Anders Pape Møller, Laboratoire de Parasitologie Evolutive, CNRS UMR 7103, Université Pierre et Marie Curie, Bât. A, 7ème étage, 7 quai St Bernard, Case 237, F-75252 Paris Cedex 05, France. Tel.: (+33) 1 44 27 25 94; fax: (+33) 1 44 27 35 16; e-mail: firstname.lastname@example.org
Senescence is the age-related deterioration of the phenotype, explained by accumulation of mutations, antagonistic pleiotropy, free radicals or other mechanisms. I investigated patterns of actuarial senescence in a sample of 169 species of birds in relation to latitude and migration, by analysing longevity records adjusted for sampling effort, survival rate and body mass. Senescence might decrease at low latitudes because of elevated adult survival rates and generally slow life histories. Alternatively, the rate of senescence might increase at low latitudes because of the greater impact of biological interactions such as parasitism, predation and competition on fitness through differential effects of age-specific mortality (e.g. because immunologically naïve young individuals and immuno-senescent old individuals might die more frequently than individuals belonging to intermediate age classes). Bird migration entails extensive exercise twice annually, with migrants spending more time in benign environments with little abiotic mortality than residents, migrants having higher adult survival rate and lower annual fecundity than residents, and migrants suffering more from the consequences of oxidative stress than residents. The rate of senescence increased with latitude, as expected because of slow life histories at low latitudes. Independently, rate of senescence decreased with increasing migration distance. These findings were robust to control for potentially confounding effects of body mass, age of first reproduction and phenotypic similarity among species because of common descent.
Senescence reflects the age-specific decrease in survival and fecundity at old age (recent reviews in Partridge, 2001; Bronikowski & Promislow, 2005). Many different theories of senescence have been proposed over the years, leading to intense research of the underlying causes (Williams, 1957; Hamilton, 1966; Rose, 1991; Partridge & Barton, 1993; Kirkwood & Austad, 2000). However, the hypothetical factors contributing to variation in rates of senescence among species are still poorly understood, especially when it comes to free-living organisms (Ricklefs, 2000; Partridge, 2001). A few attempts have been made to investigate broad patterns of senescence across species. For example, social insects appear to have a slower rate of senescence than other insects (Keller & Genoud, 1997). A subsequent study of sociality and senescence in birds revealed that the apparent effect of sociality on the rate of senescence was due to such species starting their reproduction at an older age rather than any inherent factor associated with sociality (Møller, 2006). In another example, Blanco & Sherman (2005) showed for venomous species of fish, frogs and reptiles that they attained a much older maximum age than related, non-venomous sister taxa. This finding is in accordance with evolutionary theories of senescence, although again potentially confounding factors such as differences in life history between the two groups of organisms need to be considered.
The area in which most progress has been made in terms of understanding the factors associated with the evolution of senescence is in the effects of early life-history decisions on ageing. Life-history theory accounts for the evolution of senescence through the effects of early reproduction and high rate of reproduction on the accumulation of mutations or other factors with debilitating effects (e.g. Williams, 1957; Hamilton, 1966). Any mutation that enhances reproductive success early in life at the cost of a reduction in the duration of the late part of lifespan will be favoured by natural selection, resulting in pleiotropic effects. Accordingly, a number of studies have shown that a delay in the start of reproduction reduces the rate of senescence (e.g. Gustafsson & Pärt, 1990; Partridge & Fowler, 1992; Partridge et al., 1999). Likewise, interspecific variation in age at first reproduction predicts the rate of senescence in birds, with species starting to reproduce at an older age, senescing more slowly (Møller, 2006).
The latitude at which animals live may hypothetically relate to the rate of senescence in two different ways. First, biotic interactions are generally considered to be more common and more intense close to the equator than in temperate climates (e.g. Moreau, 1944; MacArthur, 1972; Martin, 2004). For example, parasite-induced mortality is more common in the tropics, and the fitness costs of interspecific competition are also greater in the tropics when compared with the temperate zone (Janzen, 1970; Connell, 1971; Leigh, 1994; Guégan et al., 2003, 2005; Guernier et al., 2004). If biotic causes of mortality such as predation and parasitism are more common at low latitudes, as the previously cited studies suggest, we should also expect age-specific mortality to vary with latitude. For example, young individuals with naïve immune systems and immuno-senescent old individuals would be more likely to die from parasitism than individuals of intermediate age, causing biotic agents to kill off young and very old individuals more frequently at low latitudes. This would predict an increasing rate of senescence with decreasing latitude. Secondly, if mortality because of abiotic factors is particularly common at high latitudes, but less important in the more benign and stable climate of the tropics (e.g. Moreau, 1944; Lack, 1954; Ashmole, 1963; MacArthur, 1972; Newton, 1998), with a large fraction of mortality being caused by density-independent mortality linked to weather conditions, then abiotic factors would be expected to kill off very old individuals more frequently at high than at low latitudes (Medawar, 1952; Williams, 1966). Trade-offs between life-history traits are reflected in the life history of animals in the tropics being characterized by an emphasis on self-maintenance and hence extending reproductive life (Williams, 1966; Murphy, 1968). Few reproductive vacancies in the tropics also select for low rates of reproduction (Martin, 2004), with consequences for survival and hence potentially for reduced rates of senescence.
