A belief that ageing and longevity are governed by genetic factors has led to growing excitement that research on the human genome will soon uncover the genes for ageing and perhaps open new paths to longer life and health spans. Even if direct gene modification is remote, a clearer understanding of the pathways regulated by such genes may point the way to nongenetic interventions that exploit this knowledge. But what is the evidence that genes do control ageing and how realistic is it to expect that the ‘new genetics’ can secure for us a modern-day elixir of youth? And how can we accommodate the genes responsible for ageing within the framework of natural selection, when surely the decline in vitality that results from the ageing process would appear to run counter to the principle of maximizing Darwinian fitness?
Several lines of evidence support the idea that genes affect ageing. First, there is the obvious fact that different species have different lifespans, and where better to look for the underlying causes than their genomes? Secondly, there are clear lifespan differences between different inbred strains of laboratory animals, such as mice and rats. Thirdly, studies of longevity records in human populations, particularly those comparing the similarity of lifespans for monozygotic and dizygotic twin pairs, have revealed clear heritability in life expectancy . Fourthly, and the subject of much recent research, simple organisms like fruit flies and nematode worms have revealed a range of gene mutations that markedly affect the length of life. The evidence for a genetic contribution to ageing is therefore compelling, but intriguingly all these studies point to only weak genetic specification of individual lifespan. In humans, genes account for only about a quarter of what determines individual length of life . In other species, the picture is much the same . So what kinds of genes control longevity, and how come they do so to such limited extent?
Ageing is widespread amongst animal species but by no means universal, and not all species age in the gradual way that we do. Some organisms, like the freshwater Hydra, show no signs of ageing at all . Others like the Pacific salmon age all at once, just as soon as their once-in-a-lifetime chance of reproduction has come and gone. In the case of the Pacific salmon, the rapid postreproductive death of the adult appears to be driven by sex hormones. If a salmon has its gonads removed, it cannot of course reproduce, but it lives much longer .
Understanding why and how ageing evolved tells us much about the nature of the genes that are involved and how these shape the life history, particularly the relationship between reproduction and survival. In fact, we are discovering that each of the diverse life-history patterns seen in nature can be understood as variants on a single theme. This theme is that the power of natural selection to contain and control the tendency for systems to become disrupted through damage imposed from within and without and from intrinsic errors in metabolism is limited. In effect, ageing occurs because to an important extent, genes evolved to treat the body as disposable. From this central insight follow essential implications concerning the mechanisms underlying age-related frailty and disease.
Genetics and intrinsic sources of phenotypic variance
To understand why natural selection apparently fails to control the intrinsic deterioration of the body, it is helpful to consider the general problem of how organisms constrain the intrinsic sources of variation in the phenotype. The striking feature of the genetic control of metabolism and development is that systems work with a remarkable degree of reliability and reproducibility. This is all the more striking when we note that at the molecular and cellular levels, the potential for disruption is ever present. In spite of the many things that can and do go wrong, the overall process of morphogenesis is remarkably reliable. Nevertheless, small early perturbations can result in large effects, and there is evidence of significant organ size variations amongst genetically identical organisms . In addition to variations in the end-points of development, such as adult organ size, there can be significant variations in the timing of key developmental events, such as puberty. Such variations can affect outcomes as fundamental as the size of the ovary or hippocampus. For example, the size of the ovary in newborn mice of the same strain may vary threefold [7, 8], and there can be differences of 10–20% in the sizes of the hippocampi in human twin pairs .
