Genetics of human longevity: From variants to genes to pathways

The current increase in lifespan without an equivalent increase in healthspan poses a grave challenge to the healthcare system and a severe burden on society. However, some individuals seem to be able to live a long and healthy life without the occurrence of major debilitating chronic diseases, and part of this trait seems to be hidden in their genome. In this review, we discuss the findings from studies on the genetic component of human longevity and the main challenges accompanying these studies. We subsequently focus on results from genetic studies in model organisms and comparative genomic approaches to highlight the most important conserved longevity‐associated pathways. By combining the results from studies using these different approaches, we conclude that only five main pathways have been consistently linked to longevity, namely (1) insulin/insulin‐like growth factor 1 signalling, (2) DNA‐damage response and repair, (3) immune function, (4) cholesterol metabolism and (5) telomere maintenance. As our current approaches to study the relevance of these pathways in humans are limited, we suggest that future studies on the genetics of human longevity should focus on the identification and functional characterization of rare genetic variants in genes involved in these pathways.


Introduction
As humans, we are all ageing, but the exact definition of ageing is still heavily debated.However, in the field of ageing biology, most scientists would agree that the dysregulation of biological mechanisms affecting the genome, proteome or any other biological molecule is part of the definition [1].There is also discordance in the definition of the phenotypes that are commonly used to define and study the ageing process.Lifespan is defined as the number of years lived [2].Yet, for the definitions of healthspan and longevity, several slightly varying interpretations exist in the literature as the traits are more subjective and less easy to quantify.Healthspan is often defined as the number of years lived in good health or the number of years lived before the occurrence of any major debilitating disease [3,4].For the phenotype longevity, survival to an exceptionally old age is used, which is either defined as survival above a defined age, for example 95 years [5], or belonging to a birth-cohort-specific survival percentile, for example 10% longest lived [6].It will be a challenge for the field to delineate these definitions in the future to generate data that is comparable between different studies.Fortunately, the field of ageing biology agrees that ageing itself represents the major risk factor for the occurrence of most debilitating chronic diseases, such as cardiovascular diseases, metabolic diseases -for example type 2 diabetes -and neurological impairments, including Alzheimer's disease [7].Furthermore, ageing increases the risk of suffering from severe outcomes from infections as the function of the immune system declines with increasing age, resulting in ineffectiveness in fighting pathogens and detecting and destroying cancerous cells [8,9].Unfortunately, this decrease in immune function also results in a decreased effectiveness of vaccines that therefore offer less protection in aged individuals [10].
Many of us would like to live a long life, but, currently, that often means spending the last decade or two of one's life suffering from multiple debilitating diseases [2].Consequently, the advancements in medical care to prevent death due to chronic diseases (i.e.cancer and type 2 diabetes) or infections (i.e. through improved hygiene and the development of vaccines and antibiotics) and the subsequent increase in lifespan now urgently need to be matched by major advancements that increase healthspan [11].The most eminent challenge of the ageing field is to identify how we can improve health during ageing to enable people to live long and productive lives without posing a burden on the healthcare system.Thus, the incentive of current studies should not be to increase lifespan but to focus on the extension of healthspan.To this end, we need to utilize all possible tools to investigate how we can manipulate the healthspan of an individual via, for example, the environment, nutrition, microbiome and physical activity (i.e.exercise habits).Although some of these interventions may work in broad populations, the success of other interventions may depend on an individual's genetic makeup.Consequently, a personalized approach might be necessary to address the challenges of an ageing population [12,13].
The traditional approach in medicine is to study the unhealthy individuals within the population to find out what contributes to, or even causes, their disease and develop a cure or treatment for this particular disease.However, it might be more beneficial to study the other side of the spectrumin this case, the extremely healthy and long-lived.This approach may enable us to understand what we can do to not only cure diseases but also push the entire population to stay healthy in old age.Exceptionally long-lived individuals often delay or even escape the onset of most age-related diseases [14,15], thus making longevity an interesting naturally occurring case study for healthy ageing within the human population [16,17].Moreover, longevity is assumed to be a heritable trait [6].Studying the genetic make-up of the exceptionally long-lived may thus help us uncover the underlying molecular mechanisms that could be targeted to improve healthspan.
The focus of this review will be on mechanisms shared between humans and model organisms to understand what we can learn from the genetics of longevity.First, we will discuss the findings and limitations of the genetic studies performed in humans, with a focus on studies that included an extensive and well-defined study population (n ≥ 200 long-lived individuals).Second, we will describe how we can utilize model organisms and the evolved differences in the natural lifespan among them (with a focus on longevity, not median lifespan, in the context of the species) to further deepen our understanding of ageing and longevity in humans.

