ATM show high levels of genomic instability and very rapid senescence (Sekiguchi et al., 2001 and Wong et al., 2003), pointing to synergies between DSB repair, DNA damage signaling, and telomere maintenance in genomic maintenance and overall cellular viability.

In summary, there are many potential roles for NHEJ in preventing premature aging-like phenotypes, such as those observed in Ku80- and DNA-PKcs-deficient mice. Current findings suggest, however, that aging phenotypes likely do not result from a failure of DSB repair alone but instead from a loss of other functions or combinations of functions of these proteins, perhaps in concert with environmental factors, such as housing conditions and/or infection and/or background mutations. Such complexities also may likely apply to other DNA repair proteins implicated in the suppression of aging as well, since, as we shall see, many of these proteins play roles in other cellular processes besides DNA repair and, as with NHEJ, only a subset of mutants in any given DNA repair pathway show aging-related phenotypes.
Potential Roles for Rad50 and BRCA1 in HR and Aging
HR in DSB Repair

Two factors with roles in HR, Rad50 and BRCA1 (breast cancer susceptibility gene-1), appear linked to aging-like phenotypes in mouse models. HR is the other major pathway of DSB repair in mammalian cells. Unlike NHEJ, HR uses the sister or (in some cases) the homologous chromosome as a template to repair the broken chromosome.

Whereas recent data indicate that NHEJ functions in DSB repair throughout the cell cycle, HR is largely restricted to late S/G2 (Couedel et al., 2004, Mills et al., 2004, Rothkamm et al., 2003 and Takata et al., 1998). The first step in HR is processing of DSBs by a nuclease to generate 3′ ssDNA tails, which are coated with RPA protein (Figure 4). The MRN complex, composed of the proteins Mre11, Rad50, and Nbs1, is a candidate for this nuclease, although other nucleases are likely involved as well. The MRN complex plays multiple other roles in DSB repair: it functions both upstream and downstream of ATM and has been proposed to promote sister chromatid association and recombination (Stracker et al., 2004). The Rad51 protein, assisted by a number of factors including Rad52, Rad54, BRCA2, and the Rad51 paralogs (XRCC2, XRCC3, Rad51B, Rad51C, and Rad51D), forms a nucleoprotein complex with the DNA and directs the 3′ ssDNA tails to search out, invade, and pair with undamaged homologous sequences. DNA polymerases then carry out repair using the intact DNA as a template. The processes of DNA strand exchange and extension generate Holliday junctions (HJs), structures in which two dsDNA duplexes are intertwined. In mammalian DSB repair, HJs are thought to be resolved primarily via disengagement and gap repair rather than cleavage of the HJ (Valerie and Povirk, 2003).



BRCA1 in HR and Other Cellular Processes

The biology of the BRCA1 protein has proven to be very complex; BRCA1 plays roles in multiple fundamental cellular processes. Biochemically, BRCA1, together with its partner protein BARD1, possesses E3 ubiquitin ligase activity in addition to binding both DNA and multiple other proteins. Functionally, several lines of evidence link the BRCA1 protein to HR (Scully et al., 2004). BRCA1 is a phosphorylation target of ATM and the related PIKK, ATR, following DNA damage induction. BRCA1 forms a complex with Rad51 and BRCA2 (Dong et al., 2003) and can be detected with Rad51 in S phase foci thought to represent stalled replication forks. BRCA1 foci also form in response to ionizing radiation and on chromosomes during meiotic recombination. BRCA1 deficiency leads to impaired HR-mediated repair of chromosomal DSBs, hypersensitivity to many DNA-damaging agents, and genomic instability. The biochemical nature of BRCA1’s involvement in HR remains unclear; one possibility is that BRCA1 may perform a scaffolding function, potentially coordinating the formation of functional repair complexes at DSBs. The E3 ubiquitin ligase activity of BRCA1 may also be relevant in this context; BRCA1 may modify other proteins during HR to alter their functions or direct their degradation. Additionally, BRCA1 has been implicated in many other functions outside HR: transcription, G2/M checkpoint control, chromatin remodeling, and X inactivation (Scully et al., 2004). It has also been proposed that BRCA1 may play a role in NHEJ (Ting and Lee, 2004). Thus, BRCA1 plays roles in numerous cellular processes, including DNA repair.
BRCA1 and Rad50 Hypomorphs Show Aging Phenotypes

Defects in Rad50 or BRCA1 cause progeroid phenotypes. Deficiency of Mre11, Nbs1, or Rad50 is not compatible with cellular survival (Stracker et al., 2004). However, homozygosity for a Rad50 hypomorphic allele (Rad50s/s) permits viability; Rad50s/s mice show a shortened life span, cancer predisposition, and hematopoietic stem cell and spermatogenic failure (Bender et al., 2002). Genomic instability is detectable in cells derived from this animal. The attrition of the hematopoietic and male germ cell lineages occurs in large measure due to p53-mediated signaling triggered by genomic instability (Bender et al., 2002).

Homozyogous inactivation of BRCA1 results in early embryonic lethality (Valerie and Povirk, 2003); however, mice homozygous for a BRCA1 hypomorphic allele and haploinsufficient for p53 (BRCA1Δ11/Δ11/p53+/−) are viable and have many features reminiscent of accelerated aging: wasting, skin atrophy, osteopenia, and malignancy (Cao et al., 2003b). There is compelling evidence that this phenotype results from p53-dependent responses to unrepaired DNA damage (Cao et al., 2003b). Baseline p53 protein levels are higher in BRCA1Δ11/Δ11/p53+/− mice than in p53+/− control animals. Similar to some NHEJ-deficient MEFs, BRCA1Δ11/Δ11 MEFs show increased chromosomal abnormalities and premature cellular senescence, and BRCA1Δ11/Δ11 embryos show tissue SA-β-galactosidase activity, a marker of senescence. Thus, in both the Rad50s/s and BRCA1Δ11/Δ11/p53+/− mice, signs of premature aging occur as a consequence of the p53-mediated responses to unrepaired DNA damage. However, since BRCA1 and Rad50 are both involved in multiple cellular processes, the possibility exists that loss of other functions beyond their DNA repair roles may contribute to these aging-like mutant phenotypes.
Single-Stranded DNA Lesions and Aging
Nucleotide Excision Repair

DNA lesions that affect only one DNA strand are repaired via BER or NER and its subpathways. Although BER is thought to play a critical role in the repair of oxidative lesions, mutations in genes involved in this pathway do not produce aging manifestations: they are either lethal or confer no obvious phenotypes (Hasty et al., 2003). By contrast, lesions in some factors involved in NER can lead to premature aging syndromes in mice and humans. NER is activated by a wide range of helix-distorting DNA lesions, including UV-induced photoproducts, bulky chemical adducts, and certain oxidative lesions. NER can be subdivided into two pathways, global genome NER (GG-NER) and transcription-coupled NER (TC-NER), which differ with respect to the lesion detected and some of the factors involved (Mitchell et al., 2003; Figure 5).




The basic NER machinery consists of the proteins XPA through XPG, the CSA and CSB proteins, and other participants such as the basal transcription factor TFIIH. The GG-NER specific factors, XPC (in a complex with the HR23B protein) and XPE are responsible for detecting helix-distorting lesions that occur throughout the genome. The TC-NER specific factors, CSA and CSB, are involved in repair specifically in transcribed regions and, when RNA polymerase II stalls at a lesion, contribute to displacing the stalled polymerase to create access for repair machinery. Following recognition of the damaged DNA, common NER factors are recruited in both GG-NER and TC-NER. TFIIH, which contains two DNA helicases, XPB and XPD, unwinds the DNA flanking the lesion. The single-strand DNA (ssDNA) binding protein RPA binds to and stabilizes the unwound DNA strands, and XPA aids in lesion recognition. Two structure-specific endonucleases, XPG and XPF (the latter in a complex with the protein ERCC1), then make single-strand incisions on either side of the lesion to release an oligonucleotide. The resulting gap is filled by template-dependent DNA polymerization followed by ligation.