Migration is the annual movements between breeding and wintering ranges seen in many species of birds and fish, but also in mammals and insects. Three different mechanisms could link migration to senescence: (1) migration to benign environments reduces age-specific mortality rates; (2) evolutionary changes in life history of migrants decreases the rate of senescence and (3) efficient ingestion, absorption, metabolism and use of anti-oxidants reduces the rate of senescence in migrants.
First, if the ancestral state was resident behaviour (Berthold, 2001), we should expect migration to evolve when the benefits of migration exceeded the costs of intraspecific competition and other density-dependent processes such as predation and parasitism among residents. Although residents will have to cope with adverse conditions during different times of the year, migrants could potentially enjoy benign environmental conditions during a large part of their annual cycle, with some species experiencing two annual periods of peak productivity in seasonal environments in the northern and the southern hemisphere respectively. Therefore, migrants, when compared with residents, by living in more benign environments than residents would suffer less from abiotic causes of mortality, with a reduction in rates of extrinsic, condition-independent mortality, causing an increase in the contribution of old age classes to the next generation as predicted by classic evolutionary theories of ageing (Medawar, 1952; Williams, 1957).
Secondly, migrants encounter more severe constraints on reproduction than residents because of late arrival to the breeding grounds and a short breeding season. Hence, migrants generally start reproduction later during spring than residents (Böhning-Gaese et al., 2000; Berthold, 2001), allowing for fewer broods to be produced each year by migrants compared with residents. Several studies have suggested that migrants have a higher adult survival rate and a lower fecundity than resident species (Mönkkönen, 1992;Berthold, 2001; Böhning-Gaese et al., 2000; Clobert et al., 2001). Such a life history would suggest that migrants had a slower rate of senescence than residents, and that once differences in life history were controlled statistically, there would be no difference in the rate of senescence between migrants and residents.
Thirdly, migration is a long-distance movement that has resulted in a number of morphological, behavioural and physiological adaptations. There is direct physiological evidence suggesting that migrants arrive at their breeding grounds with depleted levels of anti-oxidants, with early arriving migrants differing in level of anti-oxidant depletion from late arriving individuals (Ninni et al., 2004). Similarly, there is experimental evidence from salmon that oxidative stress is an important cost of migration (Welker & Congleton, 2005). As anti-oxidants play an important role as quenchers of free radicals that are the natural by-products of metabolism caused by physical activity, reduction or depletion of such levels may cause permanent damage to DNA and other molecules because of the actions of free radicals (reviewed in Møller et al., 2000), with consequences for the rate of senescence (Finkel & Holbrook, 2000). If the level of depletion of anti-oxidants were a linear function of migration distance, this would result in rate of senescence increasing with migration distance. Such depletion would select for increased efficiency of ingestion, absorption, metabolism and use of anti-oxidants by migrants (Ninni et al., 2004). Such adaptations if found to exist would prevent damage to DNA and hence slow the rate of senescence. Therefore, we should, under this scenario, expect the rate of senescence to decrease with increasing migration distance.
The objectives of this study were to test whether interspecific differences in the rate of senescence were predicted by differences in breeding latitude and migration distance. To this end, I analysed an extensive data set on the rate of senescence in 168 species of birds from the Western Palearctic for which information was available for the rate of senescence and a number of ecological variables such as breeding latitude and migration distance (Møller, 2006). The migratory patterns of the birds included here are diverse (Berthold, 2001). These range from: (1) resident species that have completely overlapping breeding and non-breeding ranges, to; (2) partial migrants that have some populations that are resident to others than are migratory, and to; (3) migrants that have non-overlapping breeding and winter ranges. Migration distances are equally diverse ranging from: (1) residents that do not migrate, to; (2) short-distance migrants that breed and winter within the Palearctic zone, to; (3) long-distance migrants that breed in the Palearctic zone and winter in tropical or sub-tropical areas of Africa south of the Sahara or in Asia, and to (4) long-distance migrants that breed in the Palearctic zone and winter in the Antarctic, with the Arctic tern Sterna paradisaea that migrate up to 44 000 km setting the record (Berthold, 2001). Maps showing the actual breeding and winter distributions can be found in Cramp & Perrins (1977–1994).
