The evolutionary ecology of senescence

Authors

  • P. Monaghan,

    Corresponding author
    1. Division of Environmental and Evolutionary Biology, Institute of Biomedical and Life Sciences, Graham Kerr Building, University of Glasgow, Glasgow G12 8QQ, UK;
      *Correspondence author. E-mail: p.monaghan@bio.gla.ac.uk
    Search for more papers by this author
  • A. Charmantier,

    1. Centre d’Ecologie Fonctionnelle et Evolutive, CNRS U.M.R. 5175, 1919, route de Mende, 34293 Montpellier cedex 5, France;
    Search for more papers by this author
  • D. H. Nussey,

    1. Institute of Evolutionary Biology, University of Edinburgh, Kings’ Buildings, West Mains Road, Edinburgh EH9 3JT, UK; and
    Search for more papers by this author
  • R. E. Ricklefs

    1. Department of Biology, University of Missouri-St. Louis, One University Boulevard, St. Louis, MO 63130, USA
    Search for more papers by this author

*Correspondence author. E-mail: p.monaghan@bio.gla.ac.uk

Summary

  • 1Research on senescence has largely focused on its underlying causes, and is concentrated on humans and relatively few model organisms in laboratory conditions. To understand the evolutionary ecology of senescence, research on a broader taxonomic range is needed, incorporating field, and, where possible, longitudinal studies.
  • 2Senescence is generally considered to involve progressive deterioration in performance, and it is important to distinguish this from other age-related phenotypic changes. We outline and discuss the main explanations of why selection has not eliminated senescence, and summarise the principal mechanisms thought to be involved.
  • 3The main focus of research on senescence is on age-related changes in mortality risk. However, evolutionary biologists focus on fitness, of which survival is only one component. To understand the selective pressures shaping senescence patterns, more attention needs to be devoted to age-related changes in fecundity.
  • 4Both genetic and environmental factors influence the rate of senescence. However, a much clearer distinction needs to be drawn between life span and senescence rate, and between factors that alter the overall risk of death, and factors that alter the rate of senescence. This is particularly important when considering the potential reversibility and plasticity of senescence, and environmental effects, such as circumstances early in life.
  • 5There is a need to reconcile the different approaches to studying senescence, and to integrate theories to explain the evolution of senescence with other evolutionary theories such as sexual and kin selection.

Organisms change during their lifetimes in many ways, sometimes permanently as with most aspects of body growth and development, but sometimes temporarily, as with seasonal leaf loss and replacement, growth and regression of gonads, pre-migratory and pre-hibernation accumulation of fat reserves, or development of thicker fur or down in response to decreases in temperature. The predictability and reversibility of such change is of great interest to biologists, and we seek to understand how change is controlled and what fitness advantages it confers. Changes that occur in old age, generally referred to as ageing or senescence, are particularly fascinating, since these progressive and irreversible changes impair rather than improve performance, with apparently negative effects on fitness. Understanding the evolution and persistence of senescence therefore poses a particular challenge.

The typical route followed in biological research is to observe a natural process and then try to understand and explain it by investigations under controlled conditions, often in the laboratory. In this respect, research on senescence has followed an unusual path since, in many animal taxa, it was first studied in captive or laboratory conditions. The subsequent emphasis on the mechanisms responsible for impairment of performance in old age has been due, in part, to the great interest in whether we might intervene to slow or reverse senescence in humans. With this aim in mind, the majority of studies have focused on model species in laboratory conditions. These studies have greatly increased our knowledge and understanding of senescence, providing considerable insight into the genetic and cellular mechanisms (Kirkwood & Austad 2000; Bonsall 2006). This approach, however, is not appropriate for understanding the evolutionary ecology of senescence. Field and laboratory studies on a broader range of organisms are needed to understand biological variation in the processes contributing to senescence.