Heritability of longevity
Studies of long-lived families reveal that longevity can be inherited as a genetic trait [18][19][20].Although the heritability of longevity has not yet been estimated, as a placeholder, the heritability of lifespan -determined in twin and pedigreebased studies -is estimated to be between 10% and 30% [21].The success of studies investigating the genetic component of longevity is dependent on the selection criteria used to select long-lived cases.However, because the meaning of longevity as a trait is not well defined, the use of heterogeneous study cohorts to investigate longevity unfortunately complicates the interpretation of the results.Based on the analysis of two large familybased datasets, the most refined selection criteria for individuals to be considered exceptionally longlived is that they belong to the top 10% of survivors of their birth cohort and have additional family members who meet the same criteria [6].These selection criteria were further expanded by taking into account that the newly defined longevity phenotype is inherited in at least 20% of family members for at least two subsequent generations and used to create a 'longevity relatives count' score [22].An alternative longevity score that has been used in the field is the 'family longevity selection score' [23].However, thus far, both scores have only been applied to a limited number of populations, and more research is needed to prove their usability for the selection of genetically enriched long-lived cases.

Current approaches to studying the genetics of longevity in humans
Most genetic differences between humans are caused by single-nucleotide polymorphisms (SNPs), whereas other, less common, changes  [44] can affect multiple nucleotides, for example insertions [24].The primary aim of genetic studies of exceptionally long-lived individuals is to detect genetic variants that are significantly enriched or depleted in their genome in comparison to controls from the general population.The earliest genetic studies of longevity employed a hypothesis-driven approach and focused on investigating SNPs in specific genes or pathways that were previously linked to lifespan regulation in model organisms or age-related diseases in humans.The development of genotyping arrays enabled the investigation of thousands of SNPs in multiple genes at the same time using more unbiased hypothesis-free approaches.Recent advances in sequencing technologies now allow for the investigation of the whole exome or even the entire genome [25][26][27].
The most extensive genetic studies of human longevity have been performed in cohorts from populations of European, East Asian and Ashkenazi Jewish ancestry, as summarized in Table 1.
In the following, we will elaborate on some of the findings from the most rigorous studies that were more likely to identify relevant SNPs, loci and genes due to the stringent inclusion criteria of long-lived cases.An overview of all genetic variants and genes associated with human longevity is available in the LongevityMap database (http://genomics .senescence.info/longevity/)[28].

