Age is a major risk factor for many human diseases. Extremely long-lived individuals, such as centenarians, have managed to ward off age-related diseases and serve as human models to search for the genetic factors that influence longevity. The discovery of evolutionarily conserved pathways with major impact on life span in animal models has provided tantalizing opportunities to test the relevance of these pathways for human longevity. Here we specifically focus on the insulin/insulin-like growth factor-1 signaling as a prime candidate pathway. Coupled with the rapid advances in ultra high-throughput sequencing technologies, it is now feasible to comprehensively analyze all possible sequence variants in candidate genes segregating with a longevity phenotype and to investigate the functional consequences of the associated variants. A better understanding of the functional genes that affect healthy longevity in humans may lead to a rational basis for intervention strategies that can delay or prevent age-related diseases.
Rather than driven by stochastic events alone and being a simple entropic process of deterioration, the rate of aging and life span have been shown to be regulated by multiple pathways. Many single-gene mutations identified in invertebrates result in a reduced rate of aging and an increased life span (1,2). A large fraction of these mutations extends life span in mammals as well (3–5). These studies have provided important clues for discovering the evolutionarily conserved pathways of life span and their impact on the aging rate. Altering the pathways controlling metabolism, endocrine signaling, nutrient sensing, stress resistance, and telomere maintenance has been shown to prolong life span from yeast to mammals (6,7). Among those conserved pathways, the insulin/insulin-like growth factor-1 (IGF-1) signaling (IIS) pathway was the first one shown to regulate life span in animal models and also had the most profound effect on life-span extension (2,8). In brief, binding of IGF-1 and insulin activates an intracellular signaling cascade composed of IRS-1/PI3K/AKT that antagonizes the activity of Forkhead/FOXO transcription factors (Figure 1). Reduction-of-function mutations in IIS genes activate Forkhead/FOXO proteins that regulate expression of several hundred genes implicated in metabolism, stress resistance, and antimicrobial defense (9,10). Remarkably, these long-lived animals maintain their youthfulness (11,12) and display delayed signs of aging and a late onset of age-related diseases (12). Continued discoveries of beneficial effects of mutations on many aspects of aging in long-lived animals support the concept that aging is a major risk factor for many common human diseases and that targeting aging is a logical and economically sound strategy in developing new targets for prevention and early-stage treatment.
The single-gene mutations extending life span were first discovered in nematode, Caenorhabditis elegans (13,14). Hypomorphic mutations of daf-2, encoding nematode ortholog of insulin receptor/IGF-1 receptor (IGF1R; 15), and age-1, encoding a PI3K ortholog (16), result in profound extension of life span (Figure 1). The IIS pathway negatively regulates the transcriptional activity of the Forkhead transcription factor, daf-16/FOXO, and limits longevity (17,18). Mutations that attenuate signal transduction within the IIS pathway reduce phosphorylation of DAF-16, which enables DAF-16 to translocate into the nucleus, thereby modulating transcription of several hundred target genes in distinct families including antioxidants, chaperones, and antimicrobial defenses (21–23). Inhibition of IIS also activates the heat-shock transcription factor-1 (24,25) and the Nrf-like xenobiotic-response factor, SKN-1 (26). Once activated, heat-shock transcription factor-1 induces expression of small heat-shock proteins, and SKN-1 induces genes involved in phase 2 detoxification response. Overall, IIS exerts its effects on aging in C. elegans by modulating expression of downstream genes through these transcription factors. In Drosophila melanogaster, mutations in the insulin receptor homolog and the insulin receptor substrate, CHICO, also impair IIS and prolong life span in females (27–29). Similar to C. elegans, ablation of the daf-16 homolog, dFOXO, almost completely blocks IIS-mediated life-span extension (30), whereas tissue-specific overexpression of dFOXO extends life span (31,32).
