Eusocial Hymenoptera show a unique divergence in lifespan of queens and workers; queens belong to the longest lived insects while workers in most eusocial species have significantly shorter lives. The different phenotypes within a colony emerge through reproductive division of labour, which is a characteristic trait of eusocial animals. Division of labour as a measure of organismal complexity increases with colony size in eusocial species similar to the increase of complexity with size that has been shown for the whole range of living organisms. We show that queen and worker lifespan diverge in closely related species representing the transition from solitary to social life and show that queen and worker lifespan are correlated if colony size is taken into account: with increasing colony size the lifespan differential between queen and worker increases, whereas neither queen nor worker lifespan is associated with colony size. Additionally, the lifespan differential is better explained by colony size than by the weight differences between the castes. The divergence of phenotypes found is in line with the increasing specialization of subunits in larger organisms, which leads to increasing complexity. We argue that division of labour is acting to increase colony efficiency, which in turn shapes the investments made into individuals leading to short-lived workers and long-lived queens. Additionally, maintenance investments may be shaped due to the variable extrinsic risk faced by different castes. © 2013 The Linnean Society of London, Biological Journal of the Linnean Society, 2013, 109, 710–724.
Biology is largely a matter of size: larger organisms tend to be more complex and this rule applies to the whole range of living organisms and to different levels of biological organization (Bell & Mooers, 1997; Bonner, 1988, 2004; Changizi, McDannald & Widders, 2002). Often organismal complexity is measured as the number of specialized subunits (Bell & Mooers, 1997) but it may also be measured by their phenotypic divergence (Changizi et al., 2002). The increase in complexity results from increasing cooperative division of labour that appears at all levels of biological organization and which is a measure of the degree to which subunits (organelles, cells, and individuals) specialize on different tasks and thus reflects complexity (Bonner, 2004; Gadau, Fewell & Wilson, 2009; Rueffler, Hermisson & Wagner, 2011). Division of labour benefits the organism with increasing efficiency because it allows the organism to simultaneously accomplish several physiological processes (Hurd, Jeanne & Nordheim, 2007; Simpson, 2012). Once formerly autonomous units in a group shift to be intrinsically dependent upon one another they become integrated into a new hierarchical layer of biological organization representing an evolutionary transition such as that from unicellular to multicellular life or that from solitary to social life, which are accompanied by a shift in the level of selection (Smith & Szathmary, 1997; Heylighen, 2000; Wilson & Wilson, 2007; Goldsby et al., 2012).
This latest evolutionary transition is accomplished by eusocial species (Michod, 2000). The emerging colonies have therefore been termed superorganisms, a collective organism consisting of many individual ‘sub-unit’ organisms (Wheeler, 1911; Wilson & Sober, 1991; but see Strassmann & Queller, 2009). The level of the superorganism is reached when societies are bound by altruism and division of labour (Gadau et al., 2009). The colony can thus be likened to the body of an individual organism. As a result of division of labour, the colony is composed of reproductive individuals (queens) that represent the germ line and workers that represent the soma, but as compared with a multicellular organism has lower genetic relatedness between its members, which gives rise to conflict and competing levels of selection (individual- and colony-level) (Wheeler, 1911; Wilson & Sober, 1991; Ross & Matthews, 1991; Boomsma, 2009).
Here we focus on queen–worker differences in lifespan that arise with changes in colony size. Increasing colony size has a number of consequences for the social organization and the evolution of life histories of different individuals within eusocial colonies, which are driven by increasing division of labour (Bourke, 1999; Dornhaus, Powell & Bengston, 2012). The positive association of colony size and division of labour leads to increasing morphological skew as the degree of task and caste specialization increases, as well as a divergence of lifespan between queens and workers and the loss of individual totipotency (e.g. reduced reproductive capabilities of workers), all of which increase efficiency gains for larger colonies (Holldobler & Wilson, 1990; Alexander, Noonan & Crespi, 1991; Schmid-Hempel, 1998; Bourke, 1999; Carey, 2001; Changizi et al., 2002; Hurd et al., 2007). Once group living has evolved, coevolutionary changes affecting lifespan evolution emerge (Alexander et al., 1991; Carey, 2001). Disruptive selection becomes apparent once the chance or physiological ability of worker reproduction is minimized and leads to the evolution of distinct worker and queen phenotypes (Bourke, 2007). The divergence of phenotypes is a self-reinforcing process because early ageing in helpers with increasing colony sizes decreases the chance of future reproduction, promoting further phenotypic specialization (Alexander et al., 1991). This leads from a reduced reproductive potential to complete worker sterility while the fertility and lifespan of the queen increases (Holldobler & Wilson, 1990). As a result, queens belong to the longest lived insects and generally live longer than workers, although in most species they do not differ genetically and develop from the same female larvae (Keller & Genoud, 1997; Heinze & Schrempf, 2008; but see Kerr, 1950).
