Metabolic networks of longevity

Abstract

Molecular and cellular networks implicated in aging depend on a multitude of proteins that collectively mount adaptive and contingent metabolic responses to environmental challenges. Here we discuss the intimate links between metabolic regulation and longevity, and outline novel approaches for analyzing and manipulating such links to promote human healthspan.

Life expectancy has increased steadily over the last century and almost doubled in many countries and cultures. This increase is not due to an intrinsic change in human physiology but reflects large-scale improvements in public health. The major cause of mortality is shifting rapidly from infectious diseases to a variety of cardiovascular and metabolic disorders. Although treating diabetes, hypertension or hyperlipidemia generally increases lifespan, there is no evidence that global application of drug interventions (e.g. statins for heart disease) improves longevity in healthy subjects. Here, we do not consider pathways that cause either rare (e.g. progeria) or common (e.g. hypertension or hyperlipidemia) diseases. Instead we focus on molecular and cellular networks that are part of an intrinsic program that controls lifespan in the absence of overt disease.

The only intervention that consistently has been shown to increase lifespan from nematodes to primates is caloric or dietary restriction (Colman et al., 2009; Fontana et al., 2010; Lin et al., 2002). How caloric restriction (CR) extends lifespan is still not well understood (reviewed in (Canto and Auwerx, 2009), but several studies demonstrate that perception and sensing of nutrient levels is important. It is likely that the absence of specific dietary amino acids mediates the effects of CR rather than the restriction of calorie intake per se (Grandison et al., 2009; Miller et al., 2005). A diet low in the essential amino acid methionine boosts lifespan in the mouse and reduces age-related pathologies (Miller et al., 2005). Addition of methionine alone to a CR diet also averts the reduced fecundity normally associated with CR in the fruit fly Drosophila melanogaster but without affecting increased longevity (Grandison et al., 2009). Surprisingly, whereas addition of essential amino acids to a CR diet reduced lifespan, this reduction of lifespan was prevented by simply removing methionine from the diet, indicating that the interaction between methionine and the other amino acids plays a key role (Grandison et al., 2009). Mechanisms directly linked to methionine manipulation are unknown but may involve specific metabolic sensors, many of which are extremely well conserved and often act in a cell-autonomous fashion, i.e. acting within the cell. Possible sensors include the target of rapamycin (TOR), AMP-activated protein kinase (AMPK), and the sirtuin proteins that detect changes in specific metabolites such as amino acids, AMP, and NAD⁺, respectively. Such changes, reflecting the general metabolic state, could trigger metabolic adaptations, possibly through regulators such as the forkhead box transcription factors (including the FOXO and FOXA families), SMK-1 (suppressor of MEK null), and the peroxisome proliferator-activated receptor γ coactivator 1α (PGC-1α).

Interestingly, the sensory perception of nutrients is also involved in CR-induced longevity, suggesting that non-cell autonomous signaling pathways, i.e. relying on external cues, such as hormonal and neuronal pathways, also may be involved in the aging process (reviewed in (Libert and Pletcher, 2007). Blocking sensory perception in the nematode Caenorhabditis elegans or D. melanogaster by ablation of the chemosensory neurons or by deletion of the odorant receptor Or83b, respectively, increases lifespan. Although CR still extends lifespan in the Or83b mutant flies, the effect was milder than in wild-type flies. Strikingly, even exposure to food odorants dampens the CR-induced lifespan extension in flies (Libert and Pletcher, 2007). The fact that serotonin inhibitors, which in humans are used as antidepressants and often induce weight loss, also extend lifespan in C. elegans (Libert and Pletcher, 2007), suggests that specific neuronal signaling pathways may mimic the pseudo-starved state. Whether and how sensory perception is also relevant in mammals merits further research.

Elucidating the mechanisms by which both these cell-autonomous and non-cell-autonomous signaling pathways are integrated to control the response to CR, will be key for understanding longevity and its natural variation. However, the beneficial effects of CR are not uniformly mediated by these pathways, because the actions of these pathways depend largely on the context of the CR regimen and the compensatory regulatory networks that are activated. It is therefore no surprise that depending on the timing and type of the CR, certain genes and pathways are dispensable for the effect on lifespan (Kenyon, 2010). How these various signaling pathways impact longevity, whether CR-induced or natural, is still not clear.