Candidate gene-and pathway-based studies
Candidate gene-or pathway-based studies are hypothesis-driven investigations of the association between genetic variants in a gene, or multiple genes within a pathway, with a given phenotype, such as longevity.The selection of the genes is usually informed by prior knowledge of the connection between the gene and the phenotype, that is from studies in model organisms.Candidate gene-based studies were the first studies conducted to investigate the genetics of human longevity, as this approach is relatively inexpensive and methodically less challenging than hypothesis-free approaches given that only the variants of interest need to be measured.
Gene-based studies.One of the earliest candidate gene-based studies of human longevity was performed by Schachter et al. [32], who were able to identify an association of genetic variation in the APOE locus with longevity.Until today, this is the best replicated longevity locus, and its association has been confirmed in many different study populations [45][46][47].There are two main genetic variants at this locus implicated in longevity: ApoE ε2 (rs7412), which is enriched in long-lived individuals, and ApoE ε4 (rs429358), which is depleted in long-lived individuals.The only other longevity-associated locus that has been replicated in several independent studies is FOXO3, which was first identified by Willcox et al. [40] and later replicated by several other groups [30,34,36,37,[48][49][50]. Additional loci have been identified using candidate gene-based studies (see the LongevityMap database for an overview), but these could so far not be replicated in independent studies and/or populations.
Pathway-based studies.As genetic variants within different genes that belong to the same pathway may result in a shared downstream effect, the candidate gene-based approach was subsequently broadened to investigate multiple genes within, or even entire, lifespan-associated pathways.The pathways that were found to be relevant and were further investigated are as follows: (1) insulin/insulin-like growth factor 1 (IGF-1) signalling (IIS) [51], (2) mTOR signalling [52], (3) DNA-damage and repair [53] and (4) telomere maintenance [51] (Table 2).
The first study of this kind by Deelen et al. focused on 1021 SNPs in 68 genes assigned to the IIS pathway and 88 SNPs in 13 genes that are part of the telomere maintenance pathway to investigate the synergistic effect of the variants on longevity.Both pathways showed significant association with longevity.The association of the IIS pathway was distributed across nine genes (AKT1, AKT3, FOXO4, IGF2, INS, PIK3CA, SGK, SGK2 and YWHAG), whereas the gene POT1 by itself was sufficient to explain the association of the telomere maintenance pathway with longevity [51].The other extensive gene-set analysis study by Debrabant et al. was focused on an in-depth investigation of the DNA-damage response and repair pathway by testing the association of 592 SNPs in 77 genesthat can be categorized into 9 sub-pathways -with longevity [53].Although none of the tested associations passed adjustment for multiple testing, and the findings could not be replicated in an independent dataset, it is worth mentioning that the base excision repair sub-pathway was nominally associated with longevity due to genetic variation in several genes, whereas the nominally significant Genetics of Healthy Ageing [18] 17q12-q22 Genetics of Healthy Ageing [18] 19p13.3-p13.11Genetics of Healthy Ageing [18] 19q13.11-q13.32APOE (ApoE ɛ2 and ɛ4) Genetics of Healthy Ageing [18] Note: Gene names in bold indicate all homologous genes that were found in at least two independent studies by a different approach.Underlined gene names indicate genes that were found in at least two independent studies but in the same model organism.
association of the homologous recombinational repair and RecQ helicase activities pathways was primarily the result of association of only one gene (RAD52 and WRN, respectively).
The most recent pathway-based study, which used a SNP-SNP interaction approach, was performed on 1058 SNPs in 140 genes by Dato et al. [54].The focus of this study was on the IIS, DNA repair and pro/antioxidant pathways.They identified several SNP-SNP interactions associated with longevity, and the strongest ones involved genetic variants in TP53 (indicating an interaction between the DNA repair and pro/antioxidant pathways) and GHSR (indicating an interaction between the IIS and DNA repair pathways).

Linkage studies
With the development of genotyping arrays, the focus of the field subsequently shifted to hypothesis-free approaches -such as linkage studies -to identify genetic loci that are associated with longevity.Linkage studies of longevity are aimed at the identification of chromosomal regions that show cosegregation within long-lived family members and thus require a family-based study design.Contrary to the initial expectations, only very few loci were identified through such studies (Table 3), which is likely due to their relatively small sample size.In addition, none of the identified loci did, thus far, replicate between studies [18,55].

Genome-wide association studies
Another hypothesis-free approach that is now often used in the field is the identification of SNPs using genome-wide association studies (GWAS).GWAS that used longevity as the phenotypic read-out were able to identify several variants that are associated with this trait, as depicted in Table 4.However, the only genetic variants that are successfully replicated between independent studies and populations are the ones located in the APOE locus, which were already identified using candidate gene studies.Robust identification of additional variants is currently simply hampered by the limited number of cases -that is long-lived individuals -available for genetic studies.GWAS of longevity are further complicated by the absence of a proper control cohort, as the matching control individuals (i.e.those from the same birth cohorts as the cases) have already died before being collected.As a workaround to address some of these hurdles, and to further increase the number of detected variants, several recent studies combined genetic studies on longevity with those of other ageing-related phenotypes [59,60].