Underlying the more visible phenotypic variations, there is a variety of sources of intrinsic variability within organisms (Table 1). At the molecular level, subcellular processes are subject to diffusion, which generates variation associated with the random Brownian motion of molecules that will be superimposed on any directional trafficking. Transcription in eukaryotes is controlled by multiple transcription factors, each of which may be present only in very small numbers within the cell. Many genes are transcribed to produce only a few mRNAs per cell, which can cause large statistical fluctuations in biosynthesis. There is growing evidence of the intrinsic stochastic nature of gene expression and macromolecular biosynthesis [10, 11].
|Level of effect||Timing of effect|
|During development||During adulthood|
Errors in biosynthesis
Modification (e.g. glycosylation, cross-linkage)
Damage (e.g. oxidation)
Diffusion (Brownian motion)
Errors in biosynthesis
Modification (e.g. glycosylation, cross-linkage)
Damage (e.g. oxidation)
Diffusion (Brownian motion)
|Cellular||Differentiation (cell fate)|
Random segregation of constituents between daughter cells at division
|Differentiation (cell fate)|
Random segregation of constituents between daughter cells at division
Accumulation of metabolic wastes
|Organ/system||Cell number variation|
Extent of asymmetry
Injury due to toxins, infection, trauma
Genome instability, which results in somatic mutations and chromosomal abnormalities, is another important source of intrinsic variation. For example, in ageing mice, mutation frequencies as high as 10−4 per gene per cell have been reported [12, 13]. Epimutations may also occur through loss or disruption of DNA methylation patterns, affecting gene expression . Genome instability also manifests as mitochondrial DNA point mutations and deletions. These aberrations arise at a 10-fold greater rate than nuclear mutations and accumulate in certain tissues throughout life [15, 16], likely to cause impaired generation of cellular energy.
Abnormal proteins, which may resist proteolysis and accumulate over time, arise regularly as a result of damage, misfolding, denaturation, post-translational modification or errors in biosynthesis . Synthesis errors occur at rates high enough that a sizeable fraction of newly synthesized proteins (particularly larger proteins) are predicted to contain at least one sequence error. As a consequence of failures in protein homeostasis, aggregates may form which resist turnover by normal, ubiquitin-mediated, proteasome functions .
In all, intrinsic sources of phenotypic variance are thought to contribute an important part of the variation in lifespan, a clear instance of which is seen in isogenic populations of the nematode Caenorhabditis elegans where, despite having a constant genome and environment, individual worms show strikingly different longevity . Recognition of this essential role of intrinsic variation led Finch and Kirkwood  to propose an extension of the classical population genetics model developed by Fisher, Haldane and Wright, which resolves phenotypic variance (VP) within a population into two terms, that due to heredity (VH) and that due to environment (VE):
For a trait such as nematode lifespan, with sizeable VP and negligible VH and VE, it is that clear something is missing from this model. A third term VC needs to be introduced to account explicitly for intrinsic chance variations in the development of genetically identical individuals within a uniform external environment, i.e.
Optimizing the investment in somatic maintenance and repair
The potential for phenotypic variance, including the variation that arises through damage, is considerable. To reduce the adverse consequences of such variation, it is clear that an extensive array of maintenance and repair systems has evolved. This begs the question, however, of how good these systems should evolve to be. This question needs to be answered from an ecological perspective, which should take account of the fact that animal lifespans in the wild are curtailed mainly by extrinsic sources of mortality (accident, starvation, cold, predation and infection). Somatic maintenance needs only to be good enough to keep the organism in sound physiological condition for as long as it has a reasonable chance of survival in the wild. For example, as more than 90% of wild mice die in their first year , any investment in mechanisms for survival beyond this age benefits at most 10% of the population. Nearly all of the mechanisms required for somatic maintenance and repair (DNA repair, antioxidant systems, etc.) require metabolic resources. Resources are scarce, as shown by the fact that the major cause of mortality for wild mice is cold, due to failure to maintain thermogenesis . The mouse will therefore benefit by investing any spare resource into thermogenesis or reproduction, rather than into better capacity for somatic maintenance and repair, even though this means that damage will eventually accumulate to cause ageing.