During evolution, the IIS pathway has diverged from a single receptor in invertebrates to multiple receptors and pathways forming more complicated regulatory networks in mammals with a certain degree of redundancy (33). Insulin regulates metabolic pathways, while IGF-1 controls cell growth and differentiation (33,34). Therefore, translating longevity effects of IIS gene mutations discovered in invertebrates to mammals is neither plain nor simple. However, several genetic manipulations in mice have shown that interference of this pathway prolongs murine life span. Mutations that reduce growth hormone (GH) production or GH receptor activity result in small body size (dwarfism), numerous indices of delayed aging, enhanced stress resistance, and a major increase in the life span of mice (34–41). GH is the major regulator of circulating IGF-1 suggesting that GH functions are mediated at least partially through IGF-1 action. Alterations in IIS emerged as one of the likely mechanisms linking GH with aging. The most compelling model seems to be the heterozygous IGF1R knockout mouse. Whereas mice homozygous for IGF1R knockout die at birth, female heterozygous mice live 33% longer than wild-type mice and show an increased resistance to oxidative stress (42). The same genetic alteration also protected animals from Alzheimer’s-like disease symptoms, including reduced behavioral impairment, neuroinflammation, and neuronal loss (43). Adipose tissue–specific homozygous deletion of the insulin receptor (INSR) increases life span in both sexes (44). Most recently, the homozygous deletion of IRS-1, which serves as an adapter protein that transmits signals from both IGF-1 and insulin receptors, extended life span and increased resistance to various age-related pathologies only in female mice (45). These results suggest that the GH/IGF pathway is critical in murine longevity, where the effects of reduced IIS is often more profound in females (35). Similarly, as described later, the results from human genetic studies suggest that the association of IIS gene variants with human longevity is sometimes more significant in women.
IIS PATHWAY AND HUMAN LONGEVITY
Despite the parallel between the IIS in invertebrates and mammals (3,6), the importance of this regulatory network for aging and longevity in human populations is uncertain. The question is, do candidate loci with major effects on longevity show segregating allelic variation in natural populations? Indeed, even if we adopt the premise that the basic mechanisms that control longevity and determine the rate of aging are common to all multicellular organisms, it is possible that the observed effects of candidate gene variants on longevity will prove to be exclusively associated with laboratory strains, but not in the same animals in the wild. This concept is illustrated by an elegant work by Walker and coworkers (46), who exposed mixtures of mutants and wild-type age-1 worms to more natural conditions, that is, cycles of starvation. A large reduction in the frequency of the mutant allele suggested a substantial difference in relative fitness. Furthermore, it has been demonstrated that daf-2 worms with partial loss-of-function of the IGF1R-like protein exhibit dramatically reduced fitness even under laboratory conditions (47). Thus, it remains to be demonstrated whether candidate loci with major effects on longevity in model organisms such as genes involved in insulin/IGF-1 pathway will show segregating allelic variation in human populations. Moreover, humans suffer from different age-related diseases than mice. Therefore, the only valid strategy to test the possibility that genetic variation at candidate loci contributes to phenotypic variance for longevity and aging phenotypes in natural populations would be to conduct population-based studies.
MODELS OF HUMAN LONGEVITY: CENTENARIANS
Longevity is known to have a genetic component. Although debated, the heritability of average life expectancy has been estimated to be approximately 25% (48–50). Family studies on centenarians, those who aged 100 years or more, suggest that the relationship between genetics and longevity is stronger in the oldest-old adults. For example, the siblings of centenarians have a four times greater probability of surviving to age 90 than the siblings of people who have an average life expectancy (52). When it comes to living 100 years, the probability is 17 times greater for male siblings of centenarians and 8 times greater in fe` siblings of centenarians compared with average life span of their birth cohort (53). The immediate ancestors of Jeanne Calment, the longest ever lived human being who died at the age of 122 years and 164 days, were shown to have more than a 10-fold higher probability of living to 80 years or more compared with the control ancestors of a reference family (54). These studies support the utility of centenarians as a model system for studying the genetics of longevity, with extreme age as the phenotype toward the discovery of genetic variations predisposing people to longevity. Therefore, genetic studies of centenarians are based on the premise that they may lead to the identification of alleles that are either particularly enriched in these populations due to positive effects on extreme longevity or under represented due to a negative impact on human health. With the recognition that the IIS pathway regulates life span and healthy aging in animal models, genes involved in IIS become attractive candidates for investigating the genetic basis of human longevity using centenarians as a model (candidate gene approaches).