Polygenic score analyses
In addition to the study of SNPs, several studies have looked at polygenic scores (PGSs) to determine the enrichment/depletion of age-related disease-associated variants in long-lived individuals.Early studies showed that long-lived individuals have a similar burden of common diseaseassociated genetic variants [27,71,72], but more recent studies have shown a consistent depletion of PGSs for Alzheimer's disease [73,74].Hence, this could partly explain why many long-lived individuals remain cognitively healthy up until very high ages [31].However, it is still very likely that the extreme lifespan of long-lived individuals is primarily determined by variants promoting health, given that sequencing studies show that long-lived individuals seem to carry many pathogenic variants [25,[75][76][77].

Future directions of genetic studies of longevity in humans
Although the number of genetic studies has dramatically increased over the past decades, the number of identified genes that are reliably associated with human longevity has stagnated (as depicted in Fig. 1).Only two genes (APOE and FOXO3) have consistently been linked to longevity in studies of different designs and using data from different populations.Even efforts to combine the data from different studies -that is metaanalyses -were not able to significantly increase the number of longevity-associated genes.The steady increase in the ratio of sequenced long-lived individuals over identified variants supports the hypothesis that the genetic component of human longevity is likely not encoded by common genetic variants with large effect sizes, but rather by rare variants that either enable longevity autonomously or in concert.Given that rare variants can hardly be detected with our current approaches, the field needs to adapt its strategy to be able to study their effects on longevity.One way forward would be to massively increase the sample size of the genetic studies to allow the detection of variants with a small effect size, which would only be possible by setting up new longevity studies.Although this effort might eventually lead to success, this strategy is not practicable given the high costs to set up such studies and the limited number of individuals who are currently considered long-lived.Revisiting the candidate gene-/pathway-based approach offers an alternative.However, instead of focusing on common variants (as has been done previously), these studies should focus on the identification of rare variants in genes and pathways that have been associated with lifespan regulation in model organisms.The most critical step is to restrict the number of variants to be considered for subsequent experimental analyses to variants located in the most promising genes.The selection of such genes can be accomplished by only including genes belonging to pathways that are associated with longevity in multiple model organisms.
As the identified variant will be rare and the association of the variants with longevity is hard to test [78], the causality between the variant and longevity must be confirmed in the laboratory [79].Hence, it is advisable to only focus on variants in regions of the genome that are of known function and can therefore be reliably investigated (e.g.exons or enhancer and suppressor sequences).In the case of variants residing in exons, it would also be sensible to only investigate protein-altering variants to improve the likelihood of a functional consequence of the variant.We will next discuss some results from studies that already successfully applied a functional characterization approach.

In vitro characterization of genetic variants
Genetic association studies offer a fantastic tool to identify the mechanisms that facilitate longevity.However, most studies stay short of exploring the actual functional consequence of genetic variants and, consequently, establishing causality [80,81].Given that many of the identified variants cannot be confirmed in independent studies and populations of different ethnicities, one might wonder if they are indeed causal.Fortunately, previous studies have shown that functional characterization of the effects of longevity-associated genetic variants is possible in vitro [5,82,83].
Common genetic variants.To our knowledge, the common longevity-associated variants that were so far functionally characterized are the ApoE ε2and ε4-defining variants rs7412 and rs429358 [84], respectively, rs2802292, rs12206094 and rs4946935, all located in the intronic region of FOXO3 [5,83], as well as several regulatory variants in PRKCH, CLU and NFKBIA [85].The investigation of intronic variants poses its challenges and limitations, as intronic regions are usually less conserved between species, if at all, impeding the possibility of using in vivo models to functionally investigate these variants.The functional study of two of the common intronic variants in FOXO3 -rs12206094 and rs4946935 -was consequently performed by transiently introducing a luciferase reporter construct under the control of the variantspecific FOXO3 promoter sequence into the human Panc1 and Jurkat cell lines.Both variants seem to increase the expression of FOXO3 in normal cell culture conditions but result in a decrease in expression upon IGF-1, but not insulin, stimulation [5].To determine the function of rs2802292, Grossi et al. used a more elegant approach by introducing the variant into the near-haploid human embryonic stem cell line HAP-1 via CRISPR-Cas9 in addition to the use of established and primary human cell lines.They showed that this variant increases the expression of FOXO3, which is even further enhanced by the exposure of cells to stressful conditions, due to the creation of a binding site for the heat shock transcription factor HSF1 [83].
Rare genetic variants.As mentioned above, experimental validation of rare variants is even more important than for common variants due to the absence of statistical power to associate them with longevity.This approach has already successfully been applied in the case of two rare variants in the coding region of the IGF1R gene -rs777765504 (Ala67Thr) and rs34516635 (Arg437His) [82,86].
The investigation of the IGF1R variants in Igf1r null mouse embryonic fibroblasts by a rescue approach showed that the variants lead to reduced activity of IGF1R and downstream signalling (phosphorylation of AKT) upon IGF1 stimulation, resulting in reduced transcript levels of IGF1 signallinginduced genes [82].A reduction in IIS pathway activity has previously been linked to lifespan extension in model organisms [87], supporting the conclusion that the identified rare variants in the IGF1R may indeed be causal in extending lifespan.A similar approach has been successfully used to show a functional effect of rs183444295 (Ala313Ser) and rs201141490 (Asn308Lys) in SIRT6 [88].However, it has to be noted that these variants were not significantly enriched in long-lived individuals, making it less convincing that they contributed to the longevity of their carriers.