This concept, with its explicit focus on evolution of optimal levels of cell maintenance, is termed the ‘disposable soma’ theory [22, 23]. In essence, the investments in durability and maintenance of somatic (nonreproductive) tissues are predicted to be sufficient to keep the body in good repair through the normal expectation of life in the wild environment, with some measure of reserve capacity. Thus, it makes sense that mice (with 90% mortality by 10 months) have intrinsic lifespans of around 3 years, whilst humans (who probably experienced something like 90% mortality by age 50 in our ancestral environment) have intrinsic lifespans limited to about 100 years. The distinction between somatic and reproductive tissues is important because the reproductive cell lineage, or germline, must be maintained at a level that preserves viability across the generations, whereas the soma needs only to support the survival of a single generation. The idea that intrinsic longevity is tuned to the prevailing level of extrinsic mortality is supported by extensive observations on natural populations . Evolutionary adaptations, such as flight, protective shells and large brains, all of which tend to reduce extrinsic mortality, are associated with increased longevity.
The disposable soma theory makes specific predictions about the biology of ageing, as follows:
- 1Ageing results from lifelong accumulation of unrepaired cellular and molecular damage through evolved limitations in somatic maintenance and repair functions.
- 2Longevity is controlled primarily through genes that regulate the levels of somatic maintenance and repair functions (see Fig. 1).
- 3Immortality of the germline may require elevated levels of maintenance and repair in germ cells, when compared with somatic cells.
- 4The mechanisms of cellular and molecular ageing are inherently stochastic (i.e. strongly influenced by chance).
- 5There are likely to be multiple kinds of damage contributing to ageing, which will be regulated by a complex network of maintenance and repair functions.
- 6The allocation of resources to maintenance and repair is determined by evolutionary optimization, and the allocation strategy may be needed to be adaptive to respond to individual variations in circumstances of the organism during its life cycle (e.g. fluctuating food supply).
As these predictions reveal, the disposable soma theory brings a mechanistic specificity to the evolutionary theory of ageing, which previously had been couched in terms of the kinds of gene actions that might have been produced by the fact that the force of natural selection, i.e. its capacity to discriminate between alternative gene variants, tends to decline progressively with age, even if there is no intrinsic age-associated deterioration of the organism. Because, as we have seen, organisms die from a variety of causes, many of them simply accidental, the later a gene has its effect, the smaller can be its potential impact for good or ill. This led during the 1950s to the ideas that ageing results either from late-acting harmful mutations that acted too late to be cleared effectively by selection  or from genes that have opposing effects at different ages, being beneficial during earlier life but becoming harmful later . There is, however, one important difference between these earlier ideas and the disposable soma theory. In both of the earlier theories, although ageing is not a good thing per se, it is nevertheless the result of gene action. In the disposable soma theory, ageing evolved not through genes doing something, but because of genes not doing things. In other words, it is a strategy of genetic neglect.
Why ageing is not programmed
Although the evolutionary theory of ageing is by now well established, there has continued to be a tendency to seek explanation of ageing in terms of some kind of adaptive genetic programme. The attractions of this concept are easily understood. First, ageing is phylogenetically a very widely distributed trait and in species where senescence occurs, it affects every individual that lives long enough to experience its adverse impacts on fertility and vitality. Secondly, there are clear genetic effects on longevity and this leads naturally to supposing that the relevant genes specify some kind of ‘ageing clock’. In spite of these attractions, the programme theory, as a general explanation for ageing, is both logically and empirically unsound.
For programmed ageing to have evolved, it is necessary that intrinsic ageing should make a sufficient contribution to natural mortality that the hypothesized selection process is feasible.
If an individual dies before senescent effects are apparent, it makes no difference whether or not that individual is endowed with genes that programme ageing. Such a programme can only be fashioned by selection acting to realize the hypothesized benefits in individuals who survive to an age when the programme takes effect. Yet, in most natural populations, relatively few individuals survive long enough to be affected by intrinsic senescence, which is clearly seen only in protected environments [27, 28]. As noted earlier, the exception occurs in semelparous species, such as Pacific salmon, that have evolved a life-history plan in which there is only a single bout of reproduction. In such species, death of the parent usually occurs soon after reproduction. This is the result of directing all available resources to maximizing reproductive success, without regard to the subsequent survival of the adult. An important source of misunderstanding of the evolutionary theory of ageing has been to regard postreproductive death in semelparous species as an instance of programmed ageing, when in fact its evolutionary explanation is very different .