Longevity genes and metabolic pathways

The insulin/IGF1 signaling pathway is the best-characterized pathway affecting longevity (Figure 1). In C. elegans and D. melanogaster — organisms in which the insulin and IGF1 pathways converge on a single receptor — decreased insulin/IGF1 signaling increases lifespan by as much as two-fold (reviewed in (Kenyon, 2010). In mammals, the role of the insulin/IGF1 signaling network in longevity is complicated by potential involvement of multiple insulin/IGF receptors. Perturbing this pathway in mice results in severe metabolic disease leading to premature death (reviewed in (Russell and Kahn, 2007). Several studies, however, have connected insulin/IGF1 signaling to longevity in mammals. Lifespan is clearly increased by heterozygous and tissue-specific mutations in components of the insulin/IGF1 pathway (Russell and Kahn, 2007). The FOXO (forkhead box) transcription factors are key modulators of this insulin/IGF1 network (Kenyon, 2010). Phosphorylation of FOXO by AKT — a downstream kinase in the insulin/IGF1 pathway — results in translocation of FOXO to the cytosol and inactivation of its pro-longevity transcriptional targets. Such targets include genes involved in defense against oxidative stress and genes encoding molecular chaperones (Russell and Kahn, 2007). Consistent with this, deleting daf-16 (the C. elegans FOXO homolog) in worms carrying a mutated daf2 (which have no insulin/IGF1 signaling) abrogates the lifespan extension of the daf2 mutant worms (Kenyon, 2010). The importance of FOXO in CR-induced longevity is, however, still under debate. In contrast to worm daf-16, the PHA-4 protein of C. elegans, which is homologous to the mammalian family of Foxa transcription factors, mediates the effects of CR on lifespan, but is not required for the increased longevity caused by other genetic pathways that regulate aging (Panowski et al., 2007). Future research will need to address what causes these species-specific differences and how such differences can be used to influence mammalian longevity.

Figure 1. A metabolic network of aging.

A simplified scheme for metabolic proteins that are involved in longevity. There are complex interactions among individual metabolic signaling pathways, such as the insulin/IGF1, TOR, AMPK, sirtuin and FOXO pathways. These interactions impact healthspan and longevity. For instance, the TOR pathway, which is activated in situations of nutrient excess, is not only linked to other nutrient-responsive pathways, such as the insulin/IGF1 pathway, but also to the CR-induced pathway involving AMPK and Foxa activation. Longevity promoting proteins are in green, proteins that when mutated increase longevity are in red, other proteins are indicated in blue; mammalian protein names are used.

The TOR signaling pathway is also a critical player in longevity. Initially characterized for its involvement in the antifungal effects of the drug rapamycin, TOR proteins are now recognized as sensors that link nutrient availability to cellular growth (reviewed in (Stanfel et al., 2009). In yeast, worms, and flies, manipulation of TOR by genetic ablation or by pharmacological inhibition using rapamycin greatly improves lifespan (Figure 1) (Stanfel et al., 2009). The same beneficial effect of rapamycin is also seen in mice, where it increases mean and maximum lifespan, even though the treatment only started in aged animals (Harrison et al., 2009). Using various fly mutants, the effects of rapamycin on longevity have been linked to the regulatory role of TOR on autophagy and on translation. The ribosomal S6 protein kinase 1 (S6K1) and the eukaryotic translation initiation factor 4E binding protein (4EBP), both downstream effectors of TOR, are critically involved in these processes (Bjedov et al., 2010). The role of autophagy in longevity has been further substantiated by using the polyamine spermidine, a compound that induces autophagy, inhibits oxidative stress, and increases lifespan in several model organisms (Eisenberg et al., 2009). Finally, the roles of S6K1 and 4E-BP in lifespan extension have been confirmed in genetic models. Deletion of 4E-BP in D. melanogaster decreases lifespan and abrogates CR-induced longevity, whereas overexpression of 4E-BP does not enhance the effect of CR, suggesting that 4E-BP is crucial for CR-induced longevity (Zid et al., 2009). Conversely, female mice lacking S6K1 live longer, healthier lives compared with wild-type littermates (Selman et al., 2009), perhaps because loss of S6K1 mimics some aspects of CR. These mice are lean and sensitized to insulin because of a decrease in the phosphorylation of threonine residues in IRS1 (insulin receptor substrate 1) resulting in altered insulin signaling (Um et al., 2004). Based on gene expression data and the characterization of worms carrying mutant S6K1 and AMPK, it seems that the beneficial effects of the loss of S6K1 on glucose homeostasis and lifespan are principally mediated through AMPK activation (Selman et al., 2009).

