The first two authors contributed equally to the manuscript.
Alexandra Schrempf, Biology I, Evolution, Behavior and Genetics, University of Regensburg, Universitätsstraße 31, 93040 Regensburg, Germany. Tel.: ++49 941 943 3001; fax: ++49 941 943 3304; e-mail: Alexandra.firstname.lastname@example.org
Evolutionary theories of ageing predict that life span increases with decreasing extrinsic mortality, and life span variation among queens in ant species seems to corroborate this prediction: queens, which are the only reproductive in a colony, live much longer than queens in multi-queen colonies. The latter often inhabit ephemeral nest sites and accordingly are assumed to experience a higher mortality risk. Yet, all prior studies compared queens from different single- and multi-queen species. Here, we demonstrate an effect of queen number on longevity and fecundity within a single, socially plastic species, where queens experience the similar level of extrinsic mortality. Queens from single- and two-queen colonies had significantly longer lifespan and higher fecundity than queens living in associations of eight queens. As queens also differ neither in morphology nor the mode of colony foundation, our study shows that the social environment itself strongly affects ageing rate.
Evolutionary theories predict that the rate of ageing – the physiological deterioration of organisms with time – should increase with increasing extrinsic mortality (Rose, 1991), leading to a reduced intrinsic longevity in organisms with high extrinsic mortality rates. This hypothesis has found vast support throughout taxa (Stearns, 1992). For example, species that have lower predation risks and thus lower levels of extrinsic mortality are usually longer lived than animals with a similar metabolic rate but a higher predation risk. This holds both for birds (Holmes et al., 2001) and for bats (Austad & Fischer, 1991), which can escape predators by flying, but also island populations of opossums that have lower predation risks than their conspecifics on the mainland (Austad, 1993).
The enormous intraspecific variation of lifespan in perennial social insects (honeybees, ants, termites and many social wasps) also appears to be explained by evolutionary theories of ageing. In weaver ants, large workers engaging in risky foraging have a higher ageing rate than small workers that preferentially perform tasks inside the colony (Chapuisat & Keller, 2002). Similarly, the reproductives of social insects, which live in the relative safety of their nests and are sheltered from predation or pathogens, outlive most solitary insects and live up to 28 times longer than their nonreproducing nestmates (Keller & Genoud, 1997; Hartmann & Heinze, 2003). This reverses the typical trade-off between fecundity and longevity known from many solitary species (e.g. Stearns et al., 2000; Schrempf et al., 2005; Bourke, 2007; Heinze & Schrempf, 2008).
Different extrinsic mortality was also suggested to explain the difference of queen lifespan between monogynous species (colonies contain a single queen) and polygynous species (colonies may contain multiple queens). Queens of monogynous species live on average longer than queens of polygynous species (Keller & Genoud, 1997; Keller, 1998). Polygyny is often correlated with life in unstable habitats and frequent moves to new nest sites (Hölldobler & Wilson, 1977; Heinze, 1993; for other causes of polygyny see, e.g. Herbers, 1993). Lower queen lifespan in polygynous species was therefore suggested to have evolved in response to a higher risk of extrinsic mortality during moving (Keller & Genoud, 1997).
Reproductives of monogynous and polygynous ant species differ in how they disperse and found new colonies. Queen of monogynous species is generally specialized for long-distance dispersal and solitary founding. They are typically larger and have more fat reserves than queens in polygynous species, which reproduce in or near the maternal nest, assisted by workers (Keller & Passera, 1989; Tsuji & Tsuji, 1996). Consequently, differences in queen lifespan may be influenced by a large set of correlated life-history traits, which are difficult to disentangle in cross-species comparisons.
Factors other than mortality risk outside the nest may therefore influence queen lifespan. In particular, we hypothesize that queen fecundity and longevity may vary with their social environment because of behavioural or chemical competition among queens, variation in the care given by workers or resource availability in the nest. For example, fighting among queens for reproductive rights or resources might result in a lower lifespan and fecundity (e.g. Heinze & Weber, 2011), or workers might care only for queens with a high egg-laying rate. Furthermore, queens themselves or their workers might adjust queen lifespan to changed life history and reproductive tactics. To test the effect of social environment within ant colonies, we randomly assigned queens of the facultatively polygynous tramp ant Cardiocondyla obscurior to colonies with one, two or eight queens and a variable worker number. We observed both worker/queen and queen/queen interactions and determined queen longevity and egg-laying rate. We found that the number of queens per nest strongly affects queen lifespan and fecundity, as both decreased in multi-queen colonies, but in contrast to our expectations, we did not find evidence for visible queen competition or lowered provisioning of queens in multi-queen societies.