Several core NER factors have been implicated in processes aside from NER. XPB and XPD are critical in RNA polymerase II transcription; CSB associates with RNA polymerases I and II; and XPF/ERCC1 has been implicated in repair of interstrand crosslinks, homology directed repair, and processing of the 3′ G strand overhang at telomeres.
Some NER Defects Lead to Premature Aging Phenotypes

Numerous human patients and mouse strains with defects in different NER factors exist, and some have phenotypes reminiscent of premature aging (Mitchell et al., 2003). Defects in the TC-NER-specific factors CSA or CSB lead to Cockayne syndrome in humans, a severely debilitating disorder with striking progeroid features. CSA- and CSB-deficient mice show much milder phenotypes than their human counterparts (Mitchell et al., 2003). Patients with specific mutations in XPD suffer from trichothiodystrophy (TTD), a disease characterized by photosensitivity, brittle hair, skin defects, and a shortened life span. These patients show defects in transcription of hair- and skin-specific transcripts (and perhaps other mRNAs) (Bergmann and Egly, 2001). In addition, cells derived from TTD patients show NER defects, suggesting that multiple functions of XPD are defective in these individuals. Mice bearing a targeted mutation in XPD that recapitulates a human TTD mutation show similar manifestations to human TTD patients. Additionally, these XPDTTD mice show aging-associated changes such as wasting, scoliosis, osteoporosis, and melanocyte loss (de Boer et al., 2002). XPDTTD/XPA double mutants show a much more rapid degenerative phenotype, suggesting that in the setting of impaired transcription and/or TC-NER, a total lack of NER is extremely deleterious. Similarly, mice deficient in both CSB and XPA also die within a few weeks after birth, although only cerebellar defects have been described in detail in these animals (Murai et al., 2001).

The aging-like phenotypes in human CS patients and in XPDTTD, XPDTTD/XPA, and CSB/XPA mouse mutants may be explained by the fact that a failure to repair lesions in transcribed genes can result in cell death, leading to tissue attrition and aging. Stalled RNA polymerase provides a signal for activation of p53-dependent apoptosis (Ljungman and Lane, 2004). In this regard, it will be of interest to determine whether p53 deficiency rescues the aging-like features of XPDTTD, XPDTTD/XPA, and CSB/XPA mice. Alternatively, the involvement of XPD and the CSB proteins in transcription suggests that impaired transcription of critical genes may play a role in causing these progeroid phenotypes, perhaps interacting with the repair defects in a complicated fashion.
WRN and Aging

Defects in proteins with less well-defined functions in DNA repair can lead to aging phenotypes as well. The human disease WS represents the best model for premature aging in humans (Goto, 1997). WS patients develop premature graying, cataract, loss of subcutaneous fat, skin atrophy, osteoporosis, diabetes, atherosclerosis, and malignancies. WS cells senesce prematurely in culture. The gene defective in WS, WRN, encodes a helicase of the RecQ family (a group defined by its similarity to E. coli RecQ helicase). WRN also possesses an exonuclease domain. WRN plays a role in the maintenance of overall genomic stability, and WRN may be involved in multiple DNA repair pathways (Bachrati and Hickson, 2003).

Recent experiments using mice bearing targeted mutations in WRN provide evidence that, with respect to aging, the most relevant sites of WRN function are the telomeres. WRN-deficient mice do not recapitulate human WS (Bachrati and Hickson, 2003). Based on the observations that human WS cells show telomeric instability (Schulz et al., 1996 and Tahara et al., 1997) and that introduction of telomerase into WS fibroblasts can rescue their premature senescence (Wyllie et al., 2000), it was proposed that WRN-deficient mice may not demonstrate a strong phenotype due to the abundant mouse telomere reserve (Lombard et al., 2000). This hypothesis has been proven correct by the generation of mice deficient in both WRN and Terc (Chang et al., 2004 and Du et al., 2004). In the WRN/Terc double deficient animals, phenotypes reminiscent of human WS that are not observed in either single mutant are present: osteopenia, diabetes, and sarcomas. Phenotypes ordinarily seen in late-generation Terc-deficient animals also occur earlier in the compound mutants and are associated with a greater degree of telomeric dysfunction. Whether or not these phenotypes depend on p53 function is unknown; the premature senescence of human WS fibroblasts is p53 dependent (Davis et al., 2003). The latter observation suggests that cellular checkpoint functions may be involved in producing the clinical features of WS.

The exact role of WRN in telomere maintenance is currently unclear. WRN can be detected at the telomeres in the absence of telomerase function in mammalian cells (Johnson et al., 2001 and Opresko et al., 2004), and Sgs1p, the S. cerevisiae WRN homolog, is required for recombinational telomere maintenance in telomerase-deficient cells in yeast (Cohen and Sinclair, 2001, Huang et al., 2001 and Johnson et al., 2001). WRN also interacts with the telomeric protein TRF2 and can unwind and degrade telomeric D loop structures. These observations suggest that WRN may play a role in providing telomeric access to other factors involved in telomere maintenance (Machwe et al., 2004, Opresko et al., 2004 and Orren et al., 2002). Overall, the phenotype of the WRN/Terc double knockout mouse argues that defective telomere maintenance is an important factor in producing the premature aging-like aspects of WS in humans, although it does not exclude a role for other functions of WRN at nontelomeric sites.
Sir2: A Link between Metabolism, Genome Stability, and Life Span

In the models discussed above, decreased genomic stability is associated with shortened life span. The Sir2 family of proteins provides an example in which increased genomic stability extends life span (Blander and Guarente, 2004). In S. cerevisiae, the chromatin regulatory factor Sir2 (silent information regulator-2) functions as an NAD-dependent histone deacetylase (Imai et al., 2000) to suppress recombination and turns off transcription at multiple genomic loci (Blander and Guarente, 2004). One metric of aging in yeast is the number of divisions that a single mother cell undergoes. The excision and replication of rDNA circles is an important cause of mortality in yeast when life span is measured in this fashion (Sinclair and Guarente, 1997). Loss of Sir2 increases rDNA recombination and shortens life span; whereas an extra genomic copy of Sir2, which increases rDNA stability, extends life span (Kaeberlein et al., 1999).

It has been proposed that Sir2 activity ties the energy status of the yeast cell to longevity (Blander and Guarente, 2004 and Lin et al., 2000). When nutrients are scarce, yeast cells preferentially employ respiration rather than fermentation to generate ATP (Lin et al., 2002). This metabolic switch alters the metabolism of the cell, increasing the NAD/NADH ratio and/or decreasing levels of the Sir2 inhibitor nicotinamide, in turn activating Sir2 and increasing rDNA stability (reviewed in Blander and Guarente, [2004]). Strikingly, overexpression or pharmacologic activation of Sir2 in worms and flies also extends life span (Rogina and Helfand, 2004, Tissenbaum and Guarente, 2001 and Wood et al., 2004). Sir2-driven increased longevity in C. elegans requires the Daf-16 transcription factor (Tissenbaum and Guarente, 2001). Daf-16 is a critical mediator in the insulin-like signaling pathway, normally employed by worms to arrest as extremely long-lived larvae under unfavorable environmental conditions; certain mutations in this pathway confer longevity upon adult worms. The mechanisms by which Sir2 extends life span in flies are currently unclear. Sir2 family members may play a general role in mediating caloric restriction (Sohal and Weindruch, 1996), an intervention capable of extending life span in many different organisms from yeast to mammals (Cohen et al., 2004, Howitz et al., 2003, Lin et al., 2000, Rogina and Helfand, 2004 and Wood et al., 2004), although in yeast the involvement of Sir2 in CR appears to be strain specific (Kaeberlein et al., 2004).