GENETIC ASSOCIATION STUDIES IN HUMAN LONGEVITY: IIS AS CANDIDATES
Candidate gene approaches focus on the genes or pathways that are selected based on a priori hypotheses about their role in the phenotype (55). A standard linkage approach is usually not feasible due to an almost complete absence of extensive pedigrees of long-lived individuals and the inherent complexity of longevity phenotypes, which likely involve multiple genes with small effects. Association-based studies are more effective tools than linkage studies when the focus is on complex traits such as longevity because they have greater statistical power to detect genes of small effect (56–59). By comparing the frequency of genetic variants in cases (long-lived individuals, nonagenarians, or centenarians) and unrelated controls (elderly individuals), association studies evaluate the correlation between a genetic variation and a particular trait, eg, longevity (Figure 2). Single-nucleotide polymorphisms (SNPs) are the most common type of genetic variations in the human genome and have been the workhorse of association studies. As described later, the results from genetic association studies of both common SNPs and rare variants point to a role of the IIS pathway in human longevity.
Common IIS Variants and Human Longevity
In most candidate gene approaches, SNPs are selected based on linkage disequilibrium, which describes the nonrandom correlation between alleles at a pair of SNPs. The HapMap provides global coverage of the genome with SNP markers (60,61) to select nonredundant tagSNPs based on defined haplotype patterns. In this format, genetic association analysis is an indirect approach using a set of SNPs as surrogate markers to detect association between a candidate gene and longevity, whether or not the markers themselves have functional effects. The search for the causative variants would then be limited to the regions showing association with longevity. In addition, selected SNPs are necessarily common (minor allele frequency >5%–10%) to achieve statistical power to detect an association with a given sample size.
Six polymorphisms from IIS pathway genes (GHRHR, GH1, IGF-1, INS, and IRS1) that were previously reported to be associated with the gene expression or protein function were analyzed in Leiden 85 plus cohort by van Heemst and coworkers (62). The analyses were stratified for gender because reduced IIS extends life span of females more than in both D. melanogaster and mice. The GH1 SNP (rs2665802) was associated with reduced body height and relative mortality in women only, whereas most SNP alleles that were associated with reduced signaling showed only a trend with the same directionality. To address the combined effect of the six polymorphisms, they used an additive model (composite IIS6 score) in which each of the SNPs was weighted equally because they had no a priori knowledge regarding the relative effect of individual SNPs. They calculated composite IIS6 score by assigning a value of −1 for an allele expected to have reduced IIS activity or +1 for an allele predicted to cause increased IIS activity, therefore an individual bearing all six alleles predicted to have reduced IIS would have an IIS6 score of −6. Association of IIS6 score with height and longevity was observed only in women.
The IGF1R SNP (rs2229765) at nucleotide 3174 of messenger RNA has also been shown to be associated with lower plasma IGF-1 levels and is more frequent in long-lived individuals (63). For FOXO1A, two intronic SNPs (rs10507486 and rs4943794) were associated with age at death in a Genome-Wide Association Study (GWAS) of the Framingham study (64), and two other intronic SNPs (rs2755209 and rs2755213) showed association with the longevity of Chinese women (65). Recently, a systematic study was conducted to analyze 291 SNPs in 30 genes involved in the IIS pathway, showing the association of several genes, including IGF1R, FOXO3A, and AKT, with longevity in Caucasian women from the Study Osteoporotic Fracture cohort (19). The association of AKT and FOXO3A was replicated in other Caucasian populations, eg, Cardiovascular Health Study and Ashkenazi Jewish Centenarians. Among the associated IIS genes, FOXO3A appears to be the most robust longevity gene in humans because common variants in this gene showed consistent association with longevity in multiple populations with different ethnic backgrounds including Asians, Europeans, Caucasians, and African Americans (19,64–68). Even though each study selected different SNPs for genotyping, the fact that these SNPs show the same effect and many of them are in high linkage disequilibrium suggests that the FOXO3A gene plays a role in longevity.
Rare IIS Variants and Human Longevity
Although there is great enthusiasm for the candidate association approach using marker SNPs based on the expectation that statistically powerful genetic association studies are now feasible, a major criticism still remains. This involves the assumption that there is little allelic heterogeneity within loci and that susceptibilities for longevity are due to a small number of ancient polymorphisms that occur at high frequencies in all populations as argued in the common disease/common variant hypothesis (69,70). However, if late-onset phenotypes, such as longevity, are due to large numbers of rare variants at many loci, this strategy would fail, as no single haplotype would be strongly associated with longevity, and the contribution of most individual variants would be too small (71). These concerns turned out to be true even for common traits as evident from the GWAS. More than 1,000 published GWAS reported significant (p < 5 × 10−8) associations of approximately 4,000 SNPs for more than 200 traits/diseases (72). However, each locus has a surprisingly low to modest effect. Furthermore, there is a wide gap between the population variation in disease that is explained by the results of GWAS (usually <10%) and heritability estimates (often >50%; ). Much of the speculation about missing heritability from GWAS has focused on the possible contribution of rare variants (minor allele frequency < 0.5%; 74).