However, beside the effect of colony size on lifespan evolution, lifespan in the social insects may also be correlated with body size (Porter & Tschinkel, 1985; Calabi & Porter, 1989), independent of size as a response to the level of extrinsic mortality (Chapuisat & Keller, 2002) or may be task dependent (Rueppell et al., 2007). Therefore both, evolutionary (ultimate) and mechanistic (proximate) theories of ageing could explain the divergence of queen and worker lifespan (Hughes & Reynolds, 2005). Evolutionary theories of ageing assume that high levels of extrinsic mortality select for a reduced investment in physiological maintenance leading to a shorter lifespan (Medawar, 1952; Williams, 1957; Hughes & Reynolds, 2005). Following this the evolution of the observed lifespan variation between different castes in social insects has been linked to the caste-specific differences of extrinsic mortality (Keller & Genoud, 1997; Heinze & Schrempf, 2008; Parker, 2011). Evolutionarily selected worker lifespan as a response to different levels of extrinsic mortality has been shown to be preserved under laboratory conditions (Chapuisat & Keller, 2002). Extended longevity, by contrast, is associated with adaptations that reduce the risk of predation. Examples are subterranean living (Buffenstein, 2005), the ability to fly (Austad & Fischer, 1991), group or social living (Carey, 2001; Carey & Judge, 2001; Buffenstein, 2005; Keller & Jemielity, 2006) and intergenerational transfers (Lee, 2003; Amdam & Page, 2005). In most cases the queens in eusocial colonies are long-lived, because they are protected in the centre of the colony while short-lived workers take over tasks that are correlated with higher levels of extrinsic risk (Schmid-Hempel, 1998). As a result the mean lifespan of queens in fire ant (Solenopsis invicta) is 30 times longer than that of small workers (Calabi & Porter, 1989; Holldobler & Wilson, 1990). Even species with less pronounced morphological caste differences such as Diacamma rugosum show a 1.3-fold difference of maximum lifespan between workers and queens (Tsuji, Nakata & Heinze, 1996). In contrast, mechanistic theories of ageing make predictions about the relationship between metabolic rate and lifespan. Larger organisms have a lower metabolic rate per unit of mass, a slower development and greater lifespan (Collatz & Sohal, 1986; Finch, 1990). For example, the lifespan of fire ant workers is affected by both temperature and body size, which both change metabolic expenditure (Calabi & Porter, 1989).
However, mechanistic and evolutionary theories of ageing have been broadly discussed (Schmid-Hempel & Wolf, 1988; Keller & Genoud, 1997; Moret & Schmid-Hempel, 2000; Jemielity et al., 2005; Heinze & Schrempf, 2008; Parker, 2011) and therefore in this study we elaborate on the effect of colony size on changes in lifespan for both workers and queens. Highly eusocial species with larger colonies tend to have shorter lived workers (Matsuura, 1991; Schmid-Hempel, 1998) but for queens no such association was found (Holldobler & Wilson, 1990).
We first review trends in lifespan evolution associated with the transition from solitary to eusocial organization displayed by closely related species that show both solitary and social organization. Because within the Hymenoptera only bees and wasps represent this transition, we also included other eusocial species (mole-rats, gall-inducing thrips and ambrosia beetles) in this analysis. This approach can give an indication on how lifespan changes with the onset of task specialization of different individuals within a group.
Second, because colony size is correlated with social complexity and division of labour, including reproductive potential and caste differentiation, we tested the hypotheses that colony size can explain the differences between queen and worker lifespan within the eusocial Hymenoptera as proposed by Carey (2001). As phenotypic specialization increases, the lifespan differential between queens and workers should increase as a result of disruptive selection. Disruptive selection should lead to the evolution of diverging caste-specific lifespan as a response to task-specific levels of extrinsic mortality. Additionally, because the differential lifespan of workers and queens emerges from group living and the associated reproductive division of labour we expect that the differences between worker and queen lifespan should be phylogenetically independent and apply to terrestrial (wasps, bees) as well as subterranean (ant) species similarly as a response to colony size. Alternatively, lifespan differences between worker and queen castes could result from the increasing morphological differences caused by increasing division of labour in larger colonies, which in turn affect metabolic rate and longevity. If so, we should find a correlation between the differences in lifespan and the differences in weight between queen and worker castes.
To analyse the association between colony size and lifespan we collected lifespan, colony size, and dry weight data on eusocial Hymenoptera from the published literature.