Sirtuins are a class of NAD⁺-dependent deacetylases, which deacetylate both histones and a wide range of proteins (Houtkooper et al., 2010). The connection between sirtuins and longevity is based on the discovery that the effects of CR on lifespan are, at least in part, mediated by the sirtuin orthologues sir2/SIRT1 in yeast, fly, and worm. Although the exact function of sir2/SIRT1 in lifespan extension is still not known, its activating effects on mitochondrial respiration are thought to play a crucial role (Canto and Auwerx, 2009). Consistent with the link between SIRT1 function and metabolism, single nucleotide polymorphisms in SIRT1 have been associated with energy expenditure (Lagouge et al., 2006) and SIRT1 expression levels are tightly correlated with insulin sensitivity in humans (Rutanen et al., 2010). Notably, a recent study showed that transgenic mice overexpressing Sirt1 have an improved healthspan and are protected against diseases of aging, such as diabetes and cancer, but do not show an increased lifespan (Herranz et al., 2010). This striking absence of a longevity phenotype in Sirt1 transgenic mice underscores the pitfalls associated with translating the conclusions of studies in yeast, worm, and flies to mammals. Additional data concerning the possible relation of SIRT1 and mouse or human longevity are eagerly awaited. Finally, a potentially interesting association between polymorphisms in the SIRT3 gene, one of the mitochondrial sirtuins, and longevity was reported in a small human study (Bellizzi et al., 2005).

Some of the sirtuin longevity pathways involved in CR-mediated lifespan extension, most notably those involving SIRT1, are tightly intertwined with those controlled by the metabolic sensor AMPK, which has been linked to longevity in multiple ways. Like the FOXOs, the role of AMPK in CR-induced longevity is still debated (Kenyon, 2010), however, it does seem to be important in several longevity pathways. Indeed, the beneficial effect of S6K1 deficiency on lifespan is thought to involve AMPK activation (Selman et al., 2009). In addition, AMPK phosphorylates and thereby activates FOXO, possibly impacting longevity. Another mechanism by which AMPK may control longevity entails boosting NAD⁺ levels and activating SIRT1, which in turn deacetylates and activates targets, such as PGC-1α and FOXO1, leading to the induction of mitochondrial activity, respiratory metabolism, and oxidative stress responses (Canto et al., 2009; Canto et al., 2010). A role for AMPK in longevity is also supported by the fact that metformin, an AMPK agonist that is used clinically for treating type 2 diabetes, extends lifespan in mice (Anisimov et al., 2008). In addition, resveratrol, a drug that activates SIRT1 but not in a direct fashion (Beher et al., 2009; Pacholec et al., 2010) achieves this effect by mildly inhibiting mitochondrial respiration, thereby inducing energy stress. The subsequent increase in the AMP/ATP ratio activates AMPK, increases NAD⁺ levels and stimulates SIRT1, resulting in a compensatory induction of mitochondrial activity (Canto et al., 2009; Canto et al., 2010; Um et al., 2010), a process termed “mitohormesis”. In mice fed a high calorie diet, resulting in a significantly decreased lifespan, resveratrol normalized lifespan suggesting that the AMPK/SIRT1 pathway is involved (Baur et al., 2006). Although this suggests the involvement of AMPK/SIRT1 signaling in longevity, many aspects — such as the role of this signaling pathway and mitochondria in CR — remain to be investigated especially in mammals.

Many of the longevity genes identified to date have multiple effects on metabolism and appear to be intertwined in partly overlapping metabolic signaling networks that often affect the function of mitochondria (Figure 1). Supporting a potential role for mitochondria in longevity, is the fact that activation of the transcriptional cofactor PGC-1α, which in mammals controls many aspects of mitochondrial biogenesis and function, protects against diseases associated with aging (Baur et al., 2006; Lagouge et al., 2006). Definitive proof of PGC-1α’s role in longevity is, however, still lacking. The impact of certain mutations in mitochondrial genes also deserves attention. Could it be that mutations that partially impair mitochondrial function, such as the oxidative phosphorylation or ubiquinone synthesis mutants of the worm (Dillin et al., 2002; Lee et al., 2003) induce mitohormesis and longevity? The seemingly contradictory results regarding activation or inhibition of mitochondrial metabolism and increased longevity warns us that a simple genetic approach is not sufficient and that a more integrative physiological approach, including the thorough study of epigenetic and metabolic pathways, is warranted. The characterization of these pathways will undoubtedly be fertile ground for further research in the longevity field and will result in a clearer view of how mitochondria are involved in longevity and in aging-related diseases.