Materials and methods
Nests of our study species C. obscurior were collected in 2004 from a lemon plantation at Ilhéus, Bahia, Brazil. In the laboratory, ants were kept under near-natural conditions in climatic chambers with 30/25 °C temperature and 12/12 h day per night cycles (for details see Cremer & Heinze, 2003). Less than one-third of natural nests contained a single queen, whereas approximately two-thirds of the nests were polygynous with two to 15 dealate queens per colony (median: 2; quartiles 1 and 4). Worker numbers per nest ranged from 2 to 75, with a median of 13 (data J.H., see also Heinze & Delabie, 2005). Colony foundation occurs by budding, i.e. workers and queen(s) leave the mother colony together with brood and initiate a new nest close by. The extrinsic mortality rate of queens in monogynous and polygynous nests is therefore expected to be very similar in that all queens are accompanied by workers from the beginning and queens never have to raise first workers on their own. In contrast to a palaearctic clade of Cardiocondyla, in which queens show a pronounced wing diphenism (e.g. Schrempf et al., 2005), queens of C. obscurior have a monomorphic size distribution (B. Lautenschläger & S. Cremer, unpublished data). Queens in monogynous and polygynous societies therefore neither differ in morphology nor life history.
Queen lifespan and fecundity
We extended our previous work on the lifespan and fecundity of monogynous queens (Schrempf et al., 2005; N = 47 for lifespan; N = 21 for fecundity) by testing the effect of multiple queen associations on queen life expectancy and egg laying. We set up ‘low-polygyny colonies’ consisting of two queens (N = 18, 36 ♀♀) and ‘high-polygyny colonies’ with eight queens (N = 16, 128 ♀♀). The composition of experimental colonies matches natural conditions: in 21% of 124 field colonies, a single queen was found, 26.6% consisted of two queens and 10.5% contained 7–15 queens (data J. H., see also Heinze & Delabie, 2005). We allowed young queens from large stock colonies to mate before the start of the experiment (more than 99% of young queens mate within 5 days after emergence; A. Schrempf & S. Cremer, unpublished data) and thereafter used them to set up artificial colonies together with workers from different ‘donor’ colonies. To obtain our sample sizes, we had to set up a few more colonies (N = 50 for monogynous and N = 39 for the polygynous colonies) as queens in three monogynous and five polygynous colonies died within 4 weeks after the start of the experiment. These colonies were omitted, as we know from preliminary studies that colonies may collapse shortly after being set-up, likely due to handling, and thus cannot be used for the study of queen lifespan. To be able to set up sufficient numbers of colonies with queens of approximately the same known age (max. 10 days different), we had to assemble queens from different colonies in each set-up. This, however, in fact reflects natural conditions. First, observations in this and other species of Cardiocondyla suggest that queens are easily adopted by unrelated colonies (e.g. Kinomura & Yamauchi, 1987). Furthermore, Cardiocondyla obscurior is not native to Brazil but was accidentally introduced. Like in other tramp species (e.g. Passera, 1994), colonies do not have clear borders. The exchange of individuals among colonies is possible in the laboratory and likely occurs regularly in the field (Heinze et al., 2006). Finally, observations of queens from preliminary experiments with workers and queens collected from natural colonies, whose age was unknown when they were transferred into the laboratory, showed a similar result as the current study with same-aged young queens. Queens survived for additional 24 ± 5 weeks (mean ± SE; N = 13) in single-queen colonies, but only for additional 19 ± 10 weeks in 10-queen colonies (N = 54; S. Cremer, unpublished data).
Colonies were fed diluted honey and pieces of cockroaches ad libitum, and the survival of the queens as well as the number of eggs in the colonies was determined twice per week. When single queens died in the 2- and 8-queen set-ups, their lifespan was recorded and the queen was replaced by another queen that had been marked by tarsal clipping to keep queen number constant over time as long as the last of the original queens died. To avoid any positive feedback on queen fecundity from increasing worker numbers, we kept the number of workers constant over time by removing all emerging surplus workers.