In vivo characterization of genetic variants
The above-mentioned functional studies pave the way for further investigations of other longevityassociated common variants, as well as rare genetic variants identified from whole-genome and exome sequencing of long-lived individuals.Given that some of these (newly identified) variants might be conserved in model organisms, investigation of their in vivo effect could further elucidate whether the variants can indeed induce a healthier ageing phenotype and an extension in lifespan [79].
The first studies using this approach are currently underway.

Supporting evidence from studies in model organisms
To improve our selection of the most promising lifespan-associated genes, we need to combine all possible information from studies in model organisms, with a focus on evolutionarily conserved pathways, and ideally supporting evidence from human studies.This knowledge can subsequently be used to screen the genetic data obtained from long-lived individuals to identify (rare) genetic variants in these candidate genes.In the past decades, studies in model organisms have investigated various genes and their potential to increase lifespan.This research was initiated by the landmark study by Kenyon et al., who showed that mutations in daf-2 -the mammalian homolog of the insulin/IGF-1 receptor familycan more than double the lifespan of nematode worms (Caenorhabditis elegans) [89].This finding that genetic manipulations of genes involved in the IIS pathway result in lifespan extension was subsequently replicated in fruit flies (Drosophila melanogaster) and mice (Mus musculus) [90,91] and opened up the field for investigation of other conserved pathways and their potential to regulate lifespan.In the next sections, we have utilized the results from genetic studies in multicellular model organisms -specifically mice, fruit flies and nematode worms -to determine which pathways are enriched for longevity-associated genes.We also summarized the pathways and genes identified in studies using comparative genomic approaches.
Pathways that are enriched for longevity-associated genes in model organisms.To methodically determine the most relevant and conserved longevityassociated pathways that are based on experimental approaches, the longevity-associated genes for mice, fruit flies and nematode worms were collected from GenAge (https://genomics.senescence.info/genes, 4th of May 2022).All 'pro-longevity' and all 'anti-longevity' genes were used, and only unannotated genes were excluded.The gene list was uploaded to Gene Ontology (http://geneontology. org/), and the annotation dataset PANTHER Pathways was employed to enrich for pathways.The significantly enriched pathways were compared between the three organisms and reduced to the pathways that were found to be significantly enriched in all of them (Table 5).
Based on the pathway enrichment of the longevityassociated genes, the following overarching pathways could be identified: (1) cellular stress, (2) IIS, (3) endothelin signalling, (4) apoptosis signalling and (5) immune function.It should be noted that the main limitation of this analysis is that many of the lifespan studies in model organismsespecially in mice -are biased, as they were based on initial findings in lower organisms.Hence, the pathways that are only relevant for higher organisms are not picked up using this approach.
Pathways and genes associated with lifespan through comparative genomics.Many genes that are essential for the function of an organism have evolved in a shared ancestral organism and are thus largely conserved across species.We assume that genes -or at least the underlying molecular mechanisms and pathways -that govern ageing also follow this pattern.Hence, we can utilize comparative biology to learn more about human longevity.An example of this is provided by the IIS pathway (see above).The investigation of the variation in maximum lifespan between closely related animals may thus enable us to determine common mechanisms inherent to multiple long-lived species.Based on a comparative genomic study among mammals, reptiles and amphibia, de Magalhães and Toussaint concluded that the type of ageing observed in humans likely evolved early in mammalian evolution, probably as a side effect of another mammalian-specific trait, and should therefore obey similar mechanisms in all mammals [92].The extended lifespan observed in some mammals -such as naked mole rats, elephants, bats and whales -is likely caused by adaptations that were independently attained in these species [92].Consequently, shared biological adaptations that increased the lifespan of other mammals might also translate to humans.However, the question remains how longevity evolved on the genomic level?Genes that have the capacity to increase lifespan are assumed to be more conserved in longlived species [93].However, in the context of evolution, it would also make sense that the perturbation of genes that are highly conserved in longlived species is so stringently conserved because these genes cannot tolerate perturbation and, as a consequence, would shorten lifespan if mutated.Another hypothesis is that ageing is conserved in mammals, and as longevity is positively correlated with the complexity of the intraspecies relationships -meaning that in social species living in groups, longevity provides a fitness benefit to the survival of the population -it is consequently an adaptation [94].In this context, shared -or at least similar -variations in genes of long-lived species that are otherwise conserved in all mammals are of interest [95].If longevity is indeed an adaptive mechanism, then the focus of the field should shift to genes that are rapidly evolving as a consequence of positive selection [96].The findings from com-parative genomic studies that investigated different mammals at the family or species level that evolved longevity in comparison to their short-lived cousins are summarized in Table 6.