Generally data on survival in social species were scarce. Apart from a few comparative studies or book chapters that focus on lifespan or colony size (Beckers et al., 1989; Holldobler & Wilson, 1990; Keller, 1998; Schmid-Hempel, 1998; Page & Peng, 1991; Hou et al., 2010), data needed for this study were often hidden in papers focusing on other topics. We collected data for eusocial Hymenopterans on lifespan, colony size, and dry weight by searching the scientific literature. We conducted queries on Google scholar and ISI web of knowledge for the following keywords: ‘lifespan’, ‘survival’, ‘colony size’, and ‘worker/queen weight’. The collected data originated from a combination of laboratory, field observation, and anecdote (especially data on lifespan). Sources for the data used in our analysis are given in the supplemental material (Table S1).
We collected lifespan data of solitary and closely related species that represent the steps from solitary towards eusocial organization as well as lifespan data for eusocial Hymenoptera. Within the Hymenoptera only bees and wasps represent recent examples of the transition from solitary to eusocial organization and thus we added eusocial species apart from the Hymenoptera (mole-rats, thrips, and ambrosia beetles) to analyse lifespan changes with changes in social organization. While data on survival would be more appropriate but rare we had to use data on mean or maximum lifespan, which was often complicated because sample sizes, especially for queens, were often low. Another problem was that published data can be obtained from laboratory or field colonies. For queens the difference between laboratory studies and field studies should be minimal; in both cases the queen lives in the centre of the colony protected from extrinsic risks. In the case of the workers laboratory lifespan could be higher than for field studies but bees and wasps were often allowed to forage outside in laboratory studies. Because we were interested in the intrinsic lifespan differences shaped by natural selection, which in turn result from differential investments into individuals or castes as a result of different levels of extrinsic mortality (Chapuisat & Keller, 2002; Heinze & Schrempf, 2008), we preferred laboratory over wild data, if both were available (e.g. Harpagoxenus saltator). Wild data, especially for workers, were often left truncated and did not cover the whole individual lifespan (e.g. workers of Pogonomyrmex barbatus were just monitored after first appearance outside the colony). We omitted such data from the analysis. In two cases field data were used because both queen and worker lifespan were available as field data but for just one caste as laboratory data and otherwise we would have mixed wild and laboratory data for one species, which we avoided if possible (e.g. Vespa simillima). If several maximum lifespan values were obtained we used the highest reported value because maximum lifespan usually represents maximum observed lifespan, which may vary between studies mainly depending on sample size. For species with polymorphic worker castes with differing lifespan (mean or maximum lifespan) we used the highest reported values because they would represent maximum worker lifespan (e.g. Acromyrmex subterraneus brunneus, Solenopsis invicta, Apis mellifera diutinus workers); if information was only available for one caste we used this. If survival curves were published we calculated mean lifespan and used the age of the last survivor as maximum lifespan. When just life expectancy was reported (e.g. Bombus ferdivus) but no survival curve was published we used the reported value as mean lifespan. If the reported maximum lifespan value was obtained while the individual was still alive (LS > 3 years) we used the reported value (LS = 3 years) to avoid an overestimation of lifespan. Sometimes we just found estimates of queen lifespan derived from colony survival, which may be critical as unobserved queen turnover may have occurred (Keller, 1998). For some annual species lifespan was reported after the first egg was laid or colony lifespan alone is reported, so we then added the hibernation time, which is part of the adult lifespan (e.g. Bombus melanopygus). If a lifespan range was given we used the upper limit as maximum lifespan and the midpoint as mean lifespan (e.g. Tetragonisca angustula).
We collected mean queen lifespan of 72 species. Mean worker lifespan was collected for 82 species. Maximum lifespan was obtained for 79 species (queen) and 76 species (worker).
Colony sizes were collected from the published literature (Table S1). If possible we used mean colony size of mature colonies. In cases where we just found one value we used that. If we found several reported colony sizes we took the mean, and if several mean values were obtained we used the mean weighted by the sample size (e.g. Pogonomyrmex occidentalis). If a range of colony sizes was given we used the midpoint of the reported values, and in cases where just a minimum size was given we used this. If we found a statement that the colonies are smaller than a certain value we reduced this by one-third because we found that average values are 20–50% smaller than reported maximum values, and additionally we did not want to overestimate colony size.
To account for the morphological differences between queens and workers we collected dry weight data for each caste. If we obtained several dry weight measures from different studies or several values for different worker castes we calculated the mean dry weight across castes and studies. Data sources can be found in Table S1.