A second major obstacle is that the evolutionary mechanism on which programme theories depend is that of ‘group selection’. For an individual, senescence is disadvantageous because fertility and vitality are diminished. Therefore, if there is a programme for ageing, disruption of this programme by mutating the responsible genes is likely to produce a ‘selfish’ advantage. Although theoretical modelling indicates that it is possible to create a scenario in which evolution of programmed ageing can occur through group selection in a suitable spatially structured environment , the special assumptions necessary to support such examples also indicate the severity of the problems facing any general explanation of ageing in terms of a genetic programme. Finally, if it were true that ageing was programmed, it should happen occasionally that mutation should disrupt the programme, resulting in immortal mutants. None of the observed mutations that increase lifespan has been observed to remove the ageing process, only to delay it.
Implications for mechanisms of ageing
The evolutionary theory clearly predicts that ageing is driven primarily by the lifelong impact of molecular damage, which accumulates in cells and eventually gives rise to age-related frailty and disease (Fig. 2). This is a process which readily explains the life course nature of the ageing process, as damage begins to accumulate from the earliest stages of life, including in utero. Although instances of damage are random, in the sense that they may strike any of a wide range of targets within the cell, their rates of occurrence and repair are to some extent regulated through the evolutionary design of the cellular machinery for macromolecular biosynthesis and repair. It is from the genetic specification of such machinery and its level of function, that the degree of heritability of ageing and longevity is thought to arise. For instance, human centenarians are observed to have higher activity levels of poly(ADP-ribose) polymerase-1, which is a key player in the immediate cellular response to stress-induced DNA damage .
The process of ageing, as summarized in Fig. 2, clearly reveals how a variety of nongenetic factors can act to influence age-related morbidity and mortality. These factors include nutrition, lifestyle (exercise, etc.), socioeconomic status, work, and so on. In the case of nutrition, for example, a poor diet containing excess sugar and saturated fats contributes directly to the burden of damage with which cells have to deal, whereas a Mediterranean-style diet may contribute protective factors such as dietary antioxidants. Unravelling the various contributions from these nongenetic factors, in terms of both magnitude and precise nature, is a major research challenge. Studies such as the ongoing Newcastle 85+ Study on biological, medical and psychosocial factors associated with healthy ageing  represent attempts to address this challenge.
At the molecular level, evidence suggests that several of the most important mechanisms involve damage to macromolecules. Numerous studies have reported age-related increases in somatic mutation and other forms of DNA damage, suggesting that the capacity for DNA repair is an important determinant of the rate of ageing at the cell and molecular levels. When species with different longevity are compared, it is found that there is a general relationship between longevity and DNA repair . In many human somatic tissues, a decline in cellular division capacity with age appears to be linked to the fact that the telomeres, which protect the ends of chromosomes, get progressively shorter as cells divide . This is due to the absence of the enzyme telomerase, which is normally expressed only in germ cells (in the testis and ovary) and in certain adult stem cells. Some have suggested that in dividing somatic cells telomeres act as an intrinsic ‘division counter’, perhaps to protect us against runaway cell division as happens in cancer, but causing ageing as the price for this protection. Whilst the loss of telomeric DNA is commonly attributed to the so-called ‘end-replication’ problem – the inability of the normal DNA copying machinery to copy right to the very end of the strand in the absence of telomerase – it has been found that stress, especially oxidative stress, has an even bigger effect on the rate of telomere loss , telomere shortening being greatly accelerated (or slowed) in cells with increased (or reduced) levels of stress. An important connection between molecular stress and ageing is suggested by the accumulation of mitochondrial DNA (mtDNA) mutations with age  and by the accelerated ageing seen in a mouse model with enhanced levels of mtDNA mutation . Cells in which mtDNA mutation reaches a high level are likely to suffer from impaired ATP production resulting in a decline in tissue bioenergenesis.