Most prominent longevity-associated candidate pathways for future studies on human longevity
Condensing all longevity-associated genes down to the ones that were implicated by at least two of the above-described approaches in either humans or model organisms results in a list of 18 genes (Table 7).These remaining genes can be assigned to five overarching pathways: (1) IIS, (2) DNAdamage response and repair, (3) immune function, (4) cholesterol metabolism and (5) telomere maintenance.Eight of the 16 genes -namely AKT, FOXO, IGF, INSR, IRS, PIK3C, RPS6KB/S6K and SHC -can be grouped as part of the IIS pathway, although some of these genes -such as AKT and PIK3Care also part of other pathways, such as immune function.The remaining genes are characteristic for the four other pathways -with BLM, ERCC, RAD and XRCC being part of the DNA-damage response and repair, and IL6 and ADCY being genes relevant to immune function.Finally, APOE and POT1 belong to the pathways of cholesterol metabolism and telomere maintenance, respectively.
Based on this summary, the IIS pathway (Fig. 2) is the most interesting to investigate further in humans as it comprises multiple genes that are directly linked to longevity.Previous candidate pathway studies based on common [51,52], as well as rare, genetic variants [98] have already provided some evidence that the IIS pathway is indeed relevant to human longevity.Consequently, further investigation of this pathway -including a focus on rare genetic variants coming from exceptionally long-lived humans -would likely prove fruitful.Furthermore, these findings are supported by several studies that have shown that non-genetic interventions -for example dietary restriction or pharmacological treatment (i.e.rapamycin)can extend lifespan in model organisms through manipulation of the IIS pathway [99].All genes that were identified in Table 7 are highlighted in the simplified version of the IIS pathway (Fig. 2).
Although the genetic evidence based on the current literature is weaker, the other four pathways could also be of interest in future model organismand human-based studies (Table 7).There is some evidence for the involvement of these pathways in the regulation of lifespan across species, including

FOXO4 GHR GHRH IGF1 INSR IRS1 PIK3CB
PTPN1 SHC1 36 mammalian species [96] Telomere maintenance POT1 57 mammalian species [95] Note: Gene names in bold indicate all homologous genes that were found in at least two independent studies by a different approach.Gene names in red indicate human orthologues that were identified in human and model organismbased studies.For clarity reasons, we only included genes that were part of the 1157 validated genes by Farre et al.
(2021) [95] and that were already reported in pathways identified in other studies.
human longevity.It is clear from the literature that manipulation of the DNA-damage response and repair [100], the immune function [101] and telomere maintenance pathways [102] can result in changes in lifespan.Cholesterol metabolism, on the other hand, is strongly linked to human longevity via APOE.This pathway is unique in that no model organism-based genetic study has implicated cholesterol metabolism in the regulation of lifespan, which is likely due to the unique role the different APOE isoforms play in humans [45,103].