Longevity records only provide reliable information on maximum lifespan if records are adjusted for sampling effort. Among 271 species of common birds in Europe, for which longevity records were available, the total number of recoveries and recaptures of banded birds across Europe ranged from 106 to 187 764, with a total of 2 640 601 records (http://www.euring.org). Therefore, I used total number of recoveries reported as a measure of variation in sampling effort.
Species with high survival rates will, invariably, also have extreme longevity records. However, a relatively long lifespan for a given survival rate will provide information about the rate of senescence, because longevities greater than predicted for a given survival rate suggest a slow rate of senescence, whereas short records of longevity suggest a fast rate of senescence. I extracted information on adult annual survival rate from Cramp & Perrins (1977–1994) and Glutz von Blotzheim & Bauer (1985–1997). Adult annual survival estimates were not based on age-structured models because only very few such studies has been published so far. Although some estimates are based on capture–mark–recapture methods that provide rigorous ways of adjusting for heterogeneity in capture probability, others are based on less rigorous methods. However, such heterogeneity is likely to make any statistical tests conservative.
I extracted the northernmost and the southernmost distribution limits for the breeding season and the winter to the nearest 0.1° from Cramp & Perrins (1977–1994). Mean breeding latitude was estimated as the mean of the northernmost and southernmost latitudes during the breeding season. Migration distance was estimated as the mean of the northernmost and southernmost latitudes during the breeding season minus the mean of the northernmost and southernmost latitudes during winter.
The 271 species with longevity records were reduced to 169 species for which information for all variables was available. The data set is reported in electronic Appendix S1.
I controlled for similarity in the rate of senescence among species because of common ancestry by calculating standardized independent linear contrasts (Felsenstein, 1985), using the computer program caic (Purvis & Rambaut, 1995). All regressions were forced through the origin because the dependent variable is not expected to have changed when the independent variable has not changed (Felsenstein, 1985). Standardization of contrast values was checked by the examination of absolute values of standardized contrasts vs. their standard deviations (Garland et al., 1992). Plotting the resulting contrasts against the variances of the corresponding nodes revealed that these transformations made the variables suitable for regression analyses. In cases where extreme residuals were recorded, I tested for the robustness of the conclusions by excluding contrasts with studentized residuals >3.00 (Jones & Purvis, 1997). Likewise, I ranked the independent variable to test if the conclusions remained unchanged (Møller & Birkhead, 1994), and in no case did this procedure give rise to conclusions different from those obtained with the contrast values. Ranking provides a very conservative test of an hypothesis, and robustness of findings to ranking of the independent variable thus suggests that distributions of variables are not a confounding factor leading to specific conclusions.
I log10-transformed research effort, longevity, age at first reproduction, migration distance and body mass, whereas adult survival rate was square-root arcsine-transformed to obtain variables that were normally distributed.
Møller (2006) described in detail the logic underlying the analyses of senescence reported here, and I only briefly reiterate this method. Basically, longevity records adjusted for sampling effort, survival rate and body mass provide unbiased estimates of actuarial senescence. Records of longevity by necessity must be controlled for sampling effort because it is easier to record an extremely old individual in a large sample than a small sample. I used the number of recoveries to control for sampling effort, by using the log-transformed number of recoveries as an independent variable in the analyses.
A second cause of bias in the comparative analyses of senescence is that large species have greater longevity, survive better and start their first reproduction at an older age than small species (e.g. Promislow, 1991; Bennett & Owens, 2002). Hence, allometric effects must be controlled statistically in order to avoid spurious correlations, and I used log-transformed body mass in the statistical analyses to control for such allometry effects.
Relative maximum longevity, after controlling maximum longevity statistically for sampling effort and body mass, should by definition be positively related to relative survival rate after controlling for sampling effort and body mass, because a high survival rate should eventually produce extremely old individuals. However, species with a relative maximum longevity larger than predicted for a given relative survival rate could by definition be considered to senesce slowly, whereas species with a smaller relative maximum longevity would senesce rapidly. Relative longevity was estimated from a multiple regression analysis with maximum longevity as the dependent variable and sampling effort, body mass and adult survival rate as independent variables.