Mean and maximum lifespan calculation
For some species we collected both mean and maximum lifespan data but for others we obtained just mean lifespan for worker and maximum lifespan for queens or vice versa. To compare those species and to incorporate a higher variability of species into our analysis we calculated a conversion factor for both queens and workers to calculate the mean from maximum lifespan or vice versa. Instead of dividing maximum lifespan by 1.45 to obtain mean lifespan (Keller & Genoud, 1997; Hou et al., 2010) we used a conversion factor retrieved from a linear regression (weighted by the study sample sizes) between mean and maximum lifespan of species where we collected both lifespan measures for either queens or workers. We chose linear regressions through the origin because of the clear relationship between mean and maximum queen lifespan (as seen in Fig. 1). For the regression we only used data from species where both mean and maximum lifespan were obtained either from the laboratory or the wild. For queens, the conversion factor (slope) obtained was 1.34 ± 0.05 (t1, 44 = 28.32, P < 0.001, R2 = 0.95, F = 802.2) and for workers 1.66 ± 0.07 (t1, 55 = 21.15, P < 0.001, R2 = 0.89, F = 447.3) (Fig. 1). After calculating maximum lifespan from mean lifespan (Fig. 1) we used the maximum lifespan data for all correlations between lifespan and colony size; using mean lifespan revealed similar results.
Lifespan and colony size
We used linear models to analyse the relationship between worker as well as queen lifespan and colony size. To test for a correlation between queen and worker lifespan we calculated the ratio of queen and worker lifespan (queen lifespan in days/worker lifespan in days) to obtain a dimensionless measure. We used linear regressions to analyse the relationship between the lifespan ratio and colony size. To normalize the residuals of the linear models we had to log-transform both the lifespan ratio and colony size. The species used and data sources can be found in Table S1.
For the Formicidae we also compared models including the mode of colony founding (dependent/independent) or queen number (monogyny/polygyny) along with colony size as explanatory variables. We used a backward elimination modelling approach starting with a maximal model. We then eliminated stepwise non-significant terms until a minimum adequate model with significant terms was reached (Crawley, 2002). We used ordinary least squares (OLS) linear models and model comparisons were performed using ANOVAs with χ2 tests. Additionally, the models were compared using the Akaike information criterion (AIC).
Dry weight ratio
We were able to calculate the weight ratio (queen weight/worker weight) for 15 species, for which we also had the data to calculate the lifespan ratio (queen lifespan/worker lifespan). We were then able to test if either colony size or the dry weight differences between queens and workers can explain the lifespan differences between queens and workers.
We used a backward elimination modelling approach starting with a maximal model with lifespan ratio as the response variable and dry weight ratio and colony size, plus the interaction between them, as explanatory variables. We then eliminated stepwise non-significant terms until a minimum adequate model with significant terms was reached (Crawley, 2002). We used OLS linear models forced through the origin with log-transformed variables. Following a Lilliefors- (Kolmogorov–Smirnov) test the residuals were found to be normally distributed for all models. Model comparisons were done using ANOVAs with χ2 tests. Additionally, the models were compared using AIC. The used species and data sources can be found in Table S1.
We used phylogenetically independent contrast (Felsenstein, 1985; Garland, Harvey & Ives, 1992) for continuous variables to control for evolutionary relationships. Contrasts for continuous variables were obtained using the crunch option within the R-package caper (David et al., 2012) and were then subjected to regression analyses through the origin.
We used two different approaches. First we used an existing phylogeny for the Fomicidae (Moreau et al., 2006), which includes some species that appear in our dataset. We further added species if the genus was included in the existing phylogeny. We were able to include 13 ant species, the honey bee, and a wasp species as these where used as outgroups in the analysis of Moreau et al. (2006).
Second, we constructed a phylogeny from published trees of social Hymenoptera (Schmitz & Moritz, 1998; Danforth, 1999; Costa et al., 2003; Arevalo et al., 2004; Moreau et al., 2006). Because no information on branch lengths could be assigned to our composite phylogeny we set all branch lengths to the same value (Fjerdingstad & Crozier, 2006). Branch length information is not indispensable for independent contrast analyses, which are generally robust (Garland et al., 1992). We were able to include 29 species representing all genera in our analysis (the tree and species used can be found in Table S2 and Fig. S1).
All statistical analyses were performed using the R-statistical software (R Core Development Team, 2010) and the packages nortest (Gross & Ligges, 2012), caper (David et al., 2012), and ape (Paradis, Claude & Strimmer, 2004).
Lifespan changes during social evolution
In our review of the literature an increased divergence of lifespan of queens versus workers with increasing sociality was found in all groups examined (including mole-rats, thrips, and ambrosia beetles in addition to bees and wasps, which represent the social Hymenoptera).