Rare variants would not be strongly associated with any common alleles defined by the tagSNPs and are not queried in most case–control studies. Because extreme longevity is a rare phenotype (1/10,000; 75), only an exhaustive and comprehensive approach will ensure that no rare but important functional SNP escapes attention. So far, we were the first to report comprehensive resequencing analysis of IGF-1 and IGF1R genes in centenarians (n = 79) and controls (n = 161) leading to the discovery of two potentially functional missense variants in the IGF1R gene, Ala37Thr and Arg407His (76). Although rare in frequencies, these variants are significantly enriched in centenarians (n = 9) as compared with controls (n = 1) among a total of 384 centenarians and 312 controls tested (76). As increasingly recognized, most functional SNPs might be rare (77,78) and most rare variants arise as new mutations, which most likely reduce the gene function (79). These results suggest that centenarians may harbor individually rare but collectively more common genetic variations in genes encoding components of the GH/IGF signaling pathway. Additional comprehensive genetic studies in the centenarians should reveal other molecules within the IIS pathway that are important in human longevity. Unprecedented advances in sequencing technology, namely next generation sequencing, will facilitate the confirmation and expansion of the role of IIS pathway in human longevity.
FUNCTIONAL CHARACTERIZATION OF LONGEVITY-ASSOCIATED IIS GENE VARIANTS
The candidate gene approach establishes possible associations between genetic variants in the IIS pathway and longevity outcome. However, association studies typically leave open the question of whether the associated genetic variant is functionally important or serves only as a proxy, a genetic marker co-inherited with the functional allele. To complete a genetic study with solely a statistical end point is unsatisfactory in view of the uncertainties associated with the statistical interpretation. To minimize possible errors and spurious association, an integrated approach is required to assess the functional relevance of gene variations at the molecular level and to overcome the statistical limitations. Moreover, functional variants are known to be rare in frequency (80), and these rare SNPs are not queried in most case–control studies due to the insufficient statistical power (78). In this case, it is critical to directly test their impact by functional analysis. Here, we will describe later the specific approaches to test the functionality of longevity-associated IIS gene variants focusing on the IGF1R variants that we have discovered (76). In principle, these approaches can be applied to other common associated variants or potentially functional rare variants in IIS or other traits. For associated common variants of a candidate gene, a resequencing step is required to discover potentially functional variants before extensive functional analysis (Figure 2).
In Silico Analysis
Potential functionality of SNPs in the associated loci can be determined by in silico analysis. There are a number of computational tools available enabling one to predict the impact of missense SNPs on protein function and structure or regulatory SNPs on gene expression. If appropriately used, these tools greatly reduce the efforts required to test all possible missense SNPs or regulatory SNPs by allowing one to identify and prioritize candidate SNPs before an exhaustive functional assay. For coding SNPs, publicly available tools, such as SIFT (81) and PolyPhen (82), can be utilized to predict whether an amino acid substitution at a particular position in a protein will have a phenotypic effect, and SWISS-MODEL (83) can be used for automated comparative modeling of three-dimensional protein structure. For regulatory SNPs, databases such as TRANSFAC (84), Eukaryotic Promoter Database (85), or MatInspector (86) can be utilized to predict putative transcription factor binding sites and other cis-regulatory elements as well as PupaSNP (87) and ESEfinder (88) for the prediction of exonic splicing enhancers. In silico analysis of putative functionally important SNPs can pave the way for further experimental studies of the associated variants for the effects on the regulation of gene expression or protein function.