Accordingly, the papers in this special feature on the ‘Evolutionary Ecology of Senescence’ are intentionally biased towards investigations of senescence in natural conditions and/or non-model species. The papers focus on a number of new developments in the study of senescence likely to be of particular interest to ecologists. The importance of the comparative approach in both evolutionary and mechanistic studies is demonstrated by Ricklefs (2008), while Nussey et al. (2008) discuss the challenges facing field ecologists seeking to accurately measure senescence rates in free-living populations. The great advantages of data from longitudinal studies of known individuals are highlighted. Münch et al. (2008) review a remarkable body of emerging work on senescence and its causes in the honey bee (Apis mellifera), a species in which the life spans of genetically identical individuals vary by an order of magnitude or more. They show how novel insights can be gained by studying a non-model species that has a particularly interesting and relevant life-history pattern. Mangel (2008) links environmental variation with senescence patterns by developing a general life-history model linking early life conditions to degenerative disease in later life. He provides a quantitative framework that can be used to assess costs and benefits of exposure to environmental stressors in early life. Wilson et al. discuss and illustrate how a quantitative genetic approach can be used to examine the genetic basis of senescence rates and their evolution in wild populations. Their approach provides a fertile testing ground for evolutionary theory outside the laboratory. Finally, Bonduriansky et al. (2008) examine the largely unexplored links between the evolution of senescence and the evolution of sexual strategies. In this introductory paper, we discuss these and related issues more broadly. We also highlight and discuss some areas of controversy, and how these might be resolved by further research.

What is senescence?

This might seem a straight forward enough question, but in fact the terms ‘ageing’ and ‘senescence’ are used in various ways. There is obviously a distinction between age-related phenotypic change, which includes many developmental and maturational processes, and ‘ageing’ in the mature organism. In the individual, senescence is generally thought of as an inevitable, irreversible accumulation of damage with age that leads to loss of function and eventual death. This too is how the term ageing is generally used. Thus, for simplicity, the terms ageing and senescence are treated as synonymous in this special feature. We preferentially use the latter here, because there are phenotypic changes associated with advancing age, such as greying of the hair in many mammals, which might not in themselves impair performance in any way. Indeed, in some species of primates such as the gorilla (Gorilla gorilla gorilla), such greying is associated with an increase in performance, when so-called ‘silver backs’ gain mating or resource acquisition advantages (Margulis, Whitham & Ogorzalek 2003). Are these apparently trivial phenotypic changes, albeit age-related, not in fact ‘senescent’ or are they harbingers, or side effects, of deterioration at other levels in the body? To answer such questions, we need a closer integration of studies of mechanisms with studies of fitness consequences.

How might senescence have evolved?

Senescence has captured the attention of evolutionary biologists for more than a century (Finch 1990; Kirkwood & Rose 1991; Charlesworth 2000; Hughes & Reynolds 2005). The German biologist August Weissman, who developed the germ plasm theory of inheritance distinguishing germ line and somatic cell lineages, proposed in 1882 that senescence benefits the population by removing old, unproductive individuals (Weissman 1889). However, since senescence as explained by Weissman cannot benefit the individual, it should be selected against except in the special circumstances where kin or group selection can pertain. Moreover, because Weissman's hypothesis presupposes the existence of senescence, it cannot explain its origin.

Currently, there are three main, albeit related, theories for the evolution of senescence. The first can be traced to Haldane (1941) and Medawar (1952), who suggested that, because individuals die of causes that are not connected to senescence (termed extrinsic factors), the force of selection declines with age in proportion to Fisher's reproductive value (Fisher 1930), which measures the contribution of an individual to future generations. Medawar supposed that deleterious mutations expressed at older ages would accumulate in populations and reduce the survival and reproductive success of older individuals. Williams extended this idea in a second theory by proposing the existence of antagonistically pleiotropic genes that have deleterious effects in old age, but are nonetheless favoured because of their contributions to the survival and fecundity of younger individuals (Williams 1957). According to this second hypothesis, senescence is considered as a consequence of positive selection on genetic factors that happen to have negative effects later in life.