This procedure for estimating rate of senescence was cross-validated with estimates of actuarial rates of senescence obtained from life table analyses and reported by Ricklefs (1998). The senescence parameter ω scaled significantly with body mass for all estimates reported by Ricklefs (1998) for birds [F = 42.33, d.f. = 1,16, r2 = 0.73, P < 0.0001, slope (SE) = −0.063 (0.010)]. Therefore, Møller (2006, p. 686) related longevity records to sampling effort, survival rate, body mass and Ricklefs’ senescence parameter ω, showing a strong relationship between the two measures of senescence [partial regression between longevity record and ω for the species that were common to the two data sets: F = 73.98, d.f. = 1,7, r2 = 0.91, P < 0.0001, slope (SE) = −2.18 (0.25)]. Therefore, the two senescence parameters reflect the same underlying phenomenon.
I tested for a hypothetical relationship between relative longevity and breeding latitude and migration distance. This regression analysis was repeated for species-specific data and for standardized linear contrasts. In this multiple regression analysis absolute longevity was the dependent variable, whereas sampling effort, body mass, adult survival rate, age at first reproduction, mean breeding latitude and migration distance were used as independent variables, using a stepwise elimination procedure to derive the best-fit models.
Information was missing for some species for certain variables, causing sample sizes to differ slightly among analyses.
A first model of maximum longevity of different species controlled for survival rate, number of recoveries and body mass, to produce an estimate of senescence, but also included migration distance, latitude and age at first reproduction as explanatory variables. This model explained 61 % of the variance with the variables accounting for sampling effort, age at first reproduction, breeding latitude and allometry explaining most of the variance (Table 1A). Survival rate was only marginally related to longevity, but was still retained in the final model. Longevity was positively related to migration distance, accounting for 3 % of the variance (Fig. 1a). In addition, longevity was negatively related to breeding latitude, accounting for 4 % of the variance (Fig. 1b). Finally, longevity was positively related to age at first reproduction (5 % of the variance). Breeding sociality as defined in Møller (2006) was not a significant additional predictor of longevity (the model in Table 1A with breeding sociality as an additional variable, partial F = 1.78, d.f. = 1, 161, P = 0.18).
Table 1. Best-fit models of longevity in relation to migration distance, latitude, age at first reproduction, survival rate, no. recoveries, and body mass in birds, based on (A) species-specific data and (B) standardized linear contrasts. The models had the statistics F = 40.88, d.f. = 6, 161, r2 = 0.61, P < 0.0001 and F = 26.75, d.f. = 6, 160, r2 = 0.50, P < 0.0001.
Sum of squares
Age at first reproduction
Age at first reproduction
An analysis of standardized linear contrasts revealed a model that accounted for 50 % of the variance (Table 1). The relationships detected for species-specific values were reproduced in the model of contrasts. The three factors accounting for survival, sampling effort and allometry explained most of the variance (Table 1B). Survival rate was far from a significant predictor variable (Table 1B), but the overall model only changed marginally when survival rate was eliminated. In addition, longevity was positively related to migration distance (3 % of the variance), negatively related to breeding latitude (3 % of the variance) and positively related to age at first reproduction (2 % of the variance) (Table 1B). Breeding sociality as defined in Møller (2006) was not a significant additional predictor of longevity (partial F = 0.15, d.f. = 1, 159, P = 0.70).
The main findings of this study of interspecific variation in rate of senescence in birds were that (1) rate of senescence increased with increasing latitude and (2) rate of senescence decreased with increasing migration distance between the breeding grounds and the winter quarters. These findings were robust to statistical control for the confounding effects of age at first reproduction, and they were also robust to control for similarity in the rate of senescence among species because of common descent.
Bird species breeding at low latitudes senesced more slowly than species breeding at high latitudes, with clear evidence of the relationship being linear (Fig. 1a). This relationship was relatively weak, when not controlling statistically for the effect of migration distance (0.7 %), and when not controlling for the effect of either age of first reproduction or migration (1.2 %), compared with 3.7 % of the variance explained when both these confounding variables were controlled. Thus, the association between latitude and rate of senescence was the strongest when the confounding effects of age at first reproduction and migration distance had been accounted for. The importance of interspecific interactions as causes of mortality is generally considered to increase towards the tropics, with abiotic causes of mortality playing a more important role in temperate and arctic climate zones (e.g. Moreau, 1944; MacArthur, 1972; Martin, 2004). Species that bred at lower latitudes had slower rates of senescence than species breeding at high latitudes. This indirectly implies that the rate of senescence is more strongly related to abiotic than biotic causes of mortality, as predicted by classic evolutionary theories of ageing (Medawar, 1952; Williams, 1957). The observation that rate of senescence increased with increasing latitude is consistent with this hypothesis.