In mammals, there are seven Sir2 family members, designated SIRT1–SIRT7 (Frye, 2000); SIRT1 is the most highly related to S. cerevisiae Sir2. The role of SIRT1 in mammalian longevity has not yet been directly tested, since on a pure strain background SIRT1-deficient animals die very early as a consequence of multiple developmental defects (Cheng et al., 2003 and McBurney et al., 2003). Unlike yeast Sir2, which has no known targets aside from histones, SIRT1 possesses a large and growing list of targets, some of which, including p53 and forkhead transcription factors (mammalian homologs of Daf-16), modulate cellular resistance to oxidative and genotoxic stress (Blander and Guarente, 2004). Additionally, SIRT1, like Sir2, has recently been shown to directly modify chromatin and silence transcription (Vaquero et al., 2004). It is now important to determine whether SIRT1, in addition to silencing transcription, also suppresses recombination and genomic instability via chromatin effects and if so, whether such an activity could be involved in regulating aging in mammals. SIRT1 conditional alleles may allow studies of the role of this protein in aging.

The true mammalian functional ortholog of Sir2, if one exists, might also be a different mammalian Sir2 family member (or members) than SIRT1. SIRT2 and SIRT3 are unlikely to play this role, because these proteins are cytoplasmic and mitochondrial, respectively, rather than chromatin associated (Blander and Guarente, 2004). Thus far, no information has been forthcoming regarding the functions of the four remaining SIRTs, SIRT4–SIRT7, but these remain candidates for proteins that may regulate longevity through genome stabilization.
Conclusions

The hypothesis that nuclear DNA, a critically important cellular constituent that cannot be replaced, is an important target of age-related change is supported by evidence that nuclear DNA damage and mutations accumulate with age. While ROS are likely to be one important source of this damage, there are numerous other cellular and environmental sources of damage, and the impact of such lesions may be enhanced by age-related compromise of DNA repair. In the latter context, most premature aging syndromes are caused by mutations in genes encoding proteins involved in DNA repair (Karanjawala and Lieber, 2004). Accumulation of mutations in critical genes may be one general mechanism by which compromised DNA repair could contribute to aging. In addition, p53-mediated senescence and apoptosis, in response to DNA damage, also likely contribute to aging (Figure 6). Indeed, the fact that lesions in several disparate repair systems cause phenotypes that are broadly similar to one another (Table 1) is consistent with the notion that the specific chemical nature of the accumulated DNA lesions may be less important than their ability to activate the common cellular checkpoint machinery.




It remains unclear why only certain DNA repair mutants in particular pathways show progeroid phenotypes. In some cases, it may simply be that some mutants have not been scrutinized sufficiently to reveal such effects. However, genetic background effects almost certainly play an important role in modifying the aging-like manifestations of DNA repair deficiencies. Also, it must be remembered that many of the DNA repair genes and factors implicated in suppressing aging also play roles in cellular processes other than DNA repair, and therefore aging-like phenotypes might be enhanced by impairment of other cellular functions in conjunction with altered DNA repair. In addition, as discussed above, environmental factors including housing conditions, infectious agents, diet, and many other influences, also likely play a significant role in the expression of aging phenotypes in mouse models and, perhaps, in humans as well. Thus, aging is likely the outcome of a complex interplay between the genetic endowment of an organism and the stresses placed upon it by its particular environment.

All mouse models that link DNA repair to aging possess defects in DNA repair and have shortened life spans. It is important to bear in mind the potential pitfalls of such models (Hasty and Vijg, 2004 and Miller, 2004). Aging encompasses a wide spectrum of degenerative processes, many of which are quite nonspecific, both clinically and pathologically (Harrison, 1994). Thus, it is difficult to arrive at a strict, experimentally useful definition of aging. Factors implicated in organismal decline in genetic models might not play a role in the normal aging processes. A related difficulty is that premature aging models fail to recapitulate all aspects of aging but are instead “segmental progerias” (Hasty and Vijg, 2004 and Miller, 2004); that is, they reproduce in an accelerated fashion some but not all aspects of aging as it occurs in wild-type animals. In this regard, the myriad histopathologic changes of normal aging (Table 2) correspond poorly with the changes that occur in models of premature aging (Table 1). Mammalian aging is not likely a single process but rather the decline of many somatic functions, heavily influenced by the environment; this is a complex interplay that is extremely difficult to model accurately. For these reasons, genetic models of extended life span are likely to be more informative than models with reduced longevity with respect to physiologically relevant causes of decline and mortality; in such long-lived organisms, life span-limiting factors must of necessity be counteracted.

There are many outstanding questions regarding the connection between DNA repair and aging that will benefit from the application of emerging techniques in molecular biology and genetics. In models of premature aging, the most vulnerable system fails first, leading to death and precluding gain of insights from effects on other, potentially more relevant, organ systems. This difficulty might be overcome with hypomorphic alleles or via conditional knockouts of relevant genes. The latter approach, for example, might allow an evaluation of the roles of genes essential during embryogenesis in the aging process. Such an approach also could permit insights into the types of DNA lesions that are most important in causing aging in different tissues by allowing impairment of specific DNA repair pathways in defined cell populations. For example, tissue-specific deletion of DNA damage-response/checkpoint genes in DNA repair mutants showing evidence of premature aging might allow the direct effects of these DNA lesions to be distinguished from the cellular responses to them. Further insights into the basic biology of DNA repair proteins, particularly regarding the exact nature of the DNA lesions that these repair systems respond to and their roles in checkpoints, may also shed more light on the role of various DNA lesions in contributing to aging.

Eventually, the importance of DNA damage in aging might be directly tested via the generation of experimental organisms with enhanced efficiency of DNA maintenance, which would be predicted to show retarded aging. Although the plethora of different repair systems makes this a challenging, if not impossible, undertaking, the existence of Sir2, a regulator of genome stability and aging in yeast, offers encouragement for those seeking similar global regulators of genome maintenance and potentially DNA repair in mammals.

 

[ThaG]

The Future of Aging Therapies

Advances in understanding aging processes and their consequences are leading to the development of therapies to slow or reverse adverse changes formerly considered to be “normal” aging and processes that underlie multiple age-related conditions. Estimating the effectiveness of candidate aging therapies, whose effects on human aging may require many years to determine, is a particular challenge. Strategies for identifying candidate interventions can be developed through multiple approaches, including the screening of molecular targets and pathways in vitro and in animal models, informed as well by evidence from human genetic and epidemiologic data. A number of recently established programs and networks can serve as resources for such research. For all these research approaches, from in vitro molecular studies to clinical trials, contributions of cell and molecular biology are crucial and offer the prospect of therapeutic advances that address fundamental biological processes as well as the clinically important challenges of aging.

Article Outline

Main Text

Introduction
Issues and Resources for Developing New Human Aging Therapies

Identification and Evaluation of Potential Therapies
Ascertainment of Adverse as well as Beneficial Effects of Interventions
Interactions among Intervention Development Activities and Resources

Specific Therapies

Telomere Maintenance

Caloric Restriction

Interventions against Age-Related Vascular Stiffening
Alzheimer’s Disease

Conclusion

References



Main Text
Introduction

Considered most broadly, the concept of aging therapies includes prevention and treatment of the variety of conditions whose rate of occurrence increases with age. Progress against many of these conditions has been ongoing for several decades. Most of this has stemmed from research on specific age-related conditions rather than on aging per se, such as clinical trials showing effective means of preventing cardiovascular disease and diabetes in older persons. It is likely that progress on these and other age-related conditions will continue. This review deals principally with therapies focused specifically on aging processes, specifically those that contribute to disease or dysfunction. The pronounced increase with age in incidence of most conditions causing morbidity and mortality, as well as long-term increases in the proportion of individuals surviving to ages when the prevalence and severity of such conditions are high, highlight the potential health benefits of better interventions to modulate such aging processes.