Both of the above theories have been refined somewhat since they were first put forward. Notably, Hamilton (1966) clarified the nature of selection on genes expressed at different ages. He pointed out that reproductive value is not the appropriate measure of the strength of selection because this parameter reflects only an individual's contribution to future generations and not the contribution from the gene pool as a whole. Indeed, in the absence of senescence, the reproductive value of an individual alive at any age (its expectation of future reproduction) is constant, providing of course its environmental circumstances remain the same. Therefore, changes in reproductive value with age cannot explain the evolution of senescence. Rather, even in the absence of senescence, the decline in the proportion of individuals remaining alive at progressively older ages, as a consequence of extrinsic mortality factors, provides sufficient explanation for the decline in the strength of selection with age. This is because, at least for those species that reach a fixed adult size, the greater the proportion of individuals experiencing the positive or negative effects of a particular trait, the stronger will be the selective forces. Thus, Medawar's mutation accumulation theory posits that, because there are so few individuals alive in older age classes as a result of non-senescence-related mortality factors, there is only very weak selection against mutations that do not have detrimental effects until old age. The antagonistic pleiotropy idea posits that, because more individuals are alive in younger age classes, mutations that produce positive fitness benefits in young individuals are favoured by selection even when they are detrimental in old individuals. The decline in the strength of selection with age can be offset to some extent in species that grow and increase in fecundity throughout their adult lives, such as turtles and some fish, in which case selection to postpone senescence can persist into very old age (Vaupel et al. 2004); in a few extreme cases, such as Hydra vulgaris, organisms seem to show negligible senescence and apparently indefinite life span (Martinez 1998).

The third theory for the evolution of senescence, developed by Kirkwood and termed the ‘disposable soma theory’, is centred on trade-offs in the allocation of limiting resources to self-maintenance and other activities, particularly reproduction (Kirkwood 1977, 2002; Kirkwood & Holliday 1979). In this theory, decline in function results from unrepaired damage to molecules, cells, and tissues as a result of life processes, particularly the harmful byproducts of normal metabolism and the stress imposed by reproduction and other factors; this damage therefore accumulates with age (Westendorp & Kirkwood 1998; Sgro & Partridge 1999). Kirkwood emphasises that the rate of accumulation of such damage is influenced by various biochemical mechanisms that prevent or repair damage, such as antioxidants and DNA repair enzymes. These mechanisms carry costs for the individual in terms of deployment of resources that might otherwise have been allocated to provisioning offspring or avoiding environmental causes of mortality, such as predation, inclement weather, and food shortages. Accordingly, the expected rate of deterioration of an individual reflects an optimized balance between resource allocation to self-maintenance and to other competing activities (Kirkwood 2005).

The disposable soma theory posits that there is no point in maintaining the soma beyond an age that the individual can reasonably expect to attain in its particular environment. Thus, individuals that live in a safer environment have a longer expectation of life and selection should increase allocations towards better prevention and repair mechanisms. Kirkwood's hypothesis is based on conflicting demands, which can occur within a life-history stage or across different life-history stages; where the latter occurs, investment patterns that confer early life benefits are likely to be favoured by selection. The disposable soma theory can be seen as a phenotypic version of antagonistic pleiotropy; its emphasis is on the age-related consequences of resource allocation trade-offs rather than the antagonistic consequences of the expression of a gene at different ages.

Thus, we have three different explanations of why selection to maximize individual fitness has not acted to eliminate senescence. The mutation accumulation theory suggests that the forces of selection against senescence are simply too weak, which is perhaps the least convincing. The antagonistic pleiotropy theory suggests that senescence is a consequence of an unfortunate link between the benefits that some traits provide to young organisms and their negative consequences late in life. The disposable soma theory suggests that while damage is inevitable, the rate of senescence is a consequence of the extent to which selection or individual circumstances favour the allocation of resources to repair. These three evolutionary processes are not mutually exclusive and they may well operate to a variable extent in different species. There is some empirical evidence in support of each (Kirkwood & Austad 2000) and Wilson et al. (2008) provide a short review of the testing of these hypotheses.

What causes senescence?