Bird species that migrated the farthest senesced more slowly than species that migrated short distances or did not migrate at all (Fig. 1b), independent of mean breeding latitude (Table 1). This result is surprising given the vagaries that migratory species encounter on their voyages often covering thousands of kilometres annually, with a maximum of 44 000 km for the Arctic tern (Berthold, 2001). For example, it is common that migratory birds die in hundreds or thousands during periods of inclement weather while migrating (Berthold, 2001). Some of the variance in rate of senescence accounted for by migration distance was because of covariation with age at first reproduction and breeding latitude. Although the amount of variance accounted for was 2.9 % when both these factors were controlled statistically, only 2.4 % of the variance was accounted for after control for breeding latitude, and only 1.2 % when neither breeding latitude nor age of first reproduction were included as covariates. This implies that the effect of migration distance on the rate of senescence was partly concealed by correlations between migration distance and latitude and age at first reproduction respectively. I suggested three mechanisms to account for a hypothetical relationship between migration distance and senescence. First, migrants cover large distances annually, but also enjoy the benefits of living in productive environments in different parts of the world at different times of the year (Berthold, 2001). This fact allows migrants to exploit the advantages of productive environments twice during their annual cycle compared with only once for resident species. Such abundance of food may cause less tear and wear than the more restricted abundance encountered by residents remaining in the same site throughout the annual cycle, but also reduce the rate of abiotic extrinsic mortality thus increasing the contribution of old age classes to the next generation (Medawar, 1952; Williams, 1957). Secondly, migrants differ from residents with respect to several life-history traits such as fewer brood produced per year and a higher adult survival rate and a lower fecundity than resident species (Mönkkönen, 1992; Böhning-Gaese et al., 2000; Berthold, 2001; Clobert et al., 2001). Such a life history would predict a slower rate of senescence in migrants than in residents. I did include adult survival rate and age at first reproduction in the statistical analyses, implying that these life-history traits could not have contributed to the described relationship between senescence and migration distance. Thirdly, adaptations to long-distance migration could potentially delay senescence, for example, in terms of free radical scavenging (Ninni et al., 2004), or physiological repair following periods of excessive physical activity (Battley et al., 2001). Given that migratory birds and fasting individuals of the same species encounter similar changes in body composition (Battley et al., 2001), and given that fasting has been suggested to slow down the rate of senescence (e.g. Sohal & Weindruch, 1996; Vitousek et al., 2004; Masoro, 2005), such effects may allow a reduced rate of senescence. Likewise, the fact that migration is associated with elevated levels of oxidative stress (Ninni et al., 2004; Welker & Congleton, 2005) may have led to specific adaptations in terms of ingestion, absorption and metabolism of anti-oxidants. Such adaptations may include mechanisms consistent with the anti-oxidant theory of senescence (Finkel & Holbrook, 2000). There is also evidence for intraspecific variation in the rate of senescence being associated with migration. An intraspecific study of the rate of senescence in two populations of barn swallows Hirundo rustica showed that rapidly migrating male birds, but not the more slowly migrating female birds in a Danish population that migrated 9400 km annually, and that eventually reached a very old age, arrived much earlier from spring migration than control individuals that died young, whereas there was no similar difference for either male or female birds from a Spanish population that only migrated 3100 km annually (Møller et al., 2006).
The analyses presented here showed that latitude and migration distance explained small amounts of variation in senescence among species of birds (3.7 % and 2.3 % respectively). These estimated variance components are likely to be under-estimates for several reasons. First, the amounts of variance accounted for by the number of recoveries and adult survival rate should be excluded from the total sums of squares, because they are only auxiliary variables. If that is done the amounts of variance explained by latitude and migration increase to 4.6 % and 2.9 % respectively. Secondly, life history and ecological data do not necessarily derive from the same population, and although life-history traits from different populations are more similar than those of different species, such heterogeneity in the data will cause noise. Such noise will make it more difficult to discern patterns, causing variance components to be under-estimated.
In conclusion, the present comparative study of the rate of senescence in birds revealed that species breeding at low latitudes and performing long-distance migration senesce more slowly than their sister taxa. These patterns of interspecific variation in rate of senescence may have implications for our understanding of the ecological and evolutionary factors involved in the evolution of senescence.
I thank C. Spottiswoode for constructive discussions.