Aging therapies may be aimed at reversing age-related changes once they have occurred or at decelerating or preventing aging changes starting at various points in the life span. A therapy that is effective for one of these purposes may be ineffective or even counterproductive for the other. An intervention started before old age might slow a process that resulted in a particular pathology in late life but have adverse effects in individuals in whom that pathology is already present. This, for example, has been suggested as a possibility for effects of estrogens in regard to atherosclerosis and coronary heart disease.

Two therapeutic foci are of particular interest:

Changes with age previously considered to be “normal aging” but that have adverse consequences. Although many age-related changes are asymptomatic and are found in many or all persons, they are not necessarily benign. As understanding of adverse effects of a “normal” aging change progresses, the change may be recognized as a disease or risk factor. A familiar example is osteoporosis, which results in most cases from progressive bone loss starting in middle age. Its relation to fracture risk has been recognized for several decades, and insights into its pathophysiology have led to a variety of therapies. An analogous evolution has occurred recently regarding age-related loss of muscle mass: disability and other adverse conditions have been linked to muscle loss of sufficient severity, for which the term “sarcopenia” is now applied. Similarly, as discussed below, there is increasing evidence that an age-related increase in vascular stiffness may be a risk factor for cardiovascular disease and that therapies may be developed to enhance vascular compliance and reduce cardiovascular mortality.

The importance of identifying adverse effects of age-related changes and developing therapies that retard or reverse such changes is highlighted by the fact that, for many age-related conditions, currently known risk factors only explain a modest proportion of the risk. For example, chronologic age remains the strongest single risk factor for cardiovascular disease, presumably reflecting effects of specific age-related changes whose role has not yet been appreciated and which may be modifiable once identified. Thus, new therapies to counter aging changes whose role in disease risk is identified could, when added to existing therapies, reduce risk for age-related conditions (or delay their onset) much more than is currently possible.

Fundamental mechanisms responsible for multiple age-related pathologies. Several underlying processes, many discussed elsewhere in this issue, that may regulate multiple aging changes have been proposed and investigated. These include damage and repair of macromolecules and other tissue components; regulation of cell proliferation, differentiation, and death; control of cellular bioenergetics; and control of genomic stability. The potential health benefits of an intervention that modulates a mechanism that affects the development of multiple adverse changes with age have frequently been noted. In addition to the inherent efficiency of a single intervention with multiple benefits, this type of intervention could avoid problems of adverse effects of a therapy for one age-related condition on other age-related conditions (e.g., exacerbation of high blood pressure by nonsteroidal anti-inflammatory drug treatment for arthritis).
Issues and Resources for Developing New Human Aging Therapies
Identification and Evaluation of Potential Therapies

Particularly for interventions whose effectiveness may depend on starting early in life, a fundamental issue for the development of aging therapies is the long time scale of human aging. Many proposed mechanisms of aging are postulated to act over the life span, so that effects of interventions that decelerate aging changes might be expected to diminish as the age when they are begun increases. This implies that testing of some human interventions may need to begin at an early age and continue for a long time. Even interventions started in late life to reverse age-related changes may take several years to show effects in humans. Thus, selection of aging therapies for testing long-term effects in humans is especially dependent on shorter-term evaluation of potential interventions. Such strategies include the following:

Identifying pathways regulating aging changes and potential molecular targets for interventions. Choices of potential therapeutic targets can be informed by evaluating genetic effects or interventions in short-lived organisms such as yeast, nematodes, fruit flies, and mice. During the past 10 years, many genetic changes as well as interventions that increase longevity in these animal models have been reported and may provide a basis for developing promising human interventions. Although genetic manipulations may not be directly applicable to humans, results of these experiments are helpful in identifying pathways for possible pharmacological intervention.

Short-lived nematodes, fruit flies, and mice have been particularly useful for identifying genes that influence longevity. Because this area has been extensively discussed by Cynthia Kenyon in this issue of Cell (Kenyon, 2005), we will only briefly mention the most salient results. A striking observation is that major effects on life span have been achieved by mutating genes that affect the signaling pathways related to insulin-like factors (Liang et al., 2003 and Tatar et al., 2003). Significant effects have also been achieved from genetic manipulations that: (1) reduce ROS generation by mitochondrial electron transport (Feng et al., 2001, Knauf et al., 2002 and Ishii et al., 2004); (2) reduce food consumption (Miskin and Masos, 1997 and Lakowski and Hekimi, 1998); (3) increase expression of antioxidant genes (Larsen, 1993, Mitsui et al., 2002 and Ruan et al., 2002), heat shock genes (Munoz, 2003), or sirtuins (Tissenbaum and Guarente, 2001); or (4) modify the apoptotic (Migliaccio et al., 1999 and Cohen et al., 2004) or proliferative (Vellai et al., 2003) potential of cells. A limited number of potential pharmacological interventions have also been identified including antioxidants (Melov et al., 2000) and compounds that modulate protein deacetylase activity (Howitz et al., 2003 and Wood et al., 2004).

Though human intervention studies based on factors identified in animal models are, for the most part, not likely to start until well into the future, in vitro studies of human cells provide another means to identify especially promising pathways and targets for human interventions. This approach has not been used extensively to date but has considerable potential. Though such studies do not directly test the effects of interventions on human aging, they do allow testing effects of genetic factors and interventions on specific functions (e.g., oxidative stress resistance or insulin sensitivity) in human cells, including tests in specific cell types, could provide clues to their in vivo effects. These in vitro approaches may be amenable to creative and high-throughput strategies. In addition, comparative studies on cells from young and old human donors may reveal age-related differences in metabolic or signaling pathways that may suggest targets for interventions to reverse or prevent such changes.

Recently, genetic epidemiologic studies of human longevity have begun to provide another means of identifying possible pathways and molecular targets. For example, such studies have identified a relationship between exceptional longevity and variants of several genes affecting lipoproteins (Schachter et al., 1994, Barzilai et al., 2003 and Geesaman et al., 2003). The products of these genes may serve as targets for interventions.

New technologies are facilitating high-throughput strategies to identify potential therapeutic targets and promising interventions in very short-lived species. Using a dye-based assay for cell death, Gill et al. (2003) have developed a high-throughput assay for interventions that delay death in nematodes. Using an RNA interference screen of 5690 nematode genes, a large number of genes affecting mitochondrial function were found also to affect life span (Lee et al., 2003). Bauer et al. (2004) have developed a rapid method to identify interventions that increase longevity of fruit flies, based on the fact that mutations that change life span in fruit flies also change the timing of gene expression patterns. Because longer-lived flies express lethal reporter genes later than control flies, flies with normal aging rates can be killed off before flies with extended longevity, so these long-lived flies can be enriched in the population. This assay has also been successfully used to test pharmacological interventions as well as response to environmental conditions, showing, for example, that resveratrol and lipoic acid but not vitamin E can increase fruit fly longevity. In addition, after identification of molecular targets modulating aging changes, multiple compounds that could affect these targets can be screened. This approach has been applied to identify sirtuin activators that increase life span and decelerate aging in S. cerevesiae (Howitz et al., 2003), C. elegans, and D. melanogaster (Wood et al., 2004). It should be noted that NIH’s Roadmap Initiatives (http://nihroadmap.nih.gov/) include efforts to establish technologies and infrastructure for high-throughput screening of molecules that impact specific metabolic pathways. These efforts will greatly facilitate aging research in this area.

Laboratory mammalian studies testing intervention effects on outcomes. Interventions for testing in mammals can be selected based on findings from the types of invertebrate studies described above, physiologic observations in mammals themselves, or epidemiologic data suggesting an effect of a particular factor in humans. Mammalian intervention studies have been conducted on the effects of interventions in abnormally short-lived strains with specific conditions or genetic defects as well as in long-lived wild-type strains. The latter type of study is widely accepted as preferable for studies to examine potential human applicability, and the following discussion is confined to such studies.