To distinguish between these evolutionary theories of senescence, and/or identify the circumstances under which they apply, we need to know more about the proximate processes responsible for organism senescence and how these are influenced by genetic and environmental factors. Many mechanisms have been identified as contributing to age-related deterioration in function (Nemoto & Finkel 2004), and it is not our intention to review these fully here (see also papers by Ricklefs and Münch 2008). One process that is widely believed to play an important role is the accumulation of oxidative damage, now termed the free radical theory (Harman 1956; Beckman & Ames 1998; Finkel & Holbrook 2000; Nemoto et al. 2004). The normal oxidative production of ATP by the mitochondria produces as a byproduct various reactive oxygen species (ROS) that can oxidize a variety of macromolecules, including lipids, proteins and DNA, and interfere with cell and tissue function (Finkel & Holbrook 2000; Barja 2004). Such oxidative damage is a strong candidate for senescence-related changes in individuals (Stadtman 1992; Hamilton et al. 2001; Kujoth et al. 2007). The production of ROS in the mitochondria can be reduced by altering membrane proton gradients (Brand 2000; Balaban, Nemoto & Finkel 2005), but this results in reduced efficiency in producing ATP and increased overall energy requirement (Serra et al. 2003; Speakman et al. 2004; Humphries et al. 2005). There appears to be no simple connection between ROS production and life span (Perez-Campo et al. 1998; Barja 2004), and the production of ROS can be balanced to some extent by the maintenance of systems to prevent their damaging effects. These include the ROS scavenging antioxidant enzymes and other antioxidants (Sohal, Mockett & Orr 2002). However, these mechanisms presumably require energy and nutrient allocation, and potentially interfere with the roles of ROS as signalling molecules and in defence against pathogens. Balancing the generation of free radicals and defence against oxidative stress appears to be an important mechanism underpinning life-history trade-offs. Thus, the metabolic processes of organisms, and their pattern of energy expenditure, are likely to be very important in influencing senescence. Many of the genes that appear to play a role in influencing senescence and life span regulate mitochondrial function and metabolic parameters, and responses to oxidative stress. The insulin–IGF-1 pathway, which is involved in nutritional signalling and metabolic regulation, appears to be of particular importance, and is relatively well conserved across taxa (Gems & Partridge 2001;Tatar, Bartke & Antebi 2003; Nemoto et al. 2004).

Another potentially important mechanism that influences senescence, and could underlie life-history trade-offs involving life span, is telomere attrition (Blackburn 1991; Monaghan & Haussmann 2006; Haussmann et al. 2007). Telomeres are regions of non-coding highly repetitive DNA that cap the ends of chromosomes, enabling cells to distinguish chromosome ends from chromosome breaks. Because some DNA is lost at the ends of chromosomes with each replication cycle, telomeres also enable cell lines to undergo the repeated replications necessary to build an organism and maintain its tissues without loss of coding sequences. When telomere length in nuclear chromosomes declines to a certain point, DNA replication ceases and the cell enters a state of replicative senescence. Such cells either die, or remain in tissues; in the latter case, they can secrete substances that cause further damage. Furthermore, the genome tends to become unstable when telomere length is short, which is associated with many negative effects (Campisi 2005). Studies of the links between cellular and organism senescence are still relatively few (Herbig et al. 2006). Comparative studies of telomere dynamics are also very limited. Studies in wild birds have suggested a correlation between the potential longevity of individuals and telomere loss (Haussmann et al. 2003; Pauliny et al. 2006; Haussmann et al. 2007). Most telomere loss occurs early in life, and the rate of loss appears to be linked to growth rate (Hall et al. 2004). Thus, telomere loss could mediate the trade-off between growth rate and senescence (Blount et al. 2003), but more work is needed on the links between cellular and organism senescence (Monaghan & Haussmann 2006).

Genes, senescence and evolution

To understand the evolution of senescence, we do of course need to know about its genetic basis. Genetic research on senescence has tended to follow one of two paths. Molecular gerontologists have sought to identify candidate genes associated with longevity in model study systems (Kenyon et al. 1993). At the same time, evolutionary biologists have used artificial selection and quantitative genetic techniques to test for age-specific genetic effects consistent with either mutation accumulation or antagonistic pleiotropy (Rose 1984; Charlesworth & Hughes 1996). Both approaches have yielded great insight into the process of senescence. Molecular work has identified numerous genes associated with increased longevity in humans, fruit flies and nematode worms and has pinpointed metabolic pathways that are likely to play a crucial role in driving variation in longevity and senescence patterns (Kenyon et al. 1993; Lin, Seroude & Benzer 1998; Suh et al. 2008; Nemoto et al. 2004). However, the significance of many of these mutations to senescence in natural populations is unclear. Artificial selection and quantitative genetic breeding experiments on laboratory fruit flies have yielded compelling evidence for age-specific gene action consistent with both mutation accumulation and antagonistic pleiotropy (Rose & Charlesworth 1980; Rose 1984; Charlesworth & Hughes 1996; Snoke & Promislow 2003). One challenge in the emergence of an inter-disciplinary approach to senescence is the reconciliation of these two approaches both to classical model systems and to a wider range of systems both inside and outside the laboratory. The increasing availability of advanced genomic technologies and interest in the application of quantitative genetic and quantitative trait locus (QTL) mapping techniques to non-model study systems is likely to facilitate such an integration (e.g. Wilson et al. 2008; Slate 2005). Approaches such as QTL mapping could help link specific regions of the genome with variation in longevity, senescence rates and trade-offs across life stages in any study system, and tie together ultimate evolutionary processes and more proximate genetic and molecular mechanisms underpinning senescence.