In addition to damage to DNA, proteins are also subject to damage. Protein turnover is essential to preserve cell function by removing proteins which are damaged or redundant. Age-related impairment of protein turnover is indicated by the accumulation over time of damaged proteins, and there is evidence that an accumulation of altered proteins contributes to a range of age-related disorders, including cataract, Alzheimer’s disease and Parkinson’s disease. Protein turnover involves the functions of chaperones, which help to sequester and if possible restore denatured proteins, and proteasomes, which recognize and selectively degrade damaged and ubiquitinated proteins. With ageing, there is evidence for functional declines in the activities of both proteasomes  and chaperones . These declines may be part of a more general failure, through overload, of cellular ‘waste disposal’ processes .
The disposable soma theory suggests that multiple kinds of damage will accumulate in parallel within cells, as the same logic limits the investment in each of a wide range of maintenance and repair functions. Although the multiplicity of ageing mechanisms is now widely acknowledged, the reductionist nature of experimental techniques means that, in practice, most research is still narrowly focused on single mechanisms. This is severely limiting because, although for each of the kinds of damage outlined above evidence can be found that such lesions do accumulate during ageing, it is also clear that synergism and interaction between the mechanisms is likely to be important.
This interplay between mechanisms is illustrated by a recent study on the causes of the extensive cell-to-cell heterogeneity in the division potential of human diploid fibroblasts . Such cells have only limited capacity to divide, a phenomenon thought to be a manifestation of ageing at the cellular level. However, even when a cell population is derived from the multiplication of a single founder, the individual cells within the population display very different division potentials. This intrinsic variability challenges any idea of a simple cell division counter, such as that the limit on cell division is set by the progressive shortening of telomeres. Motivated by a mathematical model which showed how the heterogeneity of cell senescence could be explained quite naturally by interactions of multiple mechanisms (oxidative damage, telomere shortening and the stochastic nature of mutation to mitochondrial and nuclear DNA) , experimental study revealed a major, hitherto unexplored role for mitochondrial dysfunction in senescence.
What is still to be learnt?
The evolutionary understanding of ageing has made essential contributions but this theory needs continual refinement and development in the light of new empirical and theoretical discoveries. For example, it came as a surprise initially when single genes producing major increases in lifespan were discovered in C. elegans. This was because the ‘classic’ evolutionary theory of ageing predicted that there was likely to be a large number of genes determining longevity. This followed because the disposable soma concept applies generally to a wide array of mechanisms for somatic maintenance and repair, whereas in the case of the Medawar/Williams scenarios the theory predicts the existence of whole classes of alleles of which there might be many specific instances. In most cases, the genes with major effects on longevity have been found to regulate central aspects of metabolism, particularly with respect to organismal energetics, such as insulin signalling . Furthermore, the relevant pathways are often involved in the response to environmental modulation such as crowding or variation in nutrient abundance. With hindsight, the discovery of these gene effects should not have been surprising. The optimal allocation of metabolic resources between competing activities, such as maintenance, growth, reproduction and storage, lies at the heart of the physiological evolution of life histories (Fig. 3), and in particular the trade-off between investment in maintenance and other activities is the mainstay of the disposable soma theory. Organisms that are particularly subject to varying or unpredictable environments are likely to have evolved a regulatory gene hierarchy that can detect change and adjust metabolism accordingly to a different optimum.
An interesting question, yet to be addressed, is how the optimizing influence of natural selection with respect to somatic maintenance and repair might result in different outcomes for the different maintenance systems in different cell types. It is clear that the consequences of damage are different in stem cells, where possession of a high intrinsic division capacity means that a cell that experiences DNA damage might pose a high risk of malignancy. Presumably this is why some stem cell populations, e.g. in the intestinal epithelium, are highly sensitive to low doses of DNA damaging agents and readily initiate apoptosis, because deleting damaged stem cells has proved to be evolutionarily beneficial. In dividing cells generally, damage to DNA is more of a threat than in postmitotic cells. Conversely, accumulation of potentially toxic metabolic wastes is a greater problem for postmitotic cells, such as neurons, which cannot ‘dilute’ such wastes by synthesis of new cell materials during cell division.