Multifactorial aspects of longevity

Studies of genetically engineered animal models have proven useful for identifying longevity genes, but monogenic approaches have limitations. We also need methods that make it possible to study complex polygenic networks that interact with environmental factors to influence metabolic homeostasis and longevity. A hallmark of such complex traits is their graded and nearly continuous distribution throughout a population. Single gene variants are not associated with clear-cut Mendelian and modal patterns of inheritance. However, contributions of these variants can often be mapped to specific chromosomal regions that are called quantitative trait loci (QTLs) because they generate some fraction of the total variation of trait values in study populations. The study of complex traits has advanced rapidly and has relied both on family linkage studies and genome-wide association studies of large human cohorts. Corresponding studies in model organisms have often exploited test crosses (F2 intercrosses and backcrosses) and genetic reference panels, including sets of congenic and recombinant inbred strains. But how are these genetic approaches being applied to identify genes and molecular networks associated with longevity?

Studies of human populations have led to the identification of a number of loci associated with longevity (de Magalhaes et al., 2009). These studies generally entail an analysis of populations with exceptionally long lifespans, such as centenarians. This approach is valuable, but studies of humans are still difficult because of the limited availability of subjects and potential environmental confounds. The most important limitation is of course that prospective and experimental perturbations are usually not feasible for testing causal models.

Studies of rodents and other model organisms overcome some of these problems and allow us to add the experimental component while still accounting for genetic diversity. The most straightforward network approach uses model organisms in which a single gene is perturbed, either by genetic deletion or RNA interference, to map the compensatory mechanisms that result in a longevity trait. A more complicated method, but also more realistic in terms of modeling longevity, involves the characterization of QTLs in recombinant inbred lines. Recombinant inbred lines of D. melanogaster have been used extensively for the identification of QTLs for longevity. These studies typically use ~100 recombinant inbred fly lines to measure lifespan and link variation to specific QTLs (Curtsinger and Khazaeli, 2002). A challenge has been to achieve sufficient resolution to define responsible gene variants. A recent study used extreme QTL (X-QTL) mapping to define determinants of mitochondrial function in yeast (Ehrenreich et al., 2010). Even though this study did not assess yeast replicative or chronological lifespan, it marks an important step because X-QTL analysis not only allows mapping of major QTLs, but also minor ones that might be missed using classical methods. An interesting alternative approach involves the analysis of modular networks (Xue et al., 2007). Combining genes in various modules (e.g. differentiation, proliferation, oxidative metabolism and reductive metabolism) reveals that the aging process in the fly is characterized by a general shift from oxidative to reductive metabolism. CR clearly slows down this shift, thereby accounting for the longevity effect. Known aging genes are highly represented as “date hubs” between these modules, indicating that they serve as important regulators of aging (Xue et al., 2007). Interestingly, this unbiased network approach confirmed the importance of oxidative metabolism, implicated by monogenic studies.

Another step forward will be the switch from relatively simple organisms, such as yeast and C. elegans, to more complex mammalian systems. Genetic and genomic analyses of large mouse families, generated by intercrossing long-lived and short-lived stock, will be useful for finding new longevity genes and networks. We recently used gene expression data for adipose tissue from four large mouse F2 crosses to identify candidate longevity networks. Following the strategy of using the ‘known’ to uncover the ‘unknown’, we leveraged a seed set of 60 known longevity genes to extract a larger network, containing 742 tightly connected longevity genes (Argmann et al., 2009). Among the top 20 genes tightly connected with known longevity genes was 4E-BP, a determinant of longevity in multiple species (Argmann et al., 2009; Zid et al., 2009). When we applied pathway enrichment analysis to uncover relationships among the 742 genes contained within the longevity network, we detected enrichment in several gene ontologies with established links to aging such as the complement and coagulation cascades (inflammation), insulin signaling, and the ubiquinone pathway (i.e. oxidative stress). Importantly, other unsuspected pathways were also highly enriched, including PPAR signaling — an ontology that actually ranked highest. Using an allelic series of PPARγ mutant mice, we validated a role for PPARγ in longevity (Argmann et al., 2009; Heikkinen et al., 2009).