For African mole-rats (Bathyergidae) maximum recorded lifespan of solitary mole-rats was between 6 (Bathyergus suillus) and 11 (Georychus capensis) years (Dammann, 2006), but data were scarce and hence not as reliable as data for the social species. Maximum lifespan in the social species varied from 11 years (Cryptomys hottentotus) to 30 years (Heterocephalus glaber) (Dammann, 2006; Liang et al., 2010). Non-breeders of the genus Fukomys displayed significant differences in lifespan compared with breeders. Non-breeder lifespan was comparable to solitary species (Fukomys anselli, maximum lifespan = 8 years; Fukomys mechowii, 10 years) while breeder lifespan was prolonged (Fukomys anselli, maximum lifespan = 19 years; Fukomys mechowii, 16 years) (Dammann & Burda, 2006; Dammann et al., 2011). Breeders and non-breeders of Heterocephalus glaber, which forms the largest groups (up to 295 individuals) within the social mole-rats, showed similar lifespan (Brett, 1991; Buffenstein, 2005, 2008; Liang et al., 2010).
Several gall-inducing thrips are eusocial (Crespi, 1992; Kranz, 2005) and in the genus Kladothrips both solitary and eusocial species exist (Choe & Crespi, 1997). In solitary species (Kladothrips rugosus and K. ellobus) females lived for about 10–12 months (annual life cycle), and the same lifespan was found for foundresses in the eusocial species K. intermedius and K. waterhousei, while soldiers in the same species showed a reduced lifespan of about 3–6 months (Crespi, Morris & Mound, 2004).
In the solitary fungus-growing ambrosia beetles Dolipygus dubius and Dendroplatypus impar offspring leave the nest after emergence and individuals had lifespans of 14–16 weeks (Browne, 1962) and 1 year (Browne, 1961). Species in which cooperative breeding occurs, such as Notoplatypus elongates, reached a lifespan of more than 3 years (Kirkendall, Kent & Raffa, 1997). The only known eusocial species is Austroplatypus incompertus, which forms monogynic colonies that can last up to 37 years (Costa, 2006). Females in this species may live more than 4 years while it was not known how long the workers live (Kirkendall et al., 1997).
In the superfamily Apoidea eusociality has arisen at least eight times (Wilson, 1971). Solitary Apidae (Ceratina calcarata) had an average lifespan of approximately 14–16 months and a maximum lifespan of almost 2 years (22 months) (Rehan & Richards, 2010a, 2010b). In facultatively social Apidae (Xylocopa virginica) mean lifespan was below 11 months and maximum lifespan reached 26 months (Richards, 2011).
Wasp lifespan increased with increasing social complexity (Evans, 1958; Wilson, 1975; Carey, 2003). Solitary parasitoids had short lifespans of 14–60 days (Natocyphus), and extensive parental care in Gorytes and Stenogaster species led to an increase in lifespan (60–365 days) leading to advanced eusocial wasps (e.g. Vespula) with queen lifespans ranging from 180 to 1000 days (Evans, 1958; Wilson, 1975; Carey, 2003). However, worker lifespan in eusocial wasps (Vespula, Vespa) seemed to be similar to their related solitary species (20–65 days) (Strassmann, 1991). Data on eusocial wasps were also included in the following analyses.
Colony size effects on lifespan in eusocial Hymenoptera
Worker lifespan and colony size
We did not find a general trend in the relationship between maximum worker lifespan and colony size. All regressions of maximum worker lifespan on colony size showed significant intercepts but not slopes (Table 1; Fig. 2). Due to the small sample size the Halictidae were only included in the regression on all Hymenopterans.
|Family||Intercept||P value||Slope||P value||R2||d.f.|
|Family||Intercept||P value||Slope||P value||R2||d.f.|
Queen lifespan and colony size
We did not find a general trend in the relationship between maximum queen lifespan and colony size. The regressions of maximum queen lifespan on colony size showed significant intercepts except for the Apidae; slopes were marginally significant for the Formicidae, Apidae, and all Hymenoptera combined (Table 1). In the Apidae the R2 = 0.761 was large given the small sample size of five species. Long-lived queens were found at all colony sizes while short-lived queens do not appear at larger colony sizes (Fig. 3). Due to the small sample size the Halictidae were only included in the regression on all Hymenopterans.