An interesting recent discovery causing re-examination of the evolutionary understanding of ageing has been the finding that a form of ageing can be found in unicellular organisms. The original formulations of the classic evolutionary theory explained the evolution of somatic mortality, as opposed to the immortality of the germline. There was an expectation that unicellular organisms, which lack a soma (in the commonly understood sense), should be immortal. However, during the 1980s, the budding yeast Saccharomyces cerevisiae became established as an experimental model for research on ageing. In this case, the ‘mortal’ mother cell can, in a sense, be seen as soma, whereas the smaller bud that becomes the daughter can be seen as germline . The same argument might even include the bacterium Caulobacter crescentus which divides asymmetrically and also exhibits a form of ageing . The case became harder, however, with reports of ageing in fission yeast Schizosaccharomyces pombe  and the bacerium Escherichia coli , both of which divide symmetrically.
A resolution to this seeming challenge comes from looking more closely at the molecular and cellular bases of ageing . The origin of the ageing process in multicellular animals arose from the division of labour between germline and soma. As soon as the germline/soma distinction evolved, only the germ cells carried the responsibility for forming individuals of the next generation, freeing somatic cells to become specialists, such as neurons, muscle cells or cells in the lens of the eye. This came at a price, however, because the soma then became disposable. What the recent work showed is that a division of labour exists even in E. coli, highlighting the fundamental importance of reproductive asymmetry in creating a context for ageing to evolve. The germline/soma distinction is a sufficient instance of such asymmetry, but it is not a necessary one. Although E. coli appears to divide symmetrically, in molecular terms it does not in fact do so. One daughter cell receives the old cell pole, the other cell receiving a new pole. The difference is apparently enough to cause a decline in fitness of the daughter that receives the old pole and thus does not benefit from the complete renewal of its molecular structures.
Other evolutionary puzzles can be found in the context of some of the distinctive features of the human life history . In particular, the occurrence of menopause – the universal cessation of human female fertility at around the age of 50 years – is sometimes used to support the idea of programmed ageing. Why should a woman cease reproducing at a much earlier age relative to her biological lifespan potential than occurs in other mammals? There are two credible possibilities. First, it may be that humans do not differ from other mammals as much as it might seem. The earlier and more abrupt decline in reproduction seen in women relative to men is observed in a similar fashion in rats and mice living under the protected conditions of the laboratory . It is possible that for the vast majority of human evolutionary history, women only survived about as long as their egg supply (the depletion of which is the proximate cause of menopause), and only began experiencing menopause routinely as humans achieved longer and longer lives over the past several millennia. However, an alternative possibility is that menopause is a specific evolutionary adaptation that has its origin in the unique combination of circumstances that define the human life history.
The pressure to evolve increased lifespans was probably driven by the increase in human brain size, leading to advanced intelligence, tool use and social living, all of which will have reduced the level of extrinsic mortality and favoured increased investments in somatic maintenance. Increased neonatal brain size, however, makes giving birth riskier. The result appears to have been a compromise whereby, in comparison with other mammals, the human infant is born unusually altricial (i.e. requiring extended postnatal development before gaining independence from the mother) whilst still possessing an unusually large head. This has led to the suggestion that the menopause protects older mothers from the risks of late child-bearing, when senescence may make pregnancy and child-bearing less safe, and favours the survival of the mother to raise her existing children to independence. An alternative is that postreproductive females may gain more by contributing to the reproductive success of their offspring, through helping to care for and provision their offspring, than they would gain from attempting further reproduction of their own. Recent theoretical modelling indicates, however, that neither of the two hypotheses outlined above – the ‘maternal mortality’ and ‘grandmother’ hypotheses is in fact adequate on its own. Only when both are taken together in a combined model, they show that menopause does indeed confer an evolutionary advantage . This is important because it may explain why menopause is essentially unique to our species, in which this combination of factors has occurred. In essence, it is this combination, representing a convergence of biological and cultural evolution, that conferred sufficient biological value on older women that menopause evolved as an adaptation to reflect this value in evolving human social groups. A recent analysis of a remarkable data set on births, deaths and family structures, gathered between 1950 and 1975 in four villages in the Gambia, lends weight to the idea that menopause may indeed have an evolutionary foundation .