The rapid discovery rate of findings from very short-lived invertebrates, implicating potential therapeutic targets for longer-lived species, offers challenges as well as opportunities for mammalian studies: mammalian studies of effects on life span and on other responses that require observation over an appreciable percent of the life span require much more time than analogous invertebrate studies. Hence, it is likely that the fraction of candidate interventions suggested by invertebrate studies that can actually be tested in mammals will be quite limited. Expanding mammalian model testing could help to reduce this discrepancy. Strategies to help refine the selection of interventions suggested by in vitro and invertebrate studies for mammalian testing can also help. For example, it is likely that many factors with positive effects on life span operate on the same or convergent pathways (as has been found for the multiple longevity-related genes that affect the insulin-like peptide signaling pathways), and sorting these factors into a limited number of pathways will facilitate finding the most promising points on these pathways for targeting preclinical and clinical interventions. In addition, more information on functions of homologous proteins in invertebrate models and in mammalian models for therapeutic aging studies could help to distinguish those targets for which mammalian interventions are likely to have positive effects from those that affect pathways that are important in invertebrates but not in mammals.

Few mammalian life span studies have tested interventions to modulate specific processes implicated in the invertebrate studies discussed above. Antioxidant studies have not shown substantial life span effects (Golden et al., 2002), with the exception of moderate extension in rodents by nitrone free radical spin-trapping agents, though evidence suggests that this is due to inhibition of specific redox-sensitive signal transduction processes related to inflammation, rather than reduction in free radicals per se (Floyd et al., 2002). In studies of agents affecting mitochondrial function, neither α-lipoic acid (which has antioxidant effects) nor coenzyme Q10 (which increases electron transport and has antioxidant properties) affected life span in mice (Lee et al., 2004).

As information on pathways and molecular targets for interventions that may affect longevity accumulates, the number of promising interventions for mammalian testing will likely increase. Recently, NIA’s Interventions Testing Program (discussed below) has initiated four mouse intervention studies testing effects on longevity of two anti-inflammatory agents (aspirin and nitroflurbiprofen), 4-hydroxy-phenyl-N-tert-butyl nitrone (a free radical scavenger), and nordihydroguarietic acid (a lipoxygenase inhibitor with structural similarities to the sirtuin activator resveratrol).

In addition to life span studies, there have been shorter-term rodent studies on interventions to reverse adverse aging changes by modulating the effects of putative aging mechanisms. There have been reports of reduction of age-related declines in cognition by compounds with antioxidant properties (reviewed in Golden et al. [2002]). There also have been rodent intervention studies of the effects of agents that modulate mitochondrial function on aging changes: acetyl-l-carnitine administration to old rats increased tissue oxygen consumption and mitochondrial membrane potential and reversed age-related declines in cognition and physical activity (Ames, 2004).

Rodent model systems have limitations for evaluating the likely effect of interventions on humans, because of the considerable differences in age-related pathologies between rodents and humans. Nonhuman primates have greater similarities, but, other than three ongoing studies of effects of caloric restriction on aging and life span (Roberts et al., 2001), they have not been used in intervention studies of aging therapies. The development of other animal models of human aging pathologies may be especially valuable for experimental therapeutic studies.

Short- or medium-term human studies of intervention effects on factors affecting risk for subsequent aging outcomes. When there is sufficient evidence to support human testing of interventions, it is necessary to decide the age of the population to be studied, the duration of administration of the intervention, and the outcomes to be measured. If the intervention is expected to reverse in older persons adverse aging changes that increase their risk of clinical outcomes, it may be possible to design a study to measure the effect of the intervention on these outcomes directly. The longer the interval needed to observe such an effect, the more challenging this task becomes. However, clinical trials with treatment periods of up to 8 years have been successfully organized, and longer trials may be possible.

If, however, the expected effect of the intervention is to decelerate aging changes, requiring that it must be initiated relatively early in life and sustained for decades before clinical effects occur, the logistical challenges of conducting a trial long enough to detect these effects are extremely formidable. Except perhaps in the case of the most promising interventions with abundant preliminary data, alternative designs to determine shorter-term effects of the intervention can serve as an intermediate step to screen potential interventions for longer-term testing. Such studies can be conducted in different age groups to determine age-related differences in response and determine, for example, whether there is a maximum age above which initiating the intervention has no effect.

These studies can examine effects on known risk factors for clinical outcomes (e.g., bone density and cholesterol), including effects on rates of age-related changes in these factors. They can also determine effects on physiologic functions affected by interventions that delay or reverse aging changes in laboratory animals. In addition, there have been attempts to identify “biomarkers of aging” that could predict very long-term consequences of aging therapies in humans, including effects on longevity (Miller, 2001 and Butler et al., 2004). In mice, early life predictors of longevity have been identified (Harper et al., 2004 and Warner, 2004). Though several such human biomarkers have been proposed, these have not been validated by longitudinal epidemiologic data or other evidence regarding their long-term predictive value. However, analyses of data from very long-term longitudinal studies may identify such markers.

In both human and laboratory animal intervention studies, it is particularly useful to compare the effects of the intervention on aging outcomes with its effects on its molecular targets in vivo. Such comparisons can help in refining hypotheses about aging mechanisms as well as in guiding future experimental therapeutic studies. For example, if the in vivo molecular effects of an intervention are consistent with expectations but its clinical or physiologic effects are not, this may cause reconsideration of the role that the targeted pathway actually plays in the aging outcomes being studied. In one specific example, following considerable observational evidence that oxidized low-density lipoprotein (LDL) played a significant role in progression of atherosclerosis, a human α-tocopherol intervention study found that, although α-tocopherol administration indeed markedly reduced LDL oxidation, it had no effect on atherosclerosis progression (Hodis et al., 2002).
Ascertainment of Adverse as well as Beneficial Effects of Interventions

A critical issue regarding the identification of therapeutic targets in animal models is whether longer life span associated with genetic variants or experimental manipulations is accompanied by maintenance of healthy function during this extended lifetime. Possibilities to evaluate this in invertebrates are limited by scarcity of information on their pathologies. However, the Indy mutation in fruit flies increases longevity “without a concomitant reduction in resting metabolic rate, flight velocity, or age-specific fecundity” (Marden et al., 2003). The age-1 and daf-2 mutations extend nematode longevity and slow tissue degeneration (Garigan et al., 2002 and Herndon et al., 2002) and increase resistance to bacterial infection (Garsin et al., 2003 and Lithgow, 2003). However, not all the effects of all factors that extend life span are beneficial. For example, although the long-lived Ames dwarf mouse has lower incidence of cancers (Ikeno et al., 2002), it also has developmental abnormalities (Carter et al., 2002) and impairments in immune function (Esquifino et al., 1991).

For interventions in human studies and mammalian models, attention to adverse effects as well as to beneficial effects on aging, is crucial. Few interventions have unmitigated benefits; it is unlikely that aging therapies will be an exception. Though human studies suggest that factors contributing to exceptional longevity also contribute to exceptional health over the life span (Hitt et al., 1999 and Atzmon et al., 2004), potential adverse effects of any putative human intervention need to be explored thoroughly. Studies of effects of interventions in laboratory mammals and human cells in vitro not only can detect pathologic effects but could dissect the mechanisms responsible for them from those that produce benefits and could lead to refined interventions that slow or reverse adverse aging changes while causing minimal risk.
Interactions among Intervention Development Activities and Resources

As discussed above, the multiple approaches to developing clinical aging interventions interact with each other in several ways. This is illustrated in Figure 1. To date, such interactions have mainly been informal, depending on spontaneous collaboration and information sharing among scientists. In the future, there may be benefits from more structured ways to encourage communication and collaboration on promising lines for intervention development.