Colonies differed in initial worker number (from 20 to 80 workers), as we intended to investigate a possible effect of worker/queen ratio on productivity and life span (ranging from 2.5 to 20 workers/queen, according to natural conditions; Heinze & Delabie, 2005). A statistic analysis revealed that worker number affects neither queen lifespan (Cox proportional regression: Wald statistics: worker number: 2.25, P = 0.13; queen number: 18.82, P = 0.000014) nor fecundity (GLM: egg number in the colony: worker number: F1 = 1.63, P = 0.21; queen number: F2 = 35.84, P < 0.00001; queen number*worker number: F1 = 0.06, P = 0.80; egg laying per queen: worker number: F1 = 0.18, P = 0.67; queen number: F2 = 24.51, P < 0.00001). In addition, there are no differences in ovary state between queens kept with either more or less workers (2 ♀♀: U = 7.00, P = 0.77, 8 ♀♀: U = 107.00, P = 0.61), and queen number, not worker number, affects colony growth (multiple regression: queen number: F1 = 21.64, P = 0.00015, worker number: F1 = 0.10, P = 0.765) and behaviour (ancova: worker number: F7 = 1.88, P = 0.21; queen number: F14 = 3.70, P = 0.010). We therefore pooled all replicates for single-, two- and eight-queen colonies for the final analysis.
Ovary development, colony growth and behavioural observations
Regular egg counts provide only an estimate of the total productivity of all queens in the nest. We therefore determined whether individual queens might contribute unequally to the brood by dissecting all queens from four additional two-queen and four-eight-queen colonies 6–8 weeks after the colonies had been established. Again, surplus workers had been removed similar to the above-mentioned experimental colonies.
To determine how variation in queen fecundity affects colony growth rates, we set up the same experimental groups as above but did not remove new workers from additional colonies for a period of 90 days after experimental set-up (1 ♀: N = 7, 2 ♀♀: N = 8, 8 ♀♀: N = 8).
Lastly, we also observed the behaviour of queens and the behaviour of workers towards the queens during 10–20 behavioural scans (Martin & Bateson, 2007) (each of a duration of approximately 1–2 s) per day and queen for 4–7 days per colony (1 ♀: N = 5, 2 ♀♀: N = 6, 8 ♀♀: N = 6). Again, surplus workers were removed regularly from the colonies to keep colony size constant. The following behavioural categories could be distinguished: (i) worker behaviour: antennation of the queens, allogrooming of the queens, carrying the queens; (ii) queen behaviour: brood care, moving in the nest, resting in the nest, time spent outside of the nest.
Parametric statistical analyses were performed when data were normally distributed and of equal variance; otherwise, nonparametric analyses were applied. The data of egg laying were analysed twice (individual and total egg-laying rate), and we reduced the significance level α to 0.025, accordingly. Statistical tests were performed with spss 15.0 (IBM Corporation, Sumers, NY) and SigmaStat 2.03 (Aspire Software International, Ashburn, VA).
Queen lifespan and fecundity
The mean lifespan of queens reared in different groups differed significantly from each other (survival analysis using Gehan’s generalized Wilcoxon test: = 11.79, P = 0.0027). Queens from single-queen and two-queen colonies lived equally long (mean, min–max (weeks): 1 ♀: 26, 6–56; 2 ♀♀: 26, 5–49), whereas queens in eight-queen colonies lived significantly shorter (Fig. 1; 8 ♀♀: 20, 5–36; post hoc Mantel-Cox test: 1♀ vs. 8 ♀♀: χ2 = 12.35, P < 0.0001; 2♀♀ vs. 8 ♀♀: χ2 = 12.12, P = 0.001; 1♀ vs. 2 ♀♀: χ2 = 0.01, P = 0.91). Overall, 17% (8/47) and 16.7% (6/36) queens of the single-queen and two-queen colonies, respectively, lived longer than 36 weeks, which was the maximum lifespan of queens from the eight-queen set-ups.
As expected, the average egg number per colony increased with queen number (Fig. 2; mean ± SE: 1 ♀: 10.88 ± 0.79; 2 ♀♀: 17.43 ± 0.61; 8 ♀♀: 34.69 ± 0.84; anova: F2, 52 = 254.68, P < 0.00001; Scheffé post hoc test: 1 ♀ vs. 2 ♀♀: P < 0.00001, 1 ♀ vs. 8 ♀♀: P < 0.00001, 2 ♀♀ vs. 8 ♀♀: P < 0.00001). Under the assumption that all queens contribute equally to the brood, the average egg number per queen decreased with increasing queen number, but much stronger in 8-queen groups (Fig. 2; Kruskal–Wallis test: H2, 52 = 35.86, P < 0.001; post hoc Dunn′s test: 1 ♀ vs. 2 ♀♀: Q = 1.55, P > 0.05; 1 ♀ vs. 8 ♀♀: Q = 5.87, P < 0.05; 2 ♀♀ vs. 8 ♀♀: Q = 4.21, P < 0.05). In accordance with their longer lifespan and their higher weekly egg-laying rate, queens in single-queen colonies laid on average 283 eggs during their lives, queens in two-queen colonies laid 227 eggs and queens in eight-queen colonies laid only 86 eggs. Although the results for single queens came from a study that had been conducted earlier than the study with multiple queens (albeit under exactly the same laboratory conditions, see Materials and methods), we are confident that this does not affect the validity of our conclusions, as observations on two-queen and eight-queen colonies were made at the same time.