Developing our theoretical understanding of how natural selection acts on senescence rates is another important area in need of further research. The very fundament of the evolutionary theory of senescence – the expectation that the force of natural selection should inevitably weaken with age (Hamilton 1966) – has been questioned recently by theoreticians on the basis that, as mentioned already, in systems where size or fecundity increase through adulthood, the onset of senescence can be substantially or perhaps even indefinitely delayed (Baudisch 2005; Baudisch 2008). Theoreticians have also recently challenged Williams’ classic theoretical prediction that reduction in mortality factors that are age- and condition-independent should lead to selection for slower senescence (Williams & Day 2003; Reznick et al. 2004; Williams et al. 2006). Williams’ prediction, although widely accepted and tested, is of course dependent on the validity of separating intrinsic (senescence-related) and extrinsic (non-senescence-related) mortality risks (Abrams 1993; Williams & Day 2003). Such a separation might be possible in laboratory studies in which an age-independent hazard can be experimentally applied, but how applicable is this to natural populations? On the face of it, such a distinction seems improbable in the wild, since, for example, vulnerability to disease or predators seems likely vary with age (Williams & Day 2003). Interestingly, comparison of age-related mortality patterns in birds in captivity and in the wild does not support this, since the rate of senescence remains similar despite the absence of important extrinsic mortality factors ‘such as predation’ in captive conditions (see Ricklefs 2008). This highlights our need to know more about the causes of death, and how vulnerability to these causes does or does not change with environmental circumstances and with age. Rather than trying to separate causes of mortality into intrinsic and extrinsic mortality factors, it might be more fruitful to consider the extent to which any one factor has senescence-dependent and senescence-independent components. The number of predators in an area might be age- and condition-independent, but how fast you can run away from them is not. The inclusion of a condition-dependent environmental hazard alters the rate of age specific deterioration early and late in life, and we need now to focus on more detailed analyses of changes in the rate of senescence across the lifetime, and indeed also on the timing of its onset, which has received little theoretical and empirical consideration.

More interaction with theories of the evolution of senescence and other areas of evolutionary theory is also needed. There is currently only a very limited theoretical literature linking either kin (Bourke 2007) or sexual selection (Bonduriansky 2008) to the evolution of senescence. Bonduriansky et al., review of the latter subject reveals remarkable theoretical and empirical gaps: theoretical predictions remain unclear and data with which to test them scarce. For example, although potentially fundamental to the theories of senescence evolution, links between the intensity of sexual selection and senescence are still largely undescribed and theoretical predictions unresolved. Should higher sexual competition result in more rapid senescence due to a ‘live fast, die young’ trade-off, or, conversely, should it result in slower senescence because of stronger selection on high performance individuals? The questions and discussion presented in Bonduriansky et al.'s paper will undoubtedly stimulate multiple experimental tests in the near future.

Environmental effects on senescence

Given the diversity of mechanisms that underlie the process of senescence, it is not surprising that environmental influences are important in its expression (see Münch et al. and Mangel 2008). Generation of free radicals, for example, will be influenced by the pattern of energy expenditure, and the effectiveness of antioxidant defences influenced by dietary factors. Levels of environmental stressors, such as food availability, temperature, predators and competitors will also vary. A well-documented and robust environmental effect on senescence in laboratory studies is the increased life span associated with reduced food intake. Based on experimental results with calorie and diet restriction, Masoro (1998) developed the hormesis hypothesis (see Mangel 2008), which states that longer life under these conditions results from the effect of any number of mild caloric and nutritional stresses that bring protective and repair mechanisms into play. The hormesis hypothesis is similar to Wingfield et al.'s ideas about the effects of stress realised through the hypothalamic–pituitary–adrenal (HPA) axis, mediated by glucocorticoid secretions (Boonstra 1994, 2005; Wingfield et al. 1998). Thus, mild stresses induce protective mechanisms that enhance immediate survival but might also extend life more generally by altering the rate of senescence (Gems & Partridge 2008). Chronic severe stress has the opposite effect, possibly through continuously elevated metabolism and mobilization of energy reserves. Processes that improve resistance to stress, such as the expression of heat shock proteins, might reduce damage generation and accumulation, and thereby have life-extending effects (Ogburn et al. 1998; Kapahi, Boulton & Kirkwood 1999; Fabrizio et al. 2001; Ogburn et al. 2001; Zera & Harshman 2001; Lithgow & Walker 2002). However, in addition to the effects on senescence, we also need to know effects on other fitness-related parameters such as reproductive output. Furthermore, the distinction between factors that alter the risk of death, as opposed to the rate of senescence, is very important in this context (see below).