One means by which this can be done is by facilitating interactions and collaborative activities among investigators working with diverse approaches and in diverse model systems and by providing the resources relevant to their work in intervention development (see Figure 1): the NIA Longevity Assurance Genes Interactive Network, involving mainly studies in a wide range of laboratory animal models, continues to make important contributions to identifying possible molecular targets for interventions (Warner, 2003). The large number of genes influencing life span identified by this network is likely to continue to expand rapidly as high-throughput screening techniques are applied.

To promote rigorous controlled testing in mammals of interventions that appear particularly promising based on available data, the National Institute on Aging has begun an Interventions Testing Program for interventions that may hold promise for future human studies (Warner et al., 2000). In this program, candidate interventions are evaluated in terms of the likelihood of efficacy, and those selected are tested in replicate mouse studies in three laboratories using standardized diet and environmental conditions. Information about the program, including the process for proposing and selecting interventions, is available at http://www.nia.nih.gov/ResearchInformation/ScientificResources/InterventionsTestingProgram.htm.

More recently, the Longevity Consortium, involving large-scale human population studies working in collaboration with researchers in the biology of aging, has been established to identify polymorphisms affecting longevity, including variants in homologs of genes that have effects on longevity in laboratory animals (Hadley and Rossi, 2005). Additional information on the consortium is available at http://www.longevityconsortium.org/. This research should be useful in identifying or confirming targets or pathways for intervention. NIA has also begun a program to develop and screen compounds for new drug targets for age-related conditions, presently focused on neurodegenerative diseases, diabetes, congestive heart failure, and oncologic and immunologic diseases.

Identification of long-term predictors of exceptionally healthy aging, which could be used in short- and medium-term human intervention studies, could be facilitated by the recently organized NIA Longitudinal Data on Aging Working Group, comprising researchers from numerous epidemiologic studies, which is examining possibilities for enhanced uses and analyses of longitudinal data in aging research. The report of the Working Group’s first meeting is available at http://www.nia.nih.gov/ResearchInformation/ConferencesAndMeetings/.
Specific Therapies

This section discusses selected examples that illustrate the considerations discussed in preceding sections and also illustrate how development of new therapies is stimulated by different types of preliminary evidence, i.e., by biologic findings implying a cellular structure’s role in aging (telomeres), by laboratory animal intervention studies (caloric restriction), and by investigations of the consequences of “normal” aging changes (vascular stiffening).
Telomere Maintenance

Critical observations by Hayflick and collaborators in the 1960s demonstrated that replicative capacity of normal somatic cells is finite and that, after a critical number of cell divisions, cells reach a state in which further division cannot occur, a state termed replicative senescence (Hayflick, 1965). This observation in turn raised the question of what biological process provides the counting mechanism that monitors cell division and underlies replicative senescence. Research over the past 15 years has provided an intriguing answer to this question in identifying the role of telomere function in regulating cell replication and senescence (reviewed by McEachern et al. [2000]). DNA replication during mitosis is incomplete and results in loss of 50–200 terminal bp per cell division. This loss of DNA occurs in telomeres, the tandem hexanucleotide repeats that constitute the ends of linear chromosomes. When cumulative loss of telomeres results in critical shortening, the “capping” function of telomeres is compromised, and recognition of this altered telomere structure results in cell senescence and/or apoptotic death. A pivotal discovery was the identification of telomerase, an RNA-dependent DNA polymerase that is capable of synthesizing terminal telomere repeats and extending the length of telomeres, thus compensating for the telomere loss that occurs with cell division. The potential implications of these findings for human aging and for clinical intervention have provoked substantial interest.

The role of telomeres in regulating the capacity for cell division, and the ability to extend cell division by intervening to manipulate telomere length and function have been well established. Cultured human cells reach replicative senescence after cell division in culture and generally express little or no detectable telomerase activity by the time that they are approaching senescence. Induction of telomerase activity in normal human somatic cells has been accomplished by transduction of telomerase genes and has been shown to result in the “immortalization” of these cells, as defined by an apparently unlimited capacity for cell division (Bodnar et al., 1998). An important question now under study is whether these immortalized cells retain their normal function or are altered in clinically relevant respects, such as loss of normal function or increased risk of malignant transformation.

What is the evidence that altered telomere length or function has consequences for multicellular organisms in vivo? Direct and compelling evidence comes from seminal experiments in which telomerase activity is eliminated by inactivation of telomerase genes. These experiments have been carried out in a number of species, but of most apparent clinical relevance are the studies carried out in knockout mice as a mammalian model (Blasco et al., 1997). During successive generations of breeding, telomeres progressively shorten, resulting in sterility due to the failure of both female and male germ cell lineages. Mice that have reached these late generations of telomere shortening also show a number of abnormalities similar to some if not all aspects of premature aging (Rudolph et al., 1999).

What, then, is the evidence that altered telomere function plays a role in normal human aging or in clinical conditions? There is considerable indication that telomere shortening occurs in multiple cell lineages during normal human aging and that accelerated telomere shortening can occur in specific cell types and under disease-associated circumstances. These lines of evidence are consistent with but do not prove the existence of a causal relationship between telomere shortening and disease- or age-related changes in humans. In this regard, two recent findings are of particular note. In a study of men and women aged 60–97, it was observed that telomere length in peripheral blood cells is related to subsequent survival, with long telomeres associated with significantly lower risk of death (Cawthon et al., 2003). This correlation is intriguing, although it is not a direct demonstration that telomere length is a direct determinant of survival. The alternative possibility consistent with these findings is that genetic or life course events affect survival and that telomere length is affected in parallel but is an epiphenomenon rather than a causal determinant of health or survival. Perhaps the most compelling evidence for the existence of such a causal relationship is the inherited disease dyskeratosis congenita (DKC), the autosomal dominant variant of which is caused by a mutation in the telomerase RNA template gene and the X-linked variant of which is caused by mutations in the telomere-associated protein dyskerin (Mitchell et al., 1999 and Vulliamy et al., 2001). DKC is marked by abnormally short telomeres, defects in epithelial and hematopoietic cell lineages, and hematopoietic malignancies. These findings suggest that telomere length or telomerase activity can affect human health.

What are the strategies by which existing information could be translated into therapeutic manipulations of telomeres? The most accessible approaches at present appear to be those in which specific human cells would be manipulated ex vivo, for example, by telomerase gene transduction to express telomerase, maintain or increase telomere length, and therefore to have extended capacity for cell division. Cells so modified could then be administered to the same individual under circumstances in which cell numbers or potential for cell division would otherwise be insufficient to serve the desired outcomes. Such approaches will require close attention to potential complications such as uncontrolled division or altered function of modified cells.
Caloric Restriction

Chronic caloric restriction (CR) (“undernutrition without malnutrition”), when started in early life or adult life, substantially extends life span in rodents as well as in multiple invertebrate species. The fact that this single intervention slows multiple age-related changes, delays the onset of cancer and multiple other age-related pathologies, and extends life span is consistent with (but does not prove) the idea that one or a few mechanisms, modulated by caloric restriction, control the rate of multiple aging changes and may be potentially controllable by other interventions as well. Chronic CR in nonhuman primates has been found to produce parallel physiologic changes to those seen in rodents; studies of its effects on nonhuman primate life span have not been completed. Despite extensive research on caloric restriction’s effects on aging, we still do not understand the mechanism(s) of its effects nor their implications for humans (Hadley et al., 2001, Roberts et al., 2001, Hursting et al., 2003 and Koubova and Guarente, 2003).

There is also evidence that periodic food deprivation in mice, produced by every-other-day intermittent feeding, may induce similar physiologic effects over a period of weeks to those of caloric restriction, even when average daily intake is minimally different from ad libitum intake (Anson et al., 2003). This finding has potentially very important implications. However, a life span study in which caloric intake on an intermittent feeding regimen is documented to equal ad libitum intake and produce equivalent effects to CR’s has yet to be reported.