Conclusions
The genetic context, as well as the environment, plays a pivotal role in regulating the effect of a genetic variant on lifespan.In the absence of biomarkers that can reliably predict age at death, we have limited options to further investigate additional possible rare variants and enhance our understanding of the part genetics plays in facilitating human longevity.Ideally, we would design the perfect study cohort, enrol people when they are young and start longitudinal measurements of a high number of parameters until the death of all participants.At this point, we could go back and identify who became long-lived and who did not.Given that we enrolled enough participants that would eventually become long-lived, this data would give us the power to truly investigate longevity.However, this kind of study would take a massive amount of money, labour and -especially -time to conduct.The data collected by the UK Biobank might be close to such a dataset in the future, but, currently, this kind of data is not available.Consequently, we will need to continue with the current approaches and try to increase the statistical power to allow us to detect additional variants (likely with little success), or we can aim to identify rare genetic variants and functionally validate the causality of these variants, as described for FOXO3 and IGFR1.As each individual carries many rare variants and enrichment will not be possible due to the rarity of these variants, it will be essential to prove causality by introducing these variants into model organisms to investigate their effects in vitro and in vivo [79].
To select the most promising genetic variants, we can combine the knowledge from all studies with longevity as a primary outcome.This approach is likely the most feasible option we have until more human data that allows for the detection of rare genetic variants becomes available.
Hence, our aim with this review was to collect all currently available genetic evidence from studies in humans and model organisms -including comparative genomic approaches -to guide the selection of the most promising pathways, genes and loci.As a secondary objective, we were hop-  5) telomere maintenance (POT1), all of which had been previously implicated to play a role in human longevity.On the one hand, this result could imply that all relevant pathways for human longevity with strong effect sizes have already been identified, and we should continue to focus on identifying common and rare genetic variants in these pathways in exceptionally long-lived individuals to determine how exactly the pathway needs to be tweaked to extend human lifespan.On the other hand, our approach may be inherently biased because specific genes and pathways are more researched in the context of longevity, resulting in an overrepresentation of those, and could have other shortcomings, which are already discussed in-depth by de Magalhães and Toussaint [104].
Longevity studies in humans have so far only been able to consistently associate APOE and FOXO3 with longevity despite a significant increase in the genetic data available.This suggests that longevity and healthy ageing cannot be explained by a few common variants with high penetrance but rather by many rare variants with potentially small effects that act synergistically.Additionally, the design of the human study population itself poses at least two major shortcomings.As mentioned before, most studies use genetic data from younger related or unrelated individuals who have not yet died -that is in theory, these individuals could still become long-lived -and all of the 16 study populations included in this review are of European, East Asian or Ashkenazi Jewish decent.Therefore, this data needs to be interpreted with caution, and the meaningfulness of the results for the entire human species is limited.It is possible that some of the identified variants are not as unique as we believe and might be detected in short-lived individuals once genetic data of more diverse populations becomes available.
The pathway enrichment based on the longevityassociated genes from genetic manipulations in model organisms resulted in the identification of five major pathways, out of which two were also indicated to be relevant in humans.However, it needs to be considered that large genetic screens are not possible in mammals, and findings from non-mammalian models are less likely to translate to humans.Furthermore, genes investigated in rodent models were, at least in the past, often only investigated in one sex and a specific strain of limited genetic diversity, making the results vulnerable to bias.Consequently, the results may not be translatable to rodents of diverse genetic backgrounds, let alone humans.It also needs to be kept in mind that human ageing might be different from ageing in other mammals, and that some genes and pathways might be uniquely relevant to human ageing and diseases, such as APOE.Therefore, these genes and pathways cannot be detected by experiments performed in model organisms.The comparative genomics approach to determine variants that differ between naturally occurring long-lived and short-lived mammalian species is less hypothesis-driven compared to the results from model organisms and is therefore more likely to result in the identification of novel genes and pathways.We found four papers that met our inclusion criteria -namely that they included genomic data of several mammals with different lifespans -and reported the specific genes and pathways implicated in facilitating the extended lifespan of the comparatively longlived mammalian species.Surprisingly, we identified only one pathway -DNA-independent cell cycle regulation -that was not previously reported to be associated with longevity.The other four pathways are (1) IIS, (2) DNA-damage response and repair, (3) immune function and (4) telomere maintenance, which are well known to play a role in longevity and lifespan regulation.Although it was out of the scope of this review to link the available genetic data with other omics-based approaches, the combination of different omics datasets may enable further insight into how human ageing and longevity are regulated.Unfortunately, for most of the human study cohorts, only genetic data was collected and very sporadically metabolomic, transcriptomic or epigenetic data exists for the same individuals.
Although we cannot exclude context-dependent effects related to the experimental set-up, environment or the genetic background of any of the used study populations, we still think that the five identified pathways offer a promising starting point for further investigation of human longevity.We hope that this review will be useful in finding new ways to identify potential longevity-associated variants in already existing datasets.Furthermore, we want to emphasize the importance of combining bioinformatic knowledge with experimental studies to establish the causality of rare variants in facilitating longevity or healthy ageing and to investigate the underlying molecular mechanisms.The knowledge of the potentially shared underlying mechanisms may empower bioinformatic studies and the development of interventions and therapeutics that target the underlying cause of functional decline and the occurrence of diseases with advancing age.