In Vitro Cell Culture Models
Cell lines established from human subjects are a useful in vitro model for studying genetic control as they may reflect the functional characteristic of the donor. Lymphoblastoid cell lines established from subjects have been used for quick readouts of the potential outcomes of particular gene variants. Caution needs to be taken as they are known to be oligoclonal in origin and thus may reflect gene actions of B cells at different stages of development from stem cells to plasma cells. The utilization and value of these cell lines for studying genotype-driven molecular endpoints have been demonstrated in functional analysis of the longevity-associated IGF1R variants (76). When treated with IGF-1, lymphoblastoid cell lines established from centenarian carriers of the rare variants, Ala37Thr or Arg407His, showed reduced activities of IGF1R compared with lymphoblastoid cell lines from centenarian noncarriers. However, because complete genotype information for the IGF1R rare variant carriers is not known, functional information obtained from lymphoblastoid cell lines of these subjects is still correlational and cannot establish a cause–effect relationship between the variants and the associated phenotypes, for example, longevity. To determine whether these human longevity-associated IGF1R variants directly affect IGF-1 signaling, mouse embryonic fibroblasts expressing the different human IGF1R variants in a mouse Igf1r null background were generated. When treated with IGF-1, mouse embryonic fibroblasts expressing the human longevity-associated IGF1R variants, Ala37Thr and Agr407His, showed attenuation of IGF-1 signaling, compared with mouse embryonic fibroblasts expressing the wild-type IGF1R. This was demonstrated by a significant reduction in the phosphorylation of both IGF1R and AKT after IGF-1 treatment, reduced expression of many IGF-1-activated genes involved in mitogenesis, apoptosis and differentiation processes, and impaired cell cycle progression (89). These results convincingly demonstrated that the human longevity-associated IGF1R variants are reduced function mutations similar to the IIS mutations causing life-span extension in model organisms (3). The mechanistic parallels between the effects of IIS mutations in model organisms and those produced by longevity-associated IGF1R mutations in humans suggest that evolutionary conservation of IIS reduction and extreme longevity includes humans.
Association studies have provided a sound-basis for the involvement of IIS in human longevity. Common variants in some IIS candidate loci, including AKT and FOXO genes, are associated with longevity. Particularly, the association of FOXO3a variants is robust and has been replicated in many diverse population-based studies. The next challenge is moving from statistical association to functional mechanism. Technological advances have made it possible to use high-throughput sequencing as a primary discovery tool in genetics (72), such as whole-genome sequencing or exom sequencing, specifically for identifying rare variants. The statistical challenge of estimating the influence of rare variants can be overcome by functional analysis of these variants to ascertain their biological significance. Despite successes in identifying IIS genes associated with longevity, the functional variants and their molecular basis have only been determined in onefigure 1 has been cited after reference (8) in the text, “(70, 89)” given in the figure caption has to be renumbered for sequential order and in turn all the reference citations also have to be renumbered. study (89). Discovery of the reduced-function variants in IGF1R that are rare but collectively enriched in centenarians as compared with controls provide the mechanistic parallel of IIS mutations between model organisms and humans. Although performed in short-term cell culture models, these results further support the role of IIS in human longevity. Taken together, a combination of common and rare SNP variants in IIS likely contributes to exceptional longevity.
Understanding the mechanisms by which longevity-associated IIS variants contribute to human longevity will be critically important to facilitate the development of strategies that can delay aging and promote health span. The complexity of aging and longevity phenotypes poses a daunting yet exciting challenge to establish functionality and causality of longevity-associated IIS variants. A variety of experimental approaches in multiple model systems are needed to elucidate the functional consequences of gene variants. For regulatory variants, in vitro reporter assays can be utilized to characterize the variants in a promoter or enhancer region (Figure 2B). Genotype–phenotype relationship can be directly assessed in patient-derived cells as described for lymphoblastoid cells. Recently, induced pluripotent stem cells generated from patient-derived primary cells such as fibroblasts present potential opportunities for human gene variant modeling to interrogate the cellular and biochemical consequences of gene variants (73). The field is poised to make important advances.
Eventually, the in vivo roles of longevity-associated IIS variants will have to be determined especially for their impact on life span. Classically, the generation of knock-in models expressing human gene variants in model organisms has been explored and mice have been the model system par excellence for human genetic variants. However, caution needs to be taken in interpreting the results because human phenotypes may not be successfully replicated in mice owing to fundamental biological differences between the two species. Hopefully, identification and elucidation of molecular pathways leading to exceptional longevity in humans could provide a rational basis for intervention strategies that slow aging and delay the onset of age-related diseases.
National Institutes of Health (AG024391, AG027734, and AG17242).
C.T. is a recipient of Ellison/American Federation for Aging Research postdoctoral fellowship. Y.S. is the recipient of a Glenn Award for Research in Biological Mechanisms of Aging.