Molecular and physiologic mechanisms proposed to explain CR’s effects on life span are essentially coextensive with those proposed to regulate life span (reviewed in Masoro, 2000, Koubova and Guarente, 2003 and Ingram et al., 2004), including reduction of oxidative damage, increased metabolic efficiency, increased genomic stability, decreased apoptosis, increased apoptosis, lowered glucose levels, slowed glycolysis, lower insulin levels, lowered protein glycation, decreased body temperature, neuroendocrine responses, and sirtuin activation.

Exploration of potential human implications of the extension of life span by CR has taken two general directions. One derives from the search for the mechanisms mediating its effects in laboratory animals, which could be used to identify “CR mimetic” interventions that produce CR’s effects in the absence of caloric restriction (Hadley et al., 2001). Since almost every mechanism proposed to modulate aging changes has been proposed to modulate CR’s effects on aging, in the broadest sense, almost all studies of interventions that modulate these mechanisms could be considered CR mimetic studies. In addition to such studies discussed earlier, there have been recent CR mimetic intervention studies in rodents focused on glucose metabolism, using the glycolysis inhibitor 2-deoxyglucose and the hypoglycemic agent metformin (Ingram et al., 2004).

Another potential line of research on human implications of CR’s effects on aging is controlled human trials. For reasons discussed above, medium-term studies of its effects are advisable before considering longer-term studies. NIA is supporting randomized controlled pilot trials of CR, collectively known as CALERIE (Comprehensive Assessment of Long-term Effects of Restricted Intake of Energy Intake). CALERIE will test effects of 2–3 years of caloric restriction (20%–30% reduction) in young and middle-aged nonobese persons. CR will also be compared with exercise interventions that produce weight loss to determine how their effects differ. Major goals of CALERIE are to determine whether this magnitude and duration of CR is feasible and safe, to compare human physiologic and cellular responses with effects reported in laboratory animals, and to determine its effects on disease risk factors and predictors of longevity derived from human studies. CALERIE is encouraging proposals for ancillary studies on mechanisms underlying CR’s physiologic effects. Information about CALERIE, including ancillary study policies, is available at http://calerie.dcri.duke.edu/.

If findings from CALERIE indicate feasibility and safety, a trial of a longer CR intervention may be worthwhile, particularly if CALERIE interventions have favorable effects on known disease risk factors. Prospects for long-term CR studies may be enhanced by the rapidly increasing understanding of the regulation of food intake and the possible ensuing development of better anorexiants (Flier, 2004).
Interventions against Age-Related Vascular Stiffening

One of the most salient age-associated changes in large arteries is an increase in intimal-medial thickness, which is accompanied by an increase in stiffness. A growing body of epidemiologic evidence indicates that these changes, and the ensuing increase in systolic and pulse pressure, formerly thought to be part of “normal” aging, precede and predict a higher risk for developing clinical atherosclerosis, hypertension, and myocardial infarction or stroke (Lakatta and Levy, 2003). In other words, aging blood vessels provide a milieu in which vascular diseases can flourish. These vascular changes that accompany aging in persons who do not have a diagnosis of clinical cardiovascular disease, however, largely have remained outside the bailiwick of clinical medicine and have not been the focus of preventive measures. However, if such “risky” arterial aging leads to arterial disease, it offers potential targets for treatment and prevention, involving modification of lifestyle or pharmacotherapy, to retard its rate of progression and delay or prevent clinical disease.

Cellular and molecular mechanisms that underlie arterial intimal medial thickening and stiffening with advancing age may provide targets for such interventions. These include elevated levels or activity of molecules such as matrix metalloprotease-2 (MMP2), angiotensin II (AngII), transforming growth factor β, monocyte chemoattractant protein-1 (MCP1), interstitial cell adhesion molecule-1, and NADPH oxidase. Each of these factors is a signaling target downstream of the angiotensin AT1 receptor. AngII increases MMP2 activity in the aged arterial wall and increases transcription of TGFβ and the TGFβ II receptor. Activated MMP2 activates TGFβ, leading to enhanced transcription of fibronectin and collagen genes. AngII signaling also increases the production of MCP1 and its receptor, CCR2, in arterial endothelial and vascular smooth muscle cells. Another sign of increased inflammation is enhanced NADPH oxidase activity, which also results in part from an age-associated increase in arterial AngII signaling. Excessive NADPH oxidase produces superoxide, which reacts with nitric oxide to produce the toxic species peroxynitrite, which can lead to protein nitration and “steals” nitric oxide in the process, reducing its bioavailability.

Interactions among these factors create a metabolically active environment (Lakatta, 2003, Spinetti et al., 2004 and Wang et al., 2003). Some vascular smooth muscle cells (VSMCs) shift their phenotype from contractile to secretory, proliferative, and invasive and migrate into the thickened intima. Enhanced VSMC invasion within the older arterial wall is promoted by elaboration of the chemoattractant PDGFβ and its receptor, both of which are also downstream signaling effects of Ang II. VSMC invasion of matrix within the old arterial wall is facilitated both by activated MMP2 and by MCP1. Arterial wall elastin becomes fragmented with advancing age, and excessive collagen is synthesized and becomes nonenzymatically glycated. Endothelial dysfunction and its attendant alterations in endothelial permeability and vasotonic actions occur. These same metabolic, enzymatic, cellular, and endothelial alterations appear to play a critical role in the genesis or promotion of hypertension, atherosclerosis, vascular inflammation, vascular remodeling, and oxidant stress. In other words, many of the same factors that underlie the age-associated structural and functional alterations of the arterial intima and media are also implicated in the pathogenesis of clinical arterial disease. These and other factors are the “culprits” that underlie the “risky” component of arterial aging in humans (Lakatta, 2003).

As progress is made in further elucidating the diverse molecular mechanisms that underlie the arterial alterations that accompany advancing age, novel therapies must emerge that will specifically target these pathways and retard or reverse “unsuccessful” arterial aging. Treatments targeting structural factors have begun. Nonenzymatic crosslinks between glucose (or other reducing sugars) and amino groups that generate advanced glycation end products alter long-lived proteins, e.g., collagen and elastin, and increase with advancing age, and both cause arteries and the heart to stiffen. NIA studies have demonstrated that ALT-711, a novel thiazolium agent that breaks such crosslinks, reduces arterial stiffness both in nonhuman primates (Vaitkevicius et al., 2001) and in humans (Kass et al., 2001). Chronic inhibition of angiotensin receptor signaling also substantially retards the age-associated increase in collagen content and intimal-medial thickening and stiffness in rodents (see Lakatta [2003] for review).

The substantial variability among older persons in the degree of these vascular changes also reinforces the possibility of identifying factors that modify them (Lakatta and Levy, 2003), including lifestyle factors: it is noteworthy that arterial stiffness is inversely related to physical fitness, assessed as maximum oxygen consumption, over a broad age range (Vaitkevicius et al., 1993). To date, however, clinical trial findings on the effects of physical activity on vascular stiffness have not been reported. Dietary interventions can modulate vascular properties. Food fat content has been shown to adversely affect vascular stiffness and endothelial function; diets that are reduced in sodium are associated with reduced arterial stiffening with aging, independent of their blood pressure-lowering effects (Gates et al., 2004).

While such preventive lifestyle or pharmacological strategies may be explored even now, future genetic characterization of individuals will likely allow person-specific stratification with respect to risk, efficacy, and cost effectiveness of measures to retard vascular aging in order to reduce CV functional disability and disease at older ages.
Alzheimer’s Disease

Alzheimer’s disease (AD) is a devastating neurodegenerative disease that afflicts older men and women, with risk increasing dramatically with age. Alzheimer’s disease research has stimulated not only knowledge of the disease but also understanding of normal brain aging and the ways in which aging of the brain both contributes to and differs from the pathology of AD. Knowledge gained from multiple avenues of research illustrates the way in which genetic, molecular, and epidemiologic information can be translated into preclinical and clinical interventions.