Fig. 1
Fig. 1 Timeline of publications investigating genetic variants associated with longevity.Summary of all relevant publications of candidate genes (focused on APOE and FOXO3 and whole pathways), linkage and genome-wide association studies.The dramatic increase in the combined number of investigated long-lived individuals is not accompanied by a noticeable increase in the number of identified variants replicated among different populations.

Fig. 2
Fig. 2 Simplified depiction of the conserved insulin/insulin-like growth factor 1 signalling pathway in humanswith all longevity-associated genes highlighted.Summary of all genes for which homologous genes were found in at least two independent studies by a different approach.Gene names in bold indicate all homologous genes that were found in at least two independent studies by a different approach.Gene names in red indicate human orthologues that were identified in human and model organism-based studies.Gene names in blue indicate human paralogues that were identified in human and model organism-based studies.

Table 1 .
Summary of some of the most extensive cohorts that were designed to investigate the genetics of human longevity.

Table 2 .
Summary of results from candidate pathway-based studies of longevity in humans.

Gene names in bold indicate
all homologous genes that were found in at least two independent studies by a different approach.Gene names in red indicate human orthologues that were identified in human and model organismbased studies.

Table 3 .
Summary of results from linkage studies of longevity.

Table 4 .
Summary of results from genome-wide association studies (GWAS) of longevity.
Note: Gene names in bold indicate all homologous genes that were found in at least two independent studies by a different approach.Underlined gene names indicate genes that were found in at least two independent studies but in the same model organism.Gene names in red indicate human orthologues that were identified in human and model organism-based studies.Abbreviation: SNP, single-nucleotide polymorphism.

Table 5 .
Summary of pathway enrichment for longevity-associated genes in model organisms.

Table 5 .
Gene names in boldindicate all homologous genes that were found in at least two independent studies by a different approach.Underlined gene names indicate genes that were found in at least two independent studies but in the same model organism.Gene names in red indicate human orthologues that were identified in human and model organism-based studies.

Table 6 .
Summary of results from comparative genomics-based studies of longevity in mammals.

Table 7 .
Summary of results from genetic-based studies of longevity in humans, model organisms and mammals.

Table 6
For all genes listed homologous genes were found in at least two independent studies by a different approach.Underlined gene names indicate genes that were found in at least two independent studies but in the same model organism.Gene names in red indicate human orthologues that were identified in human and model organism-based studies.