Lifespan ratio (queen/worker) and colony size
We found a positive association between the lifespan ratio, which represents the differences in the life-history trait lifespan between queens and workers and colony size, as follows: Hymenoptera: estimated slope = 0.23 ± 0.01 (t1, 48 = 17.14, P < 0.001, R2 = 0.86, F = 293.7); Apidae: estimated slope = 0.33 ± 0.3 (t1, 4 = 6.92, P = 0.002, R2 = 0.92, F = 47.92); Vespidae: estimated slope = 0.32 ± 0.04 (t1, 9 = 7.74, P < 0.001, R2 = 0.86, F = 59.87); Formicidae: estimated slope = 0.21 ± 0.01 (t1, 30 = 16.55, P < 0.001, R2 = 0.9, F = 273.7). Figure 4 shows the results for all Hymenoptera combined. For the Formicidae we also compared models including the mode of colony founding (dependent/independent) or queen number (monogyny/polygyny) along with colony size as explanatory variables. Neither of the additional variables was significant and the model just including colony size was preferable (ANOVA (queen number): change in the residual sum of squares = −0.530, d.f. = 1 on 46, P = 0.22; ANOVA (mode of colony founding): change in the residual sum of squares = −0.114, d.f. = 1 on 24, P = 0.867). The AIC was lower for the model using colony size as main effect (AIC = 4.81) than the model with colony size and queen number as explanatory variables (AIC = 12.78) or the model including colony size and the mode of colony founding (AIC = 13.02).
Using the phylogeny of Moreau et al. (2006) revealed a slope of 0.2 ± 0.063 (t1, 13 = 3.17, P = 0.007, R2 = 0.44, F = 10.04). Using the tree created by merging different Hymenoptera revealed: slope = 0.21 ± 0.047 (t1, 24 = 4.38, P < 0.001, R2 = 0.44, F = 19.18). In both cases the relationship between the lifespan ratio and colony size did not change when controlling for phylogenetic effects.
Lifespan ratio and weight ratio
After elimination of non-significant terms, only colony size remained in the model (slope = 0.21 ± 0.02, t1, 14 = 11.07, P < 0.001, R2 = 0.89, F = 122.6). Both the interaction between colony size and dry weight ratio (ANOVA: change residual sum of squares = 0.086, d.f. = 1 on 13, P = 0.2961) and dry weight ratio as a main effect (ANOVA: change in residual sum of squares = 0.004, d.f. = 1 on 14, P = 0.8552) were not significant.
We also compared the AIC values of the minimum adequate model (with colony size as the only explanatory variable) with that of a model with dry weight ratio as the only explanatory variable (slope = 0.57 ± 0.09, t1, 14 = 6.42, P < 0.001, R2 = 0.73, F = 41.16). The AIC was much lower for the model using colony size as the explanatory variable (AIC = 6.36) than the model with only dry weight ratio (AIC = 19.97). This is a strong indication that the model with colony size is the better model to explain the lifespan ratio of queens and workers than the model including the weight differences of queens and workers.
Colony size and weight ratio
We found a positive association between colony size and weight ratio using colony size as the explanatory variable: estimated slope = 0.31 ± 0.04 (t1, 14 = 8.69, P < 0.001, R2 = 0.83, F = 75.62).
Our comparative study reveals a correlation between division of labour measured as colony size and the lifespan ratio of queens and workers in eusocial Hymenoptera (Fig. 4). The differences in lifespan increase with increasing colony size and point to a correlation between queen and worker lifespan when colony size is taken into account. Interestingly, the correlation with colony size is weak or disappears if workers and queens are considered independently (Table 1; Fig. 2, 3). While weight differences between queens and workers increase with colony size, colony size itself is a better predictor for the lifespan differences found.
Throughout the transition from solitary to eusocial organization across different species from insects to vertebrates we find a divergence in lifespan between castes with increasing sociality, as did Carey (2001), probably caused by increasing reproductive division of labour (Bourke, 1999). In most groups lifespan of the reproductive individuals increases while worker lifespan stays within the range of the comparable solitary species. However, we found two exceptions to this pattern. First, soldiers of the gall-inducing thrips show a reduced lifespan as compared with solitary species while the foundress lifespan remains constant. Another unique pattern is that soldiers in this species are larger than foundresses, while their main task is to defend the colony (Perry et al., 2004). Second, within the social mole-rats, Heterocephalus glaber is an exception by showing the same lifespan of breeders and non-breeders. This may be caused by differences in maintaining the social system between the Fukomys species (incest avoidance) (Burda, 1995) and Heterocephalus glaber (dominant control), where the queen suppresses other colony members from reproduction but workers retain the ability to reproduce within their natal colony (Faulkes & Abbott, 1996).
However, when focusing on the social hymenoptera we find no correlation of worker lifespan with colony size (Fig. 2). Worker lifespan might decrease with colony size as in wasps but only within closely related species (Matsuura, 1984). Thus, phylogenetic constraints could drive the results of Matsuura (1984) and have also been suggested for the long lifespan of Myrmecia and ponerine workers (Holldobler & Wilson, 1990). Carey (2001) argued that worker lifespan or daily birth rate determine colony size, either of which needs to increase in order to obtain larger colonies (Carey, 2001). Because worker lifespan does not change with colony size the egg-laying rate of the queen must increase with increasing colony size.