It is becoming increasingly clear that environmental circumstances and events during growth and development, experienced directly or as a result of maternal effects, can have long-term consequences for the pattern of degeneration later in life (Metcalfe & Monaghan 2001), Mangel (2008) provides a life-history approach to modelling the costs and benefits of early life effects in relation to the adult environment, likely to be generally applicable. Studies in wild populations are increasingly finding such links across life-history stages (Reid et al. 2003; Nussey et al. 2007; Keller, Reid & Arcese 2008). There are many mechanistic pathways through which such a link could occur – antagonistically pleiotropic effects creating a trade-off between rates of growth and senescence, for example. Alternatively, sustaining a high early growth rate might have damaging effects that do not manifest themselves until later in life. Environmentally induced variation in the pattern of growth is known to alter many aspects of the phenotype, including antioxidant defences (Blount et al. 2003), telomere dynamics (Hall et al. 2004; Monaghan et al. 2006; Houben et al. 2008) and stress responses (McEwen 2007), all of which can influence senescence rates.

Clearly, both genetic and environmental factors are likely to influence the pattern of senescence, and to a variable extent in different organisms. Dissecting these effects is an important focus for future interdisciplinary studies.

Identifying and measuring senescence

Recognising and measuring deterioration in performance in animals pose formidable difficulties, especially in the wild (Nussey et al. 2008). At the individual level, we tend to look for phenotypic changes with age that are linked to reduced fecundity and increased mortality risk. Such phenotypic changes can be difficult to recognise and might be minimal in some taxa (see Ricklefs 2008). At the population level, however, age-related changes in average fecundity and survival probabilities with age are often used as measures of senescence, but can lead to senescence being misidentified or missed altogether. Longitudinal studies of known individuals can get round some of the problems, but are difficult and time consuming (Nussey et al. 2008). Many studies of senescence therefore focus on age at death. However, the distinction between life span and senescence raises important considerations for evolutionary ecologists.

First, age at death by itself does not measure senescence. Variation in average age at death amongst populations or species can arise solely as a consequence of variation in the extrinsic mortality factors such as predation or disease risk, and need not be related to variation in the rate of senescence. It is also important to realise that there can be intrinsic, but still non-senescence-related, factors that alter life expectancy. Such factors, which might be environmentally induced, could alter vulnerability (sometimes called ‘frailty’) over the short or long term, producing differences in life span that are not due to differences in the rate of senescence. For example, phenotypic effects of early life conditions might elevate the risk of death at all ages, but the rate of senescence itself might be unchanged. The timing of the onset of senescence, rather than its progression, could also change, thereby affecting longevity (Hulbert et al. 2004). This kind of effect has recently been found in Drosophila: flies subjected to dietary restriction exhibit an immediate drop in the risk of death, but the age-related increase still proceeds at the same pace. Interestingly, this is not the case with changes in temperature. Reducing temperature does seems to reduce the subsequent accumulation of damage, resulting in the flies growing old more slowly in the cold (Hulbert et al. 2004).