Ovary development, colony growth and behavioural observations
Results did not reveal any difference in ovary status among queens within a colony (2 ♀♀, N = 8: median number of maturing oocytes: 2; 25%/75% quartiles: 1.5, 2.5; variance: 0.57; coefficient of variance CV: 0.378; mature eggs: 2; 1, 2.5; variance 0.70, CV: 0.445; 8 ♀♀, N = 31: median number of maturing oocytes: 2, 25% 75% quartiles 1, 2; variance: 0.44; CV: 0.413; mature eggs: 2; 1,2; variance 0.50, CV: 0.431; Fig. 3). Therefore, queens in multi-queen colonies likely contribute equally to reproduction.
As expected from the higher egg-laying rates in eight-queen colonies, colony growth (slope over 90 days) was highest in colonies with eight queens, intermediate in two-queen colonies and least in single-queen colonies (1 ♀, N = 7: 0.05 ± 0.13; 2 ♀♀, N = 8: 0.23 ± 0.14; 8 ♀♀, N = 8: 0.42 ± 0.11; anova: F2, 20 = 16.44, P = 0.00006; Scheffépost hoc tests: 1 ♀ vs. 2 ♀♀: P = 0.044; 1♀ vs. 8 ♀♀: P = 0.00006; 2♀♀ vs. 8 ♀♀: P = 0.019). The highest percentage colony growth over the 90 days was 235% in an 8-queen colony.
Workers did not treat queens differently, and queens did not differ in their behaviour within the respective set-ups (χ2 test, P > 0.05 for all comparisons). Thus, the mean of each colony was used to compare the different groups in an anova. In the case of grooming, the homology of variances was not given (Brown–Forsythe test: F = 5.30, P = 0.019), and we thus conducted a Median test.
The frequencies of worker behaviour towards the queens (antennation, grooming and carrying of queens) did not differ between the three experimental colony types (antennation: anova: F2, 14 = 1.28, P = 0.309; grooming: Median test: = 0.14, P = 0.932; carrying: anova: F2, 14 = 0.73, P = 0.498). Queen behaviour was also quite similar regardless of queen number: running, resting in the brood chamber and spending time outside the brood chamber did not differ significantly (anova: running: F2, 14 = 3.36, P = 0.064; resting: F2, 14 = 0.49, P = 0.624; time outside: F2, 14 = 2.14, P = 0.155). However, queens in eight-queen associations spent significantly less time with brood care behaviour than queens in two-queen colonies (anova: F2, 14 = 4.67, P = 0.028; Scheffé test: 1 ♀ vs. 2♀♀: P = 0.99; 1 ♀ vs. 8 ♀♀: P = 0.07; 2 ♀♀ vs. 8 ♀♀: P = 0.05).
In this study, we could clearly show that queens of Cardiocondyla obscurior lived significantly longer in experimental single- and two-queen colonies than in eight-queen colonies, despite of equal morphology, mode of colony founding and level of ‘extrinsic mortality’. Furthermore, queens from single- and two-queen colonies laid significantly more eggs per week than queens from 18-queen colonies. Both differences resulted in a much lower lifetime reproductive success (eggs/lifetime) of queens in eight-queen colonies. Not a single queen of an 8-queen association reaches roughly the maximum lifespan of queens kept alone or together with one other queen.
Our findings of fecundity and longevity changes after manipulation of queen number corroborate earlier observations that in social insects, longevity is not traded off against fecundity (see also Page & Peng, 2001; Schrempf et al., 2005; Corona et al., 2007). In addition, the fact that queens from two-queen colonies much more resemble those of single-queen colonies than those from eight-queen colonies documents that in C. obscurior, lifespan and fecundity are less affected by the fundamental dichotomy between monogyny and polygyny but instead vary more with the absolute number of queens. Several previous studies have documented that conspecific queens are considerably less fecund in polygynous than in monogynous colonies (e.g. Solenopsis invicta,Vargo & Fletcher, 1989; Vargo, 1992; Plagiolepis pygmaea,Mercier et al., 1985; Leptothorax acervorum,Bourke, 1993), and genetic analyses suggest that queen turnover may be more rapid in polygynous than monogynous colonies (Heinze et al., 1995; Bourke et al., 1997; Bargum et al., 2007). However, our study is the first to compare lifespan of queens in polygynous and monogynous societies of the same species. It reveals that social environment plays a crucial role with regard to lifespan in insect societies. Hence, it might be of similar importance in other animal societies. For example, Wasser & Sherman (2010) suggested that bird species breeding in groups live longer and attributed this to a lower extrinsic mortality rate (but see Blumstein & Møller, 2008). It would be interesting to investigate whether and how group size influences longevity at a given extrinsic mortality. It may also play a role in human societies, where telomere length with its reported role in ageing is correlated with social status (Cherkas et al., 2006). Because of other reported similarities in ageing patterns between ant societies and societies of naked mole rats (Dammann & Burda, 2006), we suggest that social group composition may also affect longevity in this eusocial species, which, however, may be difficult to document because of their long generation time.