In the early 1900s, Alois Alzheimer described the amyloid plaques and neurofibrillary tangles characteristic of AD. Sequencing of the protein component of brain amyloid in the 1980s was followed soon thereafter by identification of the gene encoding its precursor protein, amyloid precursor protein (APP). Subsequently, it was found that mutations in the APP gene are responsible for some cases of familial early onset AD; that other cases of familial early onset AD are caused by mutations in two other genes, presenilins 1 and 2; and that mutations in all three of these genes increased the production of specific forms of amyloid. These molecular and genetic discoveries focused research on the pathways leading to amyloid formation and accumulation of plaques. The enzymes β- and γ-secretase cooperate to cleave APP and produce the peptide form, β-amyloid, that preferentially constitutes plaques in the brains of AD patients, making inhibition of these enzymes a target of current drug development strategies (history reviewed in Hardy and Selkoe [2002]). The enzyme α-secretase cleaves β-amyloid internally, potentially preventing the formation of plaques; enhancing the activity of α-secretase thus represents another target (Lichtenthaler and Haass, 2004). Yet another is degradation of amyloid (Tanzi et al., 2004) or preventing the accumulation of soluble, possibly highly toxic intermediates (Klein et al., 2001). The challenge of designing effective interventions that inhibit AD-specific pathways but do not interfere with essential normal functions is illustrated by the example of γ-secretase, which, apart from its role in β-amyloid production, is also the protease responsible for cleavage and activation of notch, a receptor that is involved in a diverse set of normal differentiative pathways (Conboy et al., 2003). This raises the possibility, being actively investigated by a number of researchers, that altered presenilin function may also lead to neurodegenerative disease by other, nonamyloidogenic pathways (Thinakaran and Parent, 2004).

The identification of mutated APP and presenilin genes that are causal for AD has also led to the development of the first transgenic mouse models of this disease, providing model systems for testing of amyloid interventions in vivo. Experimental treatments tested in these models to date include the successful reduction in amyloid plaque formation by chelation (Cherny et al., 2001) and by immunization against amyloid (Schenk et al., 1999). Clinical trials of immunization with amyloid peptide were initiated in humans but were stopped due to side effects of encephalopathy (Orgogozo et al., 2003). Researchers continue to explore different approaches to immunotherapy that would reduce amyloid load without negative side effects.

Neurofibrillary tangles are another obvious target for intervention. Research on metabolism of their protein component, tau, has led to insights into phosphorylation and other modifications of tau that favor tangle formation and dissolution of the microtubule network. Again, genetics has played a role in highlighting the role of tau, since a number of mutations in the tau gene have been shown to cause frontotemporal dementia (FTDP-17) and related dementias (Ingram and Spillantini, 2002). The link between tau and amyloid pathology in AD is not yet clear, nor is the link between age-related accumulation of tau or amyloid and development of AD. Research on multiply transgenic mice combining abnormalities in amyloid and tau (Oddo et al., 2004) is beginning to elucidate the relationships. Interventions targeting tau phosphorylation and stabilization of microtubules are being developed.

Understanding of genetic factors in the much more common late onset form of AD was initiated by the discovery that the E4 allele of the apolipoprotein gene is a major risk factor gene for the disease (Strittmatter et al., 1993), leading to research on the mechanisms of this effect and possible therapeutic targets. No other risk factor genes have yet been unambiguously identified, although several regions of the genome have been pinpointed. The NIA’s genetics initiative http://www.nia.nih.gov/ResearchInfo...scienceOfAging/AlzheimersDiseaseGenetics.htm; www.ncrad.org) will make available to qualified researchers well-characterized family and case control sets presently being identified by the AD Centers (http://www.alzheimers.org/adcdir.htm) and other investigators.

Even early stages of AD are characterized by loss of synapses and neurons in entorhinal cortex and hippocampus, areas important for memory. The antecedent(s) of cell death are unknown. Possible interventions targeting cell dysfunction and death include introduction of growth factors into affected brain regions, a technique effective in nonhuman primates (Smith et al., 1999) and presently the focus of a small clinical trial in AD patients. Stimulating endogenous stem cell production of neurons may positively affect hippocampal function (Kuhn et al., 1996 and Kempermann et al., 2004). Other possible targets, common to a number of neurodegenerative diseases, include boosting the pathways clearing abnormally folded and toxic proteins and developing strategies to reduce abnormal folding (Ross and Poirier, 2004). Techniques for assessing levels of thousands of individual mRNAs will reveal underlying patterns of mRNA and protein changes (Blalock et al., 2004), identifying altered pathways early in disease development.

Both AD brain pathology and epidemiological studies have implicated oxidative stress and inflammation as causal factors in AD. Diets enriched in antioxidants, such as vitamins C and E and α-lipoic acid improve learning and memory and reduce plaque load in aged canines (Milgram et al., 2004) and rats, and anti-inflammatory agents reduce plaque load in APP transgenic mice (Lim et al., 2001 and Eriksen et al., 2003). Long-term prevention clinical trials for several antioxidants are ongoing. Participants in a recently suspended prevention trial of nonsteroidal anti-inflammatory drugs will continue to be followed for safety and efficacy for 2 years. Other risk factors for AD include several also related to cardiovascular disease—high blood pressure, cholesterol, and homocysteine levels. A statin is being tested in a clinical trial on patients who have AD, and another treatment trial tests efficacy of reduction in blood homocysteine levels by folate/B6/B12. The effect on AD risk of lowering blood glucose levels is being tested in older diabetics. Data on most of these trials can be accessed at http://www.alzheimers.org/clintrials/search.asp.

Research on AD initially focused on late stage disease, but with better understanding and more sophisticated research tools, the focus has shifted to earlier and even preclinical stages (Grundman et al., 2004). It is likely that AD pathology is initiated many years before clinical diagnosis; by the time a person has mild cognitive impairment (MCI) with memory decline, loss of key neurons in the entorhinal cortex is profound (Price et al., 2001), and loss of neuronal function is apparent in other brain regions as well (Mufson et al., 2002). Important initiatives will therefore focus on developing prevention strategies that would slow or halt the pathology before major functional impairment. Key to early identification of persons at risk is the development of neuropsychological, clinical, and imaging techniques sensitive to characteristic early pathologies such as delayed recall memory measures, loss of hippocampal volume, and lowered metabolism in affected brain regions. Identification of suitable imaging and biological makers of progression would be critical to reducing the size, time, and cost of treatment and prevention trials. The AD Neuroimaging Initiative (ADNI) (http://www.alzheimers.org/pubs/conv11n1and2.htm#new) is a major new longitudinal study of normal controls and persons with MCI and AD to identify surrogate markers for disease onset and progression and is potentially a scaffold for ancillary studies testing specific hypotheses.
Conclusion

In recent years, cellular, molecular, and genetic studies of aging in in vitro models and in short-lived invertebrates have generated an impressive pace of discovery. These discoveries, together with outcomes of epidemiologic and clinical research, have in turn helped to identify potential therapeutic targets for testing in longer-lived mammalian species. It is important that this pace of progress continue, with application of cutting-edge basic science. It will also be critical to develop strategies that will allow selection of the most promising of candidate interventions suggested by in vitro and invertebrate studies, and by laboratory mammal and human studies, for further evaluation.

As illustrated in this review, in all aspects of the development of interventions that target aging-related processes, from in vitro to clinical studies, there are needs and opportunities for innovative molecular and cellular research strategies. Aging is a fundamental biological phenomenon with wide-ranging medical and societal effects. The discovery of effective new therapies to address it is a worthy challenge for molecular and cell biology.