An additional point to bear in mind here is that we might also see age-specific changes in the risk of death due to variation in factors other than the rate of damage accumulation. Such changes can then appear to be age-related when they are not, for example, when increasing age and changing environmental conditions are confounded. It is important therefore, if possible, to study more than one cohort in the wild. It is also possible that the risk of death might change with age, without being due to senescence if individuals take on new roles or occupy different environments with age, which expose them to higher disease or predation levels. Changes in investment patterns with age could also mean a higher risk of death. For example, changes in the environment or the onset of sexual maturity might mean that an organism has to allocate more resources to, say, foraging or thermoregulation at the expense of, say, immune function. Alternatively reduced cognitive ability as a consequence of poor growth conditions, as has been found in zebra finches (Fisher, Nager & Monaghan 2006), might not affect mortality risk until an individual has the increased demand of dependent offspring. Such effects can increase the risk of death in mature individuals, but not as a consequence of accumulated damage. In some cases, such altered mortality risk could be reversible, but not where it is a consequence of permanent phenotypic change. The apparent reversibility of senescence in the honey bee challenges the view that senescence is caused by the accumulation of irreversible damage. However, it could also be due to an associated change in risk of death for other reasons, as suggested above (see Münch et al. 2008).

To understand and evaluate such effects, we need more information on causes of mortality and how these change in relation to behaviour, environment and intrinsic state. It is also important to realise that life span might be terminated by intrinsic factors without senescence having to occur. In addition to the traditional view of senescence as a progressive decline resulting from damage accumulation, which is how we usually think of senescence, it is also possible that age at death could be controlled by particular pleiotropic genes that act like time bombs, suddenly putting an end to life by triggering a catastrophic failure in key organ systems. In such species, senescence might hardly be evident at all. As highlighted by Ricklefs (2008), it is surprising how little understanding we have of causes of death in natural populations, or indeed even in captive animals, that would allow us to say whether the grim reaper slowly engulfs the organism, or suddenly ambushes it.

While changes in life span need not imply changes in rate of senescence, conversely we cannot assume that faster senescence will necessarily be manifest as a reduction in the average life span of a population since it is often the case that very few individuals in the wild live long enough to suffer senescent declines. Finally, and very importantly, to understand the fitness consequences of, and hence the evolutionary ecology of, senescence, we need to combine studies of age-specific mortality patterns with studies of age-specific fecundity. There is increasing evidence of progressive deterioration in fecundity with age within individuals and many more longitudinal studies are needed (see Nussey et al. 2008).

Where do we go from here?

A consensus amongst all the papers in this special feature is the need for more studies in natural conditions involving a broader taxonomic range of organisms. More interspecific comparisons are likely to generate new insights into both the evolution and the mechanisms of senescence. As highlighted recently by Partridge & Gems (2007), extrapolating results from controlled laboratory conditions has its limits, especially for processes affecting life-history traits highly influenced by environmental conditions. Studies in the wild also provide the opportunity to study organisms that are difficult to maintain in the laboratory. We can identify species with exceptional life spans, such as many birds, of which the medical community is unaware, but whose physiology might reveal new approaches to slowing senescence in humans. There is a surprising lack of information on senescence in plants, which surely offer many interesting avenues for studying the evolution of senescence (Thomas 2002; Munne-Bosch 2007); most studies at present focus on seasonal leaf loss, and not on senescence rate and life-history trade-offs in the plant itself. Furthermore, there have recently been major improvements in how data from natural populations can be used in studies of senescence, and there remain many promising avenues to pursue in this area. Some of the processes identified in the laboratory as being of great importance, at least to life span, but possibly also to senescence rate, such as food intake, metabolism and oxidative stress, need to be studied in a wider range of organisms and in natural populations; in addition to their effects on longevity, we need to know the consequences for other important fitness parameters such as fecundity, and fitness-related traits such as immune function. We have progressed from descriptive studies answering the question ‘Do wild animals display reproductive and actuarial senescence?’ to sophisticated methods analysing longitudinal datasets over long time scales and enabling individual variation and environmental factors to be investigated. The question ‘Why does senescence occur?’ remains a challenge. However, recent attempts to test evolutionary theories of senescence while integrating them in a wider life-history perspective are providing new insights. The use of long-term datasets in exploring evolutionary processes has posed new methodological challenges, such as the problem of individual detection probability, revealing key issues that need to be solved. A number of fundamental issues also remain to be resolved, only some of which we have discussed above. Field based genetic studies of the causes of senescence need to be integrated with evolutionary theory, since this will enable us to evaluate the importance of the three different explanations of the evolution of senescence in different taxa and environments. Without doubt, the study of senescence offers a fertile ground for future research, and one that will bring together mechanistic and evolutionary approaches, and field and laboratory studies.

Acknowledgements

We thank Charles Fox and two anonymous referees for comments on this introductory paper, and all the other authors for contributing such interesting papers to this special feature.

Ancillary