Furthermore, it is important to combine studies describing the effects of social environment on longevity and fecundity with studies on the underlying mechanisms of the observed variation, as only the combination of both ultimate and proximate approaches will lead to an understanding of the evolution of ageing in social species. For example, in our study, the attraction of workers to queens did not vary with queen number. Similarly, and again in contrast to our hypotheses, queens did not differ in their active behaviour, except that queens in eight-queen associations performed on average less brood care because of the much lower brood/queen ratio (see Fig. 2). However, these results are only marginally significant and cannot explain the pronounced differences in lifespan and fecundity. Queens also did not show visible agonistic interactions towards each other. Physical queen–queen competition as a factor influencing queen longevity is therefore unlikely. At present, we cannot exclude that the lowered lifespan of queens in eight-queen colonies is because of unnoticed, subtle behavioural differences in queen–queen or queen–worker interactions and/or subsequent stress. Furthermore, we also cannot exclude chemical competition among queens. Queen pheromones appear to play an important role in the regulation of worker behaviour and fertility, but it is debated whether such chemicals are honest signals or manipulative agents (see e.g. Heinze & d′Ettorre, 2009; Holman et al., 2010). Manipulative queen pheromones are considered to be unstable in evolution. Mutual pheromonal inhibition among queens (Vargo & Fletcher, 1989) therefore seems unlikely (Keller & Nonacs, 1993). In fact, experimental manipulation of queen number in colonies of the ant Leptothorax acervorum did not result in the predicted changes in queen fecundity (Bourke, 1993).
The observed pattern of queen lifespan variation meets expectations from life-history theory. In monogynous species, queens usually first rear numerous workers and increase colony size before switching to the production of sexuals. In contrast, under polygynous conditions, queens often begin to produce sexual offspring very early in their lives (Keller & Genoud, 1997). The short generation time of dependently founding queens may lead to higher fitness in growing populations, and selection for a long queen lifespan may thus be relaxed (Tsuji, 2006). In addition, kin selection may select for earlier programmed death when dispersal is limited (Travis, 2004; Dytham & Travis, 2006). Dispersal distance is indeed limited in C. obscurior, as most queens mate in the natal nest and shed their wings before leaving. If queens adjusted their reproductive tactics and age at death in response to the presence of other queens, they would switch earlier to the production of sexuals in the presence of other queens and die earlier. Although sexual production was not monitored in our study, previous studies show that C.obscurior queens indeed produce sexuals earlier in multi-queen colonies (Suefuji et al., 2008) and also adaptively adjust their sex ratio (Cremer & Heinze, 2002).
Despite lower per capita fecundity and shortened lifespan, eight-queen colonies grew faster than single- and two-queen colonies. Such highly polygynous colonies may therefore more quickly produce sexuals, propagate through budding and colonize habitat patches. Surprisingly, eight-queen associations in C. obscurior appear to be rare in nature (Heinze & Delabie, 2005) compared to single- and two-queen colonies. In general, polygynous colonies of facultatively polygynous species are expected to occur particularly under conditions of nest site limitation and/or when dispersal is risky (see e.g. Herbers, 1986). In our population, nest sites – ephemeral cavities in plant material, such as folded lemon leaves or the bracts of aborted coconuts – do not appear to be limited but provide colonies with highly restricted space. The observed pattern might therefore simply reflect the fact that natural nest sites are too small to harbour large colonies with numerous queens. Alternatively, the pattern in the field might indicate that the queens at least partly determine colony composition. Egg-laying rates suggest that, from an individual queen’s point of view, the optimal queen number per colony is lower than the queen number that maximizes colony productivity.
Research on Cardiocondyla obscurior from Brazil was made possible through a permit of the Brazilian Ministry of Science and Technology (RMX 004/02) and financial support from DFG (Schr 1135/1-1 and He 1623/22). Two referees made helpful comments on the manuscript.