We found marginally significant correlations of queen lifespan with colony size for the Formicidae and the Apidae (Fig. 3, Table 1). Within the Apidae a positive association between colony size and queen lifespan might exist but with our small sample size we cannot statistically confirm this. Within the Formicidae long queen lifespan occurs at all colony sizes while short-lived queens are only found in smaller colony sizes (Fig. 3). As pointed out by Keller & Genoud (1997) queens in polygyne colonies live significantly shorter than queens of monogyne species and for our dataset queen number would be a better predictor of queen lifespan than colony size. Interestingly, polygyne species appear with higher frequencies at smaller colony sizes. Short-lived queens do not occur in species with large colony sizes; this can either result from ancestral monogyny (Schrempf & Heinze, 2007) with the need for a longer lifespan in queens with larger colonies or it could be caused by a sampling bias because queen lifespan can be obtained by colony survival in monogyne but not in polygyne species (Gordon, 1991; Hou et al., 2010). Larger colony sizes lead to an increased protection of the queen against extrinsic risks. Additionally, an increasing reproductive value and the dependency of colony fitness on the survival of the single queen should select for an increasing queen lifespan and survival, which is in line with the evolutionary theories of ageing (Stearns, 1992; Keller & Genoud, 1997).
If we could control for ecological factors or larger phylogenetic trends within the hymenoptera it might be possible that a correlation between lifespan (queen and worker) and colony size would appear. For example, queen lifespan of subterranean species (ants) is much greater than for terrestrial species (bees and wasps) (Carey, 2001) and colony size is likely to be set by ecological factors (Bourke, 1999). The relative differences of lifespan (lifespan ratio) and colony size used here are likely to control for phylogenetic and ecological constraints and offer an opportunity to compare different eusocial species as the result of the independent contrast analysis suggests. In our comparison bee, wasp, and ant species fit into the observed relationship of lifespan ratio and colony size. We could even include the social mole-rats of the genus Fukomys [which show long lifespans for queens and workers, but have small colony sizes (Dammann & Burda, 2006; Dammann et al., 2011)] without changing the results.
We found that worker and queen lifespan diverge (increasing lifespan ratio) with increasing colony size in eusocial Hymenoptera. This may point to a general trend in eusocial organisms, which is already visible at the transition from solitary to social species also for species not belonging to the Hymenoptera. Two examples are the wasps and Fukomys species (mole-rats), in which worker lifespan remains at levels of the solitary ancestors while queen lifespan increases in the social species. In honey bees (Apis mellifera) the same pattern is observed throughout colony growth: in small colonies workers have longer lives when compared with workers living in larger colonies (Rueppell, Kaftanouglu & Page, 2009). Increasing colony size leads to a number of social consequences: the reproductive potential of workers decreases while the social complexity and the degree of caste differentiation increase (Alexander et al., 1991; Bourke, 1999; Changizi et al., 2002). Lifespan as a life-history trait appears to be affected by those changes.
Our finding generally points to an adaptive process of division of labour and increasing task specialization that eusocial Hymenoptera have in common and that could lead to an adaptive demography, where the demography of the colony is directly adaptive and not its causal parameters (Oster & Wilson, 1978; Holldobler & Wilson, 1990). As colony size increases the quality of individuals diverges and lifespan as an outcome of maintenance investments may be evolutionarily shaped to optimize productivity and hence fitness on the colony level by modifying the resource allocation into different castes. However, it is difficult to disentangle the effect of larger colony size on the lifespan differential. Colony size is associated with increasing division of labour, which in turn leads to morphological differences and different patterns of resource allocation into individuals possibly as a response to different levels of extrinsic mortality (Bourke, 1999; Amdam & Omholt, 2002; Muench, Amdam & Wolschin, 2008; Dornhaus et al., 2012). Thus, it is difficult to identify exclusive explanatory factors for the observed lifespan differential.