The analysis of mouse genetic reference populations is another powerful strategy to identify molecular networks that influence longevity. Genetic reference populations, such as the BXD and LXS mouse strains and the HXB rat strains, are stable families of inbred lines that can be used for large-scale collaborative phenotyping across multiple dimensions. These types of resources are ideal for studying the genetics of longevity because each genome is represented by an entire strain, making it possible to obtain far more accurate mortality data for each cohort of animals and for all genotypes under different experimental conditions and interventions. The strains within a genetic reference population collectively incorporate a level of genetic variation that can match or exceed that of human populations, making them an excellent vehicle with which to explore the generality of findings, the role of genetic background effects, and the translational relevance of gene variants to human populations and disease.

Two large mouse genetic reference populations are now being used actively in longevity research — the BXD set generated by crossing C57BL/6J and DBA/2J, and the LXS set generated by crossing ILS/IbgTejJ and ISS/IbgTejJ. Using the original subset of BXD strains, several regions in the mouse genome have been linked to longevity (De Haan and Van Zant, 1999; Gelman et al., 1988), but small sample sizes and technical limitations have hampered the analysis. However, larger and much more powerful genetic and genomic data sets can now be exploited to reanalyze original longevity data for larger genetic reference populations and to extract more definitive and precise mapping results. Following identification of a QTL, candidate genes can be identified by association studies between mRNA expression and the trait, that is, lifespan. Expression of these genes can then also be cross-referenced with the expression of other genes, allowing a network to be assembled. Online tools, such as the Kyoto Encyclopedia of Genes and Genomes (KEGG) and gene ontology (GO) analysis can help in the identification of pathways linked to candidate genes. Finally, interesting candidate genes can be validated using longevity analyses in vivo. The advantage of this approach is that it uses natural variations in gene expression (like those found in normal human populations) instead of relying on the more artificial situation that occurs in gene knockout or knockdown experiments. It is imperative to realize that the directionality of the association cannot be identified using such strategies and that additional physiological evidence is always required to back up the genetic data and establish causality and directionality. This is especially relevant as various longevity pathways obtained in lower organisms proved difficult to phenocopy in mammals (Russell and Kahn, 2007) and, in other cases, mutations in the same gene resulted in opposite effects on lifespan (Argmann et al., 2009; Heikkinen et al., 2009). These discrepancies indicate that additional efforts should be put into the identification of the mechanisms underlying these differences, and attention should be paid not only to genetic but also to epigenetic factors that could affect metabolic networks of aging.

Perspectives

Studies using single gene gain or loss to alter gene activity in various organisms, ranging from yeast to mouse, have provided a wealth of information about specific genes, such as those encoding components of the insulin/IGF1 pathway, involved in longevity. Using such reductionist approaches, an emerging theme is the overrepresentation of metabolic pathways in general, and of mitochondrial activity in particular, as determinants of longevity. In this Essay, we have outlined the known metabolic pathways impacting longevity and we have integrated these pathways in a simplified network (Figure 1).

Although it has long been appreciated that longevity is a complex process influenced by multiple gene variants, the development of complex genetics, genomics, and bioinformatics tools is now finally enabling a more thorough systems analysis of longevity networks. By using population studies in mouse and human, multiple genes and pathways have been recently linked to longevity and aging-related diseases. We have tried to give a flavor of what may soon emerge by using such synthetic approaches. In the near term, we foresee that the continued use of complex genetic strategies will gain importance in the longevity field and lead to the identification of tightly regulated longevity networks that govern intersecting metabolic pathways often affecting mitochondrial functions. Furthermore, it will prove increasingly important to extend these studies beyond a purely genetic level to include the evaluation of epigenetic factors. Such epigenetic factors impact processes ranging from DNA transcription (histone and DNA modifications), RNA (miRNAs), and protein stability and activity (post-translational modifications), to metabolite levels (metabolite ligands or cofactors for proteins). The application of such holistic research strategies to elucidate the mechanisms by which CR impacts healthspan and lifespan in a variety of model organisms will probably be a major step forward in our understanding of the aging process. Finally, given the conserved role of mitochondria across species in the determination of longevity, obtaining insights into the complex and multifactorial regulation of metabolic flexibility may in fact hold the key to unlock the metabolic underpinnings of aging. The importance of metabolic flexibility for longevity is perhaps best illustrated by the universal association of a shift from glycolytic to oxidative metabolism in CR or endurance exercise, conditions that improve healthspan, whereas the converse switch from oxidative to glycolytic metabolism occurs when healthspan is compromised, such as in cancer or ischemia. It goes without saying that identifying the crucial nodes of these metabolic longevity networks might translate into the development of targeted preventative and therapeutic strategies to improve healthspan, a goal since the quest of Ponce de Leon to find the fountain of youth.