Following the evolutionary theories of ageing maintenance, investment could be shaped according to the differences in extrinsic mortality and thus for social insects it has been argued that extrinsic mortality explains the lifespan of queens, the differences between worker and queen lifespan, as well as lifespan differences between different worker castes (Keller & Genoud, 1997; Chapuisat & Keller, 2002; Heinze & Schrempf, 2008). Following this, one possible explanation for the increasing lifespan ratio is that workers of larger colonies experience higher levels of extrinsic mortality while queens are more heavily protected by their colony. Larger colony sizes may force colonies to increase their foraging range and to defend more distant food patches. The distance each forager has to cover during a foraging trip increases and thus the risk of dying from predation (Jorgensen & Porter, 1982). Additionally, competition with other colonies might increase for species with larger foraging ranges (Gordon & Kulig, 1996). Selection may act on the colony level to ensure resources within the protected colony and minimize losses caused by the early death of a worker. Individual worker productivity should be optimized: applying a model of intergenerational transfers to honey bees suggests that task- or caste-dependent senescence patterns may occur as a result of different schedules of resource allocation to individuals dependent on the task carried out (Amdam & Page, 2005). In larger colonies selection for morphological or physiological changes that increase worker productivity should be more efficiently selected for because the benefits would be multiplied by the number of workers. This trait could be selected for at the colony level to increase colony fitness by changing the life histories of the individuals within, resulting in disposable workers (Porter & Jorgensen, 1981) and highly reproductive and long-lived queens. Additionally, redundancy and parallel operations leads to a reduced dependency of the superorganism on the precise functioning of single individuals (Oster & Wilson, 1978). As long as the required tasks are accomplished and no fitness compromises are made, selection on worker lifespan might be weak, leading to workers that expend fewer resources independent of extrinsic mortality. It has been shown that small short-lived workers of Solenopsis invicta are less costly for the colony than large workers (Calabi & Porter, 1989).
Alternatively, the differences in lifespan could be explained by the morphological differences, but in our study we find that colony size and not the increasing morphological differences measured by the dry weight ratio explain the difference in lifespan between queens and workers. Even in species with morphologically identical castes, which differ in social or reproductive status, lifespan may differ. Examples are ant species with gamergate systems such as Platythyrea punctata or Diacamma cf. rugosum (Tsuji et al., 1996; Hartmann & Heinze, 2003) or social wasps such as Roplidia marginata (Gadagkar et al., 1993), which are included in our dataset.
However, worker lifespan may be correlated with body size (Porter & Tschinkel, 1985; Calabi & Porter, 1989), independent of size as a response to the level of extrinsic mortality (Chapuisat & Keller, 2002) or may be task-dependent (Rueppell et al., 2007), and a combination of those factors may ultimately explain the interspecific differences of worker lifespan as a response to colony size. Thus, in this study we cannot be certain which factors drive the positive association of colony size and lifespan.
In general, we cannot be sure why the observed lifespan differences appear as a response to changes in colony size. But we can confirm a larger organismal trend, namely the increasing specialization with increasing size within the superorganism. As a general rule governing the evolution of organismal complexity, larger entities [organisms, nervous systems, ecosystems, and ant colonies (superorganisms)] are composed of increasing numbers of subunits (Bell & Mooers, 1997; Changizi et al., 2002). In multicellular organisms division of labour measured as the number of cell types increases with body size (Bell & Mooers, 1997), and in ant colonies the number of castes increases with colony size (Changizi et al., 2002). Here we can show for the social insects that the increasing number of phenotypes also leads to an increasing divergence and specialization of those phenotypes. Although we can confirm the increasing morphological differences with colony size measured as weight (this study) or size differences (Changizi et al., 2002), we can also show that life-history traits may also be affected. The measured trait lifespan diverges as a response to colony size and increasing complexity.
For social Hymenoptera lifespan shows a strong association with colony size once the differences between workers and queens are taken into account. The queen–worker lifespan ratio (queen lifespan/worker lifespan) is positively associated with colony size, while the association between queen or worker lifespan with colony size is not existent or weak. If we assume that the differences in lifespan between queen and worker of a species arise from increasing task specialization in larger groups we find a general pattern in eusocial Hymenoptera that confirms increasing task specialization of individuals, consistent with increasing organismal complexity found in the evolution of nervous systems, multicellular organisms, ant colonies, and ecosystems (Bell & Mooers, 1997; Changizi et al., 2002). In social insects, caste specialization and reproductive division of labour are often accompanied by different levels of extrinsic mortality, which may lead to the correlation of lifespan differences with increasing colony size. To elaborate further the link between extrinsic mortality it would be interesting to test if workers in larger colonies generally experience higher levels of extrinsic mortality.
We thank Ulrike Lange and Carolin Sempf for their help gathering the data, Alexander Scheuerlein for his help with setting up the database, Susanne Foitzik, Florian Menzel and three anonymous reviewers for helpful suggestions on earlier drafts of the manuscript, and several people who helped us by providing data or helpful information on data sources: Laurent Keller, Miriam Richards, Peter Biedermann, Saul Newman, Ruchira Sen, Bernhard Crespi, Edilberto Gianotti, Mareike Wurdack and Xavier Espadaler, plus those who offered their help after we posted a request on David Nash's INSECTS mailing list. We would also like to thank James W. Vaupel for the idea of using the lifespan differential and Owen Jones for his help with the statistical analysis. The project was funded by the Max Planck Institute for Demographic Research. The authors declare